Recent dispersal and diversification within the clingfish genus Acyrtus (Actinopterygii: Gobiesocidae), with the description of a new western Atlantic species

The genus Acyrtus (Gobiesocidae) is represented by four valid species distributed in the western Atlantic, and a recently described fifth species from the eastern Pacific. Here, we describe a new species endemic to Trindade Island, Brazil, and provide the first phylogenetic inference for the genus including all representatives. The new species can be distinguished from all its congeners by meristic and morphometric characters, as well as genetic differences. It presents low genetic diversity and, contrarily to other Trindade Island endemic fishes, shows no evidence of recent population growth. Our phylogeny reveals cryptic species and the paraphyletic nature of Acyrtus , which included Arcos nudus (western Atlantic) in a clade that separated from Arcos erythrops (tropical eastern Pacific) around 20 Mya. The three species found in the Brazilian Province, including one that remains undescribed, form a monophyletic clade which colonized the western South Atlantic around 2.6 Mya. Our study suggests that Arcos nudus should be placed in Acyrtus , and that the relationships among the closely-related Gobiesocidae genera Acyrtus (mostly from the Atlantic Ocean) and Arcos (from the Pacific Ocean) need further investigation.

In this context, the use of molecular approaches combined with traditional taxonomy has provided a better understanding of evolutionary relationships (e.g., Henriques et al., 2002;Fricke et al., 2017;Wagner et al., 2019) and identification of cryptic lineages (e.g., Henriques et al., 2002;Craig, Randall, 2008;Conway et al., 2014;Wagner et al., 2019;Torres-Hernández et al., 2020). Though recent molecular studies have provided important progress in the knowledge of this family (Conway et al., 2017(Conway et al., , 2020Fricke et al., 2017), the evolutionary history of most species and the full extent of Gobiesocidae diversity remain unknown. Phylogenetic work including the New World Gobiesocinae suggested a recent diversification, the presence of several cryptic species and paraphyletic genera (Conway et al., 2014(Conway et al., , 2017(Conway et al., , 2020Tavera et al., 2021), all of which we will explore in this study within the genus Acyrtus Schultz, 1944. Acyrtus is represented by five valid species, Acyrtus artius Briggs, 1955, Acyrtus lanthanum Conway, Baldwin & White, 2014, Acyrtus pauciradiatus Sampaio, Anchieta, 3/22 ni.bio.br | scielo.br/ni Thais L. Quintão, João Luiz Gasparini, Jean-Christophe Joyeux, Luiz A. Rocha andHudson T. Pinheiro Nunes &Mendes, 2004 andAcyrtus rubiginosus (Poey, 1868) are restricted to the western Atlantic, while Acyrtus arturo Tavera, Rojas-Vélez & Londoño-Cruz, 2021, was recently described from the eastern Pacific. These are small-bodied fishes (maximum size of less than 30 mm) with cryptobenthic habits, some presenting large secretory cells similar to those present in the venom glands of other teleost fishes (Conway et al., 2014). In the Atlantic, three species are found in the Caribbean and the fourth (A. pauciradiatus) is endemic to Fernando de Noronha Archipelago and Rocas Atoll, in Brazil. However, two other undescribed species have been found in restricted locations of the southwestern Atlantic, one on the Brazilian continental shelf and the other in the oceanic Trindade Island (Pinheiro et al., 2017).
Morphologically, Acyrtus is closely related to Rimicola Jordan & Evermann, 1896, a genus restricted to the eastern Pacific. However, a recent molecular phylogenetic study exploring Gobiesocidae has shown a close relationship between Acyrtus and Arcos Schultz, 1944 and, under low bootstrap values, has placed Arcos nudus (Linnaeus, 1758), from the western Atlantic, within the Acyrtus clade (Conway et al., 2014). This result agrees with the distribution of Arcos and Acyrtus genera since Arcos is mainly restricted to the eastern Pacific and Acyrtus to the western Atlantic. However, the recent description of A. arturo (Tavera et al., 2021) from Malpelo Island opens discussion about the evolutionary history of Acyrtus and Arcos. In addition, both genera share morphological characteristics that obscure their evolutionary proximity (e.g., Tavera et al., 2021). Species of both genera have previously been described as Gobiesox (Schultz, 1944;Conway et al., 2017), indicating that morphological similarities often lead to misidentifications (Conway et al., 2017). In this study, we provide the first phylogenetic inference for the genus Acyrtus containing all known species and including the undescribed species from Trindade Island as new, and an undescribed species from Brazilian coast. In addition, we analyzed the evolutionary history of the genus, and the phylogeographic and demographic history of the Brazilian species complex.

