Matrilysin Cleavage of Corneal Collagen Type XVIII NC1Domain and Generation of a 28-kDa Fragment

Hsin-Chiung Lin,1 Jin-Hong Chang,1 Sandeep Jain,1 Eric E. Gabison,1 Tomoko Kure,1

Takuji Kato,1 Naomi Fukai,2 and Dimitri T. Azar1

PURPOSE. To localize endostatin and collagen type XVIII inhuman corneas and to characterize the enzymatic action ofmatrix metalloproteinases (MMPs) in the cleavage of collagentype XVIII and generation of endostatin in the cornea.

METHODS. Anti-endostatin and anti-hinge antibodies were gen-erated using peptide fragments corresponding to the endo-statin region and the adjacent nonendostatin hinge region ofcollagen XVIII noncollagenous (NC)1 domain, respectively.Confocal immunostaining was performed to localize collagenXVIII in human corneas. SV40-immortalized corneal epithelialcells were immunoprecipitated and incubated with activeMMP-1, -2, -3, -7, or -9, and Western blot analysis was per-formed to study collagen XVIII cleavage. Incubation withMMP-7 was performed at various concentrations (0, 2, 4, and 6mg/ml) and time intervals (0, 1, 5, and 12 hours). Purifiedrecombinant NC1 fragment of collagen XVIII was also digestedwith MMP-7, and the cleavage product was sequenced.

RESULTS. Collagen XVIII was immunolocalized to the humancorneal epithelium, epithelial basement membrane, and Des-cemet membrane. Western blot analysis demonstrated a 180- to200-kDa band corresponding to collagen XVIII. MMP-7 (but notMMP-1, -2, -3, and -9) cleaved corneal epithelium–derived col-lagen XVIII to generate a 28-kDa endostatin-spanning fragmentin a time- and concentration-dependent fashion. MMP-7cleaved purified recombinant 34-kDa NC1 fragment of collagenXVIII in the hinge region to generate a 28-kDa fragment.

CONCLUSIONS. Collagen XVIII is present in human cornea.MMP-7 cleaves the collagen XVIII NC1 domain to generate a28-kDa fragment in the cornea. (Invest Ophthalmol Vis Sci.2001;42:2517–2524)

Collagen type XVIII is a nonfibrillar collagen belonging tothe heparan sulfate proteoglycan family.1,2 It is composed

of 10 collagen domains alternating with 11 noncollagenous(NC) domains. The N terminus (NC11) and C terminus (NC1)noncollagenous domains have a non–triple-helical structure.3

The NC1 region contains three functionally distinct regions: an

association domain (necessary for collagen XVIII oligomeriza-tion), a hinge domain (sensitive to proteases cleavage), and theendostatin domain (a 20-kDa fragment with potent anti-angio-genic properties).4

Collagen XVIII is mainly localized to the vascular and epi-thelial basement membrane.5,6 Several isoforms of collagenXVIII cDNAs have been isolated. In the mouse, three isoforms(two in humans) have been shown to be differentially ex-pressed among tissues.7 In humans, the shorter isoform isexpressed in the heart, kidney, placenta, ovaries, skeletal mus-cle, and small intestine, whereas the longer variant is liverspecific. In the eye, collagen XVIII localization has been re-ported in the retina (inner limiting membrane and pigmentepithelium) and the lens capsule but has not been reported inthe cornea.8

Cleavage of collagen types IV, XV, and XVIII by proteasesgenerates fragments possessing anti-angiogenic properties (ar-resten, restin, and endostatin, respectively).9–11 Several pro-teases, including cathepsin L12 and elastase,13 have been dem-onstrated to be involved in the cleavage of the C-terminalnoncollagenous domain (NC1) of collagen type XVIII to pro-duce endostatin.

Matrix metalloproteinases (MMPs) and other proteases mayalso be involved in the cleavage of collagen XVIII.12 The MMPfamily encompasses 24 members that share a calcium and zincdependency and a high rate of homology in the catalyticdomain.14–16 MMPs are involved in extracellular matrix (ECM)remodeling, wound healing, development, cancer invasion,and angiogenesis.17 Several MMPs (types 1, 2, 7, 8, 9, and 13)are known to degrade native collagens (types I, II, III, IV, andVI) and gelatins.18 However, cleavage of collagen XVIII byMMPs and the proteolytic release of endostatin have not beenfully characterized.

