Restorer genes for different forms of Brassica cytoplasmic male sterility map to a single nuclear locus that modifies transcripts of several mitochondrial genes

Abstract

The oilseed rape plant, Brassica napus, possesses two endogenous male sterile cytoplasms, nap and pol. Previous studies have shown that nuclear restoration of pol cytoplasmic male sterility (CMS) is conditioned by a gene, Rfp, that is also involved in modifying transcripts of the pol CMS-associated orf224/atp6 mtDNA region. We now find that the nap nuclear restorer gene Rfn apparently is identical to Mmt, a gene that conditions the modification of transcripts from several different mtDNA regions, including one that is associated with nap CMS and contains orf222, a chimeric gene related to orf224. Mmt, in turn, is found to be allelic to Rfp, suggesting that restorer genes for the two cytoplasms represent different alleles or haplotypes of a single nuclear locus. This view is supported by restriction fragment length polymorphism mapping studies that indicate that Rfn and Rfp map to the same chromosomal position. Thus, in contrast to CMS in other species, different forms of Brassica CMS are restored by alleles of a single nuclear locus, and the restoration properties of these alleles reflect their involvement in the modification of transcripts of corresponding CMS-associated mtDNA regions. A survey of 51 varieties from 8 Brassica and Sinapis species failed to find evidence of Rfn(Mmt) in other than fertility-restored, nap cytoplasm B. napus. This suggests that Rfn(Mmt) arose in Brassica with nap cytoplasm and that the necessity for fertility restoration may have provided the selective pressure for its origin and maintenance.

The control of the expression of individual mitochondrial genes by specific nuclear genes represents a key mechanism for ensuring the cooperative function of the nuclear and mitochondrial genomes (1, 2). In flowering plants, the suppression of cytoplasmic male sterility (CMS) by nuclear restorer of fertility (Rf) genes represents a striking example of this type of nuclear–mitochondrial gene interaction. CMS is a maternally transmitted failure in pollen production. In several plant species, mitochondrial gene regions have been identified whose expression is associated with CMS. These regions contain unusual ORFs that often are cotranscribed with conventional mitochondrial genes (3). In general, the nuclear restorer genes that suppress CMS specifically modify expression of the CMS-associated regions, but not other mitochondrial genes.

In certain plant species, multiple forms of CMS are found. These forms can be distinguished by their associated novel mitochondrial ORFs and by the nuclear Rf genes that restore their fertility (3, 4). In such cases, the restorers for the different forms represent distinct genes that map to different chromosomal loci. In maize, for example, there are three forms of CMS: T, S, and C. Restoration of cms-T requires two genes, Rf1 and Rf2, that map to chromosomes 3 and 9, respectively (5, 6), whereas cms-S is restored by Rf3 on chromosome 3 (7, 8) and cms-C is restored by Rf4 on chromosome 8 (9). In rice, as well, restorer genes for three different forms of CMS have been found to map to different chromosomes (10). Even when linkage is observed between restorers for different forms of CMS, on close examination these have been found to segregate as distinct loci, as in the case of the wheat Rfv1 and Rf3 genes (11). Although a single type of CMS may be restored by genes mapping to different chromosomal positions (7, 12), the converse situation, in which more than one type of CMS can be restored by a gene or genes present at a single locus, has not been shown to occur.

Two forms of CMS, designated nap and pol, are endogenous to the oilseed rape plant, Brassica napus (13). Most B. napus varieties contain the nap cytoplasm but are male-fertile because they possess a restorer gene for nap CMS, designated here as Rfn (14). The nap cytoplasm confers male sterility on a few exceptional varieties that lack Rfn, such as the cultivar “Bronowski” (13, 15). The male fertile “maintainer” strains of these varieties contain the fertile cam cytoplasm derived from the related species Brassica campestris (13, 16). Most B. napus varieties lack a restorer gene for pol CMS and, hence, are sterilized by pol cytoplasm (14). Restorer genes for pol CMS have been identified in various strains (14, 17) and map to a single nuclear locus designated Rfp (18). The relationship between the various B. napus cytoplasms, restorer genes, and male sterility or fertility are outlined in Table 1.

