Differential expression of steroid-related genes across electrosensory brain regions in two sexually dimorphic species of electric knifefish

Abstract

The production of communication signals can be modulated by hormones acting on the brain regions that regulate these signals. However, less is known about how signal perception is regulated by hormones. The electrocommunication signals of weakly electric fishes are sexually dimorphic, sensitive to hormones, and vary across species. The neural circuits that regulate the production and perception of these signals are also well-characterized, and electric fishes are thus an excellent model to examine the neuroendocrine regulation of sensorimotor mechanisms of communication. We investigated (1) whether steroid-related genes are expressed in sensory brain regions that process communication signals; and (2) whether this expression differs across sexes and species that have different patterns of sexual dimorphism in their signals. Apteronotus leptorhynchus and Apteronotus albifrons produce continuous electric organ discharges (EODs) that are used for communication. Two brain regions, the electrosensory lateral line lobe (ELL) and the dorsal torus semicircularis (TSd), process inputs from electroreceptors to allow fish to detect and discriminate electrocommunication signals. We used qPCR to quantify the expression of genes for two androgen receptors (ar1, ar2), two estrogen receptors (esr1, esr2b), and aromatase (cyp19a1b). Four out of five steroid-related genes were expressed in both sensory brain regions, and their expression often varied between sexes and species. These results suggest that expression of steroid-related genes in the brain may differentially influence how EOD signals are encoded across species and sexes, and that gonadal steroids may coordinately regulate central circuits that control both the production and perception of EODs.

1. Introduction

Gonadal steroid hormones often modulate animal communication by acting on peripheral and central motor structures that control signal production. Elevated testosterone (T) levels are often associated with increased motivation to produce aggressive or mating calls (Wetzel and Kelley 1983, Marler and Ryan 1996, Hunt et al. 1997, Solís and Penna 1997, Burmeister and Wilczynski 2001, Tokarz et al. 2002, Pasch et al. 2011, Alward et al. 2018), and T can act directly on the skeletal muscles involved in signaling (Bleisch et al. 1984, Sassoon and Kelley 1986, Brantley et al. 1993, Girgenrath and Marsh 2003, Schlinger et al. 2013). Estradiol (E2) can also influence signal production after the enzyme aromatase metabolizes T to E2 (Gurney and Konishi 1980, Schlinger and Arnold 1991, 1992, Wade and Arnold 2004, Alward et al. 2016). Steroidal regulation of signal production is enabled by the expression of androgen receptors, estrogen receptors, and steroid metabolizing enzymes, like aromatase, in the brain areas and peripheral structures that control signal production (Fischer and Kelley 1991, Wu et al. 2001, Forlano et al. 2006, Forlano et al. 2010, Fuxjager et al. 2013, Fuxjager et al. 2015, Mangiamele et al. 2016, Quispe et al. 2016, Smith et al. 2021, Zheng et al. 2021). For example, estrogen receptor (esr) expression in HVC varies across songbird species, and aromatase levels increase when male zebra finches sing (Remage-Healey et al. 2009, Frankl-Vilches and Gahr 2018). Although abundant evidence shows that androgens and estrogens modulate the production of signals via local expression of steroid receptors and aromatase in motor pathways, steroidal modulation of sensory circuits in receivers should also be critical for coordinating species and sex differences in communication.

While steroidal modulation of signal perception has been relatively less explored compared to hormonal modulation of motor circuits, hormones can also act on the sensory structures that regulate signal processing. For example, female plainfin midshipman fish have lower inner ear auditory thresholds that are tuned to male courtship hums during the breeding season (Sisneros and Bass 2003). Higher summer auditory sensitivity is mediated by changes in circulating steroid hormones, including E2 (Sisneros et al. 2004, Forlano et al. 2015). E2 is a common modulator of vertebrate auditory processing (Maney et al. 2008, Remage-Healey 2012, Caras 2013). Auditory responsiveness in frogs, measured as phonotaxis or immediate early gene expression, can also be associated with higher levels of T and E2 (Arch and Narins 2009, Chakraborty and Burmeister 2015, Wilczynski and Burmeister 2016). Whether the expression of steroid-related genes in sensory brain regions mediates seasonal changes in auditory sensitivity is less explored. In the auditory midbrain of túngara frogs, however, females have more esr mRNA, while males have more androgen receptor (ar) mRNA (Chakraborty and Burmeister 2010). Thus, expression of steroid receptors in sensory brain areas can vary by sex, which suggests steroid-related genes have the potential to influence sex-specific encoding of sensory information.

Hormonal regulation of communication has been particularly well-studied in South American weakly electric knifefishes, a speciose group with extensive variation in sexual dimorphism and signal structure. Knifefish have specialized organs in their tail that produce continuous electric organ discharges (EODs) that function in electrolocation and communication. EOD waveform and frequency (EODf) are individually consistent and often vary with species and sex (Smith 2013). In Apteronotus leptorhynchus, males have a higher EODf than females (Hagedorn and Heiligenberg 1985, Meyer et al. 1987). In contrast, Apteronotus albifrons shows the opposite pattern of sexual dimorphism, with females having a higher EODf than males (Dunlap et al. 1998, Kolodziejski et al. 2005). EODs are sensitive to gonadal steroid hormones in sexually dimorphic species. For example, androgen treatment raises EODf in A. leptorhynchus but lowers EODf in A. albifrons (Schaefer and Zakon 1996, Dunlap et al. 1998, Ho et al. 2013). Steroids and their receptors influence EODf by acting on the brain region that controls the firing rate of the electric organ, the pacemaker nucleus (Pn), and on spinal electromotor neurons, whose axons form the electric organ in apteronotid electric fishes (Meyer 1983, Schaefer and Zakon 1996). Indeed, steroid-related genes, including those encoding androgen receptors, estrogen receptors, and aromatase are expressed in the Pn; and their abundance varies across species that vary in sexual dimorphism of EODf (Smith et al. 2018, Proffitt 2022). For example, ar mRNA is more abundant in the Pn of species with sexually dimorphic EODf than in ‘Apteronotus’ bonapartii, a species that lacks a sex difference in EODf (Proffitt 2022). Whether and how hormones modulate the perception of EODs, however, is less well-understood.

