Partial Decay of Thiamine Signal Transduction Pathway Alters Growth Properties of Candida glabrata

The phosphorylated form of thiamine (Vitamin B1), thiamine pyrophosphate (TPP) is essential for the metabolism of amino acids and carbohydrates in all organisms. Plants and microorganisms, such as yeast, synthesize thiamine de novo whereas animals do not. The thiamine signal transduction (THI) pathway in Saccharomyces cerevisiae is well characterized. The ~10 genes required for thiamine biosynthesis and uptake are transcriptionally upregulated during thiamine starvation by THI2, THI3, and PDC2. Candida glabrata, a human commensal and opportunistic pathogen, is closely related to S. cerevisiae but is missing half of the biosynthetic pathway, which limits its ability to make thiamine. We investigated the changes to the THI pathway in C. glabrata, confirming orthologous functions. We found that C. glabrata is unable to synthesize the pyrimidine subunit of thiamine as well as the thiamine precursor vitamin B6. In addition, THI2 (the gene encoding a transcription factor) is not present in C. glabrata, indicating a difference in the transcriptional regulation of the pathway. Although the pathway is upregulated by thiamine starvation in both species, C. glabrata appears to upregulate genes involved in thiamine uptake to a greater extent than S. cerevisiae. However, the altered regulation of the THI pathway does not alter the concentration of thiamine and its vitamers in the two species as measured by HPLC. Finally, we demonstrate potential consequences to having a partial decay of the THI biosynthetic and regulatory pathway. When the two species are co-cultured, the presence of thiamine allows C. glabrata to rapidly outcompete S. cerevisiae, while absence of thiamine allows S. cerevisiae to outcompete C. glabrata. This simplification of the THI pathway in C. glabrata suggests its environment provides thiamine and/or its precursors to cells, whereas S. cerevisiae is not as reliant on environmental sources of thiamine.

Saccharomyces cerevisiae synthesizes TPP de novo through steps of condensation, hydrolysis, and pyrophosphorylation of HMP-PP and HET-P [4] (Fig 1A). HMP synthesis utilizes the proteins encoded by genes SNO2/3 and SNZ2/3, which are regulated by thiamine availability, to make the vitamin B 6 (also known as pyridoxal-5'-phosphate or PLP) [4,5]. Members of the THI5 gene family (THI5/11/12/13) are all functionally redundant and synthesize HMP-P from PLP [6,7]. After being formed, HMP-P is phosphorylated by either Thi20 or Thi21 to form HMP-PP; however, Thi20/Thi21 is also capable of phosphorylating HMP to HMP-P [8]. The single copy THI4 gene is involved in HET-P synthesis, and during thiamine-depleted conditions, transcription of THI4 produces one of the most abundant proteins in the cell [9]. Recently, Thi4p was discovered to act as a co-substrate and not an enzyme and can only undergo a single reaction, meaning it is a suicide thiamine thiazole synthase [10]. HET phosphorylation to form HET-P is catalyzed by an enzyme, HET kinase, that is encoded by THI6 [1]. THI6 also condenses HMP-PP and HET-P to thiamine phosphate. The final steps of TPP synthesis require the dephosphorylation and pyrophosphorylation of thiamine phosphate. The pyrophosphorylation is carried out by thiamine pyrophosphokinase and is encoded by THI80 [4,11].
In addition to being able to synthesize thiamine de novo, S. cerevisiae can utilize extracellular thiamine to synthesize TPP. There are two genes that are known to be involved in the acquisition of extracellular thiamine, THI10 (also known as THI7) and PHO3 [13][14][15]. THI10 encodes for a thiamine transporter (and there are likely additional thiamine transporters [16]) in the plasma membrane while PHO3 encodes a periplasmic acid phosphatase that allows for the intake of thiamine by dephosphorylating TPP to thiamine [4]. Both of these genes are under control of the THI signal transduction pathway and are induced during thiamine-depleted conditions. S. cerevisiae can also utilize extracellular HMP and HET to form TPP. HMP is thought to be brought into the cells via thiamine transporters but HMP also can be released from yeast cells when too much HMP is present [17]. HET is brought into the cells through diffusion and is trapped due to HET kinase-catalyzed phosphorylation [18].
