Reciprocal hybridization at different times between Senecio flavus and Senecio glaucus gave rise to two polyploid species in north Africa and south-west Asia

Authors

  • J. W. Kadereit,

    Corresponding author
    1. Institut für spezielle Botanik und Botanischer Garten, Johannes Gutenberg-Universität Mainz, D-55099 Mainz, Germany;
    • Author for correspondence: Joachim W. Kadereit Tel: +49 61 31 39 25 33 Fax: +49 61 31 39 35 24 Email: kadereit@uni-mainz.de

    Search for more papers by this author
  • S. Uribe-Convers,

    1. Institut für spezielle Botanik und Botanischer Garten, Johannes Gutenberg-Universität Mainz, D-55099 Mainz, Germany;
    Search for more papers by this author
  • E. Westberg,

    1. Institut für spezielle Botanik und Botanischer Garten, Johannes Gutenberg-Universität Mainz, D-55099 Mainz, Germany;
    Search for more papers by this author
  • H. P. Comes

    1. Institut für spezielle Botanik und Botanischer Garten, Johannes Gutenberg-Universität Mainz, D-55099 Mainz, Germany;
    2. Fachbereich für Organismische Biologie, Paris-Lodron-Universität Salzburg, A-5020 Salzburg, Austria
    Search for more papers by this author

Summary

  • • The analysis of hybrid plant taxa using molecular methods has considerably extended understanding of possible pathways of hybrid evolution.
  • • Here, we investigated the origin of the tetraploid Senecio mohavensis ssp. breviflorus and the hexaploid Senecio hoggariensis by sequencing of nuclear and chloroplast DNA, and by analysis of the distribution of taxon-specific amplified fragment length polymorphism (AFLP) fragments.
  • • Both taxa originated from hybridization between the diploid Senecio flavus and Senecio glaucus. Whereas S. glaucus was the female parent in the origin of S. mohavensis ssp. breviflorus, S. flavus was the female parent in the origin of S. hoggariensis.
  • • The distribution of AFLP fragments suggests that S. hoggariensis is an allohexaploid species with two diploid genomes of S. glaucus and one diploid genome of S. flavus. The high frequency of S. flavus-specific fragments in S. mohavensis ssp. breviflorus is explained either as the result of introgression between a primary hybrid and S. flavus or as the result of intergenomic recombination in a primary hybrid. These two alternative processes cannot easily be distinguished.

Introduction

Interspecific hybridization plays an important role in plant and animal evolution (Arnold, 1997), and can result in new taxa via introgressive hybridization, homoploid hybrid speciation, and polyploid hybrid speciation (Abbott, 1992). Although this has long been known and accepted (Stebbins, 1950, 1959), much progress in understanding these different trajectories of hybrid evolution has been made in the recent past, mainly but not exclusively through the application of molecular methods (for a review, see Hegarty & Hiscock, 2005). The widespread occurrence of introgressive hybridization was recently reviewed by Rieseberg & Wendel (1993), and detailed insights have been gained into homoploid hybrid speciation (Rieseberg et al., 1995, 2003; Rieseberg, 1997). In polyploid hybrid speciation, much progress has been made in understanding mechanisms and rates of polyploid formation (Ramsey & Schemske, 1998; Otto & Whitton, 2000) and in appreciating the recurrent origin of polyploid species formation (Soltis & Soltis, 1999). Recently, interest has focused on the early evolution of polyploids (Soltis & Soltis, 1995, 2000, 2004; Leitch & Bennett, 1997; Bayer, 1999; Wendel, 2000; Ramsey & Schemske, 2002; Hegarty et al., 2005). The existence of intergenomic recombination was first suggested by the observation of substantial segregation in, for example, allopolyploid Nicotiana (Kostoff, 1938), Gilia hybrids (Grant, 1966a,b) and the recently evolved allohexaploid Senecio cambrensis Rosser (Ingram & Noltie, 1984). However, cytogenetic characterization of the chromosome complement of hybrid species such as Nicotiana tabacum (Kenton et al., 1993) and studies of experimental allopolyploids, particularly in Brassica (Song et al., 1995; Lukens et al., 2004; Pires et al., 2004) and Triticum-Aegilops (Feldman et al., 1997; Liu et al., 1998; Levy & Feldman, 2004), were needed to demonstrate conclusively that substantial intergenomic recombination can take place, resulting in the loss of parental characters and the origin of new hybrid characters within only a few generations of hybrid offspring (but see Liu et al., 2001; Ainouche et al., 2004). The study of artificial allopolyploids in Brassica by Song et al. (1995) also suggested that the amount of genome evolution is proportional to the degree of relatedness of the parent species, with more genome change occurring in distant crosses, and that the genome of the paternal parent changes more than that of the maternal parent.

The hybrid origin of new taxa is well documented in certain groups of Senecio. In the British Isles, hybridization between the diploid (2n = 20) Senecio squalidus L., itself a hybrid between Senecio aethnensis Jan. ex DC and Senecio chrysanthemifolius Poiret (Crisp, 1972; Abbott et al., 2000, 2002), and the tetraploid (2n = 40) Senecio vulgaris L. gave rise to the tetraploid Senecio vulgaris var. hibernicus Syme (Stace, 1977; Ingram et al., 1980; Abbott et al., 1992), the tetraploid Senecio eboracensis R.J. Abbott & A.J. Lowe (Irwin & Abbott, 1992; Lowe & Abbott, 2000, 2003; Abbott & Lowe, 2004), and, recurrently, the hexaploid (2n = 60) Senecio cambrensis (Rosser, 1955; Weir & Ingram, 1980; Ashton & Abbott, 1992; Harris & Ingram, 1992). Moreover, the hexaploid (2n = 60) Canary Islands endemic Senecio teneriffae Schultz Bip. (Kadereit, 1984) is probably a hybrid derivative of S. vulgaris and the diploid (2n = 20) Senecio glaucus L. (Lowe & Abbott, 1996). The species involved in these various hybridizations are all closely related to each other (Comes & Abbott, 2001).

A much more distant cross was apparently involved in the origin of two tetraploid (2n = 40) subspecies of Senecio mohavensis A. Gray, the North American ssp. mohavensis and the south-west Asian ssp. breviflorus (Kadereit) M. Coleman. Subspecies breviflorus, originally described as Senecio flavus (Decne.) Schultz Bip. ssp. breviflorus Kadereit (Kadereit, 1984), from south-west Asia, has been shown to be a tetraploid hybrid derivative of diploid S. glaucus and diploid S. flavus through a combination of chloroplast DNA (cpDNA), internal transcribed spacer (ITS), random amplified polymorphic DNA (RAPD), isozyme and chromosomal evidence (Liston et al., 1989; Liston & Kadereit, 1995; Coleman et al., 2001, 2003; Comes & Abbott, 2001). Senecio glaucus is widely distributed over north Africa and south-west Asia, and S. flavus grows in north and south-west Africa (Alexander, 1979; Kadereit, 1984; Coleman et al., 2001). Chloroplast DNA analysis provides strong evidence that in this cross S. glaucus was the female parent (Comes & Abbott, 2001), and homogenization of ITS also took place towards S. glaucus (Comes & Abbott, 2001; Coleman et al., 2003). In the study by Coleman et al. (2003), S. flavus was found to be more closely related to other genera of Asteraceae-Senecioneae than to S. glaucus. This clearly demonstrates that the two species are only very distantly related. Indeed, a broad survey of ITS variation in Asteraceae-Senecioneae suggested that S. flavus may not even be part of a re-defined Senecio (P. B. Pelser et al., Department of Botany, Miami University, Oxford, OH, USA, personal communication). These molecular data confirm earlier experimental work. Thus, Alexander (1975, 1979) did not succeed in hybridizing S. flavus with S. glaucus, Senecio leucanthemifolius Poiret, Senecio aegyptius L. or Senecio hoggariensis Batt. & Trab., and of the three hybrid individuals between S. flavus and S. squalidus, a close relative of S. glaucus, obtained by Kadereit (1983) two were completely pollen-sterile and one had a very low percentage (3.5%) of stainable pollen grains.

Considering the differences in chromosome number among S. glaucus, S. flavus (2n = 20) and S. mohavensis (2n = 40) in combination with the very close morphological similarity between S. flavus and S. mohavensis, it remains unclear by which mechanism S. mohavensis originated. Coleman et al. (2001) considered the possibilities of an origin either via autopolyploidy (of S. flavus) followed by introgression with S. glaucus or via allopolyploidy. Here we investigate the systematic distribution of species-specific amplified fragment length polymorphism (AFLP) fragments in S. mohavensis, its putative parental taxa (S. flavus and S. glaucus), and several close relatives (S. aethnensis, Senecio vernalis Waldst. & Kit. and S. vulgaris) to further investigate this problem. In a similar way, a qualitative analysis of AFLP fragment variation recently was used to interpret hybrid evolution in Achillea by Guo et al. (2005). We will discuss the possibility that the genomic constitution of S. mohavensis, as revealed by AFLP fragment distribution, may have arisen through allopolyploid intergenomic recombination, but will show that this unusual mechanism of allopolyploid hybrid speciation may be difficult to distinguish from introgression. We also include another species of Senecio in this study, the hexaploid (2n = 60) S. hoggariensis from north Africa and south-west Asia (Hoggar Mountains, Tibesti Mountains, south-east Egypt and Sinai; Alexander, 1979). This species has morphological similarities with both S. glaucus and S. flavus and thus may be yet another hybrid derivative of these two species. In addition to AFLP variation, we analysed sequences of both nuclear ribosomal DNA (ITS) and three cpDNA gene regions (the trnT-L intergenic spacer, the trnL intron and the 3′trnK intron) in S. hoggariensis and its close relatives. As the mode of inheritance of cpDNA in Senecio is strictly maternal (Harris & Ingram, 1992), examination of relationships with both nuclear and cytoplasmic markers not only provides a more accurate picture of the phylogenetic relationships of this species, but also allows identification of the maternal and paternal parents.

