Process Performance and Bacterial Community Structure Under Increasing Influent Disturbances in a Membrane-Aerated Biofilm Reactor

Abstract

The membrane-aerated biofilm reactor (MABR) is a promising municipal wastewater treatment process. In this study, two cross-flow MABRs were constructed to explore the carbon and nitrogen removal performance and bacterial succession, along with changes of influent loading shock comprising flow velocity, COD, and NH₄-N concentrations. Redundancy analysis revealed that the function of high flow velocity was mainly embodied in facilitating contaminants diffusion and biosorption rather than the success of overall bacterial populations (p > 0.05). In contrast, the influent NH₄-N concentration contributed most to the variance of reactor efficiency and community structure (p < 0.05). Pyrosequencing results showed that Anaerolineae, and Beta- and Alphaproteobacteria were the dominant groups in biofilms for COD and NH₄-N removal. Among the identified genera, Nitrosomonas and Nitrospira were the main nitrifiers, and Hyphomicrobium, Hydrogenophaga, and Rhodobacter were the key denitrifiers. Meanwhile, principal component analysis indicated that bacterial shift in MABR was probably the combination of stochastic and deterministic processes.

Introduction

The feasibility of applying the membrane-aerated biofilm reactor (MABR) in the treatment of municipal and industrial wastewaters has been demonstrated in numerous research studies [18,29,37]. This novel system is operated by the use of a microporous membrane allowing air to penetrate through it, resulting in production of small air bubbles, thereby significantly increasing the oxygen transfer efficiency. Consequently, oxygen and pollutants such as organic compounds, nitrogen, and phosphorus transfer from the opposite sides of biofilms, leading to fundamental stratification in the biofilm if given appropriate process conditions. For example, the nitrifying bacteria that tend to grow in oxygen-rich environments might inhabit the aerobic region where the oxygen concentration is high and the organic matter concentration is low (i.e., aerobic zone; Fig. 1); in contrast, the heterotrophic and anaerobic microorganisms might survive in the oxygen-poor and pollutes-rich region (i.e., anaerobic zone; Fig. 1). Here, the fibers serve both as a supplier of air or oxygen and a carrier of biofilms. As a consequence, this unique microbial structure in the MABR is significantly distinct from those grown on conventional, inert surfaces where both oxygen and contaminants are transferred into the biofilm in the same direction [14,28,48,53]. Hereby, this innovative and intriguing system combines several advantages, including low biomass yields, high oxygen utilization efficiency [6], long solid retention time, simultaneous nitrification and denitrification [8,20,21,26,47,50,51], fewer odors production [5], and lower economical and environmental footprints.

Fig. 1.Schematic of bench-scale MABR.

It is widely realized that an effective mass transfer process plays a crucial role in the increase of COD and nitrogen removal for the MABR. Appropriate hydrodynamic conditions, such as fluid shear, can be substantially conducive to the performance of an MABR by decreasing the biofilm-liquid boundary layer thickness and increasing substrate flux [9]. Furthermore, higher velocities also effectively exert significant influence on biofilm thickness and density, and EPS production and accumulation [10] as well as population competition between nitrifiers and denitrifiers [13], and considerably strengthen the resistance of the loading shock (e.g., high influent COD or NH4-N concentration) of biofilms. However, these and similar studies generally focus on the relation between flow conditions and physical and chemical construction of biofilms. We also have reason to assume that the microbial assembly in biofilms can be easily subjected to hydraulic conditions and the performance is remarkably dependent on the microbial community composition, and it has been observed in other water treatment systems. For example, significant changes in bacterial community structure have been detected under distinct hydraulic regimes in an experimental drinking water distribution system [16]. Hence, a comprehensive understanding of the structure and succession of the bacterial community is essential to optimize and control the biofilm structure of MABRs for domestic wastewater treatment.

To date, high-throughput sequencing technology has been widely adopted to detect and acquire comprehensive information about the microbial community in various wastewater treatment systems [3,25,55]. Owing to its higher resolution in discerning minor groups in organisms compared with conventional molecular fingerprinting methods, we chose to use 454-pyrosequencing technology to investigate the bacterial community inhabiting MABR biofilms. In this study, two identical bench-scale reactors were designed and set up to explore the process performance and bacterial structure in MABRs under gradient cross-flow velocities combined with increasing influent contaminants loading. The aim of the present research was to evaluate the influence of hydraulic regimes and influent contaminant concentrations on the performance and bacterial communities of biofilms in MABRs.

