1. Introduction
The cytoplasm of
Ae. kotschyi is one of several cytoplasms among the wild relatives of wheat (Triticum sp.) capable of inducing cytoplasmic male sterility (cms) in wheat, both hexaploid and tetraploid [
1,
2,
3,
4,
5,
6]. Male sterility manifests itself only in the absence of the chromosome arm 1BS, such as in the wheat-rye translocation 1RS.1BL. So far, only one chromosome 1B of wheat tested, that of
Triticum spelta var. duhamelianum, was found to be incapable of restoring fertility; all other chromosomes 1B are capable of restoring fertility [
3,
4]. As this locus on chromosome arm 1BS is capable of restoring male fertility to several different cytoplasms, it was named
Rfmulti by Tsunewaki [
7].
Cms based on the
Ae. kotschyi cytoplasm would be of considerable value in wheat breeding, as under standard growing conditions and with the standard chromosome constitution of the nuclear genome, it does not appear to cause undesirable agronomic effects, and there are some indications that it may improve resistance to abiotic stresses [
8]. The ubiquitous presence of a fertility restorer on 1BS simplifies fertility restoration in hybrids. However, fertility restoration by a single copy of
Rfmulti reaches only about 2/3 of a full seed set. Full fertility is restored only by two copies of the locus. Hence, additional selection/manipulation would be required for effective use of the system in hybrid breeding. One of the approaches was a transfer of
Rfmulti from its normal position on chromosome arm 1BS to rye chromosome arm 1RS and the translocation of that engineered 1RS arm to chromosome arms 1AL and 1DL [
9]. Hybrids of male sterile wheats with wheats carrying such translocations are fully fertile as they carry two doses of
Rfmulti: one in its natural position on 1BS and the other on the 1RS.1AL or 1RS.1DL translocation. On the other hand, the ubiquitous
Rfmulti was removed from 1BS by chromosome engineering, creating three chromosomes: 1B
1:6, 1B
25:6, and 1B
35:6 [
9,
10], to be used in the cms maintainer lines. Despite the absence of
Rfmulti, one of these engineered chromosomes started showing a low level of fertility restoration, suggesting that more than a single locus on 1BS may be involved.
With the chromosome constitution of the nuclear genome deviating from normal, the cytoplasm of Ae. kotschyi in wheat shows several additional effects pointing to a range of nuclear–mitochondrial interactions. One such effect is the rate of growth and grain size in the absence of the chromosome arm 1S in the B genome (either 1BS or 1RS). Such plants are always small and grow much more slowly than their sisters with 1BS present. They produce very small grains, as small as 1/10 of the normal test weight, and germinate poorly. Chromosome arm 1RS, in the wheat rye translocation 1RS.1BL, fully compensates the absence of 1BS. As Rfmulti is absent from 1RS, some rye genes on 1RS must interact with the mitochondrial genome of Ae. kotschyi. The absence of other group 1S arms, such as 1AS or 1DS, has no discernible effect on the growth rate or seed size.
Even more interestingly, the cytoplasm of
Ae. kotschyi is one of several cytoplasms among wheat relatives capable of affecting various aspects of double fertilization in wheat, resulting in the production of haploids (parthenogenesis), twin seedlings (mostly haplo-diplo), and embryo-less kernels [
1,
11]. This occurs when the proper inducer is present in the genome of an alloplasmic wheat line. The first such inducer discovered was a centric wheat–rye chromosome translocation 1RS.1BL in a wheat line ‘Salmon’, a hexaploid derivative of an octoploid triticale [
12]. The haploid frequency induced by the Salmon source can exceed 90% [
4].
Haploids are very interesting and useful organisms. Upon chromosome doubling, they produce homozygous doubled haploids. These are of considerable value in genetic research, especially for mapping quantitative trait loci (QTL) in biparental populations, and in practical breeding, as they eliminate the long process of genotype stabilization after the initial hybridization, hence the speeding up of the breeding process [
13]. For these reasons, numerous approaches have been invented and improved to produce haploids on a large scale, mostly focused on the induction of microspores into the sporophytic pathway of development [
14]. Genetic means of haploid production would be even more beneficial, as it would eliminate the need for tissue/cell culture. The combination of the
Ae. kotschyi cytoplasm + 1RS inducer in wheat generates normally developed grain with normal germination rate but with some proportion of haploids or twin seedlings, of which one is usually haploid; a small proportion of seed is germless. For these three reasons: male sterility and fertility restoration, the effect on growth rate, and haploid production, this system attracted the attention of the authors, with the idea of identifying the loci responsible for the three phenomena, but specifically the 1RS locus responsible for the induction of parthenogenesis.
