Metabolic engineering of Clostridium autoethanogenum for selective alcohol production

Gas fermentation using acetogenic bacteria such as Clostridium autoethanogenum offers an attractive route for production of fuel ethanol from industrial waste gases. Acetate reduction to acetaldehyde and further to ethanol via an aldehyde: ferredoxin oxidoreductase (AOR) and alcohol dehydrogenase has been postulated alongside the classic pathway of ethanol formation via a bi-functional aldehyde/alcohol dehydrogenase (AdhE). Here we demonstrate that AOR is critical to ethanol formation in acetogens and inactivation of AdhE led to consistently enhanced autotrophic ethanol production (up to 180%). Using ClosTron and allelic exchange mutagenesis, which was demonstrated for the first time in an acetogen, we generated single mutants as well as double mutants for both aor and adhE isoforms to confirm the role of each gene. The aor1+2 double knockout strain lost the ability to convert exogenous acetate, propionate and butyrate into the corresponding alcohols, further highlighting the role of these enzymes in catalyzing the thermodynamically unfavourable reduction of carboxylic acids into alcohols.


Introduction
The deleterious environmental impact caused by the continuing extraction and exploitation of fossil fuels for energy, coupled with their inherent finite nature, are the principle drivers for the development of sustainable alternatives. One option is to develop a biological route. However, the economic conversion of non-food, cellulosic feedstocks into liquid transportation fuels through biological fermentation is proving challenging. Alternative conversion processes are required. In this regard, gas fermentation has emerged as a promising technology that converts industrial waste gases or syngas containing CO, CO 2 and H 2 into fuels without impacting on food production. It is reliant on bacterial process organisms that are able to utilise single carbon gases as a source of carbon typified by a group of strictly anaerobic bacteria known as acetogens. One such acetogen is Clostridium autoethanogenum (Abrini et al., 1994). During autotrophic growth, C. autoethanogenum employs the Wood-Ljungdahl pathway to fix CO 2 , together with H 2 as reductant, into predominantly acetic acid and ethanol. It is also able to grow on CO as a sole source of carbon and energy and synthesize ethanol, 2,3-butanediol and lactate (Köpke et al., 2011).
Insight into the metabolic capabilities of C. autoethanogenum has been gleaned from the determination of its genome sequence (Brown et al., 2014;Humphreys et al., 2015;Utturkar et al., 2015) and 'omics' data (Köpke et al., 2011;Marcellin et al., 2016;Mock et al., 2015) under different growth conditions (fructose, CO and H 2 +CO 2 ). Moreover, some understanding of how the acetogen conserves energy while generating reduced products ethanol and 2,3-butanediol autotrophically was derived through the determination of the specific activities and cofactor specificities of all relevant oxidoreductases from cell extracts (Mock et al., 2015;Wang et al., 2013). As yet, however, the biosynthetic capabilities of C. autoethanogenum, and in particular ethanol synthesis, have not been thoroughly investigated at the genetic and molecular levels. Similar to other prominent autotrophic ethanol producers (e.g., Clostridium ljungdahlii, "Clostridium ragsdalei", and Clostridium carboxidivorans), the ethanol biosynthesis pathway of C. autoethanogenum comprises two main routes ( Fig. 1): (i) the direct, two-step sequential reduction of acetyl-CoA into ethanol via acetaldehyde using bi-functional aldehyde/alcohol dehydrogenase (AdhE), CoA-dependent acetaldehyde dehydrogenase (Ald) and alcohol dehydrogenase (Adh) as found in other ethanol producing bacteria including E. coli (Membrillo-Hernandez and Lin, 1999) and; (ii) a postulated indirect route that proceeds via acetate and employs aldehyde: ferredoxin oxidoreductase (AOR) to first reduce acetic acid to acetaldehyde before ethanol synthesis via Adh (Köpke et al., 2010;Mock et al., 2015). The genome of C. autoethanogenum contains genes encoding two AOR isoforms: aor1 (CLAU_0081) and aor2 (CLAU_0099); and two AdhE enzymes: adhE1 (CLAU_3655) and adhE2 (CLAU_3656) that appear in tandem and are potentially a result of gene duplication (Humphreys et al., 2015). The same arrangement is also found in C. ljungdahlii (Köpke et al., 2010;Leang et al., 2013).
One key distinction between the two ethanol biosynthesis routes is that the indirect route reduces acetate, which is an unwanted byproduct as it limits yield and is known to be toxic at elevated concentrations. All naturally isolated acetogens form acetate as it provides an advantage through conservation of one ATP per mole of acetate via substrate level phosphorylation (SLP), which is significant under the ATP-limiting conditions of autotrophic growth. Thermodynamic and stoichiometric analyses estimated that during autotrophic growth of C. autoethanogenum on H 2 + CO 2 , the ATP yield is only 0.5 ATP/mol ethanol via acetyl-CoA reduction to acetaldehyde, in comparison to the 1.2 ATP/mol ethanol via acetic acid reduction to ethanol (Mock et al., 2015). Similarly, during growth on CO, ATP yield would be greater via the AOR route (1.875 ATP/mol ethanol) compared to the AdhE route (1.375 ATP/mol ethanol) when applying the same scheme.
In this study, the adhE and aor genes of C. autoethanogenum were inactivated to determine their roles in autotrophic ethanol production. In addition to the intron-based gene inactivation demonstrated previously (Marcellin et al., 2016;Mock et al., 2015), here we expanded the genetic tools for this acetogen to include allelic exchange. Characterization of these strains revealed the roles of AOR in support-ing alcohol production, and demonstrated that strains producing greater amounts of ethanol (up to 180% improvement) could be obtained by genetically inactivating adhE or aor2.

