The α_vβ₁ integrin plays a critical in vivo role in tissue fibrosis

Abstract

Integrins are transmembrane heterodimeric receptors that contribute to diverse biological functions and play critical roles in many human diseases. Studies using integrin subunit knockout mice and inhibitory antibodies have identified important roles for nearly every integrin heterodimer and led to the development of a number of potentially useful therapeutics. One notable exception is the α_vβ₁ integrin. α_v and β₁ subunits are individually present in numerous dimer pairs, making it challenging to infer specific roles for α_vβ₁ by genetic inactivation of individual subunits, and α_vβ₁ complex–specific blocking antibodies do not yet exist. We therefore developed a potent and highly specific small-molecule inhibitor of α_vβ₁ to probe the function of this understudied integrin. We found that α_vβ₁, which is highly expressed on activated fibroblasts, directly binds to the latency-associated peptide of transforming growth factor–β₁ (TGFβ₁) and mediates TGFβ₁ activation. Therapeutic delivery of this α_vβ₁ inhibitor attenuated bleomycin-induced pulmonary fibrosis and carbon tetrachloride–induced liver fibrosis, suggesting that drugs based on this lead compound could be broadly useful for treatment of diseases characterized by excessive tissue fibrosis.

INTRODUCTION

Integrins are present in nearly all multicellular organisms and play a conserved role in mediating cell adhesion to fixed extracellular ligands and in the maintenance of tissue integrity (1). In invertebrates, a surprisingly small number of integrin heterodimers mediate these diverse functions (2, 3). Much has been learned about the critical in vivo functions of most members of the integrin family through the use of mice with global or conditional inactivating mutations of individual subunits (4, 5) and through the use of heterodimer-specific blocking monoclonal antibodies (6, 7). One major exception is the α_vβ₁ integrin. This integrin, first identified biochemically more than two decades ago (8), is composed of α and β subunits that are both present in multiple heterodimers (5 in the case of α_v and 12 in the case of β₁) (1), which has made it difficult to generate heterodimer-specific antibodies or to infer function from gene knockout studies. As a result, this integrin has been largely ignored.

We and others have shown that two members of the integrin family, α_vβ₆ and α_vβ₈, have as their principal ligands latency-associated peptides (LAPs) of the growth factors TGFβ₁-3 (transforming growth factor–β₁-3) (9–11) and that these integrins play major roles in activation of latent forms of this growth factor that are stored in the extracellular matrix in most healthy adult tissues. In mice, inactivation of both of these integrins recapitulates all of the developmental phenotypes of loss of TGFβ₁ and TGFβ3 (12). Inhibitors of each of these integrins have identified important and distinct roles for each in multiple disease models and have provided new options for therapeutically targeting TGFβ in specific contexts, thereby avoiding potentially undesirable side effects of globally inhibiting this pleiotropic growth factor (9, 13–16).

However, in contrast to development, it is clear that there are a number of important pathologic circumstances in adults where inhibition of TGFβ is therapeutically effective, but inhibition of α_vβ₆ and α_vβ₈ is not. One of these is hepatic fibrosis (17). We recently used cremediated deletion of the integrin α_v subunit in activated fibroblasts to demonstrate that loss of all α_v integrins from these cells protects mice from fibrosis in multiple organs, including the liver, and that this effect was associated with reduced tissue TGFβ signaling (17). Tissue fibroblasts can express four α_v-containing integrins, α_vβ₁, α_vβ₃, α_vβ₅, and α_vβ₈. We found that individual deletion of α_vβ₃, α_vβ₅, or α_vβ₈ integrin either globally or conditionally in activated fibroblasts (in the case of α_vβ₈ integrin) had no effect on organ fibrosis but were unable to examine any possible contributions of the α_vβ₁ integrin because of the lack of suitable experimental tools. Our previous results could thus have been explained either by redundancy of α_v integrins (the interpretation we favored) or by a specific role for fibroblast α_vβ₁ in driving fibrosis.

To begin to identify important functions for the α_vβ₁ integrin, we used information from the solved crystal structure of other α_v and β₁ integrins (18, 19) and from the design of other small-molecule inhibitors targeting integrins (20), to develop a potent and specific small-molecule inhibitor of the α_vβ₁ integrin. We then used this inhibitor to demonstrate a previously unknown role for this integrin in activating the growth factor TGFβ and in driving tissue fibrosis in the lung and liver.

