Fluorophore-Assisted Retinal Break Detection Using Antibodies to Glial Fibrillary Acidic Protein

Abstract

purpose. To evaluate the role of fluorescent antibodies as a means of enhancing the detection of retinal breaks during vitrectomy for rhegmatogenous retinal detachment.

methods. In ex vivo studies, unfixed, porcine retinal flatmounts were incubated with Cy3 anti-GFAP. Experiments were repeated in the presence of excess soluble GFAP and after surface excimer laser ablation through the internal limiting membrane, into the Müller cell foot processes. Tissue was also incubated with trypan blue, and cross-species immunoreactivity was determined in bovine, rabbit, and human retina. In vivo studies were conducted in a porcine model of rhegmatogenous retinal detachment. Cy3 anti-GFAP was injected into the vitreous cavity of eyes with retinal breaks and then rinsed from the eye. Barrier filters were fitted to the operating microscope to allow intraoperative visualization of tissue stained with Cy3. Excitation endoillumination was provided by a 532-nm diode-pumped laser.

results. In ex vivo studies, retinal flatmounts exposed to Cy3 anti-GFAP showed minimal surface fluorescence, but exposed glial elements at the cut edge of the flatmount stained brightly, as did those exposed by excimer ablation of the Müller cell membrane. Blocking studies confirmed that binding was antigen specific. Trypan blue colocalized to the cut edge of retinal flatmounts. All species showed high levels of immunoreactivity except rabbit. In vivo studies demonstrated selective intraoperative staining of retinal breaks with a high level of specificity.

conclusions. Intraoperative vital staining of retinal breaks is possible in an animal model of retinal detachment. Ex vivo studies indicate that this occurs because the Cy3 anti-GFAP selectively binds the intermediate filaments of glial cells with damaged or destroyed cell membranes.

Rhegmatogenous retinal detachment (RRD) occurs when a full-thickness retinal defect allows vitreous fluid to pass into the subretinal space, separating the retina from the underlying retinal pigment epithelium (RPE). Over the past 20 years, there have been notable advances in the treatment of RRD including the widespread introduction of pars plana vitrectomy (PPV), indirect surgical viewing systems, highly purified silicone oil, gas tamponade, and heavy liquids. Despite these innovations, the success rate for primary repair of RRD remains the same as it was 20 years ago, ^{1 2} with 6% to 23% ^{1 2 3 4 5 6 7 8 9} of primary operations failing to reattach the retina. Failed primary RRD repair is associated with serious complications, such as proliferative vitreoretinopathy ^{10 11} and results in a reduced visual prognosis. ^{6 12}  

One of the commonest causes of failed primary RRD surgery is failure to identify all the full-thickness defects in the retina, ^{2 13 14} defined clinically as missed retinal breaks. Careful fundal examination by an experienced vitreoretinal surgeon remains the best method of detecting retinal breaks, but even in this setting, there are cases in which no break can be identified before or during surgery. ¹⁵ A technique that improves retinal break detection would therefore represent a useful advance in vitreoretinal surgery. 

The ultimate purpose of this study was to develop a vital stain that could be used to enhance retinal break detection. There are several structures that could be targeted by such an agent. Retinal breaks represent a discontinuity in the retinal surface and expose outer retinal structures to the vitreous cavity. Hence, the interphotoreceptor matrix could be targeted with an intravitreous lectin and the RPE with antibodies to pancytokeratin. Fresh retinal breaks contain an annulus of dead and damaged cells ¹⁶ that could be targeted by dead cell markers such as trypan blue. 

To avoid widespread staining of detached retinal tissue, in the present study we sought to target exposed structures at the edge of retinal breaks. In particular, we elected to target exposed glial cells with an antibody to glial fibrillary acidic protein (GFAP). Anti-GFAP targets the 10-nm diameter intermediate filaments ¹⁷ present in retinal glia. Using a commercially available fluorophore-tagged anti-GFAP, we sought first to demonstrate that it is possible to target retinal breaks selectively, and second to characterize the nature of this selective staining. 

Methods

Ex Vivo Studies

Preparation of Antibody.

Commercially available, Cy3-tagged, monoclonal, mouse anti-pig, anti-GFAP (Sigma-Aldrich, Poole, UK) was dialyzed to exchange the azide-containing vehicle with an isotonic solution designed to be nontoxic to the retina. Regenerated-cellulose dialysis tubing was prepared to remove any potentially toxic material, as recommended by the manufacturer (BioDesign Inc., Poughkeepsie, NY). Aliquots (0.5 mL) of Cy3 anti-GFAP were dialyzed in sterile, phosphate-buffered saline (PBS; sodium chloride, 120 mM; potassium chloride, 2.7 mM; phosphate buffer salts, 10 mM [pH 7.4] at 25°C; Sigma-Aldrich), and then overnight in 500 mL of sterile intraocular irrigating solution (Chauvin Pharmaceuticals Ltd., Essex, UK). Samples were removed before and after dialysis, and the absorbance at 552 nm was determined with a spectrophotometer (UV-160A; Shimadzu, Kyoto, Japan). The postdialysis concentration of Cy3 anti-GFAP was thereby calculated to be 0.76 ± 0.02 mg/mL (SD; range, 0.74–0.77). The dialyzed antibody underwent 175 Gy γ-irradiation (Nordion Int. Inc., Kanata, Ontario, Canada). 

