Cytological and Morphological Evaluation of Interspecific Hybrids between Trifolium repens and T . uniflorum

A significant constraint to the vegetative persistence of white clover (Trifolium repens L.) is summer moisture stress. A closely related species, T. uniflorum L., has a robust root system that could provide drought resistance. Hybridization of these two species leads to the generation of fertile hybrids. A study was conducted to generate F1 hybrids and first backcrosses (BC1) to white clover, and to evaluate the fertility, meiotic chromosome behavior, and plant morphology of these hybrids. Marker chromosome counts of the F1 and BC1 individuals confirmed hybridity. Meiotic configurations of five F1 and five BC1 plants indicated close homology between the two species and homoeologous pairing of white clover subgenomes. To evaluate phenotypes, clones of 56 individual genotypes (seven F1, 32 BC1, and 17 elite white clover plants) were grown in sand. After 13 mo, the leaves, stolons, and roots were measured and the dry weights of shoots and roots were determined by destructive harvest. Pattern analysis of the genotype-by-trait data identified four progeny groups. The F1 progeny were confined to the group with the lowest mean expression for most root and shoot traits. Most of the BC1 progeny grouped with the elite white clovers, but two BC1 formed a group with superior vigor and morphology, combining the best root and shoot traits of both parents. These BC1 individuals with similar shoot morphology to white clover and a robust root system, similar to T. uniflorum, are being integrated into a program to breed drought-resistant hybrids to replace white clover in dry environments. S.W. Hussain, I.M. Verry, M.Z.Z. Jahufer, and W.M. Williams, AgResearch, Grasslands Research Centre, Private Bag 11008, Palmerston North, New Zealand. Received 23 May 2017. Accepted 6 July 2017. *Corresponding author (wajid.hussain@agresearch.co.nz). Assigned to Associate Editor Ali M. Missaoui. Abbreviations: BC1, Backcross 1; DW, dry weight. Published in Crop Sci. 57:2617–2625 (2017). doi: 10.2135/cropsci2017.05.0314 © Crop Science Society of America | 5585 Guilford Rd., Madison, WI 53711 USA This is an open access article distributed under the CC BY license (https:// creativecommons.org/licenses/by/4.0/). Published August 17, 2017

The hybrids were found to be fertile and had strong root systems, suggesting potential value for breeding stronger roots into white clover.However, these hybrids were not developed further.In the research described here, new F 1 hybrids between T. repens and T. uniflorum and their backcrosses to white clover were generated and characterized both morphologically and cytologically, and their potential was revealed for the development of largescale plant populations for breeding and selection of new and novel transgressive synthetic clovers.

Hybridization and Embryo Culture
Plant material of the parent species T. uniflorum and T. repens was obtained from the Margot Forde Forage Germplasm Centre, Palmerston North, New Zealand.For F 1 production, T. uniflorum was represented by plants sampled from two accessions, one originating from a high-altitude collection site in Greece (Mount Parnes W of Athens in Abies Forest, maintained as AZ 4382 and AZ 4436) and the other from coastal Turkey (a hillside near the sea between Cheshme and Erythrea, maintained as AZ 4383 and AZ 4437).Trifolium repens was represented by individual plants sampled from diverse cultivars (Tables 1 and 2).For several reasons, including high floret numbers and ease of emasculation, hand pollination was conducted using T. repens as the female parent and T. uniflorum as the male parent.All F 1 hybrids were generated through embryo culture using the techniques of Williams et al. (2011).All but one of the BC 1 progenies were obtained without the use of embryo rescue by hand crossing F 1 plants (as either male or female) to T. repens.One BC 1 family was obtained using T. uniflorum as the recurrent parent.A list of BC 1 families is given in the Supplemental Table S1.

Cytological Observations
Somatic chromosome counts, meiotic configurations in pollen mother cells, and pollen staining of F 1 and BC 1 plants were performed using techniques described by Hussain and Williams (1997).

