Ecological associations among epidermal microstructure and scale characteristics of Australian geckos (Squamata: Carphodactylidae and Diplodactylidae)

Abstract

A first step in examining factors influencing trait evolution is demonstrating associations between traits and environmental factors. Scale microstructure is a well‐studied feature of squamate reptiles (Squamata), including geckos, but few studies examine ecology the of microstructures, and those focus mainly on toe pads. In this study, the ecomorphology of cutaneous microstructures on the dorsum was described for eight Australian species of carphodactylid (Squamata: Carphodactylidae) and 19 diplodactylid (Squamata: Diplodactylidae) geckos. We examined scale dimensions, spinule and cutaneous sensilla (CS) morphology, using scanning electron microscopy, and described associations of these traits with microhabitat selection (arboreal, saxicoline or terrestrial) and relative humidity of each species’ habitat (xeric, mesic or humid). We used a phylogenetic flexible discriminant analysis (pFDA) to describe relationships among all traits and then a modeling approach to examine each trait individually. Our analysis showed that terrestrial species tended to have long spinules and CS with more bristles, saxicoline species larger diameter CS and arboreal species tended to have large granule scales and small intergranule scales. There was high overlap in cutaneous microstructural morphology among species from xeric and mesic environments, whereas species from humid environments had large diameter CS and few bristles. Significant associations between epidermal morphology and environmental humidity and habitat suggest that epidermal microstructures have evolved in response to environmental variables. In summary, long spinules, which aid self‐cleaning in terrestrial geckos, are consistent with greater exposure to dirt and debris in this habitat. Long spinules were not clearly correlated to environmental humidity. Finally, more complex CS (larger diameter with more bristles) may facilitate better perception of environmental variation in geckos living in drier habitats.

Introduction

Skin, comprising the epidermis (outer layer) and the dermis (inner layer), provides vertebrates with direct protection from desiccation, radiation, moisture, irritants and infectious agents (Schliemann, 2015). The Oberhäutchen (the outermost layer of epidermis) of squamate reptiles expresses external relief in the form of microscopic structures collectively known as microornamentation or microstructures (Ruibal & Ernst, 1965; Ruibal, 1968; Gans & Baic, 1977; Jackson & Sharawy, 1980; Bauer & Russell, 1988; Arnold & Poinar, 2008). Microstructures have been studied for over a century (Cartier, 1872; Schmidt, 1912; Maderson, 1964), with most contributions focusing on either the taxonomic utility of these structures (Stille, 1987; Harvey, 1993; Ananjeva & Matveyeva‐Dujsebayeva, 1996; Harvey & Gutberlet, 2000; Bucklitsch et al. 2016), or their possible function (Hiller, 1968; Hazel et al. 1999; Russell, 2002; Autumn, 2006; Berthé et al. 2009; Russell et al. 2014). Some studies have demonstrated that microstructures vary adaptively with specific life‐histories; for example, subdigital (toepad) microstructures are associated with habitat use in some geckos and iguanids (Höfling & Renous, 2009; Collins et al. 2015). There have, however, been few studies on the relationship between the ecology of geckos in relation to microstructures that are not associated with locomotion. Such studies are necessary because they can tentatively demonstrate the functions of cutaneous morphological traits (Hagey et al. 2014).

Geckos are an ancient (> 100 myo) taxon of primarily nocturnal lizards, exhibiting great diversity and a worldwide distribution (Gamble et al. 2008; Garcia‐Porta & Ord, 2013), occupying a wide variety of habitats and environmental conditions (Wilson & Swan, 2013; Böhme & Sander, 2015). Their wide habitat use makes them excellent candidates for studying epidermal microstructural features, and the associations of these features with habitats and climatic zones.

The external surface of the skin of geckos exhibits small, hair‐like microstructures called spinules or spines (Ruibal, 1968), which vary in length (from 0.3 to 3.0 μm) among species (Stewart & Daniel, 1975; Rosenberg et al. 1992; Bauer, 1998; Peattie, 2009; Spinner et al. 2013a). Recent studies have demonstrated that high spinule length and density are correlated with hydrophobicity (Vucko, 2008; Hiller, 2009) and produce self‐cleaning and antibacterial epidermal characteristics (Watson et al. 2015a,2015c, 2016). Bacteriocidal properties occur because large bacteria are pierced by the spinules, whereas small bacteria are damaged by compression, stretching or tearing between the spinules (Li et al. 2016). Spinule length or density, or both, may therefore increase if geckos live in an environment exposed to dirt and debris, or with a higher load of harmful microorganisms. For example, we would expect terrestrial species to have longer or more densely packed spinules (or both) compared with arboreal or saxicoline (rock‐dwelling) species, because the accumulation of dirt and debris is more prevalent on the ground than on higher strata (Ungar et al. 1995) and terrestrial species are also more exposed to microorganisms than are arboreal or saxicoline species (Nunn et al. 2000; McCabe et al. 2015). Conversely, species from humid areas, such as rainforests, may have long or dense spinules, or both, because bacterial and fungal growth rates may be greater in rainforests, leading to an increased prevalence of these microorganisms (Bouskill et al. 2012). If microstructures are indeed adaptive and are linked to species’ habitats, we would expect to see consistent variation in spinule morphology among habitat types.

Gecko skin also features mechanoreceptors, called cutaneous sensilla (Hiller, 1976; Bauer & Russell, 1988). For example, the distribution and form of cutaneous sensilla on the dorsal surface of the feet of the Tokay gecko (Gekko gecko) aid in the correct placement of the toes to maximise adhesive toe pad function (Lauff et al. 1993). Also, the density and distribution of cutaneous sensilla on the tail of the leopard gecko (Eublepharis macularius) suggests that they mediate the location of tail breakage and the movement of the tail after breaking (Russell et al. 2014). Other authors have argued, however, that mechanoreception cannot be the sole functional explanation for the huge variation in the distribution and morphology of cutaneous sensilla in different taxa and suggest additional functions, such as the perception of humidity or temperature (Ananjeva et al. 1991; Matveyeva & Ananjeva, 1995). Therefore, although the functional significance of sensilla is still unclear, it is likely that they fulfill different functions on different parts of the body. If cutaneous sensilla do play a role in perception of temperature or humidity, their morphology or distribution should vary among species exposed to different environmental conditions. These differences should be most obvious on body regions weakly influenced by other mechanoreceptive functions, such as toe placement, tail breakage or prey consumption. For example, terrestrial species in arid environments may depend upon detailed perception of disturbances on their body surface, as they are more likely to be susceptible to desiccation by wind (Tingley & Shine, 2011). If cutaneous sensilla do play a role in detection of humidity or temperature, terrestrial species from arid biomes may have more, or more complex (i.e. larger, with more bristles, or bristles with more complex structure), cutaneous sensilla, or both. On the other hand, microhabitat use (e.g. burrow or crevice use) may confound simple correlations between habitat temperature and especially humidity, reducing correlations among skin features and microhabitats (Aguilar & Cruz, 2010).

In this study, we examine a range of Australian carphodactylid (Squamata: Carphodactylidae) and diplodactylid geckos (Squamata: Diplodactylidae), using scanning electron microscopy to describe the microstructure of the Oberhäutchen. Then, accounting for phylogeny, we test the associations between scale characteristics and habitat and between scale characteristics and the humidity of the environment, to determine whether (1) terrestrial species have longer or more densely arranged spinules than arboreal or saxicoline species do; (2) species from humid habitats have longer or more densely arranged spinules than species from more arid habitats; and (3) terrestrial species from drier habitats have more, or more complex, cutaneous sensilla than arboreal species, or species from more humid habitats.

Materials and methods

Study species

The two closely related families Carphodactylidae and Diplodactylidae originate from Australia and occupy nearly all habitats and ecosystems. Carphodactylines tend to be larger and include more species occupying humid habitats. Geckos from Queensland, Australia, were captured at night by hand, and those from South Australia were captured in 20‐L pitfall traps. Only healthy adult specimens were returned to the laboratory, and dental imprint moulds were taken from the skin 4–8 days after each individual had shed, as described by Vucko et al. (2008). Since it was not possible to wait for shedding in Diplodactylus wiru, Nephrurus laevissimus and Nephrurus levis, moulds were obtained at an unknown stage of the resting phase of their shedding cycle, but were still considered usable, as individual scales in good condition could be located and measured. In total, 27 species representing 10 genera of the Carphodactylidae and Diplodactylidae were examined (Table 1). Species were selected to include a broad range of habitats and humidity regimens.

Table 1.

