Erythropoietin Is Expressed in the Human Retina and It Is Highly Elevated in the Vitreous Fluid of Patients With Diabetic Macular Edema Free

The present study included 24 type 2 diabetic patients (12 consecutive patients with DME and 12 consecutive patients with PDR, age 61.6 ± 15.3 years) in whom a classic three-port pars plana vitrectomy was performed. Vitreous from 10 age-matched nondiabetic patients (age 59.6 ± 10.7 years) with macular hole requiring vitrectomy served as the control group (n = 10). The exclusion criteria were 1) photocoagulation in the preceding 3 months or previous vitreoretinal surgery, 2) recent vitreous hemorrhage (<3 months before vitrectomy) or intravitreous hemoglobin >5 mg/ml, 3) renal failure (creatinine ≥120 μmol/l), and 4) presence of anemia (hemoglobin <12 g/dl).

Fluorescein angiography was performed in all patients with DME in order to assess the degree of retinal ischemia denoted by nonperfusion of the retinal capillaries. For this purpose, we used the method developed by the Central Retinal Vein Occlusion Study Group (20). The photographic protocol included fluorescein angiographic views of posterior pole and midperiphery. Capillary nonperfusion was measured using a transparent disc area template placed over the angiogram. The number of disc areas was counted by the same ophthalmologist in every case, who was unaware of the clinical status of the patients. According to this protocol, eyes with <10 disc areas of nonperfusion were classified as having little or no evidence of ischemia.

The protocol for sample collection was approved by the hospital’s ethics committee, and informed consent was obtained from the patients.

In all cases, a classic three-port pars plana vitrectomy was performed. For visualization of the vitreous cavity, we used a wide-field system with a precorneal Volk lens of 130° and inversion image system by Moeller-Wedel (Hamburg, Germany).

Undiluted vitreous samples (∼1 ml) were obtained at the onset of vitrectomy by aspiration into a 1-ml syringe attached to the vitreous cutter (Alcoon Model; Ten-Thousand Ocutome, Irvine, CA) before starting intravitreal infusion of balanced salt solution. The vitreous samples were transferred to a tube, placed immediately on ice, and centrifuged at 16,000g for 5 min at 4°C. Supernatants were frozen at −80°C until assayed. For serum determinations, blood samples were collected simultaneously with the vitrectomy, then centrifuged at 3,000g for 10 min at 4°C to obtain serum, aliquoted, and stored at −80°C until assayed.

Human postmortem eyes were obtained from 16 donors: 8 nondiabetic and 8 age-matched diabetic donors without diabetic retinopathy (age 73 ± 2.3 vs. 71 ± 8 years, respectively; P = 0.23). Although erythropoietin mRNA has extraordinary stability for postmortem degradation (21), the time elapsed from death to enucleation was <4 h in all cases. After enucleation, the eyes were snap frozen in liquid nitrogen and stored at −80°C.

Erythropoietin was assessed by radioimmunoassay (Incstar-Diasorin). The intra- and interassay coefficients of variation were 4.8 and 10.6%, respectively. The lowest limit of detection was 4.4 mU/ml.

Vitreous hemoglobin levels were measured by spectophotometry (Uvikon 860; Kontron Instruments, Zurich, Switzerland), which permits to measure hemoglobin in micromolar concentration. The lowest limit of detection was 0.03 mg/ml.

Vitreal proteins were measured by a previously validated microturbidimetric method with an autoanalyzer (Hitachi 917; Boehringer, Mannheim, Germany). This method, based on the benzetonium chloride reaction, is a highly specific method for the detection of proteins and has a higher sensibility and reproducibility than the classic method of Lowry. The lowest level of proteins detected was 0.02 mg/ml. Intra- and interassay coefficients of variation were 2.9 and 3.7%, respectively.

Human neuretina and retinal pigment epithelium (RPE) were harvested under the microscopic dissection of isolated eye cups from donors. Total RNA was extracted from isolated retinal tissues using the RNeasy Mini Kit with DNAse digestion (Quiagen Distributors, IZASA, Barcelona, Spain) according to the manufacturer’s instructions. A 1-μg aliquot of total RNA was used directly for reverse transcription, which was carried out using random hexanucleotide priming and TaqMan Reverse Transcription Reagents (Applied Biosystems, Madrid, Spain) in a 50-μl reaction according to the manufacturer’s protocol.