The Sandpit Experiment
In February 2006, 8 to 12 stolon cuttings were taken from each of seven F 1 , 32 BC 1 , and 17 T. repens genotypes representing five elite cultivars (Supplemental Table S1) for propagation in the glasshouse.After 6 wk, three rooted cuttings with visually uniform root and shoot sizes were planted into a large area of coarse river sand of 45-cm depth.The experimental design was a randomized complete block design with three replicates.Plants were arranged using a 60-cm grid with two rows of plants per replicate, and each plant received a complete nutrient solution (Yates Thrive, N/P/K 25:5:8.8)twice weekly.Overhead irrigation was also applied daily.Immediately before harvest, data were recorded on the length and width of the central leaflet of the first fully expanded leaf from the tip, length of longest stolon, stolon thickness at the midpoint of the longest stolon, stolon number, node number on longest stolon, plant size (1 = very small; 9 = very large, >40 cm from the center of the plant to the edge of the dense spread), and nodal rooting (0 = no nodal rooting, 5 = very frequent nodal rooting).In April 2007, plants were destructively harvested and data were collected on root length, number of roots with 2-mm-thick, nodulation (0 = no nodules, 5 = many nodules), and root and shoot dry weight (DW).

Data Analysis
The data collected for the 13 traits measured were analyzed using the residual maximum likelihood option in GenStat 7.1  First Backcross (BC 1 ) The BC 1 seeds were obtained without the use of embryo culture.Most of the F 1 hybrids were self-compatible, and so it was desirable to use them as male parents in backcrosses to white clover to prevent self-fertilization.
Alternatively, if T. repens was used as the male parent, a white clover plant with a distinct leaf mark could be used to facilitate paternity testing.Seed set was evaluated during the development of a group of 11 BC 1 progenies, seven with white clover plants as the female parent and four with the F 1 as the female parent (Table 3).Seed set was considerably lower in the backcrosses where F 1 plants were used as the female parent (6-11 seeds 100 florets −1 ) than those where white clover was used as female parent (32-48 seed 100 florets −1 ).A single BC 1 progeny obtained by pollinating an F 1 with T. uniflorum also had low seed set (8 seeds 100 florets −1 ).Male fertility was assessed in another set of BC 1 s involving three F 1 hybrids backcrossed to white clover parents (Table 4).Variation in male fertility occurred among closely related backcross plants.
Hybrid 902-1 in combination with five white clover parents from two cultivars produced BC 1 plants ranging from 3 to 88% in pollen staining.Within a single family, Crusader-45 ´ 902-1, pollen staining varied from 16 to 88%.Some parental combinations were clearly inferior to others.Hybrid 900-5 was highly fertile (Table 2) and produced a moderately fertile BC 1 with the white clover parent Crusader-45 (Table 4).However, the backcross combination with a different white clover parent, Trophy (GenStat, 2003).The analyses were performed using mixed linear models with replicates as random and plant genotypes as fixed.The resulting means were used to construct a 56-plantgenotype ´ 13-trait data matrix that was used in pattern analysis, a combination of cluster analysis and principal component analysis (Gabriel, 1971;Kroonenberg, 1994;Watson et al., 1996).The objective of using pattern analysis was to produce a graphical summary of the 56-genotype ´13-trait data matrix.Prior to cluster analysis, the mean values for each of the respective traits were standardized to have a mean of zero and a variance of one to remove scaling effects (Cooper and DeLacy, 1994).A 32-bit PC version of the Watson et al. (1996) GEBEI package was used to conduct the clustering.To decide on an optimum level of truncation for the resulting hierarchy from cluster analysis, the increase in the sum of squares among accession groups as the number of groups increased was investigated.The group level selected was determined by the point where the percentage of accession sum of squares among groups did not improve substantially as the number of groups increased (DeLacy, 1981).

T. repens ´ T. uniflorum F 1 Crosses
Thirty-four F 1 hybrids were produced from embryo culture and grown to maturity.Twenty of these F 1 hybrids were from crosses using plants of T. uniflorum, collected from Greece, as the pollen parent.These were crossed to six T. repens plants used as female parents (Table 1).Fourteen of these were assessed for male fertility and had pollen staining ranging from 30 to 80% with a median of 67%.The remaining 14 F 1 hybrids were produced from crosses using T. uniflorum plants collected from Turkey as the male parent to two of the T. repens female parents (Table 2).Eleven of these were assessed for pollen staining, with a range from 5 to 87% and a median of 72%.Hybrids also differed in pollen abundance and, although this was not consistently recorded, it was noted when a plant was clearly divergent.For example, Plant 900-1 with 5% pollen staining was also noted as having markedly less pollen than other hybrids, whereas Plant 900-5 R4-1, produced nine plants, six of which were effectively male sterile, with very little pollen and with staining <5%.

