Immunocytochemical and Ultrastructural Evidence of Glial Cells and Hyalocytes in Internal Limiting Membrane Specimens of Idiopathic Macular Holes

Abstract

Purpose.: To provide new information on epiretinal cell proliferation and the cells' origin in idiopathic macular holes and to overcome the effects of embedding and sectioning preparation procedures on cell-distribution patterns.

Methods.: Interference and phase-contrast microscopy, immunocytochemistry, and scanning and transmission electron microscopy were performed on surgically excised whole-mounted internal limiting membrane (ILM) specimens removed from 60 eyes with idiopathic macular holes. Cell distribution and cell morphology were correlated with immunocytochemical staining characteristics. Twelve cell type–specific antibodies were used to detect glial cells, hyalocytes, retinal pigment epithelial cells, retinal ganglion cells, and immune cells. Cell viability was analyzed.

Results.: Epiretinal cell proliferation was found in all ILM specimens, irrespective of the stage of the macular hole. Cell density showed a broad variety. Immunocytochemistry frequently revealed simultaneous expression of GFAP/CD45, GFAP/CD64, GFAP/CD68, GFAP/CRALBP, and GFAP/CD90. Some cells presented with intracellular contractile filaments (anti-αSMA); others were not immunoreactive to any antibody examined. The percentage of viable cells showed a broad variety with a mean of 73% (SD 29%). Electron microscopy demonstrated glial cells, hyalocytes, and myofibroblast-like cells.

Conclusions.: The presence of epiretinal cells at the ILM in all macular hole stages strongly suggests a substantial involvement of cell migration and proliferation in the course of macular hole development. Glial cells and hyalocytes play the predominant role in epiretinal cell proliferation. Given the co-expression of glial cell and hyalocyte markers, transdifferentiation of epiretinal cells needs further elucidation, especially with respect to αSMA-positive cells leading to traction at the vitreoretinal interface.

The pathogenesis of idiopathic macular hole is still a matter of debate. There is general consensus on the role of vitreoretinal traction in macular hole formation. ^{1 –5} However, the role of epiretinal cell proliferation at the internal limiting membrane (ILM) remains controversial and is the focus of ongoing investigation. Two substantial hypotheses are currently discussed. 

On the one hand, activated glial cells, in particular Müller cells and astrocytes, are considered to dedifferentiate, proliferate, and migrate from the retinal side to the vitreal side of the ILM where they form epiretinal membranes. ^{6 –8} Epiretinal cell proliferation at the ILM causes tangential traction that leads to progression or reopening of macular holes. ⁹ Gliotic responses of the retina occur in a broad variety of pathologic retinal alterations, to protect neurons from retinal damage. ¹⁰ Within the retina, Müller cells are the main type of glial cell. The upregulation of the intermediate filament glial fibrillary acidic protein (GFAP) is the most sensitive response of Müller cells to retinal injury and can be used as a retinal stress indicator and a marker for Müller cell activation. ^6,7,11,12  

On the other hand, it has been suggested that hyalocytes become activated, thereby driving cellular proliferation at the ILM. These macrophage-like cells of the vitreous were first described in 1845 by Hannover. ¹³ Balazs et al. ¹⁴ named these cells hyalocytes. They are embedded in the vitreous cortex situated 20 to 50 μm from the ILM at the posterior retina. ³ In an incomplete posterior vitreous detachment with anteroposterior vitreoretinal traction, which has been postulated to be a primary cause of macular hole formation, the vitreous cortex remains partially left behind at the ILM containing a heterogeneous population of hyalocytes. ⁴ These cells are thought to belong to the monocyte/macrophage lineage, but derivations from the neuroglia in the retina, from mesenchymal cells of the retinal vasculature, and from cells of the hyaloid system have also been proposed. ^3,15  

Determining the exact cell types and the cell's origin has been the purpose of several previous studies using light and electron microscopic preparation techniques. During the last decades, morphologic analyses of surgically excised ILM specimens have demonstrated a variety of cells in epiretinal membranes of idiopathic macular holes that are thought to be of retinal and extraretinal origin, including hyalocytes, glial cells (Müller cells, fibrous astrocytes, microglia), retinal pigment epithelial cells, fibrocytes, and myofibrocytes. ^{16 –20} However, morphologic criteria alone appear to be inadequate for identifying the cells' origin, because proliferating epiretinal cells undergo a phenotypic transdifferentiation adopting morphologic features of other cell types. Even immunocytochemical investigation of cell type–associated antigen expression has not overcome the issue of transdifferentiation so far, because there is no specific marker of it. ^21,22  

Recently, a new preparation method of flat-mounted ILM specimens was proposed to provide more information on epiretinal cell proliferation. ^{23 –25} To overcome limitations of conventional sectioning and embedding preparation procedures, we used the flat-mount preparation procedure to investigate cell proliferation at the ILM by en face phase-contrast and interference microscopy directly combined with immunocytochemical analysis, live–dead cell viability assay, scanning electron microscopy, and transmission electron microscopy. Herein, we present new details on epiretinal cell distribution and cell density, as well as on cell viability, the cells' ultrastructural phenotype, and their antigen expression in macular hole formation. 

