1 INTRODUCTION Color patterns on animals affect their fitness in terms of reproductive success and survivorship (Caro, 2005). For example, vivid colored ornaments can increase the chance to be a potential mate by conspicuousness, and color patterns can reduce predation risk by concealment. Therefore, the study of animal coloration is important to understand animal behavior and evolution. The function of color patterns depends on the sensory perception of the viewer. Based on the colorblindness of humans in weak light, mammals (i.e., most are nocturnal species) were previously assumed to lack color vision (Jacobs, 1993). However, recent studies showed at least 24 species of non-primate mammals have color vision, including 10 that were nocturnal (Kelber et al., 2003). The finding that mammals can recognize color, suggests the possibility that they also can see fluorescence, as has been found in some other taxa (Arnold et al., 2002; Lim et al., 2007). Fluorescence is a by-product of excitation due to high energy radiation such as ultraviolet. The excitation changes invisible radiation such as ultraviolet to visible light. Although humans cannot recognize fluorescence under the normal light conditions, in a dimly lit room with black lamps, humans can see the fluorescence just like visible color. Fluorescent patterns are known to be important in mate choice of some arachnids and birds (Arnold et al., 2002; Lim et al., 2007). However, little is known about the occurrence, let alone function, of fluorescent color patterns in mammals. 2 In Metatherian mammals (marsupials), 29 species within the Family Didelphidae (New world opossums) have fluorescent patterns on their pelage (Pine and Abravaya, 1978; Pine, 1981; Meisner, 1983; Pine and Handley, 1984; Pine et al., 1985; Macrini, 2004) although their functions are unknown. Moreover, the hair of 14 species of Order Diprotodontia (kangaroos and wallabies etc.) and Eastern Quoll (Dasyurus viverrinus) (Bolliger, 1944; Nicholls and Rienits, 1971, 1973) contains fluorescent substances. There are few data regarding fluorescence in Eutherian (placental) mammals. A second potential source of color is ultraviolet reflectance. Ultraviolet is defined as an electronic radiation with a wavelength of 4 - 400 nm, but only the longer wavelengths of ultraviolet (290 - 400nm) are available and reflected as potential source of sensory information (Daniels, 1964; Koller, 1965; Robinson, 1966). Most animals that have ultraviolet vision have peak sensitivity around 360 nm wavelengths of ultraviolet (Jacobs, 1992; Tovée, 1995). As with fluorescent patterns (effectively), humans are blind to ultraviolet reflectance patterns; the color vision of human is restricted to the “visible spectrum” (380 – 760nm; Koller, 1965). Ultraviolet vision and ultraviolet reflectance patterns also function in mate-choice decisions for some animals, including some butterflies, spiders, fishes, lizards and birds (Silberglied, 1973; Brunton and Majerus, 1995; Andersson and Amundsen, 1997; Bennett et al., 1997; Hunt et al., 1999; Limbourg et al., 2004; Boulcott et al., 2005; Mougeot and Arroyo, 2006; Kemp and Rutowski, 2007; Lim et al., 2007). In contrast to “visible” (i.e., to humans) coloration and fluorescence, 3 ultraviolet reflectance patterns are less permanent. For example, nutritive condition can change the strength of ultraviolet reflectance (Lim and Li, 2006). In addition to sexual display, ultraviolet reflectance patterns also facilitate species discrimination. For instance, some species of birds that appear indistinguishable based on the visible spectrum have different ultraviolet reflectance patterns (Bleiweiss, 2004). This suggests that ultraviolet vision enables them to differentiate their own species from other species. There is no apparent relationship between ultraviolet reflectance patterns and visible color patterns (Bennett et al., 1997; Lyytinen et al., 2004), suggesting that ultraviolet vision and ultraviolet reflectance patterns confer additional (rather than redundant) information. Although researchers believed, until recently, that mammals were blind to ultraviolet radiation (Bennett et al., 1994), some mammals, such as the house mouse (Mus musculus), Mongolian jird (Meriones unguiculatus), and Botta’s pocket gophers (Thomomys bottae) were confirmed to possess ultraviolet vision (Jacobs, 1992). Moreover, Hut et al. (2000) concluded that nocturnal rodents are likely to have ultraviolet vision. Furthermore, it might be advantageous for nocturnal rodents to possess ultraviolet signals. In butterflies, for example, ultraviolet reflectance patterns occur more frequently in nocturnal species than diurnal species (Lyytinen et al., 2004), which appears to be associated with predation. Diurnal birds tend to have ultraviolet vision, whereas nocturnal birds tend not to (Koivula et al., 1997). My research investigates the taxonomic distribution and particular nature (e.g., 4 sexual dimorphism, anatomical location) of ultraviolet reflectance patterns and fluorescent patterns on rodents in an ecological hypothesis-driven framework to investigate possible adaptive functions. Although many biologists have made detailed examination of mammalian coloration with respect to the visible spectrum, there are no data on 365nm ultraviolet-induced ultraviolet reflectance and fluorescent patterns on rodents. If rodents possess these invisible (to humans) patterns, the effect of their coloration could be different from that of visible coloration. Therefore, I will assess the function of these invisible patterns by the association of life history and environmental characters. METHODS Specimens I examined 634 study skins (n = 2 per species) of 297 rodent species available at the Museum of Vertebrate Zoology (University of California, Berkeley) during April 2008 to February 2009; Dr. James Patton graciously allowed use of these specimens for this study. The specimens were chosen because they had associated collection location data (latitude, longitude and elevation), represented both sexes, and phylogenetic relationships (topology and branch lengths) had been previously assessed. Because preservation of skins does not affect the existence of ultraviolet reflection and fluorescence (Pine et al., 1985; Eaton and Lanyon, 2003; Hausmann et al., 2003), I was able to use these specimens. To maximize the number of species, I examined only one specimen of each sex from 292 species. 5 However, to investigate the possibility of the individual variation within species, I observed ultraviolet reflection and fluorescence of 10 individuals (5 males and 5 females) on five species, chose to represent four Rodentia suborders, including one species each from Sciuromorpha, Castorimorpha and Hystricomorpha and two species from Myomorpha. When many specimens of the species from the same locality were available, I chose the specimens with the best skin condition, collected at the highest elevation (to maximize exposure to ultraviolet radiation, which is blocked by aerosol [e.g., cloud], and favored most recently collected specimens. Ultraviolet reflection and Fluorescent patterns Both dorsal and ventral surfaces were examined on a non reflective plate in a handmade dark box (26×60×52cm.) for eliminating unwanted spectra from the environment (Plate 1). 6 Plate 1 Dark box 60cm 52cm 26cm Black vinyl cover Non reflectance plate Adjustable hooks Ultraviolet lamp UV protection goggle Peep-hole The filter CCD camera 7 The dorsal surface included head, ears, neck, body, fore-feet, hind-feet, fore-legs, hind-legs and tail. The ventral surface included face, neck, body, fore-legs, hind-legs, plantar tubercles and tail. In some individuals, I examined a lateral surface for observations of legs and face patterns. The plate was painted by Krylon ultra-flat camouflage Black #4290 to clearly see ultraviolet reflection from specimens. The former version of this paint, ultra-flat Black Krylon #1602, reflects 0.2 % at 350 nm (Wightman and Grum 1981). The box was made by two thick wood panels stood parallel on a thin wood panel and the top was covered by a thin wood panel. The remaining sides were covered by black vinyl which