Cytological Evaluation of F 1 and BC 1
Five F 1 and five BC 1 plants were evaluated for somatic chromosome counts (Table 5) and were all tetraploid (2n = 4x = 32).Hybridity of the F 1 plants was confirmed by the presence of three satellited chromosomes (Fig. 1), two derived from the T. uniflorum male parent and one from the T. repens female parent.Meiotic chromosome configurations of the five F 1 and five BC 1 plants consistently revealed both trivalent and quadrivalent formation (Table 5), indicating close chromosome pairing homology between the genomes of the two species.The pollen fertility of the hybrid plants was negatively correlated with the frequency of odd-numbered configurations (univalents and trivalents, r = −0.85,P < 0.01).Mean frequencies of univalent and trivalent configurations for the F 1 and BC 1 did not differ, and mean pollen staining was similar (62-63%) for both generations.Mean ´ T66-6)-5 morphological traits measured.Shoot DW for the white clover control plants ranged from 3.7 to 23.3 g plant −1 , whereas the F 1 plants were markedly smaller (0.9-3.4 g plant −1 ).The BC 1 plants ranged widely from 1.5 to 21.4 g plant −1 , the best hybrids being similar in size to the best elite white clovers.Belowground, the F 1 hybrids were poor in root DW (0.6-1.7 g plant −1 ) but had the highest ratio of thick roots to total roots (1.7-4.8 thick roots g −1 DW root).Elite white clover root DW ranged from 2.9 to 11.1 g plant −1 , but the ratios of thick roots were lower at 0.1 to 1.0 thick root g −1 DW root.The BC 1 hybrids covered a wide range of root system sizes from 1.1 to 13.3 g −1 plant and ratios of thick roots of 0.7 to 3.3 thick roots g −1 DW root.Thus, the root systems of the best BC 1 hybrids were similar in DW to those of the best elite white clovers but generally showed more thick roots.Phenotypic data for each plant are given in Supplemental Table S1.
The multivariate analysis generated four statistically distinctive morphological groups (Fig. 2, Supplemental Table S1).The plants in Group 1 were low-yielding plants that were markedly below average in all traits, except for the numbers of thick roots.All of the F 1 plants and 14 BC 1 plants were in this group.Groups 2 and 4 were higher yielding, with Group 2 being distinguished from Group 4 by being larger in most plant parts including leaves, stolons, and roots.The plants in these groups were the most white clover-like and, on average, had fewer thick roots per gram DW of root than those in Group 1. Included frequencies of bivalents per cell were higher (P < 0.001) in the BC 1 plants (12.0) than in the F 1 plants (8.0), and multivalent numbers per cell (trivalents and quadrivalents) were correspondingly lower (P < 0.001) in the BC 1 (1.9) than in the F 1 (3.9).Two of the three T. uniflorum plants exhibited only bivalents and quadrivalents, but one showed a low frequency of univalents and trivalents.This plant also had a slightly lower pollen fertility than the others (Table 5).The F 1 s and BC 1 s showed 16-16 disjunction in most anaphase I cells.However, unequal disjunction was also observed in two F 1 s and two BC 1 s (Table 5).This suggested the formation of aneuploid gametes, albeit at a low frequency, by some hybrids.

Phenotypic Analysis
Analysis of variance indicated that there were significant (P < 0.05) differences among the hybrids for all of the  in these groups were all of the white clover plants (11 in Group 2, 6 in Group 4) and 16 of the BC 1 hybrids (10 in Group 2, 6 in Group 4).Plants in Group 3 were characterized by high expressions of most of the desired features of both parent species, combining high aboveground yields with large roots and, especially, high numbers of thick roots (Fig. 2).Two BC 1 hybrids, both backcrosses to the same white clover cultivar (Will), made up this group.
A phenotypic correlation matrix (Supplemental Table S2) showed that all of the shoot and root traits except numbers of thick roots were positively correlated.Numbers of thick roots varied independently of all the other traits except for a weak correlation with stolon thickness.Stolon thickness showed weaker than average correlations with most other traits (Fig. 2, Supplemental Table S2).

