Agr Quorum Sensing influences the Wood-Ljungdahl pathway in Clostridium autoethanogenum

Acetogenic bacteria are capable of fermenting CO2 and carbon monoxide containing waste-gases into a range of platform chemicals and fuels. Despite major advances in genetic engineering and improving these biocatalysts, several important physiological functions remain elusive. Among these is quorum sensing, a bacterial communication mechanism known to coordinate gene expression in response to cell population density. Two putative agr systems have been identified in the genome of Clostridium autoethanogenum suggesting bacterial communication via autoinducing signal molecules. Signal molecule-encoding agrD1 and agrD2 genes were targeted for in-frame deletion. During heterotrophic growth on fructose as a carbon and energy source, single deletions of either gene did not produce an observable phenotype. However, when both genes were simultaneously inactivated, final product concentrations in the double mutant shifted to a 1.5:1 ratio of ethanol:acetate, compared to a 0.2:1 ratio observed in the wild type control, making ethanol the dominant fermentation product. Moreover, CO2 re-assimilation was also notably reduced in both hetero- and autotrophic growth conditions. These findings were supported through comparative proteomics, which showed lower expression of carbon monoxide dehydrogenase, formate dehydrogenase A and hydrogenases in the ∆agrD1∆agrD2 double mutant, but higher levels of putative alcohol and aldehyde dehydrogenases and bacterial micro-compartment proteins. These findings suggest that Agr quorum sensing, and by inference, cell density play a role in carbon resource management and use of the Wood-Ljungdahl pathway as an electron sink.

). In the former three, a spo0E-like gene is also present. The aforementioned species also contain the agrD2C2 operon, although the surrounding genomic context deviates for C. coskatii, C. ragsdahlii and C. scatologenes. The C. autoethanogenum's AgrD1 and AgrD2 AIPs share considerable similarity in protein sequences amongst the above examples, especially in the conserved Pro-X-X-Pro motif region that has been suggested to represent the recognition site in which AgrD AIPs interacts with AgrB1 55 . However, each species AgrD AIP has a unique 5-membered lactone ring structure which is thought to bind to respective AgrC receptors 48,56 (Supplementary information, Table S5).

Generation and complementation of agrD mutants.
To study the role of the Agr system in C. autoethanogenum, both agrD signalling genes were targeted for in-frame deletion using allelic-exchange techniques to abolish putative signalling capabilities [57][58][59] . Genes agrD1 (CLAU_0814) and agrD2 (CLAU_3176) in each agr operon were either targeted individually or in tandem. In-frame deletion methods removed approximately 80% of each agrD gene, leaving a 30-bp non-functional sequence. The starting point for this strategy was a pyrE defective C. autoethanogenum strain, which due to this defect displayed uracil auxotrophy and 5-fluorouracil resistance 60 . The reason for using this strain lay in the ease with which chromosomal complementation of the introduced agrD deletions could be achieved: repair of the pyrE defect in the final step of the procedure restored uracil prototrophy and could be done either on its own or coupled to co-integration of an intact copy of the previously targeted agrD gene immediately downstream of pyrE. Hence, the final strains carried the described agrD modifications, but were otherwise identical to the wild type (WT). Three agrD mutant permutations were created in this manner consisting of two single knockouts, and a double knockout respectively: ΔagrD1, ΔagrD2 and ΔagrD1D2. Two further ΔagrD1D2 independent mutants were created later in the study, and designated ΔagrD1D2_1 and ΔagrD1D2_2 and were compared against the original double knockout mutant and WT. As stated above, each mutant was pyrE corrected to regain uracil autotrophy before full phenotypic characterisation 57,61 (Supplementary information, Fig. S1). Chromosomal complementation was performed using agrD double mutant, ΔagrD1D2_ΔpyrE, which was conjugated with pyrE repair plasmids, pMTLCH20-D1 and pMTLCH20-D2 that contained agrD1 and agrD2 Figure 1. Schematic of the proposed Agr QS systems in C. autoethanogenum. Two separate, putative agr operons are present in the genome. Both are incomplete: operon 1 is missing the response regulator encoding gene agrA, whereas operon 2 is lacking both agrB1 and agrA homologs. The precursors of two distinct autoinducing peptides (AIPs) are encoded by agrD1 and agrD2, respectively. Export and processing of both precursors is hypothesised to be mediated by membrane associated AgrB1. AgrC1 and AgrC2 are putative histidine kinases which act as receptors for AgrD1 and AgrD2 derived AIPs, respectively. In other Agr systems, their role is to phosphorylate their cognate AgrA response regulators. In C. autoethanogenum, these may be encoded elsewhere on the chromosome. A spo0E-like gene (CLAU_0813) is located downstream of agrB1, this arrangement is also observed in several other acetogenic clostridia. www.nature.com/scientificreports/ genes, respectively. Either agrD gene was chromosomally integrated downstream of the simultaneously repaired pyrE gene and was expressed via the pyrE operon promoter instead of its native agr promoter. Successful construction of mutants and complemented mutants was routinely verified by PCR and Sanger-sequencing, and further confirmed through whole genome sequencing as described in the methods. Complemented mutants were designated ΔagrD1D2 + D1comp and ΔagrD1D2 + D2comp respectively and phenotypically characterised.
Phenotypic characterisation of agrD mutants in heterotrophic conditions. Heterotrophic growth and fermentation profile assessments of the ΔagrD1, ΔagrD2 and ΔagrD1D2 mutants against the WT were performed for cultures grown in PETC medium containing 50 mM d-fructose at 37 °C for approx. 175 h. Transition from exponential to stationary phase was typically seen after 48 h and peak optical densities were similar for all strains ( Fig. 2A).
Determining fermentation product profiles focused on ethanol and acetate titres (Fig. 2B,C), whereas fructose consumption was determined at specific time points of 12, 31, 56 and 159 h (Fig. 2D). Overall, acetate production profiles were similar for single knockout mutants ΔagrD1 and ΔagrD2 and the WT, and although titres appeared slightly lower for ΔagrD1 there were no statistical differences against the WT, with final concentrations of acetate comprising of 53.3 ± 4.4 mM (P = 0.7), 57.3 ± 7.3 mM (P = 0.9), and 55.8 ± 3.6 mM, respectively (Fig. 2B). Similarly, no major differences were observed in final ethanol concentrations of ΔagrD1 and ΔagrD2 and the WT that reached 19.5 ± 3.6 mM (P = 0.17), 16.1 ± 3.1 mM (P = 0.42) and 11.8 ± 0.9 mM, respectively (Fig. 2C). However, notable differences were observed for the ΔagrD1D2 mutant in comparison to the WT. Ethanol concentrations increased fourfold to a final concentration of 48.9 ± 10.3 mM, whereas the final acetate concentration was around 1.7-fold lower, reaching only 33.2 ± 6.4 mM. Fructose consumption was measured using HPLC at specified time points and was noted to be similar amongst all strains, although the ΔagrD1D2 had consistently consumed all Red inverted triangles, ΔagrD2 (n = 4); Green squares, ΔagrD1D2 (n = 6). Colour representations corresponds similarly to bar graph (D). Error bars indicate standard error of the mean, * equates to a ≤ 0.05 P value of statistical difference to the WT.  Fig. 2D. Other fermentative products such as lactate and 2,3-butanediol were only produced in very small amounts, (final concentrations ranging from 4 to 6 mM respectively) and showed no significant differences amongst strains (not shown). CO 2 is liberated as a by-product during fructose-based fermentation and is partially reassimilated for WLPdependent acetate formation 42 . Serum flasks were routinely monitored for total pressure with a handheld pressure gauge. Visual inspection and sampling of flasks had suggested that ΔagrD1D2 flasks in contrast to single mutants and WT, had built up higher total gas pressures towards the end of fermentation experiments. When pressure was checked in one particular experiment, total pressures of 159, 114 and 115 kPa were found for ΔagrD1D2, ΔagrD2 and WT respectively (the ΔagrD1 mutant was not analysed during this particular experiment). A GCbased analysis of the serum flask head space was performed, and CO 2 observed at significantly higher partials pressures in ΔagrD1D2 flasks which reached 99.0 ± 2.1 kPa, compared to the WT and the ΔagrD2 mutant which only reached 33.4 ± 1 kPa and, 36.3 ± 1 kPa, respectively. To provide further evidence that these results were due to the complete inactivation of Agr QS, two further double knockout mutants (ΔagrD1D2_1 and ΔagrD1D2_2) were created. The original ΔagrD1D2 mutant was obtained by introducing the ΔagrD2 mutation into mutants with a ΔagrD1 background, whereas the ΔagrD1D2_1 and ΔagrD1D2_2 mutants were generated by inactivating agrD1 in the ΔagrD2 background. The headspace composition produced by each mutant was measured after 200 h and an increase of CO 2 partial pressure was again observed for the original ΔagrD1D2 double mutant, as well as the two newly generated ΔagrD1D2_1 and ΔagrD1D2_2, with 78.2 ± 0.4 kPa (P = 0.02), 76.4 ± 3 kPa (P = 0.02) and 70.0 ± 4.6 kPa (P = 0.19), respectively compared to 56.2 ± 5.7 kPa for the WT (Fig. 3A). Traces of H 2 were also measured ( Fig. 3A), serum flasks contained a starting H 2 pressure of 3.7 kPa, but the resolution in differences in  Complementation of ΔagrD1D2 mutants. As single agrD knockout mutants showed no discernible phenotypic differences compared to the WT, only the ΔagrD1D2 double knockout mutant was complemented.
As detailed above, this was done by inserting intact copies of either agrD1 or agrD2 immediately downstream of the pyrE gene on the chromosome, thereby generating strains ΔagrD1D2 + D1comp and ΔagrD1D2 + D2comp, respectively. The tested strains were then examined in PETC medium containing 50 mM D-fructose and their fermentation profiles were compared to the WT and ΔagrD1D2 mutant. As shown in Fig. 4A, complementation with agrD1 resulted in acetate and ethanol titres similar to those observed for the WT: the final acetate and ethanol concentrations reached 63.1 ± 3 mM and 25.6 ± 0.9 mM, respectively for the ΔagrD1D2 + D1comp strain, compared to 67.4 ± 0.4 mM and 18.5 ± 0.7 mM respectively for the WT. However, titres for the agrD2 complemented strain, ΔagrD1D2 + D2comp matched closer to the original ΔagrD1D2 mutant, with 42.6 ± 1.7 mM acetate and 32.1 ± 1 mM ethanol, respectively. Similarly, headspace gas analysis at the end of the fermentation showed that CO 2 partial pressures for ΔagrD1D2 + D1comp cultures (86.5 ± 2 kPa) were closer to the WT (66.3 ± 1 kPa), whereas those for the ΔagrD1D2 + D2comp strain (115. 7 ± 3 kPa) resembled the pressures observed for the ΔagrD1D2 mutant, (108. 7 ± 6.3 kPa) (Fig. 4B). Thus, it appears that chromosomal expression of agrD1 from the pyrE locus was able to closely restore WT characteristics in the double mutant, whereas for unknown reasons this was not the case for the agrD2 complemented strain.

