Revision of the Festuca alpina group (Festuca section Festuca, Poaceae) in Europe

Transcription

1 bs_bs_banner Botanical Journal of the Linnean Society, 2012, 170, With 8 figures Revision of the Festuca alpina group (Festuca section Festuca, Poaceae) in Europe BRUNO FOGGI 1 *, GILBERTO PAROLO 2, PETR ŠMARDA 3, ANDREA COPPI 1, LORENZO LASTRUCCI 1, DMITAR LAKUŠIĆ 4, RUTH EASTWOOD 5 and GRAZIANO ROSSI 2 1 Department of Evolutionary Biology Plant Biology, University of Florence, via La Pira, 4, I-50121, Florence, Italy 2 Department of Earth and Environmental Sciences, University of Pavia, via S. Epifanio, 14, I-27100, Pavia, Italy 3 Department of Botany and Zoology, Faculty of Science, Masaryk University, Kotlářká 2, CZ-61137, Brno, Czech Republic 4 Institute of Botany and Botanical Garden, Faculty of Biology, University of Belgrade, Takovska 43, Belgrade, Serbia 5 Millennium Seed Bank Partnership, Seed Conservation Department, Royal Botanic Gardens, Kew, Wakehurst Place, Ardingly, West Sussex RH17 6TN, UK Received 7 March 2012; revised 16 April 2012; accepted for publication 1 September 2012 The Festuca alpina group includes several small fescues growing in rocky habitats across the European mountain chains. A taxonomic study of F. alpina and F. alfrediana, the two most common species of this group, is presented here. Morphological data collected from 298 specimens across 37 populations from all the major European mountain ranges, genetic inter simple sequence repeat (ISSR) data, chromosome counts and DNA ploidy information were analysed. We found that the two species are differentiated by a combination of morphological and genetic characters, which are in line with the geographical distributions. Festuca alpina is distributed from the Pyrenees and Alps to the western Carpathians, whereas F. alfrediana is found in Sardinia, the Apennines and the Dinarids. Here, F. alfrediana is split into three subspecies on the basis of quantitative morphological characters and complete geographical segregation: F. alfrediana subsp. alfrediana in the Corso-Sardinian area; F. alfrediana subsp. ferrariniana in the Apennines; and F. alfrediana subsp. durmitorea in the Balkan Peninsula. The last two taxa are newly described here. Detailed description and information on identification, distribution, ecology, illustrations, synonyms and type material are provided The Linnean Society of London, Botanical Journal of the Linnean Society, 2012, 170, ADDITIONAL KEYWORDS: analysis of variance (ANOVA) chromosome count discriminant analysis (DA) genetic inter simple sequence repeat (ISSR) morphometrics multivariate analysis of variance (MANOVA) principal coordinate analysis (PCoA) taxonomy. INTRODUCTION Since Hackel s Monographia Festucarum Europaearum (Hackel, 1882), the systematics of the genus Festuca L. in Europe has advanced as a result of the efforts of many investigators. Among these, *Corresponding author. bruno.foggi@unifi.it Saint-Yves (1909, 1913, 1930) and many festucologists belonging to the Francophone biosystematic school, including De Litardière (1923), Bidault (1964, 1969), Auquier (1974, 1977) and Kerguélen (1975), have provided valuable regional information and methodological contributions. All of these endeavours were condensed by Markgraf-Dannenberg (1980) in her treatment of this genus for Flora Europaea. This monumental work exposed many taxonomic 618