The neural circuits that control both the production and perception of electrocommunication signals are well-characterized, making it easy to identify where steroid hormones may act to influence signals in electric knifefishes. Tuberous electroreceptors distributed across the fish’s skin detect changes in the electric field created by the EOD (Bennett 1971). Electroreceptors are tuned to the match the sender’s EODf (Meyer et al. 1987). Electroreceptor best frequency (BF) is also modulated by gonadal steroids (Bass and Hopkins 1984, Meyer et al. 1987, Ferrari and Zakon 1989). For example, treatment with a nonaromatizable androgen, 5α-dihydrotestosterone, lowers the BF of electroreceptors in Sternopygus macrurus, a species in which EODf is lower in males than in females (Meyer and Zakon 1982, Ferrari and Zakon 1989). Similarly, in A. leptorhynchus, a species in which EODf is lower in females than in males, E2 both lowers EODf and reduces the BF of electroreceptors (Meyer et al. 1987). Electroreceptors project to the electrosensory lateral line lobe (ELL) in the hindbrain, whose neurons encode changes in the amplitude and timing of the EOD, including beats created by the interference of conspecific EODs with the fish’s own (Berman and Maler 1999, Vonderschen and Chacron 2011). The ELL projects to the dorsal torus semicircularis (TSd), a region in the midbrain whose neurons respond to higher-order features of electrocommunication signals (Carr et al. 1981, Vonderschen and Chacron 2011). Whether steroid hormones act centrally within electrosensory circuits is unknown. Like the Pn, however, it is possible that the expression of steroid receptors and metabolizing enzymes in the TSd and ELL modulate how electrocommunication signals are perceived.

Here, we examined the expression of five steroid-related genes in the TSd and ELL in both male and female A. albifrons and A. leptorhynchus: androgen receptor α (ar1), androgen receptor β, (ar2), estrogen receptor α (esr1), estrogen receptor β2 (esr2b), and aromatase b (cyp19a1b) (Munley et al. 2023). This study had two primary goals: (1) to determine whether steroid-related genes are expressed in electrosensory brain regions and (2) to identify whether expression of genes for steroid receptors or steroidogenic enzymes in sensory brain regions is species- and/or sex-specific. Because these genes are found in the Pn and gonadal steroids play a role in sensory processing in other systems, we hypothesized that steroid-related genes would also be expressed in the ELL and TSd of knifefishes. In addition, we hypothesized that patterns of species- and sex-related variation in the expression of these genes in the ELL and TSd would mirror patterns in the Pn. For example, aromatase mRNA was more highly expressed in females than in males in both A. leptorhynchus and A. albifrons in the Pn. We therefore might expect females to express higher levels of cyp19a1b than males in electrosensory brain regions.

2. Materials and methods

2.1. Animals

Fish included in this study were primarily used as part of two other projects (Proffitt 2022, Freiler 2023). Proffitt (2022) explored variation in steroid-related genes in the Pn; and Freiler (2023) examined variation in neuromodulator receptor gene expression in the ELL and TSd. We took advantage of unused brain tissue and cDNA from these projects to ask whether steroid-related genes are expressed in electrosensory brain regions. A subset of fish from these other projects was used in this study. From Proffitt (2022), A. leptorhynchus (N=5 M, 5 F) and A. albifrons (N=5 M, 5 F, Emark Tropical Imports, Brooklyn, NY) were collected in Colombia and transported to Indiana University in 2019 and 2020. From Freiler (2023), Colombian A. leptorhynchus (N=6 M, 6 F, Emark or AliKahn Tropical Fish, Richmond Hill, NY) and A. albifrons (N=6 M, 6 F, Emark or East Coast Transship, Clinton, MD, USA) were transported to the lab between 2018–2022. Fish from both replicates were housed individually on a 12h:12h light:dark cycle in 29-L tanks in water maintained at 25–28°C with conductivity of 200–600 μS cm⁻¹ and pH 5.6–6.6. Fish were all fed live blackworms, live earthworms, and/or frozen bloodworms three times a week. As outlined below, the behavioral and tissue collection procedures differed slightly between the two replicate years. Due to slight differences in experimental procedures and the use of different qPCR calibrators to normalize gene expression data across plates, we analyzed data from fish in the two replicates independently. We will refer to the data taken from fish in Proffitt (2022) as Replicate 1 and to the data taken from fish in Freiler (2023) as Replicate 2. All procedures were conducted in accordance with protocols approved by the Bloomington Institutional Animal Care and Use Committee (BIACUC).

2.2. EODf sampling

EODfs from Replicate 1 were recorded immediately before dissection by placing two wires next to the fish, amplifying the signal (Grass-Telefactor, West Warwick RI, USA, gain 100X), and measuring frequency with a multimeter (Fluke 187 True RMS multimeter, Fluke, Everett, WA). For Replicate 2, EODfs were measured from a 1-hr recording that started around 9:45 pm the evening before fish were dissected. A pair of carbon electrodes were placed opposite each other on either side of the fish’s tank and were connected to a Zoom H6 Handy Recorder (Zoom, Hauppauge, NY, USA) with an amplification setting between 5 and 7 on the recorder. Signals were sampled at 16 bits and 48 kHz, and EODf was determined using the frequency analysis feature in Adobe Audition (Adobe Systems, San Jose, CA, USA). Since EODf varies with temperature, EODf measurements were corrected to that expected at 26°C by using a Q_10°C value of 1.63 (Dunlap et al. 2000).

2.3. Blood sampling

In Replicate 1, blood samples were collected 1–2 weeks before dissection and tissue collection. In Replicate 2, blood was taken the morning of dissection. After a fish was removed from its home tank, it was anesthetized lightly in 0.075% 2-phenoxyethanol dissolved in tank water for 2–4 minutes. Blood (~25–50 μL) was collected from the caudal vein with a 25G × 5/8” heparinized needle attached to a 1 mL syringe and was transferred to heparinized microhematocrit tubes. Blood was centrifuged at 4400 g for 7–8 mins to separate plasma, which was removed from the hematocrit tubes with a Hamilton syringe and was stored at −80°C. In Replicate 1, fish were allowed to recover from anesthesia for ~5–10 min in a bucket with tank water before they were returned to their home tanks. These fish were anesthetized for tissue collection 1–2 weeks later as described below. In Replicate 2, fish were placed back in anesthetic solution and were prepped for dissection and tissue collection immediately after blood sampling.

