Curcumin inhibits the TGF-β1-dependent differentiation of lung fibroblasts via PPARγ-driven upregulation of cathepsins B and L

Pulmonary fibrosis is a progressive disease characterized by a widespread accumulation of myofibroblasts and extracellular matrix components. Growing evidences support that cysteine cathepsins, embracing cathepsin B (CatB) that affects TGF-β1-driven Smad pathway, along with their extracellular inhibitor cystatin C, participate in myofibrogenesis. Here we established that curcumin, a potent antifibrotic drug used in traditional Asian medicine, impaired the expression of both α-smooth muscle actin and mature TGF-β1 and inhibited the differentiation of human lung fibroblasts (CCD-19Lu cells). Curcumin induced a compelling upregulation of CatB and CatL. Conversely cystatin C was downregulated, which allowed the recovery of the peptidase activity of secreted cathepsins and the restoration of the proteolytic balance. Consistently, the amount of both insoluble and soluble type I collagen decreased, reaching levels similar to those observed for undifferentiated fibroblasts. The signaling pathways activated by curcumin were further examined. Curcumin triggered the expression of nuclear peroxisome proliferator-activated receptor γ (PPARγ). Contrariwise PPARγ inhibition, either by an antagonist (2-chloro-5-nitro-N-4-pyridinyl-benzamide) or by RNA silencing, restored TGF-β1-driven differentiation of curcumin-treated CCD-19Lu cells. PPARγ response element (PPRE)-like sequences were identified in the promoter regions of both CatB and CatL. Finally, we established that the transcriptional induction of CatB and CatL depends on the binding of PPARγ to PPRE sequences as a PPARγ/Retinoid X Receptor-α heterodimer.


Results and Discussion
Curcumin inhibits the TGF-β1-induced differentiation of human lung fibroblasts. Given both the scarcity of human clinical specimens (e.g. lung IPF biopsies) and the discordance in the expression of lung cathepsins and their endogenous inhibitors between murine bleomycin-induced fibrosis and human pulmonary fibrosis 40 , human CCD-19Lu fibroblasts appeared to be a pertinent model to decipher molecular mechanisms and signaling pathways associated with human pulmonary myofibrogenesis 29 . A hallmark in the development of fibrosis is the TGF-β1-dependent activation, proliferation and differentiation of fibroblasts to α-SMA-expressing myofibroblasts that secrete excessive amounts of collagen. First, we analyzed the cell viability of TGF-β1differentiated CCD-19Lu fibroblasts, following a curative treatment by curcumin (0-50 µM) for different times (24-96 h). Curcumin inhibited cell growth in a dose-dependent manner (Fig. 1a) at concentrations greater than 20 µM (IC 50 = 35 μM) after one-day exposure. We therefore chose a concentration range of 0-10 µM for all subsequent experiments since no cell toxicity was measured. Fibroblasts stimulated with TGF-β1 (5 ng/ml) were further treated with curcumin (0-10 µM) during 48 h, then total RNA was isolated and α-SMA expression level determined by real-time quantitative PCR (Fig. 1b). The expression of α-SMA mRNA in curcumin-treated (10 µM) myofibroblasts was reduced to less than 5% relative to untreated control. This was confirmed by an immunoblot analysis of the protein level of α-SMA (Fig. 1c,d) as well as by immunofluorescence (Fig. 1e). The level of α-SMA in curcumin-treated myofibroblasts was similar to that observed for undifferentiated CCD-19Lu fibroblasts. We reported previously a significant and similar reduction of α-SMA expression in the presence of CA-074Me (L-3-trans-(propylcarbamyl)oxirane-2-carbonyl)-L-isoleucyl-L-proline methyl ester), a cell permeable pharmacological inhibitor of CatB or after CatB silencing by a specific siRNA 29 , raising the question of a possible relationship between the antifibrotic effects of curcumin and the profibrotic activity of CatB. Since TGF-β1 regulates ECM (including type I collagen) deposition in the fibrotic tissue 37 , we analyzed in a second step the expression of both endogenous TGF-β1 and collagen following addition of curcumin. Curcumin significantly decreased the expression of TGF-β1 mRNA, down to levels similar to those observed in undifferentiated fibroblasts (Fig. 2a). Subsequently we observed reduced amounts of both 50-kDa TGF-β1 (proform) and its 25-kDa bioactive form (mature dimer) in the presence of curcumin (10 µM) (Fig. 2b). Additional immunoreactive bands corresponding to variant glycosylated forms were observed as reported previously (confirmed by treatment with endo-β-N-acetylglucosaminidase F1) 41 . Of note, we previously observed an opposite effect after the pharmacological inhibition of CatB or after its specific silencing, with increased amounts of the 50-kDa proform of TGF-β1 and impaired release of its 25-kDa form 29 . Taken together, our data likely support a cellular crosstalk between TGF-β1, CatB and curcumin. Otherwise, it is well established that the two genes encoding the α-1 and α-2 chains of type I collagen (Col1a1 and Col1a2) are upregulated during fibrotic processes (for reviews [42][43][44]  50% in curcumin-treated cells (Fig. 2c). Measurement of fibrillar collagen in conditioned media confirmed that curcumin significantly reduced the amounts of soluble collagen (Fig. 2d). Similarly, we observed a significant decrease of insoluble cell layer-associated collagen (Fig. 2e). Following curcumin treatment, the levels of both soluble and insoluble collagen were similar to their observed constitutive level for undifferentiated fibroblasts, attesting that curcumin down-regulated both TGF-β1 and collagen I expression and thus could impair or delay myofibrogenesis. In agreement with these results, it has been shown that curcumin inhibits both myofibrogenesis and collagen secretion in mice following bleomycin-induced lung injury 39,45 . Likewise, curcumin attenuated type I collagen production in an experimental pulmonary fibrosis in rats and showed beneficial antifibrotic effects in a rodent model of obstructive nephropathy 46,47 . Of note, a similar decrease of soluble and insoluble type I collagen that was previously observed in TGF-β1-differentiated lung fibroblasts after treatment by the cathepsin inhibitors E-64d (i.e. (2S, 3S)-trans-epoxysuccinyl-L-leucylamido-3-methylbutane ethyl ester) or CA-074Me 29 . This prompted us to evaluate the effect of curcumin on the expression of cathepsins. Three days after addition of TGF-β1, myofibroblasts were treated with curcumin (0-10 µM) for 48 h. Analysis of α-SMA expression was performed by quantitative real time PCR analysis of α-SMA transcripts. Data are expressed as percentage relative to untreated control. (c) In parallel, myofibroblasts layers were harvested and lysed. Samples were submitted to electrophoresis (12% SDS-PAGE, under reducing conditions) before the protein level of α-SMA was analyzed by western blot using a mouse α-SMA antibody. β-actin was used for load control. A representative sample of three independent experiments is shown. Full-length blots are presented in Supplementary Fig. 5

Recovery of cathepsins B and L expression and proteolytic activity in curcumin-treated CCD-19Lu fibroblasts.
