Diversification of Circum-Mediterranean Barbels

The Mediterranean Basin is one of the 25 most biodiverse regions on Earth. It is considered a biodiversity hotspot for its high numbers of endemic vascular plants, birds, mammals, reptiles, and amphibians, sometimes restricted to small distribution areas (Medail & Quezel, 1999; Mittermeier et al., 1998; Myers et al., 2000). The Mediterranean has had a long and complex geomorphologic history, being a relic of the Mesozoic Tethys Ocean. The Tethys had disappeared by the end of the Eocene (34 Ma) due to the collision of the Indian and Asian plates (Rogl, 1999). The orogenic movements raised new mountain ranges in the Taurides, the Hellenides, the Dinarides and finally the Alps by the Middle/early Late Miocene (Hsu et al., 1977). This orogeny separated the borning Mediterranean and a central/eastern European inland sea – the Paratethys Sea (Hsu et al., 1977; Rogl, 1999). Landbridge connections and seaway passages between the Mediterranean and Paratethys, and between them and the Indian and Atlantic oceans, were then intermittent throughout the Miocene until the final opening of the Strait of Gibraltar ending the Messinian Salinity Crisis (Agusti et al., 2006; Hsu et al., 1977; Krijgsman et al., 1999; Rogl, 1999). This complex geomorphological scenario has allowed multiple faunal and floral exchanges between neighboring regions (e.g. Agusti et al., 2006; Benammi et al., 1996; Pickford et al., 1993, 1995); this melting pot might have contributed to the extraordinary diversity observed nowadays. For instance, the Middle East has been an important region for freshwater fish interchange between Africa, Asia and Europe (Durand et al., 2002). Another relevant aspect is whether persistence (i.e. low extinction), diversification (i.e. high speciation), or both, are responsible for high species diversity in the Mediterranean (e.g. Reyjol et al., 2007). Among vertebrate groups, primary freshwater fishes probably constitute the majority of living endemisms in the Mediterranean region and include several species with restricted distribution ranges. Certain regions in the northern Mediterranean have been identified as important biodiversity hotspots for riverine fish (Reyjol et al., 2007) and the same is likely true for southern Mediterranean ones. This is explained by the limited dispersal routes of freshwater-restricted species, living within the confines imposed by salt water on one hand, and land on the other. Such qualities make primary freshwater fishes ideal models for the study of biogeographical history, landscape evolution and processes driving diversification in general (Briggs, 1995). Cyprinid fishes are a prime example. They are the most diverse family within the order Cypriniformes and naturally inhabit freshwaters of all continents except for Antarctica, Australia, and South America (Banarescu & Coad,


Improved fossil data and calibration of barbel phylogenies
Some previous studies of historical Mediterranean biogeography -of parts or the whole area -have made use of calibrated molecular phylogenies of barbels. Most of the studies relied on molecular clocks calibrated with known geological events, such as the openings of the Strait of Gibraltar and/or the Strait of Korinthos (Machordom & Doadrio, 2001b;Mesquita et al., 2007;Tsigenopoulos et al., 2003Tsigenopoulos et al., , 2010Zardoya & Doadrio, 1999). The results from the different studies varied slightly, depending on which particular node was calibrated, but most importantly they might have inadvertently biased results for accepted vicariant events or those perceived as more likely. Other studies have calibrated molecular phylogenies using published rates for other organisms (Durand et al., 2002;Tsigenopoulos & Berrebi, 2000;Tsigenopoulos et al. 2010). The use of (a range of) possible mutation rates is commonly accepted, in particular if fossils are not available for specific groups. Nevertheless, mass-specific metabolic rate and temperature influence the rate of molecular evolution in poikilotherm fishes (Estabrook et al., 2007). Using a 'universal' rate might have as a consequence over-or underestimation of the real rate of molecular evolution of the particular study organism. Interestingly, of the abovementioned studies that calibrated phylogenies either using known geological events or published rates, perhaps all but one underestimated the age of Barbus and Luciobarbus according to current fossil data. One last study calibrated a molecular phylogeny with fossil data ), but it was restricted to a very small area of the Mediterranean Basin. Therefore, given the current scenario just described, there is the need for a new analysis