Diversification of Cerastium sylvaticum and C. subtriflorum on the margin of the south-eastern Alps

The south-eastern Alps and the southerly adjacent areas are considered an important refugium for plant species during the Pleistocene glaciations; consequently, they harbour many endemic taxa. One of them is Cerastium subtriflorum, which in this area occurs sympatrically with morphologically similar, but more widespread C. sylvaticum. Here, we used amplified fragment length polymorphisms as well as ITS and plastid DNA sequences to explore phylogenetic relationships between the two species as well as their relationships to other Cerastium taxa previously deemed closely related. Our data show that both species are genetically well-differentiated sister taxa, but their relationship to other taxa remains unclear; both species rather form an independent evolutionary lineage within European Cerastium. Cerastium sylvaticum exhibits low genetic differentiation among populations in the investigated area, suggesting a single Pleistocene refugium in that part of Europe. The much more pronounced genetic differentiation within C. subtriflorum is likely a result of multiple Pleistocene refugia or more recent vicariance among the three disjunct partial distribution areas. Phylogenetic sequence data suggest hybridisation between the two taxa, resulting in a discordant phylogenetic position between ITS and plastid trees of one population of C. subtriflorum. Flow cytometric and multivariate morphometric analyses show that C. subtriflorum and C. sylvaticum differ in their genome size as well as in multiple morphological traits. We provide a taxonomic treatment for both species including an identification key and species descriptions.


Introduction
The south-eastern Alps and the southerly adjacent areas have been considered an important refugium for Alpine biota during Pleistocene glaciations (Tribsch and Schönswetter 2003;Schmitt 2009) as the southern margin of the Eastern Alps was not entirely covered with an ice sheet during the Last Glacial Maximum, thus enabling the survival of plant species (Tribsch and Schönswetter 2003;Schönswetter et al. 2005;Alvarez et al. 2009). As in other glacial refugia in the Alps (Pawłowski 1970;Tribsch and Schönswetter 2003), also the south-eastern Alps harbour many endemic species; more than 60 endemic plant taxa have been listed for this area (Mayer 1960a, b;Wraber 1990), for instance Aconitum angustifolium Bernh. ex Rchb. (Ranunculaceae), Athamanta turbith (L.) Brot. (Apiaceae), Campanula zoysii Wulfen (Campanulaceae), Cerastium julicum Schellm. (Caryophyllaceae), Gentiana froelichii Jan ex Rchb. (Gentianaceae) and Moehringia villosa (Wulfen) Fenzl (Caryophyllaceae). Not only alpine plants survived the glaciations in this part of Europe; there is also evidence for existence of broadleaved forests in the areas southerly adjacent to the southeastern Alps, which were suggested as likely refugia for oaks (Quercus sp.; Petit et al. 2002) and beech (Fagus sylvatica L.; Magri et al. 2006) as well as for several beech forest understory herbs (Willner et al. 2009 In order to disentangle relationships between the southeastern Alpine endemic C. subtriflorum and the more widespread C. sylvaticum and to infer their relationships to other taxa, we here used a combination of phylogenetic analyses, relative genome size (RGS) measurements and multivariate morphometrics. More specifically, we (1) used ITS and plastid DNA sequences as well as amplified fragment length polymorphisms (AFLPs) to explore the relationships between both species and to infer their position within Cerastium, (2) analysed whether the two species differ in their relative genome size despite having the same chromosome number and (3) searched for additional morphological characters to distinguish between both species in vegetative stage using a multivariate morphometric approach. Based on this integrative approach, we provide a taxonomic treatment for both species, including detailed descriptions and an identification key.

