Fibroblast Activation Protein (FAP) Accelerates Collagen Degradation and Clearance from Lungs in Mice*

Idiopathic pulmonary fibrosis is a disease characterized by progressive, unrelenting lung scarring, with death from respiratory failure within 2–4 years unless lung transplantation is performed. New effective therapies are clearly needed. Fibroblast activation protein (FAP) is a cell surface-associated serine protease up-regulated in the lungs of patients with idiopathic pulmonary fibrosis as well as in wound healing and cancer. We postulate that FAP is not only a marker of disease but influences the development of pulmonary fibrosis after lung injury. In two different models of pulmonary fibrosis, intratracheal bleomycin instillation and thoracic irradiation, we find increased mortality and increased lung fibrosis in FAP-deficient mice compared with wild-type mice. Lung extracellular matrix analysis reveals accumulation of intermediate-sized collagen fragments in FAP-deficient mouse lungs, consistent with in vitro studies showing that FAP mediates ordered proteolytic processing of matrix metalloproteinase (MMP)-derived collagen cleavage products. FAP-mediated collagen processing leads to increased collagen internalization without altering expression of the endocytic collagen receptor, Endo180. Pharmacologic FAP inhibition decreases collagen internalization as expected. Conversely, restoration of FAP expression in the lungs of FAP-deficient mice decreases lung hydroxyproline content after intratracheal bleomycin to levels comparable with that of wild-type controls. Our findings indicate that FAP participates directly, in concert with MMPs, in collagen catabolism and clearance and is an important factor in resolving scar after injury and restoring lung homeostasis. Our study identifies FAP as a novel endogenous regulator of fibrosis and is the first to show FAP's protective effects in the lung.

Idiopathic pulmonary fibrosis, the most common of the idiopathic interstitial pneumonias, is characterized by inexorable progressive lung injury and scarring, with eventual death within 2-4 years from the time of diagnosis in the absence of lung transplantation (1). The etiology of the disease is poorly understood, and current Food and Drug Administration-approved treatments have only limited impact on the course of the disease (2)(3)(4).
Fibroblast activation protein (FAP, 2 also known as seprase) is a 95-kDa cell surface, type II integral serine protease belonging to the post-proline dipeptidyl aminopeptidase (DPP) family (5) that is specifically induced on lung fibroblasts in patients with idiopathic pulmonary fibrosis, in particular at the leading edge of fibrosis (6). The DPP family of serine proteases cleaves amino-terminal dipeptides from polypeptides with L-proline or L-alanine at the penultimate position. FAP is unique in that it displays additional in vitro endopeptidase (7), gelatinase, and potentially collagenase activity (8,9). FAP expression is restricted, occurring at high levels on mesenchymal cells during embryogenesis (10) and then is repressed shortly after birth. In conditions associated with matrix remodeling, such as wound healing (11), fibrosis (6,12,13), and cancer (5, 14 -17), however, FAP expression is up-regulated on activated fibroblasts. FAP has also been detected on pericytes, bone marrow-derived mesenchymal stem cells (18,19), and a small population of macrophages (20,21). FAP's in vivo substrates remain unclear. Despite a lack of direct evidence, FAP is assumed to degrade ECM components, including type I collagen in vivo (9,22,23). In support of this idea, we have observed that FAP deficiency leads to increased tumor collagen content in a syngeneic transplant model of colon cancer and an endogenous K-ras-driven murine lung tumor model (24). In general, FAP expression by tumor stromal cells correlates with greater tumor aggressiveness, whereas inhibition of FAP activity curtails tumor growth and invasiveness (16, 24 -27). Not surprisingly, cancer researchers are actively exploring FAP's therapeutic potential as a stromal cell target. In regard to fibrosis, however, FAP remains a relatively understudied protein, and its place in the pathogenesis of this disease is unknown.
The studies described herein were designed to define the role of FAP in the development of pulmonary fibrosis in vivo, employing a genetic approach with global knock-in mice in which the Fap gene has been replaced by a lacZ gene that is expressed under the control of the endogenous Fap promoter (28). Two well established complementary murine models of pulmonary fibrosis, intratracheal bleomycin and thoracic irradiation (29), were used. FAP-deficient mice demonstrated increased mortality and increased lung collagen content compared with wild-type mice in both models. This phenotype was not attributable to increased myofibroblast induction, heightened collagen synthesis, or appreciable differences in MMP activity. Instead, we present evidence that loss of FAP expression directly results in defective processing of type 1 collagen and impaired ECM remodeling. In addition, although we did not find increased numbers of ␣-smooth muscle actin-positive cells by immunofluorescence staining, FAP-deficient primary mouse lung fibroblasts displayed more robust induction of a myofibroblast phenotype compared with wild-type in response to TGF-␤. Our findings are the first to demonstrate that FAP protects against the development of pulmonary fibrosis after lung injury.

Animals
Eight-to 12-week-old C57BL/6 male and female mice were purchased from Charles River Laboratories. FAP-deficient FAP LacZ/LacZ mice (28) were obtained from W. J. Rettig and A. Schnapp (Boehringer Ingelheim Pharma KG, Ingelheim, Germany) and backcrossed 12 generations to a C57BL/6 background to facilitate fibrosis studies. These FAP-null mice have been previously characterized and show no developmental nor overt adult abnormalities under homeostatic conditions (28). Mice were genotyped as described previously (24). All mice were housed in a specific pathogen-free animal facility at the Wistar Institute or at the University of Pittsburgh. The protocols used in this study were approved by the Institutional Animal Care and Use Committee at The Wistar Institute or the University of Pittsburgh, and all procedures were conducted according to ethical committee guidelines on animal welfare and the Guide for the Care and Use of Laboratory Animals (30).

Two Murine Pulmonary Fibrosis Models
Thoracic Irradiation-Mice were anesthetized and irradiated as described previously (31). In brief, 8 -12-week-old female FAP LacZ/LacZ mice and age/sex-matched controls were anesthetized with intraperitoneal xylazine/ketamine. A single fraction of 13.5 Gy was delivered to the thorax of the mice via a 250-kVp orthovoltage machine. A customized jig provided lead shielding over the animals' head/neck and abdomen/pelvis regions, exposing only the thorax to irradiation. Mice were followed for modified survival studies, and survivors were sacrificed 16 weeks after thoracic irradiation for tissue collection.
Intratracheal (i.t.) Bleomycin-8 -12-Week-old male FAP LacZ/LacZ mice and age/sex-matched controls were anesthetized with intraperitoneal ketamine/xylazine. A single dose of bleomycin (1.0 -1.75 IU/kg, depending on experiment) was administered by i.t. injection, using a STEPPER TM repetitive pipette (Tridak TM , LLC) to minimize dose variations due to pipetting error. Mice were either followed for modified survival studies or sacrificed at designated time points for tissue collection. Male mice were used as they are more bleomycin-sensitive than their female counterparts.

