A Eurasia-wide polyploid species complex involving 6x Trifolium ambiguum, 2x T. occidentale and 4x T. repens produces interspecific hybrids with significance for clover breeding

Background Trifolium ambiguum occurs as a 2x, 4x, 6x polyploid series in W Asia. The 6x form is the most agronomically desirable, having strong rhizomatous spread and drought tolerance. These traits would be potentially very valuable if they could be transferred to white clover (T. repens) which is the most important agronomic clover species. However, to-date, no fertile interspecific hybrids with 6x T. ambiguum are available. Previously, 2x T. occidentale from W Europe has produced synthetic fertile hybrids with both 2x and 4x T. ambiguum and these were inter-fertile with white clover. Here we ask whether 2x T. occidentale can form fertile hybrids with 6x T. ambiguum and act as a genetic bridge to white clover and bring these species together as part of a common gene pool. Results Ten verified F1 (6x T. ambiguum x 2x T. occidentale) hybrids were produced by embryo rescue and seven were studied further. All four investigated for chromosome number were 2n = 4x = 32 and FISH confirmed the expected 21 T. ambiguum and 8 T. occidentale chromosomes. Hybrid fertility was extremely low but 2n female gametes functioned with white clover pollen to produce seeds. Derived plants were confirmed using FISH and were successfully backcrossed to white clover to produce partially fertile breeding populations. Conclusions Although T. occidentale and 6x T. ambiguum are widely separated by geography and ecological adaptation they have maintained enough genomic affinity to produce partially fertile hybrids. Inter-fertility of the hybrids with allotetraploid T. repens showed that T. occidentale can provide a genetic bridge between 6x T. ambiguum and white clover to produce plants with new phenotypes combining the traits of all three species. Use of this information should enable potentially valuable stress tolerance traits from 6x T. ambiguum to be used in white clover breeding for the first time.


Background
The genus Trifolium has approximately 250 species distributed throughout Europe, Africa, W Asia and the Americas [1]. Within the genus, a small section (Trifoliastrum) of 14 species has been delineated by DNA sequence phylogeny [2]. The species in the section range from diploid annuals to hexaploid perennials, and are scattered separately across Europe and W Asia. Among other species, the section includes T. ambiguum (Caucasian clover), T. occidentale (western clover) and T. repens (white clover). Caucasian clover (Trifolium ambiguum) occurs as a polyploid series in continental W Asia centred around the Caucasus region. The polyploid series approximates an altitudinal gradient, with diploid populations at high altitudes, hexaploids at the lowest altitudes and tetraploids between [3]. This variation across the series suggests that the genomes of the ploidal forms carry adaptive genetic differences. In contrast, T. occidentale is a diploid, strictly maritime species confined to the gulf-stream sea coasts of W Europe from Portugal to Ireland, SW England and Wales [4][5][6][7][8]. T. repens is a tetraploid with a wide natural distribution, spanning Europe and W Asia, and is the most important of about 10-15 clover species used in world-wide agriculture [9].
Although white clover is widely used, it is a stoloniferous plant that moves across the soil surface and has a shallow, fibrous root system that makes it vulnerable to drought conditions. This vulnerability is exacerbated by the fact that white clover is also an inefficient user of water [10,11]. By contrast, T. ambiguum is of considerable agricultural interest as a pasture legume because of its deep root system, underground creeping growth habit through rhizomes, drought tolerance, pest and disease resistance and high feed quality [9]. These attributes complement several shortcomings that restrict the performance of white clover on farms, especially in dry environments.
T. repens is of hybrid (allotetraploid) origin [12] and is unusual as the genus is remarkably lacking in reticulate speciation [2]. Despite the rarity of natural species crossing, several species pairs in Section Trifoliastrum have been artificially hybridized, mainly to add new genetic diversity to clover breeding populations [9]. These crosses included those between T. repens and artificial tetraploid T. occidentale [13][14][15][16]. Hybridization of 4x T. ambiguum and 4x T. repens has also been achieved, but with very low success rates and with considerable difficulty using embryo/ovule culture [17][18][19]. These hybrids have been difficult to breed and despite several generations of population development and field selection [9,20,21] have, to-date, produced only one cultivar. This might, in part, be due to the vegetative inferiority of the tetraploid form of T. ambiguum used, and this increases the desirability of finding ways to incorporate the hexaploid form.
So far, the hexaploid form has been of greatest agricultural value. This form is vegetatively much more robust than the diploid and tetraploid forms, with large leaves, tall habit and large underground stems and roots [3] and it has been successfully deployed for forage production in the mid-western USA [22]. A potential strategy to improve the drought performance of white clover could be to hybridize it with this vigorous form of Caucasian clover. However, direct hybridization would need to overcome potential barriers of geographic distance, diverse ecological adaptations and polyploid speciation. Although F 1 hybrids between white clover and 6x T. ambiguum have been obtained by embryo rescue [23], no fertile F 1 hybrids have so-far been reported, and so the desirable hexaploid form has not been incorporated into the wider gene-pool for clover breeding.
An alternative strategy could be to use the knowledge that white clover is a hybrid between ancestral forms of T. pallescens and T. occidentale [12] and that there has been successful hybridization of 2x and 4x T. ambiguum with 2x T. occidentale. This suggested that T. occidentale could be used as a genetic bridge between white clover and T. ambiguum [24,25]. In the present case, this would require the generation of not only fertile hybrids between 6x T. ambiguum and T. occidentale, but also the ability to obtain hybrids of these with a third species, T. repens. If achieved, this hybridization would have implications, not only for speciation in the genus, but also for expanding the gene-pool for clover breeding.
We therefore undertook a study to see whether fertile hybrids between 6x T. ambiguum and 2x T. occidentale could be produced by embryo rescue. Here we provide experimental evidence that 6x T. ambiguum can form fertile hybrids with 2x T. occidentale and that the 4x F 1 hybrid plants were inter-fertile with white clover. This hybridization spanned the potentially isolating barriers of continental-scale distances, diverse ecological adaptations, and polyploid speciation, and provided evidence of a recent pan-Europe-W Asia species radiation. Furthermore, the results reveal the commercial potential for T. occidentale to be used as a genetic bridge for including 6x T. ambiguum in the gene-pool to facilitate the breeding of improved cultivars of white clover.