Genetic mapping to locate a locus of interest requires polymorphism (allelic variation). Apart from the single case of non-restoring 1BS arm of wheat described by Mukai and Tsunewaki [
3], no polymorphism for the
Rfmulti appears to be present. The locus was located on the 1BS by Tsunewaki [
7] using a set of 1RS-1BS recombinant lines, later re-located, re-mapped, and renamed by Chen [
15], and sequenced [
16]. The availability of the wheat IWGSC reference genome, pan-genome data sets, and the reference genome of rye
S. cereale “Lo7” enabled studies of the genomic interval carrying
Rfmulti in wheat and the corresponding non-restoring region in rye [
16,
17,
18]. These comparative studies pointed to several candidate genes in the region, including
TraesCS1B02G071642.1 [
16,
17] and TraesCS1B01G072300 [
18]. Both candidate genes have a strong prediction for mitochondrial localization and belong to the P subclass of the pentatricopeptide repeat protein family, to which the majority of identified restorer genes in crops belong [
19]. The candidate gene mapped by Chen [
15] on 1BS in wheat and named Rfk1 appeared to be
TraesCS1B02G197400LC. Later, the same authors proposed the same gene as a candidate for Rfd1 restorer gene for the
Ae. juvenalis cytoplasm [
20]. The
TraesCS1B02G197400LC gene encodes a pectin esterase inhibitor located to the cell wall. How a cell wall-localized pectin inhibitor protein interacts with the cms-causing genes encoded in the mitochondrial genomes remains to be studied.
To the best knowledge of the authors, no attempt was ever made to locate genetic loci on 1BS associated with the growth rate of alloplasmic wheat lines with the cytoplasm of Ae. kotschyi. As 1RS fully compensates for the absence of 1BS, recombinants 1RS-1BS are useless for this purpose and this aspect of the study remains unattended. This left the parthenogenesis-inducing factor(s) on 1RS as the primary focus. In search of allelic variants, a set of 1RS chromosome arms originating from various sources, all in a uniform genetic background of cv. ‘Pavon 76’, were crossed and backcrossed to an alloplasmic line of cv. Pavon with the cytoplasm of Ae. kotschyi. No haploids were ever detected, either in 1RS disomics in or heterozygotes with 1BS. At the same time, by fortuitous coincidence, high frequencies of haploids were detected in an alloplasmic line of triticale cv. (kot)Presto with the 1D(1B) substitution and in an alloplasmic line of a winter wheat cv. (kot)Joker. Cv. ‘Joker’ carries the wheat-rye translocation 1RS.1BL which does not appear to be related to the most common version of the translocation from cvs. ‘Aurora’ and ‘Kavkaz’ or that of Salmon (F.J. Zeller, personal communication).
When chromosome arm 1RS from cv. Presto, already present in cv. Pavon, both as a substitution 1R(1B) or translocation 1RS.1BL, was tested in the cytoplasm of
Ae. kotschyi, no haploids were detected. Suspecting an error during line development, chromosome 1R from Presto 1D(1B) and translocation 1RS.1BL from cv. Joker were transferred into (
kot)Pavon by backcrosses, and again, no haploids were detected. This indicated that cv. Pavon carries a suppressor or suppressors of parthenogenesis somewhere in the genome, but not on chromosome arm 1BS. Both the 1R(1B) substitution and 1RS.1BL translocation remove the 1BS arm. Mukai [
11] indicated that chromosome arm 1BS of wheat carries such a suppressor; it appeared to operate gametophytically (no 1B carrying eggs produced haploids) and did not completely suppress haploid production by a heterozygote 1BS + 1RS.1BL. Data by Mukai [
11] also hint at a non-1BS suppressor or suppressors: cvs. Aurora and Kavkaz differed dramatically in haploid production, with the former exceeding 90% when homozygous. Both of these cultivars carry 1RS.1BL translocation, which, by all indications, originated from the same source [
21] and appear to be identical [
22], suggesting that the absence of haploids in (
kot)Kavkaz was not caused by the absence of the inducer (on 1RS), but by the presence of some other factor suppressing their production. The presence of a parthenogenesis suppressor in cv. ‘Chinese Spring’ or an enhancer in Salmon was speculated on by Tsunewaki and Mukai [
4].
