Effects of visible implanted elastomer marking on physiological traits of frogs

Proteins and bacteria on the skin of amphibians can protect them from diseases. Marking frogs with elastomer tags had no effect on stress hormone production or protein release, but lead to a slightly greater abundance of skin microbes. This marking technique may be a comparatively safe option for amphibians.


Introduction
Amphibians are currently undergoing major global population declines, a major driver of which are emerging infectious diseases, such as chytridiomycosis and ranaviruses (reviewed by Gray et al., 2009;Kilpatrick et al., 2010). Amphibians possess innate immune defences, including antimicrobial peptides and symbiotic bacterial communities, that can protect them from such diseases (reviewed by Rollins-Smith and Conlon, 2005;Rollins-Smith et al., 2011;Bletz et al., 2013). The potential for using symbiotic bacteria for probiotic treatments against chytridiomycosis is currently being investigated (reviewed by Bletz et al., 2013), and on-going research has identified symbiotic bacteria that inhibit the growth of the chytridiomycosis fungus, Batrachochytrium dendrobatidis, from a number of amphibian species (e.g. Harris et al., 2006;Culp et al., 2007;Lauer et al., 2007;Banning et al., 2008;Brucker et al., 2008;Flechas et al., 2012;Loudon et al., 2013;Roth et al., 2013).
Events that cause disruption of symbiotic bacterial communities or deplete peptide stores could potentially increase the susceptibility of individuals to disease and may have implications for amphibians involved in probiotic trials or time course studies that investigate symbiotic bacterial communities. Wild and captive amphibians are often marked for a variety of reasons, including identification of individuals, in order to avoid resampling, or to conduct mark-releaserecapture surveys. Historically, toe-clipping was often used for identification, although less invasive techniques, such as passive integrated transponder (PIT) tagging or visible implant elastomer (VIE) dyes, are now more commonly used. The existing literature indicates that these marking techniques have few negative effects on amphibians (e.g. Kinkead et al., 2006;Phillips and Fries, 2009;Schmidt and Schwarzkopf, 2010;Sapsford et al., 2014), although a recent study showed that PIT tagging causes a proliferation in resident bacteria and fungi on the skin of frogs (Antwis et al., 2014a).
Amphibian peptides are produced and stored in granular glands and control the colonization of the skin by both pathogenic and symbiotic micro-organisms (reviewed by Conlon, 2011). It is thought that amphibian peptides have a reciprocal relationship with symbiotic bacterial communities on the skin and may play a role in maintaining the bacterial community by limiting their antimicrobial potency (reviewed by Boman, 2000;Conlon, 2011). Although peptide production and secretion occurs continuously at a low level, it increases substantially in response to stressful events, such as alarm, injury or pathogenic infection (Rollins-Smith et al., 2011;Pask et al., 2012). Peptide granular glands are surrounded by a layer of myoepithelial cells containing α-adrenoreceptors that are activated by adrenaline (epinephrine) or noradrenaline (norepinephrine) to induce the release of peptides (reviewed by Rollins-Smith, 2001;Delfino et al., 2006). The greater the stress experienced, the greater the level of peptide release by an individual; Xenopus laevis injected with increasing concentrations of noradrenaline showed a concurrent increase in release of peptides . Particularly stressful events can cause granular glands to become exhausted of peptides, and studies have shown that concurrent increases in glucocorticoids lead to the inhibition of transcription factors required to produce peptides (Simmaco et al., 1997(Simmaco et al., , 1998. This temporary absence of peptides can lead to the proliferation of bacterial communities (Simmaco et al., 1998) and may explain the results seen by Antwis et al. (2014a).
There are limited studies that have investigated the effects of marking techniques on steroid (e.g. adrenaline, noradrenaline, glucocorticoid) production in amphibians. Kinkead et al. (2006) showed that two species of salamanders (Desmognathus fuscus and Desmognathus monticola) marked with VIE or toe-clipped under varying degrees of anaesthesia did not exhibit increased noradrenalineor adrenaline levels in comparison to unmarked salamanders. Handling has been shown to increase circulating glucocorticoid concentrations in a range of amphibian species (Woodley and Lacy, 2010;Narayan et al., 2011aNarayan et al., , 2012aNarayan et al., , 2013, and toe-clipping causes a significant and prolonged (>3 days) increase in circulating corticosterone in cane toads (Rhinella marina; Narayan et al., 2011b). However, to the authors' knowledge, the effects of other marking techniques on glucocorticoid production have not been evaluated for amphibians, along with concurrent effects on peptide release and bacterial communities. Here, we determine the effects of VIE marking on culturable cutaneous microbial abundance, peptide release and faecal glucocorticoid metabolite concentrations of captive red-eyed tree frogs (Agalychnis callidryas).

