Disentangling relationships among eastern Mediterranean Cymbalaria including description of a novel species from the southern Peloponnese (Greece)

The eastern Mediterranean basin hosts a remarkably high plant diversity. Historical connections between currently isolated areas across the Aegean region and long-distance dispersal events have been invoked to explain current distribution patterns of species. According to most recent treatments, at least two Cymbalaria species occur in this area, Cymbalaria microcalyx and C. longipes. The former comprises several intraspecific taxa, treated at different ranks by different authors based on morphological data, evidencing the need of a taxonomic revision. Additionally, some populations of C. microcalyx show exclusive morphological characters that do not match any described taxon. Here, we aim to shed light on the systematics of eastern Mediterranean Cymbalaria and to propose a classification informed by various sources of evidence. We performed molecular phylogenetic analyses using ITS, 3’ETS, ndhF and rpl32-trnL sequences and estimated the ploidy level of some taxa performing relative genome size measures. Molecular data combined with morphology support the division of traditionally delimited C. microcalyx into C. acutiloba, C. microcalyx and C. minor, corresponding to well-delimited nrDNA lineages. Furthermore, we propose to combine C. microcalyx subsp. paradoxa at the species level. A group of specimens previously thought to belong to Cymbalaria microcalyx constitute a well-defined phylogenetic and morphological entity and are described here as a new species, Cymbalaria spetae. Cymbalaria longipes is non-monophyletic, but characterized by being glabrous and diploid, unlike other eastern species. The nrDNA data suggest at least two dispersals from the mainland to the Aegean Islands, potentially facilitated by marine regressions.


Introduction
The eastern Mediterranean basin comprises two major centers of biodiversity, the Balkan-Aegean area with 6500-7000 species, of which ca. 1500 are endemic, and the Handling Editor: Stanislav Španiel.
Contribution to "Plants of the Balkan Peninsula in Space and Time"

