Phylogenetics and distribution of Orchidinae s.l.
With broad sampling across Orchidinae and seven DNA markers, many new or previous overlooked phylogenetic relationships were discovered in our analyses. Orchidinae s.l. (comprising Orchidinae s.s., Habenariinae and Satyriinae of other classifications, e.g. ) are monophyletic with strong support and divided into three main clades: Satyrium (Africa and Asia), Orchidinae s.s. (mainly Northern Hemisphere), and the formerly recognised Habenariinae’ (Fig. 3; nearly cosmopolitan). Satyrium is sister to the rest of Orchidinae s.l. with strong support.
Orchidinae s.s. include two superclades and two smaller clades sister to these. Superclade I has high to moderate support and includes 12 sampled genera: Anacamptis, Chamorchis, Dactylorhiza, Galearis, Gymnadenia, Himantoglossum, Neotinea, Ophrys, Orchis s.s., Platanthera, Steveniella, and Traunsteinera. Most of these genera with ovoid/ellipsoid/globose tubers have their center of diversity in Europe and the Mediterranean region, but the two genera with palmately lobed tubers, i.e., Dactylorhiza and Gymnadenia, extend into the Far East (Dactylorhiza into North America), whereas two genera with rhizomes or fleshy rootstocks, i.e., Galearis and Platanthera, have two centers of diversity, the Pan-Himalayan region and North America.
Superclade II (97, 88, 1), according to our delimitation (discussed below), includes five genera: Hemipilia, Ponerorchis, Shizhenia, Sirindhornia, and Tsaiorchis. This is a subtropical to temperate clade with ovoid/ellipsoid/globose tubers and basal leaf (leaves) occurring almost exclusively in eastern Asia and the eastern Himalayas, with a single species (Ponerorchis cucullata (L.) X.H. Jin, Schuit. & W.T.Jin) occurring as far west as Poland. In the subtropical zone the members of this superclade is normally found above 1000 m asl., often in alpine regions [4, 31].
Two relatively isolated genera, Schizochilus and Brachycorythis, were resolved as successive sister to these two superclades with weak to strong support (Fig. 2). Schizochilus is exclusively from eastern to southern Africa, whereas Brachycorythis is most diverse in tropical and southern Africa, but also has some species in tropical Asia as far east as Taiwan. Unlike many other Orchidinae, Brachycorythis does not occur at alpine elevations in Asia ; it is essentially a tropical montane genus.
The clade formerly referred to as Habenariinae (Clades XXI–XXXI) include four superclades (III–VI) and four isolated clades. Chase et al. (2003; 2015) decided not to recognise Habenariinae because of uncertainty over phylogenetic placements of genera such as Androcorys (synonym of Herminium, see Raskoti (2016)), Brachycorythis and Holothrix, and until all of the missing genera of Orchidinae s.l. are included, it would be premature to resurrect Habenariinae and not necessary regardless of the final phylogenetic outcome. Superclade III comprises the genera Herminium and Hsenhsua and is exclusively Asian, except for a single taxon that extends into Europe (H. monorchis (L.) R.Br. ex Aiton). Most are alpine plants, including the species that probably holds the elevational record in Orchidaceae (H. pugioniforme Lindl. ex Hook.f. at c. 5200 m). Superclade IV includes Pecteilis (often included in Habenaria) and about 50 species of Habenaria mainly distributed in tropical Asia. Superclade V includes only one genus, Peristylus, sister to superclades III + IV with weak support. Peristylus is only found in tropical Asia, from the lowlands to the alpine zone. The small tropical Asian genus Diplomeris is sister to these three superclades with weak support. Superclade VI includes Bonatea and 42 species of Habenaria; this superclade is distributed in the Americas, tropical and southern Africa, and tropical Asia. A number of small clades are well supported, but their relationships with other Habenariinae are still largely unresolved. Benthamia and Tylostylis are two sister genera from Madagascar, comprising clade XXVIII, whereas clade XXIX contains the two species of Gennaria, one from the Canary Islands and the other from the Himalayan region, a remarkable disjunction. Clade XXX contains two genera from Madagascar and the Mascarenes, Cynorkis (also in southern Africa) and Physoceras and is resolved as sister to former Habenariinae with strong support. The African genus Stenoglottis, often included in Habenarrinae, is sister to the remainder of Habenariinae, but with weak support.
Phylogeny and generic delimitation
Usually there are many ways to translate a phylogenetic tree into a Linnean classification. In the interest of nomenclatural stability it is desirable that a classification is broadly accepted. One way to help ensure such acceptance is by using objective criteria for generic delimitation as far as possible. Bateman, as cited in Tang et al. (2015), proposed the following five guidelines:
Recognize only monophyletic groups (clades);
Preferentially divide the tree at branches that are relatively robust;
Preferentially divide the tree at branches that receive similar levels of statistical support (obviously, there exists considerable overlap between Rules 2 and 3);
Minimize the proportion of branches in the tree that represent more than one taxonomic rank (notably monotypic higher taxa);
Preferentially divide the tree in a way that minimizes the need (a) to create new names and/or (b) to create new combinations of existing names.
Although we would agree to a large extent with the utility of these guidelines (always allowing for exceptions), we think that one consideration is missing: there is no reference to morphology. We would suggest that a classification supported by morphology is likely to gain acceptance more readily than one based solely on molecular criteria. Therefore, we propose the following addition:
Preferentially divide the tree in such a way that the points of division correspond to morphological discontinuities (in particular those representing evolutionary innovations).
Ultimately, we are mainly interested in the topology of our phylogenetic tree. Measures of internal support inform us of clarity in the phylogenetic signal of the data analysed. Morphological criteria should also be applied; they not only help us reduce the large number of potential classification schemes based on a given tree, but also make the resulting classification of greater practical value, and thereby easier to accept. However, without a formal cladistics analysis of morphological data, such applications of morphological criteria are ad hoc and likely to be subjective and ignore conflicting characters while emphasising those preferred by one set of authors. Below we discuss a number of troublesome alliances in which consensus has not yet been reached, as well as others in which we are able to present a consensus agreed by most authors.
Bonatea, Diplomeris, Habenaria, Herminium, Pecteilis, and Peristylus
Morphologically, Diplomeris is an isolated genus in Orchidinae, characterized by its large, white flowers with an entire lip, a column with well developed, semi-circular connective about two thirds the length of the column, two elongate, oblong stigmatic lobes extending from near the entrance of the spur, and a large, shield-like midlobe of the rostellum. Ecologically, it is unusual in Orchidinae in that the species grow as lithophytes on cliffs with seasonal running water trickling down the moss-covered rock. This genus of two species is mainly distributed in tropical to subtropical regions in the Himalayas [2, 4, 33] and would be difficult to place without molecular data. It is somewhat unexpected that Diplomeris is sister to the group formed by superclades III–V, but support for this position is as yet weak only.
Peristylus consists of about 70 species mainly distributed in tropical Asia [2, 34]. It is not always easy to distinguish from Peristylus from Habenaria and Herminium [2, 10, 23, 33,34,35]. Recent studies have demonstrated that most alpine species of Peristylus actually belong to Herminium [10, 23]. With the exclusion of these species, Peristylus is distributed mainly in the lowland and montane zones of tropical and subtropical Asia. Our results indicate that, excluding the misplaced species, Peristylus is monophyletic and characterized by its elliptic tuber, more or less clearly trilobed and spurred lip, narrow connective, and two stigma lobes adnate to the entrance of the spur. The species of Peristylus analysed are distributed over four strongly supported clades: (1) P. chapaensis in subtropical limestone regions; (2) a tropical group, including P. stocksii (Hook.f.) Kraenzl.; (3) a subtropical montane subclade including P. affinis; (4) another subtropical montane subclade including P. densus (Lindl.) Santapau & Kapadia. Each subclade is to a large extent characterized by minor morphological characters, especially of the spur and lip.
Pecteilis consists of about eight species distributed in southern and eastern Asia and the Malay Archipelago [4, 30, 31]. Morphologically, it is similar to Habenaria, but differs in its concave, sessile stigma lobes, unlike the stalked stigmaphores of Habenaria [2, 30]. We find that Pecteilis is not monophyletic and deeply nested within clade XXIII (otherwise containing species currently accepted as members of Habenaria by most authors), and forms a strongly supported subclade interspersed with species of Habenaria. It seems that the stigmaphore has been lost independently at least twice in this subclade. Except for the stigmaphore, several members of Habenaria of clade XXIII share morphological characters with Pecteilis, such as an anther with a broad connective, thecae extending from base of column, and large, subrhombic to flabellate lateral lobes of the lip (Fig. 1h). On the other hand, many other species have linear to filiform lateral lobes, such as H. plurifoliata Tang & F.T.Wang, and H. propinquior Rchb.f. Species in clade XXIV, which most recent authors include in Habenaria, are mostly from alpine habitats in Asia (especially in Himalayan region), and are characterized by two basal, almost opposite leaves appressed to the ground, lateral lobes of the lip linear to filiform, green to white flowers, and petals often more or less distinctly bilobed. Some species in clade XXIII also have leaves appressed to the ground, such as H. roxburghii D.H.Nicolson and H. delavayi (the latter also with linear lateral lobes to the lip, but usually with three or more leaves in a rosette). On the other hand, the petal lobes in clade XXIV are sometimes indistinct, making the two clades ultimately difficult to tell apart. Together they form superclade IV.
Our results indicate that Habenaria is polyphyletic but with low support; it is divided into two well supported but not immediately related groups, superclades IV and VI. Superclade IV, discussed above under Pecteilis, is exclusively Asian, and is sister to superclade III (Herminium and Hsenhsua). A sister-group relationship of superclade VI with any other clade in Habenariinae (Fig. 2) has not been established, since superclade VI consisting of all African and southern American taxa sampled so far as well as some Asian ones is part of a polytomy.
There are at least two divergent options to treat the polyphyly of Habenaria. One option would be to expand the delimitation of Habenaria to include Benthamia, Bonatea, Diplomeris, Gennaria, Herminium, Hsenhsua, Pecteilis, and Tylostigma (Clades XXI–XXIX), many of which have been in the past considered members of Habenaria (e.g., many species of Benthamia, Bonatea, Gennaria, Herminium, Pecteilis, Tyolostigma have combinations in Habenaria). Herminium would be the earliest available name for this assemblage. Therefore, ironically, an attempt to keep all the species of Habenaria within a single genus would lead to the loss of the name Habenaria, unless it was conserved. This option would cause minimal nomenclatural upheaval, and the resulting genus could be diagnosed by the characters used to delimit the formerly recognised Habenariinae. A perhaps more disruptive option would be to split Habenaria into at least two genera, a decision that would likely result in no agreement over the number of segregate genera and how to recognise them.
The main problem with recognising two genera, the two larger clades into species of Habenaria fall, is that there appear to be no consistent morphological differences between the two superclades. However, perhaps it is possible to circumscribe the two groups in a more circuitous way, based on polythetic sets. For example, bilobed petals are, as far as we can tell, only found within certain subclades (in both superclades, unfortunately); similarly, pectinate or laciniate lateral lobes of the lip (Fig. 1d) only occur in some branches of our tree. When the group has been more fully studied it is perhaps possible to identify the two superclades using multiple sets of character states that are unique to each of the two.
Splitting Habenaria into more than two genera is another option to be considered. We note, however, that many character states show much homoplasy, and suspect that a more finely split classification will not be easier to use. Instead of having two difficult-to-define genera, we may end up with five or more difficult-to-define genera. This would only increase the potential for misidentifications, without having any obvious benefits, but some authors D. Szlachetko and collaborations have already started down this route, describing new genera based on only morphological characters, mostly column details, which typically results in polyphyletic circumscriptions (such as [36,37,38]). At this stage it would be premature to propose new combinations.