RESULTS

Design and synthesis of an α_vβ₁ integrin–specific inhibitor

Starting with a base compound that specifically binds to the α_v subunit in α_vβ₃ integrin, we looked to impart β₁ subunit–binding specificity through addition of a sulfonamidoproline moiety we had previously shown to bind to the β₁ subunit in α₂β₁ integrin (Fig. 1A, blue shading), which occupies a hydrophobic pocket in the β₁ chain (20).

Fig. 1. α_vβ₁ integrin inhibitor design and demonstration of potency and specificity.

(A) Design principle of α_vβ₁ integrin inhibitor by combining a positively charged guanidine moiety in α_vβ₃ integrin inhibitor (blue shading) and a sulfonamidoproline moiety in α₂β₁ integrin inhibitor (green shading). (B) Structures of α_vβ₁ integrin–specific inhibitor c8. (C) Docking model of α_vβ₁ integrin inhibitor c8 bound to α_vβ₁ integrin, where α and β subunits are, respectively, in green and blue. The model predicted that a linker length of n = 3 or 4 would have the highest affinity. (D) Dose-dependent cell adhesion assay of c8 against all α_v and related integrins. (E) Curve-fitted IC₅₀ of c8 against RGD (arginine–glycine– aspartic acid)–binding integrins in cell adhesion assay. Data represent means ± SEM; n = 3 (three experimental replicates).

Because cocrystal structures are available for the ligand-binding regions of the α_vβ₃ and α₅β₁ integrins (18, 19), we were able to construct a computational model of the α_vβ₁ integrin to further guide our inhibitor design (Fig. 1C). We then synthesized a small set of compounds, including the α_v-binding base compound and the β₁-binding sulfonamidoproline moiety separated by amide linkers of various lengths, and found outstanding geometric and electrostatic complementarity when these were docked to our model of the α_vβ₁ integrin. Potency and specificity of each compound were tested by performing cell adhesion assays with a panel of cell lines and ligands designed to isolate adhesion mediated by individual integrin heterodimers (Fig. 1D, fig. S1, and table S1). The most promising compound, c8 (Fig. 1B), is one of two compounds predicted to have the highest affinity based on its excellent fit to the modeled integrin structure. Indeed, c8 inhibited α_vβ₁ integrin– mediated cell adhesion to the previously identified ligand fibronectin (8), with a sub-nanomolar median inhibitory concentration (IC₅₀), but only minimally inhibited binding mediated by other related integrins up to concentrations five orders of magnitude higher (Fig. 1E and table S1). A related compound, c6, performed similarly well in binding assays (fig. S2). The remarkable specificity obtained for c8 and c6 demonstrated the effectiveness of our chimeric design strategy and provided us with novel reagents to examine α_vβ₁ integrin function.

Identification of widespread expression of the α_vβ₁ integrin on fibroblasts and a critical role for this integrin in activation of latent TGFβ by these cells

To determine the functional relevance of the α_vβ₁ integrin, we began by examining its possible role in the process of tissue fibrosis (12). Fibrosis is a critical contributor to many chronic diseases that eventually lead to organ failure. Despite the societal burden of fibrotic diseases, there are currently few approved therapies. As we reported for hepatic stellate cells, primary murine lung fibroblasts and human fetal and adult lung fibroblasts, as well as lung fibroblasts from patients with idiopathic pulmonary fibrosis (IPF), all clearly expressed the α_vβ₁ integrin as determined by immunoprecipitation (IP) of α_v, followed by Western blotting for β₁ (Fig. 2A and fig. S5). In contrast, whereas both the α_v and β₁ subunits were easily detectible in primary endothelial and epithelial cells, coimmunoprecipitation did not detect the α_vβ₁ heterodimer in these cells.

Fig. 2. α_vβ₁ is expressed on pulmonary and hepatic fibroblasts and mediates adhesion to TGFβ₁ LAP and activation of latent TGFβ.

