Table 4 contains data on a comparison of smaller-sized (SL < 130 mm) and larger-sized (SL > 130 mm) specimens per phenotype. Significantly size-related (p < 0.01000) are 18 characters in S. microlepis phenotype 1 and 22 characters in S. microlepis phenotype 2. Shared size-related characters are as follows: dorsal fin depth (% SL), anal fin depth (% SL), head length (% SL), head length (% body depth), head depth at nape (% HL), snout length (% HL), eye horizontal diameter (% SL), eye horizontal diameter (% HL), eye horizontal diameter (% interorbital width), interorbital width/eye horizontal diameter, snout length/eye horizontal diameter, head depth at nape/eye horizontal diameter, pectoral fin length/pectoral – pelvic-fin origin distance, predorsal length/head length. Head depth at nape and snout length increase with size while anal- and dorsal-fin depth, head length, and eye diameter decrease.

The two phenotypes are readily distinguished (phenotype 1 vs. phenotype 2; external characters on an example of middle-sized specimens see Fig. 5; Tables 2–4) by the following combinations of character states:

1 number of gill rakers: (13)14–16 (15 in lectotype of S. microlepis), mean 15.0 vs. 11–14, mean 12.6;

2 total vertebrae: commonly 42 (24+18 or 25+17) and 43 (25+18) vs. commonly 43 (24+19);

3 dorso-hypural distance: commonly falling behind the posterior eye margin (at a considerable distance from the eye margin in large-sized specimens as can be also seen in Kottelat and Freyhof (2007: figure on page 269) vs. commonly falling into the middle or posterior half of the eye when reported forward;

4 the back: usually a well pronounced discontinuity behind the head (even in small-sized individuals), a straightened back profile and the maximum body depth located just behind the head vs. smoothly convex lacking a prominent hump behind the head and the maximum body depth located at or slightly in front of the dorsal-fin origin;

5 maximum body depth: the body deepest at a vertical closer to the head than to the dorsal-fin origin and, respectively, maximum body depth exceeds 1.05–1.20 times body depth at the dorsal-fin origin vs. about equal to body depth at the dorsal-fin origin;

6 length of lower jaw (% interorbital width): 121–155% (mean 134.5%) vs. 100–121% (mean 111%); length of lower jaw (% cranium width) 95–118% (mean 105% vs. 81–101% (mean 90%);

7 head length (% SL): 29–34% (mean 31%) vs. 25–30% (mean 27%); the ranges do not overlap in larger-sized specimens (SL > 130 mm; Table 3);

8 head depth at nape (% HL) in larger-sized specimens (SL > 130 mm; Table 3): 54–61% (mean 59%) vs. 59–69% (mean 63%);

9 the upper head profile: straight vs. commonly slightly convex behind the eyes.

Besides these characters, Fig. 5B illustrates a specimen of S. microlepis phenotype 1 having the upper jaw not projecting beyond the lower jaw, the lower jaw-quadrate junction lying on the vertical through the middle of the eye, and a prominent ‘angle’ formed by the posterior end of the lower jaw; and the mouth cleft is long, straight and oblique. Phenotype 2 (Fig. 5A) is commonly characterised by the upper jaw clearly projecting beyond the lower jaw and including the tip of the lower jaw; the lower jaw-quadrate junction located about at a vertical through or slightly in front of the anterior margin of the pupil; the lower jaw posterior end not forming a prominent angle; and the mouth cleft slightly curved and more horizontal.

Comparison of the two phenotypes of S. microlepis

I A DFA based on counts and standardised direct measurements (Fig. 7A) support 100% discrimination for both groups. DFA statistics values are as follows: Wilks’ Lambda 0.5525, appr. F (19.64) = 60.297, P < 0.0000. In this analysis, the lower jaw length, number of gill rakers, head length, upper caudal-fin lobe and maximum head width contribute most to the discrimination between the phenotypes.

Figure 7. 

DFA performed for two combined samples of Squalius microlepis phenotype 1 and phenotype 2 A based on 32 standardised direct measurements and 12 counts B based on 52 proportional measurements (as in Table 2) and counts. Specimens A–K in Table 5 not included.

II A DFA based on counts and relative measurements (as in Table 2) (Fig. 7B) also support 100% discrimination for both groups (Wilks’ Lambda 0.04411, appr. F (23.63) = 59.369, P < 0.0000), and the most contributing characters are the number of gill rakers, interorbital width (% HL), ethmoid width (% pterotic cranial width), prepelvic length (% SL), and head length (% SL). A CA run for the same set of characters support perfect clusters into two groups (Fig. 8).

Figure 8. 

