Multi-omic analysis of medium-chain fatty acid synthesis by Candidatus Weimerbacter bifidus, gen. nov., sp. nov., and Candidatus Pseudoramibacter fermentans, sp. nov.

Chain elongation is emerging as a bioprocess to produce valuable medium-chain fatty acids (MCFA; 6 to 8 carbons in length) from organic waste streams by harnessing the metabolism of anaerobic microbiomes. Although our understanding of chain elongation physiology is still evolving, the reverse β-oxidation pathway has been identified as a key metabolic function to elongate the intermediate products of fermentation to MCFA. Here, we describe two chain-elongating microorganisms that were enriched in an anaerobic microbiome transforming the residues from a lignocellulosic biorefining process to short- and medium-chain fatty acids. Based on a multi-omic analysis of this microbiome, we predict that Candidatus Weimerbacter bifidus, gen. nov., sp. nov. used xylose to produce MCFA, whereas Candidatus Pseudoramibacter fermentans, sp. nov., used glycerol and lactate as substrates for chain elongation. Both organisms are predicted to use an energy conserving hydrogenase to improve the overall bioenergetics of MCFA production. IMPORTANCE Microbiomes are vital to human health, agriculture, environmental processes, and are receiving attention as biological catalysts for production of renewable industrial chemicals. Chain elongation by MCFA-producing microbiomes offer an opportunity to produce valuable chemicals from organic streams that otherwise would be considered waste. However, the physiology and energetics of chain elongation is only beginning to be studied, and we are analyzing MCFA production by self-assembled communities to complement the knowledge that has been acquired from pure culture studies. Through a multi-omic analysis of an MCFA-producing microbiome, we characterized metabolic functions of two chain elongating bacteria and predict previously unreported features of this process.

gen. nov., sp. nov. used xylose to produce MCFA, whereas Candidatus Pseudoramibacter 23 fermentans, sp. nov., used glycerol and lactate as substrates for chain elongation. Both organisms 24 are predicted to use an energy conserving hydrogenase to improve the overall bioenergetics of 25 MCFA production. 26

IMPORTANCE 27
Microbiomes are vital to human health, agriculture, environmental processes, and are receiving 28 attention as biological catalysts for production of renewable industrial chemicals. Chain elongation 29 by MCFA-producing microbiomes offer an opportunity to produce valuable chemicals from 30 organic streams that otherwise would be considered waste. However, the physiology and 31 energetics of chain elongation is only beginning to be studied, and we are analyzing MCFA 32 production by self-assembled communities to complement the knowledge that has been acquired 33 from pure culture studies. Through a multi-omic analysis of an MCFA-producing microbiome, we 34 characterized metabolic functions of two chain elongating bacteria and predict previously 35 unreported features of this process. 36

INTRODUCTION 37
Chain elongation has been proposed as a microbial process to produce valuable chemicals from 38 complex waste feedstocks (1). This bioprocess relies on the combined metabolism of an anaerobic 39 microbiome to hydrolyze complex organic substrates, ferment the hydrolyzed products to small 40 organic intermediates (2 to 3 carbon molecules), and elongate these fermentation products to 41 medium-chain fatty acids (MCFA; 6 to 8 carbon molecules) through reverse β-oxidation (1). 42 MCFA are an attractive product due to their high value, relatively low solubility in water, and 43 potential to offset fossil fuel demands for petrochemicals and other products (2, 3). Bioreactors 44 performing chain elongation also provide model systems for studying the metabolic contributions 45 of uncultured organisms to this process. To date, all but one of the known chain-elongating bacteria 46 belong to the Clostridia class within the Firmicutes phylum (1). These chain-elongating bacteria 47 primarily use lactate (4, 5), ethanol (6), or carbohydrates (7) to drive MCFA production. 