A Novel cis Element Achieves the Same Solution as an Ancestral cis Element During Thiamine Starvation in Candida glabrata

Regulatory networks often converge on very similar cis sequences to drive transcriptional programs due to constraints on what transcription factors are present. To determine the role of constraint loss on cis element evolution, we examined the recent appearance of a thiamine starvation regulated promoter in Candida glabrata. This species lacks the ancestral transcription factor Thi2, but still has the transcription factor Pdc2, which regulates thiamine starvation genes, allowing us to determine the effect of constraint change on a new promoter. We identified two different cis elements in C. glabrata - one present in the evolutionarily recent gene called CgPMU3, and the other element present in the other thiamine (THI) regulated genes. Reciprocal swaps of the cis elements and incorporation of the S. cerevisiae Thi2 transcription factor-binding site into these promoters demonstrate that the two elements are functionally different from one another. Thus, this loss of an imposed constraint on promoter function has generated a novel cis sequence, suggesting that loss of trans constraints can generate a non-convergent pathway with the same output.

We previously identified that C. glabrata, unlike most yeast species, is auxotrophic for thiamine because of a partial loss of the biosynthetic pathway, but it still upregulates 5 genes involved in biosynthesis and scavenging (CgTHI4,CgTHI20,CgTHI10,CgPET18,and CgPMU3) .50 fold in response to thiamine starvation (Iosue et al. 2016;Nahas et al. 2018). Similar to S. cerevisiae, this upregulation is dependent on the DNA binding protein, Pdc2 and its regulator, Thi3 (Iosue et al. 2016;Nahas et al. 2018). However, C. glabrata lost the transcription factor Thi2, which is necessary for the thiamine starvation response in S. cerevisiae and in the ancestor of these yeast species, suggesting that there is some rewiring of how thiamine responsive genes are regulated (Gabaldón Estevan et al. 2013;Huerta-Cepas et al. 2014) (Figure 1).
C. glabrata has also recently acquired a novel phosphatase gene (CgPMU3) regulated by thiamine starvation (Orkwis et al. 2010;Nahas et al. 2018). CgPMU3 is up regulated .50-fold in response to thiamine starvation and is essential for accessing external thiamine when it is pyrophosphorylated. Interestingly, CgPMU3 appears to have replaced the more common PHO3-related phosphatases observed in other related species (Nahas et al. 2018) (Figure 1). Because pyrophosphatase activity provides a selective advantage by allowing cells to access phosphorylated forms of external thiamine, we were able to ask the question of whether a new gene becomes integrated into a regulatory pathway in a completely novel way, or are the existing trans components used, but with modifications. We observed the unexpected result of a new promoter, regulated by thiamine starvation, acquiring multiple novel characteristics relative to the promoters that have been present over a long evolutionary time. However, the promoter behaves very similar to the ancestral promoters in terms of output and uses some of the same ancestral trans factors. Thus, the experiment of "replaying life's tape" (Gould 1990;Blount et al. 2018) by looking at a new promoter under selection suggests parallel yet very different changes and interestingly, the generation of a new DNA binding element.
In addition to observing a novel cis element, this work is motivated by defining the requirements for the CgPMU3 promoter. C. glabrata resides predominantly in mammalian gastrointestinal tracts, and is the second most common cause of candidiasis Pappas et al. 2018;Kumar et al. 2019). C. glabrata is often more resistant to anti-fungal drugs relative to C. albicans and thus the development of targeted therapies would be beneficial (Whaley and Rogers 2016). Because human serum transports thiamine primarily in the form of TPP, an understanding of CgPMU3 upregulation is critical for long-term studies about the pathogenicity of this species (Lu and Frank 2008), and targeting CgPMU3 expression is a potential avenue for antifungal development.
Here, we used truncation analysis to identify regions of THI promoters required for upregulation of expression during thiamine starvation. We identified an 11 base pair (bp) region that is essential for upregulation in CgPMU3 but surprisingly, regions similar to this in other THI promoters were not required for upregulation. Using almost base pair resolution, we identified a different 13 bp region in other THI promoters that does not share obvious similarity to the 11 bp region in CgPMU3, and these 11 bp and 13 bp regions are not interchangeable. The difference between CgPMU3 and other THI promoters is that CgPMU3 likely never evolved thiamine regulation in the presence of Thi2, unlike the other THI promoters. We examined expression of C. glabrata THI promoters in S. cerevisiae, and noted they are not regulated, but the inclusion of a putative Thi2 binding site restored upregulation of CgTHI promoters (and CgPMU3) in S. cerevisiae. However, this upregulation differed between the ancestral THI promoters and the CgPMU3 promoter. We conclude that loss of Thi2 and selection for a newly regulated gene confers a different path across the evolutionary landscape than the THI genes that are conserved within the Ascomycete lineage.

