Bayesian 50% majority-rule consensus tree from the ITS dataset, with posterior probability values (PP) and boostrap support (BS: in brackets and italics) shown near statistically supported nodes; the main clades of Lithospermeae are indicated with small squares and letters according to Cecchi & Selvi (2009). 

Context in source publication

Context 1

... 2013 on the Taurus chain (S Turkey), in a first site not far from the type locality around Gazipa ş a (Gökbel plateau; Fig. 1A–C), and in a second one ca. 35 km to the east (Karahasan pass between the villages of Ermenek and Taskent). Geographical details on these sites are given in Appendix 1, together with the full list of examined taxa and voucher information. Additional material of Arnebia species for morphological observations was obtained from herbarium collections mainly in FI, FI-W, B and G. DNA extraction and amplification— Genomic DNA of the new Arnebia accessions was extracted from silica- gel dried samples of leaf tissue using a modified 2xCTAB protocol (Doyle & Doyle 1990). Amplification of the ITS region of nuclear DNA, including ITS1, 5.8S and ITS2, and of trnL-trnF IGS followed the procedure described in Cecchi et al. (2014). Automated DNA sequencing was performed directly from the purified PCR products using BigDye Terminator v.2 chemistry and an ABI310 sequencer (PE-Applied Biosystems, Norwalk, CT, USA). Sequence alignment and phylogenetic analyses— Original sequences of the Arnebia accessions were treated as described in Cecchi et al . (2014). Three datasets were prepared for phylogenetic analyses, ITS, IGS and combined ITS-IGS, retrieving most of sequences from INSDC (accession numbers are given in Appendix 1). The ITS1-5.8S- ITS2 dataset included 46 ingroup taxa, of which 13 accessions (11 species) of Arnebia s.l. (including Huynhia ) and the others representing the great majority of old-world Lithospermeae (17 genera). The trnL-trnF IGS dataset included 28 ingroup taxa, of which ten accessions (nine species) of Arnebia s.l. and a wide range of old-world Lithospermeae involved (13 genera). Different sample size of the two datasets was because IGS sequences were available for fewer taxa than for ITS, but this did not apparently cause inconsistencies between the resulting phylogenies also due to the low resolution power of IGS. Gaps were coded as separate characters according to Simmons & Ochoterena (2000) using FastGap v.1.0.8 (Borchsenius, 2009), and appended at the end of the datasets. An additional dataset consisting of concatenated ITS-IGS sequences plus coded gaps was also prepared for a combined analysis (25 ingroup taxa). Congruence between the two single-marker datasets and respective trees was inferred from the absence of conflicting well-supported clades in the resulting trees, according to Wiens (1998). Four taxa of the tribes Echiochileae and Boragineae were selected as outgroup members, based on their relationships to Lithospermeae (Långström & Chase 2002; Weigend et al. 2013). Phylogenetic analyses were performed using Maximum Parsimony and Bayesian methods. Tree construction was first performed using PAUP 4.0 (Swofford 2000), running Heuristic searches with “tree-bisection-reconnection” (TBR) branch-swapping with accelerated transformation (ACCTRAN) optimisation to infer branch (edge) lengths; MULTREES option on, ADDSEQ = random, twenty randomised replicates. All characters were weighted equally, and character state transitions were treated as unordered. Bootstrap support for clades was obtained performing a heuristic search with 1.000 replicates, using TBR branch-swapping, 10 random taxon entries per replicate and MULTREES option on. The data sets were also analysed using Bayesian inference of phylogeny with MrBayes 3.1.2 (Ronquist & Huelsenbeck 2003). Based on jModeltest (Posada 2008), the best fitting models of nucleotide substitution were GTR for ITS, with gamma-distributed rate variation across sites, and GTR + I + Γ for trnL-trnF IGS. The analyses were performed using four incrementally heated Markov chains (one cold, three heated) simultaneously started from random trees, and run for one million cycles sampling a tree every ten generations. The stationary phase was reached when the average standard deviation of split frequencies reached 0.01. Trees that preceded the stabilization of the likelihood value (the burn-in) were discarded, and the remaining trees were used to calculate a majority-rule consensus phylogram. The trees were viewed and edited with TreeView (Page 1996), with indication of Bayesian Posterior Probabilities (PP) values for the internal tree nodes. Micromorphology (SEM)— Pollen grains from dry specimens were rehydrated in a solution of Aerosol-OT 20% and then observed with a FEI ESEM-QUANTA 200 working at 30 kV. Nuclear ITS-5.8S dataset— The aligned matrix included a total of 805 positions, with coded gaps in pos. 683~805. In the Maximum Parsimony analysis, 267 characters were constant, 177 variable but non-informative, and 361 variable and parsimony informative. The heuristic search produced 54 most parsimonious trees with L = 1662, Consistency index (CI) = 0.51 and