Nomenclature

The electronic version of this article in Portable Document Format (PDF) in a work with an ISSN or ISBN will represent a published work according to the International Code of Nomenclature for algae, fungi, and plants, and hence the new names contained in the electronic publication of a PLoS ONE article are effectively published under that Code from the electronic edition alone, so there is no longer any need to provide printed copies.

In addition, new names contained in this work have been submitted to IPNI, from where they will be made available to the Global Names Index. The IPNI LSIDs can be resolved and the associated information viewed through any standard web browser by appending the LSID contained in this publication to the prefix http://ipni.org/. The online version of this work is archived and available from the following digital repositories: PubMed Central, LOCKSS, ResearchGate (https://www.researchgate.net/), ARPI (https://arpi.unipi.it/).

Ethics statement

The plant tissues (a small portion of leaf) analyzed are from herbarium specimens. These specimens are stored in Herbaria. In addition, when the plants were sampled, no specific permissions were required for this. Finally, the sampling has been performed without damaging the populations or the species in the case of narrow endemics with a single known population.

Pinguicula sampling

The plant tissues (i.e. small portion of leaf) analyzed were sampled from herbarium specimens. These source of material and Herbaria is summarized in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. List of Pinguicula accessions sequenced in the molecular study, available karyological information, voucher information, origin, ITS_T—GC content (%) and GenBank accession no.

https://doi.org/10.1371/journal.pone.0167610.t001

A total of 38 accessions, corresponding to 13 species, was sampled. A single accession from Italy (P. cf. christinae) exhibited peculiar morphological features and its taxonomic position has been investigated in the present study (see morphometric analyses in S1 File).

Most of the sampled taxa are endemic to Italy, except for those accessions coming from Corsica (P. corsica), France (P. cf. apuana from Bendola, P. hirtiflora from Val Roya, P. reichenbachiana Schindl. also from Val Roya), Bulgaria (P. balcanica Casper) and Greece (P. hirtiflora). Sequence data of further 26 Pinguicula species (73 total accessions), many of them with a range not covering Italy, were obtained from GenBank, based on the works of Degtjareva et al. [14] and Kondo & Shimai [15] (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. List of GenBank ITS_T accessions used for the phylogenetic analysis of Pinguicula taxa (taxon, distribution, GenBank no., reference).

https://doi.org/10.1371/journal.pone.0167610.t002

DNA extraction, PCR amplification and sequence analyses

Total DNA was extracted from approx. 3 mg of fresh leaf or 2 mg of dried leaf material using ZR Plant/Seed DNA MicroPrep (ZYMORESEARCH), according to the manufacturer’s protocol, then resuspended in 20 μL sterile distilled water. The concentration was estimated by quantifying 1 μl of DNA extract using a Qubit dsDNA HS Assay Kit with the Qubit ver. 2 Fluorometer (Invitrogen, Thermo Fisher Scientific Inc., Waltham, MA, USA).

ITS_T was amplified by using plant universal primers that were reported in the literature {Forward-18S(3’) 5’-GGA GAA GTC GTA ACA AGG TTT CCG-3′, and Reverse-26S(5′)-internal 5’-TTC GCT CGC CGT TAC TAA GGG-3’} [28] and selective internal primers designed on Pinguicula sp. accessions from GenBank (Pingui_18Sfor 5’-AAG GAT CAT TGT CGA DWY Y-3’ and Pingui_26Srev 5’-TGR GGT CGC RRR IGT TGG CR-3’). The use of nested-PCR with selective primers for Pinguicula accessions can be very useful to remove possible contaminators (e.g. pollen) occurring on the flypaper traps. All oligos were synthetized by Macrogen Inc., the annealing temperature was 55°C for the two primer pairs. The volume of each polymerase chain reaction (PCR) was 25 μl, with c. 10–20 ng of DNA template, 2.5 μl of 10× DreamTaq Buffer (Thermo Fisher Scientific Inc.), 0.5 μl each of the 2.5 mM nucleotides (Promega), 0.125 μl of 50mM primer and 0.25 μl DNA DreamTaq polymerase (5Uμl^–1) (Thermo Fisher Scientific Inc.). Amplifications of recalcitrant DNA templates were performed by using nested-PCRs with internal selective primers or KAPA3G Plant PCR Kit (KAPABIOSYSTEMS). The cycling parameters of the PCRs were performed according to the manufacturer’s instructions in a SimpliAmp thermal cycler (Applied Biosystems, Thermo Fisher Scientific Inc.). Amplification products were purified, using the NucleoSpin® Gel and PCR Clean-up (Macherey-Nagel) following the manufacturer’s protocols. An aliquot of approximately 70 ng of purified DNA template was used with the Bright Dye Terminator Cycle Sequencing Kit (ICloning) following the procedure according to Di Maio & De Castro [41] and analysed with a 3130 Genetic Analyzer (Applied Biosystems, Thermo Fisher Scientific Inc.).

