Intraspecific Genetic Diversity of Cistus creticus L. and Evolutionary Relationships to Cistus albidus L. (Cistaceae): Meeting of the Generations?

Cistus (Cistaceae) comprises a number of white- and purple-flowering shrub species widely distributed in the Mediterranean basin. Within genus Cistus, many taxa are subject to various taxonomic uncertainties. Cistus creticus, a prominent member of the purple-flowered clade, is a prime case of the current taxonomic troubles. Floras and databases approve different species names and utilise different or additional/fewer synonyms. Various intraspecific classification systems based on subspecies or varieties are in use. The inconsistent determination of plant material makes it difficult to compare literature regarding the phytochemical diversity and biological activities of plant material and impedes a systematic utilization of the manifold medicinal properties of C. creticus. In the present investigation, we used DNA sequence data from one nuclear region (ITS) and two chloroplast regions (trnL-trnF, rpl32-trnL) to test the intraspecific genetic diversity of C. creticus and its evolutionary relationships to the closely related C. albidus. The combined DNA data confirmed C. creticus as a rather heterogeneous species that integrates two major evolutionary lineages with clearly different genetic characteristics. The ‘Eastern Mediterranean clade’ seems to represent old and ancestral characteristics. This lineage exhibits a close relationship to the geographically distant C. albidus, expressed by very closely related ribotypes and an interspecifically shared chlorotype. The ‘Western Mediterranean clade’ is characterized by a distinctive ITS polymorphism (co-occurring paralogous ribotypes) and more distantly related chlorotypes. The formation of the genetically complex ‘Western Mediterranean clade’ seems to have involved hybridization and recurrent formation or migration movements.


Introduction
The genus Cistus L. (Cistaceae, Malvales) comprises white-and purple-flowering shrub species that are widely distributed in the Mediterranean, on the Canary Islands, and on Madeira, often in open or disturbed, fully sun-exposed vegetation. Various monographs  Table S1). Pictures of selected plants and populations are available as supplementary material in [25]. Voucher specimens are kept at the herbarium of the Institute for Animal Nutrition and Functional Plant Compounds, University of Veterinary Medicine Vienna, Austria.
The PCR cycle profile for the amplification of nrITS included a denaturation step at 95 • C for 15 min, followed by 35 cycles of 95/58/72 • C (depending on the samples) for 45/45/60 s and a final elongation step of 7 min at 72 • C. Amplification of the trnL-trnF intergenic spacer started with an initial cycle of 95/45/72 • C for 15/1/1 min, followed by 35 cycles of 95/50/72 • C for 1/1/1 min and a final elongation step at 72 • C for 7 min. The cycle profile for the amplification of the rpl32-trnL intergenic spacer started with an initial cycle of 95/56/72 • C for 15 min/45 s/60 s, followed by 35 cycles of 95/56/72 • C for 30/60/120 s and a final elongation step with 72 • C for 10 min.
PCR products were checked on 1.4% agarose gels and purified with ExoI and FastAP (Fisher scientific, Vienna, Austria) according to the manufacturer's instructions. Sequencing was performed by an external company (Microsynth, Vienna, Austria) using the primers ITS5, f or rpl32-trnL-F as sequencing primers.
The obtained sequences were edited using Geneious 5.3.6 (www.geneious.com). Concerning nrITS, some directly sequenced accessions of C. creticus (about 27%) and C. albidus (about 1%) resulted in ambiguous sequence chromatograms characterized by the presence of single to few additive polymorphic sites and noise/divergence of the sequences after a specific alignment position. These sequence chromatogram characteristics could later be attributed to an intraspecific/intraindividual ITS sequence polymorphism, namely the simultaneous occurrence of divergent ITS paralogs differing by one to five SNPs and one indel of 6 bp in the last third of the sequences. For individuals exhibiting different paralogs in somehow equivalent copy numbers (additive polymorphic sites of the same height) the two different ribotypes were determined through careful base subtraction (based on [32]), by using clear sequence chromatograms of geographically neighboured accessions as reference sequences. In some plants solely the predominant ribotype was extracted. All ITS sequences deriving from heterogeneous individuals were labelled as such and were handled with special attention in the subsequent data analysis. Within the individuals of white-flowering Cistus species no sequence characteristics or sequencing artefacts were detected that would indicate a regular simultaneous presence of ITS variants. All sequences were deposited in GenBank (accession numbers are provided in Supplementary Table S1). For accessions with intra-individual ITS polymorphism, two accession numbers per plant individual were assigned.

