Production of Elastin-Derived Peptides Contributes to the Development of Nonalcoholic Steatohepatitis

Affecting 25% of the general population, 50% of patients with type 2 diabetes, and 75% of obese patients (1), nonalcoholic fatty liver disease (NAFLD) is usually benign. A combination of steatosis, hepatocyte injury, and inflammation give rise to a process called nonalcoholic steatohepatitis (NASH), which can cause progressive fibrosis and cirrhosis. Although NASH is very closely linked to insulin resistance, obesity, and metabolic syndrome, the exact cause of this chronic disease has not been formally elucidated. Indeed, not all patients with obesity or insulin resistance exhibit NAFLD/NASH, and, conversely, not all patients with NAFLD/NASH exhibit insulin resistance or obesity. Endogenous factors such as tissue aging and independent lifestyle factors may contribute to the progression of NAFLD into NASH (2). In addition, extracellular matrix (ECM) remodeling such as elastin degradation during aging and the production of elastin-derived peptides (EDPs) (found in systemic blood circulation) may play an important role in the development of NASH. Elastin, a major structural protein of skin, lung, and large blood vessels, confers on them the physical properties of extensibility and elastic recoil and, thus, is essential for the physiological function of these and other tissues. Polymeric elastin is extraordinarily durable; in “inflamm-aging” conditions, elastase secretion, such as the secretion of neutrophil elastase (NE), is increased and the elastin undergoes degradation, releasing elastokines (3). Elastokines are EDPs that exhibit biological activities and are considered to be aging biomarkers. The typical motif of EDP is the hexapeptide VGVAPG, whose sequence is tandemly repeated six times in human tropoelastin (4). Our laboratory has previously shown that EDPs bind to the elastin receptor complex (ERC) and contribute to the development of several diseases, such as cancer, atherosclerosis (5), and insulin resistance (6), in mice fed with a low-calorie diet. This widely expressed receptor complex comprises a peripheral EDP-binding subunit, the so-called elastin binding protein (EBP), a protective protein/cathepsin A, and neuraminidase-1 (Neu-1), whose enzymatic activity is required for ERC signaling (7). Elastin knockout mice have been shown to exhibit diabetes (8), suggesting that elastic fiber aging may influence the development of diseases characteristic of the metabolic syndrome. Nevertheless, no study has yet investigated the role of EDPs and their receptors in liver physiology and NASH development. The present work shows, for the first time, that ERC activated by EDPs contributes to lipid accumulation and inflammation by impairment of the AMPK pathway. In addition, EDPs induce a hepatic ECM remodeling leading to strong fibrosis. With regard to humans, we suggest that plasma EDPs may serve as new noninvasive biomarkers to diagnose the risk of developing NASH.

Animal treatments were performed according to the American Physiological Society’s Guiding Principles for the Care and Use of Animals. Eight-week-old C57BL/6N, db/db, and db/+ mice (Janvier, Saint-Berthevin, France) were maintained continuously with normal food and water available ad libitum. Because several studies have shown that EDP clearance in vivo is extremely fast, we injected 10 mg/kg weekly for 8 weeks (6), i.e., less than observed in diabetic mice (15 mg/kg). To study the implications of the ERC, we injected ERC inhibitors (V14 peptides [91 µg/kg/week], chondroitin sulfate [CS; 50 mg/kg/week], or 2,3-dehydro-2-deoxy-N-acetylneuraminic acid [DANA; 10 g/L/week]) (6) in parallel to kE (mixture of EDPs). We performed a treatment with metformin (35 mg/kg/day) (Sigma-Aldrich, Saint Quentin Fallavier, France) or resveratrol (2 mg/kg/day) (Sigma-Aldrich) in addition to the kE treatment.

Once reaching confluence, AML12 hepatocytes (CRL-2254; American Type Culture Collection, Manassas, VA), seeded at a density of 7,500 cells/cm², were exposed to the different treatments and their carrier vehicles during a 24-h period, which was followed by mRNA or protein extraction. The conditions of treatment were as follows: kE (50 μg/mL) ± CS sodium (200 μg/mL) or DANA, used as Neu-1 inhibitor (Merck, Molsheim, France), V14 peptide (purity >98%) (Genecust, Dudelange, Luxembourg), resveratrol (50 μmol/L), or metformin hydrochloride (1 mmol/L). The data are the mean of three independent experiments, each performed in triplicate.