Taken together, these observations raise questions about the nature of haploid induction by the 1RS arm in the
Ae. kotschyi cytoplasm: Are inducers, such as those in Salmon, Joker, and Presto, indeed rare, or are non-1BS suppressors of haploid production in the wheat genome common? Zeller and Hsam [
21] state but provide no direct evidence or reference that “most European 1B/1R wheat cultivars transferred into the cytoplasm of
Ae. kotschyi are also able to produce haploids”. The 1B/1R cultivars in the cited segment reflect the old notation of either 1R(1B) substitution or 1RS.1BL translocation.
This study was undertaken to address the issues of fertility restoration to the Ae. kotschyi cytoplasm, including a search for allelic variants of Rfmulti, to identify and locate in the genome of wheat a suppressor or suppressors of haploid production, and to establish a haplotype/haplotypes associated with the absence of suppressors, so that allelic variation of the haploidy inducers can be tested and their genetic location determined.
4. Discussion
The
Rfmulti locus has been located [
7] and identified [
31], the candidate genes have been proposed [
16,
17,
18] and there is little doubt that it is the main restorer of male fertility in the cytoplasm of
Ae. kotschyi in wheat. However, the strange behavior of chromosome 1B
1:6 raises questions about possible other fertility restoration factors that may be present on the chromosome arm. Numerous DNA sequences encoding
Restorer of Fertility-like proteins are present in the vicinity of the
Rfmulti gene on the short arm of chromosome 1BS in wheat [
16,
32]. Perhaps some of these are capable of restoring a low level of male fertility. The three engineered chromosomes of 1B, 1B
1:6, 1B
25:6, and 1B
35:6, share the same distal breakpoint from primary recombinant T-6 [
9,
10]; they differ by the proximal breakpoints. The rye segment removing
Rfmulti in 1B
1:6 is the longest [
8]; hence, it removes the largest corresponding wheat segment. The other two engineered chromosomes, 1B
25:6 and 1B
35:6, with shorter wheat segments removed, have never shown any anther dehiscence or seed set, while 1B
1:6 shows it in some seasons, usually during the fall when greenhouse conditions are milder (flowering is usually in late November to mid-December). If indeed some additional
Rf-like DNA sequences in the vicinity of the
Rfmulti locus are responsible for the low level of fertility restoration, all three engineered chromosomes, as well as some recombinants in the T- configuration with breakpoints distal to the main
Rfmulti locus, should show the same behavior, but they do not. There are
Rf -like sequences in the rye 1RS arm [
18] but no anther dehiscence or seed set at any level was ever observed in (
kot)Pavon 1RS.1BL over at least ten generations grown in various seasons. A speculative explanation is that perhaps some
Rf-like motifs present in 1RS are incapable of restoring even a trace of male fertility on their own, but when combined with some motifs on 1BS, by a fortuitous crossover event, which creates the 1B+1 breakpoint, they generate detectable anther dehiscence and a low seed set when conditions are right.
Observations presented here show clearly that the Rfmulti locus affects more than male fertility and that more than a single gene must be involved. In the absence of a chromosome arm in the 1S position of the B genome (that is, either 1BS or 1RS), the growth rate is seriously reduced, starting as early as the first 24 h of new seed development and continuing through plant maturity. The Rfmulti cannot be directly involved as it is absent from standard 1RS and from engineered 1B chromosomes. All of these chromosomes, even when present in a single dose, produce normally vigorous plants with fully developed grain upon cross-pollination. Either some other loci present on both 1BS and 1RS generate normal vigor or the Rfmulti region, both on 1BS and 1RS but not Rfmulti itself, are involved in interactions with the mitochondrial genome.
The
Rfmulti locus has a major effect on the pollen transmission rates of chromosomes 1B and 1RS.1BL, or, rather, on the competitive advantage/disadvantage of pollen grains carrying one of these chromosomes, favoring pollen with the locus present. It is possible that different transmission rates are a consequence of differences in pollen tube growth rates. A restorer to the
Ae. juvenalis cytoplasm in wheat was implicated in pollen germination and vegetative growth and proposed as
Rfd1 [
20]. At the same time, the transmission rates of individual chromosomes (or, rather, their competitive advantage/disadvantage) depend to some extent on the allelic composition of the competing chromosome arms. Pollen bearing the 1RS.1BL translocation can be competitive with pollen carrying 1BS from wheats other than Pavon (see
Table 2), producing an almost random transmission rate of the translocation. However, factors other than
Rfmulti itself (or, rather, the region in its immediate vicinity as present in engineered chromosomes 1B and 1R), must also be involved. Interestingly, the transmission rates of the 1RS.1BL translocation from heterozygotes with 1B observed here are in direct contrast to those reported before [
33], where preferential retention of 1B (or discrimination against 1RS.1BL) was observed in the eggs. Taking these observations together, the system tested here offers interesting research avenues into the nuclear–mitochondrial interaction going far beyond male sterility.
