Chemical parameters
Eleven parameters were measured for each of the reactor samples: COD (chemical oxygen demand), TOC (total organic carbon), total nitrogen content (N), electrical conductivity, TVFA (total volatile fatty acids), TS (total solids), VS (volatile solids), pH, biogas yield, and concentrations of CH4 and CO2 (Additional file 1: Table S1). Biogas yields were obtained from lab-scale batch experiments, whereas all the other parameters originated from in situ measurements of digester samples. Batch experiments were performed without adding substrates and obtained biogas yields depended only on the organic fraction within the sludge samples.
After normalizing the data, successive combinations of three parameters (permutation) were plotted in a Gnuplot multiplot (Fig. 2). The resulting data matrix included biogas production but not methane and CO2 concentration, in order to avoid redundancies. This resulted in three clearly defined clouds, each corresponding to one of the different digester facility types (Fig. 2a). SS and CD values were plotted in two opposed vertices of the plot, with LB located in an intermediate position. The yield of biogas produced is shown in Fig. 2b and the highest yields are plotted as a relatively small cloud (black dots) overlapping with the extremes of the CD cloud. As a general conclusion, parameter values were higher (corresponding in general with high nutrient contents) when biogas production was highest. In a second statistical approach, this observation was verified by a principal component analysis (Additional file 2: Figure S1), where samples coming from the same type or reactor clustered together and notably differed from those from other reactor types.
Taxonomic composition of eubacteria
Eubacteria from all samples were identified by high-throughput sequencing as described in “Material and methods” section, and phylum-level results are shown in Fig. 3. There was little variation between replicates, clearly indicating that differences in taxonomic composition accounted for the differences found between reactors and time. Similarly, different sampling times resulted in very small variations in the taxonomic profile, being the taxonomic composition of each sample very constant after 1 week. Only in one case (LB reactor in Saalfeld) that a substantial shift was detected in the amount of Bacteroidetes and Spirochaetes after 1 week.
The taxonomic composition of the samples correlated closely with reactor type. Indeed, three different profiles were observed, each corresponding to a particular facility type. CD samples were dominated by the phylum Firmicutes, with nearly 46–60 % of classified sequences assigned to Firmicutes in the first two stages and less than 20–32 % in the third stage (remnant storage); followed by Bacteroidetes, which proved mainly in the third stage, when it accounted for up to 73 % of the total identified taxa. The three CD digesters contained low amounts of Synergistetes, and the remnant storage contained moderated amounts of Actinobacteria, Proteobacteria, Spirochaetes, and Tenericutes (Fig. 3a).
The second facility type (LB) displayed a totally different microbial composition (Fig. 3b) with comparatively fewer Firmicutes reads (between 3 and 19 % of total sequences). The microbial LB communities were dominated by Spirochaetes (30 and 72 % of the total reads), along with Bacteroidetes (11 and 47 %). The third phylum, Thermothogae, reached low to moderate frequencies in LB facilities in Schmölln and Saalfeld (between 2 and 19 %), and it was absent in the six replicates of Schlossvippach. Minor counts of Actinobacteria and Proteobacteria were also detected. The third profile was associated with the sewage sludge digesters (Fig. 3c). Although the SS facilities showed certain similarities compared to the LB facilities, the overall microbial composition differed from both CD and LB reactors. In common with the LB samples, SS reactors contained high amounts of Bacteroidetes and Spirochaetes (Bacteroidetes between 13 and 51 %, Spirochaetes between 27 and 50 %). However, unlike the CD and LB facilities, SS reactors were particularly rich in Chloroflexi (9 and 39 %) and Proteobacteria (4–9 %). Besides the aforementioned taxa, small amounts of Actinobacteria, Synergistetes, and Thermotogae were also observed.
