An Interplay between Transcription Factors and Recombinant Protein Synthesis in Yarrowia lipolytica at Transcriptional and Functional Levels—The Global View

Abstract

Transcriptional regulatory networks (TRNs) associated with recombinant protein (rProt) synthesis in Yarrowia lipolytica are still under-described. Yet, it is foreseen that skillful manipulation with TRNs would enable global fine-tuning of the host strain’s metabolism towards a high-level-producing phenotype. Our previous studies investigated the transcriptomes of Y. lipolytica strains overproducing biochemically different rProts and the functional impact of transcription factors (TFs) overexpression (OE) on rProt synthesis capacity in this species. Hence, much knowledge has been accumulated and deposited in public repositories. In this study, we combined both biological datasets and enriched them with further experimental data to investigate an interplay between TFs and rProts synthesis in Y. lipolytica at transcriptional and functional levels. Technically, the RNAseq datasets were extracted and re-analyzed for the TFs’ expression profiles. Of the 140 TFs in Y. lipolytica, 87 TF-encoding genes were significantly deregulated in at least one of the strains. The expression profiles were juxtaposed against the rProt amounts from 125 strains co-overexpressing TF and rProt. In addition, several strains bearing knock-outs (KOs) in the TF loci were analyzed to get more insight into their actual involvement in rProt synthesis. Different profiles of the TFs’ transcriptional deregulation and the impact of their OE or KO on rProts synthesis were observed, and new engineering targets were pointed.

1. Introduction

Transcriptional regulatory networks (TRNs) can be defined as regulatory interactions among transcription factors (TFs) and their target genes [1,2]. A given TRN is responsive to a given stimulus and governs a specific biological process. Skillful manipulation with TRN will presumably lead to global optimization of a microbial cell towards a researcher-defined target. TRNs have not yet been described for Y. lipolytica; however, a very insightful metabolic model has been recently released [3], though not yet tested for recombinant protein (rProt) production prediction. In our studies, we are interested in defining TFs, and in further perspective, TRNs, that are involved in enhancing Yarrowia lipolytica’s capacity towards high-level production of rProts. The first step in the endeavor is identifying TFs implicated in the process. Y. lipolytica has long been used as a microbial platform for rProts production [4,5,6,7]. Thanks to recent achievements in developing an erythritol-inducible expression system [8,9,10,11,12,13], Y. lipolytica has become a potent alternative to conventional expression platforms in industrial applications (several companies are implementing this system into their portfolio).

Previously, we have shown that by applying the knowledge-driven genetic engineering of Y. lipolytica’s synthesis and secretion mechanisms, the rProt yields can be significantly enhanced [14,15,16]. By defining the genes deregulated upon production of biochemically-different rProt [16] and using them as engineering ‘helper genes’ [15], the target rProt amounts were increased by over 2-fold. Notably, the previous set of the ‘helper genes’ covered those involved in the translation, folding, and trafficking of polypeptides, as well as one TF, Hac1 (YALI0B12716; hereafter, gene signatures will be abbreviated by skipping the ‘YALI0’ prefix), which is known to initiate the ER-based unfolded protein response (UPR). Hac1 mediates the activation of hundreds of molecular events, including the increased provision of chaperones and membranes, to concertedly relieve the burdened secretory pathway, e.g., [17,18]. Hence, the Hac1-governed TRN is directly associated with the biological process ‘protein synthesis and secretion’ and matches our practical aim of enhancing rProts synthesis. Technically, co-overexpression (co-OE) of HAC1 with a reporter protein initially led to mediocre yield improvement (of ~30%; [15]), but more in-depth studies showed that the effect was indeed bifurcated, causing a nearly 7-fold drop in the retained fraction of rProt but promoting its secretion by nearly 2.5-fold [19].

The following studies on exploiting Y. lipolytica’s TRNs for researcher-defined practical aims were focused on enhancing lipid accumulation [20,21], stress resistance, and rProts synthesis [22,23]. In their high-throughput screens, [20] i.a., identified Brg1 (E31757) as a TF promoting total lipid accumulation by ~1.9-fold when OE (the highest increase read from amongst all TFs tested). Gorczyca et al. [22] proved that the increased abundance of Skn1 (D14520) enables maintaining the rProt synthesis capacity under heavy osmotic stress, while OE of the Yap-like TF-encoding gene (D07744) positively impacts the growth of Y. lipolytica under alkaline conditions. OE of GZF1 (D20482) and HSF1 (E13948) brings universal improvement in the rProts synthesis, irrespectively from the environmental conditions. Deletion (knock-out; KO) of GZF1 and HSF1 promotes the synthesis of neutral lipids under a low C/N ratio (C/N of 4; not promoting lipid accumulation). Finally, a comprehensive investigation of a library of Y. lipolytica strains over-expressing (OE-ing) one of 125 TFs under 72 different environmental conditions revealed multiple interesting phenotypes driven by the increased abundance of the TFs [23], data presented in a searchable database YaliFunTome (https://sparrow.up.poznan.pl/tsdatabase/, accessed throughout June 2024). Inter alia, the YaliFunTome database presents the phenotype of another global enhancer of rProt synthesis, Klf1 (D05041), as well as TFs acting against this biological process, Azf1 (A16841), Dep1 (F05896), and Cat8 (C19151), all of which are known to be involved in carbon metabolism [24,25,26,27,28]. Notably, several of the TFs studied there promoted growth under limited oxygen availability (LAC9 (D20460), B14443, B20944, and DAL81 (D02783)), which is the major limiting factor in Y. lipolytica growth and rProt synthesis [29,30,31]. Altogether, those studies proved the validity of the concept of using TFs as tools in driving desired phenotypes in Y. lipolytica.

With the current advent of next-generation sequencing technologies, DNA synthesis, and even miniaturization and parallelization of strain culturing and sample analysis, the repositories of ‘big biological data’ have grown enormously. This is specifically true for the model microorganisms, like Saccharomyces cerevisiae or Escherichia coli. Considering the extent of the datasets deposited in publicly accessible databases, ‘big biological data recycling’ has become a very efficient research approach in driving novel hypotheses and findings, also in the field of yeast research [32]. If approached carefully and with adequate methodology [33,34,35,36,37], reusing previous data enables substantial time- and resource-savings. Examples of successful completion of such analyses for yeast are available [3,38,39]. When compared to the model yeast species, the high-throughput datasets for Y. lipolytica deposited publicly are still limited. Nevertheless, some transcriptomics data recycling for Y. lipolytica has been very recently reported [11,19,39]; where RNAseq datasets were extracted, arranged in different combinations, and re-analyzed to answer new research questions. Those previous studies reported recycling of transcriptomics data only.

In this study, we aimed to combine and reanalyze datasets related to the expression profile of TFs and the functional effect of the TFs OE on rProt synthesis in Y. lipolytica. The previously generated datasets (transcription level of TFs and amounts of rProts triggered by TF’s OE) were juxtaposed to investigate an interplay between TRNs and rProts synthesis in Y. lipolytica at transcriptional and functional levels. Having transcriptomics data on one hand and functional screening results on the other, we wanted to test a working hypothesis of whether transcriptomics data could be used as a selection-driver of TF involved in a specific biological process, here—rProt synthesis in Y. lipolytica, and if a preliminary prognostic of its mode of implication can be inferred. For several TFs, further experimental verification followed the computational analyses.

