PATTERNS & PHENOTYPES

Regional Distribution of �9�1 Integrin Within theLimbus of the Mouse Ocular SurfaceAhdeah Pajoohesh-Ganji,1,2 Sonali Pal Ghosh,1 and Mary Ann Stepp1,3*

The epithelial basal cells of the corneal limbus are known to contain adult corneal epithelial stem cells, but theproperties of these cells are not well understood. In addition, how and when the limbal epithelium forms duringpostnatal development in mammals is not clear. To better understand the anatomy and cell biology of the limbus, awhole-mount procedure was used to show that the nasal, inferior, temporal, and superior regions of the mouse limbuscontain different numbers of �9 integrin–positive cells most of which are observed in the nasal region. We also showthat this pattern develops progressively over time from 1 to 8 weeks after birth. High magnification image projectionsand three-dimensional reconstructions of the limbal region were generated from confocal images obtained aftertissues were dual stained with �9 integrin and propidium iodide (PI) or triple stained with �9 integrin, E-cadherin, andPI. Data show that �9 integrin is present on the adult mouse cornea in the limbal basal cells and is more abundantin the apical-most cytoplasm of the limbal basal cells, where it can be found colocalized within the plasma membranewith E-cadherin. These studies are an important step toward improving our understanding of the development andcell biology of limbal basal cells. Developmental Dynamics 230:518–528, 2004. © 2004 Wiley-Liss, Inc.

Key words: integrin; �9�1 integrin; limbal stem cell; limbus; stem cell marker; whole-mount procedure

Received 8 October 2003; Revised 23 December 2003; Accepted 9 January 2004

INTRODUCTION

A relatively early event in eye devel-opment is when the lens comes incontact with the overlying surfaceectoderm and induces the forma-tion of corneal epithelium. Theevents that occur after, describedfirst in the chick by Revel and Hay(1969), lead to the development ofthe mature ocular surface with itsthree distinct epithelia: the conjunc-tival, limbal, and corneal epithelia.All three play roles in maintaining theocular surface and protecting itfrom damage while preserving clar-ity for transmission of light to the ret-ina. The corneal epithelium is unique

among stratified squamous epithe-lia, because it overlies an avascularextracellular matrix and is main-tained not by a population of pro-genitor cells within its basal cell layerbut by proliferation of adult stemcells that reside among the cells thatmake up the limbal epithelium(Schermer et al., 1986; Thoft, 1989;Lehrer et al., 1998; Pellegrini et al.,1999; Dua and Azuara-Blanco, 2000;Daniels et al., 2001).

The term limbus means a borderand, in the cornea, the limbal epi-thelium forms the border betweenthe conjunctival epithelium and thecorneal epithelium. The conjunctival

epithelium contains numerous gob-let cells and functions to producethe bulk of the mucins that make upthe tear film (Argueso and Gipson,2001; Kunert et al., 2001). It overlies avascularized stromal matrix, and it isbelieved to be maintained by astem cell population that resideswithin the fornix of the eye (Wei etal., 1995, 1996). Although conjuncti-val epithelial cells can participate inthe healing of wounds to the corneain the complete absence of the lim-bal and corneal epithelia, healing isabnormal and cells continue to ex-press conjunctival-specific keratinswhile on the corneal surface (Sha-

1Department of Anatomy and Cell Biology, The George Washington University Medical Center, Washington, DC2Department of Biological Sciences, The George Washington University, Washington, DC3Department of Ophthalmology, The George Washington University Medical Center, Washington, DCGrant sponsor: NIH/NEI; Grant number: RO1-EY13559-03.Correspondence to: Mary Ann Stepp, Ph.D., Departments of Anatomy and Cell Biology and Ophthalmology, The George WashingtonUniversity Medical Center, 2300 I Street N.W., Washington DC, 20037. E-mail: mastepp@gwu.edu

DOI 10.1002/dvdy.20050Published online 5 May 2004 in Wiley InterScience (www.interscience.wiley.com).

DEVELOPMENTAL DYNAMICS 230:518–528, 2004

© 2004 Wiley-Liss, Inc.

piro et al., 1981; Kinoshita et al., 1983;Tseng et al., 1984; Chen et al., 1994;Moyer et al., 1996). Failure or inabilityof the limbal stem cells to either di-vide and produce corneal epithelialcells or move toward the centralcornea leads to migration of theconjunctival cells onto the corneaand eventual loss of avascularityand light transmission (Pfister, 1994;Tsubota et al., 1995; Dua et al., 2000).Thus, the limbal epithelial cells serveboth as a reservoir for the produc-tion of cells to populate the cornealepithelium and as a barrier againstthe encroachment of conjunctivalcells onto the central cornea. Thisbarrier is likely to consist of epithelialand stromal elements; the extracel-lular matrix beneath the limbal basalcells is distinct from that of either thecornea or conjunctiva (Gipson,1989, 1992; Kolega et al., 1989; Ljubi-mov et al., 1995). How and when thelimbus forms during corneal devel-opment, the nature of the barrier orniche created at the limbus, and theproperties of the cells that make upthis epithelia are not yet fully under-stood.

We know that the limbal epithelialcells differ from both the conjunctivaand the corneal epithelial cells inseveral ways. They express a uniquekeratin profile (Kasper et al., 1988)and, due to the lack of connexin-43–containing gap junctions, do not in-teract with one another the sameway the surrounding conjunctivaland corneal epithelial cells do(Matic et al., 1997). Furthermore, wefound several years ago that theyhad high levels of the protein �9 in-tegrin, whereas the conjunctivaland corneal epithelial cells did not(Stepp et al., 1995). �9 integrin ex-pression was studied further inwound healing experiments, and wefound that it was turned on in thecorneal epithelial cells during cellmigration in response to injury to theocular surface (Stepp and Zhu,1997). From these results, it seemsclear that �9 integrin is expressed onthe transiently amplifying (TA) cellsat the limbus. TA cells are the prog-eny of adult stem cells in regenerat-ing tissues; their numbers are knownto increase in response to stress (Pot-ten and Morris, 1988; Lehrer et al.,1998; Lavker et al., 1998). Several

other molecules have been shownto be localized within the basal epi-thelial cells at the limbus, including�-enolase (Zieske et al., 1992a,b;Chung et al., 1995), p63 (Pellegrini etal., 2001), and TGF-� receptors I andII (Zieske et al., 2001). However, be-cause the number of cells that ex-press these proteins also increase inresponse to tissue injury, like �9 inte-grin, they are ruled out as stem cell-specific markers.

