The microbial community of 21 full-scale biogas reactors was examined using 454 pyrosequencing of 16S rRNA gene sequences. These reactors included seven (six mesophilic and one thermophilic) digesting sewage sludge (SS) and 14 (ten mesophilic and four thermophilic) codigesting (CD) various combinations of wastes from slaughterhouses, restaurants, households, etc. The pyrosequencing generated more than 160 000 sequences representing 11 phyla, 23 classes, and 95 genera of Bacteria and Archaea. The bacterial community was always both more abundant and more diverse than the archaeal community. At the phylum level, the foremost populations in the SS reactors included Actinobacteria, Proteobacteria, Chloroflexi, Spirochetes, and Euryarchaeota, while Firmicutes was the most prevalent in the CD reactors. The main bacterial class in all reactors was Clostridia. Acetoclastic methanogens were detected in the SS, but not in the CD reactors. Their absence suggests that methane formation from acetate takes place mainly via syntrophic acetate oxidation in the CD reactors. A principal component analysis of the communities at genus level revealed three clusters: SS reactors, mesophilic CD reactors (including one thermophilic CD and one SS), and thermophilic CD reactors. Thus, the microbial composition was mainly governed by the substrate differences and the process temperature.
Anaerobic digestion (AD) of organic material is a proven technology for methane formation and the recycling of nutrients (Börjesson & Mattiasson, 2008). The application of the biogas process has evolved from the treatment of mainly sewage sludge and manure toward the use of more energy-rich waste mixtures with higher methane potentials, such as slaughterhouse-, food industry-, and household wastes; glycerol; and frying fats (Deublein & Steinhauser, 2008). However, energy-rich substrates may increase the stress on the operational system, because these substrates may result in a less stable process (Salminen & Rintala, 2002; Edström et al., 2003; Cuetos et al., 2008). Such disturbances may lead to suboptimal use of the organic material challenging efforts to obtain an efficient utilization of energy-rich substrates. An improved understanding of the microbial communities and their function during the different aspects of AD may help to optimize biogas production, and molecular biology techniques offer possible tools.
The anaerobic degradation of organic matter mainly proceeds in a series of four metabolic steps: hydrolysis, fermentation, acetogenesis, and methanogenesis (c.f. Zinder, 1984). A diverse number of Bacteria take part in the hydrolysis and fermentation steps, whereas the oxidation of intermediate fermentation products to acetate is performed by either hydrogen- or formate-producing acetogens (Stams & Plugge, 2009). Finally, methane formation is mainly derived from acetate and hydrogen/CO2 by methanogenic Archaea. A balanced interaction between the microorganisms in this degradation chain is crucial for the continuous transformation of the intermediates formed and subsequently an efficient biogas production. Methanogens are especially important for the obligatory syntrophic interactions driving the acetogenic proton reduction needed for growth on, for example, fatty acids and alcohols (c.f. Stams & Plugge, 2009). However, the roles and interactions of specific microorganisms within the biogas-producing communities are very complex as shown in a number of studies (Fernández et al., 1999; McHugh et al., 2003, 2004; Leclerc et al., 2004; Klocke et al., 2007; Levén et al., 2007; Cheon et al., 2008; Lee et al., 2008; Zhang et al., 2009; Feng et al., 2010; Krakat et al., 2010). A deeper understanding of the microbial community structure and function dynamics during different settings of AD is therefore vital to improve process performance. Studies of the AD microbial community in general have included cloning and sequencing of 16S rRNA gene (c.f. McHugh et al., 2003; Cheon et al., 2008; Klocke et al., 2008; Zhang et al., 2009) as well as specifically addressing the methanogenic Archaea using the functional mcrA gene (Nettmann et al., 2008; Cardinali-Rezende et al., 2009). Furthermore, links between bacterial community analysis and metabolic function have been addressed in a few studies (Ariesyady et al., 2007; Abram et al., 2011; Regueiro et al., 2012).
The more recent development of ‘next-generation sequencing’ such as 454 pyrosequencing has made it possible to efficiently deep-sequence microbial communities in complex biological samples without the time-consuming cloning procedure. The technique has so far been used for the sequencing of metagenomes from a number of biogas reactors (Schlüter et al., 2008; Werner et al., 2011; Lee et al., 2012). Schlüter et al. (2008) described the bacterial community from a full-scale, completely stirred tank reactor (CSTR) digesting maize silage (63%) and green rye (35%) together with small amounts of chicken manure. Bacterial members of the taxonomic classes Clostridia and Bacteroidetes were most abundant. Among the Archaea, the hydrogenotrophic Methanoculleus sp. dominated, but the acetoclastic Methanosarcina sp. were also detected. Lee et al. (2012) used 454 pyrosequencing of the V1, V2, and V3 regions of the 16S rRNA gene to assess the microbial community in seven full-scale reactors over time. Six of the reactors treated waste-activated sludge (one of these in combination with smaller amounts of food waste), and one reactor treated night soil. Sequences belonging to Proteobacteria, Bacteroidetes, Firmicutes, and Chloroflexi were found to be the most abundant, and the bacterial population was influenced by the digestion temperature. Werner et al. (2011) characterized bacterial communities in nine full-scale granulated sludge reactors treating brewery waste water by targeting a part of the rRNA gene. These bacterial communities were dominated by Syntrophobacterales, Desulfuromonales, Bacteroidetes, Spirochetes, Clostridia, Chloroflexi, and Synergistia.
