Characterization of shu2∆ recombination phenotype and its physical interactions with the other Shu complex members

The Rad51 paralogs have a number of defining features that are related to promoting Rad51 filament formation, such as decreased rates of Rad51-mediated gene conversion on disruption (Sung 1997; Alvaro et al. 2007; Godin et al. 2013). Consistently, we previously found that deleting either CSM2 or PSY3 results in a decrease in gene conversion and a subsequent increase in Rad51-independent repair (Godin et al. 2013). To determine if Shu1 or Shu2 has a similar role in promoting Rad51-dependent repair, we performed a heteroallelic recombination assay in WT, shu1∆, and shu2∆ cells (Figure 1A). In this assay, a recombination event between two leu2 heteroalleles with an intervening URA3⁺ gene can generate a LEU2⁺ prototroph through repair by Rad51-dependent sister-chromatid GC (LEU2⁺ URA3⁺) or Rad51-independent intrachromosomal SSA (LEU2⁺ URA3⁻). Similar to csm2∆ and psy3∆ cells (Godin et al. 2013), we found that shu1∆ and shu2∆ cells have significantly decrease rates of Rad51-dependent GC (P ≤ 0.02 and P ≤ 0.005, respectively) with a corresponding increase in rates of error-prone SSA (Figure 1A). These results demonstrate that although Shu2 is not a Rad51 paralog, it exhibits many of the same phenotypes as the other Shu complex members.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

SHU2 mutants have phenotypes similar to the Rad51 paralogs. (A) Disruption of SHU1 or SHU2 leads to decreased rates of Rad51-dependent repair. Diagram of direct repeat recombination assay (leu2-∆EcoRI::URA3::leu2-∆BstEII) used to measure rates of direct-repeat recombination by Rad51-dependent GC (right side of diagram, Ura⁺Leu⁺ colonies) or Rad51-independent SSA (left side of diagram, Ura⁻Leu⁺ colonies). The rates of GC and SSA events in shu1∆ or shu2∆ cells were compared with WT and standard deviations shown. (B) Shu1 bridges the interaction between Shu2 and Psy3. Yeast two-hybrid analysis of pGBK-SHU2 (containing a GAL4-binding domain) with pGAD-SHU1, pGAD-CSM2, pGAD-PSY3, and pGAD-C1 empty plasmid (containing a GAL4-activating domain) in the presence or absence of the endogenous SHU1 gene. Interaction is indicated by growth on medium lacking histidine. Growth on medium lacking leucine and tryptophan selects for the individual plasmids and serves as a loading control.

The human Shu2 homolog hSWS1 interacts with multiple Rad51 paralogs either directly or indirectly through hSWSAP1 (Martin et al. 2006; Liu et al. 2011). We asked whether other yeast Shu members similarly bridge the protein-protein interactions between Shu2 and Shu1 or Psy3 (Shor et al. 2005; Ball et al. 2009). To address this question, we performed Y2H analysis of yeast expressing SHU2 in the GAL4 DNA-binding domain (pGBK-SHU2) and tested its interaction with either SHU1 or PSY3 expressed in the GAL4 DNA-activating domain (pGAD-SHU1, pGAD-PSY3) in a genetic background in which one of the four SHU genes is disrupted (Figure 1B and data not shown). We found that loss of SHU1 disrupts the Shu2-Psy3 Y2H interaction (Figure 1B). Therefore, similar to hSWS1, which interacts with the other hRad51 paralogs through hSWSAP1, the interaction of Shu2 with the Rad51 paralog Psy3 is likely stabilized by Shu1.

Shu2 orthologs are conserved across eukaryotes

Although Shu2 homologs were identified in fission yeast and humans (Shor et al. 2005; Martin et al. 2006), their phylogenetic orthology with budding yeast Shu2 has not been demonstrated, nor have Shu2 orthologs been identified in other major eukaryotic lineages, including important model organisms. To produce a complete picture of the SWS1 family, we searched for Shu2 homologs across eukaryotic and outgroup archaeal lineages. PSI-BLAST queried with yeast Shu2 yielded hits across fungal species and metazoans. PSI-BLAST queried with human SWS1 (ZSWIM7) led to hits across eukaryotes, including plants, and archaeal proteins containing a SWIM domain. In addition, we identified putative SWS1 orthologs in several early-branching eukaryote lineages, including Diplomonadida (Giardia lamblia), Euglenozoa (Leishmania and Phytomonas), green algae (Coccomyxa and Bathycoccus), Stramenopiles (Aphanomyces, Albugo, Phaeodactylum, and Ectocarpus), Alveolata (Paramecium, Plasmodium, and Oxytricha), and Ichthyosporea (Capsaspora) (sequences in File S1). These sequences clustered with known SWS1 orthologs in phylogenetic trees separated from archaea outgroup sequences. This deep diversity, in addition to known fungal, metazoan, and plant orthologs, suggests an ancient origin of the SWS1 protein family. PSI-BLAST recovered a single-protein sequence from each species, suggesting that there were few or no Shu2/SWS1 family duplications and that these orthologs are well conserved.

To define a comprehensive Shu2/SWS1 protein family, we constructed a phylogeny using Shu2 homologs from major eukaryotic and archaeal lineages (Figure 2A). In the resulting phylogeny, eukaryotic sequences were cleanly separated from the archaeal sequences with high branch support (aLRT = 0.98). Moving forward from the archaeal root, the well-supported branches between eukaryotic sequences were in agreement with accepted speciation events, thus supporting the orthology of these Shu2/SWS1 sequences. Although the short alignment (219 amino acids) of Shu2/SWS1 did not provide sufficient power to infer all branch nodes with strong support, these sequences appear to be true orthologs. Notably, this tree revealed Shu2 orthologs in D. melanogaster and C. elegans, which we will refer to as dmSws1 and ceSws-1 hereafter, as well as an Arabidopsis thaliana ortholog, AT4G33925. The archaeal SWIM domain–containing proteins served only to root the tree, and we are not able to comment on the orthology or specific function of those sequences.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

The SWS1 protein family extends across eukaryotes. (A) A phylogeny based on amino acid sequences shows the deep conservation of the SWS1 protein from Giardia lamblia, an early branching eukaryote lineage, to plants, fungi, and metazoans. The short sequence length did not allow resolution of some interior branches, but all well-supported nodes support the orthology of these sequences. (B) A partial multiple alignment shows the highly conserved SWIM domain. Absolutely conserved residues (*) include the defining CXC…X_n…CXH motif. Double-gap columns (–) indicate trimmed regions of low sequence identity. (C) The Shu2-SWS1 protein family is a monophyletic clade in this phylogeny of all proteins with SWIM domains from humans, D. melanogaster, C. elegans, and yeast. The alignment for this phylogeny contained only the 30 most highly conserved residues in the SWIM domain. These relationships further support the newly discovered Shu2-SWS1 orthologs.

A defining feature of the Shu2/SWS1 protein is the SWIM domain, a zinc-binding feature (Figure 2B) (Martin et al. 2006). To further ensure orthology between yeast Shu2 and human SWS1 (also known as ZSWIM7) in exclusion of other SWIM domain–containing proteins in humans (ZSWIM1–6 and ZSWIM8), C. elegans (pqn-55), and D. melanogaster (CG34401), we constructed a phylogeny with all SWIM domain proteins from these species and all putative eukaryotic SWS1 orthologs (Figure 2C). Homology between these proteins is limited to the SWIM domain, so their alignment is limited to a region of 30 highly conserved amino acids. As expected for a small alignment, many branching nodes were not well resolved; however, the branching pattern cleanly separates the SWS1 orthologs from the other SWIM domain proteins with moderate support (aLRT = 0.70). This topology further supports the putative orthology of the sequences in Figure 2A (i.e., they descended from a single common ancestor).

