Phaeiris (Spach, 1846) M. B. Crespo, Mart.

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 Phaeiris clade (Louisiana irises or Hexagonae irises)

Spach (1846a,b) first recognised this group by establishing I. subg. Phaeiris Spach (1846a: 46) to include I. fulva Kew Gawler (1812 : t. 1496). He stressed affinities of the vegetative structures with those of Limniris , though accurately emphasised differences in flower, fruit and seed features. This group is commonly accepted to include ca. 9 species, which are readily recognised by the compact, nodose-annulate, reddish rhizome with numerous long roots ( Fig. 2H ); falls with a pubescent central ridge; standards usually reflexed or patent, sometimes erect ( Fig. 4G ); stigma bilobed, with obtuse to triangular-acute lobes; capsules hexagonal in cross section, with 6 angles equally spaced, prominent and sometimes shortly winged ( Fig. 18E ); and seeds large, with testa thickened, corky and almost smooth ( Fig. 6K ) ( Dykes 1912). They are waterside irises, tall and with many-branched and foliose inflorescences, usually growing in swamps and marshes, a peculiar semi-aquatic environment in which their corky seeds seem to be an advantage for floatation dispersal ( Mathew 1989a). Most taxa occur in southeastern North America (from Ohio and Virginia to Louisiana and probably Texas) ( Fig. 19C ), this being the origin of their horticultural name ‘Lousiana irises’. Counts reveal main chromosome numbers 2n = 42, 43 and 44, and occassional polyploids to 2n = 84 ( Henderson 1994, 2002), which suggests a probable p = 7.

Species of this clade have usually been included among the ‘ Apogon irises’ (I. subg. Limniris ), usually treated as a subsection ( Diels 1930) or a series ( Lawrence 1953, Mathew 1989a) in sect. Limniris , to which the epithet ‘Hexagonae’ has been applied on account of the capsule features. Most authors have placed these irises in I. sect. Limniris close to other North American groups, such as I. series Longipetalae , Californicae and Laevigatae ( Dykes 1912, Diels 1930, Lawrence 1953, Rodionenko 1961, Mathew 1989a). However, molecular analyses by Wilson (2009, 2011) and Mavrodiev (2010) have shown that the Louisiana irises are not so closely related to those other groups, and the Limniris section is not monophyletic as usually accepted ( Rodionenko 2007). The Louisiana irises form an isolated well supported clade that is sister to the clade formed by the true Limniris ( Limniris s.str.) plus the group of Eremiris , Sclerosiphon and Joniris . Our recent analyses ( Fig. 1 ) are strongly congruent with that arrangement, and show Phaeiris as reliable sister to the above mentioned clades.

According to the discussed morphological data and the isolated position of this clade with regard to other groups of the former ‘ Apogon irises’, this remarkable aggregate is recognised here at the generic rank, for which the name Phaeiris is recovered.

The Louisiana irises show a high variation in floral and vegetative characters. They have been the matter of model evolutionary studies (Arnold 2000, Arnold et al. 2008, 2012, Martin et al. 2006, 2008, among others), which tried to demonstrate that morphological variation is usually derivated from reticulate evolution among taxa of this group (Arnold 2000, Arnold et al. 2008, 2010) frequently occurring in natural environments highly influenced by human disturbance, which modify gene flow among populations (e.g. Arnold 2000, Hamlin & Arnold 2014; see however Meerow et al. 2011). Taxonomic recognition of this variation in the past lead to description of a high number of species ( Small 1933, Small & Alexander 1931), many of which sometimes show weak biological value and perhaps have been treated in lower infraspecific ranks ( Foster 1937) or even as hybrids (summarised in Henderson 1994). However, experimental work demonstrated that first generation hybrids of the Louisiana irises perform as well as or even better than their parent taxa ( Burke et al. 1998), and they can establish either by sexual reproduction or vegetative multiplication. Hybridisation hence can generate phenotypic variation, which may be adaptive in rapidly changing environments ( Brothers et al. 2013). Therefore, some of those hybridogenous taxa are

36 • Phytotaxa 232 (1) © 2015 Magnolia Press

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

Phytotaxa 232 (1) © 2015 Magnolia Press • 37 in need of re-evalutation for eventual treatment as species when appropriate. For example, according to the results and summary of Arnold (2000: 1687; see also Henderson 1994, among others), the vast majority of Hexagonae species described by Small and Alexander are just the product of hybridisation processes between three taxa: Iris brevicaulis Rafinesque (1817: 20) , I. fulva , and I. hexagona Walter (1788: 66) . Even if so, as suggested for Iris nelsonii Randolph (1966: 150) (but almost no nucleotide sequence data deposited in GenBank for this presumably triple hybrid), we believe that at least nine well-established species of Louisiana irises are occurring in the United States which merit recognition, some of them being regarded as “stabilised hybrids” and most of them seriously threatened with extinction ( Trahan 2007).

It is worth mentioning that the recent molecular treatment of the Louisiana irises in Florida ( Meerow et al. 2011) basically matches Henderson’s (2002) assumption of the only presence of Iris savannarum Small (1925: 57) and I. giganticaerulea Small (1929: 5) . However, neither of Small’s taxa such as I. alabamensis Small ( Small & Alexander 1931: 354) , I. albispiritus Small (1929: 3) , I. elephantina Small & Alexander (1931: 352) , I. kimballiae Small (1925: 59) , or I. rivularis Small (1927: 11) (most of them are listed in Henderson 1994, 2002 as putative synonyms), among others, were mentioned by Meerow et al. (2011), despite of the actual ( Small & Alexander 1931) or potential occurrence of those taxa in the Floridian Peninsula and the numerous “fairly isolated and genetically homogeneous” populations of Iris series Hexagonae accurately described by those authors ( Meerow et al. 2011: 1036). In fact, Meerow et al. (2011: 1036) showed that most of the studied Floridian populations are genetically dissimilar, and at the same time these populations appear to be highly differentiated with low levels of migration. Meerow et al. (2011) correctly suggested that admixture is not common among Floridian iris populations in the field except when they are geographically proximal ( Meerow et al. 2011). From the general standpoint, the results of Meerow et al. (2011) may be easily seen as a nice modern reappraisal to the classical taxonomical treatment of Small & Alexander (1931), but further investigation is necessary to establish more accurate congruences.

The Zhaoanthus clade ( Chinenses irises)

According to Rodionenko (2007: 550), the Chinenses irises are the most interesting group within Limniris , which still seems to be poorly understood. This clade includes a group of small plants sharing slender rhizomes, usually wiry and stoloniferous; leaves prominently ribbed; stems short, with reduced leaves; falls widely spatulate, slightly reflexed towards apex, with a low wavy or almost entire crest which belongs to type I of Guo (2015); standards patent to erect-patent, spreading, broadly oblanceolate to suborbicular; and seeds globose, with fleshy raphe or aril that withers early as a wing ( Dykes 1912, Mathew 1989a) ( Figs. 4H–I , 6P , 21A ). Sometimes they produce rootnodules ( Fig. 2J ), resembling those of clover plants ( Grey-Wilson 1997b). It is currently accepted to include at least 7 species, occurring in China, Korea and Japan ( Fig. 19D ), with chromosome numbers 2n = 22, 24, 32, 44 ( Chimphamba 1973b, Zhao et al. 2000) and presumably 2n =28 ( I. henryi , according to Wilson 2009: 283), which suggest p = 7 or 8.

Dykes (1912) first referred members of this clade as ‘The Chinese Group’, a name that was validated as I. subsect. Chinenses by Diels (1930) and later treated at serial rank by most authors ( Lawrence 1953, Rodionenko 1961, Mathew 1989a). Relationships of this clade have been obscure, since many morphological traits are shared with other representatives of the ‘ Apogon irises’, such as Lophiris (the slender, stoloniferous rizhomes and the dwarf habit) or Eremiris (the rootstock clothed with much reduced leaves in some species), or Joniris (seeds features) ( Dykes 1912, Mathew 1989a). However, the combination of morphological traits is unique among all irises clades.

