Widespread bone-based fluorescence in chameleons

Fluorescence is widespread in marine organisms but uncommon in terrestrial tetrapods. We here show that many chameleon species have bony tubercles protruding from the skull that are visible through their scales, and fluoresce under UV light. Tubercles arising from bones of the skull displace all dermal layers other than a thin, transparent layer of epidermis, creating a ‘window’ onto the bone. In the genus Calumma, the number of these tubercles is sexually dimorphic in most species, suggesting a signalling role, and also strongly reflects species groups, indicating systematic value of these features. Co-option of the known fluorescent properties of bone has never before been shown, yet it is widespread in the chameleons of Madagascar and some African chameleon genera, particularly in those genera living in forested, humid habitats known to have a higher relative component of ambient UV light. The fluorescence emits with a maximum at around 430 nm in blue colour which contrasts well to the green and brown background reflectance of forest habitats. This discovery opens new avenues in the study of signalling among chameleons and sexual selection factors driving ornamentation.

scales (Fig. 1C) in most cases. Micro-CT scans reveal these raised scales to be caused by tubercular outgrowths of the underlying bones (Fig. 1A,D, Supplementary Figs S1-4). Histological sections of a fluorescent tubercle (FT, Fig. 2) demonstrate that the top of the bony tubercle is only covered by a thin layer (20-25 µm) of epidermis that functions as a 'window' through which the bone is directly visible (Fig. 2B,D,F, Supplementary Fig. S5A). The bony tubercle displaces the dermis containing melanophores and chromatophores 24 , which are present around the protuberance and identical to that of a non-tubercle scale ( Fig. 2A,C,E), rendering the layer covering the tubercle thin and transparent. A comparison of fluorescence spectra of a FT with the underlying bone showed that the emission peak of the bone covers the same range and is broadened towards longer wavelengths with a maximum at about 445 nm (Supplementary Fig. S5F,G). This indicates that the FT spectrum is included in that of the bone. Obviously the thin layer of epidermis acts as an optical filter and shifts the fluorescence towards a 'bluer' emission spectrum.
Distribution patterns of FTs. This externally visible bone-based fluorescence is not restricted to Calumma but also occurs in at least 8 of the 12 chameleon genera currently recognized (Fig. 3A). We focused on the species specificity and sexual dimorphism of Calumma. A quantitative analysis of FTs in 24 of the 34 species 25,26 (126 adult specimens in total) shows that male individuals have on average more tubercles than females in almost all species (Fig. 3B). As we were restricted to using only relatively fresh material, despite having one of the largest collections of Calumma specimens outside Madagascar at our disposal, sample sizes per species were low, and the statistical significance of sexual dimorphism is therefore limited; although ANOVA results provide significantly more tubercles in the C. brevicorne group, with p = 0.016, and C. nasutum group, with p = 0.003 (Supplementary Table S1). The trend appears to be strong across all species (Figs 3B and 4A-D), and we expect that greater sampling will only strengthen it further. FTs are concentrated around the eye and the temporal region (Supplementary Figs S1-4), which are important areas for colour signalling among chameleons. Their distribution differs however among species, and even more so among species groups (Fig. 3C). A quantitative comparison of the FTs per cranial bone (Fig. 1D) in males of different Calumma species is systematically and phylogenetically informative (Fig. 3D,E). Members of the C. parsonii group for instance differ clearly from the other species by the presence of a high number of FTs in the temporal region, caused by the broadened postorbitofrontal and squamosal bone which are fused and are densely covered with tubercles ( Supplementary Fig. S4). The differences in FT patterns among closely related species are inconsistent, with some sister-species pairs (e.g. C. brevicorne and C. crypticum 27 ) having strongly overlapping values, while others (e.g. C. tsaratananaense and C. hafahafa) are distinctly different (Supplementary Table S4). FT patterns strongly reflect species groups, and are therefore of significant systematic value at the supraspecific level within Calumma (Fig. 3E), but their role within species groups requires further study.