2.4. Hormone assays

Plasma 11-ketotestosterone (11-KT), T, and E2 concentrations were quantified using ELISA kits (Cayman Chemical, Ann Arbor, MI, USA). These kits have been validated and used previously to measure hormone concentrations in apteronotid knifefishes (Cox-Fernandes et al. 2010, Ho et al. 2013, Petzold and Smith 2016, Freiler et al. 2022). 11-KT levels were only measured in fish with testes and E2 levels were only measured in fish with ovaries because 11-KT and E2 concentrations are often below the assay detection limits in individuals with a female or a male gonadal sex, respectively. Plasma samples were diluted in assay buffer: 11-KT 1:30–1:55; T 1:17–1:50; and E2 1:3–1:9. The specific dilutions varied slightly between samples and were determined by expected hormone concentrations in these species based on previous studies (Dunlap et al. 1998, Ho et al. 2013, Smith et al. 2018) and by the amount of plasma available from each fish. All samples were run in duplicate, and controls and standards were included as specified by manufacturer’s instructions. For Replicate 1, intra-assay variation for each plate was calculated using the CV between the three duplicate wells of the fourth standard (12.5 pg mL⁻¹ for 11-KT, 62.5 pg mL⁻¹ for T, or 156.3 pg mL⁻¹ for E2) that was distributed across the plate. Inter-assay variation for T was calculated by using the CV between the two sets of three duplicate wells of the fourth standard across the two plates. For Replicate 2, an extra standard curve was made for each hormone when the first assay was run. The fourth standard in the dilution series was aliquoted and stored at −80°C for use on every plate. This standard was plated in duplicate three times across each plate. The CV among these wells within and between plates was used to calculate intra- and inter-assay variation, respectively. Because fish in both Replicates 1 and 2 were part of other, larger studies, the plasma samples for this study were distributed across several ELISA plates. For Replicate 1, intra-assay variations were 14.5% for 11-KT, 12.9 and 25.0% for T, and 26.3% for E2. Inter-assay variation for T in Replicate 1 was 18.8%. Samples were run on a single 11-KT and E2 assay for Replicate 1. In Replicate 2, intra-assay variations were between 8.8–16.7% for 11-KT, 8.9–10.7% for T, and 10.3–10.9% for E2. Inter-assay variation in Replicate 2 was 19.2% for 11-KT, 12.1% for T, and 20.8% for E2. We note that the intra-assay and inter-assay variation values are somewhat higher than ideal. Consequently, quantitative analysis of the hormone levels of the fish in this study may have elevated measurement error, and these data are primarily used to qualitatively assess the reproductive condition of the fish. The detection ranges (lowest-highest concentrations on the standard curve) for the assay kits were 0.78–100 pg mL⁻¹ for 11-KT, 3.9–500 pg mL⁻¹ for T, and 0.61–10,000 pg mL⁻¹ for E2. Sometimes the lowest concentration standard did not fit on the standard curve. In these cases, the next lowest concentration standard was used as the lower detection limit instead (1.56 pg mL⁻¹ for 11-KT, 7.8 pg mL⁻¹ for T, and 2.4 pg mL⁻¹ for E2). After correcting for sample dilution, the minimum plasma detection limits in the samples were thus 23.4–85.8 pg mL⁻¹ for 11-KT, 66.3–390 pg mL⁻¹ for T, and 1.83–21.6 pg mL⁻¹ for E2. Seven samples run on T assays and one sample on an E2 assay fell below the minimum detectable limit. For samples that fell off the curve, if the sample was within 5% of the percent max bound, the concentration was estimated at half the detection limit. If the sample had a percent bound/max bound (%B/B0) greater than 5% of the %B/B0 of the least concentrated sample included on the standard curve, then the concentration was set to zero. One sample on an 11-KT assay fell above the maximum detection limit on the standard curve, and its 11-KT concentration was estimated at the maximum detection limit.

2.5. Brain tissue collection

Each fish was weighed either immediately before dissection (Replicate 1) or three days before dissection (Replicate 2). Individuals were deeply anesthetized in either 0.1% (Replicate 1) or 0.15% (Replicate 2) 2-phenoxyethanol dissolved in tank water for ~4–8 mins. Fish were dissected on ice. The skin was removed from the top of the skull using a scalpel and forceps, and surgical scissors were used to remove the top of the skull and sever the spinal cord. Brains were removed within 3 min of surgery start and were frozen in OCT compound (Sakura, Tokyo, Japan) on dry ice. For Replicate 1, the pacemaker nucleus was dissected from the hindbrain in ice-cold molecular grade ultrapure water (Invitrogen, Waltham, MA, USA) before the rest of the brain was embedded. All embedded brains were kept at −80°C until further processing.

2.6. Reproductive condition

Immediately after brains were removed, gonads and reproductive ducts were examined, dissected, and removed. A mid-ventral incision was made through the body wall from the vent to the caudal margin of the peritoneal cavity. Gonads were visually inspected to confirm the sex of the fish. The sperm duct or oviduct was severed where it joins the intestine near the vent, and the entire membrane-encapsulated testes or ovaries along with the ducts were carefully removed. Gonads were weighed immediately after they were removed to calculate gonadosomatic index (GSI) (gonad mass/body mass × 100) as a measure of reproductive condition. Ovarian maturity was also assessed qualitatively following dissection. Nearly all ovaries examined had yolked follicles, which were characterized by a yellow color and large size (~ 1– 2 mm), indicating that the fish were in near-spawning condition. In contrast, immature ovaries are typically smaller with white follicles <0.5 mm in diameter.

2.7. Brain micropunching and histology

A stainless-steel tissue matrix (Ted Pella Inc., Redding, CA, USA) was used to prepare 1 mm slices of whole brains. Slices were then placed on a microscope slide in a cryostat at −20°C. The A. leptorhynchus brain atlas (Maler et al. 1991) was used as a reference for locating and collecting tissue punches from the ELL and TSd. Brain slices were visualized under a 1.5x LED magnifying visor, and 8–10 tissue punches (4–5 ELL, 4–5 TSd) were collected with a 0.5 mm or 1.0 mm biopsy punch (World Precision Instruments, Sarasota, FL, USA) for the ELL or TSd, respectively. Punches were placed in RNAlater solution in 0.65 mL microcentrifuge tubes, stored at 4^oC for at least one night, and then stored at −20°C.

Slices from which punches were taken were fixed in 10% formalin and stored at 4°C in a small petri dish for 1–4 days. Brain slices were then embedded in 15–20% gelatin and were stored in 10% formalin at 4°C for another 1–4 days before 25–30% sucrose was added for cryoprotection. After 1–4 days, embedded and cryoprotected brain slices were frozen with pulverized dry ice and were sectioned at 50–75 μm on a sliding microtome (Model 860, AO Optical, Buffalo, NY). The thin sections were transferred to a well plate containing PBS, were mounted on gelatin coated slides using a solution of water and Triton X-100 surfactant, and were left to dry. Slides were stained the next day to verify punch locations. Slides were rehydrated in distilled water, were stained in thionin (0.1% in acetic acid/acetate buffer) and were differentiated in 0.1% glacial acetic acid in 70% ethanol followed by 70% ethanol. The slides were allowed to dry before coverslipping with DPX mountant (Electron Microscopy Sciences, Hatfield, PA).