Determination of Exposure Times and Dilution.

Eyes were obtained from mixed-breed large-white-type pigs aged 36 to 40 weeks, killed at a local abattoir, and transported on ice. Experiments were conducted within approximately 6 hours of enucleation. Isolated, unfixed, peripheral, retinal flatmounts were incubated with Cy3 anti-GFAP diluted 1:12, 1:25, 1:100, 1:400, and 1:800 with PBS for 30 minutes at 37°C in moist, covered chambers. Tissue was then rinsed three times in PBS and mounted on a plain glass microscope slide using a glycerol-based mounting-medium (Citifluor Ltd., London, UK). Retinal flatmounts were viewed and photographed (Ektachrome 320T; Eastman Kodak, Rochester, NY) using a fluorescence microscope with a 546/580-nm filter set (Leitz, Wetzlar, Germany). Experiments were repeated with exposure times of 1, 2, 5, 30, 45, and 95 minutes using 1:12 and 1:25 Cy3 anti-GFAP. Peripheral retina from all quadrants was compared with tissue obtained from the posterior pole. Experiments were repeated in retinas that had been torn, to determine whether this injury produced a different pattern of staining to the cut tissue at the edge of the retinal flatmount. Full-thickness retinal tears were created with a 27- or 30-gauge needle in unfixed tissue. 

To assess the degree of interspecies variability in immunoreactivity, isolated retinal flatmounts were obtained from eyes of Friesian cows killed at a local abattoir and transported on ice, and from New Zealand White rabbits killed by other researchers working in the same institute as the authors. Rabbits that had undergone experiments that might affect ocular tissue were excluded. Human tissue was obtained from three human donor eyes (Keratec Eye Bank; St. Georges Hospital, London, UK) with a mean time from death to enucleation of 27 ± 7 hours (SD; range, 20–34) and mean age of 77.7 ± 5.5 years (SD; range, 72–83). Retina was incubated with Cy3 anti-GFAP, as per porcine tissue. 

Control Experiments.

Blocking studies were conducted to determine whether the pattern of fluorescence was antigen specific. A 1:25 dilution of Cy3 anti-GFAP was incubated for 30 minutes at 37°C with excess (10:1 molar ratio of antigen to antibody) soluble GFAP (Calbiochem, La Jolla, CA), serial dilutions of GFAP (1:1, 1:10, 1:100, and 1:1000 Ag:Ab), and with PBS. Porcine retinal flatmounts were exposed to these solutions for 30 minutes at 25°C. The degree of fluorescence was assessed by an observer who was unaware of the ratio of antigen to antibody. 

To confirm that anti-GFAP was targeting glial elements, ex vivo experiments were repeated using another glial marker, Cy3-tagged anti-vimentin (Sigma-Aldrich). The pattern of fluorescence obtained with anti-vimentin was compared with that from tissue stained with anti-GFAP. 

Experiments were designed to determine whether the pattern of fluorescence was due to nonspecific binding of the Cy3 fluorophore. Isolated, unfixed retina from enucleated porcine eyes was incubated with the Cy3 fluorophore alone, using the same protocol as the studies using the Cy3 anti-GFAP conjugate. Cy3 monofunctional reactive fluorophore was obtained as an N-hydroxysulfosuccinimide (NHS) ester (Amersham Pharmacia Biotech, Bucks, UK). The NHS ester was hydrolyzed by dissolving the dye in 0.1 M potassium hydroxide for 1 hour and subsequently adjusting the pH to 7.2 with hydrochloric acid. The absorption of the Cy3 fluorophore at 552 nm was measured using the spectrophotometer. A standard curve of the Cy3 anti-GFAP conjugate was also determined. The Cy3 fluorophore concentration was then adjusted with PBS to match the absorption of the Cy3 anti-GFAP conjugate at 1:3, 1:15, and 1:30 dilutions. 

Experiments were conducted to determine whether the binding pattern of Cy3 anti-GFAP colocalized with that of damaged cells. Trypan blue (0.01%) was used to identify devitalized tissue. ¹⁸ Experiments were conducted on unfixed flatmount retina, as per Cy3 anti-GFAP. Cy3 anti-GFAP experiments were repeated with tissue fixed for 30 minutes in 4% paraformaldehyde, to prevent active cell processes such as endocytosis and to determine whether this altered the binding pattern. 

Excimer Laser Retinal Ablation.

