When populations experience different sensory conditions, natural selection may favor whole sensory system divergence, from the peripheral structures to the brain. We characterized the outer eye morphology of sympatric Heliconius species from different forest types, and their first-generation reciprocal hybrids to test for adaptive visual system divergence and hybrid disruption. In Panama, Heliconius cydno occurs in closed forests, whereas Heliconius melpomene resides in more open areas. Previous work has shown that, among wild individuals, H. cydno has larger eyes than H. melpomene, and there are heritable, habitatassociated differences in the visual brain structures that exceed neutral divergence expectations. Notably, hybrids have intermediate neural phenotypes, suggesting disruption. 34

To test for similar effects in the visual periphery, we reared both species and their hybrids in 25 26 27 28 29 30 31 32 33 35 36 37 38 39 40 41 42 43 44 45 46 47 common garden conditions. We confirm that H. cydno has larger eyes and provide new evidence that this is driven by selection. Hybrid eye morphology is more H. melpomene-like despite body size being intermediate, contrasting with neural trait intermediacy. Thus, eye morphology differences between H. cydno and H. melpomene are consistent with adaptive divergence, and when combined with previous neuroanatomy data, suggest hybrid visual system disruption due to mismatched patterns of intermediacy and dominance in the visual pathway.

Keywords: sensory adaptation, selection, visual system, hybrid disruption

Sensory systems mediate the transmission of information between an organism and its surroundings (Stevens, 2013). Natural selection is expected to favor divergent sensory phenotypes across populations exposed to different sensory conditions and/or which exploit different resources, potentially leading to both pre- and post-mating mating reproductive 54

isolation (Dell9Aglio et al., 2023). Habitat-associated variation in sensory traits is well49 50 51 52 53 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 documented, particularly for vision (Webster, 2015). However, most studies of visual adaptation between populations have focused on color vision in aquatic organisms (Carleton & Yourick, 2020; Cummings & Endler, 2018), whereas other aspects of visual perception are understudied. For example, we know little about how visual signals are evaluated (Rosenthal, 2018) or how adaptive processes operate at different levels within the visual pathway. Moreover, we lack information on the visual capabilities of hybrids despite their potential to contribute to reproductive isolation between species. Here, we characterize the outer eye morphology of Heliconius butterflies, examine evidence for adaptive divergence, and consider how this may lead to the disruption of visual systems in hybrids.

The compound eyes of insects represent an easily quantifiable sensory structure that directly affects visual perception. The insect compound eye consists of numerous independent photosensitive units, ommatidia, each of which receives visual (light) information and transfers it to the brain. Variation in the total number of ommatidia, their size, and density directly affects visual perception and often correlates with temporal activity (Greiner, 2006; Land, 1997; Stöckl et al., 2017; Warrant, 2004). For example, nocturnal and crepuscular species often have larger eyes and larger facets to enhance photon sensitivity in light-poor environments, as observed across insect taxa (Freelance et al., 2021). Associations between the local environment and the visual systems of diurnal insects are less studied and often included only as a comparison to other nocturnal species (e.g., Frederiksen & Warrant, 74 75 76 77 78 79 80 81 82 83 85 86 87 88 89 90 91 92 93 95 96 97 4 2008). Nonetheless, the conditions experienced by diurnal insects can vary greatly (Endler, 1993), and may represent important adaptations to the local environment.