The aim of the present study was to characterize the microbial communities in full-scale digesters and to relate the community structures to process parameters such as substrate composition, temperature, hydraulic retention time (HRT), and organic loading rate (OLR). The digesters selected reflect typical substrates and substrate combinations used in Sweden. These include sewage sludge (SS) and codigestion (CD) mixtures of slaughterhouse, household and restaurant waste, crops, wheat stillage, and/or manure (Table 1). 454 pyrosequencing was used to investigate the microbial communities, and the set of sequences obtained were, together with the full-scale process parameters of the sampled reactors, analyzed using principle component analysis (PCA).
Table 1. Full-scale process parameters of the biogas reactors in this study. The temperature, OLR, HRT, SBP, total VFA concentration, concentration, and pH are given
OLR (kg VS m−3 day−1)
SBP (m3 kg VS−1 day−1)
VFA (mg L−1)
n.a.; not available; SBP, specific biogas production; BSS, Biological sewage sludge; FIW; food industry waste, OFMSW, organic fraction municipal solid waste; PSS, primary sewage sludge; SHW, slaughterhouse waste.
2–6% fat is included in the remaining substrate.
1.5% iron-rich sludge is included in the remaining substrate.
FIW (62) and PSS/BSS (38)
Sludge from SS1as
BSS (60) and PSS (40)
BSS (67), PSS (30), and sludge from external source (3)
All of the 21 full-scale biogas reactors sampled were CSTRs, operated at mesophilic or thermophilic conditions (see Table 1 for details regarding substrate, HRT, OLR, etc.), and had been running for more than 2 years. While minor changes in substrate composition and organic loading may occur over time, all plants reported normal operating conditions at the time of sampling, and no major changes had occurred prior to sampling. Therefore, the samples withdrawn are considered to be representative of each reactor over time. The reactors sampled are coded as follows: if more than one SS or CD reactor from the same plant was sampled, they are marked a and b. Reactors running in series are denoted by s; those running in parallel are denoted by p, while T stands for thermophilic (50–55 °C).
The reactors were sampled during March 2009, by withdrawing digester material from the internal circulation sampling loop system of each reactor after flushing the loops. The samples for the microbial analysis were collected in sterile tubes (15 mL; Sarstedt, Nürnbrecht, Germany), and the volatile fatty acid (VFA) samples were collected in 0.5-L plastic bottles. During transportation to the laboratory, the 0.5-L bottles were kept in a cooler with freezer blocks, while the 15-mL samples were frozen immediately in a cooler with dry ice. Both sample types were stored in the laboratory at −20 °C.
All process parameters (except VFA) given in Table 1 have been obtained from the records of the specific plants. The VFA content (acetate, propionate, isobutyrate, butyrate, isovalerate, valerate, isocaproate, and caproate) in the sampled reactor liquids was determined according to Jonsson & Borén (2002).
Prior to DNA extraction, all samples were thawed and shaken by hand before transferring 0.4 g to the Lysing Matrix E tubes included in the extraction kit (Fast DNA® SPIN Kit for Soil, MP Biomedicals, LLC, Solon, OH). The extraction was performed according to the manufacturer's protocol, and the DNA obtained was stored at −20 °C until further analysis. Concentrations of double-stranded DNA in the extracts were determined using the Quant-iT dsDNA HS assay kit and the Qubit fluorometer (Invitrogen, Lidingö, Sweden).
DNA amplification and 454 pyrosequencing
A 100-fold dilution of the extracted DNA was shown to give the best amplification results and hence applied in this study. DNA amplification was performed on a Primus 96 Advanced® Gradient (PEQLAB, Biotechnologie GmbH, Erlangen, Germany). The primers used for pyrosequencing were a modified 341F (CCTAYGGGRBGCASCAG) and a modified 806R (GGACTACNNGGGTATCTAAT; Yu et al., 2005), which amplified a DNA fragment of 465 bp length flanking the V3 and V4 regions of the 16S rRNA gene. Modification of primer 341F was made by changing the base 5 from the 5′ end from C to Y and of primer 806R by changing the bases 8 and 9 from the 5′ end from YV to NN to increase primer coverage of sequences available in the Ribosomal Database Project (RDP) database. The first PCR amplification (20 μL) was performed using 1X Phusion HF buffer, 0.2 mM dNTPs mixture (5Prime GmbH, Hamburg, Germany), 0.4 U Phusion High-Fidelity DNA Polymerase (Finnzymes Oy, Espoo, Finland), 0.5 μM of each of the primers 341F and 806R, and 5 μL of the diluted DNA sample. The PCR incubation conditions were as follows: 98 °C for 30 s, followed by 35 cycles of 98 °C for 5 s, 56 °C for 20 s, and 72 °C for 20 s with a final extension of 72 °C for 5 min. The separation of PCR products was performed on a 1% agarose gel stained with ethidium bromide. Bands were cut out and purified using a High-Pure PCR Cleanup Micro Kit, according to the manufacturer's instructions (Roche Diagnostics GmbH, Mannheim, Germany). DNA concentrations of the cleaned PCR products were again determined using the Quant-iT dsDNA HS assay kit and the Qubit fluorometer (Invitrogen). The concentration of total double-stranded DNA in each sample was adjusted to 5 ng μL−1. A second PCR amplification (25 μL) was performed using the method described above, except that the fusion primers 341F and 806R with adapters and tags were used (Supporting Information, Table S1) and the number of cycles was reduced to 10. Bands of 520 bp were cut out, purified, and quantified with the Qubit fluorometer as described above and with qPCR (Mx-3000, Stratagene; Larsen et al., 2010). The sample amplicons were adjusted to a concentration of 4 × 105 copies μL−1 to ensure equal representation of each sample. A two-region 454 pyrosequencing run was performed on a 70 × 75 GS PicoTiterPlate using a GS FLX pyrosequencing system according to the manufacturer's instructions (Roche).
