Relationships within Mcneillia Indicate a Complex Evolutionary History and Reveal a New Species of Minuartiella (Caryophyllaceae, Alsinoideae)

The genus Mcneillia has been recently segregated from Minuartia L. based on molecular results, also supported by morphology. However, to date, a comprehensive study on the phylogenetic relationships within this genus is lacking. In this paper, we provide a multigene phylogeny of all the species and subspecies of Mcneillia employing two nuclear and six chloroplast markers. We documented extensive gene flow between taxa, sometimes separated at specific rank. In addition, Mcneillia as currently circumscribed, is not monophyletic. In fact, Mcneillia graminifolia subsp. brachypetala, strictly endemic to Greece, truly belongs to Minuartiella, a genus otherwise limited to South-West Asia. Moreover, even after removal of this taxon, our results do not support the monophyly of the taxa included in M. graminifolia s.l., the most variable and widespread species of the genus. Further controversial subspecies of Mcneillia graminifolia, i.e., subsp. hungarica and subsp. rosanoi, are shown to deserve taxonomic recognition as separate species, whereas Mc. moraldoi is not distinct at specific rank. In addition, Mc. saxifraga subsp. tmolea is here regarded as a further distinct species. A consistent taxonomic treatment is therefore proposed with six new combinations and nomenclatural notes, providing the necessary typifications.

Dillenberger and Kadereit [26] documented rampant polyphyly in Minuartia as traditionally conceived, revealing that the main diagnostic characters for the genus (i.e., presence of three styles and three fruit valves) are indeed plesiomorphic. Even subgeneric ranks (e.g., Minuartia subg. Minuartia) were often non-monophyletic and included clades related to other genera. As a consequence, the said authors limited Minuartia to sects. Plurinerviae McNeill and Minuartia, transferring the other species to different genera, monophyletic as The systematics of the group, however, are far from being entirely accepted. A synopsis of the most relevant taxonomic treatments is provided in Table 1. In addition, the relationships between the taxa are still largely speculative (e.g., [27]), and the phylogeny almost unknown; as at present only four species (mostly with single The systematics of the group, however, are far from being entirely accepted. A synopsis of the most relevant taxonomic treatments is provided in Table 1. In addition, the relationships between the taxa are still largely speculative (e.g., [27]), and the phylogeny almost unknown; as at present only four species (mostly with single specimens) have been investigated by molecular methods [26]. In this study, we infer nuclear and plastid phylogenies of genus Mcneillia involving all its species and subspecies across their geographic distributions. This paper aims at verifying whether the genus is monophyletic, and whether the current taxonomic treatment correctly depicts phylogeny. As a contribution for the systematics of Alsinoideae, we also propose a consistent taxonomic treatment with typification of the taxa under study. Lanceolatae ser. Graminifoliae [abbr. "Mn."]. Authorships omitted for the sake of brevity. Legend: (-) not accepted, (X) not known/not treated, (1) included in the subsp. graminifolia.

ITS Phylogeny
The consensus tree from the Bayesian analysis conducted using the 290-taxa alignment of our ITS sequences, plus the ones by Dillenberger and Kadereit [26] and those by Koç et al. [35], confirmed that genus Mcneillia is not monophyletic as presently circumscribed (Figure 2a  Consensus tree from a Bayesian analysis employing ITS DNA data. Sequences are from the specimens employed by Dillenberger and Kadereit [26], Koç et al. [35], and the present study. The other 31 ITS sequences belonging to Mcneillia (Figure 2c), which encompassed both ours and those by Dillenberger and Kadereit [26], were included in a single clade (p.p. = 1). This poorly resolved group was first divided in two clades, one of which included the specimens of both subspecies of  [26] was originally labeled by them as "Minuartia graminifolia 4"; however, after examination of the voucher specimen (authors' obs.), we found that it was representative of Mc. graminifolia subsp. clandestina from Bosnia (see Figure 2c).