Overview of all species examined in this study, showing collection site, humidity and habitat category assigned, morphometrics and scale sizes

Species	Collection site	Habitat	Humidity
Carphodactylidae
Carphodactylus laevis Günther, 1897	Mount Hypipamee NP, QLD (17° 25′ 33″ S, 145° 29′ 10″ E)	Terrestrial	Hydric
Nephrurus asper Günther, 1876	Mingela Range, QLD (20° 06′ 39″ S, 146° 52′ 39″ E)	Terrestrial	Mesic
Nephrurus laevissimus Mertens, 1958	Great Victoria Desert, SA (29° 21′ 07″ S, 132° 41′ 57″ E)	Terrestrial	Xeric
Nephrurus levis De Vis, 1886	Great Victoria Desert, SA (29° 21′ 07″ S, 132° 41′ 57″ E)	Terrestrial	Xeric
Phyllurus amnicola Couper et al. 2000	Bowling Green Bay NP, QLD (19° 28′ 50″ S, 146° 58′ 59″ E)	Saxicoline	Hydric
Phyllurus nepthys Couper et al. 1993	Eungella NP, QLD (21° 08′ 43″ S, 148° 29′ 57″ E)	Arboreal	Hydric
Phyllurus ossa Couper et al. 1993	Airlie Beach, QLD (20° 20′ 08″ S, 148° 40′ 19″ E)	Saxicoline	Hydric
Saltuarius cornutus (Ogilby, 1892)	Crater Lakes NP, Lake Eacham Section, QLD (17° 17′ 12″ S, 145° 37′ 17″ E)	Arboreal	Hydric
Diplodactylidae
Amalosia rhombifer (Gray, 1845)	Airlie Beach, QLD (20° 20′ 08″ S, 148° 40′ 19″ E)	Arboreal	Mesic
Diplodactylus ameyi Couper & Oliver, 2016	Winton, QLD (22° 27′ 26″ S, 142° 57′ 39″ E)	Terrestrial	Xeric
Diplodactylus conspicillatus Lucas & Frost, 1897	Great Victoria Desert, SA (29° 21′ 07″ S, 132° 41′ 57″ E)	Terrestrial	Xeric
Diplodactylus platyurus Parker, 1926	Mingela Range, QLD (20° 01′ 56″ S, 146° 48′ 17″ E)	Terrestrial	Mesic
Diplodactylus tessellatus (Günther, 1875)	Boulia, QLD (22° 35′ 50″ S, 139° 42′ 59″ E)	Terrestrial	Xeric
Diplodactylus wiru Hutchinson et al. 2009	Ifould Lake, SA (30° 52′ 60″ S, 132° 09′ 00″ E)	Terrestrial	Xeric
Lucasium damaeum (Lucas & Frost, 1896)	Great Victoria Desert, SA (29° 24′ 13″ S, 132° 50′ 19″ E)	Terrestrial	Xeric
Lucasium immaculatum (Storr, 1988)	Winton, QLD (22° 28′ 34″ S, 142° 55′ 45″ E)	Terrestrial	Xeric
Lucasium steindachneri (Boulenger, 1885)	Mingela Range, QLD (20° 08′ 06″ S, 146° 52′ 32″ E)	Terrestrial	Mesic
Lucasium stenodactylum (Boulenger, 1896)	Great Victoria Desert, SA (29° 25′ 37″ S, 132° 56′ 42″ E)	Terrestrial	Xeric
Oedura bella Oliver & Doughty, 2016	Mt. Isa, QLD (20° 51′ 42″ S, 139° 27′ 42″ E)	Saxicoline	Xeric
Oedura castelnaui (Thominot, 1889)	Mingela Range, QLD (20° 22′ 06″ S, 146° 57′ 39″ E)	Arboreal	Mesic
Oedura cincta De Vis, 1888	Winton, QLD (22° 27′ 18″ S, 142° 58′ 18″ E)	Arboreal	Xeric
Oedura coggeri Bustard, 1966	Hidden Valley, QLD (19° 00′ 06″ S, 146° 04′ 47″ E)	Saxicoline	Mesic
Oedura monilis De Vis, 1888	Eungella NP, QLD (21° 10′ 07″ S, 148° 30′ 09″ E)	Arboreal	Mesic
Rhynchoedura ormsbyi Wells & Wellington, 1985	Winton, QLD (22° 28′ 34″ S, 142° 55′ 45″ E)	Terrestrial	Xeric
Strophurus krisalys Sadlier et al. 2005;	Mt. Isa, QLD (20° 35′ 51″ S, 139° 34′ 10″ E)	Arboreal	Xeric
Strophurus taeniatus (Lönnberg & Anderson, 1913)	Mt. Isa, QLD (20° 49′ 30″ S, 139° 27′ 42″ E)	Arboreal	Mesic
Strophurus williamsi (Storr, 1983)	Mingela Range, QLD (20° 12′ 56″ S, 146° 52′ 20″ E)	Arboreal	Mesic

Descriptive terminology

The skin of Australian Carphodactylidae and Diplodactylidae is covered by up to three types of scales, termed tubercles (enlarged, conical scales), granules (medium‐sized scales) and intergranules (smaller scales surrounding the medium‐sized scales; Fig. 2A ‐ C) in accordance with Vanderduys (2016).

Descriptive terminology for microstructures follows that of Ruibal (1968), Peterson & Williams (1981), Peterson (1984), Irish et al. (1988), Bauer & Russell (1988), Lang (1989), Harvey (1993), Harvey & Gutberlet (1995) and Arnold (2002), and includes some new terms. Hair‐like microstructures (spinules) are surrounded by connecting radial lines (struts) with small indentations (pits) in between the spinules and the struts (Fig. 1A). The long hair‐ or brush‐like structures arising from the cutaneous sensilla are termed bristles (Fig. 1B). In some species, the bristles themselves are covered with microscopic projections called setules (Fig. 1C). Uprisings of the scale surface are generally termed knobs if they are devoid of microstructures, or hillocks if they are covered by spinules (Fig. 1D). As these two structures appear otherwise similar, and as knobs arise out of hillocks in some species, we have combined them for statistical analyses.

Figure 1.

SEM images showing the microstructure terminology used for this study. (A) Long spinules, surrounded by pits and struts (Rhynchoedura ormsbyi). (B) A cutaneous sensillum (CS) of Lucasium damaeum with a single bristle (BR). (C) A cutaneous sensillum of Nephrurus levis has multiple bristles (BR), each covered by setules, and is surrounded by a moat (MO). (D) Detail of a knob of Nephrurus asper, arising from the top of a setae covered hillock.

The categories for substrate use were terrestrial, saxicoline and arboreal, and categories for environmental relative humidity were xeric, mesic and hydric. Lizards from xeric conditions were collected from deserts in central Australia, lizards from mesic conditions were collected from open savannah woodland in northern Australia and lizards from hydric conditions were collected from the rainforests of the Australian wet tropics. Geckos were assigned to habitat‐use and relative humidity groups using published data (Wilson & Swan, 2013; Cogger, 2015) and observations of habitat‐use recorded during collection (Table 1).

Epidermal microstructures

Dorsal epidermal scale moulds were taken from the mid‐dorsal region halfway between the front and the hind limbs, because microstructures (especially cutaneous sensilla) in this area may be less likely to perform specialised functions such as detecting body movement or posture, tail breakage or prey capture (Lauff et al. 1993; Russell et al. 2014). Detailed epoxy‐resin moulds of live specimens were made according to methods described by Vucko et al. (2008). Images of microstructures of each species were taken from the epoxy‐resin moulds with a JEOL JSM‐5410LV (JEOL, Tokyo, Japan) scanning electron microscope (SEM) at magnifications between ×15 and ×10 000 and analysed using imagej (V.1.36b Schneider et al. 2012). In all SEM images reproduced here, unless otherwise stated, the anterior of the animal is towards the top of the image. imagej was also used to measure all scale characteristics and microstructures observed. Measurements included the area of granules and intergranules [mm² (n = 50); Table 2], the length of spinules [μm (n = 50); Table 3, the density of spinules per 10 μm² (n = 3; Table 3), the diameter of pits [μm (n = 40); Table 3, the density of pits per 5 μm² (n = 3; Table 3), the number of cutaneous sensilla per scale (n = 10) and per mm² (Table 4), the number of bristles per sensor (n = 10) and per mm² (Table 4), the diameter of cutaneous sensilla [μm (n = 10); Table 4), and the percentage of each granule covered by knobs or hillocks, including partial and fully formed knobs and thick spinule‐like structures (n = 10), with n representing the number of features measured per individual (Table 3). Cutaneous sensilla and number of bristles per mm² on each gecko were calculated from the same images used for counts per scale.

Table 2.

Morphometric and scale measurements for all species in this study. Sample size indicates the number of specimens subjected to morphometric measurements. When two scale sizes are given, the higher number is for the scales along the dorsal midline, and the smaller number is for the remaining dorsal body area

