Noncanonical Contributions of MutLγ to VDE-Initiated Crossovers During Saccharomyces cerevisiae Meiosis

In Saccharomyces cerevisiae, the meiosis-specific axis proteins Hop1 and Red1 are present nonuniformly across the genome. In a previous study, the meiosis-specific VMA1-derived endonuclease (VDE) was used to examine Spo11-independent recombination in a recombination reporter inserted in a Hop1/Red1-enriched region (HIS4) and in a Hop1/Red1-poor region (URA3). VDE-initiated crossovers at HIS4 were mostly dependent on Mlh3, a component of the MutLγ meiotic recombination intermediate resolvase, while VDE-initiated crossovers at URA3 were mostly Mlh3-independent. These differences were abolished in the absence of the chromosome axis remodeler Pch2, and crossovers at both loci became partly Mlh3-dependent. To test the generality of these observations, we examined inserts at six additional loci that differed in terms of Hop1/Red1 enrichment, chromosome size, and distance from centromeres and telomeres. All six loci behaved similarly to URA3: the vast majority of VDE-initiated crossovers were Mlh3-independent. This indicates that, counter to previous suggestions, levels of meiotic chromosome axis protein enrichment alone do not determine which recombination pathway gives rise to crossovers during VDE-initiated meiotic recombination. In pch2∆ mutants, the fraction of VDE-induced crossovers that were Mlh3-dependent increased to levels previously observed for Spo11-initiated crossovers in pch2∆, indicating that Pch2-dependent processes play an important role in controlling the balance between MutLγ-dependent and MutLγ-independent crossovers.

repaired either by a synthesis-dependent strand annealing pathway that forms non-crossovers (NCOs), or by a pathway that forms double Holiday junction (dHJ) intermediates that are resolved as crossovers (COs) by the MutLg (Mlh1-Mlh3 and Exo1) meiosis-specific resolvase (Schwacha and Kleckner 1994;Wang et al. 1999;Khazanehdari and Borts 2000;Kirkpatrick et al. 2000;Tsubouchi and Ogawa 2000;Allers and Lichten 2001b;Allers and Lichten 2001a;Hoffmann et al. 2003;Argueso et al. 2004;Bishop and Zickler 2004;Nishant et al. 2008;Zakharyevich et al. 2010;Al-Sweel et al. 2017). In budding yeast, COs and NCOs are formed at similar levels, suggesting that roughly equal fractions of DSBs are repaired by these two pathways (Martini et al. 2006;Mancera et al. 2008). Apart from these two major pathways, a minor pathway uses mitotic resolvases (structure-selective nucleases, SSNs: Mus81-Mms4, Yen1 and Slx1-4) to form both NCOs and COs (De Los Santos et al. 2003;Argueso et al. 2004;Lynn et al. 2007;Jessop and Lichten 2008;De Muyt et al. 2012;Zakharyevich et al. 2012;Agostinho et al. 2013;Oke et al. 2014). While the proteins and enzymatic activities contributing to each of these pathways has been the subject of considerable study (reviewed in Ehmsen and Heyer 2008;Hunter 2015;Manhart and Alani 2016), the question of what roles local chromosome environment might play in pathway choice remains much less explored. Medhi et al. (2016) addressed this question using a meiosis-specific endonuclease, VDE, that cleaves a recognition sequence (VRS) at high efficiency regardless of chromosomal context (Gimble and Thorner 1992;Gimble and Thorner 1993;Nogami et al. 2002;Fukuda et al. 2003;Medhi et al. 2016;this work). Like Spo11 DSBs, VDE DSBs are processed to form single-stranded overhangs that recruit the Rad51 and Dmc1 proteins that perform strand invasion and homology search (Bishop et al. 1992;Fukuda et al. 2003;Fukuda and Ohya 2006). Medhi et al. inserted a VRS-containing recombination reporter at two loci: HIS4, present in a region with high levels of both Spo11 DSBs and Hop1 binding; and URA3, in a region with low levels of Spo11 DSBs and Hop1 binding (Pan et al. 2011;Panizza et al. 2011). Most COs at HIS4 were Mlh3-dependent, while COs at URA3 were Mlh3-independent. In pch2Δ mutants, Hop1 occupancy at HIS4 was reduced, as were the fraction of COs that were Mlh3-dependent, while at URA3 the fraction of COs that were Mlh3-dependent increased. Based on these findings, Medhi et al. suggested that the local chromosome structure, in particular levels of Hop1 enrichment, may be an important determinant of CO pathway choice.
To test the generality of the above suggestion, we inserted the same VRS recombination reporter at six new loci with varying Hop1 occupancy in their vicinity and found that VDE-initiated meiotic COs at all six new loci were predominantly Mlh3-independent. Moreover, as previously seen for inserts at URA3 (Medhi et al. 2016), pch2Δ mutation increased the fraction of COs that were Mlh3-dependent. These results indicate that, contrary to our previous suggestion, local Hop1 occupancy levels alone do not determine the mechanism of JM resolution during the formation of VDE-induced meiotic crossovers. They also suggest that, at most loci, VDE DSBs are repaired differently than are Spo11 DSBs.

