Maintenance of strains

Plasmid and cassette construction

For markerless genomic integration of gene cassettes, plasmids expressing unique gRNAs targeting the SGA1 locus or the intergenic region X-2 [33] were constructed. The plasmid backbones of puDR119 and pURD164 were obtained by PCR amplification using the primer combination 5792–5980 and plasmids pMEL11 and pMEL10 [34], respectively, as templates. The plasmid inserts of pUDR119 and pUDR164, containing the expression cassettes coding for the unique 20-bp gRNA sequences targeting SGA1 and X-2, respectively, were obtained by PCR amplification using the primer combinations 5979–7023 for SGA1, 5979–7374 for X-2, and plasmids pMEL11 and pMEL10, respectively, as templates. The assembly of plasmids pUDR119 and pUDR164 was performed in vitro using the Gibson Assembly Cloning kit (New England Biolabs, Ipswich, MA) following the supplier’s guidelines. The assembly was enabled by homologous sequences present at the 5′ and 3′ ends of the PCR-amplified plasmid backbones and inserts. In each case, 1 μL of the Gibson-assembly mix was used for E. coli XL-1 blue transformation by electroporation, performed in a Gene PulserXcell Electroporation System (Biorad, Hercules, CA). Correct assembly of plasmids was confirmed by diagnostic PCR (Dreamtaq, Thermo-Scientific) or restriction digestion. The constructed plasmids pUDR119 and pUDR164 were isolated from transformed E. coli cultures using a Sigma GenElute Plasmid kit (Sigma-Aldrich, St. Louis, MO) and used in transformations of S. cerevisiae.

For markerless deletion of GPD2, the plasmid backbone of pROS10 (URA3 marker) or pROS11 (amdS marker) was PCR amplified using primer combination 5793–5793 (double binding). The plasmid insert, containing the expression cassette coding for the unique 20-bp gRNA sequence targeting GPD2, was obtained using primer combination 6966–6966 (double binding) and plasmid pROS10 as template.

A yeast codon-optimized cassette for T. denitrificans cbbm overexpression [28] was obtained by PCR amplification using plasmid pBTWW002 as template and primer combination 7549–7550. The resulting fragment was ligated to a pJET/1.2 blunt vector (Thermo-Scientific) following the supplier’s protocol and cloned to E. coli. The resulting plasmid was used as PCR template to generate cbbm integration cassettes, using primer combinations 11206-6285, 6280–6273, 6281–6270, 6282–6271, 6284–6272, 6283–6275, 6287–6276, 6288–6277 and 6289–7075. The overexpression cassettes of cbbm (pTDH3-cbbm-tCYC1) were genetically identical, except for different overhangs present at the 5′ and 3′ ends of the fragments to allow for in vivo homologous recombination. Codon-optimized yeast expression cassettes of groEL (pTEF1-groEL-tACT1) and groES (pTPI1-groES-tPGI1) were obtained using plasmids pUD232 and pUD233 as templates and primer combinations 7076–7077 and 7078–7079, respectively.

The genomic sequence corresponding to the constitutive promoter of YEN1 [35] was obtained by PCR amplification with primer combination 7933–7295 and genomic DNA of IMX585 as template. The genomic sequence of the anaerobically inducible promoter of DAN1 [35] was obtained by PCR amplification with primer combinations 7930–7931 (integration at X-2) and 7978–7931 (integration at SGA1), using genomic DNA of IMX585 as template.

The terminator sequence of PGK1 was obtained by PCR amplification using primer combinations 7084–7934 (integration at X-2) and 7084–11,205 (integration at SGA1), using genomic DNA of IMX585 as template.

The S. oleracea prk-ORF was obtained by PCR amplification using primer combinations 7297–7081 (pYEN1-prk cassette construction), 7932–7081 (pDAN1-prk cassette construction), and plasmid pUDE046 as template. The various primer combinations resulted in prk-ORF fragments with homologous overhangs to the different promoter sequences and the terminator sequence of PGK1. The complete expression cassettes (pYEN1-prk-tPGK1 and pDAN1-prk-tPGK1) were assembled by in vivo homologous recombination after transformation to yeast and correct assembly was verified by diagnostic PCR and Sanger sequencing (Baseclear, Leiden, The Netherlands).