Hydroxyproline Assay
Collagen quantification was performed by hydroxyproline assay as previously described (24). The right lung was consistently dedicated for this assay to allow comparison. Hydroxyproline content may be converted to collagen content using the conversion factor of 1 g of hydroxyproline corresponds to 6.94 g of collagen.
Immunohistochemistry Staining for FAP and Immunofluorescence Staining for ␣-SMA FAP Immunohistochemistry-Antigen retrieval was performed on de-paraffinized lung sections using 10 mM sodium citrate buffer, pH 6.0, for 20 min at 95°C. Slides were washed at room temperature and hydrated in PBS. Endogenous peroxidase activity was then quenched with 3% hydrogen peroxide. Sections were then washed in PBS, 0.05% Tween 20 and endogenous avidin and biotin blocked using a commercially available avidin-biotin blocking kit (Vector Laboratories). Sections were incubated overnight at 4°C in biotin-conjugated sheep antihuman FAP antibody (R&D Systems; AF3715; 15 g/ml) or biotin-conjugated sheep control antibody (R&D Systems; BAF020;15 g/ml). Sections were then washed in PBS, 0.05% Tween 20, and specific signal amplification was performed using an HRP-streptavidin/biotin-XX tyramide-containing tyramide signal amplification kit (Molecular Probes), followed by detection using the Vectastain Elite ABC kit (Vector Laboratories). n ϭ 3 animals/group.
␣-SMA Immunofluorescence-Immunofluorescence staining for ␣-SMA was performed on deparaffinized lung sections as described previously (24). Prepared slides were incubated with rabbit anti-␣SMA (Abcam; ab5694; 1:100; 2 g/ml) or isotype control antibody at 4°C overnight. AlexaFluor-568 goat anti-rabbit IgG (Life Technologies, Inc.; A11036; 1:500) was used as the secondary antibody. Nuclei were stained with DAPI (Life Technologies Inc.; 300 nM) and mounted for fluorescence microscopy. An ImageJ macro was constructed to objectively quantify the amount of ␣-SMA signal found in the total area of lung injury per field. The analysis was performed in a blinded fashion by an independent observer. For each mouse, all five lung lobes were examined, with four different ϫ40 images taken per lobe; n ϭ 3-4 mice per group. The percentage of total injured lung area showing positive ␣SMA signal was quantified by computer assisted morphometry using ImageJ.

Masson Trichrome Staining
Lung sections were deparaffinized and rehydrated as described previously (24)

Generation of Antibody Specific for Murine FAP
Murine anti-murine FAP antibody 73.3 was produced and characterized in our laboratory as described (32). The specificity of this antibody was validated based on nonreactivity with tissues from FAP-null mice compared with analogous positive control tissues from wild-type mice.

Generation of Recombinant Murine FAP ECD
HEK293 cells transfected with murine FAP ECD containing a 5Ј-His and 3Ј-FLAG tag were obtained from Dr. Jonathan Cheng (Fox Chase Cancer Center, Philadelphia) and used to produce murine FAP ECD that was purified as described previously (32).
Murine FAP IP/IB-Immunoblotting for detection of murine FAP was performed after immunoprecipitation. Immunoprecipitation was performed from equal amounts of protein (1 mg of lung homogenate/sample). Samples were precleared by incubating with 50 g of isotype IgG1-conjugated agarose beads for 2 h and nutating at 4°C. The samples were then spun down at 2000 rpm at 4°C, and the supernatant was incubated with either 50 g of anti-FAP 73.3-conjugated protein A-agarose beads or 50 g of isotype IgG1-conjugated protein A-agarose beads overnight at 4°C. The samples were then spun down at 4°C at 2000 rpm; the supernatant was aspirated off, and the remaining beads were washed three times with 25 mM HEPES, pH 7.5, containing 0.1% Triton X-100, 300 mM NaCl, 0.5 mM DTT, 0.2 mM EDTA, 1.5 mM MgCl 2 , containing 20 mM ␤-glycerophosphate, 1 mM Na 3 VO 4 , 10 mM NaF, 10 mM sodium pyrophosphate, and protease inhibitor mixture (Roche Applied Science) at 25 mg/ml, followed by a final PBS wash. The beads were resuspended in 2ϫ Laemmli buffer ϩ DTT and boiled, and 20 l of each sample was resolved on an 8% SDS-polyacrylamide gel. The remainder of the protocol is as above, for immunoblotting, with the membranes incubated with primary anti-murine FAP 73.3 antibody overnight, followed by secondary HRP-goat anti-mouse antibody (Jackson ImmunoResearch).
Human FAP IP/IB-Immunoblotting for hFAP expression in our adenovirus experiments was performed after immunoprecipitation. Mouse lung samples were homogenized in immunoprecipitation lysis/wash buffer (Pierce) with complete mini-protease inhibitors (Roche Applied Science), and immunoprecipitation was performed with a commercial co-immunoprecipitation kit (Pierce) according to the manufacturer's instructions. Briefly, 20 g of mouse anti-hFAP antibody (F19, Ludwig Institute for Cancer Research) and species-matched mouse IgG isotype control (MAB002, R&D Systems) were coupled and immobilized to AminoLink Plus coupling resin included in the kit. 1 mg of the mouse lung lysates were precleared with agarose-resin from the kit and then incubated (16 h; 4°C) with antibody-bound or isotype IgG control-bound AminoLink resin. The bound protein-antibody complexes were washed with immunoprecipitation lysis/wash buffer and then eluted with the elution buffer. Samples were heated (95°C; 5 min) and separated by two-dimensional electrophoresis on 4 -12% NuPAGE BisTris gels (Invitrogen Life Technologies, Inc.) followed by immunoblotting with sheep anti-human FAP antibody (R&D Systems; AF3715; 0.5 g/ml) followed by HRPconjugated donkey anti-sheep secondary antibody (Jackson ImmunoResearch, 1:5000). All immunoblots were quantified using ImageJ.

Quantitative RT-PCR
Mouse lungs were homogenized in TRIzol (Invitrogen) and processed for RNA extraction following the manufacturer's protocol. After quantification and assessment of quality/degradation by electrophoresis, reverse transcription was performed using the standard protocol for the TaqMan reverse transcription kit (Applied Biosystems). Gene expression levels were then assayed by real time PCR on an ABI Prism 7900HT real time PCR system (Applied Biosystems) using SYBR Green reagents and procedures. The results are expressed as relative gene 1 g of total RNA was reverse-transcribed into cDNA using the high capacity cDNA reverse transcription kit (Applied Biosystems). Quantitative real time PCR was performed using the TaqMan gene expression assay on an Applied Biosystems 7900HT fast real time PCR system according to the manufacturer's instructions. TaqMan gene expression master mix (Applied Biosystems) and commercially designed human FAP (catalog no. 4331182 ID:Hs00990806_m1) and 18S (catalog no. 4333760F, for detection in mouse, rat, and human) primer/ probe sets were purchased from Applied Biosystems. ⌬CT Ct(FAP) Ϫ Ct(18S) was used to represent the expression level of human FAP in the mouse lung after adenovirus administration.

Gelatin Zymography
Mouse lungs were homogenized in MPER buffer (Pierce) in the presence of protease inhibitor mixture without EDTA (Roche Applied Science). Protein concentration was determined by BCA assay (Pierce), and 20 g of each lung homogenate were mixed with 2ϫ non-reducing sample buffer containing 2% SDS, 0.1% bromphenol blue, and 40% glycerol. The samples were incubated at room temperature for 30 min and then loaded onto 8% SDS-polyacrylamide gels containing 0.5 mg/ml gelatin (Sigma). After electrophoresis, the gels were washed four times at room temperature for 15 min with 2.5% Triton X-100 solution and then incubated overnight in activation buffer containing 50 mM Tris-HCl, 200 mM NaCl, 5 M ZnCl 2 , 5 mM CaCl 2 , and 0.02% NaN 3 , pH 7.5, for 18 h at 37°C. The gels were then stained with 0.125% Coomassie Blue in 60% methanol and 25% acetic acid for 1 h at room temperature followed by destaining in 10% acetic acid with 30% methanol until bands appeared.

