Gattenhofia, Medikus, 1790

Crespo, Manuel B., Martínez-Azorín, Mario & Mavrodiev, Evgeny V., 2015, Can a rainbow consist of a single colour? A new comprehensive generic arrangement of the ‘ Iris sensu latissimo’ clade (Iridaceae), congruent with morphology and molecular data, Phytotaxa 232 (1), pp. 448-450 : 448-450

The Gattenhofia lineage (Vernae irises)

Medikus (1790) accepted some genera segregated from Iris . Among them, he described the monotypic Gattenhofia to include I. verna Linnaeus (1753: 39) , which he separated on the basis of differences in vegetative and fruit features. This taxon is restricted to southeastern North America ( Fig. 8B) and retains some presumedly plesiomorphic characters, such as the short stems hidden into spathes, and the long scapiform tube of perigone ( Figs. 3B, 6C, 7B). However, it possesses some differential features such as the heterogenous rhizome, usually slender and torulose, sometimes with long lateral branches, densely covered with brown scale-like leaves; the minute velvety hairy band on the falls; and the testa surface longitudinally ribbed ( Fig. 6C). More recent authors treated it as a subsection in Iris sect. Iris ( Diels 1930) or in I. sect. Apogon Baker (1876c: 143) (Forster 1937), or a series in I. sect. Limniris ( Lawrence 1953, Rodionenko 1961), or more recently as a separate section in Iris ( Rodionenko 2009) . The chromosome number of I. verna is 2n = 42 ( Simonet 1934), which suggests p = 7.

Relationships of Iris verna to other relatives have usually remained obscure. Foster (1937) stressed that some vegetative characters related I. verna to the bearded irises, albeit flower structure resembled that of the beardless irises. Furthermore, its chromosome structure made it hard to relate this species to other irises ( Simonet 1934, Foster 1937). In fact, Dykes (1912) had previously pointed out that this species seemed “to stand entirely apart from all other Irises”, and it could hardly be grouped with any of them. Similarly, Mathew (1987) considered this plant to be “somewhat out on a limb and the only representative of its series”. Provided that remarkable differences exist to segregate I. verna in its own, the genus Gattenhofia is revived to accommodate this noteworthy species. A similar generic treatment had previously been adopted by Small (1933), who placed it in Neubeckia , a genus broadly heterogeneous and artificial as first conceived by Alefeld (1863).

Recent analyses ( Fig. 1) recovered Gattenhofia verna and the clade ( Belamcanda + Pardanthopsis ) in an early branching position with regard to other lineages of “bearded irises”. In the case of Gattenhofia , this placement is well supported by some shared floral characters, such as the perigone pieces similar in shape and size, the presence of a minute, velvety pubescence (giving a papillate appearance) on the central yellow band of falls ( Mathew 1989a: 197, Henderson 2000), the apically incurved standards, and the entire and rounded stigmatic lip ( Figs. 3B, 7B). In fact, Medikus (1790) had already remarked that flower features were identical in his new Gattenhofia and Iris , and Dykes (1912) had stressed “the appearance and habit similar to a small I. pumila ”. Seeds of I. verna produce a fleshy ‘raphe’ ( Fig. 6C) and are in some extent similar to those of the arillate bearded irises, I. sect. Oncocyclus (Siems.) Baker, I. sect. Regelia (Foster ex Baker) Lynch, I. sect. Psammiris (Spach) J.J.Taylor , and I. sect. Pseudoregelia Dykes. It is therefore hard to understand that Baker (1892) included it in I. subg. Pardanthopsis (Hance) Baker (see below).

The Pardanthopsis (Vesper irises) plus Belamcanda (Blackberry lilies or Leopard lilies) clade

Iris dichotoma Pallas (1776: 712) is a morphologically singular taxon growing in central Asia (Siberia, Mongolia and northern China), which shows a peculiar combination of characters not present altogether in any other group of irises. The most sound features distinguishing this species are the many-branched and many-flowered scapes producing a succession of short-lived, small flowers which shrivel spirally after anthesis and which break off just below the ovary if unfertilised, the absence of a perigone tube, and the very small style branches with two short lobes ( Mathew 1989a) ( Figs. 3C, 7C). In the light of its divergent morphology it was segregated as I. sect. Pardanthopsis Hance (1875a: 105) , and was later raised to subgeneric rank by Baker (1892) who circumscribed it in a wide sense. More recently, Lenz (1972b) treated it as an independent genus, a solution that has largely been accepted in the last decades, though alternatively it also has been treated at subgeneric rank ( Rodionenko 1961, Wilson 2011). According to Zhao et al. (2000) and despite the taxonomic treatment adopted, two closely related species can be distinguished in the Vesper irises, both occurring in easterm Asia (from southern Siberia to southeastern China) ( Fig. 9A).

A NEW GENERIC ARRANGEMENT OF IRIS SENSU LATISSIMO

Phytotaxa 232 (1) © 2015 Magnolia Press • 15 16 • Phytotaxa 232 (1) © 2015 Magnolia Press

CRESPO ET AL.

About a dozen morphological, anatomical and biological features were found ( Lenz 1972a –c, Schulze 1971a) which allowed easy separation of Pardanthopsis from other irises. The most closely related taxon seems to be Belamcanda chinensis , with which it shares some vegetative features, but from which it differs by a totally divergent floral structure and fruit and seed characteristics ( Lenz 1972a –c, Schulze 1971a, Mathew 1989a, Goldblatt et al. 1998) ( Figs. 3C–D, 6A–B, 7C–D). Accordingly, molecular phylogenies place Pardanthopsis as sister to Belamcanda , they both being presumably sisters to the true “bearded irises” ( Fig. 1; see also Tillie et al. 2001, Makarevitch et al. 2003, and Wilson 2011). A number of shared rare chemical compounds (see Kaššak 2012, Kukula-Koch et al. 2014, Wei et al. 2012, and references therein) are found in Pardanthopsis , Belamcanda , I. sect. Oncocyclus and I. sect. Chinenses (the genus Evansia ), a fact supporting their close placement in the molecular trees. However, peculiar chromosomal characteristics of Pardanthopsis (i.e. the large chromosomes, two of which show a deep subterminal constriction) allow easy separation form the bearded irises, whereas in some extent they connect that genus with the ‘Evansias’ and the ‘Tectores irises’ (cf. Simonet 1934).

Ledebour (1852) had already noted the similarity of Pardanthopsis and Belamcanda , particularly in the branching patterns and leaf morphology (reviewed in Lawrence 1953) when transferring I. dichotoma to Pardanthus Ker Gawler (1804: 246) , a later synonym of Belamcanda . Although Pardanthopsis and Belamcanda could equally be treated at generic rank or be merged in a single genus (which would retain the name Belamcanda ), we prefer maintaining Pardanthopsis in its own, on the basis of its strong morphological divergence, as suggested by Chimphamba (1973a) and others ( Goldblatt et al. 1998).

Adanson (1763) described the genus Belamcanda (as “ Belamkanda ”), to which Medikus (1790) referred the Linnaean Ixia chinensis , a species that later was used by Ker Gawler (1804) as the basis of the synonymous genus Pardanthus Ker Gawl. The genus is morphologically outstanding, and fairly different from the typical irises ( Figs. 3, 7). Even its common name among horticulturists does not recall it as a member of Iris , and consequently most authors have traditionally classified Belamcanda far from Iris , as an independent genus in “ Sisyrinchieae ” ( Bentham & Hooker 1883; Pax 1888, Baker 1892, Diels 1930) or in “ Cypelleae ” ( Klatt 1866). Only recent molecular analyses ( Tillie et al. 2001, Wilson 2004, 2009, 2011; see also Fig. 1) have revealed a close relationship with “bearded species” of the Iris core (I. subg. Iris ), and therefore formal reinclusion in a recircumscribed I. subg. Pardanthopsis (Hance) Baker has been proposed ( Wilson 2011). In fact, Goldblatt & Mabberley (2005) transferred Belamcanda to the genus Iris , and established the combination Iris domestica ( Linnaeus 1753: 952) Goldblatt & Mabberley (2005: 129) , based on the earliest available name Epidendrum domesticum L. Indeed , the name ‘ I. chinensis ’ had already been used (cf. Curtis 1797: t. 373) for a different plant currently synonymised to I. japonica Thunberg (1794: 327) . Curiously, Epidendrum domesticum is the basionym of Vanilla domestica (L.) Druce (1914: 425), and is lectotypified with an incomplete illustration by Kaempfer (1712) that was said to correspond to

A NEW GENERIC ARRANGEMENT OF IRIS SENSU LATISSIMO

Phytotaxa 232 (1) © 2015 Magnolia Press • 17 Belamcanda chinensis ( Garay 1997) . However, a careful examination of the lectotype and the accompainying Latin description do not fully support this conclusion, and the issue is in need of reappraisal.

