We first compare the descriptions of P. acuminata by Killip (1938) and Tillett (2003), P. tolimana by Harms (1894) and Killip (1938), the original description of P. metae by Bonilla Morales et al. (2016) and those of P. gleasonii by Killip (1924, 1938) and Tillett (2003). This comparison is focused on traits relative to stems, stipules, petiole, leaves, bracts, calyx, corolla, corona (outer and innermost series of filaments), operculum, limen, ovary, and fruit.

A second comparison bears on 15 of the 17 vouchers associated with the above-mentioned descriptions, focusing on the following traits: petiolar gland position, plant pubescence, bract size, and perianth parts, as observable or mentioned on the voucher label. The specimen Ducke 10528, cited in Killip’s description of P. acuminata, could not be obtained for examination. The specimen Bonilla 197 (paratype of P. metae) had not been deposited at CUVC. No voucher was associated to Tillett’s description of P. gleasonii.

We further include our own observations on living material from Ecuador (Sucumbíos), Colombia (Meta, Santander), Venezuela (Mérida, Táchira, Zulia), and Pará (Brazil), other vouchers (newly determined), as well as good and precisely geolocalized photographs from several botanists and/or passionflowers enthusiasts.

After the analyses of descriptions and reference materials showed that the four species under study could not be distinguished on a morphological basis, a geographical database was constituted, including the localizations of 78 herbarium specimens, photographs and observations listed above. These sites were mapped and a global distribution model was developed, using the MAXENT 3.4.1 software and 19 bioclimatic variable layers from the Worldclim 2.1 database at a 2’30” grid resolution (corresponding roughly to 4.4 × 4.6 km at Equator; https://www.worldclim.org/data/worldclim21.html; Fick and Hijmans 2017). MAXENT identifies potential distribution areas based on their similarity in climatic conditions compared to those at the sites where the species has already been observed, hence modeling where conditions are suitable for their development. It infers the probability distribution of maximum entropy (i.e., closest to uniform) subject to the constraint that the expected value of each environmental variable (or its transform and/or interactions) under this estimated distribution matches its empirical average (Phillips et al. 2006). A logistic threshold value equivalent to the 10^th percentile training presence was retained to separate climatically favorable areas from marginally fit areas. Thresholds of 33 and 67% training presence were used to discriminate “very good” and “excellent” climates. Furthermore, those bioclimatic covariates that most contributed to the model were extracted for the collection sites, and submitted to a Principal Component Analysis to characterize the bioclimatic spaces corresponding respectively to the previous assignations of specimens.

Three environmental parameters were extracted from specimen labels and/or observed from Google Earth (when highly precise geographical coordinates, from direct GPS measures, were available): proximity of water courses, white sands, and degree of perturbation.

The comparison between the descriptions of P. acuminata and the other three taxa is presented in Table 1. The color and pubescence of bracts, and the pubescence of sepals are not given in the table because the descriptions are incomplete regarding these criteria. These traits will be treated in the analysis of herbarium specimens. The series of filaments appear banded with white and purple in all descriptions.

As P. tolimana and P. gleasonii descriptions were based on herbarium materials, they lack information on perianth color. In the case of P. tolimana, the description underestimates the perianth size and lacks the size of the bracts and the operculum.

As shown in Table 1, the comparison of the different descriptions allows no clear distinctions that could justify the consideration of different species. Stems are rounded to angulate, which is the commonest situation in passion flowers. Stipules appear to be short, as compared to the common size in Laurifoliae, which is likely related to their observation on dry materials, and slightly longer for P. metae, but the information on stipule shape and width or with the scale given in the drawing is very inconsistent for this taxon, hampering any particular conclusion. Petioles are short too, under 2 cm, and their glands are placed close to apex, with the relative exception of Killip’s description of P. tolimana. For leaves and bracts, their variation in size, shape and distribution of marginal glands is negligible, falling well within the range of variation observed in other widespread species of series Laurifoliae (e.g. P. riparia; Rome and Coppens d’Eeckenbrugge 2019).

