Distribution of endangered Italian gudgeon Romanogobio benacensis (Cypriniformes, Cyprinidae, Gobioninae) with remarks on distinguishing morphological characters

Abstract Distribution data on many freshwater fish species in Croatia are scarce and species identifications are difficult, requiring further detailed studies. This paper presents a report of the Italian gudgeon Romanogobio benacensis from the Mirna River in the Istra Peninsula in Croatia, in the south-east from its previously known distribution range. The identification of R. benacensis in Croatia was supported by a morphological comparison with R. benacensis from Italy and Slovenia, the common gudgeon Gobio gobio, and the Danubian gudgeon Gobio obtusirostris from geographically close locations. A combination of character states (number of scales between anus and anal-fin origin, branched pectoral-fin rays, lateral-line scales, total, abdominal, and caudal vertebrae, and the size and number of lateral blotches) distinguishes R. benacensis from both G. gobio and G. obtusirostris. The phylogenetic analyses using mitochondrial sequences of cytochrome b gene confirmed that specimens from the Mirna River belong to R. benacensis. Also, Reka River system (Adriatic Sea basin) in Slovenia is inhabited by a possibly introduced Danubian gudgeon, G. obtusirostris, and not by R. benacensis.


Introduction
The richness of Croatian freshwater ichthyofauna manifests in at least 147 native fish and lamprey species, many of which are endemic (Mrakovčić et al. 2006, Jelić et al. 2008, Jelić 2011b. The number of fish species in Croatian freshwater environments is continuously increasing as a result of new species descriptions (Zupančič and Bogutskaya 2002, Marčić et al. 2011, Bogutskaya et al. 2012) and re-discovery of the previously described species (Jelić 2011a, Jelić andJelić 2015). For example, Telestes miloradi Bogutskaya, Zupančič, Bogut & Naseka, 2012, an endemic species whose description is based on material deposited in a museum, collected more than 100 year ago, and which had been considered extinct, was recently re-discovered in nature (Jelić and Jelić 2015).
Another example is the Italian gudgeon Romanogobio benacensis (Pollini, 1816), which was firstly recorded in Croatia in 2011 (Jelić 2011a). This cyprinid fish species, belonging to Palearctic subfamily Gobioninae, was originally described as Cyprinus benacensis from specimens collected in Lake Garda in the Po drainage (Italy). Later, the Italian gudgeon was considered a subspecies of the common gudgeon Gobio gobio (Linnaeus) (Bianco and Taraborelli 1984, Bianco 1988, Pizzul et al. 1993, Bănărescu et al. 1999 or a valid species Gobio benacensis (Kottelat 1997, Bianco and Ketmaier 2001, Kottelat and Persat 2005. Currently, the species is assigned to Romanogobio Bănărescu (Kottelat andFreyhof 2007, Zupančič et al. 2008). Systematic position of the Italian gudgeon within Romanogobio is supported by phylogenetic reconstructions using mitochondrial DNA (mtDNA) sequences of genes coding for cytochrome b (cytb) (Bianco and Ketmaier 2005) and the cytochrome c oxidase subunit I (COI) (Geiger et al. 2014). However, the basal node in the Romanogobio clade which shows divergence between R. benacensis and the remaining subclades was not supported, thus preventing Bianco and Ketmaier (2005) to consider Romanogobio as a supported clade in comparison with the Gobio Cuvier clade.
Based on some diagnostic morphological characters, the Italian gudgeon is considered more similar to Gobio than to Romanogobio (Bianco and Ketmaier 2005, Kottelat and Freyhof 2007, Bianco 2014. According to Bianco and Ketmaier (2005), a single character discriminating R. benacensis and G. gobio was the number of scales between the anus and the anal-fin origin, 2-4 in the former species and 4-8 in the latter. Kottelat and Freyhof (2007) added that in R. benacensis the distance between the anus and the anal-fin origin is distinctly smaller than the eye diameter, while in Gobio it is equal to or greater than the eye diameter. Further, in R. benacensis the scales on the abdomen extend only to a point between the pectoral and pelvic-fin bases, while in Gobio they sometimes extend to a level of the posterior end of the pectoral-fin base. However, systematic position of R. benacensis is controversial, since some of the morphological characters of R. benacensis do correspond to those diagnostic of the genus Gobio, while others are typical for Romanogobio (Kottelat and Freyhof 2007).