We previously reported the expression of MMPs in thecornea under normal (MMP-2 and -7) and wound-healing con-ditions (MMP-9, -13, and -14).19–22 The purpose of this study isto investigate whether collagen XVIII is present in the corneaand to identify the MMPs that are capable of degrading cornealcollagen XVIII to generate endostatin-like fragments.

MATERIALS AND METHODS

Cell Culture and Tissue

The SV40-immortalized human corneal epithelial cell line was a giftfrom Hiroshi Handa, Kyoto Institute of Technology, Kyoto, Japan.23

Corneal epithelial cells were grown to confluence, scraped, and lysedwith extraction buffer containing 8 M urea, 50 mM Tris-HCl (pH 8.5),150 mM NaCl, 1% NP-40 and protease inhibitor combination (SigmaChemical Co., St. Louis, MO).

Human corneal specimens were obtained from donor corneal tis-sue, embedded in optimal cutting temperature (OCT) compound(Miles, Elkhart, IN), and flash frozen. Eight-micrometer sections wereprocessed for immunohistochemistry.

From the 1Department of Ophthalmology, Massachusetts Eye andEar Infirmary and the Schepens Eye Research Institute, and the 2De-partment of Cell Biology, Harvard Medical School, Boston, Massachu-setts.

Supported by National Eye Institute Grant EY10101, the NewEngland Corneal Transplantation Research Fund, and the Massachu-setts Lion Eye Research Fund. DTA is the recipient of a Lew R.Wasserman Merit Award from Research to Prevent Blindness.

Submitted for publication March 14, 2001; revised May 30, 2001;accepted June 19, 2001.

Commercial relationships policy: N.The publication costs of this article were defrayed in part by page

charge payment. This article must therefore be marked “advertise-ment” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Corresponding author: Dimitri T. Azar, Corneal, External Diseaseand Refractive Surgery Services, Massachusetts Eye and Ear Infirmary,243 Charles Street, Boston, MA 02114. dazar@meei.harvard.edu

Investigative Ophthalmology & Visual Science, October 2001, Vol. 42, No. 11Copyright © Association for Research in Vision and Ophthalmology 2517

Generation and Characterization of Anti-CollagenXVIII Antibodies

The study was conducted in accordance with the ARVO Statement forthe Use of Animals in Ophthalmic and Vision Research. Female NewZealand white rabbits were immunized using synthetic peptides con-taining residues of the murine 38-kDa NC1 domain (Research GeneticsInc., Huntsville, AL). The first peptide, DDILANPPRLPDRQPYPGVPHH,contained 22 residues in the hinge domain adjacent to the endostatindomain. The second peptide, RRADRGSVPIVNLKDEVLSPSWD, con-tained 23 residues in the N-terminal of endostatin domain of NC1 (Fig.1). Rabbit anti-mouse polyclonal antibodies to the hinge and endostatindomains were generated. They were affinity purified and used forimmunohistochemical and Western blotting experiments.

Collagen XVIII immunocomplexes were prepared by incubatingmouse liver lysate with anti-collagen antibodies (anti-endostatin oranti-hinge), and Western blot analysis was performed.

Confocal Immunohistochemistry

Cryosections (8 mm) of human corneas were mounted on glass slides(Superfrosted/Plus; Fisher Scientific, Pittsburgh, PA) and kept at roomtemperature for 30 minutes. A blocking solution (1% BSA in PBS) wasapplied for 30 minutes at room temperature, followed by incubationwith primary antibodies, either anti-hinge, anti-endostatin or type IVcollagen (Southern Biotechnology Associates, Inc., Birmingham, AL) of

1: 100 dilution for 60 minutes. Samples were then washed with PBS,and the secondary antibody, either rhodamine-conjugated donkey anti-goat of 1:400 dilution or FITC-conjugated affinity purified donkeyanti-rabbit IgG of 1:100 dilution (Jackson ImmunoResearch Laborato-ries, West Grove, PA), was applied for 30 minutes. After the PBS wash,the specimens were mounted with anti-fading medium (Vectashield;Vector Laboratories, Burlingame, CA). The sections were examinedwith a confocal microscope (TCS4D; Leica, Heidelberg, Germany).Negative control samples (without primary antibody) were similarlyprocessed using the same procedure.

