Biodiversity and species competition regulate the ...download.xuebalib.com/3zariXf1ekSD.pdf · ORIGINAL ARTICLE Biodiversity and species competition regulate the resilience of microbial
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
OR I G I N A L A R T I C L E
Biodiversity and species competition regulate the resilience ofmicrobial biofilm community
1CAS Key Laboratory for Environmental Biotechnology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing, China
2Department of Chemical and Biochemical Engineering, Technical University of Denmark, Lyngby, Denmark
3Sino-Danish Center for Education and Research, Beijing, China
4University of Chinese Academy of Sciences, Beijing, China
5State Key Laboratory of Industrial Ecology and Environmental Engineering (Ministry of Education, China), School of Environmental Science and Technology,
Dalian University of Technology, Dalian, China
6State Key Laboratory of Urban Water Resource and Environment, Harbin Institute of Technology (SKLUWRE, HIT), Harbin, China
7Guangdong Provincial Key Laboratory of Microbial Culture Collection and Application, Guangdong Institute of Microbiology, Guangzhou, China
8School of Minerals Processing and Bioengineering, Central South University, Changsha, China
9College of Resources and Environment, University of Chinese Academy of Sciences, Beijing, China
10School of Environmental Science and Engineering, Sun Yat-sen University, Guangzhou, China
11Department of Microbiology and Plant Biology, Institute for Environmental Genomics, University of Oklahoma, Norman, OK, USA
is the sum of total phylogenetic branch length of OTUs in each sample.
The selected representative OTU sequences were aligned using
PyNAST (Caporaso et al., 2010) against with GreenGene data set, and
tree file was generated using FastTree program (Price, Dehal, & Arkin,
2009). The PD was calculated using Picante package in R (v.3.2.5)
(Kembel et al., 2010). Unweighted principal coordinate analysis (PCoA)
based on UniFrac matrix (Lozupone, Hamady, Kelley, & Knight, 2007;
Lozupone, Hamady, & Knight, 2006; Lozupone & Knight, 2005) was
used for microbial community structure changes. Permutational multi-
variate analysis of variance (PERMANOVA) was used to test the dis-
similarity among three groups using both Bray–Curtis and Jaccard
distance methods via vegan package (v.2.3-5) in R (v.3.2.5). Three cor-
relation approaches involving Pearson, Kendall and Spearman were
applied to correlate the recovery time with a-diversity. The signifi-
cance between two groups was determined by two-tailed Student’s t
test, and significance comparing three groups was obtained using one-
way analysis of variance (ANOVA).
2.6 | Network construction with RMT-basedapproach and network analysis
To elucidate microbial interactions in biofilms in response to pH distur-
bance, we constructed three groups of phylogenetic molecular ecologi-
cal networks (pMENs) via random matrix theory-based interface
approach in molecular ecological network analysis pipeline (MENA,
http://ieg2.ou.edu/MENA/) (Deng et al., 2012; Zhou et al., 2010,
2011). The whole process was described in a previous study compre-
hensively (Deng et al., 2012) and we briefly introduce the procedure
used here. First, only the OTUs appeared in more than nine samples
(18 samples totally for each group) were kept without log-transferring
prior to obtaining Spearman rank correlation matrix (r value). A series
of thresholds from 0.01 to 0.95 with 0.01 intervals were obtained and
applied to the matrix. Only the correlations above a specific threshold
were kept for calculating the network eigenvalues. The finest threshold
was selected when the nearest-neighbour spacing distribution fol-
lowed Poisson distribution well, which is related to specific and non-
random properties of a complex system (Luo et al., 2007). In addition,
to compare these three networks under same conditions, a uniform
threshold (0.79) was determined to generate three networks. And fur-
ther network properties including R2 of power law, average connectiv-
ity, average path length, average clustering coefficient and modularity
were all calculated in MENA pipeline (Deng et al., 2012) (Table S5).