 

Materials and Methods

Reactor Configuration and Operation

This study involved two bench-scale MABRs with volumes of 2.5 L each. Fig. 1 shows the outline of each apparatus. The hollow-fiber membranes module (Hydroking Sci.&Tech. Ltd., Tianjin, China) was installed and submerged in a rectangular and well-designed perspex container (9 cm wide, 10 cm high, 30 cm long). The membrane module was composed of 245 hydrophobic polyvinylidene fluoride membranes in a parallel configuration with the wall thickness of 150 μm, outer diameter of 400 μm, and active length of 1 m, yielding a specific surface area of 240 m2/m3, and these fibers were entwined through the rectangular channel. Two water distributors on both sides of the reactor were designed to generate the uniform distribution of bulk flow, and the flow direction was almost perpendicular to the membrane. A circulation pump was used to control the cross-flow across the membranes, and three different hydraulic regimes were 0 m/sec, 0.004-0.006 m/sec, and 0.05-0.08 m/sec, which were calculated by dividing the inlet feed flux by the intralumen cross-sectional area of the rectangular reactor. Air was provided from one end of the module and exhausted from the other end of the membrane to prevent accumulation of water condensate. The seeding sludge, with the concentration of about 6,400 mg/l was derived from a MBR device of Tianjin University (Tianjin, China). The start-up period lasted for 50 days before running. Reactor #1 was used as the control group and performed around 30 days under consistent conditions; Reactor #2 was operated over 160 days under varying velocities and influent concentrations. The detailed working operation is presented in Table 1. As for Reactor #2, seven phases were designed along with the changes of influents. Among them, Ph.1 was processed under static flow condition, followed by the second stage (Ph.2) with low flow disturbance shock. After that, the reactor was subjected to a small increase of COD concentration in Ph.3, followed by the increasing influent NH4-N (Ph.4) with unchanged flow velocity. Ph.3 and Ph.4 were designed to study the shock resistance of the reactor against a small increase of influent loading under the low flow rate. Then, a large increase of flow velocity was performed in Ph.5, Ph.6, and Ph.7; the latter two stages were designed to investigate the shock resistance ability for high influent COD and NH4-N loadings under higher flow turbulence, respectively. The applied synthetic wastewater (Table S1) was made referring to the typical compositions of COD and N in domestic wastewater [28,39,40,43,50]. The batch experiment was conducted with a 24 h hydraulic retention time (HRT) at room temperature (20 ± 2℃) and the pH was maintained between 7.6 and 8.1.

Table 1.Operational conditions for the MABR systems during the whole processing time.

Sampling, DNA Extraction, and PCR

Biomass samples were carefully removed with a razor from three sites (Fig. 1) of the membrane at the end of each stage, according to Table 1. In total, 11 samples (including the seeding sludge) were collected and immediately stored at -80℃ before further analysis. Around 0.5 g of the samples was used for genomic DNA extraction using the FastDNA Spin Kit for Soil (MP Biomedicals, Cleveland, USA) following the manufacturer’s instructions. The quantity and quality of the extracted DNA were assessed using a UV-1700 PharmaSpec UV-VIS spectrophotometer (Shimadzu, Japan).

The V3-V6 region of bacterial 16S rRNA gene fragments were amplified using the forward primer 341F (5’-CCTACGGGAGGCAGCAG) and reverse primer 1073R (5’-ACGAGCTGACGACARCCATG). Each forward primer contained a unique 10 nucleotide barcode used for distinguishing each sample. The PCR amplification mix (50 μl) included 40-50 ng of template DNA, 1 unit of FastPfu Polymerase, 10 μl of 5× FastPfu buffer, 5 μl of 2.5 μM dNTPs, 6.25 pmol of each primer, and 29 μl of ddH2O. The PCR conditions were as follows: an initial denaturation at 95℃ for 2 min, followed by 24 cycles of 95℃ for 30 sec, 56℃ for 30 sec, 72℃ for 30 sec, and a final elongation at 72℃ for 5 min. The triplicate PCR products were pooled and purified using the Gel/PCR Extraction Kit (Bioteke Corporation, Beijing, China).

Pyrosequencing and Data Processing

A total of 11 prepared PCR amplicons (Table 2) were mixed together and sent to the TEDA Institute of Biological Sciences and Biotechnology (Tianjin, China) for pyrosequencing on a massively parallel 454 GS-FLX sequencer, according to standard protocols [31].