Bacterial strains and growth conditions
The bacterial strains used are listed in Table S1. Escherichia coli strains employed for general plasmid propagation, cloning and conjugation were cultivated at 37°C in LB medium in the presence of antibiotic (25 μg/mL chloramphenicol, 100 μg/mL spectinomycin). C. autoethanogenum DSM 10061 was purchased from Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ) GmbH, Braunschweig, Germany and routinely cultivated under strict anaerobic conditions in CaGM medium (Liew et al., 2016a).

DNA manipulations
Genomic DNA from C. autoethanogenum was isolated using DNeasy Blood and Tissue kit (Qiagen) for PCR diagnostics. For Southern Blot analysis, genomic DNA of C. autoethanogenum was extracted according to Bertram and Dürre (1989). Plasmid DNA from C. autoethanogenum was isolated using QIAprep Spin Miniprep kit (Qiagen) with the supplementation of 20 mg/mL chicken lysozyme into lysis buffer and incubation at 37°C for 30 min before proceeding to downstream procedures. Polymerase Chain Reaction (PCR) was carried out using Phusion DNA polymerase (NEB) or Q5 DNA polymerase (NEB). All primers used in this study are listed in Table S2. Primers were designed using Geneious 6.1.7 (Biomatters) and synthesized by Sigma-Aldrich or Eurofins. Sanger sequencing of plasmids and amplicons was carried out by Source Bioscience Plc.
An allelic exchange plasmid, 'pMTL-AMH101' (Supplementary nucleotide sequence 1), was used for deletion of 227 bp of the Cterminus of C. autoethanogenum pyrE (CLAU_1436). It contains a heterologous pyrE (cac_0027) from C. acetobutylicum ATCC 824 (to be employed as a counter selectable marker) and comprises a 303 bp short left homology arm (LHA) and a larger 1219 bp right homology arm (RHA), with lacZα in between, as the allelic-exchange cassette. The in-frame deletion (IFD) allelic-exchange cassettes of C. autoethanogenum adhE1, adhE1+2, and aor2 consists of two homology arms of similar lengths (518 -580 bp), and were assembled using SOE-PCR and oligonucleotides listed in Table S2. All the IFD cassettes retained only the start and stop codons of the target loci without affecting the 5`-untranslated region (UTR) and 3`-UTR. Following SOE-PCR, the IFD cassettes were digested with SacII and AscI and cloned into plasmid pMTL-AMH101 to generate plasmids 'pMTL84151-ΔadhE1', 'pMTL84151-ΔadhE1+2', and 'pMTL84151-Δaor2'. For the restoration of pyrE, a plasmid called 'pMTL-AMH102' (Supplementary nucleotide sequence 2), which consists of a pyrE repair allelic exchange cassette with a 526 bp LHA and 1213 bp RHA, was employed.

Plasmid transfer into C. autoethanogenum
Plasmids were transformed into E. coli donor strain CA434 (HB101 containing the conjugative plasmid R702) and then transferred into C. autoethanogenum via conjugation using previously established methods (Mock et al., 2015;Purdy et al., 2002;Williams et al., 1990). Thiamphenicol (7.5 µg/mL) was used to select for catP-based plasmids. Trimethoprim (10 µg/mL) was used to counter select against E. coli CA434 after conjugation. For the validation of plasmid complementation strains, plasmids were isolated from C. autoethanogenum transconjugants and subsequently transformed into E. coli cells, before restriction digest analysis was carried out on the 'rescued' plasmids. The 16 s rRNA gene was also amplified from the genomic DNA of transconjugants using oligonucleotides 'univ-0027-F' and 'univ-1492-R', followed by Sanger sequenced for verification purposes.