Sequence Analysis and Phylogenetic Reconstruction
ITS, trnL-trnF, and rpl32-trnL sequences were aligned using Geneious 5.3.6, with subsequent manual correction. Variable positions in the data matrices were checked against the original sequence chromatogram files. NrITS ribotypes and cpDNA chlorotypes were defined according to obvious patterns in the condensed alignments and sequences were grouped and assigned accordingly. The distribution of nrITS-ribotypes, trnL-trnF chlorotypes, and rpl32-trnL chlorotypes was then mapped to visualize phylogeographical distribution patterns among populations. Distribution maps were compiled with Google Earth (https://www.google.com/earth/download/), with subsequent modifications.

cpDNA Variability
The trnL-trnF alignment representing the two target species C. creticus and C. albidus and their hybrid C. × canescens consisted of 157 sequences (125/30/2) and 427 bp. Eight nucleotide substitutions and three indels (two small indels related to poly-T stretches, one indel of four bp) were detected. When neglecting the poly-T stretch related indels, the remaining polymorphic sites identified five different trnL-trnF chlorotypes (variable trnL-trnF alignment positions are summarized in Table 1; geographical chlorotype distribution is displayed in Figure 1a). The frequent chlorotype A was shared by C. creticus and C. albidus (Figure 1a). Within C. albidus it was the only trnL-trnF chlorotype present in the 15 populations. Within C. creticus chlorotype A was not present in populations close to the natural distribution of C. albidus but was the dominant one in geographically distant Eastern Mediterranean areas (Greece, Turkey, Cyprus, and the Near East). Chlorotype B (a rare variant of chlorotype A) was detected in one Turkish population. The three trnL-trnF chlorotypes C, D and E were specific for C. creticus. Chlorotype C was present in the three, geographically isolated Spanish populations. Chlorotype D was distributed in Italy, Albania, Western Greece, and the Crimean Peninsula and showed sporadic presence in Croatia, Turkey and Cyprus. Chlorotype E was the main chlorotype of the Croatian populations. Of the two Spanish accessions of the hybrid of C. creticus and C. albidus (C. × canescens), one exhibited chlorotype A, the other one chlorotype C.     The rpl32-trnL alignment representing C. creticus, C. albidus and C. × canescens collected during this investigation consisted of 141 sequences (113/26/2) and 899 bp. Five nucleotide substitutions identified four rpl32-trnL chlorotypes (Table 2, Figure 2a). Chlorotype A was shared by C. creticus and C. albidus (Table 2, Figure 2a) and was detected in most plants exhibiting trnL-trnF chlorotype A (Figures 1a and 2a). Local sequence variants of the rpl32-trnL chlorotype A were present in two Israeli and one Jordanian accession (chlorotype B) as well as one Greek accession (chlorotype C). Chlorotype D was specific to C. creticus and was present in Spain, Italy, Croatia, Albania, Western Greece, the Crimean Peninsula and in one Cypriot and one Turkish population (plants exhibiting trnL-trnF chlorotypes C, D and E).   Table S2). The trnL-trnF haplotype network (Figure 1b) depicted chlorotype A (C. albidus and Eastern Mediterranean C. creticus) and the closely related, rare chlorotype B distinct and isolated from the further chlorotypes of C. creticus. Chlorotype A was connected via five mutations (plus one poly-T stretch related indel) and one chlorotype of C. heterophyllus to chlorotype C, the chlorotype of the three Spanish relict populations of C. creticus. During this investigation, chlorotype C was solely detected in Spanish C. creticus but three GenBank accessions indicated that chlorotype C might be interspecifically shared with C. albidus and C. heterophyllus. The relationship of chlorotypes C, D, and E with the three further C. heterophyllus chlorotypes appeared to be close and alternative relationships between these chlorotypes and to chlorotypes of C. ocreatus and C. osbeckiifolius were indicated. Chlorotype C of the Spanish populations was (via C. ocreatus) connected to the chlorotypes of five other purple-flowered Cistus species (except for C. crispus, all endemic to the Canary Islands). Via one substitution (plus one poly-T stretch related indel and one 4 bp indel), chlorotype C was linked to chlorotype D, the immediate ancestor of chlorotype E (one mutation plus one poly-T stretch related indel in two accessions) and of the very distinct impressive haplotype diversity of Corsica and Sardinia, where 19 accessions exhibited twelve chlorotypes (characterized by combinations of 18 unique nucleotide polymorphisms; Supplementary Table S2). The white-flowered clade was connected via C. ocreatus and C. ladanifer.