Minor variations or sub-profiles of the three main biomass-associated profiles were detected. For example, two of the three Jena CD reactors were very similar, while the third one displayed higher eubacteria diversity. This might be due to the fact that the last stage (remnants) was kept at RT instead of mesophile temperatures. Although LB and SS samples corresponded to two main profiles, one location of each type (LB-Schlossvippach and SS-Rudolstadt) exhibited a characteristic presence/absence of one particular taxon: the former typically lacked Thermotogae, which was well represented in the other two LB plants; while SS Rudolstadt was particularly rich in Chloroflexi (Fig. 3b, c). The absence of Thermotogae in the LB reactor from Schlossvippach may be due to the fact that the solid phase is mainly heated up by the leachate (without extra heating in the solids storage—“garage”), which can lead to irregularities in temperature. In the Schlossvippach sample, it took more than 1 week to heat up a newly filled garage (Christoph Bürger and Kevin Lindner personal communication).
In general, taxonomic eubacteria profiles strongly correlated with the biomass type. The differences observed between CD and SS reactors are in accordance with previous studies [13] describing an overall difference between sewage sludge and co-fermentation regarding the microbial profile. The high amount of Bacteroidetes and Firmicutes in CD reactors is also consistent with previous reports [13, 18, 22, 24]. One reason for the abundance of Firmicutes could be the high content in TS derived from plant material (Additional file 1: Table S1), which probably fosters biofilm formation. Firmicutes have been described as main degraders of cellulolytic material [24] and are abundant in biofilms of water supply systems [25, 26]. LB and SS reactors, both containing liquid substrates, had high titers of the very mobile and efficient swimmer Spirochaete, described as able to swim in high viscous gel-like liquids, such as those found in LB reactors [27]. It has to be highlighted that the observed microbial profiles for the LB samples were only those from leachate, and that the solid fraction of LB systems might be rich in Firmicutes due to the high percentage of solids. The abundance of Chloroflexi in SS reactors has previously been reported. In fact, different Chloroflexi species have been found in more than 60 sewage reactors in different European countries based on FISH experiments [28] and also in other facilities around the world [29]. The prevalence of Proteobacteria and Bacteroidetes is in accordance with the work by Wang et al. [30] on the microbial profile of domestic sewage outfalls.
The different taxonomic profiles we found correlated to biogas yield. For instance, the phylum Chloroflexi was detected in sewage plants, where very low biogas yields were measured. Also, Proteobacteria were only found in the plants with low biogas yields (digestate storage of the three-stage plant, Schlossvippach, and all sewage samples), while Firmicutes were particularly abundant in reactors with high biogas yields (CD samples). However, differences in biogas yield might also be a consequence of the concentration of TS, which is especially high in CD reactors.
In summary, our results are strongly consistent with previous reports demonstrating patchiness of the digesters in terms of the distribution of bacterial populations [31]. This strongly suggests ecological parameters (i.e. liquid/solid substrate or biomass type) are the key factors shaping microbial communities; but also reveal an important, albeit secondary, role of the facility/reactor on this mainly biomass-associated distribution of the taxonomic profiles.
Taxonomic composition of archaea
The taxonomic composition of the sampled reactors in terms of archaea contents is shown in Fig. 4. The data correspond to all but one reactor (three replicates and two time points), corresponding to the third stage of the Jena CD reactor, from which no archaeal DNA could be amplified. CD reactors were dominated by archaea belonging to the genus Methanoculleus (Fig. 4a), accounting for 59–76 % of all the sequences. A significant amount of Methanosarcina (9–24 %), Methanobacterium (10–21 %), and Methanobrevibacter (3–7 %) was detected, as well as infrequent genera such as Methanosphaera, Methanothermobacter, and Methanosaeta. In contrast, LB digesters were characterized by substantially smaller amounts of Methanoculleus (3–44 %); and by the abundance of Methanosarcina (37–95 %). One of the three LB-digesters showed a very high amount of Methanobrevibacter (31–35 %), whereas the other two reactors had very low amounts (1–2 %). Minor genera were Methanobacterium, Methanosphaera, and Methanosaeta. In the SS samples, Methanosaeta proved the most prevalent genus with a total number of reads between 42 and 88 % (Fig. 4c). While Methanosaeta was detected in high amounts in all the SS reactors, the frequency of other genera differed among SS digesters. The biogas plant in Rudolstadt was very rich in Methanomethylovorans (40–55 %), while the other two SS reactors showed a relatively high amount of Methanoculleus (1–10 %) and Methanospirillum (8–21 %).
