Cdkn2a/p16^Ink4a Regulates Fasting-Induced Hepatic Gluconeogenesis Through the PKA-CREB-PGC1α Pathway

Type 2 diabetes (T2D) is a complex metabolic disorder involving a combination of insulin resistance, impaired insulin secretion, and increased hepatic glucose production (1,2). The pathogenesis of T2D is multifactorial, involving both genetic and environmental susceptibility factors (3). During these last few years, the search for genetic determinants of T2D greatly progressed, identifying new loci contributing to T2D. A better understanding of the function of the gene products of these loci is required to identify new strategies for the prevention and treatment of T2D (3,4). Hence, recent human genome-wide association studies (GWAS) have identified a polymorphism on chromosome 9p21 (rs10811661), located ~125 kb upstream of the CDKN2B and CDKN2A genes, that is strongly and reproducibly linked to T2D (5–7), establishing genes on the CDKN2A/B locus among the strongest candidates for conferring susceptibility to T2D across different ethnicities (4).

The gene products are the cyclin-dependent kinase (CDK) inhibitors p16^Ink4a and p14^ARF for the CDKN2A locus and p15^Ink4b for the CDKN2B locus, which are tumor suppressors acting as cell cycle inhibitors (8,9). The p15^Ink4b and p16^Ink4a proteins bind to either CDK4 or CDK6, thus inhibiting the action of cyclin D and preventing retinoblastoma protein phosphorylation and subsequent release of the E2F1 transcription factor. As a consequence, the transcription of genes required for cell cycle progression to the S phase is restrained.

However, how the CDKN2A/B gene products modulate glucose metabolism is less clear. In murine models, increased expression of p15^Ink4b in pancreatic islets is associated with islet hypoplasia and impaired glucose-induced insulin secretion (10). Moreover, p16^Ink4a plays a crucial role in senescence and aging. p16^Ink4a expression increases with age in pancreatic β-cells and promotes an age-dependent decline in islet regenerative potential (11). Additionally, other cell cycle regulators, like CDK4, E2F1, and cyclin D, also play roles in glucose homeostasis through actions in the pancreas, muscle, and/or adipose tissue (12–16). However, whether the CDKN2A/B gene products modulate hepatic glucose production is unknown (17).

Glucose homeostasis is determined by the balance of its production and utilization. Impaired postprandial glucose control and the persistence of fasting hyperglycemia are hallmarks of T2D (18). Increased rates of hepatic glucose production are a major cause of fasting hyperglycemia in T2D patients (1). In physiological conditions, during prolonged fasting, hepatic gluconeogenesis is a major pathway for the maintenance of normal plasma glucose levels (19) owing to the action of different hormones, among which are glucagon and glucocorticoids, like cortisol. During starvation, low blood glucose levels induce pancreatic α-cell glucagon secretion and hypothalamic-pituitary-adrenal axis activation. In the liver, glucagon binds to its receptor, which then causes a GDP/GTP exchange, hence stimulating adenylate-cyclase activity, which converts ATP into cAMP (20). The rise in intracellular cAMP levels stimulates the dissociation of the catalytic and regulatory subunits of protein kinase A (PKA) (21). The catalytic PKA subunit then enters the nucleus where it phosphorylates the CREB at Ser¹³³, converting it into its transcriptionally active form, which induces gluconeogenic gene expression (22–24). In concert, glucocorticoids activate the glucocorticoid receptor, which binds to glucocorticoid-responsive elements in the promoters of gluconeogenic genes (25,26).

Given the strong association of the CDKN2A/B locus with T2D risk, which in large population studies is mainly established by the measurement of fasting hyperglycemia (5), we set out to study whether p16^Ink4a plays a role in hepatic glucose homeostasis using p16^Ink4a-deficient mice (p16^−/−), mouse primary hepatocytes, and mouse hepatic cell line. Our results identify p16^Ink4a as a modulator of the PKA-CREB–peroxisome proliferator–activated receptor γ coactivator (PGC1α) signaling pathway and, hence, as a regulator of fasting hepatic glucose homeostasis, independent of its function as cell cycle regulator.

p16^−/− and littermate control (p16^+/+) mice on a C57Bl6 background (>97%) were housed under standard conditions in conventional cages with free access to water and food unless indicated otherwise. Twelve-week-old male mice were killed by cervical dislocation at 9:00 a.m. after a 24-h fasting. Experimental procedures were conducted with the approval of the ethics committee for animal experimentation of the Nord Pas-de-Calais region (CEEAA022008R).

Overnight fasted mice (5:00 p.m. to 9:00 a.m.) were injected with sodium pyruvate (P4562; Sigma) (2 g/kg body wt i.p.). Blood glucose levels were measured from the tail vein at the indicated time points using an automatic glucose monitor (OneTouch; LifeScan).

Mice were anesthetized with a mixture of ketamine (100 mg/kg) and xylasine (20 mg/kg) administered intraperitoneally. Livers were perfused in situ through the inferior cava vein, with Hanks’ balanced salt solution (H9394; Sigma) containing 0.5 mmol/L EGTA and 50 mmol/L HEPES followed by Hanks’ balanced salt solution containing 0.025% collagenase (C5138; Sigma) until loss of its firm texture. The soft liver was removed and cut into pieces and the homogenate filtered and centrifuged for 2 min. The pellet was washed three times and resuspended in Williams medium supplemented with 0.1% BSA, 1% glutamine, 1% gentamycine, 100 nmol/L insulin, and 100 nmol/L dexamethasone. Cell number and viability were assessed using trypan blue. Cells were plated on six-well plates during 2 h for hepatocyte selection and then incubated in deprivation medium (1% penicillin-streptomycin, and 1% glutamine, with distinct concentrations of glucagon [0, 1, 10, and 100 nmol/L]) for 6–8 h (for RNA measurements) or 30 min (for protein analysis).

Alpha Mouse Liver 12 (AML12) (cat. no. CRL2254; American Type Culture Collection) cells were cultured in deprivation medium–Ham's F-12 supplemented with 10% FBS (Invitrogen), 5 g/mL insulin (Sigma), 5 g/mL transferrin (Sigma), 5 ng/mL selenium (Sigma), 1% glutamine, and 1% penicillin-streptomycin and maintained at 37°C under 5% CO₂. AML12 cells were transfected with small interfering RNA (siRNA) for CDKN2A (043107-00-005; Thermo Scientific [ON-TARGET plus SMART pool siRNA]), CDK4 (ON-L-040106-00-0005; Thermo Scientific [ON-TARGET plus SMARTpool siRNA]), or control (D-001810-10-20; Thermo Scientific [ON-TARGET plus nontargeting pool siRNA]) using the Dharmafect1 reagent (Thermo Scientific) according to the manufacturer’s instructions. AML12 cells were treated for the indicated times points with 10 µmol/L forskolin.

Primary hepatocytes were cultured in six-well plates in Williams medium with 0.1% BSA, 100 nmol/L dexamethasone, 1% penicillin-streptomycin, and 1% glutamine. After 2 h, the medium was replaced with 1 mL glucose-production buffer consisting of glucose-free Krebs-ringer buffer (115 mmol/L NaCl, 5.9 mmol/L KCI, 1.2 mmol/L MgCl₂, 1.2 mmol/L NaH₂PO₄, and 2.5 mmol/L NaHCO₃ pH 7.4) without phenol red, supplemented with 15 mmol/L sodium lactate and 1 mmol/L sodium pyruvate. Glucose concentrations were measured at different time points with a colorimetric glucose assay kit (Sigma). The values were then normalized to total protein content determined on whole-cell lysates.

Liver total RNA was isolated using the guanidinium isothiocyanate phenol/chloroform extraction method, and total RNA from cultured cells was extracted using the TRIzol reagent (Eurobio). One microgram of total RNA was reverse transcribed to cDNA using the High-Capacity cDNA Reverse Transcription kits (Applied Biosystems) according to the manufacturer's instructions. Reverse transcribed cDNAs were quantified by Brilliant III Ultra-Fast SYBR green-based real-time PCR using specific oligonucleotides (Supplementary Table 1) on a Stratagene Mx3005P (Agilent Technologies) apparatus. mRNA levels were normalized to Cyclophilin A expression as an internal control, and mRNA fold induction was calculated using the comparative Ct (2^−∆∆Ct) method.