Ethics statement
The Chester Zoo Ethical Committee and the University of Manchester Ethics Committee approved this study prior to starting. This study did not require a UK Home Office licence because VIE tagging is an approved method for the identification of frogs, and data were collected during routine marking of frogs for identification purposes. Frogs were checked daily throughout the study and for 2 weeks after for any signs of adverse reaction to any of the techniques used, none of which was observed.

Study animals, husbandry and experimental design
A total of 16 adult A. callidryas were used in this study; eight (four males, four females) in the control (unmarked) group and eight (four males, four females) in the marked group. All frogs were from the same clutch of F3 generation captivebred individuals and had not previously been involved in any other studies. Prior to the start of the experiment, frogs were group housed in 60 × 45 × 45 cm glass tanks (ExoTerra ® , UK) with a naturalistic set-up consisting of a LECA base for drainage and a soil layer covered with dried oak leaves, and planted with a peace lily and cuttings of devil's ivy. Two weeks prior to the start of the study, frogs were randomly assigned to 16 individual experimental housing tubs (30 × 18 × 20 cm ExoTerra ® plastic faunariums), with frogs from the same group housed tanks spread across treatment groups to allow for prior variation in bacterial communities. Individual housing was required to distinguish faecal samples between individuals and to avoid the potential for bacterial communities to be confounded by group housing during the study. Experimental tanks were lined with damp paper towels and contained a water dish and a cutting of devil's ivy. Paper towels and water dishes were changed twice weekly. All tanks were maintained with a ZooMed Reptisun ® 10.0 ultraviolet strip light with reflectors and a Philips daylight bulb with reflectors, both of which were on a 10:14 light:dark cycle. Frogs were fed three or four times weekly with black crickets (Gryllus bimaculatus) gut-loaded on fresh fruit and vegetables and dusted with Nutrobal (VetArk, UK).
After a 2 week habituation period in the individual housing, faecal samples were collected daily from day 1 until the end of the study (day 28) to measure faecal glucocorticoid metabolite concentrations (see 'Faecal glucocorticoid metabolite concentrations' section below for further details). On day 8, frogs were sampled for bacterial communities and peptide release as described below (see 'Microbial communities' and 'Peptide release' sections). This provided baseline data for each individual, which is referred to hereafter as 'start of study'. On day 14, frogs in the 'marked' group were tagged using VIE dye (see 'Marking techniques' below for details). On day 15, all frogs were sampled again for bacterial communities and peptide release, and these data are referred to as 'post-marking'. On day 28, bacterial communities and peptide release were sampled again, and these data are referred to as 'end of study'.

Marking techniques
Agalychnis callidryas in the marked group were tagged on day 14 of the study using VIE dye (Northwest Marine Technology Inc., USA). This method was chosen because frogs were due to be rehomed at another institution and it was necessary to distinguish these frogs from existing A. callidryas in the collection at the receiving institution without the requirement for individual identification. The elastomer dye was prepared according to the manufacturer's instructions. Gloves were worn throughout marking and changed between frogs to minimize cross-contamination. During marking, the frog was restrained on a flat and stable surface that had previously been sterilized, and the sterile needle containing the dye was inserted smoothly and quickly under the skin of the underside of the tibiofibular (G. Garcia, personal communication). The marking process was standardized and took ~30 s per individual. In order to determine whether the whole marking process affected physiological traits of frogs, control frogs were not sham injected or manipulated at the post-marking time point. All frogs were monitored for the remainder of the study and for 2 weeks after the end of the study for any signs of adverse reaction, none of which was observed.