For nearly two centuries after its description, R. benacensis was considered an Italian endemic species, native in the Padano-Venetian district from the Isonzo River in the north to the Marecchia River in the south (e.g. Bianco and Taraborelli 1986, Bianco 1991, Bianco and Ketmaier 2005. First finding of Italian gudgeon outside of Italian territory was reported by Povž et al. (2005) in the lower reaches of Vipava River in the Soča (Isonzo) drainage (Slovenia). Crivelli (2006) cited a personal communication by M. Povž that R. benacensis was also found in the Reka River in the Adriatic basin in Slovenia. This was published later by Zupančič (2008) and Povž et al. (2015). However, no diagnostic characters of specimens from the Reka River were given to support this identification. Kottelat and Freyhof (2007) presumed that the Italian gudgeon probably occurs elsewhere in the northern Adriatic basin. Out of the native range, the Italian gudgeon was introduced and established in the Arno, Tiber, and Ombrone rivers in central Italy (Bianco 1994, Bianco andKetmaier 2005).
In recent years, G. gobio was introduced in Italy and became invasive species in river systems down to the Badolato River in the south, making a serious threat to R. benacensis (Bianco and Ketmaier 2005). Phylogenetic inferences on mtDNA sequences of cytb (Bianco and Ketmaier 2005) and COI (Geiger et al. 2014) genes showed that the examined specimens of G. gobio from the Po drainage shared identical haplotypes with G. gobio from the Rhône drainage. However, Bianco and Ketmaier (2005) and Bianco (2009) indicated possible introductions of Danubian G. gobio in Italy (which refers to the Danubian gudgeon Gobio obtusirostris Valenciennes according to the recent taxonomic concept). Gobio gobio and G. obtusirostris are larger-sized fishes  and, if successfully established, they might cause a considerable decline in populations, and even extirpation, of R. benacensis Ketmaier 2005, Bianco 2009). The latter species is thought to be represented only by genetically "pure" populations in its native range in the Tagliamento River in Italy (Bianco and Ketmaier 2005) and the Adriatic basin in Slovenia (Crivelli 2006, Bianco 2009. Romanogobio benacensis is considered an endangered species both globally (EN B2ab(i,ii,iii,iv,v), Crivelli 2006) and in Italy (Bianco et al. 2013).
Although four species of the subfamily Gobioninae (gudgeons) have been reported in Croatia , Freyhof and Kottelat 2007, Jelić 2011a, their systematic status is uncertain (e.g., species in the R. albipinnatus group sensu Freyhof and Kottelat 2007). Mustafić et al. (2005) reported G. gobio, G. uranoscopus (Agassiz), G. albipinnatus Lukasch, and G. kesslerii Dybowski. Later, these species were assigned to G. obtusirostris, the Danubian longbarbel gudgeon R. uranoscopus, the Danubian whitefinned gudgeon R. vladykovi (Fang), and the Kessler's gudgeon R. kesslerii, respectively (Freyhof and Kottelat 2007, Jelić 2011a. In the previous studies, gudgeons in the Istra Peninsula (Mirna River; Fig. 1) were identified as G. gobio obtusirostris (Leiner et al. 1995) (= G. obtusirostris) and G. gobio . The main aim of this study was to investigate which gobionine species occurs in the Istra Peninsula (Croatia) and to confirm the presence of R. benacensis in Croatia. Gudgeon individuals collected during an ichthyological survey in the Istra Peninsula was preliminary identified as R. benacensis by using morphological characters (Jelić 2011a). In the present paper, samples of R. benacensis from different localities in Italy, Slovenia, and Croatia were described using morphological and molecular characters to support their identification in comparison with G. obtusirostris and G. gobio.