Comparison of ΔagrD1D2 and WT proteomes.
To better understand the metabolic shift to increased ethanol formation for fructose-grown cells, the ΔagrD1D2 mutant's proteome was compared to that of the WT by LC-MS/MS-based iTRAQ quantitation. Cultures were grown on 50 mM fructose in PETC medium and cells were harvested when reaching an OD 660 of approx. 1.0, as cultures established their typical fermentation patterns (compare Fig. 2). Extracted peptides from the biological duplicates of WT and ΔagrD1D2 cells were labelled using iTRAQ 4-plex and analysed with MaxQuant software (see Supplementary information). From the 4-Plex iTRAQ experiment, 843 proteins were identified with quantification information with 2 unique peptides at 1% false discovery rate (FDR). Of these 843 proteins, 282 proteins were determined to be statistically differentially regulated (P < 0.0001), with Posterior Error Probability (PEP) of < 0.05 relating to peptide expression levels. Within these 282 quantified proteins (full list in Supplementary information, Sect. 3), 39 proteins with differential abundances relating to C. autoethanogenum's metabolism were categorised into WLP, pyruvate metabolism, glycolysis, energy conservation and fermentation end-product groups (Fig. 5). Enzymes central to both branches of the WLP, the CODH/ACS with subunit complexes (CLAU_1578 and 1579) and CODH/ACS complexes (CLAU_2924 and 1566), showed reduced abundance for the ΔagrD1D2 mutant. Several hydrogenase subunits (CLAU_2717, 2719, 2721 and 2722) representing HytC, HytD, HytA, HytE2 subunits respectively, encoded by the hytCBDE1AE2 cluster, were also observed to be in low abundance alongside a subunit of the [FeFe]-hydrogenase (CLAU_1535). Furthermore, formate dehydrogenase A (CLAU_2713; FdhA), known to form a complex with HytABCDE1E2 (and encoded as part of the hyt cluster), was also found in low abundance in the mutant. To independently confirm some of these changes, the overall specific activities for CODH and hydrogenase were determined using crude cell lysates derived from fructosegrown cultures of ΔagrD1D2 and WT as described in the Methods section. Assays involving hydrogenases and CODH used methyl viologen (MV) as an electron acceptor. WT hydrogenases showed eightfold higher specific activity than ΔagrD1D2 with 28.6 ± 2 U/mg to 3.5 ± 0.7 U/mg, respectively (P = 0.00006). Similarly, CODH specific activity was twofold higher in the WT than ΔagrD1D2, with 1.32 ± 0.1 U/mg to 3.5 ± 0.7 U/mg (P = 0.003).
Enzymes involved in alcohol metabolism were divided into two distinct groups: reduced and enhanced abundances for the ΔagrD1D2 mutant. Bifunctional acetaldehyde-CoA/alcohol dehydrogenases AdhE1 and AdhE2 (CLAU_3655 and 3656), alcohol dehydrogenase (CLAU_1794) and aldehyde ferredoxin oxidoreductases (AOR) www.nature.com/scientificreports/ (CLAU_0099 and 0089) all were found in low abundance. The latter three enzymes are noted to play a role during autotrophic metabolism 31,34 . Enhanced abundance of enzymes involved in alcohol formation included alcohol dehydrogenases (CLAU_1766 and 3861) and three, mono-functional, acetaldehyde dehydrogenases (CoA acetylating) (CLAU_3204, 1783 and 1772) 59 (Fig. 5), the genes of which are all linked to clusters encoding putative bacterial microcompartments (BMC). Given that some of the enzymes predicted to contribute to ethanol production were increased in the ΔagrD1D2 mutant whereas others were reduced, overall NADH-dependent alcohol dehydrogenase activities were determined. Alcohol dehydrogenase specific activity was measured using acetaldehyde as a substrate, with NADH as an oxidative cofactor, and showed threefold higher activity in ΔagrD1D2 in comparison to the WT, with 0.015 ± 0.002 U/mg (P = 0.002) and 0.005 ± 0.001 U/mg, respectively. Specific activity measurements were all statistically significant and agreed with the observed reduced abundance of WLP enzymes. Interestingly, predicted BMC proteins encoded by one particular cluster were found to have enhanced abundances in this analysis for the ΔagrD1D2 mutant (CLAU_1769, CLAU_1772, CLAU_1773, CLAU_1775 and CLAU_1785), along with 1,2-propanediol and two ethanolamine-utilising enzymes, PduL, EutQ and EutJ (CLAU_1771, CLAU_1774 and CLAU_1779) respectively. A second BMC cluster with enhanced abundances, represented another BMC associated protein together with further predicted 1,2-propanediol and ethanolamine utilising enzymes (CLAU_3206, CLAU_3202 and CLAU_3205). Finally, uncharacterised proteins CLAU_1786, 1780 and 1788 were noted for their enhanced abundances in the mutant, and proximity to BMC clusters. The first and third were related to PadR-type transcriptional regulators, involved in phenolic-acid stress 62,63 , the second uncharacterised protein shared high similarity with C. ljungdahlii's ethanolamine utilisation cobalamin adenosyl transferase (EutT), involved in ethanolamine metabolism 64 . Phosphate acetyltransferase (Pta), CLAU_3274,  www.nature.com/scientificreports/ involved in acetate formation was expected to be in low abundance in the ΔagrD1D2 mutant, but interestingly did not appear to be significantly differentially regulated between the mutant or WT. Other notable fold differences in protein abundance were identified in several transcriptional regulators (Supplementary information, Table S6), but do not directly link to the observed ethanol/acetate and gas-assimilation phenotypes in ΔagrD1D2 mutants.

