Introduction

Plants that reproduce by gametophytic apomixis form an embryo directly from an egg cell in an unreduced embryo sac. In contrast to sexual reproduction, the processes of meiosis and fertilization are avoided, leading to the formation of genetically uniform progeny (Nogler, 1984a; Koltunow, 1993). Apomixis is therefore typically defined by function as asexual reproduction through seed (syn. agamospermy) (Nogler, 1984a). The developmental mechanisms employed by different apomictic species, however, are remarkably varied (Nogler, 1984a; Koltunow et al., 1995), possibly reflecting the apparent polyphyletic origin of this trait among flowering plants (see Carman, 1997). Known mechanisms of apomixis have been categorized in different ways. Typically genera are grouped by similarities to a type-mechanism (Asker & Jerling, 1992). From our own observations in Hieracium subgenus Pilosella it is apparent that this form of characterization can be misleading (Koltunow et al., 1998). The members of this taxon are reported to form asexual seed by apospory of the Hieracium-type (Nogler, 1984a; Asker & Jerling, 1992) as first described by Rosenberg (1906). Cytological comparisons were conducted between vegetatively propagated isolates of three closely related species, the apomicts H. aurantiacum and H. piloselloides and a sexual accession of H. pilosella. Although development in the sexual accession was found to be typical of the Polygonum-type, between the apomictic lines studied significant differences were recorded in the number and precocity of aposporous initials formed, and in the number, morphology, developmental cytology and fates of the resultant unreduced embryo sacs. It is apparent from these and other unreported data, that many of the developmental details of apospory are quite variable in Hieracium, and that this variability is expressed between apospecies, between accessions of an apospecies and even between florets on an individual capitulum.

As a complement to this work, the inheritance of apomixis was studied using these three accessions. Apomixis has been described as a heritable trait in several species (Asker & Jerling, 1992). Dominant inheritance for apospory has been described in the grass genera Panicum (Savidan, 1981), Brachiaria (Borges do Valle et al., 1994) and Pennisetum (Dujardin & Hanna, 1985). The results of Christoff (1942) indicate that dominant inheritance may also apply in Hieracium aurantiacum (reviewed by Nogler, 1994). Similarly, Gadella (1991) noted that apospory may be inherited as a dominant factor in H. pilosella. Nogler (1984b) conducted a detailed study of the inheritance of apospory in Ranunculus auricomus, also concluding that the trait was conferred by the inheritance of a single dominant determinant. That determinant, however, could only be transferred in a heterozygotic state in a diploid or polyploid gamete. Nogler noted that this mechanism favoured the involvement of a polyploid parent and ensured the formation of a polyploid, apomictic zygote.

There is a very close association between gametophytic apomixis and polyploidy (reviewed in Asker & Jerling, 1992). In common with most other taxa containing gametophytic apomicts, all naturally occurring apomictic forms of Hieracium have been recorded as polyploid (Skalinska, 1970; Gadella, 1991) yet several diploid sexual species are known. The recent recovery of a diploid, apomictic plant of H. aurantiacum (Bicknell & Borst, 1997) indicates, however, that polyploidy is not an obligate requirement for the expression of apomixis in this taxon.

The aims of the current study were to determine the patterns of inheritance in our experimental accessions of H. aurantiacum and H. piloselloides, to test for any prejudice against the mediation of haploid gametes in the transmission of dominant alleles conferring apomixis, to test for allelism between the two apomicts tested and to karyotype the chromosomes of H. pilosella.

