Ultraviolet Radiation and the Role of Matrix Metalloproteinases in the Pathogenesis of Ocular Surface Squamous Neoplasia

Abstract

purpose. Ocular surface squamous neoplasia (OSSN) is an uncommon tumor of the corneal and conjunctival epithelium associated with risk of permanent visual impairment. The purposes of this study were to (1) identify and localize potential mediators in tissue from patients with OSSN and (2) culture human dysplastic conjunctival epithelial cells (DCECs) to determine their responsiveness to ultraviolet (UV)-B radiation compared with normal conjunctival epithelial cells (NCECs).

methods. Immunohistochemical analysis was performed on OSSN (n = 23) and normal conjunctival (n = 17) tissue to identify matrix metalloproteinases (MMPs) and their tissue inhibitors (TIMPs). Cell viability as well as basal and UVB-modulated levels of MMPs and TIMPs from DCECs and NCECs was determined by immunoassays, zymography, and RT-PCR.

results. A higher proportion of diseased specimens stained for MMP-1 (83%), MMP-3 (86%), TIMP-2 (87%), and TIMP-3 (83%) compared with normal conjunctiva (41%, 41%, 47%, and 53%, respectively). UVB radiation induced cell death and apoptosis at doses ≥ 50 mJ/cm². MMP-1 and -3 mRNA and protein expression in DCECs was induced by UV and was mitogen-activated protein kinase–dependent, although the same enzymes were upregulated in NCECs only at doses that induced apoptosis. TIMP-1 and -2 levels remained relatively unchanged, except for a dose-dependent suppression of TIMP-3.

conclusions. The results suggest that MMPs and TIMPs play a significant role in the pathogenesis of OSSN and that UVB initiates and perpetuates the development of this lesion on the ocular surface.

Ocular surface squamous neoplasia (OSSN) is a dysplastic or carcinomatous lesion arising from epithelial cells in the conjunctiva, limbus, or cornea ¹ and is considered an uncommon disease with a prevalence of 0.2 to 3.5 cases per 100,000. ² The lesion is slow-growing, is usually confined to the ocular surface, and rarely metastasizes. Increased incidence of OSSN has been reported in the past decade ^{1 2 3} and despite improvements in treatment, recurrence rates range from 5% to 53%. ^{1 4}  

The pathogenesis of OSSN has yet to be definitively attributed to a specific etiological factor. Several risk factors have been reported and include chronic ultraviolet (UV) exposure, ⁵ human papilloma virus, ^{6 7} human immunodeficiency virus, ⁸ and smoking. ⁹ The role of limbal epithelial stem cells in the development of this disease is controversial. It has been speculated that OSSN may have arisen from dysfunctional limbal stem cells having been altered by mutagenic agents, such as UV radiation. ¹⁰ Despite multiple theories, evidence of UV radiation as the prime etiological trigger is likely, as epidemiologic studies show a linear relationship between incidence rates and proximity to the equator. ⁴ Of note, this disease shares some striking similarities to skin neoplasia, whose origins are defined. ¹¹ Independent of cellular and DNA injury, UV radiation can induce phosphorylation events in cytokine and growth factor receptors which in turn activate downstream transcription factors. ^{12 13 14}  

Currently, only a limited number of molecular studies have been undertaken to delineate the pathogenesis of OSSN. ^{8 15 16 17 18} An area lacking attention is the potential contribution of matrix metalloproteinases (MMPs) and their natural inhibitors, the tissue inhibitors of metalloproteinase (TIMPs). These molecules have been heavily implicated in the tissue invasion and metastasis that characterize many human neoplasias. ¹⁹ It is well documented that a fine-tuned balance between MMPs and TIMPs is necessary for regulated tissue remodeling, ²⁰ whereas discoordinate expression may result in pathologic conditions such as rheumatoid arthritis, ²¹ pterygium, ^{12 22 23 24 25} and scleritis. ²⁶ The complexity of MMP interactions is exemplified by the recent discovery of novel functions including modulation of inflammation via their ability to release ECM-bound mitogenic factors and activate cytokine receptors. ^{27 28}  

We hypothesize that there is a differential pattern of MMP and TIMP expression in OSSN that may be a consequence of UV radiation exposure at the ocular surface. In the present study, we identified a pattern of MMP and TIMP expression in OSSN tissue that implies that these molecules play a critical role in the development of this disease. Moreover, we demonstrated that cells derived from dysplastic tissue are more sensitive to UVB radiation than are normal conjunctival cells, with regard to MMP production, thereby providing preliminary clues as to pathogenesis of OSSN and potential preventative measures that can be recommended for at-risk subjects. 

Materials and Methods

Tissue Specimens

Twenty-three formalin-fixed, paraffin-embedded, histologically confirmed cases of OSSN (mild dysplasia, n = 5; moderate dysplasia, n = 3; severe dysplasia, n = 9, carcinoma in situ n = 6) were obtained from an archival tissue collection (Prince of Wales Hospital, Sydney). The inclusion criteria consisted of (1) confirmation of histopathologic diagnosis by an experienced pathologist who retrieved the original H&E sections for another review, (2) tissue collected between 2000 and 2006 from the same anatomic pathology hospital department, and (3) sufficient material in each tissue block to analyze by immunohistochemistry. Control conjunctival tissue was sourced from remnant nondiseased graft tissue from patients undergoing pterygium surgery. Control tissue was fixed in formalin, embedded in paraffin and used for immunohistochemical analysis. Demographics of patients with OSSN and control subjects are summarized in Table 1 . Positive control tissue included pterygium (n = 2), whereas negative control tissue consisted of corneal–limbal rims (n = 2) obtained from the Lions Eye Bank (Sydney, Australia). In some cases, tissue was used as explants to establish dysplastic and normal conjunctival epithelial cells in culture. Informed consent was obtained from each participant and experimental protocols were approved by the UNSW Human Research Ethics Committee (HREC-04/088) and performed in accordance with the tenets of the World Medical Association’s Declaration of Helsinki. 

