Inhibiting Microglia Expansion Prevents Diet-Induced Hypothalamic and Peripheral Inflammation

Consumption of a high saturated fat diet favors metabolic disorders by causing inflammation in peripheral organs (1–3). High-fat diet (HFD)–induced peripheral inflammation is associated with inflammation in brain areas such as the hippocampus and the hypothalamus (4–8). In particular, inflammation in the mediobasal hypothalamus, including the arcuate nucleus (ARC), happens rapidly, before obesity is established (9). Within the ARC, agouti-related protein (AgRP)– and proopiomelanocortin (POMC)–producing neurons sense and integrate nutrient and hormonal signals to guarantee energy homeostasis (10,11). Hypothalamic inflammation causes changes in ARC neurons and, in turn, is responsible for diet-induced obesity (9,12,13) and insulin resistance in peripheral organs (14). Studies have shown that activation of the inflammatory nuclear factor-κB (NF-κB) pathway and recruitment of proinflammatory microglia in the hypothalamus lead to obesity (9,12,15,16). HFD intake can also induce cell proliferation and eventually neurogenesis in the adult hypothalamus, which in turn modifies the neural circuitry regulating energy balance (17).

We hypothesized that cell proliferation plays a key role in the central inflammatory response to HFD and investigated whether modulation of cell proliferation affects adaptive behavioral and metabolic changes related to caloric overload. Thus, mice were fed a chow diet or switched to a diet high in saturated fat. To control for the impact of fat and carbohydrate content on cell proliferation, chow-fed mice also were switched to a very high-fat, very low-carbohydrate diet (VHFD). Simultaneously, we blocked cell proliferation by centrally delivering the antimitotic drug arabinofuranosyl cytidine (AraC) and evaluated changes in energy balance, microglia activity, and inflammation. These studies revealed that HFD feeding specifically induces hypothalamic microglia expansion and that inhibition of HFD-induced cell genesis blunts food intake and adiposity while preventing hypothalamic and peripheral inflammation.

Experiments were performed according to European Union recommendations (2010/63/EU) after approval from the French Ministry of Agriculture and Fisheries (authorization number 3309004) and the ethics committee of the University of Bordeaux. Seven-week-old male C57BL/6J mice (Janvier Labs, Le Genest-Saint-Isles, France) were individually housed at 22°C with a 12-h light/dark cycle (lights off at 1:00 p.m.). Animals had free access to water and chow (Standard Rodent Diet A03; SAFE, Augy, France) unless stated otherwise (Table 1). Mice were acclimated for 1 week before the start of the study. One week after intracerebroventricular surgery, mice were either maintained on chow or switched to an HFD (D12492; Research Diets, New Brunswick, NJ). Mice were fed the HFD for up to 3 weeks (Supplementary Fig. 1A). In related studies, mice were fed chow or switched to a VHFD (D11111601; Research Diets) for 3 weeks. The VHFD was chosen so that the number of calories from saturated fat was similar to that of the HFD but with limited carbohydrates content, the major source of calories in chow.

Food intake and body weight were recorded daily. Feed efficiency was calculated as (body weight gained/caloric intake) × 100. At the end of the study, mice were either anesthetized and perfused for neuroanatomical analysis or killed by cervical dislocation and tissues and blood collected for molecular and biochemical analysis. The number of animals is detailed in the figure legends.

Eight-week-old mice were anesthetized with ketamine (100 mg/kg) and xylazine (10 mg/kg) and then implanted with a cannula into the lateral ventricle (anteroposterior −0.3 mm from bregma; lateral −1 mm to bregma; dorsoventral −2.5 mm below skull) by using a stereotaxic apparatus (David Kopf Instruments). The cannula was connected to an Alzet osmotic minipump (flow rate 0.11 μL/h for 4 weeks (model 1004; Alzet Charles River) through 120-mm vinyl tubing (inner diameter 0.69 mm) filled with artificial CSF (aCSF) (Tocris Bioscience). The minipumps were filled with aCSF containing 4.5 μg/μL BrdU (Sigma) with or without AraC 15 μg/μL (Sigma) and then primed overnight at 37°C in 0.9% saline.