Recent analyses ( Fig. 1 ; see also Wilson 2009, 2011) revealed that species of I. ser. Chinenses formed a highly supported group, with the exception of I. henryi Baker (1892: 6) which nested in Limniris . Later, Guo & Wilson (2013) recovered I. henryi as a close relative of I. proantha Diels (1924: 427) using a different sequence data source, Guo JW11-35 ( RSA), this being congruent with the traditional arrangement of the cited section ( Grey-Wilson 1997b) based on morphological and geographical data. Similarly, Mizuno et al. (2012) and Lee & Park (2013) have reported placement of I. rossii as sister to other Asian members of I. ser. Chinenses , as previously suggested on morphological grounds ( Baker 1877b, Dykes 1912, Mathew 1989a, Zhao et al. 2000).

Regarding the position of I. speculatrix Hance (1875b: 196) , the available molecular data are still scarce and do

38 • Phytotaxa 232 (1) © 2015 Magnolia Press

CRESPO ET AL.

not resolve confidently its phylogenetic position. Relationships of this species have always been revealed as problematic ( Rodionenko 2007) and its monophyly is questionable. For example, Tillie et al. (2001) recovered it as strongly sister to a wide aggregate of beardless irises (the Xiphion clade sensu Wilson 2011 ), albeit it was highly sequence-divergent with regard to the rest of representatives in there. That position was said to be divergent by Hall (2013), but consistent with morphological data. Similarly, in a restricted analysis of crested irises, Guo & Wilson (2013) have recovered this Chineses species as sister to I. winogradowii Fomin in Woronow & Schelkownikow (1914: 8), a member of Iridodictyum (I. subg. Hermodactyloides ). Conversely, classical arrangements by Rodionenko (1961) and Mathew (1989a) placed it among the Evansia irises, respectively in I. sect. Crossiris and I. sect. Lophiris . However, Mathew (1989a) questioned this position based on the reduced entire crest of I. speculatrix , which seemed only a raised ridge. Important morphological characters join the latter species to other members of the Chinenses irises aggregate and put it apart from the above cited Xiphion clade, namely the slender usually wiry and stoloniferous rhizomes, the shiny leaves, the spreading falls, the central ridge on the standards which belongs to type I of Guo (2015), the oblong to triangular entire stigma, or the globose seeds with fleshy raphe or aril that withers early as a wing ( Fig. 6J, 6P ). Similarly, anatomical studies of leaves by Wu & Cutler (1985) also suggested inclusion of I. speculatrix among the Chinenses irises. This is congruent with Dyke’s (1912) inclusion of I. grijsii Maximovicz (1880: 515) , a synonym of I. speculatrix (cf. Zhao 1980), in that latter group.

According to morphological, phylogenetic and biogeographic data, I. ser. Chinenses is treated here at the genus rank under the name Zhaoanthus . Furthermore, following the traditional arrangement of modern authors ( Rodionenko 1961, Mathew 1989a) we also place here I. speculatrix on the basis of information discussed above. Nonetheless, new molecular data are still needed to fully confirm either its monophyly or definitive placement in the present genus.

The Iridodictyum clade (Reticulata irises)

Taxa included in this clade are easy to recognise on the basis of the tunicate tuberiform bulbs, constituted by a single fleshy tunic, clothed with reticulate or reticulate-hairy outer tunics ( Fig. 2A ) ( Rodionenko 1961); leaves bifacial, canaliculate, 4–6–8-angulose in section, surrounded at the base by several membranose, uncoloured sheaths; stem short, lacking well-developed green leaves; style branches divided up to their base; stigma with two deep rounded lobes ( Figs. 4J–K , 21B ); pollen grains ellipsoid with a single longitudinal furrow and finely reticulate surface; and seeds with wrinkled testa, bearing a large pale caruncle. In a narrow sense fitting Iridodictyum sect. Iridodictyum , this genus as here understood includes ca. 17 species, occurring from Turkey and the Transcaucasus to the Middle East ( Fig. 22B ). The values of x are 8 and 10 ( Simonet 1934).

At first sight, Iridodictyum is similar to Xiphion , Cryptobasis and Syrianthus due to the morphology of flower ( Figs. 4‒5 ) and seedling ( Rodionenko 1961, Tillich 2003). It also resembles Hermodactylus in leaf and seed features, and also in the presence of several sheaths clothing the stem base. Almost those same taxa were recovered as the closest relatives of Iridodictyum in recent molecular topologies ( Fig. 1 ; see also Tillie et al. 2001, Wilson 2011). Furthermore, they all share the bilobed to bifid stigma. Hall (2013) also observed in I. pamphylicum Hedge (1961: 557) , a remarkable Turkish member of Iridodictyum , that the base-plate of the bulb elongates to form nearvertical branches with terminal bulbs, similar to those in Syrianthus , a fact which would indicate a close connection. This species shows some morphological peculiarities, such as the thickened fleshy annual roots, sharp basal leaf fibres and relatively long stem, which led Mathew (1989b: 84) to describe Iris sect. Brevituba B.Mathew to accommodate it (see below). However, synapomorphies of Iridodictyum such as the bulb anatomy and the pollen morphology ( Rodionenko 1961, 1999) warrant recognition of this genus.

As originally delineated, Iridodictyum is non-monophyletic due to the placement of Iris kolpakowskiana Regel (1877: 263) (I. sect. Monolepis) ( Fig. 1 ). Taxa in the Iridodictyum clade ( Fig. 1 ) have usually been treated as a compact group in its own. Spach (1846a,b) segregated I. reticulata Marschall von Bieberstein (1808: 34) as the only member of his I. subg. Hermodactyloides , a treatment which was adopted by Mathew (1989a,b) and Goldblatt & Manning (2008), though the latter authors also included here taxa of Alatavia . Klatt (1866) merged the ‘Reticulata irises’ in an expanded genus Xiphion , a treatment followed by Baker (1871, 1877) who also created a new subgenus, Xiphion subg. Micropogon Baker (1877: 124) , to accommodate Iris danfordiae Boissier (1882: 124) ( Fig. 4K ). Later, Baker (1892) reduced Xiphion and all its members to I. subg. Xiphion (which also included Iridodictyum and Alatavia as here accepted). More recently, Dykes (1912) and Diels (1930) have recognised I. sect.

A NEW GENERIC ARRANGEMENT OF IRIS SENSU LATISSIMO

Phytotaxa 232 (1) © 2015 Magnolia Press • 39 Reticulata Dykes (1912: 220) , in which they included the Middle East taxa here referred to Iridodictyum and Alatavia , a treatment basically followed by Lawrence (1953) who included I. sect. Reticulata in I. subg. Xiphion . Finally, Rodionenko (1961) raised that section to genus rank as Iridodictyum , based on strong morphological and biogeographical differences as shown above. He however recognised two sections (I. sect. Iridodictyum and I. sect. Monolepis) on the basis of sound characters, including vegetative (e.g. leaf and sheath features) and reproductive (e.g. style and pollen) characteristics, a proposal accepted by Kamelin (1973). The latter section (I. sect. Monolepis) was later segregated ( Rodionenko 1999) as the new genus Alatavia (see below).

The wide morphological variation found in Iridodictyum s.l. together with the molecular evidence ( Fig. 1 ) points out to recognition of smaller generic groups, among which Iridodictyum is here accepted in a narrower circumscription that excludes Iridodictyum sect. Monolepis . Furthermore, the position of Iris pamphylica as sister group of the Iridodictyum clade ( Fig. 1 ), which can be explained on the basis of morphological differences, points out to recognition of the monotypic Iris sect. Brevituba ( Mathew 1989b) , though transferred to the genus Iridodictyum .

The Cryptobasis clade (Tenuifoliae irises)

This is an Asian group of xerophytic species, morphologically fairly heterogeneous but easy to recognise. They show remarkable traits in their life history and morphology of the vegetative organs ( Mavrodiev & Alexeev 2003), typically described as “small slender subvertical rhizomes” that produce a “tufted habit” (e.g., Dykes 1912, Mathew 1989a) ( Figs. 21C–D ). The latter trait results in the relatively long life of the monocarpic shoots ( Mavrodiev & Alexeev 2003), which is a feature uncommon within Iris s.l. ( Mavrodiev & Alexeev 2003). Leaves are narrow, but sometimes round in a cross-section ( Mavrodiev & Alexeev 2003), frequently tough, prominently ribbed, and their sheaths remain as long reddish-brown fibers forming an apical neck-like structure ( Fig. 2L ); the spathe valves are usually inflated, keeled, sometimes transversely nerved ( Figs. 2L , 4L ), due to the full or partial reduction of the peduncle (the main stalk of the inflorescence) ( Rodionenko, 2006b, 2013, Mavrodiev & Alexeev 2003); the perigone tube is fairly long, up to 15 cm in Cryptobasis mariae Mavrodiev (2002: 78 ; Iris juncifolia Pallas 1799: 94 nom. nud., LINN-HS92-30, “Tauria”, Pallas, s.n.) and C. loczyi ( Kanitz 1891: 58) Ikonnikov

40 • Phytotaxa 232 (1) © 2015 Magnolia Press

CRESPO ET AL.

(1972: 302) ( Figs. 21C–D ), perhaps the maximum value among Iridaceae ; the stigma is bilobed, normally with rounded lobes; the capsules show usually six prominent ribs and are shortly beaked; and seeds are angulose to subcubic, without aril, with testa irregularly wrinkled on the sides and almost smooth on the back ( Dykes 1912, Mathew 1989a). It includes at least 10 species, occurring in Crimea, SE European Russia, Kazakhstan, and central and southeastern Asia, the highest diversity being found from Afghanistan to Mongolia and eastern China ( Fig. 20B ). The known chromosome numbers (2n = 14, 20 and 28; Zhao et al. 2000, Probatova 2006) are still incomplete, and would suggest p = 7.

Members of this group have usually been included in I. sect. Apogon or I. subg. Apogon . Baker (1876c) first reported the informal “Group of I. tenuifolia ”, in which he included I. tenuifolia and I. ventricosa together with I. songarica (here treated as the only member of Sclerosiphon ) and I. macrosiphon Torrey (1857: 144) (here included in Limniris sect. Californicae ). This solution was partially adopted by Dykes (1912), who reduced that group to the two former species and I. bungei Maxim. Diels (1930) validated it as I. sect. Apogon subsect. Tenuifoliae and accepted Dyke’s circumscription. That proposal was assumed by Lawrence (1954) and Mathew (1989a) who treated it as I. ser. Tenuifoliae, albeit they reincluded I. songarica . Rodionenko (1961) revived I. subsect. Tenuifoliae and reorganised it into two series, describing I. ser. Ventricosae Rodionenko (1961: 189) to include I. ventricosa , I. bungei and I. songarica . More recently, Rodionenko (2006b) has transferred taxa of the latter series to the genus Sclerosiphon .

Molecular studies by Makarevitch et al. (2003) recovered I. tenuifolia , I. ventricosa and I. loczyi Kanitz (1891: 58) to form a compact clade, which was sister to a clade including taxa of I. subg. Xyridion (here treated as genus Chamaeiris ). This is due to the fact that their sampling was quite reduced and did not include members of I. subg. Xiphion (sensu Wilson 2011) , but the obtained relationships are congruent with morphological traits of both groups. Wilson (2004) recovered I. loczyi and I. missouriensis Nuttall (1834: 58) as strongly supported sisters (see also Mavrodiev 2010), though the position of this clade was not resolved with regard to other beardless irises (namely, the Limniris core, the Xiphion clade and I. sect. Lophiris ).

After complete re-sampling of the available cpDNA data for the presumably non-monophyletic I. loczyi (reviewed in Mavrodiev et al. 2014: 2, Appendix S1), we obtained better resolved trees in which taxa of Cryptobasis nested far apart from the monotypic Sclerosiphon as first circumscribed ( Fig. 1 ), as well as from Longipetalae irises.

The phylogenetic position of the Cryptobasis clade, however, is controversial. Conventional phylogenetic analyses place Cryptobasis as sister to Hermodactylus ( Fig. 1 ). Genera Iridodictyum , Syrianthus , Xiphion and Alatavia are recovered as next relatives, and the resulting clade is finally sister to Chamaeiris ( Fig. 1 ). These relationships are well supported by a similar general flower morphology and seedling features ( Rodionenko 1961, Tillich 2003), though many other morphological characters allow differentiation. Furthermore, members of Cryptobasis possess particular rare compounds ( Wang 2010, Kaššak 2012) which set them apart from the rest of members of Limniris . In contrast, 3TA placed Cryptobasis as a next sister to the rest of the Iris clades after Siphonostylis ( Unguiculares irises) ( Fig. 1 ).

On the basis of the noticeable morphological divergence of Cryptobasis , we adopt here the genus rank, though in an expanded circumscription to include other eastern Asian species. Our analyses, however, only included I. ventricosa , I. tenuifolia and I. loczyi ( Fig. 1 ), and therefore relationships to other related members ( I. bungei , I. qinghainica Zhao 1980: 55 , and relatives) still remain unclear and are the focus of future research.

The Hermodactylus lineage (Widow irises or Snake’s head irises)

Since described by Miller (1754), Hermodactylus has widely been accepted to form a separate genus including H. tuberosus ( Linnaeus 1753: 40) Miller (1768 : without pagination) (cf. Parlatore 1854, 1860, Alefeld 1863, Bentham & Hooker 1883, Baker 1877, 1892, Dykes 1912, Diels 1930, Rodionenko 1961, 2013, Mathew 1989a, Goldblatt et al. 1998, Crespo et al. 2013, among others). Some sound morphological characters are usually referred to for supporting generic segregation ( Mathew 1989a), such as the bifacial, quadrangular leaves, the short rhizome, ending in 2–4 oblong tubercles, ± digitate; the short, inconspicuous standards, tapering into a long point; the unilocular ovary, with parietal placentation, and the papery, unilocular capsule, producing globose seeds with a gelatinous aril ( Figs. 5A , 6H , 21E ). The single species included in this group is probably native to the eastern

A NEW GENERIC ARRANGEMENT OF IRIS SENSU LATISSIMO

Phytotaxa 232 (1) © 2015 Magnolia Press • 41 Mediterranean basin, though it is widely cultivated and naturalised in southern Europe and the British Isles (as well as North America), with some scattered localities in Morocco ( Fig. 19E ). It has usually been divided into two varieties (Baker 1877), which show some morphological and anatomical differences ( Özdemir et al. 2014).The basic chromosome number is x = 10 ( Marcucci et al. 2005, Crespo et al. 2013).

As said above, many authors accepted segregation of this genus based on a unique morphological combination of characters, of which the peculiar tuberous rootstock and the unilocular ovary and fruit were always remarked. However, Spach (1846a,b) reduced it to a subgenus of Iris , and Lynch (1904) to a section.

Recent analyses ( Fig. 1 ) place Hermodactylus close to the Reticulata irises ( Iridodictyum ≡ I. subg. Hermodactyloides ) (see also Tillie et al. 2001 and Wilson 2011), a relationship that is fully supported by the similar features of leaves and sheaths, and the arillate seeds. Its closest relative appears to be Iris pamphylica Hedge , with which it shares the solitary capsule at the stem apex as well as similar rootstock and pollen features ( Hall 2013). Due to these reasons, Goldblatt & Manning (2008) included Hermodactylus in I. subg. Hermodactyloides , though they remarked its exceptionality in the group. Furthermore, flower characteristics (e.g. the panduriform to obovate erect falls reflexed towards the upper part, and the bifid stigma) also connect Hermodactylus with members of Xiphion and Syrianthus , they all forming a well-supported clade. Consequently, Wilson (2011) largely expanded I. subg. Xiphion to accommodate all those cited genera, the resulting aggregate being morphologically heterogeneous and difficult to characterise on morphological grounds.

Despite those evident similarities which point out to a long shared evolutionary history, many other features such as the unsual digitate tubercles attached to a short rhizome, the remarkable green and black flowers with inconspicuous and long acuminate standards, usually narrowed into a canaliculate haft, and the unique unilocular ovary and capsule, allow unequivocal recognition (Crespo et al. 2013). Therefore, Hermodactylus is accepted here at genus rank as traditionally treated in the last two centuries.

The Syrianthus clade ( Syriacae irises)

This distinctive group stands alone from the rest of irises by showing a tough nearly vertical rhizome, with compact bulbiform somewhat swollen leaf bases, and bearing many needle-like bristles at the apex ( Dykes 1912). These sharp and fine bristles probably protect the plants from herbivory, mostly from rodents ( Cohen 1997), and it is also well known among horticulturists that they can produce painful injury to fingers when rootstocks are handled uncarefully ( Mathew 1989a). Leaves are isobilateral, stiff and coriaceous; flowers are solitary, with crests long, lanceolate-triangular, nearly straight; stigma bilobed, with rounded denticulate lobes; capsules are unbeaked, with pericarp subcoriaceous, and they produce seeds subglobose, necked, without aril, showing a hard testa with tuberculate surface ( Figs. 5B , 23A ). This group includes five species (sometimes reduced to 1–3 by some authors), occurring in the Middle East, from eastern Turkey and northern Iran (Kurdistan) to Israel ( Fig. 20C ), where they grow in seasonal semiarid areas which receive rains in winter and the early spring but remain extremely dry the rest of the year. The reported basic chromosome number is x = 12 ( Simonet 1934).

When describing I. grant-duffii, Baker (1892: 7) placed it in I. subg. Apogon among members of the current I. ser. Californicae (here included in the genus Limniris ). Dykes (1912) separated that species, together with I. aschersonii Foster (1902: 288) , I. masia Dykes and I. melanosticta Bornmüller (1907: 495) (all three latter species suggested to be synonyms of I. grant-duffii ) in “The Syrian Group”, an informal name that later Diels (1930) validated as I. subsect. Syriacae . It was raised to sectional rank as I. sect. Dykesiana Simonet (1934: 261), based on its chromosomal peculiarities. This group, with a similar circumscription, was treated as I. ser. Syriacae by Lawrence (1953) and Mathew (1989a), albeit accepted again at subsectional rank by Rodionenko (1961), who later mentioned that the Syriacae irises should however be treated as a section of Cryptobasis ( Rodionenko 2007: 550) but suggesting no formal nomenclatural changes.

Morphological affinities between the Syriacae irises and the Reticulata irises have largely been reported. Dykes (1912) noted that the bulbiform leaf-bases in the rootstock of I. grant-duffii are covered with a reticulated coat, and by the end of the first year they produce seedlings very much alike those of Iridodictyum reticulatum (M.Bieb) Rodionenko (1961: 202) . Similarly, Mathew (1989a) also suggested that seedlings of the former species give rise initially to a small bulb almost indistinguishable from those of the Reticulata irises, which later develops into a short rizhome with thickened fleshy roots (see also Rodionenko 2008).

42 • Phytotaxa 232 (1) © 2015 Magnolia Press

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

Phytotaxa 232 (1) © 2015 Magnolia Press • 43

Molecular phylogeny of Tillie et al. (2001) (see also Wilson 2004, 2009, 2011 and Mavrodiev 2010) recovered members of I. ser. Syriacae as monophyletic and closely related to a clade grouping the Reticulata irises, Hermodactylus and Cryptobasis , which is congruent with the close morphological similarities. Floral structure is also very similar in those aggregates ( Fig. 5 ), this also supporting the molecular relationships. Furthermore, some species (e.g. I. pamphylica , here referred to Iridodictyum ) show traces of bristles which are characteristic in the Syriacae irises ( Hall 2013), whereas some members of the latter group (e.g. I. grant-duffii and I. masia ) show reticulate outer tunics like those in Iridodictyum (cf. Mathew 1989a).

In our recentmost analyses ( Fig. 1 ) similar relationships are recovered, albeit the Syriacae irises are sister to ( Iridodictyum + ( Cryptobasis + Hermodactylus )) clade (conventional phylogenies only) or to the clade ( Iridodictyum + Hermodactylus ) (3TA topology). Furthermore, Cryptobasis shares a number of morphological characters with the Syriacae irises (e.g. the peculiar neck-like fibrous structure at the apex of the rhizome, the isobilateral and coriaceous leaves prominently ribbed, the similar flower structure, and the bilobed stigma with rounded lobes), which support their close placement in the molecular trees. Hall (2013) also reported similarities in the stamen structure among the Syriacae irises, Cryptobasis , and I. sect. Ophioiris Zhao (1980: 56) (here included in the closely related clade Chamaeiris = I. subg. Xyridion ), which support their phylogenetic closeness in our trees ( Mavrodiev et al. 2014). However, important differences are found (e.g. number and characteristics of spathe valves, perigone tube length, fruit and seeds features, etc.), which warrant recognition at the genus rank for all those aggregates. Furthermore, Simonet (1934) found that karyotype peculiarities of I. grant-duffii supported segregation of that group from the rest of Apogon irises.

According to the data discussed above, on the basis of the much divergent morphological characteristics with regard to the other members of the ‘ Xiphion clade’, the new genus Syrianthus is proposed to name this remarkable group of Middle East irises.

The Xiphion clade (Spanish irises)

Irises in this group are well characterised by the tunicate bulb with several free tunics, clothed with smooth, dark membranous or apically fibrous outer tunics, sometimes accompanied by small bulbils ( Fig. 2B ); stems leafy; leaves bifacial, canaliculate, not markedly angulose; falls panduriform or obovate-lanceolate; stigmata bifid, usually with acute narrow lobes ( Fig. 5C–D , 23B ); capsule unbeaked, with papery pericarp; and seeds globose to pyriform or angulose to narrowly winged, without aril, with the testa surface irregularly verruculose to cristate. Xiphion includes 10 species, occurring in the western and central Mediterranean basin, from Italy to Portugal, the highest diversity being in the Iberian Peninsula and northwestern Africa ( Fig. 22A ). The probable values of p are 8 and 9 (Crespo et al. 2013).

On the basis of some of the above mentioned characters Tournefort (1719) had already split this group of bulbous irises as Xiphion , a name which was validated by Miller (1754). The latter author ( Miller 1768) circumscribed Xiphion to include the bulbose irises, and therefore he also transferred here taxa of Juno . The genus Xiphion was in general use until the beginning of the 20 th century, though under different circumscriptions. Adanson (1763: 58) and Medikus (1790) accepted segregation at generic rank, though the former used the name Chamoletta Adanson (1763: 60) arguing that Xiphion was a Greek name for Gladiolus L. Other authors such as Reichenbach (1841), Parlatore (1854, 1869), Klatt (1866) or Baker (1871, 1877) keep on using the generic rank, though with different circumscriptions that in some cases included all bulbous irises ( Xiphion , Iridodictyum , Alatavia and Juno ), or even more the whole ‘beardless irises’ (including also Limniris and several taxa of Chamaeiris ). At the same time, other authors maintained it in Iris at different ranks. First, Tausch (1823) treated it as I. sect. Xiphion , though also including Gynandriris Parlatore (1854: 49) and Hermodactylus . Spach (1846a) raised it to I. subg. Xiphion , also including Iridodictyum (I. sect. Reticulata ), a circumscription that later adopted Lynch (1904) though in I. sect. Xiphion . Bentham & Hooker (1883) and Pax (1888) included most taxa of Xiphion in I. sect. Diaphane ( Salisbury 1812: 304) Bentham & Hooker (1882: 687) , together with other bulbous irises which were placed in I. sect. Juno and I. sect. Gynandriris (Parl.) Bentham & Hooker (1882: 687) . Baker (1892) adopted the subgeneric rank for all bulbous irises, namely I. subg. Xiphion (incl. the ‘Reticulata irises’), I. subg. Juno and I. subg. Gynandriris (Parl.) Baker (1892: 43) . Similarly, Diels (1930) accepted I. sect. Xiphium in a narrow sense, ranked at the same level than the rest of bulbous irises (I. sect. Reticulata , I. sect. Gynandriris , and I. sect. Juno ), a treatment that also parallels that of Foster (1892) or Dykes (1912), and which is virtually coincident

44 • Phytotaxa 232 (1) © 2015 Magnolia Press

CRESPO ET AL.

with that proposed by Lawrence (1954) and Mathew (1989a), though both latter accepted all those groups at subgeneric rank. On the basis of molecular results, Wilson (2011) proposed a rearrangement of I. subg. Xiphium to include Xiphion , Iridodictyum , Alatavia , Hermodactylus and Syrianthus , which renders the resulting aggregate heterogeneous in morphology (see Figs. 21 , 23 ). On the contrary, Rodionenko (1961, 2013) has segregated Xiphion in the strict sense of the name, a solution that was followed by Schulze (1971b), Kamelin (1973), Soják (1982), Mavrodiev (2010), Mavrodiev et al. (2014), Crespo et al. (2013, 2014) and Peruzzi et al. (2014), accepting an independent and better circumscribed Xiphion restricted to the bulbous western Mediterranean taxa.