Shu2 and its orthologs have evolutionary histories strongly correlated with recombination and meiosis-related proteins

To better define the biologic function of the Shu complex, we performed coevolutionary analysis in both budding yeast and Drosophila. This analysis exploits the observation that functionally related proteins tend to have covarying rates of evolution because they experience shared evolutionary pressures. This property can be quantified as ERC, reflecting the degree to which two proteins have rates of sequence evolution that covary between species (Clark et al. 2009; Clark and Aquadro 2010; Clark et al. 2012). ERC for a protein pair is quantified as a correlation coefficient (ranging between −1 and 1) for which higher values reflect stronger rate covariation. ERC values are typically elevated between functionally related proteins genome-wide in yeasts and Drosophila, as well as between meiosis and DNA repair proteins in mammals (Clark et al. 2012, 2013; Findlay et al. 2014). Hence an elevated ERC value for a protein pair suggests cofunctionality between them.

We first demonstrated that members of the yeast Shu complex have significantly covarying rates with each other, as we might predict given their cofunctionality (Figure 3A); their pairwise ERC values were highly elevated as a group (mean ERC = 0.52, permutation test P < 0.00001), whereas the expected mean ERC for a random gene set is zero. These ERC values were calculated across a phylogeny of 18 fungal species, including S. cerevisiae. We also found elevated ERC values between the Shu complex members and Srs2 (Figure 3A), consistent with a conserved physical and/or genetic interaction between Shu2 and Srs2 (a DNA helicase that disassembles Rad51 filaments) in both budding and fission yeast (Ito et al. 2001; Martin et al. 2006; Bernstein et al. 2011).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Coevolutionary signatures indicate that Shu2/SWS1 has highly conserved functional relationships with Rad51 paralogs and meiosis proteins. (A) Evolutionary rate covariation (ERC) values between members of the Shu complex are highly elevated (P < 1 × 10⁻⁵⁾. The degree of red shading in a cell indicates a higher ERC for that protein pair compared with the null expectation of zero. (B) Violin plots depict the distributions of ERC values between Shu2 and proteins from various functional classes. Violin width is proportional to the density at that ERC value. The genome-wide distribution (all genes) of ERC with Shu2 is centered at zero, as is Shu2 with mitosis genes (N = 107 genes). Meiosis (N = 126) and Shu complex genes (N = 4) show a strong enrichment of high ERC values with Shu2, consistent with cofunctionality between them. P-values (horizontal bars) strongly reject similarity between those distributions and the genome-wide distribution. Each of the four values in the Shu complex is also plotted with an X. (C) Drosophila Sws1 (dmSws1), ortholog of fungal Shu2, similarly shows elevated ERC values with Rad51 paralogs. (D) DmSws1 also has high ERC values with meiosis proteins (N = 116) and Rad51 paralogs (N = 4) but not with mitosis proteins (N = 129).

To determine the coevolutionary relationship of the Shu complex with broader functional groups within fungi, we studied Shu2’s ERC values with mitotic and meiotic proteins. Shu2 did not show significant rate covariation with mitotic proteins, but ERC values between Shu2 and meiotic proteins were significantly elevated, suggesting strong coevolution between them (P = 1.1 × 10⁻⁸) (Figure 3B and Table S3 and Table S4). Importantly, these results were unchanged when recombination-related genes and genes shared by the two sets were removed from analysis, so the association is not limited to HR proteins (mitosis P = 0.27; meiosis P = 1.1 × 10⁻⁶). Similar to Shu2, Psy3 also exhibits significantly increased ERC values with meiotic but not mitotic proteins (Figure S1A and Table S5 and Table S6). In contrast to Shu2 and Psy3, both Shu1 and Csm2 show significant rate covariation with both mitotic and meiotic proteins (Figure S1, B and C, and Table S7, Table S8, Table S9, and Table S10). In addition, the D. melanogaster Shu2 ortholog dmSws1 (CG34314) also showed rate covariation with meiotic proteins (P = 0.013) and with Rad51 paralogs (P = 0.0066) but not with mitotic proteins as a class (Figure 3D). The Drosophila dmSws1 results also remained unchanged after removing recombination-related and shared genes (mitosis P = 0.62; meiosis P = 0.0238). Most notably, Rad51C and Rad51D showed extremely high rates of covariation, whereas the other two Rad51 paralogs, Spn-A and Spn-B, showed modestly elevated levels (Figure 3C). Finally, we tested for coevolutionary associations of meiotic and mitotic genes with the mammalian Shu2 ortholog SWS1 (ZSWIM7). Contrary to results in fungi and Drosophila, ERC values between mammalian SWS1 and meiotic and mitotic genes were not generally elevated. Overall, results from fungi and D. melanogaster suggest that the Shu complex has conserved meiotic and mitotic roles in eukaryotic species.

