Introduction

Production of methane by anaerobic digestion (AD) is widely used to convert organic waste into biogas and forms an important part of the transition from fossil fuel to sustainable energy production. The AD process is divided into four sequential steps that are performed by specialized microbes: hydrolysis, fermentation (acidogenesis), acetogenesis (dehydrogenation) and methanogenesis (acetoclastic or hydrogenotrophic) (Angenent et al., 2004). Hence, the overall function, stability and efficiency of the AD process are dependent on tightly coupled synergistic activities of the complex microbial communities (Schink, 1997; Weiland, 2010). However, the microbial communities in AD are still poorly understood, and relatively little is known about their diversity and function (Chouari et al., 2005; Werner et al., 2011; Sundberg et al., 2013; De Vrieze et al., 2015). In addition, most of the microorganisms have no pure culture representatives, and, given the synergistic interactions of members of the community, a reductionist approach to understand the ecology of the system is not possible (Kaeberlein et al., 2002; Fuhrman et al., 2015).

The AD environment also harbours extensive diversity of previously uncharacterized bacterial phyla, often known only by their 16S rRNA gene sequence, making it an ideal environment for the study of novel bacterial lineages (Guermazi et al., 2008; Pelletier et al., 2008; Limam et al., 2014; Sekiguchi et al., 2015). New developments in single-cell genomics and metagenomics have in recent years provided a glimpse into the ecology and evolution of many novel candidate phyla (Dinis et al., 2011; Albertsen et al., 2013; Rinke et al., 2013; Brown et al., 2015; Nobu et al., 2015; Sekiguchi et al., 2015). The genomes have enabled construction of metabolic models that attempt to explain the physiology of these organisms in detail. The genome-based models form the basis of more extensive investigations, such as in situ single-cell characterization, metatranscriptomics and proteomics (Koch et al., 2014).

In this study, extensive 16S rRNA gene amplicon sequencing was used to screen anaerobic digesters for the presence of members of the Hyd24-12 lineage, which remains one of the few known candidate phyla for which no genomic information is available with nothing known about their morphology, physiology or ecology (Rinke et al., 2013). Selected samples were subjected to metagenome sequencing and used for retrieval of three near-complete genomes of Hyd24-12 through differential coverage binning. The genomes were used for detailed metabolic reconstruction and design of oligonucleotide probes for the first in situ visualization of these hitherto unrecognized players in AD.

Materials and methods

Sample collection and storage

A total of 306 biomass samples were obtained from 29 anaerobic digesters at 17 Danish wastewater treatment facilities (see Supplementary Table S1). Most digesters were mesophilic (22), whereas 7 were thermophilic. A volume of 50 ml was sampled, homogenized and stored as 2 ml aliquots at −80 °C for DNA extraction.

For fluorescence in situ hybridization (FISH) analyses, diluted biomass samples (1:4 in 1 × phosphate-buffered saline) were fixed with 4% (w/v) paraformaldehyde and stored in 50% (v/v) ethanol/1 × phosphate-buffered saline solution at −20 °C, as previously described by Daims et al. (2005).

DNA extraction

DNA was extracted from anaerobic digester sludge using the FastDNA Spin kit for soil (MP Biomedicals, Santa Ana, CA, USA), following the standard protocol except for four times increased bead beating duration and a sludge input volume of 50 μl. These digester-sample-specific modifications to the protocol were found to provide the best trade-off between DNA yield/biomass and DNA integrity (Supplementary Figure S4).

Community profiling with 16S rRNA gene amplicon sequencing

Bacterial community profiling was carried out as recommended by Albertsen et al. (2015). The bacterial primers used were 27F (AGAGTTTGATCCTGGCTCAG) (Lane, 1991) and 534R (ATTACCGCGGCTGCTGG) (Muyzer et al., 1993), which amplify a DNA fragment of ~500 bp of the 16S rRNA gene (variable V1–V3 region). PCR amplification was performed using 1 × Platinum High fidelity buffer, 400 pm dNTP, 1.5 mm MgSO4, 2 mU Platinum Taq DNA Polymerase High Fidelity, 5 μm illumina barcoded V1–V3 adaptor mix (see Supplementary Data 1), and 10 ng template DNA. PCR conditions were 95 °C for 2 min, 30 cycles of 95 °C for 20 s, 56 °C for 30 s, 72 °C for 60 s, and a final step of elongation at 72 °C for 5 min. PCR products were purified using Agencourt AmpureXP (Beckman Coulter, Brea, CA, USA) with a ratio of 0.8 bead solution/PCR solution. Barcoded amplicons were pooled and paired-end sequenced on the Illumina MiSeq platform (v3 chemistry, 2 × 300 bp). The paired-end reads were trimmed using trimmomatic (v. 0.32) (Bolger et al., 2014) and then merged using FLASH (v. 1.2.11) (Magoč and Salzberg, 2011). The reads were screened for potential PhiX contamination using USEARCH (v. 7.0.1090) (Edgar, 2010). The reads were clustered into operational taxonomic units (OTUs, sequence identity ⩾97%) using USEARCH and subsequently classified using the RDP classifier (Wang et al., 2007) with the MiDAS database (v. 1.20) (McIlroy et al., 2015). Further processing was carried out in the R environment (v. 3.1.2) using the ampvis package (Albertsen et al., 2015) (v. 1.24.0), which wraps a number of packages including the phyloseq package (v. 1.8.2) (McMurdie and Holmes, 2013), ggplot2 (v. 1.0.1), reshape2 (v. 1.4.1) (Wickham, 2007), dplyr (v. 0.4.2), vegan (v. 2.3-0), knitr (v. 1.10.5), Biostrings (v. 2.36.1), data.table (v. 1.9.4), DESeq2 (v. 1.8.1) (Love et al., 2014), ggdendro (v. 0.1–15) and stringr (v. 1.0.0). The samples were subsampled to an even depth of 10 000 reads per sample, and the fraction of reads classified as Hyd24-12 was obtained. The survey data are available at the SRA with the accession IDs ERS861217-ERS861224.

In silico analysis of Hyd24-12 source locations

The Genbank IDs of the sequences classified as Hyd24-12 in SILVA (v. 121, 1982 sequences in total) (Quast et al., 2013) were used to download the corresponding Genbank files. The fields ‘isolation source’ and ‘PUBMED’ were extracted to classify the sequences as originating from either engineered or natural systems.

Metagenome sequencing, assembly and binning

Illumina TruSeq DNA PCR free libraries were prepared for DNA extracts from three of the mesophilic digesters (Supplementary Table S1) according to the manufacturer’s protocol and paired-end sequenced on the Illumina HiSeq 2000 platform (2 × 150 bp) and Illumina MiSeq platform (v3 chemistry, 2 × 300 bp). The metagenomic assembly and binning process was carried out as described by Albertsen et al. (2013) and detailed at ‘madsalbertsen.github.io/mmgenome/’. Unmerged reads were quality-trimmed and filtered using default settings in CLC Genomics Workbench (v. 7.5.1; CLC Bio, Aarhus, Denmark). The metagenomic reads were assembled separately for each plant using default settings in CLC Genomics Workbench. Reads were mapped to the assemblies using default settings in CLC Genomics Workbench. The assemblies and mappings were exported as .fasta and .sam files, respectively. The exported files and the mmgenome workflow script ‘data.generation.2.1.0.sh’ were used to generate the files necessary for the binning process. 16S rRNA gene sequences were extracted from the assemblies using ‘rRNA.sh’ and classified using the SINA Alignment service (SILVA v 121) (Pruesse et al., 2012); essential genes were called using Prodigal (Hyatt et al., 2010). Binning was carried out using differential coverage binning in the R environment (v. 3.1.2, R Core Team, 2016) using the R package ‘mmgenome’ (github.com/MadsAlbertsen/mmgenome v. 0.4.1) (Albertsen et al., 2013). The genome bins were checked for completeness, essential single copy genes and coverage distribution using CheckM (v. 0.9.7) (Parks et al., 2015) and the metrics in the mmgenome package. Average nucleotide identity between the genome bins was calculated using JSpecies (Richter and Rosselló-Móra, 2009), and CRISPR arrays were identified with CRT (v. 1.1) (Bland et al., 2007). The genome sequence data have been submitted to DDBJ/EMBL/GenBank databases under accession numbers LKHB00000000, LKHC00000000 and LKHD00000000.

