Long neglected diversity in the Accursed Mountains of northern Albania: Cerastium hekuravense is genetically and morphologically divergent from C. dinaricum

The Balkan Peninsula is a hotspot of European biotic diversity. One of its biogeographically most peculiar but poorly explored regions are the Albanian Alps (Alpet Shqiptare/Prokletije/Accursed Mountains) on the border between Albania, Kosovo and Montenegro, characterised by a high number of endemic species. A poorly known taxon from the Albanian Alps is Cerastium hekuravense, which was described from Mt. Maja Hekurave (Albania) in 1921, but later usually merged with C. dinaricum, a widespread endemic of the Dinaric Mountains, or connected with the arctic–alpine C. alpinum. Here, we used amplified fragment length polymorphisms to explore the phylogenetic position of C. hekuravense and particularly its relationship to C. dinaricum. Our data show that both species are genetically well differentiated, but their relation to other taxa remains unclear—they are either closely related to Alpine species of C. ser. Latifolia or to species co-occurring on the Balkan Peninsula, such as C. banaticum and C. decalvans. In addition, multivariate morphometric analyses show that C. dinaricum and C. hekuravense are morphologically well differentiated. Also their relative genome sizes, estimated using flow cytometry, differ. We propose a taxonomic treatment with lectotype designation for both taxa and provide descriptions and an identification key. Last but not least, these cold-adapted species mostly growing on northerly exposed humid screes are highly threatened due to the global warming and should be ranked endangered according to IUCN criteria. Cerastium hekuravense known only from three localities is likely one of the most endangered mountain plant species of the Balkan Peninsula.


Introduction
The Balkan Peninsula is a centre of plant species diversity and endemism; it is the floristically richest area in Europe harbouring about 6500 species, of which more than onethird (c. 2600-2700) are endemic and about 400 are considered to be local endemics (Horvat et al. 1974;Kryštufek and Reed 2004;Stevanović et al. 2007). These numbers are likely an underestimate, as only for Greece, including Crete and the islands, 5752 species, of which 1278 are endemic to this country, have been reported (Dimopoulos et al. 2013). In the western Balkan Peninsula, especially the high mountain system of the Dinarides (Dinaric Mountains) is highly diverse (Horvat et al. 1974;Redžić 2011). It spans from Slovenia in the North to northern Albania in the South, where the highest peak, Maja Jezercë (2694 m), is situated. This southernmost part of the Dinarides, which is known as Albanian Alps (Alpet Shqiptare/Bjeshkët e Namuna in Albanian) or Accursed Mountains (Prokletije in the southern Slavic languages), includes ca. 40 peaks over 2000 m and 17 peaks over 2500 m (Rakaj 2009).
Here, we used amplified fragment length polymorphisms (AFLPs), relative genome size (RGS) measurements and morphometric analyses to elucidate the relationship between C. hekuravense from the Albanian Alps and the more widespread C. dinaricum. Specifically, we address the following questions: (1) Is C. hekuravense from the locus classicus (type locality) genetically differentiated from C. dinaricum?
(2) If so, do the other known and recently discovered populations from the Albanian Alps cluster with C. dinaricum or with C. hekuravense? (3) Do eventually detected genetic groups differ in RGS and are they also morphologically differentiated? (4) Based on our results, we provide a taxonomic treatment of both taxa and evaluate their conservation status following the IUCN criteria.

Plant material
Molecular analyses are based on silica gel-dried leaf material. We included four populations of Cerastium dinaricum previously analysed by Kutnjak et al. (2014), including the population from the type locality, and additionally sampled four populations from Albania tentatively ascribed to C. dinaricum s.l., including the population from the type locality of C. hekuravense (Table 1, Fig. 1). Additionally, we included all other species (Online Resource 1) classified within the C. latifolium aggregate (Niketić 2007)-C. carinthiacum (two populations), C. latifolium (two populations) and C. uniflorum Clairv. (one population), as well as the species distributed on the Balkan Peninsula, which include tetraploid populations (Niketić et al. 2013