MATERIAL AND METHODS
Morphological analysis. The eight type specimens were collected with hand nets during a field expedition to Trindade Island in June 2009. Specimens were fixed in formalin 10% for 24 h, and then transferred to alcohol 70%. Counts were performed with a stereo microscope (Leica S9i, Amplification 6.1-55x) and X-rays obtained with a radiography system Faxitron LX60. Morphological characters were measured to the nearest 0.01 mm using an ocular micrometer mounted on a dissecting scope. Measurements and counts followed the methods presented in Briggs (1955), with the addition of predorsal and preanal lengths, which are the shortest distances between the tip of the upper lip and the dorsal and anal-fin origin, respectively (see Conway et al., 2014). Vertebral counts are presented as precaudal + caudal. The anterior-most vertebra with a haemal spine was counted as the first caudal vertebra, the urostylar complex the last. Following Smith-Vaniz (1971), the principal caudal ray counts included only those rays that articulate with the hypural plate, and are provided as upper+lower counts. Procurrent caudal rays are also provided as upper+lower counts. Measurements are expressed as a percentage of either standard length (SL) or head length (HL). Type  (Conway et al., 2014;Conway et al., 2017;Pinheiro et al., 2017;Tavera et al., 2021) and applied to phylogenetic and phylogeographic analyses. The DNA sequences from the Brazilian species A. pauciradiatus were extracted and amplified following protocols detailed in Weigt et al. (2012) for COI gene. Species of the families Gobiesocidae, Pseudochromidae, and Grammatidae were added as outgroups for the molecular-clock calibration. All sequences were aligned using ClustalW algorithm implemented in MEGA 7 (Kumar et al., 2016). Accession numbers of the sequences used are shown in Tab. S1. Phylogenetic analysis and molecular clock calibration. Interspecific and intraspecific genetic divergences were calculated in MEGA 7 using Tamura-Nei model. Phylogenetic relationships among Acyrtus and Arcos were reconstructed by Bayesian Inference using Mr. Bayes 3.2.6 (Huelsenbeck, Ronquist, 1997). The analysis was performed for two parallel runs of 10 million generations, with four chains each and sampling trees every 1,000 generations. The burn-in value and the effective sample size (ESS) were assessed using Tracer 1.5 (Rambaut et al., 2018). All parameters exceed 200 in ESS values. The consensus tree was obtained from the maximum credibility clades with TreeAnnotator 1.7.5 (Drummond et al., 2012). The appropriate substitution model used was determined using PartitionFinder (Lanfear et al., 2017). We used SYM+I+G model for the first codon position, F81+I for the second position, and GTR+I+G for the third position. We used species of Gobiesox as outgroups based on Conway et al. (2017).
We estimated the divergence times among Acyrtus species using the Relaxed Clock Log Normal model and the Birth/Death prior implemented in BEAUTi & The BEAST 2.5.0 software (Bouckaert et al., 2014). Species of Gobiesocidae, Pseudochromidae, and Grammatidae family were incorporated in the alignment to implement a secondary calibration derived from the results of Near et al. (2013), which were also used by Conway et al. (2017) in the Gobiesocidae family phylogeny. We choose the nodes A (80.3 Ma), B (75.8 Ma), F (42.9 Ma), and G (23.1 Ma) from Conway et al. (2017) to represent the divergence of the last common ancestors between (A) Pseudochromidae + Grammatidae + Gobiesocidae, (B) Grammatidae + Gobiesocidae, (F) all Gobiesocidae species, and (G) Acyrtus + Arcos + Gobiesox. We constrained all node calibrations based on the topology obtained by Conway et al. (2017), which represents the most complete phylogeny obtained for the subfamily Gobiesocinae. The analysis was run in the Cipres Portal (http://www.phylo.org/) using 100 million generations and sampling every 3,000 generations. The effective sample size (ESS) and appropriate burn-in values were visualized in Tracer 1.5 (Rambaut et al., 2018). All parameters exceed 200 in ESS values. We used TreeAnnotator 1.4.3 (Drummond et al., 2012) to obtain the maximum clade credibility tree that was edited in FigTree (Rambaut, 2014) and Inkscape (Free Software Foundation, Boston, USA).