Our analyses ( Fig. 1 ) confirmed Xiphion as a reliable clade (see also Wilson 2011), which is sister to a clade including Iridodictyum s.str., Cryptobasis , Hermodactylus and Syrianthus . All those latter clades are presumably sister to Alatavia , as recovered by conventional analyses ( Fig. 1 ), but not fully confirmed by 3TA, in which Cryptobasis was placed as an early branching lineage ( Fig. 1 ) and Alatavia as presumably sister of the ‘Foetidissima irises’ ( Fig. 1 ). However, Alatavia , Cryptobasis , Hermodactylus , Iridodictyum , Syrianthus and Xiphion are mostly bulbous or tuberous irises, showing a similar flower structure usually with panduriform outer tepals, and bifacial (or very narrow isobilateral leaves), which may point to a common ancient relationship. They also show similar pollen types ( Schulze 1971b). Probably, this led Wilson (2011) to expand I. subg. Xiphion to include Hermodactylus , Iridodictyum (plus Alatavia ) and Syrianthus , a group she also referred as the “subg. Limniris IV ” clade. However, each of those clades evolved divergently to produce strong differences in rootstock structure, leaf morphology and anatomy, stigma features, and fruit and seed characteristics. All those differences warrant generic recognition of Xiphium (as well as of other cited genera), though in a restricted sense to include only the western Mediterranean irises with bulbs formed with free tunics ( Rodionenko 1961, 2013, Crespo et al. 2013).

The Alatavia lineage (Monolepis irises)

Rodionenko (1961) placed Iris kolpakowskiana and the rare and almost extinct I. winkleri Regel (1884: 677) (reviewed in Rodionenko 2013), two peculiar representatives of I. sect. Reticulata , in the newly described genus Iridodictyum . However, initially Rodionenko (1961) included both species in Iridodictyum sect. Monolepis . Mostly in external traits, these two species are related to Iridodictyum , though fairly distinct morphological characters allow easy recognition ( Fig. 23C ; see also Rodionenko 1961: 204, Rodionenko 1999). Leaves are numerous, channelled and finely sulcate, all included in a single long tubular sheath; stigmatic lip entire, rounded; pollen grains sphaerical, unfurrowed, with surface cracked; and seed with reticulate testa, lacking aril ( Rodionenko 1999). Both members of this genus occur only in the Tien Shan Mountains, central Asia ( Fig. 22C ), and share a basic chromosome number x = 10, albeit notable karyotype differences exist between both groups ( Rodionenko 1999).

Alatavia and Iridodictyum share a similar flower structure ( Fig. 5E ) and the tunicate tuberiform bulbs, with a single fleshy inner tunic and reticulate outer clothes. The single basal sheath is also present in Xiphion , Syrianthus and Cryptobasis , whereas the rounded entire stigmatic lip is unique in this aggregate. Similarly, the pollen grains show a divergent morphology ( Rodionenko 1961, 1999), they being spheroidal and lacking the typical central furrow found in all other genera of the ‘ Xiphion clade’ ( sensu Wilson 2011 ).

Due to the recent lack of both cpDNA sequence data and useful plant material for A. winkleri , molecular analyses of the Monolepis irises are still incomplete, with only one taxon, A. kolpakowskiana (Regel) Rodionenko (1999: 104) , included in the matrix we analysed. However, both 3TA and conventional phylogenies undoubtedly placed it fairly far form the Iridodictyum clade ( Fig. 1 ). Therefore, due to the strong morphological and geographic divergence among all groups in that aggregate, we accept Alatavia as an autonomous genus. From a molecular phylogenetic standpoint, Alatavia kolpakowskiana itself, however, is to be accepted at the generic rank, even if future analyses would show non-monophyly of Alatavia , as circumscribed by Rodionenko (1999) (e.g., A. winkleri would nest within Iridodictyum , etc.).

The Chamaeiris clade (Spuria irises plus Foetidissima irises)

Members of this group are well characterised by their isobilateral leaves, usually fetid when crushed; the flowers are terminal or axillary and show patent falls, markedly panduriform to obovate, and the stigmas are bifid with

A NEW GENERIC ARRANGEMENT OF IRIS SENSU LATISSIMO

Phytotaxa 232 (1) © 2015 Magnolia Press • 45 pointed lobes ( Fig. 5F–I ); capsules are coriaceous, usually sharply beaked and 6-ribbed, sometimes with ribs paired on the angles ( Fig. 23D ); and seeds are globose to subcubic, without aril, with testa papery and loose, or fleshy and coloured ( Figs. 6F–G ) ( Mathew 1989 a, Bowley 1997, Crespo et al. 2013). As here circumscribed, it includes 28 species, occurring from Europe (excepting the northernmost areas) and northestern Africa to eastern Asia, the highest diversity being from Turkey and the Caucasus to western Himalaya ( Fig. 24A ). The probable p values are 8 and 9, although p = 10 and 11 have also been suggested (Crespo et al. 2013).

Members of this group were soon separated from the typical bearded irises. Dodoens (1616) referred Iris graminea as “ Chamaeiris ” (the dwarf iris), a name later validated at generic rank by Medikus (1790) to include I. graminea , I. foetidissima and some members of the Iris spuria group. This circumscription entirely matches the modern concept of the group, but unfortunately it was neglected by later authors. Tausch (1823) grouped members of the I. spuria aggregate as I. sect. Xyridion , a name treated as subgenus by Spach (1846a,b) and as genus by Fourreau (1869). This latter author also proposed the genus Spathula for I. foetidissima . Fourreau’s proposal was partly followed by Klatt (1872a), who only accepted Xyridion though in an expanded sense which made it heterogeneous, since it also included members of Limniris s.l. A similar heterogeneous treatment was proposed by Parlatore (1854, 1860), albeit he transferred the Spuria irises to Xiphion .

In general terms, representatives of the present clade were included later in I. sect. Apogon (Baker 1876c, Dykes 1912) or I. subg. Apogon ( Baker 1877a, 1892), sometimes arranged into two different aggregates, the “Group of I. graminea and I. sibirica ” and the “Group of I. spuria ” (cf. Baker 1876c). On a similar basis, Dykes (1912) recognised “The Spuria group” (for the aggregates of I. graminea and I. spuria ) and “The scarlet-seeded iris” (for I. foetidissima ), which Diels (1930) treated them respectively as I. sect. Apogon subsect. Spuriae Diels (1930: 502) and subsect. Foetidissimae Diels (1930: 502), and Lawrence (1953) as I. sect. Spathula subsect. Apogon ser. Spuriae and subsect. Foetidissimae. Both groups were regarded as sections in I. subg. Xyridion by Rodionenko (1961), whilst Mathew (1989a) reduced them to series in I. subg. Limniris . The latter proposal has been widely followed by most recent authors.