Similarly, taking into account the effect of desiccation on herbarium descriptions, no taxa show a distinct range of variation for sepals, petals, and corona filaments. Killip’s description of P. gleasonii is unique in mentioning no awn on sepals, which would be exceptional for the whole subgenus Passiflora; and it is contradicted by Tillet’s description. In fact, minute awns can be seen on sepals of the isotypes conserved at the Gray herbarium and the Missouri Botanical Garden (897955). The respective lengths given for the corona elements clearly indicate that the two outer series of filaments are longer than sepals and petals, an observation that was reported only in Harm’s description of P. tolimana. The description of the innermost series of filaments, intermediate in size, is not explicit in the descriptions of Harms and Killip, however these authors mention their possible fusion at base, which implies that they are not as short as, and denser than, those intermediate series with few filaments 1–2 mm long. Indeed, the analysis of the P. gleasonii holotype shows that the operculum described by Killip is a fourth series of striped filaments, more or less fused. Apart from this confusion in Killip’s description, no significant variation has been reported for the operculum and limen. The mention of a 6 mm-long limen by Bonilla Morales et al. (2016) is not supported by the longitudinal flower section drawn and photographed in their paper. In fact, the limen is reduced to a ring widening the trochlea, as in the description of Tillett (2003). Finally, the only potentially significant difference is the report of a glabrous ovary for P. metae in the last row of Table 1.

Table 2 presents the 15 reference specimens cited in the descriptions of P. acuminata, P. gleasonii, P. tolimana and P. metae that could be examined. Six of them were sterile or in buds.

The type of P. acuminata shows a very degraded flower. Thus, while we perceive the presence of two equal outer series of filiform filaments, as well as pubescent bracts with glands, it is impossible to observe the pubescence of the ovary and the presence of the inner series of filaments. From the seven specimens associated to P. acuminata in Killip’s description, only Swallen 5095, Spruce 760, and Burchell 9504 exhibit a set of traits sufficient to allow full confidence in their determination, given the common presence of P. laurifolia, also having two glands at the apex of the petiole, in the Pará state of Brazil.

Among the four specimens associated with Killip’s descriptions of P. tolimana, only the type Lehmann 6060 is representative. The three other specimens, Pennel 5754, Killip 5594, and Killip 11679, present no positive criteria to confirm their identification (and the flower buds on the latter could not be dissected). On the contrary, they present glands on the middle of the petiole, which indicates that they belong to P. ambigua, now known from the Cumbre region where these samples were collected. This confusion explains why Killip’s description diverges from the original one on the position of the petiolar glands.

The description of P. metae cites the holotype and isotypes Bonilla et al. 187 (deposited at CUVC and at FAUC) and the paratype Bonilla et al. 197 (deposited at CUVC). However, the only specimen available is present at CUVC, where it is given as the holotype and referenced as Bonilla et al. 167. Its flower is in very poor condition and the androgynophore has been broken, which prevents from verification of the glabrousness of the ovary, i.e., the only potentially distinctive trait of this taxon.

Except for the three specimens associated to Killip’s description of P. tolimana, which are likely representatives of P. ambigua, all the examined specimens present leaves with two glands at the petiole apex (or just below), bracts with glandular margins, variable pubescence on the ovary and sometimes on other floral parts as well (bracts, peduncle and calyx), a hypanthium reduced to the length of the nectary chamber, two equal outer series of filaments, longer than the perianth, a fourth series of filaments almost perpendicular to the androgynophore closing the nectary chamber, and a limen reduced to a ring at the base of the trochlea. On the specimens De La Cruz 2963 and Spruce 760, the fourth, innermost, series shows filaments more or less fused together.

In total, we have analyzed 55 herbarium specimens and 23 photographed specimens of P. acuminata, including herbarium specimens that had been previously determined as representatives of P. laurifolia (eight cases), P. nitida (five cases), P. riparia (two cases), P. ambigua (one case), P. aff. guazumifolia Jussieu (one case), and two undetermined specimens. These redeterminations were mostly based on hypanthium length (reduced to the nectary chamber in P. acuminata), the inner series of filaments, about 1 cm, absent in P. laurifolia, P. riparia and P. ambigua, or parallel to the androgynophore in P. nitida (vs. perpendicular in P. acuminata), and the position of petiolar nectaries (in apical position vs. median position in P. ambigua and P. riparia).

Plate 1 presents the high similarity observed for flower and fruit traits, even among materials collected at considerable distances. Photographs A, B, C, D, respectively from Pará (Brazil), Sucumbíos (Ecuador), Mérida (Venezuela) and Guyana, show the same floral structure, which can also be clearly observed on the photograph of the holotype of P. metae presented by Bonilla Morales et al. (2016). There are two equal outer series of filaments, longer than the perianth. The inner series are aborted while the innermost series, about 1 cm long, close the entrance of the nectary chamber. The hypanthium is short and reduced to the nectary chamber. The androgynophore is twice wider at its base than above. It can be more or less punctuated with purplish red spots, as can the stamen filaments.

As mentioned by Lehmann for P. tolimana (Killip 1938), fruits are rare, however we could gather observations from Colombia, Brazil and Venezuela (Díaz 829; photographs F, G, H), showing strikingly similar fruits, elliptic to fusiform, slightly ribbed, yellowish green, with a soft epicarp and very thin mesocarp, subhexagonal in cross sectional view. This convergence on an exceptional trait provides further support to the synonymization of the four taxa.