Nuclear and Chloroplast Phylogenies
The stepping-stone analysis aimed at testing strict and relaxed clock models for both datasets produced clear evidence in favor of the strict clock: for the nuclear dataset, the mean marginal likelihood for the relaxed clock (ln) was −3018.19, whereas the mean marginal likelihood for the strict clock was −2920.93; and for the chloroplast dataset, −5663.00 and −5504.37, respectively.
The consensus tree obtained from the Bayesian analysis of the nuclear dataset (Figure 3a), rooted by using Mn. recurva subsp. condensata, showed two monophyletic groups (each one with p.p. = 1). The first was composed by Mc. saxifraga subsp. tmolea and a branch with the two individuals of subsp. saxifraga (p.p. = 0.95). The other one included all the other taxa and branched in a first clade with the three samples of Mc. graminifolia subsp. graminifolia (p.p. = 1) and in another one with all the remaining samples (p.p. = 0.7). In the latter, a group of the two southern-most specimens of Mc. stellata was sister to all the other taxa (each clade with p.p. = 1). In this larger, not completely resolved clade, the following groups were recognized: (1) the two northern specimens of Mc. stellata

Haplotype Network
The TCS network based on chloroplast data showed the occurrence of two main haplogroups separated by 10 mutation steps: one included all the Apennine-Sicilian specimens (i.e., the Italian Mc. graminifolia subsp. clandestina, Mc. graminifolia subsp. rosanoi, and Mc. moraldoi), and the other all the remaining taxa ( Figure 4). The Italian haplotypes resulted as generally separated by one or two mutations, and the haplotypes of the specimen of Mc. moraldoi and that of Mc. graminifolia subsp. rosanoi from Sicily, both from Southern Italy, were more similar to each other than to the other individuals from Central Italy. In the other haplogroup, relationships among haplotypes were more complex: most of them originated from an ancestral haplotype from Turkey and Southern Greece, now found in both Mc. saxifraga subsp. tmolea and in one specimen of Mc. stellata; from this ancestral situation, haplotypes were sorted across (or within) the various taxa in different ways. For instance, the chloroplast haplotypes found in Mc. graminifolia subsp. graminifolia and Mc. pseudosaxifraga originated directly from this ancestral one, and presented five and two to four mutations, respectively. The Balkan specimens of Mc. graminifolia subsp. clandestina showed a chloroplast haplotype closer to that of a specimen of Mc. stellata; this latter species showed a high heterogeneity of chloroplast haplotypes ( Figure 4). The haplotypes found in Mc. graminifolia subsp. hungarica and Mc. saxifraga subsp. saxifraga originated from an ancestral haplotype, not sampled or possibly extinct; the former taxon resulted separated by the ancestral haplotype by one-two mutations. The two specimens of Mc. saxifraga subsp. saxifraga from two different localities showed very different haplotypes, each separated from the ancestral/unsampled one by three or five mutations ( Figure 4).

Haplotype Network
The TCS network based on chloroplast data showed the occurrence of two main haplogroups separated by 10 mutation steps: one included all the Apennine-Sicilian specimens (i.e.,

