Mulloidichthys flavolineatus flavicaudus Fernandez-Silva & Randall (Perciformes, Mullidae), a new subspecies of goatfish from the Red Sea and Arabian Sea

Abstract The number of goatfish species has increased recently, thanks in part to the application of molecular approaches to the taxonomy of a family with conservative morphology and widespread intraspecific color variation. A new subspecies Mulloidichthys flavolineatus flavicaudus Fernandez-Silva & Randall is described from the Red Sea and Arabian Sea, including Socotra and Gulf of Oman. It is characterized by a yellow caudal fin, 25–28 gill rakers, and 37–38 lateral-line scales and it is differentiated from nominal subspecies Mulloidichthys flavolineatus flavolineatus by 1.7% sequence divergence at the mitochondrial cytochrome b gene. The morphometric examination of specimens of Mulloidichthys flavolineatus flavolineatus revealed variation in head length, eye diameter, and barbel length, in western direction from the Hawaiian Islands, South Pacific, Micronesia, and the East Indies to the Indian Ocean. The population of Mulloidichthys flavolineatus flavicaudus subsp. n. in the Gulf of Aqaba differs from that of the remaining Red Sea by shorter barbels, smaller eyes, shorter head, and shorter pelvic fins. We present a list of 26 endemic fishes from the Gulf of Aqaba and discuss the probable basis for the endemism in the light of the geological history of this region.


Introduction
The goatfish Mulloidichthys flavolineatus was described by Lacepède (1801) based on a manuscript written by Dr. Philibert Commerçon (Commerson in English). There is no type specimen and no record of the type locality (Bauchot et al. 1985). It is almost certainly Mauritius, where Commerson spent several years collecting biological specimens, including many fishes. Fricke (1999: 309) designated a neotype for M. flavolineatus from nearby Réunion, but it was later considered invalid by him (Fricke 2000: 639) as "not sufficiently in accordance with Article 75b and Article 75d of the International Code of Zoological Nomenclature." We designate and describe a neotype in the present paper ( Fig. 1) collected and photographed in Mauritius by the second author. We also illustrate a live individual from the island (Fig. 2).
Mulloidichthys flavolineatus is presently regarded as the most wide-ranging species of the family Mullidae, from the northern Red Sea (Ben-Tuvia and Kissil 1988) to the Pitcairn Islands (Nichols 1923;Randall 1999). Such a broad distribution might be expected from the unusually large size attained by the postlarvae at settlement, 60 to 80 mm SL (Randall 2005). It is also unusual for such a common and widespread species to have only two junior synonyms, Mulloides samoensis Günther, 1874, type locality, Upolu, Samoa Islands, and Upeneus preorbitalis Smith & Swain, 1882, type locality, Johnston Atoll.   Like other goatfishes, this species uses the pair of sensory barbels on its chin to locate prey, mainly in sedimentary substrata, as seen in Fig. 3 of an adult in the Hawaiian Islands and one in the Red Sea (Fig. 4). Randall (2005: 292) summarized the prey of specimens from the Hawaiian Islands as small crabs, shrimps, polychaete worms, small bivalve mollusks, hermit crabs, crab megalops, heart urchins, small gastropods, amphipods, foraminifera, and unidentified eggs. During periods of inactivity, the fish may be seen hovering in aggregations a short distance above the bottom (Fig. 5) or in groups resting on sand (Fig. 6). Myers (1999: 159) reported spawning in Palau over shallow sandy areas near the reef's edge for several days following new moon. Females in the Mariana Islands may be mature as small as 123 mm in SL, and males as small as 112 mm. The spawning season is December to September, with peaks from March to April. Large aggregations of silvery postlarvae settle out between March and June to shallow water on reef flats where they are often caught in seines or throw nets.
We, and surely others, have noticed that the population of Mulloidichthys flavolineatus in the Red Sea has only yellow caudal fin (Fig. 7), whereas in most of the Indian Ocean and in the Pacific, the caudal fin is usually gray but occasionally also yellow. This goatfish should not be confused with M. vanicolensis (Valenciennes, 1831), which also has a yellow caudal fin (lead fish of the three of Fig. 8), as well as yellow dorsal, anal, and pelvic fins, whereas pelvic and dorsal fins are whitish in M. flavolineatus. The geographic distribution of the two color morphs of M. flavolineatus matches the distribution of two distinct mitochondrial lineages with 1.7% divergence at the cytochrome b (cytb) gene (Fernandez-Silva et al. 2015).      The caudal fin continues to be yellow from the Red Sea into the Gulf of Aden and Socotra, as shown by Fig. 9, where a few individuals of Mulloidichthys flavolineatus have mixed with a school of M. ayliffe. Uiblein (2011) described the latter in a review of the species of Mulloidichthys of the Western Indian Ocean. It mimics and often schools with the snapper Lutjanus kasmira. It is an amazing example of parallel evolution with M. mimicus Randall & Guézé, 1980 of the Marquesas Islands and Line Islands in the Central Pacific, which closely mimics the stripe pattern of L. kasmira and forms aggregations with it.
