<<

Genetics: Published Articles Ahead of Print, published on August 6, 2010 as 10.1534/genetics.110.121194

A Two-pathway Analysis of Meiotic Crossing Over and in

Saccharomyces cerevisiae

Franklin W. Stahl and Henriette M. Foss

Institute of , University of Oregon, Eugene, Oregon 97403-1229

Dedicated to the Institute of Molecular Biology, established January 1, 1959.

1

Running title: Crossing over and conversion

Key words: Mlh1, Msh2, Msh4, double-strand-break repair, mismatch repair

Corresponding author: Franklin W. Stahl, Institute of Molecular Biology, 1229 University of Oregon, Eugene, Oregon 97403

Phone: (541) 346-6096

Fax: (541) 346-5891

E-mail: [email protected]

2

ABSTRACT

Several apparently paradoxical observations regarding meiotic crossing over and gene conversion are readily resolved in a framework that recognizes the existence of two recombination pathways that differ in mismatch repair, structures of intermediates, crossover interference, and in the generation of noncrossovers. One manifestation of these differences is that simultaneous gene conversion on both sides of a recombination- initiating DNA double-strand break ("two-sidedness") characterizes only one of the two pathways and is promoted by mismatch repair. Data from previous work are analyzed quantitatively within this framework, and a molecular model for meiotic double-strand- break repair based on the concept of sliding D-loops is offered as an efficient scheme for visualizing the salient results from studies of crossing over and gene conversion, the molecular structures of recombination intermediates, and the biochemical competencies of the proteins involved.

3

Eukaryotes transit from the diplophase to the haplophase via , which is associated with a number of interrelated processes, including crossing over and gene conversion. These processes involve meiosis-specific, programmed DNA double-strand breaks (DSBs) and their repair (DSBr). DSBr, in turn, is associated with mismatched base-pairs and their rectification, referred to as "mismatch repair" or MMR (BISHOP et al.

1987). Current efforts to accommodate both the genetic and molecular phenomena associated with meiotic DSBr in yeast () have been thoroughly reviewed (e.g., HOLLINGSWORTH and BRILL 2004; HOFFMANN and BORTS 2004; SURTEES et al. 2004; HUNTER 2007; BERCHOWITZ and COPENHAVER 2010), but none of the reviews commits to an overall picture with quantitative predictions. This work aims to remedy that lack. Specifically, we have made use of salient published studies to develop, step by step, a comprehensive model of meiotic DSBr and MMR. The main features of this model are summarized in Table 1.

For readers who are unfamiliar with yeast genetics and/or the known details of

MMR, we will begin by reviewing (1) the basic principles and vocabulary of tetrad analysis in yeast, which exposes the products of individual acts of meiosis, (2) the DSBR model by SZOSTAK et al. (1983) as modified by SUN et al. (1991), which has provided a basic molecular interpretation of meiotic recombination, and (3) the known roles of mismatch-repair proteins such as Msh2 and Mlh1.

Relative frequencies of tetrad types provide measures of linkage distance and crossover interference: Consider a population of diploid yeast cells heterozygous for

4

two linked sites, A/a and D/d. When meiosis proceeds without a hitch, the resulting tetrads each contain four viable haploid spores. Because the genotypes of the spores are identifiable by the phenotypes of the colonies they give rise to, each spore in the tetrad can be characterized as a crossover or a noncrossover with respect to sites

A/a and D/d. When the A/a and D/d sites are closely linked, the most frequent tetrads contain only the two genotypes that characterized the parents, i.e., they have two AD and two ad spores, and are therefore referred to as "parental ditype" tetrads

(PD). Two other types of tetrads may be found in various frequencies: "tetratype" tetrads (T), i.e., those in which the spores are all different (AD, Ad, aD, ad), and

"nonparental ditype" tetrads (NPD: Ad, Ad, aD, aD) representing the two recombinant genotypes. This type of tetrad analysis -- assessing the relative frequencies of PDs, Ts and NPDs -- allows a quantitative measure of crossing over (i.e., linkage distances and crossover interference).

Gene conversion as evidence of DSBr: The same diploid may be marked at one or two other sites, B/b (and/or C/c), closely bracketing a "DSB hotspot", i.e., a site that receives a high frequency (e.g., 20%) of programmed, meiosis-induced double-strand breaks. The B/b (and C/c) sites serve to identify tetrads that have undergone a DSBr event at the hotspot (GILBERTSON and STAHL 1996). Such tetrads are recognized by their failure to exhibit Mendelian (i.e., 2:2) segregation of the markers among the spores.

Instead, a "conversion tetrad" may contain three spores with a marker derived from one parent, and one spore with the allele from the other parent (e.g., 3 B:1 b, or 3 b:1 B). This non-2:2 distribution of genotypes is also termed "aberrant" or "non-

5

Mendelian". Furthermore, because the , a double-stranded DNA molecule, carries genetic information on each of its complementary strands, a haploid spore may give rise to a "mixed colony", consisting of both B and b (or C and c) cells. Tetrads containing such a spore will be referred to as "half-conversions" (HCs) or "5:3". They can be diagrammed as BB, BB, Bb, bb when B is in excess over b, or as bb, bb, bB,

BB when b is in excess. In this paper, we shall deal with the two most common types of conversions: the HCs or 5:3s as described above, and the "full conversions" (FCs) or 3:1s

(aka 6:2s), which may be diagrammed as BB, BB, BB, bb or bb, bb, bb, BB). In general, tetrads with an excess of B or of b are found at similar frequencies. A significant deviation from this expectation is referred to as "disparity".

Successful application of tetrad analysis requires careful placement of the markers ABCD/abcd. Sites B/b and C/c must be located within a few hundred base pairs from a hotspot to efficiently register DSBr events as gene conversions. In contrast, markers A/a and D/d, designed to monitor crossing over and crossover interference, must be far enough from the hotspot to rarely suffer gene conversion, but close enough to ensure that most crossovers between A/a and D/d are products of the DSBr events monitored by B/b and/or C/c.

Because of limitations on the availability of markers, only one of the two sites closely bracketing a hotspot has typically been in a gene that determines a conveniently scored phenotype. Screening for conversions at that site has been used to select for analysis those tetrads that have undergone a DSBr event, involving non-sister chromatids, at the marked hotspot. Within that population, the markers at the second,

6

less conveniently scored site, have then been determined by DNA analysis. The use of conversion at one marked site, guaranteeing that homologous DSBr has occurred at the hotspot, allows more meaningful scoring of the other site for 2:2 segregation (as well as for FC or HC) of the marker. By this procedure, tetrads are identified as “two-sided”

(conversions for both B/b and C/c) and “one-sided” (conversions for only the more conveniently scored site).

The Szostak/Sun model for meiotic DSBr: We can use the Szostak/Sun DSBR model (Figure 1) to illustrate how DSBr could generate a variety of the observed tetrads.

For example, Figure 1A illustrates how mismatched base pairs created by the loss of genetic information from the chromatid undergoing DSBr, and its replacement with information from the unbroken homolog, can create a HC. Should a 3’ end be degraded past the site of a marker near the DSB, the repair process is called “gap repair” and results, invariably, in FC for the marker, as shown in Figure 1B. Such gap repair presumably occurs independently of MMR proteins. MMR, too, can generate FCs, but it operates only in the presence of known MMR proteins. MMR occurring at invasion or annealing will be directed by the invading or annealing termini and result only in FCs

(Figure 1C). Should a mismatch escape MMR at invasion or annealing, it becomes subject to MMR directed by the termini created by Holliday junction resolution. Since, in this model, both the Watson and the Crick chains are cut to effect resolution, MMR can, with presumed equal probability, result in an FC or in restoration to 2:2 segregation

(Figure 1C and 1D). Insofar as 2:2 segregation for a marker close to a DSB hotspot (in a tetrad with conversion for a close marker on the other side of the DSB site) can be

7

attributed to restoration, MMR directed by junction resolution can be inferred. Note that Figure 1 does not show all four chromatids, but illustrates the fate of only the two chromatids directly involved in the DSBr process.

Many basic features of the Szostak/Sun model have survived more than two decades' worth of tests. These features include the steps leading up to the formation of the joint-molecule, ligated DSBr intermediate (Figure 1A, step 4). A notable exception is the model's thesis that this intermediate yields both crossovers and noncrossovers. The existence of this intermediate is not in doubt -- it was verified by SCHWACHA and

KLECKNER (1995) and was shown by BÖRNER et al. (2004) to be Msh5-dependent. The ligated intermediate, however, is now understood to give rise (by an, as yet, unknown mechanism) to crossovers only (ALLERS and LICHTEN 2001a; reviewed in BISHOP and

ZICKLER 2004). In this work we present further departures from the Szostak/Sun model inspired (1) by the demonstration (GETZ et al. 2008) that, in yeast, meiotic DSBr occurs via either a "pairing pathway", first proposed by ZALEVSKY et al. (1999), or a "disjunction pathway" (see STAHL et al. 2004), and (2) by the extensive data of HOFFMANN et al. (2005), which allow us to propose and quantify a set of DSBr-pathway-specific properties.

Since this analysis rests, to a large extent, on the known functions of MMR proteins in meiosis, and since much of our understanding of these proteins is based on studies of the bacterial anti-mutation proteins MutS and MutL, we summarize here some conventional wisdom on MutS, MutL and their eukaryotic homologs.

Conventional skinny on MMR in meiosis: The bacterial MutS and MutL proteins have a demonstrated role in correcting mismatches that arise during DNA

8

replication (MODRICH 1991). Several MutS and MutL homologs have been identified in Saccharomyces cerevisiae and other (reviewed in NAKAGAWA et al. 1999;

CULLIGAN et al. 2000; ARGUESO et al. 2002;). In both bacterial and eukaryotic MMR, rectification of the mismatch is directed by a discontinuity, one one strand or the other, in the DNA duplex. MMR sacrifices the mismatched marker that is on the strand with the nearby end. The functions of most eukaryotic homologs appear to be similar to those of their bacterial cousins. For example, MutS homologs recognize mismatches and attract MutL homologs to the region, and MutL homologs attract the remaining components of the mismatch repair machinery. The expected manifestations of such

MMR in meiosis are dependent on the DSBr structures in which it is presumed to operate as well as on the intrinsic repairability of the particular mismatch.

Among MutS homologs, the Msh4-Msh5 heterodimer (BURNS et al. 1994;

HOLLINGSWORTH et al. 1995; POCHART et al. 1997) stands out in several respects. First, the

Msh4-Msh5 heterodimer alone is meiosis-specific (ROSS-MACDONALD and ROEDER 1994) and required for wild-type levels of crossing over and interference (NOVAK et al. 2001) as well as for the formation of (detectable) ligated joint-molecule DSBr intermediates

(Figure 1A step 4). The meiosis-specificity is consistent with the failure of Msh4-Msh5 to affect the mitotic mutation rate. However, the absence of Msh4-Mhs5 also fails to manifest an increase in the frequency of HCs, the expected consequence of defective meiotic MMR (ROSS-MACDONALD and ROEDER 1994; HOLLINGSWORTH et al. 1995;

WANG et al. 1999). This, plus the observation that Msh4-Msh5 protein has no identified mismatch recognition sequence like that of MutS and its yeast homolog Msh2

9

(CULLIGAN et al. 2000), has been taken as evidence that Msh4-Msh5 lacks the ability to effect MMR. (We shall revisit this view below). The --induced loss of crossing over, without a detected loss of either DSBr or MMR, suggests that the DSBs that were fated to have been interhomolog crossovers have, instead, been repaired using the sister chromatid as jig and template. Armed with this background, we shall, first, review the observations that force revision of the Szostak/Sun model of DSBR and, second, develop the alternative model (Table 1) for DSBr and MMR, along with arguments demonstrating the adequacy and economy of the model.

Two pathways for meiotic DSBr: In an effort to reveal the relationship between

DSBr and MMR, GETZ et al. (2008) used tetrad analysis to monitor, at each of two hotspots, conversion of a palindrome marker (NAG et al. 1989). DSBr-induced mismatches involving such palindrome markers often fail to undergo MMR in the presence of normal MMR proteins (Figure 1A). We refer to such mismatches as "poorly repairable mismatches" or PRMs. In addition, GETZ et al. (2008) monitored crossing over between markers bracketing the palindrome and the DSB site. They conducted these crosses in the presence or absence of Msh4, a protein required for wild-type levels of crossing over and interference as described above. Their data show that the frequencies of only those crossover tetrads with FC or with non-conversion for the palindrome marker (in many of which the PRM had presumably undergone restorational repair) were dependent on the presence of Msh4 protein. In contrast, the frequencies of both crossover-and noncrossover-tetrads with HC for the marker (i.e., those in which MMR of the PRM had failed) were, to a good approximation, Msh4-independent. Similarly, in

10

wild-type yeast, the crossover-tetrads with FC or nonconversion for the palindrome marker showed positive interference, while the crossover-tetrads with HC for the palindrome lacked such interference.

