Fine-mapping of DIPLOSPOROUS

In this study, we fine-mapped diplospory in Taraxacum by extending the population of 73 plants used for initial mapping [38] to a total of 2227 offspring plants analyzed here. This population originated from a cross between a sexual maternal plant and a diplosporous, but non-parthenogenetic, paternal plant. The dominant SCARs S8 and S10 were used to screen for recombinants. Presence of the diagnostic PCR products of both, S8 and S10, indicated the presence of the DIP-chromosomal region, whereas the presence of one of the two markers only indicated a recombination within this region (Figure 1). The results showed 1068 plants (47.9%; Table 1) positive for S8 and S10, 1135 plants (51.0%) negative for both markers, and 24 plants (1.1%) recombinant. From the recombinants, nine were positive for S8 and negative for S10 and 15 showed the opposite score. This made the total frequencies of S8 and S10 fragments 48,4% and 48,6%, respectively, both lacking a significant deviation from the expected 1:1 segregation rate (S8+/S8- = 1077:1150, Chi² = 2.39, p > 0.1; S10+/S10- = 1083:1144, Chi² = 1.67, p > 0.1). Our results confirmed the dominant inheritance of S8+/S10+, respectively the DIP-chromosomal region, in a much larger population than screened before. In addition, they reduced the distance between S8 and S10 from 2.7 cM in the previous study to 1.1 cM found here, and indicated that S10 was more closely linked to DIP than S8.

The 24 recombinants between S8 and S10 were phenotyped for their mode of reproduction via crosses with sexual diploid pollen donors followed by analyzing the offspring for their ploidy levels. A total of 20 offspring plants, ten of each of two crosses, were analyzed per recombinant, and chromosome numbers were deduced from flow cytometric analysis (see Methods for exact measurements and definitions). The results showed 20 recombinants (83%; Table 1) to be diplosporous and four being meiotic, clearly deviating from the expected 1:1 segregation rate (DIP:MEI = 20:4, Chi² = 10.67, 0.01 > p > 0.001). Fourteen of the diplosporous recombinants showed strong penetrance of diplospory, that is, ≥ of the offspring plants (18 to 20 out of the 20 measured) showed DNA ratios ≥ 1.79 as compared to a diploid (~tetraploid). The other six showed reduced penetrance of diplospory varying from 80% (n = 2), to 75, 70, 63 and 50%. This was supported by two to four additional crosses for each of these recombinant plants (data not shown).

A random set of 61 plants of the mapping population were phenotyped via crosses in an earlier study [44]. This showed 27 diplosporous and 34 meiotic plants, supporting a 1:1 segregation rate (DIP:MEI = 27:34, Chi² = 0.80, p > 0.1), and a high penetrance of diplospory in the diplosporous individuals: 98.5% of the offspring showed the tetraploid chromosome set. The DIP-linked AFLPs co-segregated with the diplosporous phenotype [38]. Accordingly, we deduced the diplosporous phenotype and a high penetrance of diplospory from the S8+/S10+ genotype in the remainder of the population. The final result showed 1088 diplosporous (1068 S8+/S10+ plants plus 20 diplosporous recombinants = 48.9%; Table 1) and 1139 meiotic plants (51,1%), confirming the expected 1:1 segregation rate (DIP:MEI = 1088:1139, Chi² = 1.17, p > 0.1). In summary, our results showed a strong bias towards diplosporous individuals among the recombinants, whereas they confirmed a 1:1 segregation rate of diplospory versus meiotic reproduction in the entire mapping population. One third of the diplosporous recombinants showed incomplete penetrance of diplospory.

The marker distribution over a larger DIP-chromosomal region was investigated on the basis of 300 individuals, including the 73 plants of the initial mapping population and a random set of 227 plants chosen from the extended population used here. These plants were screened for the PCR-markers S4, Mst78a and Mst53b in addition to S8 and S10. Results are summarized in Table 2. The distance between the most widely separated markers, S4 and Mst53b, was 7.3 ± 3.0 cM, which was comparable to 9.6 cM found previously [38]. The distance between Mst78a and Mst53b was 2.3 ± 1.7 cM, which was in a better agreement with the 5-10 cM found at non-DIP homolog's in Taraxacum in earlier studies [42–44] than their close linkage in our initial DIP-map [38]. It also non-significantly deviated from the 6.3 cM between these Mst alleles at the non-DIP homolog's in the diploid sexual parent S2.125 in this study (Chi² = 1.86, p > 0.1). The distance between the Mst78 and Mst53 alleles at non-DIP homolog's in the tetraploid diplosporous parent PAX varied from 4.7 cM to 10.3 and 12.7 cM, the latter two significantly deviating from the 2.3 cM found at the DIP-homolog. Excluding the results from one of the five flower heads, because it showed a strong bias towards recombination's between two particular allele combinations, reduced the distance between the Mst alleles to 2.0 cM at the DIP-chromosome and 3.7, 4.3 and 6.7 cM at the non-DIP homolog's in PAX, lacking a significant deviation between the extremes (Chi² = 2.54, p > 0.1). In summary, our results confirmed the occurrence of recombination in the DIP-chromosomal region and lacked strong support for the existence of a 'cold spot', a region in which recombination is suppressed, at the Mst78 and Mst53 loci located at ~3.5 cM from DIP at one flank.

