Abstract
SUMOylation is a dynamic posttranslational modification, that provides fine-tuning of protein function involved in the cellular response to stress, differentiation, and tissue development. In the adrenal cortex, an emblematic endocrine organ that mediates adaptation to physiological demands, the SUMOylation gradient is inversely correlated with the gradient of cellular differentiation raising important questions about its role in functional zonation and the response to stress. Considering that SUMO-specific protease 2 (SENP2), a deSUMOylating enzyme, is upregulated by ACTH/PKA signalling within the zona Fasciculata (zF), we generated mice with adrenal-specific Senp2 loss to address these questions. Disruption of SENP2 activity in steroidogenic cells leads to specific hypoplasia of the zF, a blunted reponses to ACTH and isolated glucocorticoid deficiency. Mechanistically, overSUMOylation resulting from SENP2 loss shifts the balance between ACTH/PKA and WNT/β-catenin signalling leading to repression of PKA activity and ectopic activation of β-catenin. At the cellular level, this blocks transdifferentiation of β-catenin-positive zona Glomerulosa cells into zF cells and sensitises them to premature apoptosis. Our findings indicate that the SUMO pathway is critical for adrenal homeostasis and stress responsiveness.
The adrenal cortex is a constantly self-renewed endocrine organ composed of concentric zones, including the outermost zona glomerulosa (zG) layer producing mineralocorticoids and the innermost zona fasciculata (zF) layer producing glucocorticoids. According to the centripetal migration model occurring during postnatal development, progenitors cell populations located in the adrenal capsule (characterised by GLI1 expression) or within the zG (characterised by SHH expression) consecutively differentiate into steroid-producing zG cells, then through a process of zonal transdifferentiation, convert into zF cells before eventually undergoing apoptosis at the corticomedullary junction1–3.
Genetic models and in vitro approaches have identified two important signalling pathways for adrenal cortex homeostasis. On the one hand, the WNT/Rspondin/β-catenin pathway is necessary for the maintenance of progenitor pools and the acquisition of zG identity4–6. On the other hand, cAMP/PKA signalling, stimulated by pituitary ACTH, triggers the recruitment of progenitors by inducing the transdifferentiation of zG cells into zF cells and stimulates glucocorticoids production7,8. We and others have previously shown that these two signalling pathways antagonise each other by modulating various actors such as WNT4, PDE2A or CCDC803,6,9. Optimal response to PKA signalling and therefore zF differentiation is also subject to epigenetic programming by the histone methyl transferase EZH210. Nonetheless, the mechanisms that maintain adrenal cortex zonation and balance between these two pathways are yet to be discovered.
SUMOylation is a dynamic and one of the fastest evolving11 posttranslational modification consisting in the covalent addition of SUMO peptides on a target protein. This modification can affect various processes such as protein stability, interactions or subcellular localisation12. There are three main SUMO peptides in rodents, namely SUMO1, which shares around 50% of identity with SUMO2 and SUMO3, the two latter differing by only three amino acids. SUMOylation is achieved through an enzymatic cascade involving activation by the E1 heterodimer (SAE1/UBA2), conjugation by the sole E2 enzyme UBC9 (encoded by Ube2i) and final ligation by various E3 SUMO ligases such as members of the PIAS and TRIM families as well as RANBP2 and CBX4.13,14 SUMO peptides can be removed from SUMO-conjugated substrates by deSUMOylases belonging mainly to the Sentrin-specific proteases family (SENPs) or by the more recently discovered DeSI-1 and USPL1 making this posttranslational modification highly dynamic (Figure S1.A.B.). Several in vitro and in vivo studies have highlighted the importance of controlling SUMOylation levels to enable differentiation or maintain cellular identity15–17 and tissue homeostasis in vivo in various cell lineages18,19. The adrenal gland could provide a paradigm to study how SUMOylation dynamics can interact with the function and homeostasis of an organ, in charge of constant adaptation to stress.
We have previously shown that protein SUMOylation follows a decreasing centripetal gradient in human and mouse normal adrenal cortices. Moreover, this gradient is altered in genetic endocrine diseases with deregulated PKA or WNT signalling pathways20. Remarkably, SUMOylation is negatively and acutely regulated by ACTH in both adrenal cortex and adrenocortical cell cultures through transcriptional control of key enzymes, especially SENP2 whose upregulation by PKA correlates with transient hypoSUMOylation in zF. Interestingly, PKA-mediated up-regulation of Senp2 was previously shown to promote the progression of preadipocytes into the adipogenic program15. Taken together, these studies suggest that limiting SUMOylation may facilitate or be a prerequisite for any change in differentiation states. Conversely, an excess of WNT/β-catenin signalling in the adrenal cortex induces an expansion of zG identity that is correlated with a high SUMOylation state20. Finally, preventing in vivo SUMOylation of the transcription factor SF-1 (SF-12KR/2KR mice), the main driver of adrenogonadal cell fate, disturbs endocrine development by maintaining discrete gonadal traits in the cortex and adrenal traits in the testis21. This highlights the need to control SUMOylation during cell fate decisions leading to adrenal cortex identity.
We hypothesise that disruption of the SUMOylation gradient in the adrenal cortex may disrupt zonation and impair adaptive response to stress. In order to understand the implication of SUMO pathway on homeostatic maintenance and endocrine function, we have developed mouse models of adrenal hyperSUMOylation by conditional ablation of Senp2 in the cortex (Senp2cKO). Our report reveals that Senp2cKO mice show zone-specific adrenal atrophy, isolated glucocorticoid deficiency and blunted response to ACTH. Progressive atrophy of zF evoked by SENP2 deficiency results from a blockade of zonal transdifferentiation, early apoptosis and impaired PKA catalytic activity that cannot be rescued by genetic derepression of the PKA holoenzyme. SENP2-deficient adrenals also show increased β-catenin SUMOylation and activity that may help to antagonise PKA signalling, thus maintaining the suppression of zF identity. As Senp2 expression is itself under the control of ACTH/PKA, our data identify SUMOylation as a feedforward mechanism that readies the adrenal cortex to respond to stress and maintain functional zonation.
Results
Senp2 invalidation in the adrenal cortex leads to zF hypoplasia and adrenal dysplasia
To assess the role of SUMOylation in the adrenal cortex, we have developed a mouse model with specific deletion of the ACTH-regulated deSUMOylase SENP220 in steroidogenic cells by mating Senp2fl/fl mice22 with SF-1(Nr5a1)-Cre mice23. Senp2 conditional knock-out mice are later referred to as Senp2cKO. Senp2 deletion was confirmed in 4-week-old mouse adrenals by RT-qPCR analyses showing reduced Senp2 mRNA accumulation in both genders and by genomic PCR, confirming adrenal-specific recombination at the Senp2 locus (Figure S1.C.D.).
