Evolutionary fine-tuning of background-matching camouflage Title among geographical populations in the sandy beach tiger beetle
Author(s) Yamamoto, Nayuta; Sota, Teiji
Proceedings of the Royal Society B: Biological Sciences Citation (2020), 287(1941)
Issue Date 2020-12-23
URL http://hdl.handle.net/2433/259830
This is the accepted manuscript of the article, which has been published in final form at http://doi.org/10.1098/rspb.2020.2315; この論文は出版社版 Right でありません。引用の際には出版社版をご確認ご利用く ださい。; This is not the published version. Please cite only the published version.
Type Journal Article
Textversion author
Kyoto University 1 Evolutionary fine-tuning of background-matching camouflage among geographic
2 populations in the sandy beach tiger beetle
3
4
5 Nayuta Yamamoto and Teiji Sota
6
7 Department of Zoology, Graduate School of Science, Kyoto University, Kyoto, Japan
8
9
10 Authors for correspondence: Nayuta Yamamoto and Teiji Sota, Department of Zoology,
11 Graduate School of Science, Kyoto University, Sakyo, Kyoto 606-8502, Japan.
12 e-mail: [email protected]; [email protected].
13
1 14 Abstract
15 Background-matching camouflage is a widespread adaptation in animals; however, few
16 studies have thoroughly examined its evolutionary process and consequences. The tiger
17 beetle Chaetodera laetescripta exhibits pronounced variation in elytral colour pattern
18 among sandy habitats of different colour in the Japanese Archipelago. In this study, we
19 performed digital image analysis with avian vision modelling to demonstrate that elytral
20 luminance, which is attributed to proportions of elytral colour components, is fine-tuned
21 to match local backgrounds. Field predation experiments with model beetles showed that
22 better luminance matching resulted in a lower attack rate and corresponding lower
23 mortality. Using restriction site-associated DNA (RAD) sequence data, we analysed the
24 dispersal and evolution of colour pattern across geographic locations. We found that sand
25 colour matching occurred irrespective of genetic and geographic distances between
26 populations, suggesting that locally adapted colour patterns evolved after the colonisation
27 of these habitats. Given that beetle elytral colour patterns presumably have a quantitative
28 genetic basis, our findings demonstrate that fine-tuning of background-matching
29 camouflage to local habitat conditions can be attained through selection by visual
30 predators, as predicted by the earliest proponent of natural selection.
31
32 Keywords: camouflage, character evolution, colouration, local adaptation
33
2 34 1. Introduction
35 Organismal appearance traits such as body colour pattern are strikingly diverse and have
36 been widely studied as evidence for evolution by natural selection and biodiversity [1–3].
37 Anti-predator adaptation is among the most important evolutionary processes, and is
38 responsible for a myriad of putative camouflage patterns to avoid detection by visually
39 oriented predators [2–6]. A typical form of camouflage comprises cryptic colouration,
40 often referred to as background matching, which conceals individuals through similarities
41 in colouration with their natural background [4,5,7]. A fundamental prediction for this
42 type of camouflage is the evolution of a single optimal body colour that is fine-tuned to
43 match the background colouration, with respect to predator vision, thereby maximising
44 the efficacy of concealment [7–10].
45 Previous studies have provided empirical and comparative evidence to support the
46 hypothesis that selection via predation can lead to background matching in animal
47 colouration (e.g., [11,12]). However, few studies have quantitatively examined how
48 closely prey colour matches the background from the perspectives of predator vision and
49 the efficacy of predation risk reduction in natural conditions [13,14]. A fundamental
50 criterion of camouflage theory is that closer matching of an object to its background will
51 reduce the likelihood that it is found and attacked by predators [5]. Wild animals such as
52 avian predators have visual systems that differ from the human system in terms of the
53 number of receptor types, receptor sensitivity, and ability to perceive ultraviolet (UV)
54 light [15]. Therefore, for effective camouflage, prey animals must closely resemble their
55 background in appearance with respect to predator vision; the resemblance must directly
56 relate to survival against predators in the wild. Recent advances in optical analysis
57 techniques and knowledge of colour perception have allowed modelling of predator
58 vision and consideration of how colouration might be perceived by natural predators
59 [12,16]. Some studies have used vision models to test concealment [17–21]; however,
60 tests for optimal camouflage using actual predators remain rare [13,14]. Therefore, the
61 evolutionary fine-tuning of cryptic colouration to match local backgrounds has not yet
3 62 been fully demonstrated, although it is a fundamental prediction of anti-predator
63 adaptations.
64 Traits with multiple functions (e.g., body colouration) can be affected by both biotic
65 and abiotic selection factors, as well as by phylogenetic constraints [1]. Therefore, it is
66 crucial to examine how different types of selection and phylogenetic constraints have
67 affected the evolution of body colouration, in conjunction with natural selection for
68 camouflage [22,23]. Geographic colour variation provides a good system to study colour
69 matching with spatial replication and to resolve the relative importance of different
70 processes in phenotypic evolution [24]. The study of geographic variation can also
71 provide insight into the population genetics of phenotypic evolution [25]. Background
72 matching can occur by two processes: evolutionary fine-tuning of body colour to match
73 background patterns after the colonisation of a new habitat, and colonisation of a habitat
74 with background patterns that match body colour. The relative importance of these
75 processes can be evaluated by examining the relationships of genetic, geographic, and
76 phenotypic distances among populations.
77 Tiger beetles (Coleoptera: Cicindelidae) are a species-rich group with extreme
78 colour pattern diversity, which provide an intriguing means for investigation of colour
79 pattern evolution [26,27]. Associations between body colouration and environment are
80 observed even among subspecies of tiger beetles, and have long been recognised as
81 predator-avoidance camouflage [3,28]. Previous studies have shown differences in white
82 elytral markings and activity between cool and warm periods among subspecies of tiger
83 beetles, suggesting that thermoregulation may be another important determinant of colour
84 pattern [29,30]. Furthermore, the body colouration of tiger beetles is determined
85 genetically and indicates relatedness among species, suggesting that phylogenetic
86 constraints can affect colouration [27,31,32]. The tiger beetle Chaetodera laetescripta
87 shows pronounced geographic variation in the black and ivory colour patterns of its elytra
88 in the Japanese archipelago (figure 1a), but not in continental Asia [33,34]. In Japan, C.