Heliconius butterflies inhabit tropical and subtropical regions of the Americas and rely heavily on vision for foraging for both flowers and hostplants (Dell9Aglio et al., 2016; Gilbert, 1982), as well as finding and choosing suitable mates (Crane, 1955; Estrada & Jiggins, 2008; Hausmann et al., 2021; Jiggins et al., 2001; Merrill et al., 2019). In Panama, the closely related species Heliconius melpomene and Heliconius cydno are broadly sympatric, but occupy different forest types (Estrada & Jiggins, 2002). H. melpomene primarily lives in forest edge habitats, whereas H. cydno occurs deeper within the forests, with less light and increased habitat complexity (Fig. 1A) (DeVries, 1987; Estrada & Jiggins, 84 2002; Seymoure, 2016). Although patterns of opsin expression suggest few differences in wavelength sensitivity (McCulloch et al., 2017), recent data on brain morphologies of H. melpomene and H. cydno reported heritable differences in the size of the visual neuropils that exceed expected rates of neutral divergence (Montgomery et al., 2021). Using wild caught individuals, Seymoure et al. (2015) similarly found that i) H. cydno has larger eyes than H. melpomene and ii) that H. cydno males have larger eyes than H. cydno females (intraspecific differences in H. melpomene were non-significant). However, these results were based on individuals sampled as adults in their respective habitats and may include effects of environment-induced plasticity. Also, total ommatidia counts were measured for only two individuals for each species and sex, so statistical power to explore different eye morphology 94 traits was limited.

Given the evidence for selection acting on the visual processing regions of the brain, a more thorough examination of the visual periphery in these species is warranted. In particular, the role of plasticity and selection, and the potential for the mismatch of components of the visual system in interspecific hybrids have not yet been assessed. To 99 100 101 102 103 105 106 107 108 109 110 111 112 113 116 117 118 119 120 address this, we characterized the outer eye morphology of H. melpomene and H. cydno to test for patterns of adaptive divergence in the visual system. First, we compared the eye morphology of butterflies (15+ for each species and sex) reared under common garden conditions. We then used a quantitative genetics approach to test if the species-specific eye differences are due to selection using PST-FST analysis. Finally, we report patterns of eye 104 morphology in first-generation (F1) hybrids of H. melpomene and H. cydno.

Butterfly specimens We established outbred stocks of Heliconius cydno chioneus (C) and Heliconius melpomene rosina (M) from butterflies caught in Gamboa and the nearby Soberanía National Park, Panama between 2007-2009 and 2015-2017. We generated reciprocal F1 hybrids between H. cydno and H. melpomene by either crossing a H. cydno female with H. melpomene a male (CxM), or a H. melpomene female with a H. cydno male (MxC). All pure individuals and hybrids were reared under common garden conditions in the Smithsonian Tropical Research 114

Institute insectaries in Gamboa, and all specimens were preserved in DMSO/EDTA/NaCl and 115 stored at -80° C as described in Merrill et al. (2019).

Heritable shifts in eye morphology between species residing in different forest types In total, we sampled 15 male and 19 female H. cydno and 18 male and 15 female H. melpomene, all reared under common garden conditions. Using tibia length as a proxy for 204 body size, we found that H. cydno was larger than H. melpomene (F1, 65 = 58.77, FDR-p < 0.001), but there was no evidence for sexual size dimorphism in either species (FDR-p = 0.7).

All three fixed effects, tibia length, species, and sex, were retained in our model examining 207 facet count. After accounting for size (larger butterflies had more facets: F1, 63 = 20.44, FDRp < 0.001), H. cydno had more facets than H. melpomene (F1, 63 = 7.32, FDR-p = 0.013; Fig. 1B), and males had more facets than females (F1, 63 = 27.35, FDR-p < 0.001; Figs. 1C, S2). The interaction between species and sex was not significant (FDR-p = 0.7), implying conserved patterns of sexual dimorphism. We found similar results for corneal area (Fig. 1E, F), where all three fixed effects were also retained (Table S1). For all butterflies, larger corneal area was due to an increase in facet number, as evidenced by facet count significantly 214

affecting corneal area in the model log10(corneal area) ~ log10(facet count) * species * sex 197 198 199 200 201 202 203 205 206 208 209 210 211 212 213 215 216 217 218 219 220 (F1,64 = 235.38, FDR-p < 0.001). Importantly, the facet count x species interaction was not significant in this model (FDR-p = 0.15), suggesting no differences in facet size between species.

