ARTICLE

Received 15 Apr 2016 | Accepted 16 Nov 2016 | Published 21 Dec 2016 DOI: 10.1038/ncomms13960 OPEN

Nutrient enrichment modifies

temperature-biodiversity relationships
in large-scale field experiments
Jianjun Wang1,2,*, Feiyan Pan3,*, Janne Soininen2, Jani Heino4 & Ji Shen1

Climate effects and human impacts, that is, nutrient enrichment, simultaneously drive spatial
biodiversity patterns. However, there is little consensus about their independent effects on
biodiversity. Here we manipulate nutrient enrichment in aquatic microcosms in subtropical
and subarctic regions (China and Norway, respectively) to show clear segregation of bacterial
species along temperature gradients, and decreasing alpha and gamma diversity toward
higher nutrients. The temperature dependence of species richness is greatest at extreme
nutrient levels, whereas the nutrient dependence of species richness is strongest at
intermediate temperatures. For species turnover rates, temperature effects are strongest at
intermediate and two extreme ends of nutrient gradients in subtropical and subarctic regions,
respectively. Species turnover rates caused by nutrients do not increase toward higher
temperatures. These findings illustrate direct effects of temperature and nutrients on
biodiversity, and indirect effects via primary productivity, thus providing insights into how
nutrient enrichment could alter biodiversity under future climate scenarios.

1 StateKey Laboratory of Lake Science and Environment, Nanjing Institute of Geography and Limnology, Chinese Academic of Sciences, Nanjing 210008,
China. 2 Department of Geosciences and Geography, University of Helsinki, Helsinki FIN-00014, Finland. 3 Jiangsu Key Laboratory for Molecular and Medical
Biotechnology, Nanjing Normal University, Nanjing 210023, China. 4 Finnish Environment Institute, Natural Environment Centre, Biodiversity, Oulu FI-90014,
Finland. * These authors contributed equally to this work. Correspondence and requests for materials should be addressed to J.W.
(email: jjwang@niglas.ac.cn).

NATURE COMMUNICATIONS | 7:13960 | DOI: 10.1038/ncomms13960 | www.nature.com/naturecommunications 1

ARTICLE
S
patial patterns of biodiversity are a core topic in ecology;
however, the mechanisms driving these patterns remain
unclear. Climatic factors, especially temperature, are
regarded as the main drivers underlying diversity gradients over
broad spatial scales. For instance, the positive relationships
between temperature and species richness prevail along gradients
in elevation and latitude, which are explained by numerous
hypotheses, including the metabolic theory of ecology (MTE)13
and productivity-diversity hypothesis2. In the last 100 years, the
Earth has warmed by B0.78 !C, and global mean temperatures
are projected to increase by 4.30.7 !C by the year 2100 (ref. 4).
The changing temperatures may affect species richness because
temperature covaries with primary productivity, limits the
distribution ranges of species and drives speciation rates1,5. The
increased temperatures may favour higher species richness, but
also result in the extinction of endemic species in colder regions,
such as at high elevations and latitudes68.
In addition, human impacts, such as nutrient enrichment, have
been identified as one of the main drivers of biodiversity loss in
recent decades9. For instance, mountainous regions are becoming
increasingly impacted by settlements and transport networks10,
and are facing more intensive forestry practices, agriculture
activities, eutrophication and habitat loss. Higher temperatures
and nutrient enrichment would increase the ecosystem primary
productivity11, which could further affect species richness12.
Thus, the interactions between climate change and human
impacts on biodiversity make it difficult to predict the spatial
patterns of biodiversity13. The typical covariance between climatic
factors and human impacts14,15, such as that along elevational
gradients16, further complicates the evaluation of their
independent roles in determining biodiversity patterns17. The
independent effects of climate change and human impact on
biodiversity patterns have rarely been addressed18,19.
A promising approach to exploring climatic effects is the
use of macroecological experiments (that is, large-scale field
experiments) on mountainsides. This approach integrates
elevational gradients with experimental manipulations of nutrient
enrichment to explore the independent effects of climate and
human impacts on biodiversity20,21. For instance, De Sassi et al.22
used a natural temperature gradient along elevations, combined
with experimental nitrogen fertilization, to investigate the effects
of elevated temperature and increased anthropogenic nitrogen
deposition on the structure and phenology of a grassland
herbivore assemblage. Such field experiments along natural
climatic gradients can be used to disentangle climatic effects
from any effects of local environmental conditions over relatively
large spatial scales.
Here we conducted comparative field experiments on
two mountainsidesin Norway and Chinato examine the
independent effects of temperature and nutrient enrichment on
aquatic bacterial richness and community composition (Fig. 1a).
Along nutrient and elevation (that is, temperature) gradients,
we established sterile aquatic microcosms composed of lake
sediments and artificial lake water, then let airborne bacteria
freely colonize the sediments and water of microcosms (Fig. 1a,b).
The microcosms were left in the field for 1 month before the
sediments were collected, and sediment bacteria were examined
using high-throughput sequencing of 16S rRNA genes. We chose
bacteria as model organisms for two reasons. First, bacteria are
small, abundant, diverse, essential to virtually all biogeochemical
cycles, and important components of ecosystems response to
global change23,24. Second, bacteria can passively disperse over
long distances and adapt quickly to changing environments due
to rapid generation times and dormant-resistant stages25.
Bacterial communities allow us to examine patterns of diversity
with a high degree of experimental control and replication in
2 NATURE COMMUNICATIONS | 7:13960 | DOI: 10.1038/ncomms13960 | www.nature.com/naturecommunications
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13960
natural field conditions that are subject to real species pool effects,
experiments that cannot be conducted under laboratory
conditions or with larger organisms within feasible time
periods26,27. Moreover, our recent field survey on the study
mountains indicated that nutrients were one of the main drivers
of aquatic bacterial diversity28.
We considered three components of bacterial biodiversity:
alpha, beta and gamma diversity29. Alpha diversity referred to the
local bacterial species richness in each microcosm. Beta diversity
referred to the community differentiation among microcosms.
Gamma diversity referred to the species richness of each elevation
(that is, temperature) or nutrient level. We quantified beta
diversity with the turnover rate of the distance-decay relationship
(DDR)30,31, considering the variations in community
composition from one microcosm to another along temperature
or nutrient gradients. We addressed the following five questions:
(1) How does the temperature effect on species richness vary
along a gradient in nutrient enrichment? (2) How does the
nutrient-richness relationship (NRR) vary with elevation, as
representative of different temperature zones? (3) How does the
slope of the temperature DDR, which is the community turnover
rate along the temperature gradient, vary with the gradient in
nutrients? (4) How does the species turnover rate along the
nutrient enrichment gradient (that is, nutrient DDR) vary with
temperature? (5) How do nutrient enrichment and temperature
jointly influence bacterial communities? Our results show clear
segregation of bacterial species along temperature gradients, and
decreasing alpha and gamma diversity toward higher nutrients.
The temperature dependence of species richness is weakest at the
intermediate nutrient levels, whereas the nutrient dependence of
species richness is strongest at intermediate temperatures. Thus,
our empirical evidence illustrates how temperature and nutrients
directly affect biodiversity, and also their indirect influence via
primary productivity.
Results
Primary productivity and pH. In our experiments, linear and
quadratic models were significantly (Po0.05, F-test) fitted for
most of the relationships of temperature-primary productivity,
as represented by Chlorophyll a (Chl a) (Supplementary Fig. 1),
which shows that primary productivity was highly correlated with
temperature. Nutrient enrichment increased primary productivity
more strongly at lower elevations and in the subtropical region
(Supplementary Fig. 2). This finding shows that nutrient effects
on primary productivity were weaker at the colder temperatures,
and indicates that in a warming climate, the ecosystem
productivity could be promoted more strongly than in current
climate. Higher temperatures also resulted in higher water pH,
especially at high nutrient concentrations (Supplementary Fig. 3).
Nutrient concentrations correlated positively with water pH,
particularly at low elevations (Supplementary Fig. 4). Chl a and
pH were positively correlated at almost all nutrient levels and
elevations (Supplementary Fig. 5).
Community composition. Bacterial communities were grouped
mainly by study region (r2 0.332, Po0.01) and elevation
(r2 0.251, Po0.01) based on a permutational multivariate
analysis of variance (PERMANOVA) (Fig. 2a). Communities
were also structured by local environments. In both regions,
community variations were primarily related to elevation,
temperature, pH, Chl a and nutrients according to multiple
statistical methods (that is, multiple regression analyses
(Supplementary Table 1), Mantel tests, Pearson correlations
(Supplementary Fig. 6) and canonical correspondence
analysis (Supplementary Fig. 7)).
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13960 ARTICLE

4,000

3,500

3,000
(1,000)

2,500
(500)

2,000
1,800 (0)
China Norway

b Replicates Replicates c

0.00 11.25
Subtropical
Total nitrogen (mg N l1)

Total nitrogen (mg N l1)

Subarctic
0.45 15.75
Biodiversity Subtropical Subarctic
1.80 21.60

4.05 28.80

7.65 36.00
Elevation Nutrients

Figure 1 | The manipulation of nutrient enrichment along elevational gradients. The experiments were conducted in parallel in the mountains of the
subtropical (that is, China, left panel) and subarctic (that is, Norway, right panel) regions (a). The figures of the two mountains were created according to
the plant species and climate zones along elevational gradients. Elevations (m a.s.l.) are shown without and with parenthesis for subtropical and subarctic
regions, respectively (a). Along each mountainside, sterile microcosms with ten nutrient levels and three replicates at each level (b, field photo) were set up
at each of five elevations, indicated by the brown dots (a), and were left in the field for 1 month. The nutrient levels were indicated by nitrogen because the
ratio between nitrogen and phosphorus was consistent (b). Airborne microbes freely colonized the sterile habitats. Nutrient addition promoted the growth
of algae, which caused gradual changes in green colour with higher nutrient enrichment (b). The bacterial biodiversity was expected to be higher in the
subtropics than in the subarctic region (c, upper panel), and showed predictable patterns along elevation (that is, temperature) and nutrient enrichment
(c, lower panels). The slopes of biodiversity along elevational gradients (c, left-lower panel) and nutrient enrichment (c, right-lower panel) were expected to
vary between regions, and with nutrient levels and elevations, respectively.