Materials and methods

Plant materials

Two apomictic biotypes of Hieracium were used, a triploid accession (2n=27, x=9) of H. piloselloides (designated ‘D3’) and an aneuploid accession (2n= 3x + 4=31) of H. aurantiacum (designated ‘A3.4’). Preliminary studies using test-crosses and pollen viability stains revealed that both had a high degree of pollen fertility and that both frequently produced haploid pollen. They were also selected for their distinctive morphologies. The bright orange flowers of H. aurantiacum and the upright form and strap-shaped leaves of H. piloselloides provided useful marker characteristics for confirming hybridity in the progeny. A3.4 was originally obtained from an adventive population in Central Otago, South Island, New Zealand, and D3 was obtained from a wild population in Steiermark, Austria. Two sexual biotypes were used; a tetraploid accession of H. pilosella from Caen, France (designated ‘P4’) and an anther culture-derived diploid (Bicknell & Borst, 1996) of P4 (designated ‘P2’). Both P4 and P2 appear to be self-incompatible under cool winter conditions. Pollinations were conducted without prior emasculation.

To ensure the genetic integrity of stocks used, the plants were maintained vegetatively, either by divi- sion of mature plants or through micropropagation (Bicknell, 1994). The inheritance studies were conducted in a glasshouse at Lincoln, New Zealand, maintained with a minimum night temperature of 12°C, minimum day temperature of 18°C and maximum day temperature of 25°C. Flowering was promoted by day-length extension lighting, using high pressure sodium vapour lamps with a photoperiod of 16 h (Yeung, 1989).

Breeding scheme for the inheritance of apomixis

The breeding scheme used is detailed in Fig. 1. Inheritance studies with apomictic species are subject to some unique constraints. As the trait results in the formation of predominantly maternal seed, apomixis restricts access to the female gamete, the egg cell. Furthermore, when a rare hybridization event does occur in these plants either a reduced or an unreduced egg may be used, resulting in the formation of different progeny classes and frustrating the analysis of segregation. To avoid these difficulties crosses were designed using an apomict as the staminate parent and a sexual plant as the recipient.

Fig. 1
figure 1

The breeding scheme used in the study. (a) Utilization of the sexual diploid ‘P2’ as the pistillate parent. (b) Utilization of the sexual parent ‘P4’ as the pistillate parent. (c) Test for linkage/allelism between the AD allele of D3 and the AA allele of A3.4. *A similar scheme was used with A3.4 as the apomictic, staminate parent. **Proposed genotype. ***Transgenic for a simplex, hemizygous copy of p35S-NPTII.

Full size image

Initially it was hoped that by using a triploid apomict as the staminate parent in a cross with a diploid, sexual recipient, the inheritance of apomixis could be studied at the diploid level. Following a cross between the apomictic triploid D3 and the sexual diploid P2, how- ever, most of the progeny were found to be either triploid or aneuploid (Table 1, cross 1). The expected diploid class was only represented by three weak individuals. The triploid progeny of this cross were therefore chosen for further analysis. Representative sexual and apomictic hybrid triploid progeny were backcrossed to the sexual recurrent parent and the BC1 progeny analysed for ploidy and reproductive mode. Again, as triploids were the most common progeny class in the BC1 populations, they were analysed for the inheritance of apomixis.

Table 1 Crosses conducted to determine the mode of inheritance of apomixis in Hieracium
Full size table

Both the apomictic triploid D3 and aneuploid A3.4 were then crossed to the sexual tetraploid P4 and the progeny analysed for ploidy (Table 1, crosses 4 and 7). Triploids were recovered from both crosses in sufficient numbers to permit an assessment of the inheritance of apomixis. Backcrosses were conducted for both the sexual and apomictic triploid progeny classes of the F1, using the sexual accession P4 as the recurrent, maternal parent. The triploid BC1 progeny were analysed for the inheritance of apomixis. Linkage between the dominant alleles from H. piloselloides and H. aurantiacum (‘AD’ and ‘AA’, respectively) was tested by hybridizing A3.4 and D3, followed by a test-cross using P2 as the pistillate parent. To facilitate the recovery of rare hybrids from among the predominantly maternal progeny of A3.4, the D3 parent used was first transformed with a single copy of a chimeric pnos-NPTII-3′nos sequence (Bicknell & Borst, 1994). The progeny were then screened on a medium supplemented with 100 mg/L kanamycin and the inheritance of the marker in the resistant seedlings was confirmed by Southern analysis (data not shown). The hybrid progeny were also clearly morphologically intermediate between their dissimilar parents.

