Construction and analysis of a cell factory for terpenoid biosynthesis in Pichia pastoris via metabolic engineering and metabolomics

Abstract

Terpenoids are widely distributed in nature and have various applications in health products, pharmaceuticals, and fragrances. Despite the tremendous potential of terpenoids, traditional production methods such as plant extraction and chemical synthesis face challenges in meeting current market demands. With the developments in synthetic biology and metabolic engineering, it has become feasible to construct efficient microbial cell factories for large-scale production for terpenoids. In this work, using the yeast Pichia pastoris as the host cell, a "plug-and-play" cell factory for universal terpenoid production was constructed by enhancing the expression of the MVA pathway for common precursor synthesis and reducing branch pathway diversion strategies. We have successfully and efficiently synthesized β-elemene, β-farnesene, (+)-valencene, (−)-α-bisabolol by this cell factory. Furthermore, by analyzing metabolites in different engineered strains in terms of system biology, it was discovered that an increase in key protein copy numbers enhanced the synthesis of arginine and other metabolic pathways. The robustness of the strains and the tightly regulated metabolic network constrain rational metabolic engineering transformations. These data provide important clues for the modification and optimization of production strains.

Graphical abstract

1. Introduction

Terpenoid, a class of secondary metabolites widely present in nature, plays an indispensable role in the fields of medicine, fragrance, and energy due to their unique biological activity and application value [1,2]. For example, artemisinin as a main component of antimalarial drugs [3], and paclitaxel is a widely used chemotherapy drug for cancer treatment [4]. Both of them are representative terpenoid compounds. However, traditional methods of terpenoid production, such as plant extraction and chemical synthesis, are limited by issues such as scarce resources, high production costs, and environmental pollution, promoting the development of microbial cell factories as an alternative production method [[5], [6], [7]]. Although microbial synthesis pathways provide new opportunities for sustainable production of terpenoid, industrial application still faces challenges such as increasing yield and reducing costs [8,9].

Among many microbial cell factories, Pichia pastoris has become an ideal host for synthesizing terpenoid due to its physiological characteristics like high-density fermentation capability [10,11], powerful promoter system (AOX1), and excellent protein secretion ability [12,13]. However, despite its endogenous mevalonic acid pathway (MVA) providing precursors for terpenoid synthesis, the natural flux often fails to meet the demands of efficient production. Currently, strategies to enhance the efficiency of the MVA pathway and address bottleneck issues have become important research directions in this field, such as precursor supply and competitive branch metabolism [14]. Regulation strategies of the MVA pathway are particularly crucial. By upregulating the expression of key rate-limiting enzymes and inhibiting branch pathways, the synthesis efficiency of target products can be effectively increased. These strategies have been validated in different yeast systems, indicating that precise metabolic engineering regulation can significantly improve the efficiency of terpenoid compound production [15,16].

Furthermore, metabolomics, as a powerful tool in systems biology, plays an important role in optimizing cell metabolism. Untargeted metabolomics not only reveals the ability for global metabolic reprogramming but also identifies unknown byproducts and metabolic bottlenecks, providing important guidance for iterative engineering. Successful cases demonstrated that the systematic analysis combining metabolomics could effectively guide the engineering transformation of terpenoid production [17].

This study aims to construct an efficient platform for terpenoid compound production through three-stage engineering modifications in P. pastoris (Fig. 1). Firstly, the synthesis of target products was achieved by introducing heterologous synthetic enzymes into P. pastoris. Secondly, strain PP3 was developed by overexpressing rate-limiting enzymes of the MVA pathway including truncated HMG-CoA reductase (tHMG1), isopentenyl diphosphate isomerase 1 (IDI1), mevalonate kinase (ERG12), mevalonate diphosphate decarboxylase (ERG19) and limiting branch squalene synthase (ERG9), further enhancing the synthesis efficiency of target products. Subsequently, strains PP4 and PP5 were constructed by increasing the copy numbers of IDI1 and tHMG1 in strain PP3, leading to efficient synthesis of target products. Lastly, by utilizing untargeted metabolomics technology, a systematic metabolic analysis was conducted on the wild-type strain WT, intermediate strain PP3, and final engineered strain PP5, providing a deep understanding and guidance for subsequent engineering modifications. Through this series of engineering strategies, this study not only improved the production efficiency of terpenoid compounds but also provided new ideas and methodologies for the application of microbial cell factories in the field of biosynthesis.