(A) Coimmunoprecipitation and Western blot reveal expression of α_vβ₁ heterodimers in human and murine cell lines from the liver and lung. Cell lysates were immunoprecipated with antibodies to α_v (antibody RMV-7 for murine cells or L230 for all other cells), followed by Western blotting for either α_v (lower panels, to control for capture, antibody 611012) or β₁ (upper panels, antibody 04-1109). nhLu fb control (normal human lung fibroblasts from an uninjured control subject); IPF fb (lung fibroblasts isolated from a patient with IPF); mLu fb (murine lung fibroblasts); mHSC (murine hepatic stellate cells); WI38 (diploid human lung fibroblast); CHO [WT (wild-type)] control α₅-deficient CHO cells (which lack expression of α_vβ₁); CHO (α_v) (α₅-deficient CHO cells engineered to express the α_vβ₁); hAT2 (human alveolar type 2 cells, which lack expression of α_vβ₁); hPAEC (human pulmonary artery endothelial cells, which lack expression of α_vβ₁). IgG, immunoglobulin G. (B) WT CHO cells (lacking α_vβ₁) adhere poorly, whereas CHO cells with forced expression of α_vβ₁ (CHO α_v) and WI38 cells strongly adhere to TGFβ₁ LAP. (C) WI38 cell adhesion to TGFβ₁ LAP (0.3 µg/ml) is inhibited by c8 (IC₅₀ = 0.72 nM). (D) c8 treatment reduced activation of TGFβ by cells expressing α_vβ₁ (IC₅₀ range, 0.35 to 0.50nM). Fibroblasts (as indicated) were cocultured with TGFβ reporter (PAI1-luciferase) cell line in the presence of a range of concentrations of c8. Data represent means ± SEM, n = 3 (three experimental replicates).

We and others have reported that the closely related α_v integrins α_vβ₆ (9) and α_vβ₈ (11) can each bind to an N-terminal fragment of the TGFβ₁ and TGFβ3 gene products called the LAP, which normally forms a noncovalent complex with the active cytokine, preventing TGFβ from binding to its receptors and inducing biological effects. When mechanical force is applied to the latent complex by contraction of integrin-expressing cells, the resultant conformational change leads to release of active TGFβ₁ (21).Our previous work suggested that an α_v integrin on fibroblasts contributed to tissue fibrosis by binding to and activating TGFβ (17). To determine whether the relevant fibroblast integrin could be α_vβ₁, we performed cell adhesion assays with either primary fibroblasts, control α₅-deficient Chinese hamster ovary (CHO) cells or α₅-deficient CHO cells engineered to express the α_vβ₁ integrin (8). Both fetal lung fibroblasts (WI38 cells) and α_vβ₁ integrin–expressing CHO (CHO α_v) cells efficiently adhered to a range of concentrations of TGFβ₁ LAP, whereas control CHO (CHO wild-type) cells did not (Fig. 2B and table S2). This adhesion could be inhibited by c8 with a similar IC₅₀ as that shown for inhibition of adhesion to fibronectin (Fig. 2C and table S3). Adhesion to LAP could be inhibited by antibodies to either β₁ or α_v, but not with antibodies to either α_vβ₃, α_vβ₅, α_vβ₆, or α_vβ₈ (fig. S3). We used a coculture system to quantitatively assess TGFβ activation (22) and found that c8 potently and specifically inhibited activation of TGFβ by CHO cells engineered to express the α_vβ₁ integrin and by a variety of primary murine and human lung fibroblasts, including fibroblasts from the lungs of patients with IPF, and murine hepatic stellate cells (Fig. 2D and table S4). The IC₅₀ for c8-mediated inhibition of TGFβ activation by these cells ranged from 0.35 to 0.50 nM, close to the IC₅₀ of 0.75 nM for inhibition of WI38 cell adhesion to LAP (Fig. 2C). Together, these data suggest that α_vβ₁ is the principal integrin on fibroblasts responsible for adhesion to TGFβ₁ LAP and for activation of latent TGFβ. A biochemical association between the α_vβ₁ integrin and TGFβ₁ LAP was previously reported on the basis of affinity chromatography (23), but this is the first evidence that contractile cells can use this integrin to activate TGFβ.