Morphological analysis
Measurements were made according to Naseka and Freyhof (2004). All measurements were made point-to-point with a digital calliper and recorded to the nearest of 0.1 mm. Vertebrae counts are given according to Naseka (1996). Last two rays in dorsal and anal fins based on a single pterygiophore were counted as 1½ ray. Unbranched rays in dorsal and anal fins were counted from radiographs. In total, 30 morphometric indices were used for descriptions and statistical analyses as in Table 1 and 18 meristic characters as in Table 2 were examined. All characters were obtained from specimens of both sexes and combined in analyses and tables. A Mann-Whitney U Test and a Discriminant Function Analysis (DFA) were performed using STATISTICA v6.0 and

DNA extraction, gene amplification, and sequencing
Besides species identification based on morphological features, species were characterised using sequences of cytochrome b gene (cytb), which is commonly used mtDNA genetic marker for species affiliation of European cyprinids (e.g. Zardoya and Doadrio 1998, Bianco and Ketmaier 2005, Perea et al. 2010. Total DNA was extracted from pectoral fin tissue with the Qiagen DNeasy Blood and Tissue Kit (Qiagen, Germany) following the manufacturer protocol. After extraction, total genomic DNA was stored on -20°C until the polymerase chain reaction (PCR) was conducted. The primers used for cytochrome b were GluF and ThrR (Machordom and Doadrio 2001). The PCR was carried out with the HotStarTaq Master Mix Kit (Qiagen). PCR reactions were prepared in a total volume of 50 µL comprised of 2.5 U HotStarTaq DNA Polymerase, 1.5 mM MgCl 2 , 200 µM each dNTP, 0.2 µM of each primer and 20 ng of DNA template. The amplification process was conducted with the same conditions as described in Perea et al. (2010). Purification and sequencing of the PCR products were prepared by Macrogen Inc. (Seoul, South Korea) using the same primers used for gene amplification. Purified PCR products were sequenced on ABI 3730XL DNA Analyzer (Applied Biosystems, Foster City, USA). Sequence chromatograms were analysed using SEQUENCHER (version 5.3; Gene Codes Corp., Ann Arbor, USA) and aligned by eye.
The obtained sequences in this study (1141 base pairs long, bp) were examined using Nucleotide Basic Local Alignment Search Tool (Nucleotide BLAST; http://blast. ncbi.nlm.nih.gov/Blast.cgi) to screen for the most similar sequences in the GenBank nucleotide database (National Center for Biotechnology Information, U.S. National Library of Medicine, USA).
The Median-Joining (MJ) haplotype network (Bandelt et al. 1999) was used to infer the intraspecific relations in R. benacensis with cytb sequences obtained in this study and 342 bp sequences which were downloaded from the GenBank (Bianco and Ketmaier 2005). The MJ network was computed using PopART (Population Analysis with Reticulate Trees) v1.7 (Leigh and Bryant 2015).
Phylogenetic tree reconstructions were conducted using the cytb sequences obtained in this study and the available sequences belonging to Gobio and Romanogobio (1141 bp) from the GenBank (Briolay et al. 1998, Zardoya and Doadrio 1998, Madeira et al. 2005, Saitoh et al. 2006, Yang et al. 2006, Perea et al. 2010, Liu et al. 2010, Tang et al. 2011Table 5). Sequences originating from the tench Tinca tinca (Linnaeus), the European bitterling Rhodeus amarus (Pallas) and the stone moroko Pseudorasbora parva (Temminck & Schlegel) were used as an outgroup. Newly obtained sequences in this study were deposited in the GenBank under accession numbers shown in Table 5 (will be available for publication). Phylogenetic reconstructions were inferred using three methods (Maximum Likelihood -ML, Bayesian Inference -IB and Maximum Parsimony -MP). The best-fit evolutionary model used in ML and IB was computed using jModelTest2 (version 2.1.6; Darriba et al. 2012) with the Bayesian information criterion (BIC) as implemented on the Cipres Science Gateway (version 3.1; http://www.phylo.org; Miller et al. 2010). Best-fit model