TRAIT AND ENVIRONMENTAL VARIATION MEDIATE THE INTERACTION BETWEEN A HARMFUL PHYTOPLANK TER AND AN INVASIVE GRAZER By Jeffrey D. White A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Fisheries and Wildlife Œ Doctor of Philosophy Ecology, Evolutionary Biology a nd Behavior Œ Dual Major 2015 ABSTRACT TRAIT AND ENVIRONMENTAL VARIATION MEDIATE THE INTERACTION BETWEEN A HARMFUL PHYTOPLANK TER AND AN INVASIVE GRAZER By Jeffrey D. White Phytoplankton that form harmful algal blooms (HABs) can foul water with unpleasant odors and tastes, accumulate as visible surface scums, and produce potentially dangerous toxins. The cyanobacterium Microcystis aeruginosa is the most widespread of the freshwater HAB- forming species, and this dissertation explores the influences of variation in environmental drivers (biotic, abiotic) and vari ation in biological traits (colony size, growth rate) on its ecologyŠin particular, the interaction with a facilitator species, the invasive zebra mussel (Dreissena polymorpha). Chapter 1 quantifies the vulnerability of Microcystis to grazing by zebra mussels as a function of large variation in both Microcystis colony size (5-88 µm equivalent diameter) and zebra mussel body size (8-28 mm shell length), and re lates the findings to their size distributions in the primary study lake, Gull Lake, Mich igan. Based on colony size alone, the Microcystis population in Gull Lake can vary widely in its vulnerability to grazing within single growing seasons, and the range of ingestible colonies ( 80 µm equivalent diameter) is greater for zebra mussels than published ranges for other dominant filter-feeding grazers (e.g., Daphnia). Following a mass mortality event (~100%) of zeb ra mussels on epilimnetic sediments in Gull Lake during a relatively warm summer , evidence from a combination of in situ monitoring and experiments presented in Chapter 2 demonstr ates a causal relations hip between chronic, accumulated heat exposure (> 25 °C) and elevated zebra mussel mortality. Though these temperatures are lethal to zebra mussels , they are within the optimal range for Microcystis. Results from a long-term (13-year) study of the Gull Lake Microcystis population are presented in Chapter 3, the fi rst long-term analysis of Microcystis dynamics in a low-nutrient lakeŠan uncharacteristic niche for this species strongly facilitated by zebra mussels. Microcystis biomass and microcystin toxin we re significantly higher and peak biomass occurred significantly earlier in warmer summers, consistent with climate change projections. However, the heat- induced mass mortality event of zebra mussels (Chapter 2) resulted in a 2-year collapse of the Microcystis population during the warmest period in th e time series, highlighting the need to understand how these two strongly interacting species will respond together to climate warming. Lastly, Chapter 4 returns to the importance of large intraspecific trait variation for the ecology of Microcystis, to further understand its niche expansion into low-nutrient lakes. Laboratory growth assays of 18 colonial strains, recently isolated from 11 Michigan inland lakes spanning the entire productivity gradient (7.6-196 µg L-1 total phosphorus), show that Microcystis strains from high- nutrient lakes grow significantly faster (up to ~7 fold) than those from low-nutrient lakes, which may indicate the presence of an ecological trade- off enabling local adaptation to these widely disparate habitats. Possibly as a result of their faster growth ra tes, strains from high-nutrient lakes are also more likely to cease colony formati on in culture sooner, which has implications for the design and interpretation of lab studies of Microcystis. Taken together, the chapters within this dissertation demonstrate important ecological consequences of the biological trait variation inherent within and among populations, and illustrate how that diversity might interact with other biotic and abiotic factors, improving our understanding of species™ responses to complex global change. iv To my wife and parents, who gave and sacrificed so much to support me in this endeavor. v ACKNOWLEDGMENTS First and foremost I thank my major adviso r, Orlando ‚Ace™ Sarnelle, and my guidance committee members, Pat Soranno, Steve Hamilt on, and Gary Mittelbach, for mentoring, challenging, and supporting me throughout my gra duate program. Their expertise, constructive criticism, and humor were all greatly appreciated. I also thank the other faculty of the Michigan State University Limnology Lab, in addition to Ace and Pat, for their many years of advice, feedback, and encouragement that helped me to gain confidence, polish talks, obtain funding, pursue incredible teaching opportunities, and su ccessfully land a job: Mary Bremigan, Kendra Cheruvelil, Jo Latimore, Scott Peacor, and Lois Wolfson. Funding for this dissertation research was provided by the Environmental Protection Agency (Ecology and Oceanography of Harmful Algal Blooms/2004-Science to Achieve Results-C1, project RD83170801), the National Science Foundation (Division of Environmental Biology-0841864, Division of Environmental Biol ogy-0841944), the College of Agriculture and Natural Resources and the Graduate School at Michigan State University (Recruitment Fellowship and Dissertation Comp letion Fellowship), the Robert C. Ball and Betty A. Ball Fisheries and Wildlife Fellowship, and the Gull Lake Quality Organization. Additional funding to travel and present this research at scientif ic conferences was provided by the Department of Fisheries and Wildlife, the EEBB Program, the Co llege of Agriculture and Natural Resources, and the Graduate School at Michigan State University, as well as th e Association for the Sciences of Limnology and Oceanography. The research presented here would not have been possible without the help, hard work, and dedication of numerous lab and field technicians that I ha ve had the pleasure of working with: Joel Berry, Shelby Flemming, Theresa Ge elhoed, Mark Iadonisi, Carrie Kozel, Kelsey vi Lincoln, Megan Schuetz, and Dave Weed. The staff at the W.K. Kellogg Biological Station, especially Tyler Brownell, Nina Consolatti, Andy Fogiel, and Mark Williams were instrumental in ensuring all seven of my field seasons on Gull Lake ran smoothly. There are many other people and organizations that deserve recognition and a big ‚thank you,™ and they are acknowledged at the conclusion of each chapter for which they made specific contributions. I especially thank the graduate students and post-docs of the Limnology Lab at Michigan State University for their friendship and camarad erie which made the day-to-day pressures of graduate school tolerable: Tom Alwin, Stacie Auvenshine, Josh Booker, Paul Bourdeau, Sarah Collins, Angela DePalma-Dow, Katie Droscha, Em i Fergus, Geoff Horst, J.-F. Lapierre, Andrea Miehls, Dianna Miller, Emily Norton-Henry, Al ex Rafalski, Caren Scott, Nick Skaff, Kim Winslow, Caroline Wynne, and Heidi Ziegenmeyer. To the 547 students I had the privilege to work with in class: thank you for your enthusiasm for learning, your contri butions that made me a better educator, and for affirming my decision to be a life-long educator and learner. The Biology faculty at Saint Michael™s Colle ge inspired me to pursue graduate school and a career in undergraduate education, two major life goals that have now been realized as a result of huge leaps they helped and encouraged me to take. Last but not least, I thank my wife Jenny, parents David and Kathy, and my other family and friends for their love, kindness, and unyielding support. I could not have done this without you. vii PREFACE This dissertation was prepared in manuscript format with each chapter submitted or to be submitted as separate papers to peer-reviewed journals. Therefore, ‚we™ is used in place of ‚I™ throughout to reflect the contribution of all co-authors to this work, including project conceptualization, data collection and analyses, and feedback on earlier written drafts. Co-authors are: Chapter 1, Orlando Sarnelle; Chap ter 2, Stephen Hamilton and Orlando Sarnelle; Chapter 3, Stephen Hamilton and Orlando Sa rnelle; Chapter 4, Orlando Sarnelle. At the time this dissertation was written, Ch apters 1 and 2 were already published in the peer-reviewed literature exactly as they appear here. Complete bibliographic information for these two publications is as follows. Chapter 1: White, J. D. and Sarnelle, O. 2014. Size-structured vulnerability of the colonial cyanobacterium, Microcystis aeruginosa, to grazing by zebra mussels ( Dreissena polymorpha). Freshwater Biology 59:514-525. Chapter 2: White, J. D., Hamilton, S. K., and Sarnelle, O. 2015. Heat-induced mass mortality of invasive zebra mussels ( Dreissena polymorpha) at sublethal water temperatures. Canadian Journal of Fisheries and Aquatic Sc iences. Pagination not final. viii TABLE OF CONTENTS LIST OF TABLES x LIST OF FIGURES xii KEY TO SYMBOLS AND ABBREVIATIONS xvii PROLOGUE 1 CHAPTER 1 SIZE-STRUCTURED GRAZING VULN ERABILITY OF THE COLONIAL CYANOBACTERIUM, MICROCYSTIS AERUGINOSA, TO ZEBRA MUSSELS (DREISSENA POLYMORPHA) 5 Abstract 5 Introduction 6 Methods 9 Collection and maintenance of experimental organisms 10 Design of feeding experiments 13 Sample processing and data analysis 16 Lake sampling 19 Results 19 Experiment 1: colony size-selectivity experiment 19 Experiment 2: mussel size × colony size experiment 20 Lake sampling 24 Discussion 28 Acknowledgements 33 CHAPTER 2 HEAT-INDUCED MASS MORTALITY OF INVASIVE ZEBRA MUSSELS (DREISSENA POLYMORPHA) AT SUBLETHAL WATER TEMPERATURES 34 Abstract 34 Introduction 34 Methods 40 Study site 40 In situ mortality: caged mussels 40 In situ mortality: mussels in enclosures 41 Experiment 1: acute temperature tolerance 43 Experiment 2: chronic temperature tolerance 45 Statistical analyses 47 Results 48 In situ mortality: caged mussels 48 In situ mortality: mussels in enclosures 50 Experiment 1: acute temperature tolerance 54 Experiment 2: chronic temperature tolerance 54 ix Discussion 57 Acknowledgements 62 Supplemental Information 63 CHAPTER 3 OPPOSING RESPONSES OF STRONGLY INTERACTING SPECIES TO ELEVATED TEMPERATURES SUPPRESS THE HARMFUL PHYTOPLANKTER MICROCYSTIS 64 Abstract 64 Introduction 65 Methods 68 Study site 68 Limnological sampling 71 Lab analyses 72 Temperature data 74 Statistical analyses 76 Results 81 Responses of Microcystis biomass to interannual variation in water temperature 81 Responses of Microcystis biomass to interannual variation in other potential drivers 86 Responses of microcystin toxin 91 Discussion 91 Acknowledgements 100 CHAPTER 4 GROWTH VARIATION AMONG STRAINS OF THE HARMFUL CYANOBACTERIUM, MICROCYSTIS AERUGINOSA, ACROSS A LARGE PRODUCTIVITY GRADIENT OF LAKES 102 Abstract 102 Introduction 103 Methods 106 Isolation and maintenance of lab strains 106 Growth rate assays 111 Monitoring changes in growth habit 114 Lab analyses 114 Statistical analyses 115 Results 116 Growth rate assays 116 Monitoring changes in growth habit 120 Discussion 124 Acknowledgements 128 APPENDIX 129 LITERATURE CITED 133 x LIST OF TABLES Table 1. Median sizes of Microcystis aeruginosa colonies (Gull Lake strain 2009C) in the eight feeding suspensions used in first experiment. See text for a description of the size metrics. 14 Table 2. Median sizes of Microcystis aeruginosa colonies (Gull Lake strain G11-08) in the three feeding suspensions used in the s econd experiment. See text for a description of the size metrics. 15 Table 3. Selection of an accumulated degree hou r threshold temperature that best explains mortality of caged large (16-30 mm, n = 10) and small (8-15 mm, n = 7) Dreissena polymorpha in Gull Lake. Results are for linear regressions. Model fit was assessed with AIC, where a lower AIC indicates a better fit. AIC was then computed for each model with respect to the mode l with the lowest AIC. Models having AIC 2 (underlined) were considered to have equal statistical support as the model with the lowest AIC. 49 Table 4. Accumulated heat exposure (degree hours > 25 °C) across years and depths in the study of caged Dreissena polymorpha in Gull Lake. 51 Table 5. Summary of limnological characteristics of Gull Lake, Michigan (1998-2014, mixed layer during summer stratific ation). SE = standard error; 10th = 10th percentile; 90th = 90th percentile; n = number of observations; Chl- a = chlorophyll-a; Secchi = Secchi disk transparency; Zepi = epilimnion depth; Alk = alkalinity; TP = total phosphorus; TDP = total dissolved phosphorus; SRP = soluble reactive phosphorus; NO3- = nitrate; NH 4+ = ammonium; DIN:TP = ratio of [NO 3- + NH4+ ] to TP. 70 Table 6. Mean Gull Lake water temperatures (° C), computed annually for each seasonal time scale: spring (April-May), summer (June-August), and spring-summer. 87 Table 7. Influence of water temperature on annual means for each response variable for years when zebra mussels were pres ent in Gull Lake (excluding 2011-2012, see text). We also assessed the influence of the s easonal scale over which water temperatures were averaged on each response variable. The three seasonal time scales are: spring (April-May), summer (June-August), and spring-summer. Predictors are listed from the best to the worst fit, as assessed by AIC. Microcystis biomass was log-transformed prior to analysis. Significant ( p 0.05) results are underlined. All results are for linear regressions. 89 Table 8. Limnological data on the Microcystis aeruginosa source lakes, and a summary of the sampling, isolation, and establishment of all lab strains. A strain was considered established if the isolate gr ew and the seed culture was devoid of any other algal contaminants. TP = total phosphorus; SRP = soluble reactive phosphorus; Chl-a = chlorophyll-a; Secchi = Secchi disk transparency. 109 xi Table 9. Within-population variation in maximum intrinsic growth rate (d -1) of Microcystis aeruginosa, as determined with individual colony growth assays. The four source lakes range from oligot rophic to eutrophic (7.9-47.8 µg L -1 total phosphorus) from top to bottom. Data were pooled for all 2011 and 2013 strains assayed from the given lake population. The coefficients of variation (CV) and sample sizes ( n) are given. 119 xii LIST OF FIGURES Figure 1. Formation of a surface scum duri ng a mid-August bloom of toxic Microcystis aeruginosa in a hyper-eutrophic shallow lake in Michigan. Photo credit: Jeffrey D. White. 3 Figure 2. (A) Microcystis aeruginosa colony from Gull Lake, Michigan, photographed under a light microscope at 100×. The cells a ppear brown due to preservation in Lugol™s iodine. (B) Zebra mussels ( Dreissena polymorpha) filter-feeding in Gull Lake at a depth of 1 m. Photo credits: Jeffrey D. White. 4 Figure 3. Estimated Dreissena polymorpha filtering impact on phytoplankton in the mixed layer of Gull Lake (0-7.5 m depth). Si ze distribution data were obtained from Wilson and Sarnelle (2002). Filterin g impact was calculated using the D. polymorpha length-filtering rate regression of Kryger and Riisgård (1988). 12 Figure 4. View of a laboratory feeding experime nt with zebra mussels. Photo credit: Jeffrey D. White. 17 Figure 5. Results from experiment 1, the colony size-selectivity experiment. (A) Mean selectivity of Dreissena polymorpha across size classes of a single strain (2009C) of colonial Microcystis aeruginosa (n = 31). Selectivity was calculated as filtering rate on M. aeruginosa divided by filtering rate on the high-quality food alga, Ankistrodesmus falcatus. Error bars represent ± 1 SE. Asterisks indicate size fractions eliciting significant avoidance (selectivity significantl y < 1). (B, C) Comparison of mean filtering rates on A. falcatus and M. aeruginosa, respectively. Error bars represent ± 1 SE. An N indicates non-detectable filtration (filtering rate not significantly > 0). Median Microcystis size is reported as both equivalent diameter (ED) and maximum linear dimension (MLD). 21 Figure 6. Results from experiment 2, the mussel size × colony size experiment. Mean mass-specific filtering rates of three common size classes of Dreissena polymorpha (shell length ranges given in mm) fe eding on a size-fractioned strain of Microcystis aeruginosa (n = 27). Error bars represent ± 1 SE. Median M. aeruginosa colony size is reported as both equivalent diameter (E D) and maximum linear dimension (MLD). 23 Figure 7. Size structure of the Microcystis aeruginosa population in Gull Lake: overall colony size distribution (equivalent diameter, ED) obtained from pooling 8 summers of measurements ( n = 3,035). 25 Figure 8. Size structure of the Microcystis aeruginosa population in Gull Lake: comparison of July and August colony size di stributions by year. The horizontal dotted line at 80 µm ED is the approximate threshold above which M. aeruginosa is largely invulnerable to grazing by Dreissena polymorpha (see Fig. 5). 26 xiii Figure 9. Seasonal change in median Microcystis aeruginosa colony size (equivalent diameter, ED) in Gull Lake during three consecutive summers. The horizontal dotted line at 80 µm ED is the approximate threshold above which M. aeruginosa is largely invulnerable to grazing by Dreissena polymorpha (see Fig. 5). 27 Figure 10. (A) Recently expired Dreissena polymorpha retrieved from the nearshore area of Gull Lake immediately following the 2010 epilimnetic mass die-off event. (B) Shells of dead D. polymorpha cover the bottom of Gull Lake at a depth of 2 m in 2011, following the 2010 mass die-off event. Phot o credits: (A) Stephen K. Hamilton; (B) Jeffrey D. White. 37 Figure 11. Box plots of mean daily June-August air temperatures recorded at the Kellogg Biological Station, adjacent to Gull Lake, for the years after Dreissena polymorpha invaded Gull Lake. The die-off occurred during the shaded summer (2010), and the dotted line indicates the median ai r temperature (22.4 °C) for that summer. Whiskers extend to the lowest and highest non-outliers; open circles indicate outliers. 38 Figure 12. Monitoring of in situ zebra mussel mortality in Gull Lake, Michigan. (A) View of tethered cages in position on the lake bottom at a depth of 2m. (B) View of the 2008 enclosure experiment. Mussels were suspended within the enclosures inside cage trees similar to those visible on the platform at center right. Photo credits: Jeffrey D. White. 42 Figure 13. View of the zebra mussel acute temperature tolerance experiment, showing the nested aquarium water bath design. Th e entire experiment was housed within an incubator. Photo cred it: Jeffrey D. White. 44 Figure 14. View of the zebra mussel chronic temperature tolerance experiment. Experimental trays with nested water bath s are in the foreground; lake water holding tanks are visible in the background. The peristaltic pump is at top center. Photo credit: Jeffrey D. White. 46 Figure 15. Mortality of caged Dreissena polymorpha in Gull Lake (May-October, 2011-2014) for (A) large (16-30 mm) and (B) sma ll (8-15 mm) individuals as a function of lake degree hours > 25 °C. Results from the chronic temperature tolerance experiment (see Fig. 18) are indicated for comparison. The regression lines (significant at p < 0.05) are for caged D. polymorpha only. Mortality is given as the proportion of dead individuals. Bars are ± SE. The arrows denote accumulated degree hours in Gull Lake during the initial mass die-off event in 2010. 52 Figure 16. Mortality of Gull Lake Dreissena polymorpha in four enclosure experiments conducted in Gull Lake (July-August, 2005-2008), as a function of median air temperatures recorded nearby. Mortality is gi ven as the proportion of dead individuals. Bars are ± SE. 53 xiv Figure 17. Water temperatures and mortality of large (16-25 mm) and small (8-15 mm) Gull Lake Dreissena polymorpha in the acute temperature to lerance experiment for (A) heated and (B) control treatments. Morta lity is given as the proportion of dead individuals. Bars are ± SE. 55 Figure 18. Results from the chronic temperature to lerance experiment: (A) temperature, (B) accumulated degree hours > 25 °C, and (C) mortality of large (17 mm) Gull Lake Dreissena polymorpha in heated and control treatments. Mortality is given as the proportion of dead individuals. Bars are ± SE. 56 Figure 19. Map of Gull Lake, Michigan, indicating the locations of the nearshore (fidockfl) and central (ficenterfl) sampling stations. Depth contours are in increments of 3 meters. 69 Figure 20. Prediction of total phytoplankton dry biomass from chlorophyll- a concentration in the epilimnion of Gull Lake ( n = 8). 75 Figure 21. Relationship between daily mean air temp eratures recorded at the Long-term Ecological Research (LTER) site at the Kellogg Biological Station, and at a lakeshore laboratory on Gull Lake (n = 273). 77 Figure 22. Conversion of KBS-LTER air temper ature data to epilimnetic water temperature for Gull Lake, by s eason. (A) Spring (April-May, n = 105) and (B) summer (June-August, n = 604). Left panels illustrate the selection of the best-fitting lag time between air and water temperatures via R 2 (squares) and AIC (circles). Right panels show the relationship between observed epilimnetic water temperatures in Gull Lake (2010-2014) and the best-fitting air temperature lag of 22 days. The 22-day mean air temperature leading up to and including th e day of water temperature observation was the shortest lag having maximal statistical support ( AIC < 2; see text). 78 Figure 23. Time series plots of Microcystis parameters measured in Gull Lake, 1998- 2014. (A) Microcystis dry biomass, (B) proportion of the total phytoplankton dry biomass as Microcystis, (C) median Microcystis colony size (as equivalent diameter), and (D) particulate (cell-bound) microcys tin concentration. Microcystin was not monitored in Gull Lake until 2005, and sampli ng for all other response variables was conducted intermittently in 1999 and 2002-2004. Asterisks (*) indicate observations made during years in which zebra mussels were largely absent from the epilimnetic sediments of Gull Lake (due to heat-induced mass mortality, see text). 82 xv Figure 24. Seasonal dynamics of Microcystis biomass in Gull Lake during a representative warm year (2010; mean spring-summer water temperature = 21.41 °C, maximum biomass = 69.5 µg L-1 on 29 July) and a representative cool year (2014; mean spring-summer water temperatur e = 19.51 °C, maximum biomass = 11.4 µg L-1 on 19 August). The two series are plotted on a common time scale (day of year) to emphasize differences in the timing of first appear ance and the timing and magnitude of the biomass maximum. The mass mortality event of zebra mussels occurred between days 209 and 215 in 2010. 84 Figure 25. Microcystis biomass dynamics as a function of in terannual variation in Gull Lake water temperatures: (A) annual mean Microcystis dry biomass, and (B) day of year of maximum Microcystis biomass. Years with zebra mussels ( Dreissena polymorpha) present on epilimnetic sediments (circles, n = 11) are differentiated from years (2011 and 2012, squares) during which zebra mussels were largely absent on epilimnetic sediments as a result of heat-induced mass mortality ( see text). Results from the linear regression analysis in (A) are for years with Dreissena only, whereas those in (B) are for all years ( see text for explanation). 85 Figure 26. Microcystis biomass (A) and particulat e (cell-bound) microcystin concentration (B) in Gull Lake in the presence (n = 11, black columns) versus absence (2011-2012; n = 2, open columns) of zebra mussels ( Dreissena polymorpha) on the epilimnetic sediments. Zebra mussels were largely absent on epilimnetic sediments in 2011-2012 as a result of heat-induced mass mortality ( see text). Error bars represent ± SE. 88 Figure 27. Relationships between mean Microcystis biomass in Gull Lake and (A) mean proportion of total algal dry biomass as Microcystis (n = 13) and (B) mean particulate (cell-bound) microcystin toxin concentration (n = 10). Years with zebra mussels (Dreissena polymorpha) present on epilimnetic sediments (circles) are differentiated from years (2011 and 2012, squares) during whic h zebra mussels were largely absent on epilimnetic sediments as a result of heat-induced mass mortality ( see text). Results from linear regression analyses are for all years ( see text for explanation). 92 Figure 28. Map indicating the locations and trophic state (based on total phosphorus) of all 2011 and 2013 Microcystis aeruginosa source lakes. 107 Figure 29. The Microcystis aeruginosa culture collection. The right panel illustrates the buoyant, colonial attributes of the recently isolated strains. 108 Figure 30. Sequence of digital micrographs depicting growth of an individual colony of Microcystis aeruginosa during the course of a 6-day growth assay. Colonies were photographed at 100× and all images are shown to scale. The strain pictured is F11-05, isolated in 2011 from eutrophic Ford Lake, Michigan. Photo credits: Jeffrey D. White. 112 xvi Figure 31. Comparison of maximum growth rates of eight strains of Microcystis aeruginosa as determined from two different methods: individual colony and batch culture assays. The two methods yielded gr owth rates that were not significantly different from each other. Error bars denote ± SE. 117 Figure 32. Variation in maximum in trinsic growth rate of Microcystis aeruginosa strains isolated from lakes spanning a large la ke productivity gradient, from oligotrophic to hyper-eutrophic (as total phosphorus). Growth rates were determined from the change in volume of individual colonies during a 6- day growth assay with saturating nutrients (n = 18 strains). Growth rate data (batch culture assays of M. aeruginosa) from Wilson et al. (2006) are shown for comparison (square s). The linear regression is for the current study only. Note the log scale on the x-axis. 118 Figure 33. Variation in microcystin quota of the Microcystis aeruginosa strains (n = 18) assayed in Fig. 32 as a function of (A) s ource lake total phosphorus and (B) maximum intrinsic growth rate. 121 Figure 34. Logistic regression of growth habit (colonial or single-celled) of all Microcystis aeruginosa strains (n = 73) isolated in 2013 versus source lake total phosphorus, 1.5 years post-isolation into lab culture. All strains were colonial at the time of isolation. Data points are randomly jit tered vertically to avoid over-plotting. 122 Figure 35. Comparison of maximum growth rate s as a function of growth habit (colonial or single-celled) for twelve Microcystis aeruginosa strains. Strains were all assayed initially as colonies ( see Fig. 32, Table 9), and then again after disaggregating following a period of ~2 years in lab culture. 123 Figure 36. Locations of Michigan inland lakes for which recent Dreissena polymorpha status reports (decline or not during 2010-2013) were received during our qualitative survey of statewide lake associations. 132 xvii KEY TO SYMBOLS AND ABBREVIATIONS °C degree Celsius µeq microequivalent µg microgram µm micrometer µmol micromole AIC Akaike information criterion ANCOVA analysis of covariance ANOVA analysis of variance C cell density Chl-a chlorophyll-a CV coefficient of variation d day df degrees of freedom DIC dissolved inorganic carbon DIN dissolved inorganic nitrogen ED equivalent diameter F filtering rate g gram h hour ha hectare HAB harmful algal bloom id inner diameter xviii k light extinction coefficient KBS W. K. Kellogg Biological Station Km half-saturation constant L liter ln natural logarithm (base e) log logarithm (base 10) LTER Long Term Ecological Research m meter max maximum mg milligram MiCorps Michigan Clean Water Corps min minimum mL milliliter MLD maximum linear dimension MLSA Michigan Lake and Stream Associations mm millimeter N population size n sample size N nitrogen NH4+ ammonium NO3- nitrate p p-value P phosphorus xix r Pearson correlation R2 coefficient of determination RM repeated measures rmax maximum intrinsic growth rate SE standard error SRP soluble reactive phosphorus t t-value TDP total disso lved phosphorus TP total phosphorus V volume W watt Zepi epilimnion (mixed layer) depth ZSD Secchi disk transparency 1 PROLOGUE The quality of freshwater ecosystems is of paramount social and economic concern given their importance as drinking water sources, as well as their aesthetic and recreational values. Yet, these systems are increasingly vulnerable to and impacted by numerous anthropogenic stressors associated with global changeŠparticularly cu ltural eutrophication, climate warming, and the spread of invasive species (Carpenter et al. 1992, Ricciardi and MacIsaac 2000, Rigosi et al. 2014). Each of these stressors is known to promote species of nuisance phytoplankton that can form harmful algal blooms (HABs) (Schindl er 1974, Raikow et al. 2004, Paerl and Huisman 2008, Elliott 2012), which foul wate r with unpleasant odors and ta stes, accumulate as visible surface scums (Fig. 1), and produce potentially dangerous toxins that can poison humans, livestock, and pets (Carmichael 1994, Chorus and Bartram 1999, Huisman et al. 2005, Cronberg and Annadotter 2006). The colonial , toxin-producing cyanobacterium Microcystis aeruginosa (Fig. 2a) is the most common of the freshwate r HAB-forming species in nutrient-polluted (eutrophic) lakes, and represents a major global threat to water quality and human health (Visser et al. 2005). Despite the fact that M. aeruginosa is an asexual, prokaryotic organism, its populations are known to harbor a surprising extent of genetic (Wilson et al. 2005, Kardinaal et al. 2007a, Tanabe et al. 2007, Dyble et al. 2008) and phenotypic (Otsuka et al. 2000, Martins et al. 2009, Horst et al. 2014, White and Sarnelle 2014) variati on, which could strongly influence its ecology, human impacts, and response to global cha nge (Burkholder and Glibert 2009). However, understanding of the ecological implications of this large biological variation within M. aeruginosa, and species in general, is still limited (B olnick et al. 2011, White et al. 2011, Violle et al. 2012). 