Evidence for a Putative Isoprene Reductase in Acetobacterium wieringae

ABSTRACT Recent discoveries of isoprene-metabolizing microorganisms suggest they might play an important role in the global isoprene budget. Under anoxic conditions, isoprene can be used as an electron acceptor and is reduced to methylbutene. This study describes the proteogenomic profiling of an isoprene-reducing bacterial culture to identify organisms and genes responsible for the isoprene hydrogenation reaction. A metagenome-assembled genome (MAG) of the most abundant (89% relative abundance) lineage in the enrichment, Acetobacterium wieringae, was obtained. Comparative proteogenomics and reverse transcription-PCR (RT-PCR) identified a putative five-gene operon from the A. wieringae MAG upregulated during isoprene reduction. The operon encodes a putative oxidoreductase, three pleiotropic nickel chaperones (2 × HypA, HypB), and one 4Fe-4S ferredoxin. The oxidoreductase is proposed as the putative isoprene reductase with a binding site for NADH, flavin adenine dinucleotide (FAD), two pairs of canonical [4Fe-4S] clusters, and a putative iron-sulfur cluster site in a Cys6-bonding environment. Well-studied Acetobacterium strains, such as A. woodii DSM 1030, A. wieringae DSM 1911, or A. malicum DSM 4132, do not encode the isoprene-regulated operon but encode, like many other bacteria, a homolog of the putative isoprene reductase (~47 to 49% amino acid sequence identity). Uncharacterized homologs of the putative isoprene reductase are observed across the Firmicutes, Spirochaetes, Tenericutes, Actinobacteria, Chloroflexi, Bacteroidetes, and Proteobacteria, suggesting the ability of biohydrogenation of unfunctionalized conjugated doubled bonds in other unsaturated hydrocarbons. IMPORTANCE Isoprene was recently shown to act as an electron acceptor for a homoacetogenic bacterium. The focus of this study is the molecular basis for isoprene reduction. By comparing a genome from our isoprene-reducing enrichment culture, dominated by Acetobacterium wieringae, with genomes of other Acetobacterium lineages that do not reduce isoprene, we shortlisted candidate genes for isoprene reduction. Using comparative proteogenomics and reverse transcription-PCR we have identified a putative five-gene operon encoding an oxidoreductase referred to as putative isoprene reductase.

Soils and marine environments harboring aerobic isoprene-degrading organisms serve as isoprene sinks (4,22). The fate of isoprene under anoxic conditions was examined previously by our group (23), whereby isoprene was found to act as a 2e 2 acceptor, with one C = C bond being reduced by an anaerobic enrichment culture to predominately 2-methyl-1butene. Sequencing of 16S rRNA gene amplicons from this culture revealed enrichment of Acetobacterium to 92 to 100% relative abundance, with Comamonadaceae accounting for the rest (2 to 8%). The homoacetogenic Acetobacterium lineage dominating the H 2 -fed enrichment used 1.6 mmol isoprene h 21 as an electron acceptor in addition to HCO 3 2 . Growth of the homoacetogen with isoprene produces 40% less acetate than with H 2 and HCO 3 2 alone, suggesting that its reduction to methylbutene is coupled to energy conservation (23). More recently, these results were further validated by an independent study with a pure culture of Acetobacterium named "strain Y" (24). Strain Y also transformed isoprene to predominately 2methyl-1-butene with a similar rate of 1.74 mmol isoprene h 21 (262.3 6 21.2 mM day 21 ). However, Jin et al. suggest that isoprene hydrogenation in strain Y is a cometabolic process and is not linked to energy conservation. Besides isoprene, strain Y was also found to reduce 1,3-butadiene to 1-butene. Homoacetogens are known to reduce electron acceptors other than CO 2 , such as fumarate (25), nitrate (26), chloroethenes and chloroethanes (27), brominated aromatics (28), and acrylate derivatives (29). For instance, reduction of the functionalized C = C bond (C = C bond conjugated to an electron-withdrawing group) in caffeate by the model organism Acetobacterium woodii (29-31) is a well-studied example for CO 2 -alternative electron acceptors in Acetobacterium species. Interestingly, pure cultures of A. woodii DSM 1030, A. malicum DSM 4132, and A. wieringae DSM 1911 showed no isoprene-reducing activity (23), suggesting that the isoprene hydrogenation capability is not a mutual trait among all Acetobacterium spp.
This study aimed to identify microorganisms and their corresponding genes/enzymes involved in the reduction of the unfunctionalized C = C bond in isoprene. DNA from an isoprene-reducing enrichment culture was sequenced, and protein profiles with and without isoprene were compared. Metagenomic and comparative metaproteomic analyses implicate a putative oxidoreductase in isoprene reduction encoded in a putative five-gene operon.

