Copyright  2001 by the Genetics Society of America

The Budding Yeast Msh4 Functions in Synapsis and the Regulation of Crossover Distribution

Janet E. Novak,* Petra B. Ross-Macdonald*,1 and G. Shirleen Roeder*,†,‡ *Department of Molecular, Cellular and Developmental Biology, †Howard Hughes Medical Institute and ‡Department of Genetics, Yale University, New Haven, Connecticut 06520-8103 Manuscript received February 9, 2001 Accepted for publication April 24, 2001

ABSTRACT The budding yeast MSH4 encodes a MutS homolog produced specifically in meiotic cells. Msh4 is not required for meiotic mismatch repair or gene conversion, but it is required for wild-type levels of crossing over. Here, we show that a null mutation substantially decreases crossover interference. With respect to the defect in interference and the level of crossing over, msh4 is similar to the zip1 mutant, which lacks a structural component of the synaptonemal complex (SC). Furthermore, epistasis tests indicate that msh4 and zip1 affect the same subset of meiotic crossovers. In the msh4 mutant, SC formation is delayed compared to wild type, and full synapsis is achieved in only about half of all nuclei. The simultaneous defects in synapsis and interference observed in msh4 (and also zip1 and ndj1/tam1) suggest a role for the SC in mediating interference. The Msh4 protein localizes to discrete foci on meiotic and colocalizes with Zip2, a protein involved in the initiation of chromosome synapsis. Both Zip2 and Zip1 are required for the normal localization of Msh4 to chromosomes, raising the possibility that the zip1 and zip2 defects in crossing over are indirect, resulting from the failure to localize Msh4 properly.

EDUCTIONAL is unique match correction; instead, they are required specifically R to the first division of . Sister chromatids for wild-type levels of meiotic crossing over. In the ab- remain associated throughout this division, while ho- sence of Msh4 or Msh5, meiotic gene conversion occurs mologous chromosomes segregate to opposite poles of at approximately wild-type levels, but crossing over is the spindle apparatus. The chromosome content of a reduced two- to threefold. The MSH4 and MSH5 diploid cell is thereby reduced to the haploid number are expressed specifically in meiotic cells (Ross-Mac- of chromosomes. A prerequisite to proper chromosome donald and Roeder 1994; Chu et al. 1998). The pro- segregation at meiosis I is meiotic recombination. Cross- teins directly interact to form a complex (Pochart et ing over establishes chromatin bridges between homo- al. 1997), and they colocalize to foci on meiotic chromo- logs, called chiasmata, that ensure the proper orienta- somes (J. E. Novak, unpublished data). In the absence tion of chromosomes on the meiosis I spindle (Roeder of Msh4 (and presumably also Msh5), chromosomes 1997). The proper segregation of chromosomes in mei- sometimes fail to undergo crossing over (Ross-Mac- osis also depends on formation of the synaptonemal donald and Roeder 1994). Because nonrecombinant complex (SC), an elaborate proteinaceous structure chromosomes often missegregate at meiosis I, the msh4 that holds homologous chromosomes close together and mutants display reduced spore viability. along their lengths during the pachytene stage of mei- The recombination phenotype of msh4 and msh5 is otic prophase (Roeder 1997). Mutations in genes not unique to these mutants. Mutations in a number of encoding structural components of the SC lead to ho- other yeast genes also reduce the frequency of meiotic molog nondisjunction at meiosis I and precocious sepa- crossing over without decreasing the frequency of non- ration of sister chromatids (Roeder 1997). Mendelian segregation. These genes include ZIP1, ZIP2, Two budding yeast involved in meiotic re- ZIP3, MER3, MLH1, MLH3, and EXO1. The three ZIP combination are Msh4 and Msh5 (Ross-Macdonald genes are involved in SC formation. The Zip1 protein and Roeder 1994; Hollingsworth et al. 1995), which is present along the lengths of synapsed meiotic chromo- are homologs of the bacterial MutS mismatch repair somes and serves as a major structural component of protein. However, Msh4 and Msh5 play no role in mis- the SC (Sym et al. 1993; Sym and Roeder 1995; Tung and Roeder 1998; Dong and Roeder 2000). Zip2 and Zip3 are present on meiotic chromosomes at discrete Corresponding author: G. Shirleen Roeder, Howard Hughes Medical foci that correspond to the sites where synapsis initiates, Institute, Department of Molecular, Cellular and Developmental Biol- and these proteins are required for the normal polymer- ogy, Yale University, P.O. Box 208103, New Haven, CT 06520-8103. E-mail: [email protected] ization of Zip1 along chromosomes (Chua and Roeder 1 Current address: Bristol-Myers Squibb, Applied Genomics, P.O. Box 1998; Agarwal and Roeder 2000). The MER3 gene is 5400, Princeton, NJ 08543-5400. expressed specifically in meiotic cells and encodes a

Genetics 158: 1013–1025 ( July 2001) 1014 J. E. Novak, P. B. Ross-Macdonald and G. S. Roeder putative helicase (Nakagawa and Ogawa 1999). The significantly reduces crossover interference and results two MutL homologs, Mlh1 and Mlh3, form a hetero- in SC formation that is delayed and often incomplete. dimer specifically involved in meiotic crossing over Msh4 colocalizes with the Zip2 protein at sites of synapsis (Hunter and Borts 1997; Wang et al. 1999); Mlh1 also initiation and depends on both Zip2 and Zip1 for functions in mismatch repair both in vegetative and proper localization to chromosomes. Analysis of a msh4 meiotic cells (Kolodner and Marsischky 1999). The zip1 double mutant indicates that Msh4 and Zip1 act in Exo1 protein is a 5Ј to 3Ј exonuclease specific for dou- the same pathway of crossing over. ble-stranded DNA (Huang and Symington 1993; Fio- rentini et al. 1997). In addition to its role in meiotic MATERIALS AND METHODS crossing over (Khazanehdari and Borts 2000; Kirk- patrick et al. 2000; Tsubouchi and Ogawa 2000), Exo1 Plasmids, disruptions, and strains: Media were prepared is involved in mismatch repair and recombination in and yeast manipulations were carried out using standard pro- vegetative cells (Fiorentini et al. 1997; Tishkoff et al. cedures (Sherman et al. 1986). All yeast transformants were verified by Southern blot analysis. Yeast strain genotypes are 1997). given in Table 1. Meiotic crossovers are nonrandomly distributed along MSH4 gene disruptions were engineered as follows. chromosomes such that two crossovers rarely occur close pmsh4⌬85-2395 was derived from p6K (Ross-Macdonald and together—a phenomenon known as crossover interfer- Roeder 1994) by inserting linkers containing a NotI site into ence. Interference is generally assumed to involve the a BstBI site at nucleotide 80 and an SspI site at nucleotide 2393 within the MSH4 coding region. Appropriate NotI-BamHI transmission of an inhibitory signal from one crossover and NotI-EcoRI fragments were ligated into pUC18 cut with site to nearby potential sites of crossing over. In budding EcoRI and BamHI to yield pmsh4⌬85-2395, in which nucleo- yeast, mutations in three different genes—ZIP1, NDJ1 tides 85–2395 were replaced with GCGGCCGCAA. To create ⌬ ⌬ (a.k.a. TAM1), and MER3—have been shown to reduce pmsh4 ADE, pmsh4 85-2395 was digested with NotI; the ends were filled in with the Klenow fragment of DNA polymerase or abolish crossover interference (Sym and Roeder I and a 3.7-kb blunt-ended fragment containing the ADE2 1994; Chua and Roeder 1997; Nakagawa and Ogawa gene was inserted. The msh4::ADE2 mutation was introduced 1999). A zip1 null mutation abolishes SC formation into yeast by substitutive transformation (Rothstein 1991) (Sym et al. 1993), while an ndj1 null mutation causes a using pmsh4⌬ADE digested with NcoI and BamHI. p6H, con- substantial delay in SC formation (Chua and Roeder taining msh4::LEU2, was derived from p5E (Ross-Macdonald and Roeder 1994) by cutting with NdeI and circularizing the 1997; Conrad et al. 1997). mer3 has not been tested for fragment containing MSH4 sequences, resulting in a plasmid its effect on synapsis. The observed correlation between in which part of a transposon (including LEU2) replaces nucle- impaired synapsis and decreased interference in mu- otides 168–2290 of the MSH4 coding region. The msh4::LEU2 tants of yeast and other organisms (Moens 1969; Hav- disruption was introduced into yeast by transformation with ekes et al. 1994) is consistent with the hypothesis that p6H cut with NotI. Strains producing tagged versions of the Msh4 protein were the SC is involved in transmission of the inhibitory signal constructed as follows. In pU-Msh4-HA, a SacI-XhoI fragment responsible for interference (Egel 1978; Maguire 1988). containing the MSH4 gene marked with three copies of the The observation that Schizosaccharomyces pombe and Asper- HA tag at the 3Ј end was obtained from pJ8 (Ross-Macdonald gillus nidulans lack both SC and interference further and Roeder 1994) and inserted between the XhoI and the SacI sites of pRS306 (Sikorski and Hieter 1989). The MSH4- supports this idea (Strickland 1958; Olson and Zim- HA gene was integrated (Rothstein 1991) at URA3 by trans- mermann 1978; Egel-Mitani et al. 1982; Bahler et al. formation with pU-Msh4-HA targeted with StuI. pkan-Msh4C- 1993). 3ϫHA was designed to allow simultaneous disruption of MSH4 Meiotic crossovers are nonrandomly distributed not and insertion of MSH4-HA. To make pkan-Msh4C-3ϫHA, p10H only along chromosomes, but also among chromo- (Ross-Macdonald and Roeder 1994) was cut with BstUI and EcoRI, and the 2.4-kb fragment containing the downstream somes. In most meioses, every chromosome pair, no portion of MSH4-HA was inserted into pFA6-Kan-MX4 (Wach matter how small, sustains at least one crossover—a et al. 1994) cut with EcoRI and EcoRV. The msh4::MSH4-HA- so-called obligate crossover or obligate chiasma. Obli- kan allele was introduced into yeast by transformation with gate crossing over might be due, at least in part, to pkan-Msh4C-3ϫHA cut with BsrGI to target integration at the the fact that large chromosomes display more crossover MSH4 . pU-Zip2C-GFP contains part of the ZIP2 gene tagged with interference than do small chromosomes (Kaback et al. green fluorescent protein (GFP; Chua and Roeder 1998). 1999). By preventing excess exchanges on large chro- To generate pU-Zip2C-GFP, a 2.1-kb BamHI-SalI fragment con- mosomes, interference might ensure that crossovers taining the downstream portion of ZIP2::GFP (Chua and (generally presumed to be limited in number) are dis- Roeder 1998) was inserted between the BamHI and SalI sites tributed such that every chromosome pair undergoes of pRS306 (Sikorski and Hieter 1989). The zip2::ZIP2-GFP- URA3 allele was introduced into yeast by integrating pU-Zip2C- at least one exchange. Consistent with this hypothesis, GFP (targeted with EcoRI) at ZIP2. Other ZIP2 alleles were mutations that reduce or eliminate interference ran- introduced as described (Chua and Roeder 1998). domize the distribution of crossovers among chromo- pNDT80⌬kan was constructed from pTP77 (Tung et al. somes such that chromosomes sometimes fail to cross 2000). A 1.1-kb BglII-EcoRI fragment in NDT80 was replaced with a 1.4-kb BglII-EcoRI fragment from pFA6-Kan-MX4 (Wach over and therefore nondisjoin (Sym and Roeder 1994; et al. 1994) containing kanr. NDT80 was disrupted by transfor- Chua and Roeder 1997). mation with pNDT80⌬kan targeted with XbaI and BamHI. We have found that a msh4 null mutation (like ndj1) p11B, which contains SER1 disrupted by a transposon, was Msh4 Functions in Synapsis and Interference 1015

