PERSISTENT TRANSMISSION OF MALARIA IN COASTAL PAPUA NEW GUINEA: EFFECT OF VECTOR BITING PATTERNS AND HOST SELECTION By John Bosco Keven A DISSERTATION Michigan State University in partial fulfillment of the requirements Submitted to for the degree of Microbiology and Molecular Genetics—Doctor of Philosophy 2020 ABSTRACT PERSISTENT TRANSMISSION OF MALARIA IN COASTAL PAPUA NEW GUINEA: EFFECT OF VECTOR BITING PATTERNS AND HOST SELECTION By John Bosco Keven In Papua New Guinea where malaria is endemic, nationwide vector control campaign based on free distribution of insecticidal bed nets is presently ongoing to alleviate the burden of the disease. In the north coast of Madang province, prevalence of malaria infection in human populations in 2006 was 42.1% Plasmodium falciparum and 41.7% Plasmodium vivax. After the commencement of the bed net campaign in this region in 2008, infection prevalence dropped for both P. falciparum (9.0% in 2014) and P. vivax (12.7% in 2010). Human biting rates, the number of Anopheles bites per person-night, dropped from a range of 10–60 in 2009 to 3.0–18 in 2011. Infectious biting rates, the number of Plasmodium spp. sporozoite-infective Anopheles bites per person-year, dropped from a range of 20–350 in 2009 to 0–15 in 2011. The research described in this dissertation investigated if the ongoing efforts to intensify the bed net program has succeeded in reducing malaria transmission further or whether further reduction is hampered by vector-related ecological factors which enable transmission to persist. Barrier screen and human landing catch mosquito sampling, active-case malaria detection, and bed net survey were conducted in 2016–2017 in the north coast villages of Madang. PCR methods were used to identify Anopheles species and their blood-meal host, and to test for malaria infection in humans and mosquitoes. Human blood meals were fingerprinted by microsatellite genotyping to identify the individuals bitten. The results show that despite high bed net usage (> 80% of villagers reportedly slept under a net) in all villages, the overall infection prevalence rebounded to 18% P. falciparum and 14% P. vivax in 2016–2017. The human biting rates increased to a range of 9.0– 31 bites per person-night. The infectious biting rates also increased to a range of 6.5–159.7 infective bites per person-year. Anopheles species richness ranged from four to six species per village, but relative abundance was highly uneven within and between villages, and community composition was generally dissimilar among villages. The vectors did not exhibit strict anthropophagy but were rather opportunistic in their host selection and utilized non-human hosts when access to humans was limited by the bed nets. The frequency of blood meals obtained from different human individuals was not random but rather clustered, and a few individual villagers were utilized as hosts more frequently than others. The human biting frequency was also clustered spatially and was high in few locations and low in most locations, causing the basic reproduction rate of malaria to increase by multiplicative factors ³ 1.6. Most (29–50%) of the bites taken on humans occurred before 22:00 hr when 75–90% of people were awake and unprotected by the bed nets and 67–83% of the bites occurred before nearly everyone retired to bed at 2:00 hr. Most (> 50%) of the bites taken on humans occurred outdoors where the risk of exposure to the bed nets was low. These results show that in the north coast of Madang, malaria transmission has increased in recent years and continues to persist at high levels despite ongoing efforts to intensify the vector control program. The diverse Anopheles species, their high abundance and great dissimilarity in species composition among villages indicate that the bed net is failing to affect the vectors, allowing vector populations to revert to pre-bed net condition. The opportunistic host utilization and outdoor biting behavior of the local vectors enable them to evade the indoor-deployed bed nets by feeding on alternative hosts to supports their population size and continue to transmit malaria by feeding on humans outdoors and early in the evening when people are active and unprotected by the bed nets. By increasing the basic reproduction rate, the heterogeneity of bites on humans causes transmission to persist rather than dissipate. This dissertation is dedicated to my son Jerome and daughter Jerolyn who witnessed me go through this studious academic pursuit as they transitioned from childhood to teenagers. iv ACKNOWLEDGEMENTS The success of the research described in this dissertation and my PhD program in general was made possible through the support of various individuals. First and foremost, I would like to thank Dr. Edward Walker, Professor of Microbiology and Molecular Genetics and Entomology at Michigan State University (MSU), for accepting to be my academic advisor and for supporting me in various ways throughout my graduate program. His scientific guidance, wisdom and sense of humor were invaluable to my academic and research achievements. Ned, you’re the best and will always be a very special person to me. Secondly, I would like to acknowledge the support of Dr. Peter Zimmerman, Dr. James Kazura, Dr. Christopher King and Dr. Daniel Tisch of the Center for Global Health and Diseases at Case Western Reserve University. These individuals were instrumental in shaping my carrier path prior to and during my PhD program and facilitated financial support through the US Fogarty scholarship program during the first half of my graduate program. I would also like to acknowledge the MSU faculty who served on my guidance committee namely Dr. James Smith, Dr. Lixin Zhang, Dr. Zhiyong Xi and Dr. David Foran, as well as Dr. Shannon Manning who served as the chair of my preliminary examination. Thank you for your guidance and oversight on my research project and academic training. Dr. Foran not only served on my guidance committee but also gave me access to his laboratory and trained me on molecular forensic techniques that were relevant to my research. My research project was supported through a collaborative effort between Dr. Walker’s laboratory and Dr. Foran’s laboratory at MSU and Papua New Guinea Institute of Medical Research Vector Borne Diseases Unit (PNGIMR-VBDU). Dr. Moses Laman (current head of PNGIMR-VBDU) and Dr. Leanne Robinson (former head of PNGIMR-VBDU) both provided administrative oversight, and v together with Dr. Walker, myself and others helped secure a small grant from the Tropical Disease Research program of the World Health Organization to fund the field-component of my research. The field work was conducted with the support of scientific staff, nursing officers and field technicians in the VBDU, particularly Dr. Stephan Karl (then head of the Entomology Section of VBDU) who provided operational oversight and Miss Michelle Katusele who coordinated and conducted the field work alongside me and Dr. Rodríguez-Rodríguez who was investigating the association between human behavior and risk of malaria. Most of the laboratory assays performed on the mosquito and human blood samples were conducted in Dr. Walker’s and Dr. Foran’s laboratories by two MSU undergraduate students Miss Georgia Artzberger and Miss Mary Gilles under my supervision. Finally, I would like to give a very special thanks to my parents for making me believe in my intellectual capability and for setting the trajectory early on in my life for me to arrive at this highly unlikely stage of my life, and to my family (wife Adela, son Jerome and daughter Jerolyn) for their company and emotional support throughout the duration of my graduate program. vi TABLE OF CONTENTS LIST OF TABLES ........................................................................................................................ xi LIST OF FIGURES .................................................................................................................... xiv KEY TO ABBREVIATIONS .................................................................................................... xvii CHAPTER 1: INTRODUCTION ...................................................................................................1 1.1. Overview ...............................................................................................................................1 1.2. Biology of malaria parasites and mosquito vectors .............................................................. 1 1.2.1. Plasmodium life cycle ....................................................................................................1 1.2.2. Anopheles life cycle and their role as malaria vectors ...................................................3 1.3. Global epidemiology of malaria ...........................................................................................5 1.3.1. Geographic distribution of the parasite species .............................................................5 1.3.2. Geographic distribution of the primary malaria vector species ..................................... 6 1.3.3. Malaria endemicity ........................................................................................................ 7 1.3.4. Public health burden and control of malaria ................................................................10 1.4. Malaria in Papua New Guinea ............................................................................................12 1.4.1. Landscape and geo-climatic description ......................................................................12 1.4.2. Anopheles vectors ........................................................................................................15 1.4.3. Malaria epidemiology ..................................................................................................16 1.5. Research rationale and objective ........................................................................................18 1.5.1. Rationale ......................................................................................................................18 1.5.2. Objective ......................................................................................................................21 1.6. Research ethics approval .....................................................................................................22 REFERENCES ..........................................................................................................................24 CHAPTER 2: SPECIES COMPOSITION, ABUNDANCE AND NOCTURNAL ACTIVITIES OF FEMALE ANOPHELES IN MADANG, PAPUA NEW GUINEA: ASSESSMENT WITH BARRIER SCREEN SAMPLING ...............................................................................................34 2.1. Abstract ...............................................................................................................................34 2.2. Introduction .........................................................................................................................35 2.3. Methods ...............................................................................................................................38 2.3.1. Study sites ....................................................................................................................38 2.3.2. Mosquito sampling ......................................................................................................39 2.3.3. Data analyses ...............................................................................................................42 2.4. Results .................................................................................................................................43 2.4.1. Species composition and abundance ............................................................................43 2.4.2. Nocturnal activities of mosquitoes ...............................................................................50 2.5. Discussion ...........................................................................................................................54 2.6. Conclusion ..........................................................................................................................59 APPENDICES ...........................................................................................................................60 Appendix 2.1. Dates of barrier screen collection in the study villages ......................................61 vii Appendix 2.2. Bray-Curtis index matrices ................................................................................64 REFERENCES ..........................................................................................................................65 CHAPTER 3: PERSISTENT MALARIA TRANSMISSION IN HUMAN POPULATIONS WITH HIGH BED NET USAGE IN MADANG, PAPUA NEW GUINEA ...............................70 3.1. Abstract ...............................................................................................................................70 3.2. Introduction .........................................................................................................................71 3.3. Methods ...............................................................................................................................74 3.3.1. Study sites and mosquito sampling ..............................................................................74 3.3.2. Human blood sampling and bed net survey .................................................................75 3.3.3. PCR detection of malaria in humans and mosquitoes .................................................76 3.3.4. Data analyses ...............................................................................................................77 3.4. Results .................................................................................................................................78 3.4.1. Bed net ownership and usage .......................................................................................78 3.4.2. Malaria infection prevalence ........................................................................................79 3.4.3. Anopheles vectors ........................................................................................................82 3.4.4. Sporozoite rates ............................................................................................................84 3.4.5. Human biting rates .......................................................................................................85 3.4.6. Annual infectious biting rate ........................................................................................87 3.5. Discussion ...........................................................................................................................88 3.6. Conclusion ..........................................................................................................................92 REFERENCES ..........................................................................................................................94 CHAPTER 4: SPATIAL AND TEMPORAL HETEROGENEITY OF HUMAN EXPOSURE TO ANOPHELES BITES IN THE NORTH COAST VILLAGES OF MADANG PROVINCE, PAPUA NEW GUINEA ...............................................................................................................98 4.1. Abstract ...............................................................................................................................98 4.2. Introduction .........................................................................................................................99 4.3. Methods .............................................................................................................................101 4.3.1. Study sites, mosquito sampling and active case malaria detection ............................101 4.3.2. Data analyses .............................................................................................................101 4.4. Results ...............................................................................................................................103 4.4.1. Anopheles vectors ......................................................................................................103 4.4.2. Temporal heterogeneity of exposure to Anopheles bites ...........................................104 4.4.3. Spatial heterogeneity of exposure to Anopheles bites ................................................108 4.4.4. Spatial distribution of malaria case households .........................................................112 4.5. Discussion .........................................................................................................................117 4.6. Conclusion ........................................................................................................................121 REFERENCES ........................................................................................................................122 CHAPTER 5: PROBE-BASED MULTIPLEX QUANTITATIVE PCR IDENTIFIES BLOOD- MEAL HOSTS IN ANOPHELES MOSQUITOES FROM PAPUA NEW GUINEA ...............126 5.1. Abstract .............................................................................................................................126 5.2. Introduction .......................................................................................................................127 5.3. Methods .............................................................................................................................129 5.3.1. Samples ......................................................................................................................129 viii 5.3.2. Probe design ...............................................................................................................129 5.3.3. Development of multiplex qPCR ...............................................................................130 5.3.4. Assay validation .........................................................................................................131 5.4. Results ...............................................................................................................................132 5.5. Discussion .........................................................................................................................137 5.6. Conclusion ........................................................................................................................140 APPENDIX ..............................................................................................................................142 Appendix 5.1. Non-specific human primers ........................................................................143 REFERENCES ........................................................................................................................144 CHAPTER 6: PLASTICITY OF HOST SELECTION BY ANOPHELES VECTORS OF MALARIA IN MADANG, PAPUA NEW GUINEA: IMPLICATIONS FOR PERSISTENT MALARIA TRANSMISSION ...................................................................................................149 6.1. Abstract .............................................................................................................................149 6.2. Introduction .......................................................................................................................150 6.3. Methods .............................................................................................................................152 6.3.1. Study sites and mosquito sampling ............................................................................152 6.3.2. Identification of Anopheles species and their blood-meal hosts ................................152 6.3.3. Demographic survey ..................................................................................................153 6.3.4. Data analyses .............................................................................................................154 6.4. Results ...............................................................................................................................154 6.4.1. Mosquito blood-meal sources ....................................................................................154 6.4.2. Mosquito host selection .............................................................................................157 6.4.3. Relationship between bed net usage and anthropophagy ..........................................165 6.5. Discussion .........................................................................................................................166 6.6. Conclusion ........................................................................................................................171 APPENDIX ..............................................................................................................................172 Appendix 6.1. Tabulated blood-meal sources of mosquitoes ..............................................173 REFERENCES ........................................................................................................................179 CHAPTER 7: NON-RANDOM HOST SELECTION AND MIXED HUMAN BLOOD MEALS BY ANOPHELES VECTORS OF MALARIA IN MADANG, PAPUA NEW GUINEA: IMPLICATIONS FOR MALARIA TRANSMISSION .............................................................184 7.1. Abstract .............................................................................................................................184 7.2. Introduction .......................................................................................................................185 7.3. Methods .............................................................................................................................187 7.3.1. Mosquito and human blood samples ..........................................................................187 7.3.2. Human genotyping .....................................................................................................188 7.3.3. Data analyses .............................................................................................................189 7.4. Results ...............................................................................................................................191 7.4.1. Percent match values of human profiles ....................................................................191 7.4.2. Cryptic mixed meals and probability of interrupted feeds .........................................192 7.4.3. Non-random human host selection ............................................................................194 7.5. Discussion .........................................................................................................................201 7.6. Conclusion ........................................................................................................................207 APPENDICES .........................................................................................................................209 ix Appendix 7.1. Percent profile match values ............................................................................210 Appendix 7.2. Human blood index of Anopheles populations ................................................211 Appendix 7.3. Frequency distribution of blood meals .............................................................212 REFERENCES ........................................................................................................................213 CHAPTER 8: GENERAL CONCLUSION ................................................................................219 x LIST OF TABLES Table 2.1. Number of mosquitoes collected in each of seven villages for all years combined ....44 Table 2.2. Number of mosquitoes collected in each of five villages in April–December 2012 ...44 Table 2.3. Number of mosquitoes (excluding the species An. farauti no. 4) collected in each of four villages in August 2016–March 2017 ...............................................................................44 Table S2.1. Dates of barrier screen mosquito sampling in each village in the years 2012, 2013, 2015, 2016 and 2017 .....................................................................................................................61 Table S2.2. Bray-Curtis dissimilarity index matrix based on species abundance data for April–December 2012 sampling period ........................................................................................64 Table S2.3. Bray-Curtis dissimilarity index matrix based on species abundance data for August 2016–March 2017 sampling period ..................................................................................64 Table 3.1. SR of 12 vector populations (values outside parentheses are number of sporozoite- positive mosquitoes and inside are the SR) ...................................................................................84 Table 3.2. SR of Anopheles spp. in four villages (values outside parentheses are number of sporozoite-positive mosquitoes and inside are the SR) .................................................................85 Table 4.1. Number of Anopheles mosquitoes according to species collected in four villages ...104 Table 4.2. Chi-square test of fit of negative binomial and Poisson models to frequency distribution of Anopheles numbers collected inside and outside houses in three villages ..........111 Table 4.3. Estimated Rf for indoor and outdoor mosquito collections in three villages .............111 Table 5.1. Primers and probes for the three vertebrate host species ...........................................130 Table 5.2. Standard curve and reaction efficiency of the three qPCR assays .............................133 Table 5.3. Frequency of concordant and discordant Anopheles blood-meal results analyzed by quantitative versus conventional PCR ...................................................................................135 Table S5.1. Human primers designed to target mitochondrial genes found to be human-specific by in silico test but were nonspecific (also amplifies dog and/or pig DNA) by in vitro test ......143 Table 6.1. Number of humans, pigs and dogs in village census and blood meals of two populations of An. bancroftii along with their host selection tendency based on exact binomial comparison of the expected (Exp) and observed (Obs) blood index ..........................................159 xi Table 6.2. Number of humans, pigs and dogs in village census and blood meals of a population of An. farauti no. 4 along with their host selection tendency based on exact binomial comparison of the expected (Exp) and observed (Obs) blood index ..............................................................159 Table 6.3. Number of humans, pigs and dogs in village census and blood meals of five populations of An. farauti (s.s.) along with their host selection tendency based on exact binomial comparison of the expected (Exp) and observed (Obs) blood index ..........................................160 Table 6.4. Number of humans, pigs and dogs in village census and blood meals of six populations of An. koliensis along with their host selection tendency based on exact binomial comparison of the expected (Exp) and observed (Obs) blood index ..........................................161 Table 6.5. Number of humans, pigs and dogs in village census and blood meals of two populations of An. longirostris along with their host selection tendency based on exact binomial comparison of the expected (Exp) and observed (Obs) blood index ..........................................162 Table 6.6. Number of humans, pigs and dogs in village census and blood meals of six populations of An. punctulatus (s.s.) along with their host selection tendency based on exact binomial comparison of the expected (Exp) and observed (Obs) blood index ...........................163 Table 6.7. Two bed net metrics in each village during one or more sampling periods ..............165 Table 6.8. Results of multiple logistic regression which tested for association between mosquito anthropophagy and two predictor variables ................................................................................166 Table S6.1. An. bancroftii blood-meal sources ...........................................................................173 Table S6.2. An. farauti (s.s.) blood-meal sources .......................................................................174 Table S6.3. An. farauti no. 4 blood-meal sources .......................................................................175 Table S6.4. An. koliensis blood-meal sources .............................................................................176 Table S6.5. An. longirostris blood-meal sources ........................................................................177 Table S6.6. An. punctulatus (s.s.) blood-meal sources ...............................................................178 Table 7.1. Distribution of successfully genotyped blood meals according to Anopheles species ......................................................................................................................................................193 Table 7.2. Estimates of blood host choice parameters for six Anopheles populations ...............193 Table 7.3. Results of exact binomial test comparing the observed proportion of each sex in the blood meals with their expected proportion based on village census count ...............................198 xii Table 7.4. Results of exact binomial test comparing the observed proportion of each age group in the blood meals with their expected proportion based on village census count .........................198 Table 7.5. Logistic regression results for test of variation in likelihood of malaria infection among age and sex groups in Mirap ...........................................................................................200 Table 7.6. Logistic regression results for test of variation in likelihood of malaria infection among age and sex groups in Wasab ..........................................................................................201 Table S7.1. Human blood index (B) of six Anopheles populations ............................................211 xiii LIST OF FIGURES Figure 1.1. Map of Papua New Guinea showing three geo-climatic regions separated by grey dashed lines ...................................................................................................................................14 Figure 2.1. Map showing locations of the study villages .............................................................39 Figure 2.2. Barrier screen setup ....................................................................................................40 Figure 2.3. Relative species abundance of Anopheles mosquitoes ...............................................47 Figure 2.4. Box plots showing the distribution of Anopheles spp. numbers collected nightly per barrier screen .................................................................................................................................48 Figure 2.5. PCoA biplots of villages (black dots) with Anopheles species (blue dots) projected on the plot based on their weighted average scores ...........................................................................49 Figure 2.6. Mean number (dots) of blood-fed and unfed mosquitoes collected per barrier screen per night (y-axis) at each of the 12 hr periods of the night (x-axis, in 24 hr format) ...................51 Figure 2.7. Mean number (dots) of blood-fed and unfed mosquitoes collected per barrier screen per night (y-axis) at each of the 12 hr periods of the night (x-axis, in 24 hr format) ...................52 Figure 2.8. Plot of GLM coefficients (open circle with 95% confidence interval bars) showing the comparison of mosquito numbers captured in each of 11 hourly nocturnal periods compared to 19:00 hr (represented by the broken vertical line centered at 0) as the reference variable and on village side of the barrier screen compared to the bush side (represented by the broken vertical line) as the reference variable .......................................................................................................53 Figure 2.9. Plot of GLM coefficients (open circle with 95% confidence interval bars) showing the comparison of mosquito numbers captured in each of 11 hourly nocturnal periods compared to 19:00 hr (represented by the broken vertical line centered at 0) as the reference variable and on village side of the barrier screen compared to the bush side (represented by the broken vertical line) as the reference variable .......................................................................................................54 Figure 3.1. Prevalence of infection (all ages and sexes) of P. falciparum, P. vivax and mixed species in four villages ..................................................................................................................80 Figure 3.2. Prevalence of malaria infection in four human age groups in four villages ...............81 Figure 3.3. Plot of logistic regression odds ratio (open circles or boxes) with 95% confidence interval bars for P. falciparum (black) and P. vivax (grey) infection in four human age groups in four villages ..................................................................................................................................82 xiv Figure 3.4. Relative abundance of four Anopheles species in four villages .................................83 Figure 3.5. Bar plots of mean HBR (y axis) with standard error bar for 12 vector populations (x axis) across four villages ...............................................................................................................86 Figure 3.6. Bar plots of mean HBR with standard error bar (y axis) for Anopheles spp. in four villages (x axis) .............................................................................................................................87 Figure 3.7. Bar plots of mean IBR with standard error bar (y axis) for P. falciparum and P. vivax ........................................................................................................................................................88 Figure 4.1. Mean number (colored dots) of mosquitoes collected per person-night at each of six bihourly periods of the night (expressed in 24 hr format) in four villages .................................106 Figure 4.2. Plot of GLM coefficients (open circle) with 95% confidence interval bars of five variables (y-axis) of Anopheles in four villages .........................................................................107 Figure 4.3. Proportion of human-seeking Anopheles spp. captured at three nocturnal periods in four villages ................................................................................................................................108 Figure 4.4. Boxplots showing the distribution of Anopheles spp. numbers captured nightly inside and outside of houses in four villages .........................................................................................109 Figure 4.5. Map showing the location of houses where mosquitoes were sampled in Megiar, Mirap and Wasab villages ...........................................................................................................110 Figure 4.6. Geolocation of malaria case (colored squares) and non-case (black circles) households for P. falciparum (panel a) and P. vivax (panel c) in Bulal, along with their corresponding D statistic plot (panel b and d) ............................................................................113 Figure 4.7. Geolocation of malaria case (colored squares) and non-case (black circles) households for P. falciparum (panel a) and P. vivax (panel c) in Megiar, along with their corresponding D statistic plot (panel b and d) ............................................................................114 Figure 4.8. Geolocation of malaria case (colored squares) and non-case (black circles) households for P. falciparum (panel a) and P. vivax (panel c) in Mirap, along with their corresponding D statistic plot (panel b and d) ............................................................................115 Figure 4.9. Geolocation of malaria case (colored squares) and non-case (black circles) households for P. falciparum (panel a) and P. vivax (panel c) in Wasab, along with their corresponding D statistic plot (panel b and d) ............................................................................116 Figure 5.1. Results of conventional (the agarose gel images) and quantitative (the linear amplification plots) multiplex PCR both applied to the same 10-fold DNA dilution series ......134 xv Figure 5.2. Bar plot showing mean DNA concentration (with standard error bar) for mosquito blood meals that were detected versus undetected by conventional PCR ..................................136 Figure 6.1. Proportion of blood-meal sources of mosquitoes sampled in study villages ...........156 Figure 7.1. Percentage of cryptic mixed human meals ...............................................................194 Figure 7.2. Chi-square test of fit of zero-truncated Poisson and zero-truncated negative binomial expected distribution (red curves) to observed distribution (vertical bars) of the frequency of occurrence of unique blood-meal profiles in mosquito samples of Mirap (top row) and Wasab (bottom row) villages ..................................................................................................................196 Figure 7.3. Prevalence of infection for P. falciparum and P. vivax in residents of Mirap and Wasab villages ............................................................................................................................199 Figure S7.1. Probability of occurrence (y-axis) of different percent match values (x-axis) calculated from pairwise percent profile match analysis of 427 DNA profiles of Mirap and 164 DNA profiles of Wasab residents ...............................................................................................210 Figure S7.2. Distribution of frequency of blood meals taken on different human individuals by Anopheles mosquitoes in Mirap and Wasab village ...................................................................212 xvi KEY TO ABBREVIATIONS BLAST Basic Local Alignment Search Tool CI DNA confidence interval deoxyribonucleic acid EDTA ethylenediaminetetraacetic acid GLM HBR HLC IBR LT NDOH OR PCoA PCR PNG qPCR RC rRNA SR WHO generalized linear model human biting rate human landing catch infectious biting rate light trap National Department of Health odds ratio principal coordinates analysis polymerase chain reaction Papua New Guinea quantitative polymerase chain reaction resting collection ribosomal ribonucleic acid sporozoite rate World Health Organization xvii CHAPTER 1: INTRODUCTION 1.1. Overview This chapter presents a literature review of relevant materials as background information for the topics that are discussed in subsequent chapters. It begins with the most fundamental aspects of the biology of malaria parasites and their mosquito vectors and how these are linked to the process of transmission of the disease. This is followed by the global epidemiology of malaria, which includes its geographic distribution and the distribution of the primary mosquito vectors, the categorization of malaria epidemiology into different endemicity classes and the public health burden and control of the disease. The discussion then switches from a global to local scale by focusing on the epidemiology of malaria in Papua New Guinea (PNG). The chapter ends with the objective and statement of ethical approval of the research reported in this dissertation. 1.2. Biology of malaria parasites and mosquito vectors 1.2.1. Plasmodium life cycle Malaria is a vector-borne infectious disease of humans. It is caused by infection of the liver and red blood cells by microscopic unicellular protozoan parasites in the genus Plasmodium. The disease is transmitted from person to person by the bites of female mosquitoes in the genus Anopheles. There are many species of Plasmodium but only four are known obligate parasites of humans: Plasmodium falciparum, Plasmodium vivax, Plasmodium malariae and Plasmodium ovale. They are called human malaria because they are not known to naturally infect and cause disease in other mammalian, avian or reptilian hosts (Antinori, et al. 2012; Gilles 1993; Schlagenhauf 2004). A fifth species, Plasmodium knowlesi, is a simian malaria but it can also 1 infect humans if humans are exposed to the bites of the mosquitoes that transmit this parasite among the simian hosts (Cox-Singh, et al. 2008; White 2008). Transmission of the malaria parasites is directly linked to their life development cycle, therefore, a review of the fundamental aspects of this complex biological system is relevant. Although the development stages of the four Plasmodium species vary in fine details, they all undergo the same general life cycle which consists of two parts: a sporogonic and merogonic cycle. The sporogonic cycle includes all the sexual development stages, which take place in the mosquito vector (definitive host) whereas the merogonic cycle includes all the asexual development stages, which occur in the human host (intermediate host). The sporogonic cycle begins when an uninfected female Anopheles bites an infected person and ingests a blood meal containing the haploid stage of the parasite called the gametocyte. If both forms of gametocyte— microgametocyte (male gamete) and macrogametocyte (female gamete)—are present in the blood meal, sexual fusion takes place inside the mosquito midgut about an hour after the blood meal uptake to form a diploid cell called a zygote. About a day after the sexual fusion, the zygote transforms into a motile, partially elongated stage called an ookinete. The ookinete exits the mosquito midgut lumen by traversing the midgut epithelium and affixes itself on the outer side (the hemocoel side) of the gut. There the ookinete transforms into a round, immobile stage called an oocyst. Each oocyst undergoes sporogony, a 10–14 days process where the parasite undergoes multiple fission forming thousands of sporozoites—a motile, elongated, haploid stage of the parasite. Upon rupture of an oocyst, thousands of sporozoites are released into the hemolymph (insect blood) and migrate to the mosquito head and thorax where they inhabit the salivary glands. When the mosquito takes subsequent blood meals from humans, the sporozoites are 2 inoculated along with the mosquito saliva into those individuals, thereby infecting them (Ghosh, et al. 2000; Smith and Jacobs-Lorena 2010; Vlachou, et al. 2006). The merogonic cycle begins when a person is inoculated with the sporozoites. One to several hours after inoculation, the sporozoites leave the inoculation site and travel through the blood stream to the liver where they invade and infect hepatocytes (Mota, et al. 2001; Yamauchi, et al. 2007). Inside the liver cells, each sporozoite undergoes schizogony, a multiple fission process which gives rise to thousands of haploid parasite stage called merozoites. When released from the liver into the blood stream, each merozoite invades a red blood cell (erythrocyte) and undergo schizogony again, producing thousands of merozoites in the blood cell. Upon rupture of the infected erythrocyte, thousands of merozoites are released into the blood stream where they further invade uninfected erythrocytes and repeat the schizogony process. It is the invasion, destruction and depletion of the red blood cells that directly or indirectly cause mild and severe forms of clinical illnesses observed in malaria patients. During the blood stage schizogony, some of the schizonts differentiate into gametocytes which circulate in the blood stream and where they can be picked up by a mosquito and undergo the sporogonic cycle (Roberts and Janovy 2005; Tuteja 2007). 1.2.2. Anopheles life cycle and their role as malaria vectors There are over 3,500 known species of mosquitoes distributed over 49 different genera, but only some species in the genus Anopheles are known to transmit malaria (Foster and Walker 2019). Basic understanding of the life history of anophelines is necessary to understand their role in transmitting malaria. Like all mosquitoes, Anopheles are holometabolous insects and thus undergo four distinct stages of development: eggs, larvae, and pupae, which are aquatic stages, 3 and imago or winged adults which is non-aquatic (Romoser and Stoffolano 1998). Adult females are hematophagous, which means that they feed on blood of vertebrate hosts, primarily mammals including humans, to nourish and produce viable eggs (Black and Kondratieff 2005; Lehane 2005). After an adult male and a female Anopheles have mated, the inseminated female searches for a vertebrate host, obtains a blood meal and undergoes the egg development process (gestation). After 3–4 days of gestation it then goes in search of a suitable water body (larval habitat) to lay a batch of several hundred eggs. The eggs then hatch into first instar larvae, which undergo three more instars and a pupal stage development in the water before emerging as winged adults. The period between egg-laying and emergence of winged adults takes about 12– 15 days. Two to three days after emergence, the adult mosquitoes are ready to mate and repeat the life development cycle (Eldridge 2005; Foster and Walker 2019). While an adult male mosquito can participate in multiple mating events, a female mates only once in its lifetime, but the sperm it receives is sufficient to fertilize several batches of eggs. Every batch of eggs that a female Anopheles lays during its lifetime is preceded by a blood meal. The period between a blood meal and laying of an egg batch is called the mosquito’s gonotrophic cycle (Eldridge 2005; Foster and Walker 2019). The need for blood meals and multiple gonotrophic cycles are the key factors that enable the transmission of the malaria parasites. When a female Anopheles obtains a blood meal from an infected human it also picks up the parasites, which undergo sporogonic development inside the mosquito. When that mosquito obtains blood meals from uninfected humans during subsequent gonotrophic cycles those individuals become infected. That is how malaria is transmitted from person to person. 