Across the Arabian Sea to the south coast of Oman aggregations of Mulloidichthys flavolineatus in Oman and Maldives include many individuals with yellowish caudal fin mixed with a few gray-tailed and yellow-tailed fish (Figs 10,11 and 12). Elsewhere, caudal fins are predominantly white or light gray, although we have observed that the color of the caudal fin in individuals from South Africa to French Polynesia and Hawaiian Islands may vary from hyaline gray (predominantly) to yellow (occasionally).

Measurements and counts
Type specimens were deposited at the Bernice P. Bishop Museum, Honolulu, HI, U.S.A. (BPBM); the California Academy of Sciences, San Francisco, CA, U.S.A. (CAS); the Museum of the Hebrew University of Jerusalem, Israel (HUJ); the Senckenberg Museum, Frankfurt, Germany (SMF); and the U.S. National Museum of Natural History (NMNH). These were the primary sources of goatfish specimens examined in this study.
Lateral-line counts begin with the first pored scale completely posterior to the upper end of the gill opening and end at the base of the caudal fin (three pored scales continue onto the caudal fin). Counts of gill rakers were made on the first gill arch; they include all rudiments.
Lengths of specimens are given as standard length (SL), measured from the median anterior point of the upper lip to the base of the caudal fin (posterior end of the hypural plate); body depth is taken vertically from the base of the first dorsal-fin spine where it emerges from the body (not the internal base); body width is the maximum width measured just posterior to the gill openings; head length (HL) from the front of the upper lip to the posterior end of the opercular membrane, and snout length from the same anterior point to the nearest fleshy edge of the orbit; orbit diameter is the greatest fleshy diameter, and interorbital width the least fleshy width; upper-jaw length is taken from the front of the upper lip to the end of the maxilla; barbel length is the maximum straight length; caudal-peduncle depth is the least depth, and caudalpeduncle length the horizontal distance between verticals at the rear base of the anal fin and the caudal-fin base; length of fin spines and rays of the dorsal and anal fins are measured from where they emerge from the body to their tip; caudal-fin length is the horizontal length from the posterior end of the hypural plate to a vertical at the tip of the longest ray; caudal concavity is the horizontal distance between verticals at the tips of the shortest and longest rays; pectoral-fin length is measured from the base of the uppermost ray; pelvic-fin length is measured from the base of the pelvic spine to the tip of the longest soft ray. Proportional measurements in the text are rounded to the nearest 0.05.
Only meristic characters and measurements that vary between M. f. flavolineatus and M. f. flavicaudus subsp. n. were applied in the diagnoses and comparisons: the number of gill rakers, lateral-line scale counts, barbel length, eye diameter and head length. We also compared the length of the pectoral and pelvic fins, but these did not show differences between M. f. flavolineatus and M. f. flavicaudus subsp. n.
Because goatfishes present allometric changes in body form (Uiblein and Heemstra 2010) during ontogeny, in the current study we only included fish > 73 mm and <288 mm.