2 Surprisingly, M. aeruginosa has also expanded its niche into low-nutrient (oligotrophic) lakes, where it would not otherwise achieve appreciable abundance (Watson et al. 1997, Downing et al. 2001), as a direct result of facilitation by an invasive bivalve grazer, the zebra mussel (Dreissena polymorpha; Fig. 2b) (Vanderploeg et al. 2001, Raikow et al. 2004, Sarnelle et al. 2005). Microcystis biomass and associated microcys tin toxin concentrations are significantly elevated in these otherwise high- quality lakes, as compared to similar but non- invaded lakes (Knoll et al. 2008, Sarnelle et al. 2010). Though this pattern has been experimentally validated (Sar nelle et al. 2012), little else is known about the ecology of M. aeruginosa in these uncharacteristic low-nutrient habita ts. In particular, the consequences of ongoing climate change for Microcystis populations in low-nutrient lakes are unknown. This dissertation explores the influences of variation in environmen tal drivers (e.g., zebra mussels, temperature) and variation in biological traits (e.g., colony size, growth rate) on the ecology of Microcystis, with an emphasis on the ecological in teraction with zebra mussels. In addition to improving knowledge of the dynamics of harmful cyanobacteria, the presented results have broader implications for our understanding of population dynami cs and species interactions, and for predicting the responses of ecosystems to complex global change. 3 Figure 1. Formation of a surface scum duri ng a mid-August bloom of toxic Microcystis aeruginosa in a hyper-eutrophic shallow lake in Michigan. Photo credit: Jeffrey D. White. 4 Figure 2. (A) Microcystis aeruginosa colony from Gull Lake, Michigan, photographed under a light microscope at 100×. The cells appear br own due to preservation in Lugol™s iodine. (B) Zebra mussels ( Dreissena polymorpha) filter-feeding in Gull Lake at a depth of 1 m. Photo credits: Jeffrey D. White. A) B) 5 CHAPTER 1 SIZE-STRUCTURED GRAZING VULN ERABILITY OF THE COLONIAL CYANOBACTERIUM, MICROCYSTIS AERUGINOSA, TO ZEBRA MUSSELS ( DREISSENA POLYMORPHA) Abstract We quantified the vulnerability of colonies of the bloom-forming cyanobacterium, Microcystis aeruginosa, to grazing by the invasive filter feeding zebra mussel ( Dreissena polymorpha) as a function of size in both organisms with laboratory feeding experiments. In one experiment, size-selectivity of 16-21 mm mussels was assessed for a single M. aeruginosa strain across a wide size range (~5-88 m median equivalent diameter, ED). Consumption of colonies 80 m median ED (109 m median maximum linear dimension, MLD) was undetectable, indicating a size threshold of grazing invulnerability. Smaller colonies and single cells were consumed at rates sim ilar to a highly-palatable alga ( Ankistrodesmus ). In a second experiment, the size-selectivity of three size classes of mussels (8-11, 17-20, and 25-28 mm shell length) was assessed across three size classes of M. aeruginosa (~32-75 m median ED). There were no systematic differences in the abilities of the different mussel size classes to consume the largest colonies within this size range . An 8-year field survey of the M. aeruginosa population in Gull Lake, MI, the source of the experimental or ganisms, revealed that median colony size consistently decreased during each summer, from above to below the size threshold of effective mussel feeding we identified, which suggests major within-season shifts in the overall vulnerability of the M. aeruginosa population to mussel grazing. Variation in the size structure of M. aeruginosa may help explain highly-variable effects of D. polymorpha on the dynamics of this harmful phytoplankter w ithin and across systems. 6 Introduction Microcystis aeruginosa, a colonial cyanobacterium comm on in eutrophic freshwaters, poses a major water quality threat due to the production of toxic, harmful algal blooms (HABs) that degrade drinking water and impede recreatio nal use. Better understanding of the ecology of this species is critical for mitigating these negativ e effects. In this paper, we focus on the ability of grazers to consume and potentially control M. aeruginosa populations in nature by quantifying size-selective feeding behavior by the filter-feeding zebra mussel, Dreissena polymorpha, within a context of natural size variation in M. aeruginosa. Due to the formation of large, mucilaginous colonies and the production of toxins and other putative chemical deterrents (Fu lton and Paerl 1987b, a, Jungmann and Benndorf 1994, Agrawal 1998), M. aeruginosa is widely characterized as ‚ine dible,™ although recent studies have documented high rates of consumption and control of this species by some grazers (White et al. 2011, Chislock et al. 2013). Populations exhibit marked variation in colony size (spanning several orders of magnitude) w ithin and across lakes (Reynolds and Rogers 1976, Reynolds et al. 1981, O'Brien et al. 2004, Wilson et al. 2006). Colony size is also highly variable within and among genotypes, showing little genetic correl ation (Rico et al. 2006, Wilson et al. 2010). Despite the potential importa nce of size structure in Microcystis -grazer interactions, many studies investigating the grazing vulnerability of M. aeruginosa have employed cultured strains that are either single-celled or produce only small colonies (B aker et al. 1998, Dionisio Pires and Van Donk 2002, Dionisio Pires et al. 2004). Grazer selectivity for size-fractioned Microcystis has been investigated in Daphnia (Jarvis et al. 1987, 1988). These studies used naturally-occurring colonial Microcystis, thus avoiding the aforementioned limitations and pr oviding valuable information about the size- 7 structured grazing vulnerability of the particular Microcystis population under scrutiny. However, we recently reported that different strains of M. aeruginosa isolated from the same population can vary maximally in their vulnerabi lity to grazing independent of colony size (White et al. 2011). Consequently, factors othe r than colony size that vary among co-occurring genotypes (e.g., cellular chemistry) could confound attempts to de lineate size-vulnerability relationships in feeding experiments using natu ral seston. Thus, there is a need for further investigation as to how col ony size affects grazing vulnerability in the absence of such confounds. The ongoing infestation by the inva sive, filter-feeding zebra mussel ( Dreissena polymorpha) has led to marked changes in North Am erican freshwater ecosystems (Strayer 2008, Higgins and Vander Zanden 2010), including unexpected increases in M. aeruginosa biomass in low-nutrient lakes (Vanderploeg et al. 2001, Raikow et al. 2004, Knoll et al. 2008, Sarnelle et al. 2010). Furthermore, the effect of D. polymorpha on M. aeruginosa biomass is highly variable across invaded ecosystems, e xhibiting strong sensitivity to total phosphorus concentration (Sarnelle et al. 2012). Better understanding of this interaction is needed, and large natural variation in M. aeruginosa colony size may help explain these variable effects. Dreissena are selective feeders (Baker et al. 1998, Bastviken et al. 1998, Vanderploeg et al. 2001, Dionisio Pires et al. 2004), but there is great variation in the size ranges of ingestible particles reported for D. polymorpha. For example, D. polymorpha has been shown to reject particles > 50-80 µm in some studies (Ten Winkel and Davids 1982, Roditi et al. 1996, Naddafi et al. 2007), K. Wissing, unpubl. data) yet filter phytoplankton up to 150 µm at equal rates in another (Horgan and Mills 1997). Often, particle size-classes in feeding studies are created by varying phytoplankton species comp osition, either by using different laboratory-cultured species 8 or by fractionating natural seston (Sprung and Rose 1988, Bern 1994, Horgan and Mills 1997, Dionisio Pires et al. 2004). This introduces a confound between particle size and other factors (such as chemical composition) that vary across phytoplankton species, and it has been established that factors other than particle size have a strong influence on Dreissena feeding (Ten Winkel and Davids 1982, Bastviken et al. 1998, Naddafi et al. 2007, White et al. 2011), leaving lingering uncertainty about their size selectivity. Previous studies have provided a valuable starting point with which to specifically assess size-selectivity of D. polymorpha for M. aeruginosa, but the upper limit to ingestion has not been determined. For example, studies have fractionated pure Microcystis cultures into coarse ‚small™ versus ‚large™ size classes about an arbitrary cutoff (e.g., 53 or 60 µm (Vanderploeg et al. 2001, Dionisio Pires et al. 2005) and observed efficient clearance by musse ls on the larger fraction. We can thus expect the upper limit to in gestion to be greater than 53-60 µm. However, more precisely estimating the upper size limit of eff ective grazing requires employing many more size fractions that each contains a very narrow range of colony sizes. In addition to the above, we are not aware of any study quantifying the relative mortality imposed by different size classes of Dreissena on different size classes of phytoplankton (Naddafi et al. 2007). Body size is of fundamental importance in determining the size spectrum of ingestible particles for filte r-feeding zooplankton: larger an imals are capable of consuming larger particles (Burns 1968, Geller and Müller 1981, Bogdan and Gilbert 1984, Hansen et al. 1994). Dreissena polymorpha inhalant siphon diameter increases with body size, leading to the prediction that larger mussels s hould be able to ingest larger particles (MacIsaac et al. 1991, Naddafi et al. 2007). Bridoux et al. (2010), quantifying phytoplankton biomarkers in Lake Erie D. polymorpha tissue, found evidence for increased impor tance of larger phytoplankton taxa in 9 the diet of larger mussels. Yet, MacIsaac et al. (1995) found no systematic differences among size classes of D. polymorpha >10 mm in their ability to inge st microzooplankton up to 89 m, though these results may not be directly compar able to phytoplankton. An assessment of the relative size-selectivities of different size classes of D. polymorpha for phytoplankton, and more specifically for Microcystis, is still needed . Our study addresses two basic questions. Firstly, how does the vulnerability of M. aeruginosa to grazing by D. polymorpha vary with colony and mussel size? Secondly, how do these size-vulnerability relationships compare to size distributions of M. aeruginosa colonies in nature? The latter comparison may help to explain variable effects of D. polymorpha invasion on M. aeruginosa biomass. Size-sele ctive grazing by D. polymorpha on M. aeruginosa, if intense, might also shift the size distribution of M. aeruginosa colonies toward less vulnerable size classes. This could lead to M. aeruginosa populations dominated by la rge colonies, which should be more likely to accumulate at the surface as a scum (possibly elevating exposure risk to more concentrated doses of toxins) due to faster vertical migration ve locities (Visser et al. 1997). In general, knowledge of grazing impacts on M. aeruginosa as a function of both colony and mussel size would improve our understanding of the complex interaction with D. polymorpha and the role of large variation in colony size in the population dynamics of this HAB species (Burkholder and Glibert 2009, Pitcher 2012). Methods Size-selectivity of D. polymorpha was assayed via a ‚particle- choice™ experiment in which eight size classes of a single, palatable strain of M. aeruginosa (2009C) were individually presented to mussels along with Ankistrodesmus falcatus (mean cell dimensions: 38.5 × 2.5 µm). 10 A. falcatus was employed as a high-quality alga against which selectivity for M. aeruginosa size fractions could be assessed (White et al. 2011). Si ze selectivity as a function of both colony and mussel size was measured in a second experiment in which three size classes of mussels were fed three size classes of a singl e, palatable strain of M. aeruginosa (G11-08). In the second experiment, we used a simpler ‚no choi ce™ design because filtration rates on M. aeruginosa were found to be strongly correlated with selectivity (M. aeruginosa versus A. falcatus ) in previous experiments (White et al., 2011; this study). To relate measur ed size-selectivities to a natural population of M. aeruginosa, we also report the results of extensive monitoring of M. aeruginosa size distributions in Gull Lake, MI (2001-2011), the source of the experimental mussels and M. aeruginosa strains. Collection and maintenance of experimental organisms Gull Lake is a large (area: 822 ha, max depth: 33 m), low-nutrient lake (TP ~10 g L-1, mixed layer chlorophyll a ~1-6 g L-1) in southwest MI (Barry and Kalamazoo Counties). The biomass of M. aeruginosa increased dramatically after zebra mussels invaded in 1994 (Sarnelle et al. 2005). Gull Lake has been the site of large-scale enclosure experiments (Sarnelle et al. 2005, Sarnelle et al. 2012) aimed at unders tanding the complex interaction between D. polymorpha and M. aeruginosa. Mussels were collected with an Ekman grab and immediately sepa rated from substrata (macrophytes and small rocks). Once collected, musse ls were gently scrubbed free of attached material, acclimated to room temper ature in the lab, and fed a diet of A. falcatus (~4 g L-1 chlorophyll a). We selected animals with shell lengt hs of 16-21 mm for the first experiment because this size class was used in the aforemen tioned enclosure experiments. Based on a survey 11 of the D. polymorpha population conducted in 1999 (Wilson and Sarnelle 2002), this size class accounts for ~30% of the total Dreissena filtering impact on phytoplankton in the mixed layer (mean depth ~7.5 m) of Gull Lake during summer stratification (Fig. 3). More generally, the 16-21 mm size class is representative of the larger individuals commonly found in invaded inland lakes (Horgan and Mills 1997, Idrisi et al. 2001, Naddafi et al. 2007). For the second experiment, we sorted mussels into three size classes (8-11, 17-20, and 25-28 mm shell length) that collectively account for ~70% of th e total filtering impact in the mixed layer of Gull Lake (Fig. 3). We selected two strains of M. aeruginosa isolated from Gull Lake that were: 1) still forming colonies in culture, and 2) shown to be highly palatable to mussels (i.e., elicited filtering rates comparable to A. falcatus in previous experiments (White et al. 2011), unpublished data). The strains were maintained in batch cultures of 0.5× WC-S medium, with an inoculum transferred to fresh, sterile medium every 4-8 weeks (White et al. 2011). A. falcatus was grown in semi-continuous culture in full-strength WC-S medium, with gentle aeration and stirring. All phytoplankton were grown on a 12:12 h light:dark cycle under fluorescent lights (70 µmol m -2 s-1) at 20 °C. Before each experiment, we grew 25 L of M. aeruginosa in 1 and 2 L bottles of 0.5× WC-S, under the same growth conditions descri bed above. The large volume was required to yield sufficient biomass in all size classes fo r the feeding experiments. Bottle position was randomized every 2 d to reduce heterogeneity in light conditions during growth. Cultures were harvested for the experiment 30 d after inocul ation, at which time they were still growing exponentially (J. White, pers. obs.). 12 Figure 3. Estimated Dreissena polymorpha filtering impact on phytoplankton in the mixed layer of Gull Lake (0-7.5 m depth). Size distribution data were obtained from Wilson and Sarnelle (2002). Filtering impact was calculated using the D. polymorpha length-filtering rate regression of Kryger and Riisgård (1988). 13 Twenty-four hours prior to an experiment, the M. aeruginosa cultures were pooled into a single container and mixed thoroughly. Two liters we re set aside for controls and mussel pre- acclimation (24 h) to the M. aeruginosa strain (White et al. 2011). The remaining ~23 L was sequentially passed through sieves of decreasing pore size to gently size-fractionate the colonies (Table 1). Material collected was thoroughly rins ed from each sieve and re-suspended in filtered Gull Lake water (1 µm nominal pore size; here after, ‚filtered lake water™). Size-fractioned M. aeruginosa was stored in the dark overnight before the experiments. Design of feeding experiments Each feeding suspension contained one size class of M. aeruginosa plus A. falcatus in filtered lake water (first expe riment: mean total chlorophyll a = 3.9 µg L-1) or just one size class of M. aeruginosa (second experiment: mean chlorophyll a = 2.3 µg L-1). Chlorophyll levels were within the range of mixe d-layer conditions in Gull Lake. In the first experiment, we employed 4 replicate beakers with mussels for each of eight size-class treatments, a nd a total of 3 control beakers containing un-fractionated M. aeruginosa without mussels. The second experiment consisted of three levels of mussel size (8-12, 17-20, and 23-27 mm shell length) crossed with three size classes of M. aeruginosa (Table 2), with three replicat e beakers of each treatment combination. We also employed two repli cate beakers per mussel size class of an A. falcatus suspension (mean chlorophyll a = 3.2 µg L-1). 14 Table 1. Median sizes of Microcystis aeruginosa colonies (Gull Lake strain 2009C) in the eight feeding suspensions used in first experiment. See text for a description of the size metrics. Mesh sizes (lower, upper; µm) Median equivalent diameter ( µm) Median maximum linear dimension ( µm) --, 8* 4.8 4.8 24, 35 21.5 29.6 35, 45 30.3 41.5 45, 53 34.8 46.8 53, 63 44.3 56.7 63, 80 62.6 75.4 100, 150 80.1 109.3 150, 200 87.7 120.8 * Treatment composed of single cells. 15 Table 2. Median sizes of Microcystis aeruginosa colonies (Gull Lake strain G11-08) in the three feeding suspensions used in the second experiment. See text for a description of the size metrics. Mesh sizes (lower, upper; µm) Median equivalent diameter ( µm) Median maximum linear dimension ( µm) 35, 53 31.5 38.9 53, 100 51.7 76.1 100, -- 75.1 100.6 16 Mussels were held in filtered lake water for 3 h immediately prior to the experiments to cleanse their guts of assimilated food material an d then allocated in pairs to 1 L glass beakers containing 0.9 L of feeding suspension (Fig. 4). Beakers were gently aerated, which kept phytoplankton in suspension throughout the experiments. The feeding period began once mussels were actively filtering (siphons fully extended), wh ich generally occurred within 5 minutes of placement into beakers. Mussels fed for 1 h (f irst experiment) or 1.5 h (second experiment). Beakers were mixed thoroughly and then sampled for chlorophyll a (filtered onto Pall A/E glass-fiber filters and frozen) and algal counts (first experiment only, preserved in 1% Lugol™s iodine) immediately before mussels were added (initial ) and immediately after mussels were removed (final). In the second experime nt, each feeding suspension was sampled for measurements of colony size (preserved in 1% Lugol™s iodine) prior to allocation to beakers. Sample processing and data analysis For the first experiment, Lugol™s-preserved subsamples were settled in 10 mL phytoplankton chambers. Ankistrodesmus cells were counted at 200× with an inverted microscope (Nikon Eclipse; (Lund et al. 1958, Sandgren and Robinson 1984). The surface area and maximum linear dimension (MLD) of each M. aeruginosa colony (n ~ 25-50 colonies per sample) were measured (at 100× or 200×, depend ing on size class) from digital micrographs (SPOT, Diagnostic Instruments). Total surface area in a sample was converte d to cell density via a regression developed for Gull Lake M. aeruginosa (Sarnelle et al. 2012). The equivalent diameter (ED) of each colony was also determined by calculating the diameter of a circle with surface area equivalent to that of each measured colony. Microcystis cells in the single-cell treatment were counted and measured at 400×. 17 Figure 4. View of a laboratory feeding experiment w ith zebra mussels. Photo credit: Jeffrey D. White. 18 For the first experiment, we calculated speci es-specific filtering rates (L individual-1 d-1; i.e., mortality rates due to mussel grazing) usi ng the particle depletion method (Omori and Ikeda 1984), FlnCilnCftVN, where Ci and Cf are initial and final cell densities (cells L -1) respectively, t is the length of th e feeding period (d), V is the volume of th e feeding suspension (L) and N is the number of mussels per beaker. No correction for changes in the control beakers was needed because no changes in cell density oc curred during these short incubations (paired t-tests, p > 0.4). Mussel selectivity was determined for each beaker as Fm/Fa, where Fm and Fa are the filtering rates on the M. aeruginosa size class and A. falcatus, respectively (Jacobs 1974, Sterner 1989). For the second experiment , initial colony size of M. aeruginosa in each of the three suspensions was determined as above. Filtering rates on the unialgal suspensions were measured as the rate of chlorophyll a depletion over time as above. Chlorophyll a was measured via dark extraction of A/E filters in cold 95% ethanol fo r 24 h, followed by fluorometric analysis with a Turner Model 10-AU-005 fluorometer (Welschmeyer 1994). Statistical analyses were performed with R version 3.0.1 (R Foundation for Statistical Computing). We tested for differences in mean mussel selectivity across M. aeruginosa size classes in the first experime nt with a one-way ANOVA. In the event that this ANOVA was significant, we performed 1-tailed t-tests to check for signi ficant avoidance of M. aeruginosa (selectivity < 1) for each size class. For treatmen ts where mussels exhibite d significant selection against M. aeruginosa, we further tested whether mussel consumption of M. aeruginosa was > 0 (1-tailed t-tests). We used one-tailed tests because the a priori expectations are for mussels to prefer A. falcatus when they are selective (since A. falcatus is consumed at maximal rates), and to have non-negative filtering rates (White et al. 2011). In the second experiment, we tested for a 19 significant interaction between muss el body size and colony size of M. aeruginosa on mussel filtering rate using a two-way ANOVA. Since per capita filtering rate increases with body size in D. polymorpha (Kryger and Riisgård 1988), filtering rates were standardized to mussel mass (Wilson and Sarnelle 2002). Mass-specific filt ering rates were log-transformed to reduce heterogeneity of variances. Lake sampling Phytoplankton were sampled from the mixed layer of Gull Lake biweekly from July- August in 2001 and 2005-2008, and from June-Sep tember in 2009-2011, via two pooled casts of a depth-integrating tube sampler. Samples were collected from a near-shore station (depth = 13 m) during 2001 and 2005-2008. In 2009-2011, sample s were collected from the near-shore station as well as a central sta tion (depth = 30 m). Subsamples were preserved in 1% Lugol™s iodine and settled in 10 mL phytoplankton chambers. Colony ED of M. aeruginosa was determined as described above. Median colony size did not significantly differ between sampling stations in 2009 (paired t-test, t= 0.87, df = 7, p > 0.4), 2010 (paired t-test, t= 0.39, df = 8, p > 0.7) or 2011 (paired t-test, t= 0.0019, df = 4, p > 0.9); therefore, data were pooled across stations for those three years. Results Experiment 1: colony size-selectivity experiment Size fractionation of the M. aeruginosa culture produced a range of colony sizes spanning more than an order of magnitude across 8 treatments. Median MLD in each fraction corresponded to the mesh sizes employed more than median ED (Table 1), although we report 20 both to facilitate comparisons acr oss studies using different metric s. Initial total algal biomass did not significantly differ across treatments (ANOVA, F (7,23) = 1.58, p = 0.190). Mussel selectivity in one beaker was a large, unexplained outlier (studentized residual = 3.97) and this datum was omitted from all analyses, leaving a total n = 31 for the experiment. Mussel selectivity for M. aeruginosa differed significantly among all 8 size treatments (ANOVA, F(7,23) = 4.85, p = 0.002; Fig. 5a). Dreissena exhibited significant selection against M. aeruginosa (selectivity < 1; 1-tailed t-tests, t < -2.45, df = 3, p < 0.05) in the two treatments having median colony sizes 80.1 µm ED (109.3 µm MLD), whereas selectivity for M. aeruginosa within the remaining, sma ller fractions was not different from 1 (indicating non-selective feeding), consistent with a previous study that found strain 2009C to be highly palatable to D. polymorpha (White et al. 2011). Congruent with the selectivity results, mussel filtering rates on M. aeruginosa varied significantly across treatments (ANOVA, F (7,23) = 4.59, p = 0.002; Fig. 5c) and were not significantly greater than 0 in treatments with colonies 80.1 µm median ED (1-tailed t-tests, t > 0.21, df = 3, p > 0.09). Conversely, filtering rates on A. falcatus did not significantly vary among treatments (ANOVA, F (7,23) = 0.34, p = 0.929; Fig. 5b). All Microcystis colonies observed in the post-feeding samples appeared healthy and exhibited no signs of damage as a possible result of collection and rejection by D. polymorpha. Experiment 2: mussel size × colony size experiment There was no mussel size × colony size interaction (2-way ANOVA, F(4,18) = 2.13, p = 0.119; Fig. 6), and no effect of M. aeruginosa colony size on filtering rate for any mussel size class over the tested range of colony sizes (F(2,18) = 2.46, p = 0.114). Thus, we found no evidence for size-selective feeding by any of the three mussel size classes in this experiment. 21 Figure 5. Results from experiment 1, the co lony size-selectivity experiment. 22 Figure 5. (cont™d) (A) Mean selectivity of Dreissena polymorpha across size classes of a single strain (2009C) of colonial Microcystis aeruginosa (n = 31). Selectivity was calculated as filtering rate on M. aeruginosa divided by filtering rate on the high-quality food alga, Ankistrodesmus falcatus. Error bars represent ± 1 SE. Asterisks indicate size fracti ons eliciting significant avoidance (selectivity significantly < 1). (B, C) Compar ison of mean filtering rates on A. falcatus and M. aeruginosa, respectively. Error bars represent ± 1 SE. An N indicates non-detectable filtration (filtering rate not significantly > 0). Median Microcystis size is reported as both equivalent diameter (ED) and maximum linear dimension (MLD). 23 Figure 6. Results from experiment 2, the mussel size × colony size experiment. Mean mass- specific filtering rates of three common size classes of Dreissena polymorpha (shell length ranges given in mm) feeding on a size-fractioned strain of Microcystis aeruginosa (n = 27). Error bars represent ± 1 SE. Median M. aeruginosa colony size is reported as both equivalent diameter (ED) and maximum linear dimension (MLD). 24 Dreissena filtering rates on M. aeruginosa were similar to or greater than those on A. falcatus (Fig. 6), so the lack of evidence for size-selectivity was not due to lack of overall filtering on this M. aeruginosa strain. Small mussels have higher mass-specific filtering rates than larger mussels due to allometric scaling and so, not surprisingly, we also observed a significant effect of mussel size on mass-specific filtering rate (F (2,18) = 41.26, p < 0.001). One A. falcatus replicate (within the 8-11 mm mussel class) was omitted from analys is because mussels did not feed in that beaker. Lake sampling The overall distribution of Gull Lake colony sizes, pooled from 8 summers of data, was right-skewed with a median of 79.5 µm ED (Fig. 7, n = 3,035). However, there was considerable inter- and intra-annual variability in colony size within the Gull Lake population. In fact, median colony size decreased from July through August in every year, by as much as 27%, with July median ED being significantly greater than August median ED across all 8 years (paired t-test, p < 0.001; Fig. 8). In 2009- 2011, years where sampling was extended, the decrease in colony size over the summer was especially ap parent (Fig. 9). Most importantly, median colony ED tended to be above the mussel selectivity threshold iden tified in the colony size-selectivity experiment (~80 µm) early in the summer, but decreased to less than or equal to the threshold later in the summer, with the overall median tending to almo st exactly the same size across years (Figs. 7, 8, 9). 25 Figure 7. Size structure of the Microcystis aeruginosa population in Gull Lake: overall colony size distribution (equivalent diameter, ED) obtai ned from pooling 8 summers of measurements (n = 3,035). 26 Figure 8. Size structure of the Microcystis aeruginosa population in Gull Lake: comparison of July and August colony size distributions by year. The horizontal dotted line at 80 µm ED is the approximate threshold above which M. aeruginosa is largely invulnerable to grazing by Dreissena polymorpha (see Fig. 5). 27 Figure 9. Seasonal change in median Microcystis aeruginosa colony size (equivalent diameter, ED) in Gull Lake during three consecutive summers. The horizontal dotted line at 80 µm ED is the approximate threshold above which M. aeruginosa is largely invulnerable to grazing by Dreissena polymorpha (see Fig. 5). 