Changes in Expression by Trabecular Meshwork Cells in Response to Mechanical Stretching

Vasavi Vittal, Anastasia Rose, Kate E. Gregory, Mary J. Kelley, and Ted S. Acott

PURPOSE. Trabecular meshwork (TM) cells appear to sense to 5% of people exhibit pathologic elevations in IOP with changes in intraocular pressure (IOP) as mechanical stretching. subsequent optic nerve damage, even at advanced ages.1,2 We In response, they make homeostatic corrections in the aqueous have hypothesized that TM cells can adjust outflow resistance humor outflow resistance, partially by increasing extracellular over a timescale of hours to days by modulating trabecular ECM matrix (ECM) turnover initiated by the matrix metalloprotein- turnover and subsequent biosynthetic replacement.3–6 Manip- ases. To understand this homeostatic adjustment process fur- ulation of the trabecular activity of a family of ECM turnover ther, studies were conducted to evaluate changes in TM gene enzymes, the matrix metalloproteinases (MMPs), reversibly expression that occur in response to mechanical stretching. modulates outflow facility.7 Inhibition of the endogenous ECM METHODS. Porcine TM cells were subjected to sustained me- turnover, which is initiated by these MMPs, increases the chanical stretching, and RNA was isolated after 12, 24, or 48 outflow resistance. Therefore, ongoing ECM turnover must be hours. Changes in were evaluated with mi- necessary for homeostatic maintenance of the IOP. In addition, croarrays containing approximately 8000 cDNAs. Select mRNA laser trabeculoplasty, a common treatment for glaucoma, ap- changes were then compared by quantitative reverse transcrip- pears to owe its efficacy to producing relatively sustained tion–polymerase chain reaction (qRT-PCR). Western immuno- trabecular MMP elevations, particularly within the juxtacanal- blots were used to determine whether some of these changes icular region of the meshwork.8–10 were associated with changes in levels. Much of the aqueous humor outflow resistance appears to RESULTS. On the microarrays, 126 were significantly up- reside within the deepest portion of the TM.11 Consequently, regulated, and 29 genes were significantly downregulated at this resistance is functionally stretched like a diaphragm across one or more time points, according to very conservative sta- Schlemm’s canal. Alterations in the outflow resistance that are tistical and biological criteria. Of the genes that changed, sev- sufficient to affect IOP change the degree of stretching and eral ECM regulatory genes, cytoskeletal-regulatory genes, sig- distortion of the juxtacanalicular TM. A likely sensing mecha- nal-transduction genes, and stress-response genes were nism in TM cells, indicating that an increase or decrease in notable. These included several and matricellu- outflow resistance is needed, is that of alterations in mechan- lar ECM composed of common repetitive binding domains. The results of analysis of mRNA changes in more than ical tension or stretching. This sensing could be mediated and transduced by –ECM or similar types of interac- 20 selected genes by qRT-PCR supported the findings in the 12,13 microarray analysis. Western immunoblots of several proteins tions. We and others have found that trabecular cells can demonstrated protein level changes associated with changes in sense changes in IOP and mechanical stretching and produce 14–24 the level of mRNA. several distinctive responses. In support of our overall hypothesis of IOP , trabecular cells respond to CONCLUSIONS. The expression of a variety of TM genes is signif- icantly affected by mechanical stretching. These include sev- pressure elevations in perfused organ culture or to mechanical eral ECM proteins that contain multiple binding sites and may stretching in cell culture by increasing MMP-2 (gelatinase A) serve organizational roles in the TM. Several proteins that and MMP-14 (membrane-type-1-MMP) activity or levels, while could contribute to the homeostatic modification of aque- dramatically reducing levels of their primary tissue inhibitor, 14 ous humor outflow resistance are also upregulated or down- TIMP-2. Increased trabecular MMP activity results in reduced 7 regulated. (Invest Ophthalmol Vis Sci. 2005;46:2857–2868) outflow resistance and thus should restore IOP to normal DOI:10.1167/iovs.05-0075 levels. Therefore, the components of a self-contained trabecu- lar IOP homeostasis mechanism are present and functional. In he mechanisms that provide normal IOP homeostasis are perfused anterior segment organ culture, increased flow rates Tonly partially understood. A relatively effective homeo- produce initial elevations of IOP, which return to normal over static mechanism must exist, because only approximately 2% several days, even with sustained increases in perfusion rate.14,23 An important component of this putative IOP homeostasis From the Casey Eye Institute, Oregon Health and Science Univer- mechanism is the nature of the changes in the ECM that adjust sity, Portland, Oregon. the outflow resistance. The ECM turnover process is initiated Supported by National Institutes of Health Grants EY003279, by secretion/activation of the MMPs, which partially degrades EY008247, and EY010572 and by grants from the Glaucoma Research select ECM proteins. This phase is followed by the removal of Foundation (San Francisco, CA), Research to Prevent Blindness, and proteolytic ECM fragments, presumably facilitated by TM cell Alcon Labs (Fort Worth, TX). Submitted for publication January 21, 2005; revised April 11, endocytosis. New materials must then be synthesized by TM 2005; accepted April 22, 2005. cells to replace degraded ECM components. To adjust the Disclosure: V. Vittal, None; A. Rose, None; K.E. Gregory, None; outflow resistance, changes in the amount, composition, or M.J. Kelley, None; T.S. Acott, Alcon Labs (F) organization of this ECM are likely to be instituted. This pro- The publication costs of this article were defrayed in part by page cess must occur within the active pathway of aqueous humor charge payment. This article must therefore be marked ”advertise- outflow without allowing excessive structural disorganization. ment“ in accordance with 18 U.S.C. §1734 solely to indicate this fact. Corresponding author: Ted S. Acott, Casey Eye Institute (CERES), To begin unraveling the details of this complex process, mi- Oregon Health and Science University, 3375 SW Terwilliger, Portland, croarray analysis of TM gene expression changes occurring OR 97239-4197; [email protected]. after sustained mechanical stretching were conducted.

Investigative Ophthalmology & Visual Science, August 2005, Vol. 46, No. 8 Copyright © Association for Research in Vision and Ophthalmology 2857

Downloaded from iovs.arvojournals.org on 09/29/2021 2858 Vittal et al. IOVS, August 2005, Vol. 46, No. 8

MATERIALS AND METHODS The two types of OHSU SMC microarray chips that were used had duplicate separated spottings of approximately 5700 or 8400 human Porcine eyes were obtained from Carlton Packing (Carlton, OR); pro- cDNA clones per slide. The data presented herein are from the 8400 teinase inhibitor cocktail, phenylmethylsulfonyl fluoride (PMSF), di- chips, but very similar results were obtained earlier with the 5700 thiothreitol (DTT), and horseradish peroxidase-conjugated secondary chips. The libraries used were prepared by the IMAGE consortium32 antibodies from Sigma-Aldrich (St. Louis, MO); a DNA quantitation and distributed in sequence-verified form by Invitrogen (Carlsbad, CA). reagent (PicoGreen) from Molecular Probes (Eugene, OR); fibronectin Amplified PCR products were printed onto microarray slides (Ultra- antibody from Transduction Laboratories-BD Biosciences (San Diego, GAPS; Corning Costar, Corning, NY) using microarray printing pins CA); tenascin C antibody from Santa Cruz Biotechnology (Santa Cruz, (TeleChem 16 CMP-3) and an array printer (Cartesian PixSys 5500 XL; CA); Dulbecco’s modified Eagle’s medium (DMEM), antibiotics, and Genomic Solutions, Irvine, CA). Twenty-three plant genes (Arabidop- antimycotics from Invitrogen-Gibco (Grand Island, NY); fetal bovine sis thaliana) spotted twice per chip, were used for standard reference serum from HyClone (Logan, UT); chemiluminescence detection kits providing both a positive and a negative control. In addition, 176 spots (Super Signal) from Pierce (Rockford, IL); Falcon cell culture inserts were left blank on each chip to provide another type of negative (PET track-etched 3-␮m pore membranes in six-well format) from BD control. Scanned images were analyzed by computer (ImaGene; Bio- Biosciences (Franklin Lakes, NJ); and RNA extraction kits (RNeasy) Discovery, Marina del Rey, CA) to determine relative fluorescence from Ambion (Austin, TX). intensity for each label; to identify spot locations, shapes, sizes, and anomalies; and to determine spot-specific background intensity for TM Cell Culture, Mechanical Stretching, each label. At each of the three time points (12, 24, and 48 hours), Treatments, and Extractions three totally independent experiments were conducted. Each sample

