Epiretinal Membrane Inflammatory Cell Density Might Reflect the Activity of Proliferative Diabetic Retinopathy

Abstract

Purpose.: Diabetic retinopathy (DR) has features of chronic low-grade inflammation. The purpose of our study was to investigate whether the presence of inflammatory cells in fibrovascular membranes (FVMs) from patients with proliferative diabetic retinopathy (PDR) is associated with the activity of PDR and visual acuity improvement after vitreoretinal surgery.

Methods.: Forty FVMs from 40 patients with PDR were obtained during vitrectomy, prepared by using the agar sandwich method, and examined using light microscope and immunohistochemistry methods to define the presence and density of inflammatory cells: CD45+ cells (leukocytes), CD4+ cells (T helper lymphocytes), CD8+ cells (T cytotoxic lymphocytes), CD19+ cells (B lymphocytes), and CD14+ cells (monocytes/macrophages). For each FVM, the inflammatory cell density defined as numerical areal density was calculated. The number of vessels was defined as the volume density of vessels.

Results.: Among 40 patients with PDR, 33 patients had active PDR and 7 quiescent PDR. Significant differences in cell densities for CD4+, CD8+, and CD19+ cells were observed between patients with active and quiescent PDR. B lymphocytes were present in membranes of active PDR only. No correlation was observed between numerical areal density of inflammatory cells and the volume density of vessels. No association was found between visual acuity improvement after surgery and cell densities.

Conclusions.: Lymphocyte infiltration of FVMs might be associated with the activity of retinopathy but not with visual acuity improvement after surgery.

Introduction

Diabetic retinopathy (DR) is a late microvascular complication of diabetes mellitus and the leading cause of blindness in the working-age population. The mechanisms by which diabetes causes microvascular complications and disease progression in the retina are not fully understood, but studies in animal models and patient samples have shown that DR has features of chronic low-grade inflammation. Diabetic retinopathy is associated with significant increases in proinflammatory cytokines, chemokines, and adhesion molecules and with the recruitment of leukocytes.^1–3 Progressive capillary occlusions and ischemia with angiogenic cytokine production lead to the formation of new blood vessels in an advanced stage of DR called proliferative diabetic retinopathy (PDR). New blood vessels form fibrovascular membranes (FVMs) that can lead to vitreous hemorrhage and tractional retinal detachment due to the contraction of fibrovascular tissue. Fibrovascular membranes in PDR are mainly composed of neovascular tissue,^4,5 but various inflammatory cells have also been found in these membranes.^6–11 T lymphocytes have been found in FVMs of patients with PDR.^6–8 B lymphocytes have also been found in some FVMs of patients with PDR⁶; however, Kase and coworkers⁷ did not find any in their study. Monocytes/macrophages have also been found in neovascular tissue.^9–11 Variable numbers of inflammatory cells in FVMs reported in these studies suggest that the density of inflammatory cells in FVMs might have some clinical significance. 

The role of T lymphocytes and macrophages in retinal angiogenesis and fibrogenesis has already been demonstrated.^12,13 Angiogenesis is the main process in active PDR, while fibrosis is a prominent feature of quiescent PDR; and it is possible that FVMs from patients with active PDR and FVMs from patients with quiescent PDR differ in the extent and cellular type of leukocyte infiltrates. To our knowledge, this has not been studied yet, but could give some new insight into the factors affecting FVM formation and PDR progression. 

Pars plana vitrectomy is the only choice of treatment in patients with tractional macular detachment due to PDR. Visual prognosis after vitrectomy is in correlation with preoperative macular detachment and ischemia.¹⁴ Kase and coworkers⁷ showed that high-level infiltration of T lymphocytes in FVMs of patients with PDR correlated with poor visual prognosis. 

The purpose of our study was to investigate whether the density of inflammatory cells in FVMs from patients with PDR is associated with the activity of PDR and visual acuity improvement after vitreoretinal surgery. 