Strains
Most of the experiments were performed in C. glabrata wild-type (Cormack and Falkow 1999) and S. cerevisiae wild-type (Wykoff and O'Shea 2001) strains. Additional strains used in this study were deletions of the thiamine pathway regulators: Cgthi3Δ (DG141), and Cgpdc2Δ (DG271), Scthi2Δ (DC126), Scthi3Δ (DC143) (Iosue et al. 2016) and Scpdc2Δ, which was generated in this this study. Because PDC2 is essential in glucose-containing medium in S. cerevisiae, NATMX6 was amplified using PCR (primers in Supplemental Material, Table S1) and transformed into a diploid strain to delete ScPDC2. We covered this deletion with a URA3 + plasmid (pRS316) containing ScPDC2. Through random sporulation, we identified haploid colonies that were Scpdc2D. To construct a Scpdc2Δ strain capable of growth in glucose medium, ScPDC1 was overexpressed in this strain: ScPDC1 Figure 1 Phylogenetic relationships and presence or absence of thiamine signal transduction pathway transcription factors and thiamine pyrophosphatases (TPPases). Using a phylogeny of yeast (Gabaldón and Carreté 2016;He et al. 2017), presence or absence of genes was determined previously (Wapinski et al. 2007;Huerta-Cepas et al. 2014). C. glabrata lacks THI2 as do the other "glabrata group" yeast (not presented in figure), but only C. glabrata contains the PMU array of genes (Gabaldón Estevan et al. 2013). We believe that the S. cerevisiae THI pathway behaves similar to the ancestral pathway, and C. glabrata has lost Thi2, gained PMU3, and is unable to synthesize thiamine de novo. WGD (and the star) refers to the whole genome duplication event, and CTG clade refers to the altered codon usage of C. albicans. was amplified by PCR and cloned by homologous recombination (Corrigan et al. 2013) into a pdc2Δ strain on a LEU2+ plasmid (pRS315) under the control of the ScADH1 promoter. This strain was then grown on SD (synthetic dextrose, Sunrise Science, CA) plates with 5-FOA to select against the URA3 + plasmid containing ScPDC2. For sequencing of the CgPMU3 promoter in the SEL-seq experiment, the entire PMU gene family (PMU1, PMU2, and PMU3 promoter and open reading frame) was deleted with NATMX6 in a C. glabrata wild-type strain (Table S1).

Plasmid Construction
To assay induction of THI pathway genes, we constructed plasmids where either the full-length promoters (1000 bp) or smaller regions of the promoters of these genes were driving expression of yellow fluorescent protein (YFP). The promoters were amplified by PCR (Table S1) and cloned by homologous recombination into a HIS3+ plasmid (pRS313) containing YFP in a wild-type strain (Corrigan et al. 2013). To investigate the effects of mutations/deletions in the promoters, PCR was used to amplify the full-length promoter in two regions, with overlapping primers that incorporated the altered sequence (Table S1), and these PCR products were cloned into a YFP plasmid as previously described (Corrigan et al. 2013;Nahas et al. 2018). For some promoters, a PacI restriction enzyme site replaced the UAS so the opposite UAS could be easily introduced. Details of cloning are available upon request.

Flow cytometry
To measure induction of the THI pathway genes, fluorescence of cells containing plasmids with promoters driving YFP was quantified by flow cytometry. Cells were grown at 30°in thiamine replete SD medium lacking histidine (Sunrise Science, CA) to logarithmic growth phase (OD 600 0.2-0.5). Cells were harvested by centrifugation, washed 3 times with sterile water, inoculated into thiamine replete (0.4 mg/L) and starvation (no thiamine added) conditions in SD medium lacking histidine, and grown at 30°overnight (18 h). Mean fluorescence (in arbitrary units, a.u.) of each strain was measured using a fiow cytometer with a 533/30 FL1 filter set (Accuri C6, BD Biosciences). In almost all cases, background fluorescence was less than 12,000 a.u.; however, there is variability of fluorescence based on precise growth conditions and we included positive and negative controls in each experiment.