28 Discussion We identified a Microcystis colony-size threshold of ~80 µm median ED (~109 µm median MLD) for 16-21 mm D. polymorpha, where filtration rates essentially fell to zero. Smaller colonies and single cells were consum ed non-selectively and at maximal rates (Kryger and Riisgård 1988, White et al. 2011 ); Fig. 5). Previous experime nts measuring the grazing of D. polymorpha on M. aeruginosa have employed single-celled strains (Baker et al. 1998, Dionisio Pires and Van Donk 2002, Dionisio Pires et al. 2004) or coarse ‚small™ versus ‚large™ size fractions (Vanderploeg et al. 2001, Dionisio Pires et al. 2005), and so do not clearly identify an upper size threshold for effective D. polymorpha feeding. In the latter case, where mussel filtering rates were assessed on aggregate fractions greater than and less than an arbitrary size (e.g., 53 and 60 µm, respectively), a significant proportion of the colonies in the larger fraction may still have been within the edible range, given th at the size distributions of colonies in culture are generally highly right-skewed (J. White, pers. obs.). This may explain the lack of differences observed between filtering rates on colonial Microcystis in the < 60 µm and > 60 µm fractions (overall range: 41-722 µm) in Dionisio Pires et al. (2005). Jarvis et al. (1987, 1988) quantified the size-dependent mortality imposed on a natural M. aeruginosa population by Daphnia, among the few previous grazing studies on M. aeruginosa to more systematically account for the effect of natural size variation. Our study goes further in eliminating all biological properties that could be confounded with size by utilizing single, highly palatable strains of M. aeruginosa grown under a single set of environmental conditions. Thus, the variation in selectivity and filtering rate that we observed (Fig. 5) can be unequivocally attributed to variation in colony size. 29 Our results are similar to some previous studies of D. polymorpha size-selective feeding, although there is uncertainty in comparing our results to the literature because of the aforementioned species-for-size confound in many studies. D. polymorpha (15-30 mm) have been shown to clear particles as small as 5 µm, equivalent in si ze to single cells of M. aeruginosa (Sprung and Rose 1988, Dionisio Pires et al. 2004). Also congruent with our study, Ten Winkel and Davids (1982) observed preferential ingestion by D. polymorpha (25 mm) of diatoms averaging 10-40 µm, but rejection of larger diatoms and chrysophytes > 80 µm. However, mussels (12-20 mm) were observed to largely reject phytoplankton > 50 µm and < 7 µm in another study (Naddafi et al. 2007); in still another study, mussels (18- 22 mm) preferred natural seston in a 30-100 µm fraction over the 2-30 µm fractions (Dionisio Pires et al. 2004). Furthermore, Horgan and Mills (1997) found no diffe rences in filtering rates of small mussels (9- 11 mm) across size classes of natu ral phytoplankton ranging from 10-150 µm MLD, considerably higher than the threshold M. aeruginosa colony size identified for 16-21 mm mussels in the present study. Perhaps the rather inconsistent picture of D. polymorpha's size selectivity that emerges from the literature is in part a function of the use of different phytoplankton species as surrogates for variation in particle size. With respect to models of size-structured grazing vulnerability, quantification of mortality rates imposed on di fferent size fractions of M. aeruginosa is also more useful, compared to measurements of the maximu m particle size collected by an individual D. polymorpha, since mortality rates are needed for predicting population dynamics . Though D. polymorpha has been shown to collect phytoplankton as large as 0.9-1.2 mm, these particles are usually rejected and expelled shortly thereafter (Horgan and Mills 1997) and so are likely to experience negligible grazing mortality (particle ‚collection™ is not synonymous with particle 30 ‚ingestion™). Collect ed phytoplankton that are expelled by D. polymorpha in this manner (as pseudofeces) generally remain viable and can retu rn to the water column (Vanderploeg et al. 2001). Dreissena also filtered the same strain (2009C) empl oyed in our first experiment at near maximal rates in a previous study (White et al. 2011), although the strain was not size-fractionated in that experiment. Median colony size in that experiment was 85.3 µm ED, slightly higher than the threshold size for consumpti on found in the present study. This apparent discrepancy may arise because a large fraction of the colonies in the previous experiment were within the edible size range, since all sizes below 85 µm were present. The range of colony sizes within each size fraction in the present study was very narrow by designŠcritical for precisely quantifying grazer size-selectivity. Dreissena also consumed colonies (75.1 m median ED) during the second experiment near to, though still be low, the size threshold identified in the first experiment (80.1 m ED; Figs. 5, 6). However, the M. aeruginosa strains employed in the two experiments were different and so may have differed subtly in morphology. In addition, we note that the variability in D. polymorpha selectivity and filtering on M. aeruginosa was relatively large for the two largest size classes in the firs t experiment (Fig. 5). This high variability may signal that minor variations in size distributions near the size threshold may have major impacts on filtering rate. Thus, our estimate of the upper size threshold should be interpreted as somewhat fuzzy, although it is considerably more definitive than can be gleaned from the existing literature, especially with re spect to this important HAB species. Counter to previous expectations based on shell length-siphon diamet er relationships for D. polymorpha (MacIsaac et al. 1991, Naddafi et al. 2007), we found no ev idence for an effect of mussel size on M. aeruginosa vulnerability to grazing (Fig. 6). Inhalant siphon diameter in 16 31 mm D. polymorpha is generally 20 times wider than the M. aeruginosa colonies that were too large to be consumed in our first experiment (MacIsaac et al. 1995)(Fig. 5), suggesting that siphon diameter is not an important determinant of size selectivity. Our experiment appears to be the only direct measurement of phytoplankton size -based vulnerability to grazing as a function of D. polymorpha body size (Naddafi et al. 2007), and we employed a fairly wide range of body sizes (8-11 to 25-28 mm) relative to the body-size variation present in a natural system (Fig. 3). Consistent with our results, MacIsaac et al. (1995) found no systematic differences among D. polymorpha size classes (range: 10-22 mm) in their ab ility to ingest microzooplankton (mean size: 89 m). Conversely, Bridoux et al. (2010), quantif ying phytoplankton biomarkers in Lake Erie D. polymorpha tissue, found indirect evidence for increased importance of larger phytoplankton taxa in the diet of larger mussels. Taken together, our two experiments indicate that D. polymorpha between ~8 and ~28 mm in shell length consume partic les non-selectively from 5-75 µm ED. This higher upper threshold relative to filter-feeding zooplankton (e.g., Daphnia) indicates that standard demarcations of ‚edible™ vers us ‚inedible™ in phytoplankton studies (McCauley and Briand 1979, Cyr and Pace 1992) are not suitable for assessing phytoplankton vulnerability to Dreissena grazing. Rather, given that D. polymorpha can filter particles as small as 0.4-1.5 µm (Cotner et al. 1995, Lei et al. 1996) and the majority of fr eshwater phytoplankton biomass typically falls below 75 µm (Sprules et al. 1983), size may be generally less important relative to other factors in determining overall phyt oplankton vulnerability to Dreissena grazing, as compared to Daphnia. By putting our experimental data (Figs. 5, 6; White et al. 2011) into context and comparing them to the size distribution of M. aeruginosa in nature (Figs. 7, 8, 9), it seems 32 apparent that much of the M. aeruginosa population in Gull Lake is vulnerable to D. polymorpha grazing. If anything, the consistent seasonal decline in median colony size of M. aeruginosa from above to below 80 µm ED (Figs. 8, 9) indicates an over all increase in vulnerability during the growing season on the basis of size, and is contrary to the expectation that intense size- selective grazing by D. polymorpha would increase the proportion of large, less edible colonies. However, the relative abundance of different genotypes and chemotypes within M. aeruginosa populations is also known to shift within a seas on (Saker et al. 2005, Kardinaal et al. 2007a, Welker et al. 2007, Rinta-Kanto et al. 2009, Bozarth et al. 2010), and different genotypes of M. aeruginosa from Gull Lake can vary maximally in their vulnerability to mussel grazing irrespective of colony size within the edible range (White et al. 2011). Thus, the late-summer population, though composed of smaller colonies , may be dominated by genotypes that are grazing resistant due to other factors (e.g., chemical inhibitors). Clearly, ‚vulnerability™ should always be considered in relative terms; it is a potentially variable and dynamic attribute of M. aeruginosa populations. Although small colonies of the two Gull Lake strains we employed in our experiments were fully vulnerable to grazing, the lake population consists of other strains (White et al. 2011) and size fracti ons (this study) that are invulnerable. Variation in both colony size distributions and genotypic composition within and among M. aeruginosa populations could help to explain the dramatic differences in response of M. aeruginosa biomass to D. polymorpha invasion across systems (Smith et al. 1998, Vanderp loeg et al. 2001, Raikow et al. 2004, Knoll et al. 2008, Sarnelle et al. 2010). Size is a ‚master trait™ (Litchman a nd Klausmeier 2008) driving phytoplankton population processes and also factors into impor tant ecological trade-offs. Large phytoplankton, including large colonies of M. aeruginosa, generally have lower per capita growth rates than 33 smaller cells or colonies (Kruk et al. 2010, Wils on et al. 2010), and are also less efficient at assimilating resources (Reynolds 2006). Large size can also confer benefitsŠfor example, faster migration rates for buoyant Microcystis (Visser et al. 1997, Reynolds 2006) and, of course, lower vulnerability to consumption. Si nce variation in colony size is a characteristic trait of M. aeruginosa of clear ecological im portance, models of M. aeruginosa population dynamics should incorporate size-structured growth and loss processes, in a way that is analogous to the age-structured models of animal populations, in or der to more accurately forecast harmful algal blooms. Acknowledgements We thank J. Berry, J. Chiotti, L. Dillon, T. Geelhoed, B. Hanna, G. Horst, M. Iadonisi, C. Kissman, C. Kozel, K. Lincoln, E. Milroy, J. Nort hrop, D. Raikow, N. Sarnelle, T. Sarnelle, A. Schuerer, T. Toda, A. Wilson, D. Weed, and J. White for assistance in the field and lab. N. Consolatti, M. Williams, and S. Bassett provided tremendous logistical support. S. Hamilton and P. Soranno suggested improvements to the study desi gn and an earlier version of the manuscript. The comments of two anonymous reviewers were greatly appreciated. This research was funded by the Environmental Protection Agency (E cology and Oceanography of Harmful Algal Blooms/2004-Science to Achieve Results-C 1, project RD83170801), the National Science Foundation (Division of Environmental Bi ology-0841864, Division of Environmental Biology- 0841944), and Michigan State University (Summer Fellowship to J. White). 34 CHAPTER 2 HEAT-INDUCED MASS MORTALITY OF INVASIVE ZEBRA MUSSELS ( DREISSENA POLYMORPHA) AT SUBLETHAL WATER TEMPERATURES Abstract We observed a massive die-off of zebra mussels ( Dreissena polymorpha) on the epilimnetic sediments of Gull Lake (Michigan, USA) during the relatively warm summer of 2010, even though water temperatures were belo w widely-reported lethal temperatures of 30 °C. We followed up this obser vation with four years of in situ monitoring of caged mussels stocked across a depth-temperature gradient in Gull Lake. Mortality of caged D. polymorpha was largely explained by accumulated degree hours > 25 °C, a temperature threshold that is considerably lower than laboratory- derived lethal temperatures for D. polymorpha. We also assessed both the acute and chronic thermal tolerance of Gull Lake D. polymorpha with laboratory experiments, which confirmed higher acute tolerance (up to 32 °C) under otherwise ideal conditions but high susceptibility to prolonged exposure to fisublethalfl temperatures (exceeding 1,700 degree hours > 25 °C) as occurred in Gull Lake during the die-off. Our results indicate that the thermal tolerance of D. polymorpha under natural conditions may be lower than has been reported from laboratory studies. Lower temperature tolerance may have major implications for the dynamics, impacts, and manage ment of this invasive species given future climate change scenarios. Introduction Zebra mussels ( Dreissena polymorpha) are among the most successful and impactful aquatic invasive species in North America and Western Europe (Strayer 2008, Higgins and 35 Vander Zanden 2010). Great effort and expense are invested to understand, predict, and prevent their spread. Habitats that are most susceptible to invasion have been identified on the basis of environmental conditions and known physiological tolerances of mussels, with temperature playing an important role in delineating the potential range of D. polymorpha (Strayer 1991, Drake and Bossenbroek 2004). Given the profound impacts D. polymorpha has in invaded habitats, understanding its thermal sensitivity is important for present-day management, as well as to predict responses under future climate change scenarios (Thorp et al. 1998, Griebeler and Seitz 2007, Rahel and Olden 2008). Climate for ecasts include more frequent extreme temperature events and a 2 °C increase in mean air temperature for much of temperate North America by 2050 (Romero-Lankao et al. 2014). Obse rvations of temperate lakes have already identified responses to shorter- and longer-term climatic variation in the past few decades, including warmer water temperatures and in creased intensity and duration of thermal stratification (Schindler et al. 1990, Jankowski et al. 2006). Temperature tolerance of North American D. polymorpha has been extensively studied in the laboratory, but reported upper limits and requisite exposure times are highly variable. Lethal temperature thresholds for D. polymorpha have been reported between 29-32 °C (Garton et al. 2014), with the time to death decreasing from 4 da ys to 5 minutes as temperature increases from 30 to 43 °C (Iwanyzki and McCauley 1993, McMahon et al. 1994, McMahon and Ussery 1995, Elderkin and Klerks 2005, Beyer et al. 2011). In general, 30 °C is widely accepted as the fiupper incipient lethal temperaturefl for D. polymorpha in North America (Iwanyzki and McCauley 1993). These temperature thresholds have been ob tained largely from short-duration (minutes- days) laboratory studies that expose mussels to temperature regimes that are extreme relative to nature, typically to identify suitable, acutely lethal temperatures for rapidly purging industrial 36 intakes and boat hulls of fouling mussels. In contrast, mussels ha ve been shown to survive as long as 14-35 days at 30-32 °C when gradually acclimated (McMahon et al. 1994, Aldridge et al. 1995, Spidle et al. 1995). However, over prolonged periods (> 30 days) at elevated temperatures, D. polymorpha exhibit metabolic imbalance and negativ e somatic growth, even under otherwise ideal laboratory conditions (Aldridge et al. 1995). We observed a rapid and large (approaching 100%) mortality even t of all size classes of D. polymorpha on epilimnetic sediments in Gull Lake , Michigan (USA) in early August 2010, during a relatively warm summer (Fig. 10) although water temperatures never exceeded 29 °C at 1 m depth. Mussels survived at greater depths (lower epilimnion and metalimnion) where temperatures were 25 °C. Other variables of known importance to D. polymorpha (e.g., chlorophyll-a, dissolved oxygen, calcium, and pH) (Garton et al. 2014) were all within their typical ranges for the 16 years since D. polymorpha appeared in Gull Lake (J. White, unpubl.). Though the summer of 2010 was warm, it was neither the only warm summer nor the warmest since D. polymorpha first invaded the lake (Fig. 11), ye t no mass-mortality events of this magnitude had been observed in Gull Lake previous ly. This suggested that if temperature played a role in the 2010 die-off, its effect was likely more complex than would be predicted by simple consideration of median or maximum temperatures. 37 Figure 10. (A) Recently expired Dreissena polymorpha retrieved from the nearshore area of Gull Lake immediately following the 2010 epilimnetic mass die-off event. (B) Shells of dead D. polymorpha cover the bottom of Gull Lake at a de pth of 2 m in 2011, following the 2010 mass die-off event. Photo credits: (A) Stephe n K. Hamilton; (B) Jeffrey D. White. A) B) 38 Figure 11. Box plots of mean daily June-August air temperatures recorded at the Kellogg Biological Station, adjacent to Gull Lake, for the years after Dreissena polymorpha invaded Gull Lake. The die-off occurred during the shaded summer (2010), and the dotted line indicates the median air temperature (22.4 °C) for that summ er. Whiskers extend to the lowest and highest non-outliers; open circles indicate outliers. 39 There are other reports in the literature of large D. polymorpha mortality events that have been associated with high water temperatures, though to our knowledge these observations have been exclusive to the southernmost limits of th eir range. In the southern Mississippi River (Baton Rouge, Louisiana), water temperatures peak at 29-30 °C for several weeks during summer, with mortality of adult D. polymorpha reaching 60% in situ (Allen et al. 1999). Large die-offs have also been documented in Oklahoma reservoirs when maximum summer water temperatures reached 30 °C (Boeckman and Bidwell 2014). These in situ observations are c onsistent with the aforementioned experimentally-derived toleran ce estimates, though tolera nce is known to vary geographically, with populations in the southern Mississippi River drainage being more tolerant of warmer water temperatures than north-te mperate populations (Elderkin and Klerks 2005), possibly explaining why Gull Lake mussels might be vulnerable at relatively lower temperatures. In any case, the observed die-off in Gull Lake is not consistent with either existing lab-derived acute mortality data or field obser vations from southern populations. Based on these preliminary observations from Gull Lake, we hypothesized that sustained temperatures that are below the range of publis hed acute lethal temperatures lead to high mortality of zebra mussels. To assess the infl uence of these fisublethalfl temperatures on D. polymorpha mortality, we 1) conducted four years of in situ monitoring of caged D. polymorpha in Gull Lake (2011-2014), 2) analyzed D. polymorpha mortality data from enclosure experiments conducted in Gull Lake (2005-2008), and 3) perfor med laboratory experiments testing both the acute and chronic thermal tolerance of Gull Lake D. polymorpha. 40 Methods Study site Gull Lake is a large (822 ha), deep (33 m maximum depth, 7.7 m mean mixed-layer depth), oligotrophic hardwater lake located in Barry and Kalamazoo Counties in southwestern Michigan. Dreissena polymorpha was first observed in Gull Lake in 1994. A mean density in the littoral zone (2.5-7 m dept h) of 2,193 individuals m-2 was estimated in 1999 (Wilson and Sarnelle 2002, White and Sarnelle 2014). We made annual observations of littoral D. polymorpha during their post-establishment period in Gull Lake, and we did not observe any mass-mortality events prior to 2010 (S. Hamilton, pers. obs.). Howe ver, one enclosure experiment in 2005 was compromised by unacceptably high mortal ity of enclosed mussels (> 30%). In situ mortality: caged mussels We measured mortality of D. polymorpha in each of the four consecutive years (2011-2014) following the initial die-off event in Gull Lake. Mussels were harvested in early May with an Ekman grab from the lower epilimnion (5 m, where they survived) and stocked into cages built of rigid plastic mesh (16 cm diameter, 13 cm tall, mesh size 3.5 mm). In all years, large mussels (16-30 mm) were stocked into every cag e, with either 20 (2011-2012) or 12 (2013-2014) individuals per cage. In 2012-2014, small mussels (8-15 mm) were also stocked, either into separate cages (20 individuals each, 2012) or together with the large mussels (12 individuals each, 2013-2014). Cages were deployed in replicates of four at three depths: 2 m (upper epilimnion), 5 m (lower epilimnion), and 9 m (met alimnion). These depths span the range where D. polymorpha had previously been abundant in Gull Lake and were selected based upon inspection of temperature profile s, to provide a range of ambi ent temperatures. Cages were 41 anchored at the bottom and tethered to a surf ace buoy (Fig. 12a). Submersible loggers (Onset Corporation) recorded water temperatures at hour ly intervals just above the sediments at each depth. Cages were briefly pulled to the surface approximately weekly for inspection of stocked D. polymorpha. Cages were retrieved in October and a final mortality assessment was made. Storms in 2011 and 2014 dragged and filled the 5 m cag es with sediment, resulting in substantial mortality of enclosed mussels, so data from these cages were omitted from all analyses. In situ mortality: mussels in enclosures To assess whether mortality of Gull Lake D. polymorpha varied as a function of the wide range of temperature conditions present in the years prior to the 2010 die-off (Fig. 11), we analyzed mortality data from the summers of 2005-2008 collected during a series of enclosure experiments in Gull Lake. These experiments were designed to test the interactive effects of D. polymorpha and nutrients on the biomass of the harmful cyanobacterium, Microcystis. Experimental details can be found elsewhere (Ham ilton et al. 2009, Sarnelle et al. 2012, Horst et al. 2014). In all cases, Gull Lake mussels were ha rvested and stocked into the same plastic cages as described above and suspended within large, tubular polyethylene enclosures (diameter = 2 m, depth = 8 m, volume = 25,000 L; Fig. 12b) at a de pth of 2.5 m (always within the mixed layer). Experiments ran for 27-45 days during July and August, and concluded with an assessment of mussel mortality within the enclosures. We found no effect of nutrient treatments on mortality (ANOVA, p = 0.96, n = 21), so data from all enclosures were pooled for each experiment. Continuous water temperature data are not avai lable for Gull Lake during this time period. Instead, we used continuous KBS LTER air temperature data (http://lter.kbs.msu.edu/datatabl es/7, datatable KBS002-001.27) as a rough proxy for mixed layer 42 Figure 12. Monitoring of in situ zebra mussel mortality in Gull Lake, Michigan. (A) View of tethered cages in position on the lake bottom at a depth of 2m. (B) View of the 2008 enclosure experiment. Mussels were suspended within the enclosures inside cage trees similar to those visible on the platform at center righ t. Photo credits: Jeffrey D. White. A) B) 43 water temperature, since daily water temperature at a depth of 2 m in Gull Lake is positively related to air temperature during summer months (Pearson corr elation = 0.80). Experiment 1: acute temperature tolerance We conducted a laboratory experiment during the summer of 2012 to assess the short- term, acute thermal tolerance of Gull Lake D. polymorpha. Six 9.5 L aquaria containing filtered (1 µm nominal pore size) Gull Lake water were nested individually within 38 L aquaria containing deionized water. Submersible aquarium heaters (300 W, Marineland) were installed in 3 of the outer aquaria to create a water bath (heated treatments). A ll aquaria were housed within an incubator maintained at 23 ± 1 °C (control temperature) and a 12:12 hour light:dark cycle (Fig. 13). Temperature was recorded at hourly intervals by submer sible loggers (Onset Corporation) in the 9.5 L aquaria. All aquaria were kept vigorously aerated throughout the experiment. Zebra mussels were harvested for the experi ment as described above, gently scrubbed clean of periphyton, and maintained in the lab as described in White and Sarnelle (2014). Twelve individuals from each of two size classes (8- 15 mm and 16-25 mm shell length) were randomly allocated to each inner aquarium. Mussels were allowed to acclimate for 48 hours prior to the start of the temperature manipulation. Each day thereafter, temperature was increased 1 °C day -1 in heated treatments ( n = 3); controls were continually maintained at 23 ± 1 °C ( n = 3). Mussels were fed a satiating ration (~10-15 µg L-1) of a high-quality, palatable green alga (Ankistrodesmus falcatus) daily (White and Sarnelle 2014). Mortality was monitored daily; mussels were considered dead when gentle probing failed to elicit a shell closure response (Iwanyzki and McCauley 1993, Spidle et al. 1995), and were immediately removed without 44 Figure 13. View of the zebra mussel acute temperature tolerance experiment, showing the nested aquarium water bath design. The entire experiment was housed within an incubator. Photo credit: Jeffrey D. White. 45 replacement. The experiment was concluded wh en mortality reached 100% in heated aquaria (day 16). Experiment 2: chronic temperature tolerance In an effort to experimentally re-create temperature conditions during the mass mortality event of 2010, we exposed Gull Lake D. polymorpha to prolonged elevated temperature in the summer of 2014 using a flow-through system housed in a lakes hore laboratory. The experiment consisted of ten 32 L polyethylene trays (46 × 66 × 15 cm), each fitted with inflow and outflow spigots and a submersible temperature logger (O nset Corporation). A 38 L glass aquarium was nested within each tray, and a submersible aquari um heater (300 W, Marineland) was installed in five of these aquaria to create a water bath (heated treatments, n = 5). Fresh lake water from Gull Lake was pumped into two 1,100 L holding tanks every 3-4 days. This lake water, containing the natural Gull Lake phytoplankton community, was con tinually supplied to the experimental trays with a peristaltic pump (50 L day -1 flow rate, Ismatec/Cole-Parmer). All holding tanks and experimental trays were kept vigorously aerated throughout the experiment (Fig. 14). Mussels were harvested as described previously, and individuals with a shell length of 16-18 mm were used in the experiment. Eleven mussels were then allocated randomly to each experimental tray. This density was selected such that the collective D. polymorpha filtering rate (Kryger and Riisgård 1988) balanced the inflow rate of lake wa ter to the trays. Mussels were allowed to acclimate to experimental trays fo r 24 hours prior to the start of the temperature manipulation (day 0). On day 1, the temperature in heated trays was gradually increased from 22.0 to ~27.0 °C over 12 hours. Unheated trays were maintained at ambient room temperature (mean = 22.8, range = 20.4 - 24.9 °C). Mortality was monitored as described above. The 46 Figure 14. View of the zebra mussel chronic temperat ure tolerance experi ment. Experimental trays with nested water baths are in the foregr ound; lake water holding tanks are visible in the background. The peristaltic pump is at top center. Photo credit: Jeffrey D. White. 47 experiment was ended on day 23, once the number of degree hours achieved in heated treatments was comparable to that observe d in Gull Lake during the 2010 D. polymorpha die-off event and periods of high mortality during in situ monitoring ( see Results). Statistical analyses To explore the influence of accumulated heat exposure on in situ mussel mortality from the caged mussel study, we employed a degree h our approach. To identify the most-likely temperature threshold driving mort ality from prolonged heat exposur e, we performed a series of linear regressions of mortality versus degree hours above a threshold temperature. We started with a threshold of 23 °C, and then sequentially increased the threshold temperature by 1 °C. We assessed the relative fit of each regression model with the Akaike Information Criterion (AIC), where a lower AIC indicates a better fit, and then computed AIC for each model with respect to the model with the lowest AIC. Models having AIC 2 were considered to have equal statistical support as the model with the lowest AIC (Burnham and Anderson 2001). This procedure was performed separately for both the large and small size classes of caged mussels. Proportional mortality data we re arcsine square-root transformed before statistical analysis, but this transformation had no substa ntive effect on any result, so we present untransformed mortality data for ease of inte rpretation. We performed linear regressions of mortality on degree hours, maximum wa ter temperature, and depth for the in situ cage study, and of mortality on median air temperature for the enclosure data (combined from all years). Residual plots revealed no systematic violations of model assumptions. Because our predictors of mortality (degree hours, ma ximum temperature, and depth) are highly correlated, we compared the relative fits of the individual univar iate models with AIC. We performed analysis 48 of variance (ANOVA) to test for differences in mortality of caged mussels as a function of stocking density. To determine whether the re lationship between mortality and degree hours differed between large and small mussels in the in situ cage study, we conducted a homogeneity of slopes test, followed by analysis of covariance (ANCOVA) to test for differences in intercepts in the event the former test was not significant. To assess the effects of temperature and time on mortality in the acute and chronic temperature tolerance experiments, we performed repeated measures multivariate analysis of variance (R M-ANOVA). Statistical analyses were performed with R (version 3.0.1, The R Foundation for Statistical Computing). Results In situ mortality: caged mussels We detected no differences in mortality among cages stocked with different sizes or total biomass of mussels (ANOVA, p = 0.21, n = 10). As assessed by AIC, mortality of small mussels was explained equally well by threshold te mperatures of 23-26 °C and mortality of large mussels was explained equally well by threshold temp eratures of 23-25 °C (AIC 2, Table 3). Model fit progressively diminished (AIC increas ed) as the threshold temperature was adjusted beyond these ranges (Table 3). For ease of interpretation, we chose 25 °C as the threshold temperature of best fit, since this is the highest temperature having maximal statistical support across both size classes of musse ls. Given our hypothesis that mu ssels have a lower temperature tolerance than previously reported, choosing the highest threshold temperature that best fits the data is a conservative approach with respect to our understanding of chronic temperature stress in zebra mussels. 