Western Blot Analysis

Immunoprecipitate of collagen XVIII (using anti-hinge antibody) wererun on 10% to 20% pre-cast SDS polyacrylamide gels (Novex, SanDiego, CA), then electrotransferred onto membranes (Immobilon P;Millipore, Bedford, MA). Membranes were blotted with anti-endostatinantibodies and reacted with horseradish peroxidase (HRP)–conjugateddonkey anti-rabbit IgG secondary antibodies (New England Nuclear,Boston, MA). After a washing with TBST for 15 minutes three times,immunoblots were developed with enhanced chemiluminescence(ECL) reagent (Western blot detection reagents; NEN-Life Science Prod-ucts, Boston, MA). To test the reproducibility of our data, we used twoadditional anti-collagen XVIII antibodies (gift from Bjorn Olsen, Har-vard Medical School, Boston, MA): anti-NC1 antibody (against the

FIGURE 2. Western blot analysis andimmunoprecipitation to characterizeanti-endostatin and anti-hinge anti-bodies under reducing (A, C) andnonreducing (B) conditions. Cornealepithelial whole-cell lysate was usedin experiments (A) and (B). Liver ly-sate was used in experiment (C). Im-munoprecipitation under reducingconditions (lanes 2, 4, 7, and 8)showed a characteristic IgG heavy-chain band centered at the 50-kDaposition, whereas immunoprecipita-tion under nonreducing conditions(lane 6) showed the IgG band cen-tered at the 150-kDa position. (A)Western blot analysis and immuno-precipitation with anti-endostatin an-tibody (lanes 1, 2) and anti-hingeantibody (lanes 3, 4) showed immu-noreactivity at the 180- to 200-kDaposition consistent with collagenXVIII’s expected size. (B) Under non-reducing conditions, the IgG bandobserved after immunoprecipitation(lane 6) masked the collagen XVIIIband (180 kDa; lane 5). These conditions minimize interference of lower molecular weight nonspecific staining (as shown by the 25- to 50-kDaregion, clear of nonspecific bands). (C) Liver lysates were immunoprecipitated with anti-endostatin (blot with anti-endostatin; lane 7) or anti-hinge(blot with anti-hinge; lane 8) antibodies. The immunocomplexes were immunoblotted with either anti-endostatin (lane 7) or anti-hinge (lane 8)antibodies. Arrows: Characteristic 180- to 240-kDa bands of collagen XVIII immunostaining.7

FIGURE 1. Schematic diagram of mouse and human collagen XVIII. Two anti-mouse collagen XVIII(anti-hinge and anti-endostatin) antibodies were generated according to the peptide residues in the hinge(residues 1305-1326) and endostatin (residues 1405-1427) domains, respectively. The murine and humanpeptides show highly conserved sequences.

2518 Lin et al. IOVS, October 2001, Vol. 42, No. 11

C-terminal end of collagen XVIII which includes the endostatin do-main); and anti-NC11 (against the N-terminal end of collagen XVIII) forimmunoprecipitation or immunoblotting, respectively.

Western blots were performed under nonreducing or reducing(achieved with b-mercaptoethanol) conditions. Because the collagenXVIII band is masked by the IgG complex under nonreducing condi-

FIGURE 3. Immunolocalization ofcollagen XVIII and collagen IV inhuman cornea using confocal mi-croscopy. For double-staining ex-periments (C, F), sections were im-munostained with rabbit anti-hingeantibodies (A, D) and with goat anti-type IV collagen antibodies (B, E).Rhodamine-labeled secondary anti-rabbit antibodies and FITC-labeledsecondary anti-goat antibodies wereused. Superimposition of imagesshows colocalization of collagenXVIII and collagen IV in the epithelialbasement membrane (C) and thestromal side of the Descemet mem-brane (F). (G) Indirect immunofluo-rescence staining with anti-endo-statin antibody shows similarepithelial localization as in (A). (H)Negative control. Bar, 20 mm.