Scoring progeny for ploidy and mode of reproduction

The ploidy of progeny plants was determined by chromosome counting using aceto-orcein staining of root-tip squash preparations as previously described (Bicknell & Borst, 1997). The karyotype of P2 is illustrated in Fig. 2. For scoring, apomixis was defined as the formation of an embryo(s) at petal senescence following the decapitation of an immature capitulum. Ovule structure was visualized by ovule clearing (see below). In apomictic biotypes of Hieracium maternal embryos arise spontaneously before the final senescence of the flower (stage 10 in Koltunow et al., 1998). In sexual types, prevention of fertilization by bud decapitation (Koltunow et al., 1995) leads to the retention of a quiescent mature embryo sac up until this developmental stage, so that differences between sexual and apomictic types could be readily determined. It should be noted that this measure is a score of parthenogenesis rather than apomixis as a whole as it does not include a measure of meiosis. In a separate study the rate of female meiosis was determined to be less than 2% in this material (to be reported separately).

Fig. 2
figure 2

The ideotype of the nine chromosomes of Hieracium pilosella, compiled from more than 20 independent observations of root-tip mitotic figures of the sexual diploid P2. Chromosome 1 is a large submetacentric chromosome bearing a conspicuous single satellite on the short arm. This was the only satellite identified in the karyotype of P2; no other secondary constrictions were evident, indicating that chromosome 1 bears the sole nucleolus organizer region in this plant. Chromosome 2 is a large metacentric and chromosome 3 is subtelocentric. Chromosomes 4 and 5 are submetacentric, whereas chromosomes 6, 7, 8 and 9 are small metacentrics. No indications of B chromosomes were observed.

Full size image

Three cut capitula were scored from each plant. Immature capitula were decapitated just prior to floret opening (stages 5 or 6 in Koltunow et al., 1998) then allowed to progress to stage 10. The outer bracts were removed from the harvested heads and the tissue fixed overnight in FAA (Stelly et al., 1984). The ovaries, attached to the receptacle of the capitulum, were then treated to remove oxalate inclusions and impregnated with methyl-salicylate as previously described (Koltunow et al., 1998). The cleared ovaries were then examined by differential interference contrast on an Olympus BH2 or Leica DM R microscope and scored for the presence of embryo sacs and/or embryos. At least six ovaries were examined from each capitulum. If embryos were detected the plant was scored as a putative apomict. If embryos were absent but normally developed embryo sacs were present, the plant was scored as sexual. Finally, if only degenerate structures were seen the plant was scored as sterile. In a small number of individuals, flowers either failed to form or developed abnormally. These plants were also scored as sterile.

Statistical analysis

For each cross, progeny were assigned to a progeny class (apomictic, sexual or sterile) and the frequency of each class determined. Steriles were separately considered either as having arisen independently of apomixis/sexuality or as representing either dysfunctional sexuals or dysfunctional apomicts. In the former case they were omitted from the analysis entirely whereas in the latter cases they were either assigned to the sexual or to the apomictic progeny class, respectively.

Segregation for apomixis among the F1 and backcross (BC) progeny was examined by the chi-squared test for a fixed ratio hypothesis with application of the Yates’s continuity correction factor to compensate for the low number of observations in some data sets. The test was applied against a null hypothesis that apomixis was a dominant trait and that in each case the original apomictic donor had the genotype Aaa with respect to apomixis. Chi-squared values were computed for the three different assignments of the sterile class described above.