Genome sequence-based phylogenetic analysis

The genomes were placed within the reference genome tree of CheckM (Parks et al., 2015) (v. 0.9.7) and subsequently visualized in ARB (Ludwig et al., 2004).

Phylogeny of the 16S rRNA gene and FISH probe design

Phylogenetic analysis and FISH probe design were performed with the ARB software package (Ludwig et al., 2004). Potential probes were assessed in silico with the mathFISH software (Yilmaz et al., 2011) for hybridization efficiencies of target and potentially weak, non-target matches (Yilmaz et al., 2011). Unlabelled helper probes (Fuchs et al., 2000) were designed for calculated inaccessible regions. Unlabelled competitor probes were designed for single-base mismatched non-target sequences (Manz et al., 1992). The Ribosomal Database Project (RDP) PROBE MATCH function (Cole et al., 2009) was used to identify non-target sequences with indels (McIlroy et al., 2011). Probe validation and optimization were based on generated formamide dissociation curves (Daims et al., 2005), where average relative fluorescent intensities, of at least 50 cells calculated with ImageJ software (National Institutes of Health, New York, NY, USA), were measured for varied hybridization buffer formamide concentrations in increments of 5% (v/v) over a range of 5–50% (v/v) (data not shown). Where available, weak base mismatch non-target axenic cultures were used for probe optimization, otherwise full-scale anaerobic digester sludge was used (Table 1).

Table 1 FISH probes designed and optimized in this study
Full size table

FISH

FISH was performed essentially as described by Daims et al. (2005). Probes were applied, with recommended competitors and helpers, at the stringency conditions given in Table 1 or their original publications. The NON-EUB nonsense probe was used as a negative hybridization control (Wallner et al., 1993). Oligonucleotide probes were labelled on both the 3′ and 5′ ends with either 5(6)-carboxyfluorescein-N-hydroxysuccinimide ester (FLUOS) or with the sulphoindocyanine dyes (Cy3 and Cy5) (DOPE-FISH) (Stoecker et al., 2010). Microscopic analysis was performed with an Axioskop epifluorescence microscope (Carl Zeiss, Oberkochen, Germany).

Genome analysis

Genome annotation was performed in the ‘MicroScope’ annotation pipeline (Vallenet et al., 2009, 2013). Automatic annotations were manually curated for all genes described using the integrated bioinformatics tools and the proposed annotation rules, which include an amino acid identity of at least 40% to classify homologues and an identity of at least 25% with the support of conserved domains to determine putative homologues (Vallenet et al., 2009, 2013). The set of bioinformatics tools includes BlastP (Altschul et al., 1990) homology searches against the full non-redundant protein sequence databank UniProt (Uniprot Consortium, 2014) and against the well-annotated model organisms Escherichia coli K-12 and Bacillus subtilis 168 (Vallenet et al., 2013), enzymatic classifications based on COG (Tatusov et al., 2003), InterPro (Mitchell et al., 2015), FIGFam (Meyer et al., 2009) and PRIAM (Claudel-Renard et al., 2003) profiles, and prediction of protein localization using the TMHMM (Sonnhammer et al., 1998), SignalP (Bendtsen et al., 2004) and PSORTb (Gardy et al., 2005) tools. Synteny maps (i.e. conservation of local gene order) were used to validate the annotation of genes located within conserved operons (Vallenet et al., 2009). Metabolic pathways were subsequently identified with the assistance of the integrated MicroCyc (Vallenet et al., 2009) and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases (Kanehisa et al., 2014).

Results and discussion

Survey of 16S rRNA genes of Hyd24-12 in anaerobic digesters

The survey of 22 full-scale mesophilic and 7 thermophilic anaerobic digesters from 17 Danish wastewater treatment plants over 3 years revealed that members of the Hyd24-12 lineage were stably present in most mesophilic but no thermophilic anaerobic digesters (Figure 1). In most mesophilic digesters, they were among the five most abundant bacterial OTUs and constituted around 1–3% and, in some cases, up to 8.2% of all sequenced bacterial reads (see Supplementary Figure S1). No 16S rRNA gene sequences from Hyd24-12 were detected in the incoming surplus sludge from the activated sludge treatment plants, which demonstrates that these bacteria were actively growing in the digesters. The other abundant bacterial phyla in the mesophilic digesters were Actinobacteria, Firmicutes, Chloroflexi, Synergistetes and Bacteroidetes (Figure 1). The best (LCA) classification is shown in Figure 1, but the lack of closely related organisms in the databases and a curated taxonomy hampers taxonomic classification for a number of the most abundant OTUs. In general, the abundance stability of these top genera was high, and that may be due to relatively similar growth conditions for all digesters: feed was primary sludge and surplus activated sludge, temperature in the interval 34–37 °C, pH 7.1–8.2 and total ammonium 0.57–1.1 g N/l (see Supplementary Table S1).

Figure 1
figure 1

Heatmap of the 25 most abundant bacterial OTUs in mesophilic digesters at wastewater treatment plants along with their abundance in thermophilic digesters at 17 wastewater treatment plants. The OTU classified as belonging to the Hyd24-12 candidate phylum (purple) was detected exclusively in mesophilic reactors. Classification levels presented are phylum, class, order, family and genus and are separated by a semicolon. The field is empty where no classification at a given level could be provided. The abundance profiles show mean abundances for plants with more reactors (1–4 reactors at each WWTP) and 2–97 samples for each plant over 4 years (Supplementary Table S1). The OTUs are sorted on the basis of the mean abundance across the mesophilic samples.

Full size image

In silico analysis of 16S rRNA gene sequences within Hyd24-12 of SILVA (Quast et al., 2013) from other surveys confirmed that members of Hyd24-12 are widespread in anaerobic environments. The sequences originate from 48 separate studies, with engineered systems such as anaerobic bioreactors accounting for 10 studies, and natural systems such as marine sediments, microbial mats in hydrogen, methane-rich waters and mud volcanoes accounting for 38 studies (see Supplementary References). Furthermore, the 48 studies show that members of Hyd24-12 are globally dispersed (Supplementary Figure S3 and Supplementary Table S2) and are potentially important in many microbial ecosystems besides ADs (Mills et al., 2005; Harris et al., 2012). Some of the surveys of full-scale anaerobic digesters detected some Hyd24-12 sequences (e.g., De Vrieze et al., 2015), while others did not (Sundberg et al., 2013). This was likely because they used the RDP database, where Hyd24-12 sequences are classified as ‘unclassified bacteria’.

Recovering genomic information from Hyd24-12

Three full-scale anaerobic digesters were sampled for metagenomic analyses. To ensure differential abundance of microorganisms needed to bin genomes based on coverage profiles (Albertsen et al., 2013), biomass samples were either taken from the sludge and foam layer of reactors or from the same reactor weeks apart. More than 50 gigabases of metagenomic data were generated, and population genomes were recovered by differential coverage binning (Albertsen et al., 2013) from each of the three plants (Table 2). The three population genomes were ~2.2 Mbp with a GC content of ~64%, and the completeness of the genomes were estimated by CheckM (Parks et al., 2015) to be between 86% and 91% with less than 2.2% estimated contamination (Table 2). However, the level of completeness may be underestimated, given that members of the Hyd24-12 are distantly related to other characterized organisms, and the genes used in the marker sets might be too divergent or simply not present (Rinke et al., 2013; Brown et al., 2015; Sekiguchi et al., 2015). The three genomes each contained a single rRNA operon and shared identical 16S rRNA gene sequences, which suggests that they belong to the same species (Yarza et al., 2014). The JSpecies program determined that these three genomes shared between 99.8% and 99.9% average nucleotide identity (ANIb), supporting the close taxonomic relationship observed from the 16S rRNA gene analysis (Kim et al., 2014). In order to further evaluate the similarity between the strains, the raw metagenome reads from each digester were mapped to the assembled Hyd24-12 genomes obtained from the other two digesters. Complete coverage of all genomes with the metagenome reads from the other digester revealed that the Hyd24-12 genomes were almost identical. This also indicates, along with the high ANIb, that the genomes are more complete than estimated in Table 2 by CheckM. Indeed, the data suggested that the three strains might actually be variants of the same strain with single-nucleotide polymorphisms only. This is very interesting as the digesters were from different parts of Denmark without any exchange of sludge or feed. This could indicate that they are highly adapted to the specific AD environment in this type of mesophilic digesters.

Table 2 Genome statistics for the three Hyd24-12 genomes
Full size table

Hyd24-12 phylogeny, FISH probe design and morphology

The 16S rRNA genes obtained have a sequence identity of 86% with the original clone Hyd24-12 sequence (AJ535232) (Knittel et al., 2003) and classify to the Hyd24-12-lineage (Figure 2a). Additional phylogenetic analyses, based on the genome sequence, placed the Hyd24-12 genomes within the Fibrobacteres-Chlorobi-Bacteroidetes superphylum (Figure 3). The Hyd24-12 genomes are distantly related to all currently available genomes, supporting its status as a novel phylum.

Figure 2
figure 2

Phylogenetic analysis and design of FISH probes. (a) Maximum-likelihood (PhyML) 16S rRNA gene phylogenetic tree of target groups and selected related sequences (all >1200 bp). Phylogenetic classification is taken from the MiDAS database (Release 1.21), which is a version of the SILVA database (Release 119 NR99) (Quast et al., 2013) curated for activated sludge sequences (McIlroy et al., 2015). Clades of the Hyd24-12 lineage are shown in brackets. Probes covering clades are shown in red. The scale bar represents substitutions per nucleotide base. Bootstrap values from 100 re-samplings are indicated for branches when >50%. (b) Composite FISH micrographs of the B-1AC clade members in the Randers anaerobic digester sludge. B-1AC cells appear red (Hyd24-12_659, Cy3), other bacterial cells appear blue (EUBmix probe set (Amann et al., 1990; Daims et al., 1999), Cy5) and archaeal cells appear green (ARCH915 (Stahl and Amann, 1991), FLUOS).

Full size image
Figure 3
figure 3

Phylogenetic position of the Hyd24-12 genomes in the reference genome tree generated by CheckM. The CheckM tree is inferred from the concatenation of 43 conserved marker genes and incorporates 2052 finished and 3604 draft genomes from the IMG database (Parks et al., 2015).

Full size image

Several FISH probes were designed to target different clades within the phylum. In the MiDAS taxonomy (v. 1.21) (McIlroy et al., 2015), a version of the SILVA taxonomy (Quast et al., 2013) that is curated for activated sludge-related organisms, the Hyd24-12 lineage is delineated into four clades, designated B-1AC, zEL51, Hyd-32 and B9.18. The Hyd24-12_468 and Hyd24-12_659 probes were designed to cover the B-1AC clade, which includes the Hyd24-12 genome sequences obtained in this study (Figure 2). The former probe covers almost all the B-1AC sequences, with the closest non-target sequence match having three internal base mismatches. The Hyd24-12_659 probe is less specific, having one perfectly matched non-target sequence and several with mismatches not covered by the competitor probes. Overlap in the coverage of these two probes, labelled with different fluorochromes, allows greater confidence in their specificity. A suitable probe to cover the entire Hyd24-12 lineage was not found. However, the Hyd24-12_731 and Hyd24-12_842 probes provide good coverage of the other sequences in the phylum (see Table 1). As sequences covered by these additional probes were not detected in the full-scale anaerobic digesters studied here, optimization and assessment of these probes were not pursued.

