DEVELOPMENTAL SUCCESS, STABILITY, AND PLASTICITY IN CLOSELY RELATED PARTHENOGENETIC AND SEXUAL LIZARDS (HETERONOTIA, GEKKONIDAE)

DEVELOPMENTAL SUCCESS, STABILY, AND PLASTICY IN CLOSELY RELATED PARTHENOGENETIC AND SEXUAL LIZARDS (HETERONOTIA, GEKKONIDAE) Author(s) :Michael Kearney and Richard Shine Source: Evolution, 58(7):560-57. 004. Published By: The Society for the Study of Evolution DOI: http://dx.doi.org/0.554/03-559 URL: http://www.bioone.org/doi/full/0.554/03-559 BioOne (www.bioone.org) is a nonprofit, online aggregation of core research in the biological, ecological, and environmental sciences. BioOne provides a sustainable online platform for over 70 journals and books published by nonprofit societies, associations, museums, institutions, and presses. Your use of this PDF, the BioOne Web site, and all posted and associated content indicates your acceptance of BioOne s Terms of Use, available at www.bioone.org/page/terms_of_use. Usage of BioOne content is strictly limited to personal, educational, and non-commercial use. Commercial inquiries or rights and permissions requests should be directed to the individual publisher as copyright holder. BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors, nonprofit publishers, academic institutions, research libraries, and research funders in the common goal of maximizing access to critical research.

Evolution, 58(7), 004, pp. 560 57 DEVELOPMENTAL SUCCESS, STABILY, AND PLASTICY IN CLOSELY RELATED PARTHENOGENETIC AND SEXUAL LIZARDS (HETERONOTIA, GEKKONIDAE) MICHAEL KEARNEY, AND RICHARD SHINE School of Biological Sciences, University of Sydney, New South Wales 006, Australia Abstract. The developmental trajectory of an organism is influenced by the interaction between its genes and the environment in which it develops. For example, the phenotypic traits of a hatchling reptile can be influenced by the organism s genotype, by incubation temperature, and by genetically coded norms of reaction for thermally labile traits. The evolution of parthenogenesis provides a unique opportunity to explore such effects: a hybrid origin of this trait in vertebrates modifies important aspects of the genotype (e.g., heterozygosity, polyploidy) and may thus impact not only on the phenotype generally, but also on the ways in which incubation temperature affects expression of the phenotype. The scarcity of vertebrate parthenogenesis has been attributed to developmental disruptions, but previous work has rarely considered reaction norms of embryogenesis in this respect. We used closely related sexual and asexual races of the Australian gecko Heteronotia binoei, which include those with multiple origins of parthenogenesis, to explore the ways in which reproductive modes (sexual, asexual), incubation temperatures (4, 7, and 30 C), and the interaction between these factors affected hatchling phenotypes. The hatchling traits we considered included incubation period, incidence of deformities, hatchling survivorship, body size and shape, scalation (including fluctuating asymmetry), locomotor performance, and growth rate. Developmental success was slightly reduced (higher proportion of abnormal offspring) in parthenogenetic lineages although there was no major difference in hatching success. Incubation temperature affected a suite of traits including incubation period, tail length, body mass relative to egg mass, labial scale counts, running speed, growth rate, and hatchling survival. Our data also reveal an interaction between reproductive modes and thermal regimes, with the phenotypic traits of parthenogenetic lizards less sensitive to incubation temperature than was the case for their sexual relatives. Thus, the evolution of asexual reproduction in this species complex has modified both mean hatchling viability and the norms of reaction linking hatchling phenotypes to incubation temperature. Discussions on the reasons why parthenogenetic organisms are scarce in nature should take into account interactive effects such as these; future work could usefully try to tease apart the roles of parthenogenesis, its hybrid origin (and thus effects on ploidy and heterozygosity, etc.), and clonal selection in generating these divergent embryonic responses. Key words. Developmental stability, incubation, lizard, parthenogenesis, phenotypic plasticity, temperature. One enduring puzzle in evolutionary biology is the question of why parthenogenetic taxa (those that reproduce without fertilization) are far less common than sexual taxa (Williams 975; Maynard Smith 978; Bell 98). A parthenogenetic lineage, through dispensing with males, could potentially obtain a twofold reproductive advantage over a sexual lineage and has the additional benefit of avoiding the costs and risks associated with courtship and mating (Maynard Smith 978). Despite this advantage, only 0.% of recognized species are parthenogenetic (White 978). The answer to this paradox presumably lies in deleterious effects of parthenogenesis (or the mechanisms that generate parthenogenesis) on organismal fitness. The nature and magnitude of such effects remain obscure, but can potentially be studied through direct comparisons of closely related sexual and asexual taxa. Early development is an important phase of life history at which we might expect to see reproductive mode influence fitness. Studies of parthenogenesis have emphasized how the genetic consequences of the transition to parthenogenesis may lead to developmental disturbances. For example, parthenogenetic insects tend to have lower hatching success than related sexual species (Lamb and Willey 979; Corley and Moore 999). The degree of genetic heterozygosity is thought to be an important determinant of developmental success and Present address: Centre for Environmental Stress and Adaptation Research, La Trobe University, Bundoora, Victoria 3086, Australia; E-mail: m.kearney@latrobe.edu.au. 004 The Society for the Study of Evolution. All rights reserved. Received September 30, 003. Accepted March 5, 004. 560 stability. Lerner (954) was one of the first to emphasize that highly heterozygous hybrid lines of animals and plants often exhibit greater developmental homeostasis than their inbred, homozygous parental lines. This pattern has been upheld in many subsequent studies, although the link between heterozygosity and enhanced developmental stability and success remains controversial (e.g., Mitton and Grant 984; Mitton 993, 994; Markow 994; Zouros and Pogson 994; Polak 003). Parthenogenesis can occur through a variety of mechanisms (Suomalainen et al. 987) that have different implications for heterozygosity and therefore may affect development in different ways. When ploidy is restored through gamete duplication, for instance, the effects are equivalent to severe inbreeding and result in complete homozygosity. Parthenogenetic lineages of Drosophila that originate in this manner exhibit developmental problems (Kramer et al. 00a,b), leading to such reduced hatching success and survivorship that the potential twofold reproductive advantage of parthenogenesis may not be realized (e.g., Carson et al. 957; Carson 967; Kramer and Templeton 00). In contrast, truly clonal propagation occurs when meiosis or its effects are suppressed, leading to the perpetuation of the original level of heterozygosity. Parthenogenetically produced offspring of the cockroach Nauphotea cinerea that originate in this way suffer severe developmental problems despite the maintenance of parental levels of heterozygosity (Corley et al. 999; Corley and Moore 999; but see Jokela et al. 997a for an opposing result in snails). Extremely high levels of