of nucleotide substitution was Generalised Time Reversible (GTR) (Tavaré 1986) with a gamma distributed rate variation among sites (+G) and a significant proportion of invariant sites (+I). The ML was run using RAxML-HPC2 Workflow on XSEDE (version 8.2.8; Stamatakis 2014) on the Cipres Science Gateway with optimized parameters. For ML analysis, 200 search replicates to find the ML tree and 1000 nonparametric bootstrap replicates under the GTRGAMMA model were applied. The IB was run in MrBayes 3.2 (Ronquist et al. 2012) on the Cipres Science Gateway. Two independent runs with four MCMC chains were run for 50 million generations and sampled every 5000 generations, with temperature parameter set to 0.2 and the first 12.5 million generations discarded as burn-in. The convergence of runs was screened using AWTY (Nylander et al. 2008) while effective sample sizes of parameters were checked using TRACER 1.5 (Drummond and Rambaut 2007). The MP analysis was performed in MEGA 6.06 (Tamura et al. 2013). The MP tree was obtained using the Subtree-Pruning-Regrafting algorithm (Nei and Kumar 2000) with search level 1 in which the initial trees were obtained by the random addition of sequences (10 replicates). Nodes in phylogram which have bootstrap values P ≥ 70 in ML and MP, and posterior probabilities (pp) values ≥ 0.95 in IB were considered supported.

Comparative morphological description of R. benacensis from the Mirna River
Comparative morphological analysis of R. benacensis (n = 23), G. gobio (n = 2) and G. obtusirostris (n = 24) was performed based on number of available specimens. See Fig. 2a for general appearance and Table 1 for morphometric data. Below, only those characters demonstrating some difference between the species are discussed.
The anus is located close to the anal-fin origin, the distance between the anus and the anal-fin origin is 3-5, averaging 3% SL (3-6, averaging 4% SL, in the Po samples) in contrast to 5-10, averaging 7% SL in G. obtusirostris where the anus is usually located about the midway between the pelvic and anal-fin origins ( Table 2). The number of scales between the anus and the anal-fin origin is 1-5, commonly 2-4, in R. benacensis vs. 4-7 in G. obtusirostris and 6-9 in G. gobio.
The dorsal fin has 4 unbranched rays in all specimens from Mirna and Soča rivers and in 14 (of 19) specimens from the Po drainage (Table 2, Fig. 3). This is the first known example of a species with commonly 4 unbranched dorsal-fin rays in the Gobioninae.
Contrary, in G. obtusirostris and G. gobio, the number of unbranched dorsal-fin rays is always 3. In all species examined in this study the dorsal fin has 7½ branched rays and the anal fin has 3 simple and 6½ branched rays. The number of branched pectoral-fin rays is 12-15 in R. benacensis vs. 15-18 in G. obtusirostris and G. gobio (Table 2).
In R. benacensis from the Mirna River, scales along the midline of the belly extend forward to the middle of the pectoral-fin base. In R. benacensis from the Po drainage, scales along the midline of the belly extend forward from much behind the pectoral-fin base to the anterior end of the pectoral-fin base, commonly to the posterior end of the pectoral-fin base (Table 3). In G. gobio, the breast and throat are more scaled; scales along the midline of the belly extend forward from the middle of the pectoral-fin base to a point in front of the anterior end of the pectoral-fin base. In G. obtusirostris, the throat and breast are less scaled similarly to R. benacensis; scales along the midline of the belly extend forward from a point much behind the pectoral-fin base to the anterior end of the pectoral-fin base ( Table 3).
The lateral line is complete, with 37-39 total scales averaging 38.6 (36-40 averaging 38.8 in the Po samples). These counts are lower than in G. gobio and G. obtusirostris which have a range of 39-42 scales, and averages of 42.0 and 41.3, respectively. Other scale counts can be also seen in Table 2. No epithelial keels on scales were found in specimens of R. benacensis.