Materials and Methods

Plant material

Plant material was taken from specimens grown in the experimental glasshouse of Mainz Botanic Garden, if available, and stored at −80°C. Alternatively, herbarium material was used. Details about the origin of plant material used for AFLP and/or DNA sequence analysis are given in Table 1.

Table 1. Origin of plant material
TaxonCollectors and/or collection no. (herbarium)OriginType of materialGenBank accession numbers
3′trnK intron trnL intron trnT-L intergenic spacerITS region
  1. All accessions were used in the amplified fragment length polymorphism (AFLP) analysis, with subsets used for sequencing of the 3′trnK intron, trnL intron, and/or trnT-L intergenic spacer (ITS) regions of the chloroplast genome. In addition, three different sequences of the entire ITS region (ITS-1, 5.8S and ITS-2) of nuclear ribosomal DNA were retrieved from the single individual of S. hoggariensis after cloning (see text).

  2. NHM, Natural History Museum. Type of material: F, fresh; H, herbarium specimen; D, silica-dried.

Senecio flavus H. Merxmüller & W. Giess 28206 (M)Namibia: Hanchabfontein FarmF DQ208168 DQ208176
S. flavus D. Podlech 33668 (M)Algeria: Wilaya BecharF DQ208184
S. flavus D. Podlech 40663 (RSA)Morocco: Wadi NounF
S. mohavensis ssp. breviflorusDanin et al. 5–1940 (HUJ)Egypt: SinaiF
S. mohavensis ssp. breviflorusListon & Boaz 255–1 (HUJ)Israel: Dead SeaF DQ208178
S. mohavensis ssp. breviflorusK. Naylor 279 (E)Saudi Arabia: JabarlH
S. mohavensis ssp. breviflorusR. Western 574 (E)UAR: Abu Dhabi EmirateF DQ208170 DQ208186
S. mohavensis ssp. mohavensisEngard & Hodgson 2121 (DES)USA: Arizona, La Paz CountyF DQ208171 DQ208179 DQ208187
S. mohavensis ssp. mohavensisElias 10190 (RSA)USA: Arizona, Maricopa CountyF
S. aethnensis NHM Paris 88–167CultivationF DQ208172 DQ208180 DQ208188
S. glaucus ssp. glaucusR. J. Abbott & H. P. Comes (STA)Israel: Judean Coast, Ashdod.F DQ208166 DQ208174 DQ208182
S. hoggariensis F. Ehrendorfer 20b (MZA)Algeria: Hoggar-Massif, descent from AssekremF DQ208167 DQ208175 DQ208183 DQ208189DQ208191
S. hoggariensis D. Podlech 33320 (M)Algeria: Hoggar-Massif, Wilaya TamanrassetH
S. vernalis H. P. Comes 3Germany: Ingelheim/RheinD DQ208165 DQ208173 DQ208181
S. vernalis H. P. Comes 7Germany: Ingelheim/RheinD
S. vulgaris H. P. Comes 8Germany: MainzD DQ208177
S. vulgaris H. P. Comes 9Germany: MainzD DQ208169 DQ208185

DNA extraction

Plant material was ground in liquid nitrogen. DNA was extracted using the DNeasy‘ plant minikit (Qiagen, Hilden, Germany), following the manufacturer's instructions. DNA concentration was measured spectrophotometrically with a GeneQuant RNA/DNA calculator (Pharmacia, Cambridge, UK), or estimated visually in ethidium bromide-stained agarose gels. Extracted DNA was stored at −20°C.

AFLP procedure

Nineteen individuals representing eight taxa were included in the AFLP analysis (Table 1). Three of the S. hoggariensis individuals were offspring of the same parent (Ehrendorfer 20b; see Table 1). All reactions and polymerase chain reaction (PCR) conditions followed the procedure of Vos et al. (1995) with modifications and chemicals as described in Kropf et al. (2003), except that 150 ng of DNA was used for the restriction–ligation step. The restricted DNA was simultaneously ligated to adaptors (EcoRI: 5′-CTCGTAGACTGCGTACC-3′/5′-AATTGGTACGCAGTC-3′; MseI: 5′-GACGATGAGTCCTGAG-3′/5′-TACTCAGGACTCAT-3′) at 23°C for 14 h. A preselective amplification was carried out using the primers E+1 (5′-GACTGCGTACCAATTCA-3′) and M+1 (5′-GATGAGTCCTGAGTAAC-3′). For the selective amplifications, the three primer combinations E37/M61 (E+1 + CG/M+1 + TG), E39/M57 (E+1 + GA/M+1 + GG) and E45/M54 (E+1 + TG/M+1 + CT) were used.

