Acetobacterium woodii (DSM 1030) and Clostridium ljungdahlii (DSM13583) were obtained from the DSMZ (Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH, Brunswick, Germany). Escherichia coli XL1-Blue MRF’ [Δ(mcrA)183Δ(mcrCB-hsdSMR-mrr)173 endA1 supE44 thi-1] stemmed from Stratagene (La Jolla, CA). All plasmids used in this study are listed in Table 1.

Escherichia coli was cultivated aerobically in LB (Luria-Bertani) medium (per l: tryptone 10 g, yeast extract 5 g, NaCl 10 g) (Green and Sambrook, 2012) at 37°C on a rotary shaker (175 rpm). For generation of competent cells, E. coli was grown in modified SOB (Super Optimal Broth; per l: tryptone 20 g, yeast extract 5 g, NaCl 0.5 g, KCl 1.92 g, MgCl₂ × 6 H₂O 2.03 g) (Green and Sambrook, 2012).

Basal medium for A. woodii was modified from Balch et al. (1977) and contained per l: KH₂PO₄ 0.33 g, K₂HPO₄ 0.45 g, MgSO₄ × 7 H₂O 0.33 g, NH₄Cl 1 g, yeast extract 2 g, L-cysteine-HCl × H₂O 0.5 g, resazurin 1 mg. To this solution, 20 ml of trace element solution (per l: nitrilotriacetic acid 1.5 g, MgSO₄ × 7 H₂O 3 g, MnSO₄ × H₂O 0.5 g, NaCl 1 g, FeSO₄ × 7 H₂O 0.1 g, CoSO₄ × 7 H₂O 0.18 g, CaCl₂ × 2 H₂O 0.1 g, ZnSO₄ × 7 H₂O 0.18 g, CuSO₄ × 5 H₂O 0.01 g, KAl(SO₄)₂ × 12 H₂O 0.02 g, H₃BO₃ 0.01 g, Na₂MoO₄ × 2 H₂O 0.01 g, NiCl₂ × 6 H₂O 0.03 g, Na₂SeO₃ × 5 H₂O 0.3 mg, Na₂WO₄ × 2 H₂O 0.4 mg) and 20 ml vitamin solution (per l: biotin 2 mg, folic acid 2 mg, pyridoxine-HCl 10 mg, thiamine-HCl 5 mg, riboflavin 5 mg, nicotinic acid 5 mg, D-Ca-pantothenate 5 mg, vitamin B₁₂ 0.1 mg, p-aminobenzoic acid 5 mg, lipoic acid 5 mg (DSMZ_Medium141-1.pdf,¹) were added. Media were prepared under strictly anaerobic conditions. Heterotrophic growth was performed with 40 mM fructose as a carbon source under an atmosphere consisting of N₂ (80%) and CO₂ (20%), autotrophic growth with a CO₂ + H₂ gas mixture (33% + 67%), both at 30°C.

Clostridium ljungdahlii was cultivated in a modified medium described by Tanner et al. (1993). It contained per l: 2-(N-morpholino) ethanesulfonic acid 20 g, yeast extract 0.5 g, L-cysteine-HCl × H₂O 1 g, resazurin 1 mg. To this solution, 25 ml of mineral salts solution (per l: CaCl₂ × 2 H₂O 4 g, KCl 10 g, KH₂PO₄ 10 g, MgSO₄ × 7 H₂O 20 g, NaCl 80 g, NH₄Cl 100 g), 10 ml of modified trace element solution (American Type Culture Collection (ATCC);²; 1754 PETC medium) (per l: nitrilotriacetic acid 2 g, MnSO₄ × H₂O 0.5 g, Fe(SO₄)₂(NH₄)₂ × 6 H₂O 0.8 g, CoCl₂ × 6 H₂O 0.2 g, ZnSO₄ × 7 H₂O 1 mg, CuCl₂ × 2 H₂O 20 mg, NiCl₂ × 6 H₂O 20 mg, Na₂MoO₄ × 2 H₂O 20 mg, Na₂SeO₃ × 5 H₂O 20 mg, Na₂WO₄ × 2 H₂O 20 mg), and 10 ml vitamin solution (per l: biotin 2 mg, folic acid 2 mg, pyridoxine-HCl 10 mg, thiamine-HCl 5 mg, riboflavin 5 mg, nicotinic acid 5 mg, D-Ca-pantothenate 5 mg, vitamin B₁₂ 5 mg, p-aminobenzoic acid 5 mg, and lipoic acid 5 mg) were added. Media were prepared under strictly anaerobic conditions. Heterotrophic growth was performed with 40 mM fructose as a carbon source under an atmosphere consisting of N₂ (80%) and CO₂ (20%), autotrophic growth with a syngas mixture (50% CO, 45% H₂, and 5 CO₂), both at 37°C.