RESULTS
A putative oxidoreductase encoded by Acetobacterium wieringae is upregulated by isoprene. Cell suspension experiments with the Acetobacterium-dominated (relative abundance 16S rRNA gene amplicon sequencing of 92 to 100%) enrichment culture (23) pregrown on H 2 /HCO 3 2 indicated that isoprene reduction is induced in the presence of isoprene, H 2 , and HCO 3 2 (Fig. S1 in the supplemental material). Therefore, label-free comparative metaproteomics was performed to identify proteins and corresponding genes involved in isoprene metabolism. To generate a database for protein identification, the isoprene-reducing enrichment culture was grown on H 2 /HCO 3 2 /6isoprene, and the extracted DNA was sequenced. Over 7.5 million paired-end reads were used to assemble 338 contigs. Two nearly complete metagenome-assembled genomes (MAGs) were obtained (Table 1). No other lineages were detected. MAG ISORED-1 showed 74% average amino acid identity (AAI) and 79% average nucleotide identity (ANI) to Comamonas aquatica CJD (see Table S1 at https://doi.org/ 10.6084/m9.figshare.22012931), and ISORED-2 showed 97% AAI and ANI to Acetobacterium wieringae DSM 1911 (see Table S2 at https://doi.org/10.6084/m9.figshare.22012931). MAG ISORED-2 is dominant in the enrichment based on relative abundance (77%) and more so when cultivated in the presence of isoprene (relative abundance of 88.71%) ( Table 1). Differential metaproteomes were generated by liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis of proteins obtained from cultures grown on H 2 /HCO 3 2 with and without isoprene. A total of 1,531 proteins were identified. Consistent with the dominance of the homoacetogen in culture, 1,279 proteins (83.5%) belonged to A. wieringae ISORED-2 (Fig. 1B), and 252 proteins (16.5%) were assigned to Comamonas ISORED-1 (Fig. 1A). This is also reflected in the estimated biomass contributions, where A. wieringae FIG 1 Volcano plot of the metaproteomic data comparing cells grown on H 2 /HCO 3 2 /isoprene versus H 2 /HCO 3 2 . Significant data points (colored) are based on a LFC of 62 and an adjusted P value of #0.05. Labeling of the significant points is based on metagenome-assembled genomes ISORED-1 (A; Comamonas sp.) and ISORED-2 (B; Acetobacterium wieringae). Proteins located adjacent to each other in the genome of A. wieringae ISORED-2 (B) are highlighted. Data were obtained from 4 replicates for each growth condition (PRIDE database PXD023683). Differential expression analysis was conducted with limma (130).
Five out of 12 isoprene-responsive proteins from A. wieringae were more significantly abundant (FDR # 0.05), with a LFC of 6.2 to 10.7, than the remaining 6 (LFC of 2.5 to 4.8), and 1 protein was less expressed (VUZ24188.1, LFC of 22.56). Four (VUZ27132.1, VUZ27134.1, VUZ27135.1, and VUZ27136.1) out of these five are adjacent to each other in the ISORED-2 MAG, indicating that they might belong to the same operon ( Fig. 1B and  2A; Table 2). Protein VUZ27132.1 ( Table 2) is predicted to be a molybdopterin oxidoreductase, with the best orthologous group match ENOG4107QZ5. Protein VUZ27136.1 (ENOG4105WMM) is a HypA homolog, and VUZ27134.1 (ENOG4107RSS) is a HypB homolog. HypA and HypB proteins are typically responsible for the acquisition and insertion of nickel in the catalytic center of [NiFe]-hydrogenases; it should be noted here that the genome of A. wieringae ISORED-2 does not encode the structural genes for a [NiFe]-hydrogenase. Protein VUZ27135.1 (ENOG4105DQ9) belongs to the 4Fe-4S ferredoxin superfamily (SSF54862), but the sequence is not affiliated with any specific family. Protein VUZ23051.1 (ENOG41060FG, methyltransferase) is also highly expressed (LFC of 8.03) but is not part of the operon. Predicted protein functions from the six less isoprene-responsive proteins (LFC of 2.5 to 4.8) include acetyltransferase, xylulokinase, uroporphyrinogen decarboxylase, heat shock protein, formyltetrahydrofolate synthetase, and 50S ribosomal protein L35 (Table 2).
The putative isoprene operon. Apart from A. wieringae ISORED-2's oxidoreductase (VUZ27132.1), no isoprene-responsive protein from MAG ISORED-2 or ISORED-1 is predicted by protein function to be involved in redox processes (Table 2). This makes the oxidoreductase from the A. wieringae lineage the only likely candidate that could catalyze the isoprene hydrogenation reaction. The relevant gene of A. wieringae's oxidoreductase (VUZ27132.1) is organized in one operon together with the genes of three other isoprene-induced proteins (VUZ27134.1, VUZ27135.1, and VUZ27136.1) and an additional gene (encoding a HypA protein, VUZ27133.1) not significantly more abundant in the metaproteome (Fig. 1A). All five genes have the same orientation, with intergenic regions ranging between 7 and 71 nucleotides. Operon prediction analysis (FGENESB) suggested that the five genes are transcribed as an operon ( Fig. 2A; Fig. S2). Reverse transcription-PCR (RT-PCR) using primer sets flanking individual intergenic regions of adjacent genes (see Tables S3 and S4 at https://doi.org/10.6084/m9.figshare .22012931) also indicated that the genes are transcribed as an operon ( Fig. 2B and C). For promoter prediction analysis (BPROM), see Text S1 and Fig. S2.
