Predation Drives Interpopulation Differences in Parental Care Expression

Journal of Animal Ecology 2013, 82, 429–437 doi: 10.1111/1365-2656.12015 Predation drives interpopulation differences in parental care expression

Wen-San Huang1*, Si-Min Lin2, Sylvain Dubey3 and David A. Pike4

1Department of Zoology, National Museum of Natural Science, Taichung 404, Taiwan; 2Department of Life Science, National Taiwan Normal University, Taipei 116, Taiwan; 3Department of Ecology and Evolution, University of Lausanne, 1015 Lausanne, Switzerland; and 4School of Marine and Tropical Biology, James Cook University, Townsville, Qld 4812, Australia

Summary 1. Expressing parental care after oviposition or parturition is usually an obligate (evolved) trait within a species, despite evolutionary theory predicting that widespread species should vary in whether or not they express parental care according to local selection pressures. The lizard Eutropis longicaudata expresses maternal care only in a single population throughout its large geographical range, but why this pattern occurs is unknown. 2. We used reciprocal translocation and predator exclusion experiments to test whether this intraspecific variation is a fixed trait within populations and whether predator abundance explains this perplexing pattern. 3. Wild-caught female lizards that were reciprocally translocated consistently guarded or abandoned eggs in line with their population of origin. By contrast, most lizards raised in a common garden environment and subsequently released as adults adopted the maternal care strategy of the recipient population, even when the parents originated from a population lacking maternal care. 4. Egg predation represents a significant fitness cost in the populations where females display egg-guarding behaviour, but guarding eggs outweighs this potential cost by increasing hatching success. 5. These results imply that predators can be a driving force in the expression of parental care in instances where it is normally absent and that local selection pressure is sufficient to cause behavioural divergence in whether or not parental care is expressed. Key-words: behavioural plasticity, islands, maternal care, parental care evolution, reciprocal translocation

(Tallamy 1984; Clutton-Brock 1991; Wesoowski 1994). Introduction The general life-history circumstances favouring parental Animals have evolved a wide range of parental care strat- care thus include low egg or juvenile survival without egies to increase offspring survival, ranging from nest-site- care, relatively long egg incubation periods and relatively guarding to much more complex provisioning behaviours short juvenile stages (Clutton-Brock 1991; Klug & Bonsall (Clutton-Brock 1991; Wesoowski 1994; Klug & Bonsall 2009), while the general ecological circumstances favour- 2009). These strategies have evolved independently numer- ing parental care include food availability and predation ous times in disparate animal lineages, providing strong risk (Lack 1968; Tallamy & Wood 1986). Overall, species evidence of their success as an evolutionary strategy for are more likely to evolve parental care when the risk of increasing fitness (Tallamy 1984; Shine 1988; Clutton- predation on eggs or offspring is high (Lack 1968; Brock 1991; Gross 2005). Evolutionary theory predicts Tallamy & Wood 1986). that parental care should evolve from a nonparental Parental care is any action of the parent after oviposi- caring ancestral state when the fitness benefits accrued by tion or parturition that increases the chances of survival protecting offspring outweigh the costs of providing care of the offspring (Shine 1988; Greene et al. 2002). For example, within single populations of cichlid fish (Sarotherodon galilaeus Linnaeus 1758) and birds (Remiz *Correspondence author. E-mail: [email protected] pendulinus Linnaeus 1758), parental care can be provided

© 2012 The Authors. Journal of Animal Ecology © 2012 British Ecological Society 430 W.-S. Huang et al. by both parents or by either one of the sexes alone and increases the proportion of eggs that hatch, offspring (Persson & O¨ hrstro¨ m 1989; Balshine-Earn & Earn 1998). growth rates, and survival (Huang & Pike 2011b). Furthermore, the type and intensity of parental care However, this behaviour is costly because eggs laid inside behaviours expressed can vary with offspring development the retaining wall are exposed to predators throughout (reviewed by Magnhagen 1992) or local environmental incubation, as compared with eggs buried beneath cover conditions (Dunn & Robertson 1992; Wheelwright, objects in natural habitat (Huang & Pike 2011b). Despite Schultz & Hodum 1992; Sze´ kely & Cuthill 1999). Thus, the increased risk of predators locating eggs laid inside of many widespread species show differences within and these retaining walls, female long-tailed skinks from 12 among populations in the duration, type or intensity of other populations (11 from the island of Taiwan and one parental care (e.g. Dunn & Robertson 1992; Wheelwright, from Green Island, all nesting in similar retaining walls) Schultz & Hodum 1992; Sze´ kely & Cuthill 1999; Webb abandon their eggs immediately after oviposition (W.-S. et al. 1999), but almost all species expressing parental care Huang, unpublished). Thus, the expression of parental are known to do so throughout their entire geographical care is not a fixed trait shared among populations (i.e. range. By contrast, examples of parental care expression obligate; as it appears to be in all other species) because in some populations but not others are extremely rare; only the females nesting inside the retaining wall on known examples include a frog (Martins, Pombal & Had- Orchid Island express parental care. As in many other dad 1999) and, if we consider maternal care to occur taxa, not all females remain with the eggs for the entire throughout the entire duration of pregnancy, lizard incubation period (but they still initially express, and thus species with both oviparous and viviparous populations provide, maternal care; Huang & Wang 2009). The dura- (e.g. Heulin et al. 1999). Other rare cases include nest tion of maternal nest-guarding depends on the risk of egg abandonment in species which provision their offspring, predation: when egg-eating snakes are rare, females leave which almost always results in offspring death (e.g. Burley the nest early in incubation, but when snakes are abun- & Johnson 2002; Reynolds, Goodwin & Freckleton 2002). dant, females continue defending the nest until the eggs Parental care is thus almost always expressed throughout hatch (Huang & Wang 2009). This suggests that local the entire distribution of most species. This evolutionary predation pressure plays an important role in the evolu- pattern limits our capacity to unravel the selective tion of maternal care. We can test this hypothesis using pressures leading to the evolution of parental care (Clut- reciprocal translocation experiments to test directly ton-Brock 1991; Wesoowski 1994; Gross 2005). It also whether exposing female lizards to local ecological condi- seems to contradict the notion that because widespread tions they normally would not experience will alter the species are under different selection pressures throughout expression of parental care behaviour from that of the their geographical range, they should express parental source population. care in line with predation risk, so as to maximize An additional hypothesis for the facultative parental offspring survival. Why, then, do most widespread species care expression in long-tailed skinks is that geographical show obligate parental care expression, rather than facul- isolation has resulted in substantial genetic divergence, tative care, where adults adjust whether they provide or where the diverged population now provides maternal do not provide parental care according to the threat of care. This would mean that other populations do not have predation of the offspring? This situation would only per- the capacity to express maternal care, and thus would not sist in species where the offspring can physically survive do so, even under the ecological circumstances in which without relying on either of the parents (e.g. many rep- other (genetically diverged) individuals do provide care. tiles, amphibians and invertebrates; Shine 1988). We would then expect the long-tailed skink population on Long-tailed skinks (Eutropis longicaudata Hallowell Orchid Island to represent a highly divergent lineage 1857) are widely distributed throughout South-East Asia (or possibly even a separate, but morphologically similar, and only express parental care in a single population on species). The straight separating the island of Taiwan Orchid Island, located off the south-eastern coast of from Orchid Island is sufficiently deep (>1 km) that Taiwan (Huang 2004). In this population, mothers nest- Orchid Island has never been connected with other ing within concrete retaining walls remain at the nest dur- islands, even during low sea levels of the last glacial maxi- ing incubation and actively deter egg-eating snakes mum (LGM). Consequently, we might expect an absence (Oligodon formosanus Gu¨ nther 1872) from entering the of shared haplotypes between Orchid Island and the other nest and consuming the eggs (Huang 2006a; Supplemen- populations and perhaps the presence of distinct genetic tary Information S1). Antisnake behaviour is only lineages there. Parental care can have an underlying expressed while gravid, immediately prior to oviposition genetic component; for example, mice lacking the fosB (Huang & Pike 2011a). By contrast, females nesting in gene will not nurture their offspring (Brown et al. 1996). natural habitats nearby (beneath rocks and logs) never Although our understanding of the genetic mechanisms guard their eggs during incubation (Huang & Pike underlying parental care is preliminary and limited to 2011b). Females nesting in the retaining wall gain a mammals (a group almost exclusively comprised of spe- fitness advantage by exposing their eggs to higher incuba- cies that always express parental care), other genes and tion temperatures, which decreases incubation duration experiential learning also play important roles (Rosenblatt

