Androgen dysregulates the follicular extracellular matrix and increases pro-fibrotic gene expression in the mouse ovary

Abstract

The extracellular matrix (ECM) defines the biomechanical and biochemical microenvironment of tissues, directing cell behaviour and phenotype. In the ovary, ECM must dynamically remodel in each cycle under hormonal regulation to control follicle development and produce fertilizable oocytes. Dysregulation of this process may result in aberrant formation of ECM as seen in polycystic ovary syndrome (PCOS) whose pathology includes fibrosis of the ovary and which is a major cause of infertility. PCOS is characterised by hyperandrogenism and, here, we investigate the impact of androgens on fibrosis, cell-ECM interactions and mechanosensing. We report an altered network of gene expression related to the genesis of fibrosis. Preantral follicles from C57BL/6 mice (14–15 days postpartum) were stimulated with dihydrotestosterone (DHT, 10nM) in 24/72 hours culture. Expression of fibrosis-associated genes (Eln; Ctgf; Acta2; Plod2; Hpse) significantly increased with androgen (72 h), as did TGF-β signalling (Tgfb1; Tgfb3). We show a direct connection between androgen and mechanosensing within the ovary, with androgen upregulating the mechanosensitive Hippo pathway (Yap1; Lats1; Lats2; Stk3; Stk4; Frmd6) and downstream targets (Ctgf; Axl; Cyr61). Our results highlight hyperandrogenism as a probable driver of the fibrosis in the polycystic ovary, and emphasise the importance of ECM regulation in follicle development and fertility.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-32927-6.

Introduction

The extracellular matrix (ECM) plays a fundamental role in tissue structure and function, regulating processes including cellular differentiation, migration, proliferation and survival^1–5. The ECM itself consists of structural proteins and polysaccharides (e.g. collagens, laminins, hyaluronan)⁶, as well as matrix structural modifiers (e.g. lysyl oxidase, LOX)⁷. ECM function is impacted by the expression of molecules that control the bidirectional cellular response, including cell-ECM receptors (e.g. integrins), components of signalling pathways that mediate functions such as adhesion, mechanosensing and cytoskeletal components. The unique combination of ECM components and cellular responses defines the microenvironment, structure and function within each tissue.

Specifically within the ovary, the ECM microenvironment plays a key role in regulating the development of follicles and the oocytes within⁸. The matrix and components that modify structure, have been shown to participate in a number of key processes, including antrum formation and ovulation, as well as directly structuring the follicle via basal lamina formation^9–11. The ovarian ECM shows a unique and complex spatiotemporal structure, with intricate and extensive remodelling across the reproductive cycle. In terms of structure, ovarian matrix components have been characterised across a number of different species including cow^12,13 human^14–16, mouse¹⁷ and sheep^18,19. All species demonstrate a controlled dynamic range of expression across all follicular stages. Temporal and spatial ECM regulation is essential in promoting healthy follicle development^20,21.

Polycystic Ovary Syndrome (PCOS) is the most common hormonal disorder in women of reproductive age, characterised by the presence of clinical features including oligo/anovulation, hyperandrogenism, associated with an ovarian morphology typified by an excess of follicles and increased stromal tissue. Specifically, histological studies suggest distinct alterations in the matrix and structure of the ovaries of patients with PCOS, highlighting increases in connective tissue, subcortical and cortical stroma and an overall increased stromal volume^22,23. Furthermore there are alterations in gene expression, in the ovaries of women with PCOS, in a number of gene pathways related to ECM components involved in the formation and remodelling of the ECM microenvironment²⁴. Increased stromal hyperplasia and collagenous thickening of cortical stroma, may imply altered pathways associated with ECM formation and regulation. Moreover, fibrosis has the potential to alter the mechanics of the tissue environment, having an impact on follicle development. Such an impact is plausible given evidence that follicles alter their behaviour dependent on the stiffness of their environment^25–27, with in vitro biomaterial-based cultures indicating matrix stiffness can impact follicle growth, development, oocyte competency, antral cavity formation and relative hormone production levels^26,28.

Hyperandrogenism is one of the key clinical features of PCOS and understanding the role that androgens may play, in matrix expression within the ovary, is of key importance. An indication of androgen related regulation of the matrix layout and expression, arises from the study of female to male sex transition patients, who have undergone androgen supplementation during treatment. The ovaries demonstrate similar histological features associated with ovaries from patients with PCOS, with increased hyperplastic collagen and stromal hyperplasia²⁹. Equally, free testosterone and (Dehydroepiandrosterone sulfate) DHEAS levels are correlated to ovarian stromal volume³⁰. Elucidating the changes in ECM-related components upon androgen exposure, would help in the understanding of the pathophysiology of the disorder.

The mechanical properties of tissue microenvironments can impact cell and tissue function³¹. Matrix-mechanical effects can alter functions across a number of diverse pathways³². Such mechanosensing is mediated by receptors including integrins and Piezo1, as well as signalling molecules including the transcription factor YAP-1 and the associated Hippo pathway. Moreover, follicle growth and survival is associated with signalling through the mechanosensitive Hippo pathway^33–36. These results are clearly significant for understanding healthy follicle development, but are also potentially relevant to PCOS, given the likely matrix stiffening that accompanies PCOS-induced fibrosis. In other fibrotic conditions (e.g. in the lung and the heart), it has been proposed that matrix stiffening and the associated cell-mechanosensing response act to reinforce and amplify the fibrotic disease phenotype^31,32. This mechanism could also play a role in PCOS, but since hyperandrogenism is such a central feature of the disease, it will be critical to understand the interplay between androgen signalling and mechanosensing, and how these factors work together to control follicle development.

Here we investigate the effect of direct androgen regulation on ECM and mechanosensing pathways in the ovary, using a mouse in vitro model. Specifically, we demonstrate how androgen impacts the regulation of several extracellular matrix components, pathways associated with matrix deposition and the control of components linked with structural features and mechanotransduction. These results provide a fundamental framework to understand the effects of these changes on extracellular matrix and matrix-associated component expression properties within the mouse ovary. Aberrant regulation of ovarian ECM components provides further evidence for the change in the follicular and tissue phenotype observed in women with PCOS.

Results

Selection of target genes and the presence and absence of their expression in the mouse ovary and follicles of different developmental stages

To understand the role that androgens may have on the expression of matrix, matrix-associated genes, and mechanosensing, a selection panel of targets for investigation were selected. Since our aim is to determine how exposure to androgens impacts matrix- and mechanobiology, genes were selected to this end. This was done initially by using RNA-Seq analysis of granulosa-lutein cells (GLCs) collected from patients with/without PCOS, undergoing IVF^37,38. Due to hyperandrogenism being a key feature of the disorder, many of these genes are likely to be androgen regulated. From these, we selected a subset of genes that are relevant to the extra-cellular matrix, the cellular cytoskeleton, cell - ECM adhesion, or mechanosensing, based on Reactome pathway and gene ontology analysis (Supplemental Fig. 1). To this, we added a further selection of genes that are not specifically identified from the data as being modified in GLCs of patients with PCOS, but whose importance in matrix biology, or mechanosensing is known: Fibronectin (Fn1), Smooth Muscle Actin (Acta2), Connective Tissue Growth Factor (Ctgf) and Elastin (Eln). Figure 1A shows the selected genes with a schematic illustration of their localisation. Known interactions among these genes were determined using the Protein Interaction Database, showing that most of the genes form a large, connected network (Supplemental Fig. 2).

Fig. 1.

Representation of genes highlighted for further analysis. Representation of the genes selected for analysis, their location and proposed functions.

To elucidate whether these selected genes are present in the mouse ovary, a presence and absence RT-PCR screen was performed for different compartments within the ovary. Looking at follicles, we analysed varying sizes and developmental stages (Fig. 2A). The small-follicle group was defined to range from 86 to 117 μm; medium from 113 to 142 μm and the large-follicle group from 152 to 315 μm (Fig. 2B). Additionally, isolated oocytes and granulosa cells were included, as well as ovarian fragments for an overall picture inclusive of stromal tissue.

Fig. 2.

Presence or absence of matrix associated components within the mouse ovary. (A) Image representation of samples used for presence and absence screen (B) Pooled samples of follicles of various sizes used as small medium and large follicle samples. Box and whisker plots of follicle size; each blue circle represents one follicle measured through ImageJ. Boxes represent the 25th and 75th percentile, the line highlights the median point. Whiskers represent the range. (C) RT-PCR screen of mouse samples tissue highlighting the presence or absence of gene transcripts for matrix components highlighted in figure 1. Small medium and large follicles, ovary fragments, granulosa cells and oocytes were analysed. Equal quantities of input cDNA were loaded with each sample. Negative (–ve) sample contained no cDNA. Gel images were cropped and arranged for clarity. Original gel images are presented in Supplemental Figure 8.

The results show that mouse follicular tissue and cells express the matrix components Fn1, Fbn1, Itga9, Cav1, Lama1, Col4a1, Vcl, Eln, Acta2, Lama3, Palld and Fbln7 (Fig. 2C). Expression is ubiquitous across all sample types across each gene studied, with a few exceptions, results show little to no Cav1 expression in small follicles. Within oocytes, nine out of the twelve components are expressed (Fn1, Fbn1, Itga9, Col4a1, Vcl, Eln, Acta2, Lama3, Fbln7 and Palld), but expression of certain matrix components was undetectable (Cav1, Lama1 and Fbln7) (Fig. 2C). In terms of matrix-modifier genes, again their expression is found within almost all samples and genes analysed. Expression is present for Hpse, Plod2, Lox, Mmp19, Has2 and Ssh1 (Fig. 2C). Lastly, the two genes Ctgf and Rhou, associated with cell signalling and mechanotransduction, are shown to be expressed within all cell and tissue types within the ovary (Fig. 2C). Overall, these results validate our target gene selection and confirm that there is active gene expression, within mouse ovaries, of not only structural matrix components but also matrix modifiers and signalling molecules associated with cell adhesion and mechanotransduction.

Localisation of target proteins within mouse ovarian tissue

We demonstrate a clear expression pattern across several different ECM and ECM-associated components (ACTA2, COL4, Laminin, FN1, CTGF, VCL and LOX) (Fig. 3 and 4), in line with other published data^17,39–41. The histological data presented here indicates a clear structural feature within the ovary. Three of the proteins (ACTA2, COL4 and Laminin) form a ring surrounding each follicle (Fig. 3, ACTA2, COL4, Laminin). Expression is localised to a distinct ring of the follicular basal lamina (Laminin and parts of collagen IV) or within the stromal areas in between follicles.

Fig. 3.

Localisation of matrix components in D24 mouse ovarian tissue. Mouse ovarian tissue was labelled with various matrix components (ACTA2, COL4, FN1, Laminin) and counterstained with DAPI (blue). Images represent individual channels and then merged. Dotted line highlights a whole growing follicle. Oo, Oocyte; Gc, Granulosa Cell; Blue triangle, Antral Cavity formation; Str, Stroma.

Fig. 4.

Localisation of matrix modifiers / mechanosensitive component expression in D24 mouse ovarian tissue. Mouse ovarian tissue was labelled with various matrix components (CTGF, LOX, VCL) and counterstained with DAPI (blue). Images represent individual channels and then merged. Dotted line highlights a whole growing follicle. Oo, Oocyte; Gc, Granulosa Cell; Blue triangle, Antral Cavity formation; Str, Stroma.

Considering the localisation of LOX, which crosslinks the ECM, we note it is predominantly expressed within the granulosa cell compartment of the follicle, suggesting a possible role for collagen/elastin stiffening within this region (Fig. 4). Lastly, the expression of CTGF and VCL is shown clearly within the mouse ovary (Fig. 4). Both Ctgf (Connective Tissue Growth Factor) and Vcl (Vinculin) are associated with mechanotransduction within cells and tissues, with vinculin playing a fundamental role in cell adhesion. We show that CTGF is found across the entire ovary, with greater expression found within the granulosa cell layer. Vinculin is found universally across the entire tissue (Fig. 4). Global expression suggests that mechanical signalling can occur throughout the tissue.