SEL-seq sequencing
To perform the SEL-seq experiment with the CgPMU3 11 bp element, we constructed a plasmid that contained the ScTHI5 ORF in frame with YFP and the CgPMU3 promoter. To make the promoter, we used PCR to generate a 250 bp product that incorporated Ns in the 11 nucleotide region with 30 bp of homology to a 750 bp PCR product corresponding to the rest of the CgPMU3 promoter (from -1000 bp to -250 bp). The three PCR productstwo CgPMU3 promoter PCR products and the ScTHI5 open reading frame (Table S1)were gap repaired (Corrigan et al. 2013) into a strain lacking the wild-type CgPMU3 promoter (Cgpmu1-3ΔNATMX6 described above) and we collected 131,000 independent transformants. Approximately, 5% of transformants were judged as highly expressing during thiamine starvation (based on YFP expression). We pooled the transformants and took a time zero sample for deep sequencing of the CgPMU3 promoter. Based on sampling of unique sequences, we generated 90,000 unique sequences to query. We then grew the cultures in SD medium lacking thiamine and histidine for three successive days with 1:1000 dilution every 24 h (allowing 20 generations to pass). We monitored fluorescence by flow cytometry and observed the frequency of cells that were highly fluorescent jump from 6 to .90% in 24 h. We collected three independently grown cultures (in medium lacking thiamine) to purify DNA and amplify the CgPMU3 promoter for next generation sequencing on a MiSeq (Illumina, San Diego, CA). Sequences were extracted in Geneious, and at least 2x1750 sequences were analyzed for each sample. We sorted the sequences, identified the number of unique sequences, and quantified the percent representation of the sequence in the total sequences. We verified that extraction of a different subset did not alter the resultsi.e., the same sequences were repeatedly identified as enriched.

Data availability
All strains, plasmids, and raw data are available upon request. Table S1 lists the primers used in this study to generate strains and plasmids. Figure S1 demonstrates that ScTHI5 confers a growth advantage to C. glabrata during thiamine starvation. Figure S2 shows the frequency of abundant sequences after selection in thiamine starvation in the SEL-seq experiment. Table S2 shows the raw data from the seven samples sequenced in the SEL-seq experiment. Table S3 lists the sequences that were highly enriched after selection in thiamine starvation. Figure S3 aligns the sequences in Table S3 with C. glabrata THI promoters. Figure S4 shows a scanning mutagenesis of the 13 bp THI UAS in the CgPET18 promoter. Figure S5 is a schematic of S. cerevisiae promoters with the locations of binding sites as well as mutations and deletions made in this study. Figure S6 demonstrates that S. cerevisiae THI promoters are dependent on Pdc2, Thi2, and Thi3. Supplemental material available at figshare: https://doi.org/ 10.25387/g3.10308194.

RESULTS
The CgPMU3 promoter contains an 11 bp element required for thiamine starvation upregulation To understand the DNA sequences required for upregulation by thiamine starvation, we undertook promoter truncation experiments with portions of the CgPMU3 promoter fused to the open reading frame of yellow fluorescent protein (YFP). Induction of the promoter was quantified using flow cytometry to measure the fluorescence of YFP in the cells. First, we truncated in 100 bp increments and then in 20 bp increments from -1000 bp (referring to the location upstream of the start codon) to the start codon (data not shown). We narrowed the beginning of the upstream activating sequence (UAS) to between -260 bp and -240 bp. Performing a MEME motif discovery analysis (Bailey et al. 2009), we identified an 11 bp region that appeared to be somewhat conserved in other THI promoters ( Figure 2A). To determine whether this region was important for upregulation, we further truncated the CgPMU3 promoter and made point mutations in the 11 bp region in the context of the full-length 1000 bp promoter ( Figure 2B). These data indicate that numerous nucleotides in the 59 GACGTA-CAACG 39 sequence are critical for high-level de-repression of the CgPMU3 promoter.
To identify the importance of each nucleotide in the 11 bp sequence, we mutated individual nucleotides in the context of the 1000 bp promoter to either a T or a G ( Figure 2C). When a T or G was present in the original sequence, we mutated it to either an A or C, respectively. The trends between the two scanning mutagenesis experiments were similar, and they identified the bases in uppercase as being critically important for thiamine regulation -gACGTacaacG. However, it is clear that other nucleotides have importance, as mutation of two As (that are next to one another) to Cs also disrupts the function ( Figure 2B). While there is variable conservation of these nucleotides in other THI promoters based on the MEME analysis, using only the essential nucleotides to search the C. glabrata genome identifies too many sequences to be informative.