An overview of the mapping results is given in Figure 2. Map A is redrawn from our previous study [38] with some minor updates [[16], this study]. It has a total length of 18.6 cM and shows a regular distribution of markers over the DIP-chromosomal region. DIP is located at one third of the map and markers co-segregating with DIP are lacking. Map B zooms in to a region of ~8 cM around DIP, based on 300 plants and five PCR markers. It confirms the occurrence of recombination in this region close to DIP and lacks strong evidence for local suppression of recombination at the Mst78-Mst53 region. Map C shows the fine-map of DIP based on the 24 recombinants between S8 and S10 out of the 2227 plants analyzed and the six most closely linked AFLPs. It shows partial resolution of the AFLP cluster at one flank of DIP and no further resolution of the AFLP cluster at the other flank of DIP. The length of the smallest DIP-chromosomal region was reduced to 0.6 cM, and the markers most closely linked to DIP are S7 and the cluster with S10 at 0.4 cM and 0.2 cM, respectively.

Segregation distortion and incomplete penetrance of diplospory

In order to investigate whether the biased distribution towards diplosporous reproduction in the recombinants and the occurrence of a reduced penetrance of diplospory in some of them was correlated to a particular DIP-chromosomal region, the 24 recombinants were classified according to their place of recombination and DIP-flank involved. A total of six recombination types were possible, those between S8 and S7 (types I and VI; Figure 3), S7 and DIP (types II and V), and DIP and S10 (types III and IV), and their frequencies and accompanying average penetrance of diplospory were indicated (Figure 3). The scheme visualized the segregation distortion towards diplosporous recombinants (20 out of 24, see above) and showed that this was irrespective of the DIP-flank involved (types I versus VI, II versus V, and III versus IV). Incomplete penetrance of diplospory was associated with a region close to DIP at one flank (type III). Zooming in to the region between S7 and S10 showed an enforcement of the skewed distribution to 13 diplosporous versus one meiotic recombinant. Further zooming in to a region that covers ~0.2 cM at each side of DIP suggested that all recombinants are diplosporous here (Figure 3, boxed). Our results indicated that meiotic recombinants with a DIP-flank approaching DIP were either not formed or behaved diplosporously (type IV became included in type II and type V in type III, see discussion). They suggested that a reduced penetrance of diplospory is associated to a region close to DIP, particularly at the flank towards S10.

To further investigate the association of a reduced penetrance of diplospory to (a) particular chromosome region(s), the frequencies of offspring per chromosome number classes were depicted in a graph (Figure 4). This visualized the highest frequencies of reduced offspring in recombinants of type III (Figure 4A, black bars) and showed that these reductions were relatively severe, that is, they resulted in offspring with ≤ 24 chromosomes, indicating the formation of haploid to diploid eggs. Type III included the three plants with lowest penetrance of diplospory (50-70%), whereas the other three recombinants with reduced penetrance were of Types I and II (Figure 4A, grey bars). These plants showed less reduced penetrance of diplospory (75-80%) and also milder reductions, that is, they resulted in offspring with 24-32 chromosomes, indicating the formation of di- to triploid eggs. Recombinants of type III were depicted individually to further visualize the pattern of reduced penetrance of diplospory (Figure 4B). This showed strong reductions of diplospory in plants R04, R08 and R11 (stripes and grey bars) and normal penetrance in R02 and R21 (white and black bars). The summary of results suggested that diplosporous reproduction in T. officinale is regulated by a more complex locus rather than a single DIP gene (see discussion), with the region towards S10 particularly being important for full penetrance of diplospory.