89 laetescripta elytral colouration resembles the local habitat colouration (figure 1b), which
4 90 varies from off-white to black due to the complicated local geology [34] (figure 1a). From
91 north to south, the Japanese Archipelago is characterised by a climatic gradient that plays
92 an important role in colour divergence among insects [35,36]. Therefore, climatic factors
93 such as temperature and solar radiation may have played a role in the colour pattern
94 divergence of C. laetescripta [37]. The tiger beetle depends on sparsely vegetated sandy
95 habitats; adults emerge during the daytime in summer (June to September) [38] and are
96 therefore likely to be exposed to visually guided predators (e.g., birds and robber flies)
97 on the bare ground, as well as to heat stress associated with their colouration [31].
98 The objectives of this study were to demonstrate the occurrence of fine-tuned
99 background matching, examine the role of selection in camouflage optimisation in terms
100 of geographic colour variation, and explore the evolutionary process of geographic colour
101 variation in the tiger beetle C. laetescripta. First, we assessed how closely the elytral
102 colouration of the tiger beetle resembles its background from the perspective of avian
103 predator vision. Next, we tested the selective advantage of colour matching in artificial
104 models by exposing them to natural avian predators in the wild. Third, we analysed the
105 genetic structures of tiger beetle populations using high-throughput sequencing
106 (restriction site-associated DNA sequencing) [39] to investigate the evolutionary pattern
107 of background matching. Finally, we performed a phylogenetic comparative analysis to
108 test whether variations in colour patterns were primarily determined by background
109 colouration, climatic factors, or phylogenetic effects.
110
111
112 2. Materials and Methods
113 (a) Field sampling
114 C. laetescripta tiger beetles were collected from 12 sites, representing nearly the entire
115 range of their visual appearance and distribution in Japan (figure 1a; electronic
116 supplementary material, table S1). Internal tissues of the beetles were preserved in 99%
117 ethanol at –30°C until DNA extraction; the remaining exoskeletons were stored at –30°C
5 118 and then defrosted for the collection of photographic data. At each site, sand samples were
119 collected at < 2 cm from the surface where the tiger beetles were collected, and then air-
120 dried until they could be photographed.
121
122 (b) Digital image analysis
123 (i) Photographic data collection
124 Photographs of the tiger beetles and sand samples were taken for image analysis (n = 2–
125 20 per site; electronic supplementary material, table S1) using a Nikon D7000 camera
126 that had previously been converted for full-spectrum sensitivity by removal of the internal
127 UV filter. A Nikkor 105-mm lens attached to a Baader UV/IR Cut Filter (transmitting
128 wavelength, 400–700 nm) was used for photographs in the human visible spectrum; the
129 Nikkor 105-mm lens was attached to a Baader UV Pass Filter (300–400 nm) for UV
130 images. Illumination was provided by an MT70D EYE Colour Arc lamp (Iwasaki), from
131 which the UV blocking filter had been removed to allow a full wavelength spectrum. A
132 cylindrical sheet of natural white polytetrafluoroethylene (PTFE) was placed around the
133 sample to diffuse the light evenly. All samples were photographed from a standardised
134 distance and angle with a scale bar and a 99% Spectralon reflectance standard (Labsphere)
135 to normalise images. All images were recorded in RAW format with a fixed aperture
136 (f/10) and ISO sensitivity (400); exposure time was automatically selected for the correct
137 exposure.
138 (ii) Predator vision modelling and camouflage assessment
139 Avian visual modelling and digital image analysis were conducted using customised plug-
140 ins [17] in ImageJ software [40] to assess the efficacy of tiger beetle background-
141 matching camouflage against predators. Multispectral image stacks were first obtained
142 from sets of human visible and UV photographic data of each sample to prepare for image
143 analysis. The position of the reflectance standard was selected manually for each photo
144 for normalisation. As regions of interest, polygons of whole elytra were manually selected
145 for beetles and fixed-diameter circles (around 1000 pixels, equivalent to 15 mm) were
6 146 manually selected for sand samples, using ImageJ selection tools.
147 A cone-mapping model was generated to convert calibrated images from camera
148 photoreceptor values to predator cone-catch values by means of a polynomial mapping
149 technique [17,41] under MT70D lighting conditions in a wavelength band of 300–700
150 nm. Cone sensitivity values for the blue tit (Cyanistes caeruleus) [42] were selected
151 because small birds are the most important predators of tiger beetles [43] and because the
152 blue tit’s visual system is considered representative of those of many UV-sensitive birds
153 [13,42,44]. In addition to birds, robber flies and lizards can also be important predators
154 of tiger beetles [43,45]; however, they are unlikely to affect the elytral colouration of C.
155 laetescripta because robber flies attack beetles in the air during flight [43,45] and there
156 are no or few lizards in the habitats of this tiger beetle species in Japan.
157 Luminance (lightness of colour) perceived by avian predators in the selected region
158 of each elytra and sand image was estimated using the cone-mapping model. Correlations
159 between elytra and sand mean luminance values for each site were examined using
160 Pearson correlation tests in R ver. 3.6.2 software [46]. Just noticeable differences (JNDs)
161 were used, in accordance with the widely used Vorobyev–Oshiro receptor noise
162 discrimination model [47], to evaluate discrimination between elytral and sand
163 luminances. Each JND value was calculated from the log difference in double cone
164 catches of two samples divided by the Weber fraction [48], which was defined as 0.05 in
165 this study because it represents the most abundant cone type [47]. JND < 1.00 indicates
166 that two objects cannot be distinguished even under optimal viewing conditions, while
167 JND < 3.00 indicates that two objects cannot be distinguished under inadequate lighting
168 conditions; JND > 3.00 indicates increasing contrast and greater differences in
169 distinguishability [48].
170 (iv) Calculation of pattern coverage
171 The proportion of white elytral area on tiger beetles was measured using the Natsumushi
172 image measurement software [49] to assess the relationship between luminance and
173 colour pattern. Visible images in RAW format were converted to JPEG format using
7 174 Adobe Photoshop CC software. For each image, the whole elytra area was manually
175 selected as the region of interest using a polygon selection tool in the Natsumushi
176 software; images were then binarised using a global threshold based on Otsu’s
177 discriminate analysis [50]. The proportion of white area was calculated as the ratio of
178 pixels with higher lightness than the threshold value. Pearson’s product moment
179 correlation was used to test correlations between luminance and white elytra areas using
180 R ver. 3.6.2 software [46].
181
182 (d) Field predation experiments
183 (i) Construction of artificial tiger beetles
184 Field predation experiments using paper tiger beetle models were conducted to test for
185 the selective advantage of background matching in luminance. A well-established method
186 for field predation experiments was followed, in which a piece of printed paper and an
187 edible mealworm were used [8,51,52]. Two representative habitats showing a large
188 difference in beetle elytral colour pattern and sand colour were selected as experimental
189 sites (Site J: light background, Site F: dark background; figure 1a). Light (60% of the
190 elytral area was white) and dark models (10%) were constructed that were 15 mm in
191 length, corresponding to the adult body length of the tiger beetle [53]. The colour patterns
192 were printed using a Canon TS 9030 inkjet printer on pieces of Whatman filter paper (no.
