Fibroblast activation protein increases apoptosis, cell adhesion, and migration by the LX-2 human stellate cell line

Authors

  • Xin Maggie Wang,

    1. A. W. Morrow Gastroenterology and Liver Centre at Royal Prince Alfred Hospital, Centenary Institute of Cancer Medicine and Cell Biology and The Discipline of Medicine, University of Sydney, Sydney, Australia
    Search for more papers by this author
  • Denise Ming Tse Yu,

    1. A. W. Morrow Gastroenterology and Liver Centre at Royal Prince Alfred Hospital, Centenary Institute of Cancer Medicine and Cell Biology and The Discipline of Medicine, University of Sydney, Sydney, Australia
    Search for more papers by this author
  • Geoffrey W. McCaughan,

    1. A. W. Morrow Gastroenterology and Liver Centre at Royal Prince Alfred Hospital, Centenary Institute of Cancer Medicine and Cell Biology and The Discipline of Medicine, University of Sydney, Sydney, Australia
    Search for more papers by this author
  • Mark D. Gorrell

    Corresponding author
    1. A. W. Morrow Gastroenterology and Liver Centre at Royal Prince Alfred Hospital, Centenary Institute of Cancer Medicine and Cell Biology and The Discipline of Medicine, University of Sydney, Sydney, Australia
    • Centenary Institute of Cancer Medicine and Cell Biology, Locked bag No. 6, Newtown, NSW, 2042, Australia
    Search for more papers by this author
    • fax: (61) 2-9565-6101.


  • Potential conflict of interest: Nothing to report.

Abstract

Injury and repair in chronic liver disease involve cell adhesion, migration, apoptosis, proliferation, and a wound healing response. In liver, fibroblast activation protein (FAP) has both collagenase and dipeptidyl peptidase IV (DPIV) activities and is expressed only by activated hepatic stellate cells (HSC) and myofibroblasts, which produce and degrade extracellular matrix (ECM). FAP was colocalized with collagen fibers, fibronectin, and collagen type I in human liver. FAP function was examined in vitro by expressing green fluorescent protein FAP fusion protein in cell lines cultured on collagen-I, fibronectin, and Matrigel. Glutamates at 203 and 204 as well as serine624 of FAP were essential for peptidase activity. Human embryonic kidney 293T cells overexpressing FAP showed reduced adhesion and migration. FAP overexpression in the human HSC line LX-2 caused increased cell adhesion and migration on ECM proteins as well as invasion across transwells in the absence or presence of transforming growth factor beta-1. FAP overexpression enhanced staurosporine streptomyces–stimulated apoptosis in both cell lines. Interestingly, the enzyme activity of FAP was not required for these functions. Overexpressing FAP increased the expression of matrix metalloproteinase-2 and CD44 and reduced integrin-β1 expression in 293T cells, suggesting potential pathways of FAP-mediated impairment of cell adhesion and migration in this epithelial cell line. In conclusion, these findings further support a pro-fibrogenic role for FAP by indicating that, in addition to its enzymatic functions, FAP has important nonenzymatic functions that in chronic liver injury may facilitate tissue remodeling through FAP-mediated enhancement of HSC cell adhesion, migration, and apoptosis. Supplementary material for this article can be found on the HEPATOLOGY website (http://www.interscience.wiley.com/jpages/0270-9139/suppmat/index.html). (HEPATOLOGY 2005;42:935–945.)

Hepatic stellate cells (HSCs) are a key mediator of progressive liver fibrosis.1, 2 In the normal liver, HSCs are quiescent, but after liver injury HSCs become activated and generate fibril-forming matrix that primarily incorporates type I and type III collagens (CN-I, CN-III) and fibronectin (FN).2 These events are crucial in the development of cirrhosis.

Fibroblast activation protein (FAP) is a plasma membrane–bound, constitutively active, metal-independent serine peptidase.3 This enzyme has both dipeptidyl peptidase IV (DPIV) activity and CN-I–specific collagenase activity.4, 5 FAP is expressed by tumor stromal fibroblasts in epithelial cancers and by fibroblasts in healing wounds, but not in normal adult tissue.6, 7 In liver with cirrhosis, FAP is strongly expressed by activated HSC in the portal–parenchymal interface (PPI) and by myofibroblasts in the septum.5, 8 FAP expression intensity correlates with the severity of liver fibrosis.9 These data strongly suggest a profibrotic role for FAP in chronic liver injury.

DPIV is closely related to FAP, but several cell types constitutively express DPIV, including epithelial cells of liver, intestine, endocrine glands and kidney, endothelial cells, tumor stromal fibroblasts, and activated T cells.8, 10, 11 Abherent basolateral DPIV expression occurs on hepatocytes in liver cirrhosis,12 and tumor cells exhibit reduced expression.13 The wide range of DPIV substrates includes the CXCR4 ligand CXCL12 (stromal-derived factor-1α) and another 8 chemokines.11, 14 DPIV interacts with extracellular matrix (ECM) via FN binding15 and interferes with hepatocyte spreading on FN and CN.16

Cell adhesion and migration, proliferation, and apoptosis are central to many pathological processes involving tissue remodeling, including liver fibrosis, inflammation, angiogenesis, cancer growth, and metastasis. This study localized FAP in relation to ECM in human liver with cirrhosis and demonstrated that FAP and DPIV influence apoptosis and proliferation and cell adhesion, in vitro wound healing, and cell migration on ECM substrata. Most notably, FAP overexpression by the LX-2 human HSC cell line increased cell adhesion and migration by those cells on CN and FN.

Abbreviations

HSC, hepatic stellate cell; FN, fibronectin; FAP, fibroblast activation protein; CN-I, collagen type I; PPI, portal-parenchymal interface; DPIV, dipeptidyl peptidase IV; ECM, extracellular matrix; TGF, transforming growth factor; STS, staurosporine streptomyces; DAPI, 4′,6-diamidine-2′-phenylindole dihydrochloride; SHG, second harmonic generation; PCR, polymerase chain reaction; DMEM, Dulbecco's modified Eagle medium; FCS, fetal calf serum; GFP, green fluorescent protein; CFP, cyan fluorescent protein; DDR, discoidin domain receptor; MMP, matrix-degrading metalloproteinase; TIMP, tissue inhibitor of MMPs; PE, phycoerythrin; uPAR, urokinase plasminogen activator receptor.

Materials and Methods

Materials.

Matrigel matrix (BD Biosciences, Bedford, MA), recombinant human transforming growth factor (TGF)-β1 (R&D Systems, Minneapolis, MN), and CN-I, FN, staurosporine streptomyces (STS), and 4′,6-diamidine-2′-phe-nylindole dihydrochloride (DAPI) (Sigma, St. Louis, MO) were purchased.

Antigen and CN Fiber Localization.

Immunofluorescence (antibodies in Table 1) and second harmonic generation (SHG) on cryosections of human liver cirrhosis explants17, 18 and enzyme histochemistry, flow cytometry, and immunocytochemistry19, 20 have been described. Liver explants were obtained with informed consent and Royal Prince Alfred Hospital ethics committee approval. Imaging used a Zeiss (Heidelberg, Germany) Axioskop fluorescence microscope and AxioCam digital camera, or a Leica (Wetzlar, Germany) DMIRBE inverted stand and Leica TCS2MP confocal system. For phalloidin Alexa Fluor 594 staining, cells were incubated overnight on ECM-coated chamber slides (Nunc, Naperville, IL) then formalin fixed and permeabilized and then imaged using a Radiance Plus Confocal (BioRad, Hercules, CA) and LaserSharp 2000 software or the Axioskop.

Table 1. Antibodies
AntibodyIsotypeSupplierCatalogue No.Working Dilution
Primary Antibodies    
 Collagen type IRabbit IgGChemiconAB74510 μg/mL
 FibronectinRabbit IgGSigmaF364810 μg/mL
 Fibroblast Activation ProteinMouse IgG1W. Rettig5ATCC, CRL-27331:5
 Dipeptidyl Peptidase IVMouse IgG1T. Kähne19Clone B10/EF61.7 μg/mL
 Phalloidin-Alexa Flour 594NAMolecular ProbesA123811:40
 Matrix metalloproteinase (MMP)-2Mouse IgG1Santa CruzSC-1359520 μg/mL
 Tissue inhibitor of MMPs (TIMP)-2Rabbit IgGSanta CruzSC-553910 μg/mL
 E-cadherinMouse IgG1Santa Cruz, CASC-2179110 μg/mL
 β-cateninMouse IgG1Transduction LaboratoriesC192200.6 μg/mL
 Integrin β1 (CD29)Mouse IgG2aBD Biosciences5560482.5 μg/mL
 Integrin α3-FITCMouse IgG1SerotecMCA19410 μg/mL
 CD44-Phycoerythrin (PE)Mouse IgG2bBD Biosciences554791:10
 Annexin V-PENot applicableBD Biosciences5564211:50
 CXCR4Rabbit IgGChemiconAB184610 μg/mL
 CXCL12Mouse IgG1R&D SystemsMAB31050 μg/mL
 Discoidin domain receptor (DDR)-1Rabbit IgGSanta CruzSC-5326.7 μg/mL
 MMP9Rabbit IgGSanta CruzSC-793320 μg/mL
 ICAM-2Rabbit IgGSanta CruzSC-1073720 μg/mL
Secondary Antibodies    
 Anti-rabbit Ig-Alexa Fluor 488Goat IgGMolecular ProbesA110341:400
 Anti-mouse Ig-Alexa Fluor 594Goat IgGMolecular ProbesA110321:400
 Anti-mouse Ig-PEGoat IgGMolecular ProbesP8521:400
 Anti-rabbit Ig-PEGoat IgGMolecular ProbesP27711:400

Plasmids, Constructs, and Transfection.

