Global Transcriptome Profile of the Oleaginous Yeast Saitozyma podzolica DSM 27192 Cultivated in Glucose and Xylose

Unlike conventional yeasts, several oleaginous yeasts, including Saitozyma podzolica DSM 27192, possess the innate ability to grow and produce biochemicals from plant-derived lignocellulosic components such as hexose and pentose sugars. To elucidate the genetic basis of S. podzolica growth and lipid production on glucose and xylose, we performed comparative temporal transcriptome analysis using RNA-seq method. Approximately 3.4 and 22.2% of the 10,670 expressed genes were differentially (FDR < 0.05, and log2FC > 1.5) expressed under batch and fed batch modes, respectively. Our analysis revealed that a higher number of sugar transporter genes were significantly overrepresented in xylose relative to glucose-grown cultures. Given the low homology between proteins encoded by most of these genes and those of the well-characterised transporters, it is plausible to conclude that S. podzolica possesses a cache of putatively novel sugar transporters. The analysis also suggests that S. podzolica potentially channels carbon flux from xylose via both the non-oxidative pentose phosphate and potentially via the first steps of the Weimberg pathways to yield xylonic acid. However, only the ATP citrate lyase (ACL) gene showed significant upregulation among the essential oleaginous pathway genes under nitrogen limitation in xylose compared to glucose cultivation. Combined, these findings pave the way toward the design of strategies or the engineering of efficient biomass hydrolysate utilization in S. podzolica for the production of various biochemicals.


Introduction
According to the forecast of the United Nations, the world's population is expected to grow by~20% to 9.7 billion people within the next 30 years [1]. The demand for energy and material resources will therefore also increase. To satisfy these requirements and to circumvent economic and environmental challenges, which are associated with reliance on fossil sources, renewable and sustainable energy sources need to be developed. Microbial lipids produced by oleaginous yeasts could contribute as one of the numerous alternative energy sources. Oleaginous yeasts are able to accumulate up to 70% of their cell dry mass to the storage lipids, mainly triacylglycerols [2]. As a carbon source, yeasts are able to metabolize a large spectrum of cheap and abundant raw materials, such as lignocellulosic plant biomass or waste from the agricultural and food industry, enabling carbon recycling within a biorefinery concept [3][4][5]. Moreover, yeast cell propagation proceeds with high duplication rates and the cultivation can be easily upscaled, in contrast to microalgae. In addition, yeast cultivation is independent of season, climate and location, which is highly advantageous compared to oil seeds [6,7]. However, the molecular and evolutionary mechanisms of oleaginicity in yeasts remain unclear. By phylogenetic comparisons of oleaginous yeasts, it becomes apparent that often the property of oleagenicity is not shared

Yeast Strain and Experimental Set Up
The oleaginous yeast Saitozyma podzolica DSM 27192 was isolated from peat bog soil in the Black Forest of Germany [15]. The yeast is maintained as glycerol stock in −80 • C storage and reactivated on YM agar as previously described [10].
Cultivation of the yeast was performed in duplicates in 2.5 L benchtop bioreactors (Infors HT, Bottmingen, Switzerland; Minifors fermentor) containing 1.2 L mineral salt medium as previously described [10]. Briefly, the cultivation parameters were pH 4, 22.5 • C, 600 rpm and 1 vvm aeration rate for 144 h of process time. The first 48 h comprised a batch phase for biomass production. At the end of this phase, the concentration of glucose or xylose was~10 g/L and the nitrogen (ammonia) concentration was nearly approaching zero. At 48 h, a continuous automatic feed supply was established to maintain glucose or xylose concentration constant at~10 g/L for lipid production during the rest of the cultivation. Yeast growth and production were tracked by daily sampling and analysis of dry cell mass, carbon source and ammonia consumption and lipid and organic acid content as described in [10].
Briefly, 1 mL sample from each time point was collected and used to determine the cell dry weight (CDW) as well as the sugar, organic acid and ammonium concentrations. The sample was centrifuged at 20,000× g for 3 min. The pellets were washed using physiological saline and then centrifuged at 20,000× g for 3 min. CDW was determined by weighing the dried pellets (48 h at 60 • C) using a precision balance.
Glucose concentration was quantified by enzymatic assay using D-glucose test kit from R-Biopharm (Art. No. 10716251035, R-Biopharm AG, Darmstadt, Germany), according to the manufacturer's protocol.
For xylose, 50 • C column temperature, 5 mM H 2 SO 4 mobile phase with 0.5 mL/min flow rate were used, and the compound was detected using a refractive index detector (Agilent 1200 series, Agilent Technologies Deutschland GmbH, Böblingen, Germany). The supernatant was diluted with 20 mM KH 2 PO 4 pH 2.5 prior to organic acids quantification. A gradient mobile phase comprising 20 mM KH 2 PO 4 pH 2.5 and 100% methanol was used at 1 mL/min to separate the compounds. The oven column was set at 30 • C and the compounds were detected using a UV detector (220 nm).
Concentration of ammonium nitrogen was determined photometrically by modifying the assay volume to 300 µL per sample in microtiter plates using the Spectroquant kit (1.14752.0001, Merck KGaA, Darmstadt, Germany).