1994; Brown et al. 1996). Because parental care initially SanDiMen; this is likely due to migration outside of the study evolves from ancestors that do not provide care (Tallamy area or predation. Orchid Island contains only one lizard preda- 1984; Shine 1988; Clutton-Brock 1991), all populations of tor (the colubrid snake Elaphe carinata Gu¨ nther 1864), but San- long-tailed skinks could have genes coding for the expres- DiMen has many additional predators (including the sion of parental care, but there may be some ecological saurophagous snakes, Viperidae: Protobothrops mucrosquamatus Cantor 1839, Trimeresurus stejnegeri Schmidt 1925; Elapidae: trigger for this behaviour. Bungarus multicinctus multicinctus Blyth 1861, Naja naja atra These two pathways – local adaptation via behavioural Cantor 1862; Colubridae: Ptyas mucosus Linnaeus 1758). Females – plasticity vs. genetic divergence are not mutually exclu- from Orchid Island are naı¨ ve to these predators and may suffer sive, in that both ecological factors and genetic history higher predation rates on SanDiMen. We thus conducted a sec- could play a role in the facultative expression of parental ond reciprocal translocation experiment by releasing lizards care. We predicted that (i) parental care expression is a raised in a common garden environment. In June 2009, we behaviourally plastic trait that depends on the local envi- brought 32 gravid females from Orchid Island and 28 females ronment; consequently, females translocated among popu- from SanDiMen into the laboratory. These females were unre- lations should alter their behaviour to conform to that of lated to the ones used in our other translocation experiments, the recipient population; (ii) because the function of nest- and thus, the wild-caught and captive-reared females are indepen- guarding is to protect eggs from predation (Huang dent of one another. Females laid eggs over the next 3 months, which we incubated in individual jars containing moist vermicu- 2006a), predator abundances should be markedly higher lite at 25 °C. Hatchling lizards were placed in cages with their on Orchid Island than in other populations; (iii) the bene- clutch mates and raised until August 2010, when the lizards had fit of increased nest survival because of maternal care grown large enough to reach maturity (>98Á1 mm snout-vent should be higher in the population that expresses care length; Huang 2006b) but had not yet reproduced for the first and (iv) there will be little to no genetic divergence time. We released progeny in a fully factorial design by releasing between the population expressing parental care and other part of each clutch at the initial capture population of the mother populations not expressing care. If our predictions hold (SanDiMen or Orchid Island, serving as controls) or at the oppo- true, this will provide compelling evidence that predation site population (i.e. offspring whose mothers were collected from plays a major role in the plasticity of parental care expres- Orchid Island were moved to SanDiMen and vice versa). This sion, and thus that local selective pressures can influence design allowed us to test whether individuals from different popu- this behavioural innovation. lations, but raised under identical laboratory conditions, can alter their expression of parental care in line with the recipient population. Materials and methods reciprocal translocation experiments predator abundances and hatching success rates In our first reciprocal translocation experiment (2007), we moved 20 individually marked (using microchips) wild-caught adult To test our hypothesis that egg-eating snake predation rates on females with a history of nesting within the concrete retaining lizard eggs are higher in the insular population showing maternal wall and guarding eggs from Orchid Island to a concrete wall at care than elsewhere, we surveyed four long-tailed skink nesting SanDiMen on the island of Taiwan and moved a further 20 sites (SanDiMen on mainland Taiwan and Green Island, which females from SanDiMen to Orchid Island. In September 2010, both lack lizard maternal care, and Little Paiday Bay and Tung- we repeated this experiment by translocating 40 additional ching on Orchid Island, which both express care) to quantify females from each population to the other population. Following snake numbers and the proportion of lizard nests preyed upon by translocation, we surveyed all potential nesting sites within the egg-eating snakes. At each site, we surveyed a 2-km stretch of retaining walls at six-hour intervals (4 surveys dayÀ1) from June retaining wall (containing 1200–1500 holes used by skinks as À1 to August 2007–2011 and recorded whether recaptured females nesting sites) at 6-h intervals daily (i.e. 4 samples day ) from remained with their eggs following oviposition. We classified May to October 2001–2004 and individually marked all egg-eat- females as not having laid eggs when we saw no evidence of them ing snakes encountered and recorded snake predation rates on being gravid and did not observe them with a clutch; gravid eggs. We inferred predation by egg-eating snakes when the entire females had visibly distended abdomens and eggs that we could clutch was consumed; the only other egg predators at these sites feel by gently palpating the abdomen. In the maternal-caring are ants, which leave behind empty eggshells (Huang 2010). population, females remain with the nest for at least the first week of incubation (and can stay up to 35 days, that is, for the efficacy of parental care among populations entire incubation duration), but in the noncaring populations, females leave the nest immediately after laying eggs (typically We used a manipulative experiment to determine the relative within a few hours; Huang 2006a, 2007). We defined maternal effectiveness of maternal care in each population. We experimen- care as remaining with the eggs for >24 h after laying; these tally excluded vertebrates from accessing some nests while leaving females generally stayed at the nest for several days, but noncar- others exposed to predators at Little Paiday Bay on Orchid ing females had abandoned their nest within 6–12 h of laying Island, SanDiMen on the island of Taiwan and on Green Island. eggs. To exclude vertebrate predators (and thus document hatching Following this initial translocation, we were unable to locate rates in the absence of snake predation), we covered the entrance most of the females that were moved from Orchid Island to holes leading to freshly laid lizard nests with plastic mesh (1 mm

