Extreme natural events can have profound effects on biological systems ranging from individuals to ecosystems [98, 99] and--at least temporarily--reshape an organism's environment. However, the documentation of such events is often difficult, and precise data are rarely available to compare pre-disaster with post-disaster situations (but see [68, 100]). Classic examples of well-studied catastrophic natural disasters include the explosion of Mt. Saint Helens in 1980, which stimulated a large body of research [101–105], or the impacts of seasonal hurricanes on species like the Puerto Rican Parrot  and Anolis lizards [107, 108]. The impact of large scale floods on local adaptation and (parapatric) ecological speciation processes, however, has thus far not been examined.
Two months after the catastrophic flood in Tabasco in fall 2007, we found only little evidence for genetically detectible dislocation of individuals among the three divergent habitat types. In fact, habitat type, but not geographic distance, was the major predictor of genetic differentiation both before and after the flood. Furthermore, the pronounced genetic and life history differentiation among different habitat types (i.e., 'isolation-by-adaptation' [5, 88]) was very similar for the pre- and post-flood datasets.
Homogenization of locally adapted populations?
Our major question was whether the flood of 2007 has led to a homogenization of locally adapted populations, for which we found little evidence. A potential exception is the immigration of cave fish into the sulfur creek. It remains to be seen whether these immigrants are able to significantly contribute to the gene pool of the sulfur creek population or whether "cave alleles" will eventually vanish from the surface population with time (see below: Mechanisms of selection against immigrants).
Within the Cueva del Azufre, we did find that homogenization had occurred, since fine-scale (cave chamber-specific) genetic structure was lost after the flood in January 2008. While the small rearmost chamber XIII is usually separated by a 1.5 m waterfall from adjacent cave chambers, the flood in fall 2007 probably flooded the waterfall. Our current study provides several lines of evidence for the immigration of individuals into chamber XIII (e.g., several "newly" occurring alleles stemming from adjacent chambers). Hence, chamber XIII apparently is less separated from the other cave chambers than previously thought (e.g., [25, 77]). The evidence for increased individual dislocation between different cave chambers (setting the stage for potential gene flow) also raises interesting new questions with regard to phenotypic divergence within the cave. From front to rear chambers, there is not only a morphological gradient with variation in eye and head size [77, 109], but also heritable differentiation in opsin gene expression . To date, it is not clear how the increased connectivity within the cave affected trait expression.
On the (slightly coarser) scale of the life history analyses, we observed several shifts from pre- to post-flood samples within habitat types (as exemplified by the lower classification success of the cross-validation DFA compared to the pre-flood-only DFA). At first sight, this might be interpreted as an indication of flood-induced migration/displacement between the habitat patches. Still, there was no breakdown of trait divergence between habitat types, since the extreme ecotypes (Cueva del Azufre and El Azufre) clearly remained distinct with respect to female life history traits even after the flood (post-flood DFA). In particular, highly plastic traits like female and offspring body condition (fat content) were decreased after the flood in almost all populations. This could either be a signal of the disturbance of the ecosystems due to the rising water levels and increased flow forces or regular seasonal variation of body condition . Most importantly, the two traits known to be heritable in the Cueva del Azufre population (offspring size and fecundity [53, 55]) did not show a flood signal at all, suggesting a lack of immigrants from the other habitat types.
In the surface habitats, however, the cross-validation classification success was by far the weakest. Potentially, this could be due to a plastic response to the flood, but it more likely represents general life history differences between the two surface habitats: RA is a large river, and fish were sampled in stagnant pools on a sandbank in the river, while AB is a small, fast flowing creek. Similar differences have been documented between guppy (P. reticulata) populations from river versus creek habitats , and we also have evidence for this scenario from recent studies on general life history differences between these two habitats derived from fish sampled in January 2009 (R. Riesch, unpublished data).