The sequences were pre-processed by the TEDA Institute of Biological Sciences and Biotechnology. The raw sequences were processed with QIIME ver. 1.7.0 to obtain the optimized sequences, which removed bad sequences from raw sequences (length outside bounds of 200 and 1,000; number of ambiguous bases exceeds limit of 6; missing quality score; mean quality score below minimum of 25; maximum homopolymer run exceeds limit of 6; number of mismatches in primer exceeds limit of 0 and uncorrected barcodes). All of the optimized sequences with an average length of 733 bp were compared by performing a BLAST search via the SILVA database (ver. 106) and then clustered into operational taxonomic units (OTU). The OTU was defined as the furthest neighbor Jukes-Cantor distance of 0.03 (OTU0.03) and assigned to a taxonomy using the Ribosomal Database Project Naive Bayes classifier. For each rank assignment, the classifier automatically estimated the classification reliability using bootstrapping. Ranks where representative sequences of different OTU0.03s could not be assigned with a bootstrap confidence were displayed under an artificial “unclassified” taxon. The Shannon-Wiener diversity index and Chao1 richness estimator were generated in the QIIME program. The nucleotide sequences were deposited in GenBank under the accession number DRA003696.

Physicochemical Measurements

COD, NH4-N, NO2-N, NO3-N, TN, DO, and pH were determined according to the methods described previously [55]. Biofilm thickness was calculated according to the method described by Celmer et al. [9]. Biofilm volumetric density was obtained by the method described by Ganczarczyk and Zohid [17]. Each measurement was performed in triplicates.

Statistical Analysis

The drawing of heat maps, principal component analysis (PCA), redundancy analysis (RDA), Spearman correlation analyses between OTUs, Bray-Curtis distance, and Monte Carlo permutation test were performed using R 3.1.2 statistical computing (http://www.r-project.org/). The water-quality trend chart was made using the Origin 8.0 package.

 

Results

Performance of MABRs in Response to Different Operational Strategies

Over the course of the 160-day operational period, the performance of Reactor #2 varied in response to the different operational parameters (Fig. 2). During the first stage (Ph.1), when the bulk flow was kept in a motionless state, the average COD, NH4-N, and TN removal efficiencies were 91.6%, 21.4%, and 19.4%, respectively, and the effluent DO was 0 mg/l. In contrast, when the flow velocity was adjusted to 0.004-0.006 m/sec (Ph.2), the average COD, NH4-N, and TN removal efficiencies increased to 97.2%, 87.9% and 78.0%, respectively, and the effluent DO remarkably reached approximately 3.4 mg/l. During the third stage, when the average influent COD was increased to 200 mg/l, the NH4-N and TN removal efficiencies increased to 91.2% and 84.3%, respectively, while the COD removal rate (95.0%) showed little change. The next stage (Ph.4) was processed with a higher level of the influent NH4-N concentration (40 mg/l). The reactors exhibited more efficient functioning of pollutant removal (COD: 97.8%; NH4-N: 96.1%; TN: 91.8%). In order to evaluate the shift of the bacterial community between varied fluid shears, the flow velocity was then sharply increased to 0.05-0.08 m/sec (Ph.5). The NH4-N removal rate showed minor increment (98.4%), but the TN removal rate declined to 84.9%. Meanwhile, the COD removal rate remained about the same, and the effluent NO3-N and NO2-N concentrations increased significantly in the steady state. To increase TN removal and mitigate the impact of the high flow velocity, the average influent COD concentration was doubled to 400 mg/l in the 6th stage (Ph.6). The reactor was processed with the highest COD (98.8%), NH4-N (99.4%), and TN (95.9%) removal rates. Then the reactor was subjected to a higher NH4-N concentration shock loading (Ph.7). After large fluctuations in the initial days, the COD, NH4-N, and TN removal rates decreased gradually to 88.5%, 52.7%, and 46.7%, respectively. Apparently, the performance of the MABR was more easily influenced by the influent NH4-N concentration rather than influent COD. In general, the effluent DO had an upward trend from Ph.1 to Ph.6 along with the increasing flow velocity, suggesting that a high flow velocity might promote the degradation and removal of COD, NH4-N, and TN. As for Reactor #1, excellent and stable performance was obtained throughout the 30-day process time, with 99.5%, 96.5%, and 90.3% of COD, NH4-N, and TN removal rates, respectively.

Fig. 2.Variations in pollutants and DO at different operation conditions. (A) Influent and effluent COD concentrations and COD removal rate; (B) Influent and effluent NH4-N concentrations and NH4-N removal rate; (C) Influent and effluent TN concentrations and TN removal rate; (D) Effluent NO2-N and NO3-N concentrations; and (E) Effluent DO concentration.