Human FAP (GenBank code U09278) was cloned in-frame downstream of N-terminal green or cyan fluorescent protein in the vectors pEGFP-C2 and pECFP-C1 (BD Biosciences), and human DPIV (M80536) was cloned in upstream of the fluorescent proteins in the vectors pEGFP-N1 and pECFP-N1. Polymerase chain reaction (PCR) of the insert used Platinum Pfx Taq (Invitrogen, Carlsbad, CA) and primers designed with cloning sites SalI and KpnI and stop codon removal with annealing temperature 68°C (Table 2). The two FAP enzyme-negative mutants were engineered using point mutation primers for alanine (GCC) replacement of the catalytic serine (TCC) at position 624 and alanine replacement of the glutamic acids at 203 (GAG to GCG) and 204 (GAA to GCA). Plasmid DNA extraction from Escherichia coli DH5α cells, site-directed mutagenesis, and human embryonic kidney (HEK) 293T cell line (ATCC, CRL-11268) transfection and analysis by enzyme stain and flow cytometry have been described previously.19, 20 293T cells were assayed 40 to 48 hours after transfection.

Table 2. Primers Used in PCR
GenesForwardReverseProduct Size
EGFP/ECFP (N1)-FAP5′-GATTAGGCTGACCATCTGGAAAAATGAAGACTTGGG-3′5′-CCGATAGGTACCAAGTCTGACAAAGAGAAACACTGC-3′2280 bp
EGFP/ECFP (C2)-FAP5′-GATTAGGCTGACCATCTGGAAAAATGAAGACTTGGG-3′5′-GGGGGTACCTTAGTCTGACAAAGAGAAACACTG-3′2280 bp
EGFP/ECFP (N1)-DPIV5′-AATAATGTCGACCGACGATGAAGACACCGTGGAAG-3′5′-TATGATGGTACCAAGGGCCCTCTAGAAGGTAAAGAG-3′2298 bp
FAP Ser624Ala mutant5′-GGGTTTATGAAGCGGCAATGCTTGCTAC-3′5′-GTAGCAAGCATTGCCGCTTCATAAACCC-3′6980 bp
FAP Glu203/204Ala mutant5′-GCCATATGGGGCTGGGCCTATGGAGGATACG-3′5′-CGTATCCTCCATAGGCCCAGCCCCATATGGC-3′6980 bp

The LX-2 cell line was maintained as described.21 For transfection, 2 × 106 cells were suspended in 100 μL Nucleofector solution T (Amaxa Biosystems, Cologne, Germany) then mixed with 2 μg plasmid DNA and electrophoresed in a Nucleofector on program Y-01 (Amaxa Biosystems). Cells were recovered into 0.5 mL pre-warmed RPMI medium then after 5 minutes cultured in 2 mL Dulbecco's modified Eagle medium (DMEM) with 2% fetal calf serum (FCS) in 6-well plates and assayed 24 hours after transfection.

Cell Adhesion Assay.

The cell adhesion assay followed a published method.22 Briefly, 24-well plates were coated with CN-I, FN, or Matrigel at 5 μg/well for 1 hour at room temperature with gentle shaking. Plates were then rinsed 3 times with water, blocked with 1% bovine serum albumin in phosphate-buffered saline for 1 hour at RT, rinsed 5 times with water, then air dried. Forty hours after transfection, cells were washed then resuspended in DMEM at 4 × 105 per well. After incubation for 10 minutes at 37°C, the non-adherent cells were collected gently before collecting adherent cells. Green fluorescent protein (GFP)–expressing cells were enumerated by flow cytometry.

Wound Healing Assay.

The wound healing assay was as described23 with modifications. Six-well plates were coated as above but at 10 μg/well then washed with DMEM. Cells in DMEM/FCS were plated at 2.4 × 105 per well and transfected 24 hours later. At 40 hours after plating, the monolayer was scraped with a small pipette tip to produce “wounds” of approximately 8 × 1 mm, then 1% fresh FCS was added. After overnight incubation, the cultures were imaged by both bright-field and fluorescence stereomicroscopy. KS400 image analysis software version 3.0 (Zeiss) with automatic threshold and lowpass filter was used to measure the total area covered by cells (bright field) and the area covered by fluorescence-positive cells in wound and non-wound portions of each image. The formula used to calculate this estimate of the proportion of cells that were GFP positive was

equation image

Cell Migration Assay.

In the cell migration assay,24 CN-I, FN, or Matrigel was coated onto the lower side of a transwell insert (BD Biosciences) for 1 hour at RT then washed with DMEM. The lower chamber contained 293T or LX-2 cell conditioned medium containing 1% fresh FCS with or without TGF-β1 (2 ng/mL). After overnight serum starvation, cells were placed in the upper chamber then incubated for 2- (LX-2) or 3- (293T) days, until enumerated by flow cytometry. Little 293T cell migration occurred when serum was omitted from the lower chamber.

Apoptosis and Proliferation.

