Chromosome Transfers

I transferred Chromosomes I, II, III, VI, VIII, IX, X, XI, and XV from S. paradoxus gametes into S. cerevisiae gametes (see Materials and Methods and Figure 1). Candidate strains with the correct combination of genetic markers were initially karyotyped by PCR with species-specific primers for all possible chromosome ends from both species (64 in total; see Materials and Methods). I found that for two of the targeted chromosomes (I and IX), the first candidate strains tested were correct, whereas for other targeted chromosomes, many strains had to be screened before one with the correct karyotype was identified. The number of strains that had to be screened before the correct karyotype was found is shown in parentheses after the target chromosome number: I (one), II (18), III (eight), VI (seven), VIII (two), IX (one), X (nine), XI (four), and XV (five). The cause for rejection of all the failed candidate strains was that the PCR markers showed that other S. paradoxus chromosomes were transferred along with the targeted S. paradoxus chromosome. A linear regression found no significant relationship between the physical size of the target chromosome and the number of candidate strains that needed to be screened (regression on log-transformed values, r² = 0.14, 8 df, p = 0.32). Three chromosomes were targeted, but to date no correct transfers have been identified. These chromosomes (with the number of candidates screened in parentheses) are IV (six), XIII (eight), and XIV (11). The other four chromosomes were not been targeted either because they contain genes or markers used for the transfer method (V, ura3, can1; VII cyh2) or because no suitable auxotrophic marker has been identified in which to insert URA3 (XII, XVI).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. CHEF Gel Showing S. paradoxus Chromosome Transfers and Replacements in S. cerevisiae Gametes

Negative digital image of a CHEF gel. Replaced chromosomes are indicated by an asterisk to the left of the S. paradoxus band. S. paradoxus Chromosomes I, III, VI. and VIII are a different size than their S. cerevisiae homeologs and can be readily identified. The remaining transfers were verified by quantitative image analysis of the high-intensity band produced by the disomic chromosome (see Materials and Methods). Lanes: P, S. paradoxus YDG749; C, S. cerevisiae YDG 780; It, S. cerevisiae haploid with additional (disomic) Chromosome I transferred from S. paradoxus; Ir, S. cerevisiae haploid with Chromosome I replaced by Chromosome I from S. paradoxus; etc.

https://doi.org/10.1371/journal.pgen.0030021.g001

Chromosome Replacements

S. paradoxus Chromosome VIII spontaneously replaced its S. cerevisiae homeolog before selection was applied with the drug 5-fluororotic acid (FOA). The other eight S. paradoxus chromosomes coexisted as disomes with their homeologs until the strains were replica-plated to FOA agar, which selected for the loss of the URA3-marked native chromosome (see Materials and Methods). All then produced FOA-resistant colonies that contained 15 S. cerevisiae chromosomes and one S. paradoxus chromosome, as confirmed by using species-specific PCR markers for the chromosome ends (see Materials and Methods). The rate of FOA-resistant colony formation varied greatly, presumably inversely reflecting the ability of the URA3-marked S. cerevisiae chromosome to segregate mitotically as a disome with its S. paradoxus homeolog. In order to quantify the rate, I counted the number of FOA-resistant colonies formed in a 10% sample of each plate's area. The number of colonies formed is given in parentheses after the disomic chromosome number: I (309), II (eight), III (27), VI (176), IX (128), X (282), XI (one), and XV (347). Again, the size of the disomic chromosome was not an accurate predictor of the rate of FOA-resistant colony formation (linear regression on log-transformed values: r² = 0.047, 7 df, p = 0.61).