In a narrow sense, Belamcanda includes one species occurring in eastern Asia, Philippines and Japan ( Mathew 1989a, Goldblatt et al. 1998) ( Fig. 9B), though it is cultivated worldwide and sometimes naturalised. As said above, chemical composition also supports a close relationship among Belamcanda , Pardanthopsis and some bearded irises (Harborne & Williams 2001, Kaššak 2012, Wei et al. 2012), albeit other compounds are unique to Belamcanda ( Liu et al. 2012, Monthakantirat et al. 2005). Belamcanda chinensis and Pardanthopsis dichotoma share the same chromosome number (2n = 32) (see Chimphamba 1973a, among others), though the karyotype analysis showed differences in the chromosome type and structure ( Park et al. 2006, but see Chimphamba 1973a). Recent counts for Pardanthopsis subdichotoma (Zhao 9180: 57) M.B.Crespo, Mart.-Azorín & Mavrodiev (see below) revealed 2n = 42 ( Shen et al. 2007).

Remarkably, Belamcanda chinensis and Pardanthopsis dichotoma can mutually interbreed to produce × Pardancanda norrisii Lenz (1972a: 407) , but do not form hybrids with other irises groups. Chimphamba (1973a) and Bi et al. (2012) found that the F 1 and F 2 hybrids between those species were conservative in chromosome number and karyotype features, compared with their parents. Those hybrids were sterile when B. chinensis acted as female progenitor, albeit in the opposite crossings they showed high fertility rates. Interestingly, Bi et al. (2012) identified both pre-zygotic (e.g., non-germinating pollen grains and/or non-developing pollen tubes) and postzygotic barriers (e.g., embryo disintegration from micropyle) which resulted in sterility.

As shown above, Belamcanda shares a number of vegetative characters with Pardanthopsis , which could favour inclusion of the latter within the former. However, that solution though possible on molecular grounds would render an unnecessarily heterogeneous expanded Belamcanda , difficult to define on morphological basis. Important differences exist that allow easy separation of Belamcanda from the rest of irises ( Mathew 1989a, Goldblatt et al. 1998, Zhao et al. 2000). Flowers are actinomorphic, and tepals are completely free, lacking the typical 3-meranthium close structure, with cylindrical and short styles ( Figs. 3D, 7D); the style is not embedded in hypanthium tissue; fruits completely dehisce from apex to base, and seeds are globose, blackish, glossy, smooth and long-persisting on capsule axis after dehiscence ( Fig. 6A) resembling a blackberry, an apomorphy that is likely related to ornithochory ( Goldblatt & Mabberley 2005). Due to the fact that this is a unique combination of characters, Belamcanda is favoured here at the genus rank, placed together with the related Pardanthopsis .

18 • Phytotaxa 232 (1) © 2015 Magnolia Press

CRESPO ET AL.

The close relationship of Belamcanda and Pardanthopsis clearly requires further investigation. Relatives of both genera may be regarded as a good illustration of what Wheeler & Blackwell (1984: 12–13) referred as “ghost taxa” (namely unsampled extinct, undiscovered, taxonomically misplaced, unknown or “ghost” taxonomical entities). However, other explanations are also possible. Chimphamba (1973a) suggested the putative ancient hybrid origin of both Belamcanda and Pardanthopsis from two common progenitors and mentioned the tribe Triticeae ( Poaceae ) for similar precedents. According to this author, the alloploid origin of both Belamcanda and Pardanthopsis seems to be a more applicable alternative than the possibility of their phyletic origin from a common ancestor. Future nuclear phylogeny of the Iris s.l. clade may help to clarify an actual impact of reticulation in the natural history of the whole complex. The relatively high haploid number for both genera (x = 16) and their regular bivalent meiosis would point out to a similar allopolyploid origin from parents with x = 8, a number which is found in some groups of Iris s.l. Therefore, genomes in Belamcanda and Pardanthopsis , though being similar at first stages, subsequently would have diverged independently along with morphological (mostly reproductive) features ( Chimphamba 1973a).

The Iris s.str. clade (Bearded irises or Pogon irises)

This group includes a morphologically rather compact assemblage of taxa which show falls and standards of a similar size and shape, at least the latter being bearded ( Figs. 3E–H, 10). This is the main reason for they have usually been referred as the “bearded irises” or “pogon irises”. Other important shared morphological synapomorphies are the stigmatic lip rounded and entire (neither notched nor bifid), the capsules dehiscing through three clefts from below the top which is somewhat pointed, and the seeds large and globose, with or without a fleshy appendage ( Figs. 6Q–R). Over 100 species are currently included here, which grow through Europe, Asia and northern Africa ( Figs. 11–12). The most probable values of p are 8 and 12, though p = 10 and 11 have also been suggested (Crespo et al. 2013).

Up to five groups have been accepted at different taxonomic ranks, which are morphologically rather well defined and are mostly recovered as monophyletic ( Fig. 1). On the one hand, the typical Iris sect. Iris includes plants with stout short-branched rhizomes, many-flowered to rarely 1-flowered scapes, falls with a dense beard of long claviform hairs ( Figs. 3E, 10A), and seeds lacking fleshy appendages ( Dykes 1912, Rodionenko 2009) ( Fig. 6R). It has also been referred as I. subg. Pogoniris Spach (1846a: 48, Spach 1846b, Baker 1892), I. sect. Pogiris Tausch (1823: without pagination) or I. sect. Pogoniris Baker (1876b: 647, Lynch 1904, Diels 1930, Fedchenko 1935), or I. sect. Euiris subsect. Pogoniris (Baker) Bentham & Hooker (1883: 687 , Pax 1888). Its biodiversity centre is probably located in southern Europe ( Fig. 11), where many species have been demonstrated to have a hybrid origin ( Simonet 1934, Colasante et al. 2001), and allopolyploidy occurs frequently ( Simonet 1951, Colasante & Vosa 2001, Williams et al. 2001). On the other hand, the remaining four groups are characterised by their seeds with fleshy appendages. For that reason, they were grouped in I. subg. Arillosae Rodionenko (2009: 433), a group that however was not recovered as monophyletic ( Mavrodiev et al. 2014). First, I. sect. Oncocyclus includes the well-known “ Oncocyclus irises”, which produce reddish rhizomes with crowded growths, unbranched stems always carrying a terminal flower, tepals showing a rich-veining coloured pattern with a blackish mark (signal patch) at the basal part, an irregular band of short scattered hairs on the falls ( Figs. 3F, 10D), and the fleshy appendage usually larger than the seed itself ( Dykes 1912). It is the broadest group in the clade and it was described as the independent genus Oncocyclus Siemssen (1846: 706 , Klatt 1866) and later treated as I. subg. Susiana Spach (1846a: 70) or I. subg. Oncocyclus (Siems.) Alefeld (1863: 296, Baker 1877a, 1892), I. sect. Oncocyclus (Siems.) Baker (1876e: 787 , Lynch 1904, Dykes 1912, Diels 1930, Rodionenko 2009, Mathew 1989a) or I. subsect. Oncocyclus (Siems.) Bentham & Hooker (1883: 687 , Lawrence 1953, Pax 1888, Rodionenko 1961). It is distributed mainly in the Middle East ( Fig. 12B), and shows a homogeneous chromosome number 2n = 20 (Avishai & Zohary 1977), it being a group in which genetic barriers among species in this clade have been demonstrated to be weak (Avishai & Zohary 1980). Secondly, I. sect. Regelia corresponds to the “ Regelia irises”, a group morphologically close to I. sect. Oncocyclus from which it differs by the usually 2–3-flowered inflorescences ( Fig. 3D), a more conspicuous linear band of hairs on all six perigone segments (falls and standards) and the smaller fleshy appendage of seeds ( Fig. 6Q), among other features ( Mathew 1989a). This morphological similarity is corroborated in our analyses, the Regelias being sister to the Oncocyclus irises. Similarly, both groups show close distribution areas, though the Regelias occur eastwards to central Asia ( Fig. 12C). Although it has been