Color in alcohol.
Body overall pale, with orange blotches and bands on the dorsal side and on the head; fins hyaline; pupils hyaline with black margin; orange blotches on the iris (Fig. 1).

Coloration in life.
Based on color photographs of live specimens (Figs. 1-2): body with variable red and white bands covered by small red spots; white bands might present red blotches; bands wider anteriorly and narrowing towards the caudal fin; pupil rounded and black, with thin white margin; white and red stripes and bands    2010, but not S. poiti nor Elacatinus pridisi Guimarães, Gasparini & Rocha, 2004), possibly for the same reasons, and in addition to the risks of fishing and ornamental trade. Therefore, Acyrtus simon is recommended to be categorized as VU according to the IUCN categories and criteria (IUCN Standards and Petitions Subcommittee, 2019).
Phylogenetic analyses of Acyrtus. COI sequences of 573 bp were obtained for 73 individuals of 20 species. Our analyses suggest the northwestern Atlantic as the center of the diversification of Acyrtus, and show that the genus is composed by at least nine species, forming a non-monophyletic group of three major clades (Fig. 3). One clade is composed by Caribbean and Brazilian species, another clade by the Caribbean A. rubiginosus lineage 1, A. rubiginosus lineage 2 and Arcos nudus, and a third solely composed by A. arturo, from the tropical eastern Pacific. Although three species (A. artius, A. lanthanum and A. rubiginosus) present a high distributional overlap along most of the Caribbean, our analysis revealed two cryptic species supported by high values of posterior probability, Acyrtus aff. artius from Tobago Island, and Acyrtus aff. rubiginosus from Belize (Fig. 3). Three species are endemic to the Brazilian Province and form a monophyletic group, which has Acyrtus aff. artius as the closest related species (Fig.  3). Brazilian species present small distributions: Acyrtus simon is restricted to Trindade Island, A. pauciradiatus, is restricted to Fernando de Noronha Archipelago and Rocas Atoll, and the undescribed species (Acyrtus sp., Fig. 3 The estimated date for the most recent common ancestor of the Atlantic species and A. arturo (eastern Atlantic) is at least 15 Mya, and the origin of Arcos nudus and the Acyrtus rubiginosus clade was among the oldest diversification events of the genus within the Atlantic (Fig. 4; Tab. 4). Most of the diversification in Acyrtus is recent, occurring during the Pliocene and Pleistocene. The Brazilian clade is the youngest, diversifying around 2.55 Mya, and the divergence time between Acyrtus simon and Acyrtus sp. is around 1.7 Mya (Fig. 4). Divergences were higher among Caribbean species than among Brazilian species.  Acyrtus + Arcos + Gobiesox (see Conway et al., 2017).

Evolutionary history of Acyrtus simon.
The close relationship between Acyrtus simon, from Trindade Island, and Acyrtus sp. from the Brazilian coast was also revealed in the haplotype network (Fig. 5) and through smaller F ST values than Acyrtus pauciradiatus (Tab. 5). Acyrtus simon, represented by only three haplotypes (Fig. 5), presented low haplotype and nucleotide diversity (Hd = 0.362; π = 0.001). Neutrality tests presented negative values (Tab. 6), though the Skyline plot did not present evidence of recent population growth in Trindade Island (Fig. 6)     number; Hd = haplotype diversity; SD Hd = standard deviation of Hd; π = nucleotide diversity; SD π = standard deviation of π.

DISCUSSION
Our study presents the first phylogeny of the genus Acyrtus including all known representatives, revealing: 1) insights on the evolutionary history of the genus; 2) the current paraphyletic status of the genus; 3) the Brazilian Province species complex as a monophyletic group; 4) and the presence of undescribed and cryptic species. Although our phylogenetic inference is based on a single DNA marker, our conclusions are based on high statistical support. The absence of reliable fossil data for the family Gobiesocidae (but see Schwarzhans et al., 2017) constitutes as one of the main barriers to the understanding of its evolutionary history. The secondary calibration from the divergence times obtained by Near et al. (2013), and used by Conway et al. (2017), was used here to study the genus Acyrtus. Near et al. (2013) estimated the origin of the family Gobiesocidae in the Eocene, around 42.9 Mya, which is consistent with the emergence of most reef fish families, between 66 and 34 Mya (Bellwood, 2015). Around 39 Mya (Conway et al., 2017), shortly after the origin of the family Gobiesocidae, its New World lineage (subfamily Gobiesocinae) was formed during a period marked by great diversification in the Tethys Sea (Renema et al., 2008). It is possible that earlier Gobiesocidae lineages originated in the Tethys Sea and then colonized both the Atlantic/eastern Pacific and the Indian-western Pacific oceans (Floeter et al., 2008). However, earlier Gobiesocidae lineages could also have had their origins in any of the oceans, using the Tethys as a passage. For instance, Gobiosomatini (Gobiidae) (Thacker, 2015), also endemic of the new world, originated in a period similar to Gobiesocinae, likely via dispersal from the western Indian Ocean through the Tethys passage (Thacker, 2015). While little can be concluded about the center of origin of Gobiesocidae, it seems that the Tethys Sea was important for dispersal and early diversification in the family.
After the rise of the Gobiesocinae, Acyrtus lineages and Arcos erythrops diverged much earlier than the closure of the Isthmus of Panama, around ~21 Mya. Therefore, this event did not influence the divergence between Gobiesocinae genera as commonly seen in other groups (Lessios, 2008). Tavera et al. (2021) found a similar divergence time for Arcos erythrops, although the placement of this species did not evidence the monophyly of Arcos and Acyrtus. Despite this, our results support the topology obtained by Conway et al. (2020) based on seven different genes. Nevertheless, further investigation of the Arcos phylogeny with broader taxon sampling is needed.
Our results also show the emergence of the western Atlantic Acyrtus at least 15 Mya (the age of the last common ancestor of Acyrtus + Arcos nudus) in a period of origin and diversification of most reef fish genera (Bellwood et al., 2015). Moreover, the origin of Acyrtus coincides with the emergence of the Amazon barrier, which could have prevented an earlier dispersal from the Caribbean to the Brazilian Province. It could explain the latter diversification of Acyrtus in the history of this genus (~3.9 Mya). In this case, the glacioeustatic sea-level changes of the Pleistocene/Pliocene could have contributed to the connectivity between regions, and to the crossing of the biogeographic barrier, as suggested for many other reef fishes (Rocha, 2003).
The diversification of the Brazilian complex of species started during the Pleistocene and these three species present restricted distributions: two are endemic to oceanic islands, and a third is only known from a narrow geographic range of the Brazilian northeastern coast. There are two hypotheses for speciation and colonization routes. The first involves the colonization of the Fernando de Noronha archipelago, with posterior colonization of the Brazilian coast, which recently colonized Trindade Island. This hypothesis is corroborated by the fact that a few Caribbean species are also found in Fernando de Noronha but not along the Brazilian coast (Rocha, 2003). This archipelago could be a gateway for Caribbean species to colonize Brazilian waters. A second hypothesis would involve the colonization of the Brazilian coast first, with a subsequent and earlier colonization of Fernando de Noronha, and a more recent colonization of Trindade also from the coast. Some reef fish lineages were more widely distributed along the southwestern Atlantic in the past, persisting as relicts in restricted locations (Rocha et al., 2010;Pinheiro et al., 2017). The exposure of seamounts during low sealevels could have favored the colonization of the Vitória-Trindade Chain (Macieira et al., 2015;Pinheiro et al., 2017).
