(Diplospory) in Apomictic Dandelions (Taraxacum Officinale

Formation of Unreduced Megaspores (Diplospory) in Apomictic Dandelions (Taraxacum ofﬁcinale, s.l.) Is Controlled by a Sex-Speciﬁc Dominant Locus

Peter J. van Dijk1 and J. M. Tanja Bakx-Schotman Department of Plant Population Biology, Netherlands Institute of Ecology (NIOO-KNAW), Centre for Terrestrial Ecology, Boterhoeksestraat 48, 6666 GA Heteren, The Netherlands Manuscript received May 9, 2003 Accepted for publication October 1, 2003

ABSTRACT In apomictic dandelions, Taraxacum officinale, unreduced megaspores are formed via a modified meiotic division (diplospory). The genetic basis of diplospory was investigated in a triploid (3x ϭ 24) mapping for diplospory and meiotic reduction. This population 1:1ف population of 61 individuals that segregated was created by crossing a sexual diploid (2x ϭ 16) with a tetraploid diplosporous pollen donor (4x ϭ 32) that was derived from a triploid apomict. Six different inheritance models for diplospory were tested. The segregation ratio and the tight association with specific alleles at the microsatellite loci MSTA53 and MSTA78 strongly suggest that diplospory is controlled by a dominant allele D on a locus, which we have named DIPLOSPOROUS (DIP). Diplosporous plants have a simplex genotype, Ddd or Dddd. MSTA53 and MSTA78 were weakly linked to the 18S-25S rDNA locus. The D-linked allele of MSTA78 was absent in a hypotriploid (2n ϭ 3x Ϫ 1) that also lacked one of the satellite chromosomes. Together these results suggest that DIP is located on the satellite chromosome. DIP is female specific, as unreduced gametes are not formed during male meiosis. Furthermore, DIP does not affect parthenogenesis, implying that several independently segregating genes control apomixis in dandelions.

EXUAL plant reproduction involves a reduction of vides mitotically to form the megagametophyte or em- S the somatic chromosome number by meiosis fol- bryo sac, containing, among others, a reduced egg cell lowed by a restoration of the somatic chromosome num- and two reduced polar nuclei. Fertilization of the egg ber by fertilization. Most seed plants reproduce sexually; cell by one sperm cell of the pollen grain generates a -of angiosperm species reproduce asex- diploid embryo, while fertilization of the two polar nu %0.1ف ,however ually through seed, a process referred to as apomixis clei by the second sperm cell generates the triploid endo- (Nogler 1984; Mogie 1992). In apomictic plants, em- sperm. Gametophytic apomicts differ from sexually re- bryos develop parthenogenetically (without fertiliza- producing plants in that they produce unreduced tion) from unreduced somatic or gametophytic cells. embryo sacs. Circumvention of meiosis in gametophytic Barring mutations, apomictic offspring are genetically apomicts can be achieved either through apospory, in identical to the mother plant. Apomixis is of interest which meiosis is completely bypassed, or through diplo- for the study of the maintenance of sex, one of the spory, in which meiosis is modified (Nogler 1984; great and long-standing enigmas in evolutionary biology Asker and Jerling 1992; Koltunow 1993). In apospor- (Barton and Charlesworth 1998; West et al. 1999; ous apomicts an unreduced megagametophyte is pro- van Dijk and van Damme 2000). Apomixis also holds duced ectopically from the somatic tissue of the ovule, great promise for agriculture, because it fixes traits re- without an intervening megasporogenesis. This apospor- gardless of their complexity in genetic control (Vielle ous megagametophyte often coexists with and outcom- Calzada et al. 1996; Spillane et al. 2001). Apomixis does petes the reduced megagametophyte. In diplosporous not occur in the major crops except citrus, but it is apomicts, an unreduced gametophyte is produced from conceivable that it can be introduced into crops in the an unreduced megaspore. This unreduced megaspore future via genetic engineering. A solid understanding results from either a mitotic-like division (mitotic diplo- of the genetic regulation of apomixis is therefore impor- spory) or a modified meiosis (meiotic diplospory). In tant from both an evolutionary and an agricultural point both apospory and diplospory, the unreduced egg cell of view. develops parthenogenetically into an embryo. Endo- In sexual reproduction a diploid somatic cell, the sperm development in apomictic species may rely on megaspore mother cell, undergoes meiosis to produce fertilization of the reduced or unreduced polar nuclei four reduced megaspores. One of these megaspores di- (pseudogamy) or may proceed in the absence of fertilization (autonomous endosperm formation). Studies on the inheritance of apomixis are compli- 1Corresponding author: NIOO-KNAW, P.O. Box 40, 6666 GA Heteren, cated by the fact that the trait cannot be crossed into The Netherlands. E-mail: [email protected] sexual plants using an apomictic female. However, most