Genetic methods
During a previous phylogeographic survey of M. flavolineatus we obtained cytb sequences from 217 specimens sampled at nineteen sites throughout the Red Sea, the Arabian Sea, the Indian Ocean and the Pacific Ocean. To elucidate phylogenetic relationships we sequenced an additional fragment of the mitochondrial genome, the ATP synthetase 8 and ATP synthetase 6 (ATPase-8 and ATPase-6) regions, from individuals representative of the cytb diversity. We also sequenced an individual of M. vanicolensis and one of M. pfluegeri to use as outgroups. Briefly, DNA was extracted from fin clips and Polymerase Chain Reactions (PCR) were carried out using the primers L8331 (5'-AAA GCR TYR GCC TTT TAA GC-3') and H9236 (5'-GTT AGT GGT CAK GGG CTT GGR TC-3') (Meyer 1993). We carried out PCRs in a 15 µl volume containing 5 to 20 ng of template DNA, 0.1 µM of each primer and 5 µl of BioMix Red™ (Bioline Inc., Springfield, NJ, U.S.A.) in deionized water. PCRs were carried out with an initial denaturation step of 95 °C for 4 min, 35 cycles of denaturation (95 °C for 30 s), annealing (52 °C for 30 s) and extension (72 °C for 45 s), followed by a final extension step of 72 °C for 10 min. To clean PCR products we treated them with 0.75 units of Exonuclease I and 0.5 units of Fast Alkaline Phosphatase (ExoFAP; Thermo Fisher Scientific, Waltham, MA, U.S.A.) per 7.5 µL of PCR product, at 37 °C for 15 min, followed by deactivation at 85 °C for 15 min. We cleaned all PCR products using ExoSAP (USB, Cleveland, Ohio) and then sequenced them in the forward direction (and reverse direction, where appropriate) using a genetic analyzer ABI 3130XL (Applied Biosystems, Foster City, California) at the Hawai'i Institute of Marine Biology EPSCoR Sequencing Facility. The ATPase-8 and ATPase-6 sequences were aligned, edited, and trimmed to a common length using GENEIOUS PRO vers. 4.8.4 (Drummond et al. 2012), and the sequences were deposited in Genbank (accession numbers: KT960949-KT960972). We concatenated this alignment with the cytb sequences from the same specimens and applied Bayesian methods for phylogenetic reconstruction in BEAST vers. 1.8.0 (Drummond et al. 2012), based on Yule models of speciation and a strict molecular clock (1% per myr as per Bowen et al. 2001). We also applied Neighbor-joining distance and Maximum-Likelihood tree-building methods for phylogenetic reconstruction using MEGA (Tamura et al. 2013) and the RaxML web server at http://embnet.vital-it.ch/raxml-bb/ (Varsamos et al. 2005), respectively. Support for the trees was evaluated by bootstrapping over 1,000 replicates.

Data resources
The data underpinning the analysis reported in this paper are deposited in the Dryad Data Repository at http://dx.doi.org/10.5061/dryad.f54m5  Description. Meristics are provided in Tables 1 & 2 and measurements as % of SL in Table 4 and Fig. 14. Below, morphometric ratios are given as ratios of SL for the holotype and in parentheses for selected paratypes (n=7), except where indicated.
Mouth small, maxilla not reaching a vertical at front of orbit, upper-jaw length 12.3 (12.2-13.9) in SL; jaws with small conical teeth, in two rows with teeth more irregularly placed between both rows; no teeth on the vomer and palatines; anterior nostril small, elliptical, two-thirds eye diameter in front of eye; posterior nostril small, elliptical, at dorsoanterior corner of orbit; opercular spine flat, at mid-eye height.  These are the same data as in Table 3.
Scales very finely ctenoid; head fully scaled; scales on the base of caudal fin, other fins without scales; dorsal fin behind the vertical at fourth lateral line scale, origin of second dorsal above 18th (17th in some paratypes) scale. Pored scales on lateral line with many branching tubules. Color. Color in life silvery white to yellowish, slightly darker over lateral line; margin of each scale on upper half of body darker than scale. Yellow stripe on side of body at level of eye, from posterior margin of orbit to caudal-fin base, bordered by a narrow whitish stripe (stripe sometimes slightly blue); the stripe usually containing a black spot above posterior part of pectoral fins (under the first dorsal fin), sometimes faint due to fading, stripe anterior to spot occasionally indistinct; barbels white; dorsal fins usually transparent, sometimes first dorsal fin with yellowish tinge; pectoral, anal, and pelvic fins whitish, translucent; caudal fin yellowish or yellow. Color when fresh often pink and all fins yellow. Uniformly creamy white in preservative.
Etymology. Mulloidichthys f. flavicaudus subsp. n. is named in reference to the yellow color of the caudal fin, in contrast to the whitish gray color of the caudal fin of M. f. flavolineatus.
Distribution. Mulloidichthys f. flavicaudus subsp. n. is restricted to the NW Indian Ocean biogeographic province, where it ranges from various locations in the Red Sea (including the Gulf of Aqaba), the Gulf of Tadjoura, the Gulf of Aden, and Socotra (Fig. 9). M. f. flavicaudus subsp. n. has extended its range to Oman (Fig. 11) and probably to the Maldives (Fig. 12), where it has encountered the western distribution of M. f. flavolineatus. Underwater photographs of fish with yellow and gray caudal fins suggest overlap and interbreeding by the two subspecies. Carpenter et al. (1997) included M. flavolineatus in their catalog of fishes of the Arabian Gulf. They did not cite any voucher specimens, and the photo they used is from Mauritius.