Yeast strains
All strains (Table S1) used in this study are of SK1 background (Kane and Roth 1974) and were constructed by transformation or genetic crosses. The recombination reporter cassette with the VRS (cleavable) or VRS-103 (uncleavable) site in the ARG4 gene (Medhi et al. 2016) were inserted by ends-out transformation (for VRS-containing inserts and for VRS-103 inserts at FIR1 and HSP30, Figure S1A) or by ends-in transformation (for VRS-103 constructs at CCT6, RIM15, IMD3 and TRK2, Figure S1B) at six different locations, using primers listed in Table S2. Ends-in transformation was used for inserts at divergently transcribed loci to minimize effects on expression caused by disruption of 59 untranslated regions. Transformation was performed with overlapping DNA fragments as illustrated in Figure S1. The VRS-arg4 and VRS-103-arg4 constructs are 5.5kb and 8.6kb long, respectively, with 3kb sequence homology around the VRS site. This size difference, along with HindIII site differences, enables the detection of the parental and recombinant chromosomes on Southern blots (see Figure 2, below).

Growth and sporulation
Strains were grown in pre-sporulation SPS medium and transferred to sporulation medium as described (Goyon and Lichten 1993), with the inclusion of 10mM CuSO 4 in sporulation medium to induce VDE expression (Medhi et al. 2016). DNA samples were collected and processed as described (Allers and Lichten 2000;Jessop et al. 2005;Jessop et al. 2006).
DNA extraction and Southern hybridization DNA was extracted from samples using the CTAB extraction method (Allers and Lichten 2000;Oh et al. 2009). Genomic DNA was digested with HindIII or HindIII and PI-SceI, run on agarose gels, blotted, probed and analyzed as described (Medhi et al. 2016).

Cytology
Cells were collected, stained with DAPI, and scored by epifluoresence microscopy to follow nuclear divisions as described (Kaur et al. 2018).

Statistical analysis
GraphPad Prism was used for comparisons of mean values, using twotailed t-tests with the Holm-Sídák correction for multiple comparisons.

Data availability
All strains are available upon request. The authors affirm that all data necessary for confirming the conclusions of this article are represented fully within the article, tables, figures, and supplementary figures, tables and files. Data underlying graphs in all figures and chromosome coordinates in Figure 1 are in File S1, available at Figshare (https:// doi.org/10.25387/g3.7800728). Other supplementary files, available at the same URL, include: Figure S1: Construction of inserts Figure S2: Additional data and analyses. Includes cells completing meiosis I, timing of VDE DSBs, combined CO and NCO data, and NCO/CO ratios