Comparative analysis between DIP- and meiotic homolog's

Recombination in the DIPLOSPOROUS-chromosomal region and the absence of overall allele sequence divergence make DIPa good candidate to isolate via a map-based cloning approach

Our data indicate that suppression of recombination plays a minor role in the DIPchromosomal region in T. officinale. This is supported by the equal distribution of markers over the DIP-map (Figure 2), the absence of markers co-segregating with diplospory, and the non-significantly deviating distances between the Mst78 and Mst53 alleles at DIP and non-DIP homolog's (Table 2 and Results). This result from the large segregating population confirms the conclusion of our initial mapping study [38], whereas possible local suppression of recombination at the Mst-region close to DIP is not supported. The occurrence of recombination in the DIP-chromosomal region is unique in comparison to other mapping studies of apomeiosis to date, that all showed strong suppression of recombination and in some cases hemizygosity at these loci (see Background) [reviewed by [16, 25]]. This has frustrated map-based cloning efforts of genes associated to apomixis in the species concerned. In one of the best examples among them, FISH experiments indicated that the apospory-specific genomic region (ASGR) in Pennisetum squamulatum physically corresponds to ~50 Mbp or a quarter of a chromosome [54]. A recent mapping study in P. squamulatum, that made use of markers designed on long terminal repeat (LTR) sequences of ASGR-abundant transposon's, did also not resolve this region [55]. The occurrence of recombination around DIP in T. officinale offers a good opportunity to isolate an apomeiosis gene via further molecular marker searches based on the recombinants and BAC-walking from the markers most closely linked to DIP.

The present delimitation of DIP by S7 and S10 covers 0.6 cM. In comparison to tomato, which has a comparable genome size to T. officinale and for which 1 cM was found to correspond to 0.3-0.4 Mbp [56], the DIP-region is estimated to cover a physical length of ~200-300 Kb. With an average BAC-size of the T. officinale BAC-library of ~115 Kb [K Vijverberg et al. unpublished results] and supposing halfoverlapping BACs, this region could be closed by 4-6 BACs, making the isolation of DIP in Taraxacum feasible.

Our data show that overall ASD is lacking at the DIP-chromosomal region. This is supported by the occurrence of recombination (see former paragraph) and by comparative sequence data at three of the DIP-linked marker regions analyzed, S4, S7 and S11 (Figure 2; Table 3). At the fourth marker region, S10-flank, and also at S10, preferential amplification of the DIP-homolog was indicated. This, and the clustering of S10 with two other AFLPs, may be signs of ASD and hemizygosity at this region close to DIP, although, the severity and locality of these signs are yet unknown. Increased ASD in regions associated with asexual reproduction can be expected as a result of the accumulation of mutations in the absence of 'homogenization' with homolog's via recombination [39–41]. The strong suppression of recombination and/or hemizygosity found at most loci associated with apomixis [e.g., [29, 54]] support this hypothesis. Also in sexuals, the occurrence of ASD in some regions of the genome has been indicated, e.g., haplotypic variability in maize was shown to strongly affect the frequency and distribution of recombination events [57]. Since Taraxacum is hermaphrodite, and apomixis concerns the female reproductive pathway only, recombination during pollen meiosis offers the possibility to retard ASD. By chance, overall this is expected to occur less frequently closer to DIP. Computer simulations of the depth of asexual reproduction history along a chromosome that carries a dominant single dose DIP-allele support this model [42]. It shows that the depth of asexual reproduction history accumulates with the number of generations also under sexual-asexual cycling conditions and is deepest at and close to DIP. This depth reduces at random generation times and positions across the flanks by incidental recombination events. When ASD becomes sufficiently pronounced, it will prevent for recombination also during pollen meiosis, which can lead to further increase of ASD. In addition to this gradual process, chromosome rearrangements may arise by chance and result in suppression of recombination immediately. In a heterozygous state, these mutations and rearrangements can accumulate without effect on apomictic seed production [45]. Our data are congruent with this model by suggesting that S4 and S7 at one flank of DIP and S11 at the other flank of DIP have been homogenized by relatively recent recombination events, and the region around S10, located closer to DIP, may have a deeper history of asexual reproduction. Since recombination in the DIP-region between S7 and S10 is still possible (Figure 3), the possible ASD at the S10-region may be explained by a local rearrangement.

