Engineering of Yarrowia lipolytica for terpenoid production

Terpenoids are a group of chemicals of great importance for human health and prosperity. Terpenoids can be used for human and animal nutrition, treating diseases, enhancing agricultural output, biofuels, fragrances, cosmetics, and flavouring. However, due to the rapid depletion of global natural resources and manufacturing practices relying on unsustainable petrochemical synthesis, there is a need for economic alternatives to supply the world's demand for these essential chemicals. Microbial biosynthesis offers the means to develop scalable and sustainable bioprocesses for terpenoid production. In particular, the non-conventional yeast Yarrowia lipolytica demonstrates excellent potential as a chassis for terpenoid production due to its amenability to industrial production scale-up, genetic engineering, and high accumulation of terpenoid precursors. This review aims to illustrate the scientific progress in developing Y. lipolytica terpenoid cell factories, focusing on metabolic engineering approaches for strain improvement and cultivation optimization.


Introduction
Terpenoids or isoprenoids represent a vast and structurally diverse class of small molecules involved in specialized and general metabolism across the entire kingdom of life (Ashour et al., 2010;Gershenzon and Dudareva, 2007). A unifying feature for all terpenoids is their genesis from 5 carbon (C 5 ) units. The classification of terpenoids is based on the number of C 5 -units comprising the basic scaffolds: hemiterpenes (C 5 ), monoterpenes (C 10 ), sesquiterpenes (C 15 ), diterpenes (C 20 ), sesterterpenes (C 25 ), triterpenes (C 30 ), sesquarterpenes (C 35 ), and tetraterpenes or carotenoids (C 40 ) (Ashour et al., 2010;Sato, 2013). In addition, the isoprene scaffolds may undergo re-arrangement and other modifications, creating numerous diverse structures. Many terpenoids hold value as potent pharma-and nutraceuticals, biofuels, or as flavor, colour, or cosmetic agents (Tetali, 2018). But harvesting terpenoids from natural resources or manufacturing by chemical synthesis is often economically disadvantageous or unsustainable (Idris and Mohd Nadzir, 2021;Pateraki et al., 2017). Therefore, considerable attention has been applied to the producing terpenoids via engineered microbes. The common chassis organisms are Escherichia coli and Saccharomyces cerevisiae since they are easy to cultivate and engineer, and much is known about their biology (Moser and Pichler, 2019). Recently, the oleaginous yeast Yarrowia lipolytica has been intensely researched for its terpenoid production capabilities. This yeast offers benefits like broad substrate utilization, sequenced genomes, and available genetic engineering toolkits, with several strains having achieved GRAS-status (Darvishi et al., 2018;Dujon et al., 2004;Holkenbrink et al., 2018;Magnan et al., 2016;Turck et al., 2019). This review aims to showcase both established and nascent strategies for improving terpenoid production in Y. lipolytica.

Mevalonate pathway engineering
Engineering of the mevalonate pathway for terpenoid overproduction often involves the upregulation of MVA-pathway genes. In particular, the reduction of HMG-CoA to mevalonate catalyzed by HMGp is a popular target for upregulation (Ashour et al., 2010;Polakowski et al., 1998). This tendency is highlighted by the preponderance of the surveyed literature, which uses HMGp-upregulation as a terpenoid overproduction strategy (Table 1). The model yeast S. cerevisiae contains two HMG-genes encoding ScHMG1p and ScHMG2p, respectively (Burg and Espenshade, 2011). Both enzymes are regulated on the transcriptional level, but ScHMG1p is also regulated during translation, while ScHMG2p is post-translationally regulated by ubiquitination and endoplasmic reticulum-associated degradation. The negative feedback regulation of ScHMG1p can be overcome by truncating the membrane-anchored N-terminal (tScHMG1p), resulting in higher HMGp-activity (Polakowski et al., 1998). This led to the hypothesis that the same principle applies to Y. lipolytica HMGp, but several studies show that non-truncated HMGp outperforms tHMGp for terpenoid production in Y. lipolytica (Cao et al., 2016;Huang et al., 2018;Jia et al., 2019;Kildegaard et al., 2017;H. Liu et al., 2020;S. C. Liu et al., 2020), while a few other studies do not point to large differences (D. Li et al., 2019). While some studies utilized heterologous NADH-dependent HMGp enzymes based on a presumed large abundance of NADH in Y. lipolytica, they did not compare these enzymes to overexpression of NADPH-dependent HMGp versions (Guo et al., 2021;Y. Liu et al., 2019). Therefore, it remains unclear whether shifting HMGp dependency from NAPDH to NADH is a superior strategy for terpenoid production in Y. lipolytica.