Androgen regulated gene expression changes in ECM and ECM-Associated genes within mouse follicles

To understand the effect of androgens on ECM and ECM-Associated components within follicles we performed an isolated follicle culture in the presence of the androgen DHT (10nM). Follicles were isolated from the rest of the ovary and cultured in the presence or absence of DHT for either a 24 or 72-hour timepoint. Follicles in the presence of DHT demonstrate significant increases in growth compared to controls at each timepoint (Fig. 5B, 24 H p ≤ 0.0001 and 48 H p ≤ 0.0001, 72 H p ≤ 0.01).

Fig. 5.

Follicle growth in response to DHT (10nM) of follicles taken from D16/17 mice over a 24/72 hour culture period. Follicles were isolated, using insulin needles, from D16/17 mouse ovaries, before being cultured in 96 well plates in culture medium supplemented with/without DHT. At 24 h half the samples were removed for analysis, the remaining samples continued culture for a total of 72 h. (A) Representative images of follicles grown with or without DHT over a 72-hour period. After 24 h half of the samples were collected for analysis at this timepoint. (B) Percentage change in area of follicles treated with/without DHT (10nM) over 72 h of culture. The area of each follicle was measured using ImageJ. Follicles that were damaged / showing signs of apoptosis were excluded from analysis. n = (24 h) 77, (48 h) 38 and (72 h) 38. Statistical analysis was performed using a Mann-Whitney test where **P < 0.01, ****P < 00001.

We demonstrate that over both timepoints of culture, androgens significantly alter gene expression of several components of the matrix (Fig. 6). The changes in gene expression were also shown to not be an effect of the additional increase in size, due to DHT exposure during culture, and therefore a response to androgens as demonstrated by analysing gene expression from size matched follicles (Supplemental Fig. 3).

Fig. 6.

Log₂ fold change in matrix-associated component gene expression of 24- and 72-hour cultured follicles with/without DHT (10nM). Log₂ fold change in gene expression of components and associated components of mouse ovarian follicles treated with/without DHT (10nM). Light blue columns denote structural components of the ECM, purple columns denote modifiers of the ECM and green represents signalling pathway components. Normality of the data was calculated using the D’Agostino Pearson test and a Mann-Whitney test was performed. Individual graphs highlighted in Supplemental Figure 4 and 5. Each dot represents a sample where n of samples was between 4 and 8. *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 00001. Columns represent mean and SEM.

In the first instance, after 24 h of culture, eleven of the thirteen matrix components are significantly downregulated (Col4a1, Lama1, Lama3, Acta2, Itga9, Fbn1, Fn1, Col11a1, Eln, Cav1 and Fbln7, p ≤ 0.05) with varying degrees of reduction. The results demonstrate downregulation in nearly all structural components at 24 h. At 72 h of culture seven of the thirteen genes see significant changes in expression (Lama3, Acta2, Vcl, Fbn1, Fn1, Col11a1 and Eln p ≤ 0.05), with Eln now upregulated, while Lama3 undergoes a major downregulation. We note that several genes show consistent downregulation at both timepoints, Lama3, Fbn1, Fn1 and Col11a1 (Fig. 6, p ≤ 0.05).

In contrast to these significant changes in expression of matrix structural components, there is little or no androgen-induced change in components that are linked with ECM/matrix modification (Fig. 6).

Androgen induces a reduction in expression genes associated with the basement membrane in follicles

Considering further the changes in expression of structural components: several genes whose expression level changes upon androgen exposure are associated with the follicular basement membrane. This occurs across both timepoints of culture. At 24 h there is a reduction in Col4a1 and both laminins Lama3 and Lama1 (Fig. 6, p ≤ 0.05). Further to this at 72 h there is an over two-fold reduction in Lama3 (Fig. 6, p ≤ 0.01). Moreover Fbn1 (Fibrillin 1), downregulated across both timepoints (Fig. 6, p ≤ 0.05), is known to colocalise with perlecan at the site of basement membranes playing a role in its overall structure⁴².

Androgen regulation of gene expression indicates increases in genes associated with fibrosis within mouse follicles

We demonstrate an increase in pro-fibrotic gene expression, mediated by androgen treatment within follicles. This includes clear upregulation of a number of genes that contribute to tissue fibrosis, Eln, Rhou, Plod2 and Hpse; including an over two-fold increase in Acta2 and over a six-fold increase in Ctgf at 72 h of culture (Fig. 6, p ≤ 0.05). Moreover, both signalling components Ctgf and Rhou demonstrated significant increases in expression after prolonged culture (Fig. 6, p ≤ 0.05). Additionally, both Plod2 and Hpse genes increase in expression at 72 h (Fig. 6, p ≤ 0.05), and these genes are both are associated the pathogenesis of fibrosis^43,44. Specifically, as they are matrix modifiers, they likely play a role in fibrotic remodelling of matrix.

Androgen regulation of gene expression in genes associated with mechanotransduction within mouse follicles

There are several alterations in genes associated with cell adhesion and mechanotransduction pathways (Fig. 6). Both Vcl (Vinculin) and Itga9 (Integrin alpha-9) were differentially expressed in in response to androgen treatment (increase of Vcl at 72 h; decrease of Itga9 at 24 h, p ≤ 0.05). Vinculin is an adhesion protein, associated with focal adhesions, that helps mediate force transmission processes⁴⁵. Demonstrated within the histological images (Fig. 4), vinculin expression is ubiquitous across the ovary, suggesting an important role in ovarian functioning. Integrins are another component associated with focal adhesion structure and play a role in mechanotransduction within cells and tissues⁴⁶. Integrin alpha-9 is one of many integrin isoforms and data are limited regarding its expression within the ovary. Gene expression of Itga9 shown here is novel in terms of expression within the mouse ovary.

Genes associated with actin/actin cytoskeleton within mouse follicles are regulated by androgens

Another subset of genes that are differentially regulated by androgen treatment are those associated with actin and the actin cytoskeleton (Fig. 6). Ssh1 (Slingshot Protein Phosphatase 1) is shown to decrease in expression in response to androgens at 24 h of exposure (Fig. 6, p ≤ 0.05). Ssh1 is known to regulate actin filament dynamics by dephosphorylating and activating the actin depolymerising factor cofilin, subsequently binding actin and promoting its disassembly/depolymerisation^47,48.