The 11 bp UAS in CgPMU3 is not important for other THI promoters
To determine whether the element identified in the CgPMU3 promoter was important for upregulation in other THI promoters, we deleted the 11 bp element in CgPET18 and CgTHI20. We chose these two promoters because they shared the most sequence similarity to the CgPMU3 UAS (with 9 nucleotides identical out of 11, Figure 2A). We did not find a major defect in transcriptional induction of these genes when the putative PMU3 UAS was deleted (Figure 3). This was remarkable, given the common sequence with the CgPMU3 element. However, neither CgPET18 nor CgTHI20 have the strict xACGTx 5 G motif. We hypothesize that we have identified this motif for one of two reasons. Either 1) this motif has appeared by chance in the promoters, as the sequences are imperfect matches with the CgPMU3 UAS, or 2) this motif is present, but does not have a critical role in thiamine starvation regulation by our assay. Regardless, this suggests that CgPMU3 appears to have a different UAS requirement from the other THI genes. It is worth noting that CgPMU3 is a recent duplicate of a phosphatase gene and only acquired thiamine regulation in C. glabrata Nahas et al. 2018), whereas the other THI genes are present in multiple Ascomycota species and have likely been regulated by the same THI pathway through multiple speciation events.
SEL-seq approach to identifying the critical nucleotides in the CgPMU3 promoter Because the CgPMU3 UAS did not appear important in other THI promoters, we wanted to take a relatively unbiased approach to understand Figure 2 The CgPMU3 promoter contains an 11 bp UAS necessary for thiamine starvation dependent expression. A) After truncation analysis of the CgPMU3 promoter, a MEME analysis identified a region that appeared conserved in THI promoters (1000 bp of each THI promoter and 270 bp of the CgPMU3 promoter). Searching the C. glabrata genome for a consensus GACRNANNACG using a pattern match algorithm (Skrzypek et al. 2017), yielded 116 genes with this element in the 1 kb upstream of the start codon, including CgPMU3, but no other known THI regulated genes. The gray shading indicates nucleotides in common with CgPMU3. The number after the promoter name indicates the nucleotide (under the arrow) upstream from the start codon. B) Characterization of the 11 bp CgPMU3 UAS. The first four samples show truncation analysis and the next five samples have mutations introduced into the full-length (1000 bp) wild-type promoter. Promoter induction was assayed during thiamine starvation by measuring the fluorescence of cells containing plasmids with these promoters driving YFP. C) Scanning mutagenesis of the 11 bp CgPMU3 UAS. Single mutations were introduced into the full-length promoter, replacing the native nucleotide with either a T or a G, except when the native nucleotide was a T/G, in which case the T/G was replaced with an A/C. For this and the following figures, the data presented is the mean and standard deviation of at least three independently grown samples.
what nucleotides were important for upregulation in CgPMU3 and to determine if there were sequences that conferred regulation that might have similarity to cis elements in other THI promoters. We hypothesized that multiple versions of the CgPMU3 UAS would confer thiamine starvation regulation in this 11 bp elementi.e., there is some degeneracy in the sequence, and that potentially alterations in the CgPMU3 UAS might cause it to resemble elements in other THI promoters. To identify the important nucleotides in the 11 bp UAS using an unbiased approach, we performed a modified SEL-seq experiment (Farley et al. 2015). Using a selection with theoretically 4.2 million (4 11 ) possibilities being queried for high-level expression during thiamine starvation, we replaced the 11 bp UAS with all four nucleotides in each position (incorporated into a primer - Table S1) and selected for high-level expression during thiamine starvation. Because C. glabrata is auxotrophic for thiamine and addition of ScTHI5 restores prototrophy (Iosue et al. 2016), we could select for high-level expression of the CgPMU3 promoter by having it control ScTHI5 transcription during thiamine starvation. We confirmed that this CgPMU3p-ScTHI5 plasmid was capable of supporting growth of C. glabrata in the absence of thiamine ( Figure S1).
Using a fusion PCR method, we gap repaired the CgPMU3 promoter upstream of the ScTHI5 ORF, replacing the CgPMU3 UAS with all four nucleotides, allowing multiple sequence options to replace the UAS. The selection was successful but limited in terms of exploring the 4 million possibilities. We obtained 10 5 transformants, and observed 90,000 unique sequences with the 20 most abundant sequences representing 9.7% in our sampling sequencing prior to selection ( Figure S2 and Table S2). It is likely that PCR and primer synthesis introduced biases that led to a few abundant sequences, and a diversity of other sequences. However, after selection, the 20 most abundant sequences (which were different from the preselection sequences) represented on average 72.5% of the total sequences, indicating that some sequences conferred a strong selective advantage. To begin to eliminate sequences that might simply confer a high level expression independent of the THI pathway, we performed a parallel selection in a Cgpdc2D strain, expecting that if a sequence was abundant in the Cgpdc2D strain, that it was a sequence that allowed for higher-level expression of the ScTHI5 construct independent of the THI pathway ( Figure S2). Eight sequences were .90 fold enriched in a THI pathway dependent manner, and all contained a 59-CTG-39 motif (Table S3). Interestingly, two sequences that we identified were a 10/11 bp and 8/11 bp match for a sequence element in the CgPET18 promoter, different from the PMU3 UAS in Figure 2A, indicating that the other THI promoters contain a sequence that might functionally replace the CgPMU3 UAS (sequences A and G in Figure S3).
A 13 bp THI UAS in non-CgPMU3 THI promoters is important for thiamine starvation regulation To determine regions that are important for thiamine regulation in THI promoters other than CgPMU3, we took a parallel approach to the CgPMU3 promoter, and truncated the CgPET18, CgTHI10, CgTHI4, and CgTHI20 promoters. We narrowed down the beginning of a regulatory sequence in these promoters to a few base pairs ( Figure 4A-D). After a MEME analysis, we identified a new DNA sequence that does not appear to be present in the CgPMU3 promoter ( Figure  4E), that closely correlates with the locations of where truncation begins to decrease thiamine starvation regulation (arrows in Figure 4F), and that overlaps with the SEL-seq CgPET18 sequence ( Figure S3). Only the CgPET18 promoter sequence contains a 59-CGT-39 motif that is critical for the CgPMU3 UAS element, perhaps explaining why we enriched for CgPET18 elements in the SEL-seq experiment.
To validate that these newly identified conserved sequences were important for thiamine starvation regulation, we replaced the 13 bp UAS with a PacI restriction enzyme site in the context of the 1000 bp promoter and determined whether this element is critical for upregulation of the CgPET18 and CgTHI10 promoters ( Figure 5). In both promoters, deleting the UAS decreased expression during thiamine starvation. We also performed a scanning mutagenesis of the CgPET18 full-length promoter, mutating these 13 bp individually to A (or C if an A was in that position), and determined that the xxCCGTxxAxxTG nucleotides were important for expression ( Figure S4). There is overlap between the CgPMU3 and THI UAS in terms of both possessing a 59-CGT-39; however, CGT is not absolutely required as CgTHI10 does not contain this sequence, and the remaining nucleotides are not easily aligned with the CgPMU3 UAS. Thus, we have identified two UAS elements that do not appear related to one another: the CgPMU3 UAS ( Figure 2) and the THI UAS present in all of the other THI promoters (Figure 4). We next determined whether the CgPMU3 UAS or the THI UAS were capable of substituting for one another in promoters. To do this, we deleted the critical element with a PacI restriction enzyme site and used this PacI site to incorporate the opposite element ( Figure 6). The THI UAS is partially capable of substituting for the CgPMU3 UAS ( Figure 6A), which is not surprising, given the SEL-seq data where a randomly selected sequence in the CgPMU3 promoter is very similar to the CgPET18 13 bp UAS. However, the CgPMU3 UAS is not capable of replacing the THI UAS ( Figure 6B), suggesting that this recently evolved 11 bp promoter element does not function identically to the 13 bp element.
The THI UAS is similar to sequences in S. cerevisiae promoters and is likely the ancestral UAS Because PET18, THI10 (TH17), THI4, and THI20 are present in the genomes of Saccharomycetaceae and regulated by thiamine starvation, Figure 3 Deletion of the 11 bp CgPMU3 UAS in THI promoters does not eliminate thiamine starvation dependent expression. The putative CgPMU3 UAS (Figure 2A) was precisely deleted in the full-length promoters of CgPMU3 (and replaced with a PacI restriction site), CgPET18, and CgTHI10 and assayed for YFP expression in high and no thiamine conditions. While necessary for CgPMU3, this UAS is not important for induction of other THI promoters.
we consider these genes and promoters to have been present in the common ancestor (Byrne and Wolfe 2005;Gabaldón Estevan et al. 2013;Huerta-Cepas et al. 2014). Conversely, CgPMU3 is a novel gene present only in C. glabrata (Orkwis et al. 2010;Nahas et al. 2018). To determine whether the THI UAS in C. glabrata is similar to a UAS in S. cerevisiae, we identified through MEME-suite analysis the regions in S. cerevisiae that are most similar to the 13 bp UAS ( Figure S5). We then mutated them in the ScTHI5 and ScTHI20 promoters, and assayed the Figure 4 Fine scale truncation analysis of THI promoters uncovers a 13 bp UAS that is not present in CgPMU3 A-D) We truncated THI promoters in 100 bp intervals and then further narrowed down to regions where we observed a .90% decrease in thiamine starvation induction. E) With 50 bp regions of the THI promoters around the site of truncation, we performed a MEME analysis and identified a 13 bp region which was not present in the 1000 bp CgPMU3 promoter. The TTCCCTBTAAWTG consensus is only found in 4 promoters in the C. glabrata genome, and those genes do not appear to be regulated by thiamine starvation based on previous RNA-seq data (Nahas et al. 2018). Each promoter element has at least one mismatch from the consensus, suggesting some permissiveness in the element. The arrow indicates the nucleotide number upstream from the start codon and the gray shaded regions are conserved nucleotides. F) A schematic of the location of the two elements in the five most upregulated THI pathway promoters with the arrows indicating where a truncation reduced expression. The blue boxes correspond to the CgPMU3 UAS and the orange boxes correspond to the THI UAS.
ability of these promoters to induce expression during thiamine starvation ( Figure 7A). Surprisingly, deletion of this element did not disrupt expression of these genes during thiamine starvation. However, deletion of regions near this site (within 20 bp and spanning the putative Thi2 binding site: at -110 bp in ScTHI5p and at -170 bp in ScTHI20p) did disrupt upregulation ( Figure 7B and 7C and Figure S5). Using computational methods, there is a low confidence sequence of 59-tatatgta-39 as a Pdc2 binding site (Reddy et al. 2007;de Boer and Hughes 2012), but if there is degeneracy or error, this site could be in many locations, as we note in Figure S5. A detailed dissection of the S. cerevisiae promoters is warranted and this is in process in our laboratory. Ultimately, we were surprised that the 13 bp THI UAS was not required for expression, but given our results later with the incorporation of a Thi2 binding site into C. glabrata THI promoters, we believe that the ScPdc2 binding site may be highly degenerate, or not even required in all contexts.
Thi2 dependence -CgPMU3 UAS with ScThi2 is different from the THI UAS with ScThi2 The data in Figure 6 with the switching of the UAS elements suggest that the CgPMU3 promoter fundamentally behaves differently from the other THI promoters, but the THI and CgPMU3 promoters are still dependent on the two known transcriptional regulators CgPdc2 and CgThi3 (Iosue et al. 2016). Because CgPMU3 evolved recently, and likely did not experience selective pressures from the ancestral Thi2, we hypothesized that it may behave differently in a setting where Thi2 is important. To test this, we cloned a putative ScThi2 binding site from ScTHI20 (Nosaka 2006) into the CgTHI10 and CgPMU3 promoters 5 bp upstream of the THI or CgPMU3 UAS. We chose the ScTHI20 Thi2 binding site because it is highly conserved in ScTHI20 promoters across the Saccharomyces genus (Kellis et al. 2003). We were unsure which orientation would work, as the site appears to be in either orientation in other promoters (Nosaka 2006), therefore we cloned the Thi2 binding site in both the 59ggaaacccttagag 39 "forward" orientation and the 59 ctctaagggtttcc 39 "reverse" orientation. We had already determined that none of the C. glabrata THI promoters (including CgPMU3) were regulated in S. cerevisiae (Figure 8, and data not shown), and we asked whether the inclusion of a ScThi2 binding site altered the ability of the C. glabrata promoter to be regulated in S. cerevisiae.
Inclusion of the ScThi2 binding site allowed for regulated expression in S. cerevisiae for both the CgTHI10 and the CgPMU3 promoters ( Figure 8). However, the two promoters' dependence on the transcription factors is altered. The CgTHI10 promoter is ScTHI2 and ScPDC2 dependent and requires the 13 bp UAS in concert with Figure 6 The THI UAS is able to substitute for the CgPMU3 UAS, but the CgPMU3 UAS cannot replace the THI UAS. A) Deletion of the CgPMU3 UAS in the context of the full-length CgPMU3 promoter results in a severe defect in thiamine starvation inducible expression of YFP; however, replacement of the CgPMU3 UAS with either the CgPET18 or CgTHI10 UAS restores upregulation of the CgPMU3 promoter. B) Deletion of the THI UAS in the context of the full-length promoter results in a severe defect in thiamine starvation inducible expression of YFP; however, replacement of the CgPET18 UAS or the CgTHI10 UAS with the CgPMU3 UAS does not restore upregulation of the promoters.

Figure 5
Deletion of the THI UAS eliminates thiamine starvation inducible expression of THI promoters. The putative THI UAS ( Figure 4E) was precisely deleted, and replaced with a PacI restriction enzyme site, in the full-length promoters of CgPET18 and CgTHI10 and assayed for YFP expression in high and no thiamine conditions. the ScThi2 binding site ( Figure 8A). However, CgTHI10 cannot tolerate the ScThi2 binding site in a reverse orientation, suggesting that there is an important quaternary interaction between Thi2 and Pdc2 to position the RNA polymerase machinery. In many ways, the introduction of the ScThi2 binding site has converted the CgTHI10 promoter into a standard S. cerevisiae THI promoter, albeit not nearly as efficient, as the amount of expression is only double the background level of fluorescence.
In contrast to CgTHI10, CgPMU3 has acquired upregulation in a different manner. First, the inclusion of the ScThi2 binding site can be in either orientation to confer upregulation, although there appears to be a preference for the reverse orientation for maximal expression ( Figure 8B). Second, while the CgPMU3 promoter in S. cerevisiae requires both ScTHI2 and ScPDC2, it is unclear where ScPdc2 binds, as loss of the 11 bp UAS has no effect on the upregulation. The CgPMU3 promoter has not adopted a behavior like other THI promoters, but appears to be regulated because ScThi2 is able to bind to the promoter, and likely ScPdc2 has accompanied ScThi2 because it is in a complex with it, and ScPdc2 allows for the recruitment of the RNA polymerase machinery. Thus, we conclude the CgPMU3 promoter has a significantly different cis architecture from other THI promoters, and this is likely a consequence of the lack of co-evolution with Thi2.

DISCUSSION
We have identified two unrelated UASs in thiamine starvation-regulated promoters in C. glabrata. One UAS is likely similar to the common ancestor of THI promoters, where the transcription factor Thi2 was present during the selection for thiamine regulation. This THI UAS is likely a relatively degenerate sequence that is able to recruit Pdc2. The CgPMU3 UAS is new and likely never experienced selection with Thi2 present. We have determined that the UASs are not interchangeable for one another and that they function differently from one another based on how they behave with a ScThi2 binding site introduced.
While more work is required to understand how the architectures of these two promoters work, we hypothesize that S. cerevisiae THI promoters behave as presented previously. That is, when the intracellular TPP concentration is low, Thi2 and Pdc2 bind with Thi3 to drive transcription ( Figure 9A). However, our work suggests that Thi2 binding is the "anchoring" step, and because Pdc2 is in a complex with Thi2, Pdc2 is then able to bind to degenerate sequences nearby, leading to the recruitment of the transcriptional machinery. This alteration in the model is supported by strong conservation of a Thi2 binding site in S. cerevisiae THI promoters, but a weak conservation of the THI UAS that we identified in this study (Kellis et al. 2003). Pdc2 is still required for transcription, but there is not a clear site for its binding. The inability to gel shift Pdc2 to THI DNA elements through EMSA experiments, and the very weak interaction of the DNA binding domain of ScPdc2 with a single DNA element that is Thi2 independent, suggest there is not a high affinity DNA-transcription factor interaction Figure 7 Deletion of regions most similar to the THI UAS in S. cerevisiae promoters does not abrogate thiamine starvation regulation, but deletion of regions near the UAS reduces expression. A) Mutation of the region most similar to the THI UAS has little effect on upregulation in two S. cerevisiae promoters (see Figure S5 for details on the sequence). B) A scanning deletion of the promoter region of ScTHI5 and C) ScTHI20 uncovers 20 bp that appear important for expression. These regions span the Thi2 binding site and are near the putative CgTHI UAS. (Nosaka et al. 2012). Additionally, while deletion of THI2 removes the majority of expression of THI promoters, there is still some induction in the absence of Thi2 ( Figure S6), and overexpression of THI3 can compensate for the loss of THI2 presumably by making Pdc2 fully active while in a complex with Thi3 (Iosue et al. 2016). However, loss of PDC2 removes all induction in response to thiamine starvation, suggesting Pdc2 is core to the transcriptional response. Thi2 may be an important anchoring transcription factor in S. cerevisiae THI promoters that facilitates Pdc2 transcription factor binding. Thus, Thi2 appears to be both a specificity and high-level expression factor for THI genes, and Pdc2 is required for recruitment of the core transcriptional machinery.
We hypothesize that C. glabrata THI promoters (with the exception of CgPMU3) behave similarly to S. cerevisiae promoters but only require Pdc2 and Thi3 ( Figure 9B). This could be a consequence of the C-terminal activation domain as the two Pdc2 proteins are 80% identical in the N-terminal DNA binding domain half of the protein, but only 30% identical in the C-terminal region (Nosaka et al. 2012). This difference in Pdc2 proteins between the species could allow CgPdc2 to recruit RNA Figure 8 CgTHI10 and CgPMU3 respond differently to the introduction of a ScThi2 binding site. A) A ScThi2 binding site (forward and reverse orientation) was introduced into the CgTHI10 promoter with and without the THI UAS deleted. These plasmids were transformed into S. cerevisiae strains and assayed for fluorescence in thiamine starvation conditions. For there to be increased expression of CgTHI10 in S. cerevisiae, the Thi2 binding site must be incorporated in the forward orientation and expression requires the THI UAS. B) The CgPMU3 promoter tolerates the ScThi2 binding site in either orientation and does not require the CgPMU3 UAS to function in S. cerevisiae, but expression is still Thi2 and Pdc2 dependent. CgPMU3 with a Thi2 binding site leads to higher level expression of the promoter in S. cerevisiae relative to CgTHI10. It is unclear why the two promoters have such different expression levels.

Figure 9
Model of transcription factor binding sites in thiamine starvation regulated promoters in S. cerevisiae and C. glabrata. A) ScThi2 binding may be the "anchoring" step, and because Pdc2 is in a complex with Thi2, Pdc2 is then able to bind to degenerate sequences nearby, leading to the recruitment of the transcriptional machinery. B) THI promoters in C. glabrata (other than CgPMU3) behave similar to S. cerevisiae promoters but only require Pdc2 and Thi3. C) For the CgPMU3 promoter, it seems likely that a novel transcription factor has been co-opted into the THI pathway to act as a functional analog to Thi2, and it may bind both the CgPMU3 UAS and CgPdc2. Regardless of where Pdc2 binds, it is still required for the recruitment of the transcriptional machinery. polymerase in a THI2 independent manner and/or increase the affinity of the transcription factor for its DNA binding site.
CgPMU3 presents a novel solution to thiamine starvation regulation. Whereas we can replace the CgPMU3 UAS with the THI UAS and restore upregulation to some degree, the reverse is not true. This suggests that the CgPMU3 UAS does not specifically recruit CgPdc2, but inclusion of the THI UAS now converts CgPMU3 into a "standard" C. glabrata THI promoter. Additionally, introduction of a ScThi2 binding site into the CgPMU3 promoter does not confer the same behavior as when it is introduced into the CgTHI10 promoter. Thi2 in combination with Pdc2 confers upregulation in CgPMU3, but now the orientation of the site is irrelevant and the UAS is not required, suggesting that the only reason the CgPMU3 promoter can work in S. cerevisiae is because of Thi2 recruitment to the promoter ( Figure 8C). These data suggest two things. First, that the CgPMU3 UAS is unlikely to bind Pdc2 with a high affinity, whereas the THI UAS likely does have a high affinity for Pdc2. Second, that Thi2 may be an important anchoring transcription factor in S. cerevisiae THI promoters as opposed to the Pdc2 transcription factor. This anchoring effect is reminiscent of Pho4 and Pho2 in S. cerevisiae, which regulate the induction of phosphate starvation genes. Pho4 has a well-defined recognition motif (GAGCTC), but Pho2 has a much more permissive recognition site (Zhou and O'Shea 2011;He et al. 2017). Our work suggests that there is a great deal of flexibility in the Pdc2 binding site and that Pdc2 may be binding to a relatively degenerate sequence. We believe that the CgPMU3 promoter has acquired a novel mechanism for thiamine starvation regulation. It seems possible that a novel transcription factor has been co-opted into the THI pathway to act as a functional analog to Thi2, and it may bind both the CgPMU3 UAS and CgPdc2 ( Figure 9C). However, other possibilities exist, and we are investigating these possibilities.
Characterization of the CgPMU3 promoter UAS has uncovered a potential new mechanism to regulate thiamine starvation genes and has demonstrated an interesting aspect of cis regulatory acquisition. Often, there is the recruitment of the same transcription factors, and thus, the apparent convergent evolution of the same cis sequences to bind those factors (Dalal et al. 2016;Cvekl et al. 2017;Kuang et al. 2018). However, we observe a novel cis regulatory sequence in a promoter that is recently evolved in a different genetic milieu (i.e., lack of THI2), but still gives the same output as many other THI genes. It seems as if the simplest solution for this new promoter would be to evolve the standard ancestral THI UAS, but CgPMU3 did not acquire that solution, either because of genetic constraints or because of genetic drift. However, the selective requirement for a thiamine repressible phosphatase important for the recycling of thiamine was likely present in the history of C. glabrata (Nahas et al. 2018). Therefore, it is possible that the lack of Thi2 in C. glabrata acted as a constraint in the evolution of the thiamine starvation induction of the CgPMU3 gene, yielding the only high-fitness solution to the problem. Further dissection of how each promoter functions is required to understand the precise mechanism of CgPMU3 upregulation. However, this work suggests that caution should be taken when investigating the incorporation of new genes into an existing regulatory pathway, as gain/loss of a constraint may change cis architecture in unforeseen ways.

ACKNOWLEDGMENTS
We thank Bin He (University of Iowa) and Troy Shirangi (Villanova University) for useful comments and suggestions. The National Science Foundation (MCB 1412582 and 1921632), Villanova University, and the Dennis M. Cook Endowed Gregor Mendel Chair provided funding for this work. The authors have no financial or non-financial competing interests.