2 FESTUCA ALPINA IN EUROPE 619 problems which generated a new drive in fescue research. Since Flora Europaea, the papers having the greatest impact on the current taxonomic treatment of the genus include: Festuca section Festuca by de la Fuente & Ortuñez (1998) and a new checklist (Cebolla Lozano & Rivas Ponce, 2003) for Spain; Kerguélen & Plonka (1989) and Portal (1999) providing the complete monographs of the genus in France; a revision of F. trichophylla (Gaudin) K.Rich. and F. violacea Gaudin in Europe (Al Bermani, Catalán & Stace, 1992; Foggi, Rossi & Signorini, 1999, respectively); Festuca section Eskia Willk. in the Alps and Italy (Wallossek, 1999; Foggi et al., 2007) and in Spain (de la Fuente, Ferrero & Ortuñez, 2001); F. pallens Host in central-eastern Europe (Šmarda & Kočí, 2003; Šmarda et al., 2007); a recent revision of the F. valesiaca group in the central Alps (Arndt, 2008); and Torrecilla et al. (2003) on sections Eskia, Pseudatropis Kriv., Amphigenes (Janka) Tzvel., Pseudoscariosae Kriv. and Scariosae Hack. Concise taxonomic information is, however, still lacking in many other critical groups, creating difficulties in the preparation of reliable local floras and identification keys. One of the most critical of these is the Festuca alpina group, which comprises several small species typically found in exposed (mostly calcareous) rocky habitats in the high mountains of Europe and characterized by tiller leafsheaths completely closed or open for their upper half. Festuca alpina Suter was described with a short diagnosis: panicula secunda angustissima; spiculis 3 4-floris brevissime aristatis; fol. capillaribus glabris; culmo subnudo (Suter, 1802), reporting a citation and a series of localities in southern Switzerland from Haller (1768). Hackel (1881) included this taxon in F. ovina L. [F. ovina subsp. alpina (Suter) Hackel]. In his later and more detailed treatment, Hackel (1882) indicated F. ovina subsp. alpina as a large and homogeneous taxon widespread in the Alps, with one isolated locality in the central Apennines. A more detailed treatment of F. ovina subsp. alpina was first presented by Litardière (1922, 1923), who recognized several new varieties and subvarieties, some of which have been subsequently raised to the species level. In the treatment of Festuca in Flora von Griechenland, Markgraf-Dannenberg (1976) divided F. alpina into two subspecies: (1) F. alpina subsp. alpina in the Alps and Pyrenees, and (2) F. alpina subsp. briquetii (St.-Yves ex Litard) Markgr.-Dann.in Greece, Apennines, Sardinia and Corsica. Later, Markgraf-Dannenberg (1980) and Markgraf- Dannenberg & Pignatti (1982) treated F. alpina subsp. briquetii at the species level as F. vizzavonae Ronn. (Ronniger, 1918). This decision was later disputed (Foggi & Signorini, 1998; Signorini, Foggi & Nardi, 2003a, b) and the name F. alfrediana Foggi & Signorini was applied to F. alpina subsp. briquetii at the species level (Foggi & Signorini, 1998). Festuca alfrediana was reported by Portal (1999) for Corsica and by Catonica (2001) for the central Apennines. In his doctoral thesis on Festuca in the Dinarid mountains, Lakušić (1999) recognized a new subspecies of F. alfrediana, F. alfrediana subsp. durmitorea, which has not yet been effectively published (Art of ICBN: McNeill et al., 2006). Kerguélen, Plonka & Chas (1993) published a new subspecies, F. alpina subsp. riverae É.Chas, Kerguélen & Plonka, from the western Alps. This taxon was subsequently also reported in the Pyrenees (de la Fuente & Ortuñez, 1998; Portal, 1999) and in the central Apennines (Catonica, 2001). Pils & Prosser (1995) described F. austrodolomitica Pils & F.Prosser, a clearly distinguishable species known from the snow-bed communities of the Dolomites. Recently, an isolated population belonging to F. alpina was reported from the western Carpathians in Slovakia (Vel ká Fatra Mountains) by Šmarda & Kočí (2005). Several taxa belonging to the F. alpina group were also described in the Balkan Peninsula, but information regarding their taxonomy and distribution is still scarce. Festuca olympica Vetter (Vetter, 1929; Strid, 1991) was described from the summit of Mt. Olympus (Greece: type W!) and is known only from this locality. Festuca pirinica I.Horvat ex Markgr.-Dann. was reported from Bulgaria (Velev, 1963; Strid, 1991). Festuca olympica and F. pirinica both have open leaf-sheaths (Markgraf-Dannenberg, 1981; the authors, pers. observ.), and therefore they belong to another group. Festuca micevskiana Kostadinovski (1999, 2005) needs further study in order to understand its position in relation to the other representatives of the alpina group. For these reasons, F. austrodolomitica, F. pirinica, F. olympica and F. micevskiana will not be taken into consideration here. The ploidy of all analysed F. alpina group specimens from the Pyrenees (de la Fuente & Ortuñez, 1998), the Alps (Bidault, 1968; Parreaux, 1972; Pils, 1982; Seal, 1983), the Appenines and the western Carpathians was diploid (Šmarda & Kočí, 2005; Šmarda, 2008; Šmarda et al., 2008). The only tetraploid (2n = 28) reported was F. alfrediana in Corsica (Portal, 1999), but recently Portal (in litt. July 2009) expressed doubts concerning this count. Here, we used morphometric, ploidy and genetic data to study populations referred to as F. alpina and/or F. alfrediana across the mountains of central and southern Europe, from the central Pyrenees to the western Carpathians, including the Apennines, Dinarids and the Corsican and Sardinian mountains, in order to clarify their taxonomy.

3 620 B. FOGGI ET AL. MATERIAL AND METHODS MATERIALS We studied 37 populations (4 14 individuals per population, a total of 298 plants) from the eastern Pyrenees to the western Carpathians through the Alps, along the Apennines and Dinarids and from Corsica and Sardinia (Table 1). Twenty-five populations were sampled by us (bold in Table 1) and five by our colleagues (shown in italics in Table 1). The remaining seven populations were based on groups of herbarium specimens collected in a single locality; we considered these groups as populations. Type material of F. alpina (P!), F. alpina var. briquetii (G!), F. durmitorea (BEOU!) and F. alpina subsp. riverae (P!) was studied, but was not included in the morphometric analysis. Twenty populations were employed in the genetic analysis using inter simple sequence repeats (ISSRs) (Table 1). Plants from several populations, including those from loci classici, were cultivated in pots in the Botanic Gardens of the University of Florence to observe variation in characters under homogeneous conditions. MORPHOMETRIC ANALYSIS Fourteen morphological characters, reproductive and vegetative, considered as diagnostic in recent floras and treatments/revisions (e.g. Markgraf-Dannenberg, 1980; Kerguélen & Plonka, 1989; Wilkinson & Stace, 1991; Conert, 1998; Foggi et al., 1999; Portal, 1999; Foggi et al., 2006), were taken into account (Table 2). The length of culms (CL) and length of the tiller leaf blade (FL) were not used in the statistical analysis (see also Herrera, 2001) because they were highly variable in cultivated populations. Characters measurable on transverse leaf blade sections (width and thickness of tiller leaf blade, number of vascular bundles and number of sclerenchyma strands) were not measured systematically and are mentioned only in the description of taxa. We adopted standard measurements and terminology (cf. Foggi et al., 1999), which complies with Hackel (1882), Saint-Yves (1913), Ellis (1976) and Wilkinson & Stace (1991) in all but one case. The exception concerns the measurement of spikelets with two flowers. Where this occurred, the spikelet length was calculated as the distance from the apex of the second flower plus double the distance between the apex of the first and the apex of the second flower (excluding awn). Measurements of each character were repeated three to five times in different parts of a plant to encompass individual variation. Averaged values for each examined individual were used in the analyses (Table 2). Floral characters were observed and measured under a Zeiss stereomicroscope (Stemi SR model) at 8 20 magnification. Observations of transverse sections of leaf blades were carried out under a Reichert microscope (Univar model) at magnification. Characters were tested for pairwise correlations (Spearman rank correlation). Low correlations were found (< 0.9, cf. Lihova, Kudoh & Marhold, 2010), and so all characters were utilized in the analyses. A matrix of 298 rows (specimens) by 12 columns (variables) was transformed to a distance matrix using Gower (1971) distances as a result of the presence of qualitative and quantitative characters. This matrix was analysed by principal coordinate analysis (PCoA). Cluster analysis [unweighted pair group method with arithmetic mean (UPGMA) method] was also performed on the same matrix. To assess the goodness of fit of the UPGMA cluster, Mantel s correlations test was calculated between the initial distance matrix and the cophenetic matrix. A multivariate analysis of variance (MANOVA) was used to compare classifications based on analyses of genetic and morphological data, and results were shown using a scatter plot of a canonical variate analysis (CVA). Finally, the taxonomic arrangement proposed was tested by stepwise discriminant analysis (DA). We used DA as an inferential method to test the hypothesis derived by the first explorative analysis (Henderson, 2006). As a test for equality of group centroids derived from DA, we used Willk s l and Pillai s Trace. A confusion matrix was built by means of a linear method and by a jackknife process to ensure a cross-validation. The variation of characters showing the strongest discrimination power among the groups recognized in the multivariate analyses was shown using box-plots. Multivariate and univariate analyses were performed with the following packages: SPSS 18.0 for Windows (SPSS, Chicago, IL, USA), PAST (Hammer, Harper & Ryan, 2001; R Development Core Team, 2007) and Statistica (StatSoft, Tulsa, OK, USA). CHROMOSOME COUNTS AND PLOIDY ESTIMATIONS Chromosomes were counted using root tips from 14 cultivated plants (Table 1) representing six populations. For each locality, more than three counts were made on individuals cultivated in different pots in the Botanic Gardens of the University of Florence. The method described by Fiorini, Quercioli & Foggi (2008) to count chromosomes was employed. The ploidy of ten specimens (Table 1) was measured by flow cytometry conducted on fresh and herbarium material, using the same instrument and following the same procedure as described in Šmarda & Stančík (2006) and Šmarda (2008). Young fresh leaves of Solanum lycopersicum L. Stupické polní rané were used as an internal standard. Three herbarium specimens of

