EVOLUTION AND ECOLOGICAL SPECIALIZATION OF
A SHEWANELLA BALTICA POPULATION

By
Jie Deng

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILSOPHY
Biochemistry and Molecular Biology
2011

ABSTRACT
EVOLUTION AND ECOLOGICAL SPECIALIZATION OF A SHEWANELLA BALTICA POPULATION
By
Jie Deng
Studying how bacterial strains diverge over time and how divergence leads to
specialization to new environmental niches is important for understanding the dynamics of
environmental communities. I studied these questions using a collection of 46 Shewanella
baltica strains isolated over a period of 12 years from the Gotland Deep in the central Baltic Sea,
which is characterized by the presence of a stable redox gradient over 140 m in depth. I used
several experimental and computational methods to explore the genetic and phenotypic
profiles of these strains and determined whether these profiles could be explained by the
conditions at their sites of isolation. Specifically, genotyping by both single gene-based and
multi-locus sequence typing (MLST) indicated specialization across the redox gradient and over
time. Versatility in anaerobic respiration, for example, the ability to respire thiosulfate, was
correlated with the redox gradient, while versatility in carbon source utilization was correlated
with specialization over time. Comparative Genomic Hybridization (CGH) revealed a
heterogeneous distribution of genes across S. baltica genomes, where genes potentially
important for specialization both over the redox gradient and over time were identified.
Further mechanistic investigation of their evolution provided evidence of horizontal gene
transfer in the S. baltica genomes, whereas predicted recent recombination among strains from
two lineages was also supported by this larger dataset. The differential fitness of S. baltica
strains under particular redox conditions was determined from competition assays, where

competed strains were labeled with gfp and cfp and the population dynamics during
competition was measured by flow cytometry. The two competed genotypes were separated
by flow cytometry and their transcriptomes analyzed by RNA-Seq which showed differential
expression in response to the nitrate condition as well as to the competitor strain. By
integrating these different types of information, I determined that specialization has occurred
in this S. baltica population both over the redox gradient and over time, and that the genomic
plasticity as well as the extensive gene exchange has facilitated the specialization process.
Overall, I used S. baltica as a model to provide insights into fundamental questions about the
interactions among bacterial genomics, physiology, and habitat (ecology).

ACKNOWLEDGEMENTS
This thesis would not have been possible without the kind support and help from many
individuals who, in one way or another, contributed to the preparation and completion of this
study.
First and foremost, I owe my sincerest gratitude to my supervisor, Dr. James Tiedje, who
supported and mentored me throughout my thesis with his knowledge and patience, while at
the same time allowed me the room to work in my own way. His encouraging and optimistic
attitude has helped me through the tough and frustrating stages of my work, and has greatly
influenced my scientific and personal growth.
I am deeply grateful to a number of colleagues involved with this project. Ingrid Brettar
and Manfred Höfle supplied this precious collection of Shewanella baltica isolates. Kostas
Konstantinidis constantly provided valuable insights and continues to contribute to this project
with his specialty in bioinformatics. I also share the credit of my work with Jennifer Auchtung,
who guided me for nearly two years and helped me grow to be an independent thinker and
researcher.
In my daily work, I have been blessed with a friendly and cheerful group of lab mates,
whose critical discussions, support, and encouragement throughout the years have helped
propel me forward.
I would also thank the Department of Biochemistry and Molecular Biology, for offering
me the opportunity of studying in this university, as well as my committee, for providing
valuable advice during my Ph.D.

iv

Lastly, I am indebted to my dearest parents for their love and constant support, and for
their understanding of me pursuing advanced degrees abroad. Without the combined input and
efforts from all of these friends and colleagues, I am sure this journey would not have been this
fantastic - thank you all.

v

TABLE OF CONTENTS

LIST OF TABLES …………………………………………………………………………………………………………………….x
LIST OF FIGURES ………………………………………………………………………………………………………………….xi
LIST OF ABBREVIATIONS …………………………………………………………………………………………………….xiv
CHAPTER I
INTRODUCTION
A perspective on bacterial specialization ……………………………………………………………………………2
The versatile Shewanella ……………………………………………………………………………………………………3
The history behind the Shewanella baltica population in the Baltic Sea …………………………….5
Thesis Outline …………………………………………………………………………………………………………………….7
References………………………………………………………………………………………………………………………..10
CHAPTER II
PHYLOGENETIC ANALYSIS OF A SHEWANELLA BALTICA POPULATION USING MULTI LOCUS
SEQUENCE TYPING
Abstract …………………………………………….…………………………………………………………………………15
Introduction …………………………………………………………………………………………………………………16
Materials and Methods ………………………………………………………………………………………………..20
Bacterial strains and growth conditions ………………………………………………………………….…20
MLST Analysis ……………………………………………………………………………………………………………23
Results …………………………………………………………………………………………………………………..…….25
gyrB analysis ……………………………………………………………………………………………………….....…25
MLST analysis ………………………………………………………………………………………………………...…27
Recombination analysis on genes selected for genotyping …………………………………...….31
Discussion……………………………………………………………………………………………………………………..32
Genotyping indicates niche specialization and genetic continuity over time……………..32
Recombination was detected from sequences of the marker genes ……………………..….33
References ……………………………………………………………………………………………………………………35
CHAPTER III
COMPARATIVE GENOMIC STUDY OF THE SHEWANELLA BALTICA GENOMES USING
MICROARRAYS
Abstract …………………………………………………………………………………………………………………….…40
Introduction …………………………………………………………………………………………………………………41
vi

Materials and Methods…………………………………………………………………………………………………44
Array Design ……………………………………………………………………………………………………..………44
DNA microarray hybridization …………………………………………………………………………..………45
Gene expression studies ……………………………………………………………………………………………46
Correlation analysis on gyrB, MLST and CGH based distance matrices ………………..….…47
Results and Discussion ………………………………………………………………………………………………….49
Evaluation of probe binding efficiency and specificity ……………………………………………….49
Array normalization and setting cutoff value for determining gene presence
/absence …………………………………………………………………………………………………………………50
General statistics of CGH …………………………………………………………………………………………..51
GC bias in the core and variable genome ………………………………………………………………….55
Less conservation of intergenic regions ………………………………………………………………….…57
Functional bias in the core genome and variable genome …………………………………………59
Similar clustering patterns with MLST phylogeny ………………………………………………….…..63
Genes related to specialization over time ………………………………………………………………….65
Genes related to specialization across the redox gradient …………………………………………69
Conserved gene core in specific lineages ……………………………………………………………..……71
Genes specific to strains in Clade A and D ………………………………………………...……..72
Genes specific to Clade E ………………………………………………………………………….………74
Genes specific to Clade A ………………………………………………………………………….……..74
Genes specific to Clade D ………………………………………………………………………….……..75
Core genes in Clade B strains ……………………………………………………………………………75
High versatility in carbon source utilization was reflected from the OS223
genome …………………………………………………………………………………………………………76
Horizontal gene transfer and genomic islands ……………………………………………………..……77
Recombination among specific lineages was supported ……………………………………….…..79
References ……………………………………………………………………………………………………………………83
CHAPTER IV
CHARACTERIZATION OF CARBON SOURCE UTILIZATION AND RESPIRATORY CAPABILITIES OF
THE SHEWANELLA BALTICA STRAINS
Abstract …………………………………………………………………………………………………………………….…88
Introduction …………………………………………………………………………………………………………………89
Materials and Methods ………………………………………………………………………………………………..92
Bacterial strains and growth conditions …………………………………………………………………….92
Regression analysis on Biolog datasets ………………………………………………………………..……92
Results …………………………………………………………………………………………………………………………94
Examining diversity of carbon source utilization using the Biolog system ……………..….94
vii

Examining diversity in utilization of different electron acceptors …………………………..…97
Discussion …………………………………………………………………………………………………………………….99
Metabolic similarity in carbon source utilization is correlated with evolutionary
relatedness …………………………………………………………………………………………………………….99
Temporal distribution has a significant impact on evolution of metabolic capabilities by
S. baltica strains ………………………………………………………………………….………………………..102
Respiratory versatility within the S. baltica species is correlated with their genotypic
relatedness ………………………………………………………………………………………..…………………103
Reduction of nitrate likely via ammonification is a common phenotypic characteristic
among all S. baltica strains ……………………………………………………………………………………104
DMSO reduction may be related with specialization to anoxic deep water
environment ………………………………………………………………………………………………………...104
Inability to reduce thiosulfate is limited within Clade E ………………………………………..…105
Inability to reduce TMAO is also limited within Clade E ………………………………………..…106
References …………………………………………………………………………………………………………………107
CHAPTER V
COMPETITION ASSAYS AMONG SELECTED SHEWANELLA BALTICA STRAINS WHILE EXPOSED
TO DIFFERENT ELECTRON ACCEPTORS
Abstract …………………………………………………………………………………………………………………….111
Introduction ………………………………………………………………………………………………………………112
Materials and Methods …………………………………………………………………………………………..…115
Bacterial strains and growth conditions ………………………………………………………………….115
GFP and CFP tagging of Shewanella baltica strains ……………………………………………….…115
Competition assays …………………………………………………………………………………………………116
Experimental design for transcriptomic characterization ………………………………………..116
RNA extraction and amplification ……………………………………………………………………………117
Mapping, normalization, and statistical analysis of RNA-Seq data …………………………..117
Results and Discussion .………………………………………………………………………………………………119
Competition under different electron acceptors ………………………………………………..……119
Transcriptional characterization of S. baltica OS185 and OS195 during competition
with limited nitrate as electron acceptor ……………………………………….……………………….121
RNA processing and Illumina RNA-Seq summary ……………………………………………123
Overview of transcriptomic profiles ………………………………………………………….……126
Differential expression caused by culturing methods (pure culture vs. co-culture)
in each strain …………………………………………………………………………………………….…130
Differential expression of OS195-cfp genes in pure culture and in co-culture
(Pc::Cc) ………………………………………………………………………………………………….130
viii

Differential expression of OS185-gfp genes in pure culture and in co-culture
(Pg::Cg) …………………………………………………………………………………………………131
Comparison on expression levels of shared gene between OS185 and OS195 in
pure culture and in co-culture (Pg::Pc and Cg::Cc) ……………………………………….133
Differentially expressed genes specifically in pure culture (Pc::Pg) ………….137
Differentially expressed genes in both pure culture and co-culture (Pc::Pg,
Cc::Cg) …………………………………………………………………………………………………..138
Co-culture specific differentially expressed genes (Cc::Cg)…………………….…141
RNA-Seq technology in transcriptomic analysis ……………………………………………………….142
Mechanisms of differential fitness …………………………………………………………………………..144
References ………………………………………………………………………………………………………………...146
CHAPTER VI
THESIS SUMMARY AND OUTLOOK
References …………………………………………………………………………………………………………………159

ix

LIST OF TABLES

Table 2.1. Shewanella baltica Strains Used in this Study ……………………………………………….…….21
Table 2.2. Marker Genes Used for MLST Analysis ……………………………………………….……………….23
Table 2.3. Primers Used for MLST Analysis ……………………………………………………………….…………24
Table 2.4. Neutrality Test and Estimation of Recombination (ρ) and Mutation Rates (θ) ……28
Table 3.1. Number of Probes Designed from Each Reference Strain …………………………………….45
Table 3.2. Composition of Variable Genes ……………………………………………….………………………….52
Table 3.3. Comparison of GC Content in Core Genome and Variable Genome …………….………55
Table 3.4. Description of COG Categories …………….…………….……………………………………………….62
Table 3.5. Correlation among Three Distance Matrices for Phylogenetic Inferences ……………63
Table 3.6. Types of Differential Expression ………………………………………………………………………….68
Table 4.1. Use of Different Electron Acceptors by S. baltica Strains …………………………………….97
Table 4.2. Two Sample T Test Shows that Strains from 1998 Use Significantly More Carbon
Sources ……………………………………………………………………………………………………………………………..102
Table 5.1. Strain Description and Growth Conditions …………………………………………………………115
Table 5.2. Sample Description ……………………………………………………………………………………………123
Table 5.3. Total RNA Yield and Amplification Efficiency ……………………………………………………125
Table 5.4. Summary of Mapped RNA-Seq Reads ………………………………………………………………125
Table 5.5. Number of Genes Classified in Different Cellular Functions ………………………………136
Table 5.6. Number of Genes Classified in Different COG Categories …………………………………136

x

LIST OF FIGURES

Figure 1.1. Water chemistry profile in the Gotland Deep from summer of 1986 …………………..7
Figure 2.1. Neighbor Joining tree of 46 S. baltica strains produced from gyrB sequences ……26
Figure 2.2. (A) Neighbor Joining tree of 46 S. baltica strains from concatenated sequences of
seven genes used for MLST analysis. (B) Phylogenetic network of the MLST loci among the S.
baltica strains ……………………………………………………………………………………………………………………..29
Figure3.1. Spheres of the diagram represent the gene content of the four sequenced S. baltica
strains (OS155, OS185, OS195, and OS223) ………………………………………………………………………….45
Figure 3.2a. Box-Plot of probe signals from CGH profiles of the four reference strains against
percent identity of the probes matched to their genomes …………………………………………………..50
Figure 3.2b. Sum of false positive and false negative rates under different cut-off values ….51
Figure 3.3. Heatmap of CGH signals from all genes and intergenic regions of the 46 S. baltica
strains …………………………………………………………………………………………………………………………………53
Figure 3.4. The sizes of core genome calculated from the CGH dataset …………………………….…54
Figure 3.5. GC contents of genes that are represented in the microarray ……………………………56
Figure 3.6. Histograms illustrating frequency of presence among 46 S. baltica strains of (a) all
genes and (b) intergenic regions represented in the microarray …………………………………………58
Figure 3.7. Distribution of genes in different COG functional categories in (a) pangenome (b)
genes whose presence varies among the 46 S.baltica strains (c) genes that are present in
greater than 90% of the strains …………………………………………………………………………………………..60
Figure 3.8. Hierarchical clustering of the CGH profile …………………………………………………………..64
Figure 3.9. Comparison of conserved core genes of (a) strains isolated from the 1980s and
1998 (b) strains recovered from at or below the anoxic-oxic interface and from more oxic
water sample ………………………………………………………………………………………………………………………66
Figure 3.10. (a) COG functional categories of annotated core genes of strains isolated from
the 1980s. (b) Differential expression patterns of core genes of strains from the 1980s in
OS185 and OS195 ……………………………………………………………………………………………………………….67

xi

Figure 3.11. (a) COG functional categories of annotated core genes of strains isolated from
anoxic waters. (b) Differential expression patterns of core genes of strains from anoxic waters
in OS185 and OS195 ……………………………………………………………………………………………………………70
Figure 3.12. Core genome comparison of Clade A, D, E and J ……………………………………………….72
Figure 3.13. Number of strain-specific genes revealed from CGH …………………………………………77
Figure 3.14. The patterns of genetic exchange apply to a larger collection of S. baltica
strains …………………………………………………………………………………………………………………………………81
Figure 4.1. Heatmap of 46 S. baltica strains based on Biolog profiles …………………………………..95
Figure 4.2. Principle Coordinate Analysis (PCoA) of Biolog metabolic profiles of the S. baltica
strains …………………………………………………………………………………………………………………………………96
Figure 4.3. Regression of Biolog similarity and genomic relatedness ………………………………….100
Figure 4.4. Hierarchical clustering of the Biolog profile ………………………………………………………101
Figure 5.1. Competition patterns observed among five S. baltica isolates in minimal medium
under aerobic and anaerobic condition with nitrate as the sole electron acceptor ……………120
Figure 5.2. Growth curves of fluorescence tagged S. baltica OS185 and OS195 strains ………122
Figure 5.3. Competition pattern between fluorescence tagged S. baltica OS195 and OS185 in
minimal medium supplemented with nitrate as sole electron acceptor ………………………….…122
Figure 5.4. BioAnalyzer profile of total RNA extracted from sample Pg-2 (a) before
amplification with RNA Integrity Number of 9.3 and (b) after amplification ………………………124
Figure 5.5. Hierarchical clustering of normalized mRNA reads from the 12 samples ………..…126
Figure 5.6. Canonical Correspondence Analysis of normalized mRNA reads from the 12
samples …………………………………………………………………………………………………………………………….127
Figure 5.7. Summary of pairwise comparisons performed with RNA-seq profiles of OS185 and
OS195 under nitrate condition ……………………………………………………………………………………….…128
Figure 5.8. Testing for differentially expressed genes in the four pairwise comparisons. Scatter
plots of log2 fold changes versus average numbers of reads were drawn for (a) Cc::Cg (b)
Pc::Pg (c) Cg::Pg (d) Cc::Pc ……………………………………………………………………………………………….…129
Figure 5.9. Insertion of phage genes in OS195 chromosome ………………………………………………131

xii

Figure 5.10. COG categories of differentially expressed genes by OS185 in co-culture vs. pure
culture (Cc::Pc) ……………………………………………………………………………………………………………….…132
Figure 5.11. Heatmap of differentially expressed genes between OS185 and OS195 in coculture (Cc::Cg) and in pure culture (Pc::Pg) ………………………………………………………………….…..134
Figure 5.12. Summary of numbers of differentially expressed gene shared by both OS185 and
OS195 between the two strains in co-culture and in pure culture …………………………………..…135
Figure 5.13. Differentially expressed genes specific to Pc::Pg comparison …………………………137
Figure 5.14. Differentially expressed genes from both Pg::Pc and Cg::Cc comparisons ………140
Figure 5.15. COG categories of differentially expressed genes shared by OS185 and OS195
from Pg::Pc comparison ……………………………………………………………………………………………….……140
Figure 5.16. COG categories of differentially expressed genes shared by OS185 and OS195
from Cg::Cc comparison ……………………………………………………………………………………………….……142
Figure 6.1. COG functional comparison of S. baltica gene core and Burkholderia cenocepacia
gene core ……………………………………………………………………………………………………………………….…156

xiii

LIST OF ABBREVIATIONS

RAPD: Randomly Amplified Polymorphic DNA
MLST: Multi Locus Sequence Typing
HGT: Horizontal Gene Transfer
MLEE: Multi-Locus Enzyme Electrophoresis
EM: Electrophoretic Mobilities
RFLP: Restriction Fragment Length Polymorphism
AFLP: Amplified Fragment Length Polymorphism
ANI: Average Nucleotide Identity
PCR: Polymerase Chain Reactions
RDP: Recombination Detection Program
GARD: Genetic Algorithm for Recombination Detection
ST: Sequence Type
LMW: Low-Molecular-Weight
CGH: Comparative Genomic Hybridization
Oligoarrays: Oligonucleotide Arrays
COG: Clusters of Orthologous Groups
LEE: Locus pf Enterocyte Effacement
EPEC: Enteropathogenic Escherichia Coli
DMSO: Dimethyl Sulfoxide
TMAO: Trimethylamine-N-Oxide
xiv

PM: Phenotype Microarray
PCoA: Principle Coordinate Analysis
DNRA: Dissimilatory Nitrate Reduction to Ammonium
DMS: Dimethylsulfide
TMA: Trimethylamine
DSN: Duplex-Specific Nuclease
PF: Passed Filtering
NGS: Next Generation Sequencing
ncRNA: Non-Coding RNA

xv

CHAPTER I
Introduction

1

A perspective on bacterial specialization. Bacteria and Archaea are arguably the most
environmentally versatile organisms on earth, and represent the greatest abundance as well as
divergence in the tree of life. As the pioneer of life on this planet, originating approximately 3.7
billion years ago, prokaryotic life had evolved for over 3 billion years without eukaryotic
competition, having thus occupied most if not all habitable niches on earth (1) (2). Microbial
evolution and diversification (speciation) has been a major driver for new genotypic and
phenotypic life on Earth. Yet the principles of these processes are not well understood. Since
the principle of evolution by means of natural selection was first proposed by Darwin and later
th

accepted into more branches of biology in the 20 century (3), numerous efforts have been
focused on understanding the connections between natural environment and the biological
outcome (4). With the integration of molecular biology and population genetics, the
mechanisms underlying adaptation and speciation are also becoming clearer.
Through all the endeavors, a series of theories and concepts were incorporated into the
evolutionary framework. While natural selection is still the only known driving force of
adaptation, a number of non-adaptive causes of evolution were found including processes such
as mutation, genetic drift, genetic hitchhiking (recombination), and horizontal gene transfer,
which comprise the known engine for generation of genomic diversity (5) (6) (7) (8) (9) (10) (11).
Under this framework, genotypic traits, reflected through phenotypes of organisms, are tested
by and subjected to natural selection and random drift, which eventually leads to either fixation
or extinction of the genotype.
Prior to the genome sequencing era, the determination of genetic elements attributable
to phenotypic traits was a painstaking and time-consuming process. Microbial genetics has
2

greatly benefited from the automated sequencing technologies. Meanwhile, as more molecular
and bioinformatic tools emerged, which greatly facilitated interpretation of the large amount of
sequence data, high throughput comparative genomic studies became possible and started to
provide valuable insights into what drove the genotypic and phenotypic diversification of life (4)
(12).
Microbes are ideal candidates for investigating evolutionary trajectories as well as
mechanisms. They occupy the vastest number of niches, providing materials for assessing
adaptation in all environmental dimensions; they have relatively small genome sizes, which
saves the cost for sequencing or genotyping as well as reduces the amount of analysis efforts;
their shorter generation time also makes possible experimental evolution studies that allowed
them to evolve tens of thousands of generations in a limited time (7).
Here we report an evolutionary study of a bacterial population Shewanella baltica, using
isolates taken along an environmental gradient in an estuarine ecosystem, and sampled over
nearly two decades. Through studying specialization among these strains, we hope to provide
insights into how different the S. baltica genotypes are, how they are likely to position
themselves in an ecological gradient, and how they interact with each other as a microbial
community.

The versatile Shewanella. Shewanella genus belongs to gammaproteobacteria that are well
known for their diverse respiratory capabilities. Strains in this genus are capable of utilizing a
-

0

wide variety of electron acceptors including O2, NO3 , S , Mn (III), Fe(III), Cr(VI), V(V),
dimethylsulfoxid (DMSO), trimethylamine-N-oxide (TMAO), fumarate, and more (13). They are
3

among the most heavily studied environmental micro-organisms and have become a model
system for studying metal bioremediation, biofuel generation, as well as microbial ecology (14).
Shewanella species are known to inhabit a wide range of niches, including soil, marine,
freshwater and sediments, and are often associated with energy-rich environments where
anaerobic metabolism of organic matter is carried out by complex microbial communities (15).
Many of these environments are subjected to highly variable concentrations of organic and
inorganic electron acceptors and hence redox gradients can form, while the remarkable
electron accepting ability of Shewanellae as well as their capability to sense and taxis towards
electron acceptors allow them to be well adapted to inhabiting these organic-rich, redox
fluctuating environments (16).
Shewanella started to gain much interest after the publication describing a strain
Shewanella oneidensis MR‑1, formerly Alteromonas putrefaciens, that was shown to be capable
of reducing manganese and iron oxides (17). It was noted for its ability to transfer electrons
directly to solid metal as well as its extraordinary respiratory versatility, being able to utilize
over ten electron acceptors. The genome sequence of S. oneidensis MR‑1 was released in 2002
(18), which facilitated establishment of MR-1 as a model organism for genetic and molecular
biochemical studies of the Shewanellae genus as well as the metal-reducing bacteria in general
for the purpose of bioremediation.
The Shewanella genus currently includes 56 recognized species
(http://www.bacterio.cict.fr/), and some have been shown to occur in an even more diverse
range of environments. The genomes of 30 strains of Shewanella have either been fully
sequenced or are in the process of being sequenced or gap closure (according to NCBI Microbial
4

Genome database). Comparative genomic and proteomic analyses have been performed with a
number of Shewanella strains, and provide valuable insights into genomic and phenotypic
evolution among the Shewanella species (19). The correlation between genomic variation and
phenotypic differences among the Shewanella species was examined. Meanwhile, core and
variable gene pools were determined through comparative genomics, and it was shown that
hypothetical and mobile genetic element-related genes dominate the variable gene pool of the
species, indicative of the importance of horizontal gene transfer in evolution of Shewanella
genomes.