Besides HMGp upregulation, some studies demonstrate the benefits of overexpressing single or multiple MVA-genes for terpenoid production in Y. lipolytica. However, these modifications are typically cooverexpressed with HMGp, due to the long track record of this strategy for boosting terpenoid production. Two studies found that overexpression of ERG12p increased limonene yield (mass of product/dry biomass) 6-fold and seemingly also α-pinene titers (product concentration in cultivation broth) (Cao et al., 2016;Wei et al., 2021b). Conversely, another study showed no significant amorphadiene titers increase when ERG12p was overexpressed (Marsafari and Xu, 2020).
Another option for improving terpenoid production is to limit the flux of intermediates towards undesired side-products. In the case of non-triterpenoid or -sterol products, this may be achieved by reducing flux towards squalene. Replacement or truncation of the native SQS promoter was found to increase β-carotene titers by 2-2.5-fold (Kildegaard et al., 2017). Interestingly, some of these modifications also increased squalene titers. Contrarily, SQS-promoter replacement did not positively affect β-farnesene titer and yield (Arnesen et al., 2020). Promoter replacement of SQS with the glycerol repressible P alk -promoter resulted in decreased squalene yield and growth during cultivation on glycerol while increasing lycopene, but not total carotenoid, yield (Gao et al., 2017b). Truncation of the native SQS promoter slightly increased (− )-α-bisabolol titers and slightly decreased squalene titers (Yirong .
In summary, engineering of the MVA-pathway is a well-established strategy for increasing terpenoid production in Y. lipolytica, although the effects do vary between the specific terpenoids and strains.
Another approach to increasing the supply of IPP in the cell is to enable conversion of isoprenol to IPP via two enzymatic steps. The cell culture is then co-fed with isoprenol, which provides plentiful amounts of IPP. Expression of the Arabidopsis thaliana isopentenyl phosphate kinase (AtIPKp) and S. cerevisiae choline kinase (ScCHKp) (so called Isopentenol Utilization Pathway IUP) increased the IPP/DMAPP-pool 15.7fold in the PO1f background and lycopene yield in Y. lipolytica (Luo et al., 2020). Furthermore, introducing the IUP-pathway can enhance terpenoid biosynthesis complementary to MVA-pathway engineering. In a β-carotene producing strain overexpressing tHMG, ERG12, IDI, ERG20, and XdcrtE, the introduction of IUP increased β-carotene titers 23% (Ma et al., 2022).

Cofactor and acetyl-CoA engineering
Reducing HMG-CoA into mevalonate by HMGp in yeast requires NADPH as a cofactor (Polakowski et al., 1998). Therefore, a few studies have attempted to improve terpenoid production by increasing NADPH availability. The mannitol dehydrogenase makes NADPH from NADP+ during the conversion of mannitol into fructose (H. ). Yet, the overexpression of the mannitol dehydrogenase MnDH2p increased squalene titer slightly but did not benefit the yield (H. . Another study found that triterpenoid titers increased somewhat in some strain backgrounds when malic enzymes, Mortierella alpina EMTp or Rhodotorula toruloides Rtmep, generate NADPH by decarboxylating malate into pyruvate and CO 2, were expressed (Jin et al., 2019). For campesterol production, overexpression of a malic enzyme decreased yield by 43% (Zhang et al., 2017). Overexpression of the putative NAD + kinase POS5p failed to increase abscisic acid production (Arnesen et al., 2022). Therefore, increasing NADPH availability currently seems to only offer small potential benefits for terpenoid production in Y. lipolytica, but more research on this topic is warranted.
