Comprehensive Synthetic Genetic Array Analysis of Alleles That Interact with Mutation of the Saccharomyces cerevisiae RecQ Helicases Hrq1 and Sgs1

Most eukaryotic genomes encode multiple RecQ family helicases, including five such enzymes in humans. For many years, the yeast Saccharomyces cerevisiae was considered unusual in that it only contained a single RecQ helicase, named Sgs1. However, it has recently been discovered that a second RecQ helicase, called Hrq1, resides in yeast. Both Hrq1 and Sgs1 are involved in genome integrity, functioning in processes such as DNA inter-strand crosslink repair, double-strand break repair, and telomere maintenance. However, it is unknown if these enzymes interact at a genetic, physical, or functional level as demonstrated for their human homologs. Thus, we performed synthetic genetic array (SGA) analyses of hrq1Δ and sgs1Δ mutants. As inactive alleles of helicases can demonstrate dominant phenotypes, we also performed SGA analyses on the hrq1-K318A and sgs1-K706A ATPase/helicase-null mutants, as well as all combinations of deletion and inactive double mutants. We crossed these eight query strains (hrq1Δ, sgs1Δ, hrq1-K318A, sgs1-K706A, hrq1Δ sgs1Δ, hrq1Δ sgs1-K706A, hrq1-K318A sgs1Δ, and hrq1-K318A sgs1-K706A) to the S. cerevisiae single gene deletion and temperature-sensitive allele collections to generate double and triple mutants and scored them for synthetic positive and negative genetic effects based on colony growth. These screens identified hundreds of synthetic interactions, supporting the known roles of Hrq1 and Sgs1 in DNA repair, as well as suggesting novel connections to rRNA processing, mitochondrial DNA maintenance, transcription, and lagging strand synthesis during DNA replication.

Despite these advances in yeast RecQ research, little is known about the genetic interactions that occur between HRQ1 and SGS1 or the physical interactions between Hrq1 and Sgs1. In humans, some of the RecQ helicases are partially functionally redundant (e.g., BLM and WRN), some display complementarity (e.g., WRN and RECQL5), and others exhibit functional synergism (reviewed in (Croteau et al. 2014)). The latter is exemplified by BLM and RECQL4, where BLM promotes the retention of RECQL4 at DNA double-strand breaks (DSBs), and RECQL4 stimulates BLM activity (Singh et al. 2012). Do such connections exist between their yeast homologs Hrq1 and Sgs1? Two reports demonstrate that various combinations of hrq1 and sgs1 alleles display differential responses to DNA damage compared to single mutants Rogers et al. 2020), suggesting that functional interactions among the RecQ helicases also exist in S. cerevisiae.

Rationale for screen
The study of yeast RecQ homologs has greatly expanded our mechanistic understanding of how these enzymes function in various DNA repair pathways, but the interplay between Hrq1 and Sgs1 and their roles in other biological processes are not known. We sought to identify genes whose mutation affects the growth of hrq1 and/or sgs1 mutant cells. Because inactive alleles of DNA helicases often act as dominant negatives (Wu and Brosh 2010) and in some cases better represent disease-linked alleles, we utilized both deletion (hrq1D and sgs1D) and catalytically inactive mutants (hrq1-K318A and sgs1-K706A) of the helicases in all combinations (hrq1D, sgs1D, hrq1-K318A, sgs1-K706A, hrq1D sgs1D, hrq1D sgs1-K706A, hrq1-K318A sgs1D, and hrq1-K318A sgs1-K706A) in our screen. Many genes that encode proteins involved in genome integrity are also essential, so we performed synthetic genetic array (SGA) analysis by mating our query helicase mutant strains to both the S. cerevisiae single-gene deletion collection (Giaever and Nislow 2014) and the temperature-sensitive (TS) collection (Kofoed et al. 2015), the latter of which includes alleles of essential genes not found in the former, to generate a comprehensive set of double and triple mutant strains for SGA analysis.

