On the occurrence of Abudefduf spp . (Pisces: Pomacentridae) in the Mediterranean Sea: a critical review with new records

Fishes of the genus Abudefduf are increasingly being recorded in Mediterranean waters, with most of these occurrences being based on the sole inspection of underwater images or video footage. Here we present a critical review of these occurrences, taking into account the uncertainties related to the identification of the congeneric A. vaigiensis and A. saxatilis, and testing the validity of the external traits used so far to distinguish these two species. After a careful revision of the Mediterranean literature and based on the analyses of images mostly taken in their natural native environment, we highlight the possibility that major morphological characters previously used to distinguish A. saxatilis from A. vaigiensis are not consistent within the species and thus not sufficiently reliable for their discrimination. This also relates to the possible confusion of these two species with a third species A. troschelii , never reported in the Mediterranean Sea to date. Finally, new records of the genus Abudefduf are reported for the Adriatic Sea and Tunisian waters.


Introduction
The Mediterranean Sea is considered a hotspot of marine bioinvasions with an increasing rate of new introductions through time (Edelist et al. 2013). According to the most recent revisions, the total number of multicellular alien species in the Mediterranean is close to 1000 (Zenetos 2019). Exotic fishes are mostly represented by Red Sea species, entering through the Suez Canal, but over the last decade the number of introductions mediated by shipping and due to aquarium release has rapidly increased (Rilov and Galil 2009;Zenetos 2019), raising justified concern. This phenomenon is increasingly apparent in the Adriatic Sea, where an increasing number of new fishes entering through the Suez Canal or transported through ships, mobile platforms or aquaria  OPEN ACCESS. Pećarević et al. 2013;Lipej et al. 2008) has been reported. A similar situation exists in the Strait of Sicly , with new records regularly being reported along both the Italian (e.g. Azzurro et al. 2018a) and Tunisian (e.g. Ounifi-Ben Amor et al. 2016) coasts. Monitoring these subareas is of primary importance for supporting environmental policies at the European and Mediterranean level, as embodied within the Descriptor 2 of the European Marine Strategy Framework Directive (MSFD; EU2017/848) and Indicator 6 of the Barcelona Convention Integrated Monitoring and Assessment Programme (IMAP).
Among these new guests, the Sergeant major A. saxatilis and the Indo-Pacific sergeant A. vaigiensis have been repeatedly reported from practically all major sectors in the Mediterranean Sea (Osca et al. 2020;Lipej et al. 2019). These two species have a wide, yet disjunct native distribution, A. saxatilis being distributed in the western and eastern Atlantic and A. vaigiensis in Indo-Pacific waters (Getlekha et al. 2016;Froese and Pauly 2020). Due to their remarkable resemblance, A. vaigiensis and A. saxatilis can be easily confounded (Deidun and Castriota 2014), but some key external morphological characters (De Beaufort 1940;Allen 1991) are currently used to distinguish between these two species, especially in the Mediterranean literature, which is rich of Abudefduf records based on visual inspection only, documented by underwater photos or video footage (Osca et al. 2020).
According to De Beaufort (1940), A. saxatilis should be distinguished from other congenerics by the presence of the fifth vertical bar below the posterior end of the dorsal fin which extends on the dorsal rays in posterior part of the dorsal fin. Furthermore, Allen (1991) stated that in this species, the fourth bar usually originates immediately behind the last dorsal spine or is in contact with it and Azzurro et al. (2013) reported a personal communication by Dr. Phil Heemstra, who considered the presence of two black spots on the caudal peduncle as a further character to distinguish this species from A. vaigiensis.
Regarding the identification of A. vaigiensis, De Beaufort (1940) notes the presence of the fifth vertical bar on the caudal peduncle, behind the dorsal fin and Allen (1991) states that the fourth bar usually arises more towards the center of the soft dorsal fin, well behind the last spine. Also, Tsadok et al. (2015), investigating morphological differences between individuals of A. vaigiensis from the Red Sea and A. saxatilis introduced in the eastern Mediterranean, concluded that A. vaigiensis specimens had an interrupted extension of the fifth bar to the dorsal fin (except in a young individual) and no black spots on the caudal peduncle. Meristic counts in both species are overlapping (Allen 1991) and as such have no value for comparison between these two species. There is also a third species, A. troschelii, which is very similar to both A. vaigiensis and A. saxatilis but it has never been reported in the Mediterranean. According to Allen (1991), the only external taxonomic character useful in distinguishing A. troschelii from the other two species is the presence of scales on the inner side of the pectoral-fin base.
The aim of this study was to perform a critical review of the Mediterranean records of the genus Abudefduf, investigating the validity of external morphological characters used so far to distinguish A. saxatilis from A. vaigiensis. We also present new records of the genus Abudefduf spp. from the Adriatic Sea and Tunisian waters.

