Yeast strain, media and transformation

The yeast strains used in this study were constructed from P. pastoris KM71 (Invitrogen) (Table 1; Additional file 1: Table S1). The plasmids used in this study were generated from the pGAPZ_A and pPIC3.5K vectors (Invitrogen). Detailed description of plasmid construction is provided in Additional file 1. Yeast transformation was performed as previously described using an electroporator [26]. The following parameters were used for the electroporation: 1.5 kV, 25 μF, 200 Ω. Colony PCR and DNA sequencing were used to verify strain construction. Yeast and bacterial strains were stored in 25% glycerol at − 80 °C. E. coli was grown in Luria–Bertani medium supplemented with ampicillin (at 100 μg/L), hygromycin (at 100 μg/L), kanamycin (at 50 μg/L), or zeocin (at 100 μg/L) when required. P. pastoris was grown in YPD medium (10 g/L yeast extract, 20 g/L Bacto Peptone and 20 g/L glucose) supplemented with hygromycin (at 200 μg/L), G418 (at 100 μg/L), or zeocin (at 100 μg/L) when required. Selection of P. pastoris transformants with HIS4 was done on a yeast minimal medium [MGY (pH 6.0) containing: 13.6 g/L yeast nitrogen base without amino acids, 20 g/L glycerol, 0.1 M phosphate buffer, 0.4 mg/L d-biotin, 133.3 mg/L thiamine-hydrochloride].

Screening of KDC and ADH candidate enzymes

Characterization of engineered P. pastoris strains overexpressing various KDC and ADH enzymes was carried out in a minimal medium [MGYH (pH 6.0) containing: 13.6 g/L yeast nitrogen base without amino acids, 20 g/L glycerol, 0.1 M phosphate buffer, 0.4 mg/L d-biotin, 133.3 mg/L thiaminchloride–hydrochloride, and 20 mg/L l-histidine]. Briefly, engineered strains were pre-cultured in 5-mL aliquots in MGYH medium overnight and used to inoculate 10 mL MGYH medium (either with or without 4 g/L 2-ketoisovalerate) in 50 mL Corning tubes to achieve an initial optical density of 0.05 at 600 nm (OD₆₀₀). The cultures were grown at 30 °C and 250 rpm in an orbital shaking incubator. Samples were taken at 24 and 48 h to determine OD₆₀₀, extracellular metabolites and production of higher alcohols.

The amount of isobutanol and other extracellular metabolites were determined using high-performance liquid chromatography (HPLC). Briefly, 1 mL of culture was centrifuged at 18,000g for 5 min and the supernatant was filtered through 0.45 μm nylon syringe filter (Filtrex). The purified sample was then applied to an Agilent 1100 series HPLC equipped with an Aminex HPX-87H ion exchange column (Biorad). The LC program was performed using 5 mM H₂SO₄ as the solvent at a flow rate of 0.72 mL/min for 42 min. The column was maintained at 55 °C. All metabolites were detected with Agilent 1200 series DAD and RID detectors.

Quantification of isobutanol production in engineered strains

Characterization of engineered P. pastoris strains was carried out in a minimal medium as described in “Screening of KDC and ADH candidate enzymes” section. l-Histidine was omitted from the medium when HIS4 was used as the selection marker; this medium is designated MGY. In the study comparing the different carbon sources, glycerol was replaced with glucose, resulting in a modified MGYH with glucose as a sole carbon source (designated MGYH-glu). In the study comparing different concentrations of glucose, the media were prepared with either 2 or 10% glucose and were designated MGYH-glu (2% glucose) and MGYH-glu (10% glucose), respectively.