Construction of C. autoethanogenum ClosTron strains
Following conjugation of ClosTron retargeting plasmids into C. autoethanogenum, thiamphenicol and trimethoprim resistant colonies were transferred onto solid YTF medium supplemented with 6 µg/mL clarithromycin to select for Intron insertion in target loci. These were then repeatedly streaked onto the same selective medium until plasmid loss was evident -loss of ability to grow on medium supplemented with thiamphenicol. Genomic DNA was extracted from the clarithromycin resistant colonies and subjected to PCR screen using locus-specific flanking primers (Table S2) to identify clones (Fig. S1A & S2A) that produced an amplicon that is 1.8 kb larger than WT control (indicative of ClosTron insertion at specified DNA locus). Sanger sequencing of the ClosTron amplicons was performed to validate the location of ClosTron insertion. As final verification, Southern Blot analysis was performed using a digoxigenin (DIG) High-Prime DNA labelling and detection kit (Roche) as instructed by the manufacturer to ensure that only one ClosTron insertion had occurred in each mutant ( Fig. S1B & S2B).
2.6. Allelic-exchange procedure 2.6.1. Creation of ΔpyrE strain The procedure adopted was as previously described (Heap et al., 2012). For the construction of ΔpyrE strain, which serves as a host for further IFD of adhE1, adhE1+2, and aor2 using pyrE as a positive and negative selection marker, the plasmid pMTL-AMH101 was transferred into C. autoethanogenum via conjugation. The transconjugants were restreaked on YTF solid medium supplemented with thiamphenicol and trimethoprim to enrich and identify fast-growing singlecrossover integrant clones. Genomic DNA was isolated and subjected to PCR analysis using two different primers (ACE-plasmid-F and ACEplasmid-R) that anneal to plasmid specific sequences together with the appropriate locus-specific flanking primers (Table S2). The presence of a DNA fragment indicated that the clones were indeed single-crossover integrants, while the size was indicative of at which homology arm the recombination event had occurred. PCR verified single-crossover integrants were inoculated into CaGM liquid medium supplemented with 10 g/L fructose and thiamphenicol and allowed to grow for 2 days inside anaerobic workstation, before they were serially diluted and plated. To facilitate the screening of rare second recombination events, the CaGM solid medium had 1 g/L yeast extract replaced with 1 g/L casein acid hydrolysate, and supplemented with 1.5 mg/mL fluoroorotic acid (FOA+) and 5 μg/mL uracil. The efficacy of FOA is reduced by the presence of uracil, and yeast extract contains substantial amounts of uracil so a less enriched supplement such as casein acid hydrolysate was used instead. Incubation at 37°C was carried out inside anaerobic workstation and FOA-resistant colonies that emerged within 2-3 days were restreaked onto the same selective medium before PCR screen using locus-specific flanking primers was performed to distinguish double-crossover recombinant clones from wild-type revertant clones.
Sanger sequencing was employed to confirm the expected genotypes.

Creation of ΔadhE1, ΔadhE1 mut , and ΔadhE1+2 strains
After the loss of plasmid was demonstrated by the loss of thiamphenicol resistance, the ΔpyrE strain could serve as a host for pMTL84151-ΔadhE1, and pMTL84151-ΔadhE1+2 via conjugation for the construction of ΔadhE1 and ΔadhE1+2 strains, respectively. Single-crossover integrants and double-crossover FOA-resistant, uracil auxotrophic clones were obtained for both targets (same method as the ΔpyrE strain above). In the first attempt, Sanger sequencing revealed that in addition to the IFD of adhE1, an unintended 84 bp deletion had occurred in the promoter region of adhE2. Termed 'ΔadhE1 mut ', this strain also shown to have lost the excised plasmid, as demonstrated by loss of thiamphenicol resistance. A second attempt at generating a 'clean' 'ΔadhE1' strain without the unintended 84 bp deletion was successful but repeated attempts to isolate a clone in which the excised plasmid had been lost (shown by persistent thiamphenicol resistance) were unsuccessful. For the ΔadhE1+2 strain, Sanger sequencing revealed successful deletion of adhE1 and adhE2 without complications in the 5`-UTR of adhE1 and 3`-UTR of adhE2. However, repeated restreaking was unable to isolate thiamphenicol sensitive colonies for this strain.

Creation of aor1+2 double KO strain
For the construction of the aor1+2 double knock-out strain (herein termed 'aor1+2 KO'), the aor1 locus was first inactivated using ClosTron plasmid pMTL007C-E2::aor1_361 s in a ΔpyrE strain. Following the loss of plasmid, the IFD plasmid pMTL84151-Δaor2 was transformed and the isolation of single-crossover integrant and double-crossover recombinant clones were carried out as described above. These aor1 and aor2 double KO but uracil auxotrophic clones were transformed with plasmid pMTL-AMH102 to restore uracil prototrophy. Fast-growing thiamphenicol-resistant colonies were plated onto CaGM solid medium supplemented with 10 g/L fructose but had 1 g/L yeast extract replaced with 1 g/L casein acid hydrolysate without uracil supplementation. As final validation, PCR screening followed by Sanger sequencing was carried out using flanking primers to verify ClosTron insertion event in aor1, IFD of aor2 and restoration of pyrE. Plasmid loss was confirmed by demonstration of thiamphenicol sensitivity.

Harvest of cells for gene expression analysis
C. autoethanogenum recombinant strains were cultivated in triplicates of 500 mL pressure plus laboratory bottles (Duran), each containing 200 mL CaGM supplemented with 10g/L fructose. For strains ΔpyrE and ΔadhE1 mut , 10 µg/mL uracil was supplemented. In order to maintain plasmids in C. autoethanogenum harbouring plasmid pMTL83151-P acsA and pMTL83151-P acsA -aor1, 7.5 µg/mL of thiamphenicol was supplemented. Approximately 12 OD 600 equivalent of cells were harvested at various growth phases by centrifugation at 4°C at 3220×g for 10 min. Supernatant was removed and the cell pellet was resuspended in 1 mL RNAlater Stabilization Solution (Ambion) by pipetting. After overnight incubation at 4°C, the cell suspension was centrifuged at 3220×g at 4°C for 10 min and supernatant discarded before storage at −80°C until RNA extraction.

Total RNA extraction and cDNA synthesis
Following the addition of 1.5 mL cold TRIzol (Ambion), the thawed cell pellet was transferred into pre-chilled 2 mL microfuge tubes containing 1g of dnature 0.1 mm diameter Zirconia/Silica beads (dnature Ltd). Cell disruption was performed in 3 cycles of 1 min bead beating using Mini Beadbeater-16 (dnature Ltd), with 1 min chilling on ice in between the cycles. Following 1 min of 4°C centrifugation at 20,238×g, the supernatant was harvested and 100 μL of chloroform was added, vortexed for 20 s and then incubated at room temperature for 15 min with occasional mixing. After the centrifugation at 20,238×g (4°C) for 15 min, the aqueous phase was collected and 0.7 vol of isopropanol was added. The samples were incubated at room temperature for 10 min before centrifugation at 20,238×g (4°C) for 10 min. Supernatant was removed and the DNA pellet was washed with 700 μL of ice-cold 70% (v/v) ethanol before another round of centrifugation 20,238×g (4°C) for 10 min. Following the removal of supernatant, the RNA pellet was air-dried for 15 min before resuspension in 100 μL of RNase-free water and 1 μL of RNaseOUT (Invitrogen).
Genomic DNA was removed by the addition of TURBO DNase enzyme (Ambion) and 37°C incubation for 30 min. The DNase-treated RNA was purified using RNA Clean and Concentrator Kit (Zymo Research) as per manufacturer's instructions and stored at −80°C. The concentration and purity of isolated RNA was analyzed spectrophotometrically using Nanodrop (Thermo Scientific). To ensure the absence of residual genomic DNA in the isolated RNA, 1 μL of each RNA samples was subjected to PCR analysis using primer pairs "adhE2-662s-F" and "adhE2-662s-R". The quality of RNA was examined using 2100 Bioanalyzer (Agilent Technologies) and RNA samples with RNA integrity number (RIN) greater than 7 were used for cDNA synthesis. Two μg of total RNA was used per 20 μL SuperScript III Reverse Transcriptase reactions (Invitrogen) and diluted 10-fold with RNase-free water prior to qPCR analysis.

Quantitative reverse transcriptase polymerase chain reaction (qRT-PCR)
Primers and probe sets for target gene (adhE2) and housekeeping genes (gyrA and rho) (Table S4) were designed using the Custom TaqMan Assay Design Tool and purchased as Single-Tube Custom TaqMan Gene Expression Assays from Applied Biosystems. gyrA (CLAU_2078; encodes DNA gyrase subunit A) and rho (CLAU_2269; encodes transcriptional termination factor) were chosen as housekeeping genes because they exhibited the most stable gene expression levels for different carbon sources and stresses in closely related acetogen C. ljungdahlii DSM 13528 . The amplification efficiencies of the TaqMan probes and primers were empirically determined to be between 94.2% and 99.7% (R 2 ≥0.998) by constructing a standard curve using serially diluted cDNA as template (data not shown).
All qRT-PCR reactions were set up in 96-well Microseal PCR plates (Bio-Rad Laboratories) and performed in triplicates of 20 μL volume containing 1 μL diluted cDNA, 1 μL of 20x Custom TaqMan Gene Expression Assay, 10 μL of 2x TaqMan Gene Expression Master Mix (Applied Biosystems) and 8 μL nuclease-free water. Non-template controls (NTC) were included for each TaqMan probe and primer qRT-PCR master mixes. Each qRT-PCR runs comprised an initial denaturation and polymerase activation at 95°C for 12 min, followed by 40 cycles of denaturation at 95°C for 15 s and combined annealing and extension at 60°C for 60 s. The CFX connect Real-Time PCR Detection System (Bio-Rad Laboratories) was employed to record the accumulation of signals in each well within the PCR plate, and the accompanying CFX Manager Software was used to perform normalized gene expression analysis.

Analytical chemistry
Analysis of metabolites were performed using Varian ProStar HPLC system equipped with a RID (Refractive Index Detector) operated at 30°C and an Aminex HPX-87H column (1300×7.8 mm, particle size 9 µm) (Bio-Rad Laboratories) kept at 30°C. Slightly acidified water was used (0.005 M H 2 SO 4 ) as mobile phase with a flow rate of 0.5 mL/ min. To remove proteins and other cell residues, samples were centrifuged at 20,238×g for 5 min and the supernatant was filtered with Spartan 13/0.2 RC filters. 10 μL of the supernatant was then injected into the HPLC for analyses.

Data analysis and presentations
Statistical analysis and graphically presented results were obtained using GraphPad Prism. Two-tailed, unpaired, parametric student's ttests were employed for comparison of means.

