Hippo-vgll3 signaling may contribute to sex differences in Atlantic salmon maturation age via contrasting adipose dynamics

Abstract

Background

Sexual maturation in Atlantic salmon entails a transition in energy utilization, regulated by genes and environmental stimuli in sex-specific manner. Males require less energy, in the form of adiposity, to mature and typically mature younger than females. Maturation age is also influenced in a sex-dependent fashion by the vgll3 genotype (vestigial-like 3), a co-factor in the Hippo pathway. The underlying molecular processes of sex-dependent maturation age, and their interplay with adiposity and vgll3 genotypes, remain unclear.

Methods

To elucidate the mechanisms underlying sex- and genotype-specific maturation differences, we investigated the association of early (E) and late (L) maturation vgll3 alleles with the transcription of > 330 genes involved in the regulation of the Hippo pathway and sexual maturation, and related molecular signals in brain, adipose, and gonads.

Results

The strongest effect of vgll3 genotype was observed in adipose for females and in brain for males, highlighting sex-specific expression differences in association with vgll3 genotype. Genes related to ovarian development showed increased expression in vgll3*EE compared to vgll3*LL females. Moreover, vgll3*EE females compared to vgll3*EE males exhibited reduced markers of pre-adipocyte differentiation and lipolysis yet enhanced expression of genes related to adipocyte maturation and lipid storage. Brain gene expression further showed sex-specific expression signals for genes related to hormones and lipids, as well as tight junction assembly.

Conclusions

Overall, these sex-specific patterns point towards a greater lipid storage and slower energy utilization in females compared to males. These results suggest Hippo-dependent mechanisms may be important mediators of sex differences in maturation age in salmon.

Background

The timing of sexual maturation can have dramatic effects on survival and reproductive success [1] and is influenced by environmental cues and genetic mechanisms [2]. Optimizing maturation timing is vital in aquaculture of species like Atlantic salmon, as it directly affects growth, flesh quality, and production efficiency, making it key to sustainable farming [2]. Maturation age often exhibits sex-specific differences within species [3, 4]. In Atlantic salmon, sexual maturation is linked with seasonal environmental changes with differences between sexes [1]. Lipid allocation variation also plays a role in determining maturation age, as individuals must accumulate sufficient energy reserves to initiate maturation [5,6,7]. Genetic factors, notably the vestigial-like family member 3 (vgll3) gene, serve as major determinants of maturation age with sex-specific effects in this species [8,9,10]. Vgll3 influences maturation timing in males as early as < 1 year of age in controlled conditions [11,12,13]. Interestingly, associations with similar traits (pubertal timing, pubertal growth spurt) have been detected with the human ortholog (VGLL3) [14,15,16] and VGLL3 has also been identified as a promoter of sex-biased autoimmune diseases [17].

Previous molecular studies of vgll3 in Atlantic salmon have primarily focused on males for practical reasons such as the convenience of some males maturing as early as < 1 year of age [13, 18,19,20,21,22,23], whereas females usually take 3 or more years to mature [24]. Therefore, the underlying molecular processes of sex-specific maturation patterns have remained largely unexplored. Recent gene co-expression network analyses have indicated that the effects of vgll3 genotypes in males on genes playing a role in the reproductive axis, adipogenesis and neurogenesis are exerted through the modulation of various components of the Hippo signaling pathway [22, 23, 25]. The Hippo pathway is known for its role in biological functions such as controlling organ size in vertebrates [19, 26, 27], regulating adipocyte proliferation and differentiation [28, 29] and high fat diet -induced neural differentiation [30]. In mammals, for instance, YAP1, a major transcription co-factor of the Hippo pathway, is required for adipogenesis [28], whereas VGLL3 functions as an inhibitor of adipogenesis [29], indicating the significance of this pathway in energy acquisition processes. VGLL3 is thought to compete with YAP1, which acts as an inhibitor of the Hippo pathway [19, 31]. Moreover, the Hippo pathway has been implicated in responding to environmental cues, such as changes in diet fat and temperature, at the transcriptional level [32,33,34]. Atlantic salmon is an interesting natural model system for investigating the molecular mechanisms that directly link sexual maturation, energy acquisition, and environmental changes given that vgll3 serves as a major activating transcription co-factor of the Hippo pathway, its tight linkage to maturation and adipogenesis processes [23].

A growing body of evidence highlights the involvement of the Hippo pathway in various sex-dependent biological processes across vertebrates, such as size dimorphism in fish and reptiles [35, 36], innate immunity [37, 38], adrenal gland development in mammals [39], and gonadal development and maintenance in fish [40]. Of particular interest among the components of the Hippo pathway is VGLL3, which has been identified as a key regulator of sex-biased autoimmune responses in humans [41]. Here, VGLL3 influences a network of genes involved in various metabolic, developmental, and reproductive functions [41]. This leads to hypotheses about its potential evolutionary significance in maintaining sex-specific metabolic homeostasis under metabolic stress, with implications for autoimmune-related pathological conditions [42]. However, the molecular details of such sex-dependent roles for VGLL3 in regulating sexual maturation have not extensively been explored in any vertebrate species. Therefore, the early and late alleles of vgll3 in Atlantic salmon present a unique opportunity to investigate this aspect in greater depth.

In this study, we employed a custom-made NanoString gene expression panel to analyze the expression patterns of 333 genes, encompassing components of the Hippo pathway and their associated interacting partners, in the brain, ovary, and adipose tissue of immature female Atlantic salmon with homozygous early or late vgll3 genotypes. We compared the patterns of gene expression in females to previously published data in respective male tissues collected at a seasonal stage coinciding with the onset of sexual maturation in males carrying the vgll3*EE genotype [23]. By doing so, we provide deeper insights into the vgll3-dependent molecular processes underlying its sex-dependent effects on sexual maturation.