4 1.3. Global epidemiology of malaria 1.3.1. Geographic distribution of the parasite species Malaria is endemic in the tropical region of the world, particularly in developing countries (Hay, et al. 2004). Its restriction to the tropics is determined mostly by scantiness of the Anopheles vectors in the cold temperate regions and by inhibition of the sporogonic development in environments with temperatures < 15 °C (Dutta and Dutt 1978). However, the four malaria species vary in their geographic distribution within the tropical region. P. falciparum is the most widespread as it is found in nearly all malaria endemic regions of the world (Battle, et al. 2015; Guerra, et al. 2006; Hay, et al. 2009). Infection prevalence (proportion of surveyed humans who were infected with the parasites) of this species in children < 10 years old vary greatly from < 1% to > 70% depending on local endemicity (Battle, et al. 2015; Hay, et al. 2009). P. vivax distribution closely resembles that of P. falciparum in all continents except Africa where it is restricted to Madagascar and East African countries (Battle, et al. 2015; Guerra, et al. 2006; Howes, et al. 2016). Its absence in most parts of Africa is due to lack of P. vivax-specific Duffy binding receptor on the surface of the red blood cells of indigenous Africans which prevents the parasite from binding and infecting the red blood cells (Howes, et al. 2011; Rosenberg 2007). However, a recent study show that P. vivax is now spreading into other parts of Africa because of introduction of the Duffy-positive alleles into the human populations (Howes, et al. 2015). The infection prevalence of P. vivax in individuals of all ages vary from under 0.5% to above 10% depending on local endemicity (Battle, et al. 2015; Battle, et al. 2019; Howes, et al. 2016). P. malariae has been observed in most of the malaria endemic areas of the world but unlike P. falciparum and P. vivax, it appears patchy in its distribution with infection prevalence that rarely exceeds 2% (Cabrera and Arambulo 1977; Ghosh and Yadav 1995; Gordon, et al. 1991; Haworth 5 1988; Kawamoto, et al. 1999; Scopel, et al. 2004). Although higher prevalence of this species (15–30%) has been observed in some parts of sub-Saharan Africa and South-West Pacific (Anthony, et al. 1992; Boudin, et al. 1991; Browne, et al. 2000; Genton, et al. 1995a; Mueller, et al. 2007; Trape, et al. 1994). P. ovale is traditionally believed to be restricted to tropical Africa, the Island of New Guinea, Indonesia and Philippines (Collins and Jeffery 2005; Lysenko and Beljaev 1969). However, it has also been reported in the Indian subcontinent (Jambulingam, et al. 1989) and some parts of South East Asia (Baird, et al. 1990; Fuehrer, et al. 2010; Win, et al. 2002). In the areas where P. ovale was reported, its infection prevalence did not exceed 5% (Baird, et al. 1990; Collins and Jeffery 2005; Fuehrer, et al. 2010; Jambulingam, et al. 1989; Lysenko and Beljaev 1969; Mueller, et al. 2007; Win, et al. 2002). 1.3.2. Geographic distribution of the primary malaria vector species There are over 40 species of Anopheles that are confirmed or suspected vectors of malaria worldwide. These vectors are not globally ubiquitous, rather each species is confined within a defined geographic range (Kiszewski, et al. 2004; Sinka, et al. 2012). Thus, the same malaria species is transmitted by different vector species in different parts of the world. For instance, the vector species Anopheles gambiae (sensu lato) (s.l.), Anopheles funestus and Anopheles arabiensis are the primary vectors in sub-Saharan Africa and are not found outside this continent (Fontenille and Lochouarn 1999; Kiszewski, et al. 2004; Meillon 1951; Sinka, et al. 2012). Interestingly, An. gambiae is more abundant than An. arabiensis in areas of sub-Saharan Africa where P. falciparum is prevalent and P. vivax is absent or less prevalent. However, An. gambiae is less abundant than An. arabiensis in Madagascar and the east African countries where P. vivax is present and more prevalent than P. falciparum (Fontenille and Lochouarn 1999; Kiszewski, et 6 al. 2004). The vectors Anopheles stephensi, Anopheles culicifacies (s.l.) and Anopheles fluviatilis (s.l.) are the main vectors confined to the Indian subcontinent (Kiszewski, et al. 2004; Sinka, et al. 2012). Those confined to South America are Anopheles darlingi, Anopheles albitarsis (s.l.), Anopheles nuneztovaris (s.l.) and Anopheles aquasalis (Kiszewski, et al. 2004; Sinka, et al. 2012). The vectors Anopheles dirus (s.l.), Anopheles minimus (s.l.), Anopheles sinensis and Anopheles balabacensis are some of the many species that are restricted to South-East Asia (Kiszewski, et al. 2004; Sinka, et al. 2011; Sinka, et al. 2012). Members of the Anopheles punctulatus (s.l.) species complex are important vectors confined to Indonesia, Australia and the Pacific island countries (Kiszewski, et al. 2004; Sinka, et al. 2011; Sinka, et al. 2012). Those confined to North America, Central America, Middle East, Europe, Western Asia and North Africa are listed elsewhere (Kiszewski, et al. 2004; Sinka, et al. 2012). 1.3.3. Malaria endemicity The epidemiology of malaria classically holds to four general categories of endemicity: holoendemic, hyperendemic, mesoendemic and hypoendemic (Boyd 1949; Bruce-Chwatt 1980; MacDonald 1957; Nkumama, et al. 2017; Spencer 1963). These classes have traditionally been bounded by the so-called “spleen rate” which is the relative proportion of children found to have distended spleens by external palpation, the former having the highest and the latter the lowest. Because the spleen produces reticulocytes to replace red blood cells; and malaria parasites destroy red blood cells; spleens have to work harder in malaria-infected people and as a result become enlarged. They are easiest to palpate in children, and children are the most vulnerable to the ill effects of malaria infection as a rule. More modern methods involve estimation of the prevalence of infection in different age groups of people as done here, by detection of parasites 7 in the blood stream using such tools as microscopic examination of thick and thin blood smears, rapid diagnostic tests for parasite specific antigens circulating in blood, or use of standard or quantitative PCR methods. Areas of the world where malaria is currently holoendemic, hyperendemic, mesoendemic or hypoendemic have been mapped and presented (Dalrymple, et al. 2015; Hay, et al. 2004). However, these epidemiological categories have much more profound implications than merely to fit into categories based on enlarged spleens or prevalence of infection. In the epidemiologic settings of any of these four categories, malaria transmission persists but its relative “stability” which is manifested in the tendency to show a stable age-specific infection prevalence profile of a certain characteristic, varies across these categories. In the holoendemic settings malaria transmission is intense and is without seasonal variation, and prevalence of infection in children £ 10 years old is > 75%. The vector population remains stable with constant high human biting rate (the frequency of vector bites per person per unit time) throughout the year (Hay, et al. 2004; Hay, et al. 2008; Spencer 1963). Hyperendemicity differs slightly from holoendemicity in that although transmission in this category is as intense as in the holoendemic settings, seasonal variation exists and infection prevalence in children £ 10 years old range from 51–75%. In this endemicity, vector population size and thus the human biting rate vary seasonally. Thus, infection prevalence levels vary depending on the time of the year the survey was conducted; high prevalence is observed when transmission is in full swing and low during quiet season (Hay, et al. 2004; Hay, et al. 2008; Spencer 1963). Common to both holoendemicity and hyperendemicity is that all human individuals are continually exposed to infectious Anopheles bites from birth and throughout their life. Mortality 8 rate is highest in infants and children under 5 years old due to lack of protective immunity in this age group compared to the older individuals (Baird 1998). Those that survive death in their childhood develop partial immunity, which is boosted throughout their adult life by repeated exposure to infective vector bites (Baird 1998; Doolan, et al. 2009). As a consequence of age- dependent protective immunity in these endemicities, the infection prevalence under usual conditions (i.e., no disease intervention implemented) is highest in the younger individuals (0–5 years in holoendemicity and 0–10 years in hyperendemicity) and dramatically declines with age (Baird 1998; Doolan, et al. 2009). Immunity is lost or wanes when individuals migrate out of the high transmission area for long period of time or when vector-human contact is interrupted long term by factors such as bed nets. In mesoendemic settings, transmission intensity is moderate with more pronounced seasonal variation, and infection prevalence in children £ 10 years ranges from 11–50%. Vector population density is low most of the time and only increases during the rainy season. Unlike the two preceding endemicities, immunity in adults is weakly developed and thus age-specific prevalence of infection and risk of disease is shifted into even older age groups but mortality is still higher among children (Hay, et al. 2004; Hay, et al. 2008; Spencer 1963). In hypoendemic settings, atmospheric temperature is either low (< 15 °C), such as high altitudes, or high (> 30 °C), which prevents transmission most of the time (Dutta and Dutt 1978; Spencer 1963). Epidemics occur only occasionally, particularly during heavy rainfall. The infection prevalence in children £ 10 years can reach as high as 10% but is usually 0%. The vectors are difficult to find and occasionally increase in abundance during heavy rainfall. In this endemicity, all age groups are more or less equally susceptible to infection and disease attack due to lack of immunity (Hay, et al. 2004; Hay, et al. 2008; Spencer 1963). 9 1.3.4. Public health burden and control of malaria P. falciparum is not only widely distributed and highly prevalent but is also the most virulent of all the malaria species. Most infections by this species result in mild clinical illnesses such as high fevers and headaches. However, a small proportion of the infected individuals develops severe clinical conditions such as placental malaria, severe anemia, acidosis, respiratory distress, shock, renal failure, pulmonary edema, and cerebral pathology with convulsion and coma (Mendis and Carter 1995; Milner 2018). The other three species afflict intermittent high fevers, headaches and other mild illnesses but are rarely fatal (Mendis and Carter 1995; Milner 2018). As previously discussed, P. ovale and P. malariae have patchy or limited geographic distribution with consistently low (< 5%) prevalence of infection. In contrast, P. vivax is widely distributed with relatively high prevalence of infection and constitutes a significant amount of morbidity in humans. Thus, P. falciparum and P. vivax are both important malaria species from the standpoint of public health. According to the World Health Organization (WHO), 40% of the world’s population is at risk of malaria infection (Toure and Oduola 2004). The most vulnerable demographic groups are children under five years old, pregnant women, immunocompromised individuals, non-immune immigrants, mobile populations, and travelers. The most recent data from WHO show that in the year 2018 alone, a total of 228 million infection cases and 405,000 deaths due to malaria were reported worldwide. Most of the reported cases of infection (93%) and deaths (94%) were from the continent of Africa. Of the 405,000 reported deaths, 67% were children under 5 years old. Most of the infection cases were caused by P. falciparum; only 3.3% were by P. vivax (WHO 2019). 10 Like many other diseases, it is likely that the most effective way to control or eliminate malaria would be by means of a vaccine. However, no malaria vaccine is available at present and research is still ongoing in many laboratories around the world to develop one. Although antimalarial drugs are available, the primary means for controlling malaria in many endemic countries is through the use of insecticides that target the mosquito vectors. The insecticide- based approach for controlling the vectors is implemented in two ways: indoor residual spray and insecticidal bed nets. The indoor residual spray method involves spraying surfaces inside houses with insecticides so that mosquitoes that enter the house to feed on humans are killed when they contact the treated surfaces. The insecticidal bed nets serve as barriers, protecting humans who sleep under one from being exposed to the mosquitoes. Mosquitoes that contact the nets are killed by insecticides impregnated in the bed net fabric. By reducing mosquito abundance and frequency of bites taken on humans, the insecticide-based vector control methods interrupt malaria transmission. Although the insecticide-based methods have reduced the worldwide burden of malaria significantly (Eisele, et al. 2010; Lengeler 2004; Pluess, et al. 2010), emergence of physiological resistance to the insecticides in vector populations around the world (Quinones, et al. 2015; WHO 2016) has impeded malaria elimination and paved the way for malaria rebound. In response to this problem, novel vector control methods such as population suppression by Wolbachia-induced cytoplasmic incompatibility (Alphey 2014; Bourtzis, et al. 2014), population suppression by release of mosquitoes with dominant lethal genes (Alphey 2014; Valdeza, et al. 2011) and population replacement of wild-type mosquitoes with pathogen-refractory strains through gene-drive systems (Alphey 2014; Sinkins and Gould 2006) are being developed. Although some of these methods have proven to be successful against non-anopheline vectors of 11 arboviruses (Carvalho, et al. 2015; Harris, et al. 2012), they have not yet been demonstrated to work effectively against natural populations of Anopheles vectors of malaria. Until the novel methods are available, malaria control in many parts of the world will continue to rely on the insecticide-based methods. 1.4. Malaria in Papua New Guinea 1.4.1. Landscape and geo-climatic description PNG is a tropical country constituting the eastern half of the Island of New Guinea in the South- West Pacific. The Western half of the Island (Irian Jaya) is a province of Indonesia. The mainland of PNG is divided into northern and southern lowland (0–1,200 m altitude) plains by massive, connected mountain ranges which originate from the middle of the PNG-Irian Jaya border in the west and extends to the southeastmost tip of the mainland PNG (Figure 1.1). These ranges rise beyond 2,400 m above sea level. The northern plain consists mainly of dense tropical rainforests, lowland river valleys and flood plains of large rivers with several smaller mountain ranges. The climate there is hot and wet all year round. The southern plain consists of two geo- climatic regions. The first is a large area of continually hot and wet climate condition that begins at the foot of the mountain ranges and extends towards the south coast. The vegetation and landscape there are similar to the northern plain. The second is open monsoonal savannah which occupy thin strips of areas at the southwest coast and southeastmost tip of PNG (Figure 1.1). Between the northern and southern lowland plains are highland areas (> 1,200 m altitude). The climate condition there is mostly wet with temperate to cold temperatures (Beebe and Cooper 2002). Scattered over the northern and eastern seas of PNG is a group of islands that constitute the Bismarck Archipelago. In the southeastern sea is another group of islands that constitute the 12 Louisiade Archipelago. Rainfall in most of the lowland areas, including the savannah and the islands range from 1,300 to 7,000 mm annually (Muller, et al. 2003). Because PNG is situated close to the equator, temperature is generally stable throughout the year and differs only with respect to altitude. Temperature range from 30 °C at sea level and drops with increasing altitude to as low as 0 °C beyond 2,400 m above sea level. The entire landscape of PNG consists of a patchwork of many different ecological zones clumped within an area of 460,000 km2. These ecological zones are represented in the sea by coral reefs, offshore atolls and islands. On the coast are swamps, marshes, mangroves, lakes, grasslands, and plains. Further inland are dense rainforests, mountain ranges, plateaus and valleys. Cutting across this land divide are fast-flowing rivers with flood plains and deltas that open into the sea. Not only are these ecological zones inhabited by species-rich communities of flora and fauna that form complex ecosystems, but human populations of exceptionally diverse cultural and linguistic groups. Human populations are concentrated in the highlands (above 1,200 m altitude) and in the coastal areas (below 600 m), including the islands, with sparse distribution in the intermediate zones (Muller, et al. 2003). 13 Figure 1.1. Map of Papua New Guinea showing three geo-climatic regions separated by grey dashed lines. 14 1.4.2. Anopheles vectors Anopheles species that are common in the Eastern part of Indonesia, Northern Australia and the South West Pacific region where PNG is located are the members of the An. punctulatus (s.l.) species complex, which consists of 13 phylogenetically related cryptic species (Beebe, et al. 2000; Beebe, et al. 2015; Beebe and Saul 1995). Although most of them transmit malaria (Cooper, et al. 2009), only five—Anopheles punctulatus (sensu stricto) (s.s.), Anopheles koliensis, Anopheles farauti (s.s.), Anopheles hinesorum (formerly Anopheles farauti no. 2) and Anopheles farauti no. 4—are the primary vectors driving most of the transmission in PNG (Benet, et al. 2004; Burkot, et al. 1990; Burkot, et al. 1988; Hetzel, et al. 2016; Hii, et al. 2001; Reimer, et al. 2016; Thomsen, et al. 2017). Two non-members of the punctulatus group— Anopheles longirostris and Anopheles bancroftii—are also present in PNG and have been incriminated as minor vectors (Cooper, et al. 2009). Although these vectors can be found in sympatry, each species exhibits a defined ecogeographic distribution, which makes them main vectors in some areas but less so in other areas. The larvae of An. farauti (s.s.) can tolerate high-salinity water bodies, which gives this species a competitive advantage over the other species to occupy and breed in brackish habitats (Charlwood, et al. 1986; Sweeney 1987). As a result of this adaptation, this species is most abundant in areas within 2 km from the coastline and rapidly diminishes in numbers beyond this boundary. Because of its ability to also breed in freshwater bodies, it is also found sporadically in inland lowland areas beyond 10 km from the coastline (Cooper, et al. 2002). An. punctulatus (s.s.) utilizes small, shallow, freshwater puddles that are open to direct sunlight as larval habitats. It rapidly colonizes transient water puddles shortly after they are formed either by anthropogenic activities such as tire ruts along roadsides, logging, mining, road constructions, dams, or gardens, 15 or by natural means such as stagnant riverine puddles (Charlwood, et al. 1986). An. punctulatus (s.s.) distribution is opposite to An. farauti (s.s.) in that it is abundant in inland lowland and decreases in numbers towards the coast. It is also found sporadically in the temperate environment of the highland region. Interestingly, An. punctulatus (s.s.) is widely distributed in the northern plain but in the southern plain this species is concentrated in areas to the east and is rare in areas to the west, particularly the monsoonal savannah areas (Cooper, et al. 2002). An. koliensis also utilizes the same larval habitats as An. punctulatus (s.s.) but it most commonly breeds in semi-permanent water bodies such as watercourses, ground pools and water-filled bomb craters (Charlwood, et al. 1986; Cooper, et al. 2002). Its geographic distribution is similar to that of An. punctulatus (s.s.), but it is more widespread than An. punctulatus (s.s.) in the northern plain (Cooper, et al. 2002). An. farauti no. 4 occupies the same habitats as An. punctulatus (s.s.), particularly stagnant water puddles in the flood plains of large rivers. However, unlike An. punctulatus (s.s.), it is sporadically distributed in the inland areas of the northern plain and only occasionally found in the southern plain (Cooper, et al. 2002). An. hinesorum is less restrictive and utilizes all the larval habitat types used by the other species. Although it is more abundant in the southern than northern plain, this species is the most widely distributed of all the members of the An. punctulatus (s.l.) complex (Cooper, et al. 2002). 1.4.3. Malaria epidemiology PNG is arguably the most malaria endemic country outside the African continent. All four malaria species are found with P. falciparum and P. vivax the most prevalent species (Dijk and Parkinson 1973; Genton, et al. 1995a; Genton, et al. 1995b; Mehlotra, et al. 2002; Muller, et al. 2003). Prior to the implementation of a nationwide malaria control campaign (Hetzel 2009; 16 Hetzel, et al. 2014a; Hetzel, et al. 2012), malaria in the lowland areas (below 600 m altitude) of PNG where over 60% of the country’s population (8.6 million people based on 2018 census; https://sdd.spc.int/pg) reside was holoendemic or hyperendemic (Cattani, et al. 1986a; Cattani, et al. 1986b; Dijk and Parkinson 1973; Genton, et al. 1995a; Genton, et al. 1995b; Mehlotra, et al. 2002; Muller, et al. 2003; Parkinson 1973). Malaria was hypoendemic with occasional epidemic outbreaks in the colder areas of the highland region where 40% of the population reside (Mueller, et al. 2005; Muller, et al. 2003). In some pockets of the most holoendemic parts of Madang and East Sepik provinces, infection prevalence of Plasmodium spp. among children under 10 years old reached as high as 78% and accounted for as much as 17% of deaths in children (Muller, et al. 2003). The rate of transmission of Plasmodium spp. was as high as 1,000 sporozoite-infective Anopheles spp. bites per person-year (Burkot, et al. 1988; Reimer, et al. 2016). To alleviate the public health and socioeconomic burden of malaria in PNG, the National Department of Health embarked on a nationwide malaria control campaign, supported by the Global Fund and Rotary Against Malaria (Hetzel 2009). This program, which began in 2004 and intensified in 2006, focused primarily on free distribution of long-lasting insecticidal bed nets in rural villages throughout PNG and was supplemented with artemisinin-based antimalarial drugs and rapid antigen diagnostic test made available in local health centers (Hetzel 2009; Hetzel, et al. 2014a; Hetzel, et al. 2014b). By 2014, the national average of household bed net ownership (i.e., proportion of surveyed households owning at least one bed nets) and usage (proportion of interviewed individuals who reportedly sleep under a bed net) surpassed the target coverage of 80% compared to < 1% before the campaign. The number of people seeking antimalarial 17 treatments also increased from < 1% in 2009 to 78% in 2014, surpassing the target of 60% coverage (Hetzel, et al. 2014a). As a consequence of the disease control campaign, considerable reduction in malaria was observed in PNG. Nationally, the overall (all ages) prevalence of infection for Plasmodium spp. was reduced from an average of 11% in 2009 to 0.9% in 2014 (Hetzel, et al. 2017). In the north coast of Madang province where malaria was holoendemic, the overall prevalence of infection for P. falciparum was reduced from 42.1% in 2006 to 9% in 2014. P. vivax also dropped from 41.7% in 2006 to 12.7% in 2010 but rebounded to 19.7% in 2014. In children below 10 years of age, the infection prevalence dropped from > 60% in 2006 for both malaria species to 13% P. falciparum and 30% P. vivax in 2014 (Koepfli, et al. 2017). The rate of transmission of Plasmodium spp. in the north coast of Madang dropped from a range of 20–350 sporozoite- infective Anopheles spp. bites per person-year in 2009 to 0–15 infective bites per person-year in 2011. In the East Sepik province, the rate of transmission dropped from 175–1,050 in 2009 to 0– 25 infective bites per person-year in 2011 (Reimer, et al. 2016). 1.5. Research rationale and objective 1.5.1. Rationale Along the malaria endemicity continuum from holoendemic to hypoendemic, the overall population prevalence of infection declines as transmission rates are lower. In addition, the epidemic potential of malaria increases with this continuum as the extent of herd immunity declines as well. That is to say, malaria becomes less and less stable and more prone to epidemics with this continuum. The variations in stability of malaria are important considerations for the process of malaria control. Paradoxically, implementation of malaria 18 control such as through insecticidal interventions against vector populations particularly in the holoendemic and hyperendemic settings destabilizes malaria epidemiologically, increases the potential for epidemics, and cause changes in age-specific prevalence and parasite-host associations that result in higher risk of disease and death in certain age groups. For example, chronic anemia declines in children as parasite density goes down with fewer and fewer inoculations of parasites annually, but the frequency of severe neurologic outcomes such as cerebral malaria increases due to waning clinical immunity (Nkumama, et al. 2017). Attack rates increase in older children during their primary formative years in school (Nkumama, et al. 2017). Productive labor of the older age classes is more affected (Nkumama, et al. 2017). Therefore, malaria control implemented in malaria-stable populations does not necessarily follow the Hippocratic dictum “first do no harm” whether considered from the individual patient or that of the community, especially when it does not ensure complete elimination of the disease or prevent it from being reintroduced into the vulnerable population. The reduction of malaria transmission rate and infection prevalence after the vector control campaign in the north coast of Madang (Koepfli, et al. 2017) indicates that the epidemiology of malaria in this region shifted from stable (holoendemic) to unstable (mesoendemic) state. This means that immunity has waned in previously immune individuals, creating a condition where individuals of all ages are now vulnerable to malaria infection and disease attacks then before the bed net campaign when only the children (and pregnant women) were vulnerable. To minimize disease attacks in the vulnerable populations, the bed net distribution program was intensified and extended to the present by the Rotarians Against Malaria (https://ram.rawcs.com.au/) to further suppress the malaria transmission rate to accomplish near zero prevalence of infection. However, an important challenge facing this goal 19 is that the relationship between transmission as measured by the infectious biting rate or a surrogate (the entomological inoculation rate) is nonlinear such that large declines in transmission rate result in small changes in prevalence of infection (Smith et al. 2005). Accordingly, greater effort is required to reduce transmission rate to low levels meaningfully and sustainably, and to consolidate those gains to prevent a spring-like rebound and even epidemic (Cohen et al. 2012). Indeed, a stark lesson of malaria control programs is that a failure to consolidate gains in reduction in prevalence can increase the potential energy for rebound from such as hypoendemic or mesoendemic conditions to hyperendemic or holoendemic conditions (Cohen et al. 2012). Based on most recent observation, the rate of malaria patients visiting the local health centers in Madang province remained high despite the ongoing efforts to intensify the bed net campaign in this region. This observation points to the plausibility that further reduction of malaria transmission is being hampered and is most likely caused in large part by ecological factors that enable the mosquito vectors to evade the bed nets and continue to transmit the disease. Emergence of physiological resistance to the insecticides in the bed net fabric is one potential factor. However, this trait has not been observed in the vector populations in PNG (Katusele, et al. 2014; Keven, et al. 2010; Koimbu, et al. 2018), which implies that factors other than insecticide resistance are limiting the effectiveness of the bed nets. The situation where malaria transmission continues to persist after it was reduced by a vector control method such as bed nets is termed “residual transmission” (Killeen 2014). Residual transmission indicates the natural or biological limit of the impact of a vector control method on malaria reduction such that once the limit is reached further effort to scale up or intensify the control method will yield negligible or no further impact. At this juncture, 20 incorporation of complementary vector intervention methods is necessary to help reduce malaria further. The importance of residual transmission is its potential to cause disease resurgence when the vector control program is weakened or withdrawn as was the case for many endemic countries in the past (Cohen, et al. 2012). 1.5.2. Objective The objective of the research work described in this dissertation was to address the overarching hypothesis that in PNG further reduction of malaria transmission by ongoing bed net intervention program is hampered by ecological factors of the local vector populations that enable the mosquitoes to evade the bed nets. Thus, malaria transmission is currently persisting at higher rates compared to recent estimates. This objective was addressed by investigating five specific hypotheses outlined below. Hypothesis 1 (addressed in chapter 2). The local Anopheles mosquitoes have high abundance and species diversity despite the presently ongoing insecticidal bed net program, and the mosquito communities are highly dissimilar in their species composition even between those communities that are in close spatial proximities to each other. Hypothesis 2 (addressed in chapter 3). As a result of high vector abundance and species diversity (hypothesis 1), and ecological factors which enable the local vectors to evade the bed nets (hypotheses 3–5), the rates of transmission and prevalence of infection of malaria have rebounded rather the declined further in some areas in PNG. 21 Hypothesis 3 (addressed in chapter 4). Host-seeking mosquitoes evade the bed nets by entering the village and feeding on human hosts in the evening when people are active and disproportionately outdoors unprotected by the bed nets, thus the risk of contacting the indoor- deployed bed nets is minimal. Hypothesis 4 (addressed in chapter 5 and 6). The primary vector species in PNG feed unbiasedly on humans and non-human (pigs and dogs) hosts rather than exhibit strict anthropophagy. As a result of the opportunistic utilization of the blood hosts, the vectors are able to avoid contact with the bed nets by feeding on non-humans to support vector abundance and transmit malaria by feeding on humans outdoors and during the periods when they are awake and unprotected by the bed nets. Hypothesis 5 (addressed in chapter 4 and 7). Within villages, the per-person frequency of vectors bites clusters spatially and among human individuals rather than being randomly distributed among individuals, causing transmission to persist rather than dissipate by increasing the basic reproduction rate of malaria. 1.6. Research ethics approval All research activities described in this dissertation were reviewed for compliance with ethical requirements by appropriate authorities. The study involved consented recruitment of human volunteers to (i) donate capillary blood samples (250 µl) for PCR detection of malaria infection and human microsatellite genotyping or DNA profiling, (ii) obtain information on their age, sex, household geolocation and possession of domestic animals, (iii) interview for whether they did 22 or did not sleep under bed nets, and (iv) obtain antimalarial chemoprophylaxis to serve as a mosquito collector. The protocols for acquisition of individual consent, collection of blood samples, and questionnaire survey were approved by the institutional review board (IRB) of PNG Institute of Medical Research (IRB No. 1715), PNG Medical Research Advisory Committee (IRB No. 16.08) (https://www.pngimr.org.pg/about/institutional-review-board/) and Michigan State University (IRB No. 17-1603) (https://hrpp.msu.edu/contacts/about/irb- office.html). The letter of approval from the Michigan State University IRB has been filed with the university’s graduate school as part of the dissertation submission and approval process. 23 REFERENCES 24 REFERENCES Alphey L. 2014. Genetic control of mosquitoes. Annu. Rev. Entomol. 59:205–224. Anthony RL, Bangs MJ, Hamzah N, Basri H, Purnomo, Subianto B. 1992. Heightened transmission of stable malaria in an isolated population in the highlands of Irian Jaya, Indonesia. Am. J. Trop. Med. Hyg. 47(3):346–356. Antinori S, Galimberti L, Milazzo L, Corbellino M. 2012. Biology of human malaria Plasmodia including Plasmodium knowlesi. Mediterr. J. Hematol. Infect. Dis. 4(1). Baird JK. 1998. Age-dependent characteristics of protection v. susceptibility to Plasmodium falciparum. Ann. Trop. Med. PH. 92(4):367–390. Baird JK, Purnomo, Masbar S. 1990. Plasmodium ovale in Indonesia. Southeast Asian J. Trop. Med. Public Health 21:541–544. Battle KE, Guerra CA, Golding N, Duda KA, Cameron E, Howes RE, Elyazar IRF, Baird JK, Reiner RC, Gething PW and others. 2015. Global database of matched Plasmodium falciparum and P. vivax incidence and prevalence records from 1985-2013. Sci. Data 2:150012. Battle KE, Lucas TCD, Nguyen M, Howes RE, Nandi AK, Twohig KA, Pfeffer DA, Cameron E, Rao PC, Casey D and others. 2019. Mapping the global endemicity and clinical burden of Plasmodium vivax, 2000–17: a spatial and temporal modelling study. Lancet 394:332– 343. Beebe NW, Cooper RD. 2002. Distribution and evolution of the Anopheles punctulatus group (Diptera: Culicidae) in Australia and Papua New Guinea. Int. J. Parasitol. 32:563–574. Beebe NW, Cooper RD, Morrison DA, Ellis JT. 2000. A phylogenetic study of the Anopheles punctulatus group of malaria vectors comparing rDNA sequence alignments derived from the mitochondrial and nuclear small ribosomal subunits. Mol. Phylogenet. Evol. 17(3):430–436. Beebe NW, Russell T, Burkot TR, Cooper RD. 2015. Anopheles punctulatus group: evolution, distribution, and control. Annu. Rev. Entomol. 60:33–350. Beebe NW, Saul A. 1995. Discrimination of all members of the Anopheles punctulatus complex by polymerase chain reaction restriction fragment length polymorphism analysis. Am. J. Trop. Med. Hyg. 53(5):478–481. Benet A, Mai A, Bockarie F, Lagog M, Zimmerman P, Alpers MP, Reeder JC, Bockarie MJ. 2004. Polymerase chain reaction diagnosis and the changing pattern of vector ecology 25 and malaria transmission dynamics in Papua New Guinea. Am. J. Trop. Med. Hyg. 71(3):277–284. Black WC, Kondratieff BC. 2005. Evolution of arthropod disease vectors. In: Marquardt WC, editor. Biology of disease vectors. 2nd ed. New York: Academic Press. p 9–23. Boudin C, Robert V, Verhave JP, Carnevale P, Ambroise-Thomas P. 1991. Plasmodium falciparum and P. malariae epidemiology in a West African village. Bull. World Health Organ. 69(2):199–205. Bourtzis K, Dobson SL, Xi Z, Rasgond JL, Calvittie M, Moreira LA, Bossin HC, Moretti R, Baton LA, Hughes GL and others. 2014. Harnessing mosquito-Wolbachia symbiosis for vector and disease control. Acta Trop. 132:150–163. Boyd MF. 1949. Malariology. A comprehensive survey of all aspects of this group of diseases from a global standpoint. Malariology. A comprehensive Survey of all Aspects of this Group of Diseases from a global Standpoint 2. Browne ENL, Frimpong E, Sievertsen J, Hagen J, Hamelmann C, Dietzs K, Horstmann RD, Burchard GD. 2000. Malariometric update for the rainforest and savanna of Ashanti region, Ghana. Ann. Trop. Med. Parasitol. 19(4):15–22. Bruce-Chwatt LJ. 1980. Essential malariology. London: William Heinemann Medical Books Ltd. Burkot TR, Garner P, Paru R, Dagoro H, Barnes A, McDougall S, Wirtz RA, Campbell G, Spark R. 1990. Effects of untreated bed nets on the transmission of Plasmodium falciparum, P. vivax and Wuchereria bancrofti in Papua New Guinea. Trans. R. Soc. Trop. Med. Hyg. 84:773–779. Burkot TR, Graves PM, Paru R, Wirtz RA, Heywood PF. 1988. Human malaria transmission studies in the Anopheles punctulatus complex in Papua New Guinea: sporozoits rates, inoculation rates, and sporozoite densities Am. J. Trop. Med. Hyg. 39(2):87–298. Cabrera BD, Arambulo PV. 1977. Malaria in the Republic of the Philippines. Acta Trop. 34(3):265–279. Carvalho DO, McKemey AR, Garziera L, Lacroix R, Donnelly CA, Alphey L, Malavasi A, Capurro ML. 2015. Suppression of a field population of Aedes aegypti in Brazil by sustained release of transgenic male mosquitoes. PLoS Negl. Trop. Dis. 9(7). Cattani JA, Moir JS, Gibson FD, Ginny M, Paino J, Davidson W, Alpers MP. 1986a. Small-area variations in the epidemiology of malaria in Madang Province. PNG Med. J. 29(1):11– 17. 26 Cattani JA, Tulloch JL, Vrbova H, Jolley D, Gibson FD, Moir JS, Heywood PF, Alpers MP, Stevenson A, Clancy R. 1986b. The epidemiology of malaria in a population surrounding Madang, Papua New Guinea. Am. J. Trop. Med. Hyg. 35(1):3–15. Charlwood JD, Graves PM, Alpers MP. 1986. The ecology of the Anopheles punctulatus group of mosquitoes from Papua New Guinea: a review of recent work. PNG Med. J. 29:19–26. Cohen JM, Smith DL, Cotter C, Ward A, Yamey G, Sabot OJ, Moonen B. 2012. Malaria resurgence: a systematic review and assessment of its causes. Malar. J. 11(122). Collins WE, Jeffery GM. 2005. Plasmodium ovale: parasite and disease. Clin. Microbiol. Rev. 18:570–581. Cooper RD, Waterson DGE, Frances SP, Beebe NW, Pluess B, Sweeney AW. 2009. Malaria vectors of Papua New Guinea. Int. J. Parasitol. 39:1495–1501. Cooper RD, Waterson DGE, Frances SP, Beebe NW, Sweeney AW. 2002. Speciation and distribution of the members of the Anopheles punctulatus (Diptera: Culicidae) group in Papua New Guinea. J. med. Entomol. 39:16–27. Cox-Singh J, Davis TME, Lee K-S, Shamsul SSG, Matusop A, Ratnam S, Rahman HA, Conway DJ, Singh B. 2008. Plasmodium knowlesi malaria in humans is widely distributed and potentially life threatening. Clin. Infect. Dis. 46:165–171. Dalrymple U, Mappin B, Gething PW. 2015. Malaria mapping: understanding the global endemicity of falciparum and vivax malaria. BMC Med. 13:140. Dijk WJOMV, Parkinson AD. 1973. Epidemiology of malaria in New Guinea. PNG Med. J. Doolan DL, Dobano C, Baird K. 2009. Acquired immunity to malaria. Clin. Microbiol. Rev. 17(1):17–21. 22(1):13–36. Dutta HM, Dutt AK. 1978. Malaria ecology: a global perspective. Soc. Sci. Med. 12:69–84. Eisele TP, Larsen D, Steketee RW. 2010. Protective efficacy of interventions for preventing malaria mortality in children in Plasmodium falciparum endemic areas. Int. J. Epidemiol. 39(1). Eldridge BF. 2005. Mosquitoes, the Culicidae. In: Marquardt WC, editor. Biology of Disease Vectors. 2nd ed. New York: Elsevier. p 95–111. Fontenille D, Lochouarn L. 1999. The complexity of the malaria vectorial system in Africa. Parassitologia 41:267–271. 27 Foster WA, Walker ED. 2019. Mosquitoes (Culicidae). In: Mullen GR, Durden LA, editors. Medical and veterinary entomology. 3rd ed. United States: Academic Press. p 261–322. Fuehrer H-P, Starzengruber P, Swoboda P, Khan WA, Matt J, Ley B, Thriemer K, Haque R, Yunus EB, Hossain SM and others. 2010. Indigenous Plasmodium ovale malaria in Bangladesh. Am. J. Trop. Med. Hyg. 83(1):75–78. Genton B, al-Yaman F, Beck HP, Hii J, Mellor S, Narara A, Gibson N, Smith T, Alpers MP. 1995a. The epidemiology of malaria in the Wosera area, East Sepik Province, Papua New Guinea, in preparation for vaccine trials. I. Malariometric indices and immunity. Ann. Trop. Med. Parasitol. 89(4):359–376. Genton B, al-Yaman F, Beck HP, Hii J, Mellor S, Rare L, Ginny M, Smith T, Alpers MP. 1995b. The epidemiology of malaria in the Wosera area, East Sepik Province, Papua New Guinea, in preparation for vaccine trials. II. mortality and morbidity. Ann. Trop. Med. Parasitol. 89(4):377–390. Ghosh A, Edwards MJ, Jacobs-Lorena M. 2000. The journey of the malaria parasite in the mosquito: hopes for the new century. Parasitol. Today 16(5):196–201. Ghosh SK, Yadav RS. 1995. Naturally acquired concomitant infections of bancroftian filariasis and human plasmodia in Orissa. Indian J. Malariol. 32:32–36. Gilles HM. 1993. The malaria parasites. In: Gilles HM, Warrell DA, editors. Bruce-Chwatt’s essential malariology. 3rd ed. London: Edward Arnold. p 12–34. Gordon DM, Davis DR, Lee M, Lambros C, Harrison BA, Samuel R, Campbell GH, Jegathesan M, Selvarajan K, Lewis GE. 1991. Significance of circumsporozoite-specific antibody in the natural transmission of Plasmodium falciparum, Plasmodium vivax, and Plasmodium malariae in an aboriginal (Orang Asli) population of Central Peninsular Malaysia. Am. J. Trop. Med. Hyg. 45(1):49–56. Guerra CA, Snow RW, Hay SI. 2006. Mapping the global extent of malaria in 2005. Trends Parasitol. 22(8):353–358. Harris AF, McKemey AR, Nimmo D, Curtis Z, Black I, Morgan SA, Oviedo MN, Lacroix R, Naish N, Morrison NI and others. 2012. Successful suppression of a field mosquito population by sustained release of engineered male mosquitoes. Nat. Biotechnol. 30(9):828-830. Haworth J. 1988. The global distribution of malaria and the present control effort. In: Wernsdorfer WH, McGregor I, editors. Malaria - Principle and Practice of Malariology: Churchill Livingstone. p 137–1420. 28 Hay SI, Guerra CA, Gething PW, Patil AP, Tatem AJ, Noor AM, Kabaria CW, Manh BH, Elyazar IRF, Brooker S and others. 2009. A world malaria map: Plasmodium falciparum endemicity in 2007. PLoS Med. 6(3). Hay SI, Guerra CA, Tatem AJ, Noor AM, Snow RW. 2004. The global distribution and population at risk of malaria: past, present, and future. Lancet Infect. Dis. 4:327–36. Hay SI, Smith DL, Snow RW. 2008. Measuring malaria endemicity from intense to interrupted transmission. Lancet Infect. Dis. 8:369–378. Hetzel MW. 2009. An integrated approach to malaria control in Papua New Guinea. PNG Med. J. 52:1–7. Hetzel MW, Choudhury AAK, Pulford J, Ura Y, Whittaker M, Siba PM, Mueller I. 2014a. Progress in mosquito net coverage in Papua New Guinea. Malar. J. 13:242. Hetzel MW, Gideon G, Lote N, Makita L, Siba PM, Mueller I. 2012. Ownership and usage of mosquito nets after four years of large-scale free distribution in Papua New Guinea. Malar. J. 11:192. Hetzel MW, Pulford J, Maraga S, Barnadas C, Reimer LJ, Tavul L, Jamea-Maiasa S, Tandrapah T, Maalsen A, Makita L and others. 2014b. Evaluation of the global fund-supported national malaria control program in Papua New Guinea, 2009-2014. PNG Med J. 57:1–4. Hetzel MW, Pulford J, Ura Y, Jamea-Maiasa S, Tandrapah A, Tarongka N, Lorry L, Robinson LJ, Lilley K, Makita L and others. 2017. Insecticide-treated nets and malaria prevalence, Papua New Guinea, 2008–2014. Bull. World Health Organ. 95:695–705. Hetzel MW, Reimer LJ, Gideon G, Koimbu G, Barnadas C, Makita L, Siba PM, Mueller I. 2016. Changes in malaria burden and transmission in sentinel sites after the roll-out of long- lasting insecticidal nets in Papua New Guinea. Parasit. Vectors 9:340. Hii JLK, Smith T, Vounatsou P, Alexander N, Mai A, Ibam E, Alpers MP. 2001. Area effects of bednet use in a malaria-endemic area in Papua New Guinea. Trans. R. Soc. Trop. Med. Hyg. 95:7–13. Howes RE, Battle KE, Mendis KN, Smith DL, Cibulskis RE, Baird JK, Hay SI. 2016. Global epidemiology of Plasmodium vivax. Am. J. Trop. Med. Hyg. 95(6):15–34. Howes RE, Patil AP, Piel FB, Nyangiri OA, Kabaria CW, Gething PW, Zimmerman PA, Barnadas C, Beall CM, Gebremedhin A and others. 2011. The global distribution of the Duffy blood group. Nat. Commun. 2:266. Howes RE, Reiner RC, Battle KE, Longbottom J, Mappin B, Ordanovich D, Tatem AJ, Drakeley C, Gething PW, Zimmerman PA and others. 2015. Plasmodium vivax transmission in Africa. PLoS Negl Trop Dis 9(11). 29 Jambulingam P, Mohapatra SS, Das LK, Das PK, Rajagopalan PK. 1989. Detection of Plasmodium ovale in Koraput district, Orissa state. Indian J. Med. Res. 89:115–116. Katusele M, Gideon G, Thomsen EK, Siba PM, Hetzel M, Reimer LJ. 2014. Longlasting insecticidal nets remain efficacious after five years of use in Papua New Guinea. PNG Med. J. 57:86–93. Kawamoto F, Liu Q, Ferreira MU, Tantular IS. 1999. How prevalent are Plasmodium ovale and P. malariae in East Asia? Parasitol. Today 15(10):422–426. Keven JB, Henry-Halldin CN, Thomsen EK, Mueller I, Siba PM, Zimmerman PA, Reimer LJ. 2010. Pyrethroid susceptibility in natural populations of the Anopheles punctulatus group (Diptera: Culicidae) in Papua New Guinea. Am. J. Trop. Med. Hyg. 83(6):1259–1261. Killeen GF. 2014. Characterizing, controlling and eliminating residual malaria transmission. Malar. J. 13:330. Kiszewski A, Mellinger A, Spielman A, Malaney P, Sachs SE, Sachs J. 2004. A global index representing the stability of malaria transmission. Am. J. Trop. Med. Hyg. 70(5):486– 498. Koepfli C, Ome-Kaius M, Jally S, Malau E, Maripal S, Ginny J, Timinao L, Kattenberg JH, Obadia T, White M and others. 