An evolutionary shift to prioritizing mating over care is associated with consistently high androgen levels in male threespine stickleback

Abstract

Steroid hormones play a crucial role in regulating social behaviors in vertebrates, but the mechanisms by which hormones mediate the evolution of these traits remain an open question. Here, we use liquid chromatographymass spectrometry (LC-MS/MS) to quantify a panel of steroids in breeding males of two recently diverged populations of threespine stickleback. The common ecotype provides paternal care, whereas the white ecotype has evolutionarily lost care. Hormone levels were quantified in both ecotypes at three reproductive stages: (1) after completing a nest, (2) soon after mating, when commons begin providing care and whites remove their embryos from the nest, and (3) four days after mating, when commons are performing parental care and are no longer courting females while whites are not providing care and are courting females. Androgens declined in commons when males began providing care but remained elevated in whites across stages, possibly to maintain the production of “spiggin,” the androgen-dependent glue males use to construct nests. Progestogen levels were consistently low in whites and were lowest in commons after mating, suggesting an antagonistic relationship between progestogens and sexual behavior. Both ecotypes showed elevated glucocorticoid levels after mating, suggesting that the stress axis may not have diverged between ecotypes. Altogether, these results provide evidence that the ecotypes regulate steroid levels differently, supporting the ways in which they balance mating and parental effort. Our data suggest a variety of mechanisms by which steroid signaling and regulation can change during the early stages of evolutionary divergence between behaviorally distinct populations.

1. Introduction

Steroid hormones coordinate the expression of many traits and regulate behavior, physiology, and morphology in vertebrates. Steroids are particularly important for reproduction, as they coordinate the timing of gamete production, the expression of secondary sexual characteristics, and behaviors such as courtship and nesting with external cues, including photoperiod and the presence of potential mates (Adkins-Regan, 2005). For example, glucocorticoids regulate energy use and stress during mating (Borg et al., 1991; Villani et al., 2006) and parenting (Alonso-Alvarez and Velando, 2012), and progestogens influence male sexual behavior (Witt et al., 1994; Andersen and Tufik, 2006; Ahamed et al., 2024) and work in concert with other steroids like estrogen to coordinate parturition and mothering in females (Kohl and Dulac, 2018). Androgens can play an important role in regulating male reproductive behavior and trade-offs between aggression and care, as formulated in the Challenge Hypothesis, first described in birds (Wingfield et al., 1990) and later extended to other vertebrate taxa (Hirschenhauser and Oliveira, 2006). Although the role of steroid hormones in coordinating suites of traits in male vertebrates is relatively conserved (Hau, 2007), the components of steroid signaling, i.e., “endocrine building blocks,” can show substantial variation (Rosvall, 2022). As products of complex endocrine cascades involving the brain, pituitary, and gonads (hypothalamic-pituitary-gonadal (HPG) axis), steroid production and signaling can vary in numerous ways, including the expression of steroidogenic enzymes, binding globulins, co-factors, receptor abundance, and receptor affinity (Rosvall, 2022).

Because steroids play a central role in regulating suites of traits like spermatogenesis, aggression, parental care, and secondary sexual characteristics, the mechanisms regulating their production and signaling likely facilitate or constrain the evolution of these traits (Hau, 2007; Ketterson et al., 2009), yet how hormones mediate the evolution of behavior and other phenotypes remains an open question (Cox et al., 2016). For example, when a single hormone influences several traits, changes in the levels of that hormone can facilitate coordinated and rapid changes in those traits. However, if that hormone has beneficial effects on some traits but deleterious effects on others (i.e., antagonistic pleiotropy), then the evolution of traits may be constrained. An important first step in understanding how steroid-mediated traits evolve is determining if and how hormone levels vary among phenotypically divergent species or populations. To understand the hormonal mechanisms underlying behavioral evolution, comparisons of hormone levels within species, among divergent populations (e.g., Bergeon Burns et al., 2013; Enbody et al., 2018; Graham et al., 2018; Kitano et al., 2010, 2011; Kitano and Lema, 2013; Miles et al., 2018) are particularly powerful, as there has been less time for differences confounded with the behavior of interest to emerge (Jourjine and Hoekstra, 2021). Once steroid levels have been established, one can begin to narrow in on how these phenotypic and hormonal differences occur.

Here, we compare whole-body steroid levels in two closely related ecotypes of threespine stickleback (Gasterosteus aculeatus) from Nova Scotia, Canada, which exhibit dramatic differences in their reproductive behavior. In the Gasterosteidae family, male-only (paternal) care is ancestral (McLennan et al., 1988). Male stickleback typically bear sole responsibility for providing care to ensure their offspring’s survival (Van Iersel, 1953). In the breeding season, males establish territories, build nests, and then attract females with courtship displays. After a female lays eggs in the nest, the male fertilizes the eggs and chases the female away. Males immediately begin caring for their embryos while continuing to court females to obtain additional clutches of embryos. This will continue for several days until the embryos near hatching, at which point males will stop courting females and switch exclusively to parenting. Generally, male breeding behavior can be divided into a “sexual phase,” during which males build nests and court females, and a “parental phase,” during which they care for their offspring (Van Iersel, 1953; Jamieson et al., 1992). An atypical ecotype, known as the “white” ecotype due to its white nuptial coloration, has recently diverged from the typical “common” caregiving ecotype (Samuk, 2016). The white ecotype has evolutionarily lost paternal care (Blouw and Hagen, 1990; Blouw, 1996). Within minutes of fertilization, male whites spit their offspring out of the nest in a behavior known as embryo “dispersal” (Jamieson et al., 1992; Blouw, 1996). Unburdened by the constraints of parental care, male whites can remain in the “sexual phase,” allowing whites to invest nearly twice as much time into seeking new mates as male commons during the brief summer breeding season (Jamieson et al., 1992; Blouw, 1996). These two ecotypes differ not only in parental behavior and coloration (Blouw and Hagen, 1990; Blouw, 1996) but also in a suite of other traits, including nest architecture (Behrens et al., 2024), courtship behavior (Behrens et al., 2024), and juvenile social behavior (Neumann and Bell, 2023). These behavioral differences are heritable and persist in a common garden (Blouw and Hagen, 1990; Behrens et al., 2024). Given the important role androgens play in coordinating male reproductive physiology and behavior in vertebrates, including stickleback (Mayer et al., 2004), we hypothesized that the two ecotypes exhibit different androgen profiles resulting from differences in how they regulate androgen levels throughout the breeding season. Although we focus on androgens, in part because they are well-studied in stickleback and therefore allow for more detailed predictions (see predictions below), other steroids, such as progestogens and glucocorticoids, may also play a role in the white-common behavioral divergence. Progestogens are associated with male sexual behavior (Witt et al., 1994; Andersen and Tufik, 2006; Ahamed et al., 2024), and glucocorticoids can influence how parents manage the energetic demands and stressors of caregiving (Alonso-Alvarez and Velando, 2012). Because ecotypes differ in mating and care behavior, we might expect ecotypic differences in progestogen levels and glucocorticoid levels as well.

Stickleback exhibit a dissociated reproductive pattern in which spermatogenesis and breeding behavior occur at different times of the year. Spermatogenesis occurs in the fall and winter when androgens are low (Borg, 1982), as androgens can inhibit spermatogenesis (Borg, 1981). Courtship and breeding occur in the summer when the major androgen, 11-ketotestosterone (11-KT) (Mayer et al., 2004), shows a seasonal peak (Mayer et al., 1990a). In mammals, aromatization is often an important step for androgens to exert effects on reproductive behavior; however, in stickleback, it is the non-aromatizable 11-androgens, namely 11-KT, that most strongly influence male secondary sexual characteristics and reproductive behavior (Borg et al., 1993; Borg, 1994). Stickleback do produce aromatizable androgens such as testosterone; however, testosterone is less effective than 11-KT in stimulating secondary sexual characteristics (Borg et al., 1993), and it does not show the dramatic seasonal shifts observed in 11-KT (Mayer et al., 1990a). The role of products of androgen aromatization, like estrogen, in male stickleback is not well understood. Endocrine disruption studies do suggest that synthetic estrogens can negatively influence nesting behavior (Brian et al., 2006) and inhibit the production of spiggin (Andersson et al., 2007), the proteinaceous glue produced in males’ kidneys for nest construction.

Further evidence for the important role of 11-KT in male reproduction comes from previous work that quantified 11-KT levels in breeding males from a Swedish population of parental stickleback. They demonstrated that 11-KT levels vary within a single breeding cycle: levels peak when males are establishing territories and courting females, and decline after fertilization, when males reduce and ultimately cease their courtship efforts as parenting behavior takes over (Páll et al., 2002a). Based on this work, we predicted that whites, which maintain high levels of courtship behavior and do not provide care, do not experience a decline in androgen levels. Instead, we expected that male whites maintain high androgen levels throughout the breeding season. We predicted that male commons experience a decrease in 11-KT levels beginning after fertilization, consistent with the pattern previously documented in Swedish stickleback (Páll et al., 2002a).

Previous work in stickleback has also shown that the decline in 11-KT after fertilization is correlated with a rise in parental fanning behavior and a decline in courtship (Páll et al., 2002a). However, administration of androgens does not affect the performance of these behaviors (Páll et al., 2002b), suggesting that androgens do not directly inhibit parental care or promote courtship. Instead, it is hypothesized that the decline in androgens after mating may reflect a decreased need for spiggin (Mayer et al., 2004). Spiggin production is stimulated by androgens (Borg et al., 1993; Jakobsson et al., 1996, 1999). Therefore, a decline in androgens after mating may be related to a decrease in the demand for spiggin production and nest gluing behavior. To explore this, we tested whether ecotypic differences in gluing behavior mirror ecotypic differences in androgen levels.

Using quantitative liquid chromatography coupled with mass spectroscopy (LC-MS/MS), we quantified five androgens, one estrogen metabolite, two progestogens, and three corticosteroids in males from both ecotypes at three reproductive stages: nesting, immediately after mating (0 days post-fertilization, dpf), and 4 dpf. Sampling at these three stages allowed us to capture the transition that occurs after mating (0 dpf) in commons, from the “sexual phase,” when males prioritize courting females, to the “parental phase,” at 4 dpf, when commons exclusively focus on parenting and show high levels of fanning behavior (Behrens et al., 2024). Whites were collected at the same three stages to compare with commons. Male whites remain receptive to mating across stages and disperse their embryos from the nest at 0 dpf instead of providing care (Behrens et al., 2024). Behaviorally, whites at the 4 dpf stage were similar to whites at the nesting stage, except whites at 4 dpf had successfully mated. Our results prompt hypotheses about what endocrine building blocks may have changed during evolution to construct the divergent reproductive strategies of whites and commons.

2. Methods

2.1. Animals

We caught adult threespine stickleback fish (Gasterosteus aculeatus) of the caregiving common ecotype (“commons”) and non-caregiving white ecotype (“whites”) from two wild populations in Canada during the summers of 2018–2021. We collected 30–50 adults of each ecotype per year. Commons were collected from Cherry Burton Road, New Brunswick, Canada (N 46° 01.516′, W 64° 06.150′), and whites were collected from Canal Lake, Nova Scotia, Canada (N 44° 29′54.0″, W 63° 54′09.1″). Collections were approved by the Canadian Department of Fisheries and Oceans Maritime under permit #343930 held by Dr. Anne Dalziel and Dr. Laura Weir at Saint Mary’s University. These ecotypes are likely experiencing low hybridization rates at sites where they occur in sympatry (Blouw, 1996; Samuk, 2016). Therefore, these two collection sites, where ecotypes generally do not occur in sympatry, were selected to reduce the chances of accidentally collecting hybrids. Embryos were generated at Saint Mary’s University and shipped to the University of Illinois Urbana-Champaign, where fish were raised for at least one generation under common garden conditions in the lab before beginning experiments.

Fish were reared in a recirculating system with brackish water (10 ppt) in single-family, mixed-sex groups with up to 20 fish per tank. Tanks were outfitted with gravel and artificial plants. Fish were reared under temperature and photoperiod conditions that simulated winter and summer (16 °C, 8 L:16 D photoperiod in winter, and 20 °C, 16 L:8 D photoperiod in summer), with a gradual transition between the two conditions. Experiments were conducted over two summers, from May to July 2021 and from June to July 2022, using fish that were sexually mature and approximately one year old. Fish were fed a mixture of frozen bloodworms, mysis shrimp, and cyclops once a day at approximately 7:00 AM. During experiments, feeding time was kept consistent to ensure the fish were satiated, thereby minimizing the potential effects of hunger on behavior and hormone levels. All procedures were approved by the Institutional Animal Care and Use Committee (Protocol #18080 in 2021 and #21031 in 2022).

2.2. Behavior

We visually identified males in breeding condition by their coloration: red throats and blue eyes in commons, and white body coloration in whites. Selected males were moved from group housing to individual tanks (36 L × 33 W × 24 H cm) outfitted with gravel, a fake plant, and nesting materials (filamentous Cladophora algae and a box of sand). We assume that selected males had no prior mating experience, as they were not given access to the necessary nesting materials in group housing. Each male was randomly assigned to one of three stages prior to data collection: nesting, 0 days post-fertilization (0 dpf), or 4 days post-fertilization (4 dpf). Across the two years, we collected data on a total of 22 males per ecotype per stage, except nesting commons, for which we collected 23 males (133 males overall). Once males were relocated to their new tanks, they were left undisturbed overnight to acclimate to their new tanks and start constructing a nest before we began behavioral observations.

The following day, we began courtship trials following the methods of Behrens et al. (2024). Briefly, each male was presented with a gravid female of the same ecotype approximately once per day to check for the presence of a completed nest (for males assigned to the nesting stage) or to encourage males to mate (for males assigned to the 0 dpf and 4 dpf stages). Nesting behavior was quantified in 2021 but not in 2022 due to time constraints. Because behavior data are missing for over half of the nesting males, we focus here on the behavior of males collected at 0 and 4 dpf. A detailed account of ecotypic differences in nesting and other behaviors at these stages is reported in Behrens et al. (2024). Briefly, at the nesting stage, both ecotypes were actively maintaining their completed nests. Males had courted at least one female, but had not yet mated. At 0 dpf, both ecotypes had completed a nest, successfully courted a female, and fertilized one clutch of eggs. Whites had dispersed their embryos from the nest, and commons had begun caring for the embryos in their nest. At this stage, both ecotypes were receptive to mating again, but we did not provide them with additional mating opportunities after the first successful clutch. At 4 dpf, both ecotypes had completed a nest, successfully courted a female, and fertilized one clutch four days prior. Whites had dispersed their embryos from the nest and then resumed nest maintenance behaviors. Commons had cared for their embryos for four days, and by this time, males were exclusively focused on parenting. At this stage, both ecotypes had mating experience and had been isolated from adult conspecifics (i.e., no additional mating opportunities) for four days.