Quantitative real time using erythropoietin-specific primers (TaqMan premade gene expression assay Hs0017267; Applied Biosystems; Gene Bank RfSeq NM 000799.2) was performed taking 2 μl of the reverse-transcription reaction as template in a PCR setup with the TaqMan Universal Mastermix. The reaction was conducted as follows: 95°C for 10 min and 50 cycles of 15 s at 95°C and 1 min at 60°C in Applied Biosystems 7000 equipment. Each sample was assayed in duplicate, and control negative samples were included in each experiment. Automatic Relative Quantification data were obtained with ABI Prism 7000 SDS software (Applied Biosystems) using β-actin as endogenous gene expression control (Hs9999903_m1; Applied Biosystems).

In addition, standard PCR was performed with specific primers designed using Primer3 software in an Applied Biosystems 2720 Thermal Cycler. A β-actin 350-bp product was obtained with forward 5′-AAACTGGAACGGTGAAGGT-3′ and reverse 5′-TCAAGTTGGGGGACAAAAA-3′ primers, and a 210-bp erythropoietin product was obtained with forward 5′-CTTCTCCTGTCCCTGCTGTC-3′ and reverse 5′-TCC ATC CTC TTC CAG GCA TA-3′ primers. PCR products were electrophorezed on 2% agarose gel and stained with ethidium bromure.

We conducted this study in accordance with the Declaration of Helsinki and received approval from the ethics committee of Vall d’Hebron University Hospital.

The Kolmogorov-Smirnov test was utilized to confirm the assumption of the normality of the variables. Student’s t test and ANOVA were used to compare continous variables and the χ² test for categorical variables. Because of their skewed distribution, erythropoietin concentrations were displayed as median and range, and the statistical comparisons were performed using a nonparametric test (Mann-Whitney U test). Correlations were examined by Spearman’s rank correlation. Levels of statistical significance were set at P < 0.05.

The main clinical characteristics of subjects included in the study are summarized in Table 1. We did not observe significant differences in serum erythropoietin among groups (DME: median 12.4 mU/ml [range 7.8–46.7]; PDR: 10.6 mU/ml [2.5–19.2]; and control subjects: 8.7 mU/ml [4.1–20.8]; P = NS). Fluorescein angiography demonstrated a lack of significant retinal ischemia in all patients with DME (mean of disc areas of retinal nonperfusion 0.8 [range 0–4]).

Intravitreous protein concentration was significantly higher in diabetic patients than in control subjects (3.1 [range 1.1–7.2] vs. 0.75 mg/ml [0.4–1.7]; P < 0.001). However, no significant differences were detected between DME and PDR patients (3.9 [1.9–7.2] vs. 3.1 mg/ml [1.1–5.1]; P = 0.22).

The intravitreal erythropoietin concentration was significantly higher in the whole group of diabetic patients than in control subjects in absolute terms (326 [range 41–3,000] vs. 30 mU/ml [10–75]; P < 0.0001) and also after correcting for intravitreal proteins (103 [13–1,027] vs. 50 mU/mg [13–107]; P = 0.03). The differences remained significant when DME and PDR patients were compared separately with the control group in absolute terms (430 [41–3,000] vs. 30 mU/ml [10–75], P < 0.0001 and 302 [117–850] vs. 30 mU/ml [10–75]; P < 0.0001) and after correcting for intravitreal proteins (105 [13–967] vs. 50 mU/mg [13–107], P = 0.02 and 65 [32–1,027] vs. 50 mU/mg [13–107]; P = 0.03).

Higher intravitreous erythropoietin concentration was detected in subjects with DME in comparison with PDR patients, but the difference was not statistically significant either in absolute terms or after adjusting for intravitreal proteins (430 [range 41–3,000] vs. 302 mU/ml [117–850]; P = 0.21 and 105 [13–967] vs. 65 mU/mg [32–1,027]; P = 0.53, respectively; Fig. 1).

Erythropoietin was 30-fold higher in vitreous fluid than in serum from diabetic patients (326 [range 41–3,000] vs. 11.2 mU/ml [2.5–46.7]; P < 0.0001), whereas, in control subjects, the difference was only 3.5-fold higher (30 [10–55] vs. 8.7 mU/ml [4.1–20.8]; P = 0.012). No correlation between intravitreal and serum erythropoietin concentration was detected either in diabetic patients or in the control group.

Erythropoietin mRNA was detected in human retina from both nondiabetic control donors and diabetic donors (Fig. 2). A significantly higher expression was observed in RPE cells than in neuroretina in the whole group (1.62 ± 0.61 vs. 0.39 ± 0.13; P = 0.02). Higher expression was also observed in RPE cells than in neuroretina when analyzing diabetic donors and nondiabetic donors separately, but the differences were not statistically significant (2.21 ± 1.08 vs. 0.54 ± 0.23; P = 0.18 and 0.93 ± 0.24 vs. 0.21 ± 0.06; P = 0.07) (Fig. 2). In addition, higher expression of erythropoietin mRNA was detected in RