Integrated Multi-Omics Analyses Reveal the Molecular Basis of Tryptophol Over-Accumulation in Saccharomyces Cerevisiae

the effects of and (Phe) on tested, and the results that can signicantly facilitate accumulation, but ≥ reduced the effect depended on performed. 1) both the Ehrlich and tryptamine-dependent involved 2) Trp increased by enhancing which the


Conclusions
In this study, TOL production of S. cerevisiae was signi cantly improved, and our integrated multi-omics analyses have provided insights into the understanding of TOL over-accumulation, which will be useful for future production of TOL using metabolic engineering strategies.

Background
Tryptophol (TOL, i.e., indole-3-ethanol), with a molecular formula of C 10 H 11 NO and molecular weight of 161.2, is an aromatic alcohol widely found in wine, beer and other beverages [1]. Similar to most indole compounds, TOL has multiple biological activities. For instance, TOL is able to induce apoptosis in human leukaemia U937 cells without affecting normal lymphocytes [2], and can attenuate pathogen-induced in ammation-related responses of the cytokines TNFα and IFNγ [3]. In addition to the potential bioactivity for humans, TOL is a growth promoting factor for roots and leaves of plants [4]; it can inhibit viral replication of white spot syndrome of the shrimp Marsupenaeus japonicus [5], and it possesses antimicrobial activity against the food contaminants Campylobacter jejuni [6] and Salmonella enterica [7]. Therefore, research on TOL biosynthesis and regulation has attracted widespread attention.
TOL, similar to two other aromatic alcohols (2-phenylethanol and tyrosol), is biosynthesized through the well-known Ehrlich pathway through three enzymatic steps: transamination, decarboxylation and reduction. More speci cally, tryptophan (Trp) is rst transaminated by aromatic aminotransferases I and II, which are encoded by the genes aro8 and aro9, respectively [8]. The formed Indole-3-pyruvate (IPA) is then catalysed by an aromatic decarboxylase (ARO10) and three pyruvate decarboxylases (PDC1, PDC5 and PDC6), and alcohol dehydrogenases (ADHs) encoded by adh1−5 and sfa1 are proposed to be involved in the nal step of the Ehrlich pathway [8]. TOL can also be biosynthesized de novo from glucose [9]. In this process, phosphoenolpyruvate (PEP) from glycolysis and erythrose 4-phosphate (E4P) from pentose phosphate pathway enter shikimate pathway to form chorismate, which is then converted to Trp via ve enzymatic reactions and further catabolized to TOL through the Ehrlich pathway.
Many microorganisms including plant-bene cial bacteria, lamentous fungi and yeast can produce TOL [8,[10][11][12]. Among these microbes, yeast has the strongest TOL production capacity, with a maximum yield of approximately 580 mg/L, which was produced by Zygosaccharomyces priorianus [13]. Saccharomyces cerevisiae can also produce large amounts of TOL, with an output of up to 478 mg/L when Trp was the only nitrogen source [8]. The fact that Trp has notable effects on TOL production has also been reported in some other yeast species. For example, TOL in Debaryomyces hansenii was synthesized only when Trp was present in the medium [14], and in Candida albicans, the production of TOL increased by 2.5 times after the addition of Trp [15]. Moreover, the in uence of other nitrogen sources, not only Trp, on TOL production has long been recognized. González et al. found that nitrogen limitation strongly promoted the production of aromatic alcohols [10], whereas high ammonium conditions dramatically reduced them [16].
To clarify the molecular mechanisms behind the nitrogen effects, expression levels of genes in the Ehrlich pathway were investigated. It has been reported that aro9 and aro10 gene transcription was induced by the presence of Trp, phenylalanine (Phe), and tyrosine in the growth medium, and this induction required the transcriptional activator ARO80 and resulted in TOL, 2-phenylethanol, and tyrosol accumulation, respectively [16,17]. Importantly, the biosynthetic TOL activates ARO80 and, consequently, the expression of aro9 and aro10, resulting in a positive feedback loop [16]. Similarly, nitrogen-poor conditions promote the expression of aro9, aro10 and pdc6, whereas high ammonia or abundant nitrogen represses them [16]. Additionally, some genes involved in cofactor metabolism, transcriptional regulation, and amino acid transportation were reported to in uence the biosynthesis of aromatic alcohols. Dickinson et al. found that the titre of TOL obviously decreased in a dihydrolipoyl dehydrogenase gene (lpd1) mutant, meaning that LPD1 was required for Trp catabolism to TOL [8]. Similarly, THI3, a sensor of intracellular thiamine pyrophosphate (TPP), was required along with pyruvate decarboxylases for the alternative activity [18]. In addition to ARO80, CAT8 and MIG1 are two well-documented transcription factors. It has been reported that cat8 overexpression or mig1 deletion increased the transcription of aro9 and aro10, followed by an enhanced formation of 2-phenylethanol [19]. In addition, amino acid permease genes agp1, agp2, tat1 and tat2 are transcriptionally induced by extracellular aromatic amino acids, and GATA factors such as GLN3 and GAT1 regulate the transcription of aro9 and aro10 for nitrogen source and aromatic amino acid utilization [20,21]. To fully understand the biosynthetic process of TOL, it should be considered that genes other than those of the Ehrlich pathway and de novo biosynthetic pathway are used in an auxiliary pathway.
Recently, multi-omics analyses, including genomics, transcriptomics and metabolomics, have been used to gain an in-depth understanding of the Trp-dependent biosynthesis of indole-3-acetic acid (IAA), an auxin sharing a common direct precursor, indole-3-acetaldehyde (IAD), with TOL, in some plantassociated microbes [11]. In this work, to systematically clarify the molecular basis of TOL overaccumulation, a S. cerevisiae strain KMLY1-2 (hereafter referred to as KMLY1-2) with high TOL production ability was screened out, and its yield dynamics over varying Trp concentrations were monitored. Moreover, the differential effect of Trp and Phe on TOL biosynthesis was investigated, and the molecular mechanism behind TOL overproduction was elucidated by multi-omics analyses. Our current ndings may help to further improve the yield of TOL through metabolic engineering strategies and lay a foundation for industrial production of TOL with high e ciency.