Pathway-specific rules for MMR: The observations by GETZ et al. (2008) provide good evidence that wild-type yeast, unlike S. pombe (CROMIE and SMITH 2007), has two

DSBr pathways with different MMR properties, both of which yield crossovers.

Specifically, when DSBr creates a PRM via the Msh4-dependent, interference-generating disjunction pathway, the PRM undergoes efficient repair, yielding a crossover with either FC or nonconversion for the palindrome marker. In contrast, when the PRM is created by DSBr in the Msh4-independent, non-interference, "pairing pathway", the

PRM is refractory to repair, yielding either a HC crossover or a HC noncrossover. (Note the implication that, if any FCs for the palindrome marker arise in the absence of Msh4, they must have resulted from gap repair).

Msh2-deletion and PRMs are equivalent ways of eliminating MMR from the pairing pathway: The results of GETZ et al. (2008) provide an incentive to revisit, in the framework of two pathways for DSBr, the powerful study by HOFFMANN et al. (2005), which was also designed to explore relationships between DSBr and MMR. The methods used by HOFFMANN et al. (2005) differ from those used by GETZ et al. (2008).

Specifically, HOFFMANN et al. (2005) monitored DSBr as conversion for markers that make rather "well-repairable mismatches" (WRMs) in a wild-type background, and they created MMR deficiencies by deletion of the MutS homolog Msh2 or MutL homolog

Mlh1. GETZ et al. (2008), instead, monitored DSBr events as conversions for a marker

11

that make PRMs near a DSB site. They noted that failed vs. successful MMR for the marker was associated with Msh4-independence or dependence, respectively, which enabled them to assign DSBr events to one or the other DSBr pathway. In particular, they noted that MMR was effective only in the disjunction pathway (GETZ et al. 2008).

Lest the reader protest that comparing HOFFMANN et al. (2005) with GETZ et al. (2008) is comparing apples and oranges, we argue, based on the following observations, that failure to undergo repair of PRMs on the one hand, and deletion of Msh2 on the other, are equivalent manifestations of failed MMR in the pairing pathway, and that pathway only.

STONE and PETES (2006) demonstrated that, for a WRM near a DSB site, deletion of MSH2 causes a shift from FCs to HCs independently of Msh4. This implies that msh2- deletion mutants lack MMR in the Msh4-independent, pairing pathway, but have no conspicuous MMR in the Msh4-dependent, disjunction pathway. GETZ et al. (2008) demonstrated that, in a MSH2 background, failure of the PRM to be repaired is also independent of Msh4. This implies that the observed HCs for the PRM, too, represent failure of MMR in the pairing pathway. Together, these observations suggest that a

PRM involving a palindrome in a MSH2 background may be defined as a mismatch that is refractory to Msh2-dependent MMR. The work of WANG et al. (2003), showing that Msh2 interacts aberrantly with a palindrome mismatch, further supports our thesis that the methods of HOFFMANN et al. (2005) and those of GETZ et al. (2008) support each other.

Deletion of Mlh1 causes greater loss of MMR than does deletion of Msh2: In

12

their exploration of DSBr and MMR, HOFFMANN et al. (2005) had analyzed tetrads of the ABCD/abcd-type illustrated above, with markers his (which makes rather well repairable mismatches) and BIK (which makes WRMs) representing the generic B/b and C/c sites bracketing the hotspot. Accordingly, the authors registered

DSBr events as conversion for his and/or BIK. Scoring, initially, conversion for his, they showed that the elimination of either Mlh1 or Msh2 resulted in a shift from full- to half- conversion, as predicted by the Szostak/Sun model. However, the absence of Mlh1 caused a greater MMR deficiency, i.e., a greater increase in HC/(FC+HC), than did the absence of Msh2 (Appendix I). In a one-pathway model for DSBr, this might imply that

Mlh1 simply effects MMR more efficiently than does Msh2. In a two-pathway context, on the other hand, the quantitative difference in phenotypes suggests a requirement for

Mlh1 to effect MMR not only in the pairing pathway in conjunction with Msh2, but also, without Msh2, in the disjunction pathway. Further analysis of their data supports the hypothesis of Mlh1-dependent MMR in the disjunction pathway, as described below.

Evidence for disjunction-pathway-specific, Mlh1-dependent MMR: Like msh4- msh5 mutants, but unlike msh2 mutants, mutants show reduced meiotic crossing over (HUNTER and BORTS 1997) and reduced interference (ABDULLAH et al. 2004;

Appendix II) . Moreover, these mlh1 phenotypes are observed in a MSH4-

MSH5 background only (WANG et al. 1999; ARGUESO et al. 2004). Thus, at least with respect to crossing over and interference, MLH1 functions in the disjunction pathway.

The data of HOFFMANN et al. (2005) indicate that the observed mlh1-induced gain in HCs mentioned above represents twice the mlh1-induced loss in FCs (Appendix I). Data in

13

the same paper suggest (1) that the disjunction-pathway-specific, Mlh1-dependent FC tetrads were crossovers, while the mlh1-induced HCs were noncrossovers and (2) that these noncrossovers represented twice the number of Mlh1-dependent crossovers

(Appendix III). These observations imply that Mlh1 does, indeed, play a role in MMR in the disjunction pathway, and that, ~50% of the time, Mlh1-dependent MMR in the disjunction pathway restores Mendelian (2:2) segregation of the marker.

In the disjunction pathway, Mlh1-dependent MMR occurs only in response to junction resolution: The evidence that Mlh1-dependent, disjunction-pathway-specific

MMR yields restoration and FC tetrads at equal frequencies implies that such MMR was directed by resolution of the Holliday junctions of the ligated DSBr intermediate (which is the molecular hallmark of the disjunction pathway) (Figure 1D). A corollary of this view is that mismatches created at the invasion and/or annealing phases of disjunction- pathway DSBr fail to undergo MMR prior to being incorporated in the ligated intermediate. Work by ALLERS and LICHTEN (2001b) supports this interpretation. These authors used of DNA from a MSH2 MLH1 strain to characterize

DSBr intermediates with respect to a palindrome marker that makes PRMs near a DSB site. As expected, they found that the intermediate frequently contained the marker in mismatched, heteroduplex DNA, indicating a paucity of MMR prior to ligation of the intermediate. Moreover, in none of the intermediates were all four of the marked DNA strands derived from only one parent or the other, which would have been an indication of MMR.

14

The results of ALLERS and LICHTEN (2001b) indicate that PRMs in the disjunction- pathway generally escape MMR at the invasion and annealing phases of DSBr. As discussed above, the data of HOFFMANN et al. (2005), who used markers that make well repairable mismatches, show that disjunction-pathway-specific MMR of WRMs, too, is directed by junction resolution. Thus, in wild-type yeast, the lack of MMR at the invasion and annealing phases appears to be a regular feature of disjunction-pathway

DSBr. Yet, as demonstrated by GETZ et al. (2008), mismatches induced by DSBr in the disjunction pathway are invariably repaired. Thus, MMR in this pathway occurs always, and only in response to Holliday junction resolution.

msh2-induced lack of MMR reduces two-sidedness: As shown above, lack of

MMR in the disjunction pathway revealed DSBr events that would not have been detected in the presence of MMR. The data of HOFFMANN et al. (2005) allow us to ask if the absence of Msh2-dependent (i.e., pairing-pathway specific) MMR would produce the same result. These authors screened for tetrads with conversion at his. Within this his-conversion population, they compared conversion frequencies for their second marker, BIK, in msh2 vs. MSH2 strains. Their results showed that deletion of Msh2 caused an increase in HCs among conversion tetrads, as expected. At the same time, however, two-sidedness (the conversion frequency of BIK among his conversions) was decreased, rather than increased in response to loss of Msh2. GETZ et al. (2008) reported equivalent results from crosses in a MMR-proficient background. To select for DSBr events at the hotspot, they used a "B/b" site (close to a DSB hotspot) that made WRMs.

The crosses contained a "C/c"site (on the other side of the DSB hotspot) to assess two-

15

sidedness. When the C/c site made WRMs, most of the C/c conversions tetrads were

FCs, as expected. When they replaced the WRM C/c site with a C/c site that made PRMs, most of the conversions at the C/c site were now HC, also as expected, but two- sidedness was significantly reduced.

Migrating D-loops and transient heteroduplex: How could loss of MMR cause these apparent reductions in conversion? One way is illustrated in Figure 2. The figure focuses on the heteroduplex created at the invasion stage of a DSBr event, and on the D- loop resulting from the invasion. While, in the Szostak/Sun model (Figure 1), extension of the 3' invading end simply enlarges the D-loop, in Figure 2 extension of the invading end causes the lagging as well as the leading end of the D-loop to move towards the

DSB site (FERGUSON and HOLLOMAN 1996; HOFFMANN and BORTS 2005). As a result, a mismatch formed at invasion is undone as the invading strand is extruded to reunite with its original partner. If the mismatch undergoes MMR before it is "undone" by the migrating D-loop, full conversion for the marker will indicate that a mismatch had been created. If the mismatch fails to be repaired promptly, evidence that a mismatch had been created at invasion may be erased by migration of the D-loop.

A role for Msh4-Msh5 in disjunction-pathway MMR: The concept of transient heteroduplex at invasion, together with "use-it-or-lose-it" conversion opportunities, satisfactorily accounts for MMR-dependent two-sidedness in the pairing pathway.

[Unwinding of the invasion heteroduplex (“SDSA”, PAQUES and HABER 1999) could provide a second “use-it-or-lose-it” route to MMR-dependent two-sidedness. As such, one-sidedness due to lack of MMR would be enriched among noncrossovers relative to

16

crossovers (MERKER et al. 2003; GETZ et al. 2008)]. A migrating D-loop, causing transient heteroduplex, may well characterize disjunction pathway DSBr also. Indeed, data from

ALLERS and LICHTEN (2001b) and SCHWACHA and KLECKNER (1995) demonstrated that a conspicuous fraction of ligated (i.e., disjunction-pathway) DSBr intermediates had both

Holliday junctions on the same side of the DSB site. If the disjunction pathway does, in fact, have transient invasion heteroduplex, mismatches created at invasion will be lost without a trace, because the disjunction pathway appears to routinely forego MMR at invasion and annealing, even in wild-type crosses.

What could prevent MMR in wild-type strains from acting at invasion and annealing in the disjunction pathway? By way of answer, we suggest that the lack of

MMR prior to completion of the ligated intermediate in this Msh4-Msh5-dependent pathway is due to the absence of MutS function required for recognizing mismatches in duplex DNA. Msh2 and Ms4-Msh5, the only known candidates for this role, are both disqualified, although for different reasons: Msh2, which does recognize mismatches in duplex DNA, does not operate in the disjunction pathway (see Appendix I), while

Msh4-Msh5, which, by definition, does operate in the disjunction pathway, fails to recognize mismatches in duplex DNA. It remains to consider how, in the absence of mismatch recognition, PRMs, and WRMs too, in the disjunction pathway are nevertheless invariably repaired to yield either FC or nonconversion tetrads (GETZ et al.

2008). Work by SNOWDEN et al. (2004) suggests how the unique properties of Msh4-

Msh5 might allow the dimer to promote MMR in a joint-molecule double-Holliday- junction DSBr intermediate.

17

SNOWDEN et al. (2004), working with human MutS and MutL homologs, concluded that the behavior of Msh4-Msh5 protein at a Holliday junction is like that of a MutS protein at a mismatch in duplex DNA -- Mhs4-Msh5 binds to a Holliday junction and then slides away (ACHARYA et al. 2003). The high concentration of Msh4-

Msh5 could then attract the MutL homolog. In the case of the double-Holliday-junction intermediate, reiteration of such behavior, with sliding in either direction, could lead to a traffic jam of Msh4-Msh5 in the region between the junctions, attracting Mlh1 to the entire region. Now, when nicks are introduced to resolve a junction, every mismatch between the junctions is rectified. Whether the DNA removal and replacement required for rectification is an inevitable consequence of junction cutting or is dependent on the mismatch is not answerable at this time.