Given the sex-specific differences in eye morphology reported above, we analyzed the scaling relationships between eye morphology and body size for males and females separately (Fig 1D, G). The only significant difference in slope (³) was when comparing the scaling relationship between corneal area and tibia length for H. cydno vs. H. melpomene 225 226 228

p < 0.001; Table 1). males (FDR-p = 0.016; Table 1); all other comparisons were non-significant, confirming common slopes (FDR-p > 0.23; Table 1). In isolation, body size and facet count were 224 uncorrelated for H. cydno males (r2 = 0.004, p = 0.8), but there was no statistical difference in scaling between the species suggesting this is potentially due to increased variance in H. cydno (Table 1). For all comparisons with a common slope, tests for grade-shifts (³) were 227

non-significant (FDR-p > 0.4), but there was a significant shift along the common axis (FDRoccupy different habitats: H. cydno is found in closed forest environments, whereas H. melpomene resides in open forests. (B, C, E, F) Estimated marginal means (e.m.m.) of the minimum adequate statistical models for (B, C) facet count and (E, F) corneal area, demonstrating the significant effects of (B, E) species and (C, F) sex, while accounting for other significant terms (tibia length and sex/species). The interaction between species and sex was never significant (FDR-p > 0.13). Different letters indicate significant differences (p < 0.05), and the error bars represent 95% confidence intervals. (D, G) Major axis regressions of (D) facet count or (G) corneal area and body size, measured as hind tibia length. Double-logarithmic plots are presented to explore the allometric relationships between eye morphology and body size. Males are represented by solid shapes and solid lines, and females are represented by open shapes and dashed lines. C = H. cydno; M = H. melpomene.

³ (slope) ³ (y intercept) shift along common axis FDR-p <0.001

<0.001 <0.001 male female facet count corneal area facet count corneal area

LR melpomene males and females. No values are reported for male corneal area because tests for grade shifts (³) and shifts along the common axis are only appropriate with a common slope.

Differences in eye morphology are driven by selection Our PST-FST analyses suggest that the visual systems of H. cydno and H. melpomene have likely diverged as the result of selection rather than genetic drift. PST was significantly higher than FST for both facet count and corneal area (p < 0.001) for all comparisons (Fig. 2; Table S3), where the proportion of phenotypic variance due to additive genetic effects withinpopulations far exceeds the proportion of phenotypic variance due to additive genetic effects 254

between-populations. Qualitatively similar results were obtained regardless of the phenotypic 255

measurement evaluated (i.e., raw data, log10 transformed, allometrically corrected values; 256

Table S3) and also when each sex was examined separately (Tables S4, S5). 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 values between H. cydno chioneus and H. melpomene rosina (values from Martin et al., 2013). Here, both morphological measurements are allometrically corrected using tibia length as a body-size proxy and presented using varying c/h2 ratios (see Table S3 for PST estimates using raw and log10 transformed values). The dotted line represents the 95th percentile of the FST distribution.

Hybrid eye morphology is H. melpomene-like We examined the eye morphology of 60 F1 hybrids, including 15 male and 15 female CxM individuals, and 15 female and 15 male MxC individuals (Table S2). Tukey post hoc (with Bonferroni correction) revealed patterns of intermediacy in body size (using tibia length as a body size proxy) for both hybrid types compared to the parental species: both CxM and MxC hybrids were larger than H. melpomene (p < 0.001), MxC was smaller than H. cydno (t = 3.148, p = 0.0124), and CxM did not differ fromH. cydno (p = 1.0; Fig. 3A). However, F1 hybrid eye morphology was not intermediate to the parental species: both hybrid types had significantly fewer facets (p < 0.0035) than H. cydno but did not differ from H. melpomene (p 274