Interestingly, the bacterial communities at the higher elevations
in China were more similar to the communities in Norway than
those at lower elevations in China (lower panel of Fig. 2a,
Supplementary Fig. 8), suggesting that they had more species in
common. The elevational patterns of the community Srensen
similarity between each elevation in one region (that is, China)
and all elevations in the other region (that is, Norway) show that
the similarity significantly (Po0.05) increased and decreased
toward higher elevations for China and for Norway, respectively
(upper panels of Fig. 2a). These results indicate that the bacterial
communities at higher elevations in China were more similar to
those in Norway, and the communities at lower elevations in
Norway shared more species to those in China. This segregation
of species along elevations or climatic zones is, to our knowledge,
the first reported for microbes, and agrees well with the classic
observations of higher organisms. For instance, Linnaeus32,33
noted in his dissertation that y on the tops and sides of such a
mountain the same vegetables might grow, the same animals live,
as in Lapland and the frigid zone; and in effect we find in
the Pyrenean, Swiss, and Scotch mountains, upon Olympus,
Lebanon, and Ida, the same plants which cover the Alps of
Greenland and Lapland. Given the long distance between the
two mountains studied, our results suggest not only the high
dispersal ability of bacteria, but also that ambient environments
filter species at a local scale.
Alpha and gamma diversity. The alpha and gamma diversities,
that is, the species richness (that is, OTU number) of each sample
(n 300) and experimental site (n 10), respectively, were 1.97
times higher in China than in Norway (t-test, Po0.001, Fig. 2b).
For the Norwegian sites, both alpha and gamma diversities
decreased at high elevation, whereas hump-shaped patterns were
found for the Chinese sites (Fig. 2b, left panels). The different
patterns imply that the effects of temperature on diversity may
differ between subarctic and subtropical regions. In both regions,
nutrient enrichment had consistent effects on alpha and gamma
diversity, both of which decreased with increasing nutrients
(Fig. 2b, right panels). This finding indicates that nutrient
enrichment impoverishes microbial biodiversity, which agrees

NATURE COMMUNICATIONS | 7:13960 | DOI: 10.1038/ncomms13960 | www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13960

a Elevation in Norway (m) Elevation in China (m) b Elevation in China (m)

200 400 600 2,500 3,000 3,500 2,200 3,000 3,800 0 0.5 1 1.5
Srensen

China
0.50 6,000 China 6,000

Gamma
Norway
0.40 2
R = 0.214, P < 0.05
2
R = 0.549, P < 0.01 Norway
3,000 3,000
Elevation
0.2 20
170 1,500 1,500
350
nMDS 2

550
0.0

Alpha
750 1,000 1,000
2,286
2,580
2,915
0.2 500 500
3,505
3,822
Norway China
0 200 600 0 0.5 1 1.5
0.25 0.00 0.25 0.50
Elevation in Norway (m) Nutrients (log10)
nMDS 1

Figure 2 | Responses of community composition and diversity to elevation and nutrients. (a) Non-metric multidimensional scaling (nMDS) plot of
bacterial communities (lower panel), grouped by elevation (m a.s.l., indicated by colour, with higher elevations in warmer colours) and country (indicated by
dotted grey line). This plot illustrates that the communities at lower elevations in Norway (or higher elevations in China) were more similar to communities in
China (or Norway) than the communities at higher elevations in Norway (or lower elevations in China), which is quantitatively supported by the upper figure
panels (left: Norway; right: China) that have triangle points and linear regression lines. We calculated the community Srensen similarity along the elevational
gradient between each elevation of one region (that is, China) and all elevations of the other region (that is, Norway). The relationship between the similarity
and elevation was fit and tested with a linear model and permutation tests in the R package lmPerm (v.1.1-2). (b) Gamma diversity (upper panels) and
alpha diversity (lower panels) along elevations (left panels) and nutrient enrichment levels (right panels). For diversity-elevation and diversity-nutrient
relationships, we applied quadratic and linear models, respectively, and significances of the relationships were examined with F-statistics. For gamma
diversity-elevation relationships in Norway and China, the adjusted R2 values were 0.952 (P 0.024) and 0.957 (P 0.022), respectively. For alpha diversity-
elevation relationships in Norway and China, the adjusted R2 values were 0.518 (Po0.001) and 0.335 (Po0.001), respectively. For gamma diversity-nutrient
relationships in Norway and China, the adjusted R2 values were 0.546 (P 0.009) and 0.332 (P 0.047), respectively. For alpha diversity-nutrient
relationships in Norway and China, the adjusted R2 values were 0.047 (P 0.005) and 0.049 (P 0.004), respectively. The elevations (m a.s.l.) in
Norway (blue) and China (red) are shown along the bottom and top axes (b, left panels), respectively. The amount of NO3# (mg N l # 1) initially added to the
microcosms represents the nutrient enrichment (b, right panels). The points were jittered for better visualization (b, lower panels).

with a recent meta-analysis on richness-phosphorus relationships
of macroorganisms34, but is in contrast to the marginal response
of soil microbial diversity to nutrient enrichment at a global
scale35. Similar to community composition, alpha diversity was
correlated positively with temperature, Chl a, and pH in both
regions (Supplementary Fig. 6, Supplementary Table 1), and these
are typical drivers of microbial species richness or community
composition in lakes36,37 and the ocean2.
Effects of climate and nutrients on biodiversity. The results
above showed that temperature, which was correlated strongly
and negatively with elevation, was an important driver for both
richness and community composition (Supplementary Figs 6, 7
and Supplementary Table 1). Thus, we explored how the shape of
the biodiversity-temperature relationship was modified by
nutrient enrichment and how the effects of nutrients depended
on temperature.
We first investigated whether the effect of temperature on
species richness varied along a nutrient gradient. For the 20
temperature-richness relationships (TRRs), significant (Po0.05)
linear and quadratic models were fitted in 15 and 7 cases,
respectively (Supplementary Fig. 9). This finding supports the fact
that richness is strongly temperature dependent, and suggests that
the elevational diversity gradients in microbes can be explained
by environmental filtering or by MTE1. MTE provides a
framework to assess how temperature affects organismal
metabolisms and influences their ecology and evolution, such as
rates of evolution, community composition, gradients of diversity
and ecosystem processes1. Accordingly, log-transformed bacterial
species richness is a linear function of the inverse absolute
temperature (log10(S)pE " (1/kT), where S is species richness,
k is Boltzmans constant 8.62 " 10 # 5 eV K # 1, T is absolute
temperature in Kelvin and E is the slope or activation energy
in eV characterizing the temperature dependence of species
richness1. The slopes of the 15 significant linear TRRs, which
represent the activation energy, E (Fig. 3a) and indicate the
magnitude that species richness depends on temperature, varied
between # 0.88 and # 0.18, with a mean value of # 0.370.20.
These values are similar to microbes in forest soils38, but are
lower than the theoretical predictions of between # 0.70 and
# 0.60 (ref. 1). The lower E values of bacteria compared with
macroorganisms1 suggests that bacteria are less dependent on
temperature changes, perhaps due to their high dispersal ability,
rapid generation times and dormant-resistant stages25. The
E values were significantly (t-test, Po0.05) more negative in
Norway than in China (Fig. 3a). This finding indicates that
bacteria in the subarctic region are more sensitive to temperature
than those in the subtropics and may experience larger
temperature-related shifts in richness under future climate
scenarios.
In both regions, the temperature dependence of species
richness was mediated by nutrient enrichment, shown by the
fact that E values were closest to zero at intermediate nutrient

4 NATURE COMMUNICATIONS | 7:13960 | DOI: 10.1038/ncomms13960 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13960 ARTICLE

0.1
0.0 a b
0.0
0.1
0.3

TRR slope (E)

0.2

NRR slope
2,500 3,000 3,500
0.6
0.1 c
0.0
0.9 0.1
China: R 2 = 0.78 P = 0.005
2
Norway: R = 0.58 P = 0.048 0.2

0.0 0.5 1.0 1.5 0 200 400 600

Nutrients (log10) Elevation (m)

0.010 d 0.05 e
0.10
Temperature DDR slope

0.15

Nutrient DDR slope

0.015 0.20
0.25
2,286 2,580 2,915 3,505 3,822
0.020
0.05 f
0.10
0.025 0.15
2
China: R = 0.8 P = 0.004
0.20
Norway: R 2 = 0.64 P = 0.048
0.25
0.0 0.5 1.0 1.5 20 170 350 550 750
Nutrients (log10) Elevation (m)

Figure 3 | The variation of the temperature or nutrient dependence of biodiversity. Species richness plots (ac): the slopes of the TRR and NRR along
nutrient enrichment (a) and elevation gradients (b,c), respectively. The TRR slope (a), characterizing the temperature dependence of species richness, was
calculated according to MTE1, and log-transformed bacterial species richness is a linear function of the inverse absolute temperature (log10(S)pE " (1/kT),
where S is species richness, k is Boltzmans constant 8.62 " 10 # 5 eV K # 1, T is absolute temperature in Kelvin and E is the slope or activation energy (e) in
eV. Community similarity plots (df): The slopes of the temperature DDR and nutrient DDR along nutrient enrichment (d) and elevation gradients (e,f),
respectively. DDR was based on Srensen similarity. The nutrient DDR slopes were multiplied by 100 for better visualization. Solid dots indicate the
significant (Po0.05) relationships. Initially added NO3# (mg N l # 1) represents the nutrient enrichment. Blue and red dots represent Norway and China,
respectively.