5,25–27 from each experiment was analyzed on two separate identical micro- Porcine TM cells were cultured as previously described. By chips. On each chip, each of the 8400 genes was spotted twice in passages 3 to 5, cells were plated at a density of approximately 90% separate regions. Consequently, at each of the three time points, there confluence onto cell culture insert membranes in six-well culture 14,28 were 12 separate stretch and control data pairs for each gene. plates. When cells were densely confluent, 3 to 5 days later, they Microarray spot fluorescence intensities for control and stretch were placed in serum-free medium for 24 hours before and during samples with associated background values were then analyzed sepa- stretching experiments. To apply mechanical stretch/distortion, a rately at each of the three time points. To determine statistical signif- smooth, 5.25-mm glass bead was placed in the dish beneath the center icance, the data from each chip with the duplicate spot values were of the insert membrane. A weight was then applied to the lid of the 14,28 separated into half chips, with one stretch and one control value and plate, which forced the insert down onto the bead. This process their respective specific backgrounds for each of the approximately created a defined upward distortion of the membrane, which increased 8400 cDNAs. The stretch and control data from each half-chip were the surface area and produced mechanical stretching of the cells and then corrected for background and normalized via Lowess33 (Gene- their ECM. More than 20 different porcine cell lines, each comprising Spring software; Silicon Genetics, Redwood City, CA). These normal- cells from 10 to 20 eyes, were studied. DNA analysis (PicoGreen; ized fluorescence intensities were then combined into a single parallel Molecular Probes) was conducted in parallel wells in some studies, to data set for each time point, to identify those genes that achieved estimate cell density. Because the differences between wells were statistical significance using significance analysis of microarrays always less than 10%, the analysis was not conducted in all studies. At (SAM).34 Twelve normalized stretch and 12 normalized control fluo- the indicated times after initiation of stretching, medium was collected rescence intensities for each gene at each time point were thus used to Ϫ and stored in aliquots frozen at 20°C until use. For cellular or ECM determine which changes were statistically significant. The number of protein analysis, membranes were immediately rinsed with ice-cold permutations was set at 1000; the analysis was in a paired format; and phosphate-buffered saline and the cells lysed with a modified RIPA 29–31 the nearest number imputer was set to 10 in the SAM software (http:// buffer (2 mM EDTA, 2 mM EGTA, 1% NP-40, 0.5% sodium deoxy- www.stat.stanford.edu/ϳtibs/SAM/ provided in the public domain by cholate, 0.1% sodium dodecyl sulfate [SDS], 2 mM DTT, 1 mM PMSF, Stanford University, Stanford, CA). The false-discovery rate ⌬ parame- proteinase inhibitor cocktail, and 50 mM Tris [pH 7.5]) on ice. For RNA ter was then adjusted to produce a median value of less than 0.5 falsely analysis, the membranes were cut from the inserts after stretching and significant genes from the complete set. A separate evaluation was placed in denaturation buffer (RNeasy ; Ambion). They were then conducted by using two criteria to determine which gene changes vortexed and the lysate removed. The remaining steps were as de- achieved biological significance. To do this, all 12 of the stretch and scribed in the kit protocol, and total RNA was processed according the control fluorescence spot intensity pairs with their associated back- manufacturer’s instructions. The extracted RNA was subjected to grounds were combined into a single long data set. They were cor- DNase treatment with 20 units of RNase-free DNase (Promega, Madi- rected for background and then normalized via Lowess (GeneSpring; son, WI) for 20 minutes. This was followed by two phenol-chloroform Silicon Genetics), as though they were replicates on one large micro- extraction steps to obtain high-quality RNA with an A260 /A280 optical chip. These data were then filtered to select genes that exhibited at density ratio of 1.8 to 2.0. least a 1.5-fold average increase or at least a 50% average decrease. These were then further filtered to eliminate genes expressed at low Gene Chip Microarray levels both before and after mechanical stretching. Spots with average From each control or stretch sample, 9 ␮g of total purified RNA was normalized fluorescence intensities below 300 for both control and provided to the Oregon Health and Science University Spotted Microar- stretch, after the background was subtracted, were eliminated. This ray Core (OHSU SMC) for processing (http://www.ohsu.edu/gmsr/ lower expression threshold was selected because it was more than 2 smc/index.shtml). The stretch or the nonstretch control RNA was standard deviations above the average specific background on these labeled by reverse transcription using either fluorescein-modified or chips. The subset of genes that passed all three of these selection biotinylated nucleotides and an oligo-dT primer. These labeled probes criteria (i.e., exhibited both statistical and biological significance for at were mixed and hybridized to the spotted chips at 65°C for 16 hours least one time point) was retained. (M-series LifterSlips; Erie Scientific, Portsmouth, NH) in a deep well hybridization chamber (TeleChem, Sunnyvale, CA). Hybridized arrays Quantitative RT-PCR were then probed with Tyramide signal amplification (TSA) kits (PerkinElmer, Boston, MA), with horseradish peroxidase-conjugated Quantitative RT-PCR was performed (LightCycler and one-step Light- anti-fluorescein antibody or streptavidin. Slides were developed with cycler-RNA amplification Kit, SYBR Green I; Roche Diagnostics, India- Cy3- and Cy5-tyramine and scanned (ScanArray 4000 XL with napolis, IN).35 Primer pairs were designed in adjacent exons so that ScanArray Express software; Perkin-Elmer). amplification of potential traces of contaminating genomic DNA would

Downloaded from iovs.arvojournals.org on 09/29/2021 IOVS, August 2005, Vol. 46, No. 8 Trabecular Gene Expression with Stretch 2859

be easily identified. After the RT cycle (55°C, 10 minutes), 40 PCR cycles (94°C, 30 seconds; 50°C [primer dependent], 10 seconds; and 72°C, 14 seconds) were used. Fluorescence was acquired at the end of each extension step, and a melting curve was obtained at the end of each run. In a few cases in which multiple peaks were observed in the melting curve, the fluorescence was acquired in a separate step after each cycle at a temperature above the melting temperature of the additional nonspecific peak. After the PCR, products were analyzed on gels to verify band sizes and purity. In some cases, the product was also sequenced. Data were analyzed by the system software (LightCycler; Roche Diagnostics), by using a 10-point dilution standard curve and the crossing-point method to determine the relative template concentra- tion of samples. Because fibronectin exhibits alternative splicing, sev- eral additional primers were designed to determine which splice forms were expressed. Twenty genes were evaluated by qRT-PCR, including several that increased, several that decreased, and several that did not show any change on the microarrays at any time after stretching.

Western Immunoblots FIGURE 1. Typical R-I plot after Lowess normalization. All the 24-hour Western immunoblots, transferred electrophoretically from standard stretch and control data from the SMC8400 chip, which included three completely separate experiments with two chips for each experiment SDS-PAGE gels to nitrocellulose or PVDF membranes, were probed and with two spots per chip for each gene (approximately 192,000 with the indicated primary antibodies. Detection used the appropriate data points after individual local background subtractions), were sub- secondary antibodies with conjugated horseradish peroxidase and jected to Lowess normalization in the software (GeneSpring; Silicon chemiluminescence according to the manufacturer’s instructions (Su- Genetics, Redwood City, CA). The means of these normalized stretch perSignal; Pierce)5,14,31,36 Autoradiographs were scanned and relative and control data were then plotted, as indicated on the axes. band density analyzed37 with a densitometry program (UVP, Upland, CA). For both the qRT-PCR and the Western immunoblot data, Stu- dent’s t-test or Mann-Whitney ranked sum analysis was used to deter- Effects of Mechanical Stretch on Specific TM mine significance, when comparing treatment results. All experiments Gene Expression presented were repeated at least three times, and typical gels were selected for presentation. Twenty-seven transcripts that code for ECM molecules or for molecules integrally involved in ECM regulation or in were upregulated, and two were downregulated, RESULTS according to the stringent statistical and biological criteria described earlier (Table 1). Several were elevated at all three Normalization and Significance time points, but most showed significant increases at only one or two time points. Several of those that did not reach the When the microarray data sets were normalized by Lowess significance criteria were nonetheless clearly affected. As an after the backgrounds were subtracted, flat and well-centered 33 example, fibronectin 1 was elevated at 24 hours based on our ratio-intensity (R-I) curves were obtained. Figure 1 shows a statistical significance criteria, but only 1.33-fold, and thus is typical R-I curve for the SMC8400 chip and 24-hour data. At labeled NS (not significant) in Table 1 at this time point. each time point, a few hundred genes achieved statistically Transcript levels for 42 genes involved in cytoskeletal function 34 ϫ significant changes, when analyzed by SAM in each 12 12 or in a variety of other forms of cellular regulation were increased, ⌬ stretch–control data set. As the parameter decreased incre- and 9 were decreased at one or more of the three time points mentally, a lower plateau was reached at approximately a (Table 2). Transcript levels of 20 genes that are involved in median number of 0.5 falsely significant genes per data set. transcription or translation were increased, and levels of 13 genes Additional increments reduced the number of genes without were decreased at one or more time points (Table 3). Thirty-seven reducing the false-discovery rate. Therefore, the highest num- genes involved in cellular stress responses, cellular metabolism, or ber of genes that achieved this false-discovery rate was ac- other cellular processes were upregulated, and five were down- cepted. Further filtering of this group for biological signifi- regulated at one or more time points (Table 4). cance, in terms of expression level in the TM and magnitude of changes with mechanical stretch, gave 126 genes that achieved Effects of Mechanical Stretching on TM Genes a 1.5-fold or larger increase and 29 genes that achieved a 50% Associated with Primary Open-Angle Glaucoma or greater decrease at one or more of the three time points. Analysis of the results from the SMC5700 chips showed good We also looked specifically in the microarray data for changes agreement for the genes common to both arrays (data not in genes that have been identified as active in primary open- shown). The OHSU SMC microarray facility also conducted a angle glaucoma, (i.e., myocilin/TIGR, optineurin and detailed data analysis in which somewhat different specific WDR36)38–40 and for genes that reside in some of the chro- normalization and significance testing approaches were used. mosomal regions to which glaucoma loci have been mapped. Their lists of changed genes were very similar to those we Myocilin/TIGR (GLC1A) was not represented on this microar- obtained, as detailed earlier. ray, although it has been shown to be elevated in the TM early Filtering the data from all three time points, after subtrac- after mechanical stretch or elevated IOP.18,21,23 Optineurin tion of the local background and normalization via Lowess but (GLC1E) and WDR36 (GLC1G) were both represented on our with no other restrictions, simply requiring the mean expres- microarray. However, neither showed significant changes in sion levels to reach a lower threshold level of greater than 300 expression with mechanical stretch at any of the three time relative fluorescence units, produced a group of well over points. Although positive relative fluorescence was observed 3000 genes that are clearly expressed at significant levels by for both transcripts, neither was expressed at sufficiently high TM cells. levels in the TM under our culture conditions, with or without