In vivo role of the α_vβ₁ integrin in liver and lung fibrosis

Having established that c8 inhibits fibroblast adhesion to TGFβ₁ LAP and TGFβ activation by fibroblasts from multiple organs, we used this inhibitor to examine the role of α_vβ₁ integrin–mediated TGFβ activation in two widely used models of pathologic tissue fibrosis: carbon tetrachloride (CCl₄)–induced hepatic fibrosis and bleomycin-induced pulmonary fibrosis. To determine whether α_vβ₁ might be a promising therapeutic target, we waited until fibrosis was established in each model (21 days for CCl₄ and 14 days for bleomycin) and then began continuous subcutaneous administration of either c8 or an inactive control compound, c16. In each model, c8 caused a substantial and significant reduction in fibrosis, as measured by either total area of collagen staining or total organ hydroxyproline content (a measure of cross-linked collagen) (Fig. 3, A to F, and tables S5 and S6). These results are highly comparable to the protection we previously reported in mice with conditional deletion of all α_v integrins from activated fibroblasts (17), suggesting that α_vβ₁ was probably the major α_v integrin responsible for that effect.

Fig. 3. c8 protects from liver and lung fibrosis.

(A) Effects of c8 or c16 on mouse fibrosis model (liver). c8 or c16 (inactive control compound) was continuously delivered by Alzet pump beginning 3 weeks after intraperitoneal administration with oil (sham) or CCl₄ to induce liver fibrosis. (B and C) Treatment with c8 significantly reduced liver fibrosis, as determined by (B) Sirius red staining (collagen deposition) of liver tissue after olive oil (top panels) or CCl₄ treatment (bottom panels), (C) digital image analysis quantification of collagen staining, and hydroxyproline analysis. Sirius red (liver) n = 8; P = 0.0005. Hydroxyproline (liver) n = 10; P < 0.0001. (D) Effects of c8 or c16 on murine fibrosis model (lung). c8 or c16 was continuously delivered to mice using Alzet pumps beginning 14 days after intratracheal instillation of bleomycin (Bleo) to induce pulmonary fibrosis or water (H₂O). (E and F) Treatment with c8 significantly reduced lung fibrosis, as determined by (E) Sirius red staining (collagen deposition) of lung tissue after water (top panels) or bleomycin treatment (bottom panels), (F) digital image analysis quantification of collagen staining, and hydroxyproline analysis. Sirius red (lung) n = 10; P = 0.0022. Hydroxyproline (lung) n = 10; P = 0.0027. Data represent means ± SEM. Scale bars, 100 µm. P values were calculated using the unpaired Student’s t tests.

Likewise, phosphorylated Smad3 (pSmad3) fluorescence intensity mapping revealed significant down-regulation of canonical (Smadmediated) TGFβ signaling (Fig. 4), suggesting that antifibrotic effects of c8 are due to inhibition of α_vβ₁ integrin–mediated TGFβ activation in vivo.

Fig. 4. Reduced pSmad3 in mice treated with c8.

(A) Representative liver (top panels) and lung (bottom panels) sections from mice treated with c16 or c8 after induced fibrotic injury stained for fibroblasts [PDGFRβ (plateletderived growth factor receptor β), green] and pSmad3 (red). (B) Quantification of pSmad3 nuclear intensity within individual PDGFRβ⁺ cells documents a significant reduction in fibroblast-specific pSmad3 in fibrotic mice treated with c8. Data represent means ± SEM; n = 102 (Bleo c8), n = 254 (Bleo c16), n = 262 (CCl₄ c8), and n = 361(CCl₄ c16). For both comparisons, P < 0.0001 (shown is a representative example of the distribution of individual pSmad3 mean fluorescence intensities in PDGFRβ⁺ cells, and the average of these means, for a single sample condition). Scale bar, 100 µm. P values were calculated using the unpaired Student’s t tests.