Because integrins are involved incell adhesion, it would be logical topresume that they may be involvedeither in retaining the limbal stemcells within the microenvironment atthe limbus and/or in preventing themigration of conjunctival cells acrossthe limbus. Integrins are integralmembrane glycoproteins with � and� heterodimers that are nonco-valently associated. They mediateinteractions between the cytoskele-ton and extracellular matrix, which inturn, mediate cell adhesion. To date,18 � and 8 � subunits have beencharacterized. Various � subunits in-teract with different � subunits. Forexample, �1 subunit interacts with �subunits 1 through 11 and �V inte-grin, whereas �4 subunit only inter-acts with �6 subunit (Beauvais-Jou-neau and Thiery, 1997; Belkin andStepp, 2000). Integrin knockout micedemonstrate the importance of inte-grins in the migration events that oc-cur during development (Bouvard etal., 2001). For instance, �1 integrinknockout mice cannot develop pastembryonic day 9 (Fassler and Meyer,1995; Stephens et al., 1995). �3 inte-grin knockout mice die shortly afterbirth of kidney and lung abnormali-ties (Kreidberg et al., 1996) and, inaddition, their skin blisters due to dis-organization of the basement mem-brane at the epidermal–dermaljunction (DiPersio et al., 1997).

In an earlier study, we reportedthat �9 integrin was present on eye-lids during eyelid fusion—at around15 days after conception in themouse (Stepp, 1999). Here, for thefirst time, we orient the developingand adult mouse cornea to definethe nasal, inferior, temporal, and su-perior regions. We go on to showthat �9 integrin localization at thesesites varies regionally as does the for-mation of the anatomical niche that

defines the limbal border. To betterdetermine where within the limbalbasal cells �9 integrin was localized,we also evaluated adult and devel-oping corneas using antibodiesagainst �9 integrin and E-cadherin incolocalization studies. In combina-tion with the nuclear marker pro-pidium iodide (PI) and by usingthree-dimensional (3-D) reconstruc-tion of stacks of confocal imagestaken at high magnification, we de-termined the localization of �9 integrinwithin cells at the limbus. Results showthat the anatomical limbus developsover time after birth and reaches itsadult morphology between 6 to 8weeks of postnatal development. �9integrin is regionally localized alongthe circumference of the limbus withthe majority of the �9–positive cellspresent at the nasal region in theadult cornea. Thus, �9 integrin is pri-marily cytoplasmic within the limbalbasal cells but can be observed tocolocalize with E-cadherin–positivecell membranes.

RESULTS

Whole-Mount ProcedureRevealed That �9 Integrin IsRegionally DistributedThroughout the Limbus

To orient the eyes, a suture wasplaced in the sclera of the temporalregion after death, and the eyeswere labeled as left or right. Afterstaining for �9 integrin, four incisionswere made, the eyes were flattenedon a black filter, and the imageswere captured by using a confocalmicroscope. Montages were cre-ated representing the entire circum-ference of the mouse cornea. Asshown in the low-magnificationmontage of two typical eyes in Fig-ure 1, there are more �9 integrin–positive cells in the nasal region ofthe eye than any other region. Vari-ation in the intensity of staining andin the number of �9–positive cellswas observed in the inferior and su-perior regions as highlighted in thetwo eyes shown. Regardless of thisvariability, the increased localizationof �9–positive cells at the nasal re-gion was reproducible in more than10 eyes stained; the number of �9–positive cells decreased moving cir-

�9�1 INTEGRIN WITHIN MOUSE OCULAR SURFACE 519

cumferentially from the nasal to thetemporal region. The montagesshown also highlight the anatomicalcomplexity inherent in this tissue atthe limbal region. The border be-tween cornea and bulbar conjunc-tiva is not a distinct site. The stainingpattern of PI shows that the limbalregion is the site of an enfolding ofthe tissue, forming a ridge or ana-tomical niche, which is less pro-nounced at the inferior region thanelsewhere (Fig. 1). Although �9 inte-grin–positive cells are not exclusivelyrestricted to the anatomical niche,they are generally observed within

this region or in the corneal periph-ery. At its widest, the stained regionhas a width of approximately 300�m. Higher magnification images ofthe limbal regions, along with an im-age of the central cornea, revealthe variability and heterogeneity in�9 integrin expression at the limbus.

Entire Ocular Surface ContainsCells Positive for �9 Integrin 1Week After Birth

Next, we determined at what pointduring postnatal ocular surface de-velopment, the adult anatomy and

�9 integrin expression profiles wereestablished at the limbus. Mouseeyes are closed for the first 2 weeksafter birth; therefore, at 1 week, theeyes were fixed by placing thewhole head in the fixative, as de-scribed in Experimental Proceduressection, to preserve the corneal in-tegrity. Figure 2A,B shows the local-ization of �9 integrin adjacent to thedeveloping limbus (Fig. 2A) and to-ward the cornea (Fig. 2B); respec-tive negative controls are shown inthe insets. Because cells positive for�9 integrin were present in similarnumbers at each of the quadrantsof the ocular surface of the 1-week-old cornea and in the center of thecornea, it was not necessary topresent images of all four quadrantsof the eye. As in the adult, variabilityin the intensity of �9 expression fromcell-to-cell was present. Focusing onthe PI alone, we were unable to ob-serve an anatomical niche.

Fig. 1. Whole-mount double staining with �9 integrin (green) and propidium iodide (PI,red) on 8-week-old mouse corneas. To orient the cornea, care was taken to identify theeye as left or right and a suture was placed in the temporal position. Four incisions weremade after staining and the corneas were flattened and mounted for microscopy asdescribed in Experimental Procedures section. The inner and outer low-magnificationmontages, which were captured with a confocal microscope, represent the entire limbalregion around two different mouse eyes. The asterisks indicate the regions shown at highermagnification in the center along with the central cornea. Note that there are more �9integrin–positive cells in the nasal region in both corneas and a gradual reduction in thenumber of �9 integrin–positive cells toward the temporal region. It is also evident thatvariation exists in corneas from different animals in the relative numbers of cells that arepositive for �9 integrin, especially at the inferior and superior regions.