Analysis was performed as previously described (18).

Proteins from hypothalami were extracted and quantified and Western blots carried out as previously described (19). Nitrocellulose membranes were incubated overnight at 4°C with rabbit anti–phospho-STAT3 (p-STAT3) (Tyr705, 1/4,000; Cell Signaling) or rabbit anti-inhibitor of κBα (IκBα) (1/1,000; Cell Signaling). After washing, membranes were incubated with a horseradish peroxidase–conjugated secondary antibody (goat anti-rabbit 1/2,000; Cell Signaling). Immunoreactive (IR) bands were revealed by using enhanced chemiluminescence (ECL Plus; PerkinElmer) and then exposed to radiographic films (Amersham Hyperfilm ECL; GE Healthcare Life Sciences). Membranes were stripped with 2-mercaptoethanol and then reblotted with rabbit anti-STAT3 (1/4,000; Cell Signaling). Blot quantification was performed with ImageJ software (National Institutes of Health, Bethesda, MD).

Plasma leptin and insulin were measured using ELISA kits from Abcam (Paris, France) and Mercodia (Uppsala, Sweden), respectively. Plasma glucose was measured with a glucose assay colorimetric kit (Abcam). Measurements were performed according to manufacturer instructions. HOMA of insulin resistance index was calculated as follows: (glucose mmol/L × insulin mU/L)/22.5.

Multiplex assays were performed according to manufacturer instructions (Bio-Rad) to quantify levels of plasma interleukin 1β (IL-1β) and tumor necrosis factor-α (TNF-α) (Milliplex MAP Mouse Cytokine/Chemokine Magnetic Bead Panel; Merck Millipore, Guyancourt, France) (5,20). Data were obtained by a Bio-Plex 200 system and analyzed with Bio-Plex Manager software (Bio-Rad).

Hypothalami were homogenized and quantitative real-time PCR reactions and analyses were performed as previously described (18). Reference genes were Nono and Sdha. Primer sequences are listed in Supplementary Table 1.

Resident peritoneal macrophages were harvested by washing the peritoneal cavity with 10 mL of 1× PBS. Cells were centrifuged at 600g, and the cell pellet was suspended in PBS/BSA 0.1% for flow cytometry analysis.

Cells were washed and incubated for 45 min with the following antibodies: anti-CD11b-APC, anti-CD45-PerCP Cy5.5, and anti-CD86-FITC (eBioscience) and anti-CD36-PE (BioLegend, Saint-Quentin-en-Yvelines, France). Cells were washed and suspended in PBS/BSA 0.1% for analysis. Nonspecific, isotype-matched antibodies were used to assess nonspecific binding. A BD Biosciences LSRFortessa multicolor flow cytometer was used to determine antigen expression. For each sample and isotype-matched conjugate, 10,000 events were recorded. Data were analyzed with FlowJo software, and gating for each antibody was determined on the basis of nonspecific binding of appropriate negative isotype-stained control.

Animals were anesthetized with 0.1 mL pentobarbital i.p. and perfused transcardially with cold 0.1 mol/L PBS, pH 7.4, followed by 4% buffered paraformaldehyde. Brains were collected, postfixed at 4°C overnight, and transferred into 30% sucrose solution at 4°C. Thirty-micrometer-thick free-floating coronal sections were cut with a cryostat (Leica) and stored in cryoprotectant medium (30% ethylene glycol, 30% glycerol in Krebs PBS) at −20°C until further processing. Antibodies used are listed in Supplementary Table 2.

Single immunolabelings were performed using similar procedures. Sections were incubated in 3% H₂O₂ in 1× PBS before blocking in 5% normal goat serum (Sigma) and 0.3% Triton X-100 in PBS for 1 h. Sections were incubated overnight with primary antibody at 4°C in blocking buffer. After washes, sections were incubated with secondary antibody in 1% normal goat serum-PBS for 2 h followed by 1-h incubation in a peroxidase-conjugated avidin/biotin complex solution (VECTASTAIN Elite ABC System; Vector Laboratories), and staining was visualized using a DAB Peroxidase (HRP) Substrate Kit (Vector Laboratories). Sections were washed in PBS and coverslipped. For BrdU immunostaining, sections were first incubated in a 3% H₂O₂ solution and then washed before treatment with 2 N HCl for 30 min at 37°C. Sections were then processed as described above. For phenotyping BrdU-labeled cells, a two-step procedure was performed. Sections were first processed for ionized calcium-binding adapter molecule 1 (Iba1), glial fibrillary acidic protein (GFAP), and neuronal nuclei (NeuN) staining and then for BrdU. Sections were finally washed and coverslipped with mounting medium (ProLong Gold; Thermo Fisher Scientific).

For single 3,3′-diaminobenzidine immunodetection, digital images were captured by using a Leica DM5000 microscope. The numerical density of X-IR cells (N_v) or the total number of cells was estimated from counts made by systematic random sampling of every 10th section by a blind observer who used a semiautomatic neuron-tracing system (Neurolucida; MBF Bioscience).

The numerical density of Iba1-IR cells or BrdU-IR cells (number cells/mm³) in the left- and right-side hypothalami was determined for at least three sections per brain containing the nucleus of interest. The number of cells was divided by the sectional volume of the area containing the ARC (from bregma −1.22 to bregma −2.70 mm), the paraventricular nucleus (PVN) (from bregma −0.34 to bregma −1.10 mm), or the ventromedial nucleus (from bregma −1.06 to bregma −2.00 mm). The same area was used for the same region investigated among the experimental groups.

The numerical density of Iba1-IR cells (number cells/mm³) in the hippocampus was evaluated for at least five sections by using 50 × 50 × 10-μm frames at evenly spaced x-y intervals of 300 × 300 μm. N_v was calculated based on Eq. 1:

where Q⁻ is the sum of cells, P is the number of frames, h is the dissector height, and a(fra) is the area of the counting frame. Cells in sharp focus in the uppermost focal plane were disregarded (optical dissector principle).

The total number of BrdU-IR cells in the dentate gyrus was quantified throughout the septotemporal axis according to Eq. 2:

where 1/ssf is the inverse of the section sampling fraction (10). The dentate gyrus was used as the positive control area for cell proliferation.

Soma size of Iba1-IR cells was analyzed with ImageJ software. Images underwent automated thresholding, and the areas occupied by the somas were obtained by performing the analyze particle function and expressed as square micrometers per cell. A soma size >65 μm² was used as an indicator of activated microglia (21).

The phenotype of BrdU-IR cells was analyzed by using a confocal microscope with HeNe and Ar lasers (Leica DM2500 TCS SPE) with a 63× objective (1.4 numerical aperture) and through the z-axis at 0.6-μm intervals. Twenty-nine sections containing the ARC from five animals were analyzed. Within the ARC, 173 BrdU-IR cells and 1,095 Iba1-IR cells were analyzed. Sections were automatically scored for double labeling with ImageJ software.

Values are reported as mean ± SEM. Data were analyzed by unpaired Student t test or by one-way, two-way, or two-way repeated ANOVA followed by Fisher least significant difference post hoc analysis. P < 0.05 denotes statistical significance.