Species	Sample size	SVL (mm)	Mass (g)	Granule size	Intergranule size
Carphodactylidae
Carphodactylus laevis	2	110.24 ± 13.09	28.52 ± 2.65	0.06 ± 0.02	0.01 ± 0.003
Nephrurus asper	11	91.84 ± 7.87	18.42 ± 6.13	0.04 ± 0.01	0.01 ± 0.001
Nephrurus laevissimus	2	59.79 ± 1.58	4.06 ± 048	0.02 ± 0.004	0.001 ± 0.0003
Nephrurus levis	3	68.75 ± 13.41	8.76 ± 3.28	0.03 ± 0.01	0.004 ± 0.001
Phyllurus amnicola	1	92.00	13.42	0.06 ± 0.01	0.01 ± 0.003
Phyllurus nepthys	3	87.17 ± 10.06	11.56 ± 4.26	0.03 ± 0.01	0.003 ± 0.001
Phyllurus ossa	5	77.29 ± 7.05	8.36 ± 2.58	0.03 ± 0.01	0.003 ± 0.001
Saltuarius cornutus	13	134.27 ± 8.95	35.62 ± 9.02	0.07 ± 0.02	0.008 ± 0.002
Diplodactylidae
Amalosia rhombifer	6	56.31 ± 3.21	3.01 ± 0.16	0.03 ± 0.01	0.004 ± 0.001
Diplodactylus ameyi	3	52.02 ± 4.33	3.20 ± 0.61	0.13 ± 0.02, 0.06 ± 0.01	0.01 ± 0.003, 0.01 ± 0.002
Diplodactylus conspicillatus	2	58.13 ± 0.33	4.02 ± 0.28	0.11 ± 0.01, 0.06 ± 0.01	0.01 ± 0.002, 0.01 ± 0.002
Diplodactylus platyurus	8	47.53 ± 2.96	1.98 ± 0.33	0.08 ± 0.01, 0.04 ± 0.01	0.003 ± 0.001, 0.004 ± 0.001
Diplodactylus tessellatus	11	47.31 ± 7.36	2.22 ± 0.61	0.13 ± 0.02, 0.08 ± 0.02	0.02 ± 0.003, 0.01 ± 0.002
Diplodactylus wiru	2	55.35 ± 5.37	4.93 ± 0.19	0.18 ± 0.03, 0.07 ± 0.01	0.01 ± 0.002, 0.01 ± 0.001
Lucasium damaeum	2	57.31 ± 3.76	2.45 ± 0.48	0.05 ± 0.01	0.01 ± 0.001
Lucasium immaculatum	4	51.74 ± 5.23	2.65 ± 0.76	0.03 ± 0.01	0.004 ± 0.001
Lucasium steindachneri	26	55.72 ± 3.58	2.72 ± 0.53	0.03 ± 0.01	0.004 ± 0.001
Lucasium stenodactylum	2	46.79 ± 13.11	2.3 ± 1.45	0.05 ± 0.01	0.01 ± 0.001
Oedura bella	1	77.41	10.97	0.3 ± 0.07	0.03 ± 0.007
Oedura castelnaui	8	92.09 ± 3.11	15.36 ± 2.82	0.26 ± 0.04	0.01 ± 0.003
Oedura cincta	8	90.48 ± 5.53	13.54 ± 1.78	0.24 ± 0.05	0.02 ± 0.004
Oedura coggeri	7	79.03 ± 3.57	7.34 ± 1.82	0.3 ± 0.05	0.02 ± 0.003
Oedura monilis	15	98.05 ± 5.75	16.04 ± 2.71	0.39 ± 0.07	0.03 ± 0.008
Rhynchoedura ormsbyi	3	49.38 ± 8.64	1.92 ± 0.49	0.02 ± 0.004, 0.02 ± 0.003	0.003 ± 0.001, 0.003 ± 0.001
Strophurus krisalys	16	70.29 ± 6.19	5.32 ± 1.47	0.1 ± 0.03	0.02 ± 0.003
Strophurus taeniatus	1	43.58	1.16	0.04 ± 0.01	0.01 ± 0.001
Strophurus williamsi	19	58.52 ± 4.98	3.14 ± 0.64	0.07 ± 0.02	0.01 ± 0.002

Table 3.

Measurements of spinules, pits and percentage of area covered by knobs or hillocks (knobbiness index). Sample size indicates the number of specimens subject to morphometric measurements

Species	Sample size	Spinule length (μm) (n = 50)	Spinule density (per 10 μm²) (n = 3)	Pit diameter (μm) (n = 40)	Pit density (per 5 μm²) (n = 3)	Knobines index (%) (n = 10)
Carphodactylidae
Carphodactylus laevis	1	0.73 ± 0.07	477 ± 12.2	0.24 ± 0.03	232 ± 5.3	11.21
Nephrurus asper	2	0.28 ± 0.004	112 ± 3.8	0.25 ± 0.04	164 ± 6.5	6.85
Nephrurus laevissimus	1	0.42 ± 0.05	347 ± 10.1	0.28 ± 0.02	164 ± 12.5	11.32
Nephrurus levis	2	0.3 ± 0.07	62 ± 4.5	0.29 ± 0.03	151 ± 3.1	4.32
Phyllurus amnicola	1	0.47 ± 0.06	174 ± 18.9	0.29 ± 0.04	138 ± 5.6	10.09
Phyllurus nepthys	2	0.54 ± 0.1	170 ± 25.5	0.3 ± 0.03	186 ± 6.7	29.81
Phyllurus ossa	1	1.05 ± 0.16	75 ± 5.5	0.33 ± 0.04	144 ± 6.6	25.97
Saltuarius cornutus	1	0.46 ± 0.07	120 ± 1.7	0.28 ± 0.03	170 ± 4.7	5.53
Diplodactylidae
Amalosia rhombifer	1	0.79 ± 0.1	112 ± 3.2	0.32 ± 0.03	163 ± 6.1	0
Diplodactylus ameyi	1	0.91 ± 0.07	170 ± 1.5	0.22 ± 0.03	130 ± 5.5	0
Diplodactylus conspicillatus	1	0.99 ± 0.14	169 ± 6.03	0.25 ± 0.03	160 ± 3.5	0
Diplodactylus platyurus	1	0.98 ± 0.09	144 ± 5.5	0.36 ± 0.04	134 ± 3.1	0
Diplodactylus tessellatus	1	0.48 ± 0.05	343 ± 26.8	0.31 ± 0.03	180 ± 3	0
Diplodactylus wiru	1	0.28 ± 0.05	159 ± 12.1	0.34 ± 0.04	135 ± 10.97	0
Lucasium damaeum	2	0.98 ± 0.14	213 ± 9.6	0.26 ± 0.02	191 ± 5.03	0
Lucasium immaculatum	1	1.03 ± 0.17	84 ± 6.3	0.31 ± 0.05	152 ± 4.6	0
Lucasium steindachneri	1	0.94 ± 0.09	130 ± 7.5	0.26 ± 0.02	182 ± 6.8	0
Lucasium stenodactylum	1	0.97 ± 0.1	115 ± 3.4	0.24 ± 0.02	181 ± 4.5	0
Oedura bella	1	0.77 ± 0.09	300 ± 10.1	0.29 ± 0.03	162 ± 5.5	0
Oedura castelnaui	2	0.97 ± 0.1	269 ± 14.4	0.39 ± 0.04	127 ± 5.03	0
Oedura cincta	1	0.97 ± 0.1	274 ± 10.6	0.3 ± 0.03	174 ± 2.1	0
Oedura coggeri	3	0.84 ± 0.08	287 ± 0.6	0.25 ± 0.03	267 ± 2.6	0
Oedura monilis	2	0.86 ± 0.07	232 ± 7.4	0.29 ± 0.03	158 ± 5.03	0
Rhynchoedura ormsbyi	3	1.12 ± 0.13	77 ± 2.4	0.26 ± 0.03	233 ± 10.4	0
Strophurus krisalys	2	0.43 ± 0.07	255 ± 11.5	0.26 ± 0.03	112 ± 3.1	0
Strophurus taeniatus	1	0.96 ± 0.09	95 ± 5.8	0.23 ± 0.03	233 ± 7.4	0
Strophurus williamsi	1	0.43 ± 0.05	117 ± 4.04	0.29 ± 0.03	126 ± 13.2	0

Table 4.

Overview of the measurements taken of the cutaneous sensilla. This table also shows the absence or presence of lenticular sense organs. Sample size indicates the number of specimens subject to morphometric measurements

Species	Sample size	Cutaneous sensilla (CS) per scale (n = 10)	CS per mm²	Bristles per sensor (n = 10)	Bristles per mm²	CS diameter (μm) (n = 10)	Lenticular sense organs
Carphodactylidae
Carphodactylus laevis	1	3–8	71	2–6	143	22.60 ± 1.60	‐
Nephrurus asper	2	0–2	50	4–7	200	24.51 ± 1.68	‐
Nephrurus laevissimus	1	1–2	60	5–9	300	22.36 ± 1.23	‐
Nephrurus levis	2	1	20	4–6	80	24.25 ± 0.86	‐
Phyllurus amnicola	1	5–10	75	1	75	22.57 ± 1.14	Y*
Phyllurus nepthys	2	1–3	7	1	7	23.39 ± 1.76	‐
Phyllurus ossa	1	1–3	13	1	13	26.11 ± 0.91	‐
Saltuarius cornutus	1	0–5	63	1–2	63	21.69 ± 1.45	‐
Diplodactylidae
Amalosia rhombifer	1	1	33	0	0	19.98 ± 1.07	Y
Diplodactylus ameyi	1	1–3	33	1	33	15.82 ± 1.78	Y
Diplodactylus conspicillatus	1	2–4	167	1	167	15.60 ± 0.75	Y
Diplodactylus platyurus	1	1–3	35	1	35	15.60 ± 0.86	Y
Diplodactylus tessellatus	1	3–6	50	1–2	50	14.56 ± 1.39	Y
Diplodactylus wiru	1	2–5	100	1	100	16.00 ± 1.23	Y
Lucasium damaeum	2	0–1	133	1	133	17.89 ± 1.64	Y
Lucasium immaculatum	1	0–1	75	1	75	20.07 ± 0.74	‐
Lucasium steindachneri	1	0–1	67	1	67	20.17 ± 0.63	Y
Lucasium stenodactylum	1	1–3	13	1	13	17.49 ± 1.24	Y
Oedura bella	1	1–2	50	1	50	18.88 ± 0.45	‐
Oedura castelnaui	2	5–9	20	1	20	19.18 ± 1.26	Y
Oedura cincta	1	2–3	33	1	33	18.72 ± 0.45	Y
Oedura coggeri	3	2–4	100	1	100	19.04 ± 1.93	Y
Oedura monilis	2	11–14	14	1	14	17.78 ± 0.92	Y
Rhynchoedura ormsbyi	3	1	86	1	86	17.26 ± 0.63	‐
Strophurus krisalys	2	1–2	100	0	0	19.88 ± 1.02	‐
Strophurus taeniatus	1	1–3	36	0	0	18.20 ± 1.02	‐
Strophurus williamsi	1	1	33	1	33	17.84 ± 0.79	‐

^*In P. amnicola, lenticular sense organs are only present on the tubercle scales.