Table 1.

Nuclear–cytoplasmic interactions in Brassica napus

Cytoplasm	Restorer genotype and fertility status
	rfn, rfp	Rfn, rfp	rfn, Rfp
cam	Fertile	Fertile	Fertile
nap	Sterile	Fertile	Sterile
pol	Sterile	Sterile	Fertile

Analysis of mtDNA organization and expression in fertile, sterile, and nuclear restored lines has indicated that the pol CMS is likely to be specified by the atp6 gene region (16, 19–21). In pol mtDNA, atp6 is cotranscribed with a chimeric ORF, orf224. The Rfp restorer gene acts in a dominant manner to modify transcripts of the region (19, 20, 22). In pol CMS plants, dicistronic orf224/atp6 transcripts predominate; in the presence of Rfp, monocistronic atp6 transcripts predominate. Similar analyses have shown that nap CMS is correlated with expression of an mtDNA region containing a different chimeric ORF, orf222, that is cotranscribed with an exon of a trans-spliced gene, nad5c, and another ORF, orf139 (23). Nuclear restoration of nap CMS has both quantitative and qualitative effects on orf222/nad5c/orf139 transcripts. Unlike the ORFs associated with different forms of CMS in other plant species, orf222 and orf224 are highly similar in sequence over their entire length.

A survey of the transcripts detected by cloned DNAs comprising approximately 90% of the Brassica mitochondrial genome indicated that, as with all other nuclear restorer genes identified thus far, the effects of Rfp on mitochondrial gene expression appear to be restricted to a single mtDNA region (22). However, the recessive allele of this gene (rfp) is linked to a gene, Mmt (modifier of mitochondrial transcripts), that affects the transcripts of a pseudogene (ccl1-l; refs. 22 and 24) and the nad4 gene (22). We now find Mmt (rfp) to be indistinguishable from a gene that is responsible for both the modification of orf222/nad5c/orf139 transcripts and for nuclear restoration of nap CMS. This indicates that Rfn and Rfp are different alleles or haplotypes of a single, possibly complex, nuclear genetic locus. Mapping analysis using nuclear DNA markers linked to Rfp (18) is consistent with this possibility. In a survey of 50 different varieties of B. napus and related species, Mmt (Rfn) was found only in fertile lines with nap cytoplasm, suggesting that the sterile nap cytoplasm has provided the selection pressure for the origin and/or maintenance of the Mmt(Rfn) gene in B. napus.

MATERIALS AND METHODS

Plant Material.

Fifty strains of nine different crucifer species (the three italic letters in parentheses indicate the cytoplasm) were used in the present study (sources are available on request). Of the B. napus strains, the nap CMS and maintainer strains Bronowski (nap) and Bronowski (cam), the pol CMS strains Karat (pol) and Westar (pol), the pol fertility-restored strains Italy (pol), UM2353 (pol), and Westar-Rf (pol), and the male fertile strains Karat (nap), Regent (nap), and Westar (nap) have all been described previously (16, 18, 19, 22). The additional male fertile B. napus strains analyzed for nad4 transcripts, all of which possess the nap cytoplasm, were: AC Elect, Amazon, Arctik, Cyclone, Defender, Ebony, Falcon, GrGc 5–1, GrGc 5–2, Korell, LG3310, Pearl, and Topas. The male fertile B. campestris (synonymous with B. rapa) lines analyzed, all of which possess the cam cytoplasm, were: AC Parkland, AC Sunshine, Aiyou, Cash, CV2, Eldorodo, F6–15-8810, Goldrush, GrGc 1–9, Hysyn 110, Klondike, Reward, and SRS 753. The B. oleracea (ole) strains analyzed were: GrGc 3–8Y (fertile), GrGc 3–8W (nuclear male sterile), green cabbage, red cabbage, and broccoli. We also analyzed the B. carinata (car, fertile) strains Dodolla and GrGc 6–1, the B. juncea (jun, fertile) strains AC Vulcan, Commercial Brown, GrGc4–1, and IM41, the B. nigra (nig, fertile) strains AC Type 1 and GrGc 2–1, the Sinapis alba (sin, fertile) strain AC Pennant, and the S. arvensis (arv, fertile) strain SRS35.