Y. lipolytica is presumed to have high acetyl-CoA abundance relative to other common microbial chassis, which may have been inferred by the medium-to-high lipid accumulation obtained by some Y. lipolytica strains during particular cultivation conditions (Beopoulos et al., 2008;Kerkhoven et al., 2016). Some studies indicate that Y. lipolytica possesses higher acetyl-CoA flux and abundance than S. cerevisiae under similar cultivation conditions, but more evidence is needed to justify such generalized statements (Christen and Sauer, 2011;Dahlin et al., 2019). Nevertheless, strategies that increase acetyl-CoA flux and their effect on terpenoid production in Y. lipolytica have been studied (Fig. 2). The ATP-citrate lyase (ACLp) forms acetyl-CoA, oxaloacetate, and ADP + Pi from citrate and CoA in the cytosol (Blazeck et al., 2014;Dulermo et al., 2015). ACLp consists of two subunits, ACL1p and ACL2p, in Y. lipolytica. Overexpression of ACL1p with co-expression of a mutated version of the Salmonella enterica acetyl-CoA synthetase (SeACS L641P p), that forms acetyl-CoA, AMP, and diphosphate from acetate, CoA, and ATP, improved squalene yield by 3.2-fold (Huang et al., 2018). Individual expression of either SeACS L641P p or ACL1p did not significantly enhance squalene accumulation. Overexpression of SeACS L641P p and ACL1p also improved β-farnesene titer and yield (Arnesen et al., 2020). Furthermore, overexpression of ACL1p increased acetyl-CoA production in the PO1f background, while ACL2p-overexpression decreased acetyl-CoA accumulation (Huang et al., 2018). Curiously, the overexpression of ACL2p, but not ACL1p, increased the triterpenoid titers in some strain backgrounds, while both ACL1p and ACL2p overexpression increased lycopene production independently (Jin et al., 2019;Zhang et al., 2019). ACL2p overexpression also improved squalene titer, but decreased the yield due to a corresponding increase in biomass (H. . ACLp overexpression increased campesterol yield 1.3-fold (Zhang et al., 2017). Overexpression of the adenine monophosphate deaminase (AMPDp) increased lycopene titer and yield . AMPDp inhibit the isocitrate dehydrogenase, which increase citrate and, by extension, acetyl-CoA levels.
Alternatively, acetyl-CoA accumulation can be affected by decreasing flux towards lipid biogenesis or increasing rates of fatty acid degradation. Knocking out the diacylglycerol acyltransferase genes Δdga1 and Δdga2 reduced the lipid content from 26.3% to 8.7% and increased β-farnesene titers by 56.32% (T. . Knocking out Δdga1 or Δdga2 separately seemingly provided lesser reductions in lipid content and lesser increases in β-farnesene titers. The peroxisome acyl-CoA oxidase 2 (POX2p) overexpression increased campesterol yield 1.3-fold (Zhang et al., 2017). The growth media contained sunflower oil with abundant long-chain fatty acids, likely providing a substrate for the enhanced β-oxidation pathway. Interestingly, overexpression of POX1,4, 5,6p under the same conditions did not affect campesterol yield, while overexpression of POX3p, the multifunctional-oxidation protein (MFEp), and the peroxisomal oxoacyl thiolase (POTp) decreased campesterol yield. MFEp catalyzes hydration and dehydrogenation reactions during β-oxidation, while POT1p catalyzes the thiolytic cleavage of β-ketoacyl-CoA into acetyl-CoA and a shortened acyl-CoA molecule (Beopoulos et al., 2009;Hanko et al., 2018;Smith et al., 2000). Interestingly, overexpression of MFE1p and POT1p increased (− )-α-bisabolol titers (Yirong Ma et al., 2021). Overexpression of POT1p also enhanced α-humulene titer and yield in a strain where the MVA-and α-humulene biosynthetic pathways were targeted to the peroxisomes, while overexpression of MFE1p or the peroxisomal biogenesis factor 10 (PEX10p) did not provide substantial benefits (Guo et al., 2021). MFE1p and POT1p overexpression also increased triterpenoid titers in some strain backgrounds (Jin et al., 2019). Strain-dependent increases in triterpenoid titers were also found when the long-chain fatty acid transporter (PXA1p) or triacylglycerol lipase (TGL3p) were overexpressed. Alternatively to genetic engineering, adding the lipid biosynthesis blocking compound cerulenin to the growth media enhanced the amorphadiene titer by 231.13% with a strain-specific dependency (Marsafari and Xu, 2020). Lastly, expression of Vitreoscilla hemoglobin (Vhbp) was demonstrated to improve α-farnesene production by 12.7%, likely by enhancing oxygen delivery to the cells (Y. .

