signature=5b1015b92073f0266be741443236bf1e,Microarray karyotyping of commercial wine yeast strains r...

Microarray-based karyotyping

In the microarray karyotyping protocol employed here, genomic DNA from an experimental strain (i.e., a commercial wine yeast strain) was labeled with the fluorescent dye Cy5 (red), while genomic DNA from a wild-type (WT) reference strain (the sequenced haploid laboratory S. cerevisiae strain S288C) was labeled with the fluorescent Cy3 dye (green). The two labeled samples were mixed, then competitively hybridized to a spotted DNA microarray in which each spot is composed of PCR-amplified DNA corresponding to one gene of the reference S288C yeast S. cerevisiae [31, 61]. Under standard conditions [31, 61], hybridization to the microarray spots will generally occur between DNA molecules with 75% or more sequence identity. All experiments were performed in duplicate, i.e., the labeled DNA was hybridized to two separate microarrays. Note that the raw data for all microarray hybridizations reported in this paper can be downloaded from the website http://microarray-pubs.stanford.edu/wine_yeast/ as well as from the Stanford Microarray Database [62, 63].

Assuming that the experimental strain's genome has not experienced significant sequence divergence (i.e., greater than 25%) from that of the laboratory reference strain – a supposition which appears to be true for most S. cerevisiae wine yeast strains, which have an average of greater than 99% identity to the S288C strain [64] – any DNA copy number change in a single gene within the experimental strain will be detected by a deviation of the R/G ratio for that gene from the expected 1:1 ratio. If there are proportionally more copies of the gene in the experimental strain relative to the reference strain, the R/G ratio will thus be higher than 1:1 (ideally, it will be 2:1 for a duplication event, 3:1 for a triplication, etc.). Likewise, if there has been a deletion of the gene in the experimental strain relative to the reference strain, or if there are more copies of the gene in the reference strain than in the experimental strain (i.e., the experimental strain has experienced a "depletion" of the gene in terms of copy number), the R/G ratio will be less than 1:1 (ideally, it will be 1:2 for a single duplication in the reference strain relative to the wine strain, or for a deletion in the experimental strain of a single-copy gene in the reference strain, and so on). Thus, with only one hybridization, the DNA copy number of all 6,000 genes of a given strain can be determined relative to the reference strain.

Because microarray karyotyping detects proportional changes in DNA copy number relative to a reference strain, genome rearrangements that result in net copy number alterations can be easily detected with this method, including gene duplications or deletions, as well as larger-scale rearrangements such as non-reciprocal translocations, loss or amplification of large chromosomal regions, and ploidy changes for a single entire chromosome (aneuploidy). These large-scale rearrangements are seen as large regions of the chromosomes for which a similar bias in the ratios exists for all the genes in the region. Thus, despite the fact that complete genomic sequencing of every wine yeast strain is impractical, a great deal of information about the genomic structure of wine strains relative to laboratory strains and to each other can be achieved by microarray karyotyping.

Strains used for analysis

We chose four different commercial wine yeast strains for this study, using independent isolates from each of three different commercial and/or academic sources as shown in Table 1. The four yeast strains used were Montrachet, Pasteur Red, Pasteur Champagne, and Prise de Mousse (see Table 1 and Methods for complete list). These are among the most commonly used commercially-produced yeast strains in winemaking. Each strain has distinct fermentative qualities, and each is thought to bring unique flavor and other sensory components to the finished wine product [65, 66]. Independent isolates of each strain were obtained from two major wine yeast producers, Lalvin and Red Star, as well as from the wine yeast strain collection at Univ. of Calif., Davis (Table 1).

