Introduction

Along with the fast development of wastewater treatment plants (WWTPs), a huge amount of sewage sludge, as the by-product, has been and is being produced globally. In China, the annual production of sewage sludge with 80% moisture content has reached practically 3000 t, of which 80% of this has not acquired essential stabilization (Duan et al. 2012). Anaerobic process has been proven an effective and sustainable technology for sludge digestion: not only does it not require aeration, but also it is capable of producing biogas energy source such as methane (Demirel and Scherer 2008; Yu et al. 2016a). It is thus widely used and promoted to be used in many WWTPs (Wei et al. 2003; Appels et al. 2008). However, traditional anaerobic treatment process generally needs to be operated under a long sludge retention time (SRT) since the growth rate of anaerobic microorganisms is relatively slow, which would result in a relatively larger reactor volume and footprint.

Anaerobic membrane bioreactors (AnMBRs) incorporate anaerobic biological and membrane process. Based on the membrane design and operation, the configurations of AnMBR can be divided into three categories: external cross-flow membrane, internal submerged membrane, and external submerged membrane (Liao et al. 2006). Having the ability to produce a similar quality of effluent to aerobic treatment, AnMBRs generate useful energy and produce substantially fewer residuals as well (Liu et al. 2016; Smith et al. 2012). Furthermore, not only does AnMBR maintain high sludge concentration in the system, but the reactor size decreases (Ho and Sung 2010). AnMBRs are therefore expected to be a promising process for excess sludge treatment. Nevertheless, membrane fouling continues to be a major operational concern for AnMBRs. The primary foulants such as suspended biomass, colloidal solids, soluble microbial products (SMP), extracellular polymeric substances (EPS), and inorganic precipitates are important contributors to the membrane fouling in AnMBR systems (Smith et al. 2012).

Polyaluminum chloride (PAC) is widely used for treating different kinds of water and improving the coagulation performance (McCurdy et al. 2004; Tang et al. 2017; Wu et al. 2011; Yan et al. 2008; Yang et al. 2010). Dosing coagulants can slow down and control the membrane fouling in aerobic membrane bioreactors (Hwang et al. 2007; Song et al. 2008; Wu et al. 2006a). However, little has been done to control membrane fouling in AnMBRs using PAC. According to our previous research, PAC has positively affected membrane fouling by reducing SMP and improving filterability when AnMBRs were operated for sludge digestion (Yu et al. 2015). Since the anaerobic bioreactors are sensitive to environmental conditions and the performance of anaerobic digestion depends on the microbial community, it is crucial to investigate the microbial community in AnMBRs when dosing PAC.

Membrane bioreactor harbored distinct bacterial community structure due to the complete rejection of biomass by the membrane and the contrasting operating conditions applied (Wan et al. 2011). Previous studies found that there were significant differences in microbial diversity and richness among different SRTs (Duan et al. 2009), different operating temperatures, and cake ages in cake sludge (Gao et al. 2012). Lin et al. (2011) suggested that the microbial community of small flocs in bulk sludge was similar with that of the sludge cake layer, while the microbial community of large flocs showed rather significant differences, indicating that small flocs played a leading role in membrane fouling. Furthermore, a study also showed that there were significant differences in microbial community population density along the cake layer depth (Gao et al. 2011). Nonetheless, there has been limited research on the characteristics of microbial communities in AnMBRs digesting excess sludge in comparison with that in AnMBRs with and without PAC dosing.

In this study, two lab-scale external cross-flow AnMBRs treating excess activated sludge were operated. The principal objective of this study is, therefore, to characterize the microbial community in the digested sludge and cake layer of AnMBRs as well as the excess sludge utilizing polymerase chain reaction coupled with denaturing gradient gel electrophoresis (PCR-DGGE) and Illumina high-throughput sequencing. Bacteria at the phylum, class, and genus levels were analyzed and compared among different samples to identify the dominant colonization in AnMBRs. In addition, the influences of seed sludge and excess activated sludge (sludge influent) microbial community on the digested sludge microbial colonization were also evaluated. Beyond all that, to the best of our knowledge, this is the first time that the microbial characteristics of excess sludge, digested sludge, and cake layer were compared under the circumstances of dosing and not dosing PAC.

Materials and methods

The configuration, inoculation, and operation of AnMBRs

Two identical sets of lab-scale AnMBRs (AnMBR-A and AnMBR-B), which consisted of a completely mixed anaerobic digester (2.7 L) and an external hollow fiber membrane module, were established (Fig. S1). The membrane adopted polyvinylidene fluoride (PVDF) with a 0.3-μm nominal pore size (Origin Water, China). The temperature of the anaerobic bioreactor was maintained around 35 °C through water bathing.

The bioreactors were inoculated with anaerobic sludge (seed sludge) taken from the sludge digester of a WWTP in Beijing, China. The total solid (TS) concentration was approximately 30.16 g/L and the volatile solid (VS) concentration was around 13.22 g/L. The original sludge was sieved through a 1-mm screen in order to remove large particles before inoculated. After that, the anaerobic sludge was utilized for bioreactor start-up without any further acclimation.