Lipids were first derivatized to fatty acid methyl esters (FAMEs) by direct transesterification of yeast cell mass as described by [15,17]. Approximately 20-30 mg freezedried cell mass was acid esterified using a two-phase system. The first phase comprised 0.5 mL 2 mg/mL heptadecanoic acid in hexane and 1.5 mL pure hexane. The second phase contained equal volumes of 2 mL 15% H 2 SO 4 in methanol. A thermo-shaker (Universal Labortechnik, Leipzig, Germany) was used to incubate the mixture at 100 • C and 1000 rpm for 2 h. Samples were vortexed at 30 min intervals, and the reaction was stopped by placing the reaction tubes on ice for 10 min. The upper phase containing FAMEs was analysed by gas chromatography (GC) using the 6890N Network GC-System (Agilent Technologies Deutschland GmbH, Waldbronn, Germany) coupled with a DB-Wax column (30 m × 0.25 mm) (Art. No. 122-7032; Agilent Technologies Deutschland GmbH, Böblingen, Germany). Sample separation was done by using a temperature gradient of 40-250 • C at 8 • C/min. Final temperature was maintained at 250 • C for 10 min and the FAMEs were detected and quantified based on the RM3 FAME Mix standard (Art. No. 07256-1AMP, Sigma Aldrich, Taufkirchen, Germany).

RNA Isolation, Library Construction and Sequencing
To compare the transcriptome profiles of Saitozyma podzolica DSM 27192 under different carbon sources and over the course of the two growth phases, biomass was sampled at 22, 46, 70 and 120 h. The biomass was pelleted by centrifugation, snap-frozen with liquid nitrogen and stored at −80 • C until RNA extraction. Total RNA extraction and mRNA sequencing was performed at Microsynth AG (Balgach, Switzerland). Illumina-stranded TruSeq RNA libraries were prepared after poly(A) enrichment and sequenced based on Illumina NextSeq (2 × 150 bp as paired-end) sequencing chemistry.

Functional Analysis of Expressed Genes
Stringtie assembled transcripts were processed further using TransDecoder v5.5.0 (https://github.com/TransDecoder/TransDecoder, accessed on 20 August 2020) to obtain putative protein sequences. The candidate protein sequences obtained with TransDecoder were subsequently compared with predicted proteins from the Saitozyma podzolica DSM 27192 draft genome sequence with the aid of cd-hit v4.8.1 [30] to compile a complete set of full-length protein sequences for all expressed genes. The protein sequences were annotated using pannzer2 [31], KofamKOALA [32], eggNOG [33] and InterProScan v5 [34], respectively. The functional enrichment analysis reported here is primarily based on gene ontology (GO) annotation, unless otherwise stated. To evaluate the functional implication of differential gene expression, we performed GO term enrichment analyses using WebGestalt 0.4.4 with default setting and to reduce GO term redundancy, affinity propagation based on apcluster was applied [35].

Phylogenetic Analysis of Sugar Transport Proteins
The predicted protein sequences associated with overexpressed sugar transporters were further annotated using BLASTp against the conserved domain database [36], the curated UniProtKB/Swiss-Prot [37] and transporter classification [38] databases as well as prediction of transmembrane helices using TMHMM Server v2.0 [39]. The proteins were aligned using PSI/TM-Coffee [40] with UniRef100 homology extension and transmembrane sequence type options. A phylogenetic tree was constructed using IQ-TREE v1.6.11 [41].

Overview of Yeast Cultivation and RNA-Seq
The oleaginous yeast Saitozyma podzolica DSM 27192, cultivated on glucose and xylose ( Figure 1) was sampled for RNA-sequencing (RNA-seq) at 22, 48, 70 and 120 h with a view to evaluate the transcriptional response of the yeast during growth on the two sugars. For this study, we partitioned the cultivation into two stages. During the initial batch phase (~0-48 h) on glucose (Figure 1a), the yeast grew to a maximum cell mass of 14.03 ± 1.31 g/L at 48 h. At this time point,~91.44% of ammonium was metabolised by the yeast. Cultivation of the yeast under batch mode on xylose followed the trend observed with glucose ( Figure 1b). After 48 h growth on xylose, the yeast attains a maximum cell mass of 11.00 ± 0.85 g/L and metabolised~80.65% of the nitrogen source. The second phase (~48-144 h) involved establishment of a continuous feeding at~10 g/L sugars concentration under nitrogen limitation as previously described [10]. In both glucose and xylose cultures, ammonium was metabolised completely around 70 h (Figure 1a,b). The biomass increased steadily to a maximum of 26.78 ± 3.50 ( Figure 1a) and 25.38 ± 0.25 g/L (Figure 1b) on glucose and xylose, respectively. The latter phase is characterised by accumulation of singlecell oil (SCO), a process triggered by nitrogen limitation, with the complete depletion of the nitrogen source at 70 h [9]. For instance, compared to sampling time point 46 h, lipid accumulation, estimated based on percent FAME per cell dry weight, increased by~132 and 126% at 90 h and by~209 and 195% at 144 h for glucose and xylose cultures, respectively [9].