Plant material
Leaf material of one to five individuals per population from nine populations of C. subtriflorum and five of C. sylvaticum was collected and stored in silica gel for molecular and RGS analyses; the exception is population 5, where freshly collected herbarium material was used for RGS measurements and DNA sequencing. Altogether, ITS and the plastid DNA region ndhJ-trnT was sequenced for 14 populations, for which also RGS data were obtained, and 11 populations were included in AFLP analyses. Some additional populations of C. subtriflorum and C. sylvaticum were collected for morphometric analyses, for which also additional specimens from the herbaria LJU, WU, TSB and ZA were studied, resulting in a data set of 38 individuals of C. subtriflorum and 46 of C. sylvaticum. Voucher data of all studied populations, including the four out-group taxa sequenced, are presented in Online Resource 1. Distribution of all analysed in-group populations is shown in Fig. 1, where also the geographic position of all inspected herbarium specimens of C. subtriflorum from the herbaria IB, LJU, TSB, WU, W and ZA is shown.

Sequencing and phylogenetic analyses of ITS and ndhJ-trnT
Extraction of total genomic DNA was performed following the modified CTAB-protocol of Tel-Zur et al. (1999). Four regions (petN(ycf6)-psbM, trnL (UAG) -ndhF, ndhJ-trnT, trnQ-trnK) of the plastid genome have been tested for variation among different Cerastium species prior the study. The ndhJ-trnT region yielded the highest variability and was thus sequenced in this study. PCR and sequencing of ITS and the plastid ndhJ-trnT region were performed for one to two individuals per population (Table 1) as described by Kutnjak et al. (2014), with the exception that sequencing was carried out at Eurofins Genomics (Ebersberg, Germany). Contigs were assembled, edited and sequences were aligned using Geneious Pro 5.5.9 (Kearse et al. 2012). Base polymorphisms were coded using NC-IUPAC ambiguity codes. ITS and plastid alignments are in Online Resources 2 and 3, respectively. GenBank numbers of C. subtriflorum and C. sylvaticum sequences are presented in Table 1 and those of the out-group taxa in Online Resource 1.
Gaps (indels) in the plastid data set were coded as binary characters using SeqState version 1.25 (Müller 2005). Maximum parsimony (MP) and MP bootstrap (MPB) analyses of ITS and plastid data sets were performed using PAUP 4.0b10 (Swofford 2002 (Ronquist et al. 2012) applying the GTR (ITS) and F81 (plastid dataset) substitution models proposed by the Akaike information criterion implemented in MrAIC.pl 1.4 (Nylander 2004). In the case of the plastid analyses, the data were partitioned into a nucleotide data set and an indel data set; the latter was treated as morphological data according to the model of Lewis (2001). Values for all parameters, such as the shape of the gamma distribution, were estimated during the MrBayes analyses. The settings for the Metropolis-coupled Markov chain Monte Carlo process included four runs with four chains each (three heated ones using the default heating scheme), run simultaneously for 10,000,000 generations each, sampling trees every 1000th generation using default priors. The posterior probabilities (PP) of the phylogeny and its branches were determined from the combined set of trees, discarding the first 1001 trees of each run as burn-in.
In addition, a NeighbourNet was produced with ITS sequences of C. subtriflorum and C. sylvaticum using Split-sTree4 12.3 (Huson and Bryant 2006) and the plastid ingroup sequences were analysed using statistical parsimony as implemented in TCS (Clement et al. 2000) with the connection limit set to 95; gaps were treated as fifth character state. For this analysis, indels longer than 1 bp were reduced to single base-pair columns allowing those structural mutations to be counted as single base-pair mutations only.

AFLP analyses
The number of individuals per population used in AFLP analyses is presented in Table 1. The AFLP procedure followed Vos et al. (1995) with the modifications described in Schönswetter et al. (2009). In addition, 0.25 U of polymerase was used in the preselective and selective amplifications (0.4 U for the NED-labelled primer combination). The primer combinations for the selective PCR (fluorescent dye in brackets) were EcoRI (6-FAM)ACA/MseI-CAT, EcoRI (VIC)AAG/MseI-CTT, EcoRI (NED)AGC/MseI-CTG (6-FAM-labelled primers: Sigma-Aldrich; NED-and VIC-labelled primers: Applied Biosystems). Purification and visualisation of PCR products were done as described in Rebernig et al. (2010). Analysis of electropherograms and scoring of the AFLP fragments were performed as described in Caković et al. (2017), with the exception that the minimum intensity used in RawGeno was set to 75 relative A preliminary neighbour-joining (NJ) analysis based on a Nei-Li genetic distance matrix (Nei and Li 1979) was constructed and bootstrapped (2000 pseudo-replicates) using TREECON v.1.3b (van de Peer and Wachter 1997). Cerastium subtriflorum and C. sylvaticum formed a monophyletic clade (bootstrap support, BS, 100%) with unresolved relationships to other taxa; therefore, the data set including only these two species was further analysed, i.e. a NJ tree was generated as described above and a NeighbourNet of C. subtriflorum was produced with SplitsTree4 12.3 (Huson and Bryant 2006).

Relative genome size (RGS) estimation and statistical analyses
The RGS of silica gel-dried leaves of C. subtriflorum and C. sylvaticum was estimated using flow cytometry as described by Kutnjak et al. (2014) for one to five individuals per population (Table 1). Despite the fact that the coefficient of variation (CV) of the G0/G1 peak of the sample exceeded the 5% threshold in four populations, we included these data in further analyses as the peaks were clearly visible and the RGS values highly similar to the RGS values obtained in other populations. Statistical analyses of the RGS data were carried out with R-3.3.2 (R Development Core Team 2016) and the visualisation package 'ggplot2'. Mean RGS values for populations of both species were tested for normality using a Shapiro-Wilk test. Subsequently, RGS values were compared using Welch's two-sample t test.

Morphometric analyses
Material for morphometric analyses of most characters included 56 individuals, 25 of C. subtriflorum and 31 of C. sylvaticum (Online Resource 1). In some of these plants, fruit and seed characters were missing and were measured on 13 and 15 additional specimens of C. subtriflorum and C. sylvaticum, respectively, resulting in data set of 84 individuals (Online Resource 1). Altogether, 67 morphological characters were studied (Table 2), including 16 ratio characters calculated on the basis of measured characters. Certain characters were missing in some individuals and were thus replaced with the mean values calculated for the other studied populations of the same species.
Leaf characters with the exception of trichome characters were measured with the program imageJ (Schneider et al. 2012), stem and inflorescence characters were scored manually and the remaining characters were measured under the stereo microscope Olympus SZX9 equipped with an Olympus UC30 camera and SZX9 telephoto lens using the Olympus image analysis software analysis Pro. Petals and sepals were softened in hot water and spread on plastic foil. Analyses of morphometric data were performed with IBM SPSS Statistics (version 24). Capsule and seed characters were analysed separately, because they were partly measured on other individuals as the remaining characters (see above and Online Resource 1), resulting in a data set of 20 C. subtriflorum and 22 C. sylvaticum individuals. Due to the small number of C. subtriflorum individuals with developed stolon leaves (12 C. subtriflorum and 24 C. sylvaticum; Online Resource 1), we also analysed these characters separately.
Correlation among metric characters was tested employing Pearson or Spearman correlation coefficients, which exceed 0.90 in five pairs of characters: width of petal lobe/petal width, petal length/distance from the basis to the widest part of the petal, petal length/petal length from the basis to the incision, length of a mid-stem leaf/distance from the basis to the widest part of a mid-stem leaf, length of a stolon leaf/distance from the basis to the widest part of a lowermost leaf; second listed character was excluded from further analyses.
We produced boxplot diagrams for all characters in order to explore the variation between the two taxa. After standardisation to zero mean and unit variance, a principal component analysis (PCA) was performed. Discriminant analysis (DA) was applied to inspect the separation between C. subtriflorum and C. sylvaticum and the relative importance of characters discriminating between them. All these three analyses were performed for each of the three data sets separately (capsule and seeds, stolon leaves, other characters). Values presented in the identification key and the species descriptions correspond to the 10 and 90% percentiles, supplemented by extreme values in parentheses.