Results and discussions
3.1. Exemplification of allelic-exchange mutagenesis in C. autoethanogenum Whilst single intron insertion mutants of adhE1, adhE2, aor1 and aor2 were readily generated (Fig. S1 & S2), re-use of the ClosTron to generate double mutants requires marker re-cycling (Heap et al., 2010). This proved not to be possible in C. autoethanogenum (data not shown). In this study, therefore, we developed an allelic exchange method for C. autoethanogenum based on the use of a pseudo-suicide vector reliant on the pCD6 replicon (Heap et al., 2009) and a plasmidencoded counter selection marker composed of an orotate phosphoribosyltransferase (pyrE) gene of C. acetobutylicum. The approach taken, and the principles involved, have recently been reviewed (Minton et al., 2016).
In order for pyrE to be used as a counter selection marker, a C. autoethanogenum mutant lacking the 3′-end (227 bp) of the native pyrE gene (CLAU_1436) was made using an ACE (Allele-Coupled Exchange) vector equivalent to pMTL-YN18 (Ng et al., 2013), as described in Section 2.3. To create a double mutant, the aor1 gene was first inactivated using ClosTron mutagenesis in the ΔpyrE strain, before IFD of aor2 was undertaken by allelic exchange using the pyrEbased KO vector (pMTL84151-Δaor2) and counter selection using FOA. Following creation of an aor1+2 KO strain, the mutant pyrE allele was restored to WT (uracil prototrophy) using a specially constructed ACE correction vector, analogous to pMTL-YN1 of C. difficile (Ng et al., 2013). PCR verification, Sanger sequencing and Southern Blot analysis was performed to confirm the genotype of this strain (Fig. 2).
To explore the consequences of the deletion of both domains of adhE1, as well as adhE1 + adhE2, appropriate IFD mutants of C. autoethanogenum were sought using pyrE-directed allelic exchange. In a first attempt, the strain ΔadhE1 mut was obtained in which the adhE1 was deleted but in which 84 bp encompassing the adhE2 promoter region were also missing ( Fig. S3A & D). The deleted adhE2 84 bp promoter region was found to be flanked by two 9 bp repeat regions (Fig. S3D). Its deletion most likely placed the adhE2 gene under the transcriptional control of the stronger adhE1 promoter (Marcellin et al., 2016;Mock et al., 2015), a hypothesis corroborated by experimental comparison of adhE2 mRNA levels present in the ΔadhE1 mut strain and the wildtype (Fig. S10). A second attempt at generating a 'clean' IFD strain of adhE1 without the unintended 84 bp deletion was successful (Fig. S3B). However, the KO plasmid used to make the deletion could not be cured from the resultant ΔadhE1 strain. Similarly, the KO plasmids used to generate a strain (ΔadhE1+2) in which both adhE1 and adhE2 were successfully deleted (Fig. S3C) could also not be cured. Neither of these strains (ΔadhE1 and ΔadhE1+2) were therefore further characterized.

Metabolic engineering of the indirect, AOR ethanol pathway
As the genome of C. autoethanogenum encodes two AOR isoforms: aor1 (CLAU_0081) and aor2 (CLAU_0099), both single and a double mutant were generated and characterized during autotrophic and heterotrophic growth.

Characterization of single aor mutants
Whilst the genes encoding the two AOR isozymes are of the same length and share 78% identity, transcriptome data has shown that aor1 is expressed at 5-10x higher levels than aor2 during growth on CO (Marcellin et al., 2016;Mock et al., 2015). Both aor genes are expressed at higher levels during autotrophic growth as compared to heterotrophic growth (Marcellin et al., 2016). Consistent with the expression data, we found that inactivation of the more highly expressed aor1 gene restricted the eventual cell density of the mutant culture to half that of the WT (Fig. 3A) when grown on CO and reduced the levels of ethanol and 2,3-butanediol formed to 43% (p-value =0.019) and 23% (p-value < 0.0001) of the amounts produced by the WT, respectively (Fig. 3D & E). The levels of ethanol produced by the mutant could be restored to those of the WT through the introduction of a complementation plasmid, pMTL83151-P acsA -aor1 (Fig. S5). Growth of the aor1 KO strain was on the other hand not significantly affected on fructose as the carbon source, but ethanol production was reduced 33% (p-value =0.014) compared to the WT (Fig. S6B-D). In contrast, compared to the WT, the inactivation of aor2 consistently increased ethanol production during growth on CO (Fig. 3) or fructose (Fig. S6), by 170% (p-value=0.009) and 47% (p-value=0.003), respectively. The ratios of peak ethanol/peak acetate for aor1 KO strain and aor2 KO strains are 0.12 and 1.09, respectively, relative to 0.26 in the WT (Table 1). These results collectively suggest a contrasting role between aor1 and aor2 in ethanol production in C. autoethanogenum.
F. Liew et al. Metabolic Engineering 40 (2017) 104-114 alternative avenue for the offload of reduced ferredoxin is the reaction involving pyruvate: ferredoxin oxidoreductase (PFOR) that converts acetyl-CoA and CO 2 to pyruvate, which may subsequently alter the production of pyruvate-derived products such as 2,3-butanediol and lactate (Fig. 1). The 11-fold higher levels of lactate production (Fig. 3F) by the aor1 KO strain (but not by the aor2 KO strain) relative to WT indicated that the production of lactate, rather than 2,3-butanediol, is the predominant route for achieving redox balance in the event of aor1 inactivation. From pyruvate, the generation of lactate involves only one enzyme (lactate dehydrogenase) whereas the biosynthesis of 2,3butanediol involves three enzymes (acetolactate synthase, acetolactate decarboxylase and 2,3-butanediol dehydrogenase) (Köpke et al., , 2011 (Fig. 1).