Fertility Assessment.

Plants were grown to maturity in the McGill University Phytotron under conditions of a 16-hr photoperiod and day/night temperatures of 20/15°C. Fertility was assessed by observing five flowers per plant, at least two times during the flowering period. The overall morphology of the flowers was noted as well as the production of pollen on anthers. Both pol and nap CMS anthers have shorter filaments and poorly developed or absent pollen sacs and produce little or no pollen. In addition, flowers of pol CMS often have dramatically shrunken petals. The morphological contrast between CMS and fertile flowers was great enough that, for all the populations examined, CMS and fertility-restored plants could be discriminated without ambiguity.

RNA Purification and Analysis.

Mitochondrial RNA was isolated from flowers/inflorescences as described (23) except the LiCl precipitation step was omitted. RNAs were size-fragmented on agarose-urea gels, transferred to Gene Screen-Plus (DuPont) hybridization membranes in 10× SSC (1.5 M NaCl/0.15 M sodium citrate), and fixed by UV cross-linking. DNA probe labeling, RNA blot hybridization, and washes were performed as described (25). The clones used as probes were the from the coding regions of the B. napus atp6 (19), nad4 exon 2 (22), and orf139 (23) genes of the mitochondrial genome of B. campestris.

DNA Purification and Restriction Fragment Length Polymorphism (RFLP) Analysis.

DNA extraction, RFLP analysis, and cosegregation analysis by the mapmaker program were as described (18).

RESULTS

A Brassica Nuclear Genotype That Fails to Restore nap CMS Lacks Mmt and Rfp.

A survey of B. napus varieties indicated that a nap CMS line and its (nuclear) isogenic maintainer strain, the cultivar Bronowski, both of which lack the Rfn gene, did not possess the nad4 and ccl1-l transcript modifications conditioned by the Mmt gene. Because these modifications were found previously to be absent only in lines homozygous for pol restorer genes, we checked the ability of the nap CMS and maintainer lines to restore male fertility in the F1 in crosses with pol CMS lines. No fertility restoration was observed, and, hence, the genotype of these varieties is mmt, rfp/mmt, rfp, a configuration of genes in the Rfp-Mmt chromosomal region that had not been observed previously. Because these lines also lack the nap restorer gene Rfn, they possess recessive, maintainer alleles for both the nap and pol CMS systems. Therefore, we chose to designate their restorer genotype simply as rf/rf.

Allelism of rfp and Mmt.

nap maintainer genotypes, by definition, lack Rfn. Because Mmt and rfp are linked and because the nap maintainer lacked both Mmt and Rfp, it seemed possible that the three genes, Mmt, Rfp and Rfn, might all reside at the same genetic locus. To test this possibility we examined allelism of the three genes through the analysis of two different types of genetic populations. We first examined the allelism of rfp and Mmt through the analysis of a test-cross involving the double-recessive genotype mmt, rfp/mmt, rfp (cross I of Table 2). This provides a much more powerful allelism test than the F2 populations we analyzed previously (22). To generate this population, a Westar pol CMS plant (rfp, Mmt/rfp, Mmt) first was fertilized with pollen from the near isogenic pol fertility restorer line Westar-Rf (Rfp, mmt/Rfp, mmt). We recovered fertile individuals that contained pol cytoplasm and were highly homozygous except at the Rfp-Mmt locus.

Table 2.