Brachycorythis, Hemipilia, Ponerorchis, Tsaiorchis, Sirindhornia (superclade II)
Brachycorythis, consisting of 30–35 species, is mainly distributed in Africa, with a few species in Asia. It is characterized by leaf-like floral bracts and a more or less bipartite, spurred lip. Recent molecular studies indicate that Brachycorythis is sister to the rest of Orchidinae s.s. [23, 25, 26], and that Sirindhornia is sister to the clade formed by Ponerorchis, Hemipilia and related genera . Both findings are also supported by our results, in which Sirindhornia is sister to clades XIV–XVII. Tang et al.  inferred (but with only moderate support) that Brachycorythis is sister to Hemipilia s.l. (in which these authors include Ponerorchis, Amitostigma, Neottianthe and Tsaiorchis), and that Sirindhornia is sister to the remaining Orchidinae s.s. We have been unable to replicate this result in our phylogenetic analyses. Morphologically, Brachycorythis is less similar morphologically to the Hemipilia alliance than Sirindhornia.
Tang et al.  detected a hard incongruence involving a group of six taxa (Ponerorchis graminifolia, P. [graminifolia var.] suzukiana, P. chidori, Amitostigma lepidum, A. keiskei, and A. kinoshitae) between nrITS and four plastid markers and stated that this most likely reflected plastid capture. They concluded that their ITS results were more likely to reflect the actual phylogenetic relationships. Our results indicate that, based on plastid markers, these six taxa, which are endemic to Japan, are deeply nested within clade XIV, but in different positions in the tree. Amitostigma keiskei and A. kinoshitae are sister to the subclade formed of Ponerorchis alpestris–P. physoceras (Fig. 2). Amitostigma lepidum is sister to a subclade formed exclusively of Amitostigma (s.l.) species: A. parciflora–A.trifurcata. The remaining three species (including Ponerorchis graminifolia) are sister to the subclade formed of Neottianthe species + Amitostigma gracile (Blume) Schltr. Based on nrITS alone, according to the results of Tang et al.  and our analyses of nrITS (Additional file 2: Figure S2), these six species form a strongly supported clade sister to our clades XIV–XVI. Since they include the type species of Ponerorchis (P. graminifolia), this would imply, as Tang et al.  correctly pointed out, that our clade XIV could not be called Ponerorchis, as it would not include the type species of Ponerorchis.
Morphologically, these six species do not form a uniform group, and the scattered placement of its members in the plastid tree more accurately reflects their diverse morphology than their occurrence in a single clade. We consider it unlikely that six taxa that are all endemic to Japan and morphologically disparate, happen to form a single clade of which the members have somehow integrated the plastid genomes of at least three different, now extinct progenitors that are closely related to geographically isolated, extant Chinese taxa.
Based on our results from nuclear markers (ITS +Xdh) we infer that the six taxa are nested within a polytomy of Ponerorchis s.l., which is sister to Hemipilia + Tsaiorchis with moderate support (BSML = 78, BSMP = 67, PP = 0.72) (Additional file 4: Figure S4). They do not form a clade, and their placement does not render Ponerorchis in the sense of Jin et al.  poly- or paraphyletic.
It remains an interesting problem to explain how ITS data could produce such discordant results in the case of these six taxa. It is well known that certain molecular genetic processes involving ITS, such as ancient or recent array duplication events, genomic harbouring of pseudogenes, and incomplete intra- or inter-array homogenization, may distort results of phylogenetic inference and need to be examined . Whatever the cause for the incongruence, on morphological grounds and the plastid and combined nuclear molecular results, we prefer to continue to recognize Hemipilia, Ponerorchis and Tsaiorchis as distinct genera, while adding Shizhenia, which is sister to the three others combined.
Although most species-rich in Asia, Platanthera is also diverse in North America, and a few segregate genera have been recognized there (e.g. Blephariglottis Raf., Limnorchis Rydb., Piperia Rydb., and others). In a recent revision of the Asian taxa , the genus was taken in the broad sense, which corresponds with our clade I.
Platanthera is among the largest genera of the tribe Orchideae, consisting of some 130 species, and is mainly distributed in Asia and North America, with a few species in Europe and North Africa [2, 19]. Historically, the generic delimitation and phylogenetics of Platanthera were confusing [19, 21, 28, 40]. Recent results indicate that once a few misplaced species are transferred to the proper genera, Platanthera s.l. is monophyletic and relatively well characterized by morphology, such as fleshy, more or less tapering roots, often forming fusiform tuberoids, a spurred lip, a short and truncate column, and a broad connective [19, 23, 41] (Fig. 1c).