DEVELOPMENT IN SEXUAL AND ASEXUAL GECKOS 56 heterozygosity occur when a clonal mode of parthenogenesis arises in association with hybridization. This is true of all unisexual vertebrates so far examined (Vrijenhoek et al. 989) as well as some invertebrates (e.g., White et al. 977). Many parthenogenetic taxa also are allopolyploids, which further contributes to their high heterozygosity. This raises an interesting question: Does the increase in heterozygosity associated with the evolution of hybridity and polyploidy in these parthenogenetic organisms lead to enhanced developmental stability and success (Vrijenhoek and Lerman 98; Wetherington et al. 987)? Alternatively, do incompatibilities between the combining genomes (i.e., outbreeding depression: Lynch 99) lead to reduced developmental stability and success in such organisms? In this study we examine how overall developmental success and stability are affected in reptiles that have evolved parthenogenesis through hybridization. Within reptiles, comparative data on the relative developmental success of parthenogenetic and sexual lineages are mainly limited to lacertid lizards of the genus Darevskia (formerly Lacerta, summarized in Darevsky et al. 985). Here we extend these studies to gekkonid lizards, a group with multiple origins of parthenogenesis spanning at least four genera (Vrijenhoek et al. 989). In particular, we compare development in parthenogenetic geckos of the Heteronotia binoei complex and their sexual progenitors. We also consider how the evolution of hybrid parthenogenesis in H. binoei has affected their phenotypic sensitivity to incubation temperature. Such environmental modification of genotypic expression during development is an example of phenotypic plasticity, more properly referred to as developmental plasticity (Piersma and Drent 003). Developmental plasticity has received intensive study regarding the effect of temperature on reptilian incubation (e.g., Burger 989; Janzen and Paukstis 99; Shine and Harlow 996; Shine et al. 997) but it is rarely considered in the context of hybrid parthenogenetic organisms (see Parker 984; Zakharov and Shchepotkin 995; Jokela et al. 997b; Negovetic and Jokela 00 for examples in nonhybrid systems). Our study asks three basic questions relating to development in H. binoei: () Does hatching success differ between parthenogenetic and sexual races? () Does offspring viability differ between parthenogenetic and sexual races? (3) How does the sensitivity of development to temperature differ between parthenogenetic and sexual races? To answer these questions we incubated eggs obtained from the two triploid chromosome races of parthenogenetic H. binoei as well as from the two progenitor sexual races collected from the same region. We then compared their hatching success, incubation time, proportion of deformities, hatchling survivorship, body size and shape, scalation (including fluctuating asymmetry), locomotor performance, and growth rate. We were thus able to compare overall differences in developmental success as well as thermally induced reaction norms for phenotypic traits of hatchlings among parthenogenetic and sexual lizards, to see whether the evolution of parthenogenesis has been accompanied by modifications of embryonic thermal sensitivity. MATERIALS AND METHODS Study Species Heteronotia are small (up to 30 mm total length, 5 g) oviparous gekkonid lizards that are widely distributed through mainland Australia (Cogger 000). The genetics and evolution of parthenogenesis in the H. binoei complex have been well studied (Moritz 993 and references therein). Briefly, five distinct sexual taxa as well as triploid parthenogenetic races have been identified cytologically within the H. binoei complex. The parthenogenetic races originated through hybridization events between the CA6 and SM6 sexual races (Moritz 983, 984; Moritz et al. 989a). Mitochondrial DNA (mtdna) analysis has revealed two distinct maternal lineages, the 3N lineage, which has a CA6 maternal parent, and the 3N lineage, which has an SM6 maternal parent. Subsequent back-crossing within both of these maternal lineages of the original diploid hybrids with males of both progenitor species has produced two triploid races, one having a double dosage of the CA6 genome (form A: CA6/SM6/CA6) and the other having a double dosage of the SM6 genome (forms B and C, considered together in this paper as form : CA6/ SM6/SM6). In gross morphology, these two triploid forms resemble the sexual parental form for which they have a double genetic dosage; form A having a banded back-pattern like the CA6 sexual race, and form having a speckled back-pattern like the SM6 sexual race. Parthenogenesis in H. binoei is functionally apomictic, since no recombination has been observed (Moritz 984). The two parthenogenetic forms currently inhabit some of the driest regions of the Australian arid zone where they are broadly sympatric with three of the sexual forms, including the two progenitor sexual races, CA6 and SM6. Collection and Maintenance of Animals The eggs we used in this study were obtained from captive lizards that had been collected along a 00-km latitudinal transect through central Australia in early spring of 000 (see Appendix for collecting localities). We determined the identity of all lizards (parthenogenetic vs. sexual) using a combination of mtdna sequencing and microsatellite markers (J. L. Strasburg, unpubl. data). We housed specimens in a controlled-environment room at the University of Sydney. Geckos were maintained in either male/female pairs (sexuals) or female/female pairs (parthenogens), in plastic containers ( 3 8 cm) with plastic half-pipes for shelter and a 3 cm-deep substrate of sand. Offspring from sexual females were either fertilized by the male with which she was housed or by sperm stored from a mating prior to collection. We kept room temperature at 5 C and Flexwatt (Maryvale, TN) heat strips, located under each cage, provided a maximum substrate temperature of 35 C. This created a 0 C thermal gradient within the shelters, encompassing the geckos mean preferred temperature (3 C; M. Kearney, unpubl. data). We kept humidity high by tri-weekly wetting of the substrate to a depth of cm (necessary to prevent skin-shedding difficulties). Photoperiod was matched to that of the central Australian township of Alice Springs (33 5 E, 3 4 36 S). We fed each gecko four crickets, three times per week. Cal-