2017. Sustained malaria control over an 8-year period in Papua New Guinea: the challenge of low-density asymptomatic Plasmodium infections. JID 216:1434–1443. Koimbu G, Czeher C, Katusele M, Sakur M, Kilepak L, Tandrapah A, Hetzel MW, Pulford J, Robinson L, Karl S. 2018. Status of insecticide resistance in Papua New Guinea: an update from nation-wide monitoring of Anopheles mosquitoes. Am. J. Trop. Med. Hyg. 98(1):162–165. Lehane MJ. 2005. The biology of blood-sucking in insects. New York: Cambridge University Lengeler C. 2004. Insecticide-treated bed nets and curtains for preventing malaria. Cochrane Database Syst. Rev. 2. Lysenko AJ, Beljaev AE. 1969. An analysis of the geographical distribution of Plasmodium ovale. Bull. World Health Organ. 40(3):383–394. MacDonald G. 1957. The epidemiology and control of malaria. London: Oxford University Mehlotra RK, Kasenhagen LJ, Baisor M, Lorry K, Kazura JW, Bockarie MJ, Zimmerman PA. 2002. Malaria infections are randomly distributed in diverse holoendemic areas of Papua New Guiea. Am. J. Trop. Med. Hyg. 67(6):555–562. Press. Press. 30 Meillon BD. 1951. Species and varieties of malaria vectors in Africa and their bionomics. Bull. World Health Organ. 4:419–441. Mendis KN, Carter R. 1995. Clinical disease and pathogenesis in malaria. Parasitol. Today 2(5):1–16. Milner DA. 2018. Malaria pathogenesis. Cold Spring Harb. Perspect. Med. 8. Mota MM, Pradel G, Vanderberg JP, Hafalla JCR, Frevert U, Nussenzweig RS, Nussenzweig V, Rodriguez A. 2001. Migration of Plasmodium sporozoites through cells before infection. Science 291:141–144. Mueller I, Namuigi P, Kundi J, Ivivi R, Tandrapah T, Bjorge S, Reeder JC. 2005. Epidemic malaria in the highlands of Papua New Guinea. Am. J. Trop. Med. Hyg. 72(5):554–560. Mueller I, Zimmerman PA, Reeder JC. 2007. Plasmodium malariae and Plasmodium ovale – the ‘bashful’ malaria parasites. Trends Parasitol. 23(6):278–283. Muller I, Bockarie M, Alpers M, Smith T. 2003. The epidemiology of malaria in Papua New Guinea. Trends Parasitol. 19(6):253–259. Nkumama IN, O’Meara WP, Osier FHA. 2017. Changes in malaria epidemiology in Africa and new challenges for elimination. Trends Parasitol. 33(2):128–140. Parkinson AD. 1973. Malaria in Papua New Guinea 1973. PNG Med. J. 17:8–16. Pluess B, Tanser FC, Lengeler C, Sharp BL. 2010. Indoor residual spraying for preventing malaria. Cochrane Database Syst. Rev. 14(4). Quinones ML, Norris DE, Conn JE, Moreno M, Burkot TR, Bugoro H, Keven JB, Cooper R, Yan G, Rosas A and others. 2015. Insecticide resistance in areas under investigation by the International Centers of Excellence for Malaria Research: A challenge for malaria control and elimination. Am. J. Trop. Med. Hyg. 93(3 Suppl):69–78. Reimer LJ, Thomsen EK, Koimbu G, Keven JB, Mueller I, Siba PM, Kazura JW, Hetzel MW, Zimmerman PA. 2016. Malaria transmission dynamics surrounding the first nationwide long-lasting insecticidal net distribution in Papua New Guinea. Malar. J. 15:25. Roberts LS, Janovy J. 2005. Foundations of Parasitology. New York: The McGraw-Hill Companies. Romoser WS, Stoffolano JG. 1998. The Science of Entomology. New York: McGraw-Hill. Rosenberg R. 2007. Plasmodium vivax in Africa: hidden in plain sight? Trends Parasitol. 23(5). 31 Schlagenhauf P. 2004. Malaria: from prehistory to present. Infect. Dis. Clin. North Am. 18:189– 205. Scopel KKG, Fontes CJF, Nunes AC, Horta MF, Braga EM. 2004. High prevalence of Plasmodium malariae infections in a Brazilian Amazon endemic area (Apiacas-Mato Grosso State) as detected by polymerase chain reaction. Acta Trop. 90:61–64. Sinka ME, Bangs MJ, Manguin S, Chareonviriyaphap T, Patil AP, Temperley WH, Gething PW, Elyazar IR, Kabaria CW, Harbach RE and others. 2011. The dominant Anopheles vectors of human malaria in the Asia-Pacific region: occurrence data, distribution maps and bionomic précis. Parasit. Vectors 4:89. Sinka ME, Bangs MJ, Manguin S, Rubio-Palis Y, Chareonviriyaphap T, Coetzee M, Mbogo CM, Hemingway J, Patil AP, Temperley WH and others. 2012. A global map of dominant malaria vectors. Parasit. Vectors 5:69. Sinkins SP, Gould F. 2006. Gene drive systems for insects disease vectors. Nat. Rev. Genet. Smith RC, Jacobs-Lorena M. 2010. Plasmodium–mosquito Interactions: a tale of roadblocks and detours. Adv. Insect Physiol. 39:119–149. Spencer T. 1963. The classification of malaria endemicity. Geneva: World Health Organization. Sweeney AW. 1987. Larval salinity tolerances of the sibling species of Anopheles farauti. J. Am. 7:427–435. Mosq. Control Assoc. 3:589–92. Thomsen EK, Koimbu G, Pulford J, Jamea-Maiasa S, Ura Y, Keven JB, Siba PM, Mueller I, Hetzel MW, Reimer LJ. 2017. Mosquito behaviour change after distribution of bednets results in decreased protection against malaria exposure. J. Infect. Dis. 215(5):791–797. Toure YT, Oduola A. 2004. Focus: malaria. Nat. Rev. Microbiol. 2:276–277. Trape J-F, Rogier C, Konate L, Diagne N, Bouganki H, Canque B, Legros F, Badji A, Ndiaye G, Ndiaye P and others. 1994. The Dielmo project: a longitudinal study of natural malaria infection and the mechanisms of protective immunity in a community living in a holoendemic area of Senegal. Am. J. Trop. Med. Hyg. 51(2):123–137. Tuteja R. 2007. Malaria – an overview. FEBS J. 274:4670–4679. Valdeza MRWd, Nimmo D, Betz J, Gong H-F, James AA, Alphey L, Black WC. 2011. Genetic elimination of dengue vector mosquitoes. Proc. Natl. Acad. Sci. USA 108(12):4772– 4775. Vlachou D, Schlegelmilch T, Runn E, Mendes A, Kafatos FC. 2006. The developmental migration of Plasmodium in mosquitoes. Curr. Opin. Genet. Dev. 16:384–391. 32 White NJ. 2008. Plasmodium knowlesi: the fifth human malaria parasite. Clin. Infect. Dis. 46:172–173. WHO. 2016. World malaria report 2016. Geneva: World Health Organization. WHO. 2019. World malaria report 2019. Geneva: World Health Organization. Report nr CC BY- NC-SA 3.0 IGO. Win TT, Lin K, Mizuno S, Zhou M, Liu Q, Ferreira MU, Tantular IS, Kojima S, Ishii A, Kawamoto F. 2002. Wide distribution of Plasmodium ovale in Myanmar. Trop. Med. Int. Health 7(3):231–239. Yamauchi LM, Coppi A, Snounou G, Sinnis P. 2007. Plasmodium sporozoites trickle out of the injection site. Cell. Microbiol. 9(5):1215–1222. 33 CHAPTER 2: SPECIES COMPOSITION, ABUNDANCE AND NOCTURNAL ACTIVITIES OF FEMALE ANOPHELES IN MADANG, PAPUA NEW GUINEA: ASSESSMENT WITH BARRIER SCREEN SAMPLING Publication Keven JB, Katusele M, Vinit R, Koimbu G, Vincent N, Thomsen EK, Karl S, Reimer LJ, Walker ED. 2019. Species abundance, composition, and nocturnal activity of female Anopheles (Diptera: Culicidae) in malaria-endemic villages of Papua New Guinea: assessment with barrier screen sampling. Mal. J. 18:96 2.1. Abstract Community composition of Anopheles mosquitoes, their abundance and peridomestic behavior, are important factors affecting malaria transmission. In this study, barrier screen sampling was used to investigate species composition, abundance, and nocturnal activity of Anopheles populations in villages of PNG. Mosquitoes were sampled between 18:00 hr–6:00 hr in seven villages from 2012 to 2017. The barrier screens were positioned between the village houses and the perimeter of villages where cultivated and wild vegetation (“the bush”) grew thickly. Female Anopheles that rested on either village or bush side of the barrier screens, as they commuted into and out of the villages, were captured. Relative species abundance within villages and their compositional similarity among villages were assessed. Mosquitoes captured on village and bush sides of the barrier screens were sorted by feeding status and by hour of collection, and their numbers were compared using negative binomial generalized linear models. Females of six Anopheles species were present in the sample. Species richness ranged from two to six species 34 per village, but relative abundance was highly uneven within and among villages, and community composition was generally dissimilar among villages. More blood-fed mosquitoes were captured on the side of the barrier screens facing the village than bush whereas more unfed ones were captured on the bush than village side, suggesting commuting behavior of unfed host- seeking females into the villages from nearby bush and commuting of blood-fed females away from villages towards the bush. For six of 12 Anopheles populations studied, the majority of host-seeking mosquitoes arrived in the village before midnight when people were active and unprotected from the mosquitoes by bed nets. The uneven distribution of Anopheles species among villages, with each site dominated by different species, even among nearby villages, emphasizes the importance of vector heterogeneity in local malaria transmission and control. The nocturnal activity patterns of village entry and host-seeking predominantly occurring before midnight increases the risk of human exposure to Anopheles bites. 2.2. Introduction Sampling adult female Anopheles mosquitoes is crucial for studies of bionomics, to estimate population parameters, quantify malaria and filarial parasite transmission, and evaluate vector- targeted disease interventions (Mboera 2005; Service 1977; WHO 1975). Three common sampling methods are i) the human landing catch (HLC), ii) baited or unbaited light trap (LT), and iii) resting collection (RC). These can be conducted in various locations such as within human or animal domiciles or amongst vegetation. The HLC method involves human volunteers luring and capturing host-seeking mosquitoes as they land on exposed legs (WHO 1975). The method provides direct estimates of the human biting rate (see chapter 3) and infectious biting rate (see chapter 3), and provides a means for characterizing such important bionomic properties 35 as nocturnal periodicity of the biting cycle (Service 1977; WHO 1975). The LT method involves the use of battery-powered suction devices fitted with light bulbs and/or artificial host odors to attract and trap host-seeking adult mosquitoes. The method provides estimates of relative mosquito density, indirect estimates of the infectious biting rate, and can be used to assess species diversity, community composition, relative abundance, and distribution (Amusan, et al. 2005; Cooper, et al. 2006; Dear, et al. 2018; Eisen, et al. 2008; Hii, et al. 2000; Mboera 2005; Sriwichai, et al. 2015). The RC method involves search and capture of endophilic mosquitoes settled inside human houses or animal sheds, and exophilic mosquitoes resting in the surrounding vegetation or in intentionally placed resting shelters (Mboera 2005; Odiere, et al. 2007; WHO 1975). It may involve spraying insecticides that rapidly knock down mosquitoes in indoor spaces (Odiere, et al. 2007). Such collections are used to study mosquito resting habits, recover blood- fed mosquitoes for analysis of host selection (see chapter 6), to analyze distribution in space and time, and to evaluate the effect of residual insecticide treatments on endophilic vectors (McCann, et al. 2017; WHO 1975). All the mosquito sampling methods described above have been used to study PNG vectors. Many species of Anopheles are found in PNG (Cooper, et al. 2009; Cooper, et al. 2002; Cooper, et al. 2006), and most of them have been incriminated as vectors of malaria (Cooper, et al. 2009). However, seven of these species, namely An. farauti (s.s.), An. hinesorum, An. farauti no. 4, An. punctulatus (s.s.), An. koliensis, An. longirostris and An. bancroftii, are important vectors of malaria in lowland areas of PNG, including Madang province, where malaria is highly endemic (Burkot, et al. 1988b; Hii, et al. 2001; Reimer, et al. 2016; Thomsen, et al. 2017). The first five are major vectors whereas the last two play a minor role in the transmission of malaria in PNG (Burkot, et al. 1990; Burkot, et al. 1988b; Cooper, et al. 2009; Hetzel, et al. 2016; Hii, et 36 al. 2001; Reimer, et al. 2016; WHO 2011). Various aspects of these vector species including the human biting rate, infectious biting rate, nocturnal biting cycle, dispersal range, and survival rate have been studied using one or combinations of the sampling methods described above (Burkot, et al. 1988b; Charlwood, et al. 1985a; Charlwood, et al. 1985b; Hetzel, et al. 2016; Reimer, et al. 2016; Thomsen, et al. 2017). Studies of host selection relied on indoor and outdoor RC to recover blood-fed mosquitoes (Burkot, et al. 1988a; Charlwood, et al. 1985b; Hii, et al. 2001), but this approach has serious limitations in the PNG setting. All of the vector species in PNG tend to be exophagic (Sinka, et al. 2011), thus few individuals (mostly human-fed ones) are found resting indoors, resulting in insufficient and biased samples. The wide dispersal range of some of these species along with thick tropical vegetation makes the outdoor resting search for mosquitoes in peri-domestic environments a laborious task, often resulting in very few mosquitoes that also do not adequately represent the population (Charlwood, et al. 1985b). The barrier screen sampling method was developed as an alternative to the methods discussed above (Burkot, et al. 2013; Davidson, et al. 2018). It involves the use of agricultural shade cloth positioned vertically around villages and imposing a physical barrier suitable for temporary landing and resting of Anopheles mosquitoes as they commute into and out of the villages. Mosquito collectors visit the barrier screens at specific intervals throughout the night to capture the resting mosquitoes. Unlike the indoor resting collection, both anthropophilic and zoophilic mosquitoes may be intercepted as they rest on the barrier screen. By sampling the mosquitoes as they commute into and out of a village throughout the night, the barrier screen method overcomes the laborious task associated with outdoor resting collection. The flexibility in screen placement allows sampling in various locations which reduces bias associated with particular sampling locations. The effectiveness of this method to produce a sample of blood-fed 37 mosquitoes for estimating host selection of Anopheles vectors of malaria in PNG is reported in chapter 6. The aim of the study comprising this chapter was to use the barrier screen method to analyze species composition, abundance and nocturnal movement patterns of Anopheles species. 2.3. Methods 2.3.1. Study sites This study was conducted in seven malaria-endemic rural villages (Bulal, Dimer, Kokofine, Matukar, Megiar, Mirap and Wasab) in Madang province of PNG (Figure 2.1) in 2012–2017. These villages are located in three ecogeographic environments of the malaria-endemic region of northern PNG (coastal plain, hilly inland terrain, and inland alluvial plain) known to be inhabited by various Anopheles species, including the seven vector species listed above (Burkot, et al. 1989; Burkot, et al. 1988b; Charlwood, et al. 1985b; Charlwood and Graves 1986; Charlwood and Graves 1987; Hii, et al. 1997; Hii, et al. 2001; Reimer, et al. 2016; Thomsen, et al. 2017). The villages Matukar, Megiar and Mirap share similar features; all are situated on the coastal plain along the northern coastline just above sea level. Land cover consists of coconut plantation, secondary forest, vegetable gardens, brackish swamps, houses, foot trails, and exposed soil. The villages border sand beach and the shore of the Pacific Ocean. Bulal, Dimer and Wasab villages are situated several km inland from this coastline, on elevated hilltops about 150 m above sea level, with land cover and topography of steep-sided, forested hills and numerous streams flowing to rivers in nearby valleys. The seventh village, Kokofine, is situated on the alluvial plain of the Ramu river, 39 km from the coast and 400 m above sea level. Land cover there consists primarily of swamps, cocoa plantations and secondary forests. All villages have the same tropical climate condition of hot and wet with average atmospheric temperature of 28 °C. 38 Figure 2.1. Map showing locations of the study villages. The villages are identified by names and colors: Bulal (green), Dimer (black), Kokofine (purple), Matukar (blue), Megiar (grey), Mirap (red) and Wasab (brown) in Madang Province, PNG. Areas shaded green represent land masses and unshaded represent the ocean. 2.3.2. Mosquito sampling The structure and set up of barrier screens were similar to those described elsewhere (Burkot, et al. 2013). In this study, each barrier screen consisted of a 20 m long and 2 m wide polyethylene shade cloth (forest green, 70% shading grade) fastened on wooden poles or metal reinforcement bars and erected vertically to a height of 2.15 m (Figure 2.2a). The length of the barrier screen was chosen for efficiency of setting up the screen and the amount of time spent searching the barrier screen surface for resting mosquitoes. The height of the barrier screen was consistent with the ability of the collectors to reach for mosquitoes just above their average height (160 cm). The 39 barrier screens were placed between the village perimeter and surrounding environment or bush. Each barrier screen was positioned with one side of the screen facing the village, hereafter referred to as “village side”, and the other facing away from the village, hereafter referred to as “bush side” (Figure 2.2a). The barrier screen intercepted host-seeking and blood-fed mosquitoes as they commuted into and out of the village. The intercepted mosquitoes were captured as they temporarily rested on the barrier screens (Figure 2.2b). Figure 2.2. Barrier screen setup. (a) Barrier screen erected vertically at the edge of a village with one side facing the village and the other facing the bush. (b) Mosquito collector capturing resting mosquitoes with the aid of a mouth aspirator and a flashlight. Mosquitoes were sampled using two, eight or ten barrier screens per village per night (see Table S2.1, Appendix 2.1 for the number of barrier screens deployed in a particular night in each village). Each barrier screen was assigned two trained mosquito collectors. One collector worked from 18:00 hr in the evening to midnight and was replaced by the second collector who worked from midnight to 6:00 hr in the morning. The collectors, who were stationed 20 m from the screen, visited both sides of the screen three times within each hour with approximately 20 40 minutes sampling interval, which involved five minutes commuting between the barrier screen and collector station, another five minutes searching and aspirating resting mosquitoes on the barrier screen, and 10 minutes break before next visit to the barrier screen. As this study was among the first to test the barrier screen method, no prior information was available to guide the sampling strategy. Mosquitoes were believed to rest only temporarily on the barrier screen, thus searching the barrier screen three times per hour was intended to maximize the number of mosquitoes collected. The collector walked along the barrier screen and collected resting mosquitoes with the aid of a flashlight and a mouth aspirator (Figure 2.2b). The collectors were provided with and instructed to apply mosquito repellents on their bodies to deter mosquitoes from biting them. Mosquitoes were sampled for six nights in Bulal, seven in Matukar, six in Megiar, 64 in Mirap, 63 in Wasab, 12 in Dimer and six in Kokofine during the years 2012, 2013, 2015, 2016 and 2017 (see Table S2.1, Appendix 2.1). When only two barrier screens per village per night were used, each barrier screen was changed to a new location on each night of sampling to minimize sampling biases associated with limited sampling locations. Captured mosquitoes were placed in screened paper cups labelled with the hour of the night and the side of the barrier screen (i.e., bush or village side) on which the mosquitoes were captured. Information for each mosquito, including date and hour of collection, which side of the screen it was collected and the blood-engorgement status (blood-fed or unfed) were recorded. With the aid of a light microscope, mosquitoes were sorted by sex. Males, which were very few (ca. 1% of the mosquito sample), were identified and discarded. Female Anopheles were identified to their morphological species (Belkin 1962; Lee, et al. 1987), assigned a unique serial number, and stored dry on silica gel desiccant. In the laboratory, DNA was extracted from the abdomen (blood-fed ones) or head + thorax (unfed ones) of each mosquito using DNeasy Blood 41 & Tissue Kit (Product number: 69582; Qiagen, Valencia, CA, USA). Mosquitoes morphologically identified as members of the An. punctulatus (s.l.) complex were subjected to a PCR assay (Beebe and Saul 1995) to identify the species. 2.3.3. Data analyses Variation among years in the relative abundance of different Anopheles species in a village was tested by chi-square test. Variation among villages in the number of Anopheles spp. (i.e. regardless of species) captured nightly on a barrier screen (i.e., vector abundance) was tested by Kruskal-Wallis test. Both tests were computed using the function chisq.test and kruskal.test of the package stats in R software version 3.4.1 (https://www.R-project.org/). Similarity in species composition between pairs of villages was estimated by performing Bray-Curtis index analysis (Bray and Curtis 1957) on two mosquito abundance matrices (Table 2.2 and 2.3). The resulting Bray-Curtis dissimilarity index matrices (Table S2.2 and S2.3, Appendix 2.2) were used in principal coordinates analysis (PCoA) to produce ordination plots of the study villages. The plots were further modified as biplots of villages and Anopheles species by a posteriori projection of the mosquito species onto the PCoA axes based on their weighted average scores. The weighted average score of a species at PCoA axis 1 and axis 2 was calculated by averaging the product of its abundance (see Table 2.2 and 2.3) and the PCoA axis (1 or 2) score for all villages (Borcard, et al. 2011). Villages clustered together in the biplots are similar whereas those further apart are dissimilar in their species composition. The position of a village in the PCoA biplot was influenced by the abundance of the Anopheles species closest to it (Borcard, et al. 2011). The Bray-Curtis index matrix and PCoA biplot were computed using the R functions vegdist, cmdscale and wascores of the R package vegan (Oksanen, et al. 2018). 42 In this dissertation a vector population is defined as mosquitoes of a particular species from a specific location (village) and time (year). For each of 12 Anopheles populations encountered in this study, generalized linear model (GLM) with negative binomial distribution was used to compare number of mosquitoes in collections at each of the 12 hourly periods of the night and from the village and bush side of the barrier screen. In the GLM regression equation ln($)='(+'*(+,-.)+'/(0,1.)+ln(0), the expected mosquito number µ was modelled as a linear function of the categorical predictor variables Time, with 12 levels representing the 12 hourly periods of the night and Side, with two levels representing the two sides of the barrier screens. Number of barrier screen-hours S was included as the offset term. The GLM was performed using the R function glm.nb from the package MASS (Venables and Ripley 2002). Significance level for all statistical tests was based on type I error rate of 5%. 2.4. Results 2.4.1. Species composition and abundance A total of 13,897 female Anopheles mosquitoes of six different species namely An. bancroftii, An. farauti (s.s.), An. farauti no. 4, An. koliensis, An. longirostris and An. punctulatus (s.s.) were collected, including 4,379 (31.5%) blood-fed and 9,094 (68.5 %) unfed. Distribution of the mosquitoes of each of the Anopheles species over the study villages is shown in Table 2.1 (for all years combined), Table 2.2 (for April–December 2012) and Table 2.3 (for August 2016–March 2017). Data in Table 2.2 and 2.3 were used as mosquito abundance matrices in the Bray-Curtis species composition similarity analysis (see data analyses section). 43 Table 2.1. Number of mosquitoes collected in each of seven villages for all years combined. Bulal Dimer Kokofine Matukar Megiar Mirap Wasab Species 0 An. bancroftii 8 An. farauti (s.s.) 0 An. farauti no. 4 90 An. koliensis 2 An. longirostris An. punctulatus (s.s.) 34 0 0 1624 32 9 67 3 93 0 1759 361 816 11 6 0 9 16 179 0 141 0 2 28 19 0 1120 0 19 0 0 363 4502 3 2232 189 160 Table 2.2. Number of mosquitoes collected in each of five villages in April–December 2012. Species An. bancroftii An. farauti (s.s.) An. farauti no. 4 An. koliensis An. longirostris An. punctulatus (s.s.) Dimer 11 6 0 9 16 179 Kokofine Matukar Mirap 0 0 1520 0 0 0 127 1775 0 11 45 60 0 141 0 2 28 19 Wasab 3 4 0 13 275 147 Table 2.3. Number of mosquitoes (excluding the species An. farauti no. 4) collected in each of four villages in August 2016–March 2017. Mirap 140 945 2193 112 79 Wasab 0 15 1150 21 431 Species An. bancroftii An. farauti (s.s.) An. koliensis An. longirostris An. punctulatus (s.s.) Bulal 0 8 90 2 34 Megiar 0 1120 19 0 0 44 The relative Anopheles species abundance expressed in terms of mean proportion ± standard error collected on barrier screens per night in each village is shown in Figure 2.3. The results presented in Figure 2.3 and Tables 2.1–2.3 show that the Anopheles species were not evenly represented within and among the villages. Figure 2.3a shows the relative proportion of mosquitoes in villages where sampling was done within a single year: Dimer (2012), Matukar (2012), Bulal (2017) and Megiar (2017). In Dimer, where five different species were found, An. punctulatus (s.s.) constituted 66.5 ± 9% of the mosquito sample. In Bulal, where four species were found, An. koliensis constituted 66.6 ± 7% of the sample. An. farauti (s.s.) constituted 69.0 ± 7% of samples consisting of four species in Matukar, and 98.0 ± 1% of samples consisting of two species in Megiar. Mosquitoes were sampled in more than one year in Kokofine (two years), Wasab (four years) and Mirap (four years). In all three villages and sampling years, the Anopheles species were not evenly represented; one or more species predominated the samples whereas others were rare or not detected (Figure 2.3b–d). In Kokofine where four species were encountered (Table 2.1), the species composition changed from solely An. farauti no. 4 in 2012 to a more diverse community of four species in 2016, although An. farauti no. 4 still constituted most (49.0 ± 8%) of the 2016 sample (Figure 2.3b). The relative abundance of these four species in 2012 compared to 2016 was statistically different (chi-square test: c2 = 825.8, df = 3, P < 0.0001). In Wasab, five species were encountered (Table 2.1) and their relative abundance over the four years (Figure 2.3c) varied statistically (chi-square test: c2 = 2041, df = 12, P < 0.0001). In 2012, An. longirostris was the most abundant species (63.4 ± 6%) followed by An. punctulatus (s.s.) (32.0 ± 6%). In 2013, An. farauti (s.s.) increased in numbers, making it third most abundant species (26.3 ± 10%) after An. punctulatus (s.s.) (31.8 ± 12%) and An. longirostris (35.5 ± 13%). A 45 notable shift in the relative abundance was observed in 2015 where An. longirostris dropped to third place (12.3 ± 7%) whereas An. koliensis (28.7 ± 6) and An. punctulatus (s.s.) (48.3 ± 8%) rose to second and first place respectively. By 2016, An. koliensis was the most abundant species in Wasab, constituting 78.0 ± 2% of the sample, followed by An. punctulatus (s.s.) at 20.0 ± 2% (Figure 2.3c). The most striking trend observed in Wasab was the monotonic drop in An. longirostris from the most abundant species in 2012 to the third most abundant in 2016, and the rise of An. koliensis from the third most abundant species in 2012 to the most abundant in 2016 (Fig. 2.3c). In Mirap, six species were encountered (Table 2.1) and like Kokofine and Wasab, the relative abundance of these species over the four years (Figure 2.3d) differed statistically (chi- square test: c2 = 3678.6, df = 15, P < 0.0001). However, in the first three years, this variation was not as visually dramatic as in Wasab because the change in relative species abundance happened only in the less abundant species (i.e. those < 14%), while the proportion of the predominant species An. farauti (s.s.) over the three years remained above 76% (Figure 2.3d). A dramatic shift occurred in 2017 when An. koliensis whose abundance was below 2% in the preceding years rose to dominance, constituting 63.1 ± 2% of the sample, followed by An. farauti (s.s.) at 27.0 ± 2%. Anopheles density, measured as mean number of mosquitoes captured per barrier screen- night, varied among villages. For villages where mosquitoes were collected in the year 2012, Kokofine had the highest density (190 Anopheles per screen-night), followed by Mirap (36), Matukar (14), Wasab (11), and Dimer (10) (Figure 2.4a). This variation was statistically significant (Kruskal-Wallis test: c2 = 26.9, df = 4, P < 0.0001). For villages where mosquitoes were collected in the year 2016–2017, Mirap had the highest density (72), followed by Megiar 46 (24), Wasab (18), and Bulal (3) (Figure 2.4b). This variation was also statistically significant (c2 = 21.1, df = 3, P < 0.0001). Figure 2.3. Relative species abundance of Anopheles mosquitoes. Each bar represents mean proportion (with standard error bar) of each Anopheles species captured on barrier screens per night in a village. (a) Relative abundance of Anopheles in four villages where mosquitoes were sampled in a single year. (b) Relative abundance of Anopheles in Kokofine in two different years. (c) Relative abundance of Anopheles in Wasab in four different years. (d) Relative abundance of Anopheles in Mirap in four different years. 47 Figure 2.4. Box plots showing the distribution of Anopheles spp. numbers collected nightly per barrier screen. (a) Distribution of Anopheles spp. collected in five villages in April– December 2012. (b) Distribution of Anopheles spp. collected in four villages in August 2016– March 2017. Blue shaded squares are the means and black horizontal lines are the medians. Ordination of the villages based on the Bray-Curtis dissimilarity matrix (Table S2.2, Appendix 2.2) derived from the analysis of April–December 2012 mosquito abundance data (Table 2.2) showed that Dimer and Wasab closely clustered together in the PCoA biplot; Matukar and Mirap also clustered together but less closely; and Kokofine was separate from the other villages (Figure 2.5a). The close grouping of the two inland villages Dimer and Wasab was influenced by the similar presence and abundance of An. koliensis, An. punctulatus (s.s.) and An. longirostris in the year 2012 (other years were excluded from the analysis due to lack of sampling in some villages in those years). Similarly, the grouping of the two coastal villages 48 Matukar and Mirap was influenced by the similar abundance and dominance of An. farauti (s.s.) in both villages. However, their cluster was not as tight as the two inland villages because of An. bancroftii, which was present and abundant in Mirap, but absent in Matukar. Kokofine differed greatly from the other villages, owing to the presence of An. farauti no. 4, which was the most abundant and exclusively found in that village (Figure 2.5a). In contrast, ordination of villages based on Bray-Curtis dissimilarity matrix (Table S2.3, Appendix 2.2) derived from the analysis of August 2016-March 2017 data (Table 3.3) showed lack of clustering (Figure 2.5b), which indicates high dissimilarity of species composition among the villages. Figure 2.5. PCoA biplots of villages (black dots) with Anopheles species (blue dots) projected on the plot based on their weighted average scores. (a) PCoA biplot derived from the Bray-Curtis analysis of species abundance data of five villages between April–December 2012 (see Table 2.2). (b) PCoA biplot derived from data of four villages between August 2016– March 2017 (see Table 2.3). The two axes PC1 and PC2 were the principal coordinates with the highest proportion (percentage) of variation. 49 2.4.2. Nocturnal activities of mosquitoes The mean number ± standard error of blood-fed and unfed mosquitoes captured on the bush and village side per barrier screen-night was plotted for each of the 12 hourly periods of the night for 12 Anopheles populations: five (four An. farauti (s.s.) and one An. koliensis) in Mirap (Figure 2.6a–e), one (An. farauti (s.s.)) in Megiar (Figure 2.6f), five (two An. koliensis, two An. punctulatus (s.s.) and one An. longirostris) in Wasab (Figure 2.7a–e) and one (An. farauti no. 4) in Kokofine (Figure 2.7f). Their corresponding GLM statistical results are shown in Figure 2.8 and 2.9. Except for Wasab-2015 An. koliensis (Figure 2.9a) and Wasab-2016 An. punctulatus (s.s.) (Figure 2.9e) which had statistically similar number of mosquitoes (both blood-fed and unfed) collected throughout the night, the other 10 populations had statistically different numbers of mosquitoes (either blood-fed, unfed or both) for one or more of the hourly periods of the night compared to 19:00 hr (Figure 2.8 and 2.9). Unfed mosquitoes for the three An. farauti (s.s.) populations in Mirap (Figure 2.8a–c), one An. farauti (s.s.) population in Megiar (Figure 2.8f), and one An. koliensis and one An. longirostris population in Wasab (Figure 2.9b and f) had statistically higher number of mosquitoes in the hourly periods before midnight and less mosquitoes in the hours after midnight, after controlling for side of barrier screens. This trend was also expressed by the blood-fed mosquitoes in some of these population but not as strongly as their unfed counterparts. For all 12 populations, the blood-fed mosquitoes captured on the village side of the screens were either statistically greater than (Figure 2.8a–d and Figure 2.9b, d and f) or equal to (Figure 2.8f and Figure 2.9 a, c and e), but not less than the bush side. In contrast, the unfed mosquitoes captured on the village side were either statistically less than (Figure 2.8b, c and f) or equal to (Figure 2.8a, d and e, and Figure 2.9a, b, c, e and f) but not 50 greater than the bush side for all populations except one (Wasab-2015 An. punctulatus (s.s.); Figure 2.9d). Figure 2.6. Mean number (dots) of blood-fed and unfed mosquitoes collected per barrier screen per night (y-axis) at each of the 12 hr periods of the night (x-axis, in 24 hr format). (a–e) Plot of five Anopheles populations in Mirap. (f) Plot of an Anopheles population in Megiar. Vertical bars are the standard error of the mean. Color black represents mosquitoes collected on the bush side of the barrier screen and grey represents those collected on the village side. 51 Figure 2.7. Mean number (dots) of blood-fed and unfed mosquitoes collected per barrier screen per night (y-axis) at each of the 12 hr periods of the night (x-axis, in 24 hr format). (a–e) Plot of five Anopheles populations in Wasab. (f) Plot of an Anopheles population in Kokofine. Vertical bars are the standard error of the mean. Color black represents mosquitoes collected on the bush side of the barrier screen and grey represents those collected on the village side. 52 Figure 2.8. Plot of GLM coefficients (open circle with 95% confidence interval bars) showing the comparison of mosquito numbers captured in each of 11 hourly nocturnal periods compared to 19:00 hr (represented by the broken vertical line centered at 0) as the reference variable and on village side of the barrier screen compared to the bush side (represented by the broken vertical line) as the reference variable. (a–e) GLM coefficients of five Anopheles populations in Mirap. (f) GLM coefficients of one Anopheles population in Megiar. The covariate levels (y axis) include eleven of the twelve hourly periods of the night (expressed in 24 h format) and village side of the barrier screen. The GLM coefficients (x axis) are in logarithmic scale. 53 Figure 2.9. Plot of GLM coefficients (open circle with 95% confidence interval bars) showing the comparison of mosquito numbers captured in each of 11 hourly nocturnal periods compared to 19:00 hr (represented by the broken vertical line centered at 0) as the reference variable and on village side of the barrier screen compared to the bush side (represented by the broken vertical line) as the reference variable. (a–e) GLM coefficients of five Anopheles populations in Wasab. (f) GLM coefficients of one Anopheles population in Kokofine. The covariate levels (y axis) include eleven of the twelve hourly periods of the night (expressed in 24 h format) and village side of the barrier screen. The GLM coefficients (x axis) are in logarithmic scale. 2.5. Discussion The effectiveness of the barrier screen method to produce an adequate and unbiased sample of blood-fed mosquitoes for estimating host selection tendencies of mosquito vectors, and to study 54 timing of host-seeking and infer flight behavior, has been reported for several Anopheles species of the southwest Pacific, including PNG in chapter 6 and elsewhere (Burkot, et al. 2013; Davidson, et al. 2018). The Anopheles fauna encountered in the seven villages using this method of sampling was limited to six species, but species composition and dominance varied considerably, with certain species dominant in some villages (e.g. An. farauti no. 4 in Kokofine, An. farauti (s.s.) in Megiar and Mirap) while completely absent (An. farauti no. 4) or present but relatively less dominant or common (An. farauti (s.s.) in Dimer and Wasab) in other villages. These findings are consistent with, and extend those, of two other studies using other sampling methods (Hii, et al. 1997; Reimer, et al. 2016). In particular, the dominance of An. farauti (s.s.) in the coastal villages was reversed by An. punctulatus (s.s.), in the nearby inland villages, although the latter was not always the most abundant species at the inland sites (Figure 2.3). Anopheles longirostris was present across all sites (except Megiar) at least in some years and typically a less common species, but was dominant in samples from Wasab for two of the four years (Figure 2.3c). These findings show that the Anopheles communities of these villages show particular patterns of dominant species yet with shifting composition over time. These shifts may be due to seasonal effects, conditions established by control efforts such as distribution of insecticidal bed nets programmatically, or random processes. The observed diversity of the species combined with variation in their ecological attributes (Burkot, et al. 1988a; Charlwood, et al. 1985b; Charlwood and Graves 1986) such as outdoor and early-evening biting habits (chapter 4), and plasticity of host selection (chapter 6) can potentially attenuate single-intervention malaria vector control programmes such as insecticidal bed nets in PNG. For example, the host selection component of this study (chapter 6) showed that An. farauti (s.s.), An. koliensis, An. longirostris and An. punctulatus (s.s.) are 55 opportunistic (can feed unbiasedly on both humans and non-human hosts) in their host selection, whereas An. bancroftii was zoophagic (feeds excessively on non-human hosts). Thus, in villages like Wasab where vector diversity is high, and the species present include those that are highly flexible in their host choice, the effectiveness of bed nets on malaria transmission will be lessened. Such village level differences in species composition, even for those locations relatively close to each other, could explain the disparity in impact of vector control on malaria observed for a country-wide bed net campaign in PNG (Hetzel, et al. 2014). Notably, An. koliensis was the most affected species whose relative abundance was dramatically reduced immediately after the bed net distribution campaign (Hetzel, et al. 2016; Reimer, et al. 2016). The consistent rise of An. koliensis abundance in recent years in Wasab and to a lesser extent Kokofine suggests, among other potential causes, a decline in the effectiveness of the bed net campaign such as reduced bed net usage or increased flexibility in alternative host utilization. For six of the 12 Anopheles populations analyzed (Mirap-2012 An. farauti (s.s.), Mirap- 2013 An. farauti (s.s.), Mirap-2015 An. farauti (s.s.), Megiar-2017 An. farauti (s.s.), Wasab-2015 An. koliensis and Kokofine-2012 An. farauti no. 4) the number of host-seeking mosquitoes (i.e. the bush side, unfed mosquitoes) arriving in the villages peaked between 20:00 hr and 22:00 hr and declined towards morning (Figure 2.6 and 2.7). This relatively early arrival suggests that the adult resting sites and larval habitats for these populations were close to the villages, resulting in short commuting time between the habitats and the village. In contrast, the number of host- seeking mosquitoes in the other six populations was extended across the evening, midnight, and morning hours suggesting that the adult resting sites for these populations ranged in distances from nearby to further from the village. Interestingly, the An. farauti (s.s.) in Mirap exhibited early arrival behavior in the year 2012, 2013 and 2015 (Figure 2.6 a–c) but in 2017 the number 56 of host-seeking An. farauti (s.s.) as well as An. koliensis arriving in the village extended throughout the night (Figure 2.6d and e). Similarly, the An. koliensis mosquitoes in Wasab exhibited extended arrival in 2015 and early arrival in 2016 (Figure 2.7a and b). The change in arrival times in latter years may be caused by changes in the spatial distribution of mosquito breeding habitats around the villages. This observation reflects an earlier study in the north coast villages of Madang which found that blood-fed individuals of An. farauti (s.l.) flew < 50 m from the study villages before resting in the nearby vegetation, whereas An. punctulatus (s.s.) and An. koliensis dispersed widely (Charlwood, et al. 1985b). Those researchers attributed this variation in dispersal among the species to the proximity of their preferred larval habitats and resting sites to the villages. Similarly, in East Sepik province, biting rates of An. farauti (s.l.) and An. longirostris were highest in the evening and declined towards morning, whereas those of An. koliensis and An. punctulatus (s.s.) were lowest in the evening and peaked in the morning hours (Hii, et al. 1997). In Kokofine an early-evening biting pattern was observed for An. farauti no. 4 (Thomsen, et al. 2017). Generally, the primary host-seeking activity of most of these Anopheles populations coincides with evening activity of villagers who would, therefore, be unprotected by bed nets. The higher proportion of blood-fed mosquitoes on the village than bush side of the barrier screen and higher proportion of unfed on the bush than village side of the screen is consistent with commuting behavior of mosquitoes. For exophilic mosquito populations like those investigated here, freshly blood-fed mosquitoes must exit the village towards their resting sites in the surrounding vegetations and are therefore likely to be intercepted by the village than bush side of the barrier screen. Similarly, unfed host-seeking mosquitoes must enter the village from 57 their resting sites to seek vertebrate hosts and are therefore likely to be intercepted by the bush than village side of the barrier screen. This study has two potential limitations. First, the presence of the mosquito collectors near the barrier screens during mosquito collections was an unavoidable aspect of this sampling method but it also introduced potential sampling bias in favor of anthropophilic vectors. To minimize this bias, the collectors applied insect repellents on their bodies and positioned themselves further from the barrier screen and among the inhabitants of the hamlets when they were not visiting the barrier screen, although these measures were not completely effective as human DNA profiling analysis show that 22–31% of human blood meal samples were collectors’ blood (see chapter 7). Second, for the villages where mosquitoes were collected for several years, the months during which mosquitoes were sampled within each village was not consistent over the years (see Table S2.1, Appendix 2.1). Also, mosquitoes were not collected at the same time in the different villages. As different species of Anopheles in PNG, particularly those within the An. punctulatus (s.l.) group, exhibit affinity for specific larval habitat types and the temporal distribution and abundance of these habitat types are associated with annual rainfall pattern (Charlwood, et al. 1986), these inconsistencies in sampling time may not capture or confound any observable temporal or spatial pattern in the species composition and abundance associated with annual rainfall season. For example, An. farauti no. 4 (whose bionomics is poorly understood but is believed to be associated with riverine puddles formed along the flood plains of Ramu River after flooding) was the only species collected in June of 2012 in Kokofine, but species richness rose to four in February 2016. Although it is possible that this change in species composition in 2016 could have resulted from a major ecological change after 2012, it is likely the result of annual seasonal variation, which was uncontrolled in this study. 58 2.6. Conclusion In PNG, Anopheles species abundance and composition varied among sites, even those that are less than one km apart. Such local heterogeneity in species composition can complicate vector control efforts in PNG. For six of the 12 Anopheles populations, the majority of the host-seeking mosquitoes arrived in the village before midnight when most people were active and exposed to the mosquitoes. In areas where Anopheles exhibit this temporal host-seeking behavior, the bed net-based vector control will be ineffective against malaria. In addition to its effectiveness in sampling blood-fed mosquitoes for analysis of their host selection tendencies, the barrier screen sampling method is useful for analyzing other aspects of the vector populations. By serving as an interception device, the barrier screen sampling system permits inferences about local movement patterns including nocturnal activity of mosquito populations. 