Spatial variation in defensive traits is common in many antagonistic interactions 123. The geographic mosaic theory of coevolution suggests that variation in ecological factors may underlie this patchiness in the phenotypic outcome of interactions, resulting in cold and hotspots of coevolution 245. Despite evidence for coevolutionary cold and hotspots, very little evidence exists on how ecological conditions may enhance or constrain coevolution between antagonists. One current model suggests that coevolution might be related to gradients in resource productivity 6, with hotspots of coevolutionary selection associated with areas of high prey abundance. In specialist predator-prey interactions, predators have the greatest influence on prey phenotypes in geographic areas where the prey is most abundant whereas the predator exerts weaker selection in areas where prey are less abundant. Conditions of high prey abundance can occur when the abiotic environment is most favorable for prey metabolism or prey resources are of high quality/abundance. In the current study, we examine the effect of abiotic and biotic factors on spatial variation in the evolution of putative defensive traits of the freshwater snail Mexipyrgus churinceanus.

Mexipyrgus churinceanus as well as its polymorphic fish predator Herichthys minckleyi are endemic to the isolated Cuatro Ciénegas valley in the Mexican Chihuahuan desert (Fig. 1) and exhibit putatively co-evolved phenotypes 789. These phenotypes may currently be coevolving because Mexipyrgus exhibits striking examples of morphological differentiation in shell morphology and pigmentation among habitats [101112; Fig. 2]. Herichthys minckleyi is also unusual because this cichlid fish exhibits two alternative pharyngeal jaw morphologies [1314; Fig. 2]. In "papilliform" H. minckleyi, the gill arches are modified into a gracile pharyngeal jaw and this morph is specialized to shred aquatic plants and detritus 15. Alternatively, "molariforms" possess enlarged crushing molariform teeth and robust pharyngeal muscles 16 and are extremely proficient at crushing M. churinceanus 15. Molariform cichlids are the primary snail-crushing predator in Cuatro Ciénegas.

Given considerable spatial variation in shell pigmentation and morphology, we hypothesize that several types of snail defenses have evolved in response to predation by molariform H. minckleyi. For instance, cryptic coloration against different benthic substrates may minimize detection by molariform cichlids. The abundance of emergent aquatic vegetation, such as Nymphaea water lilies, is highly variable in these aquatic habitats, and may be primarily responsible for the coloration of the substrate. Darker substrates dominate in areas with dense Nymphaea stands and light, marl-colored substrates occur in the absence of Nymphaea (Fig. 2). Unpigmented M. churinceanus may be more cryptic and common in habitats without Nymphaea whereas, in heavily vegetated habitats, snails with extensive shell pigmentation may dominate because they effectively match these darker substrates.

Shell pigmentation is not the only defense M. churinceanus may exhibit in response to molariform predation. Crushing resistance is likely the most functionally important component of shell defense against molariform H. minckleyi 15 because it allows snails to resist predation once detected. Because previous studies of defensive traits in gastropods indicate that shell size and shape both influence crushing resistance 1718, we first examine whether size or shape is a better predictor of variation in crushing resistance. Next, we assess potential causes of spatial divergence in Mexipyrgus defensive traits. The abiotic characteristics of Cuatro Ciénegas may be unusually favorable for the production of elaborate snail shells. The spring-fed habitats are geothermally heated and high in dissolved minerals 7. For mollusks, higher temperatures permit elaboration of their shells because calcium carbonate is less soluble in warmer water 1920. Maintaining robust shells is also physiologically cheap in constant aquatic environments with elevated mineral content 2021 especially since Ca++ availability is correlated with the availability of other dissolved minerals in Cuatro Ciénegas 22. We examined whether spatial variation in temperature and conductivity are positively correlated with snail defensive traits.

Alternatively, variation in shell strength might be primarily a response to biotic influences. We address whether spatial variation in snail defenses is associated with two biotic factors: molariform frequency and resource productivity. If molariform H. minckleyi predation is the primary force driving spatial variation in snail defensive traits, we might predict that snail populations with higher frequencies of molariforms should exhibit greater crushing resistance and pigmentation. We also assess whether habitats with greater primary productivity promote escalation of defensive snail phenotypes. We determine whether, in habitats with greater abundance of Nymphaea, Mexipyrgus exhibits more shell pigmentation and crushing resistance. For Mexipyrgus, variation in resource productivity is likely to be driven by the abundance of Nymphaea because in habitats with greater primary productivity, there is a greater abundance of bacteria and fungi in the soft substrates in which Mexipyrgus feeds (Johnson, unpublished results). Increased resource availability probably allows greater investment in costly shell material, leading to the prediction that hotspots may occur in areas with greater food resources for snails.

A critical assumption of these population-based measures of phenotypic divergence is that they are statistically independent. This assumption may be violated because nearby snail populations may experience similar predation pressure because cichlids are more mobile than these brooding snails. To address this issue we assess whether the similarity of shell defensive traits among nearby populations can be attributed to either geographic proximity or genetic similarity 232425. Genealogical relationships within species based on geographic and genetic distances among populations can be used in conjunction with spatial autocorrelation approaches to determine whether defensive structures of nearby populations are positively autocorrelated. A lack of spatial autocorrelation suggests that defenses are distributed in a mosaic-like fashion across the landscape. If phenotypic divergence in crushing resistance and pigmentation is the result of predator-prey interactions with H. minckleyi, and this divergence occurs at very small spatial scales in a geographically mosaic fashion, we predict there will be no significant positive autocorrelation at small spatial and genetic distances. In the current study, we determine genetic distances between snail populations from a previously published study of mtDNA sequence variation in Mexipyrgus 26.