48 We recently described a chain-elongating microbiome that produced sufficient hexanoic 49 and octanoic acids from lignocellulosic biorefining residues to reduce the minimum selling price 50 of ethanol produced in a biorefinery (2). Using a combination of metagenomics and 51 metatranscriptomics, we characterized this microbiome as having a small set of high abundance 52 organisms (8), with two populations within the Clostridia class performing chain elongation. One 53 high abundance member of this microbiome (LCO1) (8) belonged to the Lachnospiraceae family 54 and was predicted to produce MCFA from xylose and other pentoses; the second one (EUB1) (8) 55 corresponded to the Eubacteriaceae family and was predicted to produce MCFA from lactate. 56 Here, we combined multi-omic approaches to further analyze the genomic and metabolic 57 features of these two predicted MCFA-producing organisms. A time-series gene expression 58 analysis showed that transcripts encoding proteins predicted to be involved in reverse -oxidation 59 are among the most abundant transcript RNAs after feeding the lignocellulosic biorefinery residues 60 to the microbial community. Our analysis also reveals that both organisms contain transcripts that 61 encode a proton-translocating energy conserving hydrogenase, suggesting contributions of 62 previously unreported metabolic networks to MCFA production. Based on these new results, we 63 conclude that LCO1 represents a novel genus within the Lachnospiraceae family and propose the 64 name of Candidatus Weimerbacter bifidus, gen. nov., sp. nov. Our data also predicts that EUB1 65 represents a new species within the Pseudoramibacter genus, and we propose the name 66 Candidatus Pseudoramibacter fermentans sp. nov., to represent this new species. 67

Refinement of metagenome-assembled genomes (MAGs). 69
We previously reported the construction of draft MAGs from a MCFA-producing microbiome fed 70 with lignocellulosic biorefinery residues, in which LCO1 and EUB1 represented the abundant 71 chain-elongating microorganisms (8). These draft MAGs were constructed using DNA samples 72 from the first 120 days of reactor operation. To improve the quality of these MAGs, we obtained 73 Illumina and PacBio sequencing reads obtained from the same microbiome at different times 74 during a 378-day operational period. We co-assembled 244 million Illumina Hi-seq (2x250) reads 75 from five time points (days 96, 120, 168, 252, and 378; Fig. S1) into 24,000 contigs. Contigs were 76 binned into MAGs; the MAGs with relative abundance greater than 1% were then gap-filled with 77 PacBio reads from the day 378 sample. This analysis resulted in an overall improvement in MAG 78 quality with respect to completeness, contamination, and number of scaffolds ( Table 1) The Genome Taxonomy Database (GTDB) Tool kit (9) was used to provide a taxonomic 91 classification of LCO1.1 and EUB1.1 (Fig. 1) Concurrent analysis of the media over this time period showed that xylose and glycerol were 128 consumed whereas lactate transiently accumulated in the reactor (Fig. S2). We used this multi-129 omic analysis to investigate substrate utilization, the enzymes involved in converting substrates 130 into intermediates and end-products, the potential for MCFA production via the reverse β-131 oxidation pathway, and the predicted energy conserving features of the predominant MAGs in this 132 anaerobic microbiome. 133 Chain elongation in Cand. W. bifidus. Four enzymes are known or predicted to comprise the 134 reverse -oxidation cycle (Fig. 2). In the first step, acyl-CoA acetyl transferase (ACAT) condenses 135 an acetyl-CoA with an acyl-CoA; the product of this reaction is reduced by 3-hydroxy-acyl-CoA 136 dehydrogenase (HAD), followed by a dehydration by 2-enoyl-CoA with enoyl-CoA dehydratase 137 (ECoAH) and a reduction by an acyl-CoA dehydrogenase (ACD) to form an elongated acyl-CoA. 138 In some organisms, the last dehydrogenation reaction is catalyzed by an electron-bifurcating 139 energy-conserving enzyme where the enoyl-CoA reduction with NADH is paired with the 140 reduction of ferredoxin through an ACD complex containing the electron transfer flavoproteins 8 EtfA and EtfB (13). The Cand. W. bifidus genome has a gene cluster encoding ACAT, HAD, 142 ACD, EtfA, and EtfB (Fig. 3A), while a gene predicted to encode ECoAH is located on a different 143 region of the genome. The abundance of transcripts encoding reverse -oxidation enzymes was at 144 or above the 90 th percentile in all samples analyzed (Fig. 4A). Pairwise comparisons of reverse -145 oxidation transcript abundance show high correlations for genes in the cluster encoding ACAT, 146 HAD, ACD, EtfA, and EtfB over the course of this analysis, but low correlations with the ECoAH 147 transcript (Fig. S3) suggesting that this later gene is not co-expressed with those predicted to be 148 involved in the reverse -oxidation pathway of Cand. W. bifidus. 