Statistical analyses

Supertree construction

A Matrix Representation with Parsimony (MRP) supertree was constructed using the following data and methods: Trees were derived from the literature (Donnellan et al. 1999; Hoskin et al. 2003; Han et al. 2004; Melville et al. 2004; Oliver et al. 2007, 2012; Oliver & Bauer, 2011) and an unpublished tree by P. Oliver. Only maximum‐likelihood trees from these sources were used, as these trees tended to be more resolved. Trees were encoded using both Baum–Ragan (Baum, 1992; Ragan, 1992) and Purvis (1995) coding, with the program supertree version 0.85b (Salamin et al. 2002). For each coding, tree characters were considered as both ordered and unordered, producing four tree matrices. The unordered character matrices were analysed using Wagner parsimony (Eck & Dayhoff, 1966; Kluge & Farris, 1969) and the ordered matrices were analysed using Camin–Sokal parsimony (Camin & Sokal, 1965). To build the trees, species were added to the tree in random order (10 000 times) and a heuristic search algorithm with local and global rearrangements was used to find the most parsimonious trees. The best 100 trees were saved in each analysis and an extended majority‐rule consensus tree was constructed from these trees for each matrix. The consensus trees were then pruned to include only taxa for which there were morphological data. The relationships of the Diplodactylus conspicillatus species complex were assigned using the species in Oliver et al. (2014) and Couper & Oliver (2016), and their relationships within the species complex were assumed to follow from their biogeographical ranges and locations (species with closer distributions were considered closer relatives). Of these four matrices we used the Camin–Sokal tree for subsequent analyses, in accordance with the published literature (see Results). Phylogenetic analyses were performed using the MIX and CONSENSE programs in the phylip suite v. 3.696 (Felsenstein, 2018).

The supertree approach used here does not produce branch lengths, which are necessary for phylogenetic comparative analyses. Thus, three types of arbitrary branch lengths were used: those of Grafen (1989), Pagel (1992) and Nee (cited in Pagel, 1992). For each of the three branch length types, a tree with the same topology as above was constructed and the performance of each branch length type was assessed by calculating phylogenetically independent contrasts (PICs) for each (log‐transformed) trait and set of branch lengths, and then calculating the Pearson correlation coefficient between the PICs and their standard error, following Garland et al. (1992). For each trait, the branch length type that produced the lowest (absolute value) correlation was chosen. As the regression analysis of PICs is equivalent to a phylogenetic generalised least squares (PGLS) regression assuming a Brownian motion model of evolution (Blomberg et al. 2012), the correct standardisation of PICs by an appropriate set of branch lengths means that set of branch lengths is appropriate for analyses using PGLS. Grafen's branch lengths were computed using the compute.brlen function in the ape package for R (Paradis et al. 2018; R Core Team, 2018). Pagel's and Nee's branch lengths were calculated using the PDAP:PDTREE package for mesquite (Midford et al. 2005; Maddison & Maddison, 2018).

Multivariate analyses

A discriminant function analysis (DFA) was done following Motani & Schmitz (2011). This method extends the flexible discriminant analysis (FDA) methods of Hastie et al. (1994) to include a PGLS step with a Pagel's lambda model to allow flexibly for potential phylogenetic effects in both the trait variables and the classification variables (habitat use or relative humidity). The optimal value of lambda (minimising the residual sum of squares, RSS) was quantified using the optim function in R and the plot of lambda vs. RSS was used to determine whether the relationship between these two variables was unimodal or monotonic. Confusion matrices were extracted from each analysis and we computed the classification performance measures of Garczarek (2002) using the ucpm function in the klaR package for R (Weihs et al. 2005). Canonical function plots were generated and group centroids with 50% and 95% confidence ellipses based on the standard error were calculated. The coefficients of the FDA function were used to determine the characteristics that loaded most heavily on the two discriminant functions (pFDA1 and pFDA2). The plots of the pFDA analysis were generated using the R package ggplot2 (Wickham, 2009).

Univariate analyses of individual traits

All the trait variables were log‐transformed, except for presence/absence of lenticular sense organs, which was treated as a binary variable. All continuous traits were analysed as response variables using PGLS, fitting the two explanatory variables (habitat use and relative humidity) in separate models, as the dataset was too small to include both explanatory variables in the same model. Because of the tight correlation between SVL and log Body Mass (see Fig. S1), these two explanatory variables were included and removed together in the models, thus avoiding the need for variable selection between them (Harrell, 2015). We considered these two variables together to represent body size. Four models for each explanatory variable were fitted: (1) the ‘full’ model with explanatory variables interacting with body size; (2) an ‘additive’ model which included body size, but did not interact with the other explanatory variable; (3) a model with only the explanatory variable and (4) a model with no explanatory variables (other than an intercept). In addition, we fitted evolutionary models assuming Brownian motion, Independence and Pagel's lambda model for each combination of explanatory variables, resulting in 12 models fitted per response variable. Branch lengths for each response variable were chosen based on the results of the analysis described above. Model adequacy was assessed by examining normal quantile‐quantile plots of the (normalised) residuals and by examining a plot of the residuals vs. the fitted values for each model.

For each response variable an information‐theoretic model averaging approach was used, based on the Akaike information criterion corrected for small samples (AICc; Burnham & Anderson, 2004). All models were fitted with maximum likelihood using the gls function in the nlme package for R (R core R Core Team, 2018, Pinheiro et al. 2018), with the exception of the presence/absence of lenticular organs, which was analysed with a Bayesian logistic regression using the MCMCglmm package for R (Hadfield, 2010; Hadfield & Nakagawa, 2010). Regression parameter estimates were averaged according to their AICc weights and Wald tests were used to assess statistical significance of the fully averaged parameter estimates. Model averaging was performed using the MuMIn package for R (Bartoń, 2018). Because of the large number of response variables examined (n = 13) and the large number of models for each variable (n = 12), P‐values for multiple testing were adjusted by controlling the false discovery rate using the method of Benjamini & Hochberg (1995).

Results

Morphology

All species observed in this study had several characteristic skin features in common:

• Dark lines were visible on all species, forming a web (see Fig. 4D). These represent either the cell borders of the Oberhäutchen cells, or the impression of the cell borders on the clear layer (the bottom‐most layer of the old epidermal generation), or both (Irish et al. 1988).

• All species have granules and smaller intergranules surrounding these (Figs 2A, 3A and 7A). All the Carphodactylidae and some species of the genus Strophurus also have larger tubercles (Figs 2C and 5B). The tubercles are scattered across the scale surface in all tubercled species, if not stated otherwise. All granules, intergranules and tubercles lack imbrication (i.e. overlap of scales) found in many other squamate species (Burstein et al. 1974; Peterson, 1984; Lang, 1989; Alibardi & Toni, 2006).

  Figure 2.

  

  Dorsal scales of Saltuarius cornutus (SEM images). (A) Overview showing granule (GS) and intergranule (IGS) scales. (B) Detail of a granule scale partially covered by large knobs (K) and four cutaneous sensilla (CS). (C) Tubercle scale covered with knobs (K) and cutaneous sensilla (CS) from the peak to the base. (D) Cutaneous sensillum with two bristles (BR) and a very shallow moat (MO).

  Figure 3.

  

  Dorsal scales of the genus Phyllurus (SEM images). Phyllurus nepthys is displayed in panels B–D, whereas panels (A,E,F) show Phyllurus amnicola (A) A tubercle scale with keels (K) forming a radial pattern. Cutaneous sensilla (CS) are scattered between the keels on higher elevations, whereas lenticular sense organs (LSO) can be found on the lower parts of the same areas. (B) A single granule scale of with three cutaneous sensilla (CS) and an area where large, thick spinules gradually fuse to form knobs (K) in the centre. (C) Detail of a cutaneous sensillum surrounded by a shallow moat (MO) that is level with the rest of the scale, with a very short, thick bristle (BR), with short, rounded barbs. (D) Detail of the central area of (B), showing partially formed knobs (PK). (E) A granule scale with eight cutaneous sensilla (CS) around the edges. Knobs (K) and areas of thicker and larger spinule‐like structures (TS) are more centrally located. (F) A slightly raised cutaneous sensillum surrounded by a wide moat (MO). Bristle (BR) is barbed and surrounded by longer spinules.

• The intergranules are always triangular but vary in size and exact shape.

• Scales contain spinules, surrounded by radial struts and pits, which were uniform except at the very edges of the scale (Fig. 1A). Therefore, one average measurement was calculated for each species, taken from the centre of the scales (Table 3).

• Tactile sense organs vary among species where (A) cutaneous sensilla occur in all species, but differ among species in distribution, number and type; and (B) lenticular sense organs (LSO) are present in one carphodactylid and several diplodactylid species and are reported here for the first time in geckos. All tactile sense organs are present on granules and tubercles, but never on intergranules.

• Detailed accounts of the microstructures are described for each genus separately, following the phylogenetic order given in Fig. S2. For measurements of microstructures see Tables 3 and 4.

Carphodactylidae

Carphodactylidae generally have granules, intergranules and tubercles. Knobs or hillocks, or both, appear on all granules and tubercles and can form long keels on the tubercles. The area of all scales is evenly covered with spinules, except for the areas occupied by knobs. Cutaneous sensilla appear on all granules and tubercles.