Results
Pollination of 6x T. ambiguum with 2x T.occidentale led to a change of the maternal standard petal colour from white to pink within 24-48 h. This was interpreted as evidence that fertilization had occurred. The frequency of florets developing pods was low (10-20%) and fewer than 1% produced detectable embryos. Dissection of hundreds of pods produced useful numbers of embryos that were green and at the early-torpedo stage. Eighteen of these were grown on agar plates into plantlets that were transferred to potting mix in the greenhouse. (A poor batch of medium led to losses of an unknown number during culture). Three months after transfer, 16 putative hybrid plants were growing in the greenhouse. Ten of these were verified as hybrids and are listed in Table 1 with their parentage. Six of these grew to maturity, flowered and three were studied in more detail (Table 1).
Early observations were that the hybrids plants were large plants, vegetatively closer in size to T. ambiguum than to the diminutive T. occidentale. They were similar to T. ambiguum in leaf surface texture and flowering pattern. However, they had morphological characteristics from both the parent species (Table 2), being semiprostrate or semi-erect, with robust horizontal stems with adventitious roots only at the basal nodes. Leaflets were large and pointed and flower heads resembled those of T. ambiguum. Detailed measurements on potgrown mature plants of three hybrids confirmed that the morphology of the hybrids was closer in shape and size to T. ambiguum than to T. occidentale (Tables 3, 4). Petioles of the hybrids were approximately double the length of those of T. occidentale and the leaflets were elongated and similar in shape to the leaflets of T. ambiguum. Stem thickness, leaf thickness and numbers of leaves and growing points per plant were also closer to those of T. ambiguum, while plant spread was inferior to both parents. Dry weights of all plant parts, except inflorescences (heads), were inferior to both parent controls (Table 4). Inflorescence numbers and DWs were transgressively much higher than both parent species. Flowering was terminal, as in T. ambiguum, rather than axillary, as in T. occidentale, and so the flowering stems were determinate, dying back to the basal nodes after seed maturity.
Below ground measurements on mature plants grown in pots (Table 3) showed that the three hybrids studied had thick roots similar in diameter to T. ambiguum.
Rhizomes were present in all three hybrids but were fewer in number and, in two cases, thinner than those of T. ambiguum. However, the dry weights of the roots and rhizomes were markedly lower than T. ambiguum (Table 4).