levels (that is, B4.057.65 mg N l # 1 total nitrogen (TN), Fig. 3a).
The fact that species richness is dependent on temperature has
been shown to be influenced by various factors, such as spatial
scale for plants39. However, the mediation of nutrient enrichment
on the magnitude of temperature dependence is rarely
considered. Our findings clearly indicate that richness decreased
faster with decreasing temperature at low or high nutrient levels
than it did at intermediate nutrient levels, suggesting that the
responses of bacteria to temperature changes are strongest at very
low or high levels of nutrients. Therefore, at intermediate levels of
nutrient enrichment, the communities or ecosystems may be
most resistant to climate influence. For instance, in the eutrophic
Taihu Lake in China, increased temperatures result in earlier,
longer-lasting cyanobacterial blooms40, which further decreases
aquatic biodiversity41. Nutrient concentrations (that is, TN) in
Taihu Lake, during the years 19972015, were 28.6% and 7.1%
higher, respectively, than the 4.05 and 7.65 mg N l # 1 TN
intermediate nutrient enrichments used in our experiment,
though with a high spatial heterogeneity (Supplementary
Fig. 10). A recent study in Taihu Lake showed that nutrient
reductions from intermediate levels magnified the impact of
extreme weather on bloom-plagued conditions42, which supports
our finding that low nutrient levels would increase vulnerability
of diversity to climate change. However, additional studies are
needed to confirm whether nutrient enrichment generally affects
the temperature dependence of species richness in other
ecosystems (that is, terrestrial environments) and results in
altered ecosystem resistance at intermediate nutrient levels,
because microbial communities are also structured by their
original habitat types43.
Second, we examined how the NRR vary among elevations
representing different temperature zones. Because bacterial
alpha and gamma diversity decreased at higher nutrient levels
in both regions (Fig. 2b), we used the slope of the linear
regression of NRR to represent the changes in species richness
with nutrient enrichment. We found that species richness
decreased significantly (Po0.05) with nutrient enrichment only
at intermediate elevations (Fig. 3b,c), suggesting that species
richness at intermediate elevations was more sensitive to nutrient
enrichment.
Third, we used the slopes of DDR to quantify species turnover
rates (that is, beta diversity) along temperature gradients and then
tested how these turnover rates varied with the nutrient gradients.
More negative DDR slopes indicate higher species turnover rates.
In both regions, the temperature DDRs were significant (Mantel
test, Po0.01) at all nutrient levels (Fig. 3d). In China, the species
turnover rates exhibited a shallow U-shaped pattern along
nutrient enrichment gradients (Fig. 3d). This pattern suggests
that nutrient enrichment first slightly increases the species
turnover rate until reaching intermediate nutrient levels, and
the species composition becomes more spatially homogeneous at
high nutrient levels. In Norway, however, the turnover rates
exhibited a unimodal pattern; they responded sharply to low
nutrient concentrations (that is, 0.45 mg N l # 1) with lower
turnover rates at intermediate nutrient levels (Fig. 3d). The
differing response of turnover rates along temperature gradient to

NATURE COMMUNICATIONS | 7:13960 | DOI: 10.1038/ncomms13960 | www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13960

a b
Tem Tem

0.32 0.169

Nut 0.589 Div Nut 0.583 Div

0.29 0.35
0.12 0.327
0.432 0.311

Pro Pro
c d
Tem Tem

0.163 0.748

Nut 0.589 Com Nut 0.583 Com

0.259 0.028

0.12 0.329
0.597 0.269

Pro Pro

Figure 4 | The direct and indirect effects of temperature and nutrients on biodiversity. The effects of temperature (Tem), nutrient enrichment (Nut) and
primary productivity (Pro) on bacterial diversity (Div) and community composition (Com) for Norway (a,c) and China (b,d), explored with partial least
squares path model. For diversity (a,b) and community composition (c,d), species richness and the first axis of nMDS were used as observed variables. For
temperature, water temperature and its squared value were used. For nutrient enrichment, the observed variables included the initial levels of added NO3#
and measured NO3# , NO2# and PO34 . For primary productivity, the observed variables were pH and Chl a. Shown are the path coefficients calculated after
1,000 bootstraps. Models were assessed using goodness of fit (GoF) statistic. The GoFs for A-D are 0.630, 0.520, 0.622 and 0.718, respectively.

nutrient enrichment for the two regions highlights the potentially
different community assembly mechanisms constrained by
nutrient and temperature gradients. Dissimilar mechanisms of
community assembly (for example, species-sorting and dispersal
limitation) have also been observed for temperate and tropical
forests44, and may also contribute to the strikingly different
biodiversity gradients of these two biogeographic regions
(Fig. 2b). The two patterns in turnover rate are also
inconsistent with the findings of the communities in other
habitats, such as the generally increasing turnover rate with
increasing primary productivity observed for freshwater
plankton45. The explanations for the inconsistency may be the
potential differences in productivity gradients among studies
because the nutrient gradient we considered here was extremely
long. Another reason could be the different organisms studied
(that is, bacteria, phyto- and zooplankton, representing
contrasting trophic groups).
Fourth, we examined how species turnover rates on the
nutrient gradient, quantified with nutrient DDR slopes, varied
with the temperature gradient. In both regions, the nutrient
DDRs of each elevation were typically significant (Mantel test,
Po0.05) (Fig. 3e,f). The significant DDR slopes decreased
significantly (Po0.05) toward high elevations (that is, decreasing
temperature) in China, but did not decrease significantly
(P 0.159) in Norway. This pattern indicates that the species
turnover rate resulting from nutrient enrichment did not increase
at higher temperatures. Our results therefore differ from the
findings for other organisms, such as vascular plants46, which
show that species turnover rates decrease toward high latitudes.
This difference may have occurred because we considered
temperature and nutrient enrichment as the sole primary
drivers for bacterial communities, which is unlikely under
natural conditions shaped by a higher number of covariant
environmental drivers.
Finally, to synthesize all the findings, we conducted partial least
squares path modelling (PLS-PM)47 to illustrate the direct and
indirect effects of temperature and nutrient enrichment on
richness and community composition. For richness, nutrient
enrichment had negative direct effects, while temperature had
positive effects (Fig. 4a,b). Temperature was the dominant factor
affecting primary productivity in both regions, while nutrients
and temperature indirectly affected richness through primary
productivity (Fig. 4a,b). Such consistency in the underlying
drivers of richness between the two regions agrees with the
parallel patterns observed of the effects of temperature (Fig. 3a)
and nutrient enrichment (Fig. 3b,c) on richness in Norway
and China. These results suggest that both temperature-related
kinetic mechanisms1 and productivity-diversity hypothesis2,48
may explain the variation in species richness, while the latter
appears to be the stronger factor. For community composition,
nutrients and temperature exerted indirect effects through
primary productivity, and primary productivity was the
dominant driver in the subarctic region (Fig. 4c,d). However,
in the subtropics, the direct effects of temperature were dominant
and nutrient effects were weakest (Fig. 4c,d). These contrasting
mechanisms are in agreement with the differences in the patterns
of temperature DDR slopes along nutrient gradients between the
two regions (Fig. 3d).
Discussion
To further elucidate the interactive effects of temperature
and nutrient enrichment on biodiversity, future studies
are encouraged to consider different taxonomic groups, various
habitats, and even more advanced experimental designs.