2.8. Relative gene expression

qPCR primer-probe assays were designed using the PrimerQuest Tool from Integrated DNA Technologies (Integrated DNA Technologies, Coralville, IA, USA). Primers for the five steroid-related genes of interest (ar1, ar2, esr1, esr2b, cyp19a1b) were validated previously by MRP for another experiment (Table S1; Proffitt, 2022). slc25a5, a mitochondrial carrier protein, was used as the internal reference gene. slc25a5 has also been validated previously, and slc25a5 mRNA abundance does not vary by species or sex (Proffitt 2022, Freiler 2023). For all primer sets, only a single amplicon was present following PCR. The validation protocols for qPCR primer-probe assays are outlined in more detail in Proffitt (2022). qPCR primer-probe assay efficiencies for each gene were calculated separately in each species by using the average of the individual reaction efficiencies measured from the exponential part of the amplification curve across samples and brain region with the LinRegPCR program (Ruijter et al. 2013, Untergasser et al. 2021). Efficiencies were consistent across species and replicates and ranged from 88–93% (Table S1).

RNA was extracted with Maxwell RSC SimplyRNA Tissue Kits (Promega, Madison, WI, USA) following the manufacturer’s protocols, which included a DNAse I treatment. ELL samples were eluted in 30 μL, instead of the traditional 50 μL, nuclease free water. RNA quality and concentration was tested with High Sensitivity RNA Screen Tapes on an Agilent 2200 (Replicate 1) or Agilent 4150 (Replicate 2) TapeStation System (Agilent, Santa Clara, CA, USA). RNA was DNAse-treated again and converted into cDNA with 10 ng of mRNA input using an iScript gDNA Clear cDNA Synthesis Kit (Bio-Rad Laboratories, Hercules, CA, USA). The amount of mRNA input was previously established based on the kit’s dynamic range and relative gene abundance (Proffitt 2022). All included samples had a RINe score of 7.4 or higher. The mean ± SEM for RINe scores in Replicate 1 were 8.39±0.10 (N=18) for ELL and 8.50±0.06 (N=20) for TSd. One ELL sample (A. albifrons F) was not made into cDNA because the RINe score (RINe=2.5) was too low, and one ELL sample (A. leptorhynchus F) was not used because the RNA concentration not high enough to include the target 10 ng of mRNA input into the cDNA reaction. The mean ± SEM for RINe scores in Replicate 2 were 8.80±0.06 for ELL (N=13) and 8.81±0.07 (N=23) for TSd. One TSd sample (A. albifrons M) was not used because the RINe score (RINe=6.3) was low. Three ELL samples (all A. albifrons F) were not used because the RNA concertation was too low, and eight ELL samples (3 A. albifrons M, 2 A. albifrons F, 1 A. leptorhynchus M, 2 A. leptorhynchus F) were not used because there was not enough cDNA leftover from the primary project (Freiler 2023). Calibrators were generated for both Replicates 1 and 2 from a homogenous mixture of A. albifrons and A. leptorhynchus whole brain tissue. 500 ng of mRNA input was used in the whole brain cDNA reactions. Even though all molecular protocols, stock reagents, and procedures were identical between replicates, the calibrator created for each replicate used different brains. Because different calibrators were used to normalize gene expression in each replicate, we could not quantitatively combine the 2^−ΔΔCt gene expression values from the two replicates. While −ΔCt values were similar across replicates (Table S3), the −ΔCt calculation does not include an interplate correction or normalization to the reference sample, the whole brain calibrator. We, therefore, analyzed each replicate separately with 2^−ΔΔCt gene expression values. All cDNA was stored at −20°C until qPCR. esr2b did not amplify from either ELL or TSd cDNA with PCR and was therefore not analyzed further with qPCR.

cDNA samples were plated in triplicate using 10 μL reactions with PrimeTime Gene Expression Master Mix and custom PrimeTime qPCR Probe Assays with a 5’ FAM reporter dye and a ZEN/Iowa Black FQ quencher (Integrated DNA Technologies). An ASSIST PLUS pipetting robot (INTEGRA Biosciences Corp., Hudson, NH, USA) running customized programs was used to load qPCR plates. Plates were incubated overnight at 4°C before performing qPCR on a QuantStudio 6 Flex Real-Time qPCR System (Applied Biosystems, Waltham, MA, USA) using the recommended reaction conditions for PrimeTime Gene Expression Master Mix. The 384-well plates (Applied Biosystems) were heated for 3 min at 95°C and then amplified for 40 cycles at 95°C for 15 secs and at 60°C for 1 min. Controls lacking cDNA template or reverse transcriptase did not amplify on any plates. An interplate correction was implemented using the three technical replicates of whole brain calibrator amplified with slc25a5 across both plates in both experimental replicates.

2.9. Statistics

Analyses were performed in R (R Core Team 2022, v 4.2.1), and figures were plotted with the ggplot2 package (Wickham 2016). Relative gene expression was analyzed with the 2^−ΔΔCt method (Livak and Schmittgen 2001) using the web-based Relative Quantification application in the Thermo Fischer Connect Platform (Thermo Fischer Scientific, Waltham, MA, USA). One female A. albifrons TSd sample in Replicate 1 was removed prior to statistical analysis because the punch location was off target. Normality of model residuals were determined using a Shapiro-Wilk test for all tests. All models testing variation in relative gene expression were run using log-transformed 2^−ΔΔCt values. A log or log+1 (for independent variables with zeros) transformation was performed on steroid hormone levels when assumptions of normality were not met to improve residual fit to normality. Differences in reproductive condition (GSI) and steroid levels across replicates were analyzed with Welch’s t-tests. Variation in 11-KT and E2 levels across species were assessed with one-way ANOVAs. Effects of species and sex on EODf, T levels, and relative gene expression across the four genes were determined using two-way ANOVAs. Analyses of hormone levels were included primarily to confirm measures of reproductive condition provided by GSI and to explore potential differences in steroid profiles across species, sex, and replicates. Reported differences in hormone levels, however, should interpreted cautiously given the higher assay CVs, small sample sizes, and resulting low statistical power. Two-way ANOVAs were first run with Type III ANOVAs. If the interaction term was nonsignificant, it was removed, and the test was run as a Type II ANOVA instead (Smith and Cribbie 2014). The F-value and P-value for the overall ANOVA model were extracted from the lm function. Post hoc pairwise comparisons for significant interactions were assessed with the emmeans package (Lenth 2023). In the models testing variation in ar1 expression in the TSd in Replicate 1 and T levels in Replicate 1, assumptions of normality were still violated after log transformation, so variation was analyzed with nonparametric Mann-Whitney U tests for each species and sex instead. Due to the limited sample size, statistical analyses were not run on ELL samples for Replicate 2, but we have still plotted the data for visualization and qualitative comparison. The difference in gene expression between the two androgen receptor gene paralogs (ar1 and ar2) in each species was analyzed using −ΔCt values (Table S3), which are appropriate for comparing mRNA abundance across genes within the same tissue type. ar1, ar2, and slc25a5 qPCR reactions were also run on the same plate in both replicates to minimize interplate effects in this calculation. Welch’s t-tests, or nonparametric Mann-Whitney U tests when assumptions of normality were not met following a log transformation, were used to test differences in expression between the two androgen receptor gene paralogs.