Materials and methods

Fish material and tissue sampling

Individuals used in this study include eight males at the immature-2 stage reported in Ahi et al. (2024a, b) as well as eight females reared in the same tanks as these males. Details of the rearing and male sampling can be found in Verta et al. (2020) and Ahi et al. (2024b), respectively. Briefly, individuals from the same population (Oulujoki) and cohort used in Verta et al. (2020) provided access to individuals with known vgll3 genotypes (see Verta et al. 2020 for details on crossing and rearing). Immature females reared in the same tanks were sampled at 1.5 years post-fertilization. Following euthanization by anesthetic overdose of MS222, various tissues, including visceral adipose tissue, brain, and ovary, were collected from the females during the summer (July 4–17). This time point was chosen as it represents a late immature stage, where molecular signals of maturation initiation may begin to emerge in some males, but no visible phenotypic signs of maturation are present in either sex. To avoid inconsistencies, the males included in this study had GSI values predominantly near 0 for both genotypes. Specifically, in vgll3*EE males, two individuals had a GSI of 0.01 and two had 0.111 and 0.13, while in vgll3*LL males, three had 0.01 and one had 0.13. Although GSI values for this stage can generally range up to 0.2 for males, individuals analyzed in this study were selected within the lower end of this range, ensuring they all retained an immature phenotype. However, it is important to note that while the GSI value can serve as a reliable indicator of maturation stage in Atlantic salmon [43, 44], a low GSI does not necessarily mean that immature males have not entered the onset of puberty. Even at an immature stage with a low GSI, males may have already initiated the molecular and cellular processes of puberty, which may not yet be externally visible. The female individuals had an average mass of 41.3 g (range 30.1–68.8 g) and an average length of 19.4 cm (range 15.0–20.3 cm), while the males had an average mass of 33.7 g (range 21.1–76.6 g) and an average length of 17.3 cm (range 14.0–18.5 cm). All females had GSI values below 0.12, confirming their immature status.

RNA extraction and the NanoString nCounter mRNA expression panel

In total, RNA was extracted from 24 tissue samples from 8 immature females: 8 samples each of visceral adipose tissue, brain, and ovary. RNA extraction was performed using a NucleoSpin RNA kit (Macherey-Nagel GmbH & Co. KG) as reported in Ahi et al. (2024b), and female samples were randomized among the male samples from Ahi et al. (2024a, b) used in this study. RNA extraction followed the manufacturer’s instructions, including a built-in DNase step to remove residual gDNA. The extracted RNA from each sample was eluted in 50 µl (adipose and brain) and 80 µl (ovary) of nuclease-free water. RNA quantity was measured with a NanoDrop ND-1000 (Thermo Scientific, Wilmington, DE, USA), and quality was assessed with a 2100 BioAnalyzer system (Agilent Technologies, Santa Clara, CA, USA). The RNA integrity number (RIN) was > 7 for all samples. For each extraction, 100 ng of total RNA was used for the hybridization step in the NanoString panel.

NanoString nCounter is a multiplex nucleic acid hybridization technology that enables assessment of RNA expression of several hundred genes simultaneously [45]. As it requires only small RNA amounts with lower quality than RNA-Seq, lacks an amplification step, and detects very low RNA expression levels, it is particularly attractive for ecological and evolutionary research [19, 23, 46, 47]. The NanoString panel used here extends the panel used in Kurko et al. (2020) by adding more than 140 genes for a total of 337 genes. This panel includes an extensive list of Hippo pathway components and interacting partners [19]. The panel also included probes for age-at-maturity-associated genes in Atlantic salmon: vgll3a and six6a (on chromosome 25 and 9, respectively) and their paralogs vgll3b and six6b (on chromosome 21 and 1, respectively) as well as probes for other functionally relevant genes involved in metabolism, adipogenesis, and sexual maturation (Supplementary File 1). Further details on gene/paralog selection and naming are available in Kurko et al. (2020), and gene accession numbers, symbols, full names, and functional categories as well as RNA hybridization procedures can be found in Ahi et al. (2024b).

Data analysis

In the ovary, all nine candidate reference genes in the panel, including actb, ef1aa, ef1ab, ef1ac, gapdh, hprt1, prabc2a, prabc2b and rps20, were used for data normalization due to their low coefficient of variation (CV) values across the samples. In the adipose tissue, seven genes (excluding actb and gapdh) were chosen for normalization as they also exhibited low CV values. For the brain, the same set of reference genes was used, excluding gapdh because of its high expression variation across samples in this tissue. The raw count data from the NanoString nCounter mRNA expression analysis was normalized using an RNA content normalization factor. This factor was calculated based on the geometric mean of selected reference gene counts for each tissue. Following normalization, a quality control check was conducted, and all samples met the default threshold as determined by the nSolver Analysis Software v4.0 (NanoString Technologies). During data analysis, the mean of the negative controls was subtracted, and normalization of the positive controls was performed using the geometric mean of all positive controls. A normalized count value of 20 was set as a background signal threshold. Below-average background signals were detected in 121, 82, and 77 genes across the samples in adipose, brain, and ovary, respectively, and were removed from further analyses. Sex-specific differential expression and Weighted Gene Coexpression Network Analysis (see below) were only investigated in the two tissues that are analogous between males and females (brain and adipose). Differential expression analysis was conducted using the log-linear and negative binomial model (lm.nb function) as implemented in NanoString's nSolver Advanced Analysis Module (nS/AAM). Sexes and genotypes were selected as predictor covariates in the model, as suggested by nS/AAM. Multiple hypothesis testing adjustment was performed using the Benjamini–Yekutieli method [48] within the software, and adjusted p-values < 0.05 were considered significant (Supplementary File 2).

The Weighted Gene Coexpression Network Analysis (WGCNA version 1.68) R-package (version 5.2.1) was implemented to identify gene co-expression modules, GCM [49]. Since our main interest was the comparison between sexes, all samples from both genotypes (for the adipose tissue and brain, separately) were used as biological replicates, providing sufficient statistical power for WGCNA. To identify sample relationships, hierarchical clustering of samples based on gene expression was conducted. Coexpression networks were constructed via seven steps described in [23]. Then, a conditional coexpression analysis was conducted as in Singh et al. [50]. Coexpression networks were constructed for each sex separately to identify the preservation of female modules in the male network and vice versa. A soft power of 8 was used to construct the adjacency matrix. Finally, module preservation statistics were calculated using WGCNA to test how the density and connectivity of modules defined in the reference dataset (e.g., female brain) were preserved in the query dataset (e.g., male brain) [51]. A permutation test was implemented in the query network to calculate Z-scores and individual Z-scores from 200 permutations were summarized as a Z-summary statistic.