Shake flask fermentation was also carried out in selected strains (KM71, PP100, PP200, PP300, PP310, PP302, PP303 and PP304). For this, engineered strains were pre-cultured in 5-mL aliquots in MGYH medium overnight and used to inoculate 50 mL MGYH-glu medium (at either 2 or 10% glucose as indicated) in 250 mL Erlenmeyer flasks closed with cotton wool plugs to achieve an initial optical density of 0.05 at 600 nm (OD₆₀₀). The cultures were grown at 30 °C and 250 rpm in an orbital shaking incubator. Samples were taken to determine OD₆₀₀, biomass, extracellular metabolites and production of higher alcohols. The amount of isobutanol and other extracellular metabolites were determined using HPLC as described above.

Quantification of isobutyl acetate production in engineered strains

Quantification of isobutyl acetate ester in engineered strains was performed as previously described with some modifications [27]. Engineered strains were pre-cultured in 5-mL aliquots in MGYH overnight and used to inoculate 10 mL modified MGY-glu medium in 50 mL Corning tubes as described in “Screening of KDC and ADH candidate enzymes” section. The yeast cultures were overlaid with 10 mL hexadecane (Sigma) to reduce evaporation of the acetate esters. The cultures were grown at 30 °C and 250 rpm in an orbital shaking incubator. Samples were taken to determine biomass, extracellular metabolites and production of higher alcohols and acetate esters. The amount of acetate esters dissolved in the hexadecane layer was determined using gas chromatography–mass spectrometry (GC–MS). The GC program was as follows: an initial temperature of 40 °C was maintained for 4 min, followed by ramping to 300 °C at a rate of 45 °C per min, where the temperature was held for 3 min. The injector temperature was held at 250 °C. The injection volume was 5 μL, injected at a 10:1 split ratio. Hydrogen was used as the carrier gas. The MS is a GCMS-QP2010S (Shimadzu). The ion source temperature was 200 °C, and the interface temperature was 250 °C. The solvent cut time was 2 min. The start m/z was 40, and the end m/z was 500. Mass spectra and retention times from samples were compared with authentic standards (Sigma).

RNA isolation and transcript quantification

Strains were pre-cultured in 5-mL aliquots in MGYH medium overnight and used to inoculate 10 mL modified MGYH-glu medium as described in “Screening of KDC and ADH candidate enzymes” section. After 24 and 48 h, a 5-mL aliquot of each culture was collected and centrifuged for 5 min at 3000g. The pellets were washed with 5 mL of distilled water. Total RNA was extracted using the QIAgen RNeasy Kit under the manufacturer’s protocol. Contaminating genomic DNA was removed from the RNA samples by DNaseI (NEB) digestion using the manufacturer’s protocol. The RNA quantity was analyzed using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies), and samples were stored at − 80 °C until RT-PCR analysis. cDNA was obtained using RevertAid Reverse Transcriptase (Thermo Fisher Scientific) using the manufacturer’s protocol. Relative expression levels of PpIlv6_CodOpt, PpIlv2_CodOpt, PpIlv5, PpIlv3, LlkivD_CodOpt, ScADH7 and ScATF1 were quantified using iQ SYBR Green Kit (Biorad) on CFX96 Touch Real-time PCR Detection System (Biorad). Real-time PCR was performed in triplicates, and PpACT1, a gene that encodes actin, was used to normalize the amount of the total mRNA in all samples.

Total DNA isolation and gene copy quantification

Strains were pre-cultured in 5-mL aliquots in MGY medium overnight in 50 mL Corning tubes as described in “Screening of KDC and ADH candidate enzymes” section. Total DNA was isolated using Wizard Genomic DNA purification kit (Promega) using the manufacturer’s protocol. The DNA quantity was assessed using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies), and samples were stored at − 20 °C until real-time PCR analysis. Gene copy numbers of PpIlv6_CodOpt, PpIlv2_CodOpt, PpIlv5, PpIlv3, LlkivD_CodOpt, ScADH7 and ScATF1 were quantified using iQ SYBR Green Kit (Biorad) on CFX96 Touch Real-time PCR Detection System (Biorad) as previously reported with some modifications (Marx 2009). Real-time PCR was performed in triplicate, and normalization of the data was achieved using actin as a reference (i.e. PpACT1 gene copy number = 1). Primers used for RT-PCR are listed in Additional file 1.