Monitoring of adrenal mass from 4 to 40 weeks of age revealed significant adrenal hypoplasia in mutants, occurring between 4 and 8 weeks in both sexes. After this time point, overall adrenal weight in Senp2cKO remained below that of controls in females only because of sex differences in the kinetics of adrenal mass gain (almost continuous growth in WT females contrasting with a progressive decrease in WT males over time) (Figure 1.A.). To investigate the causes of this hypoplasia, we performed H&E staining, which revealed two different histological phenotypes: either homogeneous atrophy of the cortex leaving medulla centrally located, or cortical atrophy accompanied by dysplasia due to clusters of large eosinophilic cells, usually at one pole of the gland and pushing the medulla toward the other pole (Figure 1.B. left panel).
Immunofluorescence co-staining for zonal markers showed that, compared to WT, the integrity of zG (DAB2+) did not seem affected in Senp2cKO adrenals of 8-week-old mice, while zF (AKR1B7+) was atrophic and sometimes mislocated together with the medulla (TH+) in dysplastic glands (Figure 1.B. right panels). To confirm that the zF was the most affected by Senp2 invalidation, the number of cells in each cortical zones and medulla was counted on 2D sections. This showed a dramatic and specific reduction in total cortical cell number regardless of sex, solely attributable to the 75-80% loss of zF cells (Figures 1.C. & S1.E). Laminin immunostaining was then used to examine vascular architecture in WT and mutant adrenal sections. Typical capillaries surrounding the “rosette” structures in the zG and delimiting the zF cells columns were evidenced in WT adrenal sections. Consistent with zonal markers immunostaining, a rosette-like pattern of vascularisation, typical of zG, was found in areas with atrophic zF (Figure 1.B.a.), whereas large columnar structures surrounding clusters of hypertrophic zF cells were evidenced in dysplastic Senp2 -deficient glands (Figure 1.B.b.).
To further characterise the origin of adrenal dysplasia, we introduced the Rosa26RmTmG reporter transgene24 (Figure S1.F.) into Senp2cKO background to trace Senp2 recombination using GFP immunostaining and non-recombined cells using Tomato immunostaining. WT adrenals (from Sf1-Cre/+::Senp2fl/+::R26RmTmG/mTmG triple transgenic mice) presented full recombined cortex, while in dysplastic Senp2cKO adrenals (Sf1Cre/+::Senp2fl/fl::R26RmTmG/mTmG), most of the cells belonging to the mislocated zF were GFP-negative/Tomato-positive, hence non recombined (Figure 1.D.). To address the identity of these cells escaping recombination, we performed co-staining of GFP with the canonical steroidogenic marker SF-1. Surprisingly, GFP-negative cells expressed SF-1, implying that they kept full steroidogenic identity (Figure 1.D. bottom centre panel) as also suggested by detection of zonal markers (Figure 1.B. right panels, bottom). To confirm that these cells were hypertrophic as H&E and Laminin staining suggested, we measured 2D cell areas in the zF of WT and Senp2cKO adrenal sections. This showed that in females, although GFP-positive and GFP-negative zF cell areas did not differ in the mutant, GFP-negative cells were larger than zF cells from WT adrenals (Figure 1.E.). Similar observations were made in males but with a P-value of 0.0560. Next, we assessed global SUMOylation status by western blot and found no variation in the profile of SUMO1 or SUMO2/3 conjugates as a function of the genotype (Figure S1.G.H.). However, GFP-positive zF cells retained nuclear SUMO2/3 staining whereas GFP-negative had diffuse staining in the cytoplasm (Figure S1.I.J.). Together, these results indicate that the clusters of cellular hypertrophy found in dysplastic Senp2cKO adrenals are predominantly composed of zF cells that have escaped Senp2 ablation and that this phenotype (hypertrophic zF clusters) seems more likely to occur in females. To examine further this sexual dimorphism, we assessed the proportion of mutant adrenals presenting with hypertrophic zF clusters from 4 to 40 weeks of age in both genders (Figure 1.F.). Throughout time, the frequency of adrenals with hypertrophic zF clusters increased in a sexually dimorphic fashion: indeed, for these clusters to develop in all adrenals, it took 40 weeks in males and only 24 in females. Thus, in 40-week-old mice, Senp2cKO adrenal cortex contained, compared to WT, very few GFP-positive zF cells, with GFP staining limited to zG and non-recombined cells constituting the entire zF (Figures 1.D. right panel & S1.K.).
Altogether, these results show that SENP2 is necessary for proper adrenal cortex zonation and homeostatic maintenance. Indeed, its inactivation initially leads, to zF atrophy, which is compensated over time by the recruitment of cells escaping Senp2 recombination allowing them to maintain a wild-type Senp2 zF in otherwise mutant adrenals. Importantly, we show that females more efficiently overcome zF atrophy induced by Senp2 loss. This further highlights the sexually dimorphic traits of adrenal homeostasis25,26.
Loss of SENP2 is associated with isolated glucocorticoid deficiency
Given the profound alterations and notably, the time and sex-dependant remodelling of adrenal cortex zonation in Senp2cKO, we measured changes in circulating steroid levels and assessed steroidogenic gene expression. Steroid hormones are synthesised from cholesterol through enzymatic processes resulting in corticosterone production by zF cells and aldosterone by zG cells (Figure 2.A.). Since zF was the most impacted zone by the Senp2 mutation, we anticipated a reduction in plasmatic corticosterone levels, the main glucocorticoid in rodents. Indeed, compared to control males, corticosterone levels were dramatically reduced in Senp2cKO at 4 weeks and remained lower at 8 weeks of age. Interestingly, levels normalised over time to be indistinguishable from controls by 24 weeks (Figure 2.B. top). In contrast, Senp2 inactivation had no impact on corticosterone concentrations in females, at any time point (Figure 2.B. bottom). The integrity of the zF and the corticosterone production are under the strict control of pituitary ACTH that maintains homeostasis through a negative feedback loop mediated by the glucocorticoids on the hypothalamic-pituitary-adrenals (HPA) axis. Therefore, we measured circulating ACTH in Senp2cKO and found a 8-10-fold increase at 24 weeks in both sexes. This is consistent with a dysfunction of the zF cells, resulting in a subclinical insufficiency over time (Figure 2.C.).
