The genetic bases for species-specific traits are widely sought, but reliable experimental methods with which to identify functionally divergent genes are lacking. In the Saccharomyces genus, interspecies complementation tests can be used to evaluate functional conservation and divergence of biological pathways or networks. Silent information regulator (SIR) proteins in S. bayanus provide an ideal test case for this approach because they show remarkable divergence in sequence and paralog number from those found in the closely related S. cerevisiae. We identified genes required for silencing in S. bayanus using a genetic screen for silencing-defective mutants. Complementation tests in interspecies hybrids identified an evolutionarily conserved Sir-protein-based silencing machinery, as defined by two interspecies complementation groups (SIR2 and SIR3). However, recessive mutations in S. bayanus SIR4 isolated from this screen could not be complemented by S. cerevisiae SIR4, revealing species-specific functional divergence in the Sir4 protein despite conservation of the overall function of the Sir2/3/4 complex. A cladistic complementation series localized the occurrence of functional changes in SIR4 to the S. cerevisiae and S. paradoxus branches of the Saccharomyces phylogeny. Most of this functional divergence mapped to sequence changes in the Sir4 PAD. Finally, a hemizygosity modifier screen in the interspecies hybrids identified additional genes involved in S. bayanus silencing. Thus, interspecies complementation tests can be used to identify (1) mutations in genetically underexplored organisms, (2) loci that have functionally diverged between species, and (3) evolutionary events of functional consequence within a genus.
MUCH of our inferred knowledge about human biology depends upon gene function being conserved between we humans and various model organisms, with our distant cousin, the mouse, being of special importance. To what extent is sequence conservation a reliable indicator of functional conservation, and when can it lead us astray? Although most orthologous proteins >100 AA in length with >25% sequence identity have similar structures, orthologs with 95% sequence identity can have divergent functions (Sander and Schneider 1991; Kosloff and Kolodny 2008). Recent sequence comparison methods such as highly accurate alignment algorithms [Fast Statistical Alignment (FSA); Bradley et al. 2009], Bayesian estimation of evolutionary rates [Phylogenetic Analysis by Maximum Likelihood (PAML); Yang 2007], and synteny browsers [Yeast Gene Order Browser (YGOB); Byrne and Wolfe 2006] allow the reliable determination of orthologous gene pairs between species, yet few methods exist to empirically test whether any two orthologous genes in fact perform the equivalent function(s) in each species. The traditional genetic and biochemical approaches to evaluating the conservation question are laborious and are not well suited for genome-wide comparative studies.
The genetic complementation test involves crossing two mutant strains with the same phenotype to determine if two different recessive mutations are located in the same gene. A transgenic variation on the complementation test, as commonly used in yeast genetics, identifies the wild-type gene corresponding to a recessive mutation by transformation of the mutant with a recombinant library. Such complementation assays have occasionally been used to determine whether a gene’s function is conserved across a large evolutionary distance [for example, in cloning the human CDK1 gene by library transformation of a Schizosaccharomyces pombe cdc2 mutant (Lee and Nurse 1987)]. More recently, reciprocal interspecies complementation analysis, performed by comparing the phenotypes of two interspecies hybrids that each lack one or the other allele of a common locus, has been employed to identify functional divergence of orthologous genes (Lee et al. 2008; Gerke et al. 2009; Zill et al. 2010). Systematic interspecies complementation assays for even a handful of genera would provide useful calibration of the extent to which sequence conservation between homologous genes reflects their functional conservation in the context of a whole organism.
Recent comparative studies of orthologous mutants demonstrate that interspecies comparisons can more thoroughly delineate conserved pathways and reveal additional functions of orthologous genes (Zill and Rine 2008; Wood et al. 2011). However, more systematic genetic analysis of gene, pathway, or network evolution requires the ability to conduct interspecies complementation tests rapidly, efficiently, and, ideally, without prior requirement for a cloned gene. Advances in DNA sequencing and synthesis have reduced entry barriers to genetic studies in species closely related to traditional “model organisms” (Stein et al. 2003; Clark et al. 2007; Rhind et al. 2011; Scannell et al. 2011). Within the Saccharomyces genus of budding yeasts (formerly referred to as the Saccharomyces sensu stricto clade), haploids from a given species readily hybridize with haploids of multiple other species. The resulting interspecies hybrid diploids can propagate mitotically, but fail to produce viable progeny through meiosis (Greig 2009). Recently developed genetic tools in four Saccharomyces species (Scannell et al. 2011) allow one to conduct interspecies complementation analysis in hybrid diploids at a scale suitable for classic genetic screens, given a set of defined mutants for one of the species under study.
In this study, we used genetic screens and interspecies complementation assays to test whether Sir-protein-based transcriptional silencing was conserved between two Saccharomyces species separated by DNA sequence divergence similar to that between mouse and human (Kellis et al. 2003). Sir-based silencing mechanisms are of particular interest in that they appear to be restricted to a much narrower range of species than the mechanistically distinct RNAi-based silencing mechanisms. Sir proteins silence transcription at the two mating-type loci residing at opposite ends of chromosome III, HML and HMR, a process that is critical for yeast mating behavior. Normally, haploid strains of Saccharomyces cerevisiae mate as either a or α due to having an a or α allele present at the actively expressed MAT locus, while HMLα and HMRa remain silenced (Herskowitz 1988). Haploid yeast cells can switch to the opposite mating type by recombining one of the silent alleles from HML or HMR into MAT. However, when silencing is lost, haploid cells lose their ability to mate and to switch mating types due to coexpression of the a and α alleles.
The four silencer elements that flank HML and HMR contain binding sites for the origin recognition complex (ORC), Rap1, and Abf1, which in turn recruit Sir proteins. Typically, silencer “strength” has been defined by the degree to which a single silencer can repress transcription of nearby genes. Silencer strength is determined by the specific combination of the three binding sites (a strong silencer such as HMR-E has all three, whereas a weaker silencer such as HMR-I has two of the three) and other influences not yet completely understood (Rusche et al. 2003). A new coevolutionary concept suggests that silencer sequences can vary across species in their Sir-protein recruitment potential, which likely reflects variation in the underlying binding-site affinity for Rap1 and/or ORC (Zill et al. 2010). Silencer strength can be maintained in the face of variations in affinity of Rap1 or ORC for particular silencer sequences via compensatory changes in the Sir4 protein that likely affect its affinity for Rap1 and/or ORC.
Gene silencing, as defined in S. cerevisiae, serves as a model for heterochromatin formation in eukaryotes, but its component proteins vary in their conservation patterns from highly conserved to highly divergent. Indeed, Sir2, the only Sir protein that is also an enzyme, is one of the most highly conserved histone deacetylases in nature (Frye 2000; Smith et al. 2000), with additional roles in regulating some forms of aging and cellular metabolism (Smith et al. 2007; Kaeberlein 2010). Sir3 is a noncatalytic paralog of Orc1, a highly conserved member of ORC. The SIR3/ORC1 and SIR2/HST1 gene pairs, which arose during the whole-genome duplication (Figure 1A), are examples of the duplication, degeneration, and complementation model of paralogous protein evolution (Lynch and Force 2000; van Hoof 2005; Hickman and Rusche 2007, 2010). In contrast, SIR1, which encodes an Orc1-binding protein, is a member of a rapidly evolving gene family, and yeast species within the Saccharomyces complex (family Saccharomycetaceae) have between zero and four SIR1 paralogs (Gallagher et al. 2009). SIR4, which encodes a scaffold-like protein that binds Sir2 tightly and also interacts with Rap1 and Sir1 at silencers, is one of the most rapidly evolving genes in the Saccharomyces complex (Fabre et al. 2005; Zill et al. 2010) and is under positive selection in the Saccharomyces genus (Zill et al. 2010).
Given the greater diversity of Sir1-like proteins present in S. bayanus than in S. cerevisiae (Figure 1A), and the rapid evolution of the Sir4 protein, we reasoned that the silencing mechanism had likely undergone functional changes in the recent ancestry of Saccharomyces. These observations presented an opportunity to use evolutionary genetic tools in Saccharomyces to determine the extent to which this genetic regulatory circuit was conserved.
Materials and Methods
Yeast strains, culture, and targeted genetic manipulations
Yeast strains used in this study are listed in Supporting Information, Table S1. Growth of S. bayanus and interspecies hybrid strains was performed using standard conditions for S. cerevisiae, except that plate and liquid culturing was performed at 25° (for both species). One-step gene replacement by homologous recombination was performed using the standard PCR-based method, with minor modifications (Scannell et al. 2011). All gene disruptions in both species were confirmed using PCR to examine the 5′ and 3′ ends of targeted open reading frames, testing for both the presence of the disrupted allele and the absence of the wild-type allele.
Two S. bayanus MATa hmra1Δ::K.l.URA3 (henceforth HMR::URA3) strains (JRY8788 and JRY8789) were the starting strains for the mutagenesis (Zill et al. 2010). In these strains, the S. bayanus HMRa1 open reading frame was replaced with the Kluyveromyces lactis URA3 coding sequence by homologous recombination, leaving the HMRa1 promoter intact.
Hybrid diploids were created by patch mating S. cerevisiae strains to S. bayanus strains of the opposite mating type. Diploids were selected using complementation of auxotrophic markers. In cases where marker complementation was not possible, hybrids were generated by mating single cells of each species using a micromanipulator.
Screen for silencing-defective mutants
Two independent cultures of S. bayanus MATa HMR::URA3 (JRY8788 and 8789) were mutagenized with EMS (Amberg et al. 2005). After mutagenesis, the cells were plated at low density (∼500 cells per plate) onto YPD, CSM-Ura, and, at higher density, on CSM + canavanine plates to estimate the level of mutagenesis. Colonies were allowed to form over 3 days at 25°. Fifty-two colonies that grew on CSM-Ura (candidates for mutants expressing HMR::URA3) were streaked onto another CSM-Ura plate for single colonies, which were then patched onto 5-FOA plates to rescreen for stable HMR::URA3 expression. To test for loss of silencing at HML, we attempted to mate the candidate mutants to an S. cerevisiae MATα strain (JRY2728). Mutants that were both unable to mate and expressed URA3 were considered silencing defective (Figure 1B). For a subset of the mutants, HMLα was inactivated by gene targeting. The restoration of mating confirmed that these mutants had bona fide silencing defects. These mutants were each mated to an S. bayanus MATα ura3 SIR+ strain to test dominance; all mutations were recessive. The resulting diploids were sporulated, and tetrads were dissected to assay segregation of the silencing-defective phenotype (by Ura+ or nonmating phenotypes). For select sir2, sir3, and sir4 mutants (as defined by complementation tests, below), the putatively mutated SIR gene’s coding sequence was PCR amplified using Phusion high-fidelity DNA polymerase (New England Biolabs, Beverly, MA) from genomic DNA isolated from each mutant. The resulting PCR products were sequenced to identify causative mutations, which were sir2–Gly264Cys (JRY8802); sir3–ΔG629 (1-bp indel, JRY8803); sir4–Gly738STOP (weakly complemented by Sc–SIR4, JRY8796); and sir4–Ser812STOP (not complemented by Sc–SIR4, JRY8810).
Interspecies complementation tests
Rescue of mating ability in the silencing-defective mutants isolated from the mutagenesis was attempted by transforming them with single-copy CEN–ARS plasmids containing S. cerevisiae SIR2 (pJR2025), SIR3 (pJR2026), SIR4 (pJR2027), S. bayanus SIR4, or a vector control (pRS315) (Sikorski and Hieter 1989). In patch-mating tests on two or three transformants of each mutant, restoration of mating ability by a plasmid indicated that silencing was restored.
