Patterns of Heteroduplex Formation Associated With the Initiation of Meiotic Recombination in the Yeast Saccharomyces cerevisiae

Corresponding author: Thomas D. Petes, University of North Carolina, Chapel Hill, NC 27599-3280., tompetes{at}email.unc.edu (E-mail)

Communicating editor: L. S. SYMINGTON

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The double-strand break repair (DSBR) model of recombination predicts that heteroduplexes will be formed in regions that flank the double-strand break (DSB) site and that the resulting intermediate is resolved to generate either crossovers or noncrossovers for flanking markers. Previous studies in Saccharomyces cerevisiae, however, failed to detect heteroduplexes on both sides of the DSB site. Recent physical studies suggest that some recombination events involve heterodupex formation by a mechanism, synthesis-dependent strand annealing (SDSA), that is inherently asymmetric with respect to the DSB site and that leads exclusively to noncrossovers of flanking markers. Below, we demonstrate that many of the recombination events initiated at the HIS4 recombination hotspot are consistent with a variant of the DSBR model in which the extent of heteroduplex on one side of the DSB site is much greater than that on the other. Events that include only one flanking marker in the heteroduplex (unidirectional events) are usually resolved as noncrossovers, whereas events that include both flanking markers (bidirectional events) are usually resolved as crossovers. The unidirectional events may represent SDSA, consistent with the conclusions of others, although other possibilities are not excluded. We also show that the level of recombination reflects the integration of events initiated at several different DSB sites, and we identify a subset of gene conversion events that may involve break-induced replication (BIR) or repair of a double-stranded DNA gap.

MEIOTIC recombination events in the yeast Saccharomyces cerevisiae are initiated by double-stranded DNA breaks (SZOSTAK et al. 1983 ; SUN et al. 1989 ) catalyzed by the protein Spo11 (BERGERAT et al. 1997 ; KEENEY et al. 1997 ). Until recently, in the most widely accepted version of the double-strand break repair (DSBR) model, the DNA ends resulting from the break are resected 5' to 3', and one of the resected "tails" invades the homologous chromatid, forming a region of heteroduplex (Fig 1). DNA synthesis primed from this 3' end results in a second heteroduplex, involving the second tail. The two regions of heteroduplex, one on each chromatid, are flanked by two Holliday junctions. Cleavage of these junctions can result in the flanking chromosomal regions in either crossover or noncrossover configurations. In this model, both crossovers and noncrossovers have the same molecular precursor.

View larger version (12K): In this window In a new window Download PPT slide   Figure 1. Canonical DSBR model (SZOSTAK et al. 1983 ; SUN et al. 1991 ). In this diagram, the DSB (indicated by the vertical arrow) occurs between two heterozygous markers on one of the chromatids with the mutant a and b alleles. The solid and open circles represent wild-type and mutant substitutions, respectively. The broken ends are processed 5' to 3', leaving protruding 3' ends (step 1). Step 2 shows the left end of the broken chromatid invading one of the unbroken chromatids (forming a heteroduplex with a mismatch). In step 3, DNA synthesis occurs (represented by a stippled red strand), resulting in displacement of the red strand. The displaced strand pairs with the broken right end, resulting in a second region of heteroduplex with a mismatch. DNA synthesis primed from the 3' end of the blue chromatid also occurs. Steps 4 and 5 represent alternative ways of cutting the strands connecting the two chromatids, resulting in chromatids without (step 4) or with (step 5) an associated crossover. Failure to correct the mismatches would result in a tetrad that had 5:3 segregation for both flanking markers. Correction of the mismatch would generate a 3:1 segregation event, if the correction were in the direction of the wild-type substitution, or would restore 2:2 segregation, if the correction was in the direction of the mutant substitution.

In Fig 1, we illustrate a recombination event initiated between two heterozygous genes with alleles A and a and B and b. If the heteroduplexes include the region of the gene with the mutation (as shown), then two DNA mismatches on two different chromatids would be formed. If these mismatches are not repaired, one of the four spores will have a postmeiotic segregation (PMS) event at the A locus, and a different spore in the same tetrad would have a PMS event at the B locus; PMS events are defined as the segregation of two alleles from a single meiotic product at the first mitotic division following meiosis (PETES et al. 1991 ). For some loci, these events can be detected by replica plating the spore colonies to diagnostic omission medium. Alternatively, one can detect such events using the polymerase chain reaction (PCR; PORTER et al. 1993 ). Repair of the mismatches will result in gene conversion tetrads, tetrads that have either three wild-type spores and one mutant spore or one wild-type spore and three mutant spores (PETES et al. 1991 ). For example, repair of the mismatch shown in Fig 1 could result in a tetrad with 3A:1a segregation pattern.

Although both physical and genetic evidence supports the existence of several of the intermediates shown in Fig 1 [for example, the double Holliday junction (SCHWACHA and KLECKNER 1995 )], there are also genetic and physical data that support the argument that all recombination events do not occur through the canonical form of the DSBR model. The genetic experiments used strains that were heterozygous for markers closely flanking a recombination hotspot (a site of frequent meiosis-specific DSB formation). These markers were designed to yield DNA mismatches that were inefficiently repaired if involved in heteroduplex formation, leading primarily to PMS tetrads. As discussed above, the DSBR model predicts a high frequency of tetrads in which one spore has a PMS event for one of the flanking markers and another spore in the same tetrad has a PMS event for the other marker. In two studies involving different loci, coevents of this type were very rare (PORTER et al. 1993 ; GILBERTSON and STAHL 1996 ), although events involving only one of the two markers were very common (unidirectional events).

ALLERS and LICHTEN 2001A , ALLERS and LICHTEN 2001B found physical data in conflict with the canonical DSBR model. In an ectopic recombination system, they showed that noncrossover heteroduplex products were completed before crossover heteroduplex products. Furthermore, they found that a mutation of NDT80, a meiosis-specific transcription factor, resulted in normal levels of noncrossover heteroduplexes, but greatly reduced crossovers. On the basis of these results, Allers and Lichten proposed that the heteroduplex intermediates for crossovers and noncrossovers were different: crossovers involving resolution of intermediates as shown in Fig 1 and noncrossovers occurring by a different pathway [synthesis-dependent strand annealing (SDSA)]. The SDSA pathway was originally suggested as a way of explaining mitotic gene conversion events that were unassociated with crossing over (reviewed by PAQUES and HABER 1999 ). In this model, following initial strand invasion and repair synthesis, the invading strand containing newly synthesized DNA is displaced and reannealed to the other 3' end. The net result of this event is heteroduplex formation on one side of the DSB in one chromatid with a noncrossover configuration of the flanking sequences.

In our previous study of heteroduplex formation (PORTER et al. 1993 ), we used heterozygous markers that were 1 kb from the HIS4 hotspot DSB site. Below, we describe experiments performed with a strain that has markers <250 bp from this DSB site and includes markers located 5 kb from the site. Using this new system, we identified bidirectional events with the configuration of heteroduplex predicted by the DSBR model. These events are primarily crossovers, but a significant fraction are noncrossovers. We also observed unidirectional events similar to those previously observed; most, but not all, of these events represent noncrossovers. We interpret these data as indicating that recombination in S. cerevisiae proceeds by both the canonical DSBR and the SDSA pathways.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Yeast strains:
All strains used were derived by transformation from the haploid strains AS4 ( trp1 arg4 tyr7 ade6 ura3) and AS13 (a leu2 ura3 ade6; STAPLETON and PETES 1991 ). In the descriptions of the genotypes, we note only deviations from the two parental genotypes.

