Morphometric and molecular evidence for taxonomic recognition of a new subspecies of Armeria filicaulis ( Plumbaginaceae ) por

A new subspecies of the Betic-Riffean species Armeria filicaulis is formally described –subsp. alfacarensis– from crystalline dolomitic sands NE of Granada (SE Spain). Its representatives exhibit a consistent morphological combination of characters on the inflorescence and flowers that support its taxonomic recognition. This is shown by means of a multivariate morphometric analysis. In one of the sites of the new taxon, where it is sympatric with A. villosa subsp. bernisii, nuclear ribosomal ITS sequences reveal gene flow between the two species, which is congruent with RAPD data and chloroplast trnL-F haplotypes. It is discussed that gene flow detected is not responsible for the origin of the new subspecies but a recent event. According to this, a differentiated biological entity related to other forms of A. filicaulis inhabiting dolomitic sands in eastern Andalusia, existed previously to gene flow events. The patterns detected (and inferred events) in A. filicaulis throughout its area suggest that the compilospecies model fits the evolution of this species as does that in A. villosa (as reported elsewhere) although not as closely.


Introduction
Armeria filicaulis (Boiss.)Boiss. is endemic to the SE quadrant of the Iberian Peninsula with a small disjunct area in the Moroccan Rif (Nieto Feliner, 2002).Although no intraspecific taxa were recognized in the last treatment of the genus for the Iberian Peninsula (Nieto Feliner, 1990), two subspecies were described in subsequent works to account for part of the morphological variability displayed by A. filicaulis in different areas.
Our study focused on the Sierra Nevada massif detected populations NE of Granada (roughly between Alfacar and La Peza, Fig. 1) that exhibit distinct characters as compared to other populations of A. filicaulis.Alike with subsp.trevenqueana, these populations have leaves that are indistinguishable from those in the core of the species and occur on crystalline dolomitic sands.As subsp.trevenqueana, they differ from the rest of the populations by characters in inflorescence, bracts and flowers.This paper aims at presenting the morphological evidence that justifies taxonomic recognition at the subspecific level for the populations of Alfacar-La Peza as well as the molecular information provided by three markers (nuclear ribosomal, chloroplastic and total DNA).In addition, the possible origin of these populations is discussed taking into account the background knowledge of evolutionary patterns and processes in Armeria, and specifically the compilospecies model, which fits another species, A. villosa Girard (Fuertes Aguilar & al., 1999b).

Material and methods
Sampling.In accordance with the strategy followed in other studies in Armeria (e.g., Gutiérrez Larena & al., 2002;Nieto Feliner & al., 2002), the neighbour territories have been sampled in order to minimize confounding molecular uniqueness with hybridization between related species.This is justified by the weak internal reproductive barriers in the genus (Nieto Feliner & al., 1996).Besides A. filicaulis, different populations of A. villosa subsp.longiaristata and A. villosa subsp.bernisii were sampled.The latter is sympatric to the populations of A. filicaulis in Alfacar-La Peza.Twenty individuals of the new subspecies from Sierra de Alfacar, representing 6 populations, have been used in the multivariate morphometric study.The geographic origin of those as well as the rest of the material used for the morphological and the molecular studies are shown in Table 1.
Molecular markers (ITS, trnL-F, RAPDs).Detailed results on the three molecular markers can be seen elsewhere.Protocols for DNA extraction, amplification, sequencing, data analysis, and specifically for ribotype (ITS repeat type) definition have been described in Fuertes Aguilar & Nieto Feliner (2003) and Nieto Feliner & al. (2004).Those for trnL-F in Gutiérrez Larena & al. (2002), The RAPD study is in Fuertes Aguilar & al. (unpubl.; Gutiérrez Larena, unpubl. PhD).Phenetic similarity from RAPD bands (Williams & al., 1990) was analyzed by Principal Coordinates Analysis and Cluster analysis applied to a similarity matrix constructed using the Dice coefficient.A minimum spanning tree based on the Euclidean distance was superimposed on the scatter plot of the samples against the first three PCO axes.Phenetic analyses of RAPD data were conducted using NTSYSpc (Rohlf, 2000).
Morphology.Thirty morphological characters have been used in the multivariate morphometric study (Table 2).Fifteen of them are quantitative continuous, seven are ratios, one is quantitative discrete (number of bracts) and seven are qualitative.For the continuous characters, values in the data matrix are averages of two or three measurements per specimen.Most of the metric characters were measured with the aid of a Brown & Sharpe Plus digital calliper (model 599-571-3).The morphometric study was conducted using NTSYSpc (Rohlf, 2000).Two different analyses have been carried out.A principal components analysis (PCA), based on the correlation matrix, was used as an ordination procedure to reduce the number of original variables to a few representative uncorrelated ones (the PCs).A minimum spanning tree based on the Euclidean distance was superimposed on the scatter plot of the samples against the first three PCA axes.An UPGMA cluster analysis was also conducted as a classificatory procedure, based on a similarity matrix constructed using Euclidean distance.All measurements have been made on dried specimens, which are kept in MA.Table 1).

