Identification and genetic analysis of the APOSPORY locus in Hypericum perforatum L.

Summary

The introduction of apomixis – seed formation without fertilization – into crop plants is a long-held goal of breeding research, since it would allow for the ready fixation of heterozygosity. The genetic basis of apomixis, whether of the aposporous or the diplosporous type, is still only poorly understood. Hypericum perforatum (St John’s wort), a plant with a small genome and a short generation time, can be aposporous and/or parthenogenetic, and so represents an interesting model dicot for apomixis research. Here we describe a genetic analysis which first defined and then isolated a locus (designated HAPPY for Hypericum APOSPORY) associated with apospory. Amplified fragment length polymorphism (AFLP) profiling was used to generate a cleaved amplified polymorphic sequence (CAPS) marker for HAPPY which co-segregated with apospory but not with parthenogenesis, showing that these two components of apomixis are independently controlled. Apospory was inherited as a dominant simplex gene at the tetraploid level. Part of the HAPPY sequence is homologous to the Arabidopsis thaliana gene ARI7 encoding the ring finger protein ARIADNE7. This protein is predicted to be involved in various regulatory processes, including ubiquitin-mediated protein degradation. While the aposporous and sexual alleles of the HAPPY component HpARI were co-expressed in many parts of the plant, the gene product of the apomict’s allele is truncated. Cloning HpARI represents the first step towards the full characterization of HAPPY and the elucidation of the molecular mechanisms underlying apomixis in H. perforatum.

Introduction

Apomixis, a form of asexual reproduction in which a seed is still developed, is found in more than 400 angiosperm species belonging to about 40 families (Carman, 1997; van Dijk and Vijverberg, 2005). The engineering of apomixis into sexual crop plants has long been considered highly desirable since it would allow for the fixation of favourable heterozygosity, and, in particular, of heterosis. The economic and social benefits of such a technology are likely to exceed those of the Green Revolution (Spillane et al., 2004). Achieving this goal requires a full understanding of the genetic and molecular mechanisms underlying apomixis. Apomixis occurs in two distinct forms – sporophytic (or adventitious embryony with autonomous embryo development in sporophytic tissues) and gametophytic (where a meiotically unreduced, non-fertilized egg develops into an embryo). The gametophytic type is recognized to include two subtypes, diplospory and apospory. In the former, the progenitor is the megaspore mother cell (MMC), which undergoes a shortened meiosis with anaphase I or anaphase II skipped; while in the latter embryo sacs develop from cell(s) adjacent to the MMC. In both cases, the development of a parthenogenetic embryo proceeds from an unreduced embryo sac, but endosperm development can be either autonomous or, if fertilization of the polar nucleus is required, pseudogamous (for reviews see Nogler, 1984; Koltunow, 1993; Savidan et al., 2001; Koltunow and Grossniklaus, 2003; Ozias-Akins, 2006; Hörandl et al., 2007). Apomictic species which maintain a balance between the genetic stability ensured by apomixis and the flexibility offered by segregation and/or recombination are referred to ‘facultative apomicts’.

Despite many years of apomixis research, the genetic control of apomixis remains in its infancy. Examples are known where apomixis is determined by the action of a single dominant gene (Savidan, 1980; Leblanc et al., 1995; Bicknell et al., 2000), but in other cases the pattern of inheritance is more complex. Apospory in Pennisetum and Paspalum spp. is associated with the presence of a non-recombining region of supernumerary chromatin (Ozias-Akins et al., 1998; Roche et al., 2001; Labombarda et al., 2002; Conner et al., 2008). Diplospory in the Boechera holboelli complex seems associated with the presence of a heterochromatic chromosome (Kantama et al., 2007). A growing body of evidence supports the notion that some of the components of apomixis (in particular, apospory and parthenogenesis) are independently, but rather simply, inherited (Noyes and Rieseberg, 2000; Barcaccia et al., 2000; Albertini et al., 2001; Matzk et al., 2001; van Baarlen et al., 2002; van Dijk and Bakx-Schotman, 2004; Catanach et al., 2006; Noyes et al., 2006). However, some of these components can also be under polygenic control (Matzk et al., 2005). Apomixis has evolved independently a number of times (Holsinger, 2000). It represents a short-circuiting of the sexual reproduction pathway, presumably via omission or deregulation, as a result of either mutation or de novo gene evolution (Tucker et al., 2003; Albertini et al., 2005; Ozias-Akins and van Dijk, 2007; Sharbel et al., 2009). As proposed by Nogler (1984), apomixis may be triggered by reproduction-specific gene expression activated at the wrong time and/or place (reviewed in Koltunow and Grossniklaus, 2003). Several mutants of sexual species display developmental heterochrony and apomixis-like meiotic non-reduction, parthenogenesis and autonomous endosperm formation (Huang and Sheridan, 1996; Ohad et al., 1996; Chaudhury et al., 1997; Guitton et al., 2004; Barrell and Grossniklaus, 2005, Ravi et al., 2008; d’Erfurth et al., 2009). Thus the expectation is that the expression and ultimate function of the genes critical for sexual development are perturbed in apomicts, although whether such changes are primary or causal effects has yet to be established.