Saltuarius cornutus features knobs, which are relatively long and slender and are often inclined towards the centre of the scale (Fig. 2A,B). There are few cutaneous sensilla per scale scattered across the granules (Fig. 2B), but over 20 on the tubercles (Fig. 2C). Each sensillum has one or two short and broad bristles (Fig. 2D).

In the genus Phyllurus, spinules are relatively short and dense (Fig. 3F), with the notable exception of Phyllurus ossa (Table 3). Knobs are scattered across the surface of the granules and are broad and stout in Phyllurus amnicola (Fig. 3B) but resemble those in S. cornutus in the examined congeners (Fig. 3E). In the centre of the granules, there are short, stout, spinule‐like structures (Fig. 3B,E) which gradually coalesce to form the knobs in P. nepthys and P. ossa (Fig. 3D).

Tubercles bear several cutaneous sensilla that are just below the peak and multiple keels that begin below the peak and run down the sides of the tubercle in a radial fashion (Fig. 3A). Thicker spinule‐like structures fill the areas between the keels near the peak of the tubercle and occur on the keels themselves towards the bottom of the tubercle. The base of each tubercle has several knobs resembling keels that break into smaller sections. In P. amnicola, the lower half of the tubercle bears lenticular sense organs (Fig. 3A), whereas hillocks are found at the same region in P. nepthys and P. ossa (see Fig. 2A in Vucko et al. (2008).

In P. amnicola, multiple cutaneous sensilla are located around the edges of each granule (Fig. 3E); in contrast, P. nepthys and P. ossa have fewer sensilla scattered across the granules (Fig. 3B). Each sensillum has one bristle, which is thick, short and lightly covered by setules in P. nepthys (Fig. 3C), but notably longer and narrower in P. amnicola and P. ossa (Fig. 3F).

Carphodactylus laevis has long spinules, and the granules and intergranules are less distinct from each other than in other species in this study (Fig. 4A). Granules have hillocks, covered with spinules that are slightly longer and denser than on the remaining scale surface (Fig. 4B,D). Between the hillocks there are multiple cutaneous sensilla per scale, each bearing many lateral setules, giving them the appearance of a bottle brush (Fig. 4C). Tubercular scales occur in only two rows along the dorsal midline of this species. They have smooth, rounded peaks, whereas their lower reaches are covered with hillocks and carry cutaneous sensilla similar to those on the granules.

Figure 4.

Dorsal scales of Carphodactylus laevis (SEM images). (A) Overview showing granules (GS) and intergranules (IGS). (B) Granule scale with hillocks (HI) scattered over the scale area and cutaneous sensilla (CS) in between the hillocks. (C) Cutaneous sensillum with five bottlebrush‐shaped bristles (BR) surrounded by a deep moat (MO). (D) A single hillock covered with spinules, which are slightly longer and denser compared with those in the top left and right of the image. Three darker lines splitting the hillock are separations between cells indicating the outline of the clear layer (CL) from the previous skin generation.

All species of the genus Nephrurus examined have relatively short spinules varying in density among species. Broad and stout knobs are scattered across granules and tubercles, arising from surrounding hillocks and thus forming their tips (Figs 1D and 5C). Tubercles also have a keel running along their anterior end and bear several cutaneous sensilla at their peak, arranged in a V‐shape around the keel, with the apex directed posteriorly (Fig. 5B). Cutaneous sensilla are positioned either centrally or towards the posterior end of the granules of Nephrurus asper and N. levis (Fig. 5A), whereas they are always located centrally in N. laevissimus (Fig. 5C). Whereas the number of bristles per sensillum is comparable in all three species (Table 3), their shape is distinctively different in N. asper compared with the other species. In N. asper the shafts of the bristles have short thick setules at their broadened end and appear slightly barbed, resembling a mace (Fig. 5D). In contrast, the cutaneous sensilla of N. laevissimus and N. levis have bottle brush‐shaped bristles, bearing long and thin setules from base to tip (Fig. 1C).

Figure 5.

Dorsal scales of the genus Nephrurus (SEM images). (A) Overview of Nephrurus asper showing granules (GS) and intergranules (IGS). Knobs (K) are scattered over the granule scales, with cutaneous sensilla (CS) in between. (B) Tubercle scales of N. asper are keeled towards the anterior end of the scale and surrounded by larger granule scales. Six cutaneous sensilla (CS) occur at the peak forming a V‐shape with the apex facing posteriorly while the rest of the tubercle is covered with knobs (K). (C) A single granule scale of Nephrurus levis with one centrally located cutaneous sensillum (CS), surrounded by knobs (K). (D) Detail of a cutaneous sensillum of N. asper with seven, slightly barbed bristles (BR) which are divided into a series of knobs at the end. A shallow moat (MO) surrounds the sensillum.

Diplodactylidae

The Diplodactylidae have granules and intergranules, but lack tubercles (except for some Strophurus). Scales are evenly covered with spinules, unless otherwise noted, and always lack knobs or hillocks. Cutaneous sensilla are located at the posterior ends of the granules and are almost exclusively single‐bristled. In contrast, lenticular sense organs, when present, are located on the anterior margins in all species except Oedura, where they are scattered all over the granules.

Oedura have relatively large granules with relatively long spinules (Tables 1 and 2). Cutaneous sensilla bear one smooth bristle, which is split at its distal end and is surrounded by longer spinules (Fig. 6B). There are only a few sensilla per scale in O. coggeri, O. bella and O. cincta, but O. castelnaui and O. monilis have distinctly more cutaneous sensilla per scale (Table 3).

Figure 6.

Dorsal scales of the genus Oedura (SEM images). (A) A single granule scale of Oedura castelnaui surrounded by small, wrinkled intergranules. and many lenticular sense organs (LSO) covering the entire scale surface. Cutaneous sensilla (CS) were also observed at the posterior end of the scale. (B) A Cutaneous sensillum on a granule scale of Oedura cincta in a recession surrounded by a moat (MO) filled with small, densely packed spinules. The bristle (BR) is split at the distal end and surrounded by long spinules. (C) A lenticular sense organ of O. castelnaui with a disc‐like elevation in the centre. These LSO are primarily found at the anterior margin of the scales. (D) A crate‐like LSO of O. coggeri. These LSO are more common towards the centre of the scale and at the distal margins.

Apart from O. bella, all species have lenticular sense organs (LSO) scattered across the scale surface (Fig. 6A,C), some of which appear distinctively crater‐like (Fig. 6D). These crater‐like LSO are more numerous towards the centre of scales.

Amalosia rhombifer has granules with one cutaneous sensillum at their posterior end. The sensilla are without bristles but are covered entirely with elongated spinules (Fig. 7B). Lenticular sense organs are rare in this species (Fig. 7A).

Figure 7.

Dorsal scales of Amalosia rhombifer (A,B) and Rhynchoedura ormsbyi (D–F) (SEM images). (A) Domed‐shaped granule scales surrounded by six intergranule scales, which are wrinkled and equilateral. One granule showing a lenticular sense organ (LSO) at the anterior scale margin. (B) Cutaneous sensilla of A. rhombifer are bristleless and covered in slightly denser and longer spinules. (C) Granule scales of R. ormsbyi located away from the dorsal midline are hexagonal but rounded at the posterior end and more hexagonally defined at the anterior end, and surrounded by six intergranules. Each of these granule has one cutaneous sensillum located at the posterior end. (D) Granule scales of R. ormsbyi running along the dorsal midline are elongated and oval while intergranules are very reduced and attached at the anterior and posterior ends. One cutaneous sensillum occurs on each granule at the posterior end, which is the peak of the scale.

In Lucasium, all species examined have relatively long spinules (Fig. 8B). The number of cutaneous sensilla per scale is generally small in this genus (Table 3). The bristles taper and are smooth in L. damaeum and L. stenodactylum (Fig. 8E) but are heavily covered by setules at the distal end in L. immaculatum and L. steindachneri (Fig. 8C). In the first two species, the cutaneous sensilla are surrounded by longer spinules.

Figure 8.

Dorsal scales of the genus Lucasium (SEM images). (A) Overview of Lucasium immaculatum: several granule scales with one cutaneous sensillum (CS) at the posterior end of each scale and surrounded by intergranules. Some scales have lenticular sense organs (LSO) at their anterior margins. (B) Evenly spaced long spinules among pits and struts (L. immaculatum). (C) Cutaneous sensillum of L. immaculatum with a slight depression but with no surrounding moat. The bristle (BR) is heavily barbed with setules and the surrounding spinules are sparser than those on the rest of the scale. (D) Lenticular sense organ of Lucasium steindachneri near the anterior edge of the granule scale. (E) Cutaneous sensillum of Lucasium damaeum with a distally tapered, smooth bristle (BR) surrounded by long spinules. (F) Lenticular sense organ (LSO) of L. damaeum, which is recessed into the scale.

Lenticular sense organs are absent in L. immaculatum, but occur around the anterior edges, and occasionally towards the centre of scales of L. damaeum and L. stenodactylum (Fig. 8A,D). In L. steindachneri, they occur around the edges of the scale, including posterior to the cutaneous sensilla (Fig. 8F).

In Rynchoedura ormsbyi the granules along the dorsal midline are elongated and oval (Fig. 7D) but are hexagonal on the rest of the dorsal body surface (Fig. 7C). All scales have long spinules (Fig. 1A) and each granule has a cutaneous sensillum at its posterior end, with a smooth and tapered bristle. No lenticular sense organs occurred in this species.