AML12 cells and mouse primary hepatocytes were lysed with cell lysis buffer (50 mmol/L Tris-HCl, pH 8; 137 mmol/L NaCl; 5 mmol/L Na₂EDTA; 2 mmol/L EGTA; 1% Triton; 20 mmol/L sodium pyrophosphate; 10 mmol/L β-glycerophosphate; 1mmol/L Na₃VO₄; 10 µmol/L leupeptin; and 5 µmol/L pepestatin A) (Sigma-Aldrich) on ice. Cells were scraped and transferred to 1.5-mL Eppendorf tubes and rotated for 30 min at 4°C, followed by centrifugation at 13,000g for 10 min at 4°C. The resulting supernatants were stored in aliquots at −80°C until they were required. Protein concentration in the cell lysates was determined using a BCA protein assay kit (Pierce). The cell lysates were mixed with 4X-SDS sample buffer NOVEX (Life Technologies). Samples were heated at 100°C for 10 min before loading and being separated on precasted 4–12% or 3–8% SDS-PAGE (Invitrogen). Proteins were electrotransferred to a nitrocellulose membrane (Millipore, Bedford, MA) in 1X transfer buffer (Invitrogen) using the Nupage Systeme for 1 h at 30 V. Nonspecific binding to the membrane was blocked for 1 h at room temperature with 5% nonfat milk in Tween–Tris-buffered saline (TTBS) buffer (20 mmol/L Tris, 500 mmol/L sodium NaCl, and 0.1% Tween 20). Membranes were then incubated overnight at 4°C with various primary antibodies in blocking buffer containing 5% nonfat milk at the dilution specified by the manufacturers. The following primary antibodies were used: phospho-CREB (Ser¹³³) (9198; Cell Signaling Technology), CREB (9197; Cell Signaling Technology), phospho–(S/T)-PKA substrates (9621; Cell Signaling Technology), phospho-pRb (3590; Cell Signaling Technology), pRb (9313; Cell Signaling Technology), PGC1α (sc-13067; Santa Cruz Biotechnology), GAPDH (sc-25778; Santa Cruz Biotechnology), p16^ink4a (sc-1207; Santa Cruz Biotechnology), phospho–regulatory subunit 2 of PKA (PKAR2) (ab32390; Abcam), and PKAR2 (ab-38949; Abcam). Membranes were then incubated with the secondary antibody conjugated with the enzyme horseradish peroxidase. The visualization of immunoreactive bands was performed using the enhanced chemiluminescence plus Western blotting detection system (GE Healthcare). Quantification of phospho-CREB level in mouse primary hepatocytes and AML12 cells was performed by volume densitometry using the ImageJ 1.47t software (National Institutes of Health).

Intracellular cAMP concentrations were measured using a ready-to-use competitive enzyme immunoassay kit (R&D Systems). Briefly, cells were lysed according to the manufacturer’s protocol, and 100 μL sample was mixed with 50 μL cAMP conjugated and then added to cAMP-specific antibody precoated microplate. After 2 h of incubation at room temperature, substrate solution was added for 20 min. Color development was stopped, and the absorbance at 450 nm was measured using a Dynex MRX TC Revelation Microplate Reader. PKA activity was measured by the signaTECT cAMP-Dependent Protein Kinase Assay System by using the Kemptide (LRRASLG) as a peptide substrate.

Coimmunoprecipitation of CDK4 from whole AML12 cell extracts was performed using the Thermo Scientific Pierce Crosslink Magnetic IP/Co-IP kit. Briefly, 48 h after siRNA transfection, cells were lysed and 500 μg total protein extract was incubated with 3 μg CDK4 antibody (sc-260; Santa Cruz Biotechnology) according to the manufacturer's protocol. The eluate was then subjected to Western blot analysis using PKAR2 (ab-38949; Abcam) and CDK4 (sc-260; Santa Cruz Biotechnology).

Cells were grown on cover slips. At 4 8h after siRNA transfection, cells were washed with PBS and fixed with 4% paraformaldehyde for 20 min. After fixation and permeabilization with 0.1% TRITON, cells were incubated overnight with antibodies against p16^ink4a (M-156, sc-1207; Santa Cruz Biotechnology) and phospho-PKAR2 (Ab-32390; Abcam) and subsequently incubated with a combination of Texas red–conjugated anti-rabbit IgG and FITC-conjugated anti-mouse IgG. A nuclear DAPI counterstain was also performed.

Data are expressed as means ± SEM. Results were analyzed by unpaired two-tailed Student t test or one-way ANOVA with least significant difference (LSD) Fisher post hoc test or two-way ANOVA with LSD Fisher post hoc test as appropriate using GraphPad Prism software. A P value of < 0.05 was considered statistically significant.