Page 2 of 19
Mediterranean part of the Anatolian Peninsula with ca. 5000 species, of which 30% are endemic (Thompson 2005). This richness is attributable to both moderate Pleistocene glaciation (Fady-Welterlen 2005;Wu et al. 2007) and a complex geography, with several mountain ranges and many islands (Polunin 1980;Thompson 2005). Furthermore, several species-rich Mediterranean lineages (e.g., Campanula L.: Roquet et al. 2009, Centaurea L. subgenus Centaurea: Hilpold et al. 2014Haplophyllum A.Juss.: Manafzadeh et al. 2014) have significantly diversified there. Finally, a higher degree of single-island endemism compared to the western Mediterranean contributes to the high diversity and singularity of the eastern Mediterranean (Thompson 2005).
Cymbalaria Hill (Plantaginaceae) originated ca. 4 Ma in the central-eastern Mediterranean basin and rapidly diverged in three lineages (Carnicero et al. 2017), which further diverged in ten species and seven subspecies according to the most recent publications (Sutton 1988;Bigazzi and Raffaelli 2000;Carnicero et al. 2018Carnicero et al. , 2019. It is a genus of perennial herbs distributed throughout the Mediterranean basin, although most taxa are endemics of small areas. Cymbalaria species grow in rocky habitats in a wide range of ecological conditions, from coastal cliffs to rock crevasses in the subalpine belt. Ploidy levels vary across species ranging from diploids (2n = 14) to octoploids (2n = 56). Diploids mainly occur in the Apennine and Balkan Peninsulas, with one species in the eastern Mediterranean; tetraploids (2n = 28) occur in Sicily, the Balkan Peninsula and the eastern Mediterranean basin, and a group of hexa-to octoploids (2n = 42, 56) occurs in Corsica, Sardinia and the Balearic Islands. Cymbalaria is a genus with intricate taxonomy in the eastern Mediterranean, where two out of the three main evolutionary lineages of the genus occur. Following Sutton (1988), two species occur in this area, the glabrous diploid C. longipes (Boiss. and Heldr.) A.Chev. and the hairy tetraploid C. microcalyx (Boiss.) Wettst. (Table 1). Cymbalaria longipes occurs in coastal areas from the Ionian Islands in the west to the Lebanon in the east (Fig. 1). There is little dispute about its taxonomic status due to some distinctive features, including 1) the characteristic habitat in cliffs and pebbles close to the sea; 2) glabrous vegetative organs; 3) big seeds that form a concrescent mass in an indehiscent capsule; and 4) diploidy (2n = 14; Sutton 1988).
On the other hand, C. microcalyx is an aggregate of taxa with an intricate history of taxonomic combinations and description of new entities, all having in common the presence of hairy vegetative organs (e.g., Cufodontis 1936;Greuter 1979;Speta 1986). Cymbalaria microcalyx inhabits cliffs from the sea level up to ca. 2000 m (Maroulis and Georgiadis 2005) from the southern Balkan Peninsula to western Anatolia. Sutton (1988) recognized five subspecies, but also mentioned other entities, which could deserve taxonomic recognition. Speta (1986) instead took a more analytical perspective and recognized four species within the group (Table 1, Fig. 1); his work is taken here as a reference, with a single exception (Table 1), as it seems to better agree with available phylogenetic data (Carnicero et al. 2017). Indeed, Carnicero et al. (2017) showed that C. microcalyx sensu Sutton (1988) is non-monophyletic and its subspecies are found in three distinct clades, but the scarce sampling among the eastern species prevented further detailed taxonomic conclusions in that study. The westernmost species of the aggregate is C. ebelii (Cufod.) Speta, a narrow endemic of the surroundings of Skadar Lake at the border between Albania and Montenegro, distinguishable by unique oblong-ovoid seeds with a few scarcely anastomosed longitudinal ridges. Cymbalaria microcalyx s. str. (i.e., subsp. microcalyx) occurs in the Peloponnese and is characterized by unequally sized seeds, all with prominent wing-like ridges. Cymbalaria microcalyx subsp. heterosepala (Cufod.) Speta is endemic to western Crete, distinguishable from subsp. microcalyx because seeds form a concrescent mass in the capsule. Cymbalaria minor (Cufod.) Speta occurs in the Peloponnese and surrounding islands; it is characterized by small, equal-sized, alveolate seeds. A further taxon, C. acutiloba (Boiss. & Heldr.) Speta, comprises two allopatric subspecies Table 1 Two most relevant published taxonomic treatments for eastern Mediterranean Cymbalaria. Speta (1986) is taken here as a reference, with the exception of C. microcalyx subsp. paradoxa, not included in Speta (1986). Asterisks indicate taxa which were not formally accepted by Sutton (1988) but were mentioned as potentially good taxa with distinct seed morphology. Cymbalaria acutiloba subsp. acutiloba occurs in the mountains of southwestern Anatolia and bears unequally sized seeds with prominent wing-like ridges (Speta 1986;Sutton 1988). Cymbalaria acutiloba subsp. dodekanesi (Greuter) Speta is distributed throughout numerous Aegean Islands including the Cyclades, the eastern Aegean Islands, eastern Crete, Karpathos and Rhodes. Its seeds are cristate to alveolate, not-winged and heterogeneous in size.
Finally, C. microcalyx subsp. paradoxa Greuter, a narrow endemic from Kastellorizo island, is densely hairy and has a similar seed morphology as C. longipes (i.e., big seeds that form a concrescent mass in an indehiscent capsule). Sutton (1988) did not study it in depth, and this taxon was not included in Speta (1986), but it was listed as accepted taxon in the last checklist of the Flora of Greece (Dimopoulos et al. 2013), similarly as C. microcalyx subsp. alba (Voliotis) Kit Tan. The latter is a Peloponnese endemic that allegedly differs from C. microcalyx subsp. microcalyx by its white flowers, longer spur and smaller calix (Voliotis 1990, Tan andIatrou 2001). With respect to ploidy level, the available data suggest that C. microcalyx subsp. microcalyx, C. ebelii, C. acutiloba subsp. dodekanesi, and C. minor are tetraploid (Speta 1986(Speta , 1989Sutton 1988;P. Carnicero unpubl. data). The role of polyploid speciation in the genus has been previously highlighted (Carnicero et al. 2017), but it remains uncertain how is it involved in the divergence of the taxa of this particular group.
During our studies on the genus Cymbalaria, two specimens kept in the herbaria SALA and MA and identified as C. microcalyx caught our attention. Both were collected from the same locality in the mountain range Taigetos (southern Peloponnese) and differ in several characters from any taxon described in the region, i.e., the higher number of leaf lobes, big flowers with a short and wide spur, and small cristate-alveolate seeds. During field work we found a second population showing the same characters. The morphological features suggest that this entity deserves taxonomic recognition, and it is expected that molecular and ploidy level data will help to determine its evolutionary history and taxonomic rank.
The high number of infraspecific taxa described within eastern Mediterranean Cymbalaria species, the incongruence in taxonomic treatments between authors and the existence of populations non-attributable to any described taxon, strongly emphasizes the need of a comprehensive systematic study combining molecular, morphological and ploidy data to clarify the taxonomy of the group. Moreover, the current distribution of eastern Mediterranean Cymbalaria species, which is strongly fragmented and spans major biogeographic barriers such as Rechinger's line (Rechinger 1943(Rechinger , 1950Strid 1996), suggests that this group is well suited to improve our understanding of plant evolution and biogeography in the eastern Mediterranean.
Using a combination of plastid (cpDNA) and nuclear ribosomal (nrDNA) sequences, morphological data and relative genome size, we aim to clarify the systematics of eastern  Sutton (1988), local Floras, personal field observations and herbarium vouchers. Names of most taxa follow Speta (1986) (Table 1), except for C. paradoxa and C. spetae (see "Taxonomic treatment") Mediterranean Cymbalaria species. Specifically, we aim to test whether the populations from Taigetos, previously considered C. microcalyx, belong to a new taxon and to determine its phylogenetic affinities. Finally, we aim to provide first insights into the group's spatial diversification in order to contribute new data to Aegean plant biogeography.