To further characterize the GCMs identified through WGCNA, we used WebGestalt [52] to examine similarities and differences between sexes in the biological processes associated with the genes within each module. Specifically, we tested for changes in module GO-association between the sexes as follows. We first identified GO-associations for each module using a gene-set overrepresentation test at a false discovery rate (FDR) of < 0.05, with a specific threshold for Gene Ontology/Biological Process (GO/BP) level 2 inclusion and all protein-coding genes as background. We then compared the GO/BP associations of each module between the sexes. This was performed for each tissue separately. To predict potential gene interactions and identify key genes with the highest number of interactions (interacting hubs), the identified differentially expressed genes in each comparison were converted to their conserved orthologs in humans (providing the highest amount of validated/studied interactome data in vertebrates) and used as input for STRING version 12.0 using the medium confidence level for predicting each interaction/molecular connection [53]. The predicted interactions between genes were derived from data on structural similarities, cellular co-localization, biochemical interactions, and patterns of co-regulation.

Results

Female gene expression differences between vgll3 genotypes

Assessment of expression differences between vgll3 genotypes in females revealed 15 differentially expressed genes (DEGs) in the brain, 16 genes in the ovary and 29 genes in adipose tissue. In the brain, 10 out of 15 DEGs showed higher expression in vgll3*EE genotype individuals (Fig. 1A). A query of molecular interactions revealed three genes, rhoaf, pparga and frmd6a, showing direct interaction with yap1 (specified with connecting lines between the genes in Fig. 1B). Among these three genes, frmd6a showed lower expression in vgll3*EE individuals whereas the two other genes (rhoaf and pparga) had higher expression in vgll3*EE individuals (Fig. 1B). Similarly in the ovary, 10 out of 16 DEGs showed higher expression in vgll3*EE genotype individuals (Fig. 1A). The predicted interactions revealed six genes (arrb1, amotl2a, pax3b, snai2b, tead1a and tead3a) with direct molecular interaction with yap1. In addition, three of these genes (amotl2a, tead1a and tead3a) had a direct interaction with vgll3 (Fig. 1B). Three of the interacting genes, arrb1, snai2b, and tead3a, had higher expression in vgll3*EE individuals whereas the remaining three had lower expression in this genotype. Unlike the brain and ovary, all 29 DEGs identified in adipose tissue showed lower expression in vgll3*EE genotype individuals (Fig. 1A). The interaction query identified seven adipose DEGs (ajubab, ets1e, foxo1c, lats2a, kdm5bc, tead3a and wwtr1b) with direct interactions with yap1, whereas four genes (akap11a, ets1e, tead3a and wwtr1b) had a direct interaction with vgll3 (Fig. 1B). One of these differentially expressed genes with direct connection with yap1, foxo1c, formed an interacting hub, and three of its interacting genes, pparaa, ppargc and esrrab, appeared to make further interacting hubs by connecting to other genes (Fig. 1B). We did not find any DEGs overlapping between the tissues (Fig. 1C), and adipose tissue was the only tissue where vgll3a was differentially expressed, with lower expression in vgll3*EE genotype individuals.

Fig. 1

Differentially expressed genes in three tissues of female Atlantic salmon with alternative vgll3 genotypes and their predicted interactions. Heatmaps represent differentially expressed genes between vgll3 genotypes in three tissues (A) and their respective predicted interactions in each tissue using STRING v12 (http://string-db.org/) (B). The thickness of the connecting lines between the genes indicates the probability of the regulatory/functional interactions. Blue and yellow colors in A indicate higher and lower expression, and in B, higher and lower expression in vgll3*EE individuals, respectively. A Venn diagram showing the lack of differentially expressed genes overlapping between the comparisons (C)

Sex-specific gene expression differences in the brain

To identify gene expression differences between the female and male brain, we compared individuals from both sexes at a similar developmental time point in the summer (the Immature 2 stage from Ahi et al. 2024), when some males, but no females, were starting to show phenotypic signs of maturation (see “Methods”). We identified 45 DEGs between female and male individuals when both vgll3 genotypes were combined, and when considering vgll3 genotypes separately, 23 and 33 DEGs were identified in female vs. male vgll3*LL and vgll3*EE genotypes, respectively (Fig. 2A). Furthermore, we found a general tendency for DEGs to have higher expression in the brain of immature females whereby higher expression in females was observed in all of the 45 genes for the combined genotypes, 21 out of 23 DEGs for the vgll3*LL, and 29 out of 33 DEGs for the vgll3*EE comparisons (Fig. 2A). Across all three comparisons, expression patterns of three DEGs, arhgef25b, cadm2a and frm6a, were independent of vgll3 genotype, and all these genes had a higher expression level in immature females compared to males (Fig. 2A). These results indicate higher transcriptional activity of the studied genes in the brain of immature females compared to males at this time of the life-cycle.

Fig. 2

Differentially expressed genes between female and male Atlantic salmon and their predicted interactions in the brain. Heatmaps representing differentially expressed genes between the sexes in immature individuals with alternative vgll3 genotypes pooled, and within vgll3*LL and vgll3*EE genotypes (A). Predicted interactions between the overlapping genes, with green, red and blue rings indicating the genes in the Venn diagram (B). The thickness of the connecting lines between the genes indicates the probability of the interaction. Blue and yellow colors indicate higher and lower expression in female individuals, respectively. A Venn diagram showing the numbers of differentially expressed genes overlapping between the comparisons (C). The color coding of the numbers corresponds to the gene colors shown in the list of genes within the heatmaps

We further investigated potential functional/molecular interactions between DEGs showing vgll3 genotype-specific differential expression (colored numbers and circles in Fig. 2B, C). The predicted interactions between these genes revealed extensive and complex regulatory connections between the DEGs in each genotype (Fig. 2B). From the vgll3 genotype-specific comparisons, 13 and 19 DEGs within vgll3*LL and vgll3*EE individuals, respectively, were found to be connected in the interaction network (circled in red and green in Fig. 2B). Moreover, except for aldh1a2 within the vgll3*LL network (green circled), and six1 and rpabc2a/POLR2F within the vgll3*EE network (red circled), all other genes in both networks had higher expression in females (Fig. 2B). Further, some of the genes in each genotype network had high numbers of interactions in the networks including ets1a, egr1d, foxo1c, lats1b and snai2b in the vgll3*LL interaction network, and fgf8b, rhoad, tead1a and tead3a in the vgll3*EE network (Fig. 2B). These findings imply that specific components of the Hippo pathway might be responsible for the elevated transcriptional activity observed in the brains of females compared to males.