TABLE 1 Yeast strains

Strain Genotype RM86a MATa CEN3::TRP1 his4-Bgl leu2 CAN1 URA3 HOM3 TRP2 MAT␣ CEN3 HIS4 leu2 can1 ura3 hom3 trp2 ho::lys2 trp1 lys2 ade2-Bgl msh4::LEU2 ho::lys2 trp1 lys2 ade2-Bgl MSH4 RM85a RM86 but homozygous msh4::LEU2 RM96a HIS3 SER1 ADE2 RM86 but his3-Nde ser1::mTn-3xHA/lacZ ade2-Bgl RM95a RM96 but homozygous msh4::LEU2 RM97a MATa CEN3::TRP1 his4-leu2 leu2 CAN1 URA3 HOM3 TRP2 MAT␣ CEN3 HIS4 leu2 can1 ura3 hom3 trp2 ho::lys2 trp1 zip1::LYS2 lys2 his3-Nde ADE2 msh4::LEU2 ho::lys2 trp1 ZIP1 lys2 HIS3 ade2-Bgl MSH4 RM98a RM97 but homozygous zip1::LYS2 RM99a RM97 but homozygous msh4::LEU2 RM100a RM97 but homozygous zip1::LYS2 msh4::LEU2 RM53b MATa his4-Bgl LEU2 arg4-Nsp msh4::URA3 lys2 ho::LYS2 ura3 MAT␣ HIS4 leu2-K ARG4 MSH4 lys2 ho::LYS2 ura3 RM51b msh4::ura3 RM53, but msh4::URA3 JN305b RM53, but homozygous ndt80::kanr JN304b RM51, but homozygous ndt80::kanr BR2495 MATa leu2-27 his4-280 arg4-8 thr1-4 cyh10 trp1-1 ade2-1 ura3-1 MAT␣ leu2-3, 112 his4-260 ARG4 thr1-4 CYH10 trp1-289 ade2-1 ura3-1 JN295 BR2495, but homozygous zip2::ZIP2-GFP-URA3 msh4::MSH4-HA-kanr JN294 zip1::LEU2 LYS2 SPO13 JN295, but zip1::LEU2 lys2 spo13::ura3-1 JN202 msh4::ADE2 MSH4-HA@URA3 thr1-4 his4-280 BR2495, but msh4::ADE2 MSH4-HA@URA3 THR1 HIS4 JN220 BR2495, but homozygous msh4::ADE2 MSH4-HA@URA3 zip2::URA3 JN232 BR2495, but homozygous msh4::ADE2 MSH4-HA@URA3 zip1::LYS2 lys2⌬Nhe MY63 BR2495, but homozygous zip1::LEU2 a These strains are congenic with SK1 (Kane and Roth 1974). b These strains are isogenic with SK1. constructed as described (Ross-Macdonald et al. 1997). The TT Ϫ (1 Ϫ3TT/2)2/3], and rounding the resulting figure to ser1::mTn-3ϫHA/lacZ allele was introduced into yeast using the nearest whole number. The ratio of NPDs observed to p11B targeted with NotI. pR1723, containing his3-Nde, was NPDs expected was then compared using the chi-square test derived from pRS303 (Sikorski and Hieter 1989) by first as described (Sym and Roeder 1994). To compare the NPD filling in the NdeI site in HIS3 and then inserting a SmaI-EagI ratio calculated for the wild type with that calculated for the fragment containing URA3 from YIp5 between the SmaI and msh4 mutant, the chi-square coincidence test was applied to EagI sites. The his3-Nde and ade2-Bgl alleles were introduced the values for observed and expected NPDs. by two-step transplacement using pR1723 (targeted with NheI) Cytology: Strains isogenic with BR2495 were grown and and pR943 (Engebrecht and Roeder 1990), respectively. sporulated as described (Rockmill et al. 1995). Strains iso- zip1::LEU2 and zip1::LYS2 alleles were introduced as described genic or congenic with SK1 were prepared for meiosis by first (Sym et al. 1993; Sym and Roeder 1994). growing to saturation in YPAD. Cells were then diluted 200- Њ Genetic analysis: To calculate map distance, only four-spore- fold into YPA and grown at 30 for 15 hr, to an OD600 of Cells were then rinsed in 2% potassium acetate that .0.9ف viable tetrads that did not show gene conversion of relevant markers were used. Crossover frequencies were compared as was warmed to 30Њ, pelleted, and resuspended in warm 2% described (Chua and Roeder 1997). potassium acetate at a volume equal to their volume in YPA. For interference analysis, the number of nonparental di- Cells were shaken at 30Њ, and samples were harvested at various types (NPDs) expected in a particular interval was derived by time points. Surface-spread meiotic nuclei were prepared as applying the formula of Papazian (1952), NPD ϭ 1/2[1 Ϫ described (Sym et al. 1993), except that glass slides were not 1016 J. E. Novak, P. B. Ross-Macdonald and G. S. Roeder coated with plastic. Samples taken to assess timing of cell by the number expected (Snow 1979). An NPD ratio Ј divisions were fixed and stained with 4 ,6-diamidino-2-phenyl- Ͻ1.0 indicates that crossover interference is operating. indole (DAPI) as described (Engebrecht and Roeder 1990). Immunofluorescence on spread chromosomes was per- Tetrads from wild-type and msh4 strains were dissected formed as described (Sym et al. 1993), with the following and analyzed for recombination in seven genetic inter- modifications. The solution for blocking and antibody dilu- vals representing three different chromosomes (Table tions was 3% bovine serum albumin, 1% teleostean gelatin 2). Map distances in msh4 strains were only 40–60% of (Sigma, St. Louis), and 0.02% thimerosal in phosphate-buf- those in wild type (Table 2), with the exception of the fered saline. Rabbit anti-Zip1 antibodies (Sym et al. 1993) were diluted 200-fold and detected with Cy3-conjugated secondary HOM3–TRP2 interval, which was unaffected (Table 2). antibodies (Jackson ImmunoResearch Laboratories, West In wild type, NPD ratios ranged from 0.13 to 0.34, indi- Grove, PA). Rat anti-tubulin antibodies (Kilmartin et al. cating significant interference (Figure 1); the only ex- 1982) were diluted 500-fold and detected with FITC-conju- ception was the HOM3–TRP2 interval, which did not gated secondary antibodies (Jackson ImmunoResearch Labo- display any interference in the strain background used ratories). Hemagglutinin (HA)-tagged Msh4 was detected with mouse monoclonal antibody HA-11 (Covance) diluted 5000- for this analysis. Excluding the HOM3–TRP2 interval, fold followed by secondary antibodies conjugated to either NPD ratios in the msh4 mutant were consistently higher Texas red or Cy3 (Jackson ImmunoResearch Laboratories). than those in wild type (Figure 1); this difference is GFP-tagged Zip2 was detected with rabbit anti-GFP antibodies statistically significant for the ADE2–SER1, CAN1–URA3, (CLONTECH, Palo Alto, CA) diluted 300-fold and secondary and URA3–HOM3 intervals (Table 2). antibodies conjugated with Oregon Green 488 (Molecular Probes, Eugene, OR). Red1 protein was detected using rabbit For the intervals examined on chromosomes V and anti-Red1 serum (Smith and Roeder 1997) diluted 