To prevent wasted energy, when thiamine is present in the environment, thiamine is not biosynthesized and the high-affinity transport system for thiamine (but presumably not the low-affinity transport system) is down regulated [2,19]. Three positive regulatory factors have been identified in the THI regulatory pathway, THI2, THI3, and PDC2 [16,19,20]. Both Thi2p and Pdc2p have DNA-binding domains in the N-terminus and appear to interact with Thi3p ( Fig 1B). When any of the regulatory factors are deleted, the transcriptional induction of genes in response to low thiamine environmental conditions is dramatically diminished [4].
In our studies with C. glabrata, we noted that the gene encoding the transcription factor THI2 is found in S. cerevisiae but not in C. glabrata. We also noted that some parts of the THI pathway seem to be lacking in C. glabrata relative to S. cerevisiae. C. glabrata has been demonstrated to be a thiamine and pyridoxine auxotroph previously [21][22][23] but we expected that the entire pathway would be absent. There are many examples of full pathway degradation in C. glabrata [24][25][26], but our interest was piqued by the apparent partial loss of the THI pathway. Additionally, given the evolutionary distance between the two species and the differences in natural habitats [27,28], we wished to explore further the THI pathway in C. glabrata.

Results
C. glabrata has lost the ability to synthesize PLP and HMP, leading to a defect in thiamine production relative to S. cerevisiae To determine the alterations in the thiamine biosynthetic pathway in C. glabrata relative to S. cerevisiae, we identified the presence/absence of orthologs between the two species with regards Thiamine biosynthesis pathway and regulation in S. cerevisiae and C. glabrata. (A) A schematic of the thiamine biosynthesis pathway with shared components labeled in black and those found only in S. cerevisiae labeled in green. Lack of Thi21p in C. glabrata likely does not result in the inability to phosphorylate HMP because Thi20p is present. The schematic is a compilation of data from [4,12]. All compounds are described in the text except RP and XP which are ribulose-5-phosphate and xylulose-5-phosphate (B) A schematic of the transcriptional regulation of THI genes in S. cerevisiae and C. glabrata.
to the known biosynthetic pathway in S. cerevisiae (Fig 1A) (gene names labeled in green are absent from C. glabrata). C. glabrata appeared to be unable to synthesize HMP, as there are no apparent orthologs of the ScTHI5 family present in C. glabrata. Additionally, there is a lack of the SNO/SNZ genes [22] indicating that C. glabrata is unable to synthesize PLP, the precursor to HMP. Analysis of the THI regulatory pathway also indicated that THI2 was absent from C. glabrata (Fig 1B). Through analysis of the Fungal Orthogroups and PhylomeDB websites [29,30], the pre-genome duplication common ancestor of C. glabrata and Saccharomyces species had the THI5, SNO/SNZ, and THI2 genes, suggesting that C. glabrata (and the closely related "glabrata group" of species [22]) has specifically lost these genes relative to other species (S1 Fig). Additionally, it appears that the Saccharomyces senso stricto species have increased the copy numbers of the SNO, SNZ, and THI5 genes relative to the pre-genome duplication species. Interestingly, all four of the THI5 gene family members in S. cerevisiae are sub-telomeric and there are two copies of the three gene array SNO/SNZ/THI5 in the genome, indicating a small scale duplication has increased the gene numbers [31]. The loss of conserved synteny of the SNO/SNZ/THI5 gene families and the increased numbers of orthologs of these genes suggests that there was a selective advantage in the Saccharomyces clade to having increased thiamine and PLP biosynthesis and conversely that C. glabrata does not benefit from increased thiamine production via HMP synthesis. Recent work demonstrates that the physical linkage of genes in yeast may be an adaptation against toxic intermediates [32], and the linkage of these SNO/ SNZ/THI5 genes in S. cerevisiae may be advantageous when there is a large metabolic flux through the THI pathway.