Screen design
The strains used in this study are listed in Table 1. The HRQ1 gene was deleted in Y8205 (Table 1) by transforming in a NatMX cassette that was PCR-amplified from the plasmid pAC372 (a gift from Amy Caudy) using oligonucleotides MB525 and MB526 (Table S1). The deletion was verified by PCR analysis using genomic DNA and oligonucleotides that anneal to regions up-and downstream of the HRQ1 locus (MB527 and MB528). The confirmed hrq1D strain was named MBY639. The hrq1-K318A allele was introduced into the Y8205 background in a similar manner. First, an hrq1-K318A(NatMX) cassette was PCR-amplified from the genomic DNA of strain MBY346  using oligonucleotides MB527 and MB528 and transformed into Y8205. Then, genomic DNA was prepared from transformants and used for PCR analyses of the HRQ1 locus with the same oligonucleotide set to confirm insertion of the NatMX marker. Finally, PCR products of the expected size for hrq1-K318A(NatMX) were sequenced using oligonucleotide MB932 to confirm the presence of the K318A mutation. The verified hrq1-K318A strain was named MBY644.
The SGS1 gene was deleted from Y8205 (Table 1) in the same manner as the HRQ1::natMX deletion above by transforming in a NatMX cassette that was PCR-amplified using oligonucleotides MB1395 and MB768 (Table S1). The deletion was verified by PCR analysis of genomic DNA and oligonucleotides MB373 and MB374. The confirmed sgs1D strain was named MBY640. The sgs1-K706A allele was PCR amplified from plasmid pFB-MBP-Sgs1K706A-his (Cejka and Kowalczykowski 2010) (Table 2) using oligonucleotides MB765 and MB1396. The NatMX cassette was PCR-amplified from pAC372 using oligonucleotides MB1397 and MB768 and fused to the sgs1-K706A PCR product by Gibson assembly (Gibson et al. 2009). The resultant sgs1-K706A(natMX) cassette was reamplified with MB765 and MB768 and transformed into Y8205. Genomic DNA was then prepared from transformants and used for PCR analyses of the SGS1 locus with oligonucleotides MB373 and MB374 to confirm insertion of the cassette. Finally, PCR products of the expected size were sequenced using oligonucleotide MB769 to confirm the presence of the K706A mutation. The verified sgs1-K706A strain was named MBY642.
The double mutant strains were constructed using similar techniques. Briefly, the hrq1D sgs1D and hrq1-K318A sgs1D strains were generated by deleting SGS1 in strains MBY639 and MBY644, respectively, using a URA3 cassette amplified from pUG72 (Gueldener et al. 2002) with oligonucleotides MB1395 and MB355 (Table S1). The strains verified by PCR of genomic DNA and sequencing were named MBY643 and MBY645, respectively. The hrq1D sgs1-K706A and hrq1-K318A sgs1-K706A strains were constructed by amplifying sgs1-K706A as above, amplifying the URA3 cassette with oligonucleotides MB1397 and MB355, and fusing the PCR products via Gibson assembly. The sgs1-K706A(URA3) cassette was then transformed into strains MBY639 and MBY644, and transformants were confirmed for proper integration by PCR and Sanger sequencing. The verified hrq1D sgs1-K706A and hrq1-K318A sgs1-K706A strains were named MBY674 and MBY676, respectively. Further details concerning strain construction are available upon request.

Phenotypes
Quantitative scoring of the genetic interactions was based on colony size. The SGA score measures the extent to which the size of a double or triple mutant colony differs from the colony size expected from combining the query and tester mutations together . The data includes both negative (putative synthetic sick/ lethal) and positive interactions (potential epistatic or suppression interactions) (Tables S2-17). The magnitude of the SGA score is indicative of the strength of the interaction. Based on statistical analysis, it was determined that a default cutoff for a significant genetic interaction is P , 0.05 and SGA score . |0.08| .

Verification of mutants
The top five negative and positive interactions for each query strain were confirmed by remaking and reanalyzing the double and triple mutants by hand, followed by spot dilution (Andis et al. 2018) and/or growth curve (Ononye et al. 2020) assays to compare the growth of the double or triple mutants to their parental strains and wild-type. Examples are shown in Figure S1.

Statistical analysis
Data were analyzed and graphed using GraphPad Prism 6 software. The reported values are averages of $ 3 independent experiments, and the error bars are the standard deviation. P-values were calculated as described in the figure legends, and we defined statistical significance as P , 0.01.

Data availability
Strains, plasmids, and other experimental reagents are available upon request. File S1 contains Table S1, a description of the other supplementary tables included in Files S2-S4, and Figures S1-S4. File S2 contains Tables S2-S9, File S3 contains Tables S10-S17, and File S4 contains the complete SGA data for all screens in the form of Tables S18-33. Supplemental material available at figshare: https://doi.org/ 10.25387/g3.13157519.