When applied to several full-scale anaerobic digester sludge samples, the Hyd24-12_468 and Hyd24-12_659 probes hybridized to small rods, approx. 2 × 0.4 μm in size, dispersed through the flocs (see Figure 2b). Good overlap was observed for these probes, supporting their specificity. Of the two probes, a much higher signal was observed for the Hyd24-12_659 probe. There was no observed overlap between the signal of two Hyd24-12 probes and the universal bacterial EUBmix probe set (see Figure 2b), which is supported by the absence of the target site for the probes of the latter in the Hyd24-12 sequences. Quantitative FISH was very difficult to carry out in the digesters due to high levels of background fluorescence. Instead, abundance estimates were carried out for the domains Bacteria, Archaea, Eukarya, and the Hyd24-12 lineage, based on read mapping from the PCR free metagenomes to the 16S rRNA genes of the MiDAS database. It showed that Archaea constituted 4–9% of the reads in sludge samples and 7–13% in foam samples. Reads from the Hyd24-12 lineage constituted 0.4–3.5% in the different samples (Supplementary Table S3).

Morphology and motility

The rod shape morphology of B-1AC clade organisms observed by FISH is supported by mreBCD and mrdAB operons in the Hyd24-12 genomes (see Supplementary Data 2). These operons encode proteins involved in the formation of membrane-bound actin filaments, which are essential for the biogenesis of rod-shape stabilizing peptidoglycans along the lateral cell wall of rod-shaped bacteria (Kruse et al., 2003, 2005; Osborn and Rothfield, 2007; Bendezú and de Boer, 2008).

The cell envelope characteristics of genome-sequenced bacteria can be determined based on PFAM protein families that are substantially enriched or depleted in archetypical monoderm lineages relative to archetypical diderm lineages (Albertsen et al., 2013). A search for such protein families in the Hyd24-12 genomes revealed an archetypical diderm cell envelope with lipopolysaccharides (see Supplementary Figure S2).

None of the Hyd24-12 genomes encode any flagella-related proteins, suggesting limited motility. However, genes associated with type IV pili were identified using the PilFind algorithm (see Supplementary Data 2) (Imam et al., 2011). These pili enable the bacteria to generate surface-associated twitching motility. This allows them to move effectively through environments that contain shear-thinning viscoelastic fluids, such as the extracellular polymeric substances of biofilms (Conrad et al., 2011; Jin et al., 2011). In addition to motility, type IV pili play a role in the attachment to living and non-living surfaces, including those of other bacteria (Giltner et al., 2012).

No genes associated with spore formation were detected in the Hyd24-12 genomes. This suggests that the Hyd24-12 genomes investigated represent non-sporulating bacteria.

Energy metabolism

The three genomes do not contain any genes for respiration with oxygen, nitrate/nitrite or Fe(III) and seem primarily to have a fermentative metabolism. However, the genomes indicate that the organisms may be able to use elemental sulphur as an electron acceptor, see below. The Hyd24-12 genomes encode a complete glycolysis pathway, along with the non-oxidative branch of the pentose phosphate pathway (Figure 4 and see Supplementary Data 2). This allows Hyd24-12 to potentially catabolize a wide range of hexoses and pentoses to pyruvate, thereby providing the cell with energy in the form of ATP and reducing equivalents in the form of NADH (Stincone et al., 2014). The sugars are probably obtained from the environment through a major facilitator superfamily transporter at the expense of the proton motive force (Madej, 2014; Wisedchaisri et al., 2014). The transporter does not share similarity (>30%) with any experimentally validated transporters, and it is therefore impossible to infer a specific substrate preference. It is known that primary sludge and activated sludge fed into the digesters contain many different polysaccharides (Raunkjaer et al., 1994; Frølund et al., 1996). No genes encoding for extracellular glycosylases were identified, which might indicate that Hyd24-12 is reliant on the hydrolytic action of other organisms present within the anaerobic digesters.

Figure 4
figure 4

Metabolic model of Hyd24-12 species in mesophilic anaerobic digesters, based on the annotated genome sequences. Selected metabolic pathways important for the bacteria in the anaerobic digestion process are highlighted. Numbers correspond to annotated genes in Supplementary Data 1.