Four RNA samples, one from the batch phase and three from the fed-batch phases, have been sequenced using NextSeq 2 × 150 (v2.5) sequencing chemistry. After trimming out sequencing adapters and filtering out low-quality reads, the data comprised~1.5 billion high-quality and clean reads of~224 Gb (Figure 2a). The glucose dataset comprised a range of~32.7-62.6 million reads of size ranging between~4.8 and~9.2 Gb and for xylose the dataset consisted of~38.6-61.9 million reads ranging between~5.6 and~9.0 Gb in size. Initial reads alignment using Hisat2 against the draft genome of S. podzolica DSM 27192 [20] resulted in an average read alignment rate of 95.05% with a range of~89. 47-97.19% ( Figure 2b), signifying the quality and the accuracy of the sequencing data. StringTie assembly evaluation using GffCompare [25] revealed 10,670 super-loci, including 2039 (19.1%) novel loci, which represent transcripts that are missing or novel isoforms of genes in the previous annotation of S. podzolica DSM 27192. Four RNA samples, one from the batch phase and three from the fed-batch phases, have been sequenced using NextSeq 2 × 150 (v2.5) sequencing chemistry. After trimming out sequencing adapters and filtering out low-quality reads, the data comprised ~1.5 billion high-quality and clean reads of ~224 Gb (Figure 2a). The glucose dataset comprised a range of ~32.7-62.6 million reads of size ranging between ~4.8 and ~9.2 Gb and for xylose the dataset consisted of ~38.6-61.9 million reads ranging between ~5.6 and ~9.0 Gb in size. Initial reads alignment using Hisat2 against the draft genome of S. podzolica DSM 27192 [20] resulted in an average read alignment rate of 95.05% with a range of ~89.47-97.19% (Figure 2b), signifying the quality and the accuracy of the sequencing data. StringTie assembly evaluation using GffCompare [25] revealed 10,670 super-loci, including 2039 (19.1%) novel loci, which represent transcripts that are missing or novel isoforms of genes in the previous annotation of S. podzolica DSM 27192. To improve the downstream gene expression analysis, we removed genes with inconsistent expression by implementing a stringent filter (minimum counts per million (CPM) ≥ 0.5 in a minimum of 2 libraries), resulting in 10,506 genes with various expression profiles across the samples. The distribution of normalised gene counts based on the regularized logarithm transformation (rlog) implemented in DESeq2 [26] is presented in Figure 2c,d. Clustering of the samples, using principal component analysis (PCA), showed that biological replicates for each sampling point cluster distinctly from samples of other time points. Interestingly, samples from 22 and 46 h sampling points cluster more closely together for both glucose and xylose grown yeast compared to those from 70 and 120 h samples, suggesting greater variability between the former and the latter sets. However, the first two principal components explained only 59% of the observed variability, suggesting that differences exist among transcriptome profiles of S. podzolica DSM 27192 growing on glucose and xylose, and across the four sampling time points. To improve the downstream gene expression analysis, we removed genes with inconsistent expression by implementing a stringent filter (minimum counts per million (CPM) ≥ 0.5 in a minimum of 2 libraries), resulting in 10,506 genes with various expression profiles across the samples. The distribution of normalised gene counts based on the regularized logarithm transformation (rlog) implemented in DESeq2 [26] is presented in Figure 2c,d. Clustering of the samples, using principal component analysis (PCA), showed that biological replicates for each sampling point cluster distinctly from samples of other time points. Interestingly, samples from 22 and 46 h sampling points cluster more closely together for both glucose and xylose grown yeast compared to those from 70 and 120 h samples, suggesting greater variability between the former and the latter sets. However, the first two principal components explained only 59% of the observed variability, suggesting that differences exist among transcriptome profiles of S. podzolica DSM 27192 growing on glucose and xylose, and across the four sampling time points.

Differential Gene Expression during Initial Growth under Glucose and Xylose in DSM 27192
To investigate possible genetic differences and genes shaping initial growth of Saitozyma podzolica DSM 27192 on glucose and xylose, we compared gene expression profiles of the 22 h samples using DESeq2 [26]. There were 361 differentially expressed (FDR < 0.05, and log2FC > 1.5; Figure S1a) genes (representing~3.4% of the expressed genes) observed when the yeast was grown on glucose compared to xylose. Further evaluation of the expression profiles revealed upregulation of 115 and down-regulation of 246 genes during growth on glucose relative to xylose ( Figure S1b).
To gain insight into signatures of molecular activities associated with initial growth of the yeast on glucose and xylose, we performed gene ontology (GO)-based functional enrichment analysis using WebGestalt [35]. Interestingly, various GO associated with transmembrane transport is associated with the observed overexpressed genes. For instance, overrepresentation of transmembrane transport (GO:0055085; FDR = 0.07), including at least six of the 15 genes annotated as sugar (and other) transporter genes, was observed during initial growth on glucose relative to xylose (Table 1 and Table S1). By contrast, a more diverse set of genes associated with the transmembrane transporter activity of GO:0022857 was significantly overrepresented (FDR < 0.05) among upregulated genes during growth on xylose compared to glucose (Table 1 and Table S1). As with glucose upregulated genes, the majority of these (23 out of 51 genes) have been annotated as a sugar (and other) transporter gene (Table S1). Besides the predominance of these, genes of K + transporter proteins in the glucose grown yeast, formate nitrite transporter, fungal trichothecene efflux pump, mitochondrial carrier, voltage-gated K channel and other general substrate transporter proteins in xylose grown cultures, were linked to the overrepresented transmembrane transport function (Table S1), suggesting a significant variation in the movement of the two sugars in Saitozyma podzolica DSM 27192. Overrepresentation of transmembrane transporter GOs, which indicates active transmembrane transport of a variety of substrates has been reported in many fungi [42,43]. Unlike the observed significant and specific enrichment of these GOs in glucose and xylose grown yeast observed in this study, genes associated with both GOs (GO:0022857 and GO:0055085) were reported to be up and downregulated on glucose grown Aspergillus fumigatus mutants [44].