Faecal glucocorticoid metabolite concentrations
Tanks were checked daily within an hour of the lights coming on (A. callidryas are nocturnal and so defaecate at night), and any faecal samples were frozen at −20°C for subsequent analyses. Gloves were used to handle samples and changed between to avoid contamination. Glucocorticoid metabolite analyses were conducted at the Chester Zoo's Wildlife Endocrinology Laboratory.
Glucocorticoid metabolites were extracted from faecal samples following thawing and manual homogenization using a wet-weight shaking extraction adapted from Walker et al. (2002). In brief, the individual faecal sample weight was recorded (mean = 0.138 ± 0.007 g) and then combined with 90% methanol, shaken overnight at room temperature and centrifuged for 20 min at 598g. The supernatant was decanted and evaporated to dryness under air. Faecal extracts were resuspended in 0.25 ml of 100% methanol and stored at −20°C until analysis.
Glucocorticoid metabolites (corticosterone) were analysed using a modified enzyme immunoassay (EIA) as previously described (Munro and Stabenfeldt, 1984;Watson et al., 2013). Each EIA used an antibody (polyclonal corticosterone antiserum CJM006 supplied by C. J. Munro, University of California, Davis, CA, USA), horseradish peroxidaseconjugated label (corticosterone prepared according to Munro and Stabenfeldt, 1984) and standards (corticosterone; Sigma-Aldrich, UK). The modified assay procedures were as follows for the corticosterone EIA. Antiserum was diluted at 1:15 000 in coating buffer (0.05 m NaHCO 3 , pH 9.6), and 50 µl per well was loaded onto a 96-well Nunc-Immuno Maxisorp (Thermo-Fisher Scientific) microtiter plate, which was covered with a plate sealer and left overnight at 4°C. Plates were washed five times (0.15 m NaCl, 0.05% Tween 20) and then 50 µl per well of each standard and sample (3.9-1000 pg per well or samples diluted 1:10 in EIA buffer) were loaded in duplicate with 50 µl per well of horseradish peroxidase conjugate diluted in EIA buffer (1:70 000). Following incubation in the dark for 2 h at room temperature, plates were washed five times and incubated with 100 µl per well of substrate [0.4 mm 2,2′-azino-di-(3-ethylbenzthiazoline sulfonic acid) diammonium salt, 1.6 mm H 2 O 2 and 0.05 m citrate, pH 4.0], left to develop at room temperature in the dark and measured at 405 nm at optical density 0.8-1.0. The corticosterone EIA was biochemically validated for measuring corticosterone metabolites in A. callidryas faeces through parallelism (sample percentage binding = 30.606 ± 0.8129, r 2 = 0.9209, F 1,7 = 81.461, P < 0.001) and matrix interference assessment (observed = 31.074 ± 1.522, r 2 = 0.998, F 1,7 = 4089.8, P < 0.001). The cross-reactivities for corticosterone antiserum CJM006 have been reported elsewhere (Watson et al., 2013), and the intra-and interassay variation were 7.52 and 6.33%, respectively. The assay was biologically validated by demonstrating a significant rise in faecal glucocorticoid metabolite concentrations from prior to any manipulations being conducted on frogs (i.e. days 1-8) to the highest value collected for each individual after this time (days 9-28; Student's paired t test; t = 3.213, d.f. = 15, P = 0.006).
Final corticosterone metabolite concentrations were calculated per gram of wet faecal mass. Data collected before any handling or marking events (days 1-7) were assigned as 'start of study', data collected from 1 day post-marking until 5 days post-marking (days 16-19) were assigned as 'post-marking', and data collected from day 20 until day 28 were assigned as 'end of study'. Raw data values for faecal glucocorticoid metabolite concentrations indicated an increase at around 3-5 days post-marking (days 17-19; Fig. 1), which is within the expected lag time in glucocorticoid metabolite deposition in faeces for a range of species (Wasser et al., 2000;Bamberg et al., 2001;Martínez-Mota et al., 2008;Santymire et al., 2012;Smith et al., 2012) including amphibians (Cikanek et al., 2014). Therefore these data were used for the postmarking time point as described above. Data were analysed in RStudio (using the lme4 package) for effects of treatment, time point and the interaction between treatment and time point using a generalized linear mixed model, with individual included as a random effect, with Tukey's post hoc analysis for pairwise comparisons.