A NEW GENERIC ARRANGEMENT OF IRIS SENSU LATISSIMO

Phytotaxa 232 (1) © 2015 Magnolia Press • 19 20 • Phytotaxa 232 (1) © 2015 Magnolia Press

CRESPO ET AL.

usually treated as a section ( Lynch 1904, Dykes 1912, Diels 1930, Mathew 1989 a, Rodionenko 2009), sometimes it was referred as I. subg. Regelia Foster ex Baker (1892: 20) , I. subsect. Regelia (Foster ex Baker) Rodionenko (1961: 198) . Both groups, the Oncocyclus plus the Regelia irises, are well-supported as sister to the typical I. sect. Iris , with which they share many morphological characters. Thirdly, I. sect. Pseudoregelia ( Dykes 1912, Diels 1930, Mathew 1989 a, Rodionenko 2009, Taylor 1976), also known as I. subsect. Pseudoregelia (Dykes) Lawrence (1953: 356, Rodionenko 1961), I. subg. Pseudevansia Baker (1892: 2) or I. sect. Pseudevansia (Baker) Lynch 1904: 55), forms a strongly supported clade in which plants show compact rhizomes not stoloniferous, narrower standards with blunt edges, a dense beard of club-shaped hairs only on falls which are markedly blotched darker ( Figs. 3G, 10E), and seeds with a smaller fleshy appendage resembling a flat ring ( Dykes 1912, Mathew 1989a). They are widely distributed in eastern and sutheastern Asia ( Fig. 12D) and, in some extent, resemble at first sight to some oriental irises (the Evansias, see below), though their overall morphology is more consistent with the Regelias. They are sister to the clade formed by all previous sections, though this relationship is not supported in our analyses. Lastly, I. sect. Psammiris (Spach 1846, Taylor 1976, Mathew 1989 a, Rodionenko 2009) is a small group, though widely distributed in eastern and sutheastern Asia ( Fig. 12A). It is strongly supported in our results as sister to all other sections of Iris with which it shares morphological features. Flowers resemble those of I. sect. Iris by the nature of the beard on falls ( Figs. 3H, 10B), but seeds show a fleshy appendage which is much smaller than in the Oncocyclus irises, the stouter rhizomes are shortly creeping or stoloniferous and the flowers twist up spirally while withering as in the Regelia irises ( Dykes 1912, Mathew 1989a). For those reasons, it has sometimes been regarded ( Dykes 1912) as the connecting link between I. sect. Regelia (Foster ex Baker) Lynch and I. sect. Iris . Most of the cited characters should therefore be considered as plesiomorphic in the whole group. According to the morphological divergence among those groups, and accepting the existence of frequent hybrids between members of all them (the so-called ‘Oncobreds’, ‘Oncoregelias’, ‘Oncogelias’ or ‘Regeliocyclus irises’, among others), we suggest treating those groups as sections in Iris . Although I. falcifolia Bunge (1852: 329) (I. sect. Hexapogon Bunge 1852: 329 ≡ I. subg. Hexapogon Bunge ex Alefeld 1863: 296) was included in the

A NEW GENERIC ARRANGEMENT OF IRIS SENSU LATISSIMO

Phytotaxa 232 (1) © 2015 Magnolia Press • 21 comprehensive molecular analyses of Wilson (2011), its final position among members of I. subg. Scorpiris (the Juno irises ) was questioned by Ikinci et al. (2011) based on the sampling. That species is indeed a bearded iris not related to the Junos, which actually resembles I. sect. Regelia , a reason which led Rodionenko (1961) to synonymisation of I. sect. Psammiris and I. sect. Regelia . In our analyses, the monophyletic I. sect. Hexapogon appeared as presumably sister of the strongly supported and essentially monophyletic I. sect Oncocyclus ( Fig. 1), but due to the preliminary position of Iris iberica Steven in Marschall von Bieberstein (1808: 30) ( Fig. 1) further investigation is necessary to clarify relationships between the Hexapogon and Oncocyclus irises.

The Juno clade ( Juno irises or Scorpio irises)

Taxa included in this group display an outstanding combination of morphological characters which allow easy identification, even in vegetative state (e.g., Rodionenko 1961, 1994, 2013). The Juno irises have characteristic bulbs ( Figs. 2C–D), formed with few scales attached to a reduced basal plate that produces thickened fleshy living roots, permanently visible during the whole year ( Dykes 1912, Mathew 1997, 2001, Rodionenko 1994). Leaves are truly bifacial, a character apparently unique among Iridaceae (Arber 1921, Rudall & Mathew 1993), usually glossy on the upper side and paler on the lower, and they are distichously arranged in such a way that commonly resemble the stem of a corn plant, showing often a cartilaginous or hyaline margin ( Mathew 1997). Flowers are borne at the axil of well-developed leaves and are characterised by showing erect-patent falls with a central ridge or entire crest, which belongs to type II of Guo (2015), sometimes accompained with one pair of low lateral ridges at the base. The standards are short and inconspicuous, and are placed patent to deflexed with regard to the flowers axis ( Figs. 3I, 13A), or exceptionally they are larger and slightly subpatent only in J. cycloglossa ( Wendelbo 1958: 187) Soják (1980: 79) (cf. Mathew 1997, 2001, Rodionenko 1994). These small standards are similar to the reduced ones of Limniris pseudacorus ( Linnaeus 1753: 38) Fuss (1866: 636) or L. setosa (Pallas ex Link 1820: 71) Rodionenko (2007: 552) , however they are mostly patent or bent downwards ( Rodionenko 1994) and perhaps supposed to act as gutters which allow the rain falling into the flower to drip off at their tips ( Mathew 2001). Pollen features are also unique, the grains of most species showing a clypeate type, spheroidal shape and eu-reticulate sculpturing of exine ( Schulze 1971 b, Oybak-Dönmez & Pinar 2001, Pinar & Oybak-Dönmez 2000). However, two additional models exist ( Rodionenko 1961, Ikinci et al. 2011). First, zone-aperturate grains with the reticulate exine

22 • Phytotaxa 232 (1) © 2015 Magnolia Press

CRESPO ET AL. A NEW GENERIC ARRANGEMENT OF IRIS SENSU LATISSIMO

Phytotaxa 232 (1) © 2015 Magnolia Press • 23 divided into two hemispherical halves, and secondly spiny grains with a spiral or winding aperture, the incomplete exine exhibiting club-like protuberances. The latter is only found in two Mediterranean species, J. planifolia ( Miller 1768: without pagination) Ascherson (1864: 114) and J. palaestina (Baker 1871: 108) Klatt (1872a: 498) , for which J. sect. Acanthospora Rodionenko (1961: 209) was established. Seeds have a whitish aril in some species ( Fig. 6S), lacking in others. The most probable values of p are 8 and 9, though p = 11 has also been suggested (Crespo et al. 2013).

The remarkable originality of taxa in this group led many authorities to classify them apart from other irises at different ranks. They were included in Xiphion along with other bulbous irises ( Miller 1768), sometimes as X. subg. Juno (Tratt.) Baker (1877a: 123) , or even treated as an independent genus to which the names Juno (Trattinnick 1821, Rodionenko 1961, 1994, Kamelin 1973, Soják 1982, Vvedensky 1963, 1971, Crespo et al. 2013), Thelysia Salisb. ex Parlatore (1860: 516) , Costia Willkomm (1860: 131) nom. illeg., or Coresantha Alefeld (1863: 298, Klatt 1866) were applied, the former having priority (Ascherson 1864). Since the beginning of the 20 th century, however, reinclusion into Iris has been usually adopted as I. sect. Juno (Tratt.) Maximowicz (1880: 505) ( Bentham & Hooker 1883: 687, Lynch 1904, Dykes 1912), I. subg. Scorpiris Spach (1846a: 16) ( Mathew 1997, 2001, Hall et al. 2001, Wilson 2011, Ikinci et al. 2011), or I. subg. Juno (Tratt.) Baker (1892: 44) .