An earlier study in T. officinale suggested that a certain level of ASD is associated with DIP [58, 42], by showing severe segregation distortion of the Mst78a and Mst53b alleles that are linked to DIP in haploid pollen grains, but not in diploid ones. A plausible explanation for this is that the DIP-allele is not transferred via haploid pollen due to pollen lethality as a result of linked deleterious mutations [45]. These mutations would be masked by (the) non-DIP allele(s) in a diploid or polyploid stage. A study in Tripsacum dactyloides also suggested that genes for apomixis might be linked to loci that are subjected to segregation distortion in haploid but not diploid gametes [32]. When apomixis genes cannot be transferred via haploid pollen, this could also explain the prevalence of apomixis in polyploids [42]. Our data indicate that overall ASD is absent at the DIP-locus, but we argue that recombination during pollen meiosis may have resulted in a mosaic pattern of different depths of asexual reproduction history along the DIP-chromosome (see former paragraph). In addition, local rearrangements may have accumulated in regions associated with asexual reproduction. The linkage of DIP to a mutation load can then be explained by its confinement to (a) particular region(s) close to DIP. Alternatively, the non-transfer of DIP via haploid pollen is associated to the DIP-allele itself. A pleiotropic recessive pollen lethal effect linked to apospory in Ranunculus auricomus was suggested by Nogler [1]. Possibly, a similar pleiotropic effect has evolved in the DIP-gene or is associated to asexual reproduction in general, however, this needs further investigations.

The occurrence of recombination and absence of overall ASD at the DIP-chromosomal region suggest that diplospory and apomixis are relatively recent in T. officinale. The clustering of markers found at the PAR-locus [PJ Van Dijk et al. unpublished results], however, seems to be in contrast to this conclusion. The evolution of either DIP or PAR alone is unlikely, since they lead to increased ploidy levels with each generation and haploid phenotypes, respectively, both expected to be evolutionary dead-ends in nature. The differences in marker clustering found at the DIP and PAR loci could possibly be explained by a different recombination history around these loci by chance [[42], see above], that has resulted in higher levels of ASD at the PAR- than DIP-locus. An alternative explanation is that the differences result from a difference in chromosome rearrangements at the DIP- and PAR-locus as compared to their non-DIP and non-PAR homolog's. Also differences in mitotic crossover and gene conversion may have played a role at these loci. Particularly one of these latter processes could have been involved in the origin of a duplication/inverted repeat in the DIP-locus as is postulated in Figure 5 (discussed below). The results possibly indicate that apomixis in Taraxacum is older than suggested by the DIP-chromosomal region alone.

In summary, our results suggest that ASD is absent in a wider region around DIP, non-pronounced close to DIP, and possibly present locally at the S10- and possible other regions linked to DIP as well as at the PAR-locus. This pattern of different levels of ASD linked to DIP and PAR is congruent with the outcome of computer simulations of recombination during pollen meiosis in regions associated with asexual reproduction and, in addition, different histories of local rearrangements. A practical implication of the absence of clear ASD at the DIP-locus is that non-DIP homolog's in the BAC-library can be used in the BAC-walking process in order to isolate DIP.

Segregation distortion towards diplosporous recombinants and incomplete penetrance of diplospory in some of them indicate a duplicated DIP-locus

We found a five times over-representation of diplosporous plants among the 24 recombinants (Figure 3), whereas the 1:1 segregation rate of diplosporous versus meiotic reproduction was confirmed in a random set of 61 plants of the mapping population by phenotyping via crosses. This 1:1 segregation was also confirmed in the remainder of the 2227 plants analyzed, according to deduction from the S8/S10 genotypes (Table 1). The over-representation of diplosporous plants is thus indicated to be specific for plants with a recombination between S8 and S10. The bias towards diplosporous reproduction increases in recombinants with a DIP-flank closer to DIP at both flanks towards the presumed single DIP gene (Figure 3). The apparent segregation distortion could find its origin in one of two crosses: (i) the parental cross: S2.125 (2x) × PAX (4x, D-allele) (Figure 1A) or (ii) the crosses used for phenotyping: F1 (3x, including D/d-recombinants) × sexual (2x). In the parental cross, possible explanations for this bias are a preferential transfer of DIP to recombinant offspring plants or a decreased viability of meiotic recombinant F1's. In the crosses used for phenotyping, the bias is possibly explained by a shift towards diplosporous reproduction in recombinants with a DIP-flank approaching DIP /the presence of a partial DIP-locus or by a misclassification of meiotic recombinants in diplosporous ones with reduced penetrance. In a tetraploid pollen donor, as is the case in the parental cross, normal bivalent formation and meiosis occur [e.g., [49]], making recombination events along the DIP-chromosome possible. In the diploid pollen as well as triploid offspring, a possible mutation load linked to DIP, as can be a result of long-term asexual reproduction, will be masked (see former paragraphs). If expressed, it will result in an under-representation of diplosporous individuals rather than an over-representation. If uncoupled, a mutation load could result in a reduced number of meiotic recombinants when it is selectively expressed in a sexual background. This does, however, not explain the reduced penetrance of DIP found in some of the recombinants and also not the effect found at both flanks: recombinants of both types IV and V are lacking (Figure 3). A modifier linked to DIP with a lethal effect in a sexual background, either in the pollen or early embryo stage, can be another possibility. Although uncoupling of the modifier may result in a lower expression of diplospory, also here, a single modifier does not explain the bias found at both flanks. The phenotyping of 61 random offspring plants also lacks signs of linkage of either favorable or deleterious mutations to diplospory. The segregation distortion towards diplosporous recombinants is, therefore, not strongly supported by a preferential transfer of diplospory or a decreased viability of meiotic offspring in the parental cross.

