Primary aldosteronism (PA) is the main cause of secondary hypertension, resulting from adrenal aldosterone-producing adenomas (APA) or bilateral hyperplasia. Here, we show that constitutive activation of WNT/β-catenin signalling is the most frequent molecular alteration found in 70% of APA. We provide evidence that decreased expression of the WNT inhibitor SFRP2 may be contributing to deregulated WNT signalling and APA development in patients. This is supported by the demonstration that mice with genetic ablation of Sfrp2 have increased aldosterone production and ectopic differentiation of zona glomerulosa cells. We further show that β-catenin plays an essential role in the control of basal and Angiotensin II-induced aldosterone secretion, by activating AT1R, CYP21 and CYP11B2 transcription. This relies on both LEF/TCF-dependent activation of AT1R and CYP21 regulatory regions and indirect activation of CYP21 and CYP11B2 promoters, through increased expression of the nuclear receptors NURR1 and NUR77. Altogether, these data show that aberrant WNT/β-catenin activation is associated with APA development and suggest that WNT pathway may be a good therapeutic target in PA.
Adrenal autonomous secretion of aldosterone also called primary aldosteronism (PA) is the main cause of secondary hypertension with a prevalence of 6–13% of hypertensive patients (1,2). In 95% of the cases, this pathology is caused by adrenal aldosterone-producing adenomas (APA) or bilateral hyperplasia (3). Clinical management of PA essentially relies on surgical resection and/or treatment with mineralocorticoids antagonists to minimize hypertension (4,5). However, complete normalization of blood pressure is rarely achieved and mineralocorticoids antagonists are associated with adverse endocrine side effects (6). Although new promising treatments such as aldosterone synthase inhibitors are emerging (7,8), identification of the molecular aetiology of PA is essential to allow the development of more specific and more efficient drugs. Multiple molecular alterations such as down-regulation of TASK channels (9,10) or circadian clock deregulation (11) have been proposed to play a role in the establishment of PA but their clinical relevance is still unclear. Recent data suggest that the development of a subset of APA is associated with mutations in the KCNJ5 gene, which cause dysfunction of the Kir3.4 potassium channel, or mutations in the genes encoding the ATPases ATP1A1 and ATP2B3 (12–16). In both cases, these alterations result in inappropriate membrane depolarization, which increases steroidogenic capacities of adrenocortical cells. Even though KCNJ5 and ATPases mutations account for ∼40–45% of patients with sporadic PA, the molecular mechanisms underlying the majority of APA still remain elusive (13,15,16,17,18).
We have recently shown that transgenic mice with specific constitutive activation of β-catenin in the adrenal cortex presented with increased aldosterone production (19). Here, we have investigated the hypothesis that aberrant WNT pathway activation may play a role in the development of APA in patients.
WNT/β-catenin pathway is constitutively activated in APA
To evaluate the potential role of WNT pathway in the development of APA, we analysed the status of WNT/β-catenin pathway activation in tissue samples from 47 patients diagnosed with PA caused by adrenal APA. Patients' clinical and biological characteristics are described in Supplementary Material, Table S1A. Immunohistochemistry for β-catenin showed cytoplasmic (Fig. 1Ac and d) and/or nuclear (Fig. 1Ae and f) accumulation of β-catenin in 33/47 APA (Supplementary Material, Table S1B). This suggested that WNT/β-catenin signalling was activated in 70% of APA in our series (Fig. 1A and Supplementary Material, Table S1B). To confirm these observations, WNT/β-catenin pathway target genes expression was analysed by reverse-transcription quantitative polymerase chain reaction (RTqPCR) in 35 out of the 47 APA that were analysed by immunohistochemistry (Supplementary Material, Table S1B). In these experiments, APA were compared with 7 normal adrenals (NA) and 17 non-aldosterone-producing adenomas (NAPA) that were selected on the basis of cortisol or androgen secretion and absence of histological signs of β-catenin activation. These experiments demonstrated a significant overexpression of AXIN2 (Fig. 1B) and LEF1 (Fig. 1C) in APA compared with both NA and NAPA. We then analysed expression of CYP11B2 (Fig. 1D) and performed correlation analyses with AXIN2 and LEF1 expression levels. These experiments showed a strong association between AXIN2 (Fig. 1E), LEF1 (Fig. 1F) and CYP11B2 expression levels. To further strengthen our hypothesis of an association between WNT activation and CYP11B2 overexpression, we analysed expression of CYP11B2 (by in situ hybridization) and β-catenin (by immunohistochemistry) on consecutive sections of APA. These data showed that cells within the adenoma displayed both high levels of CYP11B2 expression (Fig. 1Gb and c) and nuclear β-catenin staining (Fig. 1Gd and e). We therefore concluded that canonical WNT/β-catenin signalling was activated in a majority of APA and that this activation was correlated with increased CYP11B2 expression.