The sequences were checked for chimeras using Chimera.perseus (mothur; www.mothur.org/wiki/). Sorting and trimming were performed using the Pipeline Initial Process at the RDP Pyrosequencing Pipeline (http://rdp.cme.msu.edu/; Cole et al., 2009) with the default settings. The analysis of the sequencing data was conducted using Pyrosequencing Pipeline tools at RDP 10 (http://pyro.cme.msu.edu). The RDP Classifier was used to assign 16S rRNA gene sequences to a taxonomical hierarchy with a confidence threshold of 50%, because the DNA sequences were < 250 bp (Claesson et al., 2009). An estimation of the microbial richness across the samples using Chao1 index was performed at a distance of 3% using QIIME pipeline according to Caporaso et al. (2010).
A principal component analysis (PCA) of the data from the pyrosequencing was performed at the taxonomic phylum, class, and genus levels using software developed in-house in matlab® 7.3.0. Due to the high variability in sequences of low abundance within all of the reactor types, sequences occurring at < 1% in all samples at the phylum level and at < 10 fragments in all samples at class and genus levels were grouped into ‘others’ at the corresponding taxonomic level prior to the data analysis (Table S2). The group of ‘others’ ranged from 0.1% to 1.3% of the total sequences on the phylum level and from 0% to 0.1% and from 0.001% to 0.03% of the total fragments on the class and genus levels, respectively (Table S2). This reduced the number of represented phyla from 27 to 11, the classes from 32 to 23, and the genera from 332 to 94. The remaining variables were mean-centered and scaled to unit variance. The cohesiveness and number of major clusters in the PCA plots were investigated by hierarchical cluster analysis on the first (PC1) and second (PC2) principal components, which covered the main variance in the data set. The ‘squared Euclidean distance’ was calculated as a measure of distance between PC1 and PC2 with the method of ‘average linkage between groups’ using spss software (IBM Company). Differences in the relative abundance of Bacteria and Archaea among SS, CD, and CDT reactors were analyzed using the Mann–Whitney U-test (Rank Sum; P < 0.05). Thus, all claimed differences of phyla, classes, and genera refer to these tests, which are tabulated in the supplemental information, together with the mean, minimum, and maximum relative abundances (Tables S3–S5).
The sampled biogas plants
The process parameters (Table 1) differed considerably between the CD and SS reactors. The pH, ammonium, and VFA levels were higher in the CD reactors, except for the low VFA levels in CD1 and CD11a and CD11b. It should be noted that mesophilic CD reactors generally run at higher OLRs and with longer HRTs than SS digesters. The more energy-rich substrates in the CD reactors typically result in higher methane yields per kg of added volatile solids (VS) compared with the digestion of sewage sludge (Table 1).
Overall phylogenetic analysis
The total number of sequences obtained was 250 478, with an average length of 216 bp. The number was reduced to 164 822 after trimming, removing chimeras, and omitting sequences with a quality score < 20 and shorter than 150 bp. The average sequence length of the trimmed sequences was 239 bp. The number of sequences obtained from the different samples was between 1760 and 40 000 (Table 2). In total, 10 and 2 phyla, 20 and 3 classes, and 87 and eight genera were found among Bacteria and Archaea, respectively. Details regarding the total bacterial and archaeal sequences as well as the numbers of phyla, classes, and genera found in each reactor are given in Table 2, and details on their distribution can be found in the supplementary material (Tables S3–S5).
Table 2. The number of bacterial and archaeal sequences observed in the 21 biogas reactor samples (SS, sewage sludge-fed, CD, codigestion during mesophilic conditions, and CDT, codigestion during thermophilic conditions) is given together with percentage of Bacteria and Archaea of the total number of sequences
Number of phyla (Bacteria/Archaea)
Number of classes (Bacteria/Archaea)
Number of genera (Bacteria/Archaea)
The number of identified phyla, classes, and genera within Bacteria and Archaea domains is also listed.
