환경구배에따른미생물군집의...
TRANSCRIPT
이학박사학위논문
환경구배에 따른 미생물 군집의
메타지놈 구조 변화
Changes in Metagenome Structure of Microbial
Communities along Environmental Gradients
2018 년 2 월
서울대학교 대학원
생명과학부
송 호 경
Changes in Metagenome Structure
of Microbial Communities along
Environmental Gradients
Hokyung Song
Advisor: Professor Piotr Jablonski, Ph. D.
A Thesis Submitted for the Partial Fulfillment of the
Degree of Doctor of Science, Biological Sciences
February 2018
Graduate School of Biological Sciences
Seoul National University
i
Abstract
Ecology tackles the diversity and interactions of all life on
Earth. In recent years, it has become possible in ecology to obtain
radically new perspectives on the living world by using methods
based on nucleic acid sequencing. These methods, falling under the
categories of amplicon sequencing and shotgun metagenomic
sequencing, have revealed many hidden dimensions of microbial
ecology. Through amplicon sequencing techniques using key
taxonomic marker genes (usually rRNA genes), a vast natural
diversity of bacteria, archaea, fungi and micro-invertebrates has
been revealed from soil, sediment and waters. It has become
possible to test established ecological theories of diversity and
community assembly on microorganisms in fundamentally new ways.
Metagenomics, the direct sequencing of the collective genomes
found in a given environmental sample, has revealed important
insight into the potential functions of microbial communities from
various environments.
Studying how metagenome composition and diversity varies
along environmental gradients may improve the understanding of
the general principles of community and ecosystem structuring. A
culture system was set up using soil elutant to study the effects of
varying initial nutrient concentration, and time succession, on the
ii
community metagenome. A distinct set of nutrient-related or time-
related changes in the metagenome was found. For example, a high
nutrient (copiotrophic) strategy was associated with greater
abundance of genes related to cell division and cell cycle, while a
low nutrient (oligotrophic) strategy had greater abundance of genes
related to carbohydrate metabolism and virulence, disease and
defense. Time-related changes in the metagenome was also found,
revealing a distinct ‘r’-related strategy with greater abundance
of genes related to regulation and cell signaling, and a ‘K’
strategy rich in motility and chemotaxis-related genes. In terms of
diversity, both the richness of operational taxonomic unit and the
richness of species assignment of functional genes showed linear
correlations with functional gene richness, supporting the
hypothesis that greater taxonomic diversity is associated with
greater functional diversity, with possible implications for
ecosystem stability.
Aridity is also one of the most important factor for
structuring ecosystem. To see the effects of aridity stress on the
functional diversity of the soil biota, soil metagenomes along a
rainfall gradient across the eastern Tibetan Plateau was studied.
Lower mean annual precipitation was strongly associated with
reduced functional diversity of bacterial genes. It appears that
extreme environmental conditions associated with aridity constrain
iii
the diversity of functional strategies present in soil biota –
analogous to broad scale patterns found in plant functional diversity
along environmental gradients. In terms of specific functions, more
arid conditions were also associated with increased relative
abundance of dormancy related genes, and osmoprotectant-related
genes. Decreased relative abundance of antibiotic resistance and
virulence genes in more arid conditions suggests reduced intensity
of biotic interaction under extreme physiological conditions. These
trends parallel those seen in preliminary comparisons of
metagenomes across biomes.
In future studies, it will be interesting to elucidate changes
in metagenome structure of soil biota along different types of
environmental gradients, for example, pH, temperature, or soil
oxygen. Also, a study of metatranscriptomics may be required to
prove activity of functional genes in soil community. Development
of enrichment technology of prokaryotic mRNA would be a
prerequisite for achieving this goal.
Keyword: metagenome, functional diversity, copiotrophic,
oligotrophic, r-selection, K-selection, precipitation, aridity,
ecological stress
Student Number: 2016-30109
iv
v
Table of Contents
Abstract .................................................................................................. i
Table of Contents ................................................................................. v
List of Figures.......................................................................................x
List of Tables .....................................................................................xvi
CHAPTER 1: Application of Metagenomics towards an Assessment
of Biodiversity in Microbial Ecology: An Introduction .....................1
1. 1. The impact of advanced sequencing technology on
microbial ecology............................................................................2
1. 1. 1. Discovery of the unseen majority: soil microorganisms.2
1. 1. 2. A general procedure for soil microbial community
analysis with amplicon sequencing .................................................4
1. 1. 3. Lack of linkage between microbial community structure
and function .......................................................................................8
1. 2. An assessment of functional traits of soil microorganisms
10
1. 2. 1. Diverse methods to describe functional traits .............. 10
1. 2. 2. Metagenomic approach for assessing microbial
functional traits .............................................................................. 11
1. 3. Research objectives ............................................................16
CHAPTER 2: Bacterial Strategies along Nutrient and Time
vi
Gradients, Revealed by Metagenomic Analysis of Laboratory
Microcosms..........................................................................................19
2. 1. Introduction ..........................................................................20
2. 2. Materials and Methods ........................................................27
Sample preparation........................................................................ 27
Sequencing and sequence processing ......................................... 28
Chemical analysis of culture material.......................................... 29
Statistical analysis ......................................................................... 30
2. 3. Result ....................................................................................32
Chemical analysis of culture material.......................................... 32
Taxonomic and functional composition of communities ............ 34
Functional gene categories correlating with initial nutrient
concentration.................................................................................. 42
Functional gene categories correlating with incubation time ... 54
Taxonomic and functional diversity............................................. 61
Taxonomic annotation of functional genome .............................. 68
2. 4. Discussion.............................................................................73
Chemical analysis of culture material.......................................... 73
Functional gene categories correlating with initial nutrient
concentration (hypothesis A)....................................................... 73
vii
Functional gene categories correlating with incubation time
(hypothesis B) ............................................................................... 75
Taxonomic and functional diversity (hypothesis C).................. 79
Taxonomic annotation of functional genome .............................. 80
Possible limitations of this study and suggestions for further
work................................................................................................. 81
2. 5. Concluding remarks.............................................................86
CHAPTER 3: Environmental Filtering of Bacterial Trait Diversity
along an Aridity Gradient...................................................................87
3. 1. Introduction ..........................................................................88
3. 2. Materials and Methods ........................................................95
Soil sampling and data collection ................................................. 95
Climatic data collection and physicochemical analysis.............. 97
DNA extraction, sequencing and sequence processing............. 99
Statistical analysis ....................................................................... 100
3. 3. Result ..................................................................................102
Functional gene composition along an aridity gradient ........... 102
Functional gene composition along nutrient gradient .............. 110
Functional gene diversity vs OTU diversity............................. 114
Functional gene diversity along aridity gradients.................... 116
viii
Taxonomic annotation of functional genes ............................... 123
3. 4. Discussion...........................................................................128
Hypothesis 1) Trait diversity will decrease along an aridity
gradient ......................................................................................... 128
Hypothesis 2) Genes related to cell division and dormancy will
increase along aridity gradient................................................... 129
Hypothesis 3) Stress response genes – including those relating
to osmoregulants - will be more abundant in higher aridity
gradients ....................................................................................... 130
Hypothesis 4) There will be less incidence of antibiotic
resistance and virulence genes in more arid conditions ......... 131
Hypothesis 5) Motility related genes will increase along aridity
gradient ......................................................................................... 132
Hypothesis 6) Functional gene composition along nutrient
concentration will follow those found in the microcosm (Chapter
2) ................................................................................................... 132
Possible limitations of the study ................................................ 133
3. 5. Concluding remarks...........................................................135
CHAPTER 4: General Conclusions .................................................137
4. 1. Changes in metagenome structure of microbial
communities along environmental gradients ...........................138
ix
4. 2. Future directions ...............................................................141
4. 2. 1. Study of various environmental gradients influencing
microbial community.................................................................... 141
4. 2. 2. Metatranscriptomics....................................................... 141
Publication..........................................................................................143
Manuscript in preparation ................................................................144
References.........................................................................................145
Appendix............................................................................................160
국문초록 (Abstract in Korean) ........................................................167
x
List of Figures
Figure 1. A general procedure for soil microbial community analysis
based on amplicon sequencing of microbial marker genes........... 7
Figure 2. A general procedure to assess microbial community
structure with shotgun metagenomic sequencing ...................... 15
Figure 3. Schematic diagram of the study ....................................... 18
Figure 4. Percentage of proteins with no known function against log
(initial nutrient concentration+5) transformed nutrient
concentration in metagenome (A). Percentage of functional genes
not assignable to known genera against log (initial nutrient
concentration+5) transformed nutrient concentration in
metagenome (B). Percentage of unclassified genera against log
(initial nutrient concentration+5) transformed nutrient
concentration in 16S rRNA amplicon data (Song et al., 2016) (C).
A quadratic regression line was applied only to significant results
(p<0.05). ................................................................................... 35
Figure 5. Percentage of proteins with no known function against
incubation time in metagenome (A). Percentage of functional
genes not assignable to known genera against incubation time (B).
Percentage of unclassified genera in the 16s amplicon data
against incubation time (Song et al., 2016) (C). Linear or
quadratic regression line was applied when significant (p<0.05)
xi
(N.B. only the line with the stronger regression is presented
when it fits more than one) ....................................................... 36
Figure 6. Species rarefaction curves with error bars for three
replicates. Sequence reads over 15,000 are not shown here since
the curve already reached saturation point in all samples. ........ 37
Figure 7. Subsystem Level 3 function rarefaction curves with error
bars for three replicates. Sequence reads over 15,000 are not
shown here since the curve already reached saturation point in
all samples................................................................................. 38
Figure 8. nMDS plots of the distance between sample data points
calculated based on the composition of species assignment of
functional genes (A) and functional gene profile (B)................. 39
Figure 9. Heatmap of Subsystem Level 1 genes which show strong
correlation with nutrient gradient. ............................................. 48
Figure 10. Heatmap of Subsystem Level 1 genes which show strong
correlation with incubation time................................................. 60
Figure 11. Richness and Shannon diversity of functional genes
against log (initial nutrient concentration+5) transformed
nutrient concentration. A quadratic regression line was applied
only to significant results (p<0.05) ........................................... 62
Figure 12. Richness and Shannon diversity of functional genes
against incubation time. A quadratic regression line was applied
only to significant results (p<0.05) ........................................... 63
xii
Figure 13. Functional gene richness against richness of the species
assignments of functional genes (A: linear regression applied, B:
quadratic regression applied) and OTU richness (Song et al.
2016) (C: linear regression applied, D: quadratic regression
applied)...................................................................................... 64
Figure 14. Functional gene Shannon diversity against Shannon
diversity of species assignment of functional genes (A) and OTU
Shannon diversity (Song et al. 2015) (B).................................. 65
Figure 15. Beta diversity of functional genes calculated based on
Bray-Curtis dissimilarity between samples against log (initial
nutrient concentration+5) transformed nutrient concentration. A
quadratic regression line was applied only to significant results
(p<0.05) .................................................................................... 66
Figure 16. Beta diversity of functional genes calculated based on
Bray-Curtis dissimilarity between samples against incubation
time. A linear regression line was applied only to significant
results (p<0.05) ........................................................................ 67
Figure 17. Taxonomic annotation of functional genes that have
positive correlation with nutrient concentration. Y-axis shows
the relative abundance of 20 most abundant taxa ...................... 69
Figure 18. Taxonomic annotation of functional genes that have
negative correlation with nutrient concentration. Y-axis shows
the relative abundance of 20 most abundant taxa ...................... 70
xiii
Figure 19. Taxonomic annotation of functional genes that have
positive correlation with incubation time. Y-axis shows the
relative abundance of 20 most abundant taxa ............................ 71
Figure 20. Taxonomic annotation of functional genes that have
negative correlation with incubation time. Y-axis shows the
relative abundance of 20 most abundant taxa ............................ 72
Figure 21. Map of sample collection site ......................................... 96
Figure 22. RDA result showing sample distribution based on bacterial
functional gene composition along the environmental gradient in
Tibet. Vegetation types are indicated, with MAP of alpine
meadow>alpine steppe>desert steppe. MAP: mean annual
precipitation (mm), SM: soil moisture (g/g dried soil), SOC: soil
organic carbon (%), STP: soil total phosphorous (%). ............ 104
Figure 23. Heatmap of standardized relative abundance of Subsystem
Level 1 genes along mean annual precipitation (mm). Gene
categories that have positive/negative/no linear correlation with
mean annual precipitation was denoted by ‘(+)’/‘(-)’/‘(0)’
................................................................................................ 106
Figure 24. Heatmap of standardized relative abundance of Subsystem
Level 1 genes along soil moisture (g/g of dried soil). Gene
categories that have positive/negative/no linear correlation with
soil moisture was denoted by ‘(+)’/‘(-)’/‘(0)’.......... 107
Figure 25. Metagenome bacterial Subsystem Level 3 functional gene
xiv
richness against bacterial OTU richness (A). Metagenome
bacterial Subsystem Level 3 functional gene Shannon diversity
against bacterial OTU Shannon diversity (B). Linear regression
line was applied only when significant. Presented R-squared
value is adjusted R-squared value. ......................................... 115
Figure 26. Metagenome bacterial Subsystem Level 3 functional gene
richness against mean annual precipitation (A). Metagenome
bacterial Subsystem Level 3 functional gene Shannon diversity
against mean annual precipitation (B). Linear regression line was
applied. Presented R-squared value is adjusted R-squared value.
................................................................................................ 118
Figure 27. Bacterial OTU richness against mean annual precipitation
(A). Bacterial OTU Shannon diversity against mean annual
precipitation (B). Results not significant (P>0.05). ................. 119
Figure 28. Metagenome bacterial Subsystem Level 3 functional gene
richness against soil moisture (A). Metagenome bacterial
Subsystem Level 3 functional gene Shannon diversity against soil
moisture (B). ........................................................................... 120
Figure 29. Metagenome fungal Subsystem Level 3 functional gene
Shannon diversity against mean annual precipitation (A).
Metagenome archaeal Subsystem Level 3 functional gene
Shannon diversity against mean annual precipitation (B). Results
not significant (P>0.05)........................................................... 121
xv
Figure 30. Functional beta diversity in each vegetation type
calculated based on Bray-Curtis dissimilarity of Subsystem
Level 3 genes from group centroid. Alphabet denotes posthoc
test result of Tukey’s HSD.................................................... 122
Figure 31. Taxonomic annotation of functional genes that have
negative correlation with precipitation gradients. Y-axis shows
the relative abundance of 20 most abundant taxa .................... 124
Figure 32. Taxonomic annotation of functional genes that have
positive correlation with precipitation gradients. Y-axis shows
the relative abundance of 20 most abundant taxa .................... 126
xvi
List of Tables
Table 1. Chemical analysis of culture material after harvest (mean±
standard error).......................................................................... 33
Table 2. PERMANOVA test result presenting significant effect of
nutrient concentration, incubation time, and the interaction
between two on the composition of species assignment of
functional genes......................................................................... 40
Table 3. PERMANOVA test result presenting significant effect of
nutrient concentration, incubation time, and the interaction
between two on gene composition............................................. 41
Table 4. Subsystem Level 1 and Level 2 genes that have strong
(rho>0.5 or rho<-0.5) correlation with nutrient concentration
when controlling incubation time. Only the selected Subsystem
Level 2 genes that follow the same (whether positive or negative)
pattern with the Subsystem Level 1 category are shown. Partial
correlation results of “Clustering-based subsystem” and
“Miscellaneous” were not included. ....................................... 45
Table 5. Partial correlation result of cell markers suggested in Lauro
et al. (2009) against nutrient controlling incubation time. ......... 49
Table 6. Subsystem Level 1 and Level 2 genes that have strong
(rho>0.5 or rho<-0.5) correlation with incubation time when
controlling nutrient concentration. Only the selected Subsystem
xvii
Level 2 genes that follow the same (whether positive or negative)
pattern with the Subsystem Level 1 category are shown. Partial
correlation results of “Clustering-based subsystem” and
“Miscellaneous” were not included. ....................................... 56
Table 7. Forward selection of environmental variables, which best
explain variation in functional gene composition between samples.
A pseudo-F statistic from Monte Carlo permutation test was
derived in the same way as in ANOVA of the regression model
................................................................................................ 105
Table 8. Subsystem Level 3 genes belonging to stress response
category showing significant correlation with mean annual
precipitation............................................................................. 108
Table 9 Pearson correlation between Subsystem Level 1 genes and
soil total carbon (STC), soil organic carbon (SOC), soil total
phosphorous (STP), soil total nitrogen (STN) in the field study in
Tibetan plateau and Spearman correlation between Subsystem
Level 1 genes and initial nutrient concentration in the microcosm
study (Chapter 2). Significant positive correlations are
highlighted in green and significant negative correlations are
highlighted in red. .................................................................... 111
1
CHAPTER 1: Application of
Metagenomics towards an
Assessment of Biodiversity in
Microbial Ecology: An Introduction
2
1. 1. The impact of advanced sequencing
technology on microbial ecology
1. 1. 1. Discovery of the unseen majority: soil
microorganisms
Soil microbes are one of the most abundant taxonomic group
in terrestrial ecosystem. It has been estimated that microbial
biomass in a hectare of soil is over 1,000 kg∙C in grasslands and
forest sites and around 430 kg∙C in desert biome (Fierer et al.,
2009). These soil microorganisms play significant roles in shaping,
maintaining, or modifying environments surrounding them (Paul,
2014). They are involved in the key steps of nutrient cycles, such
as nitrogen fixation, denitrification, ammonia oxidation,
methanogenesis, carbon fixation, etc. (Fierer, 2017)
Although their role has been acknowledged for more than a
century, there was a limitation in studying microbial ecology in
natural ecosystems mainly because they are invisible to the naked
eyes. Prior to the development of culture-independent molecular
technologies, soil microbes could only be studied in detail by
culturing them under specific laboratory conditions. However, it had
been estimated that only about one percent or less of total soil
microbial organisms are culturable (Amann et al., 1995) thereby
3
limiting the scope of microbial ecology. Also, it is hard to guarantee
whether the observed characteristics in a culture system could also
be found in the natural ecosystem.
With the advent of sequencing technology, it has become
possible to disclose the unseen majority; soil microbes. The
biogeography of soil microbial community has been extensively
studied since early 2000s using next generation sequencing
(Cederlund et al., 2014; Fierer et al., 2007; Griffiths et al., 2011;
Lauber et al., 2009; Nemergut et al., 2011; Oliverio et al., 2017;
Pett-Ridge and Firestone, 2005; Rousk et al., 2010; Singh et al.,
2014; Sul et al., 2013; Tripathi et al., 2013; Tripathi et al., 2014).
Although there might be some exceptions, in most of these studies,
microbial community structure and diversity showed a predictable
pattern along environmental gradients. Among these factors, pH has
been reported as the primary abiotic factor influencing soil
microbial community structure and diversity followed by organic
carbon concentration, soil oxygen and redox potential, soil moisture,
and temperature.(Fierer, 2017; Lauber et al., 2009; Tripathi et al.,
2013).
4
1. 1. 2. A general procedure for soil microbial
community analysis with amplicon sequencing
As shown in Figure 1, analyzing soil microbial community
structure begins with collecting soil samples. Since soil environment
is extremely heterogeneous, it is hard to get a representative
sample of a soil of interest (Kirk et al., 2004; Vos et al., 2013).
There are several ways to avoid this problem, for example, by
sieving soils or by mixing point samples collected from a few
meters apart from each other.
DNA extraction follows after sample collection. In this step,
it is important to get a pure DNA and humic substances that inhibit
PCR amplification should be properly removed (Yeates et al., 1998).
In the PCR amplification step, 16S rRNA gene is generally
used as a target gene in bacterial community analysis. 16S rRNA
gene has several advantages as a phylogenetic marker. 16S rRNA
plays a fundamental role in survival and reproduction of bacterial
cells, so it is highly conserved and present in all bacterial cells.