For insights into the mechanisms of this insufficiency, we analysed the expression of steroidogenic genes in adrenals of 4-week-old mice. RT-qPCR analyses showed that mRNA levels of Star and Cyp11a1 encoding rate limiting step proteins in steroidogenesis, were decreased in both male and female Senp2cKO adrenals (Figure 2.D.). Interestingly, Cyp21a1, Hsd3b1 and Cyp11b1 transcripts were specifically downregulated in female Senp2cKO adrenals (Figure 2.D). To assess the impact of Senp2 loss on zG function, we measured mineralocorticoids plasma levels and Cyp11b2 expression. Consistently with histological observations, we did not find any negative effect of Senp2 ablation on aldosterone levels, but rather a positive trend with an increase in 18-hydroxy-corticosterone plasmatic concentration in Senp2cKO males at 24 weeks and a trend toward upregulation of Cyp11b2 expression in Senp2cKO female adrenals at 4 weeks (Figure S2.A.B.C.).
In conclusion, these results draw a picture of the differential role of Senp2 in the zonation of the adrenal cortex. While it is dispensable for the zG, its absence leads to a deficient zF struggling to produce enough corticosterone to maintain homeostasis. This effort is illustrated by elevated circulating ACTH and hypertrophic zF cells, hallmarks of isolated glucocorticoid deficiency.
Adrenal cortex lacking Senp2 shows blunted ACTH response
To understand the underpinnings of the adrenal phenotype in Senp2cKO mice, knowing that ACTH is a regulator of SUMOylation in the adrenal cortex20, we assessed the endocrine and transcriptional steroidogenic responses to acute ACTH stimulation. Plasma concentrations of steroids associated with glucocorticoid metabolism (Figure 2.A.) were determined by LC-MS/MS, in 24-week-old mice injected 2 hours before with ACTH and compared to vehicle (Figure 3.A. & Table 1.). We confirmed that, at this age, there were no differences in plasma levels of all steroids measured in basal conditions (vehicle) between WT and Senp2cKO mice. As expected, ACTH treatment induced a strong increase in adrenal steroids (e.g. corticosterone levels were 3- to 7-fold induced in females and males, respectively) in plasma from WT mice (except progesterone in females, which comes mainly from the ovaries). However, the ACTH-stimulated endocrine response was heavily blunted and at least halved in Senp2cKO mice (Table 1.). To determine whether alteration of endocrine response correlated with changes in gene expression, we measured mRNA levels of ACTH-responsive genes involved in initial steps of steroidogenesis (i.e. Scarb1 and Star) by RT-qPCR (Figure 3.B.). In males, Senp2 mutation did not alter Scarb1 and Star transcriptional responsiveness to ACTH (1.5- to 2-fold induction after 2h) but impaired their basal expression, so that mRNA levels in ACTH-treated mutant adrenals barely reached basal expression in controls. By contrast in females, although their basal expression remained unchanged, both Scarb1 and
Star genes entirely failed to respond to ACTH in Senp2cKO adrenals (Figure 3.B.). To test whether this blunted transcriptional response relied on changes in phosphorylation of PKA substrates (Figure S3.A.), we performed western blots on trans-acting factor CREB and SUMO E3 ligase TRIM28, in response to 30 min ACTH treatment, in WT and mutant mice. Ser133 CREB and Ser473 TRIM28 phosphorylation levels were similar in basal conditions, increased in WT upon ACTH stimulation but failed to respond to treatment in Senp2cKO adrenals (Figure 3.C.). This impaired response to ACTH/PKA-mediated phosphorylation was unlikely caused by altered expression of ACTH receptor and co-receptor (Mc2r and Mrap, respectively) that were unaltered by Senp2 loss (Figure S3.B.).
To determine whether the deficiency in ACTH response involved PKA holoenzyme or occurred upstream of the kinase, we assessed the capacity of genetic activation of PKA to rescue adrenal insufficiency in Senp2cKO mice by removing the RIα subunit (Prkar1a floxed allele) known to repress PKA catalytic activity7,9,27. As expected, 4-week-old Prkar1acKO mice developed large adrenals with hyperplastic zF and atrophic zG (Figure S3.C.D.). Adrenals from Senp2,Prkar1adcKO and Senp2cKO mice were both dysplastic and showed reduction in cortical cell numbers. However, dcKO adrenals showed an atrophic zG (loss of DAB2 staining) presumably resulting from the antagonistic action of PKA signalling on zG identity (Figures 3.D. & S3.C.D.)7,9. As a result, in the absence of NaCl supplementation, Senp2,Prkar1adcKO mice died prematurely from salt wasting (Figure S3.E.), whereas Senp2cKO mice only suffered from isolated glucocorticoid deficiency. Thus, genetic derepression of PKA was unable to overcome cortical atrophy and dysplasia imparted by Senp2 deficiency. This suggested that consequences of Senp2 loss, including the excess of SUMOylation, had a dominant impact on the zF homeostasis over PKA constitutive activation. Then, we explored endocrine activity of double KO mice. Although plasma corticosterone dosage showed no differences among the four genotypes, plasma ACTH concentrations were elevated in Senp2cKO and reduced in Prkar1acKO, consistent with the corresponding associated disorders i.e. glucocorticoid deficiency and ACTH-independent glucocorticoid excess, respectively (Figure 3.E.F.). By contrast, Senp2-Prkar1a double ablation restored ACTH levels to control values. This strongly suggested that the lack of Senp2 resulted, among other things, in the repression of PKA catalytic activity that the deletion of RIα regulatory subunits could partially overcome. Indeed, PKA kinase activity was decreased by 30% in Senp2cKO adrenal extracts compared to WT without affecting Cαβ protein levels (Figure 3.G.H.). To test a possible direct repressive effect of SUMOylation, we measured kinase activity in the presence of a mix of SUMO1/2 modified with vinyl sulfone (SUMOs-VS) acting as specific trap and potent inhibitors of SENPs SUMO proteases present in the extracts28. Under these conditions enhancing SUMOylation in adrenal extracts, PKA kinase activity was further decreased in Senp2cKO, reaching a 50% inhibition (Figure 3.G.H.). Together, these results present SENP2 as a mandatory actor of proper ACTH response acting most likely by limiting repressive action of SUMOylation on the catalytic activity of the PKA holoenzyme.
