A taxonomic revision of Curarea Barneby & Krukoff (Menispermaceae)

Abstract A monograph of Curarea, a neotropical genus in the plant family Menispermaceae, is presented. Curarea is distinguished from related genera by the combination of staminate flowers with sepals in two whorls and pistillate flowers with three petals, three carpels and usually elongated carpophores bearing three sessile drupelets. Nine species are recognised, amongst them two new to science, C. gentryana from Ecuador and C. barnebyana, from Ecuador and Peru. Additionally, two new combinations, C. iquitana and C. tomentocarpa, are proposed for distinct taxa recovered in a multivariate analysis of quantitative characters of the broadly distributed and morphologically variable C. toxicofera. The anatomy and morphology of species in the genus is documented, identification key, species descriptions, distribution maps and a preliminary conservation assessment for all accepted species are also provided. Of the nine species recognised here, C. barnebyana is assigned a preliminary status of Vulnerable, C. crassa (known only from the coastal Atlantic Forest in Brazil) and C. gentryana (endemic to western Ecuador) are both assigned a preliminary status of Endangered.


Introduction
Curarea Barneby & Krukoff of the plant family Menispermaceae, tribe Tiliacoreae, is a genus of understory or canopy dioecious lianas that are widely distributed in the humid forests of tropical America, mostly at low elevation from eastern Brazil to Costa Rica in Central America. Within Tiliacoreae, Curarea shares with remain-ing members of the tribe endocarps ornamented by longitudinal ribs, seeds hippocrepiform and without endosperm and fleshy cotyledons (Ortiz et al. 2016). Curarea was segregated from Chondrodendron Ruiz & Pav. to include species that differed in several floral and carpological features (Barneby and Krukoff 1971). In Curarea, the calyx of staminate flowers is composed of two whorls of three, scarcely fleshy and greenish sepals, these being variously described as sepaloid scales (Barneby and Krukoff 1971: 7) or as sepals (Barneby 1996). In Chondrodendron, the calyx is formed by four or more whorls of three, membranous sepals that are cream coloured and, hence, described as petaloid sepals (Barneby and Krukoff 1971: 7). Staminate flowers of both genera have 6 petals that are membranous and much shorter than the innermost sepals (Table 1). Pistillate flowers of Curarea have only 3 petals and 3 carpels, whilst those of Chondrodendron have 6 petals and 6 carpels. The drupelets of both genera have narrow supportive structures, but in Curarea, these are elongate carpophores that bear atop sessile fruits, while in Chondrodendron the fruits are conspicuously stipitate ( Table 1). The distinct nature of Curarea is also supported in a family-wide phylogenetic analysis of chloroplast markers (Ortiz et al. 2016), where it is recovered as monophyletic and sister to Sciadotenia Miers, in a neotropical clade that also includes Chondrodendron ( Fig. 1A; Ortiz et al. 2016). Within this neotropical clade, Curarea shares with Sciadotenia sessile drupelets borne on narrow elongated carpophores bearing six carpels in Sciadotenia and only three in Curarea. The two whorls of sepals in the staminate flowers distinguish Curarea from Chondrodendron and Sciadotenia, both of which have 3-7 whorls of sepals.
When first described, Curarea included four species: C. candicans, C. cuatrecasasii, C. tecunarum and C. toxicofera, with C. toxicofera designated as the type species of the genus (Barneby and Krukoff 1971). While C. cuatrecasasii and C. tecunarum were newly described, both C. candicans and C. toxicofera were transferred from Chondrodendron. Another species was later added to the original four with the recognition of C. crassa, an endemic new species from the Brazilian Atlantic Forest (Barneby 1996). Although the etymology of the generic epithet was never provided, the name Curarea doubtlessly refers to curare, the South American arrow-poison for which species of Curarea provide one of the ingredients.
In this first treatment of Curarea (Barneby and Krukoff 1971), which was also the most comprehensive one at that time, C. toxicofera was broadly interpreted and included entities formerly described as various taxa, rendering it a large complex with unclear patterns of morphological variation. However, subsequent works generally followed that of Barneby and Krukoff (1971) and used the five species they recognised in treatment of local, regional and national floras, e.g. Flora of Surinam (Jansen-Jacobs 1976); Flórula de Las Reservas Biológicas de Iquitos, Perú (Ortiz 1997); Flora of Ecuador (Ott 1997); Flora de Nicaragua (Ortiz 2001), amongst others. Seven species were later recognised in the most recent, taxonomic account of the genus (Ortiz-Gentry 2000) but Curarea toxicofera remained broadly circumscribed.
On the basis of shared macro-and micromorphology, Ortiz-Gentry (2000) suggested that there were two species groups in Curarea. These were not recovered in a later family-wide molecular study (Ortiz et al. 2016) where it was found that C. candicans -an outlier in the study of Ortiz-Gentry (2000) -was sister to the remaining species of the genus. Species boundaries in most Curarea are relatively well defined with the exception of C. toxicofera, which, as mentioned above, has always been broadly defined. For instance, Barneby and Krukoff (1971) treated the three former Chondrodendron species [i.e. Chondrodendron iquitanum Diels, C. polyanthum (Diels) Diels and C. tomentocarpum (Rusby) Moldenke] as conspecific with Curarea toxicofera. Moreover, the type of Cocculus toxicoferus Wedd. is a sterile specimen and, hence, lacks diagnostic features to readily distinguish the species, further complicating the application of this name and circumscription of the taxon, which like most Menispermaceae is also dioecious. To evaluate patterns of morphological variation across the entire range of Curarea toxicofera, 200 collections of the taxon were sorted into five groups based on selected features such as relative size of petioles and leaves, lamina shape and indumentum type on its abaxial surface and general features of staminate inflorescence (i.e. slender vs. coarser, pedicels and staminate flowers conspicuously larger). These groups of specimens were further evaluated using selected quantitative morphological variables. Based on the outcome of the analyses, three entities are recognised within C. toxicofera s.l., namely C. toxicofera s.s., C. iquitana and C. tomentocarpa. Two of the five initial groups within C. toxicofera, although distinguished by qualitative features, were however not recovered in the multivariate analyses. Future studies that include other lines of evidence, in addition  to morphology, might provide insights about their status and specimens associated with these groups are here tentatively placed under the species to which they are morphologically most similar, preceded by the "aff." qualifier. The present revision furthers that of Ortiz-Gentry (2000) which sought to evaluate the circumscription of species within the genus Curarea, using morphological and anatomical characters. In this study, I provide the taxonomic revision of the species recognised in that work and formally describe the new species treated by Ortiz-Gentry (2000), but were not effectively published there (viz. Melbourne Code Article 30.8). Additionally, in this study, I evaluate the taxonomic boundaries of the C. toxicofera complex based on multivariate analyses of the morphological variation across its distribution. Nine species are recognised in Curarea, of which two are newly described and two new combinations are proposed. Populations, tentatively identified here as C. aff. iquitana or C. aff. Tomentocarpa, exhibit complex variations that require additional field work in combination with molecular studies to better clarify their taxonomic boundaries. Of the newly described taxa, C. barnebyana occurs in Ecuador and Peru, while C. gentryana is endemic to western Ecuador and, in a preliminary assessment, is evaluated as endangered.

