Gli1 Expression in Human Epiretinal Membranes

Abstract

Purpose: We evaluate the expression of Gli1 in human epiretinal membranes (ERM) and correlate this with clinical data.

Methods: We prospectively recruited patients with ERM. A total of 33 human ERM specimens were immunolabeled with anti-Gli1 antibody and the number of total cells/hyperfield (HF), Gli1(+) cells/HF, and the percentage of Gli1(+) cells/total cells were calculated. We evaluated the interrelationship of cellular properties and clinical findings, such as presence of diabetic retinopathy (DR), retinal breaks, intraocular inflammation, central foveal thickness, maximal retinal thickness, retinal contraction, lamellar holes, pseudoholes, the attenuation or absence of an inner segment/outer segment (IS/OS) junction/external limiting membrane (ELM), cystic changes, and paravascular inner retinal defects.

Results: Among 33 specimens, 25 specimens (75.8%) showed nuclear Gli1 expression. The mean Gli1(+) cells/total cells was 54.0 ± 36.7% (range, 0%–92.8%). There was significantly higher expression of Gli1(+) cells in ERM specimens from patients with DR (P = 0.014), and lower expression from patients with retinal breaks (P = 0.022). Epiretinal membrane specimens from patients with alteration of IS/OS junction/ELM or cystic changes on OCT showed higher percentage of Gli1(+) cells/total cells.

Conclusions: Gli1 expression was detected in most ERM specimens. Patients who had DR or OCT findings indicating chronic retinal insults showed higher Gli1 expression. Gli1 may have a role in the pathogenesis of ERM after chronic retinal insults.

Epiretinal membrane (ERM) is a condition of cellular proliferation on the inner retinal surface. It has a wide spectrum of presentations and symptoms that comprise a large part of retinal diseases prevalent among aged people. According to large population studies, the overall prevalence of ERM was 7% to 11.8%, and the 5-year incidence was 5.3%.^1–3 Despite the high prevalence of the disease, its pathophysiology, including molecular pathways involved in the initiation and progression of ERM, has not been fully elucidated. 

Intraocular fibrotic changes, such as ERM and proliferative vitreoretinopathy, have long been speculated to be maladapted wound repair processes of retinal cells driven by various kinds of growth factors, cytokines and other signals.^4,5 Various types of cells, such as astrocytes, RPE cells, macrophages, and Müller glial cells, have been implicated in ERM formation.⁶ 

A recent study showed that Gli1(+) perivascular cells became activated and committed to the myofibroblast lineage, causing fibrotic changes after injuries in many kinds of organs, including the bone marrow, muscle, heart, lungs, liver, and kidneys.⁷ Although the role of Gli1 in the injured human retina has not yet been elucidated, it has been shown that Gli1 is involved in the formation of retinal progenitor and Müller glial cells, driven by sonic hedgehog (Shh) from ganglion cells in the developing rodent retina.^8–10 Recently, Zhao et al.¹¹ found that the expression of Shh and Gli1 was upregulated in the hyperglycemic adult rodent retina, and that blockage of the Shh pathway by cyclopamine upregulated glial fibrillary acidic protein (GFAP) expression, suggesting a potential neuroprotective role of Shh-Gli signaling in the injured retina.¹¹ 

Therefore, we hypothesized that Shh-Gli1 signaling may have a role in the development and progression of human ERM as a part of a maladapted retinal wound repair process. In the present study, we evaluated Gli1 expression in human ERM specimens and analyzed any clinical correlations. 

Methods

A prospective, open-label, nonrandomized interventional study was performed on patients who were scheduled for pars plana vitrectomy (PPV) for ERM removal. Institutional Review Board (IRB)/Ethics Committee approval was obtained before patient recruitment. The study protocol adhered to the tenets of the Declaration of Helsinki. All participants signed a consent form after a detailed explanation of the study design, associated surgical procedures for scientific purposes, and adjuvant imaging procedures. 

Patients and Clinical Data Acquisition

Patients who were scheduled for PPV for primary and secondary ERM removal were recruited from Seoul Saint Mary's Hospital, The Catholic University of Korea, between March 1, 2014 and February 28, 2015. Exclusion criteria were any pharmacologic intervention on the study eye within 6 months, panretinal photocoagulation on the study eye within 6 months, any pharmacologic intervention on the fellow eye within 3 months, any history of intraocular surgery other than uncomplicated cataract surgery on the study eye, and any history of ocular trauma on the study eye. 

All patients underwent postoperative follow-up for 6 months. Comprehensive ocular examinations, including best corrected visual acuity (BCVA), IOP measurement, slit-lamp examinations, color fundus photography, and an optical coherence tomography (OCT) scan by Spectralis spectral-domain (SD) OCT version 5 (Heidelberg Engineering, Heidelberg, Germany), were performed before PPV. Best corrected visual acuity, IOP, slit-lamp examinations, and SD-OCT were evaluated postoperatively after 1, 2, and 6 months. Best corrected visual acuity was measured using the decimal system, and then converted to logMAR units for statistical analysis. Demographic findings, such as age, sex, presence of diabetes, hypertension, diabetic retinopathy (DR), retinal breaks, and intraocular inflammation, were recorded. 

Various OCT findings were collected, such as central foveal thickness, maximal retinal thickness, retinal contraction, the attenuation or absence of the inner segment/outer segment (IS/OS) junction, attenuation or absence of the external limiting membrane (ELM), and the presence of intraretinal cysts,¹² lamellar holes, pseudoholes,¹³ and paravascular inner retinal defects, as previously described¹⁴ (Fig. 1). The diameters of the IS/OS junction defect and ELM defect were measured when the absence of those parameters was detected.¹⁵ Two masked investigators (SJ, JB) interpreted the OCT images. When there was disagreement, the third investigator was consulted for the final decision (WKL). 

Figure 1

 

Representative B-scan SD-OCT images identifying parameters used in the present study. (A) Central foveal thickness (arrow) was measured from RPE band to the internal limiting membrane at the base of the central foveal pit. Maximum retinal thickness (dotted arrow) was the maximum height of the retina detected within 3 mm width zone. Green box indicates 3 mm width zone. (B) Lamellar hole was diagnosed when there was an irregular foveal contour with inner foveal defects and schisis, but with maintenance of photoreceptor layer. (C) Retinal contraction was diagnosed when there was a wrinkling of retina under ERM. (D) Pseudohole was diagnosed when there were invaginated foveal edges with concomitant ERM showing central opening (arrow). (E) Intraretinal cysts were defined when there were ovoid hyporeflective lesions within the retina (arrow). The integrities of ELM and IS/OS junction were evaluated within 3 mm width zone, and graded according to the intensity. (dotted arrow). The line was graded as attenuated if the intensity of the line within 3 mm width zone was between 10% and 50% of maximum intensity, and absent if the intensity was below 10%. (F) Paravascular inner retinal defect was diagnosed when there were defects of inner retina adjacent to the major retinal vessels.

Surgical Procedures

Povidone-iodine 5% (wt/vol) eye drops were dropped into the conjunctival sac and lid margin after the application of a sterile drape and insertion of a lid speculum. A 23-gauge transconjunctival sutureless three-port PPV using a 23-gauge trocar and cannula (Alcon Laboratories, Inc., Fort Worth, TX, USA; Dutch Ophthalmic Research Center, Zuidland, The Netherlands) was performed by an experienced vitreoretinal surgeon (WKL). An ERM sample was removed by intraocular forceps and immediately transferred and fixed in 4% paraformaldehyde solution. If more than one ERM specimen per eye was taken, all samples were fixed and analyzed. After removal of the ERM, indocyanine green dye was injected for visualization and removal of the internal limiting membrane. No additional intraoperative pharmacologic injection was administered before and after ERM removal. A partial air-fluid exchange was performed in every patient to check leakage from sclerotomy sites and to facilitate their sealing. Topical antibiotics were prescribed and applied four times a day for 2 weeks after surgery. 