Quantitative mating assays:
S. cerevisiae and S. bayanus MATa sirΔ query strains bearing various CEN–ARS plasmids (as indicated in Table 1) were inoculated into overnight cultures selective for plasmid replication. Of each query strain culture, 0.5 OD600 was harvested and serially diluted 1:10, three times. Approximately 2 × 105, 2 × 104, and 2 × 103 cells of each query strain were plated on minimal medium together with 4 × 107 cells of the S. cerevisiae MATα his4 mating tester strain (JRY2728). The mating plates were incubated for 3 days at room temperature and diploid colonies were then counted. Mating efficiency was determined by dividing the average number of mating events (colonies) for each sirΔ strain by the average number of mating events for the wild-type control strain of the same species. The numbers reported in Table 1 reflect the average of three biological replicates (three independent transformants) for each query strain.
To test complementation of the EMS-induced silencing mutants, we mated S. bayanus MATa sir− HMR::URA3 hmlΔ strains to S. cerevisiae MATα hmrΔ (JRY5323) or MATα hmrΔ sir2Δ (JRY6754), sir3Δ (JRY8826), or sir4Δ (JRY7374) strains (Figure 2C, Table S1). A similar procedure was used for the comparisons of silencing in hybrids made between S. bayanus sir4Δ strains and other Saccharomyces species (Figure 3B and Figure 4). URA3 expression from either Sc–HMR or Sb–HMR could be readily assayed by growth of the hybrids on selective (CSM–Ura) and counterselective (5-FOA) media. For the panel of interspecies hybrids (Figure 3), a MATα HMR::URA3 sir4Δ strain (JRY9043) was mated to a MATa ura3 strain of each species.
Sir4 domain swap analysis
Using the gap-repair method (Orr-Weaver et al. 1983), portions of the Sb–SIR4 gene were swapped for corresponding portions of the Sc–SIR4 gene in CEN/ARS vector pRS316 (pJR2027). For the N-terminal swap, the plasmid was digested with BamHI and HindIII (partial) and the resulting gap repaired with a PCR product containing positions 541–2676 bp (AA 181–892) from Sb–SIR4 (replacing AA 181–842 of Sc–Sir4), plus 40 bp of flanking homology to Sc–SIR4. For the partitioning and anchoring domain (PAD) swap, the plasmid was digested with SpeI and the resulting gap repaired with a PCR product corresponding to positions 2752–3819 bp (AA 918–1273) from Sb–SIR4 (replacing AA 868–1224 of Sc–Sir4), plus 40 bp of flanking homology to Sc–SIR4. (Amino acids 843–867 in Sc–Sir4 were 100% conserved in Sb–Sir4.) The gap-repaired plasmids were rescued from their yeast hosts, transformed into Escherichia coli, and sequence verified. These domain-swapped SIR4 genes were then transformed into S. cerevisiae and S. bayanus MATa sir4Δ strains, and silencing function was tested by patch mating.
Interspecies hemizygosity screen for modifiers of an Sb–HMR silencing defect
An S. bayanus MATα HMR::URA3 cyh2 his3 lys2 ura3 strain (JRY9337) was mated with S. cerevisiae strains from the BY4741 (MATa) deletion collection (Winzeler et al. 1999; Giaever et al. 2002), and resulting Sc–yfg1Δ/Sb–YFG1 hybrid diploids were single-colony purified (“YFG1” denotes a given yeast gene). The hybrid strains were grown to saturation overnight in YPD and plated in 10-fold serial dilutions on 5-FOA, CSM–Ura, and YPD media to assay silencing and growth, respectively. S. cerevisiae ORF deletions were verified by PCR for each deletion strain tested.
Ancestral PAD sequence reconstruction
We reconstructed ancestral SIR4 PAD sequences (on codons) from five-species alignments using codeml (model = 0; NSSites = 0) in the PAML package (Yang 2007). We incorporated rate variation among sites using the gamma distribution approximated with eight rate classes. Amino-acid-based reconstructions and DNA non-codon-based reconstructions were obtained with FastML (Pupko et al. 2000) using Jukes–Cantor and Jones-Taylor-Thornton (JTT) models, respectively. As the maximum-likelihood reconstructions averaged 94% identity among methods, the codon-based reconstructions were used to guide DNA synthesis of ancestral PAD sequences at most sites. Sites at which maximum-likelihood reconstructions differed among methods were resolved by manual inspection, favoring amino-acid level conservation. Using the reconstructed S. mikatae/S. paradoxus/S. cerevisiae ancestral sequence, three PADs were generated by likelihood-based and manual filtering of ambiguous codons. The three PAD sequences, with 27 (strict filtering), 81, or 99 (liberal filtering) amino-acid replacements relative to the Sc–PAD sequence, were then synthesized by GeneArt (Life Technologies). [For comparison, Sc–PAD and Sb–PAD differed at 181 of 379 amino acids by using the Basic Local Alignment Search Tool (BLASTP).] After cloning each reconstructed PAD sequence into an Sc–SIR4 backbone, plasmid complementation of both Sc–sir4Δ and Sb–sir4Δ was tested using mating assays.
Evolutionary and population genetic analyses
Orthologs were identified using HMMER3 and syntenic context (Scannell et al. 2011). Protein alignments were produced using FSA with default parameters (Bradley et al. 2009), and DNA alignments were obtained by back-translation with RevTrans (Wernersson and Pedersen 2003). We calculated ω (dN/dS) for each set of orthologous genes using codeml in the PAML package with NSSites = 0 and α= 0. Torulaspora genomes were sequenced using Illumina paired-end reads and assembled with Short Oligonucleotide Analysis Package (SOAP, http://soap.genomics.org.cn/soapdenovo.html) (D. Scannell and M. Eisen, unpublished data ). Genes and orthologs were identified as described in (Scannell et al. 2011).