Genetic crosses used to assess allelism relationships among rfp, Rfn, and Mmt and the map position of Rfn

Cross	No. of plants		Ratio	Mmt-specific 1.6-kb nad4 transcript detected in
	Fertile	Sterile		Fertile, %	Sterile, %
pol cytoplasm
I. Westar (pol) Rfp, mmt/rfp, Mmt × Bronowski (cam) rf, mmt/rf, mmt	39	36	1:1	0	100
Nap cytoplasm
II. nap CMS* × Westar (pol) Rfp, mmt/rfp, Mmt	4	3	1:1	100	0
III. BC1: nap CMS × a fertile progeny from cross II	16	14	1:1	100	0
IV. BC2: nap CMS × a fertile progeny from cross III	17	17	1:1	100	0
V. BC2: nap CMS × [nap CMS × (nap CMS × Karat†)]	4	5	1:1	100	0
VI. BC1: nap CMS × (nap CMS × Karat)	18	10	1:1‡	ND	ND

ND, not determined. 

^*nap CMS: male sterile Brownowski (nap) rf, mmt/rf, mmt. 

^†Genotype of the variety Karat: Rfn, rfp, Mmt/Rfn, rfp, Mmt (nap cytoplasm). 

^‡Deviance from the expected 1:1 ratio is not statistically significant (χ² = 1.725). 

One such plant then was test-crossed as a female with the male fertile nap maintainer line Bronowski (rf, mmt/rf, mmt). The resulting test-cross population was expected to contain approximately equal numbers of sterile (rfp/rf) and fertile (Rfp/rf) pol cytoplasm individuals. If Mmt and rfp act as the same allele, all fertile progeny would be expected to lack the 1.6-kb nad4 transcript that defines the presence of the Mmt allele, and to possess the 1.4- and 1.3-kb orf224/atp6 Rfp-specific transcripts. In contrast, all sterile progeny should possess the 1.6-kb nad4 transcript and lack the 1.4- and 1.3-kb orf224/atp6 transcripts. The results obtained were entirely consistent with these predictions. Thirty-nine of the 75 test-cross progeny obtained were fertile, and all of these lacked the transcript, whereas 36 were sterile and possessed this transcript (cross I, Table 2). An example of the transcript analysis data is shown in the top two panels of Fig. 1. From these results we estimate that, at a 95% confidence level, rfp and Mmt map no more than 3.6 cM from one another. Thus, rfp and Mmt appear to represent the same allele or haplotype of a single restorer locus.

Figure 1.

RNA gel blot analysis of mitochondrial transcripts in test-cross or backcross populations segregating for male sterility and sterility restoration. For the top two panels, mtRNA samples from cross I progeny (Table 2) were probed with the Brassica atp6 (top panel) or nad4 (second panel) coding sequences to follow the segregation of Rfp and Mmt, respectively. For the bottom two panels, mtRNA samples from cross III progeny (Table 2) were probed with the nad4 (top panel) or orf139 (bottom panel) coding sequences. The sizes of transcripts specific to Rfp (mmt) plants (top panel) or Rfn (Mmt) plants (bottom three panels) are indicated to the right of the autoradiograms.

Allelism of Rfn, Mmt, and rfp.

To test for allelism between all three genes, it was necessary to generate nap cytoplasm populations in which the segregation of both Rfn and Mmt (rfp) could be observed. To achieve this, we first crossed a Westar (Rfp/rfp) individual as male to a nap CMS plant possessing the Bronowski nuclear genotype (rf, mmt/rf, mmt) and nap cytoplasm (cross II of Table 2). BC₁ and BC₂ nap cytoplasm populations then were generated by backcrossing fertile progeny as males with the nap CMS line (crosses III and IV of Table 2). If Mmt (rfp) were to act as Rfn and map to the Rfn locus, we would expect to obtain only two progeny classes in these populations and to obtain these in approximately equal numbers: fertile individuals heterozygous for both Rfn and Mmt and sterile individuals homozygous for rf and lacking Mmt. The results were completely consistent with these predictions: all fertile progeny possessed the 1.6-kb nad4 transcript indicative of Mmt, whereas all the sterile progeny lacked this transcript. Partial transcript analysis data are shown in the third panel of Fig. 1. Similar results were obtained upon analysis of a second BC2 population in which a different B. napus variety, “Karat,” was used as the source of Rfn (Mmt) (cross V, Table 2). Perfect cosegregation of Rfn and Mmt was observed in a total of 80 individual progeny of these crosses. Thus, within the limits of this analysis, Rfn is indistinguishable from Mmt.‖ Because the analysis of cross I indicated that rfp is indistinguishable from Mmt, the restorer genes for the nap and pol CMS systems map to a single nuclear genetic locus. The nap maintainer genotype Bronowski lacks Rfp, Rfn, and Mmt and, hence, possesses a third allele of this locus, rf. The locus thus has three alleles or haplotypes: Rfp (mmt), Rfn (Mmt), and rf (mmt).

Rfp and Rfn Map to the Same Site on the B. napus Genome.

Genetic mapping studies have provided further evidence that Rfn and Rfp map to the same locus. RFLP linkage analysis has shown that Rfp maps to B. napus linkage group 18 (18) [see Landry et al. (26)], between the markers 4ND7b and 4NB6. These mapping studies employed a pol cytoplasm BC1 population (termed the KW population) derived by crossing a pol CMS female with a restorer line to generate a fertile F₁ individual that was then used to pollinate the corresponding CMS parent. A similar strategy was used to map Rfn. The variety Karat (rfp, Rfn/rfp, Rfn), which served as the genotype of the CMS parent in the Rfp mapping cross, was used as the source of Rfn. Karat (nap) was crossed as male to the Bronowski nap CMS line, and the resulting F₁ was used to pollinate the nap CMS parent to generate a nap cytoplasm BC1 population segregating for Rfn (cross VI, Table 2). Eighteen male fertile and 10 male sterile plants were recovered; these progeny numbers fall within the statistically acceptable limits of the expected 1:1 pattern (χ² = 1.725).

If Rfp and Rfn are indeed allelic, they will map to the same nuclear chromosomal locus. Because these genes are expected to be in repulsion, the polymorphic fragment detected by a linked RFLP probe that segregates with male fertility in pol cytoplasm crosses should be different from the fragment detected by the same probe that segregates with fertility in nap cytoplasm crosses. Thirteen linkage group 18 markers were tested on the cross VI population. Four of these proved to be monomorphic and could not be mapped. The map of Rfn and the nine remaining markers, all of which segregated in a 1:1 ratio, showed that Rfn occupies the same position on linkage group 18 as Rfp. The location of Rfn with respect to the six most closely flanking markers is shown in Fig. 2A Right. The position of Rfp with respect to the same set of markers, as derived from the analysis of the KW BC₁ pol cytoplasm population of Jean et al. (18), is shown in Fig. 2 A Left.

Figure 2.

RFLP mapping of Rfp and Rfn. (A) Maps of linkage group 18 in the vicinity of Rfp and Rfn constructed from backcross populations (described in B, below), which allowed segregation of Rfp (the pol cross, Left) and Rfn (the nap cross, Right) to be followed; note the similar map locations of the two restorer genes. The marker 5NE12b was not polymorphic in the nap cross and could not be mapped in this population. (B) A single RFLP marker detects alleles associated with both Rfp and Rfn. The RFLP marker 3NF2 (18) was used to probe EcoRV digests of genomic DNA from sterile (S) and fertile (F) individuals of two B. napus backcross populations. The pol cross allowed segregation of the pol restorer Rfp to be followed. It was generated first by crossing a Karat (pol) CMS plant (lane 1) with Westar-Rf (lane 2) and then crossing an F₁ individual with the Karat (pol) parent; this corresponds to the KW population of Jean et al. (18). The nap cross (cross VI of Table 2) allowed segregation of the nap restorer Rfn to be followed. The genotype Karat (lane 4), which served as the CMS parent in the pol cross, was used as the source of the Rfn gene in a cross to male sterile Bronowski (nap) (lane 3). Note that the 10.0-kb fragment specific to Westar-Rf segregates with male fertility restoration (F) in the pol cross, where Karat was used as the CMS parent, whereas the 6.7-kb fragment specific to Karat segregates with male fertility restoration in the nap cross, where Karat was used as the fertility restored parent. Because 3NF2 maps approximately 10 cM from Rfp and Rfn, segregation with male sterility in both cases is incomplete.