The history behind the Shewanella baltica population in the Baltic Sea. The Baltic Sea
is the world’s largest brackish water environment. It is also an estuarine ecosystem with
pronounced eutrophication due to anthropogenic activity creating overloads of phosphorus and
nitrogen contaminants (20). Gotland Deep is a stratified basin 240 m deep located in the central
Baltic Sea (21). A comprehensive set of long-term chemical and hydrographical data has been
collected from this site. Vertical mixing of water is inhibited due to a halocline present at 60 to
90 meters depth. As stagnation periods prolong, the deep water displays oxygen deficiency and
H2S accumulation from the sediment (22). This may have caused the establishment of microbial
populations particularly adapted to these conditions. Denitrification was considered the major
factor counteracting eutrophication, while organic carbon derived from phytoplankton primary
production seemed to be the major electron donor (23).
To study the microbial ecology of this brackish water environment, especially the
microbial flora catalyzing this part of the nitrogen cycle in the central Baltic, water samples
5

were collected from different depths in Gotland Deep in the summers of 1986 and 1987.
Different methods were used to isolate denitrifying bacteria (22), yielding 113 S. baltica strains
representing 77% of the total population of culturable denitrifiers (24). S. baltica strains were
shown to be specifically enriched in the oxic-anoxic interface from 80 m to 140 m, where
vigorous denitrification activity was found (25), as well as strong gradients of oxygen, nitrate
and hydrogen sulfide [Figure 1.1]. Thirty-six of these 113 strains were selected for further
characterization based each having unique profiles from randomly amplified polymorphic DNA
(RAPD) genotyping (25), while we selected four of them, OS155, OS185, OS195 and OS223,
isolated from different redox zones, for whole genome sequencing.
The stratification of the redox environment in the Gotland Deep was disrupted by a
major water turn over event in 1993 (26) (27), after which a new but similar stratification was
established (28). In a cruise in 1998, ten additional Shewanella baltica strains were recovered
from the same sampling station using a similar procedure. Strikingly, phylogenetic profiles of
most of these strains clustered with those already found in the 1980s, indicating the stability
over time of the genetic structure of the S. baltica, even though the water column had been
completely mixed by the turnover event. Therefore, these strains are a valuable resource for
investigating both the evolution over space across the redox gradient, and evolution over time
across the 12 years.

6

Figure 1.1. Water chemistry profile in the Gotland Deep from summer of 1986. The
concentrations of oxygen, hydrogen sulfide and nitrate are highlighted in blue, green and red
color. Temperature (T) and salinity (S) are also shown. The yellow box indicates the depth
where S. baltica strains were isolated. Figure is modified from (25). For interpretation of the
references to color in this and all other figures, the reader is referred to the electronic version
of this thesis.

Thesis Outline. Although systematic studies of comparative physiology, genomics and
transcriptomics have been performed on Shewanella, this has been done on either a single or a
few strains of different species, which limits our ability to understand adaptation at the
population level (29) (19). In this thesis, I have chosen the Shewanella baltica species as the
model to study bacterial specialization because I had access to a large collection (46) of already
partially characterized strains that were isolated from defined positions along a stable spatial
environmental gradient. This offered the potential to further our understanding of ecology and
evolution using the Shewanella spp. as the model. I used several methods to characterize
7

specialization at the level of both genotypic and phenotypic variation, and inferred their
evolutionary trajectories and potential underlying mechanisms.
The following chapters are ordered such that the story begins by identifying genetic
relatedness at a high level among all strains, followed by comparison of gross differences and
commonalities among specific lineages, and ends with a detailed transcriptomic analysis on two
competed model strains. Specifically, Chapter 2 introduces a few genotyping systems to
characterize the phylogenetic relatedness among all sampled strains, which provides initial
evidence for specialization among the S. baltica population. Chapter 3 describes my attempt to
further exploit variation of genomic contents among these strains through Comparative
Genomic Hybridization (CGH), which aims at discovery of genomic signatures potentially
important for specialization. Comparative transcriptomic analysis on selected strains also
provides additional support for functional importance of differentially distributed genes.
Particularly, genomic contents of several lineages representing strains from different niches are
contrasted and compared, yielding important insights on specialization and evolution among
these lineages. Meanwhile, the S. baltica gene core is subjected to comparison with that of
Burkholderia cenocepacia in order to provide perspective at the species level on how well
genetics can reflect the greater ecology of bacterial species.
Chapter 4 focuses on phenotypic examination of S. baltica strains, which includes high
throughput characterization of carbon source utilization as well as testing respiratory
capabilities of several electron accepters. These phenotypic profiles are further compared with
the environmental parameters as well as genomic profiles to reveal how well ecology and
genetics can be reflected at the phenotypic level. Lastly, Chapter 5 examines differential fitness
8

of selected S. baltica strains under defined redox conditions through competition assays.
Illumina RNAseq technology was used for transcriptomic characterization of two model strains
under competition, which also demonstrates the power of Next Generation Sequencing (NGS)
platform for generating large amounts of sequence reads starting from very limited RNA
material. This chapter’s focus is even more specific, and aims at revealing adaptive strategies of
S. baltica in redox specialization. It is clear that specialization among the S. baltica population
has occurred at multiple levels, and this is summarized in Chapter 6.

9

REFERENCES

10

References
1.

Staley JT, R. W. Castenholz, R. R. Colwell, J. G. Holt, M. D. Kane, N. R. Pace, A. A. Salyers,
and J. M. Tiedje (1997) The Microbial World: Foundation of the Biosphere. The American
Academy of Microbiology.

2.

Brocks JJ, Logan GA, Buick R, & Summons RE (1999) Archean Molecular Fossils and the
Early Rise of Eukaryotes. Science 285(5430):1033-1036

3.

Provine WB (1988) Progress in Evolution and Meaning in Life. Evolutionary Progress.
University of Chicago Press:49–79.

4.

Koonin EV (2008) Darwinian evolution in the light of genomics. Nucleic Acids Research
37(4):1011-1034.

5.

Brown EW, LeClerc JE, Kotewicz ML, & Cebula TA (2001) Three R's of bacterial evolution:
How replication, repair, and recombination frame the origin of species. Environmental
and Molecular Mutagenesis 38(2-3):248-260.

6.

Christophe Fraser EJA, Martin F. Polz, Brian G. Spratt, William P. Hanage (2009) The
bacterial species challenge: making sense of genetic and ecological diversity. Science
323:6.

7.

Cohan FM & Koeppel AF (2008) The Origins of Ecological Diversity in Prokaryotes.
Current Biology 18(21):R1024-R1034.

8.

Fraser C, Hanage WP, & Spratt BG (2007) Recombination and the Nature of Bacterial
Speciation. Science 315(5811):476-480.

9.

Hanage WP, Spratt BG, Turner KME, & Fraser C (2006) Modelling bacterial speciation.
Philosophical Transactions of the Royal Society B: Biological Sciences 361(1475):20392044.

10.

Weinbauer MG & Rassoulzadegan F (2003) Are viruses driving microbial diversification
and diversity? Environmental Microbiology 6(1):1-11.

11.

Gogarten JP & Townsend JP (2005) Horizontal gene transfer, genome innovation and
evolution. Nature Reviews Microbiology 3(9):679-687.

12.

Doolittle WF & Papke RT (2006) Genomics and the bacterial species problem. Genome
Biology 7(9):116.

13.

Fredrickson JK, et al. (2008) Towards environmental systems biology of Shewanella.
Nature Reviews Microbiology 6(8):592-603.

11

14.

Tiedje JM (2002) Shewanella - the environmentally versatile genome. Nature
Biotechnology 20:2.

15.

Hau HH & Gralnick JA (2007) Ecology and Biotechnology of the GenusShewanella.
Annual Review of Microbiology 61(1):237-258.

16.

Bencharit S & Ward MJ (2005) Chemotactic Responses to Metals and Anaerobic Electron
Acceptors in Shewanella oneidensis MR-1. Journal of Bacteriology 187(14):5049-5053.

17.

Myers CR & Nealson KH (1988) Bacterial manganese reduction and growth with
manganese oxide as the sole electron acceptor. Science 240(4857):1319-1321

18.

Heidelberg JF, et al. (2002) Genome sequence of the dissimilatory metal ion–reducing
bacterium Shewanella oneidensis. Nature Biotechnology 20(11):1118-1123.

19.

Konstantinidis KT, et al. (2009) Comparative systems biology across an evolutionary
gradient within the Shewanella genus. Proceedings of the National Academy of Sciences
106(37):15909-15914.

20.

Elmgren R (1989) Man's Impact on the Ecosystem of the Baltic Sea: Energy Flows Today
and at the Turn of the Century. AMBIO:326-332.

21.

Fonselius S & Valderrama J (2003) One hundred years of hydrographic measurements in
the Baltic Sea. Journal of Sea Research 49(4):229-241.

22.

Brettar I & Rheinheimer G (1991) Denitrification in the Central Baltic: evidence for H 2Soxidation as motor of denitrification at the oxic-anoxic interface. Marine Ecology
Progress Series 77:157-169.

23.

Brettar I & Höfle MG (1993) Nitrous oxide producing heterotrophic bacteria from the
water column of the central Baltic: abundance and molecular identification. Marine
Ecology Progress Series 94:253-265.

24.

Brettar I (2002) Shewanella denitrificans sp. nov., a vigorously denitrifying bacterium
isolated from the oxic--anoxic interface of the Gotland Deep in the central Baltic Sea.
International Journal of Systematic and Evolutionary Microbiology 52(6):2211-2217.

25.

Ziemke F, Brettar I, & Höfle MG (1997) Stability and diversity of the genetic structure of
a Shewanella putrefaciens population in the water column of the central Baltic. Aquatic
Microbial Ecology 13:63-74.

26.

Håkansson B, Broman B, & Dahlin H (The flow of water and salt in the Sound during the
Baltic major inflow event in January 1993. ICES Statutory Meeting Dublin, Paper
C.M.1993/C:57.

12

27.

Matthäus W & Lass H-U (1995) The recent salt water inflow into the Baltic Sea. J. Phys.
Oceanogr. 25:280-286.

28.

Matthäus W, et al. (The Baltic Sea in 1996- Continuation of Stagnation and Decreasing
Phosphate Concentrations. Deutsche Hydrographische Zeitschrift 48(2):161-174.

29.

Beliaev AS, et al. (2005) Global Transcriptome Analysis of Shewanella oneidensis MR-1
Exposed to Different Terminal Electron Acceptors. Journal of Bacteriology 187(20):71387145.

13

CHAPTER II
Phylogenetic Analysis of a Shewanella baltica Population Using Multi
Locus Sequence Typing

Thanks to Jorge Rodrigues and Ingrid Brettar, who contributed the gyrB gene sequences.

14

Abstract
In order to determine the evolutionary relatedness among 46 representative
Shewanella baltica isolates, a Multi Locus Sequence Typing (MLST) system was developed using
seven genes previously identified as phylogenetically informative from the Shewanella gene
core. Eleven distinct clades were identified. When phylogenetic relatedness was compared with
the environmental profiles, niche specialization over space and over time was observed. Eightysix percent of strains isolated from anoxic water were grouped in one clade, suggesting
specialization for the anoxic niche, while strains isolated from other depths were distributed
throughout the phylogram. Clustering of strains isolated from different years also indicated
specialization over time. Finally, although extensive recombination was observed in four
sequenced S. baltica strains and thought to dominate their evolutionary paths, this pattern was
not strongly supported for the this larger population based upon allelic profiles of the MLST loci.

15

Introduction
In the eyes of zoologists, species is one of the basic units of biological organization, i.e.
classification. As stated in Wikipedia: “A species is often defined as a group
of organisms capable of interbreeding and producing fertile offspring.” When it comes to the
microbial world, there is no species concept but there is a species definition which is profoundly
different than for higher order biology. Although Bacteria and Archaea have the greatest
abundance as well as divergence in the tree of life and are critical to a habitable planet, the
principles explaining their ecology and evolution are poorly understood. While the key to the
species definition in zoology is the ability of exchanging genetic materials during reproduction,
the same principle is not applicable to microbes (1) (2) (3) (4). The boundaries among
microorganisms are blurred by the presence of horizontal gene transfer (HGT), which allows
one cell to acquire genetic material from cells that are distantly related, and because a
considerable portion of any strain’s genome varies within a species, i.e. the pangenome (5) (6).
Although homologous recombination is somewhat similar to the sexual behavior in higher
organisms, the frequency of recombination can vary greatly among different taxa. Moreover,
boundaries of recombination are not yet clear (7) (8) (9).
Early microbiologists used phenotypic characteristics to differentiate bacterial lineages,
such as those on the morphological level, or by specific (or combinations of) biochemical assays.
However, these assays provided limited resolution in distinguishing bacterial lineages, and
there was no way to justifiably argue which assays determined the real ecological differences
among different strains. For example, Multi-Locus Enzyme Electrophoresis (MLEE) was used to
differentiate bacterial lineages based upon different electrophoretic mobilities (EM) of core
16

metabolic enzymes (10). However, this approach has the drawback of inability to distinguish
silent mutations, which does not cause amino acid changes, and also failed to differentiate
between enzymes with different amino acids but without sufficiently different EM to generate
distinct bands. Serotyping was also established for differentiating bacterial lineages, especially
pathogenic strains, based on cell surface antigens (11). But it heavily relies on presence of a few
antigenic loci and could be biased due to unpredictability of reactivities of antibodies with
different antigenic variants. Furthermore, the promiscuous nature of bacteria, i.e., the ability of
harboring genes for various biochemical pathways through HGT from other microorganisms,
complicates many of these categorizing efforts (6). Functional features as revealed from
biological assays, although still and will always be essential for inferring ecological importance,
are not always appropriate or sufficiently resolving for demarcating bacterial lineages.
As molecular typing schemes emerged and were put to use for differentiating bacterial
lineages, much weight was given to the genetic coherence among microorganisms. Many of
these molecular typing approaches were based on similarity in genomic structures and DNA
profiling, determined by methods involving restriction digestion, PCR amplification, DNA-DNA
hybridization, etc. For example, both restriction fragment length polymorphism (RFLP) and
amplified fragment length polymorphism (AFLP) detection approaches employ digestion of DNA
samples by restriction enzymes and visualization of DNA fragment length polymorphism
through gel electrophoresis, while the latter method involves an additional amplification step
resulting in higher sensitivity (12) (13). The random amplification of polymorphic DNA (RAPD)
method, which is also based on PCR amplification, though without enzymatic digestion, can also
give genetically distinct individuals different bands from amplified products (14). Microarray
17

based pair wise DNA-DNA hybridization has also been used for determining genomic similarities
among lineages, and was used by systematists for demarcating bacterial species using the 70%
in DNA-DNA association criterion (15). However, all these methods lack either resolution,
reproducibility and/or accessibility by different labs, whereas the use of Multi Loci Sequence
Typing (MLST) overcomes these limitations.
MLST is a method that utilizes multiple gene sequences and can reveal detailed intraspecies phylogenetic relationships (16). It directly measures the nucleotide variation in
sequences of a set of genes and uses allelic profiles for genetic characterization of the strains.
In order to achieve an acceptable level of identification power while being time and cost
effective at the same time, sequences of seven or eight genes are commonly used. Since MLST
is sequence based, the information is fully reproducible, unambiguous and portable. The
sequence information can be easily exchanged between laboratories, as can primer sequences
and amplification protocols. MLST also has the potential of being standardized and automated.
Combined with bioinformatic techniques for population genetics, MLST can be efficiently used
to investigate evolutionary relationships among different bacterial lineages.
The selection of MLST loci has been based on their being unlinked in the genome and
containing conserved sites for PCR primer design. Selection of MLST genes has grown up
reflecting interests and databases of particular investigators, and varies according to the
species studied. The most commonly used MLST loci were of housekeeping genes involved in
central metabolism such as aspartate aminotransferase (aspC), malate dehydrogenase (mdh),
sigma factor of RNA polymerase (rpoS), DNA primase (dnaG), etc (17). However, the
disadvantage of past gene selection strategies is that it was unknown how well the
18

phylogenetic relationships derived from these genes approximated the real phylogeny of the
studied strains. In a comparative genomics study that aimed to address this question, genome
sequences of Salmonella, Burkholderia, and Shewanella groups were utilized, and the
phylogeny based upon each orthologous gene within each group was compared with the
phylogeny based upon a concatenation of all the common genes in the whole-genome as well
as the ANI (average nucleotide identity) of the genes in common (18). A number of genes were
selected from each genus that best reflected the whole-genome based phylogeny. In this study,
I followed this MLST strategy using genes predicted to reflect the true phyogeny to reveal intraspecies phylogenetic relationship among a S. baltica population from the Baltic Sea. Through
investigating genetic signatures of the sequences as well as association between the
phylogenetic relatedness and environmental parameters, I hope to connect the relationships
among the strains with ecology, and get a glimpse into their evolutionary stories.

19

Materials and Methods
Bacterial strains and growth conditions. Bacterial strains used are described in [Table 2.1]. LB
medium (Acumedia, Lansing, MI, USA) was used for their growth. Anaerobic M1 medium was
made under oxygen free argon gas. DNA was extracted using a modified CTAB genomic
extraction protocol (19).

20

Table 2.1. Shewanella baltica Strains Used in this Study
Strain

Isolation Depth /m

Isolation Method

Shewanella baltica OS106

90

NBNO3

a

1986

Shewanella baltica OS107

120

NBNO3

1986

Shewanella baltica OS109

120

NBNO3

1986

Shewanella baltica OS110

120

NBNO3

1986

Shewanella baltica OS117

130

NBNO3

1986

Shewanella baltica OS155

90

ZB

Shewanella baltica OS167

140

ZB

b

Isolation Year

1986
1986
c

1986

Shewanella baltica OS183

120

ZBan

Shewanella baltica OS185

120

ZBan

1986

Shewanella baltica OS187

120

ZBan

1986

Shewanella baltica OS189

130

ZBan

1986

Shewanella baltica OS190

130

ZBan

1986

Shewanella baltica OS193

140

ZBan

1986

Shewanella baltica OS195

140

ZBan

1986

Shewanella baltica OS223

120

0.1NBNO3

d

1986

Shewanella baltica OS225

130

0.1NBNO3

1986

Shewanella baltica OS230

130

0.1NBNO3

1986

Shewanella baltica OS250

120

NBNO3

1986

Shewanella baltica OS252

120

NBNO3

1986

Shewanella baltica OS286

140

0.1NBNO3

1986

Shewanella baltica OS288

130

0.1NBNO3

1986

Shewanella baltica OS625

80

THNO3

e

1987

Shewanella baltica OS628

80

THNO3

1987

Shewanella baltica OS631

80

THNO3

1987

Shewanella baltica OS638

120

THNO3

1987

Shewanella baltica OS641

120

THNO3

1987

21

Table 2.1. (Cont’d)
Strain

Isolation Depth /m

Isolation Method

Isolation Year

Shewanella baltica OS645

130

THNO3

1987

Shewanella baltica OS650

130

THNO3

1987

Shewanella baltica OS652

130

THNO3

1987

Shewanella baltica OS678

110

THNO3

1987

Shewanella baltica OS681

110

THNO3

1987

Shewanella baltica OS690

80

NBNO3

1987

Shewanella baltica OS696

120

NBNO3

1987

Shewanella baltica OS697

120

NBNO3

1987

Shewanella baltica OS700

120

NBNO3

1987

Shewanella baltica OS710

110

NBNO3

1987

Shewanella baltica OS712

110

NBNO3

1987
f

1998

Shewanella baltica BA37

120

(m) THNO3

1998

120

(m) THNO3

1998

175

(m) THNO3

1998

Shewanella baltica BA194

175

(m) THNO3

1998

Shewanella baltica BA196

175

(m) THNO3

1998

Shewanella baltica BA170

120

(m) THNO3

1998

Shewanella baltica BA175

e

1998

Shewanella baltica BA185

d

(m) THNO3

Shewanella baltica BA173

c

80

Shewanella baltica BA62

b

(m) THNO3

Shewanella baltica BA38

a

80

120

(m) THNO3

1998

Nutrient Broth medium supplemented with nitrate under anaerobic condition
ZoBell agar plate under aerobic condition
ZoBell agar plate under anaerobic condition
1:10 diluted Nutrient Broth medium supplemented with nitrate under anaerobic condition
Minimal medium supplemented with nitrate and thiosulfate under anaerobic condition

f

Modified minimal medium supplemented with nitrate and thiosulfate under anaerobic
condition

22

MLST Analysis. Seven genes (SO_0578, SO_0625, SO_1771, SO_2183, SO_2615, SO_2706,
SO_4702 from S. oneidensis MR-1) from the conserved gene core of nine Shewanella genomes
were selected for this assay based on a previous study (18). Primers for Polymerase Chain
Reactions (PCR) were designed for these genes based on genomic sequences of the four
sequenced S. balitca strains (S. baltica OS155, S. baltica OS185, S. baltica OS195, and S. baltica
OS223) [Table 2.2]. Genes were amplified by PCR [Table 2.3]. The products were purified using
ExoSap-IT (USB from Affymetrix, Cleveland, Ohio, USA), and were sequenced by the MSU
Research Technology Support Facility. Sequences were aligned with MUSCLE (20). A Neighbor
Joining tree was built using concatenated sequences of the seven genes by MEGA version 4
software (21). Phylogenetic networks were analyzed using the neighbor-net algorithm (22) and
the uncorrelated P distance with the program SplitsTree 4 (23).

Table 2.2. Marker Genes Used for MLST Analysis
Locus Tag
SO_0578
SO_0625
SO_1771
SO_2183
SO_2615
SO_2706
SO_4702

Annotation
Periplasmic metalloprotease
Periplasmic protein with Sel1-like repeats
D-glycerate transporter, GlyT
Periplasmic ErfK/YbiS/YcfS/YnhG family protein
Aminodeoxychorismate lyase, PabC
Succinylarginine dihydrolase, AstB
Glutathione reductase, Gor

23

Length/bp
696
537
594
321
474
702
903

Table 2.3. Primers Used for MLST Analysis

SO_0578
SO_0625
SO_1771
SO_2183
SO_2615
SO_2706
SO_4702
gyrB

Forward
ACGCCGCCGARMAAMGA
CGCTGCATAAATTCATCTGTCTW
CGGCGTGATYGCATTTATTGTTAT
GACTMTCGCGGCACTSRTTTCAC
CYYTMGCCATTGAATATCCACATC
CCCCWCAAGAGCGCCCAGAYC
ATTATGGWTTYGAYGTYWCTGTTA
GAAGTCATCATGACCGTTCTG

Reverse
STCGARGCGGCAAAGTTATGAT
CGGCRCACGCATTTTCYA
AAWGGYGCWGAAGGGAAGTTRG
CARCCGTCGGTCCAGTTGTATTG
ATCGCCAGCACATCCACWA
CCDCCGWTTTGCATGCTTTGTTT
TTRGTKGCGCCCATTTTCAT
AGCAGGGTACGGATGTGCGAGCC

Temperature/
54
52
53
58
53
61
54
55

Mixed Base Site Code: R=A,g; Y=C,T; M=A,C; K=g,T; S=C,g; W=A,T; D=A,g,T; N=A,C,g,T.

The DnaSP version 5.10 (24) was used to calculate the ratios of synonymous and nonsynonymous substitution rates (Ks and Ka, respectively) and to perform Tajima’s D test for
neutrality. Population mutation and recombination rates were estimated using a likelihood
based method implemented in the LDhat2.0 version 2.1 (25) using similar parameters as
described in (26). Detection of recombinational breakpoints was conducted through the RDP
v3.4.4 (recombination detection program) (27) as well as GARD (genetic algorithm for
recombination detection) (28) (29) program using the default settings. Distance matrices from
gyrB and MLST analysis were generated using maximum composite likelihood model in MEGA
4.0.

24

Results
gyrB analysis. 16S rRNA gene analysis already showed that these strains belong to the same
species (30). Sequences of the gyrB genes were also screened and analyzed. Using these
sequences, we were able to make a preliminary determination of the genetic relationships
among the S. baltica strains. From the total 1122 bp amplicon of the gyrB gene, a phylogeny of
10 well supported lineages (bootstrap value > 90%) was revealed [Figure 2.1].