Isoprene reduction and the putative isoprene-regulated operon is unique to ISORED-2 among known Acetobacterium species. While A. wieringae DSM 1911 is the same species as A. wieringae ISORED-2, it did not catalyze isoprene reduction nor  (23). Comparative pangenome analysis with eight publicly available Acetobacterium genomes (see Table S5 at https://doi.org/10.6084/m9.figshare.22012931) was used to assess the distribution of the putative isoprene-regulated operon encoding the oxidoreductase (VUZ27132.1) within the Acetobacterium genus and to find features unique to the A. wieringae ISORED-2 MAG. Pangenome analysis showed the protein-coding sequences from all nine genomes (33,035 in total) grouped into 8,190 gene clusters, based on an Markov Clustering Algorithm (MCL) inflation value of 6 (parameter controlling the granularity of the clustering) (Fig. S3). A shared set of 1,492 gene clusters (core) is shown across the nine genomes along with protein sets that are unique in each of the Acetobacterium genomes ( Fig. S3 and Table S6 at https://doi.org/10.6084/m9.figshare.22012931).
The putative isoprene-regulated operon (here referred to as the isr operon) is located between 69,745 and 75,048 bp in a 90,374-bp contig (ISORED_48, Fig. 3B). The first half of this contig contains mainly protein-coding genes of viral origin. The mean contig coverage and the coverage of the proviral portion are close to the values for the MAG, indicating no active viral replication. The provirus (Siphoviridae) showed an average amino acid identity of 59.11% with the Erysipelothrix phage U1605 (33) based on CheckV, and tBLASTx of many of the viral proteins also showed similar identities with several Streptococcus phages (34). This proviral region also appears in other Acetobacterium spp. genomes, including A. wieringae DSM 1911 and A. sp. KB-1 ( Fig. 3A and C). The contig contains three different Ser recombinases (integrases). Two of them . This high similarity is surprising considering this is a highly dynamic region due to the involvement of a prophage and a number of mobile genetic elements flanking the contig (e.g., recombinases and insertion sequences). This is also unexpected due to the geographically distant isolation sources. As the provirus and recombinases were the main sequences sharing some degree of synteny in the original contig with assembled Acetobacterium spp. genomes, the contig encoding the putative isr operon was scrutinized in more detail and the MAG reassembled to better understand the gene environment of the operon (Text S1). The reassembled ISORED-2 MAG was ;40 kbp larger, with a more polished putative isr operon contig ( Fig. 3B and D), which showed a higher degree of synteny with the genome of Acetobacterium sp. KB-1 (Fig. 3C). While the reassembly process was able to extend the putative isr operon contig, it did so at the expense of collapsing several insertion sequences (ISs). Most of the ISs in this region are suspected to appear in tandem repeats based on their higher apparent sequence coverage.
Predicted functional annotation of proteins encoded in the putative isr operon. Functional annotation to predict domains and important sites of the proteins encoded by the putative isr operon was performed with InterProScan (47) (see Table S9 at https://doi.org/10.6084/m9.figshare.22012931).
The first ;270 amino acids of protein VUZ27132.1 and protein VUZ27135.1 align to each other (BLASTp ID of ,26%), but no functional protein domain could be predicted from 1 to 270 amino acids by InterProScan or NCBI's Conserved Domain Database. The remaining part of VUZ27135.1 holds a predicted 4Fe-4S ferredoxin-type, iron-sulfurbinding domain (IPR017896) (see Table S10 (Fig. 5D). In addition, six cysteine residues were found topologically close (Fig. 5B to D), with four of them forming two distinct CxC motifs (Cys180-Cys182 and Cys254-Cys256). This is a less common Fe/S cluster-binding motif but is still found in some proteins, such as ferredoxin:thioredoxin reductase (52) or ISCA2 (53). Distances between the Fe/S clusters in electron transfer chains are usually ;8 to 15 Å to adjacent clusters (52,54,55), which is concordant with the distances measured between [4Fe-4S] clusters in IsrA (8.7 to 11.9 Å) (Fig. 5D), including between clusters IV and V (13.5 Å), which further supports the presence of an Fe/S cluster at the cluster V site as part of an electron transfer chain.

DISCUSSION
This study investigated the genetic basis for bacterial isoprene reduction activity previously observed in an isoprene-reducing enrichment (23). The bacteria in this enrichment culture were now identified as a Comamonas sp. and A. wieringae, named MAG ISORED-1 and ISORED-2, respectively. A. wieringae ISORED-2 dominates the isoprene-reducing culture (;89% relative abundance metagenome sequencing and ;94% relative biomass) ( Table 1) and like other Acetobacterium spp. encodes the Wood-Ljungdahl pathway (WLP) for autotrophic growth (56), the Na 1 -translocating ferredoxin:NAD 1 oxidoreductase (Rnf complex) (57), F 1 F 0 -ATPase (58), the electron transfer flavoproteins (59), and an electronbifurcating [FeFe]-hydrogenase (60). Although A. wieringae ISORED-2 dominates the isoprene-reducing culture, a Comamonas sp., which shows highest sequence similarity to Comamonas aquatica CJG (78.9% ANI and 74.5% AAI) (61), is also present (;11% relative abundance metagenome sequencing and ;6% relative abundance biomass) (Table 1). However, its relative abundance in H 2 /HCO 3 2 /isoprene-fed cultures (11%) was lower than in H 2 /HCO 3 2 -fed cultures (;23%) based on coverage values from metagenome sequencing ( Table 1), suggesting that these cells do not benefit from the inclusion of isoprene. Additionally, one of two proteins that were significantly more abundant following exposure to isoprene in Comamonas sp. ISORED-1 is SpoT (VUZ25726.1), a ppGpp synthetase/ hydrolase, indicating that the cells are experiencing nutrient stress. Bacteria respond to nutritional stress by producing (p)ppGpp, which triggers a stringent response, resulting in growth arrest and reallocation of cellular resources (32,62). In Escherichia coli, fatty acid starvation was found to induce (p)ppGpp accumulation synthesized exclusively by SpoT (63). SpoT interacts with acyl carrier protein (ACP) to likely induce a conformational switch that favors (p)ppGpp synthesis following fatty acid starvation (64). Interestingly, ACP was one of the significantly less abundant proteins in Comamonas sp. ISORED-1 in the presence of isoprene (Table 1), and proteins observed in the metaproteome included those for beta-oxidation of fatty acids (see Table S11 at https://doi.org/10.6084/m9.figshare .22012931). These results put forward that Comamonas sp. ISORED-1 is growing on necromass (e.g., fatty acids) and is experiencing stress in the presence of isoprene, which slows down cellular growth and metabolism via the (p)ppGpp stringent response. Taken together with the recent isolation of another isoprene-reducing A. wieringae strain (strain Y) (24) that shares high sequence similarities with A. wieringae ISORED-2 (99.5% ANI and 99.7% AAI) and is a pure isolate, it can be concluded that A. wieringae ISORED-2 is solely responsible for the isoprene reduction ability in this mixed enrichment culture.
Isoprene reduction was found to be an induced rather than constitutive trait, and comparative proteomics identified 13 significantly more abundant proteins following isoprene exposure. Apart from A. wieringae ISORED-2's oxidoreductase (VUZ27132.1), no isoprene-responsive protein from A. wieringae ISORED-2 or Comamonas sp. ISORED-1 is predicted by protein function to be involved in redox processes (Table 2). This makes the oxidoreductase from the A. wieringae lineage the only likely candidate within the 13 isoprene-responsive proteins that could catalyze the isoprene hydrogenation reaction.
The oxidoreductase is encoded in a putative five-gene operon together with the corresponding genes for three nickel-binding chaperones and one 4Fe-4S ferredoxin ( Fig. 2A). Four out of the five proteins encoded in this putative operon were also significantly higher in abundance following isoprene exposure in A. wieringae ISORED-2 ( Fig. 1) and are also found to be unique to the A. wieringae ISORED-2 genome by pangenomic comparison of selected available Acetobacterium genomes (see Table S7 at https://doi.org/10 .6084/m9.figshare.22012931). Because the closest relative A. wieringae DSM 1911 did not exhibit isoprene-reducing activity (23), it follows that genes encoding isoprene reduction most likely sit within this unique set. Apart from the corresponding genes for 4Fe-4S ferredoxin, two HypA proteins, and the oxidoreductase (VUZ27132.1), no other genes responding to isoprene are unique to the ISORED-2 MAG. Henceforth the operon will be referred to putatively as the isoprene-regulated operon (isr operon) and the oxidoreductase (VUZ27132.1) as the putative isoprene reductase or IsrA (gene name isrA).
Potential "isoprene reductase" candidates have also been shortlisted in recently discovered isoprene-reducing A. wieringae strain Y (24). A total of 44 putative ene-reductases (ERs) in strain Y were designated by Jin et al. as IsoR (standing for isoprene reductase) based on enzyme functionality predictions, but these suggestions were not substantiated by their experimental data. Using proteomic analysis, Jin et al. identified a candidate ER (LNN31_08025, which shares 100% nucleotide sequence identity with IsrA) for the isoprene reduction reaction, but it is not specifically referred to as "the" isoprene reductase since all 44 putative ERs in strain Y are named "IsoR." Like IsrA in A. wieringae ISORED-2, LNN31_08025 in strain Y is also encoded in a five-gene operon that has 100% nucleotide sequence identity with the isr operon in A. wieringae ISORED-2. Surprisingly, Jin et al. do not mention the operon nor the other proteins of the isr operon even though their proteomic data show significant abundance of all 4 proteins with LFC values among the highest in their data set (data set 3 in reference 24). Taken together, results from Jin et al. further validate that IsrA in A. wieringae ISORED-2 is the enzyme responsible for the isoprene hydrogenation reaction.