Fig. 1.

Schematic diagram of a versatile chassis for terpenoid production in P. pastoris. Red indicates that it needs to be strengthened, and blue indicates that it needs to be weakened. tHMG1, truncated HMG-CoA reductase; ERG12, mevalonate kinase; ERG19, mevalonate pyrophosphate decarboxylase. IDI1, isopentenyl diphosphate isomerase 1; ERG9, Squalene synthase.

2. Materials and methods

2.1. Strains and culture condition

Escherichia coli strain DH5α was utilized for the creation of plasmids, while P. pastoris strain GS115 served as the foundational strain. The plasmid HZP-gRNA (PARS1, Zeocin resistance gene, SERp-IntX-sgRNA) was employed for the expression of guide RNAs. IntX-TEF1 worked as monocistronic gene expression vectors, where X indicated various insertion sites. It was utilized for the construction and expression of the target genes [16].

2.2. Plasmid construction

The relevant plasmids and primers utilized in this investigation have been detailed in Supplementary Table S1 and Table S2, respectively. The gene GAS, FAS, CaVAL, and CcBOS were custom-synthesized by Nanjing GenScript Biotechnology Co., Ltd., with careful optimization of its codon usage, and the sequence information is provided in Supplementary Table S3. Critical genes involved in the mevalonate pathway, such as tHMG1, IDI1, ERG12 and ERG19, were amplified from P. pastoris genome and subsequently integrated into donor plasmids via a seamless cloning approach to generate the corresponding expression constructs. We truncated the promoter PERG9 to regulate the expression of ERG9 and marked as P_ERG9Δ100-ERG9. The method was discribed in previous article [16]. The design of the sgRNA plasmid was facilitated by the Benchling CRISPR tool (https://benchling.com/crispr/).

2.3. Strain construction

Various recombinant strains were derived through the targeted integration of the synthase gene and its biosynthesis-associated genes into the chromosomal DNA of P. pastoris utilizing the CRISPR/Cas9 technology. The gene expression cassette, inclusive of the homology arm, was co-transformed with the corresponding gRNA plasmid into the Cas9-expressing strain. The transformation of P. pastoris was achieved through an electroconversion technique [18]. The comprehensive list of all engineered recombinant strains can be found in Fig. 1 of this investigation.

2.4. Flask-scale fermentation of engineered strains and analytical methods for target products

The single colony strain of the constructed variant was introduced into a tube with YPD medium. The culture was incubated for 24 h at a temperature of 30 °C and at a speed of 220 rpm. Subsequently, it would be transferred to 50 mL of YPD medium with a 1 % inoculum for 96 h. After the initial 24 h of incubation, 10 % n-dodecane was added into flasks. For the analysis of the desired products, Gas Chromatography-Mass Spectrometry (GC-MS) was employed and the specific detection methods could be obtained in relevant articles published in the past [16,[19], [20], [21]].

2.5. LC-MS non-targeted metabolomics

This project used the LC-MS analysis platform to conduct metabolomics research on 18 samples (WT, PP3, PP5; six samples per group). The samples were first pretreated to remove impurities and extract metabolites. Then, in positive and negative modes, the information was collected by the computer to obtain the MS information of the metabolites. ProgenesisQI (Waters Corporation, Milford, USA) software was used to annotate metabolites, data preprocessing, etc., and finally obtain the metabolites list and data matrix. The differential metabolites were screened in combination with T test and VIP (OPLS-DA). The biological information of differential metabolism was further mined by advanced analyses such as pathway analysis, association analysis, and cluster analysis. Student's t-test combined with multivariate analysis of OPLS-DA was used to screen out the differential metabolites between groups (p value < 0.05).