The dog that didn't bark: If, as we propose, there is such a thing as Msh4-

Msh5/Mlh1-dependent MMR in the disjunction pathway, why then would msh4- msh5 deletion mutants not have a MMR-deficiency phenotype? This lack of msh4-msh5- induced increase in HCs is economically explained by the requirement of Msh4-Msh5 for the establishment of the disjunction pathway (BÖRNER et al. 2004). Without the disjunction pathway and its products to register the presence or absence of MMR, there can be no msh4-induced increase in HCs to signal that MMR had failed. Hence, the apparent paradox -- any MMR-deficiency phenotype of msh4-msh5 mutants should be detectable only in the presence of Msh4-Msh5 protein. Msh4-Msh5's vital contribution to the establishment of the disjunction pathway makes it impossible to challenge directly the proposal that, within the disjunction pathway, MMR is Msh4-Msh5-

18

dependent. A critical test of Msh4-Msh5's involvement in MMR may require an (as yet hypothetical) msh4-msh5 mutant that has retained the ability to form/stabilize the double-Holliday-junction DSBR intermediate, but has lost the ability to recruit Mlh1.

Perhaps such a separation-of-function mutant will be found and will exhibit the typical meiotic MMR-deficiency phenotype, viz., an increase in HCs.

Where the rubber meets the road: The DSBr and MMR data from the extensive study by HOFFMANN et al. (2005) were reported in terms of HCs and FCs for his and BIK in wild-type, msh2 and mlh1 crosses. In addition, as noted above, these authors screened tetrads for two-sidedness. These data can be used to test whether the model is capable of generating those observed values. Such a test requires that we first identify and evaluate (see Appendix I) the probabilities for each of the steps that can lead to a specified outcome. They are of two kinds -- those whose values are preset by the model

(nonadjustable parameters) and those that are specific to the HIS4 DSB hotspot and the markers in Hoffmann's strains (adjustable parameters).

Our model implies the following values for the four non-adjustable parameters:

• The probability that his is on the annealing side of any DSBr intermediate

= 1/2 (When his is not so situated, BIK is);

• The probability that a mismatch on the annealing side of the DSBr

intermediate in the disjunction pathway is repaired = 1;

• The probability that repair in the disjunction pathway leads to FC = 1/2;

• The probability that a mismatch on the invading side of DSBr in the

disjunction pathway remains within the migratory D-loop = 0.

19

The adjustable parameters are:

• g, the probability, specific for each marker, that a mismatch in the pairing

pathway becomes FC by gap repair (SZOSTAK et al. 1983) or FC-biased

"short-patch repair" (COÏC et al. 2000);

• R, the probability that heteroduplex rejection in the pairing pathway

(CHAMBERS et al. 1996; GOLDFARB and ALANI 2005) does not occur (see

Appendix I);

• m, the probability, specific for each marker, that mismatches in the pairing

pathway are repaired (always to FC rather than 2:2);

• E, the probability that, on the invasion side of a pairing pathway DSBr

event, the mismatch remains covered by the traveling D-loop so that it

appears as an HC in MMR-deficient crosses;

• P, the probability (expressed as number of DSBs per thousand tetrads)

that a DSB is repaired by way of the pairing pathway; and

• D, the probability (expressed as number of DSBs per thousand tetrads)

that a DSB is repaired by way of the disjunction pathway.

This list of parameters intentionally excludes the possibility of restoration of 2:2 segregation by MMr acting on a mismatch that is close to an initiating DSB. In so doing we minimize the number of parameters.

Sudoku: To estimate values for the adjustable parameters, we adopted the conventional strategy of starting with the parts of the puzzle that look easiest. For example, to obtain a value for D, we made use of the model's feature that DSBr

20

products contributed by the pairing pathway should be strictly the same for mlh1 as for msh2 crosses. Hence, any differences between those two crosses should lead directly to estimates of D, the only adjustable parameter in the disjunction pathway. To determine

the values of ghis, P, and E, we first estimated E based on the frequency of tetrads, in the

MMR mutants, that were simultaneously conversions for BIK and his (two-sided tetrads). We assume a single value for E, rather than assuming his- and BIK-specific values, on the grounds that, if we can fit the data with a single value, we could surely fit them with separate values.. The remaining two parameters were then chosen to give satisfactory fits to the FC and HC data for the MMR mutants (Appendix I).

To obtain a his-specific value for m and a value of R, we turned to the HC and FC

frequencies in the wild-type cross. We held D, ghis, E and P at the values deduced from the MMR-mutant crosses, and, for simplicity, assumed m to apply equally to mismatches created at invasion or annealing (Appendix I). The strategy for obtaining

values for the only remaining parameters, mbik and gbik, is described in Appendix I.

With a value for each of the parameters, it was then possible to compare the expected values for FC, HC and two-sidedness with the observed values. The summary

(Table 2) demonstrates that a single set of plausible parameter values satisfies both the

HC and FC data, as well as the two-sidedness, for each of the three genotypes, msh2, mlh1 and WT. It is gratifying that, with eight adjustable parameters, the model can account for 12 observations.

DISCUSSION

21

The data of HOFFMANN et al. (2005) and GETZ et al. (2008) call for an updated view of DSBr and MMR. Our interpretation of these data, above, has allowed us to assign specific attributes to each of two pathways for DSBr in wild-type yeast (Table 1), leading to a molecular model that illustrates how repair of the programmed, meiotic double-strand breaks might occur in each of the two pathways (Figures 3 and 4).

The model invites us to revisit several meiotic phenomena previously published and interpreted. These include the wide variation in the estimated lengths of regions of heteroduplex resulting from DSBr, MMR-dependent two-sidedness, the apparent mutual exclusivity of interference and gene conversion in Sordaria fimicola, and an unexpected DSBr intermediate. Below, we review these phenomena and offer interpretations within the framework of our two-pathway model with traveling D- loops.

Heteroduplex lengths: Several studies of DSBr in yeast have yielded data that create impressions of the length of the regions of heteroduplex created at the invasion and/or the annealing stages of DSBr. Some estimates were based on physical analyses

(e.g., microscopy or gel electrophoresis), others on genetic analyses. The problem is that the impressions appear to contradict each other. For example, genetic analyses suggest that conversion tracts (which depend on regions of heteroduplex) may often be “long”

(DETLOFF et al. 1992, FOSS, et al. 1999); Microscopy by BELL and BYERS (1983), on the other hand, indicates that double-Holliday-junction intermediates tend to be “short”. This estimate by microscopy is consistent with gel electrophoresis data reported by

SCHWACHA and KLECKNER (1995), which imply that the region of DNA between two

22

Holliday junctions in observed intermediates is usually short. Our model suggests that these apparent discrepancies reflect the differences between the pairing and disjunction pathways. Both the gel electrophoresis and the microscopy focus on double-Holliday- junction intermediates, i.e. on disjunction-pathway intermediates, which, in our model, are short (Figure 3, and see below). In contrast, the genetic studies indicating long regions of heteroduplex (Figure 4; DETLOFF et al. 1992; HILLERS and STAHL 1999; FOSS et al. 1999) were based on HC tetrads, which are manifestly products of the pairing pathway (GETZ et al. 2008; and see above). This interpretation is also consistent with the msh4-induced increase in the average length of conversion tracts, as deduced from genetic data of MANCERA et al. (2008); according to our model, the absence of Msh4 would eliminate the disjunction pathway with its short conversion tracts, thereby increasing the average conversion tract length).

It should be noted that electrophoresis studies have yielded little evidence on the structure of pairing-pathway intermediates. While this could reflect an ephemeral nature, we suggest that it reflects (instead or also) the variable and “unfinished” nature of these intermediates (Figure 4) and/or a length that frequently exceeds the distance between the restriction sites used by the investigator to liberate the intermediates from the chromosome. Furthermore, frequent erosion of the 3’ single stranded ends at the

DSB in the pairing pathway (g > 0) could confound the detection of DSBs specific to that pathway.

MMR-dependent two-sidedness: HOFFMANN et al. (2005) explained the phenomenon of MMR-dependent two-sidedness with the suggestion “... that in wild-

23

type cells the initial DSB repair event is two sided. The absence of MMR, by either mutation or use of poorly repaired palindromes, allows a second, unbiased mispair removal pathway to restore a proportion of heteroduplexes, leading to apparent one- sided events.” This proposal suffers from several problems, one of which is that the short-patch system hypothesized by HOFFMANN et al. (2005) to be responsible for this unbiased MMR is claimed by its discoverers in yeast (COÏC et al. 2000) to be biased against the marker on the invading strand, thus favoring FCs over restorations (and see

Appendix I, Bits and pieces, Disparity between the two classes of HCs). Our model suggests, instead, that a traveling D-loop in the pairing pathway allows a mismatch created at invasion only a transient opportunity to enjoy MMR (Figure 4). Thus, the

“one-sidedness” (fraction of conversions for one of the bracketing markers that are 2:2 for the other) reflects the failure of MMR to turn such a mismatch into an FC before the heteroduplex containing the mismatch is undone.

Gene conversion in Sordaria: Our model for DSBr obliges us to revisit the observation by KITANI (1978) that, unlike yeast-crossovers, individual Sordaria- crossovers exhibit either gene conversion or crossover interference, but not both. STAHL and FOSS (2008) had suggested that, in both organisms, MMR in the disjunction pathway is directed by junction resolution. In yeast such MMR would result in interfering crossovers with FC or 2:2 segregation for the marker with equal probability while, in Sordaria, junction-directed MMR would yield interfering crossovers with only

2:2 segregation. This proposal rested on the assumption that the disjunction pathways for the two organisms differ from each other in the same manner as do the pairing

24

pathways; viz. yeast has predominantly asymmetric heteroduplexes, revealed by 5:3 segregation, while Sordaria has a high frequency of symmetric heteroduplex, revealed as aberrant 4:4 segregation (KITANI and OLIVE 1967; STAHL and HILLERS 2000; STAHL and

FOSS 2008). We now suggest a second possibility as to why interference and conversion appear mutually exclusive in Sordaria. This possibility is based on our understanding that Kitani's monitored markers were not selected for proximity to a DSB hotspot. If, as for yeast, Sordaria's regions of heteroduplex in disjunction-pathway intermediates are relatively short, Kitani's markers might only rarely generate disjunction-pathway mismatches, in which case the interfering crossovers would usually lack conversion regardless of whether the intermediate had symmetric or asymmetric heteroduplex. In contrast, crossovers in the pairing pathway, with its longer conversion tracts, would be relatively more likely to involve a marker far from the DSB site in heteroduplex.

This explanation (for the frequent lack of detectable DSBr participation in the disjunction pathway for marked sites relatively remote from the DSB site) may apply also to the apparent lack of BIK’s participation in disjunction pathway DSBr (Appendix

I, Bits and pieces: The BIK data).

An unexpected DSBr intermediate: ALLERS and LICHTEN (2001b) reported a

DSBr intermediate that they labeled "JM2". JM2 had two Holliday junctions, identifying it as a disjunction-pathway intermediate. The junctions, however, were on the same side of the DSB site, beyond a palindrome marker located near the DSB site (on the right in

Figure 5). Moreover, the palindrome marker was in parental configuration, which the authors had some difficulty explaining. Within the framework of our model featuring a

25

traveling D-loop and junction-directed repair in the disjunction pathway, we offer the scheme illustrated in Figure 5. A most attractive feature of the scheme is that, in harmony with the observations of ALLERS and LICHTEN (2001a), it yields restoration intermediates but never FC ones. We note that, if JM2s had been frequent in the data of

HOFFMANN et al. (2005), our model would demand that the excess of HCs seen in the mlh1 cross would be more than twice the excess of FCs seen in the MLH1 msh2 cross.

The data do not lean that way.

Predictions: Our view of the roles of the various Mut homologs in meiotic MMR makes a number of predictions. (1) Crosses carried out with a poorly repairable mismatch close to a DSB site in a MSH2 background should have the same

HC/(FC+HC) ratios as do crosses in a msh2 background with a marker, at the same site, that makes well repairable mismatches. (2) The proposal that deletion of MSH4 in yeast eliminates a crossover pathway that yields only nonconversions and one-sided FCs implies that msh4 mutants should show a modest increase not only in the HC/(FC+HC) ratio for a palindrome marker at HIS4 (by decreasing FC crossovers as shown by GETZ et al. 2008), but also in the fraction of DSBr events that are two-sided (by decreasing the one-sided tetrads). (3) GETZ et al. (2008) noted that, in msh4 mutants, the loss of conversion crossovers appeared to result conspicuously in a gain in nonconversion noncrossovers (presumably by sister repair). The studies of HOFFMANN et al. (2005) suggest that, in contrast, deletion of MLH1, appears to turn FC interfering crossovers into HC noncrossovers. The possibility that these HC noncrossovers would interfere with each other is a heady one, but difficult to test. (4) Another test of the model

26

concerns the effect of Msh4 on HC/(FC+HC) ratios in msh2 and mlh1 mutants.