Experiments were designed to test the hypothesis that glia with damaged or destroyed areas of cell membrane may selectively stain with Cy3 anti-GFAP, if the antibody thereby has easier access to its intracellular target. The hypothesis was tested by comparing the staining pattern of intact retina with retina that had undergone excimer laser surface ablation through the Müller cell membrane, into the cell cytoplasm. 

Preliminary studies were conducted to determine the mean thickness of midperipheral ILM and retina. Semithin glutaraldehyde-fixed sections of nontapetal, bovine retina were stained with toluidine blue and the mean thickness was measured with a calibrated graticule on a light microscope, taking the mean of five readings from five sections. The mean retinal ablation per laser pulse was determined by full-thickness ablation of five midperipheral paraformaldehyde-fixed retinal flatmounts. 

A full-thickness button (retina to sclera) was obtained with a 4-mm biopsy punch (Stiefel Laboratories Ltd., Bucks, UK). This was fixed for 7 minutes in 4% paraformaldehyde and then transferred to hardened, PBS-soaked, filter paper (Whatmann International, Ltd., Maidstone, UK) in a Petri dish. The specimen was centered, inner limiting membrane (ILM) uppermost, using the HeNe aiming beam, and surface fluid was removed with a cellulose sponge (J. Weiss & Sons Ltd., Milton Keynes, UK). Retinal ablation was undertaken using a phototherapeutic keratectomy program on an excimer laser (Apex Plus; Summit Technology, Boston, MA), with a pulse repetition rate of 10 Hz and an irradiance emission of 180 mJ/cm². A 2.5-mm diameter, 3-μm-deep ablation was performed, the diaphragm was closed to 1.7 mm, and a further 10 μm was ablated. A central 1-mm diameter, full-thickness ablation was then performed. Specimens were then exposed to Cy3 anti-GFAP as in previous experiments. 

In Vivo Studies

Surgical Methods.

All procedures were approved by the Secretary of State (Home Office, Animal Procedures Section) and adhered to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Operations were performed by a single surgeon (TLJ) using a sterile surgical technique when appropriate. RRDs were created in seven, 10-week-old, mixed-breed, large-white-type pigs by anesthetic and surgical techniques described elsewhere. ¹⁹ Briefly, intubated and ventilated pigs underwent a full three-port pars plana vitrectomy with mechanical separation of the posterior hyaloid face from the ILM. The crystalline lens was left in situ. A subretinal canula was used to create an RD by injecting 2.8 mg/mL sodium hyaluronate viscoelastic (Pharmacia & Upjohn, Milton Keynes, UK) into the subretinal space. ^{20 21} Retinal breaks were then created at the apex of the RD by needle puncture, the vitrectomy cutter, or end-gripping retinal forceps. 

Cy3 anti-GFAP was introduced into the vitreous cavity, either at the time of initial surgery to create the RRD (day-0 group), or as a follow-up operation 1 or 7 days later. Hence, eyes that were followed up underwent two operations: the first to create an RRD, the second to undertake vital staining. Retinal breaks were created at initial surgery and also at the second operation, immediately before vital staining. Therefore, at the time of vital staining, there were both fresh and chronic breaks in each RRD. In the day-1 and -7 follow-up groups, fresh retinal breaks were also created in attached retina, immediately before vital staining. 

Any inflammatory or hematogenous debris were cleared with the vitrectomy cutter before the introduction of Cy3 anti-GFAP. Sclerotomy plugs were inserted into the superior sclerotomies and 0.5 mL of undilute, dialyzed Cy3 anti-GFAP was injected through the pars plana into the vitreous cavity with a 27-guage needle. The infusion line wasthen clamped. After 15 minutes, the infusion line was unclamped and Cy3 anti-GFAP was cleared from the vitreous cavity with the vitrectomy cutter set on aspiration mode. Day-1 follow-up operations were repeated using exposure times of 5 and 30 minutes. The degree and pattern of intraoperative tissue staining was assessed subjectively by the surgeon. A more formal quantification of the level of specificity was obtained, using tissue that was removed for fluorescence and confocal microscopy. 

For the purpose of description, retinal breaks were classified into three groups: (1) breaks formed immediately after creation of an RRD (fresh break in new RRD), (2) those created at days 1 and 7 in established RRD (fresh break in established RRD) and (3) those created at the time of initial surgery that were exposed to Cy3 anti-GFAP on days 1 or 7 (chronic break in established RRD). 

Optical Modifications of the Operating Microscope.

Using specific excitation illumination and barrier filters, we modified the operating microscope, to allow visualization of intraocular structures stained with Cy3 (Cy3 absorption maximum: 550 nm; emission: ≅ 565 nm). ²² In addition to the standard halogen light source available with the vitrector (Alcon Laboratory, Ltd., Herts, UK), illumination could be switched to a green (532 nm) diode pumped Nd:YV04 laser (Intelite Inc., Minden, NV). An on-off switch allowed pulsed illumination. 