10 : 1
19 : 3
57 : 6
9 : 1
18 : 3
49 : 5
9 : 1
19 : 3
60 : 5
10 : 1
19 : 2
44 : 4
9 : 1
18 : 3
52 : 6
10 : 1
20 : 2
65 : 7
9 : 1
17 : 3
56 : 5
5 : 1
7 : 2
24 : 3
10 : 1
17 : 1
45 : 1
7 : 1
12 : 2
39 : 3
6 : 1
7 : 2
18 : 3
7 : 0
12 : 0
28 : 0
9 : 1
13 : 2
36 : 3
10 : 1
15 : 2
35 : 2
8 : 1
13 : 2
41 : 2
10 : 1
16 : 2
47 : 3
10 : 1
14 : 2
43 : 2
7 : 0
13 : 0
36 : 0
9 : 1
15 : 2
48 : 3
5 : 1
9 : 1
14 : 1
8 : 1
12 : 1
25 : 2
The mean Chao1 index of SS, CD, and CDT reactors at 3% distance and with 95% confidence intervals indicates a trend of higher microbial richness in the SS reactors compared with the CD reactors. The CDT reactors contained the lowest microbial richness. This calculation was based on an equal number of sequences (1760 sequences, which was the lowest extracted in any reactor) selected randomly (bacterial and archaeal sequences together) from each reactor. However, the differences were not statistically significant (Table S6).
On average, Bacteria accounted for 95% (range 80–100%) of the sequences detected in the reactors, while the rest belonged to Archaea (Table 2). The ratio of archaeal sequences was higher in the SS reactors (mean 10%, range 2–20%) than in the CD reactors (mean 1.8%, range 0–4.4%).
Richness at the taxonomic phylum level
The relative abundance of archaeal and bacterial groups in the different anaerobic reactors was analyzed at the phylum, class (Bacteria), and genus (Bacteria and Archaea) levels including unclassified sequences. Thirty percent and eight percent of the bacterial sequences in SS and CD reactors, respectively, were designated as unclassified Bacteria, because it was not possible to assign them to specific phyla. A cluster analysis at 97% similarity of these unclassified sequences using OTUpipe integrated in QIIME pipeline (Caporaso et al., 2010) showed that the number of clusters varied among the reactors; hence, the unclassified OTUs from the different reactors were not closely related (Table S7).
Archaeal sequences belonging to the phylum Euryarchaeota were detected in all but two of the CD reactors (CD1 and CD7b). Because the coverage analyses of the general prokaryotic primers used showed that they targeted both Bacteria and Archaea, the lack of detection of Archaea in the two reactors is likely due to a low abundance compared with the Bacteria (Table 2).
The bacterial sequences were represented by 10 phyla (Tables 2 and S3) with Firmicutes and Bacteroidetes dominating with 53% and 13% of the total number of sequences, respectively. The relative occurrence of Firmicutes sequences was higher in the CD than in the SS reactors (69% vs. 25%; Table S3). Sequences belonging to Bacteroidetes were at similar levels for both the SS (15%) and the mesophilic CDs (14%), while constituting 7% of the sequences in the CDTs (Table S3). Sequences belonging to Actinobacteria, Proteobacteria, and Spirochetes accounted for 4–6% each, and Verrucomicrobium, Acidobacterium, and Chloroflexi ranged between 0.5% and 2% in the SS reactors, while their contribution was < 0.8% in the CDs. Thermotogae sequences were more prevalent in the CDT reactors than in the mesophilic SS and CD reactors (Table S3).
The PCA biplot on the microbial communities of all reactors at the phylum level distinguished two clusters: one with all SS reactors except SS3b and SS5 and one with all CD reactors (Figs 1 and S1). Close to 70% of the variation in the microbial community structures among the reactors was explained by PC1 and PC2. All the CDTs in the biplot are found in the lower right part of the CD cluster. The CD reactors gave tighter clusters than the SS reactors (Figs 1 and S1), which is related to the high number of sequences belonging to the Firmicutes and Thermotogae phyla that are characteristic of the CDs. The archaeal phylum Euryarchaeota is closer to the SS than to the CD cluster, which is a result of their higher abundance in the SS reactors (Table 2).
Archaeal abundance at genus level
Between four and seven archaeal genera sequences were found in the SS reactors, whereas one to three genera were detected in the CD reactors (Table 2). All percentages mentioned below reflect the relative abundance of detected archaeal sequences.
All of the SS reactors were dominated by sequences from the acetoclastic Methanosaeta sp. (68–87% of total archaeal sequences; Table S5) except for SS5, in which the hydrogenotrophic Methanobrevibacter sp. accounted for 98% of the sequences. Methanobrevibacter sp. was also found in SS3b and SS4ap (Table S8). Other detected hydrogenotrophic sequences encountered include Methanobacterium sp., Methanoculleus sp., Methanosphaera sp., Methanogenium sp., and Methanofollis sp. The latter three were only detected in the SS reactors (c.f. Tables S5 and S8).
The mesophilic CD reactors were dominated by hydrogenotrophic methanogens, and parallel reactors at the same plant showed the same pattern. As a result, Methanoculleus sp. was observed in a range of 94–100% in CD11ap and CD11 bp, and CD10 (Table S8), while sequences of Methanobrevibacter sp. ranged from 91% to 98% in CD4ap and bp, in CD5ap and CD5 bp, as well as in CD7a (Table S8). Meanwhile, sequences from the acetoclastic Methanosaeta sp. were present in CD4ap, CD5 bp, CD10, and CD11b up to c. 7%.