Results
Morphological characters.The main differences between populations of A. filicaulis from Sierra de Alfacar and the remaining are the following.The scape is 20-40 cm long, distinctly larger than the rest of the species (Fig. 2).The tubular involucral sheath, 14-22 mm, is also longer than in other populations.The involucre diameter, 21-28 mm, is also the largest within A. filicaulis.Involucral bracts are smooth straw-colored (not yellowish-cream and minutely rugose), the inner ones are long (> 10 mm), narrower than in subsp.filicaulis, with a narrower scarious margin.Unlike the usual spikelet bracts in A. filicaulis subspp., those in Alfacar-La Peza are longer than (or at least equalling) the inner involucral bracts.Spikelets also differ in being more or less stipitate, non sessile.The pedicels of every individual flower within each spike are longer than the standards and almost lack bracteoles, resem-bling species in the section Macrocentron such as A. hirta Willd.etc. Calyces are also longer (14-22 mm) that in other subspecies, mainly due to the larger lobule and awn.
Morphometric multivariate analyses.In the PCA, the first three axes account for 62.2% of the total variance (35.9%, 18.2% and 8.0%, respectively).The first axis is contributed significantly by almost half of the variables but mostly by calyx length, calyx-lobe length, involucre diameter, length of outer involucral bracts and spikelet bract length while the second is by ratio of the outer to the longest inner involucral bracts, width of inner involucral bracts and ratio of the involucral diameter to the length of the involucral sheath.In the scatter plot of the samples against the first three axes, specimens of A. filicaulis from Alfacar-La Peza are distinctly separated from the rest although with respect to the first axis they are intermingled with A. villosa (Fig. 3).The superimposed mini- mum spanning tree connects one of the samples of Alfacar-La Peza with one of A.villosa subsp.bernisii from Sierra de Gádor.Samples of A. filicaulis subsp.nevadensis also cluster separately in the scatter plot, while those of subsp.filicaulis (including var.minor) and subsp.trevenqueana form a lax cluster.The UP-GMA phenogram provides a representation of the taxonomic groupings (Fig. 4).Two large clusters arise, one including all samples of A. villosa (subsp.bernisii and subsp.longiaristata) and the other including all samples of A. filicaulis.Within the A. filicaulis cluster, the samples from Alfacar-La Peza are the first ones to split off, being all included in their own sub-cluster.
trnL-F.The two samples of A. filicaulis from Alfacar have a chloroplast trnL-F haplotype A (Fig. 1).This is the most frequent in the species, occurring in more than 50% of its samples (Gutiérrez Larena & al., 2002).Of the sequences of A. villosa subsp.bernisii, the two sampled from Alfacar correspond to haplotype L while those two from La Peza do to haplotype I.These are the most frequent haplotypes in A. villosa subsp.bernisii, occurring in 35% and 41 of its samples, respectively (Gutiérrez Larena & al., 2002).
RAPDs.PCO based on RAPD phenotypes provides a picture that is not readily interpretable (Fig. 5).
In the scatter plot of the samples against the first three   principal coordinate axes, samples of A. filicaulis from Alfacar-La Peza do not cluster.The minimum spanning tree connects two of those samples with one of A. villosa subsp.bernisii (see discussion).An UPG-MA cluster based on the same Dice similarity matrix provides also a complex picture in which the samples of A. filicaulis from Alfacar-La Peza are among the first ones to split off from our data set that includes both A. villosa and A. filicaulis (Fig. 6).
Taxonomic proposal.Morphological distinctiveness confirmed also by the multivariate morphometric analyses, together with the restricted distribution and the consistent results from the molecular marker distribution (see discussion), supports taxonomic recognition for the populations of A. filicaulis from Alfacar-La Peza.
Additional specimens studied are listed in Table 1.

Discussion
Taxonomic integrity and distinctiveness of subsp.alfacarensis.The integrity and distinctiveness of A. filicaulis subsp.alfacarensis is mainly based on the morphological data.It is supported by the following facts.The PCA of morphometric characters cluster the specimens sampled as a distinct entity.This implies a cohesive combination of characters in all the representatives of this taxon.This pattern is paralleled in the UPGMA phenogram, where all the samples of subsp.alfacarensis form a single group that is not included within the A. filicaulis cluster.These facts, together with the peculiarity of the substrate, support taxonomic recognition of subsp.alfacarensis per se.However, inferring the possible origin of the taxon is more complex and molecular data should be used.

