Engineering C. tropicalis for α-humulene production
Because C. tropicalis lacks an efficient autonomously replicating plasmid, exogenous genes are usually integrated into the genome for stable expression. Previous studies showed that single deletion of CAT in C. tropicalis had no impact on cells growth [31], therefore the CtCas9 expression cassette was integrated at the CAT locus to generate strain CU-207 for further facilitating genetic manipulation. After the URA3 marker was excised from C. tropicalis CU-207, the resulting uracil auxotrophic strain CU-208 was used as the platform strain for further metabolic engineering. Our previous studies showed that it is challenging to express heterologous genes in C. tropicalis without codon optimisation [25, 31]. Thus, the codon-optimised ZSS1 from Z. zerumbet was integrated into the chromosome of C. tropicalis CU-208 through CRISPR–Cas9, resulting in strain HC01 (possessing a single copy of ZSS1 at the POX5 locus) and HC02 (possessing double copies of ZSS1 at the POX5 loci). After 96 h of fermentation, α-humulene production was detected by GC–MS (Fig. 1b and Additional file 2: Fig. S1). HC-02 produced 1.41 mg/L α-humulene, roughly double that of HC-01 (0.77 mg/L), suggesting that the ZSS1 gene can be successfully expressed in C. tropicalis, and the expression level of ZSS1 might be a key factor influencing α-humulene production. Compared with CU-207, the biomass of HC01 and HC02 was not markedly different, indicating that a low level of α-humulene had little or no effect on yeast growth. Nevertheless, the α-humulene concentration was lower than that reported for S. cerevisiae (2.32 mg/L, possessing a single copy of ZSS1) [15].
Previous studies have reported that the peroxisome was more appropriate for the synthesis of limonene, α-humulene and squalene in S. cerevisiae [15, 17, 32]. Therefore, it is necessary to evaluate whether the peroxisome could be benefit for producing α-humulene in C. tropicalis. However, no confirmed peroxisome targeting signals are presented in C. tropicalis. Firstly, the function of peroxisome targeting signal-1 (PTS1, SKL) was evaluated with ZSS1-GFP as a reporter. To label peroxisome, a red fluorescence protein was fused with peroxisome membrane protein (PEX3-mScarlet) and co-expressed with ZSS1-GFP-PTS1. Fluorescence microscopy results showed that the GFP and mScarlet signals colocalised (Fig. 1c), indicating that ZSS1-GFP-PTS1 could be transported into peroxisome. However, when α-humulene synthase was directed to peroxisomes by PTS1, only 0.06 mg/L of α-humulene accumulated in the transformant possessing double copies of ZSS1-PTS1 (HP02; 0.04 mg/L for HP01 possessing one copy of ZSS1-PTS1; Fig. 1b).
Effects of HMGR and ERG10 overexpression and ERG9 repression on α-humulene production
The biosynthesis of α-humulene from acetyl-CoA in C. tropicalis requires multiple enzymes and complex metabolic regulation (Fig. 1a). Previous studies demonstrated that overexpression of HMGR and ERG10 and repression of ERG9 expression positively affect terpenoid production in S. cerevisiae and Y. lipolytica [27, 29, 33]. Therefore, the influence of these three genes on the production of α-humulene in C. tropicalis was evaluated.
First, HMGR and ERG10 were expressed in HC03 (a uracil auxotrophic derivative of HC02). The resulting strain HC05 (overexpressing HMGR) produced 1.88 mg/L α-humulene, 33.3% more than HC02 (Fig. 2, HC05 vs. HC02). By contrast, overexpressing ERG10 did not improve production of α-humulene (Fig. 2, HC04 vs. HC02).
It has been reported that the hydroxymethylglutaryl-CoA reductases of S. cerevisiae and Y. lipolytica share similar structures, with N-terminal multiple transmembrane domains and a C-terminal catalytically active domain [4, 29, 34]. Moreover, the N-terminal domain is a response element for signal regulation; its deletion can enhance protein stability. Thus, a truncated HMGR of C. tropicalis (tHMGR, lacking the N-terminal multiple transmembrane domains, Additional file 2: Fig. S2) was overexpressed in HC03. As expected, α-humulene production in HC06 (overexpressing tHMGR) was increased by 66.7% compared to HC02 (Fig. 2).
Squalene synthase (ERG9) catalyses the reductive dimerisation of two FPP moieties to form one molecule of squalene. FPP is a precursor of sesquiterpenoids in eukaryotes, while squalene plays an integral role in sterol synthesis (Fig. 1a). In order to increase the FPP flux towards α-humulene biosynthesis, a single copy of ERG9 was disrupted in HC03, generating strain HC07. Unexpectedly, the biomass and α-humulene titre of HC07 were decreased compared with HC02, although the α-humulene content was improved slightly (Fig. 2). In addition, the content of β-carotene in ERG9-disrupted C. tropicalis followed the same trend (YJ Li et al. unpublished). RT-qPCR analysis showed that ERG9 mRNA levels in strain HC07 were 43% lower than in HC02 (Additional file 2: Fig. S3), indicating that expression of ERG9 was repressed. Similarly, the lycopene production capacity of engineered Candida utilis could not be increased when two copies of ERG9 were deleted in tetraploid yeast [35]. In our previous study, one copy of the CAT gene was disrupted in C. tropicalis, and carnitine acetyltransferase mRNA levels and enzyme activity were decreased [31]. Meanwhile, the concentration of α,ω-dodecanedioic acid was not changed significantly. These data may indicate that the normal allele can completely (or mostly) cancel the mutant allele in diploid and polyploid yeast.
