Morphometric and molecular analysis of Armeria canescens aggr. (Plumbaginaceae) in the Italian Peninsula

Abstract

The genus Armeria (Plumbaginaceae), which includes perennial herbs and cushion-forming dwarf shrubs, is primarily distributed in the temperate region of the northern hemisphere, with a small representation in southern Chile, Argentina and the Falkland Islands. In a few words we can say that Armeria is a genus of diploid, self-incompatible plants with weak internal reproductive barriers, in which hybridization and introgression play a fundamental role. According to the most recent data (Conti et al. 2005), 16 species and 18 taxa (species and subspecies), with a high endemism rate, occur in Italy. One of the most critical groups is distributed in the Apennines, where at least two taxa have been involved in taxonomic problems over the last century: A. canescens (Host) Ebel and A. majellensis Boiss. The aim of this research is to investigate, by means of a multidisciplinary approach, these two species. The first one is a central-eastern Mediterranean species described from the mountains of Dalmatia, while the second one is an Italian endemic species, but its real taxonomic status is still uncertain and, according to some authors, this taxon must be included in A. canescens because of their poor morphological separation. With the aim of clarifying the real status of A. majellensis and its relationship with A. canescens, we performed extensive field collections of material referable to these two taxa, throughout their Italian distribution area, trying to cover the different habitat types in which they usually live (e.g., different soil types, altitude, climate, vegetation types). Moreover, we collected materials in the loci classici of species (Dalmatia and Majella). The gathered materials were subjected both to morphometric and molecular investigations. The original descriptions of the species and the type materials referring to our taxa were preliminary studied and the lectotypes were designated. We studied the soil and the major climate aspects of the 13 collection localities. After drying up, the soil was subjected to different analyses: pH measure, determination of carbonates and of exchangeable bases (Ca, Mg, K, Na), analysis of organic carbon and organic matter, total nitrogen and granulometric analysis. All these analyses were performed in the Soil Science Laboratory, Sapienza Università di Roma. The obtained data highlight a great variability of all the investigated parameters that may be related to the remarkable phenotypic plasticity displayed by the taxa under consideration. For all the Italian populations we have also characterized the collection sites with a phytoclimatic approach. All the 12 populations fall within the Temperate Regions, with a bioclimate Oceanic or Oceanic/Semicontinental. With the aim of exploring the phenetic affinities between the studied taxa, a morphometric multivariate approach has been followed. 21 morphological quantitative characters were measured in 148 dried specimens: 130 belonging to our 13 studied populations and 18 are herbarium specimens on loan of Italian strictly related species. The so obtained matrix (148 specimens x 21 characters) was submitted to multivariate analyses (Principal Component Analysis and Cluster Analysis) in order to detect specimen groups on the base of the measured characters. Regarding PCA, the scatterplot of the 148 specimens, analysed against the first three principal components, doesn’t show discrete groups, revealing the absence of a distinct morphological differentiation within the A. canescens/A. majellensis aggregate across the whole investigated geographical area. The only detectable exception regards the samples of the two endemic species of Tuscany (A. denticulata and A. saviana), which seem to be separated from the others in the scatterplot. These latter species are clearly distinct from A. canescens/A. majellensis by two characters that are the relative length of calix tube and limb and the outer involucral bracts longer than the inner ones. On the same morphological matrix an UPGMA cluster analysis was also conducted. The obtained phenogram provided an overall picture similar with that arisen from the PCA: individuals of the same population are dispersed in different groups, without an agreement between cluster composition and population origin. This feature reveals that, within the studied populations, it's not possible to recognize morphologically distinct entities. For the molecular study, were used dried leaves coming from our 13 populations and from herbarium specimens of Italian strictly related species and of A. canescens from the Balkan Peninsula and Greece. The patterns of molecular variation were explored using four molecular markers: the nuclear ITS and three non-coding plastidial regions (trnL-F, trnS-fM, matK). All the molecular analyses were performed in the Molecular Systematic Laboratory, Real Jardín Botánico, Madrid. The ITS matrix included 94 sequences and 616 bp. All the specimens coming from the Italian peninsula and Sicily shared a single ITS sequence. In contrast to this ribotype, all the specimens coming from the Balkan Peninsula and Greece, and the specimen of A. alpina from the Alps, shared four nucleotides; the specimens from Greece shared an additional nucleotide. These ITS sequences were also reanalysed within the framework of a previous ITS phylogenetic study, representing about 70% of the Armeria species and covering the whole geographical range of the genus, obtaining a final matrix of 226 sequences x 606 bp. The main feature, as regarding our research topic, is that all the specimens from Peninsular Italy and Sicily grouped together in a basal clade, sister to the remaining terminals of Armeria. The Balkan (Croatian and Greek) samples of A. canescens fall in another clade, the most extended geographically. As expected, based on the kind of variation found in our sampling, the rest of the tree topology is fully congruent with that of the previous phylogeny. The three chosen plastid regions were first analysed individually and then together in the same network (matrix of 72 sequences x 2680 bp). The variation in the concatenated sequences allowed distinguishing 29 different haplotypes; among those, two are the most frequent ones, called A and B. On the basis of its interior position in the haplotype network, the high number of connections within it and the high frequency in the populations, haplotype A can be considered an ancestral one in accordance with coalescent theory. The second one, B (present in the three populations on the Tyrrhenian side of the Lazio region) is several mutational steps away from haplotype A and is closer to the Croatian haplotypes and to the ones of A. saviana. Our 72 sequences were then combined with 78 sequences of the same three plastid regions obtained from a previous work, involving some other species of Armeria occurring in the Iberian Peninsula. The so obtained matrix (150 sequences and 2665 bp) leads us to recognize 52 different haplotypes. None of them are shared between the Italian populations and the Spanish and Portuguese ones. Moreover, the relations between the Italian haplotypes are mostly maintained, although some uncertainties are resolved and the Balkan samples are grouped with the ones of A. maritima and A. pubigera (both in the A. maritima clade), which is congruent with the ITS variation scheme. Furthermore, haplotype A maintains an interior position, high frequency and high number of connections, and can be considered, also in this new network, an ancestral type. In conclusion, to answer our main question, we can say that A. majellensis is indistinguishable from A. canescens, for both morphological and molecular characters (exclusivity of the Italian haplotypes and the ITS ribotype), at least in Italy. For the time being, the priority name, at the specific level, is A. canescens, since the name of Ebel dates back to 1840, while Boissier coined the name A. majellensis only in 1848. As to the relationships between Balkan and Italian plants, and pending a study based on a wider Balkan sampling, the molecular data suggest differences. Therefore, a suitable taxonomic treatment would be to treat the Balkan populations as a different subspecies (subsp. canescens) while the Italians are treated as subsp. majellensis.