Growth arrest in the active rare biosphere

Microbial diversity in the environment is mainly concealed within the rare biosphere, which is arbitrarily defined as all species with <0.1% relative abundance. While dormancy explains a low-abundance state very well, the cellular mechanisms leading to rare but active microorganisms are not clear. We used environmental systems biology to genomically and metabolically characterize a cosmopolitan sulfate reducer that is of low abundance but highly active in peat soil, where it contributes to counterbalance methane emissions. We obtained a 98%-complete genome of this low-abundance species, Candidatus Desulfosporosinus infrequens, by metagenomics. To test for environmentally relevant metabolic activity of Ca. D. infrequens, anoxic peat soil microcosms were incubated under diverse in situ-like conditions for 36 days and analyzed by metatranscriptomics. Compared to the no-substrate control, transcriptional activity of Ca. D. infrequens increased 56- to 188-fold in incubations with sulfate and acetate, propionate, lactate, or butyrate, revealing a versatile substrate use. Cellular activation was due to a significant overexpression of genes encoding ribosomal proteins, dissimilatory sulfate reduction, and carbon-degradation pathways, but not of genes encoding DNA or cell replication. We show for the first time that a rare biosphere member transcribes metabolic pathways relevant for carbon and sulfur cycling over prolonged time periods while being growth-arrested in its lag phase. Significance The microbial rare biosphere represents the largest pool of biodiversity on Earth and constitutes, in sum of all its members, a considerable part of a habitat’s biomass. Dormancy or starvation are typically used to explain a low-abundance state. We show that low-abundance microorganisms can be highly metabolically active while being growth-arrested over prolonged time periods. We show that this is true for microbial keystone species, such as a cosmopolitan but low-abundance sulfate reducer in wetlands that is involved in counterbalancing greenhouse gas emission. Our results challenge the central dogmas “metabolic activity translates directly into growth” as well as “low abundance equals little ecosystem impact” and provide an important step forward in understanding rare biosphere members relevant for ecosystem functions.


Introduction
The vast majority of microbial diversity worldwide is represented by the rare biosphere (Sogin et al., 2006;Pedrós-Alió, 2012;Lynch and Neufeld, 2015;Jousset et al., 2017). This entity of microorganisms consists of all microbial species that have an arbitrarily defned relative population size of <0.1% in a given habitat at a given time (Sogin et al., 2006;Pedrós-Alió, 2012;Lynch and Neufeld, 2015;Jousset et al., 2017). The rare biosphere is opposed by a much smaller number of moderately abundant or very abundant microbial species (≥0.1% and ≥1.0% relative abundance, respectively, Hausmann et al., 2016), which are thought to be responsible for the major carbon and energy fow through a habitat as based on their cumulative biomass. However, there is accumulating experimental evidence that the rare biosphere is not just a so-called "seed bank" of microorganisms that are waiting to become active and numerically dominant upon environmental change (Müller et al., 2014;Lynch and Neufeld, 2015), but also harbors metabolically active microorganisms with important ecosystem functions (Jousset et al., 2017).
First hints for metabolically active rare biosphere members were evident from seasonal patterns of marine bacterioplankton species. Here, many taxa that displayed recurring annual abundance changes were of low abundance and even during their bloom periods never reached numerically abundant population sizes (Campbell et al., 2011;Hugoni et al., 2013;Alonso-Sáez et al., 2015). In soil environments, removal of low-abundance species by dilution-to-extinction had a positive efect on intruding species, suggesting that active low-abundance species pre-occupy ecological niches and thus slow down invasion (van Elsas et al., 2012;Vivant et al., 2013;Mallon et al., 2015). Soil microorganisms of low relative abundance were also shown to play a role in community-wide species interactions, e.g, by being involved in the production of antifungal compounds that protect plants from pathogens (Hol et al., 2015) or by constituting the core of microorganisms that respond to the presence of a particular plant species (Dawson et al., 2017). Other examples include microorganisms with a specialized metabolism that sustain stable low-abundance populations in an ecosystem (Lynch and Neufeld, 2015). For example, N₂-fxing microorganisms in the ocean (Großkopf et al., 2012) or sulfate-reducing microorganisms (SRM) in peatlands (Pester et al., 2010(Pester et al., , 2012bHausmann et al., 2016) were shown to fulfll such gatekeeper functions. rapid aerobic or anaerobic re-oxidation of reduced sulfur species back to sulfate (Pester et al., 2012b). Since SRM generally outcompete methanogens and syntrophically associated fermenters (Muyzer and Stams, 2008), they exert an important intrinsic control function on peatland CH₄ production (Gauci et al., 2004(Gauci et al., , 2005Gauci and Chapman, 2006). This is important, since natural wetlands, such as peatlands, are estimated to be responsible for 30% of the annual emission of this potent greenhouse gas (Ciais et al., 2013;Kirschke et al., 2013;Saunois et al., 2016).
Little is known about the ecophysiology of metabolically active but low-abundance microorganisms.
This lack of knowledge is clearly founded in their low numerical abundance making it inherently difficult to study their metabolic responses or even to retrieve their genomes directly from the environment. In a preceding study, we could show that the low-abundance peatland Desulfosporosinus species mentioned above follows an ecological strategy to increase its cellular ribosome content while maintaining a stable population size when exposed to favorable, sulfatereducing conditions (Hausmann et al., 2016). This was unexpected since metabolic activity in bacteria and archaea is typically immediately followed by growth. Furthermore, this Desulfosporosinus species can be found worldwide in a wide range of low-sulfate wetlands including not only peatlands but also permafrost soils and rice paddy felds (Hausmann et al., 2016), which emphasizes its importance as a model organism for active rare biosphere members. In this study, we used an environmental systems biology approach to deepen our understanding of the ecophysiology of this rare biosphere member. In particular, we retrieved its genome directly from a combination of native and incubated peat soil and followed its transcriptional responses in peat soil microcosms, which were exposed to diferent environmental triggers that mimicked diverse in situ conditions. 2012a; Abu Laban et al., 2015;Petzsch et al., 2015;Mardanov et al., 2016).