Polymorphism data for S. cerevisiae and S. paradoxus were downloaded from http://www.sanger.ac.uk/research/projects/genomeinformatics/sgrp.html. For each gene we discarded all strains with coverage at <50% of bases and discarded codons with coverage (all three positions) in <50% of strains. We also discarded codons at which the reference strain (S288c in the case of S. cerevisae) was aligned to its ortholog in S. paradoxus with an FSA accuracy score of <0.5. Finally, codons with SNPs at multiple positions, SNPs with a Phred score of <30, and low-frequency SNPs (<10% of accepted strains) were discarded. We then counted synonymous and nonsynonymous SNPs using custom scripts.
A screen for silencing-defective mutants in S. bayanus
To identify genes required to silence the HML and HMR mating-type loci in S. bayanus, we performed a genetic screen using two independent criteria to identify mutants. An HMR::URA3 silencing reporter was generated by replacing the S. bayanus HMRa1 ORF with the K. lactis URA3 ORF, with URA3 expression driven by the normally silent a1 promoter (Figure 1B). After mutagenesis of an S. bayanus MATa HMR::URA3 strain, mutants exhibiting HMR::URA3 expression were selected by plating the mutagenized cells onto solid medium lacking uracil (Figure S1A). Fifty-two Ura+ colonies were subsequently screened for FOA sensitivity, reflecting strong silencing defects at HMR::URA3. Ura+, FOA-resistant isolates were likely to include mutants with weak silencing defects, and these (26 of the 52) were not studied further. We reasoned that Ura+, FOA-sensitive mutants that could not mate, likely had derepressed HMLα in addition to HMR::URA3, as expected from the behavior of sir mutants in S. cerevisiae (Rine and Herskowitz 1987; Liaw and Lustig 2006). Although the three Ura+, FOA-sensitive isolates that could still mate, might have had HMR-specific silencing mutations, they were not studied further because cells that had undergone a heterothallic mating-type switch leading to expression of URA3 from MAT would have the same phenotype. Twenty-three Ura+, FOA-sensitive, nonmating mutants were recovered and analyzed for complementation as described below (screen results are summarized in Figure S1B).
To evaluate whether the screen had specifically identified silencing-defective mutants, we performed additional tests on a subset of the mutants. To test whether loss of mating depended on expression of HML, this locus was deleted in six mutants representing all three complementation groups (see below). In all six cases, mating was restored. The six MATa HMR::URA3 hmlΔ sir− strains were mated to a MATα ura3 SIR+ strain. All six mutants were recessive, as the resulting diploids were FOA resistant and Ura−. The diploids were sporulated and tetrad analysis on the resulting meiotic products showed 2 Sir+:2 Sir− segregation in at least 20 tetrads analyzed from each cross, with no apparent linkage to MAT or HMR. Thus, single genes had likely been inactivated in each of these mutants.
Interspecies complementation to identify mutant SIR genes: genetic incompatibility between S. cerevisiae SIR4 and S. bayanus HML and HMR
Given that disruption of no single S. bayanus SIR1 paralog has a silencing defect (Gallagher et al. 2009) and the stringent selective criteria imposed, the leading candidates for genes identified by the screen were S. bayanus SIR2, SIR3, and SIR4, and potentially new genes. S. cerevisiae and S. bayanus share the vast majority of their protein-coding genes, and most protein orthologs have high sequence identity [83% genome-wide average (Cliften et al. 2006; Scannell et al. 2011)]. It is generally assumed that the vast majority of these species’ orthologs perform identical functions and would be interchangeable between species. We tested complementation of the S. bayanus sir mutants with plasmids bearing S. cerevisiae SIR genes, using the restoration of mating ability to assay complementation. (Henceforth, we distinguish between the two species’ orthologs using the nomenclature Sc–SIR and Sb–SIR.) To a first approximation, plasmids bearing origins of replication and centromeres from S. cerevisiae work sufficiently well in S. bayanus for such purposes (Scannell et al. 2011).
Of the 23 sir− mutants, six were complemented by Sc–SIR2 and five by Sc–SIR3 (Figure S1B). The remaining 12 mutants showed either no mating when transformed with any Sc–SIR gene or slight but reproducible improvements in mating when transformed with Sc–SIR4. These 12 mutants were most likely S. bayanus sir4 mutants rather than new SIR genes, as they were complemented by a Sb–SIR4 plasmid (data not shown).
Plasmid-based interspecies complementation experiments with “clean deletion” sirΔ mutants revealed that SIR2 and SIR3 were completely interspecies compatible in both directions (Figure 2, A and B, and Table 1). In contrast, Sc–SIR4 could not complement S. bayanus sir4Δ with respect to silencing function at both HML and HMR (Figure 2A and B, Table 1; Zill et al. 2010). The SIR4–HML/HMR incompatibility showed an evolutionary asymmetry, as Sb–SIR4 restored silencing to HML and HMR in S. cerevisiae (Figure 2A and B). The molecular basis of the SIR4 incompatibility involved an inability of Sc–Sir4 to function at the S. bayanus silencer, as shown previously (Zill et al. 2010). The work presented here focused on the origin and nature of the changes to Sc–SIR4 that effectively restricted its “species range” and additional uses of complementation tests in interspecies hybrids.
The ability of Saccharomyces species to mate and form mitotically stable hybrid diploids offers a convenient way to use the well-developed genetics of S. cerevisiae to assign mutations in S. bayanus to genes by complementation tests in interspecies hybrids. We selected six recessive S. bayanus silencing-defective mutants, two from each group based on the plasmid complementation, and deleted their HML locus to allow them to mate. The resulting sir− hmlΔ strains were then mated with S. cerevisiae sir2Δ or sir3Δ or sir4Δ mutants as an independent method for determining complementation groups. As expected from the plasmid complementation results, the S. bayanus sir− mutants fell into three pairs based on their patterns of Sb–HMR silencing ability in the resulting S. bayanus/S. cerevisiae hybrids (Figure 2C). For the first pair of mutants, the resulting hybrids had robust Sb–HMR silencing except those made using an S. cerevisiae sir2Δ strain. For the second pair, the resulting hybrids had robust Sb–HMR silencing except those made using a sir3Δ strain. Finally, for the third pair of S. bayanus mutants, the resulting hybrids—even those made with S. cerevisiae ura3 SIR+ strains—showed little or no ability to silence Sb–HMR::URA3.