Two of the thermophilic CD reactors, CD2T and CD9T, were dominated by Methanobacterium sp. sequences at 90% and 56%, respectively, plus low amounts of sequences belonging to Methanoculleus sp (Table S8). In CD3T, only sequences from Methanobrevibacter sp. were identified; while Methanoculleus sp. (46%) and Methanobacterium sp. (44%) were the main genera in CD8Tap. Sequences belonging to the thermophilic Methanothermobacter sp. were detected in all of the CDTs, except CD3T: c. 6% in CDT2 and CDT8 and 32% in CDT9 (Table S8). Methanosaeta sp. sequences were not identified in the thermophilic CD reactors.
The PCA biplot gave three clusters for the archaeal sequences at the genus level including unclassified sequences at the class and order levels (Figs 2 and S2). Six of the seven SS reactors clustered together were dominated by sequences belonging to the acetoclastic Methanosaeta sp. A second cluster encompasses three of the four thermophilic CD reactors, which are characterized by hydrogenotrophic communities. The third cluster includes all of the mesophilic CD reactors along with SS5 and CD3T. Reactors CD10 and CD11ap and CD11 bp of this cluster are dominated by sequences belonging to Methanoculleus sp., while the rest are characterized by large amounts of sequences from Methanobrevibacter sp.
Bacterial community at class level
The sequence analyses at the taxonomical bacterial class and genus levels identified 20 classes (Table S9) and 87 different genera (Table S10) in the 21 reactors. Details on the relative abundances at the class level (means, minima, and maxima) and differences among the reactor groups (SS, CD, and CDT) are given in Table S4. All percentages given below represent the relative abundance of the detected bacterial sequences.
Sequences belonging to four classes were identified in all reactors: Clostridia (15–84%), Sphingobacteria (0.1–8%), Bacilli (< 0.1–5%), and Erysipelotrichi (< 0.1–2%), as well as unclassified sequences of Firmicutes (1–14%) and Bacteroidetes (0.2–6%; Table S9).
Sequences from 17 classes were represented in all of the SS reactors (for distribution see Tables S4 and S9). The most abundant sequences came from Clostridia (15–38%), followed by Actinobacteria (up to 23%), Spirochetes (up to 16%), Bacteroidetes, and Flavobacteria (c. 10%). Proteobacteria was represented mainly by Delta-proteobacteria sequences (up to 7%), while Alpha-, Beta-, and Gammaproteobacteria contributed to < 1% of the total number of sequences. Furthermore, sequences belonging to Thermotogae, Sphingobacteria, Anaerolineae, Verrucomicrobiae, Erysipelotrichi, Acidobacteria, Deferribacteres, Bacilli, and unclassified Proteobacteria were identified in all of the SS reactors (Table S9), and the Gemmatimonadetes sequences were detected in all of the mesophilic SS reactors.
Regarding the mesophilic CD reactors, sequences from seven bacterial classes were identified in all of the reactors (Table S9). Sequences belonging to the Clostridia class dominated (40–76%), with sequences from Bacteroidetes and Flavobacteria as the next most common (0.3–21%). Furthermore, Bacilli, Sphingobacteria, Erysipelotrichi, and Mollicutes sequences were identified in all of the mesophilic CD reactors. Sequences from Actinobacteria and Spirochetes were detected in all but one of the CDs (Table S9); meanwhile, sequences belonging to Thermotogae and Alphaproteobacteria were found in eight of the ten mesophilic CD reactors (Table S9). Unclassified Proteobacteria sequences were observed in all of the mesophilic CDs, except CD11ap.
In the thermophilic CD group, all samples included sequences belonging to six bacterial classes (Table S9). Again Clostridia sequences dominated (31–84%), while Thermotogae (ranging from 0.04% to 48%; Table S9) was the second largest community, with the highest number of sequences in CD8Tap and the lowest in CD3T. Sequences belonging to Sphingobacteria were the third most abundant, followed by Actinobacteria, Bacilli, and Erysipelotrichi (Table S9). Of the thermophilic CDs, CD3T showed the highest richness with sequences detected from Acidobacteria; Alpha-, Beta-, and Gammaproteobacteria; Spirochetes; and Verrucomicrobiae (Table S9).
The SS reactors included a higher abundance of unclassified sequences at the class level as well as more unclassified sequences at the phylum level compared with the CDs (Tables S3 and S4).
The PCAs of the bacterial communities in the 21 reactors at the class level (Fig. 3) resulted in the same two clusters as the analysis at the phylum level (c.f. Fig. 1). The PC1 and PC2 accounted for 55% of the variation in the microbial community composition among the reactors (Fig. 3). The SS reactors, excluding SS3b and SS5, clustered together with the main part of the bacterial classes (Figs 3 and S3) and, thus, seem to be characterized by a higher bacterial richness compared with the CDs (Table S9). SS3b had a large number of sequences belonging to Actinobacteria, while the relatively large numbers of Thermotogae and Mollicutes sequences likely distinguished SS5 from the other SS reactors. SS2T showed a tendency to move toward the CD cluster because of its high content of Sphingobacteria sequences. The second cluster contained all CD reactors (Fig. 3). Three of the four thermophilic reactors (CD2T, CD8T, and CD9T) yielded high numbers of Thermotogae and Sphingobacteria sequences. The rest of the CD reactors were characterized by a high abundance of Clostridia and Bacilli sequences. All of the CD reactors displayed higher numbers of unclassified Firmicutes sequences than the SS reactors.
