Retinal and epiretinalglia-an immunohistochemical study · SUMMARY Immunohistochemical techniques wereusedto examinethe distribution ofcells con-taining glial fibrillary acidic protein

British Journal of Ophthalmology, 1984, 68, 698-707

Retinal and epiretinal glia-an immunohistochemicalstudyP. S. HISCOTT,' I. GRIERSON,' C. J. TROMBETTA,' A. H. S. RAHI,'J. MARSHALL,) AND D. McLEOD3

From the 'Department ofPathology, and 2Department of Clinical Ophthalmology, Institute ofOphthalmology,London, and the 'Surgical Vitreoretinal Unit, Moorfields Eye Hospital, London

SUMMARY Immunohistochemical techniques were used to examine the distribution of cells con-taining glial fibrillary acidic protein (GFAP) in normal and pathological human specimens,including 22 globes (13 of which contained epiretinal membranes 'in situ'), 16 surgically excisedepiretinal membranes, and monolayers of cells obtained from five epiretinal membranes placed intissue culture. The astrocytic cells of normal and pathological retinae stained with the glial-cellmarker, but Muller cells were GFAP-negative in normal retinae at the antisera dilutions used.Muller cells did, however, stain in retinae from glaucomatous eyes and in eyes with prolongedretinal detachment. Electron microscopy did not reveal any obvious morphological differencebetween the intermediate filaments of normal (GFAP-negative) and GFAP-positive Muller cells.Ten of the 13 epiretinal membranes 'in situ', all 16 excised membranes, and three of the fivemonolayers contained glial cells. Purely glial membranes were not associated with retinal puckeringor detachment, while all membranes causing tractional complications had a prominent fibrous,non-glial component. Our findings suggest that glial cells do not contribute significantly to thecontractile forces generated by epiretinal membranes. They may, however, provide a scaffold onwhich other cells proliferate and contract and an anchorage by means of which tangential forces aretransmitted into and through the retina.

Muller cells can be recognised easily by conventionalmorphological and ultrastructural techniques, butother types of retinal glia are far more difficult toidentify. Neuroglial stains, such as the modified delRio Hortega method, have been used to study thedistribution of accessory glial cells in the normalretina,' but these stains lack specificity2 and oftenproduce inconsistent results.34 Fortunately immuno-histochemical techniques are available which employantibodies against structural proteins in the cyto-skeleton of glial cells. One such protein-glialfibrillary acidic protein (GFAP)-is unique to glia.Thus, immunostains developed against this proteinare an extremely accurate and reproducible means oflocating glial cells in neural tissues. Anti-GFAP stain-ing has been used to investigate the distribution ofglia in the normal retina of several species includingman.5-7Correspondence to Dr P. Hiscott, Pathology Department, Instituteof Ophthalmology, 17/25 Cayton Street, London EC1V 9AT.

Retinal glia appear to be a component of cellularproliferations which develop on the retinal surface-epiretinal membranes (ERMs).522 To date, recog-nition of glia in ERMs has been based mainly onconventional histological and ultrastructural criteria.However, the cellular constituents of ERMs areactivated or metaplastic and are displaced from theirnormal location. It is notoriously difficult to identifythe origins of such ectopic cells on morphologicalcharacteristics alone. Since glial cells may haveimportant functional activities in ERMs, it is desirableto identify glia more accurately than has been possiblein the past. Three ERMs removed at surgery formassive periretinal proliferation (MPP) have recentlybeen described after immunohistochemical stainingusing antibodies against GFAP.23 We have used asimilar immunohistochemical approach to identifyglia in a much larger series of specimens includingsections (a) from normal globes, (b) from eyes con-taining ERMs, and (c) from surgically excised

698

on April 19, 2020 by guest. P

rotected by copyright.http://bjo.bm

j.com/

Br J O

phthalmol: first published as 10.1136/bjo.68.10.698 on 1 O

ctober 1984. Dow

nloaded from

http://bjo.bmj.com/

Retinal and epiretinalglia-an immunohistochemical study

membranes. Monolayers of cells grown from ERMsin tissue culture were also studied.

Materials and methods

Controls. Sections of normal human brain and opticnerve served as controls to confirm the specificity ofour technique. In addition sections from two knownoptic nerve astrocytomas were used as standards ineach immunohistochemical procedure. Monolayersof cells grown in tissue culture, including monolayersof retinal astrocytes, various ocular epithelia, andfibroblasts from normal human, bovine, and rabbiteyes, provided further positive and negative controlsfor the immunohistochemical methods employed inthis investigation.Normalglobes. In order to study the distribution of

glia in the normal retina six globes were obtainedfrom the Moorfields eye bank having been removedfrom donors within 24 hours post mortem. Thedonors' ages ranged from 14 to 66 years.