The barbel is reaching the vertical through the middle of the pupil to the posterior margin of the eye (Table 4). Similar character states are also typical to R. benacensis from the Po drainage: the barbel is reaching the vertical through the middle of the pupil to behind the posterior margin of the eye, more frequently between the posterior margin of the pupil and the posterior margin of the eye. On average, the barbel is longer in R. benacensis than in G. gobio and G. obtusirostris (30-39% HL, averaging 35 in R. benacensis from the Mirna River and 26-37% HL, averaging 32.5 in R. benacensis from the Po drainage vs. 22-28% HL, averaging 25 in G. obtusirostris from the Reka River and 24-34% HL, averaging 27.5 in G. obtusirostris from the Danube drainage) ( Table 1). In G. obtusirostris, the barbel is commonly reaching the vertical through the anterior margin of the pupil to the posterior margin of the pupil (Table 4).
Laterally (examined on both sides) with 7-8, usually 7 roundish dark blotches. The size of a blotch varies but it is relatively large: the size of the blotch below the dorsal-fin origin is about (close to) or larger than the horizontal eye diameter. The same pattern of the blotches, 5-8, usually 6 or 7, is found in the examined specimens from the Po and Soča drainages. In G. gobio and G. obtusirostris, the blotches are smaller and more numerous, 7-11, usually 8 or 9, and the size of the blotch below the dorsal-fin origin is about the half horizontal eye diameter (Fig. 2).

Statistical analysis
A Mann-Whitney U test revealed seven morphometric and nine meristic characters different on a statistically significant (0.01%) level between the samples of typical G. obtusirostris (Danube specimens) and typical R. benacensis (Po and Adige specimens): the body depth at the dorsal-fin origin, the distance between the pelvic fin and the anal-fin origin, the distance between the anus and the anal-fin origin, the dorsal-fin length, the anal-fin length, the the interorbital width, the barbel length, the number of unbranched dorsal-fin rays, the number of branched pectoral-fin rays, the numbers of scales in lateral series, total lateral-line scales and lateral-line scales to the posterior margin of hypurals, the number of circumpeduncular scales, the number of scales between the anus and the anal-fin origin, and the numbers of total and abdominal vertebrae.
These 16 distinguishing characters were used for a DFA in order to classify the Reka and Mirna samples into one of the two species. DFA statistics values are as follows: Wilks' Lambda 0.00721, approx. F (45, 78) =7.3894, p<0.0000. The Mirna specimens are the closest to Italian R. benacensis (Fig. 5)

Phylogenetic tree inference
Two unique cytb haplotypes were detected in four specimens from the Mirna River (Table 5) Haplotype 1 originates from three specimens (HDBI 1323/tissue 771, HDBI 1292/tissue ID 772, and HDBI 1292/tissue ID 773) whereas Haplotype 2 was observed in one specimen (HDBI 1292/tissue ID 774). Haplotype 1 and Haplotype 2 have 99% similarity score (1136 identical nucleotide positions in 1141 bp sequence alignment). Nucleotide BLAST search using Haplotype 1 resulted in 99% similarity score (337 identical nucleotide positions in 342 bp alignment) with GenBank entry AY641522 designated as "strain TAG" of G. benacensis in Bianco and Ketmaier (2005), validating morphological determination of R. benacensis in this study. The second top match in BLAST search using Haplotype 1 was GenBank entry AY641524 designated as "strain OMB" of G. benacensis in Bianco and Ketmaier (2005) with similarity score 328/342 (96%). The 342 bp sequence alignment of R. benacensis were analysed with the MJ haplotype network (Fig. 6). Haplotypes 1 and 2 from Croatia differ by five mutational steps from the "strain TAG" and can be considered as members of this strain. There are nine mutational steps between "strain TAG" and "strain OMB", confirming there are two strains in R. benacensis (Bianco and Ketmaier (2005). The sequences obtained in Bianco and Ketmaier (2005) (342 bp) were not used in further phylogenetic tree reconstruction in this study to avoid inclusion of significant proportion of missing sites in the final sequence alignment (1141 bp). The ML, IB and MP provided congruent trees with no supported contradictions (Fig. 7). Results of phylogenetic reconstruction indicated Romanogobio and Gobio as two statistically supported clades (P (ML) = 70, pp (IB) = 0.97, P (MP) = 91, and P (ML) = 100, pp (IB) = 1, P (MP) = 100, respectively). High statistical support was also observed for the node showing divergence between these two clades (P (ML) = 93, pp (IB) = 1, P (MP) = 80). Results of phylogenetic inference showed that R. benacensis belongs to the clade of Romanogobio (Fig. 7). All topology tests (Table 6) indicated that best topology is presented in "Tree 1: R. benacensis sister taxon for R. kesslerii and R. banaticus" vs. "Tree 2: R. benacensis sister taxon for all Romanogobio" and "Tree 3: R. benacensis sister taxon for all Gobio".