We constructed JDM173 (fus1-BX) by two-step transplacement of AS4 using KpnI-digested pMW30 (WHITE and PETES 1994 ). All constructions, unless noted otherwise, were checked by PCR. JDM179 (bik1-lop his4u-lopc ycl034W-SX) was constructed in two steps. First, PD57 (DETLOFF et al. 1992 ), an AS13 derivative with a deletion of a portion of the HIS4 upstream activating sequences (UAS), was transformed with a DNA fragment containing the wild-type UAS of HIS4 and two palindromic insertions (bik1-lop and his4u-lopc) to generate JDM148 (bik1-lop his4u-lop). This DNA fragment was generated using PCR of genomic DNA derived from AS13 cells and the primers (with the palindromic sequences underlined) his4u + lopF1 (5'-GATTCCTCATCGGAAGAGGTGGCATCCTTAACGAAAAAACGAGTACTGTATGTACATACAGTACTCTTGAAGAGGCTAATGAAAAA) and his4u + lopcR1(5'-AGTTGTGCTATGATATTTTTATGTATGTACAACACACATCGACTAGTCTAAGTACTTAGACTAGTCGGAGGTGAATATAACGTTCC). Correct transplacement of the PCR fragment results in restoration of histidine prototrophy, and histidine prototrophs were sequenced to confirm the construction. Second, the ycl034W-SX mutation was inserted into JDM148 using a two-step transplacement of SnaBI-digested pJDM4 (described below) to generate JDM179.

The plasmid pJDM4 was derived from pMW25, a plasmid with a BglII fragment containing YCL034W from pC1G-17 (PORTER et al. 1993 ) inserted into the BamHI site of YIp5 (STRUHL et al. 1979 ). pJDM4 was made by "filling in" (SYMINGTON and PETES 1988 ) the SpeI site in YCL034W, resulting in the allele ycl034W-SX.

MD229 (bik1-lop his4u-lopc his4-IR9 ycl034W-SX) was constructed by inserting the his4-IR9 mutation into JDM179 using a two-step transplacement of SnaBI-digested pDN22 (NAG and PETES 1991 ). MD248 (fus1-BX cha1::hphMX4) and MD249 (bik1-lop his4u-lopc his4-IR9 ycl034W-SX cha1::hphMX4) were constructed by inserting the hphMX4 cassette (GOLDSTEIN and MCCUSKER 1999 ), which contains a gene that confers hygromycin resistance, into the middle of CHA1 in JDM173 and MD229, respectively. The PCR synthesis of the cassette and subsequent transformation were performed as described by WACH et al. 1994 , using the plasmid pAG32 (GOLDSTEIN and MCCUSKER 1999 ) and the primers CHA-F (5'-TCCCTTCGATAATCCGGATATTTGGGAAGGACATTCATCTATGATAGATGCGTACGCTGCAGGTCGAC) and CHA-R (5'-TTTAACCTTATTCACGGAAATATGTTGCGATTTCAAATCTTGTACTATTTATCGATGAATTCGAGCTCG).

PG118 (fus1-BX rad50S) and PG119 (bik1-lop his4u-lopc ycl034W-SX rad50S) are rad50S derivatives of JDM173 and JDM179, respectively, and were constructed as described by ALANI et al. 1990 . PG138 (his4-IR9 ycl034W-SX) was constructed by inserting the ycl034W-SX mutation into DNY47 (NAG and PETES 1991 ), using pJDM4 as described above.

We made the diploid strains by mating the following haploid strains (given in parentheses after the name of the diploid): JDM1080 (JDM173 x JDM179), JDM1081 (PG118 x PG119), JDM1086 (JDM173 x MD229), JDM1091 (JDM173 x PG138), MD250 (MD248 x MD229), MD251 (JDM173 x MD249), and QF105 (FAN et al. 1995 ). The genotypes of these strains (not including the AS4- and AS13-derived markers common to all of the strains) are: JDM1080 (FUS1/fus1-BX bik1-lop/BIK1 his4u-lopc/HIS4U ycl034W-SX/YCL034W), JDM1081 (FUS1/fus1-BX bik1-lop/BIK1 his4u-lopc/HIS4U ycl034W-SX/YCL034W rad50S/rad50S), JDM1086 (FUS1/fus1-BX bik1-lop/BIK1 his4u-lopc/HIS4U his4-IR9/HIS4 ycl034W-SX/YCL034W), JDM1091 (FUS1/fus1-BX his4-IR9/HIS4 ycl034W-SX/YCL034W), MD250 (FUS1/fus1-BX bik1-lop/BIK1 his4u-lopc/HIS4U his4-IR9/HIS4 ycl034W-SX/YCL034W CHA1/cha1::hphMX4), MD251 (FUS1/fus1-BX bik1-lop/BIK1 his4u-lopc/HIS4U his4-IR9/HIS4 ycl034W-SX/YCL034W cha1::hphMX4/CHA1), and QF105 (his4-IR9/HIS4 rad50S/rad50S). The designation HIS4U signifies that the HIS4 promoter region is wild type.

Genetic analysis:
Standard materials and methods were used (SHERMAN et al. 1983 ) except where noted. Diploids were sporulated on plates at 25°, rather than 18° (the temperature used for most of our previous studies), to reduce the frequency of multiple recombination events at the HIS4 hotspot (NAG et al. 1989 ). Following tetrad dissection on plates with rich growth medium (YPD), the spore colonies were replica plated to various omission media; spore colonies on medium lacking histidine were examined microscopically to detect small sectors.

Experiments to determine whether the transcribed or nontranscribed strand of HIS4 was transferred were performed as described by NAG and PETES 1990 with the following modifications. Tetrads were dissected onto plates containing histidine omission medium (SD - his) and incubated at 30° for 10–12 hr. The spores were then examined microscopically and scored as His⁺ (5 or more cells) or His^- (1 or 2 cells). The SD - his agar slab containing the spores was transferred to a YPD plate with 0.4 ml of a 0.5% histidine solution. After incubation at 30° for 3–4 days, the colonies were replica plated to SD - his medium (as well as other omission media) to determine the pattern of sectoring. Only tetrads with a 5:3 or 3:5 pattern of aberrant segregation were scored for the remaining markers by PCR. The pattern of sectoring was then correlated with the phenotype of the spore to determine whether the transcribed or nontranscribed strand of HIS4 had been donated (NAG and PETES 1990 ). This analysis allows us to determine whether the DSB in a given recombination event occurred upstream (nontranscribed strand donated) or downstream (transcribed strand donated) of our markers in HIS4.

PCR analysis of spore colonies:
We performed PCR analysis to score the fus1-BX, bik1-lop, his4u-lopc, and ycl034W-SX markers. All PCRs were performed in 96-well trays using the GeneAmp PCR System 9700 (Applied Biosystems, Foster City, CA). PCR conditions were those suggested by the manufacturer of AmpliTaq polymerase (Applied Biosystems) with the following modifications. The reaction conditions contained 1.5 mM MgCl₂ for the PCR to score bik1-lop and 2.5 mM MgCl₂ for all other PCRs. All four dNTPs were added to a final concentration of 400 µM for the PCR to score fus1-BX and ycl034W-SX and 200 µM for the other PCRs. All PCRs contained 0.8 µM of each primer. The reactions used to score fus1-BX and ycl034W-SX contained 2.5 units of AmpliTaq, and all other reactions contained 1.75 units of AmpliTaq.

A toothpick was used to mix the cells of the spore colony on a YPD replica of the dissection plate. The cells were then transferred to 6 µl of sterile, distilled water in the 96-well tray. Each tray included three control reactions where the cells were wild type, mutant, or sectored for the marker of interest. The samples were heated to 94° for 6 min and subsequently placed at -80° for 10 min. The samples were thawed at room temperature, and the remaining 19 µl of the reaction components was added. Samples were exposed to the following conditions for 40 cycles: 94° for 1 min, 57° for 1 min, and 72° for 3 min.