WNT/β-catenin pathway activation in APA is associated with down-regulation of SFRP2
WNT/β-catenin pathway activation in adrenal tumours has mostly been attributed to activating mutations in the CTNNB1 gene (20). However, there were no such mutations in our series of 26 β-catenin-positive APA for which genomic DNA was available (Supplementary Material, Table S1B). This result was in agreement with a recent report showing no activating CTNNB1 mutations in 41 APA (21). We then evaluated the possibility of an association between WNT pathway activation and KCNJ5 mutations. Out of the 19 β-catenin-positive tumours that were available for sequencing, 9 were mutant for KCNJ5 (47%). However, 6/9 (67%) of the β-catenin-negative tumours were also mutant for KCNJ5 (Supplementary Material, Table S1B). This suggested that there was no association between KCNJ5 mutations and canonical WNT/β-catenin pathway activation in APA. In an attempt to identify WNT regulators that may be relevant to APA development, we compared expression levels of known WNT regulators (based on a compilation of data available from the Wnt Homepage and PubMed browsing) in six APA and four NA by retrospective analysis of cDNA microarray data ((22), Supplementary Material, Fig. S1). This analysis confirmed significant up-regulation of AXIN2 and LEF1 expression in APA, but surprisingly failed to identify up-regulation of positive regulators of WNT pathway activity, with the exception of Lrp6 WNT coreceptor (Supplementary Material, Fig. S1). However, this approach showed significant down-regulation of SFRP1, 2 and 4, three members of the SFRP (secreted frizzled related protein) family of WNT signalling inhibitors, in APA. These factors have the ability to bind and titrate WNT ligands and their down-regulation has been shown to stimulate WNT pathway activation in tumours (23). To further confirm our transcriptome data, we analysed expression of SFRP1, SFRP2 and SFRP4 by RTqPCR in our series of 35 APA in comparison with NAPA and NA. We reasoned that factors relevant to WNT pathway activation and aldosterone production should be deregulated in APA in comparison with both NA and NAPA. In these experiments, SFRP1 was down-regulated in APA and NAPA compared with NA (Fig. 2A). However, there was no significant difference in SFRP1 expression in APA compared with NAPA, suggesting that it was not involved in the establishment of the aldosterone-producing phenotype. In contrast with the microarray data, RTqPCR analyses of our larger series did not show down-regulation of SFRP4 in APA compared with NA and even showed up-regulation in comparison with NAPA (Fig. 2B). Interestingly, SFRP2 expression was significantly downregulated in APA compared with both NA and NAPA (Fig. 2C). This suggested that it could be involved in the establishment of deregulated WNT signalling and subsequent aldosteronism in APA patients. Consistent with this idea, SFRP2 expression was inversely correlated with both AXIN2 (Supplementary Material, Fig. S2A) and CYP11B2 expression (Fig. 2D).
Sfrp2 knockout mice have increased Wnt signalling and aldosterone production
To further explore a link between SFRP2 down-regulation and APA development, we analysed plasma aldosterone concentrations in Sfrp2 knockout mice. These mice are characterized by brachy-syndactyly, but their adrenal function was never investigated (24). Individual increases in plasma aldosterone were observed as early as 7 months in homozygous mutant females; however, significant overall increases were only observed in 12- to –13-month-old mutants compared with wild-type littermates (Fig. 2E). These were independent of renin, which was not significantly elevated in Sfrp2 knockout mice (3.77 ± 2.88 mUI/l) compared to wild-type mice (2.38 ± 1.29 mUI/l). The hormonal phenotype was specific of aldosterone, as corticosterone levels were not significantly altered by Sfrp2 knockout (Supplementary Material, Fig. S2B). Consistent with a role of Sfrp2 in restraining Wnt signalling in the adrenal, mRNA accumulation of Axin2 (Fig. 2F) and Lef1 (Fig. 2G) was increased in Sfrp2−/− adrenals. Interestingly, plasma aldosterone concentration was significantly correlated with Axin2 mRNA accumulation (Spearman r = 0.4834, P = 0.0167). We further confirmed these observations by immunohistochemical analysis of Dab2 (a molecular marker of zona-glomerulosa differentiation (25)), β-catenin and Cyp11b2 expression. Consistent with the endocrine and molecular phenotypes, immunohistological analysis of Sfrp2 knockout adrenals showed ectopic Dab2 (Fig. 2Hb versus a) and β-catenin-positive cells (Fig. 2He versus d) within the adrenal cortex and central adrenal region. These ectopic cells were also positive for Cyp11b2 (Fig. 2Hh versus g; Supplementary Material, Fig. S2C), which was reminiscent of the phenotype of ΔCat mice with adrenal-specific constitutive β-catenin activation (Fig. 2Hc, f and i (19)). This showed that Sfrp2 knockout was sufficient to increase adrenal WNT signalling and aldosterone secretion. We thus concluded that down-regulation of SFRP2 expression was one of the possible causes for deregulation of WNT signalling and development of APA in patients.
β-Catenin controls aldosterone production at multiple levels
In order to understand the role of β-catenin in aldosterone production, we used the human adrenocortical cell line H295R that has the ability to produce aldosterone in response to its natural secretagogue, Angiotensin II. However, it spontaneously expresses a constitutive active mutant form of β-catenin (S45A mutation) (26). We therefore decided to down-regulate β-catenin activity and to assess basal and Angiotensin II-induced aldosterone production. Transfection of H295R cells with small interfering RNAs to β-catenin induced a mean 60% decrease in β-catenin expression compared with unrelated GFP siRNA (Supplementary Material, Fig. S3A and B). As expected, this resulted in a significant down-regulation of AXIN2 (Supplementary Material, Fig. S3C) and LEF1 (Supplementary Material, Fig. S3D) target gene expression. Interestingly, down-regulation of β-catenin also resulted in a marked decrease in basal aldosterone output and completely abrogated response to Angiotensin II (Fig. 3A). In order to confirm these observations, we treated H295R cells with PKF115-584, a previously published inhibitor of β-catenin transcriptional activity (27). As expected, treatment with PKF115-584 induced a dose-dependent decrease in the expression of AXIN2 (Supplementary Material, Fig. S3E) and LEF1 (Supplementary Material, Fig. S3F). Consistent with our siRNA data, this treatment also induced a dose-dependent decrease in basal aldosterone production and markedly reduced Angiotensin II responsiveness at the highest dose (Fig. 3B). We thus concluded that β-catenin was required for basal aldosterone production and that it was essential to mediate Angiotensin II response in H295R cells. To identify relevant β-catenin target genes involved in the control of aldosterone production, we analysed expression of all the genes coding proteins involved in aldosterone synthesis and of the Angiotensin II Type I receptor AT1R, by RTqPCR (Fig. 3C). Treatment with β-catenin siRNAs resulted in a decrease in Angiotensin II-induced mRNA accumulation of CYP11B2 (Fig. 3D) and CYP21 (Fig. 3E) and inhibition of basal expression of AT1R (Fig. 3F). Similar results were obtained in cells treated with PKF115-584 (Fig. 3G–I). In contrast, β-catenin siRNAs had no significant effect on the expression of StAR, CYP11A1, HSD3B2 and CYP11B1 (Supplementary Material, Fig. S3G–J).