As in the eubacteria profiles, three main taxonomic combinations were found to correlate with the three reactor types. The CD samples showed a strikingly similar profile independently on the replicate, reactor, or time sampled. LB and SS reactors did exhibit sub-profiles with no variation within replicates and dependent on the sampling time (Schlossvippach and Saalfeld) or on the location sampled (Rudolstadt). The two LB facilities from Schlossvippach and Saalfeld showed an increased amount of Methanoculleus after 1 week, while the amount of Methanosarcina decreased during this period. It is likely that genus Methanoculleus is more abundant in the solid fraction of these LB systems due to the high percentage of solids. Rudolstadt samples had the typical Methanosaeta abundance of SS reactors but were characterized by an exceptionally high frequency of Methanomethylovorans. The presence of Methanosarcina and Methanoculleus correlated to high yields of biogas, while low biogas yields correlated with higher amounts of Methanosaeta.
Since methane production is solely due to the archaeal community and the different methanogenesis pathways are well known and genus-linked, we studied the expected methanogenesis pathways in each facility type according to the average taxonomic distribution (Fig. 4d–f). Interestingly, each facility type displayed a different combination of methanogenesis pathways. The CD reactors were very rich in archaea using the hydrogenotrophic pathway (Fig. 4d); LB reactors were dominated by Methanosarcina and thus with the ability to use all known pathways for methane production (Fig. 4e); and SS reactors were characterized by containing high rates of archaea using the acetoclastic pathway for methane production (Fig. 4f).
The archaea composition we describe here for the different reactor types is generally in accordance with that reported in previous studies. The prevalence of Methanoculleus in CD reactors was also found in other works with classical anaerobic digesters [22, 32, 33]. Although other studies describe a prevalence of Methanosarcina in this reactor type [34], our data is in concordance with other works linking Methanosarcina to LB reactors [35, 36]. The differences in TS levels between CD and LB reactors might be the key factor explaining their differences in microbial composition. The TS content of LB reactors was much lower (Additional file 1: Table S1), so the surface available for the growth of biofilm-forming species, such as Methanoculleus [37], was limited compared to CD reactors. Indeed, previous reports have found a prevalence of Methanoculleus in the solid fraction of LB reactors [36, 38]. Additionally, a lower number of TS may hamper the formation of spatial syntrophic relationships between acetate-oxidizing bacteria and hydrogenotrophic methanogens such as Methanoculleus. This might lead to an increase in growth of acetoclastic methanogens such as Methanosarcina, able to directly metabolize acetate (Fig. 4d–f). These findings are in concordance with previous reports on the link between high content of TS and a high frequency of hydrogen-using methanogens compared to acetoclastic methanogens [39–42].
The finding that Methanosaeta is the dominating genus in all SS digesters is consistent with other screenings [21, 43, 44]. However, the abundance of Methanomethylovorans in the SS digester in Rudolstadt might be connected to the presence of particularly high amounts of oil and alcohols such as methanol, since this particular digester was supplemented with remnants from biodiesel production, and the prevalence of this organism has been reported in sewage sludge reactors supplemented with molasses alcohol wastewater [45].
The genus Methanospirillum was more abundant in the SS reactors in Jena and Weimar but not in Rudolstadt. This genus proved, along with Methanolinea, particularly abundant in a previous SS characterization [46], suggesting that Methanospirillum and Methanosaeta are two competing genera within the anaerobic digestion process of SS sludge.