56 M. KEARNEY AND R. SHINE cium (Repcal, Rep-Cal Research Labs, Los Gatos, CA) and vitamin (Herptivite, Rep-Cal) supplements were provided once a week. Water was always available. Incubation of Eggs We monitored the stage of egg development in each female weekly by directly examining the developing follicles and eggs clearly visible through the lizards translucent abdomens. We monitored individuals in late stage gestation daily until they laid their eggs, upon which we immediately transferred the eggs to small sand-filled glass jars (one egg per jar) for incubation. Heteronotia binoei has a watertight calcareous-shelled egg akin to that of birds, as do all geckos from the subfamily Gekkoninae, making it unnecessary to provide a moist substrate for incubation (Dunson 98). Clutch size in this lizard was either one or two eggs as is typical of gekkonid lizards (Vitt 986). We partially buried eggs in the sand and then covered the jars with clear plastic wrap (permeable to O and CO but not H O) and incubated the eggs at one of three constant temperatures: 4 C, 7 C, or 30 C. We chose these incubation temperatures based on our previous experience with gekkonid eggs in the laboratory. Although in retrospect these were found to emphasize the cooler spectrum of incubation temperatures in H. binoei, this is of most ecological interest because it is usually more challenging for a lizard to find sufficiently warm nests than it is to find sufficiently cool nests, and parthenogenetic Heteronotia binoei are biased to the cooler parts of the distribution of the H. binoei complex (Kearney et al. 003). We incubated all of the first clutches laid at 7 C because we did not know initially how many clutches the geckos would produce. We randomly allocated subsequent clutches to either the 4 C or the 30 C incubation. Measurements of Hatchlings We monitored incubating eggs daily and upon hatching we recorded the following data: mass ( mg), snout-vent length (SVL), tail length, number of supra- and infralabial scales on each side of the face, and presence of deformities. We scored fluctuating asymmetry (Van Valen 96) of the labial scales by calculating the absolute value of the difference in the number of scales between the right and left sides of the face separately for the supra- and infralabial scales, and for both combined. We then housed and maintained the hatchlings in the same manner as for the adult lizards. Approximately one week after hatching, we re-weighed the lizards and measured their running speed by racing the hatchlings down a runway m long and 4 cm wide. The runway was in a room held at 5 C and we equilibrated the lizards at this temperature for 0.5 h prior to placing them at one end of the runway and chasing them down it with an artist s paintbrush. Speed was measured by photocells set at 0.5 m intervals along the runway. A computer recorded the time intervals between the lizard s arrival at each photocell. We raced each lizard three times with a 0 min break in between, and we used the maximum speed recorded over a given 0.5 m interval as the lizard s burst speed. We maintained hatchling lizards in captivity until 3 October 00 to assess survivorship and growth rates. We recorded body mass at hatching, then again on February 00, and last on May 00. At the time of the second measurement, lizards were on average 43 days old (range 7) and at the time of the third measurement lizards were on average 4 days old (range 9 5). We calculated growth rates per individual as the change in mass (mg) divided by the length of time elapsed (days). Statistical Analyses We analyzed frequency data using log-linear models, with the various levels of interaction terms considered in a hierarchical fashion (Sokal and Rohlf 995). We initially fitted a model that included all main effects and interaction terms. We then tested all terms by determining the change in loglikelihood as they were dropped from the model, beginning with the term of highest order and proceeding in a hierarchical fashion through the lower interaction terms. We included all terms of equal or lower order to the term being tested in the full and reduced models. We analyzed continuous data with ANOVA or ANCOVA, after checking that they met the assumptions of these tests; transformations or nonparametric tests were employed where necessary. We also performed a MANOVA on nine traits to determine the overall effect of reproductive mode and incubation temperature, as well as the interaction between these factors, on the phenotypes of the offspring. As mentioned above, form A parthenogens have a double dosage of the CA6 genome and resemble this sexual parent in morphology through having a banded back pattern. The same is true for form parthenogens with respect to the SM6 sexual race (i.e., they share a speckled back pattern). Such genome dosage effects extend to other phenotypic features (M. Kearney, unpubl. data) and have been reported in other unisexual vertebrates (Moore 984). Thus in most analyses we included back pattern (banded or speckled) as a factor to account for this source of variation. We did not include back pattern as a factor in the MANOVA since it rendered sample sizes too low for some traits. We did not include clutch as a factor in our analyses, because clutch size in H. binoei is small ( or eggs) and within a single clone, relatedness between full siblings may be the same as that with other members of the same clone. Thus, we considered each egg as an independent unit. RESULTS Hatching Success We incubated a total of 30 eggs from the two chromosome races of parthenogenetic H. binoei (forms A and ) and their two sexual progenitors (CA6 and SM6) at the three different temperatures (Table ). Hierarchical log-linear analysis of frequencies, with reproductive mode, back pattern, incubation temperature, and hatching success as categories revealed a significant four-way interaction ( 6.47, df, P 0.04). Visual inspection of the data indicated an interaction between back pattern and hatching success for the sexual races but not the parthenogens (Fig. ). We thus separately analyzed data for the parthenogenetic and sexual races, with back pattern, incubation temperature, and hatching success as factors. This revealed no significant two- or three-