Materials and methods
The taxonomic treatment is based upon the examination of 429 herbarium collections from: A, B, BM, BR, ECON, F, G, GH, IAN, INB, K, MG, MO, NY, P, QCNE, R, RB, U, US and USM (acronyms follow Thiers continuously updated). The virtual herbaria of MG (http://marte.museu-goeldi.br/herbario), NY (http://ssciweb.nybg. org/VirtualHerbarium.asp), US (http://collections.nmnh.si.edu/search/botany) and JSTOR Global Plants (http://plants.jstor.org) were also consulted. Additional specimens were collected during field trips to Ecuador and Peru. Label information of all specimens studied are available via TROPICOS, the Missouri Botanical Garden (MO) database, as are the images of all specimens housed at MO. Distribution maps were generated using ArcGIS Desktop Release 10 (ESRI Inc., Redlands, CA), a few specimens whose labels did not indicate coordinates, but had detailed locality information were georeferenced post facto using gazetteers, these being found within brackets in the selected specimens examined section under their respective species. A complete list of all specimens examined is presented in the Supplementary material that accompanies this article.
Species boundaries in Curarea were inferred based on qualitative and quantitative morphological data (available from the author upon request). Measurements were made from dried herbarium material for leaves, inflorescences and fruits. In Curarea, as in all Menispermaceae, development of the fruit is unequal, with the abaxial side developing more than the adaxial side such that the style scar ends up near the adaxial base of the fruit (Fig. 1B). Hence, it should be noted that, although measurements of the fruit length and width seem conventional, i.e. the length referring to the dis-tance between the apex and base of the fruit, the actual apex of the fruit is near the base (Fig. 1B) because of the extreme curvature of the fruit in the Menispermoideae. Similar criteria are followed when measuring endocarps; however, the hippocrepiform embryos are measured from one end to the other, along the outer curvature.
The flowers were rehydrated before measurement. A ruler was used to measure leaves and inflorescences, a digital caliper (Digimatic CD-6" CX, Mitutoyo, Japan) was used to measure fruits and endocarps and a dissecting stereoscope (Nikon, Japan) with a micrometric ocular was used to measure flowers. Inflorescence length includes the short peduncle. The two whorls of sepals in both the staminate and pistillate flowers are always conspicuously different in size and sometimes also in shape, hence measurements of both whorls are reported. However, a single measurement is reported for the two whorls of petals and stamens in the staminate flowers, except when noticeable variation in size and shape is present. For each character, two to three measurements per specimen were taken, the ranges of the averages being used when describing the species.
Most Curarea species are morphologically distinct, except for the geographically and ecologically widespread C. toxicofera s.l. For the latter, I carried out multivariate analyses to evaluate patterns of morphological variation across its entire range. A total of 200 collections were examined and placed in five groups based mostly on variation in general features such as leaf shape and relative size, indumentum type on abaxial leaf surface, the general structure of the staminate inflorescence (i.e. lax or compact), relative staminate flower size and staminate flowers conspicuously pedicellate or subsessile. These five groups, called: al = allpahuayo, iq = iquitana, to = toxicofera, tm = tomentocarpa and uc = ucayali, below, were further tested by means of a linear discriminant analyses (lda) of selected quantitative data (Table 2) in order to assess the relative importance of the morphological traits in separating the groups (cf. Henderson 2005). Of the 200 collections, 73 specimens were staminate, 100 were pistillate and 29 specimens were sterile. To include most of the specimens, only leaves were scored for 197 specimens (Dataset 1), a total of 53 specimens were scored for all characters of leaves + staminate inflorescences + flowers (Dataset 2), 55 for staminate inflorescences + flowers (Dataset 3) and 21 specimens for leaves + fruit characters (Dataset 4). Further, a subset from Dataset 2 was extracted and analysed after removing the toxicofera group, this subset consisted of 27 specimens. All morphological measurements were recorded on an Excel 2010 spreadsheet and log-transformed prior to the analyses. Data analyses were carried out using the package MASS in R ver. 3.1.0 (The R Foundation for Statistical Computing; <http://www.R-project.org>).
A preliminary conservation status assessment for all taxa was carried out according to the IUCN (2012) criteria. Data were analysed using the Geospatial Conservation Assessment Tool, GeoCat (http://geocat.kew.org/) with the default 2 km grid cell.
Features such as wood anatomy, foliar anatomy and leaf venation patterns were also examined. While the few samples per species available for the examination of the characters precluded their inclusion in quantitative analyses, these observations further document interspecific variation in the genus and may provide another line of evidence to distinguish species in Curarea when more samples are included. For wood anatomy studies, stem samples ca. 1.5 cm diameter were available for Curarea aff. iquitana (Diels) R. Ortiz [Ortiz 181 & 186 (MO)], C. aff. tomentocarpa (Rusby) R. Ortiz [Ortiz & Vargas 197 & 199 (MO)] and of about 5-15 cm wide for C. tecunarum [Ortiz & Vásquez 214,Ortiz et al. 220 [Ortiz & Vargas 200 (MO)]. Permanent slides were prepared following standard techniques in wood anatomy studies. Terms are defined and features are described in accordance with the IAWA committee (1989) guidelines. Table 3 show averages based on 10 measurements of randomly selected cells of the individuals examined.
Sample preparations for foliar anatomy and leaf venation patterns followed Keating (2014). Transverse sections of petiole and mid-lamina, including midrib and lamina margin, were taken. For the petiole, both the apical and basal pulvini as well as the  Cuticle preparations were obtained following methods cited in Christophel et al. (1996). Pieces of leaf of 1 cm 2 were taken from near the base of the lamina, placed in test tubes and soaked in 70-95% ethanol for about 24 hours. The ethanol was then decanted and 10 drops of 40% hydrogen peroxide and 5 drops of 90% ethanol were added to the tube and then gently boiled in a water bath until the sample turned light yellow. At this point, cuticles were peeled off and cellular material brushed away. Cleaned cuticles were then stained in 0.1% Crystal Violet, wet mounted in CaCl 2 and observed under the light microscope. Vouchers for leaf anatomy including cuticle and stomata are as follow: C. barnebyana [Ortiz & Vargas 194 & 200 (MO)], C. candicans [Jansen-Jacobs et al. 1995 (U) Scanning electron microscope (SEM) observations of leaf trichomes and stomata were made on the adaxial surface of the samples. Samples of 5 mm 2 were removed from mid-lamina of apparently adult leaves of herbarium specimens and mounted on SEM stubs with double-sided carbon conductive tabs and coated with gold using a SEM coating unit E5000. Stomatal classification followed Wilkinson (1979) and trichome description and classification followed Theobald et al. (1979). Observations and photographs were made with a Hitachi S-450 SEM @ 20KV using Polaroid panchromatic film.
The difficulty in matching collections of male and female individuals in Menispermaceae has frequently been stressed in the past (Sandwith 1930, Barneby andKrukoff 1971) and the latter authors' observations regarding the few pistillate flowers in herbarium collections is still relevant at present. In addition to being dioecious, Menispermaceae are also, for the most part, canopy lianas and so difficult to reach, hence they are usually overlooked in biological inventories. It is noteworthy that, although in herbarium collections, there are frequently more pistillate than staminate plants, these pistillate plants are mostly in fruit of different degrees of maturity and therefore are not always readily comparable across species. This is also true for Curarea, therefore species are distinguished mostly by characters of the staminate inflorescences and staminate flowers. These latter features are supplemented by characters of the fruits, when available.
In this study, species are regarded as segments of separately evolving lineages of populations, as in the general lineage species concept (de Queiroz 1998(de Queiroz , 2007. The criterion used to infer species boundaries is morphological discontinuity, species being diagnosed by unique combinations of qualitative and quantitative morphological features. These morphological features are, as indicated above, mostly based on characters of staminate plants and, for the most part, collections of pistillate plants are assigned to the recognised species if they occur in the same localities where staminate plants have been collected. This approach is rather arbitrary and could be viewed as questionable, especially when the species co-occur as is the case of C. toxicofera s.s. and the newly resurrected taxa; however, at present doing it in this manner is unavoidable in a dioecious group that is very poorly represented in herbaria collections.

Habit and stem
Species of Curarea are all canopy or understory lianas, the canopy species attaining a height of ca. 30 m, the understory species usually only reaching about 10 m high. In understory species, stems are more or less terete (flattened in C. toxicofera and ca. 3 cm wide, Grassl 10076), with rings or partial rings centric to weakly eccentric, rarely strongly eccentric, in transverse section and 1.5-4 cm in diameter/width ( Fig. 2A), while in canopy species, stems are consistently strongly flattened, the partial rings strongly eccentric in transverse section and up to ca. 40 cm in width (Fig. 2B). In both, bark of older stems is usually relatively thin, brittle and smooth, but it can be rugose with irregular shallow lengthwise fissures. Conspicuous tuberculate lenticels are frequent in C. cuatrecasasii and in C. gentryana and, to some extent, in C. iquitana.