Apomixis is often associated with extensive heterozygosity and polyploidy, although some diploid apomicts have been described (Nogler, 1982; Roy, 1995; Bicknell, 1997; Kojima and Nagato, 1997). Whether the characteristically high ploidy levels are the cause or the consequence of apomixis remains unclear (Koltunow and Grossniklaus, 2003). Chromosomal regions associated with some apomixis factors have been identified in several species, and molecular markers for diplospory and parthenogenesis have been identified (Barcaccia et al., 1998; Noyes and Rieseberg, 2000). Recombination in the region of the loci linked to apomeiosis tends to be strongly suppressed, although this was not the case for either Taraxacum (van Dijk and Bakx-Schotman, 2004) or Poa pratensis (Barcaccia et al., 1998). More recently, the possibility has been raised that epigenetic mechanisms could also be involved in the determination of apomixis (Lohe and Chaudhury, 2002; Koltunow and Grossniklaus, 2003). The presence of transposons and repetitive sequences in apomixis loci represents indirect evidence for an elevated extent of localized DNA methylation (Lohe and Chaudhury, 2002; Koltunow and Grossniklaus, 2003). A second line of evidence flows from the recognition that mutations in the direction of parthenogenesis and autonomous endosperm development involve epigenetic regulators of DNA and/or histone methylation (reviewed in Köhler and Makarevich, 2006). Some examples of the dominance of epialleles have been presented (Lohe and Chaudhury, 2002). The argument is therefore that a master gene(s) controlling apomixis can be epigenetically modified, or that regulatory factors reciprocally control epigenetic marks. The epigenetic model combines prior hypotheses surrounding mutant alleles, dominant genes, hybridization and polyploidy (Lohe and Chaudhury, 2002; Koltunow and Grossniklaus, 2003).

Hypericum perforatum L. (St John’s wort) produces pharmaceutically important metabolites with possible antidepressant, anticancer and antiviral/fungal/microbial activities (for a review see Barnes et al., 2001). Moreover, the occurrence of variable ploidy levels, facultative apospory and pseudogamy in H. perforatum L. has attracted the attention of apomixis researchers (Matzk et al., 2001, 2003). Wild populations are predominantly tetraploid (2n = 32), although both diploid (2n = 16) and hexaploid (2n = 48) forms are also known (Matzk et al., 2001; Robson, 2002; Barcaccia et al., 2006). The variation in ploidy level is thought to reflect a dynamic reproductive system. Haploidization and polyploidization are the consequences of parthenogenesis of egg cells and fertilization of non-reduced aposporous egg cells, respectively (Barcaccia et al., 2007). The result is that the species has a versatile mode of reproduction, ranging from completely sexual to nearly obligate apomictic. Along with a relatively small genome size and a short generation time, this feature has made H. perforatum a leading model for apomixis research (Matzk et al., 2001; Barcaccia et al., 2007). Here, we report the mapping and cloning of at least a part of a gene responsible for apospory in H. perforatum. We suggest that apospory in H. perforatum is most probably controlled by dominant factors contained within the cloned sequence. This advance represents the initial step necessary for the isolation of the Hypericum APOSPORY locus (HAPPY), which should provide a major insight into the molecular control of the aposporous mode of reproduction.

Results

Apospory and parthenogenesis are under independent control

The four apomictic accessions (aAn, aNo, aSi, aTo) used as pollinators exhibited between 83 and 100% apospory and 80–100% parthenogenesis (Table S1 in Supporting Information). Both the diploid (sR1, sP1, sP2, sV1, sV2 and sV3) and the tetraploid (sF₁1, sF₁2, sR1C) sexual parents were confirmed as obligate sexuals, because flow cytometry seed screen (FCSS) analysis failed to detect either apospory or parthenogenesis among their selfed progeny. Hybrids between the obligate sexuals produced uniformly obligate sexual progeny (Table 1), suggesting that the parents were all homozygous for the genes responsible for sexual reproduction. Among the triploid F₁ (4x apomictic × 2x sexual) individuals, the ratio of aposporous to sexual plants was about 3:1 (44:15), inferring the dominance of apospory at the triploid level (Table 1). This ratio is intermediate between the expected simplex 1:1 and duplex 5:1 ratios. In the F₁ populations 4x sexual (F₁1, F₁2, R1C) × 4x apomictic (aAn, aNo, aSi, aTo), 34 progeny were aposporous, at least to some degree (Table S2), and 38 were sexual, consistent with a 1:1 segregation (χ² = 0.222, P = 0.63). This 1:1 ratio was taken to indicate that apospory in the tetraploid male parent was a consequence of a simplex allelic constitution (for statistical details see Table S5).