AFLP analyses
Extraction of total genomic DNA was performed following the modified CTAB-protocol of Tel-Zur et al. (1999). AFLP fingerprinting was conducted as described by Kutnjak et al. (2014). Two blanks (DNA replaced by water) were included to test for contamination, and four samples were used as replicates between the two PCR batches to test the reproducibility. Numbers of sampled individuals per population are provided in Table 1.
Electropherograms were analysed with Peak Scanner version 1.0 (Applied Biosystems) using default peak detection parameters except employing light peak smoothing. The minimum fluorescent threshold was set to 100  Kutnjak et al. 2014). The inset in the upper right corner shows the position of the sampling area in Europe; the rectangle in the main part of the figure indicates the position of the Albanian Alps magnified in the inset in the lower left corner relative fluorescence units. Automated binning and scoring of the AFLP fragments were performed using Raw-Geno 2.0-1 (Arrigo et al. 2009) for R 2.15.2 (R Development Core Team 2012) with the following settings: scoring range = 150-500 bp, minimum intensity = 100 relative fluorescence units (rfu), minimum bin width = 1 bp and maximum bin width = 1.5 bp. Fragments with a reproducibility lower than 85% based on sample-replicate comparisons were eliminated. Fragments present/absent in only one individual were excluded. A Neighbour-joining (NJ) analysis based on Nei-Li genetic distances (Nei and Li 1979) was conducted and bootstrapped (2000 pseudo-replicates) with TREECON v.1.3b (van de Peer and De Wachter 1997). The tree was rooted with C. grandiflorum based on an ITS phylogeny (Frajman B., unpublished). Due to the simple structure in the data, no further analyses were conducted.

Genome size measurements
Flow cytometry (FCM) of 40,6-diamidino-2-phenylindole (DAPI)-stained nuclei was used to estimate relative genome size (RGS) of four newly sampled populations (all populations from Albania) as described by Kutnjak et al. (2014). The RGS was estimated for three to ten individuals per population.
Absolute genome size (AGS) was determined using FCM of propidium iodide (PI)-stained nuclei of two samples of C. dinaricum and one sample of C. hekuravense (Table 1) as described by Frajman et al. (2015), with the exception that Pisum sativum cv. Kleine Rheinländerin (2C = 8.84 pg; Greilhuber and Ebert 1994) was used as reference standard.

Morphometric analyses
Material for morphometric analyses included vouchers of all molecularly investigated populations of C. dinaricum s.l. (Table 1), supplemented with herbarium vouchers stored in the herbarium of the University of Innsbruck, IB (Frajman and Schönswetter 14631, 14632, 14633), totalling 54 individuals. Forty characters were measured or counted and 15 ratios were calculated (Table 2). Leaf characters were measured on the uppermost and one welldeveloped mid-stem leaf; the apex angle was measured only on the mid-stem leaf. Trichome characters were measured on the upper surface and the margin of a midstem leaf, as well as on the internode below the investigated leaf. Certain characters were missing in a few individuals, e.g. petals or fully developed fruits, and were thus replaced with mean values calculated for the other studied populations of the same species. Petal, sepal, bract, leaf and fruit characters were measured on images taken with a camera mounted on a Zeiss SteREO Discovery. V12 stereo microscope at 8 × magnification. Characters of trichomes were measured on magnified images taken with an Olympus UC 30 wide zoom camera mounted on an Olympus SZX9 stereo microscope with 20 × magnification.
We tested correlation among metric characters employing Pearson or Spearman correlation coefficients dependent on character distribution. After standardization to zero mean and one unit variance, principal component analysis (PCA) was performed. As Tukey HSD Post hoc test showed no discriminatory power (p values between 0.33 and 0.98) for twelve characters (1, 2, 11, 12, 18, 19, 24, 30, 48, 49, 53 and 55) and three ratios (33, 36 and 39) we excluded them from the canonical discriminant analysis (CDA), which was applied to inspect the separation between C. dinaricum and C. hekuravense and the relative importance of characters as discriminators between them. Statistical analyses were performed using the package Statistica 5.1 (StatSoft 1996). Values presented in the species descriptions and in the identification key correspond to the 10 and 90% quantiles, supplemented by extreme values in parentheses.