DISCUSSION

Here, we used a structure-based design to generate the first highly potent and specific inhibitor of the α_vβ₁ integrin. Our results demonstrating a difference of more than five orders of magnitude between concentrations of this inhibitor required to inhibit α_vβ₁ compared to six other integrins that all recognize ligands containing the same arginine–glycine–aspartic acid tripeptide suggest that this reagent is highly selective and should be broadly useful in identifying specific functional roles for α_vβ₁ in vitro and in vivo. Our finding that α_vβ₁ is the major integrin on several different primary fibroblasts responsible for binding to TGFβ₁ LAP and for mediating activation of latent TGFβ by these cells clarifies several previous reports of integrin-mediated TGFβ activation by contractile fibroblasts (17, 24). Our findings that this α_vβ₁-specific small-molecule inhibitor results in the same degree of reversal of liver and lung fibrosis as we previously reported for deletion of all α_v integrins from fibroblasts provide the first convincing evidence that the α_vβ₁ integrin is the major integrin on pathologic fibroblasts responsible for activating latent TGFβ and driving tissue fibrosis in multiple organs. Previous studies have identified expression of α_vβ₁ on a variety of immortalized cell lines, including A549 (23) and human embryonic kidney cells (293 cells) (25), but the ligand or ligands recognized by this integrin have been controversial, and the functional significance of α_vβ₁ has been difficult to determine in the absence of effective and specific inhibitors. α_vβ₁ has thus been alternately described as a receptor for vitronectin (25), fibronectin (8), and the adenovirus penton base (26). One previous study used affinity chromatography of surface-labeled lysates of A549 cells passed over a column composed of recombinant TGFβ₁ LAP cross-linked to Sepharose beads and found that α_vβ₁ could bind to TGFβ₁ LAP, but despite performing coculture TGFβ activity bioassays like the one we used here, the authors could not demonstrate evidence of TGFβ activation (23).

Recent work overexpressing integrin subunits and latent TGFβ in 293 cells also did not demonstrate evidence of TGFβ activation by α_vβ₁ (27). To try to understand the apparent differences between our results and these previous reports, we performed cell adhesion assays and coculture TGFβ bioassays with A549 and 293 cells using the same methods and reagents we used to study CHO cells and fibroblasts (fig. S4). We confirmed that A549 cells adhere well to TGFβ₁ LAP and also found evidence of TGFβ activation. Both effects were potently inhibited by concentrations of c8 as low as 1 nM. However, despite confirming that 293 cells express α_vβ₁, we did not find meaningful adhesion to TGFβ₁ LAP or TGFβ activation by these cells. Future studies examining why 293 cells are incapable of activating TGFβ using this integrin could provide insight into additional mechanisms of regulation of integrin-mediated TGFβ activation.

An important limitation of the current study is that we did not definitively prove that inhibition of TGFβ activation is the only mechanism by which inhibition of α_vβ₁ decreases the magnitude of tissue fibrosis. Although our data suggest that TGFβ activation is one important function of α_vβ₁, it certainly remains possible that other as yet unidentified functions of this integrin also have important effects on tissue fibrosis. The mechanisms that regulate formation and surface expression of the α_vβ₁ heterodimer also need to be identified, because both α_v and β₁ monomers are expressed in nearly every cell, but only a minority of cells ever express the α_vβ₁ heterodimer on their surface. Finally, our finding that 293 cells can express α_vβ₁ but are incapable of either binding to TGFβ₁ LAP or activating latent TGFβ suggests that there are additional positive and/or negative regulators of the interaction between α_vβ₁ and TGFβ that remain to be identified.

The significant therapeutic effect of our small-molecule α_vβ₁ integrin inhibitor in two different models of fibrotic disease suggest that this integrin could be a promising target for treatment of the multiple chronic diseases characterized by excessive tissue fibrosis. c8 also should be a useful tool to identify additional roles for this understudied member of the integrin family in modulating cell behavior and in vivo biology.

MATERIALS AND METHODS

Study design

The overall goals of this study were to develop and characterize a small-molecule inhibitor of the α_vβ₁ integrin and to assess the possibility that this integrin might be a reasonable therapeutic target for fibrotic diseases affecting multiple organs. To assess the potency and specificity of the candidate inhibitors that we developed, we used in vitro cell adhesion assays and TGFβ activation assays. All in vitro experiments were done at least three times with at least triplicate samples.

To evaluate the potential efficacy of our lead small-molecule c8 in vivo, we compared the effects of c8 to those of the inactive peptide analog c16 in mice treated either with vehicle alone or with the fibrogenic stimuli bleomycin (for induction of pulmonary fibrosis) or CCl₄ (for induction of liver fibrosis). Sample size was selected on the basis of extensive experience with each model to allow us to reliably detect a 50% reduction in the magnitude of fibrosis. Treatments were all delivered by continuous subcutaneous infusion beginning at a time point at which we and others had previously demonstrated that fibrosis was already present. Mice were randomized for inclusion in each of the four treatment arms, and all analyses were performed blind to inciting stimulus and treatment. All surviving mice were included for analysis.