A CA performed for two combined samples of Squalius microlepis phenotype 1 and phenotype 2, based on 52 proportional measurements (as in Table 2) and counts. Specimens A–K in Table 5 not included.

Taken together, these two analyses based on differently approached characters, clearly support the primary observations on most influential characters for distinguishing the two phenotypes (1, 6, 7 above): gill rakers count, head length, and length of lower jaw.

Classification of selected specimens A–K between the two phenotypes of S. microlepis

Character data for specimens A to K are presented in Table 5.

I A DFA classification (posterior probabilities and classification functions) based on counts and direct standardised measurements classify these specimens as follows: specimens B, F, and G are identified as phenotype 2 while others as phenotype 1 (Table 6).

II A DFA analyses based on counts and proportional measurements (as in Table 2) (posterior probabilities and classification functions) unambiguously classified specimens C, E, F (Fig. 6A), G–K as phenotype 2. Classification of specimens A and B is variable and classification of specimen D (Fig. 6B) as phenotype 1 is lower than for other specimens. In a DFA scatter plot (Fig. 9C) they are located between phenotypes 1 and 2.

Figure 9. 

DFA performed for three combined samples, Squalius tenellus, S. microlepis phenotype 1 and phenotype 2 A based on 32 standardised direct measurements and 12 counts (specimens A–K excluded) B based on 52 proportional measurements (as in Table 2) and counts (specimens A–K excluded) C same analysis as (B) but specimens A–K in Table 5 included.

So, the historical NMW sample from poljes at Vrgorac includes both phenotypes of S. microlepis. In the Tihaljina-Trebižat kartic system, most specimens were phenotype 2 while two specimens were clearly classified as phenotype 1 (Fig. 9C).

Discrimination of S. tenellus and two phenotypes of S. microlepis

I A DFA performed for three groups of samples (S. tenellus, S. microlepis phenotype 1 and S. microlepis phenotype 2 based on standardised measurements and counts; Fig. 9A) showed a perfect (100%) classification of all three groups (DFA statistics values: Wilks’ Lambda 0.00660, approx. F (48.172) = 40.519, P < 0.0000). The lower jaw length, the number of gill rakers, and the number of scales above the lateral line contribute most to the discrimination between the phenotypes. The closest are two phenotypes of S. microlepis and the most distant are S. tenellus and S. microlepis phenotype 2 (squared Mahalanobis distance equals 52.23712 and 92.95126, respectively).

II A DFA performed for the same set of samples but based on the proportional measurements and counts (Fig. 9B) also showed a perfect (100%) classification (DFA statistics values: Wilks’ Lambda 0.00668, approx. F (44.176) = 44.941, P<0.0000). Number of gill rakers, number of scales above the lateral line, number of latera-line scale, maximum head width and maximum cranium width contribute most to the discrimination between the three groups. The closest are two phenotypes of S. microlepis and the most distant are S. tenellus and S. microlepis phenotype 2 (squared Mahalanobis distance equals 57.98632 and 84.69049, respectively). When specimens A to K are included into a DFA analysis, specimens A, B, and D are closely located to each other in the morphological space and intermediate between the two phenotypes (Fig. 9C). Specimens F and G lie within the phenotype 2 while specimens C, E, and H–K lie within the phenotype 1.

Ričina-Prološko Blato-Vrljika karst system

The detailed map of this area at the border between Croatia and Bosnia and Herzegovina, its hydrographic networks, position of main discharge gauging stations and supposed groundwater flow directions are presented by Bonacci and Roje-Bonacci (2008: fig. 1) and Bonacci et al. (2013a: fig. 1). We only found individuals of the phenotype 1 in this karst drainage.

