Lens Transmittance Shapes UV Sensitivity in the Eyes of Frogs from Diverse Ecological and Phylogenetic Backgrounds

The amount of short wavelength (UV, violet and blue) light that reaches the retina depends on the transmittance properties of the ocular media, especially the lens, and varies greatly across species in all vertebrate groups studied previously. We measured the lens transmittance in 32 anuran amphibians with different habits, geographic distributions, and phylogenetic positions and used them together with eye size and pupil shape to evaluate the relationship with diel activity pattern, elevation and latitude. We found an unusually high lens UV transmittance in the most basal species, and a range that extends into the visible spectrum for the rest of the sample, with lenses even absorbing violet light in some diurnal species. However, other diurnal frogs had lenses that transmit UV light like the nocturnal species. This unclear pattern in the segregation of ocular media transmittance and diel activity is shared with other vertebrates and is consistent with the absence of significant correlations in our statistical analyses. Although we did not detect a significant phylogenetic effect, closely related species tend to have similar transmittances, irrespective of whether they share the same diel pattern or not, suggesting that ocular media transmittance properties might be related to phylogeny.

6 light using a 1000 μm guide connected to a Maya2000 spectroradiometer controlled by 121 SpectraSuite v4.1 software (Ocean Optics). The guides were aligned with the container in a 122 microbench system (LINOS, Munich, DE). The reference measurement was taken from the 123 container filled with PBS. We smoothed the curves using an 11-point running average, and 124 normalized to the highest value within the range 300-700 nm. From these data, we 125 determined λ T50 as the wavelength at which the light transmittance was 50% of the 126 maximum. The curves were cut for clarity in those cases in which the measurements at very 127 low wavelengths were too noisy due to the low sensitivity of the spectrometer in that 128 region of the spectrum. 129 We combined the lens transmittance data collected from the 32 species measured in this 130 study with those from Bufo bufo, Rhinella ornata, Lithobates catesbeianus, L. pipiens and 131 Rana temporaria that were available from a previous study [21], making a total of 37 132 species of 14 families. Given that corneal transmittance data were collected from just a 133 handful of species, they were not included in the phylogenetic comparative analyses. 134 We used eye size compiled from descriptions of the species in the scientific literature as a 135 proxy for lens optical path length. When these data were not available for a given species, 136 we obtained them from colleagues or measured it from museum specimens (see electronic 137 supplementary material S1B for the whole dataset of eye size values and sources, and S1C 138 for validation of the method). 139 For pupil shape, we visually inspected photographs available online for each of the species 140 and scored them as round or elongate (see electronic supplementary material S1D for 141 details and thresholds on scoring criteria). Even though orientation (i.e. horizontally or 142 vertically elongate) can have a differential effect on the sharpness of horizontal and vertical 7 images [22] we did not distinguish between them because vertical slit pupils are extremely 144 uncommon among anurans and would compromise the statistical analyses. 145 Diel activity pattern is somewhat labile in anurans and can vary for specific behaviours; 146 however, most species are predominantly nocturnal (solid lines in Fig. 1) [23], with only a 147 few lineages being predominantly diurnal, including the dendrobatoids (Aromobatidae + 148 Dendrobatidae), hylodids, as well as Atelopus and Brachycephalus in our study (dashed lines 149 in Figure 1) [23]. We opted to handle diel pattern as a binary variable, in line with previous 150 work that uses this approach for different types of phylogenetic analyses [23]. Following this 151 criterion, we scored Scinax ruber and Lithobates pipiens, which have been reported to be 152 arrhythmic [23], as nocturnal based on our own fieldwork experience. 153 Given that elevation and latitude contribute to shaping light habitats, we also took them 154 into account. We scored the elevation and latitude of the same specimens used to obtain 155 the lens transmittance measurements, with two caveats. First, Bombina orientalis was 156 captive bred in the pet trade in Lund, Sweden (elevation: 51 metres above sea level 157 (m.a.s.l.), latitude: 55.7°), which is within the natural elevation of the species but is 158 approximately 7° north of its northernmost distribution [24]. As such, we performed 159 analyses both including and excluding this species. Second, Kennedy and Milkman [7] did 160 not provide collection data for the Lithobates pipiens specimens they used to measure lens 161 transmittance; given that the research was conducted at Harvard University, which is within 162 the natural distribution of the species [25], we assumed they were collected nearby. 163 164

Statistical analysis 165
We performed a phylogenetic comparative analysis to evaluate the linear relationships 166 between lens λ T50 , eye size, and pupil shape as predictor variables, and diel activity pattern, 167 8 elevation, and latitude as response variables. Given the reports of a linear relationship 168 between lens λ T50 and eye size in birds and mammals [5,8], we also tested this relationship 169 explicitly. We used the phylogenetic hypothesis of Pyron [26] to control for the phylogenetic 170 non-independence of the species in our sample (see electronic supplementary material S1E 171 for details on how species missing in the tree were accommodated and 1F for the resulting 172 tree). 173 We performed all analyses in R v3.6.0 using the packages ape v5.3 [27], car v3.0-3 [28], 174 GEIGER v2.0.6.2 [29], and nlme v3.  The lens λ T50 of the 32 species measured in this study are spread throughout the UV-violet 185 part of the spectrum, covering the range 280-425 nm (Figure 1), which also contains the 186 values of the five species in our previous study [21]. The breadth of the range is similar in 187 Hyloides and Ranoides, the two major lineages of neobatrachians that contain more than 188 90% of the anuran species diversity [33], although the upper boundary for the latter seems 196 for the tree with branches lengths scaled to phylogenetic distances). All the lens λ T50 were calculated 197 in this study except those marked with an asterisk (*), which were obtained from [21].

198
We also measured the transmittance of the corneas in eight species from our sample. For 199 most of them the λ T50 was within the range ≈320-345 nm, irrespective of the lens 200 transmittance properties ( Figure 2, electronic supplementary material S1G). However, 201 Physalaemus cuvieri has a cornea λ T50 =293 nm (Figure 2), and that is probably also the case 202 for Xenohyla truncata, although the precise λ T50 value could not be calculated for the latter 203 (electronic supplementary material S1G). were obtained from a previous study [21]. The diagonal defines the regions in which light

231
We also found no significant relationship between lens λ T50 , eye size, and pupil shape 232 (predictor variables) and diel pattern, elevation, and latitude (response variables) for any of 233 the models (p > 0.09; see electronic supplementary material S1I). to which most adult terrestrial anurans are exposed. It would be interesting to investigate 245 whether the lenses of aquatic species and/or those that live in very high elevations show 246 specific patterns within this range, or significant departures from it. 247 The lower limit in the range of lens λ T50 among the anurans in our sample is intriguing, and 248 more than 20 nm shorter than the most extreme cases reported so far in vertebrates: the 249 porcupine fish Diodon hystrix (301 nm, [6]), the sand-dwelling lizard Calyptommatus nicterus 250 (303 nm, [34]), the African house snake Boaedon (Lamprophis) olivaceus (306 nm, [17]), and 251 the Japanese quail Coturnix japonica (≈310 nm, [35]). However, the unusually high lens UV 252 13 transmission in B. orientalis has no functional relevance in terms of light availability for the 253 retina, since the cornea of this frog has a λ T50 =338 nm (similar to other species with higher 254 lens λ T50 values; Figure 2 and electronic supplementary material S1G). This means that the 255 amount of ultraviolet light that can effectively reach the photoreceptors is comparable to 256 that in frogs with lens λ T50 ≈335-340 nm. Thus, in this particular species the light 257 transmittance of the eye as a whole is limited by the cornea rather than the lens (Figure 2), 258 as is the case in quails [35]. However, this is likely the exception rather than the rule and not Variation in the shape of the transmittance curves among lenses that absorb part of the UV 289 radiation is quite a common theme in vertebrates, and in particular local increases at short 290 wavelengths can be seen in lens transmittance curves from mammals [5] and snakes [17]. 291 However, the only group in which this phenomenon has been thoroughly studied are fishes, 292 for which several pigments drive these patterns [6]: fishes with curves like the ones for 293 Hylodes phyllodes and Oophaga pumilio have a pigment with peak absorption at ≈370 nm, 294 whereas other species of fishes with shoulders in their transmittance curves similar to those 295 of Brachycephalus ephippium and Craugastor fitzingeri have two different pigments with 296 peak absorptions at ≈320-330 and 360 nm. Finally, fish lenses with smooth curves and high 297 λ T50 values like the one from Leptodactylus insularum in our sample have high 298 concentrations of either the 360 nm pigment or both the 320-300 nm and 360 nm pigment 299 [6]. Curves with very subtle local increases at short wavelengths similar to those of 300 Dendropsophus microcephalus and Cochranella granulosa in our sample have not been 301 reported in fishes, but are present in some mammals such as the meerkat, in whom lens 302 pigments with absorption maxima at 360-370 nm have been extracted [5]. 303 The similarity between fishes and anurans in the overall shape of transmittance curves for 304 lenses of different species suggests that a number of pigments are involved in generating 305 those patterns in the latter, as they are in the former. However, no comparative studies of 306 lens pigmentation have been conducted in anurans. The only species for which a lens 307 pigment has been extracted is the leopard frog Lithobates pipiens; its absorbance peaks at 308 345 nm and it was not identified [7]. However, this absorbance profile does not match any 309 other pigment identified in the lenses of fishes or mammals [1], so it is very likely that its 310 chemical identity is different. 311 The presumptive presence of pigments in some anuran lenses can explain the lack of 312 correlation between lens transmittance and eye size in our analyses, as that relationship 313 holds only for unpigmented lenses [1]. It is thus possible that a relationship between the 314 two variables exists in amphibians, as it does in birds [8,35], mammals [5] and some fishes 315 [36], but is masked by the pigmented lenses in our sample. Absence of lens pigments has 316 been demonstrated for 33 species of fishes with smooth transmittance curves and lens 317 λ T50 ≈310-340 nm [6]. Interestingly, if the linear regression for our sample is performed only 318 with the six species that also have smooth transmittance curves and lens λ T50 ≈310-340 nm, 319 the relationship between lens transmittance and eye size has an excellent fit (R 2 =0.96; 320 electronic supplementary material S1J). If variation in the occurrence of pigment is 321 confirmed for frog lenses, the relationship between lens transmittance and eye size should 322 be re-tested among the species that fulfil the requirement of absence of pigment.