The versatile energy metabolism of the low-abundance Desulfosporosinus
Desulfosporosinus MAG SbF1 encoded the complete canonical pathway for dissimilatory sulfate reduction ( Fig. 1, Table S1). This encompassed the sulfate adenylyltransferase (Sat), adenylyl-sulfate reductase (AprBA), dissimilatory sulfte reductase (DsrAB), and the sulfde-releasing DsrC, which are sequentially involved in the reduction of sulfate to sulfde. In addition, genes encoding the electrontransferring QmoAB and DsrMKJOP complexes were detected, with their subunit composition being typical for Desulfosporosinus species (Abicht et al., 2011;Pester et al., 2012a;Petzsch et al., 2015;Mardanov et al., 2016). Other dsr genes included dsrD, dsrN, and dsrT (Rabus et al., 2015), with hitherto unvalidated function, as well as fdxD, which encodes a [4Fe4S]-ferredoxin, and a second set of DsrMK-family encoding genes (dsrM2 and dsrK2). SbF1 also encoded the trimeric dissimilatory sulfte reductase AsrABC (anaerobic sulfte reductase) (Huang and Barrett, 1991). SbF1 carries genes for both complete and incomplete oxidation of propionate and lactate. In addition, the ability to utilize acetate, formate, or H₂ as electron donors was encoded (Fig. 1). All enzymes necessary for propionate oxidation to the central metabolite pyruvate (including those belonging to a partial citric acid cycle) were encoded on two scafolds (Table S1). For lactate utilization, SbF1 carried three paralogs of glycolate/D-lactate/L-lactate dehydrogenase family genes.
However, the substrate specifcity of the encoded enzymes could not be inferred from sequence information alone. The transcription of lutDF and lutD_2 was stimulated by the addition of L-lactate ( Fig. 1), which indicates that these genes encode functional lactate dehydrogenases (LDH). The third acetyltransferase (PfD). Acetyl-CoA can then be completely oxidized to CO₂ via the Wood-Ljungdahl pathway (Pierce et al., 2008), which is complete in SbF1 (Fig. 1, Table S1) and present in the genomes of all other sequenced Desulfosporosinus species (Abicht et al., 2011;Pester et al., 2012a;Petzsch et al., 2015;Mardanov et al., 2016). Alternatively, acetyl-CoA may be incompletely oxidized to acetate via acetyl-phosphate by phosphate acetyltransferase (Pta) and acetate kinase (AckA). Pta and AckA are bidirectional enzymes, opening the possibility that acetate could be degraded via these two enzymes and the downstream Wood-Ljungdahl pathway to CO₂.
However, neither a glucokinase or a phosphotransferase system was found (PTS). Coupling of electron transfer to energy conservation could be mediated in SbF1 by a H⁺/Na⁺-pumping Rnf complex (RnfCDGEAB) (Buckel and Thauer, 2013) and a NADH dehydrogenase (respiratory complex I, NuoABCDEFGHIJKLMN). In addition, the complete gene set for ATP synthase (AtpABCDEFGH) was identifed ( Fig. 1, Table S1).