Results

Apomixis was inherited as a monogenic, dominant trait in H. piloselloides

One hundred and eighty-five hybrids were recovered when the triploid apomict D3 (2n=27) was crossed to the diploid sexual accession P2 (2n=18). Of these, only three were diploids (2n=18) (Table 1, cross 1), two of which were sexual and one sterile. Eighty-eight aneuploid progeny were recovered, ranging in chromosome number from 19 (2x + 1) to 26 (2x + 8), with the most common classes being 23 (2x + 5) and 24 (2x + 6) chromosomes (data not shown). Ninety-four triploids (2n=27) were recovered. For the statistical analysis it was assumed that the triploid progeny resulted from the union of a haploid gamete from the pistillate parent, with a diploid gamete from the staminate parent. As no plants were obtained with a chromosome number greater than 27, it appears that completely unreduced gametes did not contribute to the formation of this population. The statistical analysis indicates the data are consistent with the hypothesized genotype. Assignment of the steriles into either the sexual or apomictic classes did not lead to the rejection of the null hypothesis.

To determine which F1 segregant class represented the homozygous recessive phenotype, backcrosses were conducted using representative triploid apomictic and sexual F1 segregants, crossed back to the sexual parent P2 (Table 1, crosses 2 and 3). Because the apomictic parent D3 formed almost entirely maternal seed, it was not practical to conduct the reciprocal backcrosses between it and the F1 segregants. Backcrossing one of the apomictic F1 triploids to its sexual parent yielded 48 progeny. The two diploids recovered from this cross were both sterile, whereas the 30 triploid BC1 plants recovered segregated for apomixis in a manner similar to the F1 population (Table 1, cross 2). Backcrossing a triploid sexual F1 plant to the sexual parent yielded only sexual progeny (Table 1, cross 3). Sexuality therefore appears to be the phenotype of the homozygous recessive genotype, and apomixis appears to be conferred by the inheritance of a single dominant factor.

The transfer of the dominant allele was mediated by both haploid and diploid male gametes

The poor recovery of diploids from the cross between P2 and D3 clearly indicated a prejudice against the survival of products following the involvement of haploid, male gametes. It remained uncertain, however, whether this selection had occurred at the level of haploid gamete viability/efficacy, or of diploid zygote survival. To investigate this, a sexual tetraploid P4 (2n=36) was used as a pistillate parent in a cross with the apomictic triploid D3 (Table 1, cross 4). The diploid P2 was originally derived by anther culture from P4 (Bicknell & Borst, 1996), so the P2 genome represented a subset of P4. Triploid F1 progeny recovered from this cross (P4 × D3) were expected to have arisen from the union of a diploid egg with a haploid sperm nucleus, whereas tetraploids would have arisen from the involvement of a diploid sperm nucleus. A total of 99 progeny were recovered, ranging in ploidy from 3x to 4x (Table 1, cross 4). Sixty-two of the F1 plants were triploids (2n=27) and 15 were tetraploids (2n=36). Apomixis segregated among both the triploid and tetraploid F1 progeny. Segregation, tested by the chi-squared test, was found to be consistent with the expectations of a monogenic, dominant model of inheritance. It was concluded that haploid gametes could transmit the dominant allele conferring apomixis in this system.

Backcrosses were conducted to test for dominant inheritance. Triploid apomictic and sexual F1 segregants were crossed to the sexual tetraploid parent P4 (Table 1, crosses 5 and 6). The apomictic F1 backcross (Table 1, cross 5) resulted in the formation of a segregating population, with ratios of apomictic and sexual biotypes similar to the F1 (Table 1, cross 4). Conversely, backcrossing a sexual F1 plant to the sexual parent resulted in the formation of only sexual and sterile progeny (Table 1, cross 6).