Downloaded from iovs.arvojournals.org on 09/29/2021 2860 Vittal et al. IOVS, August 2005, Vol. 46, No. 8

TABLE 1. Genes Involved in ECM, ECM Regulation, or Cell Adhesion

x-Fold (Stretch/Control) Rel Flu 12 h 24 h 48 h Intensity Gene Name, Characteristics or Functions GenBank ID Location

7.165 NS* NS 815 NEL-like 2; NELL2; ECM protein NM_006159 12p13.11-q13.12 2.59 1.924 1.5 10535 Tenascin C (hexabrachion) NM_002160 9q33 1.68 1.964 2.417 14907 SPARC, secreted protein, acidic, cys-rich () NM_003118 5q31.3-q32 1.502 NS 1.5 17450 Fibronectin 1 NM_054034 2q34 1.621 NS NS 1222 Laminin, gamma 1 (formerly LAMB2) NM_002293 1q31 2.365 NS NS 8591 Collagen 1 alpha 2 NM_000089 7q22.1 NS 1.59 NS 10759 Collagen V alpha 1 NM_000093 9q34.3-q34.3 3.252 NS NS 4630 Collagen XIV alpha 1 NM_021110 8q23 1.769 NS NS 12409 4 NM_001897 15q23 1.62 1.718 1.646 14499 , small Leu-rich proteoglycan NM_002023 1q32 1.613 NS 1.5 15045 , small Leu-rich proteoglycan NM_001711 Xq28 NS 1.82 NS 10516 CD44, hyaluronan receptor NM_000610 11p13 NS 1.514 NS 9803 Syndecan 2, transmembrane proteoglycan NM_002998 8q22-q23 NS 1.57 NS 1517 -B2, EphB receptor ligand NM_004093 13q33 1.63 NS NS 737 2; hypothetical protein DRAPC1 NM_002457 11p15.5 2.684 NS NS 6881 Matrix metalloproteinase 15 NM_002428 16q13-q21 1.525 NS NS 1529 Matrix metalloproteinase 16 NM_005941 8q21 NS 2.139 NS 1658 Plasminogen activator, urokinase NM_002658 10q24 2.412 1.525 1.5 8797 Plasminogen activator inhibitor type 1B NM_006216 16q12.2 1.818 NS NS 5045 Plasminogen activator inhibitor type 1A NM_033011 8p12 9.527 NS NS 631 Growth differentiation factor 15; TGF beta/BMP family NM_004864 19p13.1-q13.2 3.719 NS 3.053 5076 Endoglin; TGF beta receptor complex protein NM_000118 9q33-q34.1 NS 1.925 NS 5323 Latent TGF beta binding protein 2 NM_000428 14q24 2.525 2.544 NS 9356 Matrix Gla protein; matrix calcification regulator NM_000900 12p13.1-p12.3 NS 1.75 NS 11553 CTGF, connective tissue NM_001901 6q23.1 2.26 NS NS 8074 Melanoma cell adhesion molecule, MUC 18, CD146 NM_006500 11q23.3 NS 1.69 NS 2443 VCAM1, vascular cell adhesion molecule 1 NM_001078 1p32-p31 NS 0.497 NS 30855 Mimecan, osteoglycin, osteoinductive proteoglycan NM_014057 9p22 0.488 NS NS 4130 Periostin, osteoblast specific factor 2, integrin ligand NM_006475 13q13.3

Rel Flu Intensity, relative fluorescence intensity. * NS indicates that it did not reach the cut-off level for both statistical and biological significance.

stretch to reach the very stringent expression level cutoff that The fibronectin microarray results were also compared with we used as part of our biological significance criteria. results from qRT-PCR and from Western immunoblots at 24 The chromosomal location of each gene that reached both hours after stretching (Fig. 3). The microarray values for fi- statistical and biological significance in our studies is included bronectin at 24 hours showed a 1.33-fold increase (Fig. 3C). in the last columns in Tables 1, 2, 3, and 4. Analysis of the Although this reached the level of high statistical significance actual chromosomal location of each possibly relevant gene by SAM, it did not achieve the cutoff filter level of a Ͼ1.5-fold relative to the markers used to map these glaucoma loci as increase and is listed as NS in Table 1. The results of the localized on the most recent human chromosomal maps, comparison of stretch and control cells evaluated by qRT-PCR showed that most of the changed genes in Tables 1, 2, 3, and (Fig. 3D) and by Western immunoblot (Figs. 3A, 3B) show 4 are not actually within any of the mapped glaucoma loci. A similar or larger increases in mRNA and protein levels. summary of changed genes that are located within several Several other sets of fibronectin PCR primers were used to mapped loci is shown in Table 5. None were within the look for possible changes in mRNA splicing with stretch (Fig. GLC1B, GLC1C, GLC1D, GLC1F, or GLC1G loci.40–44 Several 4). One pair of primers, one in the 12th and one in the 14th changed genes were identified in some of the other loci that fibronectin type III domains, were used to look for the pres- 45–47 have been mapped but not named, as indicated in Table 5. ence of the 13th type III domain, EIIIA (Fig. 4). Only the smaller (337-bp) PCR product in which this domain was Alternative Analyses of Genes Affected by spliced out was detectable in stretched (S) and in nonstretched Mechanical Stretch in Microarrays control (C) TM cells. A primer pair was made in the seventh Quantitative RT-PCR was used to evaluate mRNA levels at all and ninth fibronectin type III domains to look at splicing of the three time points of 20 of the genes that were elevated, de- eighth type III domain, EIIIB. Although at least 90% of the creased, or unaffected by stretch in the microarray analysis. transcripts had EIIIB spliced out and gave the 406-bp PCR Several examples (i.e., metallothionein 1R, tenascin C, MMP- product, approximately 10% of the transcripts gave the 678-bp 15, SPARC and mimecan transcript levels measured by microar- PCR product with the EIIIB domain spliced in. This ratio did ray and qRT-PCR) are compared at select times after stretching not change significantly with stretching. In the fibronectin in Figure 2. Some of the other genes for which mRNA levels variable region within the 17th type III domain, often referred were compared by qRT-PCR, included MMP-2, MMP-14, and to as the connecting segment (IIICS) domain, both variant (V)1 TIMP-2, which were not changed (data not shown). These also and V3 are often alternatively spliced. We made a pair of showed good agreement between the two methods. Although primers in the 16th type III domain matched with one in the the exact changes (x-fold) were seldom identical when the V2 domain to look for the V1 domain, and a pair of primers in microarray and qRT-PCR were compared, all showed similar the V2 domain matched to one in the 18th type III domain to responses to stretching by both methods. look for the V3 domain. The V2 domain is present in all the

Downloaded from iovs.arvojournals.org on 09/29/2021 IOVS, August 2005, Vol. 46, No. 8 Trabecular Gene Expression with Stretch 2861

TABLE 2. Genes Involved in the Cytoskeleton or in Cytoskeletal, and Other Cellular Regulation

x-Fold (Stretch/Control) Rel Flu 12 h 24 h 48 h Intensity Gene Name, Characteristics, or Functions GenBank ID Location

