Autophagy suppresses the pathogenic immune response to dietary antigens in cystic fibrosis

Under physiological conditions, a finely tuned system of cellular adaptation allows the intestinal mucosa to maintain the gut barrier function while avoiding excessive immune responses to non-self-antigens from dietary origin or from commensal microbes. This homeostatic function is compromised in cystic fibrosis (CF) due to loss-of-function mutations in the CF transmembrane conductance regulator (CFTR). Recently, we reported that mice bearing defective CFTR are abnormally susceptible to a celiac disease-like enteropathy, in thus far that oral challenge with the gluten derivative gliadin elicits an inflammatory response. However, the mechanisms through which CFTR malfunction drives such an exaggerated response to dietary protein remains elusive. Here we demonstrate that the proteostasis regulator/transglutaminase 2 (TGM2) inhibitor cysteamine restores reduced Beclin 1 (BECN1) protein levels in mice bearing cysteamine-rescuable F508del-CFTR mutant, either in homozygosis or in compound heterozygosis with a null allele, but not in knock-out CFTR mice. When cysteamine restored BECN1 expression, autophagy was increased and gliadin-induced inflammation was reduced. The beneficial effects of cysteamine on F508del-CFTR mice were lost when these mice were backcrossed into a Becn1 haploinsufficient/autophagy-deficient background. Conversely, the transfection-enforced expression of BECN1 in human intestinal epithelial Caco-2 cells mitigated the pro-inflammatory cellular stress response elicited by the gliadin-derived P31–43 peptide. In conclusion, our data provide the proof-of-concept that autophagy stimulation may mitigate the intestinal malfunction of CF patients.