Immunohistochemistry of ERM Specimens

Epiretinal membrane samples were immediately fixed in a 4% (vol/vol) paraformaldehyde solution and washed three times with 0.1 M PBS (pH 7.4). After permeabilization in 0.5% (vol/vol) Triton X-100 for 30 minutes and blocking in 10% normal goat serum for 2 hours, ERM samples were incubated with rat anti-Gli1 antibody (1:50; R&D Systems, Wiesbaden, Germany) at 4°C overnight and washed three times with 0.1 M PBS. Each ERM sample then was incubated with anti-rat antibody conjugated with rhodamine (1:500; Santa Cruz Biotechnology, Santa Cruz, CA, USA) at room temperature (RT) for 2 hours and washed three times. An identical staining procedure was undertaken to double-stain samples with rabbit anti-GS antibody (1:200; Abcam, Cambridge, UK)/anti-α-SMA antibody (1:100; Abcam)/anti-Vimentin (1:250; Abcam) and anti-rabbit antibody conjugated with FITC (1:500; Santa Cruz Biotechnology). Additional triple staining was done with anti-CD31 (PECAM) and allophyocyanin (APC) for an ERM sample which showed perivascular staining pattern. All staining procedures were performed with ERM specimens placed in 2.0 mL microcentrifuge tubes. We were careful not to break or aspirate each sample when staining or washing by pipetting because ERM specimens easily adhered to the inner surface of pipet tips. After staining, ERM samples were transferred onto glass slides and mounted with ProLong Gold Antifade Mountant containing 4′,6-diamidine-2′-phenylindole dihydrochloride (DAPI; Invitrogen, Carlsbad, CA, USA). Mouse IgG isotypes matching those of the primary antibodies (Sigma-Aldrich Corp., St. Louis, MO, USA) were used as negative controls. Images were taken by confocal microscope (LSM510; Carl Zeiss, Oberkochen, Germany). 

To quantify cells in ERM samples, three random images were taken at ×200 final magnification and the average number for three images was calculated and used for further analyses. The number of total cells per hyperfield (HF) was measured by the detection of DAPI-positive cells in each image. The number of Gli1(+) cells was measured by the detection of rhodamine-positive cells in each image. Finally, the percentage of Gli1(+) cells per total cells in each hyperfield was calculated. All measuring procedures were performed blinded to clinical data or OCT findings by two investigators (SJ, JB), and the average number from each investigator was used for further analysis. 

Isolation and Culture of Cells From ERM

Epiretinal membrane samples were immediately transferred and incubated with dispase (Sigma-Aldrich Corp.) for 30 minutes at 37°C in 5% CO₂, and then incubated with trypsin–EDTA (Invitrogen) for 5 minutes at 37°C with 5% CO₂. After briefly pipetting, Dulbecco's modified Eagle's medium (DMEM)/F12 (Invitrogen) supplemented with 10% fetal bovine serum (Invitrogen) was added and samples centrifuged for 5 minutes at 300g, 4°C. Cells were seeded onto fibronectin-coated (10 μg/cm²) tissue culture dishes and incubated for at least 21 days. Cultured cells were stained with anti-GS antibody (1:200, Abcam), anti-CRALBP antibody (1:1000, Abcam), anti-ZO-1 antibody (1:200, Invitrogen), and anti-RPE65 antibody (1:400, Millipore, MA, USA) for identification. 

To assess the effect of the Shh-Gli1 signaling pathway on ERM-derived cells, cells at P1 were seeded at a density of 1 × 10⁵ cells/ml in 6-well plate and treated with human Shh (100 ng/mL; Sigma-Aldrich Corp) with or without a selective Shh antagonist Cyclopamine (2.5 μM; Tocris Bioscience, Ellisville, MO, USA), or a small molecule Gli1-mediated transcription inhibitor GANT61 (5 μM; Tocris Bioscience) for 3 days. 

Real-Time Quantitative PCR (qPCR) and Western Blotting

Total RNA was extracted using TRIzol reagent (Invitrogen) according to the manufacturer's protocol. Quantification of total RNA was performed using a NanoDrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). Polymerase chain reactions were performed in triplicate using SYBR Premix Ex Taq II (TaKaRa, Shiga, Japan) and specific primers (Supplementary Material). The results were expressed as the fold-difference normalized to GAPDH using the ΔΔCt method. 

Cell lysates were subjected to SDS-PAGE, transferred to membranes, and incubated with anti-SNAI antibody (1:1000; Abcam) and anti-GAPDH (1:5000; Abcam). The membranes then were incubated with horseradish peroxidase-conjugated secondary antibodies, and proteins were visualized using a chemiluminescence substrate (Thermo Fisher Scientific). 