The primers BIK1 + 1021F (5'-ACGATTCGCTCAGTAAAGAATAC) and BIK1 + 1228R (5'-GCCGTGGTATCGACTGGTGC) produce a 208-bp product if the wild-type sequence is present and a 234-bp product if the bik1-lop sequence is present. The PCR products were digested with BsrGI (New England Biolabs, Beverly, MA), which cuts within the bik1-lop sequence. Subsequently, the digested PCRs were resolved on a 3.5% MetaPhor agarose (BioWhittaker Molecular Applications, Walkersville, MD) gel. Three patterns of bands were observed: a 208-bp fragment if only the wild-type BIK1 sequence was present in the spore colony, a pair of 115- and 119-bp fragments if only the bik1-lop sequence was present in the spore colony, and a set of 115-, 119-, 208-, and apparently 240-bp fragments if both wild-type and bik1-lop sequences were present in the spore colony (a PMS event). The fragment with the apparent size of 240 bp is likely an in vitro-generated heteroduplex fragment containing one wild-type and one bik1-lop strand.

The primers HIS4-210F (5'-CCCATGCACAGTGACTCACG) and HIS4 + 42R (5'-ATGAGGCCAGATCATCAATTAACGG) produce a 253-bp product if the wild-type sequence is present and a 279-bp product if the his4u-lopc sequence is present. The PCR products were digested with ScaI (New England Biolabs), which cuts within his4u-lopc. Upon gel analysis, three patterns (similar to those observed for bik1-lop) were seen: a 253-bp fragment if only the wild-type HIS4 sequence was present in the spore colony, a pair of 134- and 145-bp fragments if only the his4u-lopc sequence was present in the spore colony, and a set of 134-, 145-, 253-, and apparently 300-bp fragments if both wild-type and his4u-lopc sequences were present in the spore colony. Using this method to score bik1-lop or his4u-lopc, we always (10 of 10 times) detected mutant or wild-type sequences, even when they represented <10% of the DNA sample.

For most of the spores that had PMS for more than one marker, we determined whether the configuration of the markers was in cis (palindromes on the same DNA strand in the heteroduplex) or trans (palindromes on different DNA strands). Cells from the relevant spore colony were streaked onto YPD. When single colonies had formed, two independent colonies were examined by PCR or replica plating (as described above) for the relevant markers.

We scored both flanking markers, fus1-BX and ycl034W-SX, using a single PCR. The PCRs were digested simultaneously with SpeI and BclI (New England Biolabs) and resolved on 1.5% agarose gels. The primers SpeI-508F (5'-ACGCTAGAAGTGGAGTTAGC) and SpeI + 276R (5'-AACGCAGCCACCAGTTCATC) produce a fragment of 800 bp. A fragment containing wild-type sequence (YCL034W) produces fragments of 300 and 500 bp when digested with SpeI, while a fragment with the SpeI "fill in" (ycl034W-SX) remains 800 bp. The primers FUS1 + 517(I)F (5'-CCGCAGCATATACTGACACC) and FUS1 + 1514(I)R (5'-AGTCACCAGGCACAATGCCT) produce a fragment of 1 kb. A fragment containing wild-type sequence (FUS1) produces fragments of 400 and 600 bp when digested with BclI, while a fragment with the fus1-BX mutation remains 1 kb. The fus1-BX and ycl034W-SX markers generally did not exhibit sectoring, which is typical for 4-bp insertions (NAG et al. 1989 ).

Southern analysis:
Cells were harvested from rad50S diploid strains just prior to being placed on a sporulation plate (0 hr) or after 24 hr on sporulation medium. Cells were washed with 0.5 ml 10 mM Tris (pH 8.0), 1 mM EDTA, and stored at -80°. DNA isolation and Southern blot procedures were performed as described by NAG and PETES 1993 .

This procedure was used to map DSBs occurring in a 15-kb interval centered on HIS4. Probes were prepared from genomic DNA by PCR; 20-bp primers were derived from the sequence intervals described below. The HIS4-BIK1 region [Saccharomyces Genome Database (http://genome-www.stanford.edu/Saccharomyces/) chromosome III coordinates 66,644–69,621] was examined using a BglII digest and a BglII-XhoI fragment of HIS4 as a hybridization probe (NAG and PETES 1993 ). The YCL036W-YCL031C region (60,396–65,125) was examined with an SphI digest and a hybridization probe covering the coordinates 60,410–61,152. The YCL031C-HIS4 region (64,542–68,093) was examined with an NheI digest and a hybridization probe covering coordinates 66,175–67,276. The BIK1-FUS1 region (69,019–73,467) was examined using a BanI digest and a hybridization probe covering coordinates 69,073–69,817. The FUS1-YCL025C region (72,688–77,018) was examined using an NciI digest and a hybridization probe covering coordinates 72,768–73,341. Hybridization was quantitated using a PhosphorImager (Molecular Dynamics, Sunnyvale, CA), and the percentage of molecules with a DSB was calculated as described (KIRKPATRICK et al. 1999 ). All levels of DSBs were normalized to the DSB associated with the HIS4 hotspot.

Data analysis:
Statistical analysis was performed using Instat 1.12 (GraphPad Software) for the Macintosh. The Fisher's exact test with a two-tailed P value or chi-square analysis (for comparisons that involve more than two experimental measures) was used for all comparisons, and P < 0.05 was considered statistically significant.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Experimental system:
We designed related diploid strains, JDM1086 and JDM1080 (Fig 2), to examine the arrangement of heteroduplex DNA close to the HIS4 DSB site and the association of heteroduplex DNA with a crossover or noncrossover configuration of flanking sequences. Both strains were heterozygous for short palindromic insertions (bik1-lop and his4u-lopc) closely flanking the DSB site, and JDM1086 was also heterozygous for his4-IR9, a short palindromic insertion within the HIS4 coding sequence. The strains were sporulated and tetrads derived from the strains were dissected. A mismatch resulting from a heteroduplex with one wild-type DNA strand and one strand with a short palindromic insertion is inefficiently repaired to generate a gene conversion, resulting in frequent PMS events (NAG et al. 1989 ). For the bik1-lop and his4u-lopc markers, such events can be detected by PCR (details in MATERIALS AND METHODS); a PMS event involving his4-IR9 can be detected by replica plating the spore colony to medium lacking histidine. The segregation of the flanking heterozygous restriction site markers fus1-BX and ycl034W-SX was also analyzed by PCR.

View larger version (19K): In this window In a new window Download PPT slide   Figure 2. Arrangement of genetic markers in strains JDM1080 and JDM1086. The position of the DSB associated with the HIS4 recombination hotspot is indicated by the arrow. The "lollipops" indicate short palindromic insertions, markers that result in inefficiently repaired mismatches. JDM1080 lacks the his4-IR9 marker; otherwise, the strains are identical.

A summary of the segregation of markers in the two strains is in Table 1. Since (as described in the Introduction) heteroduplex formation can result in spores that have genes in which one DNA strand has wild-type and one strand has mutant information, it is convenient to describe the patterns of aberrant segregation using the nomenclature derived from eight-spored fungi. We classify gene conversion tetrads with three wild-type and one mutant or one wild-type and three mutant spore colonies as 6:2 and 2:6, respectively. Tetrads with two wild-type, one mutant, and one wild-type/mutant PMS spore colonies and those with one wild-type, two mutant, and one wild-type/mutant PMS spore colonies were classified as 5:3 and 3:5, respectively. Aberrant 4:4 tetrads have one wild-type, one mutant, and two wild-type/mutant PMS spore colonies. Several different methods of analysis were done in the JDM1086 strain. For the data designated "unselected," we examined the segregation of all markers in every spore colony in tetrads with four viable spores. The other methods of analysis are discussed further below.

Unselected tetrads of JDM1086 and JDM1080 had similar frequencies of aberrant segregation for the same markers. The frequencies of aberrant segregation of the fus1-BX, his4-IR9, and ycl034W-SX markers in JDM1086 were similar to those observed in strain JDM1091. JDM1091 is identical to JDM1086 except that it lacks the bik1-lop and his4u-lopc markers. The frequencies of crossovers between HIS4 and the linked LEU2 gene in the two strains were also very similar (Table 1). If a mutation generates a recombination hotspot in a strain heterozygous for the mutation, one finds an excess of tetrads of the 6:2 and 5:3 classes over the 2:6 and 3:5 classes (PETES et al. 1991 ). No such disparity is observed for the palindromic markers in JDM1086 or JDM1080 (Table 1), indicating that these palindromes neither stimulate nor repress recombination. We confirmed this result by mapping DSBs associated with the HIS4 hotspot in the strains JDM1081 (a rad50S derivative of JDM1080) and QF105 (a rad50S strain without the his4u-lopc and bik1-lop markers). The positions and levels of DSBs in the two strains were the same (data not shown).