Globes with elevated intraocularpressure. The glialdistribution in the retinae of three globes enucleatedfor anterior segment tumours with secondary eleva-

tion of intraocular pressure were investigated becausesome of our series of pathological globes containingERMs were glaucomatous.

Globes with epiretinal membranes in situ. Thirteenglobes known to contain non-vascularised ERMswere studied. Four of the eyes contained an attachedretina, two had localised retinal puckering, and theremaining seven had a total retinal detachment. Theywere associated with a range of pathological con-ditions (Table 1). Four of these globes had hadelevated intraocular pressure.

Excised epiretinal membranes. Sixteen specimenswere obtained during closed pars-plana microsurgery(Table 2). Ten membranes had been associated withmacular pucker and six were representative of morediffuse proliferations (MPP).

Tissue culture monolayers. Five peeled macularmembranes were placed in tissue culture (Table 2),and monolayers of cells were obtained for immuno-fluorescent investigation by a method previouslydescribed.24

TRANSMISSION ELECTRON MICROSCOPY (TEM)We compared the stability of the GFAP-containing

Table 1 Details of13 globes

Histological appearance Associated pathology Intraocular Summary ofimmunohistochemicalfindingsof retina pressure

No retinal detachment Central retinal vascular occlusion (Fig. 3) Elevated Glia in diffuse ERMs continuous with GFAP-(4 globes) positive retinal cells. Haemomacrophages

presentMultiple intraocular surgery for Chandler's Elevated Glial epipapillary membranesyndrome

Panuveitis and corneal abscess - Epiretinal glial cells continuous with OFAP-positive retinal cells. Inflammatory cellspresent

Endophthalmitis - Epiretinal glial cells continuous with GFAP-positive retinal cells. Inflammatory cellspresent

Localised retinal Central retinal vascular occlusion (Fig. le, f) Elevated Glial layer on retinal side of fibrous ERMpuckering (2 globes) and glial foci in anothcr fibrous membrane

overlying pucker. Glial ERM overlying un-distorted retina

Perforating trauma and sympathetic - Glial layer under and in fibrous ERMs. Glia con-ophthalmia (Figs. Ic, d, and 4a) tinuous with GFAP-positive retinal cells

Total retinal detachment Rhegmatogenous detachment (MPP) - Glia foci in fibrous ERMs overlying retinal(7 globes) folds. Pigmented cells present

Rhegmatogenous detachment (MPP) Elevated Glial layers in and under fibrous ERMs. Pigmentin some glia

Retinal vasculitis - Glia on ILL away from retinal folds and glialfocus in fibrous ERM over retinal fold. Pig-mented cells present

One month aftcr perforating trauma - Fibroglial layered ERM. Glia continuous withGFAP-positive cells in retina

Six weeks after perforating trauma (Fig. 4b) - Fibrous ERM extending from peripheral wound(GFAP-negative)

Endophthalmitis after perforating trauma - Fibrous ERM extending from peripheral wound(GFAP-negative)

Nine days after perforating trauma - Fibroblastic and red cells on retinal surface(GFAP-negative)

699



j.com/

Br J O


ctober 1984. Dow

nloaded from

http://bjo.bmj.com/

P. S. Hiscott, I. Grierson, C. J. Trombetta, A. H. S. Rahi, J. Marshall, and D. McLeod

Table 2 Details of21 excised epiretinal membranes

Specimens Associated pathology No. of No. ofspecimens Note on distribution ofgliaspecimens containing glia

Embedded epimacular Rhegmatogenous retinal 6 6 3 in layers, 2 in foci, 1 layerplusfocimembranes (10) detachment

Perforating trauma 2 2 Cells on ILLBlunt trauma 1 1 Layer of cellsVascular occlusion 1 1 Scattered throughout membrane

Embedded MPP membranes Rhegmatogenous retinal 5 5 2 in layers, 2 in foci, 1 with cells(6) detachment scattered throughout membrane

Retinitis (viral) 1 1 Focus of cellsTissue culture monolayers from Rhegmatogenous retinal 2 1 (2% of total cells glial)

epimacular membranes (5) detachmentPerforating trauma 2 1 (20% of total cells glial)Blunt trauma 1 1 (5% of total cells glial)

intermediate filaments after different fixation andpreparation methods. For this part of the studyportions of retinae from normal eyes were taken (i)after formol-saline fixation, (ii) after glutaraldehydefixation, and (iii) following wax embedding anddewaxing for electron microscopic processing usingthe method described by Jensen.25 In additionportions of retinae from the globes with elevatedintraocular pressures were studied to compare theintermediate filament distribution with that in retinaat normal IOP. Portions ofERMs were also examinedby TEM. Specimens were fixed in 3% bufferedglutaraldehyde, postfixed in 1% osmium tetroxide,dehydrated in graded alcohol, and embedded inAraldite. Sections 1 ,tm thick were taken from theretinal specimens and ERMs and stained withtoluidine blue for light microscopic analysis. Thinsections were stained with uranyl acetate and leadcitrate for TEM (Jeol IOOC).

