This study found that unstable environmental conditions (in R3) resulted in the development of a microbial community rich in specialist taxa (Fig. 6), with an abundance of two component regulatory systems (Fig. 8). This led to reduced resistance to a shock load, which manifested as a large shift from stochastic community assembly to deterministic community assembly (Fig. 4) and ultimately, failure of the system. In contrast, a stable environment (in R1) led to the proliferation of generalist taxa (Fig. 6) with MAGs containing less abundant two component regulatory systems (Fig. 8), which increased shock resistance and reduced the effect of deterministic community assembly. The failure of R3 was surprising as the microbial community at T8 prior to the application of the final shock load had a significantly higher alpha diversity (Fig. 1) than R1 or R2. This is contrary to a lot of research which demonstrates that increased diversity leads to increased stability, resistance and functioning in other anaerobic bioreactors treating different types of wastewaters24,25, but also other environments including the human gut 38 and soils 39–41, thus indicating that it is not diversity itself which imparts resistance, but some underlying ecological property of the community. In addition to increased diversity, there was an increase in the number of KEGG modules detected in MAGs assembled from R3 (Fig. 2), but a significant decrease in evenness (Fig. 2). The increased abundance of TCRS in R3 also suggests that it should have been better able to respond to a shock (Fig. 8). However, this was not the case and therefore, it is important to investigate the properties of the R3 community ecology and functional potential, which were altered by its unstable operational-history and led to reduced resistance to an organic shock load.
4.1 Unstable environment leads to variable community composition
The communities of R1 and R2 remained more stable than R3 throughout the course of the trial and community beta diversity fluctuated less in R1 and R2 than in R3 (Fig. 2), indicating that the unstable environment of R3 led to more changes in microbial community composition. Interestingly, beta dispersion analysis indicated that there was no significant variation between the reactors in inter-sample variability when only abundance was considered, i.e. bray-curtis (Fig. 2). However, when phylogeny (unifrac) was used to assess beta dispersion (Fig. 2), significant differences were found between R1 and R3, indicating that the unstable conditions in R3 led to more phylogenetic dispersion in temporal samples in R3 over the course of the trial.
A large proportion of the temporally stable community of R3 was made up of the fermentative Spirochaetaceae family (Midas_g_14041) (Fig. 3). Since this analysis included all timepoints (including after the shock load), only taxa which were unaffected by the shock were thus identified including Streptococcus sp., Methanolinea sp., Paludibacter sp., and Desulfovibrio sp., as well as members of the Anaerolinaceae family (Midas_g_667) and the order Cloacimonadales (Midas_g_71685) in R1. The stable community of R2 also included Methanolinea sp., and members of the family Cloacimonadaeae (Midas_g_63301) in addition to Smithella sp. and Methanomassiliicoccus sp.,. The stable community of R3 included Leptolinea sp, genera from the Bacteroidetes_vadinHA17 family (Midas_g_19) and the order Syntrophales (Midas_g_134) along with the afformentioned Spirochaetaceae family (Midas_g_14041).
The indication that phylogenetic composition (unifrac distance metric) was more variable between sampling timepoints in R3 is interesting. As the Unifrac distance metric does not consider abundance, it is a particularly useful metric for excluding the effect of the rare biosphere in diversity analysis 42. The greater phylogenetic dispersion between sampling timepoints in R3 therefore were not a result of fluctuations in rare taxa. However, rare taxa have been repeatedly suggested to play a role in resistance to environmental disturbances 43,44 and functional redundancy. Therefore, the composition of the rare biosphere was analysed in further detail, see section 2.5.
4.2 Distribution of Metagenome Assembled Genomes
Over the course of the trial, several MAGs increased in relative abundance in all three reactors, especially bin 57 of the genus Tidjanibacter, which was also the 4th most abundant bin overall (Fig. S2). Tidjanibacter species are members of the Rikenellaceae family and were first isolated from the human colon 45, although there have been suggestions that this species should be classified within the genus Alistipes 46. Tijanibacter have also been found in the digestive systems of chickens 47 and mice 48 but to the best of the authors knowledge have not been identified in anaerobic digesters. Three members of the Patescibacteria phylum (bins 69, 198 and 348), also increased in abundance in all three reactors (Fig. 7), but particularly in R1 and R2. Bin 348 was classified to order level as Moranbacteriales and was especially in R1 and R2, where its relative abundance was 7% and 5% respectively compared to < 1% in R3. The Patescibacteria are an uncultivated superphylum, which seem to be ubiquitous in many environments, but very little is known about their potential functions, other than that they have a small genome size and are thought to often be symbiotic with other microorganisms 49–51, including methanogens of the Methanothrix genus 52.
4.3 Functional Changes in Reactor Systems
The functional composition, in terms of KEGG modules annotated in recovered MAGs from all reactors clustered based on timepoint, with T1 samples clustering entirely separately from T8 samples in all reactors (Fig. 2). Individual reactors were also more distinct at T8, indicating a divergence in the community function of each reactor depending on it’s operational history (Fig. 2). There was a significant reduction in the richness of KEGG modules associated with MAGs recovered from R1 and R2, but no change in KEGG module richness in R3. At the same time the evenness of KEGG modules decreased in R3 but there was no significant change in R1 or R2. The decrease in functional evenness and increase in richness in R3 indicates a community which had become dominated by the same functions, which were perhaps more suited to the unstable environment.
In terms of functional change in the recovered MAGs, the most striking distinction between the reactors was in TCRS (Fig. 8), where R3T8 clustered separately from the other samples (Fig. 8). TCRS are the molecular mechanisms which allow microbes to sense and respond to environmental stimuli by (de)activating the transcription of associated genes (Capra and Laub, 2012). Two component systems typically consist of a membrane bound histidine kinase protein which senses environmental stimuli and a response regulator protein, which mediates differential gene expression (Stock et al., 2000). Environmental stimuli could include the presence or absences of toxins and nutrients or physical characteristics such as redox state, pH and osmotic pressure (Jacob-Dubuisson et al., 2018). Therefore it stands to reason that a microbial community which has had a more variable environment (i.e. R3) should have more TCSs. Indeed, 82 two component regulatory systems were more abundant in R3T8 than in R1T8 (Fig. 8). Several of the TCRSs which were more abundant in R3T8 have previously been associated with biofilm formation e.g. LuxQN/CqsS − LuxU − LuxO (Murugesan et al., 2023), VicK − VicR (Sun et al., 2023) and CiaH − CiaR (Wu et al., 2010). This indicates that the unstable environment in R3 may have led to increased biofilm forming capacity. Indeed, the biomass dynamics in R3 differed from R1 or R2 23 by producing much more biomass over the course of the trial (Fig. S3).
TCRS which were more abundant in R1T8 included ArcB − ArcA (anoxic redox control) and HupT − HupR (hydrogenase synthesis regulation). The Arc two component system is utilised by facultative aerobes to detect and react to changes in respiratory growth conditions 53 and regulates transcription of genes related to anaerobic growth in E.coli 54. HupT-HupR TCRS has been mostly studied in Rhodobacter species, where it functions in conjunction with HupUV sensor protein in detecting and responding to H2 concentrations by producing hydrogenases, enabling a switch to autotrophic growth 55,56. Therefore, R1 may have been more suited to dealing with changes in H2 concentrations and switching to hydrogenotrophic metabolism.
The contribution of these variations in TCRS to the failure of R3 is unclear and in fact their abundance may have had no impact at all on resistance (or the lack of). However, their predominance is indicative of the selection pressures of the applied disturbances, which has implications for further studies. Molecular diagnostics have long been suggested to be the next step in process monitoring of anaerobic digestion 57,58, but no suitable target genes have so far been identified. The increase in the presence of TCRS in MAGs recovered from R3 indicates that they may be suitable for this purpose, particularly if detected at the transcriptomic level rather than at the genomic level. This would require further, similar studies where expression can be strongly linked to a given environmental disturbance. Once that is achieved, transcriptional screening could be carried out to assess community stress responses prior to process failure.
4.4 Distinct environmental disturbances promotes the establishment of specialist taxa
The absence of any conditionally rare taxa in R3 is surprising as environmental disturbances have previously been shown to promote temporal fluctuations in the rare biosphere 17,43 and lead to the development of conditionally rare taxa 59. In addition, our analysis indicated that persistently rare taxa made up a lower proportion of the total community in R3. Therefore, it may be the case that although these rare taxa did not increase enough in abundance to meet the threshold of conditionally rare taxa, they did increase to some extent, thus resulting in a lower proportion of persistently rare taxa. This also explains the observation that the number of other rare taxa were the highest in R3 (Fig. 5).
The taxa which shifted from persistently rare to other rare taxa may have been specialised to deal with a given disturbance (e.g. ammonia), but due to the short nature of the applied disturbances did not establish enough to become conditionally rare. Indeed, the unstable operation of R3 did lead to the proliferation of a microbial community with a higher proportion of specialist taxa than the control reactor (R1) (Fig. 6). The prevalence of specialist microbes in R3 was likely caused by the variety of applied disturbances, which led to the development of a larger number of narrow niches and less overlap (Fig. S6) throughout the reactor operation. R2 had a similar number of specialist and generalist taxa to R1, indicating that the organic load increases in R2 did not lead to a proliferation of specialist microbes.
Generalist taxa have a wider niche breadth, and are able to survive under a wider set of conditions than specialists 60. In contrast, specialists are more discriminant in the environments in which they thrive 37. The relative proportion of generalists and specialists in a microbial community is thought to influence its stability in the face of a disturbance 61. Muller et al (2019) hypothesised that communities with a high proportion of specialist taxa would be more susceptible to functional breakdown in the face of a disturbance as the population performing the dominant function has a narrower niche breadth and will not be able to resist the change in conditions brought on by a disturbance. The present study provides evidence for this hypothesis in that the more specialist community of R3 failed in response to a disturbance, whereas the generalist communities of R1 and R2 were able to resist the applied shock.
The observation that different microbial life strategies evolved in the respective reactors and subsequent divergent responses of each reactor to a large disturbance indicates that a community made up of generalist microbes is more resistant to large disturbances. The proliferation of generalists was enhanced in R1 due to the stable environment. This is unexpected as generalist species are often thought to gain an advantage from environmental heterogeneity 13. However, the distinct nature of each disturbance applied to R3 (i.e. pH, NH3, temperature and sulfate) may have led to the establishment of microbes which were specialised to deal with them.
Much more taxa overlap was observed in R1 and R2 (Fig. S4, Fig. S5) indicating that more interspecies interactions may have developed, potentially as a result of more consistent environmental conditions. Indeed a recent meta-analysis including pan-genomic data indicated that genomes of generalists are enriched in functions involved with species-species interactions 62. The extent of interspecies interactions in a microbial community can impact its resistance to a disturbance as key interdependencies, e.g. sharing of public goods 63 are buffered from sudden taxa die-off and ecosystem function is less reliant on a single keystone species, which may be inhibited by a disturbance 64,65. It has also been suggested that lower numbers of interspecies interactions reduce ecosystem stability 66 and increased interspecies interactions have even been shown to increase robustness in methanogenic communities 67. Therefore, despite lower diversity, the generalist lifestyle which prevailed in R1 and R2 was crucial in improving their resistance to a major organic carbon influx.