Cytogenetic analyses of the F 1 hybrids
Somatic chromosome preparations of hybrids 5, 33, 34, 51 and 140 showed 2n = 4x = 32, each consistent with derivation from a n = 3x = 21 gamete from 6x T. ambiguum and a n = x = 7 gamete from T. occidentale. A cell of hybrid 5 is shown in Fig. 1a, revealing four satellited chromosomes and a vague indication of a 3:1 ratio in chromosome sizes.
FISH using 35S and 5S rDNA as probes showed that the chromosome complement of hybrids 33 and 140 were consistent with expectations [26] following fusion of normal gametes from both species. Figures 2a, b show a somatic cell of hybrid 33. One NOR-bearing chromosome with a 5S signal on the opposite arm was from T. occidentale and three NOR-chromosomes lacking 5S signals were from T. ambiguum. There were three chromosomes with large 5S signals from T. ambiguum and one with a very small 5S signal from T. occidentale. The T. occidentale-derived NOR was observed to be always decondensed while one or two of the T. ambiguum-derived NORs were often, but not always, condensed.

Meiotic configurations in F 1 hybrids
Metaphase I meiotic configurations were analysed for hybrids 5 and 34 ( Table 5). The results for both plants were similar, showing high numbers of univalents per cell, a few bivalents and trivalents and very few quadrivalents. Hybrid 5 was also observed at anaphase I where it showed frequent laggards and only infrequent 16-16 disjunction. Pollen fertility of the F 1 hybrids was extremely low, with pollen staining ranging from 3 to 13% (Table 6). Based on the very low pollen fertility and unsuccessful attempts to obtain seed by hand pollination, cloned copies of six hybrids were grown in pots outside and were open-pollinated by bees. Surrounding plants included diverse genotypes of both parent species along with diploid and tetraploid T. ambiguum and white clover. Five of the hybrids produced seed from open-pollination. Seed-set was determined on four hybrids and ranged from zero to 14 seeds per 100 heads (i.e. up to approximately 0.28 seeds per 100 florets, assuming an average of 50 florets per head) ( Table 6). Seed-lots from hybrids 5, 33 and 229 were germinated and the progeny were investigated.

Descriptions of OP progeny plants from hybrid 5
Hybrid 5 produced only one small normal OP seed and five very small seeds. Only the normal seed developed into a mature plant, designated 5OP-1. The seedling of 5OP-1 was raised in a greenhouse in winter and was abnormal, with a pale green unifoliolate leaf and a yellow first trifoliolate leaf. Subsequent trifoliolate leaves were pale green with very narrow elongate leaflets. The mature plant had very narrow leaflets with double-V white markings. The plant produced multiple short stems that were rooted at the basal nodes. Flower heads were T.
ambiguum-like but were white rather than pink at the base. Mature foliage was light green while leaves produced in some seasons in an unheated greenhouse were yellow-green, suggesting sensitivity to cool temperatures. 5OP-1 was moderately fertile with pollen stainability of 35% and a somatic chromosome complement of 2n = 48 (Fig. 1b). The number of satellite chromosomes was five ( Fig. 1b) as would be expected if white clover was the male parent.

Descriptions of OP progeny plants from hybrid 33
Sixteen seeds harvested from hybrid 33 after OP were germinated and 12 mature plants were obtained. Several of these flowered and showed varying degrees of fertility as shown by pollen stainability and seed-set from controlled crosses ( Table 7). There was strong phenotypic evidence that to form some of these progeny, hybrid 33 had been pollinated by white clover (Table 7). For example, progeny plants 33OP-1, 3, 13, 16, 18, 19, 20 all were stoloniferous and nodally rooted. Plants 33OP-13 and − 16 also carried purple leaf mark alleles from white clover. Progeny plant 33OP-17 was apparently rhizomatous, while several plants with horizontal stems and nodal roots only at the base could not be described as stoloniferous and might not have had a white clover parent. Plant 33OP-1 and its descendants were subjected to further study.
Plant 33OP-1 was robust and stoloniferous with elongated leaves that were obtuse, rather than acute, at the  tips. The leaves carried two white V markings and the texture was thinner and less stiff than that of 6x T. ambiguum and hybrid 33. Flowers produced in the first spring had a strong scent of coconut oil. The anthers were small, and possibly indehiscent and, while pollen was not abundant, it showed moderate stainability ( Table 5). The plant was also partially female fertile, giving 1-4 seeds/inflorescence when both open-pollinated and self-pollinated. Flow cytometry indicated that this plant was near-6x, suggestive of the union of a near-4x (i.e. unreduced) female gamete from hybrid 33 and a 2x = 16 male gamete. Somatic preparations confirmed that 33OP-1 had a chromosome complement of 2n = 51. FISH using 35S and 5S rDNA as probes (Fig. 2c, d) showed that, in addition to the full complement of marker chromosomes described above for hybrid 33, this plant had one more NORbearing chromosome with a 5S signal on the opposite arm, and a chromosome with a medium-sized 5S signal, as expected if white clover was the provider of the 2x = 16 male gamete [26]. In addition, there was a fourth, T. ambiguum-derived, chromosome carrying a large 5S signal.