4 FESTUCA ALPINA IN EUROPE 621 Table 1. Provenience of specimens tested in morphometric analysis. x = Populations used also in ISSR analysis. x = Chromosome counts. y = Ploidy level with flow cytometry. Group Area ISSR Chromosome count/ploidy level long. lat. Numb. indiv. Country Locality Altitude m herb. 2 WA ITA Moncenisio. Piemonte. G 4 CA x ITA Valle dei Vitelli, Bormio. Lombardia. PAV 5 CA x ITA Gruppo Presolana PAV 6 EA ITA Pala Munda (Pale di S.Martino). Trentino Alto Adige ROV 8 EA ITA Cadore. Veneto. FI 9 EA AUT Trins in Valle di Gschnitz. Central Tyrol FI 10 AP x ITA Bivacco Rosaro/Monte La Nuda (Fivizzano). Toscana FI 11 AP x x ITA Monte Corchia. Toscana FI 12 AP ITA Monte Focolone (Majella). Abruzzo FI 13 AP x ITA Corno Grande-Gran Sasso. Abruzzo BEO 14 SK y ITA Oliena Vette del Sitta e Bidda. Sardegna. FI 15 SK FRA Punta del Fornello. Corse G 16 AP ITA Monte Pollino. Calabria FI 17 DIN x MON Sljieme/Lomovi, Durmitor. Montenegro BEO 18 DIN x MON Kom Vasojevicki. Montenegro BEO 19 CAR x SLK Vel ka Fatra Mts., Blatnika, Western Carpathian. Slovakia. BRNO 20 EA y AUT Lienzer Dolomiten Mts. Austria. BRNO 21 EA AUT Karten, Hohe Tauern Mts., Goldberg. Austria BRNO 22 EA y AUT Niederoesterreich, Schneeberg Mts. Austria BRNO 23 CAR SLK Vel ka Fatra Mts., Blatnika, Western Carpathian. Slovakia. BRNO 24 WA ITA Valle Pesio. Piemonte. FI 25 AP x ITA Majella. Abruzzo ANC 26 SK x x FR M.Stello. Corse PAV 27 EA x y ITA Canazei. Trentino Alto Adige. PAV 28 AP x ITA Orrido di Botri. Toscana 1200 FI 29 CA x ITA Val Malenco. Lombardia. PAV 31 AP ITA Grondilice, Alpi Apuane. Toscana FI 32 WA y ITA Pyramide Calcaires Val Veny. Val d Aosta. PAV 33 WA x FR Isere. Queyras. FI 34 CA x CH Passo Gemmi. Valais PAV 35 WA x y FR Col Izoard a NE del passo. Haut Savoie PAV 36 WA x FR Col Izoard a NE del passo. Haut Savoie PAV 37 WA x y FR Col della Maddalena. Alpes Maritimes. PAV 40 WA x x FR Plateau de Bure. Superdevoluy PAV 41 PYR x ESP Piedraforca. Central Pyrenees. Huesca JACA 42 PYR ESP Vasa de la Mora. Central Pyrenees. Huesca JACA 43 CA x CH Passo Gemmi. Valais PAV AP xy ITA Via Vandelli, Alpi Apuane. Toscana. PAV CA y ITA Lago di Como. Lombardia. PAV CA xy ITA Stelvio. Lombardia. PAV PYR = Pyrenees; WA = Western Alps; CA = Central Alps; EA = Eastern Alps; CAR = Carpathian; AP = Apennines; DIN = Dinarids; SK = Sardo-Corse.