Fig. 2B shows the segregation of RFLPs detected by the marker 3NF2, which was mapped at 12.7 cM from Rfp in the KW BC1 population (18), in selected progeny from the pol and nap cytoplasm crosses used to map Rfp and Rfn, respectively. The 3NF2 probe detects polymorphic EcoRV fragments of 6.7, 10.0, and 10.2 kb specific to the nuclear genotypes of Karat, Westar-Rf, and Bronowski, respectively. Fig. 2 B Upper shows that in the pol cross used to map Rfp, where the genotype Karat was used as the CMS (rfp/rfp) parent, the 10.0-kb Westar-Rf-specific 3NF2 EcoRV fragment segregates with male fertility. In contrast, Fig. 2 B Lower shows that in the nap cross, where Karat was used to provide the Rfn restorer allele, the Karat-specific, 6.7-kb 3NF2 EcoRV fragment segregates with male fertility. This is consistent with the prediction that a single, linked probe will detect different polymorphic fragments segregating with male fertility in the two mapping populations. The RFLP mapping data therefore provide strong independent support for the view that Rfn and Rfp represent alternative alleles or haplotypes of a single nuclear locus.

Rfn (Mmt) Conditions Modification of Transcripts of the nap CMS-Associated Mitochondrial Gene Region.

Although previous studies have shown that transcripts of the nap CMS-associated orf222/nad5c/orf139 mtDNA region differ between a restored and a nap CMS variety (23), it is not known whether these transcript differences result from the presence of the restorer gene per se. The availability of the cross II, III, IV, and V populations allowed us to assess the coinheritance of orf222/nad5c/orf139 transcript modifications with Rfn-induced fertility restoration. orf139-specific probes detect a 0.8-kb transcript in fertility-restored nap cytoplasm plants that is not detected in nap CMS plants (23). This transcript was found in all the fertile progeny and in none of the sterile progeny of these crosses, indicating that Rfn (Mmt) does indeed condition the modification of transcripts of the nap CMS-associated mtDNA region. A representative portion of this analysis is shown in the bottom panel of Fig. 1. These results indicate that not only do different alleles of a single nuclear locus restore fertility to different male sterile cytoplasms but also that the same alleles of this locus are responsible for the modification of transcripts of the respective CMS-associated mitochondrial gene regions: Rfp, rfn, mmt for the pol mtDNA-specific orf224/atp6 region, and rfp, Rfn, Mmt for the nap mtDNA-specific orf222/nad5c/orf139 region.

Rfn (Mmt) Is Not a Common Gene in Most Brassica Species.

It is possible that new restorer genes are generated in response to selective pressure created by the spread of a male sterile cytoplasm in a plant population (27). If this were the case, a particular restorer gene might not be widely distributed in varieties and species in which the male sterile cytoplasm does not occur. Genetic testing for the presence of a particular restorer gene is time consuming and limited to the species in which the male sterile cytoplasm is available. With the exception of Rfn, the effects of restorer genes on mitochondrial gene expression thus far have been observed to be limited to male sterile cytoplasm-specific, CMS-associated mitochondrial gene regions. However, the apparent identity of Rfn and Mmt allowed us to easily test for the presence of Rfn in a relatively large number of strains by examining the transcripts of the nad4 gene, which is not associated with CMS, for Mmt-specific modifications.

We used this strategy to survey a total of 50 different crucifer varieties for the presence of the Mmt(Rfn) (see Materials and Methods). These included 16 fertile and 1 sterile B. napus nap cytoplasm varieties, 3 fertile and 2 sterile B. napus pol cytoplasm varieties, 1 fertile B. napus cam cytoplasm variety, 1 male sterile and 4 male fertile B. oleracea varieties, and male fertile B. carinata (2 varieties), B. juncea (4 varieties), B. nigra (2 varieties), B. rapa (13 varieties), Sinapis alba (1 variety), and Sinapis arvensis (1 variety). Only the 16 male fertile lines of B. napus with nap cytoplasm possessed the additional Mmt(Rfn)-specific transcript (not shown). The data indicate that, apart from fertility-restored nap cytoplasm B. napus lines (the majority of B. napus varieties), Mmt(Rfn) is not widely distributed in Brassica and related genera.