Characterization of double aor mutants
During growth on pure CO, the aor1+2 double KO strain exhibited a prolonged lag phase, eventually achieving a cell density that was 69% lower than the WT (p-value < 0.0001) and was only able to reduce the headspace pressure by 101 kPa over the course of the experiment, relative to a decrease of 163 kPa in WT control (Fig. 3A & B). This retarded growth and poor gas consumption highlight the important role of AOR in supporting growth and utilization of CO.
In terms of metabolite production from CO, relative to the WT, the double KO strain produced 46% less ethanol (p-value =0.034), 38%  F. Liew et al. Metabolic Engineering 40 (2017) 104-114 less acetate (p-value < 0.0001), 66% less 2,3-butanediol (p-value < 0.0001) but 9-fold higher level of lactate (p-value < 0.0001) (Fig. 3C-F). These results indicate that ethanol could still be synthesized from CO via the direct reduction of acetyl-CoA in the aor1+2 KO strain. On H 2 +CO 2 , the growth lag phase of the aor1+2 double KO strain slightly increased but was able to grow to similar cell density to the WT and reduced the same amount of headspace pressure as the WT control ( Fig. 4A & B). On a molar basis, only half the amount of reduced ferredoxin is generated from H 2 than CO (Fig. 1), which may lead to less redox imbalance and explain why the KO strain was able to grow largely unaffected on H 2 +CO 2 . Acetate production was not affected but the KO strain produced 9.2-fold less ethanol than WT (p-value < 0.0001) (Fig. 4C & D). This is also reflected in the lower peak ethanol/ peak acetate ratio of 0.02 in the aor1+2 KO strain relative to 0.16 in the WT (Table 1). No lactate or 2,3-butanediol was produced by either strain (data not shown).
The finding that a very high specific AOR activity was detected in the cell extract of H 2 +CO 2 -grown C. autoethanogenum, which was also 4-fold higher than CO-grown cell extract and 5.3-fold higher than fructose-cultivated cells (Mock et al., 2015), highlighted the significance of AOR in ethanol biosynthesis during H 2 +CO 2 conditions. Our results confirmed the prediction of Fast and Papoutsakis (2012), Bertsch and Muller (2015), and Mock et al. (2015) that very little ethanol can be generated under the ATP-limiting H 2 +CO 2 conditions without the action of AOR. Coincidentally, prominent autotrophic ethanol producers such as C. ljungdahlii (Köpke et al., 2010) and C. carboxidivorans (Bruant et al., 2010) possess AOR whereas nonethanol producing acetogens such as Acetobacterium woodii (Poehlein et al., 2012) lack a functional AOR (Bertsch and Muller, 2015;Mock et al., 2015).
Under ATP-sufficient heterotrophic growth on fructose, the growth, ethanol and 2,3-butanediol production of the aor1+2 double KO strain was not significantly affected (Fig. S6). In Pyrococcus furiosus, the deletion of its only AOR resulted in minimal ethanol production while growing on maltose (Basen et al., 2014).

Carboxylic acid reduction in aor double mutant
AOR-harbouring acetogens such as C. ljungdahlii and "C. ragsdalei" have been shown to catalytically reduce a range of carboxylic acids, such as propionic acid, butyric acid, valeric and caproic acid into the corresponding primary alcohols using CO as electron donor (Isom et al., 2015;Perez et al., 2013). To investigate whether the AOR in C. autoethanogenum is capable of catalyzing such reactions, the WT and aor1+2 double KO strain were subjected to CO growth in the presence of supplemented 60 mM acetate, 40 mM propionate and 40 mM butyrate.
The supplementation of 60 mM acetate (a physiological metabolite) had a stimulatory effect on the CO growth of C. autoethanogenum WT as the lag phase was reduced from 5 days to 1 day ( Fig. 3A and Fig.  S7A), but not the KO strain. Up to 31.3 mM acetate was consumed by the WT during early exponential phase but a net production of 79.1 mM acetate was recorded at stationary phase (Fig. S7C). In the stationary phase, up to 70.8 mM ethanol was generated by the WT (Fig.  S7D). In contrast, the aor-deficient strain was not able to consume acetate during any of the growth stages and produced only 7.2 mM ethanol ( Fig. S7C & D). The reduction of acetic acid to acetaldehyde with reduced ferredoxin is thermodynamically unfavourable under standard conditions (ΔG°′=35 kJ/mol) (Thauer et al., 1977) because of the extremely low potential reaction (E°′=−580 mV) (Loach, 1976). However, at physiological conditions with intracellular pH of 6.0 and 1000-fold higher intracellular acetate than acetaldehyde concentrations, the reaction is exergonic (Mock et al., 2015). The consumption of acetate with concomitant production of ethanol during exponential growth of C. autoethanogenum indicates that the acetogen readily catalyzes the reduction of acetic acid using CO as reductant.
Similar to the supplementation of acetate, the addition of nonphysiological substrate propionate during CO cultivation reduced growth lag phase of WT from 5 days to 2 days ( Fig. 3A and Fig.  S8A), whereas the growth lag phase of the aor1+2 double KO strain was not altered. An increase in cell density (OD 600 ) from 0.61 (no supplementation) to 1.1 (propionate supplementation) and a reduction of headspace pressure to the same level of the WT was observed for the double KO strain (Fig. S8B). The supplementation of propionate increases acetate and ethanol formation, but reduces 2,3-butanediol and lactate production of WT growing on CO ( Fig. 3 and Fig. S8). Propionate concentrations remained unchanged and no 1-propanol was detected in cultures of double KO strain (Fig. S8G & H). In contrast, in cultures of the WT strain 24.2 mM propionate was consumed and 20.9 mM 1-propanol produced during the exponential growth phase (Fig. S8G & H).
In the case of butyrate supplementation in the presence of CO, the aor1+2 double KO strain grew to a similar OD 600 as the WT and reduced headspace pressure to the same extent ( Fig. S9A & B). The KO strain produced 17% more acetate (p-value=0.019), 36% more 2,3butanediol (not statistically significant) and 2.8 mM lactate (the WT produced none) but 44% less ethanol (p-value=0.016) than the WT. Consistent with the inability to metabolize acetate and propionate, the KO strain showed no consumption of the supplemented butyrate and produced no 1-butanol ( Fig. S9G & H). In contrast, in WT cultures 7.4 mM butyrate was consumed and 6.0 mM 1-butanol produced during the stationary growth phase (Fig. S9G & H).
Taken together, these results demonstrated that the AOR of C. autoethanogenum is required for the reduction of carboxylic acids into their corresponding primary alcohols. The apparent wide substrate range of AOR in C. autoethanogenum is consistent with the finding that the crystal structure of AOR from P. furiosus identified a channel that is sufficiently spacious to accommodate a range of substrates including aliphatic and aromatic aldehydes (Chan et al., 1995). The AOR from C. autoethanogenum could therefore be heterologously expressed in butyrate-producing acetogens such as Clostridium drakei (Gössner et al., 2008), Clostridium scatologenes (Küsel et al., 2000), Eubacterium limosum (Genthner et al., 1981) and Oxobacter pfennigii (Krumholz and Bryant, 1985) to generate 1-butanol.

