Yeast strains and plasmids used in this study are shown in Table 1. A derivative of the PO1f strain of Y. lipolytica (MatA, leu2-270, ura3-302, xpr2-322, axp-2; ATCC no. MYA-2613) (Madzak et al., 2000) that we previously described was used as the base strain for all genome editing and expression experiments in this study (Schwartz et al., 2017b). DH5α E. coli was used for plasmid construction and propagation, with growth in Lysogeny Broth (LB) supplemented with 100 mg/L ampicillin. Yeast cultures were grown in YPD medium (1% Bacto yeast extract, 2% Bacto peptone, 2% glucose), while cultures for lycopene biosynthesis were grown in YPD10 medium (1% Bacto yeast extract, 2% Bacto peptone, 10% glucose). For strain development, solid synthetic defined media without leucine and uracil [SD -Leu -Ura; 0.67% Difco yeast nitrogen base without amino acids, 0.067% CSM-Leu-Ura (Sunrise Science, San Diego, CA, United States), 2% glucose, 2% agar] was used to plate and screen transformants.

Native genes in the mevalonate pathway were obtained via PCR amplification from purified genomic DNA of the PO1f strain. Genomic DNA was isolated using the Zymo Research YeaStar genomic DNA kit. Genes were amplified using Phusion DNA Polymerase with primers listed in Supplementary Table S1. Amplified genes were cloned into pUAS1B8-TEF(136)-hrGFP (Blazeck et al., 2011) digested with BssHII and NheI by Gibson Assembly yielding plasmids as shown in Supplementary Figure S1. As such, all overexpressed genes had the same promoter, the UAS1B8-TEF(136) (Supplementary Table S2). Plasmid construction for markerless integration was done as previously described, with all genes again using the UAS1B8-TEF(136) promoter (Schwartz et al., 2017b). All cloning enzymes were purchased from New England Biolabs.

Overexpression of genes in the mevalonate and lycopene pathways were carried out with linear DNA transformations using the selectable leucine marker (LEU2). Plasmids containing LEU2 and a gene of interest, derived from pIW209, were linearized with the restriction enzyme KpnI (Supplementary Figure S1). The linearized DNA was column purified using Zymo Research Clean & Concentrator. The DNA was then transformed into the yeast strain HEBI-L (Table 1) and screened by plating the transformants on minimal media agar plates (SD -Leu -Ura). Uracil auxotrophy was alleviated using an analogous methodology, with plasmids derived from pIW245. Transformation of linear DNA resulted in integration of the fragment into a random location in the genome. Transformations were conducted at stationary phase and followed a previously described procedure (Schwartz et al., 2017b). After 2 days of outgrowth at 30°C, healthy colonies were selected and used for lycopene characterization. Markerless integration enabled by CRISPR-Cas9 and homologous repair donors was used for genes identified as limiting in the mevalonate and lycopene pathways to facilitate the overexpression of more than 2 genes (Schwartz et al., 2017b). Gene disruptions were done in the HEBI-L-U strain using CRISPR-Cas9 with repair by nonhomologous end-joining, and auxotrophies alleviated after.

Cultivation of carotenoid producing yeast strains was conducted in 250 mL baffled Erlenmeyer shake flask. Strains were initially grown in YPD medium and cultured overnight at 30°C. Flasks containing 25 mL of YPD10 were then inoculated to an OD600 of 0.1 and grown in a 30°C shaker at 200 RPM. Aliquots of the shake flask cultures were taken on the 4th day and used for DCW and lycopene quantification. Extended cultures (12-day cultures) were carried out in a similar manner, with DCW not taken until the last day of the trial and glucose added via pipetting at 2 days intervals by addition of 2.5 mL of 40% glucose.

Bioreactor experiments were performed in a 1-L batch reactor (Biostat A, Sartorius, Supplementary Figure S2). The bioreactor enabled control of dissolved oxygen (25%, maintained by constant aeration with oxygen and air, and variable stir rates), temperature (30°C, maintained through a heat jacket), and pH (6.8, maintained through automated addition of 3 M NaOH). Starter cultures were grown overnight in YPD and used to inoculate 500 mL of media to an OD600 of 0.02. At 24 h intervals, 25 mL of 40% glucose was added. At each 24 h interval, 10 mL was removed and used to quantify the specific lycopene content of the culture.