Discussion
Our understanding of clostridial cell-to-cell communication systems is still limited although several common roles have been characterised. Agr-type QS systems are known to influence sporulation in the few species that have been studied 11,50,52 and in C. botulinum and C. perfringens they also contribute to the regulation of toxin production 51,54 . In two industrially relevant clostridia, C. acetobutylicum and C. saccharoperbutylacetonicum RRNPP-type systems are involved in the control of solvent metabolism and sporulation 9,10 . As will be argued below, the Agr QS system of the acetogen C. autoethanogenum may plausibly form part of a resource management mechanism through evaluating population density with respect to environmental conditions that include carbon source availability. Here we have shown that complete removal of both agrD signalling genes resulted in striking changes in both fermentative profiles and gas assimilation capabilities of the ΔagrD1D2 mutant. Moreover, notable fold differences were observed in comparative proteome analysis between the ΔagrD1D2 mutant and WT that included several key WLP related enzymes and a specific cluster of alcohol and aldehyde dehydrogenases. Initial observations showed that single knockout agrD mutants did not exhibit any discernible phenotypic differences with respect to the WT. Full length AgrD1 and AgrD2 share an overall identity of 70.2% but are predicted to be processed into unique cyclic structures that are presumably recognised by distinct AgrC receptors. This raises questions as to why two relatively similar peptide signals exist and what possible function each serves. Further insight can be derived from supplementary transcriptomic data from the study done by Marcellin et al. revealing that the RNA-seq transcript counts of agrD1 were fourfold higher in fructose-grown conditions (48.0 vs. 11.7), and sevenfold more in autotrophic conditions (162.4 vs. 22.2) than agrD2 34 . These different expressions can be speculated to act in a combinatorial mode of function, meaning that the two separate signalling peptides function separately but simultaneously, in cultures with high population density, cell diffusion and fluctuating mass transfer 65 . This can effectively improve an organism's ability to dynamically adapt and prepare in ever-changing environments through precise gene expression mediation 65,66 . An acetogen able to mediate its WLP genes during specific growth conditions may spare valuable ATP from unnecessary protein synthesis during heterotrophic growth 67 . For instance, by integrating information on population size and available organic carbon sources, C. autoethanogenum may be able to determine more precisely when to increase costly expression of WLP enzymes, before organic substrates are depleted. Regarding this study, it can be assumed that disabling only either one of these AgrD AIPs merely reduces this spectrum, leaving the Agr system largely functional. It was also reasoned that AgrD1 and AgrD2 AIPs must be processed using the same AgrB1-mediator encoded by the single agrB1 gene present within the genome, implying that AgrD2 peptides rely on expression of System 1. However, AgrD AIP detection is likely performed separately using either AgrC receptor from each respective agr operon (Fig. 1). Whether intracellular signalling pathways converge to a single, yet unidentified AgrA response regulator, or to the Spo0E-like regulator encoded downstream of agrB1, can be only speculated. A single AgrA response regulator could explain the absence of a clear phenotype in each single knockout agrD mutant by possibly allowing either AgrD AIP to compensate for the other missing AIP. Alternatively, regulatory feedback loops might lead to increased production of one of the AIPs, if the other one is lacking.
As neither single knockout mutant showed discernible phenotypic differences, efforts were focused to restore to a WT-like phenotype in the ΔagrD1D2 mutant with each agrD gene. Figure 4 showed that this restoration was partially successful with acetate, ethanol titres and gas re-assimilation for the ΔagrD1D2 + D1comp strain, but not for ΔagrD1D2 + D2comp strain. This partial complementation may be explained by changes to the respective agrD expression profiles and levels 34 , given that complementing agrD genes were expressed via the pyrE operon promoter instead of their respective native promoters. Studies examining the Agr system of C. acetobutylicum demonstrated that mutagenic disruption of agrA, agrB1 and agrC in this organism resulted in a reduction of heat-resistant endospores and loss of granulose storage compounds 11 . C. autoethanogenum, however, although a reported spore-former 22 , was not observed to produce spores under the conditions employed in this study (see "Materials and methods"), and granulose storage homologs are not present in its genome, and thus could not be used as additional phenotypic markers 68 . Phenotypic characterisation of the generated ΔagrD1D2 mutants thus primarily concentrated on fermentation profiling and autotrophic growth on CO.
With respect to WT and mutant fermentation metabolism, we reproducibly observed high ethanol titres for the three independently generated ΔagrD1D2 mutant, along with accumulated CO 2 headspace pressures. A similarly high ethanol output was noted in Liew et al. 's study after CODH subunits, encoded by acsA (CLAU_1578-79), were genetically disrupted, resulting in a twofold ethanol increase, abolished acetate production, near-absent autotrophic growth and approx. fourfold accumulated CO 2 headspace pressure 44 . The increased ethanol was reasoned to be the result of ethanol generation serving as the primary electron sink for the surplus reducing equivalents generated during glycolysis that could not be utilised by CODH/ACS. The similarities in results were intriguing but it must also be noted that ΔagrD1D2's heterotrophic growth on fructose was not affected unlike the ΔacsA mutant which exhibited 61% lower stationary phase OD compared to the WT 44 . This distinction may raise an important clue as to how CODH/ACS and other key WLP enzymes are regulated as noted in previous studies, with transcriptomic data revealing CODH/ACS having relatively high transcript counts, even during heterotrophic conditions 34 . Therefore, it can be assumed that the ΔagrD1D2's lack of peptide signalling capability prevents it from accurately mediating when to express CODH/ACS and potentially other components of the WLP pathway, allowing for unhindered growth on sugar but with a reduced capacity to fully re-assimilate CO 2 . This phenomenon was further studied by growing ΔagrD1D2 autotrophically (Fig. 3 B-D) www.nature.com/scientificreports/ considerably lower to that of the WT. Only after stepwise adaptation was the ΔagrD1D2 mutant eventually restored to WT growth and product titre levels.
Comparison of ΔagrD1D2's proteome to that of the WT supported observations of diminished CO 2 assimilation capacity with the most negative fold differences found in CODH/ACS subunits, along with several hydrogenase subunits that associated to the WLP, as well as the formate dehydrogenase alpha subunit (CLAU_2713). Interestingly yet counterintuitively, a cluster of ethanol-forming enzymes were also noted to be downregulated, some of which are expressed during autotrophic growth 34 , and involved in acetate reduction to ethanol via AORs (CLAU_0089 and 0099) 69 . Conversely, the most positive fold changes included several putative aldehyde and alcohol dehydrogenase enzymes that may explain the high ethanol yields. Further, the highest overall positive fold changes belong to BMC clusters outside of the WLP. Although unexpected, this observation is perhaps unsurprising as BMC located enzymes sequester and metabolise toxic intermediary compounds and are reasoned to be concomitantly expressed with alcohol metabolism in other species [70][71][72] . Regarding our proteomics approach, the iTRAQ method employed in this study has its own limitations resulting from complex bacterial lysate mixtures, that may result in underestimating protein abundance 73,74 . Ultimately, this analysis was used as a comparative indicator of direction of positive or negative protein fold changes. To independently verify, we employed three enzyme assays that targeted the most negative and positive fold changes from our proteomics data (with the exception for BMCs). These assays confirmed ΔagrD1D2's lower activity of CODH and hydrogenases, and slightly higher activity of alcohol dehydrogenases, and agreed with the ΔagrD1D2 mutant's phenotype.
To date, no studies have examined the role of QS in syngas-utilising acetogens despite complete and partial Agr systems being conserved across several industrially important species. Previous studies have established that these systems serve crucial functions in non-pathogenic Clostridium sp., such as the well-studied solventogen, C. acetobutylicum 10,11 . These studies emphasise the need of examining QS systems in industrially relevant biocatalysts. QS is a means of cell-to-cell signalling that ultimately aids in the organism's survival. This system can be triggered through nearby stressors that allow organisms to metabolically prepare and acclimate to changing environments [75][76][77] . Through this study it can be reasoned that C. autoethanogenum's Agr QS system plays the critical role of sensing population density in nature, allowing for prompt upregulation of enzymes required for C1 gas metabolism, well before energy-rich sugar sources are exhausted. It appears that many acetogens employ the WLP as an electron sink for redox balancing when growing with other organic compounds 42 and hence it was interesting to see that in the complete absence of Agr QS, reducing equivalents generated during fructose oxidation appear to be channelled into ethanol formation. It can be reasoned that two differentially expressed 34 AgrD peptides enable C. autoethanogenum to accurately monitor its environment and mediate the use of reducing equivalents for CO 2 reduction through appropriate expression of the necessary WLP enzymes exactly when needed.

Materials and methods
Bacterial strains and growth conditions. Laboratory stocks of C. autoethanogenum DSM 10061 used for this study were purchased from Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ) GmbH, Braunschweig, Germany, (see Supplementary information, Table S1 for full list of strains). Escherichia coli One Shot® TOP10 used in plasmid construction, cloning and conjugation, was grown in LB medium or 1.5% solid agar LB plates. Agar LB plates were incubated at 37 °C, and liquid cultures shaken at 225 RPM at 37 °C with antibiotics where appropriate. Antibiotics and working concentrations used in this study were: chloramphenicol (25 μg/mL), thiamphenicol (7.5 μg/mL), spectinomycin (100 μg/mL) and d-cycloserine (250 μg/mL).
Clostridium autoethanogenum genetic engineering was performed in an anaerobic workstation (Don Whitley Scientific, Yorkshire, UK) at 37 °C. For general cultivation methods, YTF medium was used as described previously 68 . To screen for pyrE-deficient and uracil auxotroph mutants, semi-defined minimal medium (SDMM) that was similar to YTF but used per litre: 7.2 g d-fructose, 20 g 2-(N-morpholino)ethanesulfonic acid (MES) buffer and 10 g casein amino acids in place of yeast extract and tryptone.
For growth and fermentation analysis, experiments took place in 250 mL serum flasks containing American Type Culture Collection (ATCC) PETC 1754 medium, with the following additions and modifications: 20 g/L MES and 1.5 g/L casein amino acids in place of yeast extract. Serum flasks filled with medium inside of an anaerobic chamber (COY laboratory products, USA) with a gas atmosphere of 85% N 2 , 10% CO 2 and 5% H 2 . After filling, flasks were sealed with rubber stoppers, crimped, and autoclaved. After autoclaving, 1 mL of 4% (w/v) L-cysteine was added to each serum flask. For sugar-based experiments, 90 mL of medium was dispensed into each flask with D-fructose added to a concentration of 50 mM post autoclave and placed into a New Brunswick Innova 44 shaking incubator (Eppendorf) set to 225 rpm, at 37 °C. Flasks were sealed with butyl rubber stoppers preventing any gas exchange. Analysis of C. autoethanogenum ΔagrD mutants using D-fructose used three to four technical replicates, that included a WT control, and up to six biological replicates. Starting OD 660 was approx. 0.01 and flasks were usually cultured for up to 10 days with 10-12 sampling points. For gas-grown cultures, serum flasks were filled with 50 mL PETC, pressurised with 200 kPa CO and stored horizontally at 37 °C in a static incubator. Headspace pressure was determined via a needle attached to a handheld pressure gauge (Omega Engineering, USA). Experiments began with a starting OD 660 of 0.01 with 2 biological replicates and were cultured for up to 30 days. Samples were taken approximately every 3-5 days. For all experiments growth was measured by measuring optical density at 660 nM (OD 660 ) using a Jenway 7300 spectrophotometer (Bibby Scientific, UK).

Detection of spores.
To determine sporulation activity of WT C. autoethanogenum, stocks were grown in 3 mL YTF in triplicate for up to 5 days. Two mL from each culture was spun down, washed, and resuspended in 400 μL of anaerobic PBS. This mixture was split into equal aliquots, one acting as a control, the other was heat Scientific Reports | (2022) 12:411 | https://doi.org/10.1038/s41598-021-03999-x www.nature.com/scientificreports/ shocked at 80 °C for 10 min using a SensoQuest PCR machine (SensoQuest GmbH, Germany). The experimental and control triplicates were serially diluted ranging from 10 0 to 10 -2 and plated onto YTF plates, colonies appeared after 4-5 days. C. acetobuytlicum was used as a control and was grown on clostridial basal medium (CBM). C. acetobuytlicum was subjected to identical heat-shock procedures as C. autoethanogenum and plated on either CBM or YTF agar plates respectively. Samples were observed using under a phase-contrast filter of a Nikon Eclipse Ci-E microscope (Nikon Instruments Europe B.V., UK). Visualising the presence of spores was done using Schaeffer and Fulton Spore Staining Kit (Sigma Aldrich, UK) as per manufacturer's instructions.
DNA and plasmid manipulations. Routine genomic DNA extraction used in PCR diagnostics used PureLink Genomic DNA extraction (Invitrogen/ThermoFischer). Plasmid purification from E. coli was performed using GenElute Plasmid Miniprep kits (Sigma Aldrich, UK). All purified DNA fragments and plasmids were quantified using NanoDrop Lite (Thermo Scientific, UK). High-quality, proof-read amplification of DNA fragments, for short-length sequencing and cloning used Phusion High-Fidelity DNA Polymerase (NEB, UK). Routine screening and analytical PCR procedures used DreamTaq Green DNA polymerase (Thermo Scientific, UK). Primers were synthesized by Sigma-Aldrich or MWG Eurofins. Sanger sequencing of plasmids and amplicons was carried out by Source Bioscience PLC (UK). All primers used in this study are listed in Supplementary information, Table S2.
Creation of Allelic coupled-exchange plasmids. All plasmids were based on the pMTL80000-series Modular Plasmid System 78 , specifically pMTL84151. A full list of created plasmids is shown in Table S3. For all targeted genes, inactivation was achieved using an asymmetrical homology arm design as outlined by Minton et al., allowing single and double crossover events to occur in a preferred order 79 . Plasmid pMTL84151_ΔpyrE was constructed to generate a 3' truncated version of the C. autoethanogenum pyrE gene thereby creating the pyrE-negative parent strain used for all further mutations. Its design followed the principles outlined previously 79 ; it carried a non-homologous C. acetobutylicum pyrE gene serving as a counter-selection marker for single crossover event screening, and an allelic-exchange cassette consisting of the left homology arm (LHA) (303 bp), the right homology arm (RHA) (1219 bp) of C. autoethanogenum's pyrE and a lacZα fragment (264 bp) in between.
Primer pairs used to amplify the LHA and RHA were pyrE_FSP, see Table S2 for list of primer sequences. Plasmids pMTL84151_0816 and pMTL84151_3094 were designed to inactivate ΔagrD1 or ΔagrD2, respectively and both were constructed using the original pMTL84151 backbone 60 . LHA and RHA amplicons were generated by PCR, using two sets of primer pairs for each homology arm as follows: For CLAU_0816, primer set; 0816_LHA_F + R and 0816_RHA_F + R was used. Similarly, primer pair sets for and pairs for agrD2; 3094_ LHA_F + R and 3094_RHA_F + R. Each set was combined using splice-overlapping extension PCR (SOE-PCR) to produce inserts of 1,270 bp and 1,437 bp inserts, respectively. Both inserts were subsequently ligated into the pMTL84151 backbone producing plasmids pMTL84151_0816 and pMTL84151_3094. Each in-frame deletion (IDF) region was designed to remove approx. 80% of the gene.
Plasmid pMTLCH20 60 was used to repair the pyrE defect in the generated agrD mutants, again as outlined by Minton et al., the insert of this plasmid encoded an intact pyrE gene, generated by combining LHA and RHA of 524 bp and 1212 bp, respectively 79 . Two derivatives of this plasmid, complementation vectors pMTLCH20-D1 and pMTLCH20-D2, were used to combine repair of the pyrE gene with downstream integration of intact agrD1 and agrD2 genes, respectively. Primer pairs 0816_Comp_F + R for agrD1, and 3094_Comp_F + R for agrD2 were used to amplify each agrD region respectively and the obtained PCR products were similarly digest and ligated into pMTLCH20.