Seeds and progeny from hybrids 31, 34 and 229
OP seeds were obtained from all three of these hybrids and was placed in storage. Nineteen OP progeny plants from hybrid 229 were grown in pots in a greenhouse. Only a few plants thrived, generally showing low fertility. A few had nodally rooted stolons and rounded leaves consistent with having had white clover as the male parent.
Progeny of 5OP-1 from controlled crosses 5OP-1 produced 3-4 seeds/head on self-pollination and 2-3 seeds/head when pollinated with two different white clover plants. One white clover head was pollinated with 5OP-1 and this resulted in 14 seeds.
Three seeds were germinated from the pollination of 5OP-1 with a white clover plant (BRI x WIL-1)-902, carrying the dominant R l allele for purple leaf colour in heterozygous form. All three of the progeny showed white clover traits, being stoloniferous with nodal roots and axillary inflorescences and one had purple leaves, confirming the white clover paternity. A further five progeny plants were obtained after pollination of 5OP-1 with a second white clover plant and four of these were stoloniferous with  nodal roots, again indicating successful fertilization by white clover. All eight plants obtained from crosses by white clover were partially fertile when open-pollinated among themselves, but seed-sets were low, the best two plants averaging approximately 10 and 14 seeds/head. Most of this seed was stored but a few (BC 1 F 2 ) plants were grown and further backcrossed to white clover to produce putative BC 2 seed that was also placed in storage.
Hand-pollination of 33OP-1 by a white clover plant heterozygous for red and diffuse red leaf markings (R l / R ld ) produced seeds that gave rise to verified progeny carrying one or the other leaf marking. Most of the progeny had pale green or variegated yellow/green leaflets, especially in the early seedling stages, and only about one third survived to maturity (25 seeds gave 22 seedlings of which eight grew to maturity and flowered). The mature plants were robust with thick, densely branched horizontal stems that lacked or had only basal nodal roots. Flow cytometry showed that they were near-5x. Chromosome counts on two of them confirmed numbers of 2n = 42 and 41-42. The latter of these plants, carrying a white V mark and the diffuse red leaf allele was used as the pollen parent on an unmarked white clover plant. Nine verified progeny were obtained from 10 seeds sown (one plant was a self on the white clover mother plant). All nine progeny were strong, stoloniferous, plants resembling white clover. Apart from one that failed to flower, they were moderately fertile with a mean pollen staining of 43% (range 29-69%) and none showed the pale green foliage sectors that had been characteristic of the previous generation. Flow cytometry estimates averaged 2n = 4.7 and ranged from 4.5-5.0. These plants were the products of at least two (probably three) crosses of hybrid 33 (AAAO) with white clover. Advanced generation progenies were produced for further investigation.

Progeny from controlled crosses of OP plants from hybrid 229
One OP progeny plant from hybrid 229 (229 OP-2) was crossed as male with white clover and produced nine