The combination of the
Ae. kotschyi cytoplasm and a rye 1RS inducer is well known for haploid production [
1,
4,
5] but it does not only induce parthenogenesis. It clearly affects all aspects of double fertilization in wheat: fertilization of the egg cell without fertilization of the double nucleus in the embryo sack; fertilization of the double nucleus of the embryo sack without fertilization of the egg cell producing embryo-less grain; fertilization of the double nucleus in the embryo sack and induction of the development of the egg without fertilization (parthenogenesis); fertilization of both the egg and a synergid, producing diplo–diplo twins; and fertilization of a synergid and induction of the egg development producing haplo–diplo twins. Haplo–diplo–haplo triplets are occasionally produced, but with two sperm nuclei in the pollen and a normally developed seed (triple fertilization can be ruled out unless two pollen grains are involved). However, the system clearly requires much further study as, at present, it is far from clear if allelic variation for parthenogenesis inducers exists or is obscured by the presence of suppressors and perhaps even enhancers. To this day, the authors have not found a single wheat accession with a recessive allele of
Rfmulti. Indeed, some observations are strange: as reported by Lukaszewski [
9], three accessions of
T. spelta var. duhamelianum tested for male fertility restoration were almost completely male sterile in the F
1 generation with (
kot)Pavon 1RS.1BL.
T. spelta var. duhamelianum was the only non-restoring wheat reported [
3,
4,
7] but detailed data on that observation are no longer available from the authors. The three accessions tested here, which produced almost male sterile F
1, restored male fertility in a normal fashion (that is, at the same level as any other 1B chromosome tested) in generations from BC
1 onward. When single chromosome substitutions of 1Bs from these three accessions into cv. Pavon (BC
7) were crossed as male to (
kot)Joker, one F
1 set 2.22 seeds per spikelet, typical for the restoration level by one copy of
Rfmulti. Another F
1 hybrid set 0.17 seeds per spikelet, and the third one was completely male sterile. Similarly puzzling effects were observed in three wheats from the Wheat 10+ genomes project: cvs. Arinalfor, CDC Stanley, and Mace [
34]. Their F
1 hybrids with (
kot)Pavon 1RS.1BL were almost completely male sterile (seed set per spikelet >0.1), while their euplasmic control hybrids with (
aes)Pavon 1RS.1BL were normally fertile (seed set 1.65 to 2.94 per spikelet), suggesting the absence of fertility restoration to the cytoplasm of
Ae. kotschyi. However, in BC
1 with the same chromosome constitution, 1B + 1RS.1BL and in the
Ae. kotschyi cytoplasm, the seed set was around 65% (ca. 2.0 per spikelet), typical of fertility restoration by a single dose of
Rfmulti. At present, the authors do not have any sensible explanation for this behavior.
To start untangling the issues associated with the effects of the
Ae. kotschyi cytoplasm in wheat, an attempt was made to detect and tag suppressors of parthenogenesis. The result appeared more complicated than we had hoped for. It is obvious that haploid frequencies scored here carry a certain error, especially for low frequencies. As much as a small sample appears adequate to determine a high haploid frequency, with haploids absent, there is no upper limit on the minimum sample size. With a 10% assumed haploid frequency, the minimum sample size for a 95% probability of observing haploids is 28.4; this drops to 13.6 for a 20% haploid frequency. Here, the average sample size for the entire population was 21.7. Moreover, some effect of the pollinator on haploid frequencies has been reported [
1]. This effect was ignored here as it was not possible to grow a sufficient number of the same pollinator to produce pollen over the entire flowering time of the mapping population members. If there was a pollinator effect here, it likely was minor; no major differences were observed between haploid frequencies in the maintenance backcross of (
kot)Joker and crosses of (
kot)Joker with several different wheats made during the same period.
