Epiretinal membranes (ERMs) involving the macular or perimacular regions can cause a reduction in vision, metamorphopsia, micropsia, or occasionally monocular diplopia. The presence of ERMs has been associated with various clinical conditions including proliferative diabetic retinopathy (PDR), proliferative vitreoretinopathy (PVR), vitreous haemorrhage, congenital disorders, and idiopathic or spontaneous developments [1]. The prevalence of ERMs in PDR is reported to be about 20% in Type 1 diabetes and about 5% in Type 2 diabetes [2, 3]. Histopathological studies have shown that ERMs are composed of various cell types such as glial cells, fibroblasts and endothelial cells [4, 5, 6, 7], but their pathogenic mechanisms are still unknown. In contrast, trophic factors and cytokines, such as basic fibroblast growth factor (bFGF), vascular endothelial cell growth factor (VEGF) and interleukin-8 (IL-8) were detected in vitreous fluid and ERM samples derived from PDR patients [8, 9, 10, 11, 12]. In a previous study, we examined the expression of various trophic factor receptor genes in ERMs, and showed the possibility that glial cell line-derived neurotrophic factor (GDNF) is involved in the formation of glial cell component of ERMs after PDR [13]. However, we have not determined the possible molecular mechanisms in the formation of other cell types in PDR membranes.

Nuclear factor kappa B (NF-κB) is a transcription factor that can be activated by hypoxia, bacteria, viral proteins and proinflammatory cytokines such as tumor necrosis factor-α (TNF-α) and interleukin-1β [14, 15]. NF-κB represents a group of structually related and evolutionarily conserved proteins, with five members in mammals: Rel (c-Rel), RelA (p65), RelB, NF-κB1 (p50 and its precursor p105), and NF-κB2 (p52 and its precursor p100) [16]. NF-κB/Rel proteins can exist as homo- or heterodimers. Among them, a heterodimeric complex of p65 and p50 subunits is thought to be central to the regulation of numerous inflammatory and proliferative response genes [17]. Therefore, we decided to investigate NF-κB p50 subunit using reverse transcription-polymerase chain reaction (RT-PCR) and immunohistochemical analysis. In resting cells, most NF-κB/Rel dimers are bound to IκB inhibitory proteins, but exposure to the above-mentioned stimuli generates a signal transduction cascade that leads to degradation of IκB. NF-κB released from IκB can then translocate to the nucleus and induce the transcription of genes that bear NF-κB-binding sites. The basis for the cytoplasmic localization of the inactive NF-κB/IκB complex is thought to be due to masking of the nuclear localizing signals (NLS) on the NF-κB subunits by the IκB proteins. Thus, IκB degradation leads to unmasking of the NLS, allowing free NF-κB dimers to enter the nuclear import pathway. The precise role of NF-κB in ocular diseases is not well understood. However, recent studies have shown that NF-κB regulates the expression of angiogenic factors such as IL-8 in the endothelial cells and induces neovascularization that could lead to proliferative vitreoretinal disorders such as PDR [18, 19, 20]. In addition, since endothelial cell is one of the main components of ERMs [4, 5, 6, 7], these findings suggest the possibility that NF-κB signalling is involved in the formation of PDR membranes.

In this study, we examined the expression of NF-κB p50 mRNAs in ERMs obtained from PDR and control patients, and found their high expression levels in PDR subjects. Interestingly, vascular endothelial cell component in PDR membranes were stained with NF-κB antibody raised against human p50 NLS region. Furthermore, NF-κB and IL-8 proteins were partially overlapped. We also proposed a possible relationship between NF-κB expression and ERM formation.

Subjects and methods

Subjects

This study was carried out in accordance with the tenets of the Helsinki Declaration. Informed consent was obtained from each patient for the collection of samples. Criteria for inclusion in the study were age less than 80 years old, absence of renal or haematological diseases or uremia, absent administration of chemotherapy or life-support measures, and the fewest possible chronic pathologies other than diabetes. The epiretinal membranes were surgically removed from consecutive eyes with secondary ERM following PDR (27 eyes) or idiopathic ERM (controls, 22 eyes) undergoing pars plana vitrectomy and membrane peeling in Toho University Sakura Hospital. Membranes were dissected from the retinal surface with horizontal scissors or membrane pick. Samples derived from 22 of the PDR patients (age 56±11 years, duration of diabetes 18±10 years) and 15 control subjects (age 68±7 years) were processed for reverse transcription-polymerase chain reaction (RT-PCR) analysis. The remainder of the PDR samples (age 51±16 years, duration of diabetes 15±6 years) and control samples (age 62±8 years) were processed for immunohistochemistry. These samples were embedded in optimum cutting temperature (OCT) compound (Miles Laboratories, Naperile, Ill., USA), flash-frozen in liquid nitrogen, and then stored at −80°C.