Cells were transferred 30 hours after transfection to 24-well plates at 2 × 105 per well, incubated overnight then incubated with 4 μmol/L STS for 2 to 6 hours. STS is a chemotherapeutic agent that induces cellular apoptosis.25 The cells were then stained with phycoerythrin-conjugated annexin V and propidium iodide (100 ng/mL) to enumerate the apoptotic cyan fluorescent protein (CFP)–positive cells by flow cytometry. Proliferation was assessed by a standard thymidine uptake assay. Twenty-four hours after transfection 3,000 cells per well were incubated with 0.5 μCi tritiated thymidine (Perkin-Elmer Life Sciences, Boston, MA) per well. The ratio of cpm of transfected to untransfected cell populations from 26 replicate wells from 5 experiments was termed the “Mean Proliferation Quotient.”

Statistical Analysis.

Each experiment was repeated 3 to 6 times. Results are expressed as means ± SD. Differences among groups were analyzed using Student's t test, or the non-parametric test Mann-Whitney U for proliferation assays. P < .05 was considered significant.

Results

Co-localization of CN Fibers, CN-I, and FN With FAP in Human Liver With Cirrhosis.

SHG is used to image fibers of CN-I and CN-III in unstained sections.18, 26 SHG combined with confocal imaging of FAP immunostaining showed that fine fibers of CN tend to lie adjacent to FAP-positive HSC and that dense CN fibers surround myofibroblasts in the septum (Fig. 1i). FAP was predominantly expressed in cytoplasm (Fig. 1L). Two-color immunofluorescence staining of CN-I and FAP on human liver with cirrhosis found that both CN-I and FAP had similar staining patterns (Fig. 1ii) in the PPI (Fig. 1D-F) and that FAP and CN-I were expressed by the same cells (Fig. 1G-I). CN-I and FAP co-localized in septal myofibroblasts (Fig. 1K) and co-localized in many HSC of the PPI. Single CN-I and FAP-positive cells were also observed (Fig. 1J). FN and FAP showed the same staining pattern in the PPI region (Fig. 1M-N) and often co-localized (Fig. 1O). However, FAP was predominantly in the cytoplasm (Fig. 1Q), whereas FN was located predominantly on the cell surface (Fig. 1P).

Figure 1.

Localization of CN fibers, CN-I, FN and FAP in human liver with cirrhosis. Cryosections were stained with antibodies to CN-I, FN, DPIV and α-SMA (green) and FAP (red). (i) The portal–parenchymal interface (PPI): CN fibers (green) lay adjacent to FAP-positive cells (A-C). (ii) CN-I colocalized with FAP in the PPI (D-F) in cytoplasm (G-I). (iii) Confocal microscopy showed that CN-I and FAP partially colocalized in HSC (J, arrow) and completely colocalized in myofibroblasts (MS; K, arrow). FAP predominantly localized to cytoplasm (L). (iv) FN colocalized with FAP in the PPI (M-O). FN was predominantly on the cell surface whereas FAP was predominantly cytoplasmic (Q, R). (v) DPIV localized to hepatocytes (S) whereas FAP localized to HSC and myofibroblasts (T). (vi) The SMA+FAP+ cells were predominantly septal myofibroblasts (X). A multi-photon confocal microscope configured for second harmonic generation (SHG) (i), fluorescence microscopy (ii, iv, v, vi), and confocal microscopy (iii) were used. DAPI stained the nuclei blue. CN, collagen; CN-I, collagen type I; FN, fibronectin; FAP, fibroblast activation protein; DPIV, dipeptidyl peptidase IV; α-SMA, alpha-smooth muscle actin; HSC, hepatic stellate cell.

The bile canalicular domain of epithelial cells and many lymphocytes in the lymphoid aggregates of the PPI were DPIV immunopositive. In addition, some septal myofibroblasts in 7 of 13 patients were DPIV and FAP double immunopositive (Fig. 1S-U). Identification of FAP-positive cells as HSC and myofibroblasts5 was confirmed by αSMA double staining (Fig. 1X).

Fluorescent Expression and Enzyme Activity of the Constructs FAP-GFP, DPIV-GFP, GFP-FAP, GFP-FAP Ser624Ala, and GFP-FAP Glu203/204Ala.

Because FAP was found to co-localize with ECM components CN and FN in liver, influences of FAP, compared with DPIV, on cell adhesion and migration were examined using the readily transfected 293T cell line and then the new LX-2 stellate cell line. Transient transfection was used to avoid the behavioral prejudices toward adherence, survival, and proliferation that can be imposed by selecting a stably transfected cell line.27 GFP- or CFP-derived fluorescence was used to analyze the behavior of the peptidase-expressing cells, and all assays enumerated GFP/CFP vector transfected cells as the principal negative control. DPIV trafficking is unaffected by fusion to GFP.28

The fusion protein of GFP onto the C-terminus of DPIV exhibited both green fluorescence and peptidase activity (Fig. 2B,H,N). In contrast, fusions of GFP onto the C-terminus of FAP conferred fluorescence, but not peptidase activity, on transfected cells (Fig. 2C,I,O). However, with GFP at the N-terminus of FAP, both green fluorescence and peptidase activity were seen (Fig. 2D,J,P). GFP was not placed N-terminal to DPIV because DPIV tends to shed its N-terminus.11 Transfection rates were approximately 50% to 70% in 293T cells, and the intensity of expression was comparable to that of HSC (compare Figs. 1 and 2).

Figure 2.