Overexpression of the 2-keto acid degradation pathway genes to enhance alcohol production from 2-keto acids

To metabolically engineer P. pastoris to produce higher alcohols, we employed a strategy that had been used successfully in E. coli and S. cerevisiae [8, 12–14]. Specifically, we exploited P. pastoris’ endogenous amino acid biosynthetic pathway and channeled the amino acid intermediates to the 2-keto acid degradation pathway for higher alcohol production (Fig. 1). In the 2-keto acid degradation pathway (also known as the Ehrlich pathway), 2-keto acids are converted into higher alcohols in two enzymatic steps: decarboxylation of 2-keto acids to aldehydes by 2-keto acid decarboxylase (KDC), and subsequent reduction of aldehydes to alcohols by alcohol dehydrogenase (ADH).

We first searched for suitable KDC and ADH enzymes to enhance the endogenous activity of the Ehrlich pathway and to improve the higher alcohol production. Specifically, we screened three different candidates of KDC that have demonstrated high activity in either E. coli or S. cerevisiae. These are: KIVD from Lactococcus lactis, and ARO10 and THI3 from S. cerevisiae [8]. For ADH, we screened two enzymes—both from S. cerevisiae—that have different substrate specificities and catalytic properties. These are: ADH6 and ADH7 [12]. To assess the in vivo activities of these KDC and ADH candidates in P. pastoris, we supplemented the culture medium with 2-ketoisovalerate and quantified the production of isobutanol.

Gratifyingly, we observed isobutanol production when both KDC and ADH enzymes were expressed in P. pastoris and when the culture medium was supplemented with 2-ketoisovalerate (KIV, 4 g/L) (Fig. 2). The highest specific isobutanol titer—at 48 ± 1 mg/L/OD₆₀₀—was observed when L. lactis KIVD and S. cerevisiae ADH7 were overexpressed. The specific titer decreased to approximately 40 ± 2 mg/L/OD₆₀₀ when S. cerevisiae ARO10 and ADH6 were overexpressed. We were not able to detect any isobutanol in the control strain where none of the KDC and ADH enzymes was overexpressed, nor when KIV was absent from the medium (Fig. 2).

Overexpression of l-valine biosynthetic pathway genes to further increase the pool of 2-ketoisovalerate and subsequently the production titer of isobutanol

Having determined L. lactis KIVD and S. cerevisiae ADH7 to be the most promising KDC and ADH enzyme candidates, respectively, we next turned our attention to increasing the production of the key intermediate KIV. Overproducing KIV not only obviates the need to supplement the media with expensive 2-keto acid, but has also been shown to markedly decrease the production of the other alcohols [8]. We increased the pool of KIV by upregulating a portion of l-valine biosynthesis (Fig. 3). In this pathway, acetolactate synthase condenses two molecules of pyruvate to 2-acetolactate, which is then reduced to 2,3-dihydroxyisovalerate by acetohydroxyacid reductoisomerase. Finally, 2,3-dihydroxyisovalerate is converted to KIV by dihydroxyacid dehydratase. In P. pastoris, acetolactate synthase, acetohydroxyacid reductoisomerase and dihydroxyacid dehydratase are encoded by four different genes, PpIlv2, PpIlv6 (small subunit of acetolactate synthase), PpIlv5 and PpIlv3, whereas in S. cerevisiae, the enzymes are encoded by just three genes, ScIlv2, ScIlv5 and ScIlv3. Given the smaller number of genes required for overexpression in the case of S. cerevisiae vs. P. pastoris and the demonstrated activity of the enzymes in previous reports, we first attempted to overexpress the S. cerevisiae variants in the selected engineered P. pastoris strain with the enhanced 2-keto acid degradation pathway.