Plant material
We sampled at least one specimen of all described species of Cymbalaria, with the aim of building a complete phylogenetic backbone to infer the phylogenetic position of the eastern Mediterranean taxa. The latter were more intensively sampled with the aim of studying their phylogenetic relationships and to capture as much geographic and genetic variation as possible. In total, 65 specimens were used in molecular analyses, of which 28 specimens belonging to eastern Mediterranean taxa were sequenced for the present study. The relative genome size was determined for seven specimens, and 62 were used for morphological studies. Sampling localities, voucher details, as well as specimens used in each analysis, are provided in Online Resource 1. Asarina procumbens Mill. was used as outgroup, based on previous studies (Ghebrehiwet et al. 2000;Vargas et al. 2004;Guzmán et al. 2015;Carnicero et al. 2017).

DNA extraction, amplification and sequencing
Total genomic DNA was extracted from ca. 10 to 30 mg silica gel-dried leaf material or herbarium vouchers following a CTAB protocol (Doyle and Doyle 1987) with some modifications (Tel-Zur et al. 1999). When less than 10 mg of dried material was available, no sorbitol washing was applied.
We amplified the ITS region and the conserved 3'ETS region of the nuclear ribosomal DNA (nrDNA), as well as the ndhF region and the rpl32-trnL UAG spacer of the plastid DNA (cpDNA). We used the primers ITS1 and ITS4 (Sun et al. 1994) for ITS, the primers Ast1F and 18SETS (Markos and Baldwin 2001) for 3'ETS, the primers 3'F (Eldenäs et al. 1999) and +607 (Kim and Jansen 1995) and the internal primers ndhF CymbF and ndhF CymbR (Carnicero et al. 2017) for ndhF, and the primers rpl32F and trnL UAG (Shaw et al. 2007) for rpl32-trnL. The profile used for amplification of ITS included 4-min denaturation at 95 °C, followed by 30 cycles of 90-s denaturation at 94 °C, 2-min annealing at 55 °C and 3-min extension at 72 °C, with an additional final step of 15 min at 72 °C. The profile used for amplification of the rpl32-trnL UAG spacer included 3-min denaturation at 94 °C, followed by 30 cycles of 40-s denaturation at 95 °C, 2-min annealing at 52 °C and 2-min extension at 72 °C, with an additional final step of 10 min at 72 °C. We followed the PCR profiles described in Galbany-Casals et al. (2009) for ETS and Galbany-Casals et al. (2012) for ndhF. Direct sequencing was conducted by the Macrogen Sequencing Service (Seoul, South Korea). See Online Resource 1 for information on the vouchers and the sequences.