Efimov  recognized five subgenera, which were recovered in our tree. The Platanthera minor group comprises about eight species distributed in subtropical regions in East Asia; it is characterized by several cauline leaves attenuating into bracts, a dense inflorescence, and a narrowly elliptic lip narrowed at the base. This group in subgenus Platanthera could not be assigned to a section by Efimov . The Platanthera minutiflora group displays disjunction between East Asia and North America. It was recognized by Efimov  as subgen. Platanthera sect. Lysiella (Rydberg) Efimov. The Platanthera carnosilabris group (99, 100, 1) in subgen. Platanthera includes about eight species that are characterized by small flowers, a lip more or less concave at base, and spur shorter than the lip. Efimov  could not place these in a section. This group includes the misplaced Herminium orbiculare Hook.f. Moderate support (75, 75, 0.97) is shown for subgen. Platanthera sect. Stigmatosae (Lang) Efimov (P. bakeriana–P. leptocaulon).
Both Stenoglottis and Schizochilus are small genera endemic to southern Africa. The phylogenetic positions of Stenoglottis and Schizochilus have not been fully established in previous studies, and are still uncertain. Bateman et al.  resolved Stenoglottis as sister to Habenariinae s.s. Inda et al. , on the other hand, found Stenoglottis to be sister to the rest of their tribe Orchideae (our Orchidinae s.l.). Our own results show weak support for the position inferred by Bateman et al. (< 50, 60, 0.64).
Schizochilus is here found to be sister to Orchidinae s.s., but with weak support (72, < 50, 0.6). Morphologically, Schizochilus is similar to Brachycorythis and Sirindhornia, especially in column morphology.
Satyrium (Clade XXXII)
Satyrium comprises about 90 species and is mainly distributed in temperate and montane regions of Sub-Saharan Africa, with a few species occurring in Asia, mainly in the Himalayan region . Based on its distinctive morphology, including non-resupinate flowers, a galeate lip with two spurs, stalked column, and pad-like stigmatic lobes, which sets it apart from other Orchidinae, Satyrium has been treated as a monotypic subtribe Satyriinae [2, 42]. Published molecular studies indicate that Satyrium is monophyletic and sister to Orchidinae s.l. [25, 26, 43, 44], which is also strongly supported by our results 99, 97, 1; Fig. 2). Chase et al.  assigned Satyrium to Orchidinae, which was motivated especially by the desire to minimize the number of monogeneric higher taxa. This is also in agreement with guideline 4 above (chapter 4.2). Considering the distinctive morphology of Satyrium, a case could be made to reinstate Satyriinae, but similar arguments could be advanced for many other subtribes. We would prefer not to increase the complexity of our classification if this can be avoided. Limiting the number of monogeneric higher taxa as far as possible is a means towards this end. In addition, it would appear from the work of Inda et al.  that the genus Holothrix is sister to Satyrium + the remainder of Orchidinae s.l. Therefore, recognition of a subtribe Satyriinae would also require the recognition of another subtribe for Holothrix (which might include other, related genera that have not been analysed yet). It is certainly simpler to keep them all within Orchidinae.
Gennaria (Clade XXIX)
There are two species in Gennaria, the type species G. diphylla Parl. in the Mediterranean region plus the Canary Islands, the other (the recently transferred G. griffthii (Hook.f.) X.H.Jin., L.Q.Quang, W.T.Jin & X.G.Xiang, Fig. 1e) in the Pan-Himalayan region [2, 45]. Morphologically, Gennaria is characterized by subcampanulate flowers more or less secund along the rachis, a short column with stalked lateral appendages, and convex stigmas [2, 46]. Inda et al.  found Gennaria to be deeply nested within Habenaria. Batista et al.  had Gennaria as sister to Habenaria + Bonatea, but with low support, which is similar to the analysis of Jin et al. (2014). Our new results show that Gennaria is a strongly supported clade that does not nest within Habenaria. Instead, it is one of the branches of a polytomy formed otherwise by superclade VI, Clade 26, and Clade 25. It may be an outlier in this alliance, which is also suggested by the extremely wide but fragmented distribution area of the clade, with a low number of extant species.