To provide an independent evaluation of the mutations assigned to S. bayanus SIR genes by these complementation tests, we sequenced the SIR2, SIR3, or SIR4 gene from four of the six mutants described above. In each of these mutants, a single severe missense, nonsense, or frameshift mutation was identified in the SIR gene that was deduced to be inactivated (details in Materials and Methods). Thus, the results of the genetic screen suggested that S. bayanus repressed HML and HMR using the same core silencing proteins—Sir2, Sir3, and Sir4—as S. cerevisiae. Genetic analyses of the contributions of Sb–SIR1 and its paralogs has been presented elsewhere (Gallagher et al. 2009; Zill et al. 2010).
A cladistic complementation series identified a discrete change in SIR4 function during the evolution of the Saccharomyces genus
The comparative genetic analysis identified an incompatibility involving SIR4, HML, and HMR between two species, but did not address how and when this incompatibility evolved relative to the ancestral states of these loci. We previously mapped the incompatibility determinants at HMR to its silencers, which have coevolved with Sir4 to maintain silencing in S. cerevisiae and S. bayanus (Zill et al. 2010). Given this coevolution, it was possible either that S. bayanus silencers had become limited in their ability to recognize other species’ Sir4 proteins or that S. cerevisiae Sir4 had become limited in its ability to recognize other species’ silencers. To distinguish between these possibilities, we asked whether K. lactis SIR4 was capable of silencing Sb–HML and Sb–HMR. (Due to the outgroup position of K. lactis on the phylogenetic tree, its SIR4 gene served as a proxy for the ancestral SIR4 of the Saccharomyces complex.) A plasmid bearing K. lactis SIR4 was capable of restoring silencing to Sb–HML and Sb–HMR, albeit to a lesser extent than a plasmid bearing Sb–SIR4 (Figure 3A and Table 1). K. lactis SIR4 is able to complement S. cerevisiae sir4Δ mutants, but Sc–SIR4 is not able to complement K. lactis sir4Δ mutants (Astrom and Rine 1998; Hickman and Rusche 2009). Thus, because S. bayanus silencers could serve as recruitment sites for the ancestral Sir4 protein, it appeared that Sc–Sir4 had become limited in its species range within the Saccharomyces genus. An important corollary of this conclusion was that S. bayanus silencers more closely resembled Saccharomyces ancestral silencers than did modern S. cerevisiae silencers.
To determine when during the evolution of the Saccharomyces genus the species-range-limiting changes in Sir4 occurred, we mated the S. bayanus HMR::URA3 sir4Δ strain to each of the four other Saccharomyces species and measured Sb–HMR silencing in the resulting hybrids. Notably, the S. mikatae/S. bayanus (JRY9198) and S. kudriavzevii/S. bayanus (JRY9162) SIR4/sir4Δ hybrids showed robust silencing, equivalent to the S. bayanus SIR4/sir4Δ diploid (Figure 3B). Thus, the SIR4 genes of two species that diverged from S. bayanus as long ago as did S. cerevisiae maintained the ability to complement S. bayanus sir4Δ. However, the S. paradoxus/S. bayanus hybrid showed a partial silencing defect, suggesting that the S. paradoxus Sir4 protein had lost some ability to function on S. bayanus silencers. If S. paradoxus silencing indeed represented an intermediate between S. bayanus-like silencing and S. cerevisiae-like silencing, then S. paradoxus silencers should retain some ancestral, S. bayanus-like character, and S. cerevisiae Sir4 might be partially incompatible with Sp–HML and Sp–HMR. To test this possibility, we measured the ability of Sc–SIR4 and Sb–SIR4 plasmids to complement S. paradoxus sir4Δ mutants. Although the Sb–SIR4 plasmid restored full silencing at both Sp–HML and Sp–HMR, the Sc–SIR4 plasmid restored silencing well only at Sp–HMR (Figure 3C). Thus, the Sp–Sir4 protein appeared to have a species range that was intermediate between those of Sc–Sir4 and Sb–Sir4. Similarly, S. paradoxus silencers appeared intermediate in their (presumptive) interspecies Sir4 recruitment potential, with S. cerevisiae silencers having greater attractive properties, and S. bayanus silencers being relatively less attractive to Sir4 proteins. [The locus-specific incompatibility between Sc–Sir4 and Sp–HML was likely due to differences in strength, as defined above, between HML and HMR silencers in S. paradoxus (O. Zill, unpublished results), similar to those noted for S. cerevisiae.]
As expected, robust silencing was observed in all four hybrids with both species’ SIR4 alleles intact (Figure 4A). All four Saccharomyces species were fully able to complement S. cerevisiae HMR::URA3 sir4Δ in interspecies hybrids (Figure 4B), consistent with the idea that S. cerevisiae silencers had evolved to accommodate a more limited Sir4 protein. The inability of Sc–SIR4 to complement S. bayanus sir4Δ was likely a species-wide property of S. cerevisiae, as similar Sb–HMR silencing defects were observed in hybrids made between S. bayanus sir4Δ and S. cerevisiae strains W303 (our standard laboratory strain), S288c, Σ1278b, or RM11-1A (a vineyard strain) (data not shown). Likewise, the intermediate level of Sb–HMR silencing in S. paradoxus/S. bayanus hybrids was a species-wide property of S. paradoxus, since silencing of S. bayanus HMR::URA3 sir4Δ was equivalent in independent S. paradoxus/S. bayanus hybrids made with S. paradoxus strains representing the three different subspecies groups (Figure 4C) (Cubillos et al. 2009; Liti et al. 2009). Thus, one or more species-range-limiting substitutions occurred in SIR4 in the common ancestor of S. paradoxus and S. cerevisiae, and an additional change in SIR4 likely occurred in the S. cerevisiae lineage.
Domain swaps localized Sir4 divergence to the N-terminal and PADs
To identify specific portions of Sir4 that had diverged between S. cerevisiae and S. bayanus, we constructed chimeric Sir4 proteins and tested their ability to complement an S. bayanus sir4Δ mutant (Figure 5). The chimeras were generated by substituting either a large N-terminal region including a putative regulatory domain (Moazed et al. 1997), or the PAD, from Sb–Sir4 into Sc–Sir4. The PAD serves to tether telomeres to the nuclear envelope via interactions with the Esc1 protein (Andrulis et al. 2002; Taddei et al. 2004). The PAD also interacts with the Ty5 retrotransposon integrase protein, targeting Ty5 integrations into heterochromatin (Xie et al. 2001; Brady et al. 2008). If one of these portions of Sb–Sir4 contained residues necessary for species-specific S. bayanus silencing, then silencing ability should cotransfer with that part of the protein. A chimeric Sir4 consisting of the Sb–PAD substituted into Sc–Sir4 provided a strong increase in silencing Sb–HML, as measured by patch-mating analysis (Figure 5). We also observed a slight increase in silencing ability with Sir4 containing a substitution of the Sb–N-terminal domain into Sc–Sir4. Thus, there were at least two discrete and mapable domains of species specificity between Sc–Sir4 and Sb–Sir4, each of which independently conferred partial or substantial S. bayanus silencing ability on Sc–Sir4.
We focused on the divergence in the PAD because this region of Sir4 has not previously been implicated in HML and HMR silencing in S. cerevisiae. Multiple alignments of Sir4 protein across the five Saccharomyces species revealed that sequence divergence across the PAD was similar to the average divergence level across the entire protein (data not shown). Further, the residues that are critical for the PAD–Ty5 integrase interaction in the Sc–Sir4 PAD were conserved in the other species. Therefore, there were no obvious sequence signatures of functional change. Amino-acid positions that played a role in restricting species range of Sc–Sir4 would be expected to bear distinct residues shared by Saccharomyces ancestral Sir4 and Sb–Sir4. The residues at these positions would likely be conserved among S. bayanus, S. kudriavzevii, and S. mikatae, as these species were able to provide Sb–HMR silencing function when mated to an S. bayanus sir4Δ mutant (Figure 3B). To identify candidate residues that had undergone function-altering substitutions in S. paradoxus and S. cerevisiae, we reconstructed the S. mikatae/S. paradoxus/S. cerevisiae ancestral PAD. We tested three different PAD sequences for silencing function, which were specified by different stringencies of filtering at ambiguously reconstructed sites. However, when each of these inferred ancestral sequences was inserted into the Sc–Sir4 protein, none were able to silence in S. cerevisiae (Figure S3). Thus the species-specific function of the PAD was refractory to simple refinement. More sophisticated interspecies-substitution efforts may be required to identify the critical PAD residues contributing to functional divergence in Sir4.
SIR4 continues to evolve rapidly in extant yeast populations
If the change in SIR4 function was associated with a strong selective pressure acting within a defined time window of the past, SIR4 might have undergone rapid evolutionary change leading to a high interspecies divergence relative to other genes. Since that selective pressure would have ceased to act long ago, we might expect relatively low SIR4 polymorphism levels within modern isolates of S. cerevisiae. Alternatively, if modern yeasts continue to experience selective pressure driving rapid SIR4 evolution, then we would expect relatively high levels of SIR4 polymorphism within extant yeast populations. Using population genetic data from the Saccharomyces Genome Re-sequencing Project, we tested these two possibilities, calibrating SIR4 polymorphism levels against genes of similar length, syntenic context (i.e., genes neighboring SIR4), and evolutionary rate (dN/dS). In both S. cerevisiae and S. paradoxus populations, SIR4 has continued to evolve rapidly relative to other yeast genes, and at a level similar to that expected based on interspecies divergence in SIR4 sequence (Figure 6A). These data were consistent with diversifying selection or relaxed purifying selection operating in modern Saccharomyces populations in the wild. (Details of the population genetic analyses for each species are provided in Figure S2.) The rapid evolution of SIR4 appeared to be a property deeply conserved across the Saccharomycetaceae family. In the Torulaspora genus, SIR4 evolutionary rate (dN/dS) ranked in the top 1% of 4910 orthologs across three species, similar to its position among Saccharomyces genes (Figure 6, B and C).
An interspecies hemizygosity screen for S. bayanus silencing genes
The Sb–HMR locus is weakly derepressed in S. cerevisiae/S. bayanus hybrid diploids, likely due to Sb–Sir protein sequestration by S. cerevisiae telomeres (Zill et al. 2010). We reasoned that reducing the dosage of genes important for Sb–HMR silencing might exacerbate the defect in the hybrids, allowing us to efficiently survey genes for roles in S. bayanus silencing. We tested this idea by performing a candidate-gene hemizygosity screen to find S. cerevisiae gene-knockout mutations that suppressed or enhanced the Sb–HMR silencing defect (Figure 7A). Conceptually, this approach was similar to second-site noncomplementation screens (Welch et al. 1993).
We tested 66 mutants, each carrying a deletion of a gene implicated in S. cerevisiae silencing (see Table S2, Table S3, and Table S4 for complete list of genes). Deletion of the S. cerevisiae allele of SAS2, SAS4, or SAS5 suppressed the silencing defect, consistent with the known role of the SAS-I acetyltransferase complex in counteracting silencing in S. cerevisiae (Figure 7B). Inactivation of Sc–RIF1, RLF2, or ARD1 enhanced the silencing defect, suggesting that the S. bayanus orthologs of these genes also played roles in Sb–HMR silencing. Deletion of Sb–SIR1 caused a strong loss of Sb–HMR silencing, whereas Sc–SIR1 deletion only weakly enhanced the hybrids’ silencing defect (Figure 7B). These results were consistent with a requirement for species-specific contacts between the Sb–Sir1/Kos paralogs and Sb–Sir4 to silence Sb–HMR, with Sc–Sir1 being largely incompatible with silencer-bound proteins (Zill et al. 2010).