Based on domain predictions, IsrA contains a nested FAD and NAD(P)H binding site as well as two pairs of canonical [4Fe-4S] clusters (clusters I to IV, Fig. 5D) and one extra hypothesized Fe/S cluster in a Cys 6 -bonding environment (cluster V, Fig. 5D). The bestcharacterized and only crystallized proteins in the orthologous group of IsrA are the b-subunits of NADH-dependent ferredoxin-NADP 1 -oxidoreductases (Nfn) (Fig. 4A and B, evolutionary group 3). Nfn is an electron-bifurcating enzyme (65) composed of two subunits, NfnA (32.6 kDa) and NfnB (49.8 kDa). Crystal structures from Thermotoga maritima (TM_1640) and Pyrococcus furiosus (PF1327) revealed that NfnB contains two [4Fe-4S] clusters as well as binding sites for NADPH and FAD, with FAD being the site of electron bifurcation (38,66,67) (Fig. 5A). Since IsrA is predicted to contain a FAD/NAD(P)H binding site as well as five Fe/S clusters (four [4Fe-4S] clusters plus a putative fifth Fe/S cluster), and a 4Fe-4S ferredoxin (VUZ27135.1) is encoded in the putative isr operon, bifurcation may be a reaction mechanism to contemplate for IsrA. The standard redox potential of the isoprene/methylbutene couple is not known but based on calculation using the estimation of isoprene energy of formation 197 kJ mol 21 (23,68,69) and theoretical stoichiometries with H 2 as an electron donor for the isoprene hydrogenation reaction (68,70), the standard reduction potential (E 0 ) of the isoprene/methylbutene couple calculated using the Nernst equation is estimated as The Nernst equation for standard potential of biological systems at pH 7 is where 2.3 RT/F = 0.059 at T = 289 K, F = 96,500, and R = 8.31, yielding The standard electron potential of the hydrogen electrode is EH 2 = E 0 H 2 1 (0.059/1) log (H 1 ); E 0 H 2 = 0; EH 2 = 20.059 pH E 09 ¼ E 0 2 0:059 pH ¼ 0:709 V 2 0:059 Â 7 ð Þ¼ 1 296 mV: Hypothetically, similar to caffeate reduction, NADH (E 09 = 2320 mV) derived from the [FeFe]-hydrogenase and Rnf complex could act as the reductant for the exergonic reduction of isoprene to methylbutene (E 09 = 1296 mV), which is coupled to the endergonic reduction of ferredoxin (E 09 = 2420 mV) (71). As with caffeate respiration, the reduced ferredoxin could be reoxidized at the Rnf complex to generate an Na 1 gradient (72). Out of 12 known flavin-based bifurcating enzymes (65,71), three are found in Acetobacterium spp.: the bifurcating [FeFe]-hydrogenase (60), lactate dehydrogenase/ electron transfer flavoprotein (Bf-Ldh) (73), and the caffeyl-coenzyme A (caffeyl-CoA) reductase (72). However, homology to bifurcating enzymes is not sufficient to guarantee electron bifurcating functionality (65), but as energetics and the binding sites of IsrA support the idea of an electron bifurcating process, it should be considered in future biochemical investigations of IsrA. If isoprene reduction was a linear process, reduction would have to be coupled to ATP synthesis through establishment of an ion gradient since the reduction of isoprene conserves energy (23). In contrast, Jin et al. found no difference in acetate amounts between A. wieringae strain Y cultures with and without isoprene, concluding that strain Y cannot conserve energy from isoprene reduction but cometabolizes isoprene. Jin et al. measured acetate amounts during transformation of only 100 mmol of isoprene (4 days), whereas in our previous study, acetate was measured during transformation of 800 mmol of isoprene (30 days). The difference in acetate amounts was only seen after the transformation of at least 500 to 600 mmol of isoprene (see Fig. 7 in reference 23). Until comparable results of this experiment are published, isoprene reduction is suggested to be coupled to energy conservation.
Beside IsrA, the putative isr operon contains three hyp genes (two hypA and one hypB; hydrogenase pleiotropic), which encode metallochaperones typically responsible for acquisition and insertion of nickel during maturation of [NiFe]-hydrogenases (74)(75)(76)(77)(78). The role of HypA and HypB during [NiFe]-hydrogenase maturation is well studied and first involves the metal-dependent (Ni 21 ) dimerization of HypB with one equivalent of nickel per dimer (79,80). Second, the hydrolysis of GTP facilitates the transfer of Ni 21 from HypB to HypA by weakening the binding affinity of Ni 21 to HypB and promoting the formation of the HypAB heterodimer (81) or, in some organisms such as Thermococcus kodakarensis, a heterotetramer HypAABB (82). After the Ni 21 is transferred to HypA, HypA dissociates from the complex and delivers the cofactor to the large subunit of the hydrogenase (50,83). Biosynthesis and maturation of the [NiFe]hydrogenase active site is a complex multistep process also involving a number of other accessory Hyp proteins (HypCDEF) (84,85). Yet, the genome of A. wieringae ISORED-2 does not harbor hypCDEF nor the core structural genes of the [NiFe]-hydrogenase (i.e., large and small [NiFe]-hydrogenase subunits), questioning the presence of three Hyp proteins in the isr operon. The Hyp proteins could potentially be involved with other Ni-dependent enzymes, since the genome of A. wieringae ISORED-2 encodes three other known Ni-containing enzymes: two lactate racemases (86), ISORED2_01724 (VUZ23303.1) and ISORED2_03140 (VUZ26077.1) and the CO dehydrogenase/acetyl-CoA