= 1.0; Fig. 3B). Results for corneal area were similar (Table S2; Fig. S3, S4). These patterns 275

were also evident when exploring the scaling relationships between eye morphology and 276

body size of the F1 hybrids (Fig. 3C, D; Fig. S4). (using tibia length as a proxy) including F1 hybrids, demonstrating the significant effect of species (sex was non-significant, FDR-p = 0.48). E.m.m for (B) facet count, showing the significant effect of species, while accounting for significant tibia length and sex effects. In both plots, different letters indicate significant differences (Bonferroni adjusted p < 0.05), and the error bars represent 95% confidence intervals. (C-D) Major axis regressions of facet count and body size (tibia length) for males and females separately. Double-logarithmic plots are presented to explore the allometric relationships between eye morphology and body size. C = H. cydno; M = H. melpomene; CxM = F1 hybrid of H. cydno mother crossed with H. melpomene father; MxC = F1 hybrid of H. melpomene mother crossed with H. cydno father.

When populations are exposed to different sensory conditions, natural selection may favor divergence in sensory traits, which can contribute to speciation. We compared the outer eye morphology of Heliconius butterflies that occupy different environments to test for patterns 294

of adaptive visual system divergence. Our results show that H. cydno, which occupies more 290 291 292 293 295 296 297 298 299 300 301 302 303 305 306 307 308 309 310 311 312 313 304

However, the individuals sampled for our analyses were raised under common garden visually complex closed-canopy forests, has larger eyes, and that this is a result of heritable differences in facet number. By combining our phenotypic data with genome wide estimates of FST, we additionally provide strong evidence that selection has driven the divergence of eye morphology in these butterflies. Finally, we show that F1 hybrid eye morphology is not intermediate to the parental species, contrasting with patterns for body size and neural anatomy. This suggests that visual processing in hybrids may be disrupted by mismatches in different parts of the visual pathway.

Our results are consistent with previous work by Seymoure et al. (2015), which reported larger eyes for H. cydno, and bigger eyes in H. cydno (but not H. melpomene) males. conditions, reducing the potential for environmental effects and genotype3environment interactions, which may give a distorted picture of the contribution of genetic variation on which selection can act (Brommer, 2011; Pujol et al., 2008). A potential caveat of our results is that we cut each cuticle four times to mount it on the microscope slides, which may have disrupted the semi-automated counts of individual facets. It is possible that more advanced 3D imaging techniques (e.g., Buffry et al., 2023), where cutting is not required, may give slightly higher total counts, though likely at the expense of overall sample size and associated statistical power. Regardless, our facet counts are consistent with prior work (Seymoure et al., 2015) (Fig. S1), and the close correspondence between wild and insectary-reared

butterflies further suggests that the differences in eye morphology are largely heritable.

Habitat-associated variation in eye morphology has been reported across taxa (e.g., insects: (Greiner, 2006); mammals: (Veilleux & Lewis, 2011); fish: (Lisney et al., 2020); 315 316 318 319 320 321 322 325 326 327 328 329 330 331 332 335 336 337 338 snakes: (Liu et al., 2012); primates: (Kirk, 2004)), and this variation is generally interpreted as an adaptive response to the local sensory conditions. For example, visual perception in insects is affected by the total number of ommatidia and their size/density (Greiner, 2006; Land, 1997; Warrant, 2004), and nocturnal and crepuscular species often possess larger eyes and larger facets to enhance photon sensitivity in low-light environments (Freelance et al., 2021). However, most studies do not formally evaluate the role of selection, which is a key 323

element to define adaptation (Gould & Vrba, 1982). We are aware of only one study that has 324 attempted to addresses this topic: Brandon et al. (2015) reported that eye size variation in a wild Daphnia population is associated with variation in fitness (reproductive output), suggesting that selection is operating, either directly or indirectly, on eye size variation (though the underlying mechanisms remain unknown).

