Divergent behavior amid convergent evolution: A case of four desert rodents learning to respond to known and novel vipers

Desert communities world-wide are used as natural laboratories for the study of convergent evolution, yet inferences drawn from such studies are necessarily indirect. Here, we brought desert organisms together (rodents and vipers) from two deserts (Mojave and Negev). Both predators and prey in the Mojave have adaptations that give them competitive advantage compared to their middle-eastern counterparts. Heteromyid rodents of the Mojave, kangaroo rats and pocket mice, have fur-lined cheek pouches that allow them to carry larger loads of seeds under predation risk compared to gerbilline rodents of the Negev Deserts. Sidewinder rattlesnakes have heat-sensing pits, allowing them to hunt better on moonless nights when their Negev sidewinding counterpart, the Saharan horned vipers, are visually impaired. In behavioral-assays, we used giving-up density (GUD) to gauge how each species of rodent perceived risk posed by known and novel snakes. We repeated this for the same set of rodents at first encounter and again two months later following intensive “natural” exposure to both snake species. Pre-exposure, all rodents identified their evolutionarily familiar snake as a greater risk than the novel one. However, post-exposure all identified the heat-sensing sidewinder rattlesnake as a greater risk. The heteromyids were more likely to avoid encounters with, and discern the behavioral difference among, snakes than their gerbilline counterparts.


Introduction
Deserts, and desert rodents in particular, provide a model system for studying parallel and 35 convergent evolution. Deserts around the world form at least five evolutionarily independent 36 laboratories of adaptation, ecology, and evolution [1][2][3][4][5][6]. Shared environmental conditions of 37 temperature, precipitation, and aridity force evolutionary processes in a manner that results in 38 similar adaptations in species that fill similar ecological roles. Not only do species converge, but 39 communities may too [7-10]. A good example of this can be studied in desert dunes of the 40 Mojave and of the Negev deserts. In both of these systems we find an array of plants that drop 41 their seeds onto the sand (creating a seed bank); a variety of rodent species feed on these seeds 42 [11][12][13]; and medium-sized sidewinding vipers feed on the rodents [14,15]. 43 The Mojave and Negev deserts of North America and the Middle East, respectively, 44 possess rodents with similar ecologies [5,7,13, 16,17]. These rodents are nocturnal, semi-45 fossorial, seed-eating, and seed caching. However, the heteromyid rodents of the Mojave may 46 have a constraint breaking adaptation compared to their convergent counterparts in the Negev, 47 B l e i c h e r e t a l . I n t e r v i e w s | 4 56 million years from their most recent common ancestor [20], each has evolved the same 57 locomotion method, similar coloration patterns, and a similar adaptation of scales over the eye 58 ridge, protruding as horns. However, the North American sidewinder belongs to the evolutionary 59 lineage of pit-vipers, a lineage that evolved infra-red heat sensing pits. The pit-vipers provide 60 another example of a constraint breaking adaptation compared with Saharan horned vipers. The 61 heat-sensing pits enable the sidewinder to be active on dark nights with no ambient moonlight. 62 The pits also enable safer, more precise strikes at warmer, more vulnerable locations in their 63 endothermic prey [21]. 64 We report here an intercontinental comparison for how two species of Mojave Desert 65 rodents and two species of Negev Desert rodents respond to their evolutionarily and ecologically 66 familiar versus novel snakes. We ask three questions: (1) Do gerbils and heteromyids assess risk 67 from snakes in a similar manner? That is, do they make the same choices when facing snakes 68 with and without heat-sensing pits? (2) Do both sets of rodents assess risk from novel predators 69 as equal to that of evolutionarily familiar ones? (3) Does a prolonged (two-month) exposure to 70 both snakes (in a larger and more realistic setting) diminish the perceived risk of predation from 71 horned vipers, compared with heat-sensing pit-vipers? If so, do both sets of rodents reach the 72 same conclusion, i.e. exhibit the same behavioral response?