Immunohistochemical Analysis

Serial tissue sections were cut (4 μm) and processed for immunohistochemical assessment, as previously described. ^{22 23 24 29} In brief, sections were deparaffinized and equilibrated in 0.05 M Tris-buffered saline (TBS; pH 7.6). Sections were incubated overnight at 4°C in a humidified chamber with preoptimized dilutions (∼100 ng/mL final) of commercially available mouse monoclonal antibodies (Table 2)directed against human MMP-1, -2, and -3 and TIMP-1, -2, and -3 (Calbiochem, San Diego, CA). Control reactions consisted of sections incubated with an appropriate isotype control antibody (Table 2)or in the absence of a primary antibody. Sections received a goat anti-mouse biotinylated secondary antibody (1:200 final dilution, product code E0433; Dako Cytomation, Carpinteria, CA) for 30 minutes, to amplify the antigenic signal. This amplification step was used to ensure the development of maximum immunoreactive signal (Supplementary Fig. S1). Sections were washed before the addition of HRP-conjugated streptavidin (Dako Cytomation) before the addition of 3-amino-9-ethylcarbazole (AEC; Sigma-Aldrich, St. Louis, MO), counterstained with hematoxylin, and mounted (UltraMount; Laboratory Vision Corp., Fremont, CA). Immunoreactivity was scored as either positive (present) or negative (absent) for each antigen and the results summarized in Table 3 . To determine whether disease severity correlated with MMP/TIMP protein level, an arbitrary score was assigned to the epithelium of each specimen to reflect staining intensity (1, no staining; 2, weak; 3, intermediate; 4, strong). This analysis has proven reliable for identifying MMPs in ocular surface disease ²³ and common solid tumors. ^{30 31} Direct staining comparison for an individual antigen was possible as diseased and control tissue was assessed in the one experimental run. Four observers (two of which were independent of the study and masked as to the disease status) scored each section with 100% concordance. Antigen retrieval in 10 mM citrate buffer (pH 6.0) was necessary only to unmask TIMP-3 antigenicity. 

Cell Culture Studies

The OSSN cells that were used in this study were derived from Caucasian men (age range, 60–82 years) who presented with right limbal lesions, histopathologically diagnosed as either full-thickness dysplasia (severe, n = 2) or squamous cell carcinoma in situ (n = 1). Human dysplastic (DCECs; n = 3) and normal conjunctival epithelial cells (NCECs; n = 2) were grown from tissue explants and established in pure long-term cultures, according to previously optimized protocols. ^{12 22 32} In brief, freshly resected diseased and normal conjunctiva was cut into 1- to 2-mm² segments, placed in six-well culture plates (Greiner bio-one, Frickenhausen, Germany) and incubated at 37°C in a humidified incubator set to 5% CO₂ in Eagle’s minimum essential medium (EMEM) containing 10% FBS (ThermoTrace, Melbourne, Australia), 100 U/mL penicillin, 100 μg/mL streptomycin, and 2 mM l-glutamate (ThermoTrace). Explants were removed after 10 days, cells subcultured weekly and purity established after three to four generations with a pan-cytokeratin marker (MNF-116^FITC; Dako Cytomation) and p63 (Clone 4A4; Santa Cruz Biotechnology, Santa Cruz, CA). Mouse IgG₁ ^FITC (Dako Cytomation) and mouse IgG_2a ^FITC (Jackson ImmunoResearch, West Grove, PA) were used as appropriate isotype control antibodies for the flow cytometric analysis. ³² Primary cultures reached replicative senescence between passages 15 and 20. 

For experimentation, early-generation (passages 4–10) epithelial cells were seeded (1.5 × 10⁵ and 0.5 × 10⁶) into six-well plates or 100-mm dishes (Greiner bio-one), cultured to semiconfluence, placed in serum-free medium overnight, and then exposed to physiological doses of UVB (0–60 mJ/cm²) in phosphate-buffered saline (PBS), with TL 20-W/12 RS bulbs (Philips, Sydney, Australia) in triplicate, as previously reported. ^{12 22} UVB light intensity was monitored and calibrated before each experiment with the aid of a radiometer (IL1400A; International Light, Newburyport, MA). In some experiments, the cells were not irradiated but were stimulated with known MMP inducers, including phorbol mistrate acetate (PMA, 10 ng/mL; Sigma-Aldrich) and recombinant human epidermal growth factor (rhEGF, 50 ng/mL; Sigma-Aldrich). For intracellular signaling, the cells were preincubated with pharmacologic mitogen-activated protein kinase (MAPK) inhibitors at 20 μM for ERK1/2 (PD98059; Calbiochem), 10 μM for p38 (SB203580; Calbiochem), and 1 to 5 μM for JNK (SP600125; Calbiochem) before irradiation. Supernatants were harvested at 72 hours and RNA samples at 48 hours after irradiation, stored at −80°C, and used for biochemical or molecular analysis. Of the three DCEC lines generated, a representative line (see 1 2 Figs. 3E 3F 3G 3H ; HDCE.MO.1b) was used in subsequent experiments since no significant differences were noted in growth rate, senescence, serum independence, response to UVB or the nonspecific mitogen PMA, or basal production of MMP-2 (data not shown). Moreover, cells derived from dysplastic lesions demonstrated similar phenotypic and biochemical features (see Fig. 3A 3B 3C 3D 3E 3F 3G 3H ). 