Genetics 166: 483–492 ( January 2004) 484 P. J. van Dijk and J. M. T. Bakx-Schotman apomictic species produce viable pollen and can there- test these competing models have not been published. fore be used as pollen donors in crosses with sexuals. Crosses between diploid sexuals and triploid apomictic The inheritance of apospory has been investigated in a dandelions generate low numbers of hybrids because considerable number of plant species and, remarkably, of the low frequencies of haploid and diploid pollen is always inherited as a monogenic dominant trait (Gri- grains produced by the irregular pollen meiosis of trip- manelli et al. 2001; Grossniklaus et al. 2001). Much loids. Moreover, the high loads of inviable or weak aneu- less is known about the inheritance of diplospory. To ploid pollen produced by triploids induce high selfing date the genetic control of diplospory has been exten- rates in normally self-incompatible sexual diploids sively studied in only two species: a grass species, Tripsa- (Morita et al. 1990; Tas and van Dijk 1999). The few cum dactyloides, that reproduces by mitotic diplospory hybrids obtained in these diploid sexual-triploid apo- (Grimanelli et al. 1998) and a composite species, Erig- mict crosses could be interpreted as phenotypic recom- eron annuus, that reproduces by meiotic diplospory (Noyes binants lacking different elements of apomixis (van and Rieseberg 2000). In these two species diplospory Dijk et al. 1999, 2003). Although this finding suggests is also inherited as a dominant monogenic trait. that apomixis in Taraxacum is controlled by several The genetic control of apomixis in another meiotic independent genes, the numbers of hybrids were too diplosporous apomict, the common dandelion, Taraxa- low for segregation ratio analysis. We now have circum- cum officinale, is disputed. Both diploid sexuals (2x ϭ vented this problem for diplospory by the use of a tetra- 16) and polyploid autonomous apomicts (mostly trip- ploid nonparthenogenetic diplosporous pollen donor loid, 3x ϭ 24) occur within this species (Richards 1970, (4x ϭ 32), which produces higher quality pollen. A 1973). Richards (1970, 1973) proposed a model for cross between a sexual diploid and this tetraploid pollen the control of apomixis in Taraxacum with two unlinked donor generated a large triploid population that segre- dominant genes, one for diplospory and one for parthe- gated for diplospory. Here we show that diplospory in nogenesis. This model was based on the observations Taraxacum is inherited as a monogenic dominant trait, of Sørensen and Gudjońsson (1946) and Sørensen consistent with Richards’ model. Moreover, both genetic (1958) that loss of a chromosome in disomic offspring mapping and cytogenetic analysis indicate that the diplo- (2n ϭ 3x Ϫ 1) from triploid apomicts results in the loss spory locus is located on one of the satellite chromosomes, of either diplospory or parthenogenesis, depending on as was originally suggested by Sørensen’s (1958) study. which chromosome is lost. On the basis of the chromosome terminology used by Sørensen and Gudjońsson (1946), Richards (1970) suggested that the genotypic MATERIALS AND METHODS constitutions for diplospory and parthenogenesis would Plant materials: A triploid population segregating for diplo- be Hhh and Ddd, respectively. spory was made by crossing a diploid sexual seed parent (S2- Mogie (1988, 1992) challenged the dominant model 125) with a tetraploid diplosporous pollen parent (PAX). The because of the poor quality of the previous observations: origin of these two plants is described in detail in Falque et al. (1998) and van Dijk et al. (1999). The reciprocal cross in the small numbers of disomics studied, the unreliable which tetraploid PAX was used as the seed parent produced karyology, and the ambiguous interpretation of the phe- only pentaploid (5x ϭ 40) offspring, indicating that PAX pro- notypes. Moreover, Mogie (1992) argued that all known duced unreduced egg cells that were fertilized (P. J. van Dijk, meiotic mutants in sexual species are recessive rather unpublished data). PAX eggs cells are not parthenogenetic, than dominant (Baker et al. 1976; Kaul and Murthy as PAX requires fertilization for seed set. PAX may have lost this element of apomixis by recombination during pollen mei- 1985). Therefore, Mogie suggested a single locus con- osis in its apomictic grandfather (van Dijk et al. 1999). trol of apomixis, in which diplospory is recessive and The plants of the triploid segregating population were phe- parthenogenesis is a pleiotropic effect of diplospory. notypically classified as diplosporous or nondiplosporous. This Mogie (1992) regarded parthenogenesis as “an innate was done using testcrosses in which each plant was crossed capacity which will find expression if fertilization can with a diploid pollen donor. The ploidy levels of the progeny were then evaluated. From each 3x ϫ 2x testcross a random be avoided through mechanisms such as precocity.” Pre- sample of 25 well-developed, brown seeds was germinated. cocious megagametophyte development could be a di- The nuclear DNA amounts of 10 randomly selected seedlings rect consequence of meiotic diplospory, which, ac- were determined by flow cytometry. If Ͻ10 seedlings were cording to Mogie (1992), is an “abbreviated version available, then all available seedlings were analyzed. The ma- ϫ of normal meiosis.” In Mogie’s model, the dominance ternal parent was classified as diplosporous if a 3x 2x testcross produced only near-tetraploid offspring and as non- relationships between the sexual wild-type allele and the diplosporous if a testcross produced offspring with nuclear mutant apomixis allele are dosage dependent. Triploid DNA amounts mainly in between diploid and triploids. Occa- apomicts contain one dominant wild-type allele A and sionally, incomplete nuclear restitution in a diplosporous trip- excess copies of the apomixis allele a: Aaa, whereas loid mother plant may produce offspring with a few chromo- sexual triploids are AAA or AAa. The triplex genotype somes less than the full tetraploid complement (van Dijk et al. 1999). Conversely, a nondiplosporous triploid mother plant aaa is weak or inviable due to deleterious mitotic effects may occasionally produce offspring with more chromosomes of the apomixis mutation. than the triploid complement owing to the fact that a highly Segregation studies of apomixis in Taraxacum that disturbed triploid meiosis can induce partial nuclear restitu- Diplospory in Dandelion 485