Fig. 2. Whole-mount double staining with�9 integrin (green) and propidium iodide(PI, red) on mouse cornea, 1 week afterbirth. A: Shown is the peripheral cornea atthe site of the developing limbus. B: Shownis the staining in the central cornea. Theinsets in A and B are negative controls forlimbus and cornea, respectively. The linearstaining, which is present in the negativecontrol in the inset in B, is likely nerves; thereason that more nonspecific staining fornerve is observed at 1 week is that the cor-neal epithelium is extremely thin at thisstage.

520 PAJOOHESH-GANJI ET AL.

Fig. 3. Whole-mount double staining with�9 integrin (green) and propidium iodide(PI, red) on mouse cornea at 2 weeks afterbirth. This image represents a montagetaken of the limbal region around the cor-nea of a 2-week-old mouse. The central cor-nea is shown in the center of the montage.Note that the central cornea and the nasalregion are negative for �9 integrin.

Fig. 4. Whole-mount double staining with�9 integrin (green) and propidium iodide(PI, red) on 4-week-old mouse cornea. Im-ages taken of each of the four regions (su-perior, nasal, inferior, and temporal) areshown in this figure, with the central corneashown at the center. The nasal region showsmore �9 integrin–positive cells at this stagecompared with 2 weeks but still has fewerthan the other three regions.

�9�1 INTEGRIN WITHIN MOUSE OCULAR SURFACE 521

�9 Integrin Localization Beginsto Become RegionallyRestricted by 2 Weeks AfterBirth

Figure 3 shows a montage of the lim-bal region in a 2-week-old mouseeye. Unlike at 1 week, at 2 weeks,the central cornea stained negativefor �9 integrin but the protein wasobserved along the superior, tempo-ral, and inferior aspects of the limbalregion and was absent in the nasalregion. �9 integrin staining at thecorneal periphery was observed in awider region compared with that inthe adult eye; the width was approx-imately 450–550 �m. The anatomi-cal niche is recognizable at the na-sal and the superior regions;however, we were unable to identifyit in the inferior and temporal re-gions.

In the development of the mousecornea, 2 weeks is a particularly sig-nificant time; the eyelids open be-tween 13 and 15 days of postnataldevelopment and both the overalldiameter of the cornea and the ep-ithelial cell thickness increase dra-matically within days. Eyes taken at2 weeks show more variability in �9integrin localization than at otherdevelopmental times; therefore, we

stained a total of six eyes beforeconcluding that the pattern shownin Figure 3 is typical. Only eyes thatwere opened at 2 weeks were usedfor these studies. What is most re-markable about the data obtainedat 2 weeks is that the site with thefewest �9 integrin–positive cells is thenasal region; the nasal region in theadult consistently showed more �9–positive cells than any other site.

Region-specific Staining for �9Integrin Is Present at 4 Weeks

Figure 4 shows the four quadrants ofa 4-week-old mouse cornea, as wellas the central cornea. The nasal re-gion is beginning to show positivestaining; however, the staining in thenasal region is not as intense as thatseen at 8 weeks (compare Fig. 4 withFig. 1). In addition, the superior, tem-poral, and inferior regions still retainhigh levels of �9 integrin expression.The widest region of �9 integrin stain-ing is in the superior region, and itmeasures approximately 250 �m.The central cornea remains nega-tive for �9 integrin, with some back-ground staining. The anatomicalniche is present in the nasal and su-perior regions but remains indistinctin the inferior and the temporal re-

gions. When we evaluated �9 inte-grin localization in the corneas of3-week-old mice, we observed apattern similar to that for 4 weeks(data not shown). While the nasalregion still has fewer �9 integrin–posi-tive cells at 4 weeks than do theother three regions, there is nolonger a complete lack of �9 inte-grin–positive cells at this site. In6-week-old mice, the pattern for �9integrin expression was similar to thatat 8 weeks—with the highest expres-sion in the nasal followed by thetemporal, superior, and inferior re-gions. The anatomical niche was es-tablished in the nasal, temporal, andthe superior regions at 6 weeks, but itwas not visible in the inferior region(data not shown).

�9 Integrin Is PrimarilyLocalized Within the ApicalCytoplasm of the Basal Cellsat the Limbus

The images presented in Figures 1–4are projected z-series acquired byconfocal microscope with a �20 ob-jective at 1-�m intervals. To obtainhigher resolution images, micros-copy was repeated by using a �60oil objective with images obtainedat 0.5-�m intervals. Figure 5A shows

Fig. 5. Three-dimensional reconstruction of high resolution images taken of whole-mount double staining of the 8-week-old mouse corneawith �9 integrin (green) and propidium iodide (PI, red). A: The image is a three-dimensional reconstruction of a 7.5-�m-thick block of tissueshowing the limbal basal cells; �9 integrin–positive basal cells are projected from the apical aspect of the basal cell layer. The cellindicated by the arrow in A is shown at higher magnification in B–D, as projected from the basal (B) and apical (C) aspects, as well asin cross-section (D; C.S.). The dashed lines in B,C indicate the plane through which the cross-section in D was projected. Another cell,indicated by the arrowhead in A, is shown at higher resolution in E–G, from the basal (E) and apical (F) aspects, as well as in cross-section(G). The dashed lines in E,F indicate the plane through which the cross-section in G was taken. Note that there is more �9 integrin at theapical aspect of the cells as shown most clearly in the cross-sections in D and G.Fig. 6. Two-dimensional colocalization studies indicate �9 integrin and E-cadherin can colocalize in the limbal basal cells. A single0.5-�m section taken at high magnification through the limbal basal cells was used to determine whether or not there was evidence forcolocalization of �9 integrin and E-cadherin. Eight-week-old corneas were simultaneously treated with antibodies against �9 integrin(green) and E-cadherin (blue) in whole-mount studies. A: A region through the limbus shown. ImagePro Plus software was used to visualizecolocalization. B: The colocalization mask generated is shown. The red arrows and arrowheads in A and B indicate two cells in which �9integrin partially colocalizes with E-cadherin.Fig. 7. Three-dimensional reconstruction of high magnification images of 8-week-old mouse corneas stained with �9 integrin (green),E-cadherin (blue), and propidium iodide (PI, red). The same two cells indicated by the red arrow and arrowhead in the single 0.5-�msection in Figure 6 are shown in a three-dimensional (3-D) reconstruction of a 7.5-�m-thick block of tissue. A: The �9 integrin–positive basalcells projected from their apical aspect; 15 separate 0.5-�m layers were assembled into a 3-D image using 3-D constructor in ImageProPlus. B,C: Shown in detail are projections of the cell indicated by the white arrow in A from the basal (B) and apical (C) aspects; the insetshows the projection in cross-section (C.S.) taken from the plane indicated in B and C by the dashed lines. D–F Shown in detail are 3-Dprojections of the cell indicated by the white arrowhead in A with the apical aspect of the basal cells in the front. D shows E-cadherinalone, E shows E-cadherin and �9 integrin, and F shows both proteins plus the nuclei stained using PI. Note that, while most of the �9integrin is in the cytoplasm, a portion of the integrin appears to colocalize with E-cadherin–positive cell membranes as indicated by theasterisks in the inset as well as in E.Fig. 8. A model for �9 integrin expression pattern in the developing mouse cornea. The orientation of the eye is indicated at the left topcorner: S, superior; N, nasal; I, inferior; T, temporal. The green area shows the presence of �9 integrin on the mouse ocular surface. Thethin line in the green area indicates the location of the anatomical niche relative to the �9 integrin staining, with the dashed lineshighlighting the regions in which the anatomical niche has not been established in the developing eyes or cannot clearly be identifiedin the adult.