Several sequences with slippage events for a mononucleotide repeat (polyC/G) were cloned with the CloneJET PCR Cloning Kit (Thermo Fisher Scientific Inc.) following the protocol of the manufacturer. Transformation was carried out using StrataClone SoloPack Competent Cells (Agilent Technologies). Bacteria were cultured in LB medium at 37°C for 30 min and then on LB agar plates containing 100 ug/ml ampicillin. Eight randomly selected clones from each transformation were amplified and sequenced with the pJET2.1 forward and pJET2.1 reverse primers (Thermo Fisher Scientific Inc.) which matching in the flanking regions of the pJET1.2/blunt vector polylinker.

Complete sequences of both strands of each PCR product were processed using the AB DNA Sequencing Analysis ver. 5.2 (Applied Biosystems, Life Technologies), and visually checked using Chromas Lite ver. 2.1.1. software (Chromas Lite version 2.1, http://technelysium.com.au/?page_id=13). Then, sequences were aligned using ClustalW ver. 1.4 software [42] as daughter processes of BioEdit ver. 7.2.5 software [43]. The aligned sequences were visually inspected to correct gap distributions devoid of biological meaning. The sequences obtained in this study are available from GenBank and accession numbers are provided in Table 1.

Phylogenetic analyses

One hundred and twenty accessions were analysed using two phylogenetic approaches: Bayesian inference (BI) and maximum parsimony (MP). The trees generated by these methods were checked for congruence. Outgroups were chosen after several preliminary analyses considering the literature data [7, 14, 15]: Kigelia africana (Lam.) Benth. {GenBank (GB): AY178638}, Antirrhinum majus L. (GB: EU677219) and some of its subspecies {i.e. A. majus subsp. linkianum (Boiss. & Reut.) Rothm., GB: EU677214; A. majus subsp. cirrhigerum (Ficalho) Franco, GB: EU677200; A. majus subsp. tortuosum (Vent.) Rouy, GB: EU677242; A. majus subsp. litigiosum (Pau) Rothm., GB: EU677216}. Representative of Pinguicula sister group were also included: Genlisea violacea St.-Hil. (GB: AB212116), Utricularia intermedia Hayne (GB: DQ225109) and U. minor L. (GB: AB212118).

The BI method for phylogenetic reconstruction was implemented using MrBayes ver. 3.2.2 software [44, 45] and jModelTest ver. 2.1.4 software to obtain the best nucleotide substitutions model [46, 47]. Preliminary analyses were performed to obtain the optimal BI settings. Bayesian Markov Chain Monte Carlo (MCMC) algorithm was run for 15,000,000 generations discarding the first 12% generations (burninfrac = 0.12). Four runs were performed and two heated chains were used. Bayesian posterior probabilities (PP) were obtained from the 50% majority rule consensus of 52,800,000 trees (13,200,000 trees from each runs). The general time reversible (GTR) + proportion of invariant sites (I) + gamma distribution (G) model was used in the analyses (set nst = 6 rates = invgamma), according to the results obtained with jModelTest under the Akaike Information Criterion (AIC) [48]. The same model was obtained using also the correction for sample size (AICc).

MP reconstruction were performed with TNT ver. 1.1 software [49, 50]. Settings included: gaps scored as missing data; maximum storage space of 99,000 trees; a tree storage space per iteration of 100; 100 iterations branch swapping via tree bisection-reconnection. In addition, we used also the "new technology search" (which is guaranteed to find at least one shortest tree), setting 10 hits to the shortest length and an initial level of 100. Bootstrap values [51] were calculated from 1,000 replicates.

Molecular analysis

The alignment of ITS_T sequences of all Pinguicula accessions was relatively straightforward due to not complex variation among sequences (598–625 bp, P. caerulea Walter-P. agnata Casper/P. moranensis Kunth). In the 38 accessions of Pinguicula sequenced in the present study, the length of the ITS_T ranged from 607 (P. hirtiflora) to 617 bp (P. vulgaris s.l.). The sister group taxa had relatively variable sequences in length, with ITS sequences ranging from 501 bp (Utricularia intermedia) to 650 bp (Genlisea violacea).

After introducing the necessary gaps, the ITS_T alignment resulted in a matrix of 695 characters, of which 200 constant and 495 variable (425 potentially parsimony-informative). The mean G+C content of the ITS matrix was 63.5%. Among the Pinguicula species sequenced in this study, the lowest G+C content was found in some P. hirtiflora accessions (P23, P24) (55.4%) and the highest in P. balcanica (P5, P6) (66%) (Table 1).

Phylogenetic reconstructions

Inferred phylogenies from the BI and MP analyses of the ITS_T datasets produced a very similar topology with similar statistical support (Figs 2 and 3). The MP analysis yielded 492 most-parsimonious trees with a consistency index (CI) of 0.57 and retention index (RI) of 0.89. The strict consensus tree was 1592 steps long. The bootstrap support showed high values (BS > 70%) for the 72% of the nodes (Figs 2 and 3). The Bayesian tree with posterior probabilities (PP) is also shown in Figs 2 and 3. An unresolved basal structure for some Pinguicula groups is detected, albeit many clades are well supported (Figs 2 and 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Bayesian phylogenetic tree of Pinguicula (clades A-D), reconstructed using ITS_T sequences.

Circles on the nodes show posterior probabilities (PP) from the Bayesian analysis under the GTR+I+G model (15,000,000 generations, burn-in 12%). Grey circles: PP = 100; white circles: PP = 99–95; black/white circles: PP = 90–93; black circle: PP = 85. Branches present in the MP strict consensus tree is marked with a dashed line (MP trees = 2720, steps = 1605, CI = 0.57, RI = 0.89). Bootstrap percentages (1000 replicates) are given below the branches. D or K letters after the taxon refer to the accessions obtained from Degtjareva et al. [14] or Kondo & Shimai [15], respectively (Table 2). See Table 1 for further details about our Pinguicula accessions.

https://doi.org/10.1371/journal.pone.0167610.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Bayesian phylogenetic tree of Pinguicula (clade E), reconstructed using ITS_T sequences.

Circles on the nodes show posterior probabilities (PP) from the Bayesian analysis under the GTR+I+G model (15,000,000 generations, burn-in 12%). Grey circles: PP = 100; white circles: PP = 99–95; black/white circles: PP = 90–93; black circle: PP = 85. Branches present in the MP strict consensus tree is marked with a dashed line (MP trees = 2720, steps = 1605, CI = 0.57, RI = 0.89). Bootstrap percentages (1000 replicates) are given below the branches. D or K letters after the taxon refer to the accessions obtained from Degtjareva et al. [14] or Kondo & Shimai [15], respectively (Table 2). See Table 1 for further details about our Pinguicula accessions.