Bacterial community at genus level
The PCA performed at the bacterial genus level resulted in the same clusters for PC1 and PC2 (Figs 4 and S4) as for the archaeal genus level (Fig. 2). However, the degree of explanation is lower for the bacterial than for the archaeal community biplot, that is, 36% and 63%, respectively. As in the case of the methanogens, parallel reactors displayed approximately equal patterns and were, thus, positioned closely together in the PCA plots (Fig. 4).
An extensive screening of 21 full-scale, CSTRs was performed and elicited noticeable differences between those processing sewage sludge vs. codigested substrates for both bacterial and archaeal communities.
We find, especially at the genus level, that parallel reactors clustered closely together (CD4a and b, CD5a and b, and CD11a and b; Figs 4 and S4). This tight clustering shows that the samples from the parallel reactors had similar microbial patterns as expected, because these couples (CD4a and b, CD5a and b, and CD11a and b, respectively) were running concurrently and under the same conditions. This confirms that our sampling procedure yielded representative patterns for the microbial communities in the reactors.
Earlier studies often focused on changes in one digester over time or compared process parameters, for example incubation temperature (Chouari et al., 2005; Levén et al., 2007; Zhang et al., 2009; Blume et al., 2010; Ike et al., 2010). However, Cheon et al. (2008) analyzed six full-scale anaerobic digesters and one laboratory-scale reactor using a random cloning method. Five of their sampled reactors were CSTRs: two of these digested kitchen waste at 37 and 55 °C, respectively; two, livestock waste (37 and 55 °C); and one, sewage sludge (50 °C). As in the present study, these authors found that the bacterial and archaeal community structures were affected by the substrate, as a change from sewage sludge to kitchen waste shifted the microbial composition. A relationship between the process temperature and bacterial community was also indicated. Regueiro et al. (2012) studied the bacterial community structure in relation to microbial activity in six mesophilic full-scale digesters and one laboratory-scale reactor using DGGE of PCR products from the amplification of the 16S rRNA genes of Bacteria and Archaea combined with FISH and specific activity tests. The Regueiro reactors included one treating sewage sludge; one each treating residues from brewery, dairy, sugar, and yeast industries, respectively; and two codigesters: one full-scale treating dairy and fish wastes, and one laboratory-scale treating slaughterhouse waste together with pig manure and glycerin. A cluster analysis of the bacterial and archaeal sequences obtained showed that reactors treating similar substrates grouped together, which is also supported by the results of the present study. As previously mentioned, Lee et al. (2012) found that the bacterial community composition was mainly related to the process temperature, which is in agreement with the present results Three of the reactors were run under thermophilic conditions and four under mesophilic. The authors found that the bacterial community composition of this set was mainly related to the process temperature. This is in agreement with the present results, especially regarding the CD biogas reactors, which clustered into thermophilic and mesophilic groups. It should also be noted that all reactors in the Lee study treated the same type of substrate as ours, with the exception of the night soil reactor. However, in another study, Leclerc et al. (2004) mapped the archaeal populations of 44 anaerobic digesters but found no relation between the archaeal communities and the type of substrate digested. These authors instead determined that the distribution of different Archaea was partly correlated to the process type applied, for example, CSTR, upflow anaerobic sludge blanket, and fixed bed.
The bacterial community's dominance and greater diversity compared with the archaeal community found in our study are in agreement with earlier studies (Godon et al., 1997; Fernández et al., 1999; McHugh et al., 2004; Cardinali-Rezende et al., 2009; Patil et al., 2010; Regueiro et al., 2012). The higher richness found in the CD compared with the CDT reactors is in agreement with results by Levén et al. (2007) and Pycke et al. (2011), yet Lee et al. (2012) did not find a significant difference in bacterial diversity between meso- and thermophilic conditions.
In general, the pH, ammonium, and VFA levels were higher in the CD reactors, which typically operated at higher OLRs and longer HRTs than in the SS reactors. In addition, the digestion in the CD reactors resulted in higher methane yields per kg VS added compared with the SS reactors, which is mainly a function of the more energy-rich substrates used in these reactors. The substrate and/or other process parameters of the CD reactors seemed to promote Firmicutes, because sequences belonging to this phylum dominated (70% of all sequences compared with 25% in the SS; Table S3). The dominance of Firmicutes in the CD reactors may also be related to the hygienization of the substrate (70 °C for 1 h) performed at most of the CD plants, as this treatment will likely promote the occurrence of sporulating organisms.
The higher richness indicated in the SS reactors is related to a higher proportion of unclassified Bacteria and several phyla, which are not observed at all or only at low levels in the CDs (Tables S3 and S4). One reason for this difference is likely that primary and activated sewage sludge make up the main part of the substrate for the SS reactors and that the Bacteria inherent in these two sludge types contribute to the community structures observed.