With the intent of treating excess sludge in the AnMBR system, the excess sludge from an aerobic MBR was collected from another WWTP in Beijing, China. After settling, discarding the supernatant, and filtering through a 1-mm screen, the sludge was stored at 4 °C for further usage as the treating object. The AnMBR systems were continuously fed with excess sludge. The membrane unit was operated at a constant flux mode of 9-min filtration followed by 1-min relaxation. A cross-flow velocity of 1.0 m/s was applied. The VS loading rate of the bioreactor was ranging from 1.0 to 1.3 g VS L−1·day−1. In addition, the sludge retention time (SRT) and hydraulic retention time (HRT) were maintained at 40 and 3.4 days, respectively.

AnMBR-A was operated (run 1 to run 3, from day 0 to day 405) for our previous research (Yu et al. 2015), in which the operation of AnMBR-A and AnMBR-B for this study was based and continued upon run 3 in Yu et al. (2015). At the beginning of this study (day 207 in correspondence with our previous study), half of the anaerobic sludge from AnMBR-A was transferred into AnMBR-B, and the two systems were restarted. PAC was added into AnMBR-A at the beginning of operation and readded according to the amount of sludge sampling, maintaining at a constant concentration (500 mg/L) inside the reactor. AnMBR-B was used as a control system without any PAC addition. The membranes for this study were changed to new ones at the beginning (day 207), and the two systems have been operated until day 403.

AnMBR sample collection and DNA extraction

A total number of 31 digested sludge samples (16 from AnMBR-A and 15 from AnMBR-B) were collected from day 207 to day 374 during the operation (Table S1). Corresponding to the digested sludge samples from A10 and B9, the cake layers of AnMBR-A and AnMBR-B, marked as cake A and cake B, were respectively sampled on the 270th and 255th day of operation. Corresponding to the digested sludge samples of A11–A15 and B11–B15, original excess sludge samples were collected from the sludge storage on the 336th, 340th, 344th, 364th, and 374th day of system running time and labeled as E1, E2, E3, E4, and E5, respectively.

Approximately 8 mL digested sludge and excess sludge were collected and stored in 10-mL sterile tubes. At the end of each operating cycle, the fouling cake layer was carefully scraped off and transferred into the tubes as well. The abovementioned samples were preserved at − 80 °C until DNA extraction. DNA was extracted using PowerSoil® DNA Isolation Kit (MO BIO, USA) according to the protocol provided by the manufacturer. Monitoring the absorption ratios at 260/230 and 260/280 nm, a Nanodrop spectrophotometer (Thermo 2000c, USA) was then used to assess the quantity and quality of extracted DNA.

PCR-DGGE analysis

All the samples collected above were employed for DGGE analysis. Using the primer set, 341F-GC (5′-GC-clamp-CCTACGGGAGGCAGCAG-3′) and 534R (5′-ATTACCGCGGCTGCTGG-3′), the V3 region of bacterial 16S ribosomal DNA from previous DNA extraction was amplified (Muyzer et al. 1993). A touchdown thermal profile technique was used for PCR amplification (Bacosa et al. 2013). The thermocycling steps were performed as follows: an initial denaturing at 94 °C for 5 min, followed by 20 cycles consisting of 94 °C for 1 min, annealing for 1 min, and 72 °C for 2 min. For annealing, the temperature was decreased from 65 to 55 °C by an interval of 0.5 °C for each cycle. Additional 5 cycles were carried out at 55 °C for annealing. A final extension step was performed at 72 °C for 10 min.

Adopting the nested PCR (Boon et al. 2002), the primer set, 46F (5′-YTAAGCCATGCRAGT-3′) and 1017R (5′-GGCCATGCACCWCCTCTC-3′), was used to amplify the V3 region of archaeal 16S ribosomal DNA for the first round. The following steps were performed for the first-round nested PCR: an initial denaturing at 94 °C for 5 min, followed by 30 cycles consisting of 94 °C for 1 min, annealing for 1 min, and 72 °C for 1.5 min. For annealing, the temperature was decreased from 55 to 40 °C by an interval of 0.5 °C for each cycle. Additional 5 cycles were carried out at 40 °C for annealing. A 10-min extension was performed at 72 °C in the end.

After completing the first round, the PCR products were diluted 10-fold. Taking 1 μL of the dilution as the template, the second round was performed using the primer set GC-344F/522R (5′-GC-clamp-HGCAGCAGGCGCGA-3′/5′-HGCAGCAGGCGCGA-3′). For the second round, the steps were as follows: an initial denaturing at 94 °C for 5 min, followed by 20 cycles consisting of 94 °C for 1 min, annealing for 1 min, and 72 °C for 1 min. For annealing, the temperature was decreased from 62 to 52 °C by an interval of 0.5 °C for each cycle. Additional 5 cycles were carried out at 52 °C for annealing. Finally, a 10-min extension was performed at 72 °C. All the PCR products were examined by electrophoresis on 1.5% agarose gels stained with Goldview II.

Using the Bio-Rad Dcode™ system (Bio-Rad Laboratories Ltd., USA), DGGE was performed according to the manufacturer’s instructions and protocols. An 8% polyacrylamide gel containing a 30–60% linear denaturing gradient was employed for the electrophoresis of PCR products (100% denaturant contained 7 M urea and 40% (v/v) formamide). Gels were running at 60 °C and 130 V for 1 h in 1× TAE buffer. For visualizing on Gel Doc XR system (Bio-Rad Laboratories Ltd.), the gels were stained using GelRed and then imaged. Gel images were analyzed and the density of bands was obtained using Quantity One Software Version 4.6.2 (Bio-Rad Laboratories Ltd.).