193 1001–240), which has neutral reflective properties and is often used in predation research
194 [13,54]. The image analysis described above was conducted to evaluate differences in
195 luminance between the paper model and background sand for avian viewing.
196 Photographic data were obtained for both models (n = 12 per type) and sand samples
197 collected from the experiment sites (n = 12); JND values were calculated between models
198 and sand samples. The paper model was attached to a black straw (15 mm in length, 3.5
199 mm in diameter) using double-sided tape; a dead mealworm (Tenebrio molitor larva; 10–
200 15 mm in length; frozen overnight and then thawed) was pinned into the straw with its
201 tail protruding approximately 5 mm from the end.
8 202 (ii) Model presentation
203 Experimental trials were conducted at both sites (Site J and F) from August to September
204 2018. At each site, three transects were simultaneously set at a distance of > 100 m apart.
205 Twenty-five models of each type were alternately placed at distances of > 2 m apart on
206 each transect. A model horizontal was fixed to the ground surface at a height of
207 approximately 5 mm by using a thin wire stuck into the ground. Model presentation trials
208 began at 4:00 AM and 4:00 PM (Japan Standard Time) and were completed within 30
209 min; the survivorship of the models was examined 3 h later. This procedure was repeated
210 for 5 consecutive days, except on rainy days. During the experiments, insect-eating birds
211 including sparrows (Passer montanus), meadow buntings (Emberiza cioides), and crows
212 (Corvus corone) were commonly observed to walk and search for prey on the ground at
213 both sites (electronic supplementary material, table S2); the models were attacked upon
214 disappearance of the mealworms. The final sample size was n = 2983 (Site J: 747 of each
215 model; Site F: 743 and 746 light and dark models, respectively) after exclusion of data
216 for 19 models (Site J: 3 light, 3 dark; Site F: 7 light, 4 dark) that had been lost or swarmed
217 by non-avian foragers such as ants.
218 (iii) Statistical analyses
219 Generalised linear mixed model analysis with binomial error distribution was conducted
220 to examine the effect of background resemblance on the model attack rate. Whether
221 models were attacked or not (1 or 0) was used as the independent variable; model colour
222 (light or dark) was used as the explanatory variable. Start time (4:00 AM or 4:00 PM),
223 transect number (1–3), and days (1–5) were included as random effects. These analyses
224 were conducted in R ver. 3.6.2 [46] using the ‘glmer’ function in the lme4 package [55]
225 and the ImerTest package [56] for p value calculation.
226
227 (c) Phylogenetic analysis with RAD sequencing
228 (i) RAD sequencing and phylogenetic tree estimation
229 Four to five individuals in each population from a total of 13 sites were used, including
9 230 samples from South Korea as the outgroup (figure 1a; electronic supplementary material,
231 table S1). Total genomic DNA was extracted from thorax muscles or testicles using a
232 DNeasy Blood & Tissue Kit (Qiagen). For each sample, approximately 25 ng/μL (10–30
233 ng/μL) of genomic DNA was digested using the restriction enzyme PstI in NEB Buffer4,
234 then ligated with a P1 adaptor and a unique five-base sequence. Library construction and
235 single-end 101-bp sequencing were performed using a HiSeq 2500 sequencer (Illumina)
236 at Hokkaido System Science, Sapporo, Japan.
237 The RAD sequence data were processed using the ipyrad ver. 0.9.31 software [57].
238 Sequences were identified as homologous and clustered when they had greater than 90%
239 similarity. Loci with sequences that were shared by more than 55 samples were retained
240 in the final dataset. The phylogenetic tree was estimated using the maximum likelihood
241 (ML) method using the RAxML ver. 8.2.12 software [58]. A general time-reversible
242 model with gamma probability distribution was used as the evolutionary model; node
243 support values were calculated by rapid bootstrapping analysis with 100 replications [59].
244 (ii) Mantel test and phylogenetic comparative analysis
245 The relationships among inter-population differences in body colour, genetic distance,
246 and geographical distance were investigated by means of the Mantel test [60] with 1000
247 repetitions using the ‘Mantel’ function in the ape R package [61] in R ver. 3.6.2 [46]. The
248 mean proportion of the white area in tiger beetles was calculated for each site in Japan
249 (electronic supplementary material, table S1); pairwise Euclidean distances between the
250 mean values were obtained in R. Genetic distance was calculated as pairwise FST between
251 population samples by means of the ‘fst’ function in the GENEPOP software package
252 [62], using files converted from ipyrad output using the PGDSpider conversion tool [63].
253 Geographical distance was calculated as the great circle distance (km) among sites using
254 latitude and longitude information for each site and the ‘distm’ function in the geosphere
255 R package [64].
256 The effects of background colouration and thermal environment on inter-population
257 differences in colour pattern were also examined by using phylogenetic generalised least-
10 258 squares (PGLS) analysis based on a Brownian motion model of trait evolution [65,66].
259 For the PGLS analysis, one individual per site, which had the lowest proportion of
260 missing RAD loci among individuals from the same site, was selected, and an ML
261 phylogenetic tree was constructed using the general time-reversible model with gamma
262 probability distribution in the RAxML software. The tree was converted into ultrametric
263 form using the penalised likelihood method with the ‘chronos’ function in the ape R
264 package. As climatic factors, mean temperature (°C) and mean global solar radiation (MJ
265 m–2) data for the adult activity period (June–September) [38] were used; these were
266 obtained from 1-km mesh meteorological data (Mesh Climatic Data 2000; Japan
267 Meteorological Business Support Center; electronic supplementary material, table S1).
268 The effects of hypothetical factors (mean sand luminance and the two climatic factors)
269 on the mean values of white area proportion (%) were examined by the PGLS analysis
270 using the ‘gls’ function in the nlme R package [67]. Concomitantly, the likelihoods of
271 these models were compared using Akaike’s information criterion values.
272
273
274 3. Results
275 (a) Body colour pattern and sand colour
276 We collected photographic data from 193 tiger beetles and sand samples taken at 12 sites
277 (figure 1a; electronic supplementary material, table S1 and S3). The mean luminance
278 values of elytra and sand at each site varied geographically and were correlated with each
279 other across populations (Pearson's product moment correlation: r = 0.86, P < 0.001;
280 figure 2a). The average JND between tiger beetle body and sand luminance was as low
281 as 1–5 at most sites. Thus, tiger beetles had site-specific body colouration patterns that
282 were well matched to the background colour in terms of avian vision.
