Abstract
The maize r locus encodes a transcription factor that regulates the developmental expression of the plant pigment anthocyanin. In an unusual example of gene regulatory diversity, the R-sc (Sc, strong seed color) and the R-p (P, plant color) alleles of r have nonoverlapping tissue specificity and nonhomologous 5′ flanking sequences. Heterozygotes between wild-type P and Sc mutants with Ds6 transposable element inserts (r-sc:m::Ds6 or sc:m) produce colored seed derivatives (Sc+) during meiotic recombination. The sc:m alleles with Ds6 insertion in 3′ regions of r produce crossover Sc+ derivatives. sc:m alleles with Ds6 elements inserted in 5′ regions produce rare Sc+ derivatives borne on nonrecombinant chromosomes. Among 52 such noncrossover Sc+ derivatives, 18 are indistinguishable from the Sc progenitor in phenotype and DNA sequence [Scp(+) alleles]. The remaining 34 derivatives have strong Sc+ expression, including darkly pigmented aleurone, scutellum, coleoptile, and scutellar node [Scp(e) alleles]. The coleoptile and scutellar node phenotypes are unique from either progenitor but are similar to those of some naturally occurring r alleles. Both classes of Sc+ derivatives are explained by gene conversion between the promoter region of Sc:124 and a homologous region located proximal to P. The recombinational intermediate formed between sc:m alleles and P results in deletion of the Ds6 element alone or both Ds6 and a nearby unrelated transposable element-like sequence.
ONE of several loci in maize that regulates the transcription of the genes that encode the biosynthetic enzymes for anthocyanin, the purple pigment found in plants (reviewed in Dooneret al. 1991), is r (red-1). r, and the unlinked genes b (booster-1) and in (intensifier-1), encode helix-loop-helix proteins of the myc family of transcription factors (Ludwiget al. 1989; Radicellaet al. 1991; Goffet al. 1992; Burret al. 1996). These genes determine the developmental timing and tissue specificity of anthocyanin deposition in maize (reviewed in Ludwig and Wessler 1990). Naturally occurring r alleles are unique among plant genes because of the diverse array of tissue-specific expression patterns that pigment nearly all parts of the plant and kernel. For example, R-Nj:Cudu (Navajo) controls pigmentation in the crown of the kernel, the root, coleoptile, scutellar node, silk, and anthers, whereas Lc (leaf color) pigments the leaf blade, ligule, auricle, glume, lemma, palea, pericarp, and midrib of the adult plant (Styleset al. 1973; Ludwiget al. 1990).
The r chromosomal region in many maize varieties is structurally complex, containing up to five genes per r haplotype. Individual genes in a haplotype can be arranged in either direct or inverted orientation and either closely spaced or spread over several hundred kilobase pairs (Egglestonet al. 1995; Walkeret al. 1995). This complexity is partially responsible for generating the diverse expression patterns of r alleles during maize evolution; individual r genes within a given haplotype are often associated with different tissue-specific patterns of expression. Genes in complex r haplotypes also participate in the epigenetic phenomena of paramutation (Brink 1958; reviewed in Chandleret al. 2000) and imprinting (Kermicle 1970a; reviewed in Kermicle 1996; Alleman and Doctor 2000). Both of these phenomena are related to structural complexity at r (Kermicleet al. 1995; reviewed in Alleman and Doctor 2000). Simplex (single transcriptional unit) r alleles can be isolated by out-of-register pairing and recombination between genes in complex haplotypes (McWhirter and Brink 1962; Dooner and Kermicle 1971; Robbinset al. 1991; Egglestonet al. 1995). Such r alleles are well suited for studies of gene regulation and recombination. Simplex r alleles have been used to define many aspects of r gene function, including identification of nuclear localization signals (Liuet al. 1998), gene regulation by upstream open reading frames (Wang and Wessler 1998), and cis-acting control sequences (Liuet al. 1996).
r and other kernel-specific maize genes are also used in the study of homologous recombination in higher plants. Recombinant alleles can be recovered from heteroallelic combinations by selecting phenotypic derivatives from large populations of progeny. Several novel observations have resulted from studies of recombination in plants compared to the yeast Saccharomyces cerevisiae. First, studies with the genes a (anthocyaninless-1), r, wx, and bz (bronze-1) indicate that, although conversion tracts are the same approximate length in plants vs. yeast, noncrossover products are recovered much less frequently than crossover products (Dooner and Kermicle 1986; Dooner and Ralston 1990; Xuet al. 1995; Dooner and Martinez-Ferez 1997; Okagaki and Weil 1997). This result is in contrast with similar studies in yeast in which equivalent numbers of crossover and noncrossover products are usually recovered for alleles at homologous locations (reviewed in Peteset al. 1991). Another interesting observation regarding recombination in maize genes is an absence of recombinational polarity for crossover resolution (Dooner and Martinez-Ferez 1997; Okagaki and Weil 1997). The strong polarity of conversion tracts among the many yeast genes studied is interpreted as resulting from a limited number of sites at which double-stranded breaks (DSBs) can initiate recombination (Peteset al. 1991). Thus, a lack of polarity might indicate more frequent or nonspecific sites for DSBs (Schnableet al. 1998).
This study explores recombination between two r alleles chosen for their phenotypic dissimilarity. R-sc (Sc) and R-p (P) alleles are derived as simplex alleles from the R-st (stippled) and R-r:std (standard) complexes, respectively (McWhirter and Brink 1962; Dooner and Kermicle 1971; Robbinset al. 1991; Egglestonet al. 1995). Sc alleles specify strong anthocyanin pigmentation in the aleurone and scutellum of the maize kernel and weak pigmentation of the coleoptile of the seedling. P alleles pigment the root and coleoptile of the seedling and the anthers of the mature plant, but produce colorless aleurone. The phenotypes of Sc and P alleles are shown in Figure 1, A and B. Recombinant products are readily recovered from heteroallelic combinations of wild-type P and mutant Sc alleles. These alleles retain the tissue specificities of the 5′ flanking sequences from the original progenitor. These results indicate the equivalence of 3′ coding regions of P and Sc and the location of regulatory sequences in the 5′ nontranscribed regions (Kermicle 1985, 1988).