5 622 B. FOGGI ET AL. Table 2. Morphological characters used in the morphometric analysis CL Length of culms, cm (mean of five measurements) FL Length of tiller leaf blade, cm (mean of five measurements) PL Length of the longest panicle, cm SL Length of spikelets, mm (mean of five measurements) G1 Length of lower glumes, mm (mean of five measurements) G2 Length of upper glumes, mm (mean of five measurements) WG2 Half-width of upper glumes, mm (mean of three measurements) LL Length of lemma, mm (mean of five measurements) WL Half-width of lemma, mm (mean of three measurements) AL Length of awn of lemma, mm (mean of five measurements) AS Length of anthers, mm (mean of two measurements) LP Length of palea, mm (mean of three measurements) SK Degree of scabridity of branches of panicles (1, smooth; 2, slightly scabrid; 3, scabrid; 4, strongly scabrid) MP Pubescence of palea (1, glabrous; 2, slightly pubescent; 3, pubescent; 4, highly pubescent) F. alpina of known chromosome numbers were initially measured to calculate ploidy from the ratio of S. lycopersicum/festuca to extrapolate knowledge for application to unknown samples. ISSR ANALYSES ISSR analyses were conducted on 160 plants from 20 populations (eight individuals per population, see Table 1). Genomic DNA was extracted from mg of silica gel-dried leaf tissue, using a 2 cetyltrimethylammonium bromide (CTAB) protocol with minor modifications (Mengoni et al., 2006). The quality of extracted DNA was checked using agarose gel electrophoresis (0.6% w/v) in 1 TBE buffer [0.89 M tris(hydroxymethyl)aminomethane (Tris), 2 mm ethylenediaminetetraacetic acid (EDTA), 0.89 M boric acid, ph 8.3] with ethidium bromide (1 mg ml -1 ) staining. The quantity of DNA was estimated by means of spectrophotometric readings using a BioPhotometer (Eppendorf). Three of the six primers [ISSR1, (GT) 7-YR; ISSR4, (CAA-(GA) 5; ISSR8, (CA) 6-RG] that were preliminarily screened for polymerase chain reaction (PCR) amplification yielded clear and reproducible banding patterns. PCR was performed in a total volume of 20 ml using 10 ng of DNA, 2 ml of 10 reaction buffer (Dynazyme II, Finnzyme, Espoo, Finland), 1.5 mm MgCl 2, 200 mm deoxynucleoside triphosphates, 2 ml of 10 mm primer and 1.4 U Taq DNA polymerase (Dynazyme II). Thermocycling was carried out after an initial denaturing phase of 5 min at 94 C, followed by 35 cycles each of 40 s at 94 C, 45 s at 43 C and 90 s at 72 C. A final cycle was set for 45 s at 94 C and 45 s at 42 C, with a final extension step of 5 min at 72 C. Amplification products were separated by electrophoresis using 2% agarose gel containing 1 mg ml -1 ethidium bromide. Electrophoresis was carried out at 75 W for 2 h. The resulting bands were visualized using a UV transilluminator and analysed by ImageJ software v. 1.43f (Image Processing and Analysis in Java; National Institutes of Health, Bethesda, MD, USA, 2009). All amplified bands were treated as dominant genetic markers and all ISSR profiles obtained were translated into a rectangular binary matrix. These data were also used to calculate the Euclidean distance between populations, and a nonmetric multidimensional scaling (NMDS) was performed. Using the Arlequin package (Schneider, Rosselli & Excoffier, 2000), a matrix of genetic similarity between individual plant samples was computed using Jaccard s coefficient of similarity. This method takes into account band sharing between individuals and is commonly used for the analysis of dominant markers, such as amplified fragment length polymorphism (AFLP), restriction fragment length polymorphism (RFLP) and random amplification of polymorphic DNA (RAPD) (Rieseberg, 1996; Lowe, Harris & Ashton, 2004; Coppi, Mengoni & Selvi, 2008). Using Arlequin, standard Nei s measures (Nei, 1978) were estimated, starting with single locus heterozygosity, reported as H E = 1 - (p 2 + q 2 ), where p and q represent the frequency of presence and absence, respectively, of the observed band. The mean heterozygosity within populations (H S) and the total for each species in the region (H T) were also calculated. Using Arlequin, genetic distances were estimated by computing a matrix of linearized pairwise F ST values (Slatkin, 1995), which was then used to generate a neighborjoining (NJ) dendrogram (Saitou & Nei, 1987) employing the software MEGA 3.0 (Kumar, Tamura & Nei, 2004). Analysis of molecular variance (AMOVA; Excoffier, Smouse & Quattro, 1992), as implemented in Arlequin 2000, was used to analyse the partition of total genetic variation at three different hierarchical levels (within populations, between populations and between groups of populations) and taking into consideration several combinations of geographical groups of populations. The statistical support for different hypothetical groupings of populations was

