We describe a new species of Leucostethus from Gorgona Island, a small (13 km²) island located 35 km from the Pacific coast of southern Colombia. The new species most resembles L. argyrogaster and L. fugax from western Amazonia at 340–870 m elev. in Peru and Ecuador, with which it shares pale ventral coloration and orange suffusion of the axilla, groin and concealed surfaces of the hind limb, but it is most closely related to L. bilsa from ca. 340 km SW in the southern Chocó at 420–515 m elev., northwestern Ecuador. We report miniscule white spots on the posteroventral surface of the thighs of the new species and, on the basis of our preliminary assessment of their taxonomic distribution, hypothesize that their presence is a synapomorphy of Dendrobatoidea with subsequent losses in a few groups. Given the apparent restriction of the new species to Gorgona Island, it is Vulnerable according to IUCN Red List criteria. In addition to naming the new species, we also propose the following new combinations: L. alacris (Rivero and Granados-Díaz, 1990) comb. nov., L. dysprosium (Rivero and Serna, 2000) comb. nov., and L. yaguara (Rivero and Serna, 1991) comb. nov.

Amphibia, Chocó, Colostethinae, Colostethus, Dendrobatoidea, Silverstoneia, taxonomy

Gorgona Island, located approximately 35 km from the Pacific coast of southern Colombia (Giraldo 2012), is a fragment of the Caribbean Large Igneous Province that formed in the Late Cretaceous and accreted to South America in the Eocene (Serrano et al. 2011). Although Gorgona is currently an island, the depths between it and the mainland are only ca. 60–120 m, so contiguous terrestrial habitat undoubtedly connected Gorgona and the mainland during the last glacial maximum (20–18 ka) and for a considerable time afterward (Fleming et al. 1998; Montealegre-Z et al. 2010). Indeed, Gorgona’s coral reef is estimated to date to only 2–3 ka (Glynn et al. 1982), placing a lower bound on Gorgona’s complete isolation from the mainland. Given its size (ca. 13 km²; Giraldo 2012), proximity to the mainland, and recent isolation, it is unsurprising that Gorgona’s biota is predominated by species that are widespread in the Chocó (Giraldo and Valencia 2012). Nevertheless, several putative endemics have been discovered on Gorgona, including species of Chelicerata (Lourenço and Flores 1989), Formicidae (Fernández and Guerrero 2008), Orthoptera (Montealegre-Z and Postiles 2010; Montealegre-Z et al. 2011; Baena-Bejarano and Heads 2015), and Psocodea (García Aldrete et al. 2011; Sarria-S et al. 2014; Manchola et al. 2014), although it is not uncommon for subsequent research to uncover species named from Gorgona on the mainland (e.g. Mendivil Nieto et al. 2017). Among vertebrates, the putative endemics include the catfish Trichomycterus gorgona Fernández & Schaefer, 2005, the anoles Anolis gorgonae Barbour, 1909 and A. medemi Ayala & Williams, 1988 (Phillips et al. 2019), the snake Atractus medusa Passos et al., 2009, and an undescribed species of dendrobatid frog reported as “Colostethus sp. Gorgona” by Grant et al. (2017). The objective of the present paper is to formally name and describe this new species.

We conducted fieldwork on Gorgona Island (02°58'N, 78°10'W) from 23 to 28 May 2016. See Giraldo (2012) for a general description of the island and Vásquez-Vélez (2014) for a description of the vegetation. We euthanized specimens using 20% benzocaine (Orajel; Church & Dwight Co., Inc., Ewing, New Jersey, USA) applied to the skin or by pithing (McDiarmid 1994). Following preservation of one hind limb in 96–100% ethanol for DNA analysis and/or preservation of the skin in 100% methanol, specimens were fixed in formalin and subsequently transferred to 70% ethanol and deposited in the amphibian collection of the Colección de Prácticas Zoológicas, Universidad del Valle, Cali, Colombia (CPZ-UV). We also include in the type series one specimen from the National Museum of Natural History, Smithsonian Institution, Washington DC, USA (USNM).

