Two new genera of Sapindaceae (Cupanieae) from the southern Pacific: Lepidocupania and Neoarytera

Abstract Buerki, S., J. Munzinger, P.P. Lowry II & M.W. Callmander (2020). Two new genera of Sapindaceae (Cupanieae) from the southern Pacific: Lepidocupania and Neoarytera. Candollea 75: 269–284. In English, English abstract. Phylogenetic analyses of the family Sapindaceae inferred from nuclear and plastid sequence data have revealed a high level of para- and polyphyly at the subfamilial, tribal, and generic levels. A phylogenetic study focusing on taxa in the southern Pacific belonging to tribe Cupanieae has shown that the two most species-rich genera, Arytera Blume and Cupaniopsis Radlk., are polyphyletic. This study aims to clarify generic limits among the taxa currently placed in these two genera by identifying morphological features that support monophyletic groups suitable for recognition at the generic level. Specimens deposited in major herbaria holding material of these taxa were examined to complement extensive field observations. Careful consideration of morphological features in light of previous taxonomic treatments and the results of phylogenetic analyses enabled us to propose a re-aligned generic framework for Cupanieae in which two new genera are described to accommodate species previously placed in Arytera and Cupaniopsis: viz., Lepidocupania Buerki, Callm., Munzinger & Lowry (21 species) and Neoarytera Callm., Buerki, Munzinger & Lowry (4 species). A total of 25 new combinations are made, lectotypes are designated for nine names (two first step and seven second-step), and one new synonym is established. A key to the newly circumscribed genera Arytera and Cupaniopsis, along with allied genera, is provided, accompanied by information on the distribution and ecology of each species. Received: June 26, 2020; Accepted: August 28, 2020; First published online: October 29, 2020


Introduction
Phylogenetic analyses of the family Sapindaceae inferred from nuclear and plastid sequence data have revealed a high level of para-and polyphyly at the subfamilial, tribal, and generic levels (Harrington et al., 2005;Buerki et al., 2009Buerki et al., , 2010. A new informal infra-familial classification was proposed by Buerki et al. (2009), mainly based on molecular data, to accommodate these findings and to provide a framework for developing improved generic circumscriptions that meet the criterion of monophyly. Within Sapindaceae, the Cupania group, which occurs in Australasia, Asia, South America, and Madagascar, but is absent in continental Africa, corresponds to the largest radiation in terms of the number of genera and species (Buerki et al., 2009). A forthcoming genus-level phylogeny of the family based on 353 nuclear genes will provide the basis for a new, formal tribal delimitation, in which the Cupania group corresponds to tribe Cupanieae Radlk. (Buerki et al., unpubl. data). To date, the only study that has used a phylogenetic framework to test the monophyly of genera in the Cupania group is that of Buerki et al. (2012), which focused on taxa in the islands of the southern Pacific. This study demonstrated that the two most species-rich genera, Arytera Radlk. and Cupaniopsis Radlk., were both polyphyletic. Within the Cupania group, they were placed in clade B (see Buerki et al., 2012 for more details on the composition of this clade) and their members were distributed among two subclades (B-III and B-VI), within which species of both genera were intermixed (see Fig. 1). Lepiderema Radlk. was shown to be sister to species of Arytera and Cupaniopsis in clade B-III, whereas within clade B-IV, Synima Radlk. was inferred to be sister to other taxa currently placed in Arytera and Cupaniopsis.
Cupaniopsis was described by Radlkofer (1879b) to accommodate species originally placed in various other genera, including taxa such as C. anacardioides (A. Rich.) Radlk. ( Cupania anacardioides A. Rich), which was designated as lectotype of Cupaniopsis by Reynolds (1984). The first account of the genus included 44 species (Radlkofer, 1931(Radlkofer, -1934 (Fig. 2C -D), whereas Adema (1991) recognized 60 species in his monograph of Cupaniopsis, ranging from East Malesia to Australia and several South Pacific islands (Fiji, New Caledonia, and Vanuatu), with centers of diversity in Australia, New Guinea, and New Caledonia. Radlkofer (1879b) described two sections within Cupaniopsis, in addition to the nominal section, Cupaniopsis sect. Mizopetalum Radlk. and Cupaniopsis sect. Macropetalum Radlk. Neither of these two sections were recognized by Adema (1991), even though the 18 species of Cupaniopsis sect. Mizopetalum formed a clade in the cladistic analysis he performed based on morphological characters (see Adema, 1991: 50 -51, fig. 20).