3. Results

3.1. Construction of microbial chassis for terpenoid production in P. pastoris

Due to the absence of specific geranyl diphosphate (GPP) synthase, the endogenous MVA pathway in P. pastoris can only release a negligible amount of GPP, significantly curtailing the production of terpenoid compounds [22]. We subdivided the terpenoid compound synthesis pathway into two modules for engineering, as illustrated in Fig. 1. The common precursors, isopentenyl diphosphate (IPP) and dimethylallyl diphosphate (DMAPP), are predominantly from the MVA pathway. IDI1, tHMG1, ERG12 and ERG19 are the key genes to enhance metabolic flux of MVA pathway. IDI1 and tHMG1 are acknowledged as rate-limiting enzymes in the MVA pathway, while ERG12 and ERG19 are crucial genes in the MVA pathway of yeast, which have been fully proved in many researches [[22], [23], [24]]. Consequently, we overexpressed IDI1 and tHMG1, and further augmented the expression of ERG12 and ERG19, constructing PP1 and PP2 to enhance the MVA-mediated synthesis of IPP. On the other hand, GPP serves as the direct precursor, being converted into various terpenoid compounds by different terpene synthases. To diminish the endogenous conversion of GPP, we crafted strain PP3 by truncating the natural promoter of squalene synthase ERG9 [16], effectively limiting the transcription of it. Lastly, by increasing the copy numbers of IDI1 and tHMG1, we established strains PP4 and PP5, aspiring to markedly elevate the supply of precursors.

3.2. Employing microbial chassis to improve production of β-elemene

In our aims to construct chassis cells, we started from the biosynthesis of β-elemene, a proprietary anti-tumor medication of our nation. It finds widespread application in the clinical treatment of lung cancer, breast cancer, and similar ailments [25]. Germacrene A serves as a crucial intermediate in the biosynthesis of sesquiterpenes, fully transforming into β-elemene under high temperatures. We successfully expressed Anabaena variabilis-derived germacrene A synthase gene (GAS). The strain expressing GAS in the wild-type P. pastoris was designated GAS0, serving as the progenitor strain (Fig. 2). By comparing the yield of β-elemene across different chassis cells, we discerned that the upregulation of key enzymes in the MVA pathway, alongside the restriction of GPP diversion, markedly enhanced β-elemene production. Ultimately, by systematically increasing the copy numbers of crucial enzymes within the MVA pathway, we engineered the strain PP5, which, under shake-flask fermentation conditions, produced β-elemene at a concentration of 335.0 mg/L, a yield 10.5 times higher than that of the original GAS0 strain.

Fig. 2.

Construction of a recombinant P. pastoris strain for producing β-elemene. Expression cassettes containing different genes (IDI1, tHMG1, ERG12 and ERG19) are inserted into specific loci on the genome of P. pastori. “+” indicates that the expression cassette had been integrated into the genome of the target strain. “-” indicates that the gene had not been integrated into the expression cassette of the target strain. The values of product are the average of three biological replicates.

3.3. Employing microbial chassis to improve production of β-farnesene

Concurrently, we embarked on the biotechnological synthesis of β-farnesene, which is regarded as an exemplary alternative to jet fuel. Moreover, its significance extends to the manufacture of polymers, surfactants, and lubricants [26]. Nevertheless, the natural synthesis of β-farnesene is constrained by plant growth, rendering it incapable of meeting the burgeoning market demand. Drawing upon previously published research, we expressed farnesene synthase (FAS) from Artemisia annua in diverse chassis cells. Strain expressing FAS in wild-type P. pastoris was designated as FAS0 (Fig. 3). Consistent with our findings on β-elemene, the enhanced expression of key enzymes in the MVA pathway, coupled with the restriction of GPP flux, substantially promoted the production of β-farnesene. Moreover, by incrementally increasing the copy number of critical enzymes in the MVA pathway, we engineered the strain PP5, which under shake-flask fermentation conditions produced 84.9 mg/L of β-farnesene, a 28.3-fold increase over the original FAS0 strain. However, we also discovered that, in the synthesis of β-farnesene, increasing the copy numbers of IDI1 and tHMG1 proved significantly more effective than limiting GPP diversion. This was evidenced by the fact that the yield of PP4-FAS was enhanced by 156.2 % over PP3-FAS, whereas PP3-FAS only achieved a 66.9 % improvement over PP2-FAS. This also suggests that the expression of different synthases would influence the efficacy of various engineering strategies for the high-efficiency synthesis of terpenoid products, thereby indicating the necessity for a more detailed exploration and analysis of metabolic fluxes within different chassis cells.

Fig. 3.

Employing microbial chassis to improve production of β-farnesene. Expression frames containing β-farnesene are inserted into different P. pastori (WT, PP1, PP2, PP3, PP4, PP5) to produce β-farnesene. The values of product are the average of three biological replicates.

3.4. Sustainable production of (+)-valencene and (−)-α-bisabolol in PP5

To corroborate the universality and efficiency of the engineered P. pastoris PP5 as a versatile cell factory for the heterologous expression of terpenoid, we introduced the (+)-valencene synthase gene CaVAL and the bisabolol synthase gene CcBOS into both the wild-type and PP5 strains, subsequently assessing the production of valencene and (−)-α-bisabolol. The successful expression of CaVAL and CcBOS facilitated the heterologous synthesis of (+)-valencene and (−)-α-bisabolol (Fig. 4). Remarkably, within the PP5 chassis cells, the shake flask yields of (+)-valencene and (−)-α-bisabolol reached 78.2 mg/L and 63.2 mg/L respectively, representing a 34-fold and 30-fold increase over the yields observed in the wild-type chassis cells. These findings further attested that the PP5 could work as a highly effective and versatile cell factory in P. pastoris for the production of terpenoid compounds.

Fig. 4.

Sustainable production of (+)-valencene and (−)-α-bisabolol in PP5. Expression frames containing β-farnesene are inserted into different P. pastori (WT, PP5) to produce β-farnesene. The values of product are the average of three biological replicates.

3.5. Identification of metabolic perturbations induced by chassis engineering through non-targeted metabolomics

Through the heterologous expression of β-elemene, β-farnesene, (+)-valencene and (−)-α-bisabolol, we have demonstrated that PP5 can serve as an effective cellular factory for the production of terpenoids in P. pastoris, also highlighting the universality and efficiency of traditional modification logic. However, this approach has reached its production bottleneck. To further guide the subsequent engineering constructions, we selected three strains—WT, PP3, and PP5—for non-targeted metabolomic analysis, aiming to elucidate the specific differences in metabolic flow. The metabolic products in WT, PP3, and PP5 exhibited pronounced disparities. The results of the Principal Component Analysis (PCA) vividly highlighted significant variations in the overall metabolic profiles of the three groups, with the trajectories of the groups within the direction of PC1 (50.7 %) displaying a progressive distribution from WT to PP5 (Fig. 5). Furthermore, the Venn diagram, mirroring the PCA findings, revealed a mere 124 common differential metabolites (14.99 %) across the three comparison groups. These findings suggested that targeted construction within the MVA pathway precipitated widespread metabolic changes. This phenomenon likely resulted from the combined effect of the cascading impacts within the cellular metabolic network and the regulation of robustness.

Fig. 5.

Identification of metabolic perturbations induced by chassis engineering through non-targeted metabolomics. A. Metabolite clustering analysis. A clustering analysis was performed on the top 50 differential metabolites, grouping metabolites with similar expression patterns together, and analyzing their functional correlations. B–D. KEGG Compound Classification of Difference Metabolites in PP3 vs WT (B) PP5 vs WT (C), and PP5 vs PP3 (D). The identified metabolites were compared to the KEGG Compound database to obtain an overview of metabolite classification and to create statistical plots.

