Extensive Interallelic Polymorphisms Drive Meiotic Recombination into a Crossover Pathway

Abstract

Recombinants isolated from most meiotic intragenic recombination experiments in maize, but not in yeast, are borne principally on crossover chromosomes. This excess of crossovers is not explained readily by the canonical double-strand break repair model of recombination, proposed to account for a large body of yeast data, which predicts that crossovers (COs) and noncrossovers (NCOs) should be recovered equally. An attempt has been made here to identify general rules governing the recovery of the CO and NCO classes of intragenic recombinants in maize. Recombination was analyzed in bz heterozygotes between a variety of mutations derived from the same or different progenitor alleles. The mutations include point mutations, transposon insertions, and transposon excision footprints. Consequently, the differences between the bz heteroalleles ranged from just two nucleotides to many nucleotides, indels, and insertions. In this article, allelic pairs differing at only two positions are referred to as dimorphic to distinguish them from polymorphic pairs, which differ at multiple positions. The present study has revealed the following effects at these bz heteroalleles: (1) recombination between polymorphic heteroalleles produces mostly CO chromosomes; (2) recombination between dimorphic heteroalleles produces both CO and NCO chromosomes, in ratios apparently dependent on the nature of the heteroalleles; and (3) in dimorphic heterozygotes, the two NCO classes are recovered in approximately equal numbers when the two mutations are point mutations but not when one or both mutations are insertions. These observations are discussed in light of a recent version of the double-strand break repair model of recombination that postulates separate pathways for the formation of CO and NCO products.

INTRODUCTION

Recombination experiments in yeast and maize have revealed a striking difference between the two organisms in the fraction of meiotic intragenic recombinants (IGRs) that are associated with an exchange of flanking markers. Most IGRs in maize are borne on crossover chromosomes (for review, see Dooner and Martínez-Férez, 1997), whereas in yeast, less than half of them are (for review, see Petes et al., 1991). According to the widely accepted double-strand break repair (DSBR) model for meiotic recombination (Szostak et al., 1983; Sun et al., 1991), random resolution of the double Holliday junction (DHJ) that is predicted to be the recombination intermediate would result in an equal number of crossover (CO) and noncrossover (NCO) chromosomes. The strong preference for a crossover resolution of the recombination intermediate inferred from the maize data is not explained adequately by the DSBR model.

The maize heteroalleles used in recombination experiments generally have been unrelated in origin, and because transposon-induced mutations are common and easy to map molecularly, one of the two alleles in the heterozygote often has been an insertion mutation (Dooner, 1986; Dooner and Kermicle, 1986; Patterson et al., 1995; Richter et al., 1995; Xu et al., 1995). It is precisely in those structurally polymorphic heterozygotes that the discrepancy between CO and NCO IGRs is greatest. In fact, many experiments have yielded IGRs that belong exclusively to one flanking marker class, the CO class expected based on the physical placement of the mutations. Peculiarly, however, when two different Ds insertion heteroalleles were paired against each other at either the bz or r locus, a majority of parentally marked IGRs were recovered (Dooner and Kermicle, 1986), leading to the suggestion that the presence of two related insertions in the heterozygote could affect the outcome of the recombination event. This suggestion seemed to be supported by later observations that heterozygotes between two insertions unrelated in sequence—either a Ds and a Mu1 insertion at bz (Dooner and Ralston, 1990) or a Mu1 and an rDt insertion at a1 (Xu et al., 1995)—behaved like most other maize heteroallelic combinations in yielding almost exclusively CO IGRs.

Subsequent sequencing of the progenitor alleles of the bz (Ralston et al., 1988) and r (W. Eggleston, personal communication) mutations used in the earlier recombination experiments revealed extensive differences between them. The Bz-McC and Bz-W22 progenitor alleles differ at 21 positions across the 1521 bp between the start and stop codons. Of these, 20 are single-nucleotide polymorphisms (SNPs) and one is a simple sequence repeat polymorphism. These do not occur randomly within the bz gene but are concentrated in and around the intron and in the distal (3′) part of the second exon (Figure 1). Outside of the coding region, the differences are more extensive and include, in addition to SNPs, indels ranging from a few base pairs to several hundred base pairs. The recent finding that the density of SNPs appears to affect the distribution of recombination junctions within bz (Dooner and Martínez-Férez, 1997) raised the possibility that it also could affect the outcome of recombination events.

Figure 1.

Locations of Mutations and Heterologies in Different bz Alleles.

The locations of the mutations and heterologies in all of the bz alleles used in this study are shown in a composite map of the two progenitor wild-type alleles, Bz-W22 and Bz-McC. Mutations derived from Bz-W22 are shown above the line, and those derived from Bz-McC are shown below the line. Numbers refer to the published Bz-W22 genomic DNA sequence, with 1 corresponding to the translation start site (Ralston et al., 1988). The designations E2 to E9 and EMc1 refer to the mutations bz-E2 to bz-E9 and bz-EMc1 induced by EMS in the Bz-W22 and Bz-McC progenitor alleles, respectively. The nucleotide difference between a bz-E mutation and the contrasting wild-type site is shown immediately below (E2 to E9) or above (EMc1) each mutation. The Ds insertion mutations bz-m1 and bz-m2(DI) arose in the Bz-McC progenitor allele and are represented by triangles. The number inside each triangle refers to the size of the insertion in kilobases. The location of the 8-bp direct repeat generated by the insertion is indicated in the box above each insertion. The locations of the Ac excision footprint in the bz-s30, bz-s2.1, and bz-s2.2 mutations are indicated by similar boxes. The nucleotide polymorphisms between Bz-W22 and Bz-McC are shown, respectively, above and below the lines representing the two alleles. The single intron (105 bp in Bz-W22 and 100 bp in Bz-McC) is shown by stippling. Oligonucleotides used as primers in polymerase chain reaction and sequencing are identified by arrows that indicate their 5′ to 3′ strand polarity.