To further support these observations, we decided to perform experiments in which β-catenin activity was stimulated using BIO, a known inhibitor of GSK3β. Unfortunately, the presence of a constitutive active allele of β-catenin prevented further activation of WNT signalling in H295R cells (data not shown). We thus made use of mouse Y1 adrenocortical cells that express a wild-type version of β-catenin but do not produce aldosterone. As expected, treatment with BIO induced a time-dependent increase in Axin2 transcription (Supplementary Material, Fig. S3K andL), β-catenin accumulation (Supplementary Material, Fig. S3M) and TOP-FLASH activity in transient transfection experiments (Supplementary Material, Fig. S3N). This was associated with a significant increase in Cyp11b2 and Cyp21 expression after 3 and 6 h of treatment (Fig. 3J and K). Increased expression of Cyp11b2 (Fig. 3L) and Cyp21 (Fig. 3M) was also observed in the adrenals of 10-month-old ΔCat mice, our model of constitutive β-catenin activation (19). Although we could not detect At1r expression in Y1 cells even in the presence of BIO, we observed overexpression of both murine isoforms of At1r (At1a and At1b) in ΔCat adrenals (Fig. 3N). Altogether, these data showed that β-catenin controlled aldosterone production by stimulating the expression of three actors that play an essential role in its synthesis.
β-Catenin controls CYP11B2 expression indirectly
CYP11B2 is the terminal and one of the rate-limiting enzymes for aldosterone synthesis. β-Catenin mostly acts as a transcription coactivator by interacting with transcription factors such as LEF/TCF. We thus evaluated the possibility that β-catenin-controlled CYP11B2 expression at the transcriptional level. For this, Y1 cells were treated with BIO in the presence or absence of actinomycin D, an inhibitor of transcription. RTqPCR analysis showed that induction of Cyp11b2 expression by BIO was completely abrogated by actinomycin D, suggesting that β-catenin stimulated the transcription of Cyp11b2 (Fig. 4A). Computer-assisted analysis of human CYP11B2 regulatory regions identified two putative LEF/TCF-binding sites at −768/−774 and −1266/−1273 (Fig. 4B; Supplementary Material, Fig. S4). To assess the role of these putative sites in CYP11B2 promoter activity, we transfected wild-type or mutated constructs of CYP11B2 regulatory regions in H295R cells (Fig. 4B). As previously published (28), the wild-type promoter drove luciferase expression in H295R cells and strongly responded to Angiotensin II stimulation (Fig. 4C, WT). Surprisingly, mutation of both putative LEF1-binding sites (Fig. 4B) had no effect on basal promoter activity or on Angiotensin II responsiveness (Fig. 4C, Lef mut), suggesting that β-catenin was not acting on CYP11B2 promoter regions through interaction with LEF/TCF-binding sites. The nuclear receptors NURR1 (NR4A2) and NUR77 (NR4A1) are essential regulators of CYP11B2 expression in the adrenal (28). We thus evaluated the effect of the three previously published NUR-binding sites on CYP11B2 promoter activity and on its ability to respond to β-catenin. Treatment of H295R cells with the WNT antagonist PKF115-584 inhibited both basal and Angiotensin II-induced CYP11B2 promoter activity, confirming that the effect of β-catenin on CYP11B2 expression was essentially transcriptional (Fig. 4D, WT). Mutation of the three NUR-binding sites (NBRE, Ad1, Ad5, Fig. 4B) led to decreased but still measurable levels of promoter activity (Fig. 4D, NBRE/Ad mut). This mutation abrogated promoter response to Angiotensin II and interestingly, it also prevented PKF115-584-mediated repression of promoter activity (Fig. 4D, NBRE/Ad mut). This suggested that the effect of β-catenin on CYP11B2 regulatory regions was mediated by nuclear receptors of the NUR family.
We initially hypothesized that β-catenin could work as a coactivator of NUR transcriptional activity. However, our repeated transfection experiments failed to provide support for such a theory (data not shown). Therefore, we hypothesized a role of β-catenin in controlling expression of NURs. Indeed, β-catenin knockdown resulted in a significant decrease in NURR1 (Fig. 5A) and NUR77 (Fig. 5B) basal and Angiotensin II-induced expression at the mRNA (Fig. 5A and B) and protein level (Fig. 5A and B; Supplementary Material, Fig. S3O and P) in H295R cells. This decrease was also observed in response to PKF115-584 treatment (Supplementary Material, Fig. S3Q and R). Conversely, expression of both nuclear receptors was significantly induced by BIO in Y1 cells (Fig. 5C and D). In agreement with these in vitro data, mRNAs for Nurr1 (Fig. 5E) and Nur77 (Fig. 5F) were markedly accumulated in 10-month-old ΔCat adrenals. This was further supported by increased expression of NURR1 in APA compared with NAPA (Fig. 5G). To further analyse the mechanism involving β-catenin in NUR induction, we treated Y1 cells with both BIO and actinomycin D. Actinomycin D abrogated induction of Nurr1 and Nur77 mRNA accumulation by BIO (Fig. 5I and J), suggesting that β-catenin acted by stimulating transcription from Nurr1 and Nur77 promoters. Sequence analysis of NURR1 regulatory regions identified two putative Lef/Tcf-binding sites at −57/−61 and +67/+71 that were conserved in mouse and human (Supplementary Material, Fig. S5). Such binding sites were not identified in NUR77 regulatory regions. However, a recent publication showed that β-catenin could stimulate human NUR77 expression in vitro by associating with c-Jun/c-Fos proteins at four AP1-binding sites (−3/+4, −26/−32, −169/−175 and −189/−195) that were conserved in mouse and human (29) (Supplementary Material, Fig. S6). To confirm binding of β-catenin to these regulatory regions, we performed chromatin immunoprecipitation (ChIP) experiments in H295R cells. β-Catenin bound to both NURR1 (through LEF/TCF sites) and NUR77 (through AP1 sites) promoters in untreated cells (Fig. 5K and L). Consistent with our expression analysis data, binding of β-catenin was markedly diminished after incubation with PKF115-584 (Fig. 5K and L). To further confirm these observations, we performed similar