In addition to examining unselected tetrads in JDM-1080 and JDM1086, we used several other methods of analysis for JDM1086. For tetrad data classified as "selected-1" (S1), we screened for tetrads in which the his4-IR9 marker (which can be scored by replica plating to medium lacking histidine) segregated 6:2, 2:6, 5:3, or 3:5 (indicative of a single recombination event involving the marker), and we subsequently examined the other markers (fus1-BX, bik1-lop, his4u-lopc, and ycl034W-SX) by PCR. For tetrad data classified "selected-2" (S2), we screened for tetrads that segregated 5:3 or 3:5 at the his4-IR9 locus and did not have a cosector for the bik1-lopc marker in the same spore. If the pattern was consistent with a single event initiated at the HIS4 hotspot, we examined all of the other markers. As described below, this procedure selects against tetrads in which the recombination event is initiated at a DSB that is different from the HIS4 hotspot DSB. The "strand transfer" method of analysis is described below.

Classification of recombination events:
As in previous experiments involving multiple markers located near a recombination hotspot (for example, PORTER et al. 1993 ), we find many classes of tetrads. All the data are on the website (http://www.genetics.org/supplemental/). The data are divided into five supplementary tables: Table I (single unidirectional recombination events initiated at the HIS4 hotspot), Table II (single bidirectional recombination events initiated at the HIS4 hotspot), Table III (single recombination events initiated at a site other than the HIS4 hotspot), Table IV (unambiguous multiple recombination events), and Table V (events in which the classification as single or multiple events is model dependent).

The tetrads that are most useful in exploring the nature of DSB-mediated recombination are those with a single initiating DNA lesion occurring at the HIS4 hotspot (between the bik1-lop and his4u-lopc markers). Determination of where the events were initiated and whether the recombination events were initiated by one or multiple DSBs was based on several assumptions. First, all recombination events are initiated by a DSB, occurring at the HIS4 hotspot or one of the other hotspots mapped (as described below) within an 11-kb region that includes the HIS4 hotspot. Second, recombination events involve the continuous asymmetric transfer of a single strand from one chromosome to another, resulting in the formation of heteroduplex on one side of the DSB in one chromatid, but (potentially) heteroduplex formation on the other side of the DSB in a different chromatid (Fig 1). Third, the initiating DSB occurs at one end of a heteroduplex tract. If an event is unidirectional and both ends of the heteroduplex tract correspond to meiotic DSB sites, the initiating DSB is assumed to occur at the stronger DSB site. If an event is bidirectional (involves two regions of heteroduplex on different chromatids), the DSB is assumed to occur between the two regions of heteroduplex. Fourth, Holliday junctions are resolved at the ends of the heteroduplex tracts. This last assumption will lead to some degree of underestimation of the frequency of associated crossovers, since a mismatch repair event that leads to restoration of Mendelian segregation (reviewed in PETES et al. 1991 ) would separate the detectable heteroduplex tract from the position of the crossover. We note, however, that the three palindromic insertions occur near the HIS4 DSB site, where restoration repair occurs infrequently (DETLOFF et al. 1992 ).

Of 1603 tetrads derived from strains JDM1086 and JDM1080, there were 217 tetrads with an aberrant segregation event for one or more of the five heterozygous markers in the HIS4 region; 56 are explicable as resulting from a single DSB located at the HIS4 hotspot and 77 are explicable as resulting from a single DSB located at a site other than the HIS4 hotspot. In addition, 50 tetrads had undergone multiple initiation events, and 34 can be classified as either single- or multiple-event tetrads, depending on details of the models of recombination.

Single recombination events initiated at the HIS4 DSB:
The tetrads that we classify as single recombination events initiated at the HIS4 DSB share several properties: (1) tracts of aberrant segregation are uninterrupted by markers undergoing normal Mendelian segregation, (2) markers on each side of the DSB site have aberrant segregation properties indicating involvement of a single donated DNA strand, and (3) markers on opposite sides of the DSB site involve different chromatids. In 116 unselected tetrads of JDM1086 and JDM1080, 22 (19%) had these properties. Of all tetrads examined for these strains, we classified 56 as representing single events initiated at the HIS4 hotspot.

A total of 70% of these tetrads (40 of 56) were classified as unidirectional events (Table I, classes 1–25). Unidirectional events exhibit continuous tracts of aberrant segregation on one side of the DSB (toward either HIS4 or BIK1) that are confined to a single chromatid. About 30% (10 of 34 tetrads in which the configuration of the flanking sequences could be unambiguously assigned) of these events are associated with crossovers. Examples of segregation patterns consistent with unidirectional events without and with associated crossovers are shown in Fig 3, a and b, respectively.

View larger version (28K): In this window In a new window Download PPT slide   Figure 3. Examples of segregation patterns representing different types of recombination. We illustrate segregation patterns for five markers heterozygous in JDM1086. As in Tables I–V at http://www.genetics.org/supplemental/, each row of circles represents a spore colony. For the markers bik1-lop, his4u-lopc, his4-IR9, and ycl034W-SX, solid circles represent colonies with the wild-type genotype and open circles represent the mutant genotype. For fus1-BX, solid and open circles represent mutant and wild-type genotypes, respectively. Thus, a tetrad with two rows of open circles and two rows of solid circles would indicate a nonrecombinant tetrad. Solid/open sectored circles represent PMS events. The class numbers indicated below are those used in Tables I–V, available at http://www.genetics.org/supplemental/. (a) Tetrad with unidirectional event initiated at the HIS4 hotspot unassociated with crossover (Table I, class 12). (b) Tetrad with unidirectional event initiated at the HIS4 hotspot associated with crossover between bik1-lop and his4u-lopc (Table I, class 13). (c) Tetrad with unidirectional event initiated at the HIS4 hotspot associated with crossover between bik1-lop and his4u-lopc and an incidental crossover involving two other chromatids (Table IV, class 111). (d) Tetrad with bidirectional event initiated at the HIS4 hotspot unassociated with crossover (Table II, class 30). (e) Tetrad with bidirectional event initiated at the HIS4 hotspot associated with crossover either between fus1-BX and bik1-lop or between his4-IR9 and ycl034W-SX (Table II, class 31). (f) Tetrad with single recombination event initiated at a DSB other than the HIS4 hotspot (Table III, class 64).

This number of undirectional events with associated crossovers is larger than can be explained by incidental crossovers. On the basis of data from the 116 unselected tetrads, 10% (12) of the tetrads contained an incidental crossover; incidental crossovers are defined as those in which the crossover involved is not located at the end of a tract of gene conversion/PMS. The likelihood of an incidental crossover involving the chromatid containing the aberrant segregation events is, therefore, 5% (0.10 x 0.50). Since the incidental crossover must also occur adjacent to the tract of aberrant segregation to be considered an associated crossover, the likelihood would be reduced to <5%. Given 34 unidirectional events in which the crossover could be unambiguously mapped, <2 would be expected to be scored as containing a crossover configuration of the flanking markers due to incidental crossovers.

Of the eight crossover events that could be mapped to a single interval, five involved the 380-bp interval II, one involved the 4.5-kb interval I, and two involved the 6-kb interval III. On the basis of the relative sizes of these intervals, these results suggest a strong preference for resolution of the unidirectional events as crossovers at a position near the initiating DSB. In a study similar in design to ours, L. JESSOP and M. LICHTEN (personal communication) found that 85% of the aberrant segregation events were unidirectional, and there was an even stronger bias in favor of crossover resolution near the initiating DSB.