283 The elytral white area proportion varied from 6.1% to 74.1% (electronic
284 supplementary material, table S3) and increased with increasing luminance (r = 0.96, P
285 < 0.001; figure 2b). Thus, elytral luminance was determined by the extent of white colour
11 286 marking among elytra.
287
288 (b) Camouflage effect of luminance matching
289 JND values between the luminance of two tiger beetle models used in the field experiment
290 were greater than 3 (mean ± standard deviation: 27.8 ± 1.64), suggesting that avian
291 predators were able to easily distinguish the models according to luminance. In addition,
292 the light model had a lower JND value on light sand (5.21 ± 2.19) than the dark model
293 (34.2 ± 3.84), whereas the dark model had a higher JND value on light sand (22.6 ± 2.56)
294 than the dark model (6.40 ± 4.02). Therefore, we presume that predators were able to
295 easily distinguish the colour of one model, but not the other, from the sand colour.
296 In total, 127 models were attacked at the light sand site (Site J) and 150 at the dark
297 sand site (Site F) throughout the experimental period (electronic supplementary material,
298 table S4). Light models were attacked significantly less frequently than dark models at
299 the light background site (generalised linear mixed model: z = –2.790, P = 0.005; figure
300 3, left), but more frequently than dark models at the dark background site (z = 2.521, P =
301 0.012; figure 3, right). Thus, colour patterns that better matched the background were
302 associated with higher survival rates.
303
304 (c) Colour pattern evolution
305 We obtained a total of 2.6 million RAD sequences from 64 individuals. The ML tree
306 showed monophyly of each population with > 80% bootstrap support (figure 4a). The
307 pairwise genetic distance (FST) increased with the pairwise geographic distance among
308 sampling sites (Mantel test: P = 0.009; figure 4b), revealing an isolation-by-distance
309 pattern [68]. However, genetic distance was not related to the difference in the mean
310 proportion of white area (P = 0.375; figure 4c).
311 The best model, with the smallest Akaike’s information criterion value, for the
312 proportion of white elytral area was the model that included only mean sand luminance
313 as a predictor variable, which had a positive effect on the proportion of white area (PGLS:
12 314 t = 4.951, P = 0.0006, table 1; figure 4d). Climatic factors had no significant effects on
315 the proportion of white area among models that included climatic variables (table 1).
316
317
318 4. Discussion
319 (a) Fine-tuning of body colour to local backgrounds for camouflage
320 In this study, we demonstrated that the elytral colour of C. laetescripta closely matched
321 the luminance of the local background substrate in the context of avian vision (figure 2a),
322 suggesting local adaptation for predation avoidance. The field predation experiment
323 showed that better-matched models were attacked less frequently (figure 3), suggesting
324 that better-matched colouration is effective for reduction of predation risk. These results
325 implied that predation-induced selection of elytral colour resulted in single optimal
326 colourations that were finely tuned to each visual background, supporting the
327 fundamental prediction of camouflage theory [5]. Our results confirmed the role of cryptic
328 colouration in the tiger beetles, which was initially documented by Alfred Russel Wallace
329 in the 19th century [3,28,69,70].
330 The proportion of white maculation area showed continuous variation among
331 individuals and a robust positive correlation with elytral luminance (figure 2b). The
332 maculation patterns of tiger beetles are caused by the absence of the black pigment
333 melanin and are determined genetically [31]. Thus, maculation pattern variation likely
334 has a polygenic basis, which allows beetle populations to evolve towards optimal
335 luminance against a wide range of background colouration. In nature, animal camouflage
336 with genetically determined colouration may be more widespread than the camouflage
337 determined by morphological or physiological colour changeability, which is observed in
338 cuttlefish, cephalopods, and crabs [71]. However, most studies of animal camouflage
339 showing fine-tuning to local backgrounds have dealt with species exhibiting colour
340 changeability [12,14]; the exceptions are studies of melanin-pigmented fur or skin among
341 vertebrates [21,72]. Thus, our study is one of few that have demonstrated finely tuned
13 342 background-matching camouflage in detail in a species with genetically determined
343 colouration. Notably, we demonstrated that camouflage colouration is adjusted by the
344 dark/light colour ratio, which is a common aspect of animal appearance and can also be
345 related to thermoregulation and aposematism [73,74].
346 Body colouration patterns can also influence camouflage in various ways. The
347 degree of matching in colour pattern spatial characteristics (e.g., grain and geometry) to
348 background can also affect survival [5,7,75]. In addition, background-matching
349 camouflage may contain disruptive colouration, which can interfere with the image of the
350 animal as detected by the predator, thereby enhancing predator avoidance [76]. In tiger
351 beetles, the intricate boundary between high-contrast patterns (figure 1) could function as
352 disruptive colouration, whereby some markings create false edges or blend into the
353 background [70,77]. These colour pattern components may have affected beetle survival
354 rates in a manner similar to that of elytral colour luminance (figure 2a, 3), which was a
355 major component of background-matching camouflage.
356
357 (b) Evolution of adaptive colouration
358 Despite the potential importance of body surface brightness in thermoregulation [78,79],
359 our PGLS analysis detected no significant effects of climatic factors on colour pattern
360 evolution (figure 4). This result emphasises the importance of selection pressure for
361 camouflage as a factor contributing to substantial variation in colour pattern.
362 Thermoregulation can be accomplished by various behaviours [78,80] that can surpass
363 the effect of colouration [80]. Behaviours promoting thermoregulation (e.g., basking or
364 shading) are commonly observed in various tiger beetle species including C. laetescripta
365 [27,29,45,81]. Therefore, C. laetescripta may rely on behavioural thermoregulation;
366 selection on body colour for thermoregulatory function may be negligible.
367 Although C. laetescripta occurs widely in continental Asia, it shows pronounced
368 elytral colour variation only among Japanese populations [33,34]. Therefore, it is likely
369 that colour variation originated in this species after its colonisation of Japan from
14 370 continental Asia, which is presumed to have occurred during the early to middle
371 Pleistocene [82,83]. We examined population genetic structure and found a significant
372 isolation-by-distance pattern [68] and correlation between geographic and genetic
373 distances (figure 4b), but not between phenotypic and genetic distances (figure 4c). These
374 results suggest that dispersal and colonisation have occurred somewhat gradually (i.e.,
375 stepping-stone dispersal) regardless of habitat sand colour, although local adaptation can
376 restrict the dispersal of individuals to ecologically dissimilar habitats [25,84]. In C.