Microbial communities
Microbial communities were collected using similar methods to those previously described by Antwis et al. (2014b). Briefly, frogs were rinsed to remove transient bacteria using 50 ml of ultrapure water, which was retained for peptide quantification (see 'Peptide release' below). Frogs were then swabbed on their ventral surface 20 times using sterile swabs (Eurotubo ® , Rubi, Spain). Gloves were worn throughout the swabbing process, which took ~30 s per frog, and changed between individuals to minimize cross-contamination. Swabbing methods were consistent across individuals, treatment groups and time points. Swabs were placed into 1 m NaCl 2 solution and serial dilutions constructed up to a concentration of 10 −2 . Concentrations of 10 −1 and 10 −2 were plated out on lownutrient agar (R2A; Lab M, UK). Bacterial and fungal colony abundances were counted every 2 days until negligible new growth was observed (10 days after initial plating). Bacterial and fungal counts were then multiplied by their respective dilution factors and averaged for each individual. Data for bacterial abundance and fungal abundance were analysed separately in RStudio (using the lme4 package) for effects of treatment, time point and the interaction between treatment and time point using a generalized linear mixed model with Poisson distribution, with individual included as a random effect. Tukey's post hoc analysis was conducted for pairwise comparisons.

Peptide release
Prior to sampling for cutaneous microbes, frogs were rinsed on their dorsal and ventral surfaces by pouring 50 ml of ultrapure water over each surface five times. The final volume of liquid collected was recorded, and peptides were extracted from the water using previously described techniques (Rollins-Smith et al., 2002;Woodhams et al., 2006;Sheafor et al., 2008;Conlon et al., 2011). Briefly, samples were acidified to 1% using trifluoroacetic acid (TFA; Sigma Aldrich, UK) and then passed over individual Sep-Pak C-18 cartridges (Waters Associates, Milford, MA, USA) at a rate of 5 ml/min. Sep-Pak cartridges were primed prior to use with 1 ml of 1% TFA solution as recommended by the manufacturers. Peptides were then eluted using 2 ml of elution solution composed of 70% acetonitrile (Sigma Aldrich, UK), 29.9% ultrapure water and 0.1% TFA. Eluted peptides were concentrated to dryness in a vacuum centrifuge (Genevac MiVac Modular Concentrator System) and reconstituted in 150 µl of ultrapure water.
Peptides were quantified in triplicate for each sample using a BCA Protein Assay Kit (Thermo Scientific Pierce, UK) according to the manufacturer's instructions, except that bradykinin (Sigma Aldrich, UK) was used to establish a standard curve (Rollins-Smith et al., 2002;Woodhams et al., 2006;Sheafor et al., 2008;Conlon et al., 2011). Interpolated peptide concentrations were averaged for each individual and adjusted by a scaling factor that accounted for the volume of liquid passed through the SepPak cartridges, and then divided by the mass of the corresponding frog to give a final peptide concentration per gram of frog. Peptide data were analysed in RStudio (using the lme4 package) for effects of treatment, time point and the interaction between treatment and time point using a generalized linear mixed model with Poisson distribution, with individual included as a random effect. Tukey's post hoc analysis was conducted for pairwise comparisons.

Faecal glucocorticoid metabolite concentration
There was a significant effect of time (χ 2 = 29.679, d.f. = 2, P < 0.001) but not treatment (χ 2 = 1.257, d.f. = 1, P = 0.0262) or the interaction between treatment and time (χ 2 = 1.990, d.f. = 2, P = 0.370; Figs 1 and 2) on faecal glucocorticoid metabolite concentrations. Post hoc analyses showed significant differences between control frogs at the start of the study and at the end of the study (P = 0.002) and between control frogs post-marking and at the end of the study (P = 0.001), with end-of-study faecal glucocorticoid metabolite concentrations being lower at the end of the study than at the other two time points (Supplementary Table S1; Fig. 2). Marked frogs at the end of the study had significantly lower faecal glucocorticoid metabolite concentrations than control frogs post-marking (Supplementary Table S1; Fig. 2). As time was the only statistically significant predictor of faecal glucocorticoid metabolite concentrations in the original model, a simplified generalized linear mixed model containing only this variable was run. This model was statistically significant (χ 2 = 29.282, d.f. = 2, P < 0.001), and post hoc analyses showed that faecal glucocorticoid metabolite concentrations at the end of the study were significantly lower than those at the start of the study (P < 0.001) and postmarking (P < 0.001) across all frogs, but there were no significant differences between faecal glucocorticoid metabolite concentrations at the start of the study and post-marking (P = 0.377).