using updated fossil information, across the entire Mediterranean Basin, to re-evaluate the timing and pattern of diversification of circum-Mediterranean barbels. Latest fossil data available in Böhme & Ilg (2003) were used to calibrate a molecular phylogeny. Divergence times and their credibility intervals (highest posterior density: HPD) www.intechopen.com Changing Diversity in Changing Environment 286 were estimated using a bayesian MCMC approach implemented in BEAST v1.6.1 (Drummond & Rambaut, 2007). Barbus fossils of Burdigalian age are now known from several localities in what is presently Central Europe and Anatolia (oldest: 19.0 Ma). This wide distribution area suggests that the genus had already diversified by the Early Miocene. Likewise, Luciobarbus fossils of Burdigalian age are known from Anatolia (oldest: 17.7 Ma). These dates set hard lower bounds for the diversification of each group. Additionally, Luciobarbus fossils of Messinian age are known from the Iberian Peninsula (oldest: 5.8 Ma), which represents a lower bound for the diversification of Iberian Luciobarbus as in Gante et al. (2009). The upper age is a soft bound free to vary following a lognormal distribution (Ho, 2007) set in real space with average of 1.0 and standard deviation of 0.5. Each gene used (see below) was a distinct data partition, with unlinked substitution models, and following relaxed uncorrelated lognormal clock models and a General Time Reversible model of evolution. Third codon positions were treated separately from 1st and 2nd codon positions. A speciation birth-death tree prior was used, since a Yule speciation prior assumes complete taxon sampling. Analysis was run for 25,000,000 generations, sampled every 2,500 generations, first 1,001 trees discarded as burn-in. A total of 80 taxa were analyzed for the mitochondrial regions cytochrome b (1,141 bp) and ATPsynthases 6/8 (842 bp). As target ingroup, for the reasons explained above, representatives of Barbus (n=16), Luciobarbus (n=29), Capoeta (n=3), and Aulopyge huegelii from throughout the distribution area of the group were included. Additional cyprinins (n=31) originating in Asia and Africa were included in the analysis to provide a geographic, as well as phylogenetic context. Since the birth-death tree prior used assumes balanced sampling, outgroup species with varying divergence levels were selected ( Fig. 1

Phylogeny of barbels 4.1 Relationships among genera and major groups
The phylogenetic relationships of circum-Mediterranean barbels have been thoroughly explored over the last decades. The wide phylogeny of barbels obtained here based on mitochondrial cytochrome b and ATPsynthase 6/8 genes, which includes several additional allied Asian and African cyprinines, is shown on Fig. 2. It is consistent with phylogenies obtained in previous studies based on partially overlapping sets of taxa (Durand et al., 2002;Gante et al., 2009;Machordom & Doadrio, 2001a, 2001bMesquita et al., 2007;Tsigenopoulos & Berrebi, 2000;Tsigenopoulos et al., 2002Tsigenopoulos et al., , 2003Tsigenopoulos et al., , 2010Zardoya & Doadrio, 1998. Circum-Mediterranean Barbus s. str. forms a strongly supported monophyletic group composed of two barbel lineages (Machordom & Doadrio, 2001a). These mitochondrial lineages, Barbus and Luciobarbus, now considered distinct genera, are in agreement with previous morphological evidence (Doadrio, 1990), and recent nuclear data (Gante, 2009). Barbus and Luciobarbus are sister to Aulopyge as initially suggested by Howes (1987) and Tsigenopoulos & Berrebi (2000). Altogether, they are likely sister to a group constituted by Middle Eastern Cyprinion, and Asian genera such as Schizothorax (Durand et al., 2002) and Gymnocypris (Fig. 2). This relationship is not well supported by available molecular evidence and would benefit form added sampling effort, both in terms of taxa and markers. Mitochondrial phylogenies strongly indicate that 'true' barbels are not closely related to other African and Asian barbels. Rather, African diploids are a distinct paraphyletic group, with African tetraploids nested within them, suggesting a tetraploidization event from African diploids, and independent from the one that originated circum-Mediterranean barbels ( Fig. 2; Machordom & Doadrio, 2001a;Tsigenopoulos et al., 2002). Likewise, hexaploid cyprinins found in Africa and the Middle East constitute an independent evolutionary lineage whose origin is still not well understood (Fig. 2). These have recently been lumped into Labeobarbus (Tsigenopoulos et al., 2010). Together with Labeo, they indicate multiple independent colonization events of Africa from Asian ancestors, possibly at different times.

