Testosterone treatment unveils testosterone-insensitive song in an early-branched hummingbird

Birdsong often plays a crucial role in sexual behavior, including territorial defense and mate attraction. This behavior is particularly well studied in species in which only males sing and respond to gonadal testosterone depending on the breeding season. Although hummingbirds are long known to sing in a breeding context, the sensitivity of these vocalizations to testosterone in both sexes remains unknown. Therefore, we investigated (i) which vocalizations are emitted during their breeding season, (ii) the testosterone levels in free-ranging birds, and (iii) the effect of testosterone on these vocalizations in a semi-natural environment in both sexes of the black jacobin (Florisuga fusca), an early-branched hummingbird. Females and males of black jacobin produced the same set of vocalizations (high-pitched song and chirp call) at a similar rate during the breeding season, although testosterone levels in males were significantly higher. Testosterone treatment did not affect the vocal activity or acoustic parameters of their song in both sexes. However, testosterone maintained high levels of aggression in treated birds of both sexes and hyper-developed the oviduct of females. In contrast to almost all well-studied bird species studied so far, the vocal behavior of the black jacobins in their breeding season seems to be insensitive to high testosterone levels.


INTRODUCTION
Sexual male-typical traits, such as song in Northern Hemisphere songbirds are known to be activated during the reproductive periods with high levels of testosterone (reviewed in Wingfield & Farner 1978;Gahr 2014).These traits are important for courtship (mate attraction) and agonistic encounters (territorial defense) (Catchpole & Slater 1995).Evidence of testosterone-dependent vocal performance is nevertheless not exclusive to songbirds (oscine passerines) since it has also been found in other taxa, as in suboscine passerines (Kroodsma 1984) and parrots (Brockway 1968;Nespor et al. 1996).Crowing of roosters in domestic chicken (Gallus gallus) as well as of other galliform species, such as the Japanese quail (Coturnix japonica) and the grey partridge (Perdix perdix), is another typical vocal trait known to be testosterone dependent (Hamilton 1938;Marler et al. 1962;Andrew 1963;Beani et al. 2000).A particular vocalization type, the nest-coo used in the courtship of male ring doves (Streptopelia risoria) was induced by brain implantation of testosterone (Cohen & Cheng 1982).
High levels of testosterone appear to be necessary to activate sexually relevant vocal behaviors in Northern Hemisphere songbirds during a short breeding season (Wingfield & Farner 1978;Gahr 2020), whereas, in tropical songbirds such as the silver-beaked tanager (Ramphocelus carbo), during a long breeding season (between 6 and 7 months), the testosterone levels are only high at the end of the season coinciding with higher rates of dawn-song, but this species show a higher brain sensitivity to testosterone in the beginning of the season which indicates testosterone sensitivity of their dawn-song (Quispe et al. 2016).However, testosterone sensitivity of these vocal behaviors are less clear in tropical and sub-tropical non-songbird species with protracted breeding periods or year-round territoriality (Goymann et al. 2004).Such a bird family are the hummingbirds in which males and in some species females are known to sing (Baptista & Schuchmann 1990;Ficken et al. 2000;da Silva & Vielliard 2006;Ferreira et al. 2006).
Hummingbirds have neural song control areas analog to songbirds which are active during song production (Jarvis et al. 2000) and express androgen receptors (binding site of testosterone and its converted product 5α-dihydrotestosterone, DHT) (Gahr 2000).Androgen receptors were also found in hummingbird syrinx (A.Monte unpublished data), the sound-producing organ of birds.Nevertheless, it is unknown whether song is a male-typical trait or whether any vocalization is testosterone sensitive in hummingbirds, hence our motivation to study these aspects in the black jacobin (Florisuga fusca).It is a species of the Topazes clade (tribe Topazini) which is a sister taxon of all other extant hummingbirds according to the most recent evolutionary tree (McGuire et al. 2009(McGuire et al. , 2014)).Therefore, the study of the black jacobin may give insights into the ancestral condition of testosterone-sensitive vocalizations in the hummingbird family.
The black jacobin occurs in the Atlantic forest on the extent of the Brazilian coast until Uruguay and Argentina (Sick 1997;Schuchmann & Bonnan 2019).Black jacobins produce a high-pitched song composed of very fast frequency oscillations (Olson et al. 2018;Monte et al. 2020) and uttered in the breeding season (from September until March; Ruschi 1964); however, it is unknown whether this highpitched or other vocalizations produced in the breeding season are the exclusivities of males and whether they are testosterone sensitive.To investigate the sensitivity of their vocalizations to testosterone, firstly, we measured testosterone plasma levels and body weight of free-living black jacobins from both sexes in the middle of the breeding season (December).Secondly, we carried out an experiment with wild-caught birds placed in semi-natural aviaries.We recorded the vocalizations and associated behaviors before and after testosterone implants.We established their vocal repertoire and quantified their vocal output (song and call rate), as well as the acoustic parameters of the song.Additionally, we investigated the number of attacks as a proxy for aggressiveness and their feeding behavior.We found higher circulating testosterone levels and body weight in males than females, but no sex differences in their vocalizations.The testosterone treatment induced the hypertrophy of the oviduct but did not affect the testes weight.Both sexes maintained their aggressive behavior before and after testosterone implantation whereas the control group implanted with placebos reduced aggressiveness during the study period.Unexpectedly, we could not show any changes in their vocal behaviors following testosterone implantation.These results indicate that in contrast to songbirds and most other birds studied to date, the vocalizations of hummingbirds in their breeding season seem insensitive to testosterone.

Animals
This study was carried out in the park of the Museum of Biology Prof. Mello Leitão in Santa Teresa, Espírito Santo, Brazil in the middle of the breeding season that lasts from September until March (Ruschi 1964).Free-living black jacobins (Florisuga fusca) were captured between 6:00-8:00 and 17:00-18:00 during their regular visit to feeder bottles filled with sugary water (solution of water with 20% sugar), hanging inside a trap called "Ruschi net trap" developed and optimized to capture hummingbirds (Ruschi 2009).Black jacobin's plumage is not sexually dimorphic (Ruschi 1973), the birds captured for the testosterone experiment (n = 32) were individually marked with plastic-colored tags attached to their back (Kapoor 2012) and kept randomly in outdoor aviaries until molecular sexing.Upon sexing results, we kept 14 males and 10 females that were distributed in four adjacent outdoor aviaries (2.35 m long × 2.70 m high × 3.20 m deep; A1 to A4) under natural conditions of daylight, rain, temperature, and humidity.We placed six birds per aviary.Three males and three females in aviaries A1 and A4 and four males and two females in aviaries A2 and A3.The unequal number of females and males was due to two mistakes in the molecular sexing results that were only noticed at the end of the experiment when we checked the gonads.Additionally, we caught and released 206 wild birds for blood sampling (50 females and 153 males, 3 died) and captured and sacrificed further 14 birds (seven females and seven males) for future tissue analyses.
The aviaries contained vegetation to offer shadow, branches of different thicknesses to be used as perches, two hanging pots with flowers and thin perches, three feeder bottles with watery sugar, a wooden box containing fermented fruits to attract fruit flies (part of the diet of the black jacobins), and an irrigation sprinkler activated only between observations when the temperature was above 35 ºC to spread water in half of the aviary to be used for bathing.The feeder bottles were refilled once a day before the morning observations.At the end of the aviary experiment, the birds were killed by decapitation, as for the 14 extra caught birds, organs were immediately frozen in dry ice and stored at − 80 ºC until use.
All procedures of this study were conducted in conformity with the instruction no.03/2014 of the Chico Mendes Institute for Biodiversity Conservation (ICMBio) and with the permission of the Brazilian Institute of Environment and Renewable Natural (IBAMA) under the Biodiversity Information and Authorization System (SISBio) licenses numbers 30319-1, 41794-2 and 49097-1.The bird handling and housing procedures were conducted in accordance with Law no.11.794/ 2008, as well as with the rules issued by the National Council for Control of Animal Experimentation (CONCEA) and was approved by the Ethics Committee on Animal Use of the Federal University of Para (CEUA/UFPA) under the protocol number 5937130718 (ID 000953).

Experimental procedure and testosterone implants
The first 6 days of captivity were used as a time for the housed birds (n = 24) to acclimate to the aviaries (no recorded observations), and then we started monitoring observations for 1 week (days − 6 to − 1) before the testosterone (T) or empty implant treatment for baseline activity ("before treatment").On the 7th day (day 0), a total of seven females (aviaries A1, A3 and A4 with two each; A2 with one) and five males (aviaries A1, A3 and A4 with one each; A2 with two) were implanted with testosterone ("T-treated group") and the others with placebo empty implants ("control group") and monitored for the next 2 weeks (days 1 to 12, "after treatment", Fig. S1 in Supplemental Data).Each bird was observed for 20 min twice a day, in the morning (between 5:30, after sunrise, and 12:00) and the afternoon (between 12:00 and 18:30, before sunset; except days 8 and 12, in which birds were monitored only in the morning due to logistic problems).All the birds within an aviary were observed consecutively in a randomized order.Each of the two observation sessions per bird was conducted by a different observer (A.Monte or R. da Silva Matos) to minimize the observer bias.In total, we sampled 96 hr of observations before and 176 hr after T-treatment.
The implants were self-made with a 3 mm silastic tube (Dow Corning, USA; 1.47 mm inner diameter, and 1.96 mm outer diameter) filled with ca 0.9 mg of testosterone (Sigma T1500, Sigma-Aldrich GmbH, Germany) (T implant) or left empty (placebo implant) and the endings sealed with liquid silicone sealant (ASC 40064, USA).The implant size was estimated in proportion to the body size as it was reported in other species studied (Quispe et al. 2015;Prior et al. 2016).The implants were subcutaneously inserted at the lower back of the birds after local anesthesia with lidocaine hydrochloride (Lidogel 2%, Neoquimica).

Sound recording and analyses of vocalizations
In each session we monitored the vocal behavior, feeding behavior and aggressiveness in the four aviaries following a focal-bird sampling technique (Altmann 1974;Martin & Bateson 2007).
For sound recording, we used a digital recorder Marantz PMD 671 (USA) mounted with a shotgun directional microphone (Sennheiser MKH 70 P48, Germany; Rode NTG2, Australia).The recordings were obtained in the mono channel, with frequency sample of 48 kHz and 32 bits.The audio recordings were screened manually in the software Audacity (version 2.0.3,Audacity Team 2013), and the descriptions were transcribed into audio labels that were exported as text files.The labels were quantified by a customized algorithm written on R version 3.0.3(R Core Team 2014).
We identified three vocal types based on spectral parameters as previously described for the species (Olson et al. 2018;Monte et al. 2020): high-pitched (most common, 68.1% of emission in the entire monitoring), low-pitched (0.6%) and chirps (31.3%) (Fig. S2 in Supplemental Data).
The high-pitched vocalization has harmonics up to 80 kHz with a fundamental frequency between 10 and 14 kHz.The fundamental frequency was fast-modulated and had periodic oscillation called vibrato (Monte et al. 2020).It is the most spectrally complex of black jacobin's vocalizations and is suggested as the song of the species (Olson et al. 2018).The song is composed of one or several units (or syllable) emitted in sequence with each unit composed of three (triplet), two (double) or one (single) subunit (or notes) (Fig. S3 in Supplemental Data).Chirps can be emitted alone or followed by a few repetitions with variable silent interval (single chirp or s-chirp) or followed by many repetitions with progressively longer silence intervals (multiple chirps or m-chirp) (Fig. S2 in Supplemental Data) as previously described for rufous-breasted hermit (Glaucis hirsutus) (Ferreira et al. 2006).Given that m-chirp was rare, we quantified the number of chirps in the m-and s-chirps and presented them together in the results.The low pitch was also a very rare event and was not analyzed.
We quantified the number of songs and chirps emitted per bird in each 20 min observation session (songs or chirps/20 min).We subsampled the longest song bouts with a good signalto-noise ratio and selected ~ 200 songs per bird (range: 5-1000 song bouts per bird, per session).We labeled the triplets, doubles and singles manually in the software Audacity (version 2.0.3,Audacity Team 2013) and saved the labels as text files.We quantified their repeatability (R) which represents how much of the total variance in the number of triplets, doubles and singles is attributed among birds and was calculated with the function "rpt" implemented by the package "rptR" (Stoffel et al. 2017) and using a customized algorithm written in R version 3.0.3(R Core Team 2014).For the analysis of the spectral and temporal acoustic parameters (in total 26, Table S1 in Supplemental Data), we used the function "specam" and package "warbleR" (Araya-Salas & Smith-Vidaurre 2017).We analyzed in total 3660 songs.We calculated the averages of each parameter per bird before, 1 and 2 weeks after T-treatment followed by a Principal Component Analysis (PCA) to reduce the number of parameters and maintain those with higher variance (see Statistical analyses).

Quantification of other associate behavioral observations
We quantified the number of attacks per focal bird per session (a proxy for aggressiveness), that is when the focal bird chased another bird away or pecked the bird with the bill (attacks/20 min).We also quantified the feeding events per session as the number of times the focal bird drank sugary water from the feeder or caught a fruit fly (a proxy for metabolism).These feeding counts were also a proxy for the well-being of the birds as lethargy and anorexia are signs of sickness (Hart 1988).Those two behaviors have been previously documented to be influenced by testosterone in other birds, especially in males (Wingfield et al. 2001;Soma 2006).

Blood sampling
For sex determination, a droplet of blood was collected from each bird with a heparinized capillary tube after the section of the hallux's nail as described by Ruschi (1950) diluted in Queens Lysis Buffer (0.01 M EDTA; 0.01 M Tris; 0.01 M NaCl; 1% n-Lauroylsarcosine and pH 8.0) and stored in the fridge (~ 5 ºC).For the measurement of the testosterone plasma levels, blood (mean µL, 30-80 µL) was collected from the jugular vein as described by Hoysak and Weatherhead (1991) otherwise from the trunk after decapitation.A droplet of blood was diluted in Queens Buffer for later sexing and for the rest of the blood, the plasma was separated by centrifugation (10 min at 6000 rpm) and stored in dry ice (~ − 80 ºC).

Testosterone level measurements
In total, we measured the plasma testosterone levels of 241 black jacobins.The samples for this profile were obtained within 10 consecutive days in December, during the middle of the breeding season (Ruschi 1964) for 57 females (52 adults and five juveniles) and 160 males (145 adults and 15 juveniles).First-year juveniles could be identified by differential plumage (Ruschi 1973).Since there was no significant difference between 1st-year juveniles and adults (Table S2 in Supplemental Data), they were pooled together.
The testosterone plasma levels were measured by radioimmunoassay as described by Goymann et al. (2006).In total, three assays were realized.The samples from the testosteronetreated birds were assayed 3 times, and the final testosterone value was obtained by the simple mean.These triplicates were conducted in different batches with consistent outcomes; thus, it is unlikely that the values obtained for testosterone-treated birds are an artifact.The lower detection limits were between 0.33 and 0.4 pg/mL with all samples above the detection limit.The extraction recovery of testosterone per assay was respectively 89, 90 and 87%, and the intra-assay coefficients of variation 7, 5 and 12.2%.The inter-assay coefficient of variation was 3.2%.

Statistical analyses
The statistical analyses were performed using R version 3.0.3(R Core Team 2014).All data and code are available in the online repository and accessible in this link (https://osf.io/9fs5w/).
The testosterone plasma levels (T-levels) were log 10 transformed.For wild black jacobins (57 females and 160 males), to explore the effect of sex, body weight, age and day on T-levels, and effect of sex, T-levels, age and day on body weight, we used linear models.First, we fitted models including age (1st-year juveniles × adults) which had no effect both on T-levels (estimate = 0.22 ± 0.20, t = 1.13,P = 0.262, Table S2 in Supplemental Data) and body weight (estimate = 0.18 ± 0.11, t = 1.60,P = 0.112, Table S2 in Supplemental Data).Then we fitted models without age, and we plotted 1st-year juveniles and adults together.For captive black jacobins (10 females and 14 males), to compare T-levels among sex and experimental groups (control and T-treated), we used Wilcoxon rank sum exact test (non-parametric).To compare control to wild birds, we randomly subsampled 10 T-levels values of each sex from the wild birds.
To compare gonadal and brain weight, we used Kruskal-Wallis rank sum test (nonparametric) to compare wild, control and T-treated.We used non-parametric statistics in all cases in which at least one of the groups (T-treated females) had only three data points.In these cases, we reported the median interquartile range (IQR) [25%, 75% quartiles].Following a significant Kruskal-Wallis test, we performed Dunn's Kruskal-Wallis multiple comparisons as post hoc analysis with the function "dunnTest" implemented by the package "FSA" (Ogle et al. 2020).To compare body weight, we weighed the aviaries birds every time they were handled: when they were housed (before T), on the day of the implantation (T) and at the end of the experiment (after T).Then, we used repeated measures one-way ANOVA (parametric because all the assumptions were met).In the results, we reported the mean (± SD).Following a significant ANOVA, we performed Tukey multiple comparisons of means as post hoc analysis.In both post hoc analyses, we adjusted the P-values with the Bonferroni method.Given that the body weight of males had a wide variance, we further performed the F test to compare two variances (control and T-treated) with the "var.test"function.The significance level for all tests was α = 0.01.
To test whether the testosterone treatment (T-treatment) affected the number of songs and chirps emitted by the black jacobins, their feeding behavior and aggressiveness in the aviaries, we used generalized linear mixed models (GLMMs).In all models, to account for the non-independence of the samples, we included "bird id" as a random effect.To determine the best fit model for each response variable (song, chirp, feeds or attack counts), we fitted models with all possible combinations of the predictors that were biologically meaningful for the respective response variable.Then, we compared the models using the differential Akaike's information criterion (dAIC) obtained by the function "AICtab" implemented by the "bbmle" package (Bolker & R Development Core Team 2020) and selected the most parsimonious models based on the lowest AIC (Burnham & Anderson 2016).We accessed the statistical significance of the best-fitting models' predictors via asymptotic Wald's ztests (P < 0.01).In the results, we reported the estimated mean 95% CI [lower, upper confidence interval] for song and chirp counts models and the means of the simulated values from the posterior distributions and the predicted 95% credible intervals (CrI) [lower, upper credible interval] for feeding and attack counts models.Details about the fitting of the GLMMs the assessment of their assumptions and post-hoc analysis are provided in the Supplemental Data.
We performed a Principal Component Analysis to reduce song parameters using the function "prcomp" implemented by the package "car" (Fox & Weisberg 2018).We plotted the first two Principal Components (PC1 and PC2) using a customized version of the function "ggbiplot" implemented by the package "ggbiplot" (Vu 2011).We inspected visually the PC1 and PC2 plot to identify whether the birds were grouped by treatment and/or sex at the different time points (before, 1 and 2 weeks after T-treatment).We further investigated one of the variables that were heavily loaded within the PC1 vector (mean dominant frequency of the song) and one within the PC2 vector (song duration).We compared the averages of the dominant frequency and song duration for each of these four groups: control females, T-treated females, control males and T-treated males at the time points: before, 1 and 2 weeks after T-treatment using the Friedman Rank Sum Test which is an ideal statistic to use for a repeated measures type of experiment to determine if a particular factor has an effect (Scheff 2016).We also calculated the repeatability (R) using the function "rpt" implemented by the package "rptR" (Stoffel et al. 2017).

Testosterone plasma levels and body weight differ between sexes in free-living birds during the breeding season as well as in captivity
Adult males had higher testosterone levels [estimated mean 95% CrI (lower, upper credible interval); n = 160 birds, 102.19 pg/mL (90.24, 117.18)] than adult females [n = 57, 54.39 pg/mL (44.12; 6838)] (estimate = 0.52 ± 0.14, t 213 = 3.82, P < 0.001; Fig. 1A, left Black jacobins sing a testosterone-insensitive song panel) and had a higher body weight [7.85 g (7.74, 7.96)] (estimate = 0.26 ± 0.08, t 213 = 3.33, P = 0.001; Fig. 1A, right panel) than adult females [7.47 g (7.33, 7.60)] (Table 1).The model for T-levels indicated a strong effect of body weight (estimate = 0.34 ± 0.12, t 213 = 2.90, P = 0.004; Table 1).Thus, to verify whether the sex difference found in T-levels is not a side effect of the differences in body weight, we selected males (n = 115), which fall into the weight range of the females (female weight mean ± SD) and compared T-values again.The difference remained and females still had lower T-levels than males (t = − 3.40, df = 142.27,P = 0.001).The model for T-levels also indicated a strong effect of sampling time (estimate = − 0.29 ± 0.02, t 213 = − 13.79, P < 0.001; Table 1).This temporal correlation may reflect a decline in the baseline testosterone in the second half of their breeding season.Furthermore, nearly 7% of the males (n = 11) showed T-levels above 1000 pg/mL and 30% of the females (n = 17) showed T-levels above 100 pg/mL which is expected for birds in breeding condition in the tropics (e.g.Goymann et al. 2004;Goymann & Wingfield 2014).
T-levels of captive (aviaries) control birds were inside the range of the free-living birds in both females (W = 26, P = 0.077) and males (W = 44, P = 0.968); therefore, captivity did not affect T-levels.T-levels of T-treated birds were higher than controls in both females (W = 0, P < 0.017) and males (W = 0, P = 0.001): 2.5k-fold above wild average for females and 300-fold for males (Table 2, Fig. 1B).

Captivity, but not testosterone treatment decreased body weight only in females
Captivity decreased the weight of females, but not that of males.The body weight of females changed with time [F (2,18) = 6.47,P = 0.008], but not between treatment groups (W = 16, P = 0.267).Before T application, females' weight did not differ between the control (mean ± SD, 7.1 g ± 0.3) and T-treated (7.2 g ± 0.3) groups.At the T-treatment day, weight of all females increased slightly, but not significantly compared to before (P = 0.723) (control, 7.3 g ± 0.1; T-treated, 7.3 g ± 0.4).Two weeks after the testosterone treatment (after T) weight decreased significantly compared to baseline (P = 0.041) and to T day (P = 0.008) (control, 7 g ± 0.4; T-treated 6.8 g ± 0.4) (Fig. 1C, upper panel).This effect was not observed in males, as their weight did not change in time [F (2,26) = 1.36,P = 0.276] and not between groups [F (8,4) = 0.21, P = 0.058]: Before T: control group 7.7 ± 0.4 g and T-treated group 7.6 ± 0.4 g.On the T-treatment day: control group 7.9 ± 0.4 g and T-treated group 8 g ± 0.6; after T: control group 7.9 ± 0.5 g and T-treated group 7.7 ± 1 g (Fig. 1C, lower panel).

Testosterone treatment increased oviduct weight but did not affect testes weight
The oviduct weight differed between groups (H = 6.56, df = 2, P = 0.038): wild (n = 6), captivity control (n = 3), and captivity T-treated (n = 7) (see Table 3 for medians).Ovary weight was lower in control captive birds than in the wild (P = 0.021) and T-treated females (P = 0.017), but this was not significant after Bonferroni correction (P adj.= 0.063 and 0.051, respectively) (Fig. 2A, left panel).However, the oviduct weight differed between groups (H = 11.21,df = 2, P = 0.004) with heavier oviducts in T-treated (by five and six-fold, respectively) compared to the wild (P adj.= 0.011) and to control (P adj.= 0.025) (Fig. 2A, right panel; Table 3).The testis weight was similar between groups [F (2,18) = 2.14, P = 0.147] (Fig. 2C, Table 3).Results of linear models explaining variation in the circulating testosterone plasma levels (T plasma levels) and body weight of wild black jacobins.In the wild, T plasma levels were affected by sex, body weight, and time, whereas body weight was affected by sex and T plasma levels.In bold are the significant predictors (P < 0.01).
Results of the generalized linear mixed model explaining variation in the number of songs emitted per observation session (20 min) of captive black jacobins.For a better fit, the hour of the day was transformed into linear, quadric and cubic terms.The birds tend to emit fewer songs throughout the day and males sing less than females 2 weeks after the testosterone treatment.In bold are the significant predictors (P < 0.01).5, Fig. 3D).When the birds were not emitting chirps, they were feeding in the feeders (zero-inflated effects, estimate = − 1.14 ± 0.34, z = − 3.34, P < 0.01).T-treatment and sex jointly had an effect closer to a significance for T-treated males in relation to controls and females (fixed effects interaction between T-treatment and sex, estimate = − 2.12 ± 0.87, z = − 2.44, P = 0.015).The means estimated from the best-fitting model for chirp counts show that T-treated females produced 15.67 chirps (7.26, 33.81) [estimated mean (lower, upper 95% CI)] before, 7.19 chirps (3.25, 15.9) 1 week after T and 12.73 chirps (6.02, 26.93) 2 weeks after T (Fig. 3D, Table S5 in Supplemental Data).The dispersion of the data explained by the model was greater in aviary A1 in relation to aviary A2 (dispersion effects, estimate = − 1.48 ± 0.25, z = − 5.84, P < 0.01), A3 (dispersion effects, estimate = − 0.68 ± 0.25, z = − 2.69, P < 0.01) and A4 (dispersion effects, estimate = − 1.29 ± 0.25, z = − 5.13, P < 0.01).

Testosterone treatment did not affect the acoustic parameters of the song of males and females
Black jacobin song can be composed of one (single), two (double) or three (triplets) subunits (Fig. S3A in Supplemental Data).The proportion of singles, doubles and triplets was stable before, 1 week and 2 weeks after T-treatment, because these proportions showed high repeatability within birds (R = 0.46 ± 0.11, P < 0.01).Singles were rare, most of the birds emitted triplets and three males emitted mostly only doubles (Fig. S3B in Supplemental Data).
Concerning the acoustic parameters of the song, a Principal Components Analysis (PCA) reduction revealed the first three eigenvalues to contain ~ 75% of the overall variation.PC1 (~ 36% of the variation) described mainly spectral parameters (such as frequency mean, frequency standard deviation, median, lower (25%) and upper (75%) frequency quantiles, spectral entropy, spectral flatness, dominant frequency, minimum dominant frequency, dominant frequency range, start dominant frequency, as well as peak frequency mean); PC2 (~ 23% of the variation) described temporal parameters [duration, median time, upper (75%) quantile and time interquartile range as well as time entropy] and PC3 (~ 16% of the variation) described skewness and kurtosis (Tables S6-S7 in Supplemental Data).PC1 and PC2 did not separate control from T-treated birds neither in females nor in males, although showing a slight separation between sexes (Fig. 4A).

Testosterone treatment maintained high aggressiveness in males in comparison with a decrease in placebo birds
We counted the number of attacks (chasing other birds or beak pecking on another bird) as a proxy for aggressiveness.In all observations and regardless of sexes, the observed birds attacked another bird twice or more in 55% of the observations, once in 37% and did not attack in 8%, which indicates the intrinsic high aggressiveness of the black jacobins.The number of attacks was weakly associated with song counts (r = 0.16, P < 0.001).
Black jacobins sing a testosterone-insensitive song  S1 in Supplemental Data for a list of the parameters).The first component PC1 explained 36% of the overall variation and described mainly spectral parameters such as dominant frequency.The second component PC2 explained 23% of the variation and described temporal parameters such as duration.Each data point represents the mean of the PCA loads per bird in three time points: before, 1, and 2 weeks after T-treatment.If T-treatment had affected and differentiated the acoustic parameters of the song, we expected to see a clear separation of the PCA points of T-treated birds (dashed circle) after the T-treatment in comparison to the controls (solid circle).Instead, the biplot shows a complete overlap between T-treated and controls in both sexes indicating that there are no clear sex or treatment differences in the acoustic parameters of the song before, 1 or 2 weeks after the T-treatment.Nevertheless, a small separation between sexes can be observed, because some males had consistently more positive loadings along PC1 axis and some females more negative reflecting a trend of male's song toward higher pitch independent of treatment or time.(B) The dominant frequency (left panel) and the duration (right panel) of black jacobin's song in both sexes of control and T-treated groups did not differ before, 1 or 2 weeks after T-treatment.Males consistently tended to sing higher-pitched songs than females, whereas females consistently tended to sing longer songs than males, but the differences were not significant.All box plots indicate the median and interquartile range (25 to 75%).Each data point indicates an individual song measurement (see Table 3 for details on sample size).FC, female control; FT, female T-treated; MC male control and MT, male T-treated.
The number of times the birds feed on the feeders in the aviary was not affected by T-treatment (estimate = − 0.06 ± 0.20, z = − 0.30, P = 0.761), nor by the time they spent in captivity 1 (estimate = − 0.27 ± 0.14, z = − 1.93, P = 0.054) or 2 weeks after T-treatment (estimate = 0.14 ± 0.14, z = 1.06,P = 0.290) in relation to the baseline before T (Fig. S5, Table S3 in Supplemental Data).However, T-treated males and females tended to catch more fruit flies 2 weeks after T-treatment.The best-fitting model for fruit-fly captures indicated that the number of captures was affected by the interaction of T-treatment and the time they spent in captivity after 2 weeks (estimate = 1.57± 0.43, z = 3.68, P < 0.001) and also by the number of feeding from the feeders (estimate = 0.27 ± 0.05, z = 5.39, P < 0.001, Table S3 in Supplemental Data).A post hoc analysis of interactions indicated that the control birds of both sexes captured the same number of fruit flies across time (females χ 2 = 7.34, df = 2, P = 0.102 and males χ 2 = 4.86, df = 2, P = 0.353), whereas T-treated birds tended to capture more fruit-flies as longer they spent in captivity, but this difference was non-significative (females χ 2 = 9.86, df = 2, P = 0.029 and males χ 2 = 9.71, df = 2, P = 0.031) (Fig. S5 in Supplemental Data).

DISCUSSION
This study presents the first profile of testosterone plasma levels (T-levels) and the effects of testosterone treatment (T-treatment) on the vocal behavior in females and males of a hummingbird species, the black jacobin.Although T-levels of black Table 6.
Results of the generalized linear mixed model explaining variation in the number of attacks per observation session (20 min) of captive black jacobins.For a better fit, all continuous predictors were mean-centered (z-scored).The number of attacks was affected by time 2 weeks in relation to before treatment, temperature and number of feeds on the feeder.In bold are the significant predictors (P < 0.01).jacobins in the wild were higher in males, vocal behaviors of control and testosterone treated males and females did not differ.Therefore, the vocal behavior of these hummingbirds seems insensitive to testosterone in contrast to species of other avian families in which testosterone either affects vocal output or vocal pattern or both (Gahr 2014(Gahr , 2020)).Note that T-levels tended to get lower in both sexes towards the end of the breeding season (March compared to December), indicating that they can change during the breeding season or/and throughout the year.In relation, 7% of the males had T-levels comparable to reproductively active males of other species in the tropics (Goymann et al. 2004), which, too, indicates that T-levels vary during the breeding season.The large individual differences of plasma T-levels are likely due to the long breeding period of tropical species, which leads to de-synchronized reproductive activities within populations.

Estimate
The T-levels of wild populations of hummingbirds were so far documented in only a few species given the methodological constraints to obtaining blood samples of such small birds.Previously, T-levels of hummingbirds were estimated from measurements of testosterone concentrations in the cloacal fluid of three species, all of the clade Bees, during the breeding season.In the black-chinned hummingbird (Archilochus alexandri), males had higher T-levels than females, whereas in Anna's hummingbird (Calypte anna) and the rufous hummingbird (Selasphorus rufus) males and females had similar T-levels (González-Gómez et al. 2014).Surprisingly, although these species weigh about half of black jacobin's body weight, their average T-levels were about 90-fold higher than that of black jacobins, their T-levels ranged between ~ 8 and ~ 23 ng/mL (González-Gómez et al. 2014), whereas the T-levels of wild black jacobins ranged (25 to 75% quantiles) from ~ 0.03 to 0.1 in females and ~ 0.05 to 0.2 ng/mL in males (Table 2).However, a fraction of males and females had testosterone levels typical of reproductively active male and female birds of the temperate zones with T-levels above 1 ng/mL plasma for males and 100 pg/mL plasma for females.This difference may reflect (i) the distant relationship of black jacobins (clade Topazes) and the Bees in the hummingbird family; (ii) a sampling methodology difference as T-levels quantified from the cloacal fluid can be distorted due to individual differences caused by the metabolism of hormones by gut bacteria and by matrix effects (Goymann 2012) or, more likely, (iii) methodological problems in measuring testosterone levels from cloacal fluid; such high T-levels are only found in birds after testosterone treatment (Goymann & Wingfield 2014; this study).
The T-levels of the wild black jacobin were collected in the middle of their 7 months long breeding season (Ruschi 1964), although there was no information about the exact life stage such as incubating or having off-springs of the sampled birds, the high T-levels of 7% of wild males indicate their reproductive status.Interestingly, if we calculate the median of the plasma T-levels by considering only wild males with values above 1 ng/mL (n = 11), the outcome is 1.8 ng/mL (1.4,3.3) which is rather low when compared to average plasma T-levels of northern temperate birds in the wild during the breeding season (2.8 ± 0.4 ng/mL), but on the range of tropical birds (1.3 ± 0.2 ng/ mL) (Goymann et al. 2004).For tropical birds, it is suggested that their maxima T-levels is mainly determined by the length of the breeding season and the altitude they occur: species with longer breeding seasons (above 4 months) tends to show relatively lower T-levels and species inhabiting higher altitudes, higher values (Goymann et al. 2004).Therefore, the black jacobin as a species with long breeding season that occupy relatively lower altitudes is expected to have the observed T maxima.
The T-values of the T-treated black jacobins at the end of the experiment was about 300 times higher than control in males but 33 times higher than wild males with highest T-levels (> 1 ng/mL, n = 11) and 2,500 times higher than control in females but 350 times higher than wild females with highest T-levels (> 0.1 ng/mL, n = 17).In relation, T-values of intact breeding males of zebra finches after T implantation that are ~ 4 times higher than controls has been shown to induce singing and reduce time spent in close proximity to their partner (Prior et al. 2016).In castrated males of canaries, T-values 2 days after implantation can be ~ 10 times higher than controls and induce higher singing rates (Shevchouk et al. 2019).In intact and photostimulated males of house sparrow (Passer domestics), a dose-dependent effect was observed: after low doses, T-values were up to 2 times higher than control and induced testes atrophy, whereas after high doses, T-values ranged from 2 to 10 times higher while spermatogenic activity was maintained (Turek et al. 1976).Despite the much higher circulating T-levels in both sexes of T-treated black jacobins, neither gonadal, body and brain weights nor vocal behavior were affected; testosterone only affected oviduct growth in females and the maintenance of high aggressiveness in both sexes.
It is important to keep in mind that, in some species, T-treatment induce singing in castrated or intact birds in non-breeding condition but has no effect when administered in intact birds in breeding condition because T-levels reached a certain physiological threshold (reviewed in Fusani 2008;Goymann & Wingfield 2014).In ring doves (Streptopelia risoria), for example, castrated males vocalize more after T-treatment (Cohen & Cheng 1982); however, in intact males in breeding conditions both low and high dosage T-implants did not affect any element of their courtship activity including the sexually relevant vocalizations (Fusani & Hutchison 2003).
Although T-treatment was conducted in intact and potentially breeding birds, this is unlikely the case for the black jacobin because T-treatment modulated aggressiveness and induced oviduct atrophy indicating that there was not a generalized lack of effect due to the supraphysiological T-levels in treated birds.Furthermore, a sex difference was observed in the T-levels of wild birds but not on vocal rates which reinforce the hypothesis that none of these vocalizations are a T-sensitive trait.T-implantation in castrated black jacobins would require developing surgical procedures and is, therefore, difficult to perform.Nevertheless, implants with anti-androgens in birds whose life-stages are known may help to further confirm our findings.The quantification of testosterone metabolites DHT (binds to androgen receptors) and estradiol (binds to estrogen receptors) also remains to be seen.In chickens, testosterone and estrogen act synergically on the preparation of the magnum -part of the oviduct -for normal secretion of albumen into the egg; when administered alone, testosterone increased magnum weight, but did not induce the formation of albumin secreting cells (Yu & Marquardt 1973).In black jacobin females, the presence of androgen receptors in the part of the oviduct, that is histologically similar to the magnum, may explain the testosterone-induced oviduct growth (A.Monte unpublished data).Testosterone treatment did not affect the weight of the testis of male black jacobins.This result is in line with some other species in which high T-levels did not decrease testis weight (Turek et al. 1976;Desjardins & Turek 1977) while in other species elevated T-levels regressed the testis (Chu 1940).
Hummingbirds are aggressive year-round, especially while foraging (Greenewalt 1990;Schuchmann 1999).Aggressiveness levels (number of attacks) of T-treated black jacobins were maintained during the whole experiment compared to a reduction in control birds over time in captivity.In seasonal species, aggressiveness is correlated with T-levels (Wingfield et al. 1987;reviewed in Soma 2006).To cope with the costs of high T-levels, these birds raise T-levels, particularly in periods of social instability during territorial challenges.Conversely, when their status or territory boundaries are maintained (social stability), T-levels are kept low (Wingfield et al. 1990).In the aviaries, the reduction in the number of attacks of control black jacobins across time may be explained by social stability; there was one feeder for every two birds, thus some birds might have alternated instead of disputed the same feeder.A similar use of the feeders was also observed in wild black jacobins, they awaited a free feeder instead of competing for feeders (A.Leitão personal observation).
We observed no effects of the T-treatment on the vocal rates and the acoustic parameters of the song, in contrast with other bird species investigated to date (reviewed in Gahr 2014Gahr , 2020)).In the brain, testosterone can act on the medial preoptic nucleus (POM) or directly in the vocal control system (reviewed in Gahr 2014; Shevchouk et al. 2019).For example, direct testosterone implantation in POM of castrated male canaries induced high vocal rates but increased either vocal rate and song stereotypy when administered concomitantly in POM and HVC, an important nucleus of the vocal control system (reviewed in Alward et al. 2017).In the POM and song control regions of songbirds, there are androgen receptors (AR) and estrogen receptors that mediate the activity of testosterone and its androgenic and estrogenic metabolites on the cellular level (reviewed in Ball & Balthazart 2007).Similar, AR is expressed in the putative song control region of two hummingbird species, Anna's hummingbird (Calypte anna) and amazilia hummingbird (Amazilia amazilia) (Gahr 2000).AR and estrogen receptors in the POM are a general feature of all birds (reviewed in Gahr 2020).Although the reason for life-stage dependent effects of testosterone is unknown, we can speculate that the number of AR (Fusani et al. 2000;Quispe et al. 2016) or of testosterone metabolizing enzymes such as 5ßreductase varies seasonally; birds reduce testosterone to 5ß-dhihydrotestosterone that cannot activate the AR (Floch et al. 1985).Likewise, we could assume that AR is expressed at low levels or 5ß-reductase is expressed at high levels in the putative song control brain regions of the black jacobin so that testosterone cannot affect song rate and pattern.Further, although we quantified AR in the syrinx of the black jacobin, these AR were not in muscles but in skeletal parts of the syrinx (A.Monte unpublished data), which makes an activating effect of testosterone on the syrinx and song unlikely.The only finding that indicates hormonal regulation of the song of the black jacobin is a previous report suggesting a seasonal change in the entropy of the song (Olson et al. 2018).However, these birds were not ringed and were recorded in the wild at the beginning (September) and end of the breeding season (March) (Olson et al. 2018).Thus, this seasonal change could be due to the fact that more juvenile birds were recorded at the end of the breeding season rather than an effect of testosterone.Entropy measurements of the song of our birds were not affected by testosterone (F r = 5, df = 3, P = 0.172, Fig. S6 in Supplemental Data).
The insensitivity of vocalization rates and song features to testosterone in both sexes combined with our observations of the behavior of the birds indicates that none of these vocalizations are used exclusively in a sexual context.Chirps were mainly produced in flight shortly before feeding, whereas song was mainly produced when perched or sometimes in flight when chasing another bird (Fig. S7 in Supplemental Data).Black jacobins were observed to use their songs outside of the breeding season for defending food resources (Olson et al. 2018; A. Monte personal observation).Hence, it could be conceivably hypothesized that year-round song types that are also used in non-sexual contexts are not sensitive to testosterone.Thus, the American crow (Corvus brachyrhynchos), for example, which uses its song all year long in multiple contexts to facilitate affiliative behaviors among group members (Brown 1985) is not expected to show a correlation between T-levels in contrast with species that use their song as a secondary sexual trait during breeding season in which song is modulated by testosterone as in the common chaffinch (Fringilla coelebs) (Thorpe 1958) or canaries (Fusani et al. 2000).Further studies on the seasonality of black jacobin song and the behavioral context of their emission are, nevertheless, needed to support this hypothesis.
Hummingbirds have the highest metabolic rates amongst homeotherm vertebrates (Suarez 1992), which may constrain their daily activities towards finding and defending nectar-rich food sources.Thus, we speculate that, in black jacobin, the song is used by both sexes primarily to defend food sources.Although it remains to be seen whether the testosterone-insensitivity of black jacobin's vocalizations is life-stage dependent, it is possible that the non-reproductive use of song implied the loss of the activation effect of testosterone in their vocal control system.The courtship display of the black jacobin is a joint flight of both sexes in which males chase the female around 100 m upward in a stepwise fashion followed by a synchronously straight dive (Ruschi 1949).Thus, we speculate that, in black jacobin, mate choice is primarily based on visual displays and visual sensory processing.Thus, it needs to be seen if the non-reproductive use of song implied the loss of the activation effect of testosterone in their vocal control system, and in turn, the evolution of testosterone-sensitive visual displays and visual processing.