Discussion
This research has shown that 6x T. ambiguum could be crossed with 2x T. occidentale to produce partially fertile 4x hybrids that were inter-fertile with 4x white clover (T. repens). These results have implications for both species evolution and potential introgression breeding of white clover. The fact that white clover is involved, implies potential for combining the agronomicaly desirable features of both species into new hybrid populations, perhaps leading to introgression of 6x T. ambiguum genetics into white clover. Previous studies [24,25] have shown that both 2x and 4x T. ambiguum could be hybridized with 2x T. occidentale and that the F 1 s were inter-fertile with white clover. This indicated that all four taxa were derived from a common ancestor and were part of a species radiation that now covers all of Europe and parts of W Asia. This radiation involved polyploidization and led to extremely divergent morphologies and ecological adaptations across a very wide geographic region. For both 2x and 4x T. ambiguum and T. occidentale it was apparent that the evolution of extremely divergent species had occurred in the absence of strong barriers to hybridization and potential gene flow. This study has shown that 6x T. ambiguum also shows these properties, despite further polyploidization and adaptational divergence.
Somatic chromosome counts on several F 1 s gave 2n = 4x = 32, confirming the expected number. FISH using 5S and 35S rDNA probes, carried out on two of the F 1 s also confirmed a ratio of three T. ambiguum to one T. occidentale sub-genomes, based on characterization of the parents [26].
Observations indicated no evidence of major prefertilization barriers when 6x T. ambiguum was pollinated by T. occidentale. Post-fertilization, the frequency of rescuable embryos was low but green torpedo embryos were not rare either, and so dissection of a few hundred ovules was rewarding. Most of the dissected embryos grew into large, robust plants that were vegetatively more like T. ambiguum in leaf shape and in having rhizomes, but they also weakly expressed basal nodal rooting, reflecting T. occidentale parentage. In general, the phenotypes of the F 1 plants were consistent with the 3:1 ratio of T. ambiguum to T. occidentale sub-genomes. However, clear exceptions were plant spread, which was poorer than in either parent, and inflorescence development, which was positively transgressive.
Despite the impressive vegetative and floral vigour of the F 1 s, both male fertility (pollen staining) and female fertility (seed-set) were extremely low. Of six hybrids checked, five were below 10% in pollen staining. Very low pollen staining (< 3%) was also observed in three F 1 hybrids between 2x T. ambiguum and 2x T.occidentale [24] and, similarly, it was approximately 9% in one 3x F 1 between 4x T. ambiguum and 2x T. occidentale [25]. Female fertility (OP seed-set) was, however, apparently much lower in the 6x T. ambiguum x 2x T. occidentale F 1 s than in the hybrids involving 2x and 4x T. ambiguum. One 2x hybrid produced nearly one (0.8) OP seeds/head [24], while the 3x hybrid produced 1-2 OP seeds/head [25]. In contrast, the seed-sets obtained in the present study ranged from 0.00-0.14 seeds/head ( Table 4). The very low fertilities were consistent with an expected low frequency of genomically balanced gametes following meiosis in these plants with uneven numbers of parental sub-genomes.
Meiotic analyses of two of the F 1 s provide a probable explanation for the low fertility. Data for hybrids 5 and 34 (which both produced progeny) indicated very dysfunctional meiosis, with more than half of the chromosomes per cell, on average, either unpaired (univalent) or involved in trivalents (Table 5). Unequal disjunction of these, as was observed, was likely to lead to a predominance of reduced gametes with unbalanced chromosome compositions. However, an average of one quadrivalent every two cells suggested that some A-O pairing had occurred, potentially leading to genomic recombination.
Three of the OP progeny of hybrids 5 and 33 (5OP-1, 33OP-1, 14) were checked for chromosome number and, in all cases, they were derived through the functioning of unreduced hybrid female gametes. In two cases (5OP-1 2n = 48, 33OP-1 2n = 51) the male gamete was from white clover which reliably produces gametes with 16 chromosomes. Therefore, the female gametes had, respectively, 32 and 35 chromosomes, both close to the expected unreduced number of 32. Although the gamete from 5OP-1 could have been a fully balanced 2n gamete, it is possible, given the meiotic abnormalities, that it was also unbalanced and coincidentally 2n = 32. The aneuploid 2n gamete from hybrid 33 was found to have an additional 5S rDNA chromosome from T. ambiguum. This gamete could have had a 4x + 3 constitution, but other chromosome imbalances were also possible. The third case (33OP-14, 2n =~56) was also consistent with derivation from a near-2n hybrid female gamete. These findings suggest that 2n gametes were the predominant (and perhaps only) class of functional gametes produced by these AAAO hybrids. The frequencies of seeds set were very low (e.g. hybrid 33, 0.8-1.0 seeds per 1000 florets). Assuming 1-3 ovules per floret, and that only 2n gametes were fertilized, this would produce a frequency range of successful 2n female gametes of 0.03-0.10%. Because some 2n eggs would have been unfertilized, this probably underestimated the frequency of 2n gametes, but it nevertheless falls within the known range of 2n female gamete production and at the low end for hybrids [27]. The results are therefore consistent with unreduced gametes being the predominant or only class of fertile gametes from the AAAO hybrids. Unreduced gametes were also observed to have functioned in the progeny of 2x T. ambiguum x 2x T. occidentale (AO) F 1 hybrids [24].