In addition to the 40 tetrads that had a single recombination event consistent with a unidirectional heteroduplex initiated at the HIS4 hotspot, there were an additional 6 tetrads consistent with a unidirectional heteroduplex initiated at the HIS4 hotspot plus an incidental exchange (Table IV, classes 87, 88, 110, and 111 and Table V, classes 148 and 149). An example of such a tetrad is shown in Fig 3C. These tetrads are included in Table 2, which summarizes the number of uni- and bidirectional tetrads obtained in JDM1080 and JDM1086 with different methods of analysis. Patterns of the unidirectional events are shown in Fig 4A.

View larger version (25K): In this window In a new window Download PPT slide   Figure 4. Numbers of tetrads with various patterns of aberrant segregation interpretable as unidirectional (a) and bidirectional (b) events initiated at the DSB associated with the HIS4 hotspot. In this diagram, solid circles indicate a gene conversion event, and solid/open sectored circles show a PMS event. The unidirectional events involve a single chromatid (circles connected by horizontal lines show aberrant segregation events in the same direction), whereas the bidirectional events involve different chromatids on each side of the DSB (indicated by a short vertical line connecting horizontal lines). In a, the tetrad classes (derived from Tables I–V) used for each line of data (line 1 indicating the top line) are as follows: line 1, classes 1–4; line 2, classes 5 and 87; line 3, classes 6–8 and 110; line 4, classes 9–11; line 5, classes 12–16, 111, 148, and 149; line 6, classes 17, 18, and 88; line 7, classes 19–21; line 8, classes 22 and 23; line 9, class 24; and line 10, class 25. In b, the comparable information is as follows: line 1, class 26; line 2, classes 27–31, 89, and 112–114; line 3, classes 32 and 33; line 4, classes 34 and 115; line 5, classes 35 and 90; line 6, class 116; and line 7, class 36.

Events were classified as bidirectional if they involved two tracts of aberrant segregation, one continuous tract involving markers on one side of the DSB in one spore and a second continuous tract involving markers on the other side of the DSB in a different spore (Table II, classes 26–36). To classify a tetrad as a bidirectional event, we required that at least one of the markers on each side of the DSB site had undergone a PMS event. This requirement was imposed because coconversion events involving bik1-lop and his4u-lopc could represent a recombination event initiated at a DSB site other than that located between bik1-lop and his4u-lopc. This issue is discussed further in a separate section of RESULTS.

The bidirectional events are predicted by the DSB model in Fig 1, but have been observed only very rarely in previous studies (PORTER et al. 1993 ; GILBERTSON and STAHL 1996 ). About 30% (16 of 56) of the single recombination events initiated at the HIS4 DSB are bidirectional and 64% (9 of 14 tetrads in which the configuration of the flanking sequences could be unambiguously assigned) of the bidirectional events are associated with crossovers. Examples of single bidirectional events without and with associated crossovers are shown in Fig 3D and Fig E, respectively. In 8 of the 9 tetrads (Table II, classes 26, 28, and 31–33), Holliday junction resolution involved cleavage of the strands that the DSBR model (Fig 1) predicts would contain newly synthesized DNA. Similar biases have been observed previously and are explicable as targeted cleavage of the Holliday junctions directed by the nicked strand (reviewed by FOSS et al. 1999 ).

In addition to those tetrads shown in Table II, we found seven additional tetrads in which a bidirectional event occurred in a tetrad with an incidental exchange (Table IV, classes 89, 90, and 112–116). These tetrads are included in Table 2 and Fig 4. In tetrads in which crossovers could be unambiguously mapped, 13 of 39 unidirectional events were associated with a crossover, and 13 of 19 bidirectional events were crossover associated. These levels of association are significantly different (P = 0.02, Fisher's exact test). If tetrads in which all of the aberrantly segregating markers representing conversion rather than PMS events are excluded from the analysis of the unidirectional events, then 11 of 32 unidirectional events are crossover associated, compared to 13 of 19 bidirectional events (P = 0.02, Fisher's exact test).

Since the number of bidirectional events is relatively small, one issue to consider is whether such events could reflect two independent unidirectional events. In 116 unselected tetrads, if we include the bidirectional events as representing two unidirectional events, there were 14 (12%) tetrads with aberrant segregation patterns of bik1-lop consistent with a unidirectional event initiated at the HIS4 hotspot and 14 tetrads with aberrant segregation patterns of his4u-lopc consistent with a unidirectional event initiated at the HIS4 hotspot. The predicted fraction of tetrads in which two such unidirectional events would mimic a bidirectional event is (1/2)(1/2)(0.12)(0.12) or 0.0036. The two factors of one-half reflect the probabilities that both events will be in the same direction (both 5:3/6:2 or 3:5/2:6) and that the events will involve different chromatids. The observed frequency of bidirectional events (23/1603) is at least fourfold higher than this value.

A similar conclusion can be made on the basis of a somewhat different type of argument. In 712 tetrads of JDM1086 examined by the selected-1 method, we found 10 in which the bik1-lop and his4u-lopc markers both segregated 5:3 or 3:5 in different spores as expected for bidirectional DSBR events. Only 1 tetrad had a 5:3 segregation at his4u-lopc and a 3:5 segregation at bik1-lop in a different spore, and none had a 3:5 at his4u-lopc and 5:3 at bik1-lop. Thus, the patterns expected for the bidirectional DSBR events are more common than those expected for two unidirectional events.

On the basis of the orientation of the direction of transcription of HIS4 and the 5' to 3' resection of the broken ends, recombination events initiated by a DSB at the HIS4 hotspot will involve a heteroduplex in which the nontranscribed strand of HIS4 is the donor and the transcribed strand (derived from the chromosome that received the DSB) is the recipient (NAG and PETES 1990 ). This prediction can be tested using a specialized type of tetrad analysis in which the tetrads are dissected onto plates containing medium lacking histidine (details in MATERIALS AND METHODS). After 11 hr, the spores are scored as His^- or His⁺. The medium containing the dissected spores is then transferred to plates containing excess histidine, and the histidine diffuses into the medium lacking histidine, allowing nonselective growth of the spores. When spore colonies have formed, they are replica-plated to medium lacking histidine to score His⁺/His^- sectored colonies. By correlating the aberrant segregation pattern with the scoring of the spore phenotypes on the histidine omission medium, we can determine for tetrads with a single PMS event involving his4-IR9 which strand was transferred (NAG and PETES 1990 , Figure 2). For example, if the sectored colony from a 5⁺:3^- segregation event was derived from a His^- spore, we can infer that the donor wild-type strand was nontranscribed. It should be pointed out that any recombination event initiated centromere-proximal to his4-IR9 that includes this marker in a heteroduplex event will result in donation of the nontranscribed strand (according to the canonical DSBR model), whereas events initiated by a DSB centromere-distal to the marker will result in donation of the transcribed strand.

Strand transfer analysis of a limited number of tetrads confirmed our interpretation of most of the single events initiated between the palindromes. Of six events, three unidirectional and three bidirectional, initially assigned as being initiated by the HIS4 DSB, all of the bidirectional events and two of the three unidirectional events involved transfer of the nontranscribed strand. The one unidirectional event resulting from the transfer of the transcribed strand involved postmeiotic segregation of his4-IR9 with coconversion of his4u-lopc and ycl034W-SX. An alternative interpretation of this event is that it initiated centromere-distal to the ycl034W-SX marker and terminated between bik1-lop and his4u-lopc.