Expansion of the SWIM domain to include an invariant alanine three amino acids downstream of the canonical CXC...X_n...CXH motif

The SWIM motif was originally defined as a zinc-binding motif that contains the canonical CXC...X_n...CXH sequence (Liu et al. 1995; Makarova et al. 2002; Banerjee et al. 2004). However, on analysis of our evolutionarily deep alignment, we identified an invariant alanine located three amino acids downstream from the CXH motif (Figure 4A and Figure S2). Interestingly, in humans, this invariant alanine in hSWS1 is mutated to a threonine (A108T) in a cancer patient from the COSMIC database (Shepherd et al. 2011). To determine whether alanine 108 may be functionally important, we constructed a Y2H vector containing hSWS1 mutagenized to include this mutation, SWS1-A108T, and mutations in the canonical SWIM domain residues (C85S, C87S, C103S, and H105A)(Figure 4B). Suggesting that these residues are functionally important, disruption of the canonical SWIM domain or the invariant alanine (A108) results in a reduced Y2H interaction with hSWS1’s obligate binding partner hSWSAP1 (Figure 4B). Interestingly, another highly conserved residue, F90, does not result in reduced Y2H interaction when mutated to an alanine (Figure 4B, SWS1-F90A). These results suggest that the SWIM domain in hSWS1 including the invariant alanine is likely important for hSWS1 function.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Analysis of the highly conserved SWIM domain residues in the human SWS1 protein. (A) Sequence logo of the SWIM domain reveals a highly conserved phenylalanine two residues after the CXC motif and an invariant alanine two residues after the CXH motif. (B) Mutating the SWIM domain in human SWS1 impairs its interaction with hSWSAP1. Y2H analysis of pGAD-hSWS1 with mutations in the SWIM domain (C85, C87, C103, H105), as well as F90 and A108, were assayed for interaction with human SWSAP1 (pGBD-SWSAP1) as described in Figure 1B.

The SWIM domain is important for Shu2’s functionality in vivo

To investigate the role of the SWIM domain and the invariant alanine, we created Y2H vectors harboring the analogous mutations in the budding yeast SHU2 SWIM domain (Figure 5A). Via Y2H analysis, we found that mutating the canonical SWIM domain residues (C114S, C116S, C176S, and H178S) in shu2 results in an undetectable Y2H interaction with Shu2’s binding partners Shu1 and Psy3 (Figure 5A). Interestingly, mutating alanine 181 to a threonine leads to a reduced Y2H interaction with Psy3 but not Shu1. Furthermore, while the hSWS1-F90A mutant maintains its Y2H interaction with hSWSAP1, we found that the corresponding mutation in yeast shu2, F119A, results in a reduced Y2H interaction with Psy3 but not Shu1 (Figure 5A). These results suggest that the canonical SWIM domain in yeast Shu2 is likely important for function and that the conserved alanine and phenylalanine also may be components of this domain.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