SCANNING ELECTRON MICROSCOPYPortions of retina from one globe were fixed in 3%buffered glutaraldehyde, postfixed in 1% osmiumtetroxide, dehydrated in graded alcohols and criticalpoint dried (EMscope). They were coated with goldin a sputter coater (EMscope) and examined with anHitachi S520 scanning electron microscope.

IMMUNOHISTOCHEMISTRYSpecimens intended for immunohistochemical studyby the immunoperoxidase method were fixed in 10%formol-saline or 3% glutaraldehyde and dehydratedin graded alcohol for paraffin wax embedding. Weused an immunoperoxidase staining method similarto that reviewed by Mukai and Rosai.26 Sections 10l.m thick of wax-embedded tissue, including controltissues, normal, hypertensive, and membrane-containing globes, and eight of the peeled epiretinalmembranes were dewaxed and trypsinised for

immunoperoxidase staining of glial elements.Endogenous peroxidase activity was blocked with0-5% hydrogen peroxidase in methanol. Afterthorough washing in TRIS buffered saline (TBS), pH7*6, the sections were treated with normal swineserum and then incubated for 30 minutes with anti-GFAP (DAKO Corporation, California, USA)raised in rabbit and diluted 1:300. Control sectionswere processed with omission or substitution of theprimary antibody. After three washes in TBS theywere incubated for 30 minutes with antirabbit anti-body raised in swine. After thorough washing withTBS the sections were incubated with horseradishperoxidase-rabbit antihorseradish peroxidasecomplex (PAP) for 30 minutes and washed for afurther half hour. The complex sites were shownbrown using 3,3 diaminobenzidine tetrahydrochloridewith fresh hydrogen peroxide substrate. The sectionswere counterstained with Mayer's haematoxylin andmounted for conventional and phase-contrast lightmicroscopy.Specimens for study by immunofluorescent staining

were placed in OCT embedding media (Lab-Tekproducts, Illinois) and immediately frozen in liquidnitrogen. Sections 71xm thick were cut in a cryostatfrom eight of the peeled membranes, and portionswere air dried on to glass slides for 20 minutes. Mono-layers of cells grown in tissue culture were fixed in10% formol-saline after the culture media had beenthoroughly washed off with phosphate-bufferedsaline (PBS), pH 7-6. The monolayers were treatedwith 0-025% saponin for 5 minutes and then rinsedwith PBS. Neither the frozen sections nor the cellmonolayers survived the immunoperoxidase tech-nique well. Therefore glial components were demon-strated in these specimens by an indirect immuno-fluorescent technique following a procedure similarto the immunoperoxidase method but omittingtrypsinisation and peroxidase blocking. After incuba-

700



j.com/

Br J O


ctober 1984. Dow

nloaded from

http://bjo.bmj.com/


tion with the primary antibody, sections were washedfor 30 minutes with PBS and then incubated withantirabbit fluorescent antibody. After 45 minutes ofPBS washing, they were mounted in 10% bufferedglycerol and viewed with an epifluorescent micro-scope (Zeiss).

Results

CONTROLSSections of brain showed astrocytic glial cells whichstained brown following immunoperoxidase labellingfor GFAP. Frozen sections of optic nerve and mono-layers of cells in tissue culture showed fluorescence ofonly the glial component after immunofluorescentmarking for glial cells. After omission of the primaryantisera the glial components in the controls failed tostain.

GLIA WITHIN THE RETINAImmunohistochemistry of eye-bank eyes revealedGFAP-positive cells and their processes in the innerlayers of the retina, especially in the nerve-fibre andganglion-cell layers and around blood vessels (Fig.

Fig. Ia

Fig. Id

la). At the optic disc positively staining tissue con-tinued from the inner retina into the optic nerve. Inthe peripheral retina positively staining cells werefound extending from the internal to the externallimiting membranes, especially where post-oralcystoid changes were present. Elsewhere, however,Muller cells and their processes failed to stain at theconcentration of antisera used. Formalin fixation gavethe best intensity of stain, but glutaraldehyde fixedtissue also gave good differential staining with asimilar distribution to that of the formalin fixed tissue.