Acta2, another actin component, shows an increased expression in response to androgens at 72 h (Fig. 6, p ≤ 0.001). We note that Acta2, while we have classified it as a cytoskeletal component, can exist in the extracellular space and thus effectively contribute to the mechanical properties of ECM. Here, Acta2 (Fig. 3) forms a ring around the follicle within the outer basement membrane/stromal region. While androgens cause downregulation at 24 h (Fig. 6, p ≤ 0.05), the effect is more than reversed at 72 h with an almost 3-fold increase in expression compared to the control (Fig. 6, p ≤ 0.001).

Increased expression of Hippo and Tgf-β pathways signalling components after treatment with androgens within follicles

Due to the changes observed across a number of ECM and matrix-associated genes, signalling pathways associated with fibrosis and mechanotransduction were analysed to understand their role in mediating this response. These pathways included the Hippo pathway and Tgf-β pathway.

We show here that androgen treated follicles demonstrated clear increased expression of several Hippo pathway genes, Lats1, Lats2, Stk3, Stk4, Yap1 and Frmd6 (Fig. 7A, p ≤ 0.05). It is clear that androgens play a role in the regulation of Hippo pathway components, which could alter Hippo pathway and consequently YAP activity in the follicle.

Fig. 7.

Log₂ fold change in Hippo/Tgf-β pathway component gene expression of 72-hour tube cultured follicles with/without DHT (10nM). (A) Log2 fold change in expression of components of the Hippo pathway. (B) Fold change in gene expression of downstream targets of the Hippo pathway. (C) Fold change in expression of components of the Tgf-β pathway. Normality of the data was checked using the D’Agostino Pearson test and an unpaired t-test or Mann-Whitney test was performed. Individual graphs highlighted in Supplemental Fig. 6. Each dot represents a sample where n of samples was between 6 and 9. *P < 0.05, **P < 0.01, ****P < 0.0001 columns represent mean and SEM.

To understand whether the pathway is signalling, beyond an increase in its associated components, downstream targets of Hippo signalling were assessed. These results indicate upregulation, of Ctgf, Axl and Cyr61, key downstream targets of the hippo pathway (Fig. 7B, p ≤ 0.05). Upregulation of downstream targets suggest that YAP signalling is active within the nucleus, which would drive changes in the follicle.

Androgen regulation increases certain TGF-β signalling pathway components within mouse follicles

Another well characterised pathway associated with follicle development, fibrosis and matrix remodelling / deposition is that of the Tgf-β pathway. Androgen regulation of Tgf-β components shown demonstrated significant increases in Tgfb1 (near two-fold, p ≤ 0.01) and Tgfb3 (Over two-fold, p ≤ 0.01) (Fig. 7C).

Discussion

Target gene identification and characterisation

A panel of gene targets were identified which are related to the ECM structure and function. The selected targets were characterised across various cell types and staged follicles. Their expression is expected considering their fundamental role in follicle development⁴⁹. The expression of oocyte-derived matrix components, has been previously demonstrated⁵⁰, with the present results adding to data on oocyte expressed factors. Of the components analysed for protein expression there is a clear expression to localised regions. Spatiotemporal variation of ECM protein expression is a feature previously described by Berkholtz et al.¹⁷. , who demonstrated a similar effect: localisation of collagen (I and IV), laminin and fibronectin to growing mouse follicles. Structurally, three of the key load-bearing proteins (ACTA2, COL4 and Laminin) are spatially located as a ring (either as a basal lamina or surrounding the follicle), which may have the potential to alter follicle expansion. Controlled expression of these matrix proteins is key in follicle development and aberrant expression has the potential to impact growth and competency of the developing follicle.

Androgen exposure altering gene expression

Firstly, it is clear that androgens promote enhanced follicle growth when cultured with DHT. This phenomena itself has been observed across species including cow⁵¹, mouse⁵² and rhesus monkeys⁵³. Moreover, women with PCOS, for whom hyperandrogenism is a characteristic of the disorder, exhibit a higher proportion of preantral and growing primordial follicles⁵⁴.

Our data emphasise that several genes are differentially regulated with androgen exposure, with clear upregulation and downregulation, across either timepoints. The impact of these changes is discussed herein.

Various research groups have shown changes in ECM structural and non-structural components within the ovary of women with PCOS⁵⁵. Changes in structural component expression may permit follicle expansion, a hypothesis put forward by Rodgers et al.¹⁰. In the present study, downregulation of numerous structural components does not persist at 72 h. This potentially indicates a feedback mechanism to androgen exposure over time, or that factors associated with tissue remodelling require longer time frames to promote changes.

Our results highlight little to no changes in elements associated with the modification of the ECM. This result is surprising as research by various groups has highlighted that in both follicular fluid and cultured GLCs there are higher levels of MMP-9 and MMP-2, suggesting that greater MMP activity and hence presumably matrix remodelling⁵⁶ is a feature of PCOS. It may be that such remodelling involves genes not investigated here.

Basal lamina reduction

We demonstrate an androgen dependent reduction in components associated with the basement membrane/basal lamina. The unique structure and phenotype of the basal lamina is well known to play a role in follicle function and can direct follicle and oocyte competency⁵⁷. Components of the basal lamina have also been shown to be tightly regulated throughout follicle development, indicating their specific role within follicle growth and survival¹⁰. Structural characteristics of the basal lamina also permit growth and development. The basal lamina expands in area during follicle growth, and this is accompanied by changes in expression of basal lamina components according to developmental stage^13,58. The differences in bonds between laminin and collagen fibers has led to the hypothesis that the switch in composition permits the expansion of the follicle during development¹⁰. Here, we demonstrate that decreases in laminin components (alpha 1 and 3) occur after androgen treatment; this could have the potential to lead to disruption in the function of the basal lamina and its growth.

Fibrosis

We demonstrate here androgen dependent changes in gene expression, of a number of fibrosis associated genes. Fibrosis itself is characterised by excessive deposition of matrix proteins, which can ultimately lead to tissue stiffening and impact tissue function. Structural components that contribute to fibrosis include Eln and Acta2 (Elastin and alpha smooth muscle actin); both of these showed increased expression on androgen stimulation, in the present study. Increased expression of elastin is known to be a factor in fibrosis of the liver⁵⁹, renal tissue⁶⁰ and lung⁶¹. Knockdown of Acta2 is shown to reduce fibrosis within the liver, including a reduction in collagen 1 expression⁶². Hence the here observed changes in expression of these components could demonstrate alterations towards a more fibrotic state of the ovary.