Either the halogen or laser light was transmitted to the eye through a standard, wide-angle, fiber-optic, endoillumination system (Alcon Laboratory, Ltd.) The output of the laser was measured with a light meter (model 371; Graseby, London, UK) and found to be 4.50 mW at 50 ± 0.02 mm (SD). This reduced by a factor of 8.33 when coupled to the fiber-optic endolight. Output from the endolight showed a linear decline with distance (R ² = 0.987). The cone angle of laser light emitted from the tip of the endolight was 36°. Breaks were viewed alternately with the halogen or laser for as long as necessary to make a subjective assessment of the staining pattern—approximately 5 to 15 seconds in most cases. 

The fundus was viewed with a binocular, indirect, viewing system with a stereoscopic diagonal inverter (BIOM II; Oculus, Wetzlar, Germany) fitted to an operating microscope (Omni 1; Carl Zeiss Meditec, Welwyn Garden City, UK). Two barrier-filters (Chroma Technology Corp., Brattleboro, VT) were fitted to the operating microscope below the stereoscopic inverter. The filters had 90% transmission between 555 and 610 nm and excluded light below 545 nm or above 620 nm (manufacturer’s data). A sliding mechanism allowed the surgeon to insert a filter into the light path of one eye, both eyes, or neither eye. 

Histology and Fluorescence Microscopy.

Eyes were enucleated under terminal anesthesia at the end of surgery on day 0, 1, or 7, and dissected to maximize the tissue available for light and fluorescence microscopy. The infusion line was left in situ during enucleation to facilitate orientation of the globe. Globes were positioned cornea uppermost, and the anterior segment was dissected by enlarging the superotemporal sclerotomy with a scalpel blade and extending the incision 360° circumferentially with dissecting scissors. The weight of fluid in the vitreous cavity maintained the eyecups, which were then viewed under the operating microscope with white light. Areas of detached and attached retina were identified, as were larger retinal breaks. Smaller retinal breaks identified previously during fluorophore assisted retinal break detection were located in relation to retinal vessels and then isolated as retinal flatmounts. Tissue for light microscopy was fixed in 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer containing 10 g/L calcium chloride. Tissue for frozen section was fixed in 4% paraformaldehyde in 0.1 M sodium phosphate buffer at pH 7.2 for 1 hour. Retinal breaks were viewed and photographed (Ektachrome 320T; Eastman Kodak) as unfixed flatmounts with a fluorescence microscope with a 546/580-nm filter set (Leitz) and a confocal microscope (Leitz). Paraformaldehyde-fixed specimens were transferred to sucrose and then embedding compound (Cryo-M-Bed; Bright Instruments, Huntingdon, UK) before they were frozen in a eutectic solution of isopentane. A cryostat (Anglia Scientific Instruments, Cambridge, UK) was used to cut 7- to 10-μm frozen sections that were viewed on a fluorescence microscope. Images of flatmount retina were also obtained, and the level of fluorescence of breaks was compared with background fluorescence with the software on the confocal microscope. Glutaraldehyde-fixed specimens were postfixed in 2% osmium tetroxide in 0.2 M sodium cacodylate buffer for 1 hour, dehydrated in a graded series of ethanol, and embedded in araldite epoxy resin. Semithin (1 μm) sections were stained with methylene blue and viewed on a light microscope. 

Results

Ex Vivo Studies

Pattern of Staining in Pig, Cattle, Rabbit, and Humans.

The surface of retinal flatmounts from enucleated porcine eyes did not stain appreciably with Cy3 anti-GFAP; however, the cut outer edge of the flatmount stained markedly (Fig. 1) . The specificity of edge versus surface fluorescence was high at all the dilutions tested, but was highest at the 1:25 dilution of antibody. This edge fluorescence was evident with an exposure time of 1 minute, although it was less than with longer exposure times. Fixed retina also displayed edge fluorescence, as did unfixed bovine and human retina. Rabbit tissue showed notably less edge fluorescence than the other species, with patchy inconsistent staining. An example of a retinal tear is shown in Figure 2 . Elongated, GFAP-positive fibers were visible at the base of the tear and stretched across the retinal defect. It was noted that areas of retinal damage caused by crushing or scratching during tissue manipulation also stained selectively. 

Blocking Studies and Negative Control Experiments.

Cy3 anti-GFAP binding was effectively blocked in the presence of XS soluble GFAP. Higher dilutions of antigen showing increased fluorescence. Exposure of retinal flatmounts to Cy3 dye alone did not result in edge fluorescence (Fig. 1C) . Staining of damaged tissue with trypan blue showed a distribution similar to that of Cy3 anti-GFAP, with cells at the edge of the flatmount selectively stained (Fig. 1A) . Although not formally quantified, this staining was less specific than with Cy3 anti-GFAP. The pattern of retinal staining with anti-vimentin was similar to that observed with anti-GFAP (Fig. 3) . 