100-fold XV, NPD ratios for the msh4 mutant were close to 1.0, followed by Oregon Green 488-conjugated secondary antibod- as expected in the complete absence of interference ies. All antibody binding was carried out at room temperature. (Figure 1). However, for the two intervals examined on Images were captured using a Nikon Eclipse E800 microscope chromosome III, NPD ratios were Ͻ1.0; the ratios for and a Photometrics (Tucson, AZ) Sensys CCD camera. To quantify the colocalization of Msh4 and Zip2, the follow- HIS4–CEN3 and CEN3–MAT were 0.67 and 0.38, respec- ing procedure was used. First, image contrast and brightness tively. For another test of interference on chromosome were adjusted using Adobe Photoshop to compensate for the III, NPD ratios were calculated for the HIS4–MAT inter- faintness of Msh4 foci in JN294. Msh4 and Zip2 foci that were val (which subsumes the HIS4–CEN3 and CEN3–MAT judged to be brighter than background staining were copied intervals; Figure 1). For this interval, the NPD ratio separately onto transparent sheets in order to have a defined Ͻ number of foci with distinct edges. The outlines of each spread obtained in the msh4 mutant (0.68) was 1.0, but greater were traced from the DAPI-stained DNA. Superposition of the than the corresponding NPD ratio in wild type (0.36); sheets was used to count Msh4 and Zip2 foci that overlapped. both differences are statistically significant. Together, To estimate the amount of overlap that would be found with these data indicate that the msh4 mutation abolishes Њ random positioning, the copy of Zip2 foci was rotated 180 crossover interference in some intervals and signifi- to randomize the position of the Zip2 foci with respect to the Msh4 foci. Any foci falling outside the area where the spread cantly reduces interference in other intervals. chromosomes overlapped were not included in the analysis. In theory, the observed increases in NPD ratios in the Within the overlapping region, the number of Msh4 and Zip2 msh4 mutant could be due to an acquisition of chroma- foci and the number of overlapping foci were counted. Ap- tid interference instead of a loss of crossover interfer- proximately 400 foci of each protein were scored for both ence. Chromatid interference would increase the num- JN294 and JN295. Strain MY63, which has only untagged Msh4, was used to ber of four-strand double crossovers at the expense of assess the background level of staining with antibodies against two-strand and three-strand double crossovers. [There HA. Typically, spread nuclei from MY63 showed no foci. is no chromatid interference in wild-type yeast (Morti- To quantify Msh4 staining in wild-type and mutant spreads, mer and Fogel 1974).] Chromatid interference would Imagepoint IPLab Spectrum software was used to measure have to be quite strong to account for the excess of total intensity within each spread; background intensity was subtracted. At least 17 spreads for each strain were measured. NPDs in the msh4 mutant; for the six intervals that show The average values for the strains were found to be significantly interference in the wild type, the NPD ratios in wild type different (P Ͻ 0.001) by a two-tailed t-test. average one-quarter those of the msh4 mutant. Thus, all or nearly all of the double crossovers occurring within these intervals in the mutant would have to be four- RESULTS strand double crossovers. Such strong chromatid inter- A msh4 mutant is defective in crossover interference: ference should be easily detected by examining tetrads Crossover interference can be detected as a deficit in with single crossovers in two adjacent intervals to deter- double crossovers within a genetically marked interval, mine the total number of chromatids involved. In both compared to the number expected based on the fre- msh4 and wild type, double crossovers conformed to the quency of single crossovers. In yeast, four-strand double 1:2:1 ratio of two-strand:three-strand:four-strand events crossovers generate NPD tetrads. The frequency of expected in the absence of chromatid interference (Ta- NPDs expected in the absence of interference can be ble 3). This analysis rules out the possibility that the calculated from the observed frequency of tetratype increased NPD ratios in msh4 are due to chromatid (TT) tetrads due to single crossovers (Papazian 1952). interference and reinforces the conclusion that the The NPD ratio is the number of NPDs observed divided Msh4 protein is required for crossover interference. Msh4 Functions in Synapsis and Interference 1017

TABLE 2 Tetrad analysis

Interval Strain PDa TT NPD NPD exp.b Prob.c cMd %wte ADE2–SER1 MSH4 164 238 8 34 Ͻ0.01 34.9 ADE2–SER1 msh4 591 186 6 7 0.70 14.2 41 SER1–HIS3 MSH4 313 121 1 5 0.07 14.6 SER1–HIS3 msh4 684 81 1 1 1.00 5.7 39 CAN1–URA3 MSH4 339 663 18 135 Ͻ0.01 37.8 CAN1–URA3 msh4 1019 394 17 17 1.00 17.3 46 URA3–HOM3 MSH4 342 636 28 115 Ͻ0.01 40.0 URA3–HOM3 msh4 969 439 22 22 1.00 20.0 50 HOM3–TRP2 MSH4 706 205 8 7 0.70 13.8 HOM3–TRP2 msh4 1265 395 16 14 0.59 14.6 106 HIS4–CEN3 MSH4 682 697 24 71 Ͻ0.01 30.0 HIS4–CEN3 msh4 1571 648 20 30 0.07 17.2 57 CEN3–MAT MSH4 1026 369 4 15 Ͻ0.01 14.0 CEN3–MAT msh4 1917 358 3 8 0.08 8.3 59 Strains RM86 (wild type) and RM85 (msh4) were used to examine the CAN1–URA3, URA3–HOM3, HOM3– TRP2, HIS4–CEN3, and CEN3–MAT intervals. Strains RM96 (wild type) and RM95 (msh4) were used to examine all intervals except CAN1–URA3 and URA3–HOM3. In wild type, spore viability was 94%, and 84% of tetrads contained four viable spores. In msh4, spore viability was 52%, and 34% of tetrads contained four viable spores. a Parental ditype tetrads. b NPD tetrads expected (Papazian 1952). c Probability based on a chi-square test that the difference between the observed and expected numbers of NPDs is due to chance. d Map distance in centimorgans. e The map distance in the mutant as a percentage of the map distance in wild type.