Aside from structural elements, we show changes in several factors associated with tissue remodelling. Ctgf is a member of the CCN family of proteins and its over-expression is implicated in diseases involving fibrosis⁶³. Evidence highlights its role in the increased deposition of collagen⁶⁴ as well as fibronectin and α-SMA (Alpha smooth muscle actin (ACTA2))⁶⁵. In rats, DHEA (dehydroepiandrosterone) treatment induces a significant increase in Ctgf expression in ovarian tissue, mirroring our findings⁶⁶. Changes in Rhou are less well characterised in terms of its role in fibrosis, however mice under bleomycin treatment to induce lung fibrosis, exhibited a 3.5 fold upregulation⁶⁷. Moreover, the role of Rho kinases in profibrotic pathways is becoming more evident due to their role in reorganisation and control of the actin cytoskeleton⁶⁸. We further note that Ctgf can stimulate the upregulation of Acta2⁶⁹, hence the here observed changes in Ctgf and Acta2 expression may be linked.

We further demonstrate upregulation of two enzymes linked with fibrotic progression. Hpse, an enzyme that cleaves heparan sulfate side chains, has been shown to regulate renal fibrosis in mice, while in contrast Hpse inhibition (with Roneparstat) reduced enzyme levels and reducing fibrosis⁷⁰. Additionally, Plod2, promotes collagen assembly, potentially relevant given that increased collagen deposition is a hallmark of fibrosis⁴³.

Taken together, our results thus support the hypothesis that androgen treatment can be a driver of a fibrotic phenotype within ovarian tissue. Considering PCOS, hyperandrogenism and increased fibrosis are both markers of this condition. Hence our results suggest the idea that the fibrotic phenotype, exhibited by the ovary in PCOS, is at least partially caused by excess levels of androgens. Fibrosis within the ovary has the potential to alter the mechanics of the tissue environment within the ovary itself, potentially having an impact on follicle development, of which anovulation is another hallmark of the disorder.

Alterations in pathways associated with mechanotransduction and fibrosis

To gain a better understanding of dysregulated pathways in the ovary associated with androgen exposure and tissue fibrosis/mechanotransduction, we investigated elements of the Hippo Pathway as well as the TGF-β pathway.

The Hippo signalling pathway is highly conserved and acts to limit organ size, preventing overgrowth as well as mediating mechanical signals into cellular responses⁷¹. A number of studies show the importance of the pathway in normal follicle function, e.g., active Yap1 localisation is predominantly located within proliferative granulosa cells, demonstrating its function in the mechanics of follicle growth. In the present study, our results indicate that androgen increases expression of several components of the Hippo pathway. Assessing the activity of the pathway we assessed a number of its downstream targets.

As previously mentioned, we also see increases in Ctgf, a downstream target of the Hippo pathway. Another Hippo pathway downstream target, Cyr61, is also shown to increase in expression to androgen treatment. Cyr61 is another secreted matrix associated signalling molecule as part of the CCN gene family⁷². In contrast to Ctgf’s profibrotic role, data in cultured human skin fibroblasts indicate a protective / antifibrotic role that CYR61 expression has in reducing the expression of type I procollagen and increases in the matrix enzyme MMP-1, initiating collagen fiber degradation. This anti-fibrotic effect is also shown in fibroblasts from Systemic sclerosis (SSc) patients. Overexpression of CYR61 in SSc fibroblasts results in the reduction of profibrotic genes, such as COL1A1 and ACTA2⁷³. Moreover, it increases the expression of MMPs 1 and 3, associated with collagen degradation. The protective role it may play within the androgen treated follicle is less likely to be active as additional results shown demonstrate that Acta2 expression after 72 h is significantly upregulated ~ 3-fold (Fig. 6), in contrast to the data in SSc samples. However, it may contribute to the reduction in Acta2 expression observed at 24 h, prior to perhaps the effect from ever increasing levels in profibrotic genes such as Ctgf throughout the rest of culture. These effects observed may be an interplay between a number of these factors.

The last downstream target of the Hippo pathway we studied highlights increases in Axl. The unique transmembrane receptor (part of the TAM family of tyrosine kinase receptors) has been shown across a number of organs to alter matrix synthesis and the pathogenesis of fibrosis⁷⁴. Axl has been implicated in fibrosis within various organs including renal fibrosis⁷⁵ as well as in the liver⁷⁶. Moreover, inhibitors to the receptor has been shown to significantly inhibit the fibrotic properties of IPF (Idiopathic Pulmonary Fibrosis) fibroblasts⁷⁷.

Overall, these results suggest increased YAP nuclear localisation and signalling activity, within the ovary after treatment with androgens. This has implications in the matrix environment within the ovaries of PCOS women, particularly due to the fibrotic nature of the pathway and its downstream targets. Moreover, if increased activity is linked with increased granulosa cell proliferation, then androgen regulation of YAP activity potentially could explain increased proportion of early follicular growth observed in women with PCOS. Ultimately, this disruption may then lead to the aberrant follicle development observed within the disorder.

Lastly, the Tgf-β pathways are known mediators of fibrosis. Our results indicate increases in both Tgfb1 and Tgfb3. Tgfb1 is a well characterised gene linked with various fibrotic pathologies⁷⁸. Its role in the control of matrix expression is shown through its ability to induce the expression of both collagens and fibronectin expression^79,80. The role of Tgfb3 in the progression of fibrotic matrix deposition is less clear. A number of studies indicate its function in stimulating a non-fibrotic matrix within human corneal fibroblasts⁸¹. Additionally, Tgfb3 was associated with the production of normal matrix when overexpressed within rat lung tissue, highlighting a healthy wound response, in comparison to the fibrotic deposition associated with Tgfb1 overexpression⁸². In contrast, within in vitro studies of liver tissue, Tgfb3 is indicated to promote fibrosis⁸³. It is likely however, that there are different responses within various tissues. Despite this, what is clear is the role both components have on matrix deposition and remodelling as a process.