Excimer Laser Retinal Ablation.

The mean thickness of bovine retina was 147 ± 4.38 μm (SD; range 142–153) with a mean ILM thickness of 1.47 ± 0.33 μm (SD; range, 1.05–1.89). Each laser pulse resulted in a mean ablation of 0.76 ± 0.02 μm (SD; range, 0.73–0.79) in paraformaldehyde-fixed tissue. 

The profile of retinal ablation is shown in Figure 4 . Ablation through the ILM into the Müller cell end-feet resulted in intense surface fluorescence when retina was incubated with Cy3 anti-GFAP (Fig. 5) . With deeper ablation into the nerve fiber layer, there was a reduced, speckled pattern of fluorescence. The cut outer edge of the retinal button showed characteristic edge fluorescence, but the inner edge (that bordering the central full-thickness ablation) did not. 

In Vivo Studies

After intravitreal injection of Cy3 anti-GFAP, retinal breaks were clearly evident during surgery when viewed with barrier filters and pulsed laser illumination. Small retinal breaks created with a 27-guage needle puncture were not usually visible with white light, but were easily visible with fluorophore-assisted retinal break detection. The degree of fluorescence was affected by the duration of the RRD and age of the break. Group 1 breaks (fresh break in new RRD) showed moderate fluorescence whereas groups 2 (fresh break in established RRD) and 3 (chronic break in established RRD) showed brighter fluorescence. Generally, group-2 breaks were brighter than those in group 3, and of the group-2 breaks, those created in 7-day-old RRDs stained more brightly than those in 1-day-old RRDs. There was no definite difference in the staining of group-3 breaks when comparing day 1 or 7 RRDs, although day 1 may have been brighter (Table 1) . The pattern of staining in attached retina was similar to that in detached retina. Although the morphology of retinal breaks varied with instrument used to create the break (needle puncture, vitrectomy cutter, or end-gripping retinal forceps), there was a similar pattern of staining at the margin of the break. Breaks created with end-gripping forceps had a more ragged edge than the other two types. There was negligible background staining of undamaged retina or other ocular structures. One eye with a large posterior pole hemorrhage demonstrated reduced levels of fluorescence in all retinal breaks. The posterior pole blood clot demonstrated mild fluorescence. Mild to moderate vitreous and retinal hemorrhage did not noticeably affect Cy3 retinal binding. Fluorescence microscopy of retinal breaks (Fig. 6) confirmed a high level of specificity, with edge fluorescence at least 250 times brighter than background fluorescence in all specimens tested on the confocal microscope. 

Discussion

Given that missed retinal breaks are responsible for up to 35% of failed primary operations for RRD, ²³ an agent that makes them more visible would represent a useful advance in vitreoretinal surgery. As early as 1939, Sorsby ²⁴ used Kiton fast green to highlight retinal breaks, and in 1942 Black ²⁵ observed that retinal breaks were visible as red areas when the retina was stained with methylene blue in vivo. In 1969 Kutschera ²⁶ reported success using disulphine blue as a vital stain to highlight retinal breaks. These dyes were delivered to the eye by intravenous or subretinal injections and never gained widespread acceptance. Extensive pilot studies by the present authors, using these and other chromophores, failed to reproduce these earlier findings (Jackson TL, unpublished data, 1999). 

The widespread availability of newer histologic agents with a high degree of tissue specificity means that vital staining of retinal breaks may now be possible. In addition, an increasing proportion of RRD operations are now performed by pars plana vitrectomy, ¹ which allows the surgeon to easily deliver vital stains directly into the vitreous cavity of the eye. Ophthalmologists may also be more receptive to the use of vital stains after the successful use of trypan blue ²⁷ to highlight the capsulorrhexis during cataract surgery, and indocyanine-green-assisted peeling of the ILM. ^{28 29}  

In this study, Cy3 tagged anti-GFAP was used as a vital stain to enhance the detection of retinal breaks. Cy3 anti-GFAP was found to selectively target the tissue at the edge of a retinal break. By combining appropriate laser endoillumination and barrier filters in the operating microscope, it was possible to visualize retinal breaks that were not evident during white-light illumination. 

The cut edge of retinal flatmounts, namely the outermost border, also showed marked fluorescence when incubated ex vivo with Cy3 anti-GFAP. This so-called edge fluorescence was identical with the pattern of fluorescence seen at the edge of retinal beaks created during in vivo studies. 