https://doi.org/10.1371/journal.pone.0167610.g003

A basal polytomy defined by five clades can be detected. The basally collapsed branches consist of three smaller groups (Clade A = P. agnata+P. moranensis; Clade B = P. ramosa Miyoshi+P. villosa L.+P. variegata Turcz. and Clade C = lineage of P. alpina) and a well-supported lineage (Clade D) where P. hirtiflora/P. lavalvae/P.crystallina are sister to Pinguicula species from the south-east United States of America (P. caerulea+P. lutea Walter+P. primuliflora C.E.Wood & R.K.Godfrey+P. planifolia Chapm.) and the European western-Atlantic P. lusitanica. In the next node (Clade E), which is not as well supported as other clades, four subclades can be identified. The first subclade (E1) includes P. vulgaris and allied taxa, with internal further subdivisions, including many Italian endemics (e.g. P. apuana, P. fiorii, and P. christinae), and other taxa related to P. vulgaris. The second subclade (E2) includes a single lineage with the Italian endemic P. poldinii. The third subclade (E3) includes further subdivisions of species with mostly Iberian origin, that will be discussed in more detail. Finally, the fourth subclade (E4) includes P. balcanica and P. vallisneriifolia Webb.

Comparisons to previous phylogenetic reconstructions

Despite we could not include all taxa comprised in the genus Pinguicula, our sample size and array of species was sufficient to make phylogenetic assessments on the Italian endemic taxa (Figs 2 and 3). Comparing the phylogenies obtained with the ITS_T data published by Degtjareva et al. [14] and by Kondo & Shimai [15] and the matK-trnK sequences published by Cieslak et al. [8], the overall topology of Pinguicula species is generally consistent, although showing several differences.

Cieslak et al. [8] in their matK-trnK analysis, identified five main clades, partly overlapping with our results. Among the main similarities, the deep branching position of species belonging to our clade D (P. lusitanica and P. crystallina) and the overall placement of taxa allied to P. vulgaris in a single clade (corresponding to our clade E, Fig 3). Among the incongruences, P. mundi and P. corsica are placed on different lineages, whereas in our and in the previous ITS_T studies [14, 15] the aforementioned species are always placed in the same clade. Pinguicula primuliflora and P. lutea are included in the same clade of P. lusitanica, whereas in our analysis they form a distinct clade (B). The position of P. alpina is also in contrast with our results (clade C, Fig 2), since Cieslak et al. [8] place it in an isolated position within East Asian species (our deep branching clade D). The differences between our results and those published by Cieslak et al. [8] can be explained by the use of different molecular markers, as already mentioned in the Introduction.

No major differences are detected in the tree topologies with respect to previously published ITS_T studies. Kondo & Shimai [15] detected P. crystallina and P. hirtiflora as sister to all the remaining Pinguicula species. On the contrary, we found P. lusitanica, a species occurring on western coast of the Atlantic Ocean, from northern Morocco to Scotland [52], as belonging to the same, earliest branching, clade (D) (Fig 2). The rest of the phylogenetic reconstruction of Kondo & Shimai [15] is largely consistent with our finding, but included only two Italian endemics (P. fiorii and P. poldinii).

Generally, the topologies found by Degtjareva et al. [14] are also consistent with our findings. Similarly to Kondo & Shimai [15] and Cieslak et al. [8], they identified P. crystallina and P. hirtiflora as sister group to all other Pinguicula species. Pinguicula lusitanica, once again, is placed outside this clade. Pinguicula alpina is found in the same clade as P. moranensis, but these species belong to different clades in our results (A and C, respectively) (Fig 2).

The taxonomical evaluation of P. lavalvae

All the species within clade D (Fig 2) show a tropical growth type [3], and have been classified within P. sect. Cardiophyllum [5]. Among them, P. crystallina occurs east of the Aegean sea in southern Turkey and Cyprus, while P. hirtiflora occurs west of the Aegean sea in the Balkans and in separate populations in southern Italy, mainly in Campania and a single population in Calabria [53, 54, 55, 56]. A large degree of variability and complexity is known in this group, with several infraspecific taxa described during the years for P. hirtiflora [5], and a recently described narrow-endemic to Turkey related to P. crystallina, i.e. P. habilii Yıldırım, Şenol & Pirhan [57], not included in our analysis. Given their unique basic chromosome number (x = 14) [5], it has been hypothesised that P. crystallina/hirtiflora may be the result of an ancient hybrid event involving a species similar to P. lusitanica (x = 6) and a species similar to P. corsica (x = 8). This hypothesis seems partly supported by our investigation, given the relative closeness of the P. crystallina/hirtiflora complex to P. lusitanica. Our data also confirm that P. crystallina and P. hirtiflora are separate species [7, 14].