52.76 48.1 NS* 562 Metastasis-associated 1; histone deacylase complex NM_004689 14q32.3 2.688 4.807 1.69 4233 Dynamin family of large GTP-binding/GTPase NM_002462 21q22.3 2.025 1.806 NS 8244 Transgelin 2, SM22, calponin cytoskeletal domain NM_003564 1q21-q25 1.646 NS NS 13862 Transgelin 3, neuronal calponin, and SM22-like with EF-hand NM_013259 3q13.2 NS 1.818 NS 2973 Keratin 10 NM_000421 17q21 3.87 4.01 1.7 4828 Vimentin, intermediate filament protein NM_003380 10p13 2.533 1.813 1.5 9945 Tubulin, alpha 1 NM_006000 2q35 1.835 NS NS 13397 Beta 5 tubulin NM_178014 6p21.33 1.61 NS NS 21937 Actin, gamma 2, smooth muscle, enteric NM_001615 2p13.1 NS 1.556 NS 1438 BCR-related gene, RhoGAP/RhoGEP, Rho/Rac activator NM_021962 17p13.3 NS 1.69 NS 22000 RabGAP/TBC, EVI5, ecotropic viral integration site 5 NM_005665 1p22 NS 1.621 NS 4891 Rho/rac guanine exchange factor (GEF) 2 NM_004723 1q21-q22 1.663 NS NS 8080 RAB6A, Ras family GTPase, retrograde Golgi to ER NM_002869 11q13.3 1.951 NS NS 3775 Rho GDP dissociation inhibitor (GDI) beta NM_001175 12p12.3 NS 1.554 NS 3510 G1 to S phase transition 2, peptide chain release factor NM_018094 Xp11.23-p11.21 NS 1.532 NS 6893 FER1L3, myoferlin, Ca-mediated membrane fusion NM_013451 10q24 NS 1.682 NS 4146 Pro-rich coiled-coil, enabled homolog (Drosophila) NM_018212 1q42.12 5.45 7.437 2.26 1751 Dihydrodiol dehydrogenase 1: prostaglandin redox NM_001353 10p15-p14 5.4 NS NS 2216 Sjo¨gren syndrome antigen B (autoantigen La) NM_003142 2q31.1 5.07 NS NS 1311 Chemokine (C-C motif) ligand 2 (MCP1) NM_002982 17q11.2-q21.1 4.37 11.55 NS 5484 Chemokine (C-C motif) ligand 7 (MCP3) NM_006273 17q11.2-q12 3.714 6.243 NS 6219 Chemokine (C-C motif) ligand 13 (MCP4) NM_005408 17q11.2 3.228 1.67 NS 2557 Chemokine (C-C motif) ligand 11b (eotaxin-1) NM_002986 17q21.1-q21.2 NS 2.126 NS 1594 -like 4, epithelial cell apoptosis inhibitor NM_016109 19p13.2 NS 1.709 NS 3761 TNF alpha-induced protein 2 NM_006291 14q32.32 2.812 NS NS 6781 Receptor type protein tyrosine phosphatase-N NM_002846 2q35 NS 1.556 NS 931 Lyn, Src Tyr kinase family; v-yes-1 Yamaguchi related NM_002350 8q13 4.925 NS NS 4969 Tyrosylprotein sulfotransferase 2 NM_003595 22q12.1 1.867 1.67 NS 5305 SH2-B (phospho Tyr; PH domain, SH3/SH2 adaptor) NM_015503 16p11.2 1.844 NS NS 2508 PDGF receptor alpha NM_006206 4q12 2.626 NS NS 8489 EGF receptor related, EGF activating proteinase NM_022450 16p13.3 2.488 NS NS 6934 -like growth factor-binding protein 2 NM_000597 2q33-q34 2.157 NS NS 4976 receptor (tyrosine phosphorylated) NM_000121 19p13.3-p13.2 1.676 NS NS 272 Hypothetical protein with phospho-Tyr binding domain NM_017933 2q36.3 2.015 NS NS 350 dsRNA binding protein with WW pro-rich bind domain NM_022720 22q11.2 1.91 2.846 1.5 13799 Immediate early response 3, stretch NF kappa B like NM_003897 6p21.33 3.259 1.847 NS 2076 Sterol carrier protein 2, PI transfer and distribution NM_002979 1p32 1.793 NS NS 1291 Proteasome (macropain) activator subunit 2 (PA28 beta) NM_002818 14q11.2 1.581 1.594 NS 15831 Cathepsin B, lysosomal Cys proteinase NM_001908 8p22 1.592 NS NS 15996 Cathepsin D, lysosomal aspartyl proteinase NM_001909 11p15.5 1.58 NS NS 722 Prostate proliferative shutoff associated protein NM_015032 13q12.3 1.516 NS 0.54 2572 Pregnancy-associated protein A, IGFBP4 Zn proteinase NM_002581 9q33.2 NS 0.47 NS 359 Rho-GTPase activating protein 10 NM_020824 10p12.1 NS 0.486 NS 2134 SOCS box-containing WD repeat protein SWiP-1 NM_015626 17q11.1 NS 0.468 NS 474 Chemokine (C-X-C motif) ligand 12 NM_199168 10q11.1 NS 0.413 0.41 7373 14-3-3 zeta/delta, PKC and monooxygenase activator NM_003406 8q23.1 0.481 NS NS 13332 ANK, inorganic pyrophosphate transport regulator NM_054027 5p15.1 0.468 NS NS 1245 NGF receptor (TNFRSF16) associated protein 1, adaptor NM_014380 Xq22.2 0.467 NS NS 1295 Replication protein A2, DNA damage repair and tumor NM_002946 1p35 suppressor 0.426 NS NS 15390 Platelet-derived -like (extracellular) NM_006207 8p22-p21.3

Abbreviations are as defined in Table 1.

known splice variants containing either of the other variable immunoblots on several other proteins (data not shown), also domains. Stretched cells expressed exclusively a splice variant showed expression patterns compatible with the microarray that included V1 and V3, giving, respectively, the 294- and the data. 264-bp PCR products. Nonstretched control TM cells, how- ever, expressed approximately 70% of transcripts containing V1, V2, and V3 and 30% containing V2 but without V1 and DISCUSSION V3—that is, they gave PCR products of only 219 and 171 bp, respectively. When splice variants were evaluated at the other Only 155 of approximately 8000 genes met our very stringent two time points (i.e., at 12 and 48 hours; data not shown), the biological and statistical significance criteria for change in same patterns noted at 24 hours were observed. mRNA levels with stretching at one or more time points (Ta- Western immunoblots at 24 and 48 hours showed tenascin bles 1, 2, 3, 4). This number is likely to be an underestimate of C increases (Fig. 5) that are compatible with the microarray the total number of affected genes, but the confidence that data (Table 1) and with the qRT-PCR data (Fig. 2). Western these genes changed expression levels in a biologically signif-

Downloaded from iovs.arvojournals.org on 09/29/2021 2862 Vittal et al. IOVS, August 2005, Vol. 46, No. 8

TABLE 3. Genes Involved in Transcription or Translation Regulation Affected by Mechanical Stretch

x-Fold (Stretch/Control) Rel Flu 12 h 24 h 48 h Intensity Gene Name, Characteristics, or Functions GenBank ID Location

2.41 2.152 NS* 5724 Jun B transcription factor NM_002229 19p13.2 NS 2.118 NS 2541 Interferon regulatory factor 1, transcriptional repressor NM_002198 5q31.1 NS 1.816 NS 1521 NF kappa B, 1 (p105) NM_003998 4q24 NS 1.598 NS 19320 Zinc finger protein 26 (KOX 20) transcriptional regulator NM_019591 12q24.33 NS 1.597 NS 15296 Zinc finger protein 205, transcriptional regulator NM_003456 16p13.3 NS 1.521 NS 16161 Zinc finger protein 585B (41-like) transcription regulator NM_152279 19q13.12 1.735 NS NS 14076 Zinc finger protein 36, C3H type transcription regulator NM_003407 19q13.1 1.703 NS NS 13487 Zinc finger protein 184, Kruppel-like transcription factor NM_007149 6p21.3 5.068 NS NS 517 Nuclear factor (erythroid-derived 2); NF-E2 NM_006163 12q13 2.498 NS NS 1141 Interferon-stimulated ubiquitin-like transcription factor NM_005101 1p36.33 2.23 NS NS 924 Chromobox homolog 5 (CBX5), HP1 alpha homolog NM_012117 12q13.13 2.061 NS NS 752 HOX 12, pre-B-cell leukemia transcription factor 2 NM_002586 6p21.3 2.017 2.268 1.511 11353 Forkhead box F1 (FOXF1), FREAC-1 transcription factor NM_001451 16q24 1.773 NS NS 15716 C-terminal binding protein, transcriptional corepressor NM_001328 4p16 1.769 NS NS 332 Sterol regulatory element-binding transcription factor 2 NM_004599 22q13 1.67 1.78 NS 1429 Pirin, iron-binding nuclear NFI/CTF1 interacting protein NM_003662 Xp22.2 1.826 NS NS 14672 mRNA cleavage and polyadenylation-specific factor 6 NM_007007 12q15 1.592 NS NS 3397 KIAA1404 protein, helicase family, transcription factor NM_021035 20q13.13 1.877 1.72 1.5 13638 Ubiquitin and ribosomal protein S27a, c-term extension NM_002945 2p16 1.545 NS NS 14877 Eukaryotic translation elongation factor 2 NM_001961 19pter-p12 NS 0.491 NS 1425 OVN6-2, zinc finger protein 24 (KOX 17) NM_006965 18q12 NS 0.448 NS 9374 PBX1, homeobox gene, transcription factor NM_002585 1q23 NS 0.473 NS 7706 Sp3 transcription factor NM_003111 2q31 NS 0.442 NS 1631 Histone H3 methyltransferase, SET domain protein 7 (SET7) NM_030648 4q28 0.451 NS NS 1291 Early growth response 1 (ERG1) transcription factor NM_001964 5q31.1 0.425 NS 0.37 2069 Transcription factor 4, beta catenin interacting protein NM_003199 18q21.1 NS NS 0.5 3104 RNA polymerase II A NM_000937 17p13.1 NS 0.451 0.411 782 Cyclophilin A, peptidyl-prolyl cis-trans isomerase NM_021130 7p13-p11.2 NS 0.38 NS 492 EIF-3, subunit 6 interacting protein NM_016091 22q NS 0.289 NS 1416 EIF-2B NM_003908 20pter-p12 NS 0.327 0.37 3617 CGI-90 protein NM_016033 8q21.3 0.5 NS NS 3511 Musashi homolog 2, heterogeneous nuclear RNP NM_138962 1q23.2 0.493 NS NS 9157 Transcript release factor for paused polymerase I NM_012232 17q21.2

Abbreviations are as defined in Table 1.