Extraction of carotenoids followed the method provided by Chen and coworkers with a few modifications (Chen et al., 2016; Schwartz et al., 2017b). Briefly, a 5 mL sample was taken and used for dry cell weight (DCW), and a 1 mL sample was used for extraction of lycopene by centrifuging at 5,000 g for 3 min, washing the cell pellet with water, resuspending in 1 mL 3M HCl, and incubating at 100°C for 2 min. Cells were then cooled in an ice bath for 3 min, washed with water, and resuspended in 1 mL of acetone. Two hundred-μL of 500–750 μm glass beads (Fisher Scientific) were then added, and the mixture was vortexed for 2 min. Subsequently, the mixture was centrifuged and the supernatant was subjected to analysis for lycopene. Lycopene was quantified by measuring absorbance at 472 nm and compared to a standard curve of purchased lycopene (Sigma–Aldrich, Supplementary Figure S3).

Experiments were performed in triplicate, and the mean and standard deviation are reported. Comparisons between two means were done using a two-tailed T-test (Figures 3–5), while comparisons among three or more means was done using 1-way ANOVA with post hoc Fisher’s analysis (Figures 2, 6B, 7B, 8). A p-value less than 0.05 was used to determine statistical significance.

The initial lycopene producing strain used here, HEBI-L-U (see Table 1), was generated in a previous study by using a CRISPR-Cas9-enabled markerless gene integration strategy (Schwartz et al., 2017b). As such, the strain retained both the ura3 and leu2 auxotrophies that were present in the starting Y. lipolytica PO1f strain. To investigate the effect of alleviating these auxotropies, we sought to determine the effect of restoring functional LEU2 and URA3 genes on lycopene production.

Linear DNA containing one of the selectable markers was transformed into the HEBI-L-U and randomly integrated into the genome to alleviate each of the auxotrophies. Lycopene production from the resulting strains, reported as titer, is presented in Figure 2A. As shown, alleviating either auxotrophy individually did not have an effect on lycopene titer, while alleviating both increased lycopene titer from 23 ± 1.6 mg/L to 44 ± 18.8 mg/L, an almost two-fold increase. This increase was primarily due to higher cell mass production. When lycopene production was normalized to DCW to give the specific lycopene yield (mg lycopene/g DCW), the HEBI-L-U and HEBI strains were almost identical, with both producing approximately 2 mg lycopene/g DCW (Figure 2B). Alleviating uracil and leucine auxotrophies resulted in higher overall cell mass and a corresponding increase in lycopene production, but did not affect the specific lycopene content.

A previously demonstrated strategy to increase carotenoid production in Y. lipolytica is the disruption of β-oxidation (Matthaus et al., 2014). β-oxidation is the biological process through which intracellular lipids are degraded and metabolized. As sequestration of lycopene in lipid droplets (which are composed primarily of neutral lipids) has been reported to be important for overall production of lycopene, blocking this degradation pathway may increase lycopene accumulation. To test this hypothesis in our host, we used a previously described CRISPR-Cas9 system to disrupt the MFE1 gene in the HEBI-L-U strain, yielding HEBI ΔMFE1. Transformation with pCRISPRyl_MFE1, outgrowth for 2 days, and plating on rich media enabled the identification of a strain containing an indel mutation in the MFE1 gene that rendered it non-functional (Supplementary Table S3).

The HEBI and HEBI ΔMFE1 strains were grown for 4 days and their respective lycopene yields were measured (Figure 3). The HEBI strain was found to have higher lycopene production than the HEBI ΔMFE1 strain, with HEBI ΔMFE1 yielding a 30% reduction in specific lycopene content. The reduction in lycopene yield was more apparent when the cultures were grown to 10 days, when the HEBI strain produced 1.8-fold more lycopene than HEBI ΔMFE1. Disruption of β-oxidation appeared to have a deleterious effect on specific lycopene content, especially at extended culture times once glucose is exhausted.