In addition to revealing heritable differences in eye morphology, our results suggest that eye morphology in H. cydno and H. melpomene have evolved as the result of divergent selection, as opposed to genetic drift. Although our PST-FST approach to test for evidence of selection acting on eye morphology is limited by the difficulty in approximating PST to QST (Brommer, 2011), these limitations are largely overcome by rearing our butterflies under 333

common garden conditions (Leinonen et al., 2008, 2013). Moreover, as with most insects, the 334 heritability of facet count and corneal area are unknown for the species used in this study. To 2 account for this, we used a wide range of c/h ratios, including the default assumption of c = 2 2 h (i.e., c/h = 1), where the proportion of phenotypic variance due to additive genetic effects is the same for between-population variance and within-population variance (Brommer, 2011). In all cases, PST values were higher than the 95 th percentile of the genome wide FST distribution (Tables S3-S5).

In insects, eye size may increase due to an increase in the number of ommatidia, an increase in individual ommatidia size, or both. Our results suggest that larger eye size in H. cydno, and in males of both species, is predominantly due to an increase in ommatidia number. We did not measure facet diameter directly, but we found no effect of the facet 344 count x species interaction on corneal area, suggesting no interspecific differences in facet 340 341 342 343 345 346 347 348 349 350 351 352 353 355 356 357 358 359 360 361 362 363 354 erato support this prediction (Wright et al., 2023). We note that our allometric analyses size. Seymoure et al. (2015) measured facet diameter from a subset of ommatidia in six anatomical eye regions and reported larger facet diameter in wild H. cydno compared to H. melpomene, with no differences between sexes. However, these results stemmed from an analysis of covariance including other species (H. sapho and H. erato); there were no direct comparisons between H. cydno and H. melpomene. Future studies may benefit from exploring facet diameter more directly. Regardless, the larger eyes of H. cydno, and male Heliconius, appear to be largely due to an increase in ommatidia number. In insects, increased ommatidia number is thought to contribute to higher visual acuity (Land, 1997) behavioral and morphological data revealing sexual dimorphism in the visual acuity of H. suggest part of the variation we observe is associated with body size, within and between species. However, our PST-FST analyses account for body size and still suggest a signal of selection. The pattern we observe in hybrids, where grade-shifts are clearly observed between cydno and cydno x melpomene hybrids (Figure 3C), also suggest a strong genetic component independent of body size. As such, while the behavioural impact of increased eye size likely depends on the raw numbers of facets, we conclude that selection on body size alone does not explain the increase in cydno eye size.

Multiple non-exclusive selective pressures could be driving the differences we report here. First, species differences may be explained by H. cydno occupying more complex advantageous due to e.g., increased visual acuity (Land, 1997; Wright et al., 2023). More ommatidia in males may also be due to general ecological differences between the sexes, as males actively search for and identify mates (Rutowski, 2000; Yagi & Koyama, 1963).

Interestingly, species differences persist for both sexes (Fig. S2) and when exploring the PST369

FST results for each sex separately, we still observed evidence of selection (Tables S4, S5). 365 366 367 368 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388

Taken together, our results suggest that the selective pressure on males to have more ommatidia acts in both species in concert with selection for more ommatidia in closed-forest environments. Similar patterns may exist across Heliconius, but to date, few species have been surveyed for eye morphology.

Our results also mirror neuroanatomical comparisons between species across the cydno-melpomene clade, where larger visual neuropils are reported for cydno-clade species occupying closed-forest environments, as opposed to H. melpomene (Montgomery et al., 2021). These neural differences appear to be heritable adaptations, based on similar tests of selection to those reported here. Thus, the combined results on brain morphology (Montgomery et al., 2021) and those presented here suggest whole visual system adaptation, from the sensory periphery to the brain. Similar habitat-associated differences in neuroanatomy have been reported in other Neotropical butterflies (Montgomery & Merrill, 2017; Wainwright & Montgomery, 2022), indicating a broader pattern of sensory adaptation in ecologically divergent, but closely related, butterflies.