The Juno irises form a clade that is sister to the Evansia s.l. clade (incl. Junopsis = I. subg. Nepalensis ) ( Fig. 1; see also Tillie et al. 2001, Wilson 2011), and both are connected to the large assemblage constituted by the “bearded irises” plus Belamcanda , Pardanthopsis and Gattenhofia . Morphological features supporting them were clearly shown by Hall et al. (2001). Kamelin (1973) correctly argued that Juno constitutes a lineage close to Junopsis which evolved from rhizomatous irises, though independently from the rest of bulbose irises. However, our results do not support Kamelin’s (1973) assumptions about evolutionary relationships between Juno and Hermodactylus (see below).

The internal relationships found within the Juno clade by Mavrodiev et al. (2014) are similar, but not fully coincident with those discussed by Ikinci et al. (2011) and Guo & Wilson (2013). Two main strongly supported subclades arise which correspond well to J. sect. Physocaulon Rodionenko (1961: 208) and J. sect. Juno (incl. Acanthospora and Wendelboa ) ( Fig. 1), the latter with several subclades not fitting the sectional and infrasectional classification by Rodionenko (1994) and needing further investigation (for an extensive discussion see Ikinci et al. 2011). Guo & Wilson (2013) have recognised four principal aggregates, namely clades Rosenbachiana, Bucharica, Aucheri, and Magnifica , the former corresponding to J. sect. Physocaulon and the remaining to our J. sect. Juno .

In accordance with the data shown above, Juno is regarded here at the genus rank. Following Komarov’s accurate concept of species as a morphologically defined geographical race (reviewed in Juzepcuk 1939) which was widely accepted in the classical treatment of “Flora of USSR”, Kamelin (1973) correctly stated that the number of taxa in Juno was underestimated, it probably including at least 48 species. In the generic arrangement favoured here, Juno groups ca. 60 species, occurring mainly in southwestern Asia and the Caucasus, one of them ( J. planifolia ) extending into the western Mediterranean basin through northern Africa ( Fig. 9C). Further morphological work is needed for estimation of either a more accurate infrageneric classification or even the circumscription of the whole genus.

The Evansia clade (Evansias or Bamboo-like irises)

This group of Asian irises is treated here in a narrow sense to include species with stout, many-branched, nodose rhizomes, sometimes bearing slender long creeping stolons; roots fibrous, slender; stems conspicuous, usually bamboo-like, many branched; spathe valves 3–6; perigone pieces with markedly erose-dentate, undulate margin, gradually tapering into a inconspicuous short claw, the falls bearing a central crest of a single row, coloured, usually fimbriate ( Figs. 3J, 13B), which belongs to type II of Guo (2015) and is flanked by two pairs of lateral outgrowths on both sides at the base; capsule with pericarp subcoriaceous; seeds numerous, angulose, ± flattened, irregular in shape, with small whitish aril ( Fig. 6E); testa ± corky, brownish, irregularly wrinkled, matte. At least 5 species occurring in eastern Asia belong to this group ( Fig. 14A). The known chromosome numbers are 2n = 28, 30, 32, 36, 54 and 56 ( Yu et al. 2009, Zhao et al. 2000), with probable p = 7 and 8. Additional counts 2n = 24, 31, 33, 35 and 54 have also been reported ( Brearley & Ellis 1997), albeit they most likely correspond to not truebreeding plants of probable horticultural and/or hybrid origin.

24 • Phytotaxa 232 (1) © 2015 Magnolia Press

CRESPO ET AL.

Taxa in this clade have usually been classified together with the ‘Tectores irises’ (here segregated as the new genus Tectiris ) and sometimes with the ‘Nepalensis irises’ (here treated as Junopsis ). Regarding the ‘Evansias’, they usually were treated in a wide sense and considered at different ranks with very diverse circumscriptions. Salisbury (1821) first invalidly published the generic name Evansia , later validated as Xiphion subg. Evansia Alefeld (1863: 297) . Then, it was accepted at the genus rank as Evansia (Alef.) Salisb. ex Decne. by Decaisne (1874) and Klatt (1882), or as Iris subg. Evansia (Alef.) Baker (1877a: 143) , though Spach (1846) had first applied the subgeneric rank to this group as I. subg. Crossiris Spach. The latter epithet was adopted by Rodionenko (1961) with a new wider circumscription including I. sect. Crossiris ( Tausch 1823: without page) Rodionenko (2009: 434), I. sect. Lophiris Tausch (1823 : without page) (treated here at generic rank) and I. sect. Monospatha (included here in the new genus Rodionenkoa ), the former with two series, I. ser. Japonicae Rodionenko (1961: 192) ( Evansia s.str.) and I. ser. Tectores Rodionenko (1961: 194) (here raised to the new genus Tectiris ). Other authors accepted Salisbury’s genus as I. sect. Evansia (Alef.) Baker (1876a: 36, Dykes 1912), or I. subsect. Evansia (Alef.) Bentham & Hooker (1882: 687, Pax 1888), the latter rank being followed in the classification of Lawrence (1953). Most of recent treatments follow Rodionenko’s arrangement, though including his three sections within a single I. sect. Lophiris in a widely circumscribed I. subg. Limniris . Nonetheless, many authors have usually segregated the ‘Nepalensis irises’ as an autonomous section (cf. Lynch 1904, Dykes 1912, Diels 1930), subgenus ( Lawrence 1953, Rodionenko 1961, 2009, 2013, Innes 1997, Mathew 1989 a, Goldblatt & Manning 2008) or genus (Schulze 1970), on the basis of notable morphological differences (see below).

In our analyses ( Fig. 1), Evansia (s.str. = “Crossiris”) is monophyletic, as successive sister to the clades formed by the Nepalensis (conventional phylogenies) or to the clade (Nepalensis + Tectores irises) (3TA topology) clade. The resulting group of “Evansioid irises” or Evansias ( Evansia s.str. + Nepalensis + Tectores irises) is strongly supported as sister to the Juno irises (see also Wilson 2011, and Guo & Wilson 2013). This relationship is supported by a similar general crest morphology, belonging to type II ( Guo 2015), though it being less complicated and perhaps more primitive with regard to other more elaborated crest structure present in the “Evansioid irises”

A NEW GENERIC ARRANGEMENT OF IRIS SENSU LATISSIMO

Phytotaxa 232 (1) © 2015 Magnolia Press • 25 (see below). It is fairly important that the Evansias (named “subg. Limniris V” by Wilson 2011) and the Nepalensis were placed as strongly supported sister groups, though the latter authors ( Guo & Wilson 2013) did not recover a monophyletic I. ser. Tectores, with I. milesii Baker ex Foster (1883: 231) merged with Evansia s.str. and I. tectorum Maximowicz (1871: 380) as an independent though weakly supported, isolated clade (see below for further details).

Although the Evansias, the Nepalensis and the Tectores irises share some morphological traits (i.e. the prominently 1–5-ribbed, subcoriaceous leaves or the stigma bilobed with broad lobes), strong divergences exist mostly concerning the general habit, rootstock characteristics, perigone pieces, nature of fall crest, pollen, fruit and seed features ( Hall 2013, Guo 2015, Guo & Wilson 2013, Mathew 1989 a, Rodionenko 2009, Schulze 1970, and references therein). In fact, Rodionenko (1961) suggested that both series in I. sect. Crossiris (I. ser. Japonicae and I. ser. Tectores) were “ecologically and genetically heterogeneous”, which agrees with their position as non-sister clades in our trees. Simonet (1934) has reported interesting differences in the chromosomal arrangement (both number and karyotype) of Evansia s.str. with regard to the Nepalensis and Tectores irises. Similarly, chemical constituents of Evansia are significantly diverse from those of the Tectores (cf. Kaššak 2012, Kukula-Koch et al. 2014, Wang et al. 2010) to support the autonomous position of Evansia s.str. Accordingly, we recognise here Evansia at the generic rank, though in a narrower circumscription than previous authors, excluding the Tectores irises.

