Transcriptomic shift in ethanol and amino acid metabolic genes regulated by Med15 during alcoholic fermentation Authors: David G. Cooper, Emma Grunkemeyer, Jan S. Fassler Published: 2026 ABSTRACT AbstractOrganisms that thrive in extreme environments provide natural experiments in evolution, revealing the genetic regulators that orchestrate complex phenotypic change. Wine yeast (WY) are specialized strains that are adapted to survive in the wine making environment while producing high concentrations of ethanol. In addition to large genomic changes that differentiate WY from yeast used in other industries, SNP and polyglutamine tract polymorphism in the transcriptional regulator Med15 are associated with the fermentation efficiency and stress response phenotypes of WY. In this study, we investigated the transcriptional differences during wine fermentation in transgenic lab strain yeast having integrated WY MED15 alleles. Compared to the unmodified lab strain (MED15LAB), the same strain in which the MED15 locus was replaced with a MED15 allele from yeast isolated from palm wine, the fermented sap of palm (oil, date, coconut) trees (MED15WY23), exhibited enhanced expression of amino acid biosynthesis genes as well as stress resistance and metabolic adaptation genes. Our experimental data confirm the role of arginine in efficient fermentation and suggest that certain MED15 alleles alter the expression patterns of arginine pathway genes in some cases improving carbon flux under nitrogen stress. The global benefits conferred by natural polymorphisms in a single transcriptional regulator highlight Med15 as a target for engineering of strains devoted to various types of alcohol production. Impact Statement This study identifies polymorphisms in the transcriptional regulator Med15 as one key modulator of fermentation efficiency in wine yeast (WY). By integrating WY MED15 alleles into laboratory strains, we demonstrate changes in expression of genes involved in glycolysis, fermentation and amino acid biosynthesis – particularly those in the arginine biosynthetic pathway. These findings provide mechanistic insight into how single-gene variation can drive complex phenotypic traits in industrial fermentation. The work highlights Med15 as a promising genetic target for engineering yeast strains optimized for diverse alcohol production environments, offering broad applications in biotechnology and microbial genomics and validates this approach for identifying broad drivers of evolutionary change. Data Summary All Illumina RNA-Seq datasets are available in the National Centre for Biotechnology Information under BioProject accession PRJNA1320546. Data analysis is summarized in a Microbiology Society Figshare submission at https://doi.org/10.6084/m9.figshare.32791668 and in a GitHub repository with persistent link https://doi.org/10.5281/zenodo.17047964. Introduction The production of wine requires yeast strains that can efficiently convert sugars to ethanol while also surviving the stresses of the grape juice/must environment including high osmolarity, low pH, low nitrogen availability and ultimately high ethanol concentrations. The general transcriptional regulator Med15 influences both fermentation efficiency and stress response during grape juice/must fermentation [1], and thus the MED15 gene is a good candidate for domestication niche-specific adaptive mutations leading to improved fermentation observed in wine yeast (WY) [1, 2]. Variations in alleles of the general transcriptional regulator MED15, including the length of several polyglutamine (poly-Q) tracts embedded in Med15 sequence, are associated with yeast domestication niches (beer and wine) [2]. Although various yeast species are naturally present on grapes, Saccharomyces cerevisiae dominates the fermentation environment [3]. Historically, and as a modern trend, uninoculated ‘natural fermentations’ have been used in the production of wine, although many wineries will specifically inoculate grape must with previously isolated and commercially available WY strains [4]. Fermentation reactions are characterized by an initial lag phase for yeast not pre-adapted to juice followed by a growth phase after which cells reach saturation (108 cells ml−1). Available nitrogen and ~50% of the sugars are quickly consumed during the growth phase [5]. The remaining sugar is fermented after culture saturation. During this time, the sugar is used almost exclusively to generate energy and survival compounds rather than biomass [5]. Ethanol is the main byproduct of yeast fermenting glucose for energy. Other metabolic products are made throughout the fermentation process including various acids (acetic, succinic and lactic) and volatile compounds (esters, fatty acids, aldehydes and higher alcohols) [6–8]. Industrial fermentation reactions can be reliably replicated in the lab using small-scale fermentation reactions [1, 8]. In addition to environmental stresses such as high-sugar content osmotic stress, nitrogen limitation and low pH present at the start of fermentation, accumulation of both ethanol and acetic acid are stressors specific to late fermentation cultures [3]. Correspondingly, WY exhibit high tolerance to many compounds and stressors including acetic acid, copper, sulfite, high glucose, sorbitol and ethanol [9]. Yeast respond to the initial and accumulating stresses of grape juice and fermentation through the environmental stress response (ESR) [10, 11]. The ESR allows yeast to deal with fermentation stress and moderate levels of ethanol but eventually the levels of alcohol become toxic. Ethanol toxicity is thought to occur by disrupting membrane structural integrity, while the addition of lipids or oxygenation, that is required to synthesize fatty acids, can mitigate the toxic effects [6, 7]. The S. cerevisiae yeast strains used in food and beverage production (e.g. wine, beer and bread) are phenotypically distinct from wild yeast strains and represent lineages from a small set of common ancestors [9, 12]. Prior to the isolation of specific WY strains, which are now removed from evolutionary processes by being maintained as stocks and sold to winemakers to add to grape must [13], WY diverged from wild yeast populations in numerous domestication events [4, 9, 14]. Across different alcoholic beverage industrial niches, S. cerevisiae yeast are critical for their ability to efficiently ferment various carbon sources to produce ethanol and carbon dioxide. Yeast will preferentially undergo alcoholic fermentation and accumulate ethanol even if oxygen is readily available for respiration [3], a phenomenon known as the Crabtree effect [15, 16]. After glycolysis, pyruvate is metabolized into acetaldehyde by pyruvate decarboxylases (PDCs) and acetaldehyde is converted into ethanol by alcohol dehydrogenases (ADHs). This step serves to restore NAD+ by oxidation of NADH that is produced during glycolysis. While ancestral ADH genes were optimized to produce ethanol, the whole genome duplication within the clade that includes S. cerevisiae resulted in ADH genes encoding enzymes with the ability to more efficiently consume ethanol instead of exclusively producing ethanol [3]. Specifically, Adh2 preferentially consumes ethanol with a Km for ethanol (600–800 µmol l−1) that is significantly lower than that for Adh1 (17,000–20,000 µmol l−1) [17, 18]. Thus, Adh2 works much more efficiently than Adh1 at oxidizing ethanol, while the remaining four Adh proteins (mitochondrial Adh3 and Adh4, cytoplasmic Adh1 and Adh5) preferentially produce ethanol [17]. Acetaldehyde produced by Adh2 is metabolized into acetate by aldehyde dehydrogenases (ALDs) and acetate is metabolized into acetyl CoA by acetyl CoA synthetases (ACSs) to be used in the citric acid cycle or other metabolic pathways. The loss of cis-regulatory signals for respiration genes in yeast following the whole-genome duplication may have contributed to the Crabtree effect [19]. Specifically, yeast diverging after the whole-genome duplication, including S. cerevisiae, lost regulatory signals that would promote the expression of mitochondrial genes causing these yeast to display faster anaerobic growth compared to aerobic growth [19]. The ability of S. cerevisiae to preferentially ferment sugars and to tolerate higher levels of ethanol was likely selected for, as these traits allow S. cerevisiae to outcompete other yeast species in natural fermentation environments [3]. Med15 is a component of the tail module of the RNA polymerase II mediator complex [20–23]. As a general transcriptional regulator, Med15 has global effects on stress responses and metabolism [2]. Changes in the sequence of Med15 that influence activity could thus produce broad phenotypic changes, which provide the opportunity for yeast to adapt to novel environments. Variation in MED15 sequence across strains predominantly consists of small changes in the length of simple sequence repeats that underlie glutamine