Table 1.   Crossing scheme (upper panel) and progeny analysis (lower panel) undertaken to determine the mode of reproduction of apomixis in Hypericum perforatum. Phenotype with respect to apospory scored by flow cytometry, supported by genotyping

In contrast to the apospory, the level of parthenogenesis was low (0–53%). None of the triploid progeny expressed an intermediate or high level of parthenogenesis, but a small number of the tetraploid progeny did so. The ratio of plants with and without parthenogenesis was (28 + 1):(16 + 14) among the triploids and (27 + 1):(7 + 38) among the tetraploids (Table 1), which is close to a 1:1 ratio, suggesting a simplex genetic constitution at the critical gene in the apomictic parent. Plants showing high levels of both apospory and parthenogenesis were rare (3/131) among the progeny of sexual × apomictic crosses. The occurrence of 16 triploid and seven tetraploid progeny which expressed apospory but not parthenogenesis, along with one triploid and one tetraploid which was parthenogenetic but not aposporous, shows that these two apomixis components must be independently controlled in H. perforatum. To test whether recombination between apospory and parthenogenesis had occurred, a joint segregation analysis was performed. This showed a significant deviation from the expected 1:1:1:1 ratio at both the triploid and tetraploid levels (Table 1), demonstrating the presence of linkage, estimated to be 20.1 cM (Table S5). We concluded that the two dominant alleles responsible for apospory and parthenogenesis were cis-linked.

An amplified fragment length polymorphism (AFLP) marker discriminates between accessions with contrasting modes of reproduction

An AFLP-based analysis was applied to a sample of ten apomictic and six sexual individuals (Figure 1, Table S1), employing 36 PstI/MseI primer combinations. Although there were many polymorphisms among the 16 DNAs, only one polymorphism was present in all of the apomictic but in none of the sexual templates. The same fragment also co-segregated with apospory among a sample of 40 F₁ progeny. The AFLP marker was convertible into a cleaved amplified polymorphic sequence (CAPS) marker, since the sequences of the amplicon derived from the template of aposporous and sexual accessions differed from one another by the presence/absence of an EcoRI site, which splits the 223-bp sequence into two similar sized fragments (105 and 118 bp) (Table 2). The DNA extracted from aposporous plants did appear to contain a small amount of the sexual allele, as shown by the appearance of a faint 105/118-bp product along with the well amplified 223-bp one (Table 2). However, analysis of the extended sequence of the apospory-specific allele showed that primer mismatching was responsible for the preferential amplification of the aposporous over the sexual allele (Figure S3). The association between the CAPS marker and the mode of reproduction was further tested by an analysis of all the segregating material available (157 tetraploid, triploid and diploid plants). The undigested 223-bp fragment co-segregated perfectly with the aposporous mode of reproduction, allowing the CAPS marker to be used as a replacement for FCSS analysis (Tables 2 and S2). As expected from the joint segregation between apospory and parthenogenesis, the CAPS marker was not predictive of parthenogenesis.

Figure 1.  Amplified fragment length polymorphism (AFLP) profiling of 10 apomictic (A) and six sexual accessions (S) of Hypericum perforatum. The vertical line indicates the fragment present exclusively in the profile of apomictic individuals.

Table 2.   Cleaved amplified polymorphic sequence (CAPS) analysis of 28 sexual (S/SEX) and aposporous (A/APO) Hypericum perforatum individuals. The presence of the 223-bp amplicon lacking an internal EcoRI site correlates perfectly with apospory, whereas the amplificons derived from templates of plants having the sexual mode of reproduction are digested by EcoRI into two closely migrating fragments of size ∼110 bp. The level of apospory of each plant is shown in the upper panel. Progeny analysis from various crosses with respect to apospory and CAPS marker state is shown in the lower panel

Cloning of the apospory-linked region

A BlastX analysis (Altschul et al., 1990) of the CAPS marker sequence revealed significant hits with genes encoding the ARIADNE-subclass of RING-finger proteins (Jackson et al., 2000; Mladek et al., 2003). The screen of the bacterial artificial chromosome (BAC) library with the CAPS marker produced eight positive clones, which, upon fingerprinting, all belonged to a single contig (Figure S1). The longest BAC clone (25H09) containing the marker in its central region was sequenced (GenBank accession number HM061166). The GC content of the BAC sequence was 38.8%, rising to 41.2% in the coding regions (Figure 2). The 24 predicted genes in 25H09 are depicted in Figure 2, along with their closest BlastX hits (Table S4). These include, inter alia, a variety of transcription factors and four retrotransposons.

