Alteration of reproductive behaviors by aromatase inhibition is population-dependent in an African cichlid fish

Although hormones are vital to an organism’s ability to respond to environmental stressors, they can be directly altered by the environment and impact reproductive behavior. For example, in some fishes, aquatic hypoxia (low dissolved oxygen) inhibits the aromatase enzyme that converts testosterone to estradiol. Here, we examined the effects of short-term aromatase inhibition on reproductive behavior in male Pseudocrenilabrus multicolor, a widespread African cichlid, from one normoxic river population and one hypoxic swamp population. We further tested the response of females to treated and untreated males. We predicted that aromatase inhibition would decrease courtship and competitive behaviors, but the swamp population would be less affected given generational exposure to hypoxia. Specifically, we compared competition and courtship behavior of males treated with a short-term exposure to an aromatase inhibitor with control fish from the two populations. We found that both courtship and competitive behaviors were affected by the interaction between treatment and population. River fish performed fewer courtship and competitive behaviors under the aromatase inhibition treatment while the behavior of swamp males was unaffected. Additionally, we found that females from the swamp population preferred males from the aromatase inhibition treatment and river females preferred control males. While we found behavioral effects of short-term aromatase inhibition, we did not find any effects on male nuptial coloration. Overall, these results indicate that the effects of short-term aromatase inhibition on behavior could depend on local adaptation in response to hypoxia.

Methods
Male-Male Competition Experiment

To compare differences in aggression between males treated with an aromatase inhibitor and control males and the two populations, we measured aggression during male-male competition trials (n = 19 trials; 9 trials river population, 10 trials swamp population). Specifically, we measured aggression of males when competing with an individual from the same population but exposed to the opposite treatment (e.g., aromatase males vs control male). From the 30-minute male-male competition trial videos, we quantified the number of lateral displays (male is perpendicular to the other fish and spreads all fins), charges (rapid movement towards the opposing male) as well as the time active (i.e., moving in the water column vs maintaining a stationary position) of each male in the trial.

Male Courtship & Female Preference Experiment

After the male competition trials, we compared female preference and male courtship behaviors between males in the control and aromatase inhibited group and the two populations. Specifically, after the male-competition trials were completed, we conducted a 50% water change in both male tanks (renewing the fadrozole concentration in the aromatase inhibited group). After the water change was completed, a third tank containing a single female was placed between the two tanks containing the males, with black plastic visually isolating all tanks (Supplementary Figure 1). We placed a piece of PVC pipe as a shelter in the center tank for the female to hide in. The fish were allowed to acclimate to their tanks overnight (~24 hours). After acclimation, the fish were filmed for 30 min to record all movement and interactions (n = 20/trials; 10/population). Trials where females did not leave the hiding spot were excluded in the analysis of male courtship and female preference (n =14 trials included). In each trial, we recorded the amount of time a female spent within 7 cm (~1.5 body lengths) of each male. For males, we quantified the number of and quivers (full body shaking) as well as time spent active. Lateral displays were also recorded during male courtship trials as these displays may occur during aggressive encounters with other males and during courtship of females (Alphen et al., 2004; Baerends and Roon, 1950; Gray et al., 2012). However, because males could see the other male during trials, we cannot distinguish between lateral displays directed at the other male and the focal female. All videos were analyzed using BORIS (Friard and Gamba, 2016) and the observer was blind to the population and treatment of the fish.

Aromatase Inhibition

To quantify the effect of the aromatase inhibitor on circulating hormone levels, we measured excreted estradiol and testosterone in holding water using ELISA assays according to manufacturer instructions (estradiol 501890, testosterone 582701, Cayman Chemical). The correlation between the concentrations of testosterone and estradiol in plasma and the concentration excreted by the gills has been established in male P. multicolor (Friesen et al., 2012a; Friesen et al., 2012b) as well as other species of fish (Friesen et al., 2012a; Kidd et al., 2010). Therefore, we used a holding water method that is ideal for small fish to collect hormone samples after the completion of the last behavioral trial. We placed individual males in glass containers containing 200 mL of clean treated water for one hour. After collection, the hormones from the holding water were extracted onto Sep-Pak Plus C18 SPE cartridges using a vacuum manifold and then frozen at -20 ℃ until further processing. We extracted hormones from the cartridges using 4 mL of ethyl acetate and then dried the sample under a stream of nitrogen gas. Samples were then reconstituted in assay kit buffer (582701, and 582251, Cayman Chemical) according to manufacturer instructions. Pharmaceuticals, including estradiol, have been detected in water samples at water treatment plants in Ohio even after water treatment (Dutta et al., 2024), so we measured the average baseline level of hormones in water that did not contain fish. The baseline measurement (mean ± SE; 85.3 ± 12.7 pg/ml testosterone and 119.0 ± 17.0 pg/ml estradiol), was subtracted from the sample hormone measurements. Three males had estradiol measurements that were indistinguishable from baseline (2 control males and one aromatase-inhibited male). Pooled samples were measured at the start and end of each plate to determine inter-assay variability (testosterone 27.8% CV,  n = 3 assays; estradiol 24.9% CV; n = 2 assays) and intra- assay variability (testosterone 7.9% CV; estradiol 16.4% CV). Note that the concentrations of hormones in the pooled sample were low which likely contributed to the moderately high CV, but samples were distributed equally across plates to minimize the effects of intra- and inter-assay variability.