Collagen Zymography
Collagenase activity in lung homogenates was assessed by collagen zymography as described in the literature (34). In brief, lungs were homogenized in MPER buffer (Pierce) in the presence of protease inhibitor mixture without EDTA (Roche Applied Science). Protein concentration was determined by BCA assay (Pierce) and 50 g of lung homogenate mixed with 4ϫ non-reducing sample buffer containing 200 mM Tris-HCl, pH 6.8, 8% SDS, 0.4% bromphenol blue, and 50% glycerol. The samples were then incubated at room temperature for 10 min and loaded onto 8% SDS-polyacrylamide gels containing 0.5 mg/ml rat tail type 1 collagen (BD Biosciences). After electrophoresis, the gels were washed twice for 1 h at room temperature with 2.5% Triton X-100 solution and then incubated in activation buffer containing 100 mM Tris, 5 mM CaCl 2 , 150 mM NaCl, 0.01% Brij-35 at pH 8.0 for 40 h at 37°C. The gels were then stained with 0.25% Brilliant Blue R-250 in 40% methanol and 10% acetic acid solution for 2 h at room temperature. The gels were then destained in aqueous 10% methanol and 10% acetic acid solution until the bands appeared.

Collagen Digests
Gels containing 2 mg/ml rat tail type 1 collagen (BD Biosciences) in PBS were made, using 0.1 M NaOH to neutralize the collagen pH and incubated at 37°C for 1 h to promote solidification (22). For each 45 g of collagen (i.e. 22.5 l of gel), 0.75 g of recombinant human MMP-1 (R&D Systems) in 12.5 mM sodium phosphate was added, and the mixture was incubated for 8 h at 37°C. GM6001 (50 M) was then added, and the sample was incubated for 15 min at 37°C to ensure inhibition of further MMP activity. Recombinant murine FAP ECD was then added in increasing amounts (0.25-5 g) to generate a doseresponse curve. In an additional sample, FAP-ECD (5 g) preincubated for 15 min at 37°C with PT630 (50 M, Point Therapeutics, Inc.), a pharmacologic inhibitor of both FAP and dipeptidyl peptidase IV (DPPIV), rather than FAP-ECD alone, was added to the sample to verify specificity of the assay. After an 8-h incubation at 37°C with FAP ECD Ϯ inhibitor, the collagen digests were prepared for electrophoresis by adding 4ϫ sample buffer and reducing agent and heating at 70°C for 10 min. Equal volumes of the collagen digests were separated by two-dimensional electrophoresis on a 4 -12% Bis-Tris gel (Invitrogen) with MES running buffer. A small aliquot of undigested, acid-solubilized type 1 rat tail collagen was also included as a reference for intact type I collagen. The soluble fraction of a collagen gel exposed only to FAP ECD for 8 h at 37°C (i.e. no prior digestion with MMP-1) was also included. The gel was then stained with SimplyBlue TM SafeStain Coomassie G-250 stain (Invitrogen), destained in nanopure water, and photographed.

FAP/TGF-␤ Cleavage Experiments
Recombinant murine FAP (R&D Systems) was purchased for these experiments. Enzymatic activity of the protein was confirmed by sequential type I collagen digests (i.e. MMP1 followed by recombinant murine FAP) as outlined above prior to proceeding with our TGF-␤ experiments. Recombinant proteins tested as potential candidates for FAP-mediated proteolytic cleavage were as follows: active recombinant human TGF-␤ (R&D Systems), recombinant latent human TGF-␤ (LTGF-␤, Cell Signaling Technology), and recombinant human latencyassociated peptide (R&D Systems). 1 g of each protein was incubated either alone, together with 2.5 g of recombinant murine FAP, or with 2.5 g of FAP preincubated with the FAP inhibitor, N-(quinoline-4-carbonyl)-Gly-Pro(F,F)-nitrile, 2 mM, at 15 min for 16 h at 37°C in 25 mM Tris-HCl, 0.25 M NaCl, pH 8.0, and a total sample volume of 25 l. Samples were then resolved on a 4 -12% Bis-Tris gel (Invitrogen) with MES running buffer, stained for protein using the SilverQuest TM silver staining kit (ThermoFisher Scientific), per the manufacturer's instructions, and photographed.

Lung ECM Isolation and Detection of Collagen Fragments
Lung ECM was isolated from the lungs of mice 10 days after i.t. bleomycin versus saline injection using a modification of a protocol for ECM isolation from murine left ventricles (35)(36)(37). Lungs were washed in wash buffer (nanopure water plus 20 mM EDTA and complete protease inhibitor mixture (Roche Applied Science)) for 30 min on a rocker at room temperature. Lungs were then decellularized over a 72-96-h period and rocking at room temperature, using several changes of decellularization buffer (1% SDS with complete protease inhibitor mixture and 20 mM EDTA in PBS). Lungs were then washed in wash buffer for 5 min three times, then overnight, rocking at room temperature Lungs were then snap-frozen, pulverized, then homogenized, and sonicated in Protein Extraction Reagent IV (Sigma) with 1ϫ complete protease inhibitor mixture. Protein concentration was determined by Bradford assay. 0.75 g of each ECM preparation was resolved by two-dimensional electrophoresis using 3-8% Tris acetate gels (Invitrogen) to better separate intact collagen forms, and 2.0 g of each ECM preparation was simultaneously resolved on 4 -12% Bis-Tris gels, to better separate out and detect smaller sized collagen fragments. Standard protocol for immunoblotting was performed with the exception that the PVDF membranes were cut roughly halfway between the 150-and 100-kDa markers to allow the two halves of the membrane to be incubated in primary antibody separately. This prevented the primary antibody from being consumed and bound preferentially to the much more abundant intact collagen forms in our preparations, allowing better detection of the intermediate-sized collagen fragments on the lower portion of the PVDF membrane. Polyclonal rabbit anti-mouse type 1 collagen antibody (Millipore) was used as the primary antibody, followed by HRP-conjugated goat anti-rabbit secondary antibody (Jackson ImmunoResearch). Experiments were repeated in triplicate for a total of n ϭ 3 per group.

Primary Mouse Lung Fibroblast Culture Experiments
Primary mouse lung fibroblasts were isolated and cultured as described previously (38). All experiments involving primary lung fibroblasts were performed three times with cells at P3-P4, using different animals for isolation of cells for each experiment. Primary mouse lung fibroblasts were plated at a density of 2 ϫ 10 5 cells/well on type 1 collagen-coated 6-well plates. The exception to this were experiments to detect collagen fragments in cell culture media; here, cells were plated on plastic. Recombinant human TGF-␤ (R&D Systems) stimulation was performed at 10 ng/ml. p-Smad2 and Smad 2/3 levels were evaluated after 90 min of recombinant human TGF-␤ exposure. ␣-SMA protein expression was assessed at 48 h (Fig. 8C) and 72 h (data not shown) of TGF-␤ exposure. Detection of type 1 collagen fragments in cell culture media was performed at 72 h of TGF-␤ exposure.