Synthesis of inhibitors

Preparation of (PhSO₂)Pro-(pNO₂)Phe-OtBu

Diisopropylethylamine (DIPEA) (1 eq), hydroxybenzotriazole (HOBt) (1 eq), and 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (EDC•HCl) (1 eq) were added to benzenesulfonylproline (40 mmol) in anhydrous dichloromethane (DCM) (100 ml) at 0 °C, and the solution was stirred for 1 min. Then, nitrophenylalanine t-butylester in DCM (300 ml) was added at the same temperature and warmed to room temperature (rt) and stirred overnight. The reaction mixture was washed with 1 N hydrochloric acid (HCl), 10% sodium bicarbonate (NaHCO₃), and brine successively. The volatiles were removed under reduced pressure, and the crude product was used without further purification (yellowish foam, 70%). ¹H NMR (nuclear magnetic resonance) [dimethylsulfoxide (DMSO)–d₆, 300 MHz] δ ppm 8.4 (1H, s), 8.25 (2H, br s), 7.81-7.57 (7H, m), 4.49 (1H, s), 4.09 (1H, s), 3.34-3.12 (5H, m), 1.58 (3H, br s), and 1.38 (9H, s); LRMS (low-resolution mass spectrometry) [ESI⁺ (positive electrospray ionization)] 526.4 (MNa⁺).

Preparation of (PhSO₂)Pro-(pNH₂)Phe-OtBu

A slurry of Pd/C (palladium on carbon) (0.1 eq based on weight) in MeOH (methanol)/H₂O (1:1, 10 ml) and ammonium formate (9 eq) were added to a solution of the above nitrobenzene (27.9 mmol) in MeOH/THF (tetrahydrofuran) (1:1, 300 ml) at rt. The mixture was heated at 60 °C for 3 hours and cooled to rt. The mixture was filtered through a pad of Celite and concentrated. The crude residue was dissolved in ethyl acetate and washed with saturated NaHCO₃ solution, dried over sodium sulfate (Na₂SO₄), and concentrated under reduced pressure. The crude product was used without further purification (off-white foam, 70%). ¹HNMR (CDCl₃, 300 MHz) δ ppm 7.82 (1H, d, J = 9.6 Hz), 7.62-7.53 (3H, m), 7.26-7.22 (1H, m), 6.92 (1H, d, J = 10.7 Hz), 6.57 (1H, d, J = 10.8Hz), 4.65-4.54 (1H, m), 4.2-4.05 (1H, m), 3.7-3.3 (m, 2H), 3.2-3.0 (2H, m), 2.95-2.85 (1H, m), 2.05 (1H, br s), and 1.6-1.3 (11H, br s); LRMS (ESI⁺) 496.2 (MNa⁺), 474.5 (MH⁺).

Preparation of (PhSO₂)Pro-(pNHCOCH₂CH₂CH₂CH₂NHC= NBocNHBoc)Phe-OtBu

DIPEA (1 eq), HOBt (1 eq), and EDC•HCl (1 eq) were added to the bis-Boc–protected aminopentanoic acid (11.2 mmol) (28) in anhydrous DCM (80 ml) at 0 °C, and then the amine from the previous step (1 eq) in DCM (20 ml) was added at the same temperature, warmed to rt, and stirred overnight. The reaction mixture was washed with 1 N HCl, 10% NaHCO₃, and brine successively. The solvent was removed under reduced pressure and purified by silica gel column chromatography (DCM/MeOH, 2% to 5%) to give the off-white foam (74%).

Preparation of c8

The bis-Boc–protected compound prepared from the previous step (8.3 mmol) was treated with 5% anisole in trifluoroacetic acid (TFA) (30 ml) for 3 hours at rt. The volatiles were removed under reduced pressure, and the crude product was washed with diethyl ether several times. The TFA salt was exchanged to chloride using 4 M HCl in dioxane and evaporation under reduced pressure (three times). The final salt was passed through a short pad of silica and dried under reduced pressure. ¹H NMR (TFA salt) (300 MHz, DMSO-d₆) ppm δ 9.86 (s, 1H), 8.12 (d, J = 7.8 Hz, 1H), 7.81 (d, J = 6.9 Hz, 2H), 7.70 (d, J = 7.2 Hz, 1H), 7.62–7.15 (m, 7H), 4.41 (s, 1H), 4.15 (s, 1H), 3.33 (s, 1H), 3.1–2.92 (m, 5H), 2.3 (s, 2H), and 1.59–1.48 (m, 8H); LRMS (ESI⁺) 559 (MH⁺).