Among six F 1 hybrids that were investigated for fertility, five proved to be partially fertile and produced OP progeny ( Table 4). The three progeny families that were grown were also fertile and produced progeny from further crosses. In one case (33OP-1) the male parent was confirmed by FISH to have been white clover. This plant proved to be inter-fertile with white cloverthus establishing a tri-species lineage involving 6x T. ambiguum, 2x T. occidentale and 4x T. repens. While initially aneuploid, this lineage retained fertility when taken through subsequent generations. The progeny of hybrids 5 and 229 also were inter-fertile with white clover and thus also established tri-species lineages.
The implications of these lineages are two-fold. First, they establish that these three taxa, one 2x, one 4x and one 6x form an inter-related group derived from a recent, probably diploid, common ancestor. Second, from a clover breeding viewpoint, they show that T. occidentale can be used as a genetic bridge to attempt to supplement the white clover gene-pool by introducing genomic constituents from 6x T. ambiguum, which is generally regarded as the most agronomically suitable form of this species.
Speciation has involved polyploidization along with marked phenotypic and adaptive divergence. T. occidentale is stoloniferous and adapted to saline, coastal habitats and 6x T. ambiguum is rhizomatous, and adapted to high altitude continental habitats. T. repens is an allotetraploid derivative of hybridization between alpine T. pallescens and coastal T. occidentale and its adaptation spans the habitat range of the whole group. Previous work [24,25] showed that 2x and 4x forms of T. ambiguum formed fertile hybrids with T. occidentale, also after embryo rescue. These inter-fertilities suggested that species divergence had probably not occurred from an accumulation of genetic changes in inter-fertility but rather that such changes had occurred later after the separation and independent development of the ancestral populations [24,25].
However, genetic barriers to hybridization of 6x T. ambiguum with both T. occidentale and T. repens, although incomplete, appear to be quite strongly developed. With T. occidentale, embryos were few and developed only to the early green torpedo stage and had to be rescued. Although most of the F 1 hybrids were robust, their fertilities were extremely low. Only unreduced gametes appeared to be functional with white clover and the tri-species progeny plants were near-6x and were generally also low in fertility.
Based on this small sample, it appears that the polyploidization processes that led to the formation of 6x T. ambiguum have been accompanied by accumulated genetic changes that have reduced, but not eliminated, inter-fertility with the sister taxa.
From a clover breeding viewpoint, this group, which includes other species, has been referred to as the "white clover complex" [28] and is a group of Trifolium species phylogenetically closest to white clover (T. repens). This complex is defined by presence in section Trifoliastrum [2] and ability to cross-hybridize with at least one other species in the group. Williams [18] further subdivided the species according to the ease of crossing with white clover. 4x T. ambiguum fell into the tertiary gene pool (requiring embryo rescue) while 2x and 6x T. ambiguum were placed in the quaternary gene pool (requiring both bridge crossing and embryo rescue). Infertile hybrids between 6x T. ambiguum and T. repens have been reported [23]. If these hybrids subsequently prove to be partially fertile, then 6x T. ambiguum could be moved to the tertiary gene pool.
The breeding plan to incorporate 6x T. ambiguum into the T. repens breeding pool using T. occidentale as a bridge species is given in Fig. 3. As an example, where hybrid 33 was the F 1 (AAAO), one near-6x progeny plant from pollination with T. repens [P r P r O r O r ] was 33OP-1 (2n = 51). This was derived from a near-4x (n = 35) female gamete and a normal (n = 2x = 16) white clover gamete. Further backcrosses to T. repens led to progressive reductions in ploidy towards near-4x T. repens and these would have carried chromosome additions, deletions and substitutions from 6x T. ambiguum and T. occidentale. The presence of quadrivalents at meiosis in the F 1 hybrids indicated the possibility of recombination between 6x T. ambiguum and T. occidentale chromosomes, as previously reported for hybrids involving 2x and 4x T. ambiguum and T. occidentale [24,25]. In the present case, should there have been any interspecific chromosome pairing followed by recombination at any meiosis in the breeding process then T. occidentale could function as a genetic bridge, potentially leading to introgressions from 6x T. ambiguum into T. repens backcross families.
This breeding plan indicated that significant levels of aneuploidy and mixed, unbalanced sub-genomes would lead to unpredictable genomic constitutions in the intermediate generations. In practice, the early backcross derivatives of 33 OP-1, although aneuploid, were robust plants with moderate fertilities. More advanced generations were produced in good numbers and will be investigated for evidence of addition, substitution or introgression from 6x T. ambiguum into white clover. Success would mean that the desirable features of 6x T. ambiguum that are absent from the white clover primary genepool would become available for future breeding and selection and potentially improve the performance of white clover populations in marginal, especially dry, environments.