ChIP experiments with whole adrenal glands from wild-type or ΔCat mice that overexpress both Nurr1 and Nur77 (Fig. 5M and N). Binding of β-catenin to Lef/Tcf or Ap1 sites was almost undetectable in wild-type adrenals. However, it was markedly increased in ΔCat adrenals, in which β-catenin was constitutively activated (Fig. 5M and N). Altogether, these experiments showed that β-catenin could control expression of both nuclear receptors in vitro and in vivo. It further suggested that the effect of β-catenin on CYP11B2 expression could be mediated by NURR1 and NUR77. To confirm this hypothesis, NUR binding to CYP11B2 promoter regions was analysed by in vitro and in vivo ChIP experiments. In H295R cells, NURR1 and NUR77 bound to the proximal Ad1 and Ad5 sequences (Fig. 5O, top panel) and to a lesser extent to the distal NBRE sequence (Fig. 5O, bottom panel). In both cases, NUR binding was inhibited by the Wnt inhibitor PKF115-584 (Fig. 5O). Although the NUR response elements previously described in human CYP11B2 regulatory regions were not strictly conserved in mouse Cyp11b2 regulatory regions, sequence analysis identified a putative Sf-1 response element (+156/+162, also known to bind Nur proteins) downstream of the TSS (transcription start site) and two potential nuclear receptor-binding sites (−174/−150, −233/−247) upstream of the TSS (Supplementary Material, Fig. S4). In vivo ChIP experiments showed mild binding of both Nurr1 and Nur77 to the proximal Sf-1 response element region in wild-type adrenals extracts (Fig. 5P, top panel). It also showed binding of Nur77 and to a lesser extent of Nurr1 to the distal sites (Fig. 5P, bottom panel). Binding of Nurr1 to the proximal Sf-1 response region was increased in ΔCat adrenals (Fig. 5P). There was also a mild increase in Nurr1 binding to the distal element, whereas Nur77 binding was not altered in ΔCat adrenals (Fig. 5P). Collectively, these data strongly suggested that β-catenin controlled CYP11B2 expression indirectly, through stimulation of NURR1 and NUR77 transcription.
Control of CYP21 and AT1R expression involves direct and indirect mechanisms
We then addressed the mechanisms of CYP21 regulation. As for Cyp11b2, induction by BIO in Y1 cells was abrogated by incubation with actinomycin D, suggesting that the effect of β-catenin on Cyp21 expression was transcriptional (Fig. 6A). Computer-assisted promoter analysis identified one putative LEF/TCF-binding site (−82/−86) and two putative NUR response elements (−130/−135, −62/−67) upstream of the TSS in both human and mouse CYP21 (Supplementary Material, Fig. S7). ChIP analysis in H295R cells showed binding of both β-catenin (Fig. 6B, top panel) and NURR1/NUR77 (Fig. 6B, bottom panel) to this promoter region. Binding of these three transcriptional activators was disrupted by PKF115-584 (Fig. 6B). This suggested that β-catenin could control expression of CYP21 both directly (by interacting with LEF/TCF factors) and indirectly by stimulating NUR expression. Sequence analysis of mouse Cyp21 regulatory regions identified four non-conserved distal (−629/−632, −623/−627, −608/−612, −370/−374) and three proximal putative Lef/Tcf-binding sites (−142/−146, −53/−57, −34/−30) as well as one putative Sf-1 (−280/−285) and four putative proximal Nur-binding sites (−189/−194, −122/−127, −49/−55, −4/−8) (Supplementary Material, Fig. S7). In wild-type mice, β-catenin binding was almost undetectable, whereas Nurr1 and Nur77 bound to the distal (Fig. 6C, top panel) and to a lesser extent to the proximal regions (Fig. 6C, bottom panel). Binding of β-catenin and Nurr1 to the distal and proximal regions was markedly increased in ΔCat mice compared with wild-type mice (Fig. 6C). Altogether, these experiments suggested that CYP21 was both a direct and indirect target of β-catenin in human adrenocortical cells and in mouse adrenals in vivo.
We finally addressed the regulation of AT1R expression by β-catenin. In silico analysis of AT1R regulatory regions showed one putative LEF/TCF-binding site on human AT1R promoter. This site was conserved in mouse At1a regulatory regions (Supplementary Material, Fig. S8), but not in mouse At1b promoter, which harbored one putative Ap1-binding site (Supplementary Material, Fig. S9). ChIP analysis showed binding of β-catenin in a region surrounding the putative LEF-binding site (−502/−505) on human AT1R promoter in H295R cells. As expected, binding was markedly diminished by PKF115-584 treatment (Fig. 6D). In mouse adrenals, ChIP analysis showed binding of β-catenin to the regions surrounding the putative Lef- and Ap1-binding sites on At1a and At1b promoters, respectively (Fig. 6E). This was markedly increased in ΔCat, compared with wild-type adrenals (Fig. 6E). This strongly suggested that β-catenin could directly control expression of AT1R and At1a/At1b in human and mouse adrenocortical cells, respectively.
Our study provides strong immunohistochemical and molecular evidence that WNT/β-catenin signalling is aberrantly activated in 70% of APA in a series of 47 patients, which is the most frequent molecular alteration reported in APA, to our knowledge. The observation that mice with adrenal-specific constitutive activation of β-catenin displayed increased aldosterone production suggested that aberrant WNT signalling may be driving the development of APA (19). Here, using a combination of in vitro and in vivo experiments, we show that β-catenin is essential for both basal and Angiotensin II-induced aldosterone production. We further show that this function of β-catenin relies on its transcriptional activity, which stimulates expression of three major actors of aldosterone synthesis, i.e. AT1R, CYP21 and CYP11B2, through direct and indirect mechanisms involving NURR1 and NUR77 (Fig. 7). Finally, we show that there is no obvious association between KCNJ5 mutations and activation of WNT signalling, suggesting that the two pathways act independently in APA and may promote formation of different types of tumours. Altogether, these data strongly suggest that abnormal WNT pathway activation is closely associated with the development of PA in patients.
