The effect of chronic and acute stressors, and their interaction, on testes function: an experimental test during testicular recrudescence

It is well known that stressors can have a profound negative effect on reproductive physiology and behavior (Chand and Lovejoy, 2011; Rivier and Rivest, 1991; Selye, 1946). This effect has been demonstrated in relation to a variety of stressors, including food limitation (Lynn et al., 2015), thermal stress (Hansen, 2009) and psychological stress (McGrady, 1984; Nargund, 2015). The main adaptive hypothesis for this suppressive effect states that reproduction is inhibited because the physiological, energetic and behavioral components of reproduction are costly and may directly interfere with resource and time allocation to a stress response (Breuner et al., 2008). Because the stress response is an integral part of self-maintenance and survival, the interaction between stress and reproduction constitutes an important aspect of the life history trade-off between current and future reproduction (Wingfield and Sapolsky, 2003).

Life history theory predicts that the effect of stress on reproduction should depend on the costs and benefits of both of these functions (Wingfield and Sapolsky, 2003). For example, if future reproductive success (residual reproductive value) is expected to be low, succeeding at the current reproductive effort is crucial to the fitness of the organism. It is expected that, under these conditions, reproduction should be more resistant to the suppressive effects of stressors compared with cases when future reproductive success is expected to be high. Consistent with this prediction, stress has been shown to be less suppressive of reproduction in semelparous organisms and organisms that have short breeding seasons (Wingfield and Sapolsky, 2003). At an intraspecific level, stressors have been shown to have a reduced effect on reproductive behaviors in populations that inhabit high-disturbance urban areas compared with rural habitats (Abolins-Abols et al., 2016).

While these findings are consistent with the theoretical prediction that low expected future reproductive success should reduce the effect of stressors on reproduction, we lack experimental studies that explicitly test this prediction and study which conditions lead to changes in the suppressive effect of stress on reproduction. Revealing the conditions and mechanisms that mediate effects of stress on reproduction is important for a better understanding of reproductive physiology and ecology of animals (Calisi et al., 2018), especially in the context of the human-induced rapid environmental change (Sih et al., 2011). Ultimately, such knowledge will aid in developing a predictive understanding of how and why individuals or populations respond differently to stressors.

Here, we manipulated the disturbance environment (long-term stressors) to simulate environments with different expected future reproductive success, and asked whether animals exposed to high and low disturbance differed in the degree to which an acute stressor affected their testicular function. In male vertebrates, testes produce sperm and androgen hormones, such as testosterone. Testosterone mediates multiple important reproductive functions: it regulates the development of the sexual phenotype (Hau, 2007) and is important in mediating behaviors necessary for reproduction, such as courtship (Fusani et al., 2014; Hutchinson, 1967) and territorial behavior (Soma, 2006; Wingfield et al., 1987). Testosterone synthesis is largely regulated by the hypothalamic–pituitary–gonadal (HPG) axis, where secretion of gonadotropin releasing hormone (GnRH) from the hypothalamus causes the secretion of luteinizing hormone (LH) from the pituitary, which in turn upregulates synthesis of testosterone by the Leydig cells in the testes (Farner and Wingfield, 1980; London et al., 2006). Testosterone synthesis by testes has also been shown to change in response to local signaling factors independent of brain signaling (Nogueiras et al., 2004; Tena-Sempere et al., 2002). Testosterone levels can be suppressed by both psychological (Deviche et al., 2010; Moore and Mason, 2000) as well as physical (Lynn et al., 2015; Nelson et al., 1989) stressors. This decrease may be mediated via suppressive effects of stressors on GnRH or LH secretion (Breen and Karsch, 2006; Breen et al., 2007; Nikolarakis et al., 1986), or by direct action of components of the physiological stress response, such as the hypothalamic–pituitary–adrenal (HPA) axis, on gonads (Deviche et al., 2012, 2017; Lynn et al., 2015; McGuire et al., 2013).

Gonads have been shown to be especially sensitive to stressors and other perturbations during growth (Guillette et al., 1994; Rhind et al., 2001; Zambrano et al., 2014). For example, maternal stress during fetal gonadal development can result in reduced gonadal size (Dahlöf et al., 1978), and exposure to glucocorticoids during fetal development has been shown to reduce testosterone synthesis capacity in vitro (Page et al., 2001) and delay the onset of puberty (Smith and Waddell, 2000). Such developmental programming has important implications for pathology and reproductive rates in the wild (McMillen and Robinson, 2005; Sheriff et al., 2010). If gonadal responses to stress during development are adaptive, then the effect of stressors on gonadal function should conform to the life history expectations (Sheriff and Love, 2013), as outlined above, wherein gonads are expected to be less sensitive to stressors in animals undergoing gonadal growth in a high-stress environment compared with animals developing in a benign environment.

In this study, we experimentally investigated whether a long-term disturbance environment during seasonal testicular recrudescence affected testicular physiology and gene expression in the dark-eyed junco [Junco hyemalis (Linnaeus, 1758)], a small passerine. Testicular recrudescence in seasonally breeding animals shares many parallels with development of testes in the embryo in that, during recrudescence, undifferentiated fibroblast tissue develops into mature testes with fully differentiated cell types (Nicholls and Graham, 1972). In particular, we investigated whether chronic disturbance and acute stressors, and their interaction, changed baseline testosterone levels, the overall ability to elevate testosterone following stimulation with exogenous GnRH, expression of hormone receptors that mediate testes function, and the testicular transcriptome in general. We predicted that both chronic and acute stressors should decrease testosterone levels (Deviche et al., 2010; Moore et al., 1991; Retana-Márquez et al., 2003), but that the suppressive effect of an acute stressor on testosterone synthesis would be dampened in chronically disturbed animals. We further predicted that both chronic and acute stressors would negatively affect gene expression associated with testosterone production and spermatogenesis, but upregulate expression of genes involved in cellular and hormonal stress responses, such as glucocorticoid receptor (GR) (McGuire et al., 2013) and gonadotropin-inhibitory hormone receptor (GnIHR) (McGuire and Bentley, 2010). Lastly, we predicted that the effect of an acute stressor on the testicular transcriptome would be dampened in the chronically disturbed animals.

This study was conducted in accordance with Indiana University Animal Care and Use Committee guidelines, protocol #12-050-08. We captured 36 wintering male dark-eyed juncos (Junco hyemalis hyemalis) in Bloomington (latitude: 39.144887, longitude: −86.535839), IN, USA, in December 2013 using baited mistnets and walk-in traps. We determined the sex of birds by plumage color and confirmed it by analysis of genetic sex markers (Griffiths et al., 1998). The age of birds was unknown. Before the experiment, birds were housed in free-flying groups in indoor aviary rooms, reflecting their natural winter flocking lifestyle. On 15 January 2014, we started to gradually (3 times per week) increase day length from the natural Bloomington winter photoperiod (10 h 44 min of daylight, Fig. 1). On 27 January 2014, all males were transferred to 60×60 cm metal cages (1 individual per cage, 10 cages per room, 4 rooms in total) with visual access to other birds, including females. Cages contained food, water bowls and perches. Food was provided ad libitum 3 times a week, and cages were cleaned once a week. When the experiment began on 3 February 2014, animals were experiencing 13 h 23 min of daylight and, by the end of the experiment on 26 February 2014, birds were exposed to 16 h 3 min of daylight, reflecting a summer-like photoperiod. Males were singing during the last week of the experiment.

We randomly assigned half of the birds (n=18, 2 separate rooms) to a chronic disturbance treatment, while the other half (n=18, 2 separate rooms) were treated as controls. Chronic disturbance and control rooms were separated in space to reduce the possibility that control birds were affected by the noises from the disturbance treatment. In the chronic disturbance treatment, birds were exposed to either physical disturbance (hand-waving, cage-tapping) or to a predator mount for 30 min 4 times each day for 3 weeks. During physical disturbances, observers (1 per room) either waved their hand inside the cage or tapped on the cage for 30 s, after which the observer moved on to another cage in random order. For the predator stressor, we used taxidermic study skins of Cooper's hawk (Accipiter cooperii), barn owl (Tyto alba) and fox squirrel (Sciurus niger), which were fixed to a tripod and left in the room by the observer. We had validated these treatments in a pilot study in which we investigated whether juncos exposed to hand waving or an owl mount for 30 min had higher baseline corticosterone levels than control birds (Hanauer, 2017). The pilot experiment was conducted between 17 and 21 December 2013, a month before the main experiment, on individually caged birds kept on the natural winter photoperiod. Birds used for the pilot experiment were not used in the main study. In the pilot study, hand waving (n=8) and owl presentation (n=8) treatments were carried out on different dates, with control animals (n=8 for each treatment) sampled on the same date as the treatment. Birds that experienced hand waving had significantly higher corticosterone levels [mean (x)=14.75 ng µl⁻¹, s.e.m.=2.48] than control birds (x=4.51 ng µl⁻¹, s.e.m.=0.20). Similarly, birds that were exposed to owl had significantly higher corticosterone levels (x=6.25 ng µl⁻¹, s.e.m.=0.83) than control birds (x=3.23 ng µl⁻¹, s.e.m.=0.38).

Because hand waving caused the largest increase in corticosterone in our pilot study, birds in the chronic disturbance treatment received at least 1 of these disturbances each day (average: 2.5 physical disturbances each day). Otherwise, the choice of the stressors for each disturbance bout was random. The disturbance bouts occurred at least 60 min apart for each room and were administered during daylight hours. In the control group, we did not disturb the birds except for standard animal husbandry (see above). We did not measure corticosterone increase in response to hand waving or predator presentation in the birds in the chronic disturbance treatment. We do not know, therefore, whether these birds habituated to these repeated stressors during the experiment. However, after 3 weeks of treatment, birds from the chronic disturbance treatment had significantly lower condition (mass relative to body size) and significantly lower fat deposits compared with control birds (Hanauer, 2017), an indication of chronic stress (Dickens and Romero, 2013).

We took blood samples from each bird to analyze testosterone and corticosterone concentrations before (1–2 February 2014) and after (23–24 February 2014) the 3 weeks of chronic disturbance or control treatment. In addition, blood was taken from birds 1 week after the beginning of treatment for analysis of corticosterone levels (8–9 February 2014). Each bird was captured between 7:00 and 12:00 h, and a baseline blood sample (100 µl) was taken from the brachial vein within 4 min of capture using a 26-gauge needle and microcapillary tubes. The timer for blood collection was started when we entered the room in which birds were housed. Multiple birds (up to 5) were captured at once by 1 of up to 5 observers. In order to exclude indirect effects from the capture of other birds in the study, we turned off the lights during capture and did not sample the remaining birds from the same room until the following day.

After the blood sample was taken, each bird was put in a brown paper bag, a standard acute handling stressor that is widely used to assess physiological stress response in birds (Sheriff et al., 2011). After 30 min, each bird was bled again (100 µl) to measure acute-stressor-induced hormone levels (Deviche et al., 2010; Schoech et al., 1999). Blood was stored at 4°C for up to 6 h until centrifugation. Following centrifugation, plasma was removed using a Hamilton syringe and stored at −20°C until analysis.

Both chronically disturbed and control birds increased their corticosterone levels in response to the acute stressor, but chronically disturbed birds elevated their corticosterone levels to a lesser extent (Hanauer, 2017).

To assess the ability of testes to increase testosterone, during the last sampling round at the end of the experiment (23–24 February 2014), we injected birds with 50 µl of 25 ng µl⁻¹ GnRH (chicken-derived GnRH I in 1 mol l⁻¹ PBS, American Peptide Company Inc., Sunnyvale, CA, USA, product no. 54-8-23) 35 min after capture (i.e. after the post-acute-stressor 30 min blood sample was collected). GnRH is the top regulator of testosterone synthesis and is naturally produced in hypothalamus (Ciechanowska et al., 2010). GnRH-induced testosterone levels have been shown to be repeatable (Jawor et al., 2006), and are linked to a higher probability of survival (McGlothlin et al., 2010) and with morphological characteristics, such as plumage color (Atwell et al., 2014). To administer GnRH, we cleaned the pectoral muscle using an alcohol swab, and injected GnRH using a Hamilton syringe. GnRH solutions were kept on ice before the injection to prevent degradation of the peptide. A total of 100 µl of blood were collected 30 min after injection.

After blood collection, we used a ruler to measure the size of the cloacal protuberance (CP) by measuring its length from body to the cloacal opening, to the nearest 0.5 mm. We also measured the pectoral muscle condition, fat score, and mass. Birds from the chronic disturbance treatment had significantly lower condition (mass relative to body size) and significantly lower fat deposits compared with control birds (Hanauer, 2017), an indication of chronic stress (Dickens and Romero, 2013). To minimize potential measurement bias, we randomized the order in which birds from different treatments were measured.

We measured testosterone levels in the blood plasma using a commercial enzyme immune assay (EIA) from Enzo Life Sciences (Farmingdale, NY, USA, product no. ADI-901-65) that has been previously validated for use in this species (Clotfelter et al., 2004). Testosterone was extracted from 20 µl of plasma using diethyl ether. Tritiated testosterone was added to the sample to estimate extraction efficiency (average 94.8%). Extracted hormone was reconstituted in 50 µl of 98% ethanol, followed by 300 µl of assay buffer. We followed the manufacturer's instructions for the remaining procedures. We estimated hormone concentrations in reference to a 9-point standard curve (range 0.78–200 pg well⁻¹) using a curve-fitting program (Microplate Manger, Bio-Rad Laboratories, Hercules, CA, USA). The detection limit of the assay was 0.099 ng ml⁻¹ of testosterone in plasma. Samples were distributed randomly across and within plates, and all samples and standard curves were run in duplicate. Within-plate variation was 7.01% and between-plate variation was 1.89%. We did not correct for either extraction efficiency or across-plate variation.

At 1–2 days after the final blood sampling (25–26 February 2014), birds were killed using isoflurane overdose in a 2-way factorial design: birds from both chronic disturbance and control treatments were killed either immediately after capture (baseline treatment) or after 90 min of being held in a paper bag (acute handling treatment). We selected this timing for several reasons. First, we wanted to allow sufficient time for stress-induced changes in gene expression. Second, we had preliminary data to suggest that steroidogenic gene expression is sensitive to other hormonal stimuli 90 min after administration [e.g. steroidogenic acute regulatory protein (StAR) mRNA abundance after GnRH injection; K.A.R. and E.D.K., unpublished data]. Third, literature on immediate early gene (IEG) expression and half-life suggests that assessment at 90 min following a stimulus allows us to capture both induction as well as reduction in IEG expression (Maney and Goodson, 2011). Since IEGs include transcription factors, we estimated that gene expression 90 min after capture was likely to capture both up- and downregulation of components of the testicular transcriptome, compared to expression immediately following capture. Testes were dissected and flash frozen on pulverized dry ice. To assess testes size, an observer who was unaware of our hypotheses measured testes mass (to the nearest mg) before RNA extraction.

We used a custom Nimblegen 12-plex microarray (Roche Nimblegen, Madison, WI, USA) for the dark-eyed junco (Peterson et al., 2014) to analyze the effect of treatments on the transcriptome. RNA was extracted from testes using the Trizol method (Invitrogen, Carlsbad, CA, USA) and quantified using the Nanodrop ND-2000 (ThermoScientific, Waltham, MA, USA). Sample integrity was verified using TapeStation (mean RIN: 9.3; range 8.6–9.7; Agilent Technologies, Santa Clara, CA, USA). Total RNA was converted to double-stranded cDNA using the TransPlex Complete Whole Transcriptome Amplification Kit (Sigma-Aldrich, St Louis, MO, USA, product no. WTA2) according to the manufacturer's instructions. Labeled cDNA was synthesized using a NimbleGen Dual-Color DNA Labeling Kit (Roche NimbleGen, Inc.) as described in the standard NimbleGen CGH Analysis Procedure. Briefly, 500 ng of cDNA was denatured at 98°C for 10 min. The denatured sample was snap chilled on ice and incubated with 50 units of Klenow fragment (New England Biolabs, Ipswich, MA, USA) and 10 mmol l⁻¹ dNTP mix for 2 h at 37°C. Reactions were terminated with 0.5 mol l⁻¹ EDTA and precipitated with isopropanol, rinsed in cold 80% ethanol and desiccated using a Savant DNA 120 SpeedVac Concentrator (ThermoScientific). Pellets were resuspended in nuclease-free water. Cy-labeled product was quantified using a Nanodrop ND-2000 (ThermoScientific) and 8 µg (4 µg each of Cy-3 and Cy-5) labeled product was pooled and desiccated for subsequent hybridization. The labeled product was resuspended using Sample Tracking Controls (Roche NimbleGen, Inc.). Components from the Hybridization Kit, LS (Roche NimbleGen, Inc.) were combined to prepare the hybridization solution. This solution was added to each subarray pool for a total volume of 8 µl. The sample was denatured at 95°C for 5 min and allowed to hybridize on the microarray for 18 h at 42°C in a NimbleGen hybridization system. The post-hybridization wash was conducted using a Wash Buffer Kit (Roche NimbleGen, Inc.). Arrays were agitated consecutively in Wash Buffer I (2 min), Wash Buffer II (1 min) and Wash Buffer III (15 s). The arrays were dried for 1 min using a High-Speed Microarray Centrifuge (Arrayit Corp., Sunnyvale, CA, USA). Image acquisition was attained using the NimbleGen MS 200 Microarray Scanner (Roche NimbleGen, Inc.) at 2 μm resolution. cDNA was hybridized to the microarray using a full round-robin design.

The microarray contained 33,545 assembled sequencing reads (contigs) in triplicate covering 22,765 putative genes (isogroups), which were based on dark-eyed junco transcriptome sequencing (Peterson et al., 2012). We tested for significant differences in gene expression between treatment groups (and their interaction) using R package limma (Smyth, 2005) (i.e. we compared pooled chronic disturbance versus control, and pooled acute handing versus baseline control). We calculated a global false-discovery rate across all comparisons and used a q-value threshold of 0.05 for significance (see Peterson et al., 2014 for further details). We used topGO (Alexa and Rahnenfuhrer, 2016) with the weight algorithm (Alexa et al., 2006) to identify the Gene Ontology (GO) terms (Ashburner et al., 2000) that were significantly over-represented among the significantly differentially expressed genes in each comparison.

The microarray did not include all of our candidate genes of interest. Therefore, we subsequently performed quantitative polymerase chain reaction (qPCR) to test our a priori predictions about genes related to testosterone synthesis and testicular function. Specifically, we quantified expression of steroid synthesis genes [luteinizing hormone receptor (LHR), StAR, cytochrome P450 side-chain cleavage (P450), cytochrome P450 17α-hydroxylase (CYP17), 3β-hydroxysteroid dehydrogenase/isomerase (3βHSD); we did not measure the expression of 17βHSD, a gene that catalyzes the last step in testosterone synthesis, because its isoforms are poorly characterized in birds (London and Clayton, 2010)], as well as genes that encode receptors for hormones that may regulate testicular function [sperm production: follicle-stimulating hormone receptor (FSHR); response to stressors: GR, mineralocorticoid receptor (MR), GnIHR]. Most of the primers, except FSHR primers, had been previously validated in our system (Bergeon Burns et al., 2014; Rosvall et al., 2016a,b). FSHR primers were designed using white-throated sparrow genomes using Primer-BLAST (Ye et al., 2012). Primer sequences and efficiencies are reported in Table S1. We ran qPCR reactions using PerfeCTa SYBR Green SuperMix (Quantabio, Beverly, MA, USA, product no. 95054) on the Roche LightCycler 480 platform (Roche Holding AG, Basel, Switzerland) using the same cDNA samples from the microarray. The cDNA concentration was similar between samples (17–22 ng µl⁻¹). All sample and calibrator cDNA samples for genes of interest and housekeeping genes were run in triplicate on clear LightCycler 480 384-well plates (product no. 05102430001, Roche Holding AG). Prior to sample analysis, we measured the efficiency of each primer by amplifying sequential 1:4 dilutions of a pooled cDNA sample. cDNA was diluted 1:40 in double-distilled water from the Millipore system (Merck Millipore) prior to plating. Each well contained 8 µl of PerfeCTa SYBR Green SuperMix and 2 µl of the 1:40 cDNA dilution. Plates were sealed with a clear plastic seal and centrifuged at 1200 g for 4 min to remove bubbles. The qPCR cycle program was as follows: initial denaturation for 10 min at 95°C was followed by amplification for 40 cycles (denaturation for 30 s at 95°C; annealing for 60 s at 60°C; extension for 30 s at 70°C; fluorescence measurement for 2 s at 78°C).

Following sample amplification, we investigated product size and quality using a melting curve (60 s at 95°C, 30 s at 55°C, followed by a ramp from 55–95°C at 0.11°C s⁻¹). Samples with multiple melting peaks were excluded from analysis. We observed a small extra peak with low melting temperature for most samples, which likely represents conspecific amplification or primer dimers. To preclude the possibility of this affecting our gene expression measurement, following the guidelines from the manufacturer, we moved the fluorescence measurement to 78°C (see above). This ensured that the fluorescence measurements were collected at temperatures at which the short double-stranded non-specific products were denatured.

We calculated the relative gene expression in each sample in reference to a pooled standard, and normalized this value against average expression of 2 housekeeping genes [peptidylprolyl isomerase A (PPIA) and ribosomal protein L4 (RPL4)] using LightCycler 480 software (Roche Holding AG, release no. 1.5.1.62). PPIA and RPL4 are 2 of the most stable housekeeping genes in testes of passerines (Zinzow-Kramer et al., 2014), and their expression did not differ between chronic or acute treatments. Primers for PPIA and RPL4 were designed using the white-throated sparrow genome. To calculate amplification efficiencies, we used the ‘Abs Quant/2nd Derivative Max’ method. For analysis of relative gene expression in our samples, we used the ‘Advanced Relative Quantification’ method, specifying the ‘High Confidence’ option under ‘Abs/Quant/2nd Derivative Max’, choosing the ‘All to All’ pairing rule, and using the efficiency to control for differences in amplification. We inspected all replicates for samples and the calibrator, and excluded the replicates that deviated from others by more than 0.7 crossing points (Cps). We then verified that the target/reference ratios calculated using the 2 housekeeping genes separately identified quantitatively similar relationships between the samples. If so, we reran the analysis using the ‘All to Mean’ pairing rule, which uses the mean Cp of both housekeeping genes to calculate the final target/reference ratio for each sample.

We analyzed our data in R (http://www.R-project.org/). We used linear (LM) and linear mixed effects models (LMM, package nlme, https://CRAN.R-project.org/package=nlme) to analyze the effect of chronic and acute treatments on reproductive physiology (hormone levels) and morphology (testes mass and CP size). We tested whether the time of day of hormonal sampling explained variation in hormone levels and used it as a continuous covariate in models where this effect was significant. We included individual as a random factor in all mixed effects models that included more than one measurement from the same individual (denoted as LMMs). When testing whether the change in testosterone, as a result of acute handling stress, as a result of GnRH injection, or between the beginning and end of the experiment differed between the two long-term treatments, we included an interaction term between the long-term treatment and testosterone sampling time points. To test whether the relationship between testes mass and testosterone differed between treatments, we included an interaction term between the chronic disturbance treatment and testes mass. Interactions were removed from models if they were not significant. Our sample size differed between sampling rounds and between bleeds because of low blood volume in some of the samples. When necessary, the response variables were transformed to meet model assumptions of normality. Full model tables for models with interaction terms are reported in the Supplementary Information.

We used 2 different multivariate approaches to analyze the effect of treatments on candidate gene expression (determined by qPCR). Since the expression of steroidogenic genes showed significant pairwise correlations (Table S8), we used principal components analysis (PCA) to summarize the covariation in gene expression by creating independent principal components. We then used the first principal component as a dependent variable in ANOVA to investigate whether chronic or acute treatments, or their interaction, affected steroidogenic gene expression. Following this, we used ANOVAs to analyze each gene separately and corrected for multiple comparisons using the Benjamini–Hochberg (BH) method.

Because the expression of receptors that may downregulate testicular function (GR, MR and GnIHR) was not correlated to each other and they mediate different functions in animals, we used MANOVA to analyze the effect of chronic and acute treatments on the overall sensitivity to stress signaling. Following MANOVA, we analyzed each gene independently using ANOVA and corrected for multiple comparisons using the BH method. FSHR expression was analyzed independently using an LM, because its function is not known to relate to stress signaling or steroidogenesis.

At the outset of the experiment, neither baseline testosterone (LM, n_disturb=12, n_control=12, β=−0.142, t=−7.21, P=0.479) nor testosterone after the acute handling stressor (LM, n_disturb=11, n_control=6, β=0.158, t=1.172, P=0.230) differed between the chronic and control treatment groups, and the acute handling stressor did not reduce testosterone compared to baseline levels (LMM, n=41, β=−0.049, t=−0.616, P=0.549; Fig. 2A).

Baseline testosterone increased throughout the experiment in both disturbance treatments (LMM, n_disturb=29, n_control=28, β=0.401, t=3.964, P<0.001; Table S2), consistent with recrudescence of testes in response to increasing day length. After 3 weeks of treatment, there was a significant interaction between the long-term disturbance treatment and acute handling treatment on testosterone levels (LMM, n=48, β=0.296, t=2.265, P=0.034; Table S3, Fig. 2B). Post hoc tests showed that birds from the long-term disturbance group had marginally lower baseline testosterone levels than controls (LM, n_disturb=14, n_control=13, β=−0.482, t=−1.909, P=0.068) but that testosterone levels after 30 min of the acute handling stressor did not differ between the disturbance treatments (LM, n_disturb=12, n_control=14, β=−0.015, t=−0.078, P=0.938). Post hoc analyses also showed that birds from the control treatment showed a significant reduction in their testosterone levels from baseline to post-acute-handling stress (LMM, n_handled=16, n_unhandled=14, β=−0.702, t=−2.894, P=0.015), while birds from the chronic disturbance treatment did not (LMM, n_handled=17, n_unhandled=12, β=−0.031, t=−0.187, P=0.855). Both control and chronically disturbed birds significantly increased their testosterone in response to GnRH injection compared with pre-injection levels (LMM, n_pre=26, n_post=34, β=0.214, t=9.077, P<0.001), but the GnRH-induced testosterone did not differ between treatment groups (LMM, treatment×injection interaction, n=60, β=0.016, t=0.037, P=0.971; Table S4, Fig. 2C).

Testes mass was not affected by the chronic disturbance treatment (LM, n_disturb=18, n_control=17, β=0.005, t=0.935, P=0.357; Fig. 3A). There was no relationship between testes mass and baseline testosterone levels (LM, n=27, β=7.120, t=0.949, P=0.352), but testes mass was positively related to testosterone levels after acute handling stress (LM, n=25, β=12.393, t=2.838, P=0.010) as well as to GnRH-induced testosterone levels (LM, n=34, β=34.627, t=3.127, P=0.004; Fig. 4). The relationship between testes mass and testosterone did not differ between disturbance treatments (Table S5). CP size showed significant differences between disturbance treatments, with chronically disturbed birds having larger CPs than control birds (LM, n_disturb=16, n_control=17, β=1.631, t=4.711, P<0.001; Fig. 3B).

We found that the expression of 16 transcripts was significantly affected by the chronic disturbance treatment after correcting for false discovery rate (out of 20,390 total expressed in testes in this study, 0.078%; Table S6). GO analysis identified 3 terms that were overrepresented among the significantly differentially expressed genes. These were: electron transport chain, extracellular structure organization (biological process) and glycosaminoglycan binding (molecular function) (Table 1). None of the genes that were differentially expressed between tissues in response to chronic treatment are known to link clearly to the stress response or reproductive function.

The acute handling stressor caused significant changes in the expression of 168 transcripts compared with baseline treatment (out of 20,390 total expressed in testes, 0.823%; Table S7). Among genes that were contributing to these terms were a variety of heat shock proteins (HSPB1, DNAJA4, HSPA4L, HSP90AA1), and genes associated with inflammation and cytokine signaling (IL4R, PIK3AP1, MAP3K8). GO analysis identified a number of terms that were overrepresented in the set of genes that showed significant treatment effect, including: receptor signaling protein activity, intracellular steroid hormone receptor signaling pathway, cellular protein complex assembly, response to stimulus, and ‘de novo’ posttranslational protein folding (Table 2). One unannotated transcript with an unknown function showed a significant interaction between the long-term treatments (chronic versus control) and acute treatments (handled versus baseline).

The expression of 5 genes (LHR, StAR, p450scc, CYP-17, 3bHSD) involved in testosterone synthesis was positively correlated (Table S8). In PCA, all of these genes loaded negatively on PC1, which explained 50% of the variation in expression (Table S9). However, variation in PC1 was not explained by long-term disturbance (LM, n_disturb=18, n_control=17, β=−0.294, t=−0.540, P=0.593), acute handling stressor (LM, n_handled=17, n_unhandled=18, β=−0.653, t=−1.201, P=0.239) or the interaction of these treatments (LM, n=35, β=0.670, t=0.609, P=0.547). Since testosterone levels decreased significantly in response to the acute handling stressor only in the control treatment, we conducted a separate PCA of only the control individuals (see Table S10 for loadings). This analysis showed that the acute handling stressor tended to reduce the expression of testosterone synthesis genes 90 min later, although this effect was not significant (LM, n_handled=8, n_unhandled=9, β=−1.226, t=−1.828, P=0.088). Analysis of each of the genes individually using ANOVA showed that CYP17 levels were marginally lower in handled control birds (LM, n=17, t=2.102, P=0.053; see Table S11 for other genes), although this effect was not significant after multiple comparison correction (P=0.265; Fig. S1).

Neither handling (LM, n_handled=17, n_unhandled=18, β=0.103, t=1.432, P=0.162) nor chronic disturbance (LM, n_disturb=18, n_control=17, β=−0.119, t=−1.653, P=0.108), nor their interaction (LM, n=35, β=−0.096, t=−0.661, P=0.514), affected expression of FSHR (Fig. S2).

MANOVA showed that the acute handling stressor (n=35, Wilks=0.676, F_3,30=4.791, P=0.008), but not chronic disturbance (n=35, Wilks=0.907, F_3,30=1.027, P=0.395), nor interaction between chronic treatment and the acute handling stressor (n=35, Wilks=0.855, F_3,29=1.643, P=0.201), affected expression of receptors (GR, MR, GnIHR) for hormones that are known to suppress or inhibit testosterone production. Post hoc linear models comparing handling to baseline treatment showed that the acute handling stressor significantly downregulated GR mRNA expression (LM, n_handled=17, n_unhandled=18, β=0.137, t=2.282, P=0.029; Fig. 5A) and marginally upregulated GnIHR expression (LM, n_handled=17, n_unhandled=18, β=−0.262, t=−1.875, P=0.070; Fig. 5C), although these effects were not significant after correction for multiple comparisons (Table S12).

In this study, we experimentally tested how chronic disturbance, an acute handling stressor and their interaction affected testes function in a songbird during seasonal testicular recrudescence. We showed that birds in the chronic disturbance treatment had marginally lower baseline testosterone compared with control animals. An acute handling stressor reduced testosterone levels in control animals, but not in animals from the chronic disturbance treatment. The chronic disturbance treatment did not affect the ability to produce testosterone in response to exogenous stimulation from GnRH, and testes size did not differ between chronic disturbance and control treatments. Surprisingly, birds in the chronic disturbance treatment had significantly larger CPs than birds in the control treatment. Chronic disturbance had very little effect on gene expression in the testes: neither steroidogenic enzyme expression nor expression of receptors associated with potential regulation of testes function showed significant differences between the chronic disturbance and control treatments. Overall, chronic disturbance had a significant effect on only a handful of genes in the transcriptome. The acute handling stressor, by contrast, had a comparatively strong effect on the testicular transcriptome, a marginally suppressive effect on steroidogenesis enzyme gene expression, a marginally suppressive effect on the expression of GR mRNA, and a marginally positive effect on GnIHR gene expression compared with baseline treatment. There was little evidence of an interaction between chronic disturbance and the acute stressor on expression of the testicular transcriptome. Collectively, these findings shed light on mechanisms by which short- and long-term stressors, and their interaction, affect reproductive function.

Birds from the chronic disturbance treatment had marginally lower baseline testosterone levels compared with the control treatment, agreeing with other reports in a variety of organisms that show a decrease in testosterone in response to chronic disturbance (Pickering et al., 1987; Moore et al., 1991; Retana-Márquez et al., 2003; but see Armario and Castellanos, 1984; Jones and Bell, 2004). Suppression of testosterone levels in response to a chronic stressor in the wild may result in reduced investment in reproduction, exacerbating the negative effects of disturbance on fitness.

Mechanistically, the marginal difference in testosterone levels between the chronic disturbance and control birds in this study was not explained by differences in testicular physiology: testes mass and the GnRH-induced testosterone levels were not different between the disturbance treatments, suggesting that the difference in testosterone was not caused by long-lasting differences in the ability of the testes to produce testosterone.

Furthermore, chronic disturbance did not have a significant effect on steroidogenic gene expression, expression of receptors (GnIHR, GR, MR) for potential inhibitory hormones, or the testicular transcriptome in general, suggesting that testes function was nearly unaffected by the chronic disturbance treatment. This suggests that local signaling at the level of the testes is unlikely to be the cause of lower testosterone levels or the decreased sensitivity to an acute handling stressor that we observed in the chronically disturbed animals. Instead, potential long-term differences in testosterone may be mediated by systemic signals that inhibit testicular function, such as GnIH (McGuire and Bentley, 2010) or corticosterone (McGuire et al., 2013). However, birds from the chronic disturbance treatment did not show higher corticosterone levels (Hanauer, 2017), suggesting that variation in circulating corticosterone was not responsible for differences in testosterone between the disturbance and control treatments.

The lack of effect of chronic disturbance on the testicular transcriptome may also have been due to an unintended long-term modulation of gene expression by the acute stressor treatments during blood collection in the days preceding killing. Future analysis of the post-translational effects of stressors, as well as other tissues (e.g. hypothalamus, liver) collected from these animals, may elucidate the mechanisms that caused the differences in testosterone levels between the disturbance treatments (Lynn et al., 2015).

A surprising finding in this study was that the CPs were significantly larger in the chronically disturbed birds compared to control animals. CPs in birds are sperm storage organs that develop in males during the breeding season (Salt, 1954) and are regulated by testosterone (Witschi, 1961). Males with larger CPs store more sperm (Tuttle et al., 1996) and larger CPs are hypothesized to allow faster copulation (Birkhead et al., 1993). Barring spurious results, our finding may indicate that birds in the chronic disturbance group may be investing more in sperm production compared to the control animals, perhaps to advance their reproductive readiness or enable more rapid copulation in an uncertain environment.

It is important to note that the birds in this study had not reached full reproductive capacity. Testes and CPs did not reach their full size during this study. Testes mass was 21% of mature reproductive size (Bergeon Burns et al., 2014). The capacity of birds to produce testosterone following GnRH injection was slightly lower than in wild animals during the breeding season (Rosvall et al., 2014). However, baseline testosterone concentration was similar to that in the wild (Rosvall et al., 2014), indicating that patterns observed in this study are likely to persist later in the breeding season.

Before the disturbance treatment, testosterone did not decrease in response to acute stress. However, testosterone levels at this stage were low and near the assay sensitivity limit; therefore, the absence of the effect may be due to the limitations of the hormone assay. After the treatment, birds in the control treatment showed a significant decrease in testosterone levels in response to the acute handling stressor. Suppression of testosterone in response to acute stressors has been shown in many other studies (Deviche et al., 2010, 2012; Moore et al., 1991). This effect may be due to a reduction in testosterone synthesis, or an increase in testosterone metabolism by the liver (Lynn et al., 2015). We found a marginally significant decrease in the expression of CYP17, an enzyme involved in steroidogenesis, in birds that had experienced an acute handling stressor compared with baseline treatment. Because the expression of steroidogenic enzymes is positively correlated with testosterone levels (Rosvall et al., 2016b), a decrease in expression of testosterone synthesis genes may be responsible for suppression of testosterone in response to handling.

Downregulation of steroidogenic enzyme gene expression in response to the acute handling stressor may be due to either a change in the local or systemic signaling, or because of a change in sensitivity to this signaling (Ernst et al., 2015). We found a significant overall change in hormone receptor gene expression, primarily downregulation of GR expression and upregulation of GnIHR expression, in response to the acute handling stressor. This suggests that testes in acutely stressed birds may be less sensitive to downregulation by glucocorticoids, such as corticosterone, but more sensitive to another important hormone – GnIH – that has an inhibitory effect on reproductive physiology (Tsutsui et al., 2010). Indeed, while corticosterone levels increased following the acute handling stressor (Hanauer, 2017), other studies have shown that corticosterone is unlikely to be the cause for rapid decrease in testosterone in response to acute stress in birds (Davies et al., 2016; Deviche et al. 2017). Testosterone and corticosterone levels in birds are often either unrelated or positively correlated (Klukowski et al., 1997; Buchanan et al., 2003), showing that corticosterone and testosterone do not covary in a way that is consistent with testosterone suppression by corticosterone. We could not assess the expression of GnIH in the testes or brain; therefore, we do not know whether acute or chronic stressors, in addition to GnIHR, also upregulated GnIH synthesis.

Acute handling stress caused significant changes across the testicular transcriptome compared with baseline treatment. The GO analysis did not indicate that the acute handling treatment affected testes-specific functions, such as spermatogenesis or steroidogenesis, although we note that several candidate genes were not present on this array. Instead, GO analysis suggested that acute handling stress induced a cellular stress response, resulting in upregulation of a handful of heat shock proteins. Heat shock proteins are molecular chaperones that, under normal conditions, facilitate protein assembly and are upregulated in response to stressors (Åkerfelt et al., 2010). Heat shock protein expression in testes increases during spermatogenesis and oogenesis (Neuer et al., 2000), which may ensure that the development of gametes is shielded from environmental stressors. Importantly, some of the same genes (e.g. HSP90AA1, DNAJA4) that were upregulated in response to acute handling stress in this study were also upregulated in chicken testes in response to heat stress (Wang et al., 2015), indicating that testes may have a generalized cellular stress response that is upregulated in response to a variety of stressors. Because we could only measure gene expression at one time point, we do not know the duration of the observed changes in gene expression, or whether they are linked to the observed changes in hormone levels (measured on a shorter timescale). Our data nevertheless indicate that acute handling stress has marked effects on hormone-related gene expression in the testes in the short term.

Birds from chronic disturbance and control treatments responded differently to the acute handling stressor: whereas handling reduced testosterone levels in control birds, it did not affect testosterone levels in birds from the chronic disturbance treatment. This difference can be interpreted in 2 non-exclusive ways: first, the reduced impact of the acute stressor on testosterone in birds from the chronic disturbance group might have arisen because their reproductive function was already downregulated to a degree that prohibited further decrease [‘floor effect’ sensu (Sapolsky et al., 1984)]. Second, the difference in the effect of handling on testosterone levels between chronic disturbance and control treatments could be explained by a lower sensitivity of the chronically disturbed birds to stressors compared to the control individuals. We did not find any effect of chronic disturbance on signaling in the testes that would explain this difference, suggesting that differences in sensitivity to stress may exist in the HPG axis tissues that are upstream from the testis (pituitary, hypothalamus), or other hormonal targets that interact with testosterone production.

From an ecological perspective, the first alternative (floor effect) could be interpreted as a consequence of homeostatic overload (Romero et al., 2009). Homeostatic overload refers to the cases where overexposure to stressors, and the associated increase in the frequency of stress response, results in a pathological state that leads to compromised organismal function (Romero et al., 2009). In our study, the high frequency of stressors may have resulted in a compromised ability to produce testosterone. However, because birds from both chronic and control treatments elevated their testosterone in response to GnRH to the same degree, this explanation is unlikely. Instead, the floor effect could be a result of a ‘best of a bad job’ strategy, wherein individuals of low quality or in a bad environment maximize their fitness by playing it safe (Sih and Bell, 2008). In this case, animals may opt to keep the resources diverted from reproduction to maximize their ability to respond to stressors.

The second alternative (differences in sensitivity to stress), is consistent with the life history prediction that, under conditions of low expected future reproductive success, stress should have a less negative effect on reproduction compared to situations where expected reproductive success is high (Wingfield and Sapolsky, 2003). Animals that grow their reproductive tissues in environments with a high frequency of stressors may change their physiology to reduce sensitivity of the HPG axis to stress, thus allowing them to maintain reproductive function during stressful episodes, despite the possible cost to self-maintenance (Wingfield and Sapolsky, 2003). This prediction has been supported by a study in a natural system (Abolins-Abols et al., 2016), where the reproductive phenotype of urban animals is less sensitive to stressors compared to that of rural animals.

In contrast to our testosterone findings, we did not find an interaction between chronic and acute treatments on the testicular transcriptome, although we note that the marginal effects of the chronic treatment reduce our power to explore this interaction. This finding nevertheless provides further evidence that chronic disturbance does not affect testicular capacity to produce testosterone or respond to stress. Instead, we hypothesize that the differential effect of handling stress on testosterone levels in the chronic disturbance and control treatments may be due to differences in upstream signaling from the pituitary or hypothalamus, which regulates the activity of the molecular machinery in testes. Further work will be needed to explore alternatives, such as post-translational effects of stress that are not captured at the level of gene expression. Despite these uncertainties, our results nevertheless demonstrate that testes are relatively robust to interacting effects of acute and chronic stressors on gene expression.

We found that testes mass was positively correlated with testosterone levels after acute handling stress and after GnRH injection, but not with baseline testosterone. A significant positive relationship between testes mass and the capacity of testes to produce testosterone following GnRH injection is likely due to the variation in testicular growth rate or stage in junco males: individuals with more recrudesced testes likely had more or larger Leydig cells capable of producing more testosterone (Temple, 1974). A more interesting observation is that testicular mass was positively correlated with testosterone levels after an acute stressor but not at baseline levels. This suggests that, while baseline testosterone is responsive to various environmental and social factors and stressors, these effects are superimposed on ‘true’ baseline production of testosterone that is insensitive to stimuli and is proportional to the size of the testes. In other words, suppression of testosterone may lower testosterone levels, but larger testes will nevertheless keep producing more testosterone than small testes as a result of baseline activity of the synthesis machinery.

Overall, our results show that chronic and acute stressors suppress testosterone release by the testes, but that the effect of acute stressors differs depending on the frequency of stressors in the environment. These differences in testes function between disturbance environments are unlikely to be mediated by changes in testicular gene expression, but are more likely to reside upstream of the testes. Whereas chronic treatment had a negligible effect on the testicular transcriptome, acute handling stress significantly upregulated major components of the cellular stress response and affected the expression of hormone receptors involved in the downregulation of testosterone production. These results suggest a potential mechanism for regulating testosterone decrease in response to acute stressors. Furthermore, they show that transient changes in gene expression in response to acute stressors are different from more permanent responses to chronic stressors. An important future direction is therefore to identify the mechanisms responsible for differences in testosterone levels between animals experiencing different disturbance regimes.

This study is among the first to experimentally test the mechanisms by which acute and chronic stressors interact to influence reproduction. We demonstrate patterns that, adaptive or not, are doubtlessly important in understanding how animals respond to chronic and acute stressors. We therefore urge further study on the adaptive significance and mechanisms that mediate the effect of chronic stressors on testosterone levels and other reproductive functions.

We thank A. Kimmitt, S. Slowinski, H. Kassab, A. Brenner, A. McKay, C. Taylor and N. Fletcher for assistance with disturbance manipulations, animal care and sampling. We thank L. Sloan, who extracted RNA, M. Stephens, who ran the microarray at the University of Notre Dame Genomics & Bioinformatics Core Facility, and D. Rusch and A. Fellows from Indiana University's Center for Genomics and Bioinformatics for assistance with microarray data analysis. We thank C. Bergeon-Burns for providing us with primer sequences and advice on lab work, R.A. Stewart for assistance with molecular sex determination, and the Center for the Integrative Study of Animal Behavior at Indiana University for providing laboratory facilities and equipment. We also thank the two anonymous reviewers for their feedback on the earlier drafts of this manuscript.