tion (Ramsey and Schemske 1999). Therefore the whole progeny sample was taken into consideration for the diplosporous vs. nondiplosporous classification of the mother plant. DNA flow cytometry and cytogenetics: Ploidy levels were determined with a flow cytometer (Ploidy analyzer, Partec, Mu¨nster, Germany) using 4Ј,6-diamidino-2-phenylindole as fluorescent stain (Tas and van Dijk 1999). A diploid reference plant was used as an internal standard. DNA fluorescence was calibrated with chromosome number counts in an aneuploid series (van Dijk et al. 2003), which allowed us to estimate the chromosome number. Mitotic root tip preparations were made as described in van Baarlen et al. (2000). Molecular markers: The inheritance of three codominant microsatellite markers (Falque et al. 1998) was studied. Figure 1.—The segregation of diplospory in the cross S2- MSTA72 and -78 were selected for their high levels of polymor- 125 ϫ PAX. Seed parents in a cross are to the left, pollen phism in the cross. MSTA78 was a lucky choice, as one of the parents to the right. The sizes of the rectangles are propor- alleles turned out to be associated with diplospory. A third tional to the ploidy level. The numbers above a cross fork microsatellite locus, MSTA53, was also analyzed because this indicate the number of crosses for which the ploidy levels of locus was known to be linked to MSTA78 (Falque et al. 1998). the progeny were analyzed. The numbers along the progeny DNA was extracted from young leaves according to the line indicate the total number of offspring analyzed, as protocol of Rogstad (1992). PCR conditions were identical summed over the crosses. Diplosporous plants are indicated to those described in Falque et al. (1998), with the exception in gray. that one of the primers was fluorescently labeled with Cy5 (Amersham Pharmacia Biotech). The PCR products were analyzed on an ALF express II automatic sequencer (Amersham using the BC1 module of Joinmap 3.0. In tetraploid PAX, Pharmacia Biotech). Microsatellite genotypes were scored in diplospory was mapped assuming a single dominant gene the Curve view mode of the ALFwin sequence analyzer 2.00 (DIP) controlling diplospory. In diploid S2-125, the linkages software. between a segregating rDNA-IGS fragment and the three mi- The association between diplospory and the 18S-25 S rDNA crosatellite loci were also calculated. Significance of linkage locus was also studied. Natural populations had previously was evaluated by the LOD score given by Joinmap and by revealed that rDNA-intergenic spacer (IGS) HinfI digests were Mather’s equation, ␹2[1] ϭ (a Ϫ b Ϫ c ϩ d)2/(a ϩ b ϩ c ϩ hyperpolymorphic in the 1- to 1.6-kb region due to variation d), where a, b, c, and d are the observed number of plants in in the number of 21- and 30-bp repeats (P. J. van Dijk and the four classes in the progeny (Mather 1951). L. M. King, unpublished observations). The whole 3.2-kb IGS was amplified using two conserved primers (SpaF,5Ј-GACGAC TTAAATACGCGACGG, and SpaR,5Ј-GACTACTGGCAGGA RESULTS TCAACC; Polanco and de la Vega 1994). Amplifications were performed on a Hybaid OmniGene thermal cycler, with Њ Segregation of diplospory in the mapping population: a 2-min initial denaturation step at 94 ; followed by 30 cycles The cross S2-125 ϫ PAX generated 86 F offspring of 1-min denaturation at 94Њ, 1-min annealing at 55Њ, and 1 4-min extension at 72Њ; and a final extension of 7 min at 72Њ. plants, all of which were triploids (Figure 1). No diploid ␮ The reaction mixture contained 1.0 mm MgCl2, 200 m of selfed progeny were found, indicating that PAX pollen each dNTP, 0.2 ␮m primers, 2.6 units Expand polymerase did not induce breakdown of the self-incompatibility (Boehringer Mannheim, Mannheim, Germany) and 25 ng system. Sixty-one triploid F plants were crossed with ␮ 1 template DNA. The total volume was 50 l. The amplified diploid pollen donors for testcrosses. Five plants died products were digested with HinfI restriction enzyme. DNA fragments were separated on a 2% agarose gel and stained before testcrossing and the remaining 20 plants could with ethidium bromide. not be tested due to the lack of flowering synchrony Linkage analysis: Linkage between the marker loci and the between the cross partners.

putative diplospory locus was estimated as follows. Each trip- The triploid F1 plants fell in two highly distinct classes loid offspring microsatellite genotype was split into a haploid with respect to the ploidy levels of their progeny in the egg cell genotype originating from the diploid mother S2-125 ϫ and a diploid pollen grain genotype, originating from the 3x 2x testcrosses: 27 plants produced almost exclu- tetraploid father PAX. The recombination frequencies in the sively tetraploid offspring, and 34 plants produced off- egg cells were calculated using the DH module of Joinmap 3.0 spring with aneuploid chromosome numbers in be- (van Ooijen and Voorrips 2001) using the Kosambi mapping tween diploid and triploid. This first group of 27 F1 function. The estimation of linkage between single-dose mark- plants was therefore classified as meiotic restitutional ers in polyploids is essentially equivalent to that of a diploid (diplosporous), and the second group of 34 F plants backcross (Wu et al. 1992). For the diploid pollen grains, the 1 codominant microsatellite genotypes were rewritten as single- was classified as meiotic reductional (nondiplosporous). dose genotypes for each of the four paternal alleles. The four Figure 2 shows the overall distribution of the offspring alleles of a locus were thus treated as four loci. For example, chromosome numbers obtained in 3x ϫ 2x testcrosses. for the “locus” 78a, the genotypes ab, ac, and ae were all coded The penetrance of diplospory was high. A total of 98.5% as 10 heterozygotes and the genotypes bc, be, and ce as 00 of the offspring carried the full tetraploid chromosome homozygotes. Similarly, for the locus 78b, the genotypes ab, Ͻ bc, and be were all coded as 10 heterozygotes and the genotypes set and none of the diplosporous offspring had 29 ac, ae, and ce as 00 homozygotes. The same coding was applied chromosomes. There was little overlap in chromosome to loci 78c and 78e. Linkage between all loci was calculated numbers between the diplosporous and the nondiplo- 486 P. J. van Dijk and J. M. T. Bakx-Schotman

to chromosomal linkage, we also screened the mapping population for MSTA53, which is linked to MSTA78 (at 10 cM distance; Falque et al. 1998). The segregation ratios for MSTA53 did not differ signiﬁcantly from the expected Mendelian ratios (Table 1). The MSTA53b allele showed the same strong positive association with diplospory as the MSTA78a allele, supporting the hypothesis of chromosomal linkage of the MSTA78a allele with a diplospory factor. Linkages between the three microsatellite loci were