149 Producing reduced ferredoxin during reverse -oxidation constitutes a potential energy 150 conserving process (Fig. 2) because its subsequent oxidation can be used to create an ion motive 151 force by either the RNF complex (RnfABCDEG), as proposed for the MCFA-producing C. 152 kluyveri (14), or the energy conserving hydrogenase complex Ech, as described for C.  Time-series transcriptomic analysis showed that Cand. W. bifidus produced high levels of 162 transcripts encoding the RNF and the Ech complexes when MCFA were produced after addition 163 of lignocellulosic biorefinery residues (Fig. 4). Transcripts encoding subunits of the RNF complex (RnfABCDEFG) were present above the 90 th percentile at several time points after addition of 165 lignocellulosic biorefinery residues and found to be at or above median expression levels 166 throughout the study (Fig. 4B). Transcripts for genes encoding subunits of the Ech hydrogenase 167 complex follow a different pattern, with high level abundance only for the first 12 hours after 168 lignocellulosic biorefinery residues addition (Fig. 4C). We also found that transcripts of Ech 169 complex genes were more abundant than those encoding a putative periplasmic ferredoxin 170 hydrogenase (Hyd1), the only other hydrogenase predicted to be present in the Cand. W. bifidus 171 genome. Thus, the multi-omic data supports a role for both the RNF and Ech complexes during 172 MCFA production, likely by conserving energy via generation of an ion motive force (Fig. 2). 173 The reverse β-oxidation cycle is also predicted to require either a CoA transferase (CoAT) 174 or a thioesterase to remove the CoA from the terminal acyl-CoA molecule, thereby releasing the 175 corresponding acid (Fig. 2). During the course of this experiment, Cand. W. bifidus expressed 176 genes encoding one predicted CoAT and two predicted thioesterases. Transcripts of all three genes 177 were at or above median levels throughout the time course of this experiment (Fig. 4D)

Cand. W. bifidus is predicted to use multiple routes to consume xylose as a source of carbon for 189
MCFA production. Previous studies (8) and the revised genome provided in this work predict that 190 Cand. W. bifidus can metabolize xylose and other pentoses (Fig. 5A) as the main source of energy 191 and carbon when growing in the MCFA-producing microbiome (8). In this time-series experiment, 192 we found that transcripts encoding an ABC transporter predicted to transport multiple sugars were 193 present at or above the 99 th percentile at several time points, (Fig. 5B) suggesting a role for this 194 protein in sugar uptake. 195 To investigate potential routes for sugar utilization by Cand. W. bifidus, we compared 196 transcript abundance of genes encoding enzymes predicted to function in the pentose phosphate 197 and phosphoketolase pathways after the addition of lignocellulosic biorefinery residues (Fig. 5B). 198 Patterns of transcript abundance indicate that genes encoding enzymes for both pathways are 199 expressed above median expression levels, suggesting that both pathways are used for pentose 200 utilization and its subsequent conversion to intermediates that then enter the reverse -oxidation 201 pathway to produce MCFA. 202

Predicted conservation of metabolic functions between Cand. W. bifidus and related organisms. 203
We found that all of the UBA2727 genomes, Shuttleworthia satelles, and the LCO1.1 genome 204 contained genes needed for the reverse -oxidation pathway (Supplementary Data File 2). In the 205 UBA2727, S. satelles, and Cand. W. bifidus genomes, the genes encoding ACAT, HAD, ACD, 206 EtfA, and EtfB are clustered ( Fig. 3A-F), whereas ECoAH is located on a different region of the 207 genome. The presence of genes encoding each of these enzymes and their transcript abundance 208 patterns in Cand. W. bifidus during MCFA production suggest that reverse -oxidation and the 209 electron-bifurcating ACD reaction are conserved and key features of chain elongation in each of 210 these organisms. Genome comparisons revealed that the co-location of genes encoding ACD, 211 EtfA, and EtfB is a common, although not essential, topology among known MCFA-producing 212 organisms. Examples of other MCFA producers with this gene arrangement include Megasphaera 213 elsdenii (Fig. 3G), Clostridium kluyveri (Fig. 3H) and Ruminococcaceae bacterium CPB6 214 (Fig.3I). 