Senp2 is necessary for the acquisition of zF identity
Based on the blunted response to ACTH, we decided to assess the differentiation status of the zF in Senp2cKO mice. We took advantage of cortical cells’ capacity to escape recombination to compare the intensity of differentiation markers in neighbouring cells, differing only by their recombination status. We performed triple staining for GFP, used as a proxy of Senp2 recombination, with DAB2 and AKR1B7 labelling zG and zF, respectively. We observed a consistent lower AKR1B7 staining intensity in GFP-positive cells, indicating that loss of Senp2 hinders cells from expressing zF markers compared to neighbouring GFP-negative cells (Figures 4.A.B. & S4.A.). RT-qPCR analysis revealed an increased accumulation of progenitors’ markers and higher number of NR2F2-positive capsular cells in males (Figure S4.B.C.) implying a default in cortical cell turnover which, together with the downregulated AKR1B7 expression, suggests a block in centripetal differentiation. This hypothesis was further supported by an increased proportion of cells coexpressing DAB2 and AKR1B7 in the cortex of Senp2cKO, suggesting altered zG to zF transdifferentiation (Figure 4.C.). To examine this hypothesis, we performed functional lineage tracing analysis of Senp2 -deficient cells using mTmG reporter mice and ASCre driver2, which allowed to delete Senp2 in zG cells after birth (ASCre/+::Senp2fl/fl ::R26RmTmG/mTmG). This Cre driver enabled to assess cortex cellular turnover through the percentage of GFP-stained cells progressing centripetally. As previously shown7,29, complete cortical cell took around 12 weeks in female and 40 weeks in male WT mice (Figure 4.D.). Interestingly, whereas GFP immunostaining marked the first third of the cortex (zG and upper zF) in 4 weeks WT females, GFP staining was confined to the zG and some rare stripes projecting into the zF in ASCre/+ Senp2cKO littermates (Figure 4.E.).
To accelerate lineage tracing, we mated Cre homozygosity, and thus deleted the Aldosterone Synthase gene (ASCre/Cre::Senp2fl/fl ::R26RmTmG/mTmG). Consistent with previous reports2, this enhanced trophic drive through renin-angiotensin signalling, and increased recombination rate, leading to almost full recombination cortex by 4 weeks of age in ASCre/Cre Senp2 heterozygous female adrenal. In sharp contrast, GFP staining was almost confined to the zG in ASCre/Cre Senp2cKO even though the zG was slightly expanded because of trophic stimulation by angiotensin (Figure 4.E.). A similar phenotype was observed in ASCre/+ Senp2cKO at 40 weeks of age with a recombined zF consisting of scattered stripes of GFP-positive cells (Figure 4.E.). We next quantified AKR1B7 protein accumulation in zF cells of ASCre/+ Senp2cKO of 40 weeks of age and again observed lower AKR1B7 staining in GFP-positive than in GFP-negative cells (GFP-95% CI [6887;7164] AU, GFP+95% CI [5427;5582] AU, P-value = 10−16) (Figure 4.F.G.). In contrast with Sf1-Cre mediated inactivation of Senp2, ASCre mediated deletion did not alter adrenal weight at 24 weeks of age (Figure 4.H.). However, when ACTH responsiveness was assessed over time, plasma corticosterone peaked 2 hours after ACTH treatment in WT, whereas the response was slower in ASCre/+ Senp2cKO and never reached statistical threshold (Figure 4.I.). This shows that even in the absence of adrenal hypoplasia, Senp2 ablation results in a block of zF transdifferentiation from zG cells. This causes incomplete zF differentiation in recombined cells and allows competitive selection of non-recombined cells, which cannot completely overcome the endocrine phenotype.
Senp2 deficient cells undergo apoptosis associated with DRP1 phosphorylation
We next examined the cellular mechanisms underlying the development of zF atrophy in Senp2cKO mice, by analysing the proliferation/apoptosis balance. The cortical proliferation index determined by scoring the number of Ki67- or BrdU-positive cells ruled out the contribution of a decreased proliferation rate to the hypoplastic phenotype, but rather showed a trend toward increased cell division in mutant adrenals (Figure S5.A.B.). Nonetheless, cleaved caspase-3 staining showed that the numbers of cells undergoing apoptosis was dramatically increased in Senp2cKO adrenals at 4 and 8 weeks of age in both sexes (Figure 5.A.). Whereas, according to the standard model, apoptosis is normally found at the corticomedullary junction (where adrenal cells die after centripetal migration30), in the Senp2cKO cortex, apoptosis occurred prematurely at the border between the zG and zF (Figure 5.A.).
DRP1 (Dynamin Related Protein 1) is considered the primary driver of mitochondrial fission and mitochondrial-dependant cell death31,32. Phosphorylation of DRP1 on Ser616 activates mitochondrial fission while that on Ser637 prevents the fission. The dysregulation of DRP1 phosphorylations on these two residues will result in imbalanced mitochondrial fission/fusion, a major cause of apoptotic cell death. Phosphorylation on Ser637 particularly raised our attention since it is catalysed by PKA33. Besides its participation to inhibition of mitochondrial fission, studies have shown that phosphorylated DRP1-Ser637 promotes steroidogenesis in Leydig cells34 and corpus luteum35. To assess the impact of ACTH/PKA on DRP1 activity in adrenal glands, we performed western blotting of DRP1 and its Ser637 phosphorylated form in WT or Senp2cKO adrenal treated with vehicle or ACTH for 30 min. We found no difference in total or phosphorylated DRP1 in vehicle treated mice. In constrast, ACTH treatment induced DRP1 Ser637 phosphorylation solely in WT mice whereas this response was abolished in Senp2cKO adrenals (Figure 5.B.).