59 APPENDICES 60 Appendix 2.1. Dates of barrier screen collection in the study villages Table S2.1. Dates of barrier screen mosquito sampling in each village in the years 2012, 2013, 2015, 2016 and 2017. Shaded cells represent the dates in which mosquitoes were sampled with the different colors correspond to each of the seven study villages. Numbers within each cell represents the number of barrier screens used to sample mosquitoes that night. Gradient shades represent villages where mosquitoes were sampled simultaneously. Year 2012 Dates January February March April May June July August September October November December 1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th 11th 12th 13th 14th 15th 16th 17th 18th 19th 20th 21st 22nd 23rd 24th 25th 26th 27th 28th 29th 30th 31st 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 Village color codes Dimer Kokofine Mirap Wasab Matukar & Wasab (simultaneous sampling) Dimer & Mirap (simultaneous sampling) Bulal Megiar 61 Table S2.1 (cont’d) Dates January February March April May June July August September October November December Dates January February March April May June July August September October November December Dates January February March April May June July August September October November December 1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th 11th 12th 13th 14th 15th 16th 17th 18th 19th 20th 21st 22nd 23rd 24th 25th 26th 27th 28th 29th 30th 31st Year 2013 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 Year 2015 2 2 2 2 2 2 2 2 2 2 2 2 2 1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th 11th 12th 13th 14th 15th 16th 17th 18th 19th 20th 21st 22nd 23rd 24th 25th 26th 27th 28th 29th 30th 31st 2 2 2 2 2 2 2 2 2 2 2 2 10 10 10 10 10 10 10 10 2 2 2 2 2 2 10 10 Year 2016 1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th 11th 12th 13th 14th 15th 16th 17th 18th 19th 20th 21st 22nd 23rd 24th 25th 26th 27th 28th 29th 30th 31st 2 2 8 8 8 8 8 8 8 8 8 8 8 8 8 8 62 Table S2.1 (cont’d) 1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th 8 8 8 8 8 8 Year 2017 11th 8 12th 8 13th 8 14th 8 15th 8 16th 8 17th 18th 19th 20th 21st 22nd 23rd 24th 25th 26th 27th 28th 29th 30th 31st 8 8 8 8 8 8 Dates January February March April May June July August September October November December 63 Appendix 2.2. Bray-Curtis index matrices Table S2.2. Bray-Curtis dissimilarity index matrix based on species abundance data for April–December 2012 sampling period. Values closer to 0 indicate villages with similar species composition and values closer to 1 indicate dissimilar villages. Dimer 0.000 1.000 0.791 0.909 0.460 Kokofine Matukar - 0.000 1.000 1.000 1.000 - - 0.000 0.828 0.832 Mirap - - - 0.000 0.900 Wasab - - - - 0.000 Dimer Kokofine Matukar Mirap Wasab Table S2.3. Bray-Curtis dissimilarity index matrix based on species abundance data for August 2016–March 2017 sampling period. Values closer to 0 indicate villages with similar species composition and values closer to 1 indicate dissimilar villages. Bulal Megiar Mirap Wasab Bulal 0.000 0.958 0.926 0.847 Megiar - 0.000 0.582 0.975 Mirap - - 0.000 0.503 Wasab - - - 0.000 64 REFERENCES 65 REFERENCES Amusan AAS, Mafiana CF, Idowu AB, Olatunde GO. 2005. Sampling mosquitoes with CDC light trap in rice field and plantation communities in Ogun State, Nigeria. Tanzan. Health Res. Bull. 7:112–116. Beebe NW, Saul A. 1995. Discrimination of all members of the Anopheles punctulatus complex by polymerase chain reaction restriction fragment length polymorphism analysis. Am. J. Trop. Med. Hyg. 53(5):478–481. Belkin JN. 1962. The mosquitoes of the South Pacific (Diptera, Culicidae). Berkeley and Los Angeles: University of California Press. Borcard D, Gillet F, Legendre P. 2011. Numerical ecology with R. New York: Springer. Bray JR, Curtis JT. 1957. An ordination of the upland forest communities of southern Wisconsin. Ecol. Monograph 27(4):325–349. Burkot TR, Dye C, Graves PM. 1989. An analysis of some factors determining the sporozoite rates, human blood indexes, and biting rates of members of the Anopheles punctulatus complex in Papua New Guinea. Am. J. Trop. Med. Hyg. 40(3):229–234. Burkot TR, Garner P, Paru R, Dagoro H, Barnes A, McDougall S, Wirtz RA, Campbell G, Spark R. 1990. Effects of untreated bed nets on the transmission of Plasmodium falciparum, P. vivax and Wuchereria bancrofti in Papua New Guinea. Trans. R. Soc. Trop. Med. Hyg. 84:773–779. Burkot TR, Graves PM, Paru R, Lagog M. 1988a. Mixed blood feeding by the malaria vectors in the Anopheles punctulatus complex (Diptera: Culicidae). J. Med. Entomol. 25(4):205– 213. Burkot TR, Graves PM, Paru R, Wirtz RA, Heywood PF. 1988b. Human malaria transmission studies in the Anopheles punctulatus complex in Papua New Guinea: sporozoits rates, inoculation rates, and sporozoite densities Am. J. Trop. Med. Hyg. 39(2):87–298. Burkot TR, Russell TL, Reimer LJ, Bugoro H, Beebe NW, Cooper RD, Sukawati S, Collins FH, Lobo NF. 2013. Barrier screens: a method to sample blood-fed and host-seeking exophilic mosquitoes. Malar. J. 12:49. Charlwood JD, Birley MH, Dagoro H, Paru R, Holmes PR. 1985a. Assessing survival rates of Anopheles farauti (Diptera: Culicidae) from Papua New Guinea J. Anim. Ecol. 54:1003– 1016. 66 Charlwood JD, Dagoro H, Paru R. 1985b. Blood-feeding and resting behaviour in the Anopheles punctulatus Donitz complex (Diptera: Culicidae) from coastal Papua New Guinea. Bull. Entomol. Res. 75:463–475. Charlwood JD, Graves PM. 1986. Capture-recapture studies with mosquitoes of the group of Anopheles punctulatus Donitz (Diptera: Culicidae) from Papua New Guinea. Bull. Entomol. Res. 76:211–227. Charlwood JD, Graves PM. 1987. The effect of permethrin-impregnated bednets on a population of Anopheles farauti in coastal Papua New Guinea. Med. Vet. Entomol. 1:319–327. Charlwood JD, Graves PM, Alpers MP. 1986. The ecology of the Anopheles punctulatus group of mosquitoes from Papua New Guinea: a review of recent work. PNG Med. J. 29:19–26. Cooper RD, Waterson DGE, Frances SP, Beebe NW, Pluess B, Sweeney AW. 2009. Malaria vectors of Papua New Guinea. Int. J. Parasitol. 39:1495–1501. Cooper RD, Waterson DGE, Frances SP, Beebe NW, Sweeney AW. 2002. Speciation and distribution of the members of the Anopheles punctulatus (Diptera: Culicidae) group in Papua New Guinea. J. med. Entomol. 39:16–27. Cooper RD, Waterson DGE, Frances SP, Beebe NW, Sweney AW. 2006. The anopheline fauna of Papua New Guinea. J. Am. Mosq. Control Assoc. 22(2):213–221. Davidson JR, Sukowati S, Shinta, Asih PBS, Syafruddin D, Baskin RN, Laurent BS, Hawley WA, Liu F, Burkot TR and others. 2018. Using barrier screens to characterize mosquito composition, flight activity, and abdominal status in South Lampung, Indonesia. Parasit. Vectors 11:440. Dear NF, Kadangwe C, Mzilahowa T, Bauleni A, Mathanga DP, Duster C, Walker ED, Wilson ML. 2018. Household‑level and surrounding peri‑domestic environmental characteristics associated with malaria vectors Anopheles arabiensis and Anopheles funestus along an urban–rural continuum in Blantyre, Malawi. Malar. J. 17:229. Eisen L, Bolling BG, Blair CD, Beaty BJ, Moore CG. 2008. Mosquito species richness, composition, and abundance along habitat-climate-elevation gradients in the Northern Colorado front range. J. Med. Entomol. 45(4):800–811. Hetzel M, Pulford J, Gouda H, Hodge A, Siba PM, Mueller I. 2014. The Papua New Guinea malaria control program: primary outcome and impact indicators 2009–2014. Papua New Guinea: Papua New Guinea Institute of Medical Research. Hetzel MW, Reimer LJ, Gideon G, Koimbu G, Barnadas C, Makita L, Siba PM, Mueller I. 2016. Changes in malaria burden and transmission in sentinel sites after the roll-out of long- lasting insecticidal nets in Papua New Guinea. Parasit. Vectors 9:340. 67 Hii JLK, Smith T, Mai A, Ibam E, Alpers MP. 2000. Comparision between anopheline mosquitoes (Diptera: Culicidae) caught using different methods in a malaria endemic areas of Papua New Guinea. Bull. Entomol. Res. 90:211-219. Hii JLK, Smith T, Mai A, Mellord S, Lewis D, Alexander N, Alpers MP. 1997. Spatial and temporal variation in abundance of Anopheles (Diptera: Culicidae) in a malaria endemic area in Papua New Guinea. J. Med. Entomol. 34(2):193–205. Hii JLK, Smith T, Vounatsou P, Alexander N, Mai A, Ibam E, Alpers MP. 2001. Area effects of bednet use in a malaria-endemic area in Papua New Guinea. Trans. R. Soc. Trop. Med. Hyg. 95:7–13. Lee DJ, Hicks MM, Griffiths M, Debenham ML, Bryan JH, Russell RC, Geary M, Marks EN. 1987. The Culicidae of the Australasian region. Debenham ML, editor. Canberra: Australian Government Publishing Service. Mboera LEG. 2005. Sampling techniques for adult Afrotropical malaria vectors and their reliability in the estimation of entomological inoculation rate. Tanzan. Health Res. Bull. 7:117–124. McCann RS, Messina JP, MacFarlane DW, Bayoh MN, Gimnig JE, Giorgi E, Walker ED. 2017. Explaining variation in adult Anopheles indoor resting abundance: the relative effects of larval habitat proximity and insecticide‑treated bed net use. Malar. J. 16:288. Odiere M, Bayoh MN, Gimnig J, Vulule J, Irungu L, Walker E. 2007. Sampling outdoor, resting Anopheles gambiae and other mosquitoes (Diptera: Culicidae) in Western Kenya with clay pots. J. Med. Entomol. 44(1):14–22. Oksanen J, Blanchet FG, Friendly M, Kindt R, Legendre P, McGlinn D, Minchin PR, O'Hara RB, Simpson GL, Solymos P and others. 2018. vegan: community ecology package. Version 2.5-3. Vienna, Austria: R Foundation for Statistical Computing. Reimer LJ, Thomsen EK, Koimbu G, Keven JB, Mueller I, Siba PM, Kazura JW, Hetzel MW, Zimmerman PA. 2016. Malaria transmission dynamics surrounding the first nationwide long-lasting insecticidal net distribution in Papua New Guinea. Malar. J. 15:25. Service M. 1977. A critical review of procedures for sampling populations of adult mosquitoes. Bull. Entomol. Res. 67:343–382. Sinka ME, Bangs MJ, Manguin S, Chareonviriyaphap T, Patil AP, Temperley WH, Gething PW, Elyazar IR, Kabaria CW, Harbach RE and others. 2011. The dominant Anopheles vectors of human malaria in the Asia-Pacific region: occurrence data, distribution maps and bionomic précis. Parasit. Vectors 4:89. 68 Sriwichai P, Karl S, Samung Y, Sumruayphol S, Kiattibutr K, Payakkapol A, Mueller I, Yan G, Cui L, Sattabongkot J. 2015. Evaluation of CDC light traps for mosquito surveillance in a malaria endemic area on the Thai-Myanmar border. Parasit. Vectors 8:636. Thomsen EK, Koimbu G, Pulford J, Jamea-Maiasa S, Ura Y, Keven JB, Siba PM, Mueller I, Hetzel MW, Reimer LJ. 2017. Mosquito behaviour change after distribution of bednets results in decreased protection against malaria exposure. J. Infect. Dis. 215(5):791–797. Venables WN, Ripley BD. 2002. Modern Applied Statistics with S. New York: Springer. WHO. 1975. Manual on practical entomology in malaria: part II methods and techniques. Geneva: World Health Organization. WHO. 2011. World Malaria Report 2011. Geneva: World Health Organization. 69 CHAPTER 3: PERSISTENT MALARIA TRANSMISSION IN HUMAN POPULATIONS WITH HIGH BED NET USAGE IN MADANG, PAPUA NEW GUINEA 3.1. Abstract In the absence of disease intervention, malaria infection prevalence in human populations in Madang province equilibrates between 35–45%, of which a large portion are children under 10 years old. As a result of the nationwide bed net-based vector control campaign, malaria dropped considerably in Madang. In this study, malaria infection prevalence in humans and human biting rates (HBR) and annual infectious biting rates (IBR) of vectors were quantified in villages on the north coast of Madang. Bed net survey, active case malaria detection, and human landing catch mosquito sampling of female Anopheles were conducted in the villages. A PCR method was used to test for malaria infection in humans and mosquitoes. The results show that in all villages, the proportion of villagers who reportedly slept under a bed net was > 80%. Overall (all villages combined) infection prevalence for P. falciparum rebounded two-fold from 9% in 2014 to 18% in this study. In contrast, P. vivax decreased from 19.7% in 2014 to 14% in this study. For both malaria species, older individuals were equally or more likely to be infected than children under 5 years old. The mean HBR ± standard error of Anopheles spp. among the villages ranged from 2.7 ± 0.6–31 ± 2.7 bites per person-night. The mean IBR for P. falciparum ranged from 6.5 ± 1.3–159.7 ± 13.7 and for P. vivax ranged from 0.0 ± 0.0–22.8 ± 1.96 infective Anopheles spp. bites per person-year. The current HBR and IBR were higher than most recent estimates. These results show that in Madang, and most likely other parts of PNG, malaria transmission is persisting at high rates despite ongoing efforts to intensify the bed net program. 70 3.2. Introduction Two epidemiological quantities are often used to assess the magnitude of malaria transmission: (i) prevalence of infection in the human population, often by age group; and (ii) annual infectious biting rate or IBR (aka annual entomological inoculation rate or EIR). Infection prevalence is the proportion of surveyed human individuals that were tested positive for malaria infection. IBR is the average number of sporozoite-infected Anopheles bites that an individual receives per year. It is quantified by the formula !"#=%"#∗'#∗365 , where HBR (human biting rate) is the average number of Anopheles bites per person-night, SR (sporozoite rate) is the proportion of Anopheles mosquitoes tested positive for sporozoite infection and 365 is the number of days in a year (Mboera 2005; Smith, et al. 2005). PNG is arguably the most malaria endemic country outside Africa. According to 2011 national census data from the PNG National Statistics Office nearly 61% of the country’s population live in the lowland areas (0–1,600 m above sea level) where malaria is holoendemic or hyperendemic and 39% occupy highland areas (> 1,600 m above sea level) where malaria is hypoendemic (Muller, et al. 2003). All four human malaria species are found in PNG, but P. falciparum and P. vivax are the most widespread and are responsible for most of the infection cases (Genton, et al. 1995a; Genton, et al. 1995b; Mehlotra, et al. 2002; Mehlotra, et al. 2000; Mueller, et al. 2005; Muller, et al. 2003). Numerous Anopheles species are found in PNG and most are known to transmit malaria (Cooper, et al. 2009). Five of these vectors namely An. farauti (s.s.), An. farauti no. 4, An. hinesorum, An. koliensis and An. punctulatus (s.s.) are the most widespread and are responsible for most of the transmission (Burkot, et al. 1989; Burkot, et al. 1990; Burkot, et al. 1988; Hii, et al. 2001; Reimer, et al. 2016; Thomsen, et al. 2017). 71 Active case malaria detection surveys in villages of Madang and East Sepik provinces in the years 1980–2001 showed that infection prevalence in these holoendemic populations (all ages) equilibrates at 35–45% but can reach as high as 70% or as low as 10% in some villages (Burkot, et al. 1990; Cattani, et al. 1986a; Cattani, et al. 1986b; Genton, et al. 1995a; Mehlotra, et al. 2002; Mehlotra, et al. 2000). As expected for holoendemic malaria settings, a large proportion of the infection cases in those villages were children < 10 years old who were immunologically susceptible; most adults were immune to the disease and thus were less likely to be infected (Burkot, et al. 1990; Cattani, et al. 1986a; Cattani, et al. 1986b; Genton, et al. 1995a; Mehlotra, et al. 2002; Mehlotra, et al. 2000). Malaria-associated mortality occurred mostly in children, accounting for 4–17% of deaths in children, but less so in adults mainly due to the acquired immunity in individuals of older ages (Genton, et al. 1995a; Moir, et al. 1989). With regard to transmission rates, a study conducted from 1983–1987 in 11 Madang villages located within an area of 20 km radius found that IBR ranged from 79–526 infective Anopheles spp. bites per person-year for Plasmodium spp., 44–293 infective bites per person-year for P. falciparum and 0–233 infective bites per person-year for P. vivax (Burkot, et al. 1988). To alleviate the burden of malaria in PNG, the National Department of Health (NDOH), with funding from the Global Fund to Fight AIDS, Tuberculosis and Malaria, embarked on a nationwide-wide malaria control campaign based on mass distribution of free long-lasting insecticidal bed nets between the years 2004 and 2009. The rationale for the insecticidal bed nets as a malaria control method is that by sleeping under it, vector-human contact is interrupted, resulting in reduced transmission of the disease. Treatment of the bed nets with insecticides is intended to kill vectors that come in contact with them, thereby reducing the vector population size and survival rate, and eventually the HBR. By the end of the NDOH-led campaign in 2009, 72 65% of households nationwide owned at least one bed net, but this figure was below the target of 80% household coverage (Hetzel, et al. 2012). Extension of the bed net distribution campaign by Rotary Against Malaria (https://ram.rawcs.com.au/) from 2009 to the present further increased the coverage to an estimated 82% of households nationwide (Hetzel, et al. 2014a). Further, to complement the bed nets, increased supplies of artemisinin-based antimalarial drugs and rapid diagnostic test kits were made available at public hospitals and local health centers to diagnose and treat malaria (Hetzel, et al. 2014b). Studies conducted before and after the bed net campaign showed considerable malaria decline in PNG. Nationally, Plasmodium spp. infection prevalence which was at 11% in the 2008–2009 national malaria survey, dropped to 5.1% in the 2010–2011 survey and further down to 0.9% in the 2013–2014 survey (Hetzel, et al. 2017). In the holoendemic north coast of Madang where bed net distribution began in 2008, the infection prevalence for P. falciparum dropped from 42.1% in 2006 to 9% in 2014. P. vivax also dropped from 41.7% in 2006 to 12.7% in 2010, but rebounded to 19.7% in 2014 (Koepfli, et al. 2017). The IBR for Plasmodium spp. in six villages in Madang north coast ranged from 20–350 infective Anopheles spp. bites per person- year in 2009 but was reduced to a range of 0–15 infective Anopheles spp. bites per person-year in 2011. In five villages in East Sepik province, the IBR for Plasmodium spp. ranged from 175– 1,050 in 2009 but was reduced to a range of 0–25 Anopheles spp. infective bites per person-year in 2011 (Reimer, et al. 2016). Although the levels of infection prevalence and IBR were lower after the bed net campaign than they were before the campaign, they were still high from the perspective of public health. The aim of the study comprising this chapter was to investigate if malaria transmission has been further suppressed in recent years by the ongoing vector control campaign or whether the disease has persisted or rebounded. 73 3.3. Methods 3.3.1. Study sites and mosquito sampling This study was conducted in the four Madang villages of Bulal, Megiar, Mirap and Wasab whose geographic locations (see Figure 2.1) and ecogeographic environments were described in detail in chapter 2. In brief, Megiar and Mirap are situated on the coastal plain along the seacoast where the brackish water-tolerant vector An. farauti (s.s.) is known to be predominant (Beebe and Cooper 2002; Charlwood, et al. 1986; Sweeney 1987). Bulal and Wasab are situated several km inland from the coastline, on elevated hilltops where the vectors An. punctulatus (s.s.) and An. koliensis are known to be predominant (Beebe and Cooper 2002; Charlwood, et al. 1986). Mosquitoes were sampled using the human landing catch method, which involved human volunteers luring host-seeking mosquitoes to land on exposed parts of their legs. The mosquitoes were captured immediately using a mouth aspirator aided by a flash light before they had time to pierce the skin and obtain a blood meal (WHO 1975). Twenty-four (Megiar, Mirap and Wasab) or 20 (Bulal) households were randomly selected to participate in mosquito sampling. For each village, 12 (Megiar, Mirap and Wasab) or 10 (Bulal) of the selected households were randomly chosen to participate in indoor collection (i.e., mosquitoes were collected inside the houses) whereas the other 12 or 10 households participated in outdoor collection. Mosquitoes were collected from 18:00 hr to 6:00 hr each night for four consecutive nights in Bulal (March 16th–19th, 2017), six in Megiar (February 4th–7th and March 2nd–3rd, 2017) and Mirap (January 11th–16th, 2017), and 12 in Wasab (September 5th–10th and November 4th–9th, 2016). Each household participated in mosquito sampling for two nights (Megiar, Mirap and Bulal) or four nights (Wasab) only. Two members (18–70 years old) of each household were consented to collect mosquitoes inside or outside their own house. The collectors were also 74 consented to take antimalarial prophylaxis before participating in the mosquito collection. In each house, one person collected mosquitoes for six consecutive hours (18:00 hr–0:00 hr) before being replaced by the second person who continued for the next six hours (0.00 hr–6:00 hr). Mosquitoes were captured by a mouth aspirator and transferred into screened paper cups labelled with the hour of the night, the household number, and the location (indoor or outdoor) where the mosquitoes were collected. With the aid of a light microscope, mosquitoes collected each night were separated into their respective genus the next morning. Male mosquitoes were discarded, and female Anopheles were identified to species based on morphological keys (Belkin 1962; Lee, et al. 1987). Each female mosquito was placed in a 2 ml microcentrifuge tube, assigned a unique serial number, and metadata (date, time, house location) were recorded. The mosquitoes were kept on silica gel desiccant in the field for up to 7 days and transported to the laboratory where they were stored at -20 °C. 3.3.2. Human blood sampling and bed net survey An active case malaria detection survey was implemented a few days before the commencement of the mosquito sampling in each village. Capillary blood samples (250 µl) were collected from each consented individual (ages 0.5–85 years old; consent for minors was obtained from their parents or guardians) and placed in a microtainer blood collection tube, which contained the anticoagulant potassium EDTA. The samples were kept on ice packs in the field and transported to the laboratory where they were stored in freezers. To estimate the frequency of bed net usage, consented individuals who participated in the active case detection survey were also asked whether they slept under a bed net the night before the interview. To estimate household 75 ownership of bed nets, a senior member of each household was asked for the number of bed nets owned by their family. 3.3.3. PCR detection of malaria in humans and mosquitoes Malaria parasite infection in individual human and mosquito samples was assessed as follows. A multiplex quantitative PCR (qPCR) assay containing two fluorescent-labelled TaqMan probes targeting the 18S rRNA gene of P. falciparum (forward primer: ATT GCT TTT GAG AGG TTT TGT TAC TTT; reverse primer: GCT GTA GTA TTC AAA CAC AAT GAA CTC AA; probe: FAM-CAT AAC AGA CGG GTA GTC AT-MGB) and P. vivax (forward primer: GCA ACG CTT CTA GCT TAA TCC AC; reverse primer: CAA GCC GAA GCA AAG AAA GTC C; probe: VIC-ACT TTG TGC GCAT TTT GCT A-MGB) was optimized for sensitivity (detected as low as one target gene copy/µl sample) and amplification efficiency (> 90%). The primers and probes were previously tested to be specific to their target DNA locus (Kamau, et al. 2013). Positive DNA controls were used for the assay optimization. Optimum reaction mixture (10 µl volume) consisted of 1x TaqMan Multiplex Master Mix (Catalog number: 4481882; Thermo Fisher Scientific, Waltham, MA, USA), 0.6 µM of each primer, 0.4 µM of each probe and 2 µl of DNA sample. PCR reactions were performed on a QuantStudio 7 Flex instrument (Applied Biosystems, Foster City, CA, USA) with fast cycling consisting of 1 cycle of 95 °C for 20 seconds followed by 40 cycles of 95 °C for 1 second and 60 °C for 20 seconds. Using sterile technique, the abdomen of each Anopheles mosquito was separated from the rest of the body. DNA was extracted from the abdomen-detached body part (i.e., head + thorax) of each mosquito and human blood sample (100 µl) using a DNeasy Blood and Tissue Kit (Catalog number: 69582; Qiagen, Valencia, CA, USA). Mosquitoes morphologically identified 76 as members of the An. punctulatus (s.l.) complex were analyzed in a PCR assay (Beebe and Saul 1995) to identify them to the species level. Both the human and mosquito samples were subjected to the qPCR method described above to test for P. falciparum and P. vivax infection. Samples with amplification threshold cycle (Cq) ≥ 38 were inconclusive and thus considered negative. As the mosquito DNA was isolated from the part of the body anterior to the abdomen, it was considered to be devoid of oocysts and other blood stages of the Plasmodium parasites which inhabit the mosquito midgut. Thus, the PCR positive samples were considered infected with the sporozoite stage of the parasite which inhabit the salivary glands in the head and thorax of mosquitoes and data used to estimate salivary gland, sporozoite infection rate. 3.3.4. Data analyses Bed net ownership was expressed as the proportion of surveyed households that owned one or more nets. Bed net usage was expressed as the proportion of interviewed individuals who slept under a net the night before they were interviewed. Malaria infection prevalence was expressed as the proportion of surveyed individuals who tested positive by qPCR. Comparison of the odds of P. falciparum or P. vivax infection between human groups used Fisher’s exact test of contingency tables or logistic regression of binary data. Both tests were performed using the functions fisher.test and glm from the package stats in R software. Significance level for all statistical tests was based on type I error rate of 5%. A vector population is defined here as mosquitoes of a particular species sampled from a specific location and time. The SR, HBR and IBR were estimated for each vector population as follows: The HBR was calculated for each sampling night by dividing the total number of Anopheles collected by landing catch in a night by the number of collectors on that night. It was 77 assumed that rate of landing catch equated to rate of biting, i.e., all landing mosquitoes were assumed to have bitten a human host had they been provided the opportunity before capture. The mean HBR was then estimated from these data points with night as unit of replication, rationalized because these nights were close together in time. The IBR was estimated for each night (see formula under section 3.2) and the mean was calculated with night as unit of replication. 3.4. Results 3.4.1. Bed net ownership and usage Census survey conducted during the study identified 97 households and 587 people living in Bulal, 172 households and 992 people in Megiar, 220 households and 1,283 people in Mirap, and 76 households and 395 people in Wasab. The proportion of the households that participated in the bed net survey was 88% in Bulal, 77% in Megiar, 58% in Mirap and 72% in Wasab. The proportion of the surveyed households that owned at least one bed net based on verbal response from the head of households was 100% in Bulal, 99% in Megiar, 100% in Mirap and 98% in Wasab. Mean number (± standard error) of reported bed nets per household was 4.3 ± 0.23 in Bulal, 4.3 ± 0.16 in Megiar, 3.3 ± 0.13 in Mirap and 3.6 ± 0.23 in Wasab. The proportion of interviewees who reportedly slept under a bed net (i.e., bed net usage) the night before they were interviewed was 93% (n = 357 interviewees) in Bulal, 81% (n = 478) in Megiar, 94% (n = 630) in Mirap and 90% (n = 256) in Wasab. 78 3.4.2. Malaria infection prevalence The proportion of the human population that consented to participate in the malaria survey was 65% (383/587 people) in Bulal, 44% (434/992) in Megiar, 34% (441/1,283) in Mirap, and 72% (285/395) in Wasab. Age range for the study participants was 1–84 years in Bulal, 1–82 years in Megiar, 0.5–79 years in Mirap and 1–66 years in Wasab. Sex ratio (male:female) of the study participants was 1:1.2 in Bulal, 1:1.6 in Megiar, 1:1.1 in Mirap and 1:1 in Wasab. The infection prevalence for Plasmodium spp. was 32% in Bulal, 36% in Megiar, 51% in Mirap, 9% in Wasab and 34% for all villages combined. Infection prevalence according to species (i.e., for P. falciparum and P. vivax separately) is shown for each village in Figure 3.1, and for all villages combined was 18% P. falciparum, 14% P. vivax and 2.5% mixed infection. The odds of infection with P. falciparum was statistically the same as that of P. vivax in the two inland villages—Bulal and Wasab (see Figure 2.1, chapter 2 for mapped location of these villages). In the two coastal villages (see Figure 2.1, chapter 2) the odds of infection was greater for P. falciparum than P. vivax in both Megiar (Fisher’s exact test: odds ratio (OR) = 1.69; 95% confidence interval (CI): 1.16–2.46; P = 0.005) and Mirap (OR = 1.59; 95% CI: 1.15–2.22; P = 0.005). The prevalence of both malaria species in each of four age groups (0–5, 6–10, 11–15 and >15 years old) is shown in Figure. 3.2. The odds of infection in the three older age groups were statistically the same or higher but not lower than the 0–5 years age group (reference group) for both malaria species in all four villages (Figure 3.3). In Bulal, Megiar and Mirap the odds of P. falciparum infection increased with age, but this trend was not observed for P. falciparum in Wasab or P. vivax in all villages (Figure 3.3). There was no statistical difference in the odds of infection among the two human sexes for any malaria species, even by age groups. 79 Figure 3.1. Prevalence of infection (all ages and sexes) of P. falciparum, P. vivax and mixed species in four villages. The number of consented human individuals tested for malaria infections was 383 in Bulal, 434 in Megiar, 441 in Mirap and 285 in Wasab. 80 Figure 3.2. Prevalence of malaria infection in four human age groups in four villages. The number of human individuals tested for malaria infections in each age group in each village was (in order of lowest to the highest age groups) as follows: Bulal (70, 73, 36, 203), Megiar (85, 69, 47, 223), Mirap (122, 76, 37, 205), Wasab (55, 61, 25, 144). One individual from Bulal, 10 from Megiar and one from Mirap who were also tested were excluded from the data due to unknown age. 81 Figure 3.3. Plot of logistic regression odds ratio (open circles or boxes) with 95% confidence interval bars for P. falciparum (black) and P. vivax (grey) infection in four human age groups in four villages. The covariates plotted on the y axis are three age groups, each was compared with 0–5 years group (not plotted) as the reference group, represented by the broken vertical line passing through odds ratio 1.0. 3.4.3. Anopheles vectors A total of 3,013 female Anopheles of four species were collected. Although more than one species was encountered in each village, their relative abundance varied (Figure 3.4). In Bulal, where three species were encountered, An. koliensis comprised most (81%) of the sample. In Megiar, where two species were encountered, An. farauti (s.s.) dominated (96%) the sample. Of 82 three species encountered in Mirap, An. farauti (s.s.) (32%) and An. koliensis (65%) were the abundant ones, and in Wasab where four species were represented in the sample, An. koliensis (77%) and An. punctulatus (s.s.) (19%) were the two most abundant ones. An. farauti (s.s.) was more abundant in the coastal villages than in the inland villages. In contrast, An. punctulatus (s.s.) was more abundant in the two inland than coastal villages. An. koliensis was the most common species in all villages except Megiar where An. farauti (s.s.) was more common. Figure 3.4. Relative abundance of four Anopheles species in four villages. The mosquito sample size was 86 in Bulal, 473 in Megiar, 1493 in Mirap and 961 in Wasab. 83 3.4.4. Sporozoite rates The P. falciparum and P. vivax SR for the 12 vector populations investigated in this study are presented in Table 3.1. The SR for Anopheles spp. in each village was also calculated and presented in Table 3.2. Both P. falciparum infected and P. vivax infected mosquitoes were detected in all villages, except in Bulal where no P. vivax infected mosquito was found (Table 3.2). Table 3.1. SR of 12 vector populations (values outside parentheses are number of sporozoite-positive mosquitoes and inside are the SR). Village Bulal Megiar Mirap Wasab Vector species An. farauti (s.s.) An. koliensis An. punctulatus (s.s.) Number tested* 6 70 10 An. farauti (s.s.) An. koliensis An. farauti (s.s.) An. koliensis An. longirostris An. farauti (s.s.) An. koliensis An. longirostris An. punctulatus (s.s.) 453 20 477 970 46 30 737 13 180 P. falciparum P. vivax 0 (0) 2 (0.0286) 0 (0) 2 (0.0044) 0 (0) 2 (0.0042) 19 (0.0196) 0 (0) 0 (0) 2 (0.0027) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 2 (0.0044) 0 (0) 2 (0.0042) 1 (0.001) 0 (0) 0 (0) 0 (0) 1 (0.0769) 0 (0) * Number of mosquitoes tested for sporozoite infection 84 Table 3.2. SR of Anopheles spp. in four villages (values outside parentheses are number of sporozoite-positive mosquitoes and inside are the SR). Village Bulal Megiar Mirap Wasab Number tested* 86 473 1493 960 P. falciparum 2 (0.0233) 2 (0.0042) 21 (0.0141) 2 (0.0021) P. vivax 0 (0) 2 (0.0042) 3 (0.002) 1 (0.001) * Number of mosquitoes tested for sporozoite infection 3.4.5. Human biting rates The mean HBR ± standard error for the 12 vector populations are shown in Figure 3.5. Within villages, the HBR varied among the vector species. In Bulal, An. koliensis had the highest biting rate (2.2 ± 0.5 per person-night) than the other two species which had < 0.3 bites per person- night. In Megiar, more of the bites were from An. farauti (s.s.) (9.4 ± 2.5) than An. koliensis (0.4 ± 0.2). In Mirap, where three species were present, An. koliensis had the highest HBR (20.2 ± 1.8) followed by An. farauti (s.s.) with two-fold less (9.9 ± 1.1) than the former species, followed by An. longirostris with < one bite per person-night. Most of the bites in Wasab were delivered by An. koliensis (6.6 ± 1.5) compared to the other three species which had < 2 bites per person- night. HBR was also calculated for Anopheles spp. in each village and the results (Figure 3.6) showed great variation among the villages. Mirap had the highest HBR (31 ± 2.7); Megiar (9.9 ± 2.7) and Wasab (8.6 ± 1.7) had similar HBR and were three-fold less than Mirap, followed by Bulal with 2.7 ± 0.6 bites per person-night. 85 Figure 3.5. Bar plots of mean HBR (y axis) with standard error bar for 12 vector populations (x axis) across four villages. Values above the bars are the estimated means. 86 Figure 3.6. Bar plots of mean HBR with standard error bar (y axis) for Anopheles spp. in four villages (x axis). Values above the bars are the estimated means. 3.4.6. Annual infectious biting rate Of the 12 vector populations, six had IBR > zero; the other six had zero IBR as a result of zero SR (Figure 3.7a). The mean IBR (± standard error) for An. koliensis in Bulal was 22.8 ± 5.4 for P. falciparum and zero for P. vivax. For An. farauti (s.s.) in Megiar, this quantity was 15.2 ± 4.0 for both malaria species. For the two populations in Mirap, An. farauti (s.s.) had 15.2 ± 1.7 for both malaria species and An. koliensis had 144.6 ± 12.6 for P. falciparum and 7.4 ± 0.6 for P. vivax. For the two populations in Wasab, An. koliensis had 6.5 ± 1.4 for P. falciparum and zero for P. vivax, and An. longirostris had 3.3 ± 1.6 for P. vivax and zero for P. falciparum. The mean IBR for Anopheles spp. ranged from zero to 22.9 for both malaria species across the four villages except for P. falciparum in Mirap where it was seven-fold higher (159.7) (Figure 3.7b). 87 Figure 3.7. Bar plots of mean IBR with standard error bar (y axis) for P. falciparum and P. vivax. (a) Mean IBR for six vector populations (x axis). (b) Mean IBR for Anopheles spp. in four villages (x axis). Values above the bars are the means. 3.5. Discussion The stability of malaria across the continuum from holoendemic to hypoendemic is described in detail in chapter 1 (section 1.3.3). The research presented in this chapter examines the dynamics of malaria stability, or lack thereof, after a long period of holoendemicity followed by approximately 15 years of bed net-based malaria intervention campaign. Prior to the bed net campaign, human communities living in the lowland areas of PNG exhibited the age-dependent 88 infection prevalence pattern which is typical for holoendemic settings in which infection rate is highest in children < 10 years old and declines monotonically with age (Burkot, et al. 1990; Cattani, et al. 1986a; Cattani, et al. 1986b; Genton, et al. 1995a; Mehlotra, et al. 2002; Mehlotra, et al. 2000). Data from nationwide longitudinal surveys conducted during and after the NDOH- led bed net campaign showed statistically high malaria infection (10–23%) in the younger age groups (< 10 years old) compared to the adults earlier in the campaign period (2008–2009), but continued to reduce through the course of the campaign nationally (Hetzel, et al. 2017). By 2014, the national infection prevalence had fallen below 1% and there was no statistical difference in the level of infection among age groups (Hetzel, et al. 2017). In villages on the north coast of Madang, the same region where the four villages in this study are located, the prevalence of infection was reduced substantially and became more homogeneous across age groups, particularly P. falciparum (Koepfli, et al. 2017). This research uncovered continued high use of insecticidal bed nets, with current household bed net ownership (³ 98% of households surveyed owned a net) and individual usage (³ 81% of people slept under a net) surpassing the 80% operational target set by the campaign program authorities (Hetzel, et al. 2012). Yet, despite high bed net coverage and usage, very high infection prevalence was observed in the village populations. The overall (all villages and ages combined) infection prevalence for P. falciparum was 18%, which was two times greater than in 2014. P. vivax infection prevalence was 14%, which was 1.4 times lower than in 2014 which was at 19.7%, but 1.1 times greater than in 2010 which was 12.7% and the lowest post-bed net prevalence on record in the north coast of Madang (Koepfli, et al. 2017). HBR and IBR were surveyed in three of the four study villages (Megiar, Mirap and Wasab) during the bed net campaign. In Mirap, the HBR for Anopheles spp. dropped from a 89 mean of 60 bites per person-night earlier in the campaign (2008–2009) to 18 bites per person- night later in the campaign (2010–2011). The IBR for Plasmodium spp. also dropped from a mean of 320 to 10 infective bites per person-year (Reimer, et al. 2016). Results of the current study showed that in this village the HBR has rebounded to a mean of 31 bites per person-night (two-fold increase) and the IBR to 183 infective bites per person-year (18-fold increase) in 2017 (Figure 3.6 and 3.7b) compared to the 2010–2011 results. In Megiar, the mean HBR dropped from 10 to five Anopheles spp. bites per person-night and IBR dropped from 150 to 10 infective Anopheles spp. bites per person-year (Reimer, et al. 2016). In the present study, the HBR has rebounded to a mean of 9.9 Anopheles spp. bites per person-night (two-fold increase) and IBR to 30.4 infective Anopheles spp. bites per person-year (three-fold increase) in 2017 (Figure 3.6 and 3.7b) compared to the 2010–2011 results. In Wasab, the mean HBR dropped from 11 to three bites per person-night and IBR dropped from 175 to 15 infective bites per person-year (Reimer, et al. 2016). Current results show that while the mean HBR rebounded by three-fold in 2016, the IBR remained low (Figure 3.6 and 3.7b). Although an entomological survey was not conducted in Bulal during the bed net campaign to compare with the result of the current study, it is likely that it also underwent the same transition as the other three villages. The comparison of HBR and IBR results of the current study to that conducted during the bed net campaign focused only on the results for Anopheles spp. in general. This was because in the previous study, both indices were not calculated separately for each vector species. In the current study, the contribution of the different vector species to malaria transmission was assessed (Figure 3.5 and 3.7a). In the two inland villages (Bulal and Wasab), An. koliensis was the primary vector responsible for most of the transmission, whereas An. punctulatus (s.s.) played a supplementary role. In the two coastal villages, An. farauti (s.s.) was present and had 90 the same nightly biting rate in both villages however, while in Mirap An. koliensis was more abundant and contributed most of the transmission. In the earlier study (Reimer, et al. 2016), it was An. punctulatus (s.s.) in inland villages and An. farauti (s.s.) in coastal villages that were the primary vectors. An. koliensis was present but played only a supplementary role in malaria transmission. Furthermore, in Madang and many parts of PNG, this species was the most affected by the bed nets; its abundance relative to other species was reduced dramatically during the campaign (Hetzel, et al. 2016; Reimer, et al. 2016). Its rise to dominance in present times is indicative of recent changes in the ecology of Anopheles vectors which are causing them to be resilient to the bed nets and causing malaria resurgence in these villages. There are two possible explanations for the high infection prevalence and IBR in the current study compared to the previous estimates. First, although malaria transmission was interrupted and reduced to low intensity by the bed nets in many parts of PNG including Madang (Hetzel, et al. 2017; Hetzel, et al. 2016; Reimer, et al. 2016), it may not have been interrupted in the four study villages. In contrast, it is possible that malaria transmission was reduced in these villages but has rebounded in recent years. The second case is more likely than the former based on the following: If transmission was never interrupted in these villages, the age-related immunity profile should be the same as the pre-campaign period (Hetzel, et al. 2017; Koepfli, et al. 2017) and the infection prevalence data in Figure 3.2 would show disproportionately higher infection in the younger age groups than the older ones. Instead, the data showed statistically similar prevalence across all age groups for P. vivax in all villages and higher prevalence in the older than younger age groups for P. falciparum in three of the four villages (Figure 3.2 and 3.3). This observation indicates that acquired immunity in all individuals waned as a result of long- 91 term interrupted transmission by the bed nets causing all individuals to become equally susceptible to infection. There were important caveats that can potentially affect the HBR and IBR estimates in this study. Unlike the cross-section malaria infection survey in human communitites, mosquito data are prone to environmental perturbations. Events such as heavy rain, strong wind, bushfire, or changes in the availability of larval habitat types occurring at any point in time can affect the number and types (i.e., species) of mosquitoes at that time. These effects are usually minimized through longitudinal mosquito sampling at several intervals over a few months or a year. In the current study, mosquitoes were collected for only four to six consecutive nights in each village. As a consequence, many of the Anopheles populations had very low numbers, which can greatly increase the uncertainty of the HBR estimates. Also, the number of sporozoite-infected mosquitoes is notoriously low (usually < 1% of mosquito sample tested for malaria infection), even in high prevalence settings; it requires testing hundreds of mosquitoes to find an infected one. A vector population with zero SR will also have zero IBR despite having relatively high HBR. Thus, given that both malaria species were highly prevalent in all villages in this study, vector populations that had zero IBR for P. falciparum, P. vivax or both cannot be considered as non-vectors. Rather, a sufficiently large sample size is needed to increase the chance of detecting sporozoite-positive mosquitoes. 3.6. Conclusion Increased bed net coverage is expected to reduce malaria levels. Therefore, the rise of bed net ownership and usage to > 80% in the villages investigated in this study should have resulted in further reduction of the infection prevalence, HBR and IBR relative to the levels observed in 92 2011. Instead, the results presented here showed clear evidence of malaria persistence or rebound despite the high bed net usage in all villages. As bed nets target the Anopheles vectors, the high malaria prevalence, HBR and IBR observed in this current study compared to those reported in recent studies indicate underlying ecological factors that rendered the vectors resilient to the bed nets. Some potential ecological factors that may underlie vector resilience include resistance to the insecticides in the bed net fabric, diversity and composition of vector species, tendency to seek and bite hosts outdoors more than indoors, tendency to bite earlier in the evening than later in the night, heterogeneity of human exposure to vector bites and opportunistic host selection behavior. Insecticide resistance has been investigated in Anopheles populations in PNG and no evidence of it has been found thus far (Katusele, et al. 2014; Keven, et al. 2010; Koimbu, et al. 2018); the other factors were investigated and are reported in the other chapters of this dissertation. 93 REFERENCES 94 REFERENCES Beebe NW, Cooper RD. 2002. Distribution and evolution of the Anopheles punctulatus group (Diptera: Culicidae) in Australia and Papua New Guinea. Int. J. Parasitol. 