Distinct Profiles of Putative Sugar Transporter Genes in Saitozyma Podzolica DSM 27192 Grown on Glucose and Xylose
To evaluate the potential implication of variation in the above differentially expressed sugar (and other) transporter genes during growth of the oleaginous yeast on xylose and glucose, we performed additional annotation of the predicted proteins using conserved domain database [36], the curated UniProtKB/Swiss-Prot [37] and transporter classification [38] databases as well as prediction of transmembrane helices using TMHMM Server v2.0 [39]. The analysis showed that the predicted proteins of 42 of the 65 differentially expressed transmembrane transport-associated genes under glucose and xylose (Table S1) shared significant homology (e-value, and a bit score range of 1.0 × 10 −13 -5.0 × 10 −220 and 75.1-761, respectively) with members of sugar porter family: 2.A.1.1. However, these proteins shared low homology with the sugar porter proteins, and the curated yeast transporters, with similarity values ranging between 23.3-72.5%, respectively, suggesting novelty of the putative transporters in S. podzolica DSM 27192 when compared to their closest characterised relatives.
Phylogeny-based evaluation of proteins of the above sugar porter genes showed a clustering of the 7 and 35 putative sugar (and others) transporters significantly overexpressed in glucose and xylose, respectively, in six groups. (Figure 3). Cluster 1 harbours the majority of the functionally characterised sugar transporters included in this analysis and includes 2 and 11 putative transporters whose genes were significantly overexpressed in glucose and xylose, respectively. However, within this cluster, distinct transporter pro-teins whose genes were significantly upregulated (log2FC = 2.0 and FDR = 4.0 × 10 −2 ) in glucose (EHS25_008238) and xylose (EHS25_001351; log2FC = 1. 4 [45] and showed varied expression depending on available substrate [46], indicating that the overexpressed gene in S. podzolica is likely a glucose transporter. By contrast, the gene of STL1 (2.A.1.1.38) was reported to be repressed alongside glycerol transport by glucose in Saccharomyces cerevisiae [47]. However, the low homology (30.3% similarity value) shared between EHS25_001351 and homologues of STL1 suggest that the former is a putative xylose rather than glycerol transporter but similarly repressed by glucose in S. podzolica. This is further supported by the presence, within the same cluster, of one xylose facilitator of family 2.A.1.1.40 [48] which grouped closely with the xylose overexpressed (log2FC = 2.90 and FDR = 1.4 × 10 −13 ) EHS25_002168, sharing 39% identity. In S. cerevisiae, homologues of STL1 and xylose transporters share similarity values ≤ 27% [47].
Cluster 3 also includes two putative sugar transporter genes EHS25_000057 and EHS25_009483, which were highly induced under xylose relative to glucose with log2FC values of 11.59 and 10.26 (Table S1). The proteins of these significantly overexpressed (FDR < 0.05) putative sugar transporter genes, however, grouped with a cellobiose/ cellotriose/cellodextrin/lactose transporter (2.A.1.1.83; Figure 3) but as with the predicted proteins described in clusters 1 and 2, shared low similarity values ranging between 34.4 and 41.5 with the closest characterised protein of the family 2.A.1.1.83 [53], thereby suggesting a different role for the S. podzolica orthologues of this family.  LG + F+ R6 model with -bb 1000 using IQ-TREE v1.6.11. Orange and black fonts indicate overrepresented proteins (translated genes) in glucose and xylose, respectively, while green fonts represent top matching proteins from the transporter classification database. The alignment was generated using transmembrane proteins (PSI/TM-Coffee) function of T-Coffee.
Aside from the above three clusters, which contain putative sugar transporters, highly induced in the presence of glucose or xylose the rest of sugar (and other) proteins were grouped in clusters 4, 5 and 6 ( Figure 3). Cluster 4 include a single S. podzolica protein with a significantly overexpressed (FDR < 0.05) gene under glucose relative to xylose and affiliated siderophore iron transporter 3, Str3 (2.A. 1.16.9) in the phylogeny (Figure 3). By contrast, cluster 5 comprises two proteins whose genes were overexpressed (FDR < 0.05) under xylose relative to glucose with the proteins being the most similar MFS transporter MFSG; family 2.A.1.13.27. Str3 participates in iron homeostasis in Schizosaccharomyces pombe and MFSG was associated with tolerance against glucosinolates S. cerevisiae [54].
Combined, evidence from phylogeny, sequence homology and gene expression levels strongly suggest a putative sugar transport role for proteins included in cluster 1, 2 and 3. The cache of putative sugar transporters in S. podzolica DSM 27192 could be used, therefore, to enhance production of useful bioproducts from a wide range of biomass hydrolysates. However, fermentation experiments with the closely related oleaginous yeast C. curvatus, and other oleaginous yeasts, revealed a repression of the consumption of xylose and other sugars by glucose [51,52,55] and a decrease in consumption of both glucose and xylose by other components of biomass hydrolysate [56]. The ability of our yeast to utilize a mixture of sugars or to grow directly on the hydrolysates is still under consideration. Future characterisation of the putative S. podzolica sugar transporter proteins is, however, necessary to validate the predicted roles and the precise mechanisms associated with the predicted function.