Sex-specific gene expression differences in adipose tissue

Similar comparisons to those conducted in brain tissue were conducted in adipose and seven DEGs were identified between female and male individuals when both vgll3 genotypes were pooled. When considering vgll3 genotypes separately, we found six DEGs in the vgll3*LL genotype and 23 DEGs in the vgll3*EE genotype individuals (Fig. 3A). Furthermore, we found that all DEGs in the pooled genotypes and in vgll3*LL had increased expression in females, whereas most DEGs in vgll3*EE (17 out of 23) had higher expression in males. No gene was differentially expressed in all comparisons (Fig. 3B). These results indicate significantly more pronounced and stronger sex-specific expression in the adipose tissue of individuals with the vgll3*EE genotype.

Fig. 3

Differentially expressed genes between female and male Atlantic salmon and their predicted interactions in adipose tissue. Heatmaps representing differentially expressed genes between the sexes in the immature individuals across alternative vgll3 homozygotes, and within vgll3*LL and vgll3*EE genotypes (A). A Venn diagram showing the numbers of differentially expressed genes overlapping between the comparisons and the color coding of the numbers corresponds to the gene colors shown in the list of genes within the heatmaps (B). Predicted interactions between the overlapping genes, with green and red rings indicating the genes in the Venn diagram (C). The thickness of the connecting lines between the genes indicates the probability of the interaction. Blue and yellow colors indicate higher and lower expression in female individuals, respectively

We further investigated potential functional/molecular interactions between DEGs showing vgll3 genotype-specific differential expression (colored numbers in Fig. 3B). In the vgll3 genotype-specific comparisons, two and 13 DEGs within vgll3*LL and vgll3*EE individuals, respectively, were found to be connected in the interaction network (circled in red and green in Fig. 3C). While both genes from the vgll3*LL comparison, cadm2bb and wnt5a, had higher expression in females compared to males, seven genes from the vgll3*EE comparison showed lower expression in the females (Fig. 3C). A few genes in each genotype-specific sex comparison had a high number of interactions in the networks, including wnt5a in the vgll3*LL network, as well as rhoag, foxo1c, and stk4 in the vgll3*EE network (Fig. 3C). Among the genes with a high number of interactions in the predicted network, rhoag and foxo1c showed lower expression in females. Importantly, two key components of the Hippo pathway, frmd6 and stk4/mst1, had the strongest predicted interaction with yap1 and both showed increased expression in females with the vgll3*EE genotype. Taken together, these findings suggest that the pronounced transcriptional differences observed in the adipose tissue of vgll3*EE individuals are likely mediated by key components of the Hippo pathway.

Identification of sex-specific gene coexpression modules in the brain

In order to gain a better overview of sex-specific transcriptional differences of Hippo pathway components and their known interacting genes in the brain, we applied network-based co-expression analyses in which changes between the sexes in each network could be tracked. To do this, we first built gene coexpression modules (GCMs) in the brain of each sex and then investigated the preservation of the identified GCMs between the sexes. In other words, we defined the GCM in one sex and then assessed the preservation of its modules in the other sex. We identified five brain GCMs for females (Fig. 4A, B) of which one GCM (brown) showed relatively high preservation (Zsummary > 2) in males, i.e. most of the genes in this GCM have significant expression correlations in both sexes. Three of the GCMs (yellow, green and blue) showed a low to moderate level of preservation between the sexes (Zsummary = 0–2); and the turquoise GCM, containing the highest number of genes (63 genes), showed the lowest level of preservation (Fig. 4B). Next, a gene set enrichment analysis was conducted for the GCM with the lowest level of preservation between females and males (turquoise) in order to provide insights into the biological processes associated with male vs. female differences. The most common biological processes of the genes in the least preserved (turquoise) GCM included regulation of vitamin D biosynthesis, cell–cell junction assembly, Hippo signaling pathway, cellular response to lipid and steroid hormone mediated signaling pathway (Fig. 4C). Almost half its genes in this turquoise GCM showed no coexpression preservation in males (genes lacking color in Fig. 4C). Removal of unpreserved genes in the turquoise GCM led to loss of significance of three GO terms; Hippo signaling pathway, cellular response to lipid and steroid hormone mediated signaling pathway (non-colored GOs in Fig. 4C). Knowledge-based interactome prediction using genes within the turquoise GCM was performed in order to identify potential interactions between the genes as well as hub genes with highest number of interactions [54]. The prediction of interactions between the genes within this GCM revealed that several genes among those lacking coexpression preservation directly interacted with vgll3/yap1 (represented with lines directly connecting the non-colored genes with vgll3/yap1; Fig. 4D). These genes include ets1, tead1, tead3a and wwtr1b showing direct interactions with vgll3 as well as egr1, kdm5b, pax3, rhoa, rock1, snai2b, snai1, stk3, tead1, tead3a and wwtr1b showing direct interactions with yap1 (Fig. 4D).

Fig. 4

Coexpression modules in the brain of female Atlantic salmon. Visual representation of female module preservation in male individuals. The dendrograms represent average linkage clustering tree based on topological overlap distance in gene expression profiles. The lower panels of the dendrograms represent colors that correspond to the female clustered coexpression modules (GCMs). Top: female GCMs with assigned colors. Bottom: visual representation of the lack of preservation of female GCMs genes in male individuals (A). Preservation Zsummary scores in the male GCMs for female GCMs (colors represent female GCMs). Zsummary < 0 represents lack of preservation (dotted blue line) and Zsummary 0–2 implies moderate preservation (B). The genes in turquoise GCM identified in female brain with least preservation in males. The genes without color in the module are those showing no preserved expression correlation in male individuals and the clockwise arrows above the GCM indicate the direction of genes with highest to lowest expression correlations with other genes within the GCM. In the turquoise GCM, the top enriched GOs are represented, and GOs without color were no longer enriched after removal of the genes without colors (C). Predicted interactions between the genes within the turquoise GCM. Increasing thickness level in the connecting lines between the genes indicates a higher probability of the interaction (D)