Single recombination events that include markers in the HIS4 region that are initiated at a site different from the HIS4 DSB:
From previous studies analyzing meiosis-specific DSBs on chromosome III (BAUDAT and NICOLAS 1997 ; GERTON et al. 2000 ), it is clear that there are a number of strong DSB sites near HIS4 other than the one located immediately upstream of HIS4. We mapped DSBs in the region around HIS4 in strains JDM1081 and QF105 (details in MATERIALS AND METHODS). The patterns of DSBs were the same in the two strains, indicating that the bik1-lop and his4u-lopc markers do not affect DSBs. In Fig 5, the sizes of the arrows indicate the approximate strength of the DSBs. Many of the tetrads examined in our study (77 of 1603) have patterns of aberrant segregation consistent with single initiation events at one of the DSBs that is not associated with the HIS4 hotspot (Table III, classes 37–86; Fig 3F). We define this type of tetrad as (1) those that have aberrant segregation for ycl034W-SX, fus1-BX, or his4-IR9, but not for either bik1-lop or his4u-lopc and (2) those with coevents that span the bik1-lop and his4u-lopc markers. One particularly interesting type of tetrad (Fig 5; Table III, classes 50–52) represents coconversion events that include all markers in the 10.5-kb region that includes HIS4; this type of tetrad is discussed in a separate section. Fig 5 also includes an additional 7 tetrads that were interpretable as events initiated at a site different from the HIS4 hotspot and that had an incidental exchange (Table IV, classes 91–93 and 117; Table V, classes 150–152).

View larger version (25K): In this window In a new window Download PPT slide   Figure 5. Summary of aberrant segregation events interpretable as initiating at sites other than the HIS4 hotspot-associated DSB in JDM1080 and JDM1086 tetrads. Arrows show the position and intensity of DSBs in this region. The HIS4 hotspot-associated DSB (occurring near the 3' end of BIK1) represents 5% of the DNA molecules. Normalizing this DSB to a value of 1, the other DSBs had approximate intensities of 0.5 (FUS1-associated DSB), 0.5 (DSB near YCL034W on HIS4 side), and 0.25 (doublet on other side of YCL034W). The horizontal lines show the extent of coevents (continuous tracts of conversion and/or PMS). Solid circles indicate conversion and sectored circles indicate PMS. In all cases, the events were in the same direction. The tetrad class marked with an asterisk shows a pattern of aberrant segregation that would be consistent with a unidirectional event initiated at the HIS4 hotspot, except for the observation that it involved transfer of the transcribed strand of HIS4.

Support for these classifications was provided by strand transfer experiments. Of 11 events classified as initiating from a DSB other than the HIS4 hotspot, 4 were events in which the transcribed strand of HIS4 was donated. As discussed above, such events are likely to reflect DSBs that are centromere-distal to the HIS4 hotspot DSB. Most of the other events are likely to represent events initiated from centromere-proximal DSBs. Two tetrads, however, yielded an unexpected pattern. In these two tetrads, the his4-IR9 marker, but not the his4u-lopc marker, showed aberrant segregation. We expected these two tetrads to reflect a DSB located centromere-distal to the HIS4 hotspot and, therefore, to involve donation of the transcribed strand. Both, however, involved transfer of the nontranscribed strand. The segregation patterns of these tetrads could be generated by a DSB at the HIS4 hotspot, heteroduplex formation that includes both his4u-lopc and his4-IR9, and with restoration repair of the his4u-lopc mismatch. Alternatively, these patterns could reflect recombination initiated by a DSB between his4u-lopc and his4-IR9.

We also found one tetrad (class 43, Table III) that was a coevent involving the his4u-lopc, his4-IR9, and ycl034W-SX markers, similar to one class of unidirectional events shown in Table I. This tetrad was not considered a unidirectional event initiated at the HIS4 hotspot, however, because a strand transfer experiment indicated that the donated strand was the transcribed strand of HIS4.

In summary, we conclude that 56% of the recombination events that involve his4-IR9 initiate at DSBs at sites different from the HIS4 hotspot. Our data demonstrate that the recombination activity at a specific site in the genome represents the integration of recombination activities initiated at multiple DSB sites. This conclusion, although somewhat surprising, is consistent with our previous observations that mutational changes (for example, elimination of the Rap1p binding site in the HIS4 promoter) that block DSB formation at HIS4 reduce aberrant segregation of his4-IR9 by only twofold (FAN et al. 1995 ).

At many loci, the frequency of gene conversion of a mutant site is a linear function of its position within the gene (reviewed by NICOLAS and PETES 1994 ). Such gradients of gene conversion are termed "polarity gradients." At both the ARG4 and HIS4 loci, the frequency of conversion declines from the 5' end to the 3' end, although the extent of this decline is quite different. At the ARG4 locus, the rates of conversion differ by about a factor of 10, whereas at the HIS4 locus, the difference is only a factor of 2.5 (NICOLAS and PETES 1994 ). Several types of mechanisms have been suggested to contribute to the formation of polarity gradients: (1) the extent of resection of DNA from the DSB site (SUN et al. 1991 ), (2) a distance-dependent alteration in the ratio of conversion-type to restoration-type repair (DETLOFF et al. 1992 ) perhaps directed by the resolution of Holliday junctions (FOSS et al. 1999 ), and (3) distance-dependent abortion of heteroduplexes directed by the mismatch repair system (ALANI et al. 1994 ). Our results argue that the shape of the polarity gradient may also reflect differential contributions of heteroduplexes initiated at multiple DSB sites.

Meiotic recombination events that may reflect break-induced replication or gap repair:
In 9 of 1603 tetrads (Fig 5), all five markers underwent conversion, either all 6:2 or all 2:6 (Table III, classes 50–52). These events are unusual in two ways. First, the conversion tracts were unusually long, minimally 10.5 kb. Second, the palindromic markers, which usually exhibited postmeiotic segregation, instead underwent conversion. In 9 of 10 tetrads in our study in which the flanking fus1-BX and ycl034W-SX markers coconverted, the intervening palindromic insertions also coconverted. In 20 of 22 tetrads in which the palindromic insertions, but not the flanking fus1-BX and ycl034W-SX markers, underwent coaberrant segregation (Table III, classes 64–78; Table IV, classes 92 and 93), one or more of the palindromic insertions had a PMS event. This difference is very significant (P = 0.0001).

One interpretation of this result is that such tetrads reflect a very long heteroduplex that covers all five markers. Excision tracts extending from the mismatches involving the fus1-BX and ycl034W-SX markers could result in the corepair of the mismatches resulting from the palindromic insertions. This interpretation is unlikely, however, since two-thirds of meiotic excision repair tracts are <1 kb, and none >1.8 kb were detected (DETLOFF and PETES 1992 ). In addition, of 22 tetrads (Table III, classes 41–49, 53–60, and 91) that include either fus1-BX or ycl034W-SX (but not both markers) and one or more of the palindromic sites in a coaberrant segregation event, 16 had PMS events at one or more of the palindromic insertions.

We favor the alternative possibility that the class 50–52 tetrads are gene conversion events that do not involve heteroduplex formation followed by DNA mismatch repair. We suggest two possibilities. The first is that these conversion events reflect meiotic break-induced replication (BIR) events. In BIR events, which have been invoked as a model to explain very long mitotic gene conversion tracts (reviewed by PAQUES and HABER 1999 ), a broken end of one chromosome invades a second, setting up a unidirectional replication process that proceeds to the end of the chromosome (Fig 6A). A second possibility is that the class 50–52 tetrads are a consequence of a chromatid with two closely spaced DSBs. If the broken ends derived from different DSBs are used to set up the double Holliday junctions with loss of the DNA fragment between the breaks, a gene conversion event that does not involve DNA mismatch repair would occur (Fig 6B). This model is essentially that proposed originally for the DSBR model (SZOSTAK et al. 1983 ) except that the gap is a consequence of two DSBs rather than a single DSB that is processed by degradation of both DNA strands.

View larger version (31K): In this window In a new window Download PPT slide   Figure 6. Mechanisms leading to coconversion without mismatch repair within a heteroduplex. (a) BIR. In this mechanism (reviewed by PAQUES and HABER 1999 ), one broken end invades a chromosome and replication proceeds to the end of the intact DNA molecule. Subsequently, the resulting junction is cleaved. (b) Gap repair. One chromatid receives two DSBs, with loss of the DNA fragment located between the two DSB sites. The resulting gap is filled in by repair synthesis, and the junctions are cleaved. Although we show the cleavage pattern that would generate a noncrossover configuration of flanking markers, the intermediate could also be resolved to generate a crossover.