icant manner is high. Our selection criteria, requiring an in- The genes that were changed by stretching displayed varied crease of at least 1.5-fold or a decrease of 50%, seems to temporal patterns, with some changing at only one time point provide a reasonable compromise between including changes and others at two or all three points. One gene, IGF-binding that are not really of biological significance and excluding protein 4 proteinase, was elevated 1.5-fold at 12 hours, not changes that are. Requiring either the control or the stretched significantly changed at 24 hours, and significantly decreased cellular expression levels to be several standard deviations (to 0.54-fold) at 48 hours. This temporal variation in gene above the subtracted background also seems reasonable, but is expression is not surprising and suggests that the TM cell somewhat arbitrary. The biological relevance of the absolute response to stretching is both complex and intricately coordi- abundance of a transcript, as roughly approximated by the nated. The distribution of the changes suggests a bias favoring relative fluorescence levels on the microarray and of a given only a few gene functional groups. Genes involved in ECM, x-fold change in mRNA levels is most certainly very gene cytoskeleton, or stress responses or in their regulation com- specific and highly dependent on the specific cellular context. prised a disproportionately large fraction of those we identi- Additional variables like translational efficiency, protein half- fied. The genes in this microarray are biased toward genes with life, posttranslational modifications, and the functional and known functions and include a wide representation of genes biological efficiency of each protein molecule further compli- involved in most major biological processes. A fairly predict- cate selection of these criteria. able and significant number of genes involved in transcription, The 8000 genes on this microarray represent approximately transcriptional regulation, translation, or overall cellular regu- 25% of the . Although it seems possible that a lation were significantly affected by stretching. Perhaps not too few of the porcine probe sequences do not hybridize with high surprisingly, few genes involved in apoptosis, cellular prolifer- affinity to the human cDNAs spotted on the microarrays, this ation, most aspects of cellular metabolism, or most other major number is probably small. Because human and porcine se- cellular functions were affected. Stretching is presumably a quences are generally similar, microarray hybridization condi- normal homeostatic adjustment signal requiring only a moder- tions are of only moderate stringency, and both the labeled ate and focused response from these cells. porcine probes and the spotted human cDNAs are several In the current literature, microarray and cDNA library stud- hundred base pairs in length, lack of cross-species hybridiza- ies of human TM cells subjected to elevated pressures in tion should be minimal in these studies. This selected subset of perfused organ culture are somewhat comparable to our re- 126 increased and 29 decreased mRNAs should provide useful sults.22,24,48 The TM cells in these studies, which mostly in- insights into this homeostatic process. volved experimental conditions different from ours, were

Downloaded from iovs.arvojournals.org on 09/29/2021 IOVS, August 2005, Vol. 46, No. 8 Trabecular Gene Expression with Stretch 2863

TABLE 4. Genes Involved in Stress, Metabolism and Other Processes That Are Affected by Mechanical Stretching

x-Fold (Stretch/Control) Rel Flu 12 h 24 h 48 h Intensity Gene Name, Characteristics, or Functions GenBank ID Location

114.23 82 48.51 354 Metallothionein 1G NM_005950 16q13 106.98 80.9 41.498 361 Metallothionein 1R, F, (1L, II) AF348998 16q12.2 98.44 90.25 22.991 408 Metallothionein 1H NM_005951 16q13 92.14 75.93 31.466 439 Metallothionein 1X (1L) NM_005952 16q13 89.77 77.95 40.699 511 Metallothionein 1B NM_005947 16q13 3.341 NS* NS 518 Superoxide dismutase 2, mitochondrial NM_000636 6q25.3 2.579 NS NS 825 Cytochrome P4501B1 (CYP1B1; GLC3A) NM_000104 2p21 NS 2.535 NS 3813 Glutathione peroxidase 2 (gastrointestinal) NM_002083 14q24.1 NS 1.615 NS 24808 Crystallin, alpha B NM_001885 11q22.3-q 1.743 NS NS 13894 Peroxiredoxin 2, mitochondrial protein NM_181738 19p13.2 NS 1.754 NS 20217 Legumain, Asp-specific Cys endopeptidase NM_005606 14q32.1 NS 1.739 NS 1806 ATP-dependent 26S proteosome regulatory subunit NM_152747 7p15.1 NS 1.721 1.66 4354 Ubiquitin ligase, atrogin-1, F-box atrophy protein 32 NM_058229 8q24.13 NS 1.634 NS 2836 BACE2, membrane-associated Asp proteinase NM_012105 21q22.3 4.0172 NS NS 312 Amino acid/polyamine transporter, family II NM_017728 17q25.1 2.93 2.881 NS 1589 Zinc transporter, solute carrier family 30 A1 NM_021194 1q32-q41 1.529 NS NS 1384 Potassium voltage-gated channel, shaker-related 1 NM_000217 12p13 NS NS 2 2583 Stearoyl-CoA desaturase (delta-9-desaturase) NM_005063 10q23-q24 NS NS 1.532 8780 SRY-box 5; protein-O-mannosyltransferase 1 NM_007171 9q34.1 NS 1.713 NS 338 Delta aminolevulinate synthase 1 NM_000688 3p21.1 NS 1.624 NS 2404 HesB-like domain; inorganic ion transport and metabolism NM_030940 9q21.33 NS 1.586 NS 3277 Mitochondrial proteolipid NM_004894 14q32.33 NS 1.548 NS 2834 Isopentenyl- or farnesyl-diphosphate delta isomerase NM_004508 10p15.3 2.33 1.616 NS 11655 Phosphogluconate dehydrogenase NM_002631 1p36.3-p36.13 2.3 1.601 NS 2715 Glucan (1,4-alpha-), branching enzyme 1 NM_000158 3p12.3 2.048 1.995 NS 1055 S-adenosylhomocysteine hydrolase NM_000687 20cen-q13.1 2.042 NS NS 12863 Aldolase C, fructose-bisphosphate NM_005165 17cen-q12 2.002 2.601 1.904 755 Stearoyl-CoA desaturase (delta-9-desaturase) NM_005063 10q23-q24 1.849 NS NS 12468 Lactate dehydrogenase A NM_005566 11p15.4 1.576 NS NS 1193 Phosphoribosylaminoimidazole carboxylase NM_006452 4pter-q21 1.374 NS NS 1156 CTP synthase NM_001905 1p34.1 1.765 NS 1.5 3259 Complement component 1q gamma NM_172369 1p36.11 9.849 NS NS 1587 Natural killer cell group 7 transmembrane protein NM_005601 19q13.41 5.635 NS NS 4370 PQ domain protein, probable electron transporter NM_025078 18q23 2.66 NS NS 1771 Timora, 10 transmembrane domain leu zipper protein XM_371891 7p14.3-p14.2 1.566 NS NS 8114 Pro-rich with poly Leu/Ile and Glu domains NM_014745 16q24.3 1.564 NS NS 923 Hypothetical protein RP4-622L5 NM_019118 1p36.11-p34.2 NS 0.484 NS 14188 Chaperonin-containing TCP1, subunit 4 (delta) NM_006430 2p15 0.474 NS NS 429 Transcription elongation factor A (SII)-like 4 NM_024863 Xq22.2 0.489 NS NS 635 Methylmalonyl CoA epimerase NM_032601 2p13.3 NS 0.413 NS 688 Ribosomal protein L11 NM_000975 1p36.1-p35 0.468 NS NS 2281 EST R99690 4q12

Abbreviations are as defined in Table 1.

probably responding to mechanical stretching that is function- were among the genes showing similar changes. Because of ally somewhat similar to the process we used. Matrix Gla the very complex structural organization of the cells and the protein, several metallothioneins, alpha B crystallin, alpha tu- ECM within the TM in vivo, the relative degree to which bulin, transgelin, chaperonin-containing TCP1, and periostin different TM cells are exposed to stretching and distortion with

TABLE 5. Stretch-Modified Genes and Mapped Primary Open-Angle Glaucoma Loci

Gene Location Markers GenBank ID or Name

GLC1A 1q NA Myocilin/TIGR (not on this microarray but increased dramatically) GLC1B 2cen-q13 D2S2161-D2S2264 None changed GLC1C 3q21-q24 D3S3637-D3S3694 None changed GLC1D 8q23 D8S1830-D8S592 None changed GLC1E 10p14 NA Optineurin (very low expression; 0.9, 0.8 & 1.2 fold at 12, 24 & 72 hr) GLC1F 7q35-q36 D7S2442-D7S483 None changed GLC1G 5q22-q23 D5S2084-D5S492 WDR36 (very low expression; 1.0, 1.48 & 1.65 fold at 12, 24 & 72 hr) Other 14q11-12 D14S261-D14S121 NM_002818; Other 9q22 D9S152-D9S1825 NM_002160; NM_002581; NM_030940 Other 20p12 D20S194-D20S911 NM_003908 Other 2q D2S2188-D2S2178 NM_003142; NM_003111 Other 10p D10S1477-D10S601 NM_003380; NM_020824

Downloaded from iovs.arvojournals.org on 09/29/2021 2864 Vittal et al. IOVS, August 2005, Vol. 46, No. 8

FIGURE 2. Comparison of typical mi- croarray and qRT-PCR analysis. Log base 2 plot of the ratio of stretch to control responses as analyzed by mi- croarray compared with qRT-PCR is shown for several genes at the indi- cated time points. The positions of the cutoff points used for the microarray studies (a decrease to 50% or a 1.5-fold increase) are shown by the vertical lines and are labeled above the lines for reference.

increasing IOP is difficult to predict. It seems probable that particular interest. In addition to the initiation of ECM turnover each TM cell will experience different degrees of stretching by the concerted action of MMP-2, TIMP2, and MMP-14,14,28 and distortion with elevated IOP. In our cell culture stretch subsequent biosynthetic replacement of the degraded ECM model, more uniformity is likely to be attained. However, components is likely to be an integral aspect of the homeo- different regions of the membranes in our model may also static IOP adjustment. This probably entails modifying the experience different degrees of stretch/distortion. We estimate composition or organization of the ECM to change the outflow that many of the cells in our model are experiencing degrees resistance and thus modulate IOP. One group of ECM genes and types of stretching and distortion that are similar to that of that are affected by stretching includes: NELL2, tenascin C, many of the cells in the TM in vivo. However, it is very clear SPARC, fibronectin, laminin ␥1 chain, and collagen XIV. A that additional detailed biophysical analysis of the TM is common feature of this group is that they are all modular needed to clarify this contention. matricellular ECM proteins composed of numerous repeat do- In terms of our working hypothesis of IOP homeostasis, the mains containing motifs, such as EGF-like, calcium-binding changes in ECM or ECM-related genes listed in Table 1 are of EGF-like, thrombospondin 1, fibronectin type III, von Wille-

FIGURE 3. Effects of mechanical stretching on fibronectin mRNA and protein levels. Cells were stretched for 24 hours and cellular protein or total RNA extracted and analyzed. (A) Relative band density of scans of Western immunoblots from three separate experiments. The mean Ϯ SD with paired t-test significance is shown. (B) A typical image of the Western immunoblot bands. (C) All 36 stretch and control data pairs from seven separate experiments at the 24-hour time point from all the SMC5700 and SMC8400 microarray chips were compared. The mean rel- ative spot fluorescence intensities with standard deviations and signifi- cance from paired t-test are shown. The microarray stretch/control ratio at this time point was ϳ1.3. (D) The results from 18 qRT-PCRs from two separate experiments with six differ- ent primer sets are shown with mean, standard deviations, and signif- icance from paired t-test, as indi- cated.