Discussion
Morphological data in this study confirm observations of the previous authors (Bianco and Taraborelli 1984, Bianco and Ketmaier 2005, Kottelat and Freyhof 2007 that R. benacensis differs from G. gobio and G. obtusirostris by a shorter distance between the anus and the anal-fin origin. However, this character is not completely discriminating in this study -there are specimens of both R. benacensis (including all examined specimens from the Soča River) and Gobio with 4 or 5 scales between the anus and the anal-fin origin. However, the length of the distance between the anus and the anal-fin origin is still diagnostic: in R. benacensis, the distance between the anus and the anal-fin origin (3-5% SL) is smaller than the eye diameter (6-8% SL) while in G. gobio and G. obtusirostris this distance (5−10% SL) is equal or larger than the eye diameter (5-7% SL). Kottelat and Freyhof (2007) also stated a difference in the scale pattern on the abdomen in R. benacensis; the scales extend only to a point between the pectoral and pelvic-fin bases vs. a level of the posterior end of the pectoral-fin base in Gobio. However, our data (Table 3) did not confirm this character to be clearly diagnostic for the two taxa.
Some of the morphological characters of R. benacensis do correspond to those diagnostic of the genus Gobio. As shown by Kottelat and Freyhof (2007), R. benacensis does not have epithelial crests on scales on the dorsal surface of the body, a shallow caudal peduncle, and a long distance between the anus and the anal-fin origin, the characters typical for Romanogobio. As to the position of the anus (in relation to anal- Table 5. List of species used for phylogenetic tree inference on cytb sequences. Data on catalogue numbers of analysed specimens, localities, GenBank Accession numbers, sequences lengths, and references are shown.

Species
Catalogue no. Locality fin base), it is closer to the anal fin in R. benacensis than even in Gobio; a position of the anus at (or close to) the anal-fin origin is a plesiomorphic feature in the Gobioninae (Naseka 1996). The genus Romanogobio also differs from Gobio in having the supraethmoid wide (wider than long) vs. elongated (longer than wide); a high and oval second infraorbital (Fig. 4a) vs. narrow and rod-shaped (Fig. 4b), and a relatively high number of vertebrae both total and in the regions (the modal vertebral formula 42:(11)20(5)+(3)21(18) vs. 39:(11)21 (5)+(1)18(17) (Naseka 1996, Naseka et al. 2002. As shown above, R. benacensis possess low vertebral counts which are even lower than those found in the examined Gobio species. But the shape of the supraethmoid and infraorbitals are similar to that typical of Romanogobio. As shown above in the results of the morphological analysis, the examined sample from the Reka River is G. obtusirostris, not R. benacensis. This was not expected having in mind the current hydrological features of the Reka River which drains into the Adriatic Sea. The Reka River (Notranjska Reka) originates in Croatia and flows 54 km through western Slovenia, disappears in the Škocjanske jame underground cave system and reappears again after 38 km as a part of the Timavo River in Italy flowing into the Adriatic Sea. Although results in this study indicates that R. benacensis is not present in the Reka River, further analyses using more specimens are needed for a final systematic conclusions. Also, the native status of G. obtusirostris in the Reka River is unclear. It could be a non-native species similar to an introduced chub Squalius cephalus (Linnaeus).
The morphological data discussed above confirmed that gudgeons from the Mirna River can be assigned to R. benacensis.