Statistical Analysis

We used SPSS version 15.0 for Windows (SPSS, Inc., Chicago, IL, USA) for statistical analyses. A Shapiro–Wilk test was used to assess the normality of the distribution. Pearson correlation coefficients or Spearman rank correlation coefficients were used to assess the association between continuous variables according to the normality of distribution. A 2-tailed Student's t-test was used to assess differences in the number of total cells, Gli1(+) cells, and the percentage of Gli1(+) cells/total cells according to binary variables. Independent variables significantly associated with the BCVA in univariate analyses (P < 0.05), and potentially confounding parameters were included as independent covariables in multivariate analyses using multiple regression analysis. Continuous variables were reported as mean ± SD (range). A P value < 0.05 was considered statistically significant. 

Results

Baseline Demographic and Ocular Characteristics

We enrolled 42 consecutive patients with ERM scheduled for PPV and ERM removal in this prospective interventional study. Among them, 33 ERM specimens were used for Gli1/GS immunohistochemistry. All patients underwent PPV and ERM removal as scheduled, and were followed for 6 months. Baseline demographic characteristics of the enrolled patients for Gli1 staining are described in Table 1. A total of 20 female and 13 male patients with a mean age of 64.41 ± 7.62 years was included. Five patients (15.15%) had diabetes, while 13 patients (39.39%) had hypertension that was controlled by oral medications. Ten patients (30.30%) had secondary ERM; three patients (9.09%) had diabetic retinopathy, three patients (9.09%) had a history of retinal tears or holes, while four patients (12.12%) had an intraocular inflammatory disease. Seven patients were pseudophakic, and 26 patients were phakic at the time of the PPV. Cataract surgery was done at the time of the PPV in 23 patients, while it was performed postoperatively in three patients. The mean BCVA of logMAR was 0.62 ± 0.39 before PPV, which significantly improved to 0.45 ± 0.37, 0.37 ± 0.30, and 0.35 ± 0.31, after 1, 2, and 6 months postoperatively (P = 0.003, P < 0.001, and P < 0.001, respectively). 

Table 1

 

Patient Characteristics and their Association With Cellular Properties in ERM

 

Cell Density and BCVA

Epiretinal membrane specimens showed various ranges of cell densities, from 20 to 284 cells/HF (mean ± SD, 110.2 ± 68.7 cells/HPF). We evaluated whether there is any associated factor for cellular densities. Of interest, univariate regression analysis revealed that the preoperative and postoperative 1 m BCVA were associated with the cell densities (P = 0.020 and P = 0.050, respectively). Higher cell densities correlated with worse preoperative and postoperative 1-month BCVAs. These correlations were not significant at 2 and 6 months postoperatively (P = 0.157, and P = 0.150, respectively). Age, sex, axial length, and presence of diabetes, hypertension, and preexisting intraocular diseases, such as DR, retinal breaks, or intraocular inflammation, were not significantly associated with the cell densities (Table 1). 

To evaluate whether the cell densities were independently associated with the preoperative and postoperative 1 m BCVA, we performed an additional regression analysis. A univariate regression analysis revealed that IS/OS defect diameter was associated with the preoperative and postoperative 1 m BCVA (P = 0.033, and P = 0.002, respectively). None of the other demographic characteristics or OCT parameters was associated with the preoperative and postoperative 1 m BCVA. Multivariate regression analysis revealed that the cell density was significantly associated with the preoperative BCVA after adjustment of IS/OS defect diameter, age, sex, and presence of DR (R² = 0.162, P = 0.020; Fig. 2A). In the other hand, the IS/OS defect dimeter was significantly associated with postoperative 1 m BCVA after adjustment of cell densities, age, sex, presence of DR (R² = 0.278, P = 0.002; Fig. 2B). 

Figure 2

 

Relationship between BCVA and total cell numbers/HF. (A) Total cell numbers/HF showed a positive relationship with preoperative BCVA (P = 0.020, R² = 0.162). (B) Total cell numbers/HF showed a positive relationship with postoperative BCVA at 1 month (P = 0.050, R² = 0.118). (C) A 65-year-old male patient with ERM associated with an intraretinal cyst, whose preoperative logMAR BCVA was 0.4 (total cell number/HF 129.3 cells/HF, axial length 24.73 mm). (D) A 60-year-old male patient with ERM associated with an intraretinal cyst, whose preoperative logMAR BCVA was 0.6 (total cell number/HF 215.6 cells/HF, axial length 24.59 mm). Note that their BCVAs and total cell densities were distinct when their OCT findings, such as intraretinal cystic changes involving the inner and outer nuclear layers, and the degree of outer retinal defects were similar.