allowed easy access to set specimens and flexibility for the length of specimens. Because specimens’ size varied among species, the box had an adjustable plate with hooks; for specimens too large for the box, I used a darkroom located in the Museum of Vertebrate Zoology (UC Berkeley). To detect fluorescent patterns, I used a handheld ultraviolet lamp (UVL-56, 6-watt, 365nm, 115V, UVP Inc.) as suggested by Pine (1981). To detect the ultraviolet reflectance patterns, I used the ultraviolet lamp, an ultraviolet-transmitting, visible-absorbing filter (U-360, Hoya Corporation; maximum transmittance 355 – 365nm) and a CCD Black and White Video Camera (Sony XC-EU50). Because humans cannot see the ultraviolet spectrum, I used CCDs (Charge-coupled devices) in the video camera which are specifically sensitive to ultraviolet radiation in the 360nm range. The video camera with the filter recorded only the radiation corresponding to the ultraviolet vision of rodents (355 – 365nm) (Jacobs, 1992; Tovée, 1995). 8 Although some ultraviolet reflection is measured by spectrophotometers, these machines are no better than human discrimination (Bowers, 1956). In addition, spectrophotometers cannot detect ultraviolet reflectance patterns, chiefly because the objective size on spectrophotometers is limited, for example, to 1mm in diameter (Lim and Li, 2006), 2mm in diameter (Church et al., 1998) and 3mm² (Dresp et al., 2005; Dresp and Langley, 2006). Without samples of known ultraviolet patterns, this limitation prevents accurate results (Bowers, 1956). Life history and Ecological characters I used the classification of Mammal Species of the World (Carleton et al., 2005). Life history characters were found in the Mammalian Species series and several resources (i.e., Eisenberg, 1989; Emmons, 1997; Francis, 2008; Nowak and Paradiso, 1999; Parker, 1990; Reid, 1997; Smith and Xie, 2008; Strahan, 1983). Locomotion was classified as Biped, Limited biped (the species can stand, but cannot walk bipedally), and Quadruped. Activity of species was classified as Diurnal, Nocturnal (includes crepuscular) and Diel (day and night). Predators were classified as Insects, Fish, Reptiles, Aves, Mammals and None. To classify species habitats, I used the life zone chart by Holdridge (1967). The 15 life zones (e.g., desert) were based on mean annual biotemperature, mean annual precipitation and elevation. The mean annual biotemperature, the annual average Celsius temperatures at which vegetative growth occurs, was calculated by following the method of Holdridge (1967). The weather data were collected from World Weather 9 Records (published by the Smithsonian Institution [-1940], by the United States Weather Bureau [1941-1950] ) for specimens were collected before 1950 and Monthly Climatic Data for the World (published by the National Climatic Center, Environmental Data and Information Service, NOAA) for the specimens collected after 1951. I compared the latitude and longitude of the specimens with that of available stations and then I chose the nearest station and collected the weather data (i.e. temperature and precipitation). According to Holdridge’s (1967) guidelines, each habitat fell into seven altitudinal belts (Basal belt, Premontane, Lower Montane, Montane, Subalpine, Alpine, Nival) and one of 12 humidity provinces (Semiparched, Superarid, Peratid, Arid, Semiarid, Subhumid, Humid, Perhumid, Superhumid, Semisaturated, Subsaturated, Saturated). Data Analyses Two binomial response variables regarding ultraviolet reflectance or fluorescence were considered: whether or not the trait was conspicuous and, if so, whether the trait could serve as a signal. Determination of whether ultraviolet reflection patterns were conspicuous depended on environmental context (Fig. 1). 10 Figure 1 Flowchart is describing the determination process of conspicuousness for ultraviolet reflection patterns. 