Several authors have questioned the monophyly of Arytera. Radlkofer (1931Radlkofer ( -1934 commented on the morphological similarities between species of Arytera sect. Azarytera and Cupaniopsis sect. Mizopetalum as follows: "Among the Asian and Oceanic Cupanieae possessing the calyx of Matayba and only small un-crested petals with 2 scales or nearly lacking scales, the genus [Arytera] is remarkable by its capsule [that is]  […], its unappendaged aril usually completely enclosing the seed, the compact structure of its leaves and in the species of section IV [Arytera sect. Azarytera] by the lepidote leaflets, by which character this section agrees with the genus Cupaniopsis section III Mizopetalum […]." (Radlkofer, 1931(Radlkofer, -1934(Radlkofer, : 1270; translated from Latin by R. Gereau). Turner (1995: 28)  In light of the long-recognized issues regarding the delimitation of Arytera and Cupaniopsis, coupled with the recent molecular phylogenetic analyses that have clearly shown both of them to be polyphyletic, as currently circumscribed, a thorough review of both genera seems appropriate. In the present study, we aim to revise generic limits so that they correspond to monophyletic groups, and to place all of the taxa currently assigned to Arytera and Cupaniopsis within this new generic framework, supported by morphological features that characterize the clades defined by Buerki et al. (2012). Taxa belonging to the phylogenetically closely related genera Lepiderema and Synima have also been included in an effort to clarify their relationships and taxonomic identities. Our goal is to resolve generic delimitations within this challenging group and thereby further Radlkofer's quest, which started nearly 150 years ago.

Material & Methods
In order to identify potentially informative morphological characters that support monophyletic groups and to assess affinities among the taxa currently placed in Arytera, Cupaniopsis, Lepiderema, and Synima, we examined material from the following herbaria: BM, BRI, CNS, G, K, L, MO, MPU, NOU, P, SING, and SUVA. This was complemented by accessing digital images of type specimens at other herbaria through the Global Plants website [https://plants.jstor.org]. To encompass the full range of morphological variation within species across their distributional ranges, we also consulted the descriptions provided in regional floras (i.e. Flora Malesiana, Adema et al., 1994;Flora of Australia, Reynolds, 1985b) and generic monographs (i.e. Cupaniopsis, Adema, 1991;Arytera, Turner, 1995), as well as the seminal works of Radlkofer (1879aRadlkofer ( , 1879bRadlkofer ( , 1931Radlkofer ( -1934. Data were compiled on key morphological characters for each taxon. Finally, between 2009 and 2017, we also conducted fieldwork throughout much of the range of the study group, including in Australia, Borneo, the Fijian archipelago, the Moluccas, New Caledonia, Peninsular Malaysia, Singapore, and Vanuatu, in order to observe and collect material of indigenous species, and to gain insights into their morphology and ecology, as well as their relationships with other members of the family.

Phylogenetic results and morphological characters
A key step toward resolving generic delimitations within the closely related Sapindaceae that were the focus of this study involved aligning our morphological findings with previous taxonomies and discussing the results in light of the phylogenetic framework presented in Buerki et al. (2012). Biogeography was also taken into consideration as a criterion for delimiting genera. The presentation of our findings given below follows the sequence of clades presented in Buerki et al. (2012). We also have sought to corroborate our taxonomic findings by expanding phylogenetic analyses to include DNA sequences from several key species, although the results of this work are not formally included in the present paper and will instead be the subject of a forthcoming publication focusing on the evolution and biogeography of the Cupania clade in the Pacific islands and neighboring regions. Our goal here is to provide the new generic classification required as a basis for conducting evolutionary and biogeographical analyses.