To determine whether C. glabrata utilized the THI regulatory pathway (and consequently the biosynthetic pathway), we inactivated putative regulators of the biosynthetic pathway, THI2 and THI3 in S. cerevisiae and THI3 and PDC2 in C. glabrata, to see if there was a defect in growth during thiamine starvation (Fig 2). We additionally inactivated genes in the biosynthetic pathway (S2 Fig). Deletion of ScPDC2 was not included in this analysis because it is essential for growth in high glucose conditions [16]. CgPdc2 appears to have a different transcriptional specificity relative to ScPdc2 as deletion of CgPDC2 does not result in an obvious growth defect in high glucose conditions (data not shown). As expected, there was a defect for the S. cerevisiae deletion strains relative to wild-type in the absence of thiamine. We also observed a small, but marked, defect in the ability of the C. glabrata mutants to grow relative to wild-type. The most parsimonious explanation for these data are that the thiamine biosynthetic pathway, while missing components, is still active in synthesizing additional thiamine, albeit at a much lower level than S. cerevisiae, and loss of either the transcriptional regulators or the biosynthetic genes leads to a defect in growth in the absence of thiamine.
As the sequenced genome of C. glabrata has lost genes related to PLP synthesis, we expected this species to be a PLP auxotroph in addition to a thiamine auxotroph [22,33]. When grown in the absence of thiamine and/or PLP, C. glabrata exhibits a large growth defect whereas S. cerevisiae grows well (Fig 3A and 3B). We predicted that if C. glabrata was able to synthesize HMP, growth would be enhanced in the absence of thiamine but not in the absence of PLP, as PLP is required for metabolic reactions independent of thiamine. To test this hypothesis, we generated a C. glabrata strain that had the ScTHI5 promoter and ORF integrated into the CgURA3 locus. We also generated a C. glabrata strain containing a plasmid where the ScTHI5 gene was overexpressed by the CgPGK1 promoter, which is a highly expressed gene in standard medium conditions. We anticipated that the two promoter constructs should provide a doseresponse of low and high levels of ScThi5. Growth was measured for these strains, and wildtype strains, in replete, thiamine starved, PLP starved, and thiamine/PLP starved conditions (Fig 3A and 3B). The data suggest that when a source of HMP is provided (addition of the ScTHI5 gene), C. glabrata is able to utilize this source of HMP to generate more thiamine and allow for better growth. Additionally, a higher level of expression of ScTHI5 through the CgPGK1 promoter allows for even better growth in the absence of thiamine. This growth is not present when the cells are also starved of PLP, indicating that C. glabrata is actually synthesizing HMP using PLP as a precursor.
To confirm C. glabrata was utilizing HMP as a substrate for growth, we incubated C. glabrata with another source of HMP-Scthi20ΔScthi21Δ cells. An overshoot phenomenon has been demonstrated in S. cerevisiae when too much HMP is present, resulting in an efflux of HMP from the cell. It has also been shown in S. cerevisiae that HMP can be brought into the cell when present in the environment [17]. We hypothesized that co-culturing C. glabrata with cells that are unable to make thiamine, but accumulate HMP would allow C. glabrata to grow more vigorously during thiamine starvation because C. glabrata could transport HMP into the cell and convert it into thiamine. As shown in Fig 3C, co-culturing C. glabrata wild-type with Scthi20ΔScthi21Δ cells allows for a large increase in cell density. We also deleted the thiamine transporter, THI10, to determine if this gene had a role in C. glabrata obtaining HMP from S. cerevisiae. When co-cultured with Scthi20ΔScthi21Δ cells, there was little growth for Cgthi10Δ, indicating that THI10 may be involved in bringing environmental HMP into C. glabrata cells, allowing for additional thiamine synthesis. Finally, we confirmed that the Scthi20ΔScthi21Δ cells were contributing a metabolite related to thiamine (likely HMP) because we observed very C. glabrata exhibits a growth defect relative to S. cerevisiae during thiamine starvation and utilizes the same regulatory factors. Growth during thiamine starvation of S. cerevisiae and C. glabrata wild-type strains as well as deletions of the putative regulators of the pathway: Scthi2Δ, Scthi3Δ, Cgthi3Δ, and Cgpdc2Δ. It is worth noting that in S. cerevisiae, PDC2 is required for growth in high glucose conditions, and was not included in this analysis. Cell density (OD 600 ) was measured at 12 h intervals during thiamine starvation for 48 h, and cells were diluted to remain in logarithmic growth. Error bars represent standard deviation of three experimental replicates, and this is true of all of the figures. S. cerevisiae wild-type grew to OD 600 = 314.3±8.1 in 48 h of thiamine starvation. In high thiamine conditions, all six strains grow vigorously (OD 600 > 300 in 48 h).  