An integration cassette for RPE1 overexpression (pTDH3-RPE1-tRPE1) was PCR amplified using primer combination 11593-3290 and pUD347 as a template. Similarly, integration cassettes for overexpressions of TKL1 (pPGK1-TKL1-tTKL1), TAL1 (pTEF1-TAL1-tTAL1), NQM1 (pPGI1-NQM1-tNQM1), RKI1 (pTPI1-RKI1-tRKI1) and TKL2 (pPYK1-TKL2-t-TKL2) were obtained by PCR amplification using primer combinations 5909–4068, 3274–3275, 3847–3276, 4691–3277 and 3283–11595, respectively, with plasmids pUD348, pUD349, pUD344, pUD345 and pUD346, respectively, as templates. The integration cassettes included overhang sequences to allow for in vivo assembly of overexpression cassettes of the complete non-oxidative pentose-phosphate pathway and integration at the GPD2 locus.

Yeast genome editing and strain construction

The lithium-acetate transformation protocol [36] was used for yeast transformations. Transformation mixtures were plated on synthetic medium agar plates [31] (2% Bacto Agar, BD, Franklin Lakes, NJ), supplemented with 20 g L⁻¹ glucose (final concentration) in the case of transformations with plasmids expressing the URA3 marker. In transformations with plasmids expressing the amdS marker, agar plates were prepared as described previously [37]. Confirmation of the desired genotypes in each case was performed by diagnostic colony PCR using Dreamtaq polymerase (Thermo-Scientific), according to the manufacturer’s guidelines (Additional file 1). Counter-selection of plasmids expressing URA3 was performed using 5-fluoro-orotic acid (Zymo Research, Irvine, CA), following the supplier’s guidelines. Counter-selection of plasmids expressing amdS was performed as described previously [12].

Co-transformation of pUDR119, 9 copies of the cbbm expression cassette and single copies of the expression cassettes of groEL and groES to IMX581 (after plasmid recycling from the correct mutant) yielded the RuBisCO-expressing strain IMX765. Overhangs present at the 5′ and 3′ ends of the molecules allowed for in vivo assembly of the entire construct (11 fragments) and for integration at the SGA1 locus.

Co-transformation of the pYEN1 and pDAN1 sequences, respectively, the prk-ORF and the tPGK1 fragments, along with plasmid pUDR164 to strain IMX765 yielded strains IMX773 and IMX774. For construction of strain IMX949, in which GPD2 was deleted, the two fragments of the gRNA-expressing plasmid (pROS10 backbone) and the repair oligo-nucleotides 6969–6970 were co-transformed to IMX774 (after recycling of pUDR164). For construction of strain IMX1443, in which GPD2 was deleted and the genes of the non-oxidative branch of the pentose-phosphate pathway were overexpressed, the two fragments of the gRNA-expressing plasmid (pROS11 backbone), along with the integration cassettes pTDH3-RPE1-tRPE1, pPGK1-TKL1-tTKL1, pTEF1-TAL1-tTAL1, pPGI1-NQM1-tNQM1, pTPI1-RKI1-tRKI1 and pPYK1-TKL2-tTKL2, were co-transformed to IMX774. The entire construct (6 fragments) was assembled in vivo and integrated at the GPD2 locus. Before stocking of strain IMX1443, the GPD2-targeting CRISPR plasmid was recycled by counter-selection against its amdS marker [12].

Co-transformation of the two fragments of the GPD2-targeting CRISPR plasmid (pROS11 backbone) and the non-oxidative pentose-phosphate pathway integration cassettes to strain IMX581 yielded strain IMX1472. The RuBisCO/PRK-expressing strain IMX1489 was obtained by co-transformation of pUDR103, the pDAN1, prk-ORF, tPGK1 sequences, 9 copies of the expression cassette of cbbm and the expression cassettes of groEL and groES (14 fragments) to strain IMX1472 (integration at the SGA1 locus, GPD2-targeting CRISPR plasmid recycled).