Products resulting from selective amplification were separated on a 6% polyacrylamide gel with an internal size standard (ROX 500; Applied Biosystems, Weiterstadt, Germany). The gel was run for approx. 4 h on an ABI Prism 377 automated sequencer (Applied Biosystems) and analysed with genescan software (version 3.1; Applied Biosystems). Fragments in the range from 75 to 500 bp were scored automatically as present (1) or absent (0) with the genotyper analysis software (version 2.1; Applied Biosystems) and manually corrected after visual inspection of the electrophoretograms. Ambiguous peaks were scored as missing data.

ITS sequence analysis

The entire ITS region, comprising ITS-1, 5.8S and ITS-2, was sequenced in one accession of S. hoggariensis (Table 1), using primers 18S and 28S of Muir & Schlötterer (1999). Amplifications were performed in 25-µl volumes containing 1.0 µl of 50 mm MgCl2, 2.5 µl of 10 × BioTherm PCR buffer (GeneCraft, Münster, Germany), 0.25 µl of 20 mm dNTP, 1.0 µl of each primer at 25 pmol µl−1, 0.75 U BioTherm Taq polymerase (GeneCraft), and 0.4 µl of genomic DNA (c. 5–40 ng). PCR conditions consisted of 60 s at 94°C followed by 35 cycles of 18 s at 94°C, 30 s at 55°C and 60 s at 72°C, and a post-treatment of 78 s at 55°C and 8 min at 72°C. The ITS fragments were then cloned using the TOPO TA Cloning Kit for Sequencing (Invitrogen, Carlsbad, CA, USA) following the manufacturer's instructions. Eight colonies were picked for sequencing in order to check for intra-individual variation. The ITS region was cycle-sequenced with the BigDye® Terminator Cycle Sequencing Kit v3.1 (Applied Biosystems) using the same primers as for the amplification. PCR conditions were 10 s at 96°C, followed by 4 min at 55°C repeated 30 times. The three different ITS sequences generated in this study have been deposited in GenBank (see Table 1 for accession numbers).

Chloroplast DNA sequence analysis

The trnT-L intergenic spacer, the trnL intron, and the 3′trnK intron were sequenced for eight taxa. In some cases different accessions of the same taxon were used for different cpDNA regions (Table 1). The trnT-L intergenic spacer and the trnL intron were amplified with primers a, b, c and d from Taberlet et al. (1991). The 3′trnK intron was amplified with primers 1023F and 1559R from Pelser et al. (2002). PCR conditions for all three cpDNA regions followed Pelser et al. (2002). Purified PCR products of the trnL and 3′trnK introns were cycle-sequenced in one direction using primers d and 1559R, respectively, while those of the trnT-L spacer were sequenced in both directions using the same primers as in the amplification. The sequencing reactions followed the same protocol as for the ITS region. The 24 different cpDNA sequences generated in this study have been deposited in GenBank (see Table 1 for accession numbers).

Data analyses

AFLPs  The AFLP data matrix was inspected for taxon-specific fragments that were recorded for all taxa. AFLP fragments found in only one taxon were regarded as taxon-specific (or private) even when they were not fixed in that taxon. When determining taxon-specific fragments in S. flavus and S. glaucus, fragments shared between these two species and S. mohavensis were considered private to the former two species because they are known to be the parents of S. mohavensis. The proportion of taxon-specific fragments of potential parental taxa present in S. hoggariensis, S. mohavensis ssp. mohavensis or S. mohavensis ssp. breviflorus, respectively, was then used to infer the mode of participation of parental taxa in the origin of these suspected hybrids. We assume that AFLP fragments shared among taxa are homologous but realize that this is not necessarily the case (Bussell et al., 2005; Koopman, 2005; see also Discussion).

Sequence data  All DNA sequences (ITS and cpDNA) were edited and automatically aligned using the program sequencher‘ 4.1 (Gene Codes Co., Ann Arbor, MI, USA), and adjusted visually where needed. The ITS data of S. hoggariensis were analysed together with 29 sequences of 21 Senecio species obtained from GenBank [accession numbers AJ400777, AJ400779AJ400786, AJ400789, AJ400794AJ400801, AJ400803AJ400809, AJ400812, AJ400813 (Comes & Abbott, 2001), AF457436 and AF457437 (Coleman et al., 2003)] using S. flavus as the outgroup (see Fig. 1). Phylogenetic relationships were estimated using the maximum-likelihood criterion implemented in paup* v4.0b10 (Swofford, 2002). Using the program modeltest 3.5 (Posada & Crandall, 1998), the model of sequence evolution selected was SYM+G (Zharkikh, 1994) with gamma shape parameter α = 0.1544 based on the Akaike information criterion. A heuristic search was then performed in paup* using the parameter settings obtained from modeltest, 500 random addition sequence replicates, tree bisection reconnection (TBR) branch swapping, steepest descent off, and multrees option on. Support for nodes was estimated by 100 bootstrap replicates (Felsenstein, 1985) using the model parameters obtained for the original data, 10 replications of random sequence addition, and TBR branch swapping. The aligned cpDNA sequence matrix (with all three regions combined) was analysed in paup* using maximum parsimony and the exhaustive search strategy.