Patients and Methods

Patients

From January 2010 to January 2012, forty consecutively operated patients with PDR and tractional macular detachment requiring vitrectomy (mean age 60.98 ± 11.10 years; 22 men and 18 women) were enrolled in the study. Exclusion criteria were recent retinal photocoagulation (less than 6 months), previous vitrectomy, glycated hemoglobin (HbA1c) higher than 10%, and any other ocular disease or known systemic inflammatory or hematologic disease. Informed consent was obtained from all patients. The study was approved by the National Ethical Committee (National Ethical Committee number 118/12/2011) and was performed in compliance with the Declaration of Helsinki. All patients had a complete ophthalmologic evaluation before the surgery with best-corrected visual acuity (BCVA) determination (Early Treatment Diabetic Retinopathy Study [ETDRS] visual acuity testing), slit-lamp examination, fundus examination, intraocular pressure measurement, gonioscopy, and optical coherence tomography. Based on ophthalmologic evaluation (preoperative and intraoperative), the activity of disease was noted. Clinical preoperative and intraoperative assessment of disease activity was done by two experienced vitreoretinal surgeons and medical retina specialists (MU, MGP). The activity of disease was categorized as active PDR when there were perfused capillaries in neovascular membranes, or inactive or quiescent PDR when there were only nonperfused capillaries in fibrotic membranes.¹⁵ Data regarding the patient's general condition and diabetes control were obtained from the patient and from the patient's general practitioner or diabetologist. The HbA1c was measured 1 day before the surgery. It was defined as a percentage value, which relates to the percentage of hemoglobin that is glycated and is an index of blood glucose control over preceding 6 to 8 weeks. Arterial hypertension was defined as a systolic blood pressure of 140 mm Hg or higher and/or diastolic blood pressure of 85 mm Hg or higher, or as condition treated with antihypertensive medications. Hyperlipidemia was defined as total cholesterol higher than 5 mM and/or triglycerides higher than 2 mM, or as condition treated with hypolipemic medications. Patients were followed after the surgery, and BCVA after 6 months was compared to preoperative BCVA. Visual acuity improvement was defined as gain of more than five letters on the ETDRS scale. For statistical analysis, ETDRS visual acuity was converted to logarithm of the minimum angle of resolution (logMAR). 

Agar Sandwich Method

Fibrovascular membranes were obtained during pars plana vitrectomy, gently removed from the eye, and immediately put into a vial with buffered 10% paraformaldehyde for prefixation. The agar sandwich method was modified based on previously described techniques.^16–18 The agar sandwich consisted of a solid agar disc and agar in liquid state poured over the disc. The solid and the liquid phases were prepared in advance under sterile circumstances in a laminated flow chamber. The 2.25% granulated agar was first melted by heating to almost boiling point. Each agar disc was 8 mm in diameter and approximately 1 mm thick. The agar discs were made by a trephine. The liquid agar was poured into a shallow dish to a depth of 1 mm, left to cool and harden, and cut into discs. Discs were stored in 70% ethanol at 0°C to 5°C. For the liquid phase of the agar sandwich, the melted agar was poured into a 10-mL syringe with a cap and stored cold at 0°C to 5°C. When a membrane was obtained, an agar disc was removed from the alcohol, rinsed with normal saline, and put onto a slide. A syringe of liquid phase with the cap was put into a water bath until agar melting point (90°C) and then left to cool but not harden (50°C). With the aid of a laboratory microscope, the prefixated membrane was gently straightened on the agar disc. Forceps were used only to grasp the edges of the sample. Eosin 0.1% was used to contrast the membrane as needed. The disc was dried with Surgicel (Eytec, Ripon, North Yorkshire, UK). The membrane was then attached to the disc by drops of liquid agar slowly poured from the syringe in a 0.5-mm layer to form the agar sandwich. The agar sandwich was allowed to cool and harden for 2 to 3 minutes and placed in fixative (buffered 10% paraformaldehyde). 

Immunohistochemistry

Paraffin-embedded tissue sections were produced for hematoxylin–eosin (HE) staining. The detection of different leukocytes was performed with the NovoLink Max Polymer Detection System (Leica, Wetzlar, Germany) following the manufacturer's instructions. The slides were incubated with anti-CD8 (diluted 1:100; Dako, Glostrup, Denmark), -CD4 (1:80; Dako), -CD14 (1:80; Dako), -CD45 (1:100; Dako) and -CD19 (1:200; Dako) antibodies overnight at 4°C (Fig. 1). Each cell stained immunohistochemically with different antibodies (CD8, CD4, CD14, CD45, and CD19) was defined as labeled or not labeled. 