Figure 2.  Gene annotation of the 142-kb bacterial artificial chromosome (BAC) clone 25H09 containing HpARI. HK1,2, histidine kinase; GH1,2, glycosyl hydrolase; RT1,2,3,4, retrotransposon; ExP1, expressed protein; PAT1, a member of the GRAS family; WRKY, WRKY transcription factor; DGCR, hypothetical protein; ARI, ubiquitin-mediated E3-ligase; ARI-T, a truncated version of ARI; NPH3, plastidal phototropic-responsive protein; TK, thymidine kinase; RNP, heterogeneous nuclear ribonucleoprotein; AUXR, auxin-response protein; RimL, GCN5-related N-acetyltransferase domain-containing protein; Mo25, membrane protein of the Mo-25 family; TtRNAS, threonine-tRNA ligase; RINGH2, polygalcturonase and a RING finger protein of the H2 class. Other details of the annotation are given in Fig. S4.

The ARIADNE-like gene containing the CAPS marker was a prime candidate for the HAPPY locus. In 25H09 (which contains the sexual allele), the coding region has a length of 1556 bp with no introns. Its predicted translation product shares strong homology with the 562-residue product of the A. thaliana ARIADNE 7 (ARI7) gene. The relevant genomic region was amplified from four obligate aposporous and four sexual plants to investigate sequence variation in the candidate gene. To ensure capture of a simplex locus, at least 16 independent clones of these amplicon inserts were sequenced to have a 96% chance to clone any allele (Simko, 2004). Eleven haplotypes were identified, falling into two distinct groups (Figure 3). One group contained haplotypes 300 and 311, with only the latter being specific to all aposporous plants. Among the 10 haplotypes distributed among the sexual plants, 73 single nucleotide polymorphisms (SNPs) relative to the 25H09 sequence were identified, resulting in 40 amino acid changes. The apospory-specific haplotype 311 contained 39 SNPs, some of which were responsible for a change in the translation sequence. A 6-bp deletion in haplotype 200 removes two amino acid codons but leaves the reading frame intact. Three other indels – two 2-bp deletions and a 1-bp insertion – were detected. The 2-bp deletion in haplotype 311 caused a frame-shift which resulted in a truncated gene product sharing 48 residues at the N-terminus with the sexual alleles, followed by 19 unique residues up to the stop codon. The two other indels are also apospory-specific, but occur behind the stop codon (Figure S3).

Figure 3.  HpARI haplotypes of a panel of aposporous and sexual accessions of Hypericum perforatum (upper panel) and an analysis of haplotype frequency (lower panel).

Allele-specific RT-PCR was performed to determine whether both the sexual and the aposporous alleles are expressed. As the target gene lacks introns, we first needed to successfully demonstrate the absence of contaminating genomic DNA in the cDNA template by a parallel assay of two adjacent genes (HISTIDINE KINASE 2 and Mo25) using primers which amplify across an intron. The expression of the sexual and apomictic HpARI alleles in the floral buds at megasporogenesis was indistinguishable from that at megagametogenesis (Figure 4).

Figure 4.  Expression of contrasting HpARI alleles, as estimated by RT-PCR. Both alleles are expressed in in pistils and leaves of aposporous plants.

Using the HpARI locus as a landmark for the HAPPY locus, we also examined the quantitative nature of the HpARI alleles by comparing hybridization intensities displayed by Southern blotting (Figure 5). The aposporous allele is marked by a 978-bp EcoRI fragment lacking the internal EcoRI restriction site, while the sexual allele generated a pair of fragments (583 and 395 bp) because of the presence of the internal EcoRI site. The ratio of hybridization intensity in three aposporous accessions between the 978-bp fragment and the 583-bp/395-bp pair was about 1:3, suggesting the presence of one apospory allele and three sexual alleles in the apomictic types, and four sexual alleles in the 4x sexual types. A similar Southern analysis based on either HindIII or BamHI digestion produced only a single hybridizing fragment, indicating that the HpARI locus is single copy (data not shown). As a further test, the HpARI sequence was cloned from a set of aposporous plants, and the clones digested with EcoRI. Of these, 15 lacked the restriction site (aposporous allele) and 49 possessed it (sexual allele), a ratio which conforms closely to 1:3. Thus we were able to conclude that the allelic constitution of the apospory locus in tetraploid plants is simplex. Pyrosequencing within HpARI was then used to determine the allelic constitution in the full collection of accessions. This revealed that the aposporous plants contained 20–25% of the apospory-specific HpARI allele and 70–75% of the sexual ones, while the sexual plants contain from 96–100% of the sexual allele (Figure 5). The very small proportion of apospory-specific alleles detected in the sexual plants lies within the range of error of the pyrosequencing method.

