Saccharomyces cerevisiae is one of the most studied eukaryotes. It has provided the basis for much of our understanding of the genetics, biochemistry, physiology and structure of eukaryotic cells (Botstein et al., 1997; Dolinski and Botstein, 2007; Kane and Roth, 1974; Lee and Young, 2000; Mackay, 2001; Moler et al., 2000; Myers and Kornberg, 2000; Nyberg et al., 2002; Sherman, 2005; Walberg, 2000; Wickner and Haas, 2000). The simplicity and speed of its life cycle and the facility with which it can be manipulated in the laboratory have made it the organism of choice for technologies ranging from yeast artificial chromosomes (YACs; Burke et al., 1987) to two-hybrid selections (Fields and Song, 1989), micro-arrays (Brown and Botstein, 1999) and proteomics (Hodges et al., 1999). As with many ‘model systems’, yeast laboratory strains are derived from a limited number of wild isolates. The resultant constraint on genetic variation helps make techniques reproducible and reliable in many laboratories under a variety of conditions. This also forms the basis for yeast heterothallic genetics, where a and α haploids mate with one another to form a/α diploids, which sporulate to produce 2a and 2α ascospores that can be propagated as stable haploids. Normal mating-type interconversion, which makes wild strains of yeast functionally homothallic, is prevented by mutations at the HO locus (Herskowitz, 1988). The ability to control yeast's natural alternation of generations affects every aspect of its use as a model organism.
For some purposes, a constrained genetic system is less than ideal or not useful at all. Notably, population biology and genetics depends on genetic variation at the level of species and individuals in wild populations to make inferences about ecological interactions, genetic structures, speciation and evolution. Breeding projects to produce improved varieties for agriculture and industry often depend on introgression of traits, such as insect or pathogen resistance, salt tolerance and growth or reproductive form, from the same or related species from nature.
The Saccharomyces Genome Resequencing Project (SGRP; Liti et al., 2009) provides unrestricted access to DNA sequence information (≥1× random coverage) for > 30 laboratory and wild-type strains of yeast. To begin to take advantage of this resource and to expand yeast genetics and breeding beyond the constraints of a few laboratory strains, we assembled a resource consisting of sequenced wild-type strains and unsequenced strains of high importance to industry, their F1 hybrids in all combinations and some F2 offspring. We report the construction and partial characterization of a library of F1 diploids formed from 16 strains mated in all combinations, called the Yeast Diversity Library (YDL), and a rational subset of that library arrayed into a single, 96-well microplate, called the Compact Diversity Library (CDL). We found that most of the hybrids were at least partially fertile, so F2 recombinants could be obtained. The problem of homothallism interfering with routine genetic manipulations was overcome by direct selection of complementing auxotrophic mutations following mating during ascospore germination. This approach is broadly applicable to wild strains so long as they can be induced to sporulate in the laboratory. It avoids the use of recombinant DNA techniques, which are precluded from strains destined for industrial use in many countries. As with other open access resources, the value and utility of this one should increase as it is used and augmented.
Materials and methods
Yeast strains and media
Saccharomyces cerevisiae strains used are given in Table 1, along with their assigned UA (University of Auckland) accession numbers. Derivative homozygotes are designated ##01 or ##03 for lys− or ##02 or ##04 for ura− strains. The numbering convention for F1 heterozygotes (lys−/+; ura−/+) employs the first one or two digits from the lys− parents' accession numbers (prefix) and progressive, two-digit numbers starting with 10 for the ura− parents (suffix; Table 2).
Table 1. Saccharomyces cerevisiae strains used in this study
The chart represents three 96-well microplates, labelled A, B and C, into which the library was deposited. The Accession Nos of the lys− strains are given on the left, the ura− strains above and the hybrids in their positions within the plates. Crosses between UA3602, 3904, 4104 and 4202 and UA2801–4201 were transposed into dish C to conserve space. The dashed, diagonal line represents the plane of symmetry of the collection. A selfed strain for UA4200 (S288C) was not obtained. UA4201 and 4202 vegetative cells, not ascospores, were used to make crosses. The F0 lines are in plate C, wells 9A–12H.
Growth media and conditions and tetrad dissections were as described in Guthrie and Fink (1991).