Hypothesis of ecological relevance of fluorescence in chameleons.
Phylogenetically, it seems FTs are plesiomorphic to Chamaeleonidae, but are most widespread in Calumma being found in almost all known species (Fig. 3A, Supplementary Table S4). In their sister genus Furcifer, by contrast, we found only six of 20 studied species to have FTs (Fig. 4G, Supplementary Table S6). Furcifer is typically found in more open and dryer habitats in Western Madagascar whereas Calumma are usually found in shady, humid forest habitats 28 . Also in Brookesia, which are terrestrial and forest dwelling chameleons from Madagascar, bone-based fluorescence even across the whole body is common (Fig. 4E). As shorter (UV, blue) wavelengths are scattered more strongly than longer wavelengths 29 the UV component under the diffuse irradiation in the forest shade is relatively higher compared to the direct irradiation by the sunlight. Consequently, using UV reflections for communication is apparently more common in closed habitats than in open habitats, as has been shown in chameleons of the genus Bradypodion 30 . Members of the latter genus also show fluorescence in the head region (Fig. 4F). A function of FTs as UV reflectors could therefore be hypothesised. This can be ruled out however, as these structures do not reflect UV light at 365 or 390 nm (Fig. 1B, Supplementary Fig. S6A) and at least some chameleon lenses are only transmissive above 350 nm 31 . On the other hand, the emission spectrum of FTs with a maximum at 433 nm is deep blue, a colour that is reasonably rare in a tropical forest and appears to be a conspicuous signal against the background reflectance of grey brown leaf litter or green vegetation as shown in Fig. 1 of Andersson et al. 32 . The glowing blue of the FTs is near to the maximum absorption of the pigments of the SWS (short-wave-sensitive) cones with 440-450 nm of the examined chameleons 31 . Additionally, wavelengths of around 433 nm might appear brighter to chameleons as their visual spectrum is shifted towards shorter wavelengths (from about 350 nm to 650 nm) compared to the human visual perception. Assuming that in the shade of a closed forest canopy the relative intensity of the diffuse UV-light is even higher the fluorescent part of the total reflectance might increase considerably. A quantum yield calculation revealed 0.29% of absorbed photons being emitted as fluorescence. This is a typical fluorescence quantum yield for a dye that is immobilised in a matrix (here the bone) and not in a solution (Schwager unpubl.).
Constant fluorescent patterns potentially give chameleons a secondary, stable signalling system that is not influenced by their well-known communication by colour change, and does not compromise their camouflage. Moreover, the systematic relevance of the FTs suggests that they at least correlate with trends in skull evolution, and they may have provided substrate for sexual selection. On this basis, the distribution of FTs could also be used by taxonomists as an additional character to delimitate between taxa. Sexual dichromatism based on fluorescence has been shown in birds 33 which are also highly visual animals. The use of chameleons as model organisms in the understanding of the role of visual ecology and behaviour in sexual selection 34,35 will require some adjustment to account for this newly discovered phenomenon. Important future steps will be behavioural trials as well as the creation of colour visual models (accounting for the spatial acuity of chameleon vision) in order to explore the biological relevance of this phenomenon.
Given the recent discoveries of fluorescence, especially among terrestrial vertebrates 11,36 , it is clear that fluorescence evolved by several independent and unique mechanisms. Apparently this tends to occur by the evolution of new mechanisms, rather than implementing features that already have fluorescent properties. Use of such features would be expected to be the most obvious route to fluorescence. We suspect, however, that chameleons are not unique in this regard, and that bone-based fluorescence is in fact widespread among taxa that use bones in their ornamentation, especially other squamates. Fluorescence in terrestrial vertebrates has been underestimated until now, and its role in the evolution of ornamentation remains largely unexplored, but this is a promising avenue for future research.

Materials and Methods
Investigated chameleon specimens. 160 specimens covering 31 species (Supplementary Table S4) Table S6). All other chameleon species of the ZSM were checked and exemplarily one fluorescent specimen per genus (if applicable; no specimens available for Nadzikambia) is listed in Supplementary Table S6. Detailed photographs of fluorescent tubercles (FTs) were taken using the binocular Olympus SZX12 (Olympus DF Plapo 1XPF objective) and a Sony Nex-5N digital camera (lens replaced by a camera tubus). The investigated specimen was a male C. crypticum (ZSM 503/2014, Fig. 1C). Fluorescence spectroscopy. Fluorescence excitation-emission spectra were recorded with a Fluorolog (Horiba) fluorimeter (slit width 3 nm). All fluorescence spectra were corrected for the wavelength-dependent output of the Xenon lamp. The intensity of the fluorescence signal shown here represents the detected emission signal divided by the simultaneously measured intensity of the lamp at the same wavelength.