The target alleles for this study include indistinguishable P alleles and 41 transposable element-containing mutants of the simplex Sc:124 allele. This series of r-sc:m::Ds6 mutants (or, for simplicity, sc:m) were generated using a cyclic mutagenesis scheme in which the transposable element Ds6 was mobilized by the autonomous element, Ac, to excise and reinsert back into Sc at a high frequency (Kermicleet al. 1989; Alleman and Kermicle 1993). This strategy takes advantage of the property of Ac and Ds to reinsert at positions near the donor site during transposition (Dooner and Belachew 1989; Morenoet al. 1992). We have determined, in this study and previously, that each of the sc:m mutants contains the same Ds6 transposable element. Insertion sites for the mutants are distributed from 2.5 kbp 5′ of the start of transcription to the end of the protein-coding region (Alleman and Kermicle 1993). Each allele is germinally and somatically stable in the absence of Ac but is unstable (spotted) in the presence of Ac, typical of Ds-containing alleles (McClintock 1951; Kermicleet al. 1989).
We have used heteroallelic combinations of P and 39 of the 41 sc:m to recover colored seed (Sc+) progeny. This article reports a striking difference in the recombinational properties of sc:m that are located in 3′ as compared to 5′ regions of Sc:124. Frequently, 3′-located mutants recombine with P to produce Sc+ derivatives, usually crossover in origin. In contrast, 5′-located mutants produce Sc derivatives that are reduced in total frequency and noncrossover in origin. In addition, some of the Sc+ derivatives from 5′-located sc:m have a novel phenotype in which aleurone, scutellum, coleoptile, and scutellar node are strongly pigmented, distinct from sc:m, P, or Sc:124. These derivatives are associated with loss of the Ds6 element and another nearby sequence from Sc:124. Structural attributes of the r gene in which the two groups of sc:m alleles reside are not sufficient to explain this recombinational dichotomy.
MATERIALS AND METHODS
Maize lines and genetics: All maize lines used had an inbred W22 genetic background, established by five or more generations of backcrossing. The r allele R-sc:124 (Sc:124) is a simplex allele that is the progenitor of all sc:m mutants described in this report. The phenotype of Sc:124 and other similar Sc alleles is strong nonimprinted aleurone pigmentation and weak coleoptile pigmentation. The sc:m alleles used in this research are colorless or pale seed mutations described previously (Kermicleet al. 1989). The simplex R-p (P) alleles used in this research (P:n19, P:n46, and P:n142) are equivalent in structure (Robbinset al. 1991). P alleles pigment coleoptile, root, and anthers strongly but do not produce seed pigmentation. They are collectively designated as “P.” The phenotypes of Sc and P alleles are shown in Figure 1, A and B.
Several r alleles were used in these experiments because of contrasting phenotypes or structures readily distinguished from the test alleles. The r-Δ902 allele is a deletion including 11.3 kbp of r coding region and 5′ untranslated sequences (Alleman and Kermicle 1993). The r-g:8p allele is a pale seed colorless plant derivative of R-r:std. R-Navajo:Cudu (Nj) is a simplex r allele that pigments the crown of the maize kernel, the root, coleoptile, scutellar node, silk, and anthers (Styleset al. 1973; Dellaportaet al. 1988). The R-st (stippled) haplotype is composed of four r genes [(Sc), (Nc1), (Nc2), and (Nc3)] (Egglestonet al. 1995). The R-r:std (standard) haplotype is composed of three r genes [(P), (S1), and (S2)] and one r-promoter pseudogene (q) (Robbinset al. 1991; Walkeret al. 1995).
The r-linked loci g1 (golden-1 plant color, 20 cM proximal) and mst1 (modifier of stippling due to R-st, 6 cM distal) were used as flanking markers for recombination studies. The g1 alleles that were used are G1-w22 (wild type) and g1-1 (mutant). The g1 phenotype is scored during a germination test in which g1-1 produces a light golden-colored plant tissue relative to G1-w22, which produces green tissue. The Mst1-1 mutant allele is a dominant modifier of stippling that is assayed in the presence of the r-g:nc3.5 allele (Kermicle 1970b). r-g:nc3.5 produces aleurone that is stippled with sectors of anthocyanin pigmentation in the presence of Mst1-1. The r-g:nc3.5 allele produces nearly colorless aleurone in the presence of mst1-w22. The w22 designation of alleles refers to those present in the W22 inbred background.
The phenotypes of progenitors and Sc+ derivatives from sc:m/P heteroallelic combinations. (A) The Sc allele, Sc:124; (B) the P allele, P:n46; (C) the Scp(+) type allele, R-sc:n5993; and (D) the Scp(e) type allele, R-sc:n6008. B includes abbreviations referring to parts of a maize seedling: AL, aleurone; C, coleoptile; SC, scutellum; SN, scutellar node; R, root. The aleurone is part of the triploid maize endosperm whereas the coleoptile, scutellum, scutellar node, and root are derived from the developing embryo. Maize seeds were germinated under fluorescent lights at room temperature and then incubated at 18°–20° (12 hr light/12 hr dark) for 5 days. All seedlings shown are homozygous for the specified r alleles.
The P-vv::Ac allele (variegated pericarp color) of the P locus on chromosome 1 was used as the source of the autonomous transposable element Ac (Activator) for these experiments. The unlinked recessive allele wx1-1 (waxy-1 seed) was used as a pollen contamination marker.
sc:m mutant alleles used in these experiments include 42 insertions of the transposable element Ds6 (sc:m3, m301-m328, m330-m341, and m9) and one with insertion of transposable element Ds1 (sc:m1). These alleles produce colorless or pale aleurone in the absence of transposable element Ac and spotted aleurone in the presence of Ac (Kermicle 1980; Kermicleet al. 1989). The Scp(+) type alleles produce a phenotype that is indistinguishable from Sc:124, with darkly pigmented aleurone and scutellum and weakly pigmented coleoptile. Alleles that were used for molecular analysis include: Sc:n5993 and Sc:n5991 (derived from m305), Sc:n7097 and Sc:n7098 (derived from m316), and Sc:n6732 and Sc:n7052 (derived from m330). The Scp(e) type alleles produce darkly pigmented aleurone, scutellum, coleoptile, and scutellar node. Alleles that were used for molecular analysis include Sc:n5992 (derived from m305), Sc:n6008 and Sc:n6009 (derived from m311), Sc:n7095 (derived from m316), and Sc:n7053 (derived from m330).
Isolation of Sc+ derivatives: The genetic starting material for these experiments includes 39 sc:m mutations of Sc:124 with insertion of the Ds6 element. Positions of the insertions are distributed throughout the r coding and 5′ flanking regions (Alleman and Kermicle 1993).