The segregation distortion of diplospory in the recombinants as well as the reduced penetrance of diplospory in some of them are better explained in the crosses used for phenotyping. Meiosis in these crosses occurs in triploid megasporocytes, making meiosis difficult and leading to high percentages of aneuploid gametes [e.g., [49]], whereas diplosporous non-reduction is unhindered. Although offspring plants resulting from aneuploid gametes survive (Figure 4), their number might be (slightly) reduced. The percentages of diplosporous offspring from plants with incomplete penetrance of diplospory may then, actually, be (slightly) lower. Accordingly, the six diplosporous recombinants with 50 to 85% penetrance of diplospory (Figures 3, 4), out of the total of 20, represent a minimum. In Taraxacum, unreduced, balanced triploid sperm cells can arise from a disturbed triploid true meiosis [e.g., [49, 59], own observation], although at low frequencies not exceeding the 10%. Unreduced pollen formation is also known from other polyploidy species [60], and the few existing data for unreduced eggs suggest that the natural frequency of non-reduction is similar in megasporogenesis [[60] and references therein]. Occasional formation of triploid female gametes in meiotic triploid recombinants is, therefore, expected and indicated (Figure 4A, types V and VI). However, frequencies of ≥ 50% are too high to be explained without the action of diplospory. Moreover, segregation distortion and incomplete penetrance of diplospory was found in the recombinants only and not in the other diplosporous offspring of the mapping population. Apomixis in Taraxacum is usually obligate or close to obligate [48–50], also indicating full penetrance of diplospory under natural conditions. Together, this argues against misclassification of meiotic individuals as diplosporous ones with reduced penetrance. Our data is better explained by the presence of a relatively large DIP-locus in Taraxacum that, in part, still results in diplosporous reproduction and, depending on the parts available, could result in incomplete penetrance of diplospory.

A hypothetical constitution of the DIP-locus that explains the segregation distortion as well as incomplete penetrance of diplospory in some of the recombinants is given in Figure 5. Since the over-representation of the diplosporous phenotype is irrespective of the DIP-flank involved, a locus with a minimum of two DIP-genes is postulated. The absence of plants of recombination types IV and V (Figure 3) is then explained by their misclassification in types II and III, respectively, that is, the recombined allele excludes the putative single DIP-gene (Figure 5, at 6.0 cM), but still induces diplosporous reproduction. The different penetrance of diplospory is best explained by assuming one or two enhancers or cis-regulatory elements (Figure 5). Another possibility is additional DIP-genes that may result in copy number variation in the different recombinants (not shown), however, the number of copies increases to explain our data. A certain distance between the genes, and the enhancers/regulators, is supposed in order to give rise to the different recombinants within the proposed length of ≥ 200-300 Kb between S7 and S10. Since incomplete penetrance of diplospory is most pronounced in three plants of type III (Figures 3, 4), that have their recombination between DIP and S10, essential elements for obtaining diplospory are indicated to be located in this region. We postulate that a minimum requirement for obtaining full penetrance of diplospory is the presence of a complete DIP1 gene in combination with En1 (or cis-regulator1), whereas DIP1 or DIP2 alone or DIP2 in the presence of En2 only, result in reduced penetrance of diplospory.

Our results do not support a complex DIP-locus with different functional genes. Recently, the osd1/Atrec8/Atspo11-1 triple mutant in Arabidopsis was shown to produce a highly penetrant diplosporous phenotype: mitosis replaces meiosis (MiMe) [23]. In a similar way, the DIP-locus could consist of different functional genes. Genes with a combined action to produce a complex phenotype may be tightly linked as a result of co-evolution, e.g., like the self incompatibility loci in Brassica [61, 62]. The overrepresentation of the diplosporous plants among the recombinants, however, argues against such a complex locus with different functional genes, since recombination would destroy the diplosporous phenotype.