6 NATURE COMMUNICATIONS | 7:13960 | DOI: 10.1038/ncomms13960 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13960
For instance, relevant comparison of communities of micro- and
macroorganisms49, or multiple habitats43, such as the overlying
water and sediments in our microcosms, will go a long way
toward supporting broader conclusions regarding the effects of
temperature and nutrients on biota. Although microbial
experiments are not appropriate for all ecological questions,
microbial manipulation experiments, for example, Vannette
and Fukami50, offer a complementary approach to field
and laboratory studies of macroorganisms27. Furthermore,
multiple analytical approaches of biological analysis, such as
metagenomics or Geochip51, would be helpful for understanding
the effects of temperature and nutrients on the functional
diversity and various activities of communities35,52 and,
consequently, their impacts on the ecosystem functioning and
services, which often depend on biodiversity11,53. The duration of
our experiments was one month, which is similar to that of
previous microbial manipulation field studies investigating
the underlying processes of community assembly26. Future
experiments with a high-resolution time series and longer
duration (for example, the 30-year and 150-year fertilization
experiments on plant54 and microbial biodiversity55, respectively)
and more global distribution (for example, Nutrient Network56)
would provide more evidence for the dynamic patterns of the
effects of global change on global-scale biodiversity.
Collectively, we answered five specific questions regarding the
effects of temperature and nutrient enrichment on bacterial
biodiversity. By conducting experiments along climatic gradients,
we have presented the first empirical evidence of the patterns and
pathways of the effects of temperature and nutrient enrichment
on biodiversity in subtropical and subarctic regions. For over two
centuries, ecologists have documented the relationships between
biodiversity and temperature13,32, productivity2,12,45,48,57 or
anthropogenic impacts19,35,56,58,59. The independent and inter-
active effects of these factors are central to understanding the
underlying mechanisms responsible for the generation and
maintenance of biodiversity, and furthermore, to forecasting the
effects of global changes on biodiversity19. We believe our
findings have important implications regarding these pivotal
effects on biodiversity.
First, our results highlight the fact that macroecological
experiments along environmental gradients (for example,
mountain elevation gradients) are an important tool in ecological
research because they allow for the disentangling the effects of
individual environmental drivers on biodiversity, the independent
effects of which are not be easily separated due to their covariance
in nature. The current findings using microbes as model
organisms offer strong examples of the importance of the study
of global changes using integrating experiments and natural
environmental gradients, and illustrate an emerging approach
which can be distributed globally to advance our predictive
understanding of ecological trends and responses. Second,
temperature and nutrients play pivotal roles in maintaining
elevational biodiversity patterns such that the temperature
dependence of species richness is strongest at very low and high
nutrient enrichment, while the effect of nutrients on species
richness is strongest at intermediate temperatures. We found
clear segregation of bacterial species along temperature gradients
(or climatic zones), and decreasing alpha and gamma diversity
toward higher nutrient levels. We documented the direct effects
of temperature and nutrient enrichment on biodiversity, and also
showed that both factors indirectly affected communities through
primary productivity. Thus, we fill the knowledge gaps in how
well we understand the direct and indirect effects of climate
change and human impacts on the spatial patterns of biodiversity,
and provide thoughtful insights into how nutrient enrichment
may alter biodiversity under future climate scenarios.
NATURE COMMUNICATIONS | 7:13960 | DOI: 10.1038/ncomms13960 | www.nature.com/naturecommunications
ARTICLE
Methods
Experimental design. The parallel field experiments were conducted in a subarctic
region, Balggesvarri Mountain in Norway (01,270 m a.s.l.), and in a subtropical
region, Laojun Mountain in China (2,2803,820 m a.s.l.)28, in July and
SeptemberOctober 2013, respectively (Fig. 1a). The climate in the Balggesvarri
Mountain region is subarctic, with a growing season of B3 months. The annual
temperatures ranged from # 2.90.7 !C, with July temperatures ranging from 8 to
16 !C. The tree line is located at B550 m a.s.l. The climate in the Laojun Mountain
region is subtropical. The annual temperatures ranged from 4.212.9 !C, with July
temperatures varying from 1725 !C. The tree line is located at B4,200 m a.s.l.
Along the side of each mountain, we selected unshaded locations at five
different elevations. At each elevation, we set up 30 1.5 l bottles, which included ten
nutrient levels and three replicates of each level (Fig. 1b). The elevations were
3,822, 3,505, 2,915, 2,580 and 2,286 m a.s.l. for China, and 750, 550, 350, 170 and
20 m a.s.l. for Norway (Fig. 1b). The bottles of different nutrient levels and
replicates were arranged non-randomly at each elevation (Fig. 1b). The bottom of
each bottle (B10% of the total bottle height) was buried in the local soil. We filled
each bottle with 1.2 l sterilized freshwater and 15 g sterilized sediments. The
sterilized sediments were prepared before the field experiments and were collected
from the centre of Taihu Lake in October 2012, freeze dried, and stored at # 20 !C.
The sediments were autoclaved eight times at 121 !C for 30 min, dried at 110 !C for
24 h, homogenized, and then aseptically canned with 15 g sediments per bottle for
the field experiments. The dried sediments were verified to be sterile by negative
DNA amplification using bacterial primers after DNA extraction following the
steps in the section Bacterial community analyses. No amplification results were
observed. The artificial freshwater was prepared with sterilized MilliQ water and
autoclaved at 121oC for 30 min, and the following salts were added: CaCl2
7.55 g l # 1, MgSO4 % 7H2O 6.78 g l # 1 and NHCO3 3.53 g l # 1. To facilitate the initial
colonization of heterogenetic microbes, 0.91 g l # 1 glucose was added. KNO3 was
added at rates of 0.00, 0.45, 1.80, 4.05, 7.65, 11.25, 15.75, 21.60, 28.80 and
36.00 mg N l # 1 to generate ten nutrient levels including the control of
0.00 mg N l # 1. To compensate for the nitrate additions, KH2PO4 was added so that
the N/P ratio of the overlying water was 14.93, which was similar to the annual
average ratio in Taihu Lake during 2007 (14.49). The nutrient concentrations for
the experiments were selected according to the nutrient levels of the eutrophic Lake
Taihu in China, and the highest nitrate concentration was based on the maximum
TN of Taihu in 2007 (20.79 mg N l # 1).
The bottles were left in the field for 28 and 31 days, respectively, to allow
airborne organisms (for example, bacteria) to colonize the water and sediments of
microcosms. The field setups were completed in 3 days. To keep the species
dispersal events as natural as possible, we did not cover the experimental set-ups in
case of rainfall. We checked the experimental set-ups twice during each
experimental period, and added sterilized MilliQ water to obtain a final volume of
approximately 1.2 l. Filling to a volume of 1.2 l with artificial freshwater into the
1.5 l bottles ensured that the water would not overflow due to rain or splash out in
the heavy rains during the experimental periods.
To avoid the effects of daily temperature variation, we measured the water
temperature and pH within 2 h before noon at all elevations in the day before the
final sample collection. At the end of the experimental period, we aseptically
sampled the water and sediments of each bottle. The samples were frozen at
# 20 !C after sampling until chemical and molecular analyses.
It should be noted that we analysed the sediment bacteria, but not the water
column bacteria. We did not use any specific natural aquatic bacterial communities
from ponds or lakes in the current experiments, but established new communities
via post-dispersal effects. More details on the experimental design are provided in
the Supplementary Methods.
Physicochemical and biological analyses. Water ammonium (NH4 ), nitrate
(NO3# ), nitrite (NO2# ) and dissolved inorganic phosphorus (PO34 # ) were mea-
sured with a flow injection analyser (Skalar SA1000, Breda, Netherlands). Sediment
Chl a was extracted with 90% acetone49. Sediment genomic DNA was extracted
using the phenol chloroform method, and bacterial 16S rRNA genes were amplified
in triplicate using universal bacterial primers28. Real-time qPCR quantification of
bacterial 16S rRNA genes was performed on an iCycler iQ5 thermocycler (Bio-Rad,
Hercules, CA) as described previously60. PCR products were sequenced with MiSeq
(Illumina, San Diego, CA). The sequences were processed in QIIME (v1.8)28,61.
OTUs were defined at 97% sequence similarity. The bacterial sequences were
rarefied to 18,000 per sample. More details are provided in the Supplementary
Methods.
Statistical analyses. Non-metric multidimensional scaling (nMDS) was based on
the community Srensen similarity, which is a popular beta diversity metric used in
ecological studies for DDRs31,62 and was applied in a general framework for the
distance-decay of similarity in ecological communities63. To test the hypothesis
that region and elevation structure the bacterial communities, PERMANOVA was
used64. To identify important environmental factors related to communities,
we performed Mantel tests (permutations 9,999) on the community Srensen
similarity, Pearson correlations using the first axis of nMDS, and a canonical
correspondence analysis on species abundance data.
7
ARTICLE
We used linear and quadratic models to explore the relationships between alpha
and gamma diversity with elevation and nutrient. The more appropriate model was
selected based on a lower value of Akaikes information criterion65, and F-statistic
was used to test the significance of regression. We used Pearson correlations to
explore the relationships between species richness and environmental variables.
We also applied stepwise multiple regression analyses with forward selection of
variables to identify the most important environmental factors explaining
community composition (that is, sample scores on the first axis of nMDS) and
species richness.
Water temperature was highly correlated with elevation, and was among the
strongest factors related to species richness and community composition
(Supplementary Fig. 6); therefore, we used water temperature to explore the
relationships between temperature and species richness or composition. For each
region, we fit linear and quadratic models for the TRR of each nutrient level and
found the quadratic model to be better in 7 out of 20 cases. However, significant
linear models also fit well in 15 out of 20 cases (Supplementary Fig. 9). Thus, we
examined the TRR with the MTE, where the lognormal richness is a linear function
of temperature, expressed as 1/kT, in which k is Boltzmanns constant and T is
absolute temperature in K. The slope of TRR was defined as the activation energy
(E), indicating the temperature dependence of species richness. Furthermore, the
slopes of the temperature DDR, based on Srensen similarity, were used to explore
the turnover rates of species composition across temperature gradients. Finally, the
slopes of TRR and temperature DDR were related to nutrient enrichment and the
relationships were explored with linear and quadratic models. The better model
was selected based on lower value of Akaikes information criterion. For NRR, the
slope of the linear regression was used to represent changes in species richness with
nutrient enrichment (log10). The slopes of nutrient DDR were used to investigate
the turnover rates of species composition across nutrient gradients. The
significance of DDR slopes was tested with Mantel test (permutations 9,999).
We explored the relationships between temperature, nutrient enrichment, and
bacterial communities using PLS-PM in the R package plspm (V0.4.7)47. This
method is known as the partial least squares approach to structural equation
modelling and allows for the estimation of complex cause-effect relationship
models with latent variables47, which was especially suitable for our experimental
data with strong environmental gradients. Five latent variables were used:
temperature (the measured water temperature and its squared value), nutrient
enrichment (the initially added nutrients and measured nutrients), primary
productivity (Chl a and pH), diversity (species richness), and composition (the first
axis of nMDS). We used pH as a proxy for primary productivity because of its
positive correlations with Chl a (Supplementary Fig. 5). Observed variables were
selected based on collinearity and prediction power for diversity and composition.
Most of the loadings for observed variables on latent variables were40.7