Wood anatomy
Wood anatomy of C. crassa, C. gentryana and C. toxicofera s.s. was not investigated. For the remaining species, mostly one sample per species was examined. Transverse sections of the stem of Curarea, as of several other Menispermaceae, show an irregular pattern of secondary growth consisting of successive, complete or partial rings of vascular bundles, which are separated radially by wide interfascicular rays and tangentially by pericyclic fibres and tangential cortical parenchyma. This type of growth is typical of Menispermaceae and has been termed "anomalous secondary growth" (Solereder 1908); "included (interxylary) phloem" (Chalk and Chataway 1937;Metcalfe and Chalk 1950), "xylem and phloem concentrically alternating" (Metcalfe and Chalk 1983) or "successive cambia" (Carlquist 1988(Carlquist , 1996Jacques and de Franceschi 2007).
Within Curarea, species differ moderately in their pith diameter, distribution patterns of vascular bundles and rays, occurrence of medullary sclerenchyma (e.g. fibre caps) and number and mean tangential diameter of vessels. Variation in these and other features are summarised in Table 3.
Successive and centric to weakly eccentric patterns of few (3-7) complete or partial rings are found in C. aff. tomentocarpa ( Fig. 2A) and C. cuatrecasasii and several (ca. 45) strongly eccentric arches are observed in C. tecunarum (Fig. 2B), C. candicans and C. barnebyana. Wide rays (275-340 µm) in combination with more or less straight tangential margins and vascular bundles with smaller vessel diameter are observed in C. aff. tomentocarpa (Fig. 3A), C. candicans and, to some extent, in C. cuatrecasasii. There are tangential wedge-like layers of sclereids intruding into the rays in C. tecunarum (Fig. 3B) and C. barnebyana but, in all other species examined, the rays are tangentially uniform. Relatively narrow rays (150-136 µm) with conspicuous irregularly angulate radial margins combined with relatively wide (220-228 µm) vascular bundles with large tangential diameter of vessels and strongly developed medul-  (Ortiz 197), showing wide rays B flattened stem of C. tecunarum (Ortiz & Vásquez 214) showing successive partial rings. Scale bar: 0.5 cm (A); 1.7 cm (B). lary sclerenchyma in the pith are characteristic of C. tecunarum (Fig. 3B) -this alone lacks crystalline inclusions-and C. barnebyana.
Relatively narrow rays with conspicuous irregularly angulate radial margins in combination with vascular bundles with large tangential diameters are also observed in C. cuatrecasasii. No representative of C. toxicofera s.s. was available in this study and sterile specimens provisionally identified as C. toxicofera show moderate to strongly eccentric arches of successive growth, although the width reaches only up to 3 cm.

Leaves
Within an individual, the shape and size of the lamina varies with age, position along the stem and exposure to sunlight. Thus, large blades with caudate apices are usually either juvenile and/or found in parts of the plant that are in shaded areas, while smaller blades with acute or acuminate apices are found in mature leaves that are usually exposed to direct sunlight. The latter are more conspicuous in species that reach the canopy. The type collection of C. tecunarum -i.e., Krukoff 8713 and a sterile specimen of Curarea (Cerón 2717), which remains as Curarea sp., show a few leaves with cleft apices. The margins of blades are entire, but on rare occasions, they are scarcely lobed to apically trilobed in sterile specimens of C. tecunarum (Vásquez et al. 15101, Lewis et al. 11759) and the sterile specimen L.C. Richard s.n., the type of C. candicans, also has an apically trilobed leaf blade. Leaf blades are chartaceous or subcoriaceous, sometimes thinly fleshy when young, surfaces are usually conspicuously discolorous ( Fig. 4A-B) resulting from the dense silvery indumentum covering the adaxial surface (Fig. 4B), especially in juvenile leaves. Petioles are shorter in leaves of the canopy and longer in leaves of the understory and/or shaded areas, pulvinate at both ends, with the apical one more conspicuous.  (Ortiz & Vargas 197) B moderately large vessels, sclerotic pericyclic fibres (black arrow), strongly developed medullary sclerenchyma (white arrow) and sclerenchyma strongly intruding into the rays (two white asterisks), C. tecunarum (van der Werff & Vásquez 13909). Scale bar: 0.8 mm (A, B).

Indumentum
The indumentum in Curarea consists of two-celled unbranched trichomes with a short basal cell and an elongate terminal cell ( Fig. 4C-D). Trichomes occur singly or in pairs.
Several-lobed trichomes have been reported for C. toxicofera by Wilkinson (1989), but could not be confirmed here. It is likely that the tendency for the terminal cells to collapse in an irregular fashion in dried and/or aged samples gives the impression of trichomes being lobed. Indumentum density varies with age in some species, frequently disappearing along the stems. This variation is of taxonomic use in Curarea. Thus, uniformly dense indumentum concealing the lower leaf surface is common in C. crassa, C. barnebyana and C. tecunarum (Fig. 4E), whereas the indumentum mainly localised in the areole spaces, is common in C. candicans (Fig. 4F), C. cuatrecasasii and less frequently also in C. toxicofera s.s.

Venation pattern
The venation in Curarea conforms to the acrodromous type (sensu Hickey 1973Hickey , 1979 as observed in mature leaves. Primary veins vary from 3-5(-7), (Fig. 5A-B), however in one group of species (i.e. C. candicans, C. crassa, C. barnebyana and C. tecunarum, Fig. 5B), the innermost pair runs almost the whole length of the lamina and arches towards the apex, with only one pair of loosely brochidodromous secondary veins (this pair sometimes absent) and thick and coarse veinlets (higher-order veins), (Fig. 5C).
In the other group of species which includes C. cuatrecasasii, C. gentryana, C. iquitana, C. tomentocarpa and C. toxicofera, although the innermost pair runs more than half the length of the lamina, frequently there are 2-3 pairs of strongly brochidodromous secondary veins and moderately thin and loose veinlets (Fig. 5B, D). Leaf anatomy: adaxial cuticle, transverse sections of petiole, lamina and mid-vein Most of these features were studied by Wilkinson (1989) for C. candicans and C. toxicofera s.l. when describing characters common to Tiliacoreae as a whole.
Here, I add observations of the remaining species, the main variations being summarised in Table 4. Adaxial epidermal cells in all species are squarish, rectangular or irregularly shaped. In C. candicans (Fig. 6A), C. crassa, C. barnebyana and C. tecunarum these have abundant tanniferous cells, but there is little or no tanniferous cells in C. cuatrecasasii (Fig. 6B), C. gentryana, C. iquitana, C. tomentocarpa and C. toxicofera.
Curarea species are exclusively hypostomatic, as noted by Wilkinson (1989). The stomata, between the strongly projecting veins (Fig. 6C), are concealed by a dense indumentum. They occur typically in groups of 7-102 per areole, with the lowest number observed in C. tomentocarpa and the highest in C. barnebyana (Table 4). Stomata are elliptic in outline and immediately surrounding the aperture is a distinct peristomatal rim ( Fig. 6D-E), the cell guards range in size from 26-36 × 23-35 µm. In all species, stomata are raised on a short column of ca. 4 cells (5-6 cells fide Wilkinson 1989). Wilkinson (1989) described this arrangement of the epidermal cells under the guard cells as cyclocytic. Such projecting stomata with thick peristomatal rims are frequently associated with ecological factors such as high humidity and poor soils (Solereder 1908;Wilkinson 1979), conditions typical of the habitats occupied by species of Curarea.
The apical pulvinus of the petiole is more or less circular to weakly triangular in transverse section and weakly adaxially flattened in C. candicans, C. crassa and C. tecunarum. There are no trichomes or only a few persistent basal cells in C. candicans, C. cuatrecasasii, C. gentryana, C. iquitana, C. tomentocarpa and C. toxicofera; usually there are abundant trichomes in the remaining three species. The vascular tissue is a ring of 8-13 vascular bundles. Sclerenchyma cells surround the vascular tissue, externally forming an interrupted or continuous ring of several layers. There are a few small groupings of sclereids (stone cells) in the inner parenchyma of the cortex and, in the pith, a starch sheath (i.e. endodermis) is found around the vascular tissue in all species. A pulvinus at both ends of the petiole is a common feature in Menispermaceae and they are thought to bring the lamina into the optimum position to receive light (Wilkinson 1989). Pulvini have a large volume of parenchyma; therefore, they are always swollen compared to the rest of the petiole (Esau 1967). The mid-region of the petiole is circular, from about 1 mm diameter in C. crassa and C. cuatrecasasii, to about 2.1 mm in diameter in C. iquitana. Trichomes are few to rather abundant, frequently only the basal cells are present. There are 8-14 vascular bundles surrounded by a continuous ring of thick-walled sclerenchyma cells, the layers of parenchyma cells are much reduced and frequently thick-celled parenchyma is found in the pith (Fig. 7A).
Transverse section of the lamina shows mesophyll palisade cells that are somewhat short and loosely arranged especially the inner layer in C. cuatrecasasii, C. iquitana, C. gentryana, C. tomentocarpa and C. toxicofera. Palisade cells are moderately long and more or less compactly arranged in C. barnebyana, C. candicans, C. crassa and C. tecunarum. The midrib is adaxially concave and abaxially strongly raised and trapezoid or sub orbicular but always with abundant trichomes in C. barnebyana, C. crassa and C. tecunarum (Fig. 7B). In C. cuatrecasasii, C. gentryana, C. iquitana, C. tomentocarpa and C. toxicofera, the midrib is adaxially flat to weakly convex and abaxially strongly raised, triangular or trapeziform with few or no trichomes (Fig. 7C). The midrib in C. candicans adaxially is somewhat flat to weakly concave while abaxially, it is strongly raised and somewhat trapeziform with sparse trichomes or glabrate. As noted by Wilkinson (1989), the vascular system in Curarea is frequently collateral. There are 3-9 vascular bundles distributed in an arc ( Fig. 7B-C), the three central ones being largest. The vascular bundles are surrounded by several layers of sclerenchymatous cells. Adaxial to the vascular bundles there is a continuous (e.g. C. barnebyana, C. candicans, C. crassa and C. tecunarum) or sometimes interrupted (e.g. C. cuatrecasasii, C. gentryana, C. iquitana, C. tomentocarpa and C. toxicofera) layer of chlorenchyma cells and one or several layers of sclerenchymatous cells immediately below the epidermis.
Abaxially, parenchyma cells are abundant, with few to numerous stone cells scattered throughout and below the abaxial epidermis are sclerenchymatous cells.