25

OS193
OS641
OS678
OS190
93 OS195
OS650
OS652
OS710
OS645
OS167
99 OS109
OS631
OS628

OS690
OS110
100 OS697
OS252
OS155
OS117
OS638
100 OS681
OS185
99 OS187
OS225
BA170
BA173A
OS288
100 OS107
OS700
OS250
100 OS189
OS286
OS696
99
OS230
BA194
OS625
BA62
OS106
100 BA185
BA196
BA37
99 OS183
BA175
BA173B
OS223
BA38

100

0.005

Figure 2.1. Neighbor Joining tree of 46 S. baltica strains produced from gyrB sequences. The
strain S. oneidensis MR-1 was used as the out-group but is not shown in the Figure. Bootstrap
values above 90% are shown on branches based on 1000 bootstrap replicates.
26

MLST analysis. Seven genes from the Shewanella conserved gene core were chosen for further
phylogenetic characterization of these S. baltica isolates (Locus tags as in Shewanella oneidensis
MR-1: SO_0578, SO_0625, SO_1771, SO_2183, SO_2615, SO_2706, SO_4702) [Table 2.2].
Twenty-six unique sequence types (ST) were revealed from the concatenated sequences of
these loci. Eleven well-supported clusters were identified, among which nine were confirmed
in the gyrB phylogram [Figure 2.2A]. The SO_4702, SO_0578 and SO_1771 genes account for
more than half of all polymorphic sites, whereas others provide comparatively less
discriminatory power. Ratios of synonymous and non-synonymous substitution rates (Ks and Ka,
respectively) were calculated based upon sequence information at each locus, followed by
Tajima’s test of neutrality *Table 2.4]. All the Ks/Ka ratios (except SO_2615 and SO_4702) are
much larger than 1, indicating the presence of purifying selection against amino acid changes.

27

Table 2.4. Neutrality Test and Estimation of Recombination (ρ) and Mutation Rates (θ)
SO_0578

SO_0625 SO_1771 SO_2183 SO_2615 SO_2706 SO_4702

Polymorphic Sites
Ka
Ks
Ks/Ka
Watterson θ
Estimated ρ

97
0.00667
0.13958
20.9
22.1
35.5

46
0.00261
0.12357
47.3
10.5
10.5

71
0.00188
0.08492
45.2
16.2
24

23
0.00129
0.08012
62.1
5.2
4.5

43
0.00529
0.05083
9.6
9.8
4.5

57
0.00276
0.08665
31.4
13.0
7.5

131
0.00528
0.13912
26.3
29.8
21.5

θ per site
ρ per site
Recombination to
Mutation Ratio
RDP*

0.0317
0.0510

0.0195
0.0196

0.0272
0.0404

0.0163
0.0140

0.0206
0.0095

0.0185
0.0107

0.0330
0.0238

0.804

0.502

0.743

0.430

0.230

0.289

0.361

No

No

No

No

No

No

Yes

* Only recombination signals detected by at least three methods incorporated in the RDP
program is displayed.

28

(A)
Figure 2.2.

29

Figure 2.2 (Cont’d)

Figure 2.2. (A) Neighbor Joining tree of 46 S. baltica strains from concatenated sequences of
seven genes used for MLST analysis. Strains with symbol
were isolated from 1998 while all
other strains were isolated from 1986 and 1987. The strain S. oneidensis MR-1 was used as outgroup but is not shown. Bootstrap values above 50% are shown on branches based on 1000
bootstrap replicates. Color panels correspond to the depth where strains were isolated. (B)
Phylogenetic network of the MLST loci among the S. baltica strains. The parallelogram was
constructed using the neighbor-net algorithm and depicts conflicting phylogenetic signals
among branches. (For interpretation of the references to color in this and all other figures, the reader
is referred to the electronic version of this thesis -- This sentence was already in Figure 1.1)

30

Recombination analysis on genes selected for genotyping. To investigate the influence of
recombination on the phylogeny of the ST’s, phylogenetic networks were examined using the
SplitsTree analysis [Figure 2.2B]. A number of parallel branches were found, indicating the
presence of phylogenetic incompatibilities among the MLST loci. Such incompatibilities can
result from evolutionary processes such as recombination and recurrent mutation. Based on
the amount of parallel branches, which is indicative of the number of shared evolutionary paths
among ST’s, various levels of recombination were observed among different phylogroups.
Specifically, the genotypes in Clade C showed the most extensive amount of recombination,
while there also seemed to be recent recombination events between Clade A and B. Moreover,
OS625 in Clade B, an isolate from 1987 obtained in a more oxic zone, shared common paths
with both OS187 and ST’s in Clade A, most of which were from at or below the oxic-anoxic
interface, implying the involvement of recombination during possible migration of the S. baltica
strains.
Population recombination (ρ) and mutation (θ) rates were approximated individually for
each marker gene. The recombination to mutation ratio calculated for the seven loci ranged
from 0.230 to 0.804, indicating the dominance of clonal evolution paths. The sequences were
subsequently analyzed in the RDP and GARD program for identification of recombination break
points. While RDP failed to detect any recombination signal in any of the MLST loci except for
SO_4702, recombination break points with significance level (p < 0.01) were found in all the
seven loci by the GARD program [Table 2.4]. However, it has been reported that GARD tends to
overestimate the effects of recombination in certain cases.

31

Discussion
Genotyping indicates niche specialization and genetic continuity over time. When
phylogenetic groups were compared to the geographic and temporal distribution of these
isolates, significant clustering patterns were found with regard to depths and years (p < 0.01).
From correlating the MLST phylogeny with the depth from which the strain was isolated, one
notable trend was the predominance of S. baltica strains in Clade A from the anoxic zone - 80%
of strains in Clade A were isolated from water at and below the oxic-anoxic interface.
Specifically, 75% of strains isolated from the anoxic zone in 1986 and 100% of strains isolated
from the anoxic zone in 1987 are members in this clade. Interestingly, while all strains in the
Clade A isolated in 1986 were from at and below the oxic-anoxic interface, this genotype
seemed to have expanded in 1987 and was recovered from more diverse depths (including
above the oxic-anoxic interface). Although none of the strains isolated from the anoxic water in
1998 were grouped in this clade, we speculate that different strategies of adaptation to the
oxygen depleted environment were developed in the 1998 strains considering that some of
these strains were isolated from a distinct depth of 175m. In addition, the MLST phylogeny
indicates the persistence of certain clades over time in these S. baltica lineages. While seven of
the eleven clades contain strains isolated from both sampling years of 1986 and 1987, two
clades (J, K) contain strains isolated from both 1998 and the years in the 1980’s, providing
further support for persistence of these particular genotypes in the environment. Strains OS225
and BA173A (as well as the strains in the cluster J) have very high levels of sequence similarities
(99.9% and 100%, respectively), which implies that the later strains are very likely descendants
of those from the 1980’s. Influence of the isolation conditions had been reported on selection
32

of specific genotypes based on initial genotyping using low-molecular-weight (LMW) RNA
profiling with a larger population of Shewanella baltica strains (31). While comparing
phylogenetic relatedness from the MLST analysis among the selected 46 strains with the
corresponding isolation procedures, this trend was also present yet not as significant (p = 0.1),
which is likely due to the smaller sample size. Strains in the Clade E were mostly isolated using a
nitrate supplemented nutrient broth medium, while strains in Clade I and J, mostly isolated in
1998, were isolated from a modified minimum medium supplemented with nitrate and
thiosulfate. In summary, phylogenetic inference from the MLST analysis displayed strong
correlation with the water geography and temporal distribution of these isolates, and also
showed some correlation with isolation procedures.

Recombination was detected from sequences of the marker genes. Effects of recombination
have been noticed in phylogenetic analysis of other microorganisms (17), and may result in
inaccurate representation of evolutionary relationships of the studied species. Yet this
limitation could partly be overcome by using concatenated sequences from multiple genes,
according to a previous study where significant correlation was found between lengths of
marker genes and the quality of phylogram based upon whole-genome-sequence analysis (18).
A recent study on comparative genomics of four sequenced S. baltica strains revealed
significant levels of homologous recombination (32). Similar processes were thought to occur
among more S. baltica strains, supported by the comparative genomic hybridization profiles.
The recombined gene pool included the conserved core genes as well as ecologically important
genes and operons. Specifically, one of the genes used in the MLST analysis, SO_2706, was
33

shown to be part of the recombined gene pool between OS185 and OS195, suggesting effects
of recombination present in the phylogenetic inference among these strains. Thus, sequences
of all marker genes were subjected to recombination analysis and checked for signals of
recombination [Table 2.4]. As a result, while SO_0578 has the highest recombination to
mutation ratio (0.804), SO_2706 has a much lower recombination to mutation ratio of 0.289.
This low ρ/ θ ratio for SO_2706 may be due to lack of polymorphic sites from the sequences, or
rather, that recombination was not as important in the evolutionary processes among these 46
strains as it was between the strains OS185 and OS195. Nonetheless, despite the detection of
recombination, recombination to mutation ratios of lower than 1 in all the marker genes,
suggested dominance of clonal paths in the S. baltica population. Furthermore, high agreement
between the MLST phylogeny and comparative genomic hybridization (CGH) clustering pattern
(see Chapter 3) also indicates that recombination did not heavily influence the accuracy of the
MLST genotyping method. In conclusion, although it has been reported that
recombinational/sexual rather than clonal propagation dominated evolutionary paths of a few
S. baltca strains; this result does not appear to be the case for the larger S. baltica population
studied here, at least based on sequence information of these seven MLST loci.

34

REFERENCES

35

References
1.

Christophe Fraser EJA, Martin F. Polz, Brian G. Spratt, William P. Hanage (2009) The
bacterial species challenge: making sense of genetic and ecological diversity. Science
323:6.

2.

Cohan FM (2002) What are bacterial species? Annual Review of Microbiology 56(1):457487.

3.

Doolittle WF & Papke RT (2006) Genomics and the bacterial species problem. Genome
Biology 7(9):116.

4.

Cohan FM (2001) Bacterial Species and Speciation. Systematic Biology 50(4):12.

5.

Frost LS, Leplae R, Summers AO, & Toussaint A (2005) Mobile genetic elements: the
agents of open source evolution. Nature Reviews Microbiology 3(9):722-732.

6.

Gogarten JP & Townsend JP (2005) Horizontal gene transfer, genome innovation and
evolution. Nature Reviews Microbiology 3(9):679-687.

7.

Hanage WP, Fraser C, & Spratt BG (2006) The impact of homologous recombination on
the generation of diversity in bacteria. Journal of Theoretical Biology 239(2):210-219.

8.

Fraser C, Hanage WP, & Spratt BG (2007) Recombination and the Nature of Bacterial
Speciation. Science 315(5811):5.

9.

Vos M (2009) Why do bacteria engage in homologous recombination? Trends in
Microbiology 17(6):226-232.

10.

Selander RK, et al. (1986) Methods of Multilocus Enzyme Electrophoresis for bacterial
population genetics and systematics. Applied Environmental Microbiology 51(5):873-884.

11.

Lancefield RC (1933) A serological differentiation of human and other groups of
hemolytic streptococci. The Journal of Experimental Medicine 57(4):571-595.

12.

RK Saiki SS, F Faloona, KB Mullis, GT Horn, HA Erlich and N Arnheim (1985) Enzymatic
amplification of beta-globin genomic sequences and restriction site analysis for
diagnosis of sickle cell anemia. Science 230(4732):1350-1354

13.

Pieter Vos RH, Marjo Bleeker, Martin Reijans, Theo van de Lee, Miranda Hornes, Adrie
Friters, Jerina Pot, Johan Paleman, Martin Kuiper and Marc Zabeau (1995) AFLP: a new
technique for DNA fingerprintin. Nucleic Acids Research 23(21):4407–4414.

14.

J G Williams ARK, K J Livak, J A Rafalski, and S V Tingey (1990) DNA polymorphisms
amplified by arbitrary primers are useful as genetic markers. Nucleic Acids Research
18(22):6531-6535.
36

15.

Cho JC & Tiedje JM (2001) Bacterial Species Determination from DNA-DNA Hybridization
by Using Genome Fragments and DNA Microarrays. Applied and Environmental
Microbiology 67(8):3677-3682.

16.

Martin C. J. Maiden, et al. (1998) Multilocus sequence typing: A portable approach to
the identification of clones within populations of pathogenic microorganisms.
Proceedings of the National Academy of Sciences 95(6):3140-3145.

17.

Walk ST, et al. (2009) Cryptic Lineages of the Genus Escherichia. Applied and
Environmental Microbiology 75(20):6534-6544.

18.

Konstantinidis KT, Ramette A, & Tiedje JM (2006) Toward a More Robust Assessment of
Intraspecies Diversity, Using Fewer Genetic Markers. Applied and Environmental
Microbiology 72(11):7286-7293.

19.

J G Williams ARK, K J Livak, J A Rafalski, and S V Tingey (2004) Bacterial genomic DNA
isolation using CTAB. DOE Joint Genome Institute.

20.

Edgar RC (2004) MUSCLE: multiple sequence alignment with high accuracy and high
throughput. Nucleic Acids Research 32(5):1792-1797.

21.

Tamura K, Dudley J, Nei M, & Kumar S (2007) MEGA4: Molecular Evolutionary Genetics
Analysis (MEGA) Software Version 4.0. Molecular Biology and Evolution 24(8):1596-1599.

22.

Moulton DBaV (2004) NeighborNet: An agglomerative method for the construction of
planar phylogenetic networks. Mol Biol Evol 21:255-265.

23.

Huson DH (2006) Application of Phylogenetic Networks in Evolutionary Studies.
Molecular Biology and Evolution 23(2):254-267.

24.

Librado P & Rozas J (2009) DnaSP v5: a software for comprehensive analysis of DNA
polymorphism data. Bioinformatics 25(11):1451-1452.

25.

Gil McVean PAaPF (2002) A coalescent-based method for detecting and estimating
Recombination from gene sequences. Genetics 160:1231–1241.

26.

Konstantinidis KT & DeLong EF (2008) Genomic patterns of recombination, clonal
divergence and environment in marine microbial populations. The ISME Journal
2(10):1052-1065.

27.

Rybicki DMaE (2000) RDP: detection of recombination amongst aligned sequences.
Bioinformatics 16:562-563.

28.

Kosakovsky Pond SL (2006) Automated Phylogenetic Detection of Recombination Using
a Genetic Algorithm. Molecular Biology and Evolution 23(10):1891-1901.

37

29.

Kosakovsky Pond SL, Posada D, Gravenor MB, Woelk CH, & Frost SDW (2006) GARD: a
genetic algorithm for recombination detection. Bioinformatics 22(24):3096-3098.

30.

Frank Ziemke MGH, Jorge Lalucat, and Ramon Rosselló -Mora (1998) Reclassification of
Shewanella putrefaciens Owen's genomic group II as Shewanella baltica sp. nov.
International Journal of Systematic Bacteriology 48:179–186.

31.

Höfle MG & Brettar I (1996) Genotyping of heterotrophic bacteria from the central Baltic
Sea by use of Low-Molecular-Weight RNA profiles. Applied and Environmental
Microbiology 62(4):1383–1390.

32.

Caro-Quintero A, et al. (2010) Unprecedented levels of horizontal gene transfer among
spatially co-occurring Shewanella bacteria from the Baltic Sea. The ISME Journal 5:131–
114.

38

CHAPTER III
Comparative Genomic Study of the Shewanella baltica Genomes
Using Microarrays

The data from expression arrays were contributed by Jennifer Auchtung, a former post-doc in
the lab.

Part of the work from this chapter has been published in:
Caro-Quintero, A., J. Deng, et al. (2010). "Unprecedented levels of horizontal gene transfer
among spatially co-occurring Shewanella bacteria from the Baltic Sea." The ISME Journal 5:
131–114.

39

Abstract
Comparative Genomic Hybridization (CGH) was employed to probe the genome
plasticity as well as potential evolutionary processes that act on the genomes of Shewanella
baltica strains. Microarrays contained probes from all genes and intergenic regions longer than
200 bp of four sequenced S. baltica genomes. DNA from 46 S. baltica strains was hybridized to
the arrays, through which heterogenic gene distribution in S. baltica strains was observed.
Genes differentially distributed among strains from microaerophilic vs. anoxic conditions, as
well as among strains from the 1980s vs. 1998 were extracted and contrasted to reveal genes
potentially important for redox specialization and changes in gene content over time. Core
genes of strains from several lineages were also analyzed and compared with each other to
reveal genomic differences among lineages. In conjunction with expression arrays of two model
strains (OS185 and OS195) under different redox growth conditions, where genes and operons
of potential functional importance were determined, these CGH profiles allowed us to draw
more confident conclusions about which genes are likely important in specialization of the S.
baltica population. Extensive horizontal gene transfer, reflected through genes linked to mobile
genetic elements, was revealed from CGH profiles. Recombination between OS185 and OS195
lineages was also supported.

40

Introduction
Comparative genomics provide unprecedented opportunities to probe into the plasticity,
dynamics, and evolution of genomes of interest. Through exploring similarities and differences
in genome contents and exploiting relationships of genome structures, one can gain insight into
the signatures of selection and better understand evolutionary processes that act on genomes.
In conjunction with ecological investigations, comparative genomics provides a platform for
studying directional evolution and niche adaptation (1) (2).
Oligonucleotide arrays (oligoarrays) are a sequence-based approach and a powerful tool
for high throughput characterization of many biological systems (3). Through hybridizing target
DNA or cDNA pools to arrays that contain synthesized probes designed from genes of interest,
investigators are allowed to obtain quantitative measurement of tens of thousands of genes at
the same time. Oligoarrays are still under continuing development to lower cost and improve
performance. They have been commonly applied to comparative expression studies (4) (5) (6),
but they also can be used to assay the presence of particular genes in comparative genomic and
metagenomic studies of pure culture and environmental samples (1) (7) (8). The popularity of
this approach in today’s microbial ecological studies is largely due to its relative accessibility,
cost-efficiency, highly parallel analysis capacity and flexibility of design.
Lately, oligoarrays have been employed for genomic comparisons between different
strains in both inter- and intra- species studies (1). Through hybridizing DNA of tester strains
with probes designed from genomic sequences of reference strains, differences in gene
contents can be revealed. Such oligoarrays provide a promising and convenient tool to uncover
gene-content of unknown strains and also are useful in tracking the dynamics of natural
41

populations over time and over environmental fluctuations. In a recent comparative genomic
study, oligoarrays were successfully applied to a collection of over 199 Burkholderia
cenocepacia isolates from both epidemic and environmental origins and were used to evaluate
the distribution of virulence-associated and antibiotic resistance related genes (unpublished
data). Comparative genomic hybridization analysis of environmental and clinical Escherichia coli
strains also provided gene-level evidence of niche adaptation which was further utilized to infer
E. coli pathogenic transmission and expansion in natural loci (9).
The accuracy of results from microarray experiments is subject to multiple sources of
variation including those introduced during the array manufacture processes, sample
preparation, array hybridization, and measurement of spot signal intensity. Thus, normalization
plays an important role in removing these variations. Different normalization methods may
lead to drastically different outcomes of microarray data analysis (3).
I describe here the design and implementation of an oligonucleotide microarray that
targets the pan-genome of the Shewanella baltica species. Through sets of probes designed to
target the genome fragments of four sequenced reference genomes, gene contents of 42
unsequenced strains were determined. A novel normalization method was developed, and the
hybridization specificity and sensitivity of the arrays were also tested. From the comparative
genomic analysis of these Shewanella baltica strains, I sought to answer these questions: (i)
what is the extent of gene diversity in the S. baltica species, and how is this related to their
phylogenetic relationships; (ii) is there a correlation between the particular environment of
these isolates and their gene content and, if so, which genes are potentially important for niche
adaptation, (iii) during the twelve years’ span between time of isolation, how has their gene
42

content changed and (iv) what is the composition of the core genome of the species and how is
it related to the general ecology of the species?

43

Materials and Methods
Array Design. Microarrays were made by MYcroarrays, LLC (Ann Arbor, MI, USA), using a
proprietary light-directed oligonucleotide synthesis method to synthesize 44-48 nucleotide long
probes on glass slides. Probes were designed from the genomic sequences of the four
sequenced S. baltica strains. For the purposes of probe design, those genes that shared 90%
sequence identity over 90% of their length between sequenced genomes were considered
homologous and treated as a single gene, even when represented by more than one sequence.
When homologous gene sequences diverged, probes were designed using the following
sequence preference: OS185> OS195> OS223> OS155 [Figure 3.1] [Table 3.1]. This strategy
resulted in the generation of 30,000 probes, with up to 7 probes designed for each gene, and
up to 2 probes for each unique intergenic region. An additional 720 negative control probes
were included for quality control.

44

Figure3.1. Spheres of the diagram represent the gene content of the four sequenced S. baltica
strains (OS155, OS185, OS195, and OS223). Numbers present in regions of overlap between
spheres indicate the total number of shared orthologous genes (as determined by reciprocal
best-BLAST hits). The numbers of genes unique to each strain are also indicated in the regions
of each sphere that do not overlap. Summary of analysis was done by A. Caro-Quintero and K.
Konstantinidis.

Table 3.1. Number of Probes Designed from Each Reference Strain

OS185
OS195
OS223
OS155
Total

Number of Probes
23192
2459
2112
2236
30000

Number of Genes
4228
642
532
580
5982

DNA microarray hybridization. For DNA-DNA hybridization studies, genomic DNA was
extracted as previously described (10), and sonicated to produce DNA fragments less than 3 kbp

45

in size. DNA samples were labeled with the fluorescent Cy5 dye by incorporation of
amino‐allyl‐dUTP through extension from random primers using E. coli DNA polymerase Klenow
fragment I (Invitrogen, Grand Island, NY), followed by addition of amine‐reactive Cy5.
Microarray slides were pre‐hybridized in buffer containing 0.1% SDS, 5 X SSC buffer (SigmaAldrich Corp., St. Louis, MO) and 1 mg/mL bovine serum albumin (Sigma-Aldrich) at 50 °C for 90
min, and washed with 0.5 X SSC and water. Cy5‐labeled DNA samples were mixed with the
same volume of 2 X hybridization buffer (10xSSC, 0.2% SDS, 0.2 mg/mL herring sperm DNA, and
46% formamide), heated at 95 °C for 5 min and then transferred to 68 °C. Samples were applied
to pre‐hybridized slides, which were then incubated at 50 °C for 18 h before being washed and
scanned using an Axon GenePix 4000B scanner (one‐channel hybridization).

Gene expression studies. Cells were inoculated into 25 mL of trypticase soy broth and
incubated at 22°C with agitation until the cells reached mid‐exponential phase (Optical density
at 600 nm = 0.4‐0.7). Experiments were repeated in triplicate. Cells were pelleted by
centrifugation and resuspended in 350 ml anaerobic HEPES medium (11). After 18‐19 hours of
growth at 22 °C, cells were pelleted anaerobically and resuspended in 10 mL of anaerobic
HEPES medium lacking an electron acceptor. Out of this cell suspension, 2 mL was added to
22.5 ml of HEPES medium containing either 10 mM sodium fumarate, 5 mM sodium nitrate, 10
mM sodium thiosulfate or 10 mM sodium chloride (aerobic culture). All cultures were
incubated at 22 °C. The aerobic cultures were aerated by shaking under air on an orbital shaker
at 150 rpm. RNA was extracted from cell pellets using a Qiagen RNeasy kit following the
optional protocol for better recovery of low molecular weight RNA. RNA from three
46

independent cultures of OS185 and OS195 grown in the presence of oxygen, nitrate, and
thiosulfate was used as experimental samples in hybridization experiments. RNA from all four
strains grown in each condition (oxygen, nitrate, thiosulfate, and fumarate) was used to
construct a reference RNA pool, composed of 45 μg of RNA from each condition for strains
OS185, OS195 and OS223 and 15 μg of RNA from each condition for strain OS155. RNA (10 μg)
from each experimental condition and a parallel aliquot of reference RNA were reverse
transcribed with 9 μg of random primers (Invitrogen). Reactions were incubated at 25 °C for 10
min, 42 °C for 70 min, and 70 °C for 15 min. Remaining RNA was hydrolyzed by adding sodium
hydroxide to 33 mM and incubation at 70 °C for 10 min. Labeled cDNA was purified using a
QiaQuick MinElute PCR purification column following the manufacturer’s protocol with the
exception that the sample was eluted in 12 μl of RNase‐free water (Qiagen, Valencia, CA). For
each hybridization, 10 μl of labeled experimental cDNA was mixed with an equal volume of
labeled reference cDNA and was applied to the oligoarray as described above for DNA‐DNA
studies. Cy3 and Cy5 signals for each array were normalized to the arithmetic mean of ratios for
each array using the GenePix software version 5.1. Features that had fewer than 50% of pixels
with signal more than two standard deviations above background in both Cy5 and Cy3 channels
were excluded from further analysis. Genes showing significantly increased anaerobic gene
expression (nitrate and thiosulfate) relative to aerobic growth were identified using Significance
Analysis of Microarrays (12). All experiments were repeated in triplicates.