6 FESTUCA ALPINA IN EUROPE 623 tested in terms of variance components and percentage of explained variation. RESULTS MORPHOMETRIC ANALYSIS Cluster analysis and PCoA, performed on 298 specimens from 38 populations, resulted in the delimitation of four morphologically and geographically separated groups. The first two axes of the PCoA (Fig. 1) separate the specimens from the Sardo-Corsican area, Apennines and Dinarids, positioned on the negative side of the x axis, from those from the northern parts of the study area, the Pyrenees, Alps and Carpathians. In the first group (negative side of the x axis), three subgroups are seen, which correspond to specific geographical areas: (1) Apennines; (2) Dinarids; and (3) the Sardo- Corsican area. In the second group, no separation is observed. The same pattern is also seen on the first and third axes, but these two groups are separated along the y axis. Furthermore, the separation of the populations from the Apennines, Dinarids and Sardo- Corsican area is more distinct. The separation between these four groups of populations can also be observed in the cluster (Fig. 2). In particular, populations from the Alps are grouped with those from the Pyrenees and Carpathian Mountains. In the other part of the cluster, populations from the Sardo- Corsican area form a group well separated from those from the Apennines. Populations from the Dinarids are also separated from those from the Apennines. Mantel s cophenitic correlation calculated for the UPGMA cluster method is 0.854, determining a medium to high fitness among the distance matrix and the graphical representation of the cluster. CHROMOSOME COUNTS AND PLOIDY LEVEL All the karyologically analysed populations were diploid (2n = 14). Diploidy was also confirmed in all dry samples measured by flow cytometry (see Table 1). Our analyses of several populations from Monte Stello (in the northern part of Corsica) confirm the diploid level for the F. alfrediana populations in Corsica. GENETIC ISSR The three selected ISSR primers (ISSR1, ISSR4 and ISSR8) yielded 73 reproducible bands ranging from 200 to 3000 bp in length; 100% of these were polymorphic in the 160 individual ISSR profiles obtained. Of these polymorphic bands, 7% were specific to a single population. ISSR marker numbers 17 and 18 of primer ISSR1 yielded specific bands, 1250 and 3000 bp long, for populations 37 (Col de la Maddalena) and 5 (Gruppo della Presolana), respectively. Marker 43 of primer ISSR4 was specific to population 37, and markers 65 and 69 of primer ISSR8 were found to be characteristic of the populations 35 (Col Izoard) and 10 (Bivacco Rosaro/M. La Nuda: northern Apennines), respectively. In Table 3, the essential information on the genetic diversity for each population is reported. Total heterozygosity (H T) for F. alpina s.l. was 0.307, compared with an average heterozygosity within populations (average H S) of Population heterozygosity values (H E) ranged from in population 5 (Presolana) to in population 10 from the northern Apennines. The percentage of polymorphic loci was lowest in population 5 from the central Alps (20.5%) and highest in population 17 (52.1%) from Sljieme/Lomovi, Durmitor (Montenegro). The NJ dendrogram (Fig. 3), based on Slatkin s linearized pairwise F ST matrix (see Supporting Information, Table S1), suggested the existence of four main clusters, here named A, B, C and D. Cluster A included all the Apennine populations (10, 13, 11, 28 and 25) and three Central Alps populations (4, 5 and 29). The second cluster (B) included five populations from different geographical origins (19 from the Carpathians, 27 from the eastern Alps, 33 from the western Alps, 34 from the central Alps and 41 from the Pyrenees). The two populations from the Dinarids (17 and 18) formed a third distinct cluster (C) close to cluster D. Cluster D contained one Corsican population (26) and four populations from the western Alps (35, 36, 37 and 40). AMOVA (Table 4) showed a relatively high level of genetic differentiation among populations (F ST = ). The greatest part of the total variance explained occurred among populations (57.73%); that resulting within populations was 42.27%. Five main groups were detected (Table 4), yielding the highest percentage of explained variance among groups (21.65%; P < ) and the lowest percentage of variance within groups (F SC = 0.486). These five groups were formed by populations from: (1) the western Alps and Pyreneees; (2) the Alps and Carpathians; (3) Corsica and Sardinia; (4) Apennines; and (5) Dinarids. These largely corresponded to the geographical origin and to the clusters observed in the NJ dendrogram (Fig. 3). NMDS, based on ISSR profiles (Fig. 4), showed the presence of two distinct groups of populations on the positive area of the y axis: Corsica Sardinia and Dinarids on the positive and negative parts of the x axis, respectively. Furthermore, groups of populations originating from the Alps included those from the Carpathians to Pyrenees and some from the Apennines. Populations originating from the Pyrenees were included among those from the western Alps. The Mantel test (Mantel,

7 624 B. FOGGI ET AL. Figure 1. Scattergram of the first three axes derived from the principal coordinate analysis (PCoA): +, western Alps;, Pyrenees;, central Alps; D, eastern Alps;, Carpathians; x, Apennines; o, Corso-Sardinian area; *, Dinarids.

8 FESTUCA ALPINA IN EUROPE 625 Figure 2. Cluster analysis. Unweighted pair group method with arithmetic average (UPGMA): +, western Alps;, Pyrenees;, central Alps; D, eastern Alps;, Carpathians; x, Apennines; o, Corso-Sardinian area; *, Dinarids.

9 626 B. FOGGI ET AL. Table 3. Genetic diversity data of Festuca alpina group populations Population % of polymorphic number H E H S H T loci Number of alleles per locus/loci (mean ± SD) ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Total ± Figure 3. Two-dimensional scattergram nonmetric multidimensional scaling (NMDS) based on the pairwise Euclidean distances between the 20 populations, calculated from the inter simple sequence repeat (ISSR) profiles: +, western Alps;, Pyrenees;, central Alps; D, eastern Alps;, Carpathians; x, Apennines; o, Corso-Sardinian area; *, Dinarids.