49 Table 3. Selection of an accumulated degree hour thre shold temperature that best explains mortality of caged large (16-30 mm, n = 10) and small (8-15 mm, n = 7) Dreissena polymorpha in Gull Lake. Results are for linear regressions. Model fit was assessed with AIC, where a lower AIC indicates a better fit. AIC was then computed for each model with respect to the model with the lowest AIC. Models having AIC 2 (underlined) were cons idered to have equal statistical support as the model with the lowest AIC. Large mussels Small mussels Temperature threshold, °C AIC AIC R2 p AIC AIC R2 p > 23 0.77 0.87 0.670.004-0.520.87 0.73 0.015> 24 -0.11 0.00 0.700.003-1.390.00 0.76 0.011> 25 0.95 1.05 0.670.004-1.270.11 0.76 0.011> 26 2.28 2.38 0.620.007-0.640.74 0.73 0.014> 27 4.35 4.45 0.530.0171.082.47 0.66 0.027 > 28 7.71 7.82 0.340.0772.994.38 0.55 0.056 > 29 10.53 10.64 0.120.3173.594.98 0.51 0.071 50 Natural temperature variation across years a nd cage depths provided a large range of accumulated heat exposure (Table 4). Mortality of caged D. polymorpha in Gull Lake was strongly related to accumulated degree hours > 25 °C for both large (Fig. 15a; linear regression, p = 0.004, n = 10) and small (Fig. 15b; linear regression, p = 0.011, n = 7) individuals. We found no difference in degree hours versus mortality regr ession slopes (homogene ity of slopes test, p = 0.86) or intercepts (ANCOVA, p = 0.67) between the two size classes of mussels, indicating similar susceptibility. Maximum water temperat ure and cage depth, which are both correlated with degree hours > 25 °C (r = 0.74 and -0.70, respectively), were poorer predictors of mortality of large individuals (linear regressions, p 0.04, AIC > 5.0, n = 10) and were not significant predictors of mortality of sma ll individuals (linear regressions, p 0.15, AIC > 5.4, n = 7). Notably, the highest levels of mo rtality were observed at exposures comparable to that during the die-off (> 1,500 degree hours > 25 °C, Fig. 15). In situ mortality: mussels in enclosures Mortality of D. polymorpha within experimental enclosures in Gull Lake (July-August, 2005-2008) was positively correlated with median air temperature during the experiment (r = 0.87, Fig. 16), in the absence of a continuous water temperature recor d. Notably, the highest mortality (33%) was observed during the 2005 experiment (8 July- 4 August), when the median summer air temperature was similar to that in 2010 during the mass-mortality event in Gull Lake (Fig. 11). 51 Table 4. Accumulated heat exposure (degree hours > 25 °C) across years and depths in the study of caged Dreissena polymorpha in Gull Lake. Depth (m) Year 2 5 9 20111,9091,4400 20122,804--526 20131,2487700 2014153--2 52 Figure 15. Mortality of caged Dreissena polymorpha in Gull Lake (May-October, 2011-2014) for (A) large (16-30 mm) and (B) small (8-15 mm) individuals as a function of lake degree hours > 25 °C. Results from the chronic temperature tolerance experiment ( see Fig. 18) are indicated for comparison. The regre ssion lines (significant at p < 0.05) are for caged D. polymorpha only. Mortality is given as the propor tion of dead individuals. Bars are ± SE. The arrows denote accumulated degree hours in Gull Lake during the initial mass die-off event in 2010. 53 Figure 16. Mortality of Gull Lake Dreissena polymorpha in four enclosure experiments conducted in Gull Lake (July-August, 2005-2008), as a function of median air temperatures recorded nearby. Mortality is given as the pr oportion of dead individuals. Bars are ± SE. 54 Experiment 1: acute temperature tolerance The effect of temperature on mussel mortality wa s significant, and this effect varied over time (RM-ANOVA, p < 0.001, n = 6) for both size classes of mu ssels. No mortality occurred in heated treatments until temperature reached 32 °C (day 12), at which point mortality increased rapidly (Fig. 17a). Over the first 14 days, there was zero mortality in the 23 °C controls (Fig. 17b). Temperature was further incr eased to 33 °C, the maximum attainable with our heaters, and maintained at 33 °C during the last 3 days of the experiment. Smalle r individuals (8-15 mm) appeared to be more tolerant of elevated temperature than larger individuals (16-25 mm): mortality exceeded 90% for large mussels after le ss than 24 hours at 33 °C (day 13), significantly higher than smaller mussels despit e the same exposure time (ANOVA, p = 0.04, n = 6, Fig. 17a). This same level of mortality was not observed in smaller mussels until nearly 2 days at 33 °C (day 15). Mortality was ~100% for both size clas ses after 3 days at 33 °C, at which point the experiment was ended. Average mortality over the entire experiment was < 3% in controls maintained at 23 °C (Fig. 17b). Experiment 2: chronic temperature tolerance The effect of elevated temperature (> 25 °C versus ~ 23 °C in controls) on mussel mortality also varied over time in this experiment (RM-ANOVA, p < 0.001, n = 10). Mortality of mussels in the heated treatment was low and si milar to controls up to day 10 (600 degree hours > 25 °C), then steadily increased relative to controls (Fig. 18). By the end of the experiment (1,700 degree hours > 25 °C), mortality was 2.7 times higher in heated trays (71% ± 0.045 SE) compared to unheated trays (26% ± 0.060 SE; t-test: t = 6.06, df = 8, p < 0.001). Notably, these mortality rates are highly congruent with the mortality versus heat exposure relationship 55 Figure 17. Water temperatures and mortality of la rge (16-25 mm) and small (8-15 mm) Gull Lake Dreissena polymorpha in the acute temperature tolerance experiment for (A) heated and (B) control treatments. Mortality is given as th e proportion of dead individuals. Bars are ± SE. 56 Figure 18. Results from the chronic temperature tole rance experiment: (A) temperature, (B) accumulated degree hours > 25 °C, and (C) mortality of large (17 mm) Gull Lake Dreissena polymorpha in heated and control treatments. Mortality is given as the proportion of dead individuals. Bars are ± SE. 57 observed for caged mussels in Gull Lake (Fig. 15a). Chlorophyll- a concentrations (mean = 1.28 µg L-1) were within the range found in the lake (J. White, unpubl.) and di d not differ between heated and unheated trays (paired t-test: t = 1.06, df = 7, p = 0.33). Discussion A massive die-off of Dreissena polymorpha occurred in Gull Lake, Michigan in 2010 after 2,500 accumulated degree hours > 25 °C, despit e the fact that temperatures never reached the widely-accepted, acutely lethal threshold of ~30 °C (Iwanyzki and McCauley 1993, Garton et al. 2014). In our in situ caged mussel study in which year and lake depth provided natural temperature variation, seasonal mortality of Gull Lake mussels was largely explained by accumulated degree hours > 25 °C (Fig. 15), a fisublethalfl threshold. Furthermore, using accumulated exposure times and temperatures sim ilar to those observed in Gull Lake during the 2010 die-off, our experimental results demonstrat e a causal link between this level of exposure and mortality, while simultaneously ruling out depth as a confounding factor. Earlier observations of mortality of experimental mussels in Gull Lake enclosures (Fig. 16) are also consistent with temperature being the likely driver. Populations of D. polymorpha are known to exhibit a variety of different dynamics, incl uding boom-bust, though th e influence of large interannual variation in temperatur e has generally either not been previously investigated or not found to be important in explaining these population fluctuations of D. polymorpha (Ramcharan et al. 1992, Strayer et al. 2011). We identified that periods of hi gh temperatures were likely to have had a direct role in causing a large decline in D. polymorpha density, and such declines may become increasingly important give n future climate change scenarios. 58 The distinction between measuring acute and ch ronic thermal tolerance is critical, since the acute lethal temperature will generally be hi gher than the chronic lethal temperature (Spidle et al. 1995), as we found in this study. Like many previous labora tory studies that report high survival of North American D. polymorpha at higher temperatures than those observed in Gull Lake and our chronic tolerance experiment (McMahon et al. 1994, Al dridge et al. 1995, Spidle et al. 1995), we also found high tolerance of Gull La ke mussels for relatively short-term exposure to high temperatures 30 °C in the laboratory (Fig. 17). However, many previous laboratory assessments of D. polymorpha™s thermal tolerance, upon which wide ly reported lethal thresholds of > 30 °C are largely based, were specifically designed to identify the acutely lethal threshold suitable for rapid extermination (Iwanyz ki and McCauley 1993, McMahon et al. 1995, McMahon and Ussery 1995, Beyer et al. 2011); therefore, these studies necessarily employ acclimation and treatment temperature regimes that are almost always extreme relative to nature, and therefore may not nece ssarily predict tolerance in situ . Indeed, we found that maximum water temperature was a rela tively poor predictor of in situ mortality, presumably because the maximum temperatures observed in Gull Lake during our studies ( 30.5 °C) were typically transient and not acutely lethal to D. polymorpha . With acclimation, north-temperate D. polymorpha have been shown to survive as long as 14-35 days at 30-32 °C in the laboratory (McMahon et al. 1994, Aldridge et al. 1995, Spidle et al. 1995), which exceeds the level of expos ure we identified as driving Gull Lake D. polymorpha mortality. However, in laboratory studies of thermal tolerance like these, conditions are generally otherwise ideal including feeding mussels high-quality, cultured phytoplankton. In our in situ and chronic temperature tolerance studies, musse ls fed on natural Gull Lake seston, and, in the case of the in situ study, were also subject to the full range of environmental conditions present 59 in Gull Lake. Time to death at a given temperat ure is known to increase with food quality for D. polymorpha (Stoeckmann and Garton 2001). Factors su ch as food quantity and quality may therefore influence the response to chronic temperature stress. Accumulated exposure time is clearly important for driving mortality of D. polymorpha at sublethal temperatures. For example, in our chronic tolerance experiment, mussels were initially acclimated to experimental conditions by increasing the temp erature 5 °C over 12 hours without any observable consequence. In fact, there was no discernable effect of temperature treatment on mortality for the fi rst 10 days and 600 degree hours > 25 °C of the experiment, which is consistent with studies that report a rapidly increasing survival time (from hours to days) as temperatures are dropped even as little as 1-2 °C below the acutely lethal threshold (Iwanyzki and McCauley 1993, McMahon et al. 1994) and mussels have survived at these temperatures for the entirety of relatively short-te rm (< 14 days) incubations (Spidle et al. 1995). Physiologically, chronic exposure to sublethal temperatures in this range results in a metabolic imbalance whereby feeding cannot compensate for energy expenditures (Aldridge et al. 1995), which could eventually result in starvation and death. Summer die-offs have major implications for D. polymorpha population dynamics, and these implications may vary by region. To our knowledge, all previously published observations of large-scale, apparently temperature-driven ( 30 °C) mortality occurring in nature have been restricted to warmer, southern ecosystems where D. polymorpha thermal tolerance is already known to be greater (e.g., perhaps as a result of local adaptation to warmer temperature regimes) (Allen et al. 1999, Elderkin and Klerks 2005, Boeckman and Bidwell 2014), suggesting that we should expect north-temperate populations to be vulnerable to temperatures < 30 °C in nature. Given the thermal preferences of D. polymorpha (spawning and growth occur at 12-24 °C), 60 spring-summer is the period when mussels in temperate lakes reproduce and achieve maximal somatic growth, whereas most growth in warmer ha bitats occurs in the co oler months (Karatayev et al. 1998, Allen et al. 1999, Garton et al. 2014). Summer die-offs are expected in southern populations, as temperatures are more likely to exceed thermal limits; however, these populations might be able to recover more qui ckly by allocating resources to growth and reproduction at other times of the year, which may not be possible for north-temperate populations (Allen et al. 1999). As a result, north-temperate p opulations may have longer lag times to recovery following summer die- offs, resulting in prolonged effects. The existence of a thermal refuge should al so influence the rate of population recovery from a die-off. It require d nearly three years for D. polymorpha to successfully re-establish (persist through two consecutive growing seasons) on the epilimnetic sediments of Gull Lake following the initial die-off, and densities are stil l substantially lower than historical levels (J. White, pers. obs.). This is despite the fact that cooler, deeper waters provided a thermal refuge for D. polymorpha in Gull Lake where they persisted during the die-off and periods of high mortality in epilimnetic cages (Fig. 17, Table 4) , presumably expediting recovery in shallower waters. Populations in shallow, well-mixed la kes that lack a thermal refuge may be more vulnerable to local, temperature-driven extirpations and could also experience longer lags to recovery with an increased dependence on exogenous dispersal for re-establishment. Population size structure may also interact with temper ature to determine population resilience, since larger mussels, if anything, were more vulnerable to higher temperatures than smaller mussels (Fig. 17, Table 3). This is in agreement with previous studies of D. polymorpha both in situ (Mississippi River) (Allen et al. 1999) and in the lab (McMahon et al. 1994, Elderkin 61 and Klerks 2005). Populations composed of a gr eater proportion of older, larger individuals might be more likely to experience mass mortality at elevated temperatures. Given the well-documented effects of this i nvasive species on lake ecology, large-scale die-offs of D. polymorpha should have concomitantly large e ffects on lake ecosystems. Based on literature values for N and P tissue content of D. polymorpha (0.36 and 0.015 % of dry tissue mass, respectively)(Arnott and Vanni 1996) and a biomass of ~6 g m -2 in Gull Lake (Wilson and Sarnelle 2002), we estimate that the sudden demise of th e population resulted in a pulse of ~21 mg N and ~1 mg P m -2 into the mixed layer of oligotrophic Gull Lake. Such a synchronized pulse of nutrients previously sequestered by mussels could have importa nt consequences for the lake foodweb (Strayer 2014). For example, mixed layer chlorophyll- a concentrations in Gull Lake during the month immediately following the 2010 Dreissena die-off were at their highest observed levels for the period of record from 1998-2014 (maximum of 8.65 µg L-1, relative to the 16-year summer mean of 3.72 µg L-1; J. White, unpubl.), sugge sting this nutrient pulse stimulated phytoplankton growth. In shallo wer, well-mixed systems where filter-feeding D. polymorpha exert stronger control on phytoplankton biomass (Higgins and Vander Zanden 2010), temperature-induced die-offs could reverse Dreissena™s large effects on chlorophyll concentrations and transparency (Higgins et al. 2011, Cha et al. 2013). Die-offs of Dreissena could also result in marked shifts in the com position of the associated benthic invertebrate community (Ward and Ricciardi 2007, Gergs and Rothhaupt 2015). Dreissenids are also strong promoters of the toxic cyanobacterium Microcystis aeruginosa, especially in low-nutrient lakes (Vanderploeg et al. 2001, Raikow et al. 2004, Knoll et al. 2008, Sarnelle et al. 2012). Cyanobacteria, including Microcystis, prefer warm temperatures (growth optimized at 28 °C) and are expected to increase worldwide with climate 62 change (Zehnder and Gorham 1960, Paerl and Huis man 2008). However, in low nutrient lakes like Gull Lake where the success of Microcystis depends on zebra mussels (Sarnelle et al. 2005, Sarnelle et al. 2012), Microcystis biomass and associated toxin might actually decrease if temperatures increase above the thermal tolerance of D. polymorpha. A complex interaction between D. polymorpha, Microcystis, and temperature may lead to non-monotonic responses of Microcystis to a warming climate in low nutrient lakes. Mass-mortality events, local extirpations, and range contractions have been reported in marine ecosystems (Australian coast: canopy-form ing seaweeds; Mediterranean Sea: gorgonians, sponges) in response to recent, extreme h eat waves (Coma et al. 2009, Garrabou et al. 2009, Smale and Wernberg 2013). Given documented and anticipated increases in temperatures of inland waters as a result of climate change (Sch indler et al. 1990, Jankowski et al. 2006), such events might also become more common in freshw aters. Specifically, our results suggest that summer declines or die-offs of invasive D. polymorpha may become more frequent in north- temperate lakes in the future. Therefore, it is critical to have estimates of environmental tolerances that accurately reflect processes in nature when forecasting the combined influences of species invasions and climat e change on ecosystems. Given that highly successful invasive species like D. polymorpha will likely continue to adapt to their changing environmental template, understanding and management of these sp ecies will likewise need to be responsive to ongoing global change. Acknowledgements We thank S. Flemming, T. Geelhoed, C. Kozel , and M. Schuetz for assisting with the experiments. J. Allen, T. Browne, T. Coffing, E. Fergus, A. Fogiel, E. Litchman, T. Lund, G. 63 Mittelbach, S. Peacor, D. Weed, J. White, and M. Williams provided additional assistance and/or lent equipment. Funding was provided by the Environmental Protection Agency (Ecology and Oceanography of Harmful Algal Blooms/2004- Science to Achieve Results-C1, project RD83170801), the National Science Foundation (Division of Environmental Biology-0841864, Division of Environmental Biology-0841944) and the Gull Lake Quality Organization. Long- term air temperature data were made availa ble by the NSF Long-term Ecological Research Program (DEB 1027253) at the Ke llogg Biological Station and by Michigan State University AgBioResearch. The comments of two anonymous reviewers greatly improved the manuscript. Supplemental Information For additional results pertinent to this chapter, see the Appendix (Survey of zebra mussel [Dreissena polymorpha] status in other Michigan inland lakes during recent warm summers). 64 CHAPTER 3 OPPOSING RESPONSES OF STRONGLY INTERACTING SPECIES TO ELEVATED TEMPERATURES SUPPRESS THE HARMFUL PHYTOPLANKTER MICROCYSTIS Abstract Climate change has already resulted in la rge changes in the sp atial and temporal distributions of species, with significant consequences for individual populations. However, the community- and ecosystem-level implications ma y be complex and challenging to anticipate due to the cascading effects of disrupting the interactions among species. Toxic, bloom-forming cyanobacteria like Microcystis aeruginosa are expected to increase worldwide with climate change, due in part to their relatively high temperature op tima for growth. Facilitation by invasive zebra mussels ( Dreissena polymorpha) has also resulted in increases of Microcystis in low-nutrient (oligotrophic) lakes, an uncharacteristic habitat for this harmful phytoplankter. We conducted a 13-year study of a M. aeruginosa population in a low-nutrient lake invaded by zebra mussels. In 11 of the 13 years, there was a significantly positive relationship between M. aeruginosa biomass and mean spring-summer water te mperature, which is consistent with climate change forecasts. Surprisingly, we observed very low Microcystis biomass and microcystin toxin concentrations during one of the warmest years in the time series following a heat-induced mass mortality event of zebra mussel s, which resulted in low mussel densities for two consecutive summers. Upon elimination of its facilitating species, the positive relationship between Microcystis biomass and temperature decoupled. Thus, predicting the response of a species to climate change may require, at minimu m, quantification of temperature responses of both the focal species and species that strongly interact with it. Consequently, monitoring of intact communities with respect to cl imatic variables seems essential. 65 Introduction Anthropogenic global climate change is causing rapid and large shifts in the environmental template of species, which is very likely to continue given additional increases in global mean temperatures forecasted by the end of the century (IPCC 2014). Responses of individual species to recent climatic variation have alrea dy been documented, including range shifts (Perry et al. 2005, Chen et al. 2011), changes in the phenology of critical life history stages (Edwards and Richardson 2004), mass-mortality ev ents and local extirpations (Garrabou et al. 2009, Smale and Wernberg 2013, White et al. 2015) , and potentially extinctions (Pounds et al. 2006) across a broad range of taxonomic groups (algae, plants, invertebrates, amphibians, birds) and habitats (terrestrial, mari ne, freshwater). Since not all species will respond in the same manner, this information is valuable for unders tanding and predicting the response of different populations, and for guiding appropriate manageme nt and conservation decisions. However, because species are embedded in communities com posed of complex networks of interactions, the response of an individual species to climate warming could re sult in unexpected, cascading effects on ecosystems by modifying the predicted re sponses of other interacting species, with the net result of these interactions potentially even nega ting the direct effects of climate (Suttle et al. 2007, Post and Pedersen 2008, Tylianakis et al. 2008, Van der Putten et al. 2010). The emphasis that has generally been placed to date on asse ssing the responses of individual species in isolation may therefore result in the failure to accurately predict community- and ecosystem- level responses to climate change (Gilman et al. 2010). Information about how closely interacting species respond together to large clima tic variation in natural systems, particularly in direct response to elevated temperatures, is currently limited (Gilman et al. 2010, Van der Putten et al. 2010). 66 Rising water temperatures, l onger growing seasons, and increases in the duration and stability of thermal stratification have been reported recently from north-temperate lakes (Schindler et al. 1990, Winder and Schindler 20 04, Jankowski et al. 2006, Austin and Colman 2008, Dobiesz and Lester 2009, Schneider and Hook 2010, Rösner et al. 2012), changes that are expected to exacerbate the growth of toxic, bloom-forming cyanobacteria like Microcystis aeruginosa in freshwaters worldwide with climat e change (Paerl and Huisman 2008, Paul 2008, Carey et al. 2012b, Elliott 2012, Paerl and Paul 2 012, Reichwaldt and Ghadouani 2012). Harmful cyanobacteria like Microcystis can present significant human health concerns and impair use of drinking and recreational waters (Carmichael 1994, Pilotto et al. 1997, Falconer 1999). Although there is a consensus that climate change will promote these nuisance phytoplankton in general, models that make predictions for specific ha rmful cyanobacteria species are less developed (Pitcher 2012). Furthermore and not surprisingly, attention has almost exclusively been focused on phosphorus-polluted (eutrophic) lakes (Carey et al. 2008), since the biomass and relative abundance of cyanobacteria and their toxins generally increase with lake total phosphorus (TP) concentration (Smith 1985, Trimbee and Prepas 1987, Watson et al. 1997, Downing et al. 2001, Giani et al. 2005, Kotak and Zurawell 2007, Bigham et al. 2009). However, facilitation by the strongly interacting, invasive zebra mussel ( Dreissena polymorpha) has resulted in significant increases (3.6-fold in Michigan, USA) of toxic Microcystis in low-nutrient (oligotrophic) lakesŠan uncharacteristic habitat for a harmful phytoplankter typically associated with high-nutrient ha bitats (Vanderploeg et al. 2001, Raikow et al. 2004, Knoll et al. 2008, Sarnelle et al. 2010). Consequently, Microcystis dynamics in these lakes are unlikely to be predicted well using existing models ba sed solely on nutrient concentrations, nutrient ratios, and other variables related to lake productivity (Smith 1983, 67 1986, Smith et al. 1987, Canfield et al. 1989, Do wning et al. 2001). Zebra mussels, highly successful and impactful invasive species, also continue to expand their range across North America and Europe (Strayer 2008, Higgins a nd Vander Zanden 2010). This ongoing species invasion presents an additional case of global change impacting M. aeruginosa and water quality, and it is currently unknown how Microcystis will respond to climatic warming in these socially and economically valuable low-nutrient lakes where its success is highly dependent on Dreissena. Low-nutrient Gull Lake, Michigan, where Microcystis promotion following zebra mussel invasion has been extensively studied (Sarnelle et al. 2005, Sarnelle et al. 2012, Horst et al. 2014), experienced an unprecedented, heat-induced mass-mortality event of zebra mussels on epilimnetic sediments during the warm summe r of 2010, followed by two years (2011-2012) of failed recolonization when water temperatures continued to exceed their chronic thermal tolerance (> 25 °C, with maxi mum temperatures of 29-30.5 °C; see Chapter 2) (White et al. 2015). These elevated temperatures, though lethal to its facilitating species, are within the range of laboratory-derived optimal temperatures for Microcystis (~25-32 °C) (Zehnder and Gorham 1960, Robarts and Zohary 1987, Nalewajko and Murphy 2001, Imai et al. 2009), which may lead to unexpected responses of Microcystis to increasing temperatures in low-nutrient lakes. Microcystis is now the only toxin-producing cyanobacterium that regularly achieves appreciable biomass in Gull Lake (Sarnelle et al. 2005), making it an ideal habitat for investigating the response of Microcystis to climatic variation under low-nutrient conditions. First, given the aforementioned high temper ature optima reported for cyanobacteria and Microcystis, which are expected to make them superior competitors at elevated temperatures (McQueen and Lean 1987, Fujimoto et al. 1997, Kosten et al. 2012, Rigosi et al. 2014), we 68 predicted that inter- annual variation in Microcystis biomass in Gull Lake is primarily driven by water temperature. Second, however, since D. polymorpha is already known to promote Microcystis and microcystin in Gull Lake (Raikow et al. 2004, Sarnelle et al. 2005, Knoll et al. 2008, Sarnelle et al. 2010, Sarnelle et al. 2012), we anticipated a priori that the mass-mortality event of Dreissena would have significant negative impacts on Microcystis, which might lead to non-monotonic responses of Microcystis to elevated temperatures in low-nutrient lakes. Third, we predicted that microcystin toxin concentratio ns in Gull Lake would be positively related to Microcystis biomass and available nitrogen, since mi crocystin cell quota is constrained by nitrogen availability (Horst et al. 2014). We analyzed 13 years of physical, chemical, and biological data from Gull Lake to test these predictions. Methods Study site Gull Lake is a large (822 ha), deep (33 m maximum depth, 12 m mean depth), oligotrophic hardwater lake located in Barry and Kalamazoo Counties in southwestern Michigan (Fig. 19). The lake is phosphorus-limited (Moss 1972b), and summer TP concentrations in the mixed layer average 7.59 g L-1 with a DIN:TP ratio of ~42:1 (Table 5). Details of the lake™s geologic and climatic setting ar e summarized in Moss (1972a) and Tessier and Lauff (1992). Like other low-nutrient lakes, Microcystis biomass increased dramatical ly (from ~0% to >15% 69 Figure 19. Map of Gull Lake, Michigan, indicating the locations of the nearshore (fidockfl) and central (ficenterfl) sampling stations. Depth contours are in increments of 3 meters. 70 Table 5. Summary of limnological characteristics of Gull Lake, Michigan (1998-2014, mixed layer during summer stratificati on). SE = standard error; 10 th = 10th percentile; 90 th = 90th percentile; n = number of observations; Chl- a = chlorophyll-a; Secchi = Secchi disk transparency; Z epi = epilimnion depth; Alk = alkalinity; TP = total phosphorus; TDP = total dissolved phosphorus; SRP = so luble reactive phosphorus; NO 3- = nitrate; NH 4+ = ammonium; DIN:TP = ratio of [NO3- + NH4+ ] to TP. Parameter (units) Mean (SE)10 th 90 th n Chl-a (µg L-1) 3.86 (0.12)1.406.10 231 Secchi (m) 4.18 (0.11)2.536.43 242 Zepi (m) 7.71 (0.13) 5.50 10.00 226 pH 8.53 (0.02)8.158.85 184 Alk (µeq L-1) 3,107 (45.63) 2846 _ _ 3353_ _ 80 TP (µg L-1) 7.59 (0.12)5.539.68 245 TDP (µg L-1) 4.76 (0.14)3.146.76 182 SRP (µg L-1) 1.29 (0.06)0.422.72 255 NO3- (µg L-1) 277 (7.68)116 _ _410 __ 223 NH4+ (µg L-1) 22.09 (1.20)6.6738.41 230 DIN:TP 42 _ _ (1.53)15 _ _72_ _ 211 71 relative abundance) in Gu ll Lake following invasion by D. polymorpha in 1994 (Sarnelle et al. 2005), including a visible surface scum that oc curred shortly thereafter (A. Tessier, pers. comm.). Prior to 1994, Microcystis was essentially absent from the phytoplankton community, although it was observed sporadically during a period of cultural eutrophication in the 1960s-1970s when surface blooms of Anabaena were reported (Moss 1972a, Moss et al. 1980). This eutrophication trend was reversed with the installation of a sanitary sewer around the lake in 1984, after which TP quickly decreased (Tessier and Lauff 1992) and has continued to decline to present-day oligotrophic levels (S. Hamilton, unpubl.). Limnological sampling Weekly sampling of Gull Lake was conducted from June-August in 1998-2001 and 2005-2008, and from June-September in 2009-2014. During 2000-2001 and 2009-2011, samples were collected from a near-shore station (fidockfl; dept h = 13 m) as well as a ce ntral station (ficenterfl; depth = 30 m; Fig. 19). Only the near-shore station was sampled during 2005-2008, and only the central station was sampled during 1998-1999 and 2012-2014. None of the response variables (paired t-tests, n = 21-29, all p 0.24) or predictor variables (paired t-tests, n = 15-29, all p 0.14) significantly differed between sampling stations when both sites were sampled in parallel, except for ammonium (paired t-test, n = 22, p = 0.002), which was significantly higher at the near-shore station. However, the mean differenc e in ammonium concentration between sampling stations was only 2.78 µg L-1, a relatively small difference when compared to the observed range in ammonium in Gull Lake (Table 5). Therefore, data were averaged across both stations for all variables and pooled for all analyses. 