Clade B-III identified by Buerki et al. (2012) (Fig. 1) contains all sampled taxa of Lepiderema, which form a subclade that is sister to another subclade comprising all sequenced species of Cupaniopsis sect. Mizopetalum (including 18 species, five of which were originally placed in this section by Radlkofer and the rest were subsequently added by Adema (1991) in his treatment of the genus). In this treatment, we are only recognizing 17 of these species (see taxonomic treatment below), together with all four sampled species of Arytera sect. Azarytera. Although the sampling from these groups used by Buerki et al. (2012) was limited, their phylogenetic results strongly suggested that species of Arytera sect. Azarytera were nested within Cupaniopsis sect. Mizopetalum, which precluded the possibility of recognizing either of these infrageneric taxa as a separate genus. The presence of lepidote scales on the vegetative and reproductive organs of these species represents a clear morphological synapomorphy for clade B-III. Two taxonomic interpretations are possible given this phylogenetic context: 1) recognize one large genus encompassing all the members of Lepiderema, Arytera sect. Azarytera, and Cupaniopsis sect. Mizopetalum; or 2) treat Lepiderema as one genus and place the remaining species in another genus. We prefer the second option because species of Lepiderema can easily be distinguished from the other members of clade B-III by their lack of petal and ovary scales (vs. petal and ovary scales present in the other taxa). Moreover, Lepiderema is restricted to Australia and New Guinea, whereas the other taxa occur in New Caledonia, Fiji and Samoa (with the exception of A. brackenridgei Radlk., which has a wide distribution in the Solomon Islands, Vanuatu, Fiji, Wallis and Futuna (Horn Islands), Tonga and Samoa). In order to accommodate the group that is sister to Lepiderema, which comprises the species previously placed in Arytera sect. Azarytera and Cupaniopsis sect. Mizopetalum, a new genus is required, which we described below as Lepidocupania Buerki, Callm., Munzinger & Lowry. The taxa belonging to clade B-IV of Buerki et al. (2012) (Fig. 1) can easily be distinguished from those in clade B-III by the absence of lepidote scales on their vegetative and reproductive organs. Synima is inferred to occupy a basal position in clade B-IV. This genus is restricted to Australia, with the exception of a single species in New Guinea, viz. S. cordieri (F. Muell.) Radlk. (Reynolds, 1985b). Synima is characterized by having crested scales on its petals and seeds, the latter fully covered with a sarcotesta (Reynolds, 1985a;Forster, 2006;Callmander et al., 2020). The remainder of clade B-IV comprises three subclades, which are fully aligned with previously recognized taxonomic entities. The species of Arytera subsect. Pacifica correspond to the first subclade, which is sister to the two other subclades, one containing the type of the genus (A. litoralis) and all species currently assigned to Arytera subsect. Arytera and Arytera subsect. Distylis, and a third subclade comprising the type of Cupaniopsis (C. anacardioides) and all species currently placed in this genus, with the exception of those belonging to Cupaniopsis sect. Mizopetalum (see above). Since each of these three subclades is consistent with a currently recognized taxonomic entity and is also morphologically coherent, we have opted to recognize them as a separate, well-defined genus. Two of the subclades correspond to Arytera and Cupaniopsis because they contain their respective types, whereas the third subclade represents a new genus, which we formally describe here as Neoarytera Callm., Buerki, Munzinger & Lowry.
Below we provide a key to the genera, which includes the two new genera, and we also include an appendix that presents a synopsis of currently accepted species of Arytera, Cupaniopsis, Lepidocupania, and Neoarytera, with their respective distributions. The taxonomy of the two new genera follows the comprehensive monographs by Adema (1991) and Turner (1995), with the exception of Cupaniopsis rotundifolia Adema, which is not accepted here. The synonymies proposed by Adema and Turner are not repeated here except for two names that require nomenclatural clarification: Arytera pachyphylla Radlk. and Cupaniopsis ganophloea Radlk.

A key to the newly circumscribed Arytera, Cupaniopsis and allied genera
The following key is adapted from Reynolds (1985b Turner, 1995: 151).
