Systems‐level approaches for understanding and engineering of the oleaginous cell factory Yarrowia lipolytica

Concerns about climate change and the search for renewable energy sources together with the goal of attaining sustainable product manufacturing have boosted the use of microbial platforms to produce fuels and high‐value chemicals. In this regard, Yarrowia lipolytica has been known as a promising yeast with potentials in diverse array of biotechnological applications such as being a host for different oleochemicals, organic acid, and recombinant protein production. Having a rapidly increasing number of molecular and genetic tools available, Y. lipolytica has been well studied amongst oleaginous yeasts and metabolic engineering has been used to explore its potentials. More recently, with the advancement in systems biotechnology and the implementation of mathematical modeling and high throughput omics data‐driven approaches, in‐depth understanding of cellular mechanisms of cell factories have been made possible resulting in enhanced rational strain design. In case of Y. lipolytica, these systems‐level studies and the related cutting‐edge technologies have recently been initiated which is expected to result in enabling the biotechnology sector to rationally engineer Y. lipolytica‐based cell factories with favorable production metrics. In this regard, here, we highlight the current status of systems metabolic engineering research and assess the potential of this yeast for future cell factory design development.

concentration or the presence of organic acids, high capacity to channel nutrients to commodity chemicals via the efficient flux through the tricarboxylic acid (TCA) cycle (Bilal et al., 2020), and also high potentials in biosynthesis and secretion of proteins (Soudier et al., 2019). Besides being a desirable host for bioproduction, Y. lipolytica has also been a model organism for research on the biology and physiology of dimorphism, hydrophobic substrate utilization, protein secretion, and more importantly, lipid metabolism (Beckerich et al., 1998;Beopoulos et al., 2009;Domínguez et al., 2000;Fickers et al., 2005).
Given the broad use of Y. lipolytica as an important microorganism with biological relevance and biotechnological applications, significant efforts have been made to expand the knowledge of this yeast. More recently, with the advancements in omics techniques, a quantitative systems-level insight into cellular organizations has made possible to make valuable contributions to bioprocessing (Campbell et al., 2017). Apparently, being able to analyze the entire genome as well as globally quantifying all cellular components and their interactions, does not only expand our knowledge of the biology of the system under study, but also greatly improves the quality of cell factory design. In general, there are multiple biotechnologically important metabolites produced by Y. lipolytica and the application of systems metabolic engineering is in its infancy for this yeast. Therefore, it is beneficial to provide an assessment of the state-of-the-art, to evaluate which approaches have let to promising results and where future research should be directed to. Herein, recent advances in the implementation of systems metabolic engineering in optimizing Y. lipolytica as a cell factory is being reviewed.

| Y. LIPOLYTICA AS A CELL FACTORY
The industrial application of Y. lipolytica has over half a century of history that began with the discovery of its high lipolytic and proteolytic activity.
These features enable the yeast to grow on a diverse range of lipid or protein-rich materials, where many of them can be considered as cheap substrates. The high capacity of lipase secretion was leveraged on by British Petroleum from 1950s through the 1970s, to pioneer the production of single-cell protein (SCP) from n-alkanes as substrate (Groenewald et al., 2014). SCP production based on Y. lipolytica is still available as animal feed additive but on a smaller scale (Ritala et al., 2017).
Starting from 2000s and following the development of molecular and genetic tools for engineering the yeast, research studies, as well as commercial productions, have been oriented towards leveraging the abilities of the yeast to produce more than 100 different recombinant proteins, polyalcohols, aromas, emulsifiers, and various organic acids (primarily citric acid) from industrial waste or by-products (Ledesma-Amaro & Nicaud, 2016;Miller & Alper, 2019).
In addition, the oleaginous phenotype of Y. lipolytica has been of interest as it renders the yeast a promising oleaginous cell factory that can produce oils and fatty acid-derived compounds. Two products, New Harvest™ EPA oil and Verlasso® salmon are commercial examples of this feature and contain omega-3 fatty acids for which there have been many clinical studies showing a wide range of health benefits (Xie et al., 2015).
Generally, oils and fatty acid-derived compounds from Y. lipolytica can be used as, for example, biodiesel as well as ingredients for food and cosmetic industries (Miller & Alper, 2019).
Since lipid biosynthesis is one of the most important features of Y.
lipolytica as a cell factory, a short overview of general lipid metabolism will be given (also depicted in Figure 1) while the details can be found elsewhere (Fakas, 2017;Lazar et al., 2018). Briefly, triacylglycerol (TAG) lipid species function as storage carbohydrates and can be biosynthesized either de novo or ex novo. The de novo lipid biosynthetic pathway is enhanced by metabolic imbalance upon an essential nutrient limitation in the growth media. In particular, under nitrogen starvation, the carbon flux is redirected towards TAG biosynthesis, by a mechanism where the TCA cycle enzyme isocitrate dehydrogenase is inhibited due to low levels of the allosteric activator adenosine monophosphate (AMP) that is rather broken down to yield ammonium. The ex novo pathway is based on the ability of Y. lipolytica to grow on a wide range of hydrophobic carbon sources (Beopoulos et al., 2009). Whether produced de novo or ex novo, lipids can accumulate in a dynamic storage compartment called the lipid body which mainly accommodates neutral lipids such as TAGs, while lipid precursors such as malonyl-CoA can likewise be diverted to alternative products. Metabolism of lipid biosynthesis and accumulation is well studied in Y. lipolytica, comprehensive understanding of which is vital for opting any further systems metabolic engineering approach to improve the production. It is important to note that while wild-type Y. lipolytica strains are typically inefficient in de novo lipid biosynthesis from sugars in an industrial scale (Blazeck et al., 2014), various strains have been genetically engineered to reach high lipid contents, where systems-level approaches are found promising in enhancement of production metrics and approaching the maximum theoretical yield.