Apomixis was also inherited as a monogenic, dominant trait in H. aurantiacum

Our cytological investigations of the apomicts, H. piloselloides (D3) and H. aurantiacum (A3.4) revealed that in these species distinct differences were apparent in the timing and degree of aposporous embryo sac formation. It was therefore of interest to test whether the inheritance of apomixis from A3.4 was similar to that recorded for D3. One hundred and thirty-five plants were recovered from a cross between the sexual tetraploid P4 and the apomict A3.4 (Table 1, cross 7). Of these, 102 were recorded as triploids. Segregation for apomixis among the triploid progeny was consistent with a monogenic, dominant pattern of inheritance, assuming the union of a diploid egg with a haploid sperm nucleus. Results from the backcrosses confirmed previous findings. As observed with D3, apomixis segregated when a triploid apomictic F1 was backcrossed to the sexual parent (Table 1, cross 8), but not when a triploid sexual F1 was backcrossed (Table 1, cross 9). The results indicate that the apomict A3.4 also carries a single dominant allele for apomixis and that the allele could be transmitted in either a haploid or diploid male gamete. A3.4 is an aneuploid carrying four copies of four chromosomes and three of the remaining five. Chi-squared tests indicated that the trait was monogenic in this plant; however, sample sizes are inadequate to determine between trisomic (Aaa) and tetrasomic (Aaaa) inheritance. It is therefore not possible to use these data to assign the Apo locus to a given chromosome, or chromosome set in this plant.

With the exceptions of the triploid progeny of P4 × PD3.apo (Table 1, cross 5) and the tetraploid progeny of P4 × PA3.apo (Table 1, cross 8) the treatment of sterility as an event independent of apomixis/sexuality provided the best goodness of fit for the data under the experimental assumptions. It was therefore concluded that sterility segregated as an independent trait in these experimental populations.

The dominant factors of A3.4 and D3 are closely linked

As both A3.4 and D3 could be used to transmit a dominant factor conferring apomixis, we wished to test whether they acted at the same, or at different loci in this plant system. To test for allelism it was necessary to combine the two dominant alleles in a single plant through hybridization, requiring that one of the apomicts be used as a recipient parent. To facilitate the recovery of hybrids a hemizygous, transgenic D3 parent was used (see Methods). Ten transgenic seedlings were recovered after the pollination of 30 capitula of A3.4 with transgenic pollen of D3. The number of seedlings screened was estimated at 1550, indicating a recovery rate of 0.64% and a fertilization rate of 1.3%. Of the 10 hybrids, four were tetraploids, four were triploids and two were aneuploids (3x + 4 and 3x + 5). All were either apomictic or sterile. Most of the apomicts expressed the trait at only a low level, forming only a small number of asexual seeds on each capitulum. Two of the tetraploids, however, set abundant germinable seed after the decapitation of immature buds. The best of these plants, designated ‘AD4’, was test-crossed to the sexual diploid P2 and the triploid progeny scored for apomixis. One hundred and sixty progeny were recovered. Table 2 summarizes the results of chi-squared tests conducted for three proposed models of the genotype of AD4, with the three possible assignments of the sterile class, assuming monogenic dominance for both the AA and AD alleles. Given the finding described above, that sterility does not appear to be preferentially associated with either apomixis or sexuality, close linkage, possibly allelism, is indicated as the most probable genetic model.

Table 2 Allelism test between Hieracium aurantiacum and H. piloselloides
Full size table

Discussion

Apospory in Hieracium is inherited as a dominant monogenic trait

Apospory in H. piloselloides and H. aurantiacum was inherited as a monogenic, dominant trait. This agrees with the findings of Gadella (1991) with H. pilosella and with similar findings reported for the aposporic taxa Panicum (Savidan 1981), Pennisetum (Ozias-Akins et al., 1993), Ranunculus (Nogler, 1984b) and Brachiaria (Borges do Valle et al., 1994), and for the diplosporous species Tripsacum dactyloides (Leblanc et al., 1995). It is also consistent with the finding that sexual biotypes could be recovered following the anther culture of an apomictic pentaploid of H. pilosella (Bicknell & Borst, 1996). Both D3 and A3.4 were simplex for the dominant allele at the Apo locus. Typically, seed formation in field-adapted facultative apomicts is largely maternal; very little meiotic seed is produced. As apomixis significantly reduces the availability of female gametes for hybridization, hybridization events in mixed populations of such apospecies are expected usually to involve a sexual biotype as the pistillate parent. Sexuality is the homozygous, recessive phenotype, so this mechanism will lead to the dispersion of dominant alleles. The identification of two polyploids each bearing a single dominant allele at Apo therefore is consistent with the expectation that genotypes bearing more than one dominant allele should arise only rarely in this taxon.