We found four GCMs in males and among them, the red GCM showed the lowest level of preservation (Zsummary < 0) compared to females (Fig. 5A, B). The red GCM was also the largest with 42 co-expressed genes and more than half of these genes showed no coexpression preservation in females compared to males (genes lacking color in this GCM in Fig. 5C), suggesting this GCM to be of most interest for identifying genes important for male vs. female differences. The gene set comparison of the least preserved red GCM identified four GOs including regulation of tight junction assembly, Hippo signaling pathway, developmental growth and hormone-mediated signaling pathway (Fig. 5C), suggesting these processes to be important in determining male vs. female differences. Furthermore, removal of unpreserved genes in the red GCM led to absence of significance of all these GOs (non-colored GOs in Fig. 5C).The prediction of interactions between the genes within the red GCM revealed that six genes that lost their coexpression preservation, frmd6, rhoa, rock1, sav1, snai and tead3, had direct interaction with yap1 whereas two unpreserved genes, ets1 and tead3 had direct interactions with vgll3 (Fig. 5D). Importantly, yap1 itself lost coexpression preservation in the red GCM, indicating the involvement of yap1, the major inhibitor of the Hippo pathway, in sex differences in the lowest preserved GCM. This suggests that the observed male-specific transcriptional pattern in the brain might be mediated by yap1-dependent differential activity of the Hippo pathway.

Fig. 5

Coexpression modules in the brain of male Atlantic salmon. Visual representation of male module preservation in female individuals. The dendrograms represent average linkage clustering tree based on topological overlap distance in gene expression profiles. The lower panels of the dendrograms represent colors that correspond to the male clustered coexpression modules (GCMs). Top: male GCMs with assigned colors. Bottom: visual representation of the lack of preservation of male GCMs genes in female individuals (A). Preservation Zsummary scores in the female GCMs for male GCMs (colors represent male GCMs). Zsummary < 0 represents lack of preservation (dotted blue line) and Zsummary between 2 and 6 implies moderate preservation (B). The genes in the red GCM identified in males with least preservation in females. The genes without color in the module are those showing no preserved expression correlation in female individuals and the clockwise arrows above the GCM indicate the direction of genes with highest to lowest expression correlations with other genes within the GCM. In the red GCM, the top over-represented GOs are listed, and GOs without color were no longer enriched after removal of the genes without colors (C). Predicted interactions between the genes within the red GCM. Increasing thickness level in the connecting lines between the genes indicates a higher probability of the interaction (D)

Identification of sex-specific gene coexpression modules in the adipose tissue

Gene coexpression module analyses were conducted for the adipose tissues of both sexes as described for the brain (see above). After building the GCMs for the adipose tissue of each sex, we investigated the preservation of the identified GCMs between the sexes and found no unpreserved GCM in male adipose tissue, indicating that the identified GCMs in males occur in both sexes. However, among the seven GCMs identified in the females (Fig. 6A, B), one GCM (yellow) showed a very low level of preservation (Zsummary < 0) in males and it was thus investigated further using gene-set enrichment as above. We detected two over-represented GO terms, namely regulation of cAMP-dependent protein kinase activity and Hippo signaling pathway (Fig. 6C). In the yellow GCM, we also found that almost half of the genes showed no coexpression preservation in males (9 out of 19 genes lacking color in this GCM in Fig. 6C). Removal of unpreserved genes in the yellow GCM led to loss of significance of the GO term associated with the Hippo signaling pathway (the non-colored GO in Fig. 6C). Knowledge-based interactome prediction using genes within the yellow GCM revealed several hub genes (e.g., amotl2, lats1, rhoa, fgf2 and wnt5a) (Fig. 6D). Furthermore, among those genes not showing coexpression preservation, one gene, kdm5b, had direct predicted interaction with vgll3 and three genes, kdm5b, lats1 and fgf2, had direct interactions with yap1 (represented with lines directly connecting the non-colored genes with vgll3/yap1; Fig. 6D). Among the genes directly interacting with vgll3 and/or yap1, three genes; sav1, amotl2, and lats1; show strong interactions and acting as upstream regulators of the Hippo pathway (all are upstream inhibitors of yap1). Since only lats1 does not show co-expression preservation, this suggests that lats1 may play a critical role in mediating the regulatory effects seen in this tissue. Taken together, these results suggest that the overall sex-specific differences in the expression of Hippo pathway components and its interacting partners might be influenced by lats1 transcription in the adipose tissue.

Fig. 6

Coexpression modules in female Atlantic salmon adipose tissue. Visual representation of female module preservation in male individuals. The dendrograms represent average linkage clustering tree based on topological overlap distance in gene expression profiles. The lower panels of the dendrograms represent colors that correspond to the female clustered coexpression modules (GCMs). Top: female GCMs with assigned colors. Bottom: visual representation of the lack of preservation of female GCMs genes in male individuals (A). Preservation Zsummary scores in the male GCMs for female GCMs (colors represent female GCMs). Zsummary < 0 represents lack of preservation (dotted blue line) and Zsummary 0–2 implies moderate preservation (B). The genes in yellow GCM identified in female adipose tissue with least preservation in males. The genes without color in the module are those showing no preserved expression correlation in male individuals and the clockwise arrows above the GCM indicate the direction of genes with highest to lowest expression correlations with other genes within the GCM. In the yellow GCM, the top enriched GOs are represented, and GOs without color were no longer enriched after removal of the genes without colors (C). Predicted interactions between the genes within the yellow GCM. Increasing thickness level in the connecting lines between the genes indicates a higher probability of the interaction (D)

Discussion

In this study, we explored how [sex and] the genotype of a major maturation age gene, vgll3, interplay[s] with tissue in driving sex-specific expression patterns potentially associated with the onset of maturity in Atlantic salmon. We did this by profiling the expression of known components of the Hippo pathway and their interacting partners in brain, adipose, and gonad tissues at a stage when some males begin to exhibit signs of pubertal initiation while other males and all females remain immature. In general, we observed more extensive vgll3 genotype effects on gene expression in adipose tissue of females compared to the brain and ovary. Compared to the other two tissues, expression differences in adipose tissue seemed to be more linked with Hippo pathway signaling, as several components of this pathway, including vgll3a itself, were differentially expressed between alternative vgll3 genotypes. These results are concordant with earlier research in mammals suggesting the Hippo pathway plays a pivotal role in balancing adipocyte proliferation vs. differentiation [28]. For example, vgll3 has been reported as an inhibitor of adipocyte differentiation in mice [29], and also the activity of Yap has been suggested to be indispensable during adipogenesis [28]. Additionally, a major Hippo pathway kinase, encoded by lats2, which was differentially expressed in female adipose tissue in our study, is known to promote the lipolysis process in mouse adipocytes [55]. Our recent findings in male Atlantic salmon also imply that vgll3 and its associated Hippo pathway have extensive effects on transcriptional changes in adipose tissue in relation to sexual maturation, as well as linking adipogenesis and seasonal changes in this species [23].