estimated using all F1 plants, including the nonflowering plants. Table 2 lists the significant recombination frequencies between pairs of loci and the LOD scores. MSTA78 and -53 were closely linked and both were unlinked to MSTA72, consistent with the findings of Falque et al. (1998). In PAX the allelic phases could be determined for the four homologous chromosomal segments: MSTA78a was linked with MSTA53b,-78b Figure 2.—The overall distribution of offspring chromo- with -53e,-78c with -53null, and -78e with -53c. Recombi- some numbers obtained in testcrosses with sexual diploid pol- nation was detected on three homologous chromosomal len donors for triploid F1 plants that were classified as segments, but was not detected between the diplospory- diplosporous (solid bars; N ϭ 263) or meiotic (hatched bars; ϫ ϭ ϭ associated alleles MSTA78a and -53b (81 4 324 N 172). chromosomes sampled; on average 0.009 crossovers per chromosome). Four crossovers were detected between MSTA53 and -78 during female meiosis in S2-125 (80 ϫ sporous progeny: only 2.1% of the nondiplosporous 2 ϭ 160 chromosomes; on average 0.025 crossovers per progeny had 29 or more chromosomes. Of the non- chromosome). diplosporous progeny plants, 1.45% were full tetra- The associations between MSTA78a, MSTA53b, and ploids, indicating that triploidy per se can also induce diplospory strongly suggest chromosomal linkage; for nuclear division restitution, albeit at a rate much lower the estimation of linkage, however, it was necessary to than that of the diplospory factor. first determine the diplospory genotype of tetraploid In nondiplosporous progeny, high numbers of aneu- PAX and its mode of inheritance. ploid individuals with low viability are expected due to A genetic model for the inheritance of diplospory: an unbalanced triploid meiosis. The germination rates The many different alleles present at the three microsa- in testcrosses with nondiplosporous F1 progeny were tellite loci allow one to deduce the two-allele genotypes significantly lower than those of diplosporous F1 prog- of the diploid PAX pollen grains from the F genotypes. Ϯ Ϯ Ϯ 1 eny (means SE: 0.40 0.03 and 0.75 0.06, respec- Table 3 shows that all of the six possible two-allele pollen Ͻ tively; Kruskal-Wallis, one-way ANOVA, P 0.001). genotypes at the three microsatellite loci were formed, Moreover, many nondiplosporous seedlings died at an implying the absence of strict disomic inheritance. The early stage, before they could be analyzed by flow cytom- observed numbers did not differ significantly from the etry. Therefore the average number of analyzed seed- numbers expected with random tetrasomic chromo- lings for nondiplosporous progeny was only 5.1 (range some segregation. No homozygous diploid pollen geno- 3–10) vs. 9.7 for diplosporous progeny (range 9–10). types were observed, indicating that chromatid segrega- Associations between diplospory and microsatellites: tion or double reduction did not take place. Six and five different alleles, respectively, were present Both Richards’ dominant and Mogie’s recessive mod- in the cross between S2-125 and PAX for MSTA72 and els were tested for their fit to the observed diplospory -78. Table 1 gives the observed segregation ratios of the segregation ratio. Only tetrasomic inheritance with ran- marker loci for the parents. No significant deviations dom chromosome pairing was considered, as the linked from the expected Mendelian segregation ratios were markers MSTA53 and -78 indicated the absence of diso- observed during female meiosis in diploid S2-125 or mic inheritance or chromatid segregation. In the case during male meiosis in tetraploid PAX. of the recessive model, two sexual diploid genotypes are The alleles at MSTA72 were evenly distributed over possible: AA and Aa. Both Mogie’s and Richards’ models the diplosporous and the nondiplosporous F1 progeny, were tested for a single and a double dose of the diplo- whereas at MSTA78, the a-allele was strongly associated spory allele. Table 4 gives the six possible cross combina- with diplospory. The MSTA78a frequency among diplo- tions with their goodness of fit to the observed segrega- sporous plants was 0.46 but only 0.03 among nondiplo- tion ratio. Three models were rejected, but three other sporous plants (chi-square test for heterogeneity, 30.00; genotype combinations fit the observed data with the d.f. ϭ 1; P Ͻ 0.001). Since this association could be due possible genotypes for PAX being Aaaa, AAaa,orHhhh. Diplospory in Dandelion 487

TABLE 1 Segregation ratios at marker loci in the cross S2-125 ϫ PAX

N ␹2 d.f. P S2-125 (2x) MSTA53 198 (a) 212 (d) 0.49 0.51 76 0.05 1 0.82 MSTA72 178 (f) 210 (g) 0.54 0.46 67 0.37 1 0.54 MSTA78 164 (b) 170 (d) 0.50 0.50 78 0.00 1 1.00 rDNA-IGS 1350 n 0.44 0.56 64 1.00 1 0.32

PAX (4x) MSTA53 202 (b) 210 (c) 218 (e) n 0.21 0.27 0.29 0.23 80 1.30 3 0.73 MSTA72 160 (a) 164 (b) 174 (d) 180 (e) 0.29 0.23 0.24 0.24 67 1.03 3 0.79 MSTA78 162 (a) 164 (b) 166 (c) 174 (e) 0.21 0.29 0.22 0.28 73 1.68 3 0.64 The allele sizes are in base pairs; the one-letter code is in parentheses. The observed segregation ratios are compared with the expected 0.5:0.5 and 0.25:0.25:0.25:0.25 Mendelian ratios for S2-125 and PAX, respectively, and tested for signiﬁcant deviations with a chi-square goodness-of-ﬁt test, with one and three degrees of freedom, respectively. The null alleles of MSTA53 and rDNA-IGS are indicated by n.

The three possible genotype combinations lead to No linkages between the diplospory locus and the other different predictions for the frequencies of linked mi- alleles at MSTA53 and -78 were found, supporting the crosatellite markers among diplosporous and non- hypothesis that the dominant diplospory factor is pres- diplosporous progeny (Table 5). As shown in Table 5 ent in a single dose in tetraploid PAX. the dominant Hhhh model (3) has a good ﬁt with the The association between MSTA78 and diplospory in observed allele frequencies for MSTA78, whereas both the pedigree: PAX was produced from a cross between a recessive diplospory models (1 and 2) are rejected. As- triploid diplosporous seed parent (H6-3) with a diploid suming a simplex Hhhh genotype for PAX, the recombi- sexual pollen donor. In turn, H6-3 was produced from a nation fractions between MSTA53,-78, and the diplo- cross between a diploid sexual seed parent with a triploid spory locus can be estimated (Table 2). Four plants were full apomict SE3x-6. This pedigree is shown in Figure recombinant for both microsatellite loci and for the 3. The genotypes of the apomictic founder SE3x-6 for diplospory locus. The recombination frequency be- MSTA53 and -78 are, respectively, b c null and aceand tween MSTA78a and -53b and DIP is 7 cM. Both markers that of H6-3 is bceand abe. Thus, as expected, PAX are located on the same side of the diplospory locus. inherited the diplospory-linked chromosome segment