All examined specimens from the Ričina-Prološko Blato-Vrljika karst system belong to S. microlepis phenotype 1. The NMW labels and acquisition information for the syntypes (lectotype and paralectotypes by Bănărescu and Herzig-Straschil (1998: 417)) say only ‘Imosky, Kroatien (Dalmatien), Heckel Reise 1840’ (as well as some other NMW sample, Table 1) (“Gewässer von Imosky” in the original description (Heckel 1843: 52(1042)). We suppose that the syntypes came, most probably, from Prološko Blato, which is a large swampy region in the north-western part of Imotsko Polje in modern Croatia, named after the town of Imotski (also called Imotski field, or valley, or Imotsko-Bekijsko Polje because the Herzegovinian part of the valley is called Bekija). In 19^th century, Prološko Blato was part of the year under water, and just one small part was flooded during the whole year (Proložac, or Prološko Lake). The species also occured in three other lakes close to Prološko Blato: Galipovac, Lokvičić and Knezovića lakes (A. Mikulić, pers. comm. 7 May 2008). For the first time S. microlepis was reported in the Vrljika by Katurić (1883) but it is not known how far downstream the Vrljika–Matica River this species was distributed in the past. The Vrljika originates by a spring (izvor) east of Prološko Lake and is at present connected to this lake via canal Sija. The historical NMW sample (1901) from the Vrljika is numerous and contains individuals up to 217 mm SL. Recent samples of S. microlepis collected by PZ and DJ in Imotsko Polje are only from Prološko Lake itself, at the inflow of the canal that connects it to the Vrljika. Information from local fishermen (more than ten years ago) indicates that ‘masnica’ had been rarely found in streams of Imotsko Polje but was very abundant in the lake. Further upstream, northwards from Imotsko Polje, S. microlepis occured in Ričice Reservoir, a transboundary accumulation lake constructed in the valley of the Ričina River at its confluence with the Vrbica River. It was also found by DJ and PZ in the Ričina River around Posušje (at the village Vir) and in Tribistovo Reservoir north of Posušje (built on a small tributary to the Ričina) in Bosnia and Herzegovina. However, it may be not native there: in 2008, local fishermen claimed that it had been introduced to the Ričina and the Tribistovo Reservoir from Imotski. At present, it is extremely rare in the entire Imotski area (based on the local population surveys). We failed to collect it in both Imotsko Polje and the Ričina River in 2017–2019.

There are also no recent records of any findings of a small-scaled Squalius downstream the Vrljika–Matica at present days, but S. microlepis phenotype 1 inhabits a small karstic lake, Krenica, which is located in the south of the Drinovci hill and is fed by underground waters of the Vrljika-Matica. So, it appears that Krenica Lake, populated by S. microlepis phenotype 1 and the lower reaches of the Vrljika-Matica, populated by phenotype 2, are the closest known localities of the ranges of the two phenotypes.

Some indications in literature allow to assume that S. microlepis of the Imotski area is a lacustrine species rather than a riverine one. Karaman (1928: 159–160) indicated that S. microlepis microlepis prefers ‘calm’ water and was found in Prološko Lake but not in the Vrljika River stream. All individuals ever observed by the authors of this study in the Imotski area were from Prološko Blato. Outside Imotski, Karaman (1928: 159) mentioned that Kolombatović (without an exact reference) had found this fish in Baćina lakes in lower reaches of the Neretva, in stagnant waters only, and never in the Neretva stream. Squalius microlepis (as Leuciscus turskyi microlepis) was considered as a lacustrine species by Vuković and Ivanović (1971: 150–151).

Matica-Tihaljina-Trebižat karst system

All specimens examined except two found in the karst river system of Matica-Tihaljina-Trebižat of the Neretva drainage belong to S. microlepis phenotype 2. The most upstream locality is the lower reaches of the Vrljika-Matica and the Grude Canal at its confluence with the Matica at Drinovci; this locality is close to the terminus of the river. The Vrljika-Matica originates in the northwest of Imotsko-Bekijsko Polje in Croatia. In natural conditions, the river used to go underground in a ponor (swallow hole, or sinkhole) south of the Drinovci hill, now it is accumulated in a lake, and water passes through a tunnel to the Tihaljina River some 150 m below, where a small electric power plant is constructed. The Tihaljina comes from underground very close to this point at the foot of the Jagodnica Mountain south of Drinovci as a strong karst spring, which is a continuation of the Matica underground stream (Bonacci et al. 2013b). It goes southeast to Klobuk Mountain and the spring Klokun, where it changes its name to Mlade, and from Humac to the confluence with Neretva it is called Trebižat. Tihaljina-Mlade-Trebižat is 50 km long. We are not aware of any collection samples of a small-scaled Squalius from this river section that we could additionally examine. As S. microlepis phenotype 2 is only recorded downstream to Grabovnik-Vašarovići, we suppose that it does not occur below the Kravica Waterfall (ca. 43°9'22"N, 17°36'29"E); this should be checked indeed.

Two specimens (J and K) from the Tihaljina River in the village of Tihaljina (Fig. 10) were unambiguously identified as S. microlepis s. str. (phenotype 1) using the diagnosis presented above and clear assigned to this phenotype in statistical analyses (Table 6, Fig. 9C). Our hypothesis is that individuals of phenotype 1 could penetrate from the Imotsko Polje-Vrljika system down to the Tihaljina via existing underground karst flows though the isolation between the two was enough to support the two morphologically distinct groups of populations. A similar phenomenon of migration was discovered in this karst system for sympatric Delminichthys adspersus (Palandačić et al. 2012).

Figure 10. 