We address three questions concerning the evolution of defenses in Mexipyrgus. First, we examine spatial variation in two putative defensive traits, crushing resistance and shell pigmentation. Then, we tested whether abiotic or biotic variables account for spatial patterns of crushing resistance and shell pigmentation. Finally, we determine whether variation in primary productivity accounts for small-scale variation in these defensive traits.

Pronounced geographic differences in prey defenses are probably common 527. While the unusually high crushing resistance of Mexipyrgus churinceanus in Cuatro Ciénegas and the presence of the molariform cichlids are suggestive of coevolutionary selection 815, the current study provides three important insights. First, there is considerable small-scale spatial variation in crushing resistance and pigmentation that exhibits a mosaic distribution with no significant genetic or spatial autocorrelation. Also, there was a significant negative correlation between the frequency of molariforms and both size-adjusted crushing resistance and frequency of shell pigmentation, but no relationship between these defensive traits and the abiotic variables. Lastly, crushing resistance and pigmentation are significantly higher in habitats dominated by aquatic macrophytes.

To address whether prey defenses have evolved independently across the geographic range of Mexipyrgus churinceanus requires an assessment of phenotypic variation and the spatial autocorrelation of these traits among populations at small genetic and geographic distances. There is significant spatial variation in M. churinceanus crushing resistance and pigmentation at small spatial scales suggesting that certain factors may cause this mosaic distribution of snail defensive traits. In paired populations separated by very small linear and genetic distances, elevated crushing resistance occurs in habitats containing extensive Nymphaea stands. Crushing resistance is dramatically lower in light colored substrates lacking Nymphaea. Because Tio Candido and Rio Mesquites populations represent distinct lineages based on mtDNA sequence differentiation 26, it suggests that these represent convergent patterns of increased crushing resistance and tantalizingly suggests these differences may be driven by resource availability. In habitats with greater primary productivity, there is probably greater abundance of bacteria and fungi upon which Mexipyrgus feeds extensively (Johnson, unpublished results). Increased resource availability likely allows greater investment in costly shell material, and experiments that manipulate resource availability to test its effect on Mexipyrgus shell strength would provide a further test of this hypothesis.

Although there is considerable documentation of hotspots and coldspots in coevolved antagonistic interactions, how spatial processes and community composition promote coevolution and generate selection mosaics has received less attention. A recent study of herbivorous weevils and their Camellia host plant indicate that escalation in armaments only occurred in southern latitudes 3. More northern host populations had reduced resistance but exploitation and damage by weevils was even greater in these populations, suggesting that some unknown ecological factor may constrain resistance in northern host populations. There has been speculation that the exaggerated shell form of Mexipyrgus is the result of coevolutionary interactions between endemic snails and the molluscivorous morph of H. minckleyi 8. Surprisingly, there was a highly significant negative relationship between molariform frequency and size-adjusted crushing resistance among ten populations, suggesting that molariform fitness declines as snail crushing resistance increases. Molariform cichlids may have a greater fitness advantage in resource-poor environments where snail crushing resistance is lower, and the limited availability of plant material reduces papilliform fitness. In Nymphaea habitats, increased snail crushing resistance may reduce molariform fitness due to increased costs of crushing, while papilliform fitness is higher due to increased availability of plant material for shredding. We plan to experimentally test whether these alternative cichlid morphs have fitness trade-offs in different resource environments, and more quantitatively measure resources available to Mexipyrgus.

Microspatial variation in Mexipyrgus shell pigmentation is also associated with different substrate coloration and there is no evidence of genetic or spatial autocorrelation among populations. In the three geographically-paired populations, snails with pigmented bands were significantly more common in Nymphaea habitats and unbanded snails were more common where the benthic substrate was light and marled- colored. Unbanded snails are probably more cryptic against lighter benthic substrates and banded snails are more cryptic against the darker substrates found in association with Nymphaea. This type of background matching by prey under selection from highly visual predators is common in many organisms 282930 and experimental evidence that banding patterns are cryptic against different substrate backgrounds would be an interesting line of further research.

The positive relationship between the two snail traits across environments suggest that pigmentation and crushing resistance may represent phenotypically-correlated traits. In resource-poor environments where Nymphaea is absent, investment in crushing resistance may be constrained due to resource limitation, so that crypsis is the most effective defense against molariform predation. We suspect that investment in shell pigmentation is not as costly as having more robust shells, and that snails are capable of inexpensively modifying their pigmentation in Nymphaea habitats to increase crypsis. Whether correlational selection acts on pigmentation patterns and shell strength in these habitats requires a deeper understanding of the phenotypic plasticity in both traits. One hypothesis is that increased pigmentation in these environments does not increase background matching, and pigmentation only increases due to correlational selection on crushing resistance. This hypothesis could be easily refuted if predation success on banded shells is less in Nymphaea habitats. Given evidence for inducible defenses in snails 313233, we also need to assess the heritability and phenotypic plasticity of crushing resistance and pigmentation.

Spatial variation in two prey defense traits, crushing resistance and shell pigmentation, exhibits striking variation at small spatial scales often associated with habitat differences in primary productivity and substrate coloration. These local geographic differences may result from among-habitat variation in how resource productivity interacts to promote escalation in prey defenses. Because molariform H. minckleyi undoubtedly use their jaws to crush snails, Cuatro Ciénegas may represent an ideal system for investigating mosaic predator-prey coevolution. In future studies, we plan to examine local adaptation of molariforms, escalation of molariform traits in resource rich environments, and heritability and phenotypic plasticity of both crushing resistance and pigmentation across resource gradients.

All authors conceived the study and participated in its design and conducted field work. SGJ performed statistical analysis and drafted the manuscript. CDH and FJGL helped with revisions of the manuscript. All authors read and approved the final manuscript.