| WET-LAB TOOLS TO FACILITATE SYSTEMS ANALYSIS
The availability of well-established genome modification techniques is an essential factor for optimizing microbial cell factories. In addition to wide availability of basic genetic engineering tools (i.e., host strains and markers, vectors, promoters, terminators, and replication elements) and conventional genetic modification methods, there is a continuing effort in developing modern gene/genome editing systems for Y. lipolytica that can greatly speed up the application of this yeast in systems biotechnology (comprehensively reviewed in Ganesan et al., 2019). Briefly, from 2016, several efficient CRISPR-Cas9 systems for single and multigene editing as well repression or activation of genes in Y. lipolytica are being developed (Gao et al., 2016;C. Schwartz et al., 2017C. Schwartz et al., , 2018C. M. Schwartz et al., 2016;Z. Yang et al., 2020). The gene integration efficiency of up to 55% has been achieved using a homology-independent targeted genome integration tool mediated by CRISPR/Cas9 which does not require the construction of homologous templates (Cui et al., 2021). More recently, a dual-purpose CRISPR-Cpf1 system has been suggested that is capable of simultaneous gene disruption and POORINMOHAMMAD AND KERKHOVEN | 3641 gene regulation in Y. lipolytica (Ramesh et al., 2020). Furthermore, a suite of genetic tools (EasyCloneYALI) has been generated that facilitates transformation protocols by providing a series of predesigned plasmids and oligos (Holkenbrink et al., 2018). There is still a significant opportunity to further improve CRISPR-based systems in Y. lipolytica in terms of gene integration efficiency (similar to gene disruption efficiency of about 90%), multigene editing efficiency or the need for a high-throughput screening technology after genome editing. Although the application of the CRISPR-Cas9 usage and its innovative strategies in the model yeast, Saccharomyces cerevisiae, shed light and laid the foundation of CRISPR-Cas9 implementation in other nonconventional yeasts technology, its optimizations need species-specific studies. However, there are several general factors that have been proven in this reference yeast to positively affect the efficiency of CRISPR-Cas9 genome modifications amongst which using weak to a medium-strength promoter for expressing CAS9, a priority of using natural RNAP III promoters, careful choice of the gRNA binding site and taking advantage of in silico tools for sgRNAs F I G U R E 1 Overview of lipid metabolism in Yarrowia lipolytica solid arrows: chemical conversions and transport reactions, dashed arrows: multiple chemical conversion steps, dotted lines and arrows: representing N-limitation consequences. AMP, adenosine monophosphate; Cit, citrate; DAG, diacylglycerol; DHAP, dihydroxyacetone phosphate; F6P, fructose 6-phosphate; FA, fatty acid; FBP, fructose 1,6-bisphosphate; Fum, fumarate; G3P, glycerol 3-phosphate; G6P, glucose 6-phosphate; GA3P, glyceraldehyde 3-phosphate; Icit, isocitrate; IMP, inosine monophosphate; LPA, lysophosphatidic acid; Mal, malate; Mal-CoA, malonyl coenzyme A; NH4, ammonium; OAA, oxaloacetate; PA, phosphatidicacid; Pyr, pyruvate; Suc, succinate; TAG, triacyglycerol design to decrease off-target effect can be named (Cai et al., 2019;Raschmanová et al., 2018). When optimized, these systems provide a suitable platform to be coupled with the systems-level tools to further promote the systems metabolic engineering process.
Meanwhile, and in parallel with studies to optimize CRISPR-Cas9 systems for Y. lipolytica, other genome modification strategies are being developed mainly due to the little information on CRISPR-Cas9 related molecular tools in this yeast. For instance, a Cre/lox-site-specific recombination system using only a single selection marker has been developed for targeted, repeated, and marker-less gene integration in Y.
lipolytica with above 90% efficiency. Unlike similar previous methods based on Cre/lox-site-specific recombination, their genetic tool was also proved effective for various genes, loci, and strains which suggests a promising application of Cre/lox system and strengthens its capabilities for genome modification in Y. lipolytica (Zhou et al., 2021). Moreover, as a study to overcome the challenge with building complex pathways and multiple gene integration in Y. lipolytica, Z. Guo et al. (2020) have designed a completely synthetic Y. lipolytica-specific artificial chromosome (ylAC) and successfully used it to introduce a complete cellobiose and xylose co-utilization pathway introduced to the yeast in a single step within less than one week of wet lab experimentation. The method thus is much faster and efficient in comparison to conventional multistep procedures and can be applied for multigene integration in Y. lipolytica.