Here, the issue of flanking marker recovery among IGRs at the bz locus is reexamined. In an attempt to distinguish the potentially confounding effects of insertions and multiple sequence heterologies, intragenic recombination has been analyzed in heterozygotes between a variety of mutations derived from either the same or different progenitor alleles. The mutations include point mutations induced by ethyl methanesulfonate (EMS), transposon insertions, and transposon excision footprints. Consequently, the differences between the heteroalleles studied range from just two bases to extensive sequence and structural polymorphisms. Alleles that differ at only two positions will be referred to as dimorphic and those that differ at multiple positions will be referred to as polymorphic, regardless of the nature of the differences.

This comprehensive study has enabled the following general observations to be made. (1) Most recombination between extensively polymorphic alleles results in an exchange of flanking markers, regardless of whether one of the alleles is an insertion. (2) Recombination between dimorphic heteroalleles (i.e., derived from the same progenitor and differing only at the two points between which recombination is being measured) often is not accompanied by an exchange of flanking markers. (3) Unequal recovery of the two parentally marked classes can occur in dimorphic heterozygotes but seems to be a function of the nature of the mutation, not of the relative position of the mutation within the gene.

These observations are discussed in light of a recent version of the DSBR model of meiotic recombination that postulates the existence of independent pathways for the formation of CO and NCO recombinants (Allers and Lichten, 2001). This revision of the DSBR model was made necessary by substantial genetic (Porter et al., 1993; Gilbertson and Stahl, 1996) and physical (Allers and Lichten, 2001; Hunter and Kleckner, 2001) data in yeast that cannot be accommodated easily by the canonical model (Szostak et al., 1983; Sun et al., 1991).

RESULTS

Recombination between Polymorphic Heteroalleles

Recombination between insertion and point mutations derived from different progenitor alleles yields almost exclusively CO Bz IGRs. Table 1 summarizes both published and new data obtained from 14 different heterozygotes between one of eight bz-E alleles derived from Bz-W22 and one of two bz-m mutations derived from Bz-McC. The eight bz-E mutations are either missense or nonsense mutations of Bz-W22 (Dooner and Martínez-Férez, 1997). The insertions in the bz-m1 and bz-m2(D1) mutations are different Ds transposons, measuring 1.2 and 3.3 kb, respectively. The locations of all of these mutations are indicated in Figure 1.

Table 1.

Recombination between Pairs of Polymorphic bz Heteroalleles: One Allele Carries an Insertion

Genotype (P/D)a	Population					Total	Frequency (×10−3)	CO/NCO Ratio
sh bz-m1 wxb	114,653	0	0	7	0	7	0.06	–
Sh bz-E2 Wx
sh bz-m1 wxb	87,674	2	0	27	0	29	0.33	–
Sh bz-E3 Wx
sh bz-m1 wxb	95,778	3	1	45	0	49	0.51	–
Sh bz-E4 Wx
sh bz-m1 wxb	123,632	1	2	10	0	13	0.11	–
Sh bz-E5 Wx
sh bz-m1 wxb	58,147	0	0	27	0	27	0.46	–
Sh bz-E6 Wx
sh bz-m1 wxc	18,890	0	0	8	0	8	0.42	–
Sh bz-E7 Wx
sh bz-m1 wxc	18,830	0	0	7	0	7	0.37	–
Sh bz-E8 Wx
sh bz-m1 wxc	19,870	1	0	10	0	11	0.55	–
Sh bz-E9 Wx
Sh bz-E2 Wxb	217,823	0	0	6	0	6	0.03	–
sh bz-m2(Dl) wx
sh bz-m2(Dl) wxb	221,853	0	0	12	0	12	0.05	–
Sh bz-E3 Wx
sh bz-m2(Dl) wxb	182,043	0	0	28	0	28	0.15	–
Sh bz-E4 Wx
Sh bz-E5 Wxb	194,462	0	0	5	0	5	0.03	–
sh bz-m2(Dl) wx
sh bz-m2(Dl) wxb	33,149	0	0	9	0	9	0.27	–
Sh bz-E6 Wx
sh bz-m2(Dl) wx	46,040	0	0	3	0	3	0.07	–
Sh bz-E8 Wx
Total bz-m/bz-E	–	7	3	204	0	214	–	20.4

^aP, proximal bz heteroallele; D, distal bz heteroallele.

^bData from Dooner (1986).

^cData from Dooner and Martínez-Férez (1997).