(Supplementary Fig. 11). We ran PLS-PM using 1,000 bootstraps to validate the
estimates of path coefficients and the coefficients of determination47. Path
coefficients represent the direction and strength of the linear relationships between
variables, or the direct effects. Indirect effects are the multiplied path coefficients
between a predictor and a response variable, adding the product of all possible
paths excluding the direct effect. Models with different structures were evaluated
using the goodness of fit statistic47.
Data availability. The amplicon sequences were deposited in MG-RAST under
accession number 17710. Other relevant data in this study are available from the
authors.
References
1. Brown, J. H., Gillooly, J. F., Allen, A. P., Savage, V. M. & West, G. B. Toward a
metabolic theory of ecology. Ecology 85, 17711789 (2004).
2. Fuhrman, J. A. et al. A latitudinal diversity gradient in planktonic marine
bacteria. Proc. Natl Acad. Sci. USA 105, 77747778 (2008).
3. Stegen, J. C., Ferriere, R. & Enquist, B. J. Evolving ecological networks and the
emergence of biodiversity patterns across temperature gradients. Proc. R. Soc. B
Biol. Sci. 279, 10511060 (2012).
4. Stocker, T. et al. Climate Change 2013: The Physical Science Basis
(Cambridge University Press, 2014).
5. Colwell, R. K., Brehm, G., Cardelus, C. L., Gilman, A. C. & Longino, J. T. Global
warming, elevational range shifts, and lowland biotic attrition in the wet
tropics. Science 322, 258261 (2008).
6. Thomas, C. D. et al. Extinction risk from climate change. Nature 427, 145148
(2004).
7. Pacifici, M. et al. Assessing species vulnerability to climate change.
Nat. Clim.Change 5, 215224 (2015).
8. Bellard, C., Bertelsmeier, C., Leadley, P., Thuiller, W. & Courchamp, F. Impacts
of climate change on the future of biodiversity. Ecol. Lett. 15, 365377 (2012).
9. Newbold, T. et al. Global effects of land use on local terrestrial biodiversity.
Nature 520, 4550 (2015).
10. Price, M. F. Global Change in Mountain Regions (Sapiens Publishing, 2006).
11. Isbell, F. et al. Nutrient enrichment, biodiversity loss, and consequent declines
in ecosystem productivity. Proc. Natl Acad. Sci. USA 110, 1191111916 (2013).
8 NATURE COMMUNICATIONS | 7:13960 | DOI: 10.1038/ncomms13960 | www.nature.com/naturecommunications
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13960
12. Adler, P. B. et al. Productivity is a poor predictor of plant species richness.
Science 333, 17501753 (2011).
13. Sax, D. F. & Gaines, S. D. Species diversity: from global decreases to local
increases. Trends. Ecol. Evol. 18, 561566 (2003).
14. Noges, T. Relationships between morphometry, geographic location and water
quality parameters of European lakes. Hydrobiologia 633, 3343 (2009).
15. Jeppesen, E. et al. Climate change effects on runoff, catchment phosphorus
loading and lake ecological state, and potential adaptations. J. Environ. Qual.
38, 19301941 (2009).
16. Nogues-Bravo, D., Araujo, M. B., Romdal, T. & Rahbek, C. Scale effects and
human impact on the elevational species richness gradients. Nature 453,
216219 (2008).
17. Korner, C. The use of altitude in ecological research. Trends Ecol. Evol. 22,
569574 (2007).
18. Porter, E. M. et al. Interactive effects of anthropogenic nitrogen enrichment and
climate change on terrestrial and aquatic biodiversity. Biogeochemistry 114,
93120 (2012).
19. Mantyka-pringle, C. S., Martin, T. G. & Rhodes, J. R. Interactions between
climate and habitat loss effects on biodiversity: a systematic review and
meta-analysis. Global Change Biol. 18, 12391252 (2012).
20. Sundqvist, M. K., Sanders, N. J. & Wardle, D. A. Community and ecosystem
responses to elevational gradients: processes, mechanisms, and insights for
global change. Annu. Rev. Ecol. Evol. Syst. 44, 261280 (2013).
21. Dunne, J. A., Saleska, S. R., Fischer, M. L. & Harte, J. Integrating experimental
and gradient methods in ecological climate change research. Ecology 85,
904916 (2004).
22. de Sassi, C., Lewis, O. T. & Tylianakis, J. M. Plant-mediated and nonadditive
effects of two global change drivers on an insect herbivore community. Ecology
93, 18921901 (2012).
23. Prosser, J. I. et al. The role of ecological theory in microbial ecology. Nat. Rev.
Microbiol. 5, 384392 (2007).
24. Xue, K. et al. Tundra soil carbon is vulnerable to rapid microbial decomposition
under climate warming. Nat. Clim. Change 6, 595600 (2016).
25. Lennon, J. T. & Jones, S. E. Microbial seed banks: the ecological and
evolutionary implications of dormancy. Nat. Rev. Microbiol. 9, 119130 (2011).
26. Bell, T. Experimental tests of the bacterial distance-decay relationship. ISME J.
4, 13571365 (2010).
27. Jessup, C. M. et al. Big questions, small worlds: microbial model systems in
ecology. Trends Ecol. Evol. 19, 189197 (2004).
28. Wang, J. et al. Regional and global elevational patterns of microbial species
richness and evenness. Ecography. doi: 10.1111/ecog.02216 (2017).
29. Whittaker, R. H. Vegetation of the Siskiyou Mountains, Oregon and California.
Ecol. Monogr. 30, 279338 (1960).
30. Harte, J., McCarthy, S., Taylor, K., Kinzig, A. & Fischer, M. L. Estimating
species-area relationships from plot to landscape scale using species spatial-
turnover data. Oikos 86, 4554 (1999).
31. Soininen, J., McDonald, R. & Hillebrand, H. The distance decay of similarity in
ecological communities. Ecography 30, 312 (2007).
32. Linnaeus, C. On the increase of the habitable earth. Amoenitates academicae 2,
1727 (1781).
33. Lomolino, M. V., Sax, D. F. & Brown, J. H. Foundations of Biogeography: Classic
Papers with Commentaries (University of Chicago Press, 2004).
34. Azevedo, L. B. et al. Species richnessphosphorus relationships for lakes and
streams worldwide. Glob. Ecol. Biogeogr. 22, 13041314 (2013).
35. Leff, J. W. et al. Consistent responses of soil microbial communities to elevated
nutrient inputs in grasslands across the globe. Proc. Natl Acad. Sci. USA 112,
1096710972 (2015).
36. Lindstrom, E. S., Kamst-Van Agterveld, M. P. & Zwart, G. Distribution of
typical freshwater bacterial groups is associated with pH, temperature, and lake
water retention time. Appl. Environ. Microbiol. 71, 82018206 (2005).
37. Logue, J. B. et al. Freshwater bacterioplankton richness in oligotrophic lakes
depends on nutrient availability rather than on species-area relationships. ISME
J. 6, 11271136 (2012).
38. Zhou, J. et al. Temperature mediates continental-scale diversity of microbes in
forest soils. Nat. Commun. 7, 12083 (2016).
39. Wang, Z., Brown, J. H., Tang, Z. & Fang, J. Temperature dependence, spatial
scale, and tree species diversity in eastern Asia and North America. Proc. Natl
Acad. Sci. USA 106, 1338813392 (2009).
40. Zhang, M., Duan, H., Shi, X., Yu, Y. & Kong, F. Contributions of meteorology
to the phenology of cyanobacterial blooms: implications for future climate
change. Water. Res. 46, 442452 (2012).
41. Qin, B. et al. Lake eutrophication and its ecosystem response. Chin. Sci. Bull.
58, 961970 (2013).
42. Yang, Z. et al. Nutrient reduction magnifies the impact of extreme weather on
cyanobacterial bloom formation in large shallow Lake Taihu (China). Water.
Res. 103, 302310 (2016).
43. Wang, J. et al. Phylogenetic beta diversity in bacterial assemblages across
ecosystems: deterministic versus stochastic processes. ISME J. 7, 13101321 (2013).
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13960
44. Myers, J. A. et al. Beta-diversity in temperate and tropical forests reflects
dissimilar mechanisms of community assembly. Ecol. Lett. 16, 151157 (2013).
45. Chase, J. M. & Leibold, M. A. Spatial scale dictates the productivity-biodiversity
relationship. Nature 416, 427430 (2002).
46. Qian, H. & Ricklefs, R. E. A latitudinal gradient in large-scale beta diversity for
vascular plants in North America. Ecol. Lett. 10, 737744 (2007).
47. Sanchez, G. PLS Path Modeling with R (Trowchez Editions, 2013).
48. Vallina, S. M. et al. Global relationship between phytoplankton diversity and
productivity in the ocean. Nat. Commun. 5, 4299 (2014).
49. Wang, J. et al. Contrasting patterns in elevational diversity between
microorganisms and macroorganisms. J. Biogeogr. 38, 595603 (2011).
50. Vannette, R. L. & Fukami, T. Historical contingency in species interactions:
towards niche-based predictions. Ecol. Lett. 17, 115124 (2014).
51. Zhou, J. et al. High-throughput metagenomic technologies for complex
microbial community analysis: open and closed formats. MBio. 6,
e0228802214 (2015).
52. Martiny, J. B. H., Jones, S. E., Lennon, J. T. & Martiny, A. C. Microbiomes in
light of traits: a phylogenetic perspective. Science 350, aac9323 (2015).
53. Tilman, D., Reich, P. B. & Isbell, F. Biodiversity impacts ecosystem productivity
as much as resources, disturbance, or herbivory. Proc. Natl Acad. Sci. USA 109,
1039410397 (2012).
54. Isbell, F., Tilman, D., Polasky, S., Binder, S. & Hawthorne, P. Low biodiversity
state persists two decades after cessation of nutrient enrichment. Ecol. Lett. 16,
454460 (2013).
55. Liang, Y. et al. Over 150 years of long-term fertilization alters spatial scaling of
microbial biodiversity. Mbio. 6, e0024000215 (2015).
56. Borer, E. T. et al. Finding generality in ecology: a model for globally distributed
experiments. Method. Ecol. Evol. 5, 6573 (2014).
57. Grace, J. B. et al. Integrative modelling reveals mechanisms linking productivity
and plant species richness. Nature 529, 390393 (2016).
58. Hautier, Y. et al. Eutrophication weakens stabilizing effects of diversity in
natural grasslands. Nature 508, 521525 (2014).
59. Piggott, J. J., Salis, R. K., Lear, G., Townsend, C. R. & Matthaei, C. D. Climate
warming and agricultural stressors interact to determine stream periphyton
community composition. Global Change Biol. 21, 206222 (2015).
60. Shen, J., Zhang, L., Guo, J., Ray, J. L. & He, J. Impact of long-term fertilization
practices on the abundance and composition of soil bacterial communities in
Northeast China. Appl. Soil. Ecol. 46, 119124 (2010).
61. Caporaso, J. G. et al. QIIME allows analysis of high-throughput community
sequencing data. Nat. Meth. 7, 335336 (2010).
62. Powell, J. R. et al. Deterministic processes vary during community assembly for
ecologically dissimilar taxa. Nat. Commun. 6, 8444 (2015).
63. Morlon, H. et al. A general framework for the distance-decay of similarity in
ecological communities. Ecol. Lett. 11, 904917 (2008).
64. Anderson, M. J. A new method for non parametric multivariate analysis of
variance. Austral Ecol. 26, 3246 (2001).
65. Yamaoka, K., Nakagawa, T. & Uno, T. Application of Akaikes information
criterion (AIC) in the evaluation of linear pharmacokinetic equations.
J. Pharmacokinet. Biopharm. 6, 165175 (1978).
NATURE COMMUNICATIONS | 7:13960 | DOI: 10.1038/ncomms13960 | www.nature.com/naturecommunications
ARTICLE
Acknowledgements
We are grateful to C.Y. Zhang, L.Z. Dai, X.Y. Cheng, K. Yang, J.D. Xu, Y.C. Wu,
J.Z. Zhou, L.Y. Wu, C.Q. Wen, P. Zhang, Y.L. Zhang, X.M. Tang and B.Q. Qin for field
sampling, lab analyses or data providing; to Z.J. Wang and Z.S. An for fund notice; and to
S. Langenheder and J. Stegen for comments on the manuscript. We appreciate CNERN,
Taihu Laboratory for Lake Ecosystem Research for providing data of Taihu Lake. J.W.
was supported by NSFC grants (41273088, 41571058), The Program of Global Change
and Mitigation (2016YFA0600502), Key Research Program of Frontier Sciences, CAS
(QYZDJ-SSW-DQC030), The National Geographic Air and Water Conservation Fund
(GEFC12-14), CAS oversea visiting scholarship (2011-115) and NSFC grant (40903031).
J.So and J.W. were supported by Emil Aaltonen Foundation. J.Sh was supported by 973
Program (2012CB956100) and the international partnership program for creative
research teams (KZZD-EW-TZ-08).
Author contributions
J.W. led in conceiving the ideas, with the contributions from J.So and F.P. J.W.
and F.P. performed the field experiments, sample collection and the analyses of
environmental variables. F.P. carried out the DNA preparation and sequence analyses.
J.W. and J.So performed the data analyses. J.W. led the writing, with the contributions
from the other authors. All co-authors contributed intellectual inputs and commented on
the final version of the manuscript.
Additional information
Supplementary Information accompanies this paper at http://www.nature.com/
naturecommunications
Competing financial interests: The authors declare no competing financial
interests.
Reprints and permission information is available online at http://npg.nature.com/
reprintsandpermissions/
How to cite this article: Wang, J. et al. Nutrient enrichment modifies temperature-
biodiversity relationships in large-scale field experiments. Nat. Commun. 7, 13960
doi: 10.1038/ncomms13960 (2016).
Publishers note: Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
This work is licensed under a Creative Commons Attribution 4.0
International License. The images or other third party material in this
article are included in the articles Creative Commons license, unless indicated otherwise
in the credit line; if the material is not included under the Creative Commons license,
users will need to obtain permission from the license holder to reproduce the material.
To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/
r The Author(s) 2016
9
 1 Supplementary Methods