Based on existing efforts to create orthogonal systems where the synthetic biology parts of one yeast have been applied to another (Patra et al., 2021), it is expected that in near future, the synthetic biology tools and techniques will be applied interchangeably among different species.
In all, recent advances have provided Y. lipolytica with a plethora of genetic engineering and genome editing techniques (Holkenbrink et al., 2018;Madzak, 2018), widening the perspectives for systems Under the moniker of metabolic engineering, modifications of biochemical pathways have been used for decades to achieve desired phenotypes such as overproduction of a metabolite in cell factories, either using random mutagenesis or other conventional molecular genetic engineering tools. However, only after the development of the methods and tools in the discipline of systems biology has it become practical to approach the metabolic engineering challenges from a systems perspective. The development of an optimal strain via systems metabolic engineering can be performed through two major approaches: omics data-driven (top-down) and biomolecular model-driven (bottom-up). While the bottom-up approach starts with the reconstruction of an accurate mathematical model of a specific biomolecular subsystem (i.e., metabolic, gene regulatory and signaling networks, or combinations thereof) using existing biological knowledge; the top-down approach begins with extracting information from relevant omics data sets (e.g., metabolomics, proteomics, transcriptomics, and fluxomics). Figure 2 illustrates these strategies as well as the classic metabolic engineering.
During the past decade, there have been multiple studies on biomolecular model analysis and omics data generation that aimed to improve the understanding and production of many valuable metabolites in Y. lipolytica. Genome-scale metabolic models (GEMs) are the most commonly studied biomolecular networks, the popularity of which arises not only from their innate nature which directly considers the analysis of the metabolite production phenotype, but also from the availability of powerful constraint-based modeling methods.
As GEMs can provide a global view of cellular metabolism, they are influential tools in systems metabolic engineering. Currently, there have been six GEMs published for Y. lipolytica by different research groups, where each model includes improvements over previously published models. All six GEMs have been used for evaluating metabolic engineering strategies with a major focus on over-producing lipids, for example, triacylglycerols (TAGs) and to a lesser extent for the increased dicarboxylic acids production (Kavšček et al., 2015;Kerkhoven et al., 2016;Loira et al., 2012;Mishra et al., 2018;Pan & Hua, 2012;Wei et al., 2017).
Since obtaining high-quality omics data has become progressively easier, interest in applying data-driven approaches for acquiring systems-level knowledge of Y. lipolytica have increased (Ledesma-Amaro & Nicaud, 2016). While the availability of complete genome sequences reveal its metabolic pathways and functional potential, various transcriptomics and a few proteomics analyses have provided useful information about the dynamics of biological processes, such as the transition from the growth phase to the lipid accumulation phase as well as its regulatory mechanisms (Morin et al., 2011). There have also been a few metabolomics and lipidomics analyses in Y. lipolytica, which further provide information on the metabolic pathways under study by incorporating the changes in metabolite concentrations (Zhao et al., 2015).
As individual omics analyses (e.g., transcriptomics or metabolomics) are not always able to fully elucidate the behaviors of a cellular system, combining different omics data (e.g., transcriptomics and metabolomics) may result in a more precise picture of the biological system being studied. Moreover, integration of various types of omics data following a bottom-up approach can boost the predictive capability of the model (Gu et al., 2019). Such combinatory approaches have been applied in a few studies in metabolic engineering of Y. lipolytica.
Generally, examples of systems-level understanding as well as metabolic engineering efforts in Y. lipolytica powered by systems POORINMOHAMMAD AND KERKHOVEN | 3643 biology knowledge and tools will be discussed below in relative sections. Furthermore, Table 1 summarizes the systems-level optimizations of lipid-derived as well as nonlipid products in Y. lipolytica.