High-throughput sequencing

Targeting the hypervariable V4 region of 16S ribosomal RNA genes, the extracted DNA of digested sludge, cake layer, and excess sludge samples was amplified with a primer set consisting of 515F (5′-GTGCCAGCMGCCGCGGTAA-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′). Amplicon sequencing was performed using Illumina MiSeq platform. Sequencing data was deposited in the NCBI Sequence Read Archive under the accession number SRP103643. The raw sequences were processed using the following principles: (1) remove the forward primer and barcode in the original sequences, (2) remove sequences shorter than 150 bp, (3) make sure that the quality score of the terminal bases is greater than 25, and (4) the quality score of 80% bases in one sequence should be greater than 20. The processed sequences were utilized for further analysis. The taxonomic classification of the processed sequences was performed using RDP Classifier (http://rdp.cme.msu.edu). A confidence threshold of 50% following the recommendation of RDP was adopted.

Statistical analysis

Using the band densities of PCR-DGGE obtained by Quantity One (Bio-Rad Laboratories Ltd.), clustering analysis was conducted and the dendrogram was constructed using PAST software (Version 3.16) (Hammer et al. 2001) based on unweighted pair-group method with arithmetic averages (UPGMA) and Bray-Curtis distance (Beals 1984; Bray and Curtis 1957). The obtained sequences sharing 97% or over similarity were grouped into the same operational taxonomic units (OTUs). Shannon-Weiner index (H) was calculated to evaluate the α-diversity of the microbial community. To compare the similarity (β-diversity) between two samples, Dice coefficient was calculated and utilized in the study. Detrended correspondence analysis (DCA) was performed based on the OTU composition of different samples. Additionally, canonical correspondence analysis (CCA) was conducted to discern the relationship between the OTU of samples and the environmental or operational parameters.

Analytical methods

The analysis of COD, NH4-N, and TP was determined by the Chinese NEPA standard methods (Chinese N.E.P.A. 2002). TS concentration was determined based on the standard method (APHA 1998). The pH and conductivity were measured with a pH meter (Orion 5-star, Thermo, USA) and a conductivity meter (Orion 5-star, Thermo, USA).

Results

System performance

The system performance of AnMBR-A and AnMBR-B during the operation is summarized in Table S2. A general increasing tendency was observed for TS, VS, COD, NH4-N, and TP concentrations in the two systems despite some small fluctuations for certain samples. The average VS/TS value for AnMBR-A (0.47) was slightly lower than that for AnMBR-B (0.48). With similar lowest concentration of NH4-N (195.42 mg/L for AnMBR-A and 185.41 mg/L for AnMBR-B), the highest NH4-N concentration in AnMBR-B was only accounted for 73% of that in AnMBR-A. However, the average values of COD and NH4-N concentration were rather similar between the two reactors. Moreover, the concentrations of TP varied from 4.89 to 31.73 mg/L in AnMBR-A, while it ranged from 6.64 to 68.25 mg/L in AnMBR-B; in other words, approximately 1-fold more TP concentration was accumulated in the latter one.

PCR-DGGE analysis of archaeal and bacterial community

PCR-DGGE profiles of digested sludge

During the operational period from day 207 to day 270, the PCR-DGGE profiles of digested sludge from AnMBR-A and AnMBR-B as well as the original seed sludge are shown in Fig. S2. Based on the band densities, clustering analysis was conducted and illustrated in Fig. S2. It was evident that the bacterial community in digested sludge was distinct from that in seed sludge (Figs. S2c and S2d). Represented by the intensity level of bands 9, 10, 12, 13, 14, and 15 from seed sludge, the microbial population of the corresponding species showed greater richness. Meanwhile, with stronger intensity of bands 1, 2, 4, 6, 7, 8, and 11 from digested sludge, significant variations occurred in the bacterial community during anaerobic operation.

The intensity level of bands 1 and 6 in AnMBR-A gradually decreased during the operation, namely descending abundance was observed, whereas band 8 emerged and was detected in a later period (Fig. S2a). Based on the clustering result of digested sludge from AnMBR-A, samples with close collecting days have shown greater similarity and were more likely to cluster together (Fig. S2c). AnMBR-B presented a similar situation in which the abundance of bands 1 and 7 declined whereas band 10 became more intense (Fig. S2b). Thus, a clear shift of bacterial community in both AnMBR-A and AnMBR-B has occurred. With the purpose of comparing the DGGE patterns between the two reactors, Dice index was calculated. The similarity index was 0.88 between A1 and B1. After a period of operation, nevertheless, it became merely 0.57 between A8 and B8. This was further confirmed by the clustering result among seed sludge and digested sludge from AnMBR-A and AnMBR-B (Fig. S2d). It was clear that A1 and B1 initially possessed higher similarity and thus clustered together. However, the nearby date samples in the digested sludge of AnMBR-B were more likely to present higher similarity. Unlike A1, B8 was far from other digested sludge samples from AnMBR-B, which was consistent with the Dice index calculated.