Figures 2C and 2D bottom row shows representative images of selected patients who showed different preoperative BCVAs despite similar OCT findings (logMAR 0.4 and 0.6 for Figs. 2C, 2D, respectively). These two patients showed different visual acuities, while the OCT findings, such as intraretinal cysts and outer retinal structures, were similar. Of note, the ERM specimen from the patient with the worse BCVA showed a higher cellular density than the other ERM sample, suggesting a high cell density is a potential causative factor for a poor preoperative BCVA. 

Gli1 Expression and Demographic Findings

Among 33 specimens, 25 (75.8%) showed nuclear Gli1 expression. Gli1(+) cells displayed a heterogeneous distribution among specimens, from 0 to 197 cells/HF (mean ± SD, 56.2 ± 59.7/HF). There was no significant difference in the Gli1(+) cells/HF and Gli1(+) cells/total cells between patients with primary and secondary ERMs (P = 0.232 and 0.268, respectively). Epiretinal membrane specimens from patients with DR showed a significantly higher expression of Gli1(+) cells/HF (Fig. 3A), when the number of total cells/HF showed no difference, suggesting a potential role for Gli1(+) cells in DR-associated ERM formation (P = 0.022 and P = 0.133, respectively). Figures 3B and 3C show representative images of ERM specimens from patients with DR. Many Gli1(+) cells costained with the Müller glial cell marker, GS (Fig. 3B). Interestingly, Gli1 expression was detected in a perivascular area in another part of the same specimen (Fig. 3C), supporting our hypothesis that perivascular Gli1(+) cells may be involved the pathogenesis of intraocular fibrotic changes. 

Figure 3

 

Gli1 expression in patients with DR or retinal holes. (A) The number of Gli1(+) cells/HF was significantly higher in patients with DR (P = 0.022). The percentage of Gli1(+) cells/total cells was significantly higher in patients with DR (P = 0.014). (B) A representative image of an ERM specimen from a patient with DR. (C) An image of an ERM specimen from the same patient. Note the Gli1(+) and GS (−) cells showing a vessel-like pattern, which was stained with CD31. (D) The number of Gli1(+) cells/HF were significantly lower in patients with retinal holes or tears (P = 0.001). The percentage of Gli1(+) cells/total cells was significantly lower in patients with retinal holes or tears (P = 0.022). (E) ERM in a patient with retinal holes (arrows indicate retinal holes with a laser scar). (F) Image of an ERM specimen from a patient with retinal holes. Note that there is no staining of Gli1, nor GS.

The number of Gli1(+) cells/HF was significantly lower in ERMs from patients with retinal breaks (Fig. 3D), when the number of total cells/HF was the same (P = 0.001 and P = 0.242, respectively). Figures 3E and 3F shows a representative image of an ERM specimen from a patient with retinal holes. Note the lack of expression of Gli1 or GS in this specimen (Fig. 3F). 

Gli1 Expression and OCT Findings

Epiretinal membrane specimens from patients with intraretinal cysts on OCT showed a significantly higher number of Gli1(+) cells/HF, when there was no significant difference in the total cell numbers/HF (P = 0.049 and P = 0.582, respectively; Table 2). Epiretinal membrane samples with abnormal outer retinal OCT findings, such as the absence of an IS/OS line, or the attenuation or absence of ELM, showed a higher percentage of Gli1(+) cells as a proportion of total cell numbers (P = 0.001, P = 0.002, and P = 0.001, respectively; Table 2, Fig. 4). 

Table 2

 

OCT Characteristics and Their Association With Cellular Properties in ERM

 

Figure 4

 

Gli1 expression according to OCT findings. (A) The number of Gli1(+) cells/HF and percentage of Gli1(+) cells/total cells were significantly higher in patients with intraretinal cysts (P = 0.049 and P = 0.001, respectively). (B) The percentage of Gli1(+) cells/total cells was significantly higher in patients showing an absence of IS/OS (P = 0.001), while the number of Gli1(+) cells/HF did not show a significant difference (P = 0.158). (C) The percentage of Gli1(+) cells/total cells was significantly higher in patients showing an attenuation of ELM (P = 0.002), while the number of Gli1(+) cells/HF did not show a significant difference (P = 0.065). (D) The percentage of Gli1(+) cells/total cells was significantly higher in patients showing an absence of ELM (P = 0.001), while the number of Gli1(+) cells/HF did not show a significant difference (P = 0.054).