This facilitates designing of a universal primer which targets the
whole taxa of interest. 16S rRNA also includes several variable
regions of which nucleotide sequence conversion speed generally
parallels that of phylogenetic divergence.
Traditional Sanger sequencing method requires separate
5
steps for sequencing (by applying chain-terminating
dideoxynucleotide), size separation (by electrophoresis), and
detection. Though it can produce long reads, Sanger sequencing is
time consuming and produce only one read per sample. Next
generation sequencing (NGS) methods can produce massive amount
of data in a relatively short time. There are many types of NGS
platforms exist. Iron Torrent uses semi-conductor chip sensing
hydrogen ion released from DNA when a nucleotide is incorporated.
Pyrosequencing machine use chemical reaction producing light
when pyrophosphate is released from DNA during polymerization
reaction. Illumina sequencing platforms detect fluorescent signal
from dye attached to each nucleotide. These various platforms
differ in read length, cost, error rate, throughput, and running time,
so it is important to compare and choose the proper machines which
qualifies. After PCR amplification, adaptors are ligated to each end
of the amplicons for library construction and after library
construction and quantification, we put samples to the sequencing
machine.
After sequencing, there are several steps required for
quality control. Programs such as FastQC, enable users to overview
the sequencing quality by checking metrics which has been
produced by sequencing machine. To begin with sequence
processing, adaptor, primer, and linker sequences should be
6
trimmed away. Also, low-quality sequences (which have
homopolymers or ambiguous sequences or have low quality scores)
and sequences that are too short considering the desired length
should be removed.
Once sequences pass pre-filtering steps, based on
sequence similarity, sequences are grouped into operational
taxonomic unit (OTU). There are two ways to pick OTUs. One is
alignment-based method and the other one is alignment-free
method such as uclust or CD-HIT. If most of the sequences can be
aligned with high accuracy to the alignment database, it is best to
use alignment-based methods. In both methods, chimeric
sequences, which is an artificial sequences generated during PCR
step should be removed.
The last step is microbial community analysis. This step
includes calculation of diversity indices such as Shannon diversity
or Simpson diversity. This step also includes ordination of
community data by calculating ecologically meaningful distances
(e.g. Bray-Curtis dissimilarity) or phylogenetic distances between
samples.
7
Figure 1. A general procedure for soil microbial community analysis
based on amplicon sequencing of microbial marker genes
8
1. 1. 3. Lack of linkage between microbial community
structure and function
With the advent of next generation sequencing technologies,
microbial community data has been accumulated with an unexpected
speed. It is now possible to understand the general pattern of how
microbial communities are assembled and how microbial diversity
correlates with specific environmental conditions. However, most of
the studies until now focused mainly on taxonomic information of
microbes and there exists a gap between the taxonomic information
and ecological functions of soil microorganisms.
Conventional niche theory suggests that no two species can
occupy the same niche in time and space or habitat. In this sense,
soil taxonomic diversity is important to understand overall
ecosystem productivity and resilience. In addition, it is important to
note that high taxonomic diversity is usually associated with high
ecosystem stability especially when there is no redundant function
between two groups of organisms dwelling in the same environment
(Cadotte et al., 2011; Hooper et al., 2002). However, in many cases,
several species can perform the same soil functions (functional
redundancy). Thus, measuring functional diversity could be
important for finding out the connection between biodiversity and
resilience.
9
10
1. 2. An assessment of functional traits of soil
microorganisms
1. 2. 1. Diverse methods to describe functional traits
Several methods have been developed for measuring
functional diversity (Petchey and Gaston, 2006). In most cases,
functional diversities are calculated based on collective traits
(Nunes et al., 2017; Petchey and Gaston, 2002, 2006). When
assessing functional diversity with functional or physiological traits,
it is important to choose the type of traits to be considered
(Petchey and Gaston, 2006). In the case of plants, life-cycle,
growth-form (e.g. shrub, bulb, and rosette), leaf type (e.g.
evergreen or semi-deciduous), seed mass, leaf N, fruit type, etc.
has been used (Flynn et al., 2009; Nunes et al., 2017; Petchey and
Gaston, 2002). In case of birds, weight, feeding guild, food type,
foraging location, foraging height, foraging habitat, etc. has been
used (Flynn et al., 2009; Petchey and Gaston, 2002; Sekercioglu,
2012). Similar to birds, in mammals, weight, feeding guild, food type,
activity, nesting, etc. has been used (Flynn et al., 2009; Petchey
and Gaston, 2002).
In soil microbiology, other types of methods have been
developed for measuring functional traits diversity for example,
11
assessment of catabolic potential of microbes (Degens and Harris,
1997; Torsvik and Øvreås, 2002). This type of method has been
developed mainly in culture system, by growing bacterial groups
with different type of nutrient supplied. However, these type of
analysis only focus on one subcategory of bacterial functions. There
should be a generalized methods to assess overall functions that
bacterial groups perform as one of the key member in the
ecosystem. Metagenomic profiling through shotgun sequencing can
provide a quantitative data of overall functional profile of microbes.
1. 2. 2. Metagenomic approach for assessing microbial
functional traits
The fundamental driving force of the functions that microbes
perform are the coded sequences in their DNA. In this sense, it is
natural to examine soil microbial functional diversity with functional
gene diversity. This type of approach has numerous advantages in
that the data produced from more than a million sequences are
highly quantitative and it makes it easy to compare results of
different studies.
Recent studies has started to apply shotgun metagenomic
sequencing to assess functional diversity. Burke et al. (2011)
studied algal-associated microbial communities using
12
metagenomics. They found that there was a high similarity (70%) in
functional gene composition between samples despite low similarity
(15%) in microbial species composition. Fierer et al. (2012) applied
metagenomics to assess taxonomic microbial community structure
and functional attributes of microorganisms in several different
biome. They found that several gene categories associated with
different type of biomes, for example, nutrient cycling related genes
were more abundant in mesic systems compared to desert systems.
Mendes et al. (2015) assessed functional diversity in relation to
disturbance using metagenomics and found that agricultural soil had
higher taxonomic and functional diversity than forest soil.
Figure 2 shows the basic procedure for assessing functional
attributes of microbes through shotgun metagenomic sequencing.
As in amplicon sequencing method described in Figure 1, it starts
with soil collection and DNA extraction. Unlike amplicon sequencing,
shotgun metagenomics doesn’t include amplification of marker gene.
Instead, the whole DNA extracted from soil are fragmented into
small pieces and sticky ends of DNAs are repaired for adaptor
ligation. Also, proper size of DNA should be selected in library
construction step.
After library construction and sequencing, DNA sequences
undergoes pre-filtering process as in amplicon sequencing. Using
programs such as FragGeneScan (specified for prokaryotic gene
13
calling), coding regions within the sequences are predicted. Before
annotation, proteins with high sequence similarity is clustered using
alignment-free clustering methods and a representative sequences
are BLAST against gene database for annotation.
There are several databases available for functional
annotation of genes such as GenBank (Benson et al., 2012), Cluster
of Orthologous Group (COG) (Tatusov et al., 2003), Kyoto
Encyclopedia of Genes and Genomes (KEGG) (Kanehisa, 2002), or
SEED (Overbeek et al., 2005). GenBank contains all publically
available DNA sequences and their product. COG gene database has
been constructed by finding orthologous groups in sets of complete
genome. KEGG database groups functional genes by cellular
pathways. SEED database use hierarchical way to organize gene
families into several levels of functions. These databases collect
functions of genes published based on experimental observation.
Functional genomics is the field in molecular biology which attempts
to describe gene functions and interactions. There are several ways
to prove gene functions and one example is introducing mutation to
find loss of function.
Once we get functional profile of genes through annotation,
we can calculate functional gene diversity. Besides, we can ordinate
samples by calculating dissimilarity between samples. Also, we can
regress functional categories against environmental factors to find
14
out relationships.
15
Figure 2. A general procedure to assess microbial community
structure with shotgun metagenomic sequencing
16
1. 3. Research objectives
One of the major aims of ecology is to be able to
characterize the adaptations which determine where and when each
type of organism occurs in nature (Ricklefs and Miller, 1999).
Certain combinations of features or behaviors occur predictably in
the context of certain environments and niches, and are known as
strategies (Grime, 1977; Southwood, 1977; Westoby et al., 2002).
Up until now, most of the studies on ecological strategies have
focused on macroorganisms such as plants or animals. There are
only few studies which studied microbial functions in relation to
their survival or reproductive strategies in association with
environmental conditions.
This study aims to explore the fundamental aspects of how
microbial community processes and function vary along
environmental gradients. We set hypotheses constructed based on
the studies of macrooganisms and aim to see whether the same
patterns are universally found in microbial communities. We expect
there would be specific functional gene categories correlating with
certain environmental gradients as a result of natural selection.
In the first study (Chapter 2), we constructed a laboratory
microcosm by culturing soil derived microorganisms, to study the
effects of varying initial nutrient concentration, and time succession,
17
on the community metagenome. In the second study (Chapter 3),
we set up a field study in Tibetan Plateau to understand how aridity
stress affect functional diversity of genes in the soil microbial
community.
18
Figure 3. Schematic diagram of the study
19
CHAPTER 2: Bacterial Strategies
along Nutrient and Time Gradients,
Revealed by Metagenomic Analysis of
Laboratory Microcosms
20
2. 1. Introduction
In plant and animal ecology, among the best characterized
and most widely recognized ecological strategies are the ‘r’ and ‘K’
strategies (MacArthur and Wilson, 1967). r strategists are
described as living in temporarily resource-rich environments, with
the potential to grow and mature rapidly, and produce large
numbers of young quickly, with poor ability to withstand sustained
competition or predation (Pianka, 1970). K strategists are
described as having the opposite of these characteristics, being
adapted to environments where resource supply rates are lower,
population densities are higher and competition between individuals
and between species is intense. r and K strategists have often been
described in relation to ecological succession, with r strategists
being abundant in early successional stages and K strategists
prevailing in late successional stages (Odum, 1969).
In microbial ecology, there has been little explicit discussion
of r or K strategies. Instead, attention has concentrated on the two
opposing microbial strategies of ‘oligotrophy’ (specialization to low
nutrient conditions) and ‘copiotrophy’ (specialization to high
nutrient conditions) (Koch, 2001). There has only been one
concerted attempt to define the differences between these two
strategies at the inferred functional level of cell physiology and
21
genome functions (Lauro et al., 2009). Lauro et al.’s study (2009),
while a valuable step forward, the inspiration for the present study,
involved selected examples of fully sequenced marine bacterial
species, and leaves open the need for broader testing in other
environments, and using other selection criteria.
There has been little discussion in microbial ecology of the
possible similarities and differences between the copiotrophy-
oligotrophy axis and the r – K axis of strategies. It seems, however,
that copiotrophy-oligotrophy is thought of in a static sense of
steady conditions of greater or lesser resource supply (since
population growth rate and stability is not generally discussed in the
context of this axis). The r-K axis as generally discussed in
ecology is considered as part of a time succession in which a
community develops from early conditions of sparse populations
and ready availability of resources, to later conditions of sustained
crowding and intense competition for resources.
Here we attempted an approach, which may help to elucidate
the functional strategies, which exist in the bacterial world.
Whereas field-based observations can accurately incorporate the
true complexity of nature, experimental microcosms can go some
way towards reducing the huge number of environmental factors
that vary simultaneously, and thus give further useful perspectives
(Jiang and Patel, 2008; Luckinbill, 1979). It is true that in a
22
culture-based system, there is the inevitable limitation that only a
subset of the species present in nature will grow in the system, and
that it can never precisely replicate the complexity of the natural
world. Nevertheless, the precise control of initial nutrient content,
temperature, etc. in culture systems may be useful in elucidating
general patterns that might easily be hidden within the vast
complexity of natural ecological systems.
Here, we tested for the existence of different strategies in
relation to nutrient and incubation time by combining precisely
controlled conditions, with an assessment of combined genome
capabilities. In this study, the distinctions ‘copiotrophy vs
oligotrophy’, and ‘r-strategy vs k-strategy’ have been used as a
relative form. For example, a function related to copiotrophy has
been defined as a functional gene for which relative abundances
increase along with the increase of initial nutrient concentration. We
predicted that the taxonomic assignment of functional genes and
functional gene structure of the community would differ in relation
to initial nutrient concentration and incubation time.
We tested the following predictions in relation to ecological
strategies:
A) What are the functional characteristics of oliogotrophy vs
copiotrophy amongst bacteria?
23
Hypothesis A-1) Cell division and cell cycle related genes will be
more abundant under copiotrophic conditions. It is clear that
amongst plants and corals, those growing in conditions of higher
rates of nutrient supply (and light supply, for photosynthesizers)
are innately able to grow faster and have higher potential rates of
tissue growth and extension, and more rapid life cycle completion to
reproduction (Grime, 1979; Huston, 1994). We hypothesized that,
as with larger organisms, at high nutrient concentrations bacterial
cells are able to acquire enough nutrient for their basic metabolism,
and will use supplementary nutrients for cellular
growth/reproduction.
Hypothesis A-2) Under copiotrophic conditions, genes that are
themselves related to promoting gene expression will be relatively
more abundant. In this case, we based our prediction on the
microbial literature in relation to copiotrophy. For example, rRNA
gene copy number has commonly been used as an indicator of
copiotrophy (Condon et al., 1995; Fegatella et al., 1998;
Klappenbach et al., 2000; Lauro et al., 2009). Our intention here
was to test this hypothesis under the more precisely controlled
conditions of a nutrient gradient in a culture system.
Hypothesis A-3) Nutrient accumulator genes will be more abundant
under oligotrophic conditions. Lower nutrient conditions will require
greater numbers of genes associated with accumulating nutrients.
24
Stress tolerator plants and lichens that grow under nutrient-poor
conditions are known to be very effective accumulators of nutrients
(Grime, 1979), and by extension we predicted empirical gene
function evidence for similar behavior by bacteria.
Hypothesis A-4) Dormancy and sporulation related genes will be
relatively more abundant in copiotrophic conditions. Here we were
testing for a functional gene abundance difference already predicted
by Fierer et al. (2007). Since copiotrophs cannot grow in nutrient
poor conditions, they are expected to survive in a dormant stage
until conditions improve (Fierer et al., 2007).
Hypothesis A-5) Functional gene abundance patterns will parallel
those found by Lauro et al. (2009). Using a very different approach,
based on classification of fully sequenced selected bacterial
genomes Lauro et al. (2009) compared species defined as either
oliogotrophic or copiotrophic strategists. Using this approach, they
arrived at an assemblage of functional genes, which may define the
copiotrophic strategy. We aimed to test the generality of the pattern
found by Lauro et al. (2009), by using a different approach – a
nutrient gradient in a controlled culture system.
B) What are the functional characteristics of r vs K selected
bacteria?
Hypothesis B-1) Cell division related genes will be more abundant
25
in the early successional stage. Cell division is an aspect of growth,
particularly in the prokaryotic world. Typically, in the ecology of
larger organisms, early successional species grow fast and give
many offspring, rapidly expanding their biomass (Bazzaz, 1979;
Connell and Slatyer, 1977; Odum, 1969). Thus, we expected this
pattern to hold true in bacterial succession, revealing itself in
greater relative abundance of genes related to rapid
growth/reproduction.
Hypothesis B-2) Genes related to cell-cell interactions will be
more abundant in the later successional stages. It is a generally
agreed concept in ecology that later successional ecosystems have
more intense and species-specific mutualistic and antagonistic
interactions (Morriën et al., 2017; Odum, 1969). We predicted
similar trends towards intensity of (either positive or negative)
organismic interactions in the late successional stage in the soil
bacterial systems we were studying.
Hypothesis B-3) Genes related to cell motility will be more
abundant in the early successional stage. The ‘r’ selected plant
species that thrive in early successional stages generally have high
seed dispersal ability (Huston and Smith, 1987; Peroni, 1994). Cell
motility, which functionally fulfils the same role as the seed
dispersal of ‘r’ selected plants, is important for a bacterial
population to reach newly created open patches or spaces of
26
environment that are relatively free of competition. Thus, genes for
cell motility will be more abundant in early successional microbes.
C) How does functional gene diversity relate to nutrient
concentration and incubation time?
Hypothesis C: Functional gene diversity will show a humpback
curve against initial nutrient concentration or incubation time,
resembling the pattern found in OTU diversity in the same
experimental system (Song et al., 2016).
In an earlier study on this same experimental system, Song et al.
(2016) found a ‘humpbacked’ pattern of bacterial OTU diversity in
relation to nutrient concentration, and time. We anticipated that
functional gene diversity would parallel OTU diversity, because
different taxa have different genetically based capabilities, and
coexistence of different taxa in itself will require separate
ecological niches, which are achieved partly by a range of different
cellular and metabolic functions. This taps into a wider debate of
whether functional diversity at the community level is necessarily a
correlate of taxonomic diversity (Griffiths and Philippot, 2013;
Huston, 1997; McCann, 2000; Tilman et al., 1997).
27
2. 2. Materials and Methods
Sample preparation
As described in Song et al. (2016), plates with different
nutrient concentrations were prepared. Tryptic Soy Broth was used
as a nutrient source, and agar was used as a solidifying agent. Agar
culture medium was prepared with original (nutrient A group), 10-
fold diluted (nutrient B group), 100-fold diluted (nutrient C group),
and 10,000-fold diluted (nutrient E group) Triptic Soy Broth (Difco)
concentrations. In the medium, cycloheximide (100㎍/ml) was
added to inhibit fungal growth. Sterile cellophane film was overlaid
on each culture plate to facilitate cell harvest.
Garden soil (0-5 c.m. depth) was collected from an
overgrown flowerbed on the campus of Seoul National University
(Song et al., 2016). 20 grams of soil was eluted to 190ml of 0.85%
NaCl. The sample underwent 10 min of blending followed by 30 min
of incubation in a shaking incubator at 24 ℃ at the speed of 250
r.p.m. We diluted soil solution 10-2 times, which was the
concentration that had the optimal density of derived colonies (Song
et al., 2016). We let soil particles settle for 6 hours before plating
while the water was kept chilled to 4℃ to prevent community
change. We carefully transferred supernatant into a new tube and
100 ㎕ of supernatant was spread on each prepared culture plate.
28
Plates were incubated at 25℃ for different lengths of time (Day 2,
7, 28, 56, and 84) before final sampling. Plates that showed any
fungal colonies on examination under a low power light microscope
were discarded (absence of fungi in the remaining sampled plates
was later confirmed from the analysis of metagenomes, see below,
which consist overwhelmingly of bacterial DNA). 5 plates were
pooled together and treated as one replicate. A total of 300 plates
(4 nutrient concentrations * 5 incubation time * 3 replicates for
each * 5 plates for each replicate) were used for further analysis.
DNA from the final samples was extracted using Genomic DNA
purification kit (Promega, Madison, Wisconsin, USA) following the
manufacturer’s guidelines. Extracted DNA was stored at -20℃ for
further analysis.
Sequencing and sequence processing
DNA samples were sent to Celemics (Celemics, Inc., Seoul,
Korea) for shotgun metagenomic sequencing using Illumina
HiSeq2500 sequencing platform. Sequences were uploaded to the
Metagenomics Rapid Annotation using Subsystem Technology
(MG-RAST) online server version 4.0 for taxonomic and functional
annotation (Meyer et al., 2008) under the project name “Effect of
nutrient concentration and successional stage on structure and
29
function of culturable bacteria” (project ID, 16343). The Refseq
database (Pruitt et al., 2007) was chosen for taxonomic annotation
of functional genes and the SEED database (Overbeek et al., 2005)
and the Clusters of Orthologous Groups (COG) database (Tatusov
et al., 2003) were used for functional annotation. Sequences were
annotated with default setting (maximum e-value cutoff of 10-5,
minimum percentage identity cutoff of 60%, and minimum alignment
length cutoff as 15 bp). Even though more than 99.5% of assigned
sequence reads were from bacteria, we filtered out non-bacterial
sequences for the statistical analysis.