Cytotoxicity Post-UVB Exposure

Cytotoxicity was determined in cells after exposure to various doses of UVB. The cells were photographed under phase–contrast microscopy (Eclipse TE 2000-S; Nikon, Kanagawa, Japan) with image-capture software (Image-Pro Plus, ver. 5.1.2; MediaCybernetics, Silver Spring, MD). Culture media were removed, and cells in each plate were released from their plastic substratum by enzymatic dissociation (0.05% trypsin/0.02% EDTA; ThermoTrace). Cells already in suspension and those enzymatically released, were pooled and washed in 10% FBS/EMEM, and cell viability was determined by trypan blue (Cytosystems, Sydney, Australia) exclusion. 

Apoptosis was assessed with an annexin-V detection kit (BD-Pharmingen; Franklin Lakes, NJ), according to the manufacturer’s instructions. In brief, DCECs were irradiated with the indicated doses, collected after 48 hours of enzymatic digestion, washed with 10% FBS/EMEM and resuspended in cold PBS. Cells in suspension were acquired on a flow cytometer (FACSort; BD-Pharmingen), and scatterplots were generated to discriminate necrotic cells by propidium iodine (PI) and apoptotic cells by annexin-V labeling. 

Gelatin Substrate Zymography and Reverse Zymography

Gelatin zymography (to detect MMP-2 and -9) and reverse zymography (to identify TIMP-1 and -2) was performed on undiluted supernatants as previously described. ^{12 22 29 33} Supernatants (25 μL) standardized for cell number, were mixed with nonreducing sample buffer (8 μL; 0.25 M Tris-HCl [pH 6.8], 10% SDS, 4% sucrose, and 0.1% bromophenol blue) and loaded under nondenaturing conditions into 10% SDS-PAGE resolving gels containing 1 mg/mL gelatin (Sigma-Aldrich). The procedure for reverse zymography was similar, except gelatin and proMMP-2 (250 ng/mL; Calbiochem) were copolymerized into 12% acrylamide resolving gels. Adjacent lanes were loaded with recombinant human TIMP-1 and -2 (15 ng; Calbiochem). Gelatinolytic zones (corresponding to MMPs) and protected bands (corresponding to TIMPs) were developed and semiquantified (Gel Doc 2000 and Quantity One V4.5.1 software; Bio-Rad). 

Immunoassays

The simultaneous detection and quantification of eight human MMPs (MMP-1, -2, -3, -7, -8, -9, -12, and -13) was determined in supernatants derived from irradiated DCECs and NCECs using a commercial multianalyte fluorokine profiling assay (Bioplex Fluorokine MultiAnalyte Profiling; R&D Systems, Minneapolis, MN), as instructed by the manufacturer. In brief, triplicate samples were diluted (1:1) in 96-well microplates before incubation with antigen-specific, antibody-coated, color-coded microparticles. Beads were washed, a biotin antibody cocktail was added, the mixture was incubated with streptavidin-PE, and the plates were read (Bioplex 100 Plate Reader System; Bio-Rad). 

An enzyme linked immunosorbent assay (ELISA) was used to quantify human TIMP-1 levels in culture supernatants (Biotrak ; GE Healthcare, Sydney, Australia). In brief, triplicate samples were diluted (1:4) before loading onto antibody-precoated 96-well plates, wells were washed several times, and then secondary and tertiary antibodies and reagents were added according to the manufacturer’s instructions. The plates were read at 450 nm with a spectrophotometer. Internal standards were provided to extrapolate TIMP-1 concentration. 

RNA Extraction and RT-PCR

DCEC and NCEC cells were cultured on 100-mm dishes and exposed to 0, 20, and 50 mJ/cm² of UVB. Total RNA was extracted after 48 hours with a commercially available RNA extraction kit (RNAgents Total RNA Extraction Kit; Promega, Madison, WI). The final RNA pellet was resuspended (50 μL) in DEPC-H₂O and quantified spectrophotometrically (ND-1000; Nanodrop Technologies, Wilmington, DE). One microgram of RNA from each treatment group was reverse transcribed into cDNA (Superscript III First-Stand Synthesis Kit; Invitrogen, Groningen, The Netherlands). cDNA synthesis and subsequent PCR were performed (GeneAmp PCR system 2400; Perkin Elmer, Wellesley, MA) using MMP-1 and -2; TIMP-1, -2, and -3; and GAPDH forward and reverse gene specific primers. ^{22 33} Primers for MMP-3 were (forward: 5′-GGC-TGT-ATG-AAG-GAG-AGG-CTG-3′; reverse: 5′-TGG-TCC-CTG-TTG-TAT-CCT-TTG-3′). Cycling was initiated with a 95°C hot start for 2 minutes, followed by 95°C for 30 seconds, 58°C for 30 seconds, and 72°C for 30 seconds. Each reaction was terminated with a final 2-minute extension at 72°C. The cycle number was predetermined so that the products formed fell within the linear portion of the amplification curve. Reactions were performed in triplicate, and each product was visualized on 1.5% to 3% agarose gels and semiquantified (Gel Doc 200; and the Quantity One program; Bio-Rad) after normalizing to GAPDH. A DNA ladder (Gene Ruler IV; Fermentas, Glen Burnie, MD) was included in adjacent lanes. 