Figure 1

 

With hematoxylin–eosin and immunohistochemical staining, different inflammatory cells (CD4 positive, CD8 positive, CD14 positive, CD19 positive, CD45 positive, and TNF positive) can be demonstrated in fibrovascular membranes of patients with proliferative diabetic retinopathy.

Evaluation of the Numerical Areal Density (N_A) of the CD8-, CD4-, CD14-, CD45-, and CD19-Positive Cells

The number of positive (labeled) cells per area—numerical areal densities of the CD8-, CD4-, CD14-, CD45-, and CD19-positive cells (number of positive cells/mm²)—were calculated.¹⁹ 

Volume Density of Blood Vessels (VD)

The volume density of blood vessels was morphometrically evaluated on HE-stained histologic sections. A morphometric analysis was performed on a Wild sampling microscope (Wild, Heerbrugg, Switzerland) using Weibel's B 100 double grid test system.^19,20 The volume density of blood vessels was estimated by counting the points of the grid system that hit the blood vessels (BV) (blood vessel walls or their lumen) and reference space (RS) at ×400 magnification (reference space—hits on tissue of epiretinal membrane). The volume density of blood vessels (VD) is the quotient between hits falling on vessels and hits falling on the reference space (VD = BV/RS) and is expressed in mm³/mm³. 

Two histopathologists (AMZ, DP), who were not aware of the clinical status and group of the case, evaluated the slides (histology, immunostaining, and volume density of blood vessels). The two histopathologists produced very similar results, and the differences in the number of positive cells were less than 10%. When the investigators disagreed on a section, the results of a third pathologist were taken into consideration. 

Statistical Analysis

Measurement data in our study did not meet the normality assumption, so median and range between minimum and maximum variable were used for description. Spearman's correlation was used to assess the association between variables. The Mann-Whitney U test was used for assessing differences between independent groups. A P value less than 0.05 was considered statistically significant. 

Results

Forty FVMs were removed from the eyes of 40 patients with PDR. Clinical data of patients are presented in Table 1. 

Table 1

 

Clinical Data of Patients With PDR, n = 40

	Patients With Active PDR,	Patients With Quiescent PDR,	P Value
	n = 33	n = 7
Age, y ± SD	59.27 ± 10.75	69.00 ± 9.66	0.04
Sex	20 men, 60.61%	2 men, 28.57%
	13 women, 39.39%	5 women, 71.43%
Duration of diabetes, y ± SD	13.88 ± 8.97	21.00 ± 10.03	0.06
Incidence of insulin therapy, n	26, 78.79%	6, 85.71%	0.68
Hba1c, % ± SD	7.70 ± 1.10	8.06 ± 1.03	0.40
Incidence of arterial hypertension, n	29, 87.88%	7, 100%	0.63
Incidence of hyperlipidemia, n	17, 51.52%	3, 42.86%	0.54
BMI, kg/m2 ± SD	30.43 ± 4.08	29.7 ± 4.45	0.34
Vitreous hemorrhage, n	28, 84.85%	0, 0%	<0.0001
VA improvement, n	24, 72.73%	4, 57.14%	0.42
Preoperative BCVA, logMAR ± SD	1.37 ± 0.51	0.95 ± 0.74	0.12
Postoperative BCVA, logMAR ± SD	0.86 ± 0.51	0.80 ± 0.67	0.64

 SD, standard deviation; BMI, body mass index; VA, visual acuity.

Capillary-like structures in a fibrous matrix containing several mononuclear cells were observed in all HE-stained sections of FVMs. Cell densities for each tested marker and vessel densities calculated from immunohistochemistry are presented in Table 2 as median with minimum and maximum value. 