T. repens ´ T. uniflorum Crosses
Trifolium repens and T. uniflorum are very closely related tetraploid species that have been placed together in section Trifoliastrum on the basis of DNA sequence phylogeny (Ellison et al., 2006).Despite the close relationship, only some parental genotype pairs produced interspecific hybrids.Gibson et al. (1971) obtained a few hybrid seeds after preselection of compatible genotypes (using pod swelling as a guide).Seed set would require the compatibility of the genomes of the two parent species for both embryo and endosperm development.Chen and Gibson (1971) found that this was rare and usually endosperm failure led to seed failure.The success of embryo rescue is consistent with this conclusion.Our observations have confirmed this, but we also noted that occasional fertilizations produced excellent endosperm but no embryo.Our observations over several years suggest that compatible genotype pairs are in the minority and many plant pairs produced few or no useable embryos.These results are indicative that evolutionary changes in the two species since they diverged and became isolated from each other have led to an impaired ability of the genomes to cooperate during the development of both embryo and endosperm tissues.Apparently this impairment is, so far, partial and certain genomic or genetic combinations, presumably maintained by the heterozygosity of both parents, are able to achieve cooperation.
Once obtained, the hybrids showed variable fertility.The median pollen staining of F 1 s from both Greek and Turkish T. uniflorum parents indicated that half of the hybrids were very fertile (>67 and 72%, respectively).Pandey et al. (1987) reported that the pollen staining of six F 1 hybrids ranged from 28 to 58%.When successful parental combinations produced several F 1 hybrids, the fertilities of these often varied markedly (Table 1), and similar variation occurred among the BC 1 progeny of common parents (Table 4).The most probable cause of this variation was allelic variation for factors affecting hybrid fertility and maintained by the heterozygosity of both parent species.

Chromosome Pairing and Recombination
The meiotic configuration frequencies observed in F 1 hybrids in the present work contrasted with those reported earlier.In four independent hybrids, Chen and Gibson (1972) observed a mean frequency of univalents per cell of 4.2 (compared with 1.1 in the present study), with some cells having up to 16 univalents (compared with up to 4 here).Similarly, the mean frequency of bivalents previously reported (Chen and Gibson, 1972) was 10.8 (compared with 8.0 in this study).These apparent differences were presumably related to the relatively small sample sizes and the different population sources of both parent species.
In both the F 1 and BC 1 plants, most of the chromosomes were paired in bivalent and quadrivalent configurations at meiosis (Table 5), and only three of the 32 chromosomes, on average, were involved in univalents and trivalents.Nevertheless, hybrid fertility, as revealed by pollen staining, was strongly negatively correlated with these low frequencies of odd-numbered chromosome configurations.A similar negative relationship was found by Hussain and Williams (2016) in T. repens ´ 4x T. occidentale Coombe hybrids.
The frequency of bivalents increased significantly from the F 1 to BC 1 generation, and the frequencies of multivalents decreased.This improvement in bivalent pairing in the BC 1 occurred even though the species genome balances were changing from apparently balanced (2:2 T. repens/T.uniflorum) to unbalanced (3:1 T. repens/T.uniflorum).The cause is unclear but could be a combination of factors, including natural selection for more compatible genome combinations or partial restoration of the genetic control of bivalent chromosome pairing.
In genomic terms, the T. repens ´ T. uniflorum cross can be expressed as P r P r O r O r ´ UUUU, where P r is an ancestral T. pallescens Schreb.subgenome derived from T. repens, O r is an ancestral T. occidentale subgenome derived from T. repens, and U represents the T. uniflorum subgenome (Williams et al., 2012).It was expected that the F 1 plants would have an average genomic constitution P r O r UU (one ancestral T. pallescens genome, one ancestral T. occidentale genome, and two T. uniflorum genomes).On this basis, the very low frequency of univalents (1.1 cell −1 ) indicated a very high degree of pairing among the genomes.If the U genomes were preferentially paired with each other, then a high degree of homoeologous P r -O r pairing would have occurred.Such pairing does not occur in T. repens, which is totally disomic (Corkill, 1971;Williams et al., 1998), and so any P r -O r recombination would have potential value for releasing new genetic variation for plant breeding.Alternatively, if there were frequent O r -U and P r -U pairings, then the resultant recombinations should enable the successful application of introgression breeding.High frequencies of such homoeologous pairings can be implied from the BC 1 plants, which showed an average of only 1.0 univalent per cell despite the presence, on average, of eight U chromosomes with no homologous partners.
High frequencies of P-U and O-U homoeologous pairings would imply very close relationships between the U genomes and those in white clover.Indeed, comparisons can be made with similar pairing data for T. repens ´ 4x T. occidentale hybrids where the frequencies of unpaired chromosomes (univalents) in the F 1 and BC 1 were slightly higher at 2.9 and 1.4 cell −1 , respectively (Hussain and Williams, 2016).This might tend to suggest that the O o genome (the genome of contemporary T. occidentale) paired with white clover chromosomes less strongly than the U genome.This is surprising, considering that the O r subgenome of white clover has very high synteny and collinearity with the contemporary O o genome (Williams et al., 2009).The very high pairing affinity of the U genome suggests an equally close relationship with the subgenomes of T. repens.It is unclear whether or not there is any preferential pairing and recombination of the T. uniflorum chromosomes with either of these ancestral genomes, but it is clear from the low frequency of univalents that there was no wholesale failure of any one genome to pair with the others.
Other authors (Chen and Gibson, 1972;Pandey et al., 1987) have noted that the genetic control of regular bivalent pairing that occurs in white clover breaks down in F 1 T. repens ´ T. uniflorum hybrids, leading to some multivalent formation.The mean frequency of 3.25 quadrivalents per cell was consistent with the very close pairing affinities of the subgenomes of both parents.These multivalents are further evidence that, in the hybrids, potentially valuable recombination can occur between the ancestral subgenomes of white clover.Similar frequencies of quadrivalents (3.1 cell −1 ) occurred in T. repens ´ 4x T. occidentale hybrids (Hussain and Williams, 2016).