Fluorescence, peptidase activity, and schematics of GFP and fusions of GFP with FAP and DPIV. GFP was cytoplasmic (A). DPIV-GFP (B) and GFP-FAP (C-F) fusion proteins were predominantly on the cell surface and in the Golgi of transfected 293T cells. Only the DPIV-GFP (H) and GFP-FAP (J) constructs showed peptidase activity (the H-Ala-Pro-4Methoxy-βNA product appears brown/black). FAP with GFP at the C-terminus (C,I,O) or with point mutations of the catalytic serine (E,K,Q) or of the glutamates at positions 203 and 204 (F,L,R) lacked peptidase activity. Not to scale. GFP, green fluorescent protein; FAP, fibroblast activation protein; DPIV, dipeptidyl peptidase IV.

Glu205 and Glu206 of DPIV and the homologous residue in DP8, Glu259, are essential for peptidase activity.19, 29 The homologous residues of FAP, Glu203, and Glu204, and the catalytic Ser624, were substituted with Ala and shown to be essential for peptidase activity (Fig. 2E,K,Q; F,L,R). Similar data were obtained in LX-2 cells (not shown). All of the DPIV and FAP constructs were cell surface expressed (Fig. 3).

Figure 3.

Recombinant cell surface expression of FAP and DPIV. More than 60% of cells transfected with GFP-FAP (D-F, M-O) or DPIV-GFP (G-I) expressed detectable recombinant fusion protein on the cell surface (F,H,O). Untransfected 293T cells were FAP and DPIV negative (B-C). Untransfected LX-2 cells expressed some FAP (L). FAP or DPIV transfection did not induce expression of the other enzyme (E,I,N). Cells were stained with monoclonal antibodies to DPIV (B,E,H,K,N) or FAP (C,F,I,L,O) followed by goat anti-mouse PE. FAP, fibroblast activation protein; DPIV, dipeptidyl peptidase IV; GFP, green fluorescent protein; PE, phycoerythrin.

Endogenous Expression of FAP or DPIV in 293T Cells.

To interpret effects of FAP or DPIV overexpression, measuring endogenous expression of FAP and DPIV by 293T cells is important. FAP was not detected on the surface (Fig. 3C). DPIV was not detected on the cell surface (Fig. 3B). In contrast to SKMEL28 melanoma cells, which express both DPIV and FAP after transfection with DPIV,27 in 293T cells neither FAP nor DPIV transfection induced expression of the other enzyme (Fig. 3E,I).

FAP and DPIV in 293T Cell Adhesion.

GFP-FAP, but not DPIV-GFP, expressing cells exhibited less cell adhesion on CN-I or Matrigel-coated plastic compared with cells expressing GFP alone (P < .05) (Fig. 4B). GFP-FAP–expressing cells also exhibited less cell adhesion on FN (P < .05). However, probably because DPIV binds to fibronectin,15 DPIV-GFP expression increased cell adhesion on FN (P < .01, Fig. 4B). The 2 enzyme-inactive GFP-FAP constructs indicated that the enzyme activity of FAP is not necessary for the FAP-associated effects (P < .05, Fig. 4C).

Figure 4.

Cell adhesion of FAP and DPIV overexpressing 293T cells on ECM components. Fluorescent cells adherent and non-adherent to coated plates were enumerated by flow cytometry (A). DPIV overexpressing cells exhibited increased adhesion on FN (B). Cells overexpressing FAP or its enzyme-inactive mutants showed reduced adhesion to CN-I, FN, or Matrigel (C). Data are representative of the 6 experiments that were performed in triplicate. Means and standard errors are plotted. *P < .05 compared with GFP transfection. FAP, fibroblast activation protein; DPIV, dipeptidyl peptidase IV; FN, fibronectin; CN-I, collagen type I; GFP, green fluorescent protein.

FAP and DPIV Reduced Wound Healing in 293T Cell Monolayers.

Effects of FAP and DPIV on cell migration were initially evaluated by using an in vitro monolayer wound healing assay. Fluorescence relative to the cell coverage in the wound and non-wound regions was calculated. Cells expressing either GFP-FAP or DPIV-GFP, compared with GFP, were less prevalent (P < .05) in the wound area on CN-I, FN, or Matrigel (Fig. 5C). This underrepresentation of DPIV+ and FAP+ cells in the healing wound indicates that overexpression of either of these glycoproteins impairs migration on ECM.

Figure 5.

FAP or DPIV overexpression inhibited in vitro wound healing on CN-I, FN, or Matrigel. 293T cell monolayers on CN-I–, FN-, or Matrigel-coated plates transfected with GFP or GFP fusion proteins were imaged 24 hours after wounding. Fluorescence (B) relative to the cell coverage (A) in the wound and non-wound regions was calculated (see Methods). The ratios of relative fluorescence positivity in the wound to the non-wound area from a representative experiment are depicted (C). *P < .05 compared with GFP transfected cells. Six experiments were performed in triplicate. FAP, fibroblast activation protein; DPIV, dipeptidyl peptidase IV; FN, fibronectin; GFP, green fluorescent protein.

FAP and DPIV Reduced 293T Cell Migration.

A cell invasion assay further showed impairment by FAP and DPIV of 293T cell migration on ECM. Cells expressing either GFP-FAP or DPIV-GFP exhibited less migration across the transwell membranes toward CN-I, FN, or Matrigel than did GFP-expressing cells (P < .05, Fig. 6A). Migration toward TGF-β1 was investigated because TGF-β1 can induce resident HSC to migrate into fibrotic liver regions and invade fibrillar matrix,30 and TGF-β can increase cell migration of various cell types including hepatocellular carcinoma cell line SMMC-7721.31 Compared with GFP, cells expressing GFP-FAP or DPIV-GFP exhibited reduced migration toward TGF-β (P < .01, Fig. 6B). Cell migration was reduced to similar extents when both TGF-β and CN-I or Matrigel were offered to the cells.

Figure 6.

FAP or DPIV overexpression reduced cell migration on CN-I, FN, or Matrigel. Transfected 293T cells were placed above CN-I–, FN-, or Matrigel-coated wells (A). Some wells contained TGF-β (B). The percentages of green fluorescence–positive cells in the upper and lower chambers were enumerated by flow cytometry to calculate the ratios of migrated to non-migrated recombinant protein-expressing cells. The figures depict each ratio obtained from GFP-FAP and DPIV-GFP fusion proteins as a proportion of the corresponding ratio obtained from GFP vector transfected cells. These data are representative of 8 experiments. FAP, fibroblast activation protein; DPIV, dipeptidyl peptidase IV; CN-I, collagen type I; FN, fibronectin; TGF-β, transforming growth factor beta; GFP, green fluorescent protein.

FAP, DPIV, and the Cytoskeleton.

Because the cytoskeleton is essential for cell motility in wound healing and tissue remodeling, whether FAP or DPIV are associated with the actin cytoskeleton or influence its morphology was investigated. Alexa Fluor 594-conjugated phalloidin, which binds to F-actin filaments, was used to visualize the actin cytoskeleton. Phalloidin exhibited very little co-localization with FAP in the cytoplasm (Fig. 7A). FAP was located in Golgi, endoplasmic reticulum, some small cytoplasmic vesicles, and on the cell surface (Fig. 7B). Similar results were obtained from cells on FN (Fig. 7), CN-I, or Matrigel-coated slides and from GFP and DPIV transfected and untransfected cells (data not shown). These data indicate that FAP and DPIV are exported to the cell surface via the trans-Golgi in the usual manner without associating with the cytoskeleton.

Figure 7.

FAP and the cytoskeleton. Cells transfected with GFP-FAP were plated on FN-coated slides, fixed then permeabilized and stained with phalloidin Alexa Fluor 594. FAP (green) and the actin cytoskeleton (red) were not co-localized in cytoplasm (Overlay; A). Confocal imaging of 293T cells (A-C). DAPI stained LX-2 cell nuclei blue (D). FAP, fibroblast activation protein; GFP, green fluorescent protein; FAP, fibroblast activation protein; FN, fibronectin; DAPI, 4′6-diamidine-2′-phenylindole dihydrochloride.

Cellular Proliferation and Apoptosis.

DPIV expression can influence cell proliferation and apoptosis32–35 but no such data on FAP are available. The thymidine uptake assay detected robust 293T cell proliferation, in the range 5,000 to 15,000 cpm in 5 experiments. Cells transfected with DPIV or FAP exhibited greater thymidine uptake than GFP alone (Mean Proliferation Quotients, 0.62 ± 0.11, 0.55 ± 0.07 and 0.46 ± 0.09, respectively; P < .5 × 10−4). GFP expression was an important control because it caused decreased proliferation compared with untransfected cells. Nonapoptotic CFP positive 293T cells were enumerated by flow cytometry (Fig. 8A, B). In the absence of STS, cells overexpressing FAP or DPIV exhibited a small increase in apoptosis. However, in the presence of STS, cell viability significantly decreased within 2 hours. The FAP enzyme–negative mutants were as effective as wild-type FAP at increasing STS-induced apoptosis (Fig. 8D).

Figure 8.

Overexpression of FAP or DPIV increased STS-induced cell apoptosis. Transfected 293T (A-D) or LX-2 (E) cells were incubated with STS. After flow cytometry, the cyan fluorescence-positive cell population (A) was selected for analysis. The annexin V-PE and propidium iodide double-negative cell population was enumerated (B) as nonapoptotic. Both FAP and DPIV enhanced STS induced apoptosis. FAP enzyme–negative mutants (Ser624Ala, Glu203/204Ala) both increased apoptosis as much as wild-type FAP (D-E). These experiments were performed 7 times (A-C) or 3 times (D-E). FAP, fibroblast activation protein; DPIV, dipeptidyl peptidase IV; STS, staurosporine streptomyces.

The Molecular Phenotype of FAP and DPIV Overexpressing 293T Cells.