For cultivation of A. woodii and C. ljungdahlii on solid media, a modified version of yeast tryptone fructose medium (Leang et al., 2013) was used (per l: yeast extract 10 g, tryptone 16 g, fructose 5 g, NaCl 4 g, L-cysteine-HCl × H₂O 1 g, resazurin 1 mg, agar 15 g). Agar plates were poured using anaerobic water in an anaerobic cabinet containing a N₂:H₂ (95:5%) gas atmosphere. Cell suspensions were dripped on plates and spread using glass beads. A. woodii cells were incubated at 30°C and C. ljungdahlii cells at 37°C in an incubator located in the anaerobic cabinet.

For selection of antibiotic-resistant strains, the following concentrations were used per ml: ampicillin 100 mg, clarithromycin 2.5 mg, chloramphenicol 30 mg, kanamycin 50 mg, thiamphenicol 7.5 mg.

Cell growth was monitored offline by measuring the optical density at 600 nm (Ultrospec 3100, Amersham Bioscience Europe GmbH, Freiburg, Germany) in 1-ml cuvettes (width 1 cm).

Ethanol, isobutanol, and isoamyl alcohol were determined using gas chromatograph (GC) “Clarus 680” (PerkinElmer, PerkinElmer, Waltham, MA, United States). GC was equipped with an Elite-FFAP column (i ∅ 0.32 mm × 30 m). H₂ was the carrier gas (2.25 ml min^–1, 100°C), injector temperature was 225°C, split 1:20, and flame ionization detector temperature was 300°C. Detector gases were synthetic air (450 ml min^–1) and H₂ (45 ml min^–1). A temperature profile was predefined: 90°C for 2 min, 40°C min^–1 increasing steps to 250°C (constant for 1 min). Supernatant (0.48 ml) was acidified with 0.02 mL of 2 M HCl. 1 ml was injected into the GC. Calibration was performed using defined standards of the individual components.

Acetate, 2,3-butanediol, fructose, and ketoisovalerate were determined using high performance liquid chromatography (Agilent 1260 Infinity Series HPLC, Agilent Technologies, Santa Clara, CA) equipped with a refractive index detector (for 2,3-butanediol, fructose, ketoisovalerate) operating at 35°C and a diode array detector (for acetate) at room temperature. The “CS-Chromatographie organic acid column” (CS-Chromatographie Service GmbH, Langerwehe, Germany) was kept at 40°C. 5 mM H₂SO₄ was used as mobile phase with a flow rate of 0.7 ml min^–1. 20 μl of supernatant were injected into the HPLC system for determination of compounds.

Bacterial genomic DNA was isolated using the “MasterPure^TM Gram-Positive DNA Purification Kit” (Epicentre, Madison, WI, United States). 2-mL Samples of late exponential cultures were centrifuged (18,000 g, 30 min, 4°C) and further processed according to the manufacturer’s instructions. Isolation of plasmid DNA from E. coli strains was performed with the “Zyppy^TM Plasmid Miniprep Kit” (ZYMO Research Europe GmbH, Freiburg, Germany). 4-ml Samples of overnight cultures were centrifuged (18,000 g, 1 min) and further processed according to the manufacturer’s instructions. In case of A. woodii and C. ljungdahlii, 2 ml were sampled and centrifuged (18,000 g, 30 min, 4°C). The cell pellet was suspended in 600 μl Tris−HCl (20 mM, pH 7). For effective lysis, 60 μl lysozyme (20 mg ml^–1) were added, and the solution was incubated for 1 h at 37°C before proceeding according to the manufacturer’s instructions.