Figure 5.  Determination of allele frequencies via pyrosequencing (upper panel). Southern hybridization to determine the ratio of sexual and aposporous alleles in tetraploid Hypericum perforatum (lower panel).

Genomic walking extends the apospory-specific genomic locus

When the alleles at 13 neighbouring genes along 25H09 in aposporous and sexual accessions were compared at the sequence level, no apospory-specific alleles were detected (Figure S2). Genome walking was then employed to investigate the immediate upstream and downstream regions of the aposporous HpARI allele. The former extended the sequence as far as the DGCR14 orthologue, while the latter reached beyond a truncated second copy of ARI (ARI-T). From this extended sequence, it was possible to conclude that homology between the sexual and aposporous sequences is restricted to a region closely surrounding HpARI, and disappears abruptly in the C-terminal region of DGCR14, 72 bp upstream of its predicted stop codon (Figure S4). At the downstream end, homology is lost within ARI-T, 102 bp upstream of its predicted stop codon (Figure S5). Therefore, it seems likely that HpARI is a critical component of the HAPPY locus.

Discussion

Apospory and parthenogenesis can be recombined in H. perforatum

Extensive research aimed at unravelling the genetic control of apomixis has not resulted in a consensus picture, suggesting that a number of distinct pathways have evolved in various species. This idea fits well with the notion that apomixis in general, and each apomictic trait in particular, evolved independently (van Dijk and Vijverberg, 2005). Both the mutation of genes involved in the control of sexual reproduction and the de novo evolution of genes have been proposed as underlying the determination of apospory and parthenogenesis. Studies aiming to understand the genetic and molecular factors underlying apomixis have been limited, since the asexual mode of reproduction is often associated with polyploidy and high degrees of heterozygosity, traits which make genetic and genomic analyses very difficult. In H. perforatum apomicts are able to develop embryos from aposporous fertilized egg cells, and others by parthenogenesis from meiotically reduced egg cells. This has been taken to suggest that distinct genetic factors control apospory and parthenogenesis, and that the two traits may be developmentally uncoupled, a possibility previously raised by Noack (1939). This hypothesis is further supported by the finding of H. perforatum genotypes which almost exclusively express only one component of apomixis or suppress the other (Matzk et al., 2001; Barcaccia et al., 2006).

The current experimental data have demonstrated that parthenogenetic capacity is preferentially expressed by aposporous egg cells. We have also documented the occurrence of aposporous egg cells in non-parthenogenetic individuals, as well as parthenogenetic development in meiotic egg cells. While apospory and parthenogenesis seem to be genetically linked (by ∼20cM) and so tend to be co-inherited, recombination can separate them. The analysis of segregation data reveals that the putative apospory and parthenogenesis loci are associated in a chromosome region where recombination is possible. Both traits are determined at the tetraploid level by a dominant simplex allele. While this genetic model differs from the idea that the inheritance of apomixis is governed by the allelic state at a single, fully dominant gene (Savidan, 1980; Leblanc et al., 1995; Bicknell et al., 2000), it is consistent with conclusions reached in other studies (Noyes and Rieseberg, 2000; Matzk et al., 2001; van Baarlen et al., 2002; van Dijk and Bakx-Schotman, 2004; Catanach et al., 2006; Noyes et al., 2006).

Aposporous plants are heterozygous and contain both sexual and aposporous alleles

The apospory-specific CAPS marker and the corresponding genomic locus HAPPY are both present in plants which express varying degrees of apospory. This suggests that genes required for the apospory trait are located within the HAPPY locus, but additional modifier genes might govern the level of trait expression. This model is similar to what has been proposed for the genetic basis of apomixis in Poa, where five different loci are involved, including a dominant gene which controls the initiation of apospory and a recessive one which acts to modulate the expression of the trait (Matzk et al., 2005). The segregation analysis performed here has shown that the HAPPY locus is dominantly inherited, and the apomictic parents tested are all simplex, just as is the case in Ranunculus auricomus (reviewed in Koltunow and Grossniklaus, 2003). The retention of the sexual allele allows the plant to revert to sexual seed production if apospory is prevented in some way.