This study indicates that edge fluorescence is antigen specific for the following reasons. First, and most important, it was effectively blocked in the presence of excess soluble GFAP. Second, the use of Cy3 dye alone did not result in edge fluorescence, indicating that it was not caused by nonspecific binding of the Cy3 label. Third, high-power fluorescence microscopy showed characteristic cellular morphology consistent with previous immunologic studies of glial cells. ³⁰ Last, staining with another glial marker (anti-vimentin) resulted in a similar pattern of fluorescence, consistent with experiments in cat that indicate that GFAP and vimentin have identical distributions in detached retina. ³¹  

Most immunologic protocols combine antigen and antibody for approximately 30 minutes, yet edge fluorescence occurred with exposure times as brief as 1 minute. It is not unexpected that antibodies can combine rapidly with antigens given their function in vivo. It is also possible that selective staining of damaged tissue is enhanced with brief exposure times, as this will reduce the background level of fluorescence. 

There are several possible explanations for the selective edge fluorescence observed in this study. First, the antibody may be endocytosed by glial cells, but selective staining occurred in fixed specimens, excluding this possibility. A second possibility is that damage to the glial cell membranes at the edge of retinal breaks allows the high-molecular-mass (150 kDa) antibody to gain access to its intracellular antigenic target. This hypothesis is supported by the experiments using the excimer laser. In these experiments, the surface of unablated retina did not stain with Cy3 anti-GFAP, but when an area of the Müller cell membrane was ablated, there was intense fluorescence. 

The other observations of ablated retinal tissue also support this hypothesis. When the retinal ablation was continued into the nerve fiber layer, there was reduced surface fluorescence with a speckled pattern of staining. This is consistent with other reports ^{31 32} and our own studies of retinal frozen sections stained with anti-GFAP. Both indicate that GFAP is present in highest concentration immediately below the ILM in the Müller cell end-feet, with thin filaments extending throughout the length of the cell cytoplasm (Fig. 7) . There was no edge fluorescence bordering the full-thickness central ablation. The retina surrounding the central full-thickness ablation was ablated down to the nerve fiber layer. Hence, all Müller cells in this region were exposed to the antibody, and not just those immediately surrounding the central full-thickness ablation. Selective edge fluorescence would therefore not be expected. 

The hypothesis that cell membrane damage results in selective staining is also consistent with the observation that fixed animal specimens and those from human donors both showed increased surface staining relative to unfixed animal specimens with short postmortem times (Fig. 1) . Increased postmortem times and fixation could both alter the integrity and permeability of the Müller cell end feet and increase background surface staining. 

If this hypothesis is correct, then it implies that the cell membranes of at least some glia at the margin of retinal breaks remain compromised for up to 1 week, as breaks of this duration continued to stain with Cy3 anti-GFAP. This suggests that retinal healing process is relatively protracted. 

A third possibility is that the glial cell filaments persist after the cell wall is destroyed. This hypothesis is supported by confocal images showing elongated GFAP-positive filaments at the margin of retinal breaks. As can be seen in Figure 2 , some of these filaments measured more than 300 μm long, approximately twice the retinal thickness. An intact cell of this dimension is unlikely, particularly as these ex vivo experiments did not allow time for glial hypertrophy. This suggests that aggregated intermediate glial cell filaments persist despite a loss of cell integrity, consistent with the known structure of neurofilaments. ³³ These macromolecules comprise an α-helical core that may function as the internal skeleton of neuronal cells. ³⁴ They resist trypsin digestion ³⁵ and are extremely mechanically robust. ³⁴ It is also known that they can be pulled into aggregated, highly oriented fibers. ³⁶  

The findings of this study therefore support both the second and third hypotheses, and these are not mutually exclusive. Hence, glia with damaged cell membranes may be selectively stained with Cy3 anti-GFAP, but so too are the intermediate filaments that persist after the integrity of the cell wall is destroyed. 

The finding that retinal breaks stained because the Müller cell membrane was either damaged or destroyed has implications regarding the potential clinical use of this agent as a vital stain. If fluorophore-assisted retinal break detection were possible in humans, then it may be most suitable for the acute or semiacute mechanical retinal tears that occur after posterior vitreous detachment (PVD) or intraoperative retinal injury. The ability to identify iatrogenic retinal breaks may be useful, but it is likely that partial-thickness retinal defects would also stain, if the injury disrupted an appreciable number of Müller cell membranes. Hence, if Cy3 anti-GFAP were used to locate iatrogenic retinal tears, further examination would be necessary to determine that the defect was full thickness. 

It is important to note that this study cannot conclude on the efficacy of this technique in humans. In particular, it is not known how older retinal breaks, atrophic holes, or lattice degeneration would stain. Each of these is considered in the following three paragraphs. 

Retinal break fluorescence was partly dependent on the duration of the RRD. This was consistent with the known temporal response of glial cells to injury. GFAP is expressed in the normal retina of many species ^{31 32 37 38} and our own studies confirm that there is constitutive expression in pig. ¹⁹ Levels of GFAP increase markedly in response to many types of neuronal injury, ^{30 38 39 40} , including RRD. ^{19 31} This response is well established by 24 to 48 hours ^{19 31 41} and explains why fresh breaks in healthy, attached retina stained with Cy3 anti-GFAP, but not as intensely as those created in established RRD. 

