Pempheris gasparinii sp. n. is described from five specimens, 59.1–68.0 mm in standard length. It is only known to occur in the shallow reefs of Trindade Island, 1200 km east of the Brazilian coast, in the southwestern Atlantic. Pempheris gasparinii is the third recognized species of Pempheris in the Atlantic Ocean. This new species is morphologically similar to its close relative, P. poeyi, differing by the number of lateral-line scales (51–54 in P. gasparinii vs. 47–49 in P. poeyi), scales below lateral line (10–11 vs. 9), circumpeduncular scales (11–12 vs. 13), head and caudal peduncle lengths (2.7–3.3 vs 3.5–4.0 in head length). Moreover, Pempheris gasparinii shows a 4% genetic divergence from P. poeyi at the cytochrome oxidase I locus (COI), consistent with a lineage split at the beginning of the Pleistocene. This new species represents the 12^th endemic fish species from Trindade Island.

Endemism, COI, Vitória-Trindade Chain, oceanic island, Brazil, reef fish

The genus Pempheris Cuvier contains 69 valid species (Randall and Victor 2015), with two species known from the Atlantic Ocean: Pempheris poeyi Bean 1885 and P. schomburgkii Müller & Troschel, 1848. Pempheris species generally display similar body shapes, however, these two species strongly differ by coloration (P. poeyi is silvery while P. schomburgkii is yellowish) and number of anal-fin rays (22–24 in P. poeyi vs. 31–34 in P. schomburgkii). Pempheris schomburgkii is widely distributed in the western Atlantic, from Bermuda (32°N) to Santa Catarina, Brazil (29°S). In contrast, a disjunct distribution has been previously assigned to Pempheris poeyi, with populations occurring in the Greater Caribbean and in Trindade Island, Brazil (Pinheiro et al. 2015), at least 5,000 km apart.

Pempheris poeyi was described from Cuba in 1885 (Bean 1885), and since then few specimens have been collected from localities such as the Bahamas, Grenada, Trinidad and Tobago, and Venezuela. Specimens identified as P. poeyi were found at Trindade Island in the early 20^th century (Miranda Ribeiro 1919), and rediscovered 90 years later by Pinheiro et al. (2009). There is overlap in counts of dorsal- and anal-fin rays between the Greater Caribbean and Trindade populations, a common observation among Pempheris species (Mooi and Randall 2014), which likely explains why the new Trindade species described herein was misidentified in two recent checklists (Simon et al. 2013, Pinheiro et al. 2015, Randall and Victor 2015). Recent publications revealed that there is an abundance of undescribed cryptic species in Pempheris (Koeda et al. 2013, Mooi and Randall 2014, Randall and Bineesh 2014, Randall and Victor 2014, Randall et al. 2014). In this study, the use of genetic tools allowed us to revisit the Trindade specimens and update their taxonomic status. Thus, herein we describe a third Atlantic Ocean Pempheris species, so far only known from Trindade Island, at the end of the Vitória-Trindade Chain, 1200 km off the Brazilian coast.

All specimens were collected using hand nets. Counts were performed using a microscope, and morphological characters were measured to the nearest 0.1 mm following Mooi and Randall (2014). Morphometric and meristic data for the type series are presented in Table 1. In the description, meristic values for the holotype are provided first, followed by the range of counts of paratypes in parentheses. Type specimens were deposited in the fish collections of the Universidade Federal do Espírito Santo (CIUFES), California Academy of Sciences(CAS), and Universidade Estadual de Campinas(ZUEC).

Mitochondrial Cytochrome c oxidase subunit I (COI) DNA was analyzed for the new species. DNA extraction and PCR amplification of the COI were performed following Weigt et al. (2012). DNA sequences were then compared to those of the other two Atlantic Pempheris (P. poeyi and P. schomburgkii) downloaded from GenBank. Phylogenetic relationships were assessed by Maximum Likelihood (ML, GARLI software, Zwickl 2006), and Neighbor-Joining (MP, PAUP* software, Swofford 2003) methods. For Maximum Likelihood topologies, we conducted 10 independent runs in GARLI, using default settings and the automated stopping criterion, terminating the search when the ln score remained constant for 20,000 consecutive generations. Statistical confidence in nodes was evaluated using 1,000 non-parametric bootstrap replicates (Felsenstein 1985) for the NJ method, and 100 replicates for ML in GARLI, using the automated stopping criterion set at 10,000 generations. Parapriacanthus is the only other genus of Pempheridae and was used as an outgroup as in Azzurro et al. (2015) (sequence from GenBank, accession number JF494087). Pempheris sequences are from Azzurro et al. (2015), except for P. poeyi, which is from this study. GenBank access numbers are KJ609406 (Pempheris gasparinii sp. n.), KT634057 (Pempheris poeyi), and KJ609388.1 and KJ609389.1 (Pempheris schomburgkii).

As stated by Mooi and Randall (2014), assignments of original or new names to Pempheris species is remarkably difficult because: 1) the poor condition and descriptions of type material; and 2) traditional meristics (e.g. fin ray counts) are not very informative at the species level due either to too little or too much character variability. For instance, because of its similarity to P. poeyi, P. gasparinii is misidentified in the most recent biodiversity checklists of Trindade Island (Pinheiro et al. 2009, 2015, Simon et al. 2013).

A phylogenetic tree of Pempheris of the Atlantic based on COI placed P. poeyi and P. gasparinii as sister species (Figure 4). The two species show a 4% sequence divergence at this locus, which is consistent with a speciation process initiated in the beginning of the Pleistocene. With only one sequence from each of the two species and thus no information on intraspecific genetic divergences, we cannot definitively use the genetic data to separate the two species. However, assuming that intraspecific divergence is low, as it is in P. schomburgkii from Panama and Curacao (Fig. 4), the combined genetic and morphological data easily distinguish the two species. The disjunct distribution associated with the genetic divergence of the species suggests that P. gasparinii might be a relict species. Similarly to other recently described species from Trindade Island (Pinheiro et al. 2010, Rocha et al. 2010), P. gasparinii seems to have been preserved in the Vitória-Trindade Chain while the Pleistocene sea-level changes extirpated the lineages along the adjacent Brazilian coast. This species represents the 12^th endemic fish from Trindade Island. Thus, following the diversity reported by Pinheiro et al. (2015) and criteria of reef fishes established by Floeter et al. (2008), the proportion of endemics found in the Vitória-Trindade Chain (191 species) and Trindade Island (137 species) today is 6% and 9% respectively.

We thank the Brazilian Navy (Marinha do Brasil–1° Distrito Naval), TAMAR/ICMBio Project, and Substation Curaçao for logistical support; the California Academy of Sciences for financial support; Flamar, Windive, Voz da Natureza and Fauna for diving support; R. Santos, T. Simon and R. Andrade for collaboration on dives and collecting specimens; D. Catania, J.C. Joyeux, I. Sazima, and J. Williams for helping access CAS, CIUFES, ZUEC, and USNM collections, respectively. We are especially grateful to T. Simon and A. Barber for photographs. CNPq provided a doctoral fellowship for HTP (Ciência sem Fronteiras; GDE 202475/ 2011-5). This paper is part of the projects “Vitória—Trindade Chain: testing the theory for stepping-stone connectivity in reef fishes” (CNPq grant 470725/2009-5) and “Reef fishes of the insular complex Trindade—Martin Vaz, Vitória, ES: status, recruitment and evolution of assemblages” (CNPq grant 557043/2009-3).