The genotypes of all heterozygotes in Tables 1 through 4 are entered so that the proximal bz heteroallele (P) is on the top, the distal bz heteroallele (D) is on the bottom, the telomere end of 9S (sh) is to the left, and the centromere end (wx) is to the right. The four classes of Bz IGRs are represented schematically in the following order: NCO (PP), NCO (DD), CO (PD), and CO (DP), where left and right correspond to the telomeric and centromeric markers, respectively, and P and D refer to whether the marker was associated with the proximal or distal heteroallele in the heterozygous parent. As can be seen from Table 1, only one CO class was recovered, that expected from the location of the mutations in the bz gene. Furthermore, this class occurs in a >20-fold excess over the two NCO classes combined. Similar data have been obtained at the r (Dooner and Kermicle, 1986), a1 (Xu et al., 1995), and B (Patterson et al., 1995) loci in maize.

Table 4.

Recombination between Pairs of Dimorphic bz Heteroalleles: One or Both Alleles Carries an Insertion

Genotype (P/D)	Population	Insertion Sizes (kb)					Total	Frequency (×10−3)	CO/NCO Ratio
bz-m/bz-m heterozygotes
sh bz-m1 Wxa	210,914	1.1/3.3	23	7	5	2	37	0.18	0.16
Sh bz-m2(Dl) wx
sh bz-m1 Wxa	224,012	1.1/3.8	16	3	10	2	31	0.14	0.48
Sh bz-m2(D2) wx
bz-m/bz-s heterozygotes
sh bz-m1 wx	545,780	1.1/0.008	110	165	61	49	385	0.71	0.19
Sh bz-s2.1 Wx
sh bz-m1 wx	220,800	1.1/0.008	30	54	17	14	115	0.52	0.17
Sh bz-s2.2 Wx

^aData from Dooner and Kermicle (1986).

The reciprocal CO class would have been expected to occur from either a triple crossover event (one at bz plus one in each flanking interval) or from conversion of the distal allele accompanied by exchange in the 25-centimorgan bz-wx interval. Double crossovers in the sh-bz-wx region are rare because of high chiasma interference in the region (Dooner, 1986), and, as shown in yeast, gene conversion does not interfere with an exchange in an adjacent interval (Petes et al., 1991). Therefore, the minority CO class, when it occurs, is assumed to have arisen from conversion plus a coincidental exchange and is pooled with the two NCO classes in calculations of the CO/NCO ratio.

Because of their different origins, the heteroallelic pairs in Table 1 differ not only by the presence versus the absence of a large insertion but also by multiple SNPs across the gene (Figure 1). This also was the situation in the other maize intragenic recombination studies cited above, suggesting that the observations from Table 1 are not specific to the pairs of bz heteroalleles being examined. The great excess of CO IGRs seen frequently in maize, but not in yeast, could be attributable to the presence of a large transposon insertion in one of the alleles, the high density of SNPs between the alleles, or a combination of these two factors. To attempt to distinguish the contributions of these factors, recombination between highly polymorphic heteroalleles was examined in the absence of insertion heterozygosity.

Mutant derivatives of Bz-McC with lesions either at or close to the bz-m1 and bz-m2(D1) insertion sites, but lacking the insertions, were isolated as transposon excision products from different bz-m alleles. Ac excision in maize usually produces stable null alleles because of the target site duplication footprint that is left behind at the previous insertion site. Several stable alleles have been produced by the excision of Ac from bz-m2(Ac), the progenitor of the Ds insertion mutation bz-m2(D1) (Dooner and Belachew, 1989). Two of these, bz-s2.1 and bz-s2.2, were sequenced and confirmed to have an 8-bp footprint at positions 755 to 762, which is the exact location of the Ds insertion in bz-m2(D1) (see Methods).

A somewhat different strategy was adopted to generate a bz-s allele with a lesion close to the Ds insertion site in bz-m1. Because bz-m1 arose by transposition of Ds into Bz-McC (McClintock, 1951; Martínez-Férez and Dooner, 1997) and not by internal deletion of an Ac element at bz, it lacks an autonomous progenitor allele with an Ac element at the same location as Ds (positions 161 to 168). This complicates the isolation of nonfunctional excision products from bz-m1: in the presence of a single dose of Ac to induce Ds excision, a condition imposed by Ac's strong negative dosage effect, most stable bronze seed would originate by segregation or loss of Ac, rather than by Ds excision. Therefore, bz-s derivatives were isolated from bz-m30(Ac), a new mutable allele with an Ac insertion close to the Ds insertion site of bz-m1. bz-m30(Ac) is a second-cycle derivative produced by transposition of Ac out of bz-m2(Ac) and back into the bz locus. It carries Ac at position 142-149, just 19 bp upstream of the site of insertion of Ds in bz-m1. From this allele, the stable mutant bz-s30.1 was isolated, which carries an 8-bp footprint at positions 142 to 149 (see Methods).

These new alleles enabled the study of recombination between two noninsertion heteroalleles of different origin, a bz-E heteroallele from Bz-W22 and a bz-s heteroallele from Bz-McC. Results for the heteroallelic pairs bz-s30.1/bz-E6, bz-s30.1/bz-E8, bz-s2.1/bz-E6, and bz-s2.1/bz-E8 are presented in Table 2. It is clear from the data that removal of the large insertion does not affect the distribution of flanking markers among Bz IGRs. In fact, the CO/NCO ratio of the pooled data is essentially identical to that seen among Bz IGRs from heterozygotes in which one of the heteroalleles carried a large insertion (20.8 versus 20.4; χ² = 0, 1 df, P ∼ 1). The most reasonable interpretation of these data is that the large excess of COs among IGRs is not caused by the large insertion but by the multiple heterologies that distinguish the two progenitor alleles.

Table 2.