2. Results and Discussion

2.1. Data Extraction, Juxtaposition, and Global Clustering—Clusters Overview

Transcriptomics data were extracted from the NCBI SRA (Sequence Read Archive) database (PRJNA701856 and PRJNA869113). The data represent the relative expression level of a given TF-encoding gene in a Y. lipolytica strain OE-ing one of the biochemically different rProt: (i) scSoA—highly disulfide-bonded secreted alpha-amylase from Sitophilus oryzae; (ii) scYFP—minimally modified post-translationally secreted small yellow fluorescence protein; (iii) inYFP—the same protein (YFP) in intracellular form; (iv) scTlG—highly glycosylated secreted glucoamylase from Thermomyces lanuginosus; or a strain co-OE-ing scYFP and a TF Hac1 [16,19]. The strains were maintained in steady-state, as indicated previously [14,16]. The two strains, scSoA and scYFP, were characterized by high-level synthesis and secretion of the rProts and hence were assigned names and abbreviated as high-level synthesis and secretion, HSS. The other three, inYFP, TlG, and scYFP-HAC1, produced limited amounts of the rProts, and based on transcriptomics data were shown to be facing UPR, hence they were assigned and abbreviated as UPR strains. The raw data were filtered for genes encoding TFs, according to a previously defined list of putative TFs [20]. Only these records were considered further in this study. Differential expression data, used hereafter, express fold change (FC) values calculated as a ratio of the normalized transcript counts (log2) in the OE-ing strain to a prototrophic control, maintained likewise.

Functional data on the amounts of rProt (intracellular fluorescent protein) synthesized by Y. lipolytica strains OE-ing a specific TF were extracted from the YaliFunTome database [23]. The data were read as fluorescence of the strain, expressed as total amounts of rProt or normalized per biomass. Differential phenotype data, used hereafter, denote FC values calculated as a ratio of the total or normalized fluorescence readouts in the OE-ing strain to a prototrophic control, maintained likewise.

All the TFs considered in this analysis, along with assigned or putative functions, are given in

Table 1. Y. lipolytica TFs with putative or assigned function.

Yali Signature	Assigned Name	Putative/Known Function	Reference
B12716	HAC1	Transcriptional activator of genes involved in ER-based Unfolded Protein Response (UPR).	[16,19]
E31757	BRG1	Biofilm regulator 1.	[20]
D14520	SKN7	TF involved in the activation of osmotic and oxidative stress.	[22,40]
D07744	YAP-like	TF involved in pH-dependent dimorphic transition and maintaining redox balance.	[22,41]
D20482	GZF1	Inducer of the nitrogen catabolite repression (NCR) genes.	[23,42]
E13948	HSF1	Heat shock transcription factor.	[22,23]
D05041	KLF1	Krueppel-like factor 15, regulates the expression of genes for gluconeogenic and amino acid-degrading enzymes.	[23]
A16841	AZF1	Asparagine-rich Zinc Finger protein, regulates carbon metabolism in yeast and cell wall organization.	[20,23]
F05896	DEP1	Part of the Rpd3C(L) histone deacetylase complex (HDAC). Transcriptional modulator involved in regulation of structural phospholipid biosynthesis genes.	[20,23]
C19151	CAT8	CATabolite repression TF 8. Inducer of gluconeogesis, maintains energy homeostasis, presumed to regulate formate dehydrogenases expression.	[39]
F21923	ADR1	Alcohol Dehydrogenase II synthesis Regulator, inducer of genes involved in alternative carbon utilization upon glucose starvation.	[11,39]
D20460	LAC9	LACtose regulatory protein, controls induction of the lactose-galactose regulation.	[23]
B14443	JMC2	JmjC domain family histone demethylase, promotes global demethylation of H3K4.	[23,43]
D02783	DAL81	Positive regulator of genes in multiple nitrogen degradation pathways, involved in nitrogen catabolite activation of transcription from RNA polymerase II promoter.	[23]
B19602	MGF2	Mycelial growth factor 2. TF of known roles in the dimorphic transition.	[16,41]
B21582	MHY1	Mns2/Mns4-like protein, a key regulator of yeast-to hypha dimorphic transition but not stress resistance, regulates both alkaline-pH and glucose-induced filamentation.	[44,45,46]
D01573	MGF1	Mycelial growth factor. Potential driver of the transition between morphological phases.	[16]
B05038	ZNC1	Zinc finger transcriptional factor, regulates the yeast-to-hypha transition in the dimorphic yeast.	[47]
F17886	GZF2	GATA—binding zinc finger transcription factor 2, inducer of NCR, essential for growth on simple nitrogen sources.	[42]
C22682	GZF3	GATA-zinc finger transcription factor 3, repressor of NCR.	[20,42]
E05555	GZF4	GATA-zinc finger transcription factor 4. Putative: inducer of NCR.	[42]
E16577	GZF5	Non-genuine GATA-zinc finger transcription factor.	[20,42]
E03410	ERT1-2	Positive regulator of gluconeogenesis.	[20]
E27742	GCN4	General Control Non-derepressible 4 TF. Key transcriptional activator of amino acid biosynthesis genes.	[48]
F17424	HAP1	TF responsible for oxygen sensing and signaling.	[49]
D04785	SFL1	Repressor of filamentous growth and flocculation.
F01562	EUF1	Transcription factor mediating expression of erythritol synthesis genes.	[11,50,51]
C16863	SKO1	ATF/CREB family transcription factor, repressor that mediates HOG pathway-dependent regulation of osmotic stress response, involved in protection from oxidative damage.	[40]
C14784	YAS3	Transcriptional repressor of ALK genes, de-repressed on alkanes.	[52]
C13750	MSN4	General stress response, regulates tolerance to acid-induced stress. Regulates genes involved in the antioxidant cellular response.	[53]
B13640	RIM101	pH-response transcription factor, regulator of alkaline-induced filamentation	[46]
C12364	NRG1	NRG1 repressor of erythritol utilization genes, plays a minor role in repression of filamentation.	[50,54]
D23749	ZAP1	ZAP1 involved in zinc ion homeostasis by zinc-responsive transcriptional regulation.	[20]
E30789		Putativea: transcriptional regulator of form adherence 5.
C07821		Putativea: glucose transport transcription regulator RGT1-related.
C18645	ARO80	Transcription activator required for the expression of genes involved in the catabolism of aromatic amino acids.	[55]
B08206	CRF1	Copper resistance protein transcriptional regulator.	[56]
D14872		Putative a: transcriptional regulatory protein STB4
E18304	ERT1-1	Transcription activator of gluconeogenesis.	[20]
E07942	MIG1	Controls genes involved in beta-oxidation, involved in carbon catabolite repression.	[57]
A18469	HOY1	Homeobox protein, a positive regulator of hyphae formation.	[20,58]
B06853	PUT3	PUT3 proline utilization trans-activator.
B09713	PPR1	PPR1 pyrimidine pathway regulatory protein, de novo biosynthesis.	[54,59]
D13904	LEU3	Leucine-responsive transcription regulator, regulates genes involved in branched-chain amino acid biosynthesis and ammonia assimilation.
D23045	AHR1	AHR1 adhesion and hyphal regulator 1.
E10681	WAR1	Weak Acid Resistance transcription factor.
E20449	YOX1	Homeobox protein YOX1. Transcriptional repressor of ECB-dependent genes (early cell box) to the G1/M phase.
F03157	MET32	MET32 auxiliary transcriptional regulator of sulfonate and sulfur amino acid metabolism, methionine biosynthesis, and sulfate assimilation.
F05104	TFIIIA	PZF1 general transcription factor IIIA.
F09361	U4/U6.U5 component	U4/U6.U5 tri-snRNP, spliceosomal complex—may play a role in mRNA splicing.
F09493	SAP61	Pre-mRNA-splicing factor sap61, involved in mRNA splicing, associates with cdc5 and the other cwf proteins as part of the spliceosome.
F16599	STB4	Putative transcription factor STB4—Sin Three Binding protein, involved in the transcription of transmembrane transporters.	[60]
F18788	RLM1	May function as a TF downstream of MPK1, at least some RML1 target genes are involved in cell wall biosynthesis.
E31845	PRZ1	Involved in the regulation of calcium ion homeostasis.
B15818	SterTF	Sterol transcription factor, regulating sterol biogenesis.	[61]
D05005	SEF1	Zn2-Cys6 transcription factor; regulates iron uptake.
D12628	POR1	Primary oleate regulator 1—transcriptional activator regulating genes involved in fatty acid utilization.	[62]
C02387	YAS1	Transcription factor essential for cytochrome p450 induction in response to alcanes, heteromeric Yas1p/Yas2p complex transcription factor.	[63]
B08734	REI1	Cytoplasmic pre-60S factor REI1 involved in maturation of the ribosomal 60S subunit.
E01606	OAF3	Oleate activated transcription factor 3, transcriptional inhibitor with a significantly increased number of target genes in response to oleate.
C06842	MCM1	MCM1 transcription factor involved in biofilm formation, cell adhesion, and hyphal growth.
A19778	MBP1	MluI-box Binding Protein, involved in regulation of cell cycle progression from G1 to S phase.	[39,53]
D15334	CPH2	Transcription factor that positively controls filamentous growth.
D01463	CRZ1	CRZ1 transcription regulator involved in the regulation of calcium ion homeostasis.
D13068	BUD20	BUD20 bud site selection protein 20—positioning the proximal bud pole signal; protein required for ribosome assembly.
B13354	AP-4	Transcription factor that activates viral and cellular genes.
F25861	RPN4	RPN4 transcription factor regulating proteasomal genes.
D24167	CBF1	CBF1 centromere binding factor 1, required for chromosome stability and chromosomal segregation.
C12639	SWI6	SWI6 part of a complex involved in cell-cycle-dependent transcription. SWI4 and SWI6 are required for formation of the cell-cycle box factor-DNA complex
E25960	SWI1	SWI/SNF chromatin-remodeling complex subunit SWI1.
F13321	OAF1	Oleate activated transcription factor 1, activates transcription of genes involved in fatty acid beta-oxidation.
F11487	STP1	Nutrient- and stress-responsive activator of ribosome biogenesis genes.
B05478	STP3	Possibly involved in pre-tRNA splicing and in uptake of branched-chain amino acids.	[59]
E15510		Putative a: Homeobox protein YOX1.
C22990		Putative a: ASG1 general activator of stress genes.
D04466		Putative a: Regulatory protein cys-3. Turns on the expression of structural genes which encode sulfur-catabolic enzymes.
D09647		Putative a: Arginine metabolism regulation protein II. With ARG80, ARG82 and MCM1, coordinates the expression of arginine anabolic and catabolic genes in response to arginine.
D10681		Putative a: Adhesion and hyphal regulator 1.
D18678		Putative a: Respiration factor 2. Transcription factor that regulates expression genes required for glycerol-based growth and respiration.
E17721		Putative a: Phosphatidylinositol N-acetylglucosaminyltransferase subunit P.
E31669		Putative a: Metal-binding activator 1. Copper ion-sensing transcription factor, promotes filamentous and invasive growth.
F05126		Putative a: Phosphorus acquisition-controlling protein.
F17468		Putative a: Multidrug resistance regulator 1. Acts as the central regulator of the MDR1 efflux pump.
E11693		Putative a: TF required for repression of genes during iron starvation. Represses iron-dependent and mitochondrial-localized activities including respiration, TCA cycle, amino acid metabolism, iron-sulfur-cluster and heme biosynthesis.
E10131		Putative a: Transcription elongation factor 1. Implicated in the maintenance of proper chromatin structure in actively transcribed regions.
D09757		Putative a: AP-1-like transcription factor.
F05346		Putative a: Binds a palindromic promoter element essential for induction of fungal cutinase gene.
C11858		Putative a: DNA-binding transcription factor Moc3.
B15312		Putative a: May act as a co-chaperone for HSP70.
C18667		Putative a: Transcriptional activator of the arabinanolytic system.
E16973		Putative a: Metallothionein expression activator.
F11979		Putative a: Retrograde regulation protein 1.
C19063		Putative a: Transcription factor that mediates stress and developmental response.
F18326, E05577, B08360, C20977, C13794		Putative a: sterol uptake control.