Correlation analysis on gyrB, MLST and CGH based distance matrices. Distance matrices of the
gyrB gene and concatenated sequences from MLST analysis were generated by maximum
47

composite likelihood model in Mega4.0. The distance matrix of the CGH profile is based on
Jaccard distances. All matrices were standardized before calculation of the mean squared error.

48

Results and Discussion
Evaluation of probe binding efficiency and specificity. I estimated to what extent DNA-probe
hybridization would occur if sequences were not identical by analyzing the hybridization
profiles of the four reference genomes [Figure 3]. I plotted the normalized signal of all probes in
each CGH profile against the percent identity of their best BLAST hit in the corresponding
genome. Data from all four profiles were combined and displayed by box-plot [Figure 3.2a]. The
intensity of signals decreased with sequence similarity and dropped to very low levels for
sequence similarities lower than 50%. Although there were some high signals from probes with
very low sequence identity, they are likely due to short stretches of high similarity between
probes and genome sequences which cross-hybridized in the absence of a full-length match.
Further statistical analysis showed that the relative number of these probes was very low.

49

Figure 3.2a. Box-Plot of probe signals from CGH profiles of the four reference strains against
percent identity of the probes matched to their genomes.

Array normalization and setting cutoff value for determining gene presence/absence. For
array data processing and normalization, mean signal intensity from the negative control
probes was subtracted from the signal intensity of all spots. Subsequently, the median signal
from the core probes with 100% identical matches in all four genomes was calculated for each
microarray dataset. Based on these calculations, a normalization factor was generated that
would adjust the median signal of the 100% identity core probes to the same value for each
slide. This normalization factor was then applied to all measured spot intensities minus
background. The signal value used for each gene was the median value of normalized signals
from all probes designed for that gene. Statistical approaches were used to determine the cut50

off to evaluate presence/absence of genes based upon CGH profiles. False positive values were
defined as the percentage of probes of less than 40% similarity but with signals of higher than
the proposed cutoff value, and false negative values were defined as the percentage of probes
with more than 80% identity but having signals lower than the cutoff. The best cutoff value
was determined to be 850 units, at which the sum of false positive and false negative
percentages are the lowest (<2%) [Figure 3.2b].

Figure 3.2b. Sum of false positive and false negative rates under different cut-off values. Red
line indicates the final cutoff value chosen. X-axis is the cutoff values; Y-axis shows the sum of
false positive and false negative rates under each cutoff value.

General statistics of CGH. Among the 5635 genes targeted by the 28712 probes on the array,
4075 genes (72.3%) on average showed significant signals when tested against S. baltica
genomes [Figure 3.3]. The genome of S. baltica OS195 had the largest number of genes (4397)
detected, while only 3331 genes were detected in S. baltica BA38. Among all genes represented
51

on the array, 32 were not detected in any strains, all of which targeted one plasmid in OS195,
implying loss of this plasmid during laboratory culturing of the strain. Moreover, 3395 (60.2%)
genes were shared among over 90% of all tested strains and are regarded as the core genome
of the species, while the remaining 2208 genes comprise the variable part of the pangenome
[Figure 3.4]. A slightly larger proportion of variable genes (37%) was from the OS185 genome
based-probes while about equal contributions from the four reference genomes were found in
the pool of variable genes [Table 3.2]. As expected, the number of core genes of the S. baltica
species was 28% larger than the previously estimated size of the core genome (2654 genes) of
the Shewanella genus which was determined from a comparative genomic study of ten
genomes from different species in the Shewanella genus (13).

Table 3.2. Composition of Variable Genes

OS185
OS195
OS223
OS155
Total

# of variable genes
817
453
504
434
2208

52

Figure 3.3. Heatmap of CGH signals from all genes and intergenic regions of the 46 S. baltica strains. Blue color indicates the genes
and intergenic regions with signals lower than the cut-off value. Colors from white to red represent the signal intensity from weak to
high. The color bar on the right of the heatmap represents the depths from which strains were isolated. Hierarchical clustering was
performed both by strains (vertical) and by genes (horizontal).
53

Figure 3.4. The sizes of core genome calculated from the CGH dataset. The average numbers of core genes with increasing numbers
of genomes were plotted. Error bars represent the standard deviation of core genome size when genomes were sampled randomly.

54

GC bias in the core and variable genome. While the average GC content in the four reference S.
baltica genomes range between 46.1% and 46.3%, further investigation into the GC content of
the core and variable genomes revealed much greater differences [Table 3.3] [Figure 3.5].
Genes belonging to the core genome showed a slightly higher GC content (0.48) comparing to
those that vary among genomes (0.43). This may indicate codon usage bias in some variable
genes which likely originated from more distantly related organisms. In fact, GC bias has been
used as one of the indicators for identifying horizontally transferred genomic islands in many
comparative genomic studies. A closer look into the genes with low GC content also shows their
connection to mobile genetic elements including phages and plasmids, demonstrating again the
importance of horizontal gene transfer in the evolution of the S. baltica genomes.

Table 3.3. Comparison of GC Content in Core Genome and Variable Genome
GC%
Mean
Standard
Deviation
Median

All genes
45.8

Core
47.6

Var
43.0

4.7

2.8

5.5

47.0

48.0

43.0

55

Figure 3.5. GC contents of genes that are represented in the microarray. Genes were aligned based on their original order in the
four reference genomes. The upper red line across the figure (y=0.476) marked the average GC content of the core gene pool, while
the lower red line (y=0.430) marked the average GC content of the variable core gene pool.

56

Less conservation of intergenic regions. One thousand eighteen intergenic regions were also
targeted in the microarray. An average of 493 of these intergenic regions showed positive
signals in the 46 tested genomes, while the number of intergenic regions present in each
genome varied from 287 (BA38) to 578 (OS195). Distribution of conservation of intergenic
regions was further analyzed and plotted in [Figure 3.6]. Compared with the CGH signals from
genes, a much lower level of conservation was revealed in intergenic regions, where probed
regions shared by fewer than 50% of strains accounted for a relatively larger portion. A smaller
number of intergenic regions were observed in the core genome of S. baltica strains. Only 252
intergenic regions (24.8%) were shared by greater than 90% of all S. baltica strains, which is
much lower than the 60.2% of core genes that are shared.

57

(a)

(b)
Figure 3.6. Histograms illustrating frequency of presence among 46 S. baltica strains of (a) all
genes and (b) intergenic regions represented in the microarray.
58

Functional bias in the core genome and variable genome. To further probe the functional
differences between core genes and variable genes, Clusters of Orthologous Groups (COG) of
the two gene pools were analyzed [Figure 3.7]. Although most genes in the core genome (85%)
were successfully annotated, around half of the variable genes failed to be assigned to any COG
categories. Through comparing the annotated genes in both gene pools, I observed that among
the seven COG categories that are most enriched in the variable gene pool (L, M, N, K, V, R, U,
see [Table 3.4] for description of COG groups), four categories were associated with cellular
processes and signaling, two were associated with information storage and processing, and the
other one fell into the poorly characterized class. In contrast, among the seven categories most
enriched in the core gene pool (J, E, C, H, O, F, I, P), only one category was associated with
cellular processes and signaling or information storage and processing, while five of the seven
categories were metabolism associated. Thus, there was clearly functional bias in the core and
variable gene pool, which supports the idea that, while the core genome encoded essential
genes required for basic metabolism and were under higher selective pressure, the variable
gene pool tended to be associated with specific life styles of different strains, which could
potentially be linked to the differential niche environments of individual strains.

59

R

S

K

T

E

M

C

L

P

J

O

N

H

G

U

I

V

F

Q

D

Z

Pangenome

A

B

(a)

R

S

K

T

E

M

C

L

P

J

O

N

H

G

U

I

V

F

Q

D

Z

Variable Genes

A

B

(b)
Figure 3.7.

60

Figure 3.7. (Cont’d)

R

S

K

T

E

M

C

L

P

J

O

N

H

G

U

I

V

F

Q

D

Z

Conserved Core Genome

A

B

(c)

Figure 3.7. Distribution of genes in different COG functional categories in (a) pangenome (b)
genes whose presence varies among the 46 S.baltica strains (c) genes that are present in
greater than 90% of the strains. Genes without COG annotation were not included.

61

Table 3.4. Description of COG Categories
Code

Description

Information storage and processing
J
Translation, ribosomal structure and biogenesis
A
RNA processing and modification
K
Transcription
L
Replication, recombination and repair
B
Chromatin structure and dynamics
Cellular processes and signaling
D
Cell cycle control, cell division, chromosome partitioning
Y
Nuclear structure
V
Defense mechanisms
T
Signal transduction mechanisms
M
Cell wall/membrane/envelope biogenesis
N
Cell motility
Z
Cytoskeleton
W
Extracellular structures
U
Intracellular trafficking, secretion, and vesicular transport
O
Posttranslational modification, protein turnover, chaperones
Metabolism
C
G
E
F
H
I
P
Q

Energy production and conversion
Carbohydrate transport and metabolism
Amino acid transport and metabolism
Nucleotide transport and metabolism
Coenzyme transport and metabolism
Lipid transport and metabolism
Inorganic ion transport and metabolism
Secondary metabolites biosynthesis, transport and catabolism

Poorly characterized
R
General function prediction only
S
Function unknown

62

Similar clustering patterns with MLST phylogeny. In order to study the genomic relatedness
among these Shewanella baltica isolates, a dendrogram was constructed based upon the gene
presence/absence matrix from the CGH dataset [Figure 3.8], and was compared with the
sequence based phylogeny from MLST analysis [Figure 2.2] and gyrB gene characterization
[Figure 2.1]. Pairwise mean squared error between distance matrices from the gyrB, MLST, and
the CGH datasets showed higher similarity between the MLST and CGH datasets [Table 3.5].
Specifically, ten out of eleven lineages from the MLST phylogeny were confirmed by the CGH
clustering, suggesting high agreement between these two independent phylogenetic methods.
This observation supported our hypothesis that the rate of gene flow is positively correlated
with the genetic relatedness among strains. The higher agreement observed between the MLST
phylogeny and CGH dendrogram also indicates the advantage of the MLST analysis over the
single gene-based genotyping method.

Table 3.5. Correlation among Three Distance Matrices for Phylogenetic Inference

MLST

gyrB

CGH

MLST

-

-

-

gyrB

0.669

-

-

CGH

0.757

0.647

-

63

Height
0.00

0.05

0.10

0.15

0.20

BA38
BA194
OS690
OS252
OS697
OS110
OS117
OS155
OS183
BA175
BA37
BA173B
OS223
BA185
BA196

BA62
OS106
OS286
OS189
OS250
OS230
OS696

Cluster Dendrogram

d
hclust (*, "average")

OS185
OS638
OS681
OS109
OS631
OS700
OS107
OS288

OS628
OS187
OS625
OS225
OS190
OS195
OS652
OS710
OS645
OS167
OS650
OS641
OS193
OS678
BA170
BA173A

Figure 3.8. Hierarchical clustering of the CGH profile. Gene presence and absence information
extracted from the CGH profile was used for constructing the distance matrix using the Jaccard
index. Hierarchical clustering was performed using the average linkage agglomeration method
in R software version 2.9.2.

64

Genes related to specialization over time. In order to track the changes in the conserved core
genome over time, we analyzed the core genes of strains isolated from the 1980s versus the
strains from 1998 [Figure 3.9a]. Although the 1980’s strains and 1998 strains shared the great
majority of genes, over 300 genes were apparently lost in the 1998 strains while 46 to 67 genes
(computed based on conservation in greater than 90% or 85% of strains, respectively) were
likely fixed over this time, most of which encoded hypothetical proteins or were phage related.
Within the 389 genes specifically shared among 1980s strains, 99 encoded hypothetical
proteins, and among those successfully annotated and better understood, the most enriched
categories were those related to signal transduction (T), transcription (K), amino acid transport
and metabolism (E), energy production and conversion (C) and cell wall biogenesis (M) [Figure
3.10a]. Furthermore, 89 out of these 389 genes were differentially expressed in either OS185 or
OS195 under different redox conditions [Figure 3.10b] [Table 3.6]. Specifically, 24 of these
genes were induced under anaerobic growth condition in both OS185 and OS195 strains; nine
were induced under aerobic condition and eight were induced during nitrate respiration in both
strains. Loss of these genes in 1998 strains could imply a shift in selective pressure over time.
But it needs attention that other factors such as genetic drift or mobile element transfer could
also lead to gene loss during genome evolution.

65

(a)

(b)

Figure 3.9. Comparison of conserved core genes of (a) strains isolated from the 1980s and
1998 (b) strains recovered from at or below the anoxic-oxic interface and from more oxic
water sample. Note that different levels of conservation were used for defining core genes. The
numbers in black are based on greater than 90% conservation while the numbers in red are
based on greater than 85% conservation among 46 S. baltica strains.

66

1980's specific core genes

T

S

R

K

E

C

M

N

P

G

U

H

L

O

J

V

F

Q

I

D

(a)

25

Differential expression of core genes specific to strains from
1980s

Number of genes

20
15
10
5

(b)

AnI-B
NOI-8
AI-B
NI-B
TOR-8
TI-9
AnI-8
TR-9
AI-8
NOR-8
TOI-8
NI-8
TI-B
TOR-9
NR-8
NR-9
TNI-8
TNR-9

0

Figure 3.10. (a) COG functional categories of annotated core genes of strains isolated from
the 1980s. (b) Differential expression patterns of core genes of strains from the 1980s in
OS185 and OS195. Core genes here were defined as genes shared among over 85% strains from
the 1980s or 1998.

67

Table 3.6. Types of Differential Expression
Code
AI
AnI
NI
NR
TI
TR
NOI
NOR
TOI
TOR
TNI
TNR
-8
-9
-B

Description
aerobic induced (higher in O2 relative to nitrate or thiosulfate)
anaerobic induced (higher in nitrate and thiosulfate relative to O2)
nitrate induced (higher in nitrate than in thiosulfate or O2)
nitrate repressed (higher in O2 and thiosulfate relative to nitrate)
thiosulfate induced (higher in thiosulfate relative to O2 and nitrate)
thiosulfate repressed (higher in O2 and nitrate relative to
thiosulfate)
nitrate over oxygen induced (higher in nitrate)
nitrate over oxygen repressed (higher in O2)
thiosulfate over oxygen induced (higher in thiosulfate)
thiosulfate over oxygen repressed (higher in O2)
thiosulfate over nitrate induced (higher in thiosulfate)
thiosulfate over nitrate repressed (higher in nitrate)
Only in OS185
Only in OS195
In both strains

68

Genes related to specialization across the redox gradient. Core genes were extracted from
strains isolated at and below the oxic-anoxic interface (anoxic group) as well as from strains
isolated from above the oxic-anoxic interface (microaerophilic group). Comparing to the core
gene analysis between strains from 1980s and 1998, the microaerophilic and anoxic group
shared a larger number of core genes and possessed fewer group-specific genes [Figure 3.9b].
Given that relatively fewer genes (14 under 90% conservation threshold, or 42 under 85%
threshold) were specifically shared within the microaerophilic group, specialization to the
microaerophilic condition at the level of genomic content was not well supported. In contrast, a
larger number of genes were shared among the anoxic group. Furthermore, when the threshold
of computing core genes was relaxed from 90% to 85%, the number of core genes specific to
the anoxic group increased drastically, probably due to the clustering of strains from anoxic
water from more distantly related groups.
Using the 85% threshold, 179 core genes specific to the anoxic group were revealed,
with 87 of them encoding hypothetical proteins. Among the genes successfully annotated,
many were associated with transcription regulation (K, but mostly mobile genetic elements
related), cell wall biogenesis (M), as well as amino acid transport and metabolism (E) [Figure
3.11a]. It is worth noting that a number of dehydrogenases and oxidoreductase (C), as well as
membrane ion receptors and transporters (P) were also found in the core gene set of these
strains, which suggests they may play essential roles in specialization for the anoxic zone.
However, only a small number (28) of genes in the anoxic set showed differential expression in
either OS185 or OS195. Seven of these 28 genes were up-regulated under aerobic growth
conditions, while only three were induced under anaerobic conditions in both strains [Figure
69

3.11b].

Anoxic group specific core genes
K

R

E

M

L

S

C

T

H

N

P

U

O

F

J

Q

V

(a)

8

Number of genes

7

6
5
4
3
2
1
0

(b)
Figure 3.11. (a) COG functional categories of annotated core genes of strains isolated from
anoxic waters. (b) Differential expression patterns of core genes of strains from anoxic waters
in OS185 and OS195. Core genes here were defined as genes shared among over 85% strains
from the anoxic or microaerophillic group.

70

Conserved gene core in specific lineages.
Based upon the strong correlation with the MLST phylogeny and the CGH clustering
pattern, tracking the conserved gene core of individual lineages was possible and may provide
detailed information on the genetic basis for specialization into different environmental niches
by strains from different clusters. Expression profiling of the strains OS185 and OS195 under
different redox conditions also provided additional support for whether genes differentially
distributed among the S. baltica population were likely functionally important. Pairwise
comparison of the core gene composition of strains in Clade A, D, E and J was performed
[Figure 3.12], given that these four clades potentially represent strains occupying different
niches. Clade A strains were mostly isolated from anoxic water or near the anoxic-oxic interface,
while Clade E originated from more oxic waters. Clade D was thought to have gone through
extensive genetic exchange with strains in Clade A. Furthermore, three out of four strains in
Clade J were isolated from 1998 rather than the 1980s. Thus, investigating genetic differences
in these clades may provide useful information on the impact of environments on microevolution of the S. baltica genomes.

71

Figure 3.12. Core genome comparison of Clade A, D, E and J. Core genes were calculated based
on their presence in more than 85% strains in each clade.

Genes specific to strains in Clade A and D. Among the four clades, A and D exclusively shared a
comparatively large number of genes - 302- compared to any other pairs of clades (from 9
genes exclusively shared between A and J, to 121 shared between E and J). While the majority
of these genes were either mobile element related or encoded hypothetical proteins, a number
of proteins classified in the L (22), K (21), P (15), M (15) COG categories were also revealed. The
genes shared exclusively between Cluster A and D included several membrane transporters
(Shew185_0604-0608, Shew185_3340-3345, Shew185_3346-3347, Shew185_1621, 1623),
transcriptional regulators (Shew185_4192- Shew185_4198), CRISPR-associated proteins
(clustered regularly interspaced short palindromic repeats, Shew185_3236-3241),
aminosynthases and transferases (Shew185_2970-2979, Shew185_2980-2982), as well as an F72

plasmid associated conjugative transfer system (Shew185_4429-4448). Expression of a number
of putative regulators and oxidoreductases (Shew185_1391-1395) was induced during
thiosulfate respiration by both OS185 and OS195 strains, indicating functional importance of
these genes.
Strains in Clade A and Clade D also shared a likely nitrate reductase operon
(Sbal195_3991-4002 in OS195, or Shew185_3866-3877 in OS185, including NosYFDL, NrfDCG,
and periplasmid cytochrome C), which was also found in Clade B. It was shown to be upregulated in S. oneidensis MR-1 under anaerobic growth condition using thiosulfate as electron
acceptor, thus was considered to be related with thiosulfate reduction. In agreement with this
prediction, genes in this operon were induced during thiosulfate respiration for both OS185 and
OS195 strains, and it was also induced by nitrate respiration by OS195.
Furthermore, looking at a broader range of strains, some of the genes shared between
these two clades were exclusively missing from both the OS155 clade (E) and OS183 clade (J),
including a number of genes potentially important for anaerobic respiration such as anaerobic
dehydrogenases (putative nitrate-inducible Se-containing formate dehydrogenase as in S.
oneidensis MR-1, Shew185_0099-0108), NADH oxidoreductase (Shew185_0454-0456), sulfate
transporter associated cytochrome c family proteins (strongly induced by thiosulfate in S. MR-1,
Shew185_0458, Shew185_0460), as well as other proteins involved in signal transduction and
regulation (Shew185_4192-4198). Contrasting to the isolation of, especially, most Clade A
strains from more anoxic waters (120-140 m), strains of the OS155 lineage (Clade E) and OS183
lineage (Clade J) originated from more oxic regions (80-130 m). Together with the genomic

73

signatures revealed from these strains, these two lineages appeared less adapted to anoxic
water environments.

Genes specific to Clade E. Among the 276 genes specific to Cluster E, the biggest majority were
hypothetical proteins or mobile elements related, such as prophage encoded genes and other
insertion segments (Sbal_1270- Sbal_1306, Sbal_4384- Sbal_4411). Other Clade E specific genes
and operons include those encoding proteins involved in phenazine synthesis (Sbal_1937Sbal_1938), polysaccharide biosynthesis (Sbal_2876-2893), as well as a few oxidoreductases
(Sbal_0621-0627) likely involved in sugar metabolism, etc.

Genes specific to Clade A. Although extensive genetic exchange is proposed to have occurred
between the OS185 and OS195 lineages (14), strains in the OS195 cluster (Clade A) possessed
352 specific genes that were not shared with the OS185 lineage (Clade D), and 245 of them
were exclusively found in OS195 lineage compared with more strains in Clade E and J. These
Clade A-exclusive genes include the insertion of a P2 phage (Sbal195_2888-2917), an aerobicrespiration induced polysaccharide synthesis operon (Sbal195_3017-3035), as well as a number
of cytochromes, transcriptional regulators and oxidoreductases that are likely related to
anaerobic respiration. Some of the cytochromes and transcriptional regulators (Sbal195_0256,
0257, 0258), as well as a Type I anaerobic DMSO reductase operon (dmsEFABGH,
Sbal195_2225-2230), were induced under anaerobic growth condition by OS195, indicating
their contribution to specialization of the OS195 lineage in the anoxic environment.
Interestingly, The DMSO reductase operon was flanked by integrase proteins and phage
74

elements, reflecting a potentially important role of marine phages in environmental adaptation
of S. baltica strains.

Genes specific to Clade D. Compared to strains in Clade A, E and J, strains in Clade D specifically
shared 247 core genes, 64 of which were differentially expressed by either strain OS195 or
OS185 under various respiration conditions, indicating that Clade D strains were likely well
adapted to its own redox environment. A polysaccharide synthesis operon (Shew185_28882903) shared exclusively among Clade D strains and OS631 were up-regulated under aerobic
growth in OS185; while a few genes and regulators likely related with anaerobic respiration, i.e.,
an operon potentially related to chemotaxis (Shew185_2544-2551) and cytochrome c proteincontaining operon (Shew185_2534-2538) were found to be up-regulated or at least partially
up-regulated under anaerobic condition in OS185. Genes potentially involved in phenazine
biosynthesis (Shew185_2181-2184), cobyrinic acid a,c-diamide synthase (Shew185_2564-2567)
and galactose metabolism (Shew185_3331-3339) were also found. Particularly, Clade D strains
exclusively shared a number of genes involved in DNA restriction-modification (Shew185_19761981, Shew185_2188-2198), which may be related to specific phage activities.

Core genes in Clade B strains. Strains in Clade B were closely related to those in Clade A based
on both MLST phylogeny and CGH analysis. No Clade B-specific genes were found compared
with those in Clade A, which was probably due to the lack of reference genomes from Clade B
strains for the array design. Compared with core genes revealed in Clade A, 243 genes were
seemingly lost in Clade B, including not only some of the Clade A-specific genes described above,
75

such as the anaerobic DMSO reduction operon, the CRISPR protein operon, polysaccharide
synthesis and the P2 phage insertion, but also a number of membrane transporters and signal
transduction proteins (Shew185_0604-0608, Shew185_4231-4240), as well as a potential
pathogenic island that contained a Type III secretion system (Sbal195_2248-2281).