DEVELOPMENT IN SEXUAL AND ASEXUAL GECKOS 563 TABLE. Number of eggs incubated and the number of eggs that hatched at three different incubation temperatures for two chromosome races of parthenogenetic Heteronotia binoei (A and ) and their parental sexual races (CA6 and SM6). Number of eggs (number hatched) Incubation 4 C 7 C 30 C Total A (banded) (speckled) CA6 (banded) SM6 (speckled) Total 4 (9) 49 (5) 7 (6) 8 (60) 5 (3) 54 (3) 9 (0) 08 (64) 6 (5) 5 () 9 (4) 30 () 0 (3) 5 (3) 9 (5) 54 (3) 83 (40) 43 (8) 84 (55) 30 (76) way interactions between factors for parthenogenetic races but there was a significant three-way interaction for the sexual races (Table ). Separate analyses of the two sexual forms revealed a significant interaction between hatching success and incubation temperature for the SM6 race ( 7.8, df, P 0.03) but not the CA6 race ( 3.87, df, P 0.5). In the former case, hatching success increased as incubation temperature increased (Fig. ). Because the major comparison in this study is between parthenogenetic and sexual forms, we then pooled the two chromosome races of parthenogen (since there were no differences between them) and compared them separately to the two sexual races. For the comparison with the CA6 sexual race, hatching success was independent of reproductive mode or incubation temperature, although the former result was marginal (Table 3; Fig. ). For the comparison with the SM6 sexual race, hatching success was independent of reproductive mode but was dependent on incubation temperature, with more eggs hatching at higher temperatures (Table 3; Fig.). Once all the lizards had hatched, we dissected unhatched eggs and scored whether there was a developed embryo present inside or whether development had failed altogether. Contingency-table analysis of frequencies indicated that unhatched eggs incubated at 4 C were more likely to contain a developed embryo (6.% of eggs) than those incubated at FIG.. The proportion of eggs that hatched of the two parthenogenetic chromosome races (A and ) and their sexual progenitor species (CA6 and SM6) for each incubation treatment. 7 C (9.3% of eggs) or 30 C (6.9% of eggs) ( 7.43, df, P 0.04). Incubation Period Three-way ANOVA with length of incubation as the dependent variable and with reproductive mode, back pattern and incubation temperature as factors, revealed significant main effects of all three factors (Table 4). Incubation temperature had a very strong effect, with 4 C incubation taking.7 times longer than 7 C incubation and.5 times longer than 30 C incubation (Fig. ). Reproductive mode and back pattern also affected incubation period, with parthenogens TABLE. Results of hierarchical log-linear analyses of frequencies testing independence of back-pattern (), incubation temperature (), and hatching success (HS) separately for parthenogens and sexuals. Values in bold are significant. Term excluded Parthenogens HS HS HS HS Sexuals HS HS HS HS Full model 3.78 3.78 3.78 8.349 8.349 8.349 8.38 7.85 7.85 7.85 6.90 6.90 6.90.465 Log likelihood Reduced model df P 33.004 40.539 33.855 30.69 8.960 9.859 8.349 3.9 33.8 30.9 6.47 7.0 7.038 6.90 0.44 5.5.4 3.84. 3.0 0.06 6.95 0.63 4.8 0.36.45.50 9.65 0.506 0.00 0.43 0.47 0.69 0. 0.969 0.008 0.005 0.08 0.835 0.9 0.473 0.008

564 M. KEARNEY AND R. SHINE TABLE 3. Results of hierarchical log-linear analyses of frequencies testing independence of reproductive mode (), incubation temperature (), and hatching success (HS) for separate comparisons of parthenogens (forms A and pooled) with their two sexual progenitors (CA6 and SM6). Values in bold are significant. Term excluded Parthenogens vs. CA6 HS HS HS HS Parthenogens vs. SM6 HS HS HS HS Full model 3.67 3.67 3.67 8.945 8.945 8.945 6.044 35.60 35.60 35.60 9.86 9.86 9.86 7.954 Log likelihood Reduced model df P 6.584 40.786 33.89 9.484 30.094 9.684 8.945 9.046 44.475 36.870 3.33 9.86 33.99 9.86 69.9 8.3 4.53.08.30.48 5.80 3.57 8.43 3. 3.69 0.00 8.03.66 0.000 0.000 0.033 0.583 0.30 0.447 0.055 0.000 0.000 0.073 0.58 0.986 0.08 0.64 developing faster than sexuals, and banded lizards developing faster than speckled lizards. However, there was also a significant interaction between reproductive mode and incubation temperature (Table 4). Separate two-way ANOVAs (reproductive mode and back pattern as factors) for each incubation temperature revealed that parthenogenetic lizards had briefer incubation periods than sexual lizards only at 4 C (F,36 5.8, P 0.08; all other P 0.05; Fig. ). Incidence of Abnormalities Deformities noted in hatchling lizards included abnormal pupils, damaged legs, double labial scales, kinked tails, absence of a tail, and the presence of residual yolk. Some individuals had more than one of these abnormalities. Two cases of conjoined twins were also obtained from parthenogenetic individuals from a separate study but incubated at the same time. Sample sizes were too small to test all interactions among reproductive mode, back pattern, incubation temperature, and the presence of abnormalities. Pooling over incubation temperature, hierarchical log-linear analysis of frequencies with reproductive mode, back pattern, and the presence of abnormalities as categories revealed that parthenogens were significantly more likely to be born with abnormalities (% vs. 6%; Table 5). Many of these abnormalities (40%) took the form of residual yolk in the shell and were thus not technically a deformity of the hatchling. When these cases were excluded, the interaction between the presence of abnormalities and reproductive mode was considerably weaker ( 3.7, df, P 0.054). Within the parthenogens, there was a weak tendency for the proportion of abnormal offspring (including those with residual yolk) to decrease as incubation temperature increased (proportion deformed: 33.3% at 4 C, 0.3% at 7 C,.% at 30 C) although this pattern was not statistically significant ( 4.90, df, TABLE 4. Results of three-way ANOVA of the effects of reproductive mode (), back-pattern (), and incubation temperature () on incubation period in Heteronotia binoei. Significant values are bold. Source df MS F-ratio P Error 64 53.376 5.507 60996.94 3.00 3.460 37.547 7.3 3.86 7.7 4.64 856.07 0.09 6.800.43 0.7 0.006 0.033 0.000 0.76 0.00 0.3 0.805 FIG.. The incubation period of the two parthenogenetic chromosome races (A and ) and their sexual progenitor species (CA6 and SM6) for each incubation treatment. Values at the same incubation treatment are joined by horizontal lines. Error bars represent standard errors.