Figure 1.

Maximum-likelihood internal transcribed spacer (ITS) tree rooted with Senecio flavus. Bootstrap support values are shown above branches. Note that the three different ITS sequences of Senecio hoggariensis were retrieved from a single individual after cloning (see text and Table 1).

Results

AFLPs

The three AFLP primer combinations used (E37/M61, E39/M57 and E45/M54) generated a total of 562 fragments across the 19 accessions and eight taxa analysed, with 156, 207 and 199 fragments per combination, respectively. Across the entire data set, 560 fragments were polymorphic, of which 125 were unique to individual accessions. Table 2 shows the total number of taxon-specific fragments for each taxon, and the proportion of those fragments found in S. hoggariensis and S. mohavensis subspp. mohavensis and breviflorus, respectively. A high percentage (65%) of species-specific fragments from S. glaucus were found in S. hoggariensis, compared with 22% from S. flavus, 22% from S. aethnensis, 19% from S. vernalis and 8% from S. vulgaris. In the case of S. mohavensis ssp. breviflorus, the majority of taxon-specific fragments were derived from S. flavus (41%), followed by those from S. glaucus (17%). The other species provided no major contribution (Table 2). Similar results were obtained for S. mohavensis ssp. mohavensis, with 39% from S. flavus, 19% from S. glaucus and very little contribution from the other species (Table 2).

Table 2. Comparison of numbers and percentages of taxon-specific fragments among Senecio hoggariensis, Senecio mohavensis ssp. breviflorus, Senecio mohavensis ssp. mohavensis, and potential ancestors
TaxonTaxon-specific fragmentsTaxon-specific fragments in S. hoggariensisTaxon-specific fragments in S. mohavensis ssp. breviflorusTaxon-specific fragments in S. mohavensis ssp. mohavensis
S. hoggariensis  33
S. mohavensis ssp. breviflorus 17 1 (5%)
S. mohavensis ssp. mohavensis 17 0 (0%)
S. flavus 10022 (22%)41 (41%)39 (39%)
S. glaucus ssp. glaucus 4831 (65%) 8 (17%) 9 (19%)
S. aethnensis  32 7 (22%) 0 (0%) 2 (6%)
S. vernalis  37 7 (19%) 2 (5%) 1 (3%)
S. vulgaris  45 4 (8%) 2 (4%) 2 (4%)

ITS

The sequencing of eight clones of the entire ITS region (including 5.8S) from a single individual of S. hoggariensis yielded three sequences, differing from each other in one, two and three nucleotide positions, respectively. The complete alignment, including the three ITS sequences obtained from S. hoggariensis, resulted in 651 nucleotide positions, of which 555 were constant and 44 parsimony informative. The maximum-likelihood analysis of the ITS data showed that S. hoggariensis falls into a clade with S. mohavensis, S. glaucus, and several other diploid taxa such as S. squalidus (Fig. 1). Bootstrap support is high for this ‘S. glaucus/S. mohavensis clade’, while relationships within the clade are largely unresolved. More importantly in the present context, the species of this clade are clearly more closely related to each other than to S. flavus.

The numbers of differences in nucleotide positions among S. glaucus, S. mohavensis and S. hoggariensis are roughly equal. The ITS sequence of the S. glaucus (ssp. glaucus) accession from Israel (AJ400786) differs from the three S. hoggariensis sequences in four, five and six nucleotide positions, respectively, and in five positions from those of S. mohavensis. The two subspecies of S. mohavensis have the same ITS sequence. The Moroccan accession of S. glaucus[ssp. coronopifolius (Maire) Alexander; AJ400785] differs in three, four and five positions from S. hoggariensis, and in six positions from S. mohavensis. Finally, the ITS sequence of S. mohavensis differs in five, six and seven nucleotides from the three ITS sequences of S. hoggariensis, respectively.

Chloroplast DNA

The aligned sequences of the 3′trnK intron, trnT-L spacer and trnL intron had lengths of 520, 502 and 464 bp, respectively, and contained seven substitutions and a 6-bp indel (Table 3). One short and highly variable region of the trnT-L spacer could not be used for parsimony analysis because of complicated patterns in length variation of a poly-adenine stretch. This region was thus excluded from the alignment. Of the variable positions, one substitution (trnT-L spacer: position 502) and the indel were shared by S. hoggariensis and S. flavus, and one substitution (trnT-L spacer: position 330) was common to S. vulgaris, S. aethnensis and S. glaucus (Table 3). Of the remaining substitutions, three were unique to S. hoggariensis and two to S. flavus. The maximum-parsimony analysis grouped S. flavus and S. hoggariensis as sister species with 85% bootstrap support (not shown).