Microbial communities
There was a significant effect of time (χ 2 = 12 800, d.f. = 2, P < 0.001) and the interaction between treatment and time (χ 2 = 2893, d.f. = 2, P < 0.001) on bacterial abundance, but not of treatment alone (χ 2 = 0.963, d.f. = 1, P = 0.326). Post hoc analyses showed overall that there were significant decreases in bacterial abundance over time for both treatment groups, and the bacterial abundance associated with control frogs at the end of the study was significantly lower than that of marked frogs at all time points (Supplementary Table S1; Fig. 3). In particular, marked frogs had significantly greater bacterial abundance than control frogs at the end of the study (i.e. 2 weeks post-marking).
There was a significant effect of time (χ 2 = 544, d.f. = 2, P < 0.001) on the abundance of fungi associated with the skin of frogs, but no significant effects of treatment (χ 2 = 1.309, d.f. = 1, P = 0.253) or the interaction between time and treatment (χ 2 = 0.016, d.f. = 2, P = 0.992). Post hoc analyses showed that marked frogs had a significantly greater fungal abundance at the end of the study than post-marking (Supplementary Table S1 Supplementary Table S1). Group A significance bars are all statistically different to one another and indicate a significant decrease in bacterial abundance over time for control frogs, and likewise for group B significance bars for marked frogs.

