Introduction

Most vertebrates reproduce sexually, but some do asexually (Vrijenhoek et al., 1989; Schön et al., 2009; Avise, 2015). Parthenogenesis is an asexual reproductive mode that usually produces clonal offspring. However, genetically distinct multiple clones are known in several species of all-female parthenogenetic vertebrates (Vrijenhoek, 1984, 1989). In this case, ecological differentiation between the clones is necessary for them to co-exist in narrow habitats such as islands because overlap of their niches intensifies interclonal competition (Vrijenhoek, 1984, 1989). Competition occurs chiefly through exploitation and/or interference processes (Case and Gilpin, 1974).

Lizards include unexpectedly high numbers of asexual (parthenogenetic) all-female species (Cole, 1984; Kearney et al., 2009). Two parthenogenetic clones of the lizard Cnemidophorus laredoensis (LAR-A and -B) and their sexually reproducing relative C. gularis, overlap extensively in three niche dimensions including daily activity period, microhabitat use, and diet breadth; therefore, competition is expected to be intensive between these two species (Paulissen, 2001). In this genus, sexually reproducing C. sexlineatus (particularly the male) is more aggressive than three parthenogenetic congeners: diploid C. tesselatus, triploid C. tesselatus, and diploid C. neomexicanus (Leuck, 1985). Aggressiveness is also different between two local populations (probably the different clones) of C. uniparens (Grassman et al., 1991). Thus, interference competition may be severe among these sympatric sexually reproducting and parthenogenetic Cnemidophorus species.

Lepidodactylus lugubris is an all-female parthenogenetic gecko widespread over the Indian and Pacific Ocean islands and continental Asia and America (Moritz et al., 1993; Radtkey et al., 1995; Ineich, 1999; Yamashiro et al., 2000; Bosch and Páez, 2017; Señaris et al., 2017). Multiple clones have been identified genetically and morphologically, and several clones coexist on several Pacific islands (Ineich, 1988; Ineich and Ota, 1992; Hanley et al., 1994; Radtkey et al., 1995, 1996; Yamashiro et al., 2000). Inter-clonal differences of parthenogenetic L. lugubris in spatial distribution may be due to preferred temperatures of each clone, and this may promote their coexistence on small islands (Bolger and Case, 1994). On the other hand, individuals of L. lugubris are known to interact behaviorally towards each other in the field and laboratory conditions (Brown et al., 1991; Bolger and Case, 1992; Petren and Case, 1996; Short and Petren, 2008). There are no significant differences in the frequency of behavioral interactions between Clones A and B individuals (Bolger and Case, 1992). However, Clone A initiates foraging faster and consumes more insects than Clone B, and the latter is more active but shyer than the former (Short and Petren, 2008).

In the Ogasawara Islands, Japan, another clone, Clone C, often coexists with Clone A, but the microdistribution patterns differ between these two clones (Murakami et al., 2015; Murakami, 2017). The habitats of Clone A bias towards artificial environments such as house walls, electric poles, and road guardrails, whereas those of Clone C are mainly in natural areas such as forests, rock crevices, grasslands, and beaches. On the Ogasawara Islands, a sexually reproducing gecko, Hemidactylus frenatus, is also distributed in some islands (Murakami et al., 2015). This gecko has a wide distribution range in tropical and subtropical oceanic regions (Weterings and Vetter, 2018) and co-exists with L. lugubris, both of which are often found in the human habitats (Bolger and Case, 1992; Moritz et al., 1993). However, the competitive displacement by H. frenatus of an asexual resident gecko has also been documented over a wide geographic area, and the mechanism of competition was partly due to the behavioral dominance of H. frenatus over smaller L. lugubris in areas near concentrated food (Petren et al., 1993).

In the present study, we observed behavioral interactions between combinations of two individuals of Clones A and C of L. lugubris, and males and females of H. frenatus, by inducing behavioral contests for food in a small laboratory enclosure. This is the first description of behavioral interactions of Clone C with other clones and species. We discussed how their interactive tendency could help to understand the microdistribution patterns of these sympatric geckos in nature.