DISCUSSION

Our results identify three different forms of a single Brassica napus nuclear genetic locus that we now designate as Rf-Mmt. One form, Rfp(mmt), restores male fertility to plants with pol cytoplasm and modifies transcripts of the pol CMS-associated orf224/atp6 gene region. A second form, Rfn(Mmt), restores fertility to nap cytoplasm plants and modifies transcripts of the nad4 gene, the ccl1-l pseudogene, and the nap CMS-associated orf222/nad5c/orf139 gene region. A third form, rf(mmt), is unable to restore fertility to either pol or nap cytoplasm plants and apparently is unable to condition any mitochondrial transcript alterations. Rfp(mmt) is dominant to Rfn with respect to pol restoration and modification of orf224/atp6 transcripts, and Rfn(Mmt) is dominant to Rfp(mmt) with respect to nap restoration and modification of orf222/nad5c/orf139, ccl1-l and nad4 transcripts.

In other plant species where more than one form of CMS is found, restorer genes for the different systems map to different nuclear genetic loci, and, in those cases where the locus is known to have effects on mitochondrial gene expression, these effects have been observed only on corresponding CMS-associated mitochondrial gene region. The Rf-Mmt locus therefore is novel in two respects: different forms of this single locus represent restorer genes for two distinct CMS systems, and one form of the locus, Rfn(Mmt), affects the transcripts of three different mitochondrial gene regions.

We have suggested previously that the transcript modifications conditioned by Rfp and Mmt may result from the selective destabilization of the 5′ termini of specific mitochondrial transcripts, and we proposed a model in which the specificity of these transcript modifications is conferred by the capacity of the corresponding gene products to recognize different but related hexanucleotide motifs (22). However, a sequence resembling the Rfn(Mmt) motif is not found in the orf222/nad5c/orf139 region, and the effects Rfn has on this region are qualitatively quite different from the effects Rfp has on transcripts of the orf224/atp6 region. Although it seems likely that Rfn(Mmt) acts to modify orf222/nad5c/orf139 by mediating specific RNA processing events, the mechanism and/or recognition processes through which the these transcript modifications take place therefore may be more complex than was suggested initially.

The finding that Rfn(Mmt) is, within the limits of the varieties analyzed, found only in association with the nap cytoplasm suggests that the evolutionary appearance of the nap cytoplasm and the attending male sterility may have provided the selective pressure for the origin, and possibly the continued presence, of Rfn (Mmt) in B. napus. Similarly, Rfp(mmt) could have arisen as a variant of the same locus, with a slightly different mtRNA-processing specificity, in response to the appearance of pol cytoplasm. One or the other of these genes could have evolved as a neomorph or as a variant of yet another gene capable of influencing mitochondrial RNA-processing events.

It is possible that Rfp(mmt) and Rfn(Mmt) simply represent different alleles of a single gene. It also is possible, because of its multiple associated RNA processing and nuclear restoration properties, that the Rf-Mmt locus may be more complex and contain multiple, related, tightly linked genes. In this respect, it could resemble certain complex plant disease resistance loci for which the capacity to respond to a particular pathogen or chemical stimulus is conferred through a particular configuration of individual genes encoding related proteins (28, 29). If so, the evolution of new restorer genes in response to the appearance of new male sterile cytoplasms might occur via recombination within the locus, in the same manner that new resistance specificities have been proposed to arise in response to the appearance of new races of a pathogen (29, 30). The isolation and characterization of the various forms of Rf-Mmt therefore may provide insight not only into the mechanistic basis of fertility restoration, but also into the mechanisms by which these genes have evolved.

ABBREVIATIONS

CMS

cytoplasmic male sterility

RFLP

restriction fragment length polymorphism

cM

centimorgans