We then took an in vitro approach to determine whether PKA-induced changes in DRP1 phosphorylation were involved in the apoptotic response of adrenocortical cells. The fasciculata-like ATC7 cells36 were treated with forskolin (FSK), a pharmacological activator of PKA (through increased cAMP cellular levels), one hour before a 2 hours-incubation with the proapoptotic drug staurosporine (STS). As expected, FSK alone resulted in increased Ser637 phosphorylation of DRP1, while STS alone induced caspase-3 cleavage. When combined, FSK limited STS-driven apoptosis, as depicted by the reduced accumulation of cleaved caspase-3 (Figure 5.C.). Of note, Ser616 proapoptotic phosphorylation of DRP1 showed a trend to be regulated in the exact opposite way to Ser637 (Figure 5.C.). To specifically assess the contribution of DRP1 in STS-induced apoptosis, we pretreated cells with the DRP1 specific inhibitor (Mdivi-1) before inducing apoptosis with STS37. Similar to FSK, pretreatment with Mdivi-1 resulted in protection against apoptosis as shown by reduced accumulation of cleaved caspase-3 (Figure S5.C.). Consistently in vivo, we observed an inverse correlation (τ = -0.3823, P-value = 0.0341) between cortical apoptosis rate and DRP1 Ser637 phosphorylation in Senp2, Prkar1a single and double knock-outs (Figure 5.D.E.F S5.D). Altogether, these results strongly suggest that increased apoptosis seen at the zG-zF boundary in Senp2cKO adrenals results from a deficient ability of ACTH/PKA signalling to properly phosphorylate DRP1 Ser637.
Senp2 deficiency leads to β-catenin hyperSUMOylation and mild activation of WNT pathway
To unravel new SUMOylation-sensitive pathways that could explain further the adrenal insufficiency of mice lacking SUMO protease SENP2, we performed bulk RNA sequencing on four-week-old male and female WT and Senp2cKO. We found 1337 genes to be differentially expressed in male (1115 up and 222 down) and 1235 in female (960 up and 275 down) (Figure 6.A.B.). Unsupervised clustering and principal component analysis discriminated samples based on genotype but not on sex (Figure S6.A.B.), implying that at 4 weeks of age, sex has a low impact on gene transcription. Since most of the genes were co-regulated in males and females, we chose to focus on these subsets of genes. We performed Gene Ontology (GO) functional enrichment analysis on upregulated or downregulated genes in Senp2cKO adrenals of both sexes (Figure 6.C.). The top GO terms associated with the upregulated genes were linked to neuron cells and function, which may be due to over representation of medullar chromaffin cells resulting from cortical hypoplasia. Pathways related to steroid processing were enriched in downregulated genes, consistent with the endocrine deficiency phenotype of Senp2cKO mice.
We performed Get Set Enrichment Analysis (GSEA) on the Kyoto Encyclopedia of Genes and Genomes (KEGG) database and selected only the enriched pathways that were present in both males and females (Figure 6.D.). This confirmed negative enrichment of the gene signature for steroid hormone biosynthesis pathway, in accordance with GO analysis and hormonal insufficiency characterised in Senp2cKO mice (Figure 2.). Moreover, we found negative enrichment of signatures associated with nucleotide excision repair, aminoacyl tRNA biosynthesis, ribosome, pyrimidine metabolism and spliceosome, indicating that Senp2 loss in the adrenal cortex altered basic cellular processes known to be regulated by SUMOylation12. Interestingly, among the positively enriched pathways, WNT signalling caught our attention as it is mandatory for adrenocortical maintenance and proper zonation38–40 (Figure 6.D.E.)
Based on RNA-Seq analyses and given that SENP2 has first been described as a negative regulator of β-catenin41–43, we sought out to see if the WNT/β-catenin pathway was altered in Senp2cKO adrenals. We evaluated the activation of the pathway by running RT-qPCR on its target genes. At 24 weeks of age, we observed a consistent ∼1.5 fold induction of Axin2, Lef1, Apcdd1 and Ccdc80 in the adrenals of Senp2cKO females, whereas in mutant males, only Ccdc80 was upregulated by about 3 fold (Figure 7.A.). To gain more insight into the modulation of the WNT signalling pathway, we extracted expression levels of 48 WNT target genes from RNA-seq data (Figure 7.B.). Among these genes, 31 were upregulated mainly in males Senp2cKO and 7 mainly in females while 10 genes were downregulated in both sexes.
We next carried out immunostaining on adrenal sections that revealed a nuclear localisation of non-phospho (active) β-catenin specifically in the inner cortex of mutant adrenals regardless of sex (Figure 7.C. top). Co-staining of total β-catenin with zF marker AKR1B7 confirmed zF identity of these cells, which were also GFP-positive and hence, had been targeted by Cre recombination (Figure 7.C. bottom).
Given that β-catenin can be SUMOylated by both SUMO144 and SUMO2/345, we hypothesised that it was SUMOylated in the adrenal cortex. We performed immunoprecipitation of both SUMO1 or SUMO2/3 and blotted β-catenin. We observed a doublet between 120 and 150 kDa for SUMO2/3 consistent with a previous report45 but no evidence for significant SUMO1 conjugation (Figure 7.D.). To evaluate if β-catenin SUMO2/3-ylation was affected upon Senp2 deletion, we immunoprecipitated β-catenin and blotted SUMO2/3. Although we did not see any difference in the amount of SUMOylated β-catenin, the accumulation of its native form was reduced in mutant adrenals whereas global SUMOylation by SUMO2/3 was unchanged (Figure 7.E.). Reciprocally, by immunoprecipitating SUMO2/3 and blotting β-catenin, we observed a slight but reproducible accumulation of SUMOylated β-catenin compared to native form (Figure 7.F.). To confirm this increase in β-catenin SUMO2/3-ylation, we used Proximity Ligation Assay (PLA) which enables to visualise in situ the close proximity of two proteins (Figure 7.G.). As a negative control, we first ran PLA between β-catenin and GATA6, a transcription factor present in the nucleus of all adrenocortical cells and which is not known to interact with β-catenin. As expected, there were no foci of β-catenin/GATA6 interaction in adrenal sections of all genotypes. In contrast, PLA between β-catenin and SUMO2/3 showed specific foci in the zG with similar density in both genotypes. However, in the zF of Senp2cKO, there was an increase in cells containing 2 or more β-catenin/SUMO2/3 dots per nucleus (Figure 7.G.).
In conclusion, loss of SENP2 leads to ectopic nuclear accumulation of β-catenin in the zF associated with its increased conjugation to SUMO2/3. This accumulation is correlated with a mild activation of WNT pathway targeting genes in a partially sexually dimorphic pattern.