Plasmid conjugation. Plasmids were introduced into C. autoethanogenum through previously established methods 80 , using E. coli CA434 as a conjugation donor strain. Selected plasmids were first electroporated into the donor strain, performed in a 0.2 cm wide electroporation cuvette (Biorad, UK) at 2.50 kV. After conjugation, YTF agar plates containing D-cycloserine and thiamphenicol were used to counter-select against E. coli CA434. Colonies of conjugated C. autoethanogenum typically appeared after 4-5 days. Larger, fast-growing colonies, assumed to be successful transformants, were picked and re-streaked onto fresh selective YTF containing the same antibiotics. Plasmids were introduced into C. autoethanogenum through previously established methods 80 , using E. coli CA434 as a conjugation donor strain. Selected plasmids were first electroporated into the donor strain, performed in a 0.2 cm wide electroporation cuvette (Biorad, UK). Electroporation was performed using a MicroPulser electroporator (Bio-Rad, UK) at 2.50 kV for 5 mS. After conjugation, YTF agar plates containing D-cycloserine (250 μg/mL) and thiamphenicol (25 μg/mL) were used to counter-select against E. coli CA434 and presence of the plasmid in C. autoethanogenum recipient cells, respectively. Colonies of plasmid-carrying C. autoethanogenum typically appeared after 4-5 days. Larger, fast-growing colonies, assumed to be successful transformants, were picked and re-streaked onto fresh selective YTF containing the same antibiotics.
Creation of a C. autoethanogenum_ΔpyrE mutant. First, a pyrE deficient strain of C. autoethanogenum was created which would serve as a host for later agrD IFD mutations with positive and negative selection markers 57 . To do this, C. autoethanogenum WT was conjugated with E. coli CA434 pMTL84151_ΔpyrE. Following selection on thiamphenicol, the initial, large C. autoethanogenum colonies resulting from this conjugation were picked and placed into liquid YTF containing thiamphenicol and grown for 3 days. Three PCRs were used to screen for pMTL84151_ΔpyrE singe-cross over integrants: first, using flanking primer pyrE_RSP with internal primer 84151_F, second, using flanking primer pyrE_FSP with internal primer 84151_R. The third reaction used both flanking pyrE primers, covering the entire integration region to reveal no bands as the amplicon was too large. Possible amplicons in the third reaction suggested plasmid integration was contaminated with the www.nature.com/scientificreports/ presence of WT pyrE and required repetition to obtain a pure single-cross integrant. Confirmed single-cross integrants were plated onto YTF plates containing 1 mg/mL of 5-fluoroorotic acid (FOA) and 20 μg/mL uracil to select for the double crossover event. FOA-resistant colonies appearing after 3 days were picked and replica plated onto two sets of SDMM agar plates with and without 20 μg/mL uracil. Colonies that grew exclusively on uracil supplemented SDMM plates were re-streaked again onto SDMM + /-uracil plates. Confirmed colonies were cultured in YTF for gDNA extraction and storage at -80 °C. Extracted gDNA was to amplify the now truncated pyrE region using flanking primers pyrE_FSP and pyrE_RSP. Amplicons were purified and Sanger sequenced for confirmation. Plasmid loss was determined through streaking newly created ΔpyrE strains onto YTF plates containing thiamphenicol. Lack of growth indicated pMTL84151_ΔpyrE plasmid loss.
Creation of single and double knockout agrD mutants. IFD plasmids pMTL84151_0816 and pMTL84151_3094 were conjugated separately into C. autoethanogenum_ΔpyrE to produce the respective single knock-out mutants. After the second round of selection, carried out as described above for ∆pyrE, large colonies were verified for single cross integration using. Right and left homology arm flanking primers for each target gene (0816_SCR_F + R and 3094_SCR_F + R) were paired with pMTL84151 internal primers 84151_F and 84151_R in three separate PCRs. Single-cross integrants were isolated and plated onto FOA YTF plates and grown for 3 days to promote double crossovers. Visible colonies were replica plated onto SDMM +/− uracil plates and grown for a further 4-5 days. To screen for successful target gene truncation, colonies that grew exclusively on uracil plates were screened using respective 0816 and 3094 flanking primers. Sanger sequencing confirmed in-frame truncation and mutants ΔagrD1_ΔpyrE and ΔagrD2_ΔpyrE were created. For the creation of the ΔagrD1D2 double mutants, strains ΔagrD1_ΔpyrE or ΔagrD2_ΔpyrE were used and were conjugated with either pMTL84151_3094 or pMTL84151_0816, respectively. This created three independent mutants: ΔagrD1D2_ΔpyrE, ΔagrD1D2_1 ΔpyrE and ΔagrD1D2_2 ΔpyrE.

Restoration and complementation techniques. Restoration of the pyrE locus in ΔagrD1_ΔpyrE,
ΔagrD2_ΔpyrE and ΔagrD1D2_2 ΔpyrE mutants used plasmid pMTLCH20 and followed allelic-couple exchange methods outlined earlier 61 . This plasmid was first conjugated into ΔpyrE mutant strains as described above. Fast-growing colonies were streaked directly onto SDMM +/− uracil plates and grown for 4 days. Colonies growing without uracil supplementation were selected and primers, pyrE_FSP and pyrE_RSP were used for PCR screening and Sanger sequencing to confirm pyrE restoration. A 264 bp LacZα marker was incorporated into pyrE restored mutants.