Nonetheless, internal relationships of Evansia s.str. are still in need of further investigation. One of the most enigmatic species in this group is Iris formosana Ohwi (1934: 115) , a poorly known species from Taiwan which according to the protologue is close to I. japonica . However, the description of this species in Zhao et al. (2000), as well as the data, illustration and photograph in Ying (1987, 2000), do not fit well the protologue, and the plant they describe could even correspond to a still undescribed taxon of the ‘Evansioid irises’ as treated here. Further research is therefore needed to clarify this point.

The Junopsis clade (Nepalensis irises)

Horticulturists usually include under the name ‘Nepalensis ’ a distinct group of irises well characterised by their peculiar small, inconspicuous rhizome, surrounded by many fleshy, tuberous roots which makes the plants form dense tufts ( Fig. 2G); stems simple or 1-branched, usually shorter than leaves ( Fig. 13C), sometimes inconspicuous or absent; perigone pieces with flat or slightly undulate margins, the falls with a crest of a single row, mostly with coloured fimbriae, belonging to type III ( Guo 2015), and the standards narrowed into a subcanaliculate, long haft ( Figs. 3L, 13C); capsule papery with prominent ribs, and conspicuously arillate seeds ( Fig. 6O). It is a compact aggregate with at least 8 species occurring in central Asia ( Fig. 9D), south of the main Himalayas; it is still a poorly known group in need of further studies. Chromosome data are still scarce, with counts of 2n = 28, 34 and 36 ( Simonet 1934, Malla et al. 1981), suggesting p = 7 or 9.

Foster (1892) and Lynch (1904) highlighted morphological closeness between I. decora Wallich (1830: 77) (as I. nepalensis Don 1825: 54 nom. illeg.) and Juno , opposite to other authors (e.g. Baker 1892) who placed the former within I. sect. Evansia (I. sect. Crossiris). Their reasoning was based on the similarities of the almost bulbous rootstock with fleshy roots, and some floral features. Accordingly, Lynch (1904: 57, 192) described I. sect. Nepalensis to place those Himalayan irises, noting that “the type of this section [‘ I. nepalensis ’], which has descended probably from the same ancestors as the Juno Irises , but on another line of development”. The section was accepted by Dykes (1912) and Diels (1930). More recently, Lawrence (1953) raised it to subgenus rank, and later Schulze (1970) treated it as the genus Junopsis , pointing out to the above cited relationships to the Juno irises . These connections were also highlighted by Kamelin (1973), who identified the Nepalensis irises as one of the evolutionarily closests relatives of Juno . However, Lawrence’s subgeneric treatment has widely been adopted by most taxonomists ( Rodionenko 1961, 2013, Innes 1997, Mathew 1989 a, Goldblatt & Manning 2008) so far.

Either conventional or 3TA topologies ( Fig. 1) placed the Nepalensis irises related to Evansia , a result that is congruent with previous molecular work, as said above. Some morphological traits shared with the Evansias (see above) support this placement. Baker (1876a, 1892) and Klatt (1882) had merged both groups, either as I. subg. Evansia or the genus Evansia . Although it has usually been accepted at subgeneric rank, Schulze (1970, 1971b) highlighted morphological divergences (i.e. the rootstock, stem, flower, fruit and seed characteristics) allowing

26 • Phytotaxa 232 (1) © 2015 Magnolia Press

CRESPO ET AL.

easy segregation at generic rank. Among them, pollen morphology was noticeable, and related Junopsis with Juno , this being congruent with their phylogenetic position. Simonet (1934: 271) also reported chromosomal peculiarities for I. collettii (e.g. 2n = 24, with chromosomes short, thickened, incurved, and nearly subreniform, of which 4 are longer and 2 bear satellites), which placed it in an isolated position with regard to the rest of ‘Evansioid’ irises. Similarly, the crest structure of the Nepalensis irises, as described by Guo (2015), also brings additional differences with regard to other ‘Evansioid’ lineages. In Junopsis , it includes a single central ridge usually fringed (Guo’s type III), not flanked by lateral ridges or outgrowths, which probably corresponds to a primitive structure with regard to the more elaborated crests of other related lineages (see below).

Therefore, based on the evidence shown above we here accept Junopsis at the generic rank (Schulze 1970), including the morphologically deviant Iris staintonii Hara (1974: 203) . The sistership position of the latter species is not conventionally resolved ( Fig. 1), albeit the overall morphological congruence with the Nepalensis irises recommends inclusion in that group ( Innes 1997, Hall 2013). Nevertheless, it clearly represents a diverging lineage with morphological peculiarities (e.g. the almost absent stems, short leaves of which one elongates later, small flowers with a vestigial low crest on falls, etc.), which is treated here at the sectional rank, as the new Junopsis sect. Staintonia (see below). Additional data are still needed for a more accurate taxonomical placement of this miniature iris.

The Tectiris clade (Tectores irises)

This group includes two distinct though morphologically close species, usually related to Evansia ( Fig. 1) with which they share a similar vegetative habit and some floral characters. However, important differences are found that also connect them with Junopsis . The rhizome is conspicuous, stout, lacking stolons and fleshy roots, and grows horizontally near the soil surface ( Guo & Wilson 2013); stems are branched and bear several leaves, sometimes scale-like ( Fig. 13D); flowers have a short perigone tube, with falls erose-undulate on margins and bearing a central crest of a single row ± irregularly toothed, flanked on both sides by three additional pairs of lateral ridges at the base; standards are narrowed in a long, canaliculate haft ( Fig. 3K); capsule papery with apiculate, globose-pyriform seeds bearing a small whitish aril and with a blackish, glossy, hard testa ( Fig. 6D). Both species in this clade are found in southeastern Asia and Japan ( Fig. 14B), and show chromosome numbers 2n = 24, 26, 28 and 32 ( Simonet 1934, Chimphamba 1973b), probably with p = 8.

Traditionally, the Tectores irises have been considered to belong to the Evansias (Alefeld 1863, Decaisne 1874, Klatt 1882, Baker 1877a, 1892, Spach 1846, Dykes 1912, Lawrence 1953, Rodionenko 1961, Mathew 1989a), though without any distinct treatment. Rodionenko (1961) was the first in recognising Iris ser. Tectores Rodion. to include I. tectorum and I. milesii , based on differences in the rizhome features.

Ikinci et al. (2011) found both species in I. ser. Tectores to form a strongly supported clade, apart from the Evansias and the Nepalensis, which constitute distinct clades among other Apogon irises. They however did not get resolution enough to visualise relationships among the three ‘Evansioid’ irises clades. Conversely, parallel analyses by Guo & Wilson (2013) found the latter groups as sister clades, albeit did not recover a monophyletic I. ser. Tectores. The latter authors suggested that morphology of their original accession of Iris milesii (Guo CH 10-30 RSA) from southwestern China “differs somewhat from the description of the type specimen” and therefore Guo’s collection “might represent a new species” ( Guo & Wilson 2013: 993) of Evansia s.str., and not to I. ser. Tectores (our Tectiris ). This is also congruent with its crest type II ( Guo 2015) very close to that of Evansia s.str. and strongly different form Tectiris . This fact would explain those conflicting results. Our trees ( Mavrodiev et al. 2014) resolved the position of the three clades of the ‘Evansioid’ irises in different ways, Tectores being sister to the clade ( Evansia + Junopsis ) (in the conventional topologies), or to Junopsis only (in the 3TA topology) ( Fig. 1).