Given the comparatively limited availability of genetic tools for heterologous expression of multiple genes in P. pastoris, we decided to link the individual genes in the keto-acid degradation and l-valine biosynthetic pathways together with the other members of the same pathway to create two separate modules (Fig. 4a). We chose to link the individual genes with the self-cleaving 2A peptide sequence from the foot-and-mouth disease virus (FMDV) to ensure stoichiometric production of all the proteins whose genes are linked together by the 2A peptide sequence [28]. A useful strategy expressing more than one protein, the 2A peptides allow multiple proteins to be encoded as polyproteins, which dissociate into individual proteins upon translation. Fermentation studies of the PP101 strain containing the three l-valine biosynthetic pathway genes from S. cerevisiae and the two 2-keto acid degradation pathway genes were performed in culture medium either with or without supplementation of KIV. Unfortunately, isobutanol production was observed only when KIV was supplemented to the culture media (Additional file 1: Figure S1). Negligible amounts of isobutanol were observed when the intermediate was absent from the media. Our results suggested that the l-valine biosynthetic pathway genes from S. cerevisiae could not be functionally expressed in P. pastoris.

To circumvent this problem, we overexpressed P. pastoris’ endogenous l-valine biosynthetic pathway genes in the selected engineered P. pastoris strain with the enhanced 2-keto acid degradation pathway (Fig. 4b). The two genes, PpIlv6 and PpIlv2 (each encodes a different subunit of the acetolactate synthase), were linked together by self-cleaving 2A peptide sequence and placed behind the GAP promoter in the integrative expression plasmid, pGAP-Neo. The remaining two genes, PpIlv5 and PpIlv3, were also linked together by self-cleaving 2A peptide sequence and placed behind the GAP promoter in the integrative expression plasmid, pGAP-Hyg. We observed isobutanol production in the engineered strain overexpressing the endogenous l-valine biosynthetic pathway genes. Specifically, PP300, the strain overexpressing all six genes (LlkivD, ScADH7, PpIlv6, PpIlv2, PpIlv5 and PpIlv3) produced approximately 11.8 ± 0.8 mg/L/OD₆₀₀ after 2 days in minimal yeast media with 2% glycerol, a 100-fold improvement over the levels observed in the strains overexpressing only the two keto acid degradation pathway genes, LlkivD and ScADH7 (PP100; 0.12 ± 0.04 mg/L/OD₆₀₀) (Fig. 5a, c). Low levels of isobutanol (1.6 ± 0.2 mg/L/OD₆₀₀) were also observed in the strain PP200, which overexpresses both keto acid degradation pathway genes but only two of the l-valine biosynthetic pathway genes, PpIlv5 and PpIlv3. Our results suggested that expression of all four genes from the upper portion of l-valine pathway is required to observe high levels of isobutanol production.

In order to verify that the increase in isobutanol production correlated with increased expression levels of the l-valine pathway genes and the keto-acid degradation pathway genes, we performed RT-PCR in the three engineered strains, PP100, PP200 and PP300 (Fig. 6). The expression levels of all four genes (PpIlv5, PpIlv3, PpIlv6 and PpIlv2) were higher (3- to 16-folds) in PP300 compared to the levels observed in PP100. Interestingly, the expression levels of PpIlv5 and PpIlv3 were lower in PP300 compared to PP200 (Fig. 6a, b), indicating that usage of the same promoter (i.e. P_GAP in our case) to drive the expression of multiple genes led to lower expression levels of the individual genes.