Recombination between Pairs of Polymorphic bz Heteroalleles: Neither Allele Carries an Insertion

Genotype (P/D)	Population					Total	Frequency (×10−3)	CO/NCO Ratio
bz-s/bz-E heterozygotes
sh bz-s30.1 Wx	105,860	2	1	61	0	64	0.60	–
Sh bz-E6 wx
sh bz-s30.1 Wx	61,383	2	0	21	0	23	0.37	–
Sh bz-E8 wx
Sh bz-s:2.1 wx	210,339	0	0	25	1	26	0.12	–
sh bz-E6 Wx
Sh bz-s:2.1 wx	214,457	0	0	18	0	18	0.08	–
sh bz-E8 Wx
Total bz-s/bz-E	–	4	1	125	1	131	–	20.8
bz-EMc/bz-E heterozygotes
Sh bz-EMc1 wx	70,589	0	0	14	0	14	0.20	–
sh bz-E6 Wx

A possible reservation to this conclusion is that neither bz-s heteroallele is strictly a point mutation; rather, they carry a transposon footprint that amounts to a very small insertion. Therefore, a true point mutation of Bz-McC was generated by EMS mutagenesis (see Methods and Figure 1) and paired with a point mutation from Bz-W22. The last entry in Table 2 shows the results obtained from the corresponding bz-EMc1/bz-E6 heterozygote. All of the Bz IGRs carry a CO arrangement of flanking markers, again that expected from the location of the mutations within the gene. These data reinforce the conclusion that the highly skewed recovery of COs seen among IGRs in maize is attributable to the high number of SNPs by which the two progenitor alleles differ.

Recombination between Dimorphic Heteroalleles

An examination of the published data on intragenic recombination at bz (Dooner and Kermicle, 1986; Dooner and Martínez-Férez, 1997) reveals that, in contrast to the situation for heteroalleles of different origin, NCO IGRs can be recovered in appreciable numbers from pairs of heteroalleles with a common origin. Heterozygous combinations of two insertions and of two point mutations were examined previously (Dooner and Kermicle, 1986). The proportion of NCOs was particularly high (60 to 80%) among IGRs from heterozygotes containing two Ds insertions at different positions within the gene. This observation led to the speculation that the heteroallelic Ds insertions might pair with each other, somehow affecting the resolution of the DHJ recombination intermediate postulated in the DSBR model. Now, a more systematic analysis of recombination between heteroallelic mutations derived from a common progenitor allele has been performed. Substantial additional data from pairs of point mutations have been generated, and the analysis has been extended to the critical heterozygous combination of a transposon insertion mutation and a transposon excision footprint.

Table 3 summarizes both published and new data obtained from 11 different heteroallelic pairs of bz-E point mutations derived from Bz-W22. These heterozygotes are referred to as dimorphic because they differ at only two sites, those corresponding to the EMS mutations. For every heterozygote, the NCO class on the left carries the flanking markers of the proximal heteroallele and the NCO class on the right carries the flanking markers of the distal heteroallele. Although there is some variation from one heterozygote to another in the CO/NCO ratio and in the ratio of the two NCO classes, the variation does not follow any pattern and probably represents the random fluctuation typical of small numbers (Dooner, 1998).

Table 3.

Recombination between Pairs of Dimorphic bz Heteroalleles: Neither Allele Carries an Insertion

Genotype (P/D)	Population					Total	Frequency (×10−3)	CO/NCO Ratio
Sh bz-E2 wxa	63,270	3	2	23	1	29	0.46	–
sh bz-E3 Wx
Sh bz-E2 wxa	56,560	12	3	36	1	52	0.92	–
sh bz-E4 Wx
Sh bz-E2 wx	70,520	12	12	26	2	52	0.74	–
sh bz-E6 Wx
Sh bz-E5 wx	37,265	5	12	10	0	27	0.72	–
sh bz-E6 Wx
Sh bz-E2 wx	19,540	2	0	10	0	12	0.61	–
sh bz-E8 Wx
Sh bz-E5 wx	16,370	5	1	8	0	14	0.86	–
sh bz-E8 Wx
Sh bz-E2 wxb	40,350	8	6	16	2	32	0.79	–
sh bz-E9 Wx
Sh bz-E5 wxb	37,400	3	4	14	1	22	0.59	–
sh bz-E9 Wx
Sh bz-E3 wxb	33,075	2	1	6	0	9	0.27	–
sh bz-E4 Wx
Sh bz-E3 wx	80,620	5	7	5	0	17	0.21	–
sh bz-E6 Wx
Sh bz-E4 wx	194,950	4	3	5	2	14	0.07	–
sh bz-E6 Wx
Total	–	61	51	159	9	280	–	1.3

^aData from Dooner (1986).

^bData from Dooner and Martínez-Férez (1997).

On the other hand, the cumulative data in Table 3 reveal clear trends. First, as in Tables 1 and 2, one CO class predominates among Bz IGRs (159:9), that expected from the location of the mutations in the gene. Second, the two NCO classes now occur in much larger numbers. In fact, the CO/NCO ratio, obtained by dividing the majority CO class by the sum of the three other classes, is slightly >1. The sharp difference in the recovery of the NCO class of IGRs in dimorphic versus polymorphic heterozygotes suggests that the large deficit of that class in the latter heterozygotes (Table 2) can be attributed to the multiple SNPs that distinguish one heteroallele from the other (χ² = 62.4, 1 df, P < 0.001). Third, the two NCO classes occur in similar numbers (61:51), that is, the NCO class that carries the flanking markers of the proximal allele is recovered approximately as frequently as the class that carries the flanking markers of the distal allele.