Patients and Methods

From two consecutive series of ILM specimens removed during macular hole surgery between October 2007 and October 2008, and from November 2009 to January 2010, 60 specimens from 60 eyes of 60 patients were included in the study. Specimens from 40 eyes were processed for phase-contrast microscopy, interference microscopy, and immunocytochemistry by flat-mount preparation of excised ILM. The specimens were obtained from 7 eyes with stage II macular hole, 26 eyes with stage III macular hole, and 7 eyes with stage IV macular hole, according to Gass' classification. Furthermore, flat-mounted ILM specimens obtained from 12 eyes of 12 patients (9 eyes with stage III macular hole, 3 eyes with stage IV macular hole) were processed for live–dead cell viability assay. In addition, ILM specimens of eight eyes of eight patients who underwent surgery for stage III macular holes were processed for scanning electron microscopy. Patients' records were reviewed for age, sex, previous ocular surgery and trauma, presence of metamorphopsia, and duration of symptoms before surgery. This study was approved by the Institutional Review Board and the Ethics Committee. Informed consent was also obtained from all persons involved in the project, in accordance with the tenets of the Declaration of Helsinki. 

General Procedures

Patients underwent a standard three-port pars plana vitrectomy. Before the infusion line was opened, the status of the posterior hyaloid was determined with a plano-concave contact lens. If necessary, a posterior vitreous detachment was induced by suction with the vitrectomy probe around the optic nerve head. The posterior hyaloid was detached from the retina, and the posterior vitreous detachment was extended to the periphery. In the majority of cases (51/60 specimens), Brilliant blue G (0.5 mL, 0.25%; Fluoron GmbH, Neu-Ulm, Germany) was used to visualize the ILM. No other dye was applied. The dye solution was injected into the fluid-filled globe and washed out immediately. Peeling of the ILM was performed with an end-gripping Eckardt forceps, intending to remove the ILM more than 1.5 disc diameters surrounding the macular hole. The ILM with epimacular tissue was harvested and placed into the fixation solution. After fluid–air exchange, the vitreous cavity was perfused with a mixture of 15% C2F6. Postoperatively, patients were recommended to maintain a face-down position for 4 days. 

Phase-Contrast and Interference Microscopy of Flat-Mounted ILM Specimens

Surgically excised ILM specimens of 40 patients were immediately fixed with a mixture of 2% paraformaldehyde and 0.1% glutaraldehyde in 0.1 M phosphate-buffered saline (PBS; pH 7.4) at 4°C for at least 24 hours. Under a stereomicroscope (MS 5; Leica, Wetzlar, Germany), the ILM specimens were placed onto glass slides and unfolded using glass pipettes to show the maximum area of their surface. Antifade mounting medium 4′,6-diamidino-2-phenylindole (DAPI; AKS-38448; Dianova, Hamburg, Germany) was applied for cell nuclei staining, and a coverslip was added. Specimens were analyzed with a modified fluorescence microscope (model DM 2500; Leica) that enabled us to perform both interference and phase-contrast microscopy at magnifications between ×5 and ×40. 

As previously reported, phase-contrast microscopy was performed as a contrast-enhancing optical technique that can produce high-contrast images of translucent tissue. Differences in image contrast are produced by using an optical mechanism to translate minute variations in phase into corresponding changes in amplitude. 

Interference microscopy was applied to detect changes in surface height, such as cellular proliferation compared with bare ILM. Differences in image intensity are produced by splitting the entering light into two beams that pass through the specimen and become recombined in the image plane, where the interference effects make the transparent object details become visible. 

Immunocytochemistry of Flat-Mounted ILM Specimens

Indirect immunocytochemistry was also performed on ILM specimens that were flat mounted on glass slides. After phase-contrast and interference microscopy, ILM specimens of 40 patients were further processed for immunocytochemical staining. After fixation, specimens were rinsed in 0.1 M PBS (pH 7.4) before incubation with 0.1% pepsin in 0.2 M PBS for 30 minutes at room temperature. After the sections were rinsed twice in 0.1 M PBS, they were incubated in normal donkey serum (1:20) in PBS, 0.5% BSA, 0.1% Triton X-100, and 0.1% Na-azide (PBTA) for 3 hours. Then, specimens were rinsed with PBTA three times for 5 minutes each time, and the primary antibodies were added and incubated in PBTA over night at room temperature. 

Twelve antibodies were used for glial and retinal cells (anti-glial fibrillary acidic protein [anti-GFAP]; DAKO, Hamburg, Germany; anti-α smooth muscle actin [anti-αSMA], Sigma-Aldrich, Taufkirchen, Germany; anti-cellular retinaldehyde-binding protein, [anti-CRALBP], Santa Cruz Biotechnology, Heidelberg, Germany; anti-neurofilament, [anti-NF] and anti-vimentin, DAKO; and anti-CD90, Santa Cruz, Biotechnology), for retinal pigment epithelial cells (anti-cytokeratin 8, anti-CK8, anti-pan-cytokeratin, and anti-panCK; Sigma-Aldrich), for hyalocytes (anti-CD45, anti-CD64, and anti-CD34; Santa Cruz Biotechnology), and for macrophages/microglia (anti-CD68, Santa Cruz Biotechnology), as listed in Table 1. Primary antibodies were diluted according to the manufacturer's instructions. Since the maximum number of fluorochromes used at one time was limited, and the antibody combinations were limited because of the species from which they originated, we used labeling combinations of a maximum of three antibodies. 

Table 1.