All species of the genus Diplodactylus had up to seven rows of scales running along the dorsal midline that were larger than the scales on the remainder of the dorsal surface. The scales of all species were covered evenly with spinules, which is the only type of microstructure present in D. wiru (Fig. 9A). The granules of D. tesselatus exhibit a centrally located bare area (Fig. 9B), whereas the granules of the D. conspicillatus group (D. ameyi, D. conspicillatus and D. platyurus) have honeycomb structures covering the entire scale surface, with spinules within the walls of the honeycombs. These walls are consistently narrow in D. conspicillatus (Fig. 9D), whereas they are broadened at the junction points, but vanish between junctions in D. ameyi and D. platyurus. Thus, they form triangles (Fig. 9C).

Figure 9.

Dorsal scales of the genus Diplodactylus (SEM images). (A) Granule scale of Diplodactylus wiru with lenticular sense organs (LSO) surrounding the scale and two cutaneous sensilla (CS) at the posterior end. (B) A large granule scale from along the dorsal midline of Diplodactylus tesselatus showing the centrally located bare area. Cutaneous sensilla (CS) are located at the posterior end of the scale. (C) Detail of a granule scale of Diplodactylus ameyi showing modified honeycomb structures, where the walls broaden at the meeting point, but nearly dissolve in the middle between two meeting points, forming triangulate structures. (D) Detail of a granule scale of D. conspicillatus. The whole surface is covered with honeycomb structures and spinules. (E) Detail of a cutaneous sensillum of Diplodactylus platyurus. The sensillum is slightly depressed with a barbed bristle (BR) that is tapered at the distal end. (F) Cutaneous sensillum of D. tesselatus with a smooth, tapered bristle (BR) located next to a bare area. (G) A lenticular sense organ (LSO) of D. wiru located at the edge of the scale where spinules were extremely short and sparse with clear layer outlines from the previous skin generation. (H) LSO of D. tesselatus at the anterior end of a granule.

Cutaneous sensilla occur at the posterior end of the granules. The spinules on the cutaneous sensilla are longer than the surrounding spinules (Fig. 9E,F). In D. conspicillatus and D. platyurus, the bristle of the cutaneous sensilla is covered with setules and has a tapered distal end (Fig. 9E). In contrast, the cutaneous sensilla of D. wiru and D. tesselatus have bristles that are quite smooth (without setules) but are still somewhat tapered at their distal end (Fig. 9F).

Lenticular sense organs occur around the edges of the granules (Fig. 9A). In D. conspicillatus and D. platyurus these usually lie anterior to the cutaneous sensilla and are situated towards the anterior end of the scales, whereas in D. wiru they are numerous and found around the entire edge of the scale. In these species they have a concave, slightly inverted morphology (Fig. 9G). In D. tessellatus, they are uncommon and occur around the edge of the scales towards the anterior end, displaying a disc‐like elevation (Fig. 9H).

The three species of the genus Strophurus have short spinules and no lenticular sense organs. Strophurus krisalys and Strophurus williamsi have bare areas without any microstructure at the centre of the granules (Fig. 10A), but there are no bare areas in Strophurus taeniatus (Fig. 10B). The cutaneous sensilla of S. krisalys and S. taeniatus have no bristle, but only elongated spinules (Fig. 10C), whereas those of S. williamsi have one smooth bristle that is split at its distal end (Fig. 10D).

Figure 10.

Dorsal scales of the genus Strophurus (SEM images). (A) One hexagonal granule scale of Strophurus krysalis with a bare area (BA) at the peak and one cutaneous sensillum (CS) at the posterior end. (B) One granule scale of Strophurus taeniatus with two cutaneous sensilla at the posterior end of the scale. (C) Detail of a cutaneous sensillum of S. taeniatus lacking a bristle with elongated spinules surrounded by a sparse spinulate area towards the edge of the scale. (D) Cutaneous sensillum of Strophurus williamsi with one smooth bristle (BR) divided in two at the distal end. The outline of the cell boundaries can also be seen.

Statistical analyses

Supertree

Of the analyses of the four matrices, the two coding schemes produced identical results for the Camin–Sokal analyses and known congeneric and familial relationships were preserved, but the Wagner trees did not always place known congeners together and did not reproduce monophyly for the two families (Carphodactylidae and Diplodactylidae). Hence, we used the Camin–Sokal tree for subsequent analyses (Fig. S2).

Grafen's branch lengths provided the best standardisations for the PICs for most traits, except for ecological traits (habitat use and humidity of the environment), size (mass and SVL) and cutaneous sensilla diameter, granule size and the number of sensilla per scale. For these traits, Pagel's branch lengths provided a better result (Table 5).

Table 5.

The best branch length type (Best Tree) is shown for each trait, derived from the Pearson correlation between PIC and their standard error (Garland et al. 1992). For each trait and tree, the respective r and P‐values are shown as well

Trait	Best Tree	r	P‐value
Bristles per mm²	grafen	−0.0342	0.8684
CS diameter	pagel	−0.0037	0.9855
CS per mm²	grafen	−0.1460	0.4768
Humidity	pagel	0.1220	0.5528
Habitat	pagel	−0.2814	0.1638
knobbiness index	grafen	0.1603	0.4340
LSO	grafen	0.0728	0.7239
Mass	pagel	−0.0832	0.6861
Pit density	grafen	0.1515	0.4599
Pit diameter	grafen	−0.2071	0.3100
Granule size	pagel	−0.1215	0.5543
Intergranule size	grafen	0.0705	0.7321
Spinule density	grafen	0.1587	0.4387
Spinule length	grafen	0.0943	0.6468
SVL	pagel	−0.1002	0.6261
Bristles/CS	pagel	0.0939	0.6482
CS/scale	grafen	−0.1383	0.5005

CS, cutaneous sensilla; LSO, lenticular sense organs; SVL, snout‐vent length.

Discriminant function analysis

For both habitat and relative humidity of the environment, the optimal value of lambda was close to zero and the relationship between lambda and RSS was monotonically increasing (Fig. S3). Hence, we used a lambda value of zero, which reduces the Motani and Schmitz method to the ordinary FDA method of Hastie et al. (1994).

Habitat

All 27 species were classified correctly by the analysis to the habitat from which they were collected (Table 6A) and there was a perfect separation of the groups (Fig. 11); all classification performance measures were > 0.9 (Table 7). The most important positive contributors to the first discriminant function (pFDA1) were intergranule size, spinule length and the number of bristles per sensillum, whereas granule size and knobbiness index were the most important negative contributors. For pDFA 2, the most important positive contributors were granule size, pit diameter, knobbiness index and the diameter of the cutaneous sensilla (CS). The most important negative contributors were granule size and number of bristles per sensor (Fig. 11). (For the coefficient values for all traits in the analysis see Table S1.) Terrestrial species were separated from both other groups mainly along pFDA1, whereas pFDA2 separated the saxicoline species from both of the other groups (Fig. 11). Overall, terrestrial species were more strongly separated from both arboreal and saxicoline species than these two groups were from one another.

Table 6.

Classification of habitat use (A) and the humidity of the habitat (B) of the phylogenetic flexible discriminant (pFDA) analysis of skin characteristics compared with the actual habitat or humidity use of the species, respectively. Lambda = 0 for both analysis

(A) Habitat					(B) Humidity
	Terrestrial	Arboreal	Saxicoline	% correct		Hydric	Mesic	Xeric	% correct
Predicted as					Predicted as
Terrestrial	13	0	0	100%	Hydric	4	0	0	80%
Arboreal	0	10	0	100%	Mesic	1	6	2	67%
Saxicoline	0	0	4	100%	Xeric	0	3	11	85%
Total	13	10	4	100%	Total	5	9	13	78%

Figure 11.

Plotted results of the (phylogenetic) flexible discriminant analysis of the habitat with lambda = 0. Group centroids (large symbols) are shown as well as the individual species (smaller symbols labeled with (abbreviated) species names), 95% (dotted lines) and 50% (solid lines) confidence intervals. The most important contributors are shown on the axes of the respective graphs, and the coefficients for these contributors are given here in brackets. Positive contributors for pFDA1: intergranule size (IGS): 67.27; spinule length (SL): 3.41; the number of bristles per cutaneous sensillum (Br/CS): 1.08. Negative contributors: granule size (GS): −11.55, the percentage of the scale area covered by knobs or hillocks (knobbiness; KN): −3.54. Positive contributors for pFDA2: IGS: 85.50, diameter of the pits (PDM): 4.49, KN: 3.88, diameter of the cutaneous sensilla (CSD): 3.78. Negative contributors: GS: −2.68, Br/CS: −1.55. All other measured traits had coefficients smaller than 0. Abbreviations for species names: A.rho = Amalosia rhombifer; D.con. = Diplodactylus conspicillatus; D.plat = D. platyurus; D.tes = D. tesselatus; L.dam = Lucasium damaeum; L.imm = L. immaculatum; L.stein = L. steindachneri; L.steno = L. stenodactylum; N.laev = Nephrurus laevissimus; O.cas = Oedura castelnaui; O.cog = O. coggeri; O.mon = O. monilis; P. amni = Phyllurus amnicola; P.nep = P. nepthys; R.orm = Rhynchoedura ormsbyi; S.cor = Saltuarius cornutus; S.kri = Strophurus krisalys; S.tae = S. taeniatus; S.wil = S. williamsi.

Table 7.