2 Experimental design

3 We analysed the sediment bacteria, but not the water column bacteria, for the

4 following five reasons. Firstly, the natural aquatic systems are mostly composed of

5 sediment and water; thus, the sediment is an essential part of aquatic microcosms. By

6 adding both sediment and artificial lake water, we experimentally assembled whole,

7 natural aquatic environments. Secondly, sediment acts as a minimal medium with all the

8 necessary nutrients and mineral elements for the growth of microorganisms that are not

9 contained in the artificial lake water. Thirdly, we expected that the sediment bacteria

10 would be less influenced by disturbance effects because the sediment bacteria were in

11 more stable physical conditions than the water column bacteria. The sediment was below

12 the ground surface, and thus the sediment bacteria were less exposed to ambient

13 environmental variations, such as UV radiation, than the bacteria in the water column.

14 Sediment bacteria may also face less grazing pressure from higher organisms than water

15 column bacteria, which are prone to grazing. Fourthly, we expected that the sediment

16 bacteria were less structured by stochastic processes because they were environmentally

17 selected, first by water column and then by sediment. Our previous study showed that

18 deterministic processes are more important in structuring the bacterial communities in

19 sediments than of the overlying water within aquatic environments 1. Finally, compared

20 to water column bacteria, especially those with low nutrient supply and thus with low

21 biomass, sediment bacteria were much easier to collect after siphoning out the overlying

22 water, preserve in sterile tubes in the field conditions, and examine with molecular

23 methods. We faced practical issues with sampling the bacteria in the overlying water
24 because of the very large nutrient gradients. For example, the bacterial biomass in the

25 overlying water at the lowest nutrient treatment (e.g., controls) was too low for the DNA

26 analyses with the small volume of overlying water left over after the chemical analyses.

27 Furthermore, we considered a large number of samples from each mountain, and the

28 logistical issues thus prevented us from efficiently sampling the water-column bacteria.

29 Due to the remote field conditions, we could not manually the filter water samples in situ

30 for bacteria within a reasonable time period, and it was not possible to carry the water

31 back to the laboratory. Even if the water had been transferred to the laboratory, the

32 experimental delay (e.g., in days) would likely to happen for additional water-filtering

33 manipulations.

34 In the one-month-incubation experiments, the sediment bacterial species were

35 selected by two environmental filters, the overlying water and sediments. The simplified

36 experimental ecosystems were highly suitable for the examination of the relative

37 importance of temperature and nutrient factors on bacterial communities. Because

38 bacterial species are mainly dispersed through the air, the colonized communities allowed

39 us to capture the most natural local or regional species pools possible. Due to the multiple

40 dispersal sources of bacterial species at local, regional, and global scales 2, 3, we could not

41 identify the exact origins of sediment bacteria in the microcosms. This challenge was

42 especially prevalent after the bacteria in the sediment were further filtered by the local

43 environments of the overlying water and sediment in the microcosms. Thus, we did not

44 examine bacterial communities of the samples from other local habitats, such as air dust,

45 soil, streams, and lakes, to track the sources of sediment bacteria in the microcosms.
46 We did not use any specific natural aquatic bacterial communities from ponds or

47 lakes in the current experiments, but established new communities via post-dispersal

48 effects. By using sterile microcosms, we simplified the underlying processes for the

49 observed bacterial communities by partly eliminating the legacies of historical events that

50 affect the communities in natural ponds or lakes. This is because, together with

51 contemporary environmental filtering, historical contingencies are one of the main forces

52 affecting present-day microbial communities 4. Compared to the microcosms, natural

53 aquatic environments such as ponds or lakes are influenced by numerous covariant

54 variables that are difficult to control for but that are important for organisms 5, 6, 7. (1) The

55 origin of natural aquatic environments affects the evolution and dispersal history of

56 species in their biological communities. For instance, the selected lakes might have

57 different origin and formation histories that affect their morphology and hydrological

58 conditions. (2) Catchment area, surrounding vegetation, and local physical-chemical

59 conditions also vary spatially and affect the aquatic biota. (3) The initial communities

60 based on natural communities of ponds or lakes can differ spatially thus resulting in very

61 heterogeneous initial communities, and the biological interactions with other organisms

62 are probably more complex in nature than in our experimental microcosms. (4) To have

63 consistent initial communities by inoculating microcosms with natural communities of

64 ponds or lakes (i.e., the sediment bacteria in Taihu Lake) is less feasible for such broad-

65 scale experiments due to the difficulties in preserving and transporting these natural

66 aquatic communities for large distances and long periods of time. (5) Most importantly,

67 using natural communities from ponds or lakes as the initial community does not

68 eliminate the dispersal effects from the air, and thus the potential for strong priority
69 effects 8 that would likely occur. This is because the air is one of the main dispersal

70 routes for bacterial species in nearly all surface aquatic environments 9, 10, 11.

71

72 Physicochemical analyses

73 The concentrations of ammonium (NH4+), nitrate (NO3-), nitrite (NO2-), and dissolved

74 inorganic phosphorus (PO43-) in the overlying water were measured with a flow injection

75 analyser (Skalar SA1000, Breda, Netherlands). The sediment total organic carbon (TOC)

76 was determined using a solid TOC analyser (SSM-5000A, Shimadzu, Japan). The

77 sterilized sediments before the field experiments were measured for the above variables

78 in three replicates.

79 We did not measure all the potentially important variables for the bacterial

80 communities in the sediment, such as oxygen concentration or UV radiation. This is

81 largely because we have no evidence about the potential importance of these variables on

82 sediment microbial communities. Further, we do not expect these potential effects to be

83 stronger than nutrient and temperature effects considered in our study. This is because the

84 bottles were buried in the local soil to approximately 10% of the total bottle height, and

85 the sediment in each bottle was below the ground surface. Therefore, it is unlikely that

86 UV radiation would have had strong effects on these communities that would mask the

87 temperature or nutrient effects on the bacteria. In addition, we added 15 g sterilized

88 sediment to the bottom of each bottle, which created a sediment layer of approximately 5

89 mm. Based on our previous studies on the oxygen concentrations at the sediment-water

90 interface in eutrophic lakes, e.g., Wang et al. 12, 13, the 5-mm sediment layer would be less

91 likely to be anoxic during the experimental period, especially with the shallow overlying
 92 water column (approximately 18 cm) and low total organic carbon (0.587 0.390%).