Recombination tests were performed as follows. Seeds of the genotype g1-1, P, Mst1-1/G1-w22, sc:m::Ds6, mst1-w22; wx1-1/Wx1-w22 and involving each of the sc:m alleles were planted as detasseled female rows in an isolation plot with alternating rows of homozygous g1-1, r-g:8p mst1-w22; wx1-1 as the male parent. The resulting ears were screened for Sc+ kernels. The selected kernels were progeny tested to determine Sc+ heritability and the flanking marker combination. Seedling phenotypes of P and g1 were assayed in germination trays under fluorescent lighting at 18°–20°. The mst1 phenotype was tested by crossing Sc+ individuals with the nearly colorless R-st derivative allele r-g:nc3.5 and pollinating the heterozygous plants with a colorless seed r allele. The recombinant Mst1-1, r-g:nc3.5 phenotype is spotted kernels while the mst1-w22, r-g:nc3.5 phenotype is unspotted kernels.
In most experiments, the genotype g1-1 sc:m mst1-w22/G1-w22 P (P:n19, P:n46, or P:n142) Mst1-1 was the female parent. In a few experiments involving 5′-located sc:m, seeds of the genotype g1-1 sc:m mst1-w22/G1-w22 P (p:m3 or p:m1) Mst1-1 in which a Ds insertion was also present at the P allele located in the protein-coding region were used (Dooner and Kermicle 1986; Alleman and Kermicle 1993). Data from this population are included in Table 2.
Molecular analyses: Maize DNA was isolated according to Chen et al. (1992) except that DNA extractions were performed at room temperature. DNA was digested, size fractionated, and blotted as described in Alleman and Kermicle (1993). For cloning, DNA from R-st was digested with Sau3AI (1.74 units/100 μg/ml) for 15, 30, 45, and 60 min; pooled; and partitioned on a 10–40% glycerol gradient. DNA fragments (8–15 kbp) were ligated with EMBL3 arms using T4 DNA ligase. A total of 1.2 × 106 recombinant λ phage were screened using radiolabeled r probe (Nj:1). Three clones (λMB1, λMD1, and λMF2) contained the 5′ region of the Sc gene, confirmed by restriction endonuclease mapping. Clone λR4.7 containing the P gene of R-r:std was received from Dr. T. Robbins (Robbinset al. 1991).
Regions of clones λMF2 and λR4.7 were subcloned into pBluescript vectors (Stratagene, La Jolla, CA) for DNA sequence analysis. Dideoxy chain termination sequencing using Sequenase II kits (United States Biochemical, Cleveland) was used with single-stranded templates. Sequencing strategies included subcloning, synthetic primer design, and the generation of nested deletions (Exo/Mung deletion kit, Stratagene). Subsequent sequencing of mutants and derivatives used double-stranded sequencing of plasmids or PCR products and the ABI automated DNA sequencing system (Perkin-Elmer, Norwalk, CT). All DNA sequence analysis used MacVector software (Oxford Molecular Group, Palo Alto, CA).
Synthetic oligonucleotides (Oligos or Operon, Alameda, CA) specific to regions of the Ds6 element or the r gene were used as PCR primers to generate probes, for inverse PCR, and for the mapping of the position of Ds6 elements in Sc. They include the following: B32 (CTCGATGTCTGTCAAGCGTTGG), B66 (TTCACCAAACAAAAGCTGGCGG), B971 (GCTAGCCTTGCTAATAC), B104 (TTGGCTGTACGCATTACCATGC), B129 (TTGTGTCATGTGGCTCC), B530 (GTACGTAACACGCATGG), B2261 (TAGGAGTTTGGGACTCTTCCG), F3 (TGCGACGGACAAGAAGATTTG), F4 (TTCTCGGACGTTTTCCCAGCAG), F17 (AAGCAGCAGGAGACGAAGGTAG), F59 (GACGGACGAATTGAAGAACACCTG), F66 (CCGCCAGCTTTTGTTTGGTGAA), F1178 (ATCCTCAGACGCTACACGACAC), MAR102 (GGATATACCGGTAACGAAAACGA), and MAR103 (TTCGTTTTTTACCTCGGGTTC). MAR102 and MAR103 are specific to transposable element Ac and Ds6 sequences. Other primers were from the Sc:124 sequence. The locations of Sc and P primers are shown in Figure 2.
The positions of the Ds6 insertions at the 5′ end of the Sc gene were determined using Ds6 primer MAR102 or MAR103 and a flanking r-specific primer. PCR was performed using genomic DNA as a template and Amplitaq DNA polymerase (Applied Biosystems, Foster City, CA). PCR products were purified and cloned into the pCRII vector (Invitrogen, San Diego) for sequencing.
Probes were gel-isolated restriction fragments or PCR products that were radiolabeled by random-primed incorporation of [32P]dCTP or digoxygenin-labeled dCTP (Roche Molecular Biochemicals, Indianapolis). PCR using primer combination F4/B971 and P DNA as a template was used to isolate the YL1 probe. Probes BH1415 and PX1029 were gel-purified restriction fragments from a subclone of λMF2. Nj:1 was a probe derived from clone pR-Nj:1, a BglII/HincII subclone from the R-Nj:Cudu mutant r-nj:m1 (Dellaportaet al. 1988). Probe YL202 was a PCR product generated using primers F3 and B66 and Sc:n5992 genomic DNA as a template. Probe YL210.1 was a PCR product generated using primers F3 and B530 and Sc:n6008 genomic DNA as a template. Locations of Sc regions used as probes are shown in Figure 2.
The location of P:A was determined with respect to the P gene using Southern blot analysis of restriction fragment length polymorphism (RFLP) data. Two types of recombinant alleles were used: (1) three Sc+ alleles from sc:m/P heterozygotes in which flanking marker combinations indicated a recombinant origin (Sc:n3081, Sc:n3082, and Sc:n3083 derived from sc:m308/P heterozygotes) and (2) three p:m::Ds alleles (p:m1, p:m3, or p:m9) that had the reciprocal combination of flanking markers relative to the Sc+ alleles (Kermicle 1988).
RESULTS
Recombination with P divides the Sc gene into two distinct regions: The general approach of these experiments was to use meiotic recombination to explore the structure and behavior of two distinct r progenitor alleles, Sc (seed color) and P (plant color). These two r alleles were derived from very different progenitor r haplotypes and have nonoverlapping tissue specificity. By constructing heterozygotes of 39 colorless seed sc:m mutants and wild-type P alleles, colored seed derivatives can be selected. The diagram in Figure 3 shows the composition of the heterozygotes that were used in these experiments and the strategy for derivative selection.