Materials and methods
Evaluation of the taxonomic consistency of the external characters used to distinguish A. saxatilis from A. vaigiensis The validity of the external characters most often used in the identification of Abudefduf saxatilis and A. vaigiensis, was tested through inspection of a number of underwater photos of these species taken in their donor regions: the Atlantic and the Pacific Ocean, respectively. Some photographs also featured specimens collected for scientific purposes. Images were downloaded from authoritative internet sources (mainly from Fishbase https://www.fishbase.se/search.php and Flickr https://www.flickr.com), selecting only high-quality photos with traceable author and indicated geographical location (Supplementary material Table S1). Each of these photos was visually analyzed to test for the presence of three of the most cited external characters: T1 = fourth bar originating directly behind the last dorsal spine or in contact with it in A. saxatilis. In A. vaigiensis the fourth bar usually starts well behind the last dorsal spine, more towards the middle of the soft part of dorsal fin (Allen 1991). T2 = fifth vertical bar below the posterior end of the dorsal fin, extending on the dorsal rays in the posterior part of the dorsal fin in A. saxatilis. In A. vaigiensis fifth vertical bar is on the caudal peduncle, behind the dorsal fin (De Beaufort 1940).

T3 = presence of two dark dots on the caudal peduncle in A. saxatilis.
Absence of these spots in A. vaigenisis (P. Heemstra personal communication reported by Azzurro et al. 2013).
Considering the disjunct geographical distribution of A. saxatilis vs A. vaigiensis (Froese and Pauly 2020), we expect the above-mentioned traits to be sufficient to distinguish A. saxatilis photographed in the Atlantic Ocean from A. vaigiensis occurring in Indo-Pacific waters. A. troschelii was not included in our image analysis because it has not been reported from the Mediterranean to date and since the only useful external character to distinguish this species from two congeners could not be evaluated from the photos.

Review of the Mediterranean records of Abudefduf spp.
Mediterranean records of Abudefduf spp. were primarily explored through bibliographic search carried out with the Scopus (www.scopus.com) and Google Scholar (https://scholar.google.com/) engines. The words "Abudefduf" and "Mediterranean" were searched in the title, abstract or keywords in all the peer-reviewed literature until January 2020. An analogous search, using the same keywords was performed on Google. Upon retrieval of published papers, we analyzed all documented records, extracting relevant information on geographical locations, time of observation and taxonomy. In addition, each photo of the target species appearing in published papers, previously identified as either A. vaigiensis or A. saxatilis, was carefully evaluated according to the presence or absence of the traits T1, T2 and T3 (Table S2).

New records of Abudefduf sp. in the Mediterranean Sea
On September 14 th , 2018, a sportive fisherman observed and filmed a single individual of Abudefduf sp. on the outer side of a stone pier of the harbor "Zenta" in Split (eastern middle Adriatic, Croatian coast; 43.4987°N; 16.4606°E) ( Figure 1A). Besides footage, the fisherman provided data on location and depth of the area where the fish was observed. The video was recorded with an underwater camera GoPro 4 in high-quality resolution (full HD). The quality of the footage and the fact that the fish was shot at multiple angles, which revealed important diagnostic characters, allowed the identification of the fish to the genus level. The authenticity of the location where the video was recorded was confirmed by one of the authors (P.U.), an experienced spearfisher, who recognized the area which he frequently visits during fishing activity.
On 12

Evaluation of morphological traits based on visual inspection of photographs
A total number of 38 photographs were downloaded according to the adopted criteria and subsequently analyzed by three experts (B.D., P.U. and E.A.) without prior knowledge of the location where the photos were taken in order to avoid possible bias. After excluding photos unsuitable for the analysis (mostly due to poor quality, inadequate position of the fish or appearance of multiple specimens, not sharing the same characters) a total of 22 photos were retained. Each of these photos was tested for characters T1, T2 and T3 (Figure 2). Among the Abudefduf photos taken in the Atlantic Ocean, 3 out of 8 individuals did not exhibit character T3 and one of them did not comply with all morphological characters of A. saxatilis, but instead showed T2 and T3, which are in line with that of A. vaigiensis.
To confirm the location of where the photo was taken, the author of the photo was contacted. The author confirmed that the photos were taken along the coast of Akumal (Mexico, Western Atlantic) and provided further photographs of Abudefduf sp. from the area. These additional images showed other individuals, sharing the characters of A. vaigiensis, except for the T1 character, the origin of the fourth bar (photos in the black frame in Figure 2).
Out of the 12 photographs taken in the Indo-Pacific region, only four individuals fully complied with our expectations (characters T1, T2, T3) for A. vaigiensis. On the contrary, 8 individuals showed a mixture of traits, with 3 individuals showing characters of A. saxatilis, according to the T1, T2, T3 criteria. The photos of these latter were taken in the Persian Gulf and Oman. This information was gathered by contacting the photographers themselves. Two out of three photographers confirmed the location where the photos were taken, whilst one author did not reply. Based on these findings, the identification of A. saxatilis vs A. vaigiensis cannot be assigned with certainty using only the characters T1, T2, T3. This taxonomic confusion should be also extended to A. troschelii.

Review of the Mediterranean records of Abudefduf spp.
Our online search resulted in 14 relevant publications, which reported the Mediterranean occurrence of at least 46 individuals belonging to the genus Abudefduf. The real number of observed individuals within these publications is considerably greater since some publications recorded undefined numbers of individuals (several).
Based on the findings of our photo analyses, we reassigned, as a cautionary measure, all the individuals visually identified as A. saxatilis or A. vaigiensis to the taxon Abudefduf cf. saxatilis/vaigiensis/troschelii. Clearly, this reevaluation does not affect those individuals conclusively identified through  Table S3. molecular analyses. Our revised list of sightings of Abudefduf spp. in the Mediterranean Sea is presented in Table S3 and the related map of occurrences in Figure 3. We must emphasize that the precautionary approach was also taken in few cases where all the characters (T1, T2 and T3) were lacking even though we did not find a case with the lack of all three characters in the images of Atlantic specimens (see Table S3).

New sightings of Abudefduf spp.
The video of the individual from the Adriatic Sea features an Abudefduf swimming at a depth of approx. 3-4 meters among a shoal of Chromis chromis and in the vicinity of large boulders which constitute an artificial pier. The length of the individual was approx. 10 cm TL. Another individual was observed in the lagoon of El Bibane, southern Tunisia, at a depth of approx. 0.5 m over a mixed bottom characterized by small rocks and dense vegetation composed mainly by the seagrass Cymodocea nodosa. The length of the individual was approx. 6 cm.
Both the Croatian and Tunisian individuals can be described as follows: oval shaped body, silvery-white belly and yellowish upper body with five prominent vertical black bars. The main morphological distinguishing characters (T1, T2 and T3) adopted in this study, are indiscernible from the images obtained in the Adriatic Sea since these were taken at a large distance from the fish. In the case of the Tunisian individual, T1 character is not visible in the photo, while T2 and T3 are absent. According to the provided descriptions, both the Croatian and the Tunisian individuals can be assigned to the taxon Abudefduf cf. saxatilis/vaigiensis/troschelii.