It is interesting to note that there is a remarkable ITS_T molecular variation within P. hirtiflora. In particular, the accessions from Calabria (southern Italy) (P21-P22, Fig 2 and Table 1) fall in a clade distinct from P. hirtiflora (and P. lavalvae) from Campania (P18-P20 and P25-P26, respectively, Fig 2 and Table 1), as does also the alien population from south-eastern France, Val Roya (P24, Fig 2 and Table 1) [54]. This likely attests (a) for a Balkan origin of the population from Calabria, (b) for a Balkan (or Calabrian) artificial origin for the alien population occurring in south-eastern France, known since the 2000s [54]. It can be speculated that the newly found alien populations of P. hirtiflora in the Czech Republic [58] and Switzerland [56] also derive from commercialized Balkan populations.

Pinguicula lavalvae was recently described as a narrow endemic to Sabato Valley, Mts. Picentini (Campania, southern Italy) [23]. The authors compared it with P. hirtiflora from different portions of its range, distinguishing P. lavalvae for several phenotypic features of the corolla and calyx. Despite this, the latter species showed a complete sequence identity with the accessions of topotypical P. hirtiflora from Campania [59]. The ITS sequence, along with the geographic vicinity of the two taxa, points to an assignment of P. lavalvae to an intraspecific rank within P. hirtiflora. Indeed, Fleischmann [56] considered P. lavalvae as simply one of the many white-flowered variants (P. hirtiflora f. pallida) of P. hirtiflora, as those that are found in the Balkans and elsewhere in Italy [60]. Nevertheless, the ITS_T variation in all accessions from Campania compared to the Balkans and an ecological niche shift of these populations (M. Innangi et al. in preparation) highlights the necessity of more investigation before establishing taxonomical ranks.

Subclade E1a: the clade comprising most Apennine endemics

A single clade unites four of the Pinguicula sect. Pinguicula species endemic to the Apennines: P. apuana, P. christinae, P. fiorii, and P. vallis-regiae. Only one GenBank accession of P. fiorii is falling outside this clade (Fig 3), but it must be noted that this accession was obtained from seeds [14], and may well represent a misidentification or confusion of materials. This clade includes also another Italian endemic species, the tetraploid P. mariae. On a slightly different branch within the subclade, there are two other octoploid species, P. dertosensis (Cañig.) Mateo & M.B.Crespo, endemic to south-eastern Spain [61, 62] and P. macroceras Pall. Ex Link, endemic to northern America and Japan [7]. The four species endemic to Apennines have very similar sequences, which are identical in some case (i.e. P. apuana, P. christinae P7 and P9, and P. fiorii P15-P17), clearly pointing out to a common origin. However, given the clear qualitative and quantitative combination of character-states that has been used to distinguish these taxa, coupled with allopatry [20, 21, 22], we deem reasonable to maintain them at species level. As a matter of fact, the discriminating resolution of ITS may not be enough within such level of variability. Thus, further intrapopulation approaches could be necessary. The collocation in this clade of P. vallis-regiae, morphologically very similar to P. poldinii [20], was quite unexpected. Indeed, the latter species represents a well distinct lineage in the tree (see further in the text).