We found a drastic impoverishment of genetic diversity in the Río Oxolotán after the flood. How can this pattern be explained? In the Río Oxolotán, water levels rose up to eight meters. Massive currents during the flooding probably washed away a considerable proportion of the total population. At first sight one might be tempted to interpret the observed reduction in measures of genetic diversity, above all observed heterozygosity (H
o), as an indication of a genetic bottleneck. It needs to be recalled, however, that our post-flood sampling was done only two months after the flooding, while any mollies born after the flood would have needed more time to grow to adulthood (i.e., > 6 months under laboratory rearing conditions; R. Riesch, unpublished data). Hence, our samples originated from adult fish, and are probably not offspring of the few(er) remaining individuals in such (overall, genetically impoverished) populations. Reduced values of H
o are, therefore, not as straightforward to explain as it may first seem. A possible scenario is that a big proportion of the population from our collection site in the Río Oxolotán was indeed swept away during the flood and was replaced almost entirely by fish from further upstream (the mountainous regions of Chiapas), and we hypothesize that fish in those upstream regions are characterized by lower genetic diversity. We plan to test this idea in the future by analyzing genetic diversity in more distant populations from all along the Río Grijalva drainage system.
Lower population sizes in the extreme (sulfidic) habitats generally coincide with a reduction in overall genetic variability (i.e., allelic richness and heterozygosity; for population sizes refer to [26, 112]). Such small populations are inevitably prone to loss of genetic diversity over time due to genetic drift. Indeed, reduced genetic variability appears to be a typical feature of cave fishes and has been attributed to small population sizes, founder effects, and/or repeated genetic bottlenecks [113–117].
Mechanisms of selection against immigrants
Particularly interesting in the study of ecological speciation are the mechanisms leading to and maintaining genetic differentiation, i.e., the question of how exactly divergent natural and sexual selection translate into reproductive isolation [9, 10, 118–120]. So, what are the potential mechanisms maintaining this small-scale genetic structuring even after such a catastrophic flood? Using reciprocal translocation experiments, we found natural selection against migrants between non-sulfidic and sulfidic habitats (with very high mortalities within 24 hours for fish from non-sulfidic waters when transferred into El Azufre and vice versa), whereas migrants between sulfidic cave and surface habitats did not exhibit increased mortality within the same time period . A heritable basis to higher physiological sulfide-resistance in sulfide-adapted fish was confirmed by common-garden-rearing (; reanalyzed in ). Further adaptations to survive under sulfidic, hypoxic conditions include plastic behavior, like aquatic surface respiration [122, 123] and enlarged heads (a heritable trait) allowing for a larger gill surface area . On the other hand, oxidative stress (e.g., due to down-regulated expression of antioxidant enzymes under hypoxia [124–126]) may explain the high mortality during translocation from sulfidic, hypoxic to normoxic sites ( for discussion).
But what about genetic differentiation between surface and cave habitats? This question is of particular importance because even before the flood some migrants/dislocated individuals from inside the cave into the sulfidic creek (El Azufre) were detected, and dislocation into El Azufre tended to increase due to the flood (Fig. 5). Given that genetic differentiation between both populations obviously remains stable over time ([25, 26]; this study), strong selection against immigrants at the light/dark interface must be postulated. Generally, negative phototactic behavior was hypothesized to play a role for cave colonization [127, 128]. Surface and cave fish could, theoretically, just differ in phototactic behavior, thus effectively preventing cave fish from venturing outside and vice versa. However, such a mechanism apparently plays no role in the Cueva del Azufre system, because both (at least lab-reared) surface- and cave fish are positively phototactic . It has been demonstrated though that mollies in all habitat types experience predation by a giant water bug of the genus Belostoma [79, 80, 130], and a recent prey choice experiment under semi-natural conditions found that water bugs are more likely to capture dispersers, i.e., cave fish at the surface and surface fish inside the cave . This may be due to sensory systems being maladaptive in the "wrong" habitat type. Cave mollies--living normally in perpetual darkness--have reduced eyes along with more elaborated non-visual senses like a hypertrophied head canal system of the mechano-sensory lateral line [25, 49, 77, 109, 110, 121]. This allows for earlier detection of predators in darkness compared with the surface ecotype, while smaller eyes hamper predator detection in light .
In traditional models of ecological speciation, reproductive isolation evolves incidentally as a by-product of divergent natural selection [1–3, 131]. But whenever divergent selection occurs among populations, there may also be direct (sexual) selection for premating isolation (i.e., reinforcement, [2, 132]). In fact, females of both sulfidic populations (El Azufre and chamber II of the Cueva del Azufre) discriminate against immigrant male phenotypes during mate choice . Altogether, selection against immigrants may indeed be a powerful mechanism facilitating speciation among locally adapted populations even over very small spatial distances, and we hypothesize that natural selection against immigrants plays the main role in maintaining the small-scale genetic structure.