Results And Discussion
TM1 is a suitable medium for TOL production To select a culture medium for TOL production, KMLY1-2 was cultured in transformation medium (TM) containing different carbon and nitrogen sources (Table S1). No TOL was produced in TM1 and TM2, and the highest yield (only 1.66 mg/L) was detected in TM4 (Fig. S1a). After an addition of 1 g/L Trp, the yield of TOL in TM1 and TM2, two media without nitrogen, was signi cantly increased, with TM1 resulting in the highest yield of 236.68 mg/L. In addition, TOL production decreased with the increasing nitrogen content in TM3-1Trp−TM5-1Trp, with the yield of TOL in TM5-1Trp being approximately 4.93% of that in TM1-1Trp (Fig. S1b). The data indicated that, except for Trp, high nitrogen was not conducive to TOL production, which is highly consistent with a previous report by Chen and Fink [16]. Therefore, TM1, due to the absence of nitrogen and high-yield TOL, was chosen for subsequent experiments.
TOL production is dependent on cell density and the expression of key genes Yeast growth and TOL production were measured over time. As shown in Fig. 1a, 0−24 and 24−42 h were identi ed as the exponential phase and stationary phase, respectively, according to the time−OD 600 curve.
Additionally, the TOL content sharply increased from 0−24 h and stayed constant from 24−42 h. Among them, the TOL concentration at 24 h was 211.46 mg/L. The highly consistent data of TOL production and cell growth indicated that TOL production is closely related to cell density. To clarify whether this process requires TOL biosynthetic genes, four key genes (aro8, aro9, aro10 and aro80) were selected and their expression levels were analysed. The expression of aro8, aro10 and aro80 gradually increased from 0 h to 18 h, while that of aro9 peaked at 12 h and slightly decreased at 18 h. In addition, these genes showed stable expression levels after 24 h (Fig. 1b). The pro le of key gene expression is consistent, to a certain extent, with the pattern of TOL production and growth, indicating that the phenomenon of TOL yield dependence on cell density requires the expression of aromatic aminotransferases (ARO8 and ARO9), a decarboxylase (ARO10) and a transcription factor (ARO80), which is partly congruent with reports by Chen and Fink [16]. Therefore, 24 h is a critical time point and is considered to be the fermentation time of KMLY1-2 in the subsequent experiments.