Table 3. Distribution of nucleotide substitutions and indels in the trnT-L intergenic spacer, 3′trnK intron, and trnL intron among Senecio hoggariensis, Senecio mohavensis ssp. breviflorus, Senecio mohavensis ssp. mohavensis, and potential ancestors
Taxon trnT-L intergenic spacer3′trnK intron trnL intron
PositionPositionPosition
255133050271104–109412396
S. hoggariensis CCTTT+TA
S. mohavensis ssp. breviflorusCCTAGCG
S. mohavensis ssp. mohavensisCCTAGCG
S. flavus TATTG+CG
S. glaucus ssp. glaucusCCCAGCG
S. aethnensis CCCAGCG
S. vernalis CCTAGCG
S. vulgaris CCCAGCG

Discussion

In the present study we employed sequence data for nuclear ribosomal (ITS) and chloroplast DNA as well as AFLPs to investigate the origin of the tetraploid S. mohavensis ssp. breviflorus and the hexaploid S. hoggariensis. We note that objections may be raised to the use of AFLPs in such a context, because the homology between co-migrating fragments may be difficult to establish (Guo et al., 2005). In fact, the amount of nonhomology between such fragments might be high at the sequence level, especially for distantly related taxa (Mechanda et al., 2004). However, while this source of error has the potential to produce inaccurate genetic similarity measures, namely phylogenetic reconstructions (e.g. Robinson & Harris, 1999), the major interest here is in the relative genomic contributions from putative parental taxa, in terms of taxon-specific fragments, to assumed hybrid derivatives. Hence, even if a certain percentage of, for example, the S. glaucus-specific fragments found in, for example, S. hoggariensis should be nonhomologous, such percentages are likely to vary little across all potential parent/hybrid pairs, and thus should not affect our overall proportions of genomic contributions (see Table 2).

Furthermore, it must be emphasized that the AFLP data obtained are limited in extent; in particular, our absolute values related to ‘taxon-specific fragments’ are considered as approximations because only one or a few accessions from each taxon have been analysed. Accordingly, ‘taxon-specific fragments’ are no more than ‘accession-specific fragments’. That said, all conclusions to be drawn from the analysis of these data would only be strengthened by more detailed intra(sub)specific surveys of AFLP variation. However, for S. hoggariensis, any wider sampling, at least in the Hoggar and Tibesti Mountains, is presently not possible for political reasons.

S. mohavensis ssp. breviflorus

Our qualitative analysis of AFLP fragment distribution among S. mohavensis ssp. breviflorus and its putative parents S. glaucus and S. flavus both confirms and extends existing knowledge. The analysis of cpDNA (Liston & Kadereit, 1995; Comes & Abbott, 2001), ITS (Comes & Abbott, 2001; Coleman et al., 2003), RAPD (Comes & Abbott, 2001), chromosomal (Coleman et al., 2001) and morphological variation (Kadereit, 1984; Coleman et al., 2001) previously demonstrated that S. mohavensis ssp. breviflorus originated as a polyploid hybrid taxon from a cross between S. glaucus and S. flavus, that in this cross S. glaucus was the female parent, and that the homogenization of the strongly divergent ITS sequences of the two parent species was towards S. glaucus. The major sources of evidence for the participation of S. flavus in the origin of S. mohavensis ssp. breviflorus, originally described as S. flavus ssp. breviflorus, consisted of RAPD variation, where S. mohavensis ssp. breviflorus grouped with S. flavus rather than with S. glaucus (Comes & Abbott, 2001), flower morphology, where S. mohavenis ssp. breviflorus was found to differ from S. flavus mainly in the presence of shortly ligulate capitula with epappose ray floret achenes (Kadereit, 1984), leaf shape (Coleman et al., 2001) and geographical distribution (Kadereit, 1984; Coleman et al., 2001).

The presence of species-specific AFLP fragments of both S. glaucus and S. flavus in S. mohavensis ssp. breviflorus (as well as in ssp. mohavensis; see Table 2) confirms the participation of these two species in the formation of S. mohavensis, and the higher proportion of S. flavus (41%) than S. glaucus (17%) fragments in S. mohavensis ssp. breviflorus is in complete accordance with RAPD and phenotypic variation. Our analysis of the among-species distribution of species-specific AFLP fragments clearly shows that S. mohavensis is not a simple allotetraploid species, and an explanation for the differential proportions of S. glaucus and S. flavus fragments is required.