 

Antibodies Used for Immunocytochemical Staining

Antibodies	Target Structure
Glial fibrillary acidic protein (GFAP)	Intermediate type filaments of glial cells
Vimentin	Intermediate type filaments of glial cells
Cellular retinaldehyde binding protein (CRALBP)	Glial cells/retinal pigment epithelial cells
Cytokeratin 8 (CK8)	Retinal pigment epithelial cells
Pan-cytokeratin (panCK)	Retinal pigment epithelial cells
Neurofilament	Retinal ganglion cells
CD 90	Retinal ganglion cells
CD 35	Hyalocytes
CD 45	Hyalocytes
CD 64	Hyalocytes
α-Smooth muscle actin (αSMA)	Intracellular actin filaments
CD 68	Macrophages and Microglia

After the primary antibodies were rinsed in PBTA three times for 10 minutes each, the secondary antibodies (donkey anti-mouse CY3, donkey anti-rabbit CY2, and donkey anti-goat CY5; Dianova, Hamburg, Germany) were added together, each in 1:100 PBTA, for 2 hours at room temperature. Finally, specimens were rinsed in PBTA four times for 10 minutes each and in PBS three times for 5 minutes each. 

Additional flat-mounted ILM specimens were used as a control and were not included in statistical analysis. Primary antibodies were replaced with diluent followed by incubation with secondary antibody alone. All other procedures were identical with normal immunolabeling. 

Antifade mounting medium 4′,6-diamidino-2-phenylindole (DAPI; AKS-38448; Dianova) was applied for cell nuclei staining, and a coverslip was added. For photodocumentation, we used the fluorescence microscope (model DM 2500; Leica) at magnifications between ×50 and ×400. 

Live–Dead Cell Viability Assay of Flat-Mounted ILM Specimens

Cell viability was quantified based on a two-color fluorescence assay with membrane-impermeable propidium iodide (Sigma-Aldrich) and membrane-permeable Hoechst 33342 (Intergen, Purchase, NY). All cell nuclei were labeled blue, because Hoechst 33342 dye binds to double-stranded nucleic acids (DNA) of all cells. Nuclei of nonviable cells were labeled red by propidium iodide, which is generally excluded from viable cells but penetrates cell membranes of damaged or dead cells followed by intercalation into their DNA. 

Immediately after harvesting during vitrectomy, ILM specimens of nine patients were incubated with 2.0 g/mL propidium iodide and 1.0 g/mL Hoechst 33342 for 20 minutes at 37°C for evaluation of cell viability. Specimens of three patients were incubated 24 hours after their removal during vitrectomy with storage in balanced saline at +4°C. All ILM specimens were placed onto glass slides and unfolded with glass pipettes to show their maximum surface. Documentation with the epifluorescence microscope (DM 2500; Leica) followed. Labeled nuclei were counted in fluorescence micrographs using ImageJ software (developed by Wayne Rasband, National Institutes of Health, Bethesda, MD; available at http://rsb.info.nih.gov/ij/index.html) with manual counting if the total cell count was less than 100 or with automatic cell counting if the total cell count was more than 100 cells per specimen. Viable cells were expressed as a percentage of the total nuclei in the whole specimen. 

Scanning Electron Microscopy (SEM)

Immediately after their removal, ILM specimens of eight patients were placed in 1% glutaraldehyde in 0.1 M PBS (pH 7.4) for fixation. In single cases, interference microscopy and phase-contrast microscopy were performed before dehydration in an ascending series of ethanols. Then, specimens were dried by the critical-point method, using CO₂ as the transitional fluid (Polaron Critical Point Dryer CPC E3000; Quorum Technologies, Ringmer, UK). Specimens were sputter-coated with a thin layer of 7-nm platinum by a sputtering device (Emitech K575; Quorum Technologies, Ringmer, UK) and observed by SEM (JSM 6300F; JEOL, Eching, Germany) with magnifications between ×200 and ×10,000. 

Transmission Electron Microscopy

From a proportion of patients, two or more several-part ILM specimens were peeled during vitrectomy. In these cases, four specimens of four patients (including one patient with stage II, two with stage III, and one with stage IV macular hole) were selected and processed for phase-contrast and interference microscopy. Selected specimens presented with a large area and homogeneously distributed cells. They were cut into halves. Then, one half was processed for immunocytochemical staining, and the other half was prepared for conventional light microscopy and transmission electron microscopy, to correlate flat-mount preparation with conventional sectioning preparation procedures. 

For transmission electron microscopy, specimens were placed into phosphate-buffered 4% glutaraldehyde solution for fixation. Postfixation in osmium tetroxide 2% (Dalton's fixative) and dehydration in graded concentrations of ethanol with embedding in Epon 812 followed. Semithin sections of 400 nm were stained with an aqueous mixture of 1% toluidine blue and 2% sodium borax. Ultrathin sections of 60 nm were obtained by series sectioning and were contrasted with uranyl acetate and lead citrate. Analysis and imaging of five grids (each with six to nine ultrathin sections) per specimen was performed with a light microscope and an electron microscope (Zeiss EM 9 S-2; Carl Zeiss, Jena, Germany). 

Photodocumentation and Statistical Analysis

All specimens that were processed for phase-contrast microscopy, interference microscopy, and immunocytochemistry were photodocumented using a digital camera to image the ILM specimens (ProgRes CF; Jenoptik, Jena, Germany). Images were analyzed measuring the specimen area in consideration of the magnification. Cell quantification was analyzed using ImageJ software with automatic and manual cell counting. Finally, measured areas and cell counts of specimen parts were added up from each patient. Total cell counts, cell density, and total area of specimens were analyzed considering P < 0.05 as statistically significant (SPSS 18.0; SPSS, Chicago, IL). 