Plasma was collected from 48 obese French women and men (BMI >40 kg/m²) (Supplementary Table 1) according to the ethical guidelines of the 1975 Declaration of Helsinki. The selection of patients is described in the Supplementary Materials and Methods. The patients were divided into two groups: a group without diabetes (HOMA of insulin resistance [HOMA-IR] <4.8) and a group with diabetes (HOMA-IR >4.8). Two liver disease scores were determined: the fatty liver index (FLI) was calculated according to Bedogni et al. (9) and the NAFLD fibrosis score was calculated using the formula created by Angulo et al. (10). Among 48 patients, only 16 had been diagnosed with NASH by biopsy. With regard to the function of steatosis, inflammation, and ballooning, this score ranges from 0 to 8. Plasma EDPs and NE activity were measured using the Fastin (Biocolor, Carrickfergus, U.K.) and NE Activity Assay (Abcam, Cambridge, U.K.) kits, respectively. After a 12-h overnight fast, venous blood samples were drawn to determine the levels of transaminases (alanine aminotransferase [ALT], aspartate aminotransferase [AST], and γ-glutamyltransferase), glucose, cholesterol, triglycerides (TGs), and insulin.

Cpt1/2, medium-chain acyl-CoA dehydrogenase (MCAD), and long-chain acyl-CoA dehydrogenase (LCAD) activity were analyzed from carnitine and acylcarnitine assays described by Pooya et al. (11). Malondialdehyde (MDA), superoxide dismutase (SOD), and glutathione peroxidase (GPX) were assayed as previously described by Dali-Youcef et al. (12). All elastin samples digested by NE were analyzed by nanoHPLC-nanoESI-QqTOF MS(/MS) (Thermo Fisher Scientific, Idstein, Germany and Waters/Micromass, Manchester, U.K.). To measure liver fatty acid composition, lipid assays were performed using lipid samples extracted from mouse liver by Folch procedure.

Mouse liver was fixed immediately after euthanasia in paraformaldehyde 4% and then paraffin embedded for histological staining or frozen for Oil Red O staining. Sections of 4 μm were stained with hematoxylin-eosin (H-E), picrosirius, or Oil Red O and observed under an Olympus light microscope.

Assays of serum AST, ALT, total cholesterol, HDL cholesterol, LDL cholesterol, TGs, and free fatty acids (FFAs) assays were performed by the ICS metabolic platform (ICS-IGBMC, Illkirch, France) as previously described (12).

Protein lysate (150 µg) was immunoprecipitated using 1 µg of c-Met (Cell Signaling) antibody overnight at 4°C, followed by 20% protein G–Sepharose (100 mL; Amersham Pharmacia Biotech) for 1 h. Western blotting was performed as previously described (6,11). The antibodies used for Western blot were as follows: rabbit polyclonal anti-AMPK, p-AMPK (T172), acetyl CoA carboxylase (ACC), p-ACC (S79), and p-cMet (Y1234/1235) from Cell Signaling and LKB1, p-LK1 (S431), and anti-actin from Santa Cruz Biotechnology. The sialic acid level was determined by the DIG Glycan Differentiation Kit (Sigma-Aldrich, St. Louis, MO). The analysis for qPCR was performed as previously described (6,12), and specific primers are listed in Supplementary Table 3. 36B4 and RPS26 were used as internal controls.

For correlation studies with human and mouse samples, P value and z test regressions were performed using the Statview software. Evaluations were assessed by Mann-Whitney U tests. For all analyses, at least three independent experiments were performed. Values of P < 0.05 were considered significant (*P < 0.05). The error bar represents the SEM.

Neutrophils through NE production play a crucial role during liver disease (13, 14). Accordingly, we observed an increase in NE activity during the progression of metabolic syndrome characterized by an accumulation of blood FFAs, liver TGs, and hepatic collagen concentrations in db/db mice (Fig. 1A–D) as well as in mice fed with a high-fat diet (Supplementary Fig. 1). Since the deletion of the NE gene improves liver inflammation in the NAFLD model, we hypothesized that a pharmacological treatment focused on NE activity could influence lipid metabolism in mice. Therefore, db/db mice (6 weeks) were injected with LY54439 (2 mg/kg/day for 2 weeks), which is reported to be an NE inhibitor (Fig. 1E). We found a decrease in lipid droplet accumulation in the liver (Fig. 1F) and mRNA expression coding for lipogenesis enzymes (Fig. 1G). LY544349 treatment restored activated AMPK signaling initially decreased in db/db mice (Fig. 1H). According to Mansuy-Aubert et al. (13) and Talukdar et al. (14), the data suggested a role for neutrophils in modulating hepatic metabolism through a mechanism involving NE, promoting an intense inflammatory response. Nevertheless, the relationship between NE and lipid metabolism/AMPK signaling was not clearly identified in these studies.