Introduction
Cystic fibrosis (CF) is the most frequent monogenic lethal disease affecting more than 85,000 subjects worldwide 1-4 . CF is caused by loss-of-function mutations in the gene coding for the cystic fibrosis transmembrane conductance regulator (CFTR) 5,6 , a protein with 1480 amino acid residues that belongs to the ABC transport family and functions as a cyclic AMP-regulated anion channel. CFTR is expressed in, and is relevant to the function of, many tissues, including airways, small and large intestine, pancreas, biliary tree, male reproductive tract and sweat glands 3,7 , but it is also expressed in central nervous system, leukocytes, smooth muscle and cartilage of the large airways 7 . Approximately 2000 mutations have been identified in the CFTR gene and are categorized in 6 classes according to their impact on the synthesis (class I), processing (class II), gating (class III), conductance (class IV), quantity (class V) and recycling (class VI) of the CFTR protein [8][9][10][11] . Among, these mutations, the clinically most important one is the F508del-CFTR mutation (class II), which accounts for 70-90% of CFTR cases.
CF is best known for its respiratory phenotype, as the abnormal anion transport results in increased mucin polymer cross-links and mucus viscosity [12][13][14] , leading to accumulation of thick, sticky mucus in the lung. These events cause chronic inflammation, persistent and untreatable bacterial colonization and recurrent chest infections, mostly by Pseudomonas aeruginosa, Staphylococcus aureus, and Burkholderia cepacia 15 . Chronic infection and inflammation ultimately lead to progressive lung disease with bronchiectasis and tissue destruction, culminating in respiratory insufficiency 15 . Defective CFTR function also frequently leads to intestinal problems 16,17 , including intestinal obstruction as well as an exaggerated immune response to dietary antigens [18][19][20] . Indeed, a constitutive inflammation at both airway and intestinal mucosa, is a feature of CF 17,19,20 . Moreover, CF patients often show serum antibodies against dietary antigens 18,20 .
Importantly, CFTR malfunction, TG2 activation and autophagy deficiency are engaged in a self-amplifying feed-forward loop. For this reason, inhibition of TGM2 by cysteamine is sufficient to restore autophagy and to favor the expression of functional CFTR at the epithelial surface. Indeed, treatment of neonatal mice bearing the F508del-CFTR mutation with cysteamine can prevent intestinal obstruction 30 . Moreover, cysteamine efficiently restores CFTR function and reduces lung inflammation in patients carrying at least one class II CFTR mutation 30,31 . A combination of two proteostasis regulators, cysteamine and the autophagy inducer epigallocatechin gallate (EGCG), was particularly efficient in restoring the expression and function of the mutant F508del-CFTR protein 11,31 .
CF patients exhibit a three-fold increase in the prevalence of celiac disease (CD) 18,32,33 an extremely frequent permanent intolerance to gluten/gliadin proteins that occurs in a proportion of susceptible individuals bearing the human leukocyte antigen (HLA) DQ2/DQ8 [34][35][36] . Accordingly, CFTR defective mice exhibit an increased susceptibility to the enteropathogenic effects of gliadin, a common dietary protein present in gluten from wheat, rye, and barley 18 . Gliadin inhibited the function of CFTR in human enterocytes and mouse models of CD through a direct molecular interaction involving a specific gliadinderived peptide (P31-43) with the nucleotide-binding domain-1 (NBD1) of CFTR 18 . Indeed, the effects of gliadin on enterocyte proteostasis are reminiscent of those observed in CFTR defective mice. Given the pivotal role of BECN1 and autophagy in orchestrating proteostasis in CF epithelia, we investigated whether the increased responsiveness to gliadin in CF mice may be due to defective autophagy and whether re-establishing BECN1 levels and autophagy by means of cysteamine would protect the CF intestine against the detrimental effects of gliadin.
Since gliadin is capable of impairing CFTR function in gliadin-sensitive CFTR-sufficient mice 18 , we first assessed whether the cysteamine-mediated rescue of F508del-CFTR function would persist after gliadin administration. To this aim, the small intestine from Cftr F508del/F508del mice exposed to cysteamine and/or gliadin (n = 10 per group of treatment) were mounted in Ussing chambers and CFTR function was assessed as the forskolin (Fsk)-induced increase of chloride current (Isc (μA/cm 2 )). In Cftr F508del/F508del and Cftr F508del/− mice, the 5-day pretreatment with cysteamine enhanced intestinal CFTR function, as expected 31 . This CFTR rescuing ability of cysteamine was not compromised by gliadin challenge (Fig. 1a, b). Of note, cysteamine restored CFTR function in gliadin-treated or vehicle-treated Cftr F508del/F508del and Cftr F508del/− mice, whereas it failed to do so in Cftr −/− mice (Fig. 1c), in line with the idea that the positive effect of cysteamine requires the presence of 'rescuable' F508del-CFTR protein.
Cysteamine protects Cftr F508del mice from the effects of gliadin in vivo Next, we investigated whether cysteamine would control the increased mucosal immune response that occurred in gliadin exposed Cftr F508del/F508del mice. To this aim, we measured the levels of proinflammatory cytokines in small intestine homogenates from mice fed with gliadin for 4 weeks in the presence or absence of cysteamine. Cysteamine was effective in preventing the increased production of IL-17A and IFN-γ induced by gliadin (p < 0.01 and p < 0.001) (Fig. 2a, b). In addition, cysteamine controlled the production of IL-15, a master pro-inflammatory cytokine pivotal for driving the gliadin-induced enteropathy [37][38][39][40][41] (Fig. 2c). Indeed, IL-15 is constitutively upregulated in mouse CF intestine and is significantly induced by oral gliadin challenge 18 . Again, the anti-inflammatory effect of cysteamine against gliadin-induced cytokine producton was observed in Cftr F508del/F508del , Cftr F508del/− but not in Cftr −/− mice (Supplementary Figure 1), supporting the hypothesis that cysteamine controls the gliadin- and protein (right) levels in small intestine homogenates from Cftr F508del/F508del or their Cftr WT littermates treated with vehicle or cysteamine (60 µg/kg in 100 µl saline/day for 5 days) and then challenged with gliadin for consecutive 4 weeks (5 mg/daily for 1 week and then 5 mg/daily thrice a week for 3 weeks) in the presence or absence of cysteamine (60 µg/kg in 100 µl saline/day) (n = 10 per group). Means ± SD of pooled samples assayed in triplicates. ##p < 0.01 or ### p < 0.001 Cftr WT versus Cftr F508del/F508del ; §p < 0.05 or § §p < 0.01 or § § §p < 0.001 versus cysteamine treatment; **p < 0.01, ***p < 0.001 versus gliadin challenge;°p < 0.05 or°°p < 0.01 or°°°p < 0.001 versus cysteamine + gliadin (ANOVA, Bonferroni post hoc test) induced inflammation through restoring F508del-CFTR function.
Cysteamine protects Cftr F508del mice in vivo from the increased responsiveness to gliadin through restoring BECN1 and autophagy We previously reported that the inhibition of TGM2 with the subsequent restoration of BECN1 protein levels and autophagy are pivotal for allowing cysteamine to rescue F508del-CFTR at the epithelial surface 11,18,21,31 . For this reason, we investigated whether the protective effects of cysteamine against gliadin induced immune activation would be linked to its capacity to restore autophagy. To this aim, Cftr F508del/F508del mice were backcrossed into a Becn1 haploinsufficient background (to generate Cftr F508del/F508del / Becn1 +/− mice) 31 , and gliadin was orally administered upon optional pretreatment with cysteamine. Cysteamine was unable to restore the function of the intestinal CFTR (determined in Ussing chambers) from Cftr F508del/F508del / Becn1 +/− mice, either before or after gliadin challenge (Fig. 3a). Gliadin triggered an inflammatory response in Cftr F508del/F508del /Becn1 +/− mice, similarly to that observed in Cftr F508del/F508del mice (Fig. 3b-d). Of note, cysteamine failed to mitigate the gliadin-elicited production of IL-15, IL-17A, and IFN-γ in Cftr F508del/F508del /Becn1 +/− mice (Fig. 3b-d). In conclusion, it appears that the haploinsufficiency of Becn1 (genotype: Becn1 +/− ) abrogates the antiinflammatory effects of cysteamine that is normally seen in Cftr F508del/F508del mice.