Changes in the Bacterial Community in Response to Different Operational Strategies

Forty-one phyla were retrieved from 20,830 qualified sequences (Reactor #2, Fig. 3A). The Chloroflexi phylum accounted for 5.5%, although the Proteobacteria (50.0%) constituted the most abundant bacteria in seeding sludge. In biofilms, Proteobacteria (35.3% in Ph.6 to 56.1% in Ph.7) and Chloroflexi (16.3% in Ph.7 to 35.6% in Ph.4) were the most represented. Although in the first four stages Proteobacteria remained an average 38.7%, its relative abundance increased to 47.3% on average from Ph.5 to Ph.7. In contrast, Chloroflexi remained at around 26.1% in the earlier stages (Ph.1 to Ph.3) and then increased to a peak value (35.6%) in Ph.4. After that, a remarkable decline occurred in Ph.5 (19.2%) and decreased to the lowest abundance in Ph.7 (16.3%). From the identified 40 phyla, more than 98 classes were retrieved (Reactor #2), of which Anaerolineae dominated the bacterial community in biofilms, comprising 18.0% (Ph.5) to 32.7% (Ph.4), followed by Betaproteobacteria (11.2%–19.3%), Alphaproteobacteria (8.9%–16.9%), Gammaproteobacteria (5.2%–18.5%), and Deltaproteobacteria (2.5%–6.9%) (Fig. 3B). Among the 40 known classes, a total of 118 genera were identified (Reactor #2), although an average of 62.3% of the analyzed sequences could not aligned to known genera in biofilms (Fig. 3C). Despite that, some known genera were still detected with high abundance. For example, the denitrifying bacteria such as Nitrospira and Nitrosomonas averaged 1.9% and 2.7% with major fluctuation, respectively; Hyphomicrobium showed an average relative abundance of 2.7% with minor fluctuation; Rhodobacter and Hydrogenophaga were all absent in the sludge, and showed the highest relative abundance in Ph.7 (1.3% and 2.1%, respectively); some bacteria were predominantly represented in a special stage, such as Longilinea and Dechloromonas (1.4% and 1.1%, respectively) in Ph.1, Hydrogenophaga in Ph.1 (1.3%) and Ph.7 (2.1%), and Desulfococcus in Ph.2 (1.2%). Some bacteria belonging to S47 (candidate family) dominated the community in biofilms throughout the whole process (average 16.4%).

Fig. 3.Bacterial community distributions of eight biosamples under distinctive conditions at different taxonomic levels. (A) Phyla; (B) classes; and (C) genera.

Heat maps (Fig. 4) based on the OTU0.03 level were generated in order to explore the shift of bacteria along with the changes in operation in detail (Reactor #2). As illustrated in Fig. 4A, significant bacterial shift was observed as the flow velocity and influents varied, allowing them to be roughly divided and clustered into eight groups. Group-I included bacteria mainly detected in the original sludge, such as OTU331, OTU490, and OTU2179, which were mainly affiliated to Cytophagaceae, Syntrophobacteraceae, Sinobacteraceae. Group-II included bacteria that dominated the biofilm in Ph.1, such as OTU476, OTU1402, and OTU1442, mainly belonging to Ignavibacteriaceae, Gemmataceae, and Parachlamydiaceae. Group-III comprised bacteria that predominantly inhabited in Ph.7, mainly belonging to Acinetobacter, Flavobacterium, Nitrosomonas, and Pseudomonas, including primarily OTU2094, OTU2268, OTU591, and OTU970. Group-IV included those bacteria that had the highest relative abundance in Ph.5, such as OTU2085 and OTU1625, belonging to Rhodospirillaceae and Nitrospira, respectively. Group-V showed peak abundances in Ph.4 or Ph.6, mainly belonging to Sinobacteraceae, and Rhodobacter, such as OTU536 and OTU1535. Group-VI included the bacteria that mainly inhabited the biofilms in Ph.2, Ph.4, and Ph.6, such as OTU374, OTU840, and OTU1992, which were affiliated to Alcaligenaceae and Anaerolineae. Group-VII consisted of such bacteria that were hardly detected in both Ph.5 and Ph.7, including OTU1114 (Hyphomicrobium) and OTU104 (A4b-candidate family). Group-VIII included bacteria such as OTU374 and OTU1217 that were affiliated to Alcaligenaceae and Comamonadaceae, respectively, and were mainly distributed in Ph.3, Ph.5, and Ph.6.