Expression levels of some proteins that have roles in cell adhesion may provide insights into the molecular mechanisms of the phenomena reported. In recent studies of carcinoma cell lines, DPIV overexpression was associated with increased matrix metalloproteinase (MMP)-2 and CD44.35, 36 Similarly, we observed that overexpression of either FAP or DPIV was associated with increased expression of MMP2 and CD44. DDR1 is a non-integrin CN receptor that stimulates adhesion and migration37 and is upregulated in liver with cirrhosis.38 Decreased expression of discoidin domain receptor (DDR)-1 was associated with DPIV overexpression (Table 3). However, the flow cytometry profiles of the most prominent differences indicate that these changes were small (supplemental Fig. 1), which suggests that the mechanism is multigenic.

Table 3. Molecular Phenotype of FAP- and DPIV-Overexpressing 293T Cells
 E-Cadherinβ-CateninMMP2TIMP2CD44Integrin β1CXCR4CXCL12DDR1
  • *

    Median fluorescence intensity. Values from each corresponding negative antibody control were subtracted. All data were derived only from cells in which expression of CFP was detected. ND = not done; FP, cyan fluorescent protein.

Cell surface         
 CFP3.36*0.241.960.096.984.186.614.130.52
 CFP-FAP2.86−0.281.620.798.453.126.373.162.17
 DPIV-CFP4.450.092.020.4213.25.17.092.980.29
Permeabilized         
 CFP7.0454.6−0.5927.27.04ND39.914.4193
 CFP-FAP6.8446.80.3632.77.99ND29.812.4149
 DPIV-CFP6.5752.91.3534.910.0ND34.711.6184

FAP Overexpression in LX-2 Cells.

The recent availability of a transfectable human HSC line permitted examination of effects of FAP overexpression in the liver cell type in which FAP is highly expressed in vivo. LX-2 cells express α-smooth muscle actin, vimentin, glial fibrillary acidic protein, platelet-derived growth factor receptor β, DDR2, MMP-2, TIMP-2, and MT-MMP and make CN-I in response to TGFβ1.21 We detected DDR1, integrin β1, integrin α3, β-catenin, CXCR4, CXCL12 and low levels of FAP but no DPIV on the LX-2 cell line (Figs. 3 and 9). Transfection produced similar levels of GFP-FAP expression, and no DPIV expression, on LX-2 as 293T cells (Fig. 3). Transfected LX-2 cell cultures were 70% to 90% GFP positive.

Figure 9.

Increased cell adhesion (A), in vitro monolayer wound healing (B), and migration across transwells (C) by FAP overexpressing LX-2 cells. Fluorescent cells were enumerated by flow cytometry (A, C) or quantitative image analysis (B). LX-2 migration into monolayer wounds was more rapid on CN than FN (D). Data are representative of 3 experiments performed in triplicate. Means and standard errors. *P < .05 compared with GFP transfection. FAP, fibroblast activation protein; CN, collagen; FN, fibronectin.

Interestingly, FAP overexpression in LX-2 cells was associated with increased cell adhesion (Fig. 9A) and cell migration (Fig. 9B-C) on CN and FN. FAP overexpression also increased STS-induced apoptosis (Fig. 8E). All detectable levels of GFP-FAP expression were associated with these functional effects (data not shown). These FAP activities were also evident after ablating FAP enzyme activity by point mutation. FAP did not co-localize with the actin cytoskeleton in LX-2 cells (Fig. 7D).

LX-2 cells migrated into monolayer wounds faster on CN than FN (Fig. 9D) and faster than 293T cells. Concordantly, LX-2 cells crossed transwell membranes faster than 293T cells (data not shown). More integrin β1 was expressed on migrated LX-2 cells than on non-migrated cells, and the migrated LX-2 population was enriched in β-catenin+ cells (Supplemental Fig. 2B).

Discussion

FAP is a collagenase expressed by activated HSC and myofibroblasts, and its expression level correlates with fibrosis severity.3, 5, 9 In the current study, FAP was colocalized with CN-I and FN in the tissue remodeling region, the PPI, of human liver with cirrhosis, further implicating FAP in HSC and myofibroblast interactions with the ECM. Moreover, CN fibers were adjacent to FAP-positive HSC in the PPI. Therefore, the role of FAP in cell–ECM interaction was examined in elegant in vitro models. FAP overexpression by the LX-2 HSC cell line increased cell adhesion and migration on ECM proteins. In contrast, in an epithelial cell line, FAP overexpression reduced cell adhesion and migration on ECM proteins. FAP overexpression enhanced STS-stimulated apoptosis and was not associated with the actin cytoskeleton in either cell line. We showed that the glutamates at positions 203 and 204 in FAP are essential for peptidase activity, but not for FAP-mediated effects on cell–ECM interactions or apoptosis. These data suggest an extra-enzymatic, pro-fibrotic role for FAP in pathogenesis by enhancing important stellate cell functions.