Molecular work by Tillie et al. (2001) revealed that the Spuria irises nested in the same clade as Xiphion , Alatavia and Cryptobasis ( I. bungei ), far apart from other representatives of I. subg. Limniris which was polyphyletic as usually circumscribed. This finding was congruent with data by Raycheva et al. (2011), using ISSR markers in a survey on Bulgarian irises. Similar results were obtained by Wilson (2004, 2011) after a more extensive sampling of irises groups. She found that the Spuria irises formed a strongly supported clade (100% ML BS), with I. foetidissima as sister (97% ML BS) to a clade including I. ser. Gramineae Rodionenko (1961: 192) and I. ser. Spuriae (Diels) Lawrence (1953: 361) . That group was named the “subg. Limniris II ” clade or I. subg.

46 • Phytotaxa 232 (1) © 2015 Magnolia Press

CRESPO ET AL.

Xyridion . Those internal relationships were entirely congruent with the taxonomical arrangement presented earlier by Rodionenko (2005) under the genus Xyridion . This author claimed the independence of the latter genus, in which he accepted two sections with three series morphologically well characterised, one of which was devoted to X. ludwigii ( Maximowicz 1880: 508) Rodionenko (2005: 59) ( Iris ludwigii Maxim. ). A similar arrangement has been presented by Crespo (2011), who revived the earlier name Chamaeiris , a treatment recently followed by Peruzzi et al. (2014). In fact, Makarevitch et al. (2003) found I. ludwigii as sister to I. halophila Pallas (1776: 733) in RAPDs and DNA sequencing analyses. Pollen features allow easy recognition of Chamaeiris members from other representatives in the genus ( Mitić et al. 2013), and also reflect differences between representatives of the Ch. ser. Chamaeiris (I. ser. Gramineae) and Ch. ser. Spuriae (Diels) Crespo (2011: 66), which support previous taxonomic treatments.

Our trees ( Mavrodiev et al. 2014) have however revealed that the Chinese I. anguifuga Y.T.Zhao & X.J.Xue in Zhao (1980: 56) (2n = 34) is nested among members of the Chamaeiris clade, and is not an atypical member of I. sect. Tenuifoliae as suggested by some authors ( Mathew 1989 a, Hall 2013). Zhao (1980) described I. anguifuga from the Hubei province (central-eastern China), which he placed in a new monotypic group, I. sect. Ophioiris . This species was said to be morphologically close to Sclerosiphon songaricum (Schrenk) Nevski (1937: 331) , from which it differs in having a diverse rootstock, flowers with a single spathe, and fruits very long rostrate. It is worth mentioning that I. anguifuga shows a peculiar bulb-like rootstock during summer dormancy. Afterwards, in the autumn it develops a rhizome that mostly withers away after fruit release, only remaining the apex to form again a new bulb-like resting structure (J. Murrain, pers. comm.).

Preliminary molecular results by Tillie et al. (2001) recovered I. anguifuga as sister to I. bungei ( Cryptobasis ) in a combined plastid tree, though the sampling was not exhaustive and internal branch support in the whole aggregate was low. Rodionenko (2004) raised I. sect. Ophioiris to genus rank, Ophioiris (Y.T.Zhao) Rodionenko (2004: 1359) , and justified segregation on the basis of the above exposed characters, some of which he regarded as primitive in the context of the genus. Interestingly, Rodionenko (2008) hypothesised that I. anguifuga should be regarded as the putative ancestor of the Spuria irises, and in some extent the evolutionary connecting bridge with Sclerosiphon . Our recent molecular trees ( Fig. 1 ; see also Mavrodiev et al. 2014) corroborate Rodionenko’s (2008) general conclusions. However, the clade including the aggregate of I. spuria (series Spuria ) plus I. anguifuga is strongly supported (100% ML BS) as part of the Chamaeiris clade in our trees. Morphological traits (e.g. the overall habit, the morphology of fruit and flower, and the nectarotheca structure) support this placement ( Figs. 5I , 25A ). Therefore the genus rank appears to be inadequate for I. anguifuga and hence inclusion of Ophioiris in Chamaeiris is suggested here.

Iris anguifuga forms an unresolved grade within other members of Ch. ser. Spuriae and appears to be presumably non-monophyletic on the 3TA topology ( Mavrodiev et al. 2014), therefore further work is needed. This lack of resolution within Ch. ser. Spuria may be the result of either recent rapid radiation or simply as a sampling issue. However, we believe it is worth maintaining that fairly remarkable species in its own series, Ch. ser. Ophioiris , on the basis of morphological divergences until future investigations help to clarify this issue. One interesting point is that the peculiar bulbiform rootstock of Ch. anguifuga relates the ‘ Chamaeiris clade’ to members of its sister clade, such as Syrianthus and Hermodactylus , which are also connected to the true bulbous members of the group (i.e. Alatavia , Iridodictyum and Xiphion ).

The Dielsiris clade ( Longipetalae irises or Rocky Mountain irises)

This group of western North American irises has been for a long time a matter of confusion. Members of the clade share some morphological traits ( Fig. 25B ), such as the bilobed stigma, with entire or crenate lobes, and seeds globose to pyriform (sometimes slightly flattened), apiculate, lacking fleshy appendages, and with hard testa surface (not corky), wrinkled ( Dykes 1912, Mathew 1989a). Includes at least 5 species (sometimes reduced to only two), occurring in central and western North America, namely in the Rocky Mountains area (from western Canada to northeastern Mexico ( Fig. 24B )). The avalaible counts indicate chromosome numbers 2n = 38, 86 and 88 ( Simonet 1934, Foster 1937), this suggesting p = 8 and 11.

A NEW GENERIC ARRANGEMENT OF IRIS SENSU LATISSIMO

Phytotaxa 232 (1) © 2015 Magnolia Press • 47 48 • Phytotaxa 232 (1) © 2015 Magnolia Press

CRESPO ET AL.

Members of this clade have usually been included in I. sect. Apogon (Baker 1876c, Dykes 1912), I. subg. Apogon ( Baker 1877a, 1892) or I. subg. Limniris ( Rodionenko 1961, Mathew 1989a). However, as Wilson (2004, 2009) has shown, representatives of the Limniris aggregate were nested in several clades with different origins and phylogenetic relationships. One of them was constituted in part by I. longipetala Herb. in Hooker & Arnott (1840: 395) and I. missouriensis , two related taxa from western North America. These species were frequently treated as “The Longipetala group” ( Dykes 1912), I. subsect. Longipetalae ( Diels 1930) or I. ser. Longipetalae ( Lawrence 1953), though sometimes were assigned to the “Group of I. ruthenica ” (Baker 1976c) . These species sometimes were found to be sister, albeit with a low support, to either I. tenuis Watson (1882: 380) ( Tillie et al. 2001) or I. gracilipes Gray (1858: 412) ( Wilson 2009), in a rather early branching position with regard to the rest of taxa of the “beardless irises”, the Limniris core. However, Wilson (2011) finally recovered a clade she provisionally referred as the “subg. Limniris III ” clade which included the Longipetalae irises plus a clade formed with I. gracilipes and I. tenuis . Similar results were shown by Mavrodiev et al. (2014), who found both species to form a strongly trustworthy clade that was placed either among the basal lineages of the ‘ Apogon irises’ (s.l.) in the conventional topologies together with sister Monospatha irises ( Fig. 1 ), or as sister to the ( Iridodictyum + Hermodactylus + Syrianthus + Xiphion + Chamaeiris + Alatavia kolpakowskiana ) clade in the 3TA topology ( Fig. 1 ). Morphological and biogeographical features however allow establishing some connections of the Longipetalae irises with the Monospatha irises (here regarded as the new genus Rodionenkoa ), better than the Xiphion clade. In fact, the overall flower structure of Iris missouriensis aggregate ( Fig. 5J ) resembles in some extent the Japanese I. gracilipes (namely the fringed stylar branches and the canaliculate haft of standards), and also the fruits in both groups show similar features. Furthermore, the central ridge on the outer perigone pieces in the Longipetalae irises can parallel the undissected crest found in the western North American I. tenuis . For these and other reasons, Lenz (1959) placed the latter species close to I. gracilipes , I. cristata Aiton (1789: 70) and I. lacustris Nuttall (1818: 23) among the Evansia irises.