RESULTS AND DISCUSSION
Overall results of the screen Hundreds of synthetic interactions were detected for all query strains screened through both the single-gene deletion (Table 2) and TS mutant (Table 3) collections (Tables S2-17). For the single-gene deletion collection screen, the numbers of negative and positive genetic interactions were generally the same for all query strains, except hrq1D and hrq1-K318A, which yielded approximately twice as many negative as positive interactions (Table 2). These mutants also had the fewest number of synthetic interactions by a factor of . 2.3 compared to sgs1D and sgs1-K706A. This is consistent with the generally more modest phenotypes of hrq1D and hrq1-K318A strains compared to sgs1D and sgs1-K706A for DNA damage sensitivity . The double mutant query strains yielded a greater than additive number of synthetic genetic interactions than the single mutant parental query strains, indicating that mutating both RecQ helicases had a synergistic effect. This synergism was strongest for the hrq1-K318A sgs1D mutant, which generated 880 synthetic interactions, a nearly twofold increase over the additive effect of the 132 hrq1-K318A and 312 sgs1D interactions individually (compared to 1.5to 1.sixfold increases for the other combinations).
For the TS allele collection screen, the numbers of negative and positive genetic interactions were again generally similar for all query strains (Table 3). As above, the hrq1D and hrq1-K318A mutants had the fewest number of synthetic interactions by a factor of . 2.1 compared to sgs1D and sgs1-K706A. In this case, however, the double mutant query strains yielded approximately an additive number of synthetic genetic interactions compared to the single mutant parental query strains and thus did not display the synergism described for the single-gene deletion SGA analysis. It should also be noted that the numbers of synthetic genetic interactions listed in Table 3 are inflated because several different TS alleles of the same ORF are included in the collection for many individual genes (Kofoed et al. 2015). Figure 1 shows the frequency distribution of all of the SGA scores as violin plots and separate box plots of the negative and positive synthetic genetic interactions, with outliers denoted as single points, for the single-gene deletion collection ( Figure 1A-C) and the TS collection ( Figure 1D-F). The outliers represent the mutants with the strongest synthetic phenotypes. As shown in Figures 1A and 1D, most synthetic phenotypes were mild decreases or increases in the growth of the double and triple mutant colonies. There were no significant differences in the distribution of the SGA scores among any of the mutant sets generated by crossing the query strains to the single-gene deletion collection. However, several significant differences were found in the distributions of positive SGA scores for the mutant sets yielded from the crosses to the TS collection. These includes mild differences between hrq1-K318A vs. hrq1D sgs1D (P = 0.0123) and sgs1D vs. hrq1-K318A sgs1D (P = 0.0303), intermediate differences for sgs1D vs. hrq1D sgs1D (P = 0.0016) and sgs1-K706A vs. hrq1-K318A sgs1-K706A (P = 0.0070), and strong differences between sgs1-K706A and hrq1D sgs1D, hrq1D sgs1-K706A, and hrq1-K318A sgs1D (all P , 0.0001). It is currently unclear why the strength of the positive synthetic genetic interactions significantly varied among these mutants, especially compared to the sgs1-K706A query strain, but we are actively following up on phenotypic difference among all of the hrq1 and sgs1 alleles. Regardless, as mutants giving the strongest growth effects, the outliers in Figures 1B, C, E, and F are summarized in Tables 4 and 5. For simplicity, only the negative genetic interactions are discussed in further detail below. Comparisons between the full SGA datasets for the hrq1D/hrq1-K318A, sgs1D/sgs1-K706A, and all pairwise combinations of triple mutants screens are available in Figures S2 and S3.