Analysis of the effects of carbon sources on isobutanol production

We next studied the effects of the concentration and type of carbon source on isobutanol production in our engineered yeast strains. When we switched the carbon source from 2% glycerol to 2% glucose (i.e. from MGYH to MGHY_glu), we observed an approximately fourfold improvement (from 11.8 ± 0.8 to 43.0 ± 5.3 mg/L/OD₆₀₀) in isobutanol production (Fig. 5b, d). The elevated production level corresponds to the upregulation of all isobutanol biosynthetic pathway genes in the medium with glucose as the sole carbon source (Additional file 1: Figure S2). Interestingly, we observed that isobutanol production leveled off after 1 day in this medium (Fig. 5b, d), potentially due to glucose limitation and limited oxygen availability. To increase the oxygen availability, we performed our fermentation studies in 250-mL Erlenmeyer flasks instead of 50-mL conical tubes. When we increased the glucose concentration in the MGYH_glu medium from 2 to 10% as well as extending the shake flask fermentation time to 3 days, we observed a 3.1-fold improvement (from 282 ± 7 to 885 ± 12 mg/L) in total isobutanol production (Fig. 7). The elevated production level corresponds to the upregulation of all isobutanol biosynthetic pathway genes in the medium with the increased glucose concentration (Additional file 1: Figure S3). Accumulatively, our results indicated that the engineered P. pastoris strains are able to produce isobutanol in minimal yeast medium containing glucose as the sole carbon source.

Improvement of isobutanol production by fine-tuning the expression of isobutanol biosynthetic pathway genes

To improve the isobutanol production titer in PP300, we turned our attention to optimizing the expression levels of the biosynthetic pathway genes. First, we integrated an additional copy of the PpIlv6-T2A-PpIlv2 gene construct to create strain PP310 and observed an isobutanol production titer of 1.70 ± 0.07 g/L, a 92% improvement from the level observed in PP300 (at 885 ± 12 mg/L) (Fig. 8). Strain PP310 also appears to be more efficient at consuming the substrate glucose than strain PP300, with no glucose remaining in the medium after 72 h for PP310 compared to 44 ± 1 g/L (44% of the starting concentration) for PP300 (Figs. 7, 8). These results indicate that overexpression of the first enzymatic step of the l-valine pathway, acetolactate synthesis, resulted in the necessary “pull” of the cell’s metabolism towards isobutanol production. Moreover, we hypothesize that acetolactate synthesis is a potential bottleneck in isobutanol production and warrants further optimization (Fig. 3). To this end, we examined the possibility of using an episomal plasmid-based expression system. Industrial applications of P. pastoris as a recombinant protein production host predominantly rely on the integration of foreign expression cassette(s) at a specific site within the host’s genome. In some cases, multiple integration events can occur, which can be selected for using a higher dose of antibiotics. While chromosomal integration has the advantage over plasmid-based expression in terms of genomic stability, the latter offers a broader range of foreign gene expression levels. Indeed, having a suite of plasmids with a choice of low, medium or high copy-number has enabled scientists to fine-tune the expression levels of the individual biosynthetic enzymes that resulted in large improvement in production yields of a variety of chemicals in engineered E. coli and S. cerevisiae [29, 30].

To broaden the genetic toolbox for engineering the yeast P. pastoris, we created a set of episomal expression vectors by utilizing the recently discovered autonomously replicating sequence (panARS) from Kluyveromyces lactis (Additional file 1: Figure S4) [31, 32]. Camattari et al. demonstrated that vectors containing this 452-nt element outperformed the traditional integrative plasmids in heterologous enzyme expression, both in productivity as well as clonal homogeneity [31]. We expanded this previous work by introducing the panARS sequence into four different expression plasmids: pGAPz (for zeocin selection), pGAPhyg (for hygromycin selection), pGAPneo (for G418 selection) and pGAPHis4 (for histidine auxotrophy). All of these plasmids use the strong constitutive promoter from the glyceraldehyde-3-phosphate dehydrogenase gene (P_GAP) to express recombinant proteins.

Using PP300 as the parental strain, we introduced additional copies of the PpIlv5_T2A_PpIlv3, PpIlv6_T2A_PpIlv2, and LlkivD_T2A_ScADH7 expression cassettes by placing them on the episomal expression vector pGAPHisKlARS and transforming the individual plasmids into PP300. The resulting strains, PP302, PP303 and PP304, showed superior performance to the parental strain. In particular, PP304, which harbors the episomal plasmid containing the PpIlv6_T2A_PpIlv2 expression cassette, produced the highest titer of isobutanol at 2.22 ± 0.05 g/L (isobutanol yield of 22.2 ± 0.1 mg/g glucose, specific isobutanol production titer of 153 ± 9 mg/L/OD₆₀₀). To the best of our knowledge, this is the highest reported isobutanol production titer and yield in a yeast system.

As was also the case for strain PP310, all strains harboring the episomal plasmids that contain additional copies of isobutanol biosynthetic pathway genes exhibited improved glucose consumption compared to strain PP300 (Figs. 7b, 8). Additionally, we observed elevated levels of the competing side-product ethanol (more than 10 g/L) in strains PP302, PP303 and PP304 when compared to the control strain KM71 (~ 3 g/L). The isobutanol pathway produces excess NADH (from glycolysis) and requires additional NADPH (from the reduction of isobutanal to isobutanol). Large accumulation of ethanol indicates that isobutanol production potentially caused redox cofactor imbalances, which ultimately forced the cells to use up the excess NADH by producing ethanol. Similar results have been observed in other microbial systems, particularly the Crabtree-positive S. cerevisiae. For example, Matsuda and coworkers have engineered S. cerevisiae to produce 1.6 g/L of isobutanol (16 mg/g glucose) by re-localizing the upper portion of valine metabolism to the cytosol and overexpressing the keto acid degradation pathway [11]. The authors observed that the strain also accumulated high levels of ethanol, at approximately 40 g/L, 25-folds the level of the desired product isobutanol. Several strategies have led to lower ethanol accumulation and improvement in isobutanol production in S. cerevisiae. These include: (1) downregulating the pyruvate decarboxylase enzyme [12]; (2) overexpression of a transhydrogenase or constructing a transhydrogenase-like shunt to convert excess NADH to NADPH [11]; and (3) replacing the alcohol dehydrogenase with an NADH-dependent engineered mutant [33]. These strategies could potentially improve isobutanol production in P. pastoris.

Overexpression of a broad-substrate-range alcohol O-acyltransferase to produce acetate esters of isobutanol and other higher branched-chain alcohols

To demonstrate the versatility of our yeast platform, we further engineered P. pastoris to convert isobutanol to the volatile ester isobutyl acetate. Many volatile esters exist in nature and their biosynthesis typically involves the condensation of an acyl-CoA with an alcohol. This reaction is carried out by the enzyme alcohol O-acyltransferases (ATFs) [34]. Along with the wax ester synthases/acyl-CoA:diacylglycerol acyltransferase (WS/DGAT), ATFs form a class of enzymes that use acyl-CoA units as the acid component for ester formation. The Baker’s yeast S. cerevisiae, which produces several volatile esters during beer and wine fermentation, contains four alcohol O-acyltransferase genes, ATF1, ATF2, EHT1 and EEB1 [35]. Previous work by Rodriguez and coworkers found that heterologous expression of ATF1, which has the broadest substrate scope of the four, in an engineered E. coli strain led to production of acetate ester of several alcohols including isobutanol and isopentanol [27]. Encouraged by these results, we constitutively expressed S. cerevisiae ATF1 in PP300. The resulting strain (PP400) produced isobutyl acetate at a titer of 6.0 ± 0.6 mg/L, along with smaller quantities of several other acetate esters such as isoamyl acetate (Fig. 9a, b). The low titers of isobutyl acetate ester, as well as the large accumulation of isobutanol in the production culture, indicated that the conversion of isobutanol to isobutyl acetate is a potential bottleneck, indicating the need for further strain optimization. This prompted us to overexpress ScATF1 on an episomal plasmid instead of integrating the expression cassette into the genome. The resulting strain, PP401, produced isobutyl acetate at 51.2 ± 6.8 mg/L as well as isoamyl acetate ester at 23.5 ± 3.0 mg/L (Fig. 9a, b). The episomal plasmid was stably maintained during the fermentation as confirmed by RT-PCR quantification of gene copy number (Additional file 1: Figure S6). The elevated production levels of the acetate esters correspond to the increased expression of ScATF1 and the higher gene copy number as indicated by real-time PCR (Additional file 1: Figure S5).