32:563–574. Beebe NW, Saul A. 1995. Discrimination of all members of the Anopheles punctulatus complex by polymerase chain reaction restriction fragment length polymorphism analysis. Am. J. Trop. Med. Hyg. 53(5):478–481. Belkin JN. 1962. The mosquitoes of the South Pacific (Diptera, Culicidae). Berkeley and Los Angeles: University of California Press. Burkot TR, Dye C, Graves PM. 1989. An analysis of some factors determining the sporozoite rates, human blood indexes, and biting rates of members of the Anopheles punctulatus complex in Papua New Guinea. Am. J. Trop. Med. Hyg. 40(3):229–234. Burkot TR, Garner P, Paru R, Dagoro H, Barnes A, McDougall S, Wirtz RA, Campbell G, Spark R. 1990. Effects of untreated bed nets on the transmission of Plasmodium falciparum, P. vivax and Wuchereria bancrofti in Papua New Guinea. Trans. R. Soc. Trop. Med. Hyg. 84:773–779. Burkot TR, Graves PM, Paru R, Wirtz RA, Heywood PF. 1988. Human malaria transmission studies in the Anopheles punctulatus complex in Papua New Guinea: sporozoits rates, inoculation rates, and sporozoite densities Am. J. Trop. Med. Hyg. 39(2):87–298. Cattani JA, Moir JS, Gibson FD, Ginny M, Paino J, Davidson W, Alpers MP. 1986a. Small-area variations in the epidemiology of malaria in Madang Province. PNG Med. J. 29(1):11– 17. Cattani JA, Tulloch JL, Vrbova H, Jolley D, Gibson FD, Moir JS, Heywood PF, Alpers MP, Stevenson A, Clancy R. 1986b. The epidemiology of malaria in a population surrounding Madang, Papua New Guinea. Am. J. Trop. Med. Hyg. 35(1):3–15. Charlwood JD, Graves PM, Alpers MP. 1986. The ecology of the Anopheles punctulatus group of mosquitoes from Papua New Guinea: a review of recent work. PNG Med. J. 29:19–26. Cooper RD, Waterson DGE, Frances SP, Beebe NW, Pluess B, Sweeney AW. 2009. Malaria vectors of Papua New Guinea. Int. J. Parasitol. 39:1495–1501. Genton B, al-Yaman F, Beck HP, Hii J, Mellor S, Narara A, Gibson N, Smith T, Alpers MP. 1995a. The epidemiology of malaria in the Wosera area, East Sepik Province, Papua New Guinea, in preparation for vaccine trials. I. Malariometric indices and immunity. Ann. Trop. Med. Parasitol. 89(4):359–376. 95 Genton B, al-Yaman F, Beck HP, Hii J, Mellor S, Rare L, Ginny M, Smith T, Alpers MP. 1995b. The epidemiology of malaria in the Wosera area, East Sepik Province, Papua New Guinea, in preparation for vaccine trials. II. mortality and morbidity. Ann. Trop. Med. Parasitol. 89(4):377–390. Hetzel MW, Choudhury AAK, Pulford J, Ura Y, Whittaker M, Siba PM, Mueller I. 2014a. Progress in mosquito net coverage in Papua New Guinea. Malar. J. 13:242. Hetzel MW, Gideon G, Lote N, Makita L, Siba PM, Mueller I. 2012. Ownership and usage of mosquito nets after four years of large-scale free distribution in Papua New Guinea. Malar. J. 11:192. Hetzel MW, Pulford J, Maraga S, Barnadas C, Reimer LJ, Tavul L, Jamea-Maiasa S, Tandrapah T, Maalsen A, Makita L and others. 2014b. Evaluation of the global fund-supported national malaria control program in Papua New Guinea, 2009-2014. PNG Med J. 57:1–4. Hetzel MW, Pulford J, Ura Y, Jamea-Maiasa S, Tandrapah A, Tarongka N, Lorry L, Robinson LJ, Lilley K, Makita L and others. 2017. Insecticide-treated nets and malaria prevalence, Papua New Guinea, 2008–2014. Bull. World Health Organ. 95:695–705. Hetzel MW, Reimer LJ, Gideon G, Koimbu G, Barnadas C, Makita L, Siba PM, Mueller I. 2016. Changes in malaria burden and transmission in sentinel sites after the roll-out of long- lasting insecticidal nets in Papua New Guinea. Parasit. Vectors 9:340. Hii JLK, Smith T, Vounatsou P, Alexander N, Mai A, Ibam E, Alpers MP. 2001. Area effects of bednet use in a malaria-endemic area in Papua New Guinea. Trans. R. Soc. Trop. Med. Hyg. 95:7–13. Kamau E, Alemayehu S, Feghali KC, Saunders D, Ockenhouse CF. 2013. Multiplex qPCR for detection and absolute quantification of malaria. PLoS ONE 8:8. Katusele M, Gideon G, Thomsen EK, Siba PM, Hetzel M, Reimer LJ. 2014. Longlasting insecticidal nets remain efficacious after five years of use in Papua New Guinea. PNG Med. J. 57:86–93. Keven JB, Henry-Halldin CN, Thomsen EK, Mueller I, Siba PM, Zimmerman PA, Reimer LJ. 2010. Pyrethroid susceptibility in natural populations of the Anopheles punctulatus group (Diptera: Culicidae) in Papua New Guinea. Am. J. Trop. Med. Hyg. 83(6):1259–1261. Koepfli C, Ome-Kaius M, Jally S, Malau E, Maripal S, Ginny J, Timinao L, Kattenberg JH, Obadia T, White M and others. 2017. Sustained malaria control over an 8-year period in Papua New Guinea: the challenge of low-density asymptomatic Plasmodium infections. JID 216:1434–1443. Koimbu G, Czeher C, Katusele M, Sakur M, Kilepak L, Tandrapah A, Hetzel MW, Pulford J, Robinson L, Karl S. 2018. Status of insecticide resistance in Papua New Guinea: an 96 update from nation-wide monitoring of Anopheles mosquitoes. Am. J. Trop. Med. Hyg. 98(1):162–165. Lee DJ, Hicks MM, Griffiths M, Debenham ML, Bryan JH, Russell RC, Geary M, Marks EN. 1987. The Culicidae of the Australasian region. Debenham ML, editor. Canberra: Australian Government Publishing Service. Mboera LEG. 2005. Sampling techniques for adult Afrotropical malaria vectors and their reliability in the estimation of entomological inoculation rate. Tanzan. Health Res. Bull. 7:117–124. Mehlotra RK, Kasenhagen LJ, Baisor M, Lorry K, Kazura JW, Bockarie MJ, Zimmerman PA. 2002. Malaria infections are randomly distributed in diverse holoendemic areas of Papua New Guiea. Am. J. Trop. Med. Hyg. 67(6):555–562. Mehlotra RK, Lorry K, Kastens W, Miller SM, Alpers MP, Bockarie M, Kazura JW, Zimmerman PA. 2000. Random distribution of mixed species malaria infections in Papua New Guinea. Am. J. Trop. Med. Hyg. 62(2):225–231. Moir JS, Garner PA, Heywood PF, Alpers MP. 1989. Mortality in a rural area of Madang Province, Papua New Guinea. Ann. Trop. Med. Parasitol. 83(3):305–319. Mueller I, Namuigi P, Kundi J, Ivivi R, Tandrapah T, Bjorge S, Reeder JC. 2005. Epidemic malaria in the highlands of Papua New Guinea. Am. J. Trop. Med. Hyg. 72(5):554–560. Muller I, Bockarie M, Alpers M, Smith T. 2003. The epidemiology of malaria in Papua New Guinea. Trends Parasitol. 19(6):253–259. Reimer LJ, Thomsen EK, Koimbu G, Keven JB, Mueller I, Siba PM, Kazura JW, Hetzel MW, Zimmerman PA. 2016. Malaria transmission dynamics surrounding the first nationwide long-lasting insecticidal net distribution in Papua New Guinea. Malar. J. 15:25. Smith DL, Dushoff J, Snow RW, Hay SI. 2005. The entomological inoculation rate and Plasmodium falciparum infection in African children. Nature 438:492–495. Sweeney AW. 1987. Larval salinity tolerances of the sibling species of Anopheles farauti. J. Am. Mosq. Control Assoc. 3:589–92. Thomsen EK, Koimbu G, Pulford J, Jamea-Maiasa S, Ura Y, Keven JB, Siba PM, Mueller I, Hetzel MW, Reimer LJ. 2017. Mosquito behaviour change after distribution of bednets results in decreased protection against malaria exposure. J. Infect. Dis. 215(5):791–797. WHO. 1975. Manual on practical entomology in malaria: part II methods and techniques. Geneva: World Health Organization 97 CHAPTER 4: SPATIAL AND TEMPORAL HETEROGENEITY OF HUMAN EXPOSURE TO ANOPHELES BITES IN THE NORTH COAST VILLAGES OF MADANG PROVINCE, PAPUA NEW GUINEA 4.1. Abstract Considerable gains in malaria reduction achieved by the ongoing nationwide insecticidal bed net campaign in PNG is threatened by recent resurgence of the disease in some parts of the country. This study examined the hypothesis that flexibility in mosquito host-seeking behavior enables them to access human hosts while simultaneously circumventing exposure to the insecticide in bed nets, mediated through patterns of vector-human contact in space and time. Host-seeking female Anopheles were sampled by the human landing catch method inside and outside several houses from dusk to dawn in four villages in Madang province. Blood samples were obtained from village residents and were tested for malaria infection by PCR. The results show that host- seeking behavior of Anopheles spp. (of four species total) was not homogeneous temporally or spatially. In all villages, 29–49.7% of the Anopheles catches occurred between 18:00 hr–22:00 hr) when 75–90% of people were awake and outside the bed nets, and 66.6–83% of catches occurred between 18:00 hr–2:00 hr before almost 100% of the residents retired to bed. Spatially, the frequency of outdoor catches was slightly higher than indoor catches. The number of human- biting Anopheles spp. among households was spatially heterogeneous, resulting in malaria basic reproduction rate values greater than the values under homogeneous condition by multiplicative factors ³ 1.6. Despite clustering of mosquito catches, human malaria infection cases by location of home were not clustered. These results indicate that in Madang and most likely in other parts of PNG, the majority of the mosquito vectors are able to access humans and transmit malaria 98 without contacting bed nets, thereby rendering bed nets less effective as a malaria control method. The high basic reproduction rate causes malaria to persist in these villages with the potential for rebound. 4.2. Introduction Insecticidal bed nets which target mosquito vectors are the mainstay of malaria control worldwide, particularly in endemic countries like PNG. The way bed nets work against malaria is that by sleeping under the nets, humans are barred from mosquito bites, thereby preventing or reducing the rate of transmission of the disease. By coming in contact with the nets, the mosquitoes are exposed to lethal dose of the insecticide, which results in (i) reduced population size of the vectors to levels insufficient to support transmission and (ii) reduction in the average age (i.e., longevity) of the vector populations below the duration required for the parasite to develop into the infectious stage (sporozoite) in the mosquitoes, thereby interrupting transmission of the disease (Darriet, et al. 1984; Lines, et al. 1987; Snow and Jawara 1987; Snow, et al. 1988). Despite these seemingly effective strategies offered by the bed nets, there are some ecological aspects of vector populations that have the potential to render the bed net method ineffective in preventing malaria transmission. One such factor is physiological resistance to the insecticide. This happens when the mosquitoes acquire genetic mutations in the insecticide’s target site which prevents the chemical compound from exerting its toxic effect, or when the metabolic detoxification of the insecticide is enhanced by duplication of genes that codes for detoxifying enzymes (Hemingway, et al. 2004; Hemingway and Ranson 2005). Although, physiological resistance has been reported for Anopheles populations in other parts of the world 99 (Quinones, et al. 2015), it has not been observed in the malaria vectors in PNG (Katusele, et al. 2014; Keven, et al. 2010; Koimbu, et al. 2018). As bed nets are deployed inside houses, vector populations that exhibit endophagic (tend to bite humans inside houses) behavior are easy to control as they’re likely to be exposed to the insecticides. Vector populations that exhibit exophagic (tend to bite humans outside houses) behavior are difficult to control as they are less likely to come in contact with the nets (Killeen 2014). Even if the vectors are endophagic, they must also seek human hosts late in the night and in the early morning hours when most people are asleep and protected by the bed nets in order for them to contact the nets; host-seeking in the evening when people are usually active and outside their bed nets reduces the vectors’ chance of exposure to the insecticides (Killeen 2014). Further, within a human community, human exposure to Anopheles bites can be spatially heterogeneous; that is, individuals in some locations of the village are more exposed to mosquito bites than those in other locations. This spatial heterogeneity can cause the basic reproduction rate (R0) of malaria—a measure of disease persistence—to be greater than the values under random or homogeneous exposure. Thus, human biting rates under heterogeneous condition can challenge the outcome of bed net intervention by enabling the disease to persist, with potential for rebound (Dye and Hasibeder 1986; Hasibeder and Dye 1988; Woolhouse, et al. 1997). In PNG, malaria prevalence and transmission intensity was reduced considerably to manageable levels nationally by ongoing nationwide bed net-based vector control program which commenced in 2004 (Hetzel, et al. 2015; Hetzel, et al. 2017; Hetzel, et al. 2016; Reimer, et al. 2016). However, research described in chapter 3 showed that in the north coast of Madang province, the infection prevalence and transmission intensity have both rebounded despite high bed net usage in the study villages. Prior research showed that insecticide resistance was nil in 100 the vector populations, eliminating that explanation for the resurgence (Katusele, et al. 2014; Keven, et al. 2010; Koimbu, et al. 2018). The aim of the research in this study was to investigate the temporal and spatial patterns of human exposure to Anopheles bites and how these may enable the vectors to evade bed nets and cause malaria transmission to rebound. 4.3. Methods 4.3.1. Study sites, mosquito sampling and active case malaria detection This study was conducted in the same four villages (Bulal, Megiar, Mirap and Wasab) described in chapter 3. The geographic locations and ecogeographic description of these villages; the mosquito sampling design and procedure; mosquito species identification; and human malaria infection survey were as described in chapter 3, section 3.3. 4.3.2. Data analyses Temporal and spatial heterogeneity of human exposure to Anopheles bites. The mean number of female Anopheles collected nightly by human landing catch at each of six bihourly periods of the night were calculated. Generalized linear model (GLM) with negative binomial distribution was used to compare the number of mosquitoes collected nightly at each bihourly periods. In the GLM equation ln(.)=01+03(4567)+ln('), the expected mosquito number µ was modelled as a linear function of the categorical predictor variable Time, with six levels representing the six bihourly periods. Number of house-night S was included as offset term. The frequency distribution of mosquito numbers captured in different houses was tested for fit to the Poisson and negative binomial distribution models based on chi-square test of fit of expected and observed frequency distributions. Statistically supported fit to the Poisson model indicates 101 random distribution of mosquitoes among the households, whereas fit to the negative binomial model indicates heterogeneous or clustered distribution. For villages whose mosquito frequency using the formula #8=∑ :;<=; >?@3 fitted negative binomial, the effect of spatial heterogeneity on the R0 of malaria was quantified , where vi is the proportion of vectors in house i (where i = 1 to m and m is the number of households), and hi is the proportion of humans in the house. This formula, derived by (Dye and Hasibeder 1986) and applied to empirical data by (Woolhouse, et al. 1997), estimates the multiplicative factor Rf. Whatever the true R0 value may be under spatially random condition, when vectors exhibit spatial heterogeneity the R0 increases by a multiplicative factor Rf, which is a value ³ 1.0. Statistical difference in the number of Anopheles captured inside and outside houses in each village was compared using the Mann-Whitney U test, with house-night as the unit of replication. Spatial distribution of malaria case households. The K-function, a spatial point pattern statistical analysis method, was used to test whether households with one or more human cases of malaria infections were clustered or randomly distributed in space. The theoretical basis of this statistical method is discussed elsewhere (Bivand, et al. 2013; Diggle and Chetwynd 1991; Diggle, et al. 2007). Generally, the K-function quantifies spatial intensity (number of spatial points located in a circle of radius r) of so-called point process data. In spatial epidemiology, disease cases are spatially clustered if the difference D of the K estimates of case and control households (i.e., D = Kcases - Kcontrols) is statistically greater than 0. In this study, the D statistic from K estimates of malaria infected (case) and uninfected (control) households was quantified to test for clustering. Computation of the statistical tests. All statistical tests were performed in R software. The Mann-Whitney U test was performed using the function wilcox.test of the package stats. The 102 model fit test was computed using the function fitdist of the package fitdistrplus. The GLM analysis was performed using the functions glm.nb of the package MASS. Computation for the test of spatial clustering of malaria case households was accomplished via Monte Carlo statistical procedure using algorithms (written in R codes) described on page 205–206 in (Bivand, et al. 2013). Significance level for all tests was based on type I error rate of 5%. 4.4. Results 4.4.1. Anopheles vectors A total of 3,013 female Anopheles of four different species, namely An. farauti (s.s.), An. koliensis, An. longirostris and An. punctulatus (s.s.) were collected over the sampling period. The distribution of these vector species and for Anopheles spp. (i.e., regardless of species) in each village is shown in Table 4.1. As expected, based on the findings in chapter 2, most (96%) of An. farauti (s.s.) were from the two coastal villages—Megiar and Mirap—and all (100%) of An. punctulatus (s.s.) were from the two inland villages—Bulal and Wasab (see map in Figure 2.1, chapter 2 for geographic location of the villages). Like An. farauti (s.s.), An. koliensis was present in all villages but unlike An. farauti (s.s.) it did not exhibit a bias for either coastal or inland sites. Like An. punctulatus (s.s.), An. longirostris was found in only two villages, but unlike An. punctulatus (s.s.), it did not exhibit a coastal-inland bias. 103 Table 4.1. Number of Anopheles mosquitoes according to species collected in four villages. Vector species An. farauti (s.s.) An. koliensis An. longirostris An. punctulatus (s.s.) Total Anopheles spp. Bulal Megiar Mirap Wasab 6 70 0 10 86 453 20 0 0 473 477 970 46 0 1,493 31 737 13 180 961 4.4.2. Temporal heterogeneity of exposure to Anopheles bites The mean number of mosquitoes per person-night for a specific vector species or Anopheles spp. (indoor and outdoor mosquito data were combined for this analysis) was plotted for each of the six bihourly periods of the night for each village (Figure 4.1). In the corresponding GLM plot (Figure 4.2), the number of Anopheles spp. per person-night in each bihourly period of the night was statistically compared with the second period (20:00–22:00 hr) as the reference, because the mean mosquito number was the highest in this period in all villages, except Wasab (Figure 4.1). In Bulal, the number of Anopheles spp. collected in two morning periods 0:00–2:00hr and 4:00– 6:00 was significantly lower than the reference period; the other periods were not different from the reference period (Figure 4.2). In Megiar, the number of Anopheles spp. in only the last two periods (2:00–4:00 hr and 4:00–6:00 hr) was significantly lower than the reference period (Figure 4.2). The number of Anopheles spp. in the first period in Mirap and the last period in Wasab were significantly lower than the reference period (Figure 4.2). Generally, the same trend for Anopheles spp. was observed for specific Anopheles species in all village (Figure 4.1 and 4.2). 104 The data were also evaluated in terms of the proportion of human-biting Anopheles spp. collected at each of three four-hourly nocturnal periods: evening (18:00–22:00 hr), late-night (22:00–2:00 hr) and morning (2:00–6:00 hr). In Bulal nearly half (46.5%) of the Anopheles sought humans in the evening; the other half sought humans almost equally between the late- night (28.0%) and morning (25.5%) periods (Figure 4.3). In Megiar, nearly half (49.7%) of the Anopheles sought humans in the evening whereas the other half sought humans unequally between the late-night (33.6%) and morning (16.7%) (Figure 4.3). Unlike Bulal and Megiar, the proportion of human-seeking Anopheles was generally the same in all three periods in Mirap and Wasab (Figure 4.3). 105 Figure 4.1. Mean number (colored dots) of mosquitoes collected per person-night at each of six bihourly periods of the night (expressed in 24 hr format) in four villages. Color purple red and green each represents a specific vector species whereas blue represents Anopheles spp. in general. Vertical bars are the standard error of the means. 106 Figure 4.2. Plot of GLM coefficients (open circle) with 95% confidence interval bars of five variables (y-axis) of Anopheles in four villages. The variables include five of the six bihourly periods of the night (expressed in 24 hr format), each compared with the second period 20:00- 22:00 hr (not plotted)—the reference variable (represented by the broken vertical line centered at zero). The colors purple, red, and green each represent a specific vector species whereas blue represents Anopheles spp. in general. The estimated coefficients (x-axis) are in logarithmic scale. 107 Figure 4.3. Proportion of human-seeking Anopheles spp. captured at three nocturnal periods in four villages. Mosquito sample size in each village are: Bulal (n = 86), Megiar (n = 473), Mirap (n = 1493) and Wasab (n = 961). 4.4.3. Spatial heterogeneity of exposure to Anopheles bites The mean number of Anopheles spp. per house-night collected outdoor tended to be higher than indoor collections for all villages except Bulal, which was opposite, but numbers were low (Figure 4.4). However, these variations were not significantly different for all four pairs (Mann- Whitney U tests). The total number of Anopheles spp. collected in each household throughout the sampling period is displayed as bubble plots (Figure 4.5) for all villages except Bulal which was excluded due to low sample size (n = 86). For both indoor and outdoor collections in all three villages, the number of Anopheles varied among the households. The observed heterogeneity among the households was supported by fit of the frequency distribution of household Anopheles numbers, both indoor and outdoor collections, to the negative binomial and not Poisson model (Table 4.2). Although the Anopheles numbers varied among the households in Mirap, there was 108 no clear visual spatial trend associated with this heterogeneity as most of the houses with low Anopheles numbers were found in close spatial proximity to those with larger numbers (Figure 4.5). In contrast, the other two villages exhibited observable spatial trend. In Megiar, the number of mosquitoes was lower in the houses on the south-east end of the village but higher in houses on the north-west end (Figure 4.5). In Wasab outdoor collection households, the Anopheles numbers increased from north-east to south-west; indoor collection households did not show any clear spatial trend (Figure 4.5). Assessment of the effect of spatial heterogeneity on transmission potential show that the estimated values of R0 multiplicative factor Rf were > 1.0 for both indoor and outdoor collections in all three villages (Table 4.3). Figure 4.4. Boxplots showing the distribution of Anopheles spp. numbers captured nightly inside and outside of houses in four villages. Blue shaded squares are the means, black horizontal lines are the medians, and black dots are outliers. 109 Figure 4.5. Map showing the location of houses where mosquitoes were sampled in Megiar, Mirap and Wasab villages. Blue shaded circles are sampled houses and grey points are unsampled houses. Data was plotted for both indoor (top row) and outdoor (bottom row) collections. Size of the blue circles correspond to the number of mosquitoes collected in the house 110 Table 4.2. Chi-square test of fit of negative binomial and Poisson models to frequency distribution of Anopheles numbers collected inside and outside houses in three villages. Village Location N Mean Var Model Megiar Mirap Wasab Outdoor 12 29.5 1353 Indoor 12 10.2 290 Outdoor 12 72.1 1440 Indoor 12 52.3 1458 Outdoor 12 48.0 1802 Indoor 12 25.4 279 Neg. Binomial c2 3.36 df 1 P 0.07 Poisson 2.5 x 105 Neg. Binomial 1.51 Poisson 1.9 x 104 Neg. Binomial 1.62 Poisson 1.6 x 104 Neg. Binomial 1.69 Poisson 3.5 x 1010 Neg. Binomial 3.75 Poisson 3.3 x 108 Neg. Binomial 0.14 Poisson 1.4 x 1020 2 1 2 1 2 1 2 1 2 1 2 < 0.0001 0.22 < 0.0001 0.20 < 0.0001 0.19 < 0.0001 0.053 < 0.0001 0.14 < 0.0001 Abbreviations: N, number of sampled households; Var, variance; df, degrees of freedom; P, p- value associated with the chi-square test. Table 4.3. Estimated Rf for indoor and outdoor mosquito collections in three villages. Village Megiar Mirap Wasab Rf 3.89 4.59 1.96 1.60 2.77 1.68 House location Indoor Outdoor Indoor Outdoor Indoor Outdoor 111 4.4.4. Spatial distribution of malaria case households Maps showing the geolocation of malaria infection case and non-case (negative) households in each of the four villages are presented in Figure 4.6–4.9. The results of the K-function test of spatial clustering of case households are also presented. None of the villages had D estimates significantly different from 0; the D estimates were within the 95% confidence interval limit (grey shaded area in panels b and d of Figure 4.6–4.9) at all distance r. These results indicate that the case households were randomly distributed in space like the non-case households rather than clustered. 112 Figure 4.6. Geolocation of malaria case (colored squares) and non-case (black circles) households for P. falciparum (panel a) and P. vivax (panel c) in Bulal, along with their corresponding D statistic plot (panel b and d). In the D statistic plots, the broken red line is the null condition (D = 0 for all distance r in meters) which indicates random distribution of the case households, the black line indicates the observed D statistic (y-axis) at distance r in meters (x- axis), and the grey area is the 95% confidence interval of the observed D statistic. 113 Figure 4.7. Geolocation of malaria case (colored squares) and non-case (black circles) households for P. falciparum (panel a) and P. vivax (panel c) in Megiar, along with their corresponding D statistic plot (panel b and d). In the D statistic plots, the broken red line is the null condition (D = 0 for all distance r in meters) which indicates random distribution of the case households, the black line indicates the observed D statistic (y-axis) at distance r in meters (x- axis), and the grey area is the 95% confidence interval of the observed D statistic. 114 Figure 4.8. Geolocation of malaria case (colored squares) and non-case (black circles) households for P. falciparum (panel a) and P. vivax (panel c) in Mirap, along with their corresponding D statistic plot (panel b and d). In the D statistic plots, the broken red line is the null condition (D = 0 for all distance r in meters) which indicates random distribution of the case households, the black line indicates the observed D statistic (y-axis) at distance r in meters (x- axis), and the grey area is the 95% confidence interval of the observed D statistic. 115 Figure 4.9. Geolocation of malaria case (colored squares) and non-case (black circles) households for P. falciparum (panel a) and P. vivax (panel c) in Wasab, along with their corresponding D statistic plot (panel b and d). In the D statistic plots, the broken red line is the null condition (D = 0 for all distance r in meters) which indicates random distribution of the case households, the black line indicates the observed D statistic (y-axis) at distance r in meters (x- axis), and the grey area is the 95% confidence interval of the observed D statistic. 116 4.5. Discussion Two distinct nocturnal bihourly biting patterns exhibited by the Anopheles spp. were observed. The first was in Bulal and Megiar where the mosquitoes exhibited early biting activity which peaked in the second period (20:00–22:00 hr), followed by dramatic decline in the biting activity in the later periods, although in Bulal a secondary peak of activity occurred in the fifth period (2:00–4:00 hr). The second pattern was seen in Mirap and Wasab where the biting activity also peaked in the second period however, unlike the previous two villages, the biting rate did not decline but held steady for the subsequent periods of the night before dipping in the last period. As speculated from the barrier screen data (chapter 2), these nocturnal patterns were plausibly linked to the commuting time of the mosquitoes between their resting or larval habitats and the village, wherein mosquito populations whose habitats were in close proximity to the village arrived earlier in the night to seek hosts as exhibited by the Anopheles in Bulal and Megiar. Mosquito populations whose habitats were further from the village arrived late in the night and in the morning periods; the secondary peak in Bulal represent this condition. Mosquito populations whose habitats were randomly distributed in space, ranging from closest to furthest from the village, exhibited the nocturnal activity pattern displayed by the Mirap and Wasab populations. The tendency of the mosquitoes to land on and attempt to bite humans during the period of the night when most people are active is an important factor that can render the bed nets ineffective as a vector control method (Killeen 2014; Killeen and Chitnis 2014). A survey by questionnaire of individuals in the four study villages showed that 75–90% of the people retired to bed after 22:00 hr and nearly 100% of people retired to bed after 2:00 hr (Rodriguez- Rodriguez et. al, unpublished; personal communication). In Bulal and Megiar 46.5–49.7% of the 117 Anopheles bites occurred in the evening when 75–90% of people were exposed, and 75–83.3% of the bites occurred before 2:00 hr before 100% of the people retired to bed. In Mirap and Wasab 29–34% of the bites occurred in the evening (75–90% of people exposed), and 66.6–72% of the bites occurred before 2:00 hr before 100% of the people retired to bed. The tendency of the mosquitoes to bite humans more frequently outdoors than indoors is another important factor that limits the efficacy of the bed nets. By biting outdoor, mosquitoes can evade not just the lethal contact with the indoor-deployed insecticidal bed nets but the protective barrier to their human hosts (Killeen 2014). In this study, although there was no statistical difference in the number of Anopheles collected inside versus outside houses in all the study villages, more mosquitoes were collected outdoors than indoors and conservatively it can be concluded that frequency of landing (and presumably therefore, biting) was equivalent between the two locations. The malaria vectors of PNG simply have no preference for either indoor or outdoor biting and may not even distinguish the two locations. Importantly, however, the biting (as reflected by landing catches here) is clearly peridomestic. Further, although a significant proportion of the vectors landed on the human collectors inside houses, most of them may not contact bed nets in the case of human residents of those houses, because a majority of the vector-human contact occurred during the period of the night when the human hosts were active and not under the bed nets. Assessment of spatial heterogeneity in the number of human-biting mosquitoes is important for two reasons: (i) to assess the transmission potential of malaria which is useful to guide disease intervention plans and (ii) to identify hotspots of transmission so that disease intervention can be targeted to the risk groups. One way to assess the transmission potential is through estimation of the R0, which is the number of secondary malaria cases in fully susceptible 118 individuals that ensue from one primary case during its entire infectious life span. The magnitude of the difference in the true R0 values under heterogeneous relative to homogeneous condition can be quantified in terms of Rf directly from the same data that were used to test spatial heterogeneity (Dye and Hasibeder 1986; Woolhouse, et al. 1997), as was applied in this study (see description of data analysis in Methods, this chapter). For malaria to persist (endemic), rebound, or spike (epidemic) the value of R0 theoretically must be > 1.0; values £ 1.0 will eventually lead to malaria decline (Bailey 1982). Under heterogeneous condition, malaria persistence is strengthened as the R0 increases by a multiplicative factor Rf. In the three villages that were investigated, the estimated Rf values ranged from 1.96–3.89 for indoor collections and 1.6–4.59 for outdoor collections (Table 4.3). The Rf values estimated in this study were similar to the value of 3.7 reported for Anopheles spp. in a village in East Sepik province of PNG (Woolhouse, et al. 1997). Thus, spatial heterogeneity is a strong contributing factor to resumption of high malaria transmission rates and rebound of human prevalence of infection described in chapter 3. The northwest end of Megiar and the southwest side of Wasab village where the vector-human contact rates were highest (Figure 4.5) can be considered hotspots of transmission. People in these areas are most likely to be infected (carriers) and serve as sources which amplify transmission and promote spread of infection. Thus, intervention programs such as distribution of free insecticidal bed nets and administration of antimalarial prophylactic drugs or transmission-blocking vaccines (Carter 2001), particularly in resource-limited countries like PNG, can be targeted to individuals living in those hotspots. This analysis begs the question of which is more important to inform malaria control: the locations of homes where people have malaria, which is randomly distributed, or the location where Anopheles bites on humans (as reflected by landing catches) aggregate spatially. The former requires a high target of universal 119 coverage, whereas the latter does not but the spatial structure of mosquito-human contact is not readily discernable at this stage of research. Given the spatial heterogeneity in the number of human-biting mosquitoes, particularly in Megiar and Wasab where there was clear visual directional trend, it is logical to expect the spatial distribution of households with malaria infection cases to be non-random or clustered with respect to the non-case households. The case households should cluster in the areas of the village where the numbers of human-biting mosquitoes were high (i.e., hotspots of transmission), such as the northwest end of Megiar and southwest side of Wasab. Instead, the K-function tests show that all the four study villages exhibited random rather than clustered distribution (Figure 4.6–4.9). There are two plausible reasons why there was spatial heterogeneity in the number of human-biting mosquitoes but not human infection cases. First, there must be a threshold value of biting rates below which malaria transmission is interrupted that would lead to spatial heterogeneity of infection cases. The number of human-biting mosquitoes observed in this study might be higher than the threshold level, thus allowing infection cases to be spatially homogeneous. This assertion is supported indirectly by a study which showed that an extremely low infectious biting rate is required to achieve malaria infection prevalence below 5% (Beier, et al. 1999). Second, the results discussed here were based on mosquito data from a single time point; not from longitudinal sampling. Thus, the stationarity of the spatial heterogeneity is not known but may vary temporally so that areas with a low number of human-biting mosquitoes at one time point have higher numbers at a different time point. The temporal variation of spatial heterogeneity eventually results in spatially homogeneous level of exposure to Anopheles bites and hence spatially homogeneous infection cases. 120 4.6. Conclusion The temporal and spatial patterns of vector-human contact reported in this study indicate that in Madang, and most likely other parts of PNG, bed nets may not be fully effective in preventing malaria transmission. Temporally, a large proportion of the human-biting activity of the vectors occurred in the evening, before midnight. As most people are active during this period, malaria transmission takes place before the people go to sleep under their bed nets. Spatially, a large proportion of host-seeking Anopheles bit humans outside houses where the risk of contacting the indoor-deployed insecticidal bed nets was negligible. For those that bit humans inside houses, most did so in the evening when people were not sleeping under their nets. These factors limit bed nets from fully reducing malaria transmission through suppression of vector abundance and reduction of vector longevity. The number of human-biting mosquitoes varied among households within the villages (i.e., spatial heterogeneity), resulting in Rf values that were ³ 1.6, which indicates that the disease has the potential to persist. Finally, despite the spatial heterogeneity in exposure to vector bites, this did not lead to spatial clustering of human infection cases. This inconsistency can be explained by two hypotheses: (i) the current low number of human-biting mosquitoes was insufficient to interrupt transmission and/or (ii) the spatial heterogeneity is temporally unstable and changes over time, culminating in spatially homogeneous exposure to infectious bites. These ecological factors are potential determinants underlying malaria resurgence as discussed in chapter 3. 121 REFERENCES 122 REFERENCES Springer. Bailey NTJ. 1982. The biomathematics of malaria. London: Charles Griffin & Company Ltd. Beier JC, Killeen GF, Githure JI. 1999. Entomologic inoculation rates and Plasmodium falciparum malaria prevalence in Africa. Am. J. Trop. Med. Hyg. 61(1):109–113. Bivand RS, Pebesma E, Gomez-Rubio V. 2013. Applied spatial data analysis with R. New York: Carter R. 2001. Transmission blocking malaria vaccines. Vaccine 19:2309–2314. Darriet F, Robert V, Vien NT, Carnevale P. 1984. Evaluation of the efficacy of permethrin- impregnated intact and perforated mosquito nest against vectors of malaria. World Health Organization Publication no. WHO/VBC/84.899. Diggle PJ, Chetwynd AG. 1991. Second-order analysis of spatial clustering for inhomogeneous populations. Biometrics 47:1155–1163. Diggle PJ, Gomez-Rubio V, Brown PE, Chetwynd AG, Gooding S. 2007. Second-order analysis of inhomogeneous spatial point processes using case–control data. Biometrics 63:550– 557. Dye C, Hasibeder G. 1986. Population dynamics of mosquito-borne disease: effects of flies which bite some people more frequently than others. Trans. R. Soc. Trop. Med. Hyg. 80:69–77. Hasibeder G, Dye C. 1988. Population dynamics of mosquito-borne disease: persistence in a completely heterogeneous environment. Theor. Popul. Biol. 33:31–53. Hemingway J, Hawkes NJ, McCarroll L, Ranson H. 2004. The molecular basis of insecticide resistance in mosquitoes. Insect Biochem. Mol. Biol. 34:653–665. Hemingway J, Ranson H. 2005. Chemical control of vectors and mechanism of resistance. In: Marquardt WC, editor. Biology of disease vectors. 2 ed. New York: Elsevier Academic Press. p 627–637. Hetzel MW, Morris H, Tarongka N, Barnadas Ce, Pulford J, Makita L, Siba PM, Mueller I. 2015. Prevalence of malaria across Papua New Guinea after initial roll-out of insecticide- treated mosquito nets. Trop. Med. Int. Health 20(12):1745–1755. Hetzel MW, Pulford J, Ura Y, Jamea-Maiasa S, Tandrapah A, Tarongka N, Lorry L, Robinson LJ, Lilley K, Makita L and others. 2017. Insecticide-treated nets and malaria prevalence, Papua New Guinea, 2008–2014. Bull. World Health Organ. 95:695–705. 123 Hetzel MW, Reimer LJ, Gideon G, Koimbu G, Barnadas C, Makita L, Siba PM, Mueller I. 2016. Changes in malaria burden and transmission in sentinel sites after the roll-out of long- lasting insecticidal nets in Papua New Guinea. Parasit. Vectors 9:340. Katusele M, Gideon G, Thomsen EK, Siba PM, Hetzel M, Reimer LJ. 2014. Longlasting insecticidal nets remain efficacious after five years of use in Papua New Guinea. PNG Med. J. 57:86–93. Keven JB, Henry-Halldin CN, Thomsen EK, Mueller I, Siba PM, Zimmerman PA, Reimer LJ. 2010. Pyrethroid susceptibility in natural populations of the Anopheles punctulatus group (Diptera: Culicidae) in Papua New Guinea. Am. J. Trop. Med. Hyg. 83(6):1259–1261. Killeen GF. 2014. Characterizing, controlling and eliminating residual malaria transmission. Malar. J. 13:330. Killeen GF, Chitnis N. 2014. Potential causes and consequences of behavioural resilience and resistance in malaria vector populations: a mathematical modelling analysis. Malar. J. 13: Koimbu G, Czeher C, Katusele M, Sakur M, Kilepak L, Tandrapah A, Hetzel MW, Pulford J, Robinson L, Karl S. 2018. Status of insecticide resistance in Papua New Guinea: an update from nation-wide monitoring of Anopheles mosquitoes. Am. J. Trop. Med. Hyg. 98(1):162–165. Lines JD, Myamba J, Curtis CF. 1987. Experimental hut trials of permethrin-impregnated mosquito nets and eave curtains against malaria vectors in Tanzania. Med. Vet. Entomol. 1:37–51. Quinones ML, Norris DE, Conn JE, Moreno M, Burkot TR, Bugoro H, Keven JB, Cooper R, Yan G, Rosas A and others. 2015. Insecticide resistance in areas under investigation by the International Centers of Excellence for Malaria Research: A challenge for malaria control and elimination. Am. J. Trop. Med. Hyg. 93(3 Suppl):69–78. Reimer LJ, Thomsen EK, Koimbu G, Keven JB, Mueller I, Siba PM, Kazura JW, Hetzel MW, Zimmerman PA. 2016. Malaria transmission dynamics surrounding the first nationwide long-lasting insecticidal net distribution in Papua New Guinea. Malar. J. 15:25. Snow RW, Jawara M. 1987. Observations on Anopheles gambiae Giles s.l. (Diptera: Culicidae) during a trial of permethrin-treated bed nets in The Gambia. Bull. ent. Res. 77:279–286. Snow RW, Lindsay SW, Hayes RJ, Greenwood BM. 1988. Permethrin-treated bed nets (mosquito nets) prevent malaria in Gambian children. Trans. R. Soc. Trop. Med. Hyg 82:838–842. Woolhouse MEJ, Dye C, Etard JF, Smith T, Charlwood JD, Garnett GP, Hagan P, Hii JLK, Ndhlovu PD, Quinnell RJ and others. 1997. Heterogeneities in the transmission of 124 infectious agents: implications for the design of control programs. Proc. Natl. Acad. Sci. USA 94:338-342. 125 CHAPTER 5: PROBE-BASED MULTIPLEX QUANTITATIVE PCR IDENTIFIES BLOOD- MEAL HOSTS IN ANOPHELES MOSQUITOES FROM PAPUA NEW GUINEA Publication Keven JB, Artzberger G, Gillies ML, Mbewe RB, Walker ED. 2020. Probe‑based multiplex qPCR identifies blood‑meal hosts in Anopheles mosquitoes from Papua New Guinea. Parasit. Vectors 13:111 5.1. Abstract Determination of blood-meal hosts in blood-fed female Anopheles mosquitoes is important for evaluating vectorial capacity of vector populations and assessing effectiveness of vector control measures. Sensitive molecular methods are needed to detect traces of host blood in mosquito samples, to differentiate hosts, and to detect mixed host blood-meals. This paper describes a molecular probe-based qPCR for identifying blood-meal hosts in Anopheles malaria vectors from PNG. TaqMan oligonucleotide probes targeting specific regions of mitochondrial or nuclear DNA of the three primary Anopheles hosts—humans, pigs and dogs—were incorporated into a multiplex, qPCR which was optimized for sensitivity and specificity. Amplification of serially diluted DNA showed that the qPCR detected as low as 10-5 ng/µl of host DNA. Application to field-collected, blood-fed Anopheles showed that the qPCR identified the vertebrate hosts for 89% (335/375) of mosquitoes whereas only 55% (104/188) of blood-meal samples tested in a conventional PCR were identified. Of the 104 blood-fed Anopheles that were positive in both PCR methods, 16 (15.4%) were identified as mixed blood meals by the qPCR whereas only three (2.9%) were mixed blood meals by the conventional PCR. The multiplex qPCR described here is 126 sensitive at detecting low DNA concentration and mixed host DNA in samples and useful for blood meal analysis of field mosquitoes, in particular mixed-host blood meals. 