Classification performance measures of Garczarek (2002) for both analyses (habitat and humidity)

	Habitat	Humidity
Lambda	0	0
CR	1.0000	0.7778
AC	0.9284	0.5307
AS	0.9284	0.7270
CF	0.9586	0.8416
CF vec	Terrestrial	Hydric
	0.9999	0.9213
	Saxicoline	Mesic
	0.9199	0.8454
	Arboreal	Xeric
	0.9214	0.8083

AC, accuracy; AS, ability to separate; CF, confidence; CFvec , confidence for each class; CR, correctness rate.

Terrestrial species can, thus, be separated from both of the other groups by a combination of longer spinules, more bristles per CS and, in the Carphodactylidae, smaller areas covered with knobs. Saxicoline species differ from both other groups in having pits and CS with larger diameters and, in the Carphodactylidae, larger areas covered with knobs. Arboreal species can be separated from both groups by larger granules and smaller intergranules, and have intermediate numbers of bristles per CS and areas covered with knobs (Carphodactylidae).

Relative humidity

When examining the relative humidity of the environment of origin, 78% of the species were classified correctly, with most of the misclassifications between species from mesic and xeric conditions (Table 6B). Hydric species were classified more accurately, with only one species wrongly classified as mesic and consistently had a higher performance classification coefficient [0.92 vs. 0.84 (mesic) and 0.80 (xeric)]. Overall, these performance measurements were still reasonably high (Table 7). The plot also showed high overlap between mesic and xeric groups, whereas the hydric species were well separated (Fig. 12). The most important positive factors influencing pFDA1 were intergranule size, knobbiness index, pit diameter and CS diameter, whereas granule size and bristles per CS were the strongest negative factors. For pFDA2, the most important positive factors were intergranule size, knobbiness index and bristles per CS; CS diameter and pit density were the most important negative factors. Spinule length had a weak positive influence on both axes (see Table S1). The hydric species are separated from both xeric and mesic species, mostly by pFDA1, whereas on pFDA2 only xeric and hydric species are reasonably separated, with the mesic ones falling in between and overlapping both other categories.

Figure 12.

Plotted results of the (phylogenetic) flexible discriminant analysis of the habitat with lambda = 0. Group centroids (large symbols) are shown as well as the individual species [smaller symbols labeled with (abbreviated) species names], 95% (dotted lines) and 50% (solid lines) confidence intervals. The most important contributors are shown on the axis of the respective graphs, and the coefficients for these contributors are given here in brackets. Positive contributors for pFDA1: IGS: 113.98; KN: 6.56; CSD: 2.16. Negative contributors: GS: −13.90; Br/CS: −1.11. For pFDA these are: positive: IGS: 40.81; KN: 1.88; Br/CS: 1.30. Coefficients for all other traits were smaller than 0. For abbreviations see Fig. 11.

Thus, species from a hydric environment differ from species from drier relative humidity via a combination of higher pit diameter, CS with larger diameters but fewer bristles, smaller granules and a higher density of pits. Species from mesic environments can be separated from both other groups by having smaller intergranules and slightly shorter spinules and in the Carphodactylidae they have smaller areas covered by knobs. Species from xeric environments have a lower density of pits compared with both of the other groups.

Univariate analysis of individual traits

The number of bristles and the number of cutaneous sensilla per mm² were significantly different in model averaging after controlling for the false discovery rate. There was a significant difference in the number of bristles between terrestrial and arboreal species, and terrestrial species differed significantly from arboreal and saxicoline species in terms of the number of cutaneous sensilla per mm², with both trait values being higher in terrestrial species (cf. Table 4). The model parameters for these two traits are reported in Table 8 (for non‐significant results see Data S1).

Table 8.

Model parameter for the best fitting model for morphological traits which showed significant results (P < 0.05) after adjusting for false discovery rate. For the model parameter and P‐values for all traits, see Table S1

Response variable	Model	Mode of evolution	df	Log likelihood	AICc	∆AICc	Weight	Significant differences between
Bristles per mm²	2 (habitat + SVL + Lmass)	Independent	6	−39.3	94.8	0	0.68	Terrestrial and arboreal
CS per mm²	3 (habitat)	Pagel's lambda	5	−11.884	36.6	0	1	Terrestrial and arboreal Terrestrial and saxicoline

Discussion

General results

This study describes the dorsal microstructures of carphodactylid and diplodactylid geckos for the first time for most species (but see Bauer & Russell, 1988 for N. asper and N. levis), as well as exploring associations between cutaneous microstructures and the habitat and humidity of the environment from which the geckos originated. Three hypotheses were tested: first, that terrestrial species would have longer, or more densely arranged spinules than arboreal species, which was confirmed (Fig. 12). Thus, there is an ecological association between long spinules and a terrestrial habitat in geckos. This is consistent with the idea that some functions of skin, such as self‐cleaning and bactericidal properties, would be more beneficial for terrestrial species than for arboreal or saxicoline species. Secondly, we hypothesised that species from humid habitats would have longer or more densely arranged spinules than species from drier habitats, but this was only partially supported by our analysis, in that long spinules were weakly correlated with both xeric and hydric environments (compared with mesic environments). Our final hypothesis was that terrestrial species from dry habitats would have more, or more complex, cutaneous sensilla, and this was supported in that both terrestrial species and those from dry habitats had more bristles per sensillum, whereas the sensilla tended to have smaller diameters, especially in species from dry habitats. Also, both the number of CS and bristles per mm² were significantly greater in terrestrial species after adjusting for phylogeny.

Ecological adaptation of microstructures

Although studies often discuss possible associations between microstructures and ecology (Gans & Baic, 1977; Hagey et al. 2014), they have seldom been examined analytically. Anoles (Dactyloidae) are the only reptile group for which ecological adaptations of microstructures have been studied in detail, by comparing morphology among closely related species occupying different ecological niches. Those studies, however, focused exclusively on adhesive pad morphologies in relation to different ecological niches in rainforest habitats (Macrini et al. 2003; Renous et al. 2010; Stuart et al. 2014). Similarly, the few studies of geckos discussing connections between microstructures and ecology also addressed only the evolution of adhesive pads as adaptations to different habitats (Johnson & Russell, 2009; Collins et al. 2015). In iguanid lizards, Peterson (1984) could not detect an obvious relation between spinules (occurring only in some species of this clade) and habitat, but Peterson's (1984) study was focused on the evolution of spinules and did not analyse ecological associations in detail. Thus, our study is the first to analyse ecological associations of (non‐toe pad) microstructures in detail, allowing us to address associations with habitat type (arboreal, terrestrial, saxicoline) and habitat relative humidity (xeric, mesic, hydric). Our study was limited in that only the microstructures from the dorsal mid‐body region were analysed, but this made it less likely that habitat associations would be masked by the evolutionary effects of other functional demands (i.e. feeding, movements or tail‐breakage) on cutaneous sensilla.

Interaction between habitat and relative humidity

There were associations between skin morphology (epidermal microstructures and scale size) and ecological factors (habitat type and relative humidity), although skin morphology was more strongly associated with habitat type than relative humidity (cf. Figs 11 and 12). Apparently, the occurrence of longer or more densely arranged spinules was more strongly associated with the habitat than with the humidity of the environment. Possibly, selection from habitat and humidity are not independent (Arnold, 2002). In our study, most of the species from xeric environmental conditions were terrestrial and there was only one terrestrial hydric species (C. laevis). Unfortunately for our hypothesis predicting the association between longer spinules and rainforest living (compared with species from drier habitats), any possible selection for longer spinules driven by the necessity for bacterial and fungicidal skin in rainforest environments may be overridden by even stronger selection for long spinules in dusty and (possibly) disease‐ridden terrestrial environments.

When we examined epidermal structure in relation to habitat relative humidity, we found overlap between species from mesic and xeric conditions (Fig. 12), which manifested as incorrect classifications in the analysis (Table 6B). Although open savannah woodland can range from quite mesic to quite dry, the environmental conditions (humidity, temperature) in open savannah woodland are much more similar to those of deserts than to tropical rainforests (Bureau of Meteorology, 2018). Consistent with this, some Diplodactylidae from xeric or mesic environmental conditions have changed from one relative humidity regime to the other repeatedly during their evolution (Oliver et al. 2014). In addition, geckos may avoid environmental extremes by choosing appropriate habitats (e.g. using moist burrows in drier habitats Aguilar & Cruz, 2010), which may decrease the correlation between relative humidity and values of the traits in question. The ecology of and behavior of these geckos is very poorly known and more detailed information on habitat selection and preferred conditions of humidity used by each species are required to more clearly define associations between specific features and particular relative humidity regimes.

Comparative morphology of microstructures

Cutaneous microstructures of the Carphodactylidae and Diplodactylidae are mostly consistent with those of gecko species previously examined. Spinules have been described for all geckos examined thus far (Ruibal, 1968; Peattie, 2009) and are hypothesised to be the origin of the adhesive setae present in most climbing geckos (Ruibal & Ernst, 1965; Maderson et al. 1998; Russell et al. 2015). The spinule lengths described in this study (0.28–1.12 μm) lie within, or close to, the described ranges of gecko spinule length (0.3–3 μm), although situated towards the lower end (Ruibal, 1968; Stewart & Daniel, 1975; Peterson & Bezy, 1985; Rosenberg et al. 1992; Spinner et al. 2013a). Cutaneous sensilla have also been reported for all geckos examined so far and the numbers we recorded lie within previously reported ranges (Matveyeva & Ananjeva, 1995; Russell et al. 2014).