| Lipids and fatty acid-derived products
Various studies have been performed to gain a systems-level understanding of lipid metabolism with focus on three main questions: the main source of NADPH that lipid accumulation demands; gene regulation that coincides with lipid accumulation; and transcription factor networks that are underlying the expressional changes during lipid accumulation as will be detailed below. In many cases, such studies have not only expanded our knowledge of various aspects of lipid metabolism, but also yielded proposed targets for genetic engineering for improved lipid production.
Lipid biosynthesis has a high NADPH demand that could theoretically be provided through various metabolic pathways. Based on a comparison of flux distribution obtained by 13 C metabolic flux analysis in two strains of Y. lipolytica, one of which was engineered to produce lipids at roughly twice the yield of the wild strain, Wasylenko et al. (2015) concluded that the oxidative pentose phosphate pathway (PPP) is the almost exclusive source of lipogenic NADPH when Y. lipolytica is grown on glucose. However, when cultured on acetate, another flux distribution analysis by N. Liu et al. (2016) showed the importance of gluconeogenesis on lipogenic NADPH synthesis.
To reveal if lipid accumulation is transcriptionally regulated, the first transcriptomics analysis of Y. lipolytica during its metabolic shift from growth to lipid accumulation indeed showed significant difference between the expression profile of actively dividing and lipid accumulating cells (Morin et al., 2011). However, the genes directly governing fatty acid synthesis were not transcriptionally regulated, -1.6-fold increase in productivity comparing to Saccharomyces cerevisiae.
Overexpression of POT1 M.  Citric acid