Then the constructed networks were visualized using Cytoscape 3.3.0
software (Shannon et al., 2003) for three different networks of MEC
anodic biofilm communities in response to pH disturbance.
3 | RESULTS
3.1 | The resilience of bioreactors to pHdisturbance
The generated microbial communities with different diversity were
shocked by pH disturbance for four cycles, and the resilience of
these communities was evaluated using measures of functional per-
formance and observed recovery time. Two representative parame-
ters of performance (hydrogen production rate and total energy
efficiency) are shown in Figure 1. The average hydrogen production
rates for Group A, B and C were, respectively, 1.38 � 0.17,
1.28 � 0.14 and 1.26 � 0.11 ml/(ml reactor) per day before the pH
disturbance (One-way ANOVA, p < .01). Group A had a higher
hydrogen production rate as compared with Group B (Student’s t
test, t = 2.425, p = .020) and Group C (t = 2.833, p = .007), while
Group B and Group C obtained similar hydrogen production rates
(t = 0.311, p = .757). After pH disturbance and recovering to stable
state close to original predisturbance status, similar hydrogen pro-
duction rates were obtained, 1.05 � 0.05, 0.98 � 0.05, and
1.02 � 0.05 ml/(ml reactor) per day for Group A, B and C, respec-
tively (One-way ANOVA, p = .24) (Figure 1a). In addition, pH distur-
bance significantly decreased the H2 production rate for each group
(Figure 1a). Other performance parameters (total energy efficiency,
hydrogen recovery rate and electrical energy recovery) exhibited
similar patterns in response to pH disturbance while less effect of
F IGURE 1 Hydrogen production rate (a) and total energyefficiency (b) of MEC groups in response to pH disturbance,subsidiary with one-way ANOVA analysis and Student’s t testresults, t value (p-value). Bold font means the significance at p < .05level. Pre: predisturbance; ***p < .001 of Student’s t test
pH disturbance was shown on coulombic efficiency (Figure 1b;
Fig. S2). Therefore, the three groups of MECs performed with little
difference in reactor effectiveness during the same period (predistur-
bance or recovery), and pH disturbance had negative effects on
MEC performance indicated by their effectiveness indexes.
Recovery time was used to assess the resilience of MECs in a time
scale. Five parameters related to monitoring and effectiveness were
selected to calculate the recovery time for each group (Fig. S3). Group
A with the highest initial concentration of inocula took the shortest
amount of time to reach stable stage after the pH disturbance com-
pared to Group C with lower concentration of inocula. Group B might
experience a transitional state from Group A to Group C in this experi-
ment. Additionally, the Student’s t test exhibited that four of five
recovery time showed significant difference between Group A and
Group C (p < .05, Table S1). The recovery time was 82.6 � 0.1 hr,
117.8 � 17.7 hr and 126.0 � 25.4 hr for Group A, B and C, respec-
tively, which was comparably precise and was determined as default
recovery time by statistical analysis and other calculations.
3.2 | Sequencing statistics and microbial diversity
To determine the biodiversity of biofilm communities of MECs, the V4
region of 16S rRNA gene was amplified and sequenced using high-
throughput sequencing. After quality control, a total of 2,992,109
sequences were classified into 54 biofilm samples and two activated
sludge samples (inocula). An average of 53,431 � 15,485 sequences
per sample was obtained, and the rarefaction curves indicated these
numbers of sequences were fairly enough (Fig. S4). When the opera-
tional taxonomy unit table was generated, we randomly resampled
23,741 sequences per sample for further analyses, for example micro-
bial diversity, composition and structure.
Shannon and Inverse Simpson indexes were calculated to evaluate
the a-diversity of anodic microbial communities under different condi-
tions. Group A obtained the highest a-diversity (Figure 2) and Group C
obtained the lowest, showing a significant difference for the predistur-
bance period (t = 3.585, p = .005) as the experiment was designed.