and glutamine-alanine tracts. Recently, the phenotypic stress response consequences of Med15 poly-Q variation have been investigated [1, 24, 25]. Specifically, Med15 was found to be required for resistance to the coal-cleaning chemical, 4-methylcyclohexane methanol (MCHM) and changes in poly-Q tract lengths were found in yeast resistant to MCHM relative to the endogenous MED15 allele in the lab strain [24]. Our previous work revealed that some of the fermentation efficiency characteristic of WY strains is conferred by SNPs and poly-Q tract length polymorphisms in Med15 [1]. We have found that the disorder promoted by glutamine residues is important for Med15 function and that variation in poly-Q tract length in Med15 can modulate stress response and metabolic activity [25]. The significance of poly-Q tract length variation has also been investigated in the yeast Cyc8 transcriptional co-repressor protein [26], as well as in the ANGUSTIFOLIA and Clock proteins in other organisms, where even small changes in tract length (two to four additional glutamine residues) are associated with noticeable phenotypic changes [27–29]. Here we conduct, to our knowledge, one of the first analyses of the MED15 transcriptome during white grape juice (WGJ) fermentation to identify MED15-regulated genes that are unique to the fermentation environment compared to standard lab conditions. In addition, we test the hypothesis that the MED15 alleles from WY harbouring variable length poly-Q tracts and other dispersed SNPs [1, 2] may have contributed to specialized fermentation transcriptomes by comparing the fermentation transcriptome of lab strains into which we have introduced WY MED15 alleles (labelled MED15WY##) to the unmodified lab strain (MED15LAB) background. Briefly, our results show that MED15 regulates glycolytic and fermentation pathways, with a particularly strong and distinctive regulatory effect during fermentation and that some MED15WY alleles cause subtle but biologically significant changes in metabolism when substituted for the MED15 locus in the lab strain. We report the transcriptomic basis for the improvement in fermentation rates seen in specific alcoholic beverage yeasts previously characterized [1], and although we expected changes to genes involved in glycolysis and ethanol metabolism due to the MED15WY alleles, we instead observed differences in the expression of amino acid metabolic genes, among others, compared to unmodified MED15LAB strains. These findings suggest that multiple distinct transcriptomic profiles can enhance fermentation rates. Methods Strains and plasmids Strains used in this study are derivatives of S288C, a widely used non-flocculent laboratory strain, originally designed for biochemical studies [30]. The med15 deletion strain is from the S288C MATα haploid (BY4742) deletion collection [31]. Genotypes of additional strains used in this study are listed in Table 1. WY strains were obtained from the Agricultural Research Service (ARS) culture collection (https://nrrl.ncaur.usda.gov/) (WY23, palm WY, NRRL-Y17772) or were from the Goddard laboratory collection [32] (WY7, Lalvin ICV D-254; WY15, DSM Fermichamp). Plasmids used in this study are listed in Table 2. Table 1.StrainsStrain*Relevant genotypeParental strain and /or plasmidSource or referenceBY4742MATα his3Δ1 leu2Δ0 lys2Δ0 ura3Δ0(OY235)[31](OY320†)med15Δ0::KanMX4BY4742 deletion collection[31]JF2711med15Δ0::URA3BY4742 with a URA3-marked MED15 deletion locusThis studyJF2692 MED15 LAB pJF2141 (episomal) in OY320This studyJF2641 MED15 WY7 pJF2155 (episomal) in OY320This studyJF2642 MED15 WY15 pJF2156 (episomal) in OY320This studyJF2646 MED15 WY23 pJF2158 (episomal) in OY320This studyJF2754 MED15 LAB MED15 ORF from pJF2141 integrated at MED15 locus in JF2711This studyJF2756 MED15 WY7 MED15 ORF from pJF2155 integrated at MED15 locus in JF2711This studyJF2758 MED15 WY15 MED15 ORF from pJF2156 integrated at MED15 locus in JF2711This studyJF2760 MED15 WY23 MED15 ORF from pJF2158 integrated at MED15 locus in JF2711This studyWY7Diploid WY–Lalvin ICV D254WY15Diploid WY–DSM FermichampWY23Diploid WY–NRRL Y-17772* All strains except for those designated WY carry the ura3Δ0 mutation. JF2711 carries the same mutation but is Ura+due to the URA3+ marker at the MED15 locus.† Parenthetical names (OY###) are lab-specific storage designations. Strains * All strains except for those designated WY carry the ura3Δ0 mutation. JF2711 carries the same mutation but is Ura+due to the URA3+ marker at the MED15 locus. † Parenthetical names (OY###) are lab-specific storage designations. Table 2.PlasmidsPlasmidDetailOther commentSource or referencepRS315LEU2, CEN(SV31)[96]YEp24URA3, 2μ(SV6)–pRS315 M-WTMini-MED15; the internal region of Med15 (aa 116–277) plus the Mediator association domain (aa 799–1,081) located between the promoter and terminator regions of S288C MED15 in pRS315 (LEU2, CEN)(SV286)[33]pJF2141BY4742 MED15 in pRS315 (LEU2, CEN)SV286 gap repairThis studypJF2155MED15WY7 in pRS315 (LEU2, CEN)SV286 gap repairThis studypJF2156MED15WY15 in pRS315 (LEU2, CEN)SV286 gap repairThis studypJF2158MED15WY23 in pRS315 (LEU2, CEN)SV286 gap repairThis studyParenthetical names (SV###) are lab storage designations. Plasmids Parenthetical names (SV###) are lab storage designations. Wine yeast MED15 integration A URA3-marked med15Δ lab strain (JF2711) was created to use as a starting point for integrating WY MED15 alleles. A URA3 expression cassette was amplified from YEp24 using URA3 GR (gap repair) F1 and R1 primers (Table 3). Primers were complementary to sequences outside of the MED15 ORF, allowing for homology to be used in the gap repair of BamHI- and SpeI-digested Mini-MED15 plasmid (pRS315 M-WT) [33]. The URA3 expression cassette with MED15 flanking homology was subsequently amplified from the newly constructed plasmid using MED15 F-245 and MED15 R+3498 primers (Table 3) and integrated into the genome of BY4742 at the MED15 locus to generate JF2711. WY MED15 alleles were integrated into the MED15 locus of JF2711 by homologous recombination. MED15 coding sequences with 250 bp of upstream and downstream sequence were amplified from expression plasmids (pJF2141, LAB; pJF2155, WY7; pJF2156, WY15; pJF2158, WY23) using MED15 F-245 and MED15 R+3498 primers (Table 3). MED15 PCR fragments were then transformed into JF2711 and cells plated on yeast peptone dextrose (YPD) medium. After a day of growth, plates containing 800 mg l−1 5-fluoroorotic acid (5-FOA) were used to counter-select transformants that had become Ura- due to integration of the new allele and loss of the URA3-marked MED15 locus. Integrants were stored as JF2754-JF2760. Table 3.PrimersPrimerSequence (5′–3′)UseMED15 F-245GGATGAGGATGATGAAGGTGCAmplify MED15 ORFMED15 R+3498GTACTGATGATAGTCAAGTCCATTGAmplify MED15 ORFMED15 F+384GTTCTTGAATCAGCAGGCTCScreen WY allelesMED15 R+529GTGCCACTTTCATCTGGTTCScreen and sequencingMED15 F+1924GGCACTACTTCTACTGGAAACScreen WY allelesMED15 R+2143CATTAGCCATTGCCGAATAATTAGCCACACCAGGScreen WY allelesMED15 F+2693AGATGCTTACGTCATGCACTATCCSequencingMED15 R+2990GGGTTGCCGACATCCATATTAGSequencingURA3 GR F1TCGTATAGTGCCGTACTCAAAGATCAAGGATTAAAACGCGACAGCTTATCATCGATAAGCAmplify URA3 geneURA3 GR R1GTATCAAAAGTATGGAAACTTCAAATGTTCAAGTAGCACAATCATTACGACCGAGATTCCAmplify URA3 gene Primers Colony PCR Yeast colonies were used for rapid PCR screening using Taq polymerase, colony PCR buffer (final concentrations: 12.5 mM Tris/Cl (pH 8.5), 56 mM KCl, 1.5 mM MgCl2, 0.2 mM deoxynucleotide triphosphates (dNTPs)) and primers at a final concentration of 0.2 µM. Reaction mixes were aliquoted into PCR tubes into which a small amount of yeast cells were transferred using the end of a 200 µl micropipette tip. A standard hot-start thermocycler programme with extension at 68 °C was adjusted for the Tm of the primer set and size of the amplification product. Yeast transformation Transformations of med15Δ strains were conducted using the Frozen-EZ Yeast Transformation II kit (Zymo Research) with modifications. A total of 10 ml of early log phase cells were pelleted and washed with 2 ml EZ 1 solution, repelleted and resuspended in 1 ml EZ 2 solution. Aliquots were frozen at −80 °C. A total of 0.2–1 µg DNA and 100 µl of EZ 3 solution were added to 10 µl competent cells for each transformation. Transformations were plated on selective media following incubation at 30 °C for 45 min. Transformations into all other strains were conducted using a standard lithium acetate transformation protocol [34, 35]. Small-scale fermentation reactions WGJ concentrate (Global Vinters Inc., Canada) was supplied at ~68°Bx (1°Bx is equal to 0.01 g ml–1) and was diluted to 20–21°Bx with distilled deionized water as measured using a triple scale hydrometer and confirmed with a refractometer (Milwaukee). Grape juice media was supplemented with 200 mg l−1 histidine, 300 mg l−1 leucine, 300 mg l−1 lysine and 200 mg l−1 uracil for use in growth and fermentation experiments. Ten millilitre moulded sterile clear glass vials (American Clinical Supplies Inc.) were used in fermentation experiments. A total of 5 ml of prepared grape must (or other specified media) was added to each vial which was then closed tightly with a rubber cap. A 25 G × 7/8 inch (0.5×22 mm) hypodermic needle was inserted into the cap to allow CO2 release from the system and to restrict evaporation [8]. Prior to inoculating the fermentation vials, cells were grown in SC-Leu media for 2.5 days to reach saturation and then diluted 1 : 20 into fresh SC-Leu media. 1.25×107 cells of saturated culture was added to 5 ml prepared white grape must (or other media as indicated in the figure legends) to a final cell concentration of 2.5×106 cells ml−1. SC-Leu media was added as needed to standardize the media added to each vial. Uninoculated vials with the same volume of grape must were used to control for contamination. Fermentation vials were incubated at 25±1 °C or as indicated. The temperature was maintained at 25 °C despite fluctuating room temperatures (RTs) using a heated shaker in a 4 °C cold room. Shaker speed was maintained at 165 r.p.m. In later fermentation experiments, synthetic complete media (SC) was used rather than WGJ. SC-arginine was 20% glucose and contained all amino acids except arginine. Fermentation assays Progression through the fermentation process was determined by once-daily weight loss measurements caused by CO2 release. Cumulative CO2 weight loss was calculated by subtracting daily measurements from the initial weight (day 0). To examine the impact of each MED15 plasmid, fermentation (weight loss) was evaluated simultaneously for multiple different transformants (biological replicates). Time to weight loss benchmarks were determined per transformant. The total amount of CO2 weight loss accumulated by day 7 for MED15 strains or day 8 for med15Δ strains was set as 100% weight loss. Values equal to 10, 35, 50 and 80% of the 100% value were plotted as horizontal lines on top of fermentation curves. The time, in hours, that corresponded to the intersection between the horizontal lines and fermentation curves was determined manually or by linear interpolation. Analysis of microarray datasets Microarray datasets used in the comparisons shown in Figs 3a and 5 were restricted to those that evaluated med15Δ and wild-type S288C strains after growth to mid log phase in media containing 2% glucose [36, 37]. The med15Δ strain in Hu et al. [36], reanalysed in Reimand et al. [38], was from the Open BioSystems collection in the BY4741 strain background. Cells were grown in YPD (2% glucose) and harvested at mid-log; the med15Δ strain was harvested at an OD600 of 0.88. Unaltered med15Δ data (log2 fold-change andP-values) were retrieved from provided processed data tables (E-MTAB-109 from ArrayExpress). The med15Δ strain in Kemmeren et al. [37] was constructed by the gene deletion consortium in the MATα haploid BY4742 strain [39] and was harvested in log phase growth in SC media (2% glucose) in which the authors report the wild-type strain has the same doubling time as in YPD. Cultures were harvested at an OD600 of 0.6. Unaltered med15Δ data [log2 fold-change (m-values) and P-values] was downloaded from provided supplementary files (GSE42527 from Gene Expression Omnibus). Genes with provided expression values surpassing a threshold log2 fold-change of >1 or <−1 and a P-value<0.05 were assembled into differentially regulated gene (DRG) sets that were used in the comparisons shown in Fig. 3a. All subtelomeric genes from the COS, FLO, MAL and PAU gene families, regardless of log2 fold-change and P-value cutoffs, were compiled as shown in Fig. 5. RNA extraction RNA was extracted from 10 ml yeast cultures harvested in log phase (2×107 cells ml−1). Cells were pelleted by centrifugation and resuspended in 1 ml cold water. Yeast were repelleted by centrifuging for 10 s at highest setting of tabletop centrifuge. The supernatant was aspirated, and the pellet was frozen in dry ice for 5 min before being stored at −80 °C prior to RNA extraction. RNA was extracted using a hot acid phenol protocol [40]. Frozen cell pellets were resuspended in 400 µl TES solution (10 mM Tris, 10 mM EDTA, 0.5% SDS). An equal volume of acid phenol was added to each tube and vortexed for 10 s. Tubes were incubated in a 65 °C water bath for 1 h with vortexing every 10–12 min. Phases were separated by microcentrifugation in the cold for 5 min. The aqueous layer was extracted into a new tube avoiding the DNA-enriched interface. A second round of acid phenol extraction was conducted followed by a chloroform extraction. RNA was precipitated with 300 µl of 4 M LiCl in dry ice for 20 min followed by 5 min of high-speed centrifugation at 4 °C. The pellet was washed with 500 µl ice-cold 70% ethanol and dried for 10 min at 37 °C. RNA pellets were resuspended in DEPC-treated water and stored at −20 °C. Contaminating genomic DNA was removed using the DNase Max Kit (Qiagen). 10 µl of a 1x Master Mix (5 µl 10X buffer, 5 µl water and 0.5 µl DNase I) was added to 40 µl RNA. Following a 30 min incubation at 37 °C, DNase was removed by addition of 5 µl of the DNase removal resin and a 10 min RT incubation with periodic agitation. The resin was pelleted by centrifugation for 1 min at the highest setting of a microcentrifuge. The supernatant containing DNA-free RNA was transferred to a new tube, and the concentration was determined using a NanoDrop spectrophotometer (Thermo Scientific). RNA quality was visualized on formaldehyde gels (1% agarose, 10% formaldehyde solution, 1x MOPS buffer). cDNA preparation and quantitative reverse transcription polmerase chain reaction (qRT-PCR) cDNA was prepared using Superscript III First-Strand Synthesis System (Invitrogen). Transcripts were amplified with random hexamer (Invitrogen) or anchored oligo-dT20 (Integrated DNA Technologies) primers. For anchored oligo-dT20 primers, 1 µg RNA, 50 ng primer and 0.01 µmol dNTPs were mixed in a final volume of 10 µl. Tubes were incubated for 5 min at 65 °C and then 2 min on ice. A total of 10 µl cDNA Master Mix (2 x RT buffer, 10 mM MgCl2, 0.02 M DTT, 40 U RNase OUT, 200 U Superscript III Reverse Transcriptase) or mock Master Mix lacking reverse transcriptase was added to each tube and incubated for 60 min at 50 °C and 5 min at 85 °C. To degrade RNA, 2 U RNase H was added to each tube and incubated for 20 min at 37 °C. For random hexamer primers, tubes were incubated 10 min at 25 °C and then 50 min at 50 °C. Transcript abundance was quantified for each sample relative to a normalization transcript using the PerfeCTa SYBR Green FastMix (Quantabio) in 96-well PCR plates (Hard-Shell 480 PCR Plates, Bio-Rad) measured in a LightCycler 480 (Roche). Individual 10 µl reactions consisted of 5 µl FastMix, 1 µl target-specific primer pairs (0.5 µl 10 µM forward primer and 0.5 µl 10 µM reverse primer) (Table 4), 3 µl DEPC-treated water and 1 µl cDNA, mock cDNA or water. Plates were sealed with optically clear film (PlateSeal). A standard SYBR green PCR programme was used with modifications: 95 °C for 5 min, 45 cycles of 95 °C for 10 s, 55 °C for 10 s and 72 °C for 20 s with a single fluorescence acquisition. A melting curve was conducted directly following the PCR programme to confirm the presence of individual species amplified in each well. The melting curve programme was 95 °C for 5 s, 65 °C for 1 min and ramp up to 97 °C by 0.11 °C s−1 with continuous fluorescence acquisition. ALG9 was used as a normalization transcript. All samples were measured with three technical replicates. Mock cDNA (prepared without the addition of reverse transcriptase) and water were used as negative controls. Table 4.qRT-PCR primersGene_pair#*SizeEfficiencyForward primer sequence 5′–3′Reverse primer sequence 5′–3′ACS1_1188 nt† TAGTTGGTCCACCGCAACAGTGGTCACTTTCTCCTGCTAGACS2_1194 nt CAACGCTACTGAAGGTGATGTGTTCGGCTTCGTTAGAAGCADH1_11392.03CAAGGCTATGGGTTACAGAGGTCAGTGGCCTTTAGAACAGADH2_11292.02CGAACGGTACTGTTGTCTTGGCTTCTCTGGTATCAGCTCTGADH3_1150 nt CCCTCATGGTGTCATTAACGGATGGACTTCACCACATGAGADH4_1130 nt TGGCAGGTATGGCTTTCAACGTTGGCCTCTTGAACATGAGADH5_1120 nt TAATGGCGGTTCTCATGGAGGTAAGCATGAGCTGGCATACALD2/3_1205 nt TTGGTGGAGCTAAAGGCTACCATGTGCGCTTTCTTGACATCALD4_11831.98CCCTGTTGTCACTGTAACCAACTGCGTGGTGGAAATCGALD5_1123 nt CGATATTAACAAGGCTGTTGATGACGGCCAATACCTGACTGGALD6_11402.01GCTAACTTCCAAGGTGCTATCGGTTGGTCTGATGAAGTAACCALG9_11231.98CTGAATTTCCCACTGCCTGTGCCTCTTTGTGAGGTTGTTGAGALG9_21422.02GGATAGTGGCTTTGGTGAACCAGGAAAGAACTTGGGAAGTGARO10_11682.05CACTCAAACGCTCACATCAATGTAGTGTAGCCGTTATTGTTCCATF2_12451.73ATTCGACATTCCCGAAGGTGTAGTACGGCTTGGTCGTATCBAT2_11161.87AGACTCGAACCAAGTGAATGGAATCGCAGCAGTACCAGAACFPS1_1172 nt CATCGGACCTGAGTGACTTCACCACCAAACGTTCTTTTGCGPD1_11521.94CTTGCCTGGCATCACTCTACCTGACGTGTGAATCAACATGGUT1_1116 nt GCAAATTCAAGCCGATATCTTAGTCGTTCACATCCTTGAAAGCGUT1_21192.21ATATCACCAGAGGCTCATCCCTTCAGCCAACCTTTGGATCIRC7_1145 nt GGCTCGGAAATCGAGATGAGCAGACCAGCTCTTGTGAACCLEU4_1138 nt CAATGACGAACAGTACCAAGCGATTGGACCATTACCTGTACCMAL12/32_1147 nt CCTACATCGAGAATCACGATCCCTATCTCCTGACCTTGATAGMED15_12972.02AGATGCTTACGTCATGCACTATCCGGGTTGCCGACATCCATATTAGMET10_1109 nt ACTCAATTGCCTCCTCTCAGCTTAGAAGCTTGACCGTACCPDA1_1215 nt ATCCAAGAACGATCCAATTGCAATTTGGCTTCTGGAGGAGGPDB1_1148 nt TATCCCCTGAGTTCACCTTGGATAACTTCTGCAGAGACACCPDC1_12272.01TCTCCACCATGATCAGATGGTGTCGTTGAAAGACTTGTCTTGPDC6_12172.03TCTCCACCATGATCAGATGGCTCTGAATCAGTGGTTAAGGCPTR2_11351.94CTTTTTCAGAGATCTTTGCCTCTATCGACAATGCACAACCGATTGCPTR2_2122 nt GTTTGTTCTGGTTGTGCTTCAGCAATATCGTTAGCTTTAGGTGCG* Pair number for multiple primer pair sets designed for a given gene.† nt, not tested. qRT-PCR primers * Pair number for multiple primer pair sets designed for a given gene. † nt, not tested. The relative abundance of target transcripts was determined by calculating the crossing point (CP) value for each reaction, which reflects the number of cycles of amplification required for the SYBR green signal to exceed a threshold value set to eliminate background noise. CP values were calculated using the Second Derivative Maximum method implemented in the LightCycler 480 Software (Roche). The average CP for technical replicates amplifying the target transcript with a specific RNA sample was normalized to the average CP for technical replicates amplifying the normalization transcript. This ratio was then compared across samples as a depiction of relative abundance of the target transcript in each sample. RNA sequencing Individual fermentation reactions in 5 ml supplemented WGJ were prepared as previously described for two biological replicates each of the URA3-marked med15Δ strain and integrants of the LAB, WY7, WY15 and WY23 MED15 alleles. Weight loss was monitored, and yeast were pelleted at the 35% weight loss point (~37 h for MED15 strains and 48 h for med15Δ strains). RNA was extracted as described above. RNA sample quality was assessed using an Experion automated electrophoresis system with a HighSens chip using a standard protocol (Bio-Rad). cDNA libraries were prepared using the Illumina Stranded mRNA Prep protocol (Illumina). A total of 150–200 ng of total RNA per sample and 13 cycles of amplification were used. The Illumina Dual Index Adapters set A (Integrated DNA Technologies) was ligated to the cDNA. cDNA library quality control and sequencing were performed by the Iowa Institute for Human Genetics (IIHG) Genomics Division. cDNA libraries were quantified using fluorometry with PicoGreen. The integrity of the cDNA libraries was determined using an Agilent BioAnalyzer DNA chip. A KAPA qPCR was used to get precise quantification prior to loading samples for sequencing. The pooled libraries were run with the 2×150 bp protocol on one lane of an S Prime flow cell in an Illumina NovaSeq. RNA-Seq data analysis Sequencing data was processed using a manually curated pipeline of programmes implemented in the IIHG instance of the Galaxy platform, a web-based computational workbench [41]. Adapter sequences were trimmed from reads using TrimGalore! (Galaxy tool version 0.6.3). The adapter sequence was auto-detected. Reads shorter than 50 bp after trimming were discarded. Any reads orphaned at this point were also discarded to maintain sets of fully paired reads. Paired-end reads were then mapped to the S288C reference genome (sacCer3) using Bowtie2 (Galaxy tool version 2.3.4.2) [42]. Default parameters for paired-end reads were used, the maximum paired-end length was set to 5,000 and dovetailing allowed to maximize the number of mapped reads. Counts were quantified using featureCounts (Galaxy tool version 1.6.3) [43] to tally the number of reads mapping to annotated ‘genes’ in the S288C reference general feature format file (version R64-3-1). Count normalization and differential expression analysis was done using DESeq2 (Galaxy tool version 2.11.40.2) [44] using default parameters with multiple testing correction using Benjamini–Hochberg method with a false discovery rate of 5%. Biological replicates were compared at the level of normalized counts with correlation analysis (Fig. S1, available in the online Supplementary Material). Alternative mapping and differential expression analyses included the use of limma (version 3.60.6), kallisto (version 0.46.1) and sleuth (version 0.30.2) [45–47]. Limma was used as a potentially less conservative alternative to DESeq. The featureCounts table from the Bowtie2 method was transformed using voom [48] prior to linear model fitting with limma using default methods. kallisto was used as an alternative method to increase the percentage of uniquely alignable reads. A kallisto index was prepared with default settings using the SacCer3 reference transcriptome. Trimmed reads were pseudoaligned using kallisto quant with default settings and 100 bootstraps. Differential expression analysis was done on uniquely aligned and pseudoaligned reads. kallisto abundance.h5 outputs were prepared for DESeq2 using tximport [49]. The transcript-to-gene associations were determined using biomaRt [50] ‘refseq_dna’ and ‘ensembl_gene_id’ attributes. Alternatively, kallisto abundance.h5 output files were used directly as inputs for sleuth differential expression analysis. A sleuth object was created using sleuth_prep with the extra_bootstrap_summary argument set to TRUE to take advantage of the bootstrapping done while running kallisto. The effect of the strain genotype on expression patterns was modelled using sleuth_lrt. Differential expression for each strain genotype relative to the lab strain was determined using sleuth_wt. Fig. S2 (A–D) compares the DRGs identified with the four methods we tested. Venn diagrams showing overlap of genes with adjusted P-values<0.05 for each of four methods [Bowtie2 ->DESeq2 (BowDE), Bowtie2 ->limma (BowLi), kallisto ->DESeq2 (KalDE) and kallisto ->sleuth (KalSl)] for each of the four comparisons described as part of this work (MED15LAB vs med15Δ, MED15WY7 vs MED15LAB, MED15WY15 vs MED15LAB and MED15WY23 vs MED15LAB) emphasize that the Bowtie2-DESeq2 pipeline performs as well as any of the other pipelines in terms of identifying DRGs. Pearson correlation matrices of the differential expression statistics (DESeq2 stat values, limma t-values and sleuth b-values) for the shared set of genes for each of four methods for the four comparisons are shown in Fig. S2 (E–H). Data representation Volcano plots of differentially expressed genes were prepared using VolcaNoseR [51]. Venn diagrams of overlapping differentially expressed genes were prepared using VennDiagram [52]. Gene ontology (GO)-enrichment analyses were done using AllianceMine [53]. Metabolic pathway enrichment analyses were done using BioCyc [54]. GO enrichment Differentially expressed genes were input as new lists on AllianceMine [53]. Biological process GO enrichment was done automatically. Enriched GO terms with Bonferroni adjusted P-values<0.05 were obtained. For each term, a ‘Med15 Regulated’ frequency was calculated by dividing the number of DRGs with that GO term over the total number of DRGs in the list. An ‘All Genes’ frequency was determined by dividing the number of yeast genes with that GO term by the total number of yeast genes in the ALL_Yeast_Genes list (n=7,173). Redundant or highly related enriched GO terms were manually removed. Gene set enrichment analysis Pathways enriched in differentially expressed genes, in comparisons of MED15WY alleles with MED15LAB, were identified using fgsea [55]. Wald statistics for differential expression from the DESeq2 analyses were used to determine enrichment among 1,234 yeast pathways compiled in the ConsensusPathDB [56]. These included pathways from BioCyc, KEGG and Reactome among others. For each comparison, pathways were excluded from the analysis if there were five or fewer genes with Wald statistics within that pathway. Pathways with adjusted P-values less than 0.01 were considered significantly enriched. Venn diagrams of overlapping enriched pathways were created using VennDiagram [52]. Heatmaps for gene sets that correspond to specific pathways were created using pheatmap [57] for log-transformed normalized counts. Primers All primers were synthesized by Integrated DNA Technologies and are listed in Tables 3 and 4. Primer pair efficiency was measured by qRT-PCR analysis of serial dilution of control RNA. Results Fermentation gene expression is affected in the med15Δ strain We previously observed that a med15Δ lab strain is defective in the fermentation of grape juice [1]. We hypothesized that this fermentation defect is the consequence of changes in the expression of metabolic genes involved in the production or consumption of ethanol. To determine the role of Med15 in the regulation of relevant fermentation genes, we first assembled a list of target genes by reviewing existing med15Δ microarray datasets [36, 37, 58] and identifying genes with previously described roles in ethanol metabolism and wine or beer fermentation [59–61]. We then used qRT-PCR to evaluate their expression in standard growth media. Genes we found to be regulated by MED15 had roles in ethanol metabolism and fermentation, including three (PDC1, ACS1 and ADH2) that had not previously been recognized as MED15 targets. Both positive regulation by Med15 (lower expression in the deletion strain) and negative regulation by Med15 (higher expression in the deletion strain) was observed. For example, the ADHs ADH1 and ADH3, which facilitate the conversion of acetaldehyde to ethanol, were reduced in the med15Δ strain, whereas ADH2, which catalyses the reverse reaction (ethanol to acetaldehyde), was significantly upregulated (Fig. 1). Similarly, ALD4, which is involved in the conversion of acetaldehyde to acetate, is increased in the med15Δ strain (Fig. 1) indicating that Med15 typically represses the mitochondrial, glucose-repressed ALD4. PDC1, the major isoform responsible for converting pyruvate to acetaldehyde during fermentation, was downregulated in the med15Δ strain, whereas PDC6, a minor isoform typically expressed under stress conditions, was upregulated indicating that Med15 promotes PDC1 and represses PDC6 expression to reinforce a fermentative metabolic profile. Together, the observed transcriptional effects are consistent with Med15 playing a central role in fermentation by regulating key enzymes that drive ethanol production and stress-related metabolic shifts. Influence of Med15 on expression of alcoholic fermentation pathway genes. (a) Summary of predicted impact of med15Δ on ethanol metabolism. Tested metabolic genes labelled according to changes in expression in med15Δ relative to expression in MED15 yeast. Gene names in red are upregulated (activated) by MED15, gene names in blue are downregulated (repressed) by MED15, and gene names in black are not significantly regulated by MED15. Underlined genes are mitochondrial. (b) Gene expression log2 fold-change measured by qRT-PCR normalized to ALG9 in MED15 relative to med15Δ grown in YPD with three to six technical replicates. One outlier was removed with a Grubb’s test. Relative expression values for genes with shades of red representing extent of activation by MED15, and shades of blue representing extent of repression by MED15. Significance was determined using t-tests with a Bonferroni correction for multiple testing (* <0.05; ** <0.01; *** <0.001). Genes are grouped by function: alcoholic fermentation (green), respiration (yellow), glycerol metabolism (purple), amino acid metabolism (orange) or other (grey). The MED15 transcriptome is highly specialized during alcoholic fermentation To determine the global impact of med15Δ on gene expression during alcoholic fermentation rather than in standard growth media (YPD or SC at 2% glucose) as in the datasets discussed above, an RNA sequencing experiment was conducted for cells grown in WGJ (20% glucose) at the 35% wt loss fermentation benchmark (Fig. 2a) [1, 8]. Note that weight loss benchmarks rather than time were used to coordinate sample collection in this experiment because the med15Δ strain grows and ferments much slower than strains with an intact MED15 gene. This global analysis was expected to identify all affected genes including any that we had not anticipated, and the fermentation conditions were expected to have a substantial impact on gene expression. WGJ RNA sequencing experimental setup and summary results. (a) RNA sequencing experimental overview with a description of two alternative analysis pipelines. (b) Summary mapping results from Bowtie2 [42] and kallisto [45] for each sample. Counts were then used as input for DESeq2 [44] to determine differential expression. (c) Venn diagram of DESeq2 determined DRGs (adjusted P-value<0.05) for Bowtie2 and/or kallisto methods. Scatter plot showing the relationship between DESeq2 log2 fold-change values for Bowtie2 and kallisto methods. 90% of genes (5117/5678) have log2 fold-change values that differ by less than 0.255 between the two methods. Genes with extreme variation (greater than 2.5 difference in log2 fold-change) are labelled. Due to the relatively high number of multi-mapped reads in our dataset using Bowtie2 [42], we also employed kallisto [45], an RNA-Seq quantification programme that incorporates pseudoalignment of multi-mapped reads (Fig. 2b) and thus increased the number of uniquely mapped reads to ~90%. The increase in mapped reads did not substantively change the DRG set. Seventy-nine percent of genes identified as differentially regulated were common regardless of the read-mapping strategy, and the extent of gene regulation, as expressed by the log2 fold-change, was also similar (Fig. 2c). Genes whose counts increased by an average greater than 10,000 reads using the kallisto method included members of large gene families like the small and large ribosomal proteins as well as translation factors (~50%), glycolysis genes (~15%), transport/nutrient uptake (~10%), subtelomeric genes (~10%) and chromatin/stress genes (~5–10%) (Table S1). These were not among the very few genes whose extent of regulation was drastically affected (reside outside the diagonals) in the kallisto dataset (Fig. 2c), but instead these genes had proportional increases in mapped reads across samples using kallisto. In this analysis, we found 807 genes that were differentially regulated in a comparison of WT and the med15Δ strain under the WGJ condition (log2 fold-change >1 or <−1 and adjusted P-value<0.05, Bowtie2-DESeq2 and/or kallisto-DESeq2 DRGs). 0.7% are in common with DRGs identified in the YPD microarray dataset [36, 38], and 11.5% of the WGJ DRGs are in common with DRGs identified in the SC microarray dataset [37] (Fig. 3a). The 93 DRGs in common between the WGJ and SC lists were enriched (adjusted P-value<0.05) for GO terms related to core biosynthetic metabolism, including genes involved in sulphur and nitrogen assimilation, glycolytic flux and redox balancing under fermentative conditions (Fig. 3b). Condition-specific and general metabolic pathways regulated by MED15. (a) Overlap between significant (log2 fold-change>1 or <−1 and adjusted P-value<0.05) differentially regulated genes (DRGs) in comparisons of the MED15 and med15Δ strain for log-phase growth in YPD [36] as reanalysed by Reimand et al. [38], log-phase growth in SC [37], and WGJ fermentation (this study) using the combined Bowtie2-DESeq2 and kallisto-DESeq2 pipeline gene lists. (b) Gene ontology (GO) enrichment using AllianceMine [53] of the 93 gene overlap between the WGJ and SC groups from (a). All enriched GO terms shown have a Bonferroni adjusted P-value<0.05. The DRGs shared by the closely related carboxylic acid metabolic process (most specific) GO terms are displayed. (c) GO enrichment (as for b) of the 485 genes with decreased expression in the MED15LAB strain relative to deletion for the Bowtie2-DESeq2 and/or kallisto-DESeq2 methods. All enriched GO terms shown have a Bonferroni adjusted P-value<0.05. Closely related GO-terms were manually removed. (d) GO enrichment (as for b) of the 323 genes with increased expression in the MED15LAB strain relative to deletion for the Bowtie2-DESeq2 and/or kallisto-DESeq2 methods. All enriched GO terms shown have a Bonferroni adjusted P-value<0.05. Closely related GO-terms were manually removed. Among the 485 genes negatively regulated by Med15 (significantly lower expression in MED15 vs med15Δ) within the WGJ environment the enriched GO terms included: sporulation, cell cycle, DNA replication, cell division, checkpoints, exit from mitosis and cell wall organization (Fig. 3c). A total of 323 genes that were positively regulated by Med15 in WGJ were enriched in amino acid and nucleotide (nt) biosynthesis and transport and response to stress (Fig. 3d) as well as carbohydrate transport (pentose, hexose, glucose, fructose and mannose) including several glucose transporters [high-affinity (HXT4, HXT6 and HXT7) and low-affinity (HXT1 and HXT3)]. Interestingly, two additional HXT genes, HXT10 and HXT14, were present in the negatively regulated gene list. To achieve more nuanced insight into the MED15-dependent metabolic profile during WGJ fermentation, we analysed the distribution of MED15-regulated genes across all yeast metabolic pathways using the BioCyc database [54] with the entire WT vs med15Δ WGJ fermentation dataset as input (Bowtie2-DESeq2 – including positively regulated, negatively regulated and genes not regulated by MED15). In addition to glycolytic and ethanol metabolic pathways (Fig. 4a, b), the arginine biosynthetic pathway was specifically affected during WGJ fermentation (Fig. 4c). A similar enrichment was seen for genes involved in the methionine biosynthesis (Fig. 4d). Allantoin catabolism was another significant pathway regulated by MED15 during fermentation, with all allantoin catabolism enzymes (DAL1, DAL2, DAL3 and DUR1,2) affected (Fig. 4e). Metabolic pathways regulated by Med15 in WGJ. (a) Glycolysis; (b) pyruvate fermentation; (c) arginine biosynthesis; (d) methionine metabolism and (e) allantoin metabolism. Pathways are generated using BioCyc [54]. The colour of the edges (arrows) are approximations of the sum of the effects on the individual steps (enzymes) in the pathway using Bowtie2-DESeq2 differential expression data. See key for the relative Med15 dependence of each enzyme in the pathway. Because published reports show that MED15 affects silencing at telomeres during normal growth [62, 63], subtelomerically localized genes were evaluated in the WGJ fermentation dataset. We evaluated the distal region of yeast telomeres, which contains genes belonging to a small number of gene families, including MAL, MEL, PAU, COS and FLO genes (Fig. 5a), although the S288C lab strain used here lacks MEL genes [64] and only has two partially functional MAL gene regulons [65, 66]. Most PAU genes (12/21) and COS genes (8/12) and all FLO genes (5/5) were significantly differentially regulated by MED15 under WGJ fermentation conditions (adjusted P-value<0.05, Bowtie2-DESeq2 DRGs). Virtually all differentially regulated FLO, PAU and COS genes (except for PAU9 and COS111) were more highly expressed in the med15Δ strain under WGJ fermentation conditions, a pattern also seen in the SC transcriptome. This pattern suggests that during WGJ fermentation (and growth in SC media) MED15 promotes or helps maintain silencing of subtelomeric gene families. This is in contrast to the expression pattern of subtelomeric genes during log-phase growth in YPD in which the average log2 fold-change across gene families is modestly elevated in the WT strain [36, 37, 58]. MED15 regulates silencing of subtelomeric genes during WGJ fermentation. (a) Cartoon depiction of S. cerevisiae telomeres and subtelomeric features. (b) Differential expression of subtelomeric genes of each indicated class (COS, FLO, MAL, PAU) during WGJ fermentation, or during growth in SC or YPD media. Log2 fold-change values for MED15LAB/med15Δ are plotted for WGJ (Bowtie2-DESeq2), SC and YPD [36, 37, 58]. Data were as previously described without further processing or filtering by P-value. All representative genes for the four gene families were selected from the three data tables. ITS, interspersed telomeric repeats; STR, subtelomeric repeat sequences; TAS, telomere-associated sequence; TPE, telomere position effect; TR, telomeric repeats. MED15WY - MED15LAB comparisons reveal subtle effects on the fermentation transcriptome To investigate the impact of WY MED15 alleles during WGJ fermentation, we examined the impact of three different WY MED15 alleles integrated into the lab strain genome as the sole MED15 gene. RNA-Seq analysis was conducted at 35% weight loss (Fig. 2). Overall, the fermentation genes (labelled in Fig. 6) were not significantly affected (Bowtie2-DESeq2 data). Only the ALD genes displayed absolute log2 fold-change values greater than 0.5 (Fig. 6b-d) and among these, only ALD5 was significant and this was true only in MED15WY23 (Fig. 6d). Thus, the expression of most fermentation genes was unchanged in WGJ environment regardless of MED15 allele suggesting that the enhanced fermentation phenotypes in strains with MED15WY alleles [1] is rooted in expression differences in other classes of genes. Fermentation metabolic gene expression in MED15WY23 vs MED15LAB. (a–d) Volcano plot of DRGs (Bowtie2-DESeq2) between strains expressing MED15LAB and deletion (Δ) (a), MED15WY7 (b), MED15WY15 (c) and MED15WY23 (d) are shown. Highlighted in each plot are genes that exceed both log2 fold-change and P-value thresholds represented by vertical and horizontal lines. Genes that are indicated in blue are negatively affected by the WY allele, and genes in red are positively affected by the WY allele. Each coloured gene in the MED15WY7, MED15WY15 and MED15WY23 plots is annotated. The most significantly regulated genes in the MED15LAB/med15Δ plot are annotated. Additional annotations are for fermentation genes shown to be regulated (in YDP) as in Fig. 1. To investigate the transcriptomic basis for improved WGJ fermentation in strains differing only in their MED15 allele [1], we applied GO and Pathway enrichment analyses for the DRGs from comparisons of the MED15WY and MED15LAB transcriptomes. A comparison of MED15WY23 and MED15LAB strains in the WGJ environment revealed 42 DRGs (log2 fold-change>0.5 (n=32) or <−0.5 (n=10) and adjusted P-value<0.05, Bowtie2-DESeq2 and/or kallisto-DESeq2 DRGs) (Fig. 7 and Table S2). GO enrichment analysis of genes with increased expression with MED15WY23 (n=32) pointed to an enrichment in amino acid metabolism with a specific effect on the arginine biosynthetic pathway. We also checked the genes of the urea cycle, which intersects with the arginine pathway at ornithine, a metabolic pivot for bidirectional nitrogen management. The effect of MED15WY23 on both arginine and urea cycle genes suggests an effect on nitrogen metabolism (Fig. 7). MAE1 (malate dehydrogenase) expression was also significantly upregulated in the MED15WY23 strain, alongside that of other enzymes in the gluconeogenesis pathway (Fig. 6d and Table S2). MED15WY23 differentially modulates the expression of metabolic genes during WGJ fermentation. (a) GO enrichment using AllianceMine [53] as in Fig. 3 of the 32 genes with increased expression in the MED15WY23 relative to the MED15LAB strain in the WGJ fermentation environment for the Bowtie2-DESeq2 and/or the kallisto-DESeq2 methods. All enriched GO terms shown have a Bonferroni adjusted P-value<0.05. Closely related GO terms were manually removed. (b, c) Metabolic pathways for arginine biosynthesis (b) and urea cycle (c). Metabolic pathway analysis conducted with BioCyc [54]. The colour of the edges are approximations of the sum of the effects on the individual enzymes in the pathway using Bowtie2-DESeq2 differential expression data. Transcriptomes of strains with integrated MED15WY7 or MED15WY15 alleles were very similar to the MED15LAB transcriptome at the log2 fold-change and significance thresholds shown in Fig. 6. To achieve additional resolution, we combined GO term enrichment and pathway enrichment analyses using slightly relaxed parameters (Bowtie2-DESeq2 data). We also evaluated overlap among the enriched pathways in the different MED15WY strains (compared to MED15LAB) (Fig. S3). The Venn diagram in Fig. 8 compares the pathways that are enriched in differentially expressed genes in strains differing only in their MED15 allele (MED15WY23, MED15WY7 and MED15WY15 and med15Δ) each compared to MED15LAB. We find, for example, that the reduced expression of genes in the arginine pathway that we observed in the med15 deletion strain relative to lab (Figs 4 and 8, light blue) is mirrored by the elevated expression of many of the same genes in the comparison of MED15WY23 and MED15LAB (Fig. 8, purple). This effect can also be seen in the elevated z-scores for ARG1, ARG3, ARG4, ARG5, ARG6, CPA1 and CPA2 in the MED15WY23 columns of the heatmap in Fig. 9. While the arginine pathway was not enriched in the MED15WY7 strain, there was a notable decrease in the z-scores for ARG2, ARG7 and ARG8 expression relative to MED15LAB, in contrast to the higher expression of this specific subset of arginine metabolism genes in the med15 deletion strain (Fig. 9). Venn diagram depicting shared and unique pathways that are enriched in S288C-based yeast strains with different MED15WY alleles. Genes regulated by MED15 presence/absence in the lab strain are categorized in the blue segment of the diagram. Some of the same gene categories are affected in comparisons of strains with MED15WY alleles compared to the MED15LAB allele. For example, arginine biosynthesis genes are affected by the absence of MED15 and by the MED15WY23 allele. Other affected gene categories are uniquely detected in comparisons of MED15LAB strains to multiple MED15WY alleles. For example, lysine biosynthesis is detected in MED15WY23 and MED15WY7 comparisons to MED15LAB but is not among genes detected in a comparison between MED15LAB and deletion strains. Finally, some gene categories (e.g. urea cycle genes) are unique to the MED15WY23 allele. This points to the existence of different gene expression patterns that may account for the enhanced fermentation characteristics exhibited by strains with MED15WY alleles and to the WY strains themselves. Heatmap representation of expression levels (Bowtie2) scaled per row from log2 transformed counts, in strains with different MED15 alleles for various genes in the arginine biosynthesis pathway as defined by the BioCyc database [54]. Two biological replicates (a, b) of each strain type are shown. Seven of the ten gene rows show relative downregulation in the med15 mutant, pointing to their normal activation by Med15, whereas ARG7, ARG2 and ARG8 are more highly expressed in the mutant, suggesting repression by Med15. Strikingly, these genes are under-expressed in strains with the MED15WY7 allele. Likewise, there is modest upregulation of ARG1 – CPA2 (bottom six rows) relative to ARG2, ARG7 and ARG8 in strains with the MED15WY23 allele. Similar overlap in pathways affected in the med15Δ mutant (reduced) and in strains with MED15WY7 alleles (elevated) are seen for genes involved in DNA metabolism including meiosis, cell cycle, replication and repair (Fig. 8, intersection of the blue and yellow regions). Likewise, an effect on branched chain amino acid biosynthesis is seen in the med15Δ mutant, MED15WY7, and MED15WY23 (Fig. 8, intersection of the blue, yellow and purple regions). Some affected pathways in strains with MED15WY alleles are only minimally affected in the med15Δ mutant. For example, MED15WY7 and MED15WY15 share a set of affected translation functions, ranging from translation initiation to nonsense-mediated decay and co-translational membrane targeting, that are not prominent among genes affected in the med15Δ mutant. Likewise, the effects of MED15WY23 and MED15WY7 compared with MED15LAB on lysine biosynthesis, and of MED15WY23 and MED15WY15 compared with MED15LAB on aerobic respiration and the urea cycle are not seen in the med15Δ mutant (Fig. 8) MED15-dependent arginine regulation affects fermentation efficiency Given the pronounced effect of MED15WY23 on arginine biosynthetic genes, we investigated the relevance of the arginine biosynthetic pathway in WGJ fermentation. Fermentation experiments were first conducted in minimal synthetic dextrose medium supplemented with 20% glucose and amino acids required by the strains. This media was then further supplemented with increasing amounts of arginine. As expected, we found that arginine supplementation (to 800 µg ml−1) reliably reduced the time to 50% weight loss by ~9%, and that this reduction took place regardless of MED15 genotype (Fig. S4) [67]. The impact of arginine limitation was evaluated by conducting fermentation in SC media with 20% glucose and all amino acids except for arginine (SC-Arg) (Fig. 10). We hypothesized that MED15WY alleles might tolerate the absence of arginine better than strains with the MED15LAB allele due to alterations in the expression of amino acid metabolic genes and better utilization of available nitrogen sources [1, 68]. The time to 50% weight loss in the two types of media (WGJ and SC-Arg) for each strain is given in Fig. 10b (data for WGJ is from Table 5 in Cooper et al. [1]; and was conducted on strains with episomal MED15 alleles). Interestingly, there are no significant distinctions among strains with different MED15WY alleles in WGJ. However, strains with the MED15WY7 and MED15WY23 alleles clearly outperformed strains with the MED15LAB and the MED15WY15 allele in media lacking arginine. This is consistent with the observation that arginine biosynthetic genes are elevated in the MED15WY23 strain during fermentation. Improved fermentation among strains with specific MED15WY alleles is evident in media lacking arginine. (a) Fermentation curves in SC media without arginine (containing all other amino acids) across strains with different MED15 alleles carried episomally. Significant differences between each strain and MED15LAB determined using two-tailed t-tests, **P<0.01. (b) Comparison of the fermentation efficiency of strains with different MED15WY alleles in WGJ and SC-Arg. Left, time to 50% weight loss for strains with different MED15WY alleles in WGJ (data from Table 5 in Cooper et al. [1]). Right, time to 50% wt loss for strains with different integrated MED15WY alleles in SC media with 20% glucose lacking arginine. Individual data points are plotted with mean and sd shown as horizontal lines. Significance was determined with one-way ANOVA and Tukey post-hoc tests, *P<0.05, **P<0.005, ****P<0.0001. Discussion Previously, we found that med15Δ yeast strains were compromised in the efficiency of the fermentation process, while MED15 alleles isolated from WY strains and swapped into a lab strain background in place of the MED15LAB allele conferred a modest enhancement of fermentation [1]. It is striking that a single gene from a WY strain can measurably alter fermentation when swapped into a standard laboratory strain. Yeast fermentation is a complex physiological process involving hundreds of genes that regulate metabolism, stress tolerance and nutrient uptake among other processes. The change in fermentation due to a single genetic change emphasizes that Med15 occupies a key regulatory control point for fermentation producing observable physiological differences. This effect motivated the current follow-up transcriptomic study to discern what new patterns of MED15WY-dependent gene expression might account for the fermentation booster phenotype. Transcriptional patterns in YPD are consistent with a role for MED15 in ethanol production Our analysis of the role of MED15 in fermentation was conducted in three parts. We first examined the effect of the MED15LAB allele on gene expression of specific fermentation-related genes in standard media YPD with 2% glucose (qRT-PCR, Fig. 1). Second, we looked at the effect of MED15 more globally in RNA-Seq experiments during WGJ-based alcoholic fermentation (RNA-Seq, Figs 3, 4 and 5). And third, we compared the transcriptomes of strains with MED15LAB and MED15WY alleles. Our targeted qRT-PCR data (Fig. 1b) showed that MED15 plays a broad role in regulating levels of metabolic enzymes in many steps in the glycolytic and alcoholic fermentation pathways. The expression patterns suggest that MED15 normally functions to promote ethanol production (Fig. 1, see ADH1 and ADH2). Our expression data also indicate that MED15 both positively and negatively regulates paralogous or redundant genes in multiple steps of these pathways during log-phase growth in YPD. For example, MED15 positively regulates the major isoform PDC1 and downregulates the minor isoform PDC6. MED15 positively regulates ADH1, which converts acetaldehyde to ethanol, and negatively regulates ADH2, which converts ethanol back to acetate [17]. Deletion of ADH2 has been explored for strain engineering to increase the production of bioethanol [69]. In quadruple mutants where the only ADH maintained is ADH1, yeast still produce wild-type levels of ethanol [70]. Therefore, considering the sum of MED15-dependent regulation of ADH genes, metabolism is expected to be skewed towards ethanol production. In addition to ethanol metabolism and fermentation genes, we selected genes based on their reported importance during fermentation physiology and metabolite production that influence flavour and stress adaptation [59–61, 71]. For example, glycerol, produced by the GPD1 gene, contributes to mouthfeel, sweetness and smoothness in wine and is beneficial at the start of fermentation for osmoregulation given the high sugar content of the grape or WGJ [6, 72]. Yeast strains have also been engineered to produce less ethanol by the overexpression of GPD1 to divert carbon to glycerol [61]. Finally, we found Med15 regulates amino acid metabolic genes, including ATF2, ARO10, MET10, and to a lesser degree, IRC7 during log-phase growth in YPD (Fig. 1b), which influence volatile compound profiles and sulfite production among others [73–76]. While winemakers and producers of bioethanol seek to engineer strains for ethanol production, it is possible that natural variation in the general transcriptional regulator MED15 could have permitted evolution to accomplish that same task. Transcriptional patterns under WGJ fermentation conditions The MED15 transcriptome is highly specialized during fermentation of WGJ. Less than 1% of DRGs in WGJ (20% glucose) are also identified in YPD (2% glucose). Our RNA-Seq analysis confirmed positive regulation by Med15 of many glycolytic and alcoholic fermentation genes (Figs 3 and 4). The regulation of these metabolic genes is an obvious explanation for the fermentation deficiencies we previously described for the med15Δ strain [1]. In addition to genes involved in the metabolism of glucose, MED15 regulates a variety of disparate cellular activities during fermentation including cell wall homeostasis, cell cycle, sexual reproduction, carbohydrate transport and amino acid biosynthesis (Figs 3 and 8). The regulation of these other cellular activities may also contribute to the fermentation phenotype of the med15Δ strain. For example, the yeast cell wall is known to change in composition depending on growth phase and the composition of the media [77, 78]. Specifically, growth in nitrogen-limiting media results in cell wall deformations potentially due to a reduced number of linkages between cell wall subunits [79]. Subtelomeric genes are silenced through heterochromatinization in a process referred to as the telomere position effect (TPE) [80]. Our analysis of subtelomeric gene expression during WGJ fermentation indicates that MED15 plays an enhanced role in telomeric silencing with 62% of analysed subtelomeric genes having significantly lower expression in the WT strain compared to the med15Δ strain. Med15-dependent subtelomeric silencing has been previously reported [62, 63, 81] and is observed to an extent on cells grown in minimal media laboratory conditions (SC); however, the effect on cells grown under rich media laboratory conditions (YPD) is modest and in opposition to the effect during WGJ fermentation (Fig. 5b). This suggests that either MED15-dependent telomeric silencing is induced in grape juice fermentation, and other nutrient-limited media, but not YPD, or that MED15-independent silencing is present in YPD but not grape juice fermentation. During fermentation, reinforcement of TPE may conserve cellular energy by limiting transcription of non-essential genes and stabilizing genome structure under stress. The mechanism of enhanced subtelomeric gene silencing observed during fermentation may be related to the increased levels of H4K16Ac and reduced levels of Sir2 binding at telomeres in Mediator tail subunit deletions [81]. Sir3 occupancy at telomeres is decreased in a med15Δ strain, which would correspond with a reduction in Sir2 binding [62, 63]. Another potential mechanism by which Med15 could influence telomeric silencing during fermentation is through its known transcription factor interacting partners. At least two Med15 interacting TFs have a secondary role in silencing of subtelomeric genes. Oaf1 and Msn4 exert this effect by conditionally binding to telomeric proto-silencing regions known as X-elements without their usual co-activating protein partners, usually in a condition-specific manner. The negative regulatory function of Oaf1 corresponds to Oaf1 conditionally binding X elements at telomeres while growing in media with oleate [82]. The WGJ environment may correspond to stabilization or increased binding of TFs to X-elements to promote silencing in a MED15-dependent manner. The reduced telomeric silencing in the deletion strain may be due to low levels of fatty acid production (a known phenotype of med15 mutants [83–85]) and thus less Oaf1 binding to X-elements. MED15WY23 affects expression of metabolic genes including the arginine pathway during fermentation We found previously that some MED15 alleles from WY confer enhanced fermentation traits when substituted for the MED15LAB allele in lab strains [1]. We therefore analysed expression patterns in lab strains expressing one of two MED15 alleles (MED15WY7 and MED15WY15) belonging to the European wine clade and a MED15 allele from a palm WY (MED15WY23). A set of genes differentially regulated by WY alleles were identified by GO and pathway enrichment analyses (Figs 6–8). MED15WY23 had the largest impact on the transcriptome. This allele differs from the MED15LAB allele primarily in the length of poly-Q tracts, there being only three non-synonymous substitutions outside of poly-Q tracts, and no impact of removing them was observed in our previous studies [1]. Hence, the difference in Q tract lengths in the MED15WY23 allele gene is likely accountable for the new transcriptome. The differentially expressed genes in strains with MED15WY23 compared to MED15LAB were enriched in metabolic activities as well as stress response, detoxification, transporter activity and mitochondrial/peroxisomal metabolism (Figs 6d and 7, Table S2). Especially noteworthy was the elevated expression of many arginine biosynthesis pathway enzymes (ARG1, 3–6) (Figs 6d and 7) in the MED15WY23 strain. These same arginine biosynthesis pathway enzymes were all significantly affected in comparisons of the MED15LAB strain compared to a med15 deletion strain (Fig. 4). This leads us to conclude that arginine biosynthesis from glutamine is dependent on MED15 and that the activation of these genes is enhanced by the MED1523 allele. In the row-normalized heatmap shown in Fig. 9, it was possible to detect not only the previously noted decrease in ARG1 and ARG3-6 expression in the med15 deletion mutant but also coordinated decreases in CPA1 and CPA2 gene expression, whose contribution to arginine biosynthesis is shown in Fig. 4c, and in URA2, a bifunctional protein whose N terminus comprises a carbamoyl phosphate synthetase domain that converts ammonium to carbamoyl phosphate, which feeds back into the urea cycle. Also apparent in the deletion mutant is a relative increase in expression of the ARG2, ARG7 and ARG8 genes that support the conversion of glutamate to ornithine and feeds back into the arginine biosynthetic pathway via ARG3-dependent conversion to citrulline. The heatmap also reveals that the MED15WY23 allele is not the only WY allele affecting the relative expression of genes in the arginine biosynthetic pathway. Strains with MED15WY7 show a relative reduction in the same ARG2, ARG7, ARG8 gene cluster that is more abundantly expressed in the deletion strain. These patterns suggest that the opposing expression of the two clusters of arginine biosynthetic genes is an important contributor to alcoholic fermentation and that small MED15-dependent adjustments that exaggerate the differential between the two clusters are one mechanism by which fermentation characteristics of a strain can be improved. Arginine is one of the major amino acids found in grape must and an important nitrogen source for yeast [68, 86]. Arginine is cleaved by arginase (CAR1 gene) into ornithine and urea, both of which can be used as a nitrogen source. Urea can be further catabolized to NH3 (ammonia) and CO2. This is an energy-dependent process carried out by DUR12 encoding urea amidolyase. Arginine and urea are considered secondary nitrogen sources used by the cell only when better sources like ammonia and glutamine are no longer present in the growth media. Constitutive expression of DUR12 in WY improves the degradation of urea during grape must fermentation, a reduction in fermentation time and increased ethanol production [87, 88]. Arginine is also known to have a protective effect against ethanol stress, in part by playing a role in the maintenance of cell wall integrity and in reducing oxidative stress [89], so that elevated arginine levels in strains with MED15WY alleles may also contribute to the ability of yeast strains to survive the stressful environment resulting from WGJ fermentation. Arginine biosynthesis involvement in MED15-dependent fermentation phenotypes The impact of the MED15 allele was clear in fermentation reactions conducted in media lacking arginine. Strains with the MED15WY7 or MED15WY23 allele performed significantly better than strains with the MED15LAB allele in arginine-deficient media, while strains with the MED15WY15 allele were unaffected (Fig. 10). Based on the heatmap in Fig. 9, we hypothesize that the MED15WY7 and MED15WY23 transcriptomes optimize arginine levels using two strategies. MED15WY23 promotes arginine production from glutamine (CPA1, CPA2, ARG1, 3, 4), while MED15WY7 does so by limiting ornithine production (Figs 7 and 9). Although MED15WY7 and MED15WY15 alleles had transcriptomes nearly identical to MED15LAB at the tested time point, there were a small number of differentially expressed genes, and these overlapped with differentially expressed genes in the MED15WY23 analysis, including MCH5, SNQ2, and LYS20. Since these genes are differentially regulated by at least two of the tested MED15WY alleles, they are also candidate contributors to WGJ fermentation efficiency. MCH5, a riboflavin transport gene [90] was previously identified as a fermentation stress response gene induced early in fermentation [91]. In addition to riboflavin in the environment, MCH5 is regulated by proline [92], which may be relevant to the limited nitrogen environment of WGJ. LYS20 encodes a homocitrate synthase and may be beneficial in the limited nitrogen environment of WGJ. Additionally, LYS20 functions in DNA repair [93], which may also be beneficial in the harsh WGJ environment. SNQ2 is a transporter involved in pleiotropic drug resistance, regulated by Pdr1/3 [94], and has previously been shown to serve a protective role in the fermentation environment [95]. In preliminary experiments, we found that deletion of the SNQ2 gene in strains with the MED15WY7 allele exhibited reduced fermentation (data not shown). Analysis of pathways rather than GO enrichment terms led to additional hypotheses concerning gene expression patterns that may contribute to increased fermentation rates as summarized in Fig. 8. Roles for general amino acid biosynthesis, translation, DNA metabolism and mitochondrial dynamics in fine-tuning fermentation await further analysis. One limitation of this study is that the influence of the MED15WY alleles on gene expression may vary at different times during the fermentation process. Future transcriptomic analyses of these and additional strains throughout the time course of a fermentation reaction could reveal the full effect of MED15 alleles on gene expression responsible for the boosted fermentation phenotypes we observed [1]. Supplementary material