Enabling C. glabrata to make or use HMP allows for synthesis of thiamine and growth in starvation conditions. (A) Adding back THI5 in the genome or on a plasmid restores growth to C. glabrata in thiamine starvation. Strains were grown in replete or thiamine-and PLP-starved conditions on solid agar medium. (B) The same strains as in (A) were also grown in liquid medium and cell density (OD 600 ) was measured after 24 h. (C) Co-culture of Scthi20ΔScthi21Δ strain with C. glabrata allows for vigorous growth, suggesting that S. cerevisiae supplies HMP to C. glabrata and this uptake of HMP is decreased in a Cgthi10Δ. Co-culture was required because in our hands HMP was not stable enough for conditioned medium little growth when Scthi20ΔScthi21Δ cells were co-cultured with a Cgthi3Δ, Cgpdc2Δ, or a Cgthi20Δ strain (data not shown).

HPLC analysis of thiamine vitamers in both species
We wanted to discover if the two species utilize or store thiamine in different ways, which might uncover different thiamine utilization strategies. We measured the concentration of thiamine, TMP, and TPP in both species when grown in high and no environmental thiamine conditions, using an HPLC assay. We first verified that this assay was capable of detecting the three vitamers in cell extracts (S3 Fig) and was linear for measuring thiamine vitamer concentrations (S4 Fig). We then measured the vitamer concentration on a per biomass basis (estimated by OD 600 ) for both species. S. cerevisiae and C. glabrata store approximately the same amount of thiamine and TPP during growth in high thiamine (Fig 4A and 4B). Because cells are grown at a relatively high density, we are also able to observe that concentrations of vitamers decrease over growth time, which is correlated with the diauxic shift (shifting from glucose fermentation to respiration). This is suggestive that fermentation requires more TPP than oxidative phosphorylation, but additional experiments are required to validate this hypothesis. Additionally, we observed that both species shift their total vitamer concentrations to TPP during thiamine starvation which is not surprising given that TPP is the only metabolically active vitamer (Fig 4C and 4D). Finally, we observed that cells do not maintain reserves of TMP suggesting that any TMP that is synthesized is rapidly dephosphorylated and pyrophosphorylated or simply dephosphorylated.
C. glabrata has compensated for the loss of a key transcriptional regulator of thiamine production relative to S. cerevisiae Because C. glabrata is lacking a THI2 ortholog, and a Scthi2Δ strain is a thiamine auxotroph, we wanted to understand how C. glabrata maintains any thiamine biosynthesis independent of THI2. We screened a C. glabrata genomic library [34] for clones that were able to suppress the thiamine auxotrophy of a Scthi2Δ strain, and recovered five plasmids all containing CgTHI3. No other plasmids reproducibly suppressed the Scthi2Δ auxotrophy as assayed by colony formation on solid medium lacking thiamine (data not shown). This raised the question of whether there was a function specific to CgTHI3 or if simply expressing THI3 from either species on a plasmid (and increasing the copy number) was capable of suppressing the Scthi2Δ thiamine auxotrophy. To differentiate between these two possibilities we cloned ScTHI2, ScTHI3, and CgTHI3 (promoter and ORF) into a HIS3 plasmid and determined the ability of each plasmid to suppress the thiamine auxotrophy of the thi2Δ and thi3Δ strains in both species ( Fig  5A). All of the plasmids were functional, as they were able to complement deletion of the appropriate gene. Furthermore, both ScTHI3 and CgTHI3 were capable of partially suppressing the thiamine auxotrophy of the Scthi2Δ strain (p = 0.003 and p = 0.03 respectively by Student's t-test), indicating that simple overexpression of THI3 is sufficient to suppress the loss of the THI2 and there is no apparent neofunctionalization of CgTHI3 relative to ScTHI3.