It seems clear that the allelic constitution at the HpARI homoeologues is the primary genetic basis of aposporous reproduction in H. perforatum. Only a restricted region surrounding the apospory-specific HpARI allele shows sequence similarity to the sexual alleles. Sequences beyond this limited region do not show detectable similarity between aposporous and sexual alleles nor give any hit when Blasted against all known databases. The aposporous HAPPY sequence is rather distinct from the sexual one, so that a degree of hemizygosity is associated with this region in the simplex state (Figure 6). Hemizygosity has also been described at the ASGR locus in Pennisetum, in which the sexual types lack entirely the sequence corresponding to a part of the locus, while there is sequence homology along the remainder of the locus (Ozias-Akins et al., 1998).

Figure 6.  Schematic structure of the genomic region surrounding TpARI in sexual and aposporous types. ARI, the marker-containing sequence; T, truncated version of TpARI; DGCR and NPH, respectively the upstream and downstream genes. Tetraploid apomicts possess three sexual alleles and one aposporous allele. The red arrows indicate where sequence homology between the sexual and aposporous alleles is interrupted. BlastX negative sequence in the flanking region of the aposporous allele is shown by dashed lines. The light blue box indicates the position of the amplified fragment length polymorphism (AFLP) locus used to generate the cleaved amplified polymorphic sequence (CAPS) marker, and the black bar within it the informative EcoRI site.

The HAPPY locus includes a mutated ARIADNE orthologue

The truncated HpARI gene is a part of the HAPPY locus, making it possible to suggest that its product acts in a dominant negative fashion in a simplex dosage via an interaction with the gene product of the three remaining sexual alleles. The ARIADNE proteins belong to a family of E3 ligases present in yeast, plants and animals, and thought to be involved in the control of ubiquitin-dependent protein degradation (reviewed in Vierstra, 2003). The truncated HpARI gene is expressed, so its product may act to disrupt the normal Ring/U-box complexes formed by the non-truncated products, thereby impairing the progress of gametophytic development. This hypothesis will of course need testing by mutation and biochemical analyses of the HpARI protein. The function of ARIADNE genes, including ARI7 (At2g31510, theA. thaliana orthologue of HpARI), remains obscure (Mladek et al., 2003). Arabidopsis thaliana ARIADNE genes are mainly ubiquitously expressed, although there is evidence for differentially expressed family members, with AtARI10, for example, being preferentially active in pollen and during seed development (Mladek et al., 2003; Schmid et al., 2005). Arabidopsis AtARI7, which is by far the most closely related gene to HpARI of Hypericum, is broadly expressed in many tissues. Indeed the rather ubiquitous expression of HpARI is unexpected; however, ubiquitous expression does not exclude specific functionality if another more specific factor is needed for interaction. ARIADNE genes were first cloned and characterized in Drosophila and mammals. The Drosophila ariadne-2 mutant is lethal, thus showing that it plays some essential developmental role (Aguilera et al., 2000). However, because these genes are present in the form of a multi-member gene family in both animals and plants, progress in defining the role of individual members has been necessarily slow. Arabidopsis thaliana ARI7 is a tandemly duplicated gene, unlike the single, structurally intact HpARI. Although mutant alleles which disrupt A. thaliana ARI7 have been isolated (Table S6), none have as yet shown any evidence of developmental aberration, probably because of gene redundancy. A priority in our research programme is to optimize transformation methods in H. perforatum so that the effect of introducing the aposporous HpARI allele into a sexual H. perforatum host can be examined.

While we are currently concentrating on the functional role of HpARI in apospory, it is likely that further extension of the HAPPY locus sequence will be required to fully understand the control of apospory and its putative modifiers in H. perforatum. We have initiated the cloning of the HAPPY locus by sequencing BACs from fully aposporous 4x accessions. This line of research, along with a complementary project aimed at isolating candidate genes for parthenogenesis, should provide profound insights into the molecular basis of apomixis in H. perforatum and, more generally, in other plant species.

Experimental procedures

Plant materials

The 16 accessions of H. perforatum used were obtained from a variety of collection sites and botanical gardens in Europe (Table S1). Ten were apomictic, and the other six sexually reproducing, characteristics which are clearly distinguishable by flow cytometry (Matzk et al., 2000). A segregating population of 59 triploid individuals was obtained by pollinating sexual diploid plants with obligate apomictic tetraploid accessions. A second population of 72 tetraploid individuals was obtained by crossing a chromosome doubled sexual accession (R1) with tetraploid obligate apomicts. Two sexual progeny were taken as pollinators in crosses with a number of the obligate apomictic accessions (for the crossing scheme see Table 1). The resulting tetraploid progeny from these crosses were scored for the mode of reproduction using FCSS, and for the presence/absence of a putative apospory-linked marker.