Results

Cell proliferation at the ILM was found in all ILM specimens removed from 60 eyes with idiopathic macular hole. 

Cell Count and Cell Distribution

Phase-contrast and interference microscopy showed the ILM as an intact sheet with an irregular round defect corresponding to the area of the macular hole (Fig. 1). Areas with cell proliferation were easily distinguished from areas without cell proliferation. Cell nuclei and cell extensions were clearly delineated from the ILM. Fluorescence microscopy showed the cell nuclei by DAPI staining. 

Figure 1.

 

Whole flat-mounted ILM specimens of idiopathic macular holes. (A) Interference microscopy and (B) phase-contrast microscopy demonstrate the ILM as an intact sheet with an irregular round defect corresponding to the area of the macular hole (*). Areas of cell proliferation were easily distinguished from areas without cell proliferation. (C) Cell nuclei were stained by 4′,6-diamidino-2-phenylindole (DAPI) and analyzed by fluorescence microscopy. Original magnification, ×100.

In about half of all 40 cases (n = 19) examined for cell count and cell distribution analysis, cell proliferation was seen as a continuous sheet of cells homogeneously distributed on the ILM (Figs. 2A, 2B). Specimens with homogeneous cell proliferation, including the macular hole rim, often showed that cells were not directly distributed at the macular hole edge but at a small distance surrounding the hole. In 21 eyes, the cells were inhomogeneously distributed in the ILM-forming areas of cell clusters (Figs. 2C, 2D) distant from the macular hole rim. The specimens showed a large variation in the total number of cells, ranging from 18 to 6914, with a mean of 1663 (SD 1959). The total area of removed ILM ranged from 3.2 to 22.6 mm², with a mean of 12.0 mm² (SD 4.9) (Fig. 3A). In specimens with inhomogeneous distribution patterns, areas of cell clusters covered a mean of approximately 56% (SD 22%) of the excised ILM. Cell density in specimens with homogeneous cell distribution was 95 cells/mm² compared with 365 cells/mm² in cell clusters. Specimens with cell clusters showed a significantly higher cell density than did specimens with homogeneously distributed cells (Fisher's exact test, P < 0.01; Fig. 4A). 

Figure 2.

 

Cell distribution patterns. Nuclear staining by 4′,6-diamidino-2-phenylindole (DAPI) showed different cell proliferation patterns with various cell counts on flat-mounted ILM specimens removed from idiopathic macular holes. (A, B) Nineteen eyes were seen with homogeneously distributed cells on the entire ILM specimen. (C, D) In 21 eyes, the cells were inhomogeneously located as cell clusters covering about half of the entire area of the ILM specimen. Original magnification: (A, C) ×50; (B, D) ×200.

Figure 3.

 

Cell counts according to cell distribution patterns and stage of macular hole. Specimens showed a considerable variety of total cell count and area of removed ILM comparing (A) the cell distribution and (B) the stage of disease. The total number of cells ranged from 18 to 6914, with a mean of 1663. The area of removed ILM ranged from 3.2 to 22.6 mm² with a mean of 12.0 mm². There was a tendency of cell count to increase with stage of disease.

Figure 4.

 

Cell density according to cell distribution pattern and stage of macular hole. (A) Specimens with cell clusters had significantly higher cell density than specimens with homogeneously distributed cells (Fisher's exact test, P < 0.001). (B) Cell density did not significantly increase with stage of disease.

Comparing the stages of disease (Figs. 3B, 4B), no significant differences were seen in the total cell count, peeled area of the ILM, or cell distribution patterns. However, there was a tendency of cell count to increase with stage of disease. Cluster formation increased with the duration of symptoms before surgery, but did not correlate with the presence of metamorphopsia. 

Cell Type–Associated Antigen Expression

Whenever we obtained two or more specimens from one patient, we were able to perform indirect immunofluorescence evaluation with more than one combination of three antibodies. Specimens were analyzed according to the presence of labeling and stage of macular hole, as presented in Table 2. 

Table 2.

 

Cell Type-Associated Antigen Expression in Flat-Mounted ILM Specimens of Idiopathic Macular Hole

Antigen	Antigen Expression
	Stage II MH	Stage III MH	Stage IV MH
GFAP	+	+	+
Vimentin	(+)	+	+
CRALBP	−	+	+
Neurofilament	−	−	−
Cytokeratin (CK8 + panCK)	−	(+)	(+)
CD-34	−	−	−
CD-45	−	+	+
CD-64	+	+	+
CD-90	+	−	−
αSMA	+	+	+
CD-68	−	+	+

 +, present; (+), sparsely present; −, absent.

Anti-GFAP labeling was found in all specimens and was predominant compared with all other antibodies tested. Anti-vimentin was present in all stages of macular hole, but not in all specimens. Anti-CRALBP was labeled mainly in specimens removed from stage III and IV macular holes. Anti-neurofilament was not seen in this series. Anti-CK8 and -panCK were occasionally present in stage III and IV macular holes. The hyalocyte marker anti-CD45 showed more prominent labeling than did anti-CD64. Anti-CD34 was not seen in this series. The retinal ganglion cell marker anti-CD90 was occasionally positive. Anti-αSMA was inconsistently positive, but was seen in all stages, whereas anti-CD68 was demonstrated mostly in stage III and IV macular holes. 