Confirmation of Transfer and Replacement of Targeted Chromosomes by Pulsed Field Gel Electrophoresis

As described above, many candidate strains were rejected after PCR analysis of their karyotypes because they contained additional chromosomes to those targeted. Therefore, another, independent, karyotyping method was desirable as a quality control to confirm the PCR results. Pulsed field gel electrophoresis is a robust method that allows separation of the chromosomes on a gel and direct visualization of a strain's karyotype. Figure 1 shows a typical gel containing all the chromosome transfer and replacement strains. S. paradoxus Chromosomes I, III, VI, and VIII migrate in the gel at different rates to their S. cerevisiae homeologs and can be readily identified by the different bands they form. Thus, confirmation of the successful transfer and replacement of these targeted S. paradoxus chromosomes is easily verified (see Figure 1). S. paradoxus Chromosomes II, IX, X, XI, and XV migrate to the same band in the gel as their S. cerevisiae homeologs, so a different method is required to identify them on the gel. Transfer strains that are disomic for both species' Chromosome II, IX, X, or XI will have bands that contain twice the DNA and hence are twice as intense as those in normal S. cerevisiae. Chromosome XV migrates to the same band as VII, so a strain disomic for both species' XV chromosomes will have an XV/VII band that is 50% more intense than in normal S. cerevisiae. Many of these intensity differences can be detected by the naked eye ([16] and see Figure 1), but to obtain quantitative measurement of the DNA content of bands, at least four independent preparations of each strain were made, run out independently on separate gels, and analyzed using image analysis software (see Materials and Methods for details). The results of this analysis are summarized in Figure 2, and the raw images and image analysis data are presented in Figures S1–S5 and Table S1. The mean standardized intensities for normal S. cerevisiae chromosome bands II, IX, X, XI, and XV were compared to those in their respective transfer and replacement strains by one-way analysis of variance and post hoc Tukey test (see Materials and Methods). In each case, the band for the targeted transferred chromosome was significantly more intense than the same band in normal S. cerevisiae, showing that it was disomic and transfer had been successful (p < 0.001, n = 4 for Chromosomes II, IX, X, and XI; p < 0.05, n = 10, for Chromosome XV). The XV/VII band for the XV transfer strain was not found to be significantly more intense when only four preparations were analyzed because its intensity is expected to increase by only 50%, rather than by 100% for the others. Thus, six more preparations were made and analyzed to increase the statistical power of the test. The II, IX, X, and XI bands in the (FOA-resistant) chromosome replacement strains were significantly less intense than the band in each transfer strain, showing that the native chromosome had been lost and successfully replaced by the S. paradoxus chromosome. The XV chromosome replacement band was less intense than that in the XV chromosome transfer strain (Figure 2), but it was not statistically significant. However, given that the XV transfer strain formed FOA-resistant colonies at a higher rate than any other strain (see Chromosome Replacements, above) and that in all other transfer strains the first FOA-resistant colony tested was found to have lost the targeted native S. cerevisiae chromosome, it would be very unlikely that the XV replacement strain actually remained disomic. Rather, the lack of significant difference between the XV/VII doublets in the transfer and replacement strains probably reflects the lack of power of the test to detect the relatively small difference in intensity expected in the XV/VII band (see above).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Results of Pulsed Field Gel Image Analysis

The graph shows the mean standardized band intensities (a measure of the fraction of total cell DNA) for the five targeted S. paradoxus chromosomes that comigrate to the same bands as their S. cerevisiae homeologs. The error bars represent 95% confidence intervals. Four independent replicates were produced for Chromosomes II, IX, X, and XI. Ten independent replicates were produced for Chromosome XV. Raw images of all the gels as well as a table of all the image data are included in Figures S1–S5 and Table S1.