Discussion
One of the most remarkable results of the investigations above is that Mcneillia, as currently circumscribed, is not monophyletic. In fact, one of the subspecies of Mc. graminifolia, i.e., subsp. brachypetala, is actually a species of Minuartiella.
Concerning Mcneillia s.s. (as circumscribed just above), we observe that the chloroplast phylogeny is less resolved than the nuclear one and it exhibits two main areas of incongruence, consistent with the geographic distribution of the taxa in study. The first incongruence is the sister relationships between Mc. graminifolia subsp. graminifolia from the Italian Dolomites (not far from the borders with Slovenia and Croatia) and the Balkan representatives of the subsp. clandestina in the chloroplast tree, against the occurrence of Mc. graminifolia subsp. graminifolia in a sister relationship to all the remaining taxa (barring Mc. saxifraga s.l.) in the nuclear tree. In second place, the positions in the two trees of Mc. graminifolia subsp. hungarica (Romania) and Mc. saxifraga subsp. saxifraga (Bulgaria and Northern Greece) are notably different. A strong influence of geography is also evident from the distribution of chloroplast haplotypes in the TCS network: not only a sharp separation occurs between a western (Apennine-Sicilian) haplogroup and an eastern one, but the eastern haplogroup includes different haplotypes sorted more coherently with geography than with taxonomy. Mcneillia stellata, probably because of its fragmented distribution, has retained high chloroplast variability, partly shared with Mc. saxifraga subsp. tmolea and not homogeneously distributed within the species. In addition, we can hypothesize that the eastern haplogroup started from a probable Turkish/Southern Greek ancestor, and successively the haplotypes migrated northward, likely in different times; however, a reliable phylogeographic reconstruction of these pathways is not possible. The combination of nuclear/chloroplast incongruence and strong geographical influence on chloroplast DNA topologies is rather common in numerous plant groups (e.g., [38,39]), including several Caryophyllaceae (e.g., [15,40,41]) and has been often interpreted as the consequence of chloroplast capture. This latter event, which may have originated by sympatry in refugia during Pleistocenic glaciations, may be invoked to justify some of the incongruent patterns. Chloroplast capture may have occurred in an Alpine-Dinaric refugium, involving populations of the stem lineages of Mc. graminifolia subsp. graminifolia and subsp. clandestina; and in a southern Carpathian refugium, involving populations of the stem lineages of M. saxifraga and M. graminifolia subsp. hungarica. For these reasons, in the following discussion and the consequent taxonomic treatment, albeit taking into account both trees, we paid attention to the fact that the plastid tree is severely afflicted by geography.
Regarding Mc. saxifraga s.l., the nuclear tree fully supports monophyly. Both subspecies included in this taxon share unique features within the genus (e.g., the foliaceous bracts without scarious margin, the direction of outer sepal veins, the number of leaf veins) [28,29] and occupy the eastern-most distribution area of Mcneillia. Their ecology has not been thoroughly studied, but, differently from most other Mcneillia taxa, they grow on metamorphic rocks ( [36]; authors' obs.). In addition, the nuclear tree confirms the less derived position of Mc. saxifraga within the genus, as indicated by Mattfeld [27]. Based on these considerations, Mc. saxifraga s.l. may be taxonomically segregated from the remaining taxa. However, we refrain from introducing an infrageneric division within Mcneillia, considering the substantial morphological uniformity of this small genus, and also taking into account that the plastid tree shows a strict relationship between Mc. graminifolia subsp. hungarica and Mc. saxifraga subsp. saxifraga.
The Turkish taxon Mc. saxifraga subsp. tmolea has "some claim to recognition at specific rank" [28] and remarkably differs from the European subspecies by several characters such as narrower cauline leaves, up to 2 mm vs. 3-4 mm wide, longer sepals up to 7-8 mm vs. 5-6 mm long. It has not been found elsewhere than in its locus classicus (very far from localities of subsp. saxifraga). Moreover, the plastid network indicates a different and separate position; and according to the nuclear tree, the branch length separating the two taxa is longer than those separating widely accepted species. Therefore, we propose to raise Mc. saxifraga subsp. tmolea to specific rank.
Another result of our study is the evident polyphyly of Mc. graminifolia (also after excluding subsp. brachypetala). In fact, the autonymic subspecies is completely separated, at least in terms of the nuclear DNA signal, from the other infraspecific taxa. Indeed, the phylogram in Figure 3a (Table 1), the chloroplast tree (Figure 3b) would suggest a less remote relationship of the former with subsp. clandestina; this is not surprising considering their closer geographical proximity. By a morphological standpoint, Mc. graminifolia subsp. graminifolia, even if similar to subsp. rosanoi (on account of indumentum and petal shapes) and to subsp. clandestina (on account of the shape and rigidity of leaves), readily differs by its larger sepals and petals. We therefore believe it is fully justified to regard Mc. graminifolia s.s. as a distinct species, a relic after glaciations in a small area of the South-Eastern Alps.
All the Central and Southern Italian specimens of Mcneillia, regardless of the taxonomic attribution, form a monophyletic unit in the chloroplast phylogeny. This group broadly appears in the nuclear tree as well, except for one specimen attributed to Mc. graminifolia subsp. rosanoi (APP no. 42436) from Central Italy (within the main Italian range of subsp. clandestina), which falls in a polytomy with this clade.