215 While two routes of pentose utilization (the pentose phosphate and phosphoketolase 216 pathways) are predicted to be present in the Cand. W. bifidus genome, none of the UBA2727 217 genomes, nor the S. satelles genome, contain a predicted phosphoketolase gene. This indicates that 218 the use of multiple pentose consumption pathways may be a unique feature of Cand. W. bifidus 219 when compared to other members of the UBA2727 cluster and to the nearest type strain. All of the 220 UBA2727 genomes and Cand. W. bifidus contain genes for acetate production, including 221 phosphotransacetylase and acetate kinase, indicating that all of the species in this genus may 222 produce acetate as a product of anaerobic sugar metabolism. 223 The putative phosphoketolase from Cand. W. bifidus is similar to enzymes predicted to be 224 present in other firmicutes and -proteobacteria, including three species of Megasphaera (Fig.  225   S4), a genus containing known MCFA producers. The phosphoketolase pathway (Fig. 5A), termed 226 the "bifid shunt" (21) in a Bifidobacterium (22), provides an alternative to glycolysis and the 227 pentose phosphate pathway for sugar utilization. In this pathway, the phosphoketolase enzyme 228 splits xylulose-5-phosphate into glyceraldehyde-3-phosphate and acetyl-phosphate (23). Energy 229 can then be conserved through the phosphorylation of ADP and production of acetate by acetate 230 kinase (Fig. 5A). Overall, the phosphoketolase pathway can produce more ATP than the pentose 231 phosphate pathway per mol of xylose (Fig. S5) and it directs carbon to acetate in addition to 232 producing glycolysis intermediates (Fig. 5A). 233 Chain elongation by Cand. P. fermentans. We previously predicted that Cand. P. fermentans 234 consumed lactate and produced MCFA when this microbiome was supplied with lignocellulosic 235 biorefinery residues (8). In this study, we found that transcripts encoding enzymes predicted to 236 function in the reverse -oxidation pathway were among the most abundant in Cand. P. fermentans 237 after the addition of lignocellulosic biorefinery residues (Fig. 6A). However, in the revised Cand. 238 P. fermentans genome we find that the gene encoding the putative ECoAH protein is located near 239 genes predicted to encode enzymes in the reverse -oxidation cycle (Fig. 3), unlike the genomes 240 of Cand. W. bifidus and related organisms. In addition, we find that the abundance of the transcript 241 encoding the EcoAH protein correlates well with those of Cand. P. fermentans genes predicted to 242 encode other enzymes in the reverse -oxidation cycle (Fig S6). In Cand. P. fermentans, the genes 243 encoding the predicted EtfA and EtfB proteins are not located next to those encoding the ACD 244 enzyme. Instead, the etfAB genes are in a cluster with one that encodes a homologue of a putative 245 prephenate dehydrogenase (PRDH), an enzyme that catalyzes an oxidative decarboxylation in the 246 shikimate pathway for tyrosine biosynthesis (24) (Fig. 3). A pairwise gene expression analysis of 247 transcript levels encoding the ACD, PRDH, EtfA, and EtfB proteins (Fig. 7) showed that ACD 248 had strong correlations with EtfA (r 2 =0.93) and EtfB (r 2 =0.91), suggesting co-regulation of ACD 249 and the electron transport flavoproteins, and supporting a role for an electron-bifurcating ACD in 250 the reverse -oxidation cycle of Cand. P. fermentans, as we also predict for Cand. W. bifidus. 251 The revised genome sequence of Cand. P. fermentans also contained complete sets of genes 252 for subunits of the RNF and the Ech complexes, a characteristic also shared with Cand. W. bifidus. 253 were at or near the 90 th percentile throughout the time series analyzed in this experiment (Fig. 6B), 255 whereas the transcript abundance for the genes predicted to encode the Cand. P. fermentans Ech 256 hydrogenase were below the median at most time points after the addition of lignocellulosic 257 biorefinery residues (Fig. 6C). Furthermore, the other hydrogenase predicted to be present in 258 Cand. P. fermentans, Hyd1, had transcript abundances that were higher than those of genes 259 encoding the Ech hydrogenase complex. These observations suggest that energy conservation in 260 Cand. P. fermentans primarily occurs via generation of an ion motive force by the RNF complex. 261 However, hydrogen production via Hyd1 could be important to maintain redox balance during 262 chain elongation by Cand. P. fermentans. 