522 PAJOOHESH-GANJI ET AL.

Fig. 5.

Fig. 6.

Fig. 7.

Fig. 8.

�9�1 INTEGRIN WITHIN MOUSE OCULAR SURFACE 523

staining of the limbal basal cells for�9 integrin and PI as viewed from theapical aspect of a 3-D reconstruc-tion made by using the 3-D con-structor module of Image Pro Plussoftware. The projection shown rep-resents 7.5 �m of tissue from the api-cal to basal aspects of the limbalepithelium. The cells indicated bythe arrow and arrowhead in Figure5A is shown in higher magnificationin Figure 5B–D, and Figure 5E–G, re-spectively. The two cells are pro-jected from their basal (Fig. 5B,E)and apical (Fig. 5C,F) aspect as wellas in cross-section (Fig. 5D,G). Fromthese images, we conclude thatmost of the �9 integrin is in the cyto-plasm and there is more �9 integrinat the apical aspect of the basalcells than at the basal aspect,where the cells are in contact withthe basement membrane. We can-not determine from these datawhether or not �9 integrin is presentat the plasma membrane.

Some �9 Integrin Is Found atthe Plasma Membrane Whereit Colocalizes With E-cadherin

In addition to staining with antibod-ies against �9 integrin, we alsostained developing and adult cor-neas with antibodies against E-cad-herin and �9 integrin simultaneously.E-cadherin was used as a marker forepithelial cell membranes; it local-izes within adherens junctions in epi-thelia. A typical image of an 8-week-old mouse cornea is shown in Figure6A; this image shows a single 0.5-�mlayer through the limbal basal cellsof the adult mouse cornea after tis-sues had been processed to show�9 integrin (green) and E-cadherin(blue); the image was obtained byusing a �60 oil objective. While allthe basal cells express E-cadherin,we see a typical pattern of expres-sion of �9 integrin within a subpopu-lation of cells, running diagonallyfrom left to right; the red arrow andarrowhead highlight two cells ex-pressing both �9 integrin and E-cad-herin. The 2-D colocalization was as-sessed on this image using ImageProPlus. The resulting colocalizationmask is presented in Figure 6B. Themask shows in gray scale those pixelsthat contained both blue and green

when the two pseudocolored gray-scale images were merged; the lay-ers above and below this layer gen-erated similar colocalization masks(data not shown). These data sup-port the idea that at least a portionof the �9 integrin in these cells isfound at the plasma membrane.

Figure 7A shows staining of the lim-bal basal cells for �9 integrin, E-cad-herin, and PI as viewed from the api-cal aspect of a 3-D reconstructionmade by using the 3-D constructormodule of ImagePro Plus. The pro-jection shown represents 7.5 �m oftissue from the apical to basal as-pects of the limbal epithelium. Thetwo cells indicated by the arrow andarrowhead in Figure 6 are now indi-cated by a white arrow and arrow-head in Figure 7A. A higher magnifi-cation image of the cell indicatedby the arrow in Figure 7A is projectedfrom its basal (Fig. 7B) and apical(Fig. 7C) aspect as well as in cross-section in the inset. It is clear thatmost of the �9 integrin is in the cyto-plasm but a discrete patch of colo-calization with E-cadherin is alsoseen as indicated by the asterisk.Figure 7D–F shows 3-D projections athigher magnifications of the two ad-jacent �9 integrin–positive cells indi-cated by the arrowhead in Figure7A. The projections, representing 7.5�m of tissue viewed from the apical(front) to basal (back) aspect, aretilted slightly to emphasize theirdepth. Data are shown for E-cad-herin alone (Fig. 7D), E-cadherin and�9 integrin (Fig. 7E), as well as E-cad-herin, �9 integrin, and PI (Fig. 7E);areas showing colocalization of �9integrin and E-cadherin are indi-cated by the asterisks in Figure 7E. InFigure 7D, we see that E-cadherin ismuch less abundant on the basolat-eral membranes of the limbal basalcells and is frequently present onlyon the apical plasma membrane ofthe basal cells. In Figure 7E,F, we seecolocalization both at the cell mem-branes between the two �9 integrin–positive cells and at other sites.Taken together, our data show thatthe localization of �9 integrin is de-velopmentally regulated and re-gionally distributed on the mousecornea and that, within the basalcells at the limbus, �9 integrin ispresent primarily in the apical cyto-

plasm but also is seen in the apical-most cell membrane of the basalcells.

DISCUSSION

By using a whole-mount procedure,we show, for the first time, that the �9integrin expression pattern varies re-gionally over the circumference ofthe adult mouse ocular surface. At 1week after birth, �9 integrin ispresent throughout all regions of theocular surface. At 2 weeks, �9 inte-grin becomes negative in the cen-tral cornea as it begins to be re-stricted to cells near the limbalregion; �9 integrin–positive cells re-main throughout the area of the de-veloping limbus except at the nasalregion, which is negative at 2 weeks.By 4 weeks, �9 integrin expression isincreased in the nasal region andhas begun to decrease at the tem-poral and inferior regions but re-mains distinct from that of the adult.In the adult mouse, which we defineas 8 weeks old, �9 integrin is distrib-uted in a unique pattern around thelimbus. More cells in the nasal regionare �9 integrin–positive, and thenumber of �9 integrin–expressingcells decreases toward the temporalregion.