The rpl32-trnL alignment representing C. creticus, C. albidus and C. × canescens collected during this investigation consisted of 141 sequences (113/26/2) and 899 bp. Five nucleotide substitutions identified four rpl32-trnL chlorotypes (Table 2, Figure 2a). Chlorotype A was shared by C. creticus and C. albidus (Table 2, Figure 2a) and was detected in most plants exhibiting trnL-trnF chlorotype A (Figures 1a and 2a). Local sequence variants of the rpl32-trnL chlorotype A were present in two Israeli and one Jordanian accession (chlorotype B) as well as one Greek accession (chlorotype C). Chlorotype D was specific to C. creticus and was present in Spain, Italy, Croatia, Albania, Western Greece, the Crimean Peninsula and in one Cypriot and one Turkish population (plants exhibiting trnL-trnF chlorotypes C, D and E). Table 2. Variable sites and chlorotype (CT) designation in the rpl32-trnL intergenic spacer of C. creticus (cre), C. albidus (alb) and C. × canescens (can). Only chlorotypes of plants collected during this investigation were included. Alignment positions and consensus sequence are indicated in the header. When including additional sequences (GenBank sequences and sequences created by authors) of C. creticus (19), purple-flowered C. crispus (2) and white-flowered C. ladanifer (2), C. monspeliensis (6), C. parviflorus (4), and C. salviifolius (13), the rpl32-trnL alignment consisted of 188 sequences and 916 bp. Seventy-one nucleotide substitutions and eight indels identified 23 rpl32-trnL chlorotypes (Figure 2b; Supplementary Table S3). In the rpl32-trnL haplotype network (Figure 2b) chlorotype A (immediate ancestor of the local variants B and C) is connected via three substitutions to chlorotype D. Chlorotype D functions as a link to the third included species of the purple-flowered clade, C. crispus and was identified as ancestor of the remarkable Corsican/Sardinian haplotype diversity (14 haplotypes from 19 individual plants, defined by 21 substitutions specific for the island populations). The white-flowered clade is linked via C. salviifolius and, in the absence of sequences from further, purple-flowered species, C. crispus.   Figure 2b) or indicate missing data (dark grey). (b) rpl32-trnL chlorotype network constructed from the extended sequence alignment (Table S3, Supplementary Materials). Coloured circles represent C. creticus and C. albidus chlorotypes detected during this investigation. Grey circles indicate chlorotypes from additional sequences of the enlarged alignment [GenBank sequence data (*) and the sequences of C. crispus and white-flowered Cistus species created by authors]. Light grey lines indicate alternative relationships. cre = C. creticus, cri = C. crispus, lad = C. ladanifer, mon = C. monspeliensis, par = C. parviflorus, sal = C. salviifolius.