Downloaded from iovs.arvojournals.org on 09/29/2021 IOVS, August 2005, Vol. 46, No. 8 Trabecular Gene Expression with Stretch 2865

FIGURE 4. Effects of mechanical stretching on fibronectin mRNA splice variants. Cells were stretched for 24 hours and total RNA extracted and analyzed by RT-PCR, with several PCR primer sets. A diagram is shown with the domain structure of fibronectin 1 including the alternatively spliced exons EIIIB, EIIIA, and III CS containing variant regions V1, V2, and V3 as indicated. The position of PCR primers and full-length product sizes are indicated above the respective splice variants. The size of the exon variably spliced is shown below each exon. Agarose gel with ethidium bromide-stained PCR products from cells that were stretched (S) or controls (C) for 24 hours are shown, with markers as indicated. The position of the predicted bands, with (ϩ) and without (Ϫ) the indicated spliced domains, are shown next to each gel.

brand factor C, -binding or integrin-binding RGD mo- The increase in fibronectin levels and the splicing changes, tif.49–60 Domains of these types commonly serve as specific which increase the fraction of the protein that contains V1 and binding cassettes for protein–protein, protein–glycosaminogly- V3 domains, are intriguing. Perfusion of anterior segment or- can, or protein–cell interactions and function in ECM and gan cultures with the second heparin binding (Hep II) frag- tissue structure and organization. Their elevation during ment of fibronectin—the 14th to 16th type III domains as stretching is likely to be involved in maintaining and facilitating indicated in Figure 4—reversibly increases outflow facility.57 the tissue and ECM reorganization associated with the ECM In another tissue, it has been shown that alternative splicing of turnover process initiated by the MMPs. Because they bind the V1 and V3 domains affect several specific biological activ- (GAGs) and proteoglycans, which are ities of the adjacent Hep II fragment without affecting other thought to provide at least a portion of the outflow resis- activities.56 Fibronectin has an RGDS cell-binding site in the tance,3,4,61 they could also directly affect outflow by changing 11th type III domain that binds to ␣5␤1 (Fig. 4). An resistance component orientation. ␣4␤1 integrin-binding site is found in the distal portion of the Hep II region. In the alternatively spliced variable regions of IIICS, V1 contains an LDV and V3 contains an REDV, both of which are also ␣4␤1 integrin-binding motifs.62,63 Thus, TM cells switch from a mix of splice variants, with and without these two sites, to exclusively expressing the form that con- tains these two additional cell-binding motifs. In addition, the V2 domain of fibronectin contains a GAG-binding site, and these splice variants may impact binding to CD44 and synde- can 2 (both discussed later). Although we did not see changes in fibronectin EIIIA or EIIIB exon splicing with stretching, we detected very low levels of EIIIB in both stretched and control cells. A previous report did not detect any of either of these exons in serum-free control TM cells, although both were detected after TGF-␤2 or serum treatment.64 Of interest, laminin ␥1 was increased 1.5-fold at 12 hours, but ␣2, ␣4, ␣5, ␤1, and the laminin receptor 1 were totally unaffected at all three time points. The laminins are a central component of cellular basement membranes, acting as scaf- folds for the recruitment of other ECM components.65 Because laminins 1 to 15 are different heterotrimeric combinations of the ␣1to5,␤1to4,and␥1 to 3 subunits, an increase in ␥1 may FIGURE 5. Comparison of tenascin C protein levels after 24 and 48 reflect a partial shift from one laminin form to another,58 at a hours of stretching. Extracted protein was subjected to Western im- time when the ECM is undergoing remodeling and change. The munoblot analysis to determine tenascin C levels. The relative density ␥ of the tenascin C band on Western immunoblots is compared. Repli- 1-subunit of laminin is also involved in binding to nidogen and 66 cate numbers (n) from three separate experiments and significance (P) in polymer formation. Although the primary cell-binding sites levels from paired t-tests are shown. on laminin are in the laminin ␣-subunit, cell-binding sites have

Downloaded from iovs.arvojournals.org on 09/29/2021 2866 Vittal et al. IOVS, August 2005, Vol. 46, No. 8

been identified on the ␥1 subunit of laminin.67–69 Based on between 20- to more than 100-fold at one time point. Functions relative fluorescence intensities in our microarrays, the ␣2 of these genes in other tissues include heavy metal, particularly subunit was expressed at low to negligible levels, ␤1 only at zinc, detoxification90 and homeostasis.91 This suggests that TM moderately low levels; but ␣4 and ␣5, ␥1, and laminin receptor cells are responding to elevated or released zinc or to oxidative 1 were all expressed at relatively high levels by the TM. stress in association with the stretching.92 A zinc transporter A second group of ECM genes that changed expression with and several other stress genes are also induced. The several stretching includes several proteoglycans, some of which may chemokine ligand increases and the large metastasis-associated serve a direct role in outflow resistance.3,4,6,61 Chondroitin 1 increases are similarly intriguing and similarly enigmatic. sulfate , fibromodulin, biglycan, syndecan 2, and Although the number of genes with expression levels that the part-time proteoglycan and hyaluronan GAG receptor, are significantly changed by mechanical stretch is not large, the CD44, increased with stretching, whereas mimecan decreased. potential involvement of many of them in maintaining and Presumably, one likely way to adjust the outflow resistance in adjusting the TM’s ECM and putatively the outflow resistance is response to mechanical stretching would be to adjust the likely. Understanding this complex process that appears to be proteoglycan/GAG levels or composition in the TM’s ECM. responsible for maintaining IOP homeostasis clearly requires Although it is tempting to speculate on the molecular mode by additional studies focused on the function of these proteins in which increasing one or more of these proteoglycans or of outflow pathway biology. decreasing mimecan, could decrease the outflow resistance, more information is needed to understand the process. Changes in the GAG side chains are also probable modifiers of References the outflow resistance. N Engl J Med The increase in CD44, the hyaluronan receptor, is very 1. Quigley HA. Open-angle glaucoma. . 1993;328:1097– 1106. likely to be important in this process. A relationship between 2. Quigley HA. Number of people with glaucoma worldwide. Br J hyaluronan levels and the trabecular outflow resistance is gen- 70 Ophthalmol. 1996;80:389–393. erally accepted. Several recent studies have demonstrated 3. Acott TS, Wirtz MK. Biochemistry of aqueous outflow. In: Ritch R, the involvement of CD44 in TM cellular function and have Shields MB, Krupin T, eds. The Glaucomas. Vol. 1. St. Louis: provided hypotheses explaining relationships with the outflow Mosby; 1996:281–305. 71,72 resistance. CD44 is also integrally involved and interactive 4. Acott TS. Biochemistry of aqueous humor outflow. In: Kaufman 73–75 with MMP-2 and -14, both of which are upregulated at the PL, Mittag TW, eds. Textbook of Ophthalmology. Vol. 7. London: translational but not the transcriptional level in TM cells by Mosby; 1994:1.47–41.78. 14,28 mechanical stretching. CD44 exists in forms, with and 5. Alexander JP, Samples JR, Van Buskirk EM, Acott TS. Expression of without GAG side chains, and the form with chondroitin– matrix metalloproteinases and inhibitor by human trabecular dermatan sulfates binds to collagen XIV, which is also in- meshwork. Invest Ophthalmol Vis Sci. 1991;32:172–180. creased with stretching. 6. Acott TS. Trabecular extracellular matrix regulation. In: Drance Syndecan 2, a transmembrane heparan sulfate proteogly- SM, Van Buskirk EM, Neufeld AH, eds. Pharmacology of Glau- can,76,77 was upregulated at 24 hours (Table 1). It interacts coma. Baltimore: Williams & Wilkins; 1992:125–157. with ␣5␤1 integrin in cellular signaling and with the Hep II, V1, 7. Bradley JMB, Vranka JA, Colvis CM, et al. Effects of matrix metal- and V2 regions of fibronectin. This upregulation also affects loproteinase activity on outflow in perfused human organ culture. cytoskeletal organization in an ezrin-dependent step via Rho A Invest Ophthalmol Vis Sci. 1998;39:2649–2658. GTPase activation.78,79 Syndecan 2 is phosphorylated on two 8. Parshley DE, Bradley JMB, Fisk A, et al. Laser trabeculoplasty C-terminal tyrosines by the kinase domain of EphB2 receptors induces stromelysin expression by trabecular juxtacanalicular as a step in signaling to the cytoskeleton.77 Although EphB2 cells. Invest Ophthalmol Vis Sci. 1996;37:795–804. levels were not upregulated, its ligand, ephrin-B2, was upregu- 9. Parshley DE, Bradley JMB, Samples JR, Van Buskirk EM, Acott TS. Early changes in matrix metalloproteinases and inhibitors after in lated by stretching. Thus, several proteins that have important vivo laser treatment to the trabecular meshwork. Cur Eye Res. interactions with each other are changed, further increasing 1995;14:537–544. the relevance of this process. CD44 also contains a cytoplasmic 10. Bradley JMB, Anderssohn AM, Colvis CM, et al. Mediation of laser ezrin-binding domain, linking it to this same signaling system trabeculoplasty-induced matrix metalloproteinase expression by and to the Rho A-Rho kinase–mediated cytoskeletal regula- IL-1␤ and TNF␣. Invest Ophthalmol Vis Sci. 2000;41:422–430. 80–83 tion. Therefore, the syndecan 2, ephrin-B2, and CD44 11. Bill A, Maepea O. Mechanisms and routes of aqueous humor upregulation by stretching may be closely linked in maintain- drainage. In: Albert DM, Jakotiec FA, eds. Principles and Practices ing the TM’s ECM and may be involved in, among other events, of Ophthalmology. Philadelphia: WB Saunders; 1994:206–226. the Rho A–cytoskeletal modulatory changes, which also oc- 12. Acott TS. Receptor biology and glaucoma: integrins in the eye. In: curred with mechanical stretching of TM cells (Table 2). Anderson DR, Drance SM, eds. Encounters in Glaucoma Research The cytoskeletal changes observed (Table 2) involve com- 1: Receptor Biology and Glaucoma. Milan, Italy: Foglizza Editore; ponents of all three filament systems—microfilaments, micro- 1994:97–135. tubules, and intermediate filaments—which suggests signifi- 13. Zhou LL, Zhang S-R, Yue BYJT. Adhesion of human trabecular cant cross-talk between systems.84 Early cytoskeletal changes meshwork cells to extracellular matrix proteins: roles and distri- after mechanical stretching of TM cells have been reported,15 bution of integrin receptors. Invest Ophthalmol Vis Sci. 1996;37: and sustained responses and long-term differences can be ex- 104–113. pected. Agents that manipulate the TM cell cytoskeleton have 14. Bradley JMB, Kelley MJ, Zhu XH, Anderssohn AM, Alexander JP, been shown to affect outflow facility.85–89 Whether the cy- Acott TS. Effects of mechanical stretching on trabecular matrix metalloproteinases. Invest Ophthalmol Vis Sci. 2001;42:1505– toskeletal changes are a direct cause or an effect of the ECM 1513. changes and outflow resistance adjustments remains unclear. 15. Tumminia SJ, Mitton KP, Arora J, Zelenka P, Epstein PL, Russell P. Several other intriguing gene expression changes are nota- Mechanical stretch alters the actin cytoskeletal network and signal ble, such as increases in the chemokines, metallothionines, and transduction in human trabecular meshwork cells. Invest Ophthal- metastasis-associated protein 1. The magnitude of some of mol Vis Sci. 1998;39:1361–1371. these changes suggests important, if only partially understood, 16. WuDunn D. The effect of mechanical strain on matrix metallopro- regulatory processes. Several metallothionein mRNAs increase teinase production by bovine trabecular meshwork cells. Curr Eye from low or negligible levels to quite high levels, increasing Res. 2001;22:394–397.