Fig
Fig. 1. -Circulating levels of plasma testosterone (T plasma levels) and body weight of wild and captive black jacobins.All T plasma levels are represented on a logarithmic scale (Log10).(A) Wild black jacobin is sexually dimorphic in the T plasma levels (left panel) and body weight (right panel).The outlines of the violin plot represent the proportion of the data located in each value given by kernel probability density.The data points represent individual measurements of females (n = 57, lighter points) and males (n = 160, darker points).The error bars in gray represent the fitted mean and the 95% credible interval given by the linear models.(B) Captive black jacobins under T implantation had T plasma levels increased 2000-fold for females and 300-fold for males in relation to those of the same sex that received placebo implants (controls).The error bars on the left indicate the median and interquartile range (25 to 75%) of the wild black jacobins as a reference to the physiological range.The box plots indicate the median and interquartile range (25 to 75%) of the captive black jacobins.The data points represent individual measurements of females (control, n = 3 and T-treated, n = 7; lighter points) and of males (control, n = 9 and T-treated, n = 5; darker points).(C) Females lost weight in captivity independently of T-treatment.The body weight of the captive black jacobins was obtained at three time points: on the day there were placed in the aviaries (before T), on the day of the T implantation (T), and at the end of the experiment (after T).Although T-treated black jacobins presented T levels above the physiological range, neither females nor males had their body weight affected by the treatment, instead, females were affected by the captive condition.

Fig. 2 .
Fig. 2. -Although both sexes of black jacobin have androgen receptors in their gonads, the testosterone treatment (T-treatment) affected only the gonads of the females.(A) The ovary weight (left panel) of wild and T-treated females does not differ, although the ovary of control females is slighter lighter.The oviduct weight (right panel) of T-treated females is significantly heavier than wild and control females.The data points represent individual measurements of females (wild, n = 6; control, n = 3 and T-treated, n = 7).(B) The brain weight of both sexes wild, control and T-treated black jacobins did not differ.The data points represent individual measurements of females (wild, n = 7; control, n = 3 and T-treated, n = 7; lighter dots) and males (wild, n = 7; control, n = 9 and T-treated, n = 5; darker dots).(C) The testes weight of wild, control and T-treated males also did not differ (wild, n = 7; control, n = 9 and T-treated, n = 5).All box plots indicate the median and interquartile range (25 to 75%).

Fig. 3 .
Fig. 3. -Testosterone treatment (T-treatment) did not affect the number of songs and chirps emitted by both sexes in black jacobin.(A) Spectrogram (frequency with time) and oscillogram (relative amplitude with time) of the black jacobin's song and (C) chirp.The song is composed of a single high-pitched syllable (vocal unit) that can be repeated multiple times in a song bout.The chirp is characterized by a wide frequency bandwidth in a fast duration that can also be repeated consecutively in a chirp bout.(B) The number of songs produced by both sexes in control and T-treated groups did not differ before, 1 or 2 weeks after T-treatment.(D) Although T-treated females tended to consistently produce more chirps, the number of chirps produced by both sexes in control and T-treated groups also did not differ significantly before, 1 or 2 weeks after T-treatment.The data points represent song or chirp counts obtained per bird in 20 min of observation.The error bars represent the fitted mean and the 95% confidence interval given by the generalized linear mixed models.

Fig. 4 .
Fig. 4. -Testosterone treatment (T-treatment) did not affect the acoustic parameters of the song in both sexes of black jacobin.(A) Biplot including the first two components (PC1 versus PC2) of a Principal Components Analysis including 26 acoustic parameters of black jacobin's song (see TableS1in Supplemental Data for a list of the parameters).The first component PC1 explained 36% of the overall variation and described mainly spectral parameters such as dominant frequency.The second component PC2 explained 23% of the variation and described temporal parameters such as duration.Each data point represents the mean of the PCA loads per bird in three time points: before, 1, and 2 weeks after T-treatment.If T-treatment had affected and differentiated the acoustic parameters of the song, we expected to see a clear separation of the PCA points of T-treated birds (dashed circle) after the T-treatment in comparison to the controls (solid circle).Instead, the biplot shows a complete overlap between T-treated and controls in both sexes indicating that there are no clear sex or treatment differences in the acoustic parameters of the song before, 1 or 2 weeks after the T-treatment.Nevertheless, a small separation between sexes can be observed, because some males had consistently more positive loadings along PC1 axis and some females more negative reflecting a trend of male's song toward higher pitch independent of treatment or time.(B) The dominant frequency (left panel) and the duration (right panel) of black jacobin's song in both sexes of control and T-treated groups did not differ before, 1 or 2 weeks after T-treatment.Males consistently tended to sing higher-pitched songs than females, whereas females consistently tended to sing longer songs than males, but the differences were not significant.All box plots indicate the median and interquartile range (25 to 75%).Each data point indicates an individual song measurement (see Table3for details on sample size).FC, female control; FT, female T-treated; MC male control and MT, male T-treated.

Table 2 .
Median and quartiles of the testosterone plasma levels for wild and captive black jacobins grouped by sex and treatment: birds treated with placebo (control) and birds treated with testosterone implants (T-treated).

Table 3 .
Median and quartiles of testes, ovaries, and oviduct weights for wild and captive black jacobins grouped by treatment: birds treated with placebo (control) and birds treated with testosterone implants (T-treated).

Table 5 .
Results of the generalized linear mixed model explaining variation in the number of chirps emitted per observation session (20 min) of captive black jacobins.The birds tend to emit fewer chirps across time.In bold are the significant predictors (P < 0.01).