However, this significant difference was not observed between Group
A and C at the end of recovery period (t = 2.200, p > .05), indicating
that the pH shock might eliminate the difference in a-diversity among
three groups (Figure 2). Besides, fewer effects of pH disturbance were
observed in Group B and Group C (p > .05) at the average level, but
still exhibited a decreasing tendency in response to pH disturbance
(Figure 2). The other three diversity characteristics (Chao1, Richness,
Phylogenetic diversity [PD]) all showed similar patterns to Shannon
and Inverse Simpson indexes in response to pH disturbance (Fig. S5).
These results indicated a-diversity of biofilm community was signifi-
cantly decreased by pH disturbance.
3.3 | Structure and composition of biofilm microbialcommunities
To explore the structure variance of biofilm microbial communities,
b-diversity-based statistical tools were employed. The principal
coordinate analysis (PCoA) showed that microbial communities
belonging to the same group were more closely clustered with each
other, but more distinctly separated from other groups (Figure 3).
Group A and Group C were almost separated with PCoA, while they
were totally divergent by nonmetric multidimensional scaling analysis
(NMDS) (Fig. S6). In addition, along with the arrow direction from
Group A to Group C (Figure 3), microbial communities at the end of
recovery period exhibited a tendency to cluster more tightly, which
implied that less difference was observed across groups after micro-
bial community recovered from the pH shock. It implied that pH dis-
turbance decreased the difference among groups in microbial
structure in b-diversity as well as a-diversity.
Further dissimilarity tests revealed that the significant difference
was observed within groups and across three periods. Permutational
multivariate analysis of variances (PERMANOVA) was used to con-
firm the significance results (Table 1). This method based on both
Jaccard and Bray–Curtis distance showed significant differences
F IGURE 2 Comparisons of two a-diversity indexes, Shannonindex (a) and Inverse Simpson index (b), for Group A, B and Cbiofilms at different periods in response to pH disturbance.*p < .05; **p < .01; ***p < .001 based on Student’s t test. Thecolours of asterisk correspond to different periods for pHdisturbance. The column value is the mean of the indices withineach group (n = 6), and error bars stand for standard deviation (SD)
FENG ET AL. | 6175
among groups across all three periods. On the other hand, compar-
ing the difference among three groups at same period, the significant
difference also decreased from predisturbance period
(FPERMANOVA = 6.7591, p < .001) to the end of recovery period
(FPERMANOVA = 3.6889, p = .002) based on the Bray–Curtis distance
(Table S2). Moreover, the difference between each group as indi-
cated by Bray–Curtis and Jaccard index also exerted similar decreas-
ing tendency in response to pH disturbance (Table S3)
To determine the underlying mechanisms for shedding light on
a-diversity and b-diversity previously performed across three biofilm
groups, relative abundance of micro-organisms was conducted at the
phylum/class and genus levels with a similarity of 97% for OTU clas-
sifications (Fig. S7). Obviously, all microbes at the phylum/class level
showed a dynamic equilibrium in response to pH disturbance, espe-
cially a large fluctuation at the end of disturbance compared to the
original states (Fig. S7a). For example, Deltaproteobacteria, which
was the most abundant class in MEC anodic biofilms, decreased
from 67.9% � 13.9% to 44.0% � 9.9% after the four-day pH distur-
bance in Group A. Even so, it rebounded back to 68.3% � 5.4%
after totally recovery. Similar resilient trend was observed in Group
C as well. Other phyla/classes showed a range of resilient variation
in response to pH disturbance, such as Euryarchaeota, Firmicutes
and Bacteroidetes.