Mice

Wild-type C57BL/6 mice were purchased from Jackson Laboratories. Eight- to 12-week-old sex-matched mice were housed under specific pathogen–free conditions in the Animal Barrier Facility of the University of California, San Francisco (UCSF). All experiments were approved by the Institutional Animal Care and Use Committee of the UCSF.

Fibrosis models

Lung and liver fibrosis were induced as described previously (17). In all studies, mice were randomly assigned for each treatment and analyzed in a blinded fashion. For CCl₄-induced liver fibrosis, mice were injected intraperitoneally with sterile CCl₄ (1 µl/g body weight) in a 1:3 ratio with olive oil or olive oil (sham), twice weekly for 6 weeks. Alzet osmotic pumps (Durect) were inserted after 3 weeks of treatment to deliver either c8 or the inactive control small-molecule c16, each dissolved in 50% DMSO (in sterile water) and administered at a dose of 70 mg/kg per day. Livers were harvested 24 hours after the last CCl₄ injection. For bleomycin-induced lung fibrosis, bleomycin (3 U/kg) (Bleo) or water (control) was administered by direct airway intubation with a microsprayer (Penn-Century Inc.). Alzet osmotic pumps were inserted 14 days after treatment as above, and lungs were harvested at 28 days.

Primary cell isolation

Primary murine lung fibroblasts were isolated from 4- to 12-day-old mice using a previously reported method with minor modifications (29). Mouse lungs were removed, pooled together, and digested in an enzyme solution [Hanks’ balanced salt solution (without Ca or Mg) with type I collagenase (0.3 mg/ml) (Sigma) and trypsin (0.5 mg/ml) (Sigma)] for 1 hour with removal of dispersed cells every 10 min. Dispersed cells were passed through a sterile filter (70 µm) into Dulbecco’s modified Eagle’s medium (DMEM)–Ham’s F-12 medium (Sigma) and 10% fetal bovine serum, and undigested lung tissue was placed in fresh enzyme solution. Once digestion was complete, erythrocytes were lysed at rt for 10 min using red blood cell lysing buffer (Sigma). Cells were pelleted by centrifugation and cultured in DMEM with 10% fetal bovine serum and 1% penicillin/streptomycin. Nonadherent cells were aspirated and discarded. Primary murine hepatic stellate cells were isolated and passaged as described previously (17). Mouse liver was perfused through the inferior vena cava sequentially with liver perfusion medium (Invitrogen), 0.3% pronase (Roche), and 0.02% collagenase (Sigma). The liver was excised and minced with scissors and further digested in 0.044% pronase and 0.008% deoxyribonuclease (DNase) (Roche). The cell suspension was shaken (200 to 250 rpm) at 37 °C for 10 min and strained through a sterile filter (70 µm). To remove hepatocytes, the cell suspension was centrifuged at 90g for 2 min, the supernatant was collected, and DNase was added; this procedure was repeated twice. The supernatant was centrifuged at 700g for 7 min to collect the nonparenchymal cell fraction. Collected cells were resuspended in 10 ml of complete DMEM with 10% fetal bovine serum and 1% penicillin/streptomycin and allowed to differentiate in culture into myofibroblasts before use.

Hydroxyproline assay

Mouse lung and liver tissues were homogenized with trichloroacetic acid and incubated overnight at 110 °C in HCl. Samples were reconstituted in water, and hydroxyproline content was measured using the chloramine T assay (30).

IP and Western blotting

Cells were lysed in radioimmunoprecipitation assay buffer [50 mM tris-HCl (pH 7.4), 10 mM MgCl₂, 125 mM NaCl, and 2% NP-40]. Cell lysates were centrifuged at 14,000 rpm for 10 min at 4 °C, and the supernatant was collected. Anti-α_v antibody (20 µg) was added to the supernatant, and this was rotated at 4°C for 2 hours, followed by the addition of 30 µl of prewashed protein G–Sepharose slurry (GE Healthcare) for 1 hour at 4 °C. The beads were washed three times with phosphate-buffered saline (PBS)/protease inhibitor mixture and once with PBS only. Laemmli sample buffer was added, the samples were boiled for 5 min followed by SDS–polyacrylamide gel electrophoresis, and Western blotting was performed. The antibodies that we used for α_v IP were RMV-7 for murine cell lines and L230 for all other cell lines. For α_v Western blotting, we used 611012 (1:500) (BD Biosciences), and for β₁ Western blotting, we used 04-1109 (1:500) (Millipore).