Complementation of the ΔagrD1D2 double mutant followed the same procedure but utilised pMTLCH20 derivatives pMTLCH20-D1 and pMTLCH20-D2, respectively. Conjugation took place as described previously and successful transformants were replica plated onto SDMM +/− uracil plates and grown for 4 days. Colonies growing without uracil supplementation were confirmed for pyrE locus restoration and insertion of respective agrD genes. PCR screening using pyrE_FSP and pyrE_RSP flanking primer and Sanger sequencing of the generated amplicons confirmed creation of the complemented mutants, ΔagrD1D2 + D1comp and ΔagrD1D2 + D2comp. Genome sequencing. To confirm intended mutations, genome sequencing was performed on all single and double knock-out mutants that were pyrE corrected, using Illumina next generation sequencing (Microbes NG, Birmingham). Raw genomic data was processed and assembled using CLC Genomics Workbench (v. 8.0.2), (Qiagen, Denmark) and were mapped against the Humphreys et. al. C. autoethanogenum DSM 10,061 genome 36 .
Gas chromatography (GC). Supernatant samples of 500 μL were pipetted into 1.5 mL tubes and acidified with 5 μL of 10 M H 2 SO 4 . Samples were vortexed at 21,000×g for 1 min, and 500 μL of propyl propionate containing 50 mM valeric acid was added as an internal standard and samples were vortexed once more. The extraction mixture was spun down again to remove denatured proteins. Samples then separated into a biphasic solution, from which 300 μL of the organic upper phase was removed and pipetted into a 300 μL glass insert that was placed inside a 2 mL brown sampling vial. Analysis of metabolites from prepared samples used a Focus gas chromatograph (Thermo Fischer Scientific) with a 30 m TRACE TR-FFAP column (Thermo Fischer Scientific) with 0.25 mm internal diameter. A flame ionisation detector was maintained with 350 mL/min compressed air, 35 mL/min H 2 and 30 mL/min N 2 . The H 2 carrier gas was supplied at 0.8 mL/min. Injector inlet temperature was maintained at 240 °C and detector temperature was set to 270 °C. For resolution of peaks the following column program was used: after injection, temperature was held at 40 °C for 2 min, then ramped to 150 °C at a rate of 80 °C/min, and finally to 210 °C for 1 min.
Gas headspace analysis. Serum flask headspace composition was measured with a TRACE gas chromatograph (Thermo Fischer Scientific). 1 mL of gas was removed from serum flasks with gas-tight syringes and manually injected. A 30 m TRACE TR-FFAP column (Thermo Fischer Scientific) with a 0.25 mm internal diameter was used with the following program for peak resolution: After injection, temperature was held at 40 °C for 3 min, then ramped to 250 °C at a rate of 30 °C/min, the duration of each run took 10.7 min. Standards for gas analysis, comprised of 1% CO, 1% CO 2 , 1% H 2 , 97% N 2 .
High performance liquid chromatography. Centrifuged  www.nature.com/scientificreports/ Dionex HPX-Ulti-Mate 3000 HPLC system (Thermo Fischer Scientific, UK). 20 μL volumes were injected into a Bi-Rad Aminex 87H (Biorad UK) column, maintained at 35 °C, with a 0.005 M H 2 SO 4 mobile phase at a flow rate of 0.5 mL/min. Peaks were detected through a refractive index and diode array detector at UV 210 nm.

Preparation of samples for proteomic analysis.
Cultures of WT C. autoethanogenum and C. autoetha-nogenum_ΔagrD1D2 were grown in triplicate in 250 mL serum flasks containing PETC medium with 50 mM D-fructose. Cultures were harvested at approx. OD 660 1.0, after approx. 3-4 days and centrifuged at 3000×g for 5 min and supernatant was removed. The cell pellets were stored on ice and washed twice with PBS and once with protein extraction buffer, triethyl ammonium bicarbonate (TEAB), 0.5 M pH 8.5 (Sigma Aldrich). Samples were then flash frozen in liquid nitrogen and sent to the University of Sheffield to be analysed by Dr Mahendra Raut using LC-MS/MS-based iTRAQ methods (4-plex reagents isobaric and tags for relative and absolute quantitation). See Supplementary information, Sect. 2 for full protocol.
Crude lysate preparation for enzymatic assays. Clostridium autoethanogenum WT and ΔagrD1D2 cells were grown in duplicate in 750 mL serum flasks containing 400 mL PETC with 50 mM D-fructose. Cells were harvested at early to mid-exponential phase at OD 660 0.4-0.45 in an anaerobic chamber (COY laboratory products, USA). Cultures were centrifuged at 3,000 × g for 5 min and following removal of the pellets of each strain were flash frozen in liquid N 2 and stored at -80 °C. Cell pellets were resuspended thoroughly in a combination of 750 μL of BugBuster protein extraction reagent (Merck Millipore, United States) with 750 μL 50 mM TRIS-HCl and 0.5 mM MgCl 2 , pH 7.0, this mixture was prepared anaerobically. 10 mg/mL of lysozyme, and 1 mg/mL of DNase (Sigma Aldrich) was added to the mixtures and gently aspirated. The lysis mixture was placed into a pre-evacuated 15 mL Hungate tube and sealed. The tube was placed onto a rocker at room temperature at 50 RPM for one hour. Tubes were placed in the anaerobic chamber and each tube's mixture was pipetted into a 2 mL cryovial (Sarstedt, Germany) and spun at 20,000×g at 4 °C for 20 min. The clear supernatant was injected into a 5 mL serum flask vacuum filled with pure N 2 and stored on ice ready for analysis. For full specific activity measurement methods followed published procedures 35,81 , for full methodology see Supplementary information, Sect. 2.

Data analysis and presentations. Analytics data was collected using Chromeleon 7.2 Chromatography
Data System (Thermo Fisher Scientific, UK). Data was compiled using Microsoft Excel 2010 which compared samples using used two-tailed, unpaired, parametric student's T-test. Graphical representation used GraphPad Prism 7.