^a assignment of genes’ names and genes’ functions was manually curated by cross-referencing several databases—anther, KEGG, UniProt, and NCBI and blasting (blastp) the sequences against the databases’ collections; The following TFs have no assigned function in the databases mentioned above: A10637, A12925, B00660, B20944, C01375, C03564, C05995, C09009, C09482, C13178, C15202, C16390, D01353, D01419, D06952, E17215, E18161, E18656, E24277, E31383, F06072, F13695, F15037, D09757, B04510, F22649, D17988, F03135, B22176, F11011, E19965, F15543, E14971, E20251, F13761, E34925, E23518, B13200, E10087, F16852, D02475, D06193. The references are given for functional studies in Y. lipolytica.

. Both datasets generated previously were expanded due to different sets of TFs scoring significant results in the transcriptomics (87) and functional screens (significant data and tendencies—124). Hence, here analyzed dataset is original in terms of extent and also considering the combination of functional and transcriptomics data.

Based on the overall performance profile comprising both the transcriptomics and functional data, the set of TFs was clustered into 10 clusters (Figure 1; the optimal number of clusters was defined using an elbow method; Figure S1).

As can be read from Figure 1, cluster 1 is the most abundant and contains the most variable set of TFs. It comprises 76 genes exhibiting moderate up-/down-regulation (especially in the HSS strains) or an unchanged transcriptomic profile upon co-OE of any rProt. Their OE leads to an insignificant but generally negative impact on the rProt production phenotype. This cluster comprises several TFs whose function is well-described, and their role in rProt synthesis has been confirmed. For example, Skn7 (D14520) was proved to play a role in maintaining the rProt synthesis capacity under osmotic stress infliction [22]. Here, no transcriptional or functional effects were observed as the conditions were not adequate (no osmotic stress simulation). Cluster 1 is enriched in significantly deregulated TFs of known roles in the dimorphic transition, like MGF2/B19602, MHY1/B21582, MGF1/D01573, and ZNC1/B05038 [47,53,64,65]. Their OE concertedly contributed to a slight decrease in the rProt synthesis, though they displayed an opposite deregulation pattern (two former were downregulated and the two latter upregulated in the HSS strains). Two representatives of the Gzf-family (genuine GATA—binding zinc finger family) known to be involved in nitrogen catabolite repression (NCR) were assigned to this cluster as well (GZF2/F17886—essential for growth on simple nitrogen sources, inducer of NCR; and GZF3/C22682—known repressor of NCR; [42]). But the other members of the Gzf family were clustered in either cluster 2 (GZF4/E05555—strongly upregulated under inorganic nitrogen [42], hence—putative inducer of NCR; and GZF5/E16577—function not investigated) or a self-contained cluster 8 (GZF1/D20482—known inducer of NCR; discussed hereafter). GZF2 and GZF4, although clustered separately, displayed a similar pattern of deregulation (upregulation in HSS strains); likewise, the GZF3 and GZF5 genes; however, minor changes to the OE effect contributed to their separation.

Of note, several other global regulators of carbon/nitrogen metabolism were assigned to cluster 1, i.e., ERT1-2/E03410, GCN4/E27742, ADR1/F21923, CAT8/C19151, and E10087. Strikingly, none of these genes displayed any specific deregulation pattern in the analyzed transcriptomics data, but their OE contributed universally to the inhibition of rProt synthesis (Figure 1 and Figure 2). Furthermore, several of these genes were assigned to a separate functional category upon running a statistical overrepresentation of GO terms test. Their specific function and involvement in rProt synthesis are discussed in Section 2.2. Cluster 1 also contained 4 genes (HAP1/F17424, SKN7/D14520, SFL1/D04785, EUF1/F01562) that were either slightly upregulated in the HSS strains or showed insignificant changes in transcriptomic data. OE of these 4 genes uniformly led to a slight, insignificant decrease in the rProt synthesis capacity. Due to their previously documented importance to Y. lipolytica phenotype modulation [22,23], they will be subjected to further functional studies (Section 2.4).