These cross-complementation studies indicated that we should be able to overexpress ScTHI3 in the genome of S. cerevisiae and partially suppress the thiamine auxotrophy of the Scthi2Δ strain. To test this, we engineered in the genome a high level promoter (ScPGK1) driving the expression of ScTHI3, which has homology to pyruvate decarboxylase genes but does not appear to have pyruvate decarboxylase activity, and two related homologs, ScPDC1 and ScPDC5 (49% and 50% identical to ScTHI3, respectively), not thought to be involved in the transcription of THI genes (Fig 5B) [16]. Overexpression of ScTHI3 led to increased growth of the Scthi2Δ strain in thiamine starvation and the two non-transcriptional regulators appeared to only minimally stimulate growth in the absence of thiamine. These data suggest that high levels of the Thi3 transcriptional regulator specifically can confer the ability to grow better in thiamine starvation conditions in the absence of THI2. Additionally, these data suggest that C. glabrata behaves similar to a S. cerevisiae strain that is deleted for THI2 but over expresses THI3.
C. glabrata induces a different subset of THI genes to optimize uptake and minimize thiamine biosynthesis relative to S. cerevisiae Because C. glabrata appears to be capable of utilizing an environmental source of HMP for thiamine biosynthesis (Fig 3), we hypothesized that the transcriptional derepression of the THI  genes may have different characteristics between the two species. For example, C. glabrata might induce scavenging genes to a higher level than synthesis genes relative to S. cerevisiae. To test this, we performed rt-qPCR on candidate THI genes. We chose genes that are involved in the synthesis of the HET precursor (THI4), HMP phosphorylation (THI20), and transport of thiamine into the cell (THI10). The rt-qPCR analyses suggest that S. cerevisiae synthesizes thiamine de novo when thiamine is not present by expressing THI4 at high levels while inducing THI10 expression much less (Fig 6). C. glabrata, on the other hand, may be obtaining thiamine directly from the environment and does so by expressing THI10 to high levels in an effort to uptake thiamine and HMP. A caveat to this analysis is that all of the genes are upregulated during thiamine starvation and post transcriptional mechanisms are not being considered.
Growth characteristics of S. cerevisiae and C. glabrata are altered in replete and thiamine-starved conditions To determine the consequences of having a partial decay of the THI biosynthetic and regulatory pathway in C. glabrata relative to S. cerevisiae, we utilized co-culture conditions of both species where one species expressed a yellow fluorescent protein (YFP). This experimental setup allowed for real-time measurement of cell numbers of each species by flow cytometry. We measured cell growth of wild-type C. glabrata and S.cerevisiae during exponential growth, and regardless of auxotrophic markers or presence of YFP, determined that C. glabrata outcompetes S. cerevisiae in medium containing thiamine (Fig 7A). This was not surprising as the Fig 6. C. glabrata induces a different subset of THI genes to optimize uptake and minimize thiamine biosynthesis relative to S. cerevisiae. rt-qPCR analysis of THI4, THI20, and THI10 in both S. cerevisiae and C. glabrata wild-type strains grown in thiamine depleted conditions. All expression is relative to the ACT1 gene of the respective strain.  C. glabrata out competes S. cerevisiae in the presence of thiamine but not in the absence of thiamine. (A) Using flow cytometry, the percent of cells that was either S. cerevisiae or C. glabrata was determined and plotted based on the time of the measurement. Cells were maintained in a logarithmic growth phase in SD medium containing thiamine for the entirety of the assay. The data presented are from nine separate competition assays with three different sets of strains. Because the CgADH1p-YFP was integrated into the CgURA3 locus, we confirmed that the presence or absence of uracil and the URA3 gene did not affect the results. The S. cerevisiae strain was from [36]. (B) The same assay as part (A) was performed but in the absence of thiamine. doubling time of C. glabrata (4.06 h ± 0.06) is faster that S. cerevisiae (4.44 h ± 0.23) in synthetic medium when grown individually. Within three days, over 95% of the cells in the co-culture are C. glabrata.