72 A vertical lake profile was taken from the surface to the bottom at 1 m intervals for temperature, dissolved oxygen, conductivity, and pH using a multiparameter sonde (Hydrolab Surveyor and Datasonde 4a). The temperature profile data was solely used to identify the epilimnion in the field, and to later calculate the thermocline depth and Schmidt water column stability (Idso 1973) using the rLakeAnalyzer package in R (Read et al. 2011). We measured water transparency ( ZSD) with a 20-cm Secchi disk, and estima ted the light extinction coefficient, k, as 1.7 / ZSD (Wetzel 2001). Lake water was collected from the epilimnion via two pooled casts of a depth-integrating tube sampler (12 m length × 2.5 cm i .d.). After thorough mixing and within 1 hour of collection, subsamples we re taken for phytoplankton identification and enumeration (unfiltered water preserved in 1% Lugol™s iodine), total phosphorus (unfiltered water), chlorophyll-a and particulate microcystin (retained on Pall A/E glass fiber filters, 1.0 µm nominal pore size), and dissolved nutrients (filtrate passed thr ough A/E filters: soluble reactive phosphorus, ammonium, nitrate, and major cations). Filters were frozen immediately, and water samples for nutrients were either analyzed promptly or frozen for later analysis. Lab analyses Chlorophyll-a was measured fluorometrically follo wing 12-hour dark extraction of frozen A/E filters in 10 mL of cold 95% ethanol (Welschmeyer 1994, White et al. 2011). Particulate microcystin was measured by ELI SA (enzyme-linked immunosorbant assay; Envirologix QuantiPlate Kit for Microcystin s visualized with a LabSystems Multiskan Microplate Reader) following three pooled 45 mi n, 10 mL extractions of A/E filters in 75% methanol (Harada et al. 1999, White et al . 2011). Soluble reactive phosphorus (SRP; molybdenum-blue method) and ammonium (indophe nol-blue method) were analyzed with 73 standard colorimetric techni ques and long-pathlength spectrophotometry (Murphy and Riley 1962, Solórzano 1969, Aminot et al. 1997). Total phosphorus was analyzed as above for SRP following persulfate digestion of organic materi al in an autoclave (Menzel and Corwin 1965). Nitrate was analyzed by Dionex membrane-sup pression ion chromatography (Hamilton et al. 2009). Lugol™s preserved subsamples were settled in 10 mL tubul ar plankton chambers (Hydro- Bios), followed by counting and measuring of sedimented phytoplankton via the Utermöhl inverted microscope method (Lund et al. 1958, Hasle 1978). Microcystis biomass was quantified from measurement of two-dimensional surface areas of all colonies in the chamber using digital image analysis software (Spot Advanced) at 100× . These measurements were converted to dry biomass using a regression (log 10[y] = 1.05log10[x] Œ 1.32; R2 = 0.94) that estimates cell density from colony surface area for Gull Lake Microcystis; cell density was converted to biovolume and ultimately to dry biomass using measured cell di mensions, and assuming a specific gravity of 1.0 and a wet-to-dry biomass conversion factor of 0.4 (Sarnelle et al. 2012, Horst et al. 2014). Colony size was measured as the equivalent di ameter (ED), by calculating the diameter of a circle with surface area equivale nt to that of each measured co lony (White and Sarnelle 2014). Full phytoplankton community counts were performed on 8 dates (4 per year in 2 different years), selected to represent the typical range of total phytoplankton biomass for Gull Lake as assessed from chlorophyll- a (~1.5-5.5 µg L-1). Phytoplankton were identified to genus, and cell counts were made as above at a variety of magnifications (100, 400, 1000×) to accurately assess densities of bot h large and small taxa. Approximately 50 random fields were viewed in each of two chamber regions (outer an d inner, of nearly equal surface area) to account for non-random settling of cells (Sandgren and Robinson 1984). Cell volumes (exclusive of 74 spines, horns, and sheaths) were determined from digitized measurements of geometric dimensions (Hillebrand et al. 1999, Sarnelle et al. 2005) and converted to dry biomass as above. We then performed a linear regression of log-transformed phytoplankton dry biomass on chlorophyll-a (y = 0.141x + 1.801, n = 8, R2 = 0.62, p = 0.021; Fig. 20) to enable more rapid estimation of relative Microcystis biomass. Temperature data Since variation in water temperature is likely to cause a lagged response of Microcystis biomass as a function of time-integrated condi tions, we considered annual water temperature effects at three seasonal scales: spring (April-May), summer (June-August), and the combination of the two (spring-summer). Water temperature data were averaged separately for each of these three seasonal scales to assess th eir relative explanatory power. Furthermore, because instantaneous water te mperatures observed at the time of weekly sampling do not necessarily reflect prevailing co nditions, we collected high-frequency (hourly) water temperature data by instal ling a vertical chain of data loggers (Onset Corp.) through the mixed layer of Gull Lake from April-August in 2010-2014, which was used to compute a daily mixed layer average. Since this high-frequency temperature record is unavailable for Gull Lake prior to 2010, we constructed empirical models to predict pre-2010 daily mixed layer water temperatures from daily air temperature records obt ained for the entire duration of study from the Long Term Ecological Research (LTER) site at the Kellogg Biological Station (KBS, http://lter.kbs.msu.edu/datatables /12, datatable KBS002-006.27), which is adjacent to Gull Lake. We first confirmed that KBS LTER air temper ature is a reasonable approximation of air 75 Figure 20. Prediction of total phytoplankton dry biomass from chlorophyll- a concentration in the epilimnion of Gull Lake ( n = 8). 76 temperatures over Gull Lake using nearshore da ta recorded in 2010 and 2012 (linear regression; n = 273, R2 = 0.86, p < 0.001, Fig. 21). To account for lags in the response of water temperature to variati on in air temperature when performing the conversion, we performed a se ries of linear regressions where we varied the time over which air temperatures were averag ed leading up to and including the day of water temperature observation, starting with no lag an d sequentially increasing the lag time by 1 day. We continued to increase the lag time until model fit was maximized and/or began to diminish again. We assessed the relative fit of each regression model with the Akaike Information Criterion (AIC), where a lower AIC indicat es a better fit, and then computed AIC for each model with respect to the model w ith the lowest AIC. Models having AIC 2 were considered to have equal statistical support as the model with the lo west AIC (Burnham and Anderson 2001). Since the slope of the air-water temperat ure relationship differed by season (homogeneity of slopes test, p < 0.001), we performed this procedure se parately for the spring and summer seasons (see above). Twenty-two days was the shortest lag having maximal statistical support (AIC 2) for both the spring (y = 1.587x + 5.955, AIC = 392, R2 = 0.84, n = 105, p < 0.001; Fig. 22a) and summer (y = 0.870x + 6.403, AIC = 1474, R2 = 0.72, n = 458, p < 0.001; Fig. 22b) seasons. Therefore, these are the two models we used to estimate daily Gull Lake mixed layer water temperatures in the absence of direct , continuous observations prior to 2010. Statistical analyses The effect of interannual variation in water temperature in driving variation in M. aeruginosa biomass, relative biomass, timing of the maximum biomass, median colony size, net population growth rate, and particulate microcysti n concentration were analyzed individually 77 Figure 21. Relationship between daily mean air temp eratures recorded at the Long-term Ecological Research (LTER) site at the Kellogg Bi ological Station, and at a lakeshore laboratory on Gull Lake (n = 273). 78 Figure 22. Conversion of KBS-LTER air temperature data to epilim netic water temperature for Gull Lake, by season. 79 Figure 22. (cont™d) (A) Spring (April-May, n = 105) and (B) summer (June-August, n = 604). Left panels illustrate the selection of the best-fitting lag time between air and water temperatures via R 2 (squares) and AIC (circles). Right panels show the relationship between observed epilimnetic water temperatures in Gull Lake (2010-2014) and the be st-fitting air temperature lag of 22 days. The 22-day mean air temperature leading up to and including the day of water temperature observation was the shortest lag having maximal stat istical support (AIC < 2; see text). 80 with linear regressions. Given the anticipated consequences of the ma ss-mortality event of D. polymorpha on Microcystis (see Introduction), we decided a priori to perform all statistical analyses separately for all years ( n = 13) versus excluding 2011-2012 (n = 11), during which time D. polymorpha was largely absent from the epilimnetic sediments of Gull Lake (White et al. 2015). To explore whether including other typical drivers and correlates of Microcystis biomass reported from eutrophic lakes in the literature (T P, Schmidt stability, and light availability) could explain any of the remaining variation, we pe rformed stepwise multiple regression using both forward and backward variable selection techniques. The influence of Microcystis biomass and available nitrogen (as NO 3-) on particulate microcystin conc entrations was analyzed with multiple regression. Due to the varied duration of sampling progr ams prior to 2009, the de clining phase of the Microcystis population was not fully observed in each year. Therefore, mean biomass was computed from the time of first detection th rough the population maximum. Mean biomass was then log-transformed to stabilize the variance. Relative biomass (as a proportion) was arcsine- squareroot transformed prior to analysis to meet distributional assumptions. However, this data transformation had no substantiative effect on a ny result, and so we report results for the raw data for ease of interpretation. All predictor variables, with the exception of water temperature (see above) were averaged from June-August annually. All statistical analyses were performed with R version 3.1.3 (R Foundation for Statistical Computing) a nd SYSTAT version 12. 81 Results Microcystis biomass, relative biomass, timing of maximum biomass, median colony size, and particulate microcystin all exhibited substan tial intra- and inter-annual variation in Gull Lake over the period of observation from 1998-2014 (Fig. 23). Microcystis generally first appears in the water column in June, increases to a maximum in July or August, and then declines (Fig. 24). The timing of the population maximum varied by as much as one month over the period of observation, from 26 July to 27 August. Bioma ss at the time of the population maximum ranged from 6 µg L-1 to 135 µg L-1, corresponding to ~3% and 80%, respectively, of the total phytoplankton dry biomass (Fig. 23). On average, Microcystis constituted ~2-25% of the total annual phytoplankton dry biomass during the summer in Gull Lake. Annual net population growth rates ranged from 0.04 day-1 to 0.11 day-1, and median colony size ranged from 69 µm to 138 µm equivalent diameter. Annua l mean particulate (presuma bly cell-bound) microcystin concentrations ranged from 0.02 µg L-1 to 0.34 µg L-1, and no individual observation exceeded 0.45 µg L-1 (Fig. 23). Responses of Microcystis biomass to interannual variation in water temperature On average, Microcystis biomass was significantly higher in Gull Lake during warmer years (linear regression, n = 11, R2 = 0.45, p = 0.025; Fig. 25a), but this was only true for years when zebra mussels were present on epilimnetic sediments. While zebra mussels remained absent on the epilimnetic sediments during 2011-2012, the positive effect of temperature on Microcystis biomass decoupled (linear regression; n = 13, R2 = 0.22, p = 0.11; Fig. 25a). Decoupling occurred despite the fact that this period of high mortality and failed recolonization of zebra mussels overlapped with the warmes t summer in the time seriesŠthe mean 82 Figure 23. Time series plots of Microcystis parameters measured in Gull Lake, 1998-2014. 83 Figure 23. (cont™d) (A) Microcystis dry biomass, (B) proportion of the total phytoplankton dry biomass as Microcystis, (C) median Microcystis colony size (as equivalent diameter), and (D) particulate (cell-bound) microcystin concentr ation. Microcystin was not monitored in Gull Lake until 2005, and sampling for all other response variab les was conducted intermittently in 1999 and 2002- 2004. Asterisks (*) indicate observations made duri ng years in which zebra mussels were largely absent from the epilimnetic sediments of Gu ll Lake (due to heat-induced mass mortality, see text).84 Figure 24. Seasonal dynamics of Microcystis biomass in Gull Lake during a representative warm year (2010; mean spring-summer water temp erature = 21.41 °C, maximum biomass = 69.5 µg L-1 on 29 July) and a representative cool year (2014; mean spring-summer water temperature = 19.51 °C, maximum biomass = 11.4 µg L-1 on 19 August). The two series are plotted on a common time scale (day of year) to emphasize di fferences in the timing of first appearance and the timing and magnitude of the biomass maximu m. The mass mortality event of zebra mussels occurred between days 209 and 215 in 2010. 85 Figure 25. Microcystis biomass dynamics as a function of intera nnual variation in Gull Lake water temperatures: (A) annual mean Microcystis dry biomass, and (B) day of year of maximum Microcystis biomass. Years with zebra mussels ( Dreissena polymorpha) present on epilimnetic sediments (circles, n = 11) are differentiated from years (2011 and 2012, squares) during which zebra mussels were largely absent on epilimnetic sediments as a result of heat-induced mass mortality (see text). Results from the linear regression analysis in (A) are for years with Dreissena only, whereas those in (B) are for all years ( see text for explanation). 86 spring-summer water temperature was 21.1 °C in 2012, and Gull Lake reached a maximum epilimnetic temperature of 30.5 °C (Table 6). However, mean Microcystis biomass in 2012 (3.19 µg L-1) was equivalent to that in a year with a predicted mean spring-summer water temperature of only 18.6 °CŠcooler than any of the observed years. Microcystis biomass was reduced on average by 71% during 2011-2012, as compared to years when zebra mussels were present on epilimnetic sediments (Fig. 26a). The annual Microcystis population maximum also occurred significantly earlier in the gr owing season during warmer year s, and this relationship was unaffected by the mass mortality event of zebra mussels (linear regression; n = 13, R2 = 0.38, p = 0.025; Table 7, Figs. 24, 25b). In all cases where significant temperature e ffects were found, spring-summer mean water temperature was a better predictor than either the spring or summer seasonal means alone (Table 7). We did not detect significan t temperature effects on relative Microcystis biomass, median colony size, or net population growth rate at the annual scale (linear regressions; n = 11, R2 0.32, p 0.082; Table 7). Responses of Microcystis biomass to interannual variation in other potential drivers We found no influence of annual TP, Schmidt stability, or light availability ( k) on Microcystis biomass or timing of maximum biomass (stepwise multiple regressions; p for all parameters > 0.05, irrespective of Dreissena presence). Forward and backward variable selection confirmed water temperature to be the best predictor of log mean biomass (only when Dreissena was present) and timing of ma ximum biomass (irrespective of Dreissena presence; see above) at the annual scale. The mean proportion of total phytoplankton dry biomass as Microcystis was 87 Table 6. Mean Gull Lake water temperatures (°C), computed annually for each seasonal time scale: spring (April-May), summer (June-August), and spring-summer. Year Spring Summer Spring-summer 1998 14.8524.1420.96 2000 11.9823.0319.24 2001 14.1223.5420.31 2005 11.9825.2920.72 2006 12.6724.6020.57 2007 12.1824.9720.59 2008 12.0323.8719.78 2009 11.5623.1619.18 2010 13.9625.3021.41 2011 10.0924.5719.54 2012 13.0725.3221.12 2013 10.8224.1819.60 2014 10.9523.9719.51 88 Figure 26. Microcystis biomass (A) and particulate (cell-bou nd) microcystin concentration (B) in Gull Lake in the presence ( n = 11, black columns) versus absence (2011-2012; n = 2, open columns) of zebra mussels ( Dreissena polymorpha) on the epilimnetic sediments. Zebra mussels were largely absent on epilimnetic sediments in 2011-2012 as a result of heat-induced mass mortality (see text). Error bars represent ± SE. 89 Table 7. Influence of water temperature on annual me ans for each response variable for years when zebra mussels were present in Gull Lake (excluding 2011-2012, see text). Response Predictor (seasonal scale) nR2 p AICMean Microcystis biomass Spring-summer 110.45 0.025 7.3 Summer 110.33 0.063 9.3 Spring 110.20 0.165 11.3 Day of peak Microcystis biomass Spring-summer 110.36 0.050 82.1 Summer 110.29 0.086 83.1 Spring 110.14 0.257 85.3 Relative biomass Summer 110.30 0.082 -15.1 Spring-summer 110.25 0.113 -14.4 Spring 110.05 0.512 -11.7 Colony size Spring 100.32 0.086 91.4 Spring-summer 100.06 0.485 94.7 Summer 100.01 0.799 95.2 Mean particulate microcystin Spring-summer 80.26 0.200 -8.8 Summer 80.28 0.176 -9.1 Spring 80.15 0.344 -7.8 Population growth rate Spring-summer 110.03 0.644 -39.7 Summer 110.04 0.576 -39.8 Spring 110.00 0.864 -39.5 90 Table 7. (cont™d) We also assessed the influence of the seasona l scale over which water temperatures were averaged on each response variable. The three seasonal time scales are: spring (April-May), summer (June-August), and spring-summe r. Predictors are listed from the best to the worst fit, as assessed by AIC. Microcystis biomass was log-transformed pr ior to analysis. Significant ( p 0.05) results are underlined. All results are for linear regressions.91 significantly, positively related to Microcystis dry biomass, irrespective of Dreissena presence (linear regression; n = 13, R2 = 0.74, p < 0.001; Fig. 27a). Responses of microcystin toxin Annual particulate (presumably cell-bound) mi crocystin concentrations in Gull Lake were significantly, positively related to Microcystis biomass (linear regression; n = 10, R2 = 0.79, p < 0.001; Fig. 27b). Although this relationship was unaffected by the mass mortality event of zebra mussels, particulate microcystin concentr ations were reduced on average by 80% during 2011-2012, as compared to years when zebra mussels were present on epilimnetic sediments (Fig. 26b). Adding nitrate as a covariate in addition to Microcystis biomass did not improve model fit (ANOVA goodness-of-fit test, p = 0.097). Discussion In stark contrast to existing climate change projections that predict increases in harmful cyanobacteria as a result of positive temperatur e effects on growth physiology and water column stratification (Paerl and Huisma n 2008, Paul 2008, Carey et al. 2012b), we observed a crash in Microcystis during an especially warm summer in low- nutrient Gull Lake. The primary driver of Microcystis in oligotrophic Gull Lake is unequivo cally zebra mussels, as both pre-invasion (Sarnelle et al. 2005) and post-2010 heat-induced mass mortality data demonstrate (Fig. 26). Microcystis biomass increased with temperature in Gull Lake only when zebra mussels were present (Fig. 25a), implicating Dreissena as strong facilitators of Microcystis in low-nutrient lakes ( 10-25 µg L-1 TP) (Raikow et al. 2004, Knoll et al. 2008). Zebra mussels remained absent 92 Figure 27. Relationships between mean Microcystis biomass in Gull Lake and (A) mean proportion of total algal dry biomass as Microcystis (n = 13) and (B) mean particulate (cell- bound) microcystin toxin concentration ( n = 10). Years with zebra mussels ( Dreissena polymorpha) present on epilimnetic sediments (circles ) are differentiated from years (2011 and 2012, squares) during which zebra mussels were largely absent on epilimnetic sediments as a result of heat-induced mass mortality ( see text). Results from linear regression analyses are for all years ( see text for explanation). 93 from the epilimnetic sediments during 2011-2012 following the initial 2010 mortality event because prolonged temperatures > 25 °C (maximum = 30.5 °C) continued to exceed their chronic lethal threshold (White et al. 2015) ( see Chapter 2). Despite the fact that these water temperatures were within the known optimal range for Microcystis growth and dominance (~25-32 °C) (Zehnder and Gorham 1960, Robarts and Zohary 1987, Imai et al. 2009), Microcystis dynamics during this period resembled those for the coolest years of observation (Figs. 23, 25a, 21). In fact, the magnitude of the responses of Microcystis and microcystin (3.5× and 5.1×, respectively; Fig. 26) to the Dreissena mortality event are highly congruent with comparative data from invaded and non-invaded lakes ( 3.6× and 3.3-8.0×, respectively) (Knoll et al. 2008, Sarnelle et al. 2010). Temperature, then, only modulates the actual biomass achieved in any given year in Gull Lake, since the promotion of Microcystis by zebra mussels in low-nutrient lakes is reversible upon elimination of zebra mu ssels, as shown in this study and in enclosure experiments (Sarnelle et al. 2005, Sarnelle et al. 2012). Thus, climate change might result in complex, non-monotonic responses of Microcystis to elevated temperatures in low-nutrient lakes by disrupting the critical interaction with its facil itator species, highlighting the need to consider species interactions in studies of ecological responses to climate change (Tylianakis et al. 2008, Van der Putten et al. 2010). To date, most research on the effect of climate change on species interactions has emphasized spatial and temporal mismatches in the phenologies of critical life history events (e.g., migration, germination, emergence) in pr edator-prey and mutua listic systems where species differ in the timing, direction, and magnitude of their responses (Edwards and Richardson 2004, Winder and Schindler 2004, Both et al. 2006, Schweiger et al. 2008, Yang and Rudolf 2010). We observed maximally opposed res ponses (i.e., high mortal ity versus enhanced 94 growth) by strongly interacting, apparently commensal species to the same large climatic variation within a single lake ecosystem. Furthermore, the opposi ng response of the facilitator species completely negated the positive climatic effect on Microcystis, emphasizing that understanding how a species responds in isolation to climatic drivers, particularly in an experimental setting, does not ne cessarily predict the response of the community in nature (Suttle et al. 2007, Post and Pedersen 2008). There are at least two non-mutually excl usive explanations for why zebra mussels promote Microcystis so strongly in low-nutrient lakes like Gull Lake: the selective feeding hypothesis, and the nutrient excretion hypothesis. Zebra mussels are highly selective feeders, and will reject unpalatable strains (White et al. 2011) and large colonies (White and Sarnelle 2014)(see Chapter 1) of Microcystis back into suspension, which can remain viable (Vanderploeg et al. 2001). Selective rejection of Microcystis, in concert with high mortality imposed on other competing phytoplankton, has b een posited as the primary mechanism by which zebra mussels have promoted Microcystis in the Laurentian Great Lakes (Vanderploeg et al. 2001). However, zebra mussel excretion also diverts nutrients assimilated by phytoplankton from the open water column to the benthos (H ecky et al. 2004). Although the low N:P ratios evidenced to favor cyanobacteria in eutrophic lakes (Smith 1983) are generally not encountered in the open waters of low-nutrient lakes like Gull Lake (Table 5), zebra mussels excrete at low N:P (Arnott and Vanni 1996), which may alle viate nutrient limitation and facilitate Microcystis recruitment to the plankton from the sediment s (Brunberg and Blomqvist 2003, Ståhl-Delbanco et al. 2003, Bykova et al. 2006). Interestingly, Microcystis still crashed in Gull Lake during the summers of 2011-2012 despite the fact that zebra mussels persiste d at deeper, cooler depths below the thermocline throughout the entire period of high mussel mortality on the epilimnetic 95 sediments (White et al. 2015). This implies that Microcystis promotion in low-nutrient lakes requires the direct contact with Dreissena afforded within the epilimnion, which is congruent with both the selective rejection and nutrient excretion hypotheses. Our study also constitutes the first long-term analysis of Microcystis dynamics in a low- nutrient inland lake, an uncharacteristic niche for this harmful algal bloom-forming species, and one of very few long-term ( 10 years) studies of natural populations of Microcystis in general (Utkilen et al. 1996, Liu et al. 2011). We found considerable inte rannual variation in Microcystis biomass, timing of maximum biomass, relative abundance, colony size, and microcystin in oligotrophic Gull Lake (Figs. 23, 24). The population maximum occurred significantly earlier during warmer years, irrespective of Driessena presence (Figs. 24, 25b), indicating that warmer temperatures expedite the seasonal su ccession of the phytoplankton community to Microcystis. Where we found significant temperature effects on Microcystis dynamics, spring-summer averages had better predictive power than either average spring or average summer temperatures alone (Table 7), suggesting that the influence of temperature is integrative and cumulative over the entire growing season. Dynamics of partic ulate (cell-bound) microc ystin concentrations closely tracked Microcystis biomass (Fig. 23), and Microcystis biomass was a significant predictor of both relative Microcystis biomass and annual microcystin toxin concentrations in Gull Lake irrespective of Dreissena presence (Fig. 27). Therefore, in low-nutrient lakes where Microcystis is the only common producer of microcystin, Microcystis biomass may serve as a reliable proxy for water column microcystin concentrations. Large intra-annual variation, including a seasonal decline, in Microcystis colony size has been reported previously in Gull Lake (White and Sarnelle 2014) ( see Chapter 1). 