Nonetheless, on the basis of morphological divergence, and according to their isolated position with regard to the rest of irises groups, the Longipetalae irises are here raised to genus rank, under the name Dielsiris . This group shares some convergences with members of several distant lineages, but they constitute a compact evolutionarily unit when morphology, ecology and biogeography are combined ( Mathew 1989a). However, a wider sampling is still needed for a better understanding of their true relationships.

The Rodionenkoa clade (Monospatha irises)

Irises in this group are morphologically rather diverse, though they share diagnostics characters such as the slender heterogeneous rhizome, widely creeping, which produces at the apex cord-like branches with swollen nodes ( Fig. 2K ); spathe valves 1–2, ± scarious except the base and the midrib, sometimes fused basally; outer perigone segments with a central crest ( Guo 2015), wavy to almost entire; inner segments spreading ( Fig. 5K ); stigma oblong to triangular, tongue-shapped, entire; and seeds angulose, flattened, with whitish to yellowish raphe and testa surface pitted ( Fig. 6N ). Only two species are included here, which occur disjointly in western North America (Clackamas, Oregon) and Japan (Hokkaido, Honshu and Kyushu) ( Fig. 26A ). They show chromosome numbers 2n = 28 and 36 ( Simonet 1934, Darlington & Wylie 1955), which suggest an ancestral basic chromosome number p = 7.

Both species in this group, Iris gracilipes and I. tenuis , show at first sight notable differences in their general morphology ( Brearley & Ellis 1997). The former species ( Fig. 25C ) is a slender plant with thin rhizomes producing laxe clumps, its flowers are borne on slender stalks with a single bract fused at the base to encircle the ovary ( Ohwi 1965), and the outer perigone pieces show a wavy crest raised at the tip to form a nose-like structure ( Hall 2013). The latter species produces spreading long-creeping rhizomes, not clumped; stems are deeply forked and bear flowers with two semiscarious bracts, and outer perigone segments provided with a low and entire crest, more properly a raised ridge ( Guo 2015). They have sometimes been regarded as morphologically isolated, hypothetical ancestral taxa among the beardless irises ( Hall 2013).

On the basis of such divergences, both species have been treated in different ways, probably because they show a different crest structure. In fact, Dykes (1912) placed I. gracilipes in I. sect. Evansia based on its wavy crest, albeit he left I. tenuis among members of I. sect. Apogon because its barely visible crest. A similar organization was presented by Rodionenko (1961), who described the monotypic I. sect. Monospatha to accommodate I.

A NEW GENERIC ARRANGEMENT OF IRIS SENSU LATISSIMO

Phytotaxa 232 (1) © 2015 Magnolia Press • 49 gracilipes , which along with I. sect. Lophiris (the Cristatae irises) and I. sect. Crossiris (the Evansia irises) formed I. subg. Crossiris, a group first proposed by Spach (1846) for I. japonica Thunberg (1794: 327) (as I. fimbriata Ventenat 1800 : pl. 1). However, Lenz (1959) justified placement of both I. gracilipes and I. tenuis together in I. sect. Evansia , a group so delineated to include all eastern Asian and North American crested irises. This proposal was basically followed by Mathew (1989a), who arranged them in I. subg. Limniris sect. Lophiris , a solution that has widely been accepted so far.

However, molecular phylogenies ( Tillie et al. 2001, Wilson 2004, 2011, Mavrodiev 2010, Guo & Wilson 2013) demonstrated that the Evansia irises ( sensu Rodionenko 1961 , Mathew 1989a) did not form a monophyletic group, its members being widespread in distant clades. Wilson (2011) recovered I. gracilipes and I. tenuis in a clade sister to representatives of the Longipetalae irises (the new genus Dielsiris ), and she referred the whole group as the “subg. Limniris III ” clade. In our conventional analyses the reliable Rodionenkoa clade nested among the early branching lineages of the ‘ Apogon irises’ (s.l.) sister to the Rocky Mountain irises ( Fig. 1 ), whereas in the 3TA topology it appeared as sister to the ( Limniris + Phaeiris ) clade ( Fig. 1 ).

Nonetheless, the isolated position of both species, together with the above mentioned unique combination of characters they display (namely the peculiar nature of fall crests, the oblong to triangular, entire stigma, and seeds with whitish to yellowish raphe), allow separation as an independent genus for which the name Rodionenkoa is here proposed. However, according to Rodionenko (1961) and based on the morphological differences and biogeography of both lineages in the new genus, we suggest recognition of R. sect. Monospatha to accommodate R. gracilipes (see below).

The Lophiris clade (Cristatae irises)

The eastern North American Iris cristata ( Fig. 25D ) and I. lacustris are closely related taxa which are easily distinguishable by the rizhomes heterogeneous, long slender, with cord-like branches; leaves isobilateral, slender; stems short, 1-flowered; falls with 3-ridged fringed crests on the adaxial surface (resembling a central crest of 3 parallel fringed rows), which belongs to type III of Guo (2015); and seeds with a long, coiled appendage wrapped around the seed, early withering after release ( Rodionenko 1961, Mathew 1989a). They are found in eastern North America from southern Appalachian and Ozark Mountains to the Great Lakes region ( Fig. 26B ), and have chromosome numbers 2n = 32 and 42 ( Simonet 1934) which suggests p = 7 or 8.

Both species have sometimes been connected taxonomically, I. lacustris being regarded as merely an inland,

50 • Phytotaxa 232 (1) © 2015 Magnolia Press

CRESPO ET AL.

small-sized variety ( Dykes 1912) or subspecies ( Mason & Iltis 1965) of I. cristata . Their shared unique distinctive morphology of crests, consisting of a fringed central crest flanked on both sides by a row of irregular outgrowths ( Guo & Wilson 2013, Guo 2015) ( Fig. 5L ) and the seed morphology ( Fig. 6I ) would justify such taxonomic treatments. However, morphological and karyological differences exist which warrant their recognition at specific rank (Baker 1876 a, Foster 1937, Mathew 1989a, Henderson 2002, Brearley & Ellis 1997). An attempt was made by Alefeld (1863) to segregate the species in this group into a different genus, Neubeckia p.p., though the resulting aggregate was very much artificial and extremely heterogeneous. Some authors ( Klatt 1866, Small 1933), however, adopted his treatment for taxa of the present clade. Conversely, others placed some of them together with Iris japonica (as I. chinensis Curtis 1797 : t. 373) and other taxa also showing crested falls in either I. sect. Lophiris ( Tausch 1823) , I. sect. Evansia (Baker 1876 a, Foster 1937), I. subg. Cristairis Klatt (1872b: 517), I. subg. Crossiris ( Rodionenko 1961, 2009) or the genus Evansia s.l. ( Decaisne 1874, Klatt 1882).

Wilson’s (2011) analyses, however, recovered polyphyly of the Evansia irises ( sensu Rodionenko 1961 , Mathew 1998a) and merged both Iris cristata and I. lacustris in a strongly supported (100% ML BS) early branching clade she named “ Limniris IV ”, which was well supported (80% ML BS) as sister to most of “beardless irises” ( Wilson 2011, Guo & Wilson 2013). Standard molecular approach highlights that the crested falls is a homoplasic character ( Wilson, 2006, Guo & Wilson 2013), and hence it should not be used alone as the basis for a taxonomic group. Later conventional phylogenetic analyses also placed I. cristata and I. lacustris far apart from the rest of the Evansia group and resolved the Lophiris clade as earlier divergent among the rest of irises ( Fig. 1 ). The 3TA analyses, however, placed the Cristatae irises in an early branching position with regard to most of the ‘Bearded irises’ ( Fig. 1 ).

As demonstrated by Rodionenko (1961) and supported by molecular analyses ( Fig. 1 ), I. cristata and I. lacustris form a morphologically and phylogenetically consistent group in its own that deserved taxonomic ranking, either as I. sect. Lophiris (as typified and recircumscribed by Rodionenko 1961, 2009) or I. subg. Lophiris ( Wilson 2011) . Nonetheless, the isolated position of this group together with the unique combination of characters they exhibit, allow separation as an independent genus to which the name Lophiris is here applied. As Wilson (2011) suggested, it most probably might represent an early diverging lineage among the irises.