Statistical Analysis

Data analysis was performed with the commercially available statistical software package (GraphPad Prism; ver. 5.01; San Diego, CA). Differences between MMP/TIMP immunostaining in OSSN versus control tissue were analyzed by the Fisher exact test, where significance was defined as P < 0.05. Correlation between MMP/TIMP staining levels and OSSN severity was tested using Spearman rank. In cell viability assays and immunoassays, differences between mock-irradiated and UVB-exposed cells were tested by one-way analysis of variance (ANOVA) followed by the Dunnett test for multiple comparison of treated groups with control where significance was set at P < 0.05. 

Results

Immunohistochemical Analysis

Given the reported roles of MMPs and TIMPs in neoplastic disease, ¹⁹ we attempted to localize several members of this family of proteases and their inhibitors in OSSN tissue and their distribution compared with normal conjunctiva (Figs. 1 2) . Staining on each specimen was scored as either positive (present) or negative (absent), and the results were summarized as the percentage of positively stained tissue over the total number of specimens analyzed (Table 3) . Antigenicity for each enzyme or inhibitor was predominantly confined to the epithelial component and was also identified on infiltrating inflammatory cells ^{26 33} but was rarely detected in stromal cells. Immunoreactivity for MMP-1 and -3 was prominent in the dysplastic epithelium compared with normal conjunctiva (compare Fig. 1A and 1Ewith Fig. 1B and 1F , respectively). Both enzymes were significantly overrepresented in diseased versus normal tissue (P = 0.02 and P = 0.01, respectively). When present in normal conjunctiva, MMP-1 intensity was diminished compared with OSSN and was associated with mild subepithelial inflammation (Fig. 1B , arrows). MMP-2 was identified in dysplastic (Fig. 1C)and in normal (Fig. 1D)conjunctival epithelium, with no significant difference in the proportion of positively stained diseased versus control specimens (P = 1.0). Epithelial cells within the normal cornea were occasionally illuminated with the same antibody (Fig. 1G , arrows). 

TIMP expression (Fig. 2)closely resembled the distribution of MMPs (Fig. 1) . Although TIMP-1 staining was less apparent in OSSN tissue (Fig. 2A)compared with normal conjunctiva (Fig. 2B) , the percentage of positively stained specimens did not differ (Table 3 ; P = 0.33). TIMP-2 and -3 were more likely to be present in diseased compared with normal tissue (Table 3) . Other striking differences in staining intensity (a possible indicator of enhanced antigen production) were noted (compare Figs. 2C and 2Ewith 2D and 2F, respectively). In addition, a differential pattern of TIMP-3 expression was occasionally noted within the one tissue specimen, where cells displaying normal morphology expressed significantly less antigen compared with those with typical features of dysplasia (Fig. 2E , arrows). However, no significant association was found between MMP/TIMP protein levels and disease severity (Table 4) . As well as cytoplasmic reactivity for these antigens, nuclear staining was often observed for both TIMP-3 (Fig. 2G , arrows) and TIMP-2 (Fig. 2H , arrows), a pattern previously described. ³⁴  

Cell Culture and Cytotoxicity after UVB Irradiation

Although it has been documented that domestic animal species are prone to development of OSSN-like lesions, ^{35 36} to the best of our knowledge, no experimental animal model has been described. Therefore, primary epithelial explant cultures derived from either dysplastic (Figs. 3A 3E)or normal conjunctiva (Fig. 3I)were established. In culture, morphologic differences were noted between primary (Figs. 3A 3E 3I)and passaged (Figs. 3B 3F 3J)cells; otherwise, cells of different lineage were indistinguishable. Primary cultures formed tightly adherent polygonal monolayers, whereas passaged cells were loosely adherent and more elongated. Flow cytometric analysis with a pan-cytokeratin antibody indicated epithelial purity of >95% for each cell type (Figs. 3C 3G 3K) . p63 staining confirmed that these cells were actively proliferating (Figs. 3D 3H 3L) . 

Given the extensive epidemiologic data implicating UV radiation as a triggering agent for OSSN, ^{1 2 3 4} DCECs (Fig. 4A) , and NCECs (Fig. 4B)of similar generation and cell density were plated and exposed to a range of UVB, and the morphologic appearance after exposure was recorded and cell viability estimated by trypan blue exclusion (Figs. 4A 4B ; bar graphs). Doses ranging 0 to 40 mJ/cm² caused no morphologic change in either cell type. In contrast, the cells began to detach, and some were lysed when exposed to ≥50 mJ/cm² (Fig. 4 , arrows). Stimulation with a nonspecific mitogen (PMA) caused no obvious morphologic change. 

UVB-induced apoptosis in DCECs was measured by an alternative approach that included annexin-V labeling by flow cytometry (Fig. 5) . The results demonstrated that 20 mJ/cm² was nontoxic; however, a dose of 50 mJ/cm² caused a threefold increase in early apoptosis compared with mock-irradiated cells or cells stimulated with 20 mJ/cm² UVB. 