Fig. 4.Heat-map plots drawn at the OTU level. (A) Community structural comparison and shift along with operation conditions. OTUs were present over 1% in at least one of the sequenced samples and were subjected to standard transformation. Cluster results were obtained by euclidean distance. (B) Bray-Curtis similarities between samples. (C) Spearman analysis between OTUs.

As indicated in Fig. 4B, Ph.2 and Ph.3 were generally grouped together, and so were Ph.4, Ph.6, and Ph.5. Additionally, the organisms in S.S were distinctly different from those in biofilms; among all biofilm samples, in particular, bacteria in Ph.7 and Ph.1 showed a large difference from those in other samples. As indicated in Fig. 4C, bacteria in each group mentioned above exhibited strongly positive correlations to each other. Meanwhile, there were good relationships among the eight groups. For example, the bacteria in Group-I had negative correlation to those in Group-IV; conversely, the bacteria in Group-IV had positive correlation to those in Group-VI.

Relationships Between Community Compositions and Process Parameters as well as Performance

The results of RDA for bacteria in biofilms at the class level (Fig. 5a) showed that the principal component 1 (PC1) and principal component 2 (PC2) explained 73% of the variance in overall bacterial community. Gammaproteobacteria and Betaproteobacteria had positive correlation with influent COD and NH4-N. In contrast, bacteria such as Deltaproteobacteria, Nitrospira, Chloracidobacteria, and VHS-B5-50 (candidate class) were negatively correlated with influent COD and NH4-N. Acidobacteria and Alphaproteobacteria had positive correlation with the flow velocity. Instead, Planctomycetia and Anaerolineae had negative correlation with the flow velocity. For carbon and nitrogen removal, Deltaproteobacteria, Nitrospira, Chloracidobacteria, and VHS-B5-50 were positively correlated to NH4-N and COD removal rates.

Fig. 5.Ordination plots generated by RDA based at the class level.

At the OTU level (Fig. S4), PC1 and PC2 represented 34.6% and 31.4% variance of the bacteria in biofilms, respectively. Bacteria (ellipse b) such as OTU982, OTU2094, and OTU960 had positive correlation with influent NH4-N concentration. In addition, OTU1799, OTU2085, and OTU960 (ellipse a) were positively correlated with flow velocity and influent COD concentration. In contrast, OTUs (ellipse d) such as OTU1114, OTU1485, and OTU2815 were mainly negatively correlated with influent NH4-N; OTUs (ellipse c) such as OTU935, OTU1402, and OTU1826 showed negative correlation with flow velocity and influent COD. Moreover, some bacteria had strong correlation with the removal performance. For example, bacteria (ellipse c) such as OTU1217, OTU536, and OTU840 exhibited positive correlation with NH4-N and TN removal rate. Meanwhile, bacteria (ellipse d) such as OTU2815, OTU1485, and OTU1114 showed negative correlation with COD removal rate. Moreover, it was not difficult to find that flow velocity had positive correlation with nitrogen removal.

 

Discussion

To date, limited research studies have investigated the microbiota inhabiting the biofilms in MABRs. Our study is the first to elucidate the bacterial succession and the correlations between the community composition and operational strategy as well as process efficiency along with varied flow rates and influent loadings.

At the phylum and class levels, the detection of dominant Proteobacteria (mainly in the classes Beta-, Alpha-, and Gammaproteobacteria) was unsurprising, since this phylum has been frequently detected as being predominant in various wastewater treatment systems [11,15,19]. In contrast, the large abundance of Chloroflexi (mainly in the class Anaerolineae) in biofilms showed that the biofilm conditions supplied by MABRs were more suitable for these filamentous microorganisms than that by MBRs. Additionally, these anoxygenic phototrophic and facultative anaerobic bacteria were widespread in various wastewater treatment systems and often associated with organic matter degradation [4,22,57]. At the genus level, an average of 50.8% of all the analyzed sequences were not aligned to known genera, suggesting that a wide variety of uncharacterized species exist in biofilms [30,46,54]. Among the known genera, nitrification-related species such as Nitrosomonas [56] and Nitrospira [36] and denitrification-related species such as Hydrogenophaga [7,12], Hyphomicrobium [33], and Rhodobacter [24,35,44] were the main functional bacteria in biofilms for nitrogen removal.