hrq1D interactions
The deletion of HRQ1 displayed strong negative interactions with mutations in 10 genes (Tables 4 and 5), many of which correspond to the recently described Hrq1 interactome (Rogers et al.) 1 . For instance, RECQL4 is the only human RecQ found in both the nucleus and mitochondria (Croteau et al. 2014), and Hrq1 likewise localizes to both organelles (Koh et al. 2015) and physically interacts with mitochondrial proteins (Rogers et al.) 1 . Here, we found strong negative synthetic genetic interactions between hrq1D and mutation of MRM2, a mitochondrial 2' O-ribose methyltransferase whose deletion results in mitochondrial DNA (mtDNA) loss (Pintard et al. 2002), and YSC83, a mitochondrial protein of unknown function (Sickmann et al. 2003). It is still unclear what the role of Hrq1 is in the mitochondria, but it is tempting to speculate that it is involved in mtDNA maintenance in a similar fashion to its maintenance of the nuclear genome. This role in genome integrity is highlighted by the negative interactions of hrq1D with mutation of SPO16, which is involved in the meiotic cell cycle (Shinohara et al. 2008), and RAD14, a nucleotide excision repair protein (Guzder et al. 2006) and regulator of transcription (Chaurasia et al. 2013). Deletion of HRQ1 also negatively interacted with mutation of SLX9, an rRNA processing factor (Bax et al. 2006) that additionally binds G-quadruplex (G4) DNA structures (Gotz et al. 2019). This is provocative in light of the connection of Hrq1 to rRNA processing and ribosome biogenesis (Rogers et al.) 1 , as well as the fact that G4 structures are preferred substrates for Hrq1 in vitro . Finally, mutations in YEF3, YUR1, MUP3, and PHO5 (encoding a translation elongation factor, protein glycosylase, methionine permease, and acid phosphatase, respectively), as well as the dubious open reading frame (ORF) YDR455C (Fisk et al. 2006), also negatively interacted with hrq1D.

hrq1-K318A interactions
Mutations in only two genes, RAD14 and YEF3, are shared between the lists of strong negative interactors with hrq1D and hrq1-K318A. This is not unexpected based on the ability of Hrq1-K318A to phenocopy wild-type in some pathways ). However, mutations in genes encoding proteins involved in processes shared between both sets are evident. This includes HAP2 and HAP3, which are activators of transcription (Xing et al. 1993), TCO89, a member of the TOR complex and global regulator of histone H3 K56 acetylation (Chen et al. 2012), and RAD14 as described above. Similarly, TOM70 encodes a subunit of the mitochondrial protein importer (Brix et al. 2000), which is likely important for localizing Hrq1 to the mitochondria where it may be involved in mtDNA maintenance. Genome integrity is also highlighted by CBC2, which encodes an RNA binding and processing factor involved in telomere maintenance (Lee-Soety et al. 2012). Hrq1 is known to regulate telomerase activity at both DSBs and telomeres Nickens et al. 2018;Nickens et al. 2019). Mutation of the gene encoding the Vps41 vacuolar membrane protein (Nakamura et al. 1997) also negatively interacted with hrq1-K318A.
The overall genetic interactome of HRQ1 In a companion manuscript, we present proteomic and transcriptomic data related to Hrq1 and Hrq1-K318A, highlighting similarities and differences between the wild-type and mutant to generate a wholistic picture of Hrq1 biology manuscript #401710). We also touch on the SGA data included herein. Thus, it is useful to compare some of the hrq1D and hrq1-K318A data. In total, 117 significant (P , 0.05) genetic interactions (76 negative and 41 positive) were identified between hrq1D and the single-gene deletion collection, and 119 (65 negative and 54 positive) were identified between hrq1D and the TS alleles collection (Table S2). Similarly, 132 significant (P , 0.05) genetic interactions (84 negative and 48 positive) were identified between hrq1-K318A and the single-gene deletion collection, and 102 (41 negative and 61 positive) were identified between hrq1K318A and the TS alleles collection (Table S3). When comparing the hrq1D and hrq1-K318A data sets in aggregate, there was 39% overlap between the negative genetic interactions (Figure 2A) and . 30% overlap n■  between the positive genetic interactions ( Figure 2B). However, there was very little overlap when comparing negative to positive genetic interactions and vice versa ( Figure