Downloaded from iovs.arvojournals.org on 09/29/2021 IOVS, August 2005, Vol. 46, No. 8 Trabecular Gene Expression with Stretch 2867

17. Okada Y, Matsuo T, Ohtsuki H. Bovine trabecular cells produce 40. Monemi S, Spaeth G, Dasilva A, et al. Identification of a novel TIMP-1 and MMP-2 in response to mechanical stretching. Jpn J adult-onset primary open-angle glaucoma (POAG) gene on 5q22.1. Ophthalmol. 1998;42:90–94. Hum Mol Genet. 2005;14:725–733. 18. Sato Y, Matouo T, Ohtsuki H. A novel gene (oculomedin) induced 41. Stoilova D, Child A, Trifan OC, Crick RP, Coakes RL, Sarfarazi M. by mechanical stretching of human trabecular cells of the eye. Localization of a locus (GLC1B) for adult-onset primary open angle Biochem Biophys Res Commun. 1999;259:349–351. glaucoma to the 2cen-q13 region. Genomics. 1996;36:142–150. 19. Stamer WD, Roberts BC, Epstein DL. Hydraulic pressure stimulates 42. Wirtz MK, Samples JR, Kramer PL, et al. Mapping a gene for adenosine 3Ј,5Ј-cyclic monophosphate accumulation in endothe- adult-onset primary open-angle glaucoma to 3q. Am J lial cells from Schlemm’s canal. Invest Ophthalmol Vis Sci. 1999; Hum Genet. 1997;60:296–304. 40:1983–1988. 43. Trifan OC, Traboulsi EI, Stoilova D, et al. A third locus (GLC1D) for 20. Mitton KP, Tumminia SJ, Arora J, Zelenka P, Epstein DL, Russell P. adult-onset primary open-angle glaucoma maps to the 8q23 region. Transient loss of ␣B-crystallin: an early cellular response to me- Am J Ophthalmol. 1998;126:17–28. chanical stretch. Biochem Biophys Res Commun. 1997;235:69– 44. Wirtz MK, Samples JR, Rust K, et al. GLC1F, a new primary 73. open-angle glaucoma locus, maps to 7q35–q36. Arch Ophthalmol. 21. Tamm ER, Russell P, Epstein DL, Johnson DH, Piatigorsky J. Mod- 1999;117:237–241. ulation of myocillin: TIGR expression in human trabecular mesh- 45. Wiggs J, Allingham R, Hossain A, et al. Genome-wide scan for adult work. Invest Ophthalmol Vis Sci. 1999;40:2577–2582. onset primary open angle glaucoma. Hum Mol Genet. 2000;9: 22. Gonzalez P, Epstein DL, Borras T. Genes upregulated in the human 1109–1117. trabecular meshwork in response to elevated intraocular pressure. 46. Nemesure B, Jiao X, He Q, et al. A genome-wide scan for primary Invest Ophthalmol Vis Sci. 2000;41:352–361. open-angle glaucoma (POAG): the Barbados Family Study of Open- 23. Borras T, Rowlette LL, Tamm ER, Gottanka J, Epstein DL. Effects of Angle Glaucoma. Hum Genet. 2003;112:600–609. elevated intraocular pressure on outflow facility and TIGR/MYOC 47. Wiggs JL, Lynch S, Ynagi G, et al. A genomewide scan identifies expression in perfused human anterior segments. Invest Ophthal- novel early-onset primary open-angle glaucoma loci on 9q22 and mol Vis Sci. 2002;43:33–40. 20p12. Am J Hum Genet. 2004;74:1314–1320. 24. Borras T. Gene expression in the trabecular meshwork and the 48. Gonzalez P, Zigler JS Jr, Epstein DL, Borras T. Identification and influence of intraocular pressure. Prog Retin Eye Res. 2003;22: isolation of differentially expressed genes from very small tissue 435–463. samples. BioTechniques. 1999;26:884–886, 888–892. 25. Polansky JR, Wood IS, Maglio MT, Alvarado JA. Trabecular mesh- 49. Kuroda S, Oyasu M, Kawakami M, et al. Biochemical characteriza- work cell culture in glaucoma research: evaluation of biological tion and expression analysis of neural thrombospondin-1-like pro- activity and structural properties of human trabecular cells in vitro. teins NELL1 and NELL2. Biochem Biophys Res Commun. 1999; Ophthalmology. 1984;91:580–595. 265:79–86. 26. Polansky JR, Weinreb R, Alvarado JA. Studies on human trabecular 50. Ghert MA, Qi WN, Erickson HP, Block JA, Scully SP. Tenascin-C cells propagated in vitro. Vision Res. 1981;21:155–160. splice variant adhesive/anti-adhesive effects on chondrosarcoma 27. Stamer WD, Seftor RE, Williams SK, Samaha HA, Snyder RW. cell attachment to fibronectin. Cell Struct Funct. 2001;26:179– Isolation and culture of human trabecular meshwork cells by 187. extracellular matrix digestion. Curr Eye Res. 1995;14:611–617. 51. Joester A, Faissner A. The structure and function of tenascins in the 28. Bradley JM, Kelley MJ, Rose A, Acott TS. Signaling pathways used nervous system. Matrix Biol. 2001;20:13–22. in trabecular matrix metalloproteinase response to mechanical 52. Yan Q, Sage EH. SPARC, a matricellular with impor- stretch. Invest Ophthalmol Vis Sci. 2003;44:5174–5181. tant biological functions. J Histochem Cytochem. 1999;47:1495– 29. Sambrook J, Fritsch EF, Maniatis T. Molecular Cloning: A Labora- 1506. tory Manual. 2nd ed. Cold Spring Harbor, NY: Cold Spring Harbor 53. Rhee DJ, Fariss RN, Brekken R, Sage EH, Russell P. The matricel- Laboratory Press; 1989. lular protein SPARC is expressed in human trabecular meshwork. 30. Harlow E, Lane D. Antibodies: A Laboratory Manual. Cold Spring Exp Eye Res. 2003;77:601–607. Harbor, NY: Cold Spring Harbor Laboratory Press; 1988. 54. Bornstein P, Sage EH. Matricellular proteins: extracellular modula- 31. Alexander JP, Acott TS. Involvement of protein kinase C in TNF␣ tors of cell function. Curr Opin Cell Biol. 2002;14:608–616. regulation of trabecular matrix metalloproteinases and TIMPs. In- 55. Hohenester E, Engel J. Domain structure and organisation in ex- vest Ophthalmol Vis Sci. 2001;42:2831–2838. tracellular matrix proteins. Matrix Biol. 2002;21:115–128. 32. Lennon G, Auffray C, Polymeropoulos M, Soares MB. The I.M- 56. Santas AJ, Peterson JA, Halbleib JL, Craig SE, Humphries MJ, Peters .A.G.E. Consortium: an integrated molecular analysis of genomes DM. Alternative splicing of the IIICS domain in fibronectin governs and their expression. Genomics. 1996;33:151–152. the role of the heparin II domain in fibrillogenesis and cell spread- 33. Quackenbush J. Microarray data normalization and transformation. ing. J Biol Chem. 2002;277:13650–13658. Nat Genet. 2002;32(suppl 2):496–501. 57. Santas AJ, Bahler C, Peterson JA, et al. Effect of heparin II domain 34. Tusher VG, Tibshirani R, Chu G. Significance analysis of microar- of fibronectin on aqueous outflow in cultured anterior segments of rays applied to the ionizing radiation response. Proc Natl Acad Sci human eyes. Invest Ophthalmol Vis Sci. 2003;44:4796–4804. USA. 2001;98:5116–5121. 58. Ekblom P, Lonai P, Talts JF. Expression and biological role of 35. Rasmussen R. Quantification on the LightCycler. In: Meuer S, laminin-1. Matrix Biol. 2003;22:35–47. Wittwer C, Nakagawara K-I, eds. Rapid Cycle Real-Time PCR: 59. Ehnis T, Dieterich W, Bauer M, Lampe B, Schuppan D. A chon- Methods and Applications. New York: Springer; 2001:21–34. droitin/dermatan sulfate form of CD44 is a receptor for collagen 36. Alexander JP, Bradley JMB, Gabourel JD, Acott TS. Expression of XIV (undulin). Exp Cell Res. 1996;229:388–397. matrix metalloproteinases and inhibitor by human retinal pigment 60. Gerecke DR, Foley JW, Castagnola P, et al. Type XIV collagen is epithelium. Invest Ophthalmol Vis Sci. 1990;31:2520–2528. encoded by alternative transcripts with distinct 5Ј regions and is a 37. Samples JR, Alexander JP, Acott TS. Regulation of the levels of multidomain protein with homologies to von Willebrand’s factor, human trabecular matrix metalloproteinases and inhibitor by in- fibronectin, and other matrix proteins. J Biol Chem. 1993;268: terleukin-1 and . Invest Ophthalmol Vis Sci. 1993; 12177–12184. 34:3386–3395. 61. Yue BYJT. The extracellular matrix and its modulation in the 38. Stone EM, Fingert JH, Alward WLM, et al. Identification of a gene trabecular meshwork. Surv Ophthalmol. 1996;40:379–390. that causes primary open angle glaucoma. Science. 1997;275:668– 62. Mostafavi-Pour Z, Askari JA, Whittard JD, Humphries MJ. Identifi- 670. cation of a novel heparin-binding site in the alternatively spliced 39. Rezaie T, Child A, Hitchings R, et al. Adult-onset primary open- IIICS region of fibronectin: roles of integrins and proteoglycans in angle glaucoma caused by mutations in optineurin. Science. 2002; cell adhesion to fibronectin splice variants. Matrix Biol. 2001;20: 295:1077–1079. 63–73.

Downloaded from iovs.arvojournals.org on 09/29/2021 2868 Vittal et al. IOVS, August 2005, Vol. 46, No. 8

63. Yamada KM. Fibronectin and other cell interactive . 78. Yoneda A, Couchman JR. Regulation of cytoskeletal organization In: Hay ED, ed. Cell Biology of Extracellular Matrix. New York: by syndecan transmembrane proteoglycans. Matrix Biol. 2003;22: Plenum: 1991;111–146. 25–33. 64. Li J, Tripathi BJ, Tripathi RC. Modulation of pre-mRNA splicing and 79. Granes F, Berndt C, Roy C, Mangeat P, Reina M, Vilaro S. Identifi- protein production of fibronectin by TGF-beta2 in porcine trabec- cation of a novel ezrin-binding site in syndecan-2 cytoplasmic ular cells. Invest Ophthalmol Vis Sci. 2000;41:3437–3443. domain. FEBS Lett. 2003;547:212–216. 65. Sasaki T, Fassler R, Hohenester E. Laminin: the crux of basement 80. Ponta H, Sherman L, Herrlich PA. CD44: from adhesion molecules membrane assembly. J Cell Biol. 2004;164:959–963. to signalling regulators. Nat Rev Mol Cell Biol. 2003;4:33–45. 66. Kadoya Y, Salmivirta K, Talts JF, et al. Importance of nidogen 81. Ponta H, Wainwright D, Herrlich P. The CD44 protein family. Int J Biochem Cell Biol binding to laminin gamma1 for branching epithelial morphogene- . 1998;30:299–305. sis of the submandibular gland. Development. 1997;124:683–691. 82. Hirao M, Sato N, Kondo T, et al. Regulation mechanism of ERM (ezrin/radixin/moesin) protein/plasma membrane association: 67. Kuratomi Y, Nomizu M, Tanaka K, et al. Laminin gamma 1 chain possible involvement of phosphatidylinositol turnover and Rho- peptide, C-16 (KAFDITYVRLKF), promotes migration, MMP-9 se- dependent signaling pathway. J Cell Biol. 1996;135:37–51. cretion, and pulmonary metastasis of B16–F10 mouse melanoma 83. Yonemura S, Hirao M, Doi Y, et al. Ezrin/radixin/moesin (ERM) cells. Br J Cancer. 2002;86:1169–1173. proteins bind to a positively charged amino acid cluster in the 68. Nomizu M, Kuratomi Y, Song SY, et al. Identification of cell juxta-membrane cytoplasmic domain of CD44, CD43, and ICAM-2. binding sequences in mouse laminin gamma1 chain by systematic J Cell Biol. 1998;140:885–895. peptide screening. J Biol Chem. 1997;272:32198–32205. 84. Chang L, Goldman RD. Intermediate filaments mediate cytoskeletal 69. Ponce ML, Nomizu M, Delgado MC, et al. Identification of endo- crosstalk. Nat Rev Mol Cell Biol. 2004;5:601–613. thelial cell binding sites on the laminin gamma 1 chain. Circ Res. 85. Sabanay I, Tian B, Gabelt BT, Geiger B, Kaufman PL. Functional and 1999;84:688–694. structural reversibility of H-7 effects on the conventional aqueous 70. Knepper PA, Farbman AI, Telser AG. Exogenous hyaluronidases outflow pathway in monkeys. Exp Eye Res. 2004;78:137–150. and degradation of in the rabbit eye. Invest Oph- 86. Rao P, Deng P-F, Kumar J, Epstein D. Modulation of aqueous thalmol Vis Sci. 1984;25:286–293. humor outflow facility by the Rho kinase-specific inhibitor 71. Knepper PA, Mayanil CS, Goossens W, et al. Aqueous humor in Y-27632. Invest Ophthalmol Vis Sci. 2001;42:1029–1037. primary open-angle glaucoma contains an increased level of 87. Vittitow JL, Garg R, Rowlette LL, Epstein DL, O’Brien ET, Borras T. CD44S. Invest Ophthalmol Vis Sci. 2002;43:133–139. Gene transfer of dominant-negative RhoA increases outflow facility 72. Knepper PA, Goossens W, Mayanil CS. CD44H localization in in perfused human anterior segment cultures. Mol Vis. 2002;8:32– primary open-angle glaucoma. Invest Ophthalmol Vis Sci. 1998; 44. 39:673–680. 88. Epstein D, Rowlette L-L, Roberts B. Acto-myosin drug effects and aqueous outflow function. Invest Ophthalmol Vis Sci. 1999;40: 73. Isacke CM, Yarwood H. The hyaluronan receptor, CD44. Int J Bio- 74–81. chem Cell Biol. 2002;34:718–721. 89. Clark AF, Brotchie D, Read AT, et al. Dexamethasone alters F-actin 74. Harigaya K. Role of CD44 and MT1-MMP in colorectal cancer architecture and promotes cross-linked actin network formation in invasion and metastasis [in Japanese]. Nippon Rinsho. 2003; human trabecular meshwork tissue. Cell Motil Cytoskeleton. 2005; 61(suppl 7):111–115. 60:83–95. 75. Seiki M, Mori H, Kajita M, Uekita T, Itoh Y. Membrane-type 1 90. Burdette SC, Lippard SJ. Meeting of the minds: metalloneurochem- matrix metalloproteinase and cell migration. Biochem Soc Symp. istry. Proc Natl Acad Sci USA. 2003;100:3605–3610. 2003;70:253–262. 91. Vasak M, Hasler DW. Metallothioneins: new functional and struc- 76. Couchman JR, Chen L, Woods A. Syndecans and cell adhesion. Int tural insights. Curr Opin Chem Biol. 2000;4:177–183. Rev Cytol. 2001;207:113–150. 92. Lu H, Hunt DM, Ganti R, et al. Metallothionein protects retinal 77. Couchman JR. Syndecans: proteoglycan regulators of cell-surface pigment epithelial cells against apoptosis and oxidative stress. Exp microdomains? Nat Rev Mol Cell Biol. 2003;4:926–937. Eye Res. 2002;74:83–92.

Downloaded from iovs.arvojournals.org on 09/29/2021