A significant shift of bacteria was observed along with the changes in process parameters. At the class level, influent NH4-N concentration (p = 0.144), flow rate (p = 0.289), and influent COD concentration (p = 0.358) did not exert significant influences on overall bacterial structure, based on the Monte Carlo permutation tests. At the OTU level, influent NH4-N was the most significant factor (p = 0.03*) responsible for the overall shift in bacterial community. RDA results also showed that influent NH4-N was the most important factor, followed by flow rate and influent COD. This point could also well explain why groups Ph.1, Ph.2, and Ph.3 and groups Ph.4, Ph.5, and Ph.6 generally clustered together, respectively (Figs. 4A and 4B). In other words, the same influent NH4-N seemed to contribute to similar a bacteria community for the MABR. Additionally, RDA revealed that bacteria belonging to Proteobacteria were easily subjected to the changes of external conditions. For example, influent NH4-N favored the growth of Gammaproteobacteria and Betaproteobacteria but suppressed Deltaproteobacteria. At the OTU level, some bacteria also had significant correlations with influent NH4-N. For example, influent NH4-N could increase the relative abundance of OTU591 belonging to genus Nitrosomonas, which was associated with nitrification of ammonia to nitrite in varied wastewater treatment systems [41,45]. In contrast, another nitrobacterium named Nitrospira (represented by OTU1625) seemed hardly affected by influent NH4-N. Although the flow rate was not the main external factor (p > 0.05) for bacterial shift, it probably exerted its influence on bacteria through changing the substrate transfer rate in biofilms and physical characteristics of biofilms (i.e., thickness and density; Fig. S2), which therefore produced the differences in chemical and physical gradients distribution, and hence niche formation in the biofilm. For example, some bacteria still had good association with flow rate, such as Alphaproteobacteria, which tended to grow under high flow velocity. In contrast, the anaerobic Anaerolineae was easily suppressed by flow velocity.

In this study, rapid removal efficiency of COD, NH4-N, and TN under higher flow velocity was observed in the early stage of an HRT of 24 h (Fig. S1). However, as mentioned above, the flow rate did not exert significant influence on the whole community structure. Moreover, the bacteria belonging to Detaproteobacteria, Nitrospira, Chloracidobacteria, and VHS-B5-50, which had little direct association with flow rate and negative correlations with influent COD and NH4-N concentrations, exhibited the most contributions to the discrepancy of the removal rates of COD, NH4-N, and TN in different stages (Fig. 5). Nevertheless, this did not mean that the predominant populations such as Betaproteobacteria, Alphaproteobacteria or Anaerolineae did not have any contribution to the removals of carbon and nitrogen. On the contrary, they played the key role in wastewater treatment, such as activated sludge flocculation [52], denitrification [42], biosorption [2,49], or organics degradation [23,54]. Furthermore, denitrification still proceeded from the 3rd hour to the 9th hour, although the COD concentration in bulk had been declined to about 0 mg/l, and we did not detect bacteria with the ability of autotrophic denitrification through pyrosequencing. All these indicated the strong biosorption ability of biofilms for carbon under higher flow velocity, the great contribution of predominant populations for adsorption, and the possibly endogenous respiration as carbon resource for denitrification in the oligotrophic phase at the end of the HRT.

It seemed that the bacterial assemblages in the MABR were the combined effect of deterministic (influent conditions) and stochastic processes. Given this, a control trial (Fig. S3) was designed to examine whether changes of community composition would occurr under the identical influent conditions. The PCA plot (Fig. 6) showed the obvious variability of community compositions with temporal changes under constant influent conditions, which was consistent with previous findings suggesting that perpetual fluctuations of bacterial community structure are needed for stable performance of MBRs [34]. Thus, these results supported the neutral theories, which assume that species are ecologically equivalent and that community structure is determined by random processes [1,32]. However, it is undeniable that the increasing flow rate and shock loadings of contaminants had important roles in bacterial assembly. It is probably fair to assume that the particular species inhabited and occupied a particular region in biofilms generated and were determined by influent disturbances (niche driven) and then determined by stochastic recruitment (neutrally driven) [27,38]. Hence, more efforts are needed to elucidate the trade-offs between both factors for bacterial community succession in biofilms in the MABR system.

Fig. 6.Principal component analysis based at the OTU level, showing differences in the bacterial community among 11 biosamples. Symbols representing individual samples are colored on the basis of sample type. Blue ball: seeding sludge (S.S); green ball: biofilms in Reactor #1 (control); red ball: biofilms in Reactor #2.