Recent Gene flow and co-occurring ITS repeat types.
The most likely explanation for co-occurrence of ITS ribotypes within the same genome in Armeria is gene flow, either intra-or interspecific (Nieto Feliner & al., 2004).This interpretation is particularly reliable for individuals fitting the geographical structure of ITS variation, that is, those lying in areas where two ribotypes meet.The region between Alfacar and La Peza, north of Sierra Nevada, is one of those contact areas, where ribotypes 2 and 3 meet (Nieto Feliner & al., 2004).There are 5 differing nucleotides between R2 and R3.Of these positions, one has been found to be homogenized in four of the five specimens sampled from Sierra de Alfacar that have co-occurring ribotypes.Since homogenization in ITS regions has proved to be fast in artificial hybrids of Armeria (Fuertes Aguilar & al., 1999a), to explain the maintenance of nucleotide polymorphisms (reflective of cooccurring ribotypes), either gene flow is frequent or bias in the homogenization has favoured persistence of copies of the two ribotypes within genomes.There is some evidence of bias for R1 but not for R2 or R3 (Fuertes Aguilar & al., 1999aAguilar & al., , 1999b;;Nieto Feliner & al., 2004).Therefore, it is more parsimonious to assume that gene flow affecting plants from near Alfacar, both A. filicaulis subsp.alfacarensis (3 samples, 2 co-occurring ribotypes) and A. villosa subsp.bernisii (3 samples, 3 co-occurring ribotypes), is recent.There is additional evidence for this besides the ITS.Interspecific gene flow is a frequent phenomenon in Armeria (Arrigoni, 1970;Philipp, 1974;Fuertes Aguilar & al., 1999b;Gutiérrez Larena & al., 2002) and plants of A. villosa subsp.bernisii are at most 3 km away (but probably much less) from those of A. filicaulis subsp.alfacarensis in Alfacar.Taking into account the unspecific pollinators (Woodell & Dale, 1993) and the low reproductive barriers in Armeria (Nieto Feliner & al., 1996) the possibility of recent or contemporary gene flow is high.RAPD data, specifically the minimum spanning tree, supports the suggestion that gene flow suffered by A. filicaulis subsp.alfacarensis in Alfacar and revealed by the ITS patterns, is from A. villosa subsp.bernisii, as suggested above.On this same line, based on the relative frequencies and distribution of R2 and R3 (Nieto Feliner & al., 2004), it is likely that in the inferred gene flow events between A. filicaulis subsp.alfacarensis and A. villosa subsp.bernisii near Alfacar, the former provided R2 and the latter did R3.
Origin of subsp.alfacarensis.But if recent gene flow provides a likely explanation for the co-occurring ribotypes, it appears to be insufficient to account for the morphological patterns detected and is unlikely to have been involved in the origin of A. filicaulis subsp.alfacarensis.Specifically, if standard undifferentiated populations of A. filicaulis subsp.filicaulis pre-existed in Alfacar-La Peza and only in recent times suffered gene flow from A. villosa subsp.bernisii, it would be reasonable to find standard forms of A. filicaulis subsp.filicaulis in the range too, not only subsp.alfacarensis.Also, morphological cohesiveness would be difficult to explain unless all the individuals were recent, close descendants of the individuals that suffered gene flow from A. villosa subsp.bernisii.This seems unlikely.Only specimens consistently matching the taxon here described are found in the crystalline dolomitic sands of Alfacar-La Peza (specimens cited in table 1 and personal observations).It can be therefore concluded that the origin of subsp.alfacarensis is probably not related to the occurrence of recent gene flow.In other words, there was some cohesive biological entity already differentiated prior to recent gene flow events.Further, from the eastern end of subsp.alfacarensis range, we lack sequence data, but we do have for A. villosa subsp.bernisii from this area.They reveal no co-occurring ribotypes but only the one from Sierra Nevada (R3), thus suggesting that recent gene flow events may have affected only plants from the eastern end (Alfacar).
¿What is then the origin of A. filicaulis subsp.alfacarensis?Two possibilities can be envisaged.Either populations from this area diverged from standard morphs of A. filicaulis subsp.filicaulis or the taxon was originated by hybridization and one of the progenitors was A. filicaulis subsp.filicaulis.Both possibilities match the scenario of pre-existing populations of A. filicaulis (similar to subsp.filicaulis or even subsp.trevenqueana, cf.Fuertes Aguilar & al., unpubl.)inhabiting the massif before the formation of subsp.alfacarensis.In fact, these kinds of xeric substrata exist in nearby areas of Eastern Andalusia (Cerro Trevenque in Sierra Nevada, c. 20 km south of Sierra de Alfacar; Sierra de las Guájaras, c. 50 km SE; Sierra de Almijara, c. 60 km SE).In all these spots, the predominant or exclusive species of Armeria found is A. filicaulis, the most