experimental test in a lizard - BES journal

13 downloads 0 Views 131KB Size Report
Consequences of maternal yolk testosterone for offspring development and survival: experimental test in a lizard. T. ULLER,*† L. ASTHEIMER‡ and M. OLSSON ...
Functional Ecology 2007 21, 544–551

Consequences of maternal yolk testosterone for offspring development and survival: experimental test in a lizard Blackwell Publishing Ltd

T. ULLER,*† L. ASTHEIMER‡ and M. OLSSON* *School of Biological Sciences, University of Wollongong, 2500 NSW, and ‡School of Health Sciences, University of Wollongong, 2500 NSW, Australia

Summary 1. Hormone-mediated maternal effects and developmental plasticity are important sources of phenotypic variation, with potential consequences for trait evolution. Yet our understanding of the importance of maternal hormones for offspring fitness in natural populations is very limited, particularly in non-avian species. 2. We experimentally elevated yolk testosterone by injection of a physiological dose into eggs of the lizard Ctenophorus fordi Storr, to investigate its roles in offspring development, growth and survival. 3. Yolk testosterone did not influence incubation period, basic hatchling morphology or survival under natural conditions. However, there was evidence for increased growth in hatchlings from testosterone-treated eggs, suggesting that maternal hormones have potential fitness consequences in natural populations. 4. The positive effect of prenatal testosterone exposure on postnatal growth could represent a taxonomically widespread developmental mechanism that has evolved into an adaptive maternal effect in some taxa, but remains deleterious or selectively neutral in others. 5. A broader taxonomic perspective should increase our understanding of the role of physiological constraints in the evolution of endocrine maternal effects. Key-words: maternal effects, phenotypic plasticity, reptile, Ctenophorus fordi, yolk hormones Functional Ecology (2007) 21, 544–551 doi: 10.1111/j.1365-2435.2007.01264.x

Introduction

© 2007 The Authors. Journal compilation © 2007 British Ecological Society

Cross-generational transmission of non-genetic developmental resources is an important factor in offspring development (West-Eberhard 2003; Jablonka & Lamb 2005). The evolution of inherited environmental effects (subsequently referred to as ‘maternal effects’; Rossiter 1996) is dependent on its effects on both parents and offspring, the degree of conflict between generations, and the extent to which this conflict can be resolved through modification of the adult phenotype, offspring developmental trajectories and their relationships (Rossiter 1996, 1998; Badyaev 2005). During the past decade, it has become increasingly clear that maternal effects can induce adaptive responses in the offspring, for example by preparing them for specific environmental conditions (Mousseau & Fox 1998; Agrawal, Laforsch & Tollrian 1999). Nevertheless, the dual nature of selection on maternal effects unavoidably makes some maternal effects detrimental †Author to whom correspondence should be addressed. Email: [email protected]

to offspring fitness. In the best-case scenario, such detrimental effects can be avoided by reducing the temporal or spatial overlap between generations or through modification of offspring development, for example by evolving reduced sensitivity to maternal phenotypic variation (Badyaev 2005; Uller 2006). However, some maternally transmitted developmental resources are similar to, and potentially indistinguishable from, those produced by the offspring themselves. Hormones, for example, are produced by both maternal and embryonic tissue, and there is no known mechanism whereby a developing embryo can distinguish between its own (internally produced) and maternal (externally produced) hormones. Hormones play a crucial role in development of the phenotype, and prenatal hormone exposure can influence such fundamental aspects of fitness as sex, physiology, behaviour and life history (for reviews see Moore 1995; Ketterson et al. 1996; Clark & Galef 1998; Weinstock 2001; Welberg & Seckl 2001; Dufty, Clobert & Møller 2002; Groothuis et al. 2005; see also Sakata & Crews 2004). Incomplete buffering of hormone transmission between maternal and fetal tissue may therefore interfere with the optimal 544

545 Consequences of yolk testosterone in a lizard

© 2007 The Authors. Journal compilation © 2007 British Ecological Society, Functional Ecology, 21, 544–551