10 FESTUCA ALPINA IN EUROPE 627 Table 4. Analysis of molecular variance (AMOVA) (1023 permutations; P < ). The best partition is shown in italic d.f. Sum of squares Variance components %of variation F SC F ST F CT Overall Among groups Among populations Two groups: Alps + Pyrenees + Carpathians; Corsica + Apennines + Dinarids Among groups Among populations Four groups: Alps + Carpathians; W Alps + Pyrenees + Corsica; Apennines; Dinarids Among groups Among populations Four groups: Alps + Pyrenees + Carpathians; Corsica; Apennines; Dinarids Among groups Among populations Five groups: W Alps ± Pyrenees; CE Alps ± Carpathians; Corsica; Apennines; Dinarids Among groups Among populations Six groups: Alps; Pyrenees; Carpathians; Corsica; Apennines; Dinarids Among groups Among populations Total ), comparing the matrices based on genetic distance with those based on the Gower distance, calculated on morphological data and taking into consideration the 20 populations investigated with both methods (see Table 1), was not significant (P > 0.5). Figure 4. Neighbor-joining dendrogram based on Slatkin s linearized pairwise F ST subscript matrix. DISCUSSION Morphometric and chromosome data enabled the delimitation of four geographically separated groups: 1. Populations from the Pyrenees, Alps and Carpathians: 2n = 14 (Pils, 1982; Kerguélen et al., 1993; de la Fuente et al., 2001; Šmarda & Kočí, 2005; Šmarda et al., 2008; this work). 2. Populations from the Corso-Sardinian area: 2n = 14 from Sardinia and Corsica (this work). 3. Populations from the Apennines: 2n = 14 (Šmarda et al., 2008; this work). 4. Populations from the Dinarids. The results of the ISSR partially confirmed the morphometric analyses. AMOVA (Table 4) further dis-

11 628 B. FOGGI ET AL. Figure 5. Scattergram derived from canonical variate analysis (CVA): +, Alps Pyrenees Carpathians; D, Apennines;, Dinarids;, Corso-Sardinian area. sected the F. alpina s.s. group into two groups corresponding to western (Pyrenees and western Alps) and eastern (central and eastern Alps, Carpathians) populations. This situation seems to lead to the confirmation of the presence of a western taxon (F. alpina subsp. riverae) separated from the populations from the central and eastern Alps and Carpathians (F. alpina subsp. alpina). However, the justification for the splitting of F. alpina was not supported by the morphometric analysis (Figs 1, 2) or the NJ dendrogram, which showed a group containing populations from the Pyrenees, Carpathians and parts of the Alps (Fig. 3). However, the separation between populations from the Alps, Carpathians and Pyrenees and those from southern Europe was clearly evident in the morphological and genetic data. Both genetic and morphological data supported the split of the southern European populations into three differentiated subgroups: populations from the Corso-Sardinian area; the Apennines; and the Dinarids. CVA, based on morphological data, confirmed the hypothesis of the separation between four groups of populations: Alps, Pyrenees and Carpathians; Corso-Sardinian area; Apennines; and Dinarids (scatterplot in Fig. 5). The first two components in the CVA explained more than 86% of the variation (78.44% for the first, 8.45% the second). Vectors of the characters plotted in the CVA showed that this was according to the dimension of the characters. MANOVA analyses, connected to the CVA, showed that the differences between these groups were significant (post-hoc Hotelling s test P < , with permutations). The soundness of the four groups was also tested by means of DA. The first discriminant function was positively weighted by SL (length of spikelets, P < 0.001) and G2 (length of upper glume, P < 0.05). The second was positively weighted by MP (degree of scabridity of panicle branches, P < 0.001) and negatively by AL (length of awn, P < 0.001). The third dimension was negatively weighted by LL (length of lemmas, P < 0.001) and AS (length of anthers, P < 0.001). The confusion matrix, linear and jackknifed, confirmed the perfect separation of all four hypothesized groups with 100% discrimination success of all groups (Table 5). Willk s l ( ) and Pillai s Trace (1.904) values indicated a significant difference (P < ) between the four hypothesized groups. In Figure 6A E, the character variation among the four

12 FESTUCA ALPINA IN EUROPE 629 Table 5. Confusion matrix (jackknifed shown second) for the studied populations resulting from discriminant analysis (DA) for the four hypothesized taxa A posteriori assignment A priori taxon F. alpina F. alfrediana subsp. alfrediana F. alfrediana subsp. ferrariniana F. alfrediana subsp. durmitorea F. alpina 197 (100%) F. alpina 197 (100%) F. alfrediana subsp. alfrediana 0 19 (95.4%) 3 (4.6%) 0 F. alfrediana subsp. alfrediana 0 17 (92.3%) 5 (7.7%) 0 F. alfrediana subsp. ferrariniana 0 3 (4.6%) 62 (95.4%) 0 F. alfrediana subsp. ferrariniana 0 5 (7.7%) 60 (92.3%) 0 F. alfrediana subsp. durmitorea (7.1%) 14 (92.9%) F. alfrediana subsp. durmitorea (7.1%) 13 (92.9%) recognized taxa is shown using box-plots for the best five discriminating quantitative characters. These five characters showed limited variation under cultivation and are often used in diagnostic keys at the level of species and/or subspecies. The basic descriptive statistics for the measured characters in the four recognized groups are presented in Table 6. According to the systematic concept of species and subspecies (Runemark, 1961) adopted for this genus (Foggi et al., 1999), the clear discontinuity in the variation of characters and in their distribution led us to consider the populations from the Pyrenees, Alps and Carpathian region as a different species from those occurring in the Corso-Sardinian area and those from central and southern Europe. The first group should be referred to as F. alpina Suter, whereas F. alfrediana Foggi & M.A.Signorini represents the proper name at the species level for the second group. As already stated, populations from the southwestern Alps, referred to as F. alpina subsp. riverae (Portal, 1999), cannot be distinguished on morphological grounds from those distributed in the central and eastern Alps from Switzerland to the western Carpathians. This is in accordance with Garraud (2003; Conservatoire Botanique National Alpin, Gap, pers. comm.), J.-M. Tison (L Isle-d abeau, pers. comm.) and with our personal observations of plants in the wild and under cultivation; therefore, we have opted to formally separate these two subspecies, thus giving priority to morphological data, as suggested by Henderson (2005). Festuca alfrediana could be divided according to genetic and morphological data into three subgroups, which we interpret at the subspecific level. The first subgroup includes large plants, originating from the Corso-Sardinian area, with long culms and leaves and with spikelets and anthers longer than 8 mm and 1.8 mm, respectively. This morphological description corresponds closely with that of F. alpina subvar. briquetii given by Saint-Yves (Litardière, 1922), i.e.: spiculae virides... longae 8 mm et ultra... palea... longiuscule ciliolata..., antherae...2 mm longae. This subvariety was recently raised to the species level as F. alfrediana by Foggi & Signorini (1998). The second subgroup comprises populations from the Apennines and, until now, has not been formally named and described. Here, we assign this group as a new subspecies F. alfrediana subsp. ferrariniana subsp. nov. Populations from Montenegro have already been ineffectively described in a PhD work by Lakušić (1999) under the name F. alfrediana subsp. durmitorea nom. inval., and formal validation of the name is proposed here. This taxon perhaps has a wider distribution in the Balkan Peninsula and may also refer to the populations of F. alpina subsp. briquetii from Greece (Markgraf-Dannenberg, 1976; Strid, 1991). The situation in the southern parts of the Balkans needs further investigation. The specimina visa selecta are reported in Supporting Information (Appendix S1). We presume that the separation of F. alfrediana could be interpreted as a result of a schizogenesis process that is common to many taxa occurring in the mountains of southern Europe (Favarger & Contandriopoulos, 1961). According to this interpretation, and on the basis of their low morphological separation, correlated with the geographical disjunction, the subspecies level seems appropriate. IDENTIFICATION KEY TO TAXA 1a. Spikelets mm, upper glume mm; Pyrenees, Alps and western Carpathians: F. alpina.