In contrast to adipose tissue, in the female brain, there were fewer DEGs, and none of the major components of the Hippo pathway were found to be differentially expressed between the vgll3 genotypes. This suggests a potentially significant difference in the functional role of Hippo pathway signaling between the brain and adipose tissue in females, at least at this developmental time point. In contrast, at the same immature stage, the brain of males showed very distinct transcriptional activation of the Hippo pathway between the genotypes [25]. In males, tead2, encoding a major transcription factor of the Hippo pathway, and three interacting partner genes of the Hippo pathway (kdm5b/jarid1b, mc4ra, and foxo1c) that play roles in the central regulation of the onset of puberty [56,57,58], had higher expression in the brain of individuals with the vgll3*EE genotype [25]. None of these genes were differentially expressed in the female brain between the vgll3 genotypes, indicating a sex-specific difference in the involvement of Hippo pathway components in central sexual maturation signals at this developmental timepoint (Fig. 1). However, four additional genes—dlk1b, pgra, rhoaf, and zic1b—which encode other interacting partners of the Hippo pathway, were found to have higher expression in the brains of females with the vgll3*EE genotype. This is particularly striking because the orthologs of these genes (DLK1, PGR, RHOA, and ZIC1) are all known to be involved in the central regulation of pubertal onset in human [59,60,61,62]. Furthermore, two of these genes, DLK1 and PGR, have also been demonstrated to have sex-specific roles during puberty.

In the ovary, we found a lower number of DEGs between vgll3 genotypes than in adipose tissue. However, unlike the brain, major components of the Hippo pathway were differentially expressed, such as amotl2a, arrb1, tead1a, and tead3a. Recent studies have shown that the core components of the Hippo pathway play important roles in mammalian ovarian physiology, including ovarian development, follicle development, and oocyte maturation (reviewed by Clark et al. [63]). For instance, the higher expression of arrb1 in the ovary of fish individuals with the vgll3*EE genotype could thus indicate enhanced ovarian development, as arrb1 is involved in cellular responses to hormones and growth factors and is an important marker of developing ovary [64]. Among the interacting partners of the Hippo pathway, snai2b, a known marker of primordial ovarian follicles [65], also showed higher expression in the ovary of vgll3*EE individuals. Another notable gene with increased expression in the ovary of vgll3*EE individuals was esrra (nr3b1), which encodes an orphan estrogen receptor with an important role in angiogenesis during ovarian development [66]. Moreover, we found reduced expression of cyp26a1, which encodes an enzyme that inhibits ovarian development by blocking retinoic acid (RA) signaling, in the ovary of individuals with the vgll3*EE genotype [67]. These findings suggest potential differences in Hippo pathway-mediated ovarian development between the vgll3 genotypes in Atlantic salmon, with possibly more advanced ovarian development in vgll3*EE individuals. However, future studies on ovarian development, including time series analyses from immature to mature females, similar to the recent work focused solely on vgll3 expression [68], are necessary to fully understand these differences. Such studies should explore not only the expression of the entire components of the pathway but also the changes at cellular-level, offering deeper insights into the regulatory interactions that govern ovarian maturation.

Gene expression differences suggest differences in lipolysis capacity and adipogenesis of females with distinct vgll3 genotypes

We found lower expression of vgll3a in the adipose tissue of vgll3*EE females, consistent with previous findings in males where we observed differences in lipid content and gene expression patterns in liver and adipose tissues between the vgll3 genotypes, suggesting vgll3*EE males may store larger lipid droplets in the spring/summer (many months before spawning time), whereas vgll3*LL individuals store in the autumn [23, 69]. This supports the scenario whereby vgll3*EE individuals have a higher adipogenesis capacity in the spring in both sexes. However, a closer examination of the DEGs between the genotypes adds further details to the interpretation of the results. For example, vgll3 and TAZ (WWTR1), both of which had reduced expression in vgll3*EE individuals, which are described as inhibitors of the terminal stage of adipocyte differentiation [29, 70]. On the other hand, we also found lower expression of initial-stage adipogenesis markers, such as ppargc, cebpda, ajubab and esrra, in vgll3*EE individuals [71,72,73,74]. Specifically, while pre-adipocyte differentiation might be at a lower level in vgll3*EE individuals (due to reduced ppargc, cebpda, ajubab and esrra expression), the terminal stage of adipocyte maturation might be promoted (due to reduced vgll3a and taz/wwtr1b expression) and the opposite may be true for vgll3*LL individuals. This complex transcriptional signature suggests that the genotype difference may lie in specific stages of adipogenesis (e.g., heterochrony) rather than in overall adipogenesis capacity. In other words in vgll3*EE individuals, the pre-adipocyte differentiation phase has already been completed, and the adipocytes have entered the terminal stage of differentiation. In contrast, in vgll3*LL individuals, the pre-adipocytes are still in the early stages of differentiation and have not yet progressed to their final stage. In addition, we found reduced expression of lipolysis factors, including lats2, foxo1c, and pparaa, in vgll3*EE female adipose [55, 75, 76]. This reduction may indicate an increased lipid-storing capacity in the vgll3*EE genotype, as the expression of lipolysis markers is typically reduced when lipid storage is prioritized. Together, these findings suggest reduced pre-adipocyte differentiation and lipolysis capacity, alongside enhanced adipocyte maturation and lipid storage capacity in vgll3*EE females during the summer. This is concordant to findings in males at the same developmental time point [23, 77].