One notable difference in the results obtained for eye and neural traits, however, is in the pattern of variation among F1 interspecific hybrids. The eye morphology of H. cydno and H. melpomene F1 hybrids tended to be more H. melpomene-like. This contrasts with patterns for body size (Fig. 3A) and neuroanatomy, where hybrids are intermediate for at least some traits (Montgomery et al., 2021). The observation that hybrid eye morphology is melpomenelike, but hybrid body size tends to be intermediate indicates that these two traits have 390 391 392 393 395 396 397 398 399 400 401 402 409 410 411 412 different underlying genetic architectures (i.e., eye size is not simply genetically correlated with increasing body size). Furthermore, evidence suggests that both eye morphology (this study) and neural anatomy (Montgomery et al., 2021) are under divergent selection and in the predicted direction (i.e., bigger eyes and larger visual neuropils in H. cydno). If these 394 observations truly represent adaptations (as our results suggest), then hybrids may have suboptimal visual system functioning, whereby peripheral sensory structures (number of facets) are mismatched to subsequent processing regions (optic lobe neuropils). This would predict that hybrids suffer a fitness deficit due to lower performance in visually oriented tasks, such as foraging and mate detection, but behavioral experiments are required to test these predictions. In conclusion, the adaptive differences in eye structure we observe may contribute to ecologically based pre- and post-mating reproductive barriers.

403

We thank Josephone Dessmann, Maria-Clara Melo, Moises Abanto, Adraia Tapia, Liz Evans 404 for helping raise butterflies. We thank the Smithsonian Tropical Research Institute for 405

providing research infrastructure, and the Autoridad Nacional del Ambiente for permission to 406

collect butterflies in Panama (permit no. SE/A-22-09). This work was supported by a NERC 407 Independent Research Fellowship (NE/N014936/1) to SHM, and a DFG Emmy Noether 408fellowship (Number: GZ: ME 4845/1-1) and an ERC starting grant (Grant number: 851040) to RMM. The underlying data and R-scripts supporting the findings of this study are available at 415 powerful approach to multiple testing. Journal of the Royal Statistical Society: Series B (Methodological), 57(1), 2893300.

Brandon, C. S., James, T., & Dudycha, J. L. (2015). Selection on incremental variation of eye size in a wild population of Daphnia. Journal of Evolutionary Biology, 28(11), 21123 2118. https://doi.org/10.1111/jeb.12711 421

Brommer, J. E. (2011). Whither Pst? The approximation of Qst by Pst in evolutionary and 424 Buffry, A. D., Currea, J. P., Franke, F. A., Palavalli-Nettimi, R., Bodey, A. J., Rau, C., conservation biology. Journal of Evolutionary Biology, 24(6), 116031168. Samadi, N., Gstöhl, S. J., Schlepütz, C. M., McGregor, A. P., Sumner-Rooney, L., Theobald, J., & Kittelmann, M. (2023). Evolution of compound eye morphology underlies differences in vision between closely related Drosophila species [Preprint]. 416 417 418 419 420 422 423 425 426 427 430 431 432 433 437 Crane. (1955). Imaginal behaviour of a Trinidad butterfly, Heliconius erato hydara Hewitson, with special reference to the social use of color. Zoologica, 40, 1673196. 434 Crawley, M. J. (2013). The R book (Second edition). Wiley. 435

Cummings, M. E., & Endler, J. A. (2018). 25 Years of sensory drive: The evidence and its 436

watery bias. Current Zoology, 64(4), 4713484. https://doi.org/10.1093/cz/zoy043 Dell9Aglio, D. D., Losada, M. E., & Jiggins, C. D. (2016). Butterfly Learning and the

Diversification of Plant Leaf Shape. Frontiers in Ecology and Evolution, 4. 442 DeVries, P. J. (1987). The Butterflies of Costa Rica and their natural history. New Jersey, (2023). The sensory ecology of speciation. Cold Spring Harbor Perspectives in