FCSS

The FCSS screen is based on the measurement of DNA content in the embryonic and endosperm cells (Matzk et al., 2000, 2001). In diploids, sexual reproduction generates a diploid embryo and a triploid endosperm, whereas the seed of a tetraploid plant can include one of: 4C embryo and 6C endosperm cells (the product of sexually produced seeds from reduced, double-fertilized embryo sacs); 4C embryo and 10C endosperm cells (apomictic seeds from unreduced embryo sacs, and pseudogamous endosperm formation); 2C embryo and 6C endosperm (reduced, parthenogenetic amphihaploid progeny); or 6C embryo and 10C endosperm (unreduced, double-fertilized (poly)triploid B_III progeny).

Genetic analysis

Genetic analysis treated each mapping family as a BC₁ population. The estimation of linkage between apospory and parthenogenesis was conducted separately in the two populations. The pseudo-testcross strategy (Ritter et al., 1990) employed assumed the existence of a single-dose allele (simplex locus) in the polyploid progeny. Since the simplex allele is inherited by half of the gametes, when crossed to a nulliplex individual, the segregation ratio is expected to be 1:1. Similarly a cross with a diploid generates a 1:1 segregation at the triploid level. Observed segregation ratios were tested by the χ² statistic. Genetic distances were converted from recombination frequencies to centimorgans (cM) using the Kosambi (1944) mapping function (for details see Table S5).

AFLP and CAPS analysis

Total genomic DNA was isolated with the Invisorb Spin Plant kit (Invitek, http://www.invitek.de/) and AFLP analysis was performed following Potokina et al. (2002). Selective amplification was achieved using fluorescently labelled PstI-anchored primers with two selective nucleotides and MseI-anchored ones with three selective nucleotides. The amplicons were resolved by capillary electrophoresis and fragment sizes estimated by comparison with a size standard (Genescan-500 Rox, Applied Biosystems Inc., http://www3.appliedbiosystems.com/) supplemented with five additional DNA fragments of known length (ranging from 568 to 812 bp). The data were analysed using GeneScan software v3.0 (Perkin-Elmer ABI, http://www.perkinelmer.com/). To isolate specific fragments, the amplicons were re-electrophoresed through 4.5% polyacrylamide gels, visualized by silver-staining (Bassam et al., 1991), excised from the dried gel and eluted overnight in sterile water. An aliquot of the eluate represented the template for a further PCR, and these amplicons were sequenced either following their elution from a 1.5% agarose gel and QIAquick purification (Qiagen, http://www.qiagen.com/), or were first cloned into a pGEM-T vector (Promega, http://www.promega.com/). For conversion into a CAPS marker, primers (Table S3) were designed from the sequence of the AFLP fragment, and the resulting amplicons digested overnight with EcoRI, using standard conditions for restriction. The CAPS genotype was assessed by electrophoresis through 1.5% agarose gels.

BAC library and BAC clone characterization

DNA was extracted from the leaves of 1-month-old progeny of a single diploid sexual plant, and used to construct a partial HindIII BAC library (Amplicon Express, http://www.genomex.com/). Based on a genome size of ∼630 Mbp (http://www.rbgkew.org.uk/cval; Bennett et al., 1995; Barcaccia et al., 2007), the resulting 26,000 clone library represented a genome coverage of approximately sixfold. The inserts were released from a sample of 20 random clones by NotI digestion and sized by PFGE (1% agarose gel, 0.5 × Tris-borate-EDTA buffer (TBE), 12.5°C, 6 V cm⁻¹, 5 sec initial and 15 sec final pulse time, run time 16 h). This produced an estimated mean insert size of 110 kb. The BAC library was spotted onto Hybond N+ membranes (GE Healthcare, http://www.gehealthcare.com/) using a MicroGrid II robot (BioRobotics, http://www.digilabglobal.com). A probe containing the CAPS marker sequence was labelled with ³³P-α dCTP by random hexamer priming (Feinberg and Vogelstein, 1983), and hybridized to the library following Church and Gilbert (1984). Positive BAC clones retained after a further PCR test for the presence of the probe sequence were fingerprinted in triplicate by BamHI, EcoRI, XbaI, XhoI and HaeIII restriction (Luo et al., 2003). The digested DNA was labelled with SNaPshot labelling solution, and separated by capillary electrophoresis. Fragment analysis and contig assembly were facilitated by the software packages GeneMapper v4.0 (Applied Biosystems), FPPipeliner v.2.0 (BioinforSoft LLC, http://www.bioinforsoft.com/) and fpc (Sanger Institute, http://www.sanger.ac.uk/).