Conclusions
Diploid T. occidentale and hexaploid T. ambiguum have maintained enough genomic affinity to produce partially fertile hybrids despite widely divergent geographic distributions and ecological adaptations. Inter-fertility of the hybrids with allotetraploid T. repens showed that T. occidentale can provide a genetic bridge between 6x T. ambiguum and white clover to produce plants with new phenotypes combining the traits of all three taxa. Use of this information should enable potentially valuable stress tolerance traits from 6x T. ambiguum to be used in white clover breeding.

Plant materials
All living plants used in this study were derived from cultivars and accessions conserved in long-term storage in the Margot Forde Forage Germplasm Centre, AgResearch Grasslands Research Centre, Palmerston North, New Zealand. Five hexaploid (6x) T. ambiguum plants derived from self-pollination of a single plant of the commercially available cv. Endura (PGG-Wrightson Seeds, Christchurch, New Zealand) were used as female parents (designated Endura self- 10,12,13,16,18). T. occidentale seeds (accession code OCD) were made available for research purposes only and were originally from collections made at sea level sites in N Spain under an approved Agreement for the Acquisition of Material for Plant Genetic Resources.

Hybridization and embryo rescue
The plants were brought to flower in an insect-free greenhouse at temperatures of 15-28°C and a photoperiod of 16 h. Crossing was done by hand by placing pollen from T. occidentale on the exposed stigmas of T. ambiguum using the method of Williams [29]. After 8-10 days, early-torpedo embryos were taken from the ovules and placed under aseptic conditions on CR7 medium [30] in petri dishes. Approximately 4 weeks later, the plantlets were transferred to CR5 (root initiation medium) [30]. After a further 8-12 weeks the plants were potted into a peat/sand medium in the greenhouse. Hybridity was confirmed using the isozymes SDH and PGI [31], DNA sequencing of the nuclear ITS and chloroplast TrnL intron regions [2] and by fluorescence in situ hybridization (FISH) [26].

Plant propagation, fertility assessment and progeny development
Mature plants were clonally propagated by cuttings. Flowering plants were placed outside for a full seasonal cycle to be open-pollinated (OP) by bees in the presence of the parent species and white clover.
Male fertility was estimated from pollen staining ( Table 1) and determined by extracting mature pollen, staining with 1% acetocarmine and counting the percentage of full stained grains among a minimum of 300 at 200x magnification. Female fertility was assessed using seed-set (numbers of seeds per head) following self-or cross-pollination.

Phenotypic descriptions
Three interspecific hybrids were characterized, along with plants of 6x T. ambiguum ("Endura") and 2x T. occidentale (OCD 1168-14) in a pot experiment in a greenhouse at the Grasslands Research Centre, Palmerston North. The design, conduct and analysis of the experiment were as previously described [25]. Cuttings were planted in July 2008 and measurements were taken non-destructively in the next spring (October 2008) when the plants were flowering, and then destructively in the following summer (January 2009). The qualitative comparisons made are listed in Table 2 and the quantitative traits recorded are listed in Table 3 (spring) and Table 4 (summer).

DNA analysis, cytology and molecular cytogenetics
Total DNA preparation, PCR amplification and DNA sequencing of ITS and cpDNA regions were carried out as previously described [2]. The method for ploidal estimation by flow cytometry was described in [24]. Conventional meiotic analyses were carried out on PMCs from floral buds treated with alcoholic HCl carmine [32]. Somatic chromosome preparations used the flame drying technique [26] on actively growing root tips macerated with proteolytic enzymes. These preparations were stained by Giemsa for conventional mitotic analysis and were also used for FISH. Meiotic FISH preparations were made following enzymatic maceration of floral buds. The DNA probes and procedures used to identify the chromosomal locations of 5S rDNA and 35S rDNA sequences were described in [25]. The hybridization and post-hybridization procedures were as described by [26].