Restoring BECN1 protects intestinal epithelial cells from the detrimental effects of gliadin
To complete our demonstration that BECN1 and autophagy are crucial for the inflammatory effects of gliadin in intestinal epithelial cells, we resorted to human intestinal epithelial Caco-2 cells, which are reportedly sensitive to gliadin or gliadin-derived peptides 40 . When confluent Caco-2 cells were challenged for 3 h with a peptic-tryptic digest of gliadin from bread wheat (PT gliadin; 500 μg/ml) 40,41 or the gliadin-derived peptide LGQQQPFPPQQPY (P31-43), CFTR function is inhibited 18 . Moreover, gliadin treatment of Caco-2 cells caused a reduction in BECN1 protein levels, as compared to unchallenged controls (Fig. 4a). Next, we transfected Caco-2 cells with HA-Beclin 1 and challenged them for 3 h with gliadin or P31-43. Of note, the enforced expression of BECN1 (which enhances the generation of autophagosomes and autophagy 21 ), prevented signs of gliadin-induced inflammation, as it avoided the upregulation of TGM2, the activating phosphorylation of ERK 1/2 and the downregulation of PPARγ that were induced by gliadin (Fig. 4b, c). These results suggest that BECN1 plays an active role in mitigating the gliadininduced inflammatory response of intestinal epithelial cells.

Discussion
The proteostasis network is a system of cellular adaptation to endogenous or environmental stress. Mismanaged proteostasis contributes to a number of diseases arising as a result of inherited or stress-induced defects in protein conformation 42 . Autophagy is a major player of the proteostasis network as it regulates the turnover of large protein aggregates and even entire organelles. In addition, several components of the autophagy machinery dynamically interact with multiple signalling pathways to ensure intracellular homeostasis [43][44][45] . CF is the quintessential example of a disease characterized by major alterations of the proteostasis network 11,21,23,46 . Defective CFTR function highly compromises the capacity of cells to adequately respond to endogenous stress signals as well as to external challenges arising within the respiratory and gastrointestinal tracts 11 .
The intestine from CF patients is exposed to a particularly high antigenic load due to the frequent insufficiency of the exocrine pancreas 17 . Moreover, the local overactivation of the innate immune system compromises the handling of dietary molecules, thus favouring inadequate cellular and humoral immune responses to food components. Accordingly, CF patients often exhibit increased levels of antibodies against alimentary antigens, including anti-gliadin IgA antibodies, shifts in the intestinal microbiota, elevated fecal calprotectin levels and increased intestinal permeability 17,20,47 . Moreover, the prevalence of autoantibodies against TGM2 is four times higher than in the general population, even in the absence of histological evidence of intestinal lesions 18,32,33 . Indeed, when CFTR is disabled, the intracellular milieu undergoes major pathogenic changes. CFTR inhibition results in a ROS-mediated increase in the abundance and activity of TGM2 25,26 with consequent functional sequestration of BECN1 complex and inhibition of autophagy [21][22][23] . Thus, CFTR inhibition, TGM2 activation and BECN1 inactivation act in concert to compromise proteostasis in the small intestine of CF mice, driving constitutive proinflammatory reactions that involves the activation of the NF-κB pathway and the NLRP3 inflammasome.
Here we demonstrate that functional BECN1 and autophagy are required to prevent the increased susceptibility of CF intestine to the gluten component gliadin. Indeed, the proteostasis regulator cysteamine was capable of reducing the pro-inflammatory effects of gliadin in mice bearing the most common F508del-CFTR mutant, either in homozygosity or in compound heterozygosity. Apparently, cysteamine abrogates the susceptibility of mice to oral gliadin challenge by acting as TGM2 inhibitor, thus preventing the BECN1 sequestration and autophagy impairement that normally result from CFTR inhibition. However, cysteamine fails to prevent gliadin-induced inflammation in CFTR KO mice, meaning that its effects are mediated by its ability to rescue CFTR function, as reported 23,31 . Importantly, the beneficial effects of cysteamine on F508del-CFTR mice are lost when these mice are backcrossed in an autophagy-deficient (Becn1 haploinsufficient mice) background, indicating that the restoration of BECN1 levels and autophagy are indeed required to avoid the enteropathic effects of gliadin.
In aggregate, stimulating autophagy might represent a novel option to prevent intestinal manifestations of CF. In favor of this notion, it appears that enforced expression of BECN1 in gliadin-sensitive human intestinal epithelial cells 18,40 , effectively opposes the capability of the gliadinderived P31-43 peptide 18,40,41 to induce an epithelial stress response. In this perspective, and in line with the evidence that the best option to stimulate autophagy is to interfere with the function of its endogenous inhibitors 48 , it might be attempted to neutralize BECN1 inhibitory proteins. Druggable endogenous BECN1 antagonist include proteins from the BCL2 family (which can be targeted with so-called BH3 mimetics including ABT737, navitoclax, and venetoclax) 49 , the mechanistic target of rapamycin complex-1 (mTORC1) (which are inhibited by rapamycin, everolimus or tacrolimus) 48 , as well as the acetyltransferase EP300 (which is inhibited by aspirin, epigallocatechine gallate, or spermidine) 31,46,50,51 . Moreover, there is the option to directly inhibit TGM2 by cysteamine 31,52 and to combine cysteamine with other autophagy stimulators such as EGCG 31 .
In conclusion, our data highlight the implication of CFTR in the suppression of diet-induced inflammation. CFTR may be viewed as a major stress sensor that alerts the autophagy machinery when a potentially harmful perturbation risks to perturb mucosal homeostasis. Once activated, autophagy then orchestrates the proper handling of luminal triggers by the intestinal mucosa. Pharmacological autophagy enhancement may be harnessed to prevent intestinal inflammation and to improve the nutritional status of CF patients.

Plasmids and transfection
The pcDNA3-HA-beclin 1 expression vector (a gift from N. Mizushima) was used for transfection experiments. Cells were transfected with pcDNA3-HA-beclin 1 by means of Lipofectamine 2000 (Invitrogen) in accordance with the manufacturer's instructions.

Ussing chamber
Chambers for mounting mouse tissue biopsies were obtained from Physiologic Instruments (model P2300, San Diego, CA, USA). Chamber solution was buffered by bubbling with identical Ringer solution on both sides and were maintained at 37°C, vigorously stirred, and gassed with 95%O 2 /5%CO 2 . Tissues were short circuited using Ag/AgCl agar electrodes. A basolateral-to-apical chloride gradient was established by replacing NaCl with Nagluconate in the apical (luminal) compartment to create a driving force for CFTR-dependent Cl − secretion. To measure stimulated Isc, the changed sodium gluconate solution, after stabilization, was supplied with 100 µM amiloride. Agonists (forskolin) were added to the bathing solutions as indicated (for a minimum 5 min of observation under each condition) to activate CFTR channels present at the apical surface of the epithelium (either cell surface or lumen side of the tissue) and CFTR Inh-172 (10 µM) was added to the mucosal bathing solution to block CFTR-dependent Isc. Short-circuit current (expressed as Isc (μA/cm 2 )) and resistance were acquired or calculated using the VCC-600 transepithelial clamp from Physiologic Instruments and the Acquire &Analyze2•3 software for data acquisition (Physiologic Instruments), as previously described 18,54 .

ELISA
ELISA analysis was performed on tissue samples using standard ELISA kits (R&D Systems) for IL-15, IL-17A, INF-γ, according to the manufacturer's instructions. Samples were read in triplicate at 450 nm in a Microplate Reader (BioRad, Milan, Italy) using Microplate Manager 5.2.1 software. Values were normalized to protein concentration evaluated by Bradford analysis.

Statistical analysis
GraphPad Prism software 6.01 (GraphPad Software) was used for analysis. Data are expressed as means ± SD. Statistical significance was calculated by ANOVA (Bonferroni's post hoc test) for multiple comparisons and by Student's t-test for single comparisons. We considered all P values 0.05 to be significant. The in vivo groups consisted of ten mice/group. The data reported are either representative of at least three experiments.