Anatomical and structural fea-tures may be involved in the region-alized distribution of �9 integrin. Thepresence of tears and growth fac-tors, which accumulate at the nasalregion of the eye, to be drainedthrough the nasolacrimal duct, mayprovide an explanation for thehigher number of �9 integrin–posi-tive cells at this site. An anatomicalniche or depression is well devel-oped in the adult nasal, superior,and temporal regions but cannot bedistinguished in the inferior region.This depression is due in part to thethinning of the epithelium at the lim-bus from 20–23 �m in the centralcornea to 12–15 �m at the limbus.Figure 8 shows a model depicting �9integrin expression in the developingmouse eye. The regionalized expres-sion of �9 integrin is directly respon-sible for the variability observed inthe number of �9-positive cells seenin cross-sections of the adult mouseeye.

�9 integrin can serve as a receptor

524 PAJOOHESH-GANJI ET AL.

for a variety of extracellular ligands,including tenascin-C (Yokosaki et al.,1998) and the EIIIA alternativelyspliced domain of fibronectin (Liao etal., 2002); it may also play a role inmediating cell–cell adhesion (Taookaet al., 1999). To better determine whatrole(s) �9 integrin might be playing oncells within the limbal basal cell layer,we performed colocalization analy-ses at high magnification using E-cad-herin as a marker for epithelial cellmembranes and with PI to show thelocation of the cell nuclei. The 3-D im-ages were made which encom-passed the entire limbal epitheliumfrom its apical to its basal aspect.From these images, we were able todetermine that (1) although �9 inte-grin can be found at the basolateralaspect of some basal cells, little �9integrin is localized beneath the nu-cleus at the basement membranezone; (2) most of the �9 integrin ispresent in the lateral and apical cyto-plasm; (3) occasional sites at the api-cal cell membrane showed colocal-ization of �9 integrin and E-cadherin,indicating that a portion of the �9 in-tegrin is membrane associated.

These observations suggest that�9 integrin in the normal adultmouse cornea is not functioning ex-clusively as an extracellular matrixreceptor. Rather, it appears that �9integrin may also function in cell–cell adhesion or signaling. �9 integrincouples with �1 integrin to form afunctional heterodimer; the functionof this integrin in epithelial tissues isnot known (Sheppard, 1996, 1998).The skin and cornea have abundantlevels of �2 and �3 integrins, whichalso couple with the �1 integrin sub-unit to form ��1 heterodimers. Theseother �1-family integrins are local-ized to regions of both cell–cell andcell–substrate interactions with themajority of the protein present as im-mature precursors within the cyto-plasm with lesser amounts found onthe plasma cell membrane. Whilethe function of �1-family integrins inmediating cell–cell interactions re-mains controversial, they likely playroles in growth factor signaling andthe ability of cells to respond rapidlyto injury. For epithelial keratinocytesdeficient in �1 integrin, cell adhesionand migration are impaired in vitroand in vivo (Grose et al., 2002) and

the lack of �3 integrin affects thedeposition and organization ofbasement membrane componentsboth in vitro and in vivo (Kreidberg etal., 1996; DiPersio et al., 1997). Infact, Hodivala-Dilke and colleagues(1998) propose that �3�1 integrin in-teracts with other integrins at sites ofcell–cell and cell–substrate adhe-sion to inhibit the functions of otherintegrins. It has been difficult to studythe role of �9 integrin in epithelialcells, because expression of �9 inte-grin decreases rapidly once keratin-ocytes are placed in culture (datanot shown). Although it is not clearhow the multiple �1-family integrinsfunction in epithelial cells, theyclearly play important roles.

Mouse corneal development hasbeen studied by several groups(Stepp, 1999; Saika et al., 2001; Col-linson et al., 2002; Song et al., 2003;Nagasaki and Zhao, 2003). Between6 and 8 weeks, Nagasaki and Zhao(2003) found that the corneal diam-eter increased from 2.7 to 3.0 mmand that no significant increase insize of the cornea was observed af-ter 8 weeks; they used mouse cor-neas from 7 to 11 weeks of age tostudy the natural migration of cellsfrom the limbal region to the centralcornea by using green fluorescentprotein (GFP) -labeled cells in atransgenic mouse model. Their studyextended the XYZ hypothesis origi-nally proposed by Thoft and Friend(1983), which stated that the cornearegulated the number of epithelialcells on its surface such that the pro-liferation of transient amplifying cellsin the basal cell layer (X) and therate of centripetal movement ofcells from the limbus to the centralcornea (Y) was equal to the loss ofcells from the entire apical surface(Z). Nagasaki and Zhao (2003) foundthat between 7 and 11 weeks ofage, the rate of cell movement fromthe limbal region to the central cor-nea remained constant and inter-preted this finding to mean thatthe corneal epithelial cells hadachieved a steady proliferative andmigratory state by 7 weeks of age.

Collinson and colleagues (2002)suggest that it takes 20 weeks afterbirth before the mouse ocular sur-face is fully mature. Their studieswere performed by using mice chi-

meric for expression of LacZ. Overtime after birth, radial stripes of blueor clear cells could be observed pro-ceeding toward the central corneawith the stripes being more well re-solved as the mice aged. While Na-gasaki and Zhao (2003) also ob-served radial stripes of cells developon the cornea over time in their GFP-transgenic mice, they reported nochanges in the numbers or quality ofthe stripes after 8 weeks. It is notclear why Collinson and colleagues(2002) and Nagasaki and Zhao(2003) define acquisition of adultcorneal epithelial cell phenotype inmice so differently. However, thestudies presented here support theidea that the adult mouse cornealepithelial and limbal basal cell phe-notypes are first present between 6and 8 weeks.