96 Our observations of Microcystis dynamics in low-nutrient Gull Lake in the presence of Dreissena are similar to some aspects of studies of Microcystis populations in eutrophic lakes. A long-term (11-year) monitoring study of Microcystis dynamics in eutrophic Lake Taihu, China also found that temperature was a primary driver of biomass, which increased markedly as water temperature exceeded ~20 °C, with peak biomass occurring during the summer months (Liu et al. 2011). The annual temporal pattern of Microcystis biomass and microcystin we observed in Gull Lake (population maximum in July-August) is also similar to that reported for eutrophic lakes in Germany and England (Reynolds 1973, Fastner et al. 1999), although the annual decline of the population in Gull Lake tended to occur earlier and more rapidly following the population maximum than is reported in eutrophic lake s (Reynolds and Rogers 1976, Reynolds et al. 1981, Bigham et al. 2009, Imai et al. 2009). Our results contrast markedly, however, with respect to many other critical drivers of Microcystis reported from regional, comparative studi es of eutrophic lakes. Cyanobacterial dominance of the phytoplankton community t ypically occurs at high total phosphorus concentrations (i.e., in eutrophic lakes) (Smith et al. 1987, Watson et al. 1997, Downing et al. 2001), yet we observed Microcystis achieving up to 80% dominance in a lake with a mean TP of only 7.59 µg L-1 (Fig. 23). Dominance by cyanobacteria in eutrophic lakes has been hypothesized to result from the competitive advantage of nitrogen-fixing species at low TN:TP (< 29:1) (Schindler 1977, Smith 1983). Howeve r, this open-water nutrient ratio hypothesis cannot explain the occasional dominance of Microcystis in oligotrophic Gull Lake, since Microcystis cannot fix nitrogen and the open-water DIN:TP ratio alone is ~42:1 (Table 5), which is consistent with other studies that argue TN:TP and nitroge n fixation are not adequate for explaining cyanobacterial dominance in lakes (Downing et al. 2001, Ferber et al. 2004). In fact, we did not 97 find any significant influence of TP on annual Microcystis biomass. A recent study identified nutrients over temperature as the primary driver of total cyanobacteria biovolume across lakes of all trophic states, although Microcystis biovolume specifically exhibite d a substantially larger response to temperature than nutrients relative to other cyanobacteria (Rigosi et al. 2014), highlighting the need for additional species-specific data (Pitcher 2012). Low light availability, depleted carbon dioxide concentrations, and elevated pH are also associated with or hypothesized to competitivel y favor cyanobacteria in eutrophic lakes (Smith 1986, Shapiro 1997, Bigham et al. 2009). However, annual light climate ( k range: 0.32-0.54 m-1) did not explain a significant proportion of the residual variation in Microcystis biomass. Dissolved inorganic carbon (DIC) limitation is also unlikely to explain years of high biomass given that Gull Lake always contains high concentrations of DIC (Hamilton et al. 2009) and, although the pH can be high (range: 8.1-9.0) and potentially selective for phytoplankton that can utilize bicarbonate, this capab ility is not exclusive to Microcystis . Our finding that most environmental variables that show significant in fluences at the regional scale had little or no predictive power in a single lake is perhaps not surprising, given that the among-lake variation in these studies is typically enormous relative to the inter-annual variation observed within an individual lakeŠlet alone a lake of lo w productivity (Tillmanns and Pick 2011). Of course, the relative importance of different environmental drivers can depend on the temporal and spatial scales of the analysis (Reynolds 2007, Tillmanns and Pick 2011), and our consideration of only annual means may conceal impor tant relationships that exist at weekly or even monthly scales. For example, within a single eutrophic lake in Ontario, microcystin was not significantly related to a ny environmental variable (including TP, SRP, TN, NH 4+, pH, transparency, and temperature) at the seasona l scale, but at monthly and 6-day timesteps 98 significant relationships emerged, a nd different variables were significant at different timesteps (Tillmanns and Pick 2011). Therefore, our resu lts do not imply that environmental factors besides temperature are unimportant to Microcystis in low-nutrient lakesŠonly that they are relatively poor predictors of biomass at the annual scale. Cyanobacteria toxins, like microcystin, are ge nerally associated with eutrophic habitats, yet detectable levels have been documente d by recent surveys of low-nutrient lakes and reservoirs across a broad spatial extent, incl uding lakes that are not experiencing cultural eutrophication and also lack zebra mussels. However, detection is relatively rare (12-33% of lakes) and concentrations are generally always well below the World Health Organization™s drinking water guideline of 1.0 µg L-1(Giani et al. 2005, Bigham et al. 2009, Graham and Jones 2009, Sarnelle et al. 2010). Surprisingly, the colonial cyanobacterium Gloeotrichia echinulata has also been reported to be on the rise in oligotrophic and mesotrophic lakes that lack Dreissena in the northeastern United States, where it has produced detectable levels of microcystin (Carey et al. 2008, Carey et al. 2012a). Microcystin concentrations in low-nutrient Midwestern lakes invaded by zebra mussels are also significantly higher than predicted by TP (Knoll et al. 2008, Sarnelle et al. 2010). For example, the model of Giani et al. (2005) for uninvaded lakes spanning the productivity gradient predicts an average microcystin concentration of 0.003 µg L-1 for Gull Lake (7.59 µg L-1 TP), when average annual concentrations are actually as much as two orders of magnitude higher (0.022-0.343 µg L-1). Though cyanobacteria toxins in the open waters of low-nutrient lakes are generally below levels trigge ring public health concerns, their occasional presence in these systems is still noteworthy since they are the lakes that are frequently associated with drinking and recreational uses. 99 Microcystis strains that dominate populations in oligotrophic lakes are likely to differ markedly in their growth and resource-use phys iology from those that dominate populations in eutrophic lakes, given that resource levels can vary several fold across the lake productivity gradient (Wetzel 2001). For example, strains from low-nutrient lakes, including Gull Lake, have lower maximum intrinsic growth rates than thos e from nutrient-rich lakes when grown in a common garden, indicating genetically based variation in growth (see Chapter 4). The half-saturation constant (Km) for phosphorus-dependent growth of Microcystis isolated from nutrient- rich lakes is higher (5.9 µg L-1) (Holm and Armstrong 1981, Nalewajko and Murphy 2001) than available phosphorus concentrations typically found in Gull Lake (Table 5) and other low-nutrient lakes, suggesting that Microcystis strains growing in low-nutrient lakes are either always growing well below maximal rates or possess a lower K m conducive to growth in a chronically nutrient-poor habitat. Net population growth rates observed in Gull Lake (0.02-0.11 day-1) are comparable to the median maximum intrinsic growth rate (0.15 day-1; see Chapter 4) determined for 11 different Gull Lake Microcystis strains, lending support for the latter prediction. Large intraspecific variation in Microcystis has ecological consequences (White et al. 2011, White and Sarnelle 2014) (see Chapters 1, 4), and so there is preced ent to expect it should facilitate adaptation to growth in low-nutrient lakes. Freshwater ecosystems are increasingly vulnerable to and impacted by numerous anthropogenic stressors associated with global changeŠparticularly cultural eutrophication and climate changeŠboth of which are known to prom ote nuisance levels of toxic cyanobacteria (Carpenter et al. 1992, Carpenter et al. 1998, P aerl and Huisman 2008, O'Neil et al. 2012, Rigosi et al. 2014). Consistent with climate change forecasts, Microcystis biomass was generally elevated during warmer years in a low-nutrient lake, and microcystin concentrations closely 100 tracked biomass. However, Microcystis drivers and dynamics in low-nutrient Gull Lake differ in many fundamental respects from eutrophic syst ems, resulting in large part from strong interactions with invasive zebra mussels. Zebra mussel invasion of low-nutrient lakes represents yet another ‚catalyst™ for the global increase of harmful cyanobacteria (Paerl and Huisman 2009). Interactions among abiotic and biotic drivers, particularly temperature and invasive Dreissena, can result in complex responses by Microcystis in low-nutrient lakes that are not anticipated by current limnological paradigms or climate change forecasts. Thus, predicting the response of a species to climate change may require, at minimu m, quantification of temperature responses of both the focal species and species that strongly interact with it. Consequently, monitoring of intact communities with respect to climatic variables seems essential for predicting the community- and ecosystem-level consequences of an era of unprecedented global change. Acknowledgments We thank J. Berry, J. Chiotti, A. DePalma- Dow, L. Dillon, E. Fergus, S. Flemming, T. Geelhoed, B. Hanna, D. Hopper, G. Horst, M. Iadonisi, C. Kissman, C. Kozel, K. Lincoln, E. Milroy, J. Northrop, D. Raikow, N. Sarnelle, T. Sarnelle, A. Schuerer, M. Schuetz, T. Toda, A. Wilson, D. Weed, and J. White fo r assistance in the field and lab. S. Bassett, T. Brownell, N. Consolatti, A. Fogiel, E. Litchman, and M. W illiams provided tremendous logistical support. G. Mittelbach and P. Soranno provided feedback on an earlier version of the manuscript. This research was funded by the Environmental Pr otection Agency (Ecology and Oceanography of Harmful Algal Blooms/2004-Science to Achiev e Results-C1, project RD83170801), the National Science Foundation (Division of Environmental Biology-0841864, Division of Environmental 101 Biology-0841944), Michigan State University (Summer Fellowship to J. White), and the Gull Lake Quality Organization. 102 CHAPTER 4 GROWTH VARIATION AMONG STRAINS OF THE HARMFUL CYANOBACTERIUM, MICROCYSTIS AERUGINOSA, ACROSS A LARGE PRODUC TIVITY GRADIENT OF LAKES Abstract The toxic cyanobacterium Microcystis aeruginosa now achieves non-trivial densities in many low nutrient (oligotrophic) lakes following invasion by zebra mussels ( Dreissena polymorpha), which is unexpected given that this species is typically associated with high nutrient (eutrophic) lakes. We expl ored the extent of biological variation in growth traits within this species, which might enable it to succeed in these ecologica lly disparate habitats. Using common-garden laboratory growth assays, we quantified the maximum intrinsic growth rates of 18 colonial strains of M. aeruginosa recently isolated from 11 Michigan lakes spanning the entire productivity gradient, from oligotrophi c to hyper-eutrophic (total phosphorus range: 7.6-196 g L-1). Microcystis strains possessed fixed variation in maximum growth rate as a function of source lake TP, with strains from eutrophic and hyper-eutrophi c lakes growing up to ~7 times faster than strains from oli gotrophic lakes. Strains from high-nutrient lakes also had a significantly greater probability of becoming single-celled during the first 1.0-1.5 years in lab culture, and strains grew faster as single-cells compared to their natural colonial growth habit, emphasizing that caution should be exercised when relating studies of single-celled lab strains to natural populations. Our results provide evidence that M. aeruginosa populations are genetically adapted to grow under specific local resource cond itions, which may help to broaden the species™ ecological niche and could strongly influence its response to global change. 103 Introduction Harmful algal blooms (HABs) occur worldw ide in nutrient-polluted waters, produce toxins and skin irritants, foul the taste and odor of drinking water, and can cause illness (Chorus and Bartram 1999, Huisman et al. 2005). HABs app ear to be increasing (Hallegraeff 1993) and are projected to increase with climate warming and other facets of global change (Paerl and Huisman 2008, Paul 2008, Carey et al. 2012b, Elli ott 2012, Paerl and Paul 2012). Recent work also emphasizes the importance of intraspecifi c trait variation among st rains (genotypes) in explaining population and bloom dynamics fo r HAB-forming phytoplankton (Burkholder and Glibert 2009). For example, different strains of single HAB species have been shown to constitutively vary in their competitive ability for nutrients and light (Kardinaal et al. 2007b), nutrient uptake rate (Sinclair et al. 2009), vulnerability to grazing (White et al. 2011, White and Sarnelle 2014), size (Wilson et al . 2006), toxicity (Vezie et al. 1998, Mikalsen et al. 2003, Saker et al. 2005), and growth rate (Wilson et al. 2010, Calbet et al. 2011). Therefore, holistic traits of the population might reflect the relative abundance and respective phenotypes of the different strains present (Kardinaal et al. 2007a, Kardinaal et al. 2007b, van Gremberghe et al. 2009). Large biological variation within individual HAB species may thus play a key role in the extensive spatio-temporal variat ion observed in the traits of HAB populations. Whether these ecological traits vary predictably across populations of a single HAB species according to important ecological gr adients is uncertain. Cyanobacteria, characteristic of the phytoplankton community of phosphorus-enriched (eutrophic) lakes (Watson et al. 1997), are genera lly implicated in freshwater HABs. Numerous empirical and predictive models have been deve loped to account for variation in cyanobacteria biomass production, toxicity, and response to gl obal change (Smith 1985, Trimbee and Prepas 104 1987, Downing et al. 2001, Kotak and Zurawell 2007, Bigham et al. 2009). However, models forecasting the abundance of particular HAB-for ming cyanobacteria species are less developed, and better species-specific information is st ill needed (Pitcher 2012). Of the freshwater cyanobacteria, the colonial species Microcystis aeruginosa is most widely distributed and produces the most common class of toxins, th e microcystins (Chorus and Bartram 1999). Large genetic variation has also been documented recently within M. aeruginosa (Wilson et al. 2005, Tanabe et al. 2007), though knowledge of the extent to which this raw genetic variation translates into ecologically relevant phenotypic variation, and its implications, is still limited (White et al. 2011). Despite Federally-mandated phosphorus c ontrol programs in the United States, populations of M. aeruginosa have emerged in fiunexpectedfl ha bitatsŠspecifically, oligotrophic (low-nutrient) lakes where cyanobacteria are not otherwise predicted to be abundant, following invasion by zebra mussels ( Dreissena polymorpha). Biomass of M. aeruginosa is ~4 fold and microcystins are ~3-8 fold highe r in low-nutrient lakes (< 10-20 g L-1 total phosphorus, TP) invaded by zebra mussels relative to similar, but non-invaded, lakes (Knoll et al. 2008, Sarnelle et al. 2010). Low-nutrient lakes are a fundamentally different ha bitat characterized by wholly different resource levels and phytoplankton communities from those currently assumed to be requisite for M. aeruginosa. Self-sustaining populations of M. aeruginosa in oligotrophic lakes indicate that this species ha s a much broader ecological niche than previously assumed, and perhaps this is a result of local adaptation facilitated by large intraspecific variation. Consistent with this idea, survivorship of M. aeruginosa strains isolated from low-nutrient lakes into nutrient-rich culture medium is much lower than for strains isolated from nutrient-rich lakes (Wilson et al. 2005), suggesting that strains might be adapted to grow under specific, local 105 nutrient regimes. Microcystis aeruginosa is therefore an ideal candidate for examining ecological trait variation within a species, and how such variation might mediate the responses of biodiversity to accelerating global change. Fu rthermore, better predictive models for M. aeruginosa can be constructed if variation in critical population parameters can be correlated to important ecological gradients (Rojo 1998). Given that phosphorus availability can vary by more than 2 orders of magnitude across a lake productivity gradient from oligotrophic to hyper-eutrophic (Wetzel 2001), and the degree to which different strains of harmful phytoplankton are capable of exploiting different resources, we predicted that lake nu trient status structures M. aeruginosa populations by selecting for strains with a trait repertoire that is advant ageous for the local environment (Litchman and Klausmeier 2008). Relative to strains from eutr ophic lakes, we hypothesized that strains from oligotrophic lakes have lower maximal intrinsic growth rates ( rmax ), because their growth in nature is chronically nutrient-limited which might preclude synthe sis of the cellular machinery required to sustain high rates of cell division (Klausmeier et al. 2004); whereas high rmax under eutrophic conditions should be advantageous to compensate for increased grazing pressure and competition for resources due to increased grazer and algal biomass at higher nutrients (Dillon and Rigler 1974, Hanson and Peters 1984). We te sted this hypothesis using common-garden laboratory growth assays employing numerous strain s isolated from lakes varying widely in TP, and monitored changes in growth characterist ics of the strains over time in lab culture. 106 Methods Isolation and maintenance of lab strains Water samples were collected via two pooled cas ts of an integrating tube sampler (12 m length × 2.5 cm i.d.) from the mixed layer of 11 lakes distributed across southern Michigan between 5 July and 19 August, 2011, and from 14 lakes between 6 August and 12 September, 2013 (Fig. 28, Table 8). Eight lakes were sampled in both years. The lakes ranged widely in potential primary productivity from o ligotrophic to hyper-eutrophic (7.9-196.8 g L-1 TP, Table 8). Microcystis was isolated from these samples under a dissecting microscope (16×, Leica MS5) by pipetting individual colonies through a series of six washes in sterile 0.5× WC-S growth medium within a well plate (Corni ng, Inc.), prior to being transferred into i ndividual 20 mL tubes of growth medium (White et al. 2011). For any given lake 27-100% of these isolates successfully established (i .e., grew and lacked other algal contaminants), and these were given unique strain designations identifying the originating lake, year, and strain number (e.g., F11-05; Table 8). We found no systematic variation in es tablishment success of strains as a function of source lake TP (linear regression of arcsine squareroot transformed proportions, n = 25, p = 0.36, R2 = 0.04). Once established, strains were maintained in 200 mL batch cultures of 0.5× WC-S at 23 °C and ~80 µmol m -2 s-1 on a 12:12 h light:dark cycle (Fig. 29), with an inoculum of culture transferred to fresh, sterile medium on a monthl y basis. At the time of isolation, all strains conformed to current morphological criteria for M. aeruginosa, including the production of buoyant, mucilaginous colonies with cell diameters of ~4-6 m (Otsuka et al. 2001, Wehr and Sheath 2003, Cronberg and Annadotter 2006). Thus, to the best of our ability, we identified all strains as belonging to the same species. 107 Figure 28. Map indicating the locations and trophic state (based on total phosphorus) of all 2011 and 2013 Microcystis aeruginosa source lakes. 108 Figure 29. The Microcystis aeruginosa culture collection. The right panel illustrates the buoyant, colonial attributes of the recently isolated strains. Table 8. Limnological data on the Microcystis aeruginosa source lakes, and a summary of the sampling, isolation, and establishment of all lab strains. Lake (designation) County (Michigan) TP (µg L-1)SRP (µg L-1)Chl-a (µg L-1)Secchi (m) Sample date(s) Colonies isolatedStrains established% strains established Baker (BK) Barry 28.04.633.71.52011-08-08 2013-08-07 7 106 386 30Baseline (BS) Allegan 36.13.647.11.02011-08-08 2013-08-09 9 125 556 42Bristol (BR) Barry 13.81.17.31.52013-08-06 4375 Ford (F) Washtenaw 65.010.256.20.92011-08-01 2013-08-15 9 145 456 29Gull (G) Kalamazoo 7.91.33.74.52011-07-05 2013-08-08 11 127 1064 83Kent (K) Oakland 23.63.323.12.22013-08-15 10770 Lansing (LG) Ingham 17.15.15.52.92011-08-05 2013-08-14 7 155 771 47Lee (LE) Calhoun 9.01.94.03.12013-08-06 10550 Little Long (LL) Barry 8.01.04.14.62011-07-12 2013-08-08 8 108 8100 80MSU1 (L1) Ingham 163.5155.812.22.12011-08-19 3133 MSU2 (L2) Ingham 196.87.3240.80.42011-08-16 261350 MSU3 (L3) Ingham 128.74.153.30.62011-08-06 11100 MSU4 (L4) Ingham 124.78.499.20.52013-09-12 15427 Payne (P) Barry 11.21.47.53.02013-08-07 3267 Sherman (S) Kalamazoo 13.72.89.12.92011-08-08 2013-08-07 7 106 386 30Sixteen (SX) Allegan 8.81.05.03.92013-08-07 121083 Wintergreen (W) Kalamazoo 47.82.721.42.42011-08-02 2013-08-08 9 89 5100 63109110 Table 8. (cont™d) A strain was considered established if the isol ate grew and the seed cu lture was devoid of any other algal contaminants. TP = total phos phorus; SRP = soluble reactive phosphorus; Chl-a = chlorophyll-a; Secchi = Secchi disk transparency.111 Growth rate assays Common-garden growth assays were conducted with the gene ral design as follows. Fresh 20 mL cultures of strains were initiated 7 days prior to an assay to insure M. aeruginosa was exponentially growing. Individual Microcystis colonies were then inoculated via pipette (1 µL) into randomized, separate wells containing 0.5 mL sterile 0.5× WC-S within 8-well chambered slides (Nunc Lab-Tek II Chamber Slide System) (Wilson et al. 2010). Once inoculated (day 0), colonies were photographed every 2 days for 6 days at 100× using a light microscope interfaced with a digital camera (Fig. 30; Nikon Eclipse E600 ). Measurements, added to the images with computer software (Spot Advanced, Diagnostic In struments), were made of colony surface area and depth (the straight line length perpendicular to the greatest linear dimension); colony volume (µm3) was determined as the produc t of surface area and depth (Wils on et al. 2010). Growth rate was determined as the slope of the linear re gression of natural logarithm-transformed colony volumes over time. Unless noted otherwise, one co lony per strain was employed per treatment in a given experiment. Since colonialit y is a characteristic trait of M. aeruginosa in nature (Wehr and Sheath 2003), all growth assays were performed using colonial strains that had been in culture for less than 1.5 years, unless noted otherwise. This contrasts with many previous lab studies of M. aeruginosa that have utilized old, single-celled culture collection strains. Furthermore, since all strains employed in a gi ven experiment were the same age and were recently isolated, concerns arising from evolutio n in culture were minimized (Burkholder and Glibert 2009, Lakeman et al. 2009, Demott and Mckinney 2015). 112 Figure 30. Sequence of digital micrographs depicting growth of an individual colony of Microcystis aeruginosa during the course of a 6-day growth assay. Colonies were photographed at 100× and all images are shown to scale. The st rain pictured is F11-05, isolated in 2011 from eutrophic Ford Lake, Michigan. Photo credits: Jeffrey D. White. Day 0 Day 2Day 4 Day 6113 Assaying growth of individual M. aeruginosa colonies is necessary and advantageous because, unlike batch culture assays, this pe rmits controlling for colony size and inoculation density effects on growth rate, since small colonies grow faster than large colonies (Wilson et al. 2010). To further minimize the confounding effects of initial colony size and shape on growth rate, round colonies of approximately the same equi valent diameter were selected for each strain to the fullest extent possible using the microscope™s ocular micrometer. To quantify variation in maximum growth ra te among strains, we performed a growth assay employing 19 colonial strains isolated from 11 lakes in 2011, selected in stratified random fashion according to source lake trophic state (TP range: 7.9-196.1 g L-1). Growth conditions were identical to those described for general culture maintena nce, including saturating phosphate (480 g L-1 SRP). To validate the individual colony assay technique, a randomized subset of 8 of the above strains were grown in parallel 100 mL flasks of 0.5× WC-S (batch cultures) and chambered slides (individual colonies). All strains were gr own in duplicate. The individual colony assay was conducted as above. Batch cultures were subsampled every 2 days for 8 days, starting on day 4, by filtering 15 mL of thoroughly-mixed cu lture onto A/E filters for chlorophyll- a analysis (see below). Growth rates of batch cultures were determined as above, using chlorophyll- a. We conducted a second individual colony, maxi mum growth rate assay employing all the still-available colonial strains isolated in 2013 from Gull, Lans ing, Kent, and Wintergreen Lakes to explore the range of maximum growth rates within individual populations from lakes of widely different productivity (oligotrophic to eutrophic, TP range: 7.9-47.8 µg L-1). To capture as much of this biological vari ation as possible, we pooled the data with the first experiment for 114 the 2011 strains that originated from the same f our lakes, to give a total sample size of n = 6-11 strains per M. aeruginosa population. Monitoring changes in growth habit Since M. aeruginosa typically loses the ability to form colonies over the first few years in lab culture (J. White pers. obs.)(Zhang et al. 2007), we made observations at 1.0 and 1.5 years post-isolation of the growth habit (colonial versus single-celled) for all strains isolated in 2013 (n = 73) at 100× under a light microsc ope. A strain was categorized as single-celled if and only if it was purely single-celled, since a transitioning st rain will often produce a mixture of diffuse colonies and single cells (J. White, pers. obs.). To assess any changes in maximal growth ra tes as a result of sw itching growth habits from colonial to single-celled, we re-assaye d maximal growth rates for those 2011 and 2013 strains that had become single- celled after ~2.0 years in lab culture, and whose maximal growth rates were previously determined while sti ll colonial. These assays were conducted in 2013 (2011 strains) and 2015 (2013 strains) using the batch-culture me thod described above. Results were then pooled from the two experiments. Lab analyses Subsamples of source lake water and growth medium were partitioned via filtration through A/E filters (Pall, 1.0 µm pore size) and then immediatel y frozen for later analysis. Analysis of available phosphorus (as SRP) was performed on the filtrate using the molybdenum- blue method and long-pathlength spectrophotome try (Murphy and Riley 1962). Total phosphorus 115 was analyzed on unfiltered samples as above for SR P, following persulfate digestion of organic material in an autoclave (Menzel and Corwin 1965). Chlorophyll-a was measured fluorometrically follo wing 12-hour dark extraction of frozen A/E filters in 10 mL of cold 95% ethanol (Welschmeyer 1994). Particulate microcystin was measured by ELISA (enzyme-linked immunosorba nt assay; Envirologix QuantiPlate Kit for Microcystins visualized on a LabSystems Multis kan Microplate Reader) following three pooled 45 min, 10 mL extractions of A/E filters in 75% methanol (Harada et al. 1999, White et al. 2011). Microcystis biomass was quantified from measuremen t of 2-dimensional surface areas of colonies at 100× as above. These measurem ents were converted to dry biomass using a regression that estimates cell density from col ony surface area (Sarnelle et al. 2012, Horst et al. 2014). Microcystin quotas (mg toxin mg -1 Microcystis biomass) were then calculated for lab strains. Statistical analyses We used source lake TP as the principal predictor for variation in M. aeruginosa growth traits, since TP is routinely used to classify lake trophic state (Wetzel 2001) and is also ecologically relevant to M. aeruginosa and HAB-forming cyanobacter ia (Watson et al. 1997). We performed linear regressions to test if maximum growth rates of M. aeruginosa strains (determined via the individual colony method) varied as a function of their source lake TP and initial colony size, and to test if strain microcystin quota varied as a function of either source lake TP or maximum growth rate . Microcystin quotas were log-transformed prior to analysis to meet normality requirements. Residual plots did not indicate any systematic departures from statistical model assumptions. We used the coeffi cient of variation (CV; ratio of the standard 116 deviation to the mean) to assess the relative ex tent of within-population variation in maximum growth rate. We used paired t-tests to compare growth rates determined using the individual colony versus batch culture assays, and to test for differences in maximum growth rate for strains that changed growth habit from colonial to sing le-celled over time in culture. To test whether strains from high-nutrient lakes were more likely to become single-celled during the first 1.0 and 1.5 years in culture, we performed logistic regressions. All statistical analyses were performed using R Version 3.1.3 (R Foundation for Statistical Computing). Results Growth rate assays The two assay methods (individual colony vers us batch culture) yielded similar results that were not significantly different (paired t-test, n = 8, df = 7, p = 0.71; Fig. 31), validating the individual colony method. Maximu m intrinsic growth rates (rmax ) of M. aeruginosa strains were significantly, positively related to their source lake TP (linear regression: n = 18, p = 0.030, R2 = 0.26; Fig. 32). Observed growth rates ranged nearly 7-fold from 0.08 d-1 (source lake TP = 7.9 g L-1) to 0.55 d-1 (source lake TP = 196.1 g L-1). One colony (source lake TP = 16.6 g L-1) exhibited negative growth during the experime nt and was omitted from analysis, leaving n = 18 for the experiment. As assessed with the coefficient of variation, variation in rmax within four Microcystis populations spanning the lake productivity gradient (oligotrophic to eutrophic, 7.9-47.8 g L-1) was generally similar (CV = 0.30-0.52), although stra ins from moderately productive Kent Lake (TP = 23.6 g L-1) exhibited the greatest variation (CV = 0.70; Table 9). Within a single oligotrophic lake population (Gull Lake, TP = 7.9 g L-1), rmax ranged from 0.08-0.37 d-1 117 Figure 31. Comparison of maximum growth rates of eight strains of Microcystis aeruginosa as determined from two different methods: indivi dual colony and batch culture assays. The two methods yielded growth rates that were not significantly different from each other. Error bars denote ± SE. 118 Figure 32. Variation in maximum in trinsic growth rate of Microcystis aeruginosa strains isolated from lakes spanning a large lake productivity gradient, from oligotrophic to hyper- eutrophic (as total phosphorus). Growth rates were determined from the change in volume of individual colonies during a 6-day growth assay with saturating nutrients ( n = 18 strains). Growth rate data (batch culture assays of M. aeruginosa) from Wilson et al. (2006) are shown for comparison (squares). The linear regression is fo r the current study only. Note the log scale on the x-axis. 119 Table 9. Within-population variation in maximum intrinsic growth rate (d -1) of Microcystis aeruginosa, as determined with individual colony grow th assays. The four source lakes range from oligotrophic to eutrophic (7.9-47.8 µg L-1 total phosphorus) from top to bottom. Data were pooled for all 2011 and 2013 strains assayed from the given lake population. The coefficients of variation (CV) and sample sizes ( n) are given. Lake Median (d -1) Min (d -1) Max (d -1) CV nGull 0.15 0.080.370.52 11 Lansing 0.22 0.140.300.30 6 Kent 0.21 0.010.460.70 6 Wintergreen 0.29 0.120.490.42 6 120 (n = 11 strains), and for a single eutrophic lake population (Wintergreen Lake, TP = 47.8 g L-1) from 0.12-0.49 d -1 (n = 6 strains; Table 9). We found no relationship between microcystin quota and either source lake TP or rmax (linear regressions, n = 18, p > 0.57, R2 < 0.02; Fig. 33) during exponential growth under saturating resources. Since we deliberately selected colonies of similar initial size, we also found no relationship between rmax and initial colony size in our individual colony experiment (linear regression: n = 18, p = 0.86, R2 < 0.01), obviating the need to statistically control for the effects of initial colony size on growth rate. Monitoring changes in growth habit After 1.0 and 1.5 years in culture, Microcystis strains from more productive lakes were significantly more likely to have become purely single-celled, as compared to strains isolated from less productive lakes (data fo r 1.5 years: logistic regression, n = 73, p = 0.005, odds ratio = 1.04; Fig. 34). Of the 33 strains originating from oligotrophic (TP < 10 g L-1) lakes, 79% were still producing colonies after 1.5 years in culture, whereas of the 40 strains originating from more productive lakes (11.2 TP 124 g L-1), only 50% were still producing colonies. Of the 18 strains isolated in 2011 that were initially assayed as colonies (Fig. 32), 6 became single-celled after a period of ~2 years year s in culture; likewise, 6 of the 22 previously assayed strains isolated in 2013 (Table 9) became single-celled. When these strains were assayed again as single-celled organisms, rmax was significantly higher than when colonial. Since the result was the same for both sets of strains, results were pooled across the two growth assays (paired t-test, n = 12, df = 11, p = 0.005; Fig. 35). 121 Figure 33. Variation in microcystin quota of the Microcystis aeruginosa strains (n = 18) assayed in Fig. 32 as a function of (A) source lake total phosphorus and (B) maximum intrinsic growth rate. 122 Figure 34. Logistic regression of growth habit (colonial or single-celled) of all Microcystis aeruginosa strains (n = 73) isolated in 2013 versus sour ce lake total phosphorus, 1.5 years post-isolation into lab culture. All strains were colo nial at the time of isolation. Data points are randomly jittered vertically to avoid over-plotting. 123 Figure 35. Comparison of maximum growth rates as a function of growth habit (colonial or single-celled) for twelve Microcystis aeruginosa strains. Strains were all assayed initially as colonies (see Fig. 32, Table 9), and then again after disaggregating following a period of ~2 years in lab culture. 124 Discussion We identified extensive variation in maximum intrinsic growth rate ( rmax ) among Microcystis aeruginosa strains that were recently isol ated from lakes spanning a large productivity gradient. In support of our hypothesis, strains isolated from eutrophic and hyper- eutrophic lakes grew up to ~7 times faster than strains originating from oligotrophic lakes under saturating nutrients (Fig. 32, Table 9), and these differences were not driven by colony size. Since all strains were grown in a common garden, these differences in maximal growth rates are assumed to be genetically based. Our results, obt ained from testing colonial strains, are also more relevant to natural populations of M. aeruginosa since coloniality directly affects grazing vulnerability and migration velocity in addition to growth rate (Visser et al. 1997, Wilson et al. 2010, White and Sarnelle 2014); thus, coloniality could influence fitness and ecological trade-offs (Litchman and Klausmeier 2008). Furthermore, all strains employed in a given experiment were the same age, recently isolated, and shared the same transfer regimes, limiting the potential effects of evolution in culture (Lakeman et al. 2009, Demott and Mckinney 2015). Therefore, our study provides evidence for possible local adaptation by M. aeruginosa to ecologically disparate habitats and supports the conclusion that large, genetically based trait variation in this species has ecological consequences (White et al. 2011). To our knowledge, only two previous studies (Wilson et al. 2005, Wilson et al. 2006) have explored trait variation among M. aeruginosa strains originating from lakes differing widely in productivity (12 Michigan lakes, 12.8-101.8 g L-1 TP; strains isolated in 2002). Wilson et al. (2006) reported rmax for these strains ranging from 0.17-0.46 d-1 in batch culture, congruent with those reported here (Fig. 32, Table 9), although they did not test the influence of lake productivity on growth rate. Using the rmax data of Wilson et al. (2006, their Table 1) and 125 the source lake TP data of Wilson et al. (2005, their Table 1), we found a positive, but non- significant, relationship between rmax and source lake TP (linear regression: n = 12, p = 0.1, R2 = 0.24). However, the sample size and source lake TP range in their study was smaller relative to the present study, and so the lack of statistical significance ma y be due to insufficient power. Maximum intrinsic growth rates for M. aeruginosa reported elsewhere under similar growth conditions [0.46 d-1, Fujimoto et al. (1997); 0.48 d -1, Reynolds (2006); ~0.30-0.50 d-1, Wilson et al. (2010); 0.63 d-1, Seip and Reynolds (1995); 0.88 d-1, Nalewajko and Murphy (2001)] are consistent with those reported here for strain s isolated from higher-nut rient lakes, which is expected given that most M. aeruginosa employed in growth studies originates from productive habitats. Although possessing a fast growth rate should generally be advantageous for phytoplankton, the cellular machinery required to sustain high levels of growth in high-nutrient habitats may come at an energetic cost when available nutrients, particularly phosphorus, are chronically low (e.g., in oligotrophic lakes) (Klausmeier et al . 2004). Thus, a trade-off should exist to maintain the fitness of a slow-growing strain adapted to a low-nutrient habitat. In general, growth rates tend to increase with habitat productivity for freshwater phytoplankton, but with a concomitant decrease in affinity for phosphate and an increase in the half-saturation constant (Km) for phosphate-dependent growth (Seip and Reynolds 1995, Spijkerman and Coesel 1998, Litchman and Klausmeier 2008), suggesting th at slow growth may trade-off with an improved ability to glean nutrients and subsis t at low resource levels. We predict that M. aeruginosa strains originating from o ligotrophic lakes have lower Km for phosphate and are therefore superior at maintaining positive somatic growth under conditions of chronic nutrient limitation, relative to strains originating from eutrophic and hyper-eutrophic lakes. Microcystis 126 has a reported Km of 5.9 µg L-1 for phosphate-dependent growth (Holm and Armstrong 1981), a high resource concentration that is unlikely to be encountered in oligotrophic lakes (as SRP, Table 8), which suggests we should expect large intraspecific variation in Km. The ability to make broad generalizations about these crit ical population parameters across the lake productivity gradient would be useful fo r modelling and lake management (Rojo 1998, Mieleitner and Reichert 2008), given the complex dynamics of local Microcystis populations at the strain level (Kardi naal et al. 2007a). Other ecological trade-offs could al so exist, including that between rmax and grazing resistance (Agrawal 1998, Demott and Mckinney 2015). These trade-offs could all have a genetic basis, although plastic responses to environmental gradients (such as N versus P limitation) are also common in phytoplankt on (Van Donk 1997, Van Donk et al. 1997). Nonetheless, all of these traits and trade-offs are traditionally measured between species or across broad taxonomic groups, rather than among strains of the same species (White et al. 2011). Additional work is required to identify and measure these and other possible trade-offs in M. aeruginosa, particularly as a function of important ecological grad ients like lake productivity. Interestingly, we also found a relationship between the probability of a strain becoming single-celled during the first 1.5 years in culture and source lake productivity, with the log of the odds increasing by ~1.0 for a 1.0 µg L-1 increase in lake TP (Fig. 34). If the switch in growth habit from colonial to single-celled is triggere d by mutation or is otherwise a function of the number of generations spent in culture, strains from higher-nut rient lakes may disaggregate sooner given their faster growth rates. We also found that maxi mal growth rates of individual strains were significantly higher for the single-celled growth habit than for colonies (Fig. 35), likely reflecting the growth costs incurred from being colonial (Wilson et al. 2010). These 127 observations reiterate the importan ce of characterizing traits of M. aeruginosa with recently isolated, colonial strains if the intention is to ultimately relate studies of laboratory strains to natural populations. Using neutral genetic markers, recent st udies have documented extensive genetic variation within M. aeruginosa, although these studies also report that within-lake genetic variation (which is different from ecological variation) can be similar in extent to among-lake genetic variation (Wilson et al. 2005, Tanabe et al . 2009). Yet, this may reflect the fact that a phylogenetic analysis across a large ecological gradient (e.g., oligotrophic to hyper-eutrophic lakes) has yet to be made with a sufficiently large number of strains and lakes, since most molecular studies include few, if any, M. aeruginosa strains from low-nutrient lakes (but, see Wilson et al. 2005). Therefore, molecular studies may still be underestimating the extent of genetic (and thus, perhaps, ecological) variation within M. aeruginosa. We found no evidence for variation in microcystin quotas of M. aeruginosa strains, under conditions of saturating nutrients and exponential growth, as a function of either lake TP or rmax . This suggests that any observed variation in microcystin quota is primarily driven by environmental conditions (Van de Waal et al. 2009), since many laboratory strains of M. aeruginosa, including from both oligotr ophic and eutrophic habitats, already possess the genetic capability of synthesizing microcystin (Dyble et al. 2008, White et al. 2011). Indeed, recent studies of both laboratory strains and natural populations of M. aeruginosa demonstrate the importance of nitrogen availability for regulatin g microcystin quota relative to other factors (Downing et al. 2005, Horst et al. 2014). Ecologists have recently emphasized commun ity- and ecosystem-level implications of the genetic and phenotypic diversity inherent within and among populations (Neuhauser et al. 128 2003, Whitham et al. 2003, Long and Hay 2006, Johnson and Stinchcombe 2007, Burkholder and Glibert 2009). Large genetic and phenotypic variation is obviously important from an evolutionary standpoint, as it represents adaptive potential for species like M. aeruginosa. The large selection pressures imposed by globa l change may cause complex responses of M. aeruginosa by interacting with the biodiversity it harbors. Recent studies have found that toxic strains of M. aeruginosa grow faster at elevated temper ature and phosphorus compared to non- toxic strains (Davis et al. 2009), and M. aeruginosa strains artificially subjected to increasing temperature can evolve to grow at 35 °C (Huertas et al. 2011). Perennial M. aeruginosa populations emerging and persisting in oligotrophic lakes subsequent to zebra mussel invasion (Raikow et al. 2004, Knoll et al. 2008, Sarnelle et al. 2010) is another likely example. Understanding how different strains and populations of M. aeruginosa respond to variation in environmental conditions will enhance predictions fo r the response of this HAB species to global change; more generally, accounting for the dive rsity within and among ecological populations will be critical for conserving biodiversity and predicting population and community-level responses to rapid, large-scale changes in the environment. Acknowledgements We thank C. Kozel, T. Geelhoed, and M. Schmidt for assistance sampling the source lakes, and C. Kozel, T. Geelhoed, a nd S. Flemming for maintaining the M. aeruginosa cultures. A. Wilson and C. Klausmeier provided guida nce on methodology. Funding was provided by the National Science Foundation (Division of Environmen tal Biology-0841864, Division of Environmental Biology-0841944) , the Gull Lake Quality Organization, and the Robert C. Ball and Betty A. Ball Fisheries and Wildlife Fellowship at Michigan State University. 129 APPENDIX130 APPENDIX SURVEY OF ZEBRA MUSSEL (DREISSENA POLYMORPHA) STATUS IN OTHER MICHIGAN INLAND LAKES DURI NG RECENT WARM SUMMERS This appendix describes a qualitative survey conduc ted in conjunction with the study of heat- induced zebra mussel mortality in Gull Lake, Mi chigan (Chapter 2). The survey results were not included in the published manuscript that resulted from that study. Introduction and Methods Since water temperatures in Gull Lake should covary with water te mperatures in other inland lakes in the Lower Peninsula of Michigan, we would expect similar die-offs of D. polymorpha in other lakes during recent periods of warm summer temperatures if such temperatures are an important driver of morta lity. To explore whether our observations of high D. polymorpha mortality in Gull Lake ( see Chapter 2) represented an isolated case or a widespread occurrence, we surveyed lake associations and volunteer monitoring groups through the Michigan Lake and Stream Associati on (MLSA) and Michigan Clean Water Corps (MiCorps) list-serves and memb ership newsletters during the fall of 2013. We asked for any recent (beginning summer of 2010), qualitative observations of D. polymorpha in inland lakes, including declines, die-offs, or lack thereof. We conservatively scored a report as a fidecline or die-offfl if and only if the responder clearly de scribed a very recent (summer 2010- fall 2013) and substantial (i.e., readily visible to the casual observer) change in D. polymorpha abundance, relative to prior years. If the re porter expressly noted otherwise, or if it was unclear from their description, the report was scored as a fino decline.fl We looked for a pattern between the nature and timing of the reports we received and the re gional temperature conditions (Fig. 11) to assess the generality of our hypothesis. 131 Results We received status reports on D. polymorpha from 43 inland lakes in the Lower Peninsula of Michigan. In 36 (84%) of thes e lakes, obvious declines or die-offs of D. polymorpha were independently observed during the time period from summer 2010 Œ fall 2013 (Fig. 36). These reports consistently included detailed descriptions of reduced mussel densities on dock pilings and mooring lines, and a dramatic reduction or absence of mussels on substrates at shallow depths where they had been observed previously. Sixty-eight percent of the reported declines or die-offs occurred or were first noticed during the period from 2010-2012, which are the same three consecutive warm summers (Fig. 11) during which the highest levels of mussel mortality were observed in Gull Lake (Table 4, Fig. 15). These qualitative, statewide observations suggest that the large decline of D. polymorpha we observed in Gull Lake was probably not an isolated occurrence, although we readily acknowledge that any further interpretation of these findings would be purely speculative given the nature of the information. Acknowledgements We thank S. Brown, J. Latimore, Michigan La ke and Stream Associ ations, and Michigan Clean Water Corps for disseminating our survey, and for the dozens of citizen responders who provided detailed, insightful observations from their local lakes. 132 Figure 36. Locations of Michigan inland lakes for which recent Dreissena polymorpha status reports (decline or not during 2010-2013) were received during our qualitative survey of statewide lake associations. 133 LITERATURE CITED134 LITERATURE CITED Agrawal, A. A. 1998. Algal defense, grazers, and their interactions in aquatic trophic cascades. Acta Oecologica 19:331-337. Aldridge, D. W., B. S. Payne, and A. C. Miller. 1995. Oxygen consumption, nitrogenous excretion, and filtration rates of Dreissena polymorpha at acclimation temperatures between 20 and 32 °C. Canadian Journal of Fisheries and Aquatic Sciences 52:1761-1767. Allen, Y. C., B. A. Thompson, and C. W. Ramc haran. 1999. Growth and mortality rates of the zebra mussel, Dreissena polymorpha, in the Lower Mississippi River. Canadian Journal of Fisheries and Aquatic Sciences 56:748-759. Aminot, A., D. S. Kirkwood, and R. Kérouel. 1997. Determination of ammonia in seawater by the indophenol-blue method: Evaluation of th e ICES NUTS I/C 5 questionnaire. Marine Chemistry 56:59-75. Arnott, D. L. and M. J. Vanni. 1996. Nitrogen and phosphorus recycling by the zebra mussel (Dreissena polymorpha) in the western basin of Lake Erie. Canadian Journal of Fisheries and Aquatic Sciences 53:646-659. Austin, J. A. and S. M. Colman. 2008. A century of temperature variability in Lake Superior. Limnology & Oceanography 53:2724-2730. Baker, S. M., J. S. Levinton, J. P. Kurdziel, and S. E. Shumway. 1998. Selective feeding and biodeposition by zebra mussels and their relation to changes in phytoplankton composition and seston load. Journal of Shellfish Research 17:1207-1213. Bastviken, D. T. E., N. F. Caraco, and J. J. Cole. 1998. Experimental measurements of zebra mussel (Dreissena polymorpha) impacts on phytoplankton community composition. Freshwater Biology 39:375-386. Bern, L. 1994. Particle selection over a broad size range by crustacean zooplankton. Freshwater Biology 32:105-112. Beyer, J., P. Moy, and B. Stasio. 2011. Acute upper thermal limits of three aquatic invasive invertebrates: Hot water treatment to prevent upstream transport of invasive species. Environmental Management 47:67-76. Bigham, D. L., M. V. Hoyer, and D. E. Canf ield. 2009. Survey of toxic algal (microcystin) distribution in Florida lakes. Lake and Reservoir Management 25:264-275. 135 Boeckman, C. J. and J. R. Bidwell. 2014. Dens ity, growth, and reproduction of zebra mussels (Dreissena polymorpha) in two Oklahoma reservoirs. Pages 369-382 in T. F. Nalepa and D. W. Schloesser, editors. Quagga and zebra mussels: biology, impacts, and control. CRC Press, Boca Raton. Bogdan, K. G. and J. J. Gilbert. 1984. Body size and food size in freshwater zooplankton. Proceedings of the National Academy of Sciences 81:6427-6431. Bolnick, D. I., P. Amarasekare, M. S. Araújo, R. Bürger, J. M. Levine, M. Novak, V. H. W. Rudolf, S. J. Schreiber, M. C. Urban, and D. A. Vasseur. 2011. Why intraspecific trait variation matters in community eco logy. Trends in Ecology & Evolution 26:183-192. Both, C., S. Bouwhuis, C. M. Lessells, and M. E. Visser. 2006. Climat e change and population declines in a long-distance migratory bird. Nature 441:81-83. Bozarth, C. S., A. D. Schwartz, J. W. Shepar dson, F. S. Colwell, and T. W. Dreher. 2010. Population turnover in a Microcystis bloom results in predominantly nontoxigenic variants late in the season. Applied and Environmental Microbiology 76:5207-5213. Bridoux, M., M. Sobiechowska, A. Perez-Fuentetaja, and K. T. Alben. 2010. Algal pigments in Lake Erie dreissenids, pseudofeces and sediments, as tracers of diet, selective feeding and bioaccumulation. Journal of Great Lakes Research 36:437-447. Brunberg, A.-K. and P. Blomqvist. 2003. Recruitment of Microcystis (cyanophyceae) from lake sediments: the importance of litto ral inocula. Journal of Phycology 39:58-63. Burkholder, J. M. and P. M. Glibert. 2009. The importance of intraspecific variability in harmful algae- preface to a collection of topical papers. Harmful Algae 8:744-745. Burnham, K. P. and D. R. Anderson. 2001. Kullback -Leibler information as a basis for strong inference in ecological studies. Wildlife Research 28:111-119. Burns, C. W. 1968. The relationship between body size of filter-feeding Cladocera and the maximum size of particle i ngested. Limnology & Oceanography 13:675-678. Bykova, O., A. Laursen, V. Bostan, J. Bautista, and L. McCarthy. 2006. Do zebra mussels (Dreissena polymorpha) alter lake water chemistry in a way that favours Microcystis growth? Science of th e Total Environment 371:362-372. Calbet, A., M. Bertos, C. Fuentes-Grünewald, E. Alacid, R. Figueroa, B. Renom, and E. Garcés. 2011. Intraspecific variability in Karlodinium veneficum: Growth rates, mixotrophy, and lipid composition. Harmful Algae 10:654-667. 136 Canfield, D. E. J., E. Phlips, and C. M. Duarte. 1989. Factors influencing the abundance of blue-green algae in Florida lakes. Canadian Journal of Fisheries and Aquatic Sciences 46:1232-1237. Carey, C. C., H. Ewing, K. Cottingham, K. W eathers, R. Q. Thomas, and J. Haney. 2012a. Occurrence and toxicity of the cyanobacterium Gloeotrichia echinulata in low-nutrient lakes in the northeastern United States. Aquatic Ecology 46:395-409. Carey, C. C., B. W. Ibelings, E. P. Hoffma nn, D. P. Hamilton, and J. D. Brookes. 2012b. Eco- physiological adaptations that favour freshwater cyanobacteria in a changing climate. Water Research 46:1394-1407. Carey, C. C., K. C. Weathers, and K. L. Cottingham. 2008. Gloeotrichia echinulata blooms in an oligotrophic lake: helpful insights from eutrophic lakes. Journal of Plankton Research 30:893-904. Carmichael, W. W. 1994. The toxins of cyanobacteria. Scientific American 270:78-86. Carpenter, S. R., N. F. Caraco, D. L. Correll, R. W. Howarth, A. N. Sharpley, and V. H. Smith. 1998. Nonpoint pollution of surface waters with phosphorus and nitrogen. Ecological Applications 8:559-568. Carpenter, S. R., S. G. Fisher, N. B. Gri mm, and J. F. Kitchell. 1992. Global change and freshwater ecosystems. Annual Review of Ecology and Systematics 23:119-139. Cha, Y., C. A. Stow, and E. S. Bernhardt. 2013. Impacts of dreissenid mussel invasions on chlorophyll and total phosphorus in 25 lakes in the USA. Freshwater Biology 58:192-206. Chen, I.-C., J. K. Hill, R. Ohlemüller, D. B. Roy, and C. D. Thomas. 2011. Rapid range shifts of species associated with high levels of climate warming. Science 333:1024-1026. Chislock, M. F., O. Sarnelle, L. M. Jernigan, and A. E. Wilson. 2013. Do high concentrations of microcystin prevent Daphnia control of phytoplankton? Water Research 47:1961-1970. Chorus, I. and J. Bartram, editors. 1999. Toxic cyanobacteria in water: a guide to their public health consequences, monitoring a nd management. E & FN Spon, London, UK. Coma, R., M. Ribes, E. Serrano, E. Jiménez, J. Salat, and J. Pascual. 2009. Global warming- enhanced stratification and ma ss mortality events in the Me diterranean. Proceedings of the National Academy of Sciences 106:6176-6181. 137 Cotner, J. B., W. S. Gardner, J. R. Johnson, R. H. Sada, J. F. Cavaletto, and R. T. Heath. 1995. Effects of zebra mussels ( Dreissena polymorpha) on bacterioplankton: evidence for both size-selective consumption and growth stim ulation. Journal of Great Lakes Research 21:517-528. Cronberg, G. and H. Annadotter. 2006. Manual on aquatic cyanobacteria: a photo guide and a synopsis of their toxicology. International Society for the Study of Harmful Algae; United Nations Educational, Scientific, and Cultural Organization, Copenhagen. Cyr, H. and M. L. Pace. 1992. Grazing by zooplankton and its relationship to community structure. Canadian Journal of Fisheries and Aquatic Sciences 49:1453-1465. Davis, T. W., D. L. Berry, G. L. Boyer, and C. J. Gobler. 2009. The effects of temperature and nutrients on the growth and dynamics of toxic and non-toxic strains of Microcystis during cyanobacterial blooms. Harmful Algae 8:715-725. Demott, W. R. and E. N. Mckinney. 2015. Use it or lose it? Loss of grazing defenses during laboratory culture of the digestion-resistant green alga Oocystis . Journal of Plankton Research. Dillon, P. J. and F. H. Rigler. 1974. The phosphorus-chlorophyll relationship in lakes. Limnology and Oceanography 19:767-773. Dionisio Pires, L. M., B. M. Bontes, E. Van Donk, and B. W. Ibelings . 2005. Grazing on colonial and filamentous, toxic and non-toxic cyanobacteria by the zebra mussel Dreissena polymorpha. Journal of Plankton Research 27:331-339. Dionisio Pires, L. M., R. R. Jonker, E. Van Donk, and H. J. Laanbroek. 2004. Selective grazing by adults and larvae of the zebra mussel ( Dreissena polymorpha): application of flow cytometry to natural seston. Freshwater Biology 49:116-126. Dionisio Pires, L. M. and E. Van Donk. 2002. Comparing grazing by Dreissena polymorpha on phytoplankton in the presence of toxic and non-toxic cyanobacteria. Freshwater Biology 47:1855-1865. Dobiesz, N. E. and N. P. Lester. 2009. Changes in mid-summer water temperature and clarity across the Great Lakes between 1968 and 2002. Journal of Great Lakes Research 35:371-384. Downing, J. A., S. B. Watson, and E. McCaul ey. 2001. Predicting cyanobacteria dominance in lakes. Canadian Journal of Fisheries and Aquatic Sciences 58:1905-1908. 138 Downing, T. G., C. Meyer, M. M. Gehringer, and M. van de Venter. 2005. Microcystin content of Microcystis aeruginosa is modulated by nitrogen uptak e rate relative to specific growth rate or carbon fixation rate. Environmental Toxicology 20:257-262. Drake, J. M. and J. M. Bossenbroek. 2004. The potential distribution of zebra mussels in the United States. Bioscience 54:931-941. Dyble, J., G. L. Fahnenstiel, R. W. Litaker, D. F. Millie, and P. A. Tester. 2008. Microcystin concentrations and genetic diversity of Microcystis in Saginaw Bay and western Lake Erie. Environmental Toxicology 23:507-516. Edwards, M. and A. J. Richardson. 2004. Impact of climate change on marine pelagic phenology and trophic mismatch. Nature 430:881-884. Elderkin, C. L. and P. L. Klerks. 2005. Variation in thermal tolerance among three Mississippi River populations of the zebra mussel, Dreissena polymorpha. Journal of Shellfish Research 24:221-226. Elliott, J. A. 2012. Is the future blue-green? A re view of the current model predictions of how climate change could affect pelagic fr eshwater cyanobacteria. Water Research 46:1364-1371. Falconer, I. R. 1999. An overview of problems caused by toxic bl ueŒgreen algae (cyanobacteria) in drinking and recreational water. Environmental Toxicology 14:5-12. Fastner, J., U. Neumann, B. Wirsing, J. Wecke sser, C. Wiedner, B. Nixdorf, and I. Chorus. 1999. Microcystins (hepatotoxic heptapeptides) in German fres h water bodies. Environmental Toxicology 14:13-22. Ferber, L. R., S. N. Levine, A. Lini, and G. P. Livingston. 2004. Do cyanobacteria dominate in eutrophic lakes because they fix atmo spheric nitrogen? Freshwater Biology 49:690-708. Fujimoto, N., R. Sudo, N. Sugiura, and Y. Inamori. 1997. Nutrient-limited growth of Microcystis aeruginosa and Phormidium tenue and competition under various N:P supply ratios and temperatures. Limnology and Oceanography 42:250-256. Fulton, R. S. and H. W. Paerl. 1987a. Effects of colonial morphology on zooplankton utilization of algal resourecs during blue-green algal (Microcystis aeruginosa) blooms. Limnology and Oceanography 32:634-644. Fulton, R. S. and H. W. Paerl. 1987b. Toxic and inhibitory effects of the blue-green alga Microcystis aeruginosa on herbivorous zooplankton. Journal of Plankton Research 9:837-856. 139 Garrabou, J., R. Coma, N. Bensoussan, M. Bally, P. ChevaldonnÉ, M. Cigliano, D. Diaz, J. G. Harmelin, M. C. Gambi, D. K. Kersting, J. B. Ledoux, C. Lejeusne, C. Linares, C. Marschal, T. PÉRez, M. Ribes, J. C. Romano, E. Serrano, N. Teixido, O. Torrents, M. Zabala, F. Zuberer, and C. Cerrano. 2009. Mass mortality in Northwestern Mediterranean rocky benthic communities: effects of the 2003 heat wave. Global Change Biology 15:1090-1103. Garton, D. W., R. McMahon, and A. M. Stoeckmann. 2014. Limiting environmental factors and competitive interactions between zebra and quagga mussels in North America. Pages 383-402 in T. F. Nalepa and D. W. Schloesse r, editors. Quagga and zebra mussels: biology, impacts, and contro l. CRC Press, Boca Raton. Geller, W. and H. Müller. 1981. The filtration appara tus of Cladocera: filter mesh-sizes and their implications on food selectivity. Oecologia 49:316-321. Gergs, R. and K.-O. Rothhaupt. 2015. Invasive species as driving factors for the structure of benthic communities in Lake C onstance, Germany. Hydrobiologia 746:245-254. Giani, A., D. F. Bird, Y. T. Prairie, and J. F. Lawrence. 2005. Empirical study of cyanobacterial toxicity along a trophic gradient of lakes. Canadian Journal of Fisheries and Aquatic Sciences 62:2100-2109. Gilman, S. E., M. C. Urban, J. Tewksbury, G. W. Gilchrist, and R. D. Holt. 2010. A framework for community interactions under climate change. Trends in Ecology & Evolution 25:325-331. Graham, J. L. and J. R. Jones. 2009. Microcys tin in Missouri reservoirs. Lake and Reservoir Management 25:253-263. Griebeler, E. and A. Seitz. 2007. Effects of in creasing temperatures on population dynamics of the zebra mussel Dreissena polymorpha: implications from an individual-based model. Oecologia 151:530-543. Hallegraeff, G. M. 1993. A review of harmful algal blooms and their appa rent global increase. Phycologia 32:79-99. Hamilton, S. K., D. A. Bruesewitz, G. P. Hors t, D. B. Weed, and O. Sarnelle. 2009. Biogenic calciteŒphosphorus precipitation as a negative feedback to lake eutrophication. Canadian Journal of Fisheries and Aquatic Sciences 66:343-350. Hansen, B., P. K. Bjørnsen, and P. J. Hansen. 1994. The size ratio between planktonic predators and their prey. Limnology & Oceanography 39:395-403. 140 Hanson, J. M. and R. H. Peters. 1984. Empirical prediction of crustacean zooplankton biomass and profundal macrobenthos biomass in lakes. Canadian Journal of Fisheries and Aquatic Sciences 41:439-445. Harada, K., F. Kondo, and L. Lawton. 1999. Laboratory analysis of cyanotoxins. Pages 369-405 in I. Chorus and J. Bartram, editors. Toxic cy anobacteria in water: a guide to their public health consequences, monitoring a nd management. E & FN Spon, London, UK. Hasle, G. R. 1978. The inverted-microscope method. Pages 88-96 in A. Sournia, editor. Phytoplankton manual. UNESCO, Paris, France. Hecky, R. E., R. E. Smith, D. R. Barton, S. J. Gu ildford, W. D. Taylor, M. N. Charlton, and T. Howell. 2004. The nearshore phosphorus shunt: a consequence of ecosystem engineering by dreissenids in the Laurentian Great Lakes. Canadian Journal of Fisheries and Aquatic Sciences 61:1285-1293. Higgins, S. N. and M. J. Vander Zanden. 2010. What a difference a species makes: a meta- analysis of dreissenid mussel impacts on fr eshwater ecosystems. Ecological Monographs 91:179-196. Higgins, S. N., M. J. Vander Zanden, L. N. Joppa, and Y. Vadeboncoeur. 2011. The effect of dreissenid invasions on chlorophyll and the chlorophyll:total phosphorus ratio in north-temperate lakes. Canadian Journal of Fisheries and Aquatic Sciences 68:319-329. Hillebrand, H., C. D. Durselen, D. B. Kirsch tel, U. Pollingher, and T. Zohary. 1999. Biovolume calculation for pelagic and benthic microalgae. Journal of Phycology 35:403-424. Holm, N. P. and D. E. Armstrong. 