Discussion
We have previously established that the overall SUMOylation of proteins in the adrenal cortex displays two remarkable properties: it gradually decreases as the centripetal trans-differentiation of zG cells into zF cells progresses and it is acutely downregulated by stress through ACTH/PKA signalling20. The present study was initiated to address the hypothesis that this posttranslational mechanism involved in the cellular response to various environmental stressors46 may be an integral part of the adrenal gland’s toolbox to produce steroid hormones critical for body homeostasis and stress adaptation. Adrenal-specific KO mouse models of SENP2 deSUMOylase and of PKA regulators established that this is indeed the case. Foetal or postnatal SENP2 dependent deSUMOylation is necessary for initial zF differentiation, its maintenance thoughout life and for ACTH/PKA-stimulated glucocorticoid production. As a result, foetal or postnatal deletion of Senp2 in adrenal steroidogenic cells causes postnatal hypoplasia limited to the zF or incomplete zF formation respectively, which can lead to isolated glucocorticoid deficiency. This selective atrophy and associated endocrine deficit result both from a blockage of zonal transdifferentiation preventing the switch from zG to zF identity, and from a repression of ACTH/PKA responsiveness. Interestingly, overt corticosterone deficiency is only measurable in males at 4 weeks of age, even though ACTH levels are higher at all ages tested in both genders. This implies that a subclinical deficiency is permanently established in the mutants, despite setting up compensation. This compensation to maintain physiological plasma glucocorticoid levels, which occurs in both sexes but with delay in males, most likely relies on the emergence of a population of cells that has escaped recombination and helps the mutant cortex to overcome zF atrophy. Indeed, their hypertrophy indicates that they are overreacting to the high levels of circulating ACTH, to compensate for the lack of corticosterone due to the dramatic atrophy of the Senp2 -deficient zF cell population.
We hypothesise that steroidogenic progenitor cells, which do not express the Cre recombinase for a still unknown reason, must gain a selective advantage over Senp2 -deficient cells, either by being able to proliferate more or by being less prone to death upon differentiation. Similar observations have been made in adrenals following SF-1 loss driven by ASCre2 or Cyp11a1-Cre47, where a majority of cells escaped recombination. How cells manage to express a protein but not the Cre while both genes depend on the same promoter is still unknown. One possibility could be methylation of the promoter. For instance, Ins1-Cre transgene is shown to be a poor quality driver to target β cells in the pancreas as its expression is silenced by de novo methylation even though Ins1 gene is normally expressed48. In any case, the triggering of a compensatory mechanism underlines the absolute necessity of maintaining Senp2 expression (and thus the possibility of reducing SUMOylation), in order to build a functional zona Fasciculata to maintain the individual’s ability to adapt to stress.
The fact that Senp2 -deficiency preferentially affects the zF while being expressed in all cortical areas49 can be explained by two non-mutually exclusive mechanisms. First, the zG already harbours high levels of nuclear SUMOylation compared to the zF20, therefore the expected increase in SUMOylation conjugates consecutive to lowered deconjugase pools will preferentially affect hypoSUMOylated regions. Moreover, we showed that SENP2 loss alters the PKA/WNT balance toward WNT signalling. The latter being already highly active in the zG4,6,38, it is not surprising to find an altered phenotype only in the zF where WNT/β-catenin signalling is naturally repressed9. The second reason is that low SUMOylation seems to be a prerequisite for adequate PKA signalling. Indeed, in our model, ACTH response is blunted and the zF atrophy phenotype is reminiscent, at least partially, of mouse models lacking ACTH receptor and its co-receptor i.e. Mc2r-/- or Mrap-/- whole-body knock-outs50,51. Although various mechanisms could be simultaneously affected and remain to be explored, there are converging pieces of evidence that SENP2 deletion-dependant SUMOylation alters PKA responsiveness. Indeed, first, our data show that SENP2 deletion decreases overall PKA activity in the adrenal glands and prevents the phosphorylation of specific targets (CREBSer133, DRP1Ser637, TRIM28Ser473). Secondly, since Senp2cKO phenotype is not rescued by concomitant deletion of PKA regulatory subunit (Prkar1a), it is very likely that SENP2 loss directly affects the PKA catalytic subunits function. This is in line with 1) the presence of conserved K169 lysine in the major subunits (PRKACA and PRKACB) amino acid sequence, predicted as a SUMO consensus motif (ψKxD/E where ψ is a hydrophobic residue) and 2) the decreased PKA activity caused by SUMO vinyl sulfone derivatives (figure 3.G.H.). Further studies will assess the SUMOylation of the catalytic subunits and their potential in fine-tuning PKA kinase activity.
SENP2 has first been described as a negative regulator of β-catenin stability43 but this effect was independent of SUMOylation. Our model showed increased β-catenin SUMO2/3ylation along with ectopic nuclear β-catenin localisation associated with a gene signature indicative of the canonical WNT signalling. Similar observations have been reported following direct SUMOy2/3lation of β-catenin in mammary epithelial cells45 or SUMOylation of TBL1/TBLR1 that in turn increased chromatin recruitment of β-catenin and its oncogenic activity in colon cancer52. However, the picture is not as clear since β-catenin SUMO2/3ylation can trigger its degradation in vascular smooth muscle cells53 or prevent it in mammary epithelial cells45. Targeting gain-of-function mutation of β-catenin in aldosterone-synthase expressing cells caused hyperplasia by blocking glomerulosa cells transdifferentiation into fasciculata cells3. However, in contrast to Senp2cKO mice, zF cells never became atrophic in β-catenin gain-of-function. It was thus proposed that when physiological transdifferentiation is unachievable, the zF can be maintained by an alternative cellular pathway involving progenitor cells that bypass the zG state2. Here, we report the first occurrence of β-catenin SUMOylation associated with its nuclear translocation which takes place ectopically in Senp2 -deficient zF. This leads to moderate WNT target genes activation and together with blunted PKA activity, impairs acquisition of fasciculata identity. In this context where zG-to-zF zonal transdifferentiation is impaired, the deficit in PKA signalling alters the phosphorylation profile of the major mitochondrial fission GTPase, DRP1, leading to premature apoptosis of the neoformed mutant fasciculata cells, reinforcing the atrophy of the zone over time. The block in zonal transdifferentiation induced by zG-targeted stabilised β-catenin has been ascribed to increased rosette formation (the basal cellular organisation of zG cells forming flower-like structures) and/or impaired rosette resolution (the moment when zG cells exit the rosette to form a single column of cells typical of zF)54. Conversely, adrenal cortex lacking Senp2 do not exhibit zG hyperplasia and thus should not display enhanced rosette formation. Our data indicate that mechanisms by which zG undergoes resolution during normal homeostatic turnover could be controlled by SUMOylation. Identifying what targets of Senp2 -dependent SUMOylation directly contribute to zG resolution and whether this is coupled to zF transdifferentiation will require extensive studies.