3. Results

3.1. Reproductive condition, EODf, and plasma steroid levels

Reproductive condition (measured with GSI) did not vary between replicates in female A. leptorhynchus (t-test: t(9)=0.19, P=0.86), male A. leptorhynchus (t-test: t(7.9)=0.14, P=0.89), or male A. albifrons (t-test: t(5.6)=0.50, P=0.64; Table 1). Female A. albifrons, however, were in more advanced reproductive condition in Replicate 1 than in Replicate 2 (t-test: t(7.9)=3.26, P=0.01). In both replicates, ovarian follicles were yolked, indicating that the fish were in reproductive condition. EODf and hormone levels varied across species and sexes (Table 1; Table S2). Across both replicates, A. albifrons had higher EODfs than A. leptorhynchus (Table S2). There was also a significant species by sex interaction in EODf (Table S2). Post hoc pairwise comparisons revealed that A. albifrons females had higher EODfs than males in both replicates (Replicate 1: t-ratio(16)=5.88, P<0.01; Replicate 2: t-ratio(20)=3.03, P<0.01), whereas A. leptorhynchus males had higher EODfs than females in both replicates (Replicate 1: t-ratio(16)=−4.93, P<0.01, Replicate 2: t-ratio(20)= −3.26, P<0.01).

Table 1.

Summary data for EODf, body mass, reproductive condition, and steroid levels.

	EODf (Hz)	Body mass (g)	GSI (%)	11-KT (ng/mL)	T (ng/mL)	E2 (ng/mL)
Replicate 1 (N=5 fish/group)
A. leptorhynchus males	883.9±20.3	35.2±1.6	0.37±0.01	1.75±0.31	0.76±0.13	--
A. leptorhynchus females	729.8±31.4	32.6±2.5	5.97±1.38	--	1.71±0.53*	0.08±0.07*
A. albifrons males	924.3±6.6	47.5±4.4	0.65±0.04	1.83±0.28*	0.82±0.35*	--
A. albifrons females	1108.2±22.6	38.2±2.1	11.6±1.04	--	2.27±0.29*	0.10±0.02*
Replicate 2 (N=6 fish/group)
A. leptorhynchus males	845.9±52.6	30.4±5.7	0.36±0.02	2.71±0.38	0.15±0.03	--
A. leptorhynchus females	665.0±43.5	26.2±3.5	5.58±1.55	--	0.02±0.01	0.13±0.03
A. albifrons males	944.8±19.4	39.0±1.9	0.56±0.17	1.34±0.48	0.33±0.12	--
A. albifrons females	1112.7±33.6	36.6±2.4	4.94±1.77		0.10±0.04	0.08±0.03

All values reported are mean ± SEM.

^*We did not measure hormone levels from two A. leptorhynchus females and one male and one female A. albifrons in Replicate 1. Thus, N=3 for A. leptorhynchus females and N=4 for A. albifrons females and males for the hormone measures. See Table S2 for statistical results.

11-KT levels in males did not vary by species in Replicate 1 but were higher in A. leptorhynchus males than in A. albifrons males in Replicate 2 (Table S2). T levels were higher in females than in males in both species in Replicate 1 but were higher in males than in females in both species in Replicate 2 (Table S2). E2 levels in females did not vary by species in either replicate (Table S2). 11-KT levels did not differ between replicates in A. albifrons (t-test: t(7.5)=0.87, P=0.41) but were marginally, nonsignificantly higher in Replicate 2 than in Replicate 1 in A. leptorhynchus (t-test: t(8.9)=−1.96, P=0.08). T levels were significantly higher in Replicate 1 compared to Replicate 2 in female A. albifrons (t-test: t(3.1)=7.34, P<0.01), female A. leptorhynchus (t-test: t(6.1)=6.16, P<0.01), and male A. leptorhynchus (t-test: t(4.4)=4.65, P<0.01). T levels did not differ by replicate in male A. albifrons (t-test: t(3.7)=1.31, P=0.27). E2 levels did not differ between replicates in either A. albifrons (t-test: t(7.9)=0.79, P=0.45) or A. leptorhynchus (t-test: t(3.0)=−0.73, P=0.52).

3.2. Relative gene expression

Four out of five genes (ar1, ar2, esr1, and cyp19a1b) of interest were expressed in both the TSd and ELL of males and females of both species. esr2b was not detected in the ELL or the TSd of males or females in either species.

3.2.1. ar1

ar1 expression did not vary by species or sex in Replicate 1 in either the TSd (Mann-Whitney U: [Species: W=32, P=0.32; Sex: W=50, P=0.72]) or the ELL (ANOVA: F(2,14)=0.62, P=0.55, [Species: F(1,14)=0.04, P=0.84; Sex: F(1,14)=1.16, P=0.30]) (Fig. 1 a–b). In Replicate 2, a significant sex-by-species interaction (ANOVA: F(3,19)=2.01, P=0.15, [Species: F(1,19)=3.97, P=0.06; Sex: F(1,19)=3.60, P=0.07; Species*Sex: F(1,19)=5.76, P=0.03]) was detected in the TSd. Given that the overall model was nonsignificant, interpreting this result is difficult, so no post hoc tests were run (Fig. 1c). Variation in ar1 expression in the ELL in Replicate 2 could not be assessed due to the small sample size, but qualitatively mirrored patterns from Replicate 1 (Fig. 1d).

Figure 1. ar1 mRNA abundance varied little across species and sex.

Boxplots show variation in ar1 gene expression across both species and sex in both replicates in the TSd (a,c) and the ELL (b,d). Relative expression = 2^−ΔΔCt. Gene expression was normalized with different calibrators in the two replicates, so the absolute values on the y-axes are not directly comparable across replicates. Black lines inside boxplots represent the median and whiskers extend to 1.5 x the interquartile range. Outliers are represented as single points. Aa = A. albifrons. Al = A. leptorhynchus. Females are shown in yellow, and males are shown in blue. Sample sizes in order (Aa F, Aa M, Al F, Al M): (a) N = 4,5,5,5 (b) N = 3,5,4,5 (c) N = 6,5,6,6 (d) N = 1,3,4,3.