Msh4 and Zip1 act in the same recombination path- map distances in the zip1 msh4 double mutant are similar way: To determine whether Msh4 and Zip1 operate in to both single mutants; no statistically significant differ- the same pathway, meiotic recombination and chromo- ences were observed between the double mutant and some segregation were examined in the msh4 zip1 dou- either single mutant strain for the six intervals examined ble mutant. The results of tetrad analysis indicate that (Table 4). Also, the frequency and pattern of spore

Figure 1.—The msh4 mutation reduces or eliminates interfer- ence. (A) Genetic map of the in- tervals tested. Gene order, the sizes of the intervals tested, and the overall size of each chromo- some are indicated. (B) Interfer- ence values (NPD observed/NPD expected) were calculated for the seven intervals indicated. A ratio of 1.0 (dotted line) is indicative of no interference. 1018 J. E. Novak, P. B. Ross-Macdonald and G. S. Roeder

TABLE 3 Test for chromatid interference in adjacent intervals

Intervals Strain Two strands Three strands Four strands P valuea ADE2–SER1–HIS3 MSH4 18 22 7 0.27 ADE2–SER1–HIS3 msh4 8 10 5 0.76 HIS4–CEN3–MAT MSH4 41 73 36 0.90 HIS4–CEN3–MAT msh4 13 50 22 0.27 CAN1–URA3–HOM3 MSH4 74 153 73 0.97 CAN1–URA3–HOM3 msh4 34 52 34 0.59 For the pairs of adjacent intervals indicated, all tetrads that were tetratype in both intervals were classified as two-strand, three-strand, or four-strand double crossovers according to the number of chromatids involved. a Probability, based on a chi-square test, that the ratio of two-, three-, and four-strand double crossovers observed is consistent with the 1:2:1 ratio expected in the absence of chromatid interference. viability in the double mutant were similar to those of sess spindle formation. The DNA-binding dye DAPI was the single mutants (Figure 2). In every case, tetrads used to visualize chromosomes. Nuclei were classified containing 4, 2, or 0 viable spores were present in excess, according to the extent of Zip1 localization and the compared to tetrads containing 3 or 1 viable spores; presence or absence of a meiosis I spindle as described this pattern is indicative of meiosis I nondisjunction (Figure 3, A–D; Smith and Roeder 1997). events (Sym and Roeder 1994). The frequency of chro- Nuclei with partly continuous Zip1 staining (indica- mosome III nondisjunction was estimated from the fre- tive of incomplete synapsis) appeared at the same time quency of two-spore-viable tetrads in which both spores in the mutant as in wild type, but these nuclei were are disomic for chromosome III (and therefore mating longer lived and accumulated to a higher level in msh4 incompetent due to heterozygosity at the MAT locus). (Figure 3B). By 4.5 hr, wild-type cells were at the peak of Such tetrads represented 5.2% of total tetrads in the pachytene, while msh4 cells displayed very few pachytene zip1 strain, 4.9% in the msh4 strain, and 4.4% in the zip1 nuclei (Figure 3C). In this and other time courses (not msh4 strain. Thus, by all three criteria—map distances, shown), the maximum number of pachytene nuclei was spore viability, and chromosome III nondisjunction— reduced approximately twofold, and the time point at Msh4 and Zip1 act in the same pathway. which maximal synapsis was observed was often delayed Chromosome synapsis is delayed and often incom- compared to wild type (Figure 3C). The appearance of plete in the msh4 mutant: Although pachytene nuclei nuclei containing meiosis I spindles was consistently have been observed in the msh4 mutant (Ross-Macdon- delayed in msh4 (Figure 3D). ald and Roeder 1994), the extent and timing of SC Polycomplexes are structured aggregates of SC com- formation were not quantitated. To test the possibility ponents, including Zip1. Polycomplexes are observed that the msh4 mutation reduces the efficiency of chro- in only a minority of meiotic prophase nuclei from wild mosome synapsis, SC formation was examined in msh4 type, but they are almost always present in strains over- and wild-type cells by staining spread meiotic chromo- producing Zip1 and in most mutants defective in SC somes with anti-Zip1 antibodies. Spread nuclei were si- formation (Loidl et al. 1994; Sym and Roeder 1995). multaneously stained with anti-tubulin antibodies to as- Polycomplexes can be detected as regions of intense

TABLE 4 Epistasis analysis: map distances from tetrad analysis

Map distance (cM) Strain Relevant genotype HIS4–CEN3 CEN3–MAT CAN1–URA3 URA3–HOM3 HOM3–TRP2 HIS3–ADE2 RM97 Wild type 38.3 23.1 39.8 61.2 22.2 54.0 RM99 msh4 12.8 4.5 13.2 14.6 9.8 25.5 RM98 zip1 7.9 7.9 11.8 25.5 12.6 30.8 RM1000 zip1 msh4 10.5 6.0 23.0 17.9 13.0 31.1 For each interval, 80–100 tetrads containing four viable spores were analyzed. The exception is the HIS3–ADE2 interval in strains RM97 and RM98, for which 63 and 78 tetrads were analyzed, respectively. Recombination frequencies for msh4 and zip1 are significantly different (P Ͻ 0.05) from those in wild type, except for the HOM3–TRP2 interval in the zip1 mutant (P ϭ 0.126). In no case is the recombination frequency in the msh4 zip1 double mutant significantly different from that in the msh4 or zip1 single mutant. Msh4 Functions in Synapsis and Interference 1019

Figure 2.—Spore viability in msh4 and zip1 mutant strains. The distribution of four-, three-, two-, one-, and zero- spore-viable tetrads is shown for wild-type (RM97), msh4 (RM99), zip1 (RM98), and msh4 zip1 (RM100) strains. The number of tetrads dissected was 144 for wild type, 528 for msh4, 600 for zip1, and 648 for msh4 zip1.

Zip1 staining with no DNA staining. In the strain back- Macdonald and Roeder 1994). The localization of ground used for this analysis, a polycomplex was ob- Msh4 relative to Zip1 was examined in cells (strain served in only a fraction of wild-type nuclei (Figure 3E). JN202) producing Msh4 protein tagged with the hemag- In the msh4 mutant, however, nuclei with polycomplexes glutinin epitope (Msh4-HA). This analysis was carried were far more frequent; late in meiotic prophase, nearly out in a BR2495 strain background, in which meiosis all nuclei with fully or partially synapsed chromosomes proceeds more slowly than in the SK1 strains used in contained a polycomplex (Figure 3E). the genetic experiments. Spread meiotic nuclei were The reduced number of pachytene nuclei observed prepared at 14.5 hr in meiosis; because meiosis is not in the msh4 mutant can be explained in two ways: not completely synchronous, nuclei at all stages of meiotic all msh4 cells go through pachytene, or each msh4 cell prophase are present at this time. Nuclei were classified spends an abnormally short time in pachytene. A muta- as to the extent of synapsis (as in Figure 3) and then tion that induces pachytene arrest, ndt80 (Xu et al. examined for Msh4 foci. None (0/20) of the spreads 1995), was used to distinguish these possibilities. As ex- without Zip1 staining showed any Msh4 foci. Of the pected, the ndt80 strain displayed pachytene arrest, with spreads with spotty Zip1 staining (indicative of the initia- 92% of cells in pachytene (Figure 3F). In contrast, the tion of synapsis), 72% (38/53) had Msh4 foci; the num- msh4 ndt80 strain arrested with only 24% pachytene ber of Msh4 foci appeared to be higher in nuclei in nuclei; 72% of the nuclei displayed partly linear Zip1 which Zip1 spots were more intensely staining. To- staining (Figure 3F). This result suggests that many msh4 gether, these observations suggest that Msh4 localizes cells fail to fully synapse their chromosomes, even when to chromosomes shortly after Zip1 assembly initiates. cells are held in midmeiotic prophase. 100% (50/50) of nuclei with partly continuous Zip1 Although previous results suggested the msh4 mutant staining (i.e., partially synapsed chromosomes) dis- is not delayed in meiotic nuclear division (Ross-Mac- played Msh4 foci. In addition, 100% (120/120) of nuclei donald and Roeder 1994), the defect in meiotic pro- with fully synapsed chromosomes had Msh4 foci on phase prompted a careful time course analysis of nu- chromosomes, demonstrating that Msh4 foci are pres- clear division in msh4 and wild-type strains. DAPI was ent throughout pachytene. used to visualize nuclei, and the number of mono-, bi-, Msh4 foci show extensive overlap with Zip3 foci and tetra-nucleate cells was monitored over time (Figure (Agarwal and Roeder 2000), predicting that Msh4 4). In the msh4 mutant, the production of both bi-nucle- also colocalizes with Zip2 (since Zip2 and Zip3 colocal- ate cells (products of the meiosis I division) and tetra- ize). The extent of overlap between Zip2 and Msh4 was nucleate cells (products of the meiosis II division) was assessed in cells producing Msh4-HA and Zip2 tagged hr. Thus, both meiotic divisions are de- with green fluorescent protein (Zip2-GFP). A typical 3ف delayed by layed in the msh4 mutant, as expected from the delay spread nucleus at pachytene is shown in Figure 5, A–C. in meiotic prophase progression. On average, 80% of Msh4 foci overlapped a Zip2 focus, Timing and pattern of Msh4 localization relative to and 66% of Zip2 foci overlapped a Msh4 focus. Random of Msh4 foci and %32ف Zip1 and Zip2: The Msh4 protein has been shown to positioning would cause only -of Zip2 foci to overlap (see materials and meth %26ف -foci on meiotic chromosomes (Ross 60ف localize to 1020 J. E. Novak, P. B. Ross-Macdonald and G. S. Roeder