Color. Silvery white to yellowish, slightly darker over lateral line, margins of each scale on upper half of body darker than scale. Yellow stripe on side of body at level of eye, beginning from posterior margin of orbit and ending at caudal-fin base, bordered by two whitish narrow stripes (sometimes slightly blue); the stripe usually containing a black spot above posterior part of pectoral fins (under the first dorsal fin), sometimes faint due to fading, stripe anterior to spot occasionally indistinct; barbels white; dorsal fins usually transparent, sometimes first dorsal fin with yellowish tinge; pectoral, anal, and pelvic fins whitish, translucent; caudal fin varying from usually white or light gray to occasionally yellowish or yellow. Sometimes body color pattern of broad irregular red-brown bars, especially at night. When fresh, body color can turn pink and all fins yellow. Uniformly creamy white in preservative.
Distribution. Mulloidichthys f. flavolineatus is wide-ranging from East Africa north to the Maldives and Chagos Archipelago and east to the Hawaiian, Marquesas and Pitcairn Islands, north to the Ryukyu and Bonin Islands and south to Lord Howe Island, New Caledonia and Rapa Island (Randall 2002, Uiblein 2011 (Fig. 15).
Genetics. The parsimony-based haplotype networks constructed with mtDNA cytb sequences from 217 M. flavolineatus specimens revealed a separation between individuals from the NW Indian Ocean (including the Red Sea, the Gulf of Aden and Oman) and individuals in the rest of the Indian Ocean and the Pacific Ocean (Fig. 16). Corrected genetic distance was 1.7%, with seven diagnostic mutations (Fernandez-Silva et al. 2015).
We obtained a concatenated alignment of a 715-bp segment of the cytb gene and a 731-bp segment of the ATPase-8 and ATPase-6 genes of the mitochondrial genome from seven individuals from the Red Sea (Jeddah) and five from the Pacific (Hawai'i and Okinawa). Phylogenetic reconstructions based on Bayesian inference (Fig. 17) revealed a genetic break and the presence of two well-supported monophyletic clades (posterior probability = 1): one with sequences from the Red Sea and one with the haplotypes from the Pacific. Reconstructions based on the Maximum-Likelihood and Neighbor-Joining methods were in agreement with this topology but clades had lower statistical support (results not shown).

Discussion
Higher gill-raker and lateral-line counts, smaller eyes and stable yellow coloration of the caudal fin in M. flavolineatus from the Red Sea are characters in alignment with the genetic isolation of a mitochondrial lineage in the NW Indian Ocean biogeographic province (as per Kulbicki et al. 2013) and support the subspecies designation of M. f. flavicaudus subsp. n.
Some ichthyologists, notably Gill (1999), have questioned the validity of subspecies in marine fishes, especially in reference to wide-ranging Indo-Pacific species. One could argue that the existence of subspecies should be demonstrated by intermediates between two isolated populations before they could be labeled as subspecies. Divisions of populations into two or more populations have resulted from the change in sea level caused by the variation in the size of the polar ice caps. The Indian Ocean was isolated from the Pacific, and the Red Sea from the Indian Ocean when the ice caps were very large. We assume that the yellow-tailed population of Mulloidichthys flavolineatus arose as a subspecies when the Red Sea was isolated, approximately half a million years ago assuming a molecular clock of 2% divergence per million years (as per Bowen et al. 2001). This population persisted in isolation through several Pleistocene glacial cycles (Fernandez-Silva et al. 2015) and over time extended out to Socotra, Oman and possibly Maldives, where it entered into secondary contact with the Indo-Pacific population. In the second author's book Coastal Fishes of Oman (Randall 1995), a single individual of M. flavolineatus is illustrated as Figure 620. It has a yellowish caudal fin. He wrote in the brief species account, "fins whitish, the caudal fin often yellowish." The underwater photograph of M. flavolineatus of Fig.  11 taken on the south coast of Oman shows caudal fins varying from pale greenish gray (the green part from the sea color) to a few all yellow. This photograph suggests that the two subspecies of M. flavolineatus may overlap and interbreed, hypotheses to be confirmed with genetic methods. The geographic extension of the yellow-tailed subspecies in the understudied Western Indian Ocean warrants further investigation.