However, the analysis of the metabolome revealed that the constructions targeting the MVA pathway did not fully achieve the expected redirection of metabolic flow. Two key metabolites directly related to the MVA pathway were identified (Fig. 6). Mevalonic acid exhibited a mere 1.05- and 1.04-fold increase in PP3 and PP5 compared to the wild type respectively (p < 0.05), and there was no significant change in squalene (a theoretical inhibitory node product). Considering the significant increase in the target product yields in the aforementioned examples, we hypothesized that this might be due to the low consistency of metabolite levels within the linear biosynthesis pathway. An upregulation might lead to an increase in the abundance of the final product, whereas the levels of intermediates might remain unchanged, merely reflecting an accelerated conversion rate.

Fig. 6.

Expression trend analysis between WT, PP3 and PP5. A. Trend analysis chart generated by the STEM algorithm. Each rectangle corresponded to a profile. The number in the upper-left corner of the rectangle was the profile number, ranging from 0 to 30. The broken line represented the expression trend of metabolites in different samples, and the value in the lower-left corner was the P-value. B. The KEGG enrichment analysis results of metabolites in Profile 4.

Global analysis of differential metabolites indicated that both PP3 and PP5 developed metabolic profiles conducive to the synthesis of terpenoid compounds compared to the wild type. However, the differential metabolites between PP3 and PP5 were not enriched in terpenoid-related pathways, suggesting that the final stage of MVA pathway construction did not significantly enhance the efficiency of terpenoid biosynthesis network. Expression trend cluster analysis revealed significant unexpected effects. In PP5, genes involved in the arginine biosynthesis pathway were significantly upregulated, with higher concentrations than those in PP3 and WT (p < 0.01). This phenomenon, temporally and spatially associated with MVA pathway modification, indicated that the engineering intervention at the final stage triggered a systemic reprogramming of nitrogen metabolism pathway. The activation of the arginine biosynthesis pathway, consuming the TCA cycle intermediate α-ketoglutarate, might indirectly affect the supply of MVA pathway precursor acetyl-CoA through a carbon skeleton competition mechanism, providing potential correlative evidence to explain the modest effects of over-copy modification of key genes.

4. Discussion

The P. pastoris has gradually emerged as a promising host for natural products synthesis [[27], [28], [29]]. It is widely acknowledged as an excellent recombinant protein expression host with high efficiency in protein folding and glycosylation. In recent years, P. pastoris has been successfully utilized in bioproduction for geraniol [30], ambergris [31], lycopene [32] and so on. However, the development of a universal chassis cell suitable for multi-terpenoid synthesis remains a critical issue in this field [33,34]. This study adopted a modular engineering strategy to gradually introduce key proteins in the MVA pathway, limit branch pathway protein expression, and increase the copy number of IDI1 and tHMG1 genes in P. pastoris, successfully constructing a "plug-and-play" universal chassis cell PP5, providing a novel platform for the biosynthesis of multi-terpenoid compounds.

Using β-elemene, β-farnesene, (+)-valencene and (−)-α-bisabolol as examples, the versatility and efficiency of the PP5 were validated. Flask fermentation results showed that the production of these four terpenoid compounds in PP5 reached 335.0 mg/L, 84.9 mg/L, 78.2 mg/L, and 63.2 mg/L, respectively, which were 10.5, 28.3, 34, and 30 times higher than the corresponding parental strains, demonstrating the effective adaptation of PP5 for the synthesis of different types of terpenoids. PP5 strains are beneficial for shortening the development cycle of construction for cell factories of terpenoid.

Moreover, non-targeted metabolomics analysis between the wild type, intermediate group PP3, and final group PP5, revealed reconstruction of the metabolic network from both target pathway and global metabolism perspectives. Within the target pathway, key metabolites of the MVA pathway such as mevalonate showed a significant increase (1.05- and 1.04-fold increase in PP3 and PP5 compared to the wild type respectively), and the expected competitive pathway metabolites were not significantly altered, indicating that the static overexpression strategy encountered resistance from the metabolic network robustness. However, from a global reprogramming perspective, both PP3 and PP5 exhibited metabolic profiles favorable for terpenoid synthesis compared to the wild type, with significant enrichment of differential metabolites.