synthase complex (CODH/ACS) (87) at ISORED2_03659 to ISORED2_03664, with an additional beta subunit encoded by ISORED2_01878. However, these enzymes are located in operons that include their own metallochaperones/maturation proteins, that is, larE (ISORED2_00963 and ISORED2_01065) (86) and acsF (ISORED2_03663) (87), respectively. This suggests that the two HypA and/or HypB proteins from A. wieringae ISORED-2 are unlikely to be involved with other Ni-containing enzymes but rather facilitate nickel insertion into an active site of one of the proteins in the isr operon. Genome neighborhood computation results show that out of 988 IsrA homologs, 73% are encoded within 5 genes to HypA and 56% are to HypB whereas no association of HypA or HypB could be found with the 4Fe-4S ferredoxin (see Table S12 at https://doi.org/10.6084/m9.figshare .22012931). This suggests that IsrA is the target protein for potential nickel acquisition and that this type of HypA and HypB protein may be required for the enzymatic function of many of these clade 9 oxidoreductase homologs. Known nickel-binding sites in Nicontaining proteins involve cysteine, histidine, or acidic residues, and in catalytic nickelcontaining proteins, the nickel is in the site closest to the substrate (88). Prediction for substrate transport channels into IsrA suggests a tunnel that would terminate in the immediacy of the hypothesized site of Fe/S cluster V (Fig. 5D), which would hence be closest to isoprene. Thus, Fe/S cluster V could act as a potential nickel-binding site in IsrA since in other Ni-containing enzymes, Ni 21 is found closest to the substrate (e.g., [NiFe]-CODH/ACS and [NiFe]-hydrogenase) (88). In the [NiFe]-CODH and ACS, nickel is either next to the Fe/S cluster or substituting an Fe atom in the cluster, while in the [NiFe]-hydrogenase, it located close to the Fe of the Fe(CN) 2 CO group (88). Based on the protein model, the binding site of cluster V consists of six cysteine residues and has certain conformational resemblances with the Fe/S cluster binding sites in two other enzymes; on one hand, it resembles cluster A of the acetyl-CoA synthase (a Ni-Ni-[4Fe-4S] cluster) (89)(90)(91), and, on the other hand, it resembles the site of the P-type cluster in nitrogenases (55) or the similar double cubane [8Fe-9S] clusters (92). Interestingly, enzymes containing double cubane clusters have been shown to reduce small molecules such as acetylene to ethylene (92). Whether Fe/S cluster V does indeed accommodate the nickel-binding site in IsrA or not requires further biochemical characterization.
Promotor region analysis of the isr operon also supports the idea that IsrA might be a metal-dependent enzyme; four transcription factor (TF)-binding sites could be identified 13 bp upstream of the isrA gene start codon (Fig. S2). These are suggestive of binding sites for a ferric uptake regulator (Fur) or Ni(II)-dependent transcriptional regulator (NikR) type of TF (93,94). As NikR is not encoded in A. wieringae ISORED-2, it is more likely that Fur regulates the operon (ISORED2_03031). A. wieringae ISORED-2 encodes multiple nickel import systems. NikA (nickel transport system permease) (95) and related subunits are mainly on contig ISORED2_43 (ISORED2_01669, ISORED2_03327, ISORED2_03336, ISORED2_03347, and ISORED2_03353). The lactate racemases have specific nickel importers as part of their operon(s): a three-component ATP-binding cassette (ABC) transporter lar(MN)QO (86) (ISORED2_00944-00946). Export of heavy metals from cells can be performed by a diverse number of mechanisms (96). While no Ni-specific metal exporters (97) were detected in the genome of A. wieringae ISORED-2, three P-type IB ATPases were found: ISORED2_00744 (cadA: Cd 21 , Zn 21 , and Co 21 ), ISORED2_02650 (copA: Cu 1 ), and ISORED2_03115 (ziaA: Zn 21 ). Members of the P-type ATPase subfamily IB normally transport soft Lewis acids but often have limited specificity. Based on this, CadA or ZiaA might be the best candidates for P-type ATPase Ni 21 export given that CopA transports monovalent copper. The nickel import/export systems responsible for maintaining nickel homeostasis in A. wieringae ISORED-2 are yet to be identified, but since the main enzyme complex in the WLP (56,98), the CO dehydrogenase/acetyl-CoA synthase, is a nickel-dependent enzyme (99,100) and part of Acetobacterium's core metabolism, it is expected that nickel homeostasis would be well maintained in A. wieringae ISORED-2.
Homologs of IsrA are widely distributed among anaerobic bacteria, but the putative isr operon, as observed in A. wieringae ISORED-2 and in A. wieringae strain Y, was not found in any other genome in NCBI. Acquisition of the putative isr operon via horizontal gene transfer may be one possible scenario that explains why only these two strains harbor the putative isr operon. The operon is located in a 44-kbp genomic region containing metabolic genes and is also flanked by mobile genetic elements (Fig. 3), a Siphoviridae provirus, and a series of insertion sequences in tandem, which suggests that the putative isr operon is placed in a dynamic genomic region of Acetobacterium wieringae ISORED-2. Other organisms that also encode the complete putative isr operon, from where horizontal gene transfer could have occurred, are yet to be identified.