DNA from C. thermocellum DSM 1313 was purchased from the DSMZ (Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH, Brunswick, Germany).

Sequencing of DNA was performed by GATC Biotech AG (Constance, Germany).

Standard molecular cloning techniques were performed according to established protocols (Green and Sambrook, 2012). In C. thermocellum, three gene clusters encoding a putative ketoisovalerate ferredoxin oxidoreductase (Kor) are annotated [kor1 (Clo1313_0020-0023), kor2 (Clo1313_0382-0385), kor3 (Clo1313_1353-1356)]. kor1 was PCR-amplified using primers clo0020T_fwd and clo0020T_rev, kor2 was PCR-amplified using primers clo0382T_fwd and clo0382T_rev, and kor3 was PCR-amplified using primers clo1353T_fwd and clo1353T_rev. Then, each fragment was subcloned into the pMTL83151 backbone (Heap et al., 2009), together (upstream) with the also PCR-amplified P_pta–ack promoter from C. ljungdahlii (primers: Ptaack_fwd and Ptaack_rev) and the gene encoding bifunctional butyraldehyde/butanol dehydrogenase AdhE2 from C. acetobutylicum [primers: adhE2_pka_fwd and adhE2_pka_rev; plasmid pMTL83151_ptaack_aacht_cac (Supplementary Figure 1)]. All primers used are listed in Table 2. Subcloning was performed by in-fusion cloning using the NEBuilder Assembly Tool (New England Biolabs Inc., Ipswich, MA, United States;³). The resulting plasmids pKOR1, pKOR2, and pKOR3 are shown in Supplementary Figure 2. pMTL83151 was kindly provided by Nigel Minton (University of Nottingham, United Kingdom).

The plasmids constructed for elucidation of the ketoisovalerate decarboxylase pathway were also based on the pMTL83151 backbone. The kivD gene from L. lactis was PCR-amplified using primers kivd_fwd_XhoI and kivd_rev_Eco81I (making use of the inserted restriction sites for subcloning), adhA from C. glutamicum was PCR-amplified using primers adhA_fwd_XhoI and adhA_rev_NheI (making use of the inserted restriction sites for subcloning). DNA source for amplification was plasmid pJUL34 (Supplementary Figure 3), which was kindly provided by Bastian Blombach (Technical University of Munich, Germany) and Bernhard Eikmanns (University of Ulm, Germany) (Table 1). Both fragments were subcloned together in several steps into pMTL83151. Upstream of kivD, the PCR-amplified P_pta–ack promoter fragment from C. ljungdahlii (primers: Ptaack_fwd and Ptaack_rev) was inserted. The genes ilvC (encoding ketol-acid reductoisomerase) (primers ilvC_fwd and ilvC_rev), ilvD (encoding dihydroxy-acid dehydratase) (primers ilvD_fwd and ilvD_rev), and alsS (encoding acetolactate synthase) (primers alsS_fwd and alsS_rev), all from C. ljungdahlii, were PCR-amplified (Table 2). Subcloning downstream of adhA was performed by in-fusion cloning using the NEBuilder Assembly Tool (New England Biolabs inc., Ipswich, MA, United States; see text footnote 3). The resulting plasmid was designated pKAIA (Supplementary Figure 4).

For changing the coenzyme specificity of IlvC (ketol-acid reductoisomerase) of C. ljungdahlii from NADPH to NADH, the ilvC^NADH gene of E. coli was commercially synthesized according to Brinkmann-Chen et al. (2013), with appropriate overlaps for in-fusion cloning. Codon optimization for clostridia was suggested by the commercial supplier Thermo Fisher Scientific GENEART GmbH (Regensburg, Germany). The sequence of the codon-optimized, synthesized gene is provided in Supplementary Figure 5.