The HIS4 gene is 67 kb from the left telomere of chromosome III. We constructed two strains that were isogenic with JDM1086 except for the inclusion of a heterozygous insertion of the HYG^R gene into CHA1, a gene located 16 kb from the telomere. In strains MD250 and MD251, the insertions were on the opposite homolog or the same homolog, respectively, as the palindromic insertions. In all tetrads derived from these diploids, we scored segregation of the his4-IR9 and the cha1::hphMX4 markers. In those tetrads in which the his4-IR9 marker underwent gene conversion, we examined the segregation of the palindromic insertions and the flanking markers. The data from this experiment are shown in Table 1.

Of 1429 total tetrads, we found 6 in which all five markers in the HIS4 region were coconverted (frequency of 0.4%). The cha1::hphMX4 marker underwent gene conversion in 3 of these tetrads (all derived from MD251) and segregated 2:2 in 3. Although this number of these tetrads is low, since the rate of aberrant segregation of the cha1::hphMX4 marker is only 2.1% (Table 1), it is statistically significant (P = 0.0002); in all such tetrads, four other heterozygous markers segregated 2:2, indicating that these exceptional tetrads are not likely to be false. Of these 3 tetrads, however, only 1 had the pattern shown in Fig 6A and Fig 7A. In this tetrad, three spores were fus1-BX BIK1 HIS4U HIS4 YCL034W CHA1 and one was FUS1 bik1-lop his4u-lopc his4-IR9 ycl034W-SX cha1::hphMX4, as expected for a single BIR event. In a second tetrad, one spore was fus1-BX BIK1 HIS4U HIS4 YCL034W cha1::hphMX4, two were FUS1 bik1-lop his4u-lopc his4-IR9 ycl034W-SX cha1::hphMX4, and one was FUS1 bik1-lop his4u-lopc his4-IR9 ycl034W-SX CHA1. The pattern of segregation observed in this tetrad is consistent with a single BIR event, followed by a crossover between the YCL034W and CHA1 genes (Fig 7B). In the third tetrad, we found one spore was fus1-BX BIK1 HIS4U HIS4 YCL034W CHA1, two were FUS1 bik1-lop his4u-lopc his4-IR9 ycl034W-SX CHA1, and one was FUS1 bik1-lop his4u-lopc his4-IR9 ycl034W-SX cha1::hphMX4. This pattern of segregation can be explained by a crossover between the YCL034W and CHA1 genes that preceded a BIR event (Fig 7C). It should be pointed out that all 3 of the diagnostic tetrads can also be explained as gap repair events in which one DSB occurs centromere-proximal to the markers in the HIS4 region and the other occurs centromere-distal to the cha1::hphMX4 marker. Because of the limited distance between the cha1::hphMX4 marker and the telomere, we prefer the hypothesis that they represent BIR events.

View larger version (23K): In this window In a new window Download PPT slide   Figure 7. Expected patterns of aberrant segregation after BIR alone and BIR with a crossover. Open and solid circles indicate the five heterozygous markers in the HIS4 region and the open and solid rectangles indicate the presence or absence of the cha1::hphMX4 allele. The cha1::hphMX4 marker is 50 kb from the markers in the HIS4 region. The arrow shows the position of the initiating DSB, and we assume that the centromere-distal fragment is lost. (a) Pattern expected for BIR without a crossover between the HIS4 markers and the cha1::hphMX4 marker. (b) Pattern expected for a BIR event and single crossover between the HIS4 markers and cha1::hphMX4 after the completion of BIR. (c) Pattern expected for a crossover between the HIS4 markers and cha1::hphMX4 before BIR.

In three of the six tetrads with coconversion of markers in the HIS4 region, the cha1::hphMX4 marker segregated 2:2. We suggest that these events represent the repair of a double-strand gap, as discussed above (Fig 6B). We cannot exclude the possibility that these events reflect very long heteroduplexes in which the resulting DNA mismatches are repaired by a process involving very long excision repair tracts, although no experimental evidence supports such a mechanism.

Tetrads with unambiguous multiple recombination initiation events:
Although the classification of tetrads as representing single or multiple events depends, to some extent, on what assumptions are allowed concerning heteroduplex formation (symmetric or asymmetric) and the patterns of DNA mismatch repair, tetrads with more than two recombinant chromatids (for example, Fig 3C) or that have markers on the same chromatid that segregate in opposite directions (for example, 5:3 for bik1-lop and 3:5 for his4u-lopc) must represent multiple initiation events (Table IV). We also include in this table tetrads that have two recombinant chromatids and two chromatids with one parental configuration of markers, but no chromatids with the other parental configuration. A total of 50 tetrads involving these classes of multiple events were observed (Table IV). In 23 tetrads, three chromatids were recombinant (classes 87–109); in 13, four were recombinant (classes 110–122); 5 tetrads involved either three or four chromatids (classes 123–127). We expect an 2:1 ratio of three-chromatid to four-chromatid events for double recombination events, since there are two ways of involving three chromatids, but only one way of involving four.

In addition to the 41 tetrads that have involvement of more than two chromatids in 1 tetrad, there were 9 tetrads with two recombinant chromatids, but markers that segregated in opposite directions. These tetrads represent classes 128–136 in Table IV. It should be pointed out that when tetrads could be classified as a multiple event by more than one criterion, we assigned them into one of the classes arbitrarily.

Tetrads that represent either single or multiple recombination initiation events, depending on the assumptions about the mechanism of recombination:
Thirty-four tetrads could be classified as representing either single or multiple recombination events (Table V). All tetrads in this category had no more than two recombinant chromatids and, if more than one marker underwent aberrant segregation, the markers segregated in the same direction. The different types of tetrads in this group included: (1) tetrads in which the continuity of a conversion/PMS tract was disrupted by a marker that undergoes Mendelian segregation (classes 137–147), (2) tetrads in which the crossover was separated from the aberrant segregation tract by at least one other marker that undergoes Mendelian segregation (classes 148–152), (3) tetrads that had spores with two PMS events in which the palindromes were in different DNA strands (trans events; classes 153–160), (4) tetrads with more than one PMS event for a single marker (classes 161–164), and (5) tetrads with a crossover between two markers showing aberrant segregation in the same direction (classes 165–167). Although all of these tetrads can be explained as representing multiple initiation events, all are also consistent with events initiated by a single DSB, as described below.

Noncontiguous tracts of aberrant segregation (for example, one marker segregating 2:2 with flanking markers segregating 5:3) can be explained as two DSBs giving rise to two heteroduplex regions or as a single heteroduplex in which the middle marker undergoes restoration-type repair; this type of repair of mismatches in heteroduplexes results in Mendelian segregation instead of gene conversion (conversion-type repair) or PMS (failure to repair). Although restoration-type repair events near the site of the DSB are infrequent (DETLOFF et al. 1992 ), such events occur at mismatches that are displaced from the initiating lesion (KIRKPATRICK et al. 1998 ). Restoration-type repair can also explain tetrads in which the crossover is separated from the tract of gene conversion/PMS by a marker exhibiting Mendelian segregation.