Inflorescences
As all Menispermaceae, Curarea is also dioecious, with staminate and pistillate inflorescences in different plants. Staminate and pistillate inflorescences are solitary or more frequently fascicled, cauliflorous and arising from old leafless stems, axillary, supra axillary or terminal on young shoots. Inflorescences are basically thyrsi (sensu Weberling 1992), however, in pistillate inflorescences, the primary branches are sometimes reduced to solitary flowers, thus appearing racemiform; inflorescence axes are slender or moderately stout, sometimes conspicuously ridged, variously densely pubescent, with trichomes adpressed or ascending. The inflorescence bracts subtending the primary branches are ovate to lanceolate, markedly concave and usually fleshy, the adaxial side glabrous, the abaxial side variously pubescent, the indumentum appressed or spreading.
Staminate inflorescences usually multiflorous (Fig. 8A); lax primary branches with (2-)4-6 branching orders are characteristic of C. cuatrecasasii, C. gentryana, C. iquitana, C. tomentocarpa, C. toxicofera (Fig. 8A) and C. tecunarum. Rather condensed/ compacted primary branches with 0-2 branching orders appearing umbelliform or irregularly cymose occur in C. crassa, C. barnebyana and C. candicans; occasionally, C. candicans may have 3 orders of branching. In C. tecunarum, higher branching orders frequently are reduced (i.e. they are not fully dichotomous and flowers are sessile or subsessile), occasionally in C. gentryana and C. aff. iquitana, reduced branching occuring on alternating sides along the length of the primary branches, appearing racemiform. In most species, the primary branches are moderately thick, but they are rather thin and filiform in C. cuatrecasasii and C. gentryana.
Pistillate inflorescences are usually pauciflorous (Fig. 8B), with the primary branches mostly simple dichasia subtended by an inflorescence bract. Some or all of the primary branches may be reduced to single flowers, thus resembling a raceme and these flowers are subtended by bracts. In branched inflorescences of both staminate and pistillate there may no be obvious floral bracts.

Flowers
Flowers are unisexual, actinomorphic and trimerous; usually pedicellate, less frequently sessile. The pedicels are usually terete, moderately slender and sometimes ridged. At the apex of the pedicels there are 1-3(4) minute, more or less deciduous structures, usually alternating with the outer whorl of sepals. Although their morphological nature is often unclear, to facilitate description, here I am referring to them as bracteoles. These structures are most commonly ovate-lanceolate, narrowly ovate to ovate or oblong, fleshy, glabrous adaxially, variously pubescent abaxially. They are most often absent in staminate flowers and, hence, frequently not mentioned in earlier descriptions (i.e. Barneby and Krukoff 1971).
Staminate flowers are usually light coloured, frequently described as cream, whitish, greenish, yellowish, orangish, greyish or brownish; although flowers are usually pedicellate, they can be occasionally sessile in C. toxicofera and C. barnebyana and more often so in C. tecunarum. Sepals are commonly 6(-9), free, in 2(-3) unequal whorls (sometimes spirally arranged in C. candicans), mostly ovate, narrowly ovate to obovate, oblong, rhombic, elliptic or suborbicular, usually weakly concave and scarcely fleshy, glabrous adaxially, variously pubescent abaxially, the outermost whorl being smaller than the innermost whorl. The inner sepals have valvate aestivation and, after anthesis, their tips are usually reflexed (rarely erect). Petals are (5-)6(-9 in C. tomentocarpa), free, arranged in 2(-3) usually similar whorls, they are conspicuously smaller than the inner sepals, usually obovate-trilobed to rhombic or spatulate members, weakly concave, membranous, glabrous adaxially, glabrous or tomentellous abaxially, the lateral margins strongly inflexed and clasping the filaments of the stamens immediately opposite to them. In C. gentryana, the lateral margins of the inner whorl are adaxially connate or coherent. The stamens are (3-5)6, (Fig. 8C), arranged in two, more or less similar whorls (often one in C. gentryana); the filaments are free or variously connate (connate half their length in C. aff. iquitana to fully connate in C. gentryana) and clavate, clavate-sigmoid to almost terete, glabrous (abaxially tomentellous in C. toxicofera). The anthers are tetrasporangiate and dithecous, basifixed, ellipsoid or subglobose in shape, erect or slightly incurved and dehiscing by latrorse longitudinal slits. They are yellow in colour when fresh and light brown or cream when dried. Frequently, the sporangia appear half-immersed in the dilated connective. In all species, the connective is conspicuous on both sides, however, it is frequently thicker adaxially and variously protruding and it sometimes grows beyond the thecae to form a horn-like structure, as in C. iquitana (Fig. 8D). The connective is sometimes thinner abaxially and, at the apex in C. tomentocarpa and here, each theca may separate at the apex. Pistillode absent.
Pistillate flowers are greenish, yellowish or brownish coloured; pubescent abaxially. Sepals 6-9, free, in two, or more commonly three unequal whorls, mostly ovatelanceolate to obovate, oblong, elliptic to rhombic or spatulate, usually weakly concave and scarcely fleshy to fleshy, glabrous adaxially, variously pubescent abaxially, the outermost whorl being smaller than the innermost whorl. The inner sepals have valvate aestivation and, after anthesis, their tips are usually reflexed (erect). Petals are usually in a single whorl of three, less frequently 4-6 petals are found in Curarea aff. iquitana, C. tecunarum and C. tomentocarpa, they are spatulate and weakly concave, membranous, glabrous adaxially, glabrous to scarcely pubescent abaxially. The petals are always smaller than the innermost whorl of sepals and are usually opposite the carpels. There are three (seldom four in C. tomentocarpa), free carpels, the ovaries are usually coherent when young, strongly gibbous and densely tomentose; the styles are glabrous and usually terete, weakly tapering distally, often reflexed, weakly grooved adaxially in transverse section and commonly subpersistent and located near the base of the fruit; the stigma is inconspicuous. Frequently only one carpel reaches maturity.

