Sex hormones and the development of sexual size dimorphism: 5α-dihydrotestosterone inhibits growth in a female-larger lizard (Sceloporus undulatus)

Body size is one of the most important traits of an organism, influencing numerous physiological, life-history and ecological processes (LaBarbera, 1989; Blackburn et al., 1999; Blanckenhorn, 2005). A particularly widespread phenomenon involving body size is sexual size dimorphism (SSD), in which adults of one sex are larger than the other. At higher levels of taxonomic classification, SSD tends to be predominantly female biased (Parker, 1992; Shine, 1994; Monnet and Cherry, 2002; Cox et al., 2007) or male biased (Fairbairn and Shine, 1993; Weckerly, 1998; Cox et al., 2007). Within taxa, however, SSD often varies significantly, and it not uncommon for SSD to range from female biased to male biased among species within a single family or even a single genus (Fairbairn, 1997).

To date, the majority of studies investigating SSD have focused on the ultimate selective pressures driving males to be larger in some species whereas females are larger in others (Vitt, 1986; Parker, 1992; Schuett, 1997; Weckerly, 1998; Lewis et al., 2000; Isaac, 2005; Cox et al., 2003, 2007; Fairbairn et al., 2007; Winck and Rocha, 2012). However, an informed understanding of how and why SSD evolved in any particular species is strongly dependent upon a thorough understanding of the underlying proximate mechanisms (i.e. ontogenetic, physiological, behavioral mechanisms) influencing growth and the development of SSD (Watkins, 1996). Many such studies have involved squamate reptiles, a diverse order of vertebrates in which SSD is widespread, and closely related species can be male-larger, female-larger or monomorphic (Cox et al., 2007).

The diversity of patterns of SSD within closely related groups of species is unexpected because the growth-regulatory genome is largely, if not entirely, shared between males and females of a given species (Badyaev, 2002; Chenoweth et al., 2008; Cox et al., 2009; Mank, 2009). Despite the shared growth-regulatory genome, SSD is surprisingly evolutionarily malleable, suggesting that the development of SSD is largely a result of sex-specific, epigenomic growth-regulatory mechanisms. One squamate family of particular interest for investigating these growth-regulatory mechanisms has been Phrynosomatidae – a large family of lizards in which opposite patterns of SSD occur even among closely related species (Cox et al., 2009).

Several recent studies investigating growth in male- versus female-larger phrynosomatids have focused on testosterone as an important factor influencing sex-specific growth-regulatory mechanisms. In both male- and female-larger species, sexual divergence in growth is associated with maturational increases in male testosterone levels (Haenel and John-Alder, 2002; Cox and John-Alder, 2005, 2007). Whereas testosterone stimulates growth in a male-larger species, it inhibits growth in two female-larger congeners, including Sceloporus undulatus, the subject of the present study (Abell, 1998; Cox et al., 2005a; Cox and John-Alder, 2005). These observations form the basis of the bi-potential growth regulation hypothesis for testosterone (John-Alder et al., 2007; Cox et al., 2009). Further support for the bi-potential growth regulation hypothesis comes from experimental studies on male-larger brown anoles (Anolis sagrei; Cox et al., 2009, 2014). The bi-potential growth regulation hypothesis, however, is not universally supported by evidence in squamates. Studies on geckos suggest that estrogenic hormones, rather than androgenic hormones, underlie sex differences in growth and the development of SSD (Starostová et al., 2013; Kubička et al., 2015, 2017). While further work is certainly required to investigate the generality of the ‘bi-potential growth regulation hypothesis’, a major question begs to be answered: how can testosterone promote growth in some species, but inhibit growth in others?

One potential explanation for bi-potential growth regulation by testosterone [and 5α-dihydrotestosterone (DHT)] lies within the somatotropic axis (i.e. central endocrine growth axis), in which pituitary growth hormone stimulates the synthesis and secretion of hepatic insulin-like growth factor-1 (systemic IGF-1), which then promotes cell proliferation and growth in target tissues (Le Roith et al., 2001; Wood et al., 2005; Laviola et al., 2007). Activity of the somatotrophic axis is influenced by energy, nutrients and numerous hormones including sex steroids (Reindl and Sheridan, 2012), leading Gatford et al. (1998) to remark that ‘the somatotrophic axis may be a major pathway through which gonadal steroids act to produce sex differences in growth.’