Even though we observed low genetic diversity in our results for A. simon, there was no sign of population bottlenecks or expansion in the neutrality test, differently from other VTC endemics (Pinheiro et al., 2017). Differently from species with wider distribution (Pinheiro et al., 2017), the restriction of Acyrtus to shallow reefs and its absence on the VTC seamounts may have constrained its connectivity between the coast and the islands and limited the input of new haplotypes. Recent population expansions are seen in many species along the western Atlantic (da Silva et al., 2015;Liedke et al., 2020) that seem to be related to a ~90% increase in reef area caused by the rise in sealevel following the last glacial maximum (Ludt, Rocha, 2015).
Both Conway et al. (2014) and our study show three undescribed Caribbean Acyrtus species, one of them closely related to Acyrtus artius and the two other to A. rubiginosus. The existence of many hidden lineages in the same genus may be attributed to the species small size, cryptic behavior, and a low morphological divergence. The latter, in particular, is commonly related to recent speciation and/or stabilizing selection on the ancestral phenotype (Milá et al., 2017). A similar pattern is found for Gouania (Gobiesocidae) in the Mediterranean Sea, where recent diversification and cryptic lineages were recognized by Wagner et al. (2019). These results evidence that future molecular studies for Gobiesocidae should lead to the discovery of many cryptic species. The most likely cause for the high diversification in this family (even within provinces) is related to their weak dispersal potential, small size, sedentary habit, and demersal eggs (Pires, Gibran, 2011). These life-history characteristics are among the most important drivers of speciation in the Brazilian Province (Pinheiro et al., 2018;Mazzei et al., 2021;Simon et al., 2021), and other taxonomic groups that share similar traits (e.g., Labrisomidae and Gobiidae) also show strong genetic structure and cryptic speciation within the Great Caribbean (Baldwin et al., 2011;Victor, 2014). Additionally, the distribution of the closely-related Brazilian Acyrtus species in different environments (mainland coast and oceanic islands) and localities with distinct levels of isolation suggests that ecological and peripatric speciation processes might be important drivers of diversification in Gobiesocidae (Rocha et al., 2005;Pinheiro et al., 2017;Simon et al., 2021).
Our phylogenetic inference discloses the paraphyletic nature of Acyrtus, which includes Arcos nudus in its clade. The first Acyrtus species (A. rubiginosus) was described in 1868, originally assigned to the genus Sicyases, a valid genus described based on Sicyases  (Schultz, 1944), based on Acyrtus rubiginosus and Arcos erythrops, respectively. Arcos nudus was originally described by Linnaeus (1758) as Cyclopterus, a valid and now monotypic genus described based on Cyclopterus lumpus Linnaeus, 1758 (Cyclopteridae or lumpfishes). It was later reassigned as Gobiesox nudus by Briggs (1955), and afterwards as Arcos nudus by Fernholm, Wheeler (1983). All other Arcos species are from the tropical eastern Pacific and were originally described as Gobiesox, a valid genus described based on Gobiesox cephalus Lacépède, 1800.
Our phylogenetic analyses indicate that Arcos nudus should be reassigned to Acyrtus. This species shares several distinguishing morphological characters with other Acyrtus species, including large secretory cells present inside the groove present in the subopercular spine (Conway et al., 2014). Therefore, according to our data, the genus Arcos seems to be exclusive to the eastern Pacific, while most Acyrtus are found in the western Atlantic. Additional studies including more eastern Pacific Gobiesocidae species are necessary to better assess the status of Acyrtus arturo, from Malpelo Island (Tavera et al., 2021), and Arcos nudus, from the Atlantic, which could belong to other clades. Alternatively, a more complete phylogenetic analysis including all Arcos species could also show clades uniting Acyrtus and Arcos erythrops, suggesting the unification of Acyrtus and Arcos in a single genus. Finally, our results highlight that a broad sampling in Gobiesocidae family will bring important insights about evolutionary patterns of cryptobenthic fishes.

ACKNOWLEDGMENTS
We thank the Brazilian Navy for support during scientific expeditions to Trindade Island. We also thank Raphael M. Macieira, C. L. S. Sampaio, R. G. Santos for support in the field and with samples, Fabio Di Dario, Kevin W. Conway