VDE-initiated COs are Mlh3-independent at most insert sites
To further test the hypothesis that Hop1-enrichment determines the MutLg-dependence of meiotic CO formation, six new sites were selected for VRS reporter insertion, one (HSP30) with regional Hop1 levels (average Hop1 occupancy over 10-20kb around the insert location) similar to those at URA3, four (CCT6, RIM15, TRK2 and IMD3) with Hop1 levels similar to those at HIS4, and one (FIR1) with intermediate Hop1 levels ( Figure 1B, File S1). Since it has been previously shown that Spo11-DSBs are reduced near centromeres and telomeres (Pan et al. 2011) and CO formation is regulated differently on longer and shorter chromosomes (Joshi et al. 2009;Zanders and Alani 2009), the new sites were selected such that they were on chromosomes of different sizes and were at varying distances from centromeres and telomeres ( Figure 1A, File S1). At each site, recombination products can be differentiated on Southern blots (Figure 2A, B), as was previously used to quantify DSBs, COs and NCOs (Medhi et al. 2016).
Meiotic progression of all WT and mlh3Δ strains was similar, with most cells completing the first meiotic division by 7-8h post-induction ( Figure S2A). In addition, VDE-initiated DSBs appeared and disappeared with levels and timing similar to those previously seen at HIS4 and URA3 ( Figure S2B; Medhi et al. 2016).
COs in VRS inserts ranged from 6% of total lane signal at CCT6 to 10.3% at HIS4 ( Figure 2C). As previously reported (Medhi et al. 2016), NCOs were recovered in substantial excess over COs at all insert loci ( Figure 2E), with NCO/CO ratios ranging from 2.1 to 4.8 (mean = 3.1 6 0.8; Figure S2D). The marked excess of NCOs over COs seen for VDE-initiated events differs from what is seen with Spo11-initiated events, where COs and NCOs are produced at similar levels (Martini et al. 2006;Mancera et al. 2008;Zakharyevich et al. 2012). In contrast to what was seen for VRS inserts at HIS4, where COs were reduced dramatically in mlh3Δ mutants (to 40% of wild-type levels), COs in the same sequences inserted at all other loci were only modestly affected, with COs in mlh3Δ ranging from 80% to 115% of wild type (mean = 91 6 12%; Figure 2D); NCOs were similarly unaffected ( Figure 2E, F). These results indicate that, in contrast to Spo11-initiated COs, which are reduced about twofold in mlh3Δ mutants (Wang et al. 1999;Khazanehdari and Borts 2000;Kirkpatrick et al. 2000;Tsubouchi and Ogawa 2000;Hoffmann et al. 2003;Argueso et al. 2004;Nishant et al. 2008;Al-Sweel et al. 2017;Chakraborty et al. 2017), most COs at the VDE break sites are formed independent of MutLg, irrespective of the chromosome size, distance from centromere or telomere, or Hop1enrichment in their vicinity. Thus, at most insert loci in otherwise wildtype cells, VDE-initiated recombination differs from Spo11-initiated recombination and more closely resembles mitotic recombination, in that NCOs are in excess over COs (Esposito 1978;Lichten and Haber 1989;Ira et al. 2003;Dayani et al. 2011) and, with the exception of those formed in inserts at HIS4, VDE-initiated COs are largely MutLgindependent.

VDE-initiated COs are partially Mlh3-dependent in pch2D mutants
In pch2 mutants, meiotic axis proteins are redistributed, with less pronounced differences in Hop1 occupancy distributions measured either cytologically (Börner et al. 2008;Joshi et al. 2009) or by chromatinimmunoprecipitation (Medhi et al. 2016;Subramanian et al. 2019). Previously, we found that the absence of Pch2 did not substantially alter overall NCO or CO levels at HIS4 and URA3, but the Mlh3dependence of CO formation was affected at both loci, with Mlh3independent COs increasing at HIS4 and decreasing at URA3. Because the six new VRS insert loci studied here are similar to URA3, in that most VDE-initiated COs are Mlh3-independent, we wanted to see if COs at these loci also displayed increased Mlh3-dependence in pch2Δ mutants.
Consistent with previous findings (Börner et al. 2008), meiotic divisions were delayed in pch2Δ and pch2Δ mlh3Δ mutants relative to wild type ( Figure S2A). Frequencies of NCOs at all eight VRS insert loci in the pch2Δ were similar to those seen in wild type (Figures 3C and S2C; pch2Δ/PCH2 = 111 6 10%), as were COs (Figures 3A and S2C; pch2Δ/PCH2 = 113 6 16%). Loss of Mlh3 did not substantially affect NCOs ( Figure 3C; pch2Δ mlh3Δ/pch2Δ MLH3 = 114 6 14%). However, in pch2Δ mlh3Δ double mutants, COs were reduced 20-35% relative to pch2Δ MLH3 ( Figure 3B; average pch2Δ mlh3Δ/pch2Δ = 74 6 7%), as was previously observed for inserts at URA3 and HIS4 (Medhi et al. 2016). A quantitatively similar MutLg-dependence has also been observed for Spo11-initiated COs in pch2Δ mutants, both genome-wide Spo11-initiated COs are reduced about twofold in mutants lacking MutLg; this is thought to reflect unbiased JM resolution by SSNs to form both COs and NCOs, as opposed to MutLg-mediated biased JM resolution as COs in wild type (Argueso et al. 2004;Zakharyevich et al. 2012). If the same holds true for pch2 mutants, the 25% reduction in COs seen in pch2Δ mlh3Δ would suggest that about half of the COs formed in pch2 MLH3 cells are the products of MutLg-mediated resolution, regardless of whether they were initiated by VDE or by Spo11. It therefore appears that Pch2, or processes regulated by it, prevents VDE-initiated events from forming MutLg-dependent COs.