Collagen Internalization Experiments
Type I rat tail collagen (BD Biosciences) was labeled with DyLight 650 (Pierce) as described previously (39). The dye concentration was adjusted to achieve 1 dye molecule per 3-5 collagen triple helix labeling efficiency. The unreacted dye was removed by serial collagen precipitations with 0.9 M NaCl in 0.5 M acetic acid. Purified collagen was solubilized in 2 mM HCl, characterized by electrophoresis on pre-cast 3-8% Tris acetate mini-gels (Invitrogen), and collagen concentration was determined by circular dichroism in a J810 spectrometer (Jasco) using known dilutions of unlabeled type I rat tail collagen (BD Biosciences) to generate a standard curve. Gels containing 400 g/ml DL650-labeled collagen were made in 12-well tissue culture plates and allowed to dry overnight. The following day, the gels were washed with sterile water to remove excess salts and equilibrated with 1ϫ PBS followed by DMEM washes. Mouse lung fibroblasts from wild-type or FAP LacZ/LacZ mice were pretreated with 20 M E-64d (Sigma), a lysosomal inhibitor that prevents lysosomal degradation of internalized collagen, for 1 h, then seeded at 1.4 ϫ 10 5 cells/well on the collagen gels, and incubated at 37°C for 9 h in the continued presence of E-64d. Fibroblasts were similarly plated on unlabeled collagen gels to serve as appropriate negative controls. Cells were then recovered from the gels, and surface-bound collagen was removed via a collagenase IV (Sigma) digestion followed by trypsin to achieve a single-cell suspension. Cells were then plated on fibronectin-coated glass coverslips and incubated overnight in the presence of E-64d. Cells were then fixed with 3.75% paraformaldehyde and stained with Hoechst dye for nuclei and AF488-labeled F-actin antibody. Confocal microscopy was conducted using an Olympus Fluoview 1000 confocal microscope to obtain 15-18 images from five coverslips per group. Quantitation of the integrated DL650 fluorescence per image was performed using Metamorph software, and this value was divided by the nuclei per high power field to calculate DL650 fluorescence/cell for each image. Imaging of fibroblasts seeded on unlabeled collagen gels confirmed the absence of virtually any detectable autofluorescence in the DL650 channel in our controls. Experiments were performed in triplicate. The collagen internalization experiments above were also repeated in the presence or absence of a selective FAP inhibitor, N-(quinoline-4-carbonyl)-Gly-Pro(F,F)-nitrile. FAP ϩ primary mouse fibroblasts derived from wild-type mice were preincubated with either vehicle control or N-(quinoline-4-carbonyl)-Gly-Pro(F,F)-nitrile at 1 mM concentration in cell culture media for 1 h. The cells were then seeded on DL650-labeled collagen gels, and the experiments were carried forward exactly as outlined previously except that the cells were also maintained in FAP inhibitor versus vehicle control in addition to the lysosomal inhibitor, E-64d, for the duration of the experiment. Quantification of DL650 fluorescence/cell was performed using NIS Elements on 15-16 images per condition, using fixed settings and with the binary threshold set so that there was no significant signal seen on negative control images of cells seeded on unlabeled collagen gels.

Adenoviral Reconstitution of FAP Expression
The full-length human FAP gene sequence was cloned into the shuttle plasmid, pAdlox, and adenoviruses expressing hFAP (adeno-hFAP) and empty vector (adeno-Y5) were made and purified with assistance from the Vector Core Facility at the University of Pittsburgh. FAP-null and wild-type primary mouse lung fibroblasts were then transduced with adeno-hFAP and adeno-Y5. We confirmed robust FAP expression by both IPIB and FACS (data not shown). Proceeding to in vivo experiments, 10 8 pfu of adeno-hFAP versus adeno-Y5 were administered by i.t. injection in 40 l of sterile PBS to 8 -12week-old male FAP LacZ/LacZ mice and age/sex-matched C57BL/6 controls. 72 h later, 1.0 IU/kg i.t. bleomycin was administered in 40 l of sterile normal saline to each mouse (n ϭ 6 -10/group, four groups total). Mice were later sacri-ficed at 15 days post-bleomycin for hydroxyproline assay (R lung) and histological analysis (L lung). Another experiment was conducted with mice treated with 10 8 pfu of adeno-hFAP versus adeno-Y5 alone with animals sacrificed at various intervals to examine the changing kinetics of hFAP gene and protein expression over time.

Statistics
All results are expressed as mean Ϯ S.E. Statistical analysis was performed using one-way analysis of variance with the Tukey's multiple comparison test and two-tailed Student's t test (Prism 5.0, GraphPad Software). Statistical significance of survival curves was assessed using the log-rank test. p values of less than 0.05 were considered statistically significant. Western blots were quantified using ImageJ software and analyzed for relative density of bands.

Wild-type Mice Demonstrate Rapid and Sustained Up-regulation of FAP Expression in Two Murine Models of Pulmonary
Fibrosis-We verified that FAP is induced in two separate pulmonary fibrosis models. Rapid and sustained induction of Fap gene transcription occurred within 24 h of i.t. bleomycin administration in C57BL/6 wild-type mice by RT-PCR (Fig.  1A). Fap mRNA levels remained high even 14 days after injury (Fig. 1A). Similarly, Fap gene expression was up-regulated by 24 h after thoracic irradiation (XRT) and remained elevated 4 months later, the latest time point analyzed (Fig. 1B). Fap gene expression was undetectable in the lungs of FAP LacZ/LacZ mice at all time points. Up-regulation of FAP expression in wild-type mice after bleomycin was confirmed by IP/IB of whole lung homogenates (Fig. 1C) and also by immunohistochemistry (Fig.  1D). Tissue staining (Fig. 1D) demonstrates FAP expression predominantly on stromal cells rather than on the inflammatory infiltrate or alveolar epithelial cells.

FAP-deficient Mice Show Decreased Survival Compared with Wild-type Controls in Two Murine Models of Pulmonary
Fibrosis-FAP-deficient FAP LacZ/LacZ mice and age/sexmatched wild-type controls received 1.0 IU/kg i.t. bleomycin versus saline control or a single dose of 13.5 Gy thoracic irradiation versus no treatment and were followed for survival. In the bleomycin model, there was minimal mortality in the wild-type at the 1.0 IU/kg dose but 50% mortality in the FAP LacZ/LacZ mice ( Fig. 2A). FAP LacZ/LacZ mice had roughly 50% mortality at 16 weeks after thoracic XRT compared with 25% in wild-type controls (Fig. 2B).