5.2. Introduction Identification of blood-meal hosts in blood-sucking arthropod vectors of human diseases forms an integral component of analysis of both vectorial capacity and effectiveness of vector control measures. For human malaria, analysis of female Anopheles mosquito blood meals is important for assessing the degree of feeding on humans (anthropophagy), usually expressed in terms of human blood index: the proportion of blood-engorged mosquitoes that fed on humans relative to non-human hosts. This index is used to evaluate epidemiological outcomes of vector-targeted malaria control programs since vectorial capacity, which is a measure of disease transmission potential of a vector population, varies with the human blood index (Garrett-Jones 1964a; Garrett-Jones 1964b). Blood meal analysis is also important for assessing host selection, which is the propensity of a mosquito population to feed more or less on a host species relative to other host species available to the mosquito (Boreham and Garrett-Jones 1973). Assessment of host selection is important when considering vector control methods such as insecticide-treated bed nets and zooprophylaxis which seek to divert mosquito feeding to non-human hosts. Immunological methods such as the precipitin test, latex agglutination, and enzyme- linked immunosorbent assays were the earliest methods used in mosquito blood meal analysis (Beier, et al. 1988; Boorman and Mellor 1977; Burkot, et al. 1981; Weitz 1956). These methods have produced countless valuable data (Boreham and Garrett-Jones 1973; Boreham, et al. 1979; Burkot, et al. 1989; Burkot, et al. 1988; Edman and Downe 1964; Senior-White 1952; Shemanchuk, et al. 1963) but have been superseded in recent years by nucleic acid-based 127 methods. In cases where mosquito populations are known to feed on a wide range of vertebrate hosts or when the host range has not been characterized, sequencing of amplicons amplified with universal primers followed by a BLAST (Basic Local Alignment Search Tool) search against publicly available vertebrate host sequences has been used (Hamer, et al. 2011; Hamer, et al. 2009; Logue, et al. 2016; Molaei, et al. 2006; Molaei, et al. 2010). However, this approach is expensive and sequencing facilities are not available in many laboratories. In cases where mosquito populations are known to feed on a limited range of host species, less expensive methods such as restriction-fragment length polymorphisms (Meece, et al. 2005), DNA-DNA hybridization (Fritz, et al. 2013), conventional multiplex PCR with species-specific primers (Kent and Norris 2005) or qPCR with species-specific probes (Gunathilaka, et al. 2016; Hurk, et al. 2007) can be employed. In PNG, the blood-meal host range for the Anopheles vectors of malaria is generally limited to humans, pigs and dogs (Burkot, et al. 1989; Burkot, et al. 1988; Keven, et al. 2017; Logue, et al. 2016). Thus, PCR methods with host-specific primers or probes are appropriate in this setting. Conventional multiplex PCR with primers specific to these three hosts is currently applied to mosquitoes originating from PNG (Keven, et al. 2017). However, with a large sample, the gel-electrophoresis step can be labor-intensive and time-consuming, resulting in accumulation of unanalyzed samples and delays in data availability. Furthermore, as shown in this present study, the conventional multiplex PCR has high rates of false-negative blood meals and is less sensitive at detecting mixed blood meals (see result section). The objective of this study was to develop a probe-based multiplex qPCR for identifying the common mammalian blood-meal hosts in female Anopheles mosquitoes originating from PNG. 128 5.3. Methods 5.3.1. Samples Genomic DNA isolated directly from human and animal tissues were used in the optimization of the qPCR. Pig DNA was isolated from a piglet’s tail (25 mg), removed as part of standard tail- docking routine at Michigan State University Swine Farm. Dog DNA was isolated from a blood sample (50 µl) by a laboratory that studies genetic diseases of dogs. Human DNA was isolated from a blood sample (50 µl) of a study participant in PNG during a cross-sectional malaria epidemiology survey by PNG Institute of Medical Research. Genomic DNAs isolated from abdomens of blood-fed Anopheles mosquitoes collected in a malaria-endemic village of PNG were used to test the new qPCR method. DNA extractions were performed using Qiagens DNeasy Blood & Tissue Kit. 5.3.2. Probe design Nucleotide sequences for the primers and probes described in this study are shown in Table 5.1. The probes were designed within the mitochondrial cytochrome c oxidase subunit 1 gene (cox1) for pig, mitochondrial cytochrome b gene (cytb) for dog, and intron 1 of tyrosine hydroxylase gene (th) for human using the online program PrimerQuest accessible at the Integrated DNA Technologies website. Each probe was labelled at the 5’end with reporter dye FAM, VIC or ABY and at the 3’ end with TaqMan quencher dye QSY (see Table 5.1). The primers and probes were obtained from Invitrogen and Applied Biosystems respectively through Thermo Fisher Scientific. 129 Table 5.1. Primers and probes for the three vertebrate host species. Organism Primer/probe Nucleotide sequence 5’-3’ Human Pig Dog name humanth_F humanth_R humanth_P pigcox1_F pigcox1_R pigcox1_P dogcytb_F dogcytb_R dogcytb_P GGCCTGTTCCTCCCTTATTT TACACAGGGCTTCCGAGT FAM- ATGGAGTCTGTGTTCCCTGTGACC -QSY CTGACTACTTCCACCATCCTTC TGGGCTAAGTTTCCAGCTAAA VIC- ATAGTAGAAGCCGGAGCGGGTACT -QSY TGGACAAAGCAACCCTAACA CCGGTTTCGTGTAGAAATAGGA ABY- TCATCCTCCCTTTCATCATCGCAGC -QSY Amplicon size (in bp) 101 a 108 103 a Due to presence of a microsatellite repeat sequence (5’-AATG-3’) within the human amplicon, the PCR product size varies between different human individuals. Presence of the repeat sequence does not affect PCR amplification as the primers and probe sequence do not extend into the repeat region. 5.3.3. Development of multiplex qPCR For each target organism, the optimum qPCR reaction mixture was determined by testing different concentrations of the specific probe and primers in uniplex reactions (10 µl volume) containing 1x TaqMan Multiplex Master Mix (Catalog number: 4461882; Thermo Fisher Scientific), and 2 µl of serially diluted (10-fold dilutions) template DNA samples (10 ng/µl starting concentration). PCR reactions were performed on a QuantStudio 7 Flex instrument (Applied Biosystems) with Fast Cycling condition consisting of one cycle of 95 °C for 20 seconds followed by 40 seconds of 95 °C for one second and 60 °C for 20 seconds. Results were evaluated in software QuantStudio (version 1.3). Standard curves were evaluated and the reagent concentration that gave the highest percent amplification efficiency (measured as Efficiency = (10 (-1/slope) - 1) * 100)) and lowest threshold cycle or Cq (which corresponds to highest 130 sensitivity) was considered optimum and was used as the starting concentration for optimization of multiplex qPCR. For each of the three assays, multiplex qPCR mixture that gave the highest sensitivity and percent efficiency was considered optimum. The standard curve for each assay was derived from amplification results of triplicate dilution series (10-fold) of the target DNA. 5.3.4. Assay validation Specificity of the probes and primers to their target DNA region was confirmed in silico by performing BLAST and primer-BLAST (Ye, et al. 2012) searches for potential matches to untargeted sequences in the National Center for Biotechnology Information database. Cross- reactivity of each probe to non-target organisms was also tested in vitro by subjecting DNA of the non-target organisms to uniplex qPCR containing the probe. For example, the human probe was tested for cross-reactivity to eight vertebrate species, namely pig, dog, cow, goat, horse, cat, rat and chicken; three Plasmodium species likely to be present in the blood meals, namely P. falciparum, P. vivax and P. malariae, and five species of Anopheles mosquitoes commonly found in PNG, namely An. farauti (s.s.), An. koliensis, An. punctulatus (s.s.), An. longirostris and An. bancroftii. Ten-fold serially diluted DNA samples of humans, pigs, and dogs were amplified in the multiplex qPCR and the conventional, multiplex PCR (Kent and Norris 2005) and their results were compared for sensitivity. Primers for the single gene target of the mitochondrial cytb gene, reaction mixture, and cycling condition for the conventional PCR used in this study was as described in (Keven, et al. 2017). The Anopheles blood-meal DNA samples (n = 375) were analyzed in the qPCR to validate the applicability of the method to field mosquitoes. A subset of these qPCR-tested blood-meal samples (n = 188) was also analyzed in the conventional PCR and 131 its amplification success rate was compared with that of the qPCR. Amplification results with Cq value ³ 38 were inconclusive and therefore considered negative. For mosquitoes whose blood-meal hosts were successfully identified by the qPCR, the host DNA concentration was estimated using the formula log10 (Conc) = (Cq – Intercept)/Slope where Conc is the DNA concentration in ng/µl. The slope and intercept are specific to each host species (see Table 5.2). For the single blood-meal mosquitoes, the mean host DNA concentration for those that were detected in the conventional PCR was statistically compared (two-sample t- test) with those that were undetected. Similarly, for the mixed blood meals, the mean DNA concentration for the hosts that was detected in the conventional PCR was compared with those that were undetected. 5.4. Results The three blood-meal probes performed well together in multiplex reactions and were specific to their target organism. Optimum reaction mixture for the multiplex qPCR (10 µl volume) consisted of 1x TaqMan Multiplex Master Mix, 0.6 µM of each human primer, 0.5 µM of each dog and pig primer, 0.3 µM of human probe and 0.25 µM of pig and dog probes, and DNA template (concentration vary by samples). Standard curves and reaction efficiencies for the three assays are shown in Table 5.2. The assays were sensitive at detecting low DNA concentrations; the lowest detectable DNA concentration in the dilution series was 10-4 ng/µl for human and 10-5 ng/µl for pig and dog (Figure 5.1a-d). In contrast to the qPCR, the conventional one PCR did not detect DNA concentrations below 10-3 ng/µl for human, 10-4 ng/µl for pig and 10-2 ng/µl for dog (Figure 5.1a-d). In addition to the insensitivity of the dog primers to low DNA concentration, 132 amplification of the higher DNA concentration resulted in weak PCR bands compared to that of human and pig (Figure 5.1c and d). Also, for all three hosts, the PCR band intensity in mixed DNA samples (Figure 5.1d) was weaker than single host DNA samples (Figure 5.1a-c). Table 5.2. Standard curve and reaction efficiency of the three qPCR assays. Organism Human Pig Dog Slope -3.36 -3.50 -3.51 Abbreviation: SE, standard error % Efficiency Y-Intercept R squared 98.4 93.1 92.7 0.997 0.998 0.996 25.6 22.3 23.7 SE 0.21 0.19 0.22 Of the 375 Anopheles blood meals analyzed in the qPCR, 335 (89.3%) were positive for one or more of the three vertebrate hosts whereas only 40 (10.7%) were negative. Of the 335 positive samples, 225 (67.2%) were human blood meals, 18 (5.4%) were pigs, and 41 (12.2%) were dogs, with no other hosts detected in these single host results. The remaining 15.4% of positive samples were of mixed species, including 11 (3.3%) human + pig, 32 (9.55%) human + dog, four (1.2%) pig + dog, and four (1.2%) human + pig + dog. A random subset of these blood-meal samples (n = 188) were analyzed in the conventional, multiplex PCR and the results were compared with that of the qPCR by means of concordance-discordance frequency matrix (Table 5.3). Of 104 samples that were positive in both PCR methods, 16 (15.4%) were identified as mixed blood meals by the qPCR whereas only 3 (2.9%) were mixed blood meals by the conventional PCR. Twenty-four samples that were negative in the qPCR were also negative in the conventional PCR (i.e. 100% concordance). For those blood meals that were positive in the qPCR reactions (columns of Table 5.3), only 62.3% of human, 63.6% of pig, 50% of dog, 0% 133 human + pig mix, 20% of human + dog mix and 0% of pig + dog mix were concordant with conventional PCR results. The majority of the discordant blood meals were mosquitoes that were positive by the qPCR but turned out negative in the conventional PCR (n = 60); only a few (n = 16) were actually incongruent in the blood-meal types determined by the two methods. Of these, 13 were mosquitoes with mixed blood meals by qPCR but appeared as single blood-meal types in conventional PCR (Table 5.3). Figure 5.1. Results of conventional (the agarose gel images) and quantitative (the linear amplification plots) multiplex PCR both applied to the same 10-fold DNA dilution series. (a) PCR plot and gel image for human. (b) PCR plot and gel image for pig. (c) PCR plot and gel image for dog. (d) PCR plot and gel image for mixed sample. For the qPCR curves in panel d, color blue indicates human, green indicates pig and red indicates dog as in panel a-c. The mixed samples (panel d) contain equal DNA concentration (in ng/µl) of each host. 134 Table 5.3. Frequency of concordant and discordant Anopheles blood-meal results analyzed by quantitative versus conventional PCR. Conventional PCR Quantitative PCR Human Pig Dog Human/Pig Human/Dog Pig/Dog Negative Human Pig 66 (62) 0 (0) 1 (1) 0 (0) 0 (0) 0 (0) 0 (0) 7 (64) 0 (0) 0 (0) 0 (0) 0 (0) Dog 2 (8) 0 (0) 12 (50) 0 (0) 0 (0) 0 (0) Human/ Pig 2 (33) 1 (17) 0 (0) 0 (0) 0 (0) 0 (0) Human/ Dog 9 (60) 0 (0) 1 (7) 0 (0) 3 (20) 0 (0) Pig/ Dog 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) Negative 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 39 (37) 4 (36) 10 (42) 3 (50) 2 (13) 2 (100) 24 (100) Notes: Numbers outside of parenthesis are frequency of concordant (along diagonal) or discordant (off-diagonals) blood-meal types and inside parenthesis are percentage of column- wise total. The results of the statistical comparison of the mean host DNA concentration of mosquito blood meals that were detected versus undetected in the conventional PCR are shown in Figure 5.2. As predicted, for mosquitoes with single blood-meal host (grouped according to host species), the mean DNA concentration of those that was detected was significantly higher than the undetected ones for human (t-test: t(68) = 2.15, P = 0.04), pig (t-test: t(8) = 2.9, P = 0.02) and dog (t-test: t(17) = 2.49, P = 0.02) (Figure 5.2a). Similarly, for mosquitoes with mixed blood-meal (not grouped according to host species), the mean DNA concentration for those that were detected was significantly higher than the undetected ones (t-test: t(18) = 2.84, P = 0.01) (Figure 5.2b). The mean DNA concentration for detected single blood meals (Figure 5.2a) was significantly higher for pig compared to human (Tukey HSD test: mean difference = 0.63; 95% family-wise confidence interval or CI: 0.40–0.86; adjusted P < 0.0001), higher for dog compared 135 to human (mean difference = 1.27; 95% family-wise CI: 1.09–1.45; adjusted P < 0.0001) and higher for dog compared to pig (mean difference = 0.64; 95% family-wise CI: 0.36–0.9; adjusted P < 0.0001). For undetected single blood meals, the mean DNA concentration (Figure 5.2a) were not significantly different between pig and human (mean difference = -0.008; 95% family-wise CI: -0.42–0.40; adjusted P = 0.999) but were significantly higher for dog compared to human (mean difference = 0.73; 95% family-wise CI: 0.48–0.97; adjusted P < 0.0001) and for dog compared to pig (mean difference = 0.74; 95% family-wise CI: 0.28–1.18; adjusted P = 0.0007). Figure 5.2. Bar plot showing mean DNA concentration (with standard error bar) for mosquito blood meals that were detected versus undetected by conventional PCR. (a) Comparison of mean DNA concentration for single blood meals grouped according to human (detected, n = 6; undetected, n = 38), pig (detected, n = 7; undetected, n = 3) and dog (detected, n = 12; undetected, n = 10) hosts. (b) Comparison for mixed blood meals (detected, n = 19; undetected, n = 22) not grouped according to host species. Asterisks (*) indicate significant difference (P < 0.05) between groups (Student’s t-test). 136 5.5. Discussion The qPCR method offers several advantages over conventional PCR. First, when analyzing large number of samples with conventional PCR, the post-PCR steps such as gel-electrophoresis and manual scoring of results are labor-intensive, delay availability of data, generate large quantities of waste, and increase risk of exposure to ethidium bromide, a mutagen. These concerns are eliminated when qPCR is used. Second, it may be desirable to quantify DNA template or copy number of the target gene in samples, which qPCR can accomplish. Third, while qPCR is not always more sensitive than conventional PCR (Bastien, et al. 2008), many studies including this current one have shown that the former method can detect much lower DNA template concentrations compared to the latter (Dagher, et al. 2004; Dworkin, et al. 2002; Hierl, et al. 2004; Mary, et al. 2004). Thus, sensitivity of detection is better, as was shown here. Identification of arthropod blood meals by qPCR has been applied to sand flies, biting midges, kissing bugs, fleas and mosquitoes (Graham, et al. 2012; Gunathilaka, et al. 2016; Hurk, et al. 2007; Ibanez-Cervantes, et al. 2013; Sales, et al. 2015; Van-Der-Saag, et al. 2016; Woods, et al. 2009). Most of these were SYBR green-based systems; only three were probe-based. Of the three probe-based qPCR, one was for identifying Australian mammals in Culex mosquito blood meals and did not include humans, pigs or dogs (Hurk, et al. 2007), another was for identifying blood-meal hosts of biting midges and included humans and pigs but not dogs (Van-Der-Saag, et al. 2016), and another was for identifying flea blood meals and included humans and dogs but not pigs (Woods, et al. 2009). Notably, the human probe in the latter flea blood-meal study was tested here and found to cross-react with dog DNA. Surprisingly, a thorough Google Scholar search did not find a paper describing probe-based qPCR designed specifically for identifying mammalian hosts of Anopheles blood meals. 137 When evaluating the sensitivity of the qPCR assay from amplifications of 10-fold dilution series of target DNA samples, the lowest detectable concentration for human DNA (10-4 ng/µl) was ten-fold greater than for pigs and dogs which was 10-5 ng/µl. This difference was attributed to the copy numbers of the target DNA sequences; the pig and dog probes target mitochondrial genes which exist in multiple copies per cell, whereas the human probe targets a single-copy nuclear DNA sequence. Several human probes targeting various mitochondrial gene locus were designed and tested. However, they all exhibited non-specific amplification of the two nonhuman hosts in vitro despite appearing to be target specific by in silico test (Table S5.1, Appendix 5.1). Nevertheless, the detectable limit of human DNA concentration with the current probe (10-4 ng/µl) is sufficiently low for detecting mosquito blood meals. The lower detectable limit of qPCR compared to the conventional one, particularly dog, indicates a difference in the sensitivity of the two methods. The results show that the blood-meal qPCR was more sensitive at detecting host DNA in mosquitoes (detection success rate of 89%) compared to the more commonly used conventional, multiplex PCR (detection success rate of 55%). It is possible that the 11% of mosquitoes whose blood-meal hosts were not identified by the qPCR could have fed on other host sources (e.g., chickens, cats). However, when subjected to two conventional PCR utilizing generic mammalian and avian primers, none showed a positive result, which was consistent with findings from our previous study (Keven, et al. 2017). Thus, the likelihood that host breadth was greater than the three hosts we targeted with probes here is low. A common observation in all of these unamplified blood meals was that they all contained traces of blood in their abdomens (< 0.3 µl), based on light microscopy examination of the mosquito abdomens 4–8 hours after they were collected. Given the non-nucleated status of mammalian red blood cells and disproportionately 138 low ratio of white to red blood cells, the small volumes of blood meal were likely insufficient to yield a detectable concentration of DNA. The trace blood assertion was further supported by showing that the qPCR-quantified host DNA concentration of single blood meals was significantly higher in those samples that were detected in the conventional PCR than those that were undetected for all three hosts. The mean DNA concentration for the detected samples was expected to be statistically the same between the three hosts. However, the result showed otherwise; pig DNA concentration was significantly four-folds higher than humans and two-fold less than dog (Figure 5.2a). This heterogeneity in host DNA quantity could be caused by factors such as variation among the hosts in the number of white blood cells per unit volume of blood. However, a more plausible explanation is variation in the average quantity of blood mosquitoes obtain from each host as a result of variation in host sensitivity to mosquito bites. That is, humans are more sensitive than pigs followed by dogs to mosquito bites, causing them to quickly interrupt blood-feeding mosquitoes before they had time to acquire the maximum amount of blood. Despite the variation between the three hosts in their blood-meal DNA quantity, the mean DNA concentration for mosquitoes that were undetected in the conventional PCR was expected to be the same. The expectation is based on the reasoning that below a certain DNA concentration, the qPCR amplification becomes undetectable. However, the result showed that although the mean DNA concentration was statistically the same for humans and pigs as predicted, it was five-folds higher for dog (Figure 5.2a). This indicates that the dog conventional primers are less efficient and have higher sensitivity threshold than the other two hosts, which is consistent with its behavior observed in the amplification of DNA dilution series. 139 The conventional PCR did not detect mixed blood meals sensitively; most of the samples identified as mixed blood meals by qPCR were identified as single blood meals by the conventional method (Table 5.3). The inaccurate detection of mixed blood meals was attributed to low DNA quantity of one or more of the hosts in a mixed blood-meal sample and was supported by the result which showed statistically lower mean DNA concentration for undetected compared to detected hosts (Figure 5.2b). Thus, a primary outcome of this study is the sensitivity of the probe-based, qPCR method to detect blood from different mammal species in the same blood meal. This finding indicates that a significant proportion of unidentified blood-meal sources in studies that used the conventional, multiplex PCR (Barrera, et al. 2012; Bashar, et al. 2012; Burkot, et al. 2013; Busula, et al. 2015; Keven, et al. 2017; Liu, et al. 2011; Pappa, et al. 2011) may, among other factors, be due to the low sensitivity of this method. Furthermore, the proportion of mixed blood meals may be underestimated, and single blood meals overestimated in some published studies. At the very least, such findings indicate interrupted blood feeding, which is an important variable contributing to transmission of malaria (Tedrow, et al. 2019). 5.6. Conclusion This work describes the development of a probe-based multiplex qPCR for identifying the common blood-meal hosts of Anopheles mosquitoes from PNG. The qPCR assay was sensitive at detecting smaller amounts of target DNA in samples compared to standard PCR, which makes it appropriate for analysis of mosquito blood meals that often contain low amounts of host blood. Its ability to detect mixed blood meals compared to standard PCR makes it appropriate for use in studies that investigate interrupted feeding habits of mosquitoes. Although the qPCR assay described here was intended for analysis of mosquito blood meals from PNG and other parts of 140 South West Pacific, addition of probes specific to vertebrate hosts of mosquitoes from other parts of the world (e.g., cows and goats in Africa) is possible and encouraged. 141 APPENDIX 142 Appendix 5.1. Non-specific human primers Table S5.1. Human primers designed to target mitochondrial genes found to be human-specific by in silico test but were non- specific (also amplifies dog and/or pig DNA) by in vitro test. Gene target 5' dye 3' dye Product Accession Primer/probe Oligonucleotide sequence (5'-3') Tm (°C) %GC Cytochrome b MG272719.1 reverse primer CAGGAGGATAATGCCGATGTTTC forward primer TCAATCGCCCACATCACTCG NADH dehydrogenase subunit 1 NADH dehydrogenase subunit 5 Cytochrome C oxidase 1 probe TGGCTGAATCATCCGCTACCTTCACGC forward primer TCTCACCATCGCTCTTCTACT NC_012920.1 reverse primer AGGCTAGAGGTGGCTAGAATAA probe TAGGAGGCCTAGGTTGAGGTTGACC forward primer CTCCTAGACCTAACCTGACTAGAA MG272719.1 reverse primer GGATGAGTGGGAAGAAGAAAGA probe ACAGCACCAAATCTCCACCTCCAT forward primer TTTCACCTCCGCTACCATAATC NC_012920.1 reverse primer GCACTGCAGCAGATCATTTC probe TTTAGCTGACTCGCCACACTCCAC Note: N/A, not applicable 143 60.46 55 N/A 47.83 N/A 55.6 47.6 FAM N/A 45.45 N/A length (bp) 20 23 27 21 22 25 24 22 24 22 20 24 59.5 74.8 62.28 62.25 68.27 62 62 68 62 62 68 size (bp) Source N/A N/A QSY N/A N/A 168 Woods et al. 2009 99 This study 56 45.8 45.5 50 45.5 50 54.2 FAM QSY N/A N/A N/A N/A FAM QSY N/A N/A N/A N/A FAM QSY 136 This study 104 This study REFERENCES 144 REFERENCES Barrera R, Bingham AM, Hassan HK, Amador M, Mackay AJ, Unnasch TR. 2012. Vertebrate hosts of Aedes aegypti and Aedes mediovittatus (Diptera: Culicidae) in rural Puerto Rico. Am. J. Trop. Med. Hyg. 49(4):917–921. Bashar K, Tuno N, Ahmed TU, Howlader AJ. 2012. Blood-feeding patterns of Anopheles mosquitoes in a malaria-endemic area of Bangladesh. Parasit. Vectors 5:39. Bastien P, Procop GW, Reischl U. 2008. Quantitative real-time PCR is not more sensitive than “conventional” PCR. J. Clin. Microbiol. 46(6):1897–1900. Beier JC, Perkins PV, Wirtz RA, Koros J, Diggs D, Gargan TP, Koech DK. 1988. Bloodmeal identification by direct enzyme-linked immunosorbent assay (ELISA), tested on Anopheles (Diptera: Culicidae) in Kenya. J. Med. Entomol. 25(1):9–16. Boorman J, Mellor PS. 1977. A latex agglutination test for identification of blood-meals of Culicoides (Diptera: Ceratopogonidae). Bull. Entomol. Res. 67:305–311. Boreham PFL, Garrett-Jones C. 1973. Prevalence of mixed blood meals and double feeding in a malaria vector (Anopheles sacharovi Favre). Bull. World Health Organ. 48:605–614. Boreham PFL, Lenahan JK, Bouizaguet R, Storey J, Ashkar TS, Nambiar R, Matsushima T. 1979. Studies on multiple feeding by Anopheles gambiae s.l. in a Sudan savanna area of north Nigeria. Trans. R. Soc. Trop. Med. Hyg. 73(4):418–423. Burkot TR, Dye C, Graves PM. 1989. An analysis of some factors determining the sporozoite rates, human blood indexes, and biting rates of members of the Anopheles punctulatus complex in Papua New Guinea. Am. J. Trop. Med. Hyg. 40(3):229–234. Burkot TR, Goodman WG, DeFoliart GR. 1981. Identification of mosquito blood meals by enzyme-linked immunosorbent assay. Am. J. Trop. Med. Hyg. 30:1336–1341. Burkot TR, Graves PM, Paru R, Lagog M. 1988. Mixed blood feeding by the malaria vectors in the Anopheles punctulatus complex (Diptera: Culicidae). J. Med. Entomol. 25(4):205– 213. Burkot TR, Russell TL, Reimer LJ, Bugoro H, Beebe NW, Cooper RD, Sukawati S, Collins FH, Lobo NF. 2013. Barrier screens: a method to sample blood-fed and host-seeking exophilic mosquitoes. Malar. J. 12:49. Busula AO, Takken W, Loy DE, Hahn BH, Mukabana WR, Verhulst NO. 2015. Mosquito host preferences affect their response to synthetic and natural odour blends. Malar. J. 14:133. 145 Dagher H, Donninger H, Hutchinson P, Ghildyal R, Bardin P. 2004. Rhinovirus detection: comparison of real-time and conventional PCR. J. Virol. Methods 117:113–121. Dworkin LL, Gibler TM, Gelder RNV. 2002. Real-time quantitative polymerase chain reaction diagnosis of infectious posterior uveitis. Ophthalmol 120:1534–1539. Edman JD, Downe AER. 1964. Host-blood sources and multiple feeding habits of mosquitoes in Kansas. Mosq. News 24:154–160. Fritz ML, Miller JR, Bayoh N, Vulule JM, Landgraf JR, Walker ED. 2013. Application of a reverse dot blot, DNA-DNA hydridization method to quantify host-feeding tendencies of two sibling species in the Anopheles gambiae complex. Med. Vet. Entomol. 27(4). Garrett-Jones C. 1964a. The human blood index of malaria vectors in relation to epidemiological assessment. Bull. World Health Organ. 30:241–261. Garrett-Jones C. 1964b. Prognosis for interuption of malaria transmission through assessment of the mosquito's vectorial capacity. Nature 204:1173–1175. Graham CB, Black WC, Boegler KA, Montenier JA, Holmes JL, Gage KL, Eisen RJ. 2012. Combining real-time polymerase chain reaction using SYBR green I detection and sequencing to identify vertebrate bloodmeals in fleas. J. Med. Entomol 49(6):1442–1452. Gunathilaka N, Denipitiya T, Hapugoda M, Abeyewickreme W, Wickremasinghe R. 2016. Determination of the foraging behaviour and blood meal source of malaria vector mosquitoes in Trincomalee District of Sri Lanka using a multiplex real time polymerase chain reaction assay. Malar. J. 15:242. Hamer GL, Chaves LF, Anderson TK, Kitron UD, Brawn JD, Ruiz MO, Loss SR, Walker ED, Goldberg TL. 2011. Fine-scale variation in vector host use and force of infection drive localized patterns of West Nile virus transmission. PLoS ONE 6(8). Hamer GL, Kitron UD, Goldberg TL, Brawn JD, Loss SR, Ruiz MO, Hayes DB, Walker ED. 2009. Host selection by Culex pipiens mosquitoes and West Nile virus amplification. Am. J. Trop. Med. Hyg. 80(2):268–278. Hierl T, Reischl U, Lang P, Hebart H, Stark M, Kyme P, Autenrieth IB. 2004. Preliminary evaluation of one conventional nested and two real-time PCR assays for the detection of Toxoplasma gondii in immunocompromised patients. J. Med. Microbiol. 53:629–32. Hurk AFVD, Smith IL, Smith GA. 2007. Development and evaluation of real-time polymerase chain reaction assays to identify mosquito (Diptera: Culicidae) bloodmeals originating from native Australian mammals. J. Med. Entomol 44(1):85–92. Ibanez-Cervantes G, Martínez-Ibarra A, Nogueda-Torres B, Lopez-Orduna E, Alonso AL, Perea C, Maldonado T, Hernandez JM, Leon-Avila G. 2013. Identification by Q-PCR of 146 Trypanosoma cruzi lineage and determination of blood meal sources in triatomine gut samples in Mexico. Parasitol. Int. 62:36–43. Kent RJ, Norris DE. 2005. Identification of mammalian blood meals in mosquitoes by a multiplexed polymerase chain reaction targeting cytochrome b. Am. J. Trop. Med. Hyg. 73(2):336–342. Keven JB, Reimer L, Katusele M, Koimbu G, Vinit R, Vincent N, Thomsen E, Foran DR, Zimmerman PA, Walker ED. 2017. Plasticity of host selection by malaria vectors of Papua New Guinea. Parasit. Vectors 10:95. Liu X-B, Liu Q-Y, Guo Y-H, Jiang J-Y, Ren D-S, Zhou G-C, Zheng C-J, Zhang Y, Liu J-L, Li Z-F and others. 2011. The abundance and host-seeking behavior of culicine species (Diptera: Culicidae) and Anopheles sinensis in Yongcheng city, people’s Republic of China. Parasit. Vectors 4:221. Logue K, Keven JB, Cannon MV, Reimer L, Siba P, Walker ED, Zimmerman PA, Serre D. 2016. Unbiased characterization of Anopheles mosquito blood meals by targeted high- throughput sequencing. PLoS Negl. Trop. Dis. 10(3). Mary C, Faraut F, Lascombe L, Dumon H. 2004. Quantification of Leishmania infantum DNA by a real-time PCR assay with high sensitivity. J. Clin. Microbiol. 42(11):5249–5255. Meece JK, Reynolds CE, Stockwell PJ, Jenson TA, Christensen JE, Reed KD. 2005. Identification of mosquito bloodmeal source by terminal restriction fragment length polymorphism profile analysis of the cytochrome B gene. J. med. Entomol. 42(4):657– 667. Molaei G, Andreadis TA, Armstrong PM, Anderson JF, Vossbrinck CR. 2006. Host feeding patterns of Culex mosquitoes and West Nile Virus transmission, northeastern United States. Emerg. Infect. Dis. 12:468–474. Molaei G, Cummings RF, Su T, Armstrong PM, Greg A. Williams , Cheng M-L, Webb JP, Andreadis TG. 2010. Vector-host interactions governing epidemiology of West Nile Virus in Southern California. Am. J. Trop. Med. Hyg. 83(6):1269–1282. Pappa V, Reddy M, Overgaard HJ, Abaga S, Caccone A. 2011. Short Report: Estimation of the human blood index in malaria mosquito vectors in Equatorial Guinea after indoor antivector interventions. Am. J. Trop. Med. Hyg. 84(2):298–301. Sales KGdS, Costa PL, Morais RCSd, Otranto D, Brandao-Filho SP, Cavalcanti MdP, Dantas- Torres F. 2015. Identification of phlebotomine sand fly blood meals by real-time PCR. Parasit. Vectors 8:230. Senior-White RA. 1952. Studies on the bionomics of Anopheles aquasalis Curry, 1932 (concld.). Indian J. Malariol. 6:29–72. 147 Shemanchuk JA, Downe AER, Burgess L. 1963. Hosts of mosquitoes (Diptera: Culicidae) from the irrigated areas of Alberta. Mosq. News 23:336–341. Tedrow RE, Zimmerman PA, Abbott KC. 2019. Multiple blood feeding: A force multiplier for transmission. Trends Parasitol. 35(12):949–952. Van-Der-Saag MR, Gu X, Ward MP, Kirkland PD. 2016. Development and evaluation of real- time PCR assays for bloodmeal identification in Culicoides midges. Med. Vet. Entomol. 30:155–165. Weitz B. 1956. Identification of blood meals of blood-sucking arthropods Bull. World Health Organ. 15:473–490. Woods ME, Montenieri JA, Eisen RJ, Zeidner NS, Borchert JN, Laudisoit A, Babi N, Atiku LA, Enscore RE, Gage KL. 2009. Identification of flea blood meals using multiplexed real- time polymerase chain reaction targeting mitochondrial gene fragments. Am. J. Trop. Med. Hyg. 80(6):998–1003. Ye J, Coulouris G, Zaretskaya I, Cutcutache I, Rozen S, Madden TL. 2012. Primer-BLAST: A tool to design target-specific primers for polymerase chain reaction. BMC Bioinform. 13:134. 148 CHAPTER 6: PLASTICITY OF HOST SELECTION BY ANOPHELES VECTORS OF MALARIA IN MADANG, PAPUA NEW GUINEA: IMPLICATIONS FOR PERSISTENT MALARIA TRANSMISSION Publication Keven JB, Reimer LJ, Katusele M, Koimbu G, Vinit R, Vincent N, Thomsen E, Foran DR, Zimmerman PA, Walker ED. 2017. Plasticity of host selection by malaria vectors of Papua New Guinea. Parasit. Vectors 10:95 6.1. Abstract Host selection is an important indicator of malaria transmission potential of Anopheles vectors and their response to indoor-deployed insecticidal bed nets. Despite the ongoing nationwide bed net-based malaria control campaign in PNG, little is known about the host selection of the local vectors. This study investigated the host selection of Anopheles vector populations in Madang. Blood-fed mosquitoes were sampled using the barrier screen method and blood meals analyzed for vertebrate host source with PCR methods. The abundance of common hosts was estimated in census surveys and exact binomial test was used to determine if each host was selected in proportion to its abundance relative to the other host species. Four thousand one hundred and twenty-four blood-fed Anopheles females of six species were sampled across seven Madang villages. Of these, 3,554 (86%) yielded a definitive host source; all were human, pig, or dog. Results show that hosts were not selected in proportion to their abundance, but rather were under-selected or over-selected by the mosquitoes. Low mosquito anthropophagy was associated with high bed net usage. Five species, namely An. farauti (s.s.), An. farauti no. 4, An. koliensis, 149 An. longirostris and An. punctulatus (s.s.), exhibited plasticity in their host selection, likely in response to varying levels of bed net usage and were thus considered opportunistic in their intrinsic host preference. In contrast, An. bancroftii exhibited strong affinity for pig blood despite overwhelmingly high human abundance and was considered zoophilic. By feeding on alternative hosts the likelihood of these opportunistic species contacting the bed nets inside houses is lower, making them difficult to control. By maintaining high population size on non-human blood sources, the proportion that feed on humans outdoors can sustain transmission despite high bed net usage in the village. 6.2. Introduction Host selection is an outcome of the combined effects of a mosquito’s intrinsic (genetic) host preference for a particular host species modulated by extrinsic factors (Lyimo and Ferguson 2009; Takken and Verhulst 2013). That is, even though a mosquito may intrinsically prefer a particular host species due to genetic factors, environmental factors such as availability or accessibility of the preferred host may cause the mosquito to resort to an alternative one. Therefore, in mosquito-borne disease epidemiology host selection is an important determinant of vectorial capacity, because it influences the extent to which mosquitoes in populations feed predominantly on humans or non-humans (Dye 1986; Garrett-Jones 1964). Thus, an Anopheles population whose members intrinsically prefer humans are potential vectors of malaria. However, the vectorial capacity of the mosquito population depends on whether extrinsic conditions allow the mosquitoes to actually feed on humans. Knowledge of host selection is not only important for evaluating the vectorial capacity of a vector population, but also for guiding vector-based malaria control programs, such as the 150 distribution of insecticidal bed nets. The implementation of bed nets is appropriate if we know that local vectors are sufficiently anthropophilic that bed nets will have the intended effect (Russell, et al. 2016). The inflexibility of anthropophilic species to utilize alternative hosts causes them to pursue humans inside houses and thus increases their likelihood of becoming exposed to the insecticides in the bed net fabric. The increased likelihood of contacting bed nets enables reduction of their population size to levels insufficient to support transmission. In contrast, if the mosquitoes are opportunistic and exhibit plasticity in host selection, then bed nets may have little effect because these mosquitoes can maintain high population size by feeding on non-human hosts outdoor. By maintaining high population size, the proportion of opportunistic vectors that feed on human individuals before they go under their bed nets can sufficiently sustain transmission in the community. Treating the alternative hosts with endectocides lethal to blood-feeding mosquitoes may be more appropriate for controlling such opportunistic vectors. By implementing both methods, the anthropophilic and opportunistic vectors might be successfully controlled. Human malaria is endemic to PNG (Muller, et al. 2003) and the main vectors are members of the An. punctulatus (s.l.) group (Beebe and Cooper 2002; Beebe, et al. 2015; Beebe and Saul 1995), primarily An. punctulatus (s.s.), An. koliensis, An. farauti (s.s.), An. farauti no. 4, and An. hinesorum (Benet, et al. 2004; Burkot, et al. 1989; Burkot, et al. 1988b; Cooper, et al. 2009; Hetzel, et al. 2016; Reimer, et al. 2016; Thomsen, et al. 2017). An. bancroftii and An. longirostris are also vectors of malaria in PNG (Cooper, et al. 2009). Nationwide insecticidal bed net-based vector control campaign has been ongoing in PNG since 2004 (Hetzel 2009; Hetzel, et al. 2014a; Hetzel, et al. 2012) to help alleviate the burden of malaria. However, little is known about the host selection behavior of these vectors and their relationship with bed net usage. The 151 research described in this chapter addresses this knowledge gap and provides guidance on existing as well as new vector control strategies in PNG. 6.3. Methods 6.3.1. Study sites and mosquito sampling This study was conducted in seven malaria-endemic rural villages (Bulal, Dimer, Kokofine, Matukar, Megiar, Mirap and Wasab) of Madang, PNG whose geographic locations and ecogeographic environments were described in detail in chapter 2, section 2.3. Mosquitoes were collected in each village using the barrier screen method with sampling design, procedure and time period described in detail in chapter 2, section 2.3. Only the blood-fed female Anopheles mosquitoes were used in this study. 6.3.2. Identification of Anopheles species and their blood-meal hosts The abdomen of each blood-fed mosquito was separated from the rest of the body using sterile technique and DNA was isolated from it using the DNeasy Blood and Tissue kit. Mosquitoes morphologically identified as member of the An. punctulatus (s.l.) group were further identified to their species using a PCR assay (Beebe and Saul 1995). As humans, pigs, and dogs are the main blood-meal sources of PNG mosquitoes, based on the findings in chapter 5 and elsewhere (Burkot, et al. 1989; Burkot, et al. 1988a; Logue, et al. 2016), the DNA samples were first tested for these three hosts using the multiplex qPCR described in chapter 5. Samples that did not amplify in the qPCR likely fed on other hosts (e.g., cats and chickens) and were subjected to two standard PCR with a generic mammalian primer pair (forward: 5′- CCA TCC AAC ATC TCA GCA TGA TGA AA-3′ and reverse: 5′- GCC CCT CAG AAT GAT ATT TGT CCT CA -3′) that 152 targets a 395 bp region of the mitochondrial cytochrome b gene (cytb), or an avian primer pair (forward: 5′- GAC TGT GAC AAA ATC CCN TTC CA -3′ and reverse: 5′- GGT CTT CAT CTY HGG YTT ACA AGA C -3′) which targeted a 508 bp region of avian cytb gene (Molaei, et al. 2006). The standard PCR reaction mixture consisted of ca 20 ng of DNA template added to a PCR tube (50 µl volume) containing 10 mM Tris at pH 8.3, 50 mM KCl, 1.5 mM MgCl2, 0.01% gelatin, 1.0 mM dNTP, 0.5 units of Taq polymerase, and 50 pmol of each primer. Thermal cycle consisted of 1 cycle of 95 °C for 5 minutes followed by 35 cycles of 95 °C for 1 minute, 58 °C for 1 minute, and 72 °C for 1 minute, and 1 cycle of 72 °C for 7 minutes. PCR amplicons were run on 2% ethidium bromide-stained agarose gels and visualized under ultraviolet light. Amplicons of the PCR positive samples were purified using QIAquick PCR Purification Kit (Catalog Number 28106; Qiagen) and sequenced by direct sequencing. The DNA sequences were subjected to a BLAST search against vertebrate cytb sequences in the GenBank database. A subject sequence that had ≥ 99% sequence similarity to the query sequence was considered the likely host from which the mosquito fed. 6.3.3. Demographic survey Surveys were conducted during the mosquito sampling period in each village in 2012 (Dimer, Kokofine, Matukar, Mirap and Wasab), 2016 (Mirap and Wasab) and 2017 (Bulal and Megiar). Heads of households were interviewed for information on number of people per household, their age, number of bed nets owned, and number and species of domestic animals owned by the household. Individuals were asked if they slept under a bed net the night before the interview. 153 6.3.4. Data analyses The tendency of a vector population to feed on a particular host species compared to other available hosts, was quantified by performing two-tailed exact binomial test which compared the observed blood index (i.e., the proportion of blood-fed mosquitoes that fed on that host species) and the expected blood index (the proportion of that host relative to all other host species in village census count). Only single blood-meal data were used for this analysis; mixed blood meals were excluded. A host species was considered over-selected by a mosquito population if the observed blood index was statistically greater than the expected index, under-selected if the observed was less than expected, or fed on unbiasedly (i.e., feed on the host in proportion to its relative abundance) if there was no statistical difference. Each mosquito was assigned a binary value 1 if its blood meal was from a human or 0 if it was from a pig or dog and multiple logistic regression was used to evaluate the relationship between the probability of feeding on human blood meal and two predictor variables: i) bed net usage (proportion of interviewed villagers who reportedly slept under a bed net the night before the interview) and ii) mean number of household members per bed net (households as unit of replication). Both the binomial and logistic regression tests were computed using the R functions binom.test and glm of the package stats. Significance level for all tests was based on type I error rate of 5%. 6.4. Results 6.4.1. Mosquito blood-meal sources A total of 4,124 blood-fed Anopheles mosquitoes of six different species were analyzed and the blood-meal hosts of 3,554 (86%) were identified. The 14% of the samples that did not was likely due to trace amounts of blood materials insufficient to yield a PCR-detectable DNA 154 concentration (see chapter 5 for more details). Human, pig, and dog were the only host species identified in the blood meals despite the presence of cats, chickens and ducks in all the villages. The exact number of mosquitoes of each Anopheles species that fed on a human, pig, or dog, or a mixture of these hosts (human + pig, human + dog, pig + dog or human + pig + dog) in each village are shown in supplementary Table S6.1–6.6 (Appendix). The blood-meal source data in the supplementary Table S6.1–6.6 were also expressed as relative percent proportion in Figure 6.1. In Figure 6.1, each bar represents the blood-meal sample of an Anopheles species in a village. In all 16 categories (i.e., 16 bars in Figure 6.1), there were more single than mixed blood meals; the highest proportion of mixed blood meals was in An. koliensis in Megiar (18.2%; Figure 6.1d). For single blood meals, the proportion of each of the three hosts in a sample (i.e., each bar in Figure 6.1) varied greatly. For example, Mirap An. bancroftii blood meals consisted of 12.8% humans, 76.9% pigs and 7.7% dogs (Figure 6.1a), and An. punctulatus (s.s.) in Wasab consisted of 71.9% humans, 10.4% pigs and 13.4% dogs (Figure 6.1f). Among villages, the proportions of each host in blood-meal samples of the same mosquito species varied among villages (Figure 6.1c-f). For example, the proportion of human blood meals in An. longirostris was 67.9% in Wasab but was only 15% in Mirap (Figure 6.1e). 155 Figure 6.1. Proportion of blood-meal sources of mosquitoes sampled in study villages. (a) An. bancroftii in Mirap (n = 39). (b) An. farauti no. 4 in Kokofine (n = 462). (c) An. farauti (s.s.) in Matukar (n = 52), Megiar (n = 214), Mirap (n = 1,332) and Wasab (n = 27). (d) An. koliensis in Bulal (n = 34), Megiar (n = 11), Mirap (n = 480) and Wasab (n = 548). (e) An. longirostris in Mirap (n = 19) and Wasab (n = 84). (f) An. punctulatus (s.s.) in Bulal (n = 8), Dimer (n = 34), Mirap (n = 25) and Wasab (n = 164). 