During growth in medium lacking thiamine, we observed more complex growth characteristics, but they are consistent with C. glabrata not responding to thiamine starvation as efficiently as S. cerevisiae (Fig 7B). We observe consistent short-term advantages of C. glabrata every time cultures are diluted. We do not know why this is the case, but could be a consequence of C. glabrata being able to grow better in fresh medium and grow more rapidly than S. cerevisiae for a short period. In the long-term though, C. glabrata is unable to synthesize enough thiamine for robust growth and so S. cerevisiae slowly dominates the co-culture. Interestingly, we observe a phenomenon similar to frequency-dependent selection in co-culture in both high and no thiamine conditions; the less fit species is still maintained (between 1-5%) after 10 days of culturing, even though the doubling times of individual cultures suggest that one species should completely outcompete the other. We believe that this indicates that in both conditions, there are trade-offs with regards to growth between the two species and potentially compounds can be transferred between species to allow both species to grow more vigorously.

Discussion
We demonstrate a partial decay of the THI biosynthetic pathway in C. glabrata that is a derived trait in the C. glabrata clade relative to closely related species [22]. Comparison of the THI pathways of S. cerevisiae and C. glabrata have allowed for a number of observations that suggest the two species thrive in different niches. S. cerevisiae grows on decaying plant material and C. glabrata grows on mammalian mucosa [37,38]. Of the 139 orthologs absent in C. glabrata relative to S. cerevisiae, at least nine are involved in PLP (vitamin B 6 ) and thiamine (vitamin B 1 ) biosynthesis [31]. PLP participates in core metabolism independent of thiamine metabolism, and thus, both vitamins are required for proper growth. Our results suggest that mammalian tissue is supplying PLP and HMP/thiamine, and that S. cerevisiae may not derive those compounds as readily from decomposing plant material. It is worth highlighting that it is not clear what the "natural" environment of the two species is [28,39], but given that C. glabrata is isolated from human GI tract and S. cerevisiae is often isolated from rotting plant material, our assumption is not unreasonable.
Whereas the HMP part of the thiamine biosynthetic pathway has decayed in C. glabrata, it is noteworthy that the transcriptional wiring has also changed between the two species. A priori there is no apparent need to lose THI2 just from the glabrata clade of species, given that all of the other recent common ancestors contain THI2. However, this event is similar to the loss of the absolute requirement for CgPHO2 in the phosphate signal transduction pathway [40], suggesting that C. glabrata (and related species) has simplified its transcriptional response in at least two signal transduction pathways. Further work is required to determine whether S. cerevisiae has modified the core THI pathway for the purposes of generating multiple transcriptional outputs or if C. glabrata has streamlined its genome. S. cerevisiae has amplified the ScTHI5 family relative to most species possibly indicating that the amount of thiamine required by S. cerevisiae is larger in its environment than C. glabrata. Interestingly, Nosaka et al. has demonstrated that some THI promoters in S. cerevisiae (ScTHI2 and ScTHI10) are not dependent on ScTHI2, suggesting that ScThi2p might be an accessory transcription factor that allows for an increased level of expression of the THI transcripts [4,41]. The difference in specificity of the Pdc2 transcription factor also underlies the transcriptional rewiring; ScPdc2 is critical for the regulation of pyruvate decarboxylase genes and this is not true of CgPdc2 [16].