A previous study in Y. lipolytica found that disruption of phosphatidic acid phosphohydrolase, PAH1, resulted in increased expression of proteins localized to the ER (Guerfal et al., 2013). Several genes needed for synthesis of lycopene, including HMG1, are localized to the ER, and so increasing membrane protein expression might increase flux to lycopene. To test this hypothesis, the PAH1 gene was disrupted in the HEBI-L-U strain by CRISPR-Cas9, with an indel mutation rendering the PAH1 gene non-functional (Supplementary Table S3). Lycopene production between the HEBI and HEBI ΔPAH1 strains were compared; however, no significant change in lycopene production was observed (Figure 4).

A recent work showed that disruption of glycogen synthase (GSY1) in Y. lipolytica gave an increase in lipid accumulation by diverting carbon from carbohydrate storage to lipid storage (Bhutada et al., 2017). We hypothesized that eliminating glycogen biosynthesis may increase overall acetyl-CoA production, and that some of this increased pool of acetyl-CoA might then enter the mevalonate pathway and lead to higher lycopene yields. The GSY1 gene was disrupted in the HEBI-L-U strain by CRISPR-Cas9 (Supplementary Table S3) and HEBI and HEBI ΔGSY1 were compared for lycopene production (Figure 5). It was found that under the conditions used in this study, disruption of GSY1 did not lead to an increase in lycopene production.

Lycopene biosynthesis is dependent upon the products of the mevalonate pathway as precursors, and so increasing flux through the mevalonate pathway is likely to increase lycopene production. The mevalonate pathway, from acetyl-CoA to isopentenyl diphosphate (IPP), consists of 6 enzymatic reactions (Figure 6A). First, 2 molecules of acetyl-CoA are condensed to form acetoacetyl-CoA by acetoacetyl-CoA thiolase (ERG10). A third acetyl-CoA molecule is then added to form 3-hydroxy-3-methylglutaryl-CoA (HMG-CoA) by HMG-CoA synthase (ERG13). The next step, which is reported to be rate-limiting (Donald et al., 1997), involves the conversion of HMG-CoA to mevalonate (MVA) by HMG-CoA reductase (HMG1) while also consuming 2 molecules of NADPH. Mevalonate is then phosphorylated twice, first by mevalonate kinase (ERG12) and second by phosphomevalonate kinase (ERG8), with each phosphorylation consuming 1 molecule of ATP and resulting in mevalonate-5-diphosphate (MV5PP). MV5PP is then converted to IPP by mevalonate pyrophosphate decarboxylase (MVD1), which consumes 1 molecule of ATP and removes 1 carbon from the molecule. Finally, IPP can be isomerized to dimethylallyl-pyrophosphate (DMAPP) by isopentenyl diphosphate:dimethylallyl diphosphate isomerase (IDI1).

To identify the limiting step(s) in this pathway in the context of lycopene biosynthesis, the HEBI-L-U strain was separately transformed with an expression cassette for each gene in the pathway and the auxotrophies were alleviated (Figure 6B). Production of lycopene with each gene separately overexpressed was then measured. Only one gene showed a statistically significant increase in lycopene production, the gene encoding MVD1. MVD1 overexpression yielded an increase of ∼1.9-fold in specific lycopene yield. Two additional genes, ERG8 and HMG1, appeared to increase lycopene production, although the results were not statistically significant (p∼0.1). ERG10, ERG13, and ERG12 appeared to reduce specific lycopene content (but, the reductions are not statistically significant), possibly due to the overexpressions leading to the accumulation of intermediates and an unbalanced pathway.