FIGURE 4. Western blot analysis (performed under nonreducing conditions) showing MMP-inducedcleavage of collagen XVIII. Immunoprecipitation of cultured human corneal epithelial cell lysate wasperformed using anti-hinge antibodies and immunoblot analysis was performed with anti-endostatinantibody. (A) Immunoprecipitates were incubated without MMPs (lane 1) and with MMPs-7, -2, and -9(lanes 2, 3, and 4, respectively). MMP-7 incubation (2 mg/ml) for 1 hour showed evidence of cleavage ofcollagen XVIII to generate a 28-kDa fragment (lane 2). Incubation with MMP-2 and -9 did not generate a28-kDa fragment (lanes 3, 4). (B) Immunoprecipitates incubated with active MMP-1, -2, -3, and -9 (with andwithout an additional step of APMA activation) revealed no additional change of collagen XVIII with theadditional step of APMA activation.

IOVS, October 2001, Vol. 42, No. 11 Matrilysin and Endostatin in the Cornea 2519

tions and the endostatin band hidden by IgG light chain under reduc-ing conditions, the two methods were performed depending on theband of interest.

Cleavage of Collagen XVIII by MMPs

Collagen XVIII immunocomplexes were prepared by incubating celllysates with anti-hinge antibodies. The immunocomplexes were thenwashed twice with washing buffer and once with phosphate-bufferedsaline and then incubated with active MMP-1, -2, -3, -7, or -9 enzyme insubstrate buffer (50 mM Tris-HCl [pH 7.4], 150 mM NaCl, 50 mMZnSO4) for 1 hour at 37°C. MMP-1 and -3 enzymes were obtained fromSigma Chemical Co. and MMP-2, -7, and -9 from Calbiochem (SanDiego, CA). Aminophenylmercuric acetate (APMA; 1 mM at 37°C. for 2hours; Sigma) was used to activate inactive enzymes. Combinations ofMMPs (MMP-1 and -7, MMP-2 and -7, MMP-3 and -7, MMP-9 and -7,MMP-1 and -3, or MMP-3 and -9) were also used. Various MMP-7concentrations (0, 2, 4, and 6 mg/ml) and incubation times (0, 1, 5, and12 hours) were used. The reaction was stopped by adding 23 SDS gelloading buffer and boiling for 2 minutes, and Western blot analysis wasperformed as described earlier.

RESULTS

Western Blot Analysis with Anti-endostatin andAnti-hinge Antibodies

Western blot analysis of corneal epithelial cells (with andwithout immunoprecipitation with anti-endostatin and anti-hinge antibodies) under reducing conditions revealed a 180- to200-kDa band corresponding to collagen XVIII (Fig. 2A). Im-munoprecipitation showed an IgG band at the 50-kDa position

under reducing conditions (Fig. 2A, lanes 2 and 4). Undernonreducing conditions (Fig. 2B, lane 6) the IgG band wascentered at the 150-kDa position, which masked the collagenXVIII immunoreactive band. To confirm the specificity of ourantibodies (anti-hinge and anti-endostatin antibodies), liver ly-sates were used for immunoprecipitation and further analyzedby Western blot analysis. As shown in Figure 2C under reduc-ing conditions, characteristic high-molecular-weight bandswere seen at the 180- to 240-kDa position. (These multiplereactive bands of collagen XVIII may be due to alternativesplicing and/or heparan sulfate modification in the liver.)

Immunohistochemistry of Collagen XVIII inHuman Cornea

Confocal immunostaining with anti-hinge and anti-endostatinantibodies revealed collagen XVIII immunolocalization to thecorneal epithelium and epithelial basement membrane (Fig.3A) and to the Descemet membrane (Fig. 3D). Epithelial stain-ing was uniform in all layers. No differences between collagenXVIII immunolocalization were observed between anti-hinge(Fig. 3A) and anti-endostatin (Fig. 3G) antibody staining. Colo-calization of collagen XVIII and type IV collagen was observedin the epithelial basement membrane (Fig. 3C) and the stromalside of the Descemet membrane (Fig. 3F). No staining was seenin the negative control specimens (Fig. 3H).