We have established genetic criteria for determining functional conservation or divergence for components of a regulatory circuit based on interspecies complementation. Using a genetic screen, we identified three genes required for silencing in S. bayanus. Two of the three genes (SIR2 and SIR3) were functionally conserved, based on their interspecies compatibility, and one was not (SIR4). Using outgroup complementation and a cladistic complementation series, we narrowed the time window for major changes in SIR4 to the shared S. cerevisiae/S. paradoxus branch and the S. cerevisiae branch. Domain swaps localized S. bayanus-specific silencing functions to two regions of the Sir4 protein, the N-terminal domain and the PAD. Finally, an interspecies hemizygosity screen for modifiers of an Sb–HMR silencing defect identified genes likely involved in S. bayanus silencing.
The Sir2/Sir3/Sir4 silencing machinery was conserved across Saccharomyces species, but Sir4 interactions have diverged
In Ascomycete fungi there are two known classes of epigenetic silencing mechanisms: S. pombe and N. crassa, and their close relatives use an HP-1- and-RNAi-based heterochromatin silencing mechanism, whereas S. cerevisiae, K. lactis, and their relatives lack HP-1 and most RNAi genes, and instead form heterochromatin via Sir proteins (most of which are absent from the former group of organisms). RNAi components have been found in S. castellii (Drinnenberg et al. 2009), an outgroup of the Saccharomyces clade, raising the question of how broadly the Sir silencing mechanism known from S. cerevisiae operates within the set of yeast species in which Sir proteins are found. The screen described herein was a pilot study to determine, in broad outline, whether the Sir-based silencing mechanism is conserved across the Saccharomyces genus. We identified SIR2, SIR3, and SIR4 as the primary genes required to silence HML and HMR in S. bayanus (Figure S1B). Previous work had established that S. bayanus SIR3 is required for silencing and can complement S. cerevisiae sir3Δ (Liaw and Lustig 2006). Although the screen was not saturated, based upon statistical criteria, that was not our intention. Rather we explored the feasibility of the interspecies complementation approach and used it to ascertain the phylogenetic and molecular underpinnings of divergence in Sir4.
In S. cerevisiae, subtelomeric genes are silenced by Sir proteins, with the mechanism overlapping somewhat with HML and HMR silencing. Sir2, Sir3, and Sir4 are required to silence HML, HMR, and subtelomeric genes, but Sir1 is important only for HML–HMR silencing (Pryde and Louis 1999). Additionally, Sir proteins are recruited to telomeric regions using somewhat distinct DNA elements: the telomeric terminal repeats (containing Rap1 sites) and internal sequences known as X elements (that usually contain ORC and Abf1 sites). As Sc–Sir4 can silence some S. bayanus subtelomeres but not others (Zill et al. 2010), our results may bear some relevance to S. bayanus telomeric silencing. However, further studies will be needed to address the details of how telomeric sequences and silencing have evolved in Saccharomyces. Deeper and more sensitive screens may reveal additional species-specific aspects of S. bayanus silencing (e.g., the Sir1 paralogs and their interaction partners), but the general mechanism of S. bayanus silencing at HML and HMR clearly depends on Sir proteins.
We established that S. cerevisiae mutations, in most cases, could be used to assign S. bayanus mutations to complementation groups in interspecies hybrids. Importantly, the single exception (SIR4) allowed us to identify unexpected functional divergence in Sir silencing based on the inability of S. cerevisiae SIR4 to complement recessive mutations in S. bayanus SIR4 (Figure 2B and C). Sir2 and Sir3 were bidirectionally interspecies compatible (Figure 2), suggesting that changes in interactions made by the Sir2/3/4 complex had diverged mainly with respect to Sir4-mediated interactions. As Sir2 deacetylates, and Sir3 interacts with, highly conserved histone tails, it was not surprising that these components of silent chromatin had not substantially diverged between S. cerevisiae and S. bayanus. Although Sir4 protein sequences are highly divergent relative to other Saccharomyces proteins (Zill et al. 2010), it was nonetheless surprising that Sir4 proteins were not completely compatible across species (discussed further below).
An interspecies hemizgosity screen revealed that Sas2, Sas4, and Sas5 antagonized silencing of S. bayanus HMR, similar to their roles in promoting active transcription in S. cerevisiae (Figure 7). Conversely, Rif1, Rlf2 (alias Cac1), and Ard1 all promoted robust silencing of S. bayanus HMR. The effect of Sb–SIR1 deletion was likely due to the direct action of Sb–Sir1 at Sb–HMR (Gallagher et al. 2009). The reduced Sb–HMR silencing by ARD1 deletion was notable because in S. cerevisiae ARD1 deletion perturbs only HML silencing (Whiteway et al. 1987). It appeared that either S. bayanus HMR silencing had evolved a requirement for ARD1 function or that S. cerevisiae HMR silencing had lost that requirement. Further experiments will be necessary to distinguish whether this additional regulatory role for Ard1 is specified by the Sb–HMR locus or whether it reflects global rewiring of the S. bayanus silencing machinery.
Ongoing rapid sequence evolution of SIR4 as a facilitator of discrete functional changes
Population genetic data from S. cerevisiae and S. paradoxus revealed an unusually high level of nonsynonymous polymorphism in Sir4 within species (Figure 6 and S2), suggesting that diversifying selection affects SIR4 in modern yeast populations. Our previous phylogenetic analysis of SIR4 in Saccharomyces identified rapid evolution across much of the SIR4 gene and positive selection operating at several codons (Zill et al. 2010). The rapid evolution of SIR4 orthologs in the Torulaspora (which last shared a common ancestor with Saccharomyces ∼100 MYA) suggests that the changes that gave rise to the incompatibility between S. cerevisiae and S. bayanus occurred on a background of rapid evolution at the SIR4 locus. Thus, SIR4 evolution in Saccharomyces has followed two distinct selection regimes: a pattern of long-term, continual evolutionary churning across the gene and the recent fixation of species-range-limiting substitutions at a smaller number of sites. We cannot rule out that some Sir4 amino acids, particularly in the N-terminal half of the protein, may be subject to relaxed purifying selection. However, we consider a purely neutral model of Sir4-functional change to be unlikely because Sir4 function must be maintained to ensure robust mating and mating-type switching in haploid yeasts. The Ty5 retrotransposon, whose integrase protein binds Sc–Sir4 to target Ty5 integration (Zou et al. 1996; Zou and Voytas 1997) into the “safe haven” of silent chromatin (Boeke and Devine 1998; Dai et al. 2007), is a possible candidate for a selective pressure driving the recent functional divergence of SIR4.