156 6.4.2. Mosquito host selection A mosquito population was defined here as mosquitoes of a particular species collected from a specific location and time. The results of host selection tests are presented for two populations of An. bancroftii (Table 6.1), one population of An. farauti no. 4 (Table 6.2), five populations of An. farauti (s.s.) (Table 6.3), six populations of An. koliensis (Table 6.4), two populations of An. longirostris (Table 6.5) and six populations of An. punctulatus (s.s.) (Table 6.6). The results show that host selection varied among different populations of the same Anopheles species, except An. bancroftii and An. farauti no. 4; the former did not show variation in host selection and the latter consisted of a single population and comparison of host selection among populations could not be done for this species. The two An. bancroftii populations were from Mirap village but two different sampling periods (Table 6.1). Both over-selected pigs, under- selected humans and fed unbiasedly on dogs, although sample size was low for both populations (n £ 29). Like An. bancroftii, the population of An. farauti no. 4 from Kokofine (generous sample size, n = 441) over-selected pigs but under-selected or fed unbiasedly the other two hosts (Table 6.2). Unlike An. bancroftii, the five An. farauti (s.s.) populations varied in their host selection (Table 6.3). The Matukar population fed unbiasedly on all three hosts, although the sample size was low (n = 43). For the Wasab population, none of the three hosts was over- selected; all were under-selected (pigs) or fed on unbiasedly (humans and dogs), although the sample size was also low (n = 22). The two Mirap and one Megiar populations where the sample sizes were modest (n ³ 191) all over-selected pigs and under-selected humans but varied in their selection of dogs. Of the six An. koliensis populations (Table 6.4), four (Bulal, Megiar, Mirap 2012–2013 and Wasab 2012–2013) fed unbiasedly on all three hosts, although the sample sizes of these populations were low (n £ 28; see Table 6.4). For Mirap 2015–2017 population where 157 sample size was generous (n = 409), humans were under-selected whereas both non-human hosts were over-selected. In contrast, Wasab 2015–2016 population (generous sample size, n = 453) over-selected humans whereas the non-humans were under-selected or fed on unbiasedly. For the two An. longirostris populations (Table 6.5), the Mirap 2012-2013 population, pigs were over- selected whereas humans and dogs were under-selected or fed on unbiasedly, but the sample size was low (n = 12). In contrast, the Wasab 2012–2013 population with modest sample size (n = 73) over-selected humans, under-selected dogs and unbiasedly fed on pigs. For the six An. punctulatus (s.s.) populations (Table 6.6), except for Wasab 2015–2016 population which had a modest samples size (n = 100), the rest had low samples sizes (n £ 31, see Table 6.6). Three populations, Bulal, Wasab 2012–2013 and Mirap 2012–2013 fed unbiasedly on all three hosts. The Dimer population over-selected dogs but fed unbiasedly on both humans and pigs. For Mirap 2015–2017 population, pigs were over-selected, humans were under-selected, and dogs were fed on unbiasedly. For Wasab 2015–2016 population, humans and dogs were over-selected whereas pigs were under-selected. 158 Table 6.1. Number of humans, pigs and dogs in village census and blood meals of two populations of An. bancroftii along with their host selection tendency based on exact binomial comparison of the expected (Exp) and observed (Obs) blood index. Village Period Mirap 2012–2013 Mirap 2015–2017 Host humans pigs dogs humans pigs dogs Abbreviation: NS, not significant Census Blood meals Exp 0.767 732 0.115 110 112 0.117 4 23 2 1781 181 90 1 7 1 0.87 0.09 0.04 Obs (95% CI) 0.137 (0.039–0.317) 0.793 (0.603–0.920) 0.069 (0.008–0.228) 0.111 (0.003–0.482) 0.778 (0.400–0.972) 0.111 (0.003–0.482) Host selection P < 0.0001 Under-selected < 0.0001 Over-selected NS Unbiased < 0.0001 Under-selected < 0.0001 Over-selected NS Unbiased Table 6.2. Number of humans, pigs and dogs in village census and blood meals of a population of An. farauti no. 4 along with their host selection tendency based on exact binomial comparison of the expected (Exp) and observed (Obs) blood index. Period Village Kokofine 2012–2013 Host humans pigs dogs Census Blood meals Exp 0.536 167 0.215 70 88 0.271 220 188 33 Obs (95% CI) 0.499 (0.451–0.547) 0.426 (0.380–0.474) 0.075 (0.052–0.103) Abbreviation: NS, not significant 159 P Host selection Unbiased NS < 0.0001 Over-selected < 0.0001 Under-selected Table 6.3. Number of humans, pigs and dogs in village census and blood meals of five populations of An. farauti (s.s.) along with their host selection tendency based on exact binomial comparison of the expected (Exp) and observed (Obs) blood index. Period Village Matukar 2012–2013 Megiar Mirap Mirap Wasab 2015–2017 2012–2013 2015–2017 2015–2016 Host humans pigs dogs humans pigs dogs humans pigs dogs humans pigs dogs humans pigs dogs Abbreviation: NS, not significant Obs (95% CI) 0.767 (0.614–0.882) 0.047 (0.006–0.158) 0.186 (0.084–0.334) 0.539 (0.466–0.611) 0.335 (0.269–0.407) 0.126 (0.082–0.181) 0.473 (0.428–0.519) 0.444 (0.400–0.490) 0.082 (0.059–0.110) 0.386 (0.331–0.444) 0.399 (0.344–0.457) 0.215 (0.170–0.265) 0.727 (0.498–0.893) 0.136 (0.029–0.349) 0.136 (0.029–0.349) P NS NS NS Host selection Unbiased Unbiased Unbiased < 0.0001 Under-selected < 0.0001 Over-selected NS Unbiased < 0.0001 Under-selected < 0.0001 Over-selected 0.016 Under-selected < 0.0001 Under-selected < 0.0001 Over-selected < 0.0001 Over-selected NS 0.026 NS Unbiased Under-selected Unbiased Census Blood meals Exp 432 0.751 0.117 67 76 0.132 33 2 8 734 171 86 732 110 112 1781 181 90 295 207 64 103 64 24 230 216 40 117 121 65 16 3 3 0.741 0.172 0.087 0.767 0.115 0.117 0.87 0.09 0.04 0.52 0.37 0.11 160 Table 6.4. Number of humans, pigs and dogs in village census and blood meals of six populations of An. koliensis along with their host selection tendency based on exact binomial comparison of the expected (Exp) and observed (Obs) blood index. Village Period Megiar 2017 Host humans pigs dogs Census Blood meals Exp 0.741 734 0.172 171 86 0.087 6 2 1 732 110 112 1781 181 90 211 71 48 295 207 64 9 1 0 193 65 151 11 1 0 331 69 53 Mirap Mirap Wasab Wasab 2012–2013 humans pigs dogs 2015–2017 humans pigs dogs 2012–2013 humans pigs dogs 2015–2016 humans pigs dogs Abbreviation: NS, not significant 0.767 0.115 0.117 0.87 0.09 0.04 0.639 0.215 0.145 0.52 0.37 0.11 161 Obs (95% CI) 0.667 (0.299–0.925) 0.222 (0.028–0.600) 0.111 (0.003–0.482) 0.900 (0.555–0.997) 0.100 (0.003–0.445) 0.000 (0.000–0.308) 0.472 (0.423–0.522) 0.159 (0.125–0.198) 0.369 (0.322–0.418) 0.917 (0.615–0.998) 0.083 (0.002–0.385) 0.000 (0.000–0.265) 0.731 (0.687–0.771) 0.152 (0.120–0.189) 0.117 (0.089–0.150) P NS NS NS NS NS NS Host selection Unbiased Unbiased Unbiased Unbiased Unbiased Unbiased < 0.0001 < 0.0001 < 0.0001 Under-selected Over-selected Over-selected NS NS NS Unbiased Unbiased Unbiased < 0.0001 < 0.0001 NS Over-selected Under-selected Unbiased Table 6.4 (cont’d). Village Bulal Period 2017 Host humans pigs dogs Census Blood meals Exp 0.719 489 0.135 92 99 0.146 16 4 8 Obs (95% CI) 0.571 (0.372–0.755) 0.143 (0.040–0.327) 0.286 (0.132–0.487) P NS NS NS Host selection Unbiased Unbiased Unbiased Abbreviation: NS, not significant Table 6.5. Number of humans, pigs and dogs in village census and blood meals of two populations of An. longirostris along with their host selection tendency based on exact binomial comparison of the expected (Exp) and observed (Obs) blood index. Village Wasab Mirap Period 2012–2013 2012–2013 Host humans pigs dogs humans pigs dogs Abbreviation: NS, not significant Census Blood meals Exp 0.639 211 71 0.215 0.145 48 57 14 2 732 110 112 3 8 1 0.767 0.115 0.117 Obs (95% CI) 0.781 (0.669–0.869) 0.192 (0.109–0.301) 0.027 (0.003–0.095) 0.250 (0.055–0.572) 0.667 (0.349–0.901) 0.083 (0.002–0.385) P 0.014 NS 0.001 Host selection Over-selected Unbiased Under-selected 0.0002 Under-selected < 0.0001 Over-selected NS Unbiased 162 Table 6.6. Number of humans, pigs and dogs in village census and blood meals of six populations of An. punctulatus (s.s.) along with their host selection tendency based on exact binomial comparison of the expected (Exp) and observed (Obs) blood index. Village Dimer Mirap Mirap Wasab Wasab Period 2012 2012–2013 2015–2017 2012–2013 2015–2016 Host humans pigs dogs humans pigs dogs humans pigs dogs humans pigs dogs humans pigs dogs Abbreviation: NS, not significant Census Blood meals Exp 0.712 622 0.181 158 94 0.108 20 2 9 732 110 112 1781 181 90 211 71 48 295 207 64 8 1 0 0 12 2 24 4 2 72 9 19 0.767 0.115 0.117 0.87 0.09 0.04 0.639 0.215 0.145 0.52 0.37 0.11 Obs (95% CI) 0.645 (0.454–0.808) 0.065 (0.008–0.214) 0.290 (0.142–0.480) 0.889 (0.518–0.997) 0.111 (0.003–0.482) 0.000 (0.000–0.336) 0.00 (0.000–0.232) 0.857 (0.572–0.982) 0.143 (0.018–0.428) 0.800 (0.614–0.923) 0.133 (0.038–0.307) 0.067 (0.008–0.221) 0.720 (0.621–0.805) 0.090 (0.042–0.164) 0.190 (0.118–0.281) P NS NS 0.004 NS NS NS Host selection Unbiased Unbiased Over-selected Unbiased Unbiased Unbiased < 0.0001 Under-selected < 0.0001 Over-selected NS Unbiased NS NS NS Unbiased Unbiased Unbiased < 0.0001 Over-selected < 0.0001 Under-selected 0.025 Over-selected 163 Table 6.6 (cont’d) Village Period Bulal 2017 Host humans pigs dogs Census 489 92 99 Blood meals Exp 4 0.719 1 0.135 3 0.146 Obs (95% CI) 0.500 (0.157–0.843) 0.125 (0.003–0.527) 0.375 (0.085–0.755) P NS NS NS Host selection Unbiased Unbiased Unbiased Abbreviation: NS, not significant 164 6.4.3. Relationship between bed net usage and anthropophagy The proportion of interviewed villagers that reportedly slept under a bed net and the mean number of household members per bed net in each village during one or two mosquito sampling periods are shown in Table 6.7. The results of multiple logistic regression which tested the association between the two predictor variables and the probability of obtaining a human blood meal by mosquitoes (i.e., anthropophagy) of An. farauti (s.s.), An. koliensis and An. punctulatus (s.s.) are shown in Table 6.8; the other three vector species were not assessed because of low sample size or a single population. For all of the three vector species tested, low anthropophagy was associated with high bed net usage in the village. This association was strongly significant for An. farauti (s.s.) and An. koliensis (P < 0.05), and showed the same trend but was “marginally significant” (P = 0.096) for An. punctulatus (s.s.). No statistical relationship was observed between anthropophagy and mean number of persons per net for all three vectors. Table 6.7. Two bed net metrics in each village during one or more sampling periods. Village (sampling period) Bulal (2017) Dimer (2012) Matukar (2012) Megiar (2017) Mirap (2012–2013) Mirap (2015–2017) Wasab (2012–2013) Wasab (2015–2016) Proportion of people using a bed net 0.93 0.46 0.55 0.81 0.84 0.94 0.58 0.90 Mean number of people/bed net 1.47 1.41 1.48 1.16 2.45 1.81 1.89 1.56 165 Table 6.8. Results of multiple logistic regression which tested for association between mosquito anthropophagy and two predictor variables. Species An. farauti (s.s.) An. koliensis N 1,537 920 An. punctulatus (s.s.) 219 variable Bed net usage Mean persons/net Bed net usage Mean persons/net Bed net usage Mean persons/net b -3.58 -0.15 -15.87 -0.98 -1.37 0.66 SE 0.82 0.11 3.71 0.55 0.82 0.66 P < 0.0001 0.162 < 0.0001 0.076 0.096 0.322 Abbreviations: N, number of mosquitoes; b, regression coefficient; SE, standard error of the regression coefficient; P, p-value 6.5. Discussion This study found that humans, pigs, and dogs were the primary blood-meal sources of the Anopheles vectors of malaria in these villages despite the presence of cats, chickens, and ducks in the villages. The narrow host range observed here is consistent with previous studies (Burkot, et al. 1989; Burkot, et al. 1988a; Charlwood, et al. 1985b; Hii, et al. 1997; Hii, et al. 2001; Logue, et al. 2016), although some previous studies also detected a small number of mosquitoes that had fed on cats, chickens, opossum and bats. This difference may be due to the use of different sampling methods such as outdoor resting collection by some previous studies, but those hosts were uncommon in all cases compared to the three primary hosts. A number of studies (Burkot, et al. 1989; Burkot, et al. 1988a; Charlwood, et al. 1985b; Hii, et al. 1997; Hii, et al. 2001) have described the host preference of PNG vectors based on patterns of host selection using the human blood index. A mosquito species that had consistently high human blood index was classified as anthropophilic (“human loving”), whilst those with consistently low human blood index as zoophilic (“animal loving”); one with variable human 166 blood index was considered opportunistic (i.e. selects hosts indiscriminately). However, an important limitation with this approach is that it does not take into account the relative abundance of the different host species within the mosquito foraging habitat. Various measures including the forage ratio (Hess, et al. 1968), the feeding index (Kay, et al. 1979) and the Manly resource selection ratio (Hamer, et al. 2011; Hamer, et al. 2009), have been used to evaluate mosquito host selection. These measures are inter-related, and each requires some information of the range of available hosts, their relative abundance and related attributes. The Manly resource selection ratio is preferable because it provides a statistical test of departure from random host selection with confidence intervals. However, there is no upper limit to the range of values it can take. Therefore, in this study exact binomial test which also provides a statistical test of departure from randomness with confidence intervals and whose values are constraint between zero (lower limit) and one (upper limit) was used to quantify host selection. The intrinsic host preference of the Anopheles species investigated here can be delineated based on their pattern of host selection. The tendency of the two An. bancroftii populations to feed on pigs despite overwhelming abundance of humans relative to pigs in the village indicates that this species is zoophilic in its host preference. Although the current data supporting this assertion is weak (low sample size and only two populations), this finding is consistent with a previous study (Hii, et al. 1997) which found An. bancroftii with very low human blood index even when the other species in the same village had high human blood index. Four species, namely An. farauti (s.s.), An. koliensis, An. Longirostris, and An. punctulatus (s.s.) can be considered opportunistic as they exhibited varying host selection among populations. Three of the five An. farauti (s.s.) populations exhibited a biased affinity for pig blood whereas the other two exhibited unbiased selection of the three hosts. Of the six An. koliensis populations, four 167 exhibited unbiased selection of the three hosts, one over-selected pigs and the other over-selected humans. Of the two An. longirostris populations, one over-selected humans and the other over- selected pigs. Three of the six An. punctulatus (s.s.) populations exhibited unbiased selection of the hosts and the other three exhibited biased affinity for humans, pigs or dogs. The plasticity of host selection by the four species observed here is consistent with previous studies, which found large variation in the human blood index among populations of these species (Burkot, et al. 1989; Burkot, et al. 1988a; Charlwood, et al. 1985b). A range of factors and processes influence the host selection of opportunistic mosquitoes. These include host availability, host density, physical access to hosts, differential host attractiveness, behavior of hosts in response to mosquitos’ attempts to feed, and a mosquito’s intrinsic host preference (Lyimo and Ferguson 2009; Takken and Verhulst 2013). This study emphasized relative abundance of hosts as a determining factor and found that mosquito host selection was not merely a function of relative host abundance. Rather, the strong affinity for pig or dog blood meals by many of the mosquito populations, particularly the An. farauti (s.s.), despite overwhelmingly high abundance of humans relative to the non-humans in all villages indicates that other factors impinged on it. One likely factor causing the observed host selection bias was unavailability of many human hosts due to protective effects of bed nets. The opportunistic nature of their host preference allows them to respond flexibly to host utilization under varying levels of bed net usage. This contention was supported by the logistic regression test which showed that low anthropophagy exhibited by these species was associated with higher levels of bed net usage. Although An. farauti no. 4 was found in only one village and its host selection is reported here for the first time, marked overutilization of pigs indicates that it is also 168 an opportunistic species, responding flexibly to host utilization under high bed net usage (92%) in Kokofine. The modulation of host selection by bed nets has been observed both in PNG and elsewhere. A study in Kenya showed that the majority of blood meals taken by Anopheles funestus and Culex quinquefasciatus changed from humans before permethrin-impregnated bed nets were distributed to non-humans after the bed net distribution (Bogh, et al. 1998). In a coastal village of Madang, PNG, the proportion of blood meals taken on humans by An. farauti (s.l.) dropped from 70% before permethrin-impregnated nets were distributed to 38% after the bed net distribution (Charlwood and Graves 1987). In the Wosera district of East Sepik province, PNG, increased use of insecticide-free bed nets resulted in a decline of the human blood index of Anopheles mosquitoes (Hii, et al. 2001). The use of insecticidal bed nets for controlling mosquito vectors remains the primary malaria intervention method in PNG. The reduction of malaria incidence, prevalence (Hetzel, et al. 2015; Hetzel, et al. 2017; Hetzel, et al. 2016) and transmission intensity (Reimer, et al. 2016) in PNG in recent years have been attributed to the intensification of the nationwide bed net campaign over the last 1.5 decades (Hetzel 2009; Hetzel, et al. 2014a; Hetzel, et al. 2012; Hetzel, et al. 2014b). No resistance to pyrethroids has been detected in malaria vector populations in PNG (Katusele, et al. 2014; Keven, et al. 2010; Koimbu, et al. 2018), but transmission continues to persist and very recent data indicate malaria resurgence in four of the seven study villages (chapter 3), perhaps in large part because of the opportunistic host preference of the vectors described here whose lack of dependence on human blood allows them to escape the lethal effect of the insecticides in the bed net fabric. These vector species live in sympatry and co-transmit malaria in most of the endemic areas of PNG (Burkot, et al. 1988b; Charlwood, et al. 1985a; 169 Charlwood, et al. 1985b; Charlwood and Graves 1986; Hetzel, et al. 2016; Hii, et al. 2000; Hii, et al. 1997; Hii, et al. 2001; Reimer, et al. 2016; Thomsen, et al. 2017). Therefore, while bed nets may affect the more anthropophilic and vulnerable species, transmission is still sustained by the more opportunistic and behaviorally flexible species. This condition, when combined with increased outdoor and early-evening biting observed in some vector population of PNG ((Reimer, et al. 2016; Thomsen, et al. 2017); chapter 2 and 4), presents a challenge to the bed net program in PNG as well as the rest of the South West Pacific region where these opportunistic species are the primary regional vectors. This study was not without limitations. First, although the statistical analysis of host selection was based on the total number of the three primary hosts in the village, there were variations in the relative number of these hosts among households. Therefore, the household- level variation can bias the host selection results because mosquitoes were sampled near a household throughout the night. This bias was minimized by relocating each barrier screen to a new location in the village every subsequent night to try and capture a good representation of the village (see chapter 2, section 2.3). Second, the mobility of domestic hosts, including humans, throughout the night while mosquitoes were being sampled on the barrier screen can change the actual availability of the hosts and bias the result of host selection. However, it was observed that although the animals roamed freely, dogs and pigs were found mostly close to their owner’s house during the night. Third, readers may be concerned with bias associated with indoor resting mosquitoes not sampled by the barrier screen method. The effect of this bias is minimal as members of the An. punctulatus (s.l.) group are primarily exophilic even when they feed indoors (Sinka, et al. 2011). 170 6.6. Conclusion This study investigated the host selection of different populations of six Anopheles species and delineated their intrinsic host preference based on their pattern of host selection. It was found that, with exception of An. bancroftii which was zoophilic, the other vector species were opportunistic blood feeders. The study also found that the opportunistic nature of the mosquito host preference causes them to flexibly divert blood feeding away from humans to the two non- human hosts when the level of bed net usage by the village inhabitants increases. The host selection plasticity of these opportunistic vectors can potentially limit the success of the bed net program in PNG and the Southwest Pacific region where these Anopheles mosquitoes are the primary transmitters of malaria. While malaria elimination by the bed net program is achievable from areas of the world where strict anthropophilic vectors are present, it is difficult for bed nets alone to achieve malaria elimination in PNG where most of the vectors are opportunistic. These opportunistic vectors will be resilient to bed nets and continue to transmit the disease. Thus, alternative vector control methods or strategies such as treatment of pigs and dogs with endectocides need to be developed and implemented alongside the bed net program in PNG. 171 APPENDIX 172 Appendix 6.1. Tabulated blood-meal sources of mosquitoes Table S6.1. An. bancroftii blood-meal sources. Numbers outside parenthesis are mosquito counts (n) and inside parenthesis are percent of column-wise totals. Dashes (“-”) along columns indicate villages with zero blood-meal samples for this mosquito species. Blood-meal type Host Villages Bulal n (%) Dimer n (%) Kokofine n (%) Matukar n (%) Megiar n (%) Single Mixed Human Pig Dog Human/Pig Human/Dog Pig/Dog Human/Pig/Dog - - - - - - - 1 (100.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) - - - - - - - - - - - - - - - - - - - - - 173 Mirap n (%) 5 (12.8) 30 (76.9) 3 (7.7) 0 (0.0) 1 (2.6) 0 (0.0) 0 (0.0) Wasab n (%) - - - - - - - Table S6.2. An. farauti (s.s.) blood-meal sources. Numbers outside parenthesis are mosquito counts (n) and inside parenthesis are percent of column-wise totals. Dashes (“-”) along columns indicate villages with zero blood-meal samples for this mosquito species. Blood-meal type Host Human Pig Dog Human/Pig Human/Dog Pig/Dog Human/Pig/Dog Single Mixed Villages Dimer n (%) Kokofine n (%) Matukar n (%) Megiar n (%) 2 (100.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) - - - - - - - 33 (63.5) 2 (3.9) 8 (15.4) 2 (3.9) 4 (7.7) 1 (1.9) 2 (3.9) 103 (48.1) 64 (29.9) 24 (11.2) 10 (4.7) 10 (4.7) 3 (1.4) 0 (0.0) Bulal n (%) 0 (0.0) 0 (0.0) 1 (100.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) Mirap n (%) 576 (43.2) 558 (41.9) 143 (10.7) 30 (2.3) 11 (0.8) 14 (1.1) 0 (0.0) Wasab n (%) 17 (63.0) 6 (22.2) 3 (11.1) 0 (0.0) 1 (3.7) 0 (0.0) 0 (0.0) 174 Table S6.3. An. farauti no. 4 blood-meal sources. Numbers outside parenthesis are mosquito counts (n) and inside parenthesis are percent of column-wise totals. Dashes (“-”) along columns indicate villages with zero blood-meal samples for this mosquito species. Blood-meal type Host Bulal n (%) Dimer n (%) Human Pig Dog Human/Pig Human/Dog Pig/Dog Human/Pig/Dog - - - - - - - - - - - - - - Single Mixed Villages Matukar n (%) Megiar n (%) Mirap n (%) Wasab n (%) - - - - - - - - - - - - - - - - - - - - - - - - - - - - Kokofine n (%) 220 (47.6) 188 (40.7) 33 (7.1) 10 (2.2) 8 (1.7) 3 (0.7) 0 (0.0) 175 Table S6.4. An. koliensis blood-meal sources. Numbers outside parenthesis are mosquito counts (n) and inside parenthesis are percent of column-wise totals. Dashes (“-”) along columns indicate villages with zero blood-meal samples for this mosquito species. Blood-meal type Host Human Pig Dog Human/Pig Human/Dog Pig/Dog Human/Pig/Dog Single Mixed Bulal n (%) 16 (47.1) 4 (11.8) 8 (23.5) 2 (5.9) 2 (5.9) 2 (5.9) 0 (0.00) Dimer n (%) 2 (50.0) 1 (25.0) 1 (25.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) Villages Kokofine n (%) Matukar n (%) 1 (33.3) 1 (33.3) 0 (0.0) 0 (0.0) 1 (33.3) 0 (0.0) 0 (0.0) - - - - - - - Megiar n (%) 6 (54.6) 2 (18.2) 1 (9.1) 1 (9.1) 1 (9.1) 0 (0.0) 0 (0.0) Mirap n (%) 202 (42.1) 66 (13.8) 151 (31.5) 19 (4.0) 27 (5.6) 12 (2.5) 3 (0.6) Wasab n (%) 342 (62.4) 70 (12.8) 53 (9.7) 22 (4.0) 47 (8.6) 9 (1.6) 5 (0.9) 176 Table S6.5. An. longirostris blood-meal sources. Numbers outside parenthesis are mosquito counts (n) and inside parenthesis are percent of column-wise totals. Dashes (“-”) along columns indicate villages with zero blood-meal samples for this mosquito species. Blood-meal type Host Human Pig Dog Human/Pig Human/Dog Pig/Dog Human/Pig/Dog Single Mixed Bulal n (%) - - - - - - - Dimer n (%) 3 (75.0) 0 (0.0) 1 (25.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) Villages Kokofine n (%) Matukar n (%) Megiar n (%) - - - - - - - 0 (0.0) 1 (50.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 1 (50.0) - - - - - - - Mirap n (%) 3 (15.8) 10 (52.6) 3 (15.8) 1 (5.3) 2 (10.5) 0 (0.0) 0 (0.0) Wasab n (%) 57 (67.9) 15 (17.9) 3 (3.6) 9 (10.7) 0 (0.0) 0 (0.0) 0 (0.0) 177 Table S6.6. An. punctulatus (s.s.) blood-meal sources. Numbers outside parenthesis are mosquito counts (n) and inside parenthesis are percent of column-wise totals. Dashes (“-”) along columns indicate villages with zero blood-meal samples for this mosquito species. Blood-meal type Host Human Pig Dog Human/Pig Human/Dog Pig/Dog Human/Pig/Dog Single Mixed Villages Dimer n (%) Kokofine n (%) Matukar n (%) Megiar n (%) 20 (58.8) 2 (5.9) 9 (26.5) 1 (2.9) 1 (2.9) 1 (2.9) 0 (0.0) - - - - - - - 2 (50.0) 2 (50.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) - - - - - - - Bulal n (%) 4 (50.0) 1 (12.5) 3 (37.5) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) Mirap n (%) 8 (32.0) 13 (52.0) 2 (8.0) 1 (4.0) 0 (0.0) 1 (4.0) 0 (0.0) Wasab n (%) 118 (72) 17 (10.4) 22 (13.4) 1 (0.6) 4 (2.4) 1 (0.6) 1 (0.6) 178 REFERENCES 179 REFERENCES Beebe NW, Cooper RD. 2002. Distribution and evolution of the Anopheles punctulatus group (Diptera: Culicidae) in Australia and Papua New Guinea. Int. J. Parasitol. 32:563–574. Beebe NW, Russell T, Burkot TR, Cooper RD. 2015. Anopheles punctulatus group: evolution, distribution, and control. Annu. Rev. Entomol. 60:335–350. Beebe NW, Saul A. 1995. Discrimination of all members of the Anopheles punctulatus complex by polymerase chain reaction restriction fragment length polymorphism analysis. Am. J. Trop. Med. Hyg. 53(5):478–481. Benet A, Mai A, Bockarie F, Lagog M, Zimmerman P, Alpers MP, Reeder JC, Bockarie MJ. 2004. Polymerase chain reaction diagnosis and the changing pattern of vector ecology and malaria transmission dynamics in Papua New Guinea. Am. J. Trop. Med. Hyg. 71(3):277–284. Bogh C, Pedersen EM, Mukoko DA, Ouma JH. 1998. Permethrin-impregnated bednet effects on resting and feeding behaviour of lymphatic filariasis vector mosquitoes in Kenya. Med. Vet. Entomol. 12:52–59. Burkot TR, Dye C, Graves PM. 1989. An analysis of some factors determining the sporozoite rates, human blood indexes, and biting rates of members of the Anopheles punctulatus complex in Papua New Guinea. Am. J. Trop. Med. Hyg. 40(3):229–234. Burkot TR, Graves PM, Paru R, Lagog M. 1988a. Mixed blood feeding by the malaria vectors in the Anopheles punctulatus complex (Diptera: Culicidae). J. Med. Entomol. 25(4):205– 213. Burkot TR, Graves PM, Paru R, Wirtz RA, Heywood PF. 1988b. Human malaria transmission studies in the Anopheles punctulatus complex in Papua New Guinea: sporozoits rates, inoculation rates, and sporozoite densities Am. J. Trop. Med. Hyg. 39(2):135–144. Charlwood JD, Birley MH, Dagoro H, Paru R, Holmes PR. 1985a. Assessing survival rates of Anopheles farauti (Diptera: Culicidae) from Papua New Guinea J. Anim. Ecol. 54:1003– 1016. Charlwood JD, Dagoro H, Paru R. 1985b. Blood-feeding and resting behaviour in the Anopheles punctulatus Donitz complex (Diptera: Culicidae) from coastal Papua New Guinea. Bull. Entomol. Res. 75:463–475. Charlwood JD, Graves PM. 1986. Capture-recapture studies with mosquitoes of the group of Anopheles punctulatus Donitz (Diptera: Culicidae) from Papua New Guinea. Bull. Entomol. Res. 76:211–227. 180 Charlwood JD, Graves PM. 1987. The effect of permethrin-impregnated bednets on a population of Anopheles farauti in coastal Papua New Guinea. Med. Vet. Entomol. 1:319–327. Cooper RD, Waterson DGE, Frances SP, Beebe NW, Pluess B, Sweeney AW. 2009. Malaria vectors of Papua New Guinea. Int. J. Parasitol. 39:1495–1501. Dye C. 1986. Vectorial capacity: must we measure all its components? Parasitol. Today 2(8):203–209. Garrett-Jones C. 1964. The human blood index of malaria vectors in relation to epidemiological assessment. Bull. World Health Organ. 30:241–261. Hamer GL, Chaves LF, Anderson TK, Kitron UD, Brawn JD, Ruiz MO, Loss SR, Walker ED, Goldberg TL. 2011. Fine-scale variation in vector host use and force of infection drive localized patterns of West Nile virus transmission. PLoS ONE 6(8). Hamer GL, Kitron UD, Goldberg TL, Brawn JD, Loss SR, Ruiz MO, Hayes DB, Walker ED. 2009. Host selection by Culex pipiens mosquitoes and West Nile virus amplification. Am. J. Trop. Med. Hyg. 80(2):268–278. Hess AD, Hayes RO, Tempelis CH. 1968. The use of the forage ratio technique in mosquito host preference studies. Mosq. News 28(3):386–389. Hetzel MW. 2009. An integrated approach to malaria control in Papua New Guinea. PNG Med. J. 52:1–7. Hetzel MW, Choudhury AAK, Pulford J, Ura Y, Whittaker M, Siba PM, Mueller I. 2014a. Progress in mosquito net coverage in Papua New Guinea. Malar. J. 13:242. Hetzel MW, Gideon G, Lote N, Makita L, Siba PM, Mueller I. 2012. Ownership and usage of mosquito nets after four years of large-scale free distribution in Papua New Guinea. Malar. J. 11:192. Hetzel MW, Morris H, Tarongka N, Barnadas Ce, Pulford J, Makita L, Siba PM, Mueller I. 2015. Prevalence of malaria across Papua New Guinea after initial roll-out of insecticide- treated mosquito nets. Trop. Med. Int. Health 20(12):1745–1755. Hetzel MW, Pulford J, Maraga S, Barnadas C, Reimer LJ, Tavul L, Jamea-Maiasa S, Tandrapah T, Maalsen A, Makita L and others. 2014b. Evaluation of the global fund-supported national malaria control program in Papua New Guinea, 2009-2014. PNG Med J. 57:1–4. Hetzel MW, Pulford J, Ura Y, Jamea-Maiasa S, Tandrapah A, Tarongka N, Lorry L, Robinson LJ, Lilley K, Makita L and others. 2017. Insecticide-treated nets and malaria prevalence, Papua New Guinea, 2008–2014. Bull. World Health Organ. 95:695–705. 181 Hetzel MW, Reimer LJ, Gideon G, Koimbu G, Barnadas C, Makita L, Siba PM, Mueller I. 2016. Changes in malaria burden and transmission in sentinel sites after the roll-out of long- lasting insecticidal nets in Papua New Guinea. Parasit. Vectors 9:340. Hii JLK, Smith T, Mai A, Ibam E, Alpers MP. 2000. Comparision between anopheline mosquitoes (Diptera: Culicidae) caught using different methods in a malaria endemic areas of Papua New Guinea. Bull. Entomol. Res. 90:211–219. Hii JLK, Smith T, Mai A, Mellord S, Lewis D, Alexander N, Alpers MP. 1997. Spatial and temporal variation in abundance of Anopheles (Diptera: Culicidae) in a malaria endemic area in Papua New Guinea. J. Med. Entomol. 34(2):193–205. Hii JLK, Smith T, Vounatsou P, Alexander N, Mai A, Ibam E, Alpers MP. 2001. Area effects of bednet use in a malaria-endemic area in Papua New Guinea. Trans. R. Soc. Trop. Med. Hyg. 95:7–13. Katusele M, Gideon G, Thomsen EK, Siba PM, Hetzel M, Reimer LJ. 2014. Longlasting insecticidal nets remain efficacious after five years of use in Papua New Guinea. PNG Med. J. 57:86–93. Kay BH, Boreham PFL, Edman JD. 1979. Application of the feeding index concept to studies of mosquito host-feeding patterns. Mosq. News 39:68–72. Keven JB, Henry-Halldin CN, Thomsen EK, Mueller I, Siba PM, Zimmerman PA, Reimer LJ. 2010. Pyrethroid susceptibility in natural populations of the Anopheles punctulatus group (Diptera: Culicidae) in Papua New Guinea. Am. J. Trop. Med. Hyg. 83(6):1259–1261. Koimbu G, Czeher C, Katusele M, Sakur M, Kilepak L, Tandrapah A, Hetzel MW, Pulford J, Robinson L, Karl S. 2018. Status of insecticide resistance in Papua New Guinea: an update from nation-wide monitoring of Anopheles mosquitoes. Am. J. Trop. Med. Hyg. 98(1):162–165. Logue K, Keven JB, Cannon MV, Reimer L, Siba P, Walker ED, Zimmerman PA, Serre D. 2016. Unbiased characterization of Anopheles mosquito blood meals by targeted high- throughput sequencing. PLoS Negl. Trop. Dis. 10(3). Lyimo IN, Ferguson HM. 2009. Ecological and evolutionary determinants of host species choice in mosquito vectors. Trends Parasitol. 25(4):189–196. Molaei G, Andreadis TA, Armstrong PM, Anderson JF, Vossbrinck CR. 2006. Host feeding patterns of Culex mosquitoes and West Nile Virus transmission, northeastern United States. Emerg. Infect. Dis. 12:468–474. Muller I, Bockarie M, Alpers M, Smith T. 2003. The epidemiology of malaria in Papua New Guinea. Trends Parasitol. 19(6):253–259. 182 Reimer LJ, Thomsen EK, Koimbu G, Keven JB, Mueller I, Siba PM, Kazura JW, Hetzel MW, Zimmerman PA. 2016. Malaria transmission dynamics surrounding the first nationwide long-lasting insecticidal net distribution in Papua New Guinea. Malar. J. 15:25. Russell TL, Beebe NW, Bugoro H, Apairamo A, Chow WK, Cooper RD, Collins FH, Lobo NF, Burkot TR. 2016. Frequent blood feeding enables insecticide‑treated nets to reduce transmission by mosquitoes that bite predominately outdoors. Malar. J. 15(156): Sinka ME, Bangs MJ, Manguin S, Chareonviriyaphap T, Patil AP, Temperley WH, Gething PW, Elyazar IR, Kabaria CW, Harbach RE and others. 2011. The dominant Anopheles vectors of human malaria in the Asia-Pacific region: occurrence data, distribution maps and bionomic précis. Parasit. Vectors 4:89. Takken W, Verhulst NO. 2013. Host preferences of blood-feeding mosquitoes. Annu. Rev. Entomol. 58:433–453. Thomsen EK, Koimbu G, Pulford J, Jamea-Maiasa S, Ura Y, Keven JB, Siba PM, Mueller I, Hetzel MW, Reimer LJ. 2017. Mosquito behaviour change after distribution of bednets results in decreased protection against malaria exposure. J. Infect. Dis. 215(5):791–797. 183 CHAPTER 7: NON-RANDOM HOST SELECTION AND MIXED HUMAN BLOOD MEALS BY ANOPHELES VECTORS OF MALARIA IN MADANG, PAPUA NEW GUINEA: IMPLICATIONS FOR MALARIA TRANSMISSION 7.1. Abstract Non-random distribution of blood meals and multiple blood meals taken from humans by Anopheles vectors of malaria can multiply the force of transmission of malaria and allow the disease to persist even when interventions are in place. This study investigated these phenomena in Madang villages, using barrier screen sampling of blood fed mosquitoes, and human genotyping of blood meals and villagers using microsatellite loci. Blood meals with three or more alleles in multiple microsatellite loci signaled cryptic meals from more than one person. DNA profiles of mosquito collectors and village residents were analyzed and checked for matches with the blood-meal profiles. The distribution of blood meals from individual humans (collector-fed blood meals were excluded) fitted the negative binomial better than Poisson model, indicating non-random utilization of individual human hosts. Males and individuals 15– 30 years old were statistically overrepresented in blood meals relative to females and younger or older age groups. However, the risk of malaria infection was generally similar across sexes and age groups. The proportion of cryptic mixed meals ranged from 5.5–15%. The probability of an interrupted feeding on a human leading to a second feeding on a different human ranged from 0.03–0.18 for six Anopheles populations across four villages. Because non-random human selection and cryptic mixed blood-feeding behaviors can multiply the basic reproduction number of malaria and vectorial capacity of vectors, malaria is likely to persist despite ongoing efforts to 184 suppress it by insecticidal bed net intervention programs. The biased feeding on young males emphasizes the importance of targeted rather than population-wide disease intervention. 7.2. Introduction Analysis of mosquito blood meals using molecular methods to identify vertebrate host species fed on by mosquitoes is widely used and has enabled (i) evaluation of important aspects of mosquito-host interactions such as host selection that affect vector-borne disease ecology and (ii) assessment of effectiveness of vector control measures such as insecticide-treated bed nets (Fritz, et al. 2013; Gunathilaka, et al. 2016; Hamer, et al. 2011; Hamer, et al. 2009; Hurk, et al. 2007; Kent and Norris 2005; Keven, et al. 2017; Logue, et al. 2016; Meece, et al. 2005; Molaei, et al. 2006; Molaei, et al. 2010). In recent years, analysis of blood meals has been extended to include multilocus genotyping of microsatellite (aka short tandem repeat) markers that serve as unique DNA profiles (aka DNA fingerprints) to identify or differentiate individuals of a species, particularly humans in blood-meal samples (Benedictis, et al. 2003; Chow-Shaffer, et al. 2000; Harrington, et al. 2014; Keven, et al. 2019b; Michael, et al. 2001; Scott, et al. 2006; Soremekun, et al. 2004). Identification of human individuals in mosquito blood meals has three important applications in studies of mosquito-borne disease ecology. First, if mosquito biting rates are heterogeneous so that some individuals in a community are bitten more frequently than others, these individuals are more likely to be infected and serve as reservoirs or even sinks of mosquito- borne pathogens. Thus, identification of such individuals or their characteristics can inform public health officials, particularly those in resource-limited countries, to conduct targeted rather than population-wide disease intervention. For example, distributing interventions (antimalarial 185 prophylactic drugs, insecticide treated bed nets) to those segments of the human population who are bitten in excess could reduce prevalence of infection in parasite reservoirs and thereby diminish rates of transmission, simultaneously increasing effectiveness of the intervention and reducing costs (Bousema, et al. 2010; Bousema, et al. 2012; Greenwood 2010; Nofal, et al. 2019; Noor, et al. 2010). Second, assessing patterns of mosquito biting frequencies among individuals is important to prognose the course of disease epidemiology, particularly in the context of a disease control program. Malaria prognosis is usually based on two epidemiological quantities: (i) basic reproduction number (R0), which is the number of new human infection cases in a susceptible population that arises as a result of transmission from a single primary case throughout its infective life and (ii) the vectorial capacity (V), which is the expected number of infective bites that arises from all the Anopheles that bite a single infectious person in a single day (Smith, et al. 2012). Indeed, V is a submodel of R0. Some studies have shown that the values of both R0 and V under non-random conditions of vector-human contact (whether socially or spatially) are higher than under random conditions (Dye and Hasibeder 1986; Hasibeder and Dye 1988). As malaria transmission increases with R0 and V, such non-random biting patterns may permit transmission to persist with potential for resurgence, even when a vector control program is underway (Dye and Hasibeder 1986; Hasibeder and Dye 1988). Third, molecular methods that can distinguish between different individuals of a host species can be used to estimate the proportion of blood-fed mosquitoes that had obtained a blood meal from two or more individuals of that species, hereafter referred to as “cryptic mixed meals”. The term was used to distinguish this blood-meal type from those that obtained a blood meal from two or more individuals of different host species or “patent mixed meals” (Boreham 186 and Garrett-Jones 1973; Burkot, et al. 1988). Importantly, cryptic mixed meals occur within a single gonotrophic cycle, meaning there were multiple contacts between an individual female mosquito and humans rather than assuming (as is the case with transmission models) that there is only one contact per gonotrophic cycle. Cryptic mixed meals have been shown to increase both R0 and V; an effect that has been termed a “force multiplier” (Tedrow, et al. 2019). In this study, DNA profiles of humans in mosquito blood meals were analyzed and the data were used to quantify the proportion of cryptic mixed blood meals, estimate the probability of a human feed being interrupted, and test for non-random vector-human contacts in Anopheles vectors of malaria in PNG. 7.3. Methods 7.3.1. Mosquito and human blood samples This study involved analysis of blood-fed female Anopheles mosquitoes from Bulal, Megiar, Mirap and Wasab villages. Geographic location (see Figure 2.1) and ecogeographic description of these villages are described in chapter 2. Mosquitoes were collected for six nights in Bulal (16th–21st March 2017), Megiar (4th–9th February 2017) and Mirap (11th–16th January 2017) and 12 nights in Wasab (5th–10th September and 4th–9th November 2016) using the barrier screen method and sampling procedure described in chapter 2. The mosquito species and their blood meal hosts were identified by PCR methods described in chapter 6. Census counts of male and female humans, and blood sample and age of consented individuals (including mosquito collectors) in Mirap and Wasab were obtained as described in chapter 3. Finger-pricked human blood samples (250 µl) were obtained from consented residents (ages 0.5–85 years; consent for minors obtained from parents or guardians) and the mosquito collectors working in the two 187 villages. The human blood samples were tested for P. falciparum and P. vivax infection using the qPCR method described in chapter 3. 7.3.2. Human genotyping Human genotypes (“DNA profiles”) of the village residents (including the mosquito collectors) of Mirap and Wasab and the Anopheles blood meals of all four villages were analyzed using a microsatellite genotyping assay used in forensic identification (Oostdik, et al. 2014). The assay utilized a multiplex PCR kit (PowerPlex Fusion System; catalog number, DC2402; Promega, Madison, WI, USA) which contained fluorophore-labelled primer pairs that co-amplified 22 autosomal microsatellite loci (D3S1358, D1S1656, D2S441, D10S1248, D13S317, Penta E, D16S539, D18S51, D2S1338, CSF1PO, Penta D, TH01, vWA, D21S11, D7S820, D5S818, TPOX, D8S1179, D12S391, D19S433, FGA, D22S1045), one Y-linked microsatellite locus (DYS391), and one locus serving as a sex-specific marker (Amel) (Oostdik, et al. 2014). The PCR reaction mixture (10 µl reaction volume) consisted of 6 µl of water, 1 µl of 5x Fusion Master Mix, 1 µl of 5x Fusion Primer Mix, 0.5 µl of 10x PCR Buffer (product number, N8080010; Thermo Fisher Scientific, Waltham, MA, USA), 0.5 µl of 25 mM MgCl2 (product number, N8080010; Thermo Fisher Scientific) and 1 µl of template DNA. Thermal cycling consisted of 1 cycle of 96 °C for 1 minute followed by 30 cycles of 94 °C for 10 seconds, 59 °C for 1 minute, 72 °C for 30 seconds, and 1 cycle of 60 °C for 10 minutes. For detection of amplicon size, solution mixtures consisting of 1 µl of the PCR products or Allelic Ladder (product number, DG381B; Promega) combined with 0.3 µl of WEN Internal Lane Standard 500 (product number, DG5001; Promega) and 9 µl of Hi-Di Formamide (catalog number, 4311320; Thermo Fisher Scientific) were analyzed by capillary electrophoresis (3500 188 Genetic Analyzer; Applied Biosystems). Results were analyzed using GeneMapper software version 4.1 (Applied Biosystems). Microsatellite alleles were represented by the number of sequence repeats, which can be heterozygous or homozygous. The sex-specific marker indicated a female individual if it was homozygous (X, X) or a male if heterozygous (X, Y). For each sample, the genotype of each locus was listed in a Microsoft Excel spreadsheet. The alleles of each marker were separated by a comma; for heterozygous genotypes, the alleles were listed in order from small to large. Blood- meal samples with more than two alleles at one or more loci were considered cryptic mixed meals from different human hosts. 7.3.3. Data analyses Procedure for identifying unique DNA profiles. A program was written in R that engaged loop functions and conditional statements to compare the genotype of each locus in a query sequence (i.e., a DNA profile) to its corresponding locus in a subject sequence (i.e., another DNA profile). The similarity of two DNA profiles was quantified based on their percent profile match, which is the proportion of matched loci times 100 (e.g., 18 matched loci/24 total loci x 100 = 75% profile match). In principle, DNA profiles from two blood-meal specimens that originated from a same human source will have 100% profile match, thus, match values < 100% indicate blood meals from different individuals. However, as some blood-meal specimens yield suboptimum human DNA (due to a tiny amount of engorged blood, for example), false mismatches due to allele dropouts can arise. To account for this, a value less than 100% was used as the criterion for determining a match. To determine this value, pairwise percent profile match analysis was performed on the DNA profiles of village residents. This generated n(n-1)/2 profile match results 189 or values, where n is the number of village residents. The highest value in the pairwise match output (excluding match results of monozygotic twins, which was 100%) was the criterion below which two DNA profiles were considered different. Derivation of formulae for estimating probability of a human feed being interrupted. A blood-feeding mosquito that is interrupted by its human host before obtaining a full blood meal can complete it by feeding on another human (cryptic mixed meal) or a different host species (patent mixed meal) (Burkot, et al. 1988). The probability (I) of an interrupted feed resulting in a second meal, either cryptic or patent, can be estimated from the equation != #$% ∗ () * $+ #$% ∗ () * $+ * $, , where Hp is the number of patent mixed human meals, B is the human blood index (proportion of all single blood meals that are humans), Hc is the number of cryptic mixed human meals and Hs is the number of single human meals. Hp consists of two groups of mosquitoes: those that feed on non-humans after being interrupted during feeding on a human (i.e., human to non-human), and those that feed on a human after being interrupted during feeding on a non-human (non-human to human). Because the interest is in the probability of a human feed (and not a non-human feed) being interrupted, the latter group (non-human to human) overestimates I. This is corrected by multiplying Hp with B in the equation, which eliminates the portion of Hp that consists of non-human to human patent mixed meals. The quantity I represents the probability of a human feed being interrupted that led to a second meal on another host of any species; not just humans. To estimate the probability (IH) of a second meal on humans only, the patent mixed meal component in the numerator is eliminated to derive the equation !