From populations of 13,700–46,500 progeny kernels, Sc+ derivatives were recovered in frequencies ranging from 0 to 0.233% Sc+ kernels per sc:m chromosome tested (Table 1). The 39 sc:m alleles fall into two distinct groups defined by the relative frequency of Sc+ derivatives and by the association of these derivatives with recombination of alleles of the flanking marker loci golden-1 (g1) and modifier of stippled-1 (mst1). Group I sc:m mutants produce Sc+ derivatives at relatively high frequencies, averaging 6.1 × 10−4 Sc+ per chromosome tested. Nearly all of these derivatives (416 of 434 Sc+) are carried on recombinant chromosomes carrying the 5′ flanking marker of the sc:m parent and the 3′ flanking marker of the P parent. In contrast, heterozygotes involving group II sc:m mutants produce Sc+ infrequently, averaging 6.4 × 10−5 Sc+ per chromosome tested, most of which are associated with the noncrossover flanking marker combination of the sc:m parent (21 of 28 Sc+). Spotted kernels were not present among the testcross populations, indicating absence of an active Ac element. Furthermore, germinal Sc+ revertants as well as spotted kernels were absent in populations of sc:m homozygotes (Kermicleet al. 1989). It is unlikely, therefore, that the noncrossover Sc+ class marked as parental sc:m is attributable to Ac-mediated transposition.
The positions of Ds6 insertions in the 5′ region of Sc. Differentially shaded regions represent sequence landmarks. The positions of the origin of transcription and translation (atg) are based on those determined for the Lc gene (Ludwiget al. 1989). “R common sequences” refer to regions present in both Sc and P, including 698 bp of noncoding DNA (shaded section) and all r-coding sequence (labeled in the figure). Vertical lines indicate positions of Ds6 insertions in the r-sc:m mutations. The Sc:A region is defined as Sc-specific sequences contained within λ clone MF2 and located 5′ of all r common sequences. The arrows represent the positions of PCR primers discussed in the text. Probes Nj:1, YL202, YL210.1, BH1415, and PX1029 are designated by shaded bars. Probe YL202 was synthesized by PCR using an allele (R-sc:n5992) with a deletion of the Bnot segment.
Structure of sc:m/P heterozygotes used as progenitors for Sc+ derivative selection. Shown are the r chromosomal regions containing an sc:m allele (top) and a P allele (bottom) that were used in these experiments. Outside flanking markers include g1 (golden plant parts) and mst1 (modifier of stippling by R-st). Sc and P are divided into transcribed (trc) and 5′-flanking (up) sequences. The differential shading of Sc:up and P:up denotes the nonoverlapping tissue specificity produced by these alleles. Ds6 insertions in sc:m are located in either Sc:up or Sc:trc (triangles). In some experiments, Ds6 insertions were located in P:trc. Recombination distances are designated in centimorgans. The sc:m phenotype is colorless or pale-colored seeds and colorless plant parts. The P allele phenotype is colorless seed and pigmented root, coleoptile, and anthers. Two types of Sc+ derivatives are produced by recombination between sc:m and P, designated as “a” (noncrossover) and “b” (crossover). Double crossover (DCO) progeny are marked with the P flanking markers (c).
Group II sc:m mutants are located in 5′ flanking regions of Sc: Preliminary studies indicated that group I alleles contain Ds6 insertions located in relatively 3′ regions of r whereas group II alleles are located in relatively 5′ regions, mostly upstream of the transcription start site (Alleman and Kermicle 1993). To determine the precise location of the insertions, the 5′ flanking region of the Sc gene and the insertion site of the mutants were sequenced, and 5′ portions of the Sc gene from R-st were cloned into λEMBL3 as recombinant phage λMF2, part of which was subcloned and sequenced (GenBank accession number for Sc 5′ flanking DNA: AF380388). The Sc sequence was compared with the published sequence of the lc gene, a displaced member of the r gene family, to determine sequence landmarks and the orientation of the cloned DNA (Ludwiget al. 1989).
The precise locations of the 5′-located group II alleles and alleles near the boundary of the two regions were determined by PCR analysis and DNA sequencing. PCR products overlapping the Ds6 insertion sites for each of the mutants were generated using oligonucleotide primers, one specific to Ds6 and the other using Sc sequence near the mapped positions of the sc:m alleles. Sequencing determined the insertion site for each mutant as the 8 bp of target site duplication adjacent to the Ds6 element. The insertion sites for the 19 5′-most Ds6 insertional mutants are shown on the map of Sc in Figure 2. On the basis of these results, group I and group II alleles are demarcated at an approximate position 370–470 bp 3′ of the Sc transcription start site or between the insertion sites of sc:m337 and m312. Although derivative frequencies from sc:m/P for sc:m located close to this position are infrequent, Ds6 insertion sites close to the boundary include m338 (four convertants/zero crossovers), m337 (two convertants/zero crossovers), m312 (zero convertant/one crossover), m320 (zero convertants/zero crossovers), and m308 (zero convertants/three crossovers). Crossover data for individual alleles are not delineated in Table 1.
Sc+ derivatives from G1-w22 r-sc:m mst1-w22/g1-1 R-p Mst1-1 heterozygotes
The sequence of the P region homologous to the sc:m insertion sites indicates a structurally segmented region proximal to the P allele: Because upstream regions of maize genes can be very different for alleles with distinct tissue specificities (Radicellaet al. 1992), it was important to compare the sequence of the Sc and P 5′ flanking regions to determine the location of the Ds6 insertion relative to homologous regions of P. The P allele 5′ flanking region was received as part of clone λR4.7 from Dr. T. Robbins (Robbinset al. 1991). Maize DNA from λR4.7 was subcloned and sequenced (published as GenBank accession no. AF380390). Sequence comparison of a 5646-bp region of λMF2 and a 4702-bp region of λR4.7 indicates that Sc and P are 99.7% identical in the 698 bp 5′ of the start of transcription. Further 5′ of −698, the two alleles are completely nonhomologous within the sequenced regions, including 3099 bp of 5′ flanking DNA from Sc (λMF2) and 3560 bp of 5′ flanking DNA from P (λR.47). The segment from λMF2 that is unique to the Sc gene contains the insertion site for 13 sc:m mutants and is designated Sc:A (Figure 2).