Incomplete penetrance of apomeiosis was also found in Erigeron annuus [63] and Poa pratensis [35] and suggested to be correlated to a genetic background. In the study in E. annuus, a diplosporous triploid plant was pollinated by a sexual diploid, indicating that the genetic variation originated from the paternal plant and was not the result of maternal and epigenetic effects. The penetrance of diplospory was 92% in the triploid maternal parent and varied from 41 to 89% in the tetraploid offspring plants. Noyes [63] suggested either the presence of (trans-acting) modifiers that act directly on the expression of diplospory, novel epistatic interactions, or a combination thereof. The Apospory initiator (Ait) gene in P. pratensis [35] is a dominant gene needed for apospory and leading, together with four other genes, and depending on their differences in expressivity and interactions, to apomictic reproduction. Ait is of the aposporous type, not modifying meiosis, but depending on the formation of aposporous initials. The penetrance of Ait was found to vary between 0 and > 80%: it was absent in the maternal plant, but > 50% penetrant in some of the offspring originating either from selfing or from crossing with a sexual. Since incomplete penetrance of diplospory was found in some of the recombinants only and not in the diplosporous offspring plants without a recombination between S8 and S10, a genetic background that modifies the expression of DIP is not supported. Our results are better explained by a direct effect of recombination in the DIP-locus, as proposed in Figure 5, and not as a result of trans-acting effects.

Duplicated genes have often been associated with apomixis, but in the context of polyploidy [2] and not in the form of a tandem or inverted repeat at a single chromosome. Possible exceptions to this are the recently found duplicated BABY BOOM (BBM)-like gene and some duplicated putative protein-coding regions in the ASGR [20]. Hypothesized functional roles of ploidy include epigenetic gene deregulation, e.g., through changes in methylation [64], and asynchronous gene expression [65]. Others proposed that these complexities evolved upon establishment of apomictic reproduction and may be secondary [25]. In Taraxacum [58, 42] and Tripsacum [32], segregation distortion of apomixis genes in haploid but not diploid gametes also suggest a secondary role of polyploidy. Whether a segmental duplication, as is postulated here, has a functional relationship to those hypothesized for a polyploid genome is yet unknown, but it is interesting to observe a possible gene dosage effect related to apomeiotic reproduction. Eventually map-based cloning of the DIP-locus will reveal the real structure of this locus and will give more insight into its function.

Diplospory results in FDR gametes, which are considered to be superior to 2n-gametes from second division restitution (SDR) in that they transfer all or a most of the parental heterozygosity and epistasis intact to their progeny. SDR gametes, on the other hand, result in near-homozygous offspring, varying in their chromosome constitution and possible heterozygosity only at the ends of the chromosome. Although both types allow for raising the ploidy level in artificial crosses, FDR gametes allow for pyramiding mutations or valuable characteristics in the same genetic background. DIP as well as the other apomixis related apomeiosis genes are dominant alleles (reviewed by 16). Whether they represent a new function or act via the inactivation of normal functions, i.e., deregulate the sexual pathway, is since long be debated [e.g., [3, 64]). A few genes that showed deregulation of female meiosis in sexual model species, resulting in FDR-gametes, have been reported to date. These include mob1 in Medicago sativa [66, 67], dyad in Arabidopsis [21, 22], and a non-reduction mutant (nrm) in maize [64]. In contrast to DIP, these genes are all recessive. The mutants give the desired phenotype, but possibly via a different (de-)regulation than found in apomicts. Together with the useful applications of a dominant FDR-gene in plant breeding, isolation of DIP and finding out its function is desired and highly intriguing.

Plant material

The mapping population of T. officinale used for initial mapping of DIPLOSPOROUS [38] was extended to a total of 2227 individuals. The population originated from a tetraploid diplosporous, but non-parthenogenetic, pollen donor (PAX, 2n = 4x = 32) and a sexual diploid maternal parent (S2.125, 2n = 2x = 16). PAX was the result of a cross between a non-apomictic, but diplosporous, triploid hybrid (H6-3 [51]) and a sexual diploid paternal plant. The tetraploid PAX produced sufficient diploid, reduced and recombined pollen to perform segregation studies on diplospory. The triploid offspring of PAX × S2.125 segregated c. 1:1 for diplosporous:meiotic reduction [44, 38]. Seedlings were grown in the greenhouse and screened for recombination in the DIP-chromosomal region (described below). Recombinant plants were further raised in the greenhouse, phenotyped for diplospory, and analyzed for additional molecular markers in the DIP-chromosomal region.