If one accepts that these classes represent gene conversions not accompanied by the exchange of flanking markers, then there is no indication in these data for preferential conversion of the proximal or distal allele. In yeast, gene conversion does not interfere with crossing over in an adjacent interval (Petes et al., 1991). The 9 individuals recovered in the minority CO class do not differ significantly from the 15 expected from conversion of the distal bz heteroallele plus a coincidental exchange in the bz-wx interval (χ² = 3.2, 1 df, P > 0.05). Because a fraction of the majority CO class also must have a similar origin, the true CO/NCO ratio is even smaller than that estimated above.

An even greater proportion of NCOs (60 to 80%) was found previously among IGRs from pairwise combinations of Ds insertion heteroalleles, leading to the hypothesis that the Ds insertions might affect the outcome of recombination events (Dooner and Kermicle, 1986; data reprised in Table 4). The three Ds insertion heteroalleles in that study, bz-m1, bz-m2(D1), and bz-m2(D2), all were derived from the Bz-McC progenitor allele. The subsequent realization that the Bz alleles used in the earlier studies were highly polymorphic (Ralston et al., 1988) and the finding that the density of polymorphisms within a gene could affect the distribution of intragenic exchanges (Dooner and Martínez-Férez, 1997) led to a reassessment of the effect of Ds insertions on recombination. In the previous section, it was concluded that the presence of a Ds insertion in one heteroallele does not affect the pattern of flanking marker recovery among IGRs in polymorphic heterozygotes. In this section, the effect of Ds on recombination in dimorphic heterozygotes is examined.

To determine whether the presence of a Ds insertion in both homologs had any effect on the recovery of CO and NCO IGR types, one of the transposon insertions was replaced with a transposon excision footprint at the same site. In bz-s2.1 and bz-s2.2, an 8-bp excision footprint replaced the 3.3-kb Ds2(D1) and 3.8-kb Ds2(D2) insertions present in bz-m2(D1) and bz-m2(D2), respectively. The large body of data collected from the two bz-m/bz-s heterozygotes is summarized in Table 4. Once again, NCO IGRs predominate, and the sum of two NCO classes relative to the total is approximately the same as that in bz-m/bz-m heterozygotes (χ² = 1.8, 1 df, P > 0.10).

Combined with the data in Tables 1 to 3, these data clearly show that the change in the predominant IGR type from CO to NCO reported previously at bz was not attributable to the presence of Ds insertions in one or both homologs but to a change from a polymorphic to a dimorphic heteroallelic situation. Nevertheless, a comparison of the data in Tables 3 and 4 reveals that insertions do have an effect in dimorphic heterozygotes (χ² = 146, 1 df, P < 0.001). As discussed above, approximately half of the Bz IGRs listed in Table 3 represent true COs. What fraction of the Bz IGRs listed in Table 4 represent true COs? For the bz-m/bz-m combinations, practically all apparent COs can be explained by conversion of either the proximal or the distal insertion allele and a coincidental exchange in a flanking region. For the bz-m/bz-s combinations, there is a only a slight excess of the expected CO class. Therefore, in dimorphic heterozygotes, the presence of an insertion in one or both alleles appears to interfere with the recovery of CO IGRs.

Another point about the data in Table 4 that is worth noting relates to the differential recovery of the two parentally marked classes. In contrast to the situation with bz-E/bz-E heterozygotes (Table 3), the two NCO classes are not recovered equally in any of the heterozygotes listed in Table 4. The proximal NCO class is recovered preferentially in bz-m1/bz-m2(D1) and bz-m1/bz-m2(D2) heterozygotes (χ² = 17.2, 1 df, P < 0.001). The excess of the proximal NCO class in bz-m1/bz-m2(D1) and bz-m1/bz-m2(D2) heterozygotes, which had been noted previously (Dooner and Kermicle, 1986), prompted a discussion recently on the possibility that meiotic recombination in bz might exhibit polarity (Dooner, 1998; Thijs and Heyting, 1998). However, it is the distal NCO class that is recovered preferentially in bz-m1/bz-s heterozygotes (χ² = 17.4, 1 df, P < 0.001). The reversal in the majority NCO class of IGRs recovered from combinations of heteroalleles that occupy the same sites argues against polarity as the basis for the inequality between the two NCO classes. The possible basis of this inequality is discussed below.