In male-larger species, growth-stimulatory effects of testosterone are typically associated with increases in the expression of hepatic IGF-1 message and plasma levels of IGF-1 peptide (Borski et al., 1996; Larsen et al., 2004; Riley et al., 2002; Beaudreau et al., 2011), whereas estradiol (E₂) has opposite effects (Murphy and Friesen, 1988; Hanson et al., 2014). Even within a single species, functional activity of the somatotrophic axis can be increased by testosterone and decreased by E₂ (Riley et al., 2004; Norbeck and Sheridan, 2011). In other species, however, E₂ stimulates growth and may increase the expression of hepatic IGF-1 (Tzchori et al., 2004; Goetz et al., 2009; Lynn et al., 2011), suggesting evolutionary malleability of the growth-regulatory network downstream from receptor activation.

Given the prevailing understanding of how testosterone promotes and E₂ inhibits growth through the somatotrophic axis in male-larger species, the mechanism through which testosterone inhibits growth in a female-larger species could involve sexually dimorphic conversion of testosterone into E₂, followed by inhibition of the growth-regulatory axis. In this case, regulatory networks downstream from the activated receptor would have been evolutionarily conserved, and the bi-potentiality of testosterone would derive from events prior to receptor activation. Alternatively, the mechanism of receptor activation could have been evolutionarily conserved, and the bi-potentiality of testosterone would then derive from evolutionary malleability in downstream effects on the growth-regulatory axis. The present study helps to differentiate between these alternatives.

In the present study, we tested two basic hypotheses: (1) that the effects of testosterone on growth inhibition and color development are physiologically relevant responses mediated by androgen receptors; and (2) that females share the downstream growth-regulatory network with males and will be responsive to an exogenous androgenic signal. Accordingly, we predicted that: (1) DHT would mimic the growth-inhibiting and color-enhancing effects of testosterone in a female-larger lizard, (2) removal of endogenous testosterone in males via surgical castration alone would cause growth to increase and male-typical color to diminish, and (3) females would respond to DHT much the same as males. To test these predictions, we implanted yearling male and female eastern fence lizards (S. undulatus) with Silastic^® tubules containing DHT. Sceloporus undulatus is a female-larger species in which adult males exhibit vivid blue and black abdominal and gular patches with indistinct brown chevrons against a reddish-brown dorsal surface. DHT is an end metabolite of testosterone and cannot be converted to E₂ (Frye et al., 2004; Walters et al., 2008; Sartorius et al., 2014). Therefore, DHT is considered a ‘pure’ androgen. Furthermore, testosterone and DHT both bind to the same androgen receptor, although DHT with greater affinity (Fang et al., 2003), whereas E₂ binds to estrogen receptors (Fang et al., 2000, 2003; Pereira de Jésus-Tran et al., 2006) and has a low binding affinity for androgen receptors (Fang et al., 2003). Thus, the responsiveness to DHT can be interpreted as evidence of mediation by androgen receptors (Swerdloff and Wang, 1998; Walters et al., 2008). Studies on lizards have shown circulating levels of testosterone and DHT are correlated, and that both androgens rise and fall during approximately the same months of the activity season (Courty and Dufaure, 1980; Andò et al., 1992; Baird and Hews, 2007). The level of circulating DHT, however, tends to be considerably lower, being only ∼10% of that of testosterone (Courty and Dufaure, 1980; Grassman and Hess, 1992).

Animal capture and housing was approved by the New Jersey Department of Environmental Protection, Division of Fish and Wildlife (Capture Permits no. 2012043 and no. 2013110, Housing Permits no. 2012053 and no. 2013098) and the Rutgers University Animal Care and Use Committee (Protocol no. 01-019). We captured 38 (26 males and 12 females) S. undulatus (Bosc and Daudin 1801) yearlings (age ∼10 months) from three locations near the Rutgers University Pinelands Research Station in New Lisbon, Burlington County, New Jersey, USA (41°N, 74°35′W) during early June 2012 to investigate the effects of DHT on growth. In a separate experiment during June 2013, we captured 29 (19 males and 10 females) yearling lizards (age ∼10 months) to investigate the longer-term effects of castration on growth. All lizards were captured by hand-held noose or by hand. Sex was determined by the presence or absence of enlarged post-cloacal scales, and age class was determined by body size, which is absolutely diagnostic for yearlings at that time of year. Individuals were then transported back to Rutgers University, New Brunswick, NJ, USA in cloth bags where snout–vent length (SVL, to the nearest mm) and body mass (to the nearest 0.1 g) were measured. Animals were housed individually in plastic cages (59.1×43.2×45.7 cm) with bedding of absorbent litter (Kaobed, Marcal Paper Mills Inc., Elmwood Park, NJ, USA) and two bricks arranged to form both basking and shaded sites. Opaque barriers were placed between cages to prevent visual interactions. Water was always available in a shallow dish filled with aquarium gravel, and several crickets were offered each day to ensure lizards were fed to satiety. Cages were illuminated under 40 W fluorescent lighting (Chroma 50, General Electric, Fairfield, CT, USA) for a 13.5 h:10.5 h light:dark photoperiod, and a basking period of 10.5 h:14.5 h light:dark was provided by placing a 45 W incandescent bulb (Duramax, Philips, Amsterdam, the Netherlands) over the basking site.

Using the initial measurements of SVL and body mass, lizards were assigned to one of six size-matched treatment groups: (1) castrated males receiving a DHT implant (CAS+DHT males, N=8); (2) castrated males receiving a blank implant (CAS, N=7); (3) intact males receiving sham surgery and a blank implant (CON, N=7); (4) intact males receiving sham surgery and a DHT implant (CON+DHT males, N=4); (5) intact females receiving sham surgery and a DHT implant (FEM+DHT females, N=6); and (6) intact females receiving sham surgery and a blank implant (FEM, N=5).

Using the initial measurements of SVL and body mass, lizards were assigned to one of three size-matched treatment groups: (1) intact males receiving a sham surgery (N=9); (2) castrated males (N=10); and (3) intact females receiving a sham surgery (N=10).

Prior to surgery, lizards were injected intramuscularly with ketamine (100 mg kg⁻¹; Vetus Animal Health, Columbia, MO, USA) and placed on ice to undergo cold-induced surface anesthesia until they exhibited no foot-withdrawal reflex. For all castrated males, we exposed the testes through a single ventral incision, and bilaterally removed the testes after ligating the spermatic cords with 5-0 surgical silk (Ethicon, Somerville, NJ, USA). For intact males and all females, we performed sham surgeries in which we performed the same incisions, exposed the testes or ovaries, but left the gonads completely intact. In the 2012 DHT experiment, we then inserted either a DHT implant (see below) or a blank implant into the coelomic cavity. Following all surgeries, incisions were closed with polypropylene surgical suture (6-0 Prolene, Ethicon, Somerville, NJ, USA).

Tonic-release DHT implants were made from ∼4 mm pieces of Silastic^® tubing (Dow Corning, Clarkesville, TN, USA: 1.47 mm inner diameter, 1.96 mm outer diameter) with a DHT chamber of ∼1 mm in length. We sealed one end of each tubule with silicone adhesive gel (Dow Corning) and injected 3 µl of a solution of DHT (Steraloids, Inc., Newport, RI, USA) dissolved in dimethyl sulfoxide (DMSO, 50 µg DHT µl⁻¹) into the open end. The tubules were then sealed and left for a period of 72 h to allow the DMSO to diffuse out and evaporate, leaving 150 µg of crystalline DHT in each implant. Blank implants were constructed in a similar fashion except only DMSO was injected into the tubules, which, following diffusion and evaporation, left empty implants.

All animals were given a 2-week recovery time from the date of surgery. At the conclusion of the recovery periods, measurements of SVL and body mass were recorded at approximately weekly intervals over a period of 40 days and 113 days in order to calculate growth rate (growth in mm day⁻¹) for the 2012 and 2013 experiments, respectively. Feeding rate (crickets day⁻¹) and body condition were also calculated. Body condition was calculated for each lizard as the scaled mass index using the formula: M_i×(L₀/L_i)^b_SMA, where M_i and L_i are the body mass and SVL of individual i, respectively, and L₀ is the mean SVL of all lizards (Peig and Green, 2009). The scaling exponent (b_SMA) is calculated by regressing log₁₀ body mass against log₁₀ SVL for each lizard and dividing the slope from this regression by Pearson's correlation coefficient (r). The scaled mass index was used instead of the residual index (uses residuals from regression of log₁₀ body mass and log₁₀ SVL) for calculating body condition because recent analyses by Peig and Green (2009) showed the scaled mass index to be a better overall indicator of the relative size of fat and protein reserves in several vertebrate species (five small mammals, one bird, one snake). Another reason why the scaled mass index was used rather than the residual index is because measurements of body condition should be independent of sex- and age-based variation in body size in order to be an accurate assessment of individual body condition. Peig and Green (2010) compared the scaled mass index with six other conventional body condition indices (including the residual index) and found that, unlike the scaled mass index, all six conventional methods failed to account for sex- and age-based variation in body size.

Because we did not have an established DHT assay we used measurements of dorsal and ventral coloration to corroborate the efficacies of surgical and androgen treatments for the 2012 DHT experiment. Dorsal and ventral surfaces of lizards were scanned at 600 dpi using an Epson Perfection V500 digital photo scanner (Epson America Inc., Long Beach, CA, USA) prior to surgeries and again at the termination of the experiment (mean 53 days post-treatment). We used Adobe Photoshop version CS6 (Adobe Systems Inc., San Jose, CA, USA) to estimate the hue, saturation and brightness of each animal's dorsal chevrons, dorsolateral areas, gular patches and abdominal patches in the scanned images (Fig. 1). For the chevrons, we selected the most caudal dark portion of the third chevron from the head on both the left and right sides. We used these measurements as representative of overall chevron coloration. For the dorsolateral areas, we selected the area between the second and third chevrons from the head on the left and right sides of the body and used these measurements as representative of overall dorsolateral region coloration. For the gular and abdominal patches, only the areas of blue were selected. We used the Elliptical Marquee tool to capture the area of interest and then used the Histogram tool to obtain mean red, green and blue color values of all the pixels within the selected area. We used the Color Picker tool to convert these values into corresponding measurements of hue (color reflected; measured on a standard 360 deg color wheel), saturation (purity of the color; 0%=gray, 100%=fully saturated) and brightness (relative lightness of the color; 0%=black, 100%=white). All measurements of color were adequately standardized so that observer bias could not have altered the results.

To assess measurement precision, we calculated Pearson correlation coefficients for left and right measurements of ventral and dorsal coloration. Correlations between left and right color measurements were higher for ventral surfaces (0.85–0.99) than for dorsal surfaces (0.49–0.92), suggesting lower measurement precision for the dorsal surfaces. For all subsequent analyses, we used the mean values of the animals' left and right color measurements.

In order to assess the accuracy of the color measurements, we repeated the described methodology twice for each lizard. We performed analyses of variance (ANOVA) on hue, saturation and brightness values of chevrons, dorsolateral regions, gular patches and abdominal patches. We measured repeatability as the ratio of variance within individuals to total variance (within individuals+across individuals). Repeatability was extremely high (>0.95) for all measurements of color and patch size. Although this result does not measure potential variation within individuals over time, this finding does show that measurements of coloration from any single image are highly repeatable relative to the typical variation across individuals and treatment groups.

All color values were log-transformed in order to meet the assumptions of parametric analyses. Due to the absence of ovariectomized females in the 2012 DHT study, we carried out separate analyses to compare: (1) intact males with females, and (2) intact males with castrated males. We performed a repeated-measures ANOVA to test whether SVL changed over the course of the 2012 DHT experiment, regardless of treatment group. To analyze growth rate in intact lizards, we used analysis of covariance (ANCOVA) with initial SVL as a covariate and with DHT (present/absent) and sex (male/female) as main effects. To analyze growth rate only in males, we used ANCOVA with initial SVL as a covariate and with DHT and gonad (present/absent) as main effects. We used two-way ANOVA to analyze 12 post-treatment measurements of color, feeding rate and body condition, with DHT and sex as main effects in analyses of intact males and females, and DHT and gonad (presence/absence) as main effects in analyses of intact and castrated males.

For the 2013 castration study, we used ANCOVA with initial SVL as a covariate to analyze feeding rate, body condition and growth rate for the 113-day experimental period, with sex as the main effect in analyses of females, and intact males and castration as the main effect in analyses of castrated and intact males. The homogeneity of slopes assumption of all ANCOVA was met. Statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA).

Lizards of all control and treatment groups increased in SVL over the course of the experiment (F_3,28=172.54, P<0.001; Table 1), and intact males grew more slowly than females, although this sex difference in growth rate failed to attain statistical significance (F_3,18=3.16, P=0.09; Fig. 2). Body condition (F_3,17=0.04, P=0.85) and feeding rate (F_3,17=4.07, P=0.06) did not differ between intact males and females (Table 1).

Growth was inhibited by DHT in all treatment groups. Intact males and females treated with DHT grew at a significantly lower rate compared with blank-implanted controls (F_3,18=20.77, P<0.001; Fig. 2). Body condition (F_3,18=0.00, P=0.96) and feeding rate (F_3,18=2.64, P=0.12) were not affected by DHT (Table 1). Intact and castrated males treated with DHT grew at a significantly lower rate compared with blank-implanted control males (F_3,17=14.40, P=0.001; Fig. 2). Castration alone did not have a significant effect on growth rate over the experimental period of 40 days (F_3,17=0.00, P=0.97; Fig. 2). Body condition (F_3,17=1.30, P=0.27) and feeding rate (F_3,17=2.15, P=0.16) were not affected by DHT (Table 1). Castration of males had no significant effect on body condition (F_3,18=0.06, P=0.81) or feeding rate (F_3,18=0.04, P=0.84; Table 1).

