Genetic differentiation between troglomorphic fish of Al Hoota Cave and surface-dwelling fish of Wadi Al Falahi
What do the results reveal with respect to the origin of the troglomorphic forms? The main question of the study was whether the two morphotypes of G. barreimiae are genetically differentiated and whether gene flow occurs.
In the network (Figure 4) all individuals from Al Hoota Cave belong to one haplogroup ("cave haplogroup"), but this group contains also a few individuals from adjoining Wadi Al Falahi. Hence there is no strict correspondence between haplogroups and the two ecological environments. The "cave haplogroup" shows a pattern which suggests a founder effect caused by a single invasion into the cave followed by fast expansion; this is also supported by the mismatch distribution analyses. The presence of surface-dwelling individuals in this haplogroup might be explained by incomplete lineage sorting, which we consider unlikely for the following reasons: All Al Hoota specimens belong to one haplogroup. This "cave haplogroup" is not found in any other locality except in a few individuals collected in Wadi Al Falahi at the exit of Al Hoota Cave. Moreover, the vast majority of Wadi Al Falahi samples belong to several different lineages (i.e. haplotypes 13, 14, 23, 15, 11, 27, 18, 28, 17; Figure 4) and only a single haplogroup is shared with individuals in the cave. None of the other various haplotypes found in Wadi Al Falahi was detected in Al Hoota specimens. Finally, most of the individuals of Wadi Al Falahi possessing cave haplotypes show troglomorphic characters. In summary these facts indicate that the extant populations of Al Hoota and Wadi Al Falahi are not completely separated. Two potential scenarios explain how the populations might get into contact. (i) Although most of the water bodies in the wadi dry out frequently, fishes with eyes appear within a few days after rainfalls in the surface water (observation by HS and RS). Hence, fish might retreat into the groundwater, where they might get into contact with individuals of the cave population. For the surface dwelling fish, however, these environmental conditions might not be suitable for breeding. (ii) During floods, fishes might be sometimes washed out from the cave into Wadi Al Falahi. Up to 200 m3/sec  can flow out from the Al Hoota Cave into the Wadi Al Falahi after heavy rain. Concerning the direction of gene flow, our data support a sporadic translocation of individuals from Al Hoota Cave into the Wadi Al Falahi. From the ten individuals of the "cave haplogroup" caught in Wadi Al Falahi, six had reduced eyes. If gene flow was acting in both directions, some individuals with intermediate phenotypes or with "surface" Wadi Falahi haplotypes would be expected in Al Hoota Cave. Neither of these cases was detected and thus the most likely explanation is that occasional gene flow occurs from the cave to the surface, only. This could explain both the presence of samples with "cave haplotypes" and the occurrence of intermediate phenotypes in Wadi Falahi. The question, however, is whether the blind individuals can reproduce in surface water bodies. One would assume that blind fish have a low probability of surviving long and reproducing in surface populations because of impairment by radiation or by high predator pressure (e.g., from water insects, fish and birds). Gene flow is therefore expected to be quite low between surface- and cave-dwelling fish.
Besides the various intermediate forms, four individuals of the "cave haplogroup" from the Wadi Al Falahi possessed eyes that resembled the surface phenotype (Figure 5). It remains unknown whether the eyes of the latter individuals are functional: to date no systematic morphological investigation has been performed, whether there is a plastic response of eye development in the presence or absence of light. A single report by Banister  notes that the development of the optic lobes was influenced by light. Behavioural changes due to light exposure during development have been documented by : light reared juvenile fish showed significantly higher photophilic behaviour than dark reared juvenile individuals. Should the eyes in the four above mentioned individuals be functional, then the most plausible explanation is that these individuals are hybrids between the two populations (or their descendants) that retained the cave haplotype but are able to develop functional eyes. In summary, our data indicates that the troglomorphic individuals of G. barreimiae living in Al Hoota Cave sporadically interbreed with individuals of the adjacent wadis.