Homologs of IsrA observed in other Acetobacterium spp. share only ;47 to 49% amino acid sequence identity, and these homologs are located in separate subclades (Fig. 4C) and their corresponding genes are found in different gene arrangements (Fig. S5). Together with the inability of other Acetobacterium spp. (i.e., A. woodii DSM 1030, A. malicum DSM 4132, A. wieringae DSM 1911, and A. dehalogenans DSM 11527) to reduce isoprene, as determined experimentally, the phylogenetic analysis provides further evidence that IsrA and its homologs in other Acetobacterium spp. have distinct enzymatic functions. Potential enzymatic functions to consider for IsrA homologs are the hydrogenation of unfunctionalized (conjugated) C = C bonds in other unsaturated hydrocarbons that are present in anoxic environments. For example, Jin et al. found that strain Y could, besides isoprene, also reduce 1,3-butadiene to 1-butene (24). 1,3-Butadiene is an anthropogenic compound used mainly to produce polymers (101) entering the environment via combustion processes and industrial releases (102,103). Naturally occurring substrates to consider for the IsrA homologs could be terpenes (e.g., monoterpenes [C 10 H 16 ] a-pinene, b-pinene, limonene, trans-b-ocimene, a-terpinene, myrcene, and sabinene), which consist of isoprene building blocks. This might be the case for Pelotomaculum schinkii, which encodes the most closely related homolog to IsrA (Fig. 4C). P. schinkii is a strictly anaerobic, syntrophic bacterium known to live in electron acceptordepleted environments and metabolizes propionate and must resort to using H 1 and CO 2 as electron sinks (104). Degradation of propionate is thermodynamically challenging and can only be reached if H 2 or formate are kept at very low concentrations by a syntrophic partner methanogen (104). However, due to its IsrA homolog, P. schinkii might have the ability to use unfunctionalized (conjugated) C = C bonds in unsaturated hydrocarbons as electron acceptors, which could enable them to grow axenically. As a general example, the oxidation of propionate coupled to isoprene reduction would be thermodynamically favorable (23,68): This study provides evidence for the existence of a putative isoprene reductase. The putative isoprene reductase is of particular interest because of its reduction of an unfunctionalized conjugated C = C bond. IsrA homologs are widespread among various taxonomic groups of strictly and facultatively anaerobic bacteria (Firmicutes, Spirochaetes, Tenericutes, Actinobacteria, Chloroflexi, Bacteroidetes, and Proteobacteria), suggesting that the use of unfunctionalized C = C bonds in unsaturated hydrocarbons as anaerobic electron acceptors is a form of bacterial energy harvesting not previously recognized. While more rigorous physiological/biochemical testing is required to fully understand what the functions of IsrA and its homologs are, the results have environmental relevance in the context of furthering our understanding of electron sinks in anaerobic environments and furthering our understanding of contributing mechanisms to global isoprene turnover.
Cell suspension assays. Cells from six flasks of H 2 /HCO 3 2 /isoprene-grown cultures and six flasks of H 2 /HCO 3 2 -grown cultures were pooled in an anaerobic chamber by pipetting cell aggregates into two separate 6-mL anoxic glass flasks. Cells were washed in minimal medium containing 1 mM titanium citrate (23), and optical density at 600 nm (OD 600 ; 7.5) and volumes (1.57 mL) were adjusted between the two samples. Flasks were crimp sealed and flushed with N 2 for 30 min to remove isoprene, methylbutenes, and Putative Isoprene Reductase in A. wieringae mSystems CO 2 . Headspace was measured for isoprene and methylbutene before the experiment was started. H 2 (7 Â 10 4 Pa), HCO 3 2 (60 mM), and isoprene (1 mM) were added, and cells were incubated at 30°C (with shaking at 180 rpm). Headspace (100 mL) was analyzed for isoprene depletion and methylbutene production as previously described (23). Liquid samples (0.04 mL) were analyzed for acetate. Acetate was analyzed as its ethyl ester derivative by GC-flame ionization detector (GC-FID) as previously described (23) but with reduced sample size.
DNA extraction and Illumina sequencing. DNA was extracted from isoprene-reducing cultures anaerobically grown on H 2 /HCO 3 2 /6isoprene as described previously (23). Libraries were prepared using a Nextera XT DNA sample preparation kit according to the manufacturer's protocol (Illumina). Sequencing reactions were carried out using MiSeq v2 (2 Â 150 bp) chemistry (Illumina) on a MiSeq instrument (Illumina) at the Ramaciotti Centre for Genomics at University of New South Wales (UNSW; Sydney, Australia).
RNA extraction and reverse transcription-PCR. Cell aggregates from three flasks were pooled, centrifuged at 10,000 Â g for 10 min, and disrupted in lysis buffer (400 mL) (105) with mechanical agitation (30 Hz for 10 min) in FastPrep lysis matrix A tubes (MP Biomedicals). RNA was extracted with sequential phenol-chloroform-isoamyl alcohol (25:24:1; pH 4.5), 3 M sodium acetate (pH 5.2), and chloroform treatments, precipitated with isopropanol and GlycoBlue coprecipitant (Thermo Fisher Scientific, Australia), resuspended in 35 mL of water, and stored at 220°C. Residual DNA in RNA samples was digested with RNase-free DNase (Qiagen) I and cleaned three times on a spin column from a PureLink RNA minikit (Thermo Fisher Scientific, Australia). RNA was quantified with a Qubit RNA high-sensitivity assay kit (Thermo Fisher Scientific, Australia). RNA samples were stored at 280°C until use. First-strand cDNA was synthesized from 100 ng of DNase I-treated total RNA using random hexamer primers from the RevertAid first-strand cDNA synthesis kit (Thermo Fisher Scientific) following the manufacturer's instruction. In a negative control, the M-MuLV reverse transcriptase was replaced with water. Synthesized cDNA was used as the template in PCR with the Q5 high-fidelity DNA polymerase (New England BioLabs) using intergenic region primers (see Tables S3 and S4 at https://doi.org/10.6084/m9.figshare.22012931). Chromosomal DNA was used as template for the positive control (Fig. 2C).