https://doi.org/10.1371/journal.pgen.0030021.g002

Test for Rogue Chromosome Transfer by Pulsed Field Gel Electrophoresis

I used the same pulsed field gels generated above to screen the transfer strains for additional (rogue) S. paradoxus chromosomes that might have been inadvertently transferred along with the targeted chromosome and also escaped the PCR screen. S. paradoxus I, VI, III, and VIII are readily identified by their size difference: none of these chromosomes were found (except in the strains in which each was targeted). All the other chromosomes except IV/XII were tested (described in Materials and Methods) to identify bands that were significantly more intense than in S. cerevisiae and that might therefore be disomic. For Chromosomes II, IX, X, XI, and XV, the brightest band identified was that of the targeted chromosomes, showing that the test successfully identifies known disomes (a positive control for the screen). For these chromosomes, the next brightest band was then identified for each strain, and the mean of four replicates was compared to the mean of the band in the four S. cerevisiae replicates by one-way analysis of variance. In no case was the brightest band significantly brighter than in normal S. cerevisiae (analysis not shown; all intensity data are presented as supplementary material). For Chromosomes V, XIV, XIII, XVI, and VII, in which no targeted transfers were successful, the mean of the brightest band was not significantly different from the mean of the band in S. cerevisiae. It is possible, but unlikely, that rogue disomes exist that escaped the PCR screen and also escaped the image analysis screen, either because their bands could not be quantified (IV and XII) or because they are in bands of two comigrating chromosomes and a third rogue does not increase the brightness sufficiently to be statistically significant with only four replicate samples (V, XIII, XVI, VII, and XV). However, the reliability of the PCR karyotype assay is ensured by the fact that the image intensity screen found no rogue chromosomes in the nine that can be identified with confidence.

Estimate of Total Number of Recessive Speciation Genes Causing F1 Gamete Inviability

Assuming that the chromosomes were sampled randomly with respect to the presence or absence of incompatible genes, and if loci that have incompatible alleles are randomly distributed across the genome, then we can calculate the maximum likely number of gamete-killing speciation genes on the unsampled seven chromosomes. If there was a single speciation gene, it would more likely than not avoid detection, because I sampled 43% of the genome, so it would have a probability of 0.57 (1 − 0.43) of falling on a chromosome that was not tested. But the probability of six genes that are capable of killing hybrid gametes being missed falls below the 5% boundary for statistical significance [p = (1 − 0.43)⁶ = 0.035], so we can be confident that there are between zero and five S. paradoxus genes that are incompatible with genes on the S. cerevisiae genome in such a way that they kill hybrid gametes that contain them.

Contribution of Genes Causing Hybrid Gamete Inviability to F1 Hybrid Sterility

F1 hybrid sterility could have many possible causes, among them the gamete-killing recessive speciation genes screened for in this study. We can ask, therefore, if the sole cause of hybrid sterility was this type of genetic incompatibility, how many such genes would be required to reduce diploid F1 hybrid fertility to 1%? The simplest model for the evolution of genetic incompatibility [1,2] features two interacting genes, A and B, in two genetically identical geographically isolated populations. In one population, a new allele, a, is fixed, and in the other, a new allele, b. The evolved alleles a and b are incompatible with each other. A diploid F1 hybrid will have the genotype AaBb and will be viable and fertile if the incompatible alleles are recessive. Its gametes will have viable genotypes AB, Ab, and aB and the inviable genotype ab. So, a single recessive incompatibility can reduce fertility by no more than 25%, leaving 75% of gametes viable. To reduce fertility to 1%, the fertility of the yeast hybrid, a minimum of 16 incompatibilities are required [(1 − 0.25)¹⁶ = 1%]. This is a conservative estimate assuming the most powerful form of recessive speciation gene, a two-locus completely lethal incompatibility. If incompatibilities involve more than two loci, then they will affect fewer gametes (e.g., a three-locus incompatibility would only affect 12.5% of gametes), and if incompatibilities are not completely lethal, then more would be required to have a sterilizing effect. Five incompatibilities, the maximum that could reside on the unsampled chromosomes (see Estimate of Total Number of Recessive Speciation Genes Causing F1 Gamete Inviability, above) could only reduce F1 fertility to 24% (0.75⁵), more than 20 times higher than the actual F1 hybrid fertility of 1%. Thus, the sterility of yeast F1 hybrids cannot be explained solely by gamete-killing speciation genes.

An important assumption of this analysis is that the nine chromosomes tested represent a random sample of the genome. In addition to the nine chromosomes transferred, three chromosomes (IV, XIII, and XIV) were targeted but were not successfully transferred. This is not surprising: transferring single chromosomes by this method relies on rare random events—transfers of whole individual chromosomes between separate nuclei—that are very unlikely for large chromosomes like those unsuccessfully targeted [19]. Another possibility, though, is that these three chromosomes could not be transferred because they contain incompatible genes. Such an incompatible allele would need to be dominant (it would be heterozygous when transferred), but it would interact with recessives on the haploid genome. In organisms that have chromosomal sex determination, such as Drosophila, such dominant–recessive interactions are thought to cause Haldane's rule (that males are more affected than females by hybrid incompatibilities) [20]. X-linked genes are hemizygous in XY males, so recessive incompatibilities on the X chromosome interact with dominant incompatibilities elsewhere. Recessive–dominant interactions are a good explanation for Haldane's rule, but they are much less common than recessive–recessive incompatibilities [20,21]. So, we would not expect to find recessive–dominant incompatibilities without also finding recessive–recessive incompatibilities. Furthermore, recessive–dominant interactions cannot contribute to yeast speciation, because yeast lacks chromosomal sex determination and yeast cells are found only as either complete diploids or complete haploids. Thus, it is reasonable to assume that the nine chromosomes transferred are a random sample and that the three unsuccessful transfers failed by chance, rather than because they were prevented by recessive–dominant incompatibilities.

Gamete-Killing Incompatibilities That Would Not Be Detected by the Assay

Incompatibilities between loci on homeologs of the same chromosome would not be detected because whole chromosomes (containing both compatible alleles) were transferred. However, yeast has 16 chromosomes, so the chance that a given speciation gene shares the same chromosome as the locus with which it interacts is only 1:16, on average. Moreover, incompatible loci on homeologs of same chromosome could kill the gametes produced by F1 hybrids only if they were far enough apart to be unlinked (so that the ab inviable genotype could be produced in gametes), further reducing their ability to contribute to hybrid sterility. Also able to escape detection are incompatibilities that cannot kill gametes by themselves but are lethal when combined with other such incompatibilities. However, these weak speciation genes can kill only a small proportion of gametes produced from an F1 hybrid (e.g., if two incompatibilities are required to be present in a gamete to kill it, only 6.25% of gametes will inherit both, leaving 93.75% of gametes viable). Similarly, complex interactions can kill fewer hybrid gametes than simple two-locus interactions (e.g., a three-locus incompatibility would affect only 12.5% of gametes).

Other Recessive Speciation Genes That Would Not Be Detected by the Assay

Chromosomes were transferred between mature gametes, so incompatibilities that sterilize F1 hybrids by killing their gametes only before gamete maturation would escape detection. Such incompatibilities might exist in genes expressed specifically during spore germination or even in genes expressed specifically in the late stages of meiosis and spore production, because after meiosis I, many loci can be considered hemizygous and thus recessive incompatibility could be exposed. However, this group of genes is expected to be very small compared to all the genes required for vegetative growth, so there is no reason to expect to find many incompatibilities in germination if none are found in vegetative growth. Finally, genetic incompatibilities that reduce fitness but do not actually kill gametes can still contribute to speciation. F1 hybrid fertility would not be affected (because gametes would be viable, if unfit), but the low fitness of the resulting F2 hybrids could reduce their reproductive potential, contributing to reproductive isolation. Likewise, recessive incompatibilities that are only expressed in diploids, such as genes involved in meiosis, would not affect the fertility of F1 hybrids but, when homozygous in F2 hybrids, could render them sterile or inviable (hybrid breakdown).