Many groups are trying to identifyand characterize adult epithelialstem cells and potential surfacemarkers (reviewed in Watt, 1998;Lavker and Sun, 2000). The localiza-tion of �9 integrin to the limbal basalcells suggests the possibility that itmight play such a role. Most studiessuggest that the percentage ofadult stem cells in regenerative epi-thelial tissues is generally somewherefrom 0.1 to 4% of the total cell pop-ulation (Potten and Morris, 1988; Pot-ten and Loeffler, 1990; Dua et al.,2000) and the number of �9 integrin–positive limbal basal cells at the na-sal region exceeds those values. To-gether with the fact that �9 integrinis also up-regulated during woundhealing, these data suggest that themajority of �9 integrin–positive cellsin the adult mouse limbus are TAcells. However, these facts do notrule out the possibility that �9 integrinmight also be present on both the TAand the corneal epithelial stem cells.Further studies showing whether ornot �9 integrin–positive cells are la-bel-retaining will resolve this issue.We are currently optimizing stainingprocedures to permit us to simulta-neously assess BrdU label-retainingcells and the �9 integrin–positivecells as well as other epithelial-spe-cific integrins on the developingmouse ocular surface. Understand-ing the development of the limbusand of the cells that reside there,which include the adult corneal ep-

�9�1 INTEGRIN WITHIN MOUSE OCULAR SURFACE 525

ithelial stem cells, should lead to abetter understanding of the causesof corneal epithelial stem cell defi-ciency diseases and lead to im-proved treatments for patients suf-fering from these conditions. Whilemouse and human corneas are dif-ferent in many ways, studying themouse cornea will give us the oppor-tunity to evaluate corneal stem cellsdiseases in ways not possible in pa-tients.

EXPERIMENTAL PROCEDURES

Animals

All experiments described in this arti-cle conformed to the guidelines es-tablished by the George Washing-ton University Medical CenterInstitution Animal Care and UseCommittee as well as to the ARVOstatement for the Use of Animals inOphthalmic and Vision Research. Ifthe eyelids were closed, the animalswere decapitated; otherwise theywere killed using lethal injection withsodium pentobarbital.

Immunofluorescence onWhole-Mount Corneas

One-week-old mice were killed bydecapitation, whereas from 2 to 8weeks, mice were killed by lethal in-jection. There were at least 4 eyesused for each time point, except for6 and 8 weeks in which at least 10eyes were used; we made sure thatthe results were the same in rightand left eyes. For 1-week-old mice,the whole heads were fixed for 2 hrin 4:1 dilution of prechilled 100%methanol and dimethyl sulfoxide(DMSO) at �20°C and a cut wasmade around the eye to removethe eyelids without damaging thecornea. After removing the eyelids,a suture was put in the temporal po-sition. For 2-week-old mice, only cor-neas with open eyelids were used.After suturing the eyes, they werelabeled as left or right, enucleated,and stored in 100% methanol. Theeyes were kept intact throughoutthe whole-mount procedure.

At 4, 6, and 8 weeks of age, asuture was put in the temporal posi-tion and the eyes were labeled asleft or right. The eyes were enucle-

ated and fixed in methanol andDMSO mixture for 2 hr at �20°C, andwere stored in 100% methanol at�20°C. The backs of the eyes werecut, and the retina, the lens, and theiris were removed before staining. Alleyes were transferred from 100%methanol into 1� phosphate buff-ered saline (PBS; 10� PBS was madeas follows: 14.4 g of Na2HPO4, 2.4 gof KH2HPO4, 2 g of KCl, 80 g of NaCl,in total volume of 1 liter of water, pH7.4) in a graded methanol series(70%, 50%, and 30% methanol and1� PBS, 30 min each). All washesand incubations were done at roomtemperature with gentle shaking un-less specified otherwise. After wash-ing the eyes in 1� PBS twice for 30min each, the eyes were incubatedwith blocking buffer (In 100 ml 1�PBS, add 1 g of bovine albumin, stirfor 30 min, add 1 ml of horse serum,and stir for an additional minute) for2 hr. The eyes were then incubatedwith primary antibody overnight at4°C. For some studies, colocalizationwith E-cadherin was performed;both �9 integrin and E-cadherin pri-mary antibodies were added simul-taneously when dual labeling wasrequired. Polyclonal anti-peptidesera recognizing �9 integrin wascharacterized as described previ-ously (Sta Iglesia et al., 2000), and ratanti–E-cadherin antibody (catalogno. 13-1900) was purchased fromZymed Laboratories. The next day,eyes were washed five times with 1�PBS and 0.02% Tween 20 (PBST) for 1hr and were blocked for 2 hr. Theywere incubated with goat anti-rab-bit Alexa 488 (catalog no. A-11008)and goat anti-rat (catalog no. 112-175-102, Jackson ImmunoResearchLaboratories) secondary antibodiesovernight at 4°C. Next day, the eyeswere washed three times with PBSTfor 1 hr each and were incubated inPI (1:1,000) for 5 min followed bythree washes with Millipure water 5min each. Secondary rabbit anti-body and PI (catalog no. P-1304)were purchased from MolecularProbes. To achieve the best result inflattening the eyes, four incisionswere made and the eyes were puton a black filter before adding theprolonged mounting medium andcover-slipping. The images were

captured with a confocal micro-scope.

Confocal Microscopy

Confocal microscopy was per-formed at the Center for Microscopyand Image Analysis (CMIA) at theGeorge Washington University Med-ical Center. A Bio-Rad MRC 1024confocal laser scanning microscope(Hercules, CA) equipped with kryp-ton–argon laser and an OlympusIX-70 inverted microscope (Melville,NY) was used to image the localiza-tion of Alexa 488 (488-nm laser lineexcitation; 522/35 emission filter), PI(568-nm excitation; 605/32 emissionfilter), and Cy5 (647-nm excitation;680/32 emission filter). Optical sec-tions of confocal epifluorescenceimages were acquired sequentiallyat 1-�m intervals by using a �20 ob-jective lens (NA � 0.7) or at 0.5-�mintervals by using a �60 oil objective(NA � 1.40) with Bio-Rad LaserSharpversion 3.2 software. When using the�60 objective, images were ac-quired from the apical aspect of theepithelium into the underlying stro-ma; typically, this approach re-quired capturing approximately 40optical sections.

Image Analysis

Adobe Photoshop version 7.0 soft-ware with Bio-Rad plugins was usedto both convert images from Bio-Rad PIC into tiff files and to mergeimages to create montages. Im-agePro Plus version 5.0 software(Media Cybernetics, Silver Springs,MD) was used to merge stacks oflaser confocal images to both per-form 2-D colocalization measure-ments, and to render 3-D images bymeans of an ImagePro Plus 3D Con-structor version 5.0 module. Basedon the presence of �9 integrin–posi-tive cells only in the basal cell layer,only 15 of the 41 layers acquired (7.5�m of tissue) were used for 3-D ren-dering and colocalization analyses.

ACKNOWLEDGMENTSWe thank Dr. Robyn Rufner, the di-rector of CMIA at The George Wash-ington University, and Dr. AnastasPopratiloff for help with the confocalmicroscopy and image analysis. We

526 PAJOOHESH-GANJI ET AL.

also thank Drs. Ken Brown, RobertDonaldson, Robert Oakley, FrankTurano, and Raymond Walsh forhelpful discussions and advice.M.A.S. was funded by the NIH/NEI.

REFERENCES

Argueso P, Gipson IK. 2001. Epithelial mu-cins of the ocular surface: structure,biosynthesis and function. Exp Eye Res73:281–289.

Beauvais-Jouneau A, Thiery JP. 1997. Mul-tiple roles for integrins during develop-ment. Mol Biol Cell 89:5–11.

Belkin AM, Stepp MA. 2000. Integrins asreceptors for laminins. Microsc ResTech 51:280–301.

Bouvard D, Brakebusch C, Gustafsson E,Aszodi A, Bengtsson T, Berna A, FasslerR. 2001. Functional consequences ofintegrin gene mutations in mice. CircRes 89:211–223.

Chen WY, Mui MM, Kao WW, Liu CY,Tseng SC. 1994. Conjunctival epithelialcells do not transdifferentiate in orga-notypic cultures: expression of K12 ker-atin is restricted to corneal epithelium.Curr Eye Res 13:765–778.

Chung EH, DeGregorio PG, Wasson M,Zieske JD. 1995. Epithelial regenerationafter limbus-to-limbus debridement. Ex-pression of �-enolase in stem and tran-sient amplifying cells. Invest Ophthal-mol Vis Sci 36:1336–1343.

Collinson JM, Morris L, Reid AI, Ramaesh T,Keighren MA, Flockhart JH, Hill RE, TanSS, Ramaesh K, Dhillon B, West JD. 2002.Clonal analysis of patterns of growth,stem cell activity, and cell movementduring the development and mainte-nance of the murine corneal epithe-lium. Dev Dyn 224:432–440.

Daniels JT, Dart JK, Tuft SJ, Khaw PT. 2001.Corneal stem cells in review. WoundRepair Regen 9:483–494.

DiPersio CM, Hodivala-Dilke KM, JaenischR, Kreidberg JA, Hynes RO. 1997. �3�1integrin is required for normal develop-ment of the epidermal basementmembrane. J Cell Biol 137:729–742.

Dua HS, Azuara-Blanco A. 2000. Limbalstem cells of the corneal epithelium.Surv Ophthalmol 44:415–425.

Dua HS, Saini JS, Azuara-Blanco A, GuptaP. 2000. Limbal stem cell deficiency:concept, aetiology, clinical presenta-tion, diagnosis and management. Ind JOphthalmol 48:83–92.

Fassler R, Meyer M. 1995. Consequencesof lack of �1 integrin gene expression inmice. Genes Dev 9:1896–1908.

Gipson IK. 1989. The epithelial basementmembrane zone of the limbus. Eye 3:132–140.

Gipson IK. 1992. Adhesive mechanisms ofthe corneal epithelium. Acta Ophthal-mol Suppl 202:13–17.

Grose R, Hutter C, Bloch W, Thorey I, WattFM, Fassler R, Brakebusch C, Werner S.2002. A crucial role of �1 integrins forkeratinocyte migration in vitro and dur-

ing cutaneous wound repair. Develop-ment 129:2303–2315.

Hodivala-Dilke KM, DiPersio CM, Kreid-berg JA, Hynes RO. 1998. Novel roles for�3�1 integrin as a regulator of cytoskel-etal assembly and as a trans-dominantinhibitor of integrin receptor function inmouse keratinocytes. J Cell Biol 142:1357–1369.

Kasper M, Moll R, Stosiek P, Karsten U.1988. Patterns of cytokeratin and vi-mentin expression in the human eye.Histochemistry 89:369–377.

Kinoshita S, Friend J, Thoft RA. 1983. Bi-phasic cell proliferation in transdiffer-entiation of conjunctival to corneal ep-ithelium in rabbits. Invest OphthalmolVis Sci 24:1008–1014.

Kolega J, Manabe M, Sun TT. 1989. Base-ment membrane heterogeneity andvariation in corneal epithelial differen-tiation. Differentiation 42:54–63.

Kreidberg JA, Donovan MJ, Goldstein SL,Rennke H, Shepherd K, Jones RC, Jae-nisch R. 1996. �3�1 integrin has a cru-cial role in kidney and lung organo-genesis. Development 122:3537–3547.

Kunert KS, Keane-Myers AM, Spurr-Michaud S, Tisdale AS, Gipson IK. 2001.Alteration in goblet cell numbers andmucin gene expression in a mousemodel of allergic conjunctivitis. InvestOphthalmol Vis Sci 42:2483–2489.

Lavker RM, Sun TT. 2000. Epidermal stemcells: properties, markers, and location.Proc Natl Acad Sci U S A 97:13473–13475.

Lavker RM, Wei ZG, Sun TT. 1998. Phorbolester preferentially stimulates mousefornical conjunctival and limbal epi-thelial cells to proliferate in vivo. InvestOphthalmol Vis Sci 39:301–307.

Lehrer MS, Sun TT, Lavker RM. 1998. Strat-egies of epithelial repair: modulation ofstem cell and transit amplifying cellproliferation. J Cell Sci 111:2867–2875.

Liao YF, Gotwals PJ, Koteliansky VE, Shep-pard D, Van De Water L. 2002. The EIIIAsegment of fibronectin is a ligand forintegrins �9�1 and �4�1 providing anovel mechanism for regulating celladhesion by alternative splicing. J BiolChem 277:14467–14474.

Ljubimov AV, Burgeson RE, Butkowski RJ,Michael AF, Sun TT, Kenney MC. 1995.Human corneal basement membraneheterogeneity: topographical differ-ences in the expression of type IV col-lagen and laminin isoforms. Lab Invest72:461–473.

Matic M, Petrov IN, Chen S, Wang C,Dimitrijevich SD, Wolosin JM. 1997. Stemcells of the corneal epithelium lackconnexins and metabolite transfer ca-pacity. Differentiation 61:251–260.

Moyer PD, Kaufman AH, Zhang Z, KaoCW, Spaulding AG, Kao WW. 1996.Conjunctival epithelial cells can resur-face denuded cornea, but do nottransdifferentiate to express cornea-specific keratin 12 following removal oflimbal epithelium in mouse. Differentia-tion 60:31–38.

Nagasaki T, Zhao J. 2003. Centripetalmovement of corneal epithelial cells inthe normal adult mouse. Invest Oph-thalmol Vis Sci 44:558–566.

Pellegrini G, Golisano O, Paterna P, Lam-biase A, Bonini S, Rama P, De Luca M.1999. Location and clonal analysis ofstem cells and their differentiatedprogeny in the human ocular surface.J Cell Biol 145:769–782.

Pellegrini G, Dellambra E, Golisano O,Martinelli E, Fantozzi I, Bondanza S, Pon-zin D, McKeon F, De Luca M. 2001. p63identifies keratinocyte stem cells. ProcNatl Acad Sci U S A 98:3156–3161.

Pfister RR. 1994. Corneal stem cell dis-ease: concepts, categorization, andtreatment by auto- and homo-trans-plantation of limbal stem cells. CLAO J20:64–72.

Potten CS, Loeffler M. 1990. Stem cells:attributes, cycles, spirals, pitfalls anduncertainties. Lessons for and from thecrypt. Development 110:1001–1020.

Potten CS, Morris RJ. 1988. Epithelial stemcells in vivo. J Cell Sci Suppl 10:45–62.

Revel JP, Hay ED. 1969. Fine structure ofthe developing avian cornea in mono-graphs. In: Wolsky A, Chen PS, editors.Developmental biology. Vol. 1. NewYork: S. Karger Publishing.

Saika S, Saika S, Liu CY, Azhar M, SanfordLP, Doetschman T, Gendron RL, KaoCW, Kao WW. 2001. TGF�2 in cornealmorphogenesis during mouse embry-onic development. Dev Biol 240:419–432.

Schermer A, Galvin S, Sun TT. 1986. Differ-entiation-related expression of a major64K corneal keratin in vivo and in cul-ture suggests limbal location of cornealepithelial stem cells. J Cell Biol 103:49–62.

Shapiro MS, Friend J, Thoft RA. 1981. Cor-neal re-epithelialization from the con-junctiva. Invest Ophthalmol Vis Sci 21:135–142.

Sheppard D. 1996. Epithelial integrins.Bioessays 18:655–660.

Sheppard D. 1998. Airway epithelial inte-grins: why so many? Am J Respir CellMol Biol 19:349–351.

Song J, Lee YG, Houston J, Petroll WM,Chakravarti S, Cavanagh HD, JesterJV. 2003. Neonatal corneal stromal de-velopment in the normal and lumican-deficient mouse. Invest Ophthalmol VisSci 44:548–557.

Sta Iglesia DD, Stepp MA. 2000. Disruptionof the basement membrane after cor-neal debridement. Invest OphthalmolVis Sci 41:1045–1053.

Sta Iglesia DD, Gala PH, Qiu T, Stepp MA.2000. Integrin expression during epithe-lial migration and restratification in thetenascin-C-deficient mouse cornea.J Histochem Cytochem 48:363–376.

Stephens LE, Sutherland AE, KlimanskayaIV, Andrieux A, Meneses J, PedersenRA, Damsky CH. 1995. Deletion of �1integrins in mice results in inner cellmass failure and peri-implantation le-thality. Genes Dev 9:1883–1895.

�9�1 INTEGRIN WITHIN MOUSE OCULAR SURFACE 527

Stepp MA. 1999. �9 and �8 integrin ex-pression correlates with the merger ofthe developing mouse eyelids. DevDyn 214:216–228.

Stepp MA, Zhu L. 1997. Upregulation of �9integrin and tenascin during epithelialregeneration after debridement in thecornea.JHistochemCytochem45:189–201.

Stepp MA, Zhu L, Sheppard D, Cranfill RL.1995. Localized distribution of �9 inte-grin in the cornea and changes in ex-pression during corneal epithelial celldifferentiation. J Histochem Cytochem43:353–362.

Taooka Y, Chen J, Yednock T, SheppardD. 1999. The integrin �9�1 mediates ad-hesion to activated endothelial cellsand transendothelial neutrophil migra-tion through interaction with vascularcell adhesion molecule-1. J Cell Biol 145:413–420.

Thoft RA. 1989. The role of the limbus inocular surface maintenance and re-pair. Acta Ophthalmol Suppl 192:91–94.

Thoft RA, Friend J. 1983. The X, Y, Z hy-pothesis of corneal epithelial mainte-nance. Invest Ophthalmol Vis Sci 24:1442–1443.

Tseng SC, Hirst LW, Farazdaghi M, GreenWR. 1984. Goblet cell density and vas-cularization during conjunctival trans-differentiation. Invest Ophthalmol VisSci 25:1168–1176.

Tsubota K, Toda I, Saito H, Shinozaki N,Shimazaki J. 1995. Reconstruction ofthe corneal epithelium by limbal allo-graft transplantation for severe ocularsurface disorders. Ophthalmology 102:1486–1496.

Watt FM. 1998. Epidermal stem cells:markers, patterning and the control ofstem cell fate. Trans R Soc Lond B BiolSci 353:831–837.

Wei ZG, Cotsarelis G, Sun TT, Lavker RM.1995. Label-retaining cells are prefer-entially located in fornical epithelium:implications on conjunctival epithelialhomeostasis. Invest Ophthalmol Vis Sci36:236–246.

Wei ZG, Sun TT, Lavker RM. 1996. Rabbitconjunctival and corneal epithelialcells belong to two separate lineages.Invest Ophthalmol Vis Sci 37:523–533.

Yokosaki Y, Matsuura N, Higashiyama S,Murakami I, Obara M, Yamakido M,Shigeto N, Chen J, Sheppard D. 1998.Identification of the ligand binding sitefor the integrin �9�1 in the third fi-bronectin type III repeat of tenascin-C.J Biol Chem 273:11423–11428.

Zieske JD, Bukusoglu G, Yankauckas MA.1992a. Characterization of a potentialmarker of corneal epithelial stem cells.Invest Ophthalmol Vis Sci 33:143–152.

Zieske JD, Bukusoglu G, Yankauckas MA,Wasson ME, Keutmann HT. 1992b.�-enolase is restricted to basal cells ofstratified squamous epithelium. DevBiol 151:18–26.

Zieske JD, Hutcheon AE, Guo X, ChungEH, Joyce NC. 2001. TGF-� receptortypes I and II are differentially ex-pressed during corneal epithelialwound repair. Invest Ophthalmol VisSci 42:1465–1471.

528 PAJOOHESH-GANJI ET AL.