developmental trajectory of the offspring. Furthermore, the inability of offspring to discriminate between its own and maternal hormones may affect the outcome of parent–offspring conflict in favour of the parents. However, our understanding of the consequences of hormone-mediated maternal effects for offspring fitness in natural populations is very limited. Maternal hormone transfer to embryos has been studied mainly in mammals and birds, taxa with both pre- and postnatal care (for documentation of maternal yolk hormones in several of species turtle see Janzen et al. 1998; Elf, Lang & Fivizzani 2002). The presence of postnatal care may increase the predictability of conditions experienced subsequent to parturition or hatching, thereby facilitating the evolution of adaptive maternal effects (Rossiter 1996, 1998; Donohue & Schmitt 1998). In altricial birds, for example, maternal hormone allocation at the prenatal stage may have evolved to mediate sibling conflicts at the postnatal stage (Schwabl 1993; Groothuis et al. 2005). More specifically, an increase in the testosterone content of later-laid eggs may induce faster development to reduce hatching asynchrony (Schwabl 1996; Eising et al. 2001), potentially at a disadvantage to earlier-hatched siblings. In mammals, maternal hormone transfer to offspring has traditionally been considered to be maladaptive, but a recent surge in interest in prenatal stress syndrome among evolutionary biologists has generated a number of adaptive hypotheses (Bateson et al. 2004; Gluckman & Hanson 2004). Although both mammals and birds share the potential for adjustment in relation to postnatal rearing conditions, differences in the degree of spatial and temporal overlap between developing offspring may affect the potential for allocation of hormones and other developmental resources among offspring of different resource demands, such as sons and daughters (Moore 1995; Uller et al. 2004; Young & Badyaev 2004; Badyaev et al. 2005, 2006b; Badyaev, Oh & Mui 2006a; Saino et al. 2006; Uller 2006). A broad taxonomic approach is required to address these issues. In the present study, we investigated (i) levels of testosterone of maternal origin in eggs of an oviparous species without postoviposition care, the dragon lizard, Ctenophorus fordi Storr; and (ii) whether maternal yolk testosterone has any effects on offspring development and survival in this species. The choice of model organism ensured a direct link between prenatal hormone levels and offspring traits without any modulating effect of parental care. Documentation of naturally occurring levels of yolk testosterone in eggs was followed by experimental manipulation of testosterone levels and monitoring of development, hatchling traits, and growth and survival under natural conditions. Previous studies of viviparous lizards and altricial birds have suggested that early testosterone exposure increases growth rate but may reduce survival, potentially mediated via reduced immune function (Sockman & Schwabl 2000; Uller & Olsson 2003; Müller et al. 2005). However, because of the absence of sibling interactions (the main

selective force suggested in the avian literature; Groothuis et al. 2005), we predicted lower levels and a smaller effect of maternal steroids on offspring traits in C. fordi than is typical in birds.

Materials and methods The mallee dragon, C. fordi, is a small agamid lizard widely distributed in arid and semiarid Australia. The general biology of the species has been well documented (Cogger 1969, 1974, 1978). Sexual dimorphism is minor. The reproductive season commences on emergence from hibernation in spring (September), and females produce consecutive clutches of two to five eggs (mean ≈ 2·5) that they oviposit without any further parental care. Sex determination is not sensitive to incubation temperature (Harlow 2004; Uller & Olsson 2007; T.U., unpublished data). Males are not territorial and there is little male–male aggression. The vast majority of adults aestivate from January onwards, and only ≈10– 15% of adults survive until the next reproductive season (Cogger 1978). Offspring hatch from December to April, resulting in only a minor overlap between generations. All yolk is metabolized within 3 days of hatching, hence any long-term consequences of yolk testosterone must be due to early ‘programming’ effects on offspring phenotype rather than increased levels of exogenous hormones postparturition. We captured males and females from Yathong Nature Reserve, New South Wales, Australia at the beginning of the breeding season in September 2004 and 2005, and brought them back to the laboratory at the University of Wollongong. Reproductive status (pre- or postovulation) was determined by palpation. Males and females were kept in cages (645 × 413 × 347 mm); females were kept separately whereas males were kept in groups of two to three per cage. Each cage contained sand and spinifex grass, and a 60-W spotlight provided opportunities for thermoregulation for 10–12 h daily. Female cages were provided with additional sand for oviposition. In 2004 we sampled eggs for analyses of yolk hormone content. All eggs from seven clutches and a single egg from an additional 11 clutches were sampled for analyses of yolk testosterone content (N = 28), which was subsequently used as reference for experimental elevation of yolk hormones in 2005. Whole eggs were sampled, rather than using yolk biopsies, because of the small egg size (≈0·45 g at oviposition) and to avoid bias due to variation among eggs in the degree to which they had absorbed water at sampling. Eggs from three clutches were dissected under a light microscope to establish the stage at oviposition.