High versatility in carbon source utilization was reflected from the OS223 genome. While
comparing gene content of single strains, the strain OS223 has the largest number of strainexclusive genes [Figure 3.13], the greatest majority of which encoded either hypothetical
proteins or were phage related. In addition, OS223 possessed a unique polysaccharide
biosynthesis operon (Sbal223_1470-1487), as well as a short-chain dehydrogenase operon
(Sbal223_2173-2177). OS223 also possessed a number of genes and operons related to sugar
metabolism. For example, it shared an operon involving metabolism of multiple sugar
substrates (Sbal223_2138-2162) exclusively with Clade F and BA170, and a potential arabinose
metabolism operon (Sbal223_3608-3611) with a number of strains in other clusters. These
observations were also supported by the high-resolution Biolog profiling of the four reference
strains, where OS223 showed extraordinary versatility in carbon source utilization compared to
the other three tested strains. It has been proposed that although OS223 was isolated from the
same depth as OS185, it might have occupied a slightly different ecological niche in the water
column relative to OS185 or OS195 (14), for example, being associated with sinking particles as
opposed to being planktonic (or vice versa) or being transient or allochthonous at the 120–140
m depth (see also Discussion below). In agreement with the latter hypothesis, only one other
OS223-like strain was recovered in the 1986 or 1987 isolation efforts.
76

Figure 3.13. Number of strain-specific genes revealed from CGH. Strains not shown had no
strain-specific genes according to the threshold used.

Horizontal gene transfer and genomic islands. Compared to higher organisms, in which
genomic material is transferred to offspring exclusively from parental cells through
reproduction, several different mechanisms are known in bacteria that mediate lateral gene
transfer through other individuals including DNA transduction, transformation, and conjugation
(15). One of the most important means of innovation in bacteria is to acquire genetic material
from other cells or organisms, some that are even distantly related, by horizontal gene transfer
(HGT) (16) (17) (18) (19). In aquatic environments, phages or viruses are of special importance
in mediating HGT among different organisms (20) (21) (22). In cases where functionally
associated genes and operons are transferred together and incorporated into the receptor
cell’s DNA, genomic islands can be formed (17). Being either silent or actively transcribed after
incorporation, these horizontally transferred genes are subsequently subjected to a series of
77

processes including genetic drift or natural selection, and those with adaptive or survival
advantages may become fixed through selection sweeps (23).
Comparative genomics has revealed the ecological importance of a number of genomic
islands. One of the most common examples is the prevalence of pathogenicity islands in clinical
isolates, such as the LEE (locus of enterocyte effacement) islands in enteropathogenic
Escherichia coli (EPEC) strains (24), and cci in Burkholderia cenocepacia (25). A number of
genomic islands in Shewanella were identified from a comparative genomic study of ten
Shewanella genomes across different species (13). These genomic islands together with the
insertion elements dominated the differences in gene content among genomes of the same
species, and decreased where more distantly related strains were compared. In a more recent
study that compared the genome sequences of the four reference S. baltica strains (14), mosaic
genomic structures were revealed, and genomic islands potentially associated with ecological
fitness were identified such as those with genes encoding anaerobic respiratory complexes and
associated transport and cytochromes, consistent with the low redox status of their habitat.
CGH effectively characterized patterns of gene acquisition and loss among the larger
collection of S. baltica strains studies here, and further supported the mosaic genome structure
of S. baltica genomes. These genomes were heavily impacted with phages and other mobile
elements related insertions, and the so-formed genomic islands dominated the variable part of
the S. baltica genomes. Therefore, horizontal acquisition of genomic islands was possibly the
major means of harboring new functions by these organisms, while acquired genes were
subsequently subjected to natural selection through which ecologically important genes were
prone to be fixed in the population. Among the genes that were differentially distributed
78

among these S. baltica strains, a significant fraction were potentially ecologically important,
including those associated with anaerobic life style, nutrient transport and metabolism, phage
defense, as well as those involved in signaling and transcriptional regulation, providing further
evidences for the importance of HGT in the micro-evolution of these S. baltica genomes.

Recombination among specific lineages was supported. Recombination plays important roles
during sexual reproduction of eukaryotic organisms, especially in facilitating chromosomal
crossover in the offspring. Recombination can also occur during DNA damage repair in bacteria
(26). Due to homologous recombination, the rate of recombination is largely restricted by
sequence similarity between donor and receptor DNA (27), hence this process occurs mostly
among genetically closely related strains such as within the S. baltica population.
Recombination is thought to play an important role in bacterial genome evolution by
driving species cohesion (28) (29). It has been proposed that in cases where recombination
rates within bacterial lineages are relatively high, convergent evolution is more likely to take
place (30). Thus, analysis of recombination may facilitate understanding of the evolutionary
trajectory of a studied population. In the case of S. baltica, strains OS195 and OS185 have
recently undergone extensive recombination (14). In a previous comparative genomic study of
the four reference strains, 580 non-core genes were found to be shared between OS185 and
OS195, while ~350 of them plus an additional ~10% of their core genes showed 99.5-100%
nucleotide identity, contrasting sharply with -97% identity for the rest of the genes in the
genome and <3% of high identity core genes (99.5-100%) among the remaining pairs of
genomes. In addition, the density of the S. baltica population was estimated to be about 103
79

cells per mL of seawater in both sampling years 1986 and 1987 based on most probable
number estimates using several liquid media (31) . Therefore, the genetic exchange patterns
revealed by the sequenced genomes may apply to a larger collection of strains and may be
persistent over the time (1986–1987) in the natural S. baltica population.
In order to seek support for this hypothesis, CGH hybridization signals on probes
targeting the recombined and non-recombined gene pools from other strains in the same
lineages as the four reference strains were extracted and compared. Strains in the OS195
cluster (n=10) had consistently greater hybridization signal to probes that corresponded to
recombined than non-recombined genes [figure 3.14]. Thus the larger sampling of strains in the
OS195 lineage showed extensive genetic exchange was likely to have occurred between this
lineage and the OS185 lineage. It also appeared that the isolated OS195 strain, which probably
had migrated to deeper waters after the recombination event(s) with the OS185 lineage, had
presumably adapted further to the more anoxic environment of the deeper waters. For
instance, its genome encoded additional genomic islands for anaerobic lifestyle, such as a
dimethyl sulfoxide (DMSO) reductase containing island. Furthermore, OS195-like strains were
more abundant and consistently recovered from this depth in both sampling years suggesting
its adaptation to this redox environment.

80

Figure 3.14. The patterns of genetic exchange apply to a larger collection of S. baltica strains.
Hybridization signal of all strains in the same lineages as the four sequenced strains (left panel)
were analyzed. The average normalized signal of all probes that corresponded to
non‐ recombined OS185 core (red) vs. recombined OS185 core genes with OS195 (blue) are
shown (left panel). Error bars represent one standard deviation from the mean. Note that the
latter probes show consistently greater hybridization signal only in the OS195 like strains, in
agreement with the preferential genetic exchange between the OS185 and OS195 lineages.

81

However, to what extent the patterns of genetic exchange observed between OS195
and OS185 and their sister strains apply to other natural sub-populations of S. baltica in the
Baltic Sea and what accounts for the genetic flow among those specific strains remain unknown.
To address these issues, in situ genomic studies (for example, metagenomics) and sampling of
the natural populations over time will be required.
Although microarrays are sometimes limited in terms of specificity and sensitivity, for
example, being unable to recognize pseudogenes due to frame-shift or single nucleotide
mutations, they proved an efficient, cost effective, sufficiently resolving and highly parallel
method for determination of gene contents and levels of gene expression in the many strains
important to this study. The CGH data provided information on how genes were distributed
among the S. baltica genomes, and the transcriptome revealed which genes were sensitive to
changes in redox conditions. Furthermore, these data partially overcame the problem of lack of
annotation of S. baltica genes, because genes differentially expressed were more likely to be
functionally important under the particular redox conditions, and hence allowed us to better
evaluate genes differentially distributed among lineages and niche environments.

82

REFERENCES

83

References
1.

Willenbrock H, et al. (2006) Design of a Seven-Genome Escherichia coli Microarray for
Comparative Genomic Profiling. Journal of Bacteriology 188(22):7713-7721.

2.

Dobrindt U, et al. (2003) Analysis of Genome Plasticity in Pathogenic and Commensal
Escherichia coli Isolates by Use of DNA Arrays. Journal of Bacteriology 185(6):1831-1840.

3.

Oh S, Yoder-Himes DR, Tiedje J, Park J, & Konstantinidis KT (2010) Evaluating the
Performance of Oligonucleotide Microarrays for Bacterial Strains with Increasing
Genetic Divergence from the Reference Strain. Applied and Environmental Microbiology
76(9):2980-2988.

4.

Otto M, Yoder-Himes DR, Konstantinidis KT, & Tiedje JM (2010) Identification of
Potential Therapeutic Targets for Burkholderia cenocepacia by Comparative
Transcriptomics. PLoS ONE 5(1):e8724.

5.

Murphy D (2002) Gene expression studies using microarrays: principles, problems, and
prospects. Advances in Physiology Education 26(4):256–270.

6.

Beliaev AS, et al. (2005) Global Transcriptome Analysis of Shewanella oneidensis MR-1
Exposed to Different Terminal Electron Acceptors. Journal of Bacteriology 187(20):71387145.

7.

He Z, et al. (2007) GeoChip: a comprehensive microarray for investigating
biogeochemical, ecological and environmental processes. The ISME Journal 1(1):67-77.

8.

Rich VI, Konstantinidis K, & DeLong EF (2008) Design and testing of ‘genome-proxy’
microarrays to profile marine microbial communities. Environmental Microbiology
10(2):506-521.

9.

Chengwei Luo STW, David M. Gordon, Michael Feldgarden, James M. Tiedje, and
Konstantinos T. Konstantinidis (2011) Genome sequencing of environmental Escherichia
coli expands understanding of the ecology and speciation of the model bacterial species.
Proceedings of the National Academy of Sciences 108(17):6.

10.

Richards E, Reichardt M, & Rogers S (1994) Preparation of Genomic DNA from Plant
Tissue. Current Protocols in Molecular Biology eds Ausubel M, Brent R, Kingston RE,
Moore DD, Seidman JG, John A. Smith JA, & Struhl K (John Wiley, Hoboken, NJ), pp 2.3.12.3.7.

11.

Cruz-Garcia C, Murray AE, Klappenbach JA, Stewart V, & Tiedje JM (2007) Respiratory
nitrate ammonification by Shewanella oneidensis MR-1. J Bacteriol 189(2):656-662.

12.

Tusher VG, Tibshirani R, & Chu G (2001) Significance analysis of microarrays applied to
the ionizing radiation response. Proc Natl Acad Sci U S A 98(9):5116-5121.
84

13.

Konstantinidis KT, et al. (2009) Comparative systems biology across an evolutionary
gradient within the Shewanella genus. Proceedings of the National Academy of Sciences
106(37):15909-15914.

14.

Caro-Quintero A, et al. (2010) Unprecedented levels of horizontal gene transfer among
spatially co-occurring Shewanella bacteria from the Baltic Sea. The ISME Journal 5:131–
114.

15.

Wozniak RAF & Waldor MK (2010) Integrative and conjugative elements: mosaic mobile
genetic elements enabling dynamic lateral gene flow. Nature Reviews Microbiology
8(8):552-563.

16.

Thomas CM & Nielsen KM (2005) Mechanisms of, and Barriers to, Horizontal Gene
Transfer between Bacteria. Nature Reviews Microbiology 3(9):711-721.

17.

Dobrindt U, Hochhut B, Hentschel U, & Hacker J (2004) Genomic islands in pathogenic
and environmental microorganisms. Nature Reviews Microbiology 2(5):414-424.

18.

Frost LS, Leplae R, Summers AO, & Toussaint A (2005) Mobile genetic elements: the
agents of open source evolution. Nature Reviews Microbiology 3(9):722-732.

19.

Gogarten JP & Townsend JP (2005) Horizontal gene transfer, genome innovation and
evolution. Nature Reviews Microbiology 3(9):679-687.

20.

Weinbauer MG & Rassoulzadegan F (2003) Are viruses driving microbial diversification
and diversity? Environmental Microbiology 6(1):1-11.

21.

Paul JH (2008) Prophages in marine bacteria: dangerous molecular time bombs or the
key to survival in the seas? The ISME Journal 2(6):579-589.

22.

Corina PD Brussaard SWW, Frede Thingstad, Markus G Weinbauer, Gunnar Bratbak,
Mikal Heldal, Susan A Kimmance, Mathias Middelboe, Keizo Nagasaki, John H Paul,
Declan C Schroeder, Curtis A Suttle, Dolors Vaque and K Eric Wommack (2008) Globalscale processes with a nanoscale drive: the role of marine viruses. International Society
for Microbial Ecology 2:575–578.

23.

Cohan FM (2001) Bacterial Species and Speciation. Syst. Biol. 50(4):12.

24.

Nisan I WC, Hanski E, Rosenshine I. (1998) Interaction of enteropathogenic Escherichia
coli with host epithelial cells. Folia Microbiol (Praha) 43(3):247-252.

25.

Malott RJ, Baldwin A, Mahenthiralingam E, & Sokol PA (2005) Characterization of the
cciIR Quorum-Sensing System in Burkholderia cenocepacia. Infection and Immunity
73(8):4982-4992.

85

26.

SC Kowalczykowski DD, AK Eggleston, SD Lauder, WM Rehrauer (1994) Biochemistry of
homologous recombination in Escherichia coli. Microbiology and Molecular Biology
Reviews 58(3):401–465.

27.

Fraser C, Hanage WP, & Spratt BG (2007) Recombination and the Nature of Bacterial
Speciation. Science 315(5811):476-480.

28.

Vos M (2009) Why do bacteria engage in homologous recombination? Trends in
Microbiology 17(6):226-232.

29.

Brown EW, LeClerc JE, Kotewicz ML, & Cebula TA (2001) Three R's of bacterial evolution:
How replication, repair, and recombination frame the origin of species. Environmental
and Molecular Mutagenesis 38(2-3):248-260.

30.

Hanage WP, Fraser C, & Spratt BG (2006) The impact of homologous recombination on
the generation of diversity in bacteria. Journal of Theoretical Biology 239(2):210-219.

31.

Ziemke F, Brettar I, & Höfle MG (1997) Stability and diversity of the genetic structure of
a Shewanella putrefaciens population in the water column of the central Baltic. Aquatic
Microbial Ecology 13:63-74.

86

CHAPTER IV
Characterization of Carbon Source Utilization and Respiratory
Capabilities of the Shewanella baltica Strains

The Biolog dataset was a contribution from Ingrid Brettar.

87

Abstract
Shewanellae are known for their remarkable versatility in coupling organic matter
turnover with a wide range of electron acceptors. In this chapter, we provide a high throughput
approach to effectively screen the metabolism of 95 carbon sources by 46 S. baltica isolates
obtained from different depths of the Baltic Sea over a 20 year period using the Biolog
phenotypic microarray system. Positive phenotypes were obtained for 76 substrates, with 11
being metabolized by all isolates. The carbon source metabolism patterns correlated with
genetic relatedness, and were ecologically associated with the time of isolation, i.e. strains
isolated in 1998 were able to utilize significantly more carbon sources compared to those
isolated in the 1980’s. Differential respiratory capabilities of several potential alternative
electron acceptors were observed in strains isolated from zones in the water column of
different redox status. Strains capable of reducing dimethylsulfoxide (DMSO) were mostly from
anoxic water, while the inability to reduce thiosulfate and trimethylamine-N-oxide (TMAO) was
limited to a cluster where most strains were from more oxic water. Furthermore, although
these strains were characterized as respiratory denitrifiers, denitrification activity was not
observed in any isolate; but nitrate reduction via ammonification occurred instead. Our results
indicate that phenotypic evaluation of the physiology of populations can provide a better
understanding on the ecological framework of a species.

88

Introduction
Bacteria of the Shewanella genus are facultative anaerobic organisms that are known
for their versatility in coupling decomposition of organic matter with reduction of a variety of
terminal electron acceptors. Species in this genus were reported to be able to utilize organic as
well as inorganic electron acceptors such as nitrate, thiosulfate, uranium [U(VI)], chromium
[Cr(VI)], technetium, neptunium, selenite, and nitroaromatic compounds(1) (2). Thus,
Shewanellae are of great potential for bioremediation of many environmental contaminants
and have also been used in a number of microbial fuel cell studies (3). Shewanellae are
commonly found in complex ecosystems such as fresh and marine water and sediments,
especially where there is chemical stratification of redox-controlling chemicals. Due to their
respiratory versatility, they play important roles in turnover of organic matter during redox
fluctuations and near oxic-anoxic interfaces in aquatic environments where oxygen is deficient
and alternative electron acceptors become available (4). Owing to their diverse metabolic
capabilities, Shewanellae are important players in the biogeochemical cycles of carbon,
nitrogen, and sulfur.
The remarkable respiratory versatility of Shewanellae is reflected in the genomes of
sequenced Shewanella strains, which are generally rich in cytochromes, reductases, quinones,
and iron-sulfur proteins (1) (5). As revealed by the model strain Shewanella oneidensis MR-1, a
highly diverse electron-transport system was predicted given a total of 42 putative c-type
cytochromes, 211 one-component signal transduction systems and 47 two-component systems,
as well as a number of novel transcription regulatory systems. System-level studies of multiple

89

Shewanella spp. strains further revealed a diverse range of metabolic processes in signal
sending, transduction and regulation.
The Shewanella baltica strains in this study were isolated from nutrient enriched Baltic
Sea, especially with overloads of nitrogen (6). Vigorous denitrification activity was observed
from the oxic-anoxic interface based on in vitro growth experiments, which was considered the
major factor counteracting eutrophication (7). Methods targeting denitrifying bacteria were
used to isolate bacteria possibly responsible for this process, where the S. baltica was found to
represent the major fraction of culturable denitrifying bacteria (8) (9). The depths of isolation
spanned the chemocline of oxygen, nitrate and hydrogen sulfide in the water column, and
agreed with the previous notion that Shewanellae were favored in complex redox aquatic
environments. Thus, we examined the utilization of several potential electron acceptors
relevant to the environment in the Baltic Sea, to look for correlation between electronaccepting capability and niche differentiation among these S. baltica strains.
In addition to the electron accepting abilities, capabilities in carbon resource utilization
is also essential for thriving in complex environments. Shewanella spp. are among the bacteria
commonly found during food spoilage, especially anaerobic spoilage of seafood. They were
shown to be able to use a range of carbon sources including various amino acids and peptides,
fatty acids, and even nucleotides, although their ability in degrading organic acids was mostly
limited to C1-3 compounds. Phenotypic characterization of carbon source utilization has been
used to distinguish closely related Shewanella spp. (10) (8), and was shown to be correlated to
ecological conditions. Moreover,Shewanella spp. were often considered as ‘opportunists’ in the
environment, which typically use the r-strategy in unpredictable environments. As the ability to
90

reproduce quickly during nutrition rich periods is crucial for r-strategists, the metabolic capacity
of S. baltica is probably essential for their survival in fluctuating aquatic environments.
Biolog Phenotype Microarrays (PMs) provide a high throughput phenotyping system
with the capacity to effectively measure large numbers of phenotypes at once. It incorporates
redox chemistry into growth assays, using cell respiration as a universal reporter for metabolic
activity. Tetrazolium reduction is coupled with cells’ respiration activity, reflected through color
change, which can be monitored and quantified to reveal the extent of respiration. This system
has been applied to examine metabolic differentiation of bacteria. In a recent study of several
Shewanella strains of different species, comparative Biolog analysis has been shown to reflect
niche differentiation among the strains (11). Furthermore, with the aid of genomic information,
comparative Biolog analysis also provided insights into the functional clarification of ambiguous
genes, as well as important evaluation for genome predictions. In this study, I employed the
Biolog PM system to screen the utilization of 95 carbon sources by 46 S. baltica strains. Through
investigating phenotypic traits of these strains, we expect discovery of environmental
signatures of the S. baltica species, and further understanding of the outcome of selection on
this population.

91

Materials and Methods
Bacterial strains and growth conditions. LB medium and modified M1 medium (12) were used
for bacterial growth. Anaerobic M1 medium was made under oxygen free argon gas. Sodium
nitrate (2 mM) was supplied for testing denitrification. Durham tubes were inserted into the
serum tubes for gas collection. Nitrate levels were tested after 3 days of incubation at 22°C
using nitrate test strips (Merckoquant, Darmstadt, Germany, limit of detection >160 mM).
Thiosulfate reduction was tested in Triple Sugar Iron agar (TSI agar, Acumedia, Lansing, MI).
Strains were inoculated into the bottom of test tubes containing TSI agar and incubated at 22 .
Three replicates were made for each strain. Strains were classified as incapable of reducing
thiosulfate into hydrogen sulfide (H2S) if the agar did not turn black within a week.
Dimethylsulfoxide (DMSO) and trimethylamine-N-oxide (TMAO) reduction was tested in
anaerobic M1 medium supplemented with 10 mM of DMSO and TMAO as electron acceptors,
respectively. OD600 was monitored for 3 days. Strains were characterized as not capable of
reducing DMSO or TMAO if the OD600 did not exceed 0.02 at the end of incubation. Shewanella
oneidensis MR-1 was used as the positive control.

Regression analysis on Biolog datasets. Pair-wise similarities among isolates were calculated
for the gene presence and absence matrix from the CGH profile, the Biolog dataset, as well as
the environmental parameter profile. Regression analysis was performed on Biolog similarities
with CGH similarities to estimate correlation between metabolic capability and evolutionary
relatedness. A secondary regression was then performed to estimate the effects from

92

environmental factors using the phylogenetic independent residuals against similarities from
environmental profiles, which eliminated the phylogenetic influence on metabolic capabilities.

93

Results
Examining diversity of carbon source utilization using the Biolog system. Use of 95 different
carbon sources by the S. baltica strains was screened using the Biolog-GN system. Metabolic
activities were measured and assays were scored as either growth or no growth [Figure 4.1].
The S.baltica strains were able to utilize an average of 36 compounds under aerobic condition.
Among the carbon sources tested, 19 were not used by any S.baltica strain, while the following
11 substrates were used by all strains: cis-aconitic acid, sucrose, D-gluconic acid, L-glutamic
acid, dextrin, maltose, α-D-glucose, L-serine, N-acetyl-D-glucosamine, lactic acid, and inosine.
These composed the metabolic core of the S. baltica population. Interestingly, compared to S.
oneidensis MR-1, whose carbon source utilization capability was mostly limited to 3 carbon
compounds (13), these S. baltica strains were capable of using compounds containing 5 or 6
carbons. A principle coordinate analysis of these Biolog profiles [Figure 4.2] showed that all the
1998 strains clustered together, while, strains isolated from 1986 and 1987 were not well
separated in this visualization method. However, among those from the 1980s, strains in Clade
E [Figure 2.2] appeared to share a similar pattern of Biolog profile, while clustering of Clade A
strains was not apparent.

94

Figure 4.1. Heatmap of 46 S. baltica strains based on Biolog profiles. Blue color indicates no metabolic activity detected with
respect to utilization of the tested substrates; red color indicates significant level of metabolic activity and white indicates weak
metabolic activity. Hierarchical clustering was performed both by strains (vertical) and substrates (horizontal).

95

0.2

1998

PCoA_V2

0.1

1986
0

1987
-0.1

Cluster E

-0.2
-0.3

-0.2

-0.1

-3E-16

0.1

0.2

0.3

0.4

PCoA_V1
Figure 4.2. Principle Coordinate Analysis (PCoA) of Biolog metabolic profiles of the S. baltica
strains. Distance matrix is calculated using Jaccard distances. Strains isolated from different
years are marked with different labels.

96

Examining diversity in utilization of different electron acceptors. Although these S. baltica
strains were initially considered as potential in situ denitrifiers in the Baltic Sea, formation of
nitrogen gas during nitrate respiration was not detected from any of these strains. However, all
of them were able to grow with nitrate as the sole electron acceptor and decrease the
concentration of nitrate. Out of the 46 strains tested, 13 of them (all from 1986 and 1987) were
able to grow with DMSO as the only electron acceptor [Table 4.1]. Thiosulfate and TMAO can
be used as electron acceptors by most of these isolates, except for the strains contained in
Clade E, implying the possibilities of mutation or gene loss in their recent common ancestor.