The ilvC gene was cut out of pKAIA and the codon-optimized, synthesized gene was subcloned into the linearized vector using in-fusion cloning, resulting in plasmid pKAI_NADHA (Supplementary Figure 6).

The ClosTron plasmid pMTL007-E2_ilvE (Table 1) was synthesized by the company DNA 2.0 (now ATUM, Newark, CA, United States) (see below).

In case of E. coli, competent cells (Inoue et al., 1990) were obtained by inoculating 250 ml SOB medium with a 5-ml overnight culture grown in the same medium. Incubation was performed in a 2-l Erlenmeyer flask at 18°C under aerobic conditions and shaking (60 rpm), until an optical density (600 nm) of 0.6–0.8 was reached. Then, cells were put on ice for 30 min and afterward centrifuged (4,500 g, 4°C, 10 min). The pellet was suspended in 40 ml buffer (per 125 ml: piperazine-N,N’–bis(2-ethanesulfonic acid) 0.756 g, CaCl₂ 0.42 g, pH 6.7; mixed with 125 ml containing 4.66 g KCl and 1.72 g MnCl₂ × 4 H₂O), incubated on ice for 10 min, and again centrifuged (4,500 g, 4°C, 10 min). The pellet was suspended in 10 ml of the same buffer. 1.5 ml sterile dimethyl sulfoxide were added, and the solution was split into 200 μl aliquots and stored at −80°C. For transformation, such aliquots were thawed on ice, mixed with 10 μl of a DNA solution (in-fusion reaction mix or plasmid), incubated on ice for 10 min, and then at 42°C for 1 min. After an incubation period of 10 min again on ice, 800 μl LB medium were added, and the solution was incubated aerobically for 1 h at 37°C with shaking (160 rpm). Then, cells were centrifuged (4,000 g, room temperature, 3 min), and 800 μl of the supernatant discarded. The pellet was resuspended in the remaining liquid, which was used for plating on solid media with respective selection.

Electroporation of A. woodii and C. ljungdahlii was performed using a modified protocol of Leang et al. (2013). All plastic materials were stored in an anaerobic cabinet at least 1 day before transformation to eliminate any remaining traces of oxygen. For preparation of competent cells, 100 ml medium were supplemented with 40 mM fructose and 40 mM DL-threonine, inoculated with an early exponential culture, and grown at the appropriate temperature to an optical density (600 nm) of 0.3–0.7. Cultures were centrifuged anaerobically (6,000 g, 4°C, 10 min), and the pellet was washed twice with 50 ml cool, anaerobic SMP buffer (per l: sucrose 92.4 g, MgCl₂ × 6 H₂O 0.2 g, NaH₂PO₄ 0.84 g; pH 6) and suspended in 0.6 ml SMP buffer. 120 μl cool anti-freezing buffer (mix of 20 ml SMP buffer and 30 ml dimethylformamide) were added, and aliquots stored at −80°C.

Electroporation was carried out in an anaerobic cabinet. 25 μl of electrocompetent cells were mixed with 3–5 μg of plasmid DNA, cooled, and transferred to a pre-cooled 1-mm gap electroporation cuvette (Biozym Scientific GmbH, Hessisch Oldendorf, Germany). Electric pulse was performed with 625 V, resistance of 600 Ω, and a capacitance of 25 μF using a “Gene-Pulser Xcell^TM” pulse generator (Bio-Rad Laboratories GmbH, Munich, Germany). Afterward, cells were recovered using 5 ml of the respective medium without antibiotic in a Hungate tube and incubated until the optical density (600 nm) doubled. Then, the appropriate antibiotic was added and growth monitored. If growth was detectable, cells were inoculated into fresh medium with antibiotic and, after reaching the early exponential growth phase, plated onto solid media. Single colonies of obtained transformants were picked, and successful transformation was confirmed by isolating genomic or plasmid DNA and PCR amplification of respective sequences for verification.