To some extent, the Tectores irises show morphological traits that are not found in either of its closest relatives (i.e. the long perigone tube, standards with canaliculate long haft, or seeds with blackish, glossy and hard testa), which warrant differentiation. In particular, pollen grains in I. tectorum show two longitudinal folds like those in Pardanthopsis ( Rodionenko 1961) , the rizhomes grow horizontally and are subsuperficial, and the crests on falls are prominent, dissected and show unique anatomical features ( Guo & Wilson 2013), they being flanked at the base by three pairs of outgrowths ( Guo 2015) which is likely an apomorphy not found in the other ‘Evansioid’ irises. Similarly, chemical constituents of I. tectorum are shared with representatives of the bearded irises (I. sect. Iris and

A NEW GENERIC ARRANGEMENT OF IRIS SENSU LATISSIMO

Phytotaxa 232 (1) © 2015 Magnolia Press • 27 I. sect. Oncocyclus ), Belamcanda chinensis and Pardanthopsis dichotoma ( Fang et al. 2007, Kaššak 2012, Wei et al. 2012, Kukula-Koch et al. 2014), which supports its phylogenetic connections to the clade of the ‘bearded irises’. However, many other compounds are unique or significantly diverse with regard to Evansia and Junopsis ( Boland & Donnelly 1998, Kaššak 2012), as well as the chromosome numbers and karyotype structure ( Simonet 1934, Chimphamba 1973b, Park et al. 2006, Yu et al. 2009). In fact, the chromosome numbers are lower in this group than in Evansia and Junopsis , which is congruent with its isolated basal position. All those data support the generic treatment of this clade, for which the name Tectiris is proposed.

In summary, the ‘Evansioid’ irises still clearly require further investigation ( Guo & Wilson 2013) (see, for example, the lack of the conventional support for the Tectiris clade or the conventionally unresolved sistership of Iris staintonii ). However the strong congruence of all of the three “molecular”-based patterns (Tectores irises, Nepalensis irises, and Evansia irises) ( Fig. 1) with either morphology or geographical distributions (the “fourth parallelism ” sensu Williams & Ebach (2004) , as mentioned in Mavrodiev et al. 2014), encourage us to accept all those three groups at the genus rank, with the inclusion of I. staintonii in Junopsis (see below).

The Limniris clade (Beardless irises or Apogon irises)

This group is treated here in a narrow sense to include those “beardless irises” well characterised by a rhizome usually thick, branched and clothed with dark remains of leaves; flowers with perigone segments differing in size and shape, and fused in a cylindric or cup-shaped tube; falls smooth or loosely pubescent (neither crested nor bearded); stigma triangular to linguliform, entire or sometimes shallowly emarginated; pericarp coriaceous; and seeds flattened, lacking aril and with testa usually ± corky, smooth or wrinkled ( Figs. 2E, 4A–C, 6M, 15). So defined, it includes ca. 50 species widespread in Eurasia, northern Africa and North America ( Figs. 16–17), though absent in the northernmost areas. The probable values of p are 8, 9, 10, 11 and 12 (Crespo et al. 2013), and frequent polyploidy and dysploidy processes are registered which sometimes give rise to deviant number counts ( Dyer et al. 1976).

The circumscription of this group has been very variable. Tausch (1823) first proposed Iris sect. Limniris to include Iris pseudacorus Linnaeus (1753: 38) and I. sibirica , a group that Spach (1846a,b) treated later as I. subg. Limniris and Alefeld (1863) combined as Xiphion subg. Limniris Alefeld (1863: 297) . It has later been arranged in a wider sense as I. sect. Apogon Baker (1876c: 143, Dykes 1912), I. subg. Apogon Baker (1877a: 137) , or also I. subsect. Apogon Bentham & Hooker (1882: 687, Lawrence 1953). The three latter infrageneric taxa included a highly heterogeneous aggregate of natural groups, which share the smooth condition of falls, lacking any sort of crest or beard, and the rhizomatous rootstock. The first comprehensive approach was published by Dykes (1912) who recognised 15 informal groups within I. sect. Apogon . It was later slightly modified by Lawrence (1953) by naming those groups as series in an expanded I. sect. Spathula Tausch (1823 : without page), plus a nomenclatural change and the addition of the new I. ser. Hexagonae ( Diels 1930: 502) Lawrence (1953: 362). The subsequent treatment by Rodionenko (1961) established recognition of I. subg. Limniris and I. subg. Xyridion (Tausch) Spach as independent, the former being regarded later ( Rodionenko 2007) as genus Limniris with an even wider circumscription. This generic status had previously been favoured by Reichenbach (1841), Fuss (1866) and Fourreau (1869). Most authorities have commonly accepted Lawrence’s classification with posterior changes introduced by Mathew (1989a), which therefore rendered an I. subg. Limniris with 2 sections and 16 series, it being the widest and more complex aggregate in Iris s.l.

Recent analyses have demonstrated that subg. Limniris sensu Rodionenko (1961) and Mathew (1989a) is clearly polyphyletic ( Fig. 1; see also Wilson 2009, 2011). The so-called “core Limniris ” (or “subg. Limniris I”) encompassed subsections Sibiricae Diels (1930: 501), Californicae, Tripetalae Diels (1930: 502), Laevigatae, Prismaticae, Ensatae , Ruthenicae, Hexagonae and Chinenses , plus I. songarica (I. subsect. Tenuifoliae p.p.), some of them being non-monophyletic. The remaining five groups were formally treated ( Wilson 2011) as I. subg. Xyridion (clade “ Limniris II ”), I. subg. Lophiris (clade “ Limniris IV ”), I. subg. Crossiris (clade “ Limniris V”) and I. subg. Siphonostylis (clade “ Limniris VI ”), many of them being rearranged and circumscribed in either a narrower or a wider sense than previous authors. The clade “ Limniris III ” ( Wilson 2011) (I. subsect. Longipetalae plus I. sect. Monospatha) was not formally named, since it grouped an aggregate of taxa with clear morphological divergences which was resolved with no support.

28 • Phytotaxa 232 (1) © 2015 Magnolia Press

CRESPO ET AL. A NEW GENERIC ARRANGEMENT OF IRIS SENSU LATISSIMO

Phytotaxa 232 (1) © 2015 Magnolia Press • 29

Iris masia Dykes (1910: 99) (I. subsect. Syriacae ) was merged in a much expanded I. subg. Xiphion that also included I. subg. Hermodactyloides Spach (1846: 16) ( Iridodictyum plus Alatavia ), and Hermodactylus . Wilson’s (2011) final classification of Iris recognised eight subgenera plus an apparently non-monophyletic I. subg. Scorpiris , some of which resulted heterogeneous and difficult to define on morphological grounds. It however has supposed an important contribution to the comprehensive knowledge of the internal relationships of the ‘ Iris sensu latissimo ’ clade, with interesting taxonomical consequences. One of the most striking points is the narrower circumscription of I. subg. Limniris which renders a more understandable aggregate, though still morphologically difficult to characterise ( Wilson 2009).

The most recent treatment of genus Limniris was presented by Rodionenko (2007, 2013), with a new classification including three subgenera: L. subg. Limniris , L. subg. Hexagonae (Diels) Rodionenko (2007: 552) (I. subsect. Hexagonae Diels) and L. subg. Ioniris (Spach) Rodionenko (2007: 553) (I. subsect. Ruthenicae Diels plus I. subsect. Chinenses Diels). This new arrangement fits well the major subclades of Wilson’s “core Limniris ”. However, the “core Limniris ” is incompatible with segregation of genera Eremiris and Sclerosiphon ( Fig. 1), which Rodionenko (2006a,b) had previously proposed.

According to the available morphological and molecular data, Limniris is accepted here as an independent genus, strongly supported in our conventional trees ( Fig. 1), which matches both Rodionenko’s (2007) Limniris subg. Limniris and Wilson’s (2011) Iris subg. Limniris (excl. subsections Ensatae , Ruthenicae, Hexagonae and Chinenses , as well as I. songarica ), the rest of taxa in either classification being regarded as independent genera (see below). This solution accords better to Spach’s (1846a,b) first circumscription of I. subg. Limniris .