Delimited cluster 2 is the second largest cluster, containing 43 TF-encoding genes with low/lack of transcriptional deregulation and a functional phenotype similar to the control strain but with a tendency towards an increase in rProt synthesis upon OE. That cluster contains the two previously mentioned members of the Gzf family involved in the nitrogen metabolism regulation, as well as TFs of known involvement in stress response, like SKO1/C16863, YAS3/C14784, MSN4/C13750, RIM101/B13640, or a recently described repressor of erythritol utilization, NRG1/C12364 [50]. Consistently, all the ‘stress response’ TFs were upregulated transcriptionally in at least one UPR strain, which was shown to face severe stress. Another interesting gene from this cluster is ZAP1/D23749 (Zinc-responsive transcriptional regulator 1), involved in zinc ion homeostasis by zinc-responsive transcriptional regulation. It displayed upregulation in the UPR strains, and its OE led to an increase in the rProt synthesis (FC 1.19), which makes it a candidate for a ‘helper gene’ worth further functional studies.

A single-gene cluster 3 is represented by the D01353 gene, which has a unique expression profile. The gene was strongly downregulated upon any rProt OE. Contrary to expectations, its OE had no significant impact on the rProt phenotype. Another single-gene-contained cluster, cluster 4, containing the HAC1/B12716 gene, was delimited mostly because of its strong upregulation in the scYFP-HAC1 strain, induced synthetically [19]. Due to the peculiar, unexpected outcome (lack of induction of rProts synthesis upon OE), this TF was subjected to further functional studies by investigating the Δhac1 phenotype (Section 2.4). Cluster 5 also contains only 1 gene (BRG1/E31757) displaying strong upregulation in HSS and scYFP-HAC1 strains, with no significant change in OE phenotype data. Knowing its role in the synthesis of lipids in Y. lipolytica [20], it is tempting to state that its expression was enhanced in the HSS strains in response to high demands for membranes intensively exploited in the vesicular transportation of rProts. Representatives of cluster 6 (E30789, F18326, and C07821) showed strong downregulation in the UPR strains but no change in the OE phenotype. TFs assigned to cluster 7, DEP1/F05896 and AZF1/A16841, were characterized by minimal changes in transcriptional profile, but both contributed to a distinctive, very strong limitation of rProt synthesis upon their OE. These two were also selected as the candidates for studying Δazf1 and Δdep1 phenotypes with high expectations (Section 2.4).

Cluster 8 contained only 1 gene (GZF1/D20482), which exhibits a unique performance, namely, strong downregulation in HSS strains and high over-production of rProt when OE. Gzf1 is a member of the discussed above Gzf-family, acting as an activator of NCR genes. It was previously identified as one of the universal enhancers of rProt synthesis [22]. Considering its putative role as an NCR activator, such an effect could be attributed to the enhanced nitrogen scavenging capacity required for the high-level production of rProt. But then its significant downregulation in the HSS strains is not clear, as those strains definitely encountered nitrogen limitation due to their extensive consumption for rProt synthesis. A previous report detailing the impact of Gzf family members on growth and lipid synthesis by Y. lipolytica cultivated in the presence of different nitrogen sources sheds some light on the structure of the Gzf TRN [42]. Primarily, they found that expression of GZF1, GZF2, GZF4, and GZF5 was increased when the cells were grown on ammonium as the sole nitrogen source, while the level of expression of GZF3 was not significantly affected. Translating into our data, their expression should be upregulated in the HSS, facing nitrogen depletion. It was true only for the GZF2 and GZF4 genes, while GZF5 escaped this mechanism. The highest amplitude of the upregulation due to nitrogen source type was previously noted for the GZF1. In our transcriptomics, GZF1 displayed a tremendous drop in expression in the HSS strains, which is in contrast to our expectations. Interestingly, [42] showed that when GZF1 was deleted (Δgzf1), no growth defect on any of the nitrogen sources tested was displayed, suggesting that its function was taken over by the other gene. This could be performed by either GZF2 or GZF4, whose expression was significantly upregulated in the HSS strains. It was also discovered that Δgzf3 (repressor of NCR) led to the loss of regulatability of GZF1 by the nitrogen source, i.e., GZF1 expression was elevated irrespectively of the nitrogen provided. The opposite was observed in the Δgzf2 background. In our data, GZF3 was not deregulated under any conditions, which is consistent with the findings by [42]. Though functional GZF3 and GZF2 were present in the cell. Hence, GZF1, which is of key interest to us as the potent inducer of rProt synthesis, could be regulated by the nitrogen availability. Considering increased demands for nitrogen and the upregulation of GZF2, we expected upregulation of GZF1 expression in HSS strains. The question of why GZF1 was downregulated in the HSS strains hence remains to be answered. The effect of OE and Δgzf1 genotypes on rProt synthesis in Y. lipolytica in a direct comparative experiment was investigated and is presented hereafter (Section 2.4).

Representatives of cluster 9 in general exhibited no significant deregulation in transcriptomic data, but their OE led to strong overproduction of rProt. This group of TFs is definitely of the highest practical interest, gathering candidate enhancers of rProts synthesis. Indeed, 2 genes, KLF1/D05041 and HSF1/E13948, were previously reported as global enhancers of rProt synthesis [22,23]. Interestingly, a similar level of rProt synthesis enhancement was observed for TF ARO80/C18645, involved in the catabolism of aromatic amino acids [66]. Likewise, DAL81/D02783, known for its role in nitrogen turnover, was expected to bring a positive effect to rProt synthesis. It is thus concluded that the positive impact triggered by the overrepresentation of Aro80 and Dal81 is associated with the increased supply of the rProt building blocks. By similarity to KLF1’s function, the beneficial effect of CRF1/B08206 OE may be attributed to its involvement in oxidative stress response.

An interesting expression profile was observed for the gene D14872 assigned to a self-contained cluster 10. The gene showed a strong and inverted deregulation trend in the HSS (downregulated) and UPR (upregulated) strains, however, yielding no significant impact on phenotype upon OE. Genes displaying inverted deregulation patterns were deemed to play a direct role in the analyzed biological process (rProt synthesis), which turned out to be not the case, at least for D14872. They will be combined and analyzed separately in Section 2.3.

2.2. Statistical Overrepresentation Test Delimited Two Major Categories amongst TFs

The total set of 140 TFs was subjected to a statistical overrepresentation of biological processes test [67,68]. Expectedly the majority of output records were assigned to ‘regulation of DNA-templated transcription’ or ‘regulation of RNA biosynthetic process’ categories, or related (altogether 31 categories and subcategories comprising a variable number of genes, with a significant enrichment fold from >6 to >17; Table S1).

Interestingly, six genes were assigned to a ‘cellular response to nutrient levels’ category (altogether 8 categories and subcategories comprising a variable number of genes, with a significant enrichment fold from >6 to >13; Table S1). These were transcriptional activators of gluconeogenesis ERT1-1/2 (cluster 9 E18304/ cluster 1, E03410), DAL81 (D02783; cluster 9), GCN4 (E27742; cluster 1), a MIG1-related regulatory protein (E07942; cluster 2), and unknown TF E10087 (cluster 1; Figure 2). It was the only significantly enriched category, other than related to DNA transcription/RNA biogenesis.

The majority of TFs assigned to that category showed no transcriptional deregulation due to rProt OE. Also, the induced phenotypical changes were rather tendencies. Though the patterns behind them are interesting to track. ERT1 acts as a positive regulator of gluconeogenesis and fermentable carbon utilization but mechanically is a repressor of transcription by RNA polymerase II by a nonfermentable carbon source. An increased abundance of ERT1-1 (E18304) was recorded in transcriptomes of scSoA and scYFP-HAC1; when overexpressed (OEd), it triggered enhanced rProt synthesis. The other homolog, ERT1-2 (E03410), did not show any changes at the transcriptional level in the analyzed transcriptomes, and its OE led to a slight decrease in rProt amounts. Another gene from this category, DAL81/D02783, is a positive regulator of genes in multiple nitrogen degradation pathways, involved in nitrogen catabolite activation of transcription from the RNA polymerase II promoter. Even though previously we reported on enhanced protein degradation and deregulation of nitrogen catabolism processes in the UPR strains [16], and enhanced demand for nitrogenous compounds was expected for the HSS strains, its expression level was not changed. However, functional screens revealed its minor but positive effect on rProts synthesis upon OE, which is biologically well-understood. The lack of transcriptional response but a visible effect at the functional level suggests its regulation at the posttranslational level. Another TF from this functional category, MIG1/E07942, is known to be involved in glucose repression by negative regulation of transcription by RNA polymerase II by glucose. Mig1 (together with Mig2) also induces filamentation under glucose starvation stress. The steady-state maintained cells did not face the main carbon source starvation, as its residual levels were very high. MIG1 expression was only slightly upregulated in the HSS strains, expected to be in very high demand for carbon building blocks and energy. Its OE led to a minor positive effect on rProts synthesis by Y. lipolytica. However, due to its known fundamental role in shaping cellular metabolism, the Δmig1 genotype was constructed, and its effect on rProt synthesis was tested (Section 2.4).