Figure 3.—Chromosome synapsis and spindle formation in a msh4 mutant. msh4 (RM51) and wild-type (RM53) strains were introduced into sporula- tion medium at time 0; chro- mosomes were spread at the indicated times. Synapsis was assessed by staining with anti- Zip1 antibodies, and spindles were detected by simultaneous staining with anti-tubulin anti- bodies. At least 100 nuclei were scored for each strain at each time point. (A–D) Time course of synapsis and spindle forma- tion. Nuclei were divided in- to several categories to reflect the different stages in meiotic cell cycle progression: no Zip1 staining (not shown), (A) spotty Zip1 staining, (B) partly continuous Zip1 staining (in which a mixture of spots and linear stretches of Zip1 are seen, but Zip1 does not extend along the entirety of the chro- mosomes), (C) pachytene (in which Zip1 is fully continuous along all chromosomes, except in the nucleolar region), and (D) spindles (indicative of mei- otic chromosome segregation). The kinetics of appearance and disappearance of nuclei in the last four categories are shown. (E) Polycomplexes. For each class of Zip1 and tubulin stain- ing, the fraction of spreads con- taining a polycomplex was assessed. Results shown repre- sent pooled data from the 5.5- and 6.5-hr time points in the data set of Figure 3, A–D. (F) Chromosome synapsis during ndt80 arrest. Zip1 and tubulin staining was quantified in an ndt80 strain (JN305) at 9.5 hr and a msh4 ndt80 strain (JN304) at 11.5 hr. ods). Therefore, most, but not all, Msh4 and Zip2 foci compare cells at a similar stage of meiosis, only spreads have the same subnuclear location. Zip2 foci that lack with extensive Red1 staining on chromosomes were cho- Msh4 were more frequent than Msh4 foci lacking Zip2. sen for analysis. Red1 is a component of the cores of This difference could be explained if Zip2 localization meiotic chromosomes; Red1 localization is maximal at normally precedes Msh4 localization; in this case, some pachytene (Smith and Roeder 1997). of the Zip2 foci observed in spreads could be those to At 15 hr, Msh4 foci were detectable by immunofluo- which Msh4 has not yet localized. rescence in zip1 and zip2 nuclei, but they were consis- Proper localization of the Msh4 protein depends on tently much fainter than those in wild-type nuclei (Fig- Zip1 and Zip2: To understand better the relationship ure 5, G–I). Msh4 foci in the zip2 strain tended to be between Msh4 and the Zip proteins, Msh4 localization slightly fainter than those in zip1. In a typical experi- was examined in zip1 and zip2 strains, and Zip2 localiza- ment, the intensity of Msh4 staining (measured by total tion was assessed in a msh4 strain. Wild-type and mutant intensity in CCD camera images) averaged 200,000 for strains were tested at 15 hr, roughly the time of maxi- wild-type nuclei, 70,000 for the zip1 mutant, and 53,000 mum pachytene nuclei and Msh4 localization in the for the zip2 mutant. At 20 hr, each mutant displayed BR2495 strain background used for this analysis. In this slightly brighter Msh4 staining than at 15 hr, but still strain background, both zip1 and zip2 arrest in prophase less intense compared to the wild type at 15 hr (not of meiosis I (Sym et al. 1993; Chua and Roeder 1998); shown). Thus, Msh4 localization is partly dependent on therefore, the mutants were also tested at 20 hr. To both Zip1 and Zip2. In contrast, Zip2 foci appear normal Msh4 Functions in Synapsis and Interference 1021

be caused by a simple reduction in the amount of Msh4 localized to chromosomes or it may reflect a more pro- found disruption of the normal pattern of localization. To address this issue, a zip1 mutant was examined to determine whether Msh4 still colocalizes extensively with Zip2, which localizes normally in the absence of Zip1 (Chua and Roeder 1998). Figure 5, D–F, shows the localization of Msh4 and Zip2 in a typical zip1 mutant nucleus. On average, 54% of Msh4 foci overlapped a Zip2 focus, and 48% of Zip2 foci overlapped a Msh4 focus. These values are substantially less than the corre- sponding 80% and 66% overlap observed in the wild type, though still higher than the estimate for random overlap (27% of Msh4 and 24% of Zip2 foci). In a zip1 mutant, Zip2 protein localizes to axial associations Figure 4.—Time course of meiotic nuclear divisions. Wild- (Chua and Roeder 1998), which are connections be- type (RM53) and msh4 (RM51) strains were induced to un- tween aligned homologous chromosomes seen in zip1 dergo meiosis. At the indicated times, cells were fixed with mutants (Sym et al. 1993). Msh4 did not localize to the formaldehyde and subsequently stained with DAPI. Cells were scored as mono-, bi-, or tetra-nucleate; at least 200 cells were majority of axial associations visible in a zip1 mutant counted for each sample. (data not shown). Thus, in the absence of Zip1, Msh4 localization is abnormal with respect to Zip2. in pattern and intensity in the msh4 mutant (data not DISCUSSION shown), indicating that Zip2 does not depend on Msh4 for its localization to chromosomes. A msh4 mutation impairs chromosome synapsis: Care- The abnormal localization of Msh4 in zip1 cells may ful analysis of the timing and extent of SC formation