ITS and ndhJ-trnT sequences
Topologies of the trees inferred by parsimony and Bayesian analyses were congruent in both sequence data sets. In the ITS tree (Fig. 2a), C. dinaricum and C. eriophorum were most divergent, followed by C. fontanum. Cerastium carinthiacum and C. latifolium formed a highly supported clade (MPB 85%, PP 1), which was sister to the clade including C. subtriflorum and C. sylvaticum (MPB 100%, PP 1). Within the latter clade, relationships were poorly resolved; only C. sylvaticum was weakly supported in the parsimony Width of petal lobe, mm 4 Ratio petal length/petal width 5 Distance from the basis to the widest part of the petal, mm 6 Ratio distance from the basis to the widest part of the petal/petal length 7 Petal length from the basis to the incision, mm 8 Ratio petal length from basis to the incision/petal length Sepal 9 Sepal length, mm 10 Sepal width, mm 11 Ratio sepal width/sepal length 12 Distance from the basis to the widest part of the sepal, mm 13 Ratio distance from the basis to the widest part of the sepal/sepal length 14 Width of hyaline margin of sepal, mm 15 Ratio petal length/sepal length Leaf Mid-stem leaf 16 Length of a mid-stem leaf, mm 17 Width of a mid-stem leaf, mm 18 Ratio length of a mid-stem leaf/width of a mid-stem leaf 19 Distance from the basis to the widest part of a mid-stem leaves, mm 20 Ratio distance from the basis to the widest part of a mid-stem leaves/length of a mid-stem leaf 21 Angle of the apex of a mid-stem leaf, °2 2 Angle of the base of a mid-stem leaf, °L owermost leaf 23 Length of a lowermost leaf, mm 24 Width of a lowermost leaf, mm 25 Ratio length of a lowermost leaf/width of a lowermost leaf 26 Distance from the basis to the widest part of a lowermost leaf, mm 27 Ratio distance from the basis to the widest part of a lowermost leaf/length of a lowermost leaf 28 Angle of the apex of a lowermost leaf, °2 9 Angle of the base of a lowermost leaf, °3 0 Petiole length of a lowermost leaf, mm Stolon leaf 31 Length of a stolon leaf, mm 32 Width of a stolon leaf, mm 33 Ratio length of a stolon leaf/width of a stolon leaf 34 Distance from the basis to the widest part of a lowermost leaf, mm 35 Ratio distance from the basis to the widest part of a lowermost leaf/length of a stolon leaf 36 Angle of the apex of a stolon leaf, °3 7 Angle of the base of a stolon leaf, °3 8 Petiole length of a stolon leaf, mm Capsule 39 Capsule length, mm 40 Capsule width, mm 41 Ratio capsule length/capsule width 42 Distance from the basis to the widest part of the capsule, mm 43 Ratio distance from the basis to the widest part of the capsule/capsule length tree (MPB 71%, PP 0.76). In the NeighbourNet (Fig. 2b), C. sylvaticum and C. subtriflorum were divergent; populations 5, 7 and 8 of the latter from central Slovenia were intermediate between the two species albeit connected with a shorter split with other populations of the same species. Relationships between C. subtriflorum/C. sylvaticum and the out-group taxa were poorly resolved in the plastid tree (Fig. 2c). Cerastium dinaricum was sister to a clade (MPB 99%, PP 1) including the in-group taxa and C. fontanum, but with poor support (PP 0.97). Relationships between C. fontanum, C. subtriflorum and C. sylvaticum were unresolved, as the clade including most of the C. subtriflorum accessions and C. fontanum had low support (MPB 54%, PP 0.88). Population 6 of C. subtriflorum was positioned within C. sylvaticum with strong support (MPB 99%, PP 1). All other populations of C. subtriflorum formed another well-supported clade (MPB 90%, PP 1). The differences between C. subtriflorum and C. sylvaticum were solely based on indel characters, whereas the nucleotide sequences were identical in all accessions. In the statistical parsimony network (Fig. 2d), all C. sylvaticum sequences (including population 6 of C. subtriflorum) were identical, whereas the other accessions of C. subtriflorum were variable, resulting in eight different haplotypes.