Table 2

 

Numerical Areal Density (N_A) of Different Inflammatory Cells and Volume Density of Vessels (VD) From All FVMs; Median (Minimum to Maximum)

Numerical Areal Density (NA) of Different Inflammatory Cells	Median (Minimum to Maximum)
NA of CD45+ cells per mm2	256.43 (0.00–1253.99)
NA of CD4+ cells per mm2	39.92 (0.00–535.71)
NA of CD8+ cells per mm2	60.59 (0.00–408.65)
NA of CD19+ cells per mm2	8.33 (0.00–151.37)
NA of CD14+ cells per mm2	257.62 (21.48–1529.02)
VD, mm2/mm2	0.06 (0.00–0.57)

Leukocyte density (N_A of CD45+ cells) correlated well with N_A of CD4+ cells (Spearman's ρ coefficient 0.53; P < 0.0001), N_A of CD8+ cells (Spearman's ρ coefficient 0.74; P < 0.0001), N_A of CD14+ cells (Spearman's ρ coefficient 0.72; P < 0.0001), and N_A of CD19+ cells (Spearman's ρ coefficient 0.38; P = 0.016). Strong correlations were observed between N_A of CD4+ cells, CD8+ cells, and CD19+ cells (Spearman's ρ coefficient from 0.53–0.68; P < 0.0001) and between CD8+ cells and CD14+ cells (Spearman's ρ coefficient 0.55; P < 0.0001), and moderate correlations were obtained between N_A of CD14+ cells and N_A of CD4+ cells (Spearman's ρ coefficient 0.36; P = 0.023). 

No correlation was observed between cell densities and the volume density of vessels (Spearman's ρ coefficients less than 0.19; P > 0.05; Fig. 2). 

Figure 2

 

Correlation between leukocyte density (N_A of CD45+ cells) and volume density of vessels (VD) (Spearman's ρ coefficient 0.17; P = 0.28).

Clinically, 33 patients had active PDR and 7 had quiescent PDR (Table 3). In all FVMs from patients with active PDR (n = 33), CD45+ cells (leukocytes), CD14+ cells (monocytes/macrophages), and CD4+ cells (T helper lymphocytes) were identified. CD8+ (T cytotoxic lymphocytes) were detected in 32 out of 33 FVM, and CD19+ cells (B lymphocytes) were identified in 27 out of 33 FVM. 

Table 3

 

Numerical Areal Density (N_A) of Different Inflammatory Cells and Volume Density of Vessels (VD) in FVMs From Patients With Active and Quiescent PDR; Median (Minimum to Maximum); Mann-Whitney U Test

	Active PDR,n = 33	Quiescent PDR,n = 7	P Value
NA of CD45+ cells per mm2	273.44 (38.80–253.99)	158.33 (0.00–442.07)	0.098
NA of CD4+ cells per mm2	90.85 (6.51–535.71)	20.16 (0.00–31.25)	0.004
NA of CD8+ cells per mm2	65.52 (0.00–408.65)	7.13 (0.00–64.58)	0.025
NA of CD19+ cells per mm2	13.58 (0.00–151.37)	0.00 (0.00–0.00)	0.001
NA of CD14+ cells per mm2	253.90 (21.48–1529.02)	261.36 (99.81–525.00)	0.533
VD, mm2/mm2	0.06 (0.01–0.57)	0.05 (0.00–0.16)	0.496

In all FVMs from patients with quiescent PDR (n = 7), CD14+ cells (monocytes/macrophages) were identified. CD45+ cells (leukocytes) were detected in five out of seven FVMs; CD4+ cells (T helper lymphocytes) and CD8+ (T cytotoxic lymphocytes) were identified in four out of seven FVMs. No CD19+ cells (B lymphocytes) were identified in any FVMs from patients with quiescent PDR. Two FVMs from patients with quiescent PDR were negative for all lymphocyte markers. 

In FVMs from patients with active PDR, statistically significantly higher numerical areal densities of CD4+ cells (T helper lymphocytes), CD8+ cells (T cytotoxic lymphocytes), and CD19+ cells (B lymphocytes) were found in comparison with the FVMs from patients with quiescent PDR (Table 3), whereas there was no statistically significant difference in numerical areal densities of either CD45+ cells (leukocytes) or CD14+ cells (monocytes/macrophages) (Table 3). 