Table 4.1. Use of Different Electron Acceptors by S. baltica Strains
Strain
MR1
OS106
OS107
OS109
OS110
OS117
OS155
OS167
OS183
OS185
OS187
OS189
OS190
OS193
OS195
OS223
OS225
OS230
OS250
OS252
OS286
OS288

Nitrate
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+

Thiosulfate
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
97

DMSO
+
+
+
+
+
+
+
-

TMAO
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+

Table 4.1. (Cont’d)
OS625
OS628
OS631
OS638
OS641
OS645
OS650
OS652
OS678
OS681
OS690
OS696
OS697
OS700
OS710
BA37
BA38
BA62
BA170
BA173A
BA173B
BA175
BA185
BA194
BA196

+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+

+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+

98

+
+
+
+
+
+
+
-

+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+

Discussion
Metabolic similarity in carbon source utilization is correlated with evolutionary relatedness.
One of the important goals in comparative studies of genomic and phenotypic diversity among
bacterial species is to discern relationships between the genetic content and phenotypic
capability [Figure 4.3]. We found a strong and positive correlation (P < 0.001; b = 0.3520)
between these two parameters for the 46 S. baltica strains based on the Biolog dataset and
their genomic relatedness from CGH. Thus, the signals from these rather coarse level
methodologies were strong enough to detect what must be underlying genetic-phenotype
relationships. This correlation can also be visualized through comparison of the hierarchical
clustering structures of the CGH and the Biolog profiles [Figure 4.4], where six clades were at
least partially shared between the two dendrograms. It is interesting to note that, Clade A in
the CGH (or MLST) dendrogram can be matched to two clusters, A1’ and A2’, from the Biolog
dendrogram. While A1’ contains only strains isolated from 1986, A2’ includes only the strains
isolated from 1987, which have metabolic patterns that are more similar to those revealed
from the 1998 isolates. Thus, the Biolog profiling seems to have provided further discrimination
among these strains beyond the resolution of MLST genotyping. Meanwhile, metabolic patterns
of specific clusters can be inferred. For instance, strains in Clade E used a significantly fewer
number of carbon sources compared strains in other clusters. In particular, a number of 4
carbon acids such as succinate acid, aspartic acid, and α-ketobutyric acid failed to stimulate
growth of any strains in this cluster, implying the lack of certain catalytic enzymes involved in
the pathways. Moreover, metabolic profiling of strains in Cluster A2’, comparing to those in
Cluster A1’, differs mainly at utilization of an additional eight carbon sources (D-psicose, mono99

methyl succinate, alaninamide, α-ketovaleric acid, glucose-1-phosphate, D-fructose, bromosuccinic acid and propionic acid, with p<0.1), possibly due to selective sweeps that favored
capability in utilizing these compounds in the Baltic water environment.

Figure 4.3. Regression of Biolog similarity and genomic relatedness. Pair wise genomic
relatedness was calculated from gene presence and absence matrix from the CGH profile using
formula
divided by length of the vector. Pair wise Biolog similarity was calculated
using the same formula.

100

Figure 4.4. Hierarchical clustering of the Biolog profile. Pair wise Biolog dissimilarities were
calculated using the Jaccard index. Hierarchical clustering was performed using the average
linkage agglomeration method.
101

Temporal distribution has a significant impact on evolution of metabolic capabilities by S.
baltica strains. After eliminating phylogenetic influences from the metabolic profile, the
phylogenetic independent residuals were used to examine the effects from other isolation and
environment related factors. These factors include the sites and time of isolation, culturing
methods, as well as water chemistry parameters including concentrations of nitrate, oxygen,
hydrogen sulfide, etc. Significant correlation was observed only between these residuals with
the time of isolation (p < 0.01), i.e., effects of other environmental variables are either not
significant on the genetic or phenotypic diversification, or are through influencing the
phylogenetic relatedness among these isolates.

Table 4.2. Two Sample T Test Shows that Strains from 1998 Use Significantly More Carbon
Sources
All

1980s

1980s from MM*

1998 all from MM

Average number of
substrates used

35.7

33.5

38.6

43.6

P-value from t-test
against 1998 strains

0.00104

9.07e-05

0.0230

-

* MM: Minimal Medium

A closer look at the Biolog profile revealed higher metabolic versatility in carbon source
utilization among strains isolated in the more recent year. S. baltica strains isolated from 1998
grew on 44 compounds while those isolated from the 1980s grew on only 34 compounds.
Although all the S. baltica strains from 1998 were recovered in minimal medium, the t-test

102

performed between S. baltica strains from 1998 and those isolated from 1980s also using
minimal medium confirmed a significant difference in the number of carbon sources used
[Table 4.2]. In particular, capability of utilizing certain carbon sources including D-fructose,
bromo-succinic acid, α -ketobutyric acid, glucose-1-phosphate, acetic acid, α-hydroxybutyric
acid, glucose-6-phospate and L-alanine was found in a significantly larger percentage of strains
isolated in 1998 comparing to those isolated in the 1980s (p < 0.001), implying a shift in
selective pressure in the Baltic Sea environment.

Respiratory versatility within the S. baltica species is correlated with their genotypic
relatedness. Shewanella species are among the most anaerobically versatile microorganisms
and have been reported to be able to use several organic and inorganic compounds as electron
acceptors (1). In particular, Shewanella baltica population composed the major fraction of
culturable denitrifiers during the time of isolation, suggesting the importance of their roles in
marine nitrogen cycling. In the other studies, Shewanella baltica was identified as one of the
most important H2S producing bacteria during iced storage of fish (14), implying its ability to
couple organic turnover with reduction of sulfur compounds. Furthermore, in addition to
nitrate, the water chemistry profile exhibited a gradient of hydrogen sulfide along the water
column, which accumulates from the sediment and diffuses across the anaerobic deep water to
the oxic-anoxic interface. We hypothesized that, other than nitrate, respiration of certain sulfur
compounds might as well be involved in specialization of these S. baltica strains. Thus, other
than nitrate, we tested utilization of several sulfur compounds as electron acceptors, including
thiosulfate and DMSO. We also assayed these S. baltica strains for their abilities to reduce
103

TMAO, which, in addition to H2S production, was regarded as a prominent characteristic of fish
spoiling bacteria (15) (16). Respiratory capabilities of DMSO, thiosulfate and TMAO are strongly
correlated with specific genotypes (p < 0.01). Meanwhile, when these respiratory capabilities
were compared against the presence and absence of corresponding genes and pathways from
the CGH profiles, high agreement was revealed. These results further established the
relationship between genetic background and phenotypic outcome among the S. baltica strains.

Reduction of nitrate likely via ammonification is a common phenotypic characteristic among
all S. baltica strains. Although these S. baltica strains were initially isolated by methods
targeting denitrifiers from the Baltic Sea (8) (9), denitrification genes were not found in the
sequenced S. baltica genomes. Instead, genes that participate in dissimilatory nitrate reduction
to ammonium (DNRA) (nap operon and nrfA gene) were found. Thus, our results supported the
prediction that these S. baltica strains cannot perform nitrate reduction through complete
denitrification to nitrogen gas and indicate that the respiratory nitrate ammonification
mechanisms may be involved instead.

DMSO reduction may be related to specialization to anoxic deep water environment. It has
been proposed that cooling effects caused by dimethylsulfide (DMS), a primary marine biogenic
sulfur compound, may help to counteract human-induced global warming (17) (18).
Dimethylsulfoxide (DMSO), one of the precursors of DMS, is naturally formed from the
degradation of phytoplankton in marine environments and can be used as an alternative
electron acceptor for many bacterial species where oxygen is deficient. Strains capable of
104

reducing DMSO are limited in Clade A, G and K. Capabilities of reducing DMSO by these S.
baltica strains were consistent with gene presence/absence profiles from microarray
experiments, with the exception of strain OS645, which, though not capable of reducing DMSO,
possesses all genes in the Type I DMSO reductase operon (dmsEFABGH) according to the CGH
profile. The reason for loss in expression of this operon in OS645 is not clear. Furthermore, the
mRNA expression level of this operon in strain OS195 is up-regulated under conditions of both
nitrate and thiosulfate respiration. As half (seven) of the strains capable of reducing DMSO
were isolated from at and below the anoxic-oxic interface, where both oxygen and nitrate are
almost non-detectable and alternative electron acceptors are needed for survival, we predict
that the ability of reducing DSMO may be correlated with levels of available electron acceptors
in the water environment and is thus potentially important for adaption to the anaerobic water
environment.

Inability to reduce thiosulfate is limited within Clade E. As a major product of anoxic sulfide
oxidation, thiosulfate is a key intermediate in the dissimilatory sulfur cycle of aquatic
environments (19). Since H2S, often produced during the reduction of thiosulfate, is only
detectable at and below the oxic-anoxic surface, we postulate that the ability to reduce
thiosulfate into H2S may be one of the factors causing specialization of S. baltica strains. While
compared with the MLST phylogeny, all six strains incapable of thiosulfate reduction are from
the Clade E, indicating that they may have evolved through similar evolutionary paths.
Interestingly, although these 6 strains came from different zones within the redox gradient

105

(from 80 m to the oxic-anoxic interface at 130 m), none were isolated from the anoxic water
zone, suggesting that the inability to reduce thiosulfate may select against growth in the anoxic
water.

Inability to reduce TMAO is also limited within Clade E. TMAO is a nitrogenous
osmolyte broadly found in marine organisms (20). Reduction of TMAO results in the production
of trimethylamine (TMA), which has been implicated as a major contributor to spoilage of
seafood and fish (15) (16). TMAO is also a precursor to a variety of reduced nitrogenous
biogases that are important intermediates of the marine biogeochemical cycle of nitrogen and
potential regulators of atmospheric pH (21). In previous studies, Shewanella baltica has been
identified as the major bacterial species involved in fish spoilage during iced storage (14). Our
experiments showed that all the S. baltica strains except those in the Clade E are capable of
respiring TMAO, although the CGH profiles suggest that TorECAD (22), the genes encoding
TMAO reductases (Shew185_3211-Shew185_3214), were present in all S. baltica strains.
Further comparison of genomic sequences reveal that in strain OS155, TorA (Sbal_3213),
encoding a periplasmic protein connected to the quinone pool via the cytochrome TorC,
appears to be a pseudo gene due to a shift of the open reading frame resulting from a single
nucleotide deletion at the 1243

th

position of the gene. No matter whether incapability of

reducing TMAO by the other strains in this cluster is or is not attributed to a similar mechanism,
we do see that a simple change either in regulation or structural integrity can result in nonactivity of the entire operon.

106

REFERENCES

107

References
1.

Fredrickson JK, et al. (2008) Towards environmental systems biology of Shewanella.
Nature Reviews Microbiology 6(8):592-603.

2.

Scott KHNaJ (2006) Ecophysiology of the genus Shewanella. Prokaryotes 6:19.

3.

Tiedje JM (2002) Shewanella - the environmentally versatile genome. Nature
Biotechnology 20:2.

4.

Hau HH & Gralnick JA (2007) Ecology and Biotechnology of the GenusShewanella.
Annual Review of Microbiology 61(1):237-258.

5.

Heidelberg JF, et al. (2002) Genome sequence of the dissimilatory metal ion–reducing
bacterium Shewanella oneidensis. Nature Biotechnology 20(11):1118-1123.

6.

Elmgren R (1989) Man's Impact on the Ecosystem of the Baltic Sea: Energy Flows Today
and at the Turn of the Century. AMBIO:326-332.

7.

Brettar I & Rheinheimer G (1991) Denitrification in the Central Baltic: evidence for H2Soxidation as motor of denitrification at the oxic-anoxic interface. Marine Ecology
Progress Series 77:157-169.

8.

Brettar I (2002) Shewanella denitrificans sp. nov., a vigorously denitrifying bacterium
isolated from the oxic--anoxic interface of the Gotland Deep in the central Baltic Sea.
International Journal of Systematic and Evolutionary Microbiology 52(6):2211-2217.

9.

Brettar I, Moore ERB, & Höfle MG (2001) Phylogeny and Abundance of Novel
Denitrifying Bacteria Isolated from the Water Column of the Central Baltic Sea. Microbial
Ecology 42(3):295-305.

10.

Stenström IM & G M (1990) Classification of the spoilage flora of fish, with special
reference to Shewanella putrefaciens. Journal of Appplied Bacteriology 68(6):601-618.

11.

Rodrigues JLM, Serres MH, & Tiedje JM (2011) Large-Scale Comparative Phenotypic and
Genomic Analyses Reveal Ecological Preferences of Shewanella Species and Identify
Metabolic Pathways Conserved at the Genus Level. Applied and Environmental
Microbiology 77(15):5352-5360.

12.

Cruz-Garcia C, Murray AE, Klappenbach JA, Stewart V, & Tiedje JM (2006) Respiratory
Nitrate Ammonification by Shewanella oneidensis MR-1. Journal of Bacteriology
189(2):656-662.

13.

Serres MH & Riley M (2006) Genomic Analysis of Carbon Source Metabolism of
Shewanella oneidensis MR-1: Predictions versus Experiments. Journal of Bacteriology
188(13):4601-4609.
108

14.

Vogel BF, Venkateswaran K, Satomi M, & Gram L (2005) Identification of Shewanella
baltica as the Most Important H2S-Producing Species during Iced Storage of Danish
Marine Fish. Applied and Environmental Microbiology 71(11):6689-6697.

15.

Barrett EL (1985) Bacterial Reduction of Trimethylamine Oxide. Annual Review of
Microbiology 39:19.

16.

Gram L & Huss HH (1996) Microbiological spoilage of fish and fish products.
International Journal of Food Microbiology 33:121-137.

17.

Vallina SM & Simo R (2007) Strong Relationship Between DMS and the Solar Radiation
Dose over the Global Surface Ocean. Science 315(5811):506-508.

18.

Vallina SM, Simo R, & Manizza M (2007) Weak response of oceanic dimethylsulfide to
upper mixing shoaling induced by global warming. Proceedings of the National Academy
of Sciences 104(41):16004-16009.

19.

Stefan M. Sievert RPK, and Heide N. Schulz-Vogt (2007) The Sulfur Cycle. Oceanography
20:7.

20.

Walsh BASaPJ (2002) Trimethylamine oxide accumulation in marine animals relationship
to acylglycerol storage. The Journal of Experimental Biology 205:297–306.

21.

AD H & SW G (1999) A technique for the determination of trimethylamine-N-oxide in
natural waters and biological media. Analytical Chemistry 71(21):4886-4891.

22.

Jean-Philippe Dos Santosa CI-N, Carole Couillaulta, Gérard Giordanoa and Vincent
Méjean (1998) Molecular analysis of the trimethylamine N-oxide (TMAO) reductase
respiratory system from a Shewanella species. Journal of Molecular Biology 284(2):421433.

109

CHAPTER V
Competition Assays among Selected Shewanella baltica Strains while
Exposed to Different Electron Acceptors

110

Abstract
A key strategy to understand the ecology of organisms is through determining how they
respond to their environment by altering gene expression. In order to determine redox
specialization of the Shewanella baltica strains, competition assays were performed for
identification of competitive advantages in conditions mimicking their own niches, from which
potential specialization to waters containing high nitrate level was found. Two model strains, S.
baltica OS185 and OS195, with full genomes sequenced, sharing high genomic similarity, but
isolated from depths with distinct nitrate levels, were chosen to further investigate
transcriptional response to the nitrate respiring condition. The two competed genotypes were
separated by flow cytometry and their transcriptomes analyzed by Illumina RNA-Seq.
Regulatory differences were found in their responses both to the nitrate condition and to the
competitor strain. Analysis of differentially expressed genes due to presence of the other
competitor strain revealed that OS185 appeared to have induced the expression of phage
components in OS195, while OS185 itself actively responded to the high nitrate condition by
up-regulation of redox related genes. Similarly, analysis of differentially expressed genes due to
the nitrate respiring condition also suggested higher responses to nitrate condition in OS185
compared to that in OS195 through higher expression of redox related genes in OS185. These
results collectively provide explanations for the competition outcome between OS185 and
OS195, and suggested that OS185 is better adapted to nitrate-rich conditions.

111

Introduction
In the realm of microbial ecology, it is equally important to study the microbe itself as
well as its connections with the environment. The former task was traditionally realized
through various cultivation methods followed by genetic and phenotypic characterization, and
in this era, using high throughput metagenomic sequencing to finely probe the genetic
composition of the entire community. The latter task is, however, of much higher complexity,
and demands a thorough understanding of the environment itself, the biotic activities of
microorganisms, their interactions with each other, as well as the dynamics of interactions
between microbes and the abiotic environments.
Numerous efforts have been made to highlight the importance of the environmental
components associated with each microorganism(1) (2) (3), including the proposal of ‘ecotype’
theory, which incorporates niche properties into the framework of a microbial species
definition (4) (5). As proposed in the ecotype theory, selection sweep is the key component in
formation of ecotypes, as well as the major force for species cohesion. Microbial specialization
is regarded as the processes of formation of various ecotypes within a population, which
facilitates resource partitioning in the ecosystems (6) (7). Thus, differential fitness in different
‘micro’- niches is often a direct result of specialization.
Specialization of microorganisms has been observed in many ecosystems (8) (9) (10),
among which the most extensively studied cases are those of pathogenic bacteria (11) (12) (13).
However, even in many well studied systems, the answer to the question of what makes one
genotype a better competitor than another is largely unknown. The experimental proof of local
adaptation was often challenged by difficulties including the complexity of the ecosystem itself.
112

Most bacterial specialization was revealed through characterization and comparative analysis of
genetic and phenotypic properties (14) (15) (16), while further evidence for adaptation was
often investigated through in situ and in vitro quantification of ecological contributions, and in a
few cases, as competition assays (17).
The impact of natural selection on specialization is through influencing every local
component of the biosphere and adaptation is expected to have occurred at all different levels
of scales. Two important tasks in determining adaptation of species are to determine the
attributes of selective forces in particular environments, and to identify the phenotypic
properties that are responsible for the selective outcome. As influence of natural selection is
reflected through the fitness of individual organisms, identification of the differential fitness of
microorganisms under a specific environment is direct evidence for natural adaptation. Thus,
studying adaptation would improve the understanding of the processes of natural selection and
the underlying mechanisms.
Respiratory versatility is of special importance in the ecological strategy of the genus
Shewanellae (18). Among the S. baltica strains, the capability of respiring nitrate is regarded as
being important ecologically because nitrogen was enriched in the Baltic Sea due to heavy
contamination by urban coastal areas and because the recovered S. baltica are the dominant
denitrifying species (19) (20) (21) (22). Given the redox gradient formed by various redoxactive chemical substrates including oxygen and nitrate, as well as the spacial distribution of
different S. baltica strains along the depths of the water column, I propose to test whether the
isolates exhibit differential fitness from different depths that correspond to their redox
conditions at their position of isolation.
113

In this chapter, five strains were chosen to represent the S. baltica population from
different depths to study fitness effects under different redox conditions using pair-wise
competition assays. To further investigate the mechanisms underlying differential fitness, I used
transcriptional characterization by RNA-Seq to determine how two genetically similar strains,
but isolated from different depths (redox conditions), OS185 and OS195, responded to
competition under nitrate-respiring conditions. Through these efforts, differential fitness of
these strains under particular redox conditions were identified, some of which correspond well
to the environment from where they came. Transcriptomic profiles of OS185 and OS195 under
competition further revealed different reactions of the two strains towards the same redox
condition, providing insights into their adaptive mechanisms.

114

Materials and Methods
Bacterial strains and growth conditions.
Table 5.1. Strain Description and Growth Conditions
Strain

Description

E. coli BW29427 pUX-BF13 {Tn7 transposase genes tnsAE, amp(R), oriT(IncPalpha), R6Kgamma ori (depends on
JMA1012 pir gene in BW29427 for replication) thrB pro thi rpsL
hsdS lacZ(delta) M15 RPF-1316 (delta)araBAD567
(delta)dapA134::[erm pir(wt)]}
E. coli BW29427 pURR25 {miniTn7 containing
(Plac(A1/04/03)-gfpmut3 kan(R) strp(R)) ampR, IncPalpha
JMA1018 oriT, and R6Kgamma ori} thrB pro thi rpsL hsdS
lacZ(delta)M15 RPF-1316 (delta)araBAD567
(delta)dapA1341::[erm pir(wt)]
E. coli BW29427 pURR25 {miniTn7 containing
(Plac(A1/04/03)-ecfp kan(R) strp(R)) ampR, IncPalpha
JD08031 oriT, and R6Kgamma ori} thrB pro thi rpsL hsdS
lacZ(delta)M15 RPF-1316 (delta)araBAD567
(delta)dapA1341::[erm pir(wt)]
S. baltica Strain isolated from 120 m, oxic zone with high nitrate
OS185 level
S. baltica
Strain isolated from 140 m, anoxic zone with no nitrate
OS195
S. baltica Strain isolated from 120 m, oxic zone with high nitrate
OS223 level
S. baltica
Strain isolated from 140 m, anoxic zone with no nitrate
OS167
S. baltica
Strain isolated from 80 m, oxic zone
OS631

Growth
conditions
LB medium with
DAP/300 μg/mL
Amp/100 μg/mL
at 37°C
LB medium with
DAP/300 μg/mL
Amp/100 μg/mL
at 37°C
LB medium with
DAP/300 μg/mL
Amp/100 μg/mL
at 37°C
LB or M1 medium
at 22°C
LB or M1 medium
at 22°C
LB or M1 medium
at 22°C
LB or M1 medium
at 22°C
LB or M1 medium
at 22°C

GFP and CFP tagging of S. baltica strains. The S. baltica strains used for examining fitness under
redox conditions and the Escherichia coli strains carrying necessary plasmids for fluorescent
protein tagging are listed in [Table 5.1]. The gfp and cfp delivery plasmids (pURR25-miniTn7gfpmut3 and pURR25-miniTn7-ecfp) and the helper plasmids (pUX-BF13) were introduced into
the S. baltica strains by tri-parental mating as described in (23) (24). The gfp and cfp tagged
115

strains were used for competition assays. Green fluorescence can be detected in nearly 100% of
gfp tagged cells by flow cytometry. Although cfp tagged strains appeared dark in flow
cytometry, cfp tagged strains were used as controls for possible effects on fitness due to
expression of Tn7 affiliated genes in gfp tagged strains.

Competition assays. Three replicates of gfp and cfp tagged strains were inoculated separately
in LB medium and grown aerobically overnight under 22°C, followed by 1:100 dilution into fresh
M1 minimal medium and grown aerobically overnight before competition. To start competition,
6

an equal number of cells was taken from each cell line and mixed. Around 10 cells from the
mixture were then inoculated into 10 mL of M1 medium supplemented with different electron
acceptors as indicated for the particular competition assays (no supplement for oxic growing
condition and 5 mM sodium nitrate in anaerobic nitrate-respiring condition). Cells were
propagated daily by 1:100 dilution into fresh medium. Samples were taken every day and
analyzed through flow cytometry until one cell line was outcompeted or by the end of 7 days.

Experimental design for transcriptomic characterization. Model strains S. baltica OS185 and
OS195 were selected for transcriptomic characterization during competition under nitrate
respiration. The CFP labeled OS195 strain and GFP labeled OS185 strain were competed in M1
anaerobic medium supplemented with nitrate as described above. Around 0.5 million cells from
each population were taken, mixed together, and inoculated at the beginning of Day 0 for
competition, and cells were taken for sorting by flow cytometry at the same day of inoculation
during the exponential growth phase. To provide control for the possible disturbance on the
116

expression level introduced during flow cytometry sorting and cell collection steps, the two
strains were also grown separately in the same medium in pure culture, followed by the same
procedures of flow cytometry sorting, cell collection, and all the way through illumina RNA-Seq.
Three biological replicates were made for each sample.

RNA extraction and amplification. Fresh cells were taken every 5 min from the anaerobic
serum tubes using a needle and syringe, and sorted through the BD FACSVantage SE system
directly into collection tubes filled with half the volume of RNAlater (Ambion, Grand Island, NY).
An average of 1-2 million cells was collected per sample. Cells were filtered through 13 mm
membrane filters with pore size of 0.2 μm followed by total RNA extraction using the
NucleoSpin RNA XS kit (Clontech, Mountain View, CA). Eluted RNA was checked for quality by
the Agilent BioAnalyzer using the Agilent RNA 6000 Pico kit. Extracted RNA was then amplified
with the Ovation® RNA-Seq System (Nugen, San Carlos, California), and amplified cDNA was
purified using MinElute PCR purification kit (Qiagen, Valencia, CA). Quality and concentration of
amplified cDNA was checked with Agilent BioAnalyzer using the Agilent DNA 7500 kit (Santa
Clara CA). cDNA samples were subjected to Illumina RNA-Seq library preparation, followed with
normalization using duplex-specific nuclease (DSN) for rRNA removal (Evrogen, Moscow,
Russia ).