nrITS Variability
One hundred and fifty-seven individual plants of C. creticus, C. albidus and C. × canescens (125/30/2) were directly sequenced for ITS. From 32 (22/8/2) accessions no usable ITS sequence could be obtained, in most cases due to competing fungal (e.g., Aureobasidium sp.) DNA. Among them were also the sequences of the two available hybrid individuals and no conclusion about ITS sequence additivity in putative F1 individuals could be drawn. The bulk of the 125 successfully sequenced accessions exhibited clear sequence chromatograms and was uncomplicated to edit. Twenty-eight direct sequences of C. creticus (25 accessions, 27% of the successfully sequence plants) and C. albidus (3 accessions, 1% of the successfully sequence plants), however, showed very particular characteristics, single additive polymorphic sites in ITS1 and ITS2 and a short series of ambiguous bases close to the ending of ITS2. A comparison of such ambiguous sequence chromatograms with unambiguous ones suggested the presence of intragenomic ITS variants of different lengths. The stable patterns of additive polymorphic sites and shifted nucleotides could usually be fully explained by the simultaneous occurrence of two ITS sequence variants: one ITS variant continuously present in the whole sampled area and one of solely regionally occurring ITS ribotypes. For those heterozygous plants exhibiting two different ribotypes in equivalent copy numbers, the two ITS variants were extracted by careful base subtraction and using sequences from geographically close homozygous individuals as references. For some heterozygous plants, where one ITS variant was underrepresented, the dominant ITS ribotype was extracted. Such extracted sequences of heterozygous plants were labelled and handled with special care in the subsequent analysis and result interpretation.
The ITS alignment representing the accessions of the target species C. creticus, C. albidus and C. × canescens collected during this investigation consisted of 143 sequences (118/25/0, with 38/6/0 of them derived from heterozygous plants) and 699 bp. Twenty-nine ITS ribotypes were defined based on 25 variable nucleotide positions and a 6 bp indel. One hundred and nine ITS sequences (51/13/18/27) were ascribed to one of the four main ribotypes A, B, C or D (Table 3). These four main ribotypes, each, are well backed by the high number of clear direct ITS sequence chromatograms. Thirty ITS sequences (including eight sequences that were isolated solely from heterozygous plants) exhibited the sequence characteristics, defining one main ribotype in combination with few additional individual SNPs and/or additive polymorphic sites that often seem to depict a transition between two main ribotypes. Five ITS variants from heterozygous plants exhibited individual SNPs (ribotypes E/F). Table 3. Variable sites and ribotype (RT) designation in the ITS alignment of C. creticus (cre) and C. albidus (alb). Solely sequence ribotypes detected during this investigation are included. Alignment position and consensus sequence is indicated in the header. Asterisks refer to ribotype sequences extracted solely from heterozygous plants. • = Insertion (CGTCCT, alignment position bp 569 to bp 574) that discriminates ribotype A (and its variants) from all further ribotypes.  Within C. creticus ribotype A was frequent and continuously present in the whole sampled area, with by trend stronger or even exclusive appearance in populations of the Eastern Mediterranean area. Nine local variants of ribotype A (A1-A9, Figure 3b), differing by one or two additional nucleotide substitutions or single additive polymorphic sites, were present in Italy, Albania, Greece, and Cyprus. Within C. albidus, ribotype A was present in three heterozygous plants of France and Italy, at the eastern margins of the species distribution. The regular ITS variant of C. albidus was ribotype B, inter-specific variants (B1/B2, Figure 3b) with one additional additive polymorphic site/nucleotide polymorphism were present from Portugal to France. Within C. creticus, the distribution of ribotype A overlapped with that of further species-specific ribotypes (C, D, E, F, and G; Figure 3b). Between Spain and Greece, C. creticus plants were either homozygous for ribotype A or homozygous for the regionally present co-occurring ribotype, or they were heterozygous and exhibited both, ribotype A and the regionally present second ribotype. Ribotype C was co-occurring in Albanian and Greek populations, ten rare local variants of ribotype C seem to constitute intermediate sequence variants between ribotype C and ribotype B and/or exhibited single additional mutations. Ribotype D was locally distributed in Italian and Croatian populations, and its three rare variants exhibited one additional individual mutation or seemed to constitute the intermediate sequence variant between ribotype D and ribotype B. The rare Albanian ribotype E (and its variant E1) seemed to constitute an early intermediate between ITS variant B and C that gained one additional mutation. Ribotype F, exclusively distributed in Cyprus, exhibited sequence characteristics that were not present in any other C. creticus ribotype.