Comparison of Transcriptome Profiles of Saitozyma Podzolica DSM 27192 during Continuous Feeding on Glucose and Xylose
To gain insight into the gene expression dynamics during the second phase of the cultivation (fed-batch) of S. podzolica DSM 27192 under glucose and xylose, two sets of comparison were performed ( Figure 4). First, a comparison of the expression profiles for growth under glucose and xylose at 46 h, 70 h and 120 h, was performed. A total of 2364 genes were differentially expressed (FDR < 0.05, and log2FC > 1.5) in glucose relative to xylose, over the period of the cultivation (Figure 4a). Of these, only 119 genes were common to all time points for the respective contrast between glucose and xylose profiles. The highest number (1292) of differentially expressed genes (DGEs), which includes 541 and 751 up and down regulated genes, respectively, was observed at the onset of the fed batch process (Figure 4b).
As with the expression profile from the batch cultivation and consistent with continued addition of sugars during this phase of growth, both the upregulated and downregulated genes at time point 46 h were significantly associated (FDR < 0.05) with transmembrane activity ( Table 2 and Table S2). By contrast, similar numbers of DEGs were observed at time points 70 and 120 h (Figure 4a,b). Consistently, hydrolase activity, hydrolysing O-glycosyl compounds (GO:0004553) was downregulated across the three time points in cultures grown with glucose relative to xylose ( Table 2 and Table S2) as observed under batch cultivation (Table 1). Xylose induction of carbohydrate-active enzymes as well as repression of carbon catabolite by both glucose and xylose have been reported in different fungi [57][58][59]. For instance, cellulases and xylanases have been induced using xylose in Thermoascus aurantiacus [60]. Overexpression of these genes, therefore, suggests that xylose grown S. podzolica could be utilized or improved for an enhanced production of these enzymes.
growth under glucose and xylose at 46 h, 70 h and 120 h, was performed. A total of 2364 genes were differentially expressed (FDR < 0.05, and log2FC > 1.5) in glucose relative to xylose, over the period of the cultivation (Figure 4a). Of these, only 119 genes were common to all time points for the respective contrast between glucose and xylose profiles. The highest number (1292) of differentially expressed genes (DGEs), which includes 541 and 751 up and down regulated genes, respectively, was observed at the onset of the fed batch process (Figure 4b).  Aside from these, two genes of NAD + -dependent formate dehydrogenase EHS25_002775 (FDR = 0.01 and log2FC −3.01) and EHS25_003099 (FDR = 0.02 and log2FC −1.78) were significantly downregulated in glucose relative to xylose grown cultures at 46 h (Table S2). Formate dehydrogenase oxidises formate to CO 2 and the NADH generated from the reaction may serve as a source of reducing equivalents [61,62]. Formate dehydrogenase activity has been reported in S. cerevisiae and Candida tropicalis grown on glucose and xylose, respectively, as well as in a combination of these sugars with formate [61,62]. Two possible sources of formate during fermentation of glucose or xylose in S. podzolica are either exogenous formate in the culture medium or formate produced via NAD/NADP-NADH/NADPH interconversion and other pathways [63], with the latter source being more likely. Downregulation of formate dehydrogenase genes, EHS25_002530 (log2FC = −4.79 and FDR = 2.33 × 10 −3 ), EHS25_002775 (log2FC = −3.92 and FDR = 1.7 × 10 −6 ) and EHS25_003099 (log2FC= -3.28 and FDR = 9.19 × 10 −11 ) at 46 relative to 70 h sampling point and upregulation of EHS25_003099 (log2FC = 1.33 and FDR =2.04 × 10 −2 ) at 70 relative to 120 h sampling point were also observed in culture growing under glucose (Table S2), suggesting significant formate dehydrogenase activity in glucose grown S. podzolica. Further evaluation of data revealed a significant upregulation of 3,4-dihydroxy-2-butanone 4-phosphate synthase gene (EHS25_006321; log2FC = 0.72 and FDR = 4.07 × 10 −2 ) at 46 relative to 70 h in glucose grown cultures (Table S2), suggesting a greater production of formate in glucose grown S. podzolica. 3,4-dihydroxy-2-butanone 4-phosphate synthase is a key enzyme of flavin biosynthesis catalysing the conversion of Dribulose 5-phosphate to 1-deoxy-L-glycero-tetrulose 4-phosphate by releasing formate [64]. Gene ontology-based enrichment analysis further revealed significant downregulation of genes associated with cellular respiration (GO:0045333: FDR = 7.4 × 10 −3 ) at 46 h sampling point and cellular carbohydrate metabolic process (GO:0044262; FDR = 3.6 × 10 −2 ) at 70 h sampling point in S. podzolica grown on glucose relative to xylose (Table 2). However, a more comprehensive comparison of carbohydrate metabolism under the two sugars is presented below. Glycolysis, Pentose Phosphate, and Glucuronate Interconversion Pathways and TCA Cycle during Initial Growth of Saitozyma Podzolica DSM 27192 on Glucose Relative to Xylose To further elucidate the functional difference