Conclusion

The results shown here demonstrate that androgen signalling has a profound impact on gene expression within ovarian follicles. Collectively these genes play a role in several processes associated with fibrosis as well as mechanotransduction, the actin cytoskeleton and components of the basal lamina. Taken together these results indicate that androgen signalling can contribute to matrix remodelling and a fibrotic phenotype within follicles and the ovarian environment. More broadly, the altered expression especially of matrix structural proteins and matrix modifiers will impact the mechanical and biochemical microenvironment of the ovary. Aberrant control of the environment has implications in terms of follicle development as well as the overall mechanical properties of the ovary itself, particularly in reference to fibrotic disorders. Furthermore, these results work towards defining the unique ovarian phenotype seen in women with PCOS, providing a potential explanation for the disordered follicle development associated with the disorder.

Methods and materials

Mouse ovary collection

All mice were housed in accordance with the Animals (Scientific Procedures) Act of 1986 and associated Codes of Practice. All procedures were performed in accordance with the establishment licence number (X32FDCFC1). Animals for this project were sourced from either Charles River (Harlow, UK) or Envigo (Huntingdon, UK). Days 16 and 22–29 post-partum, C57BL/6 mice, were sacrificed by cervical dislocation. Samples were collected in preheated (37 °C) Leibovitz’s LM15 medium (#11415-049; Gibco), supplemented with 1% (w/v) bovine serum albumin (BSA, #A7030, Sigma-Aldrich, Gillingham, UK). Under a dissection microscope, with a heated stage (37 °C), the samples were isolated under L-15 medium and removed of extraneous tissue (fat, oviduct, bursa membrane), using insulin needles (#U-100, Terumo). Samples were then washed with phosphate-buffered saline (PBS) and fixed in 10% neutral buffered formalin solution (#HT5012, Sigma-Aldrich, UK) for immunohistochemistry, or processed for follicle culture described below. This study is reported in accordance with ARRIVE guidelines.

Follicle treatment cultures

For follicle culture experiments, preantral follicles (with 2–3 layers of granulosa cells / ~ 70–125 μm) were isolated mechanically, using insulin needles, from D15-16 mouse ovaries in L-15 medium supplemented with 1% (w/v) BSA. Follicles were assessed for quality, with only samples containing an intact basal lamina and centrally located oocyte were chosen. Isolated preantral follicles were collected from D16/17 mice and individually transferred to single wells of a 96-well plate (#167008, Thermo Scientific). Each well contained 100 µl of MEM-α medium supplemented with 0.1% (w/v) BSA, 100 µg/mL streptomycin sulphate (#S6501, Sigma-Aldrich), 75 µg/mL penicillin (# PENK, Sigma-Aldrich), 5 µg/mL insulin, 5 µg/mL transferrin and 5 ng/mL sodium selenite (ITS, #I1884, Sigma-Aldrich). Culture medium was supplemented with DHT (10nM) or an EtOH vehicle control. The cultures were maintained in humid conditions inside an incubator in 5% CO₂ at 37 °C for either 24–72 h. These samples were then imaged at the beginning of culture and every consecutive 24 h using a Nikon digital camera DXM 1200 attached to a Nikon Eclipse TE300 light microscope. The samples were assessed for morphology and were excluded from analysis according to the set criteria: (i) the oocyte had extruded from the follicle, (ii) The oocyte was displaying signs of atresia (darkness), not circular / misshapen (iii) GCs had darkened (become atretic) (iv) The oocyte was not centrally located. Follicle area was measured using Fiji software. Samples were blinded to treatment to ensure no bias. At the end of culture, (24, 72 h) samples were then processed for RNA extraction by snap freezing in dry ice in 10 µl of PBS and storing at -80 °C.

Mouse tube follicle culture

To provide larger amounts of template RNA required for experiments looking at signalling pathways, isolated preantral follicles were collected from D16/17 mice and individually transferred to single 1.5 mL Eppendorf tubes. Each Eppendorf contained 100 µl of MEM-α medium supplemented with 0.1% (w/v) BSA, 100 µg/mL streptomycin sulphate (#S6501, Sigma-Aldrich), 75 µg/mL penicillin (# PENK, Sigma-Aldrich), 5 µg/mL insulin, 5 µg/mL transferrin and 5 ng/mL sodium selenite (ITS, #I1884, Sigma-Aldrich). Culture medium was supplemented with either DHT (10nM) or an EtOH vehicle control). 10 µl of silicone oil (#378356, Sigma-Aldrich) was added atop the medium within each tube to prevent evaporation. The cultures were maintained in humid conditions inside an incubator in 5% CO₂ at 37 °C for 72 h. Lids were left open on the Eppendorfs to allow for adequate perfusion of O₂ and CO₂. At the end of the culture the silicone oil was removed, with the lids sealed and then snap frozen in dry ice in preparation for RNA extraction.

Granulosa cell isolation

Granulosa cells (GCs) were isolated from large preantral follicles in LM15 medium supplemented with 1% (w/v) BSA. These follicles were manually punctured using insulin needles and gently pressed, causing the release of GCs. These were then immediately snap frozen in dry ice for RNA extraction before being stored at -80 °C.

Oocyte isolation

Oocytes were collected from D16 to D22 mouse ovaries during the follicle isolation procedure all in LM15 medium supplemented with 1% (w/v) BSA. Oocytes were then removed of extraneous granulosa cells by pipetting up and down. Samples were imaged immediately then snap frozen within PBS in dry ice for RNA extraction before being stored at -80 °C.

Sized follicle samples

Follicles that were used in experiments focusing on follicle sizing (Supplemental Fig. 3) were isolated from D16 to D22 mice in LM15 medium supplemented with 1% (w/v) BSA using insulin needles. Using the dissection microscope samples were approximately grouped according to size by eye. Follicles grouped together as one sample were then photographed using a Nikon Eclipse TE300 light microscope. After being photographed, follicles were snap frozen within PBS, in dry ice for RNA extraction before being stored at -80 °C. The photographs were measured for size using Fiji⁸⁴.

RNA extraction procedure

Total RNA was extracted from follicles, granulosa cells and oocytes using the RNeasyMicro Kit (#74004, Qiagen, UK), performed according to the manufacturer’s instructions. Samples that had been snap frozen and stored at -80 °C were thawed gently over ice. 4 µg/µl of Carrier RNA was added to each sample. An on-column DNase digestion was performed using a DNase digestion mix, in order to remove genomic DNA (#79254, Qiagen). Elution of the RNA was performed by adding 14 µl of RNase-free water directly to the column. RNA was then either processed immediately for cDNA synthesis or stored at -80 °C.