TABLE 2 Linkage between pairs of loci in diploid S2-125 and tetraploid PAX

Recombination Plant Locus pair ␹2 P frequency (cM) LOD score S2-125 (2x) MSTA53-MSTA78 69.31 0.000 0.05 17.17 MSTA53-rDNA 9.00 0.003 0.31 2.04 MSTA78-rDNA 5.73 0.017 0.36 1.17

PAX (4x) MSTA53b-MSTA78a 79.00 0.000 0.00 24.06 MSTA53e-MSTA78b 71.20 0.000 0.02 19.87 MSTA53n-MSTA78c 71.20 0.000 0.02 20.68 MSTA53c-MSTA78e 71.20 0.000 0.02 20.76 MSTA53b-DIP 46.05 0.000 0.07 11.79 MSTA78a-DIP 46.05 0.000 0.07 11.79 Chi-square test for signiﬁcance of linkage according to Mather (1951); d.f. ϭ 1. The recombination frequencies between pairs of loci and their LOD scores were calculated with Joinmap. 488 P. J. van Dijk and J. M. T. Bakx-Schotman

TABLE 3 Diploid pollen genotypes produced by tetraploid PAX

Locus bc be bn ce cn en N ␹ 2 P MSTA53 Obs. 15 12 7 16 11 19 80 Exp. 13.3 13.3 13.3 13.3 13.3 13.3 5.94 0.31

Locus ab ad ae bd be de N ␹ 2 P MSTA72 Obs. 12 12 14 11 8 9 67 Exp. 11.2 11.2 11.2 11.2 11.2 11.2 2.58 0.76

Locus ab ac ae bc be ce N ␹ 2 P MSTA78 Obs.1071718181080 Exp. 13.3 13.3 13.3 13.3 13.3 13.3 8.97 0.11 Tests for tetrasomic inheritance in diploid pollen from the tetraploid parent PAX. Observed and expected allele combinations in the diploid pollen grains are given. The expected values are based on equal segregation ratios of 0.25 for all four alleles. Chi-square test, d.f. ϭ 5. b-a from its apomictic grandfather SE3x-6. The non- we investigated the association between the diplospory diplospory segment c-e was also inherited from SE3x-6. and the 18S-26S rDNA locus. Digestion of the rDNA- The other two chromosome segments in PAX, e-b and IGS with HinfI produced complex polymorphic banding null-c, originated from diploid sexual parents that were patterns. Only the presence or absence of a 1350-bp S2- no longer alive. 125 fragment could be reliably scored in the F1 mapping The MSTA78 genotypes of other surviving plants of population. The segregation ratio of the presence or the second generation (G2) were also determined (Ta- absence of this fragment did not differ signiﬁcantly from ble 6). The strict association between the MSTA78a al- the expected 1:1 Mendelian ratio (Table 1). According lele and diplospory in the G2 was fully consistent with to Mather’s (1951) chi-square test, MSTA53 and -78 the linkage that was found in the mapping population are both signiﬁcantly linked to the rDNA-IGS locus, at

(G4). All nine diplosporous G2 plants carried the diplo- 0.31 and 0.36 cM, respectively (Table 2). MSTA53 and spory-linked MSTA78a allele whereas none of the three -78 are linked to the diplospory locus in PAX and are nondiplosporous triploid G2 plants carried this allele. also linked to rDNA-IGS in S2-125, providing indirect The MSTA78a allele was also absent in the four diploid, evidence that the diplospory locus is located on the nondiplosporous G2 plants. satellite chromosome. The diplospory locus is located on the satellite chro- Cytological evidence also supports this chromosomal mosome: Sørensen (1958) suggested that a diplospory location of the diplospory locus. One of the triploid G2 factor was located on a satellite chromosome. Therefore plants, H6-4, a full-sib of H6-3, is apomictic as shown by

TABLE 4 Six possible parental genotype combinations for diplospory in the S2-125 ϫ PAX cross

Expected no. of plants Parental Offspring genotype Cross genotypes ratios Diplosporous Nondiplosporous ␹2 P 1 Aa ϫ Aaaa 1AAa:2Aaa:1aaa 40.7 20.3 13.86 0.00 2 AA ϫ Aaaa 1AAa:1Aaa 30.5 30.5 0.80 0.37 3 Aa ϫ AAaa 1AAA:5AAa:5Aaa:1aaa 27.7 33.3 0.03 0.86 4 AA ϫ AAaa 1AAA:4AAa:1Aaa 10.2 50.8 31.10 0.00 5 hh ϫ Hhhh 1hhh:1Hhh 30.5 30.5 0.80 0.37 6 hh ϫ HHhh 1HHh:4Hhh:1hhh 50.8 10.2 66.68 0.00 Crosses 1–4 are according to Mogie’s recessive model; crosses 5 and 6 are according to Richards’ dominant model. The diplosporous genotypes are shown in boldface type; the lethal genotypes are underlined. The expected nos. are tested for their ﬁt to the observed 27 diplosporous and 34 nondiplosporous progeny plants (chi-square test, d.f. ϭ 1). Diplospory in Dandelion 489

TABLE 5 Expected and observed frequencies of MSTA78 alleles among diplosporous and nondiplosporous progeny for three inheritance models assuming complete linkage between the diplospory locus and MSTA78