With higher resolution of the taxonomy for micro-organisms in
biofilms, there were mainly two dominant genera, Geobacter and
dominated in all three groups and across different periods (>1% in
relative abundance), showing their potential functions for maintaining
functional stability of MECs in response to pH disturbance, and this
genus in three groups across the periods was resilient in relative
abundance. Other exoelectrogens (Pseudomonas and Arcobacter) all
varied in response to pH disturbance and their relative abundances
kept stable in each group at the end of recovery. For the other dom-
inant genus, Methanobrevibacter, a group of nonexoelectrogen, its
relative abundance increased during the pH disturbance to become a
more significant member of the community at the end of recovery in
each group. Thus, this unusual trend of Methanobrevibacter might
relate to different MEC resilience across the three groups. Moreover,
the final relative abundance across the three groups exhibited less
difference (Fig. S7b) and was supported by the comparison of simi-
larity indexes which showed less dissimilarity among groups
(Table S3), confirming that biofilm microbial communities tended to
be clustered together at the genus level after the pH disturbance
(Figure 3).
3.4 | Correlations of MEC resilience and microbialdiversity
To explore the relationship between biodiversity of biofilm microbial
communities and MEC resilient performance, the correlations of the
recovery time with five a-diversity indexes were tested using three
methods (Table 2). Five a-diversity indexes describe the biodiversity
of microbial communities from qualitative description (Richness and
Chao1), quantitative evaluation (Shannon and Inverse Simpson) and
phylogenetic perspective (PD) thus could provide comprehensive
correlation results. Even though the correlation coefficients varied
differently for each column (across three correlation methods), the
correlation results all showed significantly negative interactions
between recovery time and diversity indexes (p < .05, Table 2). This
correlation indicated that MECs with higher diversity biofilm commu-
nities could be more resilient in response to pH disturbance. For
example, Group A with higher diversity needed a shorter recovery
time compared to Group C (Figure 2, Fig. S3).
The relative abundance of the genera (Fig. S7b) was tested with
recovery time as well (Table S4). For the two dominant genera,
Geobacter did not show correlations with the recovery time
(p > .05), while Methanobrevibacter exhibited significantly a positive
correlation with the recovery time (p < .05). This result confirmed
that Methanobrevibacter might be associated with recovery of the
biofilm community in response to pH disturbance. A high relative
F IGURE 3 Principal coordinate analysis (PCoA) of biofilmmicrobial communities for Group A (red), B (blue) and C (orange) inresponse to pH disturbance based on unweighted UniFrac distanceof detected OTUs. The sample colours from light to dark stand fortime scaling coupled with different shapes: predisturbance (■),postdisturbance (●), the end of recovery (▲). The associated arrowsindicate the clustering directions for microbial samples of eachbiofilm community structure
TABLE 1 Dissimilarity tests of biofilm microbial communitiesusing PERMANOVA based on Jaccard and Bray–Curtis distance
PERMANOVA test
Jaccard Bray–Curtis
F p F p
Group (A, B, C) 6.3928 .001*** 13.3703 .001***
Period (Pre, Post, Recovery) 3.6011 .001*** 7.6736 .001***
Group*Period 1.4872 .005** 2.265 .008**
*Difference is significant at 0.05 level.
**Difference is significant at 0.01 level.
***Difference is significant at 0.001 level.
6176 | FENG ET AL.
abundance of Methanobrevibacter may increase the time it takes for
the whole community to recovery. For example, Group C with high
abundance of Methanobrevibacter showed a delayed recovery in
response to pH disturbance (Fig. S3, Fig. S7b).
3.5 | Networks reveal the recovery mechanismswith microbial interactions
To explore the microbial interactions within biofilm microbial com-
munities, phylogenetic molecular ecological networks (pMENs) were
constructed. Three groups of anodic microbial communities were
analysed, and the topological properties of networks are summarized
(Table S5). The overall topology indices showed that all network con-
nectivity distributions fitted well with the power-law model (R2 of
power law from 0.771 to 0.843), indicating that most of nodes had
fewer neighbours while fewer nodes had many neighbours in a net-
work (scale-free). Average connectivity (avgK, from 4.235 to 7.191)
measured the complexity of a network, thus Group A obtained a
more complex network than the other two groups (Fig. S8). Also,
average path lengths (GD) were varied from 2.765 to 5.362, which
were nearly close to the logarithm of total number of network size
and significantly comparable to other networks showing typical
small-world network properties (Watts & Strogatz, 1998). This indi-
cated all nodes were well connected in the networks. Finally, the
modularity value (M) for all groups varied from 0.455 to 0.682,
which were all significantly higher than the M values of correspond-
ing randomized networks, indicating all pMENs appeared to be mod-
ular. Moreover, by generating random networks from the empirical
networks, the results indicated that network indices (e.g., average
path length, average clustering coefficient, modularity) were all sig-
nificant different between any two networks of those three groups
(Table S5).