Collagen XVIII Cleavage by MMPs

For collagen XVIII cleavage experiments, nonreducing condi-tions were used. To characterize the potential role of matrixmetalloproteinases in the generation of endostatin, collagen

FIGURE 5. Comparison of MMP-7–treated immunoprecipitates of cultured human epithelial cell lysateswith lysate-free samples using Western blot analysis (A) and silver staining (B) (performed under nonre-ducing conditions). Lane 1: no lysate or immunoprecipitation (lysis buffer and protein A incubated for 1hour at 37°C); lane 2: addition of MMP-7 active enzyme to sample in lane 1; lane 3: no lysate and additionof anti-hinge antibody incubated for 1 hour; lane 4: MMP-7 incubated with anti-hinge antibody; lane 5:epithelial cell lysate immunoprecipitated with anti-hinge antibody; lane 6: incubation with MMP-7showing a 28-kDa breakdown product that stains with anti-endostatin antibody (Arrow; A). Silver staining(B) was performed on the identical samples as indicated in (A). Lanes 2, 4, and 6: addition of MMP-7 activeenzyme showed a 19-kDa band that did not stain with anti-endostatin antibody. (A; lane 6, arrow) Thepresence of the 28-kDa band suggests that this fragment is generated by MMP-7–mediated cleavage ofcollagen XVIII.

2520 Lin et al. IOVS, October 2001, Vol. 42, No. 11

XVIII was immunoprecipitated from cultured human corneal ep-ithelial cells and incubated with active MMP-2, -7, and -9 for 1hour. Western blot analysis was then performed. As shown inFigure 4A, collagen XVIII can be cleaved by active MMP-7 togenerate a 28-kDa endostatin-spanning fragment. Similar experi-ments were performed using MMP-1 and -3. Figure 4B shows that1 hour incubation of immunoprecipitated collagen XVIII withactive MMP-1, -2, -3, and -9, further activated using APMA, showedthat the greatest cleavage of collagen XVIII occurred with MMP-7.

To eliminate the possibility that the 28-kDa matrilysin-derivedendostatin-like fragment came from either immunoprecipitatedantibodies, active MMP-7, or protein A, Western blot analysis wasperformed with and without lysate (Fig. 5A). A 28-kDa fragmentwas generated only in the presence of lysate, anti-hinge antibody,and active MMP-7 (Figs. 5A; lane 6), suggesting that the origin ofthis band was collagen XVIII. This 28-kDa band was not seen onsilver staining under similar conditions (Fig. 5B), which may implya low amount of generated fragment.

FIGURE 6. Immunoprecipitation experimentswith human corneal epithelial cell lysate, usinganti-NC11 antibodies. Western blot analysiswith anti-NC1 antibodies was performed undernonreducing conditions to determine the cleav-age of collagen XVIII immunoprecipitated byMMPs. (A) MMP-7 (lane 5) degraded collagenXVIII to generate a 28-kDa fragment. MMP-1, -2,-3, and -9 (lanes 2–4, 6) did not cleave collagenXVIII. (B, C) With increasing time of incubation(B) and with increasing MMP-7 concentrations(C), the 28-kDa band showed correspondingincreases in density. (D) Western blot analysisshowing cleavage of collagen XVIII by combi-nation of MMPs. MMP-1, -2, -3, and-9 did notprovide an apparent synergistic effect withMMP-7.

IOVS, October 2001, Vol. 42, No. 11 Matrilysin and Endostatin in the Cornea 2521

To confirm the cleavage of collagen XVIII by MMPs, twoadditional anti-collagen XVIII antibodies (anti-NC1 and anti-NC11) were used (Fig. 6). Incubation with active MMP-7generated a similar 28-kDa band, but not with MMP-1, -2, -3,or -9 (Fig. 6A). MMP-7 generated this 28-kDa fragment in atime- and concentration-dependent fashion (Figs. 6B, 6C).To address whether a combination of MMPs is required forcollagen XVIII processing, a mixture of MMPs was used toassay the processing of immunoprecipitated collagen XVIII(Fig. 6D). We found that the addition of other MMPs (types1, 2, 3, and 9) to MMP-7 did not alter the cleavage pattern ofcollagen XVIII. In addition, combinations of MMP-1 and -3and MMP-3 and -9 did not generate additional fragments ofcollagen XVIII (Fig. 6D).

Cleavage and Sequencing of RecombinantCollagen XVIII NC1 Domain with MMP-7

As shown in Figure 7A, a 28-kDa endostatin-spanning moleculewas produced by the cleavage of a 34-kDa NC1 fragment byactive MMP-7 enzyme. The 28-kDa fragment was subjected toN-terminal protein sequencing and was shown to have anN-terminal amino acid sequence of LxDSNPYPRR which arelocated in the hinge region of NC1 domain (Fig. 7B). Thus thecleavage site of MMP-7 differs from that of MMP-9, elastase, andcathepsin L (Fig. 7B).