All the remaining TFs assigned to the ‘cellular response to nutrient levels’ category (based on enrichment tests or added manually—CAT8 and ADR1) collectively contributed to slightly decreased rProts synthesis upon their OE. Among them, GCN4 (E27742) is a master regulator involved in multiple biological processes, with a primary role in TRN controlling amino acid metabolism. It was also shown to be required for increased biosynthesis of translation precursors such as ribosomal proteins, amino acids, and purines, depending on the external stimuli. Upon amino acid starvation, it plays a key role in the transcriptional induction of almost all genes involved in amino acid biosynthesis (19 per 20 pathways). Notably, it was reported that OE of GCN4 leads to a reduction in protein synthesis capacity via negative regulation of ribosomal protein gene transcription [48]. Consistently with its assigned role, its OE triggered a slight decrease in the rProt synthesis levels in Y. lipolytica. It is known that its levels are strictly regulated at translational/protein and mRNA stability levels (it is constitutively expressed at a low basal level), so the generally observed lack of transcriptional deregulation upon rProts overproduction is not surprising, though the inflicted conditions seem to be relevant stimuli.

TFs CAT8/C19151 and ADR1/F21923 are known to be involved in the regulation of carbon catabolite repression, but in the opposite direction than Mig1. They share the common lack of deregulation upon any rProt OE and the negative effect on rProt synthesis upon OE. Cat8 is an inducer of gluconeogenesis, binding to cis-regulatory elements upon glucose starvation. Adr1 induces transcription of genes involved in alternative carbon utilization upon glucose starvation. Its fundamental role in controlling the Euf1-dependent erythritol utilization cluster in Y. lipolytica has been recently reported [11].

Altogether, the above-discussed transcriptional-functional profiles of the genes assigned to a specific functional category ‘cellular response to nutrient levels’ suggest that enhanced degradation of nitrogenous compounds is beneficial for rProts synthesis (Dal81), but enhanced signaling of nitrogen starvation is not (Gcn4). Likewise, the induction of alternative carbon utilization is detrimental.

2.3. Usefulness of Specific Deregulation Patterns as Selectors of TFs Involved in Specific Biological Processes

To gain more insight into the relationship between the transcriptional profile and the functional data, the dataset was further filtered to extract and group together TFs that: (i) exhibited inverted transcriptional deregulation patterns in HSS and UPR strains (Figure 3); (ii) showed a uniform deregulation profile, irrespective of the inflicted perturbation (type of rProt) (Figure 4). The phenotypic effect of their OE is shown together with their expression profile in Figure 3 and Figure 4.

The inverted deregulation pattern of the TFs (grouped in Figure 3) suggests that they are directly transcriptionally responsive to the conditions faced and differing the two groups of strains (‘HSS’ and ‘UPR’). Considering the previous results, we presumed that the amounts of rProts and the associated background biological processes (oxidative stress and UPR) were the delimiting factors [16,19], ergo, these TFs are directly involved in these processes. Indeed, HOY1/A18469, MHY1/B21582, and SKO1/C16863 are known for their role in stress response, and these were consistently upregulated in the UPR strains. In this context, both the transcriptional response and functional effect on rProts synthesis upon OE of TF B00660 are very interesting, considering its putative role in sterol uptake [20]. As evidenced [61], the enhanced sterols residence in the plasma membrane is important for enhanced stress resistance. The here presented dataset suggests that the overrepresentation of TF B00660 (either natively due to UPR or synthetically by OE) contributes to generally enhanced rProts synthesis, presumably associated with enhanced stress resistance. Unfortunately, many of the remaining genes exhibiting such inverted deregulation patterns are of unknown function, and hence the mechanism behind the phenotypes cannot be deduced.

Generally, two distinct expression profiles could be distinguished for the TFs grouped in Figure 3, those upregulated in the UPR-encountering strains (inYFP, TlG, scYFP-HAC1) and downregulated in HSS (scSoA, scYFP) (majority of the examples), and those upregulated in the former and downregulated in the latter (three TFs: E24277, C18645, and C07821). If the expression profile could be directly translated into the functional outcomes, then two distinct functional behaviors would be expected. However, that was not the case. For D14872, exhibiting significant downregulation in the HSS and significant upregulation in the UPR-facing strains, no functional effect of its OE on rProt synthesis was observed. On the other hand, TFs B00660 and HOY1, while showing the same transcriptional pattern, their OE triggered enhanced rProt synthesis in the host strain (significant or just a tendency). Corresponding enhancement in rProt synthesis was observed for TF C18645, whose expression was significantly enhanced in HSS and decreased in the UPR strains. But foremost, the majority of TFs for which the inverted deregulation pattern was observed (so its direct implication in the rProt synthesis was presumed) did not render any significant functional phenotype when OE (only tendencies; Figure 3). Such an outcome implies that our working hypothesis stating that transcriptomics data for the TFs displaying inverted deregulation patterns could be used as a selection-driver of TF involved in a specific biological process, here—rProt synthesis in Y. lipolytica, should be partly rejected (for TFs displaying such an expression profile). Furthermore, considering that the TFs exhibiting an inverted deregulation pattern exert similar functional outcomes when OE (B00660 and C18645) implies that the transcriptomics pattern cannot be used as a preliminary prognostic of the TF’s mode of implication (inducer/repressor) in the biological process under study.

Another dataset prepared to test our working hypothesis comprised TFs that were responsive to enhanced synthesis of any rProt, irrespective of the biochemical character of the rProt and the background biological processes awakened (Figure 4). These comprised TFs globally up- or down-regulated across all the transcriptomes tested (in at least 3 out of 5). As can be read from Figure 4, global responsiveness to the over-synthesis of rProt is not an optimal prognostic of direct involvement of a TF in rProt synthesis. Indeed, the majority of TFs, when OE, did not exert any significant effect on rProt synthesis. Nevertheless, some significant correlation between the TFs transcriptomics profile and the rProt amounts upon TF’s OE was noted (scSoA, r = 0.6, inYFP, r = 0.56, scTlG, r = 0.63; significant at p < 0.05; Figure S2). It suggests that the TFs that were downregulated in the transcriptomes contribute to a decrease in the rProts synthesis when OEd, and that TFs upregulated promoted rProt synthesis when OEd; so such data could be used as a careful hypothesis driver on the implication of a given TF in the biological process under study (rProt synthesis in this case).

2.4. Direct Comparison of Phenotypes Elicited by Y. lipolytica Strains Bearing OE or KO of Selected TFs

In the next step, 11 TFs representing different clusters from Figure 1 were further investigated through direct comparative phenotype reading of Y. lipolytica strains bearing their OE and KO, along with rProt OE. Altogether 11 TFs were studied using this approach, selected based on the observed transcriptional-functional profiles (Figure 1) and literature data. These were representatives of the most abundant cluster 1 (HAP1/F17424, SKN7/D14520, SFL1/D04785, EUF1/F01562), cluster 2 (MIG1/E07942), cluster 4 (HAC1/B12716), cluster 7 (DEP1/F05896, AZF1/A16841), cluster 8 (GZF1/D20482), and cluster 9 (KLF1/D05041, HSF1/E13948). After genotype construction, selected subclones were cultivated under ‘standard conditions’ (no stress factor inflicted; Section 4.1 and Section 4.2). Results on the total rProt amounts, or normalized per biomass, are shown in Figure 5.