Synopsis of the accepted genera

Figures

FIGURE 1. The summary of selected analyses, modified from Mavrodiev et al. (2014). On the left: topology resulting from both Maximum Likelihood and Bayesian Analyses (195 taxa). Statistical support mentioned above and below branches (ML Bootstrap\Bayesian posterior probabilities). On the right: topology resulting from the Three-taxon statement Analysis (80 taxa). Selected synonyms of the accepted taxa are indicated in brackets (see the text for additional names), new taxa indicated in red font. See Mavrodiev et al. (2014) for additional details.

FIGURE 2. External morphology of rootstock (bulbs, rhizomes and roots). A) Iridodictyum reticulatum, bulb with reticulate outer tunics (photo: Ken Walker); B) Xiphion heracleanum, bulb with membranous outer tunics and stem bulbil (photo: Manuel B. Crespo); C) Juno narbutii, bulb with persistent fleshy roots (photo: Ken Walker); D) Juno rosenbachiana, bulb with persistent fleshy roots (photo: Ken Walker); E) Limniris pseudacorus, rhizome with swollen annulate annual growths (photo: G.–U. Tolkiehn); F) Hermodactylus tuberosus, rhizome with pseudodigitate tubercles (photo: Attilio Marzorati); G) Junopsis barbatula, fleshy tuberous roots surrounding the inconspicuous central rhizome (photo: Ken Walker); H) Phaeiris hexagona, rhizome with fibrous remains of old leaves on scars, and fibrosus roots (photo: Ken Walker); I) Siphonostylis lazica, rhizome dichotomously branched and fibrous roots (photo: Ken Walker); J) Zhaoanthus odaesanensis, slender stoloniferous rhizomes with long, thin roots bearing nodules (photo: Christy Hensler); K) Rodionenkoa gracilipes, heterogeneous rhizome with cord-like branches (photo: Ken Walker); L) Cryptobasis bungei, short rizhomes with reddish-brown remains of leaf sheaths forming an apical neck-like structure (MO s.n.).

FIGURE 4. Flower morphology: falls, standards and stylar branches, sometimes with spathe bracts (cont.). A) Limniris sibirica (photo: Frank Hoehnel); B) Limniris chrysographes (photo: Ken Walker); C) Limniris sanguinea (photo: Andrej Baryshenko); D) Eremiris lactea (photo: Andrey Korzun); E) Sclerosiphon songaricum (photo: Vladimir Kolobincev); F) Joniris brevituba (photo: Georgy Lazkov); G) Phaeiris giganticaerulea (photo: Rodney Barton); H) Zhaoanthus henryi (photo: Ken Walker); I) Zhaoanthus minutoaureus (photo: Dennis Kramb); J) Iridodictyum reticulatum (photo: Sean Zera); K) Iridodictyum danfordiae (photo: Peter Bohnert); L) Cryptobasis mariae (photo: Yuriy Danilevsky).

FIGURE 5. Flower morphology: falls, standards and stylar branches, sometimes with spathe bracts (cont.). A) Hermodactylus tuberosus (photo: Paul Tyerman); B) Syrianthus grant-duffii (photo: Gideon Pisanty); C) Xiphion xiphium (photo: Mario Martínez- Azorín); D) Xiphion heracleanum (photo: Alicia Vicente); E) Alatavia kolpakowskiana (photo: Joyce Fingerut); F) Chamaeiris foetidissima (photo: Mario Martínez-Azorín); G) Chamaeiris graminea (photo: José Quiles); H) Chamaeiris reichenbachiana (photo: José Quiles); I) Chamaeiris anguifuga (photo: Jim Murrain); J) Dielsiris longipetala (photo: Ken Walker); K) Rodionenkoa gracilipes (photo: Ken Walker); L) Lophiris cristata (photo: Ken Walker).

FIGURE 6. Carpology. A) Belamcanda chinensis, seeds black, long persistent on the central axis of fruit (photo: Stanislav Doronenko); B) Pardanthopsis dichotoma, seeds shortly winged (photo: Ken Walker); C) Gattenhofia verna var. smalliana, seeds with fleshy raphe and ribbed surface (photo: Ken Walker); D) Tectiris tectorum, seeds glossy, pyriform, apiculate, with inconspicuous aril (photo: Ken Walker); E) Evansia wattii, seeds matte, angulose, with small aril (photo: Dennis Kramb); F) Chamaeiris foetidissima, seeds reddish, persistent on valves of the opened fruit (photo: Ken Walker); G) Chamaeiris halophila, seeds with papery, loose testa (photo: David Pilling); H) Hermodactylus tuberosus, seeds globose with gelatinous aril (photo: Ken Walker); I) Lophiris cristata, seeds with long coiled appendage (photo: Jan Sacks); J) Zhaoanthus speculatrix, seeds with aril withering early as a wing (photo: Ken Walker); K) Phaeiris giganticaerulea, seeds large, flattened, with thickened, corky testa (photo: Robert R. Pries); L) Joniris uniflora, seeds with fleshy raphe vanishing on drying (photo: Robert R. Pries); M) Limniris sibirica, seeds discoidal, without aril, with subcorky testa (photo: Ken Walker); N) Rodionenkoa gracilipes, seeds with fleshy raphe and testa surface pitted (photo: Robert R. Pries); O) Junopsis collettii, seeds oblong with conspicuous whitish aril (photo: Ken Walker); P) Zhaoanthus koreanus, seeds globosecompressed, with fleshy raphe withering early as a wing (photo: Robert R. Pries); Q) Iris hoogiana, seeds with prominent aril and testa surface reticulate-wrinkled (photo: Robert R. Pries); R) Iris pumila, seeds without aril and testa surface reticulate-wrinkled (photo: Steve Hurst); S) Juno bucharica, seeds apiculate with fleshy appendage (photo: Ken Walker); T) Siphonostylis lazica, seeds angulose with testa surface covered with sessile glands (photo: Ken Walker). Scale bars in mm.

FIGURE 18. Habit and floral details. A) Eremiris lactea (Lindley 1840: t. 1); B) Sclerosiphon songaricum (photo: Vladimir Kolobincev); C) Joniris ruthenica (photo: Aleksandr Kostjukov); D) Joniris ruthenica (Ker Gawler 1811: t. 1393); E) Phaeiris fulva (Small 1927: t. 388).

FIGURE 19. Approximate distribution areas: A) Eremiris (yellow); B) Sclerosiphon (green); C) Phaeiris (red); D) Zhaoanthus (blue); E) Hermodactylus (white).

FIGURE 20. Approximate distribution areas: A) Joniris (yellow); B) Cryptobasis (red); C) Syrianthus (green).

FIGURE 21. Habit and floral details. A) Zhaoanthus minutoaureus (Hooker 1910: t. 8293); B) Iridodictyum bakerianum (Foster 1889: t. 7084); C) Cryptobasis loczyi (photo: Georgy Lazkov); D) Cryptobasis mariae (photo: Elena Rahimova); E) Hermodactylus tuberosus (Syme 1869: t. 1496).

FIGURE 22. Approximate distribution areas: A) Xiphion (red); B) Iridodictyum (yellow); C) Alatavia (green).

FIGURE 23. Habit and floral details. A) Syrianthus grant-duffii (Baker 1889: t. 7604); B) Xiphion boissieri (Baker 1890: t. 7097); C) Alatavia kolpakowskiana (Baker 1880: t. 6489); D) Chamaeiris orientalis (Trew et al. 1773: t. 100).

FIGURE 24. Approximate distribution areas: A) Chamaeiris (red); B) Dielsiris (yellow).

FIGURE 25. Habit and floral details. A) Chamaeiris anguifuga (photo: Jim Murrain); B) Dielsiris missouriensis (Baker 1881: t. 6579); C) Rodionenkoa gracilipes (Hemsley 1903: t. 7926); D) Lophiris cristata (Small 1924: t. 320).

FIGURE 26. Approximate distribution areas: A) Rodionenkoa (red); B) Lophiris (yellow).