To better understand the interactions among different micro-
organisms, the three networks were visualized, showing quite differ-
ent network structures (Fig. S8). Group A had the most complex
network with a large number of Proteobacteria (mainly Geobacter),
and the Euryarchaeota (mainly Methanobrevibacter) were almost
entirely clustered into a single module with fewer connections to
other modules. Group B exhibited a transient state for Eur-
yarchaeota and Proteobacteria interactions. Group C showed that
the group of Euryarchaeota had negative interactions with other
groups especially the predominant Proteobacteria, indicating consid-
erable negative interactions in the network. For each group of bio-
films, all Geobacter and Methanobrevibacter and their interactions
were split into three subnetworks (Figure 4). It was clear that the
group of Geobacter with high connected interactions took important
positions in the network of Group A and the interactions between
these major nodes were mainly positive. As for Group B, members
of the Methanobrevibacter were shown up gradually and became
essential nodes with an increasing number of interactions in the net-
work. More importantly, all the interactions between Geobacter and
Methanobrevibacter were negative, which implied competitive rela-
tionships for these two genera in biofilms. This pattern was more
obvious in the network of Group C with largest negative interac-
tions between these two genera (Figure 4d). The number of Geobac-
ter nodes in the networks decreased from Group A to Group C,
while it was the opposite for Methanobrevibacter, and the number of
negative interactions between Geobacter and Methanobrevibacter
increased (Figure 4d). Positive interactions were kept inside the gen-
era, and negative interactions were the only relationship between
these two essential groups of microbes. Moreover, the positive
interactions between the predominant Geobacter and those fermen-
tative bacteria that included Geothrix, Oscillospira, Acetobacterium,
Paludibacter, Anaerovorax, Bacteroides and Proteiniclasticum were
observed in the network analysis across three groups as well
(Fig. S9), which might facilitate the functional resilience of biofilm
community.
4 | DISCUSSION
The relationship of biodiversity and ecosystem stability has been
debated in macro ecosystems for many years (Bezemer & van der
Putten, 2007; May, 2001; Steudel et al., 2012). The appropriate
experiments on this ecological theory in microbial ecosystems remain
less explored, due to the high complexity, high species diversity and
intractability; thus, tractable model systems are needed to explore
mechanisms of perturbing microbial communities in situ (Alivisatos
et al., 2015). Biofilm community of microbial electrolysis cell pro-
vides an ideal ecological model to study the relationship between
diversity and stability in response to environmental disturbances.
MECs are generally operated at room temperature and under mild
TABLE 2 Correlations between recovery time of MECs in response to pH disturbance and a-diversity of anodic microbial community forthree periods
Correlation methods
Richness Shannon Inverse Simpson Chao1 PD
Correlation p Correlation p Correlation p Correlation p Correlation p
conditions. The operation is relatively easy, requiring only a change
of the feeding substrate when used in a batch mode. Additionally,
the species composition of anodic communities was of only moder-
ate complex as compared to the initial activated sludge inoculum
(Fig. S7). More importantly, the change of operation conditions might
influence the composition of microbial communities (Fig. S7), and
such an alteration could reversely affect the bioreactor performance
(Figure 1), which bridges microscope biofilms and macroscopic reac-
tor performance together. As observed in the recovery time compar-
ison (Fig. S3), the biofilms with higher diversity showed better
resilience than low diversity ones (Figure 2), supporting the general
perspective of stability and biodiversity relationship. Therefore, as
we expected, biofilms of MECs were good systems for testing micro-
bial ecology theories.
Although pH disturbance decreased the reactor performance,
MECs still exhibited relatively high resilience that could take certain
time to rebound back to the comparably stable state (Figure 1).
Interestingly, comparing the performance among three groups of
MECs, less difference was observed after the recovery than before
the disturbance (Figure 1). This is most likely associated with the
changes of biodiversity of biofilm communities and as designed, with
Group A having significantly higher biodiversity as compared to
Group B and Group C (Figure 2). This was also confirmed by the
composition of microbial community among the three groups show-
ing a convergent trend after the pH disturbance (Figure 3, Fig. S7),
indicating the distance between each two groups of microbial com-
position decreased due to pH disturbance. A conceptual scheme was
summarized to better comprehend such observations in this study
(Figure 5). This diversity, or community composition convergence,
may be observed when the community gains or loses similar species,
especially rare species, when facing various disturbances (Avolio
et al., 2015). In a low-productivity grassland in Michigan, the species
composition was analysed in response to a one-time removal of
grassland as disturbance following solarization and herbicide applica-
tion. The community dispersion was decreased, showing a conver-
gence in a grassland ecosystem following such a disturbance
(Houseman, Mittelbach, Reynolds, & Gross, 2008). This was consis-
tent with the disturbance effect on tallgrass responses to grazing
and frequent fire, which resulted in decreased community dispersion
(Collins & Smith, 2006). Additionally, the fire could increase remark-
able similarity of dominant shrubs and subshrub density (Keeley,
Fotheringham, & Baer-Keeley, 2005), and invasive grass litter addi-
tions also changed the composition of scrubland arthropod commu-
nities with decreased community variance (Wolkovich, 2010).
Moreover, trophic forests exhibited decreased group distance over
time after cyclone disturbance (Murphy, Metcalfe, Bradford, & Ford,
2014) and a more similar or homogenized community of hoverfly
and carabid beetles was increased within agricultural intensification
(Gagic et al., 2014). In activated sludge bioreactors, an increased sim-
ilarity of bioreactor communities relative to influent was observed in
response to decreased solid retention time (SRT), which might be
due to the washing effect on established species of lower SRT
(Vuono, Munakata-Marr, Spear, & Drewes, 2016). The convergence
trend of microbial communities to steady composition was simulated
in a nitrifying bioreactor after inoculating competitive species (Louca
& Doebeli, 2016). Therefore, the ecological pattern of community
convergence has been observed widely in ecosystems in response to
disturbance, suggesting a wide application of our conceptual model.
F IGURE 4 Networks to visualize interactions between Geobacterand Methanobrevibacter in three groups of biofilm microbialcommunities in response to pH disturbance. (a), (b), (c) arecorresponding to interactions within Group A, B, C respectively. Allnodes of Geobacter (red nodes) and Methanobrevibacter (purplenodes) and their interactions were remained. Node sizes areproportional to their own node degrees (connected number withother nodes). Blue links stand for positive interactions betweennodes and red links stand for negative interactions. (d) Briefsummary of nodes and links numbers is shown in histogram forthree groups of subnetworks
6178 | FENG ET AL.
However, the biodiversity convergence was seen less in biofilm com-
munity responses to environmental disturbances, thus what we per-
formed here might complement and spread such an ecological
pattern in microbial ecosystems.
The functional stability of microbial communities for MECs could
be a result of multispecies cooperation, containing both functional
exoelectrogens and nonexoelectrogens, such as fermentative bacteria
(Fig. S7b). The functional groups in anodic biofilms were affected to a
lesser degree by pH shock, implying that they could play a positive role
in the establishment of stable microbial communities (Liu et al., 2012).