Based on the transcriptional-functional performance patterns (Figure 1), the representatives of cluster 1 (HAP1/F17424, SKN7/D14520, SFL1/D04785, EUF1/F01562) and cluster 2 (MIG1/E07942) were subjected to further studies with rather low expectations (no transcriptional deregulation, no effect of the TF’s OE on rProt synthesis). Their selection was based solely on literature data and their important role in shaping the Y. lipolytica phenotype [11,22,23,40,49,50,51,57,69,70,71]. Strikingly, the deletion of TFs from cluster 1 led to moderate (55 to 76%) but significant (p < 0.05) improvement in the rProt synthesis. Suggesting that the observed tendencies in the functional screens (Figure 1) are indeed valid observations. No significant effect (nor even a tendency) was observed when MIG1 was either OE or KO (also selected based on literature data and not experimental evidence provided here [57,72,73]). It is presumed that the OE/KO MIG1 strains should be tested under a spectrum of different nutrient types and levels to observe a valid phenotype.

Hap1 is known to play a fundamental role in managing aerobic metabolism in S. cerevisiae. It is responsible for sensing the oxygen levels via the heme signaling pathway, and activation of the oxidative stress response genes [49]. Notably, its OE had a beneficial effect on rProt synthesis in S. cerevisiae [74]. It was suggested that the effect was caused by diminishing oxidative stress responses awakened by intensive rProt synthesis. The contradictory effects observed here with Y. lipolytica may result from a different metabolism of the two species regarding oxygen demands.

Skn7 is a TF involved in protein secretion and activation of oxidative and osmotic stress responses [75,76]. Previously, its KO in Y. lipolytica had a detrimental impact on the strains’ resistance to osmotic stress, irrespective of inflicted anaerobiosis stress, pH, or adopted temperature, leading to a severe underperformance in terms of rProt synthesis once osmotic stress was inflicted [22]. On the other hand, its OE enabled the maintenance of rProt synthesis under high osmolarity and low aeration [22]. Under the current experimental setup (no stress factor, high aeration), the Δskn7 phenotype performs better than the control in terms of rProt synthesis. We presume that the adopted conditions contributed to that outcome. A corresponding effect was observed when TF SFL1/D04785 was deleted (Figure 5). Sfl1 (Suppressor Gene for Flocculation) acts as a repressor of filamentous growth and flocculation, antagonizing FLO gene action; it is also known to activate the expression of stress-responsive genes [69,70,71]. The beneficial effect of Δeuf1 is also not yet clear. Euf1 is a master regulator of the ‘erythritol utilization cluster’ [11,50,51] known to govern the expression of several genes involved in erythritol utilization once glucose or glycerol are depleted. Considering the effects observed after KO of TFs from cluster 1, the only explanation for the moment is a slight relief of the transcriptional–translational machinery due to the limited expression of several genes or inadequate conditions enabling a relevant phenotype expression. Collectively, it is tempting to state that once stress conditions are not inflicted, deletion of the global stress-response regulator is beneficial for the costly biological process—rProt synthesis.

The observed Δhac1 and HAC1-OE phenotypes escaped our expectations, yielding no significant changes. Hac1 mediates the deregulation of hundreds of genes involved in polypeptide formation, folding, and maturation, but also in lipid synthesis, membrane expansion, and many others [17,18,77,78,79]. Hence, along with Hsf1, it is the most frequent target of genetic modifications for enhanced rProt synthesis in yeast [15,19,80,81,82,83,84,85]. As presented in Figure 1, the transcriptional profile of HAC1 depends on the type of rProt being synthesized due to the background molecular mechanisms. Generally, previously conducted estimations of the rProt amounts synthesized when HAC1’s expression was elevated or not showed that the yields of rProts were higher when HAC1 expression was silenced [14,16]. In the following studies, we attempted to explain this observation. Investigation of Hac1 regulome [19] indicated that elevated Hac1 presence induces massive vacuolar proteolysis. In contrast, in other experiments, we observed a ~30% increase in the rProt synthesis (small, non-modified post-translationally protein, targeted for secretion; scYFP) [15]. The following studies, conducted with cells maintained in the steady state, proved that the effect was indeed bifurcated, causing a nearly 7-fold drop in the retained fraction of rProt but promoting its secretion by nearly 2.5-fold [19]. The protein used in the functional screens (results presented in Figure 1 and Figure 5) has a similar biochemical characteristic to that former but is destined for intracellular retention. We presume that it is the cause of the surprising lack of effect from Δhac1 and HAC1-OE strains.

Comparable functional outcomes regarding rProt amounts were expected from strains deleted for AZF1/A16841 and DEP1/F05896 (cluster 7), as both the OE strains displayed a significant decrease in rProt synthesis capacity (Figure 1). However, a directly inverted phenotype was observed solely for the DEP1-deletant (Δdep1), with no effect from the Δazf1 strain (at least in terms of rProt synthesis) (Figure 5). Notably, the Δdep1 rendered the highest increase in total rProt or normalized rProt measures, making this modification the most successful example presented here. Dep1 (Deregulated Expression of Phospholipid biosynthesis) was previously found to enhance the accumulation of lipids in Y. lipolytica [20] and specifically—activate phospholipid biosynthesis in Fusarium sp. [26]. In contrast, [86] identified Dep1 as a repressor of phospholipid synthesis genes (e.g., INO1, CHO1, and OPI3) in S. cerevisiae. It seems that Dep1 from Y. lipolytica operates in a similar way as in Fusarium rather than the model yeast species. To our interpretation, an interplay between globally promoted carbon metabolism and enhanced nitrogen scavenging (like in the case of Gzf1/Gzf2+Gzf4 discussed above) may be the molecular mechanism behind the Δdep1 phenotype.

For the second representative of cluster 7, AZF1/A16841 (asparagine-rich zinc finger protein), such a directly inverted phenotype regarding rProt synthesis was not observed. Azf1 is a carbohydrate-sensing TF, and in the presence of glucose, it activates genes involved in growth, carbon metabolism, and filamentation in S. cerevisiae and O. polymorpha [25,28]. In Y. lipolytica, the Δazf1 mutation rendered no aberrant phenotype in terms of rProt synthesis. This observation is consistent with our previous notions about this phenotype investigated under different cultivation conditions [16,19]. As in the case of Mig1, we presume that alternating carbon sources would enable a relevant phenotype display (but maybe not necessarily related to rProt synthesis).

An unexpected effect was observed upon direct comparison of the strains with either OE or KO of the ‘universal rProt synthesis enhancers’, KLF1/D05041, GZF1/D20482, and HSF1/E13948. Surprisingly, the rProts synthesis capacity was enhanced regardless of whether the genes were OE or KO. Gzf1 was selected for further phenotype studies due to its surprising expression pattern in the HSS strains and highly positive effect on rProt synthesis. Considering the data presented in Figure 1 (decreased expression in the HSS producing a high amount of rProts and enhanced production of rProts when GZF1 was OE), the data presented in Figure 5 stay in peculiar agreement—if GZF1 is downregulated (HSS strains), rProts synthesis is high, and when GZF1 is upregulated (OE)—rProt synthesis is also high. Previous analysis of OE/KO GZF1 phenotype under various environmental stress factors infliction [22] demonstrated that OE of GZF1 led to very high and universal enhancement in rProt synthesis. The effect of GZF1 KO was less uniform—under specific conditions, growth was limited, but some indications for enhanced rProt synthesis capacity were revealed, but only under specific conditions not applied here (low oxygen and hyperosmolality) [22,23]. The primary difference between those previous and current studies relates to cultivation parameters (type of vessel, aeration, buffering system), and foremost—the type of nitrogen source used, which is a relevant factor affecting GZF1 activity (previously—ammonium sulfate; at present—a 1:1 mixture of glutamic acid and ammonium sulfate). Since that former study, our protocol for Y. lipolytica has been greatly improved [31]. We presume that this new cultivation system allows for the actual genotype-driven phenotype development, and the type of nitrogen source used accounts for the observed discrepancies in the Δgzf1 phenotype.