Phylogenetic diversity of circum-Mediterranean barbels
Even without exhaustive sampling, the genera Aulopyge, Barbus and Luciobarbus show strikingly different degrees of diversity. Aulopyge is a species-poor genus composed of only one very specialized extant species, A. huegelii, compared to the species-rich Barbus and Luciobarbus. Aulopyge heugelii is an (almost) scale-less species with an elongated urogenital opening which functions as an ovipositor (Kottelat & Freyhof, 2007). It inhabits the Dinaric karst in the Balkanic region. Perhaps the lack of diversity within the genus can be explained by its biology and the degree of specialization attained and/or by its habitat, being imprisoned within the karstic labyrinth and not able to colonize other regions. The genus Barbus is composed of at least four mitochondrial lineages (Fig. 2). It is restricted to the north Mediterranean region and is particularly diverse in Greece where species belonging to three out of the four identified lineages are found. Low diversity of Barbus in central and eastern Europe is likely related to the last Ice Ages, when glacier formation drove local populations/species to extinction, followed by rapid re-colonization by a restricted pool of founders (Kotlík & Berrebi, 2001). The genus Luciobarbus is the most diverse and widespread of all. It is composed of at least seven mitochondrial lineages showing very good geographic concordance (Fig. 2). This pattern indicates that Luciobarbus speciated in loco after seeding by ancestral species. Interestingly, one of these lineages comprises the hexaploid genus Capoeta. Capoeta appears to be monophyletic and has likely evolved from ancestors of (the tetraploid?) Luciobarbus subquincunciatus. Together, they form a strongly supported monophyletic group deep in the Luciobarbus lineage (Durand et al., 2002;Tsigenopoulos et al., 2003). Luciobarbus has a very wide distribution, only not being native to central and eastern Europe, and Italy (Doadrio, 1990). It is very diverse in northern Africa, where one lineage radiated and where a colonizer from Iberia, L. setivimensis, can be found (Machordom & Doadrio, 2001b).
Luciobarbus is also relatively diverse in the Middle East and Caucasus, where at least three lineages occur, concordant with multiple colonization routes scenario (Almaça, 1990).
Capoeta is also a very species-rich genus, with about 20 species distributed from western Asia to Anatolia (Banarescu, 1999;Turan et al., 2008). Besides the inferred radiation within Africa, Luciobarbus has undergone rapid speciation early in its existence -most of the lineages identified date back to early Luciobarbus diversification (Tsigenopoulos et al., 2003). These polytomies (as the ones identified in the African lineage) do not likely represent a lack of information content in the data (soft polytomies), since the phylogenetic signal before and after these splits is very strong (Fig. 2). Therefore, these polytomies should represent legitimate radiation events (i.e. hard polytomies). Regarding regional relationships within Luciobarbus, the northern African lineage, the Middle Eastern/Caucasus lineage and the Greek L. graecus form a strongly supported group. This is in conflict with the view that Iberia could have been seeded by northern African Luciobarbus (Doadrio, 1990;Gante et al., 2009) or the other way around (Almaça, 1990). Likewise, a hypothetical relationship between Capoeta and Iberian Luciobarbus (Tsigenopoulos et al., 2003) is not supported by the data. This lends the exact origin and relationships of Iberian Luciobarbus a mystery. In contrast to this abundance of fast speciation in Luciobarbus, only a few short internodes are present in Barbus lineage. Whether this pattern reflects a difference in biology between Barbus and Luciobarbus is unclear. Interestingly, though, poorly supported nodes show some overlap in time, suggesting a possible external (environmental) driver. This hypothesis would need proper testing with a much more exhaustive taxon sampling. Regarding regional relationships within Barbus, there is a much weaker correlation between lineages and geography than that seen in Luciobarbus. Such pattern indicates less isolation between Greece, central and eastern Europe, Italy and Iberia.