DISCUSSION

In this study, collagen XVIII was localized in the corneal epi-thelium, epithelial basement membrane, and Descemet mem-brane. We believe that this is the first published report dem-onstrating the presence of collagen XVIII in the human cornea.In addition, we demonstrated matrilysin-induced proteolysis ofcollagen XVIII to release a 28-kDa fragment and characterizedthe cleavage site in the hinge region of NC1 domain.

Ingber and Folkman postulated that collagen metabolismmay play a role in regulating angiogenesis.24 This is consistentwith studies showing the anti-angiogenic activity of noncollag-enous domain proteolytic fragments of collagen types XVIII(endostatin) and XV (restin) and type IV collagen chains a1(arrestin), a2 (canstatin), and a3 (tumstatin).9–11,25–27 Amongproteases, the matrix metalloproteinases are of special interestin their ability to cleave ECM molecules and cell surface pro-teins.28 MMPs regulate the assembly of the ECM and the releaseof bioactive fragments and growth factors during growth, mor-phogenesis, tissue repair, and pathologic processes. These en-zymes are distinguished from other classes of proteinases bytheir dependence on metal ions and a neutral pH to be active.There is evidence that MMPs generate angiostatin from plas-minogen.29–35 We examined the cleavage susceptibility of im-munoprecipitated collagen XVIII to five different MMPs:MMP-1 (tissue collagenase), MMP-2 (gelatinase A), MMP-3(stromelysin 1), MMP-7 (matrilysin), and MMP-9 (gelatinase B)and showed that active MMP-7 enzyme cleaves collagen XVIII

FIGURE 7. (A) MMP-7 cleavage of purified recombinant NC1. Lane 1: Western blot analysis of purified NC1 fragment; lane 2: the effect of MMP-7incubation showing a 28-kDa molecule. The molecule generated by MMP-7 activity was sequenced. (B) Amino acid sequence of the association,hinge, and endostatin domains of human collagen XVIII NC1 domain. The cleavage site of MMP-7 was within the hinge domain and was differentfrom that of other catalytic enzymes.

2522 Lin et al. IOVS, October 2001, Vol. 42, No. 11

to generate a 28-kDa fragment. This fragment differs from thepublished endostatin molecule, which is a 20- to 22-kDa frag-ment with antiangiogenic properties.9 We have not as yetascertained the antiangiogenic potential of this 28-kDa matri-lysin-derived fragment.

Human MMP-7 is known to cleave ECM and basement mem-brane proteins, such as fibronectin, collagen type IV, laminin,elastin, entactin, and proteoglycan aggregates.36 MMP-7 alsomediates proteolytic processing of the precursor of tumornecrosis factor a and urokinase plasminogen activator and hasbeen shown to release membrane-bound Fas ligand.34,35 It hasalso been shown to cleave plasminogen and generate angio-statin, another potent antiangiogenic molecule.37 We havereported the presence of MMP-7 in the cornea and its upregu-lation after excimer laser keratectomy.21 We localized MMP-7in the epithelium and the basement membrane area, suggestinga colocalization of this enzyme with collagen XVIII, whichbelongs to the heparan-sulfate proteoglycan family of mole-cules, known to act as MMP-7 ECM docking molecules.38

Because MMP-7 is the smallest member of the MMP family,it may easily access the collagen XVIII cleavage site. Ferreras etal.39 have recently reported that the recombinant collagenXVIII NC1 domain can be processed by MMP-3, -9, -12, and -13with a relative low activity. In the recombinant NC1 fragment,cleavage sites may be more accessible to MMPs than immuno-precipitated native collagen molecules. The Western blot ex-periments used in our study suggest that, under our experi-mental conditions (exposure and concentration), MMP-7 ismore efficient than the other tested MMPs in generating a28-kDa molecule, but our experiments do not establishwhether MMP-7 plays a role in the cleavage of corneal collagenXVIII in vivo. Further investigations may allow us to assess thebiological function of this 28-kDa fragment generated by theenzymatic activity of MMP-7.

Acknowledgments

The authors thank Hiroshi Handa for providing the SV-40 immortalizedhuman corneal epithelial cells and Bjorn Olsen for providing anti-collagen XVIII antibodies.