The resilient performance of biofilms might be correlated with resilient
composition of the microbial communities for each group (Fig. S7b).
The Geobacter, as a group of well-known functional microbes in MECs
related to electron delivery (Logan, 2009), were dominant and could
contribute to the functional stability and resilience in biofilm commu-
nity. Other exoelectrogens including Pseudomonas and Arcobacter (Lu
et al., 2012) showed similar trends across groups and might assist the
resilience of biofilms in response to pH disturbance. Fermentative
groups including Geothrix, Oscillospira, Sedimentibacter, Anaerovorax
and Paludibacter could contribute to the utilization of carbohydrates
and produce organic acids (Lu et al., 2012). These fermentative species
may potentially be involved in hydrogen production, but it is not yet
clear (Varrone et al., 2014). Different species occupy specific niches
within biofilms, possibly influencing reactor performance (Logan &
Regan, 2006; Varrone et al., 2014). More importantly, not only the
above mentioned fermentative groups, but also other rare genera with
low relative abundance including Acetobacterium, Bacteroides and
Proteiniclasticum are involved in positive interactions with the domi-
nant functional genus Geobacter (Fig. S9). The resilience of the func-
tional genera could be a result of multispecies cooperations, involving
self-recovery and other species’ facilitation, thus performing stable
functions in response to environmental disturbance. These positive
interactions might contribute to the hypothesis that nonexoelectro-
gens (such as fermentative bacteria) played essential roles in the func-
tional performance of biofilm community in this specific bioreactor.
The cooperation of these complex species within biofilms in MECs
might contribute to the stable biofilms in response to pH disturbance
and the more specifically interspecies interactions required further
analysis.
Obviously, the three groups of MECs exhibited different recov-
ery time in response to pH disturbance (Figure 5, Fig. S3). Biofilms
with higher diversity took less time for recovery, while the reactors
with lower diversity took a longer time to reach stable performance
after pH disturbance. Moreover, the recovery time was significantly
correlated to a-diversity of anodic microbial communities (Table 2,
p < .05), providing potential evidence that biofilm community stabil-
ity could be maintained in response to environmental disturbance by
high biodiversity inoculations. The composition of anodic microbial
communities might be good to explain this phenomenon (Fig. S7).
Methanobrevibacter as unexpected and detrimental micro-organisms
for hydrogen production in MECs (Lee, Torres, Parameswaran, &
Rittmann, 2009) was a predominant genus in this study following
with Geobacter. The relative abundance of Methanobrevibacter varied
from Group A to Group C at end of recovery, which might correlate
with the different recovery times observed across the three groups.
Thereafter, the correlation tests showed that the high component of
Methanobrevibacter may result in the delayed recovery in Group C
(Table S4, p < .05). However, neither the ecological analysis nor the
composition of microbial community analysis could demonstrate the
intrinsic interactions between the diversity and resilience of biofilms
in response to pH disturbance, which motivated us for network anal-
ysis of those biofilm communities.
Network analysis could provide another view of microbial interac-
tions and ecological rules for community assembly (Barberan, Bates,
Casamayor, & Fierer, 2012). It could be used to reveal the intrinsic
mechanisms of microbial interactions in response to environmental
perturbations as well (Bissett et al., 2013; Deng et al., 2016; Hunt &
Ward, 2015). In a constructed network, the positive interaction is
mostly resulted from mutualism or commensalism, while negative
interaction is due to competition, predation, amensalism and so on
(Faust & Raes, 2012). But multispecies interactions might also be dri-
ven indirectly by external environmental stresses (Deng et al., 2016).