Notes. -Lepidocupania shares the presence of lepidote scales on its vegetative and reproductive organs with Lepiderema. However, Lepidocupania can easily be distinguished from Lepiderema by the presence (vs. absence) of petal and ovary scales, and it differs from Cupaniopsis by the presence (vs. absence) of glandular scales on its vegetative and fertile organs. Distribution and ecology. -According to Turner's (1995) concept, Lepidocupania arcuata is endemic to the New Caledonian archipelago, where it occurs from sea level to 200 m, predominantly on calcareous soils (Loyalty Islands, Ile des Pins), but also on sand, clay and schist. It is found in mesophyll and sclerophyll forest and scrub.
Notes. -Additional taxonomic analyses are required, especially with regard to material from the Loyalty Islands, which exhibits very peculiar indument compared to that from the main island of New Caledonia [Grande Terre]. Moreover, Turner (1995) tentatively identified a specimen from Tonga (Parks 16317: L 0468503) as L. arcuata, which, if confirmed, would significantly expand the geographic range of this species.
Notes. -This species has been assigned an IUCN risk of extinction status of "Critically Endangered" [CR] based on its narrow distribution and reduction in population size (IUCN, 2020). (Adema)  Distribution and ecology. -Lepidocupania globosa is endemic to the west coast of Grande Terre, where it is restricted to sclerophyll forest (Bouchet et al., 1995).

Lepidocupania globosa
Notes. -This species has been assigned an IUCN risk of extinction status of "Vulnerable" (IUCN, 2020). See note under L. pennelii.
Lepidocupania glomeriflora (Radlk.) Buerki, Callm., Munzinger & Lowry, comb. nov. (Fig. 3A Distribution and ecology. -Lepidocupania glomeriflora is endemic to New Caledonia, where it occurs on Grande Terre, the Ile des Pins, and the three larger Loyalty Islands (Adema, 1991;Butaud, 2014). It is found in maquis vegetation, dry to mesic forests, or forest remnants, along the coast, on hills or mountainsides, often along rivers, and occurs on calcareous substrates, serpentinite and schist, but appears not to favor peridotitic substrates and is completely absent from Grande Terre's large southern ultramafic massif. Lepidocupania glomeriflora has been recorded from 10 to 500 m.
Notes. -Nine syntype collections were cited in the protologue of Cupaniopsis glomeriflora: Balansa 153 p.p., 1447, Baudouin 354 p.p., "Culta in hort. Paris", Deplanche 83, Labillardière 169, Pancher 782, Vieillard 228, and 233. Adema (1991: 116)  Distribution and ecology. -Lepidocupania gracilipes is endemic to Grande Terre, where it grows in gallery forest, continuous tall maquis ("maquis paraforestier", see McCoy et al., 1999 for details), and thickets on (rocky) serpentine terrain, and sometimes along streams on alluvium (Turner, 1995). It seems to be restricted to serpentine, hyper-magnesium brown soils. The species has been recorded from sea level to 600 m. Distribution and ecology. -Lepidocupania grandiflora is endemic to Grande Terre, where is has a peculiar distribution and edaphic range; most collections are from the central region (Mont Do and around the Col d'Amieu), but others have been made in the Pouébo area, some 175 km to the northwest. Lepidocupania grandiflora grows in wet forest and maquis vegetation, on ultramafic and non-ultramafic substrates, from 400 to 1000 m.
Notes. -Adema (1991: 119) suggested a close affinity between this species and L. oedipoda based on vegetative characters. Their morphological similarity points toward the need for further analyses, especially focusing on collections from the Pouébo area. Distribution and ecology. -Lepidocupania guillauminii is endemic to Chuuk atoll in the Caroline Islands, where it grows in lowland evergreen forests on volcanic soil and humus (Adema, 1991).
Note. -While Adema (1991: 122) noted that the available material of this species is rather incomplete and that it resembles both L. concolor and L. samoensis, he recognized it as distinct based on several differences in the amount of indument and in features of the fruits. Distribution and ecology. -Lepidocupania inoplaea, as circumscribed by Adema (1991), is endemic to the archipelago of New Caledonia, where it occurs in the northwestern part of Grande Terre and two of the Loyalty Islands (Lifou and Maré). It is found in maquis vegetation and gallery forest on serpentinite and schist, from sea level to 700 m.