FAP-deficient Mice Demonstrate Increased Lung Collagen Content and Pulmonary Fibrosis Compared with Wild-type Controls in Two Murine Models of Pulmonary Fibrosis-To
evaluate whether the increased mortality in FAP-deficient mice was caused by increased lung fibrosis, we assessed the lung collagen content in both pulmonary fibrosis models by two independent methods as follows: 1) trichrome stain (Fig. 3, A and B), and 2) a quantitative hydroxyproline assay (Fig. 3, C and D). Trichrome staining suggested greater lung collagen content following injury in FAP-null FAP LacZ/LacZ mice compared with wild-type mice in both models. Quantification of lung collagen content by hydroxyproline assay confirmed this. Interestingly, untreated FAP LacZ/LacZ mice demonstrated a small but statistically significant increase in lung hydroxyproline content at the 16-week time point compared with untreated wild-type controls (Fig. 3D). Similar to humans, mice develop a mild degree of interstitial thickening and lung scarring with normal aging. Although not apparent in early adulthood (see saline-treated animals, 10 -14 weeks old, Fig. 3C), FAP-deficient mice experience a slight acceleration in the gradual increase in lung collagen content and fibrosis associated with aging, a finding demonstrable at 24 -28 weeks of age (see Fig. 3D, untreated).
There Is No Apparent Difference in the Myofibroblast Population in FAP-deficient and Wild-type Mice to Account for the Difference in Phenotype-In recent years, the ␣-SMA ϩ myofibroblast has commanded the attention of researchers in the fibrosis field. With its enhanced contractile properties and exuberant matrix production, the myofibroblast has been highlighted as a key player in fibrogenesis (40 -42). We therefore investigated whether increased myofibroblast numbers in the FAP-null mice compared with wild-type could be responsible for the observed phenotype. Whole lung ␣-SMA mRNA levels were determined by RT-PCR in both fibrosis models (Fig. 4, A  and B), 7 days post-i.t. administration of bleomycin versus saline and 16 weeks post-13.5Gy thoracic XRT versus no treatment. Although ␣-SMA expression was induced by bleomycin exposure, there were no significant differences between FAPnull and wild-type control mice within treatment groups (Fig.  4A). In the XRT model, there appeared to be minimal induction of myofibroblasts (Fig. 4B); in fact, lung ␣-SMA transcript levels in treated mice trended lower than those in controls. Lung ␣-SMA immunofluorescence staining (Fig. 4, C and D) was performed to confirm these mRNA findings. This was performed at 14 days post-i.t. administration of bleomycin versus saline and 16 weeks post-13.5 Gy thoracic radiation versus no treatment. Objective quantification of percent ␣-SMA staining per total injured lung area revealed no differences between FAPdeficient FAP LacZ/LacZ mice and wild-type mice after either i.t. bleomycin or thoracic irradiation exposure (Fig. 4, E and F). The FAP-null phenotype therefore did not appear to be driven by differential induction of the ␣-SMA ϩ myofibroblast population.
Increased Lung Fibrosis and Mortality in the FAP-null Mice Cannot Be Attributed to Differences in MMP Activity or Known Profibrotic Factors or Increased Collagen Synthesis-Gene expression levels of multiple MMPs, Timp1, Pai-1, and Ctgf (known players in fibrosis and collagen catabolism and turnover) were also assessed (Fig. 5). We found no significant differences between bleomycin-treated FAP-null and wild-type mice, although in the XRT model, there was a significant decrease in Timp1 and Pai-1 transcript levels in the irradiated FAP-null mice compared with wild type. Gelatin and collagen zymography performed on whole lung lysates 10 days after bleomycin administration revealed no discernable differences in overall MMP enzymatic activity (Fig. 6). We assessed TIMP1 protein levels in the bleomycin model to see if differential TIMP1 expression could be modulating MMP activity. Although bleomycin induced TIMP1 expression, there was no There were also statistically significant differences in survival between saline versus bleomycin-treated mice in each genotype, p Ͻ 0.01. B, decreased survival in female FAP-null FAP LacZ/LacZ mice versus wild-type controls after a single dose of 13.5 Gy thoracic irradiation, * indicates p Ͻ 0.05. Mice were followed for 16 weeks after treatment. There were also significant differences in survival between untreated versus thoracic-irradiated mice in each genotype, p Ͻ 0.01. The number of animals per group is indicated in parentheses.
FAP-deficient Primary Mouse Lung Fibroblasts Display More Robust Induction of the Myofibroblast Phenotype by TGF-␤ Compared with Wild Type-It is well accepted that TGF-␤ plays a pivotal role in the development of fibrosis. Produced by multiple cell types, including T cells, macrophages, neutrophils, and fibroblasts, it is perhaps the most well known and potent pro-fibrotic cytokine, functioning in some ways like a master switch (43). We looked for evidence of enhanced TGF-␤ activation in FAP-null FAP LacZ/LacZ mice by evaluating phospho-Smad2 levels in whole lung homogenates in the bleomycin model. Although there was a significant increase in phospho-Smad2 levels between saline-and bleomycin-treated animals, there was no significant difference in phospho-Smad2 levels between the two genotypes within either treatment group (Fig.  8A). We also assessed TGF-␤ production in FAP-deficient FAP LacZ/LacZ versus wild-type primary mouse lung fibroblasts in response to wounding using a mink luciferase epithelial cell reporter assay and found no differences between cell types (data not shown).
Although there was little evidence that altered TGF-␤ signaling was responsible for the observed FAP-null phenotype at the tissue level (Fig. 8A), this likely was due to the fact that the multiple cell types present drowned out any fibroblast-specific signal. We therefore proceeded to examine more cell type-spe-cific responses and explored the effect of TGF-␤ exposure on primary mouse lung fibroblasts isolated from the lungs of FAP LacZ/LacZ versus wild-type mice. This revealed a heightened increase in phospho-Smad2 levels at 90 min (Fig. 8B) and induction of ␣-SMA at 48 h (Fig. 8C) and 72 h (data not shown) in FAP-deficient primary mouse lung fibroblasts versus wild type in response to TGF-␤ stimulation. In addition, soluble fragments of type I collagen in cell culture media were increased in FAP-deficient primary mouse lung fibroblasts after 72 h of TGF-␤ exposure compared with wild type (Fig. 8D). This most likely reflects both increased type I collagen synthesis in the presence of TGF-␤ and decreased breakdown of intermediate-sized collagen fragments in the absence of FAP. To support this hypothesis, FAP-deficient primary mouse lung fibroblasts display increased ␣-SMA levels and increased amounts of soluble type I collagen in cell culture media at baseline, prior to TGF-␤ stimulation (see CON in Fig. 8, C and D). We assessed TGF␤ receptor 1 and 2 levels as well as levels of the inhibitory Smad6 and -7, but we found no differences that could account for the observed differences in TGF-␤ responses between the two fibroblast cell types (data not shown). We then explored whether FAP could cleave and thereby activate or inactivate TGF-␤ (Fig. 8E). FAP did not cleave either active or inactive TGF-␤ to any significant degree. Examination of the protein sequence of TGF-␤ did reveal a possible PPGP cleavage site in the LAP. Incubation of recombinant murine FAP with recombinant human LAP did generate two faint new smaller There was a significant increase in lung hydroxyproline content between saline-and bleomycin-treated mice for each genotype, p Ͻ 0.01. D, quantification of right lung collagen content by hydroxyproline assay 16 weeks after 13.5 Gy thoracic irradiation versus no treatment, n ϭ 4 -8 mice per group. There was also a significant increase in lung hydroxyproline content between untreated and thoracicirradiated mice in each genotype, p Ͻ 0.01. protein bands, indicating a modest degree of LAP cleavage by FAP (indicated by arrows in Fig. 8E). These bands disappeared when FAP was preincubated with the FAP inhibitor, showing specificity of the result. The significance of this finding, however, is unclear.
We investigated whether FAP-null mice might produce more collagen in response to lung injury than wild-type mice. mRNA levels of several major collagen isoforms found in the lung (i.e. collagen 1␣1, 1␣2, and 3␣1) were significantly upregulated following injury in both pulmonary fibrosis models (Fig. 9A). However, no significant differences in lung transcript levels of these collagen isoforms were found between the two genotypes after bleomycin treatment. In the XRT model, irradiated FAP-null mice actually had lower mRNA levels of these collagen isoforms compared with their irradiated wild-type counterparts, perhaps due to a negative feedback mechanism.