Chemical analysis of culture material
Dissolved organic carbon (DOC) and dissolved total nitrogen
(DTN) left in the culture material after harvest was measured to
check how the nutrient treatment changed over time. Nutrient A and
C group, incubation times 2, 28 and 84 weeks were selected for the
analysis. Culture material, which had been kept after harvest at -20℃
was melted at room temperature and sent to NICEM (Korea) for
chemical analysis. Samples were filtered with hydrophilic
polyethersulfone filter and hydrophilic polypropylene filter
(Supor®-450 47mm 0.45㎛). DOC was measured using a total
organic carbon analyzer (TOC-L, SHIMADZU, Japan) and DTN was
30
measured with an AutoAnalyzer (SYNCA, BLTEC, Japan) following
standard methods provided by the American Public Health
Association (APHA), the American Water Works Association
(AWWA), and the Water Environment Federation (WEF) (Rice et
al., 2012).
Statistical analysis
To calculate taxonomic diversity, species level assignments
of functional genes from the metagenomes were used, and for
functional diversity, Subsystem Level 3 data were used. To
normalize the number of sequences in each sample, sequences were
subsampled based on the number of least abundant sequence reads
among all the samples (1,749,521 for species level assignment of
functional genes and 530,466 for Subsystem Level 3 function).
Rarefaction curves were drawn with subsampled sequence data. To
calculate taxonomic and functional distance between each sample,
the number of metagenome sequence assigned at species level and
Subsystem Level 3 in each group was square root transformed and
Bray-Curtis dissimilarity was calculated. To test the influence of
initial nutrient concentration and incubation time to the bacterial
community structure, permutational multivariate analysis of
variance (PERMANOVA) test were performed with 999
31
permutation using ‘adonis’ function in R package ‘vegan’.
To study the gene abundance differences of bacteria in
relation to initial nutrient concentration and incubation time, partial
Spearman correlation of initial nutrient concentration (or incubation
time) and relative abundance of Subsystem Level 1 and Level 2
functional gene categories, controlling for incubation time (or initial
nutrient concentration, and vice versa) was calculated using
‘pcor.test’ in R software ‘ppcor’ package. Partial correlation was
used since we aimed to discern the nutrient effect and time effect
independently and Spearman rank correlation was used since it
does not depend on normality or linearity. Relative abundance of
genes in each sample was calculated dividing the number of
sequences assigned to each category with the sum of sequences
successfully annotated to the gene database.
32
2. 3. Result
Chemical analysis of culture material
As shown in Table 1, dissolved organic carbon (DOC)
decreased over time, in both the nutrient A and nutrient C
treatments, which we analysed. However, the DOC of nutrient C
treatment never exceeded that of nutrient A, revealing the
continuing legacy of initial conditions in terms of nutrient loading of
the culture medium. Dissolved total nitrogen (DTN) decreased over
time in nutrient C group, but not in the nutrient A group. The scale
of the difference between nutrient A and C group was much greater
in the case of DTN.
33
Table 1. Chemical analysis of culture material after harvest (mean±standard error)
Incubation time
(days)
Initial nutrient
concentration
Dissolved organic carbon
(mg/L)
Dissolved total nitrogen
(mg/L)
2 A 6869.3±866.0 1838.9±86.8
2 C 2838.0±206.9 25.9±0.8
28 A 6627.3±606.1 2184.7±77.8
28 C 2194.3±14.4 16.4±0.8
84 A 5350.7±1015.5 2214.1±187.1
84 C 1108.9±79.4 11.2±0.0
34
Taxonomic and functional composition of communities
The total number of sequences per sample obtained after
quality control ranged from 1,962,970 to 4,641,543. The proportion
of proteins with no known function and functional genes not
assignable to known genera does not increase with increasing
nutrient concentration (Figure 4-A and B), but increases with
incubation time (Figure 5-A and B). 16S amplicon based results
show a similar pattern, unclassified genera being more abundant in
later successional stage (Figure 4C and Figure 5C). The rarefaction
curve of species level assignment of functional genes and
Subsystem Level 3 function had reached its saturation point in all
samples (Figure 6-7). The composition of species level assignment
of functional genes and functional profile showed a predictable
pattern with progression in either of two distinct directions related
to nutrient gradient and incubation time (Figure 8). The
permutational multivariate analysis of variance (PERMANOVA)
result shows that nutrient concentration, incubation time, and the
interaction between nutrient concentration and incubation time all
had significant influence in relation to community structure, both in
terms of the composition of species level assignment of functional
genes and the functional gene composition (Table 2 and Table 3).
35
Figure 4. Percentage of proteins with no known function against log
(initial nutrient concentration+5) transformed nutrient concentration
in metagenome (A). Percentage of functional genes not assignable to
known genera against log (initial nutrient concentration+5)
transformed nutrient concentration in metagenome (B). Percentage
of unclassified genera against log (initial nutrient concentration+5)
transformed nutrient concentration in 16S rRNA amplicon data (Song
et al., 2016) (C). A quadratic regression line was applied only to
significant results (p<0.05).
36
Figure 5. Percentage of proteins with no known function against
incubation time in metagenome (A). Percentage of functional genes
not assignable to known genera against incubation time (B).
Percentage of unclassified genera in the 16s amplicon data against
incubation time (Song et al., 2016) (C). Linear or quadratic
regression line was applied when significant (p<0.05) (N.B. only the
line with the stronger regression is presented when it fits more than
one)
37
Figure 6. Species rarefaction curves with error bars for three replicates. Sequence reads over 15,000 are
not shown here since the curve already reached saturation point in all samples.
38
Figure 7. Subsystem Level 3 function rarefaction curves with error bars for three replicates. Sequence
reads over 15,000 are not shown here since the curve already reached saturation point in all samples
39
Figure 8. nMDS plots of the distance between sample data points calculated based on the composition of
species assignment of functional genes (A) and functional gene profile (B).
40
Table 2. PERMANOVA test result presenting significant effect of nutrient concentration, incubation time,
and the interaction between two on the composition of species assignment of functional genes
d.f MS F. model R2 p-value
Day (incubation time) 4 0.603 25.40 0.267 0.001
Nutrient concentration 3 1.373 57.87 0.457 0.001
Day (incubation time) * Nutrient concentration 12 0.128 5.40 0.171 0.001
Residuals 40 0.024 0.105
Total 59 1
41
Table 3. PERMANOVA test result presenting significant effect of nutrient concentration, incubation time,
and the interaction between two on gene composition
d.f MS F. model R2 p-value
Day (incubation time) 4 0.051 19.30 0.267 0.001
Nutrient concentration 3 0.101 38.38 0.398 0.001
Day (incubation time) * Nutrient concentration 12 0.013 4.78 0.198 0.001
Residuals 40 0.003 0.138
Total 59 1
42
Functional gene categories correlating with initial
nutrient concentration
Table 4 shows the Subsystem Level 1 and Level 2 genes
that have strong (rho>0.5 or rho<-0.5) correlation with nutrient
concentration when controlling for incubation time. Only the
selected Subsystem Level 2 genes that follow the same (whether
positive or negative) pattern with the Subsystem Level 1 category
are shown. Partial correlation results of “Clustering-based
subsystem” and “Miscellaneous” were not included in the result
section and were not discussed because most of them are putative
genes, or genes for which functions are unclear. For Subsystem
Level 1 genes, we also drew a heatmap with Z-score transformed
relative abundance data (Figure 9).
As in Table 4, cell division and cell cycle related genes were
more abundant at high nutrient concentrations. Also, genes related
to RNA metabolism, especially RNA processing and modification
related genes, were relatively more abundant at higher nutrient
concentrations. Dormancy and sporulation related genes, especially,
spore DNA protection related genes, were more abundant in
copiotrophic conditions. Many different kinds of amino acid related
genes such as genes related to arginine, urea cycle, polyamines,
glutamine, glutamate, aspartate, asparagine, lysine, threonine,
43
methionine, and cysteine were also relatively more abundant at
higher nutrient concentrations.
By contrast, genes related to energy acquisition (e.g.
carbohydrate) or usage (e.g. respiration) were abundant in the low
nutrient treatments. Carbohydrate metabolism related genes had
negative correlations with initial nutrient concentration. This was
not restricted to one type of carbohydrate, but a broad range of
molecules including amino sugars, di- and oligosaccharides,
monosaccharides, and polysaccharides. Genes related to protein
metabolism, respiration and photosynthesis (although
photosynthesis itself would not have been directly selected for, as
the cultures were kept in the dark) which include genes related to
electron transfer reaction were also abundant in oligotrophic
condition. Having high abundance of genes related to virulence,
disease, interference competition and defense, especially the genes
providing resistance to antibiotics and toxic compounds, was also
one of the noticeable characteristics of more the oligotrophic/low
nutrient treatments.
We also attempted to annotate sequences based on the
COG gene database to compare our results with those of Lauro et al.
(2009) (Table 5). Note that COG2852 gene was not found in our
metagenomes (this gene was also minimally present in Lauro et al’s
(2009) result), and were thus excluded. When we set the rho cutoff
44
value higher than 0.5 or lower than -0.5, 28% (9 out of 32 genes)
of genes in our study matched the result found in Lauro et al.
(2009). The other genes did not show a strong correlation with
initial nutrient concentration, except for COG3710, which was the
only gene that showed the opposite trend from the pattern found by
Lauro et al. (2009).
45
Table 4. Subsystem Level 1 and Level 2 genes that have strong (rho>0.5 or rho<-0.5) correlation with
nutrient concentration when controlling incubation time. Only the selected Subsystem Level 2 genes that
follow the same (whether positive or negative) pattern with the Subsystem Level 1 category are shown.
Partial correlation results of “Clustering-based subsystem” and “Miscellaneous” were not included.
Subsystem Level 1 Subsystem Level 2
Averaged
relative
abundance
rhoTrophic
strategy
Amino Acids and Derivatives - 0.1015 0.803
Copiotroph
Arginine; urea cycle, polyamines 0.0174 0.794
Glutamine, glutamate, aspartate, asparagine;
ammonia assimilation0.0099 0.638
Lysine, threonine, methionine, and cysteine 0.0251 0.658
Cell Division and Cell Cycle - 0.0082 0.696
RNA Metabolism - 0.0326 0.619
RNA processing and modification 0.0237 0.648
46
Table 4. (Continued)
Subsystem Level 1 Subsystem Level 2Averaged relative
abundancerho
Trophic
strategy
Dormancy and Sporulation - 0.0029 0.593Copiotroph
Spore DNA protection 0.0001 0.800
Respiration - 0.0368 -0.828
Oligotroph
Electron accepting reactions 0.0105 -0.537
Electron donating reactions 0.0120 -0.846
Virulence, Disease and Defense - 0.0381 -0.795
Resistance to antibiotics and
toxic compounds0.0331 -0.798
Photosynthesis - 0.0003 -0.785
Electron transport and
photophosphorylation0.0000 -0.660
- 0.1248 -0.672
Aminosugars 0.0032 -0.588
Di- and oligosaccharides 0.0113 -0.667
Monosaccharides 0.0225 -0.683
47
Table 4. (Continued)
Subsystem Level 1 Subsystem Level 2Averaged relative
abundancerho
Trophic
strategy
Carbohydrates Polysaccharides 0.0014 -0.830
Oligotroph
Protein Metabolism - 0.0633 -0.552
Protein folding 0.0065 -0.581
Protein processing and
modification0.0051 -0.658
48
Figure 9. Heatmap of Subsystem Level 1 genes which show strong correlation with nutrient gradient.
49
Table 5. Partial correlation result of cell markers suggested in Lauro et al. (2009) against nutrient
controlling incubation time.
Marker rho p-value Average relative
abundance (%)
Trophic
strategy
Trophic
strategy
(reference)
Cell motility -0.33 0.01 0.326 - Copiotroph
Signal transduction mechanisms -0.27 0.04 4.070 - Copiotroph
Defense mechanisms -0.03 0.84 2.397 - Copiotroph
Transcription 0.84 0.00 4.663 Copiotroph Copiotroph
Secondary metabolites biosynthesis,
transport and catabolism
-0.44 0.00 2.924 - Oligotroph
Lipid transport and metabolism 0.07 0.62 4.722 - Oligotroph
COG0110 (Acetyltransferase
(isoleucine patch superfamily))
0.43 0.00 0.045 - Copiotroph
COG0183 (Acetyl-CoA
acetyltransferase)
-0.28 0.03 0.366 - Oligotroph
50
Table 5. (Continued)
Marker rho p-value Average
relative
abundance (%)
Trophic
strategy
Trophic
strategy
(reference)
COG0243 (Anaerobic dehydrogenases,
typically selenocysteine-containing)
0.15 0.27 0.271 - Copiotroph
COG0318 (Acyl-CoA synthetases
(AMP-forming)/AMP-acid ligases II)
-0.34 0.01 0466 - Oligotroph
COG0483 (Archaeal fructose-1,6-
bisphosphatase and related enzymes of
inositol monophosphatase family)
-0.47 0.00 0.106 - Oligotroph
COG0583 (Transcriptional regulator) 0.61 0.00 1.366 Copiotroph Copiotroph
COG0596 (Predicted hydrolases or
acyltransferases (alpha/beta hydrolase
superfamily))
-0.81 0.00 0.567 Oligotroph Oligotroph
COG0625 (Glutathione S-transferase) -0.59 0.00 0.323 Oligotroph Oligotroph
51
Table 5. (Continued)
Marker rho p-value Average
relative
abundance (%)
Trophic
strategy
Trophic
strategy
(reference)
COG0737 (5'-nucleotidase/2',3'-
cyclic phosphodiesterase and related
esterases)
-0.11 0.42 0.041 - Copiotroph
COG0814 (Amino acid permeases) 0.30 0.02 0.004 - Copiotroph
COG1024 (Enoyl-CoA
hydratase/carnithine racemase)
-0.3` 0.00 0.390 - Oligotroph
COG1028 (Dehydrogenases with
different specificities (related to
short-chain alcohol dehydrogenases))
-0.84 0.00 1.140 Oligotroph Oligotroph
COG1228 (Imidazolonepropionase and
related amidohydrolases)
-0.69 0.00 0.225 Oligotroph Oligotroph
52
Table 5. (Continued)
Marker rho p-value Average
relative
abundance (%)
Trophic
strategy
Trophic
strategy
(reference)
COG1263 (Phosphotransferase
system IIC components,
glucose/maltose/N-
acetylglucosamine-specific)
-0.29 0.03 0.031 - Copiotroph
COG1680 (Beta-lactamase class C
and other penicillin binding proteins)
-0.25 0.06 0.148 - Oligotroph
COG1804 (Predicted acyl-CoA
transferases/carnitine dehydratase)
-0.48 0.00 0.232 Oligotroph Oligotroph
COG1960 (Acyl-CoA
dehydrogenases)
0.23 0.09 0.944 - Oligotroph
COG2124 (Cytochrome P450) -0.66 0.00 0.051 Oligotroph Oligotroph
COG2200 (FOG: EAL domain) 0.13 0.31 0.117 - Copiotroph
53
Table 5. (Continued)
Marker rho p-value Average
relative
abundance (%)
Trophic
strategy
Trophic
strategy
(reference)
COG2207 (AraC-type DNA-binding
domain-containing proteins)
0.30 0.02 0.348 - Copiotroph
COG3293 (transposase and
inactivated derivatives)
0.32 0.01 0.007 - Oligotroph
COG3325 (Chitinase) -0.01 0.96 0.002 - Copiotroph
COG3386 (Gluconolactonase) -0.41 0.00 0.046 - Oligotroph
COG3710 (DNA-binding winged-
HTH domains)
-0.65 0.00 0.017 Oligotroph Copiotroph
COG3773 (Cell wall hydrolyses
involved in spore germination)
-0.46 0.00 0.028 - Oligotroph
COG3920 (Signal transduction
histidine kinase)
-0.63 0.00 0.088 Oligotroph Oligotroph
54
Functional gene categories correlating with incubation
time
Table 6 shows the Subsystem Level 1 and Level 2 genes
that have strong (rho>0.5 or rho<-0.5) correlation with incubation
time when controlling nutrient concentration. Only the selected
Subsystem Level 2 genes that follow the same (whether positive or
negative) pattern with the Subsystem Level 1 category are shown.
Partial correlation results of “Clustering-based subsystem” and
“Miscellaneous” were not included. For Subsystem Level 1 genes,
we also drew a heatmap with Z-score transformed relative
abundance data (Figure 10).
The regulation and cell signaling gene category was more
abundant in the early successional stage. In this gene category,
programmed cell death and toxin-antitoxin systems, quorum
sensing and biofilm formation related genes were abundant in the
early successional stage. The relative abundances of genes
associated with cell division and cell cycle decreased over time, but
the correlation (rho=-0.407, p=0.001) was weaker than for other
genes presented in Table 6. The motility related genes were more
abundant in late successional stages.
There were other genes that showed a significant
correlations which had not been predicted in our hypotheses. These
included genes related to sulfur metabolism, iron acquisition and
55
metabolism, potassium metabolism, metabolism of aromatic
compounds, amino acids and derivatives, and cell wall and capsule
for the early successional stage. For the late successional stage,
photosynthesis, protein metabolism, respiration, nucleosides and
nucleotides, and DNA metabolism were relatively more abundant.
56
Table 6. Subsystem Level 1 and Level 2 genes that have strong (rho>0.5 or rho<-0.5) correlation with
incubation time when controlling nutrient concentration. Only the selected Subsystem Level 2 genes that
follow the same (whether positive or negative) pattern with the Subsystem Level 1 category are shown.
Partial correlation results of “Clustering-based subsystem” and “Miscellaneous” were not included.
Subsystem Level 1 Subsystem Level 2
Averaged
relative
abundance
rho
Density
dependent
strategy
Sulfur Metabolism - 0.0171 -0.814
r-strategy
Inorganic sulfur assimilation 0.0043 -0.683
Organic sulfur assimilation 0.0090 -0.771
Iron acquisition and metabolism - 0.0268 -0.780
Siderophores 0.0053 -0.800
Regulation and Cell signaling - 0.0151 -0.773
Programmed Cell Death and
Toxin-antitoxin Systems0.0011 -0.610
Subsystem Level 1 Subsystem Level 2 Averaged rho Density
57
Table 6. (Continued)
relative
abundance
dependent
strategy
Regulation and Cell signalingQuorum sensing and biofilm
formation0.0009 -0.751
r-strategy
Potassium metabolism - 0.0095 -0.730
Metabolism of Aromatic
Compounds- 0.0236 -0.694
Metabolism of central aromatic
intermediates0.0078 -0.550
Peripheral pathways for
catabolism of aromatic compounds0.0104 -0.810
Amino Acids and Derivatives - 0.1015 -0.605
Arginine; urea cycle, polyamines 0.0174 -0.515
Aromatic amino acids and
derivatives0.0139 -0.725
58
Table 6. (Continued)
Subsystem Level 1 Subsystem Level 2
Averaged
relative
abundance
rho
Density
dependent
strategy
Amino Acids and
Derivatives
Glutamine, glutamate, aspartate,
asparagine; ammonia assimilation0.0099 -0.601
r-strategyProline and 4-hydroxyproline 0.0048 -0.843
Cell Wall and Capsule - 0.0367 -0.504
Gram-Negative cell wall components 0.0114 -0.616
Photosynthesis - 0.0003 0.821
K-strategy
Protein Metabolism - 0.0633 0.660
Protein biosynthesis 0.0386 0.544
Protein degradation 0.0110 0.848
Motility and Chemotaxis - 0.0150 0.640
Flagellar motility in Prokaryota0.0106 0.702
Respiration - 0.0368 0.592
Electron donating reactions 0.0120 0.654
59
Table 6. (Continued)
Subsystem Level 1 Subsystem Level 2
Averaged
relative
abundance
rho
Density
dependent
strategy
Nucleosides and
Nucleotides- 0.0300 0.590
K-strategyDNA Metabolism - 0.0359 0.559
DNA replication 0.0087 0.812
60
Figure 10. Heatmap of Subsystem Level 1 genes which show strong correlation with incubation time
61
Taxonomic and functional diversity
Alpha diversity. In contrast to our expectation, neither functional
gene richness nor function gene diversity show a humpback pattern
against initial nutrient concentration (Figure 11). However,
functional gene richness showed a humpback pattern against
incubation time. (Figure 12). In terms of richness, functional gene
richness had a positive correlation with OTU richness (Song et al.,
2016) or the richness of species level assignment of functional
genes from metagenome (Figure 13). In the case of Shannon
diversity which indicates species evenness/relative abundances, as
opposed to richness, functional gene diversity had no correlation
with OTU diversity nor the diversity of species assignment of
functional genes (Figure 14).