The status of Mc. graminifolia subsp. clandestina and subsp. rosanoi is difficult to interpret. Molecular data underline a distinctiveness of Italian representatives of subsp. clandestina ('Italian clandestina', hereafter) from their Balkan counterpart ('Balkan clandestina', hereafter), despite no relevant morphological differences having been observed. This discrepancy may be interpreted as a secondary contact between two incompletely differentiated allopatric taxa. In this case, we envisage that the 'clandestina' lineage arrived one or more times in Italy and hybridized with the 'rosanoi' lineage, with consequent backcrossing toward the latter and isolation from the 'Balkan clandestina'. This hypothesis is compatible with the following observations: (a) the aforementioned morphological similarity between most samples of 'Italian clandestina' and 'Balkan clandestina', except for the leaves of the flowering stems, which are usually at least 1 2 as long as internodes in Italy and rarely more than 1 / 3 as long as internodes elsewhere (cf. [29]); (b) this single character is shared by 'Italian clandestina' with the subsp. rosanoi; (c) 'Italian clandestina'is mostly limited to the northern sector of Central Apennines, where it is genetically homogeneous, and likely represents also a migration route for its lineage to reach the Peninsula; and (d) scattered southern localities of subsp. clandestina (albeit not recently confirmed and often based on poor material not investigated by molecular means) imply, however, that the ranges of subsp. clandestina and subsp. rosanoi are largely overlapping and gene flow has occurred extensively. Overall, even if Mc. graminifolia subsp. rosanoi and subsp. clandestina, in their typical forms, are well distinguishable on account of several quantitative characters, and chiefly by the indumentum, the quantitative characters are partly overlapping, and sometimes hairy and glabrescent individuals occur in the same locality [42] (p. 498; NAP!); so, some populations throughout the Apennines, albeit attributable to one of the two taxa, show intermediate features [33]. On account of these considerations, of the overlapping geographical ranges, and, above all, of the molecular results, we chose to treat the two taxa as conspecific, but distinct at subspecific rank. Regarding the genetic distinctiveness of the 'Italian clandestina' from the 'Balkan clandestina', we refrain from proposing a split treatment of these taxa, as the only diagnostic feature hitherto found (i.e., the ratio between the cauline leaves and the internodes) is rather weak and not statistically investigated. We think that further morphometric, karyological, and molecular studies, with a more extensive sampling, might shed light on this critical issue.
Surprisingly, the same holds for Mc. moraldoi, which is nested in subsp. rosanoi in both nuclear and chloroplast analyses. Morphologically, Mc. moraldoi can be distinguished from Mc. graminifolia subsp. rosanoi especially on account of its elliptic cauline leaves and also by its more laxely casepitose habit and less rigid leaves. The latter might be possible adaptations to the different habitat (shadowy flysch rocks).
Mcneillia graminifolia subsp. hungarica has been traditionally associated to the Italian plants with densely hairy and not rigid leaves, i.e., subsp. rosanoi [27], or even regarded as not distinct from it (see Table 1). According to the nuclear tree, as said above, there is some affinity between subsp. hungarica and the Italian clade, likely due to the persistence of "ancestral" characters (see above). However, on one hand, the Romanian populations constantly differ from subsp. rosanoi by their shorter glandular hairs and petals roughly equalling the sepals [33]; on the other hand, they are reproductively isolated, because not only are their ranges sharply disjunct but are separate by the 'Balkan clandestina'. In addition, the plastid tree clearly does not recover any particular affinity with the Italian taxa, but rather with the geographically closer Mc. saxifraga subsp. saxifraga. Therefore, we consider the specific level the most appropriate.
Mcneillia pseudosaxifraga, albeit described by Mattfeld [27] as a subspecies of Mc. stellata (but with some features resembling Mc. saxifraga) is regarded as a very distinct species by Halliday [29], "perhaps more closely related" to Mc. saxifraga as well. Interestingly, a relationship with both taxa is suggested by the plastid network (Figure 4), which indicates that the haplotype of Mc. pseudosaxifraga originated by a haplotype shared by Mc. saxifraga subsp. tmolea and a specimen of Mc. stellata. The monophyly of Mc. pseudosaxifraga is evident both in the nuclear and the chloroplast trees, and it is well supported by its morphological features [29,33], in some cases autapomorphic (i.e., the stems without dead leaves). As a consequence, we regard the specific rank as appropriate.
Regarding Mc. stellata, a species reported across Greece and Southern Albania, samples from the southern part of the range are different from the ones collected in the Pindhos massif. In particular, in the nuclear tree the northern individuals are included in a different, more internal clade as compared to the samples of Mc. stellata from the south. As in the case of 'Italian clandestina', the northern populations of Mc. stellata could be regarded as the result of a contact, presumably with the 'clandestina' lineage. This is suggested by the fact that a single specimen from Pindhos (APP no. 61429) is found close to 'Balkan clandestina' in plastid data. Indeed, the southern-most sampled individuals belong to the most typical form of the species (concurring with the lectotype). Whereas Mc. stellata is rather uniform throughout the central and southern sectors of its range, in the north it shows some morphological differences (not always correlated), such as the glabrous pedicels [28] and the longer leaves [30]. Incidentally, we note that these features somehow resemble those of Mc. graminifolia subsp. clandestina. For these reasons, these populations could be suspected to be of hybrid origin. They were described as var. epirota Halácsy [43] (p. 238), a taxon consistently disregarded in time [28,33], but possibly deserving subspecific rank, also in consideration of our molecular results.