between transcriptome profiles of glucose and xylose grown S. podzolica, we, in addition to gene ontology, scanned relevant metabolic pathways associated with the overexpressed genes at 22 h time point using against KEGG ortholog [32]. Using Pathview [65], KO identities of the differentially expressed genes were mapped against several important pathways (Figure 5a,b). No significant difference was observed for the comparison of TCA cycle between the two transcriptome profiles during the initial growth of the yeast (time point 22 h: Figure 5b). By contrast, two genes each of glycolysis and pentose phosphate pathway show significant differences between the two conditions. In glycolysis, EHS25_004951 (log2FC = 1.2 and FDR = −5.0 × 10 −4 ) and EHS25_006131 (log2FC = −1.6 and FDR = 2.0 × 10 −2 ) coding acetyl-CoA synthetase (EC:6.2.1.1) and alcohol dehydrogenase, propanol-preferring (EC:1.1.1.1), respectively, were significantly induced in xylose relative to glucose condition (Figure 5a). The former catalyses the activation of acetate to acetyl-CoA while the latter catalyses the first step of alcohol degradation. Previous studies in S. cerevisiae [66] showed that the expression of acetyl-CoA synthetase gene is strongly suppressed under carbon source limitation and high glucose concentration. Contrary to our observation, however, no significant difference exists in expression of the gene between recombinant xylose-utilizing S. cerevisiae grown on glucose and xylose [67]. Our observation, suppression, and overexpression of the alcohol dehydrogenase gene under glucose and xylose, respectively, was reported in these studies [66,67]. The overexpression of acetyl-CoA synthetase gene under xylose could prove significant for developing fermentation strategies on different C-sources in S. podzolica.
Comparison of the gene expression during fed-batch cultivation for the individual sugars revealed that 5308 and 3666 genes were differentially expressed (FDR < 0.05) across the three time points in glucose and xylose, respectively (Figure 4c). Of these 8,974 genes that were differentially expressed in both sugars,~2.8% (250) were observed among all time points and both sugars (Figure 4c). Higher numbers of genes (3217), comprising 1339 and 1788 up and downregulated genes, were observed to be differentially expressed at 46 relative to the 70 h time point in glucose grown cultures compared to those grown on xylose (Figure 4d). By contrast, a greater number of genes (2258), comprising 1092 and 1166 genes were significantly (FDR < 0.05) up and downregulated, respectively, at 70 relative to 120 h time point in xylose compared to glucose grown cultures. Evaluation of the functional basis of the DEGs for the various profile comparisons, however, showed different metabolic patterns during growth of the oleaginous yeast on glucose and xylose ( Table 2). For instance, at 46 relative to 70 h time point, 115 genes associated with the carboxylic acid metabolic processes (FDR = 8.7 × 10 −6 ) were downregulated under glucose compared to 9 genes of acyl-CoA metabolic process (FDR = 3.1 × 10 −2 ) for the same time point in xylose (Table 2).
To further elucidate the functional difference between transcriptome profiles of glu-cose and xylose grown S. podzolica, we, in addition to gene ontology, scanned relevant metabolic pathways associated with the overexpressed genes at 22 h time point using against KEGG ortholog [32]. Using Pathview [65], KO identities of the differentially expressed genes were mapped against several important pathways (Figure 5a,b). No significant difference was observed for the comparison of TCA cycle between the two transcriptome profiles during the initial growth of the yeast (time point 22 h: Figure 5b). By contrast, two genes each of glycolysis and pentose phosphate pathway show significant differences between the two conditions. In glycolysis, EHS25_004951 (log2FC = 1.2 and FDR = −5.0 × 10 −4 ) and EHS25_006131 (log2FC = −1.6 and FDR = 2.0 × 10 −2 ) coding acetyl-CoA synthetase (EC:6.2.1.1) and alcohol dehydrogenase, propanol-preferring (EC:1.1.1.1), respectively, were significantly induced in xylose relative to glucose condition (Figure 5a). The former catalyses the activation of acetate to acetyl-CoA while the latter catalyses the first step of alcohol degradation. Previous studies in S. cerevisiae [66] showed that the expression of acetyl-CoA synthetase gene is strongly suppressed under carbon source limitation and high glucose concentration. Contrary to our observation, however, no significant difference exists in expression of the gene between recombinant xylose-utilizing S. cerevisiae grown on glucose and xylose [67]. Our observation, suppression, and overexpression of the alcohol dehydrogenase gene under glucose and xylose, respectively, was reported in these studies [66,67]. The overexpression of acetyl-CoA synthetase gene under xylose could prove significant for developing fermentation strategies on different Csources in S. podzolica.  [68]. Asterisk '*' indicates that D-xylulose reductase is also called xylitol dehydrogenase.