Figure 5.—Localization of Msh4 protein. (A–F) Colocalization of Msh4 and Zip2. HA-tagged Msh4 and GFP-tagged Zip2 were visual- ized by immunofluorescence. Both tagged proteins are fully func- tional, as assessed by homozygous tagged strains having wild-type spore viability (J. E. Novak, un- published data; Chua and Roeder 1997). Msh4 is shown in red (A, D); Zip2 in green (B, E). The im- ages are fused in C and F to show the extent of colocalization. A–C show a wild-type nucleus (JN295); D–F show a zip1 nucleus (JN294). (G–I) Localization of Msh4 in wild type and meiotic mutants. Images show Msh4 in red and DNA (visu- alized by DAPI) in blue. (G) wild type (JN202); (H) zip2 mutant (JN220); (I) zip1 mutant (JN232). In D, the brightness of Msh4 stain- ing has been artificially increased to permit accurate quantitation of colocalization with Zip2. Bars, 2 ␮m. 1022 J. E. Novak, P. B. Ross-Macdonald and G. S. Roeder in the msh4 mutant has revealed a defect in chromosome amount of SC formation, but this occurs in only a frac- synapsis. SC formation is delayed compared to wild type, tion of nuclei and often involves nonhomologous chro- and full synapsis is achieved in less than half of all nuclei. mosomes. These observations suggest that chromosome A polycomplex is formed in nearly all cells, including synapsis in mice is more dependent on the Msh4-Msh5 those in which chromosomes synapse fully, consistent complex than is synapsis in yeast. with a decrease in Zip1 localization to chromosomes. Msh4 functions in crossing over: Several observations The defect/delay in SC formation is accompanied by a indicate that Msh4-Msh5-Zip3-Zip2 foci correspond to delay in both spindle formation and nuclear division. the sites of meiotic recombination events. First, the for- It is likely that the msh4 delay in meiosis reflects the mation of these foci appears to require the initiation operation of checkpoint triggered by the defect in SC of meiotic recombination (Chua and Roeder 1998; formation or perhaps by an associated defect in the Agarwal and Roeder 2000; our unpublished observa- resolution of recombination intermediates (Bailis and tions). Second, in a mutant (rad50S) in which meiotic Roeder 2000). recombination is blocked at an early stage, the Zip2 Msh4 is found largely at the same sites on chromo- and Zip3 proteins colocalize with the Mre11 protein; somes as the Zip2 protein. Zip2 is required for Zip1 furthermore, Zip3 co-immunoprecipitates with Mre11 localization to chromosomes, and Zip2 foci correspond from meiotic cell extracts (Chua and Roeder 1998; to the sites at which Zip1 begins to polymerize on chro- Agarwal and Roeder 2000). Mre11 interacts with mosomes (Chua and Roeder 1998). The localization Rad50 and Xrs2 to form a complex that is involved in of Msh4 to sites of synapsis initiation can account for the formation and processing of meiotic double-strand the observed defect/delay in SC formation in the msh4 breaks (the initiators of meiotic recombination events; mutant. Although Msh4 seems to localize to chromo- Usui et al. 1998). Third, the formation of axial associa- somes slightly after Zip1, it may play a role in stabilizing tions (to which Zip2 localizes) requires two strand- the association of Zip1 with chromatin, or it may help exchange proteins, Dmc1 and Rad51 (Rockmill et al. organize chromatin-associated Zip1 into SC. Alterna- 1995). Finally, Zip3 interacts physically with Rad51 and tively, msh4 may have a defect in synapsis extension, Rad57, a cofactor to Rad51-mediated strand exchange rather than synapsis initiation. We favor this interpreta- (Agarwal and Roeder 2000). tion for two reasons. First, nuclei with partially synapsed Other observations indicate that Msh4-Msh5-Zip3-Zip2 chromosomes appear at the same time in the mutant foci correspond specifically to the sites of reciprocal as they do in wild type, but nuclei of this type persist recombination events (i.e., crossovers). Null mutations longer and accumulate to a higher level in the mutant in the genes encoding all four of the known compon- than they do in wild type (Figure 3B). Second, msh4 ents of these foci reduce crossing over two- to threefold nuclei often display chromosomes with linear stretches without affecting gene conversion. Furthermore, Msh4- of SC that fail to extend along the full length of the Msh5-Zip3-Zip2 foci, like crossovers, display interfer- chromosome. This contrasts with the situation observed ence; these complexes are nonrandomly positioned in the zip3 mutant (defective in synapsis initiation) in along chromosomes such that two foci rarely occur close which some chromosomes synapse fully while others fail together (Jennifer Fung and G. S. Roeder, unpub- to exhibit any Zip1 staining (Agarwal and Roeder lished data). 2000). Msh4 might affect Zip1 polymerization at a dis- The Msh4 protein may be directly involved in crossing tance by promoting conversion of Zip1 to a conforma- over. to MutS proteins suggests that Msh4 tion that more efficiently polymerizes along chromo- has DNA-binding activity. MutS proteins bind duplex somes. DNA, with especially high affinity for DNA containing Homologs of the MSH4 and MSH5 genes have been mismatched base pairs (Modrich 1987; Kolodner and identified in both humans and mice (Paquis-Fluck- Marsischky 1999). However, at least one of the yeast linger et al. 1997; Her and Doggett 1998; Winand et MutS proteins, Msh2, also binds strongly to Holliday al. 1998; Bocker et al. 1999; de Vries et al. 1999; Edel- junctions and promotes their resolution in vitro by puri- mann et al. 1999; Kneitz et al. 2000). The human MSH4 fied -cleaving enzymes (Alani et al. and MSH5 proteins, like the yeast proteins, have been 1997; Marsischky et al. 1999). It is tempting to specu- shown to form a complex (Winand et al. 1998; Bocker late that Msh4 (and/or its partner, Msh5) binds to Holli- et al. 1999). Similar to yeast Msh4, the mouse Msh4 day junctions and promotes their resolution in favor of protein localizes to discrete foci on meiotic chromo- crossing over. somes during the zygotene and pachytene stages of mei- The number of Msh4-Msh5-Zip3-Zip2 foci observed is significantly less than the average (55ف) otic prophase (Kneitz et al. 2000). Analysis of Msh4 and at pachytene .(90ف) Msh5 knockout mice has demonstrated an important number of crossovers that occur in a meiotic cell role for the proteins in chromosome synapsis (de Vries One explanation for this discrepancy is that the foci are et al. 1999; Edelmann et al. 1999; Kneitz et al. 2000). transient and asynchronous, such that not all foci are Knockout mice are sterile, and nuclei with fully synapsed detected at any given point in time. (Note, however, chromosomes are never observed. There is a limited that in ndt80 cells arrested at pachytene, the number Msh4 Functions in Synapsis and Interference 1023 of Zip2 foci is still significantly Ͻ90.) If Msh4 does nor- tion, a number of observations suggest that the interfer- mally act at all crossover sites, then loss of Msh4 might ence signal is transmitted along the SC. The finding that have the result that recombination intermediates that a msh4 mutation impairs both synapsis and interference would normally generate crossovers are randomly re- provides additional support for this hypothesis. The in- result in crossing over. complete and delayed SC formation observed in the %50ف solved such that only An alternative possibility is that only about half of all msh4 mutant could limit transmission of the interfer- crossovers require Msh4 (and associated proteins) and ence signal along chromosomes. The simplest version serve as sites of Msh4 localization; the remaining cross- of this model is that zippering up of the SC serves as the overs occur by a Msh4-independent mechanism (Ross- mechanistic basis for signal transmission (Egel 1978; Macdonald and Roeder 1994; Nakagawa et al. 1999; Maguire 1988). Synapsis initiates at sites designated to Zalevsky et al. 1999). In this case, in the absence of become crossovers, and SC zippering extends outward Msh4, all recombination intermediates that normally (probably in both directions) from each site. Subse- serve as Msh4 substrates would be resolved without cross- quent crossovers (or commitments to crossing over) ing over. are inhibited in nearby chromosomal regions that have If there is more than one pathway of crossing over, already synapsed. then these pathways might require different gene prod- The view that SC zippering serves as the mechanistic ucts. However, all observations to date suggest that all basis for transmission of the interference signal is a of the gene products that specifically promote meiotic specific version of a more general class of models in crossing over act in the same pathway. Epistasis tests which a crossover initiates a structural change that pre- have shown that Msh5 and Mlh1 (and presumably Mlh3) vents subsequent crossing over and is transmitted along act in the same pathway as Msh4 (Hollingsworth et al. the chromosome in a time-dependent manner. In the 1995; Hunter and Borts 1997). It has been proposed model proposed by King and Mortimer (1990), a cross- that a Msh4-Msh5 dimer interacts with the Mlh1-Mlh3 over (or commitment to crossing over) promotes poly- dimer to form a tetramer analogous to other MutS-MutL merization of a recombination inhibitor along the chro- complexes observed in eukaryotic cells (Kolodner and mosome. In the model of Kaback et al. (1999), crossing Marsischky 1999; Nakagawa et al. 1999). Exo1 and over triggers a conformational chain reaction in which probably Mer3 also act in the same pathway as Msh4 a chromosomal component undergoes an allosteric (Khazanehdari and Borts 2000; Kirkpatrick et al. change that both blocks recombination and causes a 2000; Tsubouchi and Ogawa 2000; our unpublished neighboring component to undergo the same change observations). Here, we have shown that Msh4 and Zip1 in conformation. Yet another possibility is that transmis- are in the same epistasis group; Zip3 and Zip2 presum- sion of the interference signal involves changes in chro- ably also act in the Msh4-dependent pathway, since these matin structure; Kleckner and colleagues have sug- proteins colocalize with Msh4 and Msh5. Thus, if there gested that this might occur by a crossover-induced is a second crossover pathway, gene products that act relief of physical stress (Kleckner 1996; Storlazzi et al. exclusively in this pathway have yet to be identified. 