The reference strains IME324 and IME369 were obtained by transformation of p426-TEF (empty) [38] to strains IMX581 and IMX673, respectively.

Bioreactor cultivation

Physiological characterization of S. cerevisiae strains was performed in anaerobic batch and chemostat cultures in 2-L bioreactors (Applikon, Delft, The Netherlands), with 1-L working volume. Salt solutions were sterilized by autoclaving at 120 °C for 20 min. Glucose solutions were autoclaved separately at 110 °C for 20 min and subsequently added to the sterile salt solutions. All cultures were grown on synthetic medium with vitamins [31], supplemented with 20 g L⁻¹ glucose and with sterile solutions of the anaerobic growth factors ergosterol (10 mg L⁻¹) and Tween 80 (420 mg L⁻¹), as well as with 0.2 g L⁻¹ sterile antifoam C (Sigma-Aldrich). Anaerobic conditions were maintained by sparging of a gas mixture of N₂/CO₂ (90/10%, < 10 ppm oxygen) at a rate of 0.5 L min⁻¹ and culture pH was maintained at 5 by automatic addition of 2 M KOH. All cultures were grown at a stirrer speed of 800 rpm and at a temperature of 30 °C. Oxygen diffusion in the bioreactors was minimized by equipping them with Norprene tubing and Viton O-rings, and evaporation was minimized by cooling of outlet gas to 4 °C.

To generate bioreactor inocula, two pre-culture shake flasks were grown in 500-mL flasks containing 100 mL synthetic medium (20 g L⁻¹ glucose). Initial pH was adjusted to 6 by addition of KOH. Cultures were grown, under atmospheric air, at 30 °C and shaken at 200 rpm. In each case, initial pre-culture flasks were inoculated from frozen S. cerevisiae stock cultures. After incubation for 8–12 h, cultures from these flasks were used to inoculate fresh pre-culture flasks for bioreactor inoculum propagation. In all cases, bioreactors were inoculated when pre-cultures reached mid-exponential phase (OD₆₆₀ 4–5), to a starting OD₆₆₀ of 0.15–0.25.

Analytical methods

Off-gas analysis, biomass dry weight measurements, HPLC analysis of culture supernatants and correction for ethanol evaporation in bioreactor experiments were performed as described previously [20]. Optical density was determined at 660 nm, using a Libra S11 spectrophotometer (Biochrom, Cambridge, United Kingdom). In batch cultures, yields of products were calculated from samples taken at mid-exponential phase (minimum of five samples), as described previously [39]. Biomass and product yields in chemostat cultures were determined from residual glucose, biomass and metabolite concentrations in steady-state cultures, analysed after rapid quenching of culture samples [40].

For calculation of degree of reduction (electron) balances in cultures, the degrees of reduction of biomass, CO₂, NH₄⁺ and extracellular metabolites (glucose, ethanol, glycerol, succinate, pyruvate, lactate and acetate) were defined as described in [41].

Estimations of statistical significance of differences in yields between strains were determined with two-tailed Student’s t tests. All values are represented as averages ± mean deviation of independent biological replicate cultures, performed at least in duplicate.

Enzyme-activity assays

For in vitro enzyme-activity assays of PRK [28], cells (65 mL culture volume) from exponentially growing (OD₆₆₀ 4), anaerobic shake-flask cultures (100 mL working volume in 500 mL conical shake-flasks) on glucose-containing (20 g L^-1) synthetic medium were harvested, and cell extracts were prepared as described previously [42]. The harvesting and sonication buffer contained 100 mM Tris–HCl, 20 mM MgCl₂·6H₂O and 5 mM 1,4-dithiothreitol (pH 8.2). The assay mixture [43] contained 50 mM Tris–HCl (pH 8.2), 40 mM KCl, 10 mM MgCl₂·6H₂O, 0.15 mM NADH, 1 mM ATP, 3 mM phosphoenolpyruvate, 1 mM 1,4-dithiothreitol, 5 U of pyruvate kinase (EC 2.7.1.40, Sigma-Aldrich), 6 U of l-lactate dehydrogenase (EC 1.1.1.27, Honeywell Fluka, Bucharest, Romania) and 20, 30, 40 or 50 μL cell extract in 1 mL total volume. Reactions were started by addition of d-ribulose-5-phosphate (2.5 mM final concentration), and PRK activity was measured at 30 °C using a Hitachi 100-60 spectrophotometer, by monitoring of NADH oxidation at 340 nm over time. Protein concentrations in cell extracts were quantified using the Lowry method [44].