The SWIM domain is important for the function of Shu2. (A) Mutating the SWIM domain in SHU2 impairs its interaction with SHU1 and/or PSY3. Y2H analysis of pGBK-SHU2 or pGBK-shu2 mutant interactions with pGAD-SHU1 or pGAD-PSY3 as described in Figure 1B. (B) Canonical SWIM domain mutants are nonfunctional. shu2∆ cells harboring the indicated pGBK plasmids were fivefold serially diluted onto YPD medium or YPD medium containing 0.02% MMS and incubated at 30°C for 2 days. (C) Disruption of the canonical SWIM domain suppresses sgs1∆ DNA damage sensitivity. WT, sgs1∆, sgs1∆ shu2∆, sgs1∆ shu2-C114S, sgs1∆ shu2-C116S, sgs1∆ shu2-F119A, sgs1∆ shu2-C176S, and sgs1∆ shu2-H178A cells were fivefold serially diluted onto YPD medium or YPD medium containing 50 mM HU or 0.006% MMS and incubated at 30°C for 2 days. (D) Loss of the Shu complex results in reduced spore viability. Diploid yeast with the indicated mutations were sporulated, and individual spores were tetrad dissected onto rich medium. Viable spore colonies from 22 individual tetrads were quantitated, and the average number of viable meiotic progeny was calculated. Standard deviations are shown from three experiments, and significance was determined by t-test (P < 0.05). (E) Same as D except that the shu2 SWIM domain containing mutants were analyzed.

To determine whether the impaired Y2H interactions of the SWIM domain mutants result in diminished DNA damage tolerance in yeast, we complemented shu2∆ cells with a WT SHU2- or a mutant shu2-expressing plasmid. We found that the WT SHU2 plasmid complements the MMS sensitivity of shu2∆ cells, while mutations in the canonical SWIM domain (C114S, C116S, C176S, and H178A) do not (Figure 5B). Interestingly, despite the altered protein-protein interactions observed in both shu2-F119A and shu2-A181T, both mutant plasmids complement the MMS sensitivity of a shu2∆ cell (Figure 5B). These results indicate that the interaction between Shu2 and Psy3 may be dispensable for Shu2’s mitotic function. These findings were confirmed when we stably integrated the canonical SWIM domain mutants (C114S, C116S, C176S, and H178A) as well as F119A at the endogenous SHU2 locus (Figure S3). Unfortunately, we were unable to integrate A181T into our yeast strains. Because the Shu genes were originally characterized by their ability to suppress the sensitivity of a top3∆ or sgs1∆ strain to MMS or HU treatment, we also examined whether the integrated SWIM domain alleles also would rescue sgs1∆ DNA damage sensitivity such as disruption of SHU2 and found that the canonical SWIM domain mutants do (Figure 5C). Similar findings also were observed with a SWIM domain mutation in the fission yeast gene sws1-C152S (Martin et al. 2006). Thus our analysis of the four canonical SWIM domain mutants demonstrates that the conserved SWIM domain is important for this protein family’s function in the repair of MMS-induced DNA lesions.

Given the conserved nature of the Shu complex in meiotically dividing yeast, we asked how disruption of the SWIM domain would affect meiotic outcome in S. cerevisiae. To test this, we sporulated diploids homozygous for deletion of a single Shu complex member and screened for viability of the resulting offspring. In agreement with two recent reports (Hong and Kim 2013; Sasanuma et al. 2013b), we found that loss of any member of the Shu complex causes a marked decrease in spore viability, with Csm2 and Psy3 resulting in a more severe defect (P ≤ 0.02 for all) (Figure 5D). Next, we tested whether mutations of the SWIM domain similarly would lead to decreased meiotic progeny. We found that the SWIM domain mutants C114S, C116S, C176S, and H178A are also defective for spore viability compared with WT yeast (P ≤ 0.05 for all) (Figure 5E). In contrast, the shu2-F119A allele spore viability was similar to that of WT yeast (Figure 5E). These results demonstrate that the canonical SWIM domain is necessary for Shu2’s role during meiosis.