Diploid apomixis was prevented by selection against diploid zygotes

Gametophytic apomixis has been recorded in more than 400 species of flowering plant, from 35 diversely aligned families (Carman, 1997). Despite this taxonomic diversity, however, with very few exceptions gametophytic apomicts have been recorded as polyploids (Asker & Jerling, 1992; Carman, 1997). In the small number of cases when diploid apomicts have been recorded they have typically been weak experimentally derived anomalies (Nogler, 1982; Leblanc et al., 1996; Bicknell & Borst, 1997) in species predominantly composed of polyploid races. The introduction of apomixis into crops is an important goal in apomixis research (Hanna, 1995), yet the majority of crops are diploids or stable allopolyploids, the types of species in which gametophytic apomixis is almost unknown. Before the objective of transferring apomixis can be achieved, therefore, the causes and consequences of this fundamental association need to be more fully understood.

During his pioneering work with the aposporous apomict Ranunculus auricomus, Nogler (1984b) noted that A+, the dominant determinant that conferred apomixis, could not be transferred in a homozygous state. Homozygosity of A+ was associated with gamete lethality, so A+ could only be transmitted by a diploid or polyploid gamete. Nogler noted that similar mechanisms of gamete-level selection against homozygous dominant alleles could explain both the establishment of polyploidy and also its retention in apomictic populations, despite occasional hybridization events. More recently, Grimanelli et al. (1998) recorded a similar pattern of inheritance for the transfer of diplospory from Tripsacum dactyloides into Tripsacum/maize hybrids. They further noted, however, that although a profound restriction against the simplex transmission of the dominant factor for apomixis was observed in the BC1 generation it was not apparent in the subsequent BC3 generation.

By contrast, (Mogie 1988, 1992) has proposed that apomixis is conferred by the inheritance of incompletely penetrant alleles. In polyploid biotypes increased allelic dosage would permit the expression of the trait, whereas the presence of those alleles would be masked in diploids. Carman (1997) conducted a comprehensive assessment of associations between apomixis, polyspory, polyembryony, base chromosome number and several forms of polyploidy. He concluded that apomixis arises from interactions between asynchronously controlled regulatory elements inherited from dissimilar parents. This mechanism would require the inheritance of intact, or near-intact, regulatory cascades located throughout the genomes of both parents. Carman hypothesized that this would probably act against the mediation of haploid gametes, and therefore ensure that apomictic progeny were polyploid. The proposals of Nogler and Carman, outlined above, predict that selection will occur at the level of the gamete. Haploid gametes carrying either a dominant determinant for apomixis or only a partial series of regulatory elements are expected either to fail to survive or to fail to transfer the entire trait. In Hieracium, however, this interpretation does not apply. In the current study both haploid and diploid male gametes were successfully used to mediate the transfer of the AD allele for apospory from H. piloselloides and the AA allele from H. aurantiacum. Furthermore, no evidence of segregation distortion was seen among the progeny that would implicate the action of gamete-level selection.

Crosses between the diploid sexual plant P2 and three different triploid pollen donors (D3, PD3apo and PD3sex) yielded only five diploids among the 270 progeny assessed. Of those five, three were clearly sterile and all grew very poorly. Furthermore, as the criterion used for assigning sexuality was based on the presence of a typical quiescent gametophyte at floral senescence, it was unclear whether the two ‘sexual’ diploids identified would have been capable of completing sexual seed formation. By contrast, 115 of the 270 progeny assessed were aneuploids and 150 were triploids. When a tetraploid sexual parent was used in crosses with the apomicts D3 and A3.4, triploids were frequently recovered among the progeny. This indicates that many of the haploid male gametes of both D3 and A3.4 were functional and their products viable. As apomicts were well represented among the triploid progeny of these crosses there appears to have been no difficulty in transferring the trait via these gametes. It is unlikely that the poor recovery of diploids was caused by an inviability of reduced P2 egg cells. Their reduced state and efficacy were clearly demonstrated by the recovery of other ploidy classes from these crosses, from crosses with AD4, and also by the segregation patterns for ploidy and apomixis seen in the resulting hybrids.