In an initial work, we observed that mRNA levels of cathepsins B, K and L remained stable in the presence of TGF-β1, suggesting that these proteases are not transcriptionally regulated during TGF-β1-dependent myodifferentiation 29 . On the other hand, Brömme and coworkers observed that curcumin prevented collagen deposition and lung fibrosis in the murine model of bleomycin-induced fibrosis, and reported an upregulation of lung CatK 26 . Among the cysteine cathepsins showing significant ECM-degrading activities, CatK mRNA expression was increased in human and mice fibrotic lung tissues compared to normal lung specimens 47,48 . At this point, we did not observe a conclusive increase of CatK transcripts in TGF-β1-induced myofibroblasts in the presence of curcumin ( Supplementary Fig. 1). This discrepancy probably stems from the fact that CCD-19Lu fibroblasts are a homogenous cell line at variance with lung biopsies that comprise a heterogeneous population of fibroblasts and differentiated myofibroblasts, but also contain bronchial and alveolar epithelial cells that constitutively express CatK. On the contrary, we observed that curcumin induced a compelling transcriptional upregulation of CatB in differentiated myofibroblasts (Fig. 3a). This increased expression was even greater for CatL (~10 3 -fold increase relative to control) following addition of curcumin (10 µM). Measurements of enzymatic activities of extracellular cysteine cathepsins in culture medium (i.e. fibroblast supernatants) revealed  that this upregulation allowed the recovery of the peptidase activity of secreted cathepsins (Fig. 3b), up to levels similar to those observed in undifferentiated CCD-19Lu cells 29 . This dose-dependent rescue of cathepsin activity, as a consequence of curcumin treatment (0-10 µM), was substantiated by using the irreversible activity-based probe Biotinyl-(PEG) 2 -Ahx-LVG-DMK (with: Ahx, 6-aminohexanoic acid; PEG, polyethylene glycol, DMK, diazomethylketone) that targets the nucleophilic active site thiol of cysteine cathepsins (Fig. 3b) 49 . Western blot analysis confirmed the curcumin-induced upregulation of the protein amounts of both intracellular CatB (mainly its double-chain form) and CatL (Fig. 3c,d), along with the mature forms of secreted CatB and CatL (Fig. 3e,f).
The current results are also in line with a previous report indicating that curcumin activated the protein expression of CatL in glioma cells 50 . Based on the rationale that the inhibition of extracellular ECM-degrading cathepsins relies on their regulation by secreted cystatin C during myofibrogenesis, we further investigated the pharmacological activity of curcumin with respect to the expression of cystatin C.

Impairment of cystatin C expression in curcumin-treated CCD-19Lu myofibroblasts.
We established earlier that, during TGF-β1-driven differentiation of human lung fibroblasts, TGF-β1 promotes secretion of cystatin C and drives the cystatin C-dependent inhibition of ECM-degrading cathepsins 29 . Conversely, we found that the expression level of intracellular stefin B (a.k.a. cystatin B) did not vary during differentiation. Here, qPCR analysis of cystatin C showed that curcumin treatment led to a >two-fold decrease of its transcription (p < 0.01) while the expression of stefin B mRNA remained unchanged (Fig. 4a). Quantification of extracellular cystatin C by a sandwich ELISA confirmed this downregulation following curcumin treatment (Fig. 4b). Immunoblot analysis of both cystatin C, the most potent circulating inhibitor of cathepsins, and stefin B, their major intracellular inhibitor, confirmed that curcumin impaired the expression of cystatin C, but had no effects on stefin B (Fig. 4c,d). In line with these results, a recent study showed that curcumin reduced cystatin C in a rat model of adenine-induced chronic kidney disease 51 . Interestingly, we observed that a significant enhancement of cystatin C occurred in bronchoalveolar fluids of IPF (idiopathic pulmonary fibrosis) patients 30 , supporting that cystatin C could be a specific and valuable biomarker of lung fibrosis. Likewise, an increased expression of cystatin C was testified for other fibrotic disorders (cardiac, liver and oral submucous fibrosis) (for review 23 ). Although the exact molecular mechanisms yet remain to be clarified, this led us to propose that the rise of secreted cystatin C might favor the pathogenesis of lung fibrosis, since ECM remodeling, which depends on a subtle balance between synthesis and degradation, partly relies on cysteine cathepsins. Taken together with the drop of soluble and insoluble collagens, the present data sustained that curcumin may restore the cathepsins/cystatin C balance, both by weakening the expression of cystatin C and by rescuing the proteolytic activities of cathepsins. This therefore also raised the question of identifying likely molecular partners or cellular pathways that can make the link between the adjustment of cathepsins/cystatin C balance and the antifibrotic activity of curcumin.