Diplosporous progeny Nondiplosporous progeny Parental ␹2 Cross genotypes fa fb fc fe Nfa fb fc fe N P Observed 0.46 0.17 0.09 0.28 54 0.03 0.40 0.31 0.26 68 1 AA ϫ Aa/ab/ac/ae 0.00 0.33 0.33 0.33 0.50 0.17 0.17 0.17 78.84 0.000 2 Aa ϫ Aa/Ab/ac/ae 0.20 0.20 0.30 0.30 0.33 0.33 0.17 0.17 22.02 0.001 3 hh ϫ Ha/hb/hc/he 0.50 0.17 0.17 0.17 0.00 0.33 0.33 0.33 7.63 0.178

fa, fb, fc, and fe represent the frequencies of the PAX MSTA78 alleles in the triploid progeny assuming complete linkage with the diplospory alleles at the diplospory locus. N is the no. of alleles in the progeny classes, which is twice the no. of plants because PAX donates two alleles to its triploid offspring. The observed frequencies are tested against the expected frequencies with a chi-square test. The no. of degrees of freedom is six in model 2, but ﬁve in models 1 and 3, because classes had to be combined because of expected values smaller than ﬁve. For further details, see Table 4.

fixed heterozygosity for the four studied microsatellite mic aberrant. Attempts to determine whether H6-4-4 loci (MSTA61,-64,-72, and -78;P.J.van Dijk, unpub- had lost diplospory by testcrossing it with a diploid pol- lished results) in the offspring. However, one of its apo- len donor were unsuccessful. Four flower heads were mictically produced offspring plants (H6-4-4) lacked the crossed, but were completely sterile. maternal MSTA78a allele. Moreover, flow cytometry indicated that H6-4-4 is a hypotriploid (3x Ϫ 1). Chromo- some spread preparations showed that H6-4-4 contains DISCUSSION only two instead of the normal three satellite chromo- A model for the inheritance of diplospory in Taraxa- somes (Figure 4). The simultaneous loss of one of the cum: The inheritance of diplospory and the association three MSTA78 alleles and one of the three satellite chro- with two linked microsatellite loci strongly suggests that mosomes provides cytological evidence that MSTA78 diplospory in Taraxacum is controlled by a dominant and, therefore, also the linked diplospory locus are lo- genetic factor. This diplospory gene, DIP, is present in cated on a satellite chromosome. a single dose in the triploid apomictic founder of the H6-4-4 was a slender plant that produced only a few pedigrees. The diplospory locus may be heterozygous inflorescences. H6-4-4 flower heads were smaller than or hemizygous. Hemizygosity was demonstrated for the those of H6-4 and had an irregular appearance because apospory locus in the grass species Pennisetum squamula- of the curled floret ligules, as was also described by Sørensen and Gudjońsson (1946) for the tenuis diso- TABLE 6

G2 offspring from the apomictic pedigree founder SE3x-6

Ploidy MSTA78 Plant level Breeding system genotype H2-1 2x Sexual be H3-3 2x Sexual bc H3-8 2x Sexual bc H4-1 2x Sexual bc H2-3 3x Type A, nondiplosporous bce H2-4 3x Type A, nondiplosporous bce H3-5 3x Type A, nondiplosporous bce H3-1 3x Type B, diplosporous ace H3-6 3x Type B, diplosporous abc H3-7 3x Type B, diplosporous abe H6-4 3x Type B, diplosporous abe H2-5 3x Apomict abc H3-2 3x Apomict abc H6-1 3x Apomict abc H6-2 3x Apomict abc Figure 3.—The full pedigree of the apomictic founder SE3x-6 with the genotypes for MSTA78. The siblings of H6-3 Type A are sexual, nondiplosporous plants; type B are are listed in Table 6. Diplosporous plants are indicated in diplosporous parthenogenetic plants, which lack autonomous gray. Dots represent unknown alleles. G, generation. endosperm development (van Dijk et al. 1999, 2003). 490 P. J. van Dijk and J. M. T. Bakx-Schotman

Figure 4.—Root tip metaphase in the disomic plant H6-4-4 (3x Ϫ 1 ϭ 23). This is an offspring plant from the apomictic mother H6-4. H6-4-4 is missing one of the three maternal MSTA78 alleles (see Curve view print; small arrows) and also one of the three satellite chromosomes (large arrows).

tum (Ozias-Akins et al. 1998) and Paspalum simplex ments of apomixis: diplospory, parthenogenesis, and