Dating the diversification of barbels
The dating strategy followed here differs from that of most other studies that included circum-Mediterranean barbels. Here, up to date fossil data (Böhme & Ilg, 2003) was used to calibrate a molecular clock, instead of biogeographical events or 'standard' mutation rates. As a consequence, the dates estimated in the present work are substantially older than previous estimates. For instance, the time of splitting between Barbus and Luciobarbus has been estimated to have occurred 5.5 Ma (Machordom & Doadrio, 2001b), 7.3 Ma (Tsigenopoulos et al., 2010) or ≈8 Ma (Tsigenopoulos et al., 2003;Zardoya & Doadrio, 1999) using the Messinian Salinity Crisis as the driver of speciation of the Iberian Luciobarbus lineage. It was estimated to have occurred 10.3 Ma (Tsigenopoulos et al., 2010) or 10.6-12.8 Ma (Tsigenopoulos & Berrebi, 2000) using previous estimates of mutation rates. Since the oldest known fossils of Barbus and Luciobarbus are 19.0 Ma and 17.7 Ma, respectively, those ages are certainly an underestimation of the real time of divergence between these two genera. In contrast to the abovementioned estimates, according to the calibration used here, Barbus and Luciobarbus diverged 27.6 Ma (95% HPD: 24.6-31.2 Ma). Since the calibration was applied to the nodes (without the stem), it is possible this age could be somewhat overestimated if earlier lineages diversifying within both Barbus and Luciobarbus (and represented in the fossils found) got extinct and are missing from the molecular phylogeny. Nevertheless, other sources of evidence support the new estimates shown here. For instance, divergence of Varicorhinus is estimated to have occurred 17.2 Ma (95% HPD: 13.4-21.2 Ma), which is supported by fossils of 17.8 Ma found in central Europe (Böhme & Ilg, 2003). Furthermore, the estimated time of divergence of L. setivimensis is 5.3 Ma (95% HPD: 3.7-7.1 Ma), which is exactly coincident with the re-opening of the Strait of Gibraltar by the end of the Messinian (Krijgsman et al., 1999). According to the molecular clock calibration presented here, divergence of the lineage leading to Aulopyge happened 31.6 Ma (early Oligocene, Rupelian) and divergence between Barbus and Luciobarbus occurred 27.6 Ma (late Oligocene, Chattian). After a long period of stasis (or possibly high extinction) diversification within Barbus and Luciobarbus took place 19.7 Ma and 18.6 Ma, respectively (early Miocene, Burdigalian). The lineage leading to B. haasi and B. meridionalis split 14.2 Ma (middle Miocene, Langhian-Serravalian boundary). Nevertheless, this date should be carefully interpreted since the node support is rather low (BI = 0.80). Indeed, fossil teeth and vertebrae of Barbus have been found in Iberian sediments with 16-17 Ma (Doadrio 1990

Paleogeography of the Mediterranean Basin and diversification of barbels
Seeding of the Mediterranean with a tetraploid barbel lineage most likely occurred during the late Eocene or early Oligocene (Fig. 2). This dating is consistent with an Asian origin of cyprinids and colonization of Europe at the closing of the Turgai Strait in the Eocene-Oligocene boundary (Almaça, 1990;Banarescu, 1992;Briggs, 1995;Rögl, 1999). Progression towards the west was possible due to the emergence of a large landmass that extended across the Balkans, Anatolia and Iran (Rögl, 1999). The carbonate rocks with more than 8,000 m that form the Karst Dinarides were deposited for more than 270 Ma and raised during the Alpine orogeny (Velic, 2007), suggesting that present-day karst habitat inhabited by Aulopyge was already present, in the place this oldest barbel lineage is presently found. Such a colonization scenario through southwestern Asia was also hypothesized for Leuciscins (Perea et al., 2010). The split between Barbus and Luciobarbus in the late Oligocene could have been driven by the fragmentation of this landmass (Rögl, 1999). In a time of intense tectonic activity in the Mediterranean, the opening of the Slovenian corridor is a likely candidate, fragmenting Barbus to the north and Luciobarbus to the south. The timing of diversification within these genera at 20 Ma is coincident with the closure of the Slovenian seaway (Rögl, 1999). This reunited landmass might have allowed access to regions where the oldest fossils of Barbus and Luciobarbus have been found. The time of origin