© 2012 The Authors. Journal of Animal Ecology © 2012 British Ecological Society, Journal of Animal Ecology, 82, 429–437 432 W.-S. Huang et al. diameter, too small for any vertebrate to enter). The mesh was smallest distances and ending when all haplotypes were connected large enough to allow invertebrate predators (ants) to enter the or when the distance that corresponded to the parsimony limit nest and fungus to infect the eggs (Huang 2006a) and thus was reached (Clement, Posada & Crandall 2000). allowed us to document rates of egg hatching in the absence of We used uST to assess the genetic structure between popula- both egg-eating snake predation and the mother lizard. We did tions, which takes into account the genetic distance between not cover the entrance to control nests with any form of mesh, haplotypes and the haplotype frequencies [using ARLEQUIN and thus, these nests were left completely exposed to both verte- (Excofﬁer & Lischer 2010)]. We used the Kimura two-parameter brate and ant predators. Because female lizards at Little Paiday genetic distance as the genetic model (K2P; Kimura 1980). We

Bay on Orchid Island guard their eggs upon hatching, we calculated population structure using F-statistics (FST; Wright removed the female from nests in both treatments to match the 1965) using strictly haplotype frequencies. Significance values absence of nest-guarding at the other two populations. Compar- were obtained after 1000 permutations and tested whether FST ing the difference in hatching success of nests covered with mesh values differed significantly from 0. We used AMOVA (Excoffier & (n = 30, 15 and 22 clutches from Little Paiday Bay, Green Island Lischer 2010) to compare the proportion of contribution at and SanDiMen, respectively) to a different set of control nests at among-island and within-island levels. Finally, we used BEAST ver. the same site that were not covered by mesh (n = 142, 52 and 156 1.6 (Drummond & Rambaut 2007) to estimate the time of most clutches, respectively) allowed us to determine empirically the recent common ancestor (tMRCA) for all populations. increase in hatching success that would be derived from guarding a nest in each population (note that other associated costs may be minimal; Huang 2007). Results genetic diversity reciprocal translocation experiments