Modulation of lipid storage
Some hydrophobic terpenoids like β-carotene are stored in the lipid bodies of Y. lipolytica (Larroude et al., 2018). Therefore, it may be beneficial to increase sequestration of such lipophilic products by expanding the lipid pool. Larroude et al. developed an obese Y. lipolytica platform strain by overexpression of DGA2, the glyceraldehyde-3-phosphate dehydrogenase gene (GPD1), and deletion of Δpox1-6 and Δtgl4 (Larroude et al., 2018). This obese strain accumulated 3.6-fold more lipids and the β-carotene titer and yield were boosted by 1.9-and 2.6-fold compared to the wt. Overexpression of DGA1 increased squalene yield by 2.9-fold and seemingly increased lipid accumulation (Wei et al., 2021a). Likewise, deletion of Δpex10 increased squalene production 9-fold. Furthermore, the co-deletion of Δure2, which encodes a protein putatively involved in oxidative stress responses and nitrogen metabolism, and Δpex10 improved squalene production more than the individual deletions. Overexpression of DGA1 and the acetyl-CoA carboxylase gene (ACC1) increased the lipid content by 2.97-or 1.76-fold depending on the strain background, increased the accumulation of GPP, FPP, and GGPP, and seemingly benefitting lycopene production (Luo et al., 2020). Furthermore, overexpression of either GPD1p, DGA1p, or DGA2p increased lycopene titer and yield . Presumably, engineering strategies for increased fatty acid accumulation may be advantageous specifically for intracellularly accumulated products. Indeed, deletion of Δpex10 decreased the titer and yield by >50% of the extracellularly accumulated α-farnesene, possibly due to reduced acetyl-CoA availability (S. C. .
The overexpression of the Δ9-fatty acid desaturase gene (OLE1) combined with the deletion of the diacylglycerol kinase (Δdgk1) and phosphatidic acid phosphatase (Δpah1) genes increased the lupeol titer 4.7-fold while skewing the fatty acid profile from dominantly saturated to unsaturated (J.-L. . These gene edits also increased the ratio of extracellularly to intracellularly accumulated lupeol during two-phase cultivations with an organic overlay and were shown to benefit the production of αand β-amyrin, and longifolene-type sesquiterpenes. These results were primarily attributed to increased plasma membrane permeability following lipid unsaturation allowing increased efflux into the extracellular phase. A β-carotene producing strain was used to construct a mutant library by NHEJ-mediated random mutagenesis . This approach identified four new gene targets that improved β-carotene when a leucine prototrophy conferring gene-cassette (LEU2) was integrated with proximity to or into the gene. The largest increase in β-carotene production was achieved by NHEJ-mediated insertion of LEU2 into the NDH80-locus. Reverse engineering by deletion of ΔNDT80 in the parental strain increased β-carotene production by 62%. Furthermore, the lipid content increased, while the ergosterol content decreased in the ΔNDT80-strain compared to the parental strain, which may simultaneously expand the β-carotene storage capabilities and redirect MVA-pathway flux towards carotenoid biosynthesis.