DEVELOPMENT IN SEXUAL AND ASEXUAL GECKOS 565 TABLE 5. Results of hierarchical log-linear analyses of frequencies testing independence of reproductive mode (), back pattern (), and the presence of abnormalities (AB). Significant values are bold. Term excluded AB AB AB AB Full model.698.698.698 7.74 7.74 7.74 7.46 Log likelihood Reduced model df P 35.07.535 65.0 7.88 7.846.338 7.74 6.64.67 86.65 0.8 0. 7.9 0.63 0.000 0.96 0.000 0.595 0.647 0.007 0.47 P 0.086). Only three sexual hatchlings showed abnormalities at hatching and these were all incubated at 7 C. Body Size and Shape Three-way ANOVA of the snout-vent length (SVL) of hatchling lizards, with reproductive mode, back pattern, and incubation temperature as factors, demonstrated that parthenogenetic lizards had greater mean SVLs than sexual lizards (F,64 6.780, P 0.00) and that speckled lizards had greater mean SVLs than banded lizards (F,64 3.00, P 0.00). There was also a marginal effect of incubation temperature with warmer incubation producing hatchlings with greater mean SVLs (F,64.63, P 0.076; all interactions P 0.05; Fig. 3a). The same analysis but with body mass as a dependent variable indicated significant effects only for back pattern, with speckled lizards hatching with a greater mass than banded lizards (reproductive mode, F,64.47, P 0.34; back pattern, F,64 0.6, P 0.000; incubation temperature, F,64 0.434, P 0.649; all interactions P 0.05; Fig. 3b). The same result was obtained when this analysis of body mass was repeated with SVL as a covariate (results not shown). However, when egg mass was used as a covariate, the effect of back pattern disappeared and there was a significant effect of reproductive mode and a marginal effect of incubation temperature (reproductive mode, F,64 7.598, P 0.007; back pattern, F,64.598, P 0.09; incubation temperature, F,64.990, P 0.053; egg mass, F,64 76.576, P 0.000; all interactions P 0.05). In particular, parthenogenetic hatchlings had significantly lower body mass (corrected for egg mass) than sexuals and there was also a weak tendency for hatchlings incubated at 4 C to have lower body mass (corrected for egg mass) than those incubated at 7 C and 30 C. Three-way ANOVA of the tail length of hatchling lizards, with reproductive mode, back pattern, and incubation temperature as factors, showed similar patterns to the analysis of SVL; parthenogenetic lizards had greater tail lengths than sexual lizards (F,6 7.98, P 0.006) and speckled lizards had greater tail lengths than banded lizards (F,6 0.559, P 0.00; Fig. 3c). However, in this case the effect of incubation temperature was much stronger (F,6 5.050, P 0.000; all interactions P 0.05), with tail length increasing at higher incubation temperatures. When this analysis was repeated with SVL as a covariate (P 0.000), the incubation effect remained (P 0.000) whereas the effects of reproductive mode and back pattern lost significance (reproductive mode, P 0.08; back pattern, P 0.096). Scalation and Fluctuating Asymmetry Because data on scalation as well as fluctuating asymmetry (FA) of scalation violated the assumptions of parametric AN- OVA, we used the equivalent nonparametric Kruskal-Wallis test incorporating the extension developed by Scheirer et al. (976; as described in Sokal and Rohlf 995, p. 445). We analyzed the total number of labial scales, as well as the total number of supra- and infralabial scales, with a three-way nonparametric ANOVA, with reproductive mode, back pattern, and incubation temperature as factors (Table 6). All variables were significantly affected by incubation temperature and back pattern: 4 C incubation caused higher labial scale counts, and banded lizards had more labial scales than did speckled lizards. There was a significant effect of reproductive mode for supralabial scales; however in this case, and also for total labial scales, there was an interaction between reproductive mode and incubation temperature whereby the incubation effect was weaker for parthenogens than for sexuals (Figs. 3d e). This interaction was also apparent, but nonsignificant (P 0.06), for infralabial scales (Fig. 3f). Three-way nonparametric ANOVA of fluctuating asymmetry (FA) of labial scales overall (i.e., without differentiating between supra- and infralabial scales), and of supraand infralabial scales, with reproductive mode, back pattern, and incubation temperature as factors, revealed no significant main effects or interactions. Separate Kruskal-Wallis tests for parthenogenetic and sexual lizards (pooling over back pattern) testing the effect of incubation temperature on overall FA indicated a significant effect for sexuals (H 6.964, df, P 0.03) but not for parthenogens (H.856, df, P 0.40). When these analyses were repeated for supra- and infralabial scales separately there were no significant effects of incubation temperature (all P 0.05) although this was marginal for sexuals with respect to infralabial scales (H 5.883, df, P 0.075). Locomotor Performance We also assessed the viability of hatchlings by measuring their burst (running) speeds. For these analyses we ln-transformed data on burst speed and body mass to conform to statistical assumptions. Because the sample size for sexuals

566 M. KEARNEY AND R. SHINE FIG. 3. Body sizes (a c), labial scale counts (d f), and labial scale fluctuating asymmetry (g i) of the two parthenogenetic chromosome races (A and ) and their sexual progenitor species (CA6 and SM6) for each incubation treatment. Error bars represent standard errors. incubated at 4 C was small we initially pooled over back pattern and analyzed running speed with only reproductive mode and incubation temperature as factors using ANCOVA, with body mass at birth as the covariate. This revealed body mass as a significant covariate (F,56 9.79, P 0.00, with larger lizards running faster) as well as a significant effect of incubation temperature (F,56 6., P 0.000) and an interaction between reproductive mode and incubation temperature (F,56 6.86, P 0.00). Separate comparisons of parthenogens to sexuals at each incubation temperature revealed higher burst speeds in sexuals at 7 C and 30 C (P 0.05) but no significant difference at 4 C (P 0.; Fig. 4). Comparisons of incubation temperature effects separately for parthenogens and sexuals revealed that in both cases incubation temperature had a significant effect (P 0.05) with