Our measurements of thiamine vitamer concentrations by HPLC also indicate that the amount of thiamine per biomass is varied depending on condition-i.e. during periods of slower growth there is apparently less need for TPP per cell. This may indicate that S. cerevisiae experiences times in its environment where the ability to synthesize a large amount of thiamine is beneficial. For example, during fermentation it is likely that there is large flux of glucose through the TPP requiring enzyme pyruvate decarboxylase, whereas during slower growing conditions (potentially growing on mucosal membranes) there is less TPP needed for growth as respiration is much more efficient at energy production. The expansion of the SNO/SNZ/ THI5 gene families in the Saccharomyces clade is consistent with this argument. It is worth noting that pathogenic fungi, including C. albicans, have an apparently fully functioning THI pathway, indicating that the partial loss of the THI pathway that we observe in C. glabrata is not a requirement for pathogenicity. Likewise, closely related species to C. glabrata that do not appear to be commensal with mammals have similar THI pathways to C. glabrata-i.e. the same losses of genes (S1 Fig and [22]). This loss of genes may be a function of subtly different growth niches that remain to be understood more fully. Our work is reminiscent of work with biotin biosynthesis in yeast species, where some Saccharomyces species have lost the most upstream parts of the biosynthetic pathway [42]. Interestingly, in that example, selection pressure appears to have led to horizontal gene transfer from bacteria of the upstream genes.
Finally, the characterization of the THI pathway in C. glabrata has uncovered potential future avenues of antifungal research. As current pharmacological agents in the chloroquine family have been demonstrated to target thiamine uptake in yeast and mammalian cells [43], targeting thiamine biosynthesis and uptake may be a good drug target for treatment of C. glabrata infections.

Yeast strains
Yeast strains used in this study are listed in Table 1. Primers used to make these strains are listed in S1 Table. Deletion mutants were generated using antibiotic resistance genes KANMX6 or NATMX6 and homologous recombination to precisely delete ORFs [40,44]. Deletion was confirmed by PCR. Genes were incorporated into the genome by homologous recombination to precisely replace URA3, selected with 5-FOA, and confirmed by PCR.
To construct plasmids containing genes, the promoter and the ORF of the gene were amplified (primers in S1 Table) and cloned by homologous recombination into a pRS313 (HIS3) vector [46].

Growth conditions
Yeast strains were grown in YEPD medium or synthetic dextrose (SD) medium with complete supplement mixture (CSM), either with or without histidine (Sunrise Science Products; San Diego). For induction of THI promoters and genes, strains were grown at 30°to logarithmic growth phase (OD 600~0 .2-0.5) in SD medium with thiamine. Cells were harvested by centrifugation, washed three times with medium lacking thiamine, and transferred to SD medium with thiamine (high thiamine), without thiamine (no thiamine), or without PLP and thiamine (No PLP/thiamine) at a starting OD 600 of 0.05 in 5 mL. Strains were grown in these conditions from 24 to 72 h to measure cell density (OD 600 ). During growth, cells were diluted to remain in logarithmic growth. For quantitative reverse-transcription PCR analysis, strains were grown for 8 h in these conditions.

Bioinformatic Analysis of the THI pathway in C. glabrata
Orthologs of the S. cerevisiae THI genes were identified in yeast species using the Fungal Orthogroups, PhylomeDB, and YGOB websites [29,30,47] as well as the Saccharomyces Genome Database [48] and the Candida Genome Database [49]. For generation of S1 Fig,  orthogroups were identified from reference [29,30] and a variety of pre-and post-genome duplication species were chosen to identify trends in the number of copies of genes in each family.

Growth in co-culture
After washing the cells three times in medium lacking thiamine, the volume of each culture needed to inoculate at a starting OD 600 of 0.02 for S. cerevisiae and 0.01 for C. glabrata in 5 mL was calculated. This difference in OD 600 is derived from a comparison of OD 600 to cell count of the two species by flow cytometry (data not shown), which found that there is twice the number of C. glabrata cells per O.D. as compared to S. cerevisiae cells; C. glabrata cells are smaller than S. cerevisiae.