NE is a protease that cleaves ECM proteins such as elastin and leads to the formation of potentially active EDPs. We demonstrated in vitro that potential bioactive peptides, bearing the GXXPG motif (Table 1 and Supplementary Table 4, respectively), were released from both mice and human vascular elastin by NE. The sequences of the peptides were of the GXXPG type. In parallel, we observed a high fragmentation of aortic elastic fibers in several obese and diabetic mouse models, such as db/db mice (Fig. 2A) and mice fed with a high-fat diet (Supplementary Fig. 1). Plasma EDPs were significantly increased (Fig. 2B). Interestingly, plasma EDP levels and NE activity were positively correlated (Fig. 2C). The treatment of db/db mice with NE inhibitors reduced aortic elastic fiber alterations (Fig. 2A) and plasma EDP levels (Fig. 2D) significantly. Previously, several studies have demonstrated that elastic fiber integrity is an important factor for glucose and lipid homeostasis (8,15). Therefore, we suppose that EDPs produced by NE may be involved in NAFLD progression.

We tested an EDP known to be biologically active (VGVAPG), with its scramble (VVGPGA) known to be biologically inactive, and a mixture of EDPs called kE (3) on mice. All analyses were performed after 2 months of EDP treatment (6). The plasma assays showed an increase of FFAs, TGs, and transaminases (Supplementary Fig. 2A and B), suggesting hepatic impairment in mice with accumulated EDPs when compared with the control sample. Hepatic microvesicular steatosis (Fig. 3A) present in mice treated with EDPs was characterized by an accumulation of total lipid (Supplementary Fig. 2C), lipid peroxidation (Fig. 3C), TGs (Fig. 3B and Supplementary Fig. 2D), and saturated fatty acids (Supplementary Fig. 2E), such as stearate and palmitate (Supplementary Fig. 2F). The higher density of lipid droplets observed under the influence of kE (Fig. 3A) can be explained by an increase in the expression of lipogenic genes (Fig. 3D and E). Conversely, the expression of β-oxidation genes and the lipolytic gene were greatly reduced (Fig. 3F and G). These data confirmed our hypothesis that EDPs are involved in NAFLD progression. Previously, we demonstrated that EDPs are biologically active through their binding on the ERC, which induces the modulation of several signaling pathways. Based on these conclusions, we used several inhibitors known to inhibit ERC activity. Therefore, in parallel to kE injections, we treated the mice with DANA, CS, or V14 once a week for 8 weeks. Histological approaches (H-E and Oil Red O staining) showed that the three inhibitors reduced lipid droplet accumulation in the liver (Fig. 3A) but also restored the normal expression of genes coding for lipogenesis or β-oxidation enzymes (Fig. 3D–G). Acylcarnitines were analyzed in livers from the different groups (Fig. 3H). The ratios of acylcarnitine C14/C16 and C8/C10 were significantly decreased, suggesting a reduction in the activity of LCAD and MCAD, respectively. Similarly, a significant decrease in the ratio of C0/(C16 + C18) and an increase in the ratio of (C16+C18:1)/C2 suggested that carnitine palmitoyltransferase I (CPT1) and CPT2 (12) were impaired by kE. When administrating ERC inhibitors, acylcarnitine levels were unaffected by kE. These lipid metabolisms are under the control of the AMPK pathway (Fig. 3I). The addition of ERC inhibitors protected against a decrease in AMPK phosphorylation induced by kE, thus suggesting a specific role for EDPs in NAFLD. Otherwise in db/db mice, we observed a decrease of AMPK phosphorylation associated with an increase of lipogenesis (Fig. 1), on the one hand, and an increase of endogenous EDPs (Fig. 2) on the other. Accordingly, we treated the db/db mice with the ERC inhibitors (DANA or CS) for 4 weeks. We observed that these inhibitors partially restored the AMPK pathway (Fig. 3J and K), mRNA expression involved in lipogenesis (Fig. 3L), and lipid droplet accumulation (Fig. 3M). These data suggest that EDPs produced by NE could be an intermediate link in the relationship between NE and NAFLD.