74
Study Species 75 We brought together one large and one small coexisting desert rodent from each 76 continent, two common gerbils from the Negev Desert of Israel and a kangaroo rat and a pocket 77 mouse from the deserts of the southwestern United States, to a common and controlled setting in  We used an "interview" approach [27-29] to measure the response of each rodent species 101 to the risks posed by the two snake species. We measured the response prior to exposure of the B l e i c h e r e t a l . I n t e r v i e w s | 6 102 rodents to the novel viper species and following a two-month exposure to both snake species in 103 an semi-natural arena (described in Bleicher 2014; Bleicher et al. 2016). We assessed the 104 response of each species to the vipers using a metric borrowed from foraging theory, the giving-105 up density (GUD; Brown 1988). The GUD is the amount of food a forager leaves behind 106 untouched in a resource patch to measure foraging efficiencies and costs. Most relevant for our 107 purposes is that these costs include those arising from the perceived risk of predation. Hence the 108 forager will leave a lower GUD when it perceives lower risk [31].

109
The experiments were conducted in a light-controlled room at Ben Gurion University of 110 the Negev's Blaustein Institutes for Desert Research, at Midreshed Ben Gurion, Israel (N 111 30°51'17.401", E 34°47'6.637"). We erected a total of eight, 3-compartment (henceforth room) 112 behavioral-assay systems (henceforth system), which we call interview chambers (S1). We called   Last, we averaged the GUD for each individual per snake treatment, resulting in one 162 value for at each test sequence. We then ran in Systat13© a series of generalized linear models 163 (GLM) using the mean GUD as the dependent variable. The first GLM used three independent 164 variables; rodent species, (snake) treatments, and sequence. In addition, all two and three-way 165 interactions between these variables were included. We did not use the full data set, but lowered 166 the "noise" in the data by using the mean values. This normalization means that each individual 167 provides two datapoints, one prior and one post exposure (too low for a meaningful comparison 168 on the individual level). To increase the explanatory power for each species, we ran a GLM for B l e i c h e r e t a l . I n t e r v i e w s | 9 169 each species individually as well. For the single species GLMs, we tested the independent 170 variables: snake treatment and sequence (and the two-way interaction). Post-hoc pairwise 171 comparisons were performed using Tukey's Honestly Significant Differences (THSD) tests for 172 variables that significantly affected variance. This analysis addresses a population-wise (or 173 species-wise in this case) comparison for the broader differences and not in-population variation. 174 We knowingly and purposefully removed individual ID for these reasons.

181
At first encounter, all the species ranked the snakes similarly (Friedman's test of 182 concordance X f 2 = 6.5, 2 df, p= 0.039, and W = 0.813), with lowest GUDs for the snakeless-183 control, higher GUDs for the evolutionarily novel snake, and still higher GUDs for the 184 evolutionarily familiar snake. All rodent species perceived both snakes as threats (p= 0.003, 185 0.012 for known and sidewinders and horned vipers compared to control; Fig 1).