Beta diversity. Beta diversity of functional genes showed a hump-
backed curve against initial nutrient concentration (Figure 15).
However, beta diversity of functional genes showed not so much
patterns against incubation time (Figure 16). Mantel test result
shows that the Bray-Curtis dissimilarity between samples
calculated based on taxonomic annotation of functional genes shows
significant correlation with the Bray-Curtis dissimilarity between
samples calculated based of functional genes. (Mantel static r:
0.9734, p=0.001)
62
Figure 11. Richness and Shannon diversity of functional genes
against log (initial nutrient concentration+5) transformed nutrient
concentration. A quadratic regression line was applied only to
significant results (p<0.05)
63
Figure 12. Richness and Shannon diversity of functional genes
against incubation time. A quadratic regression line was applied only
to significant results (p<0.05)
64
Figure 13. Functional gene richness against richness of the species
assignments of functional genes (A: linear regression applied, B:
quadratic regression applied) and OTU richness (Song et al. 2016)
(C: linear regression applied, D: quadratic regression applied).
65
Figure 14. Functional gene Shannon diversity against Shannon
diversity of species assignment of functional genes (A) and OTU
Shannon diversity (Song et al. 2015) (B).
66
Figure 15. Beta diversity of functional genes calculated based on Bray-Curtis dissimilarity between
samples against log (initial nutrient concentration+5) transformed nutrient concentration. A quadratic
regression line was applied only to significant results (p<0.05)
67
Figure 16. Beta diversity of functional genes calculated based on
Bray-Curtis dissimilarity between samples against incubation time.
A linear regression line was applied only to significant results
(p<0.05)
68
Taxonomic annotation of functional genome
We analysed whether the taxonomic association of each
functional gene had strong correlations with nutrient concentration
or time, to see if they matched to particular genera of bacteria that
might be pinpointed as copiotrophic/oligotrophic or r/K strategists.
However, no matter what the function is, most of the gene
sequences derived from the same set of genera that were abundant
throughout all the samples (Figure 17-20), for example,
Acinetobacter, Pseudomonas, or Sphingomonas, such that we were
unable to identify any genera as being either clearly copiotrophic vs
oligotrophic, or r vs K strategists.
69
Figure 17. Taxonomic annotation of functional genes that have positive correlation with nutrient
concentration. Y-axis shows the relative abundance of 20 most abundant taxa
70
Figure 18. Taxonomic annotation of functional genes that have negative correlation with nutrient
concentration. Y-axis shows the relative abundance of 20 most abundant taxa
71
Figure 19. Taxonomic annotation of functional genes that have positive correlation with incubation time.
Y-axis shows the relative abundance of 20 most abundant taxa
72
Figure 20. Taxonomic annotation of functional genes that have negative correlation with incubation time.
Y-axis shows the relative abundance of 20 most abundant taxa
73
2. 4. Discussion
Chemical analysis of culture material
The overall result seems to confirm that as expected, there
is some decrease in nutrient availability over successional time, but
the initial nutrient concentration is much more important overall
than the change over time. The nutrient measured in the culture
medium might be derived from the leftover of initial nutrient, or
could be derived from recycling of nutrient from dead cells. One
important point is that nutrient A had only 2 to 5 times higher DOC
compared to nutrient C. Possibly, carbon in nutrient A was removed
quickly and sequestered into extra biomass.
Functional gene categories correlating with initial
nutrient concentration (hypothesis A)
As hypothesized, cell division, RNA metabolism and cell
cycle related genes were more abundant in high nutrient
concentrations. High abundance of the RNA related gene categories
implies the possibility of rapid growth being achieved by the more
active expression of overall genetic information encoded in the DNA.
This seems to fit with the view that a copiotrophic environment
selects for the functions of growth and reproduction at the expense
74
of other gene functions. Conversely, and as we had suggested, it
appears that lower nutrient conditions require greater abundances
of genes associated with obtaining and sequestering nutrient
sources such as carbohydrates. This prioritization of acquisition of
nutrients, in different forms and at lower concentrations, suggests
that indeed such functions are a fundamental part of the oligotrophic
strategy in bacteria. It seems that oligotrophic bacterial species also
have relatively high abundance of the genes related to respiration
and electron transfer reactions that are important for the utilization
of their acquired nutrients.
We also hypothesized that copiotrophs would show a
stronger tendency to contain dormancy and sporulation genes, since
they are not adapted to actively grow in nutrient poor conditions
(Fierer et al. 2007). The result supports a view that sporulation is
in particular a strategy used by copiotrophs, to shut down the
metabolism when nutrient supply decreases, and then wait until
nutrient supply improves at some time in the future.
There were also various nutrient concentration-related
patterns in other gene categories. For example, virulence, disease
and defense related genes turned out to be abundant in oligotrophic
conditions. The negative correlation of genes related to resistance
to antibiotics and toxic compounds (one of the subcategories of
virulence, disease and defense genes) with nutrient concentration
75
supports a view of the greater importance of interference
competition amongst bacterial species at lower concentrations of
nutrients (Vance, 1984).
Overall, the correlations of COG genes with nutrient gradient
in our results – at least those genes that show significant
correlations – do largely match the patterns noted by Lauro et al.
(2009) for copiotrophy vs oligotrophy (Table 5). This is even more
striking when one considers the very different basic design of the
two studies, and the fact that the target species and overall
community type belonged to two very different environments:
Lauro et al.’s study (2009) was on marine bacteria, whereas our
study was derived from soil. This strengthens the view that many
of the genes they pinpointed really are of importance in relation to
trophic strategy, and might be universally so – since they show
similar trends in such different environments.
Functional gene categories correlating with incubation
time (hypothesis B)
As expected, cell division and cell cycle related genes were
more abundant in the earlier successional stages, although the
correlation was weaker than for some other gene categories. This
seems to fit with the theory on ‘r’ strategists (Pianka, 1970) that
76
they are selected for rapid growth and reproduction to exploit initial
conditions of lower population densities.
However, contrary to our hypotheses, genes related to
regulation and cell signaling were more abundant at the earlier time
stages of succession in our culture experiments. A priori, the
typical view in ecology would be that later (K-strategy dominated)
successional stages include more biotic interaction (Morriën et al.,
2017; Odum, 1969), whether by mutualism, interference
competition or by specialized parasitism. In our bacterial culture
systems, this principle - based upon ecological observation of
larger organisms over longer periods - does not seem to hold true.
This appears to be an example where principles that apply to
macro-organisms do not apply to the microbial world. It is possible,
for example, that these genes play a role in a strategy of rapid
colonization (in what is largely a two-dimensional static
environment across the cellophane surface of our cultures) through
biofilm formation. Quorum sensing and biofilm formation related
genes were actually more abundant in the early successional stages.
In the early successional stages, biofilm formation might involve
collaborative interactions between members of the same and
different species for the capture of space or nutrients (Shapiro,
1998). Also important in initial capture of space or nutrients may be
activities preventing other cells from growing, for example, by
77
forming a barrier which can deny access of other cells to the
nutrient source, or by antibiotic production (Shapiro, 1998).
Within the general regulation and cell signaling category, we
likewise found that the subcategory of toxin-antitoxin related
genes was more abundant in early stages. These toxin-antitoxin
systems in bacterial species are used to provide genetic stability
when bacterial species divide (Hu et al., 2010). It might be
necessary to have high abundance of these genes for faster growing
bacteria, since they have very limited time for cell division and may
have more errors than slow growing bacteria. Another known
function of toxin-antitoxin system as a mean of programmed cell
death is to provide DNA through cell lysis as a structural
component of biofilm (Bayles, 2007). It is possible, then, that in the
early successional stage, the toxin-antitoxin system might be used
to facilitate biofilm formation genes were likewise more abundant in
the early successional stage.
Motility and chemotaxis genes, and especially flagella-
related genes, were less abundant in the early successional stages,
which also disagreed with one of our hypotheses. We had predicted
that just as r-selected macroorganisms have features promoting
mobility and rapid colonization (Huston and Smith, 1987; Peroni,
1994) these would also be abundant in the early succession stages
of microbial cultures. This might be due to a systematic difference
78
between culture system and the natural environment, whereby
pioneering species are required to have high motility. In setting up
our culture system, we spread soil elutant all across the culture
plates, so early colonizers might not have needed adaptations for
dispersal. By contrast, in aquatic systems the flagellum is a key
cellular structure for acquisition of nutrients, as there are
temporarily nutrient rich patches which bacteria can reach using
flagella and chemotaxis (Jens Efsen et al., 2002). However, in a
culture system such as ours, as colonies deplete nutrients from
around themselves, slower growing bacteria might require motility
to obtain nutrient from spots that have not yet been colonized.
Many other categories of genes showed a significant trend in
relation to incubation time. For example, genes related to minor
nutrients such as sulfur, iron, and potassium showed strong
negative correlations with incubation time. Iron is often a limiting
factor for growth, and such genes might be particularly necessary
to supply enough iron for rapid growth (Church et al., 2000). K-
strategists, in contrast, already have many growth limiting factors
other than iron to contend with, so might be subject to relatively
weak ecological selection for iron uptake compared to other factors.
Another interesting time-related trend is the strong
negative correlation of genes related to metabolism of aromatic
compounds with incubation time. Aromatic compounds might be
79
used as a nutrient source providing additional metabolic versatility
for r-strategists. Even though aromatics are not likely abundant in
our culture media, this may be an incidental legacy of the fact that
these are soil bacteria that normally live in aromatic rich (e.g.
tannin and lignin-rich) environments.
Taxonomic and functional diversity (hypothesis C)
Over the years, there has been much discussion of the
hypothesis that species diversity has a positive correlation with
stability, due to greater functional diversity (McCann, 2000; Wardle
et al., 1999). However, the observed results in terms of both
functional diversity and ecosystem stability vary greatly between
different studies, with some studies supporting the hypothesis of a
link to taxonomic diversity and others contradicting it (Cadotte et
al., 2011; Hooper and Vitousek, 1997; Mayfield et al., 2005;
Petchey and Gaston, 2006). One of the main reasons for this is that
there is no consensus about how best to define functional diversity
– and indeed it is often quite vaguely defined. Functional gene
diversity, based on the proportion of reads in different categories,
is highly quantitative data, and was defined in this fairly objective
way in our study. Gene functional diversity, in terms of functional
gene richness and relative abundances showed a clear linear
80
relationship with OTU richness/richness of species within our
metagenomes – empirically supporting the supposition that more
taxonomic diversity is associated with more functional diversity. We
did not test for aspects of ecosystem stability (McCann, 2000) or
efficiency/rates of ecosystem processes. It would be interesting in
any follow-up studies to test the relationship between stability,
species diversity and functional diversity in such culture systems. It
is also possible that strength of microbial interactions – which might
vary in complex ways that do not necessarily follow overall
diversity – confounds the expected diversity-stability pattern (Ho
et al., 2016).
Taxonomic annotation of functional genome
Previous study which used amplicon sequencing gave us
possible candidates for taxa previously categorized by us as
copiotrophic or oligotrophic, on the basis of their prevalence in
different nutrient treatments (Song et al., 2016). Beyond this, our
metagenome study presented here provides functional genes
associated with bacterial strategies in different nutrient
concentration and incubation time. The taxonomic annotation of the
functional genome shows a similar distribution pattern, particularly
in terms of the most abundant genera associated with each
81
functional gene category, (Figure 17-20) which implies that the
difference between samples in terms of functional gene composition
is strongly linked to the differences in relative abundance of these
most abundant genera. It is also possible that the differences in
functional composition between samples might be caused by the
combination of many minor genera that has been selected to their
functional characteristics. Finer phylogenetic resolution (species or
strain level) may be needed to capture differences in the
community composition between treatments.
Possible limitations of this study and suggestions for
further work
It is important to acknowledge that there are various
limitations in this study, which may limit the accuracy or relevance
of the results in terms of understanding bacterial community
ecology in nature:
1) This is an artificial system. Microcosms are by their nature
derived and simplified from natural systems. This simplification has
the advantage of allowing more precise control and manipulation of
conditions, but the risk of leaving out certain crucial aspects of
natural systems which may make their results less relevant to
nature. It is possible that microcosms, which modify and monitor
82
small quantities of soil (rather than soil elutant) would give a more
representative understanding of bacterial strategies in nature. It is
also important to realize that the types of nutrients added might
have an important effect on community structure. Also, the soil
system is only one of a huge variety of environments bacteria can
live in. In other environments, different strategies might apply –
although the overlap between the genes associated with copiotrophy
in our study and those found by Lauro et al. (2009) is reassuring,
and implies that we may be observing true general patterns in
bacterial ecology.
2) We did not study metatranscriptomes. Here we studied the
metagenome (DNA), rather than metatranscriptome (RNA) which
would provide a more direct picture of cellular activity. It is
possible that part of the community composition and functional gene
composition that we see is related to a ‘legacy’ of cells that are
already dead or inactive, derived from earlier successional stages
or from the initial soil elutant. This may be tending to ‘blur’ the
functional patterns observed. On the other hand, the metagenome
can give a picture of the whole range of potentially expressed genes
present, which may be relevant to the overall strategies of these
organisms in nature, whether or not they were being expressed at
the time of sampling.
3) Our data has been collected across a prescribed (though very
83
broad) range of initial nutrient concentrations and incubation times,
so the conclusions made in this study may only be applicable within
this range. It is possible that the gene abundance patterns would
differ outside of this range.
It is likely that the more nutrient-rich treatments we used
are much more copiotrophic than the average background level of
nutrients in soils, but do correspond to occasional, spatially isolated
and often brief, bursts of very high nutrients from rotting animal
carcasses, rotting rootlets, dung, rotting fruits etc. These highly
copiotrophic environments are likely to be ecologically important
enough to select specialized communities – even if strategists that
thrive in these environments must bide their time (perhaps as
dormant or low activity forms at lower abundance within the general
soil community) between such opportunities. These bursts may also
occur on a microscopic scale (e.g. the vicinity of a single dead
nematode), but to bacteria they would be high nutrient
environments nevertheless.
We did not do chemical analysis for the “nutrient E” sample,
which has the lowest nutrient concentration. From the dilutions, we
can deduce that DOC and DTN in nutrient E should be lower than
3mg/L which is similar to environmental samples (Jones and Willett,
2006; Michalzik et al., 2001) depending on the environment that is
used for comparison.
84
4) It is difficult to separate the effects of environmental selection
and phylogeny. If a particular phylogenetic group of bacteria is
particularly abundant in (for example) copiotrophic conditions, any
genes that are consistently associated with that group will tend to
be more abundant. These genes may be fundamental to the group’s
success as copiotrophs, or alternatively they might be ‘riders’
which have no particular significance. Instead, the true ‘strategy’
might lie with subtle differences in other groups of genes – perhaps
involving finely tuned qualitative differences in behavior of
expressed proteins rather than differences in gene abundances.
5) It is unclear whether the patterns observed are always due to
selection for a particular gene function, or selection against other
functions within the genome. Since metagenome composition
measures relative abundances, a gene may become more common
not because it is actively selected for, but by the relative absence
of other sets of genes, which are selected against. This somewhat
limits the confidence that can be placed in the conclusions.
6) Since the percentage of proteins with no known function was
greater in later successional stages, it is possible that there is a
trend of increasing diversity in gene functions and functional
composition with successional time that cannot easily be discerned
using existing databases.
Nevertheless, despite all these caveats we regard this study
85
as a useful early step towards identifying the key differences,
which can explain the differences in bacterial community
composition along nutrient and time gradients.
86
2. 5. Concluding remarks
Although based on a highly simplified system, the results of
this experiment identify sets of genes, which may be part of
distinguishable community-level strategies seen amongst bacteria,
in terms of the relative abundance of particular functional gene
types. In terms of the oligotrophic-copiotrophic axis, the gene
abundance patterns we found are close to those identified by Lauro
et al. (2009), and the agreement between the two studies (based on
quite different approaches) reinforces confidence that the functional
gene differences found have relevance in determining strategies.
We were also able to identify a second axis of functional gene
variation, distinct from the oligotrophic-copiotrophic one, that
relates to time and resembles the r-K axis found in successional
ecological systems of larger organisms. However, it is interesting
that the patterns of adaptations (e.g. for dispersal, or regulation and
cell signaling) do always not match those classically described for
macro-organisms. These results, while intriguing, are merely an
initial attempt at discerning and understanding the basis of bacterial
adaptations for survival. It is hoped that this study will provide a
spur for thought and further work on this important aspect of
microbial ecology.
87
CHAPTER 3: Environmental Filtering
of Bacterial Trait Diversity along an
Aridity Gradient
88
3. 1. Introduction
Soil metagenomes can potentially reveal patterns and
general rules of function of soil biota, and from this the general
ecological patterns of community structure and ecosystem function
(Fierer, 2017). The search for community traits in metagenomes is
based upon genes considered to code for these traits, rather than
observation of the traits themselves, which would often be difficult
to observe and quantify within the soil biota in the intact system.
Various metagenetic and metagenome studies in recent
years have tested whether hypotheses based on observation of
variation in community traits of larger organisms (such as trees,
corals or birds) also apply to the microbial world. For example, Choi
et al. (2017) tested for the existence of disturbance history related
gradients in ecological strategies, analogous to the r- and K
strategies seen in larger organisms (MacArthur and Wilson, 1967).
Tripathi et al. (2017) tested for certain traits against an aridity
gradient in Israel, making various predictions about how these traits
may be expected to vary in relation to a stress gradient.
However, the study by Tripathi et al. (2017) was limited in
scope, taking only four points along an aridity gradient and using
only 3 closely clustered replicates at each point along the gradient.
There was a need for a more thorough study comparing
89
metagenomes along an aridity gradient. We chose the Tibetan
Plateau because it presents a strong aridity gradient under similar
mean annual temperature conditions.
Our primary interest was in the effects of an environmental
stress gradient on the functional diversity of the soil biota. Theory
and observation of larger organisms, notably plants, has predicted
and observed broad scale gradients in the diversity of functional
traits along climatic stress gradients (Kleidon et al., 2009; Kleidon
and Mooney, 2000; Swenson et al., 2012). These authors suggested
that extreme physiological conditions and reduced energy budgets
for organisms constrain the number of different functional
strategies, which are viable in any given environment. There were
hints of this pattern in the earlier soil metagenome study of an
aridity gradient in Israel (Tripathi et al., 2017), where higher
aridity was associated with reduced functional trait diversity in the
metagenome.
In the Israel precipitation gradient study (Tripathi et al.,
2017), authors also made various predictions regarding specific
traits. They expected increased abundance of stress response gene
and genes related to sporulation in more arid conditions. Also, they
expected genes related to nutrient cycling will be more abundant in
less arid conditions and there would be more genes related to
metabolism of complex organic compounds. These showed the
90
observed patterns in some respects (genes related to sporulation,
nutrient cycling, and metabolism of complex organic compounds)
but not others (stress response gene).