Compartmentalization, morphology, and transport engineering
Several studies have applied pathway compartmentalization for terpenoid production in yeast (Dusséaux et al., 2020;G. S. Liu et al., 2020;. Targeting the peroxisomes for α-humulene biogenesis was achieved by fusing a peroxisomal targeting signal (PTSp) peptide to the MVApathway enzymes and the Aquilaria crassna α-humulene synthase (AcHS PTS p) (Guo et al., 2021). The report demonstrated that only complete, but not partial, peroxisomal re-construction of the MVA-pathway improved α-humulene titers (50-fold) together with AcHS PTS p-expression. Further increases in α-humulene production were achieved by β-oxidation engineering (see previous chapter) and overexpression of the peroxisomal adenine nucleotide transporter gene (ANT1). This latter strategy increased the α-humulene titer by 11%, likely due to increased transport of cytoplasmic ATP into the peroxisomes. While the cited work resulted in a highly productive α-humulene strain (3.2 g/L), it is unclear whether the peroxisomal targeting strategy was superior to conventional cytoplasmic expression. However, another report does provide some evidence that compartmentalization can be beneficial for terpenoid production in Y. lipolytica. Increasing the copy number of a non-targeted astaxanthin biosynthetic fusion enzyme gene, consisting of the Paracoccus sp. β-carotene ketolase fused to the N-terminal of the Haematococcus pluvialis β-carotene hydroxylase (PsCrtW-HpCrtZ), did not affect astaxanthin production (Yongshuo ). Yet, the astaxanthin titer increased 1.62-to 1.84-fold when PsCrtW-HpCrtZp was targeted to either the lipid bodies, ER, or peroxisomes by fusion with the appropriate signal peptides. When PsCrtW-HpCrtZp was simultaneously targeted to all three compartments, the astaxanthin titer increased 4.8-fold. Expression of either the Escherichia coli resistance-nodulation-cell division family efflux pump (EcAcrBp) or the Grosmania clavigera ATP-binding cassette transporter (GcABCG1p) increased the titers of various bisabolene isomers during two-phase organic cultivation (Zhao et al., 2021).
Deleting genes involved in the single cells to hyphae transition can improve terpenoid production (M. . Deleting the protein kinase gene (ΔCLA4) involved in the transition from yeast cells to filaments increased the β-carotene yield by 81%, although pseudohyphae could still be observed. Likewise, the deletion of the transcription factor gene (ΔMHY1) abolished hyphae formation and increased the β-carotene yield by 45%. Overexpression of endosomal sorting complex subunit DID2p increased β-carotene yield by 260%. In addition, it increased the expression of specific genes involved in β-carotene biosynthesis, the pentose phosphate pathway, the tricarboxylic acid pathway, and the hexokinase encoding gene (Hxk) .
While the above examples provide exciting investigations into the topics of compartmentalization, morphology engineering, and cellular transport, these research areas remain relatively unexplored in Y. lipolytica. It would be exciting for future research to expand these areas, which show great promise.

Enzyme engineering
Some papers demonstrate the utility of engineered enzymes for terpenoid production in Y. lipolytica. A common pathway improvement strategy is fusing proteins catalyzing consecutive biocatalytic steps or linking enzymes with supporting protein partners.
Protein tagging can improve stability and solubility. For example, expression of ERG12p N-terminally tagged with maltose-binding protein (MBP) increased α-pinene 1.84-fold compared to overexpression of the untagged ERG12p (Wei et al., 2021b).
The effects of protein fusion and removal of the plastidial targeting sequences by N-terminal truncation of the A. thaliana copalyl diphosphate synthase (AtCPSp), ent-kaurene synthase (AtKSp), ent-kaurene oxidase (AtKOp) on GA-biosynthesis was investigated (Kildegaard et al., 2021). The highest GA-titers were achieved by expressing the non-fused truncated enzymes.
Structure-guided protein engineering of the R domain/lycopene cyclase from McCarRPp successfully removed substrate inhibition by lycopene (Ma et al., 2022). Structure modeling and a Position-Specific Scoring Matrix (PSSM) based on the lycopene cyclase domain and its homologs provided an evolutionary basis for selecting amino acid residues for substitution. A sampling of the variance space allowed the isolation of three McCarRPp variants, V175W, T31R-F92W, and Y27R, with improved β-carotene to lycopene ratios. The Y27R-variant had a 98% selectivity compared to 18% for the WT for β-carotene.