Both P. christinae and P. fiorii, show a certain morphological affinity with P. balcanica (endemic to Balkan Peninsula) (see [22, 18], respectively). According to our results, such relationships can be excluded, given the position of P. balcanica in a completely different clade (E4, Fig 3), as sister to the heterophyllous (i.e. bearing spring and summer rosettes significantly differing in shape and dimensions) P. vallisneriifolia, a peculiar Spanish endemic [61, 62]. Finally, P. apuana was compared by Ansaldi & Casper [21] with P. leptoceras Rchb., endemic to Alps [7] and with the circumboreal P. vulgaris [7]. The relationship with these species can also be excluded, given that both P. leptoceras and P. vulgaris fall in a different clade (E1b, Fig 3).

Subclade E1b: the clade of P. vulgaris and its allies

Albeit this clade is weakly supported and highly polytomic (Fig 3), several phylogenetic inferences can be provided. Pinguicula vulgaris, a widespread Circumboreal octoploid species, forms a clade together with similar taxa such as P. bohemica Krajina, a tetraploid/octoploid endemic to the Czech Republic [5, 63], the heterophyllous tetraploid P. reichenbachiana (endemic to the Maritime Alps), P. longifolia subsp. caussensis Casper (endemic to southern France) and the tetraploid Alpine endemic P. leptoceras [7], provisionally including the plants recognised as “P. arvetii” in SW Alps [64]. According to Conti & Peruzzi [20], P. vulgaris segregated into geographical races at its range borders. In the case of three of them, in central Italy, taxonomic recognition at subspecies level was proposed: P. vulgaris subsp. anzalonei, P. vulgaris subsp. ernica and P. vulgaris subsp. vestina. It is interesting to note that the three taxa share identical sequences, but collectively differ from other P. vulgaris accessions for two SNP’s (Single Nucleotide Polymorphism). This likely attests for their common origin, but given their clear diagnosability and allopatry, also in this case we deem the original subspecific rank as appropriate.

Concerning the accession of P. cf. apuana from Bendola (SE France), our results clearly highlight for a significant molecular differentiation of this population, within the clade including P. vulgaris and its allies. The taxonomic position of this population needs further investigation. As concerns P. cf. christinae from Val D’Aveto (Liguria), also falling in the same clade as P. vulgaris accessions, morphological evidence was gathered to support its difference from both P. christinae and P. vulgaris. For further detail, see floral morphometric analysis in S1 File. The Pinguicula population occurring in Val d’Aveto (Liguria), at first sight, shows morphological affinity with P. christinae, but also some character-states intermediate or similar to those of P. vulgaris s.l. Our molecular results highlight a phylogenetic affinity with the latter species, but also a molecular differentiation of this population in the ITS_T sequences (three SNP’s). In addition, our morphometric study of floral features clearly attests for a distinctive combination of character-states in these plants (S1 File). Accordingly, and given the allopatry of this population from both P. vulgaris s.l. and P. christinae, we describe here this population as a new narrow endemic species:

Pinguicula lattanziae Peruzzi & Gestri sp. nov. (urn:lsid:ipni.org:names:XXXXXXXX-X)

Type: Liguria, N Apennine: Monte Aiona, 1600–1700 m a.s.l., 21 June 2014, G. Gestri et C. Gavazzi s.n. (holotype, PI).

The new taxon is close to P. christinae and P. vulgaris, but distinct for its narrower corolla upper lobes (2.5 ± 0.76 mm, not 4.72 ± 0.81 mm and 3.44 ± 0.69 mm, respectively) and narrower corolla median lobe (2.08 ± 0.42 mm, not 3.31 ± 0.67 mm and 3.83 ± 0.65 mm, respectively). It shares with P. vulgaris the corolla opening angle (≤ 90°) and the spur length (5–7 mm). This new species is dedicated to Edda Lattanzi (Rome), expert in Italian flora, in occasion of her 85^th birthday.