We collected 76 Eutropis longicaudata tail muscle tissue samples All of the wild-caught adult females released from (stored in 95% ethanol before DNA extraction) from 14 insular SanDiMen onto Orchid Island abandoned their nests populations: 12 populations from the island of Taiwan plus one immediately after laying (two individual females did so in population each from nearby Green and Orchid Islands (Fig. 3). each of 2008, 2009 and 2010; Table 1). Egg-eating snakes We isolated total genomic DNA using Qiagen DNeasy Blood & consumed all of these clutches. Of the females known to 9 Tissue Kits (Qiagen, Taipei, Taiwan). DNA was suspended in 1 have expressed maternal care that were released onto San- TE buffer and stored at À20 °C. We amplified the complete DiMen, only two were found in 2008 and both females sequence of the mitochondrial cytochrome b gene (1137 bp) using were guarding their eggs (Table 1). We found a similar polymerase chain reaction (PCR). The primer pair PL (5′-aaccaa- gacctgtgayaygaa-3)/PH (5′-ggcttacaagaccarkgcttt-3′) was designed pattern after repeating this reciprocal translocation with from the consensus sequences of several lizard species in the fam- larger sample sizes: all encountered nesting females ily Scincidae. expressed or did not express maternal care in line with Reactions were conducted in a 20 lL reaction volume contain- their population of origin, rather than changing their ing 19 PCR buffer (10 mM Tris–HCl, pH 9Á0; 50 mM KCl, behaviour to match that of the recipient population 0Á01% (w/v) elatine and 0Á1% Triton X-100), 0Á8 U Taq DNA (Table 1). Rates of encountering nesting females, non- polymerase (Amersham Pharmacia Biotech Inc., Piscataway, NJ, nesting females (i.e. those that were not gravid and who USA), 0Á2 lM each primer, 0Á5mM dNTP and 50 ng template thus did not lay eggs) or not finding females were similar ° DNA. The PCR condition was set to denaturation at 94 C for between release locations (v2 = 1Á54, d.f. = 2, P = 0Á46; ° ° ° 3 min, then 35 cycles at 94 C for 30 s, 52 C for 40 s and 72 C Table 1). Adult females originating from Orchid Island for 90 s, followed by a final extension at 72 °C for 10 min using were significantly more likely to express parental care iCycler Thermal Cycler (Bio-Rad Laboratories Taiwan Ltd., than were females originating from SanDiMen, even when Taipei, Taiwan). PCR products were purified with a PCR Prod- uct Pre-sequencing Kit (USB Molecular Biology Reagents and in a novel environment (100% vs. 0%, respectively; v2 = Á = < Á Biochemicals, Santa Clara, CA, USA) and used as the template 32 0, d.f. 1, P 0 00001; Table 1). Orchid Island for direct DNA sequencing reactions with a DYEnamic ET Dye females remained with their egg clutch on SanDiMen for Terminator Cycle Sequencing Kit (Amersham Pharmacia Bio- 1–9 days (mean = 3Á5 days). This strongly suggests that tech). We sequenced products on a MegaBACE 1000 automated wild-caught adult females with prior reproductive experi- DNA sequencer (Amersham Biosciences). Sequences were deter- ence make consistent decisions about whether or not they mined in both directions, and the original signals were proofread express maternal care, regardless of the environmental using SEQUENCHER version 4.7 (Gene Codes Corporation, Ann conditions experienced at the time of nesting. Arbor, MI, USA). We compared our sequences to other pub- For the lizards born and raised to maturity in captivity, lished skink sequences to confirm PCR amplification accuracy. both experimental groups significantly altered their mater- We evaluated genetic diversity of the 14 skink populations nal care expression relative to the control groups using haplotype (h) (Nei & Tajima 1983) and nucleotide diversity (comparing lizards from Orchid Island released there vs. (p) (Nei 1987) using DNASP 5.0 (Librado & Rozas 2009). We used < Á statistical parsimony (Templeton, Crandall & Sing 1992) imple- moved to SanDiMen; Fisher’s exact test, P 0 001; mented in TCS Version 1.21 (Clement, Posada & Crandall 2000) comparing lizards from SanDiMen released there vs. 2 to define the primary uncorrected distance (p) above which there moved to Orchid Island, v = 12Á89, d.f. = 1, P < 0Á0001; is a >5% probability that the parsimony criterion is violated. All Table 1). All female lizards born to mothers collected connections were established among haplotypes, starting with the from Orchid Island and released there as adults expressed

Table 1. Results from reciprocal translocation experiments, moving lizards from the population expressing maternal care (Orchid Island) to a population not expressing maternal care (SanDiMen, located on the main island of Taiwan) and vice versa. We performed two experiments, the ﬁrst of which moved gravid wild-caught adult females from one population to the other and the second of which cap- tured a different set of gravid females from the wild and raised the progeny to maturity in captivity. We ran the ﬁrst experiment twice (in both 2007 and 2010) because of low recapture rates. The last column summarizes the percentage of females known to have expressed maternal care, given as the percentage of females that laid eggs and expressed care and then as the percentage of all females observed expressing care (i.e. including females who were not observed to be gravid and thus were not known to lay eggs). We could not assess the reproductive condition (gravid and not gravid) of females that were not recaptured

Population Number of lizards Rate of parental Not expressing care expression maternal care Not (including females Origin Recipient Expressing maternal care (not gravid) recaptured Total that did not lay eggs)

Wild-caught females translocated in 2007 and recaptured from 2007 to 2011 Orchid Island SanDiMen 2 0 18 20 100% SanDiMen Orchid Island 0 14 (10) 6 20 0% Wild-caught females translocated in 2010 and recaptured in 2011 Orchid Island SanDiMen 18 10 (10) 12 40 100% SanDiMen Orchid Island 0 25 (15) 15 40 0% Progeny raised in captivity from wild-caught females, released in 2010 Orchid Island Orchid Island 28 5 (5) 12 45 100% (87Á5%) SanDiMen SanDiMen 0 15 (12) 17 32 0% Orchid Island SanDiMen 6 10 (10) 30 46 100% (37Á5%) SanDiMen Orchid Island 13 10 (10) 9 32 100% (56Á5%) maternal care, as did the individuals moved to SanDiMen ments, body size did not play a signiﬁcant role in whether (Table 1). By contrast, female lizards originating from or not maternal care was expressed (comparing body size SanDiMen and released there were never observed to at release among the four translocation treatments and express maternal care, but 100% of individuals moved to parental care expression using clutch as a random effect:

Orchid Island displayed maternal care (Fisher’s exact test, SVL: release treatment, F3,68 = 2Á38, P = 0Á08; maternal

P = 0Á002; Table 1). This clearly suggests that females care expression F1,68 = 2Á39, P = 0Á13; mass: release treat- from both populations have the capacity to express paren- ment, F3,68 = 8Á36, P < 0Á0001; maternal care expression tal care when raised in a common garden environment, F1,68 = 0Á68, P = 0Á41). strong evidence that the expression of parental care can be a behaviourally plastic trait. egg-eating snake abundances and hatching For the lizards born and raised to maturity in captivity, success rates the duration of maternal care varied significantly among recipient populations (F2,44 = 12Á12, P < 0Á0001). Captive- Predation is the plausible trigger that has caused female raised females originating from Orchid Island and lizards to guard their nests during incubation on Orchid released there, and females originating from SanDiMen Island. If so, long-tailed skink populations lacking mater- and released to Orchid Island, stayed at the nest signifi- nal care should have lower snake abundances and thus cantly longer (mean durations of 17Á4 and 20Á3 days and higher rates of egg hatching success than Orchid Island. ranges of 8–31 and 8–33 days, respectively; Tukey’s post This is exactly what we found; egg-eating snake abundances hoc test, P < 0Á0001 in both cases) than those originating were substantially higher at the two Orchid Island sites from Orchid Island and released onto SanDiMen (mean than at two sites lacking lizard maternal care (v2 = 156Á73, duration = 3Á2 days, range: 1–5 days). d.f. = 3, P < 0Á0001; Fig. 1). The Orchid Island sites had By randomly allocating individual lizards to each treat- snake abundances 9–20 times higher than the two sites lack- ment with respect to clutch of origin, we avoided con- ing maternal care. Snake numbers corresponded with egg- founding effects of initial body size (comparing hatchling eating snake predation rates on lizard eggs; significantly body size among the four translocation treatments using higher proportions of lizard clutches were preyed upon by clutch as a random effect: snout-vent length (SVL) snakes on the Orchid Island sites than at the other two sites 2 F3,137 = 1Á69, P = 0Á17; mass F3,137 = 1Á07, P = 0Á36). (v = 69Á15, d.f. = 3, P < 0Á0001; Fig. 1). The two Orchid However, at the time of release, there was significant vari- Island sites had similar rates of lizard egg predation by ation in female body size among treatments (comparing snakes (v2 = 1Á09, d.f. = 1, P < 0Á296; Fig. 1), but despite body size among the four translocation treatments using snake abundances being 9–20 times higher at these sites, clutch as a random effect: SVL at release F3,150 = 1Á69, predation rates were only twice as high as those docu-