Compared to the study by Tripathi et al. (2017), our present
study was based on a much greater number of samples, scattered
much more widely spatially and at a range of different precipitation
values. As such, it represents a more rigorous and convincing study
of how trait diversity and characteristics vary along climate
gradients.
In the present study, we aimed to test the following hypotheses
(from now on, ‘gradient’ will always be used as a gradient from low
to high):
1) Trait diversity will decrease along climate gradient towards
increasing aridity with less inter-site variability in functional
composition and reduced functional diversity will be closely
associated with reduced taxonomic diversity. Such trends have
been found in physiological or morphological characteristics of
larger organisms (Kleidon et al., 2009; Kleidon and Mooney, 2000;
Nunes et al., 2017) and in functional gene diversity in microbes
(Tripathi et al., 2017). Extreme conditions will require a large
proportion of an organism’s energy and other resources to be
diverted into homeostasis (e.g. against extreme osmotic conditions,
extreme pH, lack of water etc.) (Lauber et al., 2009; Nunes et al.,
91
2017). This may preclude other functional traits which could
otherwise enhance survival (e.g. through additional strategies of
nutrient acquisition, mutualism, or interference competition) or
allow occupation of a wider range of different types of niches. We
also expect inter-site variability will be lower in arid conditions due
to frequent wind transport homogenizing the microbial community.
2) Genes related to cell division and dormancy will increase along
an aridity gradient. Desert soil experiences a strikingly unstable
environment. Strong and frequent wind work as a mechanical
disturbance and there tends to be an extreme daily temperature
range – especially in high elevation deserts (Belnap and Gillette,
1998; Kemp et al., 1992; van Gestel et al., 2011).
Rainfall/meltwater events in desert are rare, with soils dry most of
the time, and when the soil is wetted organisms of all types have a
brief window of time in which to grow, feed and reproduce
(Schwinning and Sala, 2004). Therefore, it is expected that
bacterial cells will be adapted to a) divide fast when they encounter
moist and moderately warm conditions, and b) stay in a dormant
state under unfavorable conditions of aridity/high temperatures/low
temperatures. This will result in greater copy numbers within
genomes of genes related to cell division (for times of growth) or
dormancy (for times when growth is not favored).
92
3) Stress response genes – especially those relating to
osmoregulants - will be more abundant in environments that are
more arid. In plant ecology, stress is defined as “external
constraints which limit the rate of dry matter production of all or
part of the vegetation” (Grime, 2006). Desert environments are
often regarded as ‘stressful’ because of high disturbance rate
through wind erosion, high salinity, low moisture content, low
nutrient concentrations, and low primary productivity. In high
elevation deserts, extreme daily temperature range and high UV
light flux are also important. Many macro-organisms living in
desert are adapted to extreme conditions by mechanisms for
example, by producing osmoprotectants (Ashraf and Foolad, 2007)
such as insoluble forms of nitrogenous waste (Hadley, 1974).
Bacterial species also cope with extreme environments by
accumulating protectant solutes, or by producing large quantities of
chaperone proteins to stabilize cellular proteins (Csonka, 1989;
Schimel et al., 2007). We expected that bacterial genes for these
and other stress responses would be more abundant towards the
more arid end of the climate gradient we studied here.
4) There will be a lower frequency of antibiotic resistance and
virulence genes in more arid conditions due to the predominance of
abiotic selection. In ecology, it is generally considered that in
extreme environments, the physiological costs and functional
93
constraints on organisms leave relatively little of their resources
for interference competition (Grime, 2006; Pueyo et al., 2008).
Additionally, slow rates of growth and population increase in
extreme environments, relative to rates of disturbance, are seen as
resulting in less intense competition (Huston, 1979). Thus, we
anticipated that there would be relatively low abundances of genes
relating to biotic interaction – either for interference competition or
for purposes of forming mutualisms. Such genes would include, for
example, antibiotic resistance genes – being less frequent in the
metagenome because of reduced interference competition by
antibiotic production (Brook, 1999; Hibbing et al., 2010).
5) Motility related genes will decrease along aridity gradient. In
plant ecology, the average seed spatial dispersal rate of the
community decreases in more arid conditions, due to differences in
investment in dispersal traits in plants (Pueyo et al., 2008; Venable
et al., 2008). Favorable growing conditions are ephemeral and rare
in the desert landscape, and thus difficult to find, and as a result
plant seeds in deserts have evolved to stay close to the mother
plant and in a dormant state rather than disperse across the
landscape to find favorable conditions (Venable et al., 2008). In the
bacterial world, we expected the same principle to operate, in that
bacterial cells should allocate more towards genes related to
94
dormancy instead of motility, since it is better to await a return of
moist conditions than to spend resources on discovering new places.
6) Functional gene composition along nutrient concentration will
follow those found in the microcosm study (Chapter 2). Although
the two studies are very different from each other systematically
(one is from precisely controlled system and the other one is from
extremely heterogeneous environment with several environmental
factors co-varying), we expect that we will be able to find
consensus patterns in the two studies.
95
3. 2. Materials and Methods
Soil sampling and data collection
We collected soil sample from 29 grassland sites across in
Tibetan plateau (Figure 21). We sampled during the growing season
in July-August. In each site, seven randomly located 0-5cm depth
soil cores (each 5 cm diameter) were collected across a single 1
square meter quadrat (m2). The seven individual soil core samples
were placed in one bag and thoroughly mixed to homogenize them –
making one sample from which the DNA for a single metagenome
would be obtained.
96
Figure 21. Map of sample collection site
97
Climatic data collection and physicochemical analysis
Climatic condition of sampling sites and physicochemical
properties of the samples are shown in Appendix Table A1. We
collected mean annual temperature (MAT), mean annual
precipitation (MAP) data of each site from the National
Meteorological Bureau of China database (Jing et al., 2015).
Growing season (from April to August) temperature (GST) and
growing season precipitation (GSP) data were compiled from the
1950-2000 records of a global climate database (Hijmans et al.,
2005; Ma et al., 2010). Potential Evapotranspiration (PET) data
were collected from Global Aridity and PET database of CGIAR-
CSI (http://cigar-csi.org) (Yang et al., 2017a). Actual
evapotranspiration (AET) was determined by comparing potential
evapotranspiration with precipitation plus available water stored in
the soil as described in Major (1963) and Kreft and Jetz (2007).
Soil moisture (SM) was measured gravimetrically by drying
soils at 105℃ for ~10 hours. Soil pH was measured using a pH
meter (Thermo Orion-868) in a 1:5 ratio of fresh soil to deionized
water slurry. Soil total carbon (STC) and soil total nitrogen (STN)
were measured by combustion of the samples in a CHN elemental
analyzer (2400 Ⅱ CHN elemental analyzer, PerkinElmer, Boston,
MA, USA). Soil total phosphorous (STP) was measured by
98
molybdenum blue method in conjugation with ultraviolet-visible
spectrophotometer (UV-2550, Shimadzu, Kyoto, Japan). Soil
CaCO3 was measured using a volumetric calcimeter (Eijkelkamp,
Netherland). Soil organic carbon (SOC) was calculated by
subtracting soil CaCO3 (SCaCO3) from soil total carbon (STC). The
density of SOC, STN, STP, SCaCO3 were calculated in the top 0-
5cm soil.
Dissolved organic carbon (DOC), dissolved total nitrogen
(DTN), and soil mineral nitrogen were extracted by mixing 10
grams of fresh soil with 50 ml of 0.5 M K2SO4, shaking for an hour,
and vacuum filtering through glass fiber filters (Fisher G4, 1.2 ㎛
pore space). Ammonium nitrogen (NH4+-N) content were assessed
colourimetrically by automated segmented flow analysis
(Bran+Luebbe AAⅢ, Germany) using salicylate/dichloroisocyanuric
acid method. Nitrate nitrogen (NO3--N) content were assessed
using the same analysis with cadmium column/sulphanilamide
reduction method. DOC and DTN was assessed by TOC-TN
analyzer (Shimadzu, Kyoto, Japan). Dissolved organic nitrogen
(DOC) was calculated by subtracting NH4+-N and NO3
-—N from
DTN (Yang et al., 2017b).
To assess aboveground biomass, plant total carbon (PTC),
plant total nitrogen (PTN), and plant total phosphorous (PTP), plant
communities in each 1-m2 plot were surveyed and harvested before
99
soil sampling (Jing et al., 2015). Plant materials were dried at 60℃
for ~12 hours and grinded. PTC, PTN, and PTP was analyzed with
the same method used for STC, STN, and STP. To assess
belowground biomass, 4-8 soil cores per 1-m2 plot (0-5 cm deep,
7 cm diameter) was collected. To calibrate root biomass, live:dead
(56:63) ratio was used.
DNA extraction, sequencing and sequence processing
DNA was extracted from 0.5 g of soil in each sample using a
FastDNA Spin kit (Bio 101, Carlsbad, CA, USA), following the
manufacturer’s instructions and stored at -40℃. Extracted DNA
samples were sent for shotgun metagenomic sequencing to
Celemics (Seoul, Korea) using Illumina HiSeq2500 sequencing
platform. Sequence data file was uploaded to MG-RAST
(metagenomics Rapid Annotation using Subsystem Technology)
server for quality control and for annotation (Meyer et al., 2008)
(reviewer access:
http://metagenomics.anl.gov/mgmain.html?mgpage=token&token=N
dd1oCi0lw4In9JuUQzlfjaXYDORTnEg2jEjWRjnsrYUaKcU6E). For
taxonomic annotation of functional genes, Refseq database was used
(Pruitt et al., 2007) and for functional annotation, SEED Subsystem
database was used (Overbeek et al., 2005). Sequences were
annotated with default setting (maximum e-value cutoff of 10-5,
100
minimum % identity cutoff of 60%, and minimum alignment length
cutoff as 15 bp). About 97% of total sequences on average
comprised of bacterial sequences, so we focused only on bacterial
functional genes and removed sequences derived from other
organisms (archaea, viruses, and eukaryotes).
Statistical analysis
Samples were subsampled to the least sequence read
number for standardization. To find environmental variables that
explain variance in functional gene composition between samples,
we performed redundancy analysis (RDA) using Canoco ver. 5.
Subsystem Level 3 functions were square-root transformed and
environmental variables were forward selected. Monte Carlo
permutation test with 999 permutations were used to test
significance of the model. To visualize the correlation between z-
score transformed relative abundance of functional gene categories
and precipitation gradients, we drew heatmap using “pheatmap”
package in R. Gene categories were clustered by their Euclidean
distance with unweighted pair group method with arithmetic mean
(UPGMA) method. To find out linear relationship between
functional gene diversity and environmental factors and to generate
101
figures, we used Sigmaplot ver. 10.0 from Systat Software, Inc.,
San Jose California USA (www.systatsoftware.com).
Functional diversity was calculated based on Subsystem
Level 3 functions. To compare functional diversity and taxonomic
diversity, bacterial community composition data was obtained from
Jing et al. (2015). To calculate beta diversity, subsampled read
numbers were square root transformed and Bray-Curtis
dissimilarity between each sample point was calculated. To compare
taxonomic and functional beta diversity, Mantel test was performed
by using “ mantel.test” function in R package “vegan” . To
compare functional beta diversity in different vegetation types
(alpine meadow, alpine steppe, and desert steppe), bacterial
functional gene composition data were square root transformed and
Bray-Curtis dissimilarity from group centroid was calculated.
102
3. 3. Result
Functional gene composition along an aridity gradient
After quality control, the total number of sequence reads per
sample ranged from 1,046,994 to 2,274,876. The RDA result
showed that mean annual precipitation was the best predictor for
functional gene composition, alone explaining 26.7% of total
variation (Figure 22, Table 7). Heatmaps of standardized
Subsystem Level 1 gene relative abundance along mean annual
precipitation (Figure 23) and soil moisture (Figure 24) show gene
categories variously having significant positive, negative correlation
or no correlation. As aridity increases along the gradient, genes
related to cell division and cell cycling, dormancy and sporulation,
DNA metabolism, RNA metabolism, protein metabolism, nucleotides
and nucleosides, and etc. became relatively more abundant. Genes
related to nutrient cycling (sulfur metabolism, potassium
metabolism, nitrogen metabolism, iron acquisition and metabolism,
fatty acids) and metabolism of aromatic compounds became less
abundant. Motility gene, regulation and cell signaling gene and
secondary metabolism related genes also decreased. Since stress
response category can be subdivided to different type of stress, we
checked Subsystem Level 3 genes, which belongs to stress
103
response category whether they have correlation with precipitation
gradients (Table 8).
104
Figure 22. RDA result showing sample distribution based on bacterial
functional gene composition along the environmental gradient in Tibet.
Vegetation types are indicated, with MAP of alpine meadow>alpine
steppe>desert steppe. MAP: mean annual precipitation (mm), SM:
soil moisture (g/g dried soil), SOC: soil organic carbon (%), STP: soil
total phosphorous (%).
105
Table 7. Forward selection of environmental variables, which best explain variation in functional gene
composition between samples. A pseudo-F statistic from Monte Carlo permutation test was derived in the
same way as in ANOVA of the regression model
Environmental variable Explains % pseudo-F P
Mean annual precipitation (MAP) 26.7 9.8 0.001
Soil moisture (SM) 6.9 2.7 0.003
Soil organic carbon (SOC) 7.4 3.1 0.002
Soil total phosphorous (STP) 4.1 1.8 0.018
Soil C:N ratio 4.0 1.8 0.019
106
Figure 23. Heatmap of standardized relative abundance of Subsystem Level 1 genes along mean annual
precipitation (mm). Gene categories that have positive/negative/no linear correlation with mean annual
precipitation was denoted by ‘(+)’/‘(-)’/‘(0)’
107
Figure 24. Heatmap of standardized relative abundance of Subsystem Level 1 genes along soil moisture
(g/g of dried soil). Gene categories that have positive/negative/no linear correlation with soil moisture was
denoted by ‘(+)’/‘(-)’/‘(0)’
108
Table 8. Subsystem Level 3 genes belonging to stress response category showing significant correlation
with mean annual precipitation
Subsystem Level 2 Subsystem Level 3 p-value Correlation
coefficient
Acid stress Acid resistance mechanisms 0.000 0.739
Cold shock Cold shock, CspA family of proteins 0.027 -0.411
Detoxification Uptake of selenate and selenite 0.000 0.736
Heat shock Heat shock dnaK gene cluster extended 0.000 -0.700
NULL Bacterial hemoglobins 0.000 0.867
Carbon Starvation 0.007 0.488
Dimethylarginine metabolism 0.034 0.395
Flavohaemoglobin 0.024 0.419
Phage shock protein (psp) operon 0.000 -0.630
109
Table 8. (Continued)
Subsystem Level 2 Subsystem Level 3 p-value Correlation
coefficient
Osmotic stress Betaine biosynthesis from glycine 0.009 -0.474
Ectoine biosynthesis and regulation 0.001 -0.584
Osmoprotectant ABC transporter YehZYXW of
Enterobacteriales
0.010 -0.469
Osmoregulation 0.000 0.638
Osmotic stress cluster 0.044 0.377
Synthesis of osmoregulated periplasmic glucans 0.000 0.607
Oxidative stress Glutathione: Non-redox reactions 0.000 0.702
Glutathione: Redox cycle 0.030 0.403
Glutathione analogs: mycothiol 0.017 -0.441
Oxidative stress 0.002 0.561
Protection from Reactive Oxygen Species 0.002 0.555
Redox-dependent regulation of nucleus processes 0.002 -0.553
110
Functional gene composition along nutrient gradient
To compare functional attributes changing along nutrient
gradients in this study with those found in the microcosm study
(Chapter 2) we tested correlation between Subsystem Level 1
functions and major nutrient components (soil total carbon, soil
organic carbon, soil total nitrogen, soil total phosphorous). As in
Table 9, several gene categories showed a different pattern in the
two studies. For example, relative abundance of carbohydrate
related genes showed negative correlation with nutrient
concentration in the microcosm study, but showed positive
correlation with nutrient concentration in the field study in Tibetan
plateau. Also, cell division and cell cycle related genes were
abundant in high nutrient condition in the microcosm study, but not
in the field study in Tibetan plateau. Genes related to motility and
chemotaxis, potassium metabolism, RNA metabolism, stress
response, virulence, disease and defense also showed an opposite
pattern.
111
Table 9 Pearson correlation between Subsystem Level 1 genes and soil total carbon (STC), soil organic
carbon (SOC), soil total phosphorous (STP), soil total nitrogen (STN) in the field study in Tibetan plateau
and Spearman correlation between Subsystem Level 1 genes and initial nutrient concentration in the
microcosm study (Chapter 2). Significant positive correlations are highlighted in green and significant
negative correlations are highlighted in red.
Subsystem Level 1 genes STC SOC STP STN Microcosm
Amino Acids and Derivatives -0.287 -0.294 -0.238 -0.332 0.803
Carbohydrates 0.525 0.558 0.377 0.554 -0.672
Cell Division and Cell Cycle -0.772 -0.791 -0.409 -0.786 0.696
Cell Wall and Capsule 0.136 0.139 -0.052 0.138 0.251
Clustering-based subsystems -0.746 -0.773 -0.365 -0.752 0.551
Cofactors, Vitamins, Prosthetic Groups, Pigments -0.556 -0.557 -0.372 -0.582 0.188
DNA Metabolism -0.762 -0.785 -0.445 -0.787 0.213
Dormancy and Sporulation -0.352 -0.345 -0.154 -0.337 0.593
Fatty Acids, Lipids, and Isoprenoids 0.641 0.655 0.219 0.637 -0.226
Iron acquisition and metabolism 0.302 0.307 0.044 0.330 -0.025
Membrane Transport 0.238 0.214 0.318 0.259 0.138
112
113
Table 9. (Continued)
Subsystem Level 1 genes STC SOC STP STN Microcosm
Metabolism of Aromatic Compounds 0.762 0.798 0.326 0.767 0.347
Miscellaneous -0.384 -0.373 -0.202 -0.348 0.105
Motility and Chemotaxis 0.673 0.694 0.310 0.661 -0.404
Nitrogen Metabolism 0.634 0.652 0.356 0.662 0.217
Nucleosides and Nucleotides -0.797 -0.793 -0.518 -0.786 0.242
Phages, Prophages, Transposable elements, Plasmids -0.616 -0.645 -0.326 -0.598 -0.070
Phosphorus Metabolism 0.172 0.217 0.086 0.248 -0.136
Photosynthesis -0.231 -0.197 -0.322 -0.173 -0.785
Potassium metabolism 0.811 0.825 0.437 0.800 -0.406
Protein Metabolism -0.724 -0.760 -0.305 -0.743 -0.552
Regulation and Cell signaling 0.419 0.434 0.154 0.435 0.209
Respiration 0.505 0.482 0.352 0.465 -0.828
RNA Metabolism -0.735 -0.755 -0.578 -0.763 0.619
Secondary Metabolism 0.422 0.473 0.225 0.501 0.091
Stress Response 0.619 0.635 0.271 0.659 -0.326
Sulfur Metabolism 0.756 0.799 0.441 0.797 0.368
Virulence, Disease and Defense 0.704 0.720 0.411 0.716 -0.795
114
Functional gene diversity vs OTU diversity
Alpha diversity. Functional gene richness increased as OTU
richness increase (Figure 25-A), but functional gene Shannon
diversity had no linear relationship with OTU Shannon diversity
(Figure 25-B).
Beta diversity. Taxonomic and functional gene beta diversity
between samples calculated based on Bray-Curtis dissimilarity
showed significant correlation. (Mantel test of 999 permutations,
Pearson correlation, Mantel static r=0.8376, p<0.001).