93

94 Bacterial community analyses

95 Genomic DNA was extracted from freeze-dried sediments using the phenol

96 chloroform method 14. Real-time qPCR quantification of bacterial 16S rRNA genes in

97 sediments was performed on an iCycler iQ5 thermocycler (Bio-Rad, Hercules, USA) as

98 described previously 15. Bacterial 16S rRNA genes were sequenced using MiSeq

99 (Illumina, San Diego, USA) as described previously 16. We amplified 16S rRNA genes in

100 triplicate using the universal bacterial primers 515F, 5-GTGCCAGCMGCCGCGGTAA-

101 3 and 806R, 5-GGACTACHVGGGTWTCTAAT-3, targeting the V4 region. Positive

102 PCR products were confirmed with agarose gel electrophoresis. PCR products from

103 triplicate reactions were combined and quantified with PicoGreen. PCR products from

104 samples sequenced in the same MiSeq run were pooled at equal molality. The pooled

105 mixture was purified with a QIAquick Gel Extraction Kit (QIAGEN Sciences,

106 Germantown, USA) and re-quantified with PicoGreen. Sample libraries were prepared

107 according to the MiSeqTM Reagent Kit Preparation Guide (Illumina, San Diego, USA).

108 We processed the sequences primarily using the QIIME pipeline (v1.8) 17 following

109 recent references e.g., 16, 18. Briefly, overlapped paired-end sequences from MiSeq were

110 assembled using FLASH 19 and poorly overlapped and poor quality sequences were

111 filtered out before de-multiplexing based on barcodes. Then, the sequences were

112 clustered into OTUs at 97% pairwise identity with the seed-based uclust algorithm 20.

113 After chimeras were removed via Uchime, and representative sequences from each OTU

114 were aligned to the Greengenes imputed core reference alignment V.201308 21 using
115 PyNAST 22. The taxonomic identity of each representative sequence was determined

116 using the RDP Classifier 23 and chloroplasts were removed.

117 We removed singletons before the subsequent analyses, which is a common

118 procedure in microbial studies, and has been highly recommended in popular microbial

119 analysis pipelines even before the generation of OTU tables (Supplementary Fig. 12). We

120 found that the removal of singletons and rare species (e.g., those occurring in less than 6

121 samples) did not notably affect the patterns in species richness in our data

122 (Supplementary Fig. 12). To keep the reported results as simple as possible, we used the

123 community OTU table without singletons for all analyses. For Archaea, we obtained 102

124 sequences from 42 out of 300 total samples. These archaeal sequences can be classified

125 into 13 OTUs related to the families Nitrososphaeraceae, Nitrosocaldaceae,

126 Methanocellaceae, Methanobacteriaceae, Methanomicrobiaceae, Methanosaetaceae,

127 Sulfolobaceae, and DHVEG-1. Due to the insufficient sequence number for Archaea, we

128 excluded all archaeal sequences from the data.

129 Supplementary Tables

130 Supplementary Table 1. Regression analyses of biodiversity with environmental

131 variables. Multiple regression analyses of bacteria diversity and community composition
132 were performed as a function of environmental variables. For diversity and community
133 composition, species richness and the first axis of non-metric multidimensional scaling,
134 respectively, were used as the response variables. The best models were identified using
135 Akaikes information criterion, and variables were selected in multiple regressions only if
136 P < 0.05. All of the variables were standardized (mean = 0; SD = 1) and are displayed
137 with increasing P-values. Temp: measured water temperature. Temp.2: the squared value
138 of measured water temperature. ADD.NO3: the initially added NO3-. NO3, NO2, and pH:
139 water NO3- and NO2-, and pH, respectively. Chl a: sediment Chl a (mgg-1). Copies: copy
140 number of 16S rRNA genes qualified by real-time qPCR (copy number per gram dry
141 sediment).
142
Model
Response Mountains Explanatory variables and Beta-weights$
r2
Diversity Norway 0.595 Temp *** Temp.2 *** NO2 *** pH *
Chl a
Diversity China 0.397 Temp *** Temp.2 *** NO2 *** *

*** *** ***

Composition Norway 0.605 Temp Temp.2 NO2 pH ***
Temp.2 NO3
Composition China 0.880 *** pH *** NH4 *** ** ADD.NO3 * Copies *
$
143 Standardized partial regression coefficients, ***P < 0.001; **P < 0.01; *P < 0.05.
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169 Supplementary Figures
China, 0 China, 0.45 China, 1.8 China, 4.05 China, 7.65

0.03

0.02



0.01












0.00

China, 11.25 China, 15.75 China, 21.6 China, 28.8 China, 36

0.03
Chl a mg g1 dry sediment

0.02






0.01

















0.00

Norway, 0 Norway, 0.45 Norway, 1.8 Norway, 4.05 Norway, 7.65

0.03
0.02



0.01

0.00

Norway, 11.25 Norway, 15.75 Norway, 21.6 Norway, 28.8 Norway, 36

0.03

0.02

0.01






0.00
.0
.5
.0
.5

.0
.5
.0
.5

.0
.5
.0
.5

.0
.5
.0
.5

.0
.5
.0
.5
39
39
40
40

39
39
40
40

39
39
40
40

39
39
40
40

39
39
40
40

1 kT (eV1)
170
171
172 Supplementary Figure 1. The relationships between sediment Chlorophyll a and
173 temperature. The relationships were analyzed for each nutrient enrichment level and
174 location, and we used the inverse absolute temperature (1/kT), where k is Boltzman's
175 constant 8.62 10 -5 eV K -1, and T is absolute temperature in Kelvin. These relationships
176 are fitted by linear and quadratic models, the significance of which (P < 0.05, F-test) are
177 shown with blue and red lines (solid or dotted). The better model, shown in solid line,
178 was selected based on lower value of Akaikes information criterion. Gray dotted lines
179 indicate non-significant relationships of a linear model (P > 0.05, F-test).
180
181
182 Supplementary Figure 2. The relationships between sediment Chlorophyll a and
183 nutrient enrichment. These relationships for each elevation band were fitted by linear and
184 quadratic models, the significance of which (P < 0.05, F-test) are shown with blue and
185 red lines (solid or dotted). The better model, shown in solid line, was selected based on
186 lower value of Akaikes information criterion. Gray dotted lines indicate non-significant
187 relationships of a linear model (P > 0.05, F-test).
188
189
190 Supplementary Figure 3. The relationships between overlying water pH and
191 temperature. These relationships for each nutrient enrichment and location were fitted by
192 linear and quadratic models, the significance of which (P < 0.05, F-test) are shown with
193 blue and red lines (solid or dotted). The better model, shown in solid line, was selected
194 based on lower value of Akaikes information criterion. Gray dotted lines indicate non-
195 significant relationships of a linear model (P > 0.05, F-test).
196
197
198 Supplementary Figure 4. The relationships between overlying water pH and nutrient
199 enrichment. These relationships for each elevation band were fitted by linear and
200 quadratic models, the significance of which (P < 0.05, F-test) are shown with blue and
201 red lines (solid or dotted). The better model, shown in solid line, was selected based on
202 lower value of Akaikes information criterion. Gray dotted lines indicate non-significant
203 relationships of a linear model (P > 0.05, F-test).
204
205
206 Supplementary Figure 5. The relationships between sediment Chlorophyll a and
207 overlying water pH. The relationships were fitted with linear models and significances
208 were examined with F-test. (A) The linear relationship between Chlorophyll a and pH for
209 each region. (B) The slopes of Chlorophyll a-pH relationships along nutrient enrichment.
210 (C, D) The Chlorophyll a-pH relationships along elevational gradients. Solid dots
211 indicate the significant (P < 0.05, F-test) relationships (B-D). Red colour for China and
212 blue colour for Norway.
213
214
215 Supplementary Figure 6. Correlations between environmental variables and bacterial
216 biodiversity. We only included the significant (P < 0.05) environmental variables related
217 to community similarity based on Mantel test (A, B), the first axis of non-metric
218 multidimensional scaling of the communities with Pearson correlations (C, D) and
219 species richness with Pearson correlations (E, F). Temperature: Water temperature in situ.
220 pH: Water pH. Chla_mg.per.g: Chl a in the sediments (mgg-1). NH4, NO2, and NO3:
221 NH4+, NO2-, and NO3- in the water at the end of experiments. ADD.NO3 and ADD.PO4:
222 initial NO3- and PO43+ added to the microcosms. NO3.diff and PO4.diff: NO3- and PO43+
223 differences between the start and end of the experiments. TC.sediment and TP.sediment:
224 total organic carbon and total phosphorus in the sediments at the end of experiments.
225 Copies.16S_per.g: copy number of 16S rRNA genes qualified by real-time qPCR (copy
226 number per gram dry sediments). Comp.1, Comp.2, Comp.3, Comp.4 and Comp.5:
227 principal components of metal concentrations (such as K, Na, Ca, Mg, Al, Fe, Zn, Cu, Cr
228 and Ni).
229
230
231 Supplementary Figure 7. Canonical correspondence analysis for the bacterial
232 communities. A is for Norway, and B is for China. Detrended correspondence analysis
233 was used to determine that the gradient length of the species abundance matrix is larger
234 than 2.0 along the first axis, thus implying a unimodal species responses. Thus,
235 considering the long environmental gradients (especially for nutrients), we applied
236 canonical correspondence analysis (CCA) to examine the relationships between bacterial
237 community compsition and explanatory variables. All environmental variables, except
238 pH, were log-transformed [by log(X + 1) or log(1000X + 1)]. The abbreviations of
239 environmental variables are given in Supplementary Figure 6. The elevations (units: m
240 a.s.l) were shown with different colors. The significance of variables was assessed with
241 permutation test (n = 1,000). *, P < 0.05; **, P < 0.01; ***, P < 0.001.
242
243
244 Supplementary Figure 8. The community Srensen similarity between elevations of the
245 two mountains. Blue: The Laojun Mountain, China. White: The Balggesvarri Mountain,
246 Norway. The elevations (unit: m a.s.l.) of each site are shown in the nodes. Thicker edges
247 indicate high Srensen similarity in bacterial composition between elevations within and
248 between regions.
249
 China, 0 China, 0.45 China, 1.8 China, 4.05 China, 7.65