The most unexpected findings were co-localizations of anti-GFAP and hyalocyte cell markers, as demonstrated in Figure 5. CD45-positive cells were found in 67% of tested specimens. When labeled with anti-GFAP as a two-antibody combination, anti-CD45 always co-localized with anti-GFAP (Fig. 5A). Anti-CD45-labeled cells were frequently found in direct proximity to αSMA-positive cells. Anti-CD64 co-localized with anti-GFAP, but never with anti-CD45 (Fig. 5B). Co-localization was also seen in GFAP- and CD68-labeled specimens (Fig. 5C). Anti-CD68 labeling was positive in half of all specimens tested and sometimes co-localized with anti-GFAP. If anti-CRALBP-positivity was found, it was co-localized with anti-GFAP, rather than with anti-CK8 or -panCK (Fig. 5D). CRALBP is known to be expressed in retinal pigment epithelial cells and Müller cells carrying 11-cis-retinol and 11-cis-retinaldehyde. ⁶ Anti-GFAP and -vimentin labeling is expected to co-localize on Müller glial cells; however, both antibodies were also found co-localized with anti-CD90 (Fig. 5E). Of note, no apparent co-expression of αSMA and GFAP cell markers was found in any specimens examined. 

Figure 5.

 

Indirect immunofluorescence evaluation of flat-mounted ILM specimens with important cell marker co-localization. GFAP-positive cells were seen in all tested specimens, dominating the co-localized cell marker combinations. (A) In 67% of specimens, cell labeling with anti-CD45 was positive and always co-localized with anti-GFAP. α-SMA labeling was frequently positive. (B) If anti-CD64 labeling was tested, it was mostly co-localized with anti-GFAP, but it was never seen to co-localize with anti-CD45. (C) Anti-CD68 is a cell marker of macrophages but was surprisingly seen co-localizing with anti-GFAP in this study. (D) Anti-CRALPB identifies glial and retinal cells. CRALPB-positive cells were demonstrated in co-localization with anti-GFAP labeling only. (E) Co-localization of glial cell markers such as anti-vimentin and anti-GFAP was expected, but they were found co-localized with anti-CD90. Original magnification: (A, B, D) ×400; (C, E) ×200.

In all control specimens, when the primary antibody was replaced by diluent, no immunoreactivity was observed, as illustrated in Figure 6. Background labeling contrast was enhanced (Photoshop CS4; Adobe) to demonstrate that there was no immunoreactivity. 

Figure 6.

 

No immunoreactivity in negative controls of whole flat-mounted ILM specimens for the experiments shown in Figure 5. Primary antibodies were replaced by diluent. Background labeling was enhanced to demonstrate that there was no immunoreactivity observed. DAPI nuclear staining in merged images demonstrates the presence of cell proliferation in all control specimens. Original magnification, ×200.

Besides the issue of co-localization of antibodies that were considered to be cell type–associated, such as anti-GFAP to identify glial cells, anti-CD45 and -CD64 to identify hyalocytes, anti-CD68 to identify macrophages/microglia, and anti-CD90 to identify retinal ganglion cells, there was a proportion of cells that did not label with any cell marker combination. These unidentified cells were not only negative for GFAP labeling but also for other cell type–specific antibodies—for example, anti-CD64 and -CD68, as demonstrated in Figure 7. Negatively labeled cells were mostly found loosely distributed as single cells at the ILM. By phase-contrast microscopy and interference microscopy, immunocytochemically unidentified cells presented with very sparse cytoplasm of star-like shape. 

Figure 7.

 

Negative labeling of anti-GFAP. Although positive anti-GFAP labeling was shown in all specimens, there was a proportion of cells within the specimens that labeled with neither anti-GFAP nor other cell markers tested, such as hyalocyte and pigment epithelial cell markers. (A) Phase-contrast micrograph combined with DAPI cell nuclear staining demonstrates two cells (arrowheads) with single elongations (B) that labeled positively with anti-GFAP and -CD64 in co-localization, but it also shows cells with sparse cytoplasm (B) that were negative for anti-GFAP, -CD64, and -CK8. (C) Interference micrograph combined with DAPI cell nuclear staining demonstrates a cluster of cells (arrows) that was positively labeled with anti-GFAP and -CD68 in co-localization (D), (C) but it also shows homogeneously distributed star-like cells with sparse cytoplasm, as shown in the higher magnification inset, that were negative (D) for anti-GFAP, -CD68, and - CK8. Original magnification: (A, B) ×400; (C, D) ×100.

Cell Viability

As presented in Table 3 and Figure 8, viable cells were expressed as a percentage of the total nuclei labeled with Hoechst 33342. Percentage of viable cells showed a broad variety from 17% to 100% with a mean of 73% (SD 29%). Specimens of two patients with stage IV macular holes showed with high cell densities of 342 and 1076 cells/mm², according to the clinical diagnosis of epiretinal gliosis made by biomicroscopic examination. However, in these two specimens, the percentage of viable cells approached the mean, with 85% and 69%. Furthermore, the usage of Brilliant Blue G during vitrectomy to visualize the ILM intraoperatively did not correlate with the lowering of the viable cell count within the specimens. The stage of disease did not show any correlation with the number of viable or dead cells. 

Figure 8.

 

Live-dead cell viability assay of flat-mounted ILM specimens removed from a patient with stage IV macular hole. Cell viability was quantified based on a two-color fluorescence assay. (A) Red: nuclei of nonviable cells labeled by propidium iodide; (B) blue: cell nuclei of all cells labeled by Hoechst 33342. (C) Interference and (D) phase-contrast micrographs display the same detail as in (A) and (B) demonstrating dense and homogeneously distributed cells proliferating at the ILM. Original magnification, ×200.

Table 3.

 

Clinical Data of Patients and Cell Counts of Live-Dead Cell Viability Assay

Pt. No.	Age/Sex/Stage of Macular Hole	Usage of Vital Dye (BBG)	Total Area of Specimen (mm2)	Total Cells (n)	Viable Cells (n)	Dead Cells (n)	Viable Cells (%)
1	67/F/III	−	1.9	4	2	2	50
2	71/F/III	+	6.3	25	25	0	100
3*	70/F/IV	−	4.4	2,157	1,831	326	85
4	61/F/III	+	3.6	76	41	35	54
5	66/M/III	+	0.9	7	7	0	100
6	69/F/III	+	0.9	20	19	1	95
7	67/F/III	+	11.1	645	107	538	17
8*	88/M/IV	−	15.8	17,000	11,734	5,266	69
9	74/F/IV	+	3.0	17	16	1	94
10†	67/F/III	−	1.0	12	5	7	42
11†	71/F/III	+	4.6	11	7	4	64
12†	63/F/III	+	3.2	58	28	30	48

*  Macular hole associated with epiretinal gliosis, as diagnosed by biomicroscopy.

†  Preparation after 24 hours.

Specimens that were not immediately incubated after removal during vitrectomy but 24 hours after surgery with storage in balanced saline at +4°C presented with lower percentages of live cells. These specimens showed a live percentage of 42% to 78%, with a mean of 58% (SD 18%). 

Observations by SEM

SEM showed epiretinal cells in all ILM specimens, as single cells and/or multilayers of cells. Collagen fibrils were also seen at the vitreal side of the ILM that were mostly found with enmeshed epiretinal cells. According to our ultrastructural observations, we demonstrate distinct cell phenotypes in Figure 9. 

Figure 9.

 

Ultrastructural observations of diverse cell morphology by SEM. (A) Single, round cells with sparse cytoplasm separately distributed on the vitreal side of the ILM. (B) Epiretinal single cell with irregular short processes near strandlike cell processes embedded in vitreous cortex collagen fibrils. (C) Epiretinal multilayer of flat, polygonal cells with broad interdigitating processes and some small microvilli. (D) A single, stretched cell of irregular shape and numerous elongations on the vitreal side of the ILM. Original magnification: (A) ×600; (B) ×2,000; (C) ×1,500; (D) ×10,000.

First, small, round or oval cells were typically found with sparse cytoplasm, being solitarily distributed as single cells at the vitreal side of the ILM (Fig. 9A). They were sometimes found with short processes variably entangled with a fine collagen fibrillar network (Fig. 9B). These cells were presumably hyalocytes. ²⁶ Polygonal cells, with or without small microvilli and broad interdigitating processes, often formed confluent epiretinal cell multilayers (Fig. 9C). They corresponded with well-established descriptions of glial cell proliferation. ²⁷ Epiretinal cells of irregular shape were seen as single cells and cell multilayers showing various appearances of their cell elongations, with or without contact to collagen fibrils.(Fig. 9D). These cells most probably are microglia. SEM observations can be correlated with interference and phase-contrast microscopic findings. In Figure 10, we present all three microscopic examinations of one specimen removed from a patient with stage III macular hole. The Interference and phase-contrast micrographs of the same area of the specimen showed massive cell proliferation at the ILM (Figs. 10A, 10B). SEM revealed epiretinal cells of polygonal, irregularly shape enmeshed in numerous elongated cell processes at the vitreal side of the ILM. These fine cell processes were arranged as unsorted multilayers extending to the surface of the ILM (Figs. 10C, 10D). 

Figure 10.

 

Specimens of the ILM removed from a patient with stage III macular hole. (A) Interference and (B) phase-contrast micrographs of the same detail show massive cell proliferation at the ILM. SEM revealed (C) epiretinal cells of polygonal, irregular shape enmeshed in numerous elongated cell processes on the vitreal side of the ILM. (D) These fine cell processes are arranged as unsorted multilayers extending to the surface of the ILM. Original magnification: (A, B) ×100; (C) ×5,000; (D) ×15,000.

Observations by Transmission Electron Microscopy

Images of flat-mount preparation procedures and conventional sectioning preparation procedures were correlated in four selected cases of specimens that were found to show large total cell counts with homogeneously distributed cells. Figure 11 presents our findings from a specimen removed from a patient with stage III macular hole. 

Figure 11.