115
Figure 25. Metagenome bacterial Subsystem Level 3 functional gene richness against bacterial OTU
richness (A). Metagenome bacterial Subsystem Level 3 functional gene Shannon diversity against
bacterial OTU Shannon diversity (B). Linear regression line was applied only when significant. Presented
R-squared value is adjusted R-squared value.
116
Functional gene diversity along aridity gradients
Alpha diversity. Functional gene richness and Shannon diversity
decreased along aridity gradient (Figure 26). The correlation was
stronger than that between OTU richness/Shannon diversity
(Figure 27). In fact, neither bacterial OTU richness nor bacterial
Shannon diversity showed significant correlation with precipitation
gradient (0.05<p<0.1) (Figure 27). Functional gene richness nor
Shannon diversity had linear relationship with soil moisture (Figure
28). However, functional gene Shannon diversity and soil moisture
had strong rank based correlation (spearman correlation, rho=0.75,
p<0.0001) which are applicable to the dependent variables with
non-normal distribution. Although our study focuses on bacterial
sequences, for more information on the functional patterns in soil
biota, we also checked whether fungal or archaeal gene Shannon
diversity have correlation with mean annual precipitation (Figure
29). We found no significant correlation.
Beta diversity. We grouped samples by vegetation type (which is
broadly representative of sections of the aridity gradient) and
compared beta diversity calculated based on Bray-Curtis
dissimilarity of functional gene composition from group centroid. In
contrast to alpha diversity, beta diversity was greatest in desert
steppe samples, which had lowest average mean annual
117
precipitation (127.87 mm) together with alpine meadow samples
(Figure 30). Thus, desert samples were on a local scale relatively
homogeneous, but at distance relatively heterogeneous.
118
Figure 26. Metagenome bacterial Subsystem Level 3 functional gene richness against mean annual
precipitation (A). Metagenome bacterial Subsystem Level 3 functional gene Shannon diversity against
mean annual precipitation (B). Linear regression line was applied. Presented R-squared value is adjusted
R-squared value.
119
Figure 27. Bacterial OTU richness against mean annual precipitation (A). Bacterial OTU Shannon diversity
against mean annual precipitation (B). Results not significant (P>0.05).
120
Figure 28. Metagenome bacterial Subsystem Level 3 functional gene richness against soil moisture (A).
Metagenome bacterial Subsystem Level 3 functional gene Shannon diversity against soil moisture (B).
121
Figure 29. Metagenome fungal Subsystem Level 3 functional gene Shannon diversity against mean annual
precipitation (A). Metagenome archaeal Subsystem Level 3 functional gene Shannon diversity against
mean annual precipitation (B). Results not significant (P>0.05).
122
Figure 30. Functional beta diversity in each vegetation type
calculated based on Bray-Curtis dissimilarity of Subsystem Level 3
genes from group centroid. Alphabet denotes posthoc test result of
Tukey’s HSD.
123
Taxonomic annotation of functional genes
To see if there are dominant taxonomic groups for each
gene category, we annotated each gene category using taxonomic
database. We found that most of Subsystem Level 1 genes that had
negative correlation with precipitation gradients belonged to
Conexibacter (Figure 31). Subsystem Level 1 genes that had
positive correlation with precipitation gradients belonged to
Mycobacterium, Rhodopseudomonas, Pseudomonas, Bradyrhizobium,
etc. (Figure 32). Overall, however, it was hard to spell out specific
taxonomic groups from each of the gene category since there were
huge overlap in the most abundant 20 genus of the taxonomic
annotation of functional genes (Figure 31-32).
124
Figure 31. Taxonomic annotation of functional genes that have negative correlation with precipitation
gradients. Y-axis shows the relative abundance of 20 most abundant taxa
125
Figure 31. (Continued)
126
Figure 32. Taxonomic annotation of functional genes that have positive correlation with precipitation
gradients. Y-axis shows the relative abundance of 20 most abundant taxa
127
Figure 32. (Continued)
128
3. 4. Discussion
Hypothesis 1) Trait diversity will decrease along an
aridity gradient
As hypothesized, functional gene diversity decreased along
the aridity gradient. By extension from gene diversity, it appears
that more arid conditions may indeed constrain the number of traits
that are viable (and hence retained in the community), as discussed
by Kleidon et al. (2009), Kleidon and Mooney (2000) for plants,
and Tripathi et al. (2017) for microbes. The relationship between
functional gene diversity and aridity was stronger than the
relationship between taxonomic diversity and aridity.
Intriguingly, functional beta diversity was paradoxically
greatest in the desert samples together with alpine meadow
samples. This may be a habitat heterogeneity effect, due to
stronger substrate rock effects in an arid environment where
weathering has exerted a weaker effect on soil chemistry. It might
also be due to an ‘oasis effect’ whereby locally more resource rich
environments produce greater patchiness in bacterial community
traits.
We found that functional gene richness also correlates with
OTU richness, but its relationship was much weaker than the effect
129
of aridity. The OTU-richness/gene richness pattern was also
weaker in comparison to the strength of the OTU-richness
relationship seen in the microcosm study (Chapter 2). In the
microcosm study, we were dealing with a rapidly selected subset of
a single common community (one soil sample from which all the
culture treatments were derived), whereas in this study across the
Tibetan Plateau the samples are a product of long-term ecological
and evolutionary selection between different places. For this reason,
we suggest that the trend seen in our study of Tibet (gene
functional diversity independent of taxonomic diversity) is more
representative of broad scale community trends in nature.
Hypothesis 2) Genes related to cell division and
dormancy will increase along aridity gradient
Agreeing with our hypothesis, genes related to cell division
and dormancy increased along aridity gradient. This suggests
bacterial species living in extreme conditions allocate more space in
their genome towards awaiting and then rapidly exploiting more
favorable growing conditions. A similar pattern has been found in
cross-biome study – showing that desert soil metagenome samples
had more abundant cell division and dormancy genes compared to
non-desert samples (Fierer et al., 2012). Also, these two gene
130
categories were most abundant in more arid regions in Tripathi et al.
(2017). The lower abundance of motility genes in more arid
conditions likewise parallels the findings of Tripathi et al. (2017),
and supports the idea that bacterial species may allocate their
energy and genome space towards dormancy rather than finding
new environments.
Hypothesis 3) Stress response genes – including
those relating to osmoregulants - will be more
abundant in higher aridity gradients
In contradiction with our expectations, overall stress
response gene relative abundance decreased along the aridity
gradient. This pattern was also found in Tripathi et al. (2017).
Within the overall category of stress response genes, this trend
was dominated by genes relating to “Synthesis of osmoregulated
periplasmic glucans” (Table 8), which are produced when there is
excessive water (Jean-Pierre and Jean-Marie, 2007). This might
then explain why moister climates were associated with greater
abundance of these genes, since average soil water content, and the
frequency of bursts of very high soil water content, is likely to be
greater. However, the subcategories of genes related to production
131
of osmoprotectants such as choline and betaine uptake or betaine
biosynthesis (“Betaine biosynthesis from glycine”, “Choline and
betaine uptake and betaine biosynthesis”, “ectoine biosynthesis and
regulation”) were more abundant in more arid conditions, as would
be expected (Csonka, 1989; Schimel et al., 2007). Hence, it seems
to be partly a case of studying the correct subset of stress genes in
relation to aridity gradients.
Hypothesis 4) There will be less incidence of
antibiotic resistance and virulence genes in more arid
conditions
As predicted, there was lower relative abundance of genes
related to cell-cell interaction (regulation of cell signaling),
antibiotic resistance and virulence genes towards the more arid end
of the gradient. As aridity increases, cell density decreases (Clark
et al., 2009), and this may be expected to decrease selection for
competitive traits, by analogy with the stress-selected state of
Grime for plants (Grime, 2006). It is possible that resource-
consuming adaptations for survival of environmental stress may
also leave fewer resources available for such competitive traits.
132
Hypothesis 5) Motility related genes will decrease
along aridity gradient
As we predicted by analogy with less investment in seed
mobility by desert plants (Pueyo et al., 2008; Venable et al., 2008),
bacterial motility genes were less abundant in arid conditions. This
suggests that for bacteria too, a more conservative strategy of
remaining in place in favorable microsites is ecologically selected in
the arid environment. The motility genes, which varied along the
aridity gradient, were mostly flagellar genes: motility in bacterial
world most often relates to flagellae, which represent an active
form of movement. It is also necessary to bear in mind that as
flagallae require water surrounding the cells, these might be of less
ecological benefit within a soil that is dry most of the time.
Hypothesis 6) Functional gene composition along
nutrient concentration will follow those found in the
microcosm (Chapter 2)
In contrast to our expectation, the field study in Tibetan
plateau and the microcosm study showed different pattern in the
functional composition along nutrient gradients. These differences
133
seems to be caused by systematic differences between the two
studies. For example, the relative abundance of cell division and cell
cycle related genes were less abundant in low nutrient
concentration in culture sample, but not in the soil samples from
Tibetan Plateau. Low nutrient conditions are usually found in desert
samples from Tibetan Plateau (Appendix Table A1), which have
higher disturbance frequency than forest sites. It is possible that
disturbance intensity in the desert steppe of Tibetan Plateau have
stronger selective force than nutrient concentration for cell division
and cell cycle related genes. Motility related genes were more
abundant in high nutrient concentration in the microcosm study, but
opposite trend was found in the field study in Tibetan Plateau.
Moisture content could be more important for the selection of
motility related genes than nutrient.
Possible limitations of the study
The functional diversity as recognized here is based only on
gene diversity, which is not necessarily the same as phenotypic
functional diversity of the whole community. Some bacterial OTUs,
while present, may be relatively more active (Blazewicz et al., 2013)
- some gene function types may be misassigned by MG-RAST. It
would be informative to study soil catabolic diversity (Degens et al.,
134
2001) to understand the full functional trait diversity of these soils
and how it varies.
It is possible that the functional profile of these soils is
significantly altered by a legacy of ‘old’ DNA from dead or lysed
cells (Carini et al., 2016; Fierer, 2017). Whether this old DNA
negates the relevance of what is seen is a moot point, as it
represents a time-integrated view of total community gene
diversity rather than a snapshot in time. In any case, it is important
to realize that old DNA would only be expected to add to the gene
diversity of the desert soils, where DNA may be preserved longer
due to slower turnover rates – yet the more arid soils in our
gradient still have lower functional gene diversity. Perhaps, without
the signal of old DNA, their gene diversity would be even lower
compared to that of moister climate soil.
135
3. 5. Concluding remarks
In this study, we were able to identify several key trends in
soil bacterial functional gene composition along aridity gradients.
The most striking of these was a gradient in total functional gene
diversity. While the question remains of whether gene diversity as
measured by MG-RAST really translates into trait diversity in a
way that is analogous to that measured in plant communities
(Kleidon et al., 2009; Kleidon and Mooney, 2000; Swenson et al.,
2012), this trend is intriguing.
It is interesting that most of the trends we predicted on the
basis of functional traits based on plant ecology show up in the
bacterial world. It would be relevant also to attempt to measure trait
diversity along an aridity gradient more directly in soil bacteria,
perhaps through catabolic profiling (Degens et al., 2001) and other
measures of soil activity (Hector and Bagchi, 2007; Jing et al.,
2015). It is also interesting to consider whether the more arid soils
with low gene function diversity are less stable or less resilient
against perturbation.
It would also be interesting to study whether other apparent
‘stress’ gradients (e.g. gradients towards extreme salinity, extreme
heat, and high concentrations of chemical poisons) are associated
with reduced gene functional diversity towards the more extreme
136
end of the gradient. Also, a systematic global comparison of
metagenomes along broader aridity gradients – expanding on fairly
small number of samples in the preliminary scale study by Fierer et
al. (2012) would be rewarding in understanding the true extent and
scale of these trends, especially if this trend actually differs on a
regional basis.
Our study did not concentrate on archaeal and fungal gene
functional diversity due to the far smaller numbers of reads
obtained, which obviated robust results. However, there are hints of
such a trend in fungi at least (Figure 29). Sequencing metagenomes
at greater depth would be interesting from the point of view of
examining functional diversity trends in these other organisms.
137
CHAPTER 4: General Conclusions
138
4. 1. Changes in metagenome structure of
microbial communities along environmental
gradients
Nutrient concentration, time succession and aridity play a
significant role in structuring ecosystem. Our study aimed to see
the effects of nutrient concentration and incubation time on
functional gene composition in a culture system (Chapter 2), and
the effects of aridity stress on functional gene composition in
natural terrestrial ecosystem (Chapter 3). The results of this study
have implications for understanding microbial community ecology in
general, and will form the basis for more sophisticated
interpretation of patterns in community composition, soil function,
and biogeochemical fluxes.
In both of the studies, we were able to identify several gene
categories that have positive/negative correlations with the
environmental gradients that generally supports hypothesis based
on the study of macroorganisms, and previous studies on
microorganisms. These different gene-based strategies may help
to explain how and why so many bacterial OTUs coexist in nature,
and the functional principles dominating natural communities.
139
In this study, we also found that the correlation between
functional genes and nutrient concentration could be different in
culture system and in natural environment. The microcosm study
has an advantage in that other factors than nutrient could be
controlled enabling us to find the pure effect of nutrient on
functional composition of bacteria. As bacteria grow with incubation,
there might be some changes in environmental factors other than
nutrient. Even if there are, those changes would be negligible
compared to the initial variation in nutrient concentration.
Microcosm study, however, cannot represent the real
natural system. There are tremendous factors co-varying in natural
environment. The field study in Tibetan plateau has its significance
with this point of view. In most cases, moisturous conditions tend to
possess large pool of nutrients in natural environment. Moisturous
conditions promote plant growth eventually increasing soil nutrient
in general. So, it might be natural to find mixed pattern in here and
the result from the field study in Tibetan plateau could be more
believable than the result from the microcosm study when it comes
to realisticity. However, as mentioned earlier, since a lot of factors
co-vary in the natural environment, it is not clear whether the
correlation between precipitation gradient and relative abundance of
140
gene categories we found in the field study in Tibetan Plateau is a
direct cause-and-effect relationship.
In both of the studies, we found strong correlation between
functional gene richness and OTU richness, supporting the
hypothesis that greater taxonomic diversity is associated with
greater functional diversity, with possible implications for
ecosystem stability. However, this relationship could be highly
dependent on sample size. In the microcosm study, functional gene
richness tended to increase as OTU richness increased, but
somehow reached a plateau at the end. If there were more samples
with large number of OTUs, there could have been a clear evidence
of functional redundancy.
Some follow-up studies or experiments might be necessary
for further justification of findings. For example, qPCR (quantitative
PCR) of functional genes of interest will be helpful since absolute
count data can give a better understanding of the intact relationship
between environmental factors and functional genes.
141
4. 2. Future directions
4. 2. 1. Study of various environmental gradients
influencing microbial community
A range of environmental factors can influence microbial
community composition. Researchers reported that the most
important factor for shaping microbial community is pH, followed by
organic carbon quality and quantity, soil oxygen and redox status,
soil moisture availability, and nitrogen and phosphorous availability
(Fierer, 2017). In a future research, it will be interesting to
investigate the functional relationship of bacterial genes with those
factors, especially pH, which has been established to have the
greatest effect on structuring microbial community.
4. 2. 2. Metatranscriptomics
The metagenome derived from DNA sequences alone gives
an incomplete picture of the functional capabilities and activity of
soil community. Not all cells are active at any one time, and not all
genes in a cell are actively expressed. In addition, dead cells may
persist, with their DNA relatively intact, even though they are not in
any sense a functional part of the community. The short-lived
metatranscriptome derived from RNA, however, gives a clearer
142
picture of which genes are active at any one time. As such, it can
provide an additional perspective that compares the activity vs
inactivity patterns within the community.
There has been only few studies, which attempted to
elucidate the whole metatranscriptome. This is very hard to do this
type of study using current technology. Ribosomal RNA consists
over 90% of cellular RNA (Wilhelm and Landry, 2009). It is easy to
enrich mRNAs when we study eukaryotes since eukaryotic mRNAs
have poly A-tail (Zhao et al., 1999), which enables us to do PCR
with oligo-dT primer targeting mRNAs. Since prokaryotic mRNAs
are not polyadenylated, there has been several methods developed
to remove ribosomal RNA from bulk RNA (He et al., 2010; Pang et
al., 2004), but they are not efficient enough and expensive to be
used wildly. With the development of mRNA enrichment technology,
a new era will begin in microbial ecology that enables us to find
intact connection between microbial functional activity and
environmental conditions.
143
Publication
Ho-Kyung Song, Woojin Song, Mincheol Kim, Binu M. Tripathi,
Hyoki Kim, Piotr Jablonski, Jonathan M. Adams; Bacterial strategies
along nutrient and time gradients, revealed by metagenomic
analysis of laboratory microcosms, FEMS Microbiology Ecology,
Volume 93, Issue 10, 1 October 2017
144
Manuscript in preparation
Ho-Kyung Song, Yu Shi, Teng Yang, Jin-Sheng He, Hyoki Kim,
Piotr Jablonski, Haiyan Chu, Jonathan M. Adams; Environmental
filtering of bacterial trait diversity along an aridity gradient. In
preparation.
145
References
Amann, R.I., Ludwig, W., and Schleifer, K.-H. (1995). Phylogenetic
identification and in situ detection of individual microbial cells
without cultivation. Microbiological reviews 59, 143-169.
Ashraf, M., and Foolad, M.R. (2007). Roles of glycine betaine and
proline in improving plant abiotic stress resistance. Environmental
and Experimental Botany 59, 206-216.
Bayles, K.W. (2007). The biological role of death and lysis in biofilm
development. Nature reviews Microbiology 5, 721-726.
Bazzaz, F.A. (1979). The physiological ecology of plant succession.
Annual Review of Ecology and Systematics 10, 351-371.
Belnap, J., and Gillette, D.A. (1998). Vulnerability of desert biological
soil crusts to wind erosion: the influences of crust development, soil
texture, and disturbance. Journal of Arid Environments 39, 133-142.
Benson, D. A., Cavanaugh, M., Clark, K., Karsch-Mizrachi, I., Lipman,
D. J., Ostell, J., & Sayers, E. W. (2012). GenBank. Nucleic acids
research, 41(D1), D36-D42.
Blazewicz, S.J., Barnard, R.L., Daly, R.A., and Firestone, M.K. (2013).
Evaluating rRNA as an indicator of microbial activity in
environmental communities: limitations and uses. The ISME Journal 7,
2061-2068.
Brook, I. (1999). Bacterial Interference. Critical Reviews in
Microbiology 25, 155-172.
Burke, C., Steinberg, P., Rusch, D., Kjelleberg, S., & Thomas, T.
(2011). Bacterial community assembly based on functional genes
146
rather than species. Proceedings of the National Academy of
Sciences, 108(34), 14288-14293.
Cadotte, M.W., Carscadden, K., and Mirotchnick, N. (2011). Beyond
species: functional diversity and the maintenance of ecological
processes and services. Journal of Applied Ecology 48, 1079-1087.
Carini, P., Marsden, P.J., Leff, J.W., Morgan, E.E., Strickland, M.S.,
and Fierer, N. (2016). Relic DNA is abundant in soil and obscures
estimates of soil microbial diversity. 2, 16242.
Cederlund, H., Wessén, E., Enwall, K., Jones, C.M., Juhanson, J., Pell,
M., Philippot, L., and Hallin, S. (2014). Soil carbon quality and
nitrogen fertilization structure bacterial communities with predictable
responses of major bacterial phyla. Applied soil ecology 84, 62-68.
Choi, S., Song, H., Tripathi, B. M., Kerfahi, D., Kim, H., & Adams, J.
M. (2017). Effect of experimental soil disturbance and recovery on
structure and function of soil community: a metagenomic and
metagenetic approach. Scientific Reports, 7.