In the three eyes with elevated intraocular pressuresecondary to anterior segment tumours the accessoryglia in the inner layers of the retina had an enhancedstaining reaction (when compared with that in thenormal globes). In addition, the whole network ofMuller cells and their processes, from external tointernal limiting membranes, was positively stainedthroughout the retina (Fig. lb). By electron micro-scopy the 10 nm intermediate filaments did not varyin either size or numbers with the various types offixatives employed (Fig. 2a, b), and the filamentswere sufficiently stable to survive the process ofdewaxing and embedding for TEM (Fig. 2c).

Fig. b Fig. Ic

Fig. le. a

Fig. I c Fig. IfFig. I Light micrographsfrom sections stained with the immunoperoxidase methodfor GFA P and counterstained withhaematoxylin (IP and H) or haematoxylin and eosin (H and E). (a) Normalhuman retina. Positive cells (brown reactionproducts) arefound only in the inner retina. (IP and H, x 137). (b) Retinafrom an eye with long-standing glaucoma. Thestaining in the inner retinal layers is more intense than in Fig. Ia and the Muller cells are positively stained. (IPand H, x 137).(c) A retinalpucker. Note the positive staining Muller cells in the retina and the overlyingfibroglial epiretinal membrane wit/iits predominantly GFA P-negativefibroblastic component. (IP and H, x27). (d) Glial cells within an epiretinal membrane.Note thefilamentous staining pattern in the cells' cytoplasm. (IPand H, x687). (e) Retina with an overlying epiretinalmembrane. The glial component in the apparentlyfibrous membrane is not clearly demonstrated. (H and E, x275). (f) Thesame specimen as Fig. le). A focus ofglial cells within the epiretinal membrane is now apparent afterstainingforGFA P.(IPand H, x275).

701



j.com/

Br J O


ctober 1984. Dow

nloaded from

http://bjo.bmj.com/

P. S. Hiscott, I. Grierson, C. J. Trombetta, A. H. S. Rahi, J. Marshall, and D. McLeod

Fig. 2 Transmission electron micrographs of: (a) Intermediatefilaments in Muiller cellsfrom a retina which wasfixed in3%buffered glutaraldehyde. (x82350). (b) Intermediatefilaments ofMuller cellsfrom a retina which wasfixed in 10% formolsaline. (x82 350). (c) Intermediatefilaments ofMuller cells in 10%formol salinefixed retina, after dewaxingandre-embedding in Araldite. (x53 530). (d) Glial cell in epiretinal membrane. The cell contains intermediatefilaments. From aspecimen which had beenfixed in 3% buffered glutaraldehyde. (x 16 470).

Although the immunoperoxidase staining reactionwas positive in Muller cells of glaucomatous eyes,there was no obvious increase in the numbers offilaments nor any change in filament appearance atTEM in these cells.

In all 13 globes containing ERMs the Muller cellswere positively staining to GFAP at the antisera

dilution used (Fig. ic). This staining pattern wassimilar to that seen in the three globes with elevatedpressure secondary to anterior segment tumours.

GLIA WITHIN EPIRETINAL MEMBRANESEnucleated globes. Immunohistochemical analysisrevealed that the four epiretinal membranes which

702



j.com/

Br J O


ctober 1984. Dow

nloaded from

http://bjo.bmj.com/

Retinal and epiretinal glia-an immunohistochemical study

Fig. 3 Scanningelectron micrographs ofan epiretinal membrane: (a) lnterdigitatingglialcellprocesses incompletely coverthe retinal surface, leavingpatches of 'bare' inner limiting lamina. (x2470). (b) Glia completely cover the retinalsurface inanother area. Glial cells are overlapping one another (arrowheads). (x2880). (c) and (d) Glial cellprocesses extend throughdefects in the inner limiting lamina (arrows). Note that the retinal surface is not distorted (asterisks). (x2470 and x 4940).

were not associated with retinal distortion or detach-ment (Table 1) essentially consisted of glia only orglia plus inflammatory cells. In three of these eyesprocesses of GFAP-positive epiretinal cells wereidentified occupying breaks in the inner limitinglamina (ILL). Scanning electron microscopy of onesuch eye showed a sheet of epiretinal glial cells with

multiple extensions through the ILL (Fig. 3a-d).The two epiretinal membranes associated with

retinal puckering and four of the seven membranesover detached, folded retina had a glial component(Fig. 1c). However, all these membranes also con-tained a prominent fibrous component-that is, fibro-blast-like, GFAP-negative cells embedded in collagen