Another question concerns the age of the cave population. An answer would provide indirect hints about the underlying mechanisms for the development of troglomorphic features. Surface and cave forms were crossbred in the laboratory (Wilkens, personal communication in ), suggesting that the two genomes are still compatible. If the eyeless condition was a genetic trait that became fixed due to relaxed selection pressure for maintenance of vision-related traits in the cave, then this would require some time. In this context the divergence time of lineages is of interest and dating of the split of the "cave haplogroup" based on genetic distances would be desirable. This is problematic because there is no reliable calibration for a molecular clock of G. barreimiae. The cave is of Plio-Pleistocene age, roughly one to three million years old. This might be set as the earliest date for the origin of the cave population. Nonetheless, the "cave" and the "surface" haplogroups may already have split some time before the intrusion into the cave. It is also possible that the cave population originated much later, and/or from a surface population with a different (more closely related) haplotype, which we either did not sample or which does no longer exist.
Despite these uncertainties, the observed genetic differentiation between sequences of the "cave haplogroup" and those from the surface populations is low (minimum 0.91% in the CR1 fragment). P distances are in the range of those found within several surface populations of G. barreimiae. Attempts have been made to calibrate a substitution rate for the cytochrome b (cytb) gene in cyprinids: 0.53% per million years  and 0.76% per million years . For the mitochondrial CR no calibration exists, but it can be assumed that this section, due to less selective constraints, has a higher substitution rate compared to cytb. Applying the rates for the cytb gene for our CR1 data set would most probably result in an overestimate of the actual divergence time of epigean and hypogean G. barreimiae. Anyway, a rough calculation based on the distance of 0.91% and the two above mentioned rates yields a split below 1 mya (0.86 or 0.60 mya, respectively). This would disagree with the results of Colli et al. , who estimated the divergence time between epigean and hypogean G. barreimiae between 0.8 - 2.0 Mya based on a 795 bp sequence of cytb. This calculation remains dubious because it is unclear whether those sequences  represent the functional cytb gene or nuclear pseudogenes (numts - copies of mitochondrial genes that have been translocated into the nuclear genome) or chimeras of both. In the study of  the cytb sequences were composed of two independently generated PCR-products. Our preliminary experiments revealed the presence of cytb numts in this genus. We used primers that bind in the flanking t-RNA genes to amplify the whole gene (1140 bp) in one fragment (data not shown). Comparing the complete cytb sequences obtained in our study with those of  shows that in the 3'-section they are almost identical; while in the 5'-section they are quite different (8% p-distance). The most plausible explanation is that  presented a sequence chimera. An extensive investigation of cytb in this genus would be necessary to draw final conclusions and to disentangle authentic mtDNA sequences and numts. In any case, it is currently impossible to use cytb data for any attempt to date splits in G. barreimiae.
The very low genetic distances between cave and surface populations (Al Hoota/Wadi Al Falahi) suggests that the Al Hoota population is quite young. This indicates an evolutionary mechanism for troglomorphic characters that differs from the simple accumulation and fixation of neutral mutations. The results favour a selection-driven hypothesis as was suggested for Astyanax, where increased shh expression proved to cause eye degeneration and enhanced forebrain and taste bud development [[5–8]]. In the evolution of troglomorphic phenotypes, however, the same genes are not necessarily always have to be involved, and selective pressures influencing different cave species may not be the same all over the world. Factors affecting a cave population can vary widely (e.g., composition of the cave coenoses, availability of trophic resources, water regime, frequency of exchanges with the surface environment, etc.). Differences in physiology or the presence of preadaptations to cave life must also be considered, along with phenotypic plasticity and epigenetic effects. The effect of different environmental pressures can be particularly strong if the development of troglomorphic traits is due to epigenetic modifications or other changes in gene regulation. The plasticity of the depigmentation in G. barreimiae, which is reversible to some extent in the presence of light (observation by HS, ), is in accordance with the presumed recent origin of the cave population.