Eight-week-old male chow-fed mice centrally received BrdU with or without AraC for 4 weeks. After 1 week from the surgery, animals either continued to eat chow or were switched to an HFD or a VHFD for 3 weeks (Supplementary Fig. 1A). After 4 weeks of central BrdU infusion, BrdU-IR cells were detected in the parenchyma surrounding the third ventricle and specifically in the ARC of chow-aCSF mice (Fig. 1A and B). Three weeks of HFD consumption doubled the number of BrdU-IR cells in the ARC (Fig. 1A and B), whereas it did not affect cell proliferation in the PVN, ventromedial nucleus, or dentate gyrus (Supplementary Fig. 1B–D). Conversely, feeding with a VHFD did not alter cell proliferation in the ARC (Supplementary Fig. 2A). Infusion of AraC within the lateral ventricle completely abolished (no BrdU-IR cells observed) cell proliferation in both the hypothalamus and the hippocampus, regardless of the diet.

Three weeks of HFD consumption increased weight, caloric intake, and fat mass while reducing lean mass gain (Fig. 1C–F). AraC treatment during HFD blunted the increase in body weight, caloric intake, and fat mass and preserved lean mass gain (Fig. 1C–F). The decrease in food intake accounted for the effect on weight because the feed efficiency was comparable between HFD-aCSF– and HFD-AraC–treated animals (Fig. 1G). Conversely, AraC had no effect in chow-fed (Fig. 1C–G) or VHFD-fed mice (Supplementary Fig. 2B–E). HFD intake altered plasma glucose, insulin, and HOMA of insulin resistance, which were not further modified by AraC (Supplementary Fig. 3).

The inhibition of hyperphagia induced by AraC in HFD-fed mice was particularly evident during the first week of HFD, when animals typically show an increase in energy intake, which wanes over time (Supplementary Fig. 4) (9,22). We analyzed changes in neuropeptides known to regulate this behavior after 1 week of HFD. AraC treatment in HFD-fed mice significantly suppressed the mRNA expression of AgRP (Fig. 2A) and tended to do so for neuropeptide Y (NPY) (Fig. 2B); both are orexigenic peptides produced in the ARC. Consistent with previous evidence (23), short exposure to HFD enhanced the hypothalamic mRNA levels of the anorectic POMC (Fig. 2C), likely in the attempt to counterregulate the caloric overload. AraC completely prevented this effect (Fig. 2C). Expression of neuropeptides produced outside the ARC, such as the corticotrophin-releasing hormone (CRH) in the PVN or the melanin-concentrating hormone (MCH) and prepro-orexin in the lateral hypothalamus, were not altered by either the diet or the AraC treatment (Fig. 2D–F). After 1 week of HFD, cell proliferation tended to increase in the ARC (BrdU-IR cells 1,558 ± 277.8 vs. 2,779 ± 829.6/mm³ in chow-aCSF vs. HFD-aCSF, respectively; t₇ = 1.256; P = 0.25; n = 4–5 mice/group). At this time, AraC had already abolished proliferation (no BrdU-IR cells observed) in the hypothalamus and hippocampus. Thus, inhibition of cell genesis affects ARC neuronal responses, which likely participate in the hyperphagia observed during HFD exposure.

A characteristic of animals that show changes in fat mass is that fat mass changes are usually linked to changes in plasma leptin levels. After 3 weeks of HFD, plasma leptin levels were significantly reduced in HFD AraC mice compared with HFD-aCSF–treated animals, and they were undistinguishable from those observed in chow-fed mice (Fig. 3A). Conversely, consumption of a VHFD modestly increased plasma leptin levels, which were not altered by AraC (Supplementary Fig. 2F). Therefore, we evaluated changes in hypothalamic leptin-dependent signaling. AraC fully prevented the typical HFD-induced increase in the expression of SOCS3, the inhibitor of leptin signaling (Fig. 3B), while partly blunting STAT3 phosphorylation (Fig. 3C), suggesting overall the reestablishment of normal hypothalamic leptin sensitivity.