Church, M.J., Hutchins, D.A., and Ducklow, H.W. (2000). Limitation of
bacterial growth by dissolved organic matter and iron in the Southern
ocean. Applied and Environmental Microbiology 66, 455-466.
Clark, J.S., Campbell, J.H., Grizzle, H., Acosta-Martìnez, V., and Zak,
J.C. (2009). Soil Microbial Community Response to Drought and
Precipitation Variability in the Chihuahuan Desert. Microbial Ecology
57, 248-260.
Condon, C., Liveris, D., Squires, C., Schwartz, I., and Squires, C.L.
(1995). rRNA operon multiplicity in Escherichia coli and the
physiological implications of rrn inactivation. Journal of Bacteriology
147
177, 4152-4156.
Connell, J.H., and Slatyer, R.O. (1977). Mechanisms of succession in
natural communities and their role in community stability and
organization. The American Naturalist 111, 1119-1144.
Csonka, L.N. (1989). Physiological and genetic responses of bacteria
to osmotic stress. Microbiological Reviews 53, 121-147.
Degens, B., and Harris, J. (1997). Development of a physiological
approach to measuring the catabolic diversity of soil microbial
communities. Soil Biology and Biochemistry 29, 1309-1320.
Degens, B.P., Schipper, L.A., Sparling, G.P., and Duncan, L.C. (2001).
Is the microbial community in a soil with reduced catabolic diversity
less resistant to stress or disturbance? Soil Biology and Biochemistry
33, 1143-1153.
Fegatella, F., Lim, J., Kjelleberg, S., and Cavicchioli, R. (1998).
Implications of rRNA operon copy number and ribosome content in
the marine oligotrophic ultramicrobacterium Sphingomonassp. strain
RB2256. Applied and Environmental Microbiology 64, 4433-4438.
Fierer, N. (2017). Embracing the unknown: disentangling the
complexities of the soil microbiome. Nat Rev Micro 15, 579-590.
Fierer, N., Bradford, M.A., and Jackson, R.B. (2007). Toward an
ecological classification of soil bacteria. Ecology 88, 1354-1364.
Fierer, N., Leff, J.W., Adams, B.J., Nielsen, U.N., Bates, S.T., Lauber,
C.L., Owens, S., Gilbert, J.A., Wall, D.H., and Caporaso, J.G. (2012).
Cross-biome metagenomic analyses of soil microbial communities
and their functional attributes. Proceedings of the National Academy
of Sciences 109, 21390-21395.
148
Fierer, N., Strickland, M.S., Liptzin, D., Bradford, M.A., and Cleveland,
C.C. (2009). Global patterns in belowground communities. Ecology
Letters 12, 1238-1249.
Flynn, D.F., Gogol‐Prokurat, M., Nogeire, T., Molinari, N., Richers,
B.T., Lin, B.B., Simpson, N., Mayfield, M.M., and DeClerck, F. (2009).
Loss of functional diversity under land use intensification across
multiple taxa. Ecology letters 12, 22-33.
Griffiths, B.S., and Philippot, L. (2013). Insights into the resistance
and resilience of the soil microbial community. FEMS Microbiology
Reviews 37, 112-129.
Griffiths, R.I., Thomson, B.C., James, P., Bell, T., Bailey, M., and
Whiteley, A.S. (2011). The bacterial biogeography of British soils.
Environmental microbiology 13, 1642-1654.
Grime, J.P. (1977). Evidence for the existence of three primary
strategies in plants and its relevance to ecological and evolutionary
theory. The American Naturalist 111, 1169-1194.
Grime, J.P. (1979). Plant Strategies and Vegetation Processes (John
Wiley & Sons, Chichester).
Grime, J.P. (2006). Plant strategies, vegetation processes, and
ecosystem properties (John Wiley & Sons).
Hadley, N.F. (1974). Adaptational Biology of Desert Scorpions. The
Journal of Arachnology 2, 11-23.
He, S., Wurtzel, O., Singh, K., Froula, J.L., Yilmaz, S., Tringe, S.G.,
Wang, Z., Chen, F., Lindquist, E.A., and Sorek, R. (2010). Validation
of two ribosomal RNA removal methods for microbial
metatranscriptomics. Nature methods 7, 807-812.
149
Hector, A., and Bagchi, R. (2007). Biodiversity and ecosystem
multifunctionality. Nature 448, 188-190.
Hibbing, M.E., Fuqua, C., Parsek, M.R., and Peterson, S.B. (2010).
Bacterial competition: surviving and thriving in the microbial jungle.
Nature reviews Microbiology 8, 15-25.
Hijmans, R.J., Cameron, S.E., Parra, J.L., Jones, P.G., and Jarvis, A.
(2005). Very high resolution interpolated climate surfaces for global
land areas. International Journal of Climatology 25, 1965-1978.
Ho, A., Angel, R., Veraart, A.J., Daebeler, A., Jia, Z., Kim, S.Y.,
Kerckhof, F.-M., Boon, N., and Bodelier, P.L.E. (2016). Biotic
interactions in microbial communities as modulators of
biogeochemical processes: methanotrophy as a model system.
Frontiers in Microbiology 7, 1285.
Hooper, D., Solan, M., Symstad, A., Diaz, S., Gessner, M., Buchmann,
N., Degrange, V., Grime, P., Hulot, F., and Mermillod-Blondin, F.
(2002). Species diversity, functional diversity and ecosystem
functioning. Biodiversity and Ecosystem Functioning: Syntheses and
Perspectives 17, 195-208.
Hooper, D.U., and Vitousek, P.M. (1997). The effects of plant
composition and diversity on ecosystem processes. Science 277,
1302-1305.
Hu, M.-x., Zhang, X., Li, E.-l., and Feng, Y.-J. (2010). Recent
advancements in toxin and antitoxin systems involved in bacterial
programmed cell death. International Journal of Microbiology 2010.
Huston, M. (1979). A General Hypothesis of Species Diversity. The
American Naturalist 113, 81-101.
150
Huston, M., and Smith, T. (1987). Plant succession: life history and
competition. The American Naturalist 130, 168-198.
Huston, M.A. (1994). Biological Diversity: the Coexistence of Species
(Cambridge, UK: Cambridge University Press).
Huston, M.A. (1997). Hidden treatments in ecological experiments:
re-evaluating the ecosystem function of biodiversity. Oecologia 110,
449-460.
Jean-Pierre, B., and Jean-Marie, L. (2007). Osmoregulation in the
Periplasm. In The Periplasm, E. M, ed. (Washington, DC: American
Society of Microbiology).
Jens Efsen, J., Jarone, P., Nicholas, B., Ulla Li, Z., and Åke,
H.Â.m. (2002). Variability in motility characteristics among marine
bacteria. Aquatic Microbial Ecology 28, 229-237.
Jiang, L., and Patel, S.N. (2008). Community assembly in the
presence of disturbance: a microcosm experiment. Ecology 89,
1931-1940.
Jing, X., Sanders, N.J., Shi, Y., Chu, H., Classen, A.T., Zhao, K., Chen,
L., Shi, Y., Jiang, Y., and He, J.-S. (2015). The links between
ecosystem multifunctionality and above- and belowground
biodiversity are mediated by climate. 6, 8159.
Jones, D.L., and Willett, V.B. (2006). Experimental evaluation of
methods to quantify dissolved organic nitrogen (DON) and dissolved
organic carbon (DOC) in soil. Soil Biology and Biochemistry 38, 991-
999.
Kanehisa, M. (2002). The KEGG database. silico simulation of
biological processes, 247, 91-103.
151
Kemp, P.R., Cornelius, J.M., and Reynolds, J.F. (1992). A simple
model for predicting soil temperatures in desert ecosystems. Soil
Science 153, 280-287.
Kirk, J.L., Beaudette, L.A., Hart, M., Moutoglis, P., Klironomos, J.N.,
Lee, H., and Trevors, J.T. (2004). Methods of studying soil microbial
diversity. Journal of microbiological methods 58, 169-188.
Klappenbach, J.A., Dunbar, J.M., and Schmidt, T.M. (2000). rRNA
operon copy number reflects ecological strategies of bacteria.
Applied and Environmental Microbiology 66, 1328-1333.
Kleidon, A., Adams, J., Pavlick, R., and Reu, B. (2009). Simulated
geographic variations of plant species richness, evenness and
abundance using climatic constraints on plant functional diversity.
Environmental Research Letters 4, 014007.
Kleidon, A., and Mooney, H.A. (2000). A global distribution of
biodiversity inferred from climatic constraints: results from a
process-based modelling study. Global Change Biology 6, 507-523.
Koch, A.L. (2001). Oligotrophs versus copiotrophs. BioEssays 23,
657-661.
Kreft, H., & Jetz, W. (2007). Global patterns and determinants of
vascular plant diversity. Proceedings of the National Academy of
Sciences, 104(14), 5925-5930.
Lauber, C.L., Hamady, M., Knight, R., and Fierer, N. (2009). Soil pH
as a predictor of soil bacterial community structure at the continental
scale: a pyrosequencing-based assessment. Applied and
Environmental Microbiology.
Lauro, F.M., McDougald, D., Thomas, T., Williams, T.J., Egan, S., Rice,
152
S., DeMaere, M.Z., Ting, L., Ertan, H., Johnson, J., et al. (2009). The
genomic basis of trophic strategy in marine bacteria. Proceedings of
the National Academy of Sciences 106, 15527-15533.
Luckinbill, L.S. (1979). Regulation, stability, and diversity in a model
experimental microcosm. Ecology 60, 1098-1102.
Ma, W., He, J.-S., Yang, Y., Wang, X., Liang, C., Anwar, M., Zeng, H.,
Fang, J. and Schmid, B. (2010), Environmental factors covary with
plant diversity–productivity relationships among Chinese grassland
sites. Global Ecology and Biogeography, 19: 233–243.
doi:10.1111/j.1466-8238.2009.00508.x
MacArthur, R.H., and Wilson, E.O. (1967). The Theory of Island
Biogeography (New Jersey, United States: Princeton University
Press).
Major, J. (1963). A climatic index to vascular plant activity. Ecology,
44(3), 485-498.
Mayfield, M.M., Boni, M.F., Daily, G.C., and Ackerly, D. (2005).
Species and functional diversity of native and human-dominated plant
communities. Ecology 86, 2365-2372.
McCann, K.S. (2000). The diversity-stability debate. Nature 405,
228-233.
Mendes, L. W., Tsai, S. M., Navarrete, A. A., De Hollander, M., van
Veen, J. A., & Kuramae, E. E. (2015). Soil-borne microbiome: linking
diversity to function. Microbial ecology, 70(1), 255-265.
Meyer, F., Paarmann, D., D'Souza, M., Olson, R., Glass, E., Kubal, M.,
Paczian, T., Rodriguez, A., Stevens, R., Wilke, A., et al. (2008). The
metagenomics RAST server – a public resource for the automatic
153
phylogenetic and functional analysis of metagenomes. BMC
Bioinformatics 9, 386.
Michalzik, B., Kalbitz, K., Park, J.-H., Solinger, S., and Matzner, E.
(2001). Fluxes and concentrations of dissolved organic carbon and
nitrogen – a synthesis for temperate forests. Biogeochemistry 52,
173-205.
Morriën, E., Hannula, S.E., Snoek, L.B., Helmsing, N.R., Zweers, H.,
de Hollander, M., Soto, R.L., Bouffaud, M.-L., Buée, M., Dimmers, W.,
et al. (2017). Soil networks become more connected and take up
more carbon as nature restoration progresses. Nature
Communications 8, 14349.
Nemergut, D.R., Costello, E.K., Hamady, M., Lozupone, C., Jiang, L.,
Schmidt, S.K., Fierer, N., Townsend, A.R., Cleveland, C.C., and
Stanish, L. (2011). Global patterns in the biogeography of bacterial
taxa. Environmental microbiology 13, 135-144.
Nunes, A., Köbel, M., Pinho, P., Matos, P., Bello, F.d., Correia, O., and
Branquinho, C. (2017). Which plant traits respond to aridity? A
critical step to assess functional diversity in Mediterranean drylands.
Agricultural and Forest Meteorology 239, 176-184.
Odum, E.P. (1969). The strategy of ecosystem development. Science
164, 262-270.
Oliverio, A.M., Bradford, M.A., and Fierer, N. (2017). Identifying the
microbial taxa that consistently respond to soil warming across time
and space. Global change biology 23, 2117-2129.
Overbeek, R., Begley, T., Butler, R.M., Choudhuri, J.V., Chuang, H.-
Y., Cohoon, M., de Crécy-Lagard, V., Diaz, N., Disz, T., Edwards, R.,
154
et al. (2005). The subsystems approach to genome annotation and its
use in the project to annotate 1000 genomes. Nucleic Acids Research
33, 5691-5702.
Pang, X., Zhou, D., Song, Y., Pei, D., Wang, J., Guo, Z., and Yang, R.
(2004). Bacterial mRNA Purification by Magnetic Capture‐
Hybridization Method. Microbiology and immunology 48, 91-96.
Paul, E.A. (2014). Soil microbiology, ecology and biochemistry
(Academic press).
Peroni, P.A. (1994). Seed size and dispersal potential of Acer rubrum
(Aceraceae) samaras produced by populations in early and late
successional environments. American Journal of Botany 81, 1428-
1434.
Petchey, O.L., and Gaston, K.J. (2002). Functional diversity (FD),
species richness and community composition. Ecology letters 5, 402-
411.
Petchey, O.L., and Gaston, K.J. (2006). Functional diversity: back to
basics and looking forward. Ecology letters 9, 741-758.
Pett-Ridge, J., and Firestone, M. (2005). Redox fluctuation structures
microbial communities in a wet tropical soil. Applied and
environmental microbiology 71, 6998-7007.
Pianka, E.R. (1970). On r- and K-Selection. The American Naturalist
104, 592-597.
Pruitt, K.D., Tatusova, T., and Maglott, D.R. (2007). NCBI reference
sequences (RefSeq): a curated non-redundant sequence database of
genomes, transcripts and proteins. Nucleic Acids Research 35, D61-
D65.
155
Pueyo, Y., Kefi, S., Alados, C.L., and Rietkerk, M. (2008). Dispersal
strategies and spatial organization of vegetation in arid ecosystems.
Oikos 117, 1522-1532.
Rice, E.W., Baird, R.B., Eaton, A.D., and Clesceri, L.S. (2012).
Standard Methods for the Examination of Water and Wastewater,
22nd edn (American Public Health Association, American Water
Works Association, Water Environment Federation).
Ricklefs, R.E., and Miller, G.L. (1999). Ecology, 4th edn (W. H.
Freeman).
Rousk, J., Bååth, E., Brookes, P.C., Lauber, C.L., Lozupone, C.,
Caporaso, J.G., Knight, R., and Fierer, N. (2010). Soil bacterial and
fungal communities across a pH gradient in an arable soil. The ISME
journal 4, 1340-1351.
Schimel, J., Balser, T.C., and Wallenstein, M. (2007). Microbial
stress-response physiology and its implications for ecosystem
function. Ecology 88, 1386-1394.
Schwinning, S., and Sala, O.E. (2004). Hierarchy of responses to
resource pulses in arid and semi-arid ecosystems. Oecologia 141,
211-220.
Sekercioglu, C.H. (2012). Bird functional diversity and ecosystem
services in tropical forests, agroforests and agricultural areas.
Journal of Ornithology 153, 153-161.
Shapiro, J.A. (1998). Thinking about bacterial populations as
multicellular organisms. Annual review of microbiology 52, 81-104.
Singh, D., Lee-Cruz, L., Kim, W.-S., Kerfahi, D., Chun, J.-H., and
Adams, J.M. (2014). Strong elevational trends in soil bacterial
156
community composition on Mt. Halla, South Korea. Soil Biology and
Biochemistry 68, 140-149.
Song, W., Kim, M., Tripathi, B.M., Kim, H., and Adams, J.M. (2016).
Predictable communities of soil bacteria in relation to nutrient
concentration and successional stage in a laboratory culture
experiment. Environmental Microbiology 18, 1740-1753.
Southwood, T.R.E. (1977). Habitat, the templet for ecological
strategies? Journal of Animal Ecology 46, 337-365.
Sul, W.J., Asuming-Brempong, S., Wang, Q., Tourlousse, D.M.,
Penton, C.R., Deng, Y., Rodrigues, J.L., Adiku, S.G., Jones, J.W., and
Zhou, J. (2013). Tropical agricultural land management influences on
soil microbial communities through its effect on soil organic carbon.
Soil Biology and Biochemistry 65, 33-38.
Swenson, N.G., Enquist, B.J., Pither, J., Kerkhoff, A.J., Boyle, B.,
Weiser, M.D., Elser, J.J., Fagan, W.F., Forero-Montaña, J., Fyllas, N.,
et al. (2012). The biogeography and filtering of woody plant
functional diversity in North and South America. Global Ecology and
Biogeography 21, 798-808.
Tatusov, R.L., Fedorova, N.D., Jackson, J.D., Jacobs, A.R., Kiryutin,
B., Koonin, E.V., Krylov, D.M., Mazumder, R., Mekhedov, S.L.,
Nikolskaya, A.N., et al. (2003). The COG database: an updated
version includes eukaryotes. BMC Bioinformatics 4, 41.
Tilman, D., Knops, J., Wedin, D., Reich, P., Ritchie, M., and Siemann,
E. (1997). The influence of functional diversity and composition on
ecosystem processes. Science 277, 1300-1302.
Torsvik, V., and Øvreås, L. (2002). Microbial diversity and function in
157
soil: from genes to ecosystems. Current opinion in microbiology 5,
240-245.
Tripathi, B.M., Kim, M., Lai-Hoe, A., Shukor, N.A.A., Rahim, R.A., Go,
R., and Adams, J.M. (2013). pH dominates variation in tropical soil
archaeal diversity and community structure. FEMS Microbiology
Ecology 86, 303-311.
Tripathi, B.M., Lee-Cruz, L., Kim, M., Singh, D., Go, R., Shukor, N.A.,
Husni, M., Chun, J., and Adams, J.M. (2014). Spatial scaling effects on
soil bacterial communities in Malaysian tropical forests. Microbial
ecology 68, 247-258.
Tripathi, B.M., Moroenyane, I., Sherman, C., Lee, Y.K., Adams, J.M.,
and Steinberger, Y. (2017). Trends in Taxonomic and Functional
Composition of Soil Microbiome Along a Precipitation Gradient in
Israel. Microbial Ecology 74, 168-176.
van Gestel, N.C., Schwilk, D.W., Tissue, D.T., and Zak, J.C. (2011).
Reductions in daily soil temperature variability increase soil
microbial biomass C and decrease soil N availability in the
Chihuahuan Desert: potential implications for ecosystem C and N
fluxes. Global Change Biology 17, 3564-3576.
Vance, R.R. (1984). Interference competition and the coexistence of
two competitors on a single limiting resource. Ecology 65, 1349-
1357.
Venable, D.L., Flores-Martinez, A., Muller-Landau, H.C., Barron-
Gafford, G., and Becerra, J.X. (2008). Seed dispersal of desert
annuals. Ecology 89, 2218-2227.
Vos, M., Wolf, A.B., Jennings, S.J., and Kowalchuk, G.A. (2013).
158
Micro-scale determinants of bacterial diversity in soil. FEMS
Microbiology Reviews 37, 936-954.
Wardle, D.A., Bonner, K.I., Barker, G.M., Yeates, G.W., Nicholson,
K.S., Bardgett, R.D., Watson, R.N., and Ghani, A. (1999). Plant
removals in perennial grassland: vegetation dynamics, decomposers,
soil biodiversity, and ecosystem properties. Ecological Monographs
69, 535-568.
Westoby, M., Falster, D.S., Moles, A.T., Vesk, P.A., and Wright, I.J.
(2002). Plant ecological strategies: some leading dimensions of
variation between species. Annual Review of Ecology and
Systematics 33, 125-159.