Summary and concluding remarks
In this study, we examined VDE-initiated meiotic recombination in a recombination reporter inserted at six loci in addition to the two loci (HIS4 and URA3) originally examined by Medhi et al. (2016). With the exception of HIS4, VDE-initiated COs at all insert loci were largely Mlh3-independent, regardless of whether inserts were at loci in Hop1-enriched or Hop1-depleted regions of the genome. Therefore, our previous hypothesis, that local Hop1 occupancy determines Figure 2 VDE-initiated crossovers at most loci are MutLg-independent. (A) Strategy for detection of VDE-initiated COs and NCOs. A cartoon of the VRS and VRS-103 inserts is shown, illustrating: white box-VRS sequences; blue arrows-HindIII restriction sites; green lines-sequences shared between the two inserts, with ARG4 coding sequences shown as a green arrow; green box-sequences used for Southern blot probes. Digestion with HindIII and PI-SceI (VDE) distinguishes parental (P1 and P2), CO and NCO products. VDE-cut inserts are not distinguished from parent P1 in these digests, but can be distinguished in digests with HindIII alone (Medhi et al. 2016). (B) Representative Southern blot containing DNA from strains with inserts at RIM15. (C) VDE-initiated COs in MLH3 and mlh3Δ cells. CO frequencies, average signal of CO1 and CO2 for 8 and 9 h samples from three independent experiments for inserts at HIS4 and from two independent experiments for inserts at all other loci. Data for inserts at URA3 and for two experiments with inserts at HIS4 are from Medhi et al. (2016). (D) fraction of COs that are MutLg-independent (ratio of CO frequencies in mlh3Δ vs. MLH3), plotted as a function of CO frequencies in MLH3 strains. CO frequencies in MLH3 and mlh3Δ differ significantly only for inserts at HSP30 and HIS4 (adjusted p values of 0.003 and 0.0001, respectively) (E,F) VDE-initiated NCOs, details as in (B) and (C); frequencies in MLH3 and mlh3Δ do not differ significantly at any locus (adjusted p values $ 0.05). Error bars in all panels denote standard deviation. See Figure S2C for summary plots with CO and NCO values for all genotypes.
mechanisms of CO formation, is inaccurate, at least for VDE-initiated recombination, in that it was based on analysis of inserts at a locus (HIS4) that appears to be exceptional. We currently do not understand why the genetic dependence of VDE-induced CO formation at HIS4 differs from that seen at the other loci examined, but cannot rule out other effects or interactions on the basis of current data.
The observation that VDE-initiated COs at most insert loci are Mlh3-independent, in turn, raises the question of whether or not VDE-initiated recombination events that occur in cells undergoing meiosis can be properly described as being "meiotic". VDE-initiated NCOs are recovered in excess of COs (2 to 5-fold, average 3.2 6 0.1), which is reminiscent of, although less than, the 5 to 20-fold excess of NCOs over COs seen in budding yeast mitotic recombination (Esposito 1978;Lichten and Haber 1989;Ira et al. 2003;Bzymek et al. 2010;Dayani et al. 2011). VDE-initiated DSB processing also resembles DSB processing in the mitotic cell cycle, in that break ends are continuously resected over time (Lee et al. 1998;Neale et al. 2002;Johnson et al. 2007), unlike the constrained resection seen with Spo11 DSBs (Mimitou et al. 2017). Finally, unlike Spo11, VDE frequently cuts both sister chromatids in a single meiosis (Gimble and Thorner 1992;Gimble and Thorner 1993;Medhi et al. 2016), and gene conversion of both sister chromatids is associated with a reduced CO/NCO ratio among HO endonuclease-initiated meiotic recombinants (Malkova et al. 2000). Further studies will be necessary to determine which of these or other factors are responsible for the marked Mlh3-independence of VDE-initiated COs at seven of the eight insert locations examined, and why the majority of VDE-initiated COs at HIS4 are Mlh3-dependent.