703



j.com/

Br J O


ctober 1984. Dow

nloaded from

http://bjo.bmj.com/

P. S. Hiscott, 1. Grierson, C. J. Trombetta, A. H. S. Rahi, J. Marshall, and D. McLeod

-at the site of membrane contraction (Fig. lc, e, f);inflammatory cells and pigment-containing cells werealso identified. The glial elements of these combinedfibrous and glial membranes ('fibroglial membranes')were sometimes distributed as layers (Fig. lc). Theglial tissue was typically sandwiched between thefibrous component and the ILL; in one eye, however,the glial layer was on the vitreal surface of the fibroustissue. Alternatively, the glial component was distri-buted in small foci which could be demonstratedclearly only by GFAP staining (Fig. le, f). The glialcomponent of two of the six fibroglial membraneswas observed to be in continuity with GFAP-positivecells in the retina via defects in the ILL (Fig. 4a).

In the three eyes with no demonstrable glial com-ponent the ERMs were composed of fibroblast-likecells, inflammatory cells, and pigment-containingcells (Fig. 4b).

Peeled epiretinal membranes. Immunohisto-chemical study of sections from surgically excisedmembranes revealed that all 16 specimens had a glialcomponent (Table 2). It was not possible to subtype

the glia immunohistochemically, but electron micro-scopy showed the presence of typical fibrous astro-cytes with their distinctive intermediate filaments(Fig. 2d). The GFAP-positive cells in thesemembranes were again distributed as foci or layers(Fig. 4c-f) except in four membranes with isolatedcells on the ILL or scattered throughout the specimen(Table 2). However, a GFAP-negative componentcomprising fibrous tissue (i.e., fibroblast-like cellsand collagen) was observed to occupy the bulk of thematerial in all specimens. Inflammatory and pig-mented cells were also present in variable numbers.Monolayers in tissue culture. Three of the five

monolayers of cells derived from peeled membranescontained GFAP-positive cells (Table 2), but theseglial cells represented a minority (between 2 and20%) of the cells in the cultures. The positively stain-ing cells had a variety of shapes, including spindle-shaped cells (Fig. 5), epithelium-like (plate-like) cells,and cells with multiple, branching processes. How-ever, the majority of the cells had the morphologicalcharacteristics of macrophages and fibroblasts and

Fig. 4c

*F:g. 't f:'

Fig. 4eFig. 4d.'S~~~~~~~''~~'

Fig. 4fFig. 4 (a) Positive phase-contrast micrograph ofretina with an overlying epiretinal membrane, stained with theimmunoperoxidase methodfor GFA P and counterstained with haematoxylin. Note the glial component ofthefibroglialmembrane continuous with intraretinalglia. (x 137). (b) Positive phase-contrast micrograph ofan epiretinal membraneassociated with a retinalpucker, stained with the immunoperoxidase methodfor GFAP and counterstained withhaematoxylin. There is no glial component in this membrane. (x 137). (c) Immunofluorescence staining pattern ofafrozensection ofa peeled epiretinal membranestainedfor GFA P. Note the layer ofglial cells in thepredominantly non-glialmembrane. (x 110). (d) Higherpower view ofsame specimen as plate 2(c). The layer ofglia can be seen juxtaposed to thenon-glial tissue. (x 412). (e) Light micrographfrom a section ofa peeled epiretinal membranestainedfor GFAP with theimmunoperoxidase method and counterstained with haematoxylin. Foci ofglia arepresent, but the bulk ofthe tissue isnon-glial. (x69). (f) Light micrograph from a section ofa peeled epiretinalmembranestainedfor GFAP with theimmunoperoxidase method and counterstained with haematoxylin. The glial component is moreprominent than in thespecimen seen in Fig. 2(e), and isforming layers. (x69).

704



j.com/

Br J O


ctober 1984. Dow

nloaded from

http://bjo.bmj.com/


were GFAP-negative. Similarly, the two monolayerswithout a demonstrable glial component consisted offibroblasts and macrophages.

Discussion

GFAP was initially isolated from cerebral tissue,27and there is good evidence that this protein is specificto glial cells.21 In addition, GFAP forms 10 nmfilaments in vitro, indicating that it is the major sub-unit of the intermediate filaments of astroglia.2"However, not all glial cells can be stained with animmunohistochemical technique using antibodies toGFAP-for example, cerebral astrocytes from new-born rats-" and the radial fibres of Muller in rat retinaonly stain following injury.6 Positively staining cellshave been found in the inner layers of the retina, atthe ora serrata, and in optic nerve head of the mouse,but Muller cells do not stain.5 Similarly, normalhuman Muller cells are GFAP-negative7 despite thefact that they contain 10 nm filaments."'Our study of normal human globes confirms that