Activation of energy metabolism is uncoupled from cell division initiation
We used metatranscriptomics to analyse gene expression changes of Desulfosporosinus MAG SbF1 in anoxic peat microcosms, which mimicked diverse in situ-like conditions. Total transcriptional activity of Desulfosporosinus MAG SbF1 was clearly stimulated by individual additions of acetate, propionate, lactate, and butyrate in combination with sulfate. In these incubations, total mRNA counts of SbF1 increased by 56-, 80-, 62-, and 188-fold as compared to the no-substrate-control, respectively, and constituted between 0.11 ± 0.13% (acetate) and 0.36 ± 0.02% (butyrate) of all transcripts in the respective metatranscriptomes after 36 days (Fig. 2a). This substrate-specifc activity was mirrored in the increased transcription of genes encoding ribosomal proteins as general activity markers ( Fig.   2b) and of all dissimilatory sulfate reduction genes, except the alternative pathway via asrABC (Fig.   3). For example, Spearman's rank correlation coefficients of dsrA and dsrB transcripts as compared to total mRNA counts were 0.91 and 0.90, respectively (FDR-adjusted p-value < 0.001). Other enzyme complexes involved in the central metabolism of SbF1 such as the ATP synthase, the NADH dehydrogenase (complex I), and ribosomal proteins followed the same transcriptional pattern (Fig. 3) with an average Spearman's rank correlation coefficients of 0.79 ± 0.07 (n = 72, FDR-adjusted pvalue < 0.05) to total mRNA counts. Interestingly, transcription of genes encoding proteins involved in general stress response were stimulated as well. In particular, genes encoding the universal stress promotor UspA and the GroSL chaperonin showed an increased transcription ( Fig. 3) with an average Spearman's rank correlation coefficients of 0.76 ± 0.06 (n = 3, FDR-adjusted p-value < 0.05) to total mRNA counts.
In addition, we screened the COG categories D, L, and M for indicator genes that encode functions in cell division (e.g., ftsZ or minE), DNA replication (e.g., gyrBA, dnaC, and dnaG) and cell envelope biogenesis (e.g., murABCDEFGI), respectively, and followed their expression patterns. Genes that unambiguously encoded such functions (Table S1) showed either no or only barely detectable but insignifcant (FDR-adjusted p-value > 0.05) increases in transcripts under these conditions (Fig. 2b, detailed in Fig. S4). Extension of this analysis to all genes belonging to COG D (n = 73), L (n = 280), and M (n = 215), which included also genes with ambiguous classifcation or unknown function, revealed that also here 96%, 99%, and 99%, respectively, were not signifcantly overexpressed under acetate, propionate, lactate, and butyrate in combination with sulfate (Table S1, Table S3).
We also analysed genes reported to be upregulated immediately after phage infection, as a potential ecological driver that controls bacterial population size. Respective genes in Bacillus subtilis encode, e.g., functions in DNA and protein metabolism and include the ribonucleoside-diphosphate reductase (nrdEF), aspartyl/glutamyl-tRNA amidotransferase (gatCAB), and the proteolytic subunit of ATPdependent Clp protease (clpP) (Mojardín and Salas, 2016). However, homologs in SbF1 were not signifcantly overexpressed (FDR-adjusted p-value > 0.05), which was refected in an average Spearman's rank correlation coefficient of 0.63 ± 0.08 (n = 5, FDR-adjusted p-value < 0.05) to total mRNA counts. The same was true when screening for active sporulation of a Desulfosporosinus subpopulation as an alternative explanation for a low population size. The identifed sporulation genes (spo0A-spoVT) did not show any signifcant increase in transcript numbers as well, with the only exception of spoIIIAD. This stage III sporulation gene was signifcantly more abundant when stimulated with propionate and sulfate, however did not correlate to total mRNA levels (Table S1).
Again, genes involved in sporulation had a low average Spearman's rank correlation coefficient of 0.44 ± 0.13 (n = 22, FDR-adjusted p-value < 0.05) to total mRNA counts.
The individual incubation regimes additionally triggered transcriptional activation of the respective substrate degradation pathways of Desulfosporosinus MAG SbF1. For example, all genes necessary for the conversion of propionate to pyruvate were overexpressed only upon addition of propionate and sulfate but not in any other incubation type. The same was true for lactate degradation, where genes encoding the lactate permease, lactate racemase and two of the detected lactate dehydrogenases were overexpressed upon addition of both lactate and sulfate, but not in incubations with lactate only (Fig. 3). Although genes encoding phosphotransacetylase and acetate kinase were overexpressed under lactate and propionate, the complete Wood-Ljungdahl pathway was overexpressed as well, which indicates that at least part of these substrates were completely degraded to CO₂ rather than to acetate and CO₂. This conclusion was supported by the overexpression of the Wood-Ljungdahl pathway in incubations amended with acetate and sulfate.
Interestingly, the Wood-Ljungdahl pathway was also overexpressed upon addition of butyrate and