SUMOylation has already proven a role in adrenocortical development with the unSUMOy-latable SF-12KR/2KR model, which results in adrenal cells expressing gonadal markers21. In ovaries which share a common foetal origin with adrenals, SUMOylation by the E3 SUMO ligase activity of TRIM28, is essential for maintenance of granulosa cell fate at the expense of Sertoli-like identity55, whereas SENP1 presence in stromal cells is necessary for proper oocyte maturation and ovulation56. From a broader perspective, our study belongs to a growing corpus of evidence associating SUMOylation with coordination of differentiation in different tissues and cell types such as white and brown adipose tissue15,17,57, induced pluripotent stem cells16, or intestine18. Seemingly counter-intuitively, the induction of the adrenal phenotype of Senp2cKO mice does not rely on an overall increase in SUMO conjugates levels but rather on specific overSUMOylation of certain cell populations (Figure S1.I.J.) or substrates belonging to signalling pathways crucial for adrenal homeostasis. These findings are consistent with recent studies exploring the consequences of increased SUMOylation in the uterine stroma (by Senp1 deletion) or decreased SUMOylation (by Ubc9 haploinsufficiency) in the intestine of Apc mutant mice that showed strong phenotypes in the absence of global changes in visible SUMO conjugate levels58,59. This illustrates the great versatility of this posttranslational modification pathway. Changes in SUMOylation capacity can have specific effects in vivo, especially when targeted steps are catalysed by various members of the same enzymatic family, such as E3 SUMO-ligases or SUMO-specific proteases, which can therefore differ in their substrate specificity. With this in mind, we speculate that specific over-SUMOylated substrates produced in Senp2 -deficient adrenals should result from the primary loss of the SUMO-protease. Their SUMOylation could be reinforced by the unrepressed (hypo-phosphorylated) E3 SUMO-ligase activity of TRIM2860,61, secondary to the blunted ACTH/PKA signalling.
Overall, the present paper demonstrates that preventing deSUMOylation by adrenocortical specific SENP2 ablation induces zF hypoplasia associated with increased premature apoptosis along the lineage conversion zG-zF process, ultimately resulting in a blockage of the physiological differentiation process, translating to isolated glucocorticoid deficiency. This is linked to dysregulation of WNT/PKA balance in favour of WNT signalling. This shift highlights the central role of SUMOylation in physiological processes such as differentiation, tissue maintenance, and stress response. Furthermore, our data could broaden the scope of expected impacts of SUMOylation alterations to endocrine pathogenesis. Although pathogenic associations have been made with alterations in SUMO enzymes or substrates, direct causal links to pathologies have very rarely been established13,62. Our work suggests that genetic alterations leading to excessive SUMOylation could be associated with isolated glucocorticoid deficiency in patients.
Methods
Cell culture
Adrenocortical tumour cell line 7 (ATC7) cells were established from an adrenal tumour derived from a mouse expressing the Simian Virus 40 large T (SV40 T) antigen under the control of the aldo-keto reductase 1b7 (Akr1b7) gene promoter specific to the adrenal cortex36,63. Cells were cultured on poly-D-lysine–coated 10 cm cell culture dishes (MilliporeSigma, Burlington, MA) in a DMEM-F12 medium (Thermo Fisher Scientific, Waltham, MA) at 37°C in the presence of 5% CO2, insulin (10 mg/mL), transferrin (5.5 mg/mL), selenium (6.7 ng/mL) (Thermo Fisher Scientific), L-glutamine (2 mM), penicillin 0.1 U/mL), streptomycin (0.1 mg/mL), 2.5% horse serum, and 2.5% foetal calf serum. Cells were seeded in 12-well plates and cultures to subconfluence and then starved by replacing medium by serum-free medium the day before the addition of forskolin (Sigma-Aldrich), staurosporine (Sigma-Aldrich) or Mdivi-1 (Merck) at the times and concentrations indicated in the figure’s legends.
Mice
Mice bred in-house and maintained on a mixed sv129-C57Bl/6 genetic background were housed on a 12-hour light/12-hour dark cycle (lights on at 7:00 am). Mice were fed normal, commercial rodent chow and provided with water ad libitum. After weaning, mice were kept in siblings with a maximum of 4 animals per cage.
Hormonal measurements
Mice were killed by decapitation around 8:30 am and blood was collected in vacuum blood collection tubes (VF-053STK, Terumo). For ACTH treatments mice were injected intraperitoneally with 0.05 mg/30g Synacthene (0.25 mg/mL, Novartis, Basel, Switzerland) 2 hours or 30 minutes prior the euthanasia. ACTH response kinetics in Figure4.I., was done with collection of the blood from the tail of the mice at 8 am, 10 am and 12 pm.
Corticosterone was measured from serum with ELISA kit (AR E-8100, LDN), ACTH was measured with Milliplex Map Kit (MPTMAG-49K, Millipore) and other steroids were measured by LC-MS/MS64.
Histology and Proximity ligation assay
Adrenals were fixed in 4% PFA for 6 hours and embedded in paraffin. 5µm sections were deparaffinised and processed for H&E. For immunohistochemistry or immunofluorescence, deparaffinised slides were submerged in antigen retrieval buffer and microwaved for 8 minutes. After being rinsed with 1X PBS, they were blocked for an hour with 2.5% horse serum (Vector) and incubated overnight at 4°C with primary antibody. After rinsing, they were incubated with ImmPRESS polymer for 30 minutes at room temperature. HRP activity was detected with NOVAred (Vector) or Alexafluor (Thermo Fisher). Primary antibodies are listed in supplementary table S1.
For PLA, blocked slides were incubated overnight at 4°C with indicated antibodies followed by Duolink in situ PLA (Sigma-Aldrich) anti-mouse (minus) and anti-rabbit (plus) probes and detection reagents according to manufacturer’s instructions.
Images were acquired with Zeiss Axioscan Z1 or Zeiss Imager M2 and analysed with QuPath software65.
RT-qPCR
Adrenal glands were removed, flash frozen on dry ice, and RNA was extracted using RNeasy micro kit from QIAGEN. After reverse transcription, PCR reaction was conducted using SYBR qPCR Premix Ex Taq II Tli RNase H+ (TAKRR820W, Takara). Primer pairs are listed in supplementary table S2.
RNA-Seq
For each sex, adrenal gene expression profiles for four 4-week-old Sf1-Cre/+::Senp2fl/fl and four WT littermates were analysed using RNA-seq. RNA-sequencing, library generation and differential expression genes analysis were performed by the GenomEast platform (IGBMC, Illkirch, France).
