DISCUSSION
In this study, we first demonstrated only a large Japanese field mouse,A. speciosus, preyed on snails, but the predation frequency was not correlated with shell colour (Figure 3; Table 1). To clarify the persistence processes of the diversified shell colour in snails, we then compared the colour variation and selection pressures of the snails between the mainland and the island. The frequency distributions of the shell colour showed that the colour variation was monomorphic on the mainland, while it was dimorphic on the island (Figure 2). The mark-recapture experiments showed that the survival rate excluding predation effects of bright-shell snails was the highest on the mainland, where predation pressure is higher than on the island (Figure 4A; Table 2).
The result of the survival rate of bright-shell snails being the highest suggests that weak stabilizing selection acts on shell colour from a factor other than predation on the mainland. This result differs from a previous study, which showed that disruptive selection acts on the shell colour on the island (Ito & Konuma, 2020). In the previous study, the disruptive selection estimated from the Bayesian method did not exclude predation effects. However, the proportion of dead snail shells from predation was relatively low on the island. Thus, it has been suggested that the cause of disruptive selection is also a factor other than predation. Hence, both stabilising selection and disruptive selection are caused by a factor other than predation. The stabilising selection favours bright snails, which could be why the shell colour variation is monomorphic on the mainland. In contrast, the disruptive selection favours dark and bright snails over intermediate-coloured snails, which could be why the shell colour variation is dimorphic on the island. The difference of the selection pressures would then result in the difference of shell colour diversity between the mainland and the island.
Strong disruptive selection acted only on juveniles on the island, and no natural selection acted on adults (Ito & Konuma, 2020). This pattern is opposite to that of the mainland. The survival rate on the mainland was lower among adults than juveniles, whereas that on the island was higher for adults than juveniles. Therefore, natural selection might only act on shell colour when the survival rate is low. The reason why natural selection acted in juvenile only on the island might be that post-zygotic isolation occurred (Wade, 2002; Coughlan & Matute, 2020). The distribution of shell colour was bimodal, in which darker and brighter snails existed sympatrically. The genetic relationship of the snails was close on the island, and hybrids of darker and brighter snails showed intermediate colour (Hayashi & Chiba, 2004; Ito & Konuma, 2020). In contrast, there was a unimodal distribution of shell colour on the mainland, in which almost all of snails were a brighter colour. These results implied that the hybrid between dark and bright shells exists only on the island and that they have a lower survival rate since there is a genetic incompatibility between them.
In adults, one of the hypotheses for the cause of natural selection is that shell colour has a function of thermal regulation (Di Lelliset al. , 2012; Schweizer et al. , 2019). For example, the brighter shell could help to maintain an optimum body temperature when exposed to sunlight in open environments, while a darker shell could be advantageous when exposed to less sunlight in closed environments. The altitude of the distributed area is higher on the islands than the mainland (Figure 1). Thus, this geographical feature of the islands might create complex environmental conditions such as heterogeneity of habitats and lead to shell colour diversification (Ożgo & Schilthuizen, 2012). In contrast, predation from birds is an alternatives hypothesis because bird predators promote shell colour diversification (Allenet al. , 1988; Kraemer et al. , 2019). In a predation scenario, for example, a brighter shell colour has an advantage of camouflage in an open habitat, while a darker shell would have the opposite effect. Although no birds preyed on the snails that we studied, some birds were filmed in our trail-camera experiments. Hence, this possibility is lower than the environmental hypothesis for this snail, as in the case of polymorphism in Cepaea nemoralis (Schilthuizen, 2013).
The high predation pressure would be a primary factor in reducing the survival rates on the mainland. Our trail-camera experiments showed that field mice were the main predators of the snails. The number of the predated snails was not correlated with the shell colour in the mark-recaptured experiments, so the field mice are unlikely to predate snails based on dark or bright shells. In fact, rodents like this mouse do not recognise colour because they are colour blind with two cone types (Conway, 2007), so predation pressure from mice could occur randomly regardless of shell colour (Rosin et al. , 2011). Our results are consistent with this idea, so it is suggested that predation from field mice can be only a factor that reduces the survival. In addition, previous studies reported that field mice are poor at climbing trees and dwell on the ground all year (Imaizumi, 1978; Nishikata, 1981). Thus, the mainland snails would face constant predation risk from field mice on the ground. The adult snails need to come down from trees to lay eggs on the ground (Inoue & Nakada, 1981), so the evidence of predation would have been more frequently observed among adults than juveniles on the mainland.
However, predation could be an indirect driver of persistence in shell-colour monomorphism. Constant predation can result in restricted habitat or microhabitat use to avoid predation (Losos et al. , 2004; Lapiedra et al. , 2018). In such a case, a trait is fitted to the adaptive optimum according to the habitat (Eklöv & Svanbäck, 2006). In contrast, a restricted trait is diversified when a release from predation occurs (Stroud & Losos, 2016). In the Izu Islands, a release from predation could have triggered the use of a variety of habitats or microhabitats and then shell colour diversification.
In conclusion, our findings suggested that monomorphic shell colour persisted in the source population due to stabilising selection. In contrast, disruptive selection acted on shell colour on the island, which was diversified in derived population. Although it is unknown whether natural selection could be changed without predation effects, the direct cause of changing natural selection was a factor other than predation, such as environmental effects. In some land snails on islands, adaptive radiation can occur, and then shell colour is rapidly diversified (Davison & Chiba, 2006; Kraemer et al. , 2019). The change of natural selection from stabilising selection to disruptive selection that we have presented in this study could explain the ultimate factor of such shell colour diversification in the process of adaptive radiation, especially in an initial stage.