Male Color

Because color can influence female preference, we also compared male color between the control and aromatase inhibition treatments by taking spectrophotometric measurements. Following euthanasia in eugenol, we took spectrophotometric measurements using a JAZ spectrophotometer (Jazz, Ocean Optics, Dunedin, FL, USA) connected to a bifurcated fiber optic cable (QR400-7-UV-VIS, Ocean Optics) following Gray et al. (2011). We measured spectral characteristics of color at three positions along the abdomen (Supplementary Figure 2). These three measurements were then averaged.

We applied a LOESS smoothing function to the spectrophotometric data to reduce noise (span = 0.2). To address negative values, we added the absolute value of the most negative number to the entire curve. We used Pavo (Maia et al., 2019) to calculate the λmax (hue or wavelength of maximum reflectance; H1 in Pavo, but restricted between 400 nm to 700 nm) and Rmax (brightness or maximum reflectance; B3 in Pavo but restricted between 400 nm to 700 nm) for each male.

Statistics

All analyses were conducted in R version 4.4.1 (R Core Team, 2024). For linear mixed models (LMM) and generalized linear mixed models (GLMM) we used the lme4 package (Bates et al., 2015; R Core Team, 2024). Results were considered significant at α < 0.05. Effect sizes were calculated as partial eta squared (η2) using the effectsize package (Ben-Shacar et al., 2020), except for GLMM’s where we estimated the overall variance explained by the model (conditional R2) and variance explained by the fixed effects (marginal R2) using the MuMIn package (Bartoń, 2023). All behavior models included the aromatase inhibition treatment (control or fadrozole), population (swamp or river), the interaction between aromatase inhibition treatment and population.

We measured the effects of fadrozole on the testosterone excretion rate, estradiol excretion rate, and the ratio of testosterone: estradiol between males in the control and aromatase inhibition treatment using LM’s with aromatase inhibition treatment and populations as fixed factors and standard length as a covariate. The testosterone excretion rate, estradiol excretion rate, and the ratio of testosterone to estradiol were log10 transformed to meet model fit assumptions.

To test whether the number of competitive behaviors (# of lateral displays + # of charges) and activity (time spent moving through the water column) were affected by the aromatase inhibition treatment or population during male competition trials, all models included treatment, population, and the interaction between treatment and population as fixed factors. Because males from the swamp population were on average slightly larger than males from the river population, standard length was included as a covariate in all behavior models (including courtship models outlined below). For all male competition models, trial was included as a random effect to account for lack of independence between male behaviors within a trial (i.e., the two males interacted with each other). A Poisson model with a log link was used to model the # of competitive behaviors. Activity was also modeled using an LMM.

We also tested whether the number of male courtship behaviors (# of lateral displays + # of quivers) and activity (time spent moving through the water column) were affected by the aromatase inhibition treatment or population during mate choice trials. A Poisson model with a log link was used to model the # of courtship behaviors. Activity was modeled using an LMM. Trial was included as a random effect to account for lack of independence of male behaviors within a trial.

We quantified female preference during the mate choice trails as the relative amount of time a female spent with the control male (time spent with control male – time spent with aromatase inhibited males). We excluded trials where the females did not leave the center of the tank and one influential point (Cooke’s distance > 1), for a total of 13 trials included in the analysis. Population, relative time spent in courtship behaviors (time control male spent in courtship behaviors – time aromatase inhibited male spent in courtship behaviors, the relative ventral λ max (λ max control male – λ max aromatase inhibited male) and relative ventral Rmax (Rmax control male – Rmax aromatase inhibited male), were also included in the female preference model.

We tested for differences in spectral characteristics of male color (λ max and R max) due to aromatase inhibition treatment or population on the ventral regions of the fish where males are typically brighter. Aromatase inhibition treatment, population, standard length, and the interaction between aromatase inhibition treatment and population were included as fixed factors in all models. For males, the ventral λ max was square root transformed.