sgs1D interactions
Over 500 genetic interactions with sgs1 alleles have been reported (see: https://www.yeastgenome.org/locus/S000004802/interaction), including most of the hits from our screen, such as the genome integrity genes MMS4, RRM3, SLX1, SLX4, SRS2, and WSS1 (Fisk et al. 2006), as well as SLX9 (see above) and EFB1, which encodes a translation elongation factor (Hiraga et al. 1993). These hits serve as internal positive controls. It should also be noted that: 1) YBR099C is a dubious ORF that completely overlaps MMS4 (Fisk et al. 2006), 2) YBR100W was an originally misannotated ORF and more recently merged with an adjacent ORF such that the coding region is now named MMS4 (Xiao et al. 1998), and 3) the pby1D strain in the singlegene deletion collection is actually a deletion of MMS4 (ölmezer et al. 2015). Thus, multiple different mms4 alleles were hits in the screen, again acting as positive controls for our approach. The individually plotted points outside of the inner fences represent outliers (i.e., interactions with mutants yielding the strongest SGA scores) and correspond to alleles whose SGA score is less than the value of the 25 th quartile minus 1.5 times the inter-quartile distance (IQR) for negative interactions and alleles whose SGA score is greater than the value of the 75 th quartile plus 1.5IQR for positive interactions. The significant differences between SGA data sets discussed in the main text were calculated using the Kruskal-Wallis test and Dunn's multiple comparisons test.
In addition to known effects, we also discovered three new negative interactions with sgs1D. These include the deletions of SWC4 and SWC5, which encode subunits of the SWR1 complex that replaces histone H2A with H2A.Z (Mizuguchi et al. 2004), preventing the spread of silent heterochromatin (Meneghini et al. 2003). This interaction could be connected to the role of Sgs1 in telomere maintenance (Huang et al. 2001;Johnson et al. 2001;Azam et al. 2006) because telomeric DNA is also silenced via the telomere position effect (Mondoux and Zakian 2005). As with hrq1D and hrq1-K318A, the yef3-f650s TS allele was also a negative genetic interactor with sgs1D (Table 5).
Mutations in the remaining genes have not previously been reported to negatively interact with sgs1D, but three of them (SPO16, YSC83, and HAP3) overlap with the hrq1 interactors described above, perhaps suggesting some overlap in function between Hrq1 and Sgs1 in the pathways related to these genes. That leaves only two genes, SUA7 and ASK10, as unique interactors here. The SUA7 gene product is the yeast transcription factor TFIIB that is needed for RNA polymerase II transcriptional start site selection (Pinto et al. 1992). This may indicate that like the human RECQL5 helicase (Aygun et al. 2008;Izumikawa et al. 2008;Saponaro et al. 2014), Sgs1 is involved in transcription, a hypothesis also put forth for Hrq1 (Rogers et al.) 1 . In support of this, the remaining interactor ASK10 encodes a glycerol channel regulator (Beese et al. 2009) that also associates with RNA polymerase II (Pagé et al. 1996).

Comparisons to previous screens
A comprehensive set of SGA data already exists for all non-essential S. cerevisiae gene deletion strains Costanzo et al. 2016), including the hrq1D and sgs1D alleles screened in this work. Indeed, these strains were included as internal controls for the catalytically inactive helicase allele screens. Comparing our data to the previously published screens, there is good overlap between the significant positive and negative genetic interactors with hrq1D and sgs1D (Fig. S4). However, the overlap is far from complete, despite the screens being performed using strains of the same genetic backgrounds and on the same equipment in the Donnelly Centre (University of Toronto). This underscores the importance of re-preforming such control screens when new alleles such as hrq1-K318A and sgs1-K706A are investigated.

CONCLUSIONS AND PERSPECTIVES
Here, we have reported a comprehensive set of synthetic genetic interactions between most of the genes in the S. cerevisiae genome and deletion and catalytically inactive alleles of the Hrq1 and Sgs1 RecQ family helicases. This data set improves upon the existing sets of known hrq1D and sgs1D interactions and expands the genetic interactome landscape of hrq1 and sgs1 mutants by including interactions with the inactive hrq1-K318A and sgs1-K706A alleles, as well as all combinations of the null and inactive double mutants. As with the five human RecQ helicases (Croteau et al. 2014), it is clear that HRQ1 and SGS1 genetically interact in yeast, and perhaps they may also physically interact. These SGA analyses have also generated testable hypotheses to drive on-going and future research. The genetic interactomes of hrq1 and sgs1 suggest links to transcription, much like the functional interaction between human RECQL5 and RNA polymerase II (Aygun et al. 2008;Izumikawa et al. 2008;Saponaro et al. 2014). Indeed, we have already shown that hrq1 cells are sensitive to the general transcription inhibitor caffeine and that hrq1 mutations alter the S. cerevisiae transcriptome (Rogers et al.) 1 . Similarly, it will be exciting to discover why double hrq1 sgs1 mutations are particularly deleterious to defects in lagging strand synthesis during DNA replication.
Obviously, our focus on the strongest negative synthetic genetic interactions in the SGA data set reported here is far from all encompassing. There are certainly important conclusions to be drawn from more subtle negative effects, considering the positive genetic interactions, and comparing the genetic interactomes between the various hrq1 and sgs1 mutants analyzed. It is our hope that these data will spur additional research in the field, both with the yeast RecQs and their human homologs RECQL4 and BLM, as well as with proteomic investigations to incorporate physical interactomes, to fully establish the roles of these enzymes in genome integrity.