The most abundant sequences at the class level originated from Clostridia (belonging to Firmicutes) in all reactors. This highly versatile class represents organotrophs, including hydrolytic strains capable of degrading proteins, lipids, and polymeric carbohydrates (c.f. Lynd et al., 2002), so their dominance is not surprising. Clostridia have also been identified in earlier studies of biogas-producing microbial communities (Levén et al., 2007; Lee et al., 2008; Schlüter et al., 2008; Cardinali-Rezende et al., 2009; Wang et al., 2009; Patil et al., 2010).
On the class level, the second most frequent phylum, Bacteroidetes, was mainly represented by sequences belonging to Sphingobacteria in the CD reactors and by Bacteroidetes sequences in the SS reactors, while Flavobacteria sequences were more equally distributed. These three classes contain mainly saccharolytic chemoorganotrophic heterotrophs (Garrity et al., 2005) and have been observed in a number of anaerobic processes (Chouari et al., 2005; Ariesyady et al., 2007; Levén et al., 2007; Rincón et al., 2008; Schlüter et al., 2008; Cardinali-Rezende et al., 2009; Lee et al., 2009; Wang et al., 2009). In addition, Firmicutes and Bacteroidetes, together with Chloroflexi and Proteobacteria, were identified as the dominating phyla in a study analyzing the diversity in anaerobic digesters from sequences available in public databases (Nelson et al., 2011). This is to some extent in agreement with the results of the present study, in which Proteobacteria sequences were detected in all reactors, however, never as the dominating phylum. The Chloroflexi sequences occurred mainly in the SS processes (Table S3). Bacteroidetes, Firmicutes, and Proteobacteria were also the three main phyla detected by Regueiro et al. (2012) and Lee et al. (2012).
Levén et al. (2007) studied the effect of process temperature on the AD of organic household waste and reported a dominance of Thermotogae and Clostridia in their thermophilic reactors, while Bacteroidetes and Chloroflexi were the main phyla in the mesophilic reactors. Their results are in agreement with ours for the thermophilic reactors (apart for CD3T), but their mesophilic community seems more related to those found in our SS reactors. In a study investigating the microbial community of seven anaerobic sludge digesters, Rivière et al. (2009) identified organisms from the Betaproteobacteria (class level), Anaerolineales (order level), Bacteroidetes (phylum level), and Synergistetes (phylum level) as ‘core’ organisms of SS processes. These results are also in accordance, to some extent, with the results obtained for the SS reactors in the present study. However, while sequences belonging to Anaerolineae (Chloroflexi) and sequences from classes of Proteobacteria (Alpha, Beta, Delta, and Gamma) are characteristic of the SS reactors, sequences of the class Bacteroidetes were present in all reactors investigated except for the thermophilic CD digesters (Table S9).
It can be suggested that there are different pathways for methane formation in the SS and CD reactors based on the distinction between them at the archaeal genus level: the dominance of sequences belonging to the acetoclastic Methanosaeta sp. in the SS reactors (except SS5) vs. the hydrogenotrophic Methanoculleus sp. and/or Methanobrevibacter sp. in the mesophilic CDs (including SS5). The ratio between Bacteria and Archaea also shows a lower presence of Archaea in the CD reactors compared with the SS reactors. The fact that no Archaea were detected in CD1 and CD7b is likely a process ramification: pyrosequencing will detect only those fragments present at concentrations above a certain threshold, which is dependent on the sequencing depth of the sample.
Archaeal vs. bacterial community ratios have been reported earlier at 4–8% in laboratory-scale biogas reactors fed with a synthetic substrate based on cellulose and egg albumin (Sundh et al., 2003), 6% in a reactor fed with fodder beet sludge (Klocke et al., 2007), 6% in a household and garden waste reactor (Goberna et al., 2009), and at 15% of the prokaryotic cells in a full-scale municipal solid waste reactor (Cardinali-Rezende et al., 2012). In contrast, Montero et al. (2008) observed as much as 40% of sequences belonging to the Archaea in a laboratory-scale thermophilic dry anaerobic reactor fed with synthetic solid waste. Additionally, Regueiro et al. (2012)reported a broad spectrum with 0–40% Archaea from their seven reactors. In the present study, the SS reactors showed ratios of 2–20%, the CDT reactors 1–4%, and the CD reactor 0–7%. Thus, different conditions seem to yield large variations in the ratios of bacterial and archaeal sequences in AD processes, but neither reasons nor general patterns can be determined.
It should be noted that there is an almost complete lack of Methanosarcina sp. sequences in the data set even though the primers include this genus. This indicates a rather low abundance of these organisms compared with other methanogens, but does not necessarily imply that they are absent in the reactors.