Mapping, normalization, and statistical analysis of RNA-Seq data. Demultiplexing was done by
the standard Illumina data processing pipeline. Raw reads that passed quality control from
Illumina RNA-Seq were mapped to S. baltica OS185 and OS195 genomes using the Bowtie
117

program (25). Reads aligned to unique positions were used in the downstream analysis.
Normalization and differential expression analysis were performed through the DESeq package
(26) in R. The significance test was based upon assumption of negative binomial distribution
followed by the Benjamini-Hochberg procedure included in the package for adjusting P-value
for multiple tests and differentially expressed genes were determined at 10% false discovery
rate.

118

Results and Discussion
Competition under different electron acceptors. In order to assess the competitive advantages
among S. baltica strains from different redox zones, a number of strains isolated from various
depths were labeled with constitutively expressed cfp or gfp reporter genes and competed
against each other under aerobic growth condition as well as anaerobic growth with nitrate as
the sole electron acceptor. The dynamics of the two competed populations were monitored
daily through flow cytometry, which allowed rapid quantification of the ratio of two cell types.
Differential fitness was revealed from these competition assays [Figure 5.1]. The strains isolated
from nitrate rich waters (OS223, OS185 and OS631) outcompeted those from nitrate-depleted
water (OS195, OS167) under the nitrate-respiring condition, indicating possible adaptation to
higher nitrate levels in their environment. Also, the strain from the microaerophillic zone
(OS185) was able to outcompete the strain from anoxic zone (OS195) under aerobic condition,
which agreed with the oxygen levels in the niches from which they were derived. However, one
strain isolated from anoxic water (OS167) was shown to have outcompeted another strain from
a more oxic zone (OS631) under aerobic condition; while the pattern between the pair of
strains OS223 and OS195 under aerobic condition was not clearly resolved.

119

Figure 5.1. Competition patterns observed among five S. baltica isolates in minimal medium
under aerobic and anaerobic condition with nitrate as the sole electron acceptor. Each arrow
represents a pair of strains competed with each other, and points from the survived strain to
the outcompeted one.

Some of the competition patterns did not agree with the niche profiles of certain strains;
however, one should note that the experimental conditions tested for competition, though
attempting to mimic conditions of the actual environment, do differ from those in nature. Thus,
these competition patterns might not represent well the Baltic Sea, but nonetheless offer a
better controlled approach to assess fitness effects in these strains by various environmental
factors. These assays opened doors to further measurement of relative fitness among a larger
collection of strains in a high-throughput manner, while an in-depth case study would promote
better understanding of potential mechanisms underlying differential fitness of strains under
different environmental conditions.

120

Transcriptional characterization of two model strains S. baltica OS185 and OS195 during
competition with limited nitrate as electron acceptor. Two of the sequenced strains, OS185
and OS195, were selected for further transcriptional characterization of adaptation to niches
with distinct nitrate conditions. OS185 was isolated from an oxic-anoxic transition zone where
nitrate was found at the highest concentration along the water column, and OS195 was isolated
from an anoxic zone characterized by a depletion of nitrate. The two strains had similar growth
kinetics when grown individually with nitrate [Figure 5.2], but when grown in co-culture with
nitrate as the limiting electron acceptor, OS185 outcompeted OS195 [Figure 5.3], suggesting
the competitive advantage of OS185 under nitrate condition. This observation is consistent with
the conditions of environments from which these strains were isolated, which may imply
specialization of the OS185 strain for the nitrate-rich conditions. Furthermore, it has been
shown that OS185 and OS195 had undergone extensive genetic exchange in the recent past,
with over 300 orthologous genes sharing 99.5-100% nucleotide identity (27). Thus, OS185 and
OS195 were ideal candidates for further characterization of response to competition at the
expression level.

121

Hours

0.00
0

6

9

12

15

18

21

log2 ( OD600)

-2.00
-4.00
-6.00

OS185-gfp
OS195-cfp

-8.00

-10.00
-12.00
Figure 5.2. Growth curves of fluorescence tagged S. baltica OS185 and OS195 strains.

Figure 5.3. Competition pattern between fluorescence tagged S. baltica OS195 and OS185 in
minimal medium supplemented with nitrate as sole electron acceptor.

122

To sense competition at an early stage, we analyzed gene expression of these two
competing strains under nitrate-respiring condition by RNA-Seq using the Illumina platform. We
used flow cytometry to sort the two populations from the co-culture based on fluorescent
signals so that we could characterize each of the transcriptomes. In addition, to provide control
for potential bias or altered expression introduced during cell sorting and collecting processes,
cells from the two strains grown individually were also processed through exactly the same
“sorting” procedure by flow cytometry. A summary of sample description is shown in [Table
5.2]. Four sets of samples were collected, from both OS185-gfp and OS195-cfp strains grown in
both pure culture and co-culture conditions, followed by total RNA extraction, amplification
and rRNA removal processes.

Table 5.2. Sample Description

Strain
OS185-gfp
OS195-cfp

Treatment
Grown in Co-Culture and
Grown in Pure Culture and
sorted through Flow
run through Flow Cytometry
Cytometry
Pg
Cg
Pc
Cc

RNA processing and Illumina RNA-Seq summary. The total RNA yield per sample varied from
480 pg to 4295 pg, and no significant RNA degradation was observed [Table 5.3] [Figure 5.4(a)].
Amplification of the total RNA by 1-10 thousand fold was achieved by this amplification method,
resulting in 3-6 μg of amplified cDNA from each sample. Lengths of amplified cDNA ranged from
50 bp to 500 bp, with peak at around 250 bp [Figure 5.4(b)] and suitable for Illumina RNA-Seq

123

library preparation. Illumina sequencing yielded 6.9 to 12.2 million raw reads from each sample
and around 80% of raw reads passed the initial quality control. 29.7%-45.9% of these reads
were successfully mapped onto coding sequences in the two genomes. mRNA reads accounted
for 94.4% to 98.4% of total mapped reads, while only 0.47% to 1.19% of rRNA reads were
recovered [Table 5.4].

Figure 5.4. BioAnalyzer profile of total RNA extracted from sample Pg-2 (a) before
amplification with RNA Integrity Number of 9.3 and (b) after amplification. RNA from other
samples shared a similar profile.

124

Table 5.3. Total RNA Yield and Amplification Efficiency

Lane Sample

1

2

3

Cc-2
Cg-2
Pc-1
Pg-1
Cc-3
Cg-3
Pc-2
Pg-2
Cc-5
Cg-5
Pc-3
Pg-3

Total RNA
Input /pg

Amplified
cDNA /ng

Fold
Amplification

Raw Reads

595
740
2560
4295
810
1115
1450
1800
1310
480
860
1640

5040
5260
4302
5004
5280
4880
5364
3096
5620
5240
6020
5660

8471
7108
1680
1165
6519
4377
3699
1720
4290
10917
7000
3451

10927941
10858169
12676525
7905916
6955956
9210444
15022982
10870701
9859209
12112591
8131929
9388052

Reads
Passed
Filtering
(PF)
8675900
8625439
10031115
6308766
5670504
7259952
11963035
8641043
8274841
9889348
6563987
7610275

Sample %
PF
79.4
79.4
79.1
79.8
81.5
78.8
79.6
79.5
83.9
81.6
80.7
81.1

Table 5.4. Summary of Mapped RNA-Seq Reads
OS185-gfp

All reads
mRNA reads
rRNA reads
% mRNA reads
% rRNA reads

Co-culture
Cg-2
Cg-3
Cg-5
2953204 2617377 4453177
2884902 2554364 4382618
13952
15677
21557
97.7
97.6
98.4
0.47
0.60
0.48

Pure Culture
Pg-1
Pg-2
Pg-3
1874399 2814860 3194192
1821382 2739971 3124396
12612
20608
18156
97.2
97.3
97.8
0.67
0.73
0.57

OS195-cfp

All reads
mRNA reads
rRNA reads
% mRNA reads
% rRNA reads

Cc-2
2898594
2735575
34532
94.4
1.19

Co-culture
Cc-3
Cc-5
2121257 3799864
2008947 3597115
23823
30929
94.7
94.7
1.12
0.81

125

Pc-2
4498426
4287371
46157
95.3
1.03

Pure Culture
Pc-1
Pc-3
3656835 2565570
3479907 2423778
40828
23968
95.2
94.5
1.12
0.93

Overview of transcriptomic profiles. Hierarchical clustering [Figure 5.5] and canonical
correspondence analysis [Figure 5.6] was performed to check the overall similarities among
these profiles. As a result, samples were well separated according to strains used; however,
they were not distinguishable by the types of growth conditions, suggesting that gene
expression patterns were highly similar in co-culture and pure culture conditions.

Figure 5.5. Hierarchical clustering of normalized mRNA reads from the 12 samples. Canberra
index was used to compute distances among sample profiles.

126

Figure 5.6. Canonical Correspondence Analysis of normalized mRNA reads from the 12
samples.

A summary of pairwise comparisons conducted between samples are depicted [Figure
5.7]. Illumina reads that mapped to all genes in OS185 were used for Cg::Pg comparison, as was
done for the Cc::Pc comparison, while reads mapped to only OS185 and OS195 shared genes
were used for Cg::Cc and Pg::Pc comparison. Differentially expressed genes were determined at
a 10% false discovery rate [Figure 5.8]. Through comparing expression profiles of each strain in
co-culture with their expression in pure culture (Cg::Pg and Cc::Pc), we expected to gain more
knowledge about which genes were altered due to introduction of the other strain; whereas
through comparing expression between these two strains under pure culture or co-culture

127

condition (Cg::Cc and Pg::Pc), we sought to answer to which genes were differentially expressed
between these two strains and how this pattern differed between the two growth conditions.

Figure 5.7. Summary of pairwise comparisons performed with RNA-Seq profiles of OS185 and
OS195 under nitrate condition.

128

(a)

(b)

(c)

(d)

Figure 5.8. Testing for differentially expressed genes in the four pairwise comparisons. Scatter
plots of log2 fold changes versus average numbers of reads were drawn for (a) Cc::Cg (b)
Pc::Pg (c) Cg::Pg (d) Cc::Pc. The red color marks genes detected as differentially expressed at 10%
false discovery rate when the Benjamini-Hochberg multiple testing adjustment was used. X-axis
represents the number of reads in each gene, and Y-axis represents the fold change in each
comparison. Dots colored in red correspond to genes that are significantly differentially
expressed.

129

Differential expression caused by culturing methods (pure culture vs. co-culture) in each
strain.
Differential expression of OS195-cfp genes in pure culture and in co-culture (Pc::Cc).
Compared to the expression profile of labeled OS195 in pure culture, 27 genes showed
differential expression while being co-cultured with OS185 strain. Two of these genes were
down-regulated in co-culture, one of which encoded a hypothetical protein, while the other
corresponded to a hydroxylamine reductase in a conserved operon, although none of any other
genes in the operon were differentially expressed. Among the 25 up-regulated genes, one
encoded a tRNA synthetase, with the rest 24 belonging to a phage-like insertion between
Sbal195_2887 (hypothetical protein) and Sbal195_2933 (putative dehydrogenase), downstream
of an operon containing genes encoding membrane transporters [Figure 5.9]. The inserted
sequences contained 45 genes (Sbal195_2888-2932), with the first part resembling components
of TP901 family phage (Sbal195_2888-2902) and the second part (Sbal195_2905-2915) more
similar to components of P2 phage in the Myoviridae family, followed by genes involved in
phage replication and transcriptional regulation (Sbal195_2919-2931). These genes were
adjacent to a phage integrase loci (Sbal195_2932), providing further evidence for phage
mediated insertion of the sequences. The induced expression of phage components included
phage capsid, tail, sheath, as well as baseplate. However, lytic cycle of the phage was not likely
initiated, since once initiated, the entire OS195 population would have probably collapsed
instantaneously, rather than being outcompeted at the end of a week of competition.
Nonetheless, expression of these phage proteins probably simultaneously introduced
disadvantage to the growth of OS195 in the co-culture, which was apparently caused by the
130

competitor strain OS185. However, at this stage, what factors, i.e., signaling agents or
transcriptional regulation pathways, from OS185 might have caused this phenomenon is not
clear and needs to be further characterized through additional biochemistry assays.

Figure 5.9. Insertion of phage genes in OS195 chromosome. Inserted genes were marked
between the red parentheses.

Differential expression of OS185-gfp genes in pure culture and in co-culture (Pg::Cg). Thirtyfive genes were differentially expressed in labeled OS185 grown in co-culture compared to
grown in pure culture, among which 25 genes were up-regulated with the other 10 genes
down-regulated. Genes encoding proteins involved in energy metabolism, ion transport and
metabolism, transcriptional regulation, and posttranslational modification accounted for over
half of all differentially expressed genes. The up-regulated genes included those encoding a
number of transporter related proteins (Shew185_2824-2825, Shew185_1499), transcriptional
regulators (Shew185_1313, Shew185_1006), outer membrane receptor (Shew185_0720) as
well as thioredoxin (Shew185_3102), an enzyme possibly involved in reducing oxidative stress;
whereas the down-regulated genes in co-culture covered a considerable number of
transcriptional regulators such as TetR, CadC, ArsR, (Shew185_2533, Shew185_3867, and
Shew185_1117, respectively) and a response regulator associated polyferredoxin domain
131

protein (Shew185_0311), as well as enzymes involved in nucleotide biosynthesis
(Shew185_1650-1651). Among these genes, 14 were also shown to be differentially expressed
in previous transcriptomic characterization of OS185 and OS195 under various redox conditions
using microarrays (Chapter 3) [Figure 5.10], indicating their functional importance in redox
fluctuated environments.

6

Number of Genes

5

Higher in pure culture
Higher in co-culture

4
3
2
1
0

Figure 5.10. COG categories of differentially expressed genes by OS185 in co-culture vs. pure
culture (Cc::Pc).

In contrast to those genes differentially expressed in OS195 under pure culture and coculture conditions (the majority of which were within a potential phage insertion), most of the
genes differentially expressed in OS185 encoded functional proteins involved in various
pathways, and a number of them were previously shown to be important for anaerobic growth
condition. This suggests that OS185 had a much higher response to the nitrate condition in co132

culture compared to OS195, and impacts from these differentially expressed genes could have
also contributed to the competition outcome between the two strains.

Comparison of expression levels of shared genes between OS185 and OS195 in pure culture
and in co-culture (Pg::Pc and Cg::Cc). In order to discern the general expression patterns
among the samples, a heatmap was generated illustrating differentially expressed genes
between OS185 and OS195 in pure culture and co-culture conditions [Figure 5.11]. The
numbers of differentially expressed genes are also summarized in [Figure 5.12]. No conflicting
expression patterns were observed for any gene (i.e., having higher expression in one strain
under pure-culture condition, while having higher expression in the other strain under coculture condition). This further supports consistency of the expression profiles.

133

Figure 5.11. Heatmap of differentially expressed genes between OS185 and OS195 in coculture (Cc::Cg) and in pure culture (Pc::Pg). Green color indicates higher expression in OS185,
and red indicates higher expression in OS195.

134

Figure 5.12. Summary of numbers of differentially expressed gene shared by both OS185 and
OS195 between the two strains in co-culture and in pure culture.

About 25% of all differentially expressed genes failed to be assigned with any COG
functional categories. Among the successfully annotated genes, and out of the three better
studied classes of COG functional categories, i.e., information storage and processing, cellular
processes and signaling and metabolism, COG categories related to metabolism accounted for a
major fraction of the differentially expressed genes [Table 5.5]. Furthermore, other than the
poorly characterized COG categories R and S, genes associated with amino acid transport and
metabolism (E) were the most frequent, followed by those associated with energy production
and conservation (C), signal transduction (T), transcription (K), and cell wall/ membrane/
envelope biogenesis (M) [Table 5.6].

135

Table 5.5. Number of Genes Classified in Different Cellular Functions
Co-Culture (Cg::Cc)
Pure Culture (Pg::Pc)
Higher
Higher
Higher
Higher
Expressed Expressed Expressed Expressed
in OS185 in OS195 in OS185 in OS195
Information storage
and processing
Cellular processes
and signaling
Metabolism

45

32

34

52

62

58

65

84

79

93

70

129

Table 5.6. Number of Genes Classified in Different COG Categories
COG
D
U
F
Q
M
O
V
I
N
L
H
J
C
G
P
R
S
K
T
E
Unknown

Cc::Cg
3
9
11
8
30
19
6
14
16
28
19
19
39
12
27
49
48
30
37
42
159

136

Pc::Pg
4
13
15
10
37
32
9
13
17
25
12
18
43
16
40
65
52
43
37
50
179

Differentially expressed genes in pure culture (Pc::Pg). Among the 317 genes differentially
expressed specifically in pure culture, 108 genes had higher expression in OS185 while the
other 209 had higher expression in OS195, out of which 39 and 50 genes, respectively, were
previously shown to be differentially expressed under redox conditions (Chapter 3) [Figure
5.13]. Although OS195 has more genes being highly induced, higher expressed genes in OS185
included more of those induced under anaerobic conditions (20 genes in OS185 compared to 11
genes in OS195, as in AnI-8, AnI-9 and AnI-B), while more highly expressed genes in OS195
included those more induced under thiosulfate respiring and aerobic conditions (10 and 14
genes in OS195, compared to 2 and 9 genes in OS185, respectively). These results suggested
that, while cultured individually, OS185 reacted more actively towards anaerobic growth
compared to OS195.

20
18
16
14
12
10
8
6
4
2
0

AI-8
AI-9
AI-B
AnI-8
AnI-9
AnI-B
NI-8
NI-B
NOI-8
NOI-B
NOR-8
NOR-9
NOR-B
NR-8
NR-9
NR-B
TI-8
TI-B
TNI-9
TNR-9
TOI-B
TOR-8
TOR-9
TR-9
TR-B

Higher in OS185
Higher in OS195

Figure 5.13. Differentially expressed genes specific to Pc::Pg comparison. The types of
differential expression were described in [Table 3.6].

137

Differentially expressed genes in both pure culture and co-culture (Pc::Pg, Cc::Cg). One
hundred seventy-eight genes showed higher expression in OS185 and 184 showed higher
expression in OS195 in both co-culture (Cc::Cg) and pure culture comparisons (Pc::Pg) [Figure
5.12]. Consistent patterns in these differentially expressed genes indicated their importance
under nitrate growth condition in both strains. These differentially expressed genes were
further compared with the previously characterized differentially expressed gene pool using
microarrays, where 38% of the genes with higher expression in OS185 were redox sensitive,
whereas a lower percentage (27%) of redox sensitive genes were found in OS195. In addition,
more anaerobic respiration related (AnI) genes were found among those with higher expression
in OS185 (18 genes), contrasting to only nine in OS195 [Figure 5.14], further supporting a highly
active transcriptional response to anaerobic conditions by OS185.
A closer look at functional characterization of differentially expressed genes reveals
similar distribution of higher expressed genes by OS185 and OS195 in most COG categories
[Figure 5.15]. However, only 4 genes with higher expression in OS185 were considered to be
associated with inorganic ion transport and metabolism (P), contrasting to 16 of those with
higher expression in OS195, including a number of membrane transporters and antiporter
(Sbal195_1047-1053, Sbal195_1296), bacterioferritin (Sbal195_3554-3556), as well as the NrfDlike reductase (Sbal195_3997) in a thiosulfate inducible operon. Meanwhile, more genes (15
genes) associated with energy production and conversion (C) had higher expression in OS185
compared to those of OS195 (11 genes), including a number of proteins involved in mediation
of electron transfer, such as cytochromes (Shew185_4167, Shew185_3505), ferredoxins
(Shew185_0311, Shew185_0967), and anaerobic dehydrogenases (Shew185_0099,
138

Shew185_0615). Particularly, the [4Fe-4S] ferredoxin iron-sulfur binding domain protein
(Shew185_0311) was expressed 12.7 fold higher in OS185, suggesting its relative importance
under nitrate respiration in strain OS185.
A number of genes were found to be strongly induced in OS185 or OS195. For example,
two antibiotic activity-related genes (Shew185_1137-1138) had 455 and 152 fold higher
expression in OS185 compared to that in OS195, respectively. Genes in a type-F conjugative
transfer system (Shew185_4435, Shew185_4439, and Shew185_4441-4442) were also strongly
induced in OS185, by 13 to 137 fold higher than in OS195. Similar patterns were also observed
in genes related to a few cell surface structures and pilus biosynthesis (Shew185_0491,
Shew185_4442, 6 and 79 fold higher in OS185), as well as several transposases and integrases
(Shew185_4403, Shew185_4407, and Shew185_0594, 26 to 360 fold higher in OS185).
Compared with OS185, although OS195 has a greater number of more highly expressed genes
in total, highly induced genes (more than 6 fold higher) were fewer than those in OS185 (29
genes in OS185, and 17 in OS195). Moreover, the most greatly induced genes in OS195 were
mostly mobile genetic element associated, such as a number of transposases (Sbal195_0589,
Sbal195_0593, 6.8 and 38.6 fold higher in OS195, respectively), putative prophage
transcriptional regulator (Sbal195_2021, 116 fold higher), and a plasmid partitioning protein
(Sbal195_4640, 1350 fold higher).

139

14
Higher in OS185
Higher in OS195

12
10
8
6
4
2

NR-9
NR-B
TI-8
TI-9
TI-B
TNI-9
TNR-9
TOI-8
TOI-9
TOI-B
TOR-8
TOR-9
TR-9
TR-B

AI-8
AI-9
AI-B
AnI-8
AnI-9
AnI-B
NI-8
NI-B
NOI-8
NOI-B
NOR-8
NOR-9
NOR-B

0

Figure 5.14. Differentially expressed genes from both Pg::Pc and Cg::Cc comparisons.

Figure 5.15. COG categories of differentially expressed genes shared by OS185 and OS195
from Pg::Pc comparison. Positive gene numbers correspond to genes with higher expression in
OS185; Negative gene numbers correspond to genes with higher expression in OS195.
140

Co-culture specific differentially expressed genes (Cc::Cg). One hundred twelve genes
exhibited higher expression in OS185 while 104 exhibited higher expression in OS195
exclusively in co-culture transcriptional comparison of the two strains [Figure 5.12]. The
numbers of more highly expressed genes that fell into most COG categories were similar,
except for the cell wall/membrane/envelope biogenesis related category (M), where 10 genes
with higher expression in OS195 were found, contrasting to only 1 with higher expression in
OS185 [Figure 5.16]. Moreover, 4 out of the 10 genes in the (M) category were previously
shown to be up-regulated under aerobic condition. The situation was quite the opposite in a
few other categories, including those associated with transcription (K), signal transduction
mechanisms (T), as well as amino acid transport and metabolism (E), where more genes with
greater expression in OS185 were found (13 vs. 2, 14 vs. 3, and 14 vs. 6 genes in OS185 and
OS195, in K, T and E categories, respectively) compared to those in OS195. Remarkably, among
these higher expressed genes in OS185, 11 transcriptional regulators, 4 periplasmic sensors
(Shew185_0052, 0253, 0259, 1002), as well as 2 chemotaxis sensory proteins (Shew185_2245,
Shew185_4113) were found. Some of the transcriptional regulators were known to be essential
under fluctuating redox conditions (28) (18), and moreover, two of them (Shew185_0096,
Shew185_2371) were previously shown to be up-regulated by OS185 under nitrate-respiring
condition. A number of nitrogen sensors and regulatory proteins were also included, i.e., a
sensor histidine kinase for nitrogen response NtrB and the corresponding transcriptional
regulator NtrC (Shew185_0258-0259), as well as the nitrogen regulatory protein GlnK_1
(Shew185_3708), suggesting highly active response of OS185 to nitrate-respiring condition in
co-culture.
141

Figure 5.16. COG categories of differentially expressed genes shared by OS185 and OS195
from Cg::Cc comparison. Positive gene numbers correspond to genes with higher expression in
OS185; Negative gene numbers correspond to genes with higher expression in OS195.