Consequences of PPARγ overexpression in curcumin-treated myofibroblasts. It was previously
suggested that curcumin may upregulate the transcription of nuclear peroxisome proliferator-activated-receptorγ (PPARγ) (for reviews [52][53][54] ), while a distinct study proposed that PPARγ could be involved in regulating the expression level of CatL in a monocytic cell line 55 . Also, it was observed that alcohol exposure increased the activities of both PPARγ and CatB in the rat pancreas, although no link could be pinpointed between the two molecules 56 . According to these statements, CCD-19Lu fibroblasts were treated with curcumin (0-10 µM) for 48 h as described earlier to examine PPARγ expression. Curcumin caused a robust upregulation of both PPARγ mRNA (Fig. 5a) and protein (Fig. 5b). Interestingly, previous reports are in line with our present finding that curcumin specifically upregulates PPARγ. Indeed it was demonstrated that PPARγ triggering may impair NF-κB signaling pathway in relation with a reduced phosphorylation of NF-κB at Ser536 (phospho-NF-κB, p-p65 form) 53,57,58 . Immunoblot analysis showed here that curcumin treatment led to a dose-dependent decrease in NF-κB phosphorylation while the basal level of non-phosphorylated NF-κB (p65 form) remained unchanged ( Supplementary  Fig. 2a,b). Then we evaluated the consequences of PPARγ inhibition on the expression level of the profibrotic marker α-SMA. Transient knockdown of PPARγ was achieved by transfection of myofibroblasts with a specific small interfering RNA (see Supplementary Fig. 3a). Alternatively, CCD-19Lu myofibroblasts were treated with 2-chloro-5-nitro-N-4-pyridinyl-benzamide (also called AT), a selective and specific PPARγ antagonist that significantly reduced (∼80%) PPARγ expression (p < 0.05) (see Supplementary Fig. 3b). Interestingly, using a rodent model of preeclampsia, treatment by 2-chloro-5-nitro-N-4-pyridinyl-benzamide led to a similar reduction of PPARγ mRNA 59 . While the expression of α-SMA remained unchanged upon treatment with a control scrambled siRNA, silencing of PPARγ induced a ~2.5 fold increase in the expression of α-SMA mRNA (Fig. 5c). Addition of 2-chloro-5-nitro-N-4-pyridinyl-benzamide (AT) resulted in a >100-fold increased transcription of α-SMA compared to PPARγ silencing (Fig. 5d). A ~two-fold increase of α-SMA protein was also observed by western blotting following either pharmacological inhibition of PPARγ or its RNA silencing (Fig. 5e,f). Likewise, the anti-fibrotic properties of curcumin were reversed by both PPARγ inhibition or its silencing by a PPARγ shRNA by western blotting. A representative sample is shown (n = 3). White arrows indicate mature forms; black arrows correspond to pro-CatB and pro-CatL. β-actin was used for load control. Full-length blots are presented in Supplementary Fig. 5. (d) Densitometric analysis of the protein level of intracellular mature CatB and CatL (normalized data relative to control without curcumin, n = 3). (e) Two days after treatment with curcumin, the protein level of extracellular CatB and CatL was analyzed by WB. A representative sample is shown (n = 3, white arrows, mature proteases). Full-length blots are presented in Supplementary Fig. 5 using renal tubular epithelial cells (HK-2 cells) 60 . Accordingly PPARγ agonists were also shown to reduce the differentiation of myofibroblasts, as well as the production of αSMA by both human lung myofibroblasts and cat corneal fibroblasts 61,62 . Likewise, transient siRNA knockdown of PPARγ induced a ~two-fold reduction of both CatB and CatL mature forms, as revealed by immunoblot analysis (Fig. 6a,b). Conversely, RNA silencing of PPARγ did not alter the expression of cystatin C mRNA ( Supplementary Fig. 4a), while an ELISA-based analysis of CCD-19Lu supernatants demonstrated that the concentration of secreted cystatin C remained unchanged ( Supplementary Fig. 4b). Thus, although the pathway is not yet elucidated, our results strongly suggest that the link sustaining inhibition of cystatin C by curcumin was not straightforwardly related to a molecular partnership with PPARγ. Present data disclosed that PPARγ inhibition restored TGF-β1-driven differentiation of human lung fibroblasts, and supported that curcumin-dependent triggering of PPARγ could enhance synthesis of cathepsins.