Discussion
Current knowledge on the mechanisms that interconnect energy metabolism, gene expression, cell division, and population growth of microorganisms are mainly based on pure cultures that can be easily maintained in the laboratory. Here, the typical lifecycle of a metabolically active microorganisms would go through an activating lag phase, an exponential growth phase, and a stationary phase upon limitation of substrate, nutrient, or space. Under ideal conditions, a single Escherichia coli cell would grow to a population with the mass of the Earth within 2 days. Clearly, this does not occur, but the discrepancy between potential and actual growth underscores that microorganisms spend the vast majority of their time not dividing (Bergkessel et al., 2016). A large fraction of these microorganisms is part of the rare biosphere. For example, in the studied peatland, the sum of all low-abundance species made up approximately 12% of the total bacterial and archaeal 16S rRNA genes (Hausmann et al., 2016). In other soils, low-abundance Alphaproteobacteria and Bacteroidetes alone constituted in sum 10% and 9% of the total bacterial population, respectively, while all low-abundance populations summed up to 37% of all bacteria (Dawson et al., 2017). Upon strong environmental change, low-abundance microorganisms often grow to numerically abundant populations and replace dominant species as observed for microbial community changes after an oil spill (Teira et al., 2007;Newton et al., 2013)  The low-abundance Desulfosporosinus MAG SbF1 represents an interesting case of the latter response type. When exposed to favorable, sulfate-reducing conditions in peat soil microcosms, it did not increase its population size but drastically increased its cellular ribosome content by one order of magnitude to 57,000-84,000 16S rRNA molecules per cell (Hausmann et al., 2016).
Throughout the incubation period of 50 days, it correlated best in its 16S rRNA response to sulfate turnover among all identifed SRM (Hausmann et al., 2016). In this study, we expanded upon this observation by genome-centric metatranscriptomics to test whether the increase in cellular ribosome content is indeed translated into increased transcriptional and, as a consequence, metabolic activity of Desulfosporosinus MAG SbF1. As expected, increases in cellular 16S rRNA content clearly corresponded to increased transcription of genes coding for ribosomal proteins (Fig. 2b;Hausmann et al., 2016). This cellular ribosome increase under sulfate-reducing conditions was correlated to an increase in total mRNA counts (Fig. 2). This is the frst time that changes in population-wide 16S rRNA levels are proven to be directly linked to transcriptional activity for a rare biosphere member. clearly overexpressed its sulfate reduction pathway under sulfate amendment when supplied with either acetate, lactate, propionate, or butyrate as compared to the no-substrate and the methanogenic controls (Fig. 3). Detailed analysis of the transcribed carbon degradation pathways showed that Desulfosporosinus MAG SbF1 is able to oxidize propionate, lactate, and acetate completely to CO₂. Under butyrate-amended conditions it presumably relied on syntrophic oxidation of acetate supplied by a primary butyrate oxidizer. This shows that Desulfosporosinus MAG SbF1 is capable of utilizing diverse substrates that represent the most important carbon degradation intermediates measured in peatlands (Schmalenberger et al., 2007;Küsel et al., 2008). Such a generalist lifestyle is of clear advantage in peat soil given the highly variable nutrient and redox conditions (Schmalenberger et al., 2007;Küsel et al., 2008). These fuctuations are caused by the periodically changing water table that steadily shifts the oxic-anoxic interface (Knorr et al., 2009;Reiche et al., 2009). In addition, the complex fow paths of water create distinct spatial and temporal patterns (hot spots and hot moments) of various biogeochemical parameters, to which peat microorganisms have to adapt (Jacks and Norrström, 2004;Knorr et al., 2009;Knorr and Blodau, 2009;Frei et al., 2012).
The question remains, which mechanisms are at work that keep the Desulfosporosinus MAG SbF1 population in a stable low-abundance state? Population sizes can be kept low by actively restricting growth. Alternatively, ongoing growth could be hidden by continuous predation, viral lysis, or active sporulation of a major subpopulation. To answer this question, we analysed expression patterns of growth-specifc genes. Compared to the strong overexpression of metabolic or ribosomal protein genes, transcription of genes essential for DNA replication, cell division, and cell envelope biogenesis did not increase or only marginally (Fig. 2b, Fig. S4). Genes encoding DNA replication or cell division typically show a largely invariable transcription in the exponential and stationary phase (e.g., Sumby Brudal et al., 2013;Sihto et al., 2014). However, there is experimental evidence that in the lag phase transcription of growth-specifc genes is not stable but increases due to the overall activation of cellular processes (Rolfe et al., 2012). In this context, the lack of an increasing transcription of growth-specifc genes would clearly indicate a state of no growth rather than an actively dividing population that is kept stable by an equally high growth and mortality or sporulation rate. This conclusion is further corroborated by the lack of overexpressed sporulation genes or genes upregulated directly after phage attack. Nevertheless, the ATP generated by the induced energy metabolism has to be utilized somehow. This could be mediated by the production of storage compounds or by counterbalancing environmental stress. We found no indication for the former scenario but observed overexpression of the universal stress promotor UspA, which is one of the most abundant proteins in growth-arrested cells (Kvint et al., 2003), and the chaperonin GroSL, which was linked previously to stress response such as low pH (Silva et al., 2005). Since the pH in the analyzed peat soil incubations varied between 4.1-5.0 (Hausmann et al., 2016), coping with a low pH would be the most likely reason that deviates ATP away from growth towards stress response. Based on the integrated fndings of our previous study (stable population over 50 days as based on 16S rRNA gene counts; Hausmann et al., 2016), and this study (no activation of the DNA replication and cell division machinery within 36 days), we propose that Desulfosporosinus MAG SbF1 was growth-arrested in the lag phase over a period of at least 50 days while being a metabolically active rare biosphere member. This fnding shows that growth arrest is not restricted to starving or otherwise limited microorganisms that persist in the environment (Bergkessel et al., 2016) but can also occur in metabolically highly active microorganisms.
Our results are important in the context of the increasing awareness that the microbial rare biosphere is not only the largest pool of biodiversity on Earth (Sogin et al., 2006;Pedrós-Alió, 2012;Lynch and Neufeld, 2015;Jousset et al., 2017) but in sum of all its low-abundance members constitutes also a large part of the biomass in a given habitat (e.g., Hausmann et al., 2016;Dawson et al., 2017). Understanding the mechanisms governing this low-abundance prevalence and its direct impact on ecosystem functions and biogeochemical cycling is thus of utmost importance.
Desulfosporosinus MAG SbF1 has been repeatedly shown to be involved in cryptic sulfur cycling in peatlands (Pester et al., 2010;Hausmann et al., 2016) -a process that counterbalances the emission of the greenhouse gas methane due to the competitive advantage of SRM as compared to microorganisms involved in the methanogenic degradation pathways (Muyzer and Stams, 2008).
This species can be found worldwide in low-sulfate environments impacted by cryptic sulfur cycling including not only peatlands but also permafrost soils, rice paddies, and other wetland types (Hausmann et al., 2016). Here, we provided proof that Desulfosporosinus MAG SbF1 is indeed involved in the degradation of important anaerobic carbon degradation intermediates in peatlands while sustaining a low-abundance population. It has a generalist lifestyle in respect to the usable carbon sources, re-emphasizing its importance in the carbon and sulfur cycle of peatlands. Our results provide an important step forward in understanding the microbial ecology of biogeochemically relevant microorganisms and show that low-abundance keystone species can be studied "in the wild" using modern environmental systems biology approaches. Hausmann et al., 2018). In brief, DNA from native peat soil (10-20 cm depth) and DNA pooled from 16 ¹³C-enriched fractions (density 1.715-1.726 g mL⁻¹) of a previous DNA-SIP experiment with soil from the same site (Pester et al., 2010) was sequenced using the Illumina HiSeq 2000 system. DNA-SIP was performed after a 73-day incubation (again 10-20 cm depth) that was periodically amended with small dosages of sulfate and frst a mixture of unlabeled formate, acetate, propionate, and lactate for two weeks and thereafter a mixture of ¹³C-labeled formate, acetate, propionate, and lactate (all in the lower µM-range) (Pester et al., 2010). Raw reads were quality fltered, trimmed, and co-assembled into one metagenomic assembly using the CLC Genomics Workbench 5.5.1 (CLC Bio). Diferential coverage binning was applied to extract the Desulfosporosinus metagenomeassembled genome (MAG) (Albertsen et al., 2013). A side efect of sequencing a DNA-SIP sample is an apparent G+C content skew, which was normalized arbitrarily to improve binning using the putative homologs of the respective database entries. In addition, classifcation according to COG (Galperin et al., 2015) or InterPro superfamilies, domains, or binding sites were used to call putative homologs in cases of an unambiguous annotation. Membership to syntenic regions (operons) was considered as additional support to call true or putative homologs.