Additionally, we wanted to find out if there is any association between visual acuity improvement and the numerical areal density of inflammatory cells. No association, however, was found between visual acuity improvement and the numerical areal density of inflammatory cells (P = 0.294). Visual acuity was better 6 months after surgery in 28 patients (70%), 24 of them having active PDR and 4 quiescent PDR. 

Discussion

In our study, T lymphocytes, B lymphocytes, and macrophages were found in FVMs from patients with PDR. B lymphocytes were present in membranes of active PDR only. Additionally, there was an association between the density of inflammatory cells in FVMs and the activity of retinopathy. To the best of our knowledge, the last two findings have not been shown before. Finally, we did not find any association between visual prognosis and the density of inflammatory cells in FVMs. 

The presence of T lymphocytes and macrophages is consistent with chronic inflammation in eyes with PDR and is in concordance with previous studies.^6,7,11 Inflammation is associated with tissue repair and scar formation. Proliferative DR is the result of a misguided attempt by the retina to repair the damage caused by low-grade chronic inflammation. The ocular pathology is not specific. It reflects a general process in diabetic disease.^12,21 The same process also affects kidneys, peripheral nerves, and the parenchyma of several other tissues. It has been shown that the infiltration of inflammatory cells precedes the process of fibrosis in diabetic nephropathy.²² Recruitment and activation of T lymphocytes is an early event in the initiation of renal fibrosis and precedes the influx of macrophages, which play a crucial role in renal fibrogenesis.²³ B lymphocytes were found to be increased in kidney glomeruli of diabetic NOD mice, and immune deposits have been described in histologic studies of diabetic nephropathy.²⁴ 

Our findings are in accordance with studies that demonstrated the role of T lymphocytes (especially CD4+ T lymphocytes) and macrophages in retinal angiogenesis and fibrogenesis.^12,13 We have found significantly higher densities of CD4+ T lymphocytes, CD8+ T lymphocytes, and B lymphocytes in FVMs with active PDR. 

Kase and coworkers⁷ found T lymphocytes in FVMs from patients with PDR, suggesting that T lymphocytes have an important role in the pathogenesis of PDR and in visual prognosis. In our study, CD4+ T lymphocytes were found in all FVMs from patients with active PDR, while CD8+ T lymphocytes were found in 32 out of 33 FVMs from patients with active PDR. Both types of T lymphocytes were also identified in 4 out of 7 (57.1%) FVMs from patients with quiescent PDR. Tang and coworkers⁶ reported CD8+ T lymphocytes in 66% and CD4+ T lymphocytes in 56% of FVMs from patients with PDR. 

In our study, B lymphocytes were present only in FVMs from patients with active PDR, and the density of B lymphocytes was relatively low in comparison to densities of other inflammatory cells. Kase and coworkers⁷ did not find any B lymphocytes in the FVMs using the marker CD20. On the other hand, Tang and coworkers⁶ found B lymphocytes in 47% of FVMs using the same marker. The marker CD19 was used in our study to detect B lymphocytes. B lymphocytes are mainly seen as sources of proinflammatory autoantibodies, but they also produce many cytokines and activate T lymphocytes in chronic inflammatory disease.²⁵ Autoantibodies are thought to play major roles in the pathogenesis of type 1 diabetes, but could have some role in the pathogenesis of type 2 diabetes as well. Cytokine-producing B lymphocytes may have an important role in either establishing or maintaining chronic inflammation in diabetes.²⁵ The role of B lymphocytes in PDR remains to be elucidated. Our findings suggest that B lymphocytes might be implicated in the pathogenesis of angiogenesis in active PDR, whereas there were no B lymphocytes in quiescent PDR. 