11 In desert and snow habitats, assumed to strongly reflect ultraviolet radiation, animals that reflected ultraviolet radiation were considered inconspicuous and those that absorbed ultraviolet radiation were considered conspicuous (to observers that see ultraviolet radiation). On the other hand, in habitats where ultraviolet radiation was primarily absorbed by vegetation (e.g. forested habitats), animals that reflected ultraviolet radiation were considered conspicuous and those that absorbed ultraviolet radiation were considered inconspicuous. Whether a conspicuous pattern constituted a potential signal further depended on which body surface relative to the locomotory mode of the species (Fig. 2). 12 Figure 2 Flowchart is showing how to classify conspicuous ultraviolet coloration and fluorescence pattern as possible signals. 13 Ultraviolet reflection patterns and fluorescence patterns on plantar tubercles or ventral surface of tail were never considered signals because these surfaces are rarely visible to other organisms. Before assessing possible adaptive significance of ultraviolet or fluorescent signal with respect to life history or environmental characters I assessed their phylogenetic independence using the Runs Test (Abouheif, 1999). The Runs Test was conducted in the program, Phylogenetic Independence Version 2.0 (Reeve and Abouheif, 2003) with completed phylogeny data modified from the super tree of Bininda-Emonds et al. (2007) in program Mesquite Version 2.6 (Maddison and Maddison, 2009). Phylogenetically non-independent data need to be tested for an association between traits (e.g., ultraviolet reflectance signal) and life history (i.e. activity and predators) or environmental (i.e. altitudinal belts, life zones, humidity provinces and elevation) variables among species by phylogenetic comparative methods (PCMs), which have been developed to reduce the effect of phylogenetic inertia (Janson, 1992). The association was tested by Generalized Estimating Equations (GEE) approach (Paradis and Claude, 2002), one of the PCMs can analyze not only non-independent data but also independent data (Paradis and Claude, 2002; Paradis, 2006). Moreover, the GEE approach can analyze binominal traits and continuous and categorical characters. The GEE approach was run using the APE package in R language (Paradis et al., 2004). When the GEE approach was applied, binominal variables were transformed to 14 “0” for absent and “1” for present. Some multi-nominal characters (i.e. altitudinal belts, life zones and humidity provinces) were reclassified to reduce the number of dummy variables (Table 1). 15 Table 1 Reclassification of multi-nominal variables for GEE approach Altitudinal belts Life zones Humidity provinces 1)Basal belt 2)Premontane, Lower Montane 3)Montane, Subalpine 1)Wet Tundra, Wet Forest, Rain Forest, Moist Forest 2)Steppe, Dry Forest, Thorn Steppe, Thorn Woodland, Very Dry Forest 3)Dry Scrub, Desert Scrub, Desert; 1)Perarid, Arid 2)Semiarid, Subhumid 3)Humid, Perhumid, Superhumid. 16 RESULTS Individual variation I observed little individual variation on the presence of ultraviolet reflection and fluorescence (Table 2). 17 Table 2 Individual variation of the presence of ultraviolet reflection and fluorescence among five rodent species was shown. Number of specimens with character present / total number of specimens, UV = ultraviolet, Dorsal = dorsal surface, Ventral = ventral surface Suborder Species Sciuromorpha Castorimorpha Myomorpha Myomorpha Hystricomorpha Tamiasciurus douglaasii Liomys salvini Zapus princeps Rattus steini Ctenomys haigi UV reflection Dorsal 0 / 10 0 / 10 8 / 10 0 / 10 9 / 10 UV reflection Ventral 1 / 10 8 / 10 10 / 10 0 / 10 10 / 10 Fluoresc ence Dorsal 0 / 10 9 / 10 0 / 10 10 / 10 10 / 10 Fluoresc ence Ventral 0 / 10 10 / 10 10 / 10 9 / 10 10 / 10 18 Phylogenetic independence Ultraviolet signal was significantly phylogenetically correlated (P=0.0003) as was sexual dimorphism in this trait (P=0.0003). Fluorescent signal was also significantly phylogenetically correlated (P=0.0003). However, sexual dimorphism in this trait was not significantly phylogenetically correlated (P=0.086). Ultraviolet signal One hundred ninety-eight species (67%) exhibited an ultraviolet signal. Ultraviolet signal was significantly associated with elevation (P < 0.001) (Fig. 3, Table 3). However, associations with life history characters (i.e. activity and predators) and other environmental variables (i.e. altitudinal belts, life zones and humidity provinces) were not significant. Among the species that exhibited ultraviolet signal, 51 species were sexually dimorphic with respect to this trait. 