A recombination-based mechanism for the production of Sc+ alleles from sc:m/P heterozygotes requires that sequences homologous to the insertion sites of Ds6 mutations must be present in P. Southern blot analysis of genomic DNA from the P genotype indicated that sequences homologous to Sc:A are present in the P allele genome (not shown). This region (called P:A) was cloned using an inverse PCR (I-PCR) strategy (Ochmanet al. 1988). Genomic DNA of R-r:std (P) and the r deletion mutant r-Δ902 was digested separately with the restriction enzymes BamHI, BglII, and XbaI. Digested DNA was diluted to 2.5 ng/μl, favoring the formation of monomeric circles, and ligated with T4 DNA ligase to circularize the fragments. PCR using primers F4 and B530 amplified the flanking sequences. Southern blots of the PCR products were hybridized with YL210.1 to confirm homology with Sc:A. No PCR products or hybridization resulted when r-Δ902 DNA was used as the template (not shown). Three PCR products from P were cloned, including pYLB, pYLG, and pYLX, denoting the restriction enzyme source of the digested DNA as BamHI (B), BglII (G), or XbaI (X). Clone pYLB was sequenced as described above.
The sequence of I-PCR clone pYLB indicates a highly homologous, yet fragmented, P:A region relative to Sc:A (diagrammed in Figure 4). Four significant regions of discontinuity exist between these two sequences: (1) P:YL1 is a 1.6-kbp region located in P:A but not in Sc:A; (2) Bnot is a 471-bp region located in Sc:A but not in P:A (published independently of Sc:A as GenBank accession no. AF380387); (3) P:YL2 is a large region of unknown length located between P:A and P, partially contained in clone λR4.7 (P); (4) A 416-bp segment of Sc:A is located adjacent to the r common region but is not present in P or P:A (Figure 4). These four insertiondeletion differences between P:A and Sc:A are defined as sequences that are present on one homolog but absent from the other. Excluding these insertion-deletion differences, the homologous segments shared by Sc:A and P:A are identical in sequence, a total of 2209 bp of DNA.
To characterize the genomic and r-linked copy numbers of insertion-deletion differences between P:A and Sc:A, Southern blot analyses were performed. Probes included Nj:1, a 600-bp region surrounding the promoter of all r genes; BH1415, a 1.4-kbp segment of Sc:A that is homologous to both P:A and Sc:A; YL1, a segment unique to P:A; PX1029, a segment of Sc:A containing part of the Bnot element; and YL202, a region of Sc:A that overlaps PX1029 but excludes Bnot. Figures 2 and 4 show the location of these probes. Results are summarized below.
r genes Sc and P share regions of discontinuous homology in the 5′ region. Alignment of the 5′ regions of Sc and P relative to Ds6 insertion sites in sc:m mutants is based on sequence data for the 5′ flanking DNA of Sc, P, and a homologous, proximally linked segment found in the P genotype (P:A). The sequences of clones λMF2 (Sc), λR4.7 (P gene, Robbinset al. 1991), and YLB (P:A) were used in the generation of these maps. Sequence landmarks and regions of sequence nonhomology are diagrammed as shaded boxes and are consistent with shading in Figure 2. Homologous stretches of DNA in the P and Sc alleles are separated by blocks of nonhomologous DNA. Four insertion/deletions comparing Sc:A and P:A are shown on the map: (1) the Bnot insertion in Sc:A; (2) the YL1 region in P:A; (3) P:YL2, the P-specific 5′ flanking region adjacent to the P allele in clone λR4.7; and (4) a 416-bp segment located in Sc:A but missing from P or P:A. All inserts are to scale except the P-specific region P:YL2, of unknown length.
First, the relative copy number of the insertion-deletion differences was determined by Southern blot analysis. The YL1 region, present in P:A but not Sc:A, was used as a probe on Southern blots containing genomic DNA from Sc:124 and P:n46. A single band appears for P DNA but not for Sc:124 DNA (Figure 5A). YL1 is, therefore, unique to the genome of P and does not exist in the Sc gene or genome. Similarly, probes were generated from Bnot and the region of Sc:A that surrounds Bnot. The probe PX1029 contains most of the 471-bp Bnot segment. Hybridization of PX1029 to Southern blots containing Sc:124 and P:n46 DNA indicates that PX1029 hybridizes to ~20 bands for DNA of either genotype (Figure 5B). A larger probe of this region excluding the Bnot segment (YL202) hybridizes to a single band on the same Southern blots (not shown). Thus, Sc:A is a single-copy segment and the Bnot element is the cause of the repetitive hybridization of probe PX1029.
To determine the copy number of Sc:A/P:A within complex r haplotypes, Southern blots containing genomic digests of Sc:124, P:n46, complex r haplotypes R-st and R-r:std Lc, were hybridized with probes Nj:1 and BH1415. R-st and R-r:std are the complex progenitor alleles of Sc and P; these r haplotypes each contain multiple r genes or pseudogenes. Previous studies indicate that R-r:std and R-st contain four Nj:1-homologous segments (Robbinset al. 1991; Egglestonet al. 1995; Walkeret al. 1995). These Southern blot data can be summarized as follows: Whereas the r promoter region probe Nj:1 hybridizes to all r genes in complex alleles, the Sc:A probe BH1415 hybridizes to a single copy sequence in R-st, R-r:std, and all derivatives. A representative Southern blot is shown in Figure 5C. These results indicate that the Sc:A/P:A region is a single-copy sequence, which is not duplicated in complex r alleles such as R-st and R-r:std.
The P:A region is located proximally to P on chromosome 10 and is not duplicated in complex r alleles: Locating P:A relative to the P transcriptional unit was also accomplished by Southern blot analysis using RFLP analysis of recombinant r alleles. Two types of recombinant alleles were used, including (1) Sc+ from sc:m/P heterozygotes that were recombinant with respect to flanking markers and (2) the reciprocal product of this event, p:m::Ds (refer to Figure 3). Southern blots were produced using digested DNA from recombinant alleles and the Sc:A region probe BH1415 to map the position of P:A relative to the P structural gene. Recombinant Sc+ alleles (for example, Sc:n3081) contain the BH1415-hybridizing restriction fragment that is associated with the Sc allele. Recombinant p:m::Ds alleles (for example, p:m1 and p:m9) contain the BH1415-hybridizing restriction fragment that is associated with the P allele. Representative Southern blot data are shown in Figure 5D. This result indicates that the BH1415 homologous region P:A is proximal to P in R-r:std and simplex P alleles. In addition, individual restriction fragments from P, even fragments >15 kbp in length in which the 3′-located restriction site is located within the P promoter, do not hybridize to both BH1415 and Nj:1 (not shown), indicative of distance between the P promoter and P:A. Thus, P:A is located at a quasi-ectopic proximal site compared to the location of Sc:A, which is immediately 5′ to the Sc promoter.