Taxonomic Treatment
The articles cited throughout the paragraph follow the Shenzen Code [44] ICN). Eremogone graminifolia is a synonym of Eremogone saxatilis (L.) Ikonn., according to POWO [31]. This latter database, however, incorrectly reports the combination by Bluff et al. [45] as "Alsine graminifolia (Ard.) Bluff, Nees & Schauer", and therefore it is wrongly reported as a further synonym of Mc. graminifolia (i.e., Alsine arduinoi Fenzl according to Bluff et al. [45] so represent a single gathering (cf. Arts. 8.2 and 9.17 of ICN). The diagnostic features of the taxon are easily observable and therefore the specimen fully supports the current use of the name: height more than 2 cm, caespitose and lax habit, leaves narrowly lanceolate, greyish-green, densely glandular, and not rigid, stems rather woody, elongated and stout but without dead leaves, inflorescences with up to five flowers, petals up to 1 2 longer than sepals (see the leftmost individual), bracts narrow, lanceolate, and not scarious on margins.
Mcneillia rosanoi (Ten.) F. Conti   Notes: The name Minuartia stellata subsp. tmolea appeared for the first time in [53], but without any diagnosis or description (invalid name, cf. Art. 38.1 of ICN). In December, Mattfeld [27] validly published the new subspecies by a diagnosis in the key at p. 132, also providing taxonomic notes at p. 133 with syntypes from Mt. Tmolus (locus classicus atque unicus): (1) a specimen or specimens by Boissier, without any indication of herbarium, and the no. 122 of Balansa's gathering in the herbarium of Haussknecht. McNeill [28] reported several syntypes of two gatherings, listing, under that by Balansa: "holo. B (destroyed?); iso. BM!, G, JE!, K!". In a successive treatment [37], it becomes clear that at BM, JE, and K specimens of both the gatherings are present, but the "holotype" in B is cited no more. In addition, McNeill located a further specimen of Boissier's gathering at LIVU. McNeill [28] chose a specimen by Balansa from B, although Mattfeld [27] explicitly cited the Haussknecht's herbarium, which has been kept at JE since the times of his owner [54]. In fact, as reported above, two duplicates by Balansa are preserved in JE, and these are the only ones traced by us as belonging to Haussknecht's herbarium. However, as the Art. 9.12 of ICN does not provide any preference between syntypes and isosyntypes in lectotypification designation, the choice by McNeill [28] is correct. Nevertheless, as the previous lectotype at B is unavailable, a new lectotype may be designated (Art. 9.11 of ICN). In this case, it seems appropriate to propose another syntype of the series already chosen by McNeill [28], i.e., Balansa n. 112, and namely in the Haussknecht herbarium at JE. There we traced two duplicates: barcodes JE00009377 (also included in the Herbarium Gaillardot), and JE 00009379 (with a print label). Both specimens were revised by Mattfeld, and the former also seen by McNeill, as indicated by modern labels. They include several fruiting individuals (fragments?). JE00009377 is more complete and shows the diagnostic features of the taxon: few-flowered inflorescences not more than 2 cm long (but cf. also P04990995!), cauline leaves strictly lanceolate with parallel veins. Other specimens belonging to the same gathering, but not marked as "Balansa n. 112" (e.g., JE00009378) are nevertheless regarded here as syntypes (Art. 8.2 of ICN). In addition to the syntypes reported by previous authors, we located abundant material elsewhere.