Phylogenetic analyses
The sequences were first examined and edited using Mega 6.06 (Tamura et al. 2013) and Finch TV 1.4 (Geospiza). Sequences were initially aligned using Clustal X 2.0.9 (Larkin et al. 2007) and adjusted manually in Mega 6.06. Ambiguous regions in the cpDNA alignment were detected and excluded using Gblocks v.0.91 (Castresana 2000;Talavera and Castresana 2007) with relaxed conditions in order to preserve as much information as possible: "Minimum Number Of Sequences For A Conserved Position" and "Minimum Number Of Sequences For A Flank Position" were half the number of sequences, "Minimum Number Of Contiguous Nonconserved Positions" was 5, "Maximum Number Of Contiguous Nonconserved Positions" was 10, "Minimum Length Of A Block" was 5, and "Allowed Gap Positions" was "With Half." We kept the entire nrDNA matrix since we detected no ambiguously aligned regions after visual inspection. Indels were coded as binary characters using the simple indel coding method (Simmons and Ochoterena 2000). The ITS region and the conserved 3'ETS region were concatenated in a single nrDNA matrix, as were the ndhF region and the rpl32-trnL UAG spacer in a single cpDNA matrix. Plastid and nrDNA matrices were analyzed separately due to the phylogenetic incongruence found between the two genomes (see "Results").
Maximum Parsimony (MP) analyses were conducted with PAUP*v.4.0a149 (Swofford 2002) for both cpDNA and nrDNA datasets. We performed 10,000 replicates of heuristic searches with random taxon addition and tree bisectionreconnection (TBR) branch swapping and retaining all most parsimonious trees; uninformative characters were excluded. The bootstrap analyses were performed with 1000 replicates, TBR branch swapping and random taxon addition with 10 replicates. Consistency Index (CI), Retention Index (RI) and Homoplasy Index (HI) were calculated from the consensus tree (Online Resource 2).
PartitionFinder (Lanfear et al. 2012) was used to find the best model of evolution and the best partitioning scheme under the Bayesian information criterion (BIC; Schwarz 1978) for the Bayesian inference (BI) analyses. The models tested were those implemented in MrBayes. A greedy search algorithm was selected for running the analysis for each dataset. Single partitions were found for the cp and nrDNA, respectively, and therefore, no partitions were defined in further analyses. The BI analysis was conducted with MrBayes v.3.2 (Ronquist et al. 2012). Coded indels were defined as separate partitions and analyzed with the restriction site (binary) model, as recommended by the program developers (Ronquist et al., 2020). We generated 10,000 trees running MrBayes for 5,000,000 generations and sampling one of every 500 generations. After ensuring that the Monte Carlo Markov chain (MCMC) reached stationarity and that the two runs converged by checking the standard deviation of split frequencies and the potential scale reduction factor, we discarded the first 2500 trees as burn-in.

Morphological analyses
Twelve characters (six vegetative and six reproductive) were scored on the basis of previous studies on Cymbalaria and the tribe Antirrhineae (Sutton 1988;Sáez and Crespo 2005;Vigalondo et al. 2015;Carnicero et al. 2018) and the observed variability within the studied taxa (Table 2). Two extra characters were calculated as ratios of different corolla size characters, in order to summarize information on the shape of the corolla. For the vegetative characters, three measurements per specimen were averaged when possible. Five semiquantitative indumentum characters were examined under a binocular stereoscopic microscope. The remaining characters were measured on scanned specimens using ImageJ (Abràmoff et al. 2004). Unfortunately, C. microcalyx subsp. heterosepala was not included in the analyses because flowers were not available in the specimens examined. Cymbalaria ebelii was excluded from quantitative analyses due to its clear differentiation using seed characters (Speta 1986(Speta , 1989, its phylogenetic position (Carnicero et al. 2017, and results of the present paper) and previous studies pointing to its divergence from C. microcalyx (Cufodontis 1936(Cufodontis , 1947Speta 1986). The final dataset comprised 62 specimens from 37 populations (see Online Resources 1 and 3). Sixty-nine additional herbarium specimens were also examined for the purpose of building the identification key (see Online Resource 3).
Quantitative analyses were conducted using a set of R functions contained in MorphoTools ver. 1.01 (Koutecký 2015). Pearson's and Spearman's correlation coefficients were computed to reveal correlation structure among the characters and to ensure that no strong correlations (> |0.9|) were present. After standardization to zero mean and unit variance, principal component analysis (zero-centered PCA based on a covariance matrix) was applied to display the overall variation pattern along the first two components.
Five seed characters were studied under a binocular stereoscopic microscope and scored (Table 2), but they were excluded from the morphometric analysis due to the high amount of missing data. Scanning electron microscopic (SEM) images were taken with the aim to better describe and illustrate the observed differences in testa sculpturing of the seeds. We analyzed mature seeds of a subset of six specimens, with special focus on taxa lacking seed micrographs in