increases in burst speed at each increment of incubation temperature for the parthenogens (post-hoc least significant difference comparisons all P 0.05). Sexual lizards incubated at 7 C and 30 C ran at similar speeds to each other (P 0.699) but were significantly faster than those incubated at 4 C(P 0.05; Fig. 4). Finally, we excluded lizards incubated at 4 C and analyzed the effect of reproductive mode, back pattern, and incubation temperature (7 C vs. 30 C) on burst speed with body mass as a covariate. This analysis revealed body mass as a significant covariate, a significant effect of reproductive mode, and interactions between reproductive mode and back pattern and between reproductive mode and incubation temperature (Table 7). The former interaction was such that individuals of the SM6 sexual race ran significantly faster than individuals of both the A and parthenogenetic races and the CA6 sexual races, whereas the latter three groups did not differ from each other (Fig. 4). The second interaction was such that the difference between parthenogens and sexuals was greater at 7 C than at 30 C (Fig. 4). Survival Rates of Hatchlings We analyzed the proportion of hatchlings that survived until October 00 when the study was ended. Hierarchical log-linear analysis of frequencies with reproductive mode and

DEVELOPMENT IN SEXUAL AND ASEXUAL GECKOS 567 TABLE 6. Results of three-way nonparametric ANOVA (Kruskal-Wallis with Scheirer-Ray-Hare extension) of mean labial scale counts, and of labial scale fluctuating asymmetry, with reproductive mode (), back pattern (), and incubation temperature () as factors. Analyses are reported for labial scales overall, and for supra- and infralabial scales. Significant values are bold. Mean scale counts Fluctuating asymmetry Factor df SS H P SS H P Total labials Error Total MS Supralabials Error Total MS Infra-labials Error Total MS 0 0 0 33.086 47975.634 8554.08 645.483 343.644 08.573 74.4 57008.963 370.35 9.408 6.94 6.833 3.99 6.467 0.55 3.05 30.388.70 0.53 0.333.646.49 9.4 0.74 3.004 69.680.63 0.036.895.988 3.30 6.48 0.74.946 5.53 9.580 36.354.304 9.68 0.303.794 0.3.457 3.875.56 5.584 0.68.840 0.850 0.000 0.000 0.077 0.044 0.87 0.378 0.09 0.00 0.000 0.9 0.008 0.860 0.408 0.57 0.000 0.00 0.7 0.06 0.90 0.398 8.3 845.60 993.85 9.569 065.488 0635.08 9388.079 6356.49 348.57 0.38 0.775 0.497.059 0.708 0.53 0.064 7.68 0.38 0.64 0.47.567 0.057.44.474 0.38 77.36 0.4 0.08 3.747 3.790 0.07 3.07 3.0.738 0.86.033.304.780.859.374 0.69 0.398.4 3.80 0.39 3.498 3.575 0.77 0.775 0.053 0.50 0.870 0. 0. 0.54 0.353 0.54 0.5 0.095 0.395 0.503 0.99 0.58 0.85 0.49 0.709 0.74 0.67 0.680 incubation temperature as categories (back pattern was not included in this analysis due to small sample sizes) revealed a significant interaction between incubation temperature and the proportion of lizards that had died, but no interactions associated with reproductive mode (Table 8). Specifically, a greater proportion of animals incubated at 4 C had died by the end of the study (57%) than those incubated at 7 C TABLE 7. Results of three-way ANCOVA of the effects of reproductive mode (), back pattern (), and incubation temperature () on burst (running) speed in Heteronotia binoei, with body mass at hatching included as a covariate. Significant values are bold. In this analysis, lizards incubated at 4 C were excluded due to small sample sizes. Burst speed and body mass were transformed with the natural logarithm. Source df MS F-ratio P Mass Error 4 0.87 0.30 0.0.530 0.576 0.07 0.003.86 3.86 6.567.70 0.087.54 4.336 0.53 0.00 8.936 0.0 0.34 0.769 0.00 0.039 0.467 0.889 0.003 (8.%) or 30 C (.7%). Those hatchlings that had died by the end of the study did so at a significantly younger age if incubated at 4 C (ANOVA: incubation temperature, F,39 6.645, P 0.003; reproductive mode, F,39 0.590, P 0.346; interaction, F,39.09, P 0.346; mean ages at death: 45.9 days at 4 C, 48.3 days at 7 C, 36.9 days at 30 C). Post-Hatching Growth Rates We analyzed growth rates of hatchling lizards using AN- COVA with reproductive mode, back pattern, and incubation temperature as factors and body mass at hatching as a covariate. This was done for the two periods at which measurements were made (Table 9). Only lizards incubated at 7 C and 30 C were included in this analysis because the low viability of hatchlings incubated at 4 C rendered sample sizes too low. Growth rates over the initial period (between hatching and approximately one to two months of age) were independent of all factors except body mass, with lizards that were larger at hatching growing faster than smaller lizards (Table 9; Fig. 5a). However, over the second period (between approximately one to two months of age and three to five months of age) there was a significant effect of reproductive mode and of incubation temperature, with sexual lizards

568 M. KEARNEY AND R. SHINE TABLE 8. Results of hierarchical log-linear analyses of frequencies testing independence of reproductive mode (), back pattern (), and the proportion of hatchlings that had died (PD) by the end of the experiment. Significant values are bold. Term excluded PD PD PD PD Full model 68.88 68.88 68.88 54.37 54.37 54.37 5.536 Log likelihood Reduced model df P 84.740 76.799 9.60 55.693 54.48 67.63 54.37 3.7 5.84 45.48.9 0.38 6.75 3.40 0.000 0.00 0.000 0.33 0.536 0.000 0.83 growing faster than parthenogenetic lizards, and lizards incubated at 30 C growing faster than those incubated at 7 C (Table 9; Fig. 5b). There was also a weak (nonsignificant) interaction between back pattern and incubation temperature whereby no effect of incubation temperature was apparent for the SM6 sexual race (Fig. 5b). Multivariate Analyses We analyzed data for nine traits with a MANOVA to identify any overall effects of reproductive mode and incubation temperature, or interactions between these factors. The nine traits were: incubation period, SVL, tail length, mass, total number of supra- and infralabial scales, fluctuating asymmetry of supra- and infralabial scales, and (log) burst speed. The MANOVA revealed significant effects of reproductive mode (Pillai Trace 0.84, F 9,44 3.545, P 0.00) and incubation temperature (Pillai Trace.039, F 8,90 3.545, P 0.000). There was also a significant interaction TABLE 9. Results of three-way ANCOVA of the effects of reproductive mode (), back pattern (), and incubation temperature (), with initial body mass as a covariate, on growth rates in Heteronotia binoei. Significant values are bold. In this analysis, lizards incubated at 4 C were excluded due to small sample sizes. Growth rates were calculated over two periods: period from age 0 to 45 days and period from 45 days to 4 days. Source df MS F-ratio P Period Mass Error Period Mass Error 83 83 0.00 0.7.064 0.30 0.65 0.3 0.0 8.6 0.834 4.65 0.007.93 0.0.35.860 0.3 3.364 0.75 0.000 0.7.476 0.363 0.78 0.373 0.05 0.33 5.883 0.00 4.09 0.66.88 3.945 0.48 4.640 0.987 0.603 0.9 0.549 0.379 0.543 0.874 0.00 0.08 0.9 0.048 0.685 0.80 0.050 0.55 0.034 between these factors: that is, the effect of incubation temperature on hatchling phenotypes differed between sexual and parthenogenetic lizards (Pillai Trace 0.34, F 8,90.38, P 0.005). When the analysis was run separately for each reproductive mode there were significant effects of incubation temperature in both cases (P 0.000). Graphical inspection of the data indicated that this effect was less pronounced in the parthenogens than in the sexuals, particularly with respect to labial scale counts and burst speed (Figs. 3d f and 4). DISCUSSION Our main purpose in this study was to determine whether the transition from sex to parthenogenesis in Heteronotia binoei has been accompanied by changes in developmental success, stability, and plasticity. Because parthenogenetic H. binoei originated via hybridization, developmental success and stability might thereby have been reduced through outbreeding depression, or increased through elevated heterozygosity FIG. 4. Burst (maximum) running speeds of the two parthenogenetic chromosome races (A and ) and their sexual progenitor species (CA6 and SM6) for each incubation treatment. Error bars represent standard errors.

DEVELOPMENT IN SEXUAL AND ASEXUAL GECKOS 569 detail in a subsequent study (M. Kearney and R. Shine, unpubl. data). FIG. 5. Growth rates of the two parthenogenetic chromosome races (A and ) and their sexual progenitor species (CA6 and SM6) for each incubation treatment. Growth rates were calculated over two periods, (a) from age 0 to 45 days and (b) from 45 days to 4 days. Error bars represent standard errors. (Vrijenhoek and Lerman 98; Moore 984; Wetherington et al. 987). We compared the two major chromosome races of parthenogenetic H. binoei with their sexual parental forms, for a variety of traits across three incubation temperatures. Our discussion is based around the three questions we posed in the introduction: () Does hatching success differ between parthenogenetic and sexual races? () Does offspring viability differ between parthenogenetic and sexual races? (3) How does the sensitivity of development to temperature differ between parthenogenetic and sexual races? Our data also reveal genome dosage effects in the two triploid races of parthenogen. This phenomenon will be considered in greater Do hatching success and viability differ between parthenogenetic and sexual H. binoei? Overall, we found no statistically significant differences in hatching success between parthenogenetic and sexual H. binoei. However, there was a marginally significant trend for parthenogens to have lower hatching success than the CA6 sexual race (Table ; Fig. ), and incubation-specific hatching success reached higher peak values for the sexual races (over 80%) than for the parthenogenetic races ( 70%). We also observed more deformities in the parthenogenetic forms, the most spectacular being two cases of conjoined twins. Although the latter were not derived from animals used in the main experiment, they were both from parthenogenetic females that had been maintained at the same time and in the same manner, and were incubated at 7 C in the same incubators. In both cases, the twins were born alive, fully formed and joined at the chest. They survived only a few days. Most of the deformities we noted, however, were relatively minor, the great majority being the presence of residual yolk either in the shell or attached to the umbilical slit of the hatchlings. Indeed, when we excluded these cases from the analyses there were no significant differences between parthenogens and sexuals in the proportion of deformities (although the P value was close to 0.05). Adherence of dried residual yolk to the posterior half of the hatchlings also may have been responsible for some other deformities, including kinked tails and damaged hind-limbs. Incomplete yolk absorption may also have caused the reduced hatchling mass (relative to egg mass) of the parthenogenetic lizards. Incomplete absorption of yolk may relate to hydric stress, because similar effects occur in flexible-shelled reptile eggs incubated on dry substrates (Shine and Brown 00). Previous studies of the lacertid lizard genus Darevskia (formerly Lacerta) have shown increased frequencies of deformed lizards (which were called monsters ) in parthenogenetic lineages compared with related sexual forms (summarized in English in Darevsky et al. 985). These included such serious deformities as conjoined twins, as in the present study. Surprisingly, most of the monsters of parthenogenetic Darevskia were males. Of the 86 parthenogenetic Heteronotia we have hatched in the laboratory thus far, our conjoined twins were the only such monsters we observed (we did not determine the sex of these individuals). This proportion (about %) is much lower than that reported for parthenogenetic Darevskia (5 35%) and similar to that found for sexual Darevskia ( 6%). We also compared levels of fluctuating asymmetry (FA) among sexual and parthenogenetic lineages of H. binoei. Fluctuating asymmetry is a popular but controversial measure of developmental instability (e.g., Polak 003 and references therein). It is a measure of the deviation from symmetry of a bilaterally symmetrical trait and reflects the ability of a given genotype to produce the same target phenotype repeatedly, on opposite sides of the body, under specified environmental conditions (Van Valen 96). Fluctuating asymmetry can be affected by both genetic and environmental