Conclusions

These results show that recessive Dobzhansky–Muller speciation genes are unlikely to be responsible for killing the gametes produced by S. paradoxus–S. cerevisiae F1 hybrids. While some types of recessive speciation genes cannot be detected by the assay used here, these are expected to be either rare or to kill only a small proportion of the gametes produced by F1 hybrids or only affect F2 hybrids. Thus, recessive speciation genes, if they exist at all, are surprisingly rare and are unlikely to be the major reason that the gametes produced by F1 hybrids are inviable.

So why are yeast F1 hybrids sterile? The most likely explanation is that of Hunter et al. [16], who showed that deleting the mismatch repair gene MSH2 increased hybrid fertility 10-fold. They proposed that yeast hybrid sterility was caused by the failure of diverged homeologous chromosomes to recombine in meiosis, in part because of the action of the mismatch repair system. This failure leads to mis-segregation, lost chromosomes, and inviable aneuploid gametes [14,15,22]. The genomes of S. cerevisiae and S. paradoxus are about 15% diverged [23]. Recent work has shown that very low levels of divergence can also affect fertility [24,25]. Thus, it is likely that simple DNA sequence divergence is the major cause of sterility in F1 hybrids formed between S. cerevisiae and S. paradoxus.

Chromosome transfer and replacement.

The method was based on that of Chambers et al. [22] and Nilsson-Tillgren et al. [26] and uses a kar1-Δ13 mutation [27] that prevents chromosome fusion after mating but allows individual chromosome transfer between nuclei [19]. S. cerevisiae strain YDG 780 (alpha, ura3-d, cyh2r, can1r kar1-Δ13), which is isogenic with S288c, was used to generate strains with different nutritional auxotrophies marking specific target chromosomes (see Protocol S1). These strains were allowed to mate overnight on YEPD agar with S. paradoxus haploid YDG 749 (ho, MATa, ura3), which is isogenic with N17 [28]. The mated cells were replica-plated to SC agar containing cycloheximide (3 mg/L) and canavanine (60 mg/L), to select against S. paradoxus haploids and hybrid diploids, and lacking the specific nutrient needed because of the auxotrophic marker on the target chromosome, to select for complementation of that marker by transfer of the wild-type S. paradoxus chromosome. Strains identified as disomic for the target chromosome (see Results and Protocol S1) were replica-plated to FOA-agar selecting for the loss of the URA3-marked S. cerevisiae target chromosome.

Verification of chromosome transfer.

The complete chromosome complement of candidate strains was determined by PCR with species-specific primers for all possible chromosome ends from both species (64 in total; see Protocol S1 for primer sequences). Strains containing the correct chromosomes were then verified physically using pulsed field gel electrophoreses as described previously [16]. Transfer and replacement of Chromosomes I, III, VI, and VIII were confirmed visually; these chromosomes migrate differently on the gel according to which species they are from (see Figure 1 and Results). Transfer of the other five chromosomes was confirmed by quantitative measurement of the appropriate band intensity using the gel analysis plugin for the open source software ImageJ (Imaging Processing and Analysis in Java; http://rsb.info.nih.gov/ij). Bands of disomic chromosomes are much brighter than those of monosomic chromosomes. Four independent pulsed field gel preparations were made from each strain. Lanes were standardized by dividing the intensity of the target chromosome band by the sum of all the band intensities in the same lane except for the IV/XII band, which varies because of spontaneous changes in rDNA tandem repeat number. The mean of the four standardized intensities of the chromosome band of interest was then compared to the mean of the four standardized intensities of that band in S. cerevisiae. Six additional preparations were made of the XV transfer, replacement, and S. cerevisiae strains in order to increase the statistical power sufficiently to detect the relatively small difference in intensity between an XV/VII band containing three chromosomes (the transferred XV S. paradoxus chromosome and the normal XV/VII chromosomes) and an XV/VII band containing two chromosomes. Statistical comparisons were performed with Minitab 14 statistical software (http://www.minitab.com). The raw images and a table of image intensity data are provided in Figures S1–S5 and Table S1.