    Following oviposition, eggs were stored at –80 °C until they were analysed. Egg contents after freezing were homogeneous, with no identifiable albumen. Thus the eggs were weighed, the total contents removed for

546 T. Uller et al.

hormone extraction, and the leathery shell and membranes reweighed to determine yolk mass. Yolk samples were diluted with 500 µl PBS buffer and homogenized by vortexing with the aid of glass beads. To determine percentage recovery, yolk from each egg was spiked with 20 µl tritiated testosterone of known activity. Testosterone was measured in yolk as described by Schwabl (1993): yolk was extracted twice in petroleum ether/diethyl ether, neutral lipids were precipitated in ethanol, and testosterone was isolated on diatomaceous earth/glycol columns and measured in duplicate by competitive binding radioimmunoassay (RIA; Wingfield & Farner 1975). Mean recovery after column chromatography was 48% (range 23–100%) and final sample concentrations were individually adjusted for this loss. To ensure that the extraction eliminated yolk compounds that may bind or otherwise interfere with antibody binding, we validated the specificity and accuracy of the assay, adapting the method described by Hayward & Wingfield (2004). Briefly, endogenous testosterone was removed from four lizard eggs by homogenizing egg contents with charcoal (10 mg ml–1) in buffer. Following centrifugation, the supernatant was spiked with testosterone (500 pg ml–1), and serially diluted across seven tubes also containing testosteronefree supernatant. These samples were then extracted and run on columns as described above. The column eluates containing testosterone were analysed by RIA, and levels were compared with the concentration of hormone originally added to the yolk, by linear regression (  4). As the slopes and elevations were not significantly different (P = 0·15 and P = 0·54, respectively), we concluded that yolk compounds did not interfere with antibody binding, and that our testosterone RIA could be used to measure yolk testosterone.

    

© 2007 The Authors. Journal compilation © 2007 British Ecological Society, Functional Ecology, 21, 544–551

In 2005, females with ovulated eggs were allowed to oviposit in the laboratory before the onset of experiments. Subsequent to oviposition, all females were mated with one male from the laboratory population. Thereafter, each female was always remated with the same male before ovulation to minimize genetic variation among offspring between clutches, which may obscure treatment effects. No male was mated to more than one female. Half the females had the testosterone content of all their eggs from the first clutch increased by experimental injection of 30 pg testosterone dissolved into 1 µl sesame oil. The amount of testosterone injected represents 1 SD of the naturally occurring levels measured in 2004 (see Results), and thus resulted in total levels well within the observed range for the majority of embryos. The second group received injections with vehicle only. The second clutch of each female received the opposite manipulation to the first clutch. Thus

each female had one control and one experimental clutch, randomized with respect to laying order. This approach was chosen, rather than using a split-clutch design, because of the low clutch size of the study organism. If mortality of the first clutch was high (one young surviving) and the female laid a third clutch, we manipulated the third clutch and excluded the offspring from the first clutch from our analyses of hatchling traits. However, all viable hatchlings from clutches with low overall hatching success were included in estimates of survival to increase sample sizes for the field study. Injections were performed through the most pointed end of the egg using a 10-µl Hamilton syringe. Eggs were incubated individually at 29·5 °C in plastic cups with sterilized water and vermiculite (1 : 7 volume ratio). At hatching, we measured offspring mass (to the nearest 0·001 g), snout–vent length (svl) and total length (to the nearest 0·5 mm), and determined offspring sex using hemipene eversion (Harlow 1996). Body condition was calculated as the residuals from a regression of mass on svl. Because of the long distance between the field site and the university (≈700 km), offspring could not be released immediately on hatching. Lizards were therefore kept individually in cages (266 × 190 × 103 mm, with a mesh lid) with sand as substrate and spinifex grass, and fed twice daily with crickets (Gryllus spp.) until release (mean ± SE days in captivity = 8 ± 0·3). All offspring were measured for svl and total length, and released into suitable habitat. The main field site is situated between two dirt roads that effectively prevent migration, as mallee dragons are rarely observed more than a few metres from spinifex grass (Cogger 1969; T.U., personal observation). The efficiency of roads as barriers was confirmed by the present data on recaptures, with no released lizard being observed in adjacent habitats despite repeated search efforts on multiple occasions. Marked posts at ≈6-m intervals allowed us to calculate dispersal for each individual. However, because of poor road conditions (floods) during one of the release sessions, this marked part of the field site could not be reached, and lizards were released into a similar habitat (sealed off by the same two roads) but lacking posts. Consequently, our sample sizes for estimating dispersal are reduced. The areas were searched intensively during two sessions on 28–31 March and 6 –11 April, just before the onset of hibernation. Upon capture, each lizard was weighed to the nearest 0·1 g, svl and total length measured to the nearest mm, and painted with a black mark on the back to avoid timeconsuming recapture. This procedure prevented us from using standard mark–recapture analyses, and all lizards were therefore classified as recaptured or not recaptured.

  All analyses were performed in  ver. 8·2. Normality and homogeneity of variances for dependent variables were checked before each analysis (Quinn & Keough

547 Consequences of yolk testosterone in a lizard

Table 1. Final statistical models for hatchling traits Snout–vent length

Mass

Relative svl

Body condition

Random effects

Estimate

χ2

Estimate

χ2

Estimate

χ2

Estimate

χ2

Family Residual

0·24 0·45

19·8***

2158 1380

55·0***

4 × 10–5 5 × 10–5

54·0***

2482 5365

25·5***

Fixed effects

ndf

ddf

F

ddf

F

ddf

F

ddf

F

Treatment Sex Clutch no. Treatment × clutch no. Sex × clutch no.