Although these two single mutants differ appreciably with respect to HC/(FC+HC) ratios at his, when the disjunction pathway is removed by deletion of MSH4 the msh2 msh4 and mlh1 msh4 double mutants should be seen to have identical values for that ratio. (5) According to our model, only the pairing pathway produces two-sided tetrads, implying that two-sided crossovers should not manifest positive interference. Expected and observed phenotypes for a variety of relevant genotypes are summarized in Table

3. A prudent investigator aiming to challenge these predictions would probably choose to work at HIS4 and with the strains of HOFFMANN et al. (2005).

ACKNOWLEDGMENTS

We thank Eva Hoffmann, John Fowler, Rhona Borts, Nancy Hollingsworth,

Elizabeth Housworth, Michael Lichten and anonymous reviewers for helpful comments. Unless otherwise specified, statistical calculations were conducted with the aid of VassarStats (http://faculty.vassar.edu/lowry/VassarStats.html).

LITERATURE CITED

ABDULLAH, M. F. F., and R.H. BORTS, 2001 Meiotic recombination frequencies are

affected by nutritional states in Saccharomyces cerevisiae. Proc. Natl. Acad. Sci. USA

98: 14526-14529.

27

ABDULLAH, M. F. F., E.R. HOFFMANN, V.E. COTTON and R.H. BORTS, 2004 A role for the

MutL homologue MLH2 in controlling heteroduplex formation and in regulating

between two different crossover pathways in budding yeast. Cytogenet. Genome

Res. 107: 180-190.

ACHARYA, S., P. L. FOSTER, P. BROOKS and R. FISHEL, 2003 The coordinated functions of

the MutS and MutL proteins in mismatch repair. Molec. Cell 12: 233 – 246.

ALLERS, T., and M. LICHTEN, 2001a Differential timing and control of noncrossover and

crossover recombination during meiosis. Cell 106: 47-57.

ALLERS, T., and M. LICHTEN, 2001b Intermediates of yeast meiotic recombination

contain heteroduplex DNA. Molec. Cell 8: 225-231.

ARGUESO, J. L., D. SMITH, J. YI, M. WAASE, S. SARIN et al., 2002 Analysis of conditional

mutations in the Saccharomyces cerevisiae MLH1 gene in mismatch repair and in

meiotic crossing over. Genetics 160: 909-921

ARGUESO, J. L., J. WANAT, Z. GEMICI, and E. ALANI, 2004 Competing crossover

pathways act during meiosis in Saccharomyces cerevisiae. Genetics 168: 1805-1816.

BELL, L. R., AND B. BYERS, 1983 Homologous association of chromosomal DNA during

yeast meiosis. Cold Spring Harbor Symp. Quant. Biol. 47: 829-840.

BERCHOWITZ, L. E., and G. P. COPENHAVER, 2010 Genetic interference: don’t stand so

close to me. Current Genomics 11: 91-102.

BISHOP, D. K., and D. ZICKLER, 2004 Early decision; meiotic crossover interference prior

to stable strand exchange and synapsis. Cell 117: 9-15.

BISHOP, D. K., M. S. WILLIAMSON, S. FOGEL and R. D. KOLODNER, 1987 The role of

28

heteroduplex correction in gene conversion in Saccharomyces cerevisiae. Nature 328:

362-364.

BÖRNER, G. V., N. KLECKNER and N. HUNTER, 2004 Crossover/noncrossover

differentiation, synaptonemal complex formation, and regulatory surveillance at the

leptotene/zygotene transition of meiosis. Cell 117: 29–45.

BURNS N., B. GRIMWADE, P. B. ROSS-MACDONALD, E-Y. CHOI, K. FINBERG et al. 1994

Large-scale analysis of gene expression, protein localization, and gene disruption in

Saccharomyces cerevisiae, Genes. Dev. 8: 1087-1105.

CHAMBERS, S. R., N. HUNTER, E. J. LOUIS and R. H. BORTS, 1996 The mismatch repair

system reduces meiotic homeologous recombination and stimulates recombination-

dependent chromosome loss. Mol. Cell. Biol. 16: 6110-6120.

COÏC, E., L. GLUCK and F. FABRE, 2000 Evidence for short-patch mismatch repair in

Saccharomyces cerevisiae. EMBO J. 19: 3408-3417.

COTTON, V. E., E. R. HOFFMANN, A. F. F. ABDULLAH and R. H. BORTS, 2009 Interaction of

genetic and environmental factors in Saccharomyces cerevisiae meiosis: the Devil is in

the details. Methods in Molec. Biol. 557: 3-20.

CROMIE, G. A. and G. R. SMITH, 2007 Meiotic recombination in Schizzosacharomyces

pombe: a paradigm for genetic and molecular analysis. Genome Dyn. Stab. DOI

10.1007/7050_2007_025.

CULLIGAN, K. M., G. MEYER-GAUEN, J. LYONS-WEILER and J. B. HAYS, 2000 Evolutionary

origin, diversification and specialization of eukaryotic MutS homolog mismatch

repair proteins. Nucleic Acids Research 28: 463-471.

29

DETLOFF, P., M. A. WHITE and T. D. PETES, 1992 Analysis of a gene conversion gradient

at the HIS4 locus in Saccharomyces cerevisiae. Genetics 132: 113-123.

FERGUSON, D. O., and W. K. HOLLOMAN, 1996 Recombinational repair of gaps in DNA

is asymmetric in Ustilago maydis and can be explained by a migrating D-loop model.

Proc. Natl. Acad. Sci. USA 93: 5419-5424.

FOSS, E., R. LANDE, F. W. STAHL and C.M. STEINBERG, 1993 Chiasma interference as a

function of genetic distance. Genetics 133: 681-691.

FOSS, H. M., K. J. HILLERS and F. W. STAHL, 1999 The conversion gradient at HIS4 in

Saccharomyces cerevisiae. II. A role for mismatch repair directed by biased resolution

of the recombinational intermediate. Genetics 153: 573-583.

GETZ, T. J., S. A. BANSE, L. S. YOUNG, A. V. BANSE, J. SWANSON, et al., 2008 Reduced

mismatch repair of heteroduplexes reveals “non”-interfering crossing over in wild-

type Saccharomyces cerevisiae. Genetics 178: 1251–1269.

GILBERTSON, L. A., and F. W. STAHL 1996 A test of the double strand break model for

meiotic recombination in Saccharomyces cerevisiae. Genetics 144: 27-41.

GOLDFARB, T., and E. ALANI, 2005 Distinct roles for the Saccharomyces cerevisiae

mismatch repair proteins in heteroduplex rejection, mismatch repair and

nonhomologous tail removal. Genetics 169: 563-574.

HILLERS, K. J. and F. W. STAHL, 1999 The conversion gradient at HIS4 of Saccharomyces

cerevisiae. I. Heteroduplex rejection and restoration of Mendelian segregation.

Genetics 153: 555-572.

HOFFMANN, E. R., and R. H. BORTS, 2004 Meiotic recombination intermediates and

30

mismatch repair proteins. Cytogenet. Genome Res. 107: 232-248.

HOFFMANN, E. R., and R. H. BORTS, 2005 Trans events associated with crossovers are

revealed in the absence of mismatch repair genes in Saccharomyces cerevisiae.

Genetics 169: 1305-1310.

HOFFMANN, E. R., E. ERIKSSON, B. J. HERBERT and R. H. BORTS, 2005 MLH1 and MSH2

promote the symmetry of double-strand break repair events at the HIS4 hotspot in

Saccharomyces cerevisiae. Genetics 169: 1291-1303.

HOLLINGSWORTH, N. M., and S. J. BRILL, 2004 The Mus81 solution to resolution:

generating meiotic crossovers without Holliday junctions. Genes and Dev. 18: 117-

125.

HOLLINGSWORTH, N. M., L. PONTE and C. HALSEY, 1995 Msh5, a novel MutS homolog,

facilitates meiotic reciprocal recombination between homologs in Saccharomyces

cerevisiae but not mismatch repair. Genes Dev. 9: 1728-1739

HOTCHKISS, R. D., 1971 Toward a general theory of in DNA.

Advances in Genetics 16: 325-348.

HUNTER, N., 2007 Meiotic recombination. Molecular Genetics of Recombination (AGUILERA

and ROTHSTEIN, Eds). New York: Springer. pp. 381–442.

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.

HUNTER, N., and N. KLECKNER, 2001 The single-end invasion: an asymmetric

intermediate at the double-strand break to double-Holliday junction transition of

meiotic tecombination. Cell 106: 59-70.

31

JESSOP, L. and M. LICHTEN, 2008 Mus81/Mms4 endonuclease and helicase

collaborate to ensure proper recombination intermediate metabolism during

meiosis, Mol. Cell 31: 313-323.

KITANI, Y., 1978 Absence of interference in association with gene conversion in Sordaria

fimicola, and presence of interference in association with ordinary recombination.

Genetics 89: 467-497.

KITANI, Y., and L. S. OLIVE, 1967 Genetics of Sordaria fimicola. VI. Gene conversion at the

g locus in mutant x wild type crosses. Genetics 57: 767–782.

LAMB, B. C., and M. R. T. WICKRAMARATNE, 1973 Corresponding-site interference,

synaptinemal complex structure, and 8+:0m and 7+:1m octads from wild-type x

mutant crosses of Ascobolus immersus. Genet. Res. 22: 113-124.

MALOISEL, L., J. BHARGAVA and G. S. ROEDER, 2004 A role for DNA polymerase in gene

conversion and crossing over during meiosis in Saccharomyces cerevisiae. Genetics

167: 1133-1142.

MANCERA, E, R. BOURGON, A. BROZZI, W. HUBER and L. M. STEINMETZ, 2008 High-

resolution mapping of meiotic crossovers and non-crossovers in yeast. Nature 454:

479-485.

MERKER, J. D., M. DOMINSKA and T. D. PETES, 2003 Patterns of heteroduplex formation

associated with the initiation of meiotic recombination in the yeast Saccharomyces

cerevisiae. Genetics 165: 47–63.

MODRICH, P., 1991 Mechanisms and biological effects of mismatch repair. Annu. Rev.

Genet. 25: 229-253.

32

NAG, D. K., M. A. WHITE and T. D. PETES, 1989 Palindromic sequences in heteroduplex

DNA inhibit mismatch repair in yeast. Nature 340: 318–320.

NAKAGAWA, T., A. DATTA and R. D. KOLODNER, 1999 Multiple functions of MutS- and

MutL-related heterocomplexes. Proc. Natl. Acad. Sci. USA 96: 14186-14188.

NOVAK, J. E., P. B. ROSS-MACDONALD and G. S. ROEDER. 2001 The budding yeast Msh4

protein functions in chromosome synapsis and the regulation of crossover

distribution. Genetics 158: 1013-1025.

OH, S. D., J. P. LAO, A. F. TAYLOR, G. R. SMITH and N. HUNTER, 2008 RecQ Helicase,

Sgs1, and XPF family endonuclease, Mus81-Mms4, resolve aberrant joint molecules

during meiotic recombination. Mol. Cell 31: 324-336.

PAPAZIAN, H. P., 1952 The analysis of tetrad data. Genetics 37: 175–188.

PAQUES, F., and J. E. HABER, 1999 Multiple pathways of recombination induced by

double-strand breaks in Saccharomyces cerevisiae. Microbiol. Mol. Biol. Rev. 63: 349–

404.

POCHART, P., D. WOLTERING and N. M. HOLLINGSWORTH, 1997 Conserved properties

between functionally distinct MutS homologs in yeast. Jour. Biol. Chem. 272: 30345-

30349.

RADFORD, S. J., S. MCMAHAN, H. L. BLANTON, and J. SEKELSKY, 2007 Heteroduplex DNA

in meiotic recombination in Drosophila mei-9 mutants. Genetics 176: 63 - 72.

RAY, A., N. MACHIN, and F.W. STAHL, 1989 A DNA double-chain break stimulates

triparental recombination in Saccharomyces cerevisiae. Proc. Natl. Acad. Sci. USA 86:

6225-6229.

33

ROSS-MACDONALD, P., and G. S. ROEDER, 1994 Mutation of a meiosis-specific MutS

homolog decreases crossing over but not mismatch correction. Cell 79: 1069-1080.

SCHWACHA, A., and N. KLECKNER, 1995 Identification of double Holliday junctions as

intermediates in meiotic recombination. Cell 83: 783-791.

SHINOHARA, M., K. SAKAI, A. SHINOHARA and D. K. BISHOP, 2003 Crossover interference

in Saccharomyces cerevisiae requires a TID1/RDH54- and DMC1-dependent pathway.

Genetics 163: 1273-1286.

SMITH, K. N., and A. NICOLAS, 1998 Recombinatio`n at work for meiosis. Curr. Op. in

Genetics & Development 8: 200-211.

SNOWDEN, T., S. ACHARYA, C. BUTZ, M. BERARDINI and R. FISHEL, 2004 hMSH4-hMSH5

recognizes Holliday junctions and forms a meiosis-specific sliding clamp that links

homologous . Mol. Cell 15: 437–451.