RT Species
When including additional sequences (GenBank sequences and sequences created by authors) of C. creticus (2), C. albidus (3), purple-flowered C. chinamadensis (2), C. crispus (2), C. heterophyllus (1), C. osbeckiifolius (2), C. symphytifolius (2), and white-flowered C. ladanifer (2), C. monspeliensis (6), C. parviflorus (4), and C. salviifolius (11), the final ITS alignment consisted of 180 sequences and 705 bp. A total of 57 distinct ITS sequence variants were defined based on nucleotide polymorphisms at 101 variable alignment positions and allover 6 indels (Supplementary Table S4). A minimum spanning network (Figure 3c) was calculated to depict ribotype relationships. The widespread ribotype A (plus its local rare variants) was connected via ribotype B, the ribotype specific for C. albidus and the supposed common ancestor of the further main ribotypes of C. creticus, the Italian/Croatian lineage D (plus its variants), the Albanian/Greek lineage C (plus its variants) and the rare Albanian lineage E/E1. The distinct Cypriot ribotype F, although devoid of the characteristic indel, was linked to ribotype A. The five additional species of the purpleflowered clade were connected via the one ITS variant of C. heterophyllus and one rare ITS variant of C. creticus designated to ITS lineage C. The four white-flowering species were split up. Cistus monspeliensis and C. parviflorus were connected via C. creticus ribotype D.
Cistus ladanifer and C. salviifolius were connected via C. heterophyllus that also linked the further, purple-flowered species.    Table 3. (c) Minimum spanning network based on the extended ITS sequence alignment (Prevosti's absolute genetic distance). Coloured circles represent the ribotypes of the C. creticus-C. albidus complex detected during this investigation (circle colours reflect those used in Figure 3a,b to display ribotype distribution). Grey circles represent ribotypes of different Cistus species or ribotypes based on GenBank sequence information (labelled with *). Circle areas reflect the sample sizes. Edge widths and shading represent relatedness.

Intragenomic ITS Diversity in C. creticus
Cistus albidus appeared as a quite homogenous species, characterized by a single chlorotype and little ribotype variability. Cistus creticus was remarkably diverse, with distinct chloro-and ribotypes and clear evidence for intraindividual ITS polymorphism in about 27% of the successfully sequenced accessions. The presence of intragenomic ITS variants in C. creticus has not explicitly been described before but GenBank sequences ending after the position of the 6 bp indel (starting point of continuously shifted bases in the direct sequence chromatogram) could indicate that previous investigations included consensus sequences from heterozygous accessions. The ITS ribotypes described here were not verified by cloning and the rare ITS variants need confirmation. However, the four frequent ribotypes (A, B, C, D) are confirmed by a bulk of clear direct sequence chromatograms. For most heterozygous accessions of C. creticus, base ambiguities, and base shifts in their direct ITS sequence chromatograms could be fully elucidated by simultaneous occurrence of the continuously widespread ribotype A and one of the two further, regionally distributed main ribotypes (C and D). These stable patterns strongly suggest the presence of persistent paralogous ribosomal loci. It was hypothesized that the location of ribosomal loci on different chromosomes might have a preventative effect against homogenization through concerted evolution [42]. A detailed investigation of 45S rDNA site number and location in Cistaceae, however, provided no evidence for the presence of more than one ribosomal locus within C. creticus [43]. Regarding the purple-flowered Cistus clade, the results of [43] indicated that, except for the basal C. crispus, all purple-flowered species diversified from a common ancestral lineage with one single rDNA site. For their investigation [13] analysed two C. creticus accessions that were also present in our study (SC19 and SC23). Both populations appeared heterogeneous, with one plant that resulted in a clear ITS sequence chromatogram and a second one that showed clear evidence for intragenomic ITS variation. Probably, rDNA site number in C. creticus was underestimated when [13] picked two homogenous accessions by chance.

Geographical Distribution of Genetic Variation
Based on the geographical patterns of plastid and ribosomal sequence variation, three main evolutionary entities can be discerned in the C. albidus-C. creticus complex: -Western Mediterranean distributed C. albidus that appeared genetically homogeneous and was typically characterized by trnL-trnF chlorotype A and ribotype B (or its rarer variants). -Eastern Mediterranean distributed C. creticus, below referred to as Eastern Mediterranean clade (EM clade), that was genetically homogeneous and exhibited trnL-trnF chlorotype A (or its rarer variants) and ribotype A (or its rarer variants). -Western and Central Mediterranean distributed C. creticus, below referred to as Western Mediterranean clade (WM clade), that appeared conspicuously heterogeneous and was characterized by trnL-trnF chlorotype B, C, or E and dissimilar co-occurring, putative paralogues ribotypes (typically ribotype A together with one regionally distributed ribotype). Based on the regionally distributed ITS paralogs, two large subentities within the WM clade became obvious. One including the populations from Spain, Italy, and Croatia (predominantly characterized by ribotype D) and the second one including populations from Albania and Greece (predominantly characterized by ribotype C).