Extensive sex-specific differences in transcription of the Hippo pathway components in the adipose tissue of vgll3*EE individuals

Assessment of adipose tissue transcriptional differences between the sexes revealed more pronounced sex differences in vgll3*EE individuals compared to vgll3*LL individuals (Fig. 3). Sex-specific expression patterns in vgll3*EE individuals were observed in components of the Hippo pathway (such as stk4, frmd6c, and pcdh18a) as well as interacting partners of this pathway that play important roles in adipogenesis and lipolysis (such as esrra, foxo1c, rhoag, and cebpda/b). Among the Hippo pathway components, stk4 is known to have an important role in adipogenesis, with increased activity leading to augmented adipose mass and obesity while reducing the energy expenditure of adipose tissue by impairing mitochondrial function [78]. Thus, the higher expression of stk4 in vgll3*EE females may indicate weaker energy expenditure performance compared to males with the same genotype, resulting in more adipose mass gain. Consistently, the lower expression of a RhoA paralog gene (rhoag) in vgll3*EE females indicates a higher capacity for adipogenesis and lipid droplet storage, as RhoA is a major suppressor of both processes in mammalian cells [79]. Moreover, a pre-adipocyte differentiation marker, esrra [71], was found to be induced in vgll3*EE females, concordant with increased adipogenesis in these females. Higher expression of a RND3 paralog gene (rnd3b), encoding a key inhibitor of lipolysis [80], and reduced expression of foxo1c, an inducer of lipolysis [75], in vgll3*EE females suggests potentially reduced lipolysis in their adipose tissue. These results suggest higher fat accumulation and adipogenesis in vgll3*EE females compared to the males with the same genotype, but potentially with a lower capacity for energy expenditure.

Sex-specific links between the Hippo pathway and cAMP-dependent protein kinase activity in adipose tissue

Our co-expression analysis revealed that only one GCM identified in the female adipose tissue was not preserved in the males (yellow module in Fig. 6). This means that most of the gene expression correlations within the yellow module were absent in males, meaning the functional relationships between these genes were not maintained in the adipose tissue of males. This GCM in females included correlated expression of several components of the Hippo pathway (such as arrb1, lats1b, amotl2a, and sav1) and genes involved in the regulation of cAMP-dependent protein kinase A (PKA) activity (such as prkar1a, prkar2a, prkar2b, cebpa, myf5, fgf2, and rhoa). PKA activity is a major signal controlling lipid metabolism, particularly lipolysis [81]. For instance, in mice, the loss of prkar1a enhances lipolysis in adipose tissue and leads to rapid weight loss [82], while prkar2b function is required for adipocyte differentiation [83]. In males, the correlation between the Hippo and PKA signals appears to be absent (Fig. 6C). The major Hippo component affected was lats1b (another inducer of lipolysis [84]), as it did not show expression correlation with the PKA components in the adipose tissue of males. Interestingly, the crosstalk between the Hippo and PKA signals is known to be mediated directly through the phosphorylation of LATS1 or indirectly through the activation of RhoA by PKA in mammals [85]. These interactions can lead to synergistic activation of both signals in various tissues. Our result here suggests that the sex-specific mechanism predicted earlier (see above), by which females exhibit higher adipogenesis and lower levels of energy expenditure compared to males, might originate from this link between the PKA and Hippo signals that exists only in the female adipose tissue. This may also indicate that while females might store a larger amount of lipid and gain more fat mass, they may delay in utilizing these reserves, leading to later maturation overall.

Gene expression differences suggest higher activity of the Hippo pathway in the brain of females compared to males

We found genotype-independent induced expression of frmd6a, encoding a major upstream regulator of the Hippo pathway, in the brain of immature females when compared to immature males (Fig. 2). FRMD6/Willin expression is tightly co-localized with GHRH (growth hormone-releasing hormone) in nerve cells, particularly in the nerves densely populated in the hypothalamus and anterior pituitary of vertebrates [86]. GHRH is one of the earliest discovered hypothalamic factors involved in the sexually dimorphic pubertal timing of mammals [87]. However, the regulatory connection between FRMD6 and GHRH during sexual maturation remains to be elucidated. FRMD6 is considered a potent inhibitor of YAP1 and an activator of the Hippo pathway [88], suggesting that the Hippo pathway is generally more active in the female brain than in males at this time point. Despite distinct transcriptional signatures between the vgll3 genotypes, both genotypes reflected higher Hippo pathway activity in the female brain (e.g., increased expression of lats1b in females with the vgll3*LL genotype, and tead1 and tead3a in females with the vgll3*EE genotype) (Fig. 2). Another noteworthy gene with genotype-independent induced expression in females was cadm2a, which encodes another conserved cell adhesion protein highly expressed during vertebrate brain development [89]. CADM2 is also known as a factor linking psychological/behavioral traits and obesity, as well as the brain and adipose tissues [90]. CADM2 has been indicated to have sex-dependent expression during brain development and function in humans [91]. Finally, the third gene identified with genotype-independent induced expression in the female brain was arhgef25b (known as GEFT in mammals), encoding a Rho-GTPase enzyme required for neurite outgrowth, which is responsible for neuronal patterning and connections [92]. A study in clownfish found arhgef25 to be one of the few sex-dependent genes during sexual transition, required during the female stage in this species [93]. These findings suggest a potential molecular axis whereby the brain senses varying energy storage status in adipose tissue and responds in a sex-specific manner. This axis may be triggered by the differential expression of cadm2a in response to adipose tissue energy status, leading to changes in cell adhesion dynamics. These changes could activate Rho-GTPase enzymes (e.g., arhgef25b/GEFT) [94], and the Rho-ROCK signaling [95]. Activated Rho-ROCK then regulates P53 [96], an apoptosis factor that can induce frmd6 expression, as observed in mammalian neural cells [97]. The induction of frmd6 activates the Hippo pathway in the hypothalamus, potentially delaying the onset of puberty in females compared to males. Although speculative, as this proposed regulatory axis is based on observations in mammalian cells, it is worthy of future testing in fishes and other vertebrates.

Sex-specific links between the Hippo pathway and lipid, hormone, and cell adhesion signals in the brain

Our co-expression analysis in the brain showed that there was at least one sex-specific gene co-expression module (GCM) in each sex i.e. gene coexpression was not preserved in the opposite sex (Figs. 4, 5). This indicates that the majority of gene expression correlations observed within the turquoise GCM in the female brain and the red GCM in the male brain were absent in the opposite sex. This suggests that the functional relationships between these genes were not maintained in the brains of males and females, respectively. In the female brain, the largest identified GCM (turquoise GCM in Fig. 4) was the least preserved in the male brain. The turquoise GCM consisted of genes encoding components of the Hippo pathway and molecular processes including signals mediated by lipid and steroid hormone as well as cell–cell junction and vitamin D biosynthesis (Fig. 4C). The identification of lipid-mediated signals in the brain is not surprising, as the brain is known to sense somatic energy storage [98]. While metabolic control of puberty has been studied for decades, the molecular links between fat storage, the brain, and sexual maturation remain underexplored [98]. Recent studies have emphasized the role of lipid-sensing signals (e.g., insulin, mTOR, AMPK) in triggering puberty and fat-induced precocious puberty [99,100,101]. Interestingly, these signals are known to interact with the Hippo pathway [30, 101, 102]. Moreover, excess lipids can directly modulate the Hippo pathway through physical interactions with TEADs [102]. Thus, our findings in the turquoise GCM indicate that there are sex-specific transcriptional signatures of the Hippo pathway components (e.g., tead1 and tead3) and factors mediating lipid sensing and hormonal signals in the brain.