Coleman et al. (2001) suggested either (1) a combination of autopolyploidy and introgressive hybridization, or (2) allopolyploidy as possible pathways for the origin of S. mohavensis. Although considering the presence of duplicated isozyme loci in S. mohavensis ssp. breviflorus in relation to S. flavus (Liston et al., 1989) as evidence in favour of allopolyploidy, Coleman et al. (2001) concluded that the molecular data available are insufficient to allow an unequivocal decision to be made between these two pathways. Three additional pathways can be considered. It is possible that (3) introgression took place between an allopolyploid hybrid and S. flavus, (4) a homoploid S. glaucus × S. flavus segregant was stabilized through autopolyploidization, or (5) intergenomic recombination in an allopolyploid S. glaucus × S. flavus hybrid resulted in fast genomic rearrangements. The last possibility was also briefly discussed by Coleman et al. (2001).

Of these five possibilities, i.e. (1) autopolyploidy combined with introgression either before or after polyploidization, (2) simple allopolyploidy, (3) allopolyploidy plus introgression, (4) homoploid segregation and autopolyploidization, and (5) allopolyploidy plus intergenomic recombination, alternatives (1), (3) and (4) require the possibly high sterility barriers between S. glaucus and S. flavus (Alexander, 1975, 1979; Kadereit, 1983) and/or ploidy barriers to be overcome, and alternative (2) appears the least likely. As clearly pointed out by Coleman et al. (2001), more or less simple additivity of parental characters would be expected in the case of simple allopolyploidy. This, indeed, has been observed in other AFLP analyses of other hybrids or hybrid taxa (O’Hanlon et al., 1999; Hedrén et al., 2001; Hodkinson et al., 2002) where, by and large, hybrids were intermediate between their parents. Considering these arguments, S. mohavensis appears to have originated by a combination of auto- or allopolyploidy and introgression, where introgression is defined as the ‘permanent incorporation of genes from one set of differentiated populations into another’ (Rieseberg & Wendel, 1993), by homoploid segregation and autopolyploidization, or by allopolyploidy and intergenomic recombination, whereby these last two processes are assumed to involve the homoploid or polyploid hybrid only. The difficulty of distinguishing these processes has previously been pointed out by, for example, Stebbins (1950) and Heiser (1973). When it is considered that ‘from a mechanistic standpoint, hybridization and introgression may be viewed as a type of recombination’ (Rieseberg & Wendel, 1993), it becomes obvious that these two possible processes differ only marginally. In both cases, recombination among chromosomes of two different parent species (or other ‘differentiated populations’) takes place, and differences are found only in the crossing scheme. Whereas in the case of introgression backcrosses to parental species or populations are involved (although not as the only type of cross possible; Rieseberg & Wendel, 1993; Rieseberg, 1997), in the case of intergenomic recombination the parental species or populations do not participate in the further evolution of the homoploid or polyploid hybrids. Accordingly, polyploidization/introgression vs polyploidization/intergenomic recombination, in whichever temporal sequence, should not be considered fundamentally different or mutually exclusive pathways of polyploid or homoploid hybrid evolution. This also suggests that distinction of these two processes may prove difficult. In the light of the possibly high sterility of diploid hybrids between S. glaucus and S. flavus, we consider allopolyploidy plus intergenomic recombination the most likely mechanism for the origin of S. mohavensis ssp. breviflorus.

S. hoggariensis

Apart from its taxonomic treatment by Alexander (1979), S. hoggariensis has not been subject to further analysis. The species is similar to S. glaucus in leaf morphology and overall habit, and to S. mohavensis in having very narrow capitula (also found in S. flavus) with short purple ligules. In combination with its chromosome number of 2n = 60 (Alexander, 1979; HPC & JWK, unpublished data), this raised the possibility that the diploid S. glaucus and the tetraploid S. mohavensis may have been involved in the origin of the hexaploid S. hoggariensis. This hypothesis, however, is not supported by our molecular data. Instead, our data clearly show that S. hoggariensis, like S. mohavensis, originated from a cross between S. glaucus and S. flavus, here with the latter species as the maternal parent.

Chloroplast DNA variation showed that S. hoggariensis is most similar to S. flavus, sharing an indel in the 3′trnK intron sequence and one substitution in the trnT-L spacer sequence (Table 3). This clearly suggests that S. flavus was the female parent. At the same time, S. hoggariensis differs from S. flavus at five positions, three of which are unique. The analysis of ITS sequence variation showed that S. hoggariensis falls into a clade with S. glaucus and S. mohavensis, and that this clade is far removed from S. flavus (Fig. 1). The ITS sequences of S. hoggariensis and S. glaucus differ in three to six nucleotide positions. Finally, our qualitative analysis of AFLP fragments (Table 2) showed that S. hoggariensis contains taxon-specific fragments from S. glaucus, S. flavus and S. mohavensis ssp. breviflorus, but also from the other four species included in the analysis. The proportional representation of species-specific fragments in S. hoggariensis, however, is very different. Whereas 65% (31 of 48) of the S. glaucus-specific fragments and 22% (22 of 100) of the S. flavus-specific fragments could be detected in S. hoggariensis, only 5% (one of 17) of ssp. breviflorus-specific fragments were detected in this species. These findings suggest that ssp. breviflorus was probably not involved in the origin of S. hoggariensis.