For the purposes of this discussion, we describe the lycopene biosynthetic pathway as the steps necessary to convert the products of the mevalonate pathway (IPP and DMAPP) to lycopene (Figure 7A). This pathway is a combination of native enzymes and heterologous activities, as Y. lipolytica does not natively synthesize lycopene. First, DMAPP and IPP are condensed to form geranyl-diphosphate (geranyl-PP) by farnesyl pyrophosphate synthetase (ERG20). The same enzyme, ERG20, then catalyzes the addition of a second IPP molecule to produce farnesyl-diphosphate (farnesyl-PP). Farnesyl-PP can then be converted to geranylgeranyl-diphosphate (GGPP) by geranylgeranyl diphosphate synthase (BTS1). In a previous work, we found that a heterologous geranylgeranyl diphosphate synthase (CrtE, from Pantoea ananatis) was more effective (Schwartz et al., 2017b), and so it is used exclusively in this work. The production of GGPP is the end of the native yeast biosynthetic pathway, the subsequent steps are heterologous. Two molecules of GGPP are condensed by phytoene synthase (CrtB, from P. ananatis) to form phytoene. Phytoene then undergoes 4 consecutive desaturation reactions catalyzed by lycopene synthase (CrtI, from P. ananatis) to yield lycopene.

Each enzyme in this pathway was separately overexpressed by random integration in the HEBI-L-U strain, and the auxotrophies were alleviated. The lycopene production from each strain was quantified (Figure 7B). Only 1 gene showed an increase in lycopene production relative to the HEBI strain: specific lycopene production was increased 1.8-fold in the strain with a second copy of CrtI. These results suggested that the final step, four consecutive desaturations of phytoene, is the limiting step in lycopene biosynthesis. ERG20, CrtE, and CrtB appeared to reduce specific lycopene content, possibly due to intermediate accumulation; however, none of the observed reductions in lycopene were statistically significance.

Through sequentially testing a variety of different genetic knockouts and pathway enzyme overexpressions, a set of changes to maximize lycopene production was identified. Alleviation of both the ura3 and leu2 auxotrophies was combined with sequential overexpressions of different combinations of the mevalonate and lycopene pathway enzymes that resulted in higher production (HMG1, ERG8, MVD1, and CrtI; Figure 8). To achieve this, HMG1 and MVD1 were integrated into the genome using CRISPR-Cas9-mediated markerless integration to yield HEBI HV -L-U. The gene coding for HMG1 was integrated into the disrupted LEU2 locus in the PO1f genome, and the gene for MVD1 was integrated in a similar manner into the pseudogene YALI0E07645g locus. Integration was confirmed by PCR. The auxotrophies of this strain were then alleviated to yield HEBI HV. Linear expression cassettes for ERG8 and CrtI overexpression were transformed into the HEBI HV -L-U strain to yield HEBI HV8 and HEBI HVI, respectively. Due to literature reports of HMG1 being limiting, a third exogenous copy of HMG1 was also integrated in the same way, yielding HEBI HHV. Finally, linear DNA fragments containing the URA3 and LEU2 selectable markers and ERG8 and CrtI were separately transformed into HEBI HV -L-U, to yield HEBI HV8I. Using this strategy, the quadruple integration gave the highest level of lycopene production, at about 2.6-fold higher than the HEBI strain. All of the triple and double integrations produced similar specific lycopene contents, indicating that the quadruple integrant had a synergistic effect to maximize flux to lycopene.

To further maximize lycopene production, the HEBI HV8I strain was grown to longer times. Figure 3 demonstrates a 2-fold increase in specific lycopene content when increasing the culture time from 4 to 10 days. We tested to see if this would be replicated in the HEBI HV8I strain. Additionally, a glucose feeding strategy was tested. For glucose feeding, every 2 days 2.5 mL of 40% glucose was added to the culture of HEBI HV8I in YPD10 media. For both feeding and non-feeding cultures, samples were taken every 2 days to determine the titer of lycopene. The results are shown in Figure 9. As shown, no feeding of glucose resulted in a lycopene titer of 155 ± 26 mg/L, while the glucose feeding strategy maximized lycopene titer throughout the culture and at day 12, reaching as high as 213 ± 8 mg/L lycopene.

In order to fully maximize specific lycopene yields, the HEBI HV8I strain was cultured in a fed batch bioreactor. Specific lycopene content measurements and cell pellet images were taken each day over the 10-day experiment. As can be seen in Figure 10A and Supplementary Figure S2, the culture turned visibly red, indicating significant lycopene production. Figure 10B shows specific lycopene measured over the course of the culture. After 10 days of growth, a specific lycopene yield of 21.1 mg/g DCW was obtained.