Wilhelm, B.T., and Landry, J.-R. (2009). RNA-Seq—quantitative
measurement of expression through massively parallel RNA-
sequencing. Methods 48, 249-257.
Yang, T., Adams, J. M., Shi, Y., Sun, H., Cheng, L., Zhang, Y., & Chu,
H. (2017a). Fungal community assemblages in a high elevation desert
environment: Absence of dispersal limitation and edaphic effects in
surface soil. Soil Biology and Biochemistry, 115, 393-402.
Yang, T., Adams, J. M., Shi, Y., He, J.-s., Jing, X., Chen, L., Tedersoo,
L. and Chu, H. (2017b), Soil fungal diversity in natural grasslands of
the Tibetan Plateau: associations with plant diversity and
productivity. New Phytol, 215: 756–765. doi:10.1111/nph.14606
Yeates, C., Gillings, M., Davison, A., Altavilla, N., and Veal, D. (1998).
Methods for microbial DNA extraction from soil for PCR amplification.
Biological procedures online 1, 40.
Zhao, J., Hyman, L., and Moore, C. (1999). Formation of mRNA 3′
159
ends in eukaryotes: mechanism, regulation, and interrelationships
with other steps in mRNA synthesis. Microbiology and Molecular
Biology Reviews 63, 405-445.
160
Appendix
Table A1. Environmental variables listed by sample ID. Long,
longitude (°E); Lat, latitude (°N), Ele.m, elevation (m); MAT,
mean annual temperature (°C); GST, growing season temperature
(°C); MAP, mean annual precipitation (mm); GSP, growing season
precipitation(mm); PET, potential evapotranspiration (mm/year);
AET, actual evapotranspiration (mm/year); SM, soil moisture (g/g
of dried soil); pH, soil pH; STC, soil total carbon (%); SOC, soil
organic carbon (%); STN, soil total nitrogen (%); STP, soil total
phosphorous (%); CaCO3, soil CaCO3 (%); SSTC, soil carbon
density (g/m-2); SSOC, soil organic carbon density (g/m-2); SSTN,
soil total nitrogen density (g/m-2); SSTP, soil phosphorous density
(g/m-2); SCaCO3, soil CaCO3 density (g/m-2); SCNratio, soil C:N
ratio; SNPratio, soil N:P ratio; DOC, dissolved organic carbon
(mg/kg of dried soil); DTN, dissolved total nitrogen (mg/kg of dried
soil); DON, dissolved organic nitrogen (mg/kg of dried soil); NO3N,
nitrate nitrogen (mg/kg of dried soil); NH4N, ammonium nitrogen
(mg/kg of dried soil); TAN, total available nitrogen (mg/kg of dried
soil); SBD, soil bulk density (g/cm-3); AGB, above ground biomass
(g/m-2); BGB, root biomass at 5cm soil depth (g/m-2); PTC, percent
total carbon in aboveground biomass (%); PTN, percent total
nitrogen in aboveground biomass (%); PTP, percent total
161
phosphorous in aboveground biomass (%); PCNratio, C:N ratio in
aboveground biomass; PNPratio, N:P ratio in aboveground biomass;
Vege.type, vegetation type (AM: alpine meadow; AS: alpine steppe;
DS: desert steppe)
162
Table A1. (Continued)
ID Long Lat Ele.m MAT GST MAP GSP PET AET SM
T1 100.9 36.3 3282 4.7 13.5 321.3 245.6 495 337 0.15
T2 96.7 32.9 4258 3 9.9 488.9 356.4 428 428 1.1
T3 96.4 33.3 4284 0.4 8.4 484.9 346.5 388 388 0.51
T4 96 33.6 4234 -1 7.6 457.9 326.8 349 349 0.71
T5 96.2 34.1 4366 -2.2 7.2 418.3 289.4 345 345 0.53
T6 99.2 35.4 4162 -0.8 8.8 330.2 243.1 373 345 0.07
T7 99.5 35.4 4089 0 9.5 342.5 254.3 398 360 0.08
T8 101 36.9 3066 3.2 12.7 386.8 269.6 469 355 0.08
T9 99.6 37.1 3241 0.3 10.6 309 234.8 412 291 0.07
T10 98.9 36.9 3297 2.2 12 224.5 182 450 240 0.14
T11 98.4 37 3130 3.8 13 219.6 166.6 476 217 0.03
T12 98.4 37.3 3425 2.7 12.6 235.9 177 461 219 0.06
T13 97.1 37.4 2918 3.9 14.5 147.6 108.9 494 155 0.01
T14 95.9 37.5 3319 3.4 14.7 78.2 61.6 489 93 0
T15 95.5 37.4 3181 3 14.5 66.4 52.1 482 81 0.03
T16 100.5 35.6 3320 2 11.1 404.6 290.1 439 401 0.17
T17 94.5 35.9 3633 2.1 11.7 127.5 102.1 439 153 0.07
T18 93.9 35.5 4598 -1.2 8.6 205.4 153 359 202 0.05
T19 92.9 34.9 4555 -4.9 5.2 291.6 215.6 322 263 0.1
T20 92.1 33.7 4697 -3.4 5.8 336.6 261.9 332 323 0.13
T21 92 32.9 5228 -2.9 6 402.5 302.3 334 334 0.15
T22 91.7 32.2 4632 -2.3 6.4 436.1 326 337 337 0.05
T23 91.3 31.5 4664 -1.2 7.3 435.4 329.3 345 345 0.03
T24 91.5 30.6 4542 1.7 9.8 509.6 374.9 410 410 0.47
T25 90.6 30.2 4557 4.2 11.2 422.3 330.9 453 420 0.63
T26 100.8 34.9 3663 -0.3 9 493.6 347.6 389 389 0.6
T27 100.4 34.5 4250 0 8.2 523.2 363.9 386 386 0.68
T28 99 34.8 4505 -2.9 7 377.1 267.7 344 344 1.05
T29 98.5 34.9 4230 -3.5 6.5 343.7 244.7 339 338 0.09
163
Table A1. (Continued)
ID pH STC SOC STN STP CaCO3 SSTC SSOC
T1 7.14 3.25 3.09 0.34 0.07 1.32 3245 3086.44
T2 6.97 15.22 15.21 1.3 0.12 0.08 15220 15210
T3 6.46 9.35 9.33 0.7 0.07 0.13 9345 9329.42
T4 7.22 11.41 11.39 0.88 0.09 0.19 11410 11387.07
T5 7.23 9.12 9.09 0.78 0.06 0.25 9120 9089.73
T6 8.4 3.08 2.07 0.16 0.06 8.46 3080 2065.12
T7 8.04 3.58 2.93 0.27 0.06 5.35 3575 2933.11
T8 8.35 3.87 3.27 0.33 0.07 4.95 3865 3270.46
T9 7.43 3.7 2.54 0.2 0.06 9.66 3695 2535.59
T10 7.91 3.34 2.12 0.19 0.07 10.13 3340 2124.49
T11 8.57 2.19 0.72 0.07 0.07 12.29 2190 715
T12 8.17 3.09 2.02 0.25 0.09 5.28 3085 2016.61
T13 8.43 2.81 0.97 0.07 0.07 15.31 2805 967.74
T14 8.66 1.85 0.66 0.02 0.06 9.92 1850 659.62
T15 7.92 1.12 0.25 0.03 0.04 7.27 1120 247.14
T16 7.8 5.09 4.85 0.45 0.08 1.96 5090 4854.52
T17 8.87 1.86 0.76 0.06 0.06 9.22 1863.33 757.53
T18 8.99 1.33 0.81 0.07 0.04 4.28 1325 811.77
T19 8.64 1.22 0.49 0.03 0.02 6.05 1215 489.34
T20 7.98 2.48 1.92 0.16 0.04 4.71 2480 1915.16
T21 7.61 2.18 2.18 0.16 0.05 0 2175 2175
T22 8.24 2.47 1.45 0.09 0.02 8.49 2465 1446.21
T23 7.27 2.14 2.14 0.2 0.04 0 2135 2135
T24 6.03 11.21 11.2 0.66 0.07 0.04 11205 11199.98
T25 8.06 7.33 7.2 0.54 0.07 1.07 7325 7197
T26 7.77 13.57 13.53 1.11 0.11 0.29 13565 13530.09
T27 6.81 9.55 9.54 0.76 0.09 0.12 9550 9535.04
T28 7.8 11.98 11 0.94 0.08 8.15 11980 11001.65
T29 9.15 2.37 1.19 0.07 0.05 9.77 2365 1192.66
164
Table A1. (Continued)
ID SSTN SSTP SCaCO3 SCNratio SNPratio DOC DTN
T1 340 65.56 738.93 9.54 5.19 98.85 55.61
T2 1295 119.62 18.97 11.75 10.83 353.67 232.25
T3 695 73.66 45.34 13.45 9.44 610.07 109.57
T4 880 86.8 76.78 12.97 10.14 410.94 83.86
T5 775 63.9 88.53 11.77 12.13 423.81 69.35
T6 155 64.41 4378.51 19.87 2.41 112.95 21
T7 265 60.27 2651.61 13.49 4.4 201.74 23.82
T8 330 67.71 3020.83 11.71 4.87 148.65 16.83
T9 200 57.26 5215.11 18.48 3.49 259.34 17.81
T10 190 67.43 5913.64 17.58 2.82 232.48 62.13
T11 65 70.69 7553.41 33.69 0.92 97.27 10.72
T12 250 91.8 2742.92 12.34 2.72 202.72 28.16
T13 70 68.27 8710.38 40.07 1.03 327.28 39.96
T14 20 61.19 7346.49 75 0.33 311.91 12.11
T15 30 40.01 4938.94 37.33 0.75 175.43 13.24
T16 445 77.06 842.38 11.44 5.77 242.77 73.32
T17 56.67 61.11 6920.21 32.88 0.93 142.97 12.29
T18 70 42.85 2300.28 18.93 1.63 146.63 8.01
T19 30 21.18 4075.41 40.5 1.42 243.52 28.87
T20 160 40.68 2821.75 15.5 3.93 261.12 34.62
T21 160 46.71 0 13.59 3.43 190.71 20.06
T22 85 21.25 4578.88 29 4 151.68 12.48
T23 195 37.54 0 10.95 5.19 163.21 13.91
T24 660 73.43 16.88 16.98 8.99 896.4 63.22
T25 535 73.38 578.7 13.69 7.29 347.12 83.49
T26 1110 106.95 111.32 12.22 10.38 320.42 190.91
T27 755 88.88 32.78 12.65 8.49 78.01 120.91
T28 940 79.55 1767.97 12.74 11.82 84 98.56
T29 65 49.77 4969.57 36.38 1.31 36.27 9.05
165
Table A1. (Continued)
ID DON NO3N NH4N TAN SBD AGB BGB PTC
T1 18.61 25.74 11.26 55.61 1.13 36.65 86.76 45.63
T2 101.65 106.1 24.51 232.25 0.45 63.91 1859.92 38.42
T3 66.77 22.92 19.88 109.57 0.69 67.49 1836.22 44.14
T4 49.69 5.03 29.14 83.86 0.77 44.45 1272.86 40.9
T5 40.31 16.3 12.74 69.35 0.93 78.02 746.5 38.16
T6 12.68 3.76 4.56 21 1.05 40.01 402.67 42.79
T7 10.09 8.8 4.94 23.82 0.99 28.49 428.44 44.24
T8 4.7 6.78 5.36 16.83 1.19 44.45 197.9 41.22
T9 6.29 5.63 5.89 17.81 1.09 110.07 106.52 45.61
T10 9.69 36.26 16.19 62.13 1.22 218.7 822.17 43.25
T11 0.99 1.69 8.04 10.72 1.23 87.08 197.32 38.92
T12 6.76 10.76 10.64 28.16 0.99 175 136.79 43.09
T13 14.12 16.65 9.19 39.96 1.25 187.12 14.37 42.96
T14 1.74 2.88 7.49 12.11 1.48 161.17 1.68 47.96
T15 3.57 1.05 8.61 13.24 1.44 59.83 1.87 41.05
T16 32.29 31.68 9.35 73.32 0.84 211.66 393.62 43.25
T17 1.27 4.01 7.01 12.29 1.59 105.74 19.96 41.5
T18 1.16 0.29 6.56 8.01 1.15 48.05 1268.63 37.6
T19 2.3 5.28 21.29 28.87 1.4 248.16 558.31 38.88
T20 8.15 17.85 8.62 34.62 1.21 118.79 824.27 35.42
T21 4.58 8.54 6.93 20.06 0.52 50.11 783.63 27.88
T22 3.18 2.32 6.97 12.48 1.22 66.13 2071.37 42.18
T23 6.17 0 7.74 13.91 1.16 54.71 3004.02 38
T24 38.13 0.33 24.75 63.22 0.88 41.39 2308.09 42.87
T25 15.66 55.54 12.28 83.49 1.08 283.09 819.99 44.33
T26 76.41 92.99 21.51 190.91 0.6 225.05 812.9 43.18
T27 31.64 62.4 26.87 120.91 0.63 123.85 575.16 42.73
T28 7.84 68.6 22.12 98.56 0.44 44.81 749.62 37.39
T29 3.02 0.5 5.53 9.05 1 54.13 155.5 36.21
166
Table A1. (Continued)
ID PTN PTP PCNratio PNPratio Vege.type
T1 2.65 0.14 17.22 19.26 AS
T2 1.78 0.16 21.58 10.93 AM
T3 1.91 0.18 23.11 10.37 AM
T4 2.74 0.24 14.93 11.52 AM
T5 2.02 0.18 18.89 10.96 AM
T6 1.19 0.1 35.96 12.08 AS
T7 1.8 0.14 24.58 12.83 AS
T8 1.77 0.1 23.29 17.57 AS
T9 1.54 0.11 29.62 13.42 AS
T10 2.49 0.12 17.37 20.08 AS
T11 1.38 0.1 28.2 13.72 DS
T12 1.46 0.12 29.51 12.03 AS
T13 1.12 0.09 38.36 12.12 DS
T14 0.65 0.08 73.78 8.04 DS
T15 1.23 0.09 33.37 13.52 DS
T16 1.9 0.2 22.76 9.7 AM
T17 1.6 0.09 25.94 17.53 DS
T18 1.97 0.14 19.09 13.76 AS
T19 1.79 0.15 21.72 12.16 AM
T20 1.43 0.16 24.77 9.02 AS
T21 1.52 0.18 18.34 8.4 AM
T22 1.69 0.18 24.96 9.63 AM
T23 1.73 0.11 21.97 15.49 AM
T24 1.89 0.12 22.68 15.31 AM
T25 1.61 0.11 27.53 14 AM
T26 1.94 0.17 22.26 11.29 AM
T27 2.07 0.15 20.64 13.89 AM
T28 1.95 0.12 19.17 15.98 AM
T29 1.49 0.14 24.3 10.96 AS
167
국문초록 (Abstract in Korean)
생태학은 생물 간 상호작용 및 생물과 무생물적 요소 사이의 관
계에 대하여 연구하는 학문이다. 최근 몇 년간 핵산 시퀀싱(sequencing)
기술의 발달에 힘입어 우리는 생물권을 향한 새로운 시각을 가지게 되었
다. 앰플리콘(amplicon) 시퀀싱과 메타지놈(metagenome)으로 대두되
는 이러한 방법들은 미생물생태학 분야의 새로운 지평을 열어주었다. 분
류학적 표지 유전자(taxonomic marker gene)을 이용하는 앰플리콘 시
퀀싱 기술을 통해 박테리아, 고세균, 균 및 미소무척추동물이 토양, 퇴적
토, 물 등의 자연 환경에서 얼마나 다양하게 존재하는지를 알 수 있었다.
미생물의 다양성 및 군집 형성과 관련한 생태학적 이론들을 완전히 새로
운 방법으로 검증하는 것도 가능해졌다. 배양 비의존적으로 주어진 환경
에서의 유전자 군을 직접적으로 시퀀싱하는 메타지노믹스
(metagenomics)는 다양한 환경 속 미생물 군집의 잠재적인 기능들을
탐구하는 데에 중추적인 역할을 한다.
메타지놈의 구성과 다양성이 환경 구배에 따라 어떻게 변화하는
지 연구하는 것은 생태계 구조 및 군집 형성에 관한 일반적인 이론을 이
해하는데 도움을 줄 수 있다. 본 연구에서는 초기 양분 농도 및 배양 시
간에 따라 군집 메타지놈이 어떻게 변화하는지 알아보기 위해 토양 용출
액을 이용한 배양 실험을 설계했다. 기능성 메타지놈(functional
metagenome)의 구조는 양분 혹은 배양 시간에 따라 다르게 변화했다.
예를 들어, 고영양(copiotrophic) 조건에서는 세포 분열 및 세포 주기와
관련된 유전자가 많은 것이 확인되었고 빈영양(oligotrophic) 조건에서
168
는 탄수화물 대사 및 독성, 질병 및 방어에 관련한 유전자가 많았다. 배
양 시간과 관련한 변화도 확인할 수 있었는데, 배양 시간이 짧을수록
(r-전략) 조절 및 세포 신호 관련 유전자가 많았고 배양 시간이 길수록
(K-전략) 이동성 및 주화성(chemotaxis) 관련 유전자가 많았다. 다양
성 측면에서는 기능성 유전자(functional gene)의 풍부도(richness)가
기능성 유전자의 분류학적(taxonomic) 풍부도 및 조작 분류 단위
(operational taxonomic unit)의 풍부도와 선형 상관관계를 가지는 것을
확인할 수 있었고, 이는 분류학적 다양성이 높을수록 기능적 다양성이
높아 생태계의 안정성을 가져다 준다고 주장하는 이론들을 뒷받침해준다.
건조함의 정도는 생태계를 구성하는 데에 있어 매우 중요한 요
소 중 하나다. 건조함의 정도가 토양 생물군의 기능적 다양성에 어떤 영
향을 주는지 알아보기 위해 티베트 고원 동부지역에서 샘플을 채취하여
강수량에 따라 토양 메타지놈의 구조가 어떻게 달라지는지 연구했다. 연
구 결과, 연평균 강수량이 낮을수록 박테리아 기능성 유전자의 다양성이
감소하는 경향이 있다는 것을 알 수 있었다. 극한 건조 조건이 토양 생
물군의 기능성 유전자의 다양성을 제한하는 것으로 보이며, 이는 환경
구배에 따른 식물의 기능적 다양성 변화 양상과 흡사하다. 또한, 건조한
지역일수록 휴면 관련된 유전자와 삼투 억제자(osmoprotectant) 관련
유전자가 많이 있는 것도 알 수 있었다. 건조한 지역일수록 항생제 저항
과 관련된 유전자 및 독성 관련 유전자는 적었는데, 이는 극한 조건에서
는 생물 간 상호작용이 크지 않다는 것을 암시한다. 본 연구의 결과는
생물군계에 따른 메타지놈 구조를 비교한 기존 연구의 결과와 흡사하다.
향후에는 산도, 온도, 토양 내 산소 농도 등 다른 종류의 환경
169
구배에 따라 메타지놈 구조가 어떻게 변하는지에 대해 연구해 보는 것도
흥미로울 것으로 보인다. 또한, 메타트랜스크립토믹스
(metatranscriptomics)에 대한 연구를 추가적으로 진행한다면 기능적
활성도(activity)에 대한 보다 정확한 정보를 알 수 있을 것으로 보인다.
이러한 목적을 이루기 위해서는 원핵생물의 전령RNA(messenger RNA)
를 농축할 수 있는 기술 개발이 선행되어야 한다.
주요어: 메타지놈(metagenome), 기능적 다양성, 고영양(copiotrophic),
빈영양(oligotrophic), r-선택, K-선택, 강수량, 건조 정도(aridity), 생
태적 스트레스(ecological stress)
학번: 2016-30109