Since GWAS revealed an association between the CDKN2A/B locus and T2D risk, primarily based on the fasting plasma glucose trait, we first measured fed and fasted blood glucose levels in 12-week-old mice. p16^−/− mice displayed a less pronounced hypoglycemia after 24 h of fasting compared with p16^+/+ mice (Fig. 1A). This effect was not due to differences in plasma glucagon levels between fasted p16^+/+ and p16^−/− mice (Supplementary Fig. 1). For evaluation of whether gluconeogenesis was influenced, a pyruvate tolerance test (PTT) was performed in fasted p16^−/− and p16^+/+ mice. Interestingly, p16^−/− mice produced higher blood glucose levels, upon pyruvate administration, suggesting an increased hepatic glucose production (Fig. 1B and C). Consistent with this, hepatic mRNA levels of gluconeogenic genes, such as G6pase and Pepck, were significantly higher in livers of fasted p16^−/− versus p16^+/+ mice (Fig. 1D and F), while Fbp1 mRNA was not different between the genotypes (Fig. 1E). Conversely, genes involved in other metabolic pathways regulated during fasting, such as glycolysis (Gk, Lpk) and β-oxidation (Cpt1a, Lcad), were not differently expressed between both genotypes upon fasting (Fig. 1G, H, J, and K), although mRNA levels of Pdk4, which block glycolysis at the level of pyruvate dehydrogenase, tended to be higher in fasted p16^−/− livers (Fig. 1I). Altogether, these data indicate that among the different hepatic metabolic pathways regulated by fasting, gluconeogenesis is the only one modulated in p16^−/− mice.

Since p16^Ink4a is a tumor suppressor and a cell cycle regulator and since hepatic proliferation and tumor growth may perturb glucose homeostasis, we investigated whether p16^Ink4a deficiency is associated with spontaneous liver tumor growth or altered hepatocyte proliferation in our experimental conditions. At the age of 12 weeks, p16^−/− mice did not display macroscopic liver abnormalities or differences in liver weight compared with p16^+/+ mice (Supplementary Fig. 2A and B). Moreover, immunohistochemical Ki-67 staining of liver sections showed no differences between p16^−/− mice and their littermate controls under fasting conditions (Supplementary Fig. 2C–H), indicating that hepatocyte proliferation is not different. These data indicate that p16^Ink4a deficiency increases fasting-induced hepatic gluconeogenesis in vivo, independent of any action on hepatocyte proliferation.

For analysis of whether the altered regulation of hepatic gluconeogenic gene expression in p16^−/− mice is a cell-autonomous phenomenon, primary hepatocytes from p16^−/− mice and their littermate controls were isolated and incubated with increasing concentrations of glucagon to mimic the fasting conditions. Basal levels of gluconeogenic gene expression were 1.5-fold higher for G6Pase (±0.15; P < 0.05), 4.4-fold higher for Pepck (±0.96; P < 0.01) and 2.3-fold higher for Fbp1 (0.17; P < 0.001) in p16^−/− compared with p16^+/+ primary hepatocytes (Fig. 2A–C). Moreover, glucagon, which activates the PKA-CREB signaling pathway, more pronouncedly induced G6pase, Pepck, and Fbp1 (Fig. 2A–C) mRNA levels in p16^−/− vs. p16^+/+ primary hepatocytes. Further, hepatic glucose production was higher in primary hepatocytes of p16^−/− than of p16^+/+ mice (Fig. 2D). Next, p16^Ink4a was silenced using a CDKN2A siRNA (which affects both p16^Ink4a and p19^ARF expression) in AML12 cells (Fig. 3A and B), a mouse hepatocyte cell line that expresses very high levels of p16^Ink4a compared with liver and primary hepatocytes (Fig. 3C). Incubation with forskolin, to activate the PKA-CREB pathway, resulted in a more pronounced increase of G6pase and Fbp1 gene expression when p16^Ink4a was silenced, while no effect was observed on Pepck gene expression (Fig. 3D–F). Moreover, although G6pase and Fbp1 gene expression only marginally increased upon forskolin treatment in p16^Ink4a-expressing AML12 cells, p16^Ink4a silencing resulted in the restoration of a strong response (Fig. 3D and E). Altogether, these results indicate that p16^Ink4a expression levels influence the response to fasting-induced stimuli both in vivo and in vitro.

For studying of the mechanism by which p16^Ink4a regulates gluconeogenic gene expression, mRNA and protein levels of PGC1α, a master regulator of the fasting adaptation process (24), were measured. The fasting response of Pgc1a mRNA was significantly more pronounced in livers of p16^−/− compared with p16^+/+ mice (Fig. 4A). In line, p16^−/− primary hepatocytes displayed a 3.4-fold increased Pgc1a mRNA level (±0.83; P < 0.01, two-tailed Student t test) and a stronger induction by glucagon compared with p16^+/+ hepatocytes (Fig. 4B). p16^Ink4a silencing in AML12 cells significantly increased Pgc1a expression at both mRNA and protein levels upon forskolin treatment (Fig. 4C and D).

To gain insight into how PGC1α is induced upon p16^Ink4a-deficiency, we first analyzed the phosphorylation status of CREB, a transcription factor inducing PGC1α expression. p-Ser¹³³-CREB was markedly higher in p16^−/− compared with p16^+/+ hepatocytes at the basal level as well as after glucagon exposure (Fig. 5A and B). Similar results were obtained upon forskolin treatment (data not shown). Likewise, p16^Ink4a silencing in AML12 cells resulted in a stronger CREB phosphorylation both at the basal level and upon forskolin treatment (Fig. 5C and D). Altogether, these data demonstrate that p16^Ink4a knockdown increases CREB phosphorylation. It is well-known that the cAMP-PKA signaling pathway regulates fasting-induced CREB phosphorylation (22,27). To test whether alterations in PKA activity may explain the increased CREB phosphorylation upon p16^Ink4a deficiency, p16^Ink4a-silenced AML12 cells were treated with H89, a specific PKA inhibitor. H89 treatment prevented CREB phosphorylation induced by p16^Ink4a silencing (Fig. 6A). Accordingly, PKA activity in p16^Ink4a-silenced AML12 cells was 1.5-fold higher compared with control (Fig. 6B). This increase was substantiated by the increase in total PKA substrate phosphorylation profiles upon p16^Ink4a silencing (Fig. 6C). Likewise, several PKA substrates were more phosphorylated in p16^−/− than in p16^+/+ primary hepatocytes both under basal conditions and after glucagon stimulation (Fig. 6D). Since PKA activity is controlled at least in part by the phosphorylation of its regulatory subunits (PKAR2), the expression and phosphorylation of PKAR2 were measured in p16^Ink4a-silenced AML12 cells and in p16^−/− primary hepatocytes. p16^Ink4a silencing or deficiency resulted in increased PKAR2 phosphorylation in AML12 cells (Fig. 7A) and in p16^−/− primary hepatocytes both at the basal state and upon glucagon stimulation (Fig. 7B). This result was confirmed by the enhanced p-PKAR2 immunostaining in p16^Ink4a-silenced AML12 cells (Fig. 7C). Noteworthy, the increased PKA activity was not due to an increase in cAMP levels (Fig. 7D). Collectively, these data demonstrate that p16^Ink4a-deficiency activates the PKA-CREB-PGC1α signaling pathway independent of changes in intracellular cAMP levels. To understand the underlying mechanism by which p16^Ink4a increases phosphorylation of PKAR2 and thereby the increase of gluconeogenic genes, we investigated the involvement of CDK4, a well-known target of p16^Ink4a. Silencing of CDK4 in p16^Ink4a-silenced AML12 cells (Fig. 8A and B) abrogated the induction of Ppc1α and Fbp1 mRNA levels by p16^Ink4a silencing (Fig. 8C and D). Moreover, coimmunoprecipitation experiments in AML12 cells after p16^Ink4a knockdown demonstrated a physical interaction between CDK4 and PKAR2 (Fig. 8E).

In recent years, a growing body of evidence supports the emerging notion that cell cycle regulatory proteins contribute to metabolic processes in addition to, or linked with, their role in cell growth (17,28). Today, these proteins are perceived as sensors of external signals that require a particular adapted metabolic response. The CDK-Rb-E2F1 pathway, which is inhibited by p16^Ink4a, has already been shown to control adipogenesis by modulating the expression of the nuclear receptor PPARγ (15,29), a master regulator of adipogenesis, as well as by controlling oxidative metabolism in adipose tissue (30). The CDK-Rb-E2F1 pathway is also a negative regulator of energy expenditure through repression of mitochondrial oxidative metabolism in muscle (16). Disruption of CDK inhibitor genes in the mouse has not revealed profound cell cycle abnormalities but does result in a specific metabolic phenotype. Mice lacking p18^Ink4c (31,32), p21^cip1, or p27^Kip1 display growth abnormalities and adipocyte hyperplasia (33). Double knockout mice (p21^−/−; p27^−/−) develop hypercholesterolemia, glucose intolerance, and insulin insensitivity (33). Surprisingly, the role of these cell cycle regulators in the liver, one of the main metabolic organs controlling glucose homeostasis, has not yet been demonstrated.