One BAC clone (25H09) was taken forward for sequencing. The BAC DNA was isolated using a Plasmid Purification Maxi kit applying the low-copy plasmid/cosmid protocol (Qiagen). The BAC DNA was randomly sheared (Hydroshear, Genomic Solutions, http://www.digilabglobal.com) and size-fractionated by agarose gel electrophoresis. Fragments of size 1–2 kb were treated with Klenow DNA polymerase (Fermentas, http://www.fermentas.com/), blunt-end ligated into pBluescriptSK (Stratagene, http://www.stratagene.com/) and sequenced. The resulting sequences were assembled using Sequencher v4.0 software (Gene Codes Corporation, http://www.genecodes.com/) set to an overlap minimum of 20 bp with 95% identity. Remaining gaps were closed by amplification based on primers derived from the flanking sequence. The final assembly consisted of 1253 sequences, giving about sixfold coverage. The accuracy of the assembly was confirmed by comparing the predicted and actual restriction digest profiles for a range of restriction enzymes. The final 141 941 bp sequence was annotated, based on the software packages Fgenesh+ with a dicot Markov model and GeneID (Parra et al., 2000), applying matrices specific for A. thaliana and tomato (Solanum lycopersicum). Refinement of these predictions were achieved by aligning the H. perforatum genomic sequence with TIGR transcript assemblies of Populus spp., Manihot esculenta, Euphorbia esula (Childs et al., 2007) and A. thaliana proteins (TAIR version 7), using GenomeThreader software (Gremme et al., 2005). Consensus gene models were derived by comparing the gene models with a reference protein database (UNIREF90; Suzek et al., 2007). Manual inspection of the consensus gene models retained four transposon-related and 20 protein-encoding genes.

Detection of SNPs

All 24 genes within 25H09 were amplified from four aposporous and four sexual plants (primer sequences given in Table S3). The resulting amplicons were cloned into pCR2.1 (Invitrogen, http://www.invitrogen.com/), and sequenced to detect any haplotype variation present.

Allele-specific expression analysis

Total RNA was isolated from both 4–5 mm and 6–8 mm long pistils as well as young leaves using an Invitek RNA isolation kit. These pistil lengths correspond to the onset of, respectively, megasporogenesis and megagametogenesis (Galla et al., in press). For first-strand cDNA synthesis, total RNA was treated with DNaseI (Invitrogen) followed by RT-PCR with RevertAid H Minus M-MuLV Revertase transcriptase (Fermentas) using random hexamer priming. The primers were designed to amplify both alleles uniformly and to enable them to be distinguished from one another by the presence of an EcoRV site in the aposporous allele (for primer sequences, see Table S3).

Allele quantification by pyrosequencing

Primers for pyrosequencing were derived using the SNP primer design v4.0 software (Biotage AB, http://www.biotage.com/). Pyrosequencing reactions were carried out using primers listed in Table S3, following the manufacturer’s standard protocols. Allele frequencies were estimated using Biotage AB software.

Southern hybridization

A 5-μg sample of genomic DNA was digested with EcoRI, separated by agarose gel electrophoresis and transferred onto a Hybond N+ membrane (GE Healthcare). The membrane was hybridized with a probe consisting of a radioactively labelled 220-bp sequence. Signal intensities (photo-stimulated luminescence mm⁻²) were quantified by a FUJIX Bas 2000 phosphoimager (Fuji Photo Film, http://fujifilm.com/). After background subtraction, the intensities of the 583-bp and 394-bp fragments were summed, and compared with that of the 977-bp fragment lacking the internal EcoRI site.

Genome walking

The genomic sequence of the target locus was extended using Genome Walker Universal kit (Clontech, http://www.clontech.com/) and Advantage2 Polymerase mix (Clontech), following the manufacturer’s instructions.

Acknowledgements

This study was primarily supported by a research innovation grant ‘Pakt für Forschung und Innovation 2007′ awarded to HB. A financial contribution towards the cost of constructing the BAC library was provided by IPK-Gatersleben. We are grateful for the skilled technical assistance of Andreas Czihal, Elke Liemann, Annett Busching and Sabine Skiebe. We thank Drs Daniela Schulte and Inge Matthies for help with, respectively, BAC fingerprinting and pyrosequencing. We also wish to thank Dr Tim Sharbel and Professor Konrad Bachmann for critical discussions. We thank smartEnglish (http://www.smartenglish.co.uk/) for linguistic help in the preparation of this manuscript.