A very remarkable finding is that the single blind fish from Al Hamra Falaj surface locality had a haplotype (25) which is highly divergent from the Al Hoota Cave haplogroup. It did not occur in any other individual and is closely related to a haplogroup comprising individuals from various surface populations (Figure 4). We assume that this adult individual, which was caught in an irrigation channel, originally derived from a yet unknown cave population where the troglomorphic phenotype has developed independently. Convergent evolution of cave fish populations from different surface populations has also been documented for Astyanax species [[17, 18, 32]]. Alternatively, the phenotype of this specimen could be ascribed to a recent mutation with pleiotropic effects, and the individual by chance survived until maturity. The latter scenario, however, seems highly unlikely because, as far as it is known from other cave fish, aborted eye development is a multifactorial trait in which at least six genes are involved and, in addition, the lack of pigmentation is caused by different genes .
Phylogeography and taxonomic considerations
Within clade 3 (Figures 3 and 5) the populations Wadi Al Falahi, Ain Al Msalla, Al Hamra Falaj, Wadi Misfat, Wadi An Nakhar and Wadi Dhum are genetically closely related and do not show any geographic structure. Within most of the localities, we found various haplotypes which are shared between different localities. This suggests frequent gene flow between localities. The wadis are usually not connected on the surface, but underground water bodies might exist which enable migration between the wadis. Flooding may also connect wadis. Another explanation would be that a former variable panmictic population was split only very recently through the separation of the wadis. The age of the youngest dripstone activity in Al Hoota Cave dates back 6,000 - 10,500 years . This date corresponds with the last more humid period, in which the contact of different populations in permanent rivers within the same river system appears likely. Although water bodies are currently separated and thus gene flow is interrupted, lineage sorting in the current populations may not be completed. The geographically more distant locality Wadi Bani Habib (35 km away from the next population sampled) is genetically homogeneous, comprising three connected haplotypes not found elsewhere. This group is separated from the next main haplotypes by a minimum of 1.5% sequence divergence (six substitutions), suggesting longer isolation of this population.
The population of Wadi Bani Khalid (clade 2 Figure 3), besides Wadi Mansah, the only sample from Al Hajar Ash Sharqi (Western Hajar), is not only geographically the most distant one (200 km), it is also genetically rather distant (average p-distance to clade 3: 4%). The individuals from this population were described as a separate subspecies, G. barreimiae gallagheri, which is morphologically close to the nominate subspecies but differs in shorter fins and a higher number of dorsal fin rays . Our results (i.e., the distinct mitochondrial lineage) are in accordance with this classification. Note, however, that many regions have not been sampled so far, for example we have no samples of the surrounding populations of Wadi Bani Khalid. It cannot be excluded that continuous geographic sampling would also reveal more continuous genetic relationships. The fact that neither this study nor any previous molecular study included fish from the type locality of G. barreimiae (Buraimi/Al Ain, which are nowadays densely populated) makes it difficult to address the intraspecific taxonomy of G. barreimiae based on the current state of knowledge.
The situation is different concerning the remarkably high genetic distances that were found between clade 1 (Wadi Hat, Wadi Bani Awf1, Wadi Bani Awf2 and Wadi Mansah) and the other clades. In fact, the genetic analyses question whether the individuals of clade 1 belong to G. barreimiae because the distances between this clade and the other lineages of G. barreimiae are in the same range as those between clade 1 and G. rufa. Our data do not resolve the phylogenetic relationships with a high statistical support. Although the specimens from these localities resemble G. barreimiae at first sight, no morphological analyses have been performed that definitely support this assignment. Thus, we cannot exclude the possibility that the populations of clade 1 represent a different species. Any taxonomic revision of Garra in Southeast Arabia should include phylogenetic analyses of samples from the topotypic locality or at least of a population close to the type locality. Nonetheless, the high genetic distances can be explained by the fact that these populations underwent long-term isolation because the water bodies they inhabit are separated by a watershed.