Infructescences and drupelets
Infructescences are similar to inflorescences, but are sometimes lenticellate or have exfoliating bark. Fruiting pedicels are clavate or terete, at times inconspicuous, slender or rather thick. As the carpel develops in C. iquitana, C. tomentocarpa, C. toxicofera, C. cuatrecasasii and C. gentryana (i.e. all understory species in which the fruits are close to the ground), the trilobed gynophore enlarges into three relatively short to moderately long stalks (Fig. 8E). This enlargement of the gynophore produces structures known as carpophores, which support the drupelets of each carpel (Forman 1975(Forman , 1986. They are weakly terete or clavate, straight or distally weakly incurved and reach 2.7-11.3 mm in length, the longest being observed in C. toxicofera. Carpophores are also found in other genera in the tribe Tiliacoreae (i.e. the neotropical Sciadotenia and some African and Indomalesian species of Tiliacora ;Forman 1975), as well as other genera outside Tiliacoreae (viz. Elephantomene of Anomospermeae, see Krukoff and Barneby 1974; and Anamirta of Coscinieae, see Forman 1975). However, in C. barnebyana, C. crassa and C. tecunarum, this enlargement does not occur and only a subglobose, weakly trilobed structure is observed at the distal end of the peduncle and, in C. candicans, this is somewhat drum-like; these four species bear fruits in the canopy.
Drupelets are oblongoid, ellipsoid, obovate or subglobose ( Fig. 8E-H) and are weakly laterally compressed when dried. They are obliquely attached to the summit of the carpophore, sessile or less frequently gradually attenuate toward the base, forming a short stipe ca. 3 mm long in some specimens of C. cuatrecasasii. The exocarp may be thin or thick, mealy-coriaceous in C. toxicofera (sensu Barneby and Krukoff 1971), the surface is smooth, rugulose or muriculate, pilosulous or velutinous-hispidulous, the trichomes being erect. The fruits are pale orange to yellow when ripe and frequently leathery outside and granular inside when dried. Although, in a strict sense, the single-layered outer epidermis of the pericarp is referred to as the exocarp, here I am following earlier characterisations of Curarea fruit by Barneby and Krukoff (1971), who described the several-layered outermost zone of the pericarp (including the epidermis) as the exocarp. It is worth noting, however, that it appears that Barneby (1996) interpreted the exocarp as the epidermis only and the remaining layers following the epidermis as the mesocarp when describing the fruit in C. crassa from dried herbarium material (e.g. Barneby 1996: 22). The mesocarp is white, fleshy and mucilaginous when fresh (Fig. 8F) and, when dried, turns into filamentous plates that adhere to the endocarp (Barneby and Krukoff 1971). Due to the eccentric development of the carpel, the scar of the style is found near the base of the developed fruit. The endocarp is oblong-ellipsoid or obovoid, hippocrepiform (i.e. horseshoe-shaped) (Fig. 8I), a shape produced by the unequal development of the abaxial side compared to the adaxial side (Botha 1980;Forman 1986); although the hippocrepiform shape is more noticeable from within rather than from the external surface. The texture is typically papyraceous (crustaceous in some specimens of C. toxicofera). Commonly the surface is fairly smooth, but muriculate or weakly reticulate in C. toxicofera (Fig. 8I). Seeds, like the endocarps, are hippocrepiform ( Fig. 8J), semicircular in transverse section and brownish when dried. Seeds in Curarea, as in most members in the tribe Tiliacoreae, lack endosperm, the cotyledons are thick, fleshy and accumbent, both are usually of equal size, but in C. barnebyana one of them may be smaller and somewhat J-shaped. The seed (embryo) is hippocrepiform (i.e. curved around the condyle). The latter has not been studied in developing carpels in Curarea, however, features of the mature endocarp suggest that it conforms to a bilaterally compressed septiform condyle as described by Ortiz (2012); the condyle corresponds with a shallow groove on the lateral sides of the external surface of the endocarp.

Phytochemistry
Several members of Menispermaceae are known to contain toxic compounds and bisbenzyltetrahydroisoquinoline alkaloids are very common in the family (Bruneton 1995, Barbosa-Filho et al. 2000. Amongst the best known is (+)-tubocurarine, a muscle relaxant produced by Chondrodendron tomentosum, a close relative of Curarea and the basis for subsequently developed synthetic forms (Bruneton 1995, Tuba et al. 2002. Curarea toxicofera produce alkaloids with curare-like activities such as (-)-curine and (+)-chondrocurine, (+)-isochondrodendrine; the latter has also been reported in C. tecunarum (Barbosa-Filho et al. 2000).

Ethnobotany
Menispermaceae are rich in medicinal and toxic compounds and many species are used to cure a variety of illnesses across their global distribution. Amongst the best-known products is curare, the South American arrow and dart poison of which an overall introduction is given by Krukoff and Moldenke (1938). Curare is a general name for a large group of poisons with muscle paralysis activities (Neuvinger 1998) and several species of Curarea are a source of this type of poison. Curarea candicans [= Chondrodendron candicans (Rich. ex DC.) Sandwith)] was reported by Sandwith (1930), based on the label information of Jenman 5199 to be used by the Warrou Indians of Guyana as a source of the "other kind of poison", a statement that was interpreted by Krukoff and Moldenke (1938) to refer to a different substance from the curare used by the Macusi Indians, also from Guyana.
Curarea tecunarum, variously identified as Chondrodendron polyanthum (Diels) Diels or as Chondrodendron limaciifolium (Diels) Moldenke, was reported as one of the main sources of curare for the Tecuna Indians from Brazil (Krukoff and Smith 1937;Krukoff and Moldenke 1938;Krukoff and Barneby 1970). The early report of Krukoff and Smith (1937) was later mentioned by Macbride (1938) in the Flora of Peru, under the name of Chondrodendron polyanthum. An extract from C. tecunarum is used as a long-term oral contraceptive for both men and women of the Denís tribe of Brazil (Barneby and Krukoff 1971;Prance 1972).
Curarea toxicofera (as Cocculus toxicoferus Wedd.) has long been known as an important source of curare for the Yaguas and Orijones of Amazonian Peru (de Castelnau 1851). Similar uses in this region have since been reported Barneby 1970, Ayala Flores 1984), also for Brazil (Prance 1972), Colombia (García-Barriga 1992) and Ecuador (Cerón 1995). Chondrodendron bioccai G.Lusina (at present provisionally included in the synonym of Curarea toxicofera), was stated to be used as a principal source in the preparation of curare by the Maku Indians from Alto Rio Negro (Lusina 1954). Similarly, the sterile specimen (Mexia 6321a), previously identified as Chondrodendron iquitanum (Krukoff and Moldenke 1938) and later as Curarea toxicofera (Barneby and Krukoff 1971), was reported to be used as a source of dart poison by Indians in the Rio Santiago area. However, as the specimen in question was not available in this study, it is unclear to which of the two species it belongs.

Geographical distribution and habitat
Curarea is relatively widespread within tropical America, ranging from Costa Rica through Panama, to Bolivia, French Guiana, western Brazil and the Atlantic Forest of South-eastern Brazil (Figs 9, 21 and 25).
Species of Curarea are typically found in lowland tropical moist forests and lower montane forest (up to 1300 m elevation). They grow in periodically flooded forest (either varzea or igapó) to upland non-flooded forest, typically in primary forest, but are also occasionally present in secondary growth. The species are apparently locally rare and mainly allopatric, however, Curarea aff. iquitana, C. toxicofera and C. tecunarum were observed growing sympatrically in Peru.