DISCUSSION

This work represents an attempt to identify general rules that govern the outcome of intragenic recombination in maize, especially with respect to the issue of flanking marker distribution. To that end, recombination was analyzed in a variety of bz heterozygotes. The differences in the bz heteroalleles used in the recombination experiments ranged from just two nucleotides (dimorphic heteroalleles) to multiple sequence and structural polymorphisms (polymorphic heteroalleles). A systematic comparison of the effects of sequence divergence on the outcome of recombination led to the following conclusions: (1) Most recombination in polymorphic heterozygotes results in an exchange of flanking markers. This is true regardless of the heteroallelic combination studied: a large insertion and a point mutation (Table 1), an 8-bp excision footprint and a point mutation (Table 2), two point mutations (Table 2), or two large insertions (Dooner and Ralston, 1990). (2) Recombination in dimorphic heterozygotes often is not accompanied by an exchange of flanking markers. The proportion of parentally marked IGRs appears to depend on the nature of the mutations in the heterozygote and can range from ∼50% in heterozygotes between two point mutations (Table 3) to >90% in heterozygotes between two insertions (Table 4). (3) The two parentally marked classes (often inappropriately referred to as convertants [Pâques and Haber, 1999]) can be recovered in unequal numbers in dimorphic heterozygotes, but this seems to be a function of the nature of the mutations, not of the relative positions of the mutations within the gene. Bias was seen in heterozygotes between either two insertions or an insertion and an excision footprint, but not in heterozygotes between point mutations. The bias can favor either the proximal or the distal allele, and at the bz locus at least, it appears to depend simply on the size of the lesion. At the r locus, bias against the same distal site was seen in heterozygotes between two insertions, two excision footprints, or an insertion and an excision footprint. Thus, at r, the bias could reflect the location of the mutation within the gene (J.L. Kermicle, personal communication).

The sharp differences in the outcomes of intragenic recombination experiments involving polymorphic and dimorphic heteroalleles can be explained in light of current versions of the original DSBR model of meiotic recombination in yeast that propose dual pathways for the formation of CO and NCO products (Gilbertson and Stahl, 1996; Pâques and Haber, 1997; Nakagawa et al., 1999; Allers and Lichten, 2001). This central modification was made necessary by several observations that were not consistent with the unitary canonical model. That model postulated that a central intermediate containing a DHJ is resolved to generate either NCO or CO recombinants (Szostak et al., 1983; Sun et al., 1991).

The Allers and Lichten (2001) model (Figure 2) provides a good framework to explain the data. The first part of this model follows the canonical DSBR model. A double-strand break is resected to expose 3′ single-strand tails, one of which invades the homolog and initiates DNA synthesis. Physical demonstration of this single-end invasion was provided recently by Hunter and Kleckner (2001). The resulting D loop has two possible fates. (1) It may capture the second end, serving as a template for additional DNA synthesis. Ligation of the newly synthesized strands to the recessed ends then results in a DHJ that is resolved to produce COs exclusively. This is, in essence, the original repair pathway, with the main modification being that resolution of the DHJ gives rise only to CO products. The mechanism by which this directed resolution occurs is not known. (2) Alternatively, the D loop intermediate may be disassembled by displacement of the newly synthesized DNA, which then can anneal with the second single-strand tail. Break repair is completed by DNA synthesis and ligation. This is the synthesis-dependent strand annealing (SDSA) pathway, which results only in NCO products (for review, see Pâques and Haber, 1999).

Figure 2.

Modified DSBR Model of Recombination with Separate Pathways for the Formation of CO and NCO Products.

The diagram shows the two interacting DNA duplexes (chromatids) as red and blue double strands. A double-strand break (DSB) formed in the blue duplex is resected by a 5′ to 3′ exonuclease to generate long 3′ single-strand overhangs. One single strand invades the homologous duplex, forming a heteroduplex and displacing a D loop, which is extended by DNA synthesis. The resulting D loop has two possible fates.

(A) In the DSBR pathway, the D loop anneals to the second 3′ end overhang, serving as a template for additional DNA synthesis. Ligation of the newly synthesized strands to the recessed ends leads to the formation of a DHJ with heteroduplex DNA flanking the double-strand break site. This recombination intermediate can be resolved by cutting the outside strands or the inside strands of each junction. The resolution always involves opposite sense cutting (two outside strands and two inside strands); only one of the two alternatives is shown here. This modified version of the DSBR pathway (Szostak et al., 1983) would generate only CO products.

(B) In the SDSA pathway (Pâques and Haber, 1999), the D loop intermediate is disassembled by displacement of the newly synthesized DNA strand, which then anneals with the second single-strand overhang. Repair of the break is completed by DNA synthesis and ligation. The SDSA pathway generates only NCO products.

(Figure adapted from Allers and Lichten, 2001.)

In this model, the decision to repair a double-strand break as a CO (via a DHJ) or a NCO (via SDSA) would occur at or soon after the initial step of strand invasion. The bz data presented here suggest that that decision is affected by the number of mismatches in the heteroduplex DNA formed by the invading strand. Polymorphic heteroalleles that differ, on average, in 1.5% of their bases yield almost exclusively CO IGRs, whereas dimorphic heteroalleles yield a mixture of CO and NCO IGRs. Observations on the effects of heterologies in yeast are consistent with these findings. Multiple heterologies (up to 1%) result in longer conversion tracts at meiosis (Schultes and Szostak, 1990) and mitosis (Nickoloff et al., 1999), which are more likely to be associated with crossing over (Aguilera and Klein, 1989).

This effect could be mediated by proteins involved in DNA mismatch repair, which are known to affect recombination (for review, see Pâques and Haber, 1999). For example, mutants deficient in MLH1 show reduced crossing over in both yeast (Wang et al., 1999) and mice (Baker et al., 1996; Woods et al., 1999). The yeast heterodimeric complexes MLH1-MLH3 and MSH4-MSH5 have been proposed to promote the processing of double-strand breaks to form DHJs, reducing alternative outcomes such as SDSA recombination (Nakagawa et al., 1999). Binding of mismatch repair proteins to a highly mismatched heteroduplex could interfere with strand displacement, forcing second-end capture and formation of the DHJ, which is resolved as a CO. In contrast, in the absence of extensive heterologies, the newly synthesized strand would be displaced readily, leading to the recovery of NCO products.