After 206 days of operation, the DGGE fingerprints and clustering results of the archaeal community from the original seed sludge and digested sludge from AnMBR-A and AnMBR-B are demonstrated in Fig. S3. Comparing the seed sludge with the digested sludge, it could be observed that the differentiation in archaeal community was not as prominent as the bacterial community from the PCR-DGGE profiles (Figs. S3a and S3b). Nevertheless, there still existed a distinction of archaeal community between the seed sludge and digested sludge, which could be seen from Figs. S3c and S3d that the similarity between the seed sludge and other digested sludge samples was the lowest. The dominating bands remained among bands 1–14 in both seed sludge and digested sludge during the operational period from day 342 to day 405; however, the intensity of bands 1–4 declined over operational time. The similarity (Dice index) of the archaeal community between A1 and B1 was 0.94, whereas it was 0.89 between A8 and B8, which indicated less variation between AnMBR-A and AnMBR-B that occurred in the archaeal community than in the bacterial community over time. This could also be demonstrated and confirmed in the clustering results of Figs. S2d and S3d, where the archaeal community of A8 and B8 showed greater similarity than the bacterial community of A8 and B8.

PCR-DGGE profiles of the cake layer

The cake layer of AnMBR-A (cake A) was sampled on day 270 during stable operation, when the corresponding digested sludge sample (A10) was collected as well. In addition, on day 255, the cake layer of AnMBR-B (cake B) and the corresponding digested sludge sample (B9) were acquired. The PCR-DGGE profiles of A10 and B9, as well as the corresponding cake layers, are presented in Fig. S4. The clustering results based on Bray-Curtis similarity are demonstrated in Fig. S4 as well. As indicated in Fig. S4c, the bacterial community between the digested sludge and cake layer within the same reactor presented higher similarity, namely A10 and cake A were clustered together; likewise, B9 and cake B were clustered together. Unlike the bacterial community, the archaeal community of cake A and cake B showed greater similarity, while the archaeal community of A10 was the most different from the other three samples (Fig. S4d).

PCR-DGGE profiles of excess sludge

The samples of excess sludge were collected on days 336, 340, 344, 364, and 374 during the operation of the reactors from sludge storage, marking them as E1, E2, E3, E4, and E5, respectively. In the meantime, the corresponding digested sludge samples from AnMBR-A and AnMBR-B were A11–A15 and B11–B15.

Figure S5 shows the PCR-DGGE fingerprints of digested sludge and excess sludge, in which the microbial community structure between the digested sludge and excess sludge exhibited obvious differences. With multiple bands showing similar intensities, for instance, the bacterial community of excess sludge was distributed rather evenly and, generally, did not change over time. By comparison, multiple noticeable dominant bacterial community (bands 2, 3, and 14) could be discerned in the digested sludge of both AnMBR-A and AnMBR-B. Having distinction of the archaeal community between excess sludge and digested sludge, for excess sludge, bands 7, 8, and 10 were evident; for the digested sludge, it is obvious for bands 15–18. Nevertheless, there still existed overlapping bands such as 3 and 4 for bacterial fingerprints and bands 4, 7, 8, 10, and 11 for archaeal fingerprints between excess sludge and digested sludge.

To further explore the relationship among samples, clustering analysis was conducted and shown in Figs. S4c and S4d. Three excess sludge samples (E1, E2, and E3) were always clustered together and away from the digested sludge samples regardless of the bacterial community or archaeal community. For the bacterial community, the digested sludge samples from AnMBR-A and AnMBR-B were individually accumulated into a cluster, suggesting that the bacterial community of the digested sludge possessed higher dissimilarity between the two reactors after operating for a long time (Fig. S4c). The situation of the archaeal community was more complex. Intriguingly, other than A11 and A12, the other digested sludge samples were more mutually similar no matter if they are from AnMBR-A or AnMBR-B (Fig. S4d).

Characterizations of bacterial community in digested sludge, cake layer, and excess sludge

Bacterial communities of digested sludge

Since limited microbial species could be detected by DGGE, the structure of the microbial community in digested sludge was further investigated and analyzed by using Illumina high-throughput sequencing. On account of little shifts in the archaeal community, the bacterial community was primarily focused for taxonomic identification. After filtering the low-quality reads and trimming the adapters, barcodes and primers, 15,804–43,268 bacterial sequence reads were obtained for each digested sludge sample. The total number of OTUs obtained at 3% distance was ranging from 863 to 1956.

The bacterial community structure of the digested sludge at the phylum level is demonstrated in Fig. 1, showing that Cloacimonetes, Chloroflexi, Bacteroidetes, Proteobacteria, Firmicutes, and Ignavibacteriae were the six dominant phyla detected in these samples. The relative abundance of Chloroflexi decreased in both AnMBR-A and AnMBR-B; additionally, that of Bacteroidetes and Proteobacteria was shown to have a declining tendency in AnMBR-B as well (Fig. 1). However, the relative abundance of Firmicutes increased from 6.8 to 28.3% in AnMBR-B along with the extension of operating time (Fig. 1).

Fig. 1
figure 1

Bacterial community structure of seed sludge and digested sludge from AnMBR-A and AnMBR-B at the phylum level. Note: “A” and “B” stand for AnMBR-A and AnMBR-B, respectively. A1 and B1 on day 207, A3 and B3 on day 219, A5 and B5 on day 231, A7 and B7 on day 243, A8 on day 252, B9 on day 255, A10 on day 270, A11 and B11 on day 329, A12 and B12 on day 333, A13 and B13 on day 339, A14 and B14 on day 345, as well as A15 and B15 on day 374 were collected as digested sludge samples in AnMBR-A and AnMBR-B, separately

Full size image

In order to thoroughly analyze the bacterial community structure and its variation characteristics, four predominant phyla, namely Chloroflexi, Bacteroidetes, Proteobacteria, and Firmicutes, were further classified at the class level (Fig. S6). Besides, shown in Fig. 2 is the bacterial community structure at the genus level.