Materials and Methods

A total of 111 females of Clone A and 92 females of Clone C of L. lugubris and 67 males and 68 females of H. frenatus were collected on Chichijima in July and August of two years, 2016 and 2017. In addition, 10 females of Clone C were collected on Mukojima in July 2017. Snout to vent length (SVL) of all captured individuals were measured to the nearest 0.1 mm, and the state of their tail (intact or regenerated tail) was recorded. Clone types of L. lugubris were identified by their dorsal color patterns (Ineich, 1999).

Fig.1.

Schematic drawing of the experimental enclosure for observation of interactions between two individuals of Clones A and C of Lepidodactylus lugubris and male and female Hemidactylus frenatus. For the methods of observation, see the text.

In the feeding experiments, we used five enclosures made with wooden boards, in which one small hole opened on one side of walls (Fig. 1). Geckos to use for the experiments were collected in the field before 7 to 15 h of video recording. They were kept together in one of the five enclosures to allow them move around within it. In darkness, two individuals selected from Clones A and C of L. lugubris and male and female H. frenatus (hereafter written as Clone A, Clone C, H-M, and H-F, respectively) were put into the empty enclosure, and the top of the enclosure was covered with a transparent acrylic board (Fig. 1). The behavior of the geckos was recorded using infrared sensitive function of the video camera (SONY FDR-AX55 or SONY DCRA-C160, Tokyo, Japan). After starting the video recording, one live small-sized mealworm (larvae in 12 to 16 mm in length of the beetle Tenebrio molitor Linnaeus, 1758) was thrown through the narrow pipe which opens at the enclosure wall. The minimum distance between the tip of the gecko's snout and the body edge of the mealworm were measured as the distance to food. We defined “contests” as feeding of a given prey animal with some aggressive behaviors between the two individuals, using the five aggressive behavioral categories by Dame and Petren (2006): arches (warping body), lunges, growls, bites, and wrestles. Therefore, we judged as no contest when one of the individuals consumed the given prey without these behavioral interactions. Individuals that fed the mealworm were judged as winners. This feeding trial using the single mealworm (a “round” hereafter) was repeated three times for the same pair, and each round continued until the mealworm was eaten by either individual. The mealworm was always eaten within 1 h. After the experiments, geckos were moved to another cage before releasing them at the capture sites in the next morning. No individuals were used twice. To prevent recapture, we changed the collecting sites every field survey. All experiments were conducted from sunset to midnight at the Ogasawara Field Research Station of Tokyo Metropolitan University on Chichijima. Geckos were maintained at 25–29°C and 50–60% in relative humidity.

All analyses were conducted using R ver. 3.5.0 (R Development Core Team, 2018). To evaluate the effects of clones and sexes of the two gecko species on the contest rate, we built a generalized linear mixed effects model (GLMM) using the function GLMER in the lme4 package (Bates et al., 2015). In this case, we are not interested in which player won in each round, but are interested in which combination of players showed aggressive behavior. In GLMM (binomial errors with logit link), occurrence of contests (0 or 1) was treated as a response variable, and “combination of taxon groups” (taxon_pair), “distance form gecko to food” (D), “absolute difference in the SVL between players” (SVL), and “state of tails of both players” (T) were used as candidates of fixed effects. Because the same pairs were observed repeatedly (three times) in the experiment, pairs of individuals were incorporated into the model as a random effect in order to prevent the problem of pseudoreplication (Hurlbert, 1984). Eleven models with different combinations of fixed effects were evaluated; only taxon_pair, only D, only SVL, only T, taxon_pair+D, taxon_pair+SVL, taxon_pair+T, taxon_pair +D+SVL, taxon_pair+D+T, taxon_pair +SVL+T, and taxon_pair+D+SVL+T. Then, model selection was made based on Akaike Information Criterion (AIC), and the model with the minimum AIC was chosen as the best model (Burnham and Anderson, 2002). Regression coefficients in the best model were tested for statistical significance using Wald test with an alpha level of 0.05.

We also built GLMM (binomial errors with logit link) to test whether the four candidates of fixed effects, “taxon of individuals” (taxon), D, SVL, and T, are significantly related to the response variable, win/lose (1 or 0). “Taxon” was determined for both winner and loser in each round. Because the same pairs were observed repeatedly, pairs of individuals were incorporated into the model as the random effect. Eleven models with different combinations of fixed effects were evaluated; only taxon, only D, only SVL, only T, taxon+D, taxon+SVL, taxon+T, taxon+D+SVL, taxo +D+T, taxon+SVL+T, and taxon+D+SVL +T. Then, model selection was made based on AIC. Because the best model included only taxon (see Results), we used this model (explanatory variables=taxon, response variables= win or lose, error distribution=binomial distribution, link function=logit) to quantify and compare the strength in conflict between them using the ‘Bradley Terry 2’ package (Firth, 2005). The Bradley-Terry model (Bradley and Terry, 1952) assumes that in a “contest” between any two “players”, say player i and player j (i, j∈{1, ..., K}), the odds that i beats j are αi/αj, where αi and αj are positive-valued parameters which might be thought as representing “fighting ability” (frequency of successful feeding in this study). Three rounds were counted as one game, and players that consume a large number of mealworms were defined as winners because Bradley-Terry model does not allow replicated data. The strength of Clone A was set to the reference value 0. The Bradley-Terry model was tested for statistical significance using Wald test, with an alpha level of 0.05.

Results

The mean SVL was 39.8 mm (n=111, SD=1.5) in Clone A, 39.7 mm (n=102, SD=1.8) in Clone C, 53.5 mm (n=67, SD=0.9) in H-M, and 54.0 mm (n=68, SD=1.4) in H-F. The proportion of individuals with intact tails was 45 of 111 Clone A (40.5%), 96 of 102 Clone C (94.1%), 35 of 67 H-M (52.2%), and 40 of 68 H-F (58.8%). Thus, most individuals of Clone C had an intact tail in the field, and the proportion of individuals with intact tails was significantly different among the four groups (χ²=69.4, df=3, P<0.0001).

A total of 522 rounds by 174 pairs were observed. Contests occurred frequently in the combinations among individuals of Clone A, H-M, and H-F, whereas infrequently in those including Clone C (Table 1). Although body size of L. lugubris was smaller than H. frenatus, the frequency that Clone A engaged in contest was similar to that of H. frenatus.

In the contest rate, the best model included only “taxon_pair” based on AIC, and “D”, “SVL” and “T” were not recognized as important factors. Among regression coefficients of “taxon_pair” in the best model, in which the contest pair of Clone A vs. Clone A was used as standard, significantly negative values were obtained in four combinations of the contest pair that include Clone C (Table 2). No significant values were detected in regression coefficients of other contest pairs.

Table 1.

Contests for food between two individuals of Clones A and C of Lepidodactylus lugubris and male and female Hemidactylus frenatus. The upper diagonal is filled by the number of observations (n) and frequency of contests (%), and the lower diagonal by the number of victories of each category. H-M: male H. frenatus, H-F, female H. frenatus.

Table 2.

The parameter estimates of regression coefficients of fixed effects in the best model of GLMM explaining the contest rates.

Clone C showed few aggressive behaviors such as arches, growls, bites, and lunges, and never exhibited wrestling, even in the interaction with the same clone (Fig. 2). Growls were the most frequent response in aggressive interactions among the combinations of Clone A, H-M, and H-F (Fig. 2). In four cases of combinations of Clone A and H-M, the tail was lost in three individuals of Clone A and one individual of H-M by biting of the opponent.

In examination of the factors affecting win/ lose rate, the best model based on AIC included only “taxon”, and “D”, “SVL”, and “T” were not recognized as important factors. Therefore, we used the Bradley-Terry model only including taxon to quantify the fighting ability as win rate. The result suggested a superiority/inferior relationship among two clones of L. lugubris and H. frenatus (H-M>Clone A>H-F>Clone C). Clone C was found to have significantly lower fighting ability than Clone A (P<0.01; Fig. 3), but no significant differences were detected between Clone A and H. frenatus.

Fig. 2.

Aggressive behaviors of players 1 and 2 when they interacted for a given prey item between two individuals of combinations of Clones A and C of Lepidodactylus lugubris and male and female Hemidactylus frenatus (H-M and H-F, respectively). Aggressive behaviors are shown in % occurrence of arches, growls, bites, lunges, and wrestles in the total number of rounds of two individuals.

Fig. 3.

Quantification of the fighting ability between two individuals of combinations of Clones A and C of Leptodactylus lugubris and male and female Hemidactylus frenatus (H-M and H-F, respectively), using the Bradley-Terry model. The strength of Clone A was set to the reference value 0. Bars indicate ±1SE. ** P<0.01.