Image analysis and base calling were performed using RTA 2.7.3 and bcl2fastq 2.17.1.14. Adapter dimer reads were removed using DimerRemover Reads were mapped onto the mm10 assembly of Mus musculus genome using STAR version 2.5.3a. Gene expression quantification was performed from uniquely aligned reads using htseq-count version 0.6.1p1. Read counts have been normalised across samples with the median-of-ratios method proposed by Anders and Huber. Differential expression have been implemented using the Bioconductor package DESeq2 version 1.16.1. Raw and processed data have been deposited in NCBI’s GEO database (GSE193480).
Gene Set Enrichement Analysis was performed using the GSEA software66 and plotted using the replotGSEA function from the Rtoolbox package (https://github.com/PeeperLab/Rtoolbox). Gene ontology analysis was performed using g:Profiler67. Data visualisation was carried out using R software (v4.1.0)68, Pheatmap package was used for heatmaps, Vennerable for Euler diagrams and ggplot2 for plots. PCA analysis was produced on read counts matrix using prcomp function from the stats package and plotted using ggplot2.
Western blot and immunoprecipitation
Proteins were extracted from snap frozen adrenals in RIPA buffer (TRIS 25 mM, EDTA 1 mM, MgCl2 5 mM, NP40 1%, glycerol 10%, NaCl 150 mM supplemented extemporaneously with phosphatase inhibitors (1 mM Na3VO4, 0.5 mM NaF), protease inhibitors (Roche, Basel, Switzerland), and SUMO proteases inhibitor N-ethylmaleimide (MilliporeSigma) (3.13 mg/mL).
For western blot, thirty µg of total denatured proteins were loaded on 10% SDS-page gel, transferred on nitrocellulose and detected with primary antibodies (Supplementary table 2). Signals were quantified with ChemiDoc MP Imaging System camera system (Bio-Rad) and Image Lab software (Bio-Rad). Expression of phosphorylated or SUMOylated proteins was normalised to the expression of the corresponding unmodified protein.
Immunoprecipitation was carried out using Automag Solution (AdemTech). 500-1000 µg of total proteins were precleared with 50 µL of beads for 30 minutes at room temperature. 10 µL of antibodies (Supplementary table 2) were crosslinked with 50 µL of beads in 20 mM DMP for 30 minutes. Precleared samples were immunoprecipitated with crosslinked antibodies for 60 minutes at room temperature, washed thrice in RIPA buffer and eluted with 50 µL of 50 mM glycine pH3. pH was brought back to neutral with 1 µL of TRIS buffer pH9 and samples were denatured in Laemmli buffer (Bio-Rad) at 95°C for 5 minutes and loaded for SDS-PAGE.
PKA activity
Proteins were extracted from snap frozen adrenals in lysis buffer (MOPS 20 mM, Betaglycerol-phosphate 50 mM, NP40 1%, DTT 1 mM, EDTA 2 mM, EGTA 5 mM supplemented extemporaneously with phosphatase inhibitors (1 mM Na3VO4, 50 mM NaF), protease inhibitors (Roche, Basel, Switzerland). HA-SUMO vinyl sulfone (R&D Systems, Minneapolis, MN) was added for the SUMO-VS condition to the concentration of 5 µM for both SUMO1 and SUMO2.
10 µg of protein from 4-week-old adrenals were used for measurement with PKA kinase activity kit (ab139435, Abcam).
Statistics
Statistics were conducted using R language68 and Comp3Moy function from sumo package (https://github.com/Damien-Dufour/sumo). Normality of populations distribution was assessed with Shapiro & Wilk test for n ∈ [7,5000] or otherwise Kolmogorof-Smirnov normality test.
If data followed a normal distribution, homoscedasticity was estimated with a Barlett test. To compare two populations, unpaired, two tailed t-test was used for normally distributed data with the same variance, Mann-Whitney for non normal distributions and Welch t-test for normally distributed data but with different variances. To compare three or more distributions : one way ANOVA for normally distributed samples with pairwise multiple t.tests or Kruskal-Wallis for non normally distributed samples with planned comparisons using Dunn’s test to determine the genotype effect or the treatment effect
Crosses on the violin plots represent the mean and lines represent the median. Error bars in barplot represent the SD unless otherwise stated. The number of samples per condition is indicated at the bottom of each plot. Corticosterone response to ACTH in Figure 4.I. was analysed with paired 2 way ANOVA to compare the effect of treatment for each genotype. Correlations in Figure 5.F. and 6.A. have been conducted using cor.test function from stat package and Kendall method as distributions did not follow normality and contained ties.
Ethical approval declarations
Mouse experiments were conducted according to French and European directives for the use and care of animals for research purposes and were approved by the Comité d’Éthique pour l’Expérimentation Animale en Auvergne (project agreement #21152-2019061912052883), C2EA-02, at Institut National de Recherche pour l’Agriculture, l’Alimentation et l’Environnement, Research Centre Clermont-Theix, France (C2E2A).
Data availability
Sequencing data has been deposited in GEO with the accession code GSE193480. All other data that supports the findings of this study is provided in the article or supplementary data. Source data is provided with this paper.
Code availability
All code for data cleaning and analysis associated with the current submission is available upon request.
Supplementary information
Authors’ contributions
DD, TD, GB and AM designed research. DTB and ETH provided ASCre-expressing and Senp2 -floxed mice, respectively. DD, ISB, MO, EP, JJW, JO, CL, AL, CSD, JCP and FRB performed experiments. DD and AM analysed data. DD and AM wrote the paper. GB, IT, DTB, PV and AMLM edited the paper.
Funding
This work was funded through institutional support from CNRS, INSERM, Université Clermont-Auvergne, the French government IDEX-ISITE initiative 16-IDEX-0001 (CAP 20-25), and grants from Ministère de l’Enseignement Supérieur, de la Recherche et de l’Innovation (to DD), Société Française d’Endocrinologie (to DD and AM), Fondation Association pour la Recherche sur le Cancer (to JW and JO), and Agence Nationale pour la Recherche (ANR-18-CE14-0012-02 Sex-Specs to AL and AM).
Competing interests
The authors declare no conflict of interest.
Supplementary figures
Acknowledgements
We thank K. Ouchen, S. Plantade, and P. Mazuel for animal care, A. Dehaze and J.P. Saru for their technical assistance, and Y. Renaud for management of the bioinformatic platform.