More variation can be observed in bracts and perianth color. Thus, we find various forms, with limited geographic consistency: long and broad bracts north of the Amazon (from the Brazilian states of Pará and Amazonas to the south of Venezuela), small and elliptical bracts in the Brazilian Amazonas, Guyana, Colombia and Ecuador, ensiform to short and wide bracts in Colombia and Venezuela. This diversity, coupled with the under-sampling and poor knowledge of this species at the time of the original descriptions, partly explains the multiplication of taxa. In their study of P. riparia and its synonyms, Rome and Coppens d’Eeckenbrugge (2019) also noted the relatively wide intraspecific variability of bract size and shape.

The color of the perianth is predominantly white. However, we found two specimens from Venezuela and Colombia with purplish petals and sepals and exhibiting all of the other characteristics of P. acuminata (Liesner 16501 and J. Restrepo’s photograph). There is also variation in the darkest color of the outer series of filaments, which can range from pink to dark purple, a variation that is also observed within P. nitida.

As observed in the specimens associated to the descriptions, there is some variation in the pubescence of the corolla elements and the ovary. While we could not verify the presence of a glabrous ovary on the type of P. metae, we could observe a weak pubescence on the ovary of specimens observed very close to the locus classicus in the Colombian department of Meta.

In contrast with several other species in the series Laurifoliae (e.g. P. riparia and P. ambigua), we have not observed flowers gathered in pseudoracemes in the analyzed materials.

As shown by the global list of examined specimens, P. acuminata has so far been documented essentially in Brazil, the country where it was first described. It appears particularly common around Belém, in northeastern Pará and northern Maranhão, but there may be a collecting bias related to the proximity of the Goeldi Museum located at Belém. It had not been collected south of the lower Amazon River (which is consistent with field observation by A. Adair, pers. com.). Elsewhere, in Colombia, Venezuela, and the Brazilian state of Acre, specimens identified as P. acuminata correspond to determinations of recent photographs in the framework of the present study, except for Liesner 13280, collected in the Venezuelan state of Zulia, close to Colombia.

A similar situation is observed for P. tolimana, described in the eastern Andes of Colombia, with observations and collections relatively concentrated around this same Tolima region and, to the south, in the Andes of Ecuador (including Holm-Nielsen1040 and Knapp 6242, as redetermined by Rome and Coppens d’Eeckenbrugge 2019), and Peru.

The name of P. gleasonii dispersed much further. Indeed, Killip described it from Guyanese specimens, however, all later collections are from Andean foothills, in Colombia and Peru, from Amazonian lowlands, in the southern Orinoco region of Venezuela and Colombia, and western Brazil (upper Amazon and Acre).

The case of P. metae is particular as the only geolocalized specimen authenticated by Bonilla Morales is the one photographed by Steven Castro very close to the locus classicus, and very close to specimens of P. acuminata with a pubescent ovary.

The analysis of the MAXENT model derived from our global dataset (all geolocalized specimens irrespective of their previous assignment) shows that the most determinant bioclimatic variables are the temperature diurnal and annual ranges, the maximum temperature of the warmest month, the mean temperature of the warmest quarter, the precipitation seasonality, with the precipitations of the driest month and those of the driest and coldest quarters. These variables were thence selected for the Principal Component Analysis. The principal plane, accounting for two thirds of the total variance, allowed a fairly good representation of the bioclimatic space of the geolocalized specimens (Fig. 1). The first and main observation is the absence of any appreciable differentiation in climatic adaptation between the four taxa under revision. Indeed, the climatic envelope occupied by specimens that had been assigned to P. acuminata encompasses that of all the other specimens. Noticeably, the small group of ten specimens showing a score above 1 on the first axis (extreme right of the plane) were all collected close to Belém (seven cases) and Santarém (three cases).

Figure 1. 

Principal plane of the Principal Component Analysis on the eight bioclimatic variables that most contributed to the P. acuminata distribution model built from 78 geolocalized specimens. Factors 1 and 2 account for 39 and 29% of total variance respectively. The symbols indicate the initial assignation of specimens.