Alternative substrate utilization
The broad substrate utilization of Y. lipolytica makes it an excellent chassis for turning unconventional low-value or waste carbon sources into high-value products. Although glucose is commonly used as a substrate for terpenoid production in Y. lipolytica (Table 1) in the lab, for industrial processes, it can be advantageous to use cheaper or more abundant feedstock. It was demonstrated that oleic acid could be used as a carbon source for α-farnesene production with the final titer of 10.2 g/ L achieved during fed-batch cultivation (Y. . Oleic acid seemed to provide slightly better titers compared to glucose. α-farnesene was also produced in the 1-2 g/L range when cultivated in shake flasks with soybean oil, either fresh or from as waste cooking oil (WCO), olive, palm, glycerol trioleate, or rapeseed oil. Likewise, substituting oleic acid for glucose at carbon equivalent concentrations increased α-humulene titers by 18.5% for a strain with the MVA-and α-humulene pathway targeted to the peroxisomes (Guo et al., 2021). Using safflower oil with an oleic acid content of 77.0% as a carbon source during fed-batch cultivation resulted in 167 mg/L astaxanthin with 48% accumulated extracellularly (N. . The production of astaxanthin per cmol of carbon when using safflower oil as a carbon source was similar to that of glucose. Using sunflower seed oil at as carbon source resulted in higher campesterol titers than glucose at carbon equivalent concentrations (Du et al., 2016). It was also observed that sunflower oil increased lipid accumulation and lipid body formation compared to glycerol or glucose. Limonene titers were higher with WCO than glycerol or glucose at carbon equivalent concentrations . Conversely, bisabolene titers were lower when using WCO than glucose at carbon equivalent concentrations, despite higher biomass accumulation when using WCO (Zhao et al., 2021). The addition of Mg 2+ to WCO-based rich media increased the bisabolene titers, which was also found for limonene production (Pang et al., 2019). It was demonstrated that WCO, sunflower, rapeseed, or soybean oil could be used for α-pinene production, with the latter noted to increase α-pinene titer 80% compared to glucose (Wei et al., 2021b). The expression of the xylose metabolism genes encoding the Scheffersomyces stipites xylitol reductase (SsXR), S. stipites xylitol dehydrogenase (SsXDH), and native xylulokinase (XK) enabled α-pinene production and highly enhanced growth on xylose-containing rich media. But the expression of the xylose metabolism genes did not benefit α-pinene production when detoxified lignocellulosic hydrolysate, with major constituents being glucose, xylose, and acetate, was used as carbon source. Another report demonstrated that expressing the genes encoding SsXR, SsXDH, and S. stipites xylulokinase (SsXK) or XK enabled growth on xylose-based rich media (Yao et al., 2020). The expression of XK yielding better growth than SsXK. A mixture of 8 g/L glucose and 32 g/L xylose provided higher limonene titers compared to 40 g/L of either pure glucose or xylose for the engineered strain. Likewise, it was demonstrated improved growth on xylose based media by expressing SsXR K270R/N272D , SsXDH, and XK . The ability of the strain to consume xylose was improved by an adaptation period on xylose containing media, after which further engineering enabled protopanaxadiol (PPD) production. Overexpression of the transketolase (TKLp) and transaldolase (TALp), which connect the xylose degradation pathway with the pentose phosphate pathway, improved the PPD titer and biomass accumulation on xylose based media. The final engineered strain exhibited the best PPD titers on pure xylose as a carbon source compared to glucose or mixed sugar compositions.
Using glycerol instead of glucose, soybean, corn oil, or oleic acid at the same concentrations resulted in the highest betulinic acid titer . Interestingly, the expression of specific MVA-pathway genes and acetyl-CoA accumulation increased during cultivation with glycerol compared to glucose. Another report found that glycerol improved limonene yield compared to glucose, citrate, fructose, maltose, sucrose, mannose, or galactose at similar concentrations (Cheng et al., 2019). Adding auxiliary carbon sources like citrate, pyruvate, malate, but not acetate increased limonene titers further.
Similarly, α-farnesene titer and yield during fed-batch bioreactor cultivation also improved when glycerol was fed instead of glucose, although the batch phase used glucose as a carbon source (S. C. . Contrarily, glycerol decreased limonene yield or α-pinene titer and yield compared to glucose at similar concentrations (Cao et al., 2016;Wei et al., 2021b).
