Androgen responsiveness to simulated territorial intrusions in Allobates femoralis males: evidence supporting the challenge hypothesis in a territorial frog

Territorial behaviour has been widely described across many animal taxa, where the acquisition and defence of a territory are critical for the fitness of an individual. Extensive evidence suggests that androgens (e.g. testosterone) are involved in the modulation of territorial behaviour in male vertebrates. Short-term increase of androgen following a territorial encounter appears to favour the outcome of a challenge. The “Challenge Hypothesis” proposed by Wingfield and colleagues outlines the existence of a positive feedback relationship between androgen and social challenges (e.g. territorial intrusions) in male vertebrates. Here we tested the challenge hypothesis in the highly territorial poison frog, Allobates femoralis, in its natural habitat by exposing males to simulated territorial intrusions in form of acoustic playbacks. We quantified repeatedly androgen concentrations of individual males via a non-invasive water-borne sampling approach. Our results show that A. femoralis males exhibited a positive behavioural and androgenic response after being confronted to simulated territorial intrusions, providing support for the Challenge Hypothesis in a territorial frog.


Introduction 34
Territoriality is a widespread behaviour across many animal taxa and provides 35 the territory holder with primary access to critical resources for individual fitness such 36 as food, shelter, breeding sites and space for mating. In many species, only males 37 engage in competitive interactions and contests with their conspecifics for the 38 acquisition of territories (Davies, 1991). There is extensive evidence that androgens 39 are involved in the modulation of typical territorial behaviours such as advertisement 40 signalling and agonistic encounters in male vertebrates (Adkins-Regan, 2005). 41 Testosterone is the main circulating androgen in most male vertebrates and, besides 42 modulating the expression of primary sexual traits, its main function related to 43 territoriality is to prepare males for social interactions, like male-male competition 44 and agonistic encounters (Wingfield et al., 2006). 45

46
In most species with a seasonal breeding pattern, androgen levels undergo a 47 products in the holding water, as it is mentioned in similar publications (Scott et al., 151 2008). Every water bath consisted of a glass container (14cmx9cmx5cm) filled with 152 40 mL of distilled water. Frogs were placed in the water bath immediately after 153 capture and removed after 60 min, and then released at the original location. In order to calculate recovery efficiency of testosterone with the SPE 163 technique, we spiked two pools with 2 different testosterone concentrations, using 164 standards of the ELISA kit (see below). Samples were extracted and processed as 165 described above and stored at 4ºC until proceeding with the assay. Water samples 166 "without frog" were also processed as blank controls to evaluate any possible 167 contamination of the holding water. To assess water-borne androgen release rate, we 168 used thirteen adult A. femoralis (Body-size mean ± SD: males=2.74 ± 0.03 cm, N=7; 169 females = 2.79 ± 0.02 cm, N=6) in January 2018 from a laboratory population kept at 170 the animal care facilities at the University of Vienna. Briefly, we manually placed 171 each frog in consecutive water baths of sampling periods of 15, 30 and 60 min. All 172 samples were collected between 08:00 and 09:00 A.M., then extracted and processed 173 as described above and, stored at 4ºC until the assay. All frogs were fed at libitum 174 with wingless fruit flies every second day. 175

Parallelism between hormone concentration in blood and holding water 177
In order to know whether water-borne androgen reflected actual levels of 178 circulating testosterone at the time of sampling, we collected eighteen free-living 179 adult A. femoralis males (Body-size mean ± SD = 2.8 ± 0.1 cm) in April 2019, from a 180 population in the vicinity of Roura, French Guiana (4º43' N -52º18' W). Frogs were 181 attracted using playbacks, captured using plastic bags and transferred into individual 182 water baths for 60 min. Water samples were processed as described above. After 183 completion of the water baths, frogs were immediately euthanized with an overdose 184 of 20% benzocaine gel and rapidly decapitated. Trunk blood was collected into 1.5 185 mL Eppendorf tubes and centrifuged at 6000 rpm for 5 min (6-position rotor) to 186 separate the plasma. Plasma volume was recorded, and samples were transferred into 187 1.5 mL eppendorf tubes prefilled with 750 uL of 96% ethanol. In the laboratory, 188 testosterone was extracted from ethanol samples three times with freeze-decanting 189 following the methodology in Goymann et al., (2007). Briefly, samples were dried 190 down with N2 at 37ºC. Dried pellets were resuspended in 4 mL of dichloromethane 191 and 100 uL of distilled water and, then incubated at 4ºC overnight for equilibration. 192 The following day, samples were shaken for 1h and then centrifuged at 4000rpm for 193 10 min to separate the aqueous and organic phase, which was transferred into a new 194 tube by freeze-decanting. This process was repeated twice, and the organic phase was 195  between 08:00 and 18:00 h. We repeated the behavioural observations at least three times, at different times of the day (i.e. morning and/or afternoon), in non-consecutive 226 days and with a minimum of three days in between observations. After every 227 behavioural observation, each frog was gently captured with a plastic bag and 228 immediately transferred into a water bath for 60 min to assess the baseline water-229 borne androgen concentration (A). Additionally, 24 females from the same population 230 were also placed into individual water baths for 60 min in order to compare water-231 borne androgen baselines between sexes. 232

233
In the post-social stimulation phase (B), we confronted focal males to STIs 234 exclusively when they were found calling. Once a focal male was located, we placed 235 the loudspeaker on the forest ground at 1 -1.5 m distance from the focal male. We 236 approach the loudspeaker and/or did not cross the perimeter before the playback was 240 finished (non-responding). In order to determine the behavioural responsiveness of 241 the frogs to the territorial challenge, we performed 3 STIs trials which were audio 242 recorded and we measured the following behavioural parameters during each trial: (1) 243 latency to the first head-body orientation towards the speaker, (2) latency to the first 244 jump and, (3) latency until the frog reached the perimeter. Frogs were not handled or 245 manipulated at least three days before any further STI. After the STIs (regardless 246 whether the males responded or not) males were caught and immediately transferred 247 to a series of three consecutive water baths of 60 min/each (1h, 2h and 3h). This 248 sequence of water baths allowed us to determine a 3h timeline of androgen secretion 249 in water. Time elapsed between the end of the STI and the beginning of the water 250 baths was always less than 10 min. Water samples were collected individually after 251 every 60 min water bath without manipulating the frog to avoid stress. For this, we 252 used two flexible polymer tubing with one end attached to the glass box and the other 253 end attached to a 20mL syringe. One tubing was used to pump the water into the glass 254 box and the other was used to suck out the sample after every 1h water-bath. Samples 255 were processed and extracted as explained in the water-borne extraction section. We 256 repeated STIs three times per focal male with at least three days between trials. 257 258

2.4.Hormone assays 259
In order to estimate androgen concentration, we used a commercial enzymatic 260 immunoassay for testosterone (ADI-900-065; Enzo Life Sciences, Farmingdale, NY, 261 USA). Reconstituted samples were brought to room temperature and shaken at 500 262 rpm for 1 h prior the assay. Samples were plated in duplicate and assays were 263 performed following the manufacturer's protocol. Plates were read at 405 nm, with 264 correction between 570 and 590 nm, using a microplate reader (Multiskan Go, 265 Thermo Fisher Scientific Oy, Finland) and androgen concentrations were calculated 266 using the Thermo Scientific SkanIt Software (version 4.1). The detection limit for the 267 assay was 5.67 pg mL -1 . The cross reactivity of the testosterone antibody with other 268 androgens was below 15% (see manufacturers manual). The average intra-and inter-269 assay coefficient of variation were 3.38% and 11.05%, respectively. 270 271

Statistical analysis 272
Prior to analysis, hormone data were log-transformed to fit normality when 273 necessary. In order to know whether water-borne androgen concentration was 274 dependent on the frogs' body size and/or body area, we first calculated the body area- To determine the parallelism between hormone concentration in blood and 288 holding water, we performed a parametric correlation between the plasma and water-289 borne androgen concentrations using the Pearson's product moment correlation 290 coefficient. In order to compare water-borne androgen levels between males and 291 females, we performed a two-sample t-test. Since time of the day, vocal and 292 locomotor activity might be interdependent with androgen concentrations (Wada,293 1986), we asked whether baseline androgen levels were dependent on natural 294 behaviours and varied across the day. For this, we first performed a LMM with water-295 borne androgen levels as response variable, time of the day (morning/afternoon) as 296 fixed factor and frog ID as the random factor. Then, we performed a Varimax 297 normalized principal component analysis (PCA) in order to minimize redundancy 298 among the behavioural variables by using the function "principal" within the R 299 package psych (Revelle, 2019). Further, we performed a series of independent LMMs with the scores of the principal components obtained as response variables, time of 301 the day (morning/afternoon) and water-borne androgen levels as fixed factors and 302 frog ID as the random factor. 303

304
In order to investigate whether A. femoralis males respond to territorial 305 challenges (STIs) with an increase in androgen levels, we first performed a LMM 306 with androgen concentration as dependent variable, and the sampling time points (0h-307 morning/afternoon baselines-, 1h, 2h and 3h water bath sampling after STIs) as fixed 308 effects. We used frog ID as the random factor to account for repeated measurements. 309 In order to compare the androgen responsiveness to STIs between responding and 310 non-responding males, we estimated the androgen responsiveness to male-male androgen concentrations on a standardized scale (Goymann et al., 2007). For this, we 317 used the function "cohens.d" within the R package misty (Yanagida, 2020). 318 Finally, in order to know whether the phonotactic approach of A. femoralis 319 males is proportional with the androgen responsiveness, we first minimized 320 redundancy among the three responsiveness latencies (latency to the first head-body 321 orientation towards the speaker, latency to the first jump and, latency until the frog 322