1 1 1 1 1

123 118 123 51·2 –

1·34 11·80*** 1·01 1·85 –

118 – 118 – –

2·12 – 9·91** – –

– 137 – – –

– 5·30* – – –

123 118 123 – 116

1·31 0·02 20·4*** – 1·83

*, P < 0·05; **, P < 0·01; ***, P < 0·001. Backward elimination was used at P > 0·25.

2002). Because none of the response variables for hatchling traits conformed to normality when each individual was treated as independent (P < 0·01), we had to resort to mean values per clutch and offspring sex, which had approximately normal distributions for all continuous response variables. Thus the models included treatment, clutch number (early vs late) and offspring sex as fixed factors, and family as a random repeated factor using a compound symmetrical covariance structure. Models were analysed using PROC MIXED (or the GLIMMIX macro for logistic models); fixed effects were tested using F-tests with the degrees of freedom estimated using Satterthwaite’s approximation; and random effects were tested using likelihood-ratio tests (Littell et al. 1996). We started with the full model and used backward elimination of factors at P > 0·25, starting with the highest-order interactions (Quinn & Keough 2002). We did not exclude main effects that also were present in an interaction of P < 0·25. We ran our models on incubation time and offspring traits at hatching twice, both with and without mean egg mass as a covariate. Including egg mass did not change the interpretation of any of the analyses, and we therefore report exclusively results from analyses without egg mass. All results are presented as mean ± SE unless otherwise specified.

Results

© 2007 The Authors. Journal compilation © 2007 British Ecological Society, Functional Ecology, 21, 544–551

The mean total amount of yolk testosterone in the eggs analysed was 19·5 ± 5·65 pg (range 0–114, SD = 30), corresponding to a mean of ≈0·05–0·07 pg mg–1 yolk. There was no significant correlation between egg mass and testosterone content (means per clutch, rs = –0·16, P = 0·54, N = 17) or between oviposition date and testosterone content (means per clutch, rs = 0·03, P = 0·92, N = 18). Analysing only the small subset of clutches from which we had multiple eggs showed significant clutch effects (nonparametric : χ2 = 14·4, P = 0·025, df = 6). The three eggs examined were all found to be at stage 26–27 (Dufaure & Hubert

1961) and thus confirmed previous reports of the early embryonic stage at oviposition in Australian agamids (for comparison, oviposition typically occurs at developmental stage 30 in lizards in general; Shine 1983; Harlow 2004). Egg mass did not differ between first and second clutches, whereas family had a strong effect on mean egg mass (family: χ2 = 12·4, P < 0·01; clutch number: F1,45 = 1·73, P = 0·20). From 377 eggs, 321 hatched out (85%), with no difference between treatments (χ2 = 1·29, P = 0·26). Hatching failure was a combined effect of embryonic mortality and infertility. There was no effect of yolk testosterone treatment or egg mass on incubation time (treatment: F1,116 = 0·09, P = 0·76), and the final model included only family, clutch number and offspring sex (family: χ2 = 41·5, P < 0·001; clutch number: F1,117 = 15·6, P < 0·001; offspring sex: F1,114 = 1·34, P = 0·25). Furthermore, yolk testosterone did not influence offspring svl, mass, body condition or relative tail length at hatching (Table 1). Daughters were generally larger and had relatively shorter tails in relation to total body length, but there was no sex effect on body mass or condition. Three offspring died before release, resulting in a total of 318 released hatchlings. There was no difference in the svl at release between treatment and control lizards (F1,125 = 0·87, P = 0·35). Of the 318 juveniles, 52 (16·4%) were recaptured. Fifteen females recruited one offspring each, whereas 16 females recruited at least two offspring. Because this resulted in highly unbalanced data at the family level, we ran our analyses twice, both with and without including family as a random effect, to ensure that our results were robust. There was no difference in recapture rates between treatment and control offspring [family: χ2 = 6·7, P = 0·01; treatment: F1,62 = 1·42, P = 0·24; 30 (15 males, 15 females) of 163 control and 22 (eight males, 14 females) of 155 testosterone-manipulated lizards were recaptured] or in dispersal (all P > 0·25). However, there was a significant interaction between hatching date and treatment on offspring svl at recapture (Table 2). Separate analyses

Table 548 2. Final statistical models for snout–vent length at the onset of hibernation

T. Uller et al. A

B

Random effects

Estimate

χ2

P

Estimate

χ2

P

Family Residual

– –





3·59 1·65

7·3