The ClosTron method was performed according to Heap et al. (2007, 2010). The aim was to create an insertion mutant in ilvE of C. ljungdahlii, which blocked the last step of valine biosynthesis. Target region for integration was determined using the respective algorithm (Perutka et al., 2004;⁴), and the respective ClosTron plasmid (pMTL007-E2_ilvE) (Table 1) was synthesized by the company DNA 2.0 (now ATUM, Newark, CA). After transformation into C. ljungdahlii, recombinant strains were verified by PCR amplification and sequencing of the respective gene region.

In the genome of C. thermocellum three gene clusters are annotated that encode a putative ketoisovalerate ferredoxin oxidoreductase (Kor), i.e., kor1 (Clo1313_0020-0023), kor2 (Clo1313_0382-0385), and kor3 (Clo1313_1353-1356). kor3 differs significantly from the other two clusters with respect to length (smaller) and gene arrangement (Supplementary Figure 7). However, all clusters consist of four genes, encoding the α, β, γ, and δ subunits of Kor. The α subunit carries the core domain of pyruvate-ferredoxin oxidoreductase, β a C-terminal thiamine pyrophosphate-binding domain, γ a catalytic domain of pyruvate/ketoisovalerate oxidoreductase, and δ a binding domain for 4Fe-4S ferredoxin. All clusters were PCR-amplified and subcloned each into the pMTL83151 backbone, together (upstream) with the P_pta–ack promoter from C. ljungdahlii and the gene encoding bifunctional butyraldehyde/butanol dehydrogenase AdhE2 from C. acetobutylicum. The resulting plasmids (pKOR1, pKOR2, and pKOR3) were transformed into A. woodii and C. ljungdahlii. Recombinant strains were then tested under heterotrophic conditions with fructose as a carbon source as well as with and without addition of ketoisovalerate (15 mM). Both, A. woodii and C. ljungdahlii did not grow with ketoisovalerate as sole carbon and energy source and also did not show an increase in optical density when fructose and ketoisovalerate were supplied together, compared to fructose alone. Wild type strains and empty plasmid-carrying strains [A. woodii (pM83) and C. ljungdahlii (pM83)] were used as controls. Without ketoisovalerate addition, recombinant A. woodii strains only formed traces of isobutanol (0.1 mM). However, all pKOR-carrying strains produced increased amounts of ethanol. With addition of ketoisovalerate, A. woodii[pKOR1] produced 0.2 mM isobutanol, A. woodii[pKOR2] 0.3 mM, and A. woodii[pKOR3] 2.9 mM (Figure 2). Some isoamyl alcohol (up to 0.3 mM) was formed in addition. The respective data for C. ljungdahlii are shown in Supplementary Figure 8. With and without ketoisovalerate addition, no isobutanol was formed. pKOR-carrying strains produced increased amounts of ethanol with addition of ketoisovalerate.

For autotrophic growth, A. woodii was cultivated under a CO₂ + H₂ atmosphere, C. ljungdahlii under syngas. Without ketoisovalerate addition, recombinant A. woodii strains formed no isobutanol. With addition, A. woodii[pKOR1] produced 0.3 mM, A. woodii[pKOR2] 1.8 mM, and A. woodii[pKOR3] 1.1 mM (Figure 3). A. woodii[pKOR1] and A. woodii[pKOR2] produced also increased amounts of ethanol. Some isoamyl alcohol (up to 0.4 mM) was formed in addition. The respective data for C. ljungdahlii are shown in Supplementary Figure 9. Without ketoisovalerate addition, no isobutanol was formed and even with addition only trace amounts [C. ljungdahlii(pKOR2) 0.1 mM; C. ljungdahlii(pKOR3) 0.2 mM]. However, C. ljungdahlii[pKOR3] showed a significant increase in ethanol formation without ketoisovalerate addition as well as C. ljungdahlii[pKOR2] with addition.