Sample Collection and DNA Extraction
We collected leaf material from 29 herbarium specimens of Mcneillia (Herbarium codes according to Thiers [55]), representing all known taxa across their whole range (Table 2, Figure 1). In this regard, we relied on the above-mentioned taxonomic treatment in POWO [31], which is the most comprehensive, to verify the taxonomic value of critical taxa, such as Mc. graminifolia subsp. hungarica and subsp. rosanoi. As the outgroup, we chose one specimen of Mn. recurva (All.) Schinz & Thell. subsp. condensata (C.Presl) Greuter & Burdet, selected according to both its position in the molecular phylogeny by Dillenberger and Kadereit [26] and availability (Table 2). An illustration of some Mcneillia taxa is provided in Figure 5. Total DNA was extracted using the GeneAll ® Exgene™ Plant SV mini kit (GeneAll Biotechnology, Seoul, Korea) following the manufacturer's protocol for dried material. Plant material was grinded to powder using Mixer Mill 300 (Retsch ® , Verder Scientific, Haan, Germany). The quality and quantity of extracted DNA was evaluated by 0.8% gel electrophoresis using the high-molecular weight marker HyperLadder™ 1 Kb (Bioline, Meridian Bioscience, Cincinnati, OH, USA).

Marker Selection, Amplification, and Sequencing
We selected eight molecular markers: two from the nuclear (ETS and ITS regions) and six from the chloroplast genome (rpoC1, rps16 intron, rps16-trnQ, rpl32-trnL, trnL-trnF, and trnH-psbA). These genomic regions were amplified by polymerase chain reaction (PCR) into a final volume of 25 µL containing: 7-10 ng DNA, 2X Kodaq PCR MasterMix (Applied Biological Materials Inc. ® , Richmond, Canada), 400 nM forward and reverse primers, and water to volume. PCR conditions and primers are listed in Table S1. The amplification products were separated by 1.5% agarose gel electrophoresis in TBE 0.5×X buffer and visualised under UV light after staining with SafeView ™ Classic (ABM ® , Richmond, Canada). The amplified products were purified using the NucleoSpin ® Gel and PCR Cleanup kit (Macherey-Nagel, Düren, Germany) and quantified on a 1.5% agarose gel. Sequencing reactions were carried out in a final volume of 5 µL using the BrightDye ® Terminator Cycle Sequencing Kit (MCLAB, Harbor Way, San Francisco, CA, USA), and purified using the the BigDye ® Xterminator ™ Purification Kit (Applied Biosystems, Thermo Fisher Scientific, Foster City, CA, USA). The ITS, rps16 intron and Total DNA was extracted using the GeneAll ® Exgene™ Plant SV mini kit (GeneAll Biotechnology, Seoul, Korea) following the manufacturer's protocol for dried material. Plant material was grinded to powder using Mixer Mill 300 (Retsch ® , Verder Scientific, Haan, Germany). The quality and quantity of extracted DNA was evaluated by 0.8% gel electrophoresis using the high-molecular weight marker HyperLadder™ 1 Kb (Bioline, Meridian Bioscience, Cincinnati, OH, USA).

Marker Selection, Amplification, and Sequencing
We selected eight molecular markers: two from the nuclear (ETS and ITS regions) and six from the chloroplast genome (rpoC1, rps16 intron, rps16-trnQ, rpl32-trnL, trnL-trnF, and trnH-psbA). These genomic regions were amplified by polymerase chain reaction (PCR) into a final volume of 25 µL containing: 7-10 ng DNA, 2X Kodaq PCR MasterMix (Applied Biological Materials Inc. ® , Richmond, BC, Canada), 400 nM forward and reverse primers, and water to volume. PCR conditions and primers are listed in Table S1. The amplification products were separated by 1.5% agarose gel electrophoresis in TBE 0.5×X buffer and visualised under UV light after staining with SafeView ™ Classic (ABM ® , Richmond, BC, Canada). The amplified products were purified using the NucleoSpin ® Gel and PCR Cleanup kit (Macherey-Nagel, Düren, Germany) and quantified on a 1.5% agarose gel. Sequencing reactions were carried out in a final volume of 5 µL using the BrightDye ® Terminator Cycle Sequencing Kit (MCLAB, Harbor Way, San Francisco, CA, USA), and purified using the the BigDye ® Xterminator ™ Purification Kit (Applied Biosystems, Thermo Fisher Scientific, Foster City, CA, USA). The ITS, rps16 intron and rps16-trnQ markers were sequenced in both directions, while the others in one direction if the signal was unambiguous; for the intergenic spacer rpl32-trnL (over 1200 bp long), we sequenced only the variable region towards the 3 end because of the lack of variation in other regions and the occurrence of long polynucleotide stretches that hampered the correct reading and assembling of reads. Capillary electrophoresis was carried out in the Applied Biosystems ® 3130 Genetic Analyzer (Applied Biosystems, Thermo Fisher Scientific, Foster City, CA, USA).