Expression of MMP/TIMP mRNA Transcripts in UVB-Exposed Epithelial Cells

Having established a viable in vitro model, we exposed epithelial cells to UVB to determine whether this treatment modulates MMP/TIMP mRNA expression. Using reverse transcription (RT)-PCR, MMP-1, -2 (not shown), and -3 and TIMP-1, -2 (not shown), and -3 transcripts were detected in DCECs (Fig. 6A)and NCECs (Fig. 6B) . Moreover, it was apparent that the MMP-1, MMP-3, and TIMP-3 transcripts were modulated by UVB in the DCECs, whereas little or no effect was observed in the NCECs. Semiquantitative analysis of the data for the DCECs showed that MMP-1 and -3 (but not MMP-2) mRNA was dose-dependently induced after UVB (Fig. 6C) . TIMP-1 and -2 transcripts were unchanged, whereas TIMP-3 was dose-dependently suppressed in the DCECs (Fig. 6D) . 

Detection of MMP and TIMP Protein in Cultured Epithelial Cells

To corroborate the mRNA analysis, supernatants from the UVB-exposed epithelial cells were analyzed for MMP content by multianalyte profiling. Of interest, only MMP-1, -2, and -3 were detected in conditioned media from diseased and normal cells. The results illustrate a dose-dependent induction of both MMP-1 and -3 in the DCECs (Figs. 7A 7C , respectively). Although both MMP-1 and -3 were induced by UVB in the NCECs, this increase was only apparent at sublethal exposures (Figs. 7B 7D , respectively). MMP-2 levels remained unchanged in both cell types (data not shown). Stimulation with PMA dramatically induced MMP-1 and -3 in both cell types. In addition to the soluble-secreted MMPs measured, constitutive cell-associated MMP-1 was localized to the same cells (Supplementary Fig. S2). 

The gelatinolytic profile of the DCECs (Fig. 8A)and NCECs (Fig. 8B)was compared by gelatin zymography. The results demonstrated that MMP-2 was constitutively expressed and not modulated by UVB exposure. Furthermore, MMP-9 was not constitutively produced, nor was this enzyme stimulated by UVB, but it was induced by PMA. Although the gelatinolytic profile for both cell types was similar, MMP-2 levels were approximately twofold higher in the NCECs than in the DCECs (Fig. 8C) . The other obvious difference was the presence of a minor active MMP-2 species in NCECs (Fig. 8B , arrowheads). The reason for the enhanced activation status and elevated production of MMP-2 in normal epithelial is currently unknown but could be related to membrane-type MMPs on the cell surface of NCECs and/or genetic variation in MMP promoters. ³⁰ Corroborating this in vitro evidence, intense MMP-2 expression was also noted in normal basal corneal epithelial cells (Fig. 1G , arrows). The same supernatants were also analyzed by reverse-zymography, which demonstrated unaffected levels of both TIMP-1 and -2 proteins after UVB irradiation (Fig. 8D) . These results were confirmed by TIMP-1 ELISA in DCECs (Fig. 8E) . 

It has been documented in the eye ^{12 22} and skin ³⁷ that the UV-mediated induction of MMPs may be via activation of intracellular signaling pathways that involve MAPKs. In an attempt to delineate the mechanism(s) responsible for the UVB-mediated induction of MMPs in our model, DCECs were preincubated with MAPK inhibitors and then exposed to a single 20 mJ/cm² dose of UVB. The ERK 1/2 inhibitor (PD98059) significantly abated MMP-1 production by more than 50%, whereas the p38 inhibitor (SB203580) showed no inhibitory effect (Fig. 9) . This was a specific interaction, as MMP-2 was unaffected by the presence of either inhibitor (Fig. 9 , zymogram gel). 

Discussion

Several unsubstantiated theories have been proposed to explain how OSSN develops. Among these are those implicating viral infection ^{6 7 8 38} and UV radiation exposure. ^{1 2 3 4} In the current investigation, several novel observations were made implicating MMPs and TIMPs in the pathogenesis of OSSN. Our study, containing the largest case–control series analyzed by an immunohistochemical technique, demonstrated higher prevalence of MMP-1 and -3 and TIMP-2 and -3 in dysplastic compared with normal ocular surface tissue (Figs. 1 2) . Furthermore, a culture model was established to simulate physiological solar radiation exposure in a controlled laboratory setting, where epithelial cells derived from dysplastic tissue were more responsive to UV radiation when compared with cells from normal conjunctiva. 

MMPs have been extensively documented in many human cancers ^{19 30 31} and are responsible for the invasive and metastatic behavior of tumor cells. Therefore, their detection in a premalignant ocular condition is not surprising. Indeed, we have identified these proteases in other inflammatory, degenerative, and destructive anterior segment disorders, including pterygia, ^{23 24 25 29} scleritis, ²⁶ cataract, ³⁹ uveitis, ⁴⁰ and bacterial keratitis. ⁴¹ Of note, in pterygia, a disease that displays characteristic features similar to those of OSSN, including its benign nature, growth on the ocular surface and association with UV radiation exposure, we consistently noted overexpression of MMPs in relation to the counter-regulatory inhibitors (TIMPs) in the proliferating epithelium. ^{12 24} This contrasted remarkably with the distribution of these effector molecules in OSSN, where both MMPs and TIMPs were highly represented (Figs. 1 2 ; Table 3 ). This apparent counteracting activity may be one reason why these lesions are slow growing and less invasive compared with pterygia in which aggressive centripetal migration of altered epithelial cells into the normal cornea is a prominent feature. ⁴² Despite these differences, one question that remains unanswered is why pterygia and OSSN develop as distinct entities; given that both arise from a similar anatomic location. Without an animal model, the answer may remain elusive, although genetic predisposition cannot be ruled out. Although several studies have demonstrated a strong correlation between disease severity and MMP expression in other cancers, ⁴³ the present study identified no such association (Table 4) . Small sample series or subtle overlapping features that distinguish the severity of this disease may be reasonable explanations. Clearer differences may have been evident if comparisons were made between benign dysplastic disease and malignant carcinoma. 