AFLP data
For the 172 individuals of the data set including other Cerastium species, 639 fragments were scored. The error rate before the exclusion of unreproducible fragments was 3.4%. Before the neighbour-joining tree of C. subtriflorum and C. sylvaticum was inferred, 311 monomorphic bands or those present or absent in only one individual were excluded. Before the NeighbourNet of C. subtriflorum was constructed, 71 additional monomorphic bands have been excluded.
The neighbour-joining tree (Fig. 3a) showed a clear separation between C. subtriflorum and C. sylvaticum with Seed width, mm 50 Ratio seed length/seed width 51 Length of tubercles on the seed, mm 52 Width of tubercles on the seeds, mm Stem 53 Stem height, mm 54 Stem width on the basis, mm 55 Length of internode adjacent to the mid-stem leaves, mm 56 Number of internodes below dichasium Inflorescence 57 Length of peduncle of terminal flower, mm 58 Length of inflorescence from the basal flower to the top, mm Trichomes 59 Number of trichomes along 1 mm on the pedunculus of the terminal flower 60 Length of trichomes on the sepal (average of six measured trichomes), mm 61 Number of eglandular hairs per 1 mm 2 on the upper epidermis of a mid-stem leaf, calculated as the average of two squares with 1 mm 2 each 62 Number of glandular hairs on the margin of a mid-stem leaf along 1 mm just below the tip of the leaf 63 Number of eglandular hairs on the margin of a mid-stem leaf along 1 mm just below the tip of the leaf 64 Length of the longest trichome on the margin of a mid-stem leaf along 1 mm just below the tip of the leaf, mm 65 Number of glandular hairs on the stem along 1 mm just below a mid-stem leaf 66 Number of eglandular hairs on the stem along 1 mm just below a mid-stem leaf 67 Length of the longest trichome on the stem along 1 mm just below a mid-stem leaf, mm 1 3 high bootstrap support (BS 100). Within C. subtriflorum, population 6 from central Slovenia was sister to all other populations, with unresolved relationships among them; only populations 1 and 3 grouped together (BS 86%). The NeighbourNet of C. subtriflorum (Fig. 3b) was star-like. Population 6 shared some splits with populations 1, 3 and 4 from western Slovenia, whereas populations 7, 8 and 9, all from the Sava river valley, were positioned on the other side of the same split. In C. sylvaticum, two main groups were resolved in the neighbour-joining tree, one including population 12 (BS 98%) from Western Slovenia and the other included all remaining populations (BS 72%).

Morphological differentiation between Cerastium subtriflorum and C. sylvaticum
For all characters excluding capsule and seed characters, measured values and ratios are presented in Online Resource 4; for capsule and seed character they are given in Online Resource 5. The first three axes of the PCA of all characters excluding capsule, seed and stolon leaf characters explained 26.14, 13.04 and 7.76% of the total variation (Fig. 5a). A clear separation between C. subtriflorum (Fig. 3c) and C. sylvaticum (Fig. 3d) was obvious along the first axis, whereas there was no visible pattern along the second and third axes. The characters contributing most to the separation along the first axis, i.e. those having highest component scores (between 0.62 and 0.90), were petal and sepal characters (petal length, width of petal lobe, ratio petal length/sepal length), leaf characters (length of a mid-stem leaf, width of a mid-stem leaf, angle of the base of a mid-stem leaf, width of a lowermost leaf, petiole length of a lowermost leaf), stem height and trichome characters (number of eglandular hairs on the margin of a mid-stem leaf along 1 mm just below the tip of the leaf, length of the longest trichome on the margin of a mid-stem leaf along 1 mm just below the tip of the leaf and number of glandular hairs on the Fig. 2 Phylogenetic relationships inferred by ITS (a, b) and plastid ndhJ-trnT (c, d) sequences. Population identifiers correspond to Online Resource 1 and Fig. 1. a,   stem along 1 mm just below a mid-stem leaf). Also the DA (Wilks' Lambda = 0.033; χ 2 = 114.08; df = 41; significance 0; p < 0.0001; Fig. 5b) based on the same set of characters showed a clear differentiation between the two species with the same characters as in the PCA contributing most strongly to the separation. The highest scores (between 0.20 and 0.32) were reached by petiole length of a lowermost leaf, number of glandular hairs on the stem along 1 mm just below a mid-stem leaf, width of petal lobe, petal length, length of the longest trichome on the margin of a mid-stem leaf along 1 mm just below the tip of the leaf and ratio petal length/ sepal length, followed by the ratio distance from the basis to the widest part of the petal/petal length, width of hyaline margin of sepal and number of eglandular hairs on the stem along 1 mm just below a mid-stem leaf. The PCA of stolon leaf (SL) characters also showed a clear separation between C. subtriflorum and C. sylvaticum. The first three PCA axes explained 52.73, 22.79 and 12.25% of the total variation (Fig. 5c). The characters contributing most to the separation along the first axis, i.e. those having highest component scores (between 0.77 and 0.88), were width of a stolon leaf, ratio length of a stolon leaf/width of a stolon leaf and petiole length of a stolon leaf. The same variables had the highest scores (between 0.56 and 0.78) in the DA (Wilks' Lambda = 0.318; χ 2 = 34.949; df = 7; significance 0.00; p < 0.0001), even though there was some overlap between the scores of the discriminant functions of C. subtriflorum and C. sylvaticum (Fig. 5d).
The PCA and the DA of capsule and seed characters (Wilks' Lambda = 0.405; χ 2 = 29.838; df = 14; significance 0.08; p > 0.0001) did not show a separation between the two species (not shown).