7.0

















6.5

6.0

China, 11.25 China, 15.75 China, 21.6 China, 28.8 China, 36

7.0













6.5
Species richness

6.0

Norway, 0 Norway, 0.45 Norway, 1.8 Norway, 4.05 Norway, 7.65

7.0





6.5










6.0

Norway, 11.25 Norway, 15.75 Norway, 21.6 Norway, 28.8 Norway, 36

7.0



6.5














6.0







.0
.5
.0
.5

.0
.5
.0
.5

.0
.5
.0
.5

.0
.5
.0
.5

.0
.5
.0
.5
39
39
40
40

39
39
40
40

39
39
40
40

39
39
40
40

39
39
40
40
1 kT (eV1)
250
251 Supplementary Figure 9. The relationships between temperature and bacterial species
252 richness. These relationships for each nutrient enrichment and location were fitted by
253 linear and quadratic models, the significance of which (P < 0.05, F-test) are shown with
254 blue and red lines (solid or dotted). Species richness was log-transformed, and we used
255 the inverse absolute temperature (1/kT), where k is Boltzman's constant 8.62 10 -5 eV K
-1
256 , and T is absolute temperature in Kelvin. The better model, shown in solid line, was
257 selected based on lower value of Akaikes information criterion. Gray dotted lines
258 indicate non-significant relationships of a linear model (P > 0.05, F-test).
259
260
261 Supplementary Figure 10. The nutrient concentrations in Taihu Lake, China, in the
262 years 1997-2015. We considered total nitrogen (TN), which was measured monthly by
263 CNERN, Taihu Laboratory for Lake Ecosystem Research. We selected six monitoring
264 sites from river mouth (THL00, THL06, THL10, and THL11), lake center (THL08), and
265 macrophyte-dominated region (THL12), which present clear nutrient variations
266 geographically. Upper panel: The relative frequency histogram of TN concentrations for
267 each site. Lower panel: The boxplot of TN through the years for each site. The red dotted
268 lines indicate the intermediate nutrient enrichment observed in this study: 4.05 and 7.65
269 mg N L-1 NO3-, respectively. These results showed that observed TN concentrations in
270 Taihu Lake, exhibiting a high spatial heterogeneity within the lake, are frequently higher
271 than the intermediate nutrient enrichment values in our experiments (i.e., ~4.05-7.65 mg
272 N L-1 NO3-).
273
274

275 Supplementary Figure 11. The loadings of observed variables for the latent variables in
276 path models. The observed variables were used for the partial least squares path modeling
277 (Fig. 4), and we showed their loadings for the latent variables. Most of the loadings were
278 selected with the threshold 0.7, as suggested in previous studies 24. The observed
279 variables included for each latent variable were shown in difference colors. There latent
280 variables are Temperature (Tem), Nutrient (Nut), Productivity (Pro) and Diversity (Div)
281 or Community composition (Com). Temperature: measured water temperature.
282 Temperature.2: the squared value of measured water temperature. ADD.NO3: the initially
283 added NO3-. NO3, NO2, and pH: water NO3- and NO2-, and pH, respectively.
284 Chla_mg.per.g: sediment Chl a (mgg-1).
285
286
287 Supplementary Figure 12. The effects of removal of singleton or other rare species on
288 species richness.
289 We removed the singletons or other rare species from the full community data set
290 (that is, with singletons and 300 total samples). Singleton was defined as a read with a
291 sequence that is present exactly once, i.e. is unique among the all sequence reads. The
292 rare species was defined continuously as the species occuring in only 1 sample, 2, 3, . ,
293 or 100 samples. For each community with species removal, gamma diversity (i.e., x-axis
294 in panel A) is the total OTU number of full data set, and mean species richness (i.e., x-
295 axis in panel B) is the average species richness of the 300 samples. Gamma diversity and
296 mean species richness were highest with singletons included, followed by the removal of
297 the rare species (i.e., occuring in only 1 sample, 2, 3, . , or 100 samples).
298 For the obtained diversity (A: gamma diversity; B: mean species richness), we
299 calculated the Pearson correlations (i.e., y-axis in panels A and B) between diversity of
300 the communities without singletons and that of the communities with singletons or the
301 communities without rare species. All Pearson correlations were significant (P < 0.01),
302 and Pearson r decreased continously along the gradients of both gamma diversity (A) and
303 mean species richness (B). For the community data (in red color) with the removal of
304 singletons and rare species occuring in less than 6 samples, the Pearson correlations were
305 quite similar, and all higher than 0.995, indicated by gray horizontal lines (A, B).
306 These results showed that the effects of OTU removal on the values of species
307 richness were low when the singletons and other rare species were excluded.
308 Furthermore, it is common in microbial studies to remove singletons before detailed
309 ecological analyses, and it is also highly recommended in microbial analysis pipelines
310 (such as usearch, http://www.drive5.com/usearch/manual/singletons.html; Qiime,
311 http://qiime.org/tutorials/open_reference_illumina_processing.html). Thus, to keep our
312 reported results as simple as possible, we finally used the OTU tables of bacterial
313 community without singletons.
314 Supplementary References
315
316 1. Wang J, et al. Phylogenetic beta diversity in bacterial assemblages across
317 ecosystems: deterministic versus stochastic processes. ISME J 7, 1310-
318 1321 (2013).
319
320 2. Liu Y, et al. Bacterial responses to environmental change on the Tibetan
321 Plateau over the past half century. Environ Microbiol 18, 1930-1941
322 (2016).
323
324 3. Barbern A, et al. Continental-scale distributions of dust-associated
325 bacteria and fungi. Proc Natl Acad Sci U S A 112, 5756-5761 (2015).
326
327 4. Martiny JBH, et al. Microbial biogeography: Putting microorganisms on
328 the map. Nat Rev Microbiol 4, 102-112 (2006).
329
330 5. Wetzel RG. Limnology. Academic Press (2001).
331
332 6. MacArthur RH, Wilson EO. The theory of island biogeography. Princeton
333 Univ Pr (1967).
334
335 7. Whittaker RJ, Fernndez-Palacios JM. Island biogeography: ecology,
336 evolution, and conservation. Oxford University Press, USA (2007).
337
338 8. Fukami T. Historical Contingency in Community Assembly: Integrating
339 Niches, Species Pools, and Priority Effects. Annu Rev Ecol Evol Syst 46,
340 1-23 (2015).
341
342 9. Hervs A, Camarero L, Reche I, Casamayor EO. Viability and potential
343 for immigration of airborne bacteria from Africa that reach high mountain
344 lakes in Europe. Environ Microbiol 11, 1612-1623 (2009).
345
346 10. Hervas A, Casamayor EO. High similarity between bacterioneuston and
347 airborne bacterial community compositions in a high mountain lake area.
348 FEMS Microbiol Ecol 67, 219-228 (2009).
349
350 11. Barbern A, Henley J, Fierer N, Casamayor EO. Structure, inter-annual
351 recurrence, and global-scale connectivity of airborne microbial
352 communities. Sci Total Environ 487, 187-195 (2014).
353
354 12. Wang J, Shen J, Zhang L, Liu E. Spatial heterogeneity of oxygen
355 exchange between sediment-water interface in lakes. J Lake Sci 21, 474-
356 482 (2009).
357
358 13. Wang J, Shen J, Zhang L, Fan C, Li W, Pan J. Sediment-water nutrient
359 fluxes and the effects of oxygen in Lake Dianchi and Lake Fuxian, Yunnan
360 Province. J Lake Sci 22, 640-648 (2010).
361
362 14. Zhou J, Bruns MA, Tiedje JM. DNA recovery from soils of diverse
363 composition. Appl Environ Microbiol 62, 316-322 (1996).
364
365 15. Shen J, Zhang L, Guo J, Ray JL, He J. Impact of long-term fertilization
366 practices on the abundance and composition of soil bacterial communities
367 in Northeast China. Appl Soil Ecol 46, 119-124 (2010).
368
369 16. Wang J, et al. Regional and global elevational patterns of microbial
370 species richness and evenness. Ecography, DOI: 10.1111/ecog.02216
371 (2017).
372
373 17. Caporaso JG, et al. QIIME allows analysis of high-throughput community
374 sequencing data. Nat Meth 7, 335-336 (2010).
375
376 18. Wang J, Soininen J, He J, Shen J. Phylogenetic clustering increases with
377 elevation for microbes. Environ Microbiol Rep 4, 217-226 (2012).
378
379 19. Mago T, Salzberg SL. FLASH: fast length adjustment of short reads to
380 improve genome assemblies. Bioinformatics 27, 2957-2963 (2011).
381
382 20. Edgar RC. UPARSE: highly accurate OTU sequences from microbial
383 amplicon reads. Nat Meth 10, 996-998 (2013).
384
385 21. DeSantis TZ, et al. Greengenes, a chimera-checked 16S rRNA gene
386 database and workbench compatible with ARB. Appl Environ Microbiol
387 72, 5069-5072 (2006).
388
389 22. Caporaso JG, Bittinger K, Bushman FD, DeSantis TZ, Andersen GL,
390 Knight R. PyNAST: a flexible tool for aligning sequences to a template
391 alignment. Bioinformatics 26, 266-267 (2010).
392
393 23. Wang Q, Garrity GM, Tiedje JM, Cole JR. Naive Bayesian Classifier for
394 Rapid Assignment of rRNA Sequences into the New Bacterial Taxonomy.
395 Appl Environ Microbiol 73, 5261-5267 (2007).
396
397 24. Sanchez G. PLS path modeling with R. Trowchez Editions (2013).
398
399