$= $+ #$% ∗ () * $+ * $,. Test of non-random human selection. To eliminate collector-feeding bias in the analysis of non-random human selection by mosquitoes, blood-meal samples whose DNA profile 190 matched that of the people who collected them were removed from the data. Cryptic mixed meals were not useful for this analysis and were also removed from the data. The number of different human individuals in the blood-meal sample was determined from the number of unique DNA profiles in the sample. The number of occurrences of each unique profile in the blood-meal sample represented the number of mosquitoes that fed on each human individual. These data were used to construct frequency distribution histograms which relate the number of different human individuals (y-axis) to the number of blood-meal occurrences (x-axis). The observed frequency distributions were fitted to zero-truncated Poisson and zero-truncated negative binomial frequency distribution models using the functions zerotrunc and rootogram of the package countreg (Kleiber and Zeileis 2016) in R software. Statistically supported fit to the Poisson model indicates random selection of human individuals by the mosquitoes, presumably reflecting no bias, whereas fit to the negative binomial indicates non-random selection, presumably reflecting bias. The observed frequency of blood meals taken on human individuals of a specific sex or age group was compared with their expected frequency (obtained from village census data) by the two-tailed exact binomial test. The likelihood of being infected with P. falciparum and P. vivax in human individuals of different age groups and sexes was tested by logistic regression of binary data. The two tests were performed using the functions binom.test and glm of the R package stats. Significance level (all tests) was based on type I error rate of 5%. 7.4. Results 7.4.1. Percent match values of human profiles DNA profiles of 427 villagers of Mirap (205 males, 222 females) and 164 villagers of Wasab (90 males, 74 females), all consented residents, were successfully analyzed. Pairwise percent match 191 analysis of these DNA profiles resulted in 90,951 and 13,366 pairwise percent match values for Mirap and Wasab, respectively. For both villages, these pairwise results consisted of 18 unique percent match values, each with a varying number of occurrences. The distribution of the probability (i.e., proportion) of occurrence of the unique values is shown in Figure S7.1 (Appendix 7.1). The lowest percent match value was 0% with 0.05 probability of occurrence, and the highest value was 71% (i.e., 17/24 matched loci) with 10-4 probability of occurrence. Although 71% was the maximum percent match value, 79% (which allows for five false mismatched loci) whose probability of occurrence was zero was chosen as the criterion below which two DNA profiles were considered as belonging to different people. 7.4.2. Cryptic mixed meals and probability of interrupted feeds Of 881 human-fed Anopheles blood meals that were genotyped, 85% (n = 753) yielded a DNA profile; the other 15% (n = 128) either did not PCR-amplify or had grossly incomplete profiles. Of those blood-meal samples that yielded a DNA profile, 18 were from Bulal, 124 from Megiar, 233 from Mirap and 378 from Wasab. The distribution of these profiles according to mosquito species, namely An. bancroftii, An. farauti (s.s.), An. koliensis, An. longirostris and An. punctulatus (s.s.), is presented in Table 7.1. The proportion of cryptic mixed meals for Anopheles spp. in each village and for three populations with sample size > 100 is shown in Figure 7.1. The numbers of cryptic mixed (Hc), patent mixed (Hp) and single (Hs) human meals for six Anopheles populations across the four villages are shown in Table 7.2; other populations (see Table 7.1) were excluded due to small sample size (n < 10). These three quantities were then used together with B (human blood index) to quantify I (probability of an interrupted blood meal on a human leading to a blood meal on another host of any species) and IH (probability of an interrupted 192 blood meal on a human leading to a second blood meal on another human) for each of the six populations (Table 7.2). The values of B in Table 7.2 were computed from blood-meal data in Table S7.1 (Appendix 7.2). The estimated values of I ranged from 0.09–0.30, and of IH ranged from 0.03-0.18 (Table 7.2). As expected, the mean of I (0.2 ± 0.08) was statistically greater than that of IH (0.1 ± 0.06) (t-test: t(9) = 2.28, P = 0.047). Bulal Wasab Table 7.1. Distribution of successfully genotyped blood meals according to Anopheles species. Anopheles species An. bancroftii An. farauti (s.s.) An. koliensis An. longirostris An. punctulatus (s.s.) 0 116 8 0 0 0 3 339 0 36 1 31 200 1 0 0 0 15 0 3 Megiar Mirap An. koliensis Table 7.2. Estimates of blood host choice parameters for six Anopheles populations. Year N Village Anopheles species IH 2017 15 Bulal 0.15 2017 116 Megiar An. farauti (s.s.) 0.06 2017 31 An. farauti (s.s.) Mirap 0.18 2017 200 Mirap 0.06 An. koliensis Wasab An. koliensis 2016 339 0.16 Wasab An. punctulatus (s.s.) 2016 36 0.03 Abbreviations: N, number of successfully genotyped human blood meals; Hs, number of single human meals; Hp, number of patent mixed human meals; Hc, number of cryptic mixed human meals; B, human blood index; I, probability of an interrupted blood meal on a human leading to blood meal on another host of any species; IH, probability of an interrupted blood meal on a human leading to blood meal on another human. B 0.57 0.54 0.29 0.47 0.72 0.78 I 0.27 0.14 0.23 0.18 0.30 0.09 Hs 10 93 21 145 227 32 Hp 3 16 5 44 60 3 Hc 2 7 5 11 52 1 193 Figure 7.1. Percentage of cryptic mixed human meals. (a) For Anopheles spp. in Bulal (n = 18), Megiar (n = 124), Mirap (n = 233) and Wasab (n = 378). (b) For three populations: An. farauti (s.s.) in Megiar (n = 116), An. koliensis in Mirap (n = 200) and An. koliensis in Wasab (n = 339). Numbers at the tip of the bar are the estimated percentage. 7.4.3. Non-random human host selection Statistical tests of non-random human selection were performed on mosquitoes from Mirap and Wasab only as it was possible to identify and eliminate the collector-fed blood meals in these samples. Also, the test was performed on Anopheles in general; analysis according to species entailed working with low sample size. Of 233 successfully genotyped Mirap blood-meal samples, 94% (n = 218) were single-human meals (i.e., not cryptic mixed meals). Of 378 successfully genotyped Wasab blood-meal samples, 87% (n = 328) were single-human meals. Of the 218 Mirap single-human blood-meal profiles, 29% (n = 63) matched the collector’s profile. 194 Of the 328 Wasab single-human blood-meal profiles, 46% (n = 152) matched the collector’s profile. After excluding the collector-fed blood-meals from the data, 88 different human individuals were identified in the Mirap blood meals (n = 155) and 79 individuals were found in the Wasab blood meals (n = 176). For each village, the number of occurrence of each human individual in the blood-meal sample was used to construct a plot which showed the frequency of occurrence (i.e., number of blood meals taken on an individual) on the x-axis and the number of individuals having a particular frequency of occurrence on the y-axis (Figure S7.2, Appendix 7.3). Fits of the frequency distribution data of both Mirap and Wasab to zero-truncated negative binomial and zero-truncated Poisson distribution models are shown in Figure 7.2. Data fit well to the negative binomial, with nonsignificant differences between observed and expected frequency distributions, and poorly to the Poisson model with significant differences between the observed and expected frequency distributions. Of the 88 different individuals identified in Mirap blood meals, most (70%) had one blood meal in the sample, 24% of people had 2–4 blood meals, and 6% had more than 4 blood meals. There were two individuals with 10 blood meals in the sample from Mirap. Of the 79 different individuals in Wasab blood meals, 68% had one blood meal, 24% of people had 2–4 blood meals, and 8% had more than 4 blood meals. There were 23 blood meals from the same person in the sample from Wasab. 195 Figure 7.2. Chi-square test of fit of zero-truncated Poisson and zero-truncated negative binomial expected distribution (red curves) to observed distribution (vertical bars) of the frequency of occurrence of unique blood-meal profiles in mosquito samples of Mirap (top row) and Wasab (bottom row) villages. The number of Anopheles spp. that fed on a female or male individual and the result of statistical analysis (exact binomial test) comparing the proportion of female-fed and male-fed blood meals to their expected proportion based on census data are shown in Table 7.3. Despite near equal proportion of both sexes in both Mirap (51% females, 49% males) and Wasab (49% females, 51% males), there were significantly more male-fed (Wasab, 69%; Mirap, 72%) and less female-fed (Wasab, 31%; Mirap, 28%) mosquitoes than expected. Twelve percent (19/155) 196 of Mirap and 24% (43/176) of Wasab Anopheles spp. blood-meal profiles were successfully matched to a village resident of known age (in years). The number of the blood-meal profiles that matched an individual from each of five age groups and the results of statistical analysis (exact binomial test) comparing the proportion of blood meals in each age group to their expected proportion are shown in Table 7.4. In both villages, the proportion of blood meals in the three older age groups did not deviate from the expected proportion whereas significant deviation was observed in the two younger age groups. Blood meals obtained from individuals < 15 years old were underrepresented whereas those obtained from 15–30 years old were overrepresented in both villages (Table 7.4). Prevalence of infection for P. falciparum and P. vivax in villagers residing in Mirap and Wasab is shown in Figure 7.3. Result of logistic regression tests which compared the likelihood of being infected with P. falciparum or P. vivax in male individuals relative to females (reference group), and for individuals in four age groups relative to those in the < 15 years group (reference group) are shown for Mirap (n = 440 study participants) in Table 7.5 and for Wasab (n = 285 participants) in Table 7.6. In Mirap, the likelihood of infection with P. falciparum was statistically higher in the age groups 15–30 years and 31–45 years compared to the < 15 years age group; no difference was observed for the two older age groups or between the sexes (Table 7.5). For P. vivax in Mirap (Table 7.5) and both malaria species in Wasab (Table 7.6), no statistical variation was observed among the age groups or between sexes. Only the main effects were evaluated, as the interaction between age and sex was not significant in either case. 197 Table 7.3. Results of exact binomial test comparing the observed proportion of each sex in the blood meals with their expected proportion based on village census count. Village Sex Blood-meal count Expected Proportion Blood-meal proportion (95% CI) P value Mirap Females Males Wasab Females Males 44 111 54 122 0.51 0.49 0.49 0.51 0.28 (0.21–0.36) 0.72 (0.64–0.79) 0.31 (0.24–0.38) 0.69 (0.62–0.76) < 0.0001* < 0.0001* < 0.0001* < 0.0001* Note: Total number of people in census data were 671 in Mirap and 292 in Wasab. Asterisks (*) indicate significant P value. Table 7.4. Results of exact binomial test comparing the observed proportion of each age group in the blood meals with their expected proportion based on village census count. Village Age group Blood-meal count Expected proportion Blood-meal proportion (95% CI) P value Mirap Wasab < 15 years 3 15–30 years 11 31–45 years 2 46–60 years 2 1 > 60 years < 15 years 11 15–30 years 19 31–45 years 9 46–60 years 3 1 > 60 years 0.48 0.23 0.19 0.07 0.03 0.48 0.25 0.16 0.07 0.02 0.16 (0.03–0.40) 0.58 (0.33–0.80) 0.11 (0.01–0.33) 0.11 (0.01–0.33) 0.05 (0.001–0.26) 0.26 (0.14–0.41) 0.44 (0.29–0.60) 0.21 (0.10–0.36) 0.07 (0.01–0.19) 0.02 (0.001–0.12) 0.005* 0.001* 0.56 0.39 0.44 0.003* 0.007* 0.40 1.0 1.0 Note: Total number of people in census data were 671 in Mirap and 292 in Wasab. Asterisks (*) indicate significant P value. 198 Figure 7.3. Prevalence of infection for P. falciparum and P. vivax in residents of Mirap and Wasab villages. (a) Prevalence of infection by age group. (b) Prevalence of infection by sex. 199 Table 7.5. Logistic regression results for test of variation in likelihood of malaria infection among age and sex groups in Mirap. Reference variable P. falciparum P. vivax Variable 15–30 years 31–45 years 46–60 years > 60 years Male 15–30 years 31–45 years 46–60 years > 60 years Male b 1.32 1.15 1.04 -14.17 0.28 -0.53 -0.16 -1.18 -0.55 -0.15 Std. Error 0.37 0.39 0.55 650.87 0.32 0.41 0.41 0.78 1.14 0.29 P value 0.0003* 0.0033* 0.0563 0.9826 0.3883 0.179 0.686 0.131 0.63 0.618 Note: Each of the age group was compared with < 15 years as the reference group and male was compared with female as the reference group. Asterisks (*) indicate significant P value. 200 Table 7.6. Logistic regression results for test of variation in likelihood of malaria infection among age and sex groups in Wasab. Reference variable P. falciparum b 1.67 x 101 1.64 x 101 -5.58 x 10-11 -5.62 x 10-11 1.73 x 101 Variable 15–30 years 31–45 years 46–60 years > 60 years Male Std. Error 1.36 x 103 1.36 x 103 3.51 x 103 6.36 x 103 1.36 x 103 P value 0.99 0.99 1.00 1.00 0.99 P. vivax 15–30 years 31–45 years 46–60 years > 60 years Male -0.59 -0.22 -15.57 -15.57 0.34 1.17 1.18 2.0 x 103 3.8 x 103 0.76 0.617 0.85 0.994 0.997 0.648 Note: Each of the age group was compared with < 15 years as the reference group and male was compared with female as the reference group. 7.5. Discussion Sources of error in matching DNA profiles is an important consideration in a study of this kind. Here, the two primary sources of error are the degree of relatedness of villagers and the allele dropout rate per locus. If villagers are closely related to each other genetically, then the ability to discriminate among blood meals taken from different villagers is reduced. If too many alleles drop out of the analysis owing to insufficient DNA or inconsistency with priming during PCR, then false homozygosity would impede matching of DNA profiling. Of the two, the former is likely to be more important to error rate as the latter still yields a profile of some value. Villagers of New Ireland, New Britain and Bougainville provinces, PNG, (similar to most villages elsewhere in the country, such as Madang Province) have low genetic diversity and are likely 201 genetically related to many other members of the village through nuclear and extended family connections (Friedlaender, et al. 2008). This means that for most people, the percent profile match with many members of the village will be high and thus require a greater number of microsatellite markers to successfully differentiate them. In this study, the pairwise percent profile match based on 24 loci found that the highest percent match between any two non- monozygotic individuals was 71%, but the probability of observing this value was 10-4 or 1 in 10, 000 pairwise trials, which is extremely low. This means that a minimum of 17 markers (16 microsatellites and one sex-determining marker), which is similar to the number of markers used in other forensic identification protocols (Beleza, et al. 2004; Gorostiza, et al. 2007; Kim, et al. 2003), is sufficient to discriminate between two individuals from a typical PNG village. By choosing 79% (with zero probability of occurrence) instead of 71% as the criterion for deciding a match (match values ³ 79%) or a mismatch (values < 79%), this decision rule allows for at most five erroneous mismatched loci of the 24 loci examined here, while maintaining the power to correctly discriminate between different individuals because of the remaining matched loci. Other studies (Benedictis, et al. 2003; Chow-Shaffer, et al. 2000; Harrington, et al. 2014; Michael, et al. 2001; Scott, et al. 2006; Soremekun, et al. 2004) that have conducted DNA profiling and matching of blood meals to human individuals have not considered this important element of the identification process and accordingly have not established this quite necessary rule of thumb, nor accurately incorporated these potential sources of error into their analysis. The frequency of Anopheles blood meals taken on individual humans in the villages was not randomly distributed but rather was clumped or clustered. One likely explanation is that the Anopheles catches in households in these same villages was not randomly distributed but rather aggregated (see chapter 4), therefore, Anopheles females are most likely to encounter humans 202 where the mosquitoes themselves aggregate. In the malaria transmission system, those few human individuals who contribute most of the blood meals could be recognized as “super- spreaders”, as they are likely to be infected and serve as reservoirs of the parasite to be spread to other members of the community (Bousema, et al. 2012; Woolhouse, et al. 1997). Non-random vector-human contact when measured by blood-meal genotyping has been observed in several other Anopheles and non-Anopheles populations and their associated mosquito-borne disease systems (Benedictis, et al. 2003; Harrington, et al. 2014; Michael, et al. 2001; Woolhouse, et al. 1997). In other villages near to the ones in the current study, people over the age of 20 who were infected with either P. vivax or P. falciparum malaria commonly harbored gametocytes (P. vivax, 40%; P. falciparum, 50%) but at lower densities of gametocytemia (P. vivax geometric mean, 2 gametocytes/ul; P. falciparum geometric mean, 0.5 gametocytes/ul) than younger age groups (Koepfli, et al. 2015). From these findings, it can be inferred that the older age groups found here to be fed upon at a greater than random encounter rate can be viewed as potential sources of malaria infection to mosquitoes rather than as sinks. Experimental feedings and other means of investigation could reveal this possibility. Certain demographic groups were overrepresented or underrepresented in mosquito host choice. In both villages, the Anopheles obtained more blood meals from male individuals of the 15–30 years old age group compared to females or other age groups. A striking finding from the blood meals was under-representation of the youngest age group (< 15 years), which constituted a large proportion of the village censuses (52% of Mirap, 50% of Wasab). Human behavior and bed net usage could explain these patterns. From observation, in typical rural villages of PNG, males are shirtless most of the time (a response to high atmospheric temperature and humidity all year round) compared to their female counterparts. Thus, males have larger exposed body 203 surface for mosquitoes to land. Adult males (> 15 years) are less likely to sleep under a bed net than are females (Hetzel, et al. 2012). Also, from observation, youth greater than 15 years of age, particularly males, often remain active into the late night and retire to bed later in the morning hours compared to other age groups and to females. Thus, male individuals in the 15–30 years old age group would have greater exposure to Anopheles bites than those of the other age groups who retire under the bed nets earlier in the night. It is also worth noting that the guidelines for the bed net campaign prioritized distribution of nets to children and pregnant women (Hetzel, et al. 2014a; Hetzel, et al. 2014b), which could also have contributed to the observed demographic representations in the mosquito blood meals here. Similar findings to those here have been reported for populations of Anopheles funestus and Anopheles gambiae in villages of Tanzania. In that setting, males and individuals ³ 20 years received more Anopheles bites than expected whereas females and those < 20 years old were bitten less frequently. In a village without bed nets, such variations were not observed (Soremekun, et al. 2004). In a Kenyan village, individuals < 20 years old received more bites than expected compared to those between 20–50 years old, but no variation was observed between sexes (Scott, et al. 2006). In five Zambian villages, the proportion of male-fed and female-fed blood meals were homogeneous with the expected proportions; data on age groups were not reported (Das, et al. 2017). Given the heterogeneous utilization of humans by the vectors, it was predicted that the likelihood of malaria infection would be higher in male than female individuals and in individuals in the youth age (15–30 years) than the other age groups. However, this prediction held true only for the age group variable in Mirap P. falciparum infections; it was not the case for either parasite species in Wasab, nor for P. vivax in Mirap. Interestingly, malaria infection for both species was not different between sexes, which would not follow from observed biases in 204 Anopheles feeding. A plausible explanation for these inconsistencies in predictions is that the human biting rates (number of mosquito bites per person per unit time) in the underutilized human groups (i.e., females and non-youth age groups) were still high enough to support transmission across these demographic groups such that the probability of infection was homogeneous in all ages and sexes despite biased mosquito feeding patterns. This assertion is supported indirectly by a study which showed that the rate of malaria transmission, measured in terms of entomological inoculation rate (the number of malaria infective Anopheles bites received per person per year), must be reduced below 1.0 before appreciable reduction in prevalence of infection in the human population can be observed (Smith, et al. 2005). The proportion of cryptic mixed human meals observed in Anopheles spp. in the four villages varied from 6–13% and in the three Anopheles populations with sufficient sample size varied from 5.5–15%. A similar study conducted in another Madang village in which the proportion of mosquitoes that fed on mother-child pair was quantified using ABO blood group markers showed that 13% of Anopheles spp., 11% of An. punctulatus (s.s.), 13% of An. koliensis and 18% of An. farauti (s.s.) had mixed meals (Burkot, et al. 1988). These results were in the same range as the ones produced in this current study. The proportion of cryptic mixed blood meals in PNG Anopheles observed in both the current and previous studies, were similar to those reported for Anopheles funestus (2–14%) and Anopheles gambiae (0–11%) in two Kenyan villages (Scott, et al. 2006), and An. gambiae (10%) in a Tanzanian village (Soremekun, et al. 2004). Similar values were also observed in non-anophelines (Benedictis, et al. 2003; Harrington, et al. 2014; Michael, et al. 2001), except for an Aedes aegypti population in Thailand where a much higher proportion (45%) was observed (Chow-Shaffer, et al. 2000). 205 The sensitivity of hosts to mosquito bites affects the outcome of a blood meal; the more intolerable a host is to mosquito bites, the more likely a blood-feeding attempt will be interrupted before a full meal is obtained (Anderson and Brust 1997; Boreham and Garrett-Jones 1973). An interrupted mosquito can either forgo further blood-feeding attempts (i.e., partial feed) or it can complete the blood meal by feeding on a different individual of the same host species or a different species. The former has important epidemiological consequences. In the case of malaria, both R0 and V increase when a considerable proportion of the vectors obtain cryptic mixed human meals (Tedrow, et al. 2019). This is because those vectors can infect more than one humans if they are sporozoite-positive, and also have more than one chance of obtaining an infectious (i.e., gametocytemic) blood meal in a single gonotrophic cycle. Before DNA profiling was available, the proportion of human blood meals that were cryptic mixed (C) was estimated indirectly from mathematical models (Boreham and Garrett- Jones 1973; Burkot, et al. 1988). According to (Burkot, et al. 1988), C = QIH, where Q is the probability of selecting a human as a host by a mosquito. Q can be calculated as Q = PH + PM [IH/(IH + IN)], where PH is the proportion of all blood meals with only human blood, PM is the proportion of all blood meals that are patent mixed human meals and IN is the probability of a non-human feed being interrupted (Burkot, et al. 1988). Substituting the right hand side of this equation for Q in the previous one gives a new equation C = (PH + PM [IH/(IH + IN)])IH. Unlike PH and PM whose values were easily estimated directly from blood-meal analysis, it was difficult to estimate IH and IN empirically and researchers depended on assumed values to estimate C (Burkot, et al. 1988). With the availability of a DNA profiling method, C can be estimated directly, as demonstrated in the current study. However, as DNA profiling is expensive, many researchers may rely on the mathematical model. Thus, some empirical estimates of IH, rather 206 than assumed values, are needed for such researchers to use in the theoretical model. In this current study a model for estimating IH, which can also be used to estimate IN, was derived (see method section) and used to estimate IH for the six Anopheles populations in PNG. The results ranged from 0.03–0.18. Researchers working with malaria vector populations in PNG can now use these IH values in the model to better estimate C. In addition to IH, a related model for calculating the probability (I) of an interrupted human feed that undergoes a second feed on any host species (not just humans) was derived. The values of I for the six Anopheles populations ranged from 0.09-0.3. The values of I were statistically greater than IH (t-test: P = 0.047), which was expected because of the high use of non-human hosts (in particular pigs and dogs) by these Anopheles species (Burkot, et al. 1989; Burkot, et al. 1988; Keven, et al. 2017; Russell, et al. 2016). 7.6. Conclusion This study employed a microsatellite-based DNA profiling method to identify different human individuals in Anopheles blood meals of four PNG villages. Analysis of the data resulted in three epidemiologically important findings. First, 6–13% of all Anopheles that feed on humans were interrupted by their host and completed their blood meal on a different individual in a single gonotrophic cycle. By feeding on more than one human, the vectorial capacity of the vector population is increased. Second, the frequency of vector bites among individuals in the villages were not random but clustered. As heterogeneous exposure to vectors is known to increase malaria R0, it is logical to postulate that non-random exposure to vector bites is a factor underlying the persistent and rebounding malaria transmission recently reported in the study villages (see chapter 3). Third, youth age male individuals were more frequently fed on by the 207 vectors compared to the females and individuals in other age groups. This is likely caused by differences in the behaviors and bed net usage among individuals in different sex and age groups. Despite overutilization of youth age males by mosquitoes, the likelihood of malaria infection was generally homogeneous among age groups and between sexes. 208 APPENDICES 209 Appendix 7.1. Percent profile match values Figure S7.1. Probability of occurrence (y-axis) of different percent match values (x-axis) calculated from pairwise percent profile match analysis of 427 DNA profiles of Mirap and 164 DNA profiles of Wasab residents. 210 Appendix 7.2. Human blood index of Anopheles populations Table S7.1. Human blood index (B) of six Anopheles populations. Village Anopheles species Bulal An. koliensis Megiar An. farauti (s.s.) Mirap An. farauti (s.s.) Mirap An. koliensis Wasab An. koliensis Total tested 28 191 120 404 440 No. of non- human meals 12 No. of human meals 16 88 85 215 122 12 103 35 189 318 42 B 0.57 0.54 0.29 0.47 0.72 0.78 Wasab An. punctulatus (s.s.) 54 Note: Mosquitoes were sampled, and the blood meal hosts were determined as described in the method section. The value of B was estimated by dividing the number of human blood meals by the total number of mosquitoes tested. 211 Appendix 7.3. Frequency distribution of blood meals Figure S7.2. Distribution of frequency of blood meals taken on different human individuals by Anopheles mosquitoes in Mirap and Wasab village. 212 REFERENCES 213 REFERENCES Anderson RA, Brust RA. 1997. Interrupted blood feeding by Culex (Diptera: Culicidae) in relation to individual host tolerance to mosquito attack. J. Med. Entomol 34(2):95–101. Beebe NW, Saul A. 1995. Discrimination of all members of the Anopheles punctulatus complex by polymerase chain reaction restriction fragment length polymorphism analysis. Am. J. Trop. Med. Hyg. 53(5):478–481. Beleza S, Alves C, Reis F, Amorim A, Carracedo A, Gusmao L. 2004. 17 STR data (AmpF/STR Identifiler and Powerplex 16 System) from Cabinda (Angola). Forensic Sci. Int. 141:193– 196. Benedictis J, Chow-Shaffer E, Costero A, Clark GG, Edman JD, Scott TW. 2003. Identification of the people from whom engorged Aedes aegypti took blood meals in Florida, Puerto Rico, using polymerase chain reaction-based DNA profiling. Am. J. Trop. Med. Hyg. 68(4):437–446. Boreham PFL, Garrett-Jones C. 1973. Prevalence of mixed blood meals and double feeding in a malaria vector (Anopheles sacharovi Favre). Bull. World Health Organ. 48:605–614. Bousema T, Drakeley C, Gesase S, Hashim R, Magesa S, Mosha F, Otieno S, Carneiro I, Cox J, Msuya E and others. 2010. Identification of hot spots of malaria transmission for targeted malaria control. J. Infect. Dis. 201(11):1764–1774. Bousema T, Griffin JT, Sauerwein RW, Smith DL, Churcher TS, Takken W, Ghani A, Drakeley C, Gosling R. 2012. Hitting hotspots: spatial targeting of malaria for control and elimination. PLoS Med. 9(1). Burkot TR, Dye C, Graves PM. 1989. An analysis of some factors determining the sporozoite rates, human blood indexes, and biting rates of members of the Anopheles punctulatus complex in Papua New Guinea. Am. J. Trop. Med. Hyg. 40(3):229–234. Burkot TR, Graves PM, Paru R, Lagog M. 1988. Mixed blood feeding by the malaria vectors in the Anopheles punctulatus complex (Diptera: Culicidae). J. Med. Entomol. 25(4):205– 213. Burkot TR, Russell TL, Reimer LJ, Bugoro H, Beebe NW, Cooper RD, Sukawati S, Collins FH, Lobo NF. 2013. Barrier screens: a method to sample blood-fed and host-seeking exophilic mosquitoes. Malar. J. 12:49. Chow-Shaffer E, Sina B, Hawley WA, Benedictis Jd, Scott TW. 2000. Laboratory and field evaluation of polymerase chain reaction-based forensic DNA profiling for use in 214 identification of human blood meal sources of Aedes aegypti (Diptera: Culicidae). J. Med. Entomol. 37(4):492–502. Das S, Muleba M, Stevenson JC, Pringle JC, Norris DE. 2017. Beyond the entomological inoculation rate: characterizing multiple blood feeding behavior and Plasmodium falciparum multiplicity of infection in Anopheles mosquitoes in northern Zambia. Parasit. Vectors 10:45. Dye C, Hasibeder G. 1986. Population dynamics of mosquito-borne disease: effects of flies which bite some people more frequently than others. Trans. R. Soc. Trop. Med. Hyg. 80:69–77. Friedlaender JS, Friedlaender FR, Reed FA, Kidd KK, Kidd JR, Chambers GK, Lea RA, Loo J- H, Koki G, Hodgson JA and others. 2008. The genetic structure of Pacific Islanders. PLoS Genet. 4(1):173–190. Fritz ML, Miller JR, Bayoh N, Vulule JM, Landgraf JR, Walker ED. 2013. Application of a reverse dot blot, DNA-DNA hydridization method to quantify host-feeding tendencies of two sibling species in the Anopheles gambiae complex. Med. Vet. Entomol. 27(4). Gorostiza A, Gonzalez-Martın A, Ramırez CL, Sanchez C, Barrot C, Ortega M, Huguet E, Corbella J, Gene M. 2007. Allele frequencies of the 15 AmpF/Str Identifiler loci in the population of Metztitlan (Estado de Hidalgo), Mexico. Forensic Sci. Int. 166:230–232. Greenwood B. 2010. Anti-malarial drugs and the prevention of malaria in the population of malaria endemic areas. Malar. J. 9:3. Gunathilaka N, Denipitiya T, Hapugoda M, Abeyewickreme W, Wickremasinghe R. 2016. Determination of the foraging behaviour and blood meal source of malaria vector mosquitoes in Trincomalee District of Sri Lanka using a multiplex real time polymerase chain reaction assay. Malar. J. 15:242. Hamer GL, Chaves LF, Anderson TK, Kitron UD, Brawn JD, Ruiz MO, Loss SR, Walker ED, Goldberg TL. 2011. Fine-scale variation in vector host use and force of infection drive localized patterns of West Nile virus transmission. PLoS ONE 6(8). Hamer GL, Kitron UD, Goldberg TL, Brawn JD, Loss SR, Ruiz MO, Hayes DB, Walker ED. 2009. Host selection by Culex pipiens mosquitoes and West Nile virus amplification. Am. J. Trop. Med. Hyg. 80(2):268–278. Harrington LC, Fleisher A, Ruiz-Moreno D, Vermeylen F, Wa CV, Poulson RL, Edman JD, Clark JM, Jones JW, Kitthawee S and others. 2014. Heterogeneous feeding patterns of the dengue vector, Aedes aegypti, on individual human hosts in rural Thailand. PLoS Negl. Trop. Dis. 8(8). 215 Hasibeder G, Dye C. 1988. Population dynamics of mosquito-borne disease: persistence in a completely heterogeneous environment. Theor. Popul. Biol. 33:31–53. Hetzel MW, Choudhury AAK, Pulford J, Ura Y, Whittaker M, Siba PM, Mueller I. 2014a. Progress in mosquito net coverage in Papua New Guinea. Malar. J. 13:242. Hetzel MW, Gideon G, Lote N, Makita L, Siba PM, Mueller I. 2012. Ownership and usage of mosquito nets after four years of large-scale free distribution in Papua New Guinea. Malar. J. 11:192. Hetzel MW, Pulford J, Maraga S, Barnadas C, Reimer LJ, Tavul L, Jamea-Maiasa S, Tandrapah T, Maalsen A, Makita L and others. 2014b. Evaluation of the global fund-supported national malaria control program in Papua New Guinea, 2009-2014. PNG Med J. 57:1–4. Hurk AFVD, Smith IL, Smith GA. 2007. Development and evaluation of real-time polymerase chain reaction assays to identify mosquito (Diptera: Culicidae) bloodmeals originating from native Australian mammals. J. Med. Entomol 44(1):85–92. Kent RJ, Norris DE. 2005. Identification of mammalian blood meals in mosquitoes by a multiplexed polymerase chain reaction targeting cytochrome b. Am. J. Trop. Med. Hyg. 73(2):336–342. Keven JB, Reimer L, Katusele M, Koimbu G, Vinit R, Vincent N, Thomsen E, Foran DR, Zimmerman PA, Walker ED. 2017. Plasticity of host selection by malaria vectors of Papua New Guinea. Parasit. Vectors 10:95. Keven JB, Walker ED, Venta PJ. 2019b. A microsatellite multiplex assay for profiling pig DNA in mosquito bloodmeals. J. Med. Entomol. 56(4):907–914. Kim Y-L, Hwang J-Y, Kim Y-J, Lee S, Chung N-G, Goh H-G, Kim C-C, Kim D-W. 2003. Allele frequencies of 15 STR loci using AmpF/STR Identifiler kit in a Korean population. Forensic Sci. Int. 136:92–95. Kleiber C, Zeileis A. 2016. Visualizing count data regressions using rootograms. Am. Stat 70(3):296-303. Koepfli C, Robinson LJ, Rarau P, Salib M, Sambale N, Wampfler R, Betuela I, Nuitragool W, Barry AE, Siba P and others. 2015. Blood-stage parasitaemia and age determine Plasmodium falciparum and P. vivax gametocytaemia in Papua New Guinea. PLoS ONE 10(5):e0126747. Logue K, Keven JB, Cannon MV, Reimer L, Siba P, Walker ED, Zimmerman PA, Serre D. 2016. Unbiased characterization of Anopheles mosquito blood meals by targeted high- throughput sequencing. PLoS Negl. Trop. Dis. 10(3). 216 Michael E, Ramaiah KD, Hoti SL, Barker G, Paul MR, Yuvaraj J, Das PK, Grenfell BT, Bundy DAP. 2001. Quantifying mosquito biting patterns on humans by DNA finger-printing of bloodmeals. Am. J. Trop. Med. Hyg. 65(6):722–728. Molaei G, Andreadis TA, Armstrong PM, Anderson JF, Vossbrinck CR. 2006. Host feeding patterns of Culex mosquitoes and West Nile Virus transmission, northeastern United States. Emerg. Infect. Dis. 12:468–474. Molaei G, Cummings RF, Su T, Armstrong PM, Greg A. Williams , Cheng M-L, Webb JP, Andreadis TG. 2010. Vector-host interactions governing epidemiology of West Nile Virus in Southern California. Am. J. Trop. Med. Hyg. 83(6):1269–1282. Nofal SD, Peto TJ, Adhikari B, Tripura R, Callery J, Bui TM, Seidlein Lv, Pell C. 2019. How can interventions that target forest‑goers be tailored to accelerate malaria elimination in the Greater Mekong Subregion? A systematic review of the qualitative literature. Malar. J. 18. Noor AM, Alegana VA, Patil AP, Snow RW. 2010. Predicting the unmet need for biologically targeted coverage of insecticide-treated nets in Kenya. Am J Trop Med Hyg 83(4):854- 860. Oostdik K, Lenz K, Nye J, Schelling K, Yet D, Bruski S, Strong J, Buchanan C, Sutton J, Linner J and others. 2014. Developmental validation of the PowerPlex Fusion System for analysis of casework and reference samples: a 24-locus multiplex for new database standards. Forensic Sci. Int. Genet. 12:69–76. Russell TL, Beebe NW, Bugoro H, Apairamo A, Cooper RD, Collins FH, Lobo NF, Burkot TR. 2016. Determinants of host feeding success by Anopheles farauti. Malar. J. 15:152. Scott TW, Githeko AK, Fleisher A, Harrington LC, Yan G. 2006. DNA profiling of human blood in anophelines from lowland and highland sites in Western Kenya. Am. J. Trop. Med. Hyg. 75(2):231–237. Smith DL, Battle KE, Hay SI, Barker CM, Scott TW, McKenzie FE. 2012. Ross, Macdonald, and a theory for the dynamics and control of mosquito-transmitted pathogens. PLoS Pathog. 8(4). Smith DL, Dushoff J, Snow RW, Hay SI. 2005. The entomological inoculation rate and Plasmodium falciparum infection in African children. Nature 438:492–495. Meece JK, Reynolds CE, Stockwell PJ, Jenson TA, Christensen JE, Reed KD. 2005. Identification of mosquito bloodmeal source by terminal restriction fragment length polymorphism profile analysis of the cytochrome B gene. J. med. Entomol. 42(4):657– 667. 217 Soremekun S, Maxwell C, Zuwakuu M, Chen C, Michael E, Curtis C. 2004. Measuring the efficacy of insecticide treated bednets: the use of DNA fingerprinting to increase the accuracy of personal protection estimates in Tanzania. Trop. Med. Int. Health 9(6):664– 672. Tedrow RE, Zimmerman PA, Abbott KC. 2019. Multiple blood feeding: A force multiplier for transmission. Trends Parasitol. 35(12):949–952. Woolhouse MEJ, Dye C, Etard JF, Smith T, Charlwood JD, Garnett GP, Hagan P, Hii JLK, Ndhlovu PD, Quinnell RJ and others. 1997. Heterogeneities in the transmission of infectious agents: implications for the design of control programs. Proc. Natl. Acad. Sci. USA 94:338–342. 218 CHAPTER 8: GENERAL CONCLUSION In chapter 2–7, a number of primary and minor conclusions were arrived at as a result of inferences made from detailed evaluation of the data. This final chapter focuses on summarizing the main conclusions to cohesively address the overarching hypothesis of this dissertation. This hypothesis was that although malaria was reduced considerably by the nationwide bed net-based vector intervention program, further reduction in the rate of transmission has halted in some parts of the country owing to ecological conditions which allow local vectors to evade the bed nets. This allows malaria transmission to persist at relatively high rates and infection prevalence to rebound despite ongoing efforts to intensify the vector intervention program. Indeed, the study reported in chapter 3 shows conclusively that in the human communities in the north coast of Madang province, the annual IBR for Plasmodium spp. has risen above 15 infective Anopheles spp. bites per person-year since it was reduced below this level in 2011, and that the prevalence of infection has rebounded after it was reduced to its lowest levels in 2014 as a result of the vector intervention program. Four vector-associated ecological factors were posited to be causing the persistent malaria transmission in this part of the country. The first was increased vector abundance and species diversity. As the purpose of the insecticide was to reduce vector abundance and consequently the IBR by killing the mosquitoes, high vector abundance signals infectiveness of the bed nets against the local vectors. The studies reported in chapter 2 and 3 show that in the north coast of Madang province, vector abundance and human biting rates have increased after plummeting to their lowest levels in 2011. Different Anopheles species or ecotypes vary in their biological attributes such as vector competence, duration of incubation period, and host 219 preference which affect their vectorial capacity. Thus, the presence of multiple vector species can increase malaria transmission potential. Vectors may also vary in host-seeking behaviors and biting cycles which can cause them to be differentially affected by the bed nets. Therefore, in communities with multiple vector species or ecotypes, those with vulnerable attributes (e.g., endophagy) can be affected by the bed nets but malaria will continue to be transmitted by the evasive ones. Thus, the presence of multiple vector species within communities and varying species composition among communities both indicate that the bed nets are not having the intended effect on the mosquitoes. Research reported in chapter 2 shows that diverse vector species live in sympatry in Madang villages and their relative composition vary even among villages that are < 1.0 km apart. Thus, it can be concluded that vector species abundance and diversity are helping to promote persistent malaria transmission in the study villages. The second ecological factor is outdoor and evening biting habits of the mosquitoes. To minimize exposure to mosquito bites and hence the likelihood of getting infected, humans must retire to bed and be protected by the bed nets during the period of the night when mosquito host- seeking activity peaks. Also, as bed nets are deployed inside houses, mosquitoes that seek humans inside houses have higher risk of exposure to the insecticide than those the feed outdoors. In this study (chapter 4), it was observed that most (29–50% of total) of the vector bites taken on humans occurred before 22:00 hr when 75–90% of people were active (awake) and were unprotected by the bed nets, and 67–83% of the bites occurred before 2:00 hr before > 95% of the people retired to bed. Most (> 50%) of the bites taken on humans occurred outdoors where the risk of exposure to the bed nets was low. Thus, it can be concluded that outdoor and evening biting behaviors of the local vectors promote persistent transmission in the study villages. 220 The third factor is heterogeneity of human exposure to the intensity of vector bites. In any environment where malaria exists the R0 of malaria has a specific value when the frequency of bites taken on individuals is homogeneous or randomly distributed. However, epidemiological models have shown that under the condition where the frequency of bites tend to cluster in a few spatial locations or on a small subset of individuals villagers, the R0 tend to increase by some quantifiable multiplicative factors. By increasing the R0, transmission tend to persist rather than dissipate. In this study, it was shown (chapter 4 and 7) that the distribution of bites by the local Anopheles spp. among spatial locations within villages and individual villagers were clustered rather than random and the R0 in these villages were increased by multiplicative factors ³ 1.6. Thus, it can be concluded that heterogeneity of human exposure to vector bites may play an important role in supporting persistent transmission in the study villages. The final factor is the propensity of vectors to feed unbiasedly on blood host species. Mosquitoes are considered non-vectors if they are strictly zoophagous because they are unable to transmit malaria. If a population is strictly anthropophagous it is considered a good vector. Strictly anthropophagic vectors are faithful to humans despite prevailing environmental conditions that may limit access to human hosts. This makes them dangerous vectors, but only under ordinary condition. Under high bed net usage, the inflexibility of the anthropophagic vectors to utilize alternative hosts will cause them to pursue humans inside houses, rendering them susceptible to the insecticide. Thus, bed net-based malaria intervention can succeed when the mosquitoes are strictly anthropophagous. If, however, the mosquito population is opportunistic in its host selection, any host that is available and accessible will be unbiasedly utilized. In this respect, inaccessibility to human hosts due to bed nets does not affect their blood meal requirement as they can utilize alternative hosts instead of pursuing humans into the house 221 and being exposed to the insecticides. By feeding on alternative hosts, populations of opportunistic mosquitoes are able to maintain high vector abundance and perpetuate transmission by feeding on humans outdoors and during the times when humans are not protected by bed nets. Studies reported in chapter 5 and 6 show that the vector species in PNG are opportunistic blood feeders and can utilize pigs and dogs as blood meal hosts. Although the main conclusions above were based on studies that were conducted in a few Madang villages, it can be extrapolated to other parts of PNG and the South West Pacific region. Thus, it can be concluded that in PNG and the South West Pacific region reliance on insecticidal bed nets alone as the primary malaria control method is insufficient to eliminate or reduce the disease to acceptably low levels. Alternative vector intervention methods, such as treatment of non-human hosts (e.g., pigs and dogs) with endectocides, need to be developed and implemented alongside the bed net program. 222