263 When considering hydrolysis of the elongated acyl-CoA molecule, the Cand. P. fermentans 264 genome encodes a CoAT and two thioesterases, each of which were expressed at or above median 265 levels when compared to other transcripts analyzed during this experiment (Fig. 6D) Cand. P. fermentans is predicted to use multiple routes for glycerol metabolism. Glycerol is 275 known to be a significant carbon source in the lignocellulosic biorefinery residues used in this and 276 earlier studies (8). Analysis of the reactor media showed that all the glycerol was removed within 277 the first 6 hours (Fig. S2) suggesting it is a favored carbon source for one or more of the microbes 278 in this microbiome. The most highly expressed gene by Cand. P. fermentans at several time points 279 after addition of lignocellulosic biorefinery residues encodes a predicted glycerol transporter (Fig.  280   8B), suggesting that glycerol is rapidly transported by this organism. 281 To assess how Cand. P. fermentans metabolizes glycerol, we monitored transcript 282 abundance of genes predicted to be involved in this process. The revised Cand. P. fermentans 283 genome also predicts that this organism contains enzymes to metabolize glycerol, and the time 284 series transcriptomics data showed that it expressed genes for three putative glycerol conversion 285 pathways (Fig. 8). The first route predicted to be active in Cand. P. fermentans (Fig. 8A, Route 1) 286 uses an ATP-dependent glycerol kinase, in which the resultant glycerol-phosphate is oxidized to 287 produce dihydroxy-acetone phosphate (DHAP). The second predicted route in Cand. P. 288 fermentans (Fig. 8A, Route 2) uses glycerol dehydrogenase to oxidize glycerol to 289 dihidroxyacetone before it is phosphorylated to DHAP. In a third predicted route in Cand. P. 290 fermentans (Fig. 8A, Route 3), glycerol is converted to 3-hydroxypropanal (reuterin), a compound 291 proposed to inhibit growth of other microbes by inducing oxidative stress (26). All of the above 292 predicted glycerol utilization genes, with the exception of glycerol dehydrogenase, are expressed 293 above median levels throughout the 36-hour time series (Fig. 8B). Thus, it is possible that multiple 294 or even all three pathways (Fig. 8A) play a role in glycerol metabolism in Cand. P. fermentans. 295 In addition to expressing genes encoding pathways for multiple glycerol utilization routes, 296 Cand. P. fermentans contains a gene cluster encoding predicted propanediol utilization body 297 proteins along with genes encoding the multiple subunits of a glycerol dehydratase (Fig. 9). Thus, 298 it is possible that these predicted protein microcompartments may help protect Cand. P. fermentans 299 from a toxin like reuterin, or others, that are potential products of glycerol metabolism by one or 300 more routes in this organism. 301 Overall, our data provides new evidence that Cand. P. fermentans directs glycerol to central 302 carbon metabolism and potentially produces toxic compounds that may impact the growth and 303 abundance of other members in this microbiome. One possible outcome of glycerol metabolism to 304 intermediates that can be further transformed via glycolysis (Fig. 8)

Cand. P. fermentans is predicted to use an electron-confurcating lactate dehydrogenase for 313
MCFA production. We previously proposed that one member of this MCFA-producing 314 microbiome consumes lactate produced by other community members (8). The Cand. P. 315 fermentans genome assembled in this study contains a gene cluster encoding a lactate transporter 316 (lactate permease), lactate dehydrogenase (LDH), and ACD (Fig. 3J). All three of these genes are 317 expressed above the 90 th percentile at all time points after the addition of lignocellulosic 318 biorefinery residue (Fig. 10), supporting their role in MCFA production from lactate. 319 The amino acid sequence of the predicted Cand. P. fermentans LDH in this gene cluster is 320 most closely related to FAD-binding oxidoreductases contained in genomes from several related 321 organisms, including other Firmicutes and organisms in the Fusobacterium phylum 322 (Supplementary Data File 8). This Cand. P. fermentans LDH clusters with proteins from 323 organisms that consume lactate anaerobically (Fig. 11), including two bacteria known to convert 324 lactate to MCFA, Ruminococcaeae CPB6 and Megasphaera elsdenii (5, 7), and with the LDH 325 from A. woodii that uses electron confurcation to couple lactate and ferredoxin oxidation with 326 NAD + reduction to overcome the thermodynamic bottleneck of lactate oxidation (27). Pairwise 327 correlation analyses of transcript abundance (Fig. 8) for