Fig. 2
figure 2

Bacterial community structure of seed sludge and digested sludge from AnMBR-A and AnMBR-B at the genus level. Note: “A” and “B” stand for AnMBR-A and AnMBR-B, respectively. A1 and B1 on day 207, A3 and B3 on day 219, A5 and B5 on day 231, A7 and B7 on day 243, A8 on day 252, B9 on day 255, A10 on day 270, A11 and B11 on day 329, A12 and B12 on day 333, A13 and B13 on day 339, A14 and B14 on day 345, as well as A15 and B15 on day 374 were collected as digested sludge samples in AnMBR-A and AnMBR-B, separately

Full size image

The relative abundance of Chloroflexi for A15 only accounted for 6.77% of A1, while Chloroflexi proportion of B15/B1 was 7.31%, which was slightly higher than the former one (Fig. 1). Six classes were identified for the phylum Chloroflexi (Fig. S6a). Among them, notably, class Anaerolineae overwhelmingly dominated at each time point, though the relative ratio reduced in both AnMBRs over time (Fig. S6a). The most abundant genera affiliated to class Anaerolineae were identified as Longilinea, Leptolinea, and Bellilinea (Fig. 2). The proportion of Bacteroidetes decreased generally along with the operation (Fig. 1). Bacteroidia and Sphingobacteria were the two primary classes identified and affiliated to phylum Bacteroidetes; however, over 40% of it could not be classified (Fig. S6b). The relative abundance of the genus Petrimonas, which was affiliated to class Bacteroidia, presented generally a decreasing tendency in AnMBR-A, but comparatively stable in AnMBR-B. Lewinella, Terrimonas, and Ferruginibacter, belonging to class Sphingobacteria, were aerobic bacteria identified as the dominant genera (Fig. 2). Proteobacteria were mainly composed of five classes, namely Alphaproteobacteria, Betaproteobacteria, Deltaproteobacteria, Gammaproteobacteria, and Epsilonproteobacteria, whereas a lower relative abundance of Epsilonproteobacteria was observed (Fig. S6c). A decline of both Alphaproteobacteria and Betaproteobacteria could be noticed in AnMBR-A and AnMBR-B; in addition, the amount of reduction was larger in AnMBR-A as well (Fig. S6c). Comamonas, Thauera, and Dechloromonas were the three major genera identified in class Betaproteobacteria (Fig. 2). Besides, Smithella and Syntrophorhabdus were the two predominant genera detected in class Deltaproteobacteria (Fig. 2), which possessed relatively higher proportion for phylum Proteobacteria in AnMBR-B (Fig. S6c). Accounting for comparatively high proportions in Proteobacteria, Gammaproteobacteria had two predominant genera—Dokdonella and Arenimonas identified (Fig. 2). With higher relative proportion of Clostridia, Firmicutes were basically classified into class Clostridia, Bacilli, Negativicutes, and Erysipelotrichia (Fig. S6d). Genus Syntrophomonas was the dominant genus detected in class Clostridia, as well as Saccharofermentans (Fig. 2).

Comparative characterization of bacterial communities in cake layer and digested sludge

The bacterial community of digested sludge and cake layer at the phylum level is given in Fig. 3, signifying that phyla Cloacimonetes (over 21.0% in all four samples) and Proteobacteria (over 16.2% in all four samples) dominated in both digested sludge and cake layer, followed by phyla Chloroflexi, Bacteroidetes, Firmicutes, and Ignavibacteriae. It could be noted that the relative abundance of Cloacimonetes from the cake layer was lower than that from the digested sludge; for Ignavibacteriae, the disparity between the two samples of AnMBR-A was not evident; however, its relative abundance from the cake layer was lower in AnMBR-B.

Fig. 3
figure 3

The bacterial community structure of digested sludge and cake layer in AnMBR-A and AnMBR-B at the phylum level. Note: “A” and “B” stand for AnMBR-A and AnMBR-B, respectively. A10 and cake A were respectively collected as the digested sludge and cake layer samples in AnMBR-A on day 270. B9 and cake B were respectively collected as the digested sludge and cake layer samples in AnMBR-B on day 255

Full size image

Comparative characterization of bacterial communities in excess sludge and digested sludge

After identifying and classifying the DNA sequences, microorganisms from excess sludge and digested sludge were shown at the phylum and genus levels (Fig. 4). Being different from the digested sludge, excess sludge was phyla Bacteroidetes and Proteobacteria based, with relative abundance range of 17.4–26.5 and 34.6–40.5%, respectively. Belonging to the abovementioned two phyla, Lewinella, Terrimonas, Ferruginibacter, and Comamonas, whose proportions in excess sludge exceeded their proportions in digested sludge, were detected as the aerobic genera, indicating that most of the aerobic microorganisms distinguished or degraded in anaerobic environment. Anaerobic microorganisms, such as phyla Cloacimonetes and Ignavibacteriae, exhibited rather low relative abundance in excess sludge. Cloacimonetes became the most dominant phylum in digested sludge as the AnMBR systems were operated over time. Furthermore, the genus Smithella was detected in excess sludge and its relative abundance increased in digested sludge.