3.2.2. ar2

ar2 gene expression in the TSd did not vary by sex but was higher in A. leptorhynchus than in A. albifrons in both Replicate 1 (ANOVA: F(2,16)=4.48, P=0.03, [Species: F(1,16)=8.88, P<0.01; Sex: F(1,16)=0.01, P=0.90]) (Fig. 2a) and Replicate 2 (ANOVA: F(2,20)=5.84, P=0.01, [Species: F(1,20)=11.68, P<0.01; Sex: F(1,20)=0.01, P=0.93]) (Fig. 2c). ar2 gene expression in the ELL did not vary by species or sex in Replicate 1 (ANOVA: F(2,15)=0.13, P=0.88, [Species: F(1,15)=0.25, P=0.63; Sex: F(1,15)=0.01, P=0.92]) (Fig. 2b). Variation in ar2 in the ELL in Replicate 2 could not be assessed statistically due to the small sample size, but qualitatively mirrored patterns from Replicate 1 (Fig. 2d).

Figure 2. ar2 mRNA abundance was higher in the TSd in A. leptorhynchus.

Boxplots show variation in ar2 gene expression across both species and sex in both replicates in the TSd (a,c) and the ELL (b,d). Asterisks represent significant species differences. Relative expression = 2^−ΔΔCt. Gene expression was normalized with different calibrators in the two replicates, so the absolute values on the y-axes are not directly comparable across replicates. Black lines inside boxplots represent the median and whiskers extend to 1.5 x the interquartile range. Outliers are represented as single points. Aa = A. albifrons. Al = A. leptorhynchus. Females are shown in yellow, and males are shown in blue. Sample sizes in order (Aa F, Aa M, Al F, Al M): (a) N = 4,5,5,5 (b) N = 4,5,4,5 (c) N = 6,5,6,6 (d) N = 1,3,4,3.

3.2.3. esr1

esr1 gene expression was higher in both male and female A. albifrons compared to both sexes in A. leptorhynchus in the TSd in both Replicate 1 (ANOVA: F(2,15)=6.25, P=0.01, [Species: F(1,15)=11.95, P<0.01; Sex: F(1,15)=0.54, P=0.47]) (Fig. 3a) and Replicate 2 (ANOVA: F(2,20)=33.00, P<0.01, [Species: F(1,20)=65.17, P<0.001; Sex: F(1,20)=0.30, P=0.59]) (Fig. 3c) and in the ELL in Replicate 1 (ANOVA: F(2,15)=13.52, P<0.01, [Species: F(1,15)=25.15, P<0.001; Sex: F(1,15)=1.88, P=0.19]) (Fig. 3b). Variation in esr1 abundance in the ELL in Replicate 2 could not be analyzed statistically, but qualitatively also supported results from Replicate 1 (Fig. 3d).

Figure 3. esr1 mRNA abundance was higher in A. albifrons.

Boxplots show variation in esr1 gene expression across both species and sex in both replicates in the TSd (a,c) and the ELL (b,d). Asterisks represent significant species differences. Relative expression = 2^−ΔΔCt. Gene expression was normalized with different calibrators in the two replicates, so the absolute values on the y-axes are not directly comparable across replicates. Black lines inside boxplots represent the median and whiskers extend to 1.5 x the interquartile range. Outliers are represented as single points. Aa = A. albifrons. Al = A. leptorhynchus. Females are shown in yellow, and males are shown in blue. Sample sizes in order (Aa F, Aa M, Al F, Al M): (a) N = 4,4,5,5 (b) N = 4,5,4,5 (c) N = 6,5,6,6 (d) N = 1,3,4,3.

3.2.4. Aromatase

In Replicate 1, there was a nonsignificant trend for greater cyp19a1b expression in the TSd of A. albifrons than in A. leptorhynchus and of females than in males (ANOVA: F(2,15)=3.14, P=0.07, [Species: F(1,15)=3.36, P=0.09, Sex: F(1,15)=2.93, P=0.11]). In Replicate 2, cyp19a1b expression in the TSd was significantly greater in A. albifrons than in A. leptorhynchus, and there was also still a nonsignificant trend for greater cyp19a1b expression in females relative to males (ANOVA: F(2,20)=16.17, P<0.01, [Species: F(1,20)=27.40, P<0.001, Sex: F(1,20)=3.92, P=0.06]) (Fig 4 a,c). In Replicate 1, cyp19a1b gene expression in the ELL was also significantly higher in A. albifrons than in A. leptorhynchus but did not vary significantly by sex (ANOVA: F(2,15)=4.52, P=0.03, [Species: F(1,15)=6.83, P=0.02; Sex: F(1,15)=2.21, P=0.16]) (Fig. 4b). Similar patterns of cyp19a1b expression in the ELL were also found in Replicate 2 (Fig. 4d). In both replicates, females of both species tended to have higher levels of cyp19a1b in the ELL as well, but these trends did not reach significance (Replicate 1) or could not be tested statistically (Replicate 2).

Figure 4. cyp19a1b mRNA abundance varied with species and sex.

Boxplots show variation in cyp19a1b gene expression across both species and sex in both replicates in the TSd (a,c) and the ELL (b,d). Asterisks represent significant species differences. There was a nonsignificant trend for higher cyp19a1b expression in females relative to males in both years and brain regions. Relative expression = 2^−ΔΔCt. Gene expression was normalized with different calibrators in the two replicates, so the absolute values on the y-axes are not directly comparable across replicates. Black lines inside boxplots represent the median and whiskers extend to 1.5 x the interquartile range. Outliers are represented as single points. Aa = A. albifrons. Al = A. leptorhynchus. Females are shown in yellow, and males are shown in blue. Sample sizes in order (Aa F, Aa M, Al F, Al M): (a) N = 4,4,5,5 (b) N = 4,5,4,5 (c) N = 6,5,6,6 (d) N = 1,3,4,3.

3.2.5. Relative expression of steroid receptor gene paralogs

Differences between ar1 and ar2 gene expression were examined with males and females combined because there were no sex differences in either ar1 or ar2 expression (Figs 1–2). In Replicate 1, ar2 expression was significantly greater than that of ar1 in the TSd of both A. albifrons (t-test: t(12.8)=15.04, P<0.01; Table S3) and A. leptorhynchus (t-test: t(7.4)=6.87, P<0.01; Table S3) and in the ELL of both A. albifrons (t-test: t(12.8)=15.04, P<0.01; Table S4) and A. leptorhynchus (Mann-Whitney U: W=81, P<0.01; Table S4). In Replicate 2, ar2 was also significantly more abundant than ar1 in the TSd of both A. albifrons (t-test: t(13.6)=29.09, P<0.01; Table S3) and A. leptorhynchus (t-test: t(21.0)=33.45, P<0.01; Table S3). The difference between ar1 and ar2 expression could not be statistically assessed in the ELL of Replicate 2 due to the limited sample size, but ar2 expression was higher than ar1 expression in every sample (Table S4). In contrast, esr1 was more abundant in both brain regions than esr2b, which was nondetectable.