In Hieracium, therefore, the formation of apomictic diploids was discouraged through selection against diploid hybrids, acting after fertilization, rather than through gamete-level selection acting prior to this event. Furthermore, the nonrecovery of apomictic diploids appears to be attributable to the poor viability of all the diploid products of the polyploid apomictic parents, rather than to any specific prejudice against the apomictic progeny class.

Mogie’s model (1988) predicts selection following fertilization and would therefore be consistent with the observations of inheritance in Hieracium. It also predicts, however, that the diploid genotype Aa would not express apomixis at an identifiable level. In a previous study (Bicknell & Borst, 1997) a diploid, apomictic form of H. aurantiacum was recovered from an experimental population. That plant, designated A2, was derived from A3.4, a genotype used in this study. As A2 appears to have formed by dihaploidy it is expected to have the genotype Aa at the Apo locus. Aposporous initials in A2 were shown to have been even more numerous than in A3.4, often causing ovule abortion because of the unrestrained competitive proliferation of multiple asexual structures. In contrast to the predictions of Mogie’s model, the genotype Aa appears to express apomictic tendencies at least as strongly as the genotype from which it was derived by reduction.

Reduced diploid hybrid viability may be primarily caused by a high genetic load in apomictic Hieracium

The poor recovery of diploids in this study may simply be a reflection of the deleterious genetic loads carried by the accessions of Hieracium studied. When a diploid sexual plant was used as the pistillate parent, vigorous hybrid progeny were recovered following fertilization with a diploid gamete from an apomict, but not following fertilization with a haploid gamete. In contrast, when a tetraploid sexual was used as the pistillate parent, seedlings arising from the mediation of a haploid gamete from an apomict were commonly recovered. With respect to both the sexual parent P2 and the apomicts D3 and A3.4, therefore, haploid gamete use was significantly enhanced by fertilization with a diploid gamete from the other parent. This requirement for fertilization by a diploid gamete is consistent with the dominance model of heterosis (Ledig, 1986) which predicts that a high load in a haploid gamete would be better masked by fertilization with a diploid, than with a haploid gamete from a highly heterozygotic partner.

The use of an anther-culture derived sexual diploid (Bicknell & Borst, 1996) may have further amplified this effect. P2 was derived from P4, a tetraploid, self-incompatible, obligate outcrossing accession of H. pilosella. Both polyploidy and outcrossing can mask deleterious, recessive mutations (Richards, 1997). A further indication of the deleterious genetic load carried by these plants was demonstrated by an apparent intolerance to inbreeding. Lower seed-set was routinely seen after the backcrossing of F1 segregants to their sexual parent (data not shown), and the percentage of sterile BC1 individuals was always greater than that seen among the corresponding F1 population (Table 1). Conversely, F1 hybrids were typically more vigorous than either of their parents.

As a form of asexual reproduction, apomixis leads to the sequential accumulation of deleterious mutations, a phenomenon often referred to as ‘Mullers Ratchet’ (Muller, 1964). In sexual diploid species, deleterious recessive mutations may be exposed to selective pressures as a result of recombination at fertilization, but probably more critically, also during haploid gametophyte formation and function (the ‘Clean Egg Hypothesis’; see Richards, 1997). In gametophytic apomicts, however, the megagametophyte is unreduced. Only microgametogenesis and pollen function, when these events are still operative, are subject to this form of gamete-level selection, which may be why reduced pollen continues to be formed by many of the autonomous apomicts of Hieracium.