References

1. Oh SP, Kamagata Y, Muragaki Y, Timmons S, Ooshima A, Olsen BR.Isolation and sequencing of cDNAs for proteins with multipledomains of Gly-Xaa-Yaa repeats identify a distinct family of collag-enous proteins. Proc Natl Acad Sci USA. 1994;91:4229–4233.

2. Oh SP, Warman ML, Seldin MF, et al. Cloning of cDNA andgenomic DNA encoding human type XVIII collagen and localiza-tion of the alpha 1(XVIII) collagen gene to mouse chromosome 10and human chromosome 21. Genomics. 1994;19:494–499.

3. Rehn M, Hintikka E, Pihlajaniemi T. Primary structure of the alpha1 chain of mouse type XVIII collagen, partial structure of thecorresponding gene, and comparison of the alpha 1(XVIII) chainwith its homologue, the alpha 1(XV) collagen chain. J Biol Chem.1994;269:13929–13935.

4. Sasaki T, Fukai N, Mann K, Gohring W, Olsen BR, Timpl R. Struc-ture, function and tissue forms of the C-terminal globular domainof collagen XVIII containing the angiogenesis inhibitor endostatin.EMBO J. 1998;17:4249–4256.

5. Saarela J, Rehn M, Oikarinen A, Autio-Harmainen H, Pihlajaniemi T.The short and long forms of type XVIII collagen show clear tissuespecificities in their expression and location in basement mem-brane zones in humans. Am J Pathol. 1998;153:611–626.

6. Saarela J, Ylikarppa R, Rehn M, Purmonen S, Pihlajaniemi T. Com-plete primary structure of two variant forms of human type XVIIIcollagen and tissue-specific differences in the expression of thecorresponding transcripts. Matrix Biol. 1998;16:319–328.

7. Muragaki Y, Timmons S, Griffith CM, et al. Mouse Col18a1 isexpressed in a tissue-specific manner as three alternative variants

and is localized in basement membrane zones. Proc Natl Acad SciUSA. 1995;92:8763–8767.

8. Halfter W, Dong S, Schurer B, Cole GJ. Collagen XVIII is a base-ment membrane heparan sulfate proteoglycan. J Biol Chem. 1998;273:25404–25412.

9. O’Reilly MS, Boehm T, Shing Y, et al. Endostatin: an endogenousinhibitor of angiogenesis and tumor growth. Cell. 1997;88:277–285.

10. Kamphaus GD, Colorado PC, Panka DJ, et al. Canstatin, a novelmatrix-derived inhibitor of angiogenesis and tumor growth. J BiolChem. 2000;275:1209–1215.

11. Ramchandran R, Dhanabal M, Volk R, et al. Antiangiogenic activityof restin, NC10 domain of human collagen XV: comparison toendostatin. Biochem Biophys Res Commun. 1999;255:735–739.

12. Felbor U, Dreier L, Bryant RA, Ploegh HL, Olsen BR, Mothes W.Secreted cathepsin L generates endostatin from collagen XVIII.EMBO J. 2000;19:1187–1194.

13. Wen W, Moses MA, Wiederschain D, Arbiser JL, Folkman J. Thegeneration of endostatin is mediated by elastase. Cancer Res.1999;59:6052–6056.

14. Murphy G, Gavrilovic J. Proteolysis and cell migration: creating apath? Curr Opin Cell Biol. 1999;11:614–621.

15. de Coignac AB, Elson G, Delneste Y, et al. Cloning of MMP-26: anovel matrilysin-like proteinase. Eur J Biochem. 2000;267:3323–3329.

16. Park HI, Ni J, Gerkema FE, Liu D, Belozerov VE, Sang QX. Identi-fication and characterization of human endometase (matrix metal-loproteinase-26) from endometrial tumor. J Biol Chem. 2000;275:20540–20544.

17. Shapiro SD. Matrix metalloproteinase degradation of extracellularmatrix: biological consequences. Curr Opin Cell Biol. 1998;10:602–608.

18. Nagase H, Woessner JF Jr. Matrix metalloproteinases. J Biol Chem.1999;274:21491–21494.

19. Azar DT, Hahn TW, Jain S, Yeh YC, Stetler-Stevensen WG. Matrixmetalloproteinases are expressed during wound healing after ex-cimer laser keratectomy. Cornea. 1996;15:18–24.

20. Ye HQ, Azar DT. Expression of gelatinases A and B, and TIMPs 1and 2 during corneal wound healing. Invest Ophthalmol Vis Sci.1998;39:913–921.