Validation and sex differences of water-borne androgens 341
Recoveries of low and high standards were 98.21% and 105.98%, 342 respectively. "Blank" water samples were below the detection limit of the assay 343 ( Figure 1A). Correlation between expected and obtained androgen concentrations in 344 2 mL aliquots was highly significant (r=0.99, P=0.006; Figure 1B), and falls within 345 the range of detectability of the assay. Androgens released in 60 min water baths was 346 significantly higher than 15 min (LMM: ß=-112.43, t=-2.33, P=0.03; Figure 1C

Daily variation of behaviours and water-borne androgen levels 355
Three components were generated with eigenvalues greater than 1 ( Table 1)

Effect of STI on the phonotactic behaviour 389
Three principal components were generated, but just one component with an 390 eigen value greater than 1, which explained the 72% of the total variance ( Table 2).  water-borne sampling might even constitute a preferable alternative to invasive 500 methods (i.e. blood sampling), offering new ways on how to study the interplay 501 between social behaviour and hormones (Bell, 2019;Narayan, 2013;Wingfield et al., 502 2006). 503

504
We observed that water-borne androgen levels and vocal activity were higher 505 in the afternoon than in the morning in A. femoralis males. In fact, previous research 506 has found a higher calling activity peak of A. femoralis in the afternoon compared to 507 the morning (between 1500-1730 h; Kaefer et al., 2012;Roithmair, 1992). Allobates 508 femoralis males use advertisement calls to engage in social interactions with 509 conspecifics (e.g. territory tenancy advertisement, inter-male spacing, courtship; previous studies also showed that castrated and androgen treated males did not 519 maintain or increase vocal behaviour (Burmeister and Wilczynski, 2001;Wetzel and 520 Kelley, 1983), suggesting that androgens are needed but not the only hormones for 521 eliciting vocal behaviour. 522 We found that elements related to courtship and foraging behaviour (e.g. # of 524 head-body orientation, # of jumps, # of feeding events, courtship call duration) did concentration in the brain depends on androgen concentrations (Boyd, 1994). The 540 synergistic effects of androgen hormones and neuropeptides on courtship behaviour 541 need further investigation in poison frogs. 542 543 Water-borne androgen was increased 1h after the STI but returned to baseline 544 levels 2h after the STI. There are at least two possible reasons for such a pattern. First, 545 short-term changes in androgen levels in non-seasonal breeders have been associated 546 with the trade-off between parental care and aggressiveness (Wingfield et al., 1990). 547 In other words, androgen levels can facultatively rise during male-male contests but 548 decrease when males are parenting the broods. Allobates femoralis males typically 549 perform tadpole transport and, although we were unable to evaluate the effect of 550 parental care before or after presenting the STIs, unpublished data suggest that 551 parenting males have significantly lower water-borne androgen compared to non-552 parenting males (Rodríguez et al., unpublished data). Second, there are costs 553 associated with maintaining high androgen levels for a prolonged period of time such 554 as the suppression of immune function, increasing of parasitic infections (Folstad and 555 Karter, 1992) and impairing parental care (Wingfield et al., 1990). Thus, the return of 556 androgens to baseline levels after a short-term increase may ease the resume of 557 ongoing activities just before the intrusion. Interestingly, water-borne androgen levels 558 went below the pre-STI baseline levels 3h after the STI. This reduction might be the 559 consequence of negative feedback of the hypothalamo-pituitary-interrenal (HPI) axis 560 (Yao & Denver, 2007). Additionally, we cannot exclude that there are some inhibitory 561 effects caused by a stress response resulting from the isolation of the frogs for a 562 prolonged period of time in the glass box. Additional research is necessary to further 563 investigate these questions in A. femoralis. 564 565

Conclusions 566 567
Our study is one of the first to support the Challenge Hypothesis in a territorial 568 frog, by using STIs and a non-invasive technique to characterize androgen levels. We 569 found that water-borne androgen is responsive to social challenges in males of the 570 highly territorial poison frog, Allobates femoralis. Since water-borne hormones 571 provide biologically and physiologically relevant information by mirroring hormone 572 levels in plasma, the integration of territorial intrusion experiments and non-invasive 573 hormone sampling may allow researchers to test the "Challenge Hypothesis" in 574 animal systems with a broad suite of life histories. 575 576 Acknowledgements 577 We thank the Nouragues research field station (managed by CNRS) which benefits from 578 "Investissement d'Avenir" grants managed by Agence Nationale de la Recherche (AnaEE 579