P = 0Á002; mass at release F3,151 = 1Á07, P = 0Á008). mented from the two populations not expressing parental Despite variation in body size at release among treat- care (Fig. 1). Snake abundances were unrelated to lizard

Fig. 1. Egg-eating snake (Oligodon formosanus) numbers and the associated percentage of lizard nests preyed upon by snakes from two insular lizard populations not expressing parental care (San- DiMen on the main island of Taiwan and Green Island) and two Fig. 2. Experimental demonstration of the fitness benefits derived insular lizard populations from Orchid Island expressing parental from maternal care in three different populations. Predation rates care (Little Paiday Bay and Tungching). Snake numbers are the of lizard nests are expressed as the proportion of long-tailed total number of snake encounters along a 2-km stretch of con- skink eggs hatching from the three populations without any form crete retaining wall from 2001 to 2004, and the total number of of nest protection (i.e. without a nest-guarding female present) = lizard nests is as follows: SanDiMen n 156, Green Island for populations on Taiwan (SanDiMen), Green Island and Little = = = n 54, Little Paiday Bay n 173 and Tungching n 37. Paiday Bay on Orchid Island, and when vertebrate predators were excluded from accessing eggs using mesh netting. The only population in which maternal care would result in a fitness benefit is from nest availability: Little Paiday Bay on Orchid Island Orchid Island, where females naturally express maternal care. (n = 173) and SanDiMen (n = 156) had large numbers of lizard nests, while Green Island (n = 54) and Tungching on genetic diversity Orchid Island (n = 37) had substantially fewer lizard nests. The 76 long-tailed skink sequences revealed five different efficacy of parental care among populations haplotypes of 1137 bp with four segregating sites (GenBank accession numbers: JF689928–JF689932; In the absence of maternal care, hatching success was Table 2; Fig. 3). Most individuals (63/76 or 83%) lowest on Orchid Island and much higher in the two pop- belonged to the most frequent haplotype H1. This haplo- ulations lacking maternal care (comparing hatched/ type appeared in all sampling sites, including Green and unhatched eggs among the mainland, Green Island and Orchid Islands (Table 2; Fig. 3). The other haplotypes Orchid Island, n = 108, 60 and 450 eggs, respectively; (H2–H5; all with a single mutation separating them from v2 = 106Á74, d.f. = 2, P < 0Á0001; Fig. 2). The two popu- H1) were rare and private and had a very restricted distri- lations lacking maternal care had similar hatching success bution (Fig. 3). Pairwise FST and corrected uST values rates (v2 = Á08, d.f. = 1, P = 0Á08; Fig. 2). Nest tempera- between Taiwan vs. Green Island, Taiwan vs. Orchid tures and relative humidity were similar among these pop- Island and Green Island vs. Orchid Island were 0Á02142, ulations and thus cannot explain this pattern À0Á00092, and 0Á00000 and 0Á02139, À0Á00085, and [mean ± SEM: SanDiMen = 30Á8 ± 0Á21 °C; 91Á5 ± 2Á4% 0Á00000, respectively, (all pairwise P-values > 0Á05). AMOVA (n = 248 nests), Green Island = 30Á3 ± 0Á30 °C; 91Á0 ± indicated that most genetic variation was within popula- 1Á9% (n = 125 nests) and Little Paiday Bay on Orchid tions (73Á91%), while variation among islands was much Island = 30Á0±0Á30 °C; 90Á9±2Á6% (n = 45 nests); lower (Table 3). We estimated the tMRCA to be between

F2,415 =1Á57, P =0Á21 for temperature; F2,415 =0Á89, 49 750 and 18 020 years ago (using divergence rates rang- P =0Á76 for relative humidity]. ing from 1% to 2Á85% per MY; Gubitz, Thorpe & Malho- Experimentally excluding vertebrate predators from tra 2000; Paulo et al. 2001), indicating that these island accessing lizard eggs (a proxy for the provision of mater- populations were differentiated roughly after the LGM. nal care) substantially increased hatching success rates on Orchid Island above that of nests without an attending Discussion female (n = 1000 eggs, v2 = 200Á89, d.f. = 1, P < 0Á0001; Fig. 2), but had no significant effect at the other two sites We sought to understand why long-tailed skinks show (mainland: n = 70 eggs, v2 = 0Á31, d.f. = 1, P = 0Á58; facultative maternal care expression throughout their geo- Green Island: n = 156 eggs, v2 = 0Á17, d.f. = 1, P = 0Á68; graphical range, which sharply contrasts the obligate Fig. 2). This confirms that Orchid Island is the only pop- expression of parental care throughout the entire ulation for which maternal nest-guarding would provide a geographical range of most other species (see Martins, direct fitness benefit via an increase in egg hatching Pombal & Haddad 1999 for an exception). We emphasize success. that many animal populations vary in the type of parental