We captured advertisement calls using a Tascam DR-40 linear pulse-code modulation (PCM) recorder and internal microphone at a sampling rate of 44.1 kHz with 16 bit encoding at 26.2°C ambient temperature. We also captured audio and video of the holotype vocalizing using a Canon EOS Rebel T3 and internal microphone and analyzed the audio (linear PCM recorder, 48 kHz sampling rate, 16 bit encoding). The calling males were observed directly and recorded at ca. 1 m distance. Acoustic analyses employed standard definitions and terminology for spectral and temporal variables (Köhler et al. 2017). We used Raven Pro 1.6 sound analysis software (Cornell Laboratory of Ornithology, NY) to score call duration (s), notes per call, note duration (ms), internote interval (ms), note repetition rate (notes/minute), number of pulses, fundamental and dominant frequencies (kHz), frequency modulation both within and among notes (kHz), and number and frequency of harmonics (kHz). We scored spectral parameters from spectrograms and power spectra (Hann window, 90% overlap, 512 point Fast Fourier Transformation resolution) and temporal parameters from expanded waveforms. The first 10–15 notes of each call were variable and differed from subsequent notes, all of which were emitted at higher amplitudes and varied little temporally and spectrally (i.e. they did not taper off toward the end of the call), so we report values for both entire calls and the first 10 and last 20 notes; as the latter values are most representative of the advertisement call, we used them for interspecific comparisons. We report statistical summaries as x̄ ± SE.

Given that individual units of sound were consistently (i.e. both in the recorded calls and the many calls that were heard, but not recorded) emitted in bouts of < 30 s duration, we classified the individual units of sound as notes grouped into calls. This treatment is consistent with the interpretation by Vigle et al. (2020) of Leucostethus bilsa Vigle et al. 2020, which emitted units of sound in bouts of < 60 s. In contrast, other species of trans-Andean Leucostethus emit sounds at regular intervals in bouts varying from less than one minute to several minutes in duration, forming continuous series of variable length (a few to more than 300; Grant and Castro 1998); as such, in those species, Grant and Castro (1998) and Marin et al. (2018) considered calls to comprise a single note. For comparison, we treat the values reported by Grant and Castro (1998) and Marin et al. (2018) for “calls” as equivalent to the values reported by Vigle et al. (2020) and us as “notes.”

Information on phylogenetic relationships was taken from Grant et al. (2017), Marin et al. (2018), and Vigle et al. (2020). Although Grant et al. (2017) included the undescribed species from Gorgona Island in their Colostethus fraterdanieli group, Marin et al. (2018) subsequently transferred that group to Leucostethus, and we follow their taxonomy here. Character definitions follow Grant et al. (2006, 2017). For hand morphology, we follow Fabrezi and Alberch (1996) in considering finger I of other tetrapods to be absent in Anura and number fingers accordingly. As such, we follow Grant et al. (2017) in referring to the swollen third finger of earlier literature (e.g. Grant et al. 2006) as swollen finger IV. The webbing formulation is that of Savage and Heyer (1967), whereby webbing is quantified by the number and proportion of free phalanges (see also Myers and Duellman 1982; Savage and Heyer 1997). Jaw muscle terminology follows Haas (2001). We took the following measurements with digital calipers to 0.1 mm:

SVL snout–vent length

FAL forearm length from proximal edge of palmar tubercle to outer edge of flexed elbow

HL hand length from proximal edge of palmar tubercle to tip of finger IV

TL tibia length from outer edges of flexed knee to heel

FL foot length from proximal edge of outer metatarsal tubercle to tip of toe IV

HW head width between angle of jaws

HLD head length diagonally from corner of mouth to tip of snout

EL eye length from posterior to anterior corner

END eye-naris distance from anterior corner of eye to center of naris

IND internarial distance between centers of nares

SL snout length from anterior corner of eye to tip of snout

IOD interorbital distance

TD tympanum diameter

Unless otherwise noted, measurements and proportions are given only for adults and are summarized as x̄ ± SE. Males with vocal slits on both sides of the mouth were scored as adults, those with only one vocal slit as sub-adults, and those lacking slits on both sides as juveniles. Females with expanded, convoluted oviducts and enlarged oocytes were considered to be adults, those with only weakly expanded, non- or weakly convoluted oviducts and poorly differentiated oocytes to be sub-adults, and those with small, undifferentiated oocytes and unexpanded, straight oviducts to be juveniles. Color in life is based on field notes and digital photographs.

We compare the new species to other species of Leucostethus sensu Marin et al. (2018) [viz. L. argyrogaster (Morales & Schulte, 1993); L. bilsa, L. brachistriatus (Rivero & Serna, 1986), L. fraterdanieli (Silverstone, 1971), L. fugax (Morales & Schulte, 1993), L. jota Marin et al., 2018, and L. ramirezi (Rivero & Serna, “1995” 2000)], to which we add L. alacris (Rivero & Granados-Díaz, 1990) comb. nov., L. dysprosium (Rivero & Serna, “1995” 2000) comb. nov., and L. yaguara (Rivero & Serna, 1991) comb. nov. Although the precise phylogenetic relationships of these three species have not yet been tested in quantitative phylogenetic analyses, Grant and Castro (1998) noted the resemblance of L. alacris and L. yaguara to L. fraterdanieli and Grant et al. (2017) considered the possibility that their Cordillera Occidental clade of L. fraterdanieli might correspond to one or both of those names. Similarly, L. dysprosium shares a complete, continuous, pale oblique lateral stripe with other species of Leucostethus (see Discussion) and appears to differ from other trans-Andean species of Leucostethus except L. alacris only in possessing more extensive toe webbing (see Rivero and Serna “1995” 2000: 49, fig. 2C). We also compare the new species to similar colostethines from the Chocó.