In the adrenal gland, constitutive WNT/β-catenin signalling has already been involved in the development of NAPA, micro- and macronodular hyperplasia and adrenocortical carcinomas (reviewed in 20). In most cases, increased activation, as assessed by immunohistochemistry or gene expression array analysis, was correlated with more aggressive disease (30–33). That WNT/β-catenin signalling plays a central role in the establishment of APA, in essence a benign adrenal tumour thus seems puzzling. However, one essential aspect of WNT signalling is its high sensitivity to dosage, which is essential for A–P axis and organ patterning. This is also reflected in different dosage requirements for tumourigenesis in transgenic mouse models (34–38). In our previously published mouse model of constitutive β-catenin activation (19), although a good number of cells with accumulation of β-catenin displayed Cyp11b2 expression, cells with highest levels of accumulation were rather undifferentiated Fig. 5 in reference 19. Furthermore, the most aggressive tumours in ΔCat mice did not overproduce aldosterone (19) (Supplementary Material, Fig. S4 in reference 19). This suggests that levels of WNT dosage may also play a central role in the development of specific types of tumours within the adrenal cortex.
The mechanisms involved in WNT pathway activation in non-aldosterone secreting adrenal tumours are not completely understood. However, 36% of adrenocortical adenomas (usually the largest ones) and 16% of carcinomas (usually the most aggressive ones) display activating mutations of the CTNNB1 gene (30,31,32). In contrast, our analysis of CTNNB1 gene sequence failed to identify any activating mutations in APA, which was in line with previously published data (21). It is therefore possible that the levels of WNT pathway activation achieved in APA differ quite significantly from the levels found in other adrenal tumours. Here, using transcriptome and RTqPCR analyses, we show down-regulation of the secreted WNT inhibitor SFRP2 (39), which is significantly inversely correlated with AXIN2 and CYP11B2 expression. We also demonstrate for the first time that Sfrp2 knockout mice have significantly increased plasma aldosterone levels that are significantly correlated with increased Axin2 expression, abnormal β-catenin accumulation and marked adrenal dysplasia. Although it is quite likely that other factors play a role in WNT pathway activation in APA, our data strongly suggest that decreased SFRP2 expression contributes to the establishment of a particular dosage of WNT signalling, which is associated with the development of APA. In light of these observations, it will be interesting to evaluate expression of WNT pathway regulators in other types of adrenal tumours, in correlation with WNT signalling activity.
SFRP2 expression is decreased in a number of tumour types where it correlates with increased WNT signalling. In most tumours, down-regulation of SFRP2 expression is the result of increased promoter DNA methylation (23,40,41). Although we cannot rule out increased methylation in a subset of cells within APA tissues, our global methylation analyses using pyrosequencing do not show an overall increase in the methylation status of SFRP2 regulatory regions (Supplementary Material, Fig. S10). This suggests involvement of some other molecular mechanisms. Interestingly, micro-RNAs miR-218 (42), miR-29 (43), miR-410, miR-433 (44) and miR-224 (45) have been proposed as regulators of SFRP2 expression and WNT signalling. It is therefore tempting to speculate that overexpression of some of these micro-RNAs could account for decreased SFRP2 expression in APA.
Our observations raise the question of the role of WNT signalling in zona-glomerulosa physiology. Active β-catenin, identified by nucleo-cytoplasmic staining in immunohistochemistry is found mostly restricted to the adrenal zona glomerulosa in wild-type mice (19,46) and in healthy subjects (21), which suggests that WNT pathway is normally activated in this specific zone. Unfortunately, mice with Sf-1:Cre-mediated ablation of β-catenin exhibit adrenal aplasia in late development, which precludes analysis of postnatal adrenal function (46). Interestingly, however, Wnt4 knockout mice have impaired zona-glomerulosa differentiation and lower plasma aldosterone concentration at birth. This effect is zone specific as Cyp21a1 expression and corticosterone levels are not altered (47). Conversely, overexpression of WNT4 in human adrenocortical cell lines increases CYP11B2 expression and aldosterone production (48). This strongly suggests that WNT4 may be essential to establish a normal pattern of β-catenin activation in adrenal zona glomerulosa. On the basis of our data, we can speculate that this pattern of activation could be involved in the establishment of zona-glomerulosa identity through stimulation of AT1R and CYP11B2 expression. Interestingly, our experiments show that β-catenin knockdown not only decreases aldosterone production but also robustly stimulates CYP17 expression (Supplementary Material, Fig. S11A) and DHEA (Supplementary Material, Fig. S11B) production at the expense of cortisol (Supplementary Material, Fig. S11C) in H295R cells. Although cell culture experiments are to be considered with caution, this suggests that alterations in WNT signalling pathway activity and β-catenin dosage can result in a switch from one hormonal production phenotype to another. It is therefore tempting to speculate that localized WNT signalling pathway activation plays an essential role in the establishment of functional adrenal zonation through both promotion of zona glomerulosa and inhibition of zona fasciculata/reticularis differentiation. Future experiments aiming at inactivating Ctnnb1 specifically in zona-glomerulosa or fasciculata cells should clarify these hypotheses.