The establishment of polyploid apomixis

Because polyploidy increases the number of alleles present in a genotype it provides a mechanism for improving the tolerance of a fixed genotype to the accumulation of recessive load. The masking of deleterious recessives will be particularly pronounced where polyploidy is established through cross-hybridization rather than by endoreduplication or self-fertilization. It is interesting to note that hybridization, particularly BIII hybridization, the fertilization of an unreduced egg cell by pollen of another genotype, has been documented in several apomictic plants, including Hieracium (Skalinska, 1973; Gadella, 1991). An apparent hybrid origin is also a feature of many apomictic groups (Asker & Jerling, 1992) indicating that it has often been an important influence in determining the composition of apomictic taxa. By providing a mechanism for increasing the tolerance of a genotype to mutation, however, polyploidy can also be expected to lead to the accumulation of a greater mutational load in these plants. Song et al. (1995) demonstrated that a genetic distance of between 2.5% and 9.6% developed between a population of F5 plants and their F2 homozygous ancestors following four generations of selfing. In the absence of the gamete-level selection mechanism discussed above, autonomous apomicts like Hieracium may accumulate mutations at an even greater rate. Furthermore, Song et al. (1995) noted that wide hybridization appeared to lead to a prejudice against paternally inherited alleles, possibly because of incompatibility between paternal nuclear and maternally inherited cytoplasmic elements. Wide hybridization in apomicts would create similar circumstances and can therefore be expected to exert similar selection pressures for gene silencing.

In summary, the transition from diploidy to polyploidy in apomictic Hieracium would confer advantages associated with masking mutations accumulated by the action of Mullers ratchet. Conversely, however, the institution of polyploidy would encourage the continued accumulation of mutations and reduce the competitiveness of any diploids that may subsequently arise. This process would rapidly favour the establishment of polyploid apomixis yet discourage any reversion to diploidy. As an apomictic taxon ages, the degree of deleterious genomic modification can be expected to increase and the impacts of this effect would become more pronounced. Haploid gamete lethality would follow, because of the exposure of recessive lethal alleles located throughout the genome during periods of haploid gametophyte formation and function. When coupling phase linkage becomes established between a lethal recessive allele and an allele for apomixis, Nogler’s hypothesis of linked lethality would apply. Grimanelli et al. (1998) noted that the prevention of simplex transmission for a dominant apomixis-factor from alloploid BC1 Tripsacum/maize plants was not apparent by the subsequent BC4 generation. They attributed this transmission ratio distortion to the operation of a trans-acting incompletely penetrant system. From our observations with Hieracium it appears that the phenomenon of transmission distortion should be more generally defined as resulting from multiple cis- and trans-acting lethal factors of varying penetrance, located throughout the fixed genome of an apomict.

Modulators of the apomictic locus

Molecular differences in the Apo locus between H. piloselloides and H. aurantiacum may account for many of the differences observed in the timing and frequency of initial cell differentiation, and in the general variety of subsequent developmental events (Koltunow et al., 1998). The inheritance data do not exclude, however, the potential influence of physiological variation, environmental effects or other genetic loci modifying the impact of a dominant determinant at the Apo locus. The modulating effects of alleles at modifier loci were indicated throughout this study by clear differences in the number and morphologies of embryos observed amongst the sibling progeny of the crosses described. Furthermore, none of the apomictic progeny examined in these experiments exhibited the high level of apomixis observed in the parents (to be published elsewhere). In a successful apomict, it is necessary but not sufficient that the events of meiosis, fertilization and parthenogenesis be uniquely regulated. They must also be co-ordinated in association with an array of other developmental functions. This was clearly demonstrated in the diploid apomict A2 (Bicknell & Borst, 1997) in which aposporous initial differentiation was prolific. Seed germination rates, however, were very low because of the frequent, premature degeneration of the ovule caused by the actions of multiple, competing asexual structures within. An important remaining question therefore is what influence do the modifier loci have on the expression of the apomictic locus in this plant? For example, do they influence spatial and temporal events of apomixis in these plants, or the frequency and penetrance of the trait in ovules, and is their presence essential to promote high yields of viable seed? In agricultural systems, seed number and viability are frequently important determinants of overall yield. The transfer of apomixis in an economic format is therefore likely to require changes in the expression of several genes. The genetic information and genotyped stocks derived from this study are now being used to conduct genetic and developmental analyses of the impacts of apomixis-modifier expression in Hieracium.