The chronicity of retinal breaks may therefore alter the degree of fluorescence. Retinal breaks created at the time of initial surgery stained brightly when exposed to Cy3 anti-GFAP 1 week later. It is not known whether later remodeling of retinal breaks would reduce retinal binding. Light and electron microscope studies of chronic experimental RD in primates ^{16 42 43} demonstrated progressive rounding of the retinal break edge associated with proliferation of glial cells at the wound edge. Both these and our own experiments in pig indicate that this chronic glial response is present within 1 week of RRD. ¹⁹ Selective staining of older retinal breaks might therefore be expected. Alternatively, the reformation of intact cell borders and the eventual degradation of residual glial filaments might reduce staining. Longer follow-up is needed to determine this outcome. 

It is not known whether atrophic holes would stain selectively. Selective staining of atrophic holes may occur, as these lesions have a histologic appearance ⁴⁴ similar to the smooth edged, 1-week-old retinal breaks that occur within 1 week of RRD. ¹⁹ Alternatively, it is possible that the lack of mechanical damage would reduce the ability of the Cy3 anti-GFAP to reach its intracellular target, as would the absence of extracellular glial elements. By the same reasoning, lattice degeneration would not be expected to stain, because there is no mechanical disruption of the glial cell membrane. Therefore, this study can only conclude on the ability Cy3 anti-GFAP to target mechanical breaks. In the common setting of RRD after PVD these are the most clinically important types of retinal break. 

It seems likely that breaks in attached or detached retina would both be stained, as noted in the in vivo studies. This finding was expected, as the increased GFAP immunoreactivity that occurs after RRD is not restricted to the area of detached retina. ¹⁹  

The viewing system used in the in vivo experiments was fitted to an operating microscope and was designed for use during PPV. In many countries PPV is increasingly used to treat RD, ^{1 2} but it would be useful to adapt this system for use with the indirect ophthalmoscope and conventional (cryobuckle) RD surgery. The optical modifications needed to do this are feasible, but delivery of Cy3 anti-GFAP to the retina requires further research. The simplest means of introducing the agent into the eye is by pars plana injection, but other options, such as iontophoresis ⁴⁵ or osmotic pumps, ⁴⁶ may become available. Although there are obstacles to overcome—most notably, delivery past the outer blood-ocular barrier—recent results have shown that transscleral delivery of immunoglobulins to the retina is possible. ⁴⁷ It is not known whether transscleral delivery of Cy3 anti-GFAP would be as effective at vital staining as intravitreous delivery, and removal, from a vitrectomized eye. 

On the basis of this study, we cannot draw conclusions about the safety of intraocular Cy3 anti-GFAP or the risk of phototoxicity from the laser light used to illuminate Cy3 stained tissue. No histologic evidence of retinal toxicity was observed during in vivo studies, but most eyes were removed within 1 hour of exposure to Cy3 anti-GFAP and may not have had time to manifest histologic changes. 

Given that our intended target is damaged glial cells at the margin of a retinal break, toxicity may be less problematic, as these areas would normally be ablated by laser or cryoretinopexy. It is possible, however, that some amount of complement fixation may occur if residual Cy3 anti-GFAP was not fully rinsed from the eye. This could be avoided by removing the constant domain of the antibody, although it is unclear whether the reduction of molecular weight would increase retinal penetration and therefore background fluorescence. Although the therapeutic and diagnostic use of antibodies has been applied in other specialties ⁴⁸ and in animal models of ocular disease, ^{49 50} any agent that was used to identify retinal breaks would have to undergo further safety testing. 

With laser endoillumination and barrier filters fitted to the operating microscope, we used Cy3 anti-GFAP as an intraoperative vital stain to enhance the detection of mechanical retinal breaks. Retinal breaks up to 1 week old were easily identified with a high level of specificity. Ex vivo studies suggested that Cy3 anti-GFAP selectively stains exposed glial elements at the edge of a retinal break. This study indicates that fluorophore-assisted retinal break detection is possible in an animal model of RRD. 

 This work forms a partial contribution to a Doctor of Philosophy (TLJ).

 Presented at the annual meeting of the Association for Research in Vision and Ophthalmology, Fort Lauderdale, Florida, May 2002, and at the 23rd Cambridge Ophthalmological Symposium, United Kingdom, September 2001.

 Supported by a research grant from the Allerton Trust and a personal training award from the Special Trustees of Guys and St. Thomas’ Hospitals, London (TLJ). Surgical equipment was donated or provided free of charge by Altomed Ltd., Tyne and Wear, UK; Stat One Services Ltd., Coldfield, UK; and Alcon Laboratories Ltd., Hertfordshire, UK.

 Submitted for publication July 25, 2003; revised September 20, 2003; accepted October 27, 2003.

 Disclosure: T.L. Jackson, Alcon Laboratories Ltd., Altomed Ltd., Stat One Services (F); J. Marshall, None

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

 

Figure 1.