Notes. -Cupaniopsis inoplea was described based on Balansa 3307. Original material at P is mounted on three sheets, two of which [P00639140, P00639141] bear the following note in Radlkofer's hand: "Cupaniopsis inoplaea m. Radlk." Here we designate the most complete and best-preserved of these two sheets [P00639140] as the lectotype.
It has come to our attention that the original spelling of the epithet, 'inoplea', was changed to 'inoplaea' by Guillaumin (1948) and Adema (1991), but the rules of nomenclature do not justify this change, and we therefore retain the original spelling, as proposed by Radlkofer (1879b). Adema (1991) did not indicate that this species grows on calcareous substrate, although two collections cited in his monograph from the Loyalty Islands (Schmid 677 and MacKee (Leg. Suprin) 43447) were gathered without doubt from calcareous sites. Further study will be needed to determine whether these collections belong to Lepidocupania inoplea or another species.
Lepidocupania lepidota (Radlk.) Buerki, Callm., Munzinger & Lowry, comb. nov. (Fig. 3C) Distribution and ecology. -Lepidocupania lepidota is endemic to Grande Terre, where it is restricted to dense humid forest on ultramafic substrates, from 10 to 915 m. Most collections are from the main ultramafic massif of the south, although it is also recorded from a few isolated localities in the north, including Mont Do and Cap Bocage (Turner, 1995), and more recently from Kantalupaik (Munzinger et al., 2018).
Notes. -Seven syntype collections were cited in the protologue of Arytera lepidota: Balansa 1445, 2841, Baudouin 134A, Pancher [Mus. Néocal.] 222, Vieillard 205, and 206. Turner (1995 designated Pancher [Mus. Neocal.] 222 as the lectotype. The original material deposited at P is, however, mounted on two sheets, necessitating the second step lectotypification designated here, for which we have selected the most complete and best-preserved sheet [P00639113]. (Guillaumin)  Distribution and ecology. -Lepidocupania myrmoctona is endemic to Grande Terre, where it is abundant along the east coast and grows in dense humid forest, mostly on nonultramafic substrate (schist), but it has also been recorded in montane ecosystems in the main southern ultramafic massif (Adema, 1991).
Cupaniopsis ganophloea was described based on four syntypes: Labillardière s.n., Pancher 777,Vieillard 227,230 and 231. The most complete and best-preserved material bearing, in Radlkofer's hand, "Cupaniopsis ganophloea m. Radlk." is [P05309776]. This collection has no label data and could represent material of either Veillard 230 or 231. Veillard 230 in P [P00639147] only has a few fruits with the note "Cupaniopsis ganophloea m. Radlk." in Radlkofer's hand. We prefer to designate the later sheet as lectotype and consider [P05309776] as a probable isolectotype. (Guillaumin)  Distribution and ecology. -Lepidocupania pennelii is endemic to Grande Terre, where it is restricted to the westcentral coastal area between Bourail and Moindou, and one site along the east coast at the same latitude, around Saint Pol. It grows in forests on limestone, from 20 to 100 m in elevation (Adema, 1991).

Lepidocupania pennelii
Notes. -Cupaniopsis pennelii was described based on Pennel 403. Material of the collection at P is mounted on three sheets, but only one of them [P00639152] bears the name of Guillaumin's new species in his own hand, and we therefore regard it as the holotype.
Lepidocupania pennelii is morphologically similar to L. globosa and L. rosea (the latter known only from the type specimen), whose ecological preferences and distributions are nearly the same. They likely form a species complex and will require further taxonomic work.

Wallis et Futuna
Lepidocupania 1 sp. Buerki, Callm., kind assistance in Queensland. Finally, we thank Bruce Gray, Idris A. Haris, Vanessa Hequet, Steve and Alison Pearson, and Garry Sankowsky for granting permission to use their photos, and Felix Forest and Fred Stauffer for improving an earlier version of this manuscript.