Discussion

In behavioral contests for food between two individuals in the small enclosure, Clone C of L. lugubris had a lower contest rate (frequency of feedings with some aggressive behaviors between individuals) and a lower fighting ability (frequency of successful feeding) than conspecific Clone A and both sexes of H. frenatus. Competition for food resources between geckos is affected by two factors: exclusive tendency (aggressive behavior towards other individuals) and foraging ability (strike distance, speed, and proportion of successive strikes) (e.g., Petren and Case, 1996). However, in the present study, we cannot distinguish all these factors. We found no significant effects of the distance to the food, body size difference in SVL, and state of tails on the contest rate and fighting ability.

Growls were the most frequent response in aggressive interactions in our experiment. However, we did not discriminate growls and clicks although clicks have been reported as a nonaggressive behavior (Dame and Petren, 2006). Therefore, our results of aggressiveness represented as growls may be overestimated. In behavioral observations, it is possible that some behaviors were actually emitted in a courtship context. However, any behavioral repertoire used in courtship (Dame and Petren, 2006) was not observed by male H. frenatus toward conspecific female in our experiments.

The behavioral tendency that Clone C was more timid than Clone A in contest rate and fighting ability could help to understand the distribution patterns of these two clones in the Ogasawara Islands. Artificial habitats in urban areas consist of typically flat building walls with artificial lights that attract insects. Therefore, insects, the lizards' primary prey, are abundant in small wall areas where competition to obtain these insects is severe. In this case, more aggressive individuals can eat more insects, and therefore timid individuals would be excluded from such an artificial area via long-term strong competition for food. In the Ogasawara Islands, the distribution of Clone A is biased towards the artificial environments such as house walls, electric poles, and road guardrails, whereas individuals of Clone C were found in natural area such as forests, rock crevices, grasslands, and beaches (Murakami et al., 2015). Our results suggest that differences in behavioral interactions may be one of the important factors affecting habitat segregation between Clones A and C. For Clone C, another cosmopolitan sexually reproducing gecko Hemiactylus frenatus may have the same impact as conspecific Clone A. If Clone C co-exists with H. frenatus, the similar distribution shift would be expected as in the case of coexisting with Clone A. However, their field distribution data are unavailable at present.

On the other hand, Clone A of L. lugubris had the high contest rate and fighting ability similar to both sexes of H. frenatus in their interactions. Aggressiveness including physical contacts may result in injury. In our observations, autotomy of the tail occurred in three individuals of Clone A and one male of H. frenatus by biting. In the field, most individuals of Clone C had intact tails, whereas a half of Clone A and H. frenatus had lost a part of their tails. This phenomenon may be explained by their different contest rates and aggressiveness.

Several researchers have reported behavioral interactions of H. frenatus and clones of L. lugubris. In Hawaii and Fiji, the proportion of H. frenatus relative to L. lugubris increases gradually in man-made arid environments, but the latter is relatively more common in more mesic and cooler environments (Case et al., 1994). Moreover, H. frenatus is more aggressive than Clones A and B of L. lugubris (Bolger and Case, 1992). If light is present, to which insects are attracted, H. frenatus dominates over L. lugubris by their aggressive interactions (Petren et al., 1993). Males of H. frenatus are more aggressive towards conspecific males, but show little aggression towards conspecific females and parthenogenetic H. garnotii (Dame and Petren, 2006). Females of H. frenatus and H. garnotii display little aggression (Dame and Petren, 2006). Our results also indicated males of H. frenatus are more aggressive than females although this is not statistically significant. Thus, H. frenatus (particularly males) is superior in interactions with parthenogenetic L. lugubris and H. garnotii. However, increases in both the dispersion of food resources and microhabitat topography have been known to reduce interspecific competition for food between H. frenatus and L. lugubris (Petren and Case, 1998). Interactions between Clones A and B of L. lugubris were also affected by habitat complexity (Short and Petren, 2008).

In our study, the superiority/inferior relationship among two clones of L. lugubris and both sexes of H. frenatus was estimated as HM> Clone A>H-F>Clone C. Clone C interacted with other individuals less frequently and showed few aggressive behaviors. Our study demonstrates that the relationship among male and female H. frenatus and clones of L. lugubris is complex. In the future, we should examine the effects of regional differences in habitat structure, prey availability, and specific/clonal combinations to unravel such complexity in behavioral interactions among these geckos.