 

Photomicrograph of retinal flatmounts. (A) Unfixed porcine retina stained with trypan blue, with the cut edge of the flatmount at left, showing an increased uptake of trypan blue relative to the remainder of the retinal surface. (B) A flatmount incubated ex vivo with Cy3 anti-GFAP and viewed with a filter set in place. The orientation is preserved with the cut edge at left. This edge showed marked fluorescence, with minimal labeling of the retinal surface. Varying the plane of focus showed that the pattern of staining was consistent with the appearance of Müller cells (see also Fig. 3 ). (C) Another retinal flatmount exposed to Cy3 dye alone. No edge fluorescence was evident. Similar images were obtained in blocking studies using excess soluble GFAP. A fixed retinal flatmount (D) shows edge fluorescence similar to that in the unfixed specimen (B). (E) Human retina. Edge fluorescence was also present. Note that there was more staining of the retinal surface in both the fixed animal specimen and unfixed human specimen (D, E) relative to the unfixed animal specimen (B). Similar images were obtained with bovine tissue. Original magnification, ×320.

Figure 2.

 

Confocal microscope image of a tear created ex vivo in unfixed bovine retina with subsequent incubation with Cy3 anti-GFAP. Note the fibrillary pattern of staining, with some fibers passing across the diameter of the tear. Bar, 100 μm.

Figure 3.

 

Fluorescence photomicrograph of a porcine retinal flatmount exposed to Cy3 anti-GFAP ex vivo (A). The edge of the flatmount is seen to the left of this image and is brightly fluorescent. (B) Tissue stained with Cy3 anti-vimentin. The pattern of edge fluorescence obtained with these agents was similar. Original magnification, ×480.

Figure 4.

 

Scanning electron micrograph of a hemisected bovine retinal flatmount that had undergone a step-wise ablation (0, 3, 10 μm, and full-thickness) with an excimer laser. Scale: central, full-thickness ablation measures 1 mm in diameter.

Figure 5.

 

Using the same ex vivo ablation profile as that shown in Figure 4 , tissue was subsequently incubated with Cy3 anti-GFAP. The unablated bovine retina (A) showed negligible surface fluorescence, whereas tissue that had undergone ablation to the level of the Müller cell end-feet showed intense surface fluorescence (B). Deeper ablation into the nerve fiber layer (C) showed staining of the Müller cell processes. No increase in fluorescence was observed at the edge of the inner, full-thickness ablation. Original magnification, ×200.

Table 1.

 

Nature of Retinal Break and Subjective Intraoperative Comparison of Levels of Fluorescence

Duration of RD	Fresh Breaks	Chronic Breaks
Day 0 n = 2 (4)	+	N/A
Day 1 n = 3 (7)†	+++	++
Day 7 n = 2 (4)	++++	++

 The assessment of the degree of fluorescence was made by the surgeon at the time of vital staining and is shown as + (lowest) to ++++ (highest). Fresh breaks were those created immediately prior to vital staining. Chronic breaks were created during the initial surgery to create the RD. Therefore, at the time of vital staining, there were both fresh and chronic breaks in each RD. Each eye had between one and three RDs. N/A, not applicable.

 *, n, number of animals, with number of RDs in parentheses.

†  Includes three cases with an exposure time of five minutes and one of 30 minutes. Other exposure times were 15 minutes.

Figure 6.

 

Iatrogenic retinal breaks created in vivo in pigs. Tissue was subsequently removed from the eye for photography on a fluorescence microscope. The four breaks in (A) were created in vivo in detached retina with a 27-gauge needle. The retinal tissue shown was exposed to intravitreous Cy3 anti-GFAP for 5 minutes during surgery, removed from the eye, and viewed as an unfixed flatmount, without further exposure to the antibody or any other reagent. (B) Higher-power view of one of the retinal breaks, visible in (A) at top left. Note that the pattern of edge fluorescence is similar to that observed at the cut edge of retinal flatmounts incubated ex vivo with Cy3 anti-GFAP (Fig. 1B) . (C) Frozen section of another retinal break in the same eye. As in the flatmounts, the only exposure to Cy3 anti-GFAP was during surgery. The grayscale (D) shows the same frozen section viewed with a white light and no barrier filter. Original magnification: (A) ×50; (B; cropped from A); (C, D) ×680.

Figure 7.

 

Frozen section of bovine retina exposed to Cy3 anti-GFAP after sectioning. The characteristic pattern of fluorescence can be seen, with peak immunoreactivity near the Müller cell end-feet. Thin GFAP-positive fibrils can be seen extending from the end-feet to the level of the external limiting membrane. Tissue was obtained from full-thickness retinal flatmounts of attached retina. Original magnification, ×670.

The authors thank Kathy Clarke for anesthetic support, Tom Williamson for assistance with surgical planning, and Jo Cunningham for technical advice and surgical assistance.