STAHL, F. W., 2008 On the “NPD ratio” as a test for crossover interference. Genetics 179:

701-704.

STAHL, F. W. and H. M. FOSS, 2008 But see KITANI (1978). Genetics 178: 1141-1145.

STAHL, F. W., and K. J. HILLERS, 2000 Heteroduplex rejection in yeast? Genetics 154:

1913-1916.

STAHL, F. W., and E. A. HOUSWORTH, 2009 Methods for analysis of crossover

interference in S. cerevisiae. Methods Molec. Biol. 557: 35-53.

STAHL, F.W. and R. LANDE, 1995 Estimating interference and linkage map distance from

two-factor tetrad data. Genetics 139: 1449-1454.

STAHL, F. W., H. M. FOSS, L. S. YOUNG, R. H. BORTS, M. F. F. ABDULLAH et al., 2004 Does

34

crossover interference count in Saccharomyces cerevisiae? Genetics: 168: 35-48.

STONE, J. E., and T. D. PETES, 2006 Analysis of the proteins involved in the in vivo repair

of base-base mismatches and four-base loops formed during meiotic recombination

in the yeast Saccharomyces cerevisiae. Genetics 173: 1223-1239.

SUN, H., D. TRECO and J. W. SZOSTAK, 1991 Extensive 3’-overhanging, single-stranded

DNA associated with the meiosis-specific double-strand breaks at the ARG4

recombination initiation site. Cell 64: 1155-1161.

SURTEES, J. A., J-L ARGUESO and E. ALANI, 2004 Mismatch repair proteins: Key

regulators of genetic recombination. Cytogenet. Genome Res. 107: 146-159.

SZOSTAK, J., T. L. ORR-WEAVER, R. J. ROTHSTEIN and F.W. STAHL, 1983 The double-

strand-break repair model for recombination. Cell 33: 25-35.

TERASAWA, M., H. OGAWA, Y. TSUKAMOTO, M. SHINOHARA, K. SHIRAHIGE et al., 2007

Meiotic recombination-related DNA synthesis and its implications for cross-over

and non-cross-over recombinant formation. Proc. Natl. Acad. Sci. USA 104: 5965-

5970.

WANG, H., Y. YANG, M. J. SCHOFIELD, C. DU, Y. FRIDMAN et al., 2003 DNA bending and

unbending by MutS govern mismatch recognition and specificity. Proc. Natl. Acad.

Sci. USA 100: 14822-14827.

WANG, T-F., N. KLECKNER and N. HUNTER, 1999 Functional specificity of MutL

homologs in yeast: Evidence for three Mlh1-based heterocomplexes with distinct

roles during meiosis in recombination and mismatch correction. Proc. Natl. Acad.

Sci. USA 96: 13914-13919.

35

ZALEVSKY, J., A. J. MACQUEEN, J. B. DUFFY, K. J. KEMPHUES and A. M. VILLENEUVE, 1999

Crossing over during Caenorhabditis elegans meiosis requires a conserved MutS-

based pathway that is partially dispensable in budding yeast. Genetics 153: 1271–

1283.

Appendix I

Assigning Parameter Values

Franklin W. Stahl

HOFFMANN et al. (2005) reported the values for the relative frequencies of two- vs. one-sidedness as well as those of HCs and FCs for his and BIK in wild-type, msh2 and mlh1 crosses. To test whether the DSBr model proposed above (Table 1,

Figures 3 and 4) can accommodate these values it was necessary to evaluate the parameters that were identified as defining the model (Table 2). This appendix describes how such evaluation was achieved.

Parameter values were assigned on the basis of data in Table 5 of HOFFMANN et al. (2005). The key observation is that the fraction of FCs among conversions in the mlh1 mutant strain (12/112) is significantly less than that in the msh2 strain (33/86; p <

0.0001). These data provide the statistical support for the view that, whereas Mlh1 is required for MMR in both the pairing and disjunction pathways, Msh2 has no role in the disjunction pathway.

To allow comparison between the observed and expected his conversion frequencies (Table 2) in msh2, mlh1 and wild-type strains, we expressed the entries in

Table 5 of HOFFMANN et al. (2005) as events per 1000 tetrads, rounding to whole

36

numbers (Table A1; a convenience that exaggerates the significance of the msh2 and mlh1 data, while reducing that of the wild-type data). In addition, we combined the four conversion classes (6:2, 2:6, 5:3, 3:5) of HOFFMANN et al. (2005) into two classes, FC and

HC, ignoring for now, the implications of some conversion disparities (Appendix IV).

Finally, we assumed that the BIK and his ends of the broken chromosome are equally likely to initiate invasion.

We now have the tools to evaluate the adjustable parameter D. Following the model, which assumes that msh2 and mlh1 have identical phenotypes in the pairing pathway, the excess of HCs in the mlh1 strain over HCs in the MLH1 msh2 strain (171 -

97= 74: Table A1) corresponds to D/2, the number of tetrads in which his is involved on the annealing side of a disjunction-pathway event. Also by the model, the excess of FCs in MLH1 msh2 over FCs in mlh1 should be D/4. That excess is 61 - 21 = 40 (Table A1).

We arrive at a value for D by comparing the two mutants with respect to HCs (D/2 =

74; D = 148) and, independently, by comparing them with respect to FCs (D/4 = 40; D =

160). We average these values to get a working estimate of D = 154.

To obtain values for P and ghis for each of the mutant strains, msh2 and mlh1, we write expressions for the number of tetrads demanded by the model to result in FC or

HC for his.

For MLH1 msh2: FC = P ghis + D/4; HC = (P/2)(E+1)(1- ghis).

For mlh1: FC = P ghis; HC = (P/2)(E+1)(1- ghis) + D/2

Based on the approximately 30% of two-sidedness (measured as the fraction his conversion tetrads that are also conversions for BIK) and the rarity of BIK FCs among

37

these two-sided events (Table 5 of HOFFMANN et al. 2005; Table A4A), we set E, the fraction of pairing pathway events in which the sliding D-loop comes to rest over the

BIK marker on the invasion side of the DSB at E = 0.3. Absent an estimate of E for his, we assume the same 0.3 value. With D = 154 and E = 0.3, values for P can be extracted from the equations. Solving the FC and HC equations for msh2, we get P = 172, while

solving the equations for mlh1 gives P = 166. We can then get values for ghis from both

the msh2 and the mlh1 data. These are 0.131 and 0.127, respectively. The P and ghis values

for the two strains are, as expected, similar, and we settled on P = 171 and ghis = 0.127 for personal reasons. (Recall that our modest goal is to determine whether there is a set of parameters that allows the model to fit the data, rather than to determine best estimates of those parameters.) Calculating the values for the FCs and HCs in the wild-type crosses (Table A3) required the evaluation of two additional parameters -- one for the

probability (mhis) that Msh2- Mlh1-dependent MMR leads to full conversion in the pairing pathway, and one for the probability (R) that heteroduplex rejection does not occur. The need for R in the wild-type cross is signaled by the otherwise puzzling observation (Table A1) that total conversions at his are lower in wild type than in the

MMR mutants despite the greater two-sidedness in wild-type. In limiting the model to eight adjustable parameters, we ignore a variety of possible additional factors with the hope that in so doing we are better exposing the skeletal features of the model. For instance, we are assuming that, for our markers, MMR in the pairing pathway is directed by the invading and annealing ends created by the DSB, with the result that all

38

MMR in that pathway results in FC. Justification for this simplification is found in the adequacy of our simple model (Table 2).

We next address the features of the model that link two-sidedness directly to

MMR (Table A4). Since BIK and his are simultaneously present in all the crosses, their conversions must depend on common values of P, D and R. Because we have assumed that E, also, is the same for his and BIK, we need pick only m and g values for BIK. If there were a large body of BIK data analogous to the his data, we might have estimated

mBIK and gBIK as we did mhis and ghis. However, the best data come from the BIK conversions among tetrads selected for being his conversions, forcing a change in

strategy. Thus, for both the MLH1 msh2 and mlh1 crosses, we sought and found a gBIK value (0.04, by trial and error) that gave satisfactory fits to these two-sidedness data as well as to the FC/(FC + HC) value for BIK among his conversions in the two MMR- defective strains (Table A4A). For the wild-type cross, the only remaining parameter to

be estimated is mBIK, which we chose to fit the two-sidedness data exactly (Table A4B).

This mBIK value proved to give a good fit to the wild-type HC/(FC + HC) ratio for BIK among his conversions, supporting the view that the Mlh1-, Msh2-dependent two- sidedness is a reflection of MMR per se.

Bits and pieces:

Disparity between the two classes of HCs: Data of HOFFMANN et al. (2005) showed disparities in the rates of conversion to his and HIS. In our Sudoku, we ignored the disparity, raising the possibility that in doing so we have concealed important information. The default hypothesis for disparity is differential rates of DSBs on the two

39

homologs, and disparity so caused would be without consequence for our analysis.

HOFFMANN et al. (2005) noted that the disparity was statistically significant only for the

HCs and attributed it to different rates of short-patch repair (COÏC et al. 2000) to restoration for the two different mismatches. This interpretation appeared to strengthen the authors’ proposal that short-patch repair, operating primarily in the absence of

Msh2 and Mlh1, was responsible for the one-sidedness seen in the msh2 and mlh1 crosses. Shortcomings of this proposal, along with support for the differential DSB hypothesis, are detailed in Appendix IV. The significance of a well-supported proposal for differential DSBs is that it undermines restoration by short-patch repair as an explanation for the one-sidedness.

The BIK data: HOFFMANN et al. (2005) noted that the rate of conversion at BIK, especially in the MMR mutants, is less than that at his. Other aspects of the BIK data combine with this observation to suggest that our Sudoku is not quite finished. Most of the BIK data, in Table 5 of HOFFMANN et al. (2005), were collected from tetrads that were pre-selected as his conversions. As such, they were two-sided and, according to our model, must be from the pairing pathway. Consequently, those BIK data were expected to be the same for the mlh1 and the msh2 crosses, which they seem to be (mlh1: 3 FC, 48

HC; msh2: 4 FC, 37 HC. p = 0.7). That’s cool, but BIK, in unselected tetrads, fails to show the HC and FC differences that characterize the his data; i.e, for BIK, there is no evidence of mlh1-specific HCs or of MLH1 msh2-specfic FCs. (mlh1: 1 FC, 10 HC; msh2: 0 FC, 9

HC). The numbers are small and could be ignored for that reason. However, the 0:9 ratio for BIK in the msh2 cross is significantly different (p = 0.02) from the corresponding

40

ratio, 33:53, for his. An economical interpretation for both the relatively low conversion rate of BIK (HOFFMANN et al. 2005) and its failure to show properties characteristic of his in the disjunction pathway is that the disjunction pathway conversion tracts usually fail to include BIK. This may be simply because conversion tracts in the disjunction pathway are short and BIK is farther (maximum ca. 600 bp) from the HIS4 hotspot than is his (maximum ca. 266 bp ) (E. R. HOFFMANN, pers. com.). This possibility, which is of no consequence for our basic Sudoku, predicts that most or all BIK conversion crossovers come from the pairing pathway and, consequently, will have weaker interference than do his conversion crossovers.

Multiple events? In all studies of conversion at recombination hotspots in yeast there are tetrads that would be interpreted on the basis of any model as due to multiple

DSBr events, and these tetrads are usually exempted from interpretation. However, as pointed out by MERKER et al. (2003), the identification of the less obvious multiple events is unavoidably model-dependent. Furthermore, the estimation of expected frequencies of multiple events is confounded by the possibility of negative interference between

DSBs at the same level (LAMB and WICKRAMARATNE 1973). Among other candidates for complex events we note two-sided HCs that are crossovers with heteroduplex on the same chromatid in the trans configuration (HOFFMANN and BORTS 2005). Such tetrads might result from a pairing-pathway noncrossover being accompanied by a not-so- incidental exchange (RAY et al. 1989) provoked by the 3’ end of a protruding SS-DNA whisker (HOTCHKISS 1971) (Figure A1).

41

ACKNOWLEDGMENT

Jette Foss provided invaluable editorial assistance.

Appendix II

Crossover interference in mlh1 crosses

Franklin W. Stahl

Deletions of Mlh1 cause a reduction in crossing over, generally to a lesser degree than do deletions of Msh4 or Msh5 (WANG et al. 1999; ABDULLAH et al. 2004; ARGUESO et al. 2004). Deletions of Mlh1 are hypostatic to deletions of Msh4 (WANG et al. 1999;

ARGUESO et al. 2004), clearly indicating that Mlh1 promotes crossing over in the disjunction pathway. Consequently, it is a strong expectation of our simple two- pathway model that mlh1Δ mutants should have reduced interference, commensurate with the degree to which they have reduced crossing over. Confirmation of this expectation is befogged by claims in the literature regarding the effect, if any, of Mlh1 on crossover interference.