Using citrate as a carbon source resulted in the highest linalool yield and titer compared to fructose, glucose, or glycerol, with the latter being the second-best carbon source (Cao et al., 2017). The addition of pyruvate further increased linalool titer and yield. Likewise, the addition of pyruvic acid also increased limonene yield (Cao et al., 2016). The addition of citrate or acetate increased squalene yield for a strain expressing ACL1p and SeACS L641P p, which respectively can convert these substrates to acetyl-CoA (Huang et al., 2018). The addition of 4 g/L citrate to YPD medium enhanced α-pinene yield and titer, which was not found for the addition of pyruvate, acetate, or malate in the range of 0-4 g/L, or lower concentrations of citrate (Wei et al., 2021b).
The ability of Y. lipolytica to utilize glucose was enhanced by overexpression of the hexokinase (HXKp), which catalyzes the phosphorylation of glucose as the initial step in the glycolysis pathway . HXKp overexpression enhanced β-carotene yield by 98% and led to faster glucose consumption.
Using palmitic acid as an addition to glucose-based rich media enhanced lycopene titers, lipid accumulation, and the accumulation of MVA-pathway metabolites like FPP and GGPP (Luo et al., 2020). Experiments using C 13 -labeled glucose showed that >90% of C16:0 and C18:0 lipids were unlabeled and therefore most likely derived from the exogenous lipid source, while ~75% acetyl-CoA was derived from the labeled glucose. These findings are supported by (Ma et al., 2022), that demonstrated abundant unlabeled IPP/DMAPP and GGPP during cultivation with labeled glucose and unlabeled stearic acid. Therefore, the addition of fatty acids could enhance intracellular terpenoid storage and contribute to the acetyl-CoA pool via β-oxidation.
In summary, carbon sources like glycerol or WCO can be used as substrates for terpenoid production in Y. lipolytica. Furthermore, the research shows that the substrate acceptance of Y. lipolytica can be widened by engineering. Various auxiliary carbon sources also show promise; but their utility may be decided by their price-to-benefit ratio. Therefore, terpenoid production by cultivating Y. lipolytica with alternative and cheap carbon sources represents a promising avenue for developing economic and sustainable bioprocesses.

Outlook
The current research has demonstrated numerous varied strategies for improving terpenoid production in Y. lipolytica. However, while some strategies achieve different results across studies, it is important to consider that the contexts vary greatly; factors like strain background, cultivation conditions, properties of the specific terpenoid product and pathway, and previous strain modifications potentially affect the outcomes. Nevertheless, direct MVA-pathway engineering has been shown in multiple studies to improve terpenoid titers several-fold. Cofactor engineering by redirecting flux towards cytosolic acetyl-CoA has also shown some utility. Conversely, increasing lipid accumulation for intracellular storage is highly effective for producing some terpenoids, although likely dependent on the product's propensity to accumulate intra-or extracellularly. Emerging evidence demonstrates new strategies like morphology or compartmentalization engineering that increase terpenoid production in Y. lipolytica. Still, more research is needed to identify in which contexts these strategies are effective. Besides modifying the native metabolism, the careful selection and engineering of heterologous pathway and supportive enzymes by protein fusion, tagging, or modifying select amino residues can lead to several-fold increased terpenoid titers in Y. lipolytica. Media optimization is a valuable addition to genetic engineering, and alternative substrates and additives have been used to modestly increase terpenoid production in some reports. In summary, there are ample well-described and nascent strategies for improving terpenoid production in Y. lipolytica, and highly productive Y. lipolytica cell factories for terpenoids with short heterologous biosynthetic pathways have been developed. Indeed, the implementation of long heterologous biosynthetic pathways in Y. lipolytica remains a challenge with a good example being the production of gibberellin plant hormones (Kildegaard et al., 2021). Before the high-level production of complex terpenoids becomes possible, greater knowledge about the expression of individual enzymes and balancing long non-native pathways in Y. lipolytica is needed. Computational approaches utilizing genome-scale modelling, omics, and machine learning have yet to substantially impact terpenoid production in Y. lipolytica. These tools may become more relevant as they develop and the knowledge of Y. lipolytica metabolism expands.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.