A Tissue Culture Assay of Corneal
Epithelial Wound Closure
Marcia M. Jumblarr and Arthur H. Neufeld
Experimental assays have been developed using cultured tissue derived from rabbit corneal epithelium
to study migration of epithelial sheets during wound closure and cell-substrate adhesion. To study wound
closure, epithelial defects, 6 mm in diameter, were produced in vitro in 24 well multiplates by a local
freezing technique, and the size of the remaining defect was quantitated over time by staining. To study
adhesion, cultured cells were labeled with 3H-leucine, suspended, and added to fresh culture plates. At
various times, adherent cells were lysed and the radioactivity of the lysate was determined. Serum
enhances the closure of experimental defects, but laminin and fibronectin have no effect. Agents which
alter mitotic rate, such as epidermal growth factor and 5-fluorouracil, do not influence the rate of wound
closure in this assay. Compounds which elevate intracellular levels of cyclic AMP inhibit wound closure
but promote cell-substrate adhesion. Thus, cultured corneal epithelial cells may be used to assay for
influences on the migratory events governing closure of superficial epithelial wounds. Invest Ophthalmol
Vis Sci 27:8-13, 1986
The corneal epithelium closes superficial wounds by
migration of the epithelial sheet over the denuded
stroma. Mitosis ceases at the wound periphery, and a
cohesive layer of epithelial cells slides toward the central
wound area.l>2 During this time, cells adhere reversibly
to the corneal surface in the absence of attachment
organelles and sustained migration is dependent upon
continued synthesis of cell surface glycoproteins.3'4
Once coverage of the wound is complete, migration
ceases, and the epithelial monolayer re-forms stable
hemidesmosomal attachments to its basement membrane. 3 Mitosis then resumes, resulting eventually in
a mature stratified corneal epithelium.1
Recently, we have established and characterized
cultures of rabbit corneal epithelial cells. These cells
require cholera toxin for optimal growth, form partially
stratified cell layers with apical microvilli, and retain
their ability to respond to /3-adrenergic agonists with
increased cyclic AMP synthesis.5 We have used these
cell cultures to describe aspects of corneal epithelial
wound healing, to study the effects of pharmacological
agents on wound closure and cell-substrate adhesion,
and to develop an assay which may prove useful as an
alternative to in vivo toxicity testing.
From the Ophthalmic Pharmacology Unit Eye Research Institute
and Department of Ophthalmology, Harvard Medical School, Boston,
Massachusetts.
This work was supported by National Institutes of Health grant
EY02360.
Submitted for publication: November 14, 1984.
Reprint requests: Dr. Arthur H. Neufeld, Eye Research Institute,
20 Staniford Street, Boston, MA 02114.
Materials and Methods
Corneal Epithelial Cultures
Rabbit corneal epithelial cell cultures were initiated
from Dispase II treated corneas as previously described.5 New Zealand white rabbits, used as a source
of corneal tissue, were treated in accordance with the
ARVO Resolution on the Use of Animals in Research.
Cultures were established in the absence of a feeder
layer in a medium consisting of equal parts of Dulbeccos Modified Eagles Medium and Hams F12 supplemented with 5% fetal bovine serum, cholera toxin
(0.1 Mg/ml), epidermal growth factor (10 ng/ml), insulin
(5 n%/m\), gentamicin (5 /ig/ml) and dimethylsulfoxide
(0.5% v/v). Epithelial cells derived from six corneas
were pooled and plated into the 6 wells of a 35-mm
diameter multiplate. Cultures were fed three times
weekly. After 7-10 days the confluent multilayer (cf
reference 5 for morphology) were subcultured, and the
cells derived from each 35-mm diameter well were
plated into a 24-well multiplate, each well of which
contained 1 ml of the above medium, except that cholera toxin was omitted. In some experiments, cells were
plated into 6-well multiplates in the presence or absence
of cholera toxin as indicated.
Wound Closure Assay
Discs, 6 mm in diameter, were cut from Millipore
HA filters, rinsed in 6 changes of distilled water, boiled
in distilled water, and dried in a laminar flow hood
prior to use. A disc was placed on the surface of each
Downloaded From: http://iovs.arvojournals.org/pdfaccess.ashx?url=/data/journals/iovs/933126/ on 06/14/2017
No. 1
CORNEAL EPITHELIAL WOUND CLOSURE / Jumblorr and Neufeld
Fig. 1. Experimental wound
closure assay: Corneal epithelial
cells were subcultured into 24well multiplates and grown to
confluency. Wounds were produced in quadruplicate by freezing at the indicated times before
fixing and staining. Control
(CON) were not wounded. Dark
areas are remaining cellular
areas; light areas are remaining
cell-free wound areas.
24
Con
32
48
hrs
culture of a 24-well multiplate, gently tapped down,
and a stainless steel probe (6 mm in diameter) was
cooled in liquid nitrogen and placed against the plastic
surface opposite the disc for 5 sec. The probe was then
removed, medium consisting of Dulbeccos Modified
Minimal Essential Medium and Hams F12 with the
addition of gentamycin and dimethylsulfoxide and with
or without other components as indicated was added,
and the disc carefully lifted out leaving a discrete circular defect in the ceil layer.
To determine the time course of wound closure,
wounding was done in replicate fashion at staggered
intervals, usually 0, 6, 24, 32 and 48 hr before the fixation step. At the time the earliest set of replicate control wounds would be almost closed, all cultures were
drained of medium, fixed with neutral buffered formalin, and stained in situ with full strength Giemsa
(see Fig. 1 for an example). The size of the remaining
defect at the staggered time intervals, as revealed by
the unstained cell-free area, was determined by projecting the plates onto a screen with an overhead projecter at a fixed distance, tracing the unstained area
onto paper, and cutting out and weighing the remaining
wound area. The weight was normalized, as a percentage, to that of the initial wound size or converted
to area in mm 2 for comparison of wound sizes. To
determine the effects of various agents on wound closure, all 24 wounds were made at 0 hr and fixed and
stained at 42 hr. Relative wound size was determined
as above.
Mitotic Rate
Cells derived from primary culture were plated in a
24-well multiplate at 1 X 104 cells/well. One ml of
culture medium containing 5-fluorouracil (5-FU) and/
or epidermal growth factor (EGF) at the indicated concentration was added to each well. At the indicated
times, cell number was determined by hemacytometer
count.
Cyclic AMP Synthesis
Synthesis of cyclic AMP by cultured corneal epithelial cells was measured in cells subcultured in both the
presence and absence of stimulators of adenylate cyclase. Medium was removed from the 6- or 24-well
culture dishes and replaced with physiological buffer
containing 0.5 mM isobutylmethylxanthine, a phosphodiesterase inhibitor.5 After 20 min, the buffer was
removed and the adherent cell layer was lysed in 0.1
M KOH at 100°C. This alkaline extract was neutralized
with an equal volume of 0.1 N HC1, and the resulting
protein precipitate was removed by centrifugation
(2000 X G, 15 min, 4°C), solubilized with 1 N NaOH
and quantitated by the method of Lowry et al.6 Cyclic
AMP concentration in the supernatant was determined
by radioimmunoassay using commercially available
reagents. Rates of cyclic AMP synthesis are expressed
as the mean ± SEM (number of observations) of pmoles
cyclic AMP/mg protein/20 min.
Downloaded From: http://iovs.arvojournals.org/pdfaccess.ashx?url=/data/journals/iovs/933126/ on 06/14/2017
10
INVESTIGATIVE OPHTHALMOLOGY & VISUAL SCIENCE / January 1986
T3
O
18 24
Hours
30
36
42
Fig. 2. Time course of wound closure in vitro: Wounded cultures
were incubated in the presence (•) or absence (O) of 5% fetal bovine
serum and the size of the wounds determined and plotted as a function
of time after healing. Mean ± SEM (n = 4).
Cell-Substrate Adhesion Assay
Corneal epithelial cells were grown to confluency
and then labeled with 1 ^Ci/ml 3H-leucine (300 mCi/
mmol) for 18 hr. Cells were removed from the culture
10-CM
Vol. 27
dish with Dispase, centrifuged, resuspended in HEPES
buffered Dulbeccos Modified Eagles Medium, triturated with a flame polished pipet, and passed through
2 layers of 120 u mesh nylon cloth. The resulting single
cell suspension was adjusted to 1-5 X 104 cells/ml in
the same medium at 4°C. Aliquots containing
1-5 X 103 cells were placed in the 24 wells of a multiplate, each well containing 1 ml of medium. Attachment proceeded for various times up to 30 min at 22°C
in room atmosphere. At the end of the attachment
period, medium and unattached cells were removed
and the adherent cells were rinsed twice with 1 ml medium. Adherent cells were lysed with 1 ml 0.1 N NaOH
containing 1% Triton X-100 and samples of the lysate,
as well as aliquots of the original cell suspension, were
placed in scintillation fluid to determine radioactivity.
Radioactivity in the adherent cells was expressed as a
percentage of that present in the initial cell aliquot.
Materials, Drugs, and Supplies
Tissue culture multiplates were obtained from Falcon (Oxnard, CA); medium and serum from Microbiological Associates (Walkersville, MD), Dispase II
from Boehringer-Mannheim (Indianapolis, IN); insulin
and epidermal growth factor from Collaborative Research (Waltham, MA); 5-fluorouracil, cholera toxin
and forskolin from Calbiochem (San Diego, CA); 3Hleucine and components of cyclic AMP radioimmunoassay from New England Nuclear (Boston, MA);
type HAfiltersfrom Millipore (Bedford, MA); laminin
and fibronectin from Bethesda Research Labs (Gaithersburg, MD) and isoproterenol, isobutylmethylxanthine, Giemsa stock solution, and Triton X-100 from
Sigma (St. Louis, MO).
E
CO
CD
C
3
O
Results
Cyclic AMP Levels in Cultured Cells
5--
0
0.1
1
% FCS
Fig. 3. Effect of fetal calf serum (FCS) on wound closure. Cultures
with experimental wounds were incubated in the presence of the
indicated serum concentrations and at 42 hr were fixed and stained.
Mean ± SEM (n = 6).
Primary corneal epithelial cultures synthesize cyclic
AMP at the rate of 22 ± 3 (6) pmol cyclic AMP/mg
prot/20 min. The continued presence of cholera toxin
in the primary tissue culture medium greatly enhances
the rate of epithelial cyclic AMP synthesis to 386 ± 47
(6) pmol cyclic AMP/mg prot/20 min. However, cells
originally grown in cholera toxin-containing medium
and subcultured into cholera toxin-free medium for
one wk have essentially basal rates of cyclic AMP synthesis 46 ± 8 (6) pmol cyclic AMP/mg prot/20 min.
We have, therefore, used cholera toxin-free subcultures,
derived from cholera toxin-containing primary cultures, to assay the effects of drugs on wound closure
and cell-substrate adhesion.
Downloaded From: http://iovs.arvojournals.org/pdfaccess.ashx?url=/data/journals/iovs/933126/ on 06/14/2017
No. 1
11
CORNEAL EPITHELIAL WOUND CLOSURE / Jumblorr and Neufeld
Wound Closure Model
The wounds obtained by local freezing are discrete
and of uniform size with an area of approximately 28
± 0.2 mm2 (mean ± SEM) for 24 wounds. The epithelial cells in the wound that are damaged adhere to
the filter disc as it is lifted away, leaving a clearly demarcated cell-free wound area (Fig. 1). The remaining
epithelial cells, moving as a cohesive sheet, gradually
cover the easily visualized and quantitated central
wound area. When wounds are made in replicate fashion and sequentially over time, a time curve can be
constructed by relating the remaining wound area to
the original wound area and the rate of closure determined (Fig. 2). When wounded cultures are incubated
at 37°C in the presence of medium containing 5% fetal
bovine serum, the central area is re-covered at a rate
of 0.8 mm2/hr over the first 30 hr, and closure is generally complete before 48 hr (Fig. 2).
Because a discrete and measurable wound remains
after 42 hr of closure, the size of the wound at that
interval was used as a measure to define the basal conditions for wound closure. In serum-free medium, epithelial wounds closed slowly; over 30% of the original
wound remained at 42 hr (Fig. 2) and continued incubation for up to 72 hr did not lead to further closure.
The effect of fetal bovine serum is dose dependent (Fig.
3) and cannot be mimicked by 5 or 20 mg of the serum
attachment factor fibronectin or by 10 mg laminin
(data not shown). Thus, serum is necessary for complete wound closure in this assay and is included at a
concentration of 5% (v/v) in the medium in experiments designed to determine the effect of drugs on
wound closure.
Table 1. Effects of growth modulators on cell
density and wound closure
Modulator
Cell Number*
(X104)
Wound Area\
(mm2)
Control
EGF
EGF + 5FU
1.8 ±0.2
8.9 ± 1.1
1.4 ±0.2
4.3 ± 0.5
4.2 ± 0.4
4.2 ± 0.7
* Determined in triplicate 7 days after innoculation of 1 X I04 cells.
\ Determined in quadruplicate 42 hr after wounding.
with the /3-subunit of cholera toxin (data not shown)
or isoproterenol (10~5 M) had unchanged rates of cyclic
AMP synthesis and wound closure.
Cell Adhesion
After 18 hr in medium containing 3H-leucine, confluent cultures of epithelial cells had incorporated label
at a rate that produced 1-3 counts/min/cell, providing
a simple method for subsequent estimation of the
number of adherent cells. Thirty minutes after addition
of suspended cells to HEPES buffered medium at am-
5--
Effect of Mitosis on Wound Closure
Corneal epithelial mitotic rates were determined in
cultures grown in the presence of an epithelial growth
promoter, EGF, and an antimitotic agent, 5-FU. EGF
at 10 /ig/ml enhanced the cellular mitotic rate; whereas,
5-FU at 10 Mg/ml totally suppressed EGF stimulated
mitosis. Nevertheless, when added to cultures at the
time of wounding, these agents had no effect on wound
closure (Table 1).
Effect of Cyclic AMP Synthesis on Wound Closure
Corneal epithelial cultures were wounded and exposed to test agents for 42 hr, at which time cyclic
AMP synthesis and wound closure were measured.
Cholera toxin (0.1 Mg/ml) and forskolin (10~5 M) increased cyclic AMP synthesis and slowed wound closure as shown in Figure 4. Neither of these agents,
however, completely prevented migration, and substantial closure occurred in all cases. Cultures treated
CON
ISOP
CTX
FS
CON
ISOP
CTX
FS
220
I 200
0
^
180
160
1 140
°> 120
100
a
80
<
o 60|
40
Q
20-
Fig. 4. Effect of cyclic AMP synthesis on wound closure. Cultures
were wounded and exposed for 42 hr to cholera toxin (CTX), isobutylmethylxanthine (IBMX), forskolin (FS), or isoproterenol (ISOP)
added to medium containing 5% FCS. At this time, remaining wound
area and cyclic AMP synthesis were determined. Mean ± SEM
(n = 4).
Downloaded From: http://iovs.arvojournals.org/pdfaccess.ashx?url=/data/journals/iovs/933126/ on 06/14/2017
12
INVESTIGATIVE OPHTHALMOLOGY & VISUAL SCIENCE / January 1986
70T
Minutes
Fig. 5. Effect of cholera toxin on cell substrate adhesion. Epithelial
cell cultures grown in cholera toxin free medium were exposed to
cholera toxin (1 Mg/ml) and labelled with 3H-leucine for 18 hr. Cells
were then suspended and added to replicate tissue culture wells. At
the indicated times, medium and unattached cells were removed and
the adherent cells lysed. Radioactivity of the lysate was determined
as a percentage of the total radioactivity of the cell suspension. Cholera
toxin treated (•), control (O). Mean ± SEM (n = 6).
bient temperature, 30-50% of cells derived from cholera toxin supplemented cultures had adhered to the
plastic substrate; whereas, during the same time only
10-15% of the cells derived from cholera toxin-free
cultures adhered (Fig. 5). When collagen coated wells
were used as a substrate, 30% of cells grown in the
absence of cholera toxin adhered in the 30 min-assay
period. When cells grown in the presence of cholera
toxin were exposed to the collagen substrate, there was
no further increase in adhesion of these cells as compared to cells exposed to the plastic substrate.
Discussion
Superficial epithelial defects close by epiboly, migration of the epithelial cell sheet over the basal lamina.
The corneal epithelium, which consists of stratified
squamous cells and rests on an avascular matrix, is a
particularly simple system in which to study epithelial
migration. Corneal epithelial wound closure has been
studied both in vivo and in organ culture.'' 2A7>8 In the
present study, we have demonstrated that cultured cells
of the corneal epithelium in culture retain the ability
to re-cover a model defect. The rate of wound closure
in vitro in the presence of serum, as determined by
linear regression analysis of the rate curve in Figure 2,
is 0.8 mm 2 /hr and is similar to that observed in experimental wounds in vivo.7-9
For wound closure to occur in vitro, serum is a necessary component of the medium. Recently, a serum
Vol. 27
protein, epibolin, was isolated which, when added to
epidermal organ cultures, promotes migration of the
epithelial sheet.10 Whole serum, added to corneal organ
cultures, promotes migration of the epithelium, and
the attachment factor fibronectin, a serum component,
has been identified in the basement membrane zone
of healing corneas." 12 However, our results demonstrate that the permissive factor for migration in serum
is unlikely to be fibronectin or laminin.
In vivo, wounds close initially by migration of epithelial cells from a peripheral site rather than proliferation within the wound site. The addition of EGF,
which is mitogenic for the cultured epithelial cells, to
the experimental medium has a negligible effect on
closure. Conversely, the antimitotic agent, 5-FU, added
in a concentration sufficient to suppress EGF-induced
mitosis, has no effect on wound closure. We therefore
conclude that migration, and not mitosis, is the dominant process by which wound closure occurs in this
in vitro model.
To cover superficial defects, migratory cells must
continuously make and break cell-substrate adhesions.
The "stickiness" of cells keeps them adherent to their
substrate and prevents inadvertent damage to the healing tissue by mechanical trauma, such as the movement
of the eyelid over the surface of the cornea during
blinking. Only after migration ceases do epithelial cells
re-form stable hemidesmosomal attachments to their
basement membrane. 3
Cell substrate adhesion during migration is probably
dependent upon glycoproteins on the cell surface. In
the cornea, glycoproteins are normally synthesized by
migrating cells, and tunicamycin prevents both glycoprotein synthesis and epithelial migration. 413 The
present study shows that cholera toxin stimulated cyclic
AMP synthesis increases the ability of cultured corneal
epithelial cells to adhere to both plastic and collagenous
substrates. Recent evidence demonstrates that dermal
epithelial cells cultured in the presence of cholera toxin
produce a specific adhesion protein, and that stimulated
cyclic AMP synthesis increases the rate of secretion of
epithelial adhesion glycoproteins.1415 Synthesis of the
glycoprotein fibronectin by full thickness corneas is
enhanced by cyclic AMP. 16 Thus, cholera toxin and
forskolin, via cyclic AMP synthesis, may increase the
secretion of adhesive cell glycoproteins, thereby increasing cell substrate adhesion. This may act to slow
the rate of migration in culture as cyclic AMP, either
added exogenously or synthesized endogenously, slows
migration of several cell types.17 The lack of effect of
isoproterenol is probably due to the inability of this
drug to sustain elevated cyclic AMP levels.
The paradoxical effect of cholera toxin to stimulate
wound closure in vivo may therefore be consistent with
Downloaded From: http://iovs.arvojournals.org/pdfaccess.ashx?url=/data/journals/iovs/933126/ on 06/14/2017
No. 1
CORNEAL EPITHELIAL WOUND CLOSURE / Jumblorr ond Neufeld
its inhibitory effect in the culture model. Originally,
we observed that pretreatment in vivo with topical
cholera toxin produced an increased rate of closure of
superficial corneal defects; whereas, we now report that,
in vitro, cholera toxin slows closure.56 We also observed
earlier that cholera toxin slows wound closure in a full
thickness corneal organ culture model in which the
epithelium migrates over native substrate to close a
superficial defect.8 Thus the differences observed in vivo
and in vitro are unlikely to reflect the sensitivity of
epithelial cells to differing substrates. If cyclic AMP
promotes adhesiveness of migrating cells in vivo, as it
does in vitro, this mechanism will protect responding
cells from mechanical disruption due to blinking and
therefore appear to promote wound closure. In vitro,
this mechanism will slow but not prevent migration.
An alternative explanation for cholera toxin stimulation of wound closure in vivo may include enhanced
epithelial proliferation. Cholera toxin does potentiate
the mitogenic effects of insulin and EGF. 5 Cholera
toxin may potentiate the influences of other cell types
in vivo or mediators that are absent in epithelial tissue
culture. Our data does not permit us to distinguish
between these hypothetical mechanisms.
Cultured corneal epithelial cells provide a useful
model system for investigating the events governing
closure of superficial epithelial defects and for assaying
the actions of exogenous agents on these events. Tissue
culture offers several advantages over in vivo and organ
culture models of wound closure: (1) The extracellular
milieu including nutrients, growth factors and substrate
is easily manipulated; (2) Closure of defects occurs in
the absence of other cell types, such as nerves, mesenchymal fibroblasts, or inflammatory cells; (3) A large
number of cultures, in this case 24, can be derived
from a single cornea; and (4) Cells can be easily labelled
and used for cell-substrate adhesion determinations.
Multiplate cultures of corneal epithelium should prove
useful for investigating the influence of a variety of
modulators, drugs and chemicals on wound closure
under defined conditions. Establishment and routine
assay of corneal epithelial wound closure in multiwell
plates may also prove useful as an in vitro, toxicological
test and an alternative to animal testing.
Key words: cornea, epithelium, adenosine cyclic monophosphate, wound closure, migration, tissue culture
13
Acknowledgments
We thank Gail Raymond and Nancy McLaughlin for their
excellent technical assistance.
References
1. Hanna C: Proliferation and migration of epithelial cells: during
corneal wound repair in the rabbit and the rat. Am J Ophthalmol
61:55, 1966.
2. Kuwabara T, Perkins DG, and Cogan DG: Sliding of the epithelium in experimental corneal wounds. Invest Ophthalmol 15:
4, 1976.
3. Buck RC: Hemidesmosomes of normal and regenerating mouse
corneal epithelium. Virchows Arch (Cell Pathol) 41:1, 1982.
4. Gipson IK, Kiorpes TC, and Brennan SJ: Epithelial sheet movement: effects of tunicamycin on migration and glycoprotein synthesis. Dev Biol 101:212, 1984.
5. Jumblatt MM and Neufeld AH: /3-Adrenergic and Serotonergic
responsiveness of rabbit corneal epithelial cells in culture. Invest
Ophthalmol Vis Sci 24:1139, 1983.
6. Lowry OH, Rosenbrough NJ, Farr AL, and Randall RJ: Protein
measurement with the folin phenol reagent. J Biol Chem 193:
265, 1951.
7. Jumblatt MM, Fogle JA, and Neufeld AH: Cholera toxin stimulates adenosine 3',5'-monophosphate synthesis and epithelial
wound closure in the rabbit cornea. Invest Ophthalmol Vis Sci
19:1321, 1980.
8. Jumblatt MM and Neufeld AH: Effects of cyclic AMP and calcium on corneal epithelial wound closure in vitro. ARVO Abstracts. Invest Ophthalmol Vis Sci 22(Suppl):25, 1982.
9. Jumblatt MM and Neufeld AH: Characterization of cyclic AMPmediated wound closure of the rabbit corneal epithelium. Curr
Eye Res 1:189, 1981.
10. Stenn KS: Epibolin: a protein of human plasma that supports
epithelial cell movement. Proc Natl Acad Sci (USA) 78:6907,
1981.
11. Nishida T, Nakagawa S, Ohasi Y, Awata T, and Manabe R:
Fibronectin in corneal wound healing: appearance in cultured
rabbit cornea. Jpn J Ophthalmol 26:410, 1982.
12. Fujikawa LS, Foster CS, Harris TJ, Lanigan JM, and Colvin RB:
Fibronectin in healing rabbit corneal wounds. Lab Invest 45:
120, 1981.
13. Gipson IK and Kiorpes TC: Epithelial sheet movement: protein
and glycoprotein synthesis. Dev Biol 92:259, 1982.
14. Okamoto Y, Sakai H, Sato J, and Akamatsu N: Effects of dibutyryl cyclic AMP on the synthesis of dolichol-linked saccharides
and glycoproteins in cultured hepatoma cells. Biochem J 212:
859, 1983.
15. Roberts GP and Jenner L: Glycoproteins and glycosaminoglycans
synthesized by human keratinocytes. Biochem J 212:355, 1983.
16. Nishida T, Tanaka H, Nakagawa S, Sasahe T, Awata T, and
Manabe R: Fibronectin synthesis by the rabbit cornea: effects of
mouse epidermal growth factor and cyclic AMP analogs. Jpn J
Ophthalmol 28:196, 1984.
17. Dunlap MK: Cyclic AMP levels in migrating and nonmigrating
newt epidermal cells. J Cell Physiol 104:367, 1980.
Downloaded From: http://iovs.arvojournals.org/pdfaccess.ashx?url=/data/journals/iovs/933126/ on 06/14/2017