A locality where two phenotypes of Squalius microlepis co-occur: Tihaljina River at Tihaljina, Bosnia and Herzegovina (7 July 2011).

According to local fishermen information, after a severe drought some ten years ago, S. microlepis has not been found in the Tihaljina near the village of Tihaljina, and S. tenellus was introduced to the Tihaljina from Buško Lake but did not establish (N. Ančić, pers. comm. 2011–2019).

Poljes at Vrgorac and Gradac

Historical NMW material includes specimens from at Gradac and Vrgorac, some indicating karst poljes’ names (Jezero and Dusina). Polje Jezero is a wetland (blato) with a periodical lake and the sinking stream Matica [Vrgoračka Matica, not to be confused with Vrijeka-Matica in Imotsko Polje] as a part of the right-hand tributary system of the Neretva. The Dusina area, where some karstic streams form temporary lakes, is located near Polje Jezero and belongs to the same karst drainage system. Squalius microlepis was often reported from Polje Jezero and ‘Lake of Dusina’ in the past since its original description based on NMW specimens (e.g., Heckel and Kner 1858: 206, Canestrini 1865: 67, Canestrini 1866: 111, Kolombatović 1886: 16, Car 1911: 64). Mrakovčić et al. (1996) indicate the occurrence of Squalius microlepis in the lower part of the Matica in Polje Jezero; however, only one specimen was collected by him many years ago (M. Mrakovčić, pers. comm.).

There were only five specimens from this area available for examination added by two more specimens we supposedly attribute to it. This sample is quite morphologically heterogeneous. Specimens C and E unambiguously belong to S. microlepis s. str. (phenotype 1) and F and G (Fig. 6A) to phenotype 2 (Table 6, Fig. 9C). Specimens A (Fig. 6C), B and D (Fig. 6B) are intermediates between the two phenotypes.

Specimen B (NMW 49427) is labelled as ‘Narenta, Heckel Reise 1840’ but there is no clarifying information on the exact locality. We failed to find collection specimens or reliable records on S. microlepis from the Neretva main stream (a long list of publications checked by us can be requested from the corresponding author). We speculate that “Neretva” as a locality can refer to streams near Vrgorac or the drainage in general; for example, Seeley (1886: 169–170) and Car (1911: 64) mentioned “river Neretva near Vergorac”. Karaman (1928) indicated that he had never found S. microlepis in the Neretva main stream.

We hypothesise that both phenotypes could co-occur in the poljes near Vrgorac in the past or individuals of the phenotype 2 from the upstream karstic system of the Tihaljina could migrate downstream to the poljes at Vrgorac. They could probably hybridise as some specimens are of intermediate morphology. The Matica [not to be confused with Vrijeka-Matica in Imostko Polje] is a part of the right-hand tributary system of the Neretva and connected to the Tihaljina system in its northernmost (upper) part (Bonacci et al. 2013b). We failed to find this fish during intensive field trips in karst poljes near Vrgorac in 2017–2019.

Two specimens NMW 49228 (as 49227 in Bănărescu and Herzig-Straschil (1998: 419)) labelled ‘Zara’ (an Italian name for Zadar, a town on the Croatian coast, ca. 44°7'19"N, 15°16'20"E), are also identified by our analyses as S. microlepis phenotype 1. Bănărescu and Herzig-Straschil (1998) supposed that these two specimens do not belong to this species but did not offer an alternative hypothesis. However, no other specimens of a small-scaled Squalius are known from this area considerably remote from the main range of distribution of S. microlepis s. l. In the vicinities of Zadar, there was a lake, Bokanjačko Jezero, dried up long ago. At present, the Baštica River and two artificial lakes in that region are inhabited by Rutilus aula (Bonaparte) and a wide range of introduced species (e.g., Rutilus rutilus (L.), Squalius cephalus (L.), Lepomis gibossus (L.), Carassius gibelio (Bloch), Cyprinus carpio L., Ameiurus melas (Rafinesque) (unpublished data of DJ and PZ). Most probably, the label does not refer to Zadar, but the sample might have been sent to NMW from Zadar by Kolombatović. The NMW collection contains two more samples labelled as “Kolombatović, Zara” (no date), of Chondrostoma knerii (Heckel, 1843) and Scardinius plotizza (Heckel & Kner, 1858) that indicates the Neretva drainage.

The Cetina River is also sometimes included into the range of S. tenellus (Freyhof and Kottelat 2008; Ćaleta et al. 2015), but it is not clear, if the species is considered as introduced or native. We could not find a published morphological description of the Cetina fish that supports this opinion. On the contrary, the native Cetina Squalius was identified as S. microlepis by some earlier authors (Kolombatović [Kolombatowitch] 1886, Brusina 1907). Later, it was considered a new undescribed species (Zupančič 2007: Squalius sp. 4) but a formal description did not follow. In the most recent review of Croatian freshwater fishes (Ćaleta et al. 2019) the presence of S. microlepis in the Cetina is considered as not confirmed.