Phylogenetic affinities
In phylogenetic analyses of the family using plastid markers, Curarea was recovered as sister to Sciadotenia, in a clade that also includes Chondrodendron (Ortiz et al. 2016).
These affinities are congruent with earlier suggestions (Barneby and Krukoff 1971: Triclisieae sensu Diels 1910; Tiliacoreae sensu Forman 1982, Ortiz et al. 2016. With Sciadotenia, Curarea shares palmatinerved or plinerved venation and more or less sessile drupelets borne on elongate carpophores. With Chondrodendron, Curarea shares stomata aggregated in areolar depressions and raised above the epidermal cells on a short column of cells and a thyrsoid staminate inflorescence. In the study of Ortiz et al. (2016), C. candicans is recovered as sister to a clade in which C. cuatrecasasii, C. barnebyana, C. tecunarum and C. toxicofera s.l. are all unresolved (Fig. 1B).
Morphological characters suggest two species groups within Curarea (Table 5), although these are not congruent with the results from analysis of molecular data. In the group including C. barnebyana, C. candicans, C. crassa and C. tecunarum, the innermost pair of primary veins of mature leaves, which are in the canopy, conforms to an acrodromous perfect venation, with all but C. tecunarum also sharing staminate inflorescences with compact primary branches that have few branching orders. All the above listed species, except for C. candicans, share villous indumentum on their abaxial leaf surface and short subglobose carpophores. Wood and leaf anatomy features shared amongst these species are summarised in Table 5. Curarea candicans differs from the other species in the first group by its web-like indumentum and its unique drum-like carpophores, its narrow vessel elements being more similar to species of the second group. In the latter group, which includes C. cuatrecasasii, C. gentryana, C. iquitana, C. tomentocarpa and C. toxicofera (Table 5), the innermost pair of primary veins on mature leaves are acrodromous imperfect, the species have a strigillose-tomentellous indumentum on their abaxial leaf surface (weblike/tomentellous in C. gentryana), staminate inflorescences with lax primary branches, these with 4-6 branch orders, elongate carpophores and a few other anatomical features (Table 5). However, C. cuatrecasasii shares wide vessels with species in the first group.
The principal component analyses (PCA) of the datasets indicates no clear pattern in the morphological variation of the C. toxicofera complex (not shown). Of the five datasets, the linear discriminant analyses (lda) of dataset 2, is presented here. The analysis of this latter dataset, which corresponds to a combination of leaves + staminate inflorescences + staminate flowers, was significant (Wilk's lambda = 0.0043335, P < 0.001) and correctly assigned all individuals to their previously defined groups. The first two canonical discriminant functions recovered three clusters, namely groups to, tm + uc and al + iq (Fig. 10A). Moreover, the toxicofera (to) group is recovered as an independent cluster by the first canonical discriminant function, whereas al, tm, iq and uc, show various levels of overlap (Fig. 10B). The first  two discriminant functions explained 68% and 16% of the variation respectively. Traits strongly correlated with the first discriminant function are inner whorl anther length, outer whorl anther length and inner whorl petal length. On the other hand, leaf width, leaf length and bract length are only weakly correlated with the second function. A cross validation (leave-one-out) classification showed that, overall, 70% of the specimens were correctly classified and individuals of the tm and to groups were 79% and 92% correctly classified, respectively. Analysis of the subset of dataset 2, which excluded the toxicofera group, recovered three clusters, al, iq and tm + uc, however the Wilk's test was not significant (Wilk's lambda = 0.000022, p = 0.2368). Similarly, a cross validation analysis indicated that, overall, only 37% of the specimens were correctly classified (not shown).
Hence based on the results of the lda analysis of the quantitative traits in dataset 2, a conservative approach is to recognise: the toxicofera (to) group, the tomentocarpa (tm) group (including the ucayali [uc] group) and the iquitana (iq) group (including the allpahuayo [al] group) as separate entities (i.e. as three distinct species).
The groups, here evaluated, appear to vary across an ecological gradient and the al and uc groups may represent separate entities that could be recognised when more material becomes available. Of the groups here recognised, one corresponds to C. toxicofera s.s., the other two (iq and tm) were also previously described and two new combinations are here proposed.
Distribution and ecology. Andean foothills of eastern Ecuador and eastern Peru (Fig. 9), at elevations of 200-450 m in tropical wet forest. Staminate flowering specimens were collected in January, June and October; fruiting and old pistillate flowering specimens were collected in January and June.
Eponymy. The specific epithet honours the late Dr. Rupert C. Barneby whose work has laid the foundations for all subsequent taxonomic studies of neotropical Menispermaceae.  Conservation status. The calculated Extent of Occurrence (EOO) based on ten collections representing six localities of C. barnebyana resulted in 72,674 km 2 , whereas the Area of Occupancy (AOO) was estimated as 24 km 2 . Of the six sub populations representing four locations, two of the latter are found nearby private or national protected areas in eastern Ecuador and is very likely that more individuals will be found within these areas. Of the two other locations found across the border, in Peru, one of them is in a rather continuous tract of forests, whereas the other is found in a nearby area where there is increasing deforestation, which may result not only in a decline in habitat quality, but also likely reduction of the geographic range of the species in the future. Based on these considerations, C. barnebyana is assigned a preliminary status of "Vulnerable " [VU, A3c + B1b(i,ii,iii,iv) + B2b(i,ii,iii,iv)].
Discussion. C. barnebyana is recognised by its large obovoid or ellipsoid and weakly laterally flattened drupelets covered with a dark brown villous tomentose indumentum and borne on claviform fruiting pedicels. Similar indumentum is found in C. tecunarum, but the primary branches of the staminate inflorescences of the latter are laxly branched, while the primary branches of the inflorescences of C. barnebyana are condensed similar to those of C. crassa and C. candicans (see discussion under C. crassa). Shared anatomical features amongst these species, which make up group 1, are summarised in Table 5. In the fruiting condition, C. barnebyana loosely resembles C. crassa, but it is readily distinguished by its relatively long (ca. 8 cm in length) infructescence axis, clavate fruiting pedicels and narrowly obovoid to ellipsoid drupelets that are weakly laterally flattened and have a dark brown indumentum. Curarea crassa, on the other hand, has a short (ca. 0.5-2 mm long) infructescence axis, terete fruiting peduncles and broadly obovoid drupelets with a dense, golden villous indumentum.
Paratypes   -30955]). Note: the original material of Abuta? pullei comprises juvenile leaves and is mounted on two sheets at U. Both appear to have Diels' handwriting and the sheet U-30956 has one, more or less, complete leaf and two fragmentary leaves. The second sheet, labelled as U-30955, is a much younger leaf and it appears that its measurements were not included in the description of the species. Hence the U-30956 specimen has been considered the holotype and annotated as such by previous authors, a designation that is followed in this study.
Common names and uses. Guyana: "teteabo (Arawak), "granny's backbone" (Creole), (Sandwith 1930 Etymology. The epithet "candicans" doubtlessly stems from the silvery indumentum on the abaxial surface of leaves, which is dense and matted in young individuals. Conservation status. Analysis of the seven collections representing seven localities resulted in an Extent of Occurrence (EOO) of 209,650 km 2 and an Area of Occupancy (AOO) of 28 km 2 . Of the seven subpopulations, the most recent collection -a sterile specimen-was made in 2004. While the species has not been collected during the past decade, which may be suggestive of population decline, it is also likely that, due to its climbing habit, the species might have been overlooked by collectors. Additionally, one of the seven individuals occurred in a nature reserve in Venezuela and the locality where one collection was made in French Guiana in the early 1990s, has also become a nature reserve. Based on these observations and the results of the assessment, C. candicans is assigned a preliminary status of "Least Concern" (LC).
Discussion. The drum-like carpophores of C. candicans (Fig. 14B) are unique in the genus. In all other species, these are elongated or subglobose. Vegetatively, C. candicans is also distinctive on account of the sparse appressed or ascending brownish trichomes on main veins and dense silvery web-like indumentum on the abaxial surface of leaves from young shoots. This indumentum later changes to a tomentellous cover that usually becomes restricted to the areoles with age. The only other species with similar web-like indumentum is C. gentryana, described below, but this lacks the brownish appressed trichomes on the main veins of C. candicans and, moreover, its web-like indumentum persists on old leaves. When leaves of C. candicans are tomentellous and this indumentum is not yet restricted to the areoles, they are indistinguishable from those of C. toxicofera and C. cuatrecasasii. However, neither of the two occurs in the Guianas and both have laxly branching rather than compact primary branches of the inflorescences of C. candicans.
In a family-wide phylogenetic analysis, C. candicans is recovered as sister to the remaining sampled species and support for this placement is high (Ortiz et al. 2016). It shares similar narrowly branched secondary axes in the staminate inflorescences with C. crassa and C. barnebyana.
The type of Abuta candicans, the basionym of Curarea candicans (Rich. ex DC.) Barneby & Krukoff, is a sterile and unnumbered Richard collection from French Guiana (deposited in the P herbarium). Although the specimen in question shows features usu-ally not associated with the remaining specimens referred to C. candicans, such as leaves with bilobulate apex, minutely undulate margins and penninerved venation, these features have sporadically been observed in a few sterile specimens of other Curarea species, although not in the same combination on the same specimen. I follow earlier workers in accepting this sterile specimen as the type of the basionym of Curarea candicans; as described by de Candolle (1818), the leaves are abaxially "glabris candicantibus", hence the specific epithet, which highlights a distinctive feature of this species. However, in order to unequivocally fix the application of the name, an epitype is being designated is this study.
On another unnumbered collection of Richard, also at P, one of the four labels has the annotation "Type coll. 2" made by Krukoff in 1968. This specimen is rather dissimilar from the type material: its leaf blades are elliptic, the apices are not bilobed and the secondary veins arise beyond the middle of the leaf, towards the apex; however, the abaxial surface is whitish. A comparison of this second specimen with other collections from the region suggests that it is conspecific with the fertile representatives of the species. However, there is no indication in de Candolle's original description of the existence of another specimen and he described the lamina as having a bilobed apex; thus this second specimen should not be considered part of the original material.
The sterile type material of Cocculus dichroa at M, image from JStore, has the leaf blades ovate, with a long-acuminate apex and the adaxial surface somewhat bullate. While ovate and long-acuminate leaves are characteristic of juvenile leaves of all species of Curarea, however a somewhat bullate adaxial surface has not been observed in C. candicans. This vegetative feature might turn out to be a distinguishing character when studied in more specimens and the extent of morphological variation C. candicans is better understood. At this time however, C. candicans is the only other species known to occur in Para and, in this study, I hesitantly follow earlier workers in the family in including C. dichroa as a synonym of C. candicans.
Distribution and ecology. Curarea crassa is known only from the coastal Atlantic Forest of Bahia in Brazil (Fig. 9), at ca. 76 m in elevation. Staminate flowering specimens were collected in May, whereas a pistillate flowering specimen was collected in February and immature fruiting specimens were collected in July and December.
Common names and uses. Brazil: "buta" (Fróes 12701/67, st). Etymology. Not explained in protologue, but likely the name is in reference to its thick exocarp (mesocarp in original description). Conservation status. The five collections representing three localities resulted in an estimated EOO of 45 km 2 and an AOO of 12 km 2 . The three localities correspond to two subpopulations and two locations, of which one is found within a protected area. However, ongoing deforestation in the largely degraded Brazilian Atlantic forest may negatively affect the already localised subpopulations of C. crassa, further reducing suitable habitat. Based on these considerations, C. crassa is here assigned a preliminary status of "Endangered" [EN B1ab(i,ii,ii,iv,v) + B2ab(i,ii,ii,iv,v)].
Discussion. Curarea crassa is vegetatively conspicuous because of its broadly elliptic or suborbicular leaves covered with a dense golden or creamy villous indumentum on the abaxial leaf blade surface. Similar indumentum may sometimes also be observed on the abaxial leaf blade surfaces of C. barnebyana and C. tecunarum; from C. barnebyana, it differs by its woolly, greyish-golden indumentum in the staminate inflorescences (vs. dark-brown). From C. tecunarum, C. crassa differs by its primary branches with few branching orders in the staminate inflorescence (vs. primary branches with several branching orders in C. tecunarum). Also, the large and broadly obovoid drupelets are unique to C. crassa.
The thick layer, from which the specific epithet is likely derived, was described as the mesocarp (Barneby 1996), however it is likely that the mesocarp in C. crassa, as in C. barnebyana, C. tecunarum and C. aff. iquitana, is mucilaginous and no longer noticeable in dried fruits, as was the case reported for C. tecunarum (Barneby and Krukoff 1971).
On morphological grounds, C. crassa is placed with the species of Group I. Micro-morphological characters shared amongst species in Group I are summarised in Table 5. Curarea crassa shares its staminate inflorescences with condensed branch orders with species of that group. They differ in the inflorescence indumentum colour, this being cream or golden in C. crassa vs. dark brown or greyish in C. barnebyana. The primary branches of the staminate inflorescences are also condensed in C. candicans, but there, trichomes are strigillose-tomentellous and not villous as in C. crassa.
In the protologue of C. crassa, Barneby (1996) described staminate flowers as having 9 sepals, but only 6 sepals were found in the only staminate plant available in this study (Thomas et al. 10900). This discrepancy is likely due to infraspecific variation as was also occasionally observed in C. candicans and in C. aff. iquitana.
Distribution and ecology. From Costa Rica throughout Panama to northwestern Colombia (Fig. 9), at elevations of 10-650(-1100) m. In wet tropical lowland to premontane forests. Staminate flowering specimens were collected in March, April, May, June and September; pistillate flowering specimens are unknown; fruiting specimens were collected in January, March and July-December.  (Forero 666, st).
Eponymy. As per Barneby and Krukoff (1971), "the specific epithet honors Dr. José Cuatrecasas, who contributed greatly to the knowledge of the Colombian Flora and was a co-collector of the holotype".
Conservation status. The calculated Extent of Occurrence (EOO) based on 26 collections representing 25 localities is 184,445 km 2 , whereas the Area of Occupancy (AOO) is estimated as 100 km 2 . Of the 23 subpopulations, 12 occurred in protected areas in Panama and Costa Rica and, although the species is not abundant where it occurs, it has a broad distribution. Hence, C. cuatrecasasii is assigned a preliminary category of "Least Concern" (LC).
Discussion. Curarea cuatrecasasii is distinguished from its congeners by the combination of slender staminate inflorescences bearing filiform primary branches, petals weakly to strongly recurved shortly above the base, connectives forming an adaxial hump at the apex of thecae when older and ovate or ovate-elliptic, 3-5-veined leaves with silvery strigillose-tomentellous indumentum on the abaxial surface. The staminate inflorescence of this species resembles those of C. gentryana and is discussed under the latter species.
In fruit, C. cuatrecasasii is indistinguishable from C. iquitana, C. tomentocarpa and C. toxicofera and they all share elongate carpophores, but they can be separated geographically: C. cuatrecasasii is restricted to the Pacific side of the Andes from North Eastern Colombia to Costa Rica and the other species are known only from the eastern side of the Andes in Colombia, Ecuador, Peru, Brazil and Bolivia.
The vessels width of C. cuatrecasasii is like that of C. tecunarum and C. barnebyana in being, on average, larger than the remaining species in group II (Table 5).

5.
Curarea gentryana R.Ortiz, sp. nov. urn:lsid:ipni.org:names:77185799-1 Fig. 19 Diagnosis. The species is distinguished from its congeners by its staminate flowers with the lateral margins of inner petals adaxially connate or connivent, also by its large broadly obovoid or ellipsoid drupelet that has a silvery tomentellous indumentum.
Note. The earlier listing of the name in Ortiz-Gentry (2000) does not constitute effective publication as per article 30.8 of the Melbourne Code (MacNeill et al. 2012), and is therefore here being validated.
Common names and uses. Ecuador. "granadilla" (Rubio & Quelal 1434, imm fr). Eponymy. The specific epithet honours the late Dr. Alwyn H. Gentry, a dedicated and extraordinary botanist and inspiring mentor, who died tragically in a plane crash in August 1993 while surveying a dry forest reserve in western Ecuador.
Conservation status. The species is known only from four collections from three localities from northwestern Ecuador. Assessment based on these collections resulted in an Extent of Occurrence (EOO) of 92.6 km 2 and an Area of Occupancy (AOO) of 12 km 2 . The three localities represent three subpopulations, each found in a small communal protected area (Reserva Etnica Awá), the surrounding area of which has been subject to increased land conversion. Therefore, it is expected that the species will in the future be negatively affected by loss of its habitat quality that may lead to reduction in the population and potentially threatening its survival. Thus, C. gentryana is assigned a preliminary status of Endangered [EN,B1ab(i,ii,iii,iv,v) + B2ab(ii,iii,iv,v)].
Discussion. The velutinous tomentellous indumentum covering the large obovoid or ellipsoid drupelets and the web-like indumentum on the abaxial surface of the broadly ovate leaves are unique to Curarea gentryana. The staminate inflorescences of C. gentryana somewhat resemble those of C. cuatrecasasii. However, C. gentryana has the spatulate inner petals with the lateral margins adaxially connate; spatulate inner petals may also be found in C. cuatrecasasii, but the lateral margins are not adaxially connate.
Note. Krukoff and Moldenke (1938) cited Chondrodendron iquitanum (Curarea iquitana in this study) as a "source of poison for darts of Indians" based on the label of the sterile Mexia 6321a. This specimen was later identified as Curarea toxicofera (Barneby and Krukoff 1971). I have not examined the referred specimen and hence I have not been able to confirm its identity.
Etymology. Presumably in reference to Iquitos, the largest city in Peruvian Amazonia and in which general area the type specimen was surely collected.
Conservation status. The assessment was based on 37 collections corresponding to 18 localities that yielded an Extent of Occurrence (EOO) of 279,551 km 2 and an Area of Occupancy (AOO) of 72 km 2 . The 18 localities represent 17 subpopulations of which four occur within protected areas (Allpahuayo-Mishana National Reserve in the Iquitos area) and three are found on private lands. Although Curarea iquitana is not abundant where it occurs, it is however broadly distributed and hence it is assigned a preliminary status of "Least Concern" (LC).
Discussion. Curarea iquitana is here resurrected from synonymy under C. toxicofera. Curarea iquitana, as defined here, encompasses a broad range of morphological variation and includes, at least provisionally, what is labelled the allpahuayo group (al) in the linear discriminant analysis. While the iquitana (iq) and allpahuayo (al) groups may represent different entities, at present it is premature to recognise them as different species, given that the morphological quantitative characters evaluated in this study partially or completely overlap between the two groups (Fig. 10B).
Additionally, features of pistillate flowers are still fragmentary or lacking, hence the extent of variation, if any, in these features, remains unknown. Collections from the foothills of central-eastern Peru in the Amazonas department from 200-800 m elevation closely resemble the type of Chondrodendron iquitanum Diels (Tessmann 4196), which was collected in the same general area -in the basin of the Marañon River, at 160 m in elevation -around which my concept of Curarea iquitana is centred. These collections tend to have conspicuously large, ovate or broadly ovate leaves with 5-7 main veins. The staminate inflorescences have a brownish to silvery strigillose indumentum and small flowers that range from 1.6 to 1.8 mm long, (mostly greyish villous and 1.3-1.8 mm long in the allpahuayo group). The adaxial horn-like protrusion of the connective at the apex of anthers, characteristic of Curarea iquitana, is variable amongst the studied collections, appearing as a horn (long and weakly incurved), as an apicaladaxial keel or as an adaxial hump, the latter more frequently being observed in the allpahuayo group; anthers are for the most part immersed in the connective. An old and fragmentary pistillate inflorescence of C. iquitana s.s. is a thyrse with brownish or silvery villous indumentum and the rugulose or muriculate drupelets have a golden-brownish velutinous-hispidulous indumentum. The only pistillate inflorescence known in the allpahuayo group has cymose primary branches proximally, distally these being reduced to single flowers giving the appearance of being racemose. The immature condition of the drupelets in most collections studied precludes their use in analyses of quantitative variation of the groups. Collections from non-flooded areas around Iquitos in eastern Peru,