It is well known that an increased rate of hepatic gluconeogenesis contributes to fasting hyperglycemia observed in T2D patients. Genetic analysis in GWAS identified an association of the CDKN2A/B locus with T2D risk (5,34,35). The association is based on the measurement of fasting glycemia and confers to the CDKN2A/B locus a high susceptibility to T2D across different ethnicities. In this study, we tried to elucidate the mechanism by which a product of CDKN2A/B, i.e., p16^Ink4a, may influence the hepatic gluconeogenic program and thereby be implicated in T2D pathogenesis. We found that p16^Ink4a deficiency raises PKAR2 phosphorylation leading to an increased PKA activity. The increased PKA activity enhances CREB-PGC1α signaling, regulating the gluconeogenic gene expression program. Since the p16^Ink4a protein shares several ankyrin repeat domains, which are involved in protein-protein interaction, we assessed whether p16^Ink4a may associate with the PKA complex. Immunoprecipitation of endogenous p16^Ink4a in AML12 cells failed to demonstrate an interaction of p16^Ink4a with the PKA regulatory subunit or the PKA catalytic subunit (data not shown), suggesting the existence of other proteins able to connect p16^Ink4a to PKA complex. One good candidate bridging p16^Ink4a to the PKA complex was CDK4, a well-known interaction partner of p16^Ink4a. Indeed, siRNA knockdown of CDK4 in AML12 cells abrogates the effect of p16^Ink4a deficiency on the expression of gluconeogenic genes, like Pgc1α and Fbp1, suggesting a direct role of CDK4 in the regulation of PKA activity by p16^Ink4a. Moreover, it has already been shown that CDK4 can displace the interaction between cyclin D and the PKAR2–A-kinase anchoring protein/AKAP95 complex when CDK4 is activated (36). Other studies have demonstrated that CDK1 also phophorylates PKAR2 (37), suggesting that other CDKs than CDK4 can have the same activity on PKAR2 (38). All of these data support the existence of a dynamic complex including p16^Ink4a-(CDK4/cyclin D)-PKA-AKAP95 involved in the control of hepatic glucose production (Fig. 8F).

In summary, GWAS identified SNPs near CDKN2A/B associate with fasting glycemia and the risk of T2D development. Our study establishes that the p16^Ink4a gene product of this locus modulates hepatic glucose production by increasing hepatic gluconeogenic gene expression. Further, we provide evidence that p16^Ink4a acts via the PKA-CREB-PGC1α signaling pathway. Although the functional role of several cell cycle regulators (CDK4, pRb, E2F, p21^cip1, p27^kip1, and p18^Ink4c) in metabolic control has been described in tissues such as adipose tissue and the pancreas (28), this is the first study that demonstrates a role of a cell cycle regulator, p16^Ink4a, in the liver, a master organ regulating glucose homeostasis in a manner independent of its function in cell proliferation. Thus, altered p16^Ink4a activity may contribute to the association between the GWAS locus and the risk of developing T2D.

Acknowledgments. The authors thank P. Krimpenfort for providing the p16-deficient mice and J. Dumont for her assistance.

Funding. This work was supported by grants from the European Genomic Institute for Diabetes (ANR-10-LABX-46). The authors also thank Cost Action (BM0602), Conseil régional Nord Pas-de-Calais, and Fonds Européens de Développement Régional (FEDER). K.B. was supported by a postdoctoral fellowship from Fondation pour la Recherche Médicale (FRM). S.-A.H. was supported by a doctoral fellowship from Université Lille 2/Conseil régional Nord Pas-de-Calais and a FRM grant (FDT20130928340).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. K.B. and S.-A.H. performed experiments, designed experiments, analyzed data, and wrote the manuscript. S.C.-H., E.V., M.B., A.L., and E.B. performed experiments. R.P., A.T., and B.S. designed experiments, analyzed data, and wrote the manuscript. B.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.