Of the fifteen genes associated with pentose phosphate and glucuronate interconversions, glycolysis pathways and TCA cycle, only two genes have been overexpressed in glucose relative xylose grown yeast over the duration of the batch fed process ( Figure 5). By contrast, five genes of the pentose and glucuronate interconversion and two of pentose phosphate pathways were significantly suppressed under glucose compared to growth under xylose. As expected, significant overrepresentation of genes encoding D-xylulose Aside from the dynamics of AMPD and ICDH above, several studies have reported the expression profiles of the 'oleagenic' genes in yeast during oil accumulation under various conditions. For instance, in Trichosporon oleaginosus grown under two carbon sources and nitrogen or phosphate limitations, only ACL and FAS genes were significantly overrepresented (FDR < 0.05) in the yeast grown in xylose under nitrogen limitation relative to the control cultures and only ACL gene significantly overrepresented (FDR < 0.05) in N-acetyl glucosamine grown yeast under nitrogen limitation relative to the control cultures [83]. By contrast, the current study did not reveal significant differences in the expression of these genes in glucose compared to xylose grown S. podzolica under nitrogen limitation (sampling time 76 h; Figure 1). Apparently, the observed overexpression (FDR < 0.05) of the ACL gene (also reported in [83]) and suppression (FDR < 0.05) of the ICDH gene likely underpin lipid accumulation under nitrogen limitation in S. podzolica. Aside from the dynamics of AMPD and ICDH above, several studies have reported the expression profiles of the 'oleagenic' genes in yeast during oil accumulation under various conditions. For instance, in Trichosporon oleaginosus grown under two carbon sources and nitrogen or phosphate limitations, only ACL and FAS genes were significantly overrepresented (FDR < 0.05) in the yeast grown in xylose under nitrogen limitation relative to the control cultures and only ACL gene significantly overrepresented (FDR < 0.05) in N-acetyl glucosamine grown yeast under nitrogen limitation relative to the control cultures [83]. By contrast, the current study did not reveal significant differences in the expression of these genes in glucose compared to xylose grown S. podzolica under nitrogen limitation (sampling time 76 h; Figure 1). Apparently, the observed overexpression (FDR < 0.05) of the ACL gene (also reported in [83]) and suppression (FDR < 0.05) of the ICDH gene likely underpin lipid accumulation under nitrogen limitation in S. podzolica.

Temporal Dynamics of S. podzolica DSM 27192 Genes Grown under Glucose and Xylose
To identify the gene expression profile changes over the duration of S. podzolica DSM 27192 growth on glucose and xylose under fed-batch condition (46, 70 and 120 h), we implemented the maSigPro algorithm based on alfa threshold value of 0.5. The analysis revealed 4782 genes (~47% of the expressed genes) showing significant differential expres-

Temporal Dynamics of S. podzolica DSM 27192 Genes Grown under Glucose and Xylose
To identify the gene expression profile changes over the duration of S. podzolica DSM 27192 growth on glucose and xylose under fed-batch condition (46, 70 and 120 h), we implemented the maSigPro algorithm based on alfa threshold value of 0.5. The analysis revealed 4782 genes (~47% of the expressed genes) showing significant differential expression (FDR < 0.05) during the continuous cultivation experiment. The co-expressed genes were partitioned based on the K-means cluster method (k = 6 and R 2 ≥ 0.7) into six distinct expression clusters (Figure 7). Clusters 1 and 5, comprising 867 (18.13% of the co-expressed genes) and 1076 (22.5% of the co-expressed genes) genes, respectively, show opposite trajectories, with expression values in the former cluster decreasing at 70 h before then peaking again at 120 h. Evaluation of genes in cluster 1 revealed an overrepresentation (FDR < 0.05) of two biological processes (BP), ribosome biogenesis, (GO:0042254) and RNA metabolic process (GO:0016070), as well as one cellular component (CC) function, nuclear lumen (GO:0031981), suggesting a possible suppression of translational machinery in response to nitrogen limitation. Previous studies have reported the influence of nutrients availability, including nitrogen, on ribosome biogenesis [84][85][86]. Specifically, proteome study of the oleaginous yeast Rhodotorula toruloides grown in glucose and xylose revealed significant depletion of genes associated with ribosome biogenesis [87].