The four species of subgenus Cistus were described to freely hybridize with each other [7,17,33]. The here observed distribution patterns of chloro-and ribotype variants indicated an intertaxon genetic exchange in two geographical regions. The presence of ribotype A in two native populations of France and an Italian C. albidus Seedbank accession might indicate gene flow from C. creticus into C. albidus. The distribution of C. albidus and C. creticus was described to overlap in this geographical region [11]. However, during our plant collection along the Ligurian and Tyrrhenian Sea, no sites with close co-occurring C. creticus and C. albidus were located. The presence of a C. creticus specific ribotype in C. albidus could therewith also be a relic of a common ancestor species that was preserved in the easternmost populations of C. albidus. In parts of mainland Greece and the Peloponnesus there was evidence for genetic exchange between the WM and EM clade of C. creticus. Populations from the provinces Central and East Macedonia in Greece exhibited trnL-trnF chlorotype A that was attributed to the EM clade together with diverse rare variants of ribotype C, characteristic for the WM clade. Additionally, the elevated number of individual ITS sequence characteristics and rare ribotypes in Greek populations might be indicative of a higher genetic diversity when the characteristics of the two lineages run into each other.

A Hypothesis about Evolutionary Relationships in the C. albidus-C. creticus Complex and Biogeographic Considerations
Molecular clock calculations have dated the last common ancestor of the closely related C. albidus, C. creticus and C. heterophyllus with 0.19 (0.33-0.01) Myr to the Chibanian age. The diversification point of C. albidus and the WM clade of C. creticus (the in Albania/Greece distributed sub-entity) was dated to 0.04 (0.13-0.002) Myr [10] and therewith as a rather recent event that, when emanating from the proposed 0.04 Myr, for example, would coincide with the arrival of the anatomically modern humans in Europe.
Based on earlier hypotheses [8,10,11] and the distribution of genetic variation observed here, we act on the assumption of a primordial, widely distributed Cistus species whose continuous distribution in Europe has split sometime during the Pleistocene. Cistus albidus and Eastern Mediterranean C. creticus sharing chlorotype A might represent the progeny of peripheral relict populations [44,45], with C. albidus as a direct descendant in the West and the EM clade of C. creticus as the corresponding counterpart in the East. In the case of Eastern Mediterranean C. creticus, this hypothesis seems to be confirmed by its genetic homogeneity and its disjunct distribution apart from its closest genetic relative and the diversity centre of the genus. Due to reproductive isolation (beside C. creticus, solely three Cistus species from the white-and whitish-pink-flowered Cistus clade are native to Turkey and the Near East) and short diversification time, Eastern Mediterranean C. creticus could have kept its original characteristics.