570 M. KEARNEY AND R. SHINE stress, an example of the former being the degree of heterozygosity (e.g., Vrijenhoek and Lerman 98; Zakharov and Shchepotkin 995). Our measures of FA were derived from labial scale counts, which have previously been shown in geckos to exhibit FA in proportion to levels of stress caused by habitat fragmentation (Sarre 996), presumably via inbreeding and subsequently reduced heterozygosity. In our study, however, we found no evidence that FA of labial scales has been affected by the transition to parthenogenesis via hybridization. Similarly, Vrijenhoek and Lerman s (98) studies of poeciliid fishes found that homozygosity from population bottlenecks can increase FA in sexual populations, but the increased heterozygosity afforded by hybrid origins of unisexuality did not translate to further reduced (or increased) FA in these fishes. They also made the important point that clonal organisms such as H. binoei have heterozygosity assurance. This means that, at times of low population density, they are less susceptible to developmental instability and other problems associated with the erosion of heterozygosity. This could be an important benefit to parthenogenetic H. binoei considering their climatically harsh environments (Kearney et al. 003). In laboratory conditions, parthenogenetic H. binoei have lower fecundity than their sexual progenitors (means of 3.0 vs. 4.4 eggs per female per year: M. Kearney and R. Shine, unpubl. data). This disparity suggests that the reproductive advantage enjoyed by parthenogenetic H. binoei is less than twofold and in fact approaches.5-fold. The reduced developmental success observed in parthenogenetic H. binoei in the present study could further reduce their twofold advantage. Although fecundity and viability are important aspects of overall reproductive rate, differences in generation time also can have a major impact (Lewontin 965). For example, although parthenogenetic lines of Drosophila mercatorum had lower fecundity and viability than their sexual progenitors, earlier maturation in the former resulted in similar intrinsic rates of increase (Kramer and Templeton 00). Our study indicates that parthenogenetic H. binoei have slower rates of growth than sexual forms (Fig. 5), which may also influence their lifetime reproductive output. Although our results provide some evidence for developmental disturbance in parthenogenetic H. binoei, the severity of the disturbance is much less than that observed in laboratory-evolved unisexual lineages (Wetherington et al. 987; Corley et al. 999; Corley and Moore 999; Kramer and Templeton 00; Kramer et al. 00a,b). These laboratory studies suggest strong constraints on the formation of parthenogenetic lineages (Vrijenhoek 989). Indeed, mtdna analysis of parthenogenetic H. binoei suggests that their origins were extremely restricted in space, presumably because hybridization between populations of the parental sexual species only rarely produced successful parthenogenetic offspring (Moritz et al. 989b; Moritz 99). In a situation where parthenogenesis arises by multiple hybridization events, many genetic combinations can be tested, thereby greatly increasing the odds of successful lineages becoming established. Parthenogenetic H. binoei exhibit considerable clonal diversity, much of which can be attributed to multiple hybrid origins (Moritz et al. 989a). The successful clones we see today are likely to be a highly biased subset of those that arose initially; that is, only those that were capable of successful development and reproduction. How does the sensitivity of development to temperature differ between parthenogenetic and sexual races of H. binoei? The two principal environmental attributes that affect reptilian incubation are temperature and moisture (Packard and Packard 988), both of which can vary greatly among lizard nests. Geckos from the subfamily Gekkoninae (including H. binoei) are unusual among lizards in that they lay eggs with a watertight calcareous shell, rendering incubation independent of moisture (Dunson 98). Thus in this study we have focused only on temperature effects. Our data demonstrate strong effects of incubation temperature on hatchling phenotypes, congruent with previous research on reptiles (e.g., Shine et al. 997; Qualls and Andrews 999; Shine 999). The traits affected include incubation time, tail length, body mass (when corrected for egg mass), labial scale count, running speed, growth rate, and hatchling survival. Of the three incubation temperature treatments we used (4 C, 7 C and 30 C) the coldest incubation was clearly the most stressful. Lizards incubated at 4 C took about five months to hatch whereas those incubated at 30 C took only two months. Additionally, hatchlings from the 4 C treatment ran slowly (Fig. 4) and had poor survival. They also had shorter tails (Fig. 3c) and more labial scales (Fig. 3d f), although the significance of these effects for fitness is unknown. Growth rate was also reduced in individuals incubated at 7 C compared to those incubated at 30 C. All hatchlings were free to thermoregulate between 5 C and 35 C as they were growing, so we do not know whether the effect of incubation temperature on growth rate was due to altered behavioral thermoregulation, or direct effects on physiology. Moreover, because 7 C incubations were biased toward first clutches (see Materials and Methods), this apparent incubation effect on growth might actually be a clutch number effect, with hatchlings from earlier clutches growing more slowly. Hatching success was also generally reduced at lower temperatures but this effect was most pronounced for the SM6 sexual race (Fig. ). Overall, 4 C is a stressful temperature for development in H. binoei. Overall, our data show that developmental success and stability were broadly similar between parthenogens and sexuals. Where differences occurred, parthenogens generally had reduced success. More striking was a consistent trend for the effects of incubation temperature on developmental rates and hatchling phenotypes to be weaker for parthenogens than for their sexual progenitors, as evidenced by the highly significant interaction term between reproductive mode and incubation temperature in the MANOVA reported above. These results suggest that development in parthenogens may be more buffered against the effects of temperature (particularly, cold stress), perhaps as a result of polyploidy, interclonal selection, or the high heterozygosity stemming from their hybrid origin. ACKNOWLEDGMENTS We thank P. Comber, B. Phillips, D. O Connor and C. Moritz for discussion and assistance in the field and the many