The almost exclusive occurrence of CO IGRs seen regularly in maize, but not in yeast, could be explained simply by the much higher degree of polymorphisms between the heteroalleles that have been used in maize recombination experiments. Although different yeast laboratories may work with different strains, experiments in any one laboratory tend to be with mutant lines isolated within the same strain (Pâques and Haber, 1999). In contrast, maize geneticists have tended to analyze combinations of highly polymorphic heteroalleles, often of different geographic origins, that may or may not have been introgressed into a common genetic background. The present study examined the effect of mismatches at a single level and distribution of heterologies, those existing between the Bz-W22 and Bz-McC parental alleles (Figure 1). An intriguing question raised by this study concerns the lowest level of heterology that leads to an almost exclusive recovery of CO IGRs.

Other observations in yeast and Escherichia coli have indicated a role for the mismatch repair system in controlling recombination. In E. coli, the mismatch repair proteins MutS and MutL, which are known to act as a barrier to interspecific recombination (Rayssiguier et al., 1989), inhibit homologous recombination between repeats that are just 2.2% divergent (Petit et al., 1991) as well as RecA-catalyzed strand transfer between phage sequences that are 3% divergent (Worth et al., 1994). Similarly, in budding yeast, MSH2 and MSH3 suppress mitotic homologous recombination with related sequences from fission yeast (Selva et al., 1997). Therefore, the mismatch repair system appears to act as a checkpoint to prevent recombination between divergent sequences. Likewise, the approximately twofold reduction in recombination between bz alleles that are 1.5% divergent could be a reflection of this regulatory role of the mismatch repair system.

The data in this report document two effects of insertions on the outcome of intragenic recombination. First, insertions appear to decrease the ratio of COs recovered in dimorphic heterozygotes (cf. Tables 3 and 4). This finding suggests that, in the absence of single base pair mismatches, insertions may promote strand displacement and, therefore, the NCO pathway. It is known that different MSH and mLH complexes interact with base/base mispairs and insertion/deletion mispairs (Nakagawa et al., 1999; Kearney et al., 2001), so the insertion effect possibly is mediated by components of the mismatch repair system. Second, the size of the insertion appears to affect the ratio of the two NCO classes in dimorphic heterozygotes (Table 4). The preferential recovery of the parentally marked class corresponding to the smaller insertion, regardless of the latter's location in the gene, suggests that mismatch correction simply is more efficient for smaller rather than larger insertions.

The present data also bear on the potential suppressive effect of insertions on recombination. The overall frequency of Bz IGRs is approximately fourfold higher in bz-m/bz-s than in bz-m/bz-m heterozygotes (Table 4). This indicates that, in dimorphic heterozygotes, the large Ds insertion in the middle of the gene suppresses recombination when the other allele also carries an insertion. No comparable data are available at this time for the effect on recombination frequency of the 1.2-kb sDs insertion in bz-m1, which lies close to the 5′ end of the bz gene (Figure 1). However, further removal of that insertion may not have much of an effect, because a previous comparison of recombination rates in bz-m/bz-m heterozygotes and dimorphic insertionless heterozygotes, such as those listed in Table 3, also revealed a fourfold difference between them (Dooner and Martínez-Férez, 1997). Possibly, the suppressive effect of insertions on recombination in dimorphic heterozygotes is seen only when both alleles carry an insertion. Clarification of this point will have to await the isolation of bz-s derivatives from bz-m1 that can be paired alternatively with bz-m2(D1), bz-m2(D2), and a corresponding bz-s excision footprint allele (either bz-s2.1 or bz-s2.2).

Polymorphic heterozygotes represent a completely different situation, because the effects of insertions are confounded by the effects of SNPs. Nevertheless, it is possible to use transposon excision derivatives to determine if removal of an insertion increases the frequency of IGRs, most of which are COs. The bz-m1/bz-E6 and bz-m1/bz-E8 heterozygotes listed in Table 1 can be compared with the bz-s30.1/bz-E6 and bz-s30.1/bz-E8 heterozygotes listed in Table 2 to examine the effect of the 1.2-kb sDs insertion on recombination. As will be recalled, bz-s30.1 carries an 8-bp addition footprint just 19 bp upstream of the site where sDs inserted in bz-m1. Similarly, the bz-m2(D1)/bz-E6 and bz-m2(D1)/bz-E8 heterozygotes listed in Table 1 and the bz-s2.1/bz-E6 and bz-s2.1/bz-E8 heterozygotes listed in Table 2 provide a test of the effect of the 3.3-kb Ds2(D1) insertion on recombination.

As can be seen from the data, the overall frequency of Bz IGRs did not increase after the removal of either insertion. The frequency of Bz IGRs is essentially identical in the two comparisons involving bz-E8. It is slightly higher in bz-s30.1/bz-E6 than in bz-m1/bz-E6 and lower in bz-s2.1/bz-E6 than in bz-m2(D1)/bz-E6, but none of the frequency differences is significant (Stevens, 1942). The bz-m2(D1)/bz-E6 value appears disproportionately high compared with that of other heterozygotes involving the bz-m2(DI) allele and may represent an overestimate.