care expressed or in the duration of parental care, and yet nearly all of these populations still express parental care to some degree. Our reciprocal translocation experiments revealed two exciting patterns: (i) adult females living in the wild consistently displayed the same behaviour as their natal population, even after reproducing in a new population up to three consecutive years following translocation (n = 2 females), but that (ii) individuals born and raised to maturity in a common garden environment have the capacity to express parental care, regardless of their population of origin (Table 1). Consequently, the low levels of genetic divergence between Orchid Island and other populations (Table 3, Fig. 3) do not adequately explain underlying geographical differences in maternal care expression (see Heulin et al. 1999 for a similar out- come comparing viviparous and oviparous populations of Lacerta vivipara Lichtenstein 1823). Given the lack of genetic divergence between populations (Table 3; Fig. 3), we might expect the ecological factors on Orchid Island (e.g. elevated predation pressure on lizard eggs) to elicit otherwise noncaring females to remain with their eggs during incubation. This is exactly what we found; females with prior nesting experience consistently expressed or did not express parental care depending on their population of origin, whereas captive-raised females lacking prior nesting experience from both populations of origin expressed maternal care on Orchid Island. Thus, once a female expresses parental care initially, it is possible that she continues doing so during future reproductive events, Fig. 3. Sampled populations of Eutropis longicaudata, showing despite any change in selection pressure to do so. F values between the island of Taiwan and Green and Orchid ST How do nesting females ‘decide’ whether or not to guard Islands. Also shown are haplotype frequencies within populations and a statistical parsimony network of haplotypes (H1–H5). Pop- their eggs during incubation? Interactions with predators ulations are numbered as in Table 1: 1 = Orchid Island, during ontogeny cannot play a role – our lizards raised in 2 = Green Island and 3–14 = Island of Taiwan. captivity never came in contact with any potential predators until after we released them as adults. Because some of these naı¨ ve individuals expressed maternal care, individual experience with predators immediately before nesting (com- bined with the propensity for individuals to learn) is a likely Table 2. Long-tailed skink (Eutropis longicaudata) molecular mechanism. However, we cannot discount the fascinating data, showing sample sizes (n), number of haplotypes (Nhap), hap- possibility that lizards are eavesdropping on the nest-guard- lotype diversity (h) and nucleotide diversity (p) ing of conspecifics and using the social behaviour of others as a cue to their own behaviour. Further experiments are Sample locality Population nNhap h p necessary to disentangle these hypotheses. Orchid Island 1 10 1 0Á0000 0Á0000 The Orchid Island females that expressed maternal care Green Island 2 10 2 0Á2000 0Á0002 on SanDiMen remained only briefly at the nest-site Island of Taiwan 56 4 0Á3695 0Á0004 (1–5 days) as compared with those located on Orchid BeiNan 3 5 1 0Á0000 0Á0000 – TaiMaLi 4 3 1 0Á0000 0Á0000 Island (8 35 days). We suspect that this (and our decreased DaWu 5 1 1 0Á0000 0Á0000 ability to locate these females; Table 1) is because these LaiYi 6 10 2 0Á5556 0Á0005 females are unable to recognize many of the new predators TaiWu 7 10 2 0Á5333 0Á0005 they could be encountering there (e.g. Downes & Adams WanAn 8 2 1 0Á0000 0Á0000 2001). The only lizard predator on Orchid Island is the SanDiMen 9 3 1 0Á0000 0Á0000 TianLiao 10 10 2 0Á3333 0Á0003 snake Elaphe carinana, which long-tailed skinks readily ShanLin 11 1 1 0Á0000 0Á0000 recognize and flee from (Huang & Wang 2009). By con- HsinHua 12 5 1 0Á0000 0Á0000 trast, SanDiMen supports five additional saurophagous DaNei 13 4 1 0Á0000 0Á0000 snakes in three families (Colubridae, Elapidae and Viperi- SanJiePu 14 6 1 0Á0000 0Á0000 dae). Thus, we suspect that females attempting to stay at Total 76 5 0Á3060 0Á0003 the nest-site to provide maternal care at SanDiMen either

Table 3. Analysis of molecular variance results

Genetic variation Sum of squares % of Variation Fixation indices Signiﬁcance tests