One observation worthy of mention is the significant expression of the TIMPs in OSSN. Although these inhibitors counteract MMP activity, their overrepresentation in OSSN is curious and may indicate a natural tumor response to enzyme production. Alternatively, it has been documented that these inhibitors are required for MMP activation ^{44 45} and possess mitogenic ^{44 46} as well as antiapoptotic ⁴⁷ activity, thereby providing an alternative mechanism for tumor promotion. In a model proposed by Jiang et al., ⁴⁸ it was suggested that, although the antiproteolytic activity of TIMPs may play an inhibitory role during late-stage tumor progression, their growth-promoting and antiapoptotic functions may provide a stimulatory microenvironment during early tumor formation. Although speculative, this model could explain why OSSN remains arrested in a premalignant stage and rarely metastasizes. 

Etiological factors for OSSN have been derived from numerous epidemiologic and clinical studies, ^{1 2 3 4 5 9} with UV radiation exposure the most likely triggering agent. In addition to skin, the other major organ continuously exposed and particularly vulnerable to the effects of UV is the human eye. ^{49 50} Solar radiation reaching the Earth’s surface is a continuum of electromagnetic radiation spanning the UVA and UVB spectrum of wavelengths. Photobiological processes are wavelength-dependent and each has distinct biological effects. ⁵¹ In this study, we chose to irradiate ocular epithelial cells with UVB, which is less penetrating than UVA and hence more likely to affect surface keratinocytes. Our in vitro data indicate that while MMP-1 and -3 were constitutively expressed in cultured DCECs, they were markedly upregulated by UVB radiation in a dose-dependent manner (Figs. 6 7) . In contrast, while the NCECs displayed similar basal levels of both MMP-1 and -3, these enzymes were not induced by UVB between 0 and 40 mJ/cm². We can only speculate as to what may account for the apparent sensitivity differences between the two cells types. Using an alternative NCECs line, we previously demonstrated a similar affect. ¹² It is likely that NCECs have mechanisms to counter the phototoxic effects of UV, thus providing an explanation as to the relatively rare occurrence of conjunctival neoplasia. Alternatively, the observed differential sensitivity may be related to the density of cell-surface receptors, which are known transmitters of stress signals. ⁵² Indeed, it could be argued that the conjunctival-derived epithelial cells used in the present study were not the appropriate comparative cell type to use, and future investigations should include corneal/limbal epithelial cells. However, harvesting sufficient normal limbal tissue for tissue culture purposes is ethically challenging. We have performed similar studies on cultured limbal-derived epithelial cells and demonstrated no response to a single dose (20 mJ/cm²) of UVB compared to pterygium cells. ⁵³  

UVB is known to activate several membrane-bound growth factor and cytokine receptors, notably EGFR, TNFR, and IL-1R, causing receptor clustering, phosphorylation, internalization, intracellular signaling, and downstream production of cytokines and MMPs, either in a ligand-dependent or ligand-independent manner. ^{11 12 37 54} Although growth factor receptors were not localized, we provided some preliminary clues as to the potential intracellular pathway activated after UVB exposure. The involvement of the EGFR in neoplasia has been the topic of considerable investigation, particularly in the setting of skin cancer, ⁵⁵ and recently in OSSN. ¹⁶ Of note and relevant to OSSN, the EGFR is known to regulate keratinocyte proliferation, differentiation, and survival. ⁵⁶ EGFR is likely to orchestrate the UVB-induced MMP expression in DCECs, as we have recently demonstrated a similar role for this receptor in pterygium’s pathogenesis. ¹² Moreover, the UVB-mediated induction of MMP-1 in DCECs was exclusively suppressed by an ERK inhibitor, implying activation of an MAPK pathway. 

Angiogenesis is a significant prognostic determinant in patients with cancer ⁵⁷ ; however, such investigations have not been performed in OSSN. Although these lesions are known to sustain a vascular reaction, clinically, angiogenesis is conspicuous compared with other ocular lesions, ^{23 58} and one molecule that may influence this pathogenic process is TIMP-3. Qi et al. ⁵⁹ recently demonstrated angiogenic abrogation, as TIMP-3 was identified as the factor responsible for inhibiting the binding of VEGF to its receptor. Our in vivo (Fig. 2 , Table 3 ) and in vitro (Figs. 6A 6D)data for TIMP-3 were not in complete agreement. On the one hand, TIMP-3 was highly represented in OSSN tissue and occasionally overexpressed in dysplastic epithelial compared with cells displaying normal morphology (Fig. 2E , arrows); on the other hand, TIMP-3 mRNA was suppressed by UVB exposure. TIMP-3 protein levels were not measured in the present study as this member is typically sequestered on ECM components ⁶⁰ and is difficult to measure as a soluble protein by immunoassays. Recently, it was discovered that methylation events contribute to transcriptional repression of TIMP-3 in neoplastic human ⁶¹ and mouse ⁶² cells. Nonetheless, the mechanism(s) responsible for TIMP-3 suppression in dysplastic cells after UVB requires further elucidation. It is difficult to ascertain whether TIMP-3 expression is similarly modulated in vivo but tissue-based studies should be interpreted with caution as they represent only a snapshot of tumor evolution. 