After 113 days, growth rate was significantly higher in castrated males than intact males (F_2,13=4.76, P=0.048; Fig. 3). Neither body condition nor feeding rate was significantly different between castrated and intact males (body condition: F_2,13=2.53, P=0.14; feeding rate: F_2,13=0.41, P=0.54). Females grew at significantly higher rates than intact males (F_2,16=52.83, P<0.001; Fig. 3). Body condition did not differ between males and females (F_2,16=0.13, P=0.73). However, feeding rate was higher in females than in intact males (F_2,16=26.06, P<0.001).

Abdominal patches were more vivid in intact males than females (saturation: F_3,18=14.01, P=0.002; brightness: F_3,18=8.26, P=0.010; Fig. 4A). These males also had significantly more vivid gular patches (saturation: F_3,18=12.53, P=0.002; brightness: F_3,18=6.56, P=0.020; Fig. 4B) compared with females. Hue, however, did not differ between sexes (abdominal: F_3,18=0.95, P=0.34; gular: F_3,18=1.24, P=0.28).

Treatment with DHT made abdominal patches more vivid by increasing saturation and decreasing brightness, without significantly affecting hue, in intact males and females compared with blank-implanted controls (saturation: F_3,18=9.25, P=0.007; brightness: F_3,18=46.12, P<0.001; hue: F_3,18=3.99, P=0.06; Fig. 4A). Treatment with DHT made abdominal patches more vivid by increasing saturation and decreasing brightness, without significantly affecting hue, in intact and castrated males compared with blank-implanted control males (saturation: F_2,18=8.37, P=0.009; brightness: F_2,18=32.80, P<0.001; hue: F_2,18=0.53, P=0.48; Fig. 4A).

Treatment with DHT made gular patches more vivid by increasing saturation and decreasing brightness, without significantly affecting hue, in intact males and females compared with blank-implanted controls (saturation: F_3,18=17.03, P<0.001; brightness: F_3,18=7.42, P=0.014; hue: F_3,18=0.92, P=0.35; Fig. 4B). Treatment with DHT made gular patches more vivid by increasing saturation and shifting hue from blue towards green, without significantly affecting brightness, in intact and castrated males compared with blank-implanted control males (saturation: F_2,18=11.49, P=0.003; brightness: F_2,18=0.15, P=0.70; hue: F_2,18=7.92, P=0.012; Fig. 4B).

Castration did not influence abdominal and gular patch hue (abdominal: F_2,18=0.06, P=0.82; gular: F_2,18=1.34, P=0.26), saturation (abdominal: F_2,18=0.14, P=0.71; gular: F_2,18=0.02, P=0.90; Fig. 4) or brightness (abdominal: F_2,18=1.53, P=0.23; gular: F_2,18=0.27, P=0.61; Fig. 4).

All dorsal coloration data are shown in Table 2. Females had darker, more conspicuous chevrons (hue: F_3,18=6.57, P=0.020; brightness: F_3,18=5.64, P=0.029) than intact males. Chevron saturation was not significantly different between intact males and females (F_3,18=1.13, P=0.30). Dorsolateral coloration was not different between the sexes (hue: F_3,18=0.21, P=0.65; saturation: F_3,18=0.51, P=0.48; brightness: F_3,18=2.48, P=0.13).

Treatment with DHT made chevrons less conspicuous by decreasing hue, without significantly impacting saturation and brightness, in intact males and females compared with blank-implanted controls (hue: F_3,18=5.85, P=0.026; saturation: F_3,18=0.20, P=0.66; brightness: F_3,18=2.78, P=0.11). Treatment with DHT made chevrons less conspicuous by decreasing hue, without significantly impacting saturation and brightness, in intact and castrated males compared with blank-implanted control males (hue: F_2,18=4.70, P=0.044; saturation: F_2,18=0.27, P=0.61; brightness: F_2,18=0.58, P=0.46).

Dorsolateral coloration was not significantly affected by DHT in intact males and females (hue: F_3,18=0.01, P=0.93; saturation: F_3,18=0.84, P=0.37; brightness: F_3,18=2.12, P=0.16) and in intact and castrated males (hue: F_2,18=0.02, P=0.89; saturation: F_2,18=2.13, P=0.16; brightness: F_2,18=1.11, P=0.31) compared with blank-implanted controls.

Castration had no significant effect on any measurement of chevron coloration (hue: F_2,18=0.83, P=0.37; saturation: F_2,18=1.97, P=0.18; brightness: F_2,18=1.82, P=0.19). However, whereas castrated males were similar to intact males for dorsolateral hue (F_2,18=1.70, P=0.21) and saturation (F_2,18=1.89, P=0.19), intact males had brighter dorsolateral coloration (F_2,18=5.28, P=0.034).