TOL production is dependent on Trp and Phe concentrations
To explore the effects of Trp and Phe on TOL biosynthesis, TOL production was monitored after KMLY1-2 was incubated in TM1 with different concentrations of Trp and Phe. As shown in Fig. 2a, TOL yield increased proportionally as the Trp concentration increased but was more or less at (231.02−266.31 mg/L) when ≥0.6 g/L Trp was supplied to the medium. Meanwhile, increasing amounts of residual Trp accumulated accordingly. These data indicated that the ability of KMLY1-2 to convert Trp into TOL was saturated from 0.6 g/L Trp. However, Phe did not affect TOL production when it was the sole nitrogen source in the medium (data not shown), which was ascribed to Phe being a direct precursor for 2phenylethanol biosynthesis [9]. In addition, TOL content was signi cantly reduced when KMLY1-2 was cultured in TM1 containing Trp and Phe as nitrogen sources, and the reduction was strongly dependent on the Phe concentration (Fig. 2b). This may be attributed to nitrogen catabolite repression (NCR) in which high ammonia restricts TOL production by repressing the transcript level of genes in the Ehrlich pathway [10,16]. However, the mechanism of Phe affecting the conversion of Trp to TOL is still unclear. In contrast with the results of extracellular TOL, a signi cantly different effect of 0.6 and 1.5 Trp on intracellular TOL production was observed (Fig. 2c). This discrepancy was mainly due to differences in biomass, as the intracellular TOL yield was calculated by weight normalization.