Flow cytometry
Relative genome size (RGS) measurements were taken with the aim of inferring the ploidy level of taxa without published chromosome counts. However, due to the lack of fresh mature seeds able to germinate and the low quality of measurements performed on dry leaf material, only four taxa were analyzed. Fresh leaf material of germinated seeds from C. microcalyx subsp. paradoxa and the putative new taxon from Taigetos was used (Online Resource 1). Additionally, C. microcalyx subsp. microcalyx and C. acutiloba subsp dodekanesi were measured in order to confirm previously documented ploidy levels (Sutton 1988; P. Carnicero unpubl. data). Diploid C. muralis G.Gaertn., B.Mey. & Scherb. from the Botanical Garden of the University of Innsbruck, was also measured as a reference. Flow cytometry (FCM) of 4',6-diamidino-2-phenylindole (DAPI; final concentration b 0.036 M)-stained nuclei was used to estimate RGS and to assess DNA ploidy levels (Suda and Trávníček 2006). The standard used to determine DNA amounts was Solanum pseudocapsicum L. (2C = 2.58 pg; Temsch and Greilhuber 2010). Fresh leaf tissue was chopped with leaf material of the standard and processed as described in Suda et al. (2007). The relative fluorescence intensity of 3000 particles was recorded using a Partec CyFlow Space flow cytometer (Partec GmbH, Münster, Germany). Partec FloMax software was used to evaluate the histograms, which were manually gated. RGS was calculated as ratio between the relative fluorescence of sample and standard. The reliability of the measurements was assessed by calculating coefficients of variation (CV) for the G1 peaks of both the analyzed sample and the reference standard. Analyses yielding a CV threshold of > 5% were discarded, and the samples measured again. Ploidy levels were estimated by comparing RGS of samples with that of diploid C. muralis (2n = 14; Sutton 1988).
In the cpDNA analyses, Cymbalaria was divided into two clades (Fig. 3). The first main clade grouped a subclade of all the specimens of uncertain identity from Taigetos (BS = 96%, PP = 1) with some of the central Mediterranean species (BS = 79%, PP = 1), and the second main clade grouped all remaining Cymbalaria specimens, although with low statistical support (PP = 0.94). Statistical supports for the basal branches in the second clade were generally Fig. 2 Bayesian inference phylogram of Cymbalaria derived from the concatenated nrDNA matrix (ITS and 3'ETS). Dotted lines indicate branches that have been manually shortened. Bayesian posterior probabilities ≥ 0.80/bootstrap support values ≥ 60% are indicated. Names of eastern taxa follow Speta (1986) (Table 1), except for C. paradoxa and C. spetae (see "Taxonomic treatment"). The non-accepted taxon C. microcalyx subsp. alba is indicated in parentheses. Gray boxes indicate clades composed only by eastern Mediterranean species. Letters next to the species names indicate the geographic origin of eastern Mediterranean specimens (A Anatolia, B Balkans, C Crete, Kt Karpathos, Kz Kastelorizo, P Peloponnese, R Rhodes, S Samos) ◂ 1 3 low. Within the eastern taxa, only C. minor and C. paradoxa (three specimens from a single population) were monophyletic (BS = 77%, PP = 0.97 and PP = 0.98, respectively). Most eastern taxa were split in two or more clades. The most remarkable relationships congruently inferred based on nrDNA and cpDNA were: 1) the monophyletic specimens from Taigetos; and 2) the Eastern Aegean Clade (PP = 0.99). Cymbalaria longipes 1 clustered with C. longipes 2 (BS = 98%, PP = 1), and this clade was sister to C. minor (PP = 0.99).

Morphological analyses
Correlation coefficients did not exceed 0.95 for any pair of characters; therefore, all characters were retained for further analyses (Table 2). In the PCA, the first axis accounted for 50% of the variation and the second axis for 17%. The ordination diagram (Fig. 4a) showed a strong separation of the specimens of uncertain identity from Taigetos (named C. spetae in the figure) from the remaining eastern species. Its clear divergence is illustrated with boxplots for the characters Calyx length and ratio spur length/corolla length (Fig. 4b, c). Cymbalaria longipes also constituted a welldelimited group. Cymbalaria microcalyx s. l. specimens formed a cloud with less clear separation among infraspecific entities. However, three groups with limited overlap could be observed: 1) C. acutiloba subsp. acutiloba and C. microcalyx subsp. paradoxa (C. paradoxa in the figure, according to the final taxonomic treatment proposed here); 2) C. minor and C. acutiloba subsp. dodekanesi; and 3) C. microcalyx subsp. microcalyx (including subsp. alba).
The examination of seeds under the binocular stereoscopic microscope, SEM images (Fig. 5) and previous studies (Speta 1986(Speta , 1989Sutton 1988) supported the groups visualized in the PCA. Details on seed morphology are summarized in Table 3.

Flow cytometry
Mean RGS for the known tetraploids C. acutiloba subsp. dodekanesi and C. microcalyx subsp. microcalyx were 0.63 (0.62-0.64) and 0.60, respectively. Both exhibited, respectively, 3.15 and 2.99 times the monoploid RGS of the diploid C. muralis, although chromosome counts confirmed them as tetraploids (Speta 1986;Carnicero unpubl. data). The specimens of uncertain identity from Taigetos and C. paradoxa showed a mean RGS of 0.61 and 0.7 (0.70-0.71), respectively, 3.1 and 3.5 times the monoploid RGS of the diploid C. muralis. Since the ratios to C. muralis are close to those shown by the two known tetraploids, we speculate that they both are likely tetraploids. All the tetraploid RGS values measured here are lower than expected if genome size increased in direct proportion with polyploidy, which can be explained by heterogeneity in the monoploid genome size across the studied species and genome downsizing, an often reported phenomenon in non-recently formed polyploids (Leitch and Bennett 2004;Frajman et al. 2015;Lazarević et al. 2015).