Although phylogenetic analyses strongly place Leucostethus siapida sp. nov. as part of the trans-Andean Leucostethus fraterdanieli group (the Colostethus fraterdanieli group of Grant et al. 2017; i.e. all species, except L. argyrogaster and L. fugax), its precise phylogenetic relationships are unclear. Grant et al. (2017) found it to be the lowland sister of the remainder of the otherwise montane trans-Andean Leucostethus, whose species occur from ca. 1000–2700 m elev. in the Andes and Cauca River valley. In contrast, Marin et al. (2018) and Vigle et al. (2020) found it to be nested inside the montane clade. It is unclear if this contradiction is due to differences in character evidence (e.g. inclusion of phenomic evidence by Grant et al. 2017), taxon sampling (e.g. inclusion of additional Leucostethus species by Marin et al. 2018 and Vigle et al. 2020), or analytical methods (DNA sequence alignment and optimality criterion). Nevertheless, there seems little reason to doubt that L. siapida sp. nov. is sister to L. bilsa, which was recently described from ca. 340 km SW in the Reserva Biológica Bilsa at 420–515 m elev., northwestern Ecuador (Vigle et al. 2020).

Vigle et al. (2020) estimated a divergence time of approximately 3 Ma for Leucostethus bilsa and L. siapida sp. nov. Given that Gorgona has only been isolated from the mainland for at most 18–20 ka, and probably considerably less (Fleming et al. 1998; Montealegre-Z et al. 2010), isolation on the island appears not to have been causally related to the speciation event, suggesting that this is a relictual population of a formerly widespread species. Moreover, given the recency of the terrestrial connection between Gorgona and the mainland, we are optimistic that mainland populations will be discovered. Indeed, given its close resemblance to species of Silverstoneia, it is possible that specimens misidentified as S. nubicola already reside in natural history collections.

Among the species of Leucostethus, L. siapida sp. nov. most resembles L. argyrogaster and L. fugax from western Amazonia at 340–870 m elev. in Ecuador and Peru. In addition to lacking dark ventral coloration, these three species share orange flash marks (reported for L. argyrogaster by Morales and Schulte 1993; shown for L. fugax by Grant et al. 2017: fig. 33B), and L. siapida sp. nov. and L. argyrogaster further share the lack of conspicuous swelling of finger IV in adult males. Despite their resemblance, phylogenetic analyses consistently place the cis-Andean species as sister taxa (for which the orange flash coloration is an apparent synapomorphy) that are far removed from L. siapida sp. nov.

Scoring the occurrence of swelling on finger IV of adult males is unproblematic in species that exhibit conspicuous swelling (developmental variation and preservation artifacts notwithstanding), as do most species of Leucostethus; however, distinguishing between absence and weak swelling is notoriously difficult (Grant et al. 2017). For example, although Grant et al. (2017) scored finger IV as unswollen in L. siapida sp. nov., subsequent detailed examination of external morphology indicated that it is weakly swollen, and histological analysis revealed the presence of a specialized mucous gland found exclusively in the swollen finger IV (I.R.S. Cavalcanti and TG, unpublished data). As such, extreme caution is required when employing absence versus weak swelling to diagnose similar species. For this reason, we did not distinguish between the reported absence of swelling in L. argyrogaster and L. bilsa and the weak swelling in L. siapida sp. nov.

A curious characteristic shared by Leucostethus siapida sp. nov. and its sister species L. bilsa is the lateral variation in testis melanization. Although ontogenetic variation in testis pigmentation is common, individual variation among adults is rare, and, prior to L. bilsa (Vigle et al. 2020), unilateral melanization in dendrobatoids had only been documented for Colostethus panamansis (Dunn, 1933) by Grant (2004). Golberg et al. (2020) found that testicular melanization and germ cell differentiation proceed in parallel in four anuran species, but the significance of intra-individual variation is unknown.