Overexpressing the entire α-humulene synthesis pathway to improve α-humulene production
Although strains HC02 and HP02 could produce α-humulene, production was very low. This might be due to the inefficiency of the native MVA pathway, which is tightly regulated in yeast. Firstly, a short synthetic terminator (Tsynth7, 32 bp) [36], which functions in S. cerevisiae and Y. lipolytica, was functionally verified in C. tropicalis by the GFP reporter system (Additional file 2: Fig. S4). Then the strong promoters PGAP1 and PFBA1, and terminators Tsynth7, TENO1, TPGK1 and TADH2, were used to control gene expression. To further enhance α-humulene production, genes encoding the entire α-humulene synthesis pathway (ERG10, ERG13, tHMRG, ERG12, ERG8, ERG19, IDI1, ERG20 and ZSS1) were constitutively overexpressed in cytoplasmic and peroxisome fractions of CU-208. The α-humulene titre of the peroxisome engineered strain DP-H01 (expressing double copies of the nine genes) was 2.42 mg/L, 43.33-fold higher than HP02, and 1.70-fold higher than the strain expressing one copy of each of the nine genes in peroxisomes (Fig. 3a). These results indicate that the peroxisome-targeted α-humulene biosynthetic pathway could enhance production in C. tropicalis. Similarly, previous studies reported that targeting biosynthetic pathways to peroxisomes can enhance productivity and inhibit by-product formation [32, 37]. However, the α-humulene titre in the engineered C. tropicalis strain was much lower than that reported for S. cerevisiae [15]. Interestingly, strains overexpressing the entire α-humulene synthesis pathway in the cytoplasm exhibited a remarkable increase in α-humulene production. In strain DC-H01 expressing double copies of the α-humulene synthesis pathway genes in the cytoplasm, α-humulene production was improved more than fivefold compared with DP-H01, reaching 12.89 mg/L (Fig. 3a). A similar result was obtained for strain SC-H01 expressing only one copy of each α-humulene synthesis pathway gene in the cytoplasm (Fig. 3a). In addition, the cell growth of the engineered strains was significantly inhibited compared with the initial strain (Fig. 1b, a)
The peroxisome subcellular organelle is nonessential for yeast growth, and a series of studies have focused on peroxisome engineering of yeast for terpene production [15,16,17]. However, our current results showed that the peroxisome of C. tropicalis is not an ideal subcellular location for α-humulene production. Therefore, strain DC-H01 was chosen for further genetic modification to improve α-humulene biosynthesis.
Identifying rate-limiting steps in the α-humulene pathway
Many researchers have demonstrated that cytoplasmic-engineered S. cerevisiae can be used for terpenoid production with high efficiency [38, 39]. Moreover, β-carotene production in cytoplasmic-engineered Y. lipolytica reached 6.5 g/L [40]. Our current results showed that the α-humulene titre of DC-H01 was significantly higher than that of SC-H01, indicating a bottleneck in the α-humulene biosynthesis pathway of the SC-H01 strain. However, the α-humulene titre was significantly lower (12.89 mg/L, Fig. 3a). Therefore, we hypothesised that α-humulene biosynthesis in DC-H01 may be limited by one or several steps in the pathway.
To confirm this hypothesis, three gene expression cassettes (cassette 1 for ERG10, ERG13 and tHMGR expression; cassette 2 for ERG12, ERG8, ERG19 and IDI1 expression; cassette 3 for ERG20 and ZSS1 expression) were constructed and transformed into strain DC-H02 (a URA3 pop-out derivative of DC-H01), generating strain DC-H03S (expressing one copy of cassette 1), DC-H03D (expressing double copies of cassette 1), DC-H05S (expressing one copy of cassette 2), DC-H05D (expressing double copies of cassette 2), DC-H07S (expressing one copy of cassette 3) and DC-H07D (expressing double copies of cassette 3). Engineered strain DC-H07S produced nearly 50% more α-humulene compared than DC-H01 (19.33 mg/L vs 12.90 mg/L; Fig. 3b), whereas strain DC-H07D produced 32.68 mg/L of α-humulene, indicating that overexpression of ERG20 and ZSS1 enhanced α-humulene production. Further studies indicated that the increase in α-humulene titre was mainly due to expression of ZSS1 (Fig. 3b, strain DC-H09D overexpressing ZSS1 vs DC-H07D). However, the effects of co-expression of ERG10, ERG13 and tHMGR, and ERG12, ERG8, ERG19 and IDI1 were limited (DC-H03S, DC-H03D, DC-H05S and DC-H05D vs. DC-H01), indicating that the steps catalysed by these enzymes are not the bottlenecks for α-humulene production in strain DC-H01. Compared with DC-H01, expression of cassette 2 inhibited cell growth of strains DC-H05S and DC-H05D. It was previously reported that IPP and DMAPP are toxic to mitochondria, and higher levels of these pyrophosphorylated intermediates can inhibit the growth of cells [33]. Overexpression of ERG12, ERG8, ERG19 and IDI1 genes can lead to accumulation of IPP and DMAPP in DC-H05S and DC-H05D, and they may be transported from the cytoplasm to the mitochondria [41], where they disrupt mitochondrial function and inhibit cell growth.