13 630 B. FOGGI ET AL A LL 8,5 8,0 B SL 7, , , , ,5 5, , ,0 4,5 C F.alpina F. alfred. subsp. ferr. F. alfred. subsp. durm. AL F. alfred. subsp. alfred. 4,0 4,8 4,6 D F.alpina F. alfred. subsp. ferr. F. alfred. subsp. durm. G2 F. alfred. subsp. alfred. 4,0 4,4 4,2 3,5 4,0 3,0 3,8 2,5 3,6 2,0 3,4 3,2 1,5 3,0 1,0 2,2 2,0 E F.alpina F. alfred. subsp. ferr. F. alfred. subsp. durm. AS F. alfred. subsp. alfred. 2,8 F.alpina F. alfred. subsp. ferr. F. alfred. subsp. durm. F. alfred. subsp. alfred. 1,8 1,6 1,4 1,2 1,0 0,8 0,6 F.alpina F. alfred. subsp. ferr. F. alfred. subsp. durm. F. alfred. subsp. alfred. Figure 6. A E, Variation of the four best discriminating characters according to the discriminant analysis (DA) among the four recognized taxa: A, length of lemmas (LL); B, length of the spikelets (SL); C, length of lemmas awn (AL); D, length of upper glume (G2); E, length of anthers (AS). Box-plots show median (strike), 25% 75% quartile range (box), 10% 90% percentile range (whiskers) and extreme (, 1% 99%).

14 FESTUCA ALPINA IN EUROPE 631 Table 6. Basic statistics (value maximum, minimum, medium and standard deviation) for characters in the four recognized taxa CL FL PL SL G1 G2 WG2 LL WL AL AS LP MP SK F. alpina (197 specimens) Min Max Mean SE Variance SD Median th percentile th percentile F. alfrediana subsp. alfrediana (22 specimens) Min Max Mean SE Variance SD Median th percentile th percentile F. alfrediana subsp. ferrariniana (65 specimens) Min Max Mean SE Variance SD Median th percentile th percentile F. alfrediana subsp. durmitorea (14 specimens) Min Max Mean SE Variance SD Median th percentile th percentile SD, standard deviation; SE, standard error.