Metabolomic pro les
The chemical pro le of the KMLY1-2 endometabolome was generated by liquid chromatography/mass spectrometry, and a total of 4473 metabolites with de nite names were identi ed. Principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA) showed that the sample replicates were tightly clustered and the samples from different media were clearly separated ( Fig. S2), suggesting that the metabolomic data were highly reproducible. Of these metabolites, 1011, 1201, 281 and 694 differential metabolites (DMs), including 30, 30, 14 and 11 compounds of Trp and its derivatives, were identi ed in TM vs. TM-06T, TM vs. TM-15T, TM-06T vs. TM-15T, and TM-06T vs. TM-TP, respectively (Table S2). The Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis of DMs showed that biosynthesis of antibiotics (ko01130) was the only shared enriched pathway in TM vs. TM-06T, TM vs. TM-15T and TM-06T vs. TM-TP (Table S2), suggesting that the addition of Trp and Phe to the medium may affect the biosynthesis of some antibiotics [22]. In addition, the abundance of DMs in the Ehrlich pathway and its bypass of Trp metabolism was analysed. As shown in Fig. 3, intracellular L-Trp abundance, as expected, increased with the increase in exogenous Trp concentration, which was not different between TM-06T and TM-TP, indicating that Phe had no effect on Trp transportation. A similar abundance trend for intracellular Trp was observed in IPA, IAD, tryptamine and IAA, suggesting that both the Ehrlich pathway and the tryptamine-dependent pathway (Trp→tryptamine→IAD) are involved in TOL biosynthesis of S. cerevisiae, as reported for the fungus Neurospora crassa [23], although the functional genes in the tryptamine pathway have not been identi ed in S. cerevisiae. In addition, the ranges of the increases in IPA and IAD abundance from TM to TM-06T were respectively smaller than those of L-Trp and IPA, while the decline ratios of IPA and IAD abundance from TM-06T to TM-TP were larger than that of L-Trp (Fig. 3a). The former indicates that transamination and decarboxylation are two rate-limiting steps in TOL biosynthesis, while the latter shows that these steps are susceptible to Phe, which may result in the NCR phenomenon of TOL biosynthesis due to Phe or phenylpyruvate (PPA) competing with Trp or IPA for the active centre of transaminase and decarboxylase, respectively. For the analysis of TOL abundance, the peak value was identi ed in TM-06T, which is highly consistent with the intracellular highperformance liquid chromatography (HPLC) data in Fig. 2c. The relatively low abundance in TM-15T may be ascribed to the reversible conversion of TOL to IAD by alcohol dehydrogenase, followed by convertion to IAA by aldehyde dehydrogenase. Different from the effect of Phe on the abovementioned metabolites, Phe promoted indolelactate accumulation (Fig. 3b), and the reason should be further studied.
General features of the KMLY1-2 genome PacBio sequencing generated an assembled nuclear genome containing 31 contigs with 11.79 Mbp (∼113 × coverage) and 38.1% GC content, similar to previously reported S. cerevisiae strains [24]. A total of 5539 protein-coding sequences (CDSs) with an average length of 1476 bp were predicted, representing 69.31% of the genome. In addition, 3 rRNA and 315 tRNA were identi ed in the genome. Among all the predicted proteins, 5537, 5527, 2733 and 3795 CDSs were allocated to the non-redundant protein (Nr), swiss-prot, eukaryotic orthologous groups of proteins (KOG), and KEGG databases, respectively, based on sequence homologies, yielding 2376 shared annotated CDSs in total. According to the KEGG annotation, the pathways for glycolysis, citrate cycle, pentose phosphate cycle, amino acids biosynthesis, and purine and pyrimidine metabolism were complete. Furthermore, the candidate genes involved in TOL biosynthesis in KMLY1-2 and S288C (a model S. cerevisiae strain) were subjected to a comparative analysis. As shown in Table 1, a complete Ehrlich pathway, containing seven aminotransferases, four decarboxylases, and seven dehydrogenases, which showed su cient homology (98.66−100%, except for ADH3), were identi ed. Genomic blast analysis showed that the adh3 gene in KMLY1-2 was highly homologous to four consecutive genes in S288C (adh3, YMR084W, YMR085W and seg1), indicating that it is a tetrafunctional polypeptide. Additionally, nine proteins (ARO1−ARO4 and TRP1−TRP5) with 97.93−100% identity were responsible for Trp biosynthesis from E4P and PEP via chorismate, meaning that both KMLY1-2 and S288C have the complete de novo biosynthetic pathway of TOL. Similarly, genes in the auxiliary pathway showed high homology (83.8−99.89%) between KMLY1-2 and S288C, suggesting that these genes are also conserved in yeast. Nevertheless, the TOL yield of KMLY1-2 was approximately 1.57 times than that of S288C when they were cultivated in TM1 supplemented with 1.5 g/L Trp, which may be attributed to the different regulatory mechanisms during TOL biosynthesis.
Transcriptome sequencing and analysis of differentially expressed genes (DEGs) With the aim of dissecting the molecular mechanism of yeast TOL biosynthesis and regulation, transcriptome sequencing and analysis of TM, TM-06T, TM-15T and TM-TP were performed. Approximately 36623732−89449834 clean reads (99.82−99.93% of raw reads) were generated in 12 RNAseq libraries, and 89.74−96.06% of the total reads mapped to the KMLY1-2 reference genome, among which 45.69−81.24% were CDS mapped reads (Table S3) (Fig. S3b), which may be closely associated with the fact that Trp facilitated TOL accumulation and Phe reduced TOL production. Using the short time series expression miner (STEM) algorithms, these shared DEGs can be clustered in 17 pro les (Fig. S3c−e). The transcript levels of DEGs in pro le 1 were in a pattern of "increase-keep-decrease" from TM to TM-06T and TM-15T to TM-TP, which perfectly matched the data of extracellular TOL yield in Fig. 2a and b. Further analyses showed that one aminotransferase (HIS5), two decarboxylases (ARO10 and PDC5), a chorismate synthase (ARO2), and the transcriptional activator ARO80, which have been proven to participate in aromatic alcohols biosynthesis [8,9,20], were contained in pro le 1 (Table S4). Additionally, eight and two genes speculated to participate in the Ehrlich, de novo, and auxiliary biosynthetic pathways of TOL were identi ed in pro le 9 and pro le 14, respectively (Table S4). Considering the fact that genes with similar expression patterns might be functionally correlated [25], it is reasonable to assume that genes in pro les 1 and 9, pro les 1 and 14 may be highly related to TOL production and the NCR phenomenon, respectively. In fact, eight and ve TPP metabolism genes, coding for essential cofactors for PDC1, PDC5 and ARO10 [26], were respectively identi ed in pro les 1 and 9. Additionally, an aldehyde dehydrogenase, which functions in the conversion of IAD to IAA [11] and showed a relatively low expression level in TM-TP, gives an alternative reason why Phe decreased IAA abundance (Fig. 3b) and has been identi ed in pro le 14 (Table S4). These data further con rmed that genes in pro les 1 and 9 were closely related to aromatic alcohol biosynthesis and regulation, and genes in pro le 14 were responsible for NCR in S. cerevisiae.