Leptin is a proinflammatory cytokine implicated in the low-grade inflammation observed in diet-induced obesity (24,25). Therefore, we assessed changes in other proinflammatory cytokines and found that AraC treatment completely prevented the HFD-dependent increase in plasma IL-1β (Fig. 4A) so that HFD-AraC and chow-aCSF mice had similar levels of this cytokine. Chow-AraC–treated animals were not included in this analysis because they were behaviorally and molecularly indistinguishable from chow-aCSF animals (Figs. 1–3). Differently from IL-1β, no change in circulating TNF-α was observed in response to either the diet or the treatment (Fig. 4B). Because macrophages are a main source of cytokines (26), we evaluated phenotypic changes of macrophages from the peritoneal cavity, a preferred site for the collection of naive tissue-resident macrophages (27). Peritoneal proinflammatory CD86-IR macrophages were reduced in HFD-AraC compared with HFD-aCSF mice (Fig. 4C), although the pool of anti-inflammatory CD36-IR macrophages was unchanged (Fig. 4D).

Having found that AraC treatment inhibited peripheral inflammation in HFD-fed mice, we focused on hypothalamic inflammation. HFD consumption caused microglia accumulation in the ARC (Fig. 5A and B), with the microglia, labeled with Iba1, displaying the typical morphology of activated cells with enlarged soma size (Fig. 5A–C). AraC during HFD decreased the number of Iba1-IR cells (Fig. 5A and B) and blunted the characteristic morphological changes of microglial activation (Fig. 5C). Of note, the number of ARC Iba1-IR cells correlated with plasma leptin levels (Pearson r = 0.57; P < 0.05). Microglia number or morphology was not altered in the hippocampus, suggesting that microglia activation was not generalized to the whole brain (Supplementary Fig. 5A and B). Besides, HFD-aCSF–treated mice displayed a strong expression of TNF-α in the ARC, highly colocalized with Iba1 (Fig. 5D and E). This HFD-dependent increase in TNF-α was prevented by AraC treatment (Fig. 5D and E). Accordingly, hypothalamic protein levels of IκBα, a negative regulator of NF-κB inhibited by TNF-α (28), were decreased by HFD and tended to be restored by AraC (Fig. 5F).

We then determined the phenotype of the newly generated cells in the ARC of HFD-fed mice. Double labeling of BrdU-IR cells with Iba1 (Fig. 6A–C) revealed strong colocalization (43 ± 9% of BrdU-IR cells were labeled Iba1). In contrast, BrdU-IR cells did not express GFAP (Fig. 6A), a marker of astrocytes, or NeuN, a marker of mature neurons (Fig. 6A).

Altogether, these results suggest that HFD consumption causes rapid expansion and proinflammatory activation of microglia in the ARC and that blockade of this process prevents both hypothalamic and peripheral inflammation while blunting excessive caloric intake and body weight gain.

Both central and peripheral inflammation contribute to diet-induced obesity, insulin resistance, and type 2 diabetes (29). Consumption of an HFD causes inflammation in the hypothalamus (4,9,12,30); however, the underlying mechanisms remain poorly understood.

We demonstrate that exposure for a brief period (up to 3 weeks) to a diet rich in saturated fats induces cell genesis in the ARC. Stimulation of cell proliferation leads to new microglia with a proinflammatory phenotype, which is associated with increased caloric intake, body weight gain, and peripheral inflammation, all changes that could be prevented in part or completely by inhibiting cell proliferation through the central delivery of AraC.