 

Correlation of flat-mount preparation with serial sectioning preparation procedures. Specimen of the ILM removed from a patient with stage III macular hole. (A–D) One half of the specimen was processed for phase-contrast microscopy, interference microscopy, and immunocytochemistry. (E–H) The other half was embedded for conventional light microscopy and transmission microscopy. (A) Cell nuclei staining by DAPI showed large cell count and homogeneously distributed cells at the ILM. Arrowheads: cutting line. (B) Only a small cell cluster was positively marked by immunocytochemical staining. (C) A cell cluster labeled with anti-GFAP and -CD45 in co-localization, but negative for anti-CK8. All other cells outside of the cluster were not labeled at all by this antibody combination. (D) Phase-contrast micrograph combined with DAPI presents same detail as image (C). (E) Light micrograph of the ILM prepared by serial sectioning shows the ILM (*) with epiretinal cell proliferation (arrowhead) and interposition of collagen strand (arrow). (F) Transmission electron micrograph demonstrates the vitreal side of the ILM (*) with native vitreous collagen (arrow) and epiretinal cell proliferation. (G) Epiretinal cell multilayer predominantly composed of myofibroblasts (arrowhead) characterized by aggregates of subplasmalemmal cytoplasmic filaments (white arrowhead, higher magnification inset). (H) Multilayered cell proliferation situated on a collagen strand (arrow) composed of fibroblast-like cells with abundant rough endoplasmic reticulum, mitochondria, and absence of intracytoplasmic filaments. Arrowhead, higher magnification inset: rough endoplasmic reticulum. Original magnification: (A, B) ×50; (C, D) ×400; (E) ×1000; (F) ×1800; (G, H) ×4800.

Irrespective of the stage of disease, by serial sectioning, we found cell multilayers at the vitreal side of the ILM in all specimens. Native vitreous collagen was always present and was characterized by a fibril diameter of 16 nm and regular arrangement. In two specimens, vitreous collagen fibrils were sorted as collagen strands. Newly formed collagen, irregularly arranged and embedded in proliferated cells, was seen in specimens from two patients. Myofibroblasts and fibroblasts were the predominant cell types. Myofibroblasts were characterized by aggregates of subplasmalemmal cytoplasmic filaments, rough endoplasmic reticulum, and fusiform nucleus. The fibroblasts were characterized by abundant rough endoplasmic reticulum, prominent Golgi complexes, and absence of intracytoplasmic filaments. Retinal pigment epithelial cells and macrophages were not found. 

Discussion

In this study, epiretinal cells were present in all stages of macular holes, and cell proliferation was seen on all surgically excised ILM specimens examined by flat-mount preparation techniques. Flat-mount preparation allows en face visualization of the whole ILM specimen. In contrast to conventional sectioning preparation procedures, the flat-mount preparation method detects even single small cell cluster formations that might be missed by serial-sectioning preparation methods of ILM specimens. Therefore, this study presents new details on cell density and cell distribution in macular holes. It was previously hypothesized that epiretinal cell proliferation represents a secondary event in macular hole pathogenesis once vitreoretinal traction has occurred, causing a full-thickness macular defect. ^9,18 Cell proliferation was thought to increase with the stage of disease. However, given the fact that cell proliferation was seen in all stage II macular holes of this study and the fact that we did not find a difference of significance in cell density and cell distribution patterns when comparing all stages of macular hole, we hypothesize that cell migration and proliferation occur early in the course of this disease. Epiretinal cell proliferation may be substantially involved in the development of idiopathic macular holes. 

Epiretinal cells presented with various phenotypes. Corresponding to the commonly used ultrastructural characteristics of epiretinal cell types, ¹⁹ transmission electron microscopy mostly revealed myofibroblast-like cells and native vitreous collagen at the vitreal side of the ILM, which is in accordance with previous ultrastructural evaluations of the vitreoretinal interface in idiopathic macular holes. ^18,20 SEM and indirect immunofluorescence examinations demonstrated cells with features of glial cells and hyalocytes that were similarly described in the past. ^26,27 However, by immunocytochemistry of flat-mounted ILM specimens, we found a proportion of epiretinal cells showing simultaneous expression of GFAP and hyalocyte cell markers, such as CD45 and CD64, which have never been shown in humans before. Further cell marker co-localizations were occasionally detected, such as GFAP and CD68, GFAP and CD90, and GFAP and CRALBP. 

Regarding our findings of cell marker co-localization, reconsideration of GFAP expression patterns in epiretinal cell proliferation may become necessary. In the past, GFAP was believed to be unique to glia. ²⁸ Therefore, immunolabeling against this intermediate filament protein appeared to be specific for detecting glial cells in epiretinal membranes. It was assumed that all glia-derived cells in epiretinal tissue would be GFAP-positive and that GFAP-negative cells would not be of glial origin. However, more recent studies demonstrated positive GFAP staining in cell populations other than glia. Several species were found to present GFAP-positive hyalocytes, including porcine, ²⁹ pectineal, ³⁰ and bovine hyalocyte cell lines. ³¹ This study provides data on co-localization of GFAP and hyalocyte cell markers in humans. According to our results, we postulate that a proportion of the GFAP-positive epiretinal cells in idiopathic macular holes are not of glial origin. We hypothesize that these cells constitute hyalocytes rather than glial cells. Hyalocytes were shown to be immunoreactive for CD45 and CD64, to belong to the monocyte/macrophage lineage and to derive from bone marrow. ^15,32 They were found to be replaced in the vitreous under physiological conditions within 7 months. ³³ This may be an explanation of why we found a very heterogeneous population of cells with immunoreactivity for hyalocyte cell markers. From an ultrastructural point, hyalocytes are described as resembling macrophages that usually possess a lobulated nucleus, a well-developed Golgi complex, a rough and smooth endoplasmic reticulum, and many large lysosomal granules and phagosomes. ^3,26,33 However, different morphologic features of hyalocytes can be found in different cells of the same hyalocyte population. ³⁴ It is not known, whether this heterogeneity is related to different origins of cells or to different states of cell metabolism and activity. Concerning the co-localization of GFAP and hyalocyte cell markers, it remains questionable whether these cells, presumably hyalocytes, are immunoreactive due to an endogenous expression of GFAP or due to phagocytosis of GFAP-positive debris or apoptotic cells. If GFAP labeling in hyalocytes results from endogenous expression, one might hypothesize that these cells have some progenitor potential. If GFAP labeling results from phagocytic activity, these cells may have engulfed Müller cell end feet, activated Müller cells, apoptotic microglia, or astrocytes. 