Discussion
In this study, we found no evidence that VIE marking of frogs leads to an increase in faecal glucocorticoid metabolite concentrations in comparison to unmanipulated control animals. It has previously been shown that marking Desmognathus salamanders with VIE does not lead to increased noradrenaline or adrenaline in comparison to handling alone (Kinkead et al., 2006), but that handling leads to an increase in glucocorticoid concentrations above baseline levels in blood or urine from a range of amphibian species (Woodley and Lacy, 2010;Narayan et al., 2011bNarayan et al., , 2012aNarayan et al., , b, 2013. There is evidence of a small increase in faecal glucocorticoid metabolite concentrations from all frogs in this study after marking or handling (Fig. 1), although the individuals used in this study were captive bred and so may have been accustomed to some degree of manipulation. There was an overall decrease in faecal glucocorticoid metabolite concentrations across all frogs over the course of the study, with significantly lower concentrations at the end of the study compared with the other time points (Fig. 2). This may be due to the frogs becoming more familiar with the new environment or individual housing over the course of the study. Cikanek et al. (2014) found similar results in Panamanian harlequin frogs (Atelopus certus and Atelopus glyphus), with frogs maintained individually exhibiting lower faecal glucocorticoid metabolite concentrations than those maintained in groups of two or eight, and those maintained in group housing exhibiting a decline in glucocorticoid metabolite concentrations over the 4 week study period.
Faecal samples have been used to measure glucocorticoid metabolite concentrations in a range of taxa (e.g. Wasser et al., 2000;Bamberg et al., 2001;Martínez-Mota et al., 2008;Santymire et al., 2012;Smith et al., 2012), including amphibians (Cikanek et al., 2014). The values for faecal glucocorticoid metabolite concentrations obtained in the study presented here are similar to those obtained from A. certus and A. glyphus, and adrenocorticotrophic hormone challenges in these species indicated a similar lag time in glucocorticoid metabolite deposition in the faeces as for A. callidryas in this study (Cikanek et al., 2014). Changes in glucocorticoids are observed much more rapidly in urine and blood (within minutes to 2 days; e.g. Woodley and Lacy, 2010;Narayan et al., 2011aNarayan et al., , b, 2012aNarayan et al., , 2013Graham et al., 2013). Faecal sampling, however, offers an almost completely non-invasive method for monitoring adrenal activity in comparison to blood or urine collection, which requires handling of individuals to collect samples and thus can confound the data. Moreover, there is evidence from studies with mammals that faecal glucocorticoid metabolites remain relatively stable over a number of days at room temperature (Washburn and Millspaugh, 2002;Evans et al., 2013). The suitability of faecal or urine sampling for different species will be dependent on how a given species separates the deposition of glucocorticoid metabolites in waste products (e.g. Wasser et al., 2000;Bamberg et al., 2001). For example, attempts to validate glucocorticoid metabolites biochemically in the urine of A. callidryas and Agalychnis moreletii by our research group were unsuccessful (R. Purcell, personal observation), indicating that these species may primarily deposit glucocorticoid metabolites in their faeces.
Overall, there was a decrease in the abundance of bacteria on the skin of all frogs throughout the study period in both treatment groups, which may be related to the change in environment for the frogs (from group housing in naturalistic vivaria to individual housing in more sterile experimental enclosures). Küng et al. (2014)   microbial communities over time, and environmental conditions in captivity are known to affect the bacterial communities associated with A. callidryas Michaels et al., 2014). However, at the end of study the marked frogs had a significantly greater bacterial abundance than the control frogs, with an elevated fungal abundance in comparison to themselves post-marking. Although the variation around the fungal abundance data for marked frogs at the end of the study is considerable (Fig. 4), fungi were cultured from four of eight frogs in the marked group at this time point, in comparison to one frog post-marking and no frogs at the start of the study. These results indicate there may be some delayed microbial growth on the skin of frogs in response to VIE marking, although it is not clear if this continued beyond the end of the study or whether the bacterial communities had undergone a proliferation between the post-marking and end-of-study time points, and the results seen at the end of the study were the microbial communities returning to normal. However, no obvious skin infections were observed in frogs for 2 weeks after the study had finished.
The cause of the delayed greater abundance of microbes (although minor) on the skin of marked frogs is unclear. The VIE tags were unlikely to be completely sterile due to the preparation methods, and the greater abundance of microbes may reflect a minor infection from the tags. Alternatively, it is possible that the act of marking caused a disruption in the dynamics of the bacterial communities that may have had delayed effects for the microbial community or elicited a physiological response in the frogs. For example, studies in rats and humans have shown increases in immunoglobulins, lymphocytes and macrophages in response to injection of silicone (as contained in VIE tags; Smalley et al., 1995;Hill et al., 1996). The implications of the greater microbial abundance is not known, and it would be of interest to investigate the susceptibility of amphibians to B. dendrobatidis after tagging with VIE and other marking systems. Moreover, culturing methods are known to underestimate microbial diversity greatly (reviewed by Amann et al., 1995), and molecular techniques (e.g. next-generation sequencing) are required to characterize the community more fully and determine whether VIE marking and other techniques affect the nonculturable portion of the microbiome.
The results presented here are in contrast to the results found by Antwis et al. (2014a), which demonstrated a major and rapid proliferation in culturable bacteria on the skin of A. moreletii 1 day after marking with PIT tags. The needle used for VIE tagging is much narrower (30 gauge or 0.31 mm outer diameter) than for PIT tagging (18 gauge or 1.27 mm), and the 'foreign body' inserted under the skin is much smaller; therefore, frogs undergoing PIT tagging may experience greater adrenal activity than those that are VIE tagged, leading to differences in the responses of the microbial community (Simmaco et al., 1997(Simmaco et al., , 1998Antwis et al., 2014a). However, definitive conclusions are difficult to draw because the two studies were conducted on different Agalychnis species and the restraint time was slightly longer during PIT tagging than VIE marking (~1 minute and 30 s, respectively).
There were no changes in the quantity of peptides released by frogs throughout the study period for either treatment group, indicating that VIE marking has no effect on availability or release of peptide stores of A. callidryas. This result is expected given that VIE marking also had no effect of faecal glucocorticoid metabolite concentrations, which have the potential to inhibit the production of new peptides for storage (Simmaco et al., 1997(Simmaco et al., , 1998. To our knowledge, this is the first study to quantify the effects of a marking technique on the production of peptides by amphibians, and further research into the effects of other marking techniques (particularly PIT tagging) on peptide production and release is required.
Overall, we have shown that there is no effect of VIE marking on adrenal response (represented by faecal glucocorticoid metabolite concentrations) or peptide release of A. callidryas, although there was evidence of a minor increase in microbial abundance on the skin. This indicates that VIE may be a preferable marking technique to PIT tagging (which causes a rapid and major proliferation of skin microbes), particularly in the context of probiotic trials or time course studies that investigate symbiotic bacterial communities. Although more work is required to determine differences between host species in response to different marking techniques, as well as the effects of marking on the susceptibility of amphibians to disease, these results together with those of Antwis et al. (2014a) indicate that captive amphibians should not be released immediately after marking, and that where possible, probiotic treatments should not be applied to individuals less than 3 or 4 weeks after marking. Moreover, marking of wild amphibians should be carefully considered.

Supplementary material
Supplementary material is available at Conservation Physiology online.