In the biparental mapping population Joker x Pavon 1RS.1BL
jok, five genomic regions appeared to be significantly associated with haploid production. Four of those, located on the chromosome arms 1DS, 2DL, 2AL, and 7AL, suppressed haploid production. One locus, located on chromosome 5AL, appeared to enhance it. The main effect of suppression was strongly associated with the region on chromosome 2DL. Chromosome arm 1BS with its known parthenogenesis suppressor was absent in the mapping population as all individuals were homozygous for the 1RS.1BL translocation from cv. Joker. At present, it is not clear if genetic variation exists for either the inducer (1RS) or the 1BS suppressor. Some casual observations of the effects of various 1BS arms, including data from the Tsunewaki group [
1,
2,
3,
4,
7,
11,
26,
33], suggest that chromosome arms 1BS may carry different alleles at the suppressor locus, which may either completely inhibit haploid production, such as 1BS of Pavon, or only partially so, as 1BS of cv. Chinese Spring. Tsunewaki et al. [
1] performed a monosomic analysis of the haploid induction rate in a cross (
umbelulata)Salmon x (CS monosomics x Salmon). Haploid frequencies ranged from 5 to 38%. With only two progenies per combination scored and with small samples, there was no conclusive evidence of a strong suppressor or suppressors. However, it appeared quite convincing that 1BS of Chinese Spring did not completely suppress haploid production. The fact that wheats such as Salmon, Joker, and Aurora [
11] can produce haploids at a rate of 90% and higher clearly suggests that allelic variation at the suppressor loci, other than that on 1BS, is present. How frequent is unclear at the moment.
Tsunewaki and Mukai [
4] proposed the designations
Spg (
suppression of parthenogenesis) and
Ptg (
parthenogenesis) for loci involved in haploid production. The former was located on chromosome arm 1BS; the latter on rye chromosome arm 1RS. They also suspected that the wheat line, Salmon, may carry a recessive gene that promotes haploid formation, or that Chinese Spring, used in their studies, carried a suppressor, “or both”. This study suggests that the worst-case scenario suspected by Tsunewaki and Mukai [
4] may in fact be true: other
Spg loci are present in wheat genomes. Five genome regions affecting haploid production were identified here (
Table 3). All four suppressor loci were contributed by cv. Pavon; the presumed enhancer locus originated from cv. Joker. Perhaps this is the enhancer locus speculated on by Tsunewaki and Mukai [
4] or perhaps some other effect, such as heterozygosity for two alleles at the locus. To follow the convention proposed by Tsunewaki and Mukai [
4], the suppression locus on chromosome 1B should be renamed
Spg-1B, and those detected here would be
Spg-1D,
Spg-2D, and
Spg-7A. If the locus on 5A indeed enhances haploid production, it is proposed to name
it Epg-5A (
enhancement of parthenogenesis). Data collected here suggest that a very high frequency of haploids among progeny is a consequence of a combination of an effective inducer, on rye chromosome arm 1RS, an enhancer on chromosome 5A, and the absence of suppressors.
It is not clear if allelic variation of the parthenogenesis inducer,
Ptg, exists or if this inducer is ubiquitous. Clearly, 1RS arms present in Salmon, Joker, and triticale Presto carry highly effective inducers. Based on data by Mukai [
11] the 1RS.1BL translocation in cv. Aurora also carries
Ptg. If so, then very likely cv. Kavkaz and all Veery lines of CIMMYT, also carry such an inducer, as they all originate from the same rye source [
35]. Aurora and Kavkaz carry 1RS.1BL translocations originating from the same source, cv. Neuzucht [
21] and appear identical [
22]. The fact that wheats Salmon, Aurora, Joker, and triticale Presto are capable of producing over 90% haploid progeny in some seasons suggests that they may also carry the enhancer on 5AS; wheats Salmon, Joker, and Aurora do not carry the suppressor on 2D; cv. Presto does not have it by virtue of its chromosome constitution. Beyond that, it is impossible to speculate on the extent allelic variation of the other loci affecting haploid frequency, but the fact that genetic mapping was possible in this study indicates that such variation does exist. Unclear are the frequencies of various alleles and their individual contributions. This would require a considerable effort of setting up suitable genetic stocks for tests, unless credible haplotypes are established for the loci involved, to select genetic backgrounds free of suppressors. The authors have created a large set of introgressions of 1RS chromosome arms, mostly as 1RS.1BL translocations but also as whole chromosome substitutions and centric translocations to other group-1 chromosomes, but only in cv. Pavon, which, with its array of suppressors, is quite unsuitable for tests. Chinese Spring, with its wide range of cytogenetic stocks would be an obvious choice, but it carries the 2D suppressing haplotype. Cv. Joker is not a sensible option as it requires a very long vernalization period, which makes two growth cycles per year almost impossible.