Discussion
Our study recovered the same early divergent lineages for the genus Cymbalaria reported in Carnicero et al. (2017) and confirm that the eastern taxa are in comparison more phylogenetically diverse than the central and western species, by occurring in two out of the three main lineages (Fig. 2). Our extended molecular sample and the addition of morphological data, allow us to confirm that the most recent taxonomic treatment of the genus (Sutton 1988), adopted by recent publications (e.g., Dimopoulos et al. 2013), underestimate the extant diversity of the eastern Mediterranean Cymbalaria and should therefore be updated. With the notable exception of a hitherto unrecognized species, which we formally describe below, our data largely support Speta's (1986) taxonomic treatment for the group, i.e., the recognition of C. longipes and the split of a variable and widely distributed C. microcalyx (sensu Sutton 1988;Dimopoulos et al. 2013), into at least three groups. These correspond to (1) C. ebelii, (2) C. microcalyx as well as (3) a group  (Stuessy et al. 2014) and is also supported by diagnostic characters allowing the species' identification. However, our data also show the need of further studies to resolve the circumscription and adscription of C. acutiloba and C. minor. For these two taxa, current taxonomy does not reflect their evolution, but the resolution and sampling obtained here were not sufficient to propose a definitive solution. In addition, we confirm the previously suggested high diagnostic value of seed morphology in the group, as reported for other members of the tribe Antirrhineae (Elisens 1985;Vigalondo et al. 2015). Cymbalaria ebelii, endemic to the Skadar Lake area in Montenegro and Albania, did not group with any eastern Mediterranean species in the molecular analyses (Figs. 2, 3). Instead, it clustered with moderate support with the Sicilian endemic C. pubescens in the nrDNA analysis. Indeed, the area occupied by C. ebelii has closer floristic affinities with the Apennine Peninsula and the northern Balkan coast than with the southern Balkan Peninsula, the Aegean region and Anatolia (Takhtajan 1986). Our molecular results are congruent with other phylogenetic and phylogeographic studies in the area showing a strong biogeographic split between the northern and the southern Balkan Peninsula (Bardy et al. 2010;Surina et al. 2011;Hilpold et al. 2014) and with the hypothesis of a polyploid origin of C. ebelii and C. pubescens from a central Mediterranean ancestor (Carnicero et al. 2017). It remains open though, whether the two species originated from a single polyploid speciation and later diverged, or alternatively two independent events from closely related ancestors led to the origin of each of the species, and whether they originated through allo-or autopolyploidy. The separate position of the two species in the cpDNA tree (Fig. 3) as well as the ploidy level and distribution data available, suggests two independent events at two edges of the distribution of a C. muralis-related ancestor (i.e., southern Apennine Peninsula and northern Balkans). That would be congruent with the often observed pattern, that polyploidy often entails niche differentiation from their diploid relatives, and that polyploids rarely succeed where the parental species occur (Ramsey 2011; Baniaga et al. 2020). Morphologically, C. ebelii has unique oblong-ovoid seeds with a few low, scarcely anastomosed, longitudinal ridges (Speta 1986;Fig. 5) and capsules with folded pericarp reducing the space for the few seeds contained (Speta 1989). Thus, both molecular and morphological data support the recognition of C. ebelii as a separate species, which agrees with Speta (1986).
The recognition of C. microcalyx sensu Speta (1986) is supported by the nrDNA tree (Fig. 2) and morphological analyses (Fig. 4a). Our data further support to split it into two allopatric subspecies, i.e., subsp. microcalyx from the Peloponnese and subsp. heterosepala from western Crete. The moderately supported paraphyly of C. microcalyx subsp. microcalyx in the nrDNA tree (Fig. 2) suggests the Peloponnese as the ancestral area for the species followed by further dispersal to Crete leading to the origin of C. microcalyx subsp. heterosepala. This dispersal event might have been facilitated by the reduced distance between the continent and Crete during the marine regressions in the Pleistocene (Van Andel and Shackleton 1982). A similar case has been revealed in Campanula subgenus Roucela (Crowl et al. 2015). That distribution might explain the split of the two subspecies in the cpDNA tree: the position of C. microcalyx subsp. heterosepala might result from chloroplast capture after gene flow with other Aegean taxa (e.g., C. longipes or C. acutiloba subsp. dodekanesi, which occur in Creta as well, Fig. 3). Alternatively, that pattern could be explained by incomplete lineage sorting of ancestral variation in the cpDNA, which might be suggested by the close position of several taxa from completely different geographic regions (Fig. 3). These two patterns are more often impossible to disentangle with certainty (Pelser et al. 2010), and the lack of resolution of the cpDNA analyses prevents us from formulating further hypotheses. In contrast, C. microcalyx subsp. alba seems to have no taxonomic value, since (1) it is nested within C. microcalyx subsp. microcalyx and (2) its diagnostic characters (i.e., white flowers, smaller calix and longer spurs; Voliotis 1990, Tan andIatrou 2001) can be observed in some individuals of other studied C. microcalyx subsp. microcalyx populations (P. Carnicero and Th. Constantinidis, pers. obs.). Seed characters best distinguish the two first mentioned subspecies. In C. microcalyx subsp. heterosepala at least some seeds form a concrescent mass, similarly as in C. longipes and C. microcalyx subsp. paradoxa, while in subsp. microcalyx the seeds are free (Fig. 5). Cymbalaria microcalyx subsp. heterosepala was originally described based on the different length of the lobes of the calyx (Rechinger 1943). This feature, however, is not even constant in all flowers of the holotype, and it is also found in some populations of C. microcalyx subsp. microcalyx (Greuter and Rechinger 1967). Therefore, it cannot be used as diagnostic character.
Cymbalaria longipes, C. acutiloba, C. minor and C. microcalyx subsp. paradoxa are clearly phylogenetically distinct from C. microcalyx (Fig. 2), but the lack of resolution both within the Circum-Aegean clade and with sister clades in both phylogenetic trees (Figs. 2, 3) prevents reconstruction of exact relationships. However, a few conclusions can be drawn. (1) The apparent split between C. microcalyx and that group, clearly shows that C. microcalyx subsp. paradoxa as circumscribed in C. microcalyx is misleading, and we therefore combine it at the species level (see Taxonomic treatment). Its phylogenetic position in both trees (Figs. 2, 3) and ploidy level (tetraploid based on RGS) indicates a polyploid origin of C. paradoxa, most probably with the diploid C. longipes as parental species. Polyploidy confers immediate isolation of the newly created lineage from their diploid relatives (Stebbins 1950;Grant 1981;Husband and Sabara 2003;Soltis et al. 2014), allowing its persistence as narrow endemic, which could otherwise be threatened by gene flow with the close populations of C. longipes in the island. Morphological data also support the strong relationship between C. longipes and C. paradoxa, by sharing similar capsules and seeds and their close position in the PCA (Fig. 4a). This adds a third polyploid speciation event in the genus Cymbalaria apart from those previously discussed (Carnicero et al. 2017). (2) The topology of the Circum-Aegean Clade in the nrDNA tree (Fig. 2) suggests a mainland origin in Greece followed by eastward dispersal, which first led to the origin of C. acutiloba subsp. dodekanesi in the Aegean Islands and subsequently to that of the Eastern Aegean Lineage (Fig. 2). However, due to the low support in the basal branches of the clade in the nrDNA tree and the low supports and incongruence observed in the cpDNA tree (Fig. 3), this hypothesis remains highly speculative. For the same reasons, and the unknown ploidy level of C. acutiloba subsp. acutiloba, the exact role of polyploidy in the group is still uncertain. Further studies at the populational level for eastern Aegean species combined with comprehensive RGS might contribute to disentangle the evolutionary history and systematics of this group. (3) The presence of C. acutiloba subsp. dodekanesi in both the western and the eastern Aegean Islands, which are considered to be floristically isolated by Rechinger's line (Rechinger 1950;Strid 1996), implies that the species crossed this biogeographic barrier, because it is much younger than the barrier (Carnicero et al. 2017). Rechinger's line has been partly confirmed as a genetic break zone with molecular data for some groups (Bittkau and Comes 2005;Crowl et al. 2015), but Rechinger (1950) himself reported that some taxa occur at both sides of the barrier (e.g., Campanula delicatula Boiss., Helichrysum orientale (L.) Gaertn.), suggesting a relative permeability. (4) Although Greuter and Rechinger (1967) considered C. longipes a "young" taxon of the C. microcalyx aggregate, ploidy data seriously challenge this statement as C. minor and C. paradoxa are tetraploid, while C. longipes is diploid (Greuter and Rechinger 1967). In addition, morphological divergence-it is the only glabrous taxon in the eastern Mediterranean region-suggests recognizing this taxon at the rank of species is appropriate. As mentioned above, the incongruence found between cp and nrDNA trees might either indicate gene flow and/or persistence of ancestral genetic variation (Figs. 2, 3).
As already shown in Carnicero et al. (2017), polyploid evolution played an important role in the evolution of the genus Cymbalaria, but its incidence on eastern species was only confirmed for C. ebelii. Our results stress its particular importance in the origin of the eastern taxa, by adding at least two confirmed polyploidy events to account for the origin of (1) the C. microcalyx-C. spetae clade and (2) C. paradoxa. At least a third event should be invoked to explain the tetraploid level of C. minor and C. acutiloba, but all the above-mentioned uncertainties regarding this group prevent the formulation of any detailed hypotheses.