Phylogenetic reconstruction in this study (Fig. 7) indicated Gobio and Romanogobio as two statistically supported clades; with R. benacensis as a member of Romanogobio clade. Although results of Bianco and Ketmaier (2005) and Geiger et al. (2012) questioned the phylogenetic recognition of Romanogobio as a distinct clade in respect to Gobio, the monophyletic status of Romanogobio was shown by Madeira et al. (2005) and in comprehensive studies on Gobionine phylogeny (Yang et al. 2006, Tang et al. 2011. Difference in statistical supports for clades Romanogobio and Gobio among studies in which cytb was applied Table 6. Comparison of user trees ("Tree 1: Romanogobio benacensis sister taxon for Romanogobio kesslerii and Romanogobio banaticus", "Tree 2: R. benacensis sister taxon for all Romanogobio", and "Tree 3: R. benacensis sister taxon for all Gobio"). The columns show the results and p-values of the following tests: 1sKH -one sided KH test based on pairwise SH tests (Shimodaira and Hasegawa 1999, Goldman et al., 2000, Kishino and Hasegawa 1989; SH -Shimodaira-Hasegawa test (1999); ELW -Expected Likelihood Weight (Strimmer and Rambaut 2002); 2sKH -two sided Kishino-Hasegawa test (1989). Plus signs denote the confidence sets. Minus signs denote significant exclusion. All tests used 5% significance level. 1sKH, SH, and ELW performed 1000 resamplings using the RELL method. 1sKH and 2sKH are correct to the 2nd position after the the decimal point of the log-likelihoods.  Geiger et al. (2012) could be less informative than a longer cytb fragment used in this study. A better resolution is expected using longer sequences in phylogenetic reconstructions although non-hierarchical relations can characterize some phylogenies no matter of the length of used sequences (Strimmer and von Haeseler 1997). The Mirna population represents a single area of occurrence of R. benacensis out of its known distribution range in the south-east of the Soča (Isonzo) drainage. If the species is native in the Mirna River, this is the only occurrence of a native species belonging to the genus Romanogobio and the subfamily Gobioninae along the Croatian section of the Adriatic coast (the Dalmatia freshwater ecoregion sensu Abell et al. 2008). It may be an evidence of past connections between Istrian rivers and the paleo-Po drainage. A similar "paleo-Po" distribution is reported for the triotto Leucos aula (Bonaparte), the Padanian barbel Barbus plebejus Bonaparte, the alborella Alburnus arborella (Bonaparte) and some other fish species (Kottelat and Freyhof 2007), also for an amphibian, the Italian agile frog Rana latastei Boulenger (Gasc et al. 1997) and a freshwater decapod crustacean, the white-clawed crayfish Austropotamobius pallipes (Lereboullet) (Jelić et al. 2016b). However, a human mediated translocation of R. benacensis cannot be excluded; rather it is a supposition to be further investigated. For example, G. obtusirostris in the Ričica River (Lika Region, Adriatic basin) probably originates from the Danube drainage (Jelić et al. 2016a).
Only four specimens of R. benacensis were collected in spite of an intensive sampling effort. Since no other gudgeon species was reported in the Mirna River, an ongoing population extirpation by competition (e.g. with G. gobio as reported by Bianco and Ketmaier (2005) in Italy) can be excluded as a reason for low population density. Nevertheless, competition with non-native cyprinid species such as P. parva, or a pre- dation by allochthonous piscivorous fish such as the pike Esox lucius (Linnaeus) or the pike-perch Sander lucioperca (Linnaeus) should not be excluded.
Therefore, having in mind that small-sized populations are more prone to extirpation due to genetic drift, extinction vortex, etc., it is necessary to implement systematical monitoring on the present R. benacensis population, accompanied with a more intense sampling in order to reveal possible remaining populations and to characterize the gene pool of this endangered species, both crucial issues for further management and conservation. Vernacular name. Romanogobio benacensis does not have any Croatian name as it has been only recently discovered in Croatian national territory. We offer "Talijanska krkuša" as its Croatian name, which originate from translation of vernacular name on English (the Italian gudgeon) and Italian (il gobione Italiano).