β-Catenin is involved in the control of steroidogenesis in the ovary, where its activity relies on physical and functional interactions with the nuclear receptor SF-1 on StAR and CYP19A1 promoter (49–53). In contrast, we have shown that β-catenin controls adrenal expression of CYP21 and CYP11B2 through a novel molecular mechanism involving stimulation of the transcription of the NGFIB family nuclear receptors, NURR1 and NUR77. NUR77 was previously identified as a target gene of β-catenin in colon cancer lines (29). Here, we extend these findings to the adrenal cortex in vitro and in vivo. We also show that NURR1, another member of the NGFIB family of nuclear receptors, is a transcriptional target for β-catenin. Interestingly, both NUR77 and NURR1 have been shown to inhibit the activity of β-catenin either by inducing its degradation (54) or by decreasing its activity as a transcriptional coactivator (55). These observations are suggestive of a regulation loop in which β-catenin induces NGFIB family members expression which in turn negatively feedback on the transcriptional activity of β-catenin. Our mechanistic findings are thus likely to extend to a number of tissues such as the pancreas, lung or cervix in which both β-catenin and NGFIB nuclear receptors have been involved in tumourigenesis (56–58).
Treatment of a majority of PA cases relies on mineralocorticoid-receptor antagonists. However, most of the time, these inhibitors do not achieve normalization of blood pressure and are associated with endocrine side effects, which decrease patients' compliance with treatment (4–6). We have conducted a pilot study in which ΔCat mice were fed a control or 0.2% quercetin-enriched diet (Supplementary Material, Fig. S12). After 90 days on the specific diets, mice receiving quercetin showed a significant and specific decrease in aldosterone production (Supplementary Material, Fig. S12A and B), which was correlated with decreased Wnt pathway activity (Supplementary Material, Fig. S12C). Quercetin is the most abundant dietary flavonoid endowed with the capacity to inhibit multiple intracellular signalling pathways, amongst which β-catenin interaction with LEF/TCF (59,60). Interestingly, increased quercetin consumption is associated with a decrease in blood pressure in animal models (61) and in hypertensive patients (62). Although multiple molecular mechanisms have been proposed to account for these activities (62), on the basis of our data, we propose that some of the hypotensive effects of quercetin may be mediated by inhibition of WNT signalling and aldosterone secretion in the adrenal cortex. The broad spectrum of quercetin activities is probably incompatible with clinical use as an inhibitor of WNT signalling. However, small molecule drugs targeting β-catenin are emerging in Phase I and II clinical trials (63). Their therapeutic potential for the treatment of PA remains to be evaluated.
In conclusion, our data provide novel molecular insights that account for a large number of APA in patients. They show that aberrant WNT pathway activation plays a central role in the aetiology of this disease and suggest that it may be a relevant pharmacological target to allow normalization of aldosterone production in patients with APA.
MATERIALS AND METHODS
Studies on patients' samples were conducted in agreement with the principles of the declaration of Helsinki and were approved by CPP (Comité de Protection des Personnes) Ile de France II. All participants gave informed consent for the study. All animal studies were approved by Auvergne Ethics committee and were conducted in agreement with international standards for animal welfare (protocols CE 76-12 and CE 77-12).
We obtained formalin-fixed and paraffin-embedded adrenals through the COrtico- et MEdullo-surrénale Tumeurs Endocrines (COMETE) network from 31 patients and 17 patients who had undergone surgery for APA at Hôpital Européen Georges Pompidou and Cochin Hospital, respectively (Plan Hospitalier de Recherche Clinique Grant AOM06179 to the COMETE-INCa network). The clinical and biological characteristics of the patients are summarized in Supplementary Material, Table S1A. Methods for screening and criteria for diagnosing PA and APA were in accordance with institutional guidelines. For controls, we selected 17 secreting NAPA, which were diagnosed in the Endocrinology Department of Cochin Hospital. These did not have nucleocytoplasmic accumulation of β-catenin, which was considered as the criteria for absence of constitutive Wnt pathway activation. Our retrospective transcriptome analyses were performed with the datasets from six APA and four control adrenals of the Cochin/COMETE series that were previously analysed on cDNA microarrays (22). These analyses were carried out using ‘R’ (http://www.r-project.org/). The LIMMA test was used for group comparisons (LIMMA R package) and P values were adjusted using the Benjamini–Hochberg correction method. All data are available on ArrayExpress web site (http://www.ebi.ac.uk/arrayexpress, experiment E-TABM-311, samples CIT-HS-ACT-EC-ACA17, CIT-HS-ACT-EC-ACA19, CIT-HS-ACT-EC-ACA29, CIT-HS-ACT-EC-ACA39, CIT-HS-ACT-EC-ACA40, CIT-HS-ACT-EC-ACA42, CIT-HS-ACT-EC-NA1, CIT-HS-ACT-EC-NA2, CIT-HS-ACT-EC-NA3, CIT-HS-ACT-EC-NA4). RTqPCR analyses were performed on the same 6 APA and 11 NAPA as well as on 29 additional APA from the HEGP/COMETE series and 6 NAPA from the Cochin/COMETE series (see Supplementary Material, Table S1B). Seven NA were obtained from radical nephrectomies (Cochin/COMETE and Rouen/COMETE). All patients gave written informed consent to participate to the study.
Immunohistochemistry for β-catenin, Dab2 and Cyp11b2 was performed on tissues embedded in paraffin as previously described (19,26,64). For patients' samples, the entirety of β-catenin-stained sections was examined. Immunohistochemical labelling was evaluated by a trained pathologist for the presence of membrane, cytoplasmic and nuclear staining by a qualitative assessment. The intensity of staining was not scored. Images were acquired with a Zeiss Axioplan 2 microscope and Axiocam HR camera. They were minimally processed for global levels and white balance using Adobe Photoshop®.
In situ hybridization
In situ hybridization analyses for human CYP11B2 detection were conducted as previously described (21).
Patients' DNA sequencing
For identification of KCNJ5 mutations, the DNA coding for amino acids 122–199 was sequenced as described in (17). Mutations in the third and fifth exons of CTNNB1 gene were analysed as described in (21).