RFLP and Southern blot analysis of the Sc:A/P:A regions of r. Maize genomic DNA was digested with restriction enzymes and analyzed by Southern blot hybridization, using 32P- or digoxygenin-labeled probes shown in Figures 2 and 3. Southern blots were probed with BH1415 (A and B), Nj:1 (B), YL1 (C), and PX1029 (D). (A) DNA from alleles P:n46 (lanes 1 and 2) and Sc:124 (lanes 3 and 4) were digested with BamHI (lanes 1 and 3) or EcoRI (lanes 2 and 4) and hybridized with digoxygenin-labeled YL1 probe. (B) DNA was digested with BamHI and HindIII and hybridized with 32P-labeled PX1029. The alleles included are Sc:124, Sc:n6008, and P:n46. The repetitive region from probe PX1029 is designated Bnot and is presumed to be a novel maize transposable element. (C) DNA was digested with HindIII and hybridized with 32P-labeled Nj:1 (lanes 1–5) or BH1415 (lanes 6–10). Alleles included in lanes 1–5, and repeated in lanes 6–10, are as follows: Sc:124, R-st, R-r Lc, R-g:1, and P:n46. The r gene composition for these alleles is as follows: Sc:124, [Sc]; R-st, [Sc, Nc1, Nc2, Nc3]; R-r Lc, [P, q, S1, S2, Lc]; R-g:1, [q, S1, S2]; and P:n46, [P]. Band assignment shown is based on published results for R-r:std and R-st (Robbinset al. 1991; Egglestonet al. 1995). (D) DNA was digested with HindIII and hybridized with 32P-labeled BH1415. Alleles included in lanes 1–7 are as follows: the sc:m mutants sc:m1, sc:m9, and sc:m308; the sc:m308/P recombinant Sc:n3081; the sc:m1/P and sc:m9/P recombinant alleles p:m1::Ds1 and p:m9::Ds6; and the P progenitor allele P:n46.
Bnot is a non-LTR-transposable element: Several features of the Bnot segment from Sc:A are consistent with this element being a maize transposable element. First, Bnot is homologous to a multicopy sequence in the maize genome (Figure 5B). In addition, the Bnot sequence in Sc:A is flanked by a 22-bp imperfect direct repeat sequence [TCTTGCTTTTTTCTACTT (T/G) TTT] that is present once in the P:A sequence. This 22-bp sequence is assumed to represent the target site that was duplicated during insertion of Bnot into the Sc gene of R-st during maize evolution.
The 449-bp Bnot sequence, excluding the 22-bp flanking repeats, was compared to sequence databases using the BLASTN program (Altschulet al. 1997). A strongly homologous sequence of 447 bp (83% sequence identity, BLAST expect value = 3 × 10−75) was identified in the maize genome several kilobase pairs downstream of the Adh1-F allele (Tikhonovet al. 1999). Nine of 10 bp near the termini of this element are duplicated and may represent part or all of the direct repeat produced during Bnot insertion near adh1-F (not shown). In addition, each of the six reading frames for Bnot was translated (MacVector software; Oxford Molecular Group, Genetics Computer Group) and compared with protein database sequences using BLASTP. One open reading frame of 107 amino acids identified homology to four putative plant reverse transcriptase sequences, including three rice sequences and one Arabidopsis sequence (Linet al. 1999). Overlap among the four sequences defines a total length of 82 amino acids. These sequences were identified as non-LTR-transposable elements with a range in homology to the Bnot open reading frame. Expect values of 0.26–4.8 with 76–50% protein sequence similarity were produced for the four database sequences.
Frequency and phenotype of Sc derivatives from heterozygotes of sc:m involving 5′ Ds6 inserts with P alleles
Recombination in sc:m/P heterozygotes involving the group II sc:m alleles produces a unique class of Sc derivatives: Initial populations of progeny from sc:m/P heterozygotes produced a dichotomy between the derivatives of group I vs. group II sc:m mutants. In these experiments, few Sc+ were obtained from group II (NCO-generating) alleles. To determine a possible unique relationship between derivatives from group I vs. group II sc:m and the basis for the difference between these groups, larger populations of progeny kernels were generated from group II mutants. Two types of parental populations were used. These included sc:m/P (n19, n46, or n142) as described in Table 1 and sc:m/p:m::Ds (p:m3 or p:m1). The latter populations, in which a Ds insertion was located in the 3′ region of P (see diagram in Figure 3), cannot generate Sc+ products by reciprocal recombination and were excluded from the data presented in Table 1. A total of 52 Sc+ were generated from sc:m/P or sc:m/p:m heterozygotes. Table 2 summarizes these data for individual alleles in group II mutants and those close to the border between groups II and I. A novel result emerges. Sc+ derivatives from eight of the group II sc:m are of two distinct phenotypic classes. The first class, Scp(+), is indistinguishable from Sc:124 crossover-derived Sc+ alleles from group I alleles and Sc+ revertants isolated in the presence of Ac. The second class, Scp(e) or enhanced pigment alleles, has the darkly pigmented aleurone and scutellum of Sc:124 but also has strongly pigmented coleoptile and scutellar node, unique from Sc:124, from the immediate progenitors sc:m and P, and from other Sc+ derivatives. Figure 1, C and D, shows the phenotypes of representative members of both classes of Scp alleles.