Ultraviolet photography. UV photographs of an ethanol-preserved
For the comparison of fluorescence spectra of a FT and the underlying bone parts of the skin were removed on the temporal region of a Calumma crypticum (ZSM 503/2014) and emission of the externally visible bone was measured ( Supplementary Fig. S5F,H). Quantum yield calculation. Using the Fluorolog fluorimeter (see above) the following four spectra were recorded with the same spectrometer settings: for the determination of the absorption, two emission spectra of the excitation signal in the range 340-365 nm (max. fluorescence excitation at 353 nm) of an FT region and non-FT region (as background), considering the same irradiation area; and for determination of the fluorescent signal, two emission spectra from 365-680 nm (max. fluorescence emission at 433 nm) of an FT region and non-FT (as background), considering the same irradiation area. The quantum yield is calculated according to  Table S1). The residuals of all species groups were checked for normality, and no significant deviations were observed. Due to our unbalanced sampling (variable number of specimens available per species), ANOVAs returned marginally different results depending on whether species or sex was given first in the formula, but this did not have any impact on significance results. Only one species, C. capuroni, has considerably more tubercles in females than in males, but the sample size of this species is low (2 specimens).
X-ray micro-CT. For skeletal morphology, X-ray micro-computed tomography scans (micro-CT scans) of the head (or full body) of 70 specimens were prepared. Methods followed Prötzel et al. 43 , using diamond or standard target, with the scanner set to 130 kV and 80 µA with a timing of 500 ms, and scans consisting of 1800 projections (i.e. 5 images per degree of rotation) or similar settings. For detailed information, see Supplementary Table S7. Histology. A FT from the temporal region (Fig. 1C) of a male Calumma crypticum (ZSM 503/2014, stored in 70% ethanol) was dissected with a razorblade. The excised FT was washed in 0.1 M phosphate buffer, stained in buffered 1% Osmium tetroxide (OsO 4 ) for one hour on ice, dehydrated in a graded acetone series, embedded in epoxy resin 44 , serial sectioned in 279 planes à 1.54 µm using a RMC MT-7000 ultramicrotome with a Diatome Histo Jumbo diamond knife, mounted on glass slides, and partly stained with a 1:1 mixture of methylene blue and Azure II for approx. 5 s at 80 °C 45 . Unstained sections ( Supplementary Fig. S5A,B) were sealed under coverslips with DPX mounting medium (Agar Scientifics). The glass slides with slice-ribbons were imaged in bright field illumination using an Olympus BX61VS light microscope with UPlanSApo 10 × NA 0.4 objective and CX10 digital camera, and the program VS-ASW FL (Olympus v 2.7) for virtual slide acquisition. Images of single slices were then extracted with OlyVIA software (Olympus v 2.9; 1267 × 2119 px, 24 bit RGB, 0.98 µm/px) for subsequent volume rendering. In addition selected slices were photographed using an Olympus CX 41 light microscope with a DP25 digital camera (objectives: Olympus Plan C 10 × NA 0.25, Olympus UPlanSApo 40 × NA 0.95, and Olympus UPlanSApo 60 × NA 1.2 W); for details see Supplementary Table S8.
Volume rendering. The colour image stack from OlyVIA was pre-processed in Adobe ® Photoshop ® CS6 as follows: converted to 8 bit greyscale, brightness and contrast autoscaled, folds in epoxy resin outside the tissue were cropped and replaced by a uniform white background to get a clear view onto the cuticle in 3D renderings (however, some remaining folds inside the tissue cannot be avoided). The stack of greyscale images was imported in Amira 5.6 (FEI), consecutive images were aligned, ROI cropped resulting in a 3D volume of 1011 × 1092 × 297 voxels (voxel size 0.98 × 0.98 × 1.54 µm 3 ), LUT inverted, volume rendering applied, LUT re-inverted, zoomed and rotated in appropriate perspectives to show both sides of the stack with bounding box (Fig. 2A,B), screenshots were saved in JPEG format (8 bit greyscale, 1390 × 1065 px, 1.28 µm/px).
Transmission Electron microscopy. Subsequent to the semithin section series some ultrathin sections (70 nm) were cut with a diamond knife and mounted on formvar-coated copper slot-grids (Agar Scientific G2500C). The slices were double-stained with 8% uranyl acetate and lead citrate 46 and imaged by using a FEI Morgagni 268 TEM at 80 kV.