Protein extraction and LC-MS/MS analysis. Cells were grown in 8 flasks with H 2 /HCO 3 2 /isoprene and 8 flasks with H 2 /HCO 3 2 . To increase cell mass, cells from two flasks were pooled from 8 to 4 samples for each condition, that is, 4 replicates for each condition. Cell aggregates were transferred into 2-mL tubes inside the anaerobic chamber, centrifuged at 10,000 Â g for 10 min, and stored at 220°C until use. Harvested cells suspended in 100 mL of lysis buffer (105) were mechanically disrupted in FastPrep lysis matrix A tubes (MP Biomedicals) at 30 Hz for 10 min. Crude extracts were passed through a 30-kDa Amicon Ultra 0.5-mL centrifugal filter and washed 6 times with 200 mL of 50 mM NH 4 HCO 3 buffer (pH 6.9). Protein concentrations were determined with the Quick Start Bradford protein assay following the manufacturer's instructions (Bio-Rad Laboratories, Australia) and adjusted to 2 mg mL 21 ; 10 mL (20 mg) was used for filter-aided sample preparation (FASP) (106)(107)(108). Samples were treated with 5 mM dithiotreitol (DTT) at 37°C for 30 min. Protein lysates were then transferred to 30-kDa Amicon Ultra 0.5-mL centrifugal filters and treated following the FASP method involving an alkylation step (100 mL of 50 mM iodoacetamide). Trypsin solution (1 mL of a 200 ng mL 21 stock) was added for digestions at 37°C overnight. Peptides were eluted in 2 Â 20 mL 50 mM NH 4 HCO 3 buffer and stored at 220°C until LC-MS/MS analysis.
A survey scan m/z 350 to 1,750 was acquired in the orbitrap (resolution = 120,000 at m/z 200, with an accumulation target value of 400,000 ions) and lockmass enabled (m/z 445.12003). Data-dependent tandem MS analysis was performed using a top-speed approach (cycle time of 2 s). MS2 spectra were fragmented by high-energy collisional dissociation (HCD; Normalised Collision Energy [NCE] = 30) activation mode, and the ion trap was selected as the mass analyzer. The intensity threshold for fragmentation was set to 25,000. A dynamic exclusion of 20 s was applied with a mass tolerance of 10 ppm.
MS data analysis. The raw MS data were processed using MaxQuant software (version 1.6.2.1) (126) and searched against a custom database of all predicted proteins in the metagenome of the isoprenereducing culture (6,517 sequences). Enzyme specificity was set to trypsin/P, cleaving C terminus to lysine and arginine, and a maximum number of two missed cleavages allowed. Carbamidomethylation of cysteine was set as a fixed modification, and oxidation of methionines and acetylation of protein N termini were set as variable modifications. The minimum peptide length was set to 7 amino acids, and a maximum peptide mass was 4,600 Da. The minimal score for modified peptides was 40, and the minimal delta score for modified peptides was 6. Peptide intensities were normalized using MaxLFQ (127). Downstream analysis was performed in R v3.5.1 with the package DEP v1.4.0 (128). First, MaxQuant output data were filtered, retaining only proteins detected by at least two unique peptides and detected in all replicates. To reduce the influence of the changing community composition and their relative contributions toward the total metaproteomic data, the metaproteomic data were partitioned based on the source MAG and analyzed separately in DEP as follows. Label-free quantification (LFQ) intensities were normalized with vsn (129), and missing values were imputed by left-censored imputation (MinProb function). Differential expression analysis was conducted with limma (130). Proteins were considered differentially expressed if they had an adjusted false-discovery rate (FDR) P value of #0.05 and a log 2 fold change (LFC) of $2 or #22.
The numbers of peptide spectrum matches per protein were used to quantify the biomass contribution of each organism to the community (131).
Ligand binding sites for FAD and NADPH were predicted with the COACH-D webserver (146). Prediction of ligand channels was performed on the Caver Web v1.2 (147).
First, each of the protein sequences was subjected to a BLAST search against UniProt to search for homologous proteins using EFI-EST with the following parameters: E value of 1 Â 10 25 , maximum number of sequences retrieved of 1,000, and "Superkingdom: Bacteria," "Superkingdom: Archaea," and "Superkingdom: Eukaryota" as taxonomy filters. The corresponding sequence similarity network (SSN) was then generated with an alignment score threshold of 30%. The SSN was used as input to generate a GNN with a neighborhood size of 5 and a cooccurrence lower limit of 20%. The SSN cluster hub nodes from the GNN outputs are listed in

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. TEXT S1, DOCX file, 0.03 MB.