By contrast, two CC functions, cytoplasm (GO:0005737) and proteasome complex (GO:0000502; also included in GO:0005737) and one BP, ribose phosphate metabolic process (GO:0019693; also included in GO:0005737) were significantly overrepresented (FDR < 0.05) among genes belonging to cluster 5 (Figure 7). Upregulation of these processes provides insight into the innate evolutionary strategy adapted by the yeast to salvage the effects of nitrogen-starvation-induced stress. For instance, overrepresentation of the proteasome complex indicates increased destructive activity of the proteasome system on cyclin, which prevents organelle dismemberment and cell destruction under nitrogen limitation [88]. The potential of cyclin activity is also indicated by the upregulation of genes of cyclin (EHS25_000231) and cyclin binding protein (EHS25_007811), included among genes associated with cytoplasmic function (Figure 7; Table S3). In addition, the increased expression of genes linked to energy production and other metabolic pathways (Table S3) (Figure 7). Evaluation of the temporal dynamics represented in this cluster revealed a predominance of genes linked to mannosyltransferase activity (GO:0000030), organelle compartment (GO:0031984), transport and cytoplasm GOs with depletion of nitrogen and the onset of lipogenesis (Figure 7). Specifically, there was a decrease in the expression of several key genes linked to protein Nglycosylation [91][92][93] such as the putative α-1,2-, α-1,3-, α-1,6-, and β-1,4-mannosyltransferases under nitrogen limitation in cultures grown on both glucose and xylose (Table S3). Apparently, these genes are involved in Glycosylphosphatidylinositol (GPI) anchoring as suggested by a corresponding reduction in the expression of genes propagating the enriched organelle compartment, transport, and cytoplasm GOs. For instance, genes of Similar reverse trajectories were also observed between clusters 2 (599 genes; 12.5% of the co-expressed genes) and 3 (1426 genes; 29.8% of the co-expressed genes) showing trends of decrease and increase, respectively, during growth of S. podzolica DSM 27192 on glucose and xylose. Cluster 2 is significantly (FDR < 0.05) enriched in two CC, one BP and an MF, structural constituent of ribosome (GO:0003735), while cluster 3 shows significant overrepresentation (FDR < 0.05) in genes linked to one BP (GO:1903506; regu-lation of nucleic acid-templated transcription) and one MF (GO:0003677; DNA binding; Figure 7). As discussed above, nitrogen limitation induces the downregulation of ribosome biogenesis. However, in contrast to the trajectory of cluster 1, the depletion of the structural constituent of ribosome and peptide biosynthetic process (GO:0043043) persisted in cluster 2, corresponding to the phase of maximum oil accumulation (72-120-h time point) in S. podzolica DSM 27192 [10]. Although not conclusive at present, our study indicates a clearer association of lipogenesis and downregulation of ribosomes, contrary to the suggestion of greater association with growth activity in R. toruloides [87]. This hypothesis is further supported by various 'omics' studies under nitrogen limitation in Yarrowia lipolytica [89,90].
Cluster 4 with 604 (12.6%) genes showed a similar rapid drop in expression values as with cluster 1 at time point 70, but the expression levels remained low through to time point 120 h in both glucose and xylose cultures (Figure 7). Evaluation of the temporal dynamics represented in this cluster revealed a predominance of genes linked to mannosyltransferase activity (GO:0000030), organelle compartment (GO:0031984), transport and cytoplasm GOs with depletion of nitrogen and the onset of lipogenesis (Figure 7). Specifically, there was a decrease in the expression of several key genes linked to protein N-glycosylation [91][92][93] such as the putative α-1,2-, α-1,3-, α-1,6-, and β-1,4-mannosyltransferases under nitrogen limitation in cultures grown on both glucose and xylose (Table S3). Apparently, these genes are involved in Glycosylphosphatidylinositol (GPI) anchoring as suggested by a corresponding reduction in the expression of genes propagating the enriched organelle compartment, transport, and cytoplasm GOs. For instance, genes of phosphatidylinositolglycan biosynthesis S protein (EHS25_000085 ; Table S3), a key component of GPI transamidase complex [94], transport protein bet1 (EHS25_004488) and endoplasmic reticulum (ER)-Golgi intermediate compartment (ERGIC) protein (EHS25_004458), both involved in shuttle between endoplasmic reticulum and Golgi membrane [95,96] further support the downregulation of ER-to-Golgi activities during nitrogen-limited lipid accumulation. However, the precise implication of these findings is subject to further investigation.
Finally, cluster 6 (210 genes; 4.4%) exhibited uniform expression values across the three time points in both sugars (Figure 7). However, higher gene expression values were observed for culture grown on xylose relative to glucose, consistent with the observed significant upregulation of hydrolase activity, hydrolysing O-glycosyl compounds (GO:0004553) throughout the cultivation period in xylose compared to glucose grown cultures (discussed above; Tables 1 and 2).

Conclusions
Previous studies have reported the ability of S. podzolica to produce various biochemicals, including single-cell oils and sugar acids from different carbon sources [10]. Thus, we evaluated the transcriptome profiles in Saitozyma podzolica DSM 27192 cultivated on glucose and xylose under batch and continuous processes by RNA-seq analysis. Our transcriptomic analysis revealed a wide range of transcriptome level differences between S. podzolica grown on glucose and xylose. We showed that several putative sugar transporter genes, including transporters of other substrates, were overexpressed distinctively on either glucose or xylose, albeit in greater numbers in the latter. This study also reports xylose induction of genes associated with carbohydrate utilization, a feature which could be harnessed potentially for enhanced cultivation of the yeast on complex substrates. We also discussed the presence of putative genes associated with the initial steps of the Weimberg pathway, which likely underpin the efficient utilization of xylose and production of xylonic acid by S. podzolica. However, the basis of carbon flux partitioning between the above pathway and the non-oxidative pentose phosphate pathway remains unclear, given that genes coding for enzymes of both pathways were significantly expressed under xylose cultivation. Our study contributes to a better understanding of the genetic basis of glucose and xylose growth and production of single cell oil by S. podzolica. These data will contribute to further development of efficient fermentation strategies and development of S. podzolica strains with superior features for enhanced biotechnological applications.

Supplementary Materials:
The following are available online at https://www.mdpi.com/article/10 .3390/jof7090758/s1, Figure S1: Differential gene expression during initial growth (22 h time point) of Saitozyma podzolica DSM 27192 on glucose compared to xylose, Table S1: Differentially expressed transporter genes in Saitozyma podzolica DSM 27192 grown on glucose compared to xylose, Table S2: Functional enrichment of differentially expressed genes during continuous cultivation of Saitozyma podzolica DSM 27192 on glucose compared to xylose, Table S3