Cell adhesion and migration are important in many pathological processes involving tissue remodeling, including liver fibrosis, inflammation, angiogenesis, cancer growth, and metastasis. In liver fibrosis, activated HSC produce ECM that provides cells with positional information and a mechanical scaffold for adhesion and migration, partly by binding to certain growth factors/cytokines, MMPs, and processing enzymes.2 CN-I and FN can be chemoattractant stimuli for HSC migration.30 The increased adhesion and migration by FAP overexpressing LX-2 cells toward CN-I and FN in transwells suggests that FAP expression by activated HSC in the PPI may assist HSC migration in injured liver. In addition, the pro-proliferative and pro-apoptotic activities of FAP shown here are consistent with the needs of cells in a tissue remodeling microenvironment such as the regenerating liver. Similarly, FAP expression by stromal fibroblasts in solid epithelial tumours6–8 perhaps functionally assists these cells as the tumour mass increases.

The collagenase and peptidase activities of FAP both depend on the catalytic serine (Ser624).4 Glu205 and Glu206 are required for DPIV peptidase activity,29 probably by positioning substrates in the catalytic pocket.3 We show here that mutating the equivalent glutamates in FAP, Glu203, and Glu204 ablated peptidase activity but did not alter FAP-dependent effects on adhesion, migration, or apoptosis, indicating that these effects are mediated by protein–protein binding rather than binding or hydrolysis of a substrate.

β-Catenin has several functions.39 The increased proportion of β-catenin+ cells in the population of migrated LX-2 cells suggests that LX-2 migration involves β-catenin–mediated adhesion. The increased expression of integrin β1 on migrated LX-2 cells suggests that LX-2 migration on CN was integrin rather than DDR1 dependent. Because FAP is known to associate with integrin α3β1 on CN-stimulated cells40 and the LX-2 cells expressed integrin α3β1, perhaps FAP acted on the integrin pathway to increase LX-2 cell migration. However, expression of neither β-catenin nor integrin β1 was altered in FAP+ LX-2 cells (Supplemental Fig. 3).

Although FAP is a cell surface molecule, immunostaining indicated that most FAP molecules lie in cytoplasm, where they might interact with cytoskeleton-associated proteins, such as vinculin, actinin, talin, and paxillin. Indeed, nischarin inhibits cell motility by binding to an integrin cytoplasmic domain.41 However, FAP did not co-localize with the actin cytoskeleton or alter cytoskeletal morphology. Thus, it appears unlikely that FAP influences cell adhesion or migration by interacting with the cytoskeleton.

No ligand of FAP has been identified despite the close structural similarity of FAP with DPIV.42 FAP is an integral membrane glycoprotein that co-localizes with integrin α3β1 and the urokinase plasminogen activator receptor (uPAR) on the LOX melanoma cell line.40, 43 The formation of the FAP–uPAR membrane complex is dependent on the cytoskeleton and integrin β1.43 The uPAR ligand uPA converts plasminogen to plasmin, which degrades fibrin and certain ECM proteins. Therefore, interactions between the extracellular domain of FAP and extracellular molecules potentially have a role in adhesion and migration on ECM.3 Cell adhesion and migration are associated with up-regulation of MMPs that proteolytically degrade ECM.44 CD44 is a hyaluronic acid receptor and thus has roles in cell migration and adhesion. Integrins are the major class of cell surface molecules linking the cytoskeleton to the ECM. We found that overexpressing FAP up-regulated MMP2 and CD44 and down-regulated integrin β1 on 293T cells. These data indicate possible pathways for impairing adhesion and migration by FAP in 293T cells.

FAP significantly enhanced STS-stimulated apoptosis in both 293T and LX-2 cells, using an enzyme activity independent process. A very recent report indicates that FAP also increases apoptosis in the mouse B16 melanoma cell line.45 HSC apoptosis contributes to liver healing,46, 47 so FAP could have a role in HSC apoptosis in liver.

The DPIV data concords with prior in vitro studies showing FN–DPIV binding15 and DPIV-mediated inhibition of invasion by melanoma cells48 and enhanced apoptosis.32, 35 Neither the protease activity nor the cytoplasmic domain of DPIV are required for these DPIV functions. Translating these data into an understanding of the roles of DPIV in liver is confounded by the several cell types, including hepatocytes, endothelial cells, and activated lymphocytes, that express DPIV.

Previous studies of DPIV overexpression had interpretive difficulties because of the expression of FAP by DPIV-transfected cells.27, 35 However, in the 293T and LX-2 cell lines, neither FAP nor DPIV induced the expression of the other enzyme, so these cell lines provided in vitro models in which to study the individual enzymes. Moreover, transfected cells were unselected and uncloned, so that effects of FAP or DPIV on cell adhesion, proliferation, or apoptosis did not preselect the cell population studied. Instead, measurements of cell behaviors were restricted to peptidase-expressing cells by fluorescent tagging. These assay design features maximized the specificity of our observations.

In conclusion, FAP overexpression in a human stellate cell line increased cell adhesion, migration, and invasion on ECM components and increased induced apoptosis independently of its enzyme activity. These data have implications for HSC and myofibroblasts in liver and are consistent with a pro-fibrotic role for FAP in chronic liver injury.

Acknowledgements

The authors thank Scott Friedman for LX-2 cells and Ellie Kable, Guy Cox, and Dennis Dwarte for advice on confocal imaging and image analysis.

Ancillary