In the Western Mediterranean, the situation appears to be more complex. Spain and Morocco with twelve Cistus species each, constitute the actual diversity centre of the genus. Four of the nine purple-flowered Cistus species, the four members of subgenus Cistus, are native to the Iberian Peninsula and Morocco. Among them, C. crispus (distributed from North Africa to Italy) is distinct whereas C. albidus and the in Spain exceedingly rare C. creticus and C. heterophyllus (distributed in North Africa and rarely in Spain) appear to be nearer associated [11]. The close relationship of the latter three species can conclusively be explained by recent or rapid diversification from a common ancestor but might furthermore be rooted in reticulation/hybridization. European C. albidus appeared genetically homogeneous and exhibited no clear evidence for a hybridogeneous speciation. In C. × clausonii nothosubsp. crespoi P.P. Ferrer and E. Laguna, the one hybrid of C. heterophyllus × C. albidus native in Spain, chloroplast heteroplasmy was revealed that was hypothesized to indicate ancient hybridization with an unidentified Cistus taxon [33]. Within the WM clade of C. creticus the simultaneous occurrence of distantly related ribotypes and the interspecifically shared trnL-trnF chlorotype C strongly indicate hybridogeneous (reticulate) speciation. From the observed genetic characteristics, a specific contribution of potentially involved taxa could not clearly be identified. The in C. creticus continuously distributed ribotype A was attributed to an ancient widespread ancestor species or immediate descendants thereof, based on the dominant appearance of this ribotype in the EM clade. The regionally co-occurring ITS paralogs appear to be a younger characteristic that seems to have derived from the main ribotype of C. albidus. The chlorotypes of Spanish and Central Mediterranean C. creticus are (according to Genbank sequences) either shared with C. albidus and C. heterophyllus or are strongly associated with further chlorotypes of C. heterophyllus. Beside the cryptic combination of genetic characteristics, Western Mediterranean C. creticus exhibits furthermore a conspicuous phytochemical specialty. The species of subgenus Cistus are usually characterized by flavonols [24,25]. Aside from these character compounds, populations of the WM clade, especially the one sub-entity present in Spain, Italy and Croatia, exhibit higher amounts of punicalagin derivatives [25], a chemical characteristic that is unique in this form in subgenus Cistus and is usually associated with white-flowered Cistus species [24,25,46]. The spectrum of non-volatile compounds of Cistus species was suspected to be strongly related to evolutionary events [24,25]. The specific phytochemical characteristic of Western Mediterranean C. creticus could constitute the preservation of a very ancient, primordial feature or could indicate hybridization and a therewith (re-)acquired phytochemical trait. The regionally distributed ribotypes of Western Mediterranean C. creticus seem to have also preserved ancient or acquired nucleotide polymorphisms. Ribotypes C and D are characterized by nucleotide polymorphisms that are not present in C. albidus and Eastern Mediterranean C. creticus but are shared with C. heterophyllus and white-flowered C. salviifolius (ribotype C) or C. monspeliensis (ribotype D). Interestingly, also in the EM clade, a distinctive phytochemical trait was supplemented by specific local ITS characteristics. Cypriot C. creticus exhibited also high amounts of punicalagin derivatives whereas these compounds seem to be of marginal importance in populations from Eastern Mediterranean mainland localities [25]. Two of the Cypriot accessions exhibited nucleotide polymorphisms that were not present in any other Cistus sample or were shared with C. crispus, C. heterophyllus and the white-flowered species. It is conceivable that populations on the biogeographically isolated island of Cyprus have rather preserved ancient traits than continental mainland populations. Cypriot populations might be a rewarding object of study for deeper insights into the evolutionary processes within subgenus Cistus.
The geographical origin of the WM clade was located to the Iberian Peninsula as this is the region where C. creticus and its possible progenitors or hybridization partners C. albidus and C. heterophyllus occur geographically close. Support comes also from [8] and [10] who concluded that dispersal and colonization of Cistus species across the Mediterranean basin has taken place after divergence and species formation in its West. When regarding chloroplast data, especially the slightly more informative trnL-trnF data, these hypotheses seem to apply to the observed distribution patterns. The migration movement of the WM clade of C. creticus appears to be a continuous one, starting with the basic chlorotype C in Spain and continuing eastwards with chlorotype D to Corsica/Sardinia, Italy, Albania and Greece and chlorotype E from Italy to Croatia. In sharp contrast to this, the ribotype distribution in Western Mediterranean C. creticus is evidently a discontinuous one. The strictly regionally distributed ribotypes C and D share none of their characteristic variable nucleotide positions and both ribotypes appear to have originated separately from the basic ribotype of C. albidus. The decreasing relevance of punicalagin derivatives in C. creticus populations eastwards of Albania [25], underlines the genetic divergence between the two sub-entities of the WM clade. Based on these