This comparison suggests that, in polymorphic heterozygotes in which only one allele carries an insertion, the insertion does not affect the frequency of recovery of IGRs (which are mostly COs). Therefore, most of the reduction in recombination detected in polymorphic heterozygotes relative to dimorphic heterozygotes (Dooner and Martínez-Férez, 1997) can be attributed to the multiple SNPs that differentiate the two heteroalleles and not to the presence of the insertion. The sDs insertion in bz-m1 was reported to suppress recombination in a short interval immediately adjacent to the insertion site (Dooner and Martínez-Férez, 1997). At present, the distribution of recombination junctions among Bz IGRs from bz-s30.1/bz-E6 and bz-s30.1/bz-E8 heterozygotes is being analyzed to determine whether that localized suppression is alleviated in the absence of the insertion.

METHODS

Description of bz Alleles

All of the alleles used in this study were in the common genetic background of the inbred maize (Zea mays) line W22. The aleurone phenotypes conditioned by the various alleles in the presence of all of the complementary factors for anthocyanin pigmentation are given in parentheses. A graphic summary of the locations and natures of the mutations is presented in Figure 1.

Bz-W22 (purple). This is the normal Bz allele carried in the W22 inbred line.

bz-E2 to bz-E9 (bronze). These are ethyl methanesulfonate–induced mutations from Bz-W22 (Dooner, 1986; Dooner and Martínez-Férez, 1997).

Bz-McC (purple). This is the normal progenitor allele of the bz-m2(Ac) mutation.

bz-EMc1 (bronze). This is an ethyl methanesulfonate–induced mutation from Bz-McC, generated by the same procedure used to obtain the bz-E1 to bz-E10 mutations from Bz-W22.

bz-m2(Ac) (purple spots on a bronze background). This is an allele arising from the insertion of the 4.6-kb Ac element at position 755-762 in the second exon of Bz-McC. The eight bases between 755 and 762 (TGGGGCAG) are duplicated on either side of Ac (McClintock, 1955; Ralston et al., 1988; GenBank accession number AF355378).

bz-s2.1 and bz-s2.2 (bronze). These are stable mutant derivatives of bz-m2(Ac) arising from independent excisions of Ac (Dooner and Belachew, 1989). The mutations carry the excision footprints TGGGGCAC/AGGGGCAG and TGGGGCAC/TGGGGCAG, respectively, at positions 755 to 762. The underlined nucleotides represent transversions in the flanking direct repeat of bz-m2(Ac). The net addition of 8 bp at the previous site of Ac creates stable frameshift mutations.

bz-m2(D1) (bronze in the absence of Ac; spotted in its presence). This is derivative 1 from bz-m2(Ac), harboring a 3.3-kb internally deleted Ds element at the same position as Ac in bz-m2(Ac) (McClintock, 1955; Dooner et al., 1986).

bz-m2(D2) (bronze in the absence of Ac; spotted in its presence). This is derivative 2 from bz-m2(Ac), harboring a 3.7-kb internally deleted Ds element at the same position as Ac in bz-m2(Ac) (McClintock, 1955; Schiefelbein et al., 1985; Yan et al., 1999).

bz-m1 (bronze in the absence of Ac; spotted in its presence). This allele arose from the insertion of a 1.2-kb sesquiDs (sDs) element at position 161-168 of Bz-McC (McClintock, 1951; Martínez-Férez and Dooner, 1997).

bz-m30(Ac) (purple spots on a bronze background). This allele arises from the transposition of Ac from the tac2094 site in the mkk1 gene located 108 kb proximal to bz (Dooner and Belachew, 1989; Fu et al., 2002) to position 142-149 in the first exon of Bz-McC. The eight bases between 142 and 149 (TTCCTCTG) are duplicated on either side of Ac. This allele is a second-cycle bz-m derivative from bz-m2(Ac) that also carries the in-frame, 6-bp addition footprint CGGCGA at the original site of insertion of Ac.

bz-s30.1 (bronze). This is a stable mutant derivative of bz-m30(Ac) arising from the excision of Ac. The new mutation carries the frameshift 8-bp excision footprint TTCCTCTA/ATCCTCTG at the previous site of Ac insertion in bz-m30(Ac) as well as the 6-bp addition footprint CGGCGA at the original site of insertion of Ac in bz-m2(Ac).

bz-R (bronze). This is the bz reference allele, associated with a 340-bp deletion that extends from within the single intron to the second exon of bz and includes the Ac insertion site in bz-m2(Ac) (Rhoades, 1952; Ralston et al., 1988).

Markers

The mutations sh (shrunken endosperm) and wx (waxy endosperm) were used as markers flanking bz. They map, respectively, ∼3 centimorgan distal and 25 centimorgan proximal to bz in 9S. The sh-wx region exhibits very high chiasma interference (Dooner, 1986), so double crossovers in the region are rare.

Selection and Analysis of Purple-Kernel Intragenic Recombinants

bz heteroallelic combinations were hand-pollinated with a sh bz-R wx tester in an isolated detasseling plot. Putative Bz intragenic recombinants were selected as single purple seed in ears otherwise containing only bronze seed. The selections were classified for outside markers and backcrossed to the male parent to verify their heritability and the recovery of the pollen markers.

DNA Extraction and Sequencing

Leaf DNA isolation, primer sequences, polymerase chain reaction amplification, and DNA sequencing were as described previously (Dooner and Martínez-Férez, 1997). Figure 1 shows the locations of all of the primers used in either amplification or sequencing.

Accession Number

The GenBank accession number for bz-m2(Ac) is AF355378.