Multi-omics analysis
Predicted overexpression and knock-out candidates. -4-fold increase in yield.
TAG GEM reconstruction, FBA and gene knockout/overexpression simulations 55.5% increase in production rate. rather than the transcriptional regulation of lipid metabolism itself.
By integrating proteome, phosphoproteome and metabolome data, Pomraning et al. (2016) suggested posttranslational modification of key lipogenesis enzymes as the governing regulatory mechanism.
Specifically, ATP-citrate lyase and acetyl-CoA carboxylase (ACC) were not differentially expressed but their level of phosphorylation increased when cells were grown in media with low nitrogen availability, suggesting that this contributes to increased lipid accumulation, which would be orthologous to the phosphorylation sites of S.
cerevisiae Acc1 that have been leveraged to enhance lipogenesis in that species (Shi et al., 2014). In Y. lipolytica, later studies showed that genes related to branched-chain amino acid metabolism were downregulated upon nitrogen limitation, while alternative nitrogen supply pathways such as protein turnover and autophagy genes were upregulated among which an important factor that affects lipid synthesis was reported to be the metabolism of leucine (Kerkhoven et al., 2016. These findings were similar the results from a multiomics study of another oleaginous yeast, Rhodosporidium toruloides (Zhu et al., 2012). . It is thus noteworthy that effective systems biology approaches such as multiomics studies (e.g., integration of transcriptome and metabolome data) integrated with in silico metabolic model simulations could be used to analyze regulatory dynamics (Hubbard et al., 2018), and from this transpired that lipid accumulation is not merely an effect of isocitrate dehydrogenase inhibition during nitrogen starvation.
Beyond employing systems-level approaches to gain understanding on lipid metabolism, it has been used to find systems-level solutions that favor the oleaginous phenotype in terms of improving its yield, productivity, and titer as well as broadening the substrate range. Nonetheless, most studies that have improved lipid production can be classified as classical metabolic engineering, via over- production in the productive phase giving the hypothesis that reduced aeration rate, which is also favored in industry, might induce lipid accumulation. Moreover, to eliminate citrate excretion, a reduced glycolytic flux was calculated while maintaining lipid synthesis rate. Combination of these findings were used to design a fermentation strategy which resulted in 80% increase in the lipid content of biomass and more than fourfold yield improvement (Kavšček et al., 2015). In a later model-based approach, various gene knockouts and overexpression targets for improved TAG biosynthesis were predicted by analyzing a comprehensive GEM reconstructed by