astrocytic cells and their processes in the inner retinallayers stain positively for GFAP, but Muller cells donot stain (at the concentration ofantisera used) exceptat the ora serrata. By contrast, eyes with elevatedintraocular pressure, retinal detachment, or an in-flammatory reaction involving the vitreal cavity showpositive staining of Muller cells. Indeed, it is difficultin these specimens to distinguish Muller cells fromastrocytes and their processes, although perivascularglial cells appear to stain most intensely. Foos andGloor showed that, during the healing of experi-mental retinal scars, the cytoplasmic filaments ofastrocytic cells progressively increase in number,32and this probably explains the increase in staining ofastrocytes in our pathological specimens comparedwith normal eyes. However, such a mechanism doesnot explain the difference in staining pattern ofMullercells. Presumably either an unmasking process ormacromolecular reorganisation causes the Mullercell filaments to stain positively, but this remainsspeculative.Our immunohistochemical studies of both epi-

Fig. 5 An areafrom a cultured monolayershowing a spindle-shaped cell which is positive after immunofluorescence stainingfor GFA P. A filamentous stainingpattern can be seen in the cell's cytoplasm. Surrounding cells are GFA P-negative andtherefore cannot be seen (x 1090).

705



j.com/

Br J O


ctober 1984. Dow

nloaded from

http://bjo.bmj.com/

P. S. Hiscott, 1. Grierson, C. J. Trombetta, A. H. S. Rahi, J. Marshall, and D. McLeod

retinal membranes 'in situ' and surgically excisedspecimens confirm that glial cells contribute to epi-retinal membranes associated with a wide range ofpathological disorders. We were also impressed bythe similarity ofdistribution of glial elements betweenpeeled and 'in situ' specimens. It is unlikely, there-fore, that the glial cells demonstrated in peeled ERMshad been avulsed from the retina during surgery,though it is recognised that portions of ILL andMuller's end-feet are often present in suchspecimens.33The specificity of the GFAP marker ensures that,

when staining occurs, it is a reliable indicator of thepresence of glial cells in the tissue. Cells containingpigment are a potential hazard during interpretationof immunoperoxidase staining with DAB (where pig-ment can be confused with the brown reactionproduct) and immunofluorescent staining (where pig-ment autofluorescence may occur). However, thestaining pattern for GFAP-positive cells tends to befilamentous or homogeneous, and we found this easyto differentiate from the granular pigment.

Since glia may assume a different structuralappearance (metaplasia) when displaced from theirnormal location, morphological criteria are ofteninadequate to determine their true presence orabsence at the abnormal site. Epiretinal membranesfrequently contain such metaplastic cells which intissue culture may adopt an even greater variety ofmorphological appearances, including spindle-forms.24 Nevertheless, the glia still retain their GFAPcontent whatever their morphology in tissue culture.It is probable, therefore, that all glia-derived cells inERMs would be GFAP-positive and that negatively-staining cells in ERMs are not of glial origin.

FUNCTIONAL SIGNIFICANCE OF GLIA IN ERMSThe role of the glial component of ERMs is open tospeculation. Pure glial epiretinal membranes havebeen described in human globes by Foos,9 who calledthem 'simple' epiretinal membranes because theyproduced only subtle lesions and no symptoms.Furthermore, injection of retinal astrocytes into therabbit vitreous,3 while resulting in collagen synthesis,does not lead to retinal detachment, although otherexperimental work suggests that glial cells may beresponsible for full-thickness retinal folding.35

In our study the four globes without tractionalretinal complications had only glia or glial plus in-flammatory cells in the membranes. Every membraneassociated with a retinal pucker or detachment had afibroblastic (GFAP-negative) component. Indeed, inthe fibroglial membranes the bulk of the tissue wasnon-glial, and some of the membranes associatedwith such complications had no glial element. Ourfindings do not, therefore, support the concept that

purely glial membranes can cause a severe tractionalevent such as macular pucker or more widespreadfull-thickness retinal folding. Although purely glialmembranes may be associated with wrinkling of theILL,"' we believe it is the non-glial tissue whichprovides most of the tractional force in MPP andmacular pucker. Indeed, in at least one of our globespredominantly fibrous tissue was associated withretinal puckering, while purely glial portions of themembranes overlay undistorted retina.The distribution of the GFAP-positive cells in some

ERMs suggests an alternative role for glia in ERM's.These cells often form a layer on the retinal side ofthe membrane upon which non-glial fibrous tissueadheres as a second lamina. The glial component inthese fibroglial membranes may thus act as a scaffoldon which other cells, such as perivascular fibroblasts,could migrate, proliferate, contract, or producecollagen. Immunohistochemical staining also high-lights the continuity established between the glialcomponent of an ERM and the glial 'sustentaculararchitecture' of the retina via dehiscences in the ILL.It is tempting to speculate that this might provide aconsiderable anchorage by means of which tangentialtraction forces on the retinal surface may be trans-mitted through the retina to produce full-thicknessretinal folding.33Whatever the capabilities of glial cells within

ERMs, the frequency of their occurrence in theseextraretinal scars suggests that they play an importantrole in membrane pathobiology. Indeed, control ofglial proliferation may be one key to future pharma-cological manipulation of epiretinal membranes.

We thank Mr R. J. Cooling, Mr P. K. Lcaver, Mr Z. Grcgor, and MrA. H. Chignell, who providcd many of the surgical spccimcns, andProfcssor W. R. Lee, who supplicd sevcral of the wholc eyes.Technical assistance was providcd by Mr R. C. Howes, Miss E.

Robins, MrJ. Prasad, and Mr R. A. Alcxander. Sccretarial assistanccwas providcd by Mrs P. Goodwin and Miss K. C. Betts.Our rescarch has been fundcd by the TFC Frost Charity Trust, the

Moorfields Locally Organised Rcscarch Scheme, the MoorfieldsEndowment Fund, the Wcllcomc Trust, and the MuirheadSettlement.

References

1 Woltcr JR. The astroglia of the human retina and other glialclemcnts of the rctina undcr normal and pathologic conditions.Am J Ophthalmol 1955; 40: 88-99.

2 Lessells S, Kuwabara T. Rctinal ncuroglia. Arch Ophihalmol1963; 70:671-8.

3 Ogden TE. Nervc fibre layer astrocytes of the primatc rctina:morphology, distribution and dcnsity. Itivest Ophthalmol VisualSci 1978; 17: 499-51(0.

4 Jiang Q, Fratkin JC, Blodi FC. Localization of glial fibrillaryacidic protein in retinal astrocytoma: an immunohistochcmicalstudy. Curr Eye Res 1983; 8: 523-7.

706



j.com/

Br J O


ctober 1984. Dow

nloaded from

http://bjo.bmj.com/

Retinal and epiretinal glia-an immunohistochemical study

5 Bromhcrg JS, Schachncr M. Localization of ncrvous systematntigcns in retina hy immunohistochemistry. Invest OphthalmnolVisual Sci 1978; 17: 920-4.

6 Bignami A, Dahl D. The radial glia of Midler in thc rat rctina andtheir rcsponsc to injury. An immunofluorcsccncc study withantihodies to glial fihrillary acidic (GFA) protcin. E,rp Eye Res1979; 28: 63-9.

7 Kumpulatinen T, Dahl D, Korhoncn LK, Nystr6im SHM.Immunolabhelling of catrhonic anhydrasc isocnzymc C and glialfibrillary acidic protcin in paraffin-cmheddcd tissuc sections ofhuman hraiin and rctina. J Histochetn Cvtochein 1983; 31: 879-86.

8 Manschot WA. Pcrsistent hyperplastic primairy vitrcous: specialrefcrcncc to prcrctinal glial tissuc ais a pathological chairatctcristicand to the dcvelopmcnt of the primairy vitrcous. Arch Oph-tli(1ltol|11958; 59: 188-21)3.

9 Foos RY. Vitrcorctinali juncturc-simpic cpirctinail mcmbrancs.Albrecht von Graefes Arc/h Klitn Ophthaibnol 1974; 189: 231-5(0.

1() Bcllhorn MB, Friedman AH, Wisc GN, 1-cnkind P. Ultra-structurc aind clinicopathologic corrclation of idiopathic prc-retinal maicular fibrosis. Ain J Ophthalhnol 1975; 79: 366-73.

11 Kcnyon KR, Pcderson JE, Grecn WR, Maumcnec AE. Fibro-glial prolifcration in pairs planitis. Tratis Ophthlalunol Soc UK1975; 95: 391-7.

12 Rcntsch FJ. The ultrastructurc of prcrctinall macular fihrosis.Alhrecht von Graefes A rch Klitn Opihthalunol 1977; 202: 32 1-37.

13 Kcnyon KR, Michels RG. Ultrastructurc of cpirctinal mcmbrancrcmoved hy pars plana vitrcorctinal surgcry. Ain J Oplthhalmol1977; 83: 815-23.

14 Vain Hiorn DL, Aahcrg TM, Machcmer R, Fcnzl R. Glial ccllprolifcrattion in human rctinail dctachmcnt with massivc peri-retinal prolifcration. An J Olphthalmnol 1977; 84: 383-93.