Furthermore, PP5 specifically activated the arginine biosynthesis pathway. The further increase in the copy number of IDI1 and tHMG1 genes might exert the pressure on cell nitrogen metabolism or energy status, triggering a compensatory mechanism for arginine synthesis. The arginine synthesis pathway is closely linked to the TCA cycle, and its upregulation may affect TCA cycle flux, indirectly impacting the availability of acetyl-CoA [35]. This might be part of the global metabolic reprogramming and related to the stress response to engineering pressure. Excessive activation of the arginine pathway would consume a large amount of carbon sources and ATP, potentially becoming a new bottleneck limiting terpenoid synthesis efficiency or affecting cell growth. Further validation of its impact was needed.

The terpenoid biosynthesis chassis cells we have constructed in Pichia pastoris are merely the rudimentary version, lacking in comparison to specialized cellular factories engineered in other yeasts for the production of target terpenoid products. Using β-elemene as an example, Ogataea polymorpha achieved the highest yield with 509 mg/L and 4.7 g/L in shake flask fermentation and fed-batch fermentation, respectively [36]. Ji et al. achieved efficient biosynthesis of β-elemene in Yarrowia lipolytica by metabolic engineering and combinatorial strategies, yielding 784.61 mg/L and 5.08 g/L in shake flask fermentation and fed-batch fermentation, respectively [37]. Xie et al. optimized the mevalonic acid pathway and screened geranyl diphosphate synthases (GASs) from plants and microorganisms to construct a cellular factory in Saccharomyces cerevisiae for (−)-β-elemene production, with a yield of 1152 mg/L [38]. Although the yield needs to be increased, this study replaced different terpenoid synthases to produce different products, demonstrating the universality of the chassis cell PP5 in terpenoid synthesis. Furthermore, metabolomic analysis revealed significant changes in the key intermediate Mevalonic acid of the MVA pathway in the engineered strain compared to the wild type (p < 0.05), indicating a response of the MVA metabolic pathway to the engineered modification. Additionally, we observed a significant enrichment of the arginine biosynthesis pathway in the PP5 strain, suggesting that an increase in the copy numbers of IDI1 and tHMG1 may promote the flow of precursor metabolites into other pathways, which could be one of the limiting factors for yield, pointing towards future rational optimization directions.

The universal chassis cell PP5 constructed in this study can efficiently synthesize various terpenoids without the need for metabolic pathway readjustment, providing new ideas and methods for the industrial production of terpenoid compounds. However, metabolomic data also revealed the limitations of static pathway modification, as merely increasing the expression levels of key genes in the MVA pathway is difficult to overcome the regulatory constraints of the metabolic network. Future research needs to further explore the interaction mechanisms between metabolic pathways and other pathways such as arginine synthesis, and combine dynamic control strategies to break through the robustness limitations of the metabolic network, thereby further enhancing the production of terpenoid compounds and promoting the widespread application of P. pastoris cell factories in the field of terpenoid synthesis.

5. Conclusions

Here, we constructed a universal terpenoid P. pastoris cell factory by overexpressing key genes in the MVA pathway and attenuating the branch pathway. Subsequently, we used this cell factory to biosynthesize β-elemene, β-farnesene, (+)-valencene, (−)-α-bisabolol, and the yield of these terpenes increased by 10–34 times compared with the original strain. This indicates that the cell factory has great potential for efficient synthesis of terpenoids.

CRediT authorship contribution statement

Chenfan Sun: Writing – original draft, Investigation, Conceptualization. Cuifang Ye: Investigation, Conceptualization. Xiaoqian Li: Investigation. Jiabin Xu: Methodology, Investigation. Huiru Yu: Methodology, Investigation. Jucan Gao: Writing – review & editing, Investigation. Chengran Guan: Writing – review & editing, Investigation, Conceptualization. Jintao Cheng: Writing – review & editing, Conceptualization.

Declaration of competing interest

The authors declare that they have no competing interests.

Appendix A. Supplementary data

The following is the Supplementary data to this article: