Experimental evolution of Saccharomyces cerevisiae for caffeine tolerance alters multidrug resistance and target of rapamycin signaling pathways

Abstract Caffeine is a natural compound that inhibits the major cellular signaling regulator target of rapamycin (TOR), leading to widespread effects including growth inhibition. Saccharomyces cerevisiae yeast can adapt to tolerate high concentrations of caffeine in coffee and cacao fermentations and in experimental systems. While many factors affecting caffeine tolerance and TOR signaling have been identified, further characterization of their interactions and regulation remain to be studied. We used experimental evolution of S. cerevisiae to study the genetic contributions to caffeine tolerance in yeast, through a collaboration between high school students evolving yeast populations coupled with further research exploration in university labs. We identified multiple evolved yeast populations with mutations in PDR1 and PDR5, which contribute to multidrug resistance, and showed that gain-of-function mutations in multidrug resistance family transcription factors Pdr1, Pdr3, and Yrr1 differentially contribute to caffeine tolerance. We also identified loss-of-function mutations in TOR effectors Sit4, Sky1, and Tip41 and showed that these mutations contribute to caffeine tolerance. These findings support the importance of both the multidrug resistance family and TOR signaling in caffeine tolerance and can inform future exploration of networks affected by caffeine and other TOR inhibitors in model systems and industrial applications.


Introduction
Caffeine is an alkaloid produced by plants including coffee and cacao.It is desirable for its effects as a stimulant and also inhibits the master signaling regulator target of rapamycin (TOR) (Reinke et al. 2006).Saccharomyces cerevisiae yeast are used to ferment coffee and cacao pulp (Ludlow et al. 2016), and there they are exposed to concentrations of caffeine ranging from approximately 20-80 mM (3.5-16 mg/g) (Pandey et al. 2000;Asfew and Dekebo 2019;Melo et al. 2021).Thus, S. cerevisiae is capable of growing in high concentrations of caffeine, although laboratory and baking strains have much lower caffeine tolerance (Peter et al. 2018;Moresi et al. 2023).
Previous experimental evolution of haploid S. cerevisiae established that increased caffeine tolerance in a laboratory strain can be selected for in 48 passages; the one sequenced evolved clone from this study had mutations in PDR1, PDR5, and RIM8, indicating a role for the multidrug resistance (MDR) pathway in caffeine tolerance (Sürmeli et al. 2019).Caffeine can be exported from yeast cells by MDR transporters Pdr5 and Snq2 (Tsujimoto et al. 2015), which are regulated by transcription factors Pdr1, Pdr3, and Yrr1 (Cui et al. 1998).Pdr5 and Snq2 are highly homologous, but Snq2 has higher activity for caffeine efflux (Tsujimoto et al. 2015).
We also previously conducted experimental evolution of diploid S. cerevisiae for fewer than 20 passages and found evolved clones had increased caffeine tolerance and mutations in many genes including PDR1 (Moresi et al. 2023).However, we did not identify any genes with mutations in multiple evolved populations, which made differentiation of causative vs passenger mutations and prioritization of follow-up studies beyond PDR1 challenging.Work remains to more fully characterize inputs and regulators of TOR in yeast, so studying mutations that increase resistance to TOR inhibitors such as caffeine could identify additional sites, domains, and relationships that are critical for TOR signaling.Many components of TOR signaling are conserved across eukaryotes (González and Hall 2017), so studying TOR regulation in yeast can also help us understand conserved processes and identify differences across species.
In order to identify more genes and processes that contribute to caffeine tolerance, we performed experimental evolution of haploid S. cerevisiae to increase our ability to see the effects of loss-of-function and recessive mutations.As in our prior study on diploid yeast, we conducted experimental evolution as part of yEvo ("yeast Evolution"), an authentic research experience for high school students (Taylor et al. 2024).Previous yEvo studies on the effects of the antifungal drug clotrimazole identified new mutations, genes, and pathways involved in clotrimazole resistance (Taylor et al. 2022).Through yEvo, we were able to generate 45 independently evolved populations and involve high school students in scientific research.By studying the effects of mutations in caffeine-evolved yeast, we identified crucial roles for specific members of the MDR family and downstream effectors of TOR signaling.

Yeast strains and media
All evolution experiments were performed with MATɑ S288C derivative BY4742.All strain genotypes are listed in Supplementary Table 1, plasmids in Supplementary Table 2, and primers and oligos in Supplementary Table 3. Strains used in evolution experiments carried a 2-µm plasmid with KanMX, which provides resistance to the general antibiotic G418, and a pigment production pathway that gives each strain a unique color (courtesy of the Boeke Lab at the New York University); see Taylor et al. 2022 for additional description.Strains from the yeast deletion collection (Tong et al. 2001;Pan et al. 2004) were confirmed by PCR amplification followed by Sanger sequencing of their barcode (GENEWIZ from Azenta Life Sciences) (Supplementary Table 3).All experiments unless otherwise noted were performed in YPD (10-g yeast extract, 20-g peptone, and 20-g dextrose per liter).Caffeine (Sigma-Aldrich) was added to YPD for a final concentration of 40 mM.G418 disulfate (KSE Scientific, prepared as 200-mg/mL stock in water) was added to the media (200 mg/L) for strains carrying pigment plasmids.

Experimental evolution
Evolution experiments were carried out via batch transfer in YPD + G418 media.One to two times per week yeast were transferred using a sterile cotton swab from a saturated culture to a tube of 2-to 3-mL fresh media containing caffeine.All experiments began in 10 mM caffeine, and concentration was raised stepwise to 20, 30, and 40 mM; the maximum concentration reached in each experiment is listed in Supplementary Table 4. Evolved clones YMD4687-4712 were isolated from populations grown in 10-40 mM caffeine for 7 weeks in a 30°C incubator with no shaking, and YMD4821-4826 were grown in 10-40 mM caffeine for 8 weeks in a 30°C shaking incubator (Bio-Rad, 70 rpm).Evolved clones YMD4666-4686 were isolated from populations grown at room temperature in 10 mM caffeine for 5 weeks; YMD4684-4686 were then grown at 20 mM for an additional 5 weeks.At endpoint, glycerol stocks of each population were saved (1:1 yeast culture:50% glycerol), frozen at −20°C, and sent to the University of Washington for long-term storage at −72°C.Three clones were isolated from each population by streak purification and growth in caffeine was measured as detailed below; 1 or 2 clones with the highest growth rate in the presence of caffeine were selected from each population for further experimentation.Clones isolated from the same population are noted in Supplementary Table 5.

Growth measurements
To assay growth in the presence of caffeine or clotrimazole, clones were grown in 200-µL YPD medium in 96-well plates for 24 h at 30° C. Cultures were resuspended, and 2 µL of each culture was transferred to 198-µL YPD+-indicated concentration of caffeine or clotrimazole.For evolved clones and ancestors, which contain pigment plasmids, 200-mg/L G418 was also added.The growth of these cultures was monitored in a BioTek Synergy H1 microplate reader for 48 h at 30°C with orbital shaking.The average growth rate of all strains with the same genotype was calculated by linear fit to logarithmic growth phase with R 2 cutoff of 0.85 and plotted in R; script is available on GitHub (https://github.com/reneegeck/DunhamLab/blob/main/platereader_growthplotter.R).
For growth with rapamycin, clones were grown overnight in 2-mL YPD medium, with 200-mg/L G418 for strains containing pigment plasmids, at 30°C in a roller drum.This larger culture format was used because when in 96-well cultures, even high concentrations of rapamycin (>1 µM) did not fully inhibit the growth of ancestor strains, in contrast to prior reports that 25 nM leads to growth arrest (Barbet et al. 1996); growth in the larger cultures more closely matched expected effects.Experiment cultures were started at OD = 0.05 in 2-mL YPD ± 5 nM rapamycin (LC Laboratories, 10-mg/mL stock in ethanol), +200 mg/L G418 if carrying pigment plasmid, and grown for 24 h at 30°C in a roller drum.Optical density at 600 nM was measured on a Bio-Rad SmartSpec 3000.
To assay petite status, growth on YPD agar was observed, and clones were additionally transferred to YPG agar plates (glycerol as carbon source) and the presence or absence of growth was noted after 48 h at 30°C.

Whole-genome sequencing
Sample preparation, sequencing, determination of copy number variation, and identification of SNPs and indels were performed as previously described (Taylor et al. 2022).Briefly, DNA was purified using a phenol-chloroform and ethanol precipitation methodology.This DNA was tagmented using a modified Illumina Tagmentation protocol (Baym et al. 2015).Evolved clones were sequenced on an Illumina NextSeq 550 to an average coverage of 75× (range 10-178).Variants present in the ancestral strain YMD4612 (sequenced to 165×) were removed from consideration.Transposable element insertions were identified using McClintock v.2.0.0 (Nelson et al. 2017), filtered for nonancestral insertions, and inspected if they were identified by at least 3 component methods per clone.All point mutations, copy number variations, and transposable element insertions were manually inspected in the Integrative Genomics Viewer (Robinson et al. 2011) to ensure they were correctly called and not present in the ancestor.Mutations, transposon insertions, and copy number alterations for each clone are listed in Supplementary Table 5.Sequencing reads are deposited in the NCBI Sequence Read Archive (SRA) under BioProject PRJNA1101923.

CRISPR-Cas9 construction of point mutations
The nearest NGG polyspacer adjacent motif (PAM) sequence to the mutation of interest was manually identified and confirmed that it could be mutated by a synonymous mutation to stop Cas9 from cutting after a successful edit.Guide RNA (gRNA) sequence was selected as 20-bp upstream of the PAM.Oligos (Supplementary Table 3) were ordered from Integrated DNA Technologies to assemble the gRNA sequence into pNA0525 (Lauer et al. 2023) at the NotI cut site.Oligos were annealed and filled by 2 cycles of PCR, 5 ng assembled into 90-ng NotI-digested pNA0525 using Gibson Assembly Master Mix (New England Biosciences E2611), transformed into NEB5ɑ Escherichia coli (New England Biosciences C2987), and selected on LB + ampicillin agar plates.Successful integration of gRNA was confirmed using Sanger sequencing (GENEWIZ from Azenta Life Sciences).Donor templates with 40 bp of homology on each side of the desired edits were ordered and annealed as with gRNA.Strain YMD4042 was transformed with Cas9 expression plasmid pCTC019 (Zhao et al. 2023) and selected on c-leu to obtain strain YMD4909.For each mutation, an overnight culture of YMD4909 was transformed with 500n-g pNA0525-gRNA and 500-ng donor template using a lithium acetate protocol and selection on c-his-leu agar plates.Successful edits were confirmed using colony PCR and Sanger sequencing (Supplementary Table 3).

MDR transporter transcription reporters
Strains containing mutations in MDR transcription factors were transformed with a CEN plasmid containing URA3 and E. coli gene lacZ under the control of S. cerevisiae promoters from PDR5, SNQ2, or YOR1 (Katzmann et al. 1994;Kolaczkowska et al. 2008) or a control plasmid containing URA3 but no lacZ gene (p416CYC1) (Mumberg et al. 1995).Resultant strains (YMD5073-86 and YMD5138-49; Supplementary Table 1) were grown overnight in c-ura, diluted to OD = 0.1, and 5-µL spotted onto c-ura agar plates containing 0.07 M KPO 4 pH7 and 40 mg/L X-gal (American Biorganics, prepared as 2-mg/mL stock in dimethylformamide).Plates were photographed after 3 days of growth at 30°C.

Statistical analysis
All statistical analysis was performed in R. Code for statistical analysis and generating figures is available at https://github.com/reneegeck/DunhamLab/tree/main/Geck2024_CaffeineYeast.

Evolution of caffeine-tolerant yeast
Students and university lab members evolved 45 populations of S. cerevisiae in increasing concentrations of caffeine for 5-10 weeks (approximately 10-20 transfers).Three clones were isolated from each population and tested for their ability to grow in the presence of caffeine relative to the ancestral strains; based on this, 1 or 2 clones per population were selected for further replicates of the growth assay (n = 3 per clone) and downstream analysis for a total of 53 clones (Fig. 1; Supplementary Table 4).The majority of clones (43/53, from 38/45 populations) had a significantly faster growth rate than the ancestor strains in caffeine (Supplementary Fig. 1), indicating successful selection of caffeinetolerant clones; some grew more slowly than the ancestor in the absence of caffeine, but not significantly slower (Supplementary Fig. 1).Raising the concentration of caffeine in the evolution experiment selection was critical for selection of caffeine-tolerant clones, since 9/10 clones that did not grow better than the ancestor in caffeine were only grown in 10 mM caffeine for the evolution process.Stepwise increasing the concentration also improved the ability to grow at higher concentrations of caffeine, but clones that only experienced 20-30 mM caffeine during evolution still had significantly increased growth in 40 mM caffeine (Fig. 1).

Genetic characterization of caffeine-evolved yeast
To investigate the genetic changes contributing to increased caffeine tolerance, we sequenced all 53 evolved clones.Of the 10 clones without increased caffeine tolerance, most either had no mutations or only 1 mutation in a gene not previously found connected to caffeine or TOR (Supplementary Table 6); 3 that lost mitochondrial DNA (mtDNA) but lacked nonsynonymous mutations did grow slightly faster in caffeine than ancestor strains (Supplementary Fig. 2a), in line with previous reports that loss of mtDNA increases growth in the presence of caffeine (Bard et al. 1980).These 10 clones that did not grow significantly faster than the ancestor in caffeine were excluded from further analysis to focus only on mutations likely to contribute to caffeine tolerance.
Every caffeine-tolerant clone had at least 1 nonsynonymous point mutation, 17/43 also lost mtDNA, and 7/43 also had a Ty insertion or focal copy number gain (Fig. 2a; Supplementary Table 5).Only 2 evolved clones had copy number alterations (Supplementary Fig. 2b).Clone YMD4666 did not have increased caffeine tolerance (Supplementary Fig. 1); clone YMD4679 had a duplication of chrIII: 15,2000-167,000.Several genes fall within this region, but one, the nucleotide pyrophosphatase NPP1, has been previously connected to caffeine tolerance through deletion set screens where npp1Δ strains were more sensitive to caffeine (Dudley et al. 2005;Brown et al. 2006).
The recurrently mutated genes PDR1, PDR5, and SIT4 are involved in MDR; Pdr5 is a drug efflux pump, and Pdr1 and Sit4 modulate drug pump expression (Miranda et al. 2010;Buechel and Pinkett 2020).All but 3 caffeine-tolerant evolved clones had a mutation in one or more of these genes (Fig. 2c).Mutations in PDR1, PDR5, and SIT4 were found in populations evolved by slightly varied strategies in different classrooms (see "Materials and methods"), emphasizing that the experimental evolution strategy is robust to variation.We did not have enough evolved clones in each group to determine if the effects of different combinations of PDR1, PDR5, and SIT4 mutants had different effects on growth in caffeine, except that ones with mutations in both PDR1 and PDR5 grew faster in caffeine than ones with only PDR1 mutations (Supplementary Fig. 2c).The 3 clones without mutations in these genes (YMD4678, YMD4679, and YMD4705) contained nonsynonymous mutations in CAT8, SKY1, TIP41, and YRR1.Mutations in Sky1, a downstream effector of TOR, were found in 7 evolved populations (Fig. 2b).This included a nonsense mutation SKY1 Q621X , indicating that SKY1 loss of function is likely an important contributor to caffeine tolerance.This is supported by previous high-throughput findings that sky1Δ increases caffeine and rapamycin tolerance; Sky1 is also phosphorylated dependent on TOR (Brown et al. 2006;Huber et al. 2009).

Cross-resistance conferred by mutations in MDR family genes
We hypothesized that many of the MDR mutations in our caffeine-evolved clones were not specific to caffeine but could confer resistance to other drugs.We previously evolved and characterized yeast with resistance to the antifungal drug clotrimazole, many of which also had gain-of-function mutations in transcription factor Pdr1 or amplification of efflux pump Pdr5 (Taylor et al. 2022).To determine if any mutations found in caffeine-tolerant clones confer cross-resistance to an unrelated drug, we grew clones in varying concentrations of clotrimazole.In general, growth in clotrimazole did not correlate strongly with growth in caffeine, although the clones that grow best in clotrimazole all have mutations in PDR1 or PDR5 (Supplementary Fig. 3a  and b and Table 8).
In our previous study of clotrimazole-evolved clones, we observed many mutations affecting PDR1, PDR3, and PDR5, but never PDR1 and PDR3 mutations in the same clone, suggesting that having mutations in both transcription factors may be redundant (Supplementary Fig. 3c).In contrast, a higher proportion of caffeine-evolved clones had nonsynonymous mutations in both transcription factor Pdr1 and efflux pump Pdr5, whereas PDR5 amplification was seen in clotrimazole-evolved clones alone or together with a PDR1 or PDR3 mutation.Additionally, while many caffeine-tolerant clones had PDR1 mutations, none had a mutation in PDR3, suggesting these transcription factors may play different roles in response to caffeine.Like the mutations in PDR1 and PDR3 from clotrimazole-evolved yeast, PDR1 mutations in the caffeine-tolerant yeast were also gain of function since they also increased caffeine tolerance in a heterozygous diploid (Supplementary Fig. 3d and Table 9).To investigate why we did not find any mutations in PDR3 in our caffeine-evolved yeast, we grew yeast with gain-of-function mutations in PDR1, PDR3, or both in caffeine, including 2 caffeine-evolved strains with PDR1 G280A or PDR1 L1056H and 2 clotrimazole-evolved strains with PDR1 F479I or PDR3 T949A .When grown in clotrimazole, PDR1 F479I and PDR3 T949A confer equal resistance (Taylor et al. 2022).However, PDR3 mutations were significantly less able to confer increased caffeine tolerance than PDR1 mutations (Fig. 3a; Supplementary Fig. 3e and Table 10).Therefore, most mutations in MDR genes conferred general drug resistance, but some had stronger effects on clotrimazole resistance than caffeine tolerance.

Pdr1, Pdr3, and Yrr1 regulate caffeine tolerance
The 2 main efflux transporters for caffeine are SNQ2 and PDR5 (Fig. 3b) (Tsujimoto et al. 2015).To determine which targets were most affected by the gain-of-function mutations in transcription factors Pdr1 and Pdr3, we used transcriptional reporters for PDR5, SNQ2, and YOR1.Each reporter contains the E. coli lacZ gene under control of the promoter of PDR5, SNQ2, or YOR1 (Kolaczkowska et al. 2008), so if the transcription is increased, yeast colonies will turn blue when grown on media containing X-gal.We observed that PDR1 G202A and PDR1 F749I led to a small increase in transcription from the PDR5 promoter and PDR3 T949A led to a greater increase (Fig. 3c; Supplementary Fig. 3f).This is in line with our conclusion that these mutations increase function since PDR5 is a target of Pdr1 and Pdr3 (Kolaczkowska and Goffeau 1999), but does not explain the difference in caffeine tolerance provided by PDR1 mutations over PDR3 mutations (Fig. 3a), since PDR3 T949A led to more transcription from PDR5 than PDR1 G202A .The effects of the mutation on expression from the SNQ2 promoter were also contrary to expectations: while PDR1 G202A conferred the fastest growth in caffeine (Fig. 3a), only PDR1 F749I and PDR3 T949A increased expression from the SNQ2 promoter (Fig. 3c; Supplementary Fig. 3f).Further investigation of this transcriptional network and the targets of these Pdr1 and Pdr3 mutants is needed to understand their differential contributions to caffeine tolerance.
We also wanted to investigate if any mutations in caffeinetolerant yeast had a greater effect on SNQ2 expression.While Pdr1 and Pdr3 are important for basal expression of SNQ2, increased activation in response to drugs is mainly mediated through another MDR transcription factor, Yrr1 (Cui et al. 1998) (Fig. 3b).YRR1 was previously found to have increased expression in a caffeine-tolerant evolved yeast strain (Sürmeli et al. 2019), and we found a mutation in YRR1 in one of our caffeine-tolerant clones (Supplementary Table 6).We hypothesized that our observed YRR1 T699P mutation leads to increased Yrr1 activity and thus contributes to MDR.We constructed the YRR1 T699P mutation using CRISPR-Cas9 and observed that it was sufficient to increase growth in caffeine in a haploid (Fig. 3d; Supplementary Fig. 3g and Table 11) and in a heterozygous diploid (Supplementary Fig. 3h and Table 11).Transcription from the PDR5 promoter was basally active, and the mutation did not increase it, but YRR1 T699P did increase transcription from SNQ2 and YOR1 promoters (Fig. 3e; Supplementary Fig. 3f), supporting that this mutation increases Yrr1 activity toward its targets.This is a similar transcriptional effect as seen by other gain-of-function YRR1 mutations (Kodo et al. 2013).Thus, Pdr1, Pdr3, and Yrr1 activities are all important for caffeine tolerance, likely by their effects on caffeine efflux transporters Pdr5 and Snq2.

Many loss-of-function mutations contribute to caffeine and rapamycin tolerance
To investigate mutations that are more specific to caffeine effects on TOR signaling, we also looked at cross-resistance to rapamycin.Both caffeine and rapamycin bind the FRB domain of Tor1 (Reinke et al. 2006), so mutations that confer caffeine tolerance by affecting TOR signaling should also confer rapamycin resistance.We confirmed that evolved clone YMD4711, which has a mutation in the FRB domain of Tor1, has increased growth in caffeine (Supplementary Fig. 1), and its growth is not inhibited by rapamycin (Supplementary Table 12).Across all strains, growth in caffeine weakly correlates with growth in rapamycin, although the majority of evolved clones are highly resistant to rapamycin and double rapidly in caffeine (Supplementary Fig. 4a and Table 12).
Two of the recurrently mutated genes, SIT4 and SKY1 (Fig. 2b), have been previously shown to affect rapamycin tolerance (López-Mirabal et al. 2008;Huber et al. 2009).We identified likely loss-of-function mutations in both genes in caffeine-tolerant clones that did not have mutations in PDR1 or PDR5, suggesting the SIT4 and SKY1 loss of function may be conferring caffeine tolerance.To investigate more genes that may be implicated in caffeine and rapamycin resistance, we focused on all genes in caffeine-tolerant strains with likely loss-of-function mutations (EBS1 I18indel , ESL1 W658X , SIT4 N269indel , SKY1 Q621X , SVF1 F284indel , TIP41 S162X , TOM20 S36indel , and USA1 G37indel ).Of these 9 genes, ESL1 is the only essential gene, but hypomorphic alleles have been constructed (Breslow et al. 2008), so ESL1 W658X is likely hypomorphic.Deletions of the other genes were available in the yeast deletion collection (Tong et al. 2001;Pan et al. 2004), so we measured their growth in caffeine and rapamycin.Deletions of EBS1, SIT1, SKY1, SVF1, TIP41, and USA1 led to increased growth in caffeine, indicating that the loss-of-function mutations we observed in these genes are not passengers from the evolution experiment (Fig. 4a; Supplementary Fig. 4b and Table 13).Only ebs1Δ, sit4Δ, svf1Δ, and tip41Δ also increased growth in rapamycin (Fig. 4b; Supplementary Table 14), indicating that they contribute to caffeine tolerance through TOR signaling and that the role of Sit4 in caffeine tolerance may be through both the MDR pathway and TOR signaling (López-Mirabal et al. 2008;Miranda et al. 2010).However, in a previous screen of the systematic deletion set, svf1Δ increased sensitivity to rapamycin (Xie et al. 2005), so its effects may be dose or assay dependent.There is also literature support that ebs1Δ and sky1Δ confer resistance to rapamycin (Ford et al. 2006;Huber et al. 2009), although we only observed a slight effect of ebs1Δ, suggesting our assay may not capture smaller effects.
TOR effectors Sky1, Sit4, and Tip41 contribute to caffeine tolerance SIT4 and TIP41 are involved in PP2A-like signaling downstream of TOR, as is RRD1, in which we found a RRD1 D143H mutation (Fig. 5a).Loss of SIT4 and RRD1 has been previously shown to increase caffeine tolerance in high-throughput screens (Kapitzky et al. 2010;Hood-DeGrenier 2011), but TIP41 has not.To confirm the contribution of TIP41 S162X , we constructed the mutation using CRISPR-Cas9 and observed that TIP41 S162X increased growth in caffeine and rapamycin (Fig. 5b and c; Supplementary Fig. 5a and Tables 11 and 14).The TIP41 S162X mutation had no effect on caffeine tolerance in a heterozygous diploid, supporting that it is loss of function and recessive (Supplementary Fig. 5b and c).We also constructed a point mutation in SIT4, SIT4 T95R , which increased growth in the presence of caffeine and rapamycin (Fig. 5b and d; Supplementary Fig. 5a and Tables 11 and 14), indicating that inhibition of PP2A-like signaling is important for caffeine tolerance.The SIT4 T95R mutation is likely loss of function since it had no effect on caffeine tolerance in a heterozygous diploid (Supplementary Fig. 5b and c).
Sky1 also functions downstream of TOR, but unlike Tip41 and Sit4, it directly associates with Tor1 instead of with Tap42 (Fig. 5a) (Huber et al. 2009;Breitkreutz et al. 2010).We constructed a SKY1 E596K point mutation using CRISPR-Cas9 and observed that it increased growth in caffeine in a haploid (Fig. 5e; Supplementary Fig. 5a and Table 11) but not heterozygous diploid (Supplementary Fig. 5b and c), indicating that the mutation likely decreases Sky1 function to increase caffeine tolerance.As seen with sky1Δ, SKY1 E596K did not significantly increase growth in 5 nM rapamycin (Fig. 5c; Supplementary Table 14).

Discussion
We successfully evolved laboratory yeast to have increased caffeine tolerance and identified contributing mutations in genes related to MDR and TOR signaling.In every caffeine-tolerant evolved clone, we identified at least 1 likely causative mutation: either in MDR pathway genes PDR1, PDR5, and YRR1 or TOR pathway genes TOR1, SKY1, TIP41, and SIT4.This is in agreement with a previous experimental evolution study in caffeine, where 1 clone was sequenced and found to have mutations in PDR1, PDR5, and RIM1 (Sürmeli et al. 2019).In many genes, we identified multiple mutations in caffeine-tolerant clones, some affecting the same or nearby sites.Future work focusing on the effects of groups of mutations could elucidate important functional and structural properties of these proteins.
The majority of our caffeine-evolved yeast clones had a mutation in a MDR pathway gene, and characterizing these mutations provides insight into functional interactions and important regions of MDR proteins.We observed many mutations in PDR1, but none in PDR3.Pdr1 may be more important than Pdr3 for the expression of main caffeine efflux transporter genes PDR5 and SNQ2, since pdr1Δ leads to a much greater decrease in PDR5 and SNQ2 transcripts than pdr3Δ (Mahé et al. 1996), although we did not observe strong activation of SNQ2 transcription and saw stronger activation of PDR5 transcription by our Pdr3 gain-of-function mutant than our Pdr1 mutants.This may be specific to this small number of mutants studied, but more quantitative measurement of the effects of these mutations and broader study of their effects on transcription of other genes beyond PDR5 and SNQ2 are needed to identify possible reasons for the difference in effect strength of Pdr1 and Pdr3 mutants on caffeine tolerance.We also added to the known gain-of-function mutations in YRR1 that increase SNQ2 expression and caffeine tolerance (Kodo et al. 2013;Rong-Mullins et al. 2018).Several of these mutations are near the activation domain of Yrr1 (Gallagher et al. 2014), including the YRR1 T699P mutant we identified, suggesting that mutations in this region are important for conferring resistance to various drugs.By characterizing mutations in MDR pathway genes, we are able to identify more gain-of-function mutants, which increase our understanding of functional domains in these proteins.
The high percentage of evolved clones with a mutation in a MDR gene somewhat hampered our ability to study TOR signaling using caffeine, since most clones had a mutation that could increase caffeine efflux, so the selection pressure for rewiring of intracellular signaling was not as strong.In the future, evolving a strain lacking key MDR components could lead to selection for different types of mutations, as done previously for the evolution of rapamycin-resistant pdr1Δ yeast (Wride et al. 2014).
While we did not identify new relationships between TOR signaling effectors, we did show that loss of function of Tip41 increased caffeine tolerance.We also observed many mutations in components of the PP2A-like complex, but none in PP2A complex members.Few PP2A-like specific downstream processes have been annotated, as many are presumed to be shared with PP2A (Jiang 2006), but specific alpha-arrestins and factors involved in cell cycle progression and budding have been proposed to be downstream of PP2A-like signals (Fernandez-Sarabia et al. 1992;Watanabe et al. 2009;Talaia et al. 2017;Bowman et al. 2022).Different regulatory strength of PP2A and PP2A-like could also lead to increased tolerance or benefit of mutations in PP2A-like factors that spare processes proposed to be only downstream of PP2A, such as autophagy (Yorimitsu et al. 2009).Studying the differential regulation of these downstream processes could increase our understanding of how yeast responds to caffeine and other drugs and stressors that affect TOR signaling.Yeast that are used for coffee and cacao fermentation are exposed to high concentrations of caffeine, and studying natural variation in their TOR pathway components compared to other domesticated yeast could address how critical different factors are for growth in Fig. 5. Contributions of TOR effectors to caffeine tolerance.a) TOR signaling pathway (Huber et al. 2009;Numamoto et al. 2015;Ferrari et al. 2017;Ariño et al. 2019) annotated with mutations found in caffeine-tolerant clones in this study and prior literature support for effects of deletions on caffeine tolerance (Rempola et al. 2000;Brown et al. 2006;Reinke et al. 2006;Banuelos et al. 2010;Kapitzky et al. 2010;Hood-DeGrenier 2011).b) Doubling time for strains engineered to have synonymous TIP41 Q160= , with or without TIP41 S161X mutation.c) Growth after 24 h in 5 nM rapamycin for engineered strains.d) Doubling time for strains engineered to have SIT4 mutation; both have synonymous SIT4 L91= mutation.e) Doubling time for strains with SKY1 mutation; both have synonymous SKY1 T599= mutation.Difference from strain with only synonymous mutation by t-test: **P < 0.01 and ****P < 0.0001.caffeine.Many domesticated yeast have been isolated and sequenced (Peter et al. 2018), but few from coffee and cacao have had their whole genomes sequenced that would be necessary for this further analysis (Ludlow et al. 2016).
The strength of our study was in the number of independent evolution experiments and ability to compare it to other experiments carried out in the same yEvo system but under a different selective pressure.Firstly, we showed that caffeine tolerance in haploid yeast can be selected for in far fewer than 48 passages, as was previously done (Sürmeli et al. 2019), since some of our caffeine-tolerant clones were derived from populations that were passaged only 10 times.Additionally, by comparing our selection for caffeine tolerance to prior yEvo selections for clotrimazole resistance, we identified notable similarities and differences in the mutations that were selected for in each regime.Yeast lacking mtDNA were frequently selected for in both caffeine and clotrimazole, confirming previous reports that mtDNA loss increases growth in the presence of azoles and caffeine, at least partly by increased Pdr3 activity (Bard et al. 1980;Hallstrom and Moye-Rowley 2000;Ferrari et al. 2011;Taylor et al. 2022).A notable difference between conditions is that while clotrimazole-evolved yeast exhibited copy number changes, including whole chromosome gains, and mutations in PDR3 (Taylor et al. 2022), copy number changes were rare (2/45 populations) and PDR3 mutations absent in caffeine-evolved yeast.Most mutations that affected efflux transporter PDR5 in clotrimazole-evolved clones were copy number gains (11/13), whereas caffeine-evolved clones had many PDR5 point mutations.Since both of these experiments started with the same strains and used the same experimental evolution protocol of serial transfers, we can conclude that the differences are due to the differential ability of these mutations to confer caffeine tolerance compared to clotrimazole resistance.As we carry out more experiments through yEvo, we are compiling a rich data set to investigate these differences and similarities between the effects of different selective pressures.