The present study shows that DHT mimics testosterone in that it inhibits growth and promotes male-typical coloration in both males and females of the female-larger lizard species, S. undulatus. The observed reductions in growth rate induced by DHT cannot be attributed to body condition or feeding rate, which were not affected by DHT in any of the treatment groups. Furthermore, it is unlikely that these responses could have been mediated by estrogen receptors, a conclusion supported by two lines of evidence: (1) the purity of the DHT was very high (>98%; Steraloids, Inc.), and (2) DHT has very low binding affinity to estrogen receptors (relative binding affinity=0.03% for DHT versus 100% for E₂; Fang et al., 2000; Matthews et al., 2000), so it is unlikely that the effects reported here are due to DHT binding to estrogen receptors. In other words, DHT would have to be roughly 3500× higher than E₂ in order to bind to estrogen receptors. Thus, our results suggest that the inhibitory effect of testosterone on growth is mediated by androgen receptors. One caveat to the present study is that we did not directly measure circulating DHT and, therefore, we cannot independently ascertain experimental plasma DHT levels vis-à-vis physiological levels. However, Hews and colleagues (Hews et al., 1994; Hews and Moore, 1995) used similar implants of testosterone and DHT in hatchling tree lizards (Urosaurus ornatus) and reported physiologically relevant levels of both androgens.

In the present study, we demonstrate a growth-inhibiting effect of DHT, the androgenic metabolite of testosterone, in males of S. undulatus, a female-larger lizard species. Several previous studies have reported similar growth-inhibiting effects of testosterone in species of squamate reptiles with female-larger SSD (Crews et al., 1985; Hews et al., 1994; Hews and Moore, 1995; Abell, 1998; Klukowski et al., 1998; Lerner and Mason, 2001; Cox and John-Alder, 2005; Cox et al., 2005a), but few previous studies have examined the effects of DHT on growth (Hews and Moore, 1995). Interpretation of the effects of testosterone is inherently ambiguous, because testosterone may act either directly or after conversion to DHT through androgen receptors, or it may be converted into E₂ and then act through estrogen receptors. The present results suggest that androgen receptors mediate downstream effects of testosterone and thus help clarify the androgenic mechanism(s) regulating growth in female-larger lizards. The question remains, however, how does testosterone inhibit growth in some species, but stimulate growth in others?

It is plausible that within female-larger species the typical stimulatory effect of testosterone on output of the somatotropic axis (as described previously in the Introduction) is reversed. Instead, testosterone may cause the ‘atypical’ effect of growth inhibition in S. undulatus by causing a decrease in the activity of the somatotrophic axis, including a decrease in the production of IGF-1 (John-Alder et al., 2007). Results of the present study, in which non-aromatizable DHT mimics the effects of testosterone, indicate that conversion of testosterone to E₂, which then acts through estrogen receptors, is unlikely. Instead, it is more likely that after binding their ligand, androgen receptors elicit different effects on the somatotropic axis in female-larger versus male-larger species. For example, Duncan (2011) demonstrated that testosterone decreased the expression of hepatic IGF-1 message in juvenile male and female S. undulatus and that this was correlated with decreased growth rate in both sexes. Furthermore, castration of yearling males in that study led to a 3-fold increase in hepatic IGF-1. The inhibitory effect of T on IGF-1 expression is contrary to what is observed in male-larger species of other taxa (Larsen et al., 2004; Norbeck and Sheridan, 2011; Reindl and Sheridan, 2012). In a study by Vaughan et al. (1994), surgical castration of Syrian hamsters (Mesocricetus auratus), a female-larger species, elevated IGF-1 levels compared with intact control and testosterone-replaced males. While these findings in S. undulatus and M. auratus are promising, they are nonetheless preliminary and thus require further investigation.

Female S. undulatus administered DHT exhibited phenotypic responses (e.g. decreased growth rate, development of male-typical ventral coloration) that were similar to or greater in magnitude than those observed in males. This indicates that despite pronounced differences in body size and color between males and females of S. undulatus, females have retained the functional androgen-regulatory network required for male-typical growth and color development. While female lizards have functional androgen receptors and associated regulatory networks, they simply lack the necessary androgenic signal (e.g. DHT) to initiate the development of male-typical characteristics. Therefore, the evolution of androgen-mediated sexual dimorphisms in S. undulatus is likely through the linkage of trait expression with activation of the androgen receptor; females have not lost this linkage but only the activating signal.

Cox et al. (2014) reported that testosterone had similar effects on growth and coloration in male and female brown anoles (A. sagrei) and concluded that sex differences in growth and color development between males and females are likely more through sex differences in circulating androgens and less so through tissue responsiveness to androgens. Male and female green anoles (Anolis carolinensis) have similar expression of androgen receptors in regions of the brain associated with reproductive behaviors, and testosterone is important for male agonistic behaviors and female receptivity (Rosen et al., 2002; Wade, 2011). Additionally, Golinski (2013) concluded that the percentage of androgen receptors in the amygdala, ventromedial hypothalamus and medial preoptic area of the brain were not different among males and females in two species of geckos. Treatment with exogenous testosterone, however, induced male sexual behaviors and the development of hemipenes in females (Golinski et al., 2011, 2014). In S. undulatus, sex differences in growth and color development between males and females are likely through sex differences in circulating androgen levels, similar to what occurs in A. sagrei, but sexual dimorphisms may also arise in part through sex differences in androgen-receptor densities. Hews et al. (2012) and Moga et al. (2000), for example, reported that while both male and female S. undulatus had androgen receptors present in regions of the brain, males had higher receptor densities and higher circulating testosterone than their female counterparts.

We failed to detect a significant effect of castration on male growth over the short term, but demonstrated a long-term growth-promoting effect of castration on male growth. Cox et al. (2005a) reported a similar pattern with S. undulatus in a large field enclosure in which castrated and intact males both had low levels of circulating testosterone and grew at similar rates during the 45 day and 56 day experimental time periods. However, following recapture 418 days later, castrated males had grown larger in body size and were virtually indistinguishable from intact females. Taken together, the long-term effects of castration on male growth observed in the present study and in the study by Cox et al. (2005a) demonstrate that, given adequate time, castration alone can promote growth in S. undulatus.

It is unclear why an effect of castration on male growth is apparent only after an extended time. However, in the present experiments, one possibility for the delayed castration effect is that all male lizards, both intact and castrated, had low circulating levels of androgens due to being held under experimental conditions, and only after prolonged acclimation to laboratory conditions did circulating androgen levels increase in intact males. Circulating testosterone levels have been shown to be atypically low in S. undulatus removed from natural home ranges and confined in field and laboratory enclosures (Cox et al., 2005a,b). In tree lizards (U. ornatus), circulating testosterone levels decreased significantly in males when housed in individual cages in the laboratory (Moore et al., 1991). This explanation demonstrates the environmental sensitivity of sex-specific growth regulation through hormones, and raises interesting questions regarding the energetic trade-offs between androgen-based reproductive behaviors and male growth (Cox et al., 2006). Another possibility is that all males had low circulating androgen levels because they had not yet reached physiological maturity. Therefore, there would not have been a significant source of androgens to remove via castration, and only following attainment of physiological maturity would differences in circulating androgen levels and growth be observed in intact versus castrated males. In a study on juvenile Sceloporus virgatus, Cox and John-Alder (2005) found that testosterone levels of small, intact males were similar to those of castrated males and females, and that castration did not affect growth in small males, but only affected growth in the larger males.

Recently, it has been suggested that testosterone may not universally be the key mechanism behind the sexual divergence in growth between male and female lizards. Studies on female-larger (Aeluroscalabotes felinus; family Eublepharidae) and male-larger (Paroedura picta; family Gekkonidae) species suggest that estrogenic hormones, rather than androgenic hormones, may underlie sex differences in growth and the development of SSD. In A. felinus, testosterone treatment decreased growth rates in castrated males and females, thus inducing male-like growth patterns. This is consistent with the bi-potential growth regulation hypothesis, but contrary to this is that male castration had no effect on growth (Kubička et al., 2013). More convincing evidence comes from three studies involving P. picta. Kubička et al. (2015) found that whereas castrated and intact males differed in circulating androgen levels, they did not differ in growth rates. Furthermore, Starostová et al. (2013) reported that castration and testosterone replacement had no effects on growth in males, yet interestingly, testosterone and ovariectomy significantly increased growth and body size of females, preventing the typical sex differences in growth rate and body size. In a more direct examination of the effects of estrogens on growth, Kubička et al. (2017) found that ovariectomized females grew at very similar rates compared with males and ovariectomized females administered testosterone or DHT. When E₂ was replaced in ovariectomized females, growth rates resembled those of intact females. Comparison of these studies on Gekkota (e.g. Gekkonidae, Eublepharidae) with studies on Iguania (e.g. Phrynosomatidae, Dactyloidae) suggests that testosterone may be a bi-potential regulator of growth in iguanians, but perhaps not in gekkotans. Within Gekkota, estrogens (e.g. E₂), as opposed to androgens (e.g. testosterone, DHT), may be the more important growth regulator. This is an interesting possibility, considering that the evolutionary split between Gekkota and Iguania is basal and occurred ∼175 million years ago (Wiens and Lambert, 2014).