Full size image

Hyd24-12 encodes for the complete pathway for glycogen biosynthesis and catabolism (Figure 4 and see Supplementary Data 2) (Preiss et al., 1983; Wilson et al., 2010). Hence, glycogen may serve as a carbon and energy storage which can be utilized to mitigate fluctuations in substrate availability. The Hyd24-12 genomes did not encode for pathways for other storage compounds such as trehalose or polyhydroxyalkanoates.

There are limited catabolic options for the pyruvate formed, for example, by glycolysis. The tri-carboxylic acid cycle of Hyd24-12 is incomplete (8 of 10 key enzymes are missing) and probably non-functional. However, pyruvate can be converted into acetyl-CoA by a pyruvate ferredoxin oxidoreductase, providing additional reducing equivalents in the form of reduced ferredoxin (Figure 4 and see Supplementary Data 2) (Menon and Ragsdale, 1997). Acetyl-CoA can then be converted into acetate by the action of phosphate acetyltransferase and acetate kinase, thus providing the bacterium with additional ATP (Latimer and Ferry, 1993; Mai and Adams, 1996).

All three Hyd24-12 genomes also encode for two aldehyde ferredoxin oxidoreductases (Figure 4 and see Supplementary Data 2). These may be used to oxidize formaldehyde and acetaldehyde to formate and acetate, respectively, providing the cell with energy in the form of additional reduced ferredoxin (Mukund and Adams, 1991). However, the enzyme may also be used in the reverse reaction to regenerate oxidized ferredoxin. The presence of a membrane-embedded, energy-conserving hydrogenase allows the cell to establish a proton motive force, based on the energy-rich reduced ferredoxin, which reduces H+ to H2 in the process (Strittmatter et al., 2009). The energy stored in the proton motive force may then be harvested through an ATP synthase to yield ATP.

High concentrations of H2 inhibit glycolysis and acidogenesis due to thermodynamic considerations (Huang et al., 2015). Hyd24-12 therefore needs a way to remove excess H2. This can be achieved by syntrophic association with other microorganisms, or internally by the action of a cytosolic hydrogenase, which couples the oxidation of H2 with the reduction of NAD+ (Figure 4 and see Supplementary Data 2). Alternatively, Hyd24-12 may employ a sulfhydrogenase to couple the oxidation of H2 to H+ with the reduction of elemental sulphur (S0) or polysulphide to hydrogen sulphide (H2S) as is seen for Pyrococcus furiosus (Mukund and Adams, 1991). The genomes do not indicate a potential for sulphate reduction. Elemental sulphur is continuously produced in the digesters because activated sludge fed into the digesters contains oxidized iron (Fe(III)), which in the presence of sulphide produces S0 and black iron sulphide (FeS) (Rasmussen and Nielsen, 1996; Nielsen et al., 2005; Omri et al., 2011). Sulphide is a normal compound in digesters and is produced from amino acids and reduction of sulphate. Notably, other studies have also detected members of the Hyd24-12 phylum in sulphur-rich environments such as hydrothermal vents, sulphur-rich springs and sediments (Elshahed et al., 2003; Schauer et al., 2011; Pjevac et al., 2014). Thus, Hyd24-12 related organisms potentially play a role in sulphur transformations in digesters and other environments. Such a role requires further investigation.

The Hyd24-12 genomes do not contain the genes required for fatty acid β-oxidation or for the catabolism of amino acids. Sugars are therefore considered the primary energy source of the Hyd24-12 in anaerobic digesters.

Whereas Hyd24-12 is able to take up carbon in the form of amino acids, carbohydrates, etc., it is unable to carry out fixation of CO2 as such genes are missing.

Amino acid and nitrogen metabolism

Based on the genome annotations, Hyd24-12 is only predicted to be able to synthesize few amino acids (glycine, serine, cysteine, threonine, asparagine, aspartate, glutamate and glutamine). Accordingly, Hyd24-12 might rely on amino acids present within the environment. As most amino acids are found as proteins, which cannot be taken up by the bacterium, Hyd24-12 needs a way to degrade these polymers, and this is achieved by the action of multiple extracellular proteases encoded in the genome, which are likely secreted in a Sec- or Tat-dependent mechanism (Natale et al., 2008) (see Supplementary Data 2). The cells may subsequently import the amino acids using ABC-transporters encoded in the genome. Owing to the lack of experimentally validated homologues from closely related species, it is not possible to predict the substrate specificity of these transporters. A reduced capacity of microorganisms for synthesizing amino acids is known from strict symbionts and, recently, also from a number of candidate phyla with very small genomes (<1 Mbp) (Brown et al., 2015). However, the relatively large size of the Hyd24-12 genomes (~2.2 Mbp) and their dispersed growth in the anaerobic sludge suggest that they are not strict symbionts.