Table 1 Wine Yeast Strains used in Study

The genomes of all four wine yeast strains are highly similar to that of S. cerevisiae S288C

We performed microarray karyotyping, as described in Methods, on each of the wine yeast strains. Because some of the strains are listed by suppliers as either S. bayanus or S. cerevisiae/S. bayanus, we also performed microarray karyotyping experiments using genomic DNA from a known type strain of S. bayanus to determine what possible signal may be generated when hybridizing S. bayanus genomic DNA to S. cerevisiae microarrays. We performed a similar experiment using S. cerevisiae DNA as a control (a "self-self" hybridization).

Figure 1 shows a graphical representation of the microarray karyotyping data for the S. cerevisiae and S. bayanus strains presented as "karyoscope" diagrams, made using the Java TreeView program [67]. To make a karyoscope diagram the hybridization ratio for each gene is mapped onto its corresponding position on each chromosome of the reference strain of S. cerevisiae. The height of each red or green vertical bar is proportional to the log2 of the red:green (R/G) ratio for a gene; if the ratio is greater than 1 (i.e., a positive log2 value), the bar will be red and drawn above the chromosome; a red bar thus represents an over-representation (amplification) of that gene in the wine strain relative to S288C. For R/G ratios less than 1 (i.e., a negative log2 value), the bar will be green and drawn below the chromosome; a green bar thus represents an under-representation (deletion or depletion, i.e., lower copy number) of that gene in the wine strain relative to S288C.

Figure 1

Karyoscope views of S288C-S288C and S. bayanus -S288C microarray hybridizations. Microarray hybridizations were performed as described in the text. In both panels the S. cerevisiae laboratory strain S288C was used as a reference sample, labeled with Cy3 dye (green). In panel A, the Cy5-labeled (red) DNA was also S288C, giving a self-self hybridization. In panel B, the Cy-5 labeled DNA was S. bayanus. The labeled DNAs were competitively hybridized to spotted microarrays bearing full-length ORFs from the S288C S. cerevisiae strain; the data thus obtained is displayed here in graphical form as a karyoscope. Red bars indicate red:green ratios above 1.0 and are graphed on a log scale; green bars indicate red:green ratios below 1.0, also graphed on a log scale. The chromosomes are shown in each panel in numerical order with chromosome 1 at the top and chromosome 16 at the bottom, and are aligned by their centromeres, with their left arms extending to the left. Both karyoscopes are drawn to the same scale, i.e., the bar heights in panels A and B proportionally represent the same amplitude of change.

In the "self-self" hybridization control, the genomic DNA from the reference S. cerevisiae laboratory strain was labeled independently with Cy3 and Cy5 and then competitively hybridized against itself. We observed robust hybridization across the entire array, and after a global mean normalization (see Methods) there were equal red and green hybridization signals (R/G ratios of 1), with no obvious deviations for any particular gene, as expected (Fig. 1A). We found, however, that the S. bayanus DNA did not hybridize to the arrays in the same manner as did the S. cerevisiae genomic DNA. The S. bayanus hybridization signals (Fig 1B) were significantly weaker and much more variable across the array, as compared to the S. cerevisiae self-self hybridization (Fig 1A).

For all 12 of the wine yeast isolates, we observed hybridization patterns much more similar to that of the "self-self" experiment, i.e., with robust signals across most of the array, and the majority of ratios near a value of 1.0. This indicates that each of the wine strains possesses an essentially complete complement of the S. cerevisiae genome, with no observed aneuploidy. Note that we cannot rule out that there may be additional genomic DNA in these wine strains that comes from other Saccharomyces species (i.e., if the strain is a hybrid organism) whose hybridization signal would be overshadowed by the signal generated by the S. cerevisiae genomic complement. However, we do not think this is likely, as control spots on the microarrays containing S. bayanus genomic DNA did not show a hybridization signal, which we consistently observe when using known S. cerevisiae – S. bayanus hybrid yeasts (B. Dunn, unpublished). Thus, we do not expect any potential non cerevisiae DNA in any of these wine strains to have any great effect on our observed ratios.

Genomic differences between wine yeast strains and the laboratory strain

We observed a distinct set of genomic variations that occur in every wine strain relative to the laboratory strain; these are shown in Figure 2, and an abbreviated list of these variations is given in Table 2. Figure 2 represents a consensus karyoscope-type plot generated by the program CGH-Miner [68], where the data from the 4 wine strains (taken as averages of the data for the 3 isolates of each strain) have been consolidated into statistically-significant amplifications or depletions relative to multiple self-self hybridizations of the laboratory S288C strain.

Table 2 "Commercial Wine Strain Signature" Gene Lista

Figure 2

Changes in Gene Copy Number Shared by all Wine Strains: the "Commercial Wine Yeast Signature". A consensus plot pooling the results of all the wine strains relative to the lab strain S288C was generated by the program CGH-Miner [68]. This plot is similar to a karyoscope in that it displays values along each chromosome, and the chromosomes are shown in numerical order from top to bottom. However, in this plot, a determination of which regions were statistically significantly altered in copy number for each wine strain (i.e., averaged array values for each of the three isolates within a strain) relative to the laboratory S288C strain were determined. For each of these regions the number of wine strains showing a significant change were then added together to give a metric of how well-shared among the wine strains that particular change was. This is shown as tall red bars for significantly amplified regions among all wine yeast strains relative to the laboratory strain, and as tall green bars for significantly depleted regions in all the wine strains relative to the laboratory strain. Each gray "shadow line" above or below the chromosome represents 20% of the number of wine strains showing a significant change in that region (see legend). The group of changes comprising the tallest (>75%) red and green bars represents the "commercial wine strain signature"; these are listed in Table 2.

The most distinctive genomic variations seen in Fig. 2 are the many prominent green bars which are seen on most of the chromosomes. These shared green bars represent genes or regions in the wine strains, taken together as a whole, that are depleted in copy number relative to the laboratory strain. In most cases these shared depleted regions consist of genes that are moderately or highly repeated in the laboratory strain, with the majority representing Ty1 elements (a yeast retrotransposon family). Also depleted are the tandemly-repeated ASP3 (asparaginase) genes on chromosome 12, and the tandemly-repeated ENA (sodium transport) genes on chromosome 4; note: general descriptions and functions for genes described in this paper were obtained from the Saccharomyces Genome Database (SGD) [69, 70]. The CUP1 (copper binding) tandemly-repeated genes are also significantly depleted in 3 of the 4 strains (Fig. 2, Table 2) and additionally show some intra-strain variation in copy number (Fig. 5B); depletion of the CUP1 genes has been observed previously in a wine strain [50]. Because these sequences are all moderately repetitive and highly conserved, we can only interpret these results to indicate that in all of the wine strains we investigated, there are either fewer or no copies of these repeated genes relative to the laboratory strain.

Because the CGH-Miner program utilizes a moving-average method for calculating significance, copy number changes of single isolated genes are not detected. However, inspection of clustered data and karyoscopes allowed us to identify single genes whose copy numbers are highly altered relative to the reference strain. Among these, we found that the single-copy drug response genes SNQ2 and PDR15 are depleted in all of the wine strains we examined (Fig 5B). In addition, PDR3, a single-copy zinc-finger transcription factor involved in pleiotropic resistance to drugs [71], appears to be entirely deleted in all four wine strains due to the very low R/G ratios it exhibits (Table 2, Fig. 5B). This gene resides adjacent to one of the Ty1 elements that is depleted in the wine strains. Additional altered-copy number genes identified by inspection of karyoscopes are listed in Fig. 5B.

In addition to the shared depleted regions, there are shared regions that appear to be slightly, but consistently, amplified in all four wine yeast strains; the most prominent is located at the right end of chromosomes 1; other regions are on the right arms of chromosomes 2 and 12, and the left end of chromosome 16 (Fig. 2, Table 2). These shared amplified regions contain some genes that are members of homologous gene families in the reference strain, such as the IMD genes (involved in GMP production) and PHO genes (acid phosphatases), but they also include unique genes such as SAM3 and SAM4 (S-adenosyl methionine transport and S-AM metabolism, respectively), PET8 (mitochondrial S-AM transport), RDS1 (response to xenobiotic drugs, including fungicides, herbicides, and pesticides), TPO1 (multidrug and polyamine transporter and drug detoxification), AAD3 (aryl-alcohol dehydrogenase), ADH7 (alcohol dehydrogenase), and FDH1 and FDH2 (formate dehydrogenase) (Table 2, Fig. 5B). Also occurring in an amplified region is the un-named and uncharacterized gene YAR068W (Table 2, Fig. 5B). This gene codes for a membrane protein which when overexpressed results in resistance to the drug ketoconazole [72]; additionally, it has been found to have one of the most elevated Ka/Ks ratios among the genomes of the Saccharomyces sensu stricto group [73], signifying that it may be a gene that is actively undergoing positive selection [74]. A homologue to this gene, YHR214W-A, is also amplified in most of the wine yeast strains.

These results indicate that all of the wine strains that we examined have experienced a characteristic increase or decrease in copy number, relative to that of the laboratory strain, of the genes listed in Table 2. Although some of these changes, particularly the low copy number (or absence) of the ENA and ASP3 genes and of Ty1 elements, are consistently observed in other industrial and non-S288C laboratory yeast strains ([51, 52, 57–60], B. Dunn, unpublished), most of the other changes, particularly the amplifications, are novel and thus far appear to be unique to commercial wine yeasts.

Genes whose copy number changes are shared by all wine strains

Interestingly, there are many transporter and permease genes, including some involved in drug response, that show altered copy numbers in the all the wine strains relative to the laboratory strain. As described above, PDR3, PDR15, RDS1, SNQ2, TPO1, and YAR068W show alterations in their copy numbers that are shared by all the wine strains. In addition, HXT9 and HXT11 are depleted in three of the four strains (Fig. 5B), and AYT1 is depleted in 2 strains and amplified in the other two. All of these genes are involved in some aspect of drug response [71, 72, 75–78].

It thus appears that a major group of shared genomic differences found among all the wine strains is composed of genes, especially transporters and permeases, involved in drug resistance pathways. Since vineyards are often treated with herbicides, pesticides, and fungicides, wine yeasts may be exposed to these agents on a routine basis and their genomes may have evolved to better tolerate such exposure. Alternatively – and perhaps more likely given the fact that these commercial strains may have been isolated prior to the widespread use of herbicides, pesticides and fungicides – the variations in transporter gene copy number seen here may have arisen to fine-tune the fermentation properties of the wine yeasts by altering the types and amounts of fermentable substrates that are taken up. These variations may also reflect an increased tolerance to, or excretion of, plant phenolic compounds. Thus, any changes in drug response pathways in these strains may be a secondary effect and not due to direct selection for drug resistance or sensitivity. This view is supported by reports that alterations in PDR1 and PDR3 gene expression, as well as mutations in the ABC transporter genes that they regulate, result in better fermentation performance by sake yeasts [79, 80].

Inter-strain variation

We also observed inter-strain genomic variations that are specific to each wine strain. Examples of such inter-strain genomic variation are highlighted as regions circled in black on the Karyoscopes in Figure 3A–D. For an idea of scale in these karyoscopes, the tallest bars on the figure represent ratios of approximately 2:1 (2-fold enhanced) for red, and 1:16 (16-fold depleted) for green. Each panel in Fig. 3 represents the average of the hybridization log ratios for all three independent isolates of the given strain; since each isolate's DNA was hybridized in duplicate, each panel thus represents the average of six arrays, making the data very robust. Some examples of inter-strain variation are also shown in more detail in Figure 5A, in which the hybridization data are represented as colored boxes, with the most intense green representing severe depletion (R/G ratios much less than 1), black representing no change in copy number (R/G ratio of approximately 1), and the most intense red representing significant amplification (R/G ratios greater than 1).

Figure 3

Wine strain karyoscopes. Microarray ratios from each of the three isolates of a given wine strain were averaged to yield an "average" microarray karyotype for each wine strain. Chromosomes 9 through 16 for the 4 strains are shown in panels A – D as follows: A. "Mont" = Montrachet (average of strains GSY1-3). B. "PDM" = Prise de Mousse (average of strains GSY10-12). C. "Champ" = Pasteur Champagne (average of strains GSY7-9). D. "Red" = French Red (average of strains GSY4-6). Circled regions highlight areas that vary in characteristic ways between the different wine strains. All karyoscopes are drawn to the same scale.

One distinct inter-strain variation is seen in the copy numbers of HXT (hexose transporter) genes. As shown by the black-circled regions on chromosomes 9 and 10 in Fig. 3A–D, and in the top panel of Fig. 5A, the Prise de Mousse isolates show no depletion of the HXT9, HXT11, and HXT12 genes, whereas the Pasteur Red isolates show a slight depletion of these genes. On the other hand, the Montrachet and Champagne isolates show extreme depletion of these genes. Since these three HXT genes are all virtually identical to each other and can cross-hybridize, this result is most likely explained by a loss of one or more of the three genes in the depleted strains. It is impossible, however, to know which particular gene(s) are missing without further investigation.

Another class of inter-strain differences is of particular interest because the differences are seen in genes that are single-copy in the laboratory strain. Depletions of these genes in a given wine strain as measured by array karyotyping may thus reflect a true deletion of that gene relative to the other wine strains, rather than just a decrease in copy number. Three instances of this type of inter-strain variation were found: the AYT1 gene (chromosome 12), the ENB1 gene (chromosome 15), and the un-characterized ORF YPL257W (chromosome 16). The circled regions on the relevant chromosomes in Fig. 3A–D highlight these single-gene strain differences, while Fig. 5A shows the data in more detail for these three genes. AYT1 encodes an acetyltransferase that was first characterized in Fusarium and subsequently identified in S. cerevisiae by homology [75]. It plays a role in de-toxifying endotoxins of the tricothecene family in Fusarium, but its function in S. cerevisiae is unknown. Although it has been shown that the S. cerevisiae AYT1 gene product can acetylate tricothecene in vivo, cells that are deficient for this gene show no increased sensitivity to the compound, and in fact show no phenotype at all [75]. ENB1 encodes an enterobactin transporter of the major facilitator superfamily involved in iron uptake [81], while YPL257W has no known function, although its predicted product has homology to membrane proteins.

Intra-strain variation

We observed relatively little genomic variation between independent isolates of each of the different strains. A major exception to this was seen, however, among the three different isolates of Montrachet. As shown in Figure 4, the Montrachet isolate obtained from Lalvin (GSY1) has a fairly extensive region near the left end of chromosome 7 that exhibits copy-number depletion (relative to the S288C reference strain), seen as a distinct cluster of green bars (a thick black line underlines this region). This region spans approximately 37 contiguous genes, all of which show a ratio lower than 1 (shown as green bars). This is in contrast to the other two Montrachet isolates (GSY2 and GSY3 from Red Star and UCD, respectively), which show no copy-number depletion in this region (Fig. 4). Because so many contiguous genes are involved in this block of genes, and because almost all of the genes exhibit a similar direction and magnitude of deviation, we believe that this region has been lost from either one or both copies of chromosome 7 in the (presumably diploid) Lalvin isolate, yielding a net depletion in the copy numbers of these genes relative to the reference strain. It is formally possible that sequence divergence in the region caused the decreased hybridization to the reference strain, but this is unlikely because sequence divergence rarely, if ever, occurs in such a dense and homogeneous pattern, nor is it ever constrained to only one region of the genome. Since there are at least two essential genes in this region (see below), it is most likely that only one of the chromosomes in this presumably diploid strain has lost this region.

It is likely that the region of depletion extends out to the end of the chromosome, despite the fact that the two distal-most genes appear as red (amplified) bars; both genes are sub-telomeric genes that are repeated in the genome and it is thus not possible to determine whether the chromosome 7 copies are specifically present or not. Therefore, the event that caused the depletion of this large region of chromosome 7 in the Lalvin isolate could have been a non-reciprocal translocation resulting in the net loss of the distal end of chromosome 7. This is supported by the fact that the inner (proximal) boundary of the depleted region occurs exactly at a tRNA (tV(AAC)G3). It has been shown that non-reciprocal translocations and other types of gross chromosomal rearrangement events that result in better-adapted genotypes occur frequently during adaptive evolutionary growth, and that these rearrangements are almost always bounded by tRNAs or Ty1 elements [31, 56, 82–84].

Figure 4

Karyoscopes of Chromosome 7 from individual Montrachet isolates. The larger black bar on the left, placed under the left end of chromosome 7 from the Lalvin Montrachet isolate (GSY1), shows a 37-ORF region that has been depleted (or deleted) with respect to the S288C laboratory strain and with respect to the chromosome 7's of each of the other two Montrachet isolates. The smaller black bar on the right, under the UCD522 (GSY3) chromosome, indicates a region of 3 ORFs (including MAL11 and MAL13) that are present at "wild-type" (S288C) copy number in the UCD522 strain but are depleted or deleted in the other two Montrachet strains. Note that this same MAL region is also present at "wild-type" copy number in UCD725, but is depleted or deleted in all remaining wine yeast strains. Again, all karyoscopes are drawn to the same scale.

Figure 5

Unique combinations of genomic differences between different strains. Panel A: Selected portions of the averaged data shown in Karyoscope form in Figure 3 are shown here in color-block form. To generate this figure the averaged data for all four wine strains were sorted for the Montrachet strain from highest green values (greatest negative number) to highest red values (greatest postitive number) and visualized as a "clustergram" using Java TreeView (note that the data were sorted, but however, were not clustered). The topmost (most green) 25 genes are shown in the top panel; the lower two panels show the sorted data surrounding two of the genes (AYT1 in the middle panel, YPL257W in the lowest panel) whose copy number varies greatly on a strain-to-strain basis. Red asterisks mark the genes whose copy numbers vary in unique combinations between the different strains. Panel B shows a matrix of gene copy number changes for each of the 12 wine strain isolates, along with the drug response profiles of each strain isolate at the bottom. Within the copy number data portion, a dark red cell with a "++" symbol indicates a red:green log ratio greater than 0.8, light red with "+" indicates a ratio between 0.2 and 0.8, and gray with "+/-" indicates a ratio between 0.2 and -0.2. Likewise dark green with "--" indicates a red:green log ratio less than -0.8, light green with "-" indicates a ratio between -0.2 and -0.8.

The depleted region of Chromosome 7 in the Lalvin Montrachet isolate extends from Open Reading Frame ("ORF") YGL226W through YGL261W. Included in this region are three genes, two essential and one non-essential, involved in protein-nucleus export (BRR6, CSE1, KAP114), two genes involved in cell cycle control (SAP4, DOC1), as well as genes involved in secretion (SEC15), and respiratory growth (HAP2, MTO1). Also included in this region is a transporter protein for zinc uptake (ZRT1). Of possible significance in terms of winemaking are genes that are involved in glucose/fructose (HXK2) [55] and alcohol (ADH4) metabolism [85], as well as one gene (YGL258W) involved in velum formation in a flor (Sherry) yeast strain. Interestingly, YGL258W and ADH4 are both induced under zinc-deficient conditions [86, 87], which may have significance with respect to the associated depletion of ZRT1. If significant fermentative or organoleptic differences exist between the Lalvin Montrachet isolate and the other two Montrachet isolates, it would be interesting to determine whether adding back genes from the depleted region (e.g. on a stable single-copy CEN plasmid) to the Lalvin isolate would cause its fermentative and/or organoleptic characteristics to now mimic those of the other two isolates.

A second intra-strain difference we observed is a depletion of four genes near the right end of chromosome 7 that is seen in both the Lalvin and Red Star Montrachet isolates but not in the UCD522 Montrachet isolate (Fig. 4, shorter black bar underlining the region at the right telomere). This group of four genes includes a maltose permease gene (MAL11) and a maltose-activator protein (MAL13); loss of these genes usually indicates a defect in maltose fermentation, but they also appear to be involved in drug response pathways since deletions of MAL11 lead to nystatin sensitivity [88]. We observed that this same group of MAL genes is depleted in all of the other strain isolates except in the French Red isolate UCD725 (GSY5). In other words, of all 12 isolates that we studied, only UCD522 (GSY3) Montrachet and UCD725 (GSY5) French Red do not show a depletion of this MAL region relative to the reference S288C strain (Fig. 3A–D, Fig. 5B).

Finally, two genomic differences were seen in the UCD725 French Red isolate (GSY5) when compared to the other two French Red strains (GSY4 and GSY6). First, as mentioned just above, the same group of MAL genes that is present at wild-type copy number in the UCD522 Montrachet isolate is also present in this UCD725 isolate, but is depleted in the other two French Red isolates. Secondly, YPL257W is depleted in the French Red strains GSY4 and GSY6, but is present at wild-type copy number in GSY5 (data not shown). This gene, encoding a putative membrane protein, is one of the set of genes that exhibit a characteristic copy number genotype between the different wine yeast strains (see below), but it apparently varies in its genotype among the French Red isolates. Overall, however, aside from the fairly small differences among the Montrachet and French Red strains, there is remarkably little intra-strain genomic variation among the strains we studied, indicating that these wine strains are relatively stable genetically, at least as assayed by microarray karyotyping.

Transporters and permeases are part of the shared group of copy number differences, and also show inter- and intra-strain differences

Figure 5B shows a matrix of relative copy numbers for each of the 12 individual wine strain isolates. The first group of genes shown in the matrix represents all of the Major Facilitator Superfamily (MFS) transporter genes for which a large copy number deviation relative to the reference strain (defined as log2 (R/G) value greater than 0.8 or less than -0.8, corresponding to a 1.74 fold change) was seen for at least one isolate. A "-" or "--" sign indicates moderate or large depletion of the gene's copy number, respectively; assuming that these are diploid strains, a "-" would indicate a loss of one of the two copies and a "--" would indicate loss of both copies. Likewise, "+" or "++" indicates moderate or large amplification of the gene, respectively, which would indicate an increase from 2 to 3 copies for "+" and to 4 or more copies for "++". Remarkably, of the total of 27 MFS transporter genes annotated as such in the Saccharomyces Genome Database [69, 70], almost half (13) show large deviations (as defined above) in copy number among the wine strain isolates; all 13 are shown in Fig. 5B. Most MFS transporters show a depletion in copy number relative to the wild-type strain; TPO1, which can transport cycloheximide and other drugs [78], is the major exception, showing amplification in all strains.

Figure 5B also shows copy number data for known permeases and other transporters, as well as any genes annotated as having a role in drug response that show a large deviation in at least one isolate. Finally, genes which do not fall into a transporter/membrane protein class, but which nonetheless stood out in the clustered data as being significantly altered in copy number relative to S288C are included (e.g., the CUP1 locus and members of the S-adenosylmethione metabolism pathway). At the bottom of the figure are shown the results of the drug halo assays for each of the isolates (see below).

Drug resistance/sensitivity phenotypes in wine strains

The prevalence of genes involved in drug response and drug resistance occurring in both the shared group of copy number changes, as well as in inter- and intra-strain differences, directed us to investigate the drug response profiles of the wine yeast strains. We chose three different drugs with which to assay the cells, representing three different general classes of drugs: cycloheximide, "CYH" (a protein synthesis inhibitor); clotrimazole, "CTZ" (member of the azole class of antifungal agents which inhibit ergosterol synthesis); and sulfomethuron methyl, "SMM" (a sulfonylurea herbicide). As described in Methods, halo assays were performed on all 12 wine strains listed in Table 1 as well as on six other control yeast strains mentioned in Methods. These additional six strains represent either control wild-type strains (FY1679, S288C, RW2802) or strains with single-gene deletions of various drug resistance genes whose copy numbers had been observed to be altered in the wine yeast strains: an AYT1 deletion strain (ScAYT1Δ; [75]), a PDR5 deletion strain (JG436; [75]), and a YPL257W deletion strain (GSY28).

Figures 6A and 6B show the actual halos on the plates, while Fig. 6C tabulates the halo results in chart form. The most notable feature of the chart is that the majority of the 12 wine strain isolates (Fig. 6A, 6C) are very sensitive to SMM compared to the WT diploid laboratory strain. However, there are three exceptions to the SMM sensitivity, with two of the three Montrachet isolates, Lalvin Montrachet (GSY1) and Red Star Montrachet (GSY2), as well as the UCD725 French Red isolate (GSY5), showing much less sensitivity or even low resistance to SMM.

Figure 6

Drug resistance phenotypes of wine strains. Halo assays were performed as described in the text to test the response of the wine strains to various drugs. Panels A and B show photographs of the halos; on the far left of each panel are shown the halo assays for the WT-diploid (FY1679, a S288C-based diploid, see Methods), against which all other strains were compared. CYH = 500 ng cycloheximide; CTZ = 5 μg clotrimazole; SMM = 20 μg sulfomethuron methyl. Panel A: strains are as described in Methods and Table 1; "WT-Dip" refers to FY1679. Panel B: Strains are as described in Methods; "WT-Dip" refers to FY1679; "YPL257 Δ" to GSY28; "ayt1 Δ" to ScAYT1Δ, "pdr5-" to JG436; and "WT-Hap" to S288C. Panel C: Graphical results of halo assays. Annular area (i.e., area of total halo minus area of disc) was calculated for each halo, then expressed as log2 of the ratio of the annular area to that of the WT-Dip (FY1679) annular area.

It can also be seen in Figs. 5B, 6A and 6C that the individual isolates of Champagne (GSY7-9) and Prise de Mousse (GSY10-12) wine strains show more consistency in their drug response profiles than do the individual isolates of Montrachet (GSY1-3) and French Red (GSY4-6). Interestingly, it is the individual isolates of both the Montrachet and the French Red wine strains that also showed the most variability in their genome structures (Fig. 5B).

Halo assays for the control yeast strains are shown in Fig. 6B. As expected, the pdr5 strain, known to be hypersensitive to drugs, is extremely sensitive to CYH and to CTZ. It is very resistant to SMM, however, with no halo at all forming around the SMM disc. The ayt1 Δ strain's response to CTZ or CYH is similar to that of the WT diploid, but like the pdr5 strain it is very resistant to SMM. The YPL257W Δ strain shows moderate resistance to SMM compared to the WT diploid, with no variant response to CTZ or CYH. Note that both of the haploid wild-type control strains, S288C ("WT-Hap" in figure) and RW2802, show greater resistance to SMM than does the WT diploid strain; S288C is moderately resistant and RW2802 is extremely resistant, with no halo detected in the presence of SMM. This may indicate that haploid strains are for some reason more resistant to SMM than diploid strains. It also appears that strain differences (i.e., compare RW2802 and S288C) may affect the degree of resistance to SMM.