(Pupilli et al. 2001; Labombarda et al. 2002). autonomous endosperm development. The G1 plant Diplospory affects only female meiosis, resulting in H6-3 was diplosporous and exhibited autonomous en- unreduced megaspores. Male meiosis in diplosporous dosperm development, but lacked parthenogenesis plants is reductional and allows for genetic mapping of (van Dijk et al. 1999). Parthenogenesis did not reappear the trait. The weak, but significant, linkage between the in later generations (PAX and its F1 progeny) and was rDNA locus and the diplospory-linked markers MSTA53 thus permanently lost. Some siblings of H6-3 were and -78 in the sexual plant S2-125 is consistent with a diplosporous and parthenogenetic, but lacked autono- satellite chromosome location of the diplospory gene. mous endosperm development (van Dijk et al. 2003). The physical location of the DIP locus was cytogeneti- This suggests that the three elements of apomixis in cally confirmed in a disomic plant (3x Ϫ 1 ϭ 23) H6- Taraxacum can be inherited independently and can be 4-4 that lacked both the diplospory-linked MSTA78a separated by recombination during male meiosis. Again allele and a satellite chromosome. Unfortunately, this this is inconsistent with a single-locus pleiotropic model plant was sterile and it could not be directly determined for apomixis in Taraxacum. whether this plant also lacked diplospory. Nevertheless Comparisons with other apomictic species: Dominant both the linkage and the cytogenetic evidence presented monogenic control of apomeiosis has been reported in in this article support the suggestion by Sørensen (1958) all apomictic species studied so far (see Grimanelli et that the DIP gene is located on the satellite chromo- al. 2001; Grossniklaus et al. 2001, for reviews). Most some. Given the weak linkage between MSTA53 and -78 of these species are aposporous. T. officinale, after E. and the rDNA locus, it is likely that the diplospory locus annuus (Noyes and Rieseberg 2000), is now the second is located on the short arm, opposite the nucleolar orga- species in which meiotic diplospory is shown to be inher- nizer region. ited as a monogenic dominant trait. Our results corroborate Richards’ two dominant loci Taraxacum and Erigeron are both members of the model for apomixis in Taraxacum. Mogie’s alternative Asteraceae (Compositae) family, but belong to different model of a single recessive gene for diplospory with a subfamilies (Bremer 1994). Diplospory in the two spe- pleiotropic effect on parthenogenesis is incompatible with cies is cytogenetically very similar (Bergman 1950). our results and therefore has to be rejected. Diplospory in triploid Erigeron is inherited disomically In Richards’ model the putative diplospory gene is (Noyes and Rieseberg 2000), suggesting that the diplo- named H and the putative parthenogenesis gene is spory chromosome does not pair with the two other named D, referring to Sørensen and Gudjońsson’s homologous chromosomes. In contrast, diplospory in (1946) chromosome terminology. However, this let- Taraxacum is inherited tetrasomically, without obvious tering is counterintuitive and therefore confusing. Now preferential chromosome pairing. that the diplospory locus has been mapped, we suggest In most species apomixis is inherited as a single domi- the name DIPLOSPOROUS, abbreviated to DIP, in which nant trait (Grossniklaus et al. 2001), whereas in both the dominant D allele controls diplospory and the reces- Erigeron and Taraxacum diplospory and parthenogene- sive d allele controls normal meiotic reduction. sis are inherited independently. Likewise, apospory and The apomictic founder SE3x-6 expressed three ele- parthenogenesis are regulated independently in Poa pra- Diplospory in Dandelion 491 tensis (Albertini et al. 2001). It is therefore conceivable, in Poa pratensis: a cytological investigation. Sex. Plant Reprod. 14: 213–217. in species where apomixis is inherited as a monogenic Asker, S. E., and L. Jerling, 1992 Apomixis in Plants. CRC Press, trait, that apomixis is in fact controlled by a complex Boca Raton, FL. of closely linked genes with different functions. Baker, B. S., T. C. Carpenter, M. S. Esposito, R. E. Esposito and L. Sandler, 1976 The genetic control of meiosis. Annu. Rev. What is the function of the DIP gene? Many meiotic Genet. 10: 53–134. mutants have been described in sexual species like Barton, N. H., and B. Charlesworth, 1998 Why sex and recombi- maize and Arabidopsis. These mutations affect meiotic nation? Science 281: 1986–1990. Bergman, B., 1950 Meiosis in two different clones of the apomictic processes such as chromosome pairing, recombination, Chondrilla juncea. Hereditas 36: 297–320. and chromatid cohesion (Kaul and Murthy 1985; Bhatt, A. M., C. Canales and H. G. Dickinson, 2001 Plant meiosis: Bhatt et al. 2001; Mercier et al. 2001). the means to 1N. Trends Plant Sci. 6: 114–121. Bremer, K., 1994 Asteraceae. Cladistics and Classification. Timber Press, Two cytological mechanisms of diplospory have been Portland, OR. described in Taraxacum: (i) first division restitution Fagerlind, F., 1947 Makrosporogenese und embryosacbildung bei (FDR) and (ii) pseudohomeotypic division. In both agamospermischen Taraxacum-biotypen. Svensk Botanisk Tidsk- rift 41: 365–390. cases the univalents move to the equatorial plane to Falque, M., J. Keurentjes, J. T. M. Bakx-Schotman and P. J. van form a metaphase plate. In the case of FDR, the uni- Dijk, 1998 Development and characterization of microsatellite valent group undergoes strong contraction to form a markers in the sexual-apomictic complex Taraxacum officinale (dandelion). Theor. Appl. Genet. 97: 283–292. so-called contraction nucleus (Gustafsson 1934; Fager- Grimanelli, D., O. Leblanc, E. Espinosa, E. Perotti, D. G. De lind 1947; Bergman 1950), which is followed by a nor- Leon et al., 1998 Mapping diplosporous apomixis in tetraploid mal meiosis II. In contrast, in the pseudohomeotypic Tripsacum: One gene or several genes? Heredity 80: 33–39. Grimanelli, D., O. Leblanc, E. Perotti and U. Grossniklaus, division, the univalents at the metaphase plate of meiosis 2001 Developmental genetics of gametophytic apomixis. Trends I divide, and the chromatids move to opposite poles Genet. 17: 597–604. and form two nuclei (Gustafsson 1934). Meiosis II is Grossniklaus, U., G. A. Nogler and P. J. van Dijk, 2001 How to avoid sex: the genetic control of gametophytic apomixis. Plant omitted in this type of diplospory. Cell 13: 15–21. Diplospory in Taraxacum is characterized by the Gustafsson, A˚ ., 1934 Primary and secondary association in Taraxa- (near) absence of bivalents. Fagerlind (1947) did not cum. Hereditas 20: 1–31. Kaul, M. L. H., and T. G. K. Murthy, 1985 Mutant genes affecting observe chromosome pairing during early prophase, higher plant meiosis. Theor. Appl. Genet. 70: 449–466. which suggests asynapsis rather than desynapsis. Both Koltunow, A. M., 1993 Apomixis: embryo sacs and embryos formed FDR and pseudohomeotypic division produce a dyad without meiosis or fertilization in ovules. Plant Cell 5: 1425–1437. Labombarda, P., A. Bust,M.E.Caceres,F.Pupilli and S. Arcioni, of two megaspores. A number of mutants have been 2002 An AFLP marker tightly linked to apomixis reveals hemizy- described in which dyads rather than tetrads are formed gosity in a portion of the apomixis-controlling locus in Paspalum during meiosis. The dyad mutation in Arabidopsis affects simplex. Genome 45: 513–519. Mather, K., 1951 The Measurement of Linkage in Heredity. Methuen, only female meiosis, as with diplospory in Taraxacum; London. however, the dyads result from the omission of meiosis Mercier,R.,M.Grelon,D.Vezon,C.Horlow and G. Pelletier, II (Siddiqi et al. 2000). The Spo13 mutation in Saccharo- 2001 How to characterize meiotic functions in plants? Biochi- mie 83: 1023–1028. myces cerevisiae also produces dyads, affecting meiosis I, Mogie, M., 1988 A model for the evolution and control of generative but no Arabidopsis Spo13 ortholog has been found so apomixis. Biol. J. Linn. Soc. 35: 127–153. far (Mercier et al. 2001). At present we are not aware Mogie, M., 1992 The Evolution of Asexual Reproduction in Plants. Chap- man & Hall, London. of any meiotic mutant in sexual plants with a phenotype Morita, T., S. B. J. Menken and A. A. Sterk, 1990 Hybridization similar to that of DIP. Moreover, all known meiotic mu- between European and Asian dandelions (Taraxacum section Ru- tants are recessive and generally highly sterile whereas deralia and section Mongolica). 1. Crossability and breakdown of self-incompatibility. New Phytol. 114: 519–529. DIP in Taraxacum is dominant and diplosporous plants Nogler, G. A., 1984 Gametophytic apomixes, pp. 475–518 in Embry- are highly fertile. This makes the DIP gene especially ology of Angiosperms, edited by B. M. Johri, Springer Verlag, Berlin. intriguing and interesting for further characterization. Noyes, R. D., and L. H. Rieseberg, 2000 Two independent loci control agamospermy (apomixis) in the triploid flowering plant We acknowledge Hans de Jong and Henny Verhaar for the cytoge- Erigeron annuus. Genetics 155: 379–390. netic analyses performed in this study and for providing Figure 4. We Ozias-Akins, P., D. Roche and W. W. Hanna, 1998 Tight clustering thank Kim Boutilier, Hans de Jong, Peter van Baarlen, Kitty Vijverberg, and hemizygosity of apomixis-linked molecular markers in Penni- and the reviewers of Genetics for their comments, which helped to setum squamulatum implies genetic control of apospory by a diver- improve the manuscript considerably. Part of this work was funded gent locus that may have no allelic form in sexual genotypes. by the European Union as part of the project “Natural apomixis as Proc. Natl. Acad. Sci. USA 95: 5127–5132. Polanco, C., and M. P. de la Vega, 1994 The structure of the rDNA a novel tool in plant breeding (ApoTool),” contract number QLG2- intergenic spacer of Avena sativa L.—a comparative study. Plant 2000-00603 of the Quality of Life and Management of Living Resources Mol. Biol. 25: 751–756. section. This is publication 3226 of the Netherlands Institute of Ecol- Pupilli, F., P. Labombarda, M. E. Caceres, Q. L. Quarin and S. ogy (NIOO-KNAW). Arcioni, 2001 The chromosome segment related to apomixis in Paspalum simplex is homoeologous to the telomeric region of the long arm of rice chromosome 12. Mol. Breed. 8: 53–61. Ramsey, J., and D. W. Schemske, 1999 Pathways, mechanisms and LITERATURE CITED rates of polyploid formation in flowering plants. Annu. Rev. Ecol. Syst. 29: 467–501. Albertini, E., A. Porceddu, F. Ferranti, L. Reale, G. Barcaccia Richards, A. J., 1970 Eutriploid facultative agamospermy in Taraxa- et al., 2001 Apospory and parthenogenesis may be uncoupled cum. New Phytol. 69: 761–774. 492 P. J. van Dijk and J. M. T. Bakx-Schotman