In the male brain, the red GCM (the least preserved in the female brain) comprised genes encoding components of the Hippo pathway and three other molecular processes involved in hormone-mediated signals, developmental growth, and tight junction assembly (Fig. 5). The factors underlying tight junction assembly in the brain are crucial for the formation and integrity of the blood–brain barrier, BBB [103, 104], and sex hormones influence the tightness of the BBB [105, 106]. Importantly, the BBB is vital for brain permeability during puberty, and sex-specific differences in BBB permeability have been demonstrated in mammalian studies [106,107,108]. The Hippo pathway is emerging as a key player in BBB formation, homeostasis, and regeneration through the regulation of tight junction proteins [109,110,111]. Therefore, the distinct sex-specific transcriptional signatures of these components observed in the salmon brain suggests potential Hippo pathway-dependent differences in BBB tightness, which may underlie sex differences in BBB permeability and lead to distinct brain responses to maturation-related circulating molecules (e.g., lipids) also in this species.

Potential limitations and future directions

While this study provides valuable insights into gene expression differences, additional validation using complementary techniques such as qPCR or in situ hybridization could further support the findings. The NanoString technology employed here is highly sensitive, particularly for detecting low-expressed genes, and offers paralog-specific probe design and absolute quantification. Given these advantages, further validation is not strictly required. However, incorporating other validation steps at spatial and temporal levels in future studies would be beneficial where feasible. Another important point is the use of whole tissue samples, which contain multiple cell types, making it difficult to determine the specific cellular sources of differentially expressed genes. Bulk RNA expression profiling lacks spatial and cell-type resolution, and while we relied on known cell-type markers to infer cellular-level changes, we deliberately avoided strong cell-type-specific interpretations. To address this in future studies, methods like single-cell RNA sequencing can help to gain more refined cellular resolution. Moreover, techniques such as spatial transcriptomics, immunofluorescence, or RNA in situ hybridization will allow a more precise localization of differentially expressed genes in specific cell populations. Also, integrating multi-omics approaches (e.g., proteomics, epigenomics) alongside transcriptomics could further enhance our understanding of the regulatory mechanisms driving maturation processes.

Another potential limitation of comparing vgll3*EE males and vgll3*EE females at this late immature stage is that observed gene expression differences may be influenced by both genotype and early transcriptional changes preceding the onset of maturation. Although we specifically selected EE males with GSI values close to 0 to minimize potential maturation effects, it is well established that gene expression differences can occur before phenotypic changes become apparent [20, 22, 26]. While none of the males in this study exhibited external signs of sexual maturation, such as gonadal development, coloration, or behavioral changes, we acknowledge that entry into puberty is a distinct process that may begin at lower GSI values before external markers become visible. Studies on Atlantic salmon have demonstrated that puberty onset can occur even at GSI values as low as 0.05–0.1%, characterized by increased plasma levels of 11-ketotestosterone, Fshb protein production by the pituitary, spermatogonial proliferation, and Sertoli cell activation [44, 112,113,114,115,116,117,118]. Given this, we cannot rule out the possibility that some males in our study, despite their low GSI values, may have already initiated puberty at the molecular and cellular levels. Thus, the possibility that some gene expression differences reflect early molecular changes associated with the onset of maturation rather than genotype alone cannot be ruled out. To address this in future studies, a developmental time series including both sexes across multiple stages (e.g., from early to late immature stages, as well as the maturing stage) would provide a better temporal overview of transcriptional changes. Such an approach would help distinguish genotype-driven differences from maturation-related signals and reduce potential confounding effects. Furthermore, to precisely confirm the gonadal developmental stage and, more specifically, to determine the potential entry into puberty, further characterization should be included. This would involve histological analysis to assess germ cell stages, proliferation, and differentiation activity [44, 112,113,114], along with plasma sex steroid and gonadotropins measurements along with GSI values [44, 115,116,117,118]. These combined approaches would provide a more comprehensive assessment of the transition from the immature stage to the initiation of puberty.

Perspectives and significance

The findings of this study shed light on fundamental biological processes with broad implications across fields such as developmental biology, endocrinology, and aquaculture science. By uncovering sex-specific interactions between lipid metabolism, brain signaling, and the Hippo pathway, this research emphasizes the complex interplay between genetic and physiological factors in determining sexual maturation timing. These insights extend beyond Atlantic salmon, offering a framework for understanding how energy allocation strategies may evolve differently between sexes in other species. The discovery of sex-based differences in lipid storage and central sensing mechanisms also raises intriguing questions about the broader roles of the Hippo pathway in regulating metabolic and developmental traits. From an applied perspective, this knowledge could inform strategies to optimize growth and reproduction in aquaculture, while contributing to evolutionary models of energy trade-offs in sexual development. Future work may explore the translational potential of these mechanisms in other taxa and investigate their implications for broader ecological and conservation challenges.

Conclusions

This study provides significant molecular evidence linking Hippo pathway to sex specific differences in the brain and adipose tissue, which may explain how male sexual maturation often occurs earlier in Atlantic salmon. Our results suggest that females may have higher adipogenesis and lower energy expenditure (lipolysis capacity) compared to males, likely due to sex-specific interactions between PKA and Hippo signaling pathways. In males, increased expression of lipolysis markers in adipose tissue may result in greater energy release, which is sensed in the brain. We also found Hippo-dependent differences in expression of genes encoding tight junction proteins, potentially contributing to greater brain permeability in males. This increased lipid release and potential changes in the brain permeability may underlie sex differences in central lipid sensing process, influencing sex-specific pubertal timing. However, further detailed functional assessments are necessary to validate these suggested differences.