| Specific oleochemicals
While classic and systems metabolic engineering efforts continue to improve lipid production in Y. lipolytica by maximizing flux through TAG biosynthesis, production of specific valuable oleochemicals are being studied in this yeast mostly via synthetic pathway construction.
Many commercially interesting oleochemicals, for example, medium-chain lipids, are less common in natural resources and are therefore ideal candidates for systems and synthetic biologybased production due to their high value (Blazeck et al., 2014).
These oleochemicals can be used as nutraceuticals or pharmaceuticals, such as polyunsaturated fatty acids ( Due to its oleaginous phenotype, Y. lipolytica accommodates a highly active TCA cycle. However, since biosynthesis of OAs is highly linked to cellular energy metabolism, reaching efficient production metrics will require more complex engineering approaches. This renders systems metabolic engineering profitable tools to optimize OA production. Nonetheless, also for this group of products has classical metabolic engineering through rational genetic modifications and optimization of fermentation condition shown to enhance the productivity (reviewed in Finogenova et al., 2005). Meanwhile, efforts to elucidate the underlying systems-level production mechanisms are crucial to optimize cell metabolism favorable for OA production. In terms of citric acid (CA) production, Sabra et al. (2017) have focused on transcriptome and fluxome data in a citrate producing strain of Y.
lipolytica. While controlling the oxygen supply on mixed substrates proved instrumental to improve the citrate production titer more than three fold, multiomics data analysis have provided pentose phosphate pathway and glyoxylate cycle genes as targets for metabolic engineering of CA (Sabra et al., 2017). In another study, after reporting Y. lipolytica as the best CA producer among more than 40 yeast strains, Kamzolova et al. (2017), studied the underlying overproduction mechanism using growth limitation assessment and enzyme activity measurements which helped in fermentation design for enhanced production (85 g/L).
A more recent screening also showed that the production of CA strongly strain dependent, therefore, screening for overproducers and studying the underlying molecular mechanisms is of great importance (Carsanba et al., 2019). Noteworthy is that due to the relations between lipid and CA metabolism, knowledge of the regulation surrounding lipid accumulation as well as nitrogen limitation can have considerable relevance for CA production studies.
The production of succinic acid (SA) in Y. lipolytica has also undergone metabolic engineering efforts (Abdel-Mawgoud et al., 2018). Unlike SA-producing bacteria, which are among the most successful producers to date, Y. lipolytica can tolerate low pH and thrives under acidic conditions. This simplifies downstream processes as the bacterial fermentation broth is rather kept near-neutral pH and requires an additional acidification step for SA purification which causes byproduct formation (Ahn et al., 2016). Although high titers have been achieved (160 g/L) in an engineered strain lacking active succinate dehydrogenase (SDH) to prevent oxidation of succinic acid to fumaric acid, low yield and productivity hinders the implementation of Y. lipolytica for industrial production. SDH deficient mutants of Y. lipolytica are growth defective in cultures using glucose source (Yuzbashev et al., 2010). To circumvent this problem Yuzbashev et al. have isolated an improved SA producer from combined induced mutagenesis and metabolic evolution, which could more efficiently utilize glucose (Yuzbashev et al., 2016). 13 C flux analysis on SDH deficient mutants was performed to investigate the contribution of various metabolic pathways in SA formation (Yuzbashev et al., 2016), however, to fully reveal the molecular mechanism by which adaptive evolutionary approaches have resulted in a beneficial phenotype would require genome sequence analysis of the evolved and unevolved strains. A further problem hindering improvement of SA production is the accumulation of the by-product acetate (Cui et al., 2017). omics-based studies . Similar approaches could be applied to Y. lipolytica to yield industrially feasible SA production levels in this yeast.
α-ketoglutarate (α-KG) is another OA for which Y. lipolytica seems a promising production host. The main strategy being used to improve its production is still based on fermentation design as well as a few genetic engineering in the direct biosynthetic enzymes (H. Guo et al., 2016). Therefore, while systems-level insight is still desired to investigate key metabolic processes influenced and regulated by the known factors that positively affect α-KG production, there are a few studies based on omics data. In a study to improve α-KG production, Guo et al., have assessed the roles of specific transporters (H. Guo et al., 2015).
To this end, they have identified six endogenous putative transporter genes using a comparative genomics approach and by expressing them in an α-KG-producing wild-type strain, the production of this OA has improved while reduced pyruvic acid (a byproduct) concentration was observed. Their study showed whole genome analysis for transporters can be helpful in improving the accumulation of specific organic acids. Another comparative genomics method with the aim of investigating the regulatory mechanisms controlling α-KG production, showed the positive role of genes associated with the regulation of mitochondrial biogenesis and energy metabolism in α-KG production, whereas they have concluded genes related to transformation between keto acids and amino acids as reducers of the accumulation of α-KG (Zeng et al., 2016). As another strategy to accumulate α-KG, is to take advantage of its thiamineauxotrophic trait. Y. lipolytica is unable to biosynthesize thiamine which is the co-factor for α-ketoglutarate dehydrogenase which converts α-KG to succinyl-CoA. Therefore, a culture medium lacking this vitamin will likely favor the accumulation of α-KG when carbon source is in excess. However, enhanced accumulation of α-KG by disruption of genes directly affecting TCA cycle will undoubtedly cause cellular growth and survival defects circumventing which urges deeper molecular knowledge. In this regard, similar omics data-based findings of the study by Walker et al. on the effects of thiamine deficiency on cellular metabolism of thiamine auxotroph Y. lipolytica which was discussed in lipid production improvement strategies, can be helpful (Walker et al., 2020).
Beyond native OAs, Y. lipolytica has been used to produce the nonnative itaconic acid (IA). IA is naturally produced by several Aspergillus species and is a replacement for petroleum derived products (Robert & Friebel, 2016). Due to shortcomings with the native producer in terms of poor growth and shear stress sensitivity, its production is studied in alternative hosts such as Y. lipolytica. IA is synthesized by cis-aconitic acid decarboxylase (CAD) (Kanamasa et al., 2008) and IA production in Y. lipolytica has been established by heterologous expression of CAD  resulting in IA yields of 0.06 g/(g glucose) at natural low pH. Although these studies can serve as a promising stepping-stone toward enhancing IA production in Y. lipolytica, to establish this yeast as an industrially competitive IA producer, further efforts are required to improve production metrics. Even while bottom-up systems-level approaches have been used for IA production in other microorganisms (Harder et al., 2016), there is still no report on systems metabolic engineering in Y. lipolytica.  Yang et al., 2015). This information could be leveraged on for future reverse engineering of Y. lipolytica strains for polyols biosynthesis.
Production of many pharmaceuticals and nutraceuticals employing Y. lipolytica as a host has been studied mostly using classic metabolic engineering . Oleanolic acid is pentacyclic triterpenoid with antioxidation activity and its heterologous production have been studied in Y. lipolytica. Based on intracellular metabolome data, a direct precursor of the triterpenoid was found to be accumulating. Guided by the analysis of differential transcriptional levels of genes, it transpired that enhancement of electron transfer by fusion expression of P450 enzyme and the last step of the biosynthetic pathway could be used to improve substrate conversion efficiency by 28% .