In contrast, in pch2Δ strains, VDE-initiated COs show the same Mlh3-dependence as Spo11-initiated COs, regardless of wild-type Hop1 occupancy levels around insert loci. It therefore seems unlikely that Hop1 redistribution in pch2Δ mutants is the only factor responsible for the increased Mlh3-dependence of COs at most insert loci and the decreased Mlh3-dependence of COs at HIS4. Homolog synapsis, recombinant formation and meiotic divisions are all delayed in pch2Δ mutants; pch2Δ mutants also display a more even distribution of the Zip1 central element protein along chromosomes and reduced CO interference (Börner et al. 2008;Joshi et al. 2009;Zanders and Alani 2009). These or other pch2Δ mutant defects might delay either recruitment of factors promoting MutLg action at Spo11-initiated events or implementation of CO interference, thus increasing the window of opportunity for these factors to act at VDE-initiated events. It also has been suggested that Pch2-dependent remodeling affects the stability of recombination intermediates (Deshong et al. 2014), and it is possible that this might differentially affect events not initiated by Spo11. Exogenous DNA damage is unable to fully rescue spo11 mutant phenotypes in several organisms (Thorne and Byers 1993;Celerin et al. 2000;Romanienko and Camerini-Otero 2000;Loidl and Mochizuki 2009;Carofiglio et al. 2018), and budding yeast responds differently during meiosis to DSBs induced by Spo11 and to DSBs formed by exogenous DNA damage (Cartagena-Lirola et al. 2008;reviewed in Longhese et al. 2008). Our current data suggest that Pch2 might implement mechanisms that distinguish Spo11-initiated recombination events from those initiated by other forms of DNA damage.
In summary, the data presented here indicate that VDE-initiated recombination events are treated differently than are those initiated by For inserts at CCT6, IMD3, FIR1 and RIM15, 9 h values are from a single experiment. Data for inserts at HIS4 and URA3 are from Medhi et al. (2016). (B) fraction of COs that are MutLg-independent (ratio of CO frequencies in mlh3Δ vs. MLH3), plotted as a function of CO frequencies in MLH3 strains. CO frequencies in MLH3 pch2Δ and mlh3Δ pch2Δ differ significantly for inserts at all loci (adjusted p values # 0.03) except CCT6 and RIM15 (adjusted p values of 0.06 and 0.07, respectively). (C, D) VDE-initiated NCOs, as in panels (A) and (B). NCO frequencies in MLH3 pch2Δ and mlh3Δ pch2Δ do not differ significantly for any locus (adjusted p values $ 0.05). Error bars in all panels denote standard deviation. See Figure  S2C for summary plots with CO and NCO values for all genotypes.
Spo11 during wild-type meiosis. VDE-initiated events produce an excess of NCOs over COs and, at seven of eight loci examined, form COs by MutLg-independent mechanisms, and thus their outcome more closely resembles those of DSB repair events that occur during the mitotic cell cycle. We conclude that the full spectrum of meiotic recombination processes that occur at Spo11-initiated DSBs do not occur at VDE-initiated DSBs, and, by inference, DSBs formed during meiosis by other nucleases. Thus, our findings call for caution in the use of DSBs formed by this nuclease, or by other exogenous means, for inferring factors that control normal meiotic recombination.

ACKNOWLEDGMENTS
We thank Jean Paul Ouyan, Seyoun Kim, and Matan Cohen for help in strain construction, and Jasvinder Ahuja, Needhi Bhalla, Matan Cohen, Julia Cooper, and Martin Xaver for comments and discussion. This work was supported by the Intramural Research Program of the NIH through the Center for Cancer Research at the National Cancer Institute.