The most remarkable observation from this study was the discovery of lenticular sense organs in the Diplodactylidae and on the tubercular scales of P. amnicola (Carphodactylidae). Lenticular sense organs have never been reported for geckos and are normally associated with non‐gekkotan squamates. Tubercle‐like sense organs, which resemble the LSO we report, occur on the heads of some snakes (Jackson, 1977; Jackson & Sharawy, 1980), and similar sense organs, termed ‘sensory pits’ by some authors, are described for the Cordylidae, Gerrosauridae (Harvey & Gutberlet, 1995), Phrynosomatidae (Sherbrooke & Nagle, 1996), Varanidae (Bucklitsch et al. 2016) and Xantusidae (Peterson & Bezy, 1985; Harvey, 1993). Lenticular sense organs are also well known for the Agamidae and some families of the Iguanidae (Ananjeva et al. 1991, 2001; Ananjeva & Matveyeva‐Dujsebayeva, 1996). For both these clades, Matveyeva & Ananjeva (1995) report that species or genera have either cutaneous sensilla or lenticular sense organs, and they suggest that when both types of sense organs are observed on the same area of a single species these are actually artefacts associated with observing skin structure in different stages in the shedding cycle of preserved specimens. The results presented here, which originate from moulds of freshly shed living specimens, demonstrate unequivocally that lenticular sense organs do co‐occur with cutaneous sensilla, at least in geckos.

Comparison between Carphodactylidae and Diplodactylidae

Overall, microstructures varied greatly between Carphodactylidae and Diplodactylidae, strongly separating the two clades morphologically, consistent with a very early evolutionary divergence, between 66 and 102 mya (Gamble et al. 2012). Cutaneous sensilla were generally single‐bristled in the Diplodactylidae, whereas they could be multi‐bristled in the Carphodactylidae, consistent with the pattern detected by Bauer & Russell (1988), who described single‐bristled sensilla for the Diplodactylidae they examined from New Zealand and New Caledonia, and multi‐bristled sensilla in the Carphodactylidae. Multiple bristles per sensillum also occur in the pygopodid genus Lialis (Spinner et al. 2013a) and are widespread but sporadic within the Gekkomorpha (Schmidt, 1912; Hiller, 1976; Ananjeva et al. 1991; Duisebayeva, 1995; Nikitina & Ananjeva, 2003; Yonis et al. 2009; Darwish, 2012; Russell et al. 2014). Lauff et al. (1993) described bristleless sensilla and sensilla with branched bristles co‐occurring, with simple, unbranched bristles on the feet of Gekko gecko. Thus, different sensilla morphologies occur even within a single species, although our findings suggest that sensilla morphology does not vary on the dorsal surface.

Additionally, cutaneous sensilla of the Carphodactylidae not only have more than one bristle, they tend to have large diameters (21–27 μm compared with 14–20 μm in the Diplodactylidae). Single‐bristled sensilla in other gekkotans are in the same size range as the sensilla of the Diplodactylidae for most species (Lauff et al. 1993; Nikitina & Ananjeva, 2003), but some larger species such as Gekko gecko (Gekkonidae) or Tarentola chazaliae (Phyllodactylidae) have sensilla diameters of 29–30 μm (Hiller, 1971). Also, the multi‐bristled sensilla of gekkonid and sphaerodactylid geckos are similar in size to those of the single‐bristled sensilla of the Diplodactylidae reported here (Hiller, 1971; Duisebayeva, 1995).

The microstructures that most clearly distinguish the Diplodactylidae and the Carphodactylidae are the absence or presence of knobs and hillocks, which occur only in the Carphodactylidae. Bauer & Russell (1988) also found knobs in Nephrurus. As knobs or hillocks are absent in the Pygopodidae (the sister taxon of the Carphodactylidae; Gamble et al. 2012), these traits can be regarded as a synapomorphy of the Carphodactylidae. Most geckos whose microstructures have been examined so far do not have knobs or hillocks (Ruibal & Ernst, 1965; Stewart & Daniel, 1975; Spinner et al. 2013a; Russell et al. 2015). Only Peterson & Bezy (1985) report hillocks (which they termed microtubercles) for the gekkonids Pachydactylus bibronii (now Chondrodactylus bibronii) and Hemidactylus brookii, and Yonis et al. (2009) describe hillocks on the dorsal head and trunk scales of Tropiocolotes tripolitanus. These traits have most likely evolved (or re‐evolved) independently in the Carphodactylidae and in these three species, as the latter are clearly nested within the Gekkonidae (Gamble et al. 2012).

Although sparse in geckos, knob‐like structures are reported repeatedly in other squamate reptiles. Some of the ‘tooth‐like structures’ described for leaf‐chameleons (Riedel et al. 2015) appear similar to the knobs in species of Saltuarius and Phyllurus examined in this study. Gans & Baic (1977) described knob‐like structures on the scales of uropeltid snakes, terming them ‘cones’, and Peterson & Williams (1981) report them on the subdigital scales of Anoles (which they termed ‘bosses’).

Function of microstructures

Apart from basic mechanosensitive abilities (Hiller, 1968; Düring & Miller, 1979), the details of the function of cutaneous sensilla are not fully understood, especially with regard to their morphological variation (Ananjeva et al. 1991; Matveyeva & Ananjeva, 1995). Results from this study suggest that at least some of the traits of sensilla (diameter and the number of bristles per sensillum) are associated with environmental humidity. This is consistent with the hypothesis suggested by Ananjeva et al. (1991) that the sensilla may detect humidity. However, more detailed analysis of the morphology of the sensilla in relation to these factors is necessary.

The adaptive significance of knobs and hillocks within the Carphodactylidae may be to increase camouflage. The leaf‐tailed geckos (Phyllurus and Saltuarius), in particular, are very cryptic species relying heavily on camouflage. Arnold (2002) showed that Lacertids with smooth scales and few microstructures were shinier than species with highly structured scales and more microstructures, because their skin reflected light directly, whereas the skin scatters light in species with raised skin topography. Also, recent studies of snakes have revealed microstructures enhancing camouflage of the Gaboon viper (Bitis rhinoceros) (Spinner et al. 2013b). Although the microstructures of this snake are far more hierarchically structured than those of the geckos we studied, knobs and hillocks found on gecko scales could still enhance their crypsis by scattering light more than uniformly flat, spinule‐covered surfaces (Arnold, 2002). This contention is consistent with a similar pattern present in the Chamaeleonidae, in which the extremely cryptic leaf‐chameleons (Palleon, Brookesia, Rieppeleon and Rhampholeon) have developed a bumpy skin surface, with knobs and hillocks (Riedel et al. 2015). If this is correct, we predict that the equally cryptic Malagasy leaf‐tailed geckos of the genus Uroplatus should also exhibit functionally similar microstructures. The tubercle scales, which were also more common in the Carphodactylidae than in the Diplodactylidae, and also appear in many other gecko species, may have a similar effect.

Conclusion

This study described the epidermal microstructures of a range of Australian geckos, and established associations between epidermal microstructures and aspects of their ecology. These associations suggest that microstructures may be adaptations to these environmental factors, although further studies are required to test this hypothesis directly. As predicted, spinule length was associated with dry and dirty environments, consistent with the proposed functions of spinules in dirt‐shedding and killing bacteria. Species from terrestrial habitats are likely exposed to more dirt and debris and potentially harmful microorganisms than are saxicoline or arboreal groups (Ungar et al. 1995; Nunn et al. 2000; McCabe et al. 2015), thus self‐cleaning and bactericidal properties associated with long spinules could be important drivers of their evolution (Watson et al. 2015b). Although only weakly supported by our analysis, the anti‐bacterial function proposed in recent studies (Watson et al. 2015a, 2016; Li et al. 2016) may also be a driver for the evolution of long spinules in rainforest species, which may be exposed to more fungi and bacteria than species from drier habitats are (Bouskill et al. 2012). Further studies should address non‐independent selection for long spinules on geckos from dry habitats and those with high relative humidity (Arnold, 2002). In addition, we found that certain features of cutaneous sensilla were associated with both the habitat and the relative humidity (diameter and number of bristles per sensillum) and their density (bristles and sensilla per mm²). Although difficult to interpret because the functions of the different morphologies of cutaneous sensilla are not known, we expect differences in skin sensory abilities required to negotiate different habitats. For example, small nocturnal animals in xeric habitats may need to detect minute wind currents or moisture differentials and this may select for an increase in sensilla number or in the number of bristles per sensillum. The many types of sensilla we detected and the differences among species in their distributions and morphologies are consistent with the hypothesis that mechanoreception alone cannot explain the high variation in sensilla morphology among geckos (Ananjeva et al. 1991; Matveyeva & Ananjeva, 1995). It is important to remember that many characters, some of which may have similar functions in different combinations, may produce similar evolutionary outcomes through convergent evolution from different evolutionary origins (Sherbrooke et al. 2007), but examining direct correlations between morphological features and habitat provides useful information on possible evolutionary associations.

Author contributions

Concept and design: J.R., M.J.V., S.K.A.R., L.S. Acquisition of data: M.J.V., S.P.B. Data analysis/interpretation: J.R., S.P.B., M.J.V., L.S. Drafting of the manuscript: J.R., M.J.V. Critical revision of the manuscript: M.J.V., S.K.A.R., S.P.B., L.S. All authors approved the article.

Supporting information

Fig. S1 Correlation between SVL and body mass.

Fig. S2 Phylogenetic tree of the gecko species used in this study.

Fig. S3 The optimal lambda values are plotted against the residual sum squares (rss) for the pFDA analysis of (A) the habitat, and (B) the humidity of the environment.

Table S1 Coefficients of the pFDA analysis.

Data S1 Overview showing all model parameters, including the non‐significant ones.