Measurement of RNA integrity and concentration

The quality and quantity of RNA extracted was calculated using the Agilent Technologies TapeStation, under manufacturers guidelines (Agilent TapeStation 2200, Agilent Technologies, US). Briefly, samples were combined with high sensitivity sample buffer (#5067–5580, Agilent Technologies) using the high sensitivity RNA ScreenTape (5067–5579, Agilent Technologies). Sample analysis of concentration and RNA integrity (determined RIN values) was produced using the Agilent 2200 TapeStation software (Agilent Technologies).

cDNA synthesis

Extracted RNA was converted to cDNA (Complementary DNA) using the SuperScript IV first strand synthesis kit (#18091050, Invitrogen) following the manufacturers guidelines. Briefly per reaction, the RNA template was combined with random hexamer primers, a dNTP mix and annealed to the template RNA. The reaction mixture was heated to 65 °C for 5 min and cooled on ice for 1 min. Annealed RNA was then added to the reverse transcriptase mix consisting of SSIV Buffer, DTT, RNaseOUT™ Recombinant RNase Inhibitor and SuperScript^® IV Reverse Transcriptase. The combined reaction mix was incubated at 23 °C for 10 min, 55 °C for 10 min and 80 °C for 10 min using the TC3000 PCR machine (Techne, UK).

Real-time quantitative polymerase chain reaction (RT-qPCR)

RT-qPCR was performed using a SYBR green detection system used to measure transcript concentrations. To quantify the gene expression the 2^−ΔΔCT method was used. A reaction mixture was made consisting of nuclease free H₂O, primers and Power SYBR™ Green PCR Master Mix (#4368577, Applied Biosystems). To each of the wells within a 384 well plate (#4309849, Applied Biosystems) the reaction mix was added to template cDNA or nuclease free H₂O (negative control). Plates were then sealed using adhesive PCR plate seals (#4311971, Applied Biosystems) followed by centrifugation of the plate at 1500 rpm for 2 min. RT-qPCR was performed on an Applied Biosystems 7900HT Fast machine. Cycling parameters included a 10-minute polymerase activation step at 95 °C followed by 40 cycles of PCR (Denature at 95 °C for 15 s and an extension at 60 °C for 1 min). Each sample was then subjected to a melt curve analysis to confirm correct product specific amplification. The internal reference genes used were Gapdh (Primer Design) and Atp5b (Primer Design) and a geomean calculated. Expression levels were normalised to the geomean of the internal reference genes and calculated as fold change relative to experimental control using the 2^−ΔΔCT method (Livak and Schmittgen⁸⁵. ΔCt values were used for statistical calculation. To ensure correct primer amplicon sizes, PCR products were separated by gel electrophoresis. Primers were analysed for amplification efficiency using LinRegPCR software^86,87. Efficiency of between 87 and 110% was considered acceptable.

Primer sequences.

GeneSym	RefSeq	Forward Sequence	Reverse Sequence	Product Length (bp)	GeneID
Acta2	NM_007392	CATCTTTCATTGGGATGGAG	TTAGCATAGAGATCCTTCCTG	96	11475
Cav1	NM_007616	CTCAGTTCTCTTAAATCACAGC	CATACACTTGCTTCTCAGTC	191	12389
Col4a1	NM_009931	TTCTCTTCTGCAACATCAAC	GAATCTGAATGGTCTGACTG	198	12826
Col11a1	NM_007729	AGAGGCAAATATTGTGGATG	TACAGAATATCCTCGGAAGTG	185	12814
Ctgf	NM_010217	GAGGAAAACATTAAGAAGGGC	AGAAAGCTCAAACTTGACAG	74	14219
Eln	NM_007925	TTCTCCCATTTATCCAGGTG	GAAGATCACTTTCTCTTCCG	146	13717
Fbln7	NM_024237	ACACGGTTAACAAAATGACC	AGTACTTGCTTCCAAACTTC	92	70370
Fbn1	NM_007993	GTGAAGATATTGACGAGTGC	GACATTTGCAGAAGTAGCTG	89	14118
Fn1	NM_010233	CCTATAGGATTGGAGACACG	GTTGGTAAATAGCTGTTCGG	167	14268
Frmd6	NM_028127	CAAGATGAGGAAATCGAGATG	GTAGATACACATCTCTGGGC	95	319710
Has2	NM_008216	GATTATGTACAGGTGTGTGAC	CCTCTAAGACCTTCACCATC	75	15117
Hpse	NM_152803	AGCCTTTACCTGATTACTGG	GTGACATTATGGAGGTTCAG	186	15442
Itga9	NM_001113514	AACATTACTCTCCAGGTCTAC	CTTTGTAGAGAGCAGTTACC	157	104099
Lama1	NM_008480	ATTTAGCCAATGGAAAGTGG	TTTTCTTACAAAGACACGGC	176	16772
Lama3	NM_010680	GGATACAGACAATAGCTACAC	TTTTCAATTTGTCTGGGACG	97	16774
Lats1	NM_010690	TGGATTTCAGTAACGAATGG	TGTATATCCTGTTCGCAGTAG	166	16798
Lats2	NM_153382	AAAAAGCTCTCAGGGAAATC	AATATGCATCTGGTCACATC	130	50523
Lox	NM_010728	CACCGTATTAGAAAGAAGCC	GTCCTTCCTACTTAAGCTAATC	93	16948
Mmp19	NM_001164197	TGTTTAAGGGCTCAGGATAC	GTTCCTTGATTGGTTTAGGG	80	58223
Palld	NM_001081390	CAAGCTCAGAAGAAAACAAC	TCGGCAGAGATATGAAGTAG	137	72333
Plod2	NM_001142916	AGCCCTTCAACTCTTTATTC	ATCACACTTTTCATCCTGAC	180	26432
Rhou	NM_133955	TCAGCTACACCACTAACG	AAACTCATCCTGTCCTGC	130	69581
Ssh1	NM_198109	GTGACTTCTAAGGAAATCCG	AAAGATGGTCAAAGATGAGG	138	231637
Stk3	NM_019635	GGAAGAAAACTCGGATGAAG	AACATGGTGCTGTTATGTTC	137	56274
Stk4	NM_021420	GAAGGAACCATGAAAAGAAGAG	GAGAGCCAGATACATTCTTG	123	58231
Tead1	NM_001166585	AGTAGAAAAAGTAGAGACGGAG	TATATTCACACATTGGCGAG	85	21676
Vcl	NM_009502	ACTGTAGAAAAGATGAGTGC	CTGGTTCAATTTGGAGTCTATG	140	22330
Wwc1	NM_170779	CGCTGAAGAAAATTGATGAG	GGTCTTGTTTCTCCTTTTCTC	131	211652
Wwtr1	NM_001168281	GGATACAGGTGAAAATTCCG	GATTACAGCCAGGTTAGAAAG	194	97064
Yap1	NM_001171147	CAATACGGAATATCAATCCCAG	GATCGGAACTATTGGTTGTC	164	22601
Axl	NM_001190974	GTGAAGACCATGAAAATTGC	TCAAATTCCTTCATGCAGAC	82	26362
Cyr61	NM_010516	AGAGGCTTCCTGTCTTTG	GTTGTCATTGGTAACTCGTG	148	16007
Gli2	NM_001081125	AGACTATTACCACCAGATGAC	TGGACTAGAGAATCGTGATG	141	14633
Tgfb1	NM_011577	GGATACCAACTATTGCTTCAG	TGTCCAGGCTCCAAATATAG	160	21803
Tgfb2	NM_009367	GAGATTTGCAGGTATTGATGG	CAACAACATTAGCAGGAGATG	114	21808
Tgfb3	NM_009368	CTCAGTGGAGAAAAATGGAAC	GGTCGAAGTATCTGGAAGAG	116	21809