The Scp(+) and Scp(e) phenotypes are due to differences in gene conversion events in sc:m/P heterozygotes: The structures of the Scp derivatives were determined using PCR amplification and DNA sequence analysis. DNA from several derivatives of both Scp(+) and Scp(e) classes were amplified via PCR using primers F17 and B2261 and sequenced directly. A total of six Scp(+) and five Scp(e) alleles were sequenced. The sequences of representative derivatives and progenitor alleles are shown in Figure 6. In all Scp(+) that were sequenced, the Ds6 element is deleted, consistent with the revertant Sc+ phenotype. In each case, the Ds6 element as well as the 8-bp target site duplication is excised precisely, restoring the progenitor Sc:124 sequence in the 517 bp flanking the original site of Ds6 insertion. In contrast, in addition to the loss of Ds6, the Scp(e) derivatives contain precise deletions of Bnot and one copy of the 22-bp direct repeat flanking it. Thus, there is a one-for-one correlation between enhanced expression of the Scp(e) alleles and deletion of Bnot. Production of Scp(+) and production of Scp(e) derivatives are presumed to occur by the same mechanism, gene conversion events in which the recombinational intermediate can be repaired in two ways, one deleting only Ds6 [Scp(+)] and the other deleting Ds6 and Bnot [Scp(e)].
The structure of Scp derivative alleles. Shown are the sequences surrounding the Ds6 insertion site (A) and the position of the Bnot element (B) for two of the sc:m alleles (sc:m305 and m330), the progenitor allele (Sc:124), one Scp(+) type derivative for each sc:m [Sc:n5993 (m305) and Sc:n6732 (m330)], one Scp(e) type derivative for each sc:m [Sc:n5992 (m305) and Sc:n7053 (m330)], and P. The underlined sequences are the 8-bp target site duplication caused by the insertion of Ds6 (A) and the 22-bp direct repeat flanking the Bnot element (B). Boxes denote the entire 2.1-kbp Ds6 element or the 471-bp Bnot element.
A conversion tract of at least 200 bp is necessary to remove both Bnot and Ds6 from the sc:m. To determine if the YL1 region was included in the conversion tract, primers F4 and B2261 were used to amplify DNA from the same 11 Scp derivatives. The presence of the YL1 sequence in any of the Scp derivatives would indicate coconversions of both this region and the Bnot region during meiotic recombination. In each case, amplification of a 1.5-kbp product [Scp(+)] or a 1-kbp product [Scp(e)] indicated that YL1 was not present in the Scp alleles (not shown).
Thus, a region of the Sc gene containing Ds insertional mutants is able to undergo gene conversion with a proximal segment of homology located upstream of the promoter of the P gene from R-r:std. The positions of the Ds6 insertions and the types and numbers of Sc+ derivatives that are produced from sc:m alleles show a strong correlation (Figure 2 and Table 2). Of the 14 mapped group II sc:m alleles, the 8 sc:m for which the insertion is located within or 5′ of the Bnot element generated Sc+ derivatives of both Scp(+) and Scp(e) types (m305, m311, m316, m322, m328, m330, m333, and m341). Three insertions located 3′ of the Bnot element in a 416-bp segment that is not present in P:A (segment 4, Figure 4) produced no Sc+ derivatives (m315, m331, and m332). Three alleles that were located 3′ of the Sc:A region produced only Scp(+) derivatives (m323, m337, and m338).
DISCUSSION
The maize r locus is unique among plant genes because of the diverse array of expression patterns of variant r alleles affecting nearly all parts of the seed and plant. r allele phenotypic diversity is the result of the complex structure of r alleles, including variation among the genes in an r haplotype and regulatory complexity of single transcriptional unit r genes (Kermicle 1985, 1988; Robbinset al. 1991; Egglestonet al. 1995). Two such divergent r alleles with nonoverlapping tissue specificity are Sc (seed color) and P (plant color). The approach of this research is to use mutations of these alleles to elucidate aspects of plant intragenic recombination and gene regulation.
We have used heteroallelic combinations of P and 39 Ds6 insertion mutations (sc:m) in the maize Sc:124 allele to recover colored seed (Sc+) progeny. Three principal results emerge from these studies.
A recombinational dichotomy occurs at a position ~370–470 bp from the start of r transcription. sc:m alleles located 5′ of that position (group II alleles) produce colored seed (Sc+) variants by gene conversion. sc:m mutants located 3′ of that position (group I alleles) produce Sc+ variants at a 10-fold higher rate, mostly in association with crossing over.
Eleven of the 14 group II sc:m mutants are in a region of Sc that is not present adjacent to the P allele promoter but is located at a proximal ectopic site on the P chromosome 10. The other 3 group II mutants are located near the Sc promoter.
The conversion event between most of these sc:m and P results in two different Sc+ products, produced by deleting either one or two polymorphic sites from the sc:m (Ds6 alone or Ds6 and the Bnot element). Deletion of Bnot from the sc:m alleles produces altered tissue specificity of anthocyanin deposition, unique from either progenitor allele.
Structural explanations for the recombinational behavior of sc:m/P heterozygotes: Two discrete recombinationally based regions of Sc are demarcated at a point between the m337 and m312 insertion sites, near the start of r gene translation. sc:m/P combinations involving group I alleles result in reciprocal crossover events at rates similar to intragenic recombination frequencies for other genes in maize and with double crossover frequencies consistent with expected values (Kermicle 1970b). The total length of the region delimited by the 3′ mutants is ~26 kbp/cM (J. Kermicle, M. Alleman and C. Illingworth, unpublished data). This length is consistent with the published range of 14–50 kbp/cM for maize genic regions (Dooner and Kermicle 1986; Brown and Sundarasan 1991; Civardiet al. 1994; Dooner and Martinez-Ferez 1997; Okagaki and Weil 1997). Thus, the unusual question presented by these data is what property of r alleles Sc and P causes the nearly exclusive production of noncrossover Sc+ from the group II sc:m alleles.
One explanation for the recombinational dichotomy observed in these data is that DNA structure or sequence influences crossing over in sc:m/P heterozygotes. Possible structural parameters might include the position of insertion/deletion differences between Sc and P or the distribution of single-base-pair polymorphisms. The location of P:A is unknown but is at least 15 kbp proximal to the P structural gene. Because of this distance, P:A may behave as an ectopic segment. According to this model, recombination events involving P:A preferentially produce noncrossover progeny whereas recombination events between “homologous” regions would produce primarily crossover progeny. Ectopic recombination has been studied extensively in yeast and in mammals and is shown to proceed through normal meiotic pathways (Lichtenet al. 1987; Kupiec and Petes 1988; Nag and Petes 1990; Steeleet al. 1991; Cooperet al. 1998). Although ectopic recombination per se does not influence reciprocal crossing over in yeast or in mice, studies indicate bias favoring conversion (Jackson and Fink 1981; Peteset al. 1991).