Protein extraction and proteomics analysis

For proteomics analysis, cells (5 mL culture volume) were harvested from mid-exponential-phase (OD₆₆₀ 2), anaerobic shake-flask cultures on synthetic medium (20 g L⁻¹ glucose or 20 g L⁻¹ galactose), washed with ice-cold MilliQ H₂O, and subsequently stored at − 80 °C. Frozen cells were lysed using mechanical disruption in a Precellys-24 homogeniser (Bertin Technologies, Montigny-le-Bretonneux, France) in 0.5 mL cold methanol (− 20 °C, Sigma-Aldrich). The protein concentration of the disrupted cell suspension was measured using a Qubit 2.0 fluorometer (Thermo-Scientific). A total of 250 μg protein was taken from each methanol suspension, and 10 μg bovine-serum albumin was spiked to all samples for quality control. Proteins were extracted from the disrupted cell suspension using chloroform (Sigma-Aldrich) and 20% TCA (Sigma-Aldrich). The obtained protein pellet was dissolved in 100 mM NH₄HCO₃ buffer (pH 7) to a final concentration of 0.5 g L⁻¹. In each sample, 5 μL of 500 mM Tris-(2-carboxyethyl)phosphine hydrochloride solution (TCEP, Sigma-Aldrich) was added, and samples were incubated at 55 °C for 30 min to facilitate disulphide reduction. Alkylation was performed through the addition of 5 μL of 550 mM iodoacetamide and subsequent incubation at 25 °C in the dark for 30 min.

Proteolysis was carried out overnight at 37 °C with Trypsin Gold (Promega, WI, USA), which specifically cleaves C-terminally at lysine and arginine, at an enzyme to substrate ratio of 1:25. Gradient elution of peptides was performed on a C18 (Acquity UPLC CSH C18 Column, 130 Å, 1.7 µm, 2.1 mm × 100 mm, Ultimate 3000) (Thermo-Scientific). 20 μL of injected peptides were separated using a gradient ratio of mobile phase A (99.9% water and 0.1% formic acid; VWR) to 20% B (99.9% acetonitrile and 0.1% formic acid; VWR) for 20 min, and to 50% B for 30 min (60 min total duration).

Data acquisition was carried out using a data-dependent method using a Q Exactive Plus mass spectrometer (Thermo-Scientific). The top 15 precursors were selected for tandem-MS/MS (MS2) analysis after higher-energy collisional dissociation (HCD) fragmentation. Full MS scans covering a mass range of 400–1600 were acquired at a resolution of 70,000 (at m/z 200), with a maximum fill time of 75 ms, and an automatic gain control (AGC) target value of 3 × 10⁶. MS2 scans were acquired at a resolution of 17,500 (at m/z 200), with a maximum fill time of 75 ms, and an AGC target value of 10⁵. An isolation window of 2 m/z with a fixed first mass of 110 m/z was applied in all experiments. HCD fragmentation was induced with a normalized collision energy of 27 for all peptides. Charge-state exclusion was set to ignore unassigned 1 charge. Isotope exclusion was enabled and peptide match was preferred.

All LC–MS/MS results were searched against the S. cerevisiae protein database, to which the amino acid sequences of the heterologous enzymes (PRK, CbbM, GroEL, GroES) were manually added, in Proteome Discoverer 1.4 Sequest HT (Thermo-Scientific). The cleavage preference of trypsin was selected, allowing for up to two missed cleavages (C-Term K/R restrict P). Dynamic modifications were set to carbamidomethyl (C), deamidation (N/Q) and oxidation (M). Precursor mass tolerance was set to 10 ppm and fragment mass tolerance to 0.6 Da. Following peptide identification, their q values were calculated based on a target decoy approach with a 1% false discovery rate.

Spot plate assay

Spot plates on synthetic medium (pH 6) were prepared as described previously [45]. Sterile solutions of glucose (180 g L⁻¹) and of the anaerobic growth factors ergosterol (10 mg L⁻¹) and Tween 80 (420 mg L⁻¹) were additionally supplemented. All plates were inoculated with serial dilutions of exponentially growing shake-flask cultures in sterile demineralized water, prepared as described above. Plates were incubated under anaerobic conditions (10% CO₂) at 30 °C for 48 h.

Ploidy determination by flow cytometry

For determination of yeast ploidy, ca. 10⁷ cells were harvested from mid-exponential phase shake-flask cultures on synthetic medium (20 g L⁻¹ glucose), washed twice with demineralized water and stored in 70% ethanol at 4 °C. Sample preparation and staining was performed as described previously [46]. Samples were processed using a BD Accuri C6 flow-cytometer (BD Biosciences, San Jose, CA) and analysed using the FlowJo software package (Flowjo LLC, Ashland, OR).

Genome sequencing

DNA was isolated from yeast cells harvested from shake-flask cultures of strain IMX774 on synthetic medium (20 g L⁻¹ glucose) using a Qiagen Blood & Cell Culture DNA kit (Qiagen, Germantown, MD), following manufacturer’s specifications. Paired-end sequencing (22 mln reads) was performed on a 350-bp PCR-free insert library using an Illumina HiSeq PE150 sequencer (Novogene Company Limited, Hong Kong) with a sample size of 3.3 Gb, accounting for a total coverage of 275×. Sequence data was mapped to the CEN.PK113-7D genome [30], to which the sequences of the pDAN1-prk-tPGK1, pTDH3-cbbm-tCYC1, pTEF1-groEL-tACT1, and pTPI1-groES-tPGI1 cassettes were manually added. Data processing and chromosome copy number analysis were carried out as described previously [47–51].

Impact of PRK expression levels on in vivo CO₂ reduction via the RuBisCO pathway in glucose-grown batch cultures

In the engineered strain used for the first demonstration of the effect of expression of the Calvin-cycle enzymes RuBisCO and PRK on the anaerobic physiology of S. cerevisiae, the coding sequence of S. oleracea prk was placed under the control of the galactose-inducible GAL1 promoter [28]. Use of galactose as an inducer of gene expression in S. cerevisiae is, however, not a realistic option in large-scale industrial fermentations for ethanol production due to the price of galactose and repression of the GAL1 promoter by glucose [52]. Furthermore, this strain expressed the T. denitrificans RuBisCO gene cbbm, as well as the E. coli chaperones groEL/groES, from a centromeric plasmid. Expression from plasmids with auxotrophic markers limits applicability in industrial processes [53] and the use of a centromeric vector restricted the number of cbbm-cassettes per cell to 1–2 [54]. The low RuBisCO activity in cell extracts of strain IMU033 [4.6 ± 0.3 nmol (mg protein)⁻¹ min⁻¹] [28] suggested that introduction of additional copies of the cbbm cassette might be relevant for improved strain performance.

In vivo tandem assembly by homologous recombination and CRISPR-mediated targeted integration at a single locus was previously shown to be an effective way to introduce multiple copies of expression cassettes without the use of multi-copy plasmids [48, 55]. To construct a galactose-independent RuBisCO-expressing platform strain with an increased number of cbbm cassettes, nine copies of a cbbm overexpression cassette, along with single expression cassettes of groEL/groES, were first integrated at the SGA1 locus of IMX581 using CRISPR/Cas9 single-step transformation and assembly [34], yielding strain IMX765. Since high-level expression of heterologous PRK in microbes has been previously shown to be toxic [56, 57], two expression cassettes were constructed, in which the prk open reading frame was either placed under the control of pYEN1 (low-level constitutive expression under a wide range of cultivation conditions, [35]) or under the control of pDAN1 (medium-level expression induced under anaerobic conditions, [35]). These expression cassettes were integrated at the X-2 locus [33] of strain IMX765, yielding strains IMX773 and IMX774, respectively.