To our knowledge, the miniscule, white spots on the posteroventral surface of the thighs have not been reported previously, possibly because they are only evident in life, but they are widespread in Dendrobatoidea. Although an exhaustive review was beyond the scope of the current study, they also occur in additional species of Leucostethus (e.g. L. bilsa; Vigle et al. 2020: figs 2–5) and a least some species of the aromobatid genera Allobates, Anomaloglossus, Aromobates, Mannophryne, and Rheobates and the dendrobatid genera Colostethus, “Colostethus” ruthveni, Epipedobates, Hyloxalus, and Silverstoneia, and we are unaware of their occurrence in Ectopoglossus, Paruwrobates, Phyllobates, Dendrobatini, or potentially close relatives of Dendrobatoidea, including bufonids, hylodids, and Thoropa. Further study is required to determine if the presence of these spots is a synapomorphy of Dendrobatoidea with a limited number of informative losses, as available data suggest, or if their evolutionary history is more complicated.

In addition to describing the new taxon Leucostethus siapida sp. nov., we transferred L. alacris, L. dysprosium, and L. yaguara from Colostethus to Leucostethus. Vigle et al. (2020: 368) also considered the generic placement of C. alacris and C. yaguara but concluded that “placing them in the genus Leucostethus requires more evidence or data (e.g. molecular and/or morphological) from type series or topotypical samples.” Nevertheless, it should not be overlooked that the persistence of these species in Colostethus for the past 15 years was not due to empirical evidence, but rather (1) their placement in Colostethus prior to its partitioning into multiple genera by Grant et al. (2006, 2017) and Marin et al. (2018), and (2) their overall resemblance to the species now referred to Leucostethus. Although referral of these species to Leucostethus is not based on either quantitative phylogenetic analysis or clear synapomorphies and is, therefore, more predictive than explanatory, it is not entirely unsubstantiated by morphology. All three species share with other species of Leucostethus the presence of a complete (i.e. groin to eye), continuous, pale oblique lateral stripe (incomplete in some specimens of L. bilsa). In contrast, the oblique lateral stripe is either absent (C. thorntoni, C. ucumari, and “C.” ruthveni; for phylogenetic relationships of the last species, see Grant et al. 2017), continuous but incomplete (extending from groin midway along flank in C. furviventris, C. imbricolus, C. inguinalis, C. latinasus, C. panamansis, and C. pratti), or complete but formed by a series of spots (C. agilis, C. lynchi, and C. mertensi).

Despite our optimism that additional populations of Leucostethus siapida sp. nov. will be found on the mainland, based on current knowledge, L. siapida is endemic to Gorgona Island. Although Gorgona Island is protected as Parque Nacional Natural (PNN) Gorgona, the fact that the only known population is confined to an island of only ca. 13 km² is sufficient to categorize this species as Vulnerable according to the IUCN Red List criteria (criterion D2; IUCN 2012). Nevertheless, we emphasize that this categorization should not prevent studies of this species from being undertaken, as many of the most basic aspects of its morphology (e.g. larval morphology, ontogenetic variation) and behavioral ecology (e.g. courtship and amplexus, oviposition site, parental care) are unknown, and studies detailing its fine-scale distribution and abundance, age-sex structure, and the frequency of morphological abnormalities, such as the testicular anomaly reported herein, will be crucial for managers of PNN Gorgona to monitor and protect the species. Ultimately, the occurrence of this endemic vertebrate adds to an already lengthy list of reasons to consider Gorgona to be a crucial locality for biodiversity conservation (Giraldo et al. 2014).

We are grateful to David Andrés Velásquez-Trujillo and Andrés Felipe Gómez Fernández for assistance during fieldwork and María Ximena Zorrilla (Director of PNN Gorgona), Luis Fernando Payan, Hector Chirimia Gonzalez, and all the PNN Gorgona functionaries for both logistic support and sharing their extensive knowledge of Gorgona and its fauna and flora. We also thank Alan Giraldo and the members of the Grupo de Investigación en Ecología Animal de la Universidad del Valle for support. Adolfo Amézquita first drew our attention to the existence of a Silverstoneia-like species on Gorgona, and Marvin Anganoy, Isabela Cavalcanti, Julián Faivovich, Rachel Monetesinos, Marco Rada, and Geven Rodríguez shared their knowledge and insights on anuran morphology and systematics. Ralph Saporito tested the new species for the presence of skin alkaloids. The article benefitted from critical reviews by Luis Coloma and an anonymous reviewer. Hector Chirimia Gonzalez was extremely helpful in choosing the specific epithet for the new species. Kevin de Queiroz, Steve Gotte, Roy McDiarmid, and Addison Wynn authorized and enabled examination of specimens at USNM. Permission to collect in PNN Gorgona was granted by Parques Nacionales Naturales de Colombia (No. 010 de 2016). Funding was provided by the São Paulo Research Foundation (FAPESP Procs. 2012/10000-5, 2018/15425-0), the Brazilian National Council for Scientific and Technological Development (CNPq Proc. 306823/2017-9), and the Fundación Univalle.