Hybrid Phenotypes and Backcross Breeding
The aim of crossing white clover with T. uniflorum was to combine the best characters of both species.With regard to the traits measured here, this involved combining the deep, thick roots of T. uniflorum with all of the aboveground characteristics of white clover.The F 1 hybrid phenotypes were an apparent backward step, as they were clearly inferior to white clover for nearly all aboveground traits.However, they were distinctive in having an approximately sixfold higher mean number of thick roots per gram of roots than white clover.After the first backcross, the BC 1 plants, on average, moved towards the recurrent parent but dispersed widely in phenotype.Some (Group 1) remained poor in most traits but retained a high ratio of thick roots.Those in Groups 2 and 4 were more white clover like, whereas the two plants in Group 3 were of special interest, as they combined many of the best characteristics of both parent species.It is plants like this, showing the outstanding features of white clover, accompanied by the enlarged root system from T. uniflorum, that will provide a base population for selection of a new interspecific hybrid clover to meet the breeding objectives.
The donor species (T.uniflorum) (Pandey et al., 1987;Nichols et al., 2014b) and F 1 hybrids were distinctly smaller than the elite white clover plants and, based on phenotype, were not an attractive proposition for a plant breeder aiming to retain vigor while improving drought resistance.Nevertheless, a single backcross to white clover, in many cases, restored trait expression to recurrent parent levels and, in some cases, led to transgressive plants with characteristics superior to both parents.This illustrates the importance of basing breeding decisions on the genotype (breeding value) rather than the phenotype of the base material.Thus, exotic introductions should never be disregarded until a genotype evaluation for all traits of interest has been performed.This could be progeny testing (the testing of backcross families), possibly aided by the use of genomic methods including quantitative trait loci analysis and genomic selection.Exotic introductions should always be retained in a seed bank because interest may arise in a previously untested trait for which the introduction may carry undetected, useful alleles.Some of the BC 1 hybrids derived from F 1 plants reported here (Table 1) have also been assessed in other studies and been found to exhibit modified root systems (Nichols et al., 2015b(Nichols et al., , 2016)), drought tolerance (Nichols et al., 2014a(Nichols et al., , 2015a)), and tolerance of low soil nutrients, especially phosphate (Nichols et al., 2014c).Because of the very small inflorescences of T. uniflorum, a potential problem with these hybrids could be seed production.However, Naeem et al. (2017) established that seed production could be restored to hybrid populations in only two generations of selection.
These results indicate that backcrossing of the F 1 hybrids to T. repens resulted in BC 1 progeny superior in the expression of key traits to the F 1 parents.This provides evidence that key morphological traits can be improved through successful interspecific introgression followed by backcrossing.In the future, backcrosses derived from F 1 s generated from crossing diverse germplasm of parental species are expected to provide novel genotypic recombinants and variation.This novel BC 1 genetic variation will provide a breeding pool for further targeted genotypic selection.
in 15 cells † PMC, pollen mother cell.

Table 1 .
Parentage and pollen staining of T. repens ´ T. uniflorum F 1 hybrids obtained by using AZ 4382 †/AZ 4436 T. uniflorum (from Greece) as the pollen parent.

Table 2
. Parentage and pollen staining of T. repens ´ T. uniflorum F 1 hybrids obtained by using AZ 4383/AZ 4437 T. uniflorum (from Turkey) as the pollen parent.

Table 3 .
Numbers of seeds obtained after the first backcrosses.

Table 4 .
Pollen staining of the BC 1 families derived from three F 1 hybrids.