Enzyme-activity assays in cell extracts of anaerobic, glucose-grown shake-flask cultures of strains IMX773 and IMX774 showed PRK activities of 0.14 ± 0.01 and 0.68 ± 0.33 μmol (mg protein)⁻¹ min⁻¹, respectively. These activities were 100- and 20-fold lower than previously measured in cell extracts of strain IMU033 under galactose-induced conditions [14.4 ± 1.5 μmol (mg protein)⁻¹ min⁻¹] [28]. Analysis of protein abundance of RuBisCO and PRK in strains IMU033 (pGAL1-prk cbbm) and IMX774 revealed tenfold higher CbbM levels and ninefold lower PRK levels in the latter, newly engineered strain (Fig. 1).

Strain IMX773, in which prk was expressed from the weak constitutive YEN1 promoter, did not show significant differences in glycerol or ethanol yields relative to the reference strain IME324 (Table 3). This result confirms that functional expression of PRK is essential for the use of CO₂ as an electron acceptor for NADH oxidation by the engineered S. cerevisiae strains. In contrast, strain IMX774, which expressed prk from the anaerobically induced, medium-strength pDAN1 promoter, exhibited a 31% lower glycerol yield and a 10% higher ethanol yield than the reference strain (Table 3). Furthermore, the glycerol production per gram biomass of strain IMX774 was 38% lower than that of the reference strain (Table 3). These observations indicated that the engineered PRK/RuBisCO pathway significantly contributed to NADH oxidation in anaerobic cultures of this engineered strain.

The impact of the RuBisCO pathway on NADH oxidation is negatively correlated with the specific growth rate

At a dilution rate of 0.05 h⁻¹, strain IMX774 showed glycerol and ethanol yields on glucose of 0.005 and 0.451 g g⁻¹, respectively. These yields were 90% lower and 7% higher, respectively, than in chemostat cultures of the reference strain IME324 grown at the same dilution rate (Table 4; Fig. 2). In these slow-growing chemostat cultures, the glycerol production per gram biomass of strain IMX774 was only 0.66 mmol (g biomass)⁻¹, which was 90% lower than observed in cultures of strain IME324 (Fig. 2). These results indicate that, at this low specific growth rate, the RuBisCO pathway almost completely replaced reoxidation of ‘excess’ NADH via glycerol formation, in agreement with previous results on IMU033 glucose/galactose-grown chemostat cultures on the same dilution rate [28].

The reference strain IME324 showed no significant differences in glycerol yield on glucose or in glycerol production per gram biomass when grown at a dilution rate of either 0.05 or 0.15 h⁻¹ in anaerobic, glucose-limited chemostat cultures (Fig. 2). In contrast, strain IMX774 showed a fivefold higher glycerol yield on glucose and glycerol production per gram biomass when grown at a dilution rate of 0.15 h⁻¹ than in cultures grown at 0.05 h⁻¹ (Fig. 2). The glycerol production per gram biomass of strain IMX774 at 0.15 h⁻¹ [3.88 mmol (g biomass)⁻¹] was only 50% lower than that of strain IME324 grown at the same dilution rate (Fig. 2). These results demonstrated that, in strain IMX774, higher specific growth rates, which coincided with a higher glycolytic flux, resulted in a smaller contribution of the engineered PRK/RuBisCO pathway to NADH reoxidation, thereby reducing its beneficial impact on (by)product formation.

Deletion of GPD2 improves CO₂ reduction to ethanol in anaerobic batch cultures of RuBisCO/PRK-expressing S. cerevisiae