Metatranscriptomics from single-substrate incubations
We analysed total RNA from anoxic peat soil slurry microcosms that were described previously (Hausmann et al., 2016(Hausmann et al., , 2018. In brief, anoxic microcosms were incubated at 14 °C in the dark for 50 days and regularly amended with either low amounts of sulfate (76-387 µM fnal concentrations) or incubated without an external electron acceptor. Formate, acetate, propionate, lactate, butyrate (<200 µM), or no external electron donor was added to biological triplicates each. RNA was extracted from the native soil, and after 8 and 36 days of incubations, followed by sequencing with the Illumina HiSeq 2000/2500 system. Raw reads were quality-fltered as described previously (Hausmann et al., 2018) and mapped to the combined metagenomic assembly using Bowtie 2 (Langmead and Salzberg, 2012). Counting of mapped reads to protein-coding genes (CDS) was performed with featureCounts 1.5.0 (Liao et al., 2014). We used an unsupervised approach to identify CDS stimulated by sulfate and the diferent substrates regimes. First, we applied the DESeq2 R package (Love et al., 2014;R Core Team, 2017) to identify diferentially expressed CDS. Treatments without external sulfate added and samples after 8 days of incubations had too little transcript counts to be used for a statistical approach. Therefore, we limited our analysis to pairwise comparison of sulfatestimulated microcosms after 36 days of incubations. We compared each substrate regime to the nosubstrate controls and each other. The set of all signifcantly diferentially expressed CDS (FDRadjusted p-value < 0.05) were further clustered into response groups. For clustering, we calculated pairwise Pearson's correlation coefficients (r) of variance stabilized counts (cor function in R), transformed this into distances (1−r), followed by hierarchical clustering (hclust function in R).
Variance stabilisation was performed using the rlog function of the DESeq2 package.