this Cand. P. fermentans LDH with EtfA 328 (r 2 =0.77) and EtfB (r 2 =0.65) suggest a role for this dehydrogenase and the electron transfer 329 flavoproteins in anaerobic lactate oxidation. Therefore, we hypothesize that in Cand. P. Cand. P. fermentans is similar to that found in the other two genomes available for 335 Pseudoramibacter organisms (Fig. 3). In all three cases, the gene encoding ECoAH clusters with 336 other genes in the reverse -oxidation pathway, while the etfAB genes are found in a separate 337 region of the genome. In all three organisms, the reverse -oxidation genes are located near the 338 genes encoding fatty acid biosynthesis (Fig. 3). While others have suggested a potential role for 339 fatty acid biosynthesis genes in MCFA production (28), our data do not provide support for this hypothesis since transcripts for gene encoding enzymes involved in reverse -oxidation were 341 orders of magnitude more abundant than fatty acid biosynthesis genes (Fig. S7). 342 The ability of Cand. P. fermentans to metabolize glycerol appears to be common in the 343 Pseudoramibacter genus, as genes for the three proposed routes of glycerol metabolism in Cand. 344 P. fermentans are also found in the other two Pseudoramibacter genomes. This is also the case for 345 the presence of genes predicted to produce protein microcompartments in a cluster with genes 346 encoding glycerol dehydratase (Fig. 9). The production of polyhedral protein microcompartments 347 thought to encase a diol dehydratase when consuming 1,2-propanediol has been described in other 348 organisms, such as Salmonella enterica (a -Proteobacterium) and Acetonema longum (Fig. 9D), 349 a member of the Negativicutes class within the Firmicutes phylum (29, 30). In addition to the two 350 other Pseudoramibacter genomes which contain identical glycerol gene clusters to Cand. P. 351 fermentans (Fig. 9B-C), a similar glycerol utilization cluster was identified in A. longum (Fig.  352   9D). 353

Summary descriptions of novel organisms 354
Description of genus Candidatus Weimerbacter, genus, nov. (Named for Paul J. Weimer, 355 a pioneer in using rumen microbes to produce valuable chemicals). The genus Weimerbacter 356 belongs to the Lachnospiraceae family. This genus contains bacteria from the cattle rumen and a 357 lignocellulose-fed anaerobic bioreactor. An isolate from this genus was obtained as part of the 358 Hungate 1000 project (10). Species within this genus contain genes needed for utilization of 359 sugars, acetate production, lactate production, reverse -oxidation, pyruvate flavodoxin 360 oxidoreductase, the RNF complex, and hydrogen production. Our analysis predicts that organisms 361 in the Weimerbacter genus can produce MCFA when using xylose as an organic substrate. A 362 predicted Weimerbacter energy conserving mechanism used during MCFA production involves 363 an electron-bifurcating acyl-CoA dehydrogenase (ACD); the reduced ferredoxin derived from this 364 activity is used by either the RNF complex or the Ech hydrogenase complex for ion translocation 365 that contributes to the creation of an ion motive force for ATP production. 366 Weimerbacter bifidus, sp. nov. (uses the bifid shunt, bifidus). 367

Description of Candidatus
Candidatus Weimerbacter bifidus is represented by the LCO1.1 metagenome assembled genome 368 (SAMN12235634). When a bioreactor community is fed lignocellulosic biorefinery residues, 369 Cand. W. bifidus is predicted to use the phosphoketolase pathway (bifid shunt) to degrade xylose 370 and produce butyrate and medium-chain fatty acids via reverse -oxidation. Cand. W. bifidus is 371 predicted to maintain redox balance by using a ferredoxin-dependent hydrogenase or an energy 372 conserving Ech hydrogenase to produce hydrogen. While other organisms within the Cand. 002396065.1), and it is differentiated from the P. alactolyticus species by an average nucleotide 379 identity lower than 95%. Unlike P. alactolyticus, both the EUB1.1 and the Pseudoramibacter sp. 380 (GCA 002396065.1) contain genes that code for a proton-translocating energy conserving 381 hydrogenase (EchABCDEF). When lignocellulosic biorefinery residues are fed to a bioreactor 382 community, Cand. P. fermentans is predicted to consume glycerol that is present in the residue 383 and lactate that can be produced by other microbial community members. This species is predicted 384 to produce MCFA, via reverse -oxidation, using lactate and glycerol as the organic substrates. 385 Cand. P. fermentans contains genes for multiple glycerol utilization routes, including conversion 386 of glycerol to reuterin. Lactate utilization is predicted to involve an electron-confurcating lactate 387 dehydrogenase (LDH). In addition, an electron-bifurcating acyl-CoA dehydrogenase (ACD) is 388 predicted to be used for energy conservation, with reduced ferredoxin produced by this enzyme 389 used by the RNF complex for the creation of an ion motive force to support ATP production. Cand. 390 P. fermentans is predicted to produce hydrogen via a ferredoxin-dependent hydrogenase as a 391 mechanism for balancing internal redox conditions. 392