The following and key question was therefore to identify the type of molecular interactions directly associating PPARγ with the overexpression of both CatB and CatL.
Upregulation of cathepsins B and L depends on binding of PPARγ to their promoter regions. Cysteine cathepsins may be controlled in various ways (namely at the transcriptional, translational, and post-translational levels), and transcriptional activation could be one of the possible mechanisms leading to cathepsin overexpression (e.g. transcription factor EB, STAT signaling pathways); also regulatory sites are found in the promoter region of some cathepsin genes including Sp1 and Sp3 binding sites, as well as interferon-stimulated response element (IRSE) 18,22,63 . The nucleotide sequences of human cathepsins were first subjected to a bioinformatic analysis (Genomatix MatInspector software, http://www.genomatix.de/; Dragon PPAR Response Element (PPRE) Spotter v.2.0, http://www.cbrc.kaust.edu.sa/ppre/) that allowed us to identify putative PPRE-like sequences in the promoter regions of both CatB and CatL (confidence score >0.9). Conversely, we did not identify hypothetical PPRE sequences in the CatK promoter region, in agreement with the lack of upregulation for CatK in fibroblasts. Consistently the expression of CatK is primarily modulated by RANKL (receptor activator of NF-κB ligand) (see 64 ). The functionality of the PPRE-like sequences for CatB and CatL was further assessed by an electrophoretic mobility shift assay (EMSA), using human recombinant PPARγ and RXRα (Retinoid X Receptor-α). Incubation of biotin-labeled CatB-or CatL-PPRE oligonucleotides with recombinant PPARγ and RXRα resulted in the formation of complexes with reduced electrophoretic mobility (Fig. 7, lane 2). These specific complexes were not observed with the mutated PPRE-like oligonucleotides (Fig. 7, lane 5) that were used as negative controls. The binding specificity of PPARγ to these PPRE sites was validated by a competitive inhibition assay with unlabeled wild-type oligonucleotides (Fig. 7, lane 3); conversely unlabeled mutated CatB-PPRE and CatL-PPRE oligonucleotides did not prevent PPARγ binding (Fig. 7, lane 4). Additionally supershift assays were performed using an anti-human PPARγ antibody, and assessed by the binding of PPARγ and RXRα to labeled consensus PPRE (positive control 65,66 ). The presence of a supershifted complexes demonstrated that both PPARγ and RXRα bound as a heterodimer to the PPRE-like sequences of the promoter regions (Fig. 7, lane 8). Of note, we consistently observed two shifted complexes with the CatB-and CatL-derived PPRE-like sequences. It could suggest the presence of distinct modes of binding with different stoechiometries 67 , which might be allowed by the greater length of the PPRE-like sequences relative to the consensus PPRE sequence that we used as positive control. Taken together, these data established that PPRE-like sequences in the promoter Concluding remarks. During these last few years we have paid a peculiar attention to the understanding of the role of cysteine cathepsins (primarily cathepsins B, L and K) in fibrotic processes. We established that TGF-β1 induces the secretion of cystatin C that -in turn -impairs the activity of extracellular matrix-degrading cathepsins, by using primary fibroblasts from IPF patients and a validated model of human lung CCD-19Lu fibroblasts. We also demonstrated a relevant increase of cystatin C in IPF bronchoalveolar lavage fluids that may reflect dysregulation of proteolytic activity in lung, and proposed that cystatin C could be used as a clinical biomarker of lung fibrosis. Otherwise Brömme and his collaborators demonstrated that curcumin, a potent anti-inflammatory and anti-proliferative nutraceutical, is an effective anti-fibrotic compound, using the murine model of bleomycin-induced lung fibrosis 26 . However, signaling pathways and accurate molecular mechanisms are still poorly described. Here we