19 Fig.3 Association between ultraviolet signal and elevation 20 Table 3 Association between ultraviolet signal and life history or environmental variables by Generalized Estimating Equations (GEE) approach was shown. n = number of species, dfP = phylogenetic degrees of freedom, S.E. = standard error, *Test of the significance: F = 8.1459, df = 1, p=0.001 Variable Activity n 284 dfP 49.78 Reptilian predators Avian predators Mammalian predators Altitudinal belts 101 101 101 297 21.98 21.98 21.98 51.25 Life zones 297 51.25 Humidity provinces 297 51.25 Elevation 297 51.25 Estimate 0.3077 0.2145 0.5708 1.3434 -0.2760 -0.1794 -0.0112 0.2625 -0.3471 0.9709 0.6261 0.0002 S.E. 0.1665 0.3326 0.3961 0.2406 0.2417 0.1687 0.2728 0.1968 0.2762 0.1907 0.1993 6.1606 t 1.8481 0.6448 1.4411 5.5847 -1.1419 -1.0634 -0.0410 1.3336 -1.2570 5.0915 3.1422 2.8541 p 0.0708 0.5221 0.1643 1.5344 0.2664 0.2928 0.9675 0.1885 0.2147 5.5888 2.8374 0.0063* 21 Fluorescent signal Two hundred fifty-eight species (87%) exhibited a fluorescent signal that was significantly associated with avian predators (P < 0.01) (Fig. 4, Table 4). However, associations with other life history characters (i.e. activity, reptilian predators and mammalian predators) and environmental variables were not significant. Among the species that exhibited a fluorescent signal, 11 species were sexually dimorphic with respect to this trait. 22 Fig.4 Association between fluorescent signal and avian predators 23 Table 4 Association between fluorescence signal and life history or environmental variables by Generalized Estimating Equations (GEE) approach was shown. n = number of species, dfP = phylogenetic degrees of freedom, S.E. = standard error, *Test of the significance: F = 6.1536, df = 1, p = 0.01 Variable Activity n 284 dfP 49.78 Reptilian predators Avian predators Mammalian predators Altitudinal belts 101 101 101 297 21.98 21.98 21.98 51.25 Life zones 297 51.25 Humidity provinces 297 51.25 Elevation 297 51.25 Estimate 0.4976 -0.6607 -0.1822 0.7067 0.0120 -0.1070 0.0518 -0.1833 0.3349 -0.6559 -0.4496 -0.0001 S.E. 0.5081 0.3664 0.4441 0.2849 0.2875 0.2405 0.3202 0.2295 0.4340 0.2203 0.3334 0.0001 t 0.9795 -1.8031 -0.4103 2.4806 0.0418 -0.4451 0.1619 -0.7986 0.7718 -2.9772 -1.3485 -0.9464 p 0.3323 0.0777 0.6858 0.0217* 0.9670 0.6582 0.8721 0.4283 0.4439 0.0045 0.1837 0.3485 24 DISCUSSION Ultraviolet signal Among other taxa, conspicuous ultraviolet coloration is used for intra- or inter-specific communication like visual coloration (ex. Avilés et al., 2006; Viitala et al., 1995). However, ultraviolet coloration may be more influenced by radiation condition than visual coloration because ultraviolet is more scattered than longer wavelength, light (Bennett and Cuthill, 1994). “Altitude effect”, the increase in solar radiation with increasing altitude, could be expected to affect the presence of ultraviolet signal due to the increases in visibility at higher altitudes. In particular, radiation of 370 nm wavelength increases nine percent per 1000 m (Blumthaler et al., 1997). Consistent with this hypothesis, I found a trend toward increasing frequency of species with ultraviolet signal at increasing elevations. The rodents possessing ultraviolet signal may choose their habitat to maximize the conspicuousness of the signal. For example, some lek birds {i.e., cock-of-the rocks (Rupicola rupicola), white-throated manakins (Corapipo gutturalis), and white-fronted manakins (Lepidothrix serena)} apparently choose specific light conditions for their display to maximize visual contrast (Endler and Théry, 1996). On the contrary, the solar radiation stimulated the employment of ultraviolet signal among rodents living in high elevation. For example, among