The CD14 marker was used in our study as a monocyte/macrophage marker to differentiate macrophages from other CD45+ cells. The density of CD14+ cells in FVMs was not significantly different between patients with active and quiescent PDR (median 253.90 and 261.36 cells/mm²; P = 0.53). It is known that CD14 is also expressed by fibroblasts.²⁶ We speculate that the abundance of CD14+ fibroblasts in FVMs from patients with quiescent PDR affected our findings, since we could not discriminate between macrophages and CD14+ fibroblasts. Additionally, it is known that microglial cells are resident macrophages in the central nervous system and retina,²⁷ and it has been shown that microglia are the major cells in the central nervous system that express CD14.²⁸ Activated microglial cells were reported in FVMs from patients with PDR by using microglial markers HLA-DR, CD68, and CD45,¹⁰ and they were implicated in the pathogenesis of new vessel formation in PDR by producing proinflammatory molecules.²⁹ When activated, however, microglia resemble hematogenous macrophages invading the retina, and there are no specific markers to distinguish activated microglia from hematogenous macrophages. In addition, it has been shown that activated microglia surround retinal vasculature, which probably makes this distinguishing even more difficult.¹⁰ 

In our study, the volume density of vessels was similar in FVMs from patients with active PDR and quiescent PDR. Moreover, no association between densities of inflammatory cells and the density of vessels was demonstrated. Blood vessels in FVMs were seen as channels of varying diameters on HE-stained histologic sections. Some channels were formed by viable endothelial cells, while others were acellular. Similarly, Wallow et al.⁴ described “ghost vessels,” which are a prominent feature of quiescent PDR, as acellular channels with fibrous material filling part or all of their lumen. By evaluating the volume density of vessels, we did not make a distinction between channels formed by viable endothelial cells and ghost vessels, which explains our results. 

Finally, we were interested in whether the density of inflammatory cells in FVMs affects the visual prognosis after operation. We did not find any association between the density of inflammatory cells in FVMs and visual acuity improvement 6 months after surgery as a surrogate of visual prognosis. With respect to our finding, Kase and coworkers⁷ demonstrated a high-level infiltration of T lymphocytes in FVMs in correlation with a poor visual prognosis. They concluded that visual deterioration after vitrectomy was due to marked reproliferation of the epiretinal membranes. We speculate that in those patients, the active form of the disease (active PDR) was present. Accepting this speculation, the findings of Kase and coworkers⁷ and our findings are in accordance, since in both studies, a higher density of T lymphocytes was demonstrated in active PDR. In our study, however, there were no patients with postoperative reproliferation of FVMs. In patients with PDR, visual acuity after vitrectomy depends on many factors, but probably the most important is the preservation of integrity and function of retinal neurons and photoreceptors in the macula. The macula integrity and function are altered in patients with PDR by preoperative macular detachment, macular edema, ischemia, and neuronal and photoreceptor damage.²⁹ Because all these variables could additionally contribute to intraocular inflammation present in patients with DR,^14,29–33 it could be presumed that the presence and density of inflammatory cells might be a marker for visual acuity prediction after vitrectomy. According to our findings, the density of inflammatory cells infiltrating FVMs does not correlate with the function of the macular retina, and consequently cannot be used as a predictor of visual prognosis after vitrectomy. 

Fibrovascular membrane analysis gives us more insight into the pathogenesis of PDR, and can be useful in developing potential new therapeutic strategies and to predict optimal therapy for individual patients. The suppression of the inflammation/immune response could be important in preventing the development or progression of FVMs. There is a paucity of clinical data in this field of research, and studies analyzing human FVMs are few and have been done on relatively small samples. Our study was conducted on 40 FVMs, the largest sample so far; inflammatory cells (T lymphocytes, B lymphocytes, and macrophages) were shown to be present in FVMs from patients with PDR. Moreover, B lymphocytes were present in membranes of active PDR only. We analyzed 33 FVMs from patients with active PDR and only 7 FVMs from patients with quiescent PDR. 

Finally, these data suggest that the density of lymphocyte infiltration might be associated with the activity of PDR, but not with visual acuity improvement after vitrectomy; however, further studies are needed to conclusively demonstrate this relationship. 

Acknowledgments

Supported by a grant from the Slovenian Research agency ARRS (P3-0333). 

The authors thank Polona Sajovic and Petra Nussdorfer for the preparation of histologic slides, and Brina Beškovnik, BA, for revising the English. 

Disclosure: M. Urbančič, None; Š. Štunf, None; A. Milutinović Živin, None; D. Petrovič, None; M. Globočnik Petrovič, None