Discussion
Proper recognition of individuals belonging to the genus Abudefduf in the Mediterranean Sea is complicated because of the morphological similarities between A. saxatilis, A. vaigiensis and A. troschelii, which hamper conclusive identification of these species from underwater photos or videos as the ones revised here.
Here we tested the validity of the morphological characters used so far to distinguish A. saxatilis from A. vaigiensis, revealing a significant inconsistency in their applicability, which would impede the reliable identification of these species. In addition, the character indicated by Allen (1991) to identify A. saxatilis -namely, the fourth bar usually originates directly behind the last dorsal spine or is in contact with it -can be considered as unsuitable to distinguish with certainty the Atlantic A. saxatilis from its Indo-Pacific analogue A. vaigiensis. Indeed, this trait has been also observed in Abudefduf photographed in the Indo-Pacific area, which is the native region of A. vaigiensis. The same character is present in A. vaigiensis specimens barcoded by Tsadok et al. (2015) and by Vella et al. (2016a). Tsadok et al. (2015) reported morphological differences between A. saxatilis and A. vaigiensis -namely, extension of fifth bar to dorsal and anal fin and presence of two spots on the caudal peduncle, in agreement with the molecular identification of the species, but the study was conducted on a relatively small number of individuals collected from only two areas. The same author stated that the extension of the fifth vertical bar to the dorsal fin is also present in young A. vaigiensis specimens. Moreover, Randall (1995) presents a photo of A. vaigiensis from Oman exhibiting morphological characters usually used to identify A. saxatilis, including the two black spots on the caudal peduncle.
This evidence strongly supports the idea that the major external morphological characters used so far to distinguish A. saxatilis from A. vaigiensis are overlapping and cannot be used to distinguish with certainty these two species, confirming doubts which have been already expressed in the literature (e.g. Deidun and Castriota 2014;Osca et al. 2020). Indeed, Deidun and Castriota (2014) as well as Osca et al. (2020) provide records of Abudefduf cf. saxatilis, exhibiting a lack of taxonomical certainty and a high degree of caution, reflecting the rationale within G.R. Allen's pers. comm., that "there are no consistent meristic or morphometric differences between the two congeners". Alternatively, the results of our photo analysis could be explained by the existence of undocumented introduction events of A. saxatilis and A. vaigiensis in the areas of Persian Gulf (Iran and Oman) and possibly Mexico, respectively, allowing both species to be photographed. In addition, the possibility of hybridization between Abudefduf spp. congeners in the areas of sympatry, including the Mediterranean, should be taken into consideration. This scenario has been documented in the area of Hawaii, where recently introduced A. vaigiensis formed hybrids with native A. abdominalis (Quoy and Gaimard, 1825) which resulted in the appearance of intermediate phenotypes in successive generations (Coleman et al. 2014).
These conclusions have a significant impact on our understanding of Abudefduf spp. occurrences in the Mediterranean region. In a review provided by Osca et al. (2020), it is shown that great number of records of Abudefduf genus in the Mediterranean are based on observed or photographed specimens. This also relates to the possible confusion with a third species A. troschelii (Gill, 1862) never reported to date in the Mediterranean Sea. This species, distributed in the Eastern Pacific from the Gulf of California to the northern Peru (Froese and Pauly 2020), can be distinguished by A. saxatilis and A. vaigiensis only by the presence of scales on the inner side of the pectoral-fin base (Allen 1991).