confirmed that curcumin inhibited TGF-β1-dependent lung fibroblast differentiation of human lung CCD-19Lu fibroblasts. Accordingly, the expression of the profibrotic marker α-SMA was down regulated, and the levels of both soluble and insoluble collagen were decreased to values similar to those observed with undifferentiated fibroblasts. Otherwise curcumin impaired the amount of secreted cystatin C, but not stefin B, and up-regulated the mRNA and protein expression levels of both CatB and CatL, leading to a restoration of the "cathepsins/cystatin C balance". Consequently, curcumin could promote the ECM-degrading activities of cathepsins, thereby relieving the detrimental accumulation of extracellular matrix. Potential molecular pathways were further examined. We found that curcumin participates to the up-regulation of nuclear PPARγ and identified functional PPARγ response element-like sequences in the promoter regions of both cathepsins B and L, that may elicit their transcriptional activation (see Fig. 8: synthetic diagram). However, curcumin is a multifaceted molecule that may interact with various molecular partners. For instance, the anti-inflammatory activity of curcumin is associated to the repression of signalling pathways including NF-κB, STAT3, Nrf2, and COX-2 (for review 68 ). In the present work, we have partially deciphered its pharmacological mechanism of action on human lung fibroblasts. Nevertheless, taking into account its versatile properties, we cannot exclude that its anti-fibrotic properties, which as demonstrated here closely relate to PPARγ activation that in turn triggers the expression of cathepsins, also involve other cellular signaling pathways.   Immunofluorescence. Control and treated CCD-19Lu cells were seeded into 8-well LabTek chamber slides. Cells were fixed in 4% paraformaldehyde and permeabilized with 0.3% TritonX-100 in PBS as described previously 29 . Photomicrographs were acquired by using an inverted fluorescence microscope (EVOS fl) from Advanced Microscopy Group (Mill Creek, WA, USA) at 200× magnification.

Methods
Western blot analysis. At each time point, culture media were harvested in the preservative buffer A (0.1 M sodium acetate buffer, pH 5.5), containing protease inhibitors (0.5 mM Pefabloc, 0.5 mM EDTA, 1 mM S-methyl thiomethanesulfonate, 0.04 mM pepstatin A). The culture medium was centrifuged for 5 min (10,000 g, 4 °C) to remove cell debris and concentrated 50-fold (Vivaspin concentrator tube, exclusion limit 2,000, Sartorius AG, Göttingen, Germany). Cell layers were washed once in ice-cold PBS and harvested by scraping in ice-cold buffer A. Following three freeze-thaw cycles using liquid nitrogen, soluble proteins were retrieved by centrifugation for 10 min (10,000 g, 4 °C). The remaining membrane pellets were resuspended in buffer A and stored at −80 °C. Protein concentrations were determined by Bradford assay (Bio-Rad). Samples were prepared in Laemmli buffer and boiled for 5 min. Concentrated culture media (60 μg/well) and cell layer lysates (10 μg/well) were loaded onto 12% SDS-PAGE, and electrophoresis was carried out under reducing conditions. Prestained molecular weight standards (Precision Plus Protein Standards) were supplied by Bio-Rad. The separated proteins were transferred to a nitrocellulose membrane (Amersham Biosciences, Buckinghamshire, UK). The membranes were blocked with 5% nonfat powdered milk in PBS, 0.1% Tween 20 (PBS-T). Following incubation with the primary antibodies (overnight at 4 °C under agitation), the secondary antibodies (1:5000) were added for 1 h at room temperature. Proteins were visualized by chemiluminescence (ECL Plus Western blotting detection system; Amersham Biosciences) according to the manufacturer's instructions. Constant loading in proteins was checked by incubation with a monoclonal anti-β-actin antibody. Bands were quantified by densitometric analysis using the ImageJ software (National Institutes of Health, Bethesda, MD, USA).