Altogether, the sequencing results substantiate acetoclastic methanogenesis (performed by Methanosaeta sp.) as the main pathway of acetate utilization in the SS reactors, apart from SS5. The absence or low abundance of acetoclastic methanogens in the CD reactors might be a consequence of the relatively high levels in these reactors (c.f. Schnürer et al., 1999; Angenent et al., 2002; Schnürer & Nordberg, 2008). Under such conditions, syntrophic acetate oxidation (SAO) to hydrogen and carbon dioxide has been shown to take over (Schnürer & Nordberg, 2008). That this pathway might take place in the CD reactors investigated is supported by the high abundance of Methanoculleus sp. sequences in reactors CD10, CD11ap, and CD11 bp. These species are reported to act as hydrogenotrophs associated with SAO (Schnürer et al., 1999). In addition, several of the CD reactors included in the present investigation have been shown to be dominated by SAO by L. Sun, B. Müller, M. Westerholm, and A. Schnürer, (manuscript in preparation), thus supporting the results obtained from our 454-pyrosequencing analysis. However, no sequences belonging to the so far identified SAO organisms (Westerholm et al., 2011) were detected in our samples by a search in MG-RAST (Meyer et al., 2008) and Silva SSU database (Pruesse et al., 2007). This indicates either that Bacteria still not identified as SAO are responsible for the acetate oxidation in the herein studied CD reactors or that the microorganisms performing SAO were not possible to target with our general primers.
The dominance of sequences from Methanobrevibacter sp. in samples from CD3T, CD4a and b, CD5a and b, and CD7a might be linked to the addition of manure in the substrate for these reactors (Table 1 and Fig. 2). It should also be noted that CD3T is separated from the rest of the thermophilic CD reactors (Figs 2 and S2). The mesophilic CD reactors not receiving manure were instead dominated by Methanobacterium sp. and/or Methanoculleus sp. and the thermophilic CD reactors by Methanobacterium and Methanothermobacter. Earlier studies on archaeal populations under thermo- and mesophilic conditions do not suggest any clear relationship between specific genera and process temperature (Hori et al., 2006; Ariesyady et al., 2007; Levén et al., 2007; Rivière et al., 2009; Ike et al., 2010), yet our results reflect both the effect of temperature on the methanogenic community and the importance of the substrate's direct and/or indirect effects on the variation in diversity.
Differences in microbial community structures
The separation within the SS reactors at the microbial phylum, bacterial class, and archaeal genus levels is most likely related to differences in the process parameters. For Bacteria, the separation of SS5 is most likely linked to the long HRT of SS5 (45–55 days) compared with 10–28 days for the other SS reactors (Table 1). According to Westerholm et al. (2011), long HRTs are needed for the growth of the SAO consortia, which might explain the inability to detect sequences of acetoclastic methanogens in SS5. This suggests that SAO is the dominating acetate pathway leading to methane formation in this reactor (c.f. discussion above). Additionally, the variation in bacterial composition in SS3b compared with SS1, SS2, SS3a, and SS4 might be related to the HRT, as SS3b has a shorter retention time (10 days) compared with the other mesophilic SS reactors (16 days or longer, Table 1). The clustering of SS2T with the mesophilic SS cluster indicates that the influence of the substrate on the bacterial and archaeal communities in this case is more important than the process temperature. The clustering of SS1 within the SS cluster throughout the analyzed phylogenetic levels is a bit surprising as its main substrate is food industry waste (62%) with primary and/or activated sludge making up for the rest. Most likely, the inherent microbial communities of the primary and the activated sludge outweigh the possible influence of the food industry waste.
The similarity in the patterns of the biplots at the genus level for Bacteria and Archaea indicates a covariation between groups of Bacteria and Archaea. This is likely a manifestation of the microbial populations cooperating and interacting during AD. For instance, the interspecies hydrogen/formate transfers between proton-reducing Bacteria and methanogens allowing for the transformation of fermentation products to acetate and hydrogen should result in such patterns.
To conclude, SS and CD reactors clearly varied in their bacterial and archaeal community compositions. These two reactor groups differed in substrate composition, and in general, pH, ammonium, and VFA levels were higher in the CD compared with the SS reactors. At the phylogenetic phylum level, Actinobacteria, Proteobacteria, Chloroflexi, Spirochetes, and Euryarchaeota sequences mainly characterized the SS reactors, while Firmicutes sequences were most prevalent in the CD reactors. The presence of Thermotogae sequences was linked to the thermophilic processes, with the exception of CD3T. A higher bacterial diversity was found in the reactors treating sewage sludge. The tight clustering of the CD reactors was, however, surprising given the large differences in substrate mixes and process parameters among these reactors (Table 1). However, the hygienization pretreatment may be an important factor behind this result, as it likely increases the abundance of sporulating Bacteria in these reactors.
The PCAs at the bacterial and archaeal genus levels formed three main clusters: SS reactors, mesophilic CD reactors (including one thermophilic CD reactor and one SS reactor), and thermophilic CD reactors. This shows that the substrate has a strong influence on the microbial composition and, furthermore, that the process temperature affects the community structure. Sequences belonging to acetoclastic methanogens (Methanosaeta sp.) were only found in the SS reactors. Their absence or low abundance in the CD reactors indicates that the main pathway for acetate-based methane formation in these reactors takes place via SAO even though no known SAO was found.
The authors would like to thank the personnel at the biogas plants for assistance during sampling and for providing data on process parameters. Thanks are also given to Riley Töörn for linguistic correction of the manuscript. The research was funded by the Swedish Energy Agency.