1996). By any of these models, if the proposed structural The role of the Zip1 protein in crossing over has been change in the chromosome must occur within the con- controversial (Sym et al. 1993; Sym and Roeder 1994; text of SC, then the msh4 defect in interference might Storlazzi et al. 1996; Chua and Roeder 1997; Roeder still be attributed to the defect in synapsis. Alternatively, 1997). Does this SC building block also participate di- Msh4 might be involved in initiating (rather than trans- rectly in recombination or does the defect in crossing mitting) the structural change that travels along the over in the zip1 mutant result indirectly from an alter- chromosome. Msh4 could play such a role regardless ation in meiotic chromosome structure? Our analysis of whether signal transmission involves SC zippering, of Msh4 localization provides a plausible explanation polymerization of a recombination inhibitor, conforma- for the role of Zip1 in crossing over. Since Zip1 is re- tional change in a chromosomal protein, or release of quired for localization of Msh4 in wild-type amounts tension. and to the correct locations, the zip1 defect in crossing Another possible explanation for the effect of msh4 over might be a consequence of a failure of Msh4 func- on interference assumes the operation of two distinct tion. pathways of crossing over. According to this model, Msh4 plays a role in crossover interference: Our data crossovers in one pathway require Msh4 and display indicate that the Msh4 protein plays an important role in interference, whereas crossovers in the other pathway crossover distribution. In the absence of Msh4, crossover are independent of Msh4 and do not exhibit interfer- interference is abolished in some intervals and reduced ence. The msh4 mutation would thus eliminate interfer- in others. Below, we consider several possible models ence by inactivating the first pathway. This model ac- for how Msh4 might mediate interference. Note that counts neatly for the observation that the HOM3–TRP2 these models are not mutually exclusive. interval shows no reduction in crossing over in msh4 One possibility is that msh4 affects interference by and no interference in wild type; presumably, all of the impairing SC formation. As described in the Introduc- crossovers in this interval are of the Msh4-independent 1024 J. E. Novak, P. B. Ross-Macdonald and G. S. Roeder variety. In this context, it is interesting to note that structures in meiotic prophase of fission yeast: a cytological analy- sis. J. Cell Biol. 121: 241–256. mutation of the homolog of MSH4 Bailis, J. M., and G. S. Roeder, 2000 The pachytene checkpoint. or MSH5 completely abolishes crossing over (Zalevsky Trends Genet. 16: 395–403. et al. 1999; Kelly et al. 2000), indicating that all cross- Bocker, T., A. Barusevicius, T. Snowden, D. Rasio, S. Guerrette et al., 1999 hMSH5: a human MutS homologue that forms a overs occur by a Msh4-dependent mechanism in this novel heterodimer with hMSH4 and is expressed during sperma- organism. Furthermore, interference in worms (and togenesis. Cancer Res. 59: 816–822. many other eukaryotes) is much stronger than it is in Chu, S., J. DeRisi, M. Eisen, J. Mulholland, D. Botstein et al., 1998 The transcriptional program of sporulation in budding yeast. yeast. This difference in the strength of interference Science 282: 699–703. across species can be economically explained by propos- Chua, P. R., and G. S. Roeder, 1997 Tam1, a telomere-associated ing that Msh4-dependent crossovers (which display in- meiotic protein, functions in chromosome synapsis and crossover interference. Genes Dev. 11: 1786–1800. terference) are diluted by Msh4-independent crossovers Chua, P. R., and G. S. Roeder, 1998 Zip2, a meiosis-specific protein (that lack interference) in budding yeast but not in required for the initiation of chromosome synapsis. Cell 93: 349– worms (or most eukaryotes). 359. Conrad, M. N., A. M. Dominguez and M. E. Dresser, 1997 Ndj1, A very different explanation for crossover interfer- a meiotic telomere protein required for normal chromosome ence is suggested by the counting model of Stahl and synapsis and segregation in yeast. Science 276: 1252–1255. colleagues (Foss et al. 1993; Foss and Stahl 1995). de Vries, S. S., E. B. Baart, M. Dekker, A. Siezen, D. G. de Rooij et al., 1999 Mouse MutS-like protein Msh5 is required for proper According to this model, adjacent crossovers must be chromosome synapsis in male and female meiosis. Genes Dev. separated by a specific number of noncrossover recom- 13: 523–531. bination events (two in the case of yeast; Foss et al. Dong, H., and G. S. Roeder, 2000 Organization of the yeast Zip1 protein within the central region of the synaptonemal complex. 1993). A test of this model indicated that data from J. Cell Biol. 148: 417–426. budding yeast cannot be adequately explained by such Edelmann, W., P. E. Cohen, B. Kneitz, N. Winand, M. Lia et al., a counting mechanism (Foss and Stahl 1995). How- 1999 Mammalian MutS homologue 5 is required for chromo- some pairing in meiosis. Nat. Genet. 21: 123–127. ever, if only a subset of crossovers exhibits interference Egel, R., 1978 Synaptonemal complex and crossing-over: structural (as proposed above), then the counting model could support or interference? Heredity 41: 233–237. still apply. Counting might be due to the clustering of Egel-Mitani, M., L. W. Olson and R. Egel, 1982 Meiosis in Aspergil- lus nidulans: another example for lacking synaptonemal com- multiple recombination intermediates (three in yeast) plexes in the absence of crossover interference. Hereditas 97: to produce a single late recombination nodule, with 179–187. only one intermediate in each such nodule being re- Engebrecht, J., and G. S. Roeder, 1990 MER1, a yeast gene re- quired for chromosome pairing and genetic recombination, is solved in favor of crossing over (Stahl 1993). In the induced in meiosis. Mol. Cell. Biol. 10: 2379–2389. context of the counting model, Msh4 might be required Fiorentini, P., K. N. Huang, D. X. Tishkoff, R. D. Kolodner and for clustering, to promote crossing over within a nodule, L. S. Symington, 1997 Exonuclease I of functions in mitotic recombination in vivo and in vitro. Mol. Cell. or to designate which of the intermediates within a Biol. 17: 2764–2773. cluster is chosen to be resolved in favor of crossing over. Foss, E., R. Lande, F. W. Stahl and C. M. Steinberg, 1993 Chiasma Summary: Msh4, a protein known to be involved in interference as a function of genetic distance. Genetics 133: 681– 691. meiotic crossing over, is also required for crossover in- Foss, E. J., and F. W. Stahl, 1995 A test of a counting model for terference. Furthermore, Msh4 localizes to sites of syn- chiasma interference. Genetics 139: 1201–1209. apsis initiation and promotes SC formation. Msh4 there- Havekes, F. W., J. H. de Jong, C. Heyting and M. S. Ramanna, 1994 Synapsis and chiasma formation in four meiotic mutants fore adds to the growing list of proteins that promote of tomato (Lycopersicon esculentum). Chromosome Res. 2: 315–325. both interference and synapsis, providing support for Her, C., and N. A. Doggett, 1998 Cloning, structural characteriza- the hypothesis that the SC is required for interference. tion, and chromosomal localization of the human orthologue of Saccharomyces cerevisiae MSH5 gene. Genomics 52: 50–61. We are grateful to Christopher Sarnowski for developing programs Hollingsworth, N. M., L. Ponte and C. Halsey, 1995 MSH5,a used for genetic analysis; to Jennifer Fung, Laurent Maloisel, Beth novel MutS homolog, facilitates meiotic reciprocal recombina- Rockmill, and Hideo Tsubouchi for comments on the manuscript; tion between homologs in Saccharomyces cerevisiae but not mis- and to Franklin Stahl, Penelope Chua, and Pedro San-Segundo for match repair. Genes Dev. 9: 1728–1739. Huang, K. N., and L. S. Symington, 1993 A 5Ј-3Ј exonuclease from comments on a previous version of the manuscript. This work was Saccharomyces cerevisiae is required for in vitro recombination be- supported by National Institutes of Health (NIH) grant GM28904 to tween linear DNA molecules with overlapping homology. Mol. G.S.R., by the Howard Hughes Medical Institute, and NIH National Cell. Biol. 13: 3125–3134. Research Service Award GM17654 to J.E.N. Hunter, N., and R. H. Borts, 1997 Mlh1 is unique among mismatch repair proteins in its ability to promote crossing over during meiosis. Genes Dev. 11: 1573–1582. Kaback, D. B., D. Barber, J. Mahon, J. Lamb and J. You, 1999 Chro- LITERATURE CITED mosome size-dependent control of meiotic reciprocal recombina- tion in Saccharomyces cerevisiae: the role of crossover interference. Agarwal, S., and G. S. Roeder, 2000 Zip3 provides a link between Genetics 152: 1475–1486. recombination enzymes and synaptonemal complex proteins. Kane, S. M., and R. Roth, 1974 Carbohydrate metabolism during Cell 102: 245–255. ascospore development in yeast. J. Bacteriol. 118: 8–14. Alani, E., S. Lee, M. F. Kane, J. Griffith and R. D. Kolodner, Kelly, K. O., A. F. Dernburg, G. M. Stanfield and A. M. Villeneuve, 1997 Saccharomyces cerevisiae MSH2, a mispaired base recognition 2000 Caenorhabditis elegans msh-5 is required for both normal and protein, also recognizes Holliday junctions in DNA. J. Mol. Biol. radiation-induced meiotic crossing over but not for completion of 265: 289–301. meiosis. Genetics 156: 617–630. Bahler, J., T. Wyler, J. Loidl and J. Kohli, 1993 Unusual nuclear Khazanehdari, K. A., and R. H. Borts, 2000 EXO1 and MSH4 Msh4 Functions in Synapsis and Interference 1025

differentially affect crossing-over and segregation. Chromosoma rescue: integrative DNA transformation in yeast. Methods Enzy- 109: 94–102. mol. 194: 281–301. Kilmartin, J. V., B. Wright and C. Milstein, 1982 Rat monoclonal Sherman, F., G. R. Fink and J. B. Hicks, 1986 Methods in Yeast antitubulin antibodies derived by using a new nonsecreting rat Genetics: A Laboratory Manual. Cold Spring Harbor Laboratory cell line. J. Cell Biol. 93: 576–582. Press, Cold Spring Harbor, NY. King, J. S., and R. K. Mortimer, 1990 A polymerization model of Sikorski, R., and P. Hieter, 1989 A system of shuttle vectors and chiasma interference and corresponding computer simulation. yeast host strains designed for efficient manipulation of DNA in Genetics 126: 1127–1138. Saccharomyces cerevisiae. Genetics 122: 19–27. Kirkpatrick, D. T., J. R. Ferguson, T. D. Petes and L. S. Symington, Smith, A. V., and G. S. Roeder, 1997 The yeast Red1 protein local- 2000 Decreased meiotic intergenic recombination and in- izes to the cores of meiotic chromosomes. J. Cell Biol. 136: 957– creased meiosis I nondisjunction in exo1 mutants of Saccharomyces 967. cerevisiae. Genetics 156: 1549–1557. Snow, R., 1979 Maximum likelihood estimation of linkage and inter- Kleckner, N., 1996 Meiosis: how could it work? Proc. Natl. Acad. ference from tetrad data. Genetics 92: 231–245. Sci. USA 93: 8167–8174. Stahl, F. W., 1993 Genetic recombination: thinking about it in phage and fungi, pp. 1–9 in The Chromosome, edited by J. S. Hes- Kneitz, B., P. E. Cohen, E. Avdievich, L. Zhu, M. F. Kane et al., lop-Harrison and R. B. Flavell. BIOS Scientific Publishers, 2000 MutS homolog 4 localization to meiotic chromosomes is London. required for chromosome pairing during meiosis in male and Storlazzi, A., L. Xu, A. Schwacha and N. Kleckner, 1996 Synap- female mice. Genes Dev. 14: 1085–1097. tonemal complex (SC) component Zip1 plays a role in meiotic Kolodner, R. D., and G. T. Marsischky, 1999 Eukaryotic DNA recombination independent of SC polymerization along the chro- mismatch repair. Curr. Opin. Genet. Dev. 9: 89–96. mosomes. Proc. Natl. Acad. Sci. USA 93: 9043–9048. Loidl, J., F. Klein and H. Scherthan, 1994 Homologous pairing Strickland, W. N., 1958 An analysis of interference in Aspergillus is reduced but not abolished in asynaptic mutants of yeast. J. Cell nidulans. Proc. R. Soc. Lond. Ser. B 149: 82–102. Biol. 125: 1191–1200. Sym, M., and G. S. Roeder, 1994 Crossover interference is abolished Maguire, M. P., 1988 Crossover site determination and interfer- in the absence of a synaptonemal complex protein. Cell 79: 283– ence. J. Theor. Biol. 134: 565–570. 292. Marsischky, G. T., S. Lee, J. Griffith and R. D. Kolodner, 1999 Sym, M., and G. S. Roeder, 1995 Zip1-induced changes in synaptone- Saccharomyces cerevisiae MSH2/6 complex interacts with Holliday mal complex structure and polycomplex assembly. J. Cell Biol. junctions and facilitates their cleavage by phage resolution en- 128: 455–466. zymes. J. Biol. Chem. 274: 7200–7206. Sym, M., J. Engebrecht and G. S. Roeder, 1993 ZIP1 is a synaptone- Modrich, P., 1987 DNA mismatch correction. Annu. Rev. Biochem. mal complex protein required for meiotic chromosome synapsis. 56: 435–466. Cell 72: 365–378. Moens, P. B., 1969 Genetic and cytological effects of three desynap- Tishkoff, D. X., A. L. Boerger, P. Bertrand, N. Filosi, G. M. Gaida tic genes in the tomato. Can. J. Genet. Cytol. 11: 857–869. et al., 1997 Identification and characterization of Saccharomyces Mortimer, R., and S. Fogel, 1974 Genetical interference and gene cerevisiae EXO1, a gene encoding an exonuclease that interacts conversion, pp. 263–275 in Mechanisms in Recombination, edited with MSH2. Proc. Natl. Acad. Sci. USA 94: 7487–7492. by R. Grell. Plenum Press, New York. Tsubouchi, H., and H. Ogawa, 2000 Exo1 roles for repair of DNA Nakagawa, T., and H. Ogawa, 1999 The Saccharomyces cerevisiae double-strand breaks and meiotic crossing over in Saccharomyces cerevisiae. Mol. Biol. Cell 11: 2221–2233. MER3 gene, encoding a novel helicase-like protein, is required Tung, K.-S., and G. S. Roeder, 1998 Meiotic chromosome morphol- for crossover control in meiosis. EMBO J. 18: 5714–5723. ogy and behavior in zip1 mutants of Saccharomyces cerevisiae. Genet- Nakagawa, T., A. Datta and R. D. Kolodner, 1999 Multiple func- ics 149: 817–832. tions of MutS- and MutL-related heterocomplexes. Proc. Natl. Tung, K.-S., E.-J. E. Hong and G. S. Roeder, 2000 The pachytene Acad. Sci. USA 96: 14186–14188. checkpoint prevents accumulation and phosphorylation of the Olson, L. W., and F. K. Zimmermann, 1978 Meiotic recombination meiosis-specific transcription factor Ndt80. Proc. Natl. Acad. Sci. and synaptonemal complexes in Saccharomyces cerevisiae. Mol. Gen. USA 97: 12187–12192. Genet. 166: 151–159. Usui, T., T. Ohta, H. Oshiumi, J.-I. Tomizawa, H. Ogawa et al., Papazian, H. P., 1952 The analysis of tetrad data. Genetics 37: 175– 1998 Complex formation and functional versatility of Mre11 of 188. budding yeast in recombination. Cell 95: 705–716. Paquis-Flucklinger, V., S. Santucci-Darmanin, R. Paul, A. Sau- Wach, A., A. Brachat, R. Pohlmann and P. Philippsen, 1994 New nieres, C. Turc-Carel et al., 1997 Cloning and expression anal- heterologous modules for classical or PCR-based gene disruptions ysis of a meiosis-specific MutS homolog: the human MSH4 gene. in Saccharomyces cerevisiae. Yeast 10: 1793–1808. Genomics 44: 188–194. Wang, T.-F., N. Kleckner and N. Hunter, 1999 Functional speci- Pochart, P., D. Woltering and N. Hollingsworth, 1997 Con- ficity of MutL homologs in yeast: evidence for three Mlh1-based served properties between functionally distinct MutS homologs heterocomplexes with distinct roles during meiosis in recombina- in yeast. J. Biol. Chem. 272: 30345–30349. tion and mismatch correction. Proc. Natl. Acad. Sci. USA 96: Rockmill, B., M. Sym, H. Scherthan and G. S. Roeder, 1995 Roles 13914–13919. for two RecA homologs in promoting meiotic chromosome synap- Winand, N. J., J. A. Panzer and R. D. Kolodner, 1998 Cloning and sis. Genes Dev. 9: 2684–2695. characterization of the human and Caenorhabditis elegans homo- Roeder, G. S., 1997 Meiotic chromosomes: it takes two to tango. logs of the Saccharomyces cerevisiae MSH5 gene. Genomics 53: Genes Dev. 11: 2600–2621. 69–80. Xu, L., M. Ajimura, R. Padmore, C. Klein and N. Kleckner, 1995 Ross-Macdonald, P., and G. S. Roeder, 1994 Mutation of a meiosis- NDT80, a meiosis-specific gene required for exit from pachytene specific MutS homolog decreases crossing over but not mismatch in Saccharomyces cerevisiae. Mol. Cell. Biol. 15: 6572–6581. correction. Cell 79: 1069–1080. Zalevsky, J., A. J. MacQueen, J. B. Duffy, K. J. Kemphues and A. M. Ross-Macdonald, P., A. Sheehan, G. S. Roeder and M. Snyder, Villeneuve, 1999 Crossing over during Caenorhabditis elegans 1997 A multipurpose transposon system for analyzing protein meiosis requires a conserved MutS-based pathway that is partially production, localization, and function in Saccharomyces cerevisiae. dispensable in budding yeast. Genetics 153: 1271–1283. Proc. Natl. Acad. Sci. USA 94: 190–195. Rothstein, R., 1991 Targeting, disruption, replacement and allele Communicating editor: M. Lichten