Expression pro le analyses of the DEGs involved in TOL biosynthesis
As shown in Fig. 4, when Trp was used as the sole nitrogen source, four of ve aminotransferases, two of four PPA or pyruvate decarboxylases, and two of three alcohol dehydrogenases in the Ehrlich pathway showed an upward trend, indicating that these enzymes played important roles in the biosynthesis of TOL; while the transcript levels of these transaminase and decarboxylase genes decreased when Trp and Phe were present in the same medium, which was similar to the fact that abundant nitrogen downregulated aro9 and aro10 gene expression as reported by Chen and Fink [16]. An alternative, but non-exclusive, explanation for NCR can be given based on the expression patterns of aro9 and aro10. Phe, Trp, and TOL (a degradation metabolite of Trp) can upregulate the expression of aro9 and aro10 [16]; hence, the promoting effect of Trp on aro9 and aro10 transcripts was stronger than that of Phe. When Phe and Trp are present simultaneously, they will compete for metabolic enzymes in the Ehrlich pathway of S. cerevisiae, and Phe is a preferentially used amino acid [27], causing the pathway of Trp to IAD to be blocked, meaning that the expression of aro9 and aro10 is only regulated by Phe. Therefore, the gene expression level in TM-TP is lower than that in TM-06T and TM-15T. For the de novo biosynthesis pathway, the transcription levels of six DEGs rst increased and then decreased sharply, with TM-06T as the turning point, which indicated that these genes tended to make cells synthesize more Trp at low concentrations of extracellular Trp, while the expression of these genes was suppressed by feedback in the presence of high concentrations of amino acids. In addition, the expression of gene 3726 (aro1) was more susceptible to exogenous Trp, and its expression level was inhibited once the exogenous environment contained Trp. For the auxiliary pathway, Trp promoted the expression abundance of three amino acid transporters, but there was no difference between TM-06T and TM-TP, which perfectly matched the metabolomics data in Fig. 3a, indicating again that Phe and Trp do not compete with the cell transport system. In addition, the expression levels of lpd1 (1892) and aro80 (4004) showed a trend consistent with the production of extracellular TOL, suggesting that they did play an important role in the biosynthesis of TOL in S. cerevisiae. However, the transcript levels of mig1 (5390), cat8 (3322) and gln3 (4745) decreased with the increase in nitrogen concentration, indicating that they were negatively correlated with the biosynthesis of TOL. This was partly the same as and partly contrary to the results reported by Wang et al. [19], and the reasons need to be further determined.