Representatives of cluster 9, KLF1/D05041 and HSF1/E13948, did not display any significant transcriptional deregulation profile in the analyzed transcriptomes, but their OE led to significant enhancement in rProts synthesis capacity (Figure 1). When deleted, Δklf1 showed no changes vs. the control strain in terms of rProts production. The same effect was observed previously, using a slightly different cultivation protocol [23]. Klf1 was shown to be responsive to external nitrogen levels in Schizosaccharomyces pombe, and its regulome was implicated in cell wall renewal, oxidative stress response, glycolysis, nutrient uptake, RNA-mediated chromatin silencing, glycosidation, and methylation [87]. It was demonstrated to specifically react to oxidative stress via relocalization to the nucleus [88]. That would explain why the rProt synthesis-promoting effect of KLF1 OE could be seen, even though its expression remained at the control level (Figure 1).

HSF1/E13948, as the key regulator of global stress response, is the most intensively studied TF in terms of its effect on rProt synthesis in yeast. rProts production was significantly enhanced in S. cerevisiae upon OE of HSF1 in its native or a constitutively active form, HSF1-R206S [89,90,91]. Previously, we demonstrated a universal, promoting effect of wild-type HSF1 OE on rProt synthesis in Y. lipolytica [22,23]. Genotype Δhsf1 displayed severely impaired growth under both ‘stressful’ and ‘optimal’ growth conditions. The effect was associated with limited production of rProts, but mainly when harsh stress was inflicted. The results here obtained were hence surprising and difficult to explain even with the differences in the growth conditions. Definitely, more studies focused on this aspect must be conducted to get an insight into the molecular mechanisms behind that observation, specifically, whether the other global regulators of stress response were not induced in the Δhsf1 Y. lipolytica strain.

3. Summary and Conclusions

Integration of multidimensional, essentially diverse biological data imposes a challenge, as the biological processes are rarely linear. However, with the current accumulation of biological knowledge, such a data-recycling approach is reasonable and, if followed with adequate methodology, may drive novel hypotheses and interesting findings that otherwise could be missed. The approach has become popular with the model organisms, and now, considering how much information is stored in public repositories, it is feasible with nonconventional species like Y. lipolytica.

In our recent studies, we investigated the possibility of using TFs as tools for global optimization of rProts production in Y. lipolytica. Here, we recycled and integrated datasets related to the expression profile of TFs and the functional effect of their OE on rProt synthesis. Specifically, we were testing a hypothesis of whether transcriptomics data could be used as a selection driver of TF involved in this particular biological process.

The conducted analyses showed that many of the TFs with previously documented implications in rProt synthesis did not respond at the transcriptional level (e.g., Skn7, Hap1). We presume that it was due to conditions under which the transcriptomics samples were collected, which were not activating a given TF’s TRN. This statement is corroborated by the results of further experimental studies. The deletion of TFs from cluster 1 (showing no transcriptional deregulation upon rProt synthesis) led to significant improvement in the rProt synthesis capacity. Likewise, representatives of cluster 7, showing no transcriptional response to rProts synthesis, were shown to strongly decrease this process when OE. Specifically, the deletion of DEP1 rendered the highest increase in rProt production, demonstrating on one hand its direct implication in this process and on the other—making this modification the most successful example presented here.

The direct correlation between the transcriptional profiles and functional effects of the TF’s OE was either not significant or weak. It was found true only for the TFs showing a uniform pattern of deregulation. In that specific case, TFs that were downregulated in the transcriptomes contributed to a decrease in the rProt synthesis when OEd, and TFs upregulated—promoted rProt synthesis when OEd. Hence, such data could be used as a careful hypothesis driver on the implication of a given TF in the biological process under study (rProt synthesis in this case). In contrast, examples of TFs showing different deregulation patterns but the same functional effect (e.g., TFs of known roles in the dimorphic transition, or B00660 vs. C07821 or C18645) and the opposite behavior (similar transcriptional pattern but different functional effect; e.g., HOY1 vs. MHY1) were predominant. The intrinsic pleiotropic activity of TFs is the key explanation for this observation (e.g., Gcn4, which is an activator and a repressor).

Nevertheless, the presented data integration led to a new hypothesis and pointed to new directions for experimental studies. For example, the juxtaposition of transcriptional profiles and functional OE results highlighted the putative importance of TF ZAP1/D23749 for rProt synthesis in Y. lipolytica. Likewise, C18645, a TF of yet undefined function, was clustered together with the most potent rProt enhancers Hsf1 and Klf1, identified previously. A new research question that remains to be answered through further insightful experimentation is why GZF1, an efficient enhancer of rProt synthesis, was indeed transcriptionally downregulated in the strains producing rProts in high quantities. Though it is known to be an activator of NCR. The analysis of transcriptional profiles and functional outcomes of the genes involved in nitrogen metabolism in Y. lipolytica implied that the enhanced degradation of nitrogenous compounds is beneficial for rProts synthesis (e.g., Dal81, Aro80), but enhanced signaling of nitrogen starvation is not (Gcn4). Likewise, the induction of alternative carbon utilization is detrimental (e.g., Adr1, Cat8, Ert1-2).

Altogether, this investigation demonstrates the feasibility and potential of biological data recycling and integration for Y. lipolytica. It appears that the amount of knowledge accumulated is sufficient, and such an approach leads to novel findings that were not discovered when the datasets were analyzed individually.

4. Materials and Methods

4.1. Microbial Strains Used for Transcriptomics and Functional OE Screens Data Acquisition

Strains used for transcriptomics and functional OE screen data acquisition were constructed previously [14,16,19,20,23,92]. Briefly, the ‘transcriptomics’ strains are all derivatives of Po1h (Genotype: MatA, ura3-302, xpr2-322, axp1-2, leu2-270::LEU2, Phenotype: Ura-, ΔAEP, ΔAXP, Suc+). The control strain (Po1h_Ura3) was transformed with a solo URA3 marker cassette to generate a prototroph. Recombinant Y. lipolytica strains were transformed with Golden Gate Assembly [93,94] cassettes bearing a single transcription unit each with one target gene: alpha-amylase—SoA, glucoamylase—TlG, fluorescent YFP (intracellular in-/secretory sc-), expressed individually under the control of a synthetic hybrid promoter 4UASpTEF, or co-OEing scYFP and HAC1 under the control of the pTEF promoter [14,16,19]. Secretory enzymatic reporters were transcriptionally fused with a signal peptide sequence native to exo-1,3-beta-glucanase (B03564) [95]. Strains for the ‘phenotype’ screens were constructed by [20,92]. Briefly, a collection of 125 Y. lipolytica strains OEing individually one of the TFs and a reporter protein (RedStar2) was created in the background of the JMY2566 (MATa, ura3::pTEF-RedStar2-LEU2ex-Zeta, leu2-270, xpr2-322) strain. Both genes were cloned under a constitutive promoter pTEF and were integrated in a zeta platform at the URA3 locus. Strain JMY2810 (complete prototroph OEing the reporter r-Prot solely; MATa, ura3::pTEF-RedStar2-LEU2ex-Zeta-URA3ex-pTEF, leu2-270, xpr2-322) was used as the control for data normalization and calculating FC values. The methodology of the above indicated strain cultivation strategy, sample acquisition, and data analysis was detailed previously [14,16,23,31,96].

4.2. Microbial Strains Used for Functional Studies—KO in a Specific TF loci

Microbial strains used for functional studies were partly constructed previously (Gzf1, Hsf1, Skn7, Klf1, Azf1, Dep1) [22,23], and partly were constructed originally in this research (Mig1, Euf1, Sfl1, Hap1, Hac1). All the strains were (re)cultured under the improved cultivation protocol [23,31] to gain a direct comparison of the phenotypes.