RNA extraction and amplification by RT-PCR

Total cellular RNA was prepared as previously reported [13]. 0.1 µg of RNA extracted from each sample was reverse transcribed into first-strand cDNA using the Superscript Preamplification System (Gibco, Paisley, Scotland) and oligo-dT primers.

RT-PCR analysis was carried out as previously described [13, 21]. Complementary DNA reverse transcribed from total RNA was amplified by using primers specific for human NF-κB p50 (sense: 5′-CAC TTA TGG ACA ACT ATG AGG TCT CTG G-3′; antisense: 5′-CTG TCT TGT GGA CAA CGC AGT GGA ATT TTA GG-3′) and glyceraldehyde-3-phosphate dehydrogenase (G3PDH) (sense: 5′-ACC ACA GTC CAT GCC ATC AC-3′; antisense: 5′-TCC ACC ACC CTG TTG CTG TA-3′). PCR was done after initial denaturation at 94°C for 3 min. Each cycle consisted of a heat-denaturation step at 94°C for 15 seconds, annealing of primers at 61°C for 2 min, followed by polymerization at 72°C for 2 min. The expected sizes of the amplified cDNA fragments of NF-κB and G3PDH were 406 and 452 bp, respectively. Human cDNA acquired from testis (Takara, Kyoto, Japan) was used as a positive control. Negative controls for PCR were done using "templates" derived from reverse transcription reactions lacking either reverse transcriptase or total RNA. After 35 cycles, 15 µl of each reaction mixture were electrophoresed on a 2% Tris borate-EDTA agarose gel and stained with ethidium bromide.

Immunohistochemistry

Frozen sections (7-µm thick) were cut by a cryostat, mounted on 3-aminopropyltriethoxysilane-coated glass slides, and air-dried at room temperature. For immunohistochemical analysis, the sections were fixed in ice cold acetone and then washed with phosphate buffered saline (PBS). The sections were incubated with normal donkey serum for 30 min to block non-specific staining. They were stained overnight at 4°C with an affinity-purified rabbit polyclonal antibody raised against amino acids 350 to 363 mapping within the nuclear location signal (NLS) region of human NF-κB p50 [22] (Santa Cruz Biotechnology, Santa Cruz, Calif., USA; 2.0 µg/ml). For double-labelling immunofluorescence studies, they were then incubated with a goat polyclonal antibody against IL-8 (Santa Cruz Biotechnology; 1.0 µg/ml) or von Willebrand factor (vWF) (Santa Cruz Biotechnology; 1.0 µg/ml) [23, 24, 25]. NF-κB was visualized with Cy3-conjugated donkey anti-rabbit IgG (Amersham Pharmacia Biotech, Piscataway, N.J., USA; 1000×), and IL-8 or vWF was visualized with FITC-conjugated donkey anti-goat IgG (Jackson Immunoresearch, West Groove, Pa, USA; 200×). The sections were examined by a confocal laser scanning microscope (Olympus, Tokyo, Japan). Primary antibody preabsorbed by blocking peptide (Santa Cruz Biotechnology) was used for negative controls. The adjacent section of each specimen was stained with haematoxylin and eosin.

Assessment of NF-κB-positive cell rate

NF-κB-positive cells were counted by a single observer who had no prior knowledge of the specimen information. Using a comparison with the adjacent section stained with haematoxylin and eosin, activated NF-κB-positive cells were counted in an immunohistochemistry profile. Measurements were performed with a calibrated reticle at 20X magnification as previously reported [26]. Since virtually all activated NF-κB is localized in the nucleus, cells were considered positive if any nuclear staining was present. The NF-κB-positive cell rate was defined as the percentage of activated NF-κB-positive cells relative to the total number of counted cells. In order to reduce sampling error, 60 to 100 cells were analyzed in each section.

Statistics

Data are shown as means ± SEM except as noted. The χ2 test was used to test for significance of the difference between presence of NF-κB mRNAs in PDR versus controls. For comparison of NF-κB-positive cell rate between PDR and controls, the Mann-Whitney's U test was used. A p of less than 0.05 was considered statistically significant.

Results

NF-κB p50 mRNA expression in ERMs was examined by RT-PCR analysis. NF-κB mRNA was detected in 20 of 22 (91%) PDR patients, but in only 9 of 15 (60%) idiopathic ERM (control) patients (p<0.05) (Table 1). Immunohistochemical analysis was carried out to identify the activated form of NF-κB p50 protein in ERMs after PDR. NF-κB protein was detected in all the five PDR subjects examined (Figs. 1A, 2A) and the immunoreactivity was completely abolished when primary antibody was preabsorbed with blocking peptide (data not shown). NF-κB protein was detected in five of the seven control subjects (data not shown), but its expression level was low. Quantitative analysis (Table 2) showed that the percentage of NF-κB-positive cells was 42±9% in PDR subjects, but only 10±3% in control subjects (p<0.01).