Among groups 0Á378 À29Á46 ΦCT = À0Á29458 P < 0Á001, d.f. = 2 Among populations within groups 5Á160 55Á54 ΦSC = 0Á42905 P < 0Á001, d.f. = 11 Within populations 6Á633 73Á91 ΦST = 0Á26086 P = 0Á49, d.f. = 62 abandon their nest after encounters with these predators 2005), and thus, we cannot discount the possibility that (sensu Huang & Wang 2009) or become depredated. the differences in parental care expression among popula- Over a 4-year period, we found substantially more egg- tions has a genetic component. Additional population eating snakes on Orchid Island than elsewhere, and these genetic studies are needed to infer migration and coloniza- snakes were responsible for higher egg predation rates tion scenarios. According to current knowledge of this liz- (Fig. 1). One of the main ecological characteristics associ- ard and available genetic evidence, this behaviour must ated with parental care is the low survival rates of eggs in have evolved within a relatively short time span. The the absence of care (Clutton-Brock 1991; Klug & Bonsall results from our reciprocal translocation of females raised 2009), which is in line with the interpopulation differences in captivity, along with predator abundances, are consis- we observed. When female long-tailed skinks guard their tent with the hypothesis that predators play a major role nests on Orchid Island, they are extremely effective at in the expression of parental care. deterring egg-eating snakes (which pose no threat to the Our results – the lack of any discernable genetic differ- mother lizard; Huang 2006a; Supporting Information S1). ences among populations, wild-caught female lizards This seems to contradict the notion that egg-eating snakes expressing parental care in line with their population of ori- are responsible for the seemingly high predation levels on gin, some captive-reared lizards adopting parental care Orchid Island, where females should be able to protect expression of the recipient population, and predators and their eggs from predation. However, on Orchid Island, predation rates being higher in the population with mater- snake abundances were 9–20 times higher than elsewhere, nal care – provide compelling evidence that predation has but egg predation was only twice that of other populations played a major role in the evolution of parental care (Fig. 1). Hence, nest-guarding females are quite effective expression in a single population of this widespread species. at deterring these snakes, in addition to the snakes relying Why there are substantially more egg-eating snakes on on additional food sources (e.g. sea turtle eggs; Huang Orchid Island as compared with other populations is not et al. 2011). Most instances of egg-eating snakes consum- entirely clear, but further ecological studies may help eluci- ing clutches on Orchid Island can be narrowed down to (i) date this. This is one of only a few species known to contra- snakes predating eggs when females are foraging away dict the general paradigm that parental care is an evolved from the nest (Huang 2007), (ii) lizard predators (e.g. the response consistently expressed throughout the entire range snake Elaphe carinata) scaring the female away (Huang of a species (even though the type and duration of parental 2006a) or (iii) the female abandoning the nest prior to the care can vary geographically) and shows that parental care eggs hatching (Huang 2007). can be a behaviourally plastic trait among individuals and We found low levels of genetic diversity among popula- populations. Although this pattern may be more common tions, with an absence of any defined structure between than reflected by the literature, virtually all endotherms Orchid Island and the other populations (Fig. 3). Clearly, provide parental care (Clutton-Brock 1991; Klug & Bonsall this indicates that the observed phylogeographical pattern 2009), so ectotherms are most likely to display this pattern of long-tailed skinks on these islands is very recent and (e.g. our study; Martins, Pombal & Haddad 1999). Given that there are no distinct genetic lineages present on the the overall paucity of research on parental care in ecto- different islands. The lack of phylogeographical structure therms (especially terrestrial ectotherms; Tallamy 1984; could be due to the recent colonization of these islands Shine 1988; Clutton-Brock 1991; Stahlschmidt 2011), the after the LGM, which substantially altered regional fauna lack of additional reports is unsurprising. and flora distributions in the region (Dynesius & Jansson 2000; Harrison et al. 2001). Global cooling during the Acknowledgements LGM could have eliminated long-tailed skinks from the region, which subsequently recolonized the area following We thank C.H. Chang, R.K. Lee, Y.B. Lin and E.A. Roznik for assis- tance. The Kuo Wu Hsiu Luan Culture and Education Foundation and warming after the LGM. Reptiles can travel long distances the National Science Council of Taiwan (NSC 99-2621-B-178-001-MY3) across oceans by rafting on floating debris, which may provided funding. Animal ethics permits were issued by the Taiwanese have aided recolonization (Calsbeek & Smith 2003; Vidal National Museum of Natural Science (NMNSHP02-002). et al. 2008). This has been reported in other Taiwanese reptiles (Lin et al. 2008) and is consistent with long-tailed References skinks reaching their northernmost range limit in southern Balshine-Earn, S. & Earn, D.J.D. (1998) On the evolutionary pathway of Taiwan (Shang, Yang & Li 2009). Fast, adaptive genetic parental care in mouth-brooding cichlid fish. Proceedings of the Royal changes have been documented in the wild (Hairston et al. Society B: Biological Sciences, 265, 2217–2222.