We observed and quantified the behavior of males assigned to the 0 and 4 dpf stages to (1) ensure males were at the correct stage to be euthanized for hormone collection (i.e., had successfully mated at 0 dpf, and were actively caring for embryos at 4 dpf (commons only)) and (2) explore the proposed relationship between gluing behavior and 11-KT (Mayer et al., 2004; Páll et al., 2005). Males assigned to 0 dpf or 4 dpf were exposed to gravid females daily until they successfully mated. When a mating occurred, the female was immediately removed, and males were observed for 15 min. We continued to observe males assigned to the 4 dpf stage for another four days after mating, once a day for 15 min, resulting in a total of five observations per male. We quantified behavior using Jwatcher (Blumstein et al., 2006). In this study, we focus on gluing behavior because gluing may be associated with the decline in androgens in caring stickleback (Mayer et al., 2004; Páll et al., 2005). During gluing, males glide over the nest, tail raised, and deposit threads of a sticky protein (spiggin) to fasten nesting material (Van Iersel, 1953). We quantified gluing by counting the number of times males performed the behavior during the 15-min observation period. Because actual spiggin deposition can be challenging to observe, we counted a glue each time a male glided over his nest, with the characteristic tail-raised posture. Fanning, the primary measure of parenting in stickleback (Van Iersel, 1953), also differs between ecotypes (Behrens et al., 2024), but this behavior is not sensitive to changes in androgens (Páll et al., 2002b), and initial inspection of the hormone data indicated that it was not related to corticosteroids or progestogens either. Therefore, we chose not to present data on fanning behavior in the current study.

All males were euthanized at their assigned stage via rapid decapitation between 9:00 and 13:30 CST to minimize hormonal variations due to time of day. We euthanized nesting males when they had completed their nests but had not mated. A nest was considered complete when males showed nesting behaviors in the presence of a female (i.e., during a courtship trial) (Behrens et al., 2024). We then waited 24 h after males had seen a female to euthanize them, so they were not influenced by the recent experience of courtship. We sacrificed 0 dpf males 45 min after they had successfully courted a female and fertilized eggs. We sacrificed 4 dpf males four days after they had fertilized eggs. For all males, brains were dissected out and set aside for another study, and bodies (the entire carcass minus the brain) were immediately flash-frozen on dry ice and stored at −80 °C for hormone extraction.

2.3. Steroid hormone extraction and quantification

Hormones were extracted from whole bodies (minus the brain) and analyzed via LC-MS/MS. Unlike immunoassays, where the cross-reactivity of antibodies to multiple steroids can confound interpretation, LC-MS/MS measures steroids of interest with great specificity and allows for the measurement of several steroids simultaneously within a single sample (Soldin and Soldin, 2009; Ketha et al., 2014).

Here, we took advantage of LC-MS/MS to quantify a panel of steroids in whites and commons. Due to the small size of our stickleback (in the lab, adult fish typically weigh about one gram), we chose to quantify steroids in bodies instead of plasma. Steroid levels in whole bodies are correlated with plasma (Graham et al., 2018; Nouri et al., 2020), suggesting this method provides a good estimate of circulating hormone levels when a sufficient volume of plasma cannot be collected. However, it is important to keep in mind that steroids in whole-body extracts reflect not only circulating steroids but also steroids present in tissues, including gonads. For example, 11-KA is the most prevalent androgen in testes but is present at low levels in blood, where 11-KA is converted to 11-KT (Borg et al., 1989; Mayer et al., 1990b). Therefore, whole-body levels of 11-KA (and most other androgens) likely primarily reflect testicular levels, whereas 11-KT levels likely primarily reflect blood levels.

Hormone extraction and quantification were performed according to the methods of McGhee et al. (2020), except for the homogenization step, which we modified to accommodate our sample type (whole bodies rather than eggs). Extractions were performed after samples from both years had been collected. To extract hormones, we coarsely chopped the bodies with scissors and ground them into a fine powder using a mortar and pestle with liquid nitrogen. The weight of each sample was recorded. Steroids were then extracted from each body with methanol. We added 4 mL of methanol to the powdered sample, then vortexed for 60 s. Samples were stored overnight at 20 °C to precipitate neutral lipids and proteins. The following day, samples were spun down (2000 rpm, 4 °C) for 20 min. The methanol supernatant was collected and diluted with MilliQ water to bring the volume to 50 mL for each sample. Then, hormones were extracted from the methanol-water solution using solid-phase extraction. C18 Sep-pak cartridges (Waters Cat# WAT020515) were charged with 5 mL of methanol and then rinsed with 5 mL of MilliQ water. Samples were added to the cartridges and pulled through at approximately 2 mL/min with a vacuum. Hormones were eluted with 5 mL of diethyl ether and dried in a fume hood overnight until the ether had evaporated. Dried samples of extracted hormones were stored at −80 °C until analysis.

Samples were analyzed by LC-MS/MS at the Carver Metabolomics Core of the Roy J. Carver Biotechnology Center, University of Illinois Urbana-Champaign, with a targeted steroids assay, as previously described (McQueen et al., 2024). We tested a panel of 35 steroids (see Supplementary Tables 1 and 2 for a complete list of the steroids on this panel). Only free steroids, i.e., unbound to carrier proteins and thought to be biologically active, were quantified. Before extraction, samples were spiked with labeled surrogate internal standards. Following extraction, a 10 μL injection was executed on an Agilent 1290 Infinity II UHPLC system (Agilent, Santa Clara, CA, USA), with a Waters Acquity C18 BEH 2.1 × 150 mm, 1.7um; Column Temp − 60 °C; flow rate 400 μL/min. Mobile phases consisted of 0.1 % formic acid in water (A) and 0.1 % formic acid in acetonitrile (B). Data was collected on a Sciex 6500+ triple quadrupole mass spectrometer (Sciex, Framingham, MA, USA). Data was acquired in positive mode using Sciex Analyst v1.7.3. Peak integration and quantitation were performed using calibration curves adjusted for internal standards in MultiQuant v3.1. Steroid chemical standards used to create calibration curves for quantification and as internal standards were acquired from Cayman Chemical (Ann Arbor, MI, USA) and Sigma-Aldrich (St. Louis, MO, USA).

2.4. Data analysis

Statistical analyses and data visualization were conducted using R Version 4.4.2 (R Core Team, 2024). Steroids were converted to concentrations (ng of steroid per gram of fish homogenate) based on the mass of each fish homogenate recovered after grinding it into a powder for extraction. We focused our analyses on steroids that were detected in >50 % of males per ecotype per stage per year. Of the 35 steroids in our panel, 11 met this criterion. This included five androgens (androstenedione (A4), 11-OH-androstenedione (11-OHA4), testosterone (T), 11-ketoandrostenedione (11-KA), 11-ketotestosterone (11-KT)), one estrogen metabolite (2-methoxyestradiol (2-ME2)), two glucocorticoids (cortisol, cortisone), one mineralocorticoid (aldosterone), and two progesterone metabolites (i.e., progestogens, 11β-OH-progesterone (11β-OHP), pregnanolone). Sample sizes for each steroid for each year, ecotype, and stage are provided in Supplementary Tables 1 and 2. One individual, a white collected at 4 dpf in 2021, was a statistical outlier across all 11 hormones and was therefore excluded from the analysis and the Supplementary Tables. When an individual had missing data for a steroid (refer to Supplementary Tables 1–2 and figure captions for sample sizes), the missing value was treated as “NA”; we did not perform missing value imputation. For the 11 steroids that met our criterion, one common was missing a value for 11-OHA4, two whites were missing values for pregnanolone, and three whites and one common were missing values for 11β-OHP. Missing data occurred most often for hormones with low median values, suggesting that missing values reflect hormone levels below the LC-MS/MS detection threshold.

First, we asked how these 11 steroids varied across reproductive stages and between ecotypes. All hormones were first log-transformed, except 11-KT, A4, and 11β-OHP, which were square-root transformed so that residuals would better fit a normal distribution. Note that although steroids were transformed for analysis, all figures present the untransformed values to allow for easier interpretation and visual comparison among steroids. We ran separate linear mixed models for each hormone, including ecotype, stage, and their interaction as fixed effects and year as a random effect using the lmer function in the lme4 package (Bates et al., 2015). When the interaction between ecotype and stage was non-significant, it was dropped from the model. We tested for the significance of fixed effects using the anova function in the lmerTest package (Kuznetsova et al., 2017) with the Satterthwaite method to estimate degrees of freedom. P-values from these final models were corrected using the Bonferonni-Holm method to account for the fact that we ran multiple separate models on steroids measured in the same individuals. All model p-values remained significant after correction, so we proceeded without correcting the p-values from linear mixed models (i.e., p-values reported in Tables 1–3). We investigated significant effects with post hoc pairwise comparisons using estimated marginal means (EMM) with the emmeans function in the emmeans package (Lenth, 2024). The resulting p-values from post hoc tests were adjusted for multiple testing using the Tukey method. Cohen’s d effect sizes were calculated for each pairwise comparison using the eff_size function in the emmeans package. Post hoc test results and effect sizes are reported in Supplementary Tables 3–5.

Table 1.

Results from linear mixed models testing for effects of ecotype and reproductive stage on the levels of five androgens and one estrogen. Hormone levels were transformed (A4 and 11-KT were square-root transformed and all others were log-transformed) before analysis to better fit the assumptions of normality, and models included a random effect of year because data were collected over two years. P-values from all models survived correction; here, we present uncorrected p-values. Significant p-values (p < 0.05) are bolded. Degrees of freedom (numerator, denominator) were estimated using the Satterthwaite method. Post hoc test results are reported in Supplementary Table 3.

Steroid	F	df	P
Androstenedione (A4)
ecotype	29.041	1, 125.02	<0.001
stage	3.530	2, 125.01	0.032
ecotype*stage	6.967	2, 125.00	0.001
11-OH-Androstenedione (11-OHA4)
ecotype	64.870	1, 124.00	<0.001
stage	10.175	2, 124.00	<0.001
ecotype*stage	17.151	2, 124.00	<0.001
11-Ketoandrostenedione (11-KA)
ecotype	0.388	1, 128.00	0.534
stage	4.579	2, 128.00	0.012
11-Ketotestosterone (11-KT)
ecotype	33.201	1, 125.00	<0.001
stage	5.576	2, 125.00	0.005
ecotype*stage	9.321	2, 125.00	<0.001
Testosterone (T)
ecotype	26.068	1, 125.01	<0.001
stage	7.450	2, 125.01	0.001
ecotype*stage	6.792	2, 125.00	0.002
2-Methoxyestradiol (2-ME2)
ecotype	58.453	1, 125.00	<0.001
stage	9.075	2, 125.00	<0.001
ecotype*stage	15.882	2, 125.00	<0.001

Table 3.

Results from linear mixed models testing for effects of ecotype and reproductive stage on the levels of three corticosteroids. Hormone levels were log-transformed before analysis to better fit the assumptions of normality, and models included a random effect of year because data were collected over two years. P-values from all models survived correction; here, we present uncorrected p-values. Significant p-values (p < 0.05) are bolded. Degrees of freedom (numerator, denominator) were estimated using the Satterthwaite method. Post hoc test results are reported in Supplementary Table 5.

Steroid	F	df	P
Cortisol
ecotype	0.762	1, 127.03	0.385
stage	10.806	2, 127.03	<0.001
Cortisone
ecotype	1.073	1, 128.00	0.302
stage	4.901	2, 128.00	0.009
Aldosterone
ecotype	0.412	1, 127.02	0.522
stage	4.511	2, 127.01	0.013

Given the multivariate nature of our data, we conducted several exploratory analyses to determine if a multivariate approach was appropriate before ultimately selecting the univariate approach presented here. We tested for correlations among steroid levels, predicting that there would be strong positive correlations within steroid classes (e. g., androgen levels highly positively correlated with each other, and so on). The results generally confirmed our expectations (an exception with 11-KA is described in the Results). We also tested for correlations within different subsets of data (e.g., visually comparing correlation plots in commons at the nesting stage to whites at the nesting stage). We found that these were typically consistent with the results from data pooled across ecotypes, stages, and years (one interesting exception is described in the Supplement). These correlation results are reported in the Supplement (Supplementary Figs. 1–2). We conducted an exploratory multivariate analysis using Principal Components Analysis (PCA) to reduce the dimensionality of the data, but found that combining data across different hormones made it difficult to interpret ecotype and stage effects on specific hormones of interest (such as 11-KT and 11-KA). We also attempted a MANOVA but found this inappropriate for our data, as the number of response variables exceeded the sample size per group when year, ecotype, and stage were included in the model. Ultimately, the univariate approach was the best fit for our data and questions.

Although we take a univariate approach here, we acknowledge that steroids are interdependent, as each steroid serves as a precursor and metabolite in steroidogenesis. Steroids can also interact with one another. For example, the behavioral effects of progestogens —such as whether they inhibit the actions of androgens or act synergistically to promote male sexual behavior —may depend on the relative levels of progestogens to androgens (Wagner, 2006). Therefore, we took advantage of the fact that several androgens and progestogens were measured in the same individuals in our study to calculate androgen and progestogen ratios (see Sections 3.1 and 3.3). Linear mixed models were used to test for the effects of ecotype, stage, and their interaction on ratios. Ratios were log-transformed to better fit a normal distribution. Post hoc tests were performed as previously described. Because we view the ratio analysis as exploratory, we only briefly mention the results here; the full results are provided in the Supplement (11-KA ratios in Supplementary Fig. 3 and Supplementary Tables 6–7, and progestogen-to-androgen ratios in Supplementary Fig. 4 and Supplementary Tables 8 and 9).

Next, we leveraged behavioral data from the males euthanized at 4 dpf to test the hypothesis that 11-KT levels are associated with levels of gluing behavior, as previously proposed (Mayer et al., 2004; Páll et al., 2005). First, we explored how gluing behavior changed over time in whites and commons, predicting that if gluing behavior followed 11-KT levels, gluing would decline over time in commons but remain steady in whites. To test this, we ran a negative binomial generalized linear mixed model testing for effects of ecotype, days post-fertilization (dpf), and their interaction on the number of times a male glued from 0 to 4 dpf. We included a random effect of year and a random effect of male ID to account for repeated measures. Gluing behavior was measured five times in each male, from 0 dpf, immediately after mating, until 4 dpf, the day males were euthanized. Significant effects were investigated with post hoc tests as described above. Post hoc test results and effect sizes from the analysis of gluing behavior are reported in Supplementary Table 10.

Finally, we examined relationships between gluing and 11-KT at the individual level by testing for correlations between these two variables using the cor.test function in the base stats package with the Spearman method. We focus on gluing behavior in males collected at 0 and 4 dpf. Because gluing behavior changes over time in commons and differs between ecotypes (see above), we ran separate correlations for each ecotype and stage.

3. Results

3.1. Levels of androgens (except 11-KA) declined as commons began providing care, but remained high in whites

We predicted that levels of androgens would decline after mating in the caregiving common ecotype and remain high in the non-caregiving white ecotype. Consistent with this prediction, levels of four androgens (T, 11-KT, A4, 11-OHA4) and one estrogen metabolite (2-ME2) declined across reproductive stages after fertilization in commons, but not whites (Table 1, Fig. 1). In commons, levels of these hormones were significantly higher when males were nesting than when parenting (Supplementary Table 3). Additionally, levels of T were greater in nesting commons than post-mating commons that were just beginning to provide care (Supplementary Table 3), and levels of A4, 11-OHA4, and 2-ME2 were greater in commons immediately after mating than when parenting (Supplementary Table 3).

Fig. 1.

Levels of four androgens and one estrogen metabolite declined in commons after mating; these steroids remained elevated across all stages in whites. 11-KA was detected at much higher levels than other androgens (note the scale on the Y axis) and showed a different pattern across stages. Levels of each hormone (A-F) are plotted beside a proposed steroidogenic pathway for teleosts, focusing on the six steroids of interest. A4 ➔ 11-OHA4 ➔ 11-KA ➔ 11-KT is the main steroidogenic pathway in stickleback testes (Borg et al., 1989), although the final step (11-KA ➔ 11-KT) primarily occurs in blood (Mayer et al., 1990b). Androstenedione can be metabolized into other steroids in stickleback, including T (Borg et al., 1992). One of these alternative pathways in teleosts (A4 ➔ estrone ➔ T) is shown here (Tenugu et al., 2021). To our knowledge, the biosynthesis of 2-ME2 from 17β-estradiol is not established in fishes. Therefore, here we show the proposed pathway for mammals (Kumar et al., 2016). Steroids measured in our study are shown in bold font in the pathway; steroidogenic enzymes are in gray italic font. Sample sizes for all hormones except 11-OHA4 were the same: n = 23 nesting commons, n = 22 nesting whites, n = 22 0 dpf commons, n = 22 0 dpf whites, n = 22 4 dpf commons, and n = 21 4 dpf whites. For 11-OHA4, due to missing data for one male, n = 21 for 4 dpf commons. For this and subsequent graphs, boxplots show the median (thick black bar), the first and third quartiles (bottom and top edges of box), and 1.5 times the interquartile range (whiskers). Commons are shown in orange, and whites are shown in gray. Each point represents an individual male. Different letters denote significant differences (p < 0.05) between groups in estimated marginal means (EMM) post hoc tests after Tukey correction.

In contrast, levels of these five hormones (T, 11-KT, A4, 11-OHA4, and 2-ME2) remained elevated across stages in whites, resembling the levels detected in nesting commons (Table 1, Fig. 1). Whites had higher levels of 11-OHA4 and 2-ME2 than commons after mating (Supplementary Table 3) and higher levels of all five hormones at 4 dpf when commons were parenting (Supplementary Table 3).

Levels of 11-KA were approximately 10× higher than levels of the other detected androgens. For example, the median level of 11-KA detected across all males was 592.91 ± 39.12 compared to 43.88 ± 4.53 for 11-KT (median ± standard error ng/g). 11-KA was also the only androgen for which we did not detect a significant interaction between ecotype and stage (Table 1, Fig. 1). Androgens upstream and downstream of 11-KA showed a decline in commons after mating. However, levels of 11-KA did not follow this pattern. Indeed, 11-KA levels were only weakly correlated with the other androgens (Supplementary Fig. 1). 11-KA levels differed across stages but not between ecotypes (Table 1); 11-KA was higher at the nesting stage than at 0 dpf in both ecotypes (Supplementary Table 3).

To explore why 11-KA levels do not differ between ecotypes at 4 dpf, but its precursor (11-OHA4) and metabolite (11-KT) show a decline across stages in commons, we calculated ratios of 11-KA relative to 11-OHA4 and 11-KT. The ratios suggest that although levels of 11-OHA4 decline in commons at 4 dpf, resulting in a reduction in the ratio of 11-OHA4 to 11-KA as less 11-KA is produced, there is also a reduction in the conversion of 11-KA to 11-KT, reflected in the increased ratio of 11-KA to 11-KT across stages. This reduction in both the conversion to 11-KA from 11-OHA4 and from 11-KA to 11-KT may explain the apparent lack of change in 11-KA levels at 4 dpf in commons (Supplementary Tables 6 and 7, Supplementary Fig. 3).

3.2. A link between 11-KT and gluing behavior

11-KT is known to promote spiggin (glue) production in stickleback (Borg et al., 1993; Jakobsson et al., 1996, 1999). Therefore, we expected the decline in androgens over time in commons to be associated with a decline in gluing behavior. To explore this, we leveraged behavioral data on individual males observed once daily from 0 dpf to 4 dpf.

We detected a significant interaction between ecotype and stage (dpf) on levels of gluing behavior (Fig. 2A: Type III Wald Chi-square = 16.370, df = 4, p = 0.003); gluing behavior declined after mating in commons but not whites. Common males glued less at 4 dpf than at 0 dpf, but levels of gluing behavior in male whites did not significantly change over time (Supplementary Table 10). Commons tended to glue less than whites at 4 dpf (Supplementary Table 10), which is consistent with the ecotypic differences in 11-KT at 4 dpf.

Fig. 2.

In the common ecotype, gluing behavior declined over time as 11-KT levels decreased, and gluing behavior was positively correlated with 11-KT at 4 dpf. (A) Data show the number of glues performed by each male during a once-daily 15-min observation. Gluing behavior declined over time in commons but not in whites. Each thin line represents the gluing behavior of a single male tracked from the day they mated (0 dpf) until males were euthanized and collected for hormone quantification (4 dpf). Thick lines represent ecotype averages over time. Only the behavior of males euthanized at 4 dpf is shown, as the other males were sacrificed at or before 0 dpf. (B) Commons with higher 11-KT at 4 dpf glued their nests more; no significant relationship was detected in whites at 0 or 4 dpf, or commons at 0 dpf. Each point represents one male, and the gray shadow represents 95 % confidence intervals.

We examined patterns at the individual level to further explore a possible link between levels of 11-KT and gluing behavior. We tested for correlations between gluing and 11-KT levels in males of each ecotype at 0 dpf and 4 dpf. Commons with higher 11-KT glued more at 4 dpf, but not at 0 dpf (Fig. 2B: 0 dpf, rho = −0.078, p = 0.730, n = 22; 4 dpf, rho = 0.487, p = 0.022, n = 22), providing some support that 11-KT promotes spiggin production and gluing in commons. We did not detect a relationship between 11-KT and gluing in whites (Fig. 2B: 0 dpf, rho = −0.076, p = 0.738, n = 22; 4 dpf, rho = 0.389, p = 0.082, n = 21).

3.3. Levels of progestogens were generally lower in whites and after mating (0 dpf)

Progestogens may promote sexual behavior (Witt et al., 1994; Andersen and Tufik, 2006; Ahamed et al., 2024). Therefore, we predicted progestogen levels would differ between male whites and commons, which prioritize mating differently across stages. The two progestogens we detected, pregnanolone and 11β-OHP, showed similar but not identical patterns across stages and ecotypes (Table 2, Fig. 3). Pregnanolone levels were lower in whites than commons (Table 2, Fig. 3B inset). 11β-OHP levels were also lower in whites than commons (Fig. 3A), but this effect was detected only at the nesting stage and at 4 dpf (Supplementary Table 4). Across stages, pregnanolone levels were lower at 0 dpf than at the other two stages (Supplementary Table 4). Levels of 11β-OHP were also lower at 0 dpf than the other two stages, but only in the common ecotype (Supplementary Table 4). Levels of 11β-OHP did not significantly differ across stages in whites.

Table 2.

Results from linear mixed models testing for effects of ecotype and reproductive stage on the levels of two progestogens. Hormone levels were transformed (11β-OHP was square-root transformed and pregnanolone was log-transformed) before analysis to better fit the assumptions of normality, and models included a random effect of year because data were collected over two years. P-values from all models survived correction; here, we present uncorrected p-values. Significant p-values (p < 0.05) are bolded. Degrees of freedom (numerator, denominator) were estimated using the Satterthwaite method. Post hoc test results are reported in Supplementary Table 4.

Steroid	F	df	P
11β-OH-Progesterone (11β-OHP)
ecotype	17.984	1, 121.00	<0.001
stage	17.564	2, 121.07	<0.001
ecotype*stage	3.108	2, 121.02	0.048
Pregnanolone
ecotype	12.043	1, 125.00	0.001
stage	16.560	2, 125.01	<0.001

Fig. 3.

Progestogens differed between ecotypes and across stages. (A) 11β-OH-Progesterone (11β -OHP) showed a significant interaction effect between stage and ecotype. (B) Pregnanolone showed a main effect of stage (main plot) and ecotype (inset). Different letters denote significant differences between groups. Due to missing data, sample sizes for these two hormones differed slightly. For 11β -OHP n = 23 nesting commons, n = 22 nesting whites, n = 22 0 dpf commons, n = 19 0 dpf whites, n = 21 4 dpf commons, and n = 21 4 dpf whites. Sample sizes were the same for pregnanolone except for 0 dpf whites (n = 20) and 4 dpf commons (n = 22).

We also explored progestogen-to-androgen ratios (levels of 11β-OHP and pregnanolone relative to 11-KT) because the effects of progestogens on male sexual behavior may depend on androgen levels. We calculated the progestogen-to-androgen ratio by dividing the levels of each progestogen (11β-OHP and pregnanolone) by 11-KT, the major androgen in stickleback (Mayer et al., 2004). Progestogen-to-androgen ratios were low when levels of male sexual behavior were high, i.e., across all stages in whites and at the nesting and 0 dpf stages in commons (Supplementary Table 8, Supplementary Fig. 4). In contrast, progestogen to androgen ratios were relatively high (approximately 7× higher) when levels of sexual behavior were low, i.e. at 4 dpf in parenting commons, which had low levels of 11-KT and high levels of progestogens (Supplementary Table 7, Supplementary Fig. 4). This difference in relative progestogen and androgen levels in whites and commons at 4 dpf likely explains the interesting observation that correlations between androgens, 11-KA, and progestogens flip at this stage in commons (Supplementary Fig. 2). In commons and whites, males with higher levels of T, 11-KT, A4, 11-OHA4, and 2-ME2 also generally have higher levels of 11-KA and progestogens except for commons at 4 dpf, when levels of T, 11-KT, A4, 11-OHA4, and 2-ME2 decline, and males with higher levels of these steroids have lower levels of 11-KA and progestogens. This provides evidence that the relationship between progestogens and androgens changes in parenting commons in a way that is unique to this ecotype and could support their transition to parenting.

3.4. Corticosteroids were elevated after mating (0 dpf) in both ecotypes

Parenthood is associated with changes in energy homeostasis and stress response, which are regulated by glucocorticoids (Milla et al., 2009). Therefore, we predicted that commons may show elevated glucocorticoids relative to whites at 0 and/or 4 dpf, if corticosteroids are involved in mobilizing energy for parenting. Alternatively, whites may show elevated glucocorticoids relative to commons at 0 dpf if a stress response triggers offspring abandonment in whites. We found no evidence to support either prediction.

We detected two glucocorticoids (cortisol and cortisone) and one mineralocorticoid (aldosterone). Unlike the androgens and progestogens, we did not detect a significant interaction between ecotype and stage for any corticosteroid, but we did detect a significant main effect of stage (Table 3). Levels of all three corticosteroids were significantly elevated in both ecotypes after mating compared to 4 dpf (Supplementary Table 5, Fig. 4). For cortisol only, levels were also significantly higher at 0 dpf than at the nesting stage (Supplementary Table 5).

Fig. 4.

Corticosteroid levels were elevated after mating in both ecotypes. Patterns were similar across all three corticosteroids detected: (A) cortisol, (B) cortisone, and (C) aldosterone. Levels of corticosteroids did not differ significantly between ecotypes at any stage; different letters denote significant differences between stages, pooled across ecotypes. Sample sizes for the three hormones were the same: n = 23 nesting commons, n = 22 nesting whites, n = 22 0 dpf commons, n = 22 0 dpf whites, n = 22 4 dpf commons, and n = 21 4 dpf whites.

4. Discussion

4.1. Evidence for ecotypic differences in androgen regulation

Androgens play an important role in male reproductive behavior and trade-offs between parental care and mating (Wingfield et al., 1990; Oliveira et al., 2002). To understand how androgens and other steroids are associated with the evolution of different reproductive strategies, we can compare hormone levels in closely related, behaviorally divergent populations. Here, we take advantage of a pair of recently diverged threespine stickleback ecotypes to gain insights into how endocrine building blocks, specifically steroid levels, may have changed in the evolutionary loss of parental care. We predicted that males of the white ecotype, which show high rates of courtship behavior (Haley et al., 2019; Behrens et al., 2024) and abandon their offspring soon after fertilization (Blouw and Hagen, 1990; Blouw, 1996), would maintain consistently high androgen levels. We predicted that males of the common ecotype, which stop courting females a few days after mating to focus exclusively on caring for their offspring (Jamieson et al., 1992), would show a decline in androgens after mating. Consistent with our hypothesis, we found that levels of most, but not all, androgens declined in commons after mating and remained high in whites across reproductive stages. These ecotypic differences in how androgen levels change over time may reflect how each ecotype balances the costs and benefits of androgen-mediated traits and the trade-off between mating and parental care. Commons, which invest substantial time and energy into parenting, likely benefit from reducing androgen levels during parenting to mitigate the deleterious effects of androgens on immunity and other traits. In contrast, for the white ecotype, the benefits of maintaining elevated androgen levels and sustaining their mating efforts likely outweigh any androgen-associated costs (see Section 4.2).

Consistent with our predictions and previous work in stickleback (Páll et al., 2002a), 11-KT declined in commons after mating as they transitioned into caregiving. 11-KT remained elevated in whites across stages. Ecotypic differences in androgen regulation may support their different reproductive trajectories. Except for 11-KA, all other androgens (A4, T, 11-OHA4) and one estrogen metabolite (2-ME2) downstream of A4 followed this pattern, which suggests that at least some of these changes occurred early in the steroidogenic pathway in the gonad and/or upstream of the gonad. This pattern prompts at least four non-mutually exclusive evolutionary changes that could contribute to the observed population-level difference in androgen levels.

First, changes in gonadotropin signaling could have contributed to the observed ecotypic differences in androgen levels. Gonadotropin pituitary hormones — luteinizing hormone (LH) and follicle-stimulating hormone (FSH) — promote androgen production in the testes (Planas, 1995). Previous work in stickleback has demonstrated that after mating, parental male stickleback show declines in the levels of FSH and LH (Páll et al., 2005). Caregiving commons may show this pattern, in which androgen-promoting signals from the pituitary decline after mating, leading to reduced androgen production. Evolutionary changes in whites’ pituitaries may cause them to maintain LH and/or FSH signaling post-mating, resulting in consistently high androgen production in the testes.

Second, another possibility is that the site of divergence between whites and commons lies in the brain. There might be differences in androgen receptor affinity, binding, or density in the hypothalamus, for example, leading to ecotypic differences in gonadotropin release from the pituitary and, ultimately, androgen levels. Previous work has also found that whites and commons differ in the activation of specific populations of hypothalamic neurons. Male commons show increased activation of preoptic oxytocin neurons after mating, but whites do not (Maciejewski et al., 2025). Commons also have higher pituitary oxytocin peptide levels (Bell et al., 2025). Oxytocin neurons in the preoptic area of the hypothalamus project to the pituitary and can influence gonadotrophs (Trudeau et al., 2024). Therefore, it is possible that activation of preoptic oxytocin neurons in commons after mating leads to changes in gonadotroph signaling, and ultimately a reduction in androgens. Whites have lost this change in oxytocin activation, which could contribute to a loss of change in gonadotroph activity, and ultimately the maintenance of elevated androgens after mating.

Third, it is also possible that whites and commons have diverged at the level of the sensory systems that detect cues and trigger parental behavior. Previous work in stickleback suggests that changes in androgen levels after mating may be induced by mating or cues from eggs, as males that mate with more females and have more eggs show more rapid declines in 11-KT levels after mating (Páll et al., 2002a). This suggests that external cues may trigger the post-mating decrease in androgens observed in caregiving commons. The lack of androgen decline in whites could result from a change in how male whites respond to external cues —for example, a change in the white ecotype’s olfactory system that alters the ability to detect egg cues. Or, perhaps female whites produce eggs that lack the necessary olfactory cues to trigger a response in males.

Fourth, there is recent evidence that steroidogenic enzymes in blood are an underappreciated but potentially potent mechanism that can modulate the effects of gonadal steroids on the brain and contribute to behavioral variation. For example, ruffs exhibit alternative mating strategies that harbor alleles differing in the efficiency of the 17-beta hydroxysteroid dehydrogenase 2 (hsd17b2) enzyme, which converts testosterone to androstenedione (Loveland et al., 2025). The behavioral strategies characterized by low testosterone (satellites and faeders) harbor allelic variants of the hsd17b2 enzyme that are more efficient and, thereby, more effective in preventing testosterone from reaching the brain. Interestingly, hydroxysteroid dehydrogenase converts 11-KA to 11-KT in blood in male stickleback (Mayer et al., 1990b). The decline in 11-KT (and the increase in the ratio between 11-KA and 11-KT) in commons after mating suggests that commons may have lower hydroxysteroid dehydrogenase activity in blood after mating. Whites may maintain high 11-KT through higher enzymatic expression and/or a more efficient form of the enzyme than commons due to genetic variation at that locus, as in the ruffs (Loveland et al., 2025). Future work quantifying steroidogenic enzyme expression in blood and testes and comparing allelic variation in steroidogenic enzymes between ecotypes can shed more light on the origins of ecotypic differences in androgen levels and the lack of ecotypic differences in 11-KA.

Changes at all these levels of biological organization are not mutually exclusive; indeed, several mechanisms may be at play. For example, the Japan Sea and Pacific Ocean ecotypes of stickleback that show divergent courtship behavior (Kitano et al., 2008) and 11-KT levels (Kitano et al., 2011) differ in the expression of LH receptor and steroidogenic enzymes in testes (Kitano et al., 2011). Likewise, a combination of mechanisms may be needed to explain how steroid levels have diverged in whites and commons. Interactions between androgens and other steroids should also be considered (for example, progestogens, see Section 4.3). Future studies should explore these and other possible modulators of androgen levels to understand how commons reduce their androgen levels after mating, whereas whites do not.

4.2. Are differences in androgen regulation related to the need for nest glue (spiggin)?

Changes in 11-KT levels (or a lack thereof in whites) may be related to ecotypic differences in the need for spiggin (Mayer et al., 2004). After mating, caregiving males show decreased courtship, decreased gluing, and increased parental fanning (Páll et al., 2002a, 2005; Behrens et al., 2024). Male whites maintain courtship and gluing behavior after mating and display no parental fanning (Jamieson et al., 1992; Behrens et al., 2024). The negative correlation between 11-KT levels and parenting behavior may initially lead us to conclude that high androgens are incompatible with parental care, as is the case for several species of birds (Silverin, 1980; Hegner and Wingfield, 1987; Oring et al., 1989; Ketterson et al., 1992; De Ridder et al., 2000; Stoehr and Hill, 2000). However, androgen manipulation studies have shown that while spiggin production is androgen-mediated (Páll et al., 2005), fanning and courtship are not (Páll et al., 2002b). We found that gluing behavior declined in commons after mating as they transitioned into parenting, and commons with higher 11-KT glued more at 4 dpf, consistent with work in a different stickleback population (Páll et al., 2005). For the first time, we show that the positive relationship between gluing behavior and 11-KT levels also holds for the white ecotype; whites maintained steady levels of both across stages. This suggests that differences in androgen levels after mating in whites and commons may be related to the costs of maintaining androgens and androgen-mediated traits, like spiggin or coloration (Borg et al., 1993; Jakobsson et al., 1996, 1999; Kurtz et al., 2007). Spiggin is energetically costly to produce (Östlund-Nilsson, 2001), and bright nuptial coloration increases males’ risk of predation (Johnson and Candolin, 2017). Androgens can also negatively impact immunity (Kurtz et al., 2007). Reducing androgen levels after mating may benefit males by minimizing these costs. In the “parental phase,” commons no longer require bright coloration to attract mates or spiggin to build and maintain nests after mating. In fact, parental stickleback dismantle their nests in the days following mating, presumably to allow for increased embryo ventilation (Van Iersel, 1953). Minimizing energetically demanding androgen-mediated traits, such as spiggin, may also enable males to reallocate resources to parental care, which is itself energetically costly (Smith and Wootton, 1999). Unlike commons, male whites remain in the “sexual phase” after mating. Whites need to maintain their coloration to attract females and need spiggin to keep their nests ready for potential mates (Jamieson et al., 1992; Blouw, 1996). For whites, the benefits of increased reproductive success conferred by maintaining high androgens likely outweigh the costs (Wingfield et al., 2001).

4.3. Low progestogen levels may promote sexual behavior

In addition to androgens, we quantified two metabolites of progesterone (pregnanolone and 11β-OH-progesterone (11β-OHP)). Both metabolites showed relatively consistent differences between ecotypes and across stages. Levels were generally lower in whites than commons and were lowest at 0 dpf in both ecotypes. This is consistent with a growing literature that implicates progesterone in male reproduction and androgen-dependent sexual behavior (Witt et al., 1994; Andersen and Tufik, 2006). Generally, the effects of progesterone on male sexual behavior appear to be dose-dependent; low doses of progesterone can facilitate male sexual behavior, and high doses can inhibit it (Wagner, 2006). Consistent with this, the progestogens we detected were lowest in both ecotypes immediately after mating (0 dpf), suggesting that these metabolites or their precursors could promote mating behavior when present at low levels.

Interestingly, the effects of progesterone may depend on the relative levels of androgens. Work in green anoles (Young et al., 1991) and zebrafish (Ahamed et al., 2024) suggests that male sexual behavior is facilitated by progesterone only when progesterone levels are low relative to androgen levels. Our LC-MS/MS panel enabled us to explore this idea by measuring both 11-KT and progestogens in each male. Our data support the idea that a low progestogen-to-androgen ratio facilitates sexual behavior, and a high ratio may inhibit it. Progestogen metabolite levels were low relative to 11-KT levels across all stages in whites and at the nesting and 0 dpf stages in commons, times when males are expected to show sexual behavior. In commons at 4 dpf, when sexual behavior subsides and parental behavior dominates (Jamieson et al., 1992; Behrens et al., 2024), progestogens were high relative to 11-KT. Perhaps, ecotypic differences in the balance of progestogens and androgens may help whites to keep sexual behavior “on” across stages, and help commons to turn sexual behavior “off” after 0 dpf as commons transition to caregiving. It should be acknowledged that these ratios were calculated using progesterone metabolites, not DHP, the major progestogen in teleosts (Scott et al., 2010); therefore, further work is needed to test whether these patterns hold for DHP. Simultaneously manipulating 11-KT and progestogen levels in breeding males would be a powerful approach to determine how concerted action by these steroids may influence ecotypic differences in male sexual behavior.

4.4. No evidence for ecotypic differences in energy homeostasis or stress response

Corticosteroids, unlike androgens and progestogens, showed no differences between ecotypes but did differ across stages. Glucocorticoids mediate energy homeostasis and stress response in fishes (Milla et al., 2009), so the lack of ecotypic differences in corticosteroids suggests that the evolutionary loss of care in whites is not associated with changes in energy homeostasis or the stress response.

Parenting is energetically demanding and physiologically costly (Smith and Wootton, 1999; Alonso-Alvarez and Velando, 2012), and in mammals, particularly mothers, the onset of parenting is associated with changes in stress response and energy homeostasis (Kaplan et al., 2025). Evidence in fathers is mixed (Saltzman and Ziegler, 2014). There is also evidence in birds (Kotrschal et al., 1998; Golet et al., 2004), frogs (Fischer and O’Connell, 2020), and fish (Magee et al., 2006) for elevated glucocorticoid levels in parents relative to non-parents. Because high glucocorticoid levels may support energy mobilization to meet the demands of caring, we might expect commons to show higher glucocorticoid levels than whites at 4 dpf, when commons’ parental effort peaks (Behrens et al., 2024). Alternatively, if high glucocorticoids reflect high stress, which can lead to offspring abandonment (Kaplan et al., 2025), we might expect whites to show higher corticosteroids than commons after mating when whites spit their offspring out of the nest. Contrary to our expectations, cortisol and cortisone were elevated in both ecotypes at 0 dpf. This suggests that glucocorticoid levels may respond to a shared experience by whites and commons, such as mating, as shown in mammals (Borg et al., 1991; Villani et al., 2006), and/or courtship. Previous work in stickleback found that cortisol levels were elevated in socially isolated stickleback after brief exposure to a conspecific (Bell et al., 2007). Perhaps the elevated glucocorticoids in our fish were a response to their interaction with a female during courtship, as males were otherwise housed alone.

Also worth noting is that aldosterone, a mineralocorticoid considered absent, or at least not present at any significant level, in fishes (Gilmour, 2005; Milla et al., 2009), was detected in all the males in our study, a testament to the power of LC-MS/MS. Aldosterone has been detected in fishes in at least one other LC-MS/MS study (Nouri et al., 2020). This warrants further investigation to determine what role aldosterone plays in the physiology and behavior of stickleback and fishes more generally.

5. Conclusions

Here, we describe differences in steroid levels across reproductive stages in two behaviorally divergent ecotypes of threespine stickleback as a first step toward understanding how endocrine mechanisms evolve when parental care is lost. Leveraging LC-MS/MS, we quantified a panel of eleven steroids. We demonstrate that this behavioral divergence was not limited to a single key hormone, such as 11-KT. Rather, seven of eleven measured steroids differed between ecotypes, including four androgens, one estrogen metabolite, and two progestogens. Corticosteroid levels differed across stages, peaking after mating, but did not differ between ecotypes, suggesting that, in the evolution of reproductive behavior, steroids in the HPG axis may be more likely to be targeted than those in the hypothalamic-pituitary-adrenal (HPA) or hypothalamic-pituitary-interrenal (HPI) axes in fishes. However, further research is required to determine if other aspects of corticosteroid signaling, such as crosstalk between the HPG and HPI (Rosvall et al., 2016) or corticosteroid receptors, may differ between ecotypes.

We found that androgen levels declined in commons across stages but remained elevated in whites, suggesting that the regulation of androgen levels over time has diverged between ecotypes. Although androgens showed the expected negative relationship with care behavior in commons, androgens are not causally related to care in stickleback (Páll et al., 2002b); therefore, the loss of care in whites is unlikely to be a direct result of high androgens. But consistently high androgen levels in whites may support the maintenance of sexual behavior as 11-KT does promote nuptial coloration, spiggin production, and possibly gluing behavior (Borg et al., 1993; Jakobsson et al., 1996, 1999; Kurtz et al., 2007). Progestogen levels also remain low in whites across stages, which could, alongside high androgen levels, facilitate the maintenance of high levels of sexual behavior. Although our work suggests many interesting relationships between steroids and the loss of care, a significant limitation of our study is its correlational design. To establish causal relationships between gluing and androgens, more work is needed. For example, pharmacological manipulations of androgens and/or experimental removal of clutches can be used in future studies to test for effects on gluing behavior. Future work will also explore potential mechanisms underlying ecotypic differences in androgens, such as variation in hormone receptors, gonadotropin levels, and steroidogenic enzyme expression, including hsd17B. This will ultimately further our understanding of how endocrine building blocks are modified by evolution in the early stages of behavioral divergence.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.yhbeh.2025.105866.

Data availability

Data will be made available on request.