377 laetescripta, colonisation success may have primarily depended on factors other than sand
378 colour, such as the presence of suitable microhabitats for the larval stage [85]; the
379 evolution of observed patterns of colour variation may have followed colonisation under
380 selection pressure from visual predation on elytral colour. Although previous studies of
381 cryptic colouration have investigated clinal variation among populations spanning less
382 than a few hundred km [86,87], our study focused on populations with discontinuous
383 colour variation scattered over an area 1400 km in width (figure 1a). Thus, the findings
384 of our study demonstrate that local adaptation has played a major role in the evolution of
385 background-matching colouration across a wide geographic range of a species. Similar
386 cases of locally fine-tuned camouflage have been suggested in other tiger beetles with
387 wide geographic ranges [69,88].
388
389 5. Conclusion
390 In this study, we applied integrated image analysis, field experiments, and phylogenetic
391 analysis to demonstrate that selection via predation has driven the evolution of finely
392 tuned background-matching camouflage in the tiger beetle C. laetescripta, following the
393 colonisation of different habitats. Our results showed that fine-tuned background
394 matching could be achieved by an evolutionary change in body colour driven by natural
395 selection from visual predators, providing new insight into the evolutionary process and
396 consequences of animal camouflage. These findings will deepen our understanding of the
397 generation and maintenance of great diversity in organismal appearance in nature.
15 398 Ethics
399 We followed the ethical guidelines for the treatment of animals in the study of animal
400 behaviour [89] in our field predation experiments.
401
402 Data accessibility
403 The raw RAD sequence data are deposited at DNA Data Bank of Japan (DDBJ):
404 BioProject accession number PRJDB10308 (SAMD00239266–SAMD00239329).
405
406 Authors' contributions
407 N.Y. and T.S. conceived the project; N.Y. performed sampling, experiments and data
408 analyses; N.Y. and T.S. wrote the manuscript.
409
410 Competing interests
411 We declare we have no competing interests.
412
413 Funding
414 This research was supported by JSPS KAKENHI (grant numbers: 15H02637, 18H04010
415 to T.S.).
416
417 Acknowledgements
418 We thank N. Tsurusaki, T. Ishikawa, S. Ino, Y. Kishida, S. Shiraiwa, M. Ujiie, A. Mizuta,
419 and T. Yamamoto for sampling; and K. Watanabe, S. Yamamoto, M. Hori, and Y. Enokido
420 for discussion. We are also grateful to Hitachi Seaside Park Office (Ministry of Land,
421 Infrastructure and Transport), Izu Islands Park Ranger Office (Ministry of the
422 Environment), Mie Prefectural Division of Environment, Uratomi Park Ranger Office
423 (Ministry of the Environment), Agency for Cultural Affairs, Tottori Prefectural Board of
424 Education, Tottori City Board of Education, and Uminonakamichi Seaside Park Office
425 (Ministry of Land, Infrastructure and Transport) for research permission.
16 426 References
427 1. Cuthill IC et al. 2017 The biology of color. Science. 357, eaan0221.
428 (doi:10.1126/science.aan0221)
429 2. Darwin E. 1794 Zoonomia; or The laws of organic life. London, UK: J. Johnson.
430 (doi:10.5962/bhl.title.25710)
431 3. Wallace AR. 1867 Mimicry and other protective resemblances among animals.
432 Westminster Foreign Q. Rev. 32, 1–43.
433 4. Cott HB. 1940 Adaptive coloration in animals. London, UK: Academic Press
434 Inc.
435 5. Stevens M, Merilaita S. 2011 Animal camouflage: mechanisms and function.
436 Cambridge, UK: Cambridge University Press.
437 6. Poulton EB. 1980 The colours of animals, their meaning and use, especially
438 considered in the case of insects. New York, NY: D. Appleton and Company.
439 7. Endler JA. 1978 A predator’s view of animal color patterns. Evol. Biol. 11, 319–
440 364. (doi:10.1007/978-1-4615-6956-5_5)
441 8. Michalis C, Scott-Samuel NE, Gibson DP, Cuthill IC. 2017 Optimal background
442 matching camouflage. Proc. R. Soc. B Biol. Sci. 284, 20170709.
443 (doi:10.1098/rspb.2017.0709)
444 9. Stevens M, Stevens M, Merilaita S. 2009 Animal camouflage: current issues and
445 new perspectives. Philos. Trans. R. Soc. London B Biol. Sci. 364, 423–427.
446 (doi:10.1098/rstb.2008.0217)
447 10. Thayer AH. 1918 Camouflage. Sci. Mon. 7, 481–494.
448 11. Vignieri SN, Larson JG, Hoekstra HE. 2010 The selective advantage of crypsis in
449 mice. Evolution. 64, 2153–2158. (doi:10.1111/j.1558-5646.2010.00976.x)
450 12. Stevens M, Broderick AC, Godley BJ, Lown AE, Troscianko J, Weber N, Weber
451 SB. 2015 Phenotype-environment matching in sand fleas. Biol. Lett. 11,
452 20150494. (doi:10.1098/rsbl.2015.0494)
453 13. Walton OC, Stevens M. 2018 Avian vision models and field experiments
17 454 determine the survival value of peppered moth camouflage. Commun. Biol. 1,
455 118. (doi:10.1038/s42003-018-0126-3)
456 14. Duarte RC, Stevens M, Flores AAV. 2018 The adaptive value of camouflage and
457 colour change in a polymorphic prawn. Sci. Rep. 8, 16028. (doi:10.1038/s41598-
458 018-34470-z)
459 15. Cuthill IC. 2006 Color perception. In Bird coloration (eds GE Hill, KJ McGraw),
460 pp. 3–40. Cambridge, UK: Harvard University Press.
461 16. Johnsen S. 2016 How to measure color using spectrometers and calibrated
462 photographs. J. Exp. Biol. 219, 772–778. (doi:10.1242/jeb.124008)
463 17. Troscianko J, Stevens M. 2015 Image calibration and analysis toolbox – a free
464 software suite for objectively measuring reflectance, colour and pattern. Methods
465 Ecol. Evol. 6, 1320–1331. (doi:10.1111/2041-210X.12439)
466 18. Théry M, Casas J. 2002 Predator and prey views of spider camouflage. Nature
467 415, 133–133. (doi:10.1038/415133a)
468 19. Price N, Green S, Troscianko J, Tregenza T, Stevens M. 2019 Background
469 matching and disruptive coloration as habitat-specific strategies for camouflage.
470 Sci. Rep. 9, 7840. (doi:10.1038/s41598-019-44349-2)
471 20. Cournoyer BL, Cohen JH. 2011 Cryptic coloration as a predator avoidance
472 strategy in seagrass arrow shrimp colormorphs. J. Exp. Mar. Bio. Ecol. 402, 27–
473 34. (doi:10.1016/J.JEMBE.2011.03.011)
474 21. Stuart-Fox DM, Moussalli A, Johnston GR, Owens IPF. 2004 Evolution of color
475 variation in dragon lizards: Quantitative tests of the role of crypsis and local
476 adaptation. Evolution. 58, 1549–1559. (doi:10.1111/j.0014-3820.2004.tb01735.x)
477 22. Marshall KLA, Philpot KE, Stevens M. 2015 Conspicuous male coloration
478 impairs survival against avian predators in Aegean wall lizards, Podarcis
479 erhardii. Ecol. Evol. 5, 4115–4131. (doi:10.1002/ece3.1650)
480 23. Mullen LM, Vignieri SN, Gore JA, Hoekstra HE. 2009 Adaptive basis of
481 geographic variation: genetic, phenotypic and environmental differences among
18 482 beach mouse populations. Proceedings. Biol. Sci. 276, 3809–18.
483 (doi:10.1098/rspb.2009.1146)
484 24. Miller ET, Leighton GM, Freeman BG, Lees AC, Ligon RA. 2019 Ecological
485 and geographical overlap drive plumage evolution and mimicry in woodpeckers.
486 Nat. Commun. 10, 1602. (doi:10.1038/s41467-019-09721-w)
487 25. Orsini L, Vanoverbeke J, Swillen I, Mergeay J, De Meester L. 2013 Drivers of
488 population genetic differentiation in the wild: isolation by dispersal limitation,
489 isolation by adaptation and isolation by colonization. Mol. Ecol. 22, 5983–5999.
490 (doi:10.1111/mec.12561)
491 26. Shelford VE. 1917 Color and color-pattern mechanism of tiger beetles. Illinois
492 Biol. Monogr. 3, 395–592.
493 27. Pearson DL, Vogler AP. 2001 Tiger beetles, the evolution, ecology, and diversity
494 of the cicindelids. New York, NY: Cornell University Press.
495 28. Townsend CHT. 1886 Protective coloration in the genus Cicindela. Can.
496 Entomol. 18, 46–49. (doi:10.4039/Ent1846-3)
497 29. Acorn JH. 1992 The historical development of geographic color variation among
498 dune Cicindela in western Canada. In The Biogeography of Ground Beetles of
499 Mountains and Islands (eds GR Noonan, GE Ball, NE Stork), pp. 217–233.
500 Andover, UK: Intercept.
501 30. Schultz TD, Hadley NF. 1987 Structural colors of tiger beetles and their role in
502 heat transfer through the integument. Physiol. Zool. 60, 737–745.
503 (doi:10.2307/30159990)
504 31. Knisley CB, Schultz TD. 1997 The biology of tiger beetles and a guide to the
505 species of the South Atlantic states. Virginia, VA: Virginia Museum of Natural
506 History.
507 32. Tsuji K, Hori M, Phyu MH, Liang H, Sota T. 2016 Colorful patterns indicate
508 common ancestry in diverged tiger beetle taxa: Molecular phylogeny,
509 biogeography, and evolution of elytral coloration of the genus Cicindela
19 510 subgenus Sophiodela and its allies. Mol. Phylogenet. Evol. 95, 1–10.
511 (doi:10.1016/j.ympev.2015.11.006)
512 33. Ashida H. 1998 Geographical distribution and variation of Chaetodera
513 laetescripta (Coleoptera : Cicindelidae) in Japan. Gekkan-Mushi 333, 5–9. (In
514 Japanese)
515 34. Enokido Y. 1991 Notes on the observation and collection of Cicindelidae. Nature
516 and Insects 26, 16–21. (In Japanese)
517 35. Tsurui K, Honma A, Nishida T. 2010 Camouflage effects of various colour-
518 marking morphs against different microhabitat backgrounds in a polymorphic
519 pygmy grasshopper Tetrix japonica. PLoS One 5, e11446.
520 (doi:10.1371/journal.pone.0011446)
521 36. Noriyuki S, Osawa N. 2015 Geographic variation of color polymorphism in two
522 sibling ladybird species, Harmonia yedoensis and H. axyridis (Coleoptera:
523 Coccinellidae). Entomol. Sci. 18, 502–508. (doi:10.1111/ens.12147)
524 37. Satoh M. 1988 Chaetodera laetescripta. Insectarium 25, 301. (In Japanese)
525 38. Ohno M. 2000 Bibliographical monograph of the selected important animals
526 occurring in Japan (54) Chaetodera laetescripta (1). J. Toyo Univ. Gen. Educ. ,
527 163–201. (In Japanese)
528 39. Baird NA, Etter PD, Atwood TS, Currey MC, Shiver AL, Lewis ZA, Selker EU,
529 Cresko WA, Johnson EA. 2008 Rapid SNP discovery and genetic mapping using
530 dequenced RAD markers. PLoS One 3, e3376.
531 (doi:10.1371/journal.pone.0003376)
532 40. Schneider CA, Rasband WS, Eliceiri KW. 2012 NIH Image to ImageJ: 25 years
533 of image analysis. Nat. Methods 9, 671–5.
534 41. Stevens M, Parraga C a, Cuthill IC, Partridge JC, Troscianko TS. 2007 Using
535 digital photography to study animal coloration. Biol. J. Linn. Soc. 90, 211–237.
536 (doi:Doi 10.1111/J.1095-8312.2007.00725.X)
537 42. Hart NS, Partridge JC, Cuthill IC, Bennett ATD. 2000 Visual pigments, oil
20 538 droplets, ocular media and cone photoreceptor distribution in two species of
539 passerine bird: the blue tit (Parus caeruleus L.) and the blackbird (Turdus merula
540 L.). J. Comp. Physiol. A Sensory, Neural, Behav. Physiol. 186, 375–387.
541 (doi:10.1007/s003590050437)
542 43. Pearson DL. 1985 The function of multiple anti-predator mechanisms in adult
543 tiger beetles (Coleoptera: Cicindelidae). Ecol. Entomol. 10, 65–72.
544 44. Stoddard MC, Stevens M. 2011 Avian vision and the evolution of egg color
545 mimicry in the common cuckoo. Evolution. 65, 2004–2013. (doi:10.1111/j.1558-
546 5646.2011.01262.x)
547 45. Kameyama T, Kato Y. 2019 Chaetodera laetescripta in Mie Pref. In Tiger beetles
548 of Japan -Phylogeny, ecology and conservation- (ed M Hori), pp. 206–236.