Internal relationships in the Limniris clade are well resolved in our trees ( Fig. 1). Limniris sect. Californicae (Diels) Rodionenko (2007: 551) (the ‘Pacific Coast irises’) is sister to a clade formed with the sisters L. sect. Limniris (the ‘ Sibiricae irises’) and L. ser. Chrysographes ( Simonet 1934: 371) Rodionenko (2007: 551) , both latter groups recovered as monophyletic. Morphological, karyological and biogeographical characters support treatment as closely related but taxonomically separate groups, both at sectional rank. Taxa in L. sect. Californicae are mostly distributed in mountain areas of western USA ( Fig. 16C), and are easy to recognise on the basis of a variety of characters. They are generally small and compact, with tough rhizomes and wiry roots; leaves are narrow and slender, and stems are mostly unbranched; flowers tend to show falls spreading horizontally, not sharply reflexed downwards ( Fig. 15B); and seeds irregular in shape, mostly angulose, usually wrinkled ( Lenz 1958, Mathew 1989a). Furthermore, 2n = 40 is the only known chromosome number in this group ( Simonet 1934).

30 • Phytotaxa 232 (1) © 2015 Magnolia Press

CRESPO ET AL.

Limniris sect. Limniris is usually defined in a wide sense to include taxa occurring in temperate areas of Eurasia, mostly in central and eastern Asia ( Fig. 16A), which produce many branched, stout, slender rizhomes; leaves grassy, deciduous, in close tufts; stems mostly hollow; and seeds smooth, usually flattened semicircular to cubical (Dykes 1913, Mathew 1989a). However, two clades can be recognised showing distinct distributions and chromosome numbers ( Grey-Wilson 1997a). First, a few species ( L. sibirica , L. sanguinea (Hornem. ex Donn 1811: 17) Rodionenko 2007: 551 , and L. typhifolia ( Kitagawa 1934: 94) Rodionenko 2007: 551 ) occurring mostly through central Europe, Siberia, northern and eastern China and Japan, which have 2n = 28 chromosomes ( Simonet 1934, Lu et al. 1994) and produce standards erect and incurved and falls with a reticulate pattern on claw ( Figs. 4C, 15A). Secondly, the remaining eight species grow mainly in southwestern China ( Fig. 16B), they show 2n = 40 chromosomes ( Simonet 1934, Zhao et al. 2000) and standards divergent or even subpatent and falls without a reticulate pattern on claw ( Figs. 4B, 15D) ( Fig. 1; see also Tillie et al. 2001, Wilson 2009, 2011, and Mizuno et al. 2012). Both latter aggregates were already segregated by Simonet (1934) and Rodionenko (2007) as the ‘ Limniris’ (= Sibiricae ) and ‘ Chrysographes’ irises. Provided the differences among all three aggregates as well as the existence of hybrids among them (the “Cal-Sibe irises”), here we propose treating them at sectional rank.

The clade ( Limniris sect. Californicae (L. sect. Chrysographes +L. sect. Limniris )) is sister to a clade constituted by L. sect. Laevigatae (Diels) Rodionenko (2007: 552) (plus I. subsect. Tripetalae) plus L. sect. Prismaticae (Diels) Rodionenko (2007: 552) ( Fig. 1). Differences between these latter two groups were clearly exposed by Dykes (1912) and Mathew (1987), and recognition at sectional rank also seems to be appropriate. In our analyses, I. subsect. Tripetalae (the ‘Tripetalous irises’) is not recovered as monophyletic, their members being nested in L. sect. Laevigatae ( Fig. 1; see also Tillie et al. 2001, Wilson 2009, 2011, Mizuno et al. 2012, and Wheeler & Wilson 2014). This is congruent with morphological traits, and supports Rodionenko’s (2007) treatment of the latter section. Indeed, the most remarkable diagnostic character of the ‘Tripetalous irises’ is perhaps the inconspicuous setiform standards ( Dykes 1912), a feature that is also found in most taxa of L. ser. Laevigatae such as L. pseudacorus , L. ensata or L. versicolor , which display an increasing trend to standard size reduction ( Fig. 15C). Therefore, an expanded treatment of L. sect. Laevigatae to include L. subsect. Tripetalae is favoured here, the resulting group being widespread in Eurasia and North America ( Fig. 17). Furthermore, Wheeler & Wilson (2014) have found that internal relationships in series of Limniris do not reflect geographic patterns, the Asian and North American members not forming regional clades. This finding refutes the hypothesis that Asian taxa are the earliest diverging lineage within the clade, and points out to past multiple exchanges between Asia and North America ( Wheeler & Wilson 2014).

A NEW GENERIC ARRANGEMENT OF IRIS SENSU LATISSIMO

Phytotaxa 232 (1) © 2015 Magnolia Press • 31 32 • Phytotaxa 232 (1) © 2015 Magnolia Press

CRESPO ET AL.

The only conflicting point in our results is still the unexpected position of L. sibirica (the type of L. sect. Limniris ), merged with members of L. sect. Laevigatae ( Fig. 1), which apparently is contrary to morphological and karyological data. The sequence of that species used in our analyses is the same as that produced by Wilson (2009, 2011). On the contrary, Tillie et al. (2001) had previously recovered both I. sect. Chrysographes and I. sect. Limniris as monophyletic when using a different sample of L. sibirica , a result which is fully congruent with morphological traits. A similar result was reported by Makarevitch et al. (2003), who found I. sibirica (also from a different sample) as strongly supported sister of I. sanguinea (both from Iris ser. Sibiricae = Limniris sect. Limniris ), and they all successively sister to I. setosa (I. subsect. Ensatae , here treated as Eremiris ), a clade with members of I. subsect. Ensatae (here treted as Eremiris ), and I. sect. Laevigatae, albeit all those relationships were weakly supported. Provided that sampling by all those previous authors was incomplete, we cannot be sure about eventual misidentifications of any of those samples. Therefore, additional full-checked samples of Limniris sibirica (L.) Fuss (1866: 637) should be sequenced to newly evaluate the phylogenetic position of this species at infrageneric level, with eventual nomenclatural implications.

The Eremiris clade

This small group includes 3–4 species from central and eastern Asia, from Afghanistan to Korea ( Fig. 19A), with a probable p = 10 (2n = 40, 50) ( Zhao et al. 2000, Yu et al. 2009). They are easy to distinguish by a combination of distinctive features, such as the slender rizhome densely clothed with leaf remains; the leaves narrow and slender, finely-ribbed; the subcupuliform perigone tube, up to 3 mm long; the narrow ovary, slightly twisted; the chartaceous capsule, long beaked and narrowly fusiform, with six equidistant ribs, borne on a very long pedicel; and the seeds globose to pyriform, shiny, smooth and lacking fleshy appendages ( Figs. 4D, 18A). That combination of characters places them apart from the remaining groups of irises ( Dykes 1912).

Different classifications have been suggested for this group. First, Spach (1846a,b) separated Iris pallasii Fischer ex Treviranus (1821: 2) , I. triflora Balbis (1804: 322) and I. doniana Spach (1846a: 34) in I. subg. Eremiris , remarking most of the above cited characters. Those three species were later treated as synonyms in Iris lactea Pallas (1776: 713) s.l. This group was later treated as I. sect. Imberbis Turcz., or as Xiphion subg. Eremiris (Spach) Alefeld (1863: 296) or included in a widely delineated I. sect. Apogon (Baker 1876c) or I. subg. Apogon ( Baker 1892) , sometimes as a member of the informal “Group Tenuifolia” (Baker 1876c), as the also informal “Ensata group” ( Dykes 1912), as I. subsect. Ensatae ( Diels 1930) or as ser. Ensatae ( Lawrence 1953). This latter

A NEW GENERIC ARRANGEMENT OF IRIS SENSU LATISSIMO

Phytotaxa 232 (1) © 2015 Magnolia Press • 33 treatment has widely been adopted in the last decades, though referred to I. subg. Limniris sect. Limniris ( Rodionenko 1961, Mathew 1989 a, SGBIS 1997). More recently, Rodionenko (2006a) has segregated it at the genus rank to include two species, Eremiris lactea (Pall.) Rodionenko (2006a: 1708) and E. oxypetala ( Bunge 1833: 63) Rodionenko (2006a: 1708) . However, according to Doronkin (2012) E. biglumis ( Vahl 1805: 149) Doronkin (2012: 458) and E. pallasii (Fisch. ex Trevir.) Doronkin (2012: 459) merit recognition as autonomous species.