Until the issues highlighted in our study are fully resolved, we suggest that all the identifications of A. saxatilis and A. vaigiensis based on visual evidence assessment (including photos and/or videos) could represent either A. saxatilis, A. vaigiensis or A. troschelii. This consideration will imply a more cautious taxonomical assignment of A. vaigiensis occurrences reported from Italy, Israel and Libya (Tardent 1959;Goren and Galil 1998;Vacchi and Chiantore 2000;Osca et al. 2020) and of A. saxatilis from Spain, Malta and Turkey and Slovenia (Azzurro et al. 2013;Vella et al. 2016a;Bilecenoğlu 2016;Lipej et al. 2019). This taxonomical reassignment does not affect those identifications based on molecular evidence. This is the case of A. vaigiensis spotted in Lebanon, Malta and Libya (Bariche et al. 2015;Vella et al. 2016a;Osca et al. 2020) and A. saxatilis reported by Tsadok et al. (2015) along the Israel coasts. Therefore, we can still affirm that both A. vaigiensis and A. saxatilis are present in the Mediterranean Sea.
Noteworthy, Abudefduf spp. have been mostly spotted in places where ship transports or aquarium releases are likely to occur (Zenetos et al. 2017) and multiple introductions can be envisaged for these species. Regarding the status of Abudefduf species in the Mediterranean, it has been claimed that Malta hosts an established population of A. saxatilis and A. hoefleri (Vella et al. 2016b;Deidun et al. 2016), whilst according to Tsadok et al. (2015), large populations of Abudefduf sp. would exist along the Israeli coasts. Finally, the occurrence of a single individual of Abudefduf sexfasciatus has been recently reported by Giovos et al. (2018) in Greece.
This study represent the first record of Abudefduf cf. saxatilis/vaigiensis/ troschelii in Tunisia and its second documented record from the Adriatic Sea. The first Adriatic record of the genus Abudefduf was based on a record of A. saxatilis based on the visual inspection of photographs of a single specimen observed in 2019 in the North Adriatic (Gulf of Trieste, Italy) and recently reported by Lipej et al. (2019).
The most probable vector of Abudefduf cf. saxatilis/vaigiensis/troschelii introduction into the Adriatic Sea is shipping activity. This hypothesis is supported by the fact that members of the family Pomacentridae have been often found to shelter under large ships and could be transferred via ballast waters (Vella et al. 2016a and references therein). Also, aquarium-related introductions could be considered as plausible, especially releases or escapes from aquaria on cruise liners (Vella et al. 2016a and references therein). In our case, the individual was observed near a large passenger harbor (less than 1 NM away), which hosts many cruise ships during summer. Occurrences of other alien fish species have already been reported from the area (e.g. Dulčić and Dragičević 2013a, b;Dulčić et al. 2010). The Tunisian individual was detected in El Bibane lagoon, which is located not far from a commercial port. Therefore, it is plausible to assume that this individual may have been transported via vessel traffic, which is the case for most non-indigenous species reported from the Gulf of Gabès (Ounifi-Ben Amor et al. 2016). Alien species, before spreading further, enter the lagoon which might provide favourable trophic conditions for their growth and reproduction.
The correct identification of non-indigenous species is of primary importance, especially when taxonomical inspections are limited to the analysis of underwater videos or photos. Considering the important contribution of divers, fishers (e.g. Azzurro et al. 2018b) and social networks (Langeneck et al. 2017;Azzurro et al. 2013;Giovos et al. 2019) in monitoring marine bio-invasions, the existence of valid characters to unambiguously identify these species is an essential pre-requisite for proper use of this information, either at the national, European or Mediterranean level. We therefore keep encouraging documentation of species occurrences through underwater photos and videos, especially in the light of marine bioinvasions where such information proved to be essential in number of instances. In our cases, characters previously considered as taxonomically valid were here demonstrated to be insufficient to reach the species-level identification with certainty. Molecular analysis combined with detailed morphological descriptions and with quality images will be therefore extremely useful for documenting the future Mediterranean occurrences of the species of Abudefduf genus considered in this study and to reach a certain identification.