Integrated metabolomics and transcriptomics analyses
To fully understand the molecular mechanism of TOL overproduction in S. cerevisiae, the metabolism of amino acids, especially Trp, and the central carbon metabolism (glycolysis, pentose phosphate pathway, and citrate cycle) containing DMs and/or DEGs were summarized and described in Fig. 5. As expected, the abundance of metabolites and expression levels of most genes in the Ehrlich pathway in TM-06T and TM-15T were signi cantly increased compared with TM. For instance, the contents of TOL, IAD and IPA in TM-06T and TM-15T were 3.97−253.69 times more than those in TM. Consistently, the transcript levels of aro9, pdc5, aro10, adh2 and adh5 increased 2.19−376.46 times (Table S5). The results indicated that the addition of Trp increased TOL biosynthesis by enhancing the Ehrlich pathway, and genes with large changes, such as aro9 (353.15 to 376.46 folds), pdc5 (30.2 to 49.16 folds), and aro10 (205.74 to 222.96 folds), may have made important contributions. However, some metabolites and genes in the Ehrlich pathway showed a decreased trend after Phe was added to TM-06T (i.e., sample TM-TP), in which the content of TOL and the expression levels of his5, aat1, aro10, and pdc5 in TM-TP were 34.26% and 19.63−43.81% of those in TM-06T, respectively (Table S5). The results suggested that Phe addition weakened the Ehrlich pathway of TOL biosynthesis, which may be attributed to the inhibition of Trp to TOL by Phe competition because Phe is a preferred nitrogen source in S. cerevisiae.
In addition, the abundance of most metabolites in other branches of Trp metabolism in TM-06T or TM-15T was, as expected, signi cantly higher than that in TM. For example, a 2.74−34.76-fold increase was identi ed for indole-3-acetonitrile, 5-hydroxy-tryptophan and N-formylkynurenine (Table S5). Most strikingly, except for gene bna2 (0177), which showed 5.83 and 6.11 times higher transcript levels in TM-06T and TM-15T, respectively, compared with those in TM, the transcript levels of other related genes mostly decreased to different degrees (Table S5). The somewhat inconsistent results between metabolomic and transcriptomic data has often been reported, which might be related to the complex post-transcriptional mechanisms after gene transcription [28,29].
For other amino acids, compared with TM, 10 amino acids, including serine, arginine and others displayed 2.05-74.19-folds increases in TM-06T or TM-15T. Moreover, alanine, valine and tyrosine increased their abundance by 3.24-, 2.56-and 5.06-fold, respectively, in the comparison of TM-06T vs. TM-TP (Table S5). These increases may be due to the addition of Trp and Phe providing cells with more energy and precursors, which promotes the biosynthesis of other amino acids and consequently results in cell growth. The fact that KMLY1-2 biomass increased signi cantly after Trp and Phe were added to TM1 (Fig. S4) further supports this speculation. In this context, the rapid growth of yeast cells will undoubtedly consume more carbon sources. Indeed, compared with TM, the abundance of many intermediate metabolites in the glycolysis, the pentose phosphate pathway and the citrate cycle was signi cantly reduced in TM-06T or TM-15T, and the expression levels of corresponding genes also showed a decreasing trend (Fig. 5). The lack of signi cant changes in glucose content may be attributed to hexokinase (1820), an important regulatory enzyme in central carbon metabolism [30], whose transcription level decreased signi cantly in TM-06T and TM-15T (Table S5), thus limiting the e cient utilization of glucose.

Conclusions
In summary, we designed a suitable transformation medium for TOL yield evaluation, and found that TOL production was dependent on cell density and the expression of key genes of S. cerevisiae. To improve TOL output, we added Trp and Phe to TM1 and found that TOL production increased proportionally as the exogenous Trp concentration increased, while Phe attenuated the stimulating effect of Trp. In addition, the effect of Trp on TOL was saturated from ≥0.6 g/L supplemental Trp, and 231.02−266.31 mg/L TOL was obtained in this situation. We also performed a multi-omics analysis to understand how yeast cells over-accumulated TOL. Our data revealed that the Ehrlich pathway was the main pathway for yeast TOL biosynthesis, in which the steps of transamination and decarboxylation played key roles in the biosynthesis of TOL and in the phenomenon of nitrogen catabolite repression conferred by Phe. In addition, TOL over-accumulation was inseparable from the help of the auxiliary pathway and many other genes. In short, our ndings provide a rich genetic resource for the subsequent studies on the biosynthesis of Trp metabolites in S. cerevisiae.