The domain-swap analysis identified two independently acting determinants of the divergence in Sir4 silencing function and localized a major portion of the divergence to the PAD (Figure 5). The role of the PAD in S. bayanus silencing is interesting because the Sc–Sir4 PAD has never been directly implicated in silencing HML and HMR, although it plays an indirect role in telomeric silencing by mediating the interactions of telomeres with Esc1 and the nuclear envelope (Taddei et al. 2004). It is possible that HM silencing in S. bayanus, in contrast to S. cerevisiae, requires association with “silent chromatin domains” at the nuclear envelope (Gartenberg et al. 2004; Gasser et al. 2004). Additionally, the Sc–Sir4 PAD mediates the interaction with Ty5 integrase. That substitutions in this domain affected silencing function was consistent with a model in which Sc–Sir4 had given up an ancestral silencing property to protect the yeast genome from harmful Ty5 retrotransposition events. By this model, the S. cerevisiae silencers then “caught up” by evolving higher-affinity Rap1 and ORC binding sites to allow a limited Sc–Sir4 to continue to silence HML and HMR (Zill et al. 2010; O. Zill, unpublished results).
Regardless of the evolutionary force driving the high levels of Sir4 sequence divergence and polymorphism, Sir4 presents a curious example given that it is essential for the yeast sexual cycle and has multiple interaction partners. Sir4’s central role in heterochromatin formation may require it to have some evolutionary pliability, which would confer the flexibility needed to consistently stabilize other heterochromatin proteins on rapidly evolving DNA templates. In general, certain scaffold proteins may evolve more rapidly to maintain regulatory allosteric interactions with multiple coevolving client proteins (Good et al. 2009, 2011). This evolutionary pliability may also allow them to adopt different interaction partners given the right set of selective conditions.
Using interspecies complementation to identify divergence in ortholog function
To what extent can we extrapolate function over a genetic distance similar to that between mouse and human? Most orthologous proteins between mouse and human have highly similar sequences [85% median similarity (Jaillon et al. 2004; Dujon 2006)]. It is generally assumed that the vast majority of these orthologs perform identical functions, and most of the few human genes tested complement the orthologous mouse and yeast mutants. However, to pursue the basis of species-specific traits, tools for systematically identifying functionally diverged orthologs are increasingly necessary. In this study we describe an example of orthologous proteins that do not function identically in two yeast species, as shown by an asymmetric failure of interspecies complementation, despite the two species being separated by a genetic distance similar to that between human and mouse.
The case of SIR4 highlights that evidence of rapid sequence evolution, including that provided by current methods for detecting positive selection, is not sufficient to infer functional divergence. S. kudriavzevii SIR4 was as diverged from S. bayanus SIR4 as was S. cerevisiae SIR4, yet one complemented S. bayanus sir4Δ and the other did not. Furthermore, highly diverged SIR4 orthologs—those of S. bayanus and K. lactis—complement the S. cerevisiae sir4Δ mutant. SIR4 evolves at a fourfold higher rate than the average Saccharomyces gene (Figure 6B). However, this rate is uniformly high along all phylogenetic branches in the genus (Zill et al. 2010), whereas SIR4 functional divergence occurred specifically in the lineages of S. paradoxus and S. cerevisiae.
In considering the available evidence, perhaps the only possible predictors of SIR4 functional divergence were its limited phylogenetic distribution and changes in paralog number among the Sir1-like proteins that interact with Sir4 (although the correlation with Sir4 compatibility was not perfect). Rapidly evolving, clade-specific genes such as SIR1 and SIR4 may be the most likely candidates for genes with species-specific functions and would therefore represent genetic determinants of the “exceptional biology” that make up the key innovations of a particular species (Eichler 2001).
Asymmetric noncomplementation across species, in general, may reveal significant evolutionary events that have shaped the biology of individual species or clades. Here we have provided a general and conceptually simple genetic framework for identifying conservation and divergence of genes between closely related species and for defining when critical function-altering changes took place. Systematic functional studies using yeast interspecies hybrids could help to derive guidelines for reliably predicting true functional divergence of orthologs between closely related species. Additionally, studies of fusion hybrids formed between human amniocytes and mouse muscle cells suggest that interspecies hybrid complementation analysis might be used more broadly to identify conserved and diverged regulatory proteins in mammals (Blau et al. 1983).
We thank Barbara Meyer, Mike Eisen, Chris Hittinger, Gianni Liti, Ed Louis, Laura Lombardi, and members of the Rine lab for helpful discussions. We thank Logan Fink for conducting the interspecies hemi-zygosity screen and Chris Hittinger for generously providing ura3Δ-marked haploid S. kudriavzevii strains. We thank Gianni Liti and Ed Louis for providing ura3Δ-marked haploid versions of the three representative subspecies strains of S. paradoxus. This work was supported by an National Science Foundation Graduate Fellowship and National Institutes of Health (NIH) training grant T32-HG00047 (O. Zill), NIH grant HG002779 (D. Scannell), NIH grant P50-GM081879, and NIH grant GM31105 (J.R.).
Communicating editor: C. D. Jones
- Received April 30, 2012.
- Accepted August 15, 2012.
- Copyright © 2012 by the Genetics Society of America