Sequence Alignment and Exploratory Data Analysis
Electropherograms were visually inspected for ambiguities, and then forward and reverse sequences were assembled in contigs before individually aligning them with the ClustalW algorithm [56] as implemented in the BioEdit v7.2.6 software [57]. Standard IUPAC ambiguity codes were used when base peaks overlapped, or the lower peak was at least one-third in height as the highest one. All sequences were deposited in Gen-Bank/DDBJ (see Data availability section for details). Nuclear and chloroplast loci were then separately concatenated in two matrices using Mesquite v3.51 [58]. The best fitting evolution models were computed for each dataset in jModelTest v2.1.3 [59] using the corrected Akaike information criterion [60]. Exploratory Bayesian phylogenetic analyses were then individually carried out in MrBayes v3.2.6 [61] on both nuclear and chloroplast Mcneillia datasets, in two replica runs of four chains (one of which heated) for 1,000,000 generations, sampling chains every 1000 generations, under the default relaxed clock. Convergence and effective sample sizes (ESS) for all parameters were investigated in Tracer v.1.7 [62], the latter considered acceptable when >200. The first 10% of the samples was discarded as burn-in. The majority-rule consensus trees were visualized using FigTree v1.4.3 (http://tree.bio.ed.ac.uk/software/figtree/, accessed on 19 September 2021).
By observing branch lengths from the investigations above, the samples of Mc. graminifolia subsp. brachypetala resulted as surprisingly different from the rest of the ingroup. We therefore decided to further verify the phylogenetic position of the said taxon. To this aim, we integrated our ITS sequences with the 255 sequences employed by Dillenberger and Kadereit [26]. The sequences were de novo aligned and the resulting alignment was trimmed at the same length of our sequences. After a Bayesian investigation, we discovered that Mc. graminifolia subsp. brachypetala was sister group to the representatives of genus Minuartiella (abbreviated as Ml. from now onwards) included in Dillenberger and Kadereit [26]. We therefore added the five Minuartiella ITS sequences published by Koç et al. [35] (MK089560.1 to MK089564.1, KF737436.1, and KF737437.1) to the previous dataset, for a total of 290 accessions.
A Bayesian analysis was then carried out in MrBayes on the said ITS dataset with two replica runs and four chains for 5,000,000 generations and sampling chains every 5000 under the JC + G model.

Analysis of the Final Datasets
The three accessions of Mcneillia graminifolia subsp. brachypetala were then removed from the Mcneillia datasets, which were re-aligned and subjected to a novel computation of the evolution model as indicated above, but in this case separately for each nuclear and chloroplast DNA region. Incongruence between the nuclear and chloroplast DNA was evaluated both by the observation of conflicting topologies with a posterior probability ≥ 0.99 [63] and by a formal incongruence length difference (ILD) test [64,65] carried out in PAUP* v4.0a build 169 [66] starting with 100 replications and increasing the search up to 10,000 trees. A test of the clockwise behaviour of each dataset was carried out with two stepping-stone analyses testing strict and relaxed clock models, run in MrBayes in two replicas, sampling 200 steps of 50,000 generations each.
Bayesian analyses were carried out again on both nuclear and chloroplast Mcneillia datasets with the above-mentioned software for 1,000,000 generations, sampling chains every 1000 generations, under a strict clock, evaluating the quality of run parameters and visualizing trees as indicated above. To evaluate the relationships among the chloroplast haplotypes, we inferred a TCS network [67] in the software PopART [68] using a parsimony threshold of 95% for the calculation of the statistical parsimony algorithm [69].