The negative interaction in microbial systems was mainly regarded as
competition (Deng et al., 2016; Faust & Raes, 2012) and in this study,
microbial interactions especially negative ones exerted significant
effects on biofilm community recovery. Our results indicated that
although the composition of microbial communities among three
groups of MECs tended to be similar (Figure 3, Fig. S7), the overview
of the three networks still showed different patterns from a holistic
view of microbial communities (Fig. S8). The group of Euryarchaeota
(mainly Methanobrevibacter) showed fewer interactions with other
taxa in Group A and more negative interactions to Proteobacteria
(mainly Geobacter) in Group C, which indicated the competitive rela-
tionship might be reinforced in lower diversity communities. This pat-
tern was more clearly observed by selecting the two major genera in
each group associated with their own interactions (Figure 4). Geobac-
ter was always present at vital position in each network, which was
consistent with the composition analysis (Fig. S7b), while Methanobre-
vibacter gradually grew from a marginal position in Group A to the
F IGURE 5 Schematic of conceptual visions of resilience relatedto reactor performance and biodiversity of biofilm communities inresponse to pH disturbance
FENG ET AL. | 6179
central position in Group C (Figure 4d). In addition, there were fewer
negative interspecies interactions between the two genera in Group A,
but with the increased presence ofMethanobrevibacter in Group C, the
negative interactions between Geobacter and Methanobrevibacter
increased and the competition might become fierce, therefore the
alteration of this negative interaction or competition might be a poten-
tial mechanism in biofilm communities in response to pH disturbance
associated with the diversity. As performed previously, the relative
abundance of Methanobrevibacter increased from Group A to Group C,
opposite the overall trend of microbial diversity (Figure 2, Fig. S7b),
which could reinforce the internal resource or spatial competition with
Geobacter (Figure 4), thus the functional recovery of biofilm communi-
ties in Group C was delayed by this competitive interaction between
the dominant genera in response to environmental disturbance. Simi-
larly, the competitive interactions of microbial populations were also
enhanced by emulsified vegetable oil injection and showed the impor-
tance of groundwater microbial community responses to environmen-
tal perturbations (Deng et al., 2016). Moreover, the environmental
alteration like pH disturbance might affect the microbial fitness or sur-
vival following with competitors in the formation of biofilms (Rendue-
les & Ghigo, 2015). Through this result, networks exhibited the
potential importance in individual species interactions to reveal intrin-
sic mechanisms and were confirmed as complementary analysis tools
for ecological theory exploration.
Overall, the resilience of biofilm community responses to envi-
ronmental disturbances might be highly correlated with a-diversity
and the composition of microbial communities coupling with micro-
bial interactions. The similar performance across three groups of
MECs might be due to the resilient exoelectrogens during the recov-
ery period. Moreover, the presence or the high abundance of
Methanobrevibacter showed negative effects on resilience of biofilms
as revealed by network analysis. The community with high diversity
might lessen the competitive effect of Methanobrevibacter for sub-
strate resources or spatial positions. Thus, this study implies that
improving the biodiversity may enhance the resilience of microbial
biofilm community especially in response to environmental distur-
bances, providing theoretical and experimental evidence for biofilm
assembly and colonization establishment in the future.
ACKNOWLEDGEMENTS
This project was supported by Key Research Program of Frontier
Sciences, CAS (QYZDB-SSW-DQC026), National Water Pollution
Control and Treatment Science and Technology Major Project
(2015ZX07206), Strategic Priority Research Program of the Chinese
Academy of Sciences (XDB15010302) and CAS 100 talent program.
The author is very grateful to Dr. James Walter Voordeckers for
careful edition on the final version.
CONFLICT OF INTEREST
The authors declare no conflict of interest.
DATA ACCESSIBILITY
Raw sequence data were submitted to NCBI with the Sequence
Read Archive (SRA) Accession no. SRP091356.
AUTHOR CONTRIBUTION
Y.D. and K.F. designed the experiments. K.F., Z.Z., W.C., W.L. H.Y.
and A.W. performed the experiment. K.F., Y.D., M.X. and Z.H. con-
tributed to the data analysis. K.F., Y.D. and Z.H. wrote the manu-
script. All authors read and approved the manuscript.