Discussion
South-eastern Alpine endemic Cerastium subtriflorum and more widespread C. sylvaticum are sister species, which are genetically (Figs. 2, 3) and morphologically (Fig. 5) clearly divergent and differ significantly also in their genome size (Fig. 4). The relationships to other Cerastium taxa remain unclear, but they clearly do not belong to the C. alpinum L. (= C. lanuginosum; tetraploid C. eriophorum from this group was included in the analyses) and C. latifolium alliances (Fig. 2) as suggested previously (Reichenbach 1841;Graebner and Correns 1918;Niketić 2007). In the ITS tree C. fontanum was positioned among the out-group taxa, whereas in the plastid tree it was included in the clade with C. subtriflorum and C. sylvaticum. However, in a phylogenetic tree based on a denser taxonomic and population sampling (Frajman, unpublished results) accessions of the C. fontanum alliance appeared more distant. Cerastium subtriflorum and C. sylvaticum thus do not belong to C. sect. Caespitosa as suggested by Božo Friedrich (1969), but rather form an independent evolutionary lineage within European tetraploid Cerastium.
Whereas C. sylvaticum is a typical forest understory species, C. subtriflorum inhabits exposed cliffs and rocks in the forests, but also rocky habitats above the timberline in the Julian Alps. Whereas the distribution of C. sylvaticum in the central part of its distribution area is more or less continuous, that of C. subtriflorum is split into three main disjunct areas, in the Julian Alps, in the Kamnik area and in the Sava valley (populations 1-2, 5-6 and 7-9, respectively, Fig. 1). The continuous distribution likely enables gene flow in C. sylvaticum, resulting in low genetic divergence in ITS sequences and presence of only one haplotype within the central part of the species' distribution. It is also likely that C. sylvaticum persisted Pleistocene glaciations in a single refugium in this area. Magri et al. (2006) suggested that beech (Fagus sylvatica) forests spread to Central and Northern Europe from their refugium in the northernmost Balkan Peninsula and it is plausible that also C. sylvaticum followed the same route, along with other understory species such as Cyclamen purpurascens (Slovak et al. 2012), Veronica chamaedrys  and Knautia drymeia (Rešetnik et al. 2016). To get a better understanding of the evolutionary history of C. sylvaticum, additional populations from the entire distribution area would need to be sampled.
Despite the fact that the area of C. subtriflorum sampled for genetic analyses is comparable to that of C. sylvaticum (Fig. 1), there is much higher diversity of ribotypes and haplotypes in this species (Fig. 2). Whether this difference between C. sylvaticum and C. subtriflorum results from the presently continuous vs. fragmented distribution ranges or rather from survival in one vs. several glacial refugia remains unclear. As populations of C. subtriflorum are small and fragmented, stochasticity of genetic drift (faster fixation or elimination of mutations) is likely more pronounced (Freeland et al. 2011;Masel 2011). On the other hand, multiple Pleistocene refugia have been suggested for several species in the western Balkan Peninsula (e.g. Surina et al. 2011) and the present three disjunct distribution areas might have been a result of vicariance in the Pleistocene. Whereas the sequence data are inconclusive regarding the relationships among the three population groups, the AFLP NeighbourNet shows that population 6 is nested within populations 1-4 from the Julian Alps, whereas the Sava valley populations (7-9) are separated by several shared splits (Fig. 3b). High genetic divergence of population 8 from the Sava valley indicated by a long split from the other populations as well as 100% BS support in the AFLP NJ tree (Fig. 3a, b) is difficult to explain, as the species is rather continuously distributed in this area of the Sava river valley. Some specimens collected from this locality deposited at the herbarium W have intermediate characters between C. subtriflorum and C. sylvaticum, and it might be that ancient hybridisation caused the genetic peculiarity of this population, which was probably maintained due to limited gene flow with other surrounding populations.
Both studied species often grow in close vicinity (authors' personal observation) in the areas where their distribution ranges overlap (Fig. 1). Hybridisation between Cerastium species was suggested to be relatively common (Buschmann 1938;Khalaf and Stace 2000;Niketić 2007) and is evidenced also by our data. Whereas there is no strong indication of hybridisation in the ITS and AFLP data, in the plastid tree (Fig. 2) population 6 of C. subtriflorum from central Slovenia is positioned within the C. sylvaticum clade. Such a discordant pattern between the nuclear and the plastid phylogeny might result from chloroplast capture (Rieseberg and Soltis 1991;Tsitrone et al. 2003;Frajman et al. 2009) via hybridisation, that is, recent introgression of C. sylvaticum into population 6 of C. subtriflorum. Also in the AFLP NJ tree this population is the most divergent, which might also be caused by hybridisation with C. sylvaticum. There are also some morphological features, which might indicate a hybrid origin of this population. The petal length is close to the minimal length for C. subtriflorum (C. sylvaticum has shorter petals than C. subtriflorum), the lowermost leaves on a lateral shoot are slightly petiolate (which is characteristic for C. sylvaticum) and glandular hairs on a leaf margin are absent; the latter is characteristic for C. sylvaticum, but present also in some C. subtriflorum individuals from other populations.