Materials And Methods
Chemicals, yeast strain and media L-Trp, L-Phe, TOL and 2-chlorophenylalanine were purchased from Sigma-Aldrich (St. Louis, MO), and chromatographic grade methanol and acetonitrile were obtained from Sangon Biotech Co., Ltd.

HPLCdetection of TOL production
The seed culture of KMLY1-2 (10 6 CFU/mL) was inoculated (1%, v/v) in 50 mL MRS medium and incubated at 35 °C for 12 h in an agitating incubator (150 rpm). Cells were harvested by centrifugation at 5000 rpm for 10 min and re-suspended in 5 mL sterile ddH 2 O after washing twice with sterile ddH 2 O. The cell suspension was inoculated (1%, v/v) in 100 mL TM1−5 and cultivated with a shaking speed of 150 rpm at 35 °C for 24 h. Cell-free supernatant (CFS) from 1 mL culture was prepared by centrifugation (12000 rpm for 10 min; 4 °C) and sterile ltration using a 0.45 μm lter (Millipore, Billerica, MA), and used for TOL content determination using an HPLC system (Agilent Technologies, Santa Clara, CA). Brie y, aliquots of 10 μL were loaded into an Acclaim Explosives E2 column (4.6 × 250 mm, 5 μm, 120 Å; Thermo Scienti c, Waltham, MA). Isogradient elution was performed using methanol/H 2 O solution (70:30, v/v) for 20 min at a temperature of 25 °C and a ow rate of 0.5 mL/min. TOL were monitored at 210 nm, and their concentrations were determined by integrating the calibration curves obtained from the standards. Each sample was performed at least in triplicate.
Time course analysis of cell growth, TOL production, and expression of key genes KMLY1-2 was pre-cultured in MRS medium and re-suspended in sterile ddH 2 O as described above. The cell suspension was inoculated (1%, v/v) in 100 mL TM1 supplemented with 0.5 g/L Trp and incubated for 42 h. OD 600 , representing cell growth, was monitored every 6 h. Meanwhile, the CFS and cell pellets were collected and used for TOL content measurement and quantitative PCR (qPCR) assay, respectively. Total RNA extraction, cDNA synthesis, and qPCR were performed according to our previous study [31]. The primers are listed in Table S6, and differences in expression of genes aro8, aro9, aro10 and aro80 were calculated according to the 2 −ΔΔCT method [32] using the actin gene (act1) as the reference.
TOL production under Trp and Phe conditions KMLY1-2 was pre-treated as described above and incubated in TM1 supplemented with 0−2.5 g/L Trp. To evaluate the effects of both Trp and Phe on TOL production, KMLY1-2 was cultured in TM1 supplemented with 0.6−1.5 g/L Trp and 0−2.25 g/L Phe. After incubation for 24 h, the CFS was prepared as described above and analysed by HPLC to quantify the produced TOL and the residual Trp. Accordingly, cell pellets from TM1 and TM1 supplemented with 0.6 g/L Trp, 1.5 g/L Trp, or 0.6 g/L Trp and 1.75 g/L Phe were collected to generate the samples of TM, TM-06T, TM-15T, and TM-TP, respectively. To test the intracellular TOL concentration, all cell pellets were weighed, ground with liquid nitrogen, and resuspended with 0.95 mL sterile ddH 2 O. After centrifugation and ltration treatment, the TOL level in the ltrate was re ected by the value of TOL/weight (μg/g). All assays were carried out at least in triplicate.