1981. Role of nutrient limitation and competition in controlling the populations of Asterionella formosa and Microcystis aeruginosa in semicontinuous culture. Limnology and Oceanography 26:622-634. Horgan, M. J. and E. L. Mills. 1997. Clearance rates and filtering activity of zebra mussel (Dreissena polymorpha): implications for freshwater lake s. Canadian Journal of Fisheries and Aquatic Sciences 54:249-255. Horst, G. P., O. Sarnelle, J. D. White, S. K. Hamilton, R. B. Kaul, and J. D. Bressie. 2014. Nitrogen availability increases the toxin quota of a harmful cyanobacterium, Microcystis aeruginosa. Water Research 54:188-198. Huertas, I. E., M. Rouco, V. López-Rodas, and E. Costas. 2011. Warming will affect phytoplankton differently: evidence through a me chanistic approach. Proceedings of the Royal Society B: Biological Sciences 278:3534-3543. 141 Huisman, J., H. C. P. Matthijs, and P. M. Visser, editors. 2005. Harmful Cyanobacteria. Springer, Dordrecht. Idrisi, N., E. L. Mills, L. G. Rudstam, and D. J. Stewart. 2001. Impact of zebra mussels (Dreissena polymorpha) on the pelagic lower trophic levels of Oneida Lake, New York. Canadian Journal of Fisheries and Aquatic Sciences 58:1430-1441. Idso, S. B. 1973. On the concept of lake stability. Limnology and Oceanography 18:681-683. Imai, H., K.-H. Chang, M. Kusaba, and S.-i . Nakano. 2009. Temperature- dependent dominance of Microcystis (Cyanophyceae) species: M. aeruginosa and M. wesenbergii. Journal of Plankton Research 31:171-178. IPCC. 2014. Climate Change 2014: Synthesis Report. Contribution of Working Groups I, II and III to the Fifth Assessment Report of the In tergovernmental Panel on Climate Change. IPCC, Geneva, Switzerland. Iwanyzki, S. and R. W. McCauley. 1993. Upper le thal temperatures of adult zebra mussels (Dreissena polymorpha). Pages 667-673 in T. F. Nalepa and D. W. Schloesser, editors. Zebra mussels: biology, impacts, and cont rol. Lewis Publishers, Boca Raton. Jacobs, J. 1974. Quantitative measurement of food selection: a modification of the forage ratio and Ivlev™s electivit y index. Oecologia 14:413-417. Jankowski, T., D. M. Livingstone, H. R. Buhr er, R. Forster, and P. Niederhauser. 2006. Consequences of the 2003 European heat wave for lake temperature profiles, thermal stability, and hypolimnetic oxygen depletion: implications for a warmer world. Limnology & Oceanography 51:815-819. Jarvis, A. C., R. C. Hart, and S. Combrink. 1987. Zooplankton feeding on size fractionated Microcystis colonies and Chlorella in a hypereutrophic lake (Hartbeesport Dam, South Africa): implications to resource utili zation and zooplankton succession. Journal of Plankton Research 9:1231-1249. Jarvis, A. C., R. C. Hart, and S. Combrink. 1988. Cladoceran filtration rate-body length relations: model improvements developed for a Microcystis -dominated hypertrophic reservoir. Journal of Plankton Research 10:115-131. Johnson, M. T. J. and J. R. Stinchcombe. 2007. An emerging synthesis between community ecology and evolutionary biology. Trends in Ecology & Evolution 22:250-257. 142 Jungmann, D. and J. Benndorf. 1994. Toxicity to Daphnia of a compound extracted from laboratory and natural Microcystis spp ., and the role of microcystins. Freshwater Biology 32:13-20. Karatayev, A. Y., L. E. Burlakova, and D. K. Padilla. 1998. Physical factors that limit the distribution and abundance of Dreissena polymorpha (Pall.). Journal of Shellfish Research 17:1219-1235. Kardinaal, W. E. A., I. Janse, M. Kamst-van Agterveld, M. Meima, J. Snoek, L. R. Mur, J. Huisman, G. Zwart, and P. M. Visser. 2007a. Microcystis genotype succession in relation to microcystin concentrations in fresh water lakes. Aquatic Microbial Ecology 48:1 - 12. Kardinaal, W. E. A., L. Tonk, I. Janse, S. Hol, P. Slot, J. Huisman, and P. M. Visser. 2007b. Competition for light between toxic and nontoxi c strains of the harmful cyanobacterium Microcystis. Applied and Environmental Microbiology 73:2939-2946. Klausmeier, C. A., E. Litchman, T. Daufresn e, and S. A. Levin. 2004. Optimal nitrogen-to- phosphorus stoichiometry of phytoplankton. Nature 429:171-174. Knoll, L. B., O. Sarnelle, S. K. Hamilton, C. E. H. Scheele, A. E. Wilson, J. B. Rose, and M. R. Morgan. 2008. Invasive zebra mussels ( Dreissena polymorpha) increase cyanobacterial toxin concentrations in low-nutrient lakes. Canadian Journal of Fisheries and Aquatic Sciences 65:448-455. Kosten, S., V. L. M. Huszar, E. Bécares, L. S. Costa, E. van Donk, L.-A. Hansson, E. Jeppesen, C. Kruk, G. Lacerot, N. Mazzeo, L. De Meeste r, B. Moss, M. Lürling, T. Nõges, S. Romo, and M. Scheffer. 2012. Warmer clim ates boost cyanobacterial dominance in shallow lakes. Global Change Biology 18:118-126. Kotak, B. G. and R. W. Zurawell. 2007. Cyanob acterial toxins in Canadian freshwaters: A review. Lake and Reservoir Management 23:109-122. Kruk, C., V. L. M. Huszar, E. T. H. M. Peeters, S. Bonilla, L. Costa, M. Lürling, C. S. Reynolds, and M. Scheffer. 2010. A morphol ogical classification capturing functional variation in phytoplankton. Freshwater Biology 55:614-627. Kryger, J. and H. U. Riisgård. 1988. Filtration rate capacities in 6 species of European freshwater bivalves. Oecologia 77:34-38. Lakeman, M. B., P. von Dassow, and R. A. Ca ttolico. 2009. The strain co ncept in phytoplankton ecology. Harmful Algae 8:746-758. 143 Lei, J., B. S. Payne, and S. Y. Wang. 1996. Filtration dynamics of the zebra mussel, Dreissena polymorpha. Canadian Journal of Fisheries and Aquatic Sciences 1995:29-37. Litchman, E. and C. A. Klausmeier. 2008. Tr ait-based community ecology of phytoplankton. Annual Review of Ecology, Evolution, and Systematics 39:615-639. Liu, X., X. Lu, and Y. Chen. 2011. The effects of temperature and nutrient ratios on Microcystis blooms in Lake Taihu, China: An 11- year investigation. Harmful Algae 10:337-343. Long, J. D. and M. E. Hay. 2006. When intraspecific exceeds interspecific variance: effects of phytoplankton morphology and grow th phase on copepod feeding and fitness. Limnology & Oceanography 51:988-996. Lund, J. W. G., C. Kipling, and E. D. Cren. 1958. The inverted microscope method of estimating algal numbers and the sta tistical basis of estimati ons by counting. Hydrobiologia 11:143- 170. MacIsaac, H. J., C. J. Lonnee, and J. H. Leach. 1995. Suppression of microzooplankton by zebra mussels: importance of musse l size. Freshwater Biology 34:379-387. MacIsaac, H. J., W. G. Sprules, and J. H. Leach. 1991. Ingestion of small-bodied zooplankton by zebra mussels ( Dreissena polymorpha): can cannibalism on la rvae influence population dynamics? Canadian Journal of Fi sheries and Aquatic Sciences 48:2051-2060. Martins, J., M. Saker, C. Moreira, M. Welk er, J. Fastner, and V. Vasconcelos. 2009. Peptide diversity in strains of the cyanobacterium Microcystis aeruginosa isolated from Portuguese water supplies. Applied Microbiology and Biotechnology 82:951-961. McCauley, E. and F. Briand. 1979. Zooplankton grazing and phytoplankton species richness: field tests of the predation hypothesis. Limnology and Oceanography 24:243-252. McMahon, R. F., M. A. Matthews, T. A. Ussery, R. Chase, and M. Clarke. 1994. Further studies of heat tolerance in zebra mussels: Effect s of temperature acclimation and chronic exposure to lethal temperatures. Pages 251-272 in Proceedings: Fourth International Zebra Mussel Conference. University of Wisconsin Sea Grant Institute, Madison, WI. McMahon, R. F., M. A. Matthews, T. A. Ussery, R. Chase, and M. Clarke. 1995. Studies of heat tolerance of zebra mussels: effects of temp erature acclimation and chronic exposure to lethal temperatures. United States Army Corps of Engineers Technical Report EL-95-9., Washington, D.C. 144 McMahon, R. F. and T. A. Ussery. 1995. Thermal tolerance of zebra mussels ( Dreissena polymorpha) relative to rate of temperature increa se and acclimation temperature. United States Army Corps of Engineers Techni cal Report EL-95-10., Washington, D.C. McQueen, D. J. and D. R. S. Lean. 1987. Influence of water temperature and nitrogen to phosphorus ratios on the domina nce of blue-green algae in Lake St. George, Ontario. Canadian Journal of Fisheries and Aquatic Sciences 44:598-604. Menzel, D. W. and N. Corwin. 1965. The measurem ent of total phosphorus in seawater based on the liberation of organically bound fractions by persulfate oxidation. Limnology and Oceanography 10:280-282. Mieleitner, J. and P. Reichert. 2008. Modelling functional groups of phytoplankton in three lakes of different trophic state. Ecological Modelling 211:279-291. Mikalsen, B., G. Boison, O. M. Skulberg, J. Fastner, W. Davies, T. M. Gabrielsen, K. Rudi, and K. S. Jakobsen. 2003. Natural variation in the microcystin synthetase operon mcyABC and impact on microcystin production in Microcystis strains. Journal of Bacteriology 185:2774-2785. Moss, B. 1972a. Studies on Gull Lake, Michigan. I. Seasonal and depth distribution of phytoplankton. Freshwater Biology 2:289-307. Moss, B. 1972b. Studies on Gull Lake, Michigan. II. Eutrophication- evidence and prognosis. Freshwater Biology 2:309-320. Moss, B., R. Wetzel, and G. H. Lauff. 1980. Annual productivity and phytoplankton changes between 1969 and 1974 in Gull Lake, Michigan. Freshwater Biology 10:113-121. Murphy, J. and L. P. Riley. 1962. A modified single solution method for the determination of phosphate in natural waters. Analytica Chimica Acta 27:31-36. Naddafi, R., K. Pettersson, and P. Eklöv. 2007. The effect of seasonal variation in selective feeding by zebra mussels ( Dreissena polymorpha) on phytoplankton community composition. Freshwater Biology 52:823-842. Nalewajko, C. and T. P. Murphy. 2001. Effects of temperature, and availability of nitrogen and phosphorus on the abundance of Anabaena and Microcystis in Lake Biwa, Japan: an experimental approach. Limnology 2:45-48. Neuhauser, C., D. A. Andow, G. E. Heimpel, G. May, R. G. Shaw, and S. Wagenius. 2003. Community genetics: expanding the synthesis of ecology and genetics. Ecology 84:545- 558. 145 O'Brien, K. R., D. L. Meyer, A. M. Waite, G. N. Ivey, and D. P. Hamilton. 2004. Disaggregation of Microcystis aeruginosa colonies under turbulent mixi ng: laboratory experiments in a grid-stirred tank. Hydrobiologia 519:143-152. O'Neil, J. M., T. W. Davis, M. A. Burford, and C. J. Gobler. 2012. The rise of harmful cyanobacteria blooms: The potential roles of eutrophication and climate change. Harmful Algae 14:313-334. Omori, M. and T. Ikeda. 1984. Methods in ma rine zooplankton ecology. Wiley-Interscience, New York. Otsuka, S., S. Suda, R. Li, S. Matsumoto, and M. M. Watanabe. 2000. Morphological variability of colonies of Microcystis morphospecies in culture. The Journal of General and Applied Microbiology 46:39-50. Otsuka, S., S. Suda, S. Shibata, H. Oyaizu, S. Matsumoto, and M. M. Watanabe. 2001. A proposal for the unification of five species of the cyanobacterial genus Microcystis Kützing ex Lemmermann 1907 under the rules of the Bacteriological Code. International Journal of Systematic and Evolutionary Microbiology 51:873-879. Paerl, H. W. and J. Huisman. 2008. Blooms like it hot. Science 320:57-58. Paerl, H. W. and J. Huisman. 2009. Climate change : a catalyst for global expansion of harmful cyanobacterial blooms. Environmental Microbiology Reports 1:27-37. Paerl, H. W. and V. J. Paul. 2012. Climate change: Links to global expansion of harmful cyanobacteria. Water Research 46:1349-1363. Paul, V. J. 2008. Chapter 11: Global warming an d cyanobacterial harmfu l algal blooms. Pages 239-257 in H. K. Hudnell, editor. Cyanobacteria l harmful algal blooms: State of the science and research needs. Perry, A. L., P. J. Low, J. R. Ellis, and J. D. Reynolds. 2005. Climate change and distribution shifts in marine fishes. Science 308:1912-1915. Pilotto, L. S., R. M. Douglas, M. D. Burch, S. Cameron, M. Beers, G. J. Rouch, P. Robinson, M. Kirk, C. T. Cowie, S. Hardiman, C. Moore, and R. G. Attewell. 1997. Health effects of exposure to cyanobacteria (blueŒgreen algae) during recreational waterŒrelated activities. Australian and New Zealand Journal of Public Health 21:562-566. Pitcher, G. C. 2012. Harmful algaeŠThe requireme nt for species-specific information. Harmful Algae 14:1-4. 146 Post, E. and C. Pedersen. 2008. Opposing plant community responses to warming with and without herbivores. Proceedings of the National Academy of Sciences 105:12353-12358. Pounds, A. J., M. R. Bustamante, L. A. Coloma, J. A. Consuegra, M. P. L. Fogden, P. N. Foster, E. La Marca, K. L. Masters, A. Merino-Viteri, R. Puschendorf, S. R. Ron, G. A. Sanchez- Azofeifa, C. J. Still, and B. E. Young. 2006. Widespread amphibian extinctions from epidemic disease driven by global warming. Nature 439:161-167. Rahel, F. J. and J. D. Olden. 2008. Assessing the effects of climate change on aquatic invasive species. Conservation Biology 22:521-533. Raikow, D. E., O. Sarnelle, A. E. Wilson, a nd S. K. Hamilton. 2004. Dominance of the noxious cyanobacterium Microcystis aeruginosa in low nutrient lakes is associated with exotic zebra mussels. Limnology and Oceanography 49:482-487. Ramcharan, C. W., D. K. Padilla, and S. I. Dodson. 1992. A multivariate model for predicting population fluctuations of Dreissena polymorpha in North American lakes. Canadian Journal of Fisheries and Aquatic Sciences 49:150-158. Read, J. S., D. P. Hamilton, I. D. Jones, K. Mu raoka, L. A. Winslow, R. Kroiss, C. H. Wu, and E. Gaiser. 2011. Derivation of lake mixing a nd stratification indices from high-resolution lake buoy data. Environmental Modelling & Software 26:1325-1336. Reichwaldt, E. S. and A. Ghadouani. 2012. Effects of rainfall patterns on toxic cyanobacterial blooms in a changing climate: Between si mplistic scenarios and complex dynamics. Water Research 46:1372-1393. Reynolds, C. 2006. Ecology of phytoplankton. Cambridge University Press, Cambridge. Reynolds, C. S. 1973. Growth and buoyancy of Microcystis aeruginosa Kutz. emend. Elenkin in a shallow eutrophic lake. Proceedings of the Royal Society of London Series B- Biological Sciences 184:29-50. Reynolds, C. S. 2007. On the interannual variability in phytoplankton production in freshwaters. Pages 187-221 in P. J. l. B. Williams, D. N. Thomas, and C. S. Reynolds, editors. Phytoplankton Productivity: Carbon Assimilation in Marine and Freshwater Ecosystems. Blackwell Science Ltd, Oxford. Reynolds, C. S., G. H. M. Jaworski, A. Cmiech, and G. F. Leedale. 1981. On the annual cycle of the blue-green alga Microcystis aeruginosa Kütz. Emend. Elenkin. Philosophical Transactions of the Royal Society of London Series B 293:419-477. 147 Reynolds, C. S. and D. A. Rogers. 1976. Seasonal variations in the vertical distribution and buoyancy of Microcystis aeruginosa Kütz. Emend. Elenken in Rostherne Mere, England. Hydrobiologia 48:17-23. Ricciardi, A. and H. J. MacIsaac. 2000. Recen t mass invasion of the North American Great Lakes by PontoŒCaspian species. Tr ends in Ecology & Evolution 15:62-65. Rico, M., M. Altamirano, V. López-Rodas, and E. Costas. 2006. Analysis of polygenic traits of Microcystis aeruginosa (Cyanobacteria) strains by Restricted Maximum Likelihood (REML) procedures: 1. Size and shape of colonies and cells. Phycologia 45:237-242. Rigosi, A., C. C. Carey, B. W. Ibelings, a nd J. D. Brookes. 2014. The interaction between climate warming and eutrophication to promot e cyanobacteria is dependent on trophic state and varies among taxa. Limnology and Oceanography 59:99-114. Rinta-Kanto, J. M., E. A. Konopko, J. M. DeBruyn, R. A. Bourbonniere, G. L. Boyer, and S. W. Wilhelm. 2009. Lake Erie Microcystis : relationships between microcystin production, dynamics of genotypes and environmental para meters in a large lake. Harmful Algae 8:665-673. Robarts, R. D. and T. Zohary. 1987. Temperatur e effects on photosynthetic capacity, respiration, and growth rates of bloom-forming cyanobact eria. New Zealand Journal of Marine and Freshwater Research 21:391-399. Roditi, H. A., N. F. Caraco, J. J. Cole, and D. L. Strayer. 1996. Filtrat ion of Hudson River water by the zebra mussel ( Dreissena polymorpha). Estuaries 19:824-832. Rojo, C. 1998. Differential attributes of phytoplankton across the trophic gradient: a conceptual landscape with gaps. Hydrobiologia 369-370:1-9. Romero-Lankao, P., J. B. Smith, D. J. Davidson, N. S. Diffenbaugh, P. L. Kinney, P. Kirshen, P. Kovacs, and L. Villers Ruiz. 2014. Climate change 2014: Impacts, adaptation, and vulnerability. Part B: regional aspects. . Pages 1439-1498 in V. R. Barros, C. B. Field, D. J. Dokken, M. D. Mastrandrea, K. J. Mach, T. E. Bilir, M. Chatterjee, K. L. Ebi, Y. O. Estrada, R. C. Genova, B. Girma, E. S. Kissel, A. N. Levy, S. MacCracken, P. R. Mastrandrea, and W. L.L., editors. Contri bution of working group II to the fifth assessment report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom. Rösner, R. R., D. C. Müller-Navarra, and E. Zori ta. 2012. Trend analysis of weekly temperatures and oxygen concentrations during summer stratification in Lake Plußsee: A long-term study. Limnology and Oceanography 57:1479-1491. 148 Saker, M. L., J. Fastner, E. Dittmann, G. Ch ristiansen, and V. M. Vasconcelos. 2005. Variation between strains of the cyanobacterium Microcystis aeruginosa isolated from a Portuguese river. Journal of Applied Microbiology 99:749-757. Sandgren, C. D. and J. V. Robinson. 1984. A stratified sampling approach to compensating for non-random sedimentation of phytoplankton ce lls in inverted microscope settling chambers. British Phycological Journal 19:67-72. Sarnelle, O., J. Morrison, R. Kaul, G. Ho rst, H. Wandell, and R. Bednarz. 2010. Citizen monitoring: testing hypotheses about the interactive influences of eutrophication and mussel invasion on a cyanobacterial toxin in lakes. Water Research 44:141-150. Sarnelle, O., J. D. White, G. P. Horst, and S. K. Hamilton. 2012. Phosphorus addition reverses the positive effect of zebra mussels ( Dreissena polymorpha) on the toxic cyanobacterium, Microcystis aeruginosa. Water Research 46:3471-3478. Sarnelle, O., A. E. Wilson, S. K. Hamilton, L. B. Knoll, and D. E. Raikow. 2005. Complex interactions between the zebra mussel, Dreissena polymorpha, and the harmful phytoplankter, Microcystis aeruginosa. Limnology and Oceanography 50:896-904. Schindler, D. W. 1974. Eutrophication and recovery in experimental lakes: Implications for lake management. Science 184:897-899. Schindler, D. W. 1977. The evolution of phosphorus limitation in lakes. Science 195:260-262. Schindler, D. W., K. G. Beaty, E. J. Fee, D. R. Cruikshank, E. R. Debruyn, D. L. Findlay, G. A. Linsey, J. A. Shearer, M. P. Stainton, and M. A. Turner. 1990. Effects of climatic warming on lakes of the centr al boreal forest. Science 250:967-970. Schneider, P. and S. J. Hook. 2010. Space observations of inland water bodies show rapid surface warming since 1985. Geophysical Research Letters 37:L22405. Schweiger, O., J. Settele, O. Kudrna, S. Klotz, and I. Kühn. 2008. Climate change can cause spatial mismatch of trophical ly interacting species. Ecology 89:3472-3479. Seip, K. L. and C. S. Reynolds. 1995. Phytoplankton functional attributes along trophic gradient and season. Limnology and Oceanography 40:589-597. Shapiro, J. 1997. The role of carbon dioxide in the initiation and maintenance of blue-green dominance in lakes. Freshwater Biology 37:307-323. 149 Sinclair, G., D. Kamykowski, and P. M. Glib ert. 2009. Growth, uptake, and assimilation of ammonium, nitrate, and ur ea, by three strains of Karenia brevis grown under low light. Harmful Algae 8:770-780. Smale, D. A. and T. Wernberg. 2013. Extreme cl imatic event drives ra nge contraction of a habitat-forming species. Pr oceedings of the Royal Society of London B: Biological Sciences 280. Smith, T. E., R. J. Stevenson, N. F. Carac o, and J. J. Cole. 1998. Changes in phytoplankton community structure during the zebra mussel ( Dreissena polymorpha) invasion of the Hudson River (New York). Journal of Plankton Research 20:1567-1579. Smith, V. H. 1983. Low nitrogen to phosphorus favors dominance by bluegreen algae in lake phytoplankton. Science 221:669-671. Smith, V. H. 1985. Predictive models for the biom ass of blue-green algae in lakes. Water Resources Bulletin 21:433-439. Smith, V. H. 1986. Light and nutrient effects on the relative biomass of blue-green algae in lake phytoplankton. Canadian Journal of Fisheries and Aquatic Sciences 43:148-153. Smith, V. H., E. Willen, and B. Karlsson. 1987. Pr edicting the summer peak biomass of four species of blue-green algae (Cyanophyta/Cyanobacteria) in Swedish lakes. Water Resources Bulletin 23:397-402. Solórzano, L. 1969. Determination of ammonia in natural water by the phenolhypochlorite method. Limnology and Oceanography 14:799-801. Spidle, A. P., B. May, and E. L. Mills. 1995. Limits to tolerance of temperature and salinity in the quagga mussel (Dreissena bugensis) and the zebra mussel ( Dreissena polymorpha). Canadian Journal of Fisheries and Aquatic Sciences 52:2108-2119. Spijkerman, E. and P. M. Coesel. 1998. Ecophysio logical characteristics of two planktonic desmid species originating from trophi cally different lakes. Hydrobiologia 369-370:109- 116. Sprules, W. G., J. M. Casselman, and B. J. Shut er. 1983. Size distribution of pelagic particles in lakes. Canadian Journal of Fisheries and Aquatic Sciences 40:1761-1769. Sprung, M. and U. Rose. 1988. Influence of food size and food quantity on the feeding of the mussel Dreissena polymorpha . Oecologia 77:526-532. 150 Ståhl-Delbanco, A., L.-A. Hansson, and M. Gyllström. 2003. Recruitment of resting stages may induce blooms of Microcystis at low N:P ratios. Jour nal of Plankton Research 25:1099-1106. Sterner, R. W. 1989. The role of grazer s in phytoplankton succession. Pages 107-170 in U. Sommer, editor. Plankton ecology: succession in plankton communities. Springer-Verlag, Berlin, Germany. Stoeckmann, A. M. and D. W. Garton. 2001. Flex ible energy allocation in zebra mussels (Dreissena polymorpha) in response to different environmental conditions. Journal of the North American Benthological Society 20:486-500. Strayer, D. 2014. Understanding how nutrient cycl es and freshwater mussels (Unionoida) affect one another. Hydrobiologia 735:277-292. Strayer, D., N. Cid, and H. Malcom. 2011. Long-te rm changes in a population of an invasive bivalve and its effects. Oecologia 165:1063-1072. Strayer, D. L. 1991. Projected distribution of the zebra mussel, Dreissena polymorpha, in North America. Canadian Journal of Fisheries and Aquatic Sciences 48:1389-1395. Strayer, D. L. 2008. Twenty years of zebra mu ssels: lessons from the mollusk that made headlines. Frontiers in Ecology and the Environment 7:135-141. Suttle, K. B., M. A. Thomsen, and M. E. Powe r. 2007. Species interactions reverse grassland responses to changing climate. Science 315:640-642. Tanabe, Y., F. Kasai, and M. M. Watanabe. 2007. Multilocus sequence typing (MLST) reveals high genetic diversity and clonal population structure of the toxic cyanobacterium Microcystis aeruginosa. Microbiology-Sgm 153:3695-3703. Tanabe, Y., F. Kasai, and M. M. Watanabe. 2009. Fine-scale spatial and temporal genetic differentiation of water bloom-forming cyanobacterium Microcystis aeruginosa: revealed by multilocus sequence typing. Environmental Microbiology Reports 1:575-582. Ten Winkel, E. H. and C. Davids. 1982. Food selection by Dreissena polymorpha Pallas (Mollusca: Bivalvia). Freshwater Biology 12:553-558. Tessier, A. J. and G. H. Lauff. 1992. The Gull Lake story. The Michigan Riparian February:12-14. 151 Thorp, J. H., J. E. Alexander, B. L. Bukaveckas, G. A. Cobbs, and K. L. Bresko. 1998. Responses of Ohio River and Lake Erie dreissenid molluscs to changes in temperature and turbidity. Canadian Journal of Fisheries and Aquatic Sciences 55:220-229. Tillmanns, A. R. and F. R. Pick. 2011. The effect of sampling scales on the interpretation of environmental drivers of the cyanotoxin mi crocystin. Lake and Reservoir Management 27:183-193. Trimbee, A. M. and E. E. Prepas. 1987. Evaluation of total phosphorus as a predictor of relative biomass of blue-green algae with an empha sis on Alberta lakes. Canadian Journal of Fisheries and Aquatic Sciences 44:1337-1342. Tylianakis, J. M., R. K. Didham, J. Bascom pte, and D. A. Wardle. 2008. Global change and species interactions in terrestrial ecosystems. Ecology Letters 11:1351-1363. Utkilen, H., O. M. Skulberg, B. Underdal, N. Gjølme, R. Skulberg, and J. Kotai. 1996. The rise and fall of a toxigenic population of Microcystis aeruginosa (Cyanophyceae/Cyanobacteria)Ša decade of observa tions in Lake Akersvatnet, Norway. Phycologia 35:189-197. Van de Waal, D. B., J. M. H. Verspagen, M. Lürling, E. Van Donk, P. M. Visser, and J. Huisman. 2009. The ecological stoichiome try of toxins produced by harmful cyanobacteria: an experimental test of the carbon-nutrient balance hypothesis. Ecology Letters 12:1326-1335. Van der Putten, W. H., M. Macel, and M. E. Visser. 2010. Predicting species distribution and abundance responses to climate change: why it is essential to include biotic interactions across trophic levels. Philosophical Transactions of the Royal Society B: Biological Sciences 365:2025-2034. Van Donk, E. 1997. Defenses in phytoplankton against grazing induced by nutrient limitation, UV-B stress and infochemicals. Aquatic Ecology 31:53-58. Van Donk, E., M. Lürling, D. O. Hessen, and G. M. Lokhorst. 1997. Altered cell wall morphology in nutrient-deficient phytoplankton and its impact on grazers. Limnology and Oceanography 42:357-364. van Gremberghe, I., P. Vanormelingen, K. Van de r Gucht, C. Souffreau, W. Vyverman, and L. De Meester. 2009. Priority effects in experimental populations of the cyanobacterium Microcystis. Environmental Microbiology 11:2564-2573. Vanderploeg, H. A., J. R. Liebig, W. W. Carmichael, M. A. Agy, T. H. Johengen, G. L. Fahnenstiel, and T. F. Nalepa. 2001. Zebra mussel ( Dreissena polymorpha) selective 152 filtration promoted toxic Microcystis blooms in Saginaw Bay (Lake Huron) and Lake Erie. Canadian Journal of Fisheries and Aquatic Sciences 58:1208-1221. Vezie, C., L. Brient, K. Sivonen, G. Bertru, J. C. Lefeuvre, and M. Salkinoja-Salonen. 1998. Variation of microcystin content of cyanobact erial blooms and isolated strains in Lake Grand-Lieu (France). Microbial ecology 35:126-135. Violle, C., B. J. Enquist, B. J. McGill, L. Jiang, C. H. Albert, C. Hulshof, V. Jung, and J. Messier. 2012. The return of the variance: intraspecific variability in community ecology. Trends in Ecology & Evolution 27:244-252. Visser, P., B. Ibelings, L. Mur, and A. Walsby. 2005. The ecophysiology of the harmful cyanobacterium Microcystis : Features explaining its success and measures for its control. Pages 109-142 in J. Huisman, H. Matthijs, and P. Vi sser, editors. Harmful cyanobacteria. Springer Dordrecht. Visser, P. M., J. Passarge, and L. R. Mur. 1997. Modelling vertical migration of the cyanobacterium Microcystis . Hydrobiologia 349:99-109. Ward, J. M. and A. Ricciardi. 2007. Impacts of Dreissena invasions on benthic macroinvertebrate communities: a meta-a nalysis. Diversity and Distributions 13:155-165. Watson, S. B., E. McCauley, and J. A. Downing. 1997. Patterns in phytoplankton taxonomic composition across temperate lakes of di ffering nutrient status. Limnology and Oceanography 42:487-495. Wehr, J. D. and R. G. Sheath, editors. 2003. Freshwater algae of North America: Ecology and classification. Academic Press, San Diego. Welker, M., L. −ejnohová, D. Némethová, D. Hans von, J. Jarkovský, and B. Marıálek. 2007. Seasonal shifts in chemotype composition of Microcystis sp. communities in the pelagial and the sediment of a shallow re servoir. Limnology and Oceanography 52:609-619. Welschmeyer, N. A. 1994. Fluorometric analysis of chlorophyll a in the presence of chlorophyll b and phaeopigments. Limnology and Oceanography 39:1985-1992. Wetzel, R. G. 2001. Limnology: lake and river ecosystems. Academic Press. White, J. D., S. K. Hamilton, and O. Sarnelle. 2015. Heat-induced mass mo rtality of invasive zebra mussels ( Dreissena polymorpha) at sublethal water temperatures. Canadian Journal of Fisheries and Aquatic Sciences. 153 White, J. D., R. B. Kaul, L. B. Knoll, A. E. Wilson, and O. Sarnelle. 2011. Large variation in vulnerability to grazing within a population of the colonial phytoplankter, Microcystis aeruginosa. Limnology and Oceanography 56:1714-1724. White, J. D. and O. Sarnelle. 2014. Size-structured vulnerability of the colonial cyanobacterium, Microcystis aeruginosa, to grazing by zebra mussels ( Dreissena polymorpha). Freshwater Biology 59:514-525. Whitham, T. G., W. P. Young, G. D. Martinsen, C. A. Gehring, J. A. Schweitzer, S. M. Shuster, G. M. Wimp, D. G. Fischer, J. K. Bailey, R. L. Lindroth, S. Woolbright, and C. R. Kuske. 2003. Community and ecosystem genetics: a co nsequence of the extended phenotype. Ecology 84:559-573. Wilson, A. E., R. B. Kaul, and O. Sarnelle. 2010. Growth rate consequences of coloniality in a harmful phytoplankter. PLoS ONE 5:e8679. Wilson, A. E. and O. Sarnelle. 2002. Relations hip between zebra mussel biomass and total phosphorus in European and North Ameri can lakes. Archiv fur Hydrobiologie 153:339- 351. Wilson, A. E., O. Sarnelle, B. A. Neilan, T. P. Salmon, M. M. Gehringer, and M. E. Hay. 2005. Genetic variation of the bloom-forming cyanobacterium, Microcystis aeruginosa, within and among lakes: implications for harmful algal blooms. Applied and Environmental Microbiology 71:6126-6133. Wilson, A. E., W. A. Wilson, and M. E. Hay. 2006. Intraspecific variation in growth and morphology of the bloom-forming cyanobacterium Microcystis aeruginosa. Applied and Environmental Microbiology 72:7386-7389. Winder, M. and D. E. Schindler. 2004. Climate change uncouples trophic interactions in an aquatic system. Ecology 85:2100-2106. Yang, L. H. and V. H. W. Rudolf. 2010. Phenology, ontogeny and the effects of climate change on the timing of species in teractions. Ecology Letters 13:1-10. Zehnder, A. and P. R. Gorham. 1960. Factors influencing the growth of Microcystis aeruginosa Kütz. Emend. Elenkin. Canadi an Journal of Microbiology 6:645-660. Zhang, M., F. Kong, X. Tan, Z. Yang, H. Cao, and P. Xing. 2007. Biochemical, morphological, and genetic variations in Microcystis aeruginosa due to colony disaggregation. World Journal of Microbiology and Biotechnology 23:663-670.