Retention index (RI) = 0.67. The strict consensus was topologically largely congruent with the 50 majority-rule consensus tree produced by the Bayesian analysis, which is shown in Fig. 2 with bootstrap (BS) and posterior probability values (PP). Arnebia was not retrieved as a monophyletic clade because of the position of A. purpurea as sister to H. pulchra . The two species clustered together in a monophyletic clade with good bootstrap (87% BS) and Bayesian (0.98 PP) support. They shared five SNPS and one 2-bp positions in the ITS1 region that were not present in anyone of the other species of Arnebia . This clade was suggested as sister (0.71 PP) to a well-supported assemblage of genera such as Cerinthe , Neatostema , Mairetis and others (clade B), but this relationship did not receive bootstrap support. All other members of Arnebia were retrieved in a clade with moderate support from Bayesian inference (0.89 PP), but not from boostrap analysis. This consisted of two main sister branches: the first one (0.96 PP) included the annual members A. linearifolia , A. decumbens , A. coerulea and A. tubata , although the affinity of the latter species to this group did not receive bootstrap support; the second one was well supported (91% BS, 0.99 PP) and formed by the perennial species partly referred to genus Macrotomia in past times, such as A. densiflora , A. benthamii and A. euchroma . The other well-supported groups were those of Onosma/Maharanga/Echium (clade A), Moltkia (clade D) Alkanna/ Podonosma (clade F) and Lithospermum/Glandora/Buglossoides/Aegonychon (clade C). Plastid trnL-trnF IGS dataset— The IGS alignment included a total of 995 positions, of which 72 coded gaps at the end of the matrix. Constant characters were 695, and 117 characters were parsimony informative. Heuristic search yielded 40 most parsimonious trees with L=426, CI=0.79 and RI=0.70. The 50-majority rule bootstrap tree and the consensus phylogram from Bayesian analysis were both strongly polytomized and therefore not shown here. Only the annual species of Arnebia formed a well supported clade (100% BS, 1.00 PP), and their sistership to A. guttata received very weak support (67% BS, 0.51 PP). Arnebia purpurea and H. pulchra shared a single SNP but were unresolved, as well as A. szechenyi Kanitz and the clade of A. benthamii/A.euchroma (57% BS, 0.99 PP). Combined ITS-IGS dataset— The aligned matrix consisted of 1783 characters, including coded gaps (pos. 1610~1783), of which 1018 were constant and 407 parsimony-informative. Maximum parsimony analysis retrieved only two trees with L=1716; CI=0.61 and RI=0.56, the consensus of which (Fig. 3b) was largely congruent with the 50% majority rule consensus tree from Bayesian analysis (Fig. 3a). Arnebia pupurea and H. pulchra clustered again in the same clade (78% BS; 0.90 PP), that was sister to clade B (inclusive of species of the genera Neatostema, Lithodora, Mairetis, Cerinthe ) but with low support (0.72 PP). In the MP and Bayesian trees, the rest of Arnebia was included in a single clade, though without strong support (59% BS, 0.89 PP). Both analyses confirmed the early split of this group in two well supported subclades, a first one with the annual species A. linearifolia, A. tubata and A. decumbens (96% BS, 1.00 PP) and a second one with the perennials A. szechenyi, A. guttata, A. benthamii and A. euchroma (87% BS, 1.00 PP). Morphology— General and detailed characters of vegetative parts, flowers and fruits of Huynhia and Arnebia are already well known from various literature descriptions, while only two standard descriptions are available for A. purpurea . Herbarium material of this species is very scarce. Here, two characters are worth of mention because of their taxonomic significance, one concerning the androecial arrangement and one the stigma structure. In Huynhia and A. purpurea the epipetalous stamens are inserted through very short filaments at different heights within the corolla tube, e.g. three higher at the throat and two lower down in the tube (Fig. 4C,D). This character is less evident in A. purpurea , where the anthers of the two lower stamens reach ca. half of the anther length of the three higher stamens, rather than less than half in Huynhia . Based on our observations and current knowledge, the stamens in Arnebia (incl. Macrotomia ) are whorled in the upper part of the corolla tube (Fig. 4A,B). Also, the style of most Arnebia s.l. species is shortly 2- or 4-forked, with each branch ending in a stigmatic portion (Fig. 4A), while it is entire and with capitate, shallowly bilobed stigma in Huynhia and A. purpurea (Fig. 4C,D). An intermediate condition is found in A. densiflora , where the cleft dividing the two lobes of the stigma reaches the style only shallowly (Fig. 4B). Heterostyly was not observed in A. purpurea , though it frequently occurs in Arnebia s.l. Pollen characters provide other elements of taxonomic significance and these are summarized in Table 1. Arnebia purpurea , here investigated for the first time, turned out to be palynologically closer to Huynhia than to Arnebia (incl. Macrotomia ). The key character shared by the two ...