Table 1. Number (%) of NF-κB mRNA-positive patients
Full size table
Fig. 1A–D.
figure 1

Expression of NF-κB (A) and von Willebrand factor (vWF) (B) in the ERM derived from a PDR patient age 49 years with an 8-year history of diabetes. (C) Double-labelling method shows the co-expression of NF-κB and vWF (arrow heads). (D) Haematoxylin and eosin staining of the adjacent section. Bar represents 100 µm

Full size image
Fig. 2A–D.
figure 2

Expression of NF-κB (A) and IL-8 (B) in the ERM derived from a PDR patient age 49 years with an 8-year history of diabetes. (C) Double-labelling method demonstrates the co-expression of NF-κB and IL-8 protein. (D) Haematoxylin and eosin staining of the adjacent section. Bar represents 100 µm

Full size image
Table 2. Percentage of NF-κB-positive cells
Full size table

We next examined whether NF-κB protein is expressed in the neovascular region in PDR subjects. To assess this possibility, we used a vascular endothelial cell-specific marker, von Willebrand factor (vWF). ERMs contained many vWF-positive cells (Fig. 1B) and some of them were double-labelled with NF-κB (Fig. 1C). Since NF-κB could induce angiogenesis by regulating angiogenic factors such as IL-8, we also examined IL-8 protein expression. ERMs contained IL-8-positive cells (Fig. 2B) and some cells were double-labelled with NF-κB (Fig. 2C). Such IL-8 immunoreactivity was completely abolished when primary antibody was pre-absorbed with blocking peptide for IL-8 (data not shown).

Discussion

This study shows high expression levels of NF-κB mRNA in ERMs after PDR. In addition, we showed that NF-κB protein is partially overlapped with vascular endothelial cell-specific marker, vWF. Since NF-κB could regulate the expression of angiogenic factors [18, 19, 20], we also examined IL-8 protein expression in PDR subjects. Interestingly, in some cells, IL-8 was co-expressed with NF-κB. Together with previous findings that endothelial cell is one of the main components of ERMs [4, 5, 6, 7], NF-κB might be involved in the formation and development of neovascular component in PDR membranes.

One important issue is that NF-κB mRNA expression was detected in both primary and secondary ERMs (Table 1) although the number of NF-κB-positive cells was smaller in idiopathic (nonvascular) ERMs (Table 2). Previous studies have shown that IL-8 induced by NF-κB plays a key role in retinal angiogenesis after PDR [18, 19, 20]. In this context, it seems to be reasonable that NF-κB-positive cell number was larger in PDR (vascular) membranes. However, some NF-κB-positive cells were IL-8- or vWF-negative (Fig. 1 and Fig. 2). NF-κB is activated in glial cells, vascular endothelial cells, and macrophage/microglia, all of which are involved in neovascularization in diabetic retinopathy [19, 20]. Thus, NF-κB could be involved in the proliferation of various cell types other than vascular endothelial cells. In fact, we have determined that PDR membranes contain a large area composed of glial cells [13]. In any case, a more detailed role of NF-κB in the process of ERM formation should be determined. Another issue is that we collected ERM samples from patients whose disease had already manifested for several years. So, we cannot exclude the possibility that NF-κB activation in ERMs might be a secondary event. We will compare NF-κB expressions between newly-formed and relatively old ERM samples in future experiments.

In a previous study, we found high expression levels of GDNF receptor (GFRα2) in glial cell component and suggested the possibility that inhibiting the activity of this receptor could prevent the development of PDR membranes [13]. Since glial cells produce and store growth factors which activate endothelial cells [29, 30, 31, 32, 33, 34], this method might also inhibit the formation of neovascular component in ERMs. In contrast, the first step of the ERM formation is thought to be the neovascularization following the interaction of retinal vessels with the vitreous. In this process many angiogenic factors play an important role [9, 35, 36, 37]. Thus, blocking these angiogenic factors could inhibit ERM formation. In this context, IL-8 is one of the potential therapeutic targets. Recent studies have shown that transfection of IκB mutant, which blocks NF-κB activation [38], can inhibit angiogenesis by suppressing the production of IL-8 [39, 40]. These results suggest the possibility that inhibiting biological activity of NF-κB by specific blocker or antisence oligonucleotide could be useful in preventing the formation of PDR membranes, especially in avoiding membrane recurrence. However, before determining this possibility, further investigations about the precise role of NF-κB on retinal cells and all cell types in PDR membranes will be needed.