A second structural feature of the r gene that might explain the recombinational behavior of upstream vs. downstream sc:m mutants is the distribution of small polymorphic sites in Sc vs. P. Studies in yeast using engineered restriction site polymorphisms show that sequence divergence decreases the recombination efficiency and, in many cases, skews the recovery of products toward gene conversion events (Borts and Haber 1987, 1989; Symington and Petes 1988). There is strong evidence that the mismatch repair system interferes directly with heteroduplex formation for nonidentical sequences (reviewed in Modrich and Lahue 1996; Chen and Jinks-Robertson 1999).
Neither of these models for the recombinational dichotomy between group I and group II mutants is consistent with our data. If the position of P:A were the basis for the nearly exclusive production of gene convertants from group II mutants, these structural parameters are not predicted to affect recombination that involved inserts located within or near the transcribed region of Sc. Three mutants (m323, m337, and m338) located near the Sc promoter produce primarily convertant progeny. It is, therefore, difficult to explain a recombinational dichotomy using the distribution of insertion/deletion differences between Sc and P. Similarly, single-base heterologies and insertion-deletion differences are scattered through Sc and P with no apparent positional dichotomy separating group II and group I alleles.
Transposition of Ds6 or Bnot as an explanation for production of Scp derivatives: An alternative explanation for the production of Scp alleles from 5′-located sc:m is that either Ds6 or Bnot excision occurs by transposition. Transposition of Ds6 in the sc:m/P heterozygotes is highly unlikely in the absence of autonomous element Ac. sc:m alleles are completely stable as homozygotes in the absence of the autonomous transposable element Ac, as described for Ds insertion mutants in other maize genes (McClintock 1951; Kermicleet al. 1989; Scottet al. 1996). The Scp derivatives occur exclusively as progeny of somatically stable sc:m/P heterozygotes. In addition, a Ds6 excision footprint, characteristic of typical Ds or Ac transposition events, is not present in any of the Scp alleles that were sequenced (Fedoroff 1989).
Bnot is a novel maize transposable element of the non-LTR class of retroelements but its excision by transposition does not contribute to our recombination data. Non-LTR retroelements transpose extremely rarely (Schmidt 1999). To use Bnot transposition to explain the existence of Scp(e) derivatives, transposition must occur frequently, only in sc:m/P heterozygotes, and only for alleles in the 5′ end of the Sc allele. Excision of Bnot does not occur during Ac-induced reversion of sc:m alleles, evidenced by the fact that none of 680 Sc+ revertants from sc:m, P-vv::Ac stocks had the Scp(e) phenotype of darkly pigmented scutellar node (M. Alleman, unpublished data). In addition, Bnot excision does not occur in sc:m/P heterozygotes involving group I alleles (not shown). We believe that, although Bnot is a transposable element, its transposition is rare or nonexistent and does not contribute to these recombinationally based data.
Production of Scp derivatives by heteroduplex repair during gene conversion at r: A second aspect of these results bears on the types of products obtained via recombination involving different sc:m alleles. Heteroallelic combinations of sc:m and P presumably produce Sc+ derivatives via the formation and repair of recombinational intermediates (heteroduplex) involving the two alleles. The proposed heteroduplex tract would include sequences from sc:m and either P or P:A depending on the location of the sc:m insert. sc:m mutants located within Sc:A and producing both Scp(+) and Scp(e) derivatives are informative. These sc:m (m305, m311, m316, m328, and m330) are able to generate Scp(+) or Scp(e) alleles. The two sites are located from 108 to 159 bp apart dependent on the sc:m allele used (Figure 2). Our data do not require discontinuous conversion tracts, only the formation of two types of recombinational intermediates or differential repair of polymorphic sites. It is possible that there is either incomplete repair of the heteroduplex or initiation of DSB in different positions. The reciprocal event, transfer of Ds6 and Bnot to P:A, or loss of Bnot alone, could not be identified in our screen.
Generation of gene regulatory differences by transposable element insertion, genetic recombination, and chromosomal rearrangement: These data support the evolution of maize genes by transposable element insertion, rearrangement, deletion, and genetic recombination. Most of the first 3.5 kbp 5′ of the start of transcription of Sc and P are nonhomologous, consistent with the structure of alleles of the maize b (booster-1) gene. Two well-characterized alleles, B-Peru (strong seed pigmentation) and B-I (strong leaf blade pigmentation), although homologous in coding region sequence, are highly divergent in the 5′ flanking sequences (Chandleret al. 1989; Radicellaet al. 1992).
A putative P-like progenitor would have undergone rearrangement to bring the P:A region into contact with the r gene, deletion of the YL1 region, and insertion of the Bnot transposable element during maize evolution. Because a P phenotype appears in most teosinte lines tested thus far (Hansonet al. 1996), it is presumed that a P gene represents the r gene progenitor. The insertion of Bnot into Sc is assumed to have changed the phenotype from strongly expressed in the coleoptile and scutellar node to the Sc:124 phenotype, weakly pigmented coleoptile and colorless scutellar node. This hypothesis is consistent with the structure and phenotype of another naturally occurring variant allele, the Scm gene from R-marbled (R-mb; Weyers 1961). Scm is phenotypically indistinguishable from the Scp(e) alleles and has the general structure of an Sc:A segment at the same position as for Sc:124 and containing YL1 but lacking Bnot (J. Bernot and M. Alleman, unpublished data).
In addition, the flanking region of P contains several known repetitive elements, including ones with homology to the maize retroviral element Ji-6 (SanMiguelet al. 1996). LTR and non-LTR retrotransposons are ubiquitous elements in plants, contributing strongly to the repetitive fraction of the genome (reviewed in Schmidt 1999). Retroviral-related transposons have expanded the intergenic regions and changed the sequence context of the maize gene as compared to the more compact genomes of rice and sorghum (Whiteet al. 1994; Bennetzenet al. 1998). Through the study of r and other related genes in plants, we are developing a view of gene regulatory evolution to include the opportunistic use of transposable element insertions and genomic recombination events.
Acknowledgments
We thank Dr. Timothy Robbins for the λR4.7 clone. We also thank Beverly Oashgar, David Heller, and Darla West for technical assistance. United States Department of Agriculture grant 93-37301-8878 (M.A.), National Science Foundation grant 9603747 (M.A.), and Department of Energy grant DE-FG02-86ER-13539 (J.L.K.) supported this work.
Footnotes
-
Communicating editor: J. A. Birchler
- Received February 22, 2001.
- Accepted August 19, 2001.
- Copyright © 2001 by the Genetics Society of America