Notably, the age of split of the Mulloidichthys flavolineatus subspecies is older than the radiation that gave rise to M. vanicolensis, M. mimicus, M. dentatus (Gill, 1862) and M. martinicus (Cuvier, 1829) less than 350,000 years ago (unpublished results).
It is remarkable that individuals of Mulloidichthys f. flavicaudus subsp. n. from the Gulf of Aqaba have consistently smaller eyes, longer head, and longer barbels than fish from the Red Sea proper (Fig. 14). Pelvic fins are also shorter in the Gulf of Aqaba (mean length in SL = 5.17) than in the rest of the Red Sea (4.40 in SL). However, both populations extensively share cytb haplotypes and the analyses of haplotype frequencies do not support genetic differentiation, although this comparison is based on mitochondrial markers only (Fernandez-Silva et al. 2015). In the northern tip of the Gulf of Aqaba, M. f. flavicaudus was among the 11 most common species on the shallow sandy habitat, but all specimens were juveniles or subadults (maximum length: 15 cm TL) (Golani 1993;Golani and Lerner 2007). The Gulf of Aqaba has remarkably high endemism. Twenty-six species of fishes, including the goatfish Up-  (Table 5). Although further research may result in range extensions for some of these fishes to the Northern Red Sea, the number of endemics is very high for an area of only 160 × 24 km. Environmental differences could explain this isolation. The Gulf of Aqaba is much deeper (1850 m) than the Red Sea to the south, and seawater temperature is considerably lower (20-27°C) and salinity higher (40-41‰) than in the Red Sea proper (25-31°C; 37-41‰) (Oren 1962). Moreover, the Gulf may have acted as a glacial refuge for reef fauna during Pleistocene low sea level stands, when most of the Red Sea was too saline for coral reef development. Geological and paleoclimatic research suggest that during these periods the Gulf of Aqaba, owing to rainfall and fluvial intake, maintained lower salinity levels and that environmental conditions were favorable to sustain coral reefs and associated fauna (DiBattista et al. 2016). Therefore, the Gulf of Aqaba served as a refuge for marine life from the harsh marine environment to the south. Parapatric speciation processes reinforced by selection may account for the elevated endemism in the region (Golani 1993;Por 2008;Tikochinski et al. 2013).
Our range-wide phylogeographic survey of Mulloidichthys flavolineatus (Fernandez-Silva et al. 2015) indicated the genetic isolation of the Hawaiian population (including Johnston Atoll) from the remainder of the Indo-Pacific. Uiblein (2011) indicates that Pacific Ocean M. flavolineatus have shorter barbels than those in the Indian Ocean, but he includes the Hawaiian Islands with the rest of the Pacific in this study. We found the Hawaiian population has shorter barbels, shorter head, smaller eyes, higher gill-raker counts, and higher lateral-line scale counts than all other populations examined, and that there is a range of variation as we move from Hawai'i to other islands of Oceania, the West Indies, the Western Indian Ocean, and the Red Sea (Tables 1-3, and Fig. 14).
Sinclair-Taylor, E. Tong, Yamada-san and D. Uyeno. R. Toonen, R. Kosaki, M. Berumen, S. Nakachi, M. Obuchi, D. Pence, T. Naruse, O. Takama, E. Kawai, E. Mason and N. Prevot helped or provided logistic support to the collection of specimens in the range-wide study leading to this work. We are also grateful for continued support to the members and alumni of the ToBo Lab at the Hawai'i Institute of Marine Biology (HIMB), the California Academy of Sciences, Bernardi's lab at UC Santa Cruz, MISE and Tachihara's Labs at the University of the Ryukyus, the Biological Institute at Kuroshio, the Coral Reef Ecology Lab at KAUST, and David Posada's research group at the University of Vigo, as well as administrative staff at these institutions. Helen Randall has also our heartfelt gratitude for her patience, support and mangoes. This work was financially supported by a U.S. National Science Foundation (NSF) grant OCE-0929031 to Brian. W. Bowen, the KAUST -OCRF under Award No. CRG-1-2012-BER-002 and baseline research funds to Michael L. Berumen, a Fulbright Postdoctoral Fellowship and a Marie Curie -7th European Community Framework Programme to I.F.-S., and a National Geographic Society Grant 9024-11 to Joseph D. DiBattista.