Regarding epiretinal cells being immunoreactive for GFAP but not for hyalocyte cell markers, it is conceivable that these cells constitute activated and migrated retinal Müller glial cells. Retinal Müller cells are thought to become activated by anteroposterior vitreoretinal traction in the context of incomplete posterior vitreous detachment, then to migrate and proliferate, thereby driving epiretinal cell proliferation at the ILM. ⁶ Since pores of the ILM have been reported to be a rare finding, ³⁵ it remains speculative whether retinal Müller cells migrate through ILM-pores from the retinal onto the vitreal side of the ILM. However, retinal Müller cells were shown to upregulate GFAP in response to retinal injury and cell activation, although they do not label with anti-GFAP under normal conditions except at their endfeet. ^6,7,11,12 In this study, we found some cells with double labeling for anti-GFAP and -CRALBP that most probably represent Müller glial cells, as previously shown. ³⁶ Furthermore, positive anti-GFAP and -vimentin double labeling is an expected co-localization of activated Müller cells that was found in our series as well. In addition, we found GFAP and vimentin co-localized with CD90, which also suggests the presence of Müller cells because CD90 was reported to become strongly expressed by Müller cells in retinal ganglion cell death and absence of retinal neurons. ³⁷ CD90 is thought to be expressed by Müller cells when they loose contact with neurons, as occurs in proliferation and migration. Thus, our findings support the hypothesis that Müller glial cells are an important component of epiretinal cell proliferation in idiopathic macular hole. 

However, Müller glial cells are not the only type of glial cells involved in epiretinal cell proliferation. By ultrastructural and immunocytochemical analysis, we found evidence of astrocytes and microglial cells. Microglial cells of the adult human retina can possess characteristics of macrophages and were previously presented to be immunoreactive for CD68. ³⁸  

As mentioned, a proportion of cells was not labeled with any cell marker combination used in the study. Consequently, these cells were nonviable, or they constituted dedifferentiated progenitor cells, or they were transdifferentiated and therefore were not detected with the commonly used immunocytochemical markers. Transdifferentiation of cells with changes in phenotype and antigen expression still complicates the determination of cell type and the cells' origin. Myofibroblast-like transdifferentiation with positive αSMA expression was shown for both hyalocytes and glial cells. ^31,32,39 In our series, ultrastructural examinations revealed confluent multilayers of glial cells by SEM and myofibroblast-like appearance of epiretinal cells by transmission electron microscopy. Moreover, we frequently found αSMA-positive cells in direct proximity, but never in co-localization, with GFAP and with hyalocyte cell markers. This is in accordance with other studies reporting on loss of GFAP with coincident gains of αSMA immunoreactivity. ⁴⁰ Since we found αSMA immunoreactivity in all macular hole stages, transdifferentiation appears to take place early in disease development. Consequently, we hypothesize that epiretinal cell proliferation in all macular hole stages possesses the potential to exert tangential traction at the vitreoretinal interface. Based on our findings of glial cells and hyalocytes as predominating cell types in macular holes, we presume that both glial and hyalocyte cell populations are possible candidates for myofibroblast-like transdifferentiation. 

In summary, cellular proliferation at the ILM appears to be present in all idiopathic macular holes, irrespective of the stage of disease. Our results strongly suggest a substantial involvement of cell migration and proliferation in the course of idiopathic macular hole development. Ultrastructural and immunocytochemical findings point to the predominance of glial cells and hyalocytes. However, according to our results, expression of the intermediate filament GFAP in epiretinal cells needs to be reconsidered, since GFAP-positive immunostaining alone no longer allows determination with certainty of whether cells are of retinal glial origin. Given the co-localization of GFAP and hyalocyte cell markers, some thoughts concerning the origin of cells should be taken into consideration. First, these cells most probably represent hyalocytes. Second, positive GFAP labeling in these hyalocytes may result either from phagocytic activity of glial cell debris and apoptotic glial cells or from endogenous expression representing some progenitor potential of this cell subpopulation. By immunocytochemistry alone, it is not possible to differentiate between these hypotheses. Further investigations are needed to elucidate whether hyalocytes contain GFAP from other origin or they express GFAP endogenously. 

Footnotes

 Presented in part at the Annual Meeting of the Deutsche Ophthalmologische Gesellschaft, Berlin, Germany, 2009, and at the World Congress of Ophthalmology, Berlin, Germany, 2010.

Footnotes

 Part of the data will be presented in an as yet unpublished thesis.

Footnotes

 Disclosure: R.G. Schumann, None; K.H. Eibl, None; F. Zhao, None; M. Scheerbaum, None; R. Scheler, None; M.M. Schaumberger, None; H. Wehnes, None; A.K. Walch, None; C. Haritoglou, None; A. Kampik, None; A. Gandorfer, None