ABDULLAH et al. (2004) wrote, “ … as might be predicted for genes of the MSH4 pathway, interference was abolished by deletion of MLH1 and MLH3.” In the same year, ARGUESO et al. (2004) wrote, “In wild type, interference was significant at all intervals analyzed in chromosome XV… These values did not significantly change in mlh1 … strains, which were shown previously to maintain interference.” These two authors have reached extreme, opposite conclusions, neither of which is in concordance with the two-pathway model (GETZ et al. 2008).

42

Shortcomings in the analyses by ABDULLAH et al. (2004): ABDULLAH et al. (2004) presented a three-factor test for interference on chromosome III. The data revealed interference in wild type and fit well with the null hypothesis of no interference in the mlh1 deletion mutant. The conclusion that mlh1Δ lacked interference was embraced by the authors as being expected. However, the authors appear to have similar data for chromosome VII, but they present no analysis of those data for interference. The authors make no tests of their two factor data for either chromosome.

Shortcomings in the analysis by ARGUESO et al. (2004): ARGUESO et al. (2004) conducted both three-factor and two-factor tests for interference and reported significant levels of interference in mlh1Δ crosses, leading them to conclude that deletion of MLH1 is without effect on interference. However, their analyses, too, have several problems:

1. The authors represent their tests as supporting the conclusion that “[interference] did not significantly change in mlh1 … strains…”. However such a conclusion requires statistical tests of mlh1 versus wild-type data, while the only tests presented are of mlh1 and wild type versus the null hypothesis of no interference.

2. To test their two-factor data for interference, the authors used an inefficient method

(PAPAZIAN 1952) for calculating the expected frequency of NPDs. In 2004, shortcomings of statistical tests based on that expectation were not generally appreciated. More recently (STAHL 2008) such tests were shown to give false positives—i.e., to give Chi- square p values that are too small. Two of Argueso’s conclusions of significance at the

5% level seem to have fallen victim to that shortcoming of the test (LYS-HIS and ADE-

43

HIS, Table A5). Further undermining the usefulness of these data for concluding that interference did not significantly change in mlh1 strains is that the ADE-HIS interval comprises a major fraction of the LYS-HIS interval, so those two observations are not independent.

3. As a test for interference revealed by three-factor crosses, ARGUESO et al. (2004) calculated coefficients of coincidence according to SHINOHARA et al. (2003). These “COC tests” look convincing for the presence of interference in the mlh1 strain (their Table 4) for the URA-LEU-LYS intervals, but not for the LEU-LYS-ADE intervals, where the tetrad data do not indicate interference at the 5% level of significance. Their “Spore” data for the LEU-LYS-ADE intervals, on the other hand, are reported as manifesting interference significant at the 5% level. However, the description provided by ARGUESO et al. (2004) of their COC test for “Spore” data suggests that it has incorrectly indicated significance. Since these data are simply their disaggregated tetrad data, supplemented by spore data from tetrads with fewer than four viable spores, the conclusion of significance is justified only if the analysis recognizes that many of the recombinants in the data set arose in pairs. Since the “RANA” software employed assumes that all spores arise from independent events (S. E. ZANDERS and E. ALANI, pers. com.), p values based on those data are underestimated. Furthermore, as a test for an mlh1-induced change in interference, COCs suffer from being a function both of the map distances involved and the distribution of crossovers with respect to each other (STAHL and

HOUSWORTH 2009).

44

Reanalyzing the data: We test our view that deletion of MLH1 reduces, but does not eliminate, interference using the two-factor data of ABDULLAH et al. (2004) and

ARGUESO et al. (2004). For a set of tetrads that contains at least one NPD, we assess the magnitude of the deviation of the data set from the hypothesis of no interference by determining the counting number (m; FOSS et al. 1993; STAHL and LANDE 1995).

Although the counting model was written for integer values of m, the calculator at Stahl

Lab Online Tools (http://www.molbio.uoregon.edu/~fstahl/tetrad.html) allows estimates of non-integer values of m by visual interpolation. If we do this for the wild type, mlh1 and msh4/5 data of Abdullah and of Argueso, we expect to see that m goes down as crossing over goes down, reaching zero for msh4/5. This analysis is tabulated in Table A5 along with p values for the null hypothesis of no interference.

The m values for mlh1 are between the wild type and msh4 values everywhere but once (Table A5), where interference was very low in the wild type. We conclude that mlh1 has interference that is intermediate between that of wild type and msh4, as expected for a strain that is missing a fraction of its disjunction pathway crossovers (and for which the linkage-map distances between the remaining disjunction pathway exchanges may have become more variable). Data by WANG et al. (1999) are fully compatible with this conclusion (Table A6). For two intervals and the inclusive interval, the m values for both mlh1 and deletion strains are less then those for wild type, indicating reduced interference. The three p values for wild type are all less than 0.05, while the mlh1 and mlh3 p values for one interval (in which wild-type interference is weak) are greater than 0.05, while for the other interval and the inclusive interval they

45

both are less than 0.05. Thus, interference, although weakened, is abolished in neither the mlh1 nor mlh3 mutants.

This analysis implicates Mlh3 as the partner to Mlh1 in promoting crossing over in the disjunction pathway and raises the question of which MutL homolog cooperates with Mlh1 in disjunction-pathway MMR. Mlh3 would surely be the prime suspect were it not for data that failed to demonstrate an MMR phenotype for mlh3 mutants (WANG et al. 1999). However, in order for a test marker to manifest an Mlh1-dependent MMR phenotype, the marker must participate in conversion in the disjunction pathway. As pointed out in Appendix I, BIK appears to be a marker that, while participating in conversion in the pairing pathway, fails to do so in the disjunction pathway, perhaps because it is rarely included in the short intermediate. Reported failures (e.g. WANG et al.

1999) to see a meiotic MMR phenotype for mlh3 may simply mean that the few markers monitored for meiotic MMR in those crosses happened to be of that sort.

Appendix III

Deletion of MLH1 changes some one-sided crossovers into one-sided noncrossovers

Franklin W. Stahl

HOFFMANN et al. (2005) scored tetrads from wild-type, msh2 and mlh1 crosses for one-vs. two-sidedness for conversion at sites B/b and C/c and for being crossed over (or not) with respect to linked sites A/a and D/d (their Table 7). They then compared each of the mutants to wild type and to each other with respect to the distribution of events

46

among the four classes scored. They wrote, “Both MMR mutant strains showed a difference in the distribution of events into those four classes compared to the wild-type strain (P < 0.05; G-test of homogeneity), reflecting that the mlh1Δ and msh2Δ strains contain more one-sided events. When we compared the distribution of mlh1Δ to that of the msh2Δ strain, we did not observe a significant difference.”

Had HOFFMANN et al. (2005) thought to compare the mutants with respect to crossover vs. noncrossover frequencies for the one-sided tetrads only, they would have found a significant difference in crossover/noncrossover ratios between the mlh1 and the MLH1 msh2 strains (Table A7). Among the one-sided tetrads, the loss of Mlh1 resulted in an mlh1-specific loss of one-sided crossvers accompanied by a gain of twice as many noncrossovers. Since the loss of Mlh1 results in a major failure of MMR, these data imply that, among disjunction-pathway one-sided tetrads, the mlh1- induced increase in HCs represents twice the number of FCs lost. The larger data set in

Appendix I supports this suggestion. We propose that the two-fold mutation-induced excess of HCs gained over FCs lost reflects the existence of disjunction-pathway-DSBr intermediates whose mismatches, in the presence of Mlh1, would have been rectified equally to either 2:2 (restoration) or FC of the his marker. Furthermore, MMR resulting in restoration must have been directed by the junction because, as posited in Appendix

I, the invasion and annealing opportunities for MMR would have yielded only FCs for a marker so close to the DSB. The further implication of this interpretation, that disjunction-pathway mismatches usually persist throughout the formation of the ligated double Holliday junction intermediate, is supported by the observations of

47

ALLERS and LICHTEN (2001) and is consistent with our hypothesis that disjunction- pathway mismatches are repaired only and inevitably via Mlh1-dependent, resolution- directed MMr (and see GETZ et al. 2008; STAHL and FOSS 2008).

Appendix IV

Disparity and one-sidedness

Franklin W. Stahl

In the total data of HOFFMANN et al. (2005) there are 182 conversions to HIS and

252 to his, a clear indication of disparity (p = 0.0009). In the MMR-defective crosses, 3:5

(HIS/his) tetrads were about twice as frequent as 5:3 (HIS/his) tetrads, while the 2:6 and

6:2 tetrads (at HIS) showed a statistically insignificant disparity in the same direction

(Table A8). HOFFMANN et al. (2005) reported that the wild-type cross showed no significant disparity among HCs, FCs or total conversions. To account for their observations, HOFFMANN et al. (2005) proposed that the G:G HIS/his mismatches that would give 5:3 segregation in the absence of Msh2 or Mlh1 be subject to unbiased short- patch repair rather than terminus directed MMR (RADFORD et al. 2007; but see COÏC et al.

2000, who reported that short-patch repair in yeast favored FCs). We note, however, that the wild-type HC data of HOFFMANN et al. (2005) are not significantly different from either of the two mutant data sets or from the sum of those two sets. Similarly, for none of the crosses are the HC data significantly more disparate than the FC data. More troubling for the unbiased short- patch-repair explanation for the disparity is the observation that the bias that is shown by the FCs, albeit insignificant, is in the direction opposite to that predicted. In their model, selective removals of 5:3 HIS/his mismatches,

48

while generating disparity in the HCs, would result in a disparity in the FCs which was half as great, but in the opposite direction, i.e., favoring 6:2 HIS/his. Since the disparities in the FCs and HCs are in the same direction, a differential frequency of DSBs on the two homologs has to be the favored explanation.

The simple explanation of differential DSBs is further supported by the disparity observed at BIK. If the disparity in HCs at his were due to mismatch-specific restoration of incipient 5:3 tetrads, there would be no expectation that unselected BIK conversions would show a related disparity. However, the unselected “subset” BIK conversions do manifest disparity [20 (2:6 + 3:5) and 9 (6:2 + 5:3)] which is significant (p = 0.03) and in the direction expected if the disparities at the two loci both resulted from the homologs being unequally subject to DSBs.

What might account for a differential frequency of DSBs on the two homologs?

Since sporulations were conducted with little or no growth following mating, it is plausible that the HIS4 hotspots on the two chromosomes were in different states with respect to Spo11 sensitivity (ABDULLAH and BORTS 2001; COTTON et al. 2009). HOFFMANN et al. (2005) physically measured DSBs at HIS4 and found no differences between the homologs. However, the physical studies were conducted on an established HIS/his diploid culture in which the homologs could not be expected to suffer a physiology- dependent difference.

We conclude that the work of HOFFMANN et al. (2005) provides no support for their thesis that one-sidedness reflects restoration by short-patch repair in MMR- defective crosses. At the same time, their work contradicts a proposal by GETZ et al.

49

(2008) that one-sidedness is caused by resolution-directed MMR in the disjunction pathway, effected by an unidentified MMR enzyme (as imagined in Figure 1D).

FIGURE LEGENDS

FIGURE 1. Classic DSBR. The model of SZOSTAK et al. (1983) as modified by SUN et al.

(1991) related double-strand-break repair, crossing over, gene conversion and mismatch repair in a well defined series of steps with the following features: (A) A meiosis- specific double-strand break is followed by resection of the 5’ ends created. (1) With the aid of RecA-like proteins, one of the two resulting 3’-ended single strands invades an intact homolog, displacing the complementary strand into a D-loop. (2) The invading 3’ end, acting as a primer, uses the homolog as a template to replace DNA lost by the resection. This synthesis further displaces the complementary strand, enlarging the D- loop, which then anneals with the other 3’-ended single-stranded DNA. (Here, and in subsequent figures, the DSB site is indicated by a black vertical line; the invasion step is shown to the left of the line, while the annealing step is to the right.) (3) Once annealed, this 3’ end, too, primes DNA synthesis across the break site creating a joint molecule that contains two duplexes worth of DNA with the break site now bracketed by two regions of heteroduplex DNA. (4) Ligation completes the classic, double-Holliday- junction structure, which yields crossovers for markers bracketing the DSBr event when the two junctions are resolved in the opposite sense – the Crick strands being cut at one junction and the Watson strands at the other. If the junctions are resolved in the same sense, noncrossovers result. In the absence of mismatch repair (MMR), any marked site between the two Holliday junctions will result in a half-conversion (HC) tetrad. (B)

50

When a 3’ end is degraded past the site of a marker, strand extension following invasion or annealing (as shown here) generates a full conversion (FC) tetrad by “gap repair”. (C) A FC tetrad can result from rectification of the mismatch by MMR during the invasion or, as shown here, the annealing step. (DNA removed and replaced by

MMR is shown speckled.) (D) Mismatches that survive into the completed double

Holliday structure can by rectified by MMR that is directed by the nicks created by resolution of a Holliday junction. Depending on which strands are cut, a mismatch may yield a FC (shown here on the invasion side) or a restoration (shown here on the annealing side).