Concluding remarks 393
Our results reveal several previously unexplored metabolic and energetic features of chain-394 elongating bacteria. Multi-omic analysis of Cand. W. bifidus suggest that this organism may have 395 a previously undescribed ability to use both the pentose phosphate and the phosphoketolase 396 pathways for pentose consumption and production of acetyl CoA that is needed for MCFA 397 synthesis. Further, both Cand. W. bifidus and Cand. P. fermentans may use multiple hydrogenases, 398 including a proton-translocating energy conserving hydrogenase (EchABCDEF) to support MCFA 399 production. Although both chain elongators contained genes for the RNF complex, the 400 transcriptomic evidence suggests that this complex may be more important for generating an ion 401 motive force in Cand. W. bifidus. Our data also predict that Cand. P. fermentans uses several 402 routes to consume glycerol as a carbon source, with potentially toxic intermediates sequestered in 403 protein microcompartments, and a thermodynamic analysis supports the ability of Cand. P. 404 fermentans to produce MCFA from this substrate. Finally, our data implicates an electron-405 confurcating LDH in providing carbon skeletons needed to support MCFA production from lactate 406 by Cand. P. fermentans. Further work is necessary to elucidate the implications of the genomic 407 features uncovered in this study on the bioenergetics of MCFA production, to assess whether 408 similar processes are involved in the production of MCFA from other substrates, and to develop a 409 better understanding of microbial pathways for production of additional valuable products from 410 renewable organic materials. 411

METHODS 412
Bioreactor operation. We operated a bioreactor containing 150 mL of liquid. Lignocellulosic 413 biorefinery residues, prepared as described previously (2) was added into the bioreactor and reactor 414 liquid was pumped out of the bioreactor every hour to maintain a residence time in the reactor of 415 6 days. The pH of the reactor was controlled at 5.5 by adding 5M KOH through a pump attached 416 to a pH controller. The temperature of the bioreactor was maintained at 35 °C using a water bath. 417 For the 36-hour time series experiment, 28 mL of liquid was removed from the reactor and 28 mL 418 of lignocellulosic biorefinery residues was added one time, bringing the starting liquid volume to 419 150 mL. 420 Reactor sampling. Prior to feeding the reactor with 28 mL of lignocellulosic biorefinery residues, 421 samples were collected for metagenomic and metatranscriptomic analyses. Samples for DNA 422 sequence analysis were collected in 2 mL centrifuge tubes and centrifuged at 10,000 g for 10 423 minutes. After decanting the supernatant, cell pellets were stored at -80 °C until DNA was 424 extracted. Samples for RNA were collected in 2 mL centrifuge tubes and centrifuged at 10,000 g 425 for 1.5 minutes. After decanting the supernatant, samples for RNA were flash frozen in liquid 426 nitrogen then stored at -80 °C until RNA was extracted. Three samples for RNA extraction and 427 sequencing were collected as a control prior to adding conversion residue. At each time point after 428 adding lignocellulosic biorefinery residues, one sample was collected for RNA extraction and the 429 supernatant from these samples was used for HPLC and GC-MS analyses, as described previously 430 (2). 431 DNA sequencing. DNA was extracted from biomass samples using a phenol-chloroform 432 extraction method described previously (8 Taxonomic assignments based on single-copy marker genes were made with GTDB tool kit (9) 447 and phylogenetic trees were constructed using RAxML (37). Average nucleotide identities were 448 calculated with JSpecies using the ANIb algorithm (38).   Cand. P. fermentans. The genes are ordered as they are predicted to be ordered in the Cand. P. 733 fermentans genome (Fig. 3J). 734 735