21. Lu PC, Ye H, Maeda M, Azar DT. Immunolocalization and geneexpression of matrilysin during corneal wound healing. InvestOphthalmol Vis Sci. 1999;40:20–27.

22. Ye HQ, Maeda M, Yu FS, Azar DT. Differential expression ofMT1-MMP (MMP-14) and collagenase III (MMP-13) genes in normaland wounded rat corneas. Invest Ophthalmol Vis Sci. 2000;41:2894–2899.

23. Araki-Sasaki K, Ohashi Y, Sasabe T, et al. An SV40-immortalizedhuman corneal epithelial cell line and its characterization (seecomments). Invest Ophthalmol Vis Sci. 1995;36:614–621.

24. Ingber D, Folkman J. Inhibition of angiogenesis through modula-tion of collagen metabolism. Lab Invest. 1988;59:44–51.

25. Colorado PC, Torre A, Kamphaus G, et al. Anti-angiogenic cuesfrom vascular basement membrane collagen. Cancer Res. 2000;60:2520–2526.

26. Maeshima Y, Colorado PC, Torre A, et al. Distinct antitumor prop-erties of a type IV collagen domain derived from basement mem-brane. J Biol Chem. 2000;275:21340–21348.

27. Petitclerc E, Boutaud A, Prestayko A, et al. New functions fornon-collagenous domains of human collagen type IV: novel inte-grin ligands inhibiting angiogenesis and tumor growth in vivo.J Biol Chem. 2000;275:8051–8061.

28. Werb Z. ECM and cell surface proteolysis: regulating cellular ecol-ogy. Cell. 1997;91:439–442.

29. O’Reilly MS, Wiederschain D, Stetler-Stevenson WG, Folkman J,Moses MA. Regulation of angiostatin production by matrix metal-loproteinase-2 in a model of concomitant resistance. J Biol Chem.1999;274:29568–29571.

30. Dong Z, Kumar R, Yang X, Fidler IJ. Macrophage-derived me-talloelastase is responsible for the generation of angiostatin inLewis lung carcinoma. Cell. 1997;88:801–810.

IOVS, October 2001, Vol. 42, No. 11 Matrilysin and Endostatin in the Cornea 2523

31. Lijnen HR, Ugwu F, Bini A, Collen D. Generation of an angiostatin-like fragment from plasminogen by stromelysin-1 (MMP-3). Bio-chemistry. 1998;37:4699–4702.

32. Gately S, Twardowski P, Stack MS, et al. The mechanism of cancer-mediated conversion of plasminogen to the angiogenesis inhibitorangiostatin. Proc Natl Acad Sci USA. 1997;94:10868–10872.

33. Stathakis P, Lay AJ, Fitzgerald M, Schlieker C, Matthias LJ, Hogg PJ.Angiostatin formation involves disulfide bond reduction and pro-teolysis in kringle 5 of plasmin. J Biol Chem. 1999;274:8910–8916.

34. Cornelius LA, Nehring LC, Harding E, et al. Matrix metalloprotein-ases generate angiostatin: effects on neovascularization. J Immu-nol. 1998;161:6845–6852.

35. Patterson BC, Sang QA. Angiostatin-converting enzyme activities of

human matrilysin (MMP-7) and gelatinase B/type IV collagenase(MMP-9). J Biol Chem. 1997;272:28823–28825.

36. Wilson CL, Matrisian LM. Matrilysin: an epithelial matrix metallo-proteinase with potentially novel functions. Int J Biochem CellBiol. 1996;28:123–136.

37. O’Reilly MS, Holmgren L, Shing Y, et al. Angiostatin: a novelangiogenesis inhibitor that mediates the suppression of metastasesby a Lewis lung carcinoma. Cell. 1994;79:315–328.

38. Yu WH, Woessner JF Jr. Heparan sulfate proteoglycans as extra-cellular docking molecules for matrilysin (matrix metalloprotein-ase 7). J Biol Chem. 2000;275:4183–4191.

39. Ferreras M, Felbor U, Lenhard T, Olsen BR, Delaisse J. Generationand degradation of human endostatin proteins by various protein-ases. FEBS Lett. 2000;486:247–251.

2524 Lin et al. IOVS, October 2001, Vol. 42, No. 11