Isolation of random ascospores
A sample of 50–100 µl overnight YPD cultures was spread onto SM (1% KOAc, 0.1% yeast extract, 0.05% glucose, 2% agar) plates and incubated for 5–14 days at 28 °C until the cultures had sporulated. Cells were suspended in 15 ml water (3 × 5 ml), transferred to 50 ml Falcon tubes and pelleted (RCF = 4000). The pellets were washed three times with 20 ml water by centrifugation (5 min; RCF = 4000) and suspended in 3 ml water. Dithiothreitol (0.5 ml, 0.5 M) and lyticase (0.5 ml Sigma L-4025 at 80 units/ml in 50 mM dithiothreitol) were added, the volume was adjusted to 5 ml with water and the sample was incubated for 12–16 h at 28 °C with shaking (100–200 rpm). Sodium dodecylsulphate was added to 1%, the sample was vortexed and the cells were pelleted and washed three times in 20 ml STD (0.1% NaCl, 0.05% Triton X-100, 1 mM dithiothreitol). The sample was suspended in 3 ml STD, 30 µl ß-glucuronidase (Sigma G-0876; ∼110 000 U/ml) was added, and the cells were incubated for 12–16 h at 28 °C. The sscospores were pelleted, washed with STD (20 ml × 3), counted in a haemocytometer, suspended at 109 spores/ml, and stored at 4 °C.
EMS mutagenesis and mutant selection
Mutagenesis was carried out as described by Lawrence (2002). Overnight cultures of the strains given in Table 1 (5 ml YPD) were centrifuged for 5 min (RCF = 4000), washed twice with water and suspended in 5 ml 0.1 M NaPO4, pH 7.0. EMS (150 µl) was added, the tubes were vortexed and incubated for 1 h at 28 °C on a rotator. The cells were pelleted, washed three times with 1 ml 5% Na2S2O3 and suspended in 2 ml water. We did not measure the viability of the primary population of mutagenized diploids, as recessive lethal mutations would be masked, but used mutagenic conditions that led to 10–50% survival of most haploid strains. In order to resolve heterozygous mutations, samples were sporulated and ascospores were isolated. Ascospores (5 × 107) were inoculated into 2 ml YPD, grown to stationary phase at 28 °C on a rotator to allow mating-type switching and promote self-fertilization, and plated onto either 5-FOA plates (Boeke et al., 1984) for selection of ura− mutants or AAD plates (Chatoo et al., 1979) for selection of lys− mutants. Colonies appearing after 2–5 days were picked and replicated onto minimal medium (SD) and SD supplemented with lysine or uracil or both. lys− or ura− strains were sporulated and single spores from dissected tetrads were used to produce F0, a/α diploids (due to mating-type switching and subsequent mating). These strains and their derivatives may contain unselected mutations, so the YDL and CDL contain congeners for each F1 combination across the axis of symmetry. The probability of congeners containing the same unselected mutations is nil. Thus, comparison of congeners will detect any effects due to unknown genetic lesions.
Mating and selection
F0 diploids were sporulated and ascospores were isolated. F1 hybrids were produced by mixing 107 ascospores from ura− strains with 107 spores from lys− strains in a microfuge tube containing 0.5 ml YPD. The spores were centrifuged for 1 min in a microfuge (RCF = 10 000) to compact them, then incubated at 28 °C for 12–16 h without shaking. These conditions were intended to promote interactions between cells of different mating types and auxotrophies. The cells were then vortexed, pelleted, washed twice with 0.1% NaCl, suspended in 50 µl 0.1% NaCl and 10 µl was streaked onto quadrants of SD plates. Single colonies were selected after 2 days.
Strain identities were confirmed by microsatellite analysis, using a multiplex of 10 hyper-variable microsatellites plus the MAT locus (Richards et al., 2009).
Microtitre dishes with 100–200 µl/well YPD containing different concentrations of ethyl alcohol were inoculated from overnight cultures grown in YPD by using either a multiprong replicator (∼3 µl/well) or a multipet (5–20 µl/well) and incubated for 18–20 h at 28 °C. The density of cells in the inocula of the different strains was ± 10%, as estimated by OD650. Wells were mixed by pipetting up and down and OD650 was determined using a microplate reader.
Measurement of ethanol concentrations
Ethyl alcohol concentrations were estimated by a modification of the method of Lim and Buttery (1977). AMP buffer (0.65 M 2-amino-2 methyl-propan-1-ol; 0.1% v/v Tween-20, pH 8.5, 50 µl) was pipetted into standard microplate wells and 1 µl ADH suspension (15 mg Sigma A-7011 suspended in 1 ml 45.5% (NH4)2SO4, 3% Na4PPi·H2O) and 20 µl colour reagent (80 mg iodonitrotetrazolium chloride, 20 mg phenazine methosulphate, 200 mg NAD in 40 ml water) were added. The reagents could be mixed prior to dispensing into microplates without altering the results. Reactions were initiated by adding 1 µl diluted sample (or standards), incubating for > 15 min at room temperature (∼25 °C) and 50 µl 1 M HCl was added. A590 was measured in a microplate reader and ethanol concentrations were estimated by interpolation of a standard curve.
YPD plates inoculated with the CDL strains by using a multiprong replicator were grown at 4 °C, 10 °C, 15 °C, 28 °C, 37 °C and 42 °C for up to 2 weeks to determine the survival temperature extremes. Microtitre dishes with 200 µl/well YPD were then inoculated by using a multipet (5 µl/well) from overnight cultures of the CDL grown in YPD, and then incubated for 18 h at various temperatures.
Construction of the yeast diversity library
Selection of strains
We souhgt to breed new S. cerevisiae ('yeast') varieties with superior properties for wine and ethanol production. To this end, we made a collection of well-characterized strains, a majority of which have been sequenced to ≥ 1× by the SGRP (Liti et al., 2009), that may prove useful for breeding and valuable for the study of the population genetics and biology of this important, ubiquitous species. We confirm here that yeast homothallic genetics is fast and facile, even without employing recombinant DNA technology, and with or without introduction of selective genetic markers.
We examined 35 SGRP strains for their ability to: (a) grow on yeast minimal medium (SD); and (b) form numerous four-spored asci on sporulation medium (SM). All but W303 grew on SD, but only about half of the strains sporulated reproducibly within 1 week. We selected 14 strains for the collection for their diversity, as inferred from DNA sequences (Liti et al., 2009; sanger.ac.uk/about/press/2009/090211.html; sanger.ac.uk/research/projects/genomeinformatics/browser.html). Two additional, non-sequenced strains (F15 and BA11) were chosen under the same criteria and in addition for their importance as wine yeasts, a total of 16 strains. These were assigned the accession numbers given in Table 1 to simplify the nomenclature of derivative strains.
Formation of F1 hybrids and reconstructed parents
We mated the 16 strains described above in all combinations to facilitate the study of genetic interactions between divergent wild-types. Yeast strains for use in winemaking in New Zealand and elsewhere cannot contain recombinant DNA or be derived from recombinant lines, even lines where precise gene replacements have been made that leave no residual foreign DNA. Thus, plasmid-mediated disruption of MAT to inactivate mating type switching and produce a and α haploids was not a choice for this study. Unselected matings done by mixing ascospores and germinating them in constrained conditions, for example in dried spots on the surface of agar medium or as pellets submerged in growth medium (see Materials and methods), are reasonably efficient (∼15% crossed with two strains as tested with unselected markers; unpublished results). Nevertheless, use of markers simplifies and speeds up selection of hybrids in crosses between homothallic strains. We therefore decided to introduce ura− and lys− mutations into each of the chosen strains, because null mutations in URA3 or LYS2/5 can be selected for on media toxic to wild-type strains (FOA and AAD, respectively; Boeke et al., 1984; Chattoo et al., 1979). Subsequent reversion at these loci can be selected for on minimal medium if the auxotrophies become undesirable.
We obtained lysine- and uracil-requiring strains by selection of EMS-mutagenized cells on AAD or FOA medium. EMS-mutagenized diploids were made homozygous by germination of ascospores in liquid culture under conditions favouring selfing after mating-type switching. Colonies were tested for the desired auxotrophy and an acceptable reversion frequency, sporulated, and single spores selected from dissected tetrads. We gave the accession numbers of lysine-requiring strains odd suffixes (01, 03, etc.) and uracil-requiring strains even suffixes (02, 04, etc.), e.g. UA2401 (lys−) and UA4104 (ura−) (Table 2, plate C, columns 9–12).
We isolated ascospores from each marked strain and mated them in all combinations under conditions that favour heterologous interactions between cells having different mating types and nutritional requirements (see Materials and methods). We obtained crossed strains in all cases, as shown by production of prototrophic colonies and verified by microsatellite analysis, and assigned them accession numbers based on the lys− parent (first two digits) and the ura− parent (last two digits). We refer to the crossed strains as ‘F1 hybrids,’ or ‘F1s’. We refer to the selfed strains as ‘reconstructed parents’ or ‘RPs’. The RPs form an axis of symmetry in the library, as shown in Table 2. Reciprocal strains across the boundary of symmetry (e.g. UA210 UA111; UA2710 UA120) should be identical except for incidental mutations introduced prior to selection and perhaps divergent mitochondrial genomes. This collection of F1 hybrids and RPs, which we refer to as the ‘Yeast Diversity Library’ (YDL), is displayed in the 96-well microplate format shown in Table 2. Copies are available by request to email@example.com, and the collection has been deposited with ATCC.
In many laboratories, utilization of the YDL might present a problem of numbers. For example, a single experiment involving five treatments in triplicate would involve 45 plates (3 × 5 × 3). Without some automation this could represent a significant challenge in sample handling and data acquisition and analysis. It would also be expensive. For this reason, we developed a smaller collection derived from the YDL, which we refer to as the ‘Compact Diversity Library’ (CDL).
The composition of the CDL is shown in Table 3. It contains 94 hybrids along with the appropriate RP controls. The strains were chosen as follows: Two strains were included because of their importance in winemaking (F15/UA100 and BA11/UA4100), although their genomes have not been sequenced; two were selected because they are common laboratory strains (SK1/UA2400 and S288C/UA4200); and the remaining five were chosen based on their varying genetic distances (inferred from DNA sequence differences) from SK1 and S288C (Table 1; Liti et al., 2009; sanger.ac.uk/research/projects/genomeinformatics/browser.html). Thus, there is complete representation of nine strains, forming a 9 × 9 array. Additional isolates derived from S288C and SK1 were included to increase diversity and fill out the microtitre plate. Our intention is that initial results from the CDL will suggest ways to use the YDL to advance experimentation.
Table 3. Compact diversity library (CDL)
The organization of the collection is shown in the lower part of the figure and its representation within the YDL (see Figure 1) in the upper part of the figure (greyed cells). The dashed line represents the plane of symmetry and identifies the eight reconstructed parents present in the library. Strain UA4232 (UA4201 × UA4202) was not obtained, so its well is empty (marked N.D., not determined).
Preliminary characterization of strains: F2 viability
Although construction of the YDL showed that diploid hybrids could be formed between all the starting strains, the effects of heterozygosity on meiosis and ascospore formation were unknown. The fraction of viable spores produced by a strain and the number of asci with four viable spores is sensitive to genetic factors such as chromosomal rearrangements and masked or suppressed lethal mutations. These might be common in wild strains. We therefore selected 10 parental strains (CDL + UA1000/L1528; Tables 1 and 3) and their hybrids and selfs, sporulated each strain, and dissected tetrads (five or more) from each. Table 4 shows the results and the genetic distances between parents, expressed as polymorphisms/104 bp (from SGRP). Figure 1 is a summary. The mean genetic distance between strains was 53.8 polymorphisms/104 bp. The distances formed an approximately Normal distribution with a high of 76 and a low of 8.8 (Figure 1, centre panel).
Table 4. F1 fertility
The UA strains shown were crossed in all combinations, the F1s sporulated and five or more tetrads were dissected and tested for viability on YPD. The number of asci in each category (zero to four viable spores) is at the top of each cell. The colour-coded quadrant reports the total fraction of viable ascospores (top) and the fraction of asci with four viable spores (bottom). The lower right quadrant gives the strain designation from Table 2 (upper number) and the genetic distance between the two parents inferred from DNA sequence differences (polymorphisms/104 BP, lower number). N.K., not known. A legend summarizing the organization of the information is given at the bottom of the table.
The fraction of viable spores (YPD colony forming units/number of spores) for crosses varied from 0 (e.g. UA2901/UWOPS 05 227 2 × UA2702/UWOPS 87 242 1) to 1 (e.g. UA4201/ S288C × UA102/L1374). However, the fraction of viable spores for selfed strains (RPs) varied from 0.89 to 1, similar to results obtained from dissection of asci of autodiploids. The distribution of strains vs. viability (Figure 1, right panel) was not correlated with the distribution of polymorphic differences.
The fraction of asci with four viable spores also varied from 0 (e.g. UA3402/UWOPS 83 787 3 × UA2101/DBVPG6040) to 1 (e.g. UA4104/BA11 × UA2101/DBVPG6040), with the RPs in the range 0.6–1. The distribution of strains that produced four viable asci (Figure 1, left panel) is an approximate inverse of the viable spore distribution and was not correlated with the distribution of polymorphisms.
Although most strains produced viable ascospores and at least some four-spored asci, there were two exceptions, apparent in Table 4: UA2900/ UWOPS 05 227 2 and UA3400/UWOPS83 787 3. UA2900 as either the ura− or lys− parent produced no four-spored asci except when selfed, in which case it produced 7/7 asci with four viable spores. The average ascospore viability, excluding the self (UA2922), was 0.09. UA3400 showed a similar, although less extreme, phenotype. The two UA3400 × UA4100 hybrids each produced a single ascus with four viable spores (UA4125 1/5, UA3431 1/10), while 50–60% of spores were viable. One UA2700 hybrid (UA2725) also produced one ascus with four viable spores but its congener did not. Whatever the cause(s) of this reduced fertility (e.g. reciprocal translocations) it appears to be different in the two strains, because their hybrids (UA2925 and UA3422) produced no four-spored asci, although UA3422 did produce 40% viable spores. The genetic distance between the two strains was close to the average for the collection (64.5 polymorphisms/104 bp; Table 4).
The results show that the YDL and CDL provide access to a virtually limitless number of F2 recombinants for use in experimentation and breeding.
To be most useful for analysis and breeding, the collection should display greater variation in parameters of growth and development than that of the parents. To determine whether this was true, we examined three parameters of relevance to yeast biology and its deployment in industry: alcohol production, alcohol tolerance, and growth at different temperatures. Figure 2 shows the production of ethanol by the CDL strains grown in 1 ml YPD in deep-well microplates without agitation (anaerobic conditions). The results illustrate an advantage of the CDL for experimental analysis: eight RPs and 86 hybrids can be tested simultaneously in one plate, along with two controls (wells 11A and 11B). The RPs and hybrids had almost identical mean levels of alcohol production and overlapping standard deviations (SDs). The highest-producing strains (RP = 0.69, F1 = 0.65) and lowest-producing strains (RP = 0.15, F1 = 0.10) had similar ethanol levels. Thus, this experiment provides no evidence for enhanced variation among the F1 hybrids as compared to the RP controls, although there was substantial variation in both populations.
Yeast varieties can grow in high concentrations of ethanol (5% to > 20%), and this is an important property for alcohol production. We therefore compared growth of the CDL strains in YPD medium lacking or containing 9% ethanol, a concentration we had determined inhibited growth of the strains in the collection by a moderate amount. Figure 3 shows the results from growth at 28 °C for 20 h. In the absence of ethanol, over half of the strains had approached stationary phase while a few were still in mid- or early-log phase. The F1s and RPs had almost identical means although the SD of the larger, F1 population was somewhat greater than that of the RPs. In the presence of 9% ethanol, the mean for the RPs was reduced to 0.49 from 0.75 without alcohol. The mean for the F1s was also reduced, but only to 0.60. Nevertheless, the difference between the RP and F1 means was not significant [t(8) = 2.56, p < 0.05].
We examined variability further by growing the CDL in YPD at different temperatures or ethanol concentrations (Figure 4). Variation as measured by SD was maximal at 10% ethanol and decreased to almost zero at 16% ethanol, where the cells barely grew. The RPs (white dots) showed variation similar to the F1s (black bars). There was little variation between RPs and F1s in the range 15–37 °C. Again, the growth values for the RPs tracked those for the F1s.
The temperature and ethanol growth results are summarized in Figure 5. Growth in increasing concentrations of ethanol decreased to zero at 18%. The SD also decreased somewhat, as expected, although when normalized to the mean it reached a high of ∼1 at 14% ethanol. The low value reached zero at 14% ethanol, whereas the high value did not do so until 18%. At 10% ethanol the ratio of high to low was ∼5 but reached much greater values at higher ethanol concentrations, due at least in part to the divisor effect.
Growth increased from 10 °C to 20 °C and then decreased at 28 °C and 37 °C. The low value paralleled the mean, whereas the high valued reached a maximum at 20 °C and then remained about constant. The ratio of high to low varied from ∼5 at 10 °C to ∼2 at 20 °C, whereas the SD was almost constant, and the SD/mean had a maximum of 2.3 at 10 °C, a high value at least partially attributable to the divisor effect. The identity of the high and low strains was not the same at different ethanol concentrations or temperatures, so it appears that selection of strains for breeding or analysis should be based on replicated results from the specific conditions of interest and not inferred from other related experiments.
A useful property of the YDL and CDL is that they are symmetrical, containing two F1s from each parental pair (e.g. UA2110 = UA117; UA3419 = UA2425), and familial, with one common parent for each column and row. Thus, hereditary relationships should be observable in symmetrical and familial patterns. To test this, we pooled temperature and ethanol growth data, grouped them by columns and rows (9 × 9 for CDL), calculated means, and analysed them graphically in Figure 6. The scatter plot shows that there is close correlation between the mean value of rows and columns (r = 0.98, SE = 0.06, m = 1.00, b = 0.00). This feature opens up interesting opportunities for experimental design and data analysis.
Heterosis and overdominance
We are interested in the possibility of exploiting F1 overdominance and heterosis for yeast breeding. In plant breeding, detection of heterosis may entail numerous replicated experiments under a variety of conditions to separate genetic (G) and environmental (E) effects, and this could be true for yeast as well, especially where E is poorly controlled, for example in wineries. Although heterosis was reported in yeast many years ago (Lindegren et al., 1953), little is known about how it might be used in structured breeding programmes. As a first step toward determining what might be required to identify and utilize heterosis in yeast, we grew the CDL strains in YPD medium containing 9% ethanol (replicated three times), tabulating the growth results (OD650) by RPs and F1s, as shown here:
and so on.
The values for the parents are taken as those for the selfed strains (i.e. UA100 = UA110 and so forth). To present the data in two dimensions, we used a standard mathematical approach: the OD values were normalized to those of Parent 1 by division, making all the Parent 1 values = 1. We then plotted ODs for Parent 2 (abscissa) vs. ODs for the F1s (ordinate) in Figure 7. Here, the horizontal line at 1.0 is the value for the normalized parent, whereas the diagonal line is Parent 2 vs. itself. Therefore, the white areas represent regions between the values for the two parents, whereas the grey areas represent regions either higher or lower than both parents, regions that represent areas where overdominance or heterosis may be affecting the F1s. The data points are distributed in a way consistent with the one-dimensional data shown in Figure 3, but show that many individual F1 strains grow under stress more than three SDs from the mean. To examine the significance of the positions of points within the graph, we tested strains for growth at four different concentrations of ethanol. Representative results are shown in the inset graphs. Strain UA2119 falls on the Parent 2 line and, as shown in the inset, has a tolerance curve similar to its ‘parents’, RPs UA2117 and UA2419. Similarly, UA4119, which falls in the upper white region, behaves much like its RPs. By contrast, UA2919, which occurs in the upper grey region, is significantly more tolerant of ethanol than either RP, consistent with heterosis. Although most of the data points were verified by this extended analysis, UA2719 illustrates the importance of further confirmation. Originally, in 9% ethanol it performed worse than its RPs. However, as shown in the inset, when tested at different ethanol concentrations it is halfway between the RPs at 9% ethanol. At higher concentrations of ethanol it grows more than either parent. Thus, this type of analysis provides an initial screen for candidates under one condition, but further analysis should be incorporated to provide a complete picture, as is generally the case with moderate to high throughput screening.
We analysed these data in another way to determine whether there were patterns of heterotic interactions, analogous to heterotic groups in maize. First, we defined a function for the coefficient of heterotic effect (CH) as shown in Figure 8. The function adds the differences between the two parents and the F1 hybrid in the numerator as an absolute value, because the function is agnostic with respect to the value of the trait, i.e. whether or not it is advantageous to have more robust growth in the presence of 9% ethanol. The denominator then normalizes the function to the differences between the two parents, again as agnostic absolute values. The behaviour of the function is shown in Figure 8 for two parents with arbitrary performance values of 5 and 10 and F1 hybrids with performance values of 1–50. F1 values falling between the parents (additive effect) are < 0.5, whereas those either below or above both parents (heterosis or overdominance effect) are 0.5–1.0. As is evident in Figure 8, the function is more sensitive to small changes than large ones, so that values > 0.9 represent a major departure from additivity.
Figure 9 shows the results for calculating the heterosis coefficients for the hybrids from the CDL (wells 1A–8H) from the data shown in Figure 7 by averaging the data from congeners. The UA4200/S288C series, for which there is no RP, was not included in the analysis. The order of the strains on both axes has been changed to cluster the data within the ranges given in the key to the figure. Six of the F1s showed only an additive effect (white cells), while eight showed strong overdominance (black cells). Four strains (UA2117, 2720, 2922 and 3929) showed a strong heterotic effect (>0.8) when mated with one another. These results do not reflect whether the heterotic effect was positive or negative, are for one condition (9% ethanol), and may not be predictive for other conditions.
The strains derived from the dissected ascospores reported in Table 4 represent another potential resource for variation, analysis, and breeding. We tested a sample of these and found that sometimes the lys− and ura− traits showed Mendelian behaviour (although often not, presumably because of problems of chromosome alignment and segregation). This suggests that at least some traits, wild or introduced, will be amenable to genetic manipulation. However, the behaviour of probable quantitative traits such as growth at low or high temperatures and alcohol tolerance through meiosis may be much more complex and difficult to exploit than simple Mendelian traits. We have begun to examine some of these derived strains (F2s) for traits relevant to winemaking. One result is shown in Table 5. The F1 formed between UA102 and UA2101 (UA2110) grew at a rate at 10 °C intermediate to the RPs. By contrast, all 16 of the F2s (homozygous diploids) exhibited less growth, with a maximum OD650 of 0.21 and an average of 0.11 ± 0.04, less than half of the hybrid and below the low RP. The values for the F2s are not distributed into discrete groups, consistent with quantitative inheritance. The results suggest that ability to grow at low temperature is a quantitative trait and is optimized in the wild-type parents. In this model, recombination would uncouple beneficial gene interactions, leading to poorer growth in the offspring.
Table 5. Effect of low temperature and ethanol on F2 strains
UA2110 (UA2101 × UA102) and its RPs were grown either at 10 °C (upper panel) or in 9% ethanol (lower panel). UA2110 was sporulated, and four tetrads (1–4, spores A–D) were also tested for growth, yielding the figures shown. SDs for 16 determinations are shown
UA2110 also grew intermediate to the RPs in the presence of 9% ethanol, as did many of the F2s, although about half grew less than UA2117, the low RP. Spore 2B was exceptional in this experiment (but unexceptional at 10 °C), in that its growth was similar to the high RP, UA110. The average for the entire sample was 0.27 ± 0.21. The large SD was indicative of the high variability (0.09–0.79). The sample distribution was disperse, but might be interpreted as bimodal, with populations separated between 0.3 and 0.4. However, this pattern is not replicated in tetrads, so its significance, if any, is uncertain. As with low temperature, most of the F2s were less robust than the parents and the F1 hybrid, consistent with segregation of additive quantitative traits that are closer to optimal in the parents and F1 hybrid.
The resequencing of S. cerevisiae wild-type strains (Liti et al., 2009) opens up opportunities for the study of the genetics, physiology and population biology of this species, which serves both as the premier eukaryotic model organism (Botstein et al., 1997; Moler et al., 2000; Sherman, 1997; Zeyl, 2000; Hodges et al., 1999) and one of the most important industrial microorganisms (Higgins et al., 2003; Bisson et al., 2002, 2007). We have taken advantage of the sequenced strains to make a collection of F1 hybrids that has many possible uses, e.g. selection of strains with enhanced properties, study of gene interactions, elucidation of the genetic basis of heterosis and overdominance, and production of an almost limitless number of recombinant F2 strains, representing much of the known genetic diversity of S. cerevisiae, for experimentation and breeding. The Yeast Diversity Library contains 250 hybrids and 15 selfed, a/α diploids that are congenic with the lys− and ura− strains used for construction of the library, referred to as ‘reconstructed parents’ (RPs), that can serve as controls for the hybrids. The genomes of 13 of the 15 RPs have been sequenced as have both genomes in 182 hybrids. Two hybrids are unsequenced, and in 56 one genome is sequenced and the other not. We encountered no difficulty in making any of the hybrids and RPs, except for UA4232, which would be the a/α diploid of UA4200 (S288C), a laboratory standard haploid. We attempted this by incubating lys− and ura− cells together, as in our standard mating procedure, and plating them under selective conditions, without success. If it becomes critical, construction of the strain could be accomplished by using recombinant DNA techniques where that is not a consideration, or as strain X2180, a spontaneous diploid of S288C, or by reconstitution from derivative strains X2180-1A and -1B (Mortimer and Johnston, 1986). The observation that we could form so many hybrids without difficulty indicates that most wild strains can be treated similarly with expectation of success.
Fungal genetics typically exploits heterothallic species, e.g. Neurospora crassa (Herskowitz, 1988; Davis, 2000; Metzenberg and Glass, 1990), Podospora anserina (Zickler et al., 1995; Coppin et al., 1997; Raju and Perkins, 1994), Schizosaccharomyces pombe (Herskowitz, 1988; Kelly et al., 1988) and Ustilago maydis (Kahmann et al., 1995; Martinez-Espinosa et al., 1993), so that mating between individuals can be controlled. Development of the genetic system of S. cerevisiae has largely depended on the inactivation of HO so that strains are functionally heterothallic, i.e. unable to switch mating types. However, the genetics of Aspergillus (Emericella) nidulans, a homothallic species, has never depended on having separate mating types (Timberlake and Marshall, 1989), demonstrating that heterothallism per se is not a prerequisite for development of a sophisticated genetic system. Here we confirm that yeast is well suited for homothallic genetics based on the formation of ura− and lys− mutants, which can be selected easily and directly, as can their revertants. Mutagenesis requires an extra step (sporulation) as compared to haploid strains, to make the mutations homozygous prior to selection. However, one mutagenesis produces enough ascospores for numerous selections over many years. Mating, most often carried out by mixing a and α haploid cells, can also be done by mixing ascospores from strains containing complementing mutations. With mating-type switching, all combinations of mating occur, but only diploids with complementing mutations grow on minimal medium. Co-pelleting the ascospores in growth medium is an easy and efficient way to obtain hybrids. Crossing can also be done without selection, but at lower frequency, and more effort is needed to identify hybrids by microsatellite analysis (Bradbury et al., 2005; Legendre et al., 2007; Richards et al., 2009).
Meiosis and spore formation might be more challenging than hybrid formation for yeast, because even benign chromosomal arrangements in one strain can have catastrophic effects on hybrid meiosis. Other types of lethality can also be envisaged, e.g. lethal genes with linked suppressers that become uncoupled during heterozygous recombination. The data presented here show that there are indeed barriers to completion of the sexual cycle. The RPs had fertilities similar to those found in standard laboratory diploids. A few of the hybrids had full or high fertility (e.g. UA4210, UA2417), whereas many others had reduced spore viability and few or no asci with four viable spores. Two strains, UA2900 and UA3400, showed extreme meiotic incompatibility with all other strains, including each other. The basis for this incompatibility is unknown but was not reflected in hybrid formation, where both strains were fully xenogamous. These two isolates aside, with the majority of strains it was possible to obtain asci with viable tetrads and useful numbers of random ascospores for analysis and strain construction. Thus, the basic mechanics for genetic manipulation of wild strains are in place, without dependence on recombinant DNA technology, which is precluded from many breeding projects.
Variation, of course, is another requirement of a useful genetic system. On the one hand, the wild-type strains used in this study, having been selected for their ability to grow in the laboratory on minimal medium, are expected to have reduced variation. On the other hand, traits important to the organism in nature may be neutral in the laboratory and thus retained, to be revealed under unusual conditions of growth or development. We made a preliminary analysis of three traits that may be important in nature and are critical to the human use of yeast: alcohol production, alcohol tolerance, and temperature tolerance. Our purpose was to examine variation among the RPs and hybrids and to determine whether hybridization enhanced or depressed variation. Further, we wished to see whether hybrid-induced variation might be exploited in breeding programmes.
We measured ethanol production by the Compact Diversity Library (CDL) strains under laboratory conditions (partial anaerobiosis; 2% glucose; 28 °C). The hybrid and RP populations had essentially the same mean alcohol concentration and variance. The same was true when the cells were grown for 20 h with agitation, but the populations showed some divergence when challenged with 9% ethanol. Here, although the nominal means differ, the hybrid population exhibited no more variation than the RP population. Similar results for different challenges (ethanol or temperature) led us to conclude that hybridization does not enhance variation, at least for this set of strains and the measurements we made.
This conclusion relates to the population, not to individual hybrids, whose traits might differ significantly from the parents. For quantitative traits in particular, detecting differences in individuals could be challenging, due to experimental variation. As a first step in examining individual variation, we chose strains that appeared to resemble their RPs or differ from them significantly with respect to growth in 9% ethanol. Replicates of each strain were then tested for growth at various concentrations of ethanol. Three of the four strains tested gave results consistent with their position in Figure 7, whereas one, a strain > 3 SDs below the mean, did not. Thus, this approach (paired RPs vs. their heterozygotes) shows promise for identifying significant variation in individuals.
Another possible source of variation is in F2s (and beyond), because the outcomes of recombination and assortment are displayed in the homozygous diploids that form from individual ascospores as a consequence of mating-type switching and mating. In addition, the power of tetrad analysis and the existence of F2 tetrads for most of the CDL strains suggest that this could be a particularly useful analytical tool for the collection. Table 5 presents the results for four tetrads from a single cross, with all the relevant strains being grown either at 10 °C or in the presence of 9% ethanol. In each case the growth of the hybrid was intermediate to that of the RPs. However, in the case of temperature challenge, all the F2s showed less growth than either RP or the hybrid, perhaps due to homogenization of an optimized quantitative trait. The results for 9% ethanol were similar, except that several offspring grew more than the low RP and an exceptional individual grew about the same as the high RP. Further studies will be required to unravel the genetics of this and other crosses with the wild strains.
A feature of the YDL and CDL is that each hybrid is represented twice: once as lys−/+× ura− and once as + /ura−× lys−/+. Although not identical, the paired strains should be very similar and thus serve as reciprocal controls. Further, the collections are organized into families that are replicated in columns and rows. The means of columns and rows are highly correlated when the cells are grown under a variety of temperatures or ethanol concentrations, showing the familial effect. Thus, these collections facilitate a number of experimental and analytical approaches, have been subjected to quality controls, and are formatted for robotic deployment. The open format of the collection may facilitate the development of shared databases that accelerate research.
Gianni Liti and Ed Louis, University of Nottingham, supplied strains in advance of publication. This work was supported by the New Zealand Foundation for Research Science and Technology (Grant No. UOAX0404), the University of Auckland, and New Zealand Winegrowers (Grant No. 07-301).