A new phylogenetically and morphologically well-delimited species
The specimens of uncertain identity sampled from Taigetos in the southern Peloponnese constituted a monophyletic group in the molecular analyses (Figs. 2, 3), are morphologically well-distinguishable, and they are therefore below described as a new species. According the nrDNA, it is closely related to C. microcalyx, with which it shares ploidy level (Fig. 2), suggesting a common origin of both taxa through polyploid speciation. However, as mentioned above, the present data do not allow us to infer with certainty a single common origin of both taxa nor to identify the diploid ancestors. The close relatedness of the new species to the central Mediterranean diploids C. pallida and C. muralis subsp. visianii in the cpDNA analyses (Fig. 3), points them or a related ancestor as potential parental taxa, while C. microcalyx might have had a different uncertain origin, more distantly related to the extant central Mediterranean diploids. However, the observed pattern is congruent as well with a common origin and posterior divergence of the cpDNA, which could also be caused by gene flow of C. microcalyx with other eastern species. Morphologically, the specimens from Taigetos show several divergent characters with respect to all other eastern species, namely a higher number of leaf lobes, a longer calyx, a proportionally shorter spur, a high number of seeds per capsule and much smaller seeds (Figs. 5,6 and Table 3). In fact, the most similar species to this entity is C. glutinosa Bigazzi & Raffaelli from the southern Apennine Peninsula, from which it can be clearly distinguished by having a longer calyx (2.3-3.7 mm vs. ca. 2 mm) and often a longer corolla (9-14.2 mm vs. 6-12(13) mm). Moreover, these two species are distantly related and have different ploidy levels. Plants from Taigetos are tetraploid as inferred from RGS estimation, while C. glutinosa is diploid (Bigazzi and Raffaelli 2000). Altogether, this entity, which is currently known from only three populations, clearly deserves taxonomic recognition at the species level and is formally described below.