Screening for recombination in the DIP-chromosomal region

DNAs were screened for recombination between two PCR-markers that were derived from the dominant SCARs S8 and S10, spanning DIP in a total of 2.7 cM [38]. Total DNAs were isolated from c. 5 mg leaf tips via the 'RETCH method', using the CTAB procedure as described by Rogstad [68] and the modifications by Vijverberg et al. [38]. In this procedure, leaf tips were transferred to tubes (Costar 1.2 ml tubes 4408 and caps 4418) that contained two steel balls (bicycle, 1/8" in 60 gross, Tiangin, China) and were placed in racks (Costar) on ice. After harvesting, 500 μl CTAB-buffer was added to each tube, and leafs were homogenized via shaking the racks 4× 1 min. at 30 rps in a RETCH shaker (MM200, Düsseldorf, Germany). At the end of the isolation procedure, DNAs were resuspended in 100 μl T_0.1E (10 mM Tris, 0.1 mM EDTA, pH = 8.0). Quality and quantity were checked by measuring the optical densities at 260 and 280 nm and analyzing 2 μl on 1% agarose gels.

PCRs were performed in 96-wells plates, using 2-10 ng DNA, 2.5 μl 10× PCRbuffer, 2.5 mM MgCl₂, 0.5 mM dNTPs, 1% polyvinylpyrrolidone (PVP-40, Sigma- Aldrich), 0.2 mM of each primer, and 0.5 Units Taq DNA Polymerase (Expand High Fidelity PCR Systems, Roche Applied Science) in a total volume of 25 μl. One drop of mineral oil (Sigma-Aldrich) was used to overlay the reactions and PCRs were performed following 1× 2 min. 94°C, 38× (30 sec. 94°C, 30 sec. annealing temperature, and 1 min. 72°C), 1× 5 min. 72°C, and a cooling down to 12°C. Annealing temperatures of S8 and S10 primer pairs were 60°C and 57°C, respectively, and primer sequences are described in Vijverberg et al. [38]. PCR products were analyzed on 1.5% agarose gels, with diagnostic bands showing lengths of 446 bp and 168 bp respectively (Figure 1).

DIP-phenotyping

Plants that showed recombination between S8 and S10 were phenotyped for diplosporous versus meiotic reduction via crosses, using sexual diploids as pollen donors [44]. The offspring was analyzed for their ploidy levels as deduced from flow cytometric analysis [59]. Diplosporous recombinants were expected to produce tetraploid offspring only, being the sum of a non-reduced triploid egg cell plus a haploid sperm cell [[51], type C plants]. They showed a DNA ratio = 2.0 (= 32 chromosomes) as compared to a diploid reference species. Meiotic recombinants produced a range of di-, tri-, and up to a few tetraploid offspring, representing the result of an aberrant meiotic reduction, typical for a triploid, plus a haploid male genome [[51], type A plants]. This was indicated by DNA ratios = 1.0 to 1.5 and some higher values. Ten offspring plants were analyzed of each of two crosses per recombinant. The penetrance of diplospory was calculated as the percentage of offspring out of the 20 plants measured that showed a DNA ratio ≥ 1.79 (≥ 29 chromosomes). This was after correction for the triploid recombinant parent by setting it's ratio to 1.5 (original ratios varied from 1.48 to 1.58). In natural triploids, the percentage of tetraploid offspring as a result of aberrant meiosis during microsporogenesis does not exceed 10% [[44], own observations]. As a 'wide' threshold, recombinants that showed < 30% tetraploid offspring in addition to reduced offspring were regarded meiotic, whereas the remainder was defined as diplosporous with a complete or reduced penetrance.

Genetic fine-mapping of DIP

Plants that showed a recombination between S8 and S10 were analyzed for AFLPs known to be linked to DIP [[38], Figure 2A]. DNA was isolated from 0.5-0.6 g fresh leaves, using the CTAB protocol described by Rogstad [68], with the modifications suggested by Vijverberg et al. [38]. Isolated DNA was resuspended in 500 μl T_0.1E, checked for content and purity as described above, and 200 ng was used for AFLP analyses. AFLPs were generated according to the protocol of Keygene N.V. (Wageningen, The Netherlands) version 2.2 described by Vos et al. [69], including the modifications described earlier [38]. AFLPs tested included those of the unresolved marker groups most closely linked to DIP: A3-E42M50-440 (= S7), A3-E37M51-495 (= S8), and A3-E42M54-295 at one flank of DIP and A4-E35M52-235 (= S9), A4-E38M48-215 (= S10), and A4-E45M53-090 at the other flank of DIP. Marker names represent the AFLP groups (A3, A4), enzymes (E = EcoRI, M = MseI), selective nucleotides in alphabetical order (31 = AAA, 32 = AAC, 33 = AAG, etc.), and length of the fragments in base pairs. Seven markers at larger distances from DIP were analyzed as controls: A1-E40M60-505 (= S4), A2-E40M52-310, A2-E46M52- 435, A5-E37M59-135, A6-E41M57-85 and the microsatellite alleles A6-Mst78a and A6-Mst53b. AFLPs were analyzed on a LI-COR IR² automated sequencer (Biosciences, USA) and the relevant bands scored manually on the gel-output. Msts were analyzed on an ALF DNA sequencer (Pharmacia) and scored manually.