Fig. 2 .
Fig. 2. Common mutations in caffeine-tolerant clones.a) Types of mutations in evolved clones with significantly increased caffeine tolerance.b) Genes with mutations in caffeine-tolerant evolved lineages and their prior support for connection to caffeine tolerance or TOR signaling.c) Doubling time of caffeine-tolerant clones grouped by mutations in MDR genes.Difference from ancestor by ANOVA with Tukey's honestly significant difference: ****P < 0.0001.

Fig. 3 .
Fig. 3. Gain-of-function mutations in MDR family genes contribute to cross-resistance to clotrimazole.a) Doubling time of WT (YMD5096) compared to clones with PDR1 and PDR3 mutations from caffeine (YMD4681 and YMD4684) or clotrimazole evolution experiments (YMD5097 and YMD5098).b) MDR transporters and their regulation by transcription factors based on Cui et al. (1998) and Kolaczkowska and Goffeau (1999), with effect on caffeine efflux from Tsujimoto et al. (2015).Dashed line arrows indicate basal activity; solid arrows indicate drug-induced activity.c) Reporter assay of PDR1 and PDR3 mutant strains expressing lacZ under the control of indicated promoters, grown 72 h on media containing X-gal.d) Growth in caffeine of CRISPR engineered strains with synonymous YRR1 T696= mutation, with or without YRR1 T699P mutation.e) Reporter assay of engineered YRR1 strains expressing lacZ under the control of indicated promoters, grown 72 h on media containing X-gal.Difference from wild type (WT) or synonymous by ANOVA with Tukey's honestly significant difference: **P < 0.01, ***P < 0.001, and ****P < 0.0001.