FAP Participates in Type 1 Collagen Catabolism after Prior MMP-mediated Cleavage of Intact Collagen to Its 3 ⁄ 4-and 1 ⁄ 4-
Length Fragments-The data above revealed increased collagen accumulation in FAP-null mice compared with controls in the absence of evidence for increased collagen synthesis or alterations in other proteases/factors typically associated with collagen turnover. We therefore postulated that FAP itself plays an essential role in collagen proteolysis and ECM remodeling so that absence of FAP activity leads to impaired collagen break- . FAP does not regulate ␣-SMA expression following lung injury. A, ␣-SMA mRNA levels were higher 7 days after i.t bleomycin (Bleo) in both wild-type and FAP LacZ/LacZ mice compared with controls by RT-PCR, but there was no significant difference in ␣-SMA transcript levels between the two genotypes when comparing within treatment groups. ␣-SMA mRNA expression is normalized to ␤-actin. n ϭ 3/group. B, trend toward a decrease in ␣-SMA mRNA expression in mice of both genotypes by RT-PCR 16 weeks after 13.5 Gy thoracic irradiation compared with no treatment, but these differences were not statistically significant. ␣-SMA mRNA expression is normalized to ␤-actin. n ϭ 3-5/group. C, representative ␣-SMA immunofluorescence staining of lungs from FAP LacZ/LacZ and wild-type mice 14 days after i.t bleomycin. Sections from saline-treated animals are not shown as they had virtually no fibroblast-associated ␣-SMA staining. ϫ20, scale bar, 200 m. n ϭ 4/group. D, representative ␣-SMA immunofluorescence staining of lungs from FAP LacZ/LacZ and wild-type mice 16 weeks after 13.5 Gy thoracic irradiation. Sections from untreated animals are not shown as they had virtually no fibroblast-associated ␣-SMA staining. ϫ20, scale bar, 200 m. n ϭ 4/group. E, quantification of ␣-SMA staining in the bleomycin-treated mice by ImageJ. All five lung lobes from each mouse were examined with four ϫ40 images taken per lobe and analyzed in blinded fashion by an ImageJ macro designed to quantify the area positive for ␣-SMA staining within areas of lung injury. Total area positive for ␣-SMA immunofluorescence was divided by total injured area for each mouse. n ϭ 4/group. F, quantification of ␣-SMA staining in the irradiated mice by ImageJ in the same manner as described in E. n ϭ 3-4/group. down and clearance after lung injury. We analyzed the sequence of type I collagen for defined consensus FAP target sequences (endopeptidase, DGESGP and DRGETGP; DPP, PPGP) and found innumerable potential DPP cleavage sites for FAP along the length of the 1␣1 chain of the type I collagen fibril (data not shown). Analogous sites were also mapped to the 1␣2 chain of type I collagen and the 3␣1 chain of type III collagen (data not shown).
In vitro studies sought to confirm a reported lack of FAP collagenase activity (Fig. 9B) (22). Incubation of type 1 collagen gels with purified recombinant murine FAP ECD alone (Fig. 9B, lane 2) failed to yield soluble collagen fragments. Collagen processing by FAP did require prior cleavage of collagen by MMP. Although unable to release fragments from intact collagen gels, FAP readily degraded MMP-generated 3 ⁄ 4 and 1 ⁄ 4 length collagen fragments to smaller fragments in a dose-dependent manner (Fig. 9B, lanes 4 -7). The specificity of this finding was confirmed as pre-incubation with PT630, a FAP and DPPIV inhibitor (24), prevented subsequent digestion of the MMPgenerated 3 ⁄4 and 1 ⁄4 length collagen fragments by FAP (Fig.  9B, lane 8).
These data indicate a direct role for FAP, in an ordered sequence with collagenase MMPs such as MMP1, in collagenolysis. Similar results were obtained with type III collagen (data not shown). MMP and MMP-FAP collagen digests were also resolved by HPLC and the fractions analyzed by gel electrophoresis (data not shown). This analysis demonstrated some degree of preservation of the quaternary structure of collagen, particularly in the MMP-only digests while also, to a lesser extent, following sequential digestion with MMP and FAP-ECD. HPLC also confirmed generation of an array of small collagen fragments from cleavage of 3 ⁄ 4 and 1 ⁄ 4 length fragments by FAP-ECD.

Intermediate-sized Collagen Fragments, Detectable in the Lungs of Wild-type Mice after Bleomycin, Are Present in the Lungs of FAP-deficient Mice at Baseline, Indicating a Defect in
Collagen Turnover and Clearance-We sought evidence for FAP-dependent ordered proteolysis of collagen fragments in vivo. We postulated that intermediate-sized collagen fragments (i.e. 3 ⁄ 4 and 1 ⁄ 4 length fragments, for example) may persist longer in the extracellular matrix of FAP-null mice due to impaired collagen processing and turnover. To test this hypothesis, total lung ECM was isolated from mice 10 days after treatment with either saline or bleomycin via an established SDS-based decellularization protocol (37). Resolved lung ECM proteins were probed with a type I collagen antibody that recognizes varioussized collagen fragments as well as intact collagen. This analysis revealed increased intermediate-sized collagen fragments (ϳ70 -120 and ϳ30 -45 kDa) in the ECM isolated from FAPnull mice at baseline (Fig. 10A, lane 2). In contrast to lung ECM extracts from FAP-null mice, lung ECM extracts from untreated wild-type mice, which had similar levels of intact collagen, did not contain detectable levels of intermediate-sized collagen fragments at baseline (Fig. 10A, 1st lane). After bleomycin, however, intermediate-sized collagen fragments were detectable in the wild-type lung ECM extracts (Fig. 10A, 3rd  lane). Interestingly, further increase in intermediate-sized collagen fragments was not evident in the FAP-null mice after bleomycin treatment (Fig. 10A, 4th lane). Immunoblots for type I collagen in whole lung homogenates (Fig. 10B) from mice in our bleomycin experiments provided data very similar to our decellularized lung ECM extracts in Fig. 10A. Although intact collagen would not be present in these homogenates due to solubility issues, smaller, partially degraded collagen fragments should be soluble in standard lysis buffers and therefore recoverable. These blots do show an increase in type 1 collagen fragments (see ϳ60-kDa fragment) in the FAP-null mice at baseline, with a further increase in the ϳ60-kDa fragment with bleomycin treatment in both groups (Fig. 10B). These results echo what we saw in the ECM preparations made from decellularized lungs solubilized in protein extraction reagent 4 from Sigma ( Fig 10A). We again, however, saw no significant difference in type I collagen fragment burden between the two bleomycin-treated groups. Bleomycin, in addition to causing a fibrotic response, also induces a strong inflammatory reaction. We suspect that in the bleomycin-treated animals, the influx of inflammatory cells, including phagocytes such as macrophages, may partially curb the accumulation of collagen fragments to some extent and make it difficult to appreciate subtle differences in the quantities of less abundant intermediate-sized collagen fragments.
FAP ϩ Fibroblasts Demonstrate More Efficient Collagen Internalization than FAP-null Fibroblasts-Cleavage of intact collagen to smaller fragments facilitates its uptake into macrophages and fibroblasts and thereby accelerates matrix turnover. We postulated that collagen cleavage by cells expressing FAP on their surface contributes to efficient collagen internalization. Type I collagen was labeled with DyLight 650 to eliminate any issues with background autofluorescence from the primary lung fibroblasts or from collagen itself. We established that after labeling, the collagen was still amenable to cleavage by proteases (Fig 11A). FAP ϩ wild-type primary mouse lung fibroblasts and FAP-null FAP LacZ/LacZ primary mouse lung fibroblasts were seeded on DL650 collagen gels in the presence of lysosomal inhibitor E-64d and later recovered from the gels by collagenase/trypsin digest and seeded on fibronectincoated glass coverslips for confocal microscopy examination. FAP ϩ wild-type fibroblasts demonstrated greater uptake of DL650-labeled collagen by quantitative analysis (Fig. 11B). Treatment with a selective pharmacologic FAP inhibitor, N-(quinoline-4-carbonyl)-Gly-Pro(F,F)-nitrile, significantly decreased DL650-labeled collagen uptake by FAP ϩ wildtype primary mouse lung fibroblasts (Fig. 11D). We evaluated for differences in expression of Endo180 (also known as uPARAP), the major receptor through which both intact and proteolytically cleaved fibrillary collagen is internalized and cleared by fibroblasts (44 -47). Levels of Endo180 were similar between FAP ϩ wild-type and FAP-null FAP LacZ/LacZ primary lung fibroblasts (Fig 11C). Processing and cleavage of intermediate-sized (i.e. 3 ⁄4 and 1 ⁄4) collagen fragments to smaller fragments by FAP thus directly facilitates collagen internalization by macrophages and fibroblasts, promoting matrix remodeling and restoration of lung homeostasis.