Although the precise contribution of MMPs and TIMPs in OSSN remains to be clearly defined, to our knowledge, the present study is the largest of its kind on the localization of relevant effector molecules. In the absence of an animal model, our in vitro system is the first to model one component of OSSN (the epithelium) using UV radiation as the stimulatory agent to determine its role on the expression of proteolytic enzymes relevant to neoplasia. Furthermore, the genetic switch, transforming dysplastic lesions into neoplastic carcinomas has yet to be determined. An investigation into differentially expressed genes using gene microarray analysis in dysplasia versus malignant carcinoma may yield clues as to the likely therapeutic targets. 

 Supported by National Health and Medical Research Council of Australia Grant 455358 and the Ophthalmic Research Institute of Australia.

 Submitted for publication March 6, 2008; revised May 11 and July 1, 2008; accepted October 2, 2008.

 Disclosure: J. Ng, None; M.T. Coroneo, None; D. Wakefield, None; N. D i Girolamo, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Table 1.

 

Patient Demographics

	Dysplastic Tissue	Normal Conjunctiva
Subjects, n	23	17
Male, n (%)	17 (74)	12 (71)
Female, n (%)	6 (26)	5 (29)
Average age, y (range)	62.7 (21–91)	53.1 (21–79)
Right eyes, n (%)	13 (56.5)	7 (41)
Left eyes, n (%)	10 (43.5)	10 (58)

Table 2.

 

Primary Antibodies used for Immunohistochemistry

Antibody	Source	Catalog No.	Clone No.	Dilution
MMP-1	Calbiochem	IM35L	41-1E5	1:600
MMP-2	Calbiochem	IM33	42-5D11	1:100
MMP-3	Calbiochem	IM36L	55-2A4	1:15
TIMP-1	Calbiochem	IM41L	147-6D11	1:60
TIMP-2	Chemicon	MAB3310	67-4H11	1:200
TIMP-3	Calbiochem	IM43L	136-13H4	1:200
Mouse IgG1	Dako	X0931	—	1:100
Mouse IgG2a	Dako	X0943	—	1:600

 All MMP/TIMP antibodies were raised in the mouse and directed against human proteins.

Table 3.

 

Immunohistochemical Analysis

	MMP-1	MMP-2	MMP-3	TIMP-1	TIMP-2	TIMP-3
OSSN (n = 23)	19 (83)	19 (82)	20 (86)	14 (61)	20 (87)	19 (83)
Normal conjunctiva (n = 17)	5 (29)	14 (83)	7 (41)	13 (76)	8 (47)	9 (53)
Fisher exact test P	0.02	1.00	0.01	0.33	0.01	0.01

 Samples were screened as either positive or negative for the presence or absence of the respective antigen. The percentage of positively staining samples from diseased (OSSN) or control tissue is shown in parentheses.

Figure 1.

 

Localization of MMPs in dysplastic and control conjunctival tissue. Dysplastic (A, C, E) and control (B, D, F) conjunctival tissues were serially sectioned and analyzed immunohistochemically with antibodies directed against human MMP-1 (A, B), -2 (C, D, G), and -3 (E, F). In some sections, the primary antibody was omitted (data not shown) or replaced with an appropriate isotype control (IgG_2a; A, B, inset). Immunoreactivity was denoted by the red cytoplasmic staining, while cell nuclei were distinguished with hematoxylin. The photomicrograph displayed in (G) is derived from a normal corneal–limbal rim and illustrates occasional MMP-2 reactivity in basal epithelial cells (arrows). (B, arrows) Mild subepithelial inflammation associated with MMP-1 reactivity. Original magnification, ×1000.

Figure 2.

 

Localization of TIMPs in dysplastic and control conjunctival tissue. Diseased (A, C, E, G, H) and control (B, D, F) conjunctival tissue was serially sectioned and analyzed immunohistochemically with antibodies directed against human TIMP-1 (A, B), -2 (C, D, H), and -3 (E, F, G). In some sections, the primary antibody was omitted or replaced with an appropriate isotype control (data not shown). Immunoreactivity and counterstaining was as described in Figure 1 . (E, arrows) A region of abrupt transition from epithelial cells with normal morphology (left), into dysplastic epithelium with loss of polarity (right). (G, H, arrows) Cells with nuclear-associated TIMP-3 and -2 reactivity, respectively. Original magnification, ×1000.

Figure 3.

 

Culture and characterization of ocular surface epithelial cells. Diseased epithelial cells derived from a patient with severe dysplasia (A–D), carcinoma in situ (E–H), or normal conjunctiva from remnant graft tissue harvested from a patient undergoing pterygium resection (I–L) were grown from explants (A, E, I). The cells were subsequently passaged (B, F, J), and cytokeratin (red; C, G, K) as well as p63 (red; D, H, L) expression was determine by flow cytometry in conjunction with appropriate isotype control antibodies (blue). Original magnification, ×100.