TGFβ activation assay

Test cells were plated at 50,000 cells per well in 96-well plates together with mink lung epithelial cells expressing firefly luciferase downstream of the TGFβ-sensitive portion of the plasminogen activator inhibitor 1 promoter (15,000 to 25,000 cells per well) (22). Cells were cocultured for 16 hours, and TGFβ activity was calculated by measurement of luminescence in the presence and absence of TGFβ-blocking antibody 1D11.

Integrin-specific adhesion assays

The effects of c8 and c16 on cell adhesion mediated by α₅β₁, α₈β₁, α_vβ₁, α_vβ₃, α_vβ₅, α_vβ₆, and α_vβ₈ were measured using pairs of cell lines and ligands selected to isolate the effect of each individual integrin. For α₅β₁, we used the colon carcinoma cell line SW480 plated on fibronectin (0.3 µg/ml); for α₈β₁, we used SW480 cells transfected with human α₈ adhering to recombinant human TGFβ₁ LAP (1 µg/ml) (31); for α_vβ₁, we used CHOα_v (α₅-deficient CHO cells engineered to express α_vβ₁) adhering to fibronectin (0.3 µg/ml) (8); for α_vβ₃, we used SW480 cells transfected with human β₃ adhering to fibrinogen (1 µg/ml); for α_vβ₅, we used wild-type SW480 cells adhering to vitronectin (0.1 µg/ml); for α_vβ₆, we used SW480 cells transfected with human β₆ adhering to recombinant human TGFβ₁ LAP (0.01 µg/ml); and for α_vβ₈, we used the glioma cell line (SNB19) adhering to recombinant human TGFβ₁ LAP (1 µg/ml). In every case, we confirmed that adhesion could be inhibited by blocking antibodies to the relevant integrin (complex-specific blocking antibodies in all cases except α_vβ₁ for which we showed equivalent effects of blocking α_v). Cells were resuspended in DMEM for 15 to 30 min at 4°C with 10-fold dilutions of c8 with a starting concentration of 10 µM. Each sample was then added to triplicate wells of a 96-well plate that had been coated overnight at 4°C with the relevant ligand, washed with PBS, blocked by 1-hour incubation with 1% bovine serum albumin (BSA), and washed again. Cells were allowed to attach for 60 min at 37°C. After incubation, nonadherent cells were removed by discarding the medium and spinning the plate topside down at 500 rpm for 5 min. Cells were then fixed and stained with 40 µl of 0.5% crystal violet, 1% formaldehyde, 20% MeOH in double-distilled water for 30 min and lysed with 40 µl of 2% Triton-X in PBS. Absorbance was measured at 595 nm in a microplate reader. For all assays, concentration-response curves were constructed by nonlinear regression analysis, and IC₅₀ values were calculated using GraphPad Prism software.

Tissue staining

Paraffin-embedded sections were processed as described previously (17). Sections (5 µm) were stained with hematoxylin and eosin or with Sirius red. Pictures were taken from random fields from each section at a final magnification of ×10. Staining area was calculated by pixel counting with National Institutes of Health (NIH) ImageJ. For fluorescence microscopy, fixed livers and lungs were transferred to 30% sucrose in PBS overnight, embedded in optimum cutting temperature compound, and then cryosectioned at 5 µm. Cryosections were permeabilized and blocked with 0.3% Triton X-100 and 3% BSA in PBS. Sections were incubated with primary antibodies (rabbit anti-pSmad3, Epitomics, 1880-1, 1:100; rat anti-PDGFRβ, eBioscience, 14-1402, 1:100) overnight at 4 °C and then with fluorophore-conjugated secondary antibodies (Invitrogen). Confocal imaging was performed on a Zeiss LSM 5 Pascal microscope. pSmad immunofluorescence staining was quantified as described (32).

Statistical analysis

All results are presented as means ± SEM. Exact P values were calculated using the unpaired Student’s t test for normally distributed data using GraphPad Prism 6.0. P values and n numbers are shown in the figure legends.