The lower impact of RuBisCO/PRK expression​ on product formation at high specific growth rates identified the glycerol-3-phosphate dehydrogenases Gpd1 and Gpd2 as potential engineering targets for increasing the contribution of the engineered PRK/RuBisCO pathway to NADH reoxidation. Deletion of GPD2 was previously reported to decrease glycerol formation in other engineered S. cerevisiae strains, without affecting osmotolerance [16–19, 22]. GPD2 was therefore deleted in strain IMX774 (GPD1 GPD2 pDAN1-prk cbbm), yielding strain IMX949 (GPD1 gpd2Δ pDAN1-prk cbbm). This deletion did not affect the specific growth rate in anaerobic bioreactor batch cultures grown on 20 g L⁻¹ glucose (Table 3). Since deletion of GPD2 has a strong negative effect on anaerobic growth of wild-type S. cerevisiae in the absence of an external electron acceptors [59, 60], this result further supported our conclusion that the RuBisCO pathway can effectively contribute to redox cofactor balancing in fast-growing anaerobic S. cerevisiae cultures. In these anaerobic bioreactor batch cultures, the glycerol and ethanol yields on glucose of strain IMX949 were 62% lower and 13% higher, respectively, than those of the reference strain IME324 (GPD1 GPD2) (Table 3). Furthermore, glycerol production per gram biomass of strain IMX949 was 65 and 43% lower than that of strains IME324 and IMX774, respectively (Table 3). These results clearly indicated that deletion of GPD2 enables a higher contribution of the engineered PRK/RuBisCO pathway to anaerobic NADH reoxidation in engineered S. cerevisiae strains.

Optimization of precursor supply to the RuBisCO pathway further decreases glycerol yield and enables wild-type specific growth rates in anaerobic cultures

In S. cerevisiae, the substrate of PRK, ribulose-5-phosphate, can be formed either by NADPH-generating oxidative decarboxylation of 6-phosphogluconate, or from glyceraldehyde-3-phosphate and fructose-6-phosphate via the re-arrangement reactions of the non-oxidative pentose-phosphate pathway (PPP, [28]). If ribulose-5-phosphate used in the RuBisCO pathway were exclusively derived from 6-phosphogluconate, this would cause an NADPH/NADP⁺ imbalance when the RuBisCO pathway completely replaces glycerol formation [10]. While formation of ribulose-5-phosphate via the non-oxidative PPP does not present such a redox constraint, extensive research on metabolic engineering of S. cerevisiae for pentose fermentation indicates that this pathway has a limited capacity in wild-type strains [61–64].

To test if the fermentation performance of strain IMX949 (GPD1 gpd2Δ pDAN1-prk cbbm) could be further improved by optimization of the ribulose-5-phosphate supply, overexpression cassettes for the non-oxidative PPP genes RPE1, TKL1, TAL1, NQM1, RKI1 and TKL2 were simultaneously integrated at the GPD2 locus of IMX774, yielding strain IMX1443 (GPD1 gpd2Δ non-ox PPP↑ pDAN1-prk cbbm). In anaerobic, glucose-grown bioreactor batch cultures, grown under identical conditions to the previously discussed RuBisCO/PRK-expressing strains, the specific growth rate of strain IMX1443 was virtually identical to that of the reference strain IME324 (GPD1 GPD2) and 36% higher than that of its parental strain IMX774 (GPD1 GPD2 pDAN1-prk cbbm) and of strain IMX949 (Table 3; Additional file 2). Furthermore, strain IMX1443 showed a 9% higher biomass yield on glucose than the reference strain IME324 (Table 3), which closely corresponds to the maximum theoretical increase for a RuBisCO/PRK-expressing strain of 13.5% [28]. These observations showed that the reduced growth rates of strains IMX774 and IMX949 were not primarily caused by accumulation of ribulose-1,5-bisphosphate or ATP depletion, resulting from an imbalance of the in vivo activities of PRK and RuBisCO. Instead, they indicate that the reduced growth rates of these strains resulted from a reduced intracellular pool of ribulose-5-phosphate, which is a key precursor for the formation of the PPP-derived biosynthetic building blocks ribose-5-phosphate and erythrose-4-phosphate.

The glycerol yield on glucose of strain IMX1443 (GPD1 gpd2Δ non-ox PPP↑ pDAN1-prk cbbm) was 81 and 87% lower than the yields of its parental strain IMX774 (GPD1 GPD2 pDAN1-prk cbbm) and of the reference strain IME324 (GPD1 GPD2), respectively (Table 3). Consistent with an almost complete replacement of redox cofactor balancing via glycerol production by the RuBisCO pathway, its glycerol production per gram biomass was 88% lower than that of strain IME324 and closely matched the phenotype observed in slow-growing glucose-limited chemostat cultures of strain IMX774 (Table 4). Furthermore, the ethanol yield on glucose of strain IMX1443 was 15 and 5% higher than that of the reference strain IME324 and of its parental strain IMX774, respectively. The phenotype of strain IMX1443 thereby approaches the theoretical maximum benefits in glycerol reduction and increased ethanol yield, without a reduction of its specific growth rate in anaerobic, glucose-grown batch cultures. Further, the osmotolerance of strain IMX1443 was not impacted by these modifications, as shown by plate growth tests on high osmolarity (1 M glucose) medium (Additional file 3).