HFD specifically increases cell proliferation in the ARC, but not in other hypothalamic nuclei or well-known cell proliferative and neurogenic structures, such as the dentate gyrus of the hippocampus (31). This cell proliferative response, in turn, results in microglia expansion. Microglia are known to take up residence in the CNS during development (32) and can self-renew in the adult brain (33–35). Differently from other brain areas, the ARC is vascularized by fenestrated vessels that facilitate the passage of circulating signaling molecules (36). This anatomical characteristic probably explains why inflammatory responses are readily observed in this brain area compared with other brain structures. This explanation also holds true for obese humans, where exacerbated microglia dystrophy is observed in the mediobasal hypothalamus but not in the cortex or amygdala (9,37). Evidence suggests that microglia work in the ARC as a sensor of saturated fatty acids (38). Such function is particularly fitting, considering the ability of ARC neurons to rapidly sense and integrate nutrient changes to modify behavior (10). Consequently, different nutrient-related signals would be expected to differentially affect cell proliferation in the ARC. This conclusion is supported by our observation that microglia expansion in the ARC depends on the type of diet consumed. Indeed, when animals ate a VHFD with a content of saturated fats comparable to that of the HFD but with only 5% of calories from carbohydrates, which are abundant in chow, cell proliferation was indistinguishable from that observed in chow-fed mice. Consumption of saturated fat, therefore, likely increases cell proliferation only when associated with appropriate intake (>5%) of carbohydrates. Besides, inhibition of cell proliferation in chow did not alter energy balance or inflammatory responses, implying that this process becomes relevant only when engaged as a counterregulatory mechanism in response to physiologic or environmental changes (39).

The type of diet used and the relatively short exposure to it could explain why we observed proliferation of microglia but not of neurons or astrocytes in the ARC. This point is critical because of conflicting evidence about the role and nature of hypothalamic cell proliferation. In fact, some studies have shown that inhibition of hypothalamic cell proliferation decreases body weight (39–41), which agrees with the current findings, whereas others describe an increase in body weight (42,43). In particular, neurogenesis has been found in the ARC (44,45). However, HFD feeding can suppress neurogenesis (41,46), not alter the neuronal fate rate (43), or induce (∼6% increase) the generation of new neurons in this area in a sex- and age-dependent manner (40,41). Moreover, astrogliosis is present in the ARC of obese animals and humans (9,13,37) but could be a consequence of signals from activated microglia.

Several technical factors such as the route and length of BrdU administration, the time point of the neuroanatomical analysis, and the age and sex of the animals studied could help to explain these seemingly contradictory findings. Besides, although BrdU labels all proliferating cells, some of these cells are immature or dying (47).

The current data demonstrate that cell proliferation is required to observe the HFD-associated inflammatory response in the ARC because AraC treatment in HFD-fed mice prevented the increase in hypothalamic microglia activation and blunted the recruitment of the NF-κB pathway. Of note, HFD consumption does not alter microglia functionality because microglia from HFD-fed animals retain full ability to respond to additional stimuli (i.e., lipopolysaccharide) and react to changes in their surroundings (7,37). However, microglia activation not only is the result of the action of specific saturated fatty acids introduced with the diet but also can be linked to other nutrient (i.e., related to high-sucrose intake) and hormonal signals, for which changes in circulating levels are observed in conditions other than HFD consumption (7,16,38,48,49). In particular, increased levels of circulating leptin and cytokines induce microglia activation (16,50–54). We observed augmented plasma leptin and IL-1β levels in HFD-fed mice, which might have activated existing microglia in the ARC. In support of this interpretation, a positive correlation was found between plasma leptin and number of ARC Iba1-IR cells. As for IL-1β, its release from microglia can be stimulated by leptin (52), and hypothalamic increase in IL-1β signaling due to early overnutrition is associated with defective melanocortin signaling, which favors food intake (55).

The current results differ from those of previous studies where HFD-induced hypothalamic inflammation was observed in the absence of peripheral inflammation (7,9). This difference may rely on the techniques used and signals investigated because we studied circulating levels of cytokines by immunoassay and looked at peritoneal macrophage phenotypes by FACS, whereas others have performed quantitative real-time PCR for inflammatory markers in peripheral tissues (7,9). Alternatively, diet might have a direct effect on peripheral immune cells because monocytes have nutrient-related receptors of which expression correlates with a range of metabolic and inflammatory markers (56,57).