To determine the direct effect of EDPs on NAFLD, we incubated hepatocytes with kE with or without ERC inhibitors (DANA, CS, or V14). In parallel, we compared the effects induced by these inhibitors with cells treated with kE and metformin or resveratrol, both known to respectively stimulate lipolysis and block lipogenesis. Except for resveratrol, all tested drugs reversed the effects of kE, i.e., they inhibited the induction of lipogenesis by the expression of ACC, fatty acid synthase (FAS), or SREBP1c and restored β-oxidation by expression of MCAD and CPT1a (Fig. 4A). Interestingly, these enzymes are targets of AMPK signaling, which was activated in the presence of ERC inhibitors (Fig. 4B). The literature mentions that AMPK phosphorylation can be induced, for instance, by metformin (activation of AMPK by LKB1) or by resveratrol (which activates AMPK by SIRT1 stimulation). Despite the presence of kE, metformin increased phosphorylation of AMPK, whereas resveratrol had a limited effect, suggesting that kE inhibited AMPK phosphorylation independently of SIRT1. These results were also observed in vivo (Figs. 1H and 3J). We tested several stimuli to explain AMPK phosphorylation. Whereas the AMP-to-ATP ratio did not seem be affected (data not shown), AMPK could also be activated by the phosphorylation of LKB1 (16). We observed that kE alone inhibited LKB phosphorylation, which was restored by DANA, CS, V14, and metformin, whereas resveratrol had no effect in vitro or in vivo. Interestingly, the LKB1/AMPK cascade was regulated by hepatic growth factor receptor (HGFR), also called c-Met (16), and the decrease of c-Met activity was associated with NAFLD progression (17–19). In our cell culture model (Fig. 4C), the EDP decrease of the phosphorylation of c-Met could be explained by desialilation of this receptor.

Any chronic liver aggression, such as lipid accumulation, induces oxidative stress and inflammation, which causes fibrous scarring of tissue according to the two-hit theory (19). First, the results showed in the EDP group an increase of oxidative stress (decrease of SOD and GPX activity) (Fig. 5A) and of inflammation (Fig. 5B), suggested by Kupffer cell marker expressions (F4/80). The expression of cytokines (Fig. 3B), which increased in the group treated with kE, is known to induce fibrosis. Chronic injections of kE or VGVAPG increased the picrosirius staining (Fig. 5C and Supplementary Fig. 3A) and collagen (Fig. 5D and E and Supplementary Fig. 3B and C) and significantly decreased protease expressions (Fig. 5F and Supplementary Fig. 3D). In terms of matrix remodeling, the expression of elastic fiber components (tropoelastin, fibulin-5, and fibrillin-1) was significantly increased (Fig. 5E and Supplementary Fig. 3C). However, the absence of orcein staining failed to demonstrate elastic fiber formation (data not shown). Interestingly, ERC inhibitors used in parallel with kE injections restored the expression of proteases, collagen, and elastin in a similar way to those observed in the control mice (Fig. 5). These data suggest that the inhibition of the EDP-ERC pathway may limit NASH progression. The fibrosis observed with EDP injections is interesting because the models of NAFLD commonly used, such as mice fed with a high-fat diet or db/db, do not show spontaneous fibrosis. Therefore, we decided to induce a moderate NASH with db/db mice, which were fed either a methionine- and choline-deficient (MCD) diet for 4 weeks or treated with thioacetamide (TAA; 200 mg/kg body weight three times per week) for 3 weeks to determine whether EDPs could potentialize the NASH and whether treatment with an ERC inhibitor could be sufficient to block the onset of fibrosis. Unfortunately, CS, which has previously given the best effects, has only partially reduced inflammation markers such as F4/80 (Fig. 5G) as well as fibrosis labeling (Fig. 5H) and collagen expression (Fig. 5I). We suppose that this decrease was associated with fibrosis induced by EDPs and the majority was associated with either diet or TAA. Nevertheless, in parallel with TAA injection, db/db mice treated with an NE inhibitor dramatically reduced fibrosis and inflammation markers (Supplementary Fig. 4).