We found that ventral coloration was made more vivid by the administration of DHT and less vivid by castration (see also Hews and Moore, 1995). The effect of DHT was especially pronounced in females, which normally do not express significant amounts of ventral coloration. Furthermore, DHT increased ventral patch sizes in both males and females. These findings indicate that sexual dichromatism in males of S. undulatus is mediated via androgen receptors and that females possess functional androgen receptors and have retained sensitivity to androgens capable of initiating the development of male-typical coloration. Several previous studies have reported that testosterone stimulates color development in lizards (Kimball and Erpino, 1971; Cooper et al., 1987; Rand, 1992; Hews and Moore, 1995; Sinervo et al., 2000; Calisi and Hews, 2007; Mills et al., 2008). In Sceloporus, Cox et al. (S. undulatus, Cox et al., 2005b; S. jarrovii, Cox et al., 2008) demonstrated that castration of juvenile male lizards decreased male-typical blue ventral coloration whereas exogenous testosterone restored this coloration in castrated males and led to the development of male-typical ventral coloration in females. Only Hews and Moore (1995), however, have reported the effects of DHT on color development in phrynosomatid lizards. They found that in tree lizards (U. ornatus) DHT had a greater effect than testosterone on the development of abdominal and gular patch color in both hatchling and adult females.

The effects of testosterone and DHT on the development of male-typical blue and black ventral coloration likely involve the distribution and density of various pigment-containing cells, such as iridophores and melanophores. Iridophores contain light-reflecting platelets that give rise to blue colors based on platelet shape, size, number and orientation (Morrison et al., 1995). Melanophores are light-absorbing cells and influence the brightness of colors (Morrison et al., 1995; Kuriyama et al., 2006). Blue coloration develops when blue wavelengths are reflected by iridophores and other wavelengths are absorbed by the melanophores (Kuriyama et al., 2006). Black coloration, however, develops when iridophores are absent and all wavelengths of light are absorbed by melanophores (Kuriyama et al., 2006). Testosterone has been shown to increase dermal melanization, resulting in increased vividness of blue colors in phrynosomatid lizards (Quinn and Hews, 2003), and other studies have shown different colors to be expressed based on the distribution of iridophores and melanophores (Morrison et al., 1995; Macedonia et al., 2000; Kuriyama et al., 2006).

To our knowledge, our study is the first to provide direct experimental evidence that administration of DHT causes dorsal coloration to become more male-like in females of S. undulatus, and that surgical castration has the opposite effect in males. These results are not as resoundingly significant as those for ventral coloration, however, because left and right color measurements are less highly correlated for dorsal than for ventral measurements. Significant effects of DHT were not observed in the dorsolateral region, although castrated males had somewhat darker dorsolateral regions (more female like) than intact males. When examining the effects of testosterone on measurements of dorsal coloration in S. undulatus, Cox et al. (2005b) found a similar pattern, where castrated males had darker, more vivid chevrons and darker dorsolateral regions than intact males. Testosterone has been shown to influence dorsal coloration in several other lizard species as well (Cooper and Ferguson, 1972; Cooper et al., 1987; Cooper and Crews, 1987; Rand, 1992). The effects of testosterone and DHT on dorsal coloration may be indirect, involving activation of the sympathetic nervous system (John-Alder et al., 2009). For example, melanophore migration leading to aggregation (darkening) or dispersion (lightening) is tied to β-adrenergic and α-adrenergic receptor stimulation, respectively (Hadley and Oldman, 1969; Cooper and Greenberg, 1992).

In conclusion, results of the present study clearly show that exogenous DHT inhibits growth and stimulates male-typical color development in both males and females of the sexually dichromatic, female-larger phrynosomatid lizard, S. undulatus. These results suggest that the effects of testosterone on growth and color development are mediated through androgen receptors. Thus, the bi-potentiality of testosterone as a growth inhibitor and a growth stimulator is likely due to evolutionary changes in the molecular growth-regulatory network downstream from activated androgen receptors. However, we still lack a complete understanding on how testosterone and DHT directly affect the physiological and molecular mechanisms regulating color development and how they can have differential effects on growth depending on the direction of SSD. Because body size and color in species exhibiting SSD and sexual dichromatism may also be correlated with several other traits, such as dominance, immunocompetence and age, it is possible for future studies in this area to become incorporated into a larger ecological and evolutionary framework of behavior, physiology, morphology and life history.

We would like to thank A. Bhattacharjee for field and laboratory assistance and animal care.