FIGURE 1
187 Desert pocket mice showed increased GUDs in response to snake presence but did not (S2). The non-significant difference between snakes in GP pre-exposure did not alter this 192 finding. Post-exposure, the rodents showed complete concordance according to snake species: 193 they all foraged least in the presence of the rattlesnake. (X f 2 = 6.5, 2 df, p= 0.039, and W= 0.813).
194 TABLE 1 195 Assessing the activity of each species of rodent, using proportion of patches of this 196 treatment in which foraging activity took place, similar patterns emerge. We found that the 197 species each exhibited different activity preferences (Table 1A). None of the rodents foraged a 198 greater proportion of trays before the snake exposure than after. However, for all four rodents the 199 exposure changed the willingness to forage in difference snake treatments (Tables 1 B-E). Pre-200 exposure, three species were active in more compartments with the novel snake than with the 201 evolutionarily familiar one (Fig 2 A). Contrarily, GP investigated more compartments with the 202 familiar snakes than novel snakes, and DM foraged in more compartments with novel snakes 203 than in snake-less compartments. Post-exposure, all rodents foraged most in the snake-free-204 control over the snake treatments (Fig 2 B). Three of the four species foraged least in the 205 compartment with sidewinder rattlesnakes. The GAs, foraged in more horned viper treatments 206 than near sidewinders. GA's activity pattern did not vary between pre-and post-exposure  The GLM combining all four species showed that each species foraged differently in the 211 interview chambers ( Table 2). The heteromyids foraged less than the gerbils. The pocket mice 212 (CP), and kangaroo rats (DM) foraged to a mean GUDs (±SE) of 1.34±0.019g and 1.32±0.019g, B l e i c h e r e t a l . I n t e r v i e w s | 11 213 respectively. Allenby's gerbils (GA) and the Egyptian gerbils (GP) foraged to mean GUD of 214 1.24±0.02g and 1.29±0.027g, respectively. In response to the snake treatments, the rodents 215 overall foraged least in the presence of the rattlesnake, and most in the control treatment (Fig 3,   216 Sup. Appendix 2). Post-hoc pairwise comparison (THSD) found a significant difference 217 between, the control and horned viper (p=0.009), the control and sidewinder (p<0.001), and 218 control and between the horned viper and sidewinder (p=0.006). After two months of exposure, 219 all four species exhibited a similar trend of decreased foraging in the presence of the sidewinder 220 rattlesnake (Fig 3) as shown in each of the single species models (Table 2).  The interpretation these studies gave are based in both locomotion and signaling resulting from B l e i c h e r e t a l . I n t e r v i e w s | 13 258 the kangaroo rats' bipedal agility. On the opposite side the evolutionary strength of the pocket 259 mice is attributed to torpor which they apply to minimize risk and survive harsh weather events.

260
In addition, the kangaroo rats are able to ward off snakes using warning signals, foot drumming verging on being dare-devils, is best exhibited by the increased resource use and patch activity in 278 the treatment with the novel snake (greater than the control). These strong differences in anti-279 predator adaptations, both behavioral and physical, are likely the evolutionary mechanism that 280 allows for these species to coexist in the great basin deserts.

Gerbils 282
The competition between GA and GP is a major model system for the study of the roles of 283 competition, predation risk and parasitism in community structure. It the behavioral differences 284 between these gerbils that allow them to coexist. They differ in habitat preference [44][45][46]

295
This experiment revealed that the gerbils were attentive to the type of predation-risk 296 present and their response to that risk is relatively plastic. Pre-exposure, both gerbils recognized 297 the novel sidewinder, as a risk (higher GUDs than the control) but not as great a risk as the During the pre-exposure interviews all rodents feared their evolutionarily familiar snake 320 equally or more than the novel one. In particular, gerbils showed higher GUDs in response to 321 greater Saharan horned vipers, and the heteromyids showed higher GUDs in response to the 322 sidewinder rattlesnake. This coincides with the snake species that each species evolved with.

323
However, this may also reflect the predator to which each of the rodents has individually been 324 exposed to previously since all animals in these experiments were wild-caught. Overall, the 325 gerbelline species were willing to take more risk investigating the predators, while the 326 heteromyids preferred to avoid both species. This reluctance to take risk in the heteromyids, B l e i c h e r e t a l . I n t e r v i e w s | 16

327
suggests an overall greater "respect" to risk posed by snakes, possibly due to their evolution 328 alongside snakes that have heat-sensing capabilities [32].

329
All four rodents, showed an ability to varying extent, to differentiate between the snake 330 species and to categorize the heat-sensing sidewinder as a greater threat post-exposure. This 331 could be attributed to two explanations. First, the rodents may have learned to identify the musk 332 produced by each species, as kangaroo rats are known to do [36]. Second, given the dark 333 conditions, aimed to highlight the difference between the snakes, the rodents may have been 334 responding most strongly to the sidewinders as they were more active in their cages.

335
The fur-lined cheek pouches, hypothesized to give the heteromyids an advantage in   The rodents were exposed to Saharan horned 547 vipers from the Negev and sidewinder rattlesnakes from the Mojave. In each frame the diagonal 548 striped bar reflects the snake that is evolutionarily novel, the gray bar is the known snake and the 549 white bar is the snake-less control.