549 Tokyo, Japan: Hokuryukan. (In Japanese)
550 46. R Core Team. 2019 R: A language and environment for statistical computing. R
551 Found. Stat. Comput. See https://www.r-project.org.
552 47. Vorobyev M, Osorio D. 1998 Receptor noise as a determinant of colour
553 thresholds. Proc. R. Soc. B Biol. Sci. 265, 351–358.
554 (doi:10.1098/rspb.1998.0302)
555 48. Siddiqi A, Cronin TW, Loew ER, Vorobyev M, Summers K. 2004 Interspecific
556 and intraspecific views of color signals in the strawberry poison frog
557 Dendrobates pumilio. J. Exp. Biol. 207, 2471–85. (doi:10.1242/jeb.01047)
558 49. Tanahashi M, Fukatsu T. 2018 Natsumushi: Image measuring software for
559 entomological studies. Entomol. Sci. 21, 347–360. (doi:10.1111/ens.12315)
560 50. Otsu N. 1979 A threshold selection method from gray-level histograms. IEEE
561 Trans. Syst. Man. Cybern. 9, 62–66. (doi:10.1109/TSMC.1979.4310076)
562 51. Cuthill IC, Stevens M, Sheppard J, Maddocks T, Párraga CA, Troscianko TS.
563 2005 Disruptive coloration and background pattern matching. Nature 434, 72–74.
564 (doi:10.1038/nature03312)
565 52. Stevens M, Cuthill IC, Windsor AM., Walker HJ. 2006 Disruptive contrast in
21 566 animal camouflage. Proc. R. Soc. B Biol. Sci. 273, 2433–2438.
567 (doi:10.1098/rspb.2006.3614)
568 53. Ueno S-I, Kurosawa Y, Satoh M. 1985 The Coleoptera of Japan in color, Vol. II.
569 Osaka: Hoikusha. (In Japanese)
570 54. Finkbeiner SD, Salazar PA, Nogales S, Rush CE, Briscoe AD, Hill RI, Kronforst
571 MR, Willmott KR, Mullen SP. 2018 Frequency dependence shapes the adaptive
572 landscape of imperfect batesian mimicry. Proc. R. Soc. B Biol. Sci. 285,
573 20172786. (doi:10.1098/rspb.2017.2786)
574 55. Bates D, Mächler M, Bolker BM, Walker SC. 2015 Fitting linear mixed-effects
575 models using lme4. J. Stat. Softw. 67, 1–48. (doi:10.18637/jss.v067.i01)
576 56. Kuznetsova A, Brockhoff PB, Christensen RHB. 2017 lmerTest package: tests in
577 linear mixed effects models. J. Stat. Softw. 82, 1–26. (doi:10.18637/jss.v082.i13)
578 57. Eaton DAR, Overcast I. 2020 ipyrad: Interactive assembly and analysis of
579 RADseq datasets. Bioinformatics 36, 2592–2594.
580 (doi:10.1093/bioinformatics/btz966)
581 58. Stamatakis A. 2014 RAxML version 8: a tool for phylogenetic analysis and post-
582 analysis of large phylogenies. Bioinformatics 30, 1312–1313.
583 (doi:10.1093/bioinformatics/btu033)
584 59. Stamatakis A, Hoover P, Rougemont J. 2008 A rapid bootstrap algorithm for the
585 RAxML web servers. Syst. Biol. 57, 758–771.
586 (doi:10.1080/10635150802429642)
587 60. Mantel N. 1967 The detection of disease clustering and a generalized regression
588 approach. Cancer Res. 27, 209–220.
589 61. Paradis E, Claude J, Strimmer K. 2004 APE: Analyses of phylogenetics and
590 evolution in R language. Bioinformatics 20, 289–290.
591 (doi:10.1093/bioinformatics/btg412)
592 62. Rousset F. 2008 GENEPOP’007: A complete re-implementation of the
593 GENEPOP software for Windows and Linux. Mol. Ecol. Resour. 8, 103–106.
22 594 (doi:10.1111/j.1471-8286.2007.01931.x)
595 63. Lischer HEL, Excoffier L. 2012 PGDSpider: An automated data conversion tool
596 for connecting population genetics and genomics programs. Bioinformatics 28,
597 298–299. (doi:10.1093/bioinformatics/btr642)
598 64. Hijmans RJ. 2019 geosphere: spherical trigonometry. R Packag. version 1.5-10.
599 See https://cran.r-project.org/package=geosphere.
600 65. Grafen A. 1989 The phylogenetic regression. Philos. Trans. R. Soc. B Biol. Sci.
601 326, 119–157. (doi:10.1098/rstb.1989.0106)
602 66. Freckleton RP, Harvey PH, Pagel M. 2002 Phylogenetic analysis and
603 comparative data: a test and review of evidence. Am. Nat. 160, 712–26.
604 (doi:10.1086/343873)
605 67. Pinheiro J, Bates D, DebRoy S, Sarkar D, Team RC. 2019 nlme: linear and
606 nonlinear mixed effects models. R Packag. version 3.1-141. See https://cran.r-
607 project.org/package=nlme.
608 68. Wright S. 1943 Isolation by distance. Genetics 28, 114–138. (doi:10.1016/B978-
609 0-12-374984-0.00820-2)
610 69. Hadley NF, Schultz TD, Savill A. 1988 Spectral reflectances of three tiger beetle
611 subspecies (Neocicindela perhispida): correlations with their habitat substrate.
612 New Zeal. J. Zool. 15, 343–346. (doi:10.1080/03014223.1988.10422624)
613 70. Schultz TD. 1986 Role of structural colors in predator avoidance by tiger beetles
614 of the genus Cicindela (Coleoptera: Cicindelidae). Bull. Entomol. Soc. Am. 32,
615 142–146. (doi:10.1093/besa/32.3.142)
616 71. Duarte RC, Flores AA V, Stevens M. 2017 Camouflage through colour change:
617 mechanisms, adaptive value and ecological significance. Philos. Trans. R. Soc.
618 Lond. B. Biol. Sci. 372, 20160342. (doi:10.1098/rstb.2016.0342)
619 72. Hoekstra HE. 2006 Genetics, development and evolution of adaptive
620 pigmentation in vertebrates. Heredity (Edinb). 97, 222–234.
621 (doi:10.1038/sj.hdy.6800861)
23 622 73. Hegna RH, Nokelainen O, Hegna JR, Mappes J. 2013 To quiver or to shiver:
623 increased melanization benefits thermoregulation, but reduces warning signal
624 efficacy in the wood tiger moth. Proc. R. Soc. B Biol. Sci. 280, 20122812.