Brown, J.R., Ye, H.R., Bronson, T., Dikkes, P. & Greenberg, M.E. (1996) Lack, D. (1968) Ecological Adaptations for Breeding in Birds. Methuen and A defect in nurturing in mice lacking the immediate early gene fosB. Co, London, UK. Cell, 86, 297–309. Librado, P. & Rozas, J. (2009) DnaSP v5: A software for comprehensive Burley, N.T. & Johnson, K. (2002) The evolution of avian parental care. analysis of DNA polymorphism data. Bioinformatics, 25, 1451–1452. Philosophical Transactions of the Royal Society B, 357, 241–250. Lin, H.-C., Li, S.-H., Fong, J. & Lin, S.-M. (2008) Ventral coloration dif- Calsbeek, R. & Smith, T.B. (2003) Ocean currents mediate evolution in ferentiation and mitochondrial sequences of the Chinese Cobra (Naja island lizards. Nature, 426, 552–555. atra) in Taiwan. Conservation Genetics, 9, 1089–1097. Clement, M., Posada, D. & Crandall, K. (2000) TCS: a computer program Magnhagen, C. (1992) Parental care and predation risk in fish. Annales to estimate gene genealogies. Molecular Ecology, 9, 1657–1660. Zooloologici Fennici, 29, 227–232. Clutton-Brock, T.H. (1991) The Evolution of Parental Care. Princeton Uni- Martins, M., Pombal, J.P. & Haddad, C.F.B. (1999) Escalated aggressive versity Press, Princeton, NJ. behaviour and facultative parental care in the nest building gladiator Downes, S.J. & Adams, M. (2001) Geographic variation in anti-snake frog, Hyla faber. Amphibia-Reptilia, 19,65–73. behaviour: the evolution of scent mediated defence responses in a ge- Nei, M. (1987) Molecular Evolutionary Genetics. Columbia University kkonid lizard. Evolution, 55, 605–615. Press, New York, NY, USA. Drummond, A.J. & Rambaut, A. (2007) BEAST: Bayesian evolutionary Nei, M. & Tajima, F. (1983) Maximum likelihood estimation of the number of analysis by sampling trees. BMC Evolutionary Biology, 7, 214. nucleotide substitutions from restriction sites data. Genetics, 105,207–217. Dunn, P.O. & Robertson, R.J. (1992) Geographic variation in the impor- Paulo, O.S., Dias, C., Bruford, M.W., Jordan, W.C. & Nichols, R.A. tance of male parental care and mating systems in tree swallows. Behav- (2001) The persistence of Pliocene populations through the Pleistocene ioral Ecology, 3, 291–299. climatic cycles: evidence from the phylogeography of an Iberian lizard. Dynesius, M. & Jansson, R. (2000) Evolutionary consequences of changes Proceedings of the Royal Society of London, Series B, 268, 1625–1630. in species’ geographical distributions driven by Milankovitch climate Persson, O. & O¨ hrstro¨ m, P. (1989) A new avian mating system: ambisex- oscillations. Proceedings of the National Academy of Sciences of the ual polygamy in the penduline tit Remiz pendulinus. Ornis Scandinavica, USA, 97, 9115–9120. 20, 105–111. Excoffier, L. & Lischer, H.E.L. (2010) Arlequin suite ver 3.5: A new series Reynolds, J.D., Goodwin, N.B. & Freckleton, R.P. (2002) Evolutionary of programs to perform population genetics analyses under Linux and transitions in parental care and live bearing in vertebrates. Philosophical Windows. Molecular Ecology Resources, 10, 564–567. Transactions of the Royal Society B, 357, 269–281. Greene, H.W., May, P.G., Hardy Sr D.L., Sciturro, J.M. & Farrell, T.M. Rosenblatt, J.S. (1994) Psychobiology of maternal behavior: contribution (2002). Parental behavior by vipers. Biology of the Vipers (eds G.W. to the clinical understanding of maternal behavior among humans. Acta Schuett, M. Hoggren, M.E. Douglas & H.W. Greene), pp. 179–225. Paediatrica Suppl, 397,3–8. Eagle Mountain Publications, Utah. Shang, K.S., Yang, Y.J. & Li, P.H. (2009) Field Guide to Amphibians and Gross, M.R. (2005) The evolution of parental care. Quarterly Review of Reptiles in Taiwan. Owl Press, Taipei, Taiwan. Biology, 80,37–46. Shine, R. (1988) Parental care in reptiles. Biology of the Reptilia, vol. 16 Gubitz, T., Thorpe, R.S. & Malhotra, A. (2000) Phylogeography and nat- (eds C. Gans & R.B. Huey), pp. 275–330. Alan R. Liss, New York. ural selection in the Tenerife gecko Tarentola delalandii: testing histori- Stahlschmidt, Z.K. (2011) Taxonomic chauvinism revisited: insight from cal and adaptive hypotheses. Molecular Ecology, 9, 1213–1221. parental care research. PLoS ONE, 6, e24192. Hairston Jr, N.G., Ellner, S.P., Geber, M.A., Yoshida, T. & Fox, J.A. Sze´ kely, T. & Cuthill, I.C. (1999) Brood desertion in Kentish plover: the (2005) Rapid evolution and the convergence of ecological and evolu- value of parental care. Behavioral Ecology, 10, 191–197. tionary time. Ecology Letters, 8, 1114–1127. Tallamy, D.W. (1984) Insect parental care. BioScience, 34,20–24. Harrison, S.P., Yu, G., Takahara, H. & Prentice, I.C. (2001) Diversity of Tallamy, D.W. & Wood, T.K. (1986) Convergence patterns in subsocial temperate plants in east Asia. Nature, 413, 129–130. insects. Annual Review of Entomology, 31, 369–390. Heulin, B., Surget-Groba, Y., Guiller, A., Guillaume, C.P. & Deunff, J. Templeton, A.R., Crandall, K.A. & Sing, C.F. (1992) A cladistic analysis (1999) Comparisons of mitochondrial DNA (mtDNA) sequences (16S of phenotypic associations with haplotypes inferred from restriction rRNA gene) between oviparous and viviparous strains of Lacerta vivi- endonuclease mapping and DNA sequence data. III. Cladogram estima- para: a preliminary study. Molecular Ecology, 8, 1627–1631. tion. Genetics, 132, 619–633. Huang, W.-S. (2004) Reptile ecology and the evolution of parental care on Vidal, N., Azvolinsky, A., Cruaud, C. & Hedges, S.B. (2008) Origin of tropical a tropical Asian island. PhD thesis. Cornell University, New York. American burrowing reptiles by transatlantic rafting. Biology Letters, 4, 115–118. Huang, W.-S. (2006a) Parental care in the long-tailed skink, Mabuya longi- Webb, J.N., Houston, A.I., McNamara, J.M. & Sze´ kely, T. (1999) Multi- caudata on a tropical Asian island. Animal Behaviour, 72, 791–795. ple patterns of parental care. Animal Behaviour, 58, 983–993. Huang, W.-S. (2006b) Ecological characteristics of the skink, Mabuya lon- Wesoowski, T. (1994) On the origin of parental care and the early evolution of gicaudata, on a tropical East Asian island. Copeia, 2006, 293–300. male and female parental-roles in birds. American Naturalist, 143,39–58. Huang, W.-S. (2007) Costs of egg caring in the skink, Mabuya longicauda- Wheelwright, N.T., Schultz, C.B. & Hodum, P.J. (1992) Polygyny and ta. Ecological Research, 22, 659–664. male parental care in savannah sparrows: effects on female fitness. Huang, W.-S. (2010) Foraging behaviors of two sympatric ant species in Behavioral Ecology and Sociobiology, 31, 279–289. response to lizard eggs. Zoology, 113,85–90. Wright, S. (1965) The interpretation of population structure by F-statistics Huang, W.-S. & Pike, D.A. (2011a) Does maternal care evolve through with special regard to systems of mating. Evolution, 19, 395–420. egg recognition or directed territoriality? Journal of Evolutionary Biol- ogy, 24, 1984–1991. Received 22 May 2012; accepted 20 September 2012 Huang, W.-S. & Pike, D.A. (2011b) Climate change impacts on fitness Handling Editor: Tim Coulson depend on nesting habitat in lizards. Functional Ecology, 25, 1125–1136. Huang, W.-S. & Wang, H.-Y. (2009) Predation risks and anti-predation parental care behavior: an experimental study in a tropical skink. Ethol- Supporting Information ogy, 115, 273–279. Huang, W.-S., Greene, H.W., Chang, T.-J. & Shine, R. (2011) Territorial Additional Supporting Information may be found in the online version Oligodon formosanus Proceedings behavior in Taiwanese kukrisnakes ( ). of this article. of the National Academy of Sciences of the United States of America, 108, 7455–7459. Kimura, M. (1980) A simple method for estimating evolutionary rate of Movie S1. On Orchid Island, Taiwan, nest-guarding female base substitution through comparative studies of nucleotide sequences. skinks (Eutropis longicaudata) guard their eggs from intruding – Journal of Molecular Evolution, 16, 111 120. egg-eating snakes (Oligodon formosanus), a behaviour not present Klug, H. & Bonsall, M.B. (2009) Life history and the evolution of parental in other populations. care. Evolution, 64, 823–835.