15 632 B. FOGGI ET AL. 1b. Spikelets mm, upper glume mm; Apennines, Corsica, Sardinia and mountains of Balkan Peninsula: 2. 2a. Spikelets mm and anthers mm; Sardinia and Corsica: F. alfrediana subsp. alfrediana. 2b. Spikelets mm and anthers mm: 3. 3a. Lemmas mm; Apennines: F. alfrediana subsp. ferrariniana. 3b. Lemmas mm; Dinarids (and Greece?): F. alfrediana subsp. durmitorea. Note that the characters in the identification key must be used in combination. The principal diagnostic characters are illustrated in Figure 7. Following Smith, Figueiredo & Moore (2011) and Knapp, McNeill & Turland (2011), we provide Latin and English diagnoses for the new names. TAXONOMIC TREATMENT Festuca alpina Suter Helvet. Fl., 1: 55 (1802) Type: An Festuca.... En. Helv. p. 216 (?) Gen. VIII n. 16 (?). A.... (?)/Botan. p.b. nemorosa (?) varietas in Gemmio reperta? (manu Haller senior), lectotype here designed in P-Haller! F. ovina subsp. alpina (Suter) Hack., Bot. Centralbl. 8: 406 (1881). F. ovina subsp. alpina var. suteri St.-Yves in Litardière, Bull. Soc. Sc. Hist. et Nat. Corse 42: 201 (1922) = F. alpina var. gaucheri St.-Yves in Litardière, Bull. Soc. Sc. Hist. et Nat. Corse 42: 201. = F. alpina var. gaucheri (St.-Yves) Kerguélen, Lejeunia n.s. 75: 151 (1975). = F. alpina subsp. riverae Chas, Kerguélen & Plonka, Lejeunia n.s. 142: 3 (1993). Notes on typification of F. alpina Suter: Festuca alpina was first recognized by Haller (1762: 51, 1768: 216), but without a Latin binomial. The first binomial was published by Suter (1802) who wrote that the plant was found as Frequens, in altioribus alpibus: Gemmio, Javernaz, Fouly, Enzeindaz, Richard, sur Champ.. The typification of the name F. alpina can be made from the original collection of Haller (senior) in P (Laujouw & Stafleau, 1956; Kerguélen & Plonka, 1989). Through our research in P, we found a folder with several specimens of F. alpina. As the lectotype, we have selected the only specimen bearing a label with Haller s handwriting (cf. Burdet, 1979). The chosen specimen corresponds to the description of Suter (1802). This specimen was collected on Mt. Gemmi (near Leukerbad, Switzerland), one of the localities reported in Haller (1762, 1768), and also cited in Suter (1802). Description: More or less densely tufted grass, sometimes with prostrate-ascending basal vegetative shoots. Vegetative tillers intravaginal. Culms 3 21 cm long, smooth and glabrous along entire length. Tiller leaf-sheaths papyraceous, clear, shiny, glabrous, fused for entire length, sometimes decaying into fibres. Ligule very short. Tiller leaf blades mm long, setaceous, generally straight, soft or slightly rigid, smooth. Transverse section of tiller leaf blades V-shaped or slightly closed polygonal, mm in diameter, thickness mm; with three to five vascular bundles; three adaxial ribs, with two lateral ribs generally less developed, with few hairs; three small sclerenchyma strands, sometimes with two others submarginal (the median strand is generally slightly larger than the others). Panicle cm long, dense, with short branches that are generally scabrid. Spikelets mm long, with (two )three four( five) flowers. Lower glume subulate mm long. Upper glume subulate mm long, mm wide. Lemma subulate mm long, mm wide; awn mm long. Palea mm long, glabrous or sparsely finely hairy on the carena. Anthers mm long. Distribution: France (Kerguélen & Plonka, 1989; Portal, 1999), Switzerland (Lauber & Wagner, 2001), Germany (Conert, 1998), Austria (Englmaier, 2005), Croatia (Markgraf, 1933; Nikolic, 2004 onwards; Mareković et al., 2009), Slovenia (Jogan, 1999), Slovakia (Šmarda & Kočí, 2005). In Italy, the species is reported in several localities in Friuli-Venezia Giulia (Poldini, 2002) and in Sud-Tirol (Wilhalm, Niklfeld & Gutermann, 2005). Populations from the Pyrenees reported as F. alpina subsp. riverae (de la Fuente & Ortuñez, 1998; Cebolla Lozano & Rivas Ponce, 2003; Foggi & Müller, 2009) should be referred to as F. alpina. The southeasternmost population of F. alpina subsp. riverae reported from Gran Sasso in central Italy (Catonica, 2001) should be referred to as F. alfrediana subsp. ferrariniana. A distribution of the analysed populations is given in Figure 8. Ecology and phytosociology: Festuca alpina prefers basophytic rocks, but it can occasionally be found on other substrates (Saint-Yves, 1913). It is a mesophytic species, preferring the northern shady sides of mountains in boreal and alpine belts from 1500 to 3000 m (Credaro & Pirola, 1975; Markgraf-Dannenberg, 1981; Mucina, 1993; Oberdorfer, 1994; Portal, 1999; Aeschimann et al., 2004). It prefers rock fissure plant communities (class Asplenietea trichomanis), composed of few and scattered cushion plants, covering less than 10% of the rocky walls (Oberdorfer, 1994). Oberdorfer (1977) and Oberdorfer et al. (2001) report F. alpina in the Androsacetum helveticae (Potentillion

16 FESTUCA ALPINA IN EUROPE 633 Figure 7. Key characteristics of the reproductive parts of the four taxa proposed here: A, Festuca alpina (Mt. Gemmi Switzerland: locus classicus); B, F. alfrediana subsp. alfrediana (Mt. Stello Corsica; locus classicus); C, F. alfrediana subsp. ferrariniana (Mt. Alto northern Apennines); D, F. alfrediana subsp. durmitorea (Mt. Durmitor Montenegro; locus classicus).

17 634 B. FOGGI ET AL. Figure 8. Distribution map of the recognized taxa based on the 37 studied populations (abbreviations of localities as in Table 1; +, Alps Pyrenees Carpathians; D, Apennines;, Dinarids;, Corso-Sardinian area). caulescentis) community. Sometimes it occurs in the vegetation of thin carbonate screes of Drabion hoppeane (English et al., 1993; Rivas-Martinez et al., 2002) or Noccaeion rotundifoliae (Julve, 1998). Festuca alfrediana Foggi & M.A.Signorini Parlatorea 2: 128 (1997) Synonyms: F. ovina var. briquetii St.-Yves in Litardière, Bull. Soc. Hist. Nat. Corse 42: 201 (1922). F. alpina subsp. briquetii (St.-Yves) Markgr.-Dann., Veröff. Geobot. Inst. Rübel (Zürich) 56: 134 (1976). F. alpina var. briquetii (St.-Yves) Gamisans, Candollea 29(1): 48 (1974). Type: Punta del Fornello. F. Halleri All. 12 juin rochers élevés. M. Stello, Corse. Lectotype in G! designated by Kerguélen & Plonka (1988). Description: More or less densely tufted grass, sometimes with prostrate-ascending basal vegetative shoots. Vegetative tillers intravaginal. Culms 5 30 cm long, smooth and glabrous for their entire length. Tiller leaf-sheaths papyraceous, clear, shiny, glabrous, completely fused, sometimes decaying into fibres. Ligule very short. Tiller leaf blades cm long, setaceous, generally straight, soft or slightly rigid, smooth. Transverse section of tiller leaf blades V-shaped or slightly closed polygonal, mm in diameter, thickness mm; with (three)five vascular bundles; three adaxial ribs of which two lateral ribs are generally only slightly developed, with few hairs; three narrow sclerenchyma strands, sometimes with two others submarginal (the median strand is generally slightly larger than the others). Panicle cm long, dense, with short branches that are generally scabrid. Spikelets mm long, with