gonodactyloid stomatopods, the visual system of each species adapts its habitat light condition respectively (Cronin et al., 2000). Some avian predators such as Eurasian kestrels (Falco tinnunculus), 25 rough-legged buzzards (Buteo lagopus) and great grey shrikes (Lanius excubitor) detect their prey by using ultraviolet cues (Viitala et al., 1995; Koivula and Viitala, 1999; Probst et al., 2002). Although I found that more than half of the rodent species investigated possessed ultraviolet signal, this trait was not associated with avian or other predators, suggesting ultraviolet signal may not affect risk of predation. Although ultraviolet signal is an important cue for mate choice in some animals such as jumping spiders (Cosmophasis umbratica), zebra finches (Taeniopygia guttata), bluethroats (Luscinia s. svecica) and starlings (Sturnus vulgaris) (Bennett et al., 1996; Andersson and Amundsen, 1997; Bennett et al., 1997; Lim et al., 2007), only 51 species out of 198 species showed sexual dimorphism, suggesting ultraviolet signal was not primarily related to sexual ornamentation. However, most of the 51 species are promiscuous like other taxa which have conspicuous sexual dimorphism. Fluorescent signal The function of fluorescent signal is not clear especially in the terrestrial environment (Andrews et al., 2007; Mazel and Fuchs, 2003). Although fluorescent signal was considered as a warning against predators (orchid bees (Eufriesea niveofasciata); Nemésio, 2005), a threat to other males and predators (Mantis shrimp (Lysiosquillina glabriuscula); Mazel et al., 2004), a lure for prey (deep-sea siphonophores (Erenna); Haddock et al., 2005), and a cue for species recognition or intraspecific communication (loliginid squid (Alloteuthis subulata and Loligo vulgaris); Mäthger and Denton, 2001; reef fishes; Michiels et al., 2008), only the 26 function for courtship among jumping spiders (Cosmophasis umbratica) and budgerigars (Melopsittacus undulatus) (Arnold et al., 2002; Lim et al., 2007) was experimentally tested. Most rodents possessed fluorescent signal, especially those with avian predators. Avian hunting can be divided into visual hunting by diurnal species and acoustic hunting by nocturnal species (Mikkola, 1983; Sherman, 1985; Honkavaara et al., 2002). Therefore, not only the presence of avian predators but also the activity pattern of avian predators may affect predation risk. Interestingly, the species, either male or female has fluorescence on dorsal body, which I consider the most conspicuous fluorescent pattern, are likely the prey of diurnal avian species. This finding was somewhat counterintuitive and suggests the presence of a trade-off. If signal increases vulnerability to predators, it must also confer some benefit exceeding this cost such as ornamentation. Indeed, the species displaying sexual dimorphism in fluorescent signal more frequently had diurnal avian predators. Moreover, the species that fluoresced its dorsal body in both sexes tended to have nocturnal avian predators or not to have avian predators. This suggested that, like loliginid squid (Alloteuthis subulata and Loligo vulgaris) ( Mäthger and Denton, 2001), rodents may use fluorescent signal for species recognition or intraspecific communication in the absence of the pressure of predation risk. 27 CONCLUSION This research clearly showed that ultraviolet signal and fluorescent signal exist among rodents. The evolution of ultraviolet signal is probably encouraged more strongly in the presence of greater ultraviolet radiation especially in the reduced predation risk. Therefore, although the adaptive function of ultraviolet signal in rodents is probably different from that in other taxa, ultraviolet coloration in rodents, as with birds (Finger and Burkhardt, 1994; Parker, 2005), may be caused by the structure of hair or skin. My findings point to the possibility that fluorescence among rodents emerged by predator-prey interaction like spiders (Andrews et al., 2007). Therefore, rodents may mainly use fluorescent signal as a communication tool hidden from their predators.