AFLP analyses
We scored 335 AFLP fragments for 96 individuals; 42 bands found in only one individual were excluded. For the 36 individuals of Cerastium dinaricum s.l. we scored 196 fragments, of which 43 found in only one individual were excluded from further analyses.
The neighbour-joining tree of AFLP profiles (Fig. 2) resulted in a strongly supported cluster (bootstrap support, BS 100) containing all species except C. grandiflorum, which was used for rooting. Within this cluster, three moderately to well-supported groups were resolved; the relationships among them were unresolved. One group included all accessions of C. dinaricum s.l. with strong support (BS 87), falling in two strongly supported clusters (both with BS 100). One of them contained accessions of C. dinaricum studied by Kutnjak et al. (2014) including the population from the locus classicus, as well as one newly sampled population from the summit of Maja Jezercë in Albania. The other cluster contained the population from Maja Hekurave, locus classicus of C. hekuravense, as well as the populations from Buni Jezercë and Maja Kakisë. The two remaining major groups contained species of C. ser. Latifolia from the Alps (C. carinthiacum, C. latifolium, C. uniflorum; BS 91) and Balkan accessions (BS 67) of species belonging to C. ser. Alpina (C. decalvans and C. eriophorum) and to C. ser. Cerastium (C. banaticum).  Fig. 3). The AGS of C. dinaricum ranged from 2.5674 pg (population 4 from the locus classicus) to 2.5713 pg (population 2), whereas the AGS of C. hekuravense (population 7 from the locus classicus) was 3.0311 pg.

Morphological differentiation between Cerastium dinaricum and C. hekuravense
The measured and counted values for the characters, as well as the calculated ratios, are presented in Online Resource 2. As the correlation coefficients did not exceed 0.9 for any character pair, we included all characters listed in Table 2 in further analyses. The PCA (first three axes explaining 15.78, 10.98 and 8.06% of the total variation; Fig. 4a) showed a clear separation between C. dinaricum and C. hekuravense along the first two axes. The characters contributing most to the separation along the first axis, i.e. those having highest component scores in the coefficient matrix, describe habit (SH, SH/IN, IL), shape of the sepals (CLL/CLW), as well as shape of the middle leaves and bracts (MLL/MLW, BL/BW). Leaf, bract and trichome characters (ULLMW, MLLMW, MLLMW/ MLL, BLMW, BLMW/BL, GHLM, EHLM) contributed most to the separation along the second axis. The CDA (Fig. 4b) showed a clear differentiation between the two species with the strongest contribution of the sepal characters (CLL/CLW, CLLMW, CLW, CLLMW/CL). Still, even for characters with high discriminatory power (Tukey HSD post hoc test, P < 0.001) boxplots (Fig. 5) show overlap in character states between both taxa.