Construction of the KO Strains—Deletion Cassette and CRISPR-Cas9 Targeting

Standard molecular biology protocols were followed in this study [97]. The deletant strains (KO-TF) were constructed in the background of the JMY2810 strain by disrupting the indicated TF locus. The deletion cassettes were designed using a GoldenGate scaffold indicated in [93] limited to three fragments cloning: (i) ARM up, (ii) NATr (nourseothricin), and (iii) ARM down, flanked with A and B and C and M overhangs. DNA fragments to be cloned were amplified using Phire Hot Start II DNA Polymerase (Thermo Fisher Scientific, Waltham, MA, USA). The cassettes were assembled using a previous protocol for the GoldenGate reaction [93]. White colonies were verified for correctness of the assembly by PCR of adjacent elements and restriction digestion of isolated plasmids. After release from the pSB1A backbone by NotI digestion, the cassettes were used for the transformation of Y. lipolytica JMY2810.

To increase targeted integration, the CRISPR-Cas9 vector JME4580 [98,99] with the TF-targeting sgRNA oligonucleotide was co-transformed along with the disruption cassette. The methodology followed the previously described protocol [22]. The sgRNA oligonucleotides were designed using the CRISPR design tool from the Benchling platform (https://benchling.com/ (accessed on 31 July 2024Day)). Targeting regions were selected close to the center of the coding sequences. sgRNA oligonucleotides with the highest efficiency scores and lowest number of off-target sites were selected. The 20 bp long target sequences were flanked with BsmBI recognition sites and with 4-bp overhangs, enabling their correct integration in a plasmid [99]. Before cloning, the sgRNA oligonucleotides were annealed and then ligated to the JME4580 plasmid using the GoldenGate thermal profile, using BsmBI and T4 DNA ligase from New England Biolabs (NEB Ltd., Ipswich, MA, USA). The reaction was then transformed into E. coli DH5alpha, and transformants were selected on LB ampicillin agar plates. White clones were screened for correctness by PCR and restriction digestion with BglII (Thermo Fisher Scientific, Waltham, MA, USA). Correct constructs were propagated, plasmids were isolated (Plasmid Mini, A&A Biotechnology, Gdynia, Poland), and used for transformation. Y. lipolytica strain was transformed using a standard lithium acetate method [100]. Transformants were selected as instructed by [99], so the co-transformation reactions were inoculated into 9 mL of YPD-hygromycin-nourseothricin liquid medium and cultured at 28 °C for 48 h with shaking at 150 rpm. The following selection conditions were used: hygromycin B at 400 μg L⁻¹, or nourseothricin (both from Sigma-Aldrich-Merck-Millipore, Darmstadt, Germany) at 250 μg L⁻¹, supplemented to YPD medium (liquid or solidified). One mL of such cultures was then transferred into 9 mL of YPD-nourseothricin medium and incubated at 28 °C for 24 h with shaking at 150 rpm to allow plasmid curing (dropping-off JME4580). Finally, the culture was diluted and plated on YPD-nourseothricin agar. Clones appearing after 48 h of incubation at 28 °C were verified for correct integration of the deletion cassettes by PCR and sequencing. All the strains were deposited as 30% glycerol stocks at −80 °C for long-term storage.

4.3. Functional Studies—Direct Comparative Study of OE and KO Strains

4.3.1. Cultivation Conditions

The yeast strains were revived from glycerol stocks and then routinely maintained at 28 °C in rich YPD (g L-1: yeast extract, 5 (BTL, Łódź, Poland); peptone, 10 (BTL); glucose, 20 (POCH, Gliwice, Poland); solidified with agar, 15 (BTL)) or in minimal YNB medium (g L-1: glucose, 10 (POCH); yeast nitrogen base, 1.7 (Sigma-Aldrich, St. Louis, MI, USA); ammonium sulfate, 5 (POCH); solidified with agar, 15 (BTL)). Two to six subclones were cultivated in biological duplicates according to a high-throughput cultivation protocol developed in our laboratory [23].

The cultivation medium and conditions were as follows [g L-1]: yeast nitrogen base without amino acids and ammonium sulfate, 5.1 (Sigma-Aldrich); ammonium sulfate and glutamic acid, both at 7.5 (POCH); glucose, 35 (POCH); buffered with 0.2 M maleic acid at pH 5.0 (POCH). Precultures were developed in media composed of a yeast nitrogen base, 5.1 (Sigma-Aldrich); ammonium sulfate, 15, (POCH); glucose, 25 (POCH); buffered with 0.2 M maleic acid at pH 5.0. All the media were filter-sterilized with 0.22-μm filters (Merck-Millipore, Darmstadt, Germany).

4.3.2. Samples Analysis

Samples were withdrawn at 48 h and assessed for optical density (OD) and fluorescence (FL). Before reading the FL from the reporter protein (RedStar2), samples were diluted in 0.75% NaCl (POCH) to match a linear range of the methods. Absorbance was measured at 600 nm in transparent 96-well plates (Costar; Merck-Millipore, Darmstadt, Germany). FL was determined under ex/em wavelengths 550/595 nm in black opaque plates (Thermo Fisher Scientific). Both measurements were performed using a Tecan Spark automatic plate reader (Tecan Group Ltd., Mannedorf, Switzerland).

Fold change values were calculated by dividing the raw readouts for the TF-engineered strain by the result read for the control strain (overexpressing only in RedStar2 protein, with no TF modification).

4.4. Omics and Functional Data Acquisition

Transcriptomics data were extracted from the NCBI SRA database (PRJNA701856 and PRJNA869113). Data were filtered for gene identifiers of the TF-encoding genes based on [20]. For the non-significantly deregulated genes, a default value of 0 was set.

Phenotype OE screen data were extracted from the YaliFunTome database (https://sparrow.up.poznan.pl/tsdatabase/, accessed throughout June 2024) and accompanying publication [23]. The phenotype data were chosen from a variant of high oxygen availability, pH 5, temperature 28 °C, glucose as a carbon source, and the average of ammonium sulfate (AS) and casamino-acid hydrolysate (CH) as nitrogen sources.

4.5. Data Processing and Mathematical Analysis

The results for each TF gene are presented in the form of a heatmap. K-means clustering was performed on these data, with the number of clusters = 10 (determined by an ‘elbow method’, Supplementary Material—Figure S1), random state = 42, scaled with StandardScaler (sklearn package version 1.3.2). The genes are sorted according to clusters, to which they have been assigned. Clusters are separated with bold horizontal lines (Figure 1). All the analyses were performed in Visual Studio Code (version 1.92.2) in the Python programming language (version 3.11.9) with corresponding data-processing and machine-learning packages.

The correlation matrix for the TFs displaying a universal deregulation pattern in the transcriptomics datasets (presented in Figure S2) was calculated based on the TFs’ transcriptional profile in the five strains and on TF-co-OEing (each in two parameters: total FC rProt synthesis and normalized to biomass FC rProt synthesis). The statistical significance of each correlation was calculated with the Pearson r-coefficient at a p-value < 0.05.

A statistical overrepresentation test was run using Panther [67,68]. Analysis Type: PANTHER Overrepresentation Test (Released 20240226). Annotation Version and Release Date: GO Ontology Database DOI: 10.5281/zenodo.10536401 Released 2024-01-17. Analyzed List—YALI signatures of 139 TFs under study (Table 1). Reference List—The reference gene list for the test. Y. lipolytica (all genes in database, 6448), 138 user-defined IDs were mapped to the genome. Annotation Dataset—GO biological process complete—Complete GO biological process annotations, including both manually curated and electronic annotations. Electronic annotations were generated by computer algorithms based on sequence similarity. This is the test that was performed using Fisher’s exact test with FDR correction.

Statistical significance of the differences observed between OE and KO strains vs. the control strain (JMY2810) was assessed by the ANOVA analysis of variance and post hoc Tukey test at a significance level of p < 0.05 (Visual Studio Code, version 1.92.2).