Reconstitution of FAP Expression by Adenoviral Gene Delivery Rescues FAP-deficient Mice and Decreases the Degree of Lung Fibrosis after Bleomycin Back to Levels of Wild-type
Controls-Up to this point, we had shown that FAP deficiency predisposes to more severe lung fibrosis after injury in our bleomycin and thoracic irradiation models. To more directly show that FAP expression is protective in the murine lung, we performed reconstitution experiments. First, primary mouse lung fibroblasts derived from FAP LacZ/LacZ mice and wild-type controls (data not shown) were transduced with replication-deficient adenovirus expressing recombinant human FAP (adeno-hFAP) and empty vector (adeno-Y5). We confirmed robust hFAP expression 72 h later at a multiplicity of infection of 10 with further augmentation at a multiplicity of infection of 20 (Fig. 12A). No significant cellular toxicity was noted. We then established the effective dose of adeno-hFAP in vivo. 10 8 pfu of adeno-hFAP given via intratracheal injection was sufficient to cause robust hFAP expression at the mRNA (Fig. 12B) and protein (Fig. 12C) level by day 3 after adenovirus administration, which was relatively sustained, although there was some evidence that protein levels began to drop off by day 10 and definitely by day 15 (Fig. 12C, better appreciated in the FAP LacZ/LacZ IPIB, upper blot). This dose of adenovirus was well tolerated by the mice. A reconstitution experiment was then performed. FAP LacZ/LacZ mice and age/sex-matched wild-type controls were given 10 8 pfu adeno-hFAP versus adeno-Y5 via intratracheal injection followed by intratracheal bleomycin administration 72 h later. There was a significant reduction, back to levels similar to wild-type, in lung hydroxyproline content in the bleomycin-treated FAP LacZ/LacZ mice receiving i.t. adeno-hFAP compared with bleomycin-treated FAP LacZ/LacZ mice receiving adeno-Y5 (Fig. 12D), amounting to rescue of the FAP-deficient phenotype with FAP overexpression. Bleomycin-treated wild-type mice receiving adeno-hFAP had equivalent lung hydroxyproline content to bleomycin-treated wildtype mice receiving adeno-Y5, so FAP overexpression beyond a certain point did not seem to confer additional benefit. Some limitations of the adenoviral overexpression approach may have affected our findings, namely the induction of ectopic FAP expression in lung epithelial cells as well as possibly other cell types besides lung fibroblasts, and also the unavoidable acute bystander inflammation induced by the adenoviral infection itself.

Discussion
In our ever-aging population, morbidity and mortality from pulmonary fibrosis continues to rise, with overall mortality from pulmonary fibrosis now outstripping that associated with several malignancies such as bladder cancer, acute myelogenous leukemia, and multiple myeloma (48). Without question, there is a compelling need for better insight into the events that govern the conversion of what begins as a normal healing process after lung injury into an uncontrolled fibroproliferative response resulting in irreversible scarring, tissue distortion, and progressive decline in lung function.
The scientific literature supports a role for aberrant regulation of cell surface and matrix-associated proteases in the pathogenesis of pulmonary fibrosis (49 -52). MMPs (49,50,53), neutrophil elastase (54,55), and proteinases of the coagulation cascade (56 -60) have all been implicated in the disease. Although the balance of the literature to date indicates a profibrotic action of these particular proteases, we have found that FAP, a serine protease in the DPP family, exerts a protective anti-fibrotic effect in the setting of lung injury. This may at first seem counterintuitive as proteins up-regulated in disease tend to serve a pathologic role. However, the scientific literature catalogues a growing number of proteins up-regulated in disease that work to counter the disease process, reestablish homeostasis, and return the body to its original state of health (61)(62)(63), and we have found that FAP behaves in such a fashion in pulmonary fibrosis. This study is the first to demonstrate that the absence of FAP worsens the development of pulmonary fibrosis after lung injury, establishing a protective role for FAP in the lung. In designing the murine studies that ultimately led to this conclusion, we elected to conduct both thoracic irradiation and i.t. bleomycin experiments because the two fibrosis models complement each other. Bleomycin gives us insight into events earlier in the development of pulmonary fibrosis, when dysregulated lung remodeling and matrix deposition may be reversible, whereas thoracic irradiation allows us to study chronic and more advanced stages of disease. We did not necessarily expect the two models to yield the same outcomes. In the end, however, FAP-deficient mice experienced decreased survival and increased fibrosis compared with wild type in both models, which validates and strengthens our findings.
Very simplistically, fibrosis can be understood as an imbalance between collagen synthesis and collagen catabolism and clearance. Although much ongoing effort has been appropriately directed at studying the activated fibroblast/myofibroblast as the prime mediator(s) of collagen overproduction in fibrosis, less examination has been given to the competing events of matrix remodeling, collagen turnover, and scar resorption. Collagen turnover has been described to occur through two processes, somewhat interrelated. The first is the extracellular/pericellular proteolytic cleavage of collagen, classically by MMPs with collagenase (i.e. MMP1, -8, -13, and -14) followed by those with gelatinase (i.e. MMP2 and -9) activity, although other proteases likely participate as well, including FAP as indicated by our current study (64 -67). The second event involves endocytosis of collagen through interaction with cell surface receptors, in particular ␣ 2 ␤ 1 integrin and uPARAP/Endo180, with subsequent degradation of the internalized collagen within the lysosomal compartment (45, 47, 68 -70). Although intact collagen may be endocytosed, fragmentation of collagen greatly speeds its rate of internalization and clearance (44,71). Collagen internalization through uPARAP/Endo180 has been implicated in both tumorigenesis and the development of fibrosis in the liver and lung in vivo (46,47,72).
Our study establishes an important role for FAP in collagen clearance and matrix turnover. Although FAP is unable to cleave intact type 1 collagen, our in vitro assay clearly demonstrates the ability of FAP to process 3 ⁄ 4 and 1 ⁄ 4 length collagen fragments generated by prior MMP exposure into smaller degradation products, facilitating their clearance. We were able to detect intermediate-sized collagen fragments in total lung ECM preparations as well as in whole lung homogenates from FAPnull mice at baseline, although these mid-sized collagen fragments were absent in lung ECM extracts and less abundant in whole lung homogenates from saline-treated wild-type mice. These data suggest that animals lacking FAP activity have a defect in collagen catabolism even under homeostatic conditions, hence collagen fragments that are normally not present in detectable quantities or at least present at very low levels are found accumulating in the FAP-null mice. Both wild-type and FAP-null mice demonstrated the presence of intermediate-sized collagen fragments in lung ECM extracts and whole lung homogenates after i.t. bleomycin, likely related to increased collagen turnover in the setting of lung injury. In wild-type mice, we reason that these fragments appear after bleomycin treatment because the accelerated rate of collagen turnover in the setting of inflammation, scarring, and increased collagen production and fibrosis exceeds the enzymatic limits of FAP and other gelatinases present in tissues. This allows these intermediate-sized collagen fragments to be transiently seen, whereas under normal circumstances they would not be detectable.
Many might question whether FAP plays a significant role in collagen degradation in vivo, because multiple MMPs possessing gelatinase activity similarly cleave 3 ⁄ 4 and 1 ⁄ 4 length collagen fragments into smaller fragments indicating some redundancy of function. However, the detection of intermediate-sized collagen fragments in lung ECM isolated from FAP-null mice but not from B6 wild-type mice indicates that FAP indeed plays a significant role in collagen turnover in vivo, because its absence alters collagen composition. A recent study from our laboratory also supports a role for FAP in collagen turnover as CD26 tumors in FAP-null mice demonstrate dramatically increased amounts of collagen stroma as do tumors in wild-type mice in which the enzymatic activity of FAP has been pharmacologically inhibited (24). Furthermore, our finding that FAP ϩ fibroblasts more effectively internalize collagen compared with FAP-null cells confirms an important functional role for the protein in fibrogenesis. Our data are supported by a recent independent study where a genome-wide RNA interference screen in Drosophila S2 cells identified FAP as one of 22 candidate genes associated with increased collagen uptake (73). Finally, our data showing that FAP expression appears to modulate TGF-␤-mediated myofibroblast differentiation suggests that FAP may be very important in regulating the fibrogenic response after lung injury. On the basis of these results, we propose that FAP plays a larger role in ECM remodeling in vivo than has been previously appreciated, commensurate and in concert with the more widely studied MMPs.
Our study has several important implications. It points to a previously unrecognized, essential role for FAP in matrix remodeling and collagen clearance in the lung and identifies FAP as a novel endogenous regulator of fibrosis. Although a better understanding of myofibroblast biology might allow one to turn off collagen production by this cell type, the ability to accelerate collagen degradation could mean not only halting further scar formation but achieving resorption of collagen and reversal of established fibrosis. This has major implications for the field of pulmonary medicine, in particular interstitial lung disease. Having shown FAP to play a protective, homeostatic role in the lung, promoting resorption of scar by direct participation in collagen catabolism and clearance and dampening myofibroblast induction in response to TGF-␤, we now may consider ways to increase FAP expression, enhance enzymatic activity, and/or target downstream effectors in the future to try to minimize the development of lung fibrosis in certain scenarios.
Our study has important ramifications for cancer biology. FAP is already a protein of great interest in the cancer field, with many groups working on potential means of safe pharmacologic inhibition of FAP to target the tumor microenvironment or stroma. In light of our findings, however, one must proceed with somewhat heightened caution in such endeavors. Many patients with primary lung cancers or other intrathoracic malignancies undergo adjuvant radiation therapy for local tumor control and/or chemotherapy. If future treatment regimens include a pharmacologic inhibitor of FAP, we need to ensure that patients will not be at increased risk of developing radiation-induced or chemotherapy-related interstitial lung disease. A better understanding of the role of FAP in human fibrosing conditions is required.
The potential biological significance of the intermediatesized collagen fragments that accumulate in FAP-null animals at baseline and in both genotypes after lung injury should also be mentioned. Previous work by others has shown significant biological activity of peptides generated from proteolytic degradation of collagens, for example the chemotactic properties of the matrikine, PGP, or the antifibrotic actions of endostatin, a product of proteolytic degradation of collagen XVIII (61,74,75). Indeed, we did identify increased neutrophilic inflammation in BAL fluid from FAP-null mice compared with wild type in our bleomycin model as well as a more prominent inflammatory infiltrate overall in the BAL from FAP-null mice versus wild type after thoracic FIGURE 11. FAP ؉ wild-type primary lung fibroblasts internalize type I collagen more efficiently than FAP-null fibroblasts. A, DL650-labeled collagen (intact, unpolymerized, 1st lane) versus MMP-1-digested (2nd lane) versus MMP-1 followed by FAP ECD-digested (3rd lane) DL650-labeled collagen were resolved on a 4 -12% Bis-Tris gel, and the gel was imaged in the Cy5 channel by a Typhoon imager. Fluorescent labeling did not impede proteolytic degradation of the collagen. B, primary mouse lung fibroblasts derived from FAP ϩ wild-type and FAP-null FAP LacZ/LacZ mice were seeded on DL650-labeled collagen gels for 9 h, in the presence of E-64d, a lysosomal inhibitor, then recovered by collagenase/trypsin digest, seeded on fibronectin-coated glass coverslips, and examined by confocal microscopy. DL650 fluorescence/cell was quantified by Metamorph software. Red, internalized DL650-labeled collagen; blue, Hoechst dye; green, F-actin. C, cell lysates from untreated primary mouse lung fibroblasts (P3) derived from FAP ϩ wild-type and FAP-null FAP LacZ/LacZ mice were probed for the collagen receptor, Endo180; reference protein GAPDH. n ϭ 5 independent samples per group. D, pretreatment of primary wild-type mouse lung fibroblasts with a selective FAP inhibitor, N-(quinoline-4-carbonyl)-Gly-Pro(F,F)-nitrile, significantly reduces collagen internalization by FAP ϩ fibroblasts. Primary mouse lung fibroblasts derived from FAP ϩ wild-type mice were pretreated for an hour and then maintained in the selective FAP inhibitor, N-(quinoline-4-carbonyl)-Gly-Pro(F,F)-nitrile, or vehicle control at 1 mM concentration for the duration of the experiment. After pretreatment, primary mouse lung fibroblasts were seeded on DL650-labeled collagen gels for 9 h, in the presence of E-64d, a lysosomal inhibitor, then recovered by collagenase/trypsin digest, seeded on fibronectin-coated glass coverslips, and examined by confocal microscopy. DL650 fluorescence/cell was quantified by NIS Elements software. Red, internalized DL650-labeled collagen; blue, Hoechst dye; green, F-actin.
irradiation (data not shown). So beyond the simple physical accumulation of collagen and collagen fragments within the interstitium of the lung in the absence of FAP, one must also consider the possibility that the presence or absence of certain products of ECM turnover and their biological effects may contribute to enhanced fibrogenesis as well (see proposed schema, Fig. 12E). Finally, it is possible, and perhaps even likely, that FAP may cleave other essential proteins involved in the pathogenesis of pulmonary fibrosis besides collagens and influence the development of the disease through yet another mechanism.