Fig. 2 presents the geographic space corresponding to this climatic envelope, i.e., the potential distribution of P. acuminata across South America and Central America. This wide distribution appears to be split between two major regions: to the east, from the Guianas to the lower Amazon River, up to the Rio Negro and southeastern Venezuela, favorable climates are mostly found at low elevations, well under 200 m (except for de Granville 1340, Wurdack 4954, and Watson’s observation, from French Guiana, Guyana and Venezuela, at elevations around 500 m); to the west, in the upper Orinoco region and along the Andean foothills and valleys, from western Venezuela to Peru, P. acuminata has mostly been reported at elevations comprised between 200 and 2200 m. However, it is also present below 200 m, in Andean valleys and low Orinoquian and Amazonian regions of Venezuela and Colombia. These major, eastern and western, regions are linked by extensive lowland areas where climates are mostly classified as marginal to the species and its presence appears sporadic. However, some of these areas are of difficult access and have been poorly explored (e.g. basins of the Javari River in western Brazil and the upper Amazon River in Peru, except for the vicinity of Iquitos).

Figure 2. 

Potential distribution model for P. acuminata; derived from the 78 geolocalized specimens of the present study (pink circles) and MAXENT bioclimatic modeling. Background colors indicate climate suitability: marginal (dark green), favorable (light green; above 10% training omission threshold), very favorable (yellow; above 33% omission threshold); excellent (orange; above 67% omission threshold).

On the other hand, several areas presenting favorable to excellent climates according to the MAXENT model, are not supported by effective observations of the species. In the eastern region, the absence of the species south of the lower Amazon, as observed by A. Adair (pers. com.) is not precisely explained by the model, as the climate appears favorable there. More strikingly, while several highly favorable areas are predicted along the shores of Suriname and French Guiana, the presence of P. acuminata has never been confirmed in northern French Guiana despite extensive collecting activity (Rome and Coppens d’Eeckenbrugge 2016, 2020). In the western region, the distribution of specimens mostly correlates with the MAXENT model results for areas along the northern Andes (Colombia and Ecuador). The relative rarity of observations in the climatically favorable upper Orinoco can be easily explained by the lack of botanical exploration in this part of Colombia (Ocampo Pérez et al. 2010). But such an explanation does not hold for the absence of observations in the Colombian Western Cordillera as well as in the Colombian, Ecuadorian and Panamanian Chocó, and further north into Central America, all contiguous areas whose climates appear highly favorable for P. acuminata. Indeed, P. nitida and P. ambigua have been repeatedly collected there, but no P. acuminata.

The MAXENT bioclimatic model also points to very favorable areas along the Central Andes, southward to Bolivia and southern Brazil, but they are not met by any observations beyond Peru.

The lack of observations of P. acuminata in several extensive areas that combine highly favorable climates and a relatively good level of botanical prospection suggests that the species distribution is significantly constrained by non-climatic factors, justifying further ecological analyses. Before such studies are carried out, our MAXENT distribution model can be considered as a guide for collecting more data.

The ecological information associated with herbarium specimens and the precise geolocalization associated with iNaturalist photograph databases allow inquiring on habitat parameters at a smaller scale, either from label information or from available geographic information (e.g. aerial views).

As for P. riparia (Rome and Coppens d’Eeckenbrugge 2019) and P. nitida, P. acuminata appears associated to riparian habitats. Out of our 78 records, 69 provide information on humidity, with 48 positive cases (close proximity to streams, lakes, ponds, or temporarily flooded areas). Another frequent element of habitat descriptions is the mention of sandy soils, mostly white sand savanna or “campina” habitat (31 cases), particularly in Amazonian and Orinoquian lowlands. Although not signaled in the specimen label, these white sands are very common in the Guyanese region of the locus classicus of P. gleasonii. In contrast, only one case of clay soil and one of rocky soil are signaled. A rocky soil is also likely for Idrobo 868, collected on the Cordillera Macarena slopes in Colombia. Finally, 51 records point to disturbed vegetations, in contrast with only eleven simply mentioning a forest habitat. The latter is not surprising for a liana, benefitting from clearings in the forest, and easier to observe and collect along roadsides and in open habitats.

This first ecogeographical approach has provided no evidence of differential climatic adaptation among specimens previously assigned to P. acuminata, P. tolimana, P. gleasonii or P. metae. The available herbarium information points to a common association with riparian habitats and/or white sand soils, particularly in lowland areas. Further studies, involving soil information, and more collections, are needed to identify non-climatic factors limiting the presence of P. acuminata in a few climatically favorable areas.

All previous observations confirm that it is not possible to differentiate several species among the four taxa under study on morphological and/or ecological bases. Thus, it is necessary not only to resurrect P. acuminata but also to place the other species as synonyms of the latter. Furthermore, as the type of this species is not sufficiently explanatory, we propose an epitype, Swallen 5095 (Plate 2), presenting all the elements necessary for the morphological understanding of the species, following the suggestion of the international code of nomenclature (Turland et al. 2018).

Plate 2. 

Specimen Swallen 5095 [US01194023], designated as epitype of Passiflora acuminata (copyright US National Herbarium).