| CONCLUSIONS AND PERSPECTIVES
For decades, Y. lipolytica has been known as a promising biotechnological chassis not only for its high oleaginous flux, but also for the production of a variety of native and nonnative chemicals. However,

| Combination of top-down and bottom-up approaches
As classical metabolic engineering results typically show, the overexpression of a gene target and in vivo production improvement are rarely linearly correlated, which is believed to be caused by various layers of cellular control. To this end, a combination of both the topdown and bottom-up approaches seems promising to elucidate the effect of genetic interventions. As GEMs allow for incorporating gene, transcript, protein and reaction relationships, they are capable to be efficiently used as scaffolds for omics data integration to refine their reconstruction as well as defining accurate constrains on flux predictions. This will consequently result in not only improved predictions of cellular phenotype or product formations but also will help gaining insight into regulation of metabolism in response to environmental or genetic perturbations. In this regard, advancements in metabolomics data generation and analysis in yeasts (Sailwal et al., 2020)

| Advances in computational approaches
Regardless of the incomplete genome annotation, the current knowledge can be used to improve the existing computational models. A robust and accurate metabolic model facilitates successful application but relies on the availability of sufficient experimental data. While the Y. lipolytica GEM has been improved , this should continue akin to the continuous curation and incorporation of new experimental data in the S. cerevisiae GEM.
Specifically, the S. cerevisiae GEM has been advanced by integrating enzymatic constraints as well as protein 3D structures, now rendering this model capable of accurately predicting how single nucleotide variations translate to phenotypic traits (Lu et al., 2019).
Such strategies are expected to also be of value for nonconventional yeasts such as Y. lipolytica.
The lack of kinetic parameters has precluded wider use of kinetic models of Y. lipolytica, but this impediment could be countered through ensemble modeling (EM). Instead of settling on specific parameter values, in EM samples of kinetic parameters are repeatedly drawn from an allowable distribution to generate an ensemble of kinetic models that best describe the system under study (Tran et al., 2008). EM has been used in systems metabolic engineering to among others identify targets for improved ethanol overproduction in Synechocystis sp. PCC 6803 (Nishiguchi et al., 2019;Nishiguchi et al., 2020); overproduction of L-lysine in Escherichia coli (Contador et al., 2009); as well as improved ethanol production in Clostridium autoethanogenum (Greene et al., 2019).
Applying this approach in Y. lipolytica might be a valuable strategy to further the understanding of its metabolic dynamics and identify novel strategies for improved production, although it should be noted that selection of ensemble models depends on the availability of (dynamic) experimental data from diverse conditions, which is currently relatively limited for Y. lipolytica. In particular dynamic metabolomics measurements would contribute valuable criteria to narrow down the solution space described by ensemble models.
Alternatively, existing GEMs could be expanded to allow for integration of further omics data, thereby yielding more accurate predictions of gene modification targets. A classical GEM can, for example, be expanded by introducing enzymatic constraints, where the usage and efficiency of enzymes is explicitly considered (Sánchez et al., 2017). Such enzyme constrained models (ec-models) do not only yield flux distributions that physiologically more relevant, but they also enable the integration of proteomics data to result in condition-specific models. However, ec-models require enzyme efficiencies as k cat values that are sparsely available for Y. lipolytica.
While k cat values could be estimated from related organisms, it is anticipated that machine-learning will be able to give more accurate predictions (Heckmann et al., 2018). GEMs could also be further expanded towards genome-scale models of metabolism and expression (ME-models), enabling the integration of gene expression and protein synthesis data in GEMs. This approach has been used in, for example, E. coli (O′brien et al., 2013) and S. cerevisiae (Oftadeh et al., 2021), but has not been widely applied to non-model microorganisms primarily due to the lack of species-specific detailed knowledge of, for example, protein synthesis.
In sum, the increasing availability of systems-level tools and techniques such as genome modification tools, metabolic models and state-of-the-art "omics" studies, are able to enhance our understanding of unique Y. lipolytica features and provide guidance for its improvement. However, it remains imperative that a deeper knowledge of regulation and metabolism in Y. lipolytica will help to advance it towards a versatile industrial cell-factory.
Largely, this can be accomplished by orthogonal application of systems-level and synthetic biology tools that have been very successful to advance model organisms such as S. cerevisiae into promising cell factories, while its unique characteristics can give Y. lipolytica an advantage. Moreover, as Y. lipolytica can be regarded as a promising host for the biosynthesis of various products, it is expected that elucidation of systems-level behaviors for one product will also positively impact on this yeast as a future host for other products.

ACKNOWLEDGMENT
We acknowledge funding from the Novo Nordisk Foundation (grant no. NNF10CC1016517) and the Research Council for Environment, Agricultural Sciences and Spatial Planning (Formas, grant no. 2018-00597).

CONFLICT OF INTERESTS
The authors declare that there are no conflict of interests.

AUTHOR CONTRIBUTIONS
Naghmeh Poorinmohammad conceived, wrote, and edited the manuscript. Eduard J. Kerkhoven critically revised and edited the manuscript.

DATA AVAILABILITY STATEMENT
Data sharing not applicable to this article as no data sets were generated or analyzed during the current study ORCID Eduard J. Kerkhoven https://orcid.org/0000-0002-3593-5792