Fig. 4
figure 4

The bacterial community structure of digested sludge in AnMBR-A and AnMBR-B and excess sludge at a the phylum level and b genus level. Note: “A” and “B” stand for AnMBR-A and AnMBR-B, respectively. A11 and B11 on day 329, A12 and B12 on day 333, A13 and B13 on day 339, A14 and B14 on day 345, as well as A15 and B15 on day 374 were respectively collected from AnMBR-A and AnMBR-B as the digested sludge samples. “E” stands for excess sludge. E1 on day 336, E2 on day 340, E3 on day 344, E4 on day 364, and E5 on day 374 were separately collected from excess sludge

Full size image

Microbial diversity of digested sludge, cake layer, and excess sludge

The microbial diversity in digested sludge, cake layer, and excess sludge is presented in Table S3 in terms of Shannon-Wiener index. Since AnMBR-A and AnMBR-B started up with the identical seed sludge, the diversity of microbial community was quite similar at the initial stage (day 207). During the start-up phase, the systems were not stabilized and easily affected by the microbial community that existed in excess sludge, where the highest microbial diversity was observed (Table S3). Following the operation and stabilization of the systems, the microbial diversity gradually decreased, indicating that some of the microorganisms distinguished during digestion. AnMBR-A possessed higher microbial diversity in comparison with AnMBR-B. In addition, compared to digested sludge, the microbial diversity in cake layer was much higher (Table S3).

Microbial similarity among digested sludge, cake layer, and excess sludge

Presented in Fig. S7, microbial similarity was evaluated through performing DCA. It could be noted that the seed sludge possessed longer distance with other samples, indicating that the structure of the microbial community distinctly shifted after the long-time operation of AnMBR. Despite the fact that samples distributed relatively concentrated, still, certain patterns could be discovered (Fig. S7a). Since two systems started up with identical seed sludge, the distance between A1 and B1 was quite close; (b) at the beginning of the start-up phase, A3 and B3 still appeared in the same region showing consistent variation between the two systems; (c) during the long run, a clear distinction was discovered between the two systems, which were divided into different regions, namely A5–A8 and B5–B8; (d) during the stable state, little variation of microbial structure was observed within the same system, where A11–A15 and B11–B15 both fell into their own regions.

The similarity between excess sludge and digested sludge is shown in Fig. S7b. Existing obvious distance with digested sludge samples, excess sludge samples gathered together in a small region, exhibiting that the microbial community was clearly different.

Influence of environmental factors on microbial community in digested sludge

To analyze the factors that affected the microbial community, CCA was performed and displayed in Fig. 5. Note that in the results, both TS and TP had a significant influence on microbial community. In addition to the influences of TS and TP, reflecting the concentration of substrate and metabolite, COD also had a certain influence on microbial community. However, a lower impact of conductivity and ammonia concentration on microbial community was observed as well.

Fig. 5
figure 5

Canonical correspondence analysis (CCA) for environmental factors and bacterial community in AnMBR-A and AnMBR-B. Note: TS, COD, and TP stand for total solid, chemical oxygen demand, and total phosphorus, respectively. “A” and “B” stand for AnMBR-A and AnMBR-B, respectively. A1 and B1 on day 207, A3 and B3 on day 219, A5 and B5 on day 231, A7 and B7 on day 243, B9 on day 255, A10 on day 270, A12 and B12 on day 333, A13 and B13 on day 339, A14 and B14 on day 345, as well as A15 and B15 on day 374 were collected as digested sludge samples in AnMBR-A and AnMBR-B, separately

Full size image

Discussion

Based on the system performance data, AnMBR-A and AnMBR-B had almost identical mixed liquor conditions except for the TP concentration (Table S2), indicating that PAC affected the digesting performance negligibly. This was partially consistent with the results of Hwang et al. (2007), where no obvious differences of COD, total nitrogen (TN), and TP were observed between the control and adding flocculants. However, in our study, an obvious distinction of TP was found between the two reactors. Given that the only difference between the two systems during the operation was the PAC addition, it could have impacts on the different variations of TP in the two systems, as well as slight reduction of VS/TS ratio (Table S2). According to our previous study on AnMBRs, the amount of bioavailable phosphorus was reduced through PAC addition (Yu et al. 2015). Therefore, PAC not only affected the TP concentration but also the bioavailable nutrients in AnMBR-A and AnMBR-B. It thus further influenced the availability of substrate for microorganisms and the microbial structure in AnMBR systems.

To better understand the shifts in PCR-DGGE profiles, that represented the shifts of microbial structure, Dice indices were calculated and clustering analysis was conducted. The reduction of Dice indices and the similarity in clustering results for the bacterial community between A1 and B1, as well as between A8 and B8, confirmed the different tendency of digested sludge variation between AnMBR-A and AnMBR-B. The discrepancy of these two systems might be caused by additional PAC dosing since other operational parameters were identical. In addition, the Dice index and similarity in clustering results for the archaeal community decreased after a period of operational time but not as significant as the bacterial community, suggesting that the variation of the archaeal community was not as obvious as the bacterial community. No evident difference of the archaeal and bacterial community between digested sludge and cake layer in the same reactor was observed from PCR-DGGE profiles (Fig. S4) mainly because samples from the same system were homologous, and the microbial community of cake layer was originally from digested sludge. The PCR-DGGE profiles of digested sludge and excess sludge (Fig. S5) suggested that after being through the anaerobic digestion process, most of the microorganisms from the original seed sludge diminished. Yet, overlapping bands might have come from the same microorganism species, indicating that a part of the microorganisms from excess sludge remained a portion of the digested sludge microbial community.

The bacterial communities of digested sludge displayed in Fig. 1 showed that Cloacimonetes, Chloroflexi, Bacteroidetes, Proteobacteria, Firmicutes, and Ignavibacteriae were the dominant phyla. Phyla Chloroflexi, Bacteroidetes, Proteobacteria, and Firmicutes, possessing vital functions in degradation and transformation of organic compounds, had been commonly detected in anaerobic reactors (Nelson et al. 2011; Ma et al. 2013; Xie et al. 2014). Higher reduction of Chloroflexi relative abundance in AnMBR-A was observed in Fig. 1. Durban et al. (2016) have proposed that aluminum salts could lead to a growth limitation for phylum Chloroflexi. In this case, since the only difference between two reactors was the PAC addition, the growth of phylum Chloroflexi was inhibited to a much greater extent in AnMBR-A than in AnMBR-B. Among the six classes identified for phylum Chloroflexi (Fig. S6a), most abundant genera affiliated to class Anaerolineae were identified as Longilinea, Leptolinea, and Bellilinea (Fig. 2). As for the mesophilic and thermophilic anaerobic bacteria, their growth was observed with different carbon resources, while weak growth occurred with the existence of casamino acids and other proteins (Yamada et al. 2006, 2007). Concerning a large amount of protein that existed in the supernatant of digested sludge (Yu et al. 2012), the growth of Chloroflexi could be also suppressed as the time went on, which led to Chloroflexi reduction in both reactors.

As for aerobic bacteria, Lewinella, Terrimonas, and Ferruginibacter (Fig. 2), belonging to class Sphingobacteria from phylum Bacteroidetes, might come from the original seed sludge (Khan et al. 2007; Lim et al. 2009; Zhang et al. 2012). On the one hand, the proportion of Bacteroidetes decreased along with the operation since these aerobic bacteria were substituted under their unfavorable anaerobic condition; on the other hand, according to Pan et al. (2014), phylum Bacteroidetes was negatively correlated with aluminum; hence, this might lead to the decreasing tendency in AnMBR-A as well. Genus Petrimonas, belonging to the mesophilic anaerobic bacteria of class Bacteroidia, can produce acetic acid, hydrogen gas, and carbon dioxide when organic compounds are metabolized, and those products provided important substrates for methanogens (Grabowski et al. 2005). In addition, aluminum could lead to a higher inhibition (over 50%) to methanogenic and acetogenic bacteria (Cabirol et al. 2003). The relative abundance of Petrimonas presented a reducing trend in AnMBR-A but showed comparative stability in AnMBR-B along with the operation. Thus, the methanogenic and acetogenic bacteria belonging to phylum Bacteroidetes could be reduced under PAC addition over time.

The reduction of Alphaproteobacteria and Betaproteobacteria could be noticed in both AnMBR-A and AnMBR-B, but the degree of declining was greater in AnMBR-A (Fig. S6c). A Previous study has shown that most phosphate-accumulating organisms (PAOs) in WWTPs were Alphaproteobacteria and Betaproteobacteria (Mehlig et al. 2013). In the context of PAC addition, the reduction and lower concentration of TP was observed in AnMBR-A (Table S2); in a way, less bioavailable phosphorous could be provided for PAOs. This had been also confirmed by our previous published results (Yu et al. 2015), where the ratio of bioavailable phosphorus declined gradually as the dosage of PAC addition increased. Therefore, the limited resource of bioavailable phosphorus became a greater restriction for the growth of Alphaproteobacteria and Betaproteobacteria in AnMBR-A. As aerobic and facultative anaerobic bacteria and identified in class Betaproteobacteria (Fig. 2), the genus Comamonas, Thauera, and Dechloromonas could be detected in both activated sludge and anaerobic digester and utilize simple organic compounds like acetic acid (Anders et al. 1995). No growth or poor growth occurs at 37 °C for some species of the genus Thauera (Anders et al. 1995). Thus, the mesophilic anaerobic condition in this study might lead to the decreasing abundance of Betaproteobacteria in both systems (Fig. S6c). Being able to convert propionate and aromatic compounds into acetate through syntrophic association with hydrogenotrophic methanogens, Smithella and Syntrophorhabdus were two predominant genera detected in class Deltaproteobacteria (de Bok et al. 2001; Qiu et al. 2008). Since propionate was the main product during the hydrolytic acidification phase of anaerobic digestion, therefore, Deltaproteobacteria possessed a relatively higher proportion in AnMBR-B (Fig. S6c). The Genus Dokdonella and Arenimonas were identified in Gammaproteobacteria (Fig. 2). Bacosa and Inoue (2015) demonstrated that polycyclic aromatic hydrocarbons (PAHs) could be potentially degraded by Dokdonella; thus, it might play a key role in the process of degrading complex organic matters. Some species from Arenimonas could also use propionate as substrate; accordingly, this might result in relatively high ratio of Gammaproteobacteria in the system (Young et al. 2007).

The overall relative abundance of phylum Firmicutes was higher in AnMBR-B than in AnMBR-A (Fig. 1). It was reported that Firmicutes were positively correlated with phosphorus (Pan et al. 2014). Due to the high TP concentration that existed in AnMBR-B, more abundant Firmicutes were observed. Therefore, the PAC addition affected the abundance of Firmicutes through reducing the TP concentration. Classes Clostridia, Bacilli, Negativicutes, and Erysipelotrichia were identified in phylum Firmicutes (Fig. S6d). Syntrophomonas and Saccharofermentans, which could use complex organic matters, produce acetate, and had similar optimal temperature with mesophilic methanogens, were the primary genus detected in class Clostridia (Fig. 2; Chen et al. 2010). Acetate is the essential substrate for anaerobic methanogens; hence, genus Saccharofermentans enrichment is beneficial to sludge digestion. Genus Syntrophomonas, which has the same optimal temperature with mesophilic methanogens, could be syntrophically in co-culture with Methanobacteria and use a sole substrate when grown independently (Wu et al. 2006b). Thus, Syntrophomona could be enriched during the process of anaerobic sludge digestion.

Phyla Cloacimonetes and Proteobacteria dominated in both digested sludge and cake layer, followed by phyla Chloroflexi, Bacteroidetes, Firmicutes, and Ignavibacteriae (Fig. 3). Nevertheless, Ersahin et al. (2016) suggested that Firmicutes, Bacteroidetes, Proteobacteria, Chloroflexi, and Acidobacteria were the five most predominant bacterial phyla in seed sludge, digested sludge, and cake layer from the submerged anaerobic dynamic membrane bioreactor. The enrichment of Cloacimonetes and Ignavibacteriae in digested sludge was potentially due to the abundant organic matters in the supernatant that provided a favorable proliferation condition. As mentioned above, the growth of phylum Chloroflexi might be suppressed in digested sludge since a large amount of proteins existed in the supernatant; in contrast, it was much easier to enrich in the cake layer. Although a similar bacterial structure was observed between cake A and cake B, the relative abundance of Chloroflexi was obviously higher in cake B (Fig. 3). On the one hand, as mentioned above, aluminum salts could limit the growth of phylum Chloroflexi (Durban et al. 2016), and it was thus less abundant in cake A because of the PAC addition; on the other hand, the bacterial communities in the cake layer accumulated during the whole operational period, and the discrepancy of phylum Chloroflexi was more evident in the cake layer than in the digested sludge. A previous study has proven that phyla Bacteroidetes, Verrucomicrobia, and Acidobacteria were negatively correlated with aluminum (Pan et al. 2014). That is, the existing higher aluminum would reduce their abundance. In our results, these three phyla were all less abundant in cake A and presented higher relative abundance in cake B. Therefore, the PAC addition negatively affected the relative abundance of phyla Bacteroidetes, Verrucomicrobia, and Acidobacteria in the cake layer.

Distinct structures of the bacterial community between the digested sludge and excess sludge were observed (Fig. 4). Anaerobic bacteria exhibited a rather low relative abundance in excess sludge, since facultative anaerobic microorganisms could survive in aerobic activated sludge where to some extent they also partially existed in the anaerobic microenvironment. Cloacimonetes becoming the most dominant phylum in the digested sludge was due to, for one thing, the proliferation of Cloacimonetes and, for another thing, the inoculation and accumulation of Cloacimonetes from the original excess sludge. In addition, genus Smithella was detected but possessed a relatively low abundance in excess sludge. As mentioned before, Smithella can convert propionate into acetate and symbiotically associated with hydrogenotrophic methanogens. Its growing relative abundance in digested sludge was because of the contribution from excess sludge and the anaerobic condition.

The diversity of the digested sludge gradually decreased in both AnMBRs, whereas the cake layers exhibited a higher diversity compared to the corresponding digested sludge samples (Table S3), which is more likely because the cake layer provided a relatively stable environment for microorganisms and was in favor of the enrichment of different microorganisms. Having the highest diversity in excess sludge and declining after being digested indicates that some of the microorganisms were distinguished, whereas some of the dominant ones got enriched (Table S3). The results of the study of Yu et al. (2016b) also showed that Shannon indices exhibited decreasing bacterial diversity during excess sludge digestion. For the digested sludge, AnMBR-A showed higher microbial diversity in comparison with AnMBR-B. This was opposite with the result of Faoro et al. (2010), where Al3+ and phosphorus significantly decreased the biodiversity. However, the situation in our study was different because excess sludge was added into the AnMBRs several times during the operation. The PAC addition has resulted in the reduction of excess sludge digesting efficiency in AnMBR-A. Since excess sludge possessed higher diversity and less excess sludge has been digested in AnMBR-A, higher bacterial diversity was detected in AnMBR-A. Higher diversity in AnMBR-A was probably caused by the excess sludge, instead of the digested sludge. The PAC addition eventually and indirectly resulted in the variation of bacterial diversity in AnMBR.

The influence of environmental factors on the microbial community is displayed in Fig. 5. The increase of TS concentration could be explained by the accumulation of inorganics and nonbiodegradable organics, as well as PAC addition. Having an impact on the interaction between the substrate and microorganisms, the increase of sludge concentration would interfere with the metabolism of microorganisms, furthermore leading to the structural shift of the microbial community. The divergence of TP mainly originated from PAC addition, which could result in the reduction of substrate that could be utilized by microorganisms and thus affected the metabolic process of the anaerobic community. The structure of the microbial community consequently varied. In addition to that, the activity of microorganisms would be affected by very low or very high concentration of substrate or metabolite accumulation, which was reflected by COD concentration.

PAC addition, on one hand, changed the proportion of bioavailable nutrient such as phosphorus in the AnMBR systems and further affected the interaction between the microbial community and substrate; consequently, the structure of the microbial community is transformed and the aluminum concentration in AnMBR-A is changed, thus affecting certain microbial communities. Besides, PAC dosing indirectly influenced the bacterial diversity of AnMBR as well.