Genomic DNA methylation analysis
Methylation analyses were conducted on genomic DNA from 4 NA, 15 NAPA devoid of β-catenin activation and 11 APA (see Supplementary Material, Table S1B). Genomic DNA was extracted from tumour samples using standard procedures and 500 ng of sample DNA were bisulfite treated by EpigenDx (Hopkinton, MA, USA) using a proprietary bisulphite salt solution. Briefly, DNA was diluted to 45 and 5 μl of 3 N NaOH were added, followed by a 30-min incubation at 42°C to denature the DNA. One hundred microliters of bisulfite salt solution were added to the DNA and incubated for 14 h at 50°C. Bisulfite-treated DNA was purified using Zymogen DNA columns and was eluted 20 μl of TE, pH 8.0. One microliter of the converted DNA was used for each PCR reaction. The PCR was performed with 0.2 μm of each primer. One of the PCR primers was biotinylated to purify the final PCR product using Sepharose beads. The PCR product was bound to Streptavidin Sepharose HP (GE Healthcare), and the Sepharose beads containing the immobilized PCR product were purified, washed and denatured using a 0.2-m NaOH solution. They were the washed using the Pyrosequencing Vacuum Prep Tool (Pyrosequencing, Qiagen) as recommended by the manufacturer. Then 0.2 μm pyrosequencing primer was annealed to the purified single-stranded PCR product. Ten microliters of the PCR products were sequenced by Pyrosequencing PSQ96 HS System (Biotage AB) following the manufacturer's instructions (Pyrosequencing, Qiagen). The methylation status of each locus was analysed individually as a T/C SNP using QCpG software (Pyrosequencing, Qiagen). The percentage of methylation at each CpG site was compared between normal adrenal, NAPA and APA using Anova followed by Tukey's post hoc test.
All animal studies were approved by Auvergne Ethics committee and were conducted in agreement with international standards for animal welfare (protocols CE 76-12 and CE 77-12). Sfrp2 mice (on a Sv129EV/C57Bl6 mixed background) were kindly provided by Dr Brendan Lee and were previously described (24). ΔCat mice (on a pure C57Bl6 background) were generated in our lab and were previously characterized (19). Female Sfrp2 knockout mice were fed a standard diet (SAFE, U8958v106). Blood was collected by retro-orbital puncture in 2 µl EDTA at different time points and used for plasma aldosterone concentration measurement. Individual increases in plasma aldosterone were found as early as 7 months but were only significant in the knockout group at 13 months. After the experiment, mice were euthanized by decapitation and blood was collected on vacuum blood collection tubes (VF-053STK, Terumo). Adrenals were either frozen in liquid nitrogen or fixed in 4% PFA. Ten-month-old female ΔCat mice were fed with either a standard control diet (Safe, A03) or a 0.2% quercetin-enriched diet (based on Safe A03) for 3 months. Blood and tissues were collected and processed as described above. Littermate control animals were used in all experiments.
Human adrenocortical cancer H295R cell line was grown as previously described with DMEM/Ham's F12 supplemented with 10% fetal bovine serum (S1800-500, Biowest), 2 mml-glutamine (25030, Gibco), 50 U/ml penicillin, 100 µg/ml streptomycin and 1× insulin transferrin selenium (41400-045, Gibco). Mouse adrenocortical tumour Y1 cell line was grown with DMEM/Ham's F12 supplemented with 10% fetal bovine serum (S1800-500, Biowest), 2 mml-glutamine (25030, Gibco), 50 U/ml penicillin and 100 µg/ml streptomycin.
Statistical analysis of RTqPCR, western-blot and transfection data was performed with GraphPad Prism5 software. When t-test was applied, variance was always evaluated. In case of unequal variances, Welch's t-test was performed. In all boxplot and scatter plot representations whiskers show the minimum and maximum values for each population. For methylation analyses, outliers are represented as dots detached from the boxplots.
TOP- and FOP-FLASH vectors were a kind gift of Dr Perret (Institut Cochin). Reporter plasmids hCYP11B2 (pB2-1521) WT and NBRE/Ad5/Ad1M were kindly provided by William E. Rainey (Department of Physiology, Medical college of Georgia, Augusta, GA, USA) (28). For the LEF/TCF mutant construct, the two LEF/TCF sequences 5′-CTTTGCTG-3′ (−1266/−1273) and 5′-CTTTGAAAAGG-3′ (−768/−774) were changed to 5′-agggtcgG-3′ and 5′-agggtccAAGG-3′, respectively, using the QuickChange II kit (Stratagene). All bases were numbered relative to hCYP11B2 transcriptional start site.
Cell transfections and treatment
H295R cells were seeded at a density of 8 × 105 cells per well in 6-well plate. The day after seeding, 1 µg of reporter plasmid DNA was transiently transfected using Transfast reagent (Promega) according to manufacturer's instructions. Following transfection, cells were incubated with 2 ml minimum medium for 16 h (DMEM/F12 medium containing l-glutamine and antibiotic). Where indicated, cells were pre-treated for 24-h with 1.0 µm PKF115-584 (WNT/β-catenin inhibitor, El-198-0100, Enzo Life Sciences), followed by 10 nm of Angiotensin II (A9525, Sigma-Aldrich) alone or in combination with PKF115-584 for 6 h. Control cells were treated with DMSO, which was used as the vehicle for PKF115-584. Cells were then lysed in reporter lysis buffer (E3971, Promega) and assayed for luciferase activity with the GenofaxA luciferase assay system (Yelen). All experiments were performed in triplicate and repeated at least thrice.
To evaluate the effect of Wnt pathway inhibition, H295R cells were seeded in 6-well plate at a density of 8 × 105 cells per well plate and cultured in minimum medium overnight before pre-treatment for 24-h with 0.5 or 1.0 µm PKF115-584 or DMSO (control group). After this, the cells were treated with 10 nm of Angiotensin II alone or in combination with PKF115-584 for 6 h.
For knockdown experiments, siRNA sequences targeting β-catenin 5′-AGCUGAUAUUGAUGGACAG-3′ and GFP 5′-ACUACCAGCAGAACACCCCUU-3′ were used. H295R cells were seeded in 6-well plate at a density of 4 × 105 and siRNAs were transfected twice at 24 h intervals, using Effectene reagent (Qiagen) according to manufacturer's protocol. Five days after the first transfection, cells were deprived of serum overnight and were induced for 6 h with 10 nm Angiotensin II.
For induction of Wnt pathway activity, Y1 cells were seeded in 6-well plate at a density of 3 × 105 cells per well and cultured in minimum medium overnight, before treatment with 500 nm BIO (B1686, Sigma) or DMSO (control group) in the presence or absence of 100 nm actinomycin D for the indicated amount of time. For TOP/FOP-FLASH assays, 3 × 105 Y1 cells were transfected with 1 µg of each plasmid using Metafectene (Biontex) in complete medium for 6–7 h. They were then cultured in minimum medium overnight and treated with DMSO or BIO as described above, for 6 h.
Reverse-transcription quantitative PCR
Cells were harvested in lysis buffer and total mRNAs were extracted using RNAII nucleotide extraction kit (Macherey Nagel) according to manufacturer's instructions. One microgram of mRNA was reverse transcribed for 1 h at 42°C with 5 pmol of random hexamers primers, 200 units reverse transcriptase (M-MLV RT, M1701, Promega), 2 mm dNTPs and 20 units RNAsin (N2615, Promega). One microliter of a one-tenth dilution of cDNA was used in each quantitative PCR. This was conducted with Taqman® probes of the Gene Expression Assay pool (Applied Biosystems, see below) or MESA GREEN Mastermix Plus (RT-SY2X-06+WOUFL, Eurogenetec). Primers and probes are listed in Supplementary Material, Tables S2 and S3.
Thirty micrograms of total proteins were loaded on a 10% SDS–PAGE gel, transferred onto nitrocellulose and detected with an antibody to β-catenin 1/1000 (610153, BD Biosciences Pharmingen), NURR1 1/500 (sc-991, Santa Cruz) and NUR77 1/1000 (sc-990, Santa Cruz). Expression of these proteins was normalized to expression of β-actin (A2066, Sigma) and signals were quantified with a DNR MF ChemiBis 3.2 camera and Multi Gauge software suite (Fujifilm).
H295R cells were seeded at a density of 5 × 106 per 10 cm dish and treated with 1.0 µm PKF115-584 for 30 h. Following treatment, they were fixed by addition of 1% formaldehyde to the culture medium for 10 min. Fixation was stopped by addition of 125 µm glycine. Cells were rinsed three times with PBS and scraped in CLB buffer (5 mm PIPES, 85 mm KCl, 0.5% NP40). After centrifugation, nuclei pellet was resuspended in NLB buffer (50 mm Tris–HCl, 10 mm EDTA, 1% SDS) and incubated on ice for 45 min. Chromatin was then sheared to an average length of 500 pb and submitted to immunoprecipitation with 5 µg of antibodies directed against β-catenin (610153, BD Biosciences Pharmingen), NURR1 (sc-991, Santa Cruz), NUR77 (sc-990, Santa Cruz) or non-immune IgG controls (Millipore). Chromatin-antibodies complexes were collected with protein A or G magnetic beads (Dynabeads, Invitrogen). After seven washes in IP buffer (150 mm NaCl, 50 mm Tris–HCl, 5 mm EDTA, 0.5% NP40, 1% Triton X-100), beads were resuspended in 100 µl of 10% Chelex slurry and DNA was recovered as previously described (65).
For in vivo ChIP, six WT or ΔCat adrenals gland were cross-linked in PBS with 1% formaldehyde for 10 min at room temperature. Fixation was stopped with 125 µm glycine and adrenals were rinsed three times with PBS. Tissues were then disrupted in CLB with a glass tissue grinder. The next steps of the protocol were as described above for H295R cells. Primer pairs used for ChIP are listed in Supplementary Material, Table S4. For all ChIP experiments, specificity controls were performed with primers hybridizing within the coding sequence of each candidate gene (Supplementary Material, Table S4).
Aldosterone, DHEA, cortisol and corticosterone were quantified in plasma and cell medium using an aldosterone ELISA kit (CAN-ALD-450, Diagnostic Biochem Canada), a DHEA ELISA kit (CAN-DH-490, Diagnostic Biochem Canada), a Cortisol ELISA kit (CAN-C-270, Diagnostic Biochem Canada) and a corticosterone 3H kit (MP Biomedicals) according to manufacturers' instructions. For cell medium quantifications, hormone concentrations were normalized to total cellular proteins measured by Bradford assays (Bio-Rad). Renin was measured by direct chemiluminescence with the Liaison Direct Renin kit (Diasorin), according to manufacturer's instructions.
This work was supported through institutional support from Centre National de la Recherche Scientifique, Institut National de la Santé et de la Recherche Médicale; Université Blaise Pascal; Université d'Auvergne; Ministère de la Recherche et de l'Enseignement Supérieur and grants from Fondation ARC pour la Recherche sur le Cancer (SFI 20101201397; young researcher grant to A.B.); La Ligue Contre le Cancer, Allier committee; Agence Nationale pour la Recherche (013-01 and 08-GENO-021); Fondation pour la Recherche sur l'Hypertension Artérielle; Fondation pour la recherche médicale (ING20101221177) and Programme Hospitalier de Recherche Clinique (AOM 06179).
We thank B. Terris at Cochin Tumour Bank and the COMETE (COrtico MEdullo Tumeurs Endocrines) network for the provision of APA tissue samples, Dr Brendan Lee (Baylor College of Medicine, Houston, USA) for kindly allowing us to use Sfrp2 knockout mice, Dr William E Rainey (Medical college of Georgia, USA) for providing CYP11B2 reporter constructs and Dr Ed Laufer (Columbia University, New York) for provision of antibodies. We also thank Khirredine Ouchen, Sandrine Plantade and Philippe Mazuel for animal care.
Conflict of Interest statement. None declared.