We found many tetrads in which more than one marker underwent PMS in the same direction (both 5:3 or both 3:5). For most such tetrads, we determined whether the event involved transfer of the same DNA strand (as described in MATERIALS AND METHODS); this analysis was done on all of the unselected tetrads and tetrads examined by the strand transfer method of analysis and more than half of the tetrads examined by the S-1 method. Although most of these co-PMS events involved transfer of the same DNA strand (cis), we found 12 tetrads in which the palindromic insertions were in different strands (trans); 2 of these tetrads were classified as representing double events for other reasons (classes 94 and 95, Table IV), whereas 10 were classified as double events solely as a consequence of the trans configuration (classes 153–160, Table V). Such trans events were found previously (PORTER et al. 1993 ; GILBERTSON and STAHL 1996 ). The trans events could represent a double SDSA event. In such double events, we suggest that one end invades and is used as a primer for DNA synthesis. The invading strand is then displaced. The second end then invades, is used as a primer for DNA synthesis, and is removed. Alternatively, these events could represent single events initiated at the HIS4 hotspot in which only one Holliday junction is cut and the other junction branch migrates (Fig 5 in GILBERTSON and STAHL 1996 ). Four of the 10 tetrads with the trans configuration had an associated crossover, indicating that whatever intermediates are involved in producing the trans configuration, they must have the option of resolution as a crossover.

In some tetrads, one or more markers exhibited aberrant 4:4 segregation (one wild-type spore colony, one mutant spore colony, and two sectored spore colonies). This pattern of segregation can be a consequence of formation of symmetric heteroduplexes, in which a single initiating event generates heteroduplexes at the same site on two different chromatids (HOLLIDAY 1964 ), or can reflect two independent initiation events, each involving asymmetric formation of heteroduplexes. Since the frequency of aberrant 4:4 tetrads in S. cerevisiae is roughly that expected for two independent events, it has been argued that symmetric heteroduplex formation is infrequent (FOGEL et al. 1981 ; PETES et al. 1991 ). This conclusion was supported by investigations of the frequency of aberrant 4:4 tetrads in strains with different levels of hotspot activity at the HIS4 locus (FAN et al. 1995 ). GILBERTSON and STAHL 1996 and STAHL and HILLERS 2000 , however, pointed out that the mechanism by which a recombination intermediate was resolved would influence the ability to detect symmetric heteroduplexes. For example, a symmetric heteroduplex intermediate that was resolved by topoisomerase, rather than cleavage of Holliday junctions, would not be detectable as an aberrant 4:4 segregation. In addition, HILLERS and STAHL 1999 found that some classes of aberrant 4:4 tetrads had patterns of associated crossovers consistent with symmetric heteroduplex formation, whereas others did not. In our view, the most likely interpretation of the existing data is that most observable aberrant 4:4 segregation events represent two initiation events associated with asymmetric heteroduplexes, although some reflect symmetric heteroduplexes. It should be pointed out that, of the 17 tetrads with markers that had double PMS or double gene conversion events, 13 were tetrads in which there were more than two recombinant chromosomes, strongly suggesting that they represent double initiation events.

We also classified tetrads in which a crossover occurred within a tract of aberrant segregation as representing ambiguous multiple events. Such events could also be explained as a single recombination intermediate in which branch migration moved the Holliday junction into the tract of aberrant segregation (following DNA mismatch repair). On the basis of the low frequency of aberrant 4:4 tetrads, which argues against extensive branch migration (FOGEL et al. 1981 ), we expect this class to be infrequent.

Although the tetrads depicted in Table V could represent either single or double events, it is likely that the majority of these tetrads represent single initiation events. This conclusion is based on the fraction of these tetrads in which there is involvement of two, three, or four chromatids. Assuming that there is no positive or negative chromatid interference for the initiation of recombination events, one would expect that double events would involve two, three, or four chromatids in an approximate ratio of 1:2:1. Thus, three-quarters of the double events would be expected to be three- or four-chromatid events. If we consider all 84 tetrads in Tables IV and V, we find that 41 represent three- and four-chromatid events, and 43 are two-chromatid events. The simplest way of explaining the excess of two-chromatid events is that many of the ambiguous "multiple" events in Table V represent single initiations.

Because of the ambiguities involved in the interpretation of these tetrads and others that may represent multiple initiation events, our discussion of models of recombination emphasizes those tetrads that can be easily explained as resulting from a single initiating DNA lesion.

Crossovers unassociated with aberrant segregation:
Although most of our analysis was done with tetrads that were screened for aberrant segregation of his4-IR9, we also nonselectively examined 116 tetrads. Seven of these tetrads had crossovers without aberrant segregation of any of the five markers in the HIS4 region. In tetrads derived from JDM1086, we found 2, 1, and 1 tetrad with crossovers in regions I, II, and IIIb, respectively. We also found 3 tetrads in strain JDM1080 with a crossover in the interval IIIa/IIIb.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Our results, as well as those of others, indicate the difficulty and, perhaps, the futility of explaining all meiotic recombination activities on the basis of a single model. At the HIS4 recombination hotspot, we suggest that there are at least three types of recombination events, all initiated by DSBs: (1) events that occur through the canonical DSBR pathway, (2) SDSA events, and (3) BIR and/or gap repair events. Each of these classes is discussed separately below.

Canonical DSBR pathway of recombination:
Some events represent formation and resolution of double Holliday junctions as predicted by a slightly modified form of the canonical DSBR model shown in Fig 1. Since very few of these events were observed in our previous study in which the flanking markers were 700–1000 bp from the DSB site (PORTER et al. 1993 ), we suggest that the initial strand invasion produces a region of heteroduplex that is variable, but often <200 bp. Since preliminary studies indicate that the resection of the DSB at the HIS4 hotspot is usually >400 bases, we argue that the size of the heteroduplex is not determined solely by the extent of resection. The heteroduplex resulting from primed synthesis of the invading strand results in a second, and much more extensive, region of heteroduplex (Fig 8). We previously proposed that the unidirectional events resulted from asymmetric resection of the DNA ends produced by the DSB (PORTER et al. 1993 ). Although we cannot exclude this model, on the basis of preliminary observations of symmetrically resected ends at the HIS4 hotspot (data not shown), we suggest that the extent of heteroduplex is not directly related to the extent of resection.

View larger version (21K): In this window In a new window Download PPT slide   Figure 8. Two mechanisms that generate the unidirectional recombination events observed at the HIS4 hotspot. We suggest that the unidirectional events have two sources. Both types of events involve the same early steps: DSB formation, followed by DSB processing (step 1), and single-ended invasion, followed by primed DNA synthesis (step 2). On the left part of the diagram, the second broken end forms a heteroduplex (step 3), and the resulting intermediate is processed to yield either a noncrossover (step 5) or a crossover (step 6). On the right part of the diagram, the single-ended invasion is reversed without crossing over (step 4).

The observed asymmetry in heteroduplexes flanking the DSB site can also be explained by other versions of the DSBR model. For example, it is possible that an extensive heteroduplex region is formed by the strand invasion, and the limited heteroduplex region results from limited DNA synthesis primed by the invading strand. Although we cannot rule out this model, we favor the first model for two reasons: (1) physical data argue that HIS4 DSBs are resected by 600 bp (NAG and PETES 1993 ; our unpublished data), an amount too limited to account for the very long (2.7 kb) heteroduplexes observed at HIS4 (DETLOFF et al. 1992 ), and (2) large (>1 kb) heterozygous insertions within the HIS4 gene are readily incorporated into heteroduplexes, an observation more compatible with DNA replication-driven heteroduplex formation than with passive DNA strand invasion or DNA branch migration.

ALLERS and LICHTEN 2001B physically demonstrated the existence of DNA molecules with heteroduplexes flanked by double Holliday junctions ("JM1" intermediates) as predicted by the DSBR model. In addition, they observed a recombination intermediate in which the double Holliday junctions were located on one side of the DSB, and the marker at the DSB site was not in a heteroduplex ("JM2" intermediates). They explained this intermediate by a model [the strand-displacement model; Fig 4C of ALLERS and LICHTEN 2001B ] in which strand invasion and DNA synthesis occur on only one side of the DSB. The invading strand is partially displaced and pairs with the other end of the broken chromosome. The net result of these events is that the double Holliday junction is located to one side of the DSB and no heteroduplex is observed on that side of the DSB. There is, however, a region of heteroduplex on the opposite side of the DSB. Although we found a few tetrads with the segregation pattern expected for this type of event (classes 148 and 149), such tetrads were rare. Many of the crossovers