RNA-Seq technology in transcriptomic analysis. One of the key strategies to understand the
ecology of an organism is through determining how it responds to its environment by altering
gene expression. High-throughput Illumina sequencing of cDNAs provides a platform for
comprehensively and quantitatively assessing the cellular response at the transcription level
(29). Studies using this method have already altered our views of the extent and complexity of
diverse transcriptomic systems (30) (31) (32). Being known as one of the Next Generation
Sequencing (NGS) technologies and dubbed "a revolutionary tool for transcriptomics", Illumina
RNA-Seq is capable of generating millions of short reads from little cDNA and offering much
greater sequencing depth. In contrast to hybridization-based methods, sequence-based
142

approaches directly measure cDNA products at base-scale resolution. Compared to microarrays,
RNA-Seq produces much lower background noise given that sequencing reads can be
unambiguously mapped to specific regions of the genome. Also due to the great sequencing
capacity, the RNA-Seq approach can detect a much larger dynamic range of differentiated
expression levels.
In the first publication of bacterial expression profiles derived from RNA-Seq by Illumina
NGS, transcriptional responses of two Burkholderia cenocepacia strains of the same lineage to
two different conditions were examined (33). The two B.cenocepacia isolates (one from soil and
the other from the sputum of a cystic fibrosis patient) were each grown under conditions
mimicking soil and CF sputum, followed by sequencing of their transcriptomes. As a result,
differences were not only found between the expression profiles of cells grown in two different
media, but also found between expression profiles of the two strains under the same
conditions, suggesting probable strain specific adaptations to the niches from which they came.
A number of previously unknown putative non-coding RNAs (ncRNAs) were also found to be
preferentially up-regulated under conditions mimicking the soil environment, indicating
potential roles of ncRNAs in specializing to the soil environment.
In this study, Illumina sequencing of whole transcriptomes of the two S. baltica strains
yielded 15 to 35 times coverage (based on number of mapped reads) of the whole genome of
OS185 and OS195, making possible detection of transcripts with low copy numbers. Fold
change between differentially expressed genes of up to 1350 times was identified, compared to
maximum fold change of 648 times in differential expression analysis using microarrays. Over
96.6% of all genes in OS195 and 97.4% of all genes in OS185 were represented by at least a
143

single read, while fewer than 100 singletons were revealed in each replicate sample, further
demonstrating the great sequencing depth of this method. Due to efficient removal of RNA
with secondary structure, reads that mapped to rRNA, tRNA and ncRNAs were very limited
(fewer than 6% in total in each replicate sample), rendering biased any quantification of these
types of RNAs. Thus, we were not able to quantitatively assess the relative abundance of these
RNA types between samples. On the other hand, given a large number of reads from mRNA
sequences, these mRNA reads should be able to well represent the true transcriptomic profiles
of the samples.

Mechanisms of differential fitness. Similar to the RNA-Seq study of the two B.cenocepacia
isolates, we also aimed to investigate the mechanisms underlying specialization in their own
environmental niches by these two S. baltica strains. In addition, we took into consideration the
relative fitness of the two strains while grown together and competed with each other, which
leads to interpreting the specializing outcome of the two strains. Given that OS185 was able to
outcompete OS195 by the end of the 7th day (45 generations), the mean relative fitness of
OS185 over OS195 was estimated to be 1.05 (34), meaning that OS185 was able to grow under
the nitrate condition at a rate 5% faster than OS195 on a daily base. Although it is by far not
clear which factors caused this difference between the two strains under the nitrate amended
condition, several possibilities were proposed based on the transcriptomic profile of the strains.
Strikingly, one possibility of OS195 being outcompeted may be associated with induction of
expression of phage components by OS185 in the co-culture. However, based upon previous
knowledge, phage induction, often associated with initiation of the lytic cycle, was most
144

commonly triggered by exposure to UV radiation or other physico-chemical stress factors (35)
(36), whereas the case of phage induction by another organism is rarely known. Also, although
it has long been known that bacteria can produce toxins and release them into their
surroundings to inhibit growth of other bacteria (37), introducing competitive disadvantage via
altering expression in the other organism, i.e., inducing expression of phage components, has
not been reported. Thus, further experimental validation and investigation would provide
important evidence to these possibilities, and may contribute to a re-evaluation of the role of
bacteriophages in competition in the natural environment. The other possibility that might
have contributed to the competition outcome was more effective transcriptional regulation in
the strain OS185. From all levels of comparisons, i.e., Cc::Cg, Pc::Pg, as well as Cg::Pg, genes
potentially important for redox specialization were found to be differentially regulated in
OS185. Moreover, compared with transcription profiles of OS195, OS185 appeared to have
more active response to the nitrate respiring condition on the gene expression level, which
further explains the adaptation of OS185 to a nitrate-rich environment. Although detailed
comparison on the gene level was often limited by the quality and quantity of gene annotation
available, and verification of importance of differentially expressed genes through molecular
biochemical studies is still needed, this study establishes the groundwork for future mechanistic
investigation of niche adaptation of the S. baltica population, and meanwhile provided an
effective method for studying interaction among microorganisms.

145

REFERENCES

146

References
1.

Cohan FM & Koeppel AF (2008) The Origins of Ecological Diversity in Prokaryotes.
Current Biology 18(21):R1024-R1034.

2.

Koonin EV (2008) Darwinian evolution in the light of genomics. Nucleic Acids Research
37(4):1011-1034.

3.

Weinbauer MG & Rassoulzadegan F (2003) Are viruses driving microbial diversification
and diversity? Environmental Microbiology 6(1):1-11.

4.

Cohan FM & Perry EB (2007) A Systematics for Discovering the Fundamental Units of
Bacterial Diversity. Current Biology 17(10):R373-R386.

5.

Koeppel A, et al. (2008) Identifying the fundamental units of bacterial diversity: A
paradigm shift to incorporate ecology into bacterial systematics. Proceedings of the
National Academy of Sciences 105(7):2504-2509.

6.

Hunt DE, et al. (2008) Resource Partitioning and Sympatric Differentiation Among
Closely Related Bacterioplankton. Science 320(5879):1081-1085.

7.

Johnson ZI (2006) Niche Partitioning Among Prochlorococcus Ecotypes Along OceanScale Environmental Gradients. Science 311(5768):1737-1740.

8.

Coleman ML (2006) Genomic Islands and the Ecology and Evolution of Prochlorococcus.
Science 311(5768):1768-1770.

9.

Dufresne A, et al. (2008) Unravelling the genomic mosaic of a ubiquitous genus of
marine cyanobacteria. genome biology 9(5):R90.

10.

Swingley WD, et al. (2008) Niche adaptation and genome expansion in the chlorophyll dproducing cyanobacterium Acaryochloris marina. Proceedings of the National Academy
of Sciences 105(6):2005-2010.

11.

Jacobsen A, Hendriksen RS, Aaresturp FM, Ussery DW, & Friis C (2011) The Salmonella
enterica Pan-genome. Microbial Ecology.

12.

Jaureguy F, et al. (2008) Phylogenetic and genomic diversity of human bacteremic
Escherichia coli strains. BMC Genomics 9(1):560.

13.

Mahenthiralingam E, Baldwin A, & Dowson CG (2008) Burkholderia cepacia complex
bacteria: opportunistic pathogens with important natural biology. Journal of Applied
Microbiology 104(6):1539-1551.

147

14.

Kaiser S, Biehler K, & Jonas D (2009) A Stenotrophomonas maltophilia Multilocus
Sequence Typing Scheme for Inferring Population Structure. Journal of Bacteriology
191(9):2934-2943.

15.

Stomp M, et al. (2004) Adaptive divergence in pigment composition promotes
phytoplankton biodiversity. Nature 432:104-107.

16.

Denef VJ, et al. (2010) Inaugural Article: Proteogenomic basis for ecological divergence
of closely related bacteria in natural acidophilic microbial communities. Proceedings of
the National Academy of Sciences 107(6):2383-2390.

17.

Belotte D, Curien J-B, Maclean RC, & Bell G (2003) An experimental test of local
adaptation in soil bacteria. Evolution 57(1):27-36.

18.

Fredrickson JK, et al. (2008) Towards environmental systems biology of Shewanella.
Nature Reviews Microbiology 6(8):592-603.

19.

Elmgren R (1989) Man's Impact on the Ecosystem of the Baltic Sea: Energy Flows Today
and at the Turn of the Century. AMBIO:326-332.

20.

Höfle MG & Brettar I (1996) Genotyping of heterotrophic bacteria from the central Baltic
Sea by use of Low-Molecular-Weight RNA profiles. Applied and Environmental
Microbiology 62(4):1383–1390.

21.

Brettar I (2002) Shewanella denitrificans sp. nov., a vigorously denitrifying bacterium
isolated from the oxic--anoxic interface of the Gotland Deep in the central Baltic Sea.
International Journal of Systematic and Evolutionary Microbiology 52(6):2211-2217.

22.

Ziemke F, Brettar I, & Höfle MG (1997) Stability and diversity of the genetic structure of
a Shewanella putrefaciens population in the water column of the central Baltic. Aquatic
Microbial Ecology 13:63-74.

23.

Victor De Lorenzo, Marta Herrero, Ute Jakubzik, & Timmis AKN (1990) Mini-TnS
Transposon Derivatives for Insertion Mutagenesis, Promoter Probing, and Chromosomal
Insertion of Cloned DNA in Gram-Negative Eubacteria. Journal of Bacteriology 172:65686572.

24.

Lambertsen L, Sternberg C, & Molin S (2004) Mini-Tn7 transposons for site-specific
tagging of bacteria with fluorescent proteins. Environmental Microbiology 6(7):726-732.

25.

Langmead B, Trapnell C, Pop M, & Salzberg SL (2009) Ultrafast and memory-efficient
alignment of short DNA sequences to the human genome. genome biology 10(3):R25.

26.

Simon Anders WH (2010) Dierential expression analysis for sequence count data.
genome biology 11(10):R106.

148

27.

Caro-Quintero A, et al. (2010) Unprecedented levels of horizontal gene transfer among
spatially co-occurring Shewanella bacteria from the Baltic Sea. The ISME Journal 5:131–
114.

28.

Beliaev AS, et al. (2005) Global Transcriptome Analysis of Shewanella oneidensis MR-1
Exposed to Different Terminal Electron Acceptors. Journal of Bacteriology 187(20):71387145.

29.

Wang Z, Gerstein M, & Snyder M (2009) RNA-Seq: a revolutionary tool for
transcriptomics. Nature Reviews Genetics 10:57-63.

30.

Tariq MA, Kim HJ, Jejelowo O, & Pourmand N (2011) Whole-transcriptome RNAseq
analysis from minute amount of total RNA. Nucleic Acids Research 39(18):e120-e120.

31.

Malone JH & Oliver B (2011) Microarrays, deep sequencing and the true measure of the
transcriptome. BMC Biology 9(34):1741-7007.

32.

Croucher NJ & Thomson NR (2010) Studying bacterial transcriptomes using RNA-seq.
Current Opinion in Microbiology 13(5):619–624.

33.

Otto M, Yoder-Himes DR, Konstantinidis KT, & Tiedje JM (2010) Identification of
Potential Therapeutic Targets for Burkholderia cenocepacia by Comparative
Transcriptomics. PLoS ONE 5(1):e8724.

34.

Lenski RE, M. R. Rose, S. C. Simpson, and S. C. Tadler (1991) Long-term experimental
evolution in Escherichia coli. I. Adaptation and divergence during 2,000 generations.
American Naturalist 138:1315-1341.

35.

Baluch J & Sussman R (1978) Correlation between UV dose requirement for lambda
bacteriophage induction and lambda repressor concentration. journal of Virology
26(3):595-602.

36.

Choi J KS, Goel R (2010) Various physico-chemical stress factors cause prophage
induction in Nitrosospira multiformis 25196--an ammonia oxidizing bacteria. Water
Research 44(15):4550-4558.

37.

Hibbing ME, Fuqua C, Parsek MR, & Peterson SB (2009) Bacterial competition: surviving
and thriving in the microbial jungle. Nature Reviews Microbiology 8(1):15-25.

149

CHAPTER VI
SUMMARY AND OUTLOOK

150

With the improving ability to differentiate and characterize bacterial strains, as well as
the development of powerful genomic technologies, microbiologists today are able to make
more significant advances in the fast paced field of environmental microbiology. The work
presented in this thesis has taken advantage of past and current technologies, along with the
unique biological resources and ecological data, aiming to provide further understanding of
bacterial specialization.
The Shewanella represent an interesting group of bacteria due to their respiratory
versatility, their ability to thrive in many environments, as well as their genetic diversity and
genomic flexibility (1). All these properties provide Shewanella with greater ability to adapt
especially to fluctuating redox environments. While a number of Shewanella strains have been
well studied and characterized, in part due to their tremendous bioremediating potential, a
number of new strains continue to be discovered.
S. baltica represents a bacterial population that was persistent in the Baltic Sea where it
played a major role in the nitrogen cycle in the last half of the 20

th

century when nitrogen and

organic carbon inputs to the Baltic Sea increased greatly (2) (3). With their important roles in
the in the Baltic environment and their great potential for use in studying short-term evolution,
a number of S. baltica genomes were sequenced by DOE’s Joint Genome Institute, with four
genomes finished and released during 2007-2008, and an additional five released during 20102011. Although analysis of the newly available genomes was not included in this study, these
genomes will continue to provide valuable insights into mechanistic studies of S. baltica
evolution.

151

Chapter 1 framed the comparative genomic, phenotypic and transcriptomic analyses for
the subsequent chapters, and provided an overview of Shewanella as well as the history behind
Shewanella baltica species. Chapter 1 also supplied an evolutionary perspective on bacterial
specialization, and signified the value of the S. baltica isolate collection in advancing
understanding of evolutionary strategies and mechanisms.
Chapter 2 examined the genetic relatedness among S. baltica lineages, and made the
initial correlation of these isolates with the environmental gradients. It also tackled the
question of the levels of recombination within the S. baltica population, and extended a
previous finding on two strains on which extensive recombination has occurred in the recent
past. This estimation is important for population genetics studies because it reflects the
evolutionary paths, i.e., divergent or convergent evolution, of the population. Thus, the S.
baltica population in general, can be regarded as undergoing diversification. Further, through
comparing MLST phylogeny with the broader scale similarity of CGH presented in Chapter 3, the
new gene selection strategy used for this MLST analysis was substantiated, which could benefit
MLST study in other microbial systems as well.
Chapter 3 coped with the question of genomic variability within the S. baltica species. In
addition to the CGH assays, transcriptomic characterization also helped identify genes of
functional importance under fluctuating redox conditions. Core genome analyses were
performed at levels of both the entire S. baltica population and within specific lineages
representing isolates from potentially different environmental niches. Although the analysis
was to some extent limited by lack of quality gene annotations, the large proportion of genes,

152

both hypothetical and annotated, that are associated with mobile genetic elements indicated
the capacity of these organisms for harboring new genes and functions through means of HGT.
Chapter 4 described the efforts in seeking for evidence of specialization on the
phenotypic level. Given that Shewanella species are often associated with organic-rich, electron
acceptor varying environment, I examined the carbon source utilization and respiratory
versatility of these S. baltica strains. Importantly, the strains isolated in 1998 used more carbon
sources than those from 1986/87, which could reflect a change in nutrient availability with time
in the Baltic Sea. One other important task in microbial genetics is to reveal to which extent
genomic information can reflect phenotypic properties. Chapter 4 attempted to correlate
phenotypic similarity with genetic relatedness at the whole genome level and at the individual
gene level, and found a few features that appeared to be explained by particular genes. Beyond
this thesis, phenotypic profiles will continue to assist functional discovery of genes and
pathways, and to improve gene annotation of Shewanella genomes (4).
For Chapter 5, I sought to determine the basis of competitive advantage in competing
strains under conditions mimicking their own environments, and found differential fitness of
strains under aerobic and anaerobic nitrate respiring condition. These model studies of course
do not accurately reflect the organisms’ natural environment, but they do better control
conditions and isolate variables that affect fitness of strains. Probably the most pace-setting
aspect of this work was to extend the competition analysis to the transcriptome level, which
revealed distinguishable patterns of differential gene regulation in the competed strains. While
analysis of these data is not complete, I did find induction of genes for phage production in the
out-competed strain and higher expression of genes related to anaerobic growth in the favored
153

strain, which was consistent with the competitive outcome. In addition, a phage repressor
(Sbal195_2927) and one other transcriptional regulator (Sbal195_2888) were identified in the
phage island. Hence, a next step would be to do bioinformatic searches to find potential
regulatory mechanisms, i.e., regulatory proteins or nucleotides that may bind to these regulator
genes, if found, their function could then be tested by biochemical assays.
Further work would likely enhance the explanatory connections among the genomics,
phenotypes, and the ecology. For examples, in order to further investigate the relation
between genomics and phenotypes, one can obtain the Biolog metabolic profile and try to
match this physiology with metabolic pathways encoded in the genomes. This was recently
successfully accomplished for different species of Shewanella (4), and it should be extendable
to strains within a species, namely, the S. baltica isolates. Meanwhile, the CGH dataset of S.
baltica strains will provide a useful reference and additional support for functional predictions
of genes and pathways based on this analysis. This strategy was shown to accelerate functional
discovery of genes and will be also utilized to refine gene annotation of the S. baltica species,
which will no doubt improve our understanding about to what extent genomic information can
be used to infer phenotypic properties.
Investigation in gene composition of populations occupying different niches will provide
important insights in evaluating the impact of ecology on bacterial population, as reflected
from their genomes, and determining whether we can use genomics to predict the
environmental niches of individual strains. To further illustrate this potential, Erick Cardenas
(who studies Burkholderia) and I compared the conserved gene core of Shewanella baltica and
Burkholderia cenocepacia. Burkholderia are Betaproteobacteria with incredible phenotypic
154

diversity, likely owing to their large, multi-replicon genomes. Although B. cenocepacia is
primarily known as a pathogen in cystic fibrosis patients, many of B. cenocepacia strains are of
environmental origin and are able to colonize diverse niches such as plant rhizospheres, soil,
water, insects and even mammals (5) (6) (7) (8). The gene core of B. cenocepacia was also
obtained through CGH in a parallel study (Cardenas, unpublished data), and was used to
compare with that of the S. baltica strains as described in Chapter 3. The analysis is so far only
at the level of comparing COG functional categories, but has already revealed interesting
signatures of functional differences [Figure 6.1]. Given the versatility in inhabiting diverse
environments by B. cenocepacia, it is not surprising that, compared to the gene core of S.
baltica, B. cenocepacia has more genes involved in transcription (K), amino acid transport and
metabolism (E), and carbohydrate transport and metabolism (G), as well as secondary
metabolism (Q). On the other hand, the core of S. baltica has more genes involved in signal
transduction (T), energy production (C), coenzyme transport and metabolism (H), cell motility
(N), and post-translational modification, protein turnover and chaperones (O), which agreed
with the nature of Shewanella of being respiratory specialists and occupying niches with rich
and fluctuated energy sources. These results indicated that the ecology of a bacterial
population can be reflected from particular groups of functional genes in their genomes. To
gain perspectives of functional specialization in more bacterial groups that inhabit more diverse
niches, we plan to extend this analysis to more bacterial species, such as Escherichia coli and
certain Prochlorococcus groups.

155

Figure 6.1. COG functional comparison of S. baltica gene core and Burkholderia cenocepacia
gene core.

A larger set of genome sequences is needed to further evaluate how ecology impacts
the evolutionary paths of the S. baltica population. In this regard, the sequenced genomes of
five more strains of S. baltica are now available. Recombination analysis on the single gene level
can be biased in inferring evolutionary paths, while on the whole genome level, it can more
conclusively reveal evolutionary trajectories. For example, in a recent comparative genomic
study of Escherichia coli strains, higher levels of recombination were found among strains from
more similar environments (9). Thus, through analyzing the extent of recombination among S.
baltica strains within similar and between more different redox zones, we may learn more
156

about how stratification of the water environment in the Baltic Sea impacted the population’s
genomic structure.
In advancing our understanding of the connections between phenotypes and ecology of
S. baltica, larger-scale competition assays can be performed to test differential fitness among a
larger number of strains. These assays can be conducted in a high throughput manner, for
example, by culturing mixed cells in 96-well plates in anaerobic chamber, with medium
containing desired electron acceptors. Results from these assays will help us to know about
which environmental parameters are of importance in determining differential fitness among
the strains.
Finally, as we continue to explore specialization of microorganisms, it is important to
remember that microbes do not live by themselves. Rather, communities are composed of
coexisting organisms including Bacteria, Archaea, eukaryotes, and viruses. Hence it is possible
that through learning more about associated communities, as well as understanding those
organisms in the communities, we will come to more comprehensive understanding about the
studied microorganisms. For instance, in this case, 16S rRNA pyrotag sequencing of the
community DNA from the original water samples can provide information on composition of
the bacterial community and an estimation of the proportion of the entire bacterial population
that belongs to the S. baltica species (10). Meanwhile, pyrotag sequencing of functional genes,
for example, those involved in nitrate reduction pathways, should provide insights into the
diversity of nitrate-reducing genes and, especially in conjunction with 16S pyrosequencing
community structure information, will be informative about which bacterial groups play
important roles in nitrate reduction in the Baltic Sea (11) (12).
157

Shotgun metagenomic sequencing should provide even greater depth of information on
the Balic Sea community, but is more resource demanding both for sequencing and massive
data analysis. Nonetheless, it would provide much more sequence and with some assembly,
larger fragments of community DNA, enabling further analysis of a larger number of functional
genes. In particular, shotgun metagenomic sequencing in combination with bait capture
technology (13) using S. baltica DNA would enable the sequencing to be targeted to only the S.
baltica genomes from the community DNA, which should allow further analysis of the diversity
of S. baltica genes as well as of recombination between S. baltica genomes. In summary,
metagenomics in its ultimate vision should provide further understanding of the function,
structure and ecology of the natural community.

158

REFERENCES

159

References
1.

Fredrickson JK, et al. (2008) Towards environmental systems biology of Shewanella.
Nature Reviews Microbiology 6(8):592-603.

2.

Elmgren R (1989) Man's Impact on the Ecosystem of the Baltic Sea: Energy Flows Today
and at the Turn of the Century. AMBIO:326-332.

3.

Brettar I & Höfle MG (1993) Nitrous oxide producing heterotrophic bacteria from the
water column of the central Baltic: abundance and molecular identification. MARINE
ECOLOGY PROGRESS SERIES 94:253-265.

4.

Rodrigues JLM, Serres MH, & Tiedje JM (2011) Large-Scale Comparative Phenotypic and
Genomic Analyses Reveal Ecological Preferences of Shewanella Species and Identify
Metabolic Pathways Conserved at the Genus Level. Applied and Environmental
Microbiology 77(15):5352-5360.

5.

Coenye T & Vandamme P (2003) Diversity and significance of Burkholderia species
occupying diverse ecological niches. Environmental Microbiology 5(9):719-729.

6.

Mahenthiralingam E, Baldwin A, & Dowson CG (2008) Burkholderia cepacia complex
bacteria: opportunistic pathogens with important natural biology. Journal of Applied
Microbiology 104(6):1539-1551.

7.

Vial L, Chapalain A, Groleau M-C, & Déziel E (2011) The various lifestyles of the
Burkholderia cepacia complex species: a tribute to adaptation. Environmental
Microbiology 13(1):1-12.

8.

Mahenthiralingam E, Urban TA, & Goldberg JB (2005) The multifarious, multireplicon
Burkholderia cepacia complex. Nature Reviews Microbiology 3(2):144-156.

9.

Chengwei Luo STW, David M. Gordon, Michael Feldgarden, James M. Tiedje, and
Konstantinos T. Konstantinidis (2011) Genome sequencing of environmental Escherichia
coli expands understanding of the ecology and speciation of the model bacterial species.
Proceedings of the National Academy of Sciences 108(17):6.

10.

Petrosino JF, Highlander S, Luna RA, Gibbs RA, & Versalovic J (2009) Metagenomic
Pyrosequencing and Microbial Identification. Clinical Chemistry 55(5):856-866.

11.

Demaneche S, et al. (2008) Characterization of Denitrification Gene Clusters of Soil
Bacteria via a Metagenomic Approach. Applied and Environmental Microbiology
75(2):534-537.

12.

Iwai S, et al. (2009) Gene-targeted-metagenomics reveals extensive diversity of aromatic
dioxygenase genes in the environment. The ISME Journal 4(2):279-285.
160

13.

Gnirke A, et al. (2009) Solution hybrid selection with ultra-long oligonucleotides for
massively parallel targeted sequencing. Nature Biotechnology 27(2):182-189.

161