Further studies will clarify whether the decreased peripheral inflammation induced by AraC is due to the decrease in fat mass observed or whether other mechanisms that directly link central with peripheral inflammatory responses might instead be involved. The latter hypothesis is supported by the observation that although AraC treatment only partly inhibits fat mass gain in HFD-fed mice, it completely prevents increases in plasma leptin and IL-1β levels.

Blockade of cell proliferation during HFD exposure may also alter microglia-neuron interaction. The typical increase in caloric intake during the first week of HFD (9,22) is blunted by AraC treatment. This reduction in caloric intake is associated with changes in the expression of neuropeptides produced by ARC neurons. Thaler et al. (9) found that feeding rats an HFD (the same one we used) for 1 week increased Iba1 staining and caused neuronal stress as assessed by the increased expression of the chaperon protein Hsp72 in ARC neurons. Valdearcos et al. (38) described similar results after 4 weeks of HFD or even short-term (3 days) administration of saturated fatty acids by gavage. In the latter study, microglia depletion abolished Hsp72 expression in neurons, suggesting that neuronal stress is triggered by increased microglia activity. Neuronal stress in response to HFD is associated with increased NF-κB signaling (12). Conversely, genetic deletion of Iκβ kinase subunit β (a component of the NF-κB pathway) in AgRP neurons decreases HFD intake (12). In the current study, HFD-fed mice had increased hypothalamic expression of SOCS3, which is a marker of augmented NF-κB activity (12), and decreased IκBα protein, a negative regulator of NF-κB (58). These changes were prevented by AraC. Thus, blockade of cell genesis, by reducing microglia proliferation and activity, probably blunts neuronal stress responses, resulting in changes in neuropeptides known to affect feeding.

Altogether, the evidence we provide linking consumption of a high saturated fat diet with hypothalamic proinflammatory microglia proliferation and body weight gain underscores the physiological relevance of our findings to humans, many of whom consume diets rich in carbohydrates and saturated fats.

Acknowledgments. The authors thank Philippe Alzieu and Florian Monteuil (INSERM U1215) and the personnel of the animal facilities of the Neurocentre Magendie for mouse care. The microscopy was done in the Bordeaux Imaging Center, a service unit of CNRS INSERM and Bordeaux University, a member of the national infrastructure France BioImaging, and supported by the LabEX BRAIN project (Bordeaux Region Aquitaine Initiative for Neuroscience). The authors also acknowledge the help of Sébastien Marais and Fabrice Cordelières (University of Bordeaux). The FACS analysis was performed at the cytometry platform (TransBioMed, University of Bordeaux).

Funding. This study was supported by a Mexican Consejo Nacional de Ciencia y Tecnología (CONACYT) postdoctoral fellowship (O.G.-Q.); INSERM (D.N.A. and D.C.); LabEX BRAIN Agence Nationale de la Recherche ANR-10-LABX-43 (D.N.A., S.L., and D.C.), ANR Blanc program ANR-2010-1414 (D.N.A., S.L., and D.C.), and ANR-13-BSV4-0006-01 (D.C.); Aquitaine Region (D.C.); and Fondation Francophone pour la Recherche sur le Diabète (FFRD), which is sponsored by Fédération Française des Diabétiques (AFD), AstraZeneca, Eli Lilly, Merck Sharp & Dohme, Novo Nordisk, and Sanofi (D.C.).

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

Author Contributions. C.A., O.G.-Q., C.R., J.R.-B., S.C., and A.C.-J. carried out the experiments, analyzed the data, and reviewed/edited and approved the manuscript. C.A., D.N.A., and D.C. wrote the manuscript. E.L. and T.L.-L. carried out experiments and reviewed/edited and approved the manuscript. A.N. contributed to the discussion and reviewed/edited and approved the manuscript. D.N.A. and S.L. analyzed data, contributed to the discussion, and reviewed/edited and approved the manuscript. D.C. conceived and supervised the study and reviewed/edited and approved the manuscript. D.C. is the guarantor of this work and, as such, has 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.