Plasma EDPs are positively correlated with the degree of hepatic steatosis in patients with NASH. In order to transpose the findings obtained in mice to human pathology, we analyzed 48 French patients with or without type 2 diabetes. The individual characteristics of the patients are listed in Supplementary Table 1. Although the NE activity was not significantly modified, we observed, as in the diabetic mice, that EDP concentrations were increased in patients with diabetes when compared with patients without diabetes (Fig. 6A). Plasma EDP concentrations had a 60.2% correlation with NE activity (Fig. 6B). A positive correlation was observed between NE activity (Supplementary Fig. 5) and plasma EDPs (Fig. 6C) and BMI, glycemia, HOMA-IR, plasma TGs, FFAs, C-reactive protein (CRP), and neutrophil concentrations. Obesity, hyperglycemia, and high levels of FFAs and CRP encompass an alteration of the hepatic metabolism. Therefore, the FLI (9) and the NAFLD fibrosis score (10) were calculated. They correlated with both NE activity (Supplementary Fig. 5) and plasma EDP levels (Fig. 6C). However, multiple regression analyses showed that EDPs were a determinant of FLI (P = 0.0041) and NAFLD fibrosis (P = 0.0024), suggesting that EDP levels may be a better plasma biomarker of liver diseases than NE activity. Among our cohort, 16 patients had been diagnosed with NASH after liver biopsies (Supplementary Table 2). Although no differences in NE activity were observed, multiple regression analyses showed that EDPs are determinants of the degree of fibrosis (P < 0.0001). Indeed, the plasma EDP level was increased in tandem with the degree of steatosis (Fig. 6D) and hepatic fibrosis (Fig. 6E). These histopathology data confirm that endogenous EDPs might be considered as plasma biomarkers to facilitate the diagnosis of liver injuries.

Several studies have shown that metabolism, in particular hepatic metabolism, not only depends on a normal genetic/epigenetic status of the genes regulating fat and glucide use but also involves ECM homeostasis. Elastin, found in large blood vessels, skin, and lungs, represents 90% of the composition of the elastic fibers (17). During aging, mechanical vascular stresses and local actions of proteases such as NE lead to degradation of the elastic fibers, i.e., elastolysis, and production of EDPs (18). Although these elastolysis mechanisms are still poorly understood, recruitment of elastases, such as the NE observed in our present study, could explain EDP production in murine models of obesity and also in obese patients. EDPs, recognized as aging markers (20), can accelerate metabolic syndrome features such as NAFLD/NASH (Fig. 6F). We demonstrate here, for the first time, that EDPs, potential products of the increase of NE activity, allow the accumulation of hepatic TGs. In contrast, the inhibition of NE activity protects mice from accumulating fat. From a clinical perspective, our study shows that EDPs, rather than NE activity, are determinants of the degree of hepatic steatosis and fibrosis. During the last decade, our team has deciphered the signaling pathways triggered by the ERC upon its binding to the canonical human sequence VGVAPG (21). This receptor has a very singular composition that comprises three subunits: a peripheral subunit called EBP and two membrane-associated proteins, a protective protein/cathepsin A (PPCA) and a neuraminidase (Neu-1), which is essential for signal transduction by the ERC. ERC activity can be inhibited by several molecules or peptides. V14 blocks EDP binding with EBP, CS inhibits EBP function, and DANA is an inhibitor of Neu-1 activity (21). Interestingly, based on our current data, we identified that these ERC inhibitors restore the normal expression of mRNA coding for lipogenesis, β-oxidation, and lipolysis impaired by EDPs. Futhermore, the LKB1-AMPK pathway, dramatically reduced by EDPs in hepatocytes, can be restored by the presence of an ERC inhibitor similarly to metformin treatments described in the literature (22). This alteration of the AMPK pathway associated with hepatic lipid accumulation has been demonstrated in several studies (22) but is still poorly understood. Several stimuli can be involved in AMPK activation, such as the activation of the leptin receptor, the reduced AMP-to-ATP ratio, or SIRT1 activation. In our study, the data obtained with db/db mice (leptin receptor deficient), whether or not treated with resveratrol (SIRT activator) or metformin (LKB1-AMPK activator), suggest that EDPs influence the activation of LKB1 before AMPK phosphorylation (Fig. 6F). In the context of metabolic disorders, the activation of c-Met/LKB1/AMPK signaling has been shown in NASH progression (23) and to promote hepatocyte proliferation and viability (16). Upon HGF binding, Tyr-1234/1235 residues in the catalytic domain of c-Met are phosphorylated and the receptor is activated. Once activated, c-Met activates multiple downstream signaling pathways, including the PI3K/Akt and MEK/ERK pathways. Through these intermediary pathways, HGF/c-Met governs a diversity of important cellular responses, such as protective actions against lipid accumulation, lipotoxicity, oxidative stress, or ECM remodeling in the liver (17–19). Nevertheless, as described for the insulin receptor (6), EGFR, and PDGFR (24), c-Met activity may be dependent on N-glycan chains fixed on the receptor ectodomain (25,26). However N-glycosylation modifications regulating the enzymatic properties of c-Met are unknown. In our study, we observed that the ERC activated by EDP is able to cleave sialic acids on the N-glycosylation chain of c-Met, thus decreasing the c-Met activity (Fig. 6F). Previously, we demonstrated (27) that removing sialic acid impairs the N-glycan chain dynamic and might impair receptor behavior with its ligands or catalytic domain.