Considering that increasing the expression of the ZSS1 gene can significantly increase the production of α-humulene (DC-H09D vs DC-H01 and DC-H07D vs DC-H01), we speculated that increasing the copy number of the ZSS1 gene may further improve the yield of α-humulene. Since the GAP1 promoter is one of the strongest promoters (more than twofold stronger than the FBA1 promoter) [25], we chose this promoter to overexpress the ZSS1 gene. The ZSS1 expression cassette was integrated at the D-lactate dehydrogenase gene (DLD1b) and/or the lipid phosphate phosphatase gene (LPP2) locus of strain DC-H08 (a URA3 pop-out derivative of DC-H07D) to increase the copy number of ZSS1, resulting in strains DC-H11S, DC-H11D, DC-H13S and DC-H13D. As shown in Fig. 3c, compared with DC-H07, the titre of α-humulene was significantly improved with increasing ZSS1 copy number. The maximum α-humulene levels in DC-H11S, DC-H11D, DC-H13S and DC-H13D reached 55.25, 77.00, 92.87 and 99.62 mg/L, respectively (Fig. 3c). Moreover, the engineered strains showed a slight increase in cell growth compared with DC-H07.
Lipid phosphate phosphatase is one of the main contributors to phosphate phosphatase activity in yeast. Deleting the LPP1 gene can increase sesquiterpene levels in S. cerevisiae [39, 42]. However, our results showed that deleting the LPP2 gene did not increase the α-humulene titre in C. tropicalis (Fig. 3c, DC-H15D vs DC-H07D). Compared with DC-H07D, the farnesol titre of strain DC-H15D was not changed significantly (data not shown). Indeed, in addition to the LPP2 alleles, there are other phosphate phosphatases (at least three isozymes of diacylglycerol pyrophosphate phosphatase) in C. tropicalis.
Previous studies have shown that HMGR is the first rate-limiting enzyme in the MVA pathway, and NADH-dependent HMG-CoA reductase (NADH-HMGR) from S. pomeroyi has better performance for the production of sesquiterpenoid in yeast [38, 43]. To further investigate the rate-limiting step of DC-H13D for α-humulene synthesis, tHMGR and NADH-HMGR from S. pomeroyi were overexpressed. Compared with tHMGR, NADH-HMGR achieved a more significant increase in α-humulene titre (Fig. 3d, DC-H17D vs DC-H13D and DC-H19D vs. DC-H13D). When both NADH-HMGR and ERG10 genes were overexpressed in DC-H13D (generating strain DC-H21D), an α-humulene titre of 119.07 mg/L was achieved, ~19.5% higher than that of DC-H13D (Fig. 3d).
Fed-batch fermentation for α-humulene production
In order to improve α-humulene production of strain DC-H21D, three different types of medium, nitrogen stress medium [23] with 100 g/L glucose, YPD60 medium and Y20P40D60 medium, were tested in shake flasks prior to fed-batch fermentation. The α-humulene titre of DC-H21D was increased to 171.50 mg/L and 195.31 mg/L in YPD60 and Y20P40D60 medium, an increase of 44.0% and 64.0% compared with YPD medium (Fig. 4). Moreover, biomass was also improved. However, cell growth and α-humulene production in nitrogen stress medium were significantly lower than in YPD medium (Fig. 4).
To further characterise α-humulene production in C. tropicalis, strain DC-H21D was employed for fed-batch fermentation in a 5-L bioreactor (Bailun Co., Shanghai, China) with 2 L YPD60 or Y20P40D60 medium. As shown in Fig. 5, the strain DC-H21D grew continuously in both fermentations. Finally, the maximum titre of α-humulene reached 1957.28 mg/L and 3144.37 mg/L from the YPD60 and Y20P40D60 medium, respectively, at 216 h (Fig. 5a, b). In order to further increase the titre of α-humulene, scale-up experiment was performed in a 30-L bioreactor (INFORS, Switzerland) with 12 L Y20P40D60 medium (Fig. 5c). In this fed-batch culture, glucose was quickly consumed within 16 h, and feeding was initiated at ~ 16 h after fermentation. The biomass (OD600) of DC-H21D increased gradually until 156 h, then fluctuated between 460 and 480 until the end of the fermentation. The concentration of α-humulene steadily increased throughout the cultivation period, and a maximum titre of 4115.42 mg/L was achieved in 264 h of fermentation. These results demonstrate the enormous potential of C. tropicalis to produce α-humulene and other terpenoids.