Metabolic engineering of direct, AdhE ethanol pathway
AdhE typically consists of an N-terminal acetylating Ald domain followed by a C-terminal Fe-type Adh domain (Extance et al., 2013;Membrillo-Hernandez et al., 2000). Since deletion studies and the characterization of the separate AdhE domains indicate that the Ald and Adh domains are functionally autonomous (Arnau et al., 1998;Chen et al., 2004;Espinosa et al., 2001), the Ald domain and Adh domains of adhE1 in C. autoethanogenum was independently disrupted using ClosTron, generating the strains 'adhE1a KO' and 'adhE1b KO', respectively. For the 'adhE2' KO strain, only the Ald domain was targeted. Whilst a number of IFD were also successfully made, the ΔadhE1 and ΔadhE1+2 mutants were not explored further as the KO plasmids used to make the deletions could not be cured (Section 3.1). In another mutant, ΔadhE1 mut , an 84 bp region encompassing the adhE2 promoter was deleted, placing this gene under the control of the stronger adhE1 promoter (Marcellin et al., 2016;Mock et al., 2015) (Fig. S3).

Heterotrophic growth of adhE mutants
Growth of both the adhE1a KO and adhE1b KO strains on fructose was characterized by a slightly longer lag phase than the WT, but the cells eventually grew to a similar OD 600 (Fig. 5A). In contrast, the final OD 600 of the adhE2 KO strain was 28% lower (p-value < 0.0001) than the WT (Fig. 5A). Even after 13 days of incubation, 0.92 g/L of fructose was detected in the adhE2 KO strain culture, whereas all the other strains completely exhausted the substrate prior to day 3 (data not shown). All three adhE KO strains reached peak acetate levels of 72.2-76.5 mM, which are 31-43% higher than the WT (p-values < 0.05) (Fig. 5B). When compared to the WT, both adhE1 KO strains produced similar amounts of ethanol but the adhE2 KO strain only generated 37% of the WT ethanol titres (p-value=0.0035) (Fig. 5C). All three adhE KO strains produced less than half of the 2,3-butanediol recorded in the WT culture (p-values < 0.05) (Fig. 5D). During growth on fructose, the ΔadhE1 mut strain produced similar amounts of ethanol as the WT, and qRT-PCR results showed that adhE2 mRNA levels are higher in the mutant strain (Fig. S10).
The finding that ethanol production from all three adhE1 inactivation strains (adhE1a KO, adhE1b KO and ΔadhE1 mut ) was not impaired during heterotrophic growth contradicts the finding of Leang et al. (2013), who showed that the deletion of C. ljungdahlii adhE1 (but not adhE2) resulted in a strain that produced 6-fold less ethanol than the WT control. Furthermore, our results in C. autoethanogenum demonstrated that adhE2 inactivation resulted in 63% lower ethanol concentration than the WT.
While C. autoethanogenum and C. ljungdahlii are very similar on a genetic level (Brown et al., 2014;Marcellin et al., 2016), phenotypic differences that include different ethanol production profiles, are known (Brown et al., 2014;Cotter et al., 2009;Köpke et al., 2010;Liew et al., 2016b;Marcellin et al., 2016;Martin et al., 2016). The AdhE1 and AdhE2 enzymes of C. autoethanogenum and C. ljungdahlii differ by three and eight amino acids (AA), respectively. It is possible that one of these substitutions results in modification of substrate and cofactor specificities. For example, a single AA change in the Fe-Adh domain of the C. thermocellum AdhE changed its cofactor from NADH to NADPH (Brown et al., 2011). One of the C. autoethanogenum AdhE2 changes relative to C. ljungdahlii resides in the NADH binding site of Adh domain. A change in cofactor specificity would be expected to have significant impact on electron and carbon flows because NADH is commonly used in catabolic reactions whereas NADPH is usually employed as reductant in anabolic processes (Alberts et al., 2002). Another possible explanation for the contradictory phenotypes is that C. autoethanogenum may possess other ethanologenic enzymes that compensate for the loss of AdhE activities during fructose growth.
RNA-sequencing experiments in both C. autoethanogenum (Marcellin et al., 2016) and C. ljungdahlii (Nagarajan et al., 2013;Tan et al., 2013) showed that adhE1 is transcribed at significantly higher levels when growing on fructose compared to autotrophic growth, which suggests a more important role for this gene under heterotrophic than autotrophic conditions. Unfortunately, the C. ljungdahlii adhE1 mutant (Leang et al., 2013) has only been characterized during heterotrophic growth (with 5 g/L fructose rather than 10 g/L as in this study) but not for autotrophic growth.