The physiological benefit of RuBisCO/PRK-expression in S. cerevisiae is independent of strain ploidy

In the context of another study, the ploidy of strains IMX765, IMX773, and IMX774 was analysed by flow cytometry. Surprisingly, strain IMX765, the parental strain of all RuBisCO/PRK-expressing strains constructed in this study, was found to have undergone a whole-genome duplication (Additional file 4). To determine whether this diploidization was accompanied by any other chromosomal copy number variations or rearrangements, the genome of strain IMX774 (GPD1 GPD2 pDAN1-prk cbbm) was sequenced and compared to that of the haploid congenic reference strain CENPK.113-7D [30]. This analysis showed that a ‘clean’ genome duplication had occurred, without chromosomal or segmental aneuploidies (Additional file 5).

The differences in glycerol and ethanol yields between strains IME324 (haploid, GPD1 GPD2 reference) and IMX774 (diploid, GPD1 GPD2 pDAN1-prk cbbm) were not expected to be influenced by ploidy variation, as biomass formation and requirements for NADH oxidation are stoichiometrically linked and the biomass yields on glucose of these strains were not significantly different (Table 3). However, as ploidy variation might affect the growth rate [65], two additional strains were constructed to investigate whether ploidy differences affected the interpretation of our results.

Strain IME369 was constructed by transformation of p426-TEF (empty) to IMX673, thereby generating a new diploid reference strain (GPD1/GPD1 GPD2/GPD2). In addition, the genetic modifications introduced in the best performing RuBisCO/PRK-expressing strain (IMX1443) were reconstructed in a haploid background, resulting in strain IMX1489 (Additional file 4). Anaerobic growth of both strains was analysed in bioreactor batch cultures, under the same conditions used for the other strains analysed in this study (Table 5; Additional file 2). The new diploid reference strain IME369 showed no significant differences in specific growth rate, biomass or ethanol yields on glucose or glycerol production per gram biomass when compared to the haploid reference strain IME324 (GPD1 GPD2) (Tables 3 and 5). Furthermore, the specific growth rates of the engineered strains IMX1489 (haploid) and IMX1443 (diploid) were the same, while their biomass and product yields also very closely corresponded (Tables 3 and 5, Fig. 3). Similarly to strain IMX1443, the osmotolerance of IMX1489 did not differ from that of a GPD1 GPD2 reference strain (Additional file 3). These results indicate that the impact of the engineering strategy presented in this study does not differ between haploid and diploid S. cerevisiae strains.

Strain	IME369	IMX1489
Relevant genotype	GPD1 GPD2	GPD1 gpd2Δ pDAN1-prk cbbm non-ox PPP↑
μ (h−1)	0.31 ± 0.00	0.30 ± 0.01
Y biomass/glucose (g g−1)	0.091 ± 0.009	0.096 ± 0.001*
Y ethanol/glucose (g g−1)	0.376 ± 0.005	0.421 ± 0.002*
Y glycerol/glucose (g g−1)	0.107 ± 0.004	0.014 ± 0.000**
Glycerol produced/biomass [mmol (g biomass)−1]	12.189 ± 1.080	1.669 ± 0.082**

Cultures were grown on synthetic medium containing 20 g L⁻¹ glucose (pH 5). Specific growth rates and stoichiometries were calculated from sample points during the mid-exponential growth phase. Values represent averages ± mean deviations of measurements on independent duplicate cultures. * (p < 0.02) and ** (p < 0.01) denote statistical significance of differences between IME324 (Table 3) and strains IME369 and IMX1489 in Student’s t tests. Degree of reduction balances constructed over the exponential growth phase yielded electron recoveries between 96 and 100%