xerophyitic one in Andalusia and probably the whole Iberian Peninsula.There are other spots with crystalline dolomitic sands in the province of Málaga, where another species of Armeria occur, A. malacitana Nieto Fel.Interestingly, this taxon has a very similar involucre to A. filicaulis subsp.alfacarensis although its leaves are wider; those in the new taxon are indistinguishable from A. filicaulis subsp.filicaulis.Furthermore, the basionym for A. malacitana is A. filicaulis var.longifolia, and it was suggested that both A. filicaulis and A. hirta were involved in its origin (Nieto Feliner, 1987).Also noteworthy is the spikelets structure in subsp.alfacarensis, with scarce bractlets, and the size and shape of the involucre, both characters resembling those in A. hirta.A detailed scenario for this link between A. malacitana and A. filicaulis subsp.alfacarensis cannot be provided, but one possibility is that in both dolomitic areas (Sierra de Alfacar in Granada, Sierra de Mijas, etc. in Málaga) there were forms affine to A. hirta that hybridized with A. filicaulis.The occurrence of endemics to some of these dolomitic areas (Anthyllis tejedensis Boiss., Convolvulus boissieri Steudel, Arenaria armerina Bory subsp.caesia (Boiss.)C. Díaz, C. Morales & F. Valle) indicate floristic links between those areas that are supported by the high floristic similarity values found between some of them (Mota & al., 1993(Mota & al., , 2002)).
Armeria filicaulis as a compilospecies.The possibility of a hybrid origin for A. filicaulis subsp.alfacarensis, the third of three subspecies described in the last years raises the possibility that A. filicaulis follows an evolutionary model -compilospecies (Harlan & De Wet, 1963)-such as the one proposed for A. villosa (Fuertes Aguilar & al., 1999b).As A. villosa, A. filicaulis has a relatively wide distribution from the Rif mountains to, at least, Sierra de Alcaraz in Albacete.In addition, samples of A. filicaulis also display different ITS repeat copies depending on the geographic origin, namely ribotype 1 (R1) in Morocco, R2 in Sierras de Tejeda or Almijara, R3 in Sierra Nevada, R4 in Cazorla-Segura, Calar del Mundo or Murcian ranges, and a mixture of two ribotypes in intermediate areas (Nieto Feliner & al., 2004).Therefore, the behaviour of the different samples from the species is representative of the geographic structure exhibited by the ITS variation at the genus level and thus reveals important amounts of gene flow (Nieto Feliner & al., 2001).The chloroplast trnL-F haplotypes exhibited by A. filicaulis seem to be influenced by the sympatric congeners (Gutiérrez Larena & al., 2002) and thus part of them are explained by gene flow, e.g., populations from Sierra de Tejeda-Almijara, or Sierra Nevada populations of A. filicaulis subsp.nevadensis.However, such an interpretation is not always straightforward.The narrow endemic A. filicaulis subsp.trevenqueana displays three different haplotypes in a very restricted area, which might be consistent with a gene flow scenario on itself.But one of those haplotypes is a singleton (Gutiérrez Larena & al. 2002) suggesting that subsp.trevenqueana has remained isolated for a long period.The morphological evidence is not as revealing of hybridization and introgression as in the case of A. villosa.If morphological features of subsp.nevadensis do reveal its likely hy-brid origin from A. splendens and A. villosa subsp.bernisii, a hybrid explanation for diagnostic features of subsp.trevenqueana and alfacarensis is unclear, as is for pink-flowered populations from Sierra Tejeda named var.minor.In summary, although not so much of the intraspecific diversification of A. filicaulis can be unequivocally attributed to introgressive hybridization as in A. villosa, traces of this phenomenon are solid and thus part of the morphological and ecological diversity shown by A. filicaulis does seem to have been acquired by this means.Therefore, the species fit to a greater or lesser extent the compilospecies model.

Fig. 3 .
Fig. 3. Principal components analysis of Armeria spp.from SE Spain based on 30 morphometric characters.Plot of 205 specimens in the space defined by the first three axes, with a superimposed minimum spanning tree (based on Euclidean distance).

Fig. 4 .
Fig. 4. UPGMA phenogram of 205 specimens of Armeria spp.from SE Spain based on Euclidean distance matrix computed from 30 morphometric characters from SE Spain

40
Fig. 5. Principal Coordinate analysis of RAPD bands of Armeria from SE Spain based on a Dice similarity matrix.Plot of 167 samples in the space defined by the first three principal coordinate axes with a superimposed minimum spanning tree (based on Euclidean distance).

Fig. 6 .
Fig. 6.UPGMA phenogram of 167 samples of Armeria from SE Spain based on a Dice similarity matrix constructed from RAPD bands.Samples of Armeria filicaulis subsp.alfacarensis marked with a solid circle.

Table 2 .
Morphometric characters used in the multivariate analyses