Five patients (15.15%) showed pseudoholes, and one patient displayed a lamellar hole (3.03%). Epiretinal membrane specimens from patients with pseudoholes showed no difference in total cell or Gli1(+) cell numbers/HF or percentage of Gli1(+) cells when compared to those from patients without pseudohole (Table 2; Fig. 5A). An ERM specimen from a patient with lamellar hole showed a high density of cells. Also, localized clusters of Gli1(+) cells were detected and these cells showed costaining with GS, supporting the previous theory that Müller glial cells contribute to the formation of lamellar hole-associated epiretinal proliferation (Fig. 5B).^16,17 

Figure 5

 

A representative image of an ERM specimen from a pseudohole and lamellar hole-associated epiretinal proliferation (LHEP). (A) Representative images from a 64-year-old female with ERM-associated pseudoholes (axial length 23.90 mm). Yellow arrows show Gli1/GS double positive cells. (B) Representative images from a 74-year-old male with LHEP (axial length 23.75 mm). Note that the cluster of cells stained for Gli1 and/or GS.

Response of ERM-Derived Cells to Shh

To explore whether Gli1(+) cells were involved in the fibrotic changes, we stained ERM specimens with Gli1 and fibroblast markers (Fig. 6A). We found that the Gli1(+) cells expressed α-SMA and vimentin, suggesting the Gli1(+) cells were involved in the fibrotic changes. Gli1 is transcriptionally induced by Shh signaling, and is not expressed in its absence.¹⁸ We observed whether ERM cells responded to exogenous Shh treatment. Cells from the ERM were dissociated with dispase and plated on a fibronectin-coated cell culture dish (Fig. 6B). Attached cells proliferated in primary culture P0 to P1, but could not be successfully subcultured further. Therefore, we used P0 or P1 cells to evaluate the response to Shh. Cultured P1 cells expressed GS, a Müller glial cell marker and CRALBP, which was expressed in Müller glial cells and RPE cells, but not ZO-1 or RPE65, which are expressed in RPE cells; these data suggested that P1 cultured cells are Müller glial cells. Sonic hedgehog treatment (100 ng/mL for 3 days) significantly upregulated Gli1 mRNA expression, the progenitor marker SOX2, and the epithelial-mesenchymal transition (EMT) marker SNAI1 when compared to a dimethyl sulfoxide (DMSO) control, suggesting a potential role for the Shh-Gli1 signaling axis in the pathogenesis of ERM (Fig. 6C). To understand the role of the Shh-Gli1 signaling axis in the EMT process, we evaluated the response of ERM-derived cells to Shh by blocking the Shh-Gli1 signaling axis with the selective Shh antagonist, cyclopamine, or the Gli1 inhibitor, GANT61. Exogenous Shh upregulated SNAI1 at the protein level, but the expression of SNAI1 was downregulated by blocking Shh signaling via cyclopamine (Fig. 6D). Blocking Gli1 also resulted in the downregulation of SNAI1, suggesting Gli1 mediated the Shh-induced EMT process in ERM-derived Müller glial cells. 

Figure 6

 

Culture of ERM cells and response to Shh treatment. (A) Double staining of Gli1 and fibroblastic markers in ERM specimens. (B) Light microscopy of dissociated cells from an ERM specimen which were attached on a fibronectin-coated tissue culture slide. These cells were positive for GS and CRALBP but negative for ZO-1 and RPE65, suggesting these cells are Müller glial cells. (C) Shh treatment (100 ng/mL) for 3 days significantly upregulated Gli1 mRNA expression, along with SNAI1 and SOX2 mRNA expression. Relative folds expression levels for indicated genes were determined by RQ-PCR with normalization to GAPDH levels (mean ± SEM, from 3 experiments, *P < 0.05). (D) Western blot analysis of SNAI1 expression after Shh with or without Cyclopamine (SMO antagonist) and GANT61 (Gli1 antagonist). Representative blot from 3 independent experiments are shown.