Table 4.

 

Correlation between MMP/TIMP Expression and Disease Severity

Case	Histologic Diagnosis	MMP-1	MMP-2	MMP-3	TIMP-1	TIMP-2	TIMP-3
1	Mild	4	3	3	2	3	2
2	Mild	4	3	3	3	3	3
3	Mild	3	2	2	1	2	2
4	Mild	4	3	3	3	2	3
5	Mild	3	3	2	2	3	2
6	Moderate	4	2	2	1	2	2
7	Moderate	1	1	1	1	3	3
8	Moderate	4	4	3	3	3	3
9	Severe	1	1	1	1	1	1
10	Severe	1	1	2	1	1	1
11	Severe	4	2	3	2	2	3
12	Severe	1	2	1	1	1	2
13	Severe	3	3	3	2	3	3
14	Severe	4	4	4	2	4	4
15	Severe	3	2	2	1	2	1
16	Severe	3	2	3	2	3	4
17	Severe	4	2	3	1	2	*
18	CIN	4	4	3	2	2	2
19	CIN	4	3	3	2	2	3
20	CIN	3	3	3	2	3	3
21	CIN	4	3	4	3	3	4
22	CIN	3	4	3	2	4	4
23	CIN	2	1	2	1	2	3
	P	0.6209	0.7174	0.2538	0.8026	0.8326	0.182

 Numbers in each column represent staining intensity (1, no staining; 2, weak; 3, intermediate; 4, strong). Two-tailed P was calculated with the Spearman rank test. CIN, carcinoma in situ.

*  Missing value (specimen not stained).

Figure 4.

 

Cell viability after UV exposure. Phase-contrast microscopy was used to morphologically assess DCECs (A) and NCECs (B) 72 hours after exposure to various doses of UVB radiation (see labels on individual micrographs). Arrows in the micrographs labeled (50) and (60) identify some necrotic and floating cells. Adherent and floating diseased and normal cells were collected, cell viability was estimated by trypan blue exclusion, and the data were reported graphically. Each bar represents the mean (±SD) percentage viability from three separate experiments, and four individual counts were performed for each. **P < 0.05. Original magnification, ×100.

Figure 5.

 

UVB-induced apoptosis in cultured dysplastic epithelial cell. Annexin V- and PI-labeled cells were analyzed by flow cytometry, and scatterplots were generated. Early (EA, bottom right) and late (LA, top right) apoptosis was investigated after exposure to (A) 0, (B) 20, and (C) 50 mJ/cm².

Figure 6.

 

MMP and TIMP mRNA expression in UVB-exposed epithelial cells. DCECs (A, C, D) and NCECs (B) were exposed to UVB radiation, and total RNA was extracted and reverse transcribed and the products formed from PCR reactions were displayed in ethidium bromide-stained gels (A, B). (C, D) Data summarize semiquantitative results of three separate experiments. Bars represent change in MMPs (C) and TIMPs (D) compared with the control after they were normalized to GAPDH. Negative controls consisted of reactions that were not reverse transcribed (−RT).

Figure 7.

 

Multianalyte profiling of secreted MMPs. DCECs (A, C) and NCECs (B, D) were exposed to various doses of UVB radiation and supernatants analyzed by multianalyte fluorokine profiling. MMP-1 (A, B), -2 (not shown), and -3 (B, D) were the only detectable proteases. Data represent the mean (±SD) enzyme levels after treatment. Note the differential dose-dependent response to UVB between the dysplastic compared with normal cells. *Statistical significance (P < 0.05) particularly at the noncytotoxic 20 mJ/cm² dose in DCECs. MMP-2 levels were unchanged in both cell types after undergoing the same irradiation protocol (data not shown).

Figure 8.

 

MMP-2 and TIMP production in irradiated diseased and normal epithelial cells. DCECs (A, C, D) and NCECs (B) were exposed to various stimuli (as indicated on the labels) the supernatants were collected and analyzed by gelatin zymography (A, B), reverse zymography (C), or ELISA (D). For zymography and reverse zymography (C), adjacent lanes were loaded with an appropriate protein ladder (MW) as well as conditioned medium from pterygium epithelial cells (C⁺², known to contain MMP-2) and human primary macrophage supernatants (C⁺⁹, known to secrete MMP-9). These gels are representative of three independent experiments. (B, arrows) An active MMP-2 species. (C) Gelatinolytic bands were scanned and semiquantitatively assessed. The same DCEC supernatants demonstrated little or no variation in the level of TIMP-1 and -2 (D). The bars in (E) represent mean (±SD) TIMP-1 concentrations from three independent experiments.

Figure 9.

 

Intracellular signaling pathway implicated in enhancing MMP production. The DCECs were preincubated in selective MAPK inhibitors and exposed to a single 20-mJ/cm² dose of UVB. Supernatants were collected after 72 hours and MMP-1 levels detected by immunoassay. The same supernatants were analyzed by gelatin zymography. Each bar represents the mean (± SD) MMP-1 concentration in three independent experiments.

Supplementary Materials

Supplementary Figure S1 - (PDF) 

Supplementary Figure S2 - (PDF) 

The authors thank Jeanie Chui for technical assistance with flow cytometry, statistical analysis and critical appraisal of the manuscript, and Roger Crouch (Anatomic Pathology, Prince of Wales Hospital, Sydney) for confirming the initial diagnosis of each case included in the study.