HPLC analysis
The methods to analyze the vitamer concentrations are adapted largely from [50]. Cultures were inoculated at OD 600 = 0.1 at t = 0 and grew to an OD 600 > 3 after 24 h. The vitamers were extracted from the yeast cells (each sample = 5 OD 600 ) through bead beating in 0.1 M HClO 4 , adapted from methods in [51]. Samples for the calibration curve were dissolved in 0. 1 6 ] to allow for fluorescence detection of the vitamers at 366 nm excitation and 436 nm emission. Before subjecting samples to HPLC, samples were filtered with a 0.2 μm syringe filter and 10 μL was injected on a RP-amide C16 column (Supelco) with 80% 0.4 M phosphate/20% methanol as the isocratic mobile phase at 1 mL/min for a 10 min sample analysis time [50]. Calibration curves were analyzed each day at the beginning and end of runs to account for any changes in instrument sensitivity. Approximate retention times are thiamine at 7.3 min, TMP at 4.2 min, and TPP at 3.1 min. To calculate concentration of the unknown samples, the integration area of the samples was converted into pmol of vitamer based on appropriate calibration curves and then normalized to OD 600 .

Identification of C. glabrata THI2 analog
To screen for the C. glabrata THI2 analog, a C. glabrata genomic DNA library [34] was transformed into a Scthi2Δ strain. Approximately 16,500 transformants were plated and then replica plated onto SD plates lacking uracil and thiamine to identify clones that were able to suppress the thiamine auxotrophy of the Scthi2Δ strain. Five plasmids were recovered using a yeast plasmid DNA purification protocol and transformation into Escherichia coli cells. The plasmids were then sequenced and all 5 contained DNA near the CgTHI3 gene [24].
Quantitative reverse-transcription PCR (rt-qPCR) RNA was extracted using a standard phenol-chloroform protocol [52] and converted to cDNA by a reverse-transcription reaction (Bio-Rad iScript cDNA synthesis kit). Quantitative PCR was performed with a CFX qPCR machine (Bio-Rad) using SybrGreen I dye in a 50-μL reaction. Primers were designed for S. cerevisiae and C. glabrata THI genes THI4, THI20, THI10, and ACT1 (primers in S1 Table). The amount of transcript for each gene was normalized to ACT1, which does not change expression in differing thiamine conditions. Each gene was also amplified using 10-fold genomic DNA dilutions as an amplification control.

Co-culture competition assay
Each of the strains were grown overnight in nutrient rich medium and inoculated at starting OD 600 of 0.002 for S. cerevisiae strains and 0.001 for C. glabrata strains in 5mL of media in triplicate in either high or no thiamine media. The mixtures were subjected to flow cytometry (Accuri C6, BD biosciences) at time-points 0, 4, 8, 12, and 24 h after the initial pipetting. Every 24 h the mixture was back-diluted into 5 mL of medium to an OD 600 of 0.003. Flow cytometry allowed for the absolute number of cells to be counted as well as species based on which species contained YFP. We confirmed that the ADH1pr-YFP constructs did not alter the viability of the strains by competing them with the appropriate parental strain for 100 h.
Supporting Information S1 Fig. Presence, absence, and number of homologs of each gene in the THI pathway in a diversity of Ascomycetes. Using YGOB, PhylomeDB, and Fungal Orthogroups [29][30][31], we identified likely orthologs and paralogs of the S. cerevisiae genes involved in thiamine biosynthesis, uptake, and regulation of the biosynthetic pathway. We have highlighted in red genes where the number is in conflict through the different analyses. The species are presented as a consensus phylogeny combining the work of [22,29]. Shading of the table is based on the number of homologs to highlight where we believe there has been gain/loss of genes in the THI pathways of specific clades.