Subclade E2: the status of Pinguicula ser. Prealpicae

Casper in Ansaldi & Casper [21] formally described Pinguicula ser. Prealpicae, to accommodate P. poldinii (type species), P. mariae and–putatively–P. vallis-regiae. The shared character-states used to justify this proposal were: homophyllous rosettes (i.e. spring and summer rosettes not significantly differing in shape and dimensions), widely open corollas with long spur and tetraploid status. The same authors explicitly exclude P. reichenbachiana from this series for being heterophyllous, regardless of other morphological similarities. According to our phylogenetic results, P. mariae and P. vallis-regiae represent two distinct lineages, placed in a large clade where also P. apuana, P. christinae, and P. fiorii occur (subclade E1a, Fig 3). On the contrary, the type species P. poldinii lies in a very distinct, albeit unresolved, lineage (subclade E2, Fig 3), to which eventually should be restricted the application of the ser. Prealpicae. However, this is not justified at all by morphology, and this highlights that certain characters in Pinguicula (i.e. homophyllous vs. heterophyllous rosettes, chromosome number) are useful taxonomic markers at species level, but cannot be safely used per se to establish systematic/evolutionary relationships.

Subclade E3: the phylogenetic placement of P. sehuensis and allied species, with some remarks on P. longifolia s.l.

The discovery of a new Pinguicula in Sardinia was defined as “the most important finding of a butterwort during the past 50 years in Europe” (J.S. Casper, in litt.). Pinguicula sehunensis was recently described by Bacchetta et al. [1] as morphologically close to P. dertosensis, P. nevadensis (H.Lindb.) Casper–both endemic to Spain [61, 62]–and the western European P. grandiflora Lam. [7]. According to our results, P. sehuensis, endemic to Sardinia, shares instead a common origin with P. corsica, endemic to Corsica (Fig 3). This phylogenetic structure parallels the biogeographic closeness between the two species, which also share a diploid status. Whereas P. dertosensis is placed in a completely different subclade (E2a, Fig 3), P. nevadensis and P. grandiflora belong to the same clade including also P. corsica+P. sehuensis, together with P. mundi Blanca, Jamilena, Ruiz Rejón & Reg.Zamora, also endemic to south-eastern Spain [61, 62]. The position of P. mundi as sister species to P. corsica and P. sehuensis is biogeographically significant, as until 25 Ma, in the late Oligocene, Sardinia and Corsica started to drift from modern Provence and reached their current geographical position in the middle Miocene (ca. 15 Ma) [64]. The ploidy level and ITS sequences of both P. corsica and P. sehuensis suggest that both species derived from a common diploid ancestor in the geologically and ecologically stable environments of the Hercynian massif of Corsica and Sardinia sometime between 25 and 15 Ma. Accordingly, and following the criteria proposed by Siljak-Yakovlev & Peruzzi [65], it is possible to hypothesize for P. corsica and P. sehuensis the status of patro-schizoendemic taxa.

Pinguicula grandiflora s.l., P. nevadensis and P. longifolia subsp. longifolia fall within the same subclade, and they are tetraploid. As argued for P. grandiflora, a species with Lusitanian distribution, its ploidy level and current distribution has been shaped by palaeoclimatic Quaternary events [66]. Our results further suggest that Quaternary glaciations played a prominent role in the evolution of polyploid butterworts currently scattered and isolated in the mountains.

Finally, some remarks can be done for P. longifolia s.l. The three different subspecies of P. longifolia are clearly paraphyletic, as already demonstrated by Cieslak et al. [8]. While P. longifolia subsp. reichenbachiana, endemic to few locations on the Maritime Alps in France and Italy, was already considered as a distinct species by several authors [1, 20, 21, 22], this is not the case for P. longifolia subsp. caussensis. The latter taxon, endemic to Les Causses in southern France, is clearly not related to P. longifolia s.str. and it has several quali-quantitative features that distinguish it as much as it happens for P. reichenbachiana [67, 68]. Degtjareva et al. [14] already pointed out the necessity of redefining the taxonomic state of P. longifolia subspecies. Thus, we deem appropriate to change its taxonomic status to species level as P. caussensis (Casper) Innangi, De Castro & Peruzzi stat. nov. (Bas.: P. longifolia subsp. caussensis Casper, Biblioth. Bot. 31(127–128): 154. 1966).