Discussion

In the present study, we found various degrees of Gli1(+) cells in ERM. Previous studies revealed that retinal Müller glial cells are involved in the development and progression of ERM.^6,19,20 We also found GS(+) Müller glial cells made up various proportions of the overall cell population, and part of these cells costained with Gli1. Müller glial cells can act as a source of proliferating progenitors to regenerate neurons in birds,¹⁸ zebrafish,²¹ and rodents,^22,23 and which can be activated by Shh signaling.²⁴ While it is not likely that Müller glial cells can regenerate retinal cells in humans, Müller glial cells respond to retinal injury by changing their morphology and biochemical activity during so-called reactive gliosis.²⁵ These processes can be beneficial to neurons by preventing glutamate toxicity and releasing various neuroprotective factors. However, prolonged gliosis may interfere with normal retinal function,²⁶ as seen in ERM. Based on our data, it seems that Shh–Gli1 signaling has a role in the activation of retinal cells, including Müller glial cells, and, therefore, ERM formation. 

Patients with DR showed significantly higher Gli1 expression. In addition, patients showing intraretinal cystic and outer retinal changes, such as a disrupted IS/OS or ELM on OCT, exhibited higher Gli1 expression, suggesting chronic retinal insults were associated with Gli1 expression. In addition, in vitro study revealed that Shh-activated ERM cells exhibited increased mRNA expressions of Gli1 and the retinal progenitor cell marker, SOX2, and the EMT marker, SNAI1. The upregulation of SNAI1 by Shh was confirmed at the protein level, and these changes were diminished by blocking either Shh and Gli1. Based on these data, it seems that the secretion of Shh from chronic retinal insult activates retinal cells, including Müller glial cells, by intracellular Gli1 upregulation and eventually contributes to the formation of ERM. 

While many of the Gli1(+) cells also were positive for the Müller glial cell marker GS, Gli1(+)/GS(−) cells also were present. Due to the small size of ERM specimens, we were unable to evaluate other cell markers for such cells as RPE cells, macrophages, or astrocytes. Of note, the ERM-derived cells that proliferated in culture expressed Müller glial cell markers. It is possible that other types of cells initially attached to the culture dish, but could not proliferate successfully due to the difference in total cell number or proliferative activity as compared to Müller glial cells. 

Gli1 staining was detected in a perivascular pattern in one ERM specimen obtained from a patient with DR. It has been shown that perivascular Gli1(+) cells undergo proliferation and differentiation into myofibroblasts after injury in many organs.⁷ The same mechanism would be involved in the pathogenesis of ERM, which is bona fide intraocular fibrosis. Further study is needed to identify such Gli1(+)/GS(−) cell populations and their role in ERM formation. 

The presence of retinal holes or tears is a frequent cause of secondary ERM formation. In the present study, lower Gli1 expression was observed in ERM specimens obtained from patients with retinal holes or tears, suggesting distinct pathophysiologic mechanisms within secondary ERMs. It has been speculated that RPE cells migrating through retinal breaks contribute to the formation of ERM in patients with retinal holes or tears. Our results supported the hypothesis that ERMs with retinal holes and tears result primarily from migrated RPE cells, unlike ERMs with a DR or lamellar hole, which are likely formed by activated Müller glial cells. 

Interestingly, the total cell number/HF was associated with preoperative BCVA, but such differences were not significant thereafter. We speculated that the large number of cells and the extracellular matrix from these cells would block light and contribute to poor preoperative BCVA in our patients. Such findings will contribute to an understanding of those patients whose preoperative visual acuity is not explained by other OCT findings as seen in Figure 2. 

Epiretinal membrane is a complex disease entity that is caused by various retinal insults and is comprised of multiple cell types. The Gli1 expression may reflect a tissue damage response resulting from destruction of retinal structure by ERM, rather than a causal factor for ERM. An exploration of the cells consisting of ERM can help us understand the complex pathophysiologic mechanisms behind this disease. Further study with a larger sample size would aid in understanding the role of Gli1 in the pathogenesis of ERM. An increased understanding of ERM pathophysiology and the molecular pathways involved may lead to new therapeutic approaches. 

Acknowledgments

Supported by Research Fund of Seoul St. Mary's Hospital, The Catholic University of Korea. 

Disclosure: S. Jeon, None; J. Baek, None; W.K. Lee, Novartis, (C, R, S), Bayer (C, R, S), Allergan (C, R, S), Alcon (C, R, S), Santen (C, R, S)