Hyd24-12 does not have the necessary pathways for fixation of nitrogen. The nitrogen metabolism of Hyd24-12 is generally limited. Amino acids may also represent a source of nitrogen. However, nitrogen can also be obtained from ammonium assimilation via the glutamine synthetase/glutamate synthase pathways (Bravo and Mora, 1988).

Oxidative stress protection

The three Hyd24-12 genomes each contains a gene cluster encoding for a superoxide reductase, nitric oxide reductase and ferroxidase. These genes are probably involved in resistance against oxidative stress, and may allow the bacteria to survive in the presence of oxygen. However, 16S rRNA gene sequences from Hyd24-12 have only been observed in oxygen-depleted environments.

Ecological significance and concluding remarks

This study applied metagenomic sequencing to obtain genomes from the candidate phylum Hyd24-12 and provides the first morphological and physiological information for the lineage. Members of the phylum were shown to be very abundant and stably present in mesophilic anaerobic digesters, occasionally accounting for the most abundant OTU in the samples, but absent in thermophilic reactors. This indicates that they are likely to play a substantial role in the ecology of mesophilic AD systems at wastewater treatment plants fed with primary sludge and surplus activated sludge. Metabolic reconstruction based on the genomic information showed that members of Hyd24-12 are likely to be fermenters relying on simple sugars. In addition, they may also use elemental sulphur as an electron acceptor, thus forming part of the microbial cycling of sulphur in anaerobic systems and partly responsible for production of hydrogen sulphide. Sulphide is unwanted in the biogas due to toxicity and corrosion (Syed et al., 2006), but will also provide more elemental sulphur by reacting with incoming Fe(III). In that case, members of Hyd24-12 may compete with the methanogens for organics. An in silico investigation of environmental 16S rRNA gene surveys suggests that members of the phylum are present in anaerobic environments, often associated with sulphurous compounds and methane production, such as sediment mats and anaerobic bioreactors. The fact that the genomes are auxotrophic for several amino acids and lacking putative secreted glycoside hydrolases also indicates a strict reliance on other organisms for nutrients. The genomes generated in this study provide the foundation for future detailed analyses of members of the phylum, such as metatranscriptomics and metaproteomics. The design of FISH probes for the phylum also revealed their morphology and spatial arrangement in anaerobic digesters and will also facilitate future in situ investigations of the phylum in digesters and other environments.

Phylogenetic and genomic analyses of the three Hyd24-12 genomes classified them as a single species within a novel phylum located within the Fibrobacteres-Chlorobi-Bacteroidetes superphylum.

We propose the following taxonomic names for the novel genus and species of Hyd24-12:

  • ‘Candidatus Fermentibacter’ gen. nov.

  • ‘Candidatus Fermentibacter daniensis’ gen. et sp. nov.

Based on this, we propose the following names for the phylum, class, order, and family:

  • ‘Candidatus Fermentibacteria’ phyl. nov.

  • ‘Candidatus Fermentibacteria’ classis nov.

  • ‘Candidatus Fermentibacterales’ ord. nov.

  • ‘Candidatus Fermentibacteraceae’ fam. nov.

Etymology

Fermentibacter (Fer.men.ti.bac'ter. M.L. n. ferment -um to ferment, Gr. dim. n. bakterion a small rod, M.L. neut. n. Fermentibacter a small fermenting rod-shaped bacterium). Fermentibacter daniensis (da.ni.ensis. M.L. fem. adj. daniensis, pertaining to Dania, the Medieval Latin name for the country of Denmark, where the species was first discovered).