4. Discussion

We found that four out of five steroid-related genes of interest (ar1, ar2, esr1, and cyp19a1b) were expressed in electrosensory brain regions of apteronotid knifefishes. In addition, we found that expression of genes for androgen and estrogen receptors and aromatase varied by species. While GSI and hormone levels varied significantly across individuals, combined, these measures indicated that the fish in this study were in relatively advanced reproductive condition. We also confirmed that EODf is sexually dimorphic in the expected direction across species. As has been reported previously (Dunlap et al. 1998), EODf was higher in females in A. albifrons, but was higher in males in A. leptorhynchus. Overall, these results highlight that in addition to gonadal steroid hormones regulating sex differences in the production of EODs, these hormones also have potential to modulate species- and sex-specific sensory processing of electrocommunication signals. Patterns of steroid-related gene expression also varied between electrosensory and electromotor circuits, which we discuss below.

4.1. Androgen receptors

Androgens commonly regulate the production, and in some cases the perception, of communication signals. For example, ar is found in the vocal-acoustic circuitry, including the torus semicircularis (TS), of midshipman fish and geckos (Tang et al. 2001, Forlano et al. 2010). Androgen receptor (AR) protein is also expressed in the TSd of both breeding and nonbreeding male Brachyhypopomus gauderio, a nonapteronotid knifefish species (Pouso et al. 2010). We found that both ar1 and ar2 mRNA was expressed in both the TSd and ELL of A. albifrons and A. leptorhynchus. ar2 mRNA was significantly more abundant than ar1 mRNA in the ELL and TSd of both species, suggesting one receptor subtype is driving most androgen-dependent responses in electrosensory pathways. ar1 abundance largely did not vary across species in either the TSd or the ELL. ar2 expression was significantly greater in A. leptorhynchus than in A. albifrons in the TSd but did not differ between species in the ELL. Sensory encoding of higher order signal features may, therefore, be more androgen-dependent in A. leptorhynchus relative to A. albifrons, which had lower levels of ar2 mRNA in the TSd.

Neither ar1 nor ar2 expression was sexually dimorphic in the TSd or the ELL. Similarly, in the Pn, gonadal steroid receptor abundance varied substantially across species but was rarely sexually dimorphic (Proffitt 2022). In contrast, males express more ar mRNA in the auditory midbrain in túngara frogs, and male singing mice have more AR immunoreactive cells in the hypothalamus and vocal-acoustic circuitry (Chakraborty and Burmeister 2010, Zheng et al. 2021). Sex differences in expression of steroid receptors in the brain can lead to sex differences in the responsiveness of behaviors to these hormones. However, sex differences in the levels of circulating gonadal steroid hormones themselves can also drive behavioral sex differences in the absence of variation in receptor abundance. The lack of sex differences in expression of androgen receptor genes suggests that the ELL and TSd may be similarly responsive to androgens in males and females. Sex differences could still exist in the local metabolism of androgens. Genes for the enzymes 11β-hydroxysteroid dehydrogenase (hsd11b), which in conjunction with 11β-hydroxylase, synthesizes 11-KT from T, and 5α-reductase, which synthesizes 5α-dihydrotestosterone from T, are expressed in the Pn and may enhance the potency of T in masculinizing EOD production (Smith et al. 2018, Proffitt 2022). It is not known whether these enzymes are expressed in electrosensory brain regions, however. hsd11b type 2 mRNA staining is very diffuse in the TS of zebrafish (Alderman and Vijayan 2012), and it is not known whether hsd11b is expressed in sensory brain regions in other teleosts. Nevertheless, sexually dimorphic concentrations of androgens (11-KT, T) may still result in sex differences in electrosensory processing of communication signals by acting on ARs in these brain areas. 11-KT concentrations are typically below the level of assay detection in female knifefish and are much greater in males (Dunlap et al. 1998). T levels were sexually dimorphic in both experimental replicates but in opposite directions. In Replicate 1, females had higher T levels than males, while males had higher T levels in Replicate 2. Because 11-KT is the primary androgen in fishes, it is not uncommon for T levels to be similar across sex or even higher in females (Hirschenhauser et al. 2004, Desjardins et al. 2006, Ho et al. 2010). Although androgen concentrations varied substantially across sex and years, ar expression in the TSd and ELL was sexually monomorphic, suggesting ar expression might be independent of peripheral androgen levels. Indeed, even when sex differences in receptor abundance are present, they can develop independently of circulating steroids (Gahr and Metzdorf 1999). These results, therefore, likely reflect standing, constitutive variation in mRNA ar abundance in electrosensory brain regions of knifefishes.

4.2. Estrogen receptors and aromatase

Estrogen receptors are often expressed in sensory structures, and estrogens can mediate seasonal and sex-specific differences in behavior and signal perception. Estrogens locally synthesized in the brain regulate nonbreeding aggression in nonapteronotid knifefishes (Zubizarreta et al. 2020, Zubizarreta et al. 2023). Estrogen receptors and aromatase are also both widely expressed across acoustic sensorimotor pathways of sonic fish (Forlano et al. 2001, Goodson and Bass 2002, Forlano et al. 2005, Fergus and Bass 2013). The TS in sonic fish, however, only expresses esr2 (Fergus and Bass 2013). esr2 is also more highly expressed in the vocal motor nucleus of females and silent, sneaker male sonic fish than in vocalizing males (Fergus and Bass 2013). In contrast, in mice, esr1 is more highly expressed than esr2 in the mammalian homolog of the TS, the inferior colliculus (Charitidi et al. 2010). While we did not detect esr2b in either apteronotid species, esr1 was expressed in both the TSd and ELL of both apteronotid species, highlighting the potential for estrogens to regulate perception of electrocommunication signals. esr1 expression was consistently higher in A. albifrons than in A. leptorhynchus across both replicates and brain regions. While esr subtypes often have tissue-specific and brain region specific expression patterns (Österlund et al. 1998, Socorro et al. 2000, Choi and Habibi 2003), the specific functions of each estrogen receptor in regulating signal production and perception are not well understood. Brain aromatase, however, has been more well-studied and often has more plastic expression that is linked to variation in sensory systems.