Taxonomic treatment
In agreement with previous studies (Peverina 1982;Jalas 1993;Vreš 1996Vreš , 2007, the main morphological characters distinguishing the two species are petal size and the shape of lowermost stem and stolon leaves, which are petiolate in C. sylvaticum and sessile in C. subtriflorum. Moreover, C. subtriflorum individuals growing in lowlands have mainly glandular hairs on the stem, but those growing above the timberline usually lack them, having only eglandular ones, which renders them similar to C. sylvaticum. Some additional characters with discriminatory value between the two species are shown in Online Resource 6 and are included in the identification key and the species descriptions below. In cases when both length and width are listed, they always appear in this order.

Identification key
Although trait values of some characters overlap, a combination of characters allows straightforward discrimination between the two taxa. The most discriminating characters are bold.

Information on Electronic Supplementary Material
Online Resource 1. Studied populations of Cerastium subtriflorum and C. sylvaticum and four out-group taxa. The distribution of populations used in phylogenetic, relative genome size and morphometric analyses and all inspected populations of C. subtriflorum (from herbaria IB, LJU, TSB, W, WU and ZA) is shown in Fig. 1. Online Resource 2. ITS alignment. Identifiers following the species names are GenBank numbers. Online Resource 3. Plastid alignment. Identifiers following the species names are GenBank numbers. Online Resource 4. Character states of Cerastium subtriflorum and C. sylvaticum from morphometric analyses of all characters excluding capsule and seed characters. Online Resource 5. Character states of Cerastium subtriflorum and C. sylvaticum from morphometric analyses of capsule and seed characters. Online Resource 6. Boxplot diagrams of morphological characters discriminating between Cerastium subtriflorum and C. sylvaticum.