Next, the isobutanol synthesis pathway via ketoisovalerate decarboxylase and alcohol dehydrogenase in acetogens was elucidated. The kivD gene from L. lactis and adhA from C. glutamicum were PCR-amplified and subcloned together into the pMTL83151 backbone, controlled by the also PCR-amplified P_pta–ack promoter from C. ljungdahlii. In order to increase the carbon flux from pyruvate to ketoisovalerate, genes ilvC (encoding ketol-acid reductoisomerase), ilvD (encoding dihydroxy-acid dehydratase), and alsS (encoding acetolactate synthase), all from C. ljungdahlii, were PCR-amplified and subcloned downstream of adhA. The resulting plasmid pKAIA was transformed into A. woodii and C. ljungdahlii. In case of C. ljungdahlii, it was also possible to construct an insertion mutant in ilvE (encoding an aminotransferase), which blocked the last step of valine biosynthesis. This strain, C. ljungdahlii:ilvE was also transformed with pKAIA. Recombinant strains were then tested under heterotrophic conditions with fructose as a carbon source as well as with and without addition of ketoisovalerate (15 mM). Wild type strains and empty plasmid-carrying strains were used as controls. Without ketoisovalerate addition, only A. woodii[pKAIA] formed some isobutanol (0.2 mM). With addition, the same strain produced 0.4 mM (Supplementary Figure 10). Some isoamyl alcohol (up to 0.2 mM) was formed in addition. The respective data for C. ljungdahlii are shown in Supplementary Figure 11. Without ketoisovalerate addition, no isobutanol was formed. With addition, C. ljungdahlii[pKAIA] and C. ljungdahlii:ilvE[pKAIA] produced up to 2.4 mM isobutanol during the exponential growth phase.

Under autotrophic growth conditions, recombinant A. woodii strains formed no isobutanol. With addition of 15 mM isovalerate, A. woodii[pKAIA] produced 1.7 mM isobutanol and 0.8 mM isoamyl alcohol (Figure 4). The respective data for C. ljungdahlii are shown in Figure 5. Only C. ljungdahlii[pKAIA] and C. ljungdahlii:ilvE[pKAIA] formed isobutanol (and isoamyl alcohol) without ketoisovalerate addition. The latter strain formed three times more acohols (up to 0.4 mM isobutanol). Addition of ketoisovalerate increased production of both alcohols in the pKAIA-carrying strains (up to 1 mM in C. ljungdahlii:ilvE[pKAIA]).

The ilvC gene product of C. ljungdahlii, the ketol-acid reductoisomerase, uses NADPH for reduction of acetolactate to 2,3-dihydroxyisovalerate. However, NADPH might be a limiting factor in a catabolic pathway with high amounts of products. Therefore, this gene in pKAIA was replaced by a similar gene encoding a NADH-dependent IlvC enzyme from E. coli (Brinkmann-Chen et al., 2013). The nucleotide sequence of Ec_IlvC^P2D1–A1 was codon-optimized for clostridia, PCR-amplified, and subcloned into pKAIA by replacing the C. ljungdahlii ilvC gene. The resulting plasmid was designated pKAI_NADHA (Supplementary Figure 6). This plasmid was transformed into C. ljungdahlii:ilvE, and the recombinant tested in heterotrophic and autotrophic fermentations. With fructose as a carbon source, C. ljungdahlii:ilvE[pKAI_NADHA] produced isobutanol only upon addition of ketoisovalerate. The production was slightly higher than with the corresponding pKAIA-carrying strain (1.6 vs. 1.4 mM). Under autotrophic conditions without isovalerate supplementation, C. ljungdahlii:ilvE[pKAI_NADHA] formed only trace amounts of isobutanol (0.1 mM). Even with isovalerate addition, lower amounts were produced than by C. ljungdahlii:ilvE[pKAIA] (0.8 mM compared to 1 mM).