Tissue fixation and processing

Whole ovaries were fixed in 10% neutral buffered formalin solution (#HT5012, Sigma-Aldrich, UK) for 3 h and then transferred to 70% ethanol. Ovaries were then processed for paraffin embedding under standard protocols. Tissue was processed through increased gradients of ethanol, 70% ethanol for 1 h, 90% ethanol for 1 h and 100% ethanol for 3 × 1 h, then placed in Histoclear (# HSM200, National Diagnostics) overnight. The following day samples were incubated for 1 h in fresh Histoclear. Samples were then transferred to clear moulds (#03015, Surgipath) and embedded in paraffin wax (#36107, VWR) for 2 h at 65 °C. Cassettes (#M480-2, Simport) were then placed over the top of the moulds and blocks were allowed to cool. The blocks (containing tissue) were then serially sectioned (5 μm) (Leica RM 2135 microtome) and mounted onto SuperFrost glass slides (#631 − 0108, VWR). Slides were left to dry overnight at 37 °C.

Immunofluorescence

Chosen sectioned tissue were dewaxed in Histoclear (2 × 5 min) and then placed through decreasing concentrations of ethanol (5 min each, 100% 95%, 70%), then placed in distilled H₂O for 5 min. An antigen retrieval step was performed using a citrate buffer (pH 6.0). Samples were then washed in PBS (2 × 5 min). Dependent on the secondary antibody species it was raised in, the matching serum was used for blocking, normal goat serum (#120316, SAFC Biosciences) or nomal donkey serum (#D9663, Sigma-Aldrich). To prevent non-specific binding the samples were blocked by incubation with 5 or 10% (v/v) (NDS (10%), NGS (5%)) species specific serum supplemented with 4% (w/v) BSA (#A7030, Sigma-Aldrich) in PBS for 30 min. Slides were then incubated overnight at 4 °C with the primary antibody; brackets display (Manufacturer, Product Code; Concentration used, RRID of antibody obtained from https://www.antibodyregistry.org/). LOX (Abcam, #Ab174316; 1:200, AB_2630343), CTGF (Abcam, #Ab6992; 1:100, AB_305688), VCL (Abcam, #Ab129002; 1:50, AB_11144129), Collagen IV (Millipore, #AB756P; 1:100, AB_2276457), Laminin (Abcam, #Ab11575; 1:100, AB_298179), Fibronectin (Abcam, #Ab2413; 1:100, AB_2262874), ACTA2 (Sigma-Aldrich, #A2547; 1:100, AB_476701) or non-immune immunoglobulin (IgG) control isotype (Sigma-Aldrich, #I814), diluted in 1% or 5% (v/v) serum (1% if blocked with NDS / 5% if blocked with NGS) (demonstrated in Supplemental Fig. 7). After the overnight incubation, slides were washed using PBS for 3 × 10 min. Slides were then incubated with the resulting AlexFluor conjugated secondary antibody (AlexaFluor 488, Invitrogen #A-11008; AlexaFluor 555 Invitrogen #A31572) diluted in PBS (1:200) for 60 min at room temperature. Slides were then washed for 2 × 5 min in PBS before then being counterstained with 1 µg/ml 4’,6 M diamidinoM2Mphenylindole (DAPI, #D8417, Sigma-Aldrich) for 3 min. Sections were then mounted with coverslips using Prolong Gold Antifade reagent containing DAPI (#P36931, Invitrogen) and left to cure for 24 h at room temperature in the dark. Coverslips were then sealed with nail varnish and stored at 4 °C in the dark. Fluorescently stained slides were imaged using a confocal laser-scanning microscope Leica inverted SP5 confocal laser-scanning microscope (Leica Microsystems, Wetzlar, Germany).

Statistical analysis

All statistical analyses were performed using GraphPad Prism (V 8.0) (Graphpad Inc, San Diego, CA). Data were assessed for normal distribution using a D’Agostino & Pearson normality test. Data not normally distributed were analysed by non-parametric Kruskal-Wallis or Mann-Whitney test. Normally distributed data were analysed by an unpaired t test. Statistical tests are highlighted in the corresponding figure legends. In all cases significance is determined by a p-value less than 0.05.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

KH, SF and IED conceived the study; KH, SF and IED supervised the study; TH and AL carried out experiments; TH analysed data; TH, SF, KH and IED wrote and revised the manuscript.

Data availability

Data is available upon request to the corresponding author.

Declarations

Competing interests

The authors declare no competing interests.