342
In this scenario, it comes as no surprise that there are diurnal anuran species in our sample 343 on both sides of the UV transmission axis and no significant correlation of lens 344 transmittance with diel pattern (and by extension, with the remaining variables that 345 influence intensity and spectral composition of the light environment). 346 Despite their shared propensity to transmit at least part of the incoming UV radiation 347 through their ocular media, the benefits might differ among nocturnal species from 348 different vertebrate groups. Nocturnal vision in vertebrates is driven by rod photoreceptors, 349 which typically have a peak spectral sensitivity outside the UV range at approximately 500 350 nm [37]. In addition, rods, as well as all other vertebrate photoreceptors, have a secondary 351 lower, broader peak in the UV range (the β-band) [37] whose contribution to overall photon 352 catch can become relevant and improve visual sensitivity in dim light when the total number 353 of photons is extremely limited. In the case of amphibians, their rods are generally bigger-354 and thus more sensitive-than those of other vertebrates [38], so the contribution of UV 355 light to overall visual sensitivity might not be as crucial as in other groups; indeed, the lenses 356 of many nocturnal frogs are close to the boundary between UV-transmissive and UV-357 absorbing (e.g. some hylids and ranids, Figure 5) and absorb almost all light in the region of 358 the β-band [21]. However, anurans and some caudates are unique among vertebrates in 359 having a second rod type with peak sensitivity at ≈435 nm, in addition to the typical one at 360 ≈500 nm [37,39]. This dual rod system allows frogs to retain the ability to discriminate 361 colours down to light intensities in which other vertebrates become colour-blind [40,41], 362 and its proper functioning might be relevant for many of the ≈80% of anuran species that 363 are nocturnal [23]. In this context, it becomes crucial that the lens does not absorb too 364 much short wavelength light; λ T50 =403 nm already reduces a significant amount of the light 365 that can reach the retina in Rana temporaria and removes almost completely the β-bands of 366 both rods' spectral sensitivity curves [21], and higher values could affect the sensitivity peak 367 of the blue-sensitive rods, becoming detrimental to the performance of the visual system of 368 nocturnal frogs in the dimly lit environments they inhabit. 369 As is the case with other vertebrates, there is no clear reason why some of the diurnal frogs 370 in our sample depart from the expected UV-absorbing lenses. Filtering short-wavelength 371 radiation can help reduce scattering and chromatic aberrations, thus improving spatial 372 resolution, as has been suggested for animals that depend on sharp vision, such as raptors 373 [8,20] and gliding snakes [17], and share UV-absorbing lenses. In the case of frogs, enhanced 374 spatial resolution could be advantageous to species that use visual displays. All the diurnal 375 representatives in our samples use them [42][43][44][45], suggesting that the optical problems 376 caused by UV light are not serious enough-probably due to the small size of their eyes-to 377 drive the selective pressure towards longer-wavelength shifted lens λ T50 values in all cases. 378 An alternative explanation could be that, in some species, chromatic aberrations, if relevant, 379 are dealt with by multifocal lenses rather by than UV-absorbing ones. Multifocality has only 380 been tested in two anurans: the bufonid Rhinella marina (formerly Bufo marinus) has a 381 multifocal lens, while the dendrobatid Phyllobates bicolor does not [10]. These data 382 complement the differences in UV transmittance between the diurnal bufonid (Atelopus) 383 and dendrobatids in our sample but are too limited to speculate about potential 384 generalities. It would be interesting to obtain information about both lens transmittance 385 and focal optics in the same species for a variety of anuran lineages, which would enable 386 well-grounded hypotheses to be formulated about potential relationships between the two 387

variables. 388
There is the possibility that UV light carries information valuable to some species in ways 389 that are beyond both our knowledge of their visual ecology and our ability to imagine, given 390