15 C(larkson JG, Green WR, Massof D. A histopaithic review of 168cascs of prcretinal mcmbranc. Ain J Oplitaliltnol 1977; 84: 1-17.

16 Grccn WR, Kincaid MC, Michels RG, Pcderson JE, KenyonKR, Maumenec AE. Pars planitis. Trans Ophthalmnol Soc UK198 1; 101: 361-7.

17 Hlarada T, Chauvaud D, Pouliqucn Y. An cicctron microscopicstudy of cpirctinal mcmbranc of human cyes. Albrecht von

Graefes Arch Kliti ()phthalmnol 1981; 215: 327-39.18 Kaimpik A, Kcnyon KR, Michcls RG, Grccn WR, dc la Cruz

ZC. Epirctinal and vitrcous membrancs. Comparative study of56 cases. A rc/h Ophthialnol 198 1 ; 99: 1445- 54.

19 Kampik A, Grecn WR, Michels RG, Ricc TA. Epirctinalcmcmbraincn nach photokoagulation (postkoagulativc maculo-paithic). Ber Dtsch Ophthalmnol Ges 1981; 78: 593-8.

2(0 Szamier RB. Ultrastructurc of the prerctinal mcmbranc inrctinitis pigmcntosat. Inve.st Ophthalmiol Visual Sci 1981;21:227-36.

21 Hiamilton CW, Chandler D, Klintworth GK, Machemer R. Atransmission and scianning ccctron microscopicall study ofsurgically cxcised prcrctinal mcmbranc prolifcrations in diabctesmcllitus. AinJ Ophthalhnol 1982; 94: 473-88.

22 Baincs PS, Hiscott PS, McLcod D. Postcrior non-vascularizcdprolifcraitivc cxtrairctinopaithy and pcriphcral nodular retinaltelangicctasis. Tratis Ophthialmnol Soc UK 1982; 102: 487-91.

23 Rodrigucs MM, Ncwsomc DA, Machcmer R. Further char-actcrization of cpiretinal mcmbrancs in human maissivc peri-rctinal prolifcraition. Curr Ekve Re.s 198 1; 6: 311-5.

24 Hiscott PS, Gricrson 1, Hitchins CA, Rahi AHS, McLcod D.Epirctinal mcmbrancs in vitro. Trans Ophthaltniol Soc UK 1983;103: 89-10(2.

25 Jcnscn OA. Prcparation of paraffin-cmbcddcd tissuc for ccctronmicroscopy. Exp Eye Res 1974; 18: 417.

26 Mukai K, Rosai J. Applications of immunoperoxidasc tcchniqucsin suirgical pathology. In: Fcnoglio CM, Wolff M, cds. Progressi.n surgical path/ology. Ncw York: Masson Publishing, 1980(1 15-49.

27 Eng LF, Vanderhaeghen JJ, Bignami A, Gcrstl B. An acidicprotcin isolatcd from tibrous astrocytes. Brraii Res 1971; 28:351-4.

28 Bock E. Ncrvous system specific protcins. J Neurochemn 1978; 30:7-14.

29 Rucger DC, IIuston JS, DathI D, Bignami A. Formation of 1(1) Afilamcnts from purificd glial fibrillary aicidic protcin in vitro. JMol Biol 1979; 135: 53-68.

3(0 Bignami A, Dahl D. Astrocytc-spccitic protcin and radial glia in

the ccrchral cortex of newhorn rat. Nature 1974; 252: 55-6.31 Hogan MJ, Alvarado JA, Weddcll JE. Histology of the human

elve-an atla.s and textbook. Philadelphia: Saunders, 1971: 393-522.

32 Foos RY, Gloor BP. Vitrcorctinal juncturc: healing of cxpcri-mcntal wounds. Albrecht von Graefes Arch Klitl Ophitialmnol1975; 196: 213-3(0.

33 McLcod D, Marshall J, Gricrson 1. Epimacular mcmbranc pecl-ing. Tratis Opithialtnol Soc UK 198 1; 101: 17(0-80(.

34 Burkc JM, Kowcr HS. Collagcn synthcsis hy rahbit ncural retinain vitro and in vivo. Exp E,ve Re.s 1980(; 31: 213-26.

35 Laiqua 11, Machcmer R. Gliall cell prolifcration in retinal dctach-mcnt (massivc pcrirctinail prolifcration). An]J Opphthaltnol 1975;80: 6(02-18.

707



j.com/

Br J O


ctober 1984. Dow

nloaded from

http://bjo.bmj.com/

Documents

Retinal and epiretinalglia-an immunohistochemical study · SUMMARY Immunohistochemical techniques wereusedto examinethe distribution ofcells con-taining glial fibrillary acidic protein