The c-Met defect was involved with lipotoxicity, oxidative stress, and inflammation, which are the principal factors mentioned in the transition from NAFLD to NASH (19). In our study, we demonstrate for the first time that EDPs induce the accumulation of nonesterified saturated FFAs known to be hepatotoxic (28,29), an increase in oxidative stress and cytokine expression. These parameters are known to be involved in ECM remodeling. Thus, we show that EDPs essentially induce an accumulation of type I collagen and, surprisingly, an expression of tropoelastin, fibulin-5, and fibrillin-1, as mentioned by Pellicoro et al. (30). Oxidative stress, inflammation, and ECM remodeling demonstrate that EDPs contribute to the NAFLD-NASH transition but also that ERC inhibitors are able to restore normal hepatic phenotypes. To find out whether ERC inhibitors such as CS can reduce fibrosis, we used two mouse models of NASH in which db/db mice were either injected with TAA or fed with an MCD diet for 3 weeks. Unfortunately, we did not observe any real benefit after the administration of CS. We believe that the fibrosis induced by these procedures is extremely important and may mask the fibrosis induced by endogenous peptides in db/db mice. Finally, it is rather difficult to estimate the effectiveness of the treatments because of the heterogeneity of the studies (population and objectives) and their lack of power. Several molecules have been evaluated but the results are inconclusive to date. We think that reducing elastin aging, EDP production (i.e., by inhibition of NE), or the activation of ERC may be a new strategy to limit NASH progression.

In conclusion, a vicious cycle induced by blood vessel defects may be responsible for NASH and NASH progression. Indeed, we show for the first time that the EDPs found in blood circulation have a systemic effect and play an important role in lipid metabolism and hepatic ECM remodeling. According to our clinical results, the identification of plasma EDPs is a new determinant of NASH. Finally, the inhibitors of elastin fragmentation, EDPs and ERC, which were used in this study to demonstrate the role of EDPs on hepatic lipid metabolism and fibrosis, suggest that aging and preservation of elastic fibers may be a potential therapeutic target for NASH.

Acknowledgments. The authors thank Aude Bressonot-Marchal (Histology Department, University Hospital, Reims, France) and Dr. Christian Garbar (Godinot Institute, Reims, France) for excellent help in histology and Fanja Rabenoelina, Olivier Bocquet, Annie Carlier, and Marie-Line Sowa (CNRS/URCA UMR7369 MEDyC) for technical help. The authors thank Bart Staels and Réjane Paumelle-Lestrelin (INSERM U545, Institut Pasteur, Lille, France) for help with the AML12 cell culture. The authors thank Bart Staels, Philippe Valet (INSERM U1048, Toulouse, France), and Catherine Postic (INSERM 1016, Institut Cochin, Paris, France) for comments and Dr. Jacqueline Van de Walle (Institut des Sciences de la Vie, Louvain-la-Neuve, Belgium) and Scribendi (Chatham-Kent, Canada) for reviewing the article for English grammar.

Funding. This work was supported by funding from the CNRS and the University of Reims Champagne-Ardenne as well as the German Research Foundation (DFG) (HE 6190/1-2 to A.H.), the LEO Foundation Center for Cutaneous Drug Delivery (2016-11-01 to A.H.), and the Fraunhofer Internal Program (Attract grant 069-608203 to C.E.H.S.). A.G. was supported by research grants from the Région Champagne Ardenne.

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

Author Contributions. B.R. and S.B. wrote the manuscript and researched data. C.I., H.S., A.H., C.E.H.S., E.B., and T.H. researched data and reviewed the manuscript. R.G., A.G., J.-L.G., J.-M.A., J.-P.B., and J.A. researched data. J.J., L.Du., P.M., A.B., and L.De. reviewed the manuscript. L.M. and V.D. contributed to discussion. S.B. 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.