While expression of another aromatase gene, cyp19a1a, was not detected with in situ hybridization in the midbrain or hindbrain of nonbreeding A. leptorhynchus (Shaw and Krahe 2018), we found consistently detectable but highly variable levels of cyp19a1b mRNA expression in both A. leptorhynchus and A. albifrons. Notably, the individuals in our study were also in comparatively more advanced reproductive condition, which could also explain the contrasting findings. The females in our study had GSIs between 5–10% and the males had GSIs between 0.35–0.65%, while Shaw and Krahe (2018) used fish with GSIs <0.30% for both sexes. In Gymnotus cf. carapo, GSIs over 5% in females represent mature ovaries that are prepared for spawning (Rotta et al. 2023). Male Sternarchogiton nattereri collected in the field during the rainy season, when fish breed, have GSIs ~0.5–0.7% (Cox-Fernandes et al. 2010). Thus, the GSIs in our sample fell well within the typical range found in spawning knifefish. cyp19a1b expression varied across both species and sexes. When there was a detectable species difference, cyp19a1b mRNA expression in the TSd and ELL was greater in A. albifrons than in A. leptorhynchus. Greater aromatase expression in TSd and ELL in A. albifrons suggests that more androgens are being converted to estrogens than in A. leptorhynchus. This is consistent with estrogens serving as potent modulators of auditory processing in birds and frogs (Yovanof and Feng 1983, Pinaud and Tremere 2012, Remage-Healey 2012, Caras 2013). A higher rate of E2 synthesis in the TSd and ELL in A. albifrons is also consistent with the finding that A. albifrons had greater esr1 expression. In contrast to A. leptorhynchus, which expressed more ar2 in the TSd and may be maintaining higher levels of T in electrosensory brain areas, sensory processing in A. albifrons may be more strongly modulated by estrogens. A. albifrons and A. leptorhynchus differ in the direction of sexual dimorphism and steroidal regulation of EODf (Dunlap et al. 1998). Differences between these species in the relative expression of genes encoding androgen receptors, estrogen receptors, and aromatase suggest that steroidal regulation of perception of electrocommunication signals might also differ between these species.

Brain aromatase expression often differs between sexes (Balthazart et al. 2011). Aromatase often regulates male-specific behaviors by converting T to E2 locally in the brain (Balthazart and Foidart 1993, Brooks et al. 2020). In teleosts, however, brain aromatase is expressed at higher levels than in many other taxa, is often more abundant in females, and can vary with reproductive state and sex steroid levels (Diotel et al. 2010, Le Page et al. 2010, Okubo et al. 2011, Maruska et al. 2020). For example, in plainfin midshipman fish, females and silent, sneaker males have significantly higher levels of aromatase levels in the vocal hindbrain than vocal males do (Schlinger et al. 1999). While we did not demonstrate a statistical sex difference in cyp19a1b expression due to the limited sample size and amount of variation in cyp19a1b abundance, there was still a consistent trend across replicates for females to have more mRNA for this aromatase gene. This is consistent with aromatase expression in the Pn of these species, where females also consistently expressed more cyp19a1b than males (Proffitt 2022). Even though esr1 abundance in the TSd and ELL was sexually monomorphic, females may be locally metabolizing more E2 and have a greater proportion of bound estrogen receptors (ERs) relative to males. Greater aromatase expression in the TSd and ELL in females may also reduce androgenic signaling by locally converting T to E2.

4.3. Steroid-related gene expression in electrosensory versus electromotor regions

Interestingly, patterns of variation in steroid receptor mRNA abundance in the electrosensory brain regions differed from that in the electromotor region that controls EODf, the Pn. In the Pn, ar1 is expressed at much higher levels than ar2 in both A. albifrons and A. leptorhynchus (Proffitt 2022, Proffitt et al. 2023). We observed the opposite pattern in the TSd and ELL, where ar2 was more highly expressed than ar1. ar1 encodes for androgen receptor α (ARα) and ar2 encodes for androgen receptor β (ARβ). Interestingly, the sequence of the ARα ligand binding domain in A. leptorhynchus differs substantially from that in the ancestral AR or the A. albifrons ARα, rendering the A. leptorhynchus ARα unable to bind androgens (Proffitt et al. 2023). In the Pn in A. leptorhynchus, ar1 and ar2 gene expression is tightly correlated, suggesting ARα function might depend on heterodimerization with ARβ (Proffitt et al. 2023). In contrast, in A. albifrons, both AR paralogs may function independently. We, however, did not observe strong correlations between expression of ar paralogs in the A. leptorhynchus ELL or TSd. Thus, another potential explanation for why ar2 is expressed at higher levels in the TSd in A. leptorhynchus relative to A. albifrons is that higher levels of ar2 compensate for the fact that the product of ar1 (i.e., ARα) is not directing androgen-dependent changes in gene transcription. As with androgen receptors, different estrogen receptor paralogs are also expressed in electrosensory and electromotor brain regions. esr1 encodes estrogen receptor α (ERα) and esr2b encodes estrogen receptor β2 (ERβ2). In the Pn of both A. albifrons and A. leptorhynchus, esr2b was the primary esr expressed, and esr1 was undetectable (Proffitt 2022), whereas in the ELL and TSd, esr1 was the primary esr, and esr2b was undetectable. Overall, ARβ and ERα may modulate sensory circuits more strongly than ARα and ESRβ2, which are stronger regulators of electromotor function instead.

Due to the whole genome duplication in teleosts, paralogous receptors often subfunctionalize and exhibit tissue-specific gene expression (Ogino et al. 2009, Arnegard et al. 2010). For example, in knifefishes, of the two sodium channels that were ancestrally expressed in skeletal muscle, one diverged to facilitate a persistent sodium current and is primarily expressed in electric organ, which allowed for the diversification of EOD waveform (Thompson et al. 2014, Thompson et al. 2018). ARs often serve different functions as well. For example, in African cichlids, the AR paralogs differentially influence social status. ARα more strongly regulates behavior while ARβ directs testes growth and dominant coloration (Alward et al. 2020). In medakas, ARα regulates tooth enlargement and courtship displays, while ARβ regulates male fin morphology (Ogino et al. 2023). The specificity of steroid receptor expression in motor versus sensory regions and the degree of sequence divergence between receptor subtypes, especially between ARα and ARβ in A. leptorhynchus (Proffitt et al. 2023), might suggest that androgen and estrogen receptor paralogs are playing different roles in mediating electrocommunication interactions.

4.4. Conclusions

Here, we demonstrate for the first time that steroid-related genes are expressed in electrosensory brain regions in electric knifefishes. This finding is, in and of itself, a novel contribution. While ample evidence highlights hormonal modulation of signal production, the role of steroids in influencing sensory perception of signals is more recently being appreciated. We also found that the expression of androgen and estrogen receptors varied by species and that the abundance of different receptor subtypes is modular across electrosensory regions and differs from that in electromotor regions. Overall, our data support the idea that androgens and estrogens have the potential to influence species- and maybe sex-specific sensory processing mechanisms in the electric sense. These results also highlight that the mechanisms by which hormones influence signal production versus signal perception may vary.

Highlights.

• Gonadal steroid hormones often modulate signal production and perception

• Steroid-related genes are expressed in electrosensory brain regions

• Androgen and estrogen receptor and aromatase gene expression varied across species

• Steroids have the potential to modulate species-specific electrosensory processing

Data availability

Raw data is publicly available on Figshare (10.6084/m9.figshare.24855759).