First molecular analyses ( Makarevitch et al. 2003; see also Wilson 2011 and Mizuno et al. 2012) placed E. lactea close to Iris songarica , and both as sister to Joniris . These relationships are now fully resolved ( Fig. 1) and rather congruent with gross morphology. Likely, morphological similarities had previously led Klatt (1872a) to merge all of them in Joniris , a genus he circumscribed in a wide sense, resulting in a very heterogeneous aggregate. Similarly, Baker (1876c) placed I. lactea together with I. tenuifolia Pallas (1776: 714) , and remarked the close morphological affinities of both species. However, Dykes (1912) suggested connections of I. lactea to the North American I. ser. Longipetalae , which in the light of the molecular data are to be interpreted as a result of convergence, without phylogenetic significance.

Nonetheless, Rodionenko (2006 a, 2013) demonstrated that despite morphological affinities, overall divergence is strong enough to accept Eremiris at the genus rank. Contrarily to members of Joniris and Iris songarica , and in addition to characters listed above, nectar drops are placed at the base of hafts of the outer tepals. This is a unique feature connecting Eremiris to I. subg. Xyridion (treated here as Chamaeiris ), a group in which nectar drops occur on the outer apical surface of the perigone tube. However, Chamaeiris is not phylogenetically close to Eremiris , and therefore morphological affinities are due to convergence. We adopt here Rodionenko’s proposal and accept Eremiris as a separate genus.

The Sclerosiphon lineage

Iris songarica was segregated by Nevski (1937) as the unique member of the new genus Sclerosiphon . It is a remarkable taxon showing a large distribution through central and eastern Asia, which is characterised by its clump-forming growth, resembling tussock grass, with slender wiry rizhomes covered with leaf remains spirally disposed; leaves are very narrow (up to 3 mm) and long, some of which sheath the stem base; flowering stems are robust and usually branched, with showy heads of large flowers; crests are narrow and long acuminate ( Figs. 4E, 18B); and capsules are long and beaked, though not strongly ribbed ( Dykes 1912, Mathew 1989a). It grows in mountainous areas of central and western Asia, from Afghanistan to Manchuria ( Fig. 19B). The proposed basic chromosome number is x = p = 9 ( Chakhgari et al. 2013).

Based on morphological features, Dykes (1912) placed Iris songarica within ‘The Spuria Group’, and regarded it as a “connecting link between the spuria and the tenuifolia groups”. Certainly, flowers resemble at first sight those of Chamaeiris , with a similar colour pattern, although different in flower structure. Similarly, Klatt (1866) had previously placed it in a widely defined I. subg. Euxiphion Alef. ex Klatt (1866: 612), namely among species of Chamaeiris , and later he included it in an expanded Joniris as J. songarica (Schrenk) Klatt (1872a: 502) , a genus which resulted very much heterogeneous. Furthermore, Baker (1876c) stressed later the morphological closeness of Sclerosiphon and Eremiris lactea , both groups sharing a similar clump-forming growth and constituting a believable clade in our analyses ( Fig. 1). However, strong morphological divergences exist that suggest recognition of Sclerosiphon and Eremiris as independent taxa. The former produces rhizomes slender, nodose, nearly vertical, capsules are coriaceous, reticulate-nerved (not strongly ribbed), and seeds are subcubic to pyriform, irregularly wrinkled or sometimes shortly winged on angles. Conversely, the latter shows rhizomes thick and compact, capsules are papery, strongly ribbed, and seeds are globose and shiny. In fact, both groups were commonly arranged in different series of I. subg. Limniris , respectively I. ser. Tenuifoliae (Diels) Lawrence (1953: 360) and I. ser. Ensatae (Diels) Lawrence (1953: 362) (cf. Mathew 1989a).

In our analyses, Limniris s.str. is sister to a clade including Sclerosiphon and Eremiris plus Joniris (see below) ( Fig. 1), which confirms previous treatments based on morphology ( Mathew 1989a). Therefore, Rodionenko’s (2006b) recircumscription of Sclerosiphon to include I. ventricosa Pallas (1776: 712) and I. bungei Maximowicz (1880: 509) is not accepted here. These two latter taxa are indeed related to I. tenuifolia , which is treated here as member of the genus Cryptobasis Nevski (see below). Both generic aggregates belong to divergent molecular and

34 • Phytotaxa 232 (1) © 2015 Magnolia Press

CRESPO ET AL.

morphological lineages ( Fig. 1), they also markedly differing in their habit, perigone tube length, crest morphology, and capsule and seed characters, among other features.

The Joniris clade (Ruthenicae irises)

Taxa in this group are easy to recognise because their small capsules, lacking ribs and opening suddenly to shed all seeds almost at once. The three valves curve strongly backwards on ripening, and the capsule remains completely opened. Seeds are pyriform and show a fleshy appendage that shrivels fast after capsule dehiscence ( Dykes 1912, Mathew 1989a). As circumscribed here, it includes 4 species occurring from eastern Europe (Transylvania) to eastern Asia (the Korean Peninsula and eastern China) ( Fig. 20A), for which chromosomes numbers 2n = 32, 42, 48, 84 ( Doronkin & Krasnikov 1984, Probatova 2006, Shatokhina 2006) have been reported, this suggesting a probable p = 8.

Affinites of Iris ruthenica Ker Gawler (1808 : t. 1123), and even its true circumscription, have been a matter of discussion. Spach (1846a,b) included that species in his monotypic I. subg. Joniris , and also stressed differences between I. ruthenica and I. humilis Marschall von Bieberstein (1808: 33) ( Chamaeiris pontica (Zapał. 1906: 191) M.B. Crespo 2011: 66 ), a plant with which had been erroneously synonymised. Alefeld (1863) and Klatt (1867) placed I. ruthenica in Neubeckia , a new genus in which the former author merged a heterogeneous, morphologically diverse aggregate of irises, only sharing the very long perigone tube. They however synonymised it to N. humilis (M.Bieb.) Alefeld (1863: 297) , a plant clearly differing in floral and reproductive characters ( Janka 1868, Dykes 1912). A similar treatment was later published by Klatt (1872a), who raised Joniris to genus rank, though still establishing a heterogeneous generic aggregate that mixed morphologically very divergent taxa. Baker (1876d) placed I. ruthenica in a separate group of his I. subg. Apogon , together with some members of the American Limniris , a treatment similar to that adopted later by Bentham & Hooker (1883) and Pax (1888). Dykes (1912) placed Iris ruthenica in its own group among the “ Apogon irises”, close to Siphonostylis ( I. unguicularis ) and Chamaeiris ( I. foetidissima Linnaeus 1753: 39 ). This solution has been widely followed to date, accepting the ‘Ruthenicae group’ as I. subsect. Ruthenicae ( Diels 1930), I. ser. Ruthenicae ( Lawrence 1953, Mathew 1989a) or I. subg. Limniris sect. Ioniris (Spach) Rodionenko (1961: 190) .

A NEW GENERIC ARRANGEMENT OF IRIS SENSU LATISSIMO

Phytotaxa 232 (1) © 2015 Magnolia Press • 35

Recent analyses ( Wilson 2009, Mavrodiev et al. 2014) fully support placement of I. ruthenica as sister to I. uniflora Pall. ex Link (1820: 71) , two taxa morphologically close ( Rodionenko 1961, Dykes 1912) which sometimes have been treated as synonyms ( Dykes 1912). They belong to the ‘Ruthenica group’ ( sensu Dykes 1912 ) and are strongly supported as sister to the clade formed by Eremiris plus Sclerosiphon ( Wilson 2009, Mavrodiev et al. 2014). These relationships can also be explained on the basis of floral morphological traits such as the standards lanceolate and tapering into a canalicutate claw, and the crests almost straight, usually with crenate margins ( Figs. 4F, 18C–D). However, strong differences exist (see Rodionenko 2006a,b) among all three groups (e.g. rootstock morphology, stigmata outline, capsule traits, seed characteristics, etc.), which warrant recognition at generic rank. However, Joniris is here treated in a more restrictive way as initially defined by Klatt (1872a), and therefore an amended circumscription is here established.