findings, it could be hypothesized that formation or range expansion of the Western Mediterranean clade was a repeated event. Ribotype C and ribotype D may represent consecutive migration waves out of refugia on the Iberian Peninsula that followed consecutive cooling periods or are coincided with massive regional ecological changes as e.g., could have been caused by a super-eruption at the Phlegrean fields that buried wide areas of South-East Europe under massive volcanic ash deposits about 0.04 Ma ago [47,48]. Alternatively, ribotype C and ribotype D might represent the progeny of geographically distinct relict populations that have re-expanded from different south-European refugia (Sicily, Peloponnese). Very recently then, increasing anthropogenic influence on ecosystems of coastal near areas might have established favourable conditions for a further range expansion. On the Istrian peninsula (Croatia), human impact on coastal ecosystems began about 5000 years ago and was described as one of the main drivers of a long-term ecological change that led to the replacement of coastal-near woods by 'shrubland' dominated, amongst others, by Cistus [49]. Three populations of C. creticus unexpectedly exhibited genetic characteristics that did not fit well in the overall distribution patterns of chloro-and ribotypes. Plants raised from seeds that originated from the Crimean Peninsula and Cyprus, as well as one Turkish accession, exhibited ribotype D (ribotype information for the one Turkish accession is missing) and trnL-trnF chlorotype D, a combination of characteristics that was, with one exception detected in Croatia, strongly restricted to Italy. Ancient Greek settlements are well documented for both areas, the Crimean Peninsula and Cyprus. The presence of Mediterranean Flora on the Crimean Peninsula appears to be the result of a complex process involving biogeographic processes and human-aided introduction of species [50]. When considering the disjunct presence of genetic characteristics, it can be hypothesized that emigrants from the historical Magna Graecia, actively or passively, brought along plant material or seeds from the South-Italian coast.

Concluding Remarks
Cistus albidus and C. creticus can be clearly differentiated and identified when combining sequence information from nuclear ITS and chloroplast trnL-trnF and considering the intraspecific genetic diversity of C. creticus. Genetically, C. creticus was confirmed as a heterogeneous species that integrates two major evolutionary lineages with clearly different genetic characteristics. The EM clade of C. creticus seems to represent old and ancestral characteristics. The genetically more complex WM clade seems to be the result of hybridization and recurrent or parallel migration movements. The genetic disparity among EM and WM clade appears to be reflected by differences in the non-volatile compound composition but could not clearly be associated with consistent macromorphological tendencies towards one of the two lineages. Traits such as the absence/presence or density of non-glandular and glandular trichomes or leaf characteristics varied highly within and between natural populations of Spain, Italy, Croatia, Albania, and Cyprus. This high morphological diversity is not surprising when considering the high level of genetic diversity within the WM clade and the indicated complex evolutionary history, but this impedes the identification of relevant discriminating characteristics. Potentially, thorough multivariate analysis of trait variation can aid identification of relevant distinctive characters. Based on the nuclear and the two chloroplast markers used, besides the inter-regional disparity of genetic characteristics, within C. creticus, there was no evident correlation between DNA characteristics and the designation of plant material to varieties or subspecies based on morphological characteristics. These results are in accordance with a previous investigation that related small scale morphological diversity more to local ecological factors (as e.g., bedrock) than to characteristics of chloroplast markers [16]. Based on these results, one could debate whether the currently often applied species concept of C. creticus with its three recognized subspecies creticus, corsicus and eriocephalus [13] must be better reworked. Our results argue for the recognition of two (genetically but not morphologically distinguishable) subspecies within C. creticus corresponding to the two major evolutionary lineages, here referred to as WM clade (C. creticus subsp. hesperius) and EM clade (C. creticus subsp. orientalis). As some Greek accessions with combined genetic characteristics indicated genetic exchange, this subdivision into two subspecies seems more adequate than a two-species concept. The three morphological and (regarding volatile content and composition) chemotypic variants presently treated as subspecies are hard to grasp. [15] were able to differentiate between C. creticus ssp. corsicus and subsp. eriocephalus from Corsica/Sardinia based on volatiles and ISSR markers, but their results are difficult to integrate as accessions from these islands appear as very distinct group in our chlorotype networks. As both densely glandular and viscid plants (presently recognized as subsp. corsicus and subsp. creticus) and much less or not glandular but conspicuously pubescent plants (presently recognized as subsp. eriocephalus) were described within both evolutionary lineages, we would propose to downgrade their taxonomic status from subspecies to form level.