A historical specimen, collected by Kolombatović in the Cetina (MZUF No. 13512, donated from Kolombatović, June 1880; see Nocita and Vanni 1999: 214) was only examined by us from a photo (Fig. 11). The specimen is damaged, the number of scales in the lateral series can be calculated by the scale pockets and remaining scales, and it is 69. So, it cannot be identified as S. tenellus but is similar to S. microlepis by this count though being quite different from the latter by its general appearance and may be Telestes ukliva (Heckel, 1843) which is a species endemic to the Cetina.

Figure 11. 

Specimen MZUF 13512 (identified as S. microlepis by Kolombatović), Cetina River. Photo credit: Saulo Bambi, Sistema Museale dell’Università degli Studi di Firenze, Sez. di Zoologia “La Specola”, Italy.

The three small-scaled entities, S. tenellus, S. microlepis s. str. (phenotype 1) and S. microlepis phenotype 2, appear much better morphologically differentiated from each other than species within the S. cephalus group (see, e.g., Doadrio et al. 2007; Turan et al. 2009; Bogutskaya and Zupančič 2010; Özuluğ and Freyhof 2011). Four published cytb sequences of S. microlepis, two from the Krenica Lake and two from the Trebižat River (Freyhof et al. 2005; Perea et al. 2010; Schönhuth et al. 2018), show some genetic difference between the two localities, 0.53–0.67% (R. Šanda, pers. comm.). We did not have the possibility to examine the voucher specimens, but the Krenica specimens are most probably a true S. microlepis (phenotype 1) and the Trebižat specimens might represent a S. microlepis phenotype 2. No variability was found between five published sequences of CO1 (Perea et al. 2010, Geiger et al. 2014, Schönhuth et al. 2018) – three from the Krenica-Imotski area and two from the Tihaljina-Trebižat (R. Šanda, pers. comm.).

Readily morphologically diagnosable entities cannot always be taxonomically discriminated using molecular markers due to very rapid events of speciation (i.e., species radiations) and specific factors driving them, such as niche evolution or morphological key innovations (e.g., Bickford et al. 2007; Martin et al. 2016) forming species complexes or polymorphic species. For example, the CO1 marker did not provide resolution in at least 17 complexes of “closely related” conventional (clearly morphologically distinct) species in the subfamily Leuciscinae (Geiger et al. 2014: table S1-C). On the other hand, many intraspecific morphological differences can occur and express themselves, for example, as ecological variability or geographic variation. Polymorphic populations are more the rule than the exception in fish (Skulason and Smith 1995) as differences between habitats of fishes (e.g., related to flow regime or foraging opportunities) create selective pressures resulting in morphological divergence between conspecific populations (Langerhans et al. 2003; Senay et al. 2014).

The key issue is how to interpret the morphological differentiation in these groups – either as reflecting different nominal species or as representing varieties or (eco-) phenotypes within a single species. As very limited molecular data exist on the two phenotypes of S. microlepis, we refrain from any taxonomic and nomenclatural conclusions until new molecular approaches (and new markers) are used, the polymorphism is properly sampled, and much more specimens are available for genetic phylogenetic analyses. However, as shown above, we can hypothesise that the phenotype 1 might represent a lacustrine morph of the species while the phenotype 2 is a riverine one.

Our study emphasises the fact that S. microlepis, either a group of two putative species or two habitat-related phenotypes, has become extirpated or extremely rare in the most part of its range since 2004–2011. A reason of the dramatic decline may be due to introductions of Perca fluviatilis Linnaeus, Squalius cephalus Linnaeus and Esox lucius Linnaeus established throughout the region. Hence, the phenotypic diversity described in the paper has been already largely lost and a critical investigation of its conservation status is severely required based on population genetic data. We applied the IUCN criteria (3.1) and suppose that the Red List status of the species should be Critically Endangered (CR: A2ce) based on 90% population reduction estimated in the last 15 years (ca. three generations). Sub-criteria: (c) population size reduction observed through the decline in the area of occupancy (AOO) and the extent of occurrence (EOO), and (e) effects of introduced taxa, pollutants and competitors are in place. Exact causes of the reduction are not yet known and may have not ceased. Remaining EOO has been estimated as approximately 250 km² and AOO only around 20 km² (five 2 x 2 km cells), although the lack of data since 2011 makes the situation even more critical.