Sequence data availability
The MAG SbF1 is available at MicroScope (https://www.genoscope.cns.fr/agc/microscope/) and is also deposited under the ENA accession number OMOF01000000. Metagenome and -transcriptomic data is available at the Joint Genome Institute (https://genome.jgi.doe.gov/) and is also deposited under the NCBI accession numbers PRJNA412436 and PRJNA412438, respectively.  Fig. 1 Metabolic model of Desulfosporosinus MAG SbF1. Gene expression stimulated by specifc substrates in combination with sulfate is indicated by coloured points. Paralogous genes are indicated by an underscore followed by a number. Plus signs indicates proposed protein complexes. Details for all genes are in given in Table S1 and transcription patterns are shown in Fig. 3. For the citric acid cycle and anaplerotic reactions, carriers of reducing equivalents and further by-products are not shown.

Figures
The following abbreviations were used. X: unknown reducing equivalents, e.g., NAD⁺ or ferredoxin.

Table S2
Characteristics and coverage of all scafolds belonging to Desulfosporosinus MAG SbF1. The two scafolds with the highest coverage encode the 23S and 16S rRNA genes, respectively.

Table S3
Expression levels of selected CDS in the analysed anoxic peat soil microcosms given in FPKM (mean ± one standard deviation). Loci are sorted as in Table S1. Headers display the individual treatments used in the peat soil microcosms: without and with external sulfate added; amended substrate; and days of incubation. White circles represent unclassifed scafolds. Only scafolds >10 000 nt length are shown, except when belonging to SbF1. Scafolds encoding selected genes in SbF1 are labelled accordingly.

Fig. S3
Two-way average amino and nucleic acid identities between Desulfosporosinus and Desulftobacterium species genomes (in%, written into cells). The dendrogram is based on Fig. S2b.

Fig. S4
Time-resolved changes of all unambiguously identifed genes related to cell division (a), DNA replication (b) and cell envelope biogenesis (c); dsrA is included for reference, analogous to Fig. 2.