Investigating the marker distribution over a wider DIP-chromosomal region

A subset of 300 individuals randomly chosen from the mapping population was, apart from S8 and S10, also screened for the polymorphic PCR markers S4, Mst78, and Mst53 that were located at larger distances from DIP, spanning a total of 9.6 cM. PCRs of S4 were performed as described for S8 and S10, using an annealing temperature of 53°C. Subsequently, 5 μl products were digested overnight at 55°C with 1 unit BseGI (Fermentas GMBH, Germany) in 1× accompanying buffer in a reaction volume of 15 μl. Digests were analyzed on 1.5% agarose gels, showing undigested products of c. 430 bp and digested ones of 270 and 160 bp. The DIP-associated S4 products lacked the BseGI recognition site, so that diplosporous individuals showed both, the undigested as well as digested products, whereas meiotic individuals showed digested products only. PCRs of Msts were performed as described by Falque et al. [43] and products were analyzed on a LI-COR IR² automated sequencer (Westburg, Leusden, The Netherlands). Since Mst-markers are co-dominant markers, with Mst78a and Mst53b being linked to DIP, they have the advantage to also resolve information about the meiotic homolog's involved.

Data analysis and mapping

AFLP-, Mst-, and phenotypic data was collected in a binary data base in Excel. Linkage analyses were performed on datasets: A = the previous data set of 73 plants and 36 markers; B = the medium data set of 300 plants and five (to 36) markers; C = the entire data set of 2227 and two (to 36) markers. The data of each of the three data sets was sorted manually in such a way that recombination events were minimized. Linkage analyses were also performed by using Joinmap^® 3.0 [70], with either all data or only the recombinants included. For the Joinmap analysis, the BC population type was used, with the Kosambi mapping function and threshold values for LOD > 1.0 and REC < 0.40. For some sets of recombinants, LOD and REC threshold values were relaxed to > 0.3 and < 0.45, respectively. An updated map A was drawn on the basis of the Joinmap output [[16], this study], confirming our previous results [38]. Maps B and C were drawn on the basis of the manual sortings and manual calculations of recombination frequencies. Since Joinmap analyses for these two data sets were inaccurate due to many missing values for most individuals, this was used as a guidance only. Confidence intervals of the map distances were calculated according to Wu et al. [71], regarding the chance that the true recombination fraction (r) is in the 95% interval under ± 1.96 √(r[1-r]/n) conditions, with n being the number of plants analyzed.

Comparison of marker regions of DIP- versus non-DIPhomolog's

Sequences were obtained for marker regions S3, S4, S7, S10-flank, A4-flank and S11, covering 300-400 bp regions located at different distances from DIP at both flanks (Figure 2A; Table 3). Individuals analyzed were the sexual diploid maternal parent S2.125, representing two meiotic homolog's (d1 and d2), a diplosporous triploid offspring plant, F1.15, representing the DIP-chromosome, one meiotic homolog originating from S2.125, and a second meiotic homolog (D, d1 and d3), and a natural triploid apomict, A68, containing the DIP-chromosome plus two meiotic homolog's (D, d4 and d5). PCR products were obtained as described above and in Vijverberg et al. [38], using as S7 reverse primer: 5'-TACTCACTTCCCGATCAACTCAC-3', and as S10-flank forward and reverse primers: 5'-TAGACTTGATGCTCCTTGACTTGG-3' and 5'-AAGGCTTCACATGAGGCAGTAAG-3', respectively. Products were cloned by using the pGem T Vector system II (Promega), according to the manuals description. Per individual, between 10 and 20 clones were sequenced per marker region (BaseClear, Leiden, The Netherlands), using an M13 primer. Sequences were compared by using the DNAstar software (Lasergene, London, UK) or CLC sequence viewer v6.2 (CLCbio, Aarhus, Denmark).