Cymbalaria spetae
Etymology: Named after the late Franz Speta, in recognition of his outstanding work on the taxonomy of Cymbalaria.
Habitat and altitudinal range: Rock crevices of shady limestone cliffs, 200-800 m a. s. l.
Distribution area: Local endemic of the Taigetos mountain range in the Peloponnese (only three populations known).

Information on Electronic Supplementary Materials
Online Resource 1. List of specimens used in molecular analyses with information on the specimen codes used in the text and figures, locality, herbarium voucher and accession numbers. Online Resource 2. Characteristics of sequences and results of phylogenetic analyses. Online Resource 3. List of additional specimens examined. Online Resource 4. Concatenated nrDNA alignment used for the phylogenetic analyses in nexus format. Online Resource 5. Concatenated cpDNA alignment used for the phylogenetic analyses in nexus format.

Supplementary material 1Supplementary material 2Supplementary material 3 Supplementary material 4Supplementary material 5Acknowledgements
This research was funded in part by the Spanish government (CGL2010-18631/BOS and the Flora iberica project, CGL2011-28613-C03-01) as well as by the Catalan government (2014-SGR 514 and 2017-SGR1116). We are grateful to the herbaria that provided material or kindly sent scanned images for examination (ATHU, B, BM, E, G, K, FI, W) and specially to the team of BC. We are also grateful to R. Vilatersana for collecting material from the Aegean Islands. We thank P. Schönswetter for valuable comments on the final manuscript. Pau Carnicero benefited from the support of a PIF Ph.D. student fellowship from the Universitat Autònoma de Barcelona. We thank M. Gassner from the Institute of Botany of Innsbruck for performing relative genome size measures. We thank the gardeners of the Botanical Garden of Barcelona for successfully cultivating accessions of Cymbalaria. We are also grateful to A. Crowl and an anonymous reviewer for their constructive and helpful comments on the first version of the manuscript. Availability of data and material The DNA sequence generated and analyzed during the current study is available in GenBank (http://www. ncbi.nlm.nih.gov/genba nk/). Other datasets (alignment, morphological data, morphometric analyses scripts) are available from the corresponding author on reasonable request.

Compliance with ethical standards
Conflicts of interest The authors declare that they have no conflict of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.