Autotrophic growth of AdhE mutants
During growth on pure CO, all three adhE KO strains (adhE1a, adhE1b, and adhE2) displayed significant growth deficiencies in the form of prolonged lag phase and 47-55% lower cell density than WT (p-values < 0.01) (Fig. 6A), which suggests inefficiency in recycling reducing equivalents. Despite the low biomass, all three adhE KO strains consistently generated 154-183% higher titres of ethanol while growing on CO. Specifically, the adhE1a KO strain produced 53.4 mM ethanol, 183% more than WT (p-value =0.0005). The adhE1b KO strain produced 171% more ethanol (not statistically significant) and the adhE2 KO strain produced 154% more ethanol than WT (p-value =0.021) (Fig. 6C). The enhanced ethanol production of these adhE mutants is also reflected in the peak ethanol/peak acetate ratio of 0.58−1.11, in comparison to 0.26 in WT (Table 1). The substantial improvements in ethanol production were partially offset by a reduction of 48-68% in 2,3-butanediol titres (p-values < 0.004) (Fig. 6D). Given the similarities in phenotypes between adhE1a KO strain and adhE1b KO strain, the position of ClosTron insertion within adhE1 (at Ald domain or Adh domain) played an insignificant role in the overall phenotype of the mutant. The ΔadhE1 mut strain, which exhibited an upregulated adhE2 mRNA expression (Fig. S10D), produced 27% more ethanol than the parental strain (not statistically significant; Fig.  S11C).
The marked increase in ethanol production exhibited by the adhE inactivation strains while growing on CO is in agreement with the hypothesis that the ATP-efficient, indirect ethanol formation route employing AOR is more favourable for autotrophic ethanol biosynthesis (Bertsch and Muller, 2015;Fast and Papoutsakis, 2012;Mock et al., 2015). It has been hypothesized by Mock et al. (2015) that the CoA-linked acetaldehyde dehydrogenase activity measured in the H 2 +CO 2 -grown C. autoethanogenum physiologically only facilitate the reuse of the ethanol formed. In the presence of high ethanol concentration and low H 2 concentration, ethanol oxidation to acetyl-CoA is hypothesized to be coupled to the reduction of 2 CO 2 to acetate (Mock et al., 2015). In support of this notion, C. autoethanogenum WT growing on H 2 +CO 2 transiently produced 10.3 mM ethanol during exponential growth but thereafter there was a steep decline to 1.8 mM F. Liew et al. Metabolic Engineering 40 (2017) 104-114 during stationary phase. In addition to the two adhE genes, there are 3 other mono-functional ald genes (CLAU_1772, 1783 in the genome of C. autoethanogenum. Accordingly, the generation of a triple ald KO strain may further channel carbon and electrons towards acetate synthesis and ethanol formation via AOR.

Conclusion
Conventional strategies that seek to enhance ethanol production commonly employ the introduction or overexpression of AdhE (Peng et al., 2008;Thapa et al., 2015;Yao and Mikkelsen, 2010). However, given the unique metabolism of acetogens living on the thermodynamic edge of life, we found that the inactivation of adhE a better strategy due to diversion of carbon and reducing equivalents towards the ATPyielding acetate formation (Figs. 4D and 6C). The acetic acid can be reduced to acetaldehyde (via AOR, using reduced ferredoxins) and then ethanol via NAD(P)H-dependent Adh (Fig. 1). Using this strategy we have generated strains that produce up to 180% more ethanol and also accumulate up to 38% less of the undesired acetate by-product. The indirect ethanol pathway has been postulated before (Köpke et al., 2010;Mock et al., 2015) but proof on genetic level was missing. Generation of aor mutants confirmed the important role of this route in autotrophic ethanol production and also exhibited that the two isozymes in C. autoethanogenum have different roles. To enable this work, we have adapted an allelic exchange strategy that allows generation of stable deletion mutants of multiple genes which has been a limitation in advancing acetogens as platform organisms. Interestingly, lactate but not 2,3-butanediol was found to be the major sink of additional reducing power and further optimization for autotrophic ethanol production could come by combining deletion of lactate dehydrogenase gene ldh  with deletions of aor2 and adhE genes identified in this study.