Richards, A. J., 1973 The origin of Taraxacum agamospecies. Bot. van Dijk, P. J., and J. M. M. van Damme, 2000 Apomixis-technology J. Linn. Soc. 66: 189–211. and the paradox of sex. Trends Plant Sci. 5: 81–84. Rogstad, S. H., 1992 Saturated NaCl-CTAB solution as a means of van Dijk, P. J., I. C. Q. Tas, M. Falque and J. M. T. Bakx-Schotman, field preservation of leaves for DNA analysis. Taxon 41: 701–708. 1999 Crosses between sexual and apomictic dandelions (Taraxa- Siddiqi, I., G. Ganesh, U. Grossniklaus and V. Subbiah, 2000 The cum). II. The breakdown of apomixis. Heredity 83: 715–721. dyad gene is required for progression through female meiosis in van Dijk, P. J., P. van Baarlen and J. H. De Jong, 2003 The occur- Arabidopsis. Development 127: 197–207. rence of phenotypically complementary apomixis-recombinants Sørensen, T., 1958 Sexual chromosome-aberrants in triploid apo- in crosses between sexual and apomictic dandelions (Taraxacum mictic Taraxaca. Bot. Tidsskr. 54: 1–22. officinale). Sex. Plant Reprod. 16: 71–76. Sørensen, T., and G. Gudjońssen, 1946 Spontaneous chromosome- van Ooijen, J. W., and R. E. Voorrips, 2001 Joinmap 3.0, Software aberrants in apomictic Taraxaca. Morphological and cyto-geneti- for the Calculation of Genetic Maps. Plant Research International, cal investigations. K. Dan. Vidensk. Selsk. Biol. Skr. 4 (2): 3–48. Wageningen, The Netherlands. Spillane, C., A. Steimer and U. Grossniklaus, 2001 Apomixis in Vielle-Calzada, J.-P., C. F. Crane and D. M. Stelly, 1996 Apo- agriculture: the quest for clonal seeds. Sex. Plant Reprod. 14: mixis. The asexual revolution. Science 274: 1322–1323. 179–187. West, S. A., C. M. Lively and A. F. Read, 1999 A pluralist approach Tas, I. C. Q., and P. J. van Dijk, 1999 Crosses between sexual and to sex and recombination. J. Evol. Biol. 12: 1003–1012. apomictic dandelions (Taraxacum). I. The inheritance of apo- Wu, K. K., B. Burnquist, M. E. Sorrells, T. L. Tew, P. H. Moore mixis. Heredity 83: 707–714. et al., 1992 The detection and estimation of linkage in polyploids van Baarlen, P., P. J. van Dijk, R. F. Hoekstra and J. H. De Jong, using single dose restriction fragments. Theor. Appl. Genet. 83: 2000 Meiotic recombination in sexual diploid and apomictic 294–300. triploid dandelions (Taraxacum officinale L.). Genome 43: 827– 835. Communicating editor: J. Birchler