Distribution and ecology.
Lowland Amazonia in Brazil, Colombia, and Peru ( Fig. 21 -fertile and only a few sterile ones are mapped), at elevations of 120-300 m (550 m in Bagua, based on Vásquez et al. 19467, a sterile collection from Bagua, in Peru, here identified tentatively) in tropical wet forest. It is also expected in the eastern lowlands of Ecuador (see below). Staminate flowering specimens were collected in February, July and September-November; the only pistillate flowering specimen was collected in November and fruiting specimens were collected in February and March.
Conservation status. The Extent of Occurrence (EOO), calculated for the 67 collections corresponding to 45 localities of C. tomentocarpa, is 984,251 km 2 and its Area of Occupancy (AOO) is 176 km 2 . Although the species is not abundant where it occurs and, of the 39 subpopulations, 10 are found within protected areas across its distribution in Bolivia, Ecuador and Peru. This suggests that its AOO may be larger than the one estimated here; therefore, C. tomentocarpa is assigned a preliminary category of "Least Concern" (LC).

Tribute
This contribution is dedicated to the memory of Alwyn H. Gentry, an extraordinary field biologist, inspiring mentor, dear friend and colleague who was always eager to share his knowledge and who died tragically in a plane crash during the summer of 1993, while surveying a dry forest in western Ecuador.

Acknowledgments
Several years in the making, this contribution is, for the most part, based on the study completed in partial fulfilment of requirements for the Masters of Science degree at the University of Missouri-St. Louis. Species boundaries in Curarea were reinterpreted in light of more data available that allowed for further analyses to be performed. I thank Peter F. Stevens and John F. Pruski for their critical reading of earlier versions of the manuscript. Gordon McPherson, Mike Grayum and Kanchi Gandhi kindly advised on nomenclatural issues. Guidance and support from Henk van der Werff, Mick Richardson, Robert Marquis, Charlotte M. Taylor and Roy Gereau, in early stages of the work is greatly appreciated. I thank Oyomoare Ozuwa-Peters, Dilys Vela Diaz and Cynthia Hong-Wa for their advice with data analyses. I am grateful to Cynthia for her assistance with the conservation status analyses. Alba L. Arbeláez made the fine illustrations for several of the species. The comments and suggestions provided by Marco Pellegrini, David Neill, Frédéric M.B. Jacques and the Phytokeys subject editor Cliff Morden, greatly contributed to improve this manuscript. Financial support was provided by the National Geographic Society Grant No. 5472-95 to Oliver Phillips and Rodolfo Vásquez, a posthumous grant to Alwyn H. Gentry from the National Tropical Botanical Garden and a grant from the Missouri Botanical Garden Alumni Fund. Mike Veith, formerly at Washington University in St. Louis, assisted with the SEM. Richard Keating at the Missouri Botanical Garden, kindly helped with the leaf anatomy and leaf clearing preparations. Gracielza Dos Santos and Regis Miller, formerly at the Forest Product Laboratory, USDA Forest Service, Madison, Wisconsin, provided assistance with wood anatomy sample preparations. I also thank the several people who in many ways contributed to my field work, especially to David Neill and Mercedes Asanza in Ecuador and Rodolfo Vásquez in Peru. Jim Solomon, curator of the herbarium at the Missouri Botanical Garden (MO) is thanked for requesting on loan the specimens on which most of this study is based. Similarly, thanks are also due to the curators of the aforementioned herbaria for loaning the Curarea specimens on which this study was based.