Intracellular metabolomics analysis
Cell pellets of TM, TM-06T, TM-15T, and TM-TP (each 50 mg) were re-suspended in 1000 μL of cold acetonitrile-methanol-water (2:2:1, v/v) containing 1 μg/mL 2-chlorophenylalanine as an internal standard. The extraction and determination of metabolites were conducted exactly as described by Xu et al. [33]. The raw data detected by UHPLC-Q-Exactive-Orbitrap-MS was converted to the mzML format and analysed by R package XCMS (version 3.2) for peak identi cation and matching. The peak area of each metabolite was normalized to the internal standard, and the normalized data from ve replicates for each sample was subjected to multivariate statistical analysis, including PCA and OPLS-DA, using the SIMCA-P package (Umetrics, Umea, Sweden). DMs between two samples were identi ed according to the restrictive conditions of fold change ≥2 and a corrected p-value <0.05.

Genome sequencing, assembly and annotation
Genomic DNA of KMLY1-2 was extracted using a yeast genome DNA extraction kit (Tiangen, China) and sequenced on the PacBio RS II platform. Brie y, the quali ed genomic DNA was fragmented with Covaris G-tubes and end-repaired to prepare single-molecule real-time (SMRT) bell DNA template libraries (insert size of >10 Kb) according to the manufacturer's speci cations (PacBio, Menlo Park, CA). SMRT sequencing was performed on the PacBio RSII sequencer using P4-C2 chemistry. The long seed reads were corrected and assembled as described by Frank et al. [34]. Finally, the assembled sequences (scaffolds) were deposited in GenBank under accession number JADIFZ000000000, and applied as the draft genome sequence of KMLY1-2. CDSs were predicted using AUGUSTUS [35], and their annotations were performed using BLASTP search against the Nr, Swiss-Prot, KOG, and KEGG databases. In addition, rRNA and tRNA prediction was performed by RNAmmer [36] and tRNAscan-SE [37], respectively.

RNA extraction, library construction and sequencing
Total RNAs from the samples TM, TM-06T, TM-15T, and TM-TP were extracted using a trizol reagent kit (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions, and 12 high-quality RNA samples were used to construct RNA-seq libraries using the TruSeq RNA sample preparation kit (Illumina, San Diego, CA) following the manufacturer's protocol. Then, the libraries were sequenced using Illumina HiSeq 4000 to generate 100 nt paired-end raw reads. The raw data were deposited in the short read archive database under accession number SRR12995594-SRR12995605. Clean reads were selected by a Perl program according to the criteria described by Wu et al. [38] and mapped to the genome of KMLY1-2 using HISAT2 [39] with default settings.

Analysis of DEGs
Aligned reads were quanti ed using an FPKM method [40]. DEGs between two samples were identi ed using DESeq2 [41] following two criteria: an absolute value of log 2 fold change ≥1 and a false discovery rate (FDR) <0.05. The DEGs were then subjected to clustering analysis by using the STEM algorithms [42].   Figure 1 Kinetics curves of TOL production and cell growth (a), and relative expression levels of TOL key biosynthetic genes (b). Different lowercase letters (a to e) indicate signi cant differences (p <0.05). The expression level at 24 h was set as 1, and # denotes the fold change relative to 24 h ≥2.

Figure 2
Effect of Trp and Phe on TOL production. a Extracellular TOL and residual Trp; b Extracellular TOL; c Intracellular TOL. Signi cant differences (p <0.05) are indicated by different lowercase letters above the bars (a to e or a' to g').
Page 23/26 Expression levels of DEGs involved in TOL biosynthesis. a the Ehrlich pathway; b the de novo synthetic pathway; c the auxiliary pathway. Means marked with different lowercase letters (a to c, a' to c' or a'' to c'') differ signi cantly (fold change ≥2 and FDR <0.05). The description of genes represented by gene ID is shown in Table 1. Schematic diagram of TOL biosynthetic pathways in S. cerevisiae. The up-and downregulated metabolites in TM-06T and/or TM-15T relative to TM are indicated in red and green fonts, respectively. If a metabolite is up-or downregulated in TM-06T vs. TM-TP, it is marked with a red or green underline, respectively. The expression pattern of each gene is represented by heat maps (as log2 fold change).
Detailed information of DMs and DEGs in this picture can be found in Table S5.
This is a list of supplementary les associated with this preprint. Click to download.