Considering all three molecular data sets together, we can speculate that S. hoggariensis originated in one of two ways: (i) in an initial step, S. flavus as female parent and S. glaucus as male parent gave rise to an allotetraploid hybrid (now extinct or unknown) that in turn became the female parent in another cross with S. glaucus, resulting in the formation of the allohexaploid S. hoggariensis; (ii) an initial triploid hybrid could have formed as a result of fertilization of a reduced egg of S. flavus by an unreduced pollen grain of S. glaucus, followed by chromosome doubling. In either instance, homogenization of ITS repeats probably occurred towards S. glaucus at an unknown stage during polyploid formation. While there is no evidence allowing us to decide between the above scenarios, both are favoured here because, in contrast to other possible pathways, they predict that the proportional representation of parental AFLP fragments in the allohexaploid S. hoggariensis should approximate 66% of fragments from S. glaucus (four haploid chromosome sets) and 33% of fragments from S. flavus (two haploid chromosome sets). These proportions conform remarkably well to those observed, with S. hoggariensis containing 65% of fragments from S. glaucus and 22% of fragments from S. flavus (Table 2). Moreover, these proportions close to expectation imply that no major genomic rearrangements, as discussed above as a possibility for the origin of S. mohavensis ssp. breviflorus, appear to have taken place in S. hoggariensis. However, we cannot exclude the possibility that the divergent homoeologous subgenomes of the latter species have not undergone significant epigenetic changes since their reunion, such as gene (in)activation through alterations in cytosine methylation (see Levy & Feldman, 2004), and this might be further explored in the future using methylation-sensitive AFLP (MSAP; Reyna-Lopez et al., 1997; Ainouche et al., 2004).

Although species-specific fragments of S. aethnensis (seven of 32; 22%), S. vernalis (seven of 37; 19%) and S. vulgaris (four of 45; 8%) were found in S. hoggariensis, we do not assume that these species were involved in the origin of S. hoggariensis. Instead, we prefer to postulate that either these fragments are nonhomologous or that they are plesiomorphic and the result of incomplete lineage sorting. It is also conceivable that not all of them are truly species-specific. Wider sampling of S. glaucus might have detected at least some of these fragments.

The relative ages of S. mohavensis ssp. breviflorus and S. hoggariensis

Our data seem to indicate that S. hoggariensis is of somewhat greater age than S. mohavensis ssp. breviflorus. In the chloroplast sequences, S. hoggariensis differs from S. flavus as its maternal parent in five nucleotide positions, three of which are unique to S. hoggariensis. In contrast, S. mohavensis differs from S. glaucus (Israel) as its maternal parent in only one position, which is not unique. In the ITS sequences the difference in nucleotides between S. glaucus (Morocco/Israel) and S. mohavensis to that between S. glaucus and S. hoggariensis is roughly equal (see also Fig. 1). In the AFLP data (Table 2), S. hoggariensis contains 33 species-specific fragments in comparison to 17 specific fragments in S. mohavensis ssp. breviflorus. The number of specific fragments found in S. hoggariensis is comparable to the numbers found in S. glaucus (35), S. aethnensis (29), S. vernalis (35), and S. vulgaris (40).

The split between a S. vernalis/S. vulgaris clade and a S. glaucus/S. mohavensis clade (see also Fig. 1) was estimated to have taken place 3.25 ± 0.96 Ma, the S. glaucus/S. mohavensis clade was estimated to have diversified 1.02 ± 0.55 Ma, and S. mohavensis was estimated to have originated less than 1 Ma (not exactly dated) by Coleman et al. (2003). Comparing the above approximations for species-specific fragments under the assumption of at least roughly clock-like accumulation of mutations in AFLPs (as demonstrated for other dominant fingerprint markers; Espinasa & Borowsky, 1998), and considering the date of origin of S. glaucus, we here postulate that S. hoggariensis originated very soon after the formation of the S. glaucus lineage around 1 Ma. According to Frenzel et al. (1992), xerophilous vegetation covered substantially larger areas around the Saharan high mountains during the last glacial maximum than it does today. Assuming that this also was the case during earlier Quaternary glacials, the highly disjunct distribution of S. hoggariensis in the Tibesti and Hoggar Mountains and the Sinai peninsula still must have involved substantial long-distance dispersal, but over shorter distances than those separating the three disjunct distribution areas of this species today.

Acknowledgements

We wish to thank all collectors of plant material listed in Table 1, and are particularly grateful to F. Ehrendorfer (Vienna) for the collection of Senecio hoggariensis. We also gratefully acknowledge very helpful comments and suggestions made by two anonymous reviewers.

Ancillary