625 (doi:10.1098/rspb.2012.2812)
626 74. Kawakami Y, Yamazaki K, Ohashi K. 2013 Geographical variations of elytral
627 color polymorphism in Cheilomenes sexmaculata (Fabricius) (Coleoptera:
628 Coccinellidae). Entomol. Sci. 16, 235–242. (doi:10.1111/ens.12005)
629 75. Merilaita S, Dimitrova M. 2014 Accuracy of background matching and prey
630 detection: predation by blue tits indicates intense selection for highly matching
631 prey colour pattern. Funct. Ecol. 28, 1208–1215. (doi:10.1111/1365-2435.12248)
632 76. Merilaita S, Lind J. 2005 Background-matching and disruptive coloration, and
633 the evolution of cryptic coloration. Proc. R. Soc. B Biol. Sci. 272, 665–670.
634 (doi:10.1098/rspb.2004.3000)
635 77. Stevens M, Merilaita S. 2009 Defining disruptive coloration and distinguishing
636 its functions. Philos. Trans. R. Soc. B Biol. Sci. 364, 481–488.
637 (doi:10.1098/rstb.2008.0216)
638 78. Stuart-Fox D, Newton E, Clusella-Trullas S. 2017 Thermal consequences of
639 colour and near-infrared reflectance. Philos. Trans. R. Soc. B Biol. Sci. 372,
640 20160345. (doi:10.1098/rstb.2016.0345)
641 79. Shi NN, Tsai C-C, Camino F, Bernard GD, Yu N, Wehner R. 2015 Keeping cool:
642 Enhanced optical reflection and radiative heat dissipation in Saharan silver ants.
643 Science. 349, 298–301. (doi:10.1126/science.aab3564)
644 80. Turner JS, Lombard AT. 1990 Body color and body temperature in white and
645 black Namib desert beetles. J. Arid Environ. 19, 303–315. (doi:10.1016/S0140-
646 1963(18)30795-X)
647 81. Pearson DL, Lederhouse RC. 1987 Thermal ecology and the structure of an
648 assemblage of adult tiger beetle species (Cicindelidae). Oikos 50, 247.
649 (doi:10.2307/3566008)
24 650 82. Satoh A, Sota T, Uéda T, Enokido Y, Paik JC, Hori M. 2004 Evolutionary
651 history of coastal tiger beetles in Japan based on a comparative phylogeography
652 of four species. Mol. Ecol. 13, 3057–3069. (doi:10.1111/j.1365-
653 294X.2004.02274.x)
654 83. Tsuji K, Sota T, Hori M. 2019 The routs tiger beetles came to Japan. In Tiger
655 beetles of Japan -Phylogeny, ecology and conservation- (ed M Hori), pp. 32–60.
656 Tokyo, Japan: Hokuryukan. (In Japanese)
657 84. Nosil P, Egan SP, Funk DJ. 2008 Heterogeneous genomic differentiation
658 between walking-stick ecotypes: ‘Isolation by adaptation’ and multiple roles for
659 divergent selection. Evolution. 62, 316–336. (doi:10.1111/j.1558-
660 5646.2007.00299.x)
661 85. Satoh A, Hori M. 2005 Microhabitat segregation in larvae of six species of
662 coastal tiger beetles in Japan. Ecol. Res. 20, 143–149. (doi:10.1007/s11284-004-
663 0028-3)
664 86. Pfeifer SP, Laurent S, Sousa VC, Linnen CR, Foll M, Excoffier L, Hoekstra HE,
665 Jensen JD. 2018 The evolutionary history of Nebraska deer mice: Local
666 adaptation in the face of strong gene flow. Mol. Biol. Evol. 35, 792–806.
667 (doi:10.1093/molbev/msy004)
668 87. Mullen LM, Hoekstra HE. 2008 Natural selection along an environmental
669 gradient: A classic cline in mouse pigmentation. Evolution. 62, 1555–1570.
670 (doi:10.1111/j.1558-5646.2008.00425.x)
671 88. Wills HL. 1967 Bionomics and zoogeography of tiger beetles of saline habitats in
672 the central United States (Coleoptera: Cincindelidae). Univ. Kansas Sci. Bull. 47,
673 145–313.
674 89. ASAB/ABS. 2018 Guidelines for the treatment of animals in behavioural
675 research and teaching. Anim. Behav. 135, I–X.
676 (doi:10.1016/J.ANBEHAV.2017.10.001)
677
678
25 679 Figure legends
680
681 Figure 1. (a) Sampling sites, sand colouration and elytral colour pattern of the tiger beetle
682 Chaetodera laetescripta. (b) Resemblance of C. laetescripta body colour to sand
683 background at Site F (left) and Site K (right). Photographs by N. Yamamoto.
684
26 685
686 Figure 2. (a) Correlation of luminance between elytra of Chaetodera laetescripta and the
687 sand of its habitats, from the perspective of avian predator vision. Error bars indicate
688 standard deviation. Grey lines are contour lines for given JND values. (b) Correlation
689 between the mean percentage of white area and luminance among C. laetescripta elytra.
690
27 691
692 Figure 3. Percentages of paper tiger beetle models attacked by birds on light sand (left;
693 Site J) and dark sand (right; Site F) during the experiments (n = 2981).
694
28 695
696 Figure 4. (a) Maximum-likelihood tree of restriction site-associated DNA (RAD)
697 sequences. See electronic supplementary material, table S1 for details of samples. (b)
698 Relationships of genetic distance (FST) with geographic distance and (c) difference in
699 mean percentage of white area in the elytra of Chaetodera laetescripta. (d) Mean
700 percentage of elytral white area in the elytra of C. laetescripta from each locality (right),
701 with phylogenetic relationships among populations (left).
29 702 Table
703 Table 1. Effects of mean sand luminance (Lum) and climatic environmental factors on
704 the proportion of white elytral area of Chaetodera laetescripta. Temp, mean temperature
705 (°C) in June–September; Radi, mean total global radiation (MJ m–2) in June–September.
Candidate models ΔAIC Factor t P
Lum 0 Lum 4.951 0.0006**
Lum 4.765 0.0010* Lum + Temp 1.65 Temp -0.514 0.6197
Lum 3.918 0.0035* Lum + Radi 1.98 Radi -0.135 0.8955
Lum 3.482 0.0083*
Lum + Temp + Radi 3.55 Temp 0.262 0.7996
Radi -0.537 0.6056
Temp 11.92 Temp 1.668 0.1263
Temp 1.926 0.0862 Temp + Radi 12.62 Radi -1.015 0.3367
Null model 12.87
Radi 14.76 Radi 0.295 0.7742
**P < 0.01, *P < 0.05
706
30