of the lineage leading to L. subquincunciatus and Capoeta, which are found in the Middle East and Caucasus, is coincident again with a transitory fragmentation of this landmass. In the middle Miocene, seaway corridors opened between Arabia, south Anatolia and eastern Anatolia, and possibly along a suture between the Balkanides and the Rhodopes (Rögl, 1999). The branching out of Capoeta occurred at the time when oceanic circulation between the Indian and Atlantic oceans stopped, in the Serravalian (Rögl, 1999). By then, the main lineages within Barbus and Luciobarbus had already originated. Explaining their current distribution is no easy task with current paleogeographical and paleontological evidence. Fossils of Barbus of Burdigalian age are found from Turkey to Iberia (Böhme & Ilg, 2003;Doadrio, 1990), which corresponds to the present distribution of the genus, except for Italy. Colonization of Italy during the Messinian Salinity Crisis is a likely scenario. Equally old fossils of Luciobarbus have been only found in Turkey and it is not before the Tortonian they are found in central Europe (Böhme & Ilg, 2003). This could be due to taxonomic bias, since Luciobarbus has not been recognized for as long as Barbus, or it could reflect a real trend of Luciobarbus biogeography. Nevertheless, the presence of fossils in central Europe in Tortonian times, where it is now absent, opens new routes for Luciobarbus dispersal. In particular, it is known that during Alpine orogeny, marine influence in the North Alpine Molasse ended in the middle Miocene (Langhian; Hsü et al., 1977;Krenmayr, 1999;Rögl, 1999). Barbels could have used this basin as a means of southwestward dispersion to Iberia, independent from the colonization of northern Africa. Alternatively, they could have used slightly different pathways via the Gomphotherium landbridge connecting Africa and Eurasia (Rögl, 1999), as suggested by Perea et al. (2010) to explain vicariance of Peloponessus and Magreb Tropodophoxinellus. Since the distribution areas of Leuciscinae and Cyprininae are similar, as well as inferred dates of groups occupying those regions, it is likely they shared common migratory routes. The subsequent radiation of northern African Luciobarbus is likely related to complex paleogeomorphology of the Rif massif (e.g. Alvinerie et al., 1992;Machordom & Doadrio, 2001a). Starting around 7.8 Ma the marine corridors between Iberia and northern Africa became restricted until the establishment of a land bridge around 5.6 Ma (Messinian) (Garcés et al., 1998(Garcés et al., , 2001Krijgsman et al., 1999;Martín et al., 2001;van Assen et al., 2006). Dispersal and subsequent vicariance of L. setivimensis between Betic and Riffian massifs has occurred during the Messinian Salinity Crisis. Nevertheless, speciation events are not concentrated in this period, nor are the inferred radiations. Other authors have recently ruledout a "Lago Mare dispersal" for leuciscin cyprinids (Levy et al., 2009;Perea et al., 2010). This period seems to have been used for transfer between adjacent areas (e.g., Iberia -northern Africa, central Europe -Italy) rather than a Mediterranean-wide colonization by barbels.

Conclusion
According to the fossil calibrated molecular phylogeny presented here, divergence of the circum-Mediterranean barbel lineages occurred during the Oligocene. Divergence within Barbus and Luciobarbus took place throughout the Miocene, including spreading to new areas. Altogether, colonization of the Mediterranean region by barbels must have been a very dynamic process we are just starting to understand, as indicated by the presence of many fossils in regions where the genera are presently not found. A good example is the presence of several Luciobarbus fossils in Libya, Italy, Austria, and Slovakia. Greater insight will likely continue coming from paleontological and paleogeographical data, and that should be accompanied by new biological data of extant species. In particular, all of these scenarios are only based in non-recombinant mitochondrial DNA markers. The coming decade should see the rise of nuclear phylogenies and an improved understanding of barbel biogeography in the Mediterranean region.

Acknowledgment
This work has greatly benefited from the interactions with several people over the last decade. Namely, I would like to thank Maria Judite Alves, Thomas E.