Discussion
Cerastium hekuravense is a distinct species, which is genetically and morphologically clearly differentiated from C. dinaricum (Figs. 2, 4, 6, 7). The two species also have different genome sizes (Fig. 3) and even where they occur in relatively close vicinity, such as on the summit of Maja Jezercë (C. dinaricum) and close to Buni Jezercë (C. hekuravense) no traces of gene flow have been detected in the AFLP data. Despite the fact that Niketić (1999Niketić ( , 2007 treated C. hekuravense as a variety or even as a form of C. dinaricum, which he suggested to be scattered throughout the distribution range of the latter, our data clearly show that C. hekuravense is endemic to the Albanian Alps, thus additionally underlining the importance of this mountain range as an endemic-rich area (Rakaj 2009; see also Introduction).
In spite of the peculiarity of its rich flora, Albania is one of the botanically least explored regions of Europe (Markgraf 1932;Frajman et al. 2014). After the last editions of the national flora (Paparisto et al. 1988;Qosja et al. 1992Qosja et al. , 1996Vangjeli et al. 2000) including 3758 taxa (3250 species) of vascular plants, a multitude of species new for Albania have been published (e.g. Barina and Pifkó 2008;Rakaj 2009;Ball 2011;Meyer 2011;Barina et al. 2013;Frajman et al. 2014). Even if C. hekuravense was described from the territory of Albania (Jávorka 1921), it was neglected in all recent Albanian floras, where only C. dinaricum was listed (Demiri 1983;Paparisto et al. 1988;Vangjeli 2003). In addition to clarifying the status of the known populations from Maja Hekurave and Maja Kakisë (the latter was published as C. dinaricum by Hayek 1924), we discovered an additional population of C. hekuravense close to Buni Jezercë (an indication of its occurrence there was kindly provided by M. 10.10.2013, personal comm. with B. Frajman) as well as the only known population of C. dinaricum in Albania, on the western summit crest of Maja Jezercë.
Spatially explicit modelling of viable habitat for C. dinaricum suggested a decrease of about 37% by the year 2050 and 70% by the year 2080 due to global warming, and the Number of glandular hairs on the margin of mid-stem leaves along 1 mm just below the tip of the leaf GHLM 51 Number of eglandular hairs on the margin of mid-stem leaves along 1 mm just below the tip of the leaf EHLM 52 Length of the longest trichome on the margin of mid-stem leaves along 1 mm just below the tip of the leaf TLM 53 Number of glandular hairs on the stem along 1 mm just below a mid-stem leaf pair GHS 54 Number of eglandular hairs on the stem along 1 mm just below a mid-stem leaf pair EHS 55 Length of the longest trichome on the stem along 1 mm just below a mid-stem leaf pair TLS Cerastium hekuravense is divergent from C. dinaricum 1 3 Fig. 2 Neighbour-joining tree derived from AFLP data. Population identifiers correspond to Table 1, Fig. 1 and Online Resource 1 predicted habitat loss could result in range-wide extinction of the species in the very near future (Kutnjak et al. 2014). Thus, following the criterion B2 of the IUCN (2012) for endangered species, the following applies for C. dinaricum: (a) area of occupancy is estimated to be less than 500 km 2 , (b) area of occupancy is severely fragmented and (c) continuing decline in the area, extent and/or quality of habitat has been inferred/projected. Therefore, we deem C. dinaricum endangered (EN) according to IUCN (2012).
The situation is similar, but likely more severe for C. hekuravense, which has a much narrower distribution and is currently known from only three localities, where it inhabits northerly exposed humid screes with extended snow cover. The ecology of C. hekuravense is thus similar to that of C. dinaricum, which, however, thrives also in rock crevices in the summit areas of some mountains (e.g. Kom Kučki in Montenegro and Maja Jezercë in Albania) and thus has a broader ecological niche. Field observations further suggest that C. hekuravense prefers more humid and colder screes than C. dinaricum (P. Schönswetter and B. Frajman, personal observations). In all three localities the species was rare, the smallest population being that on Maja Kakisë, where only a few dozen individuals were found. Although we do not have climatic niche modelling data for C. hekuravense at hand, extrapolation of the results obtained for C. dinaricum suggests that the species should be treated at least as endangered (EN).

Conservation status: Endangered (EN).
Note: In the protologue also a collection from Malovan (Velebit) is given, but the specimen has not been seen.  Distribution: Endemic to the Albanian Alps (Alpet Shqiptare), where it is only known from three localities-Buni Jezercë, Maja Hekurave and Maja Kakisë.

Key to the species of Cerastium dinaricum s.l
Even if there is a strong overlap in character states between both species, it is possible to discriminate between them using a combination of characters given in the key. The most discriminating character, although not always distinct on herbarium specimens, is bold. Both species and their habitats are shown in Figs. 6 and 7.