FIGURE 2. The traveling D-loop. (1) A D-loop is created by invasion. (2) The D-loop moves towards the DSB site as the invading strand elongates (bricks of blue), reducing the region of heteroduplex to the left of the DSB. (3, 3’) Annealing brings migration of the junction to a halt early (3) or later (3’). See Figures 3 and 4 for the steps specific to the proposed disjunction and pairing pathways, respectively.

FIGURE 3. Proposed features of the disjunction pathway of DSBr in MLH1 meiosis. (1)

The 3’ strand invades the homolog (blue) creating a D-loop (HUNTER and KLECKNER

2004). (2, 3) Strand extension, shown as bricks of blue, often moves the trailing edge of the D-loop across the DSB site, undoing any mismatch that might have been created by invasion. D-loop migration is then blocked by annealing. The absence of MMR at invasion and annealing is characteristic of wild-type DSBr in the disjunction pathway.

51

(4, 5) Formation of the right-hand junction, promoted by Msh4-Msh5 binding, limits the length of the four-stranded structure, often to a region shorter than the resection. (6, 6’)

Junction resolution by cutting gives crossover products that are 4:4 on the invasion side and either FC or 4:4 on the annealing side, with the stretch of DNA that has lost information by MMR shown speckled. The arrows identify the junctions that are cut

“vertically” and “horizontally.” “Vert. first” and “Hori. first” indicates whether the vertical or the horizontal cut was made first, respectively. When resolution is by 6’, the

DNA synthesis is on different product duplexes, as reported by TERASAWA et al. (2007).

(In an mlh1 mutant, the resulting products would be noncrossovers that are 4:4 on the invasion side and HC on the annealing side.)

FIGURE 4. Proposed features of the pairing pathway of DSBr in msh2 and mlh1 meioses.

A. Noncrossovers: (1) A D-loop is initiated by strand invasion on one side of the DSB.

(2, 3, 3’) Under the impetus of strand extension, the D-loop migrates, causing the region of heteroduplex arising from invasion to be reduced in extent (3) or even eliminated (3’) before annealing brings the D-loop migration to a stop. (4) DNA synthesis proceeds leftward as far as the junction. (4’, 5) Withdrawal and reannealing (“SDSA” of PAQUES and HABER 1999) reconstitute the duplexes. (5, 5’) In the absence of gap repair, the tetrads are HC on both sides or 4:4 on the invasion side and HC on the annealing side.

In the event of gap-repair (or of MMr in wild-type cells), the invading side and/or the annealing side would be FC. B. Crossovers: Through steps 3 and 3’, events are as for noncrossovers. (4, 4’) Strand extension continues on the annealing side, which expands

52

the D-loop. (5, 5’) Strand extension is followed by trimming as necessary. (6, 6’; 7, 7’)

Following a proposal by CROMIE and SMITH (2007), the intermediates are drawn without a pair of Holliday junctions and are presumed to be resolved by Mus81-Mms4 as pictured in Figure 3A,B of HOLLINGSWORTH and BRILL (2004). (7, 7’) In the absence of gapping, the resulting crossovers, like the noncrossovers, are HC on both sides or 4:4 on the invasion side and HC on the annealing side of the DSB. In the event of gap repair

(or of MMR in wild-type cells), the invading side and/or the annealing side would be

FC. The role of extended synthesis in the production of crossovers derives from

MALOISEL et al. (2004), who proposed “ …a model inwhich DNA synthesis determines the length of strand exchange intermediates and influences their resolution toward crossing over.” Unbridled creation and extension of structures 4 and 4’, accompanied by that creates invasive 3’ ends could lead to the accumulation of multi- molecular structures in meioses lacking the endonuclease Mus81-Mms4 and the helicase Sgs1 (OH et al. 2008; JESSOP and LICHTEN 2008).

FIGURE 5. The JM2 DSBr intermediate of ALLERS and LICHTEN (2001). Steps 1 and 2 are as in Figure 3. (3, 4) If the traveling D-loop overshoots the marker to the right of the DSB site before being halted by annealing, the marker is in heteroduplex composed of an old red strand and a new blue strand. (5, 6) Prior to the completion of the Holliday junction on the right, the unligated junction, posing as a nicked junction, directs Mlh1- dependent MMR of the mismatch, restoring the parental state (speckled) of this disjunction pathway intermediate. SCHWACHA and KLECKNER (1995) reported a high

53

frequency of JM2s (their Type IIA). Prediction: double-Holliday-junction intermediates isolated from a strain lacking Mlh1would lack JM2s.

Figure A1. Crossovers that are trans-HC on the same chromatid. Steps 1-4 are as steps 1-

4 in Figure 4. (4, 5) An intermediate in the pairing pathway has enjoyed D-loop expansion but unwinds and re-anneals. (6) The 3’ whisker arising by branch migration on the resulting noncrossover product may promote crossing over with the blue chromatid or a third chromatid.

54 Stahl and Foss, Table 1

TABLE 1

Proposed properties of two DSBr pathways

Features Pairing pathway Disjunction pathway

Products Crossovers and noncrossovers Crossovers only

Crossover Interference No positive interference Positive interference

Msh4-Msh5 dependence None Total

Bimolecular intermediate Long with junctions not fully ligated Short with fully ligated Holliday junctions

Invasion heteroduplex Partly ephemeral Ephemeral

MMR at invasion and annealing Dependent on Msh2 and Mlh1 None

MMR near the DSB site: Directed by 3’ invading and annealing ends Mlh1-dependent; directed by junction resolution

Role of Msh2 in MMR Recognizes mismatches and attracts Mlh1 None

Role of Msh4-Msh5 in MMR None Attracts Mlh1

p1, Stahl and Foss Table 2

TABLE 2

Summary of Sudoku

A. Values for adjustable parameters at his

Parameter Value a

D, disjunction pathway events per 1000 154

tetrads

P, pairing pathway events per 1000 171

tetrads

E, fraction of P retaining heteroduplex on 0.3

invasion side

R, probability that heteroduplex rejection 0.643

does not occur

ghis, Probability of gap repair at his 0.127

gBIK, Probability of gap repair at BIK 0.04

mhis, Probability of mismatch rectification 0.726

to FC at his in pairing pathway

mBIK, Probability of mismatch rectification 0.96

to FC at BIK in pairing pathway

B. Conversion at his

Class Obs.b Calc.c

mlh1 HC 171 174

mlh1 FC 21 21.8

msh2 HC 97 97

msh2 FC 61 60.3

WT HC 17 17.1

WT FC 120 122

C. Fraction of two-sided his conversions

Cross Obs.d Calc.e

p2, Stahl and Foss Table 2 mlh1 0.26-0.47 0.31

msh2 0.22-0.46 0.39

WT 0.58-0.77 0.71

D. FC/(FC+HC) for BIK among his conversions

Cross Obs.f Calc.e

mlh1 4/41 0.10

msh2 3/51 0.10

WT 89/92 0.97

a From Tables A2, A3 and A4

b From Table A1

c From Tables A2 and A3

d 95% confidence interval on data from Table 5 of HOFFMANN et al. (2005)

e From Table A4

f From Table 5 of HOFFMANN et al. (2005)

Stahl and Foss Table 3 1

TABLE 3

Mutant phenotypes of MutS and MutL homologs

A. Effects on crossing over and interference

Genotype Meiotic phenotype Explanation

msh4 Reduced crossing over, interference eliminateda Disjunction pathway eliminatedb

msh2 None NA

mlh1 Reduced crossing over and interferencec Turns disjunction pathway COs into NCOs

msh2 msh4 Like msh4d Msh2 does not affect crossing over or interference.

mlh1 msh4 Like msh4e Mlh1 does not affect crossing over or interference in the

absence of the disjunction pathway.

B. Effects on MMR for well repairable mismatch near a DSB site

Conventional wisdom Model

Genotype Phenotype Explanation Predicted/explained Explanation

phenotype

msh4 None obviousb,d,f,g Msh4 not a MMR protein Increased HC/(HC+FC) Loss of disjunction pathway, in

which all conversions are FCs Stahl and Foss Table 3 2

msh2 Increased HC/(HC+FC) h Reduces meiotic MMR Increased HC/(HC+FC) Eliminates MMR in the pairing

for NCOs and pathway only, resulting in both

noninterfering COs more HCs and fewer

conversions

mlh1 Increase in HC/(HC+FC) Removal of Mlh1 reduces Increase in Eliminates MMR in both

greater than for msh2h MMR more efficiently than HC/(HC+FC) greater pathways.

does removal of Msh2 than for msh2h

msh2 msh4 Like msh2d Msh4 not a MMR protein. Sum of msh2 and msh4 The mutants affect distinct

phenotypes pathways.

mlh1 msh4 Not known NA Like msh2 msh4 In the absence of the

disjunction pathway, mlh1 has a

phenotype only in the pairing

pathway.

C. Effects on MMR for poorly repairable mismatch near a DSB site Stahl and Foss Table 3 3

Wisdom according to GETZ et al. (2008) Model

Genotype Phenotype Explanation Predicted/explained Explanation

phenotype

msh4 Increased HC/(HC+FC) Eliminates disjunction Increased HC/(HC+FC) Loss of disjunction pathway, in

pathway which all conversions are FCs

msh2 Not known NA No effecti PRMs not repaired by Msh2j

mlh1 Not known NA Increased HC/(HC+FC) Turns disjunction pathway FC

for NCOs and 4:4 COs into HC NCOs

.

msh2 msh4 Not known NA Like msh4 Eliminates disjunction pathway.

mlh1 msh4 Not known NA Like msh2 msh4 = like PRMs not repaired in pairing

msh4 pathway; Disjunction pathway is

gone.

Abbreviations: DSBr, double-strand-break repair; MMr, mismatch rectification; MMR, mismatch repair; COs, crossovers; NCOs, noncrossovers;

HC, half conversion; FC, full conversion; NA, not applicable a NOVAK et al. (2001) Stahl and Foss Table 3 4

b GETZ et al. (2008) c ABDULLAH et al. (2004) d STONE and PETES (2006) e HUNTER and BORTS (1997) f ROSS-MACDONALD and ROEDER (1994) g HOLLINGSWORTH et al. (1995) h HOFFMANN et al. (2005) i If PRMs escape heteroduplex rejection in wild type j WANG et al. (2003)

Stahl and Foss Table A1

TABLE A1

Conversions at hisa

Conversion type per thousand tetradsb Tetrads (FC + HC)% HC/(FC + HC)%

6:2 2:6 FCc 5:3 3:5 HCd

WT 56 64 120 8 9 17 1731 13.7 12.4 msh2 28 33 61 31 66 97 545 15.8 61.4 mlh1 9 12 21 60 111 171 585 19.2 89.1

Adapted from Table 5 of HOFFMANN et al. (2005) a Conversions of the marker his4-ATC (“his”), close to the DSB site. b Ignores a few rare tetrad classes. Conversion types are rounded to the nearer whole number. c 6:2 + 2:6 d 5:3 + 3:5 1 Stahl and Foss Table A2,

TABLE A2

Work sheet for confirming parameter values for his in MMR mutants

Pairing pathway: P = 171a Disjunction pathway: D = 154b

Invasion: 1/2 Annealing: 1/2 Invasion: 1/2 Annealing: 1/2

FC HC FC HC FC HC FC HC msh2 g E(1-g) g 1-g 0 0 1/2 0

0.127 0.262 0.127 0.873

N 10.9 22.4 10.9 74.6   38.5 

Summed conversions Calculated Observed

FC 10.9 + 10.9 + 38.5 = 60.3 61

HC 22.4 + 74.6 = 97 97 mlh1 g E(1-g) g 1-g 0 0 0 1

0.127 0.262 0.127 0.873

N 10.9 22.4 10.9 74.6    77

Summed conversions Calculated Observed

FC 10.9 + 10.9 = 21.8 21

HC 22.4 + 74.6 + 77 = 174 171

a Estimated number of tetrads per thousand that enjoyed a DSB at HIS4 and were repaired on a homolog via the pairing pathway.

b Estimated number of tetrads per thousand that enjoyed a DSB at HIS4 and were repaired on a homolog via the disjunction pathway.

2 Stahl and Foss Table A2, g = 0.127 is the probability of gap repair (or short-patch repair to FC in the pairing pathway). E = 0.3 (see

Table A4) is the probability that the marker (BIK or his) on the invasion side of the DSB remains within the heteroduplex. N is the number, per thousand tetrads, expected by the model evaluated with these parameter values. The observed numbers are from Table A1. Assigning the same probability, 1/2, to the invasion and annealing sides of the DSB implies an assumed lack of left vs. right sequence preference in the invasion events.

Stahl and Foss Table A3

TABLE A3

Work sheet for confirming parameter values for his in wild type

Pairing Pathway: RP = 0.643 x 171 Disjunction Pathway: D = 154

Invasion: 1/2 Annealing: 1/2 Invasion: 1/2 Annealing: 1/2

FC HC FC HC FC HC FC HC

g+m(1-g) E(1-g)(1-m) g+m(1-g) (1-g)(1-m) 0 0 1/2 0

0.760 0.0718 0.760 0.239

N. 41.8 3.9 41.8 13.2 0 0 38.5 0

Summed conversions Calculated Observed

FC 41.8 + 41.8 + 38.5 = 122 120

HC 3.9 + 13.2 = 17.1 17

P, D, g and E values are as in Table A2. R = 0.643 and m = 0.726 were determined by fitting the model to the observed HC and FC numbers (per thousand tetrads, Table A1) for this wild type cross. Stahl and Foss Table A4 Page 1

TABLE A4

Work sheet for testing parameter values for BIK and his with regard to two-sidedness and conversion types for BIK

A. MMR-defective pathways

Pairing pathway: P = 171 Disjunction pathway: D =154

Invasion: 1/2 Annealing: 1/2 Invasion: 1/2 Annealing: 1/2

FC HC FC HC FC HC FC HC

msh2 g E(1-g) g 1-g 0 0 1/2 0

BIK: 0.04 (3.4) BIK: 0.288 (24.6) BIK: 0.04 (3.4) BIK: 0.96 (82.1) BIK: 0.5 (38.5)

his: 0.127 (10.9) his: 0.262 (22.4) his: 0.127 (10.9) his: 0.873 (74.6) his: 0.5 (38.5)

  

BIK: 0.328 (28.0) BIK: 1 (85.5) BIK: 0.5 (38.5)

: 0.389 (33.3) his: 1 (85.5) his: 0.5 (38.5)

Properties of BIK conversions Calculated Observed

Freq. 2-sided tetrads among his conversions (28.0 + 33.3)/(33.3 + 85.5 + 38.5) = 0.39 20/61 = 0.22-0.46a

b FC/(FC + HC) BIK among his conversions (3.4/28)(1/1.389) + (3.4/85.5)(0.389/1.389) = 0.10 4/41

b FC/(FC + HC} BIK in unselected tetrads (6.8 + 38.5)/(113.5 + 38.5) = 0.30 0/9

b (FC + HC) BIK /(FC + HC) his (113.5 + 38.5)/(118.8 + 38.5) = 0.97 (9/101)/(96/545) = 0.51 ± 0.33

mlh1 g E(1-g) g 1-g 0 0 0 1

BIK: 0.04 (3.4) BIK: 0.288 (24.6) BIK: 0.04 (3.4) BIK: 0.96 (82.1) BIK: 0.5 (77)

his: 0.127 (10.9) his: 0.262 (22.4) his: 0.127 (10.9) his: 0.873 (74.6) his: 0.5 (77) Stahl and Foss Table A4 Page 2    

BIK: 0.328 (28.0) BIK: 1 (85.5) BIK: 0 BIK: 0.5 (77)

: 0.389 (33.3) his: 1 (85.5) his: 0 his: 0.5 (77)

Properties of BIK conversions Calculated Observed

Freq. 2-sided tetrads among his conversions (28.0 + 33.3)/(33.3 + 85.5 + 77) = 0.31 29/81 = 0.26-0.47a

b FC/(FC + HC) BIK among his conversions As in msh2 = 0.10 3/51

b FC/(FC + HC} BIK in unselected tetrads (6.8)/(113.5 + 77) = 0.04 1/11

b (FC + HC) BIK /(FC + HC) his (113.5 + 77)/(118.8 + 77) = 0.97 (14/106)/(116/585) = 0.67 ± 0.34

B. WT Pairing pathway: RP = 0.643 x 171 Disjunction pathway: D = 154

Invasion: 1/2 Annealing: 1/2 Invasion: 1/2 Annealing: 1/2

FC HC FC HC FC HC FC HC

g+m(1-g) E(1-g)(1-m) g+m(1-g) (1-g)(1-m) 0 0 1/2 0

BIK: 0.962 (52.9) BIK: 0.0115 (0.6) BIK: 0.962 (52.9) BIK: 0.0384 (2.1) BIK: 0.5 (38.5)

his: 0.761 (41.8) his: 0.0718 (3.9) his: 0.761 (41.8) his: 0.239 (13.2) his: 0.5 (38.5)

  

BIK: 0.973 (53.5) BIK: 1.0 (55.0) BIK: 0.5 (38.5)

his: 0.833 (45.8) his: 1.0 (55.0) his: 0.5 (38.5)

Properties of BIK conversions Calculated Observed Stahl and Foss Table A4 Page 3 Freq. 2-sided tetrads among his conversions (53.5 + 45.7)/(45.7 + 55.0 + 38.5) = 0.71 61/90 = 0.57-0.77a

b FC/(FC + HC) BIK among his conversions 1 - (0.6/53.5)(1/1.831) - (2.1/55.0)(0.831/1.831) = 0.97 89/92 = 0.97

b HC/(FC + HC} BIK in unselected tetrads (2.7)/(108.5 + 38.5) = 0.02 0/9

b (FC + HC) BIK /(FC + HC) his (108.5 + 38.5)/(100.7 + 38.5) = 1.06 (10/107)/(243/1731) = 0.67 ± 0.40

Values for P, D, R and E, as well as for mhis and ghis, are as in Tables A2 and A3. E = 0.3 and gBIK = 0.04 were picked to fit the (FC + HC) and the FC/(FC + HC) data for BIK among his conversions. m BIK = 0.96 was then selected to fit the WT sidedness data. Expected numbers (N) are in parentheses after each expected conversion frequency. a Table 6 of HOFFMANN et al. (2005). b Table 5 of HOFFMANN et al. (2005). Stahl and Foss Table A5 1

TABLE A5

Map distances and interference

Chromosome XV (ARGUESO et al. 2004)

ADE-HIS URA-LYS LYS-HIS URA-LEU

wt mlh1 msh5 wt mlh1 msh5 wt mlh1 msh5 wt mlh1 msh5

PD 343 400 496 264 351 513 278 344 465 607 486 643

T 709 211 215 759 261 300 744 261 242 456 128 76

NPD 16 5 9 45 4 7 48 11 13 5 2 1

cM 37.7 19.6 18.7 48.2 23.1 16.8 47.8 26.5 22.2 22.8 11.4 5.7

m 4.5 0.8 0.0 3.1 1.9 0.9 2.7 0.7 0.0 2.1 0.5 0.0

P <0.0001 0.06 0.73 <0.0001 0.0006 0.01 <0.0001 0.06 0.91 <0.0001 0.35 0.94

Chromosome III (ABDULLAH et al. 2004)

LEU-MAT HIS-LEU

wt mlh1 msh4 msh5 wt mlh1 msh4 msh5

PD 595 570 213 153 744 583 193 121

T 722 370 62 46 496 191 39 16

NPD 51 14 3 1 20 5 1 1

cM 37.6 22.0 14.4 13.0 24.4 14.2 9.7 8.0

m 0.6 0.6 < 0 0.3 0.6 0.2 0.0 <0

p 0.0044 0.042 0.55 0.66 0.025 0.46 0.94 0.19

1 Stahl and Foss Table A5 2

Chromosome VII (ABDULLAH et al. 2004)

TRP-CYH CYH-MET

wt mlh1 msh4 msh5 wt mlh1 msh4 msh5

316 487 413 141 1039 781 514 190

PD 1023 451 128 61 395 140 25 9

T 113 25 4 2 6 2 1 0

NPD

cM 58.6 31.2 13.9 17.9 15.0 8.2 2.9 2.3

m 2.0 0.6 0 0.3 0.9 0.1 <0 ----

p <0.0001 0.028 0.82 0.62 0.011 0.59 0.042 0.82

Map lengths (cM) for the intervals are calculated on the assumption that the number of exchanges per bivalent in any interval is not more than two (PERKINS 1949). m, is an index of the strength of interference (STAHL and LANDE 1995; STAHL and HOUSWORTH 2009); m =

0 implies no interference. p is the probability that such an observed deviation of the data from the hypothesis of no interference would occur by chance alone, calculated according to STAHL (2008). The map for chromosome XV is URA-LEU-LYS-ADE-HIS. The intervals on chromosomes III and VII are all non-overlapping.

2 Stahl and Foss Table A6

TABLE A6

Map distances and interference

Chromosome III (WANG et al. 1999)

URA-HISLEU HISLEU-MAT URA-MAT

wt mlh1 msh3 wt mlh1 msh3 wt mlh1 msh3

PD 483 564 697 506 603 689 258 391 450

T 483 345 370 451 301 388 646 502 585

NPD 7 8 6 24 12 17 72 37 47

cM 27.0 21.4 18.9 30.3 20.4 21.9 55.2 38.9 40.1

m 2.7 0.9 1.0 0.6 0.3 0.3 1.1 0.6 0.5

p <0.0001 0.004 0.0015 0.027 0.34 0.24 0.0004 0.016 0.027

See Table A5 for explanations Stahl and Foss Table A7

TABLE A7a

mlh1 phenotype in two- and one-sided tetrads

Two-sided tetrads One-sided tetrads Total tetrads

CO NCO pb CO NCO pb pc

WT 35 19 1 9 8 0.067 0.39 1731

mlh1 15 9 10 29 0.007 585

MLH1 msh2 9 6 0.86 17 15 0.027 0.76 545

Among one-sided tetrads, the mlh1 mutation causes the mlh1-specific loss of approximately 17-

10 = 7 crossovers (COs) coupled with the appearance of 29-15 = 14 noncrossovers (NCOs). The similarity in total population size of the mlh1 and msh2 data sets justifies this direct comparison of observed numbers. a Data from Table 7 of HOFFMANN et al. (2005) b Probability by two-tailed Fisher Exact Probability Test that these data and those for mlh1 could have been drawn randomly from the same universe. c Probability by two-tailed Fisher Exact Probability Test that the one-sided and two-sided data could have been drawn randomly from the same universe. Stahl and Foss Table A8

TABLE A8

Disparity at hisa

FC HC pb pc

6:2 2:6 5:3 3:5

wild type 96 111 14 15 1.0 --

mlh1 5 7 35 65 0.75 0.28

msh2 15 18 17 36 0.31 0.23

mlh1+msh2 20 25 52 101 0.27 0.21

Σ three crosses 116 136 66 116 0.053 --

pd 0.23 0.0003

a Data from Table 5 of HOFFMAN et al. (2005). b Chi2 probability that a difference in disparity between FCs and HCs that is this great or greater could arise by chance alone. c Chi2 probability that a difference in disparity between the mutant strain and WT that is this great or greater could arise by chance alone. d Chi2 probability for the data summed over the three crosses that an observed disparity this great or greater could arise by chance alone. Fig 1

A B C D

1

2

3

4

Figure 2

Noncrossovers 4:4 HC Figure 1 1 Noncrossovers 4:4 HC

3p 1 2 3p 2 1 3 4p 3’

3 4p 2 5p 5p

6p 1p 2p 3 Canonical intermediate 6p 1p 2p 7p

4 7p

8p

8p

4:4 HC 4:4 HC

Interfering crossovers 4:4 HC Noninterfering crossovers4:4 HC

Interfering crossovers Noninterfering crossovers

Figure 3

1

2

3

4

5 5’

6 6’ Vert. First 4:4 FC 4:4 FC Hori. First 4:4 4:4 4:4 4:4

Figure 4

Noncrossovers A 4:4 HC B Noncrossovers 1 1 4:4 HC

3p 2 2 3p

3 4p 3’ 3 3’ 4p

5p 4 4’5p 4 4’4’ 4’ 6p 1p 2p 6p 1p 2p 5 5’ 5’ 5’ 5 5’ 7p 7p 6 6’ 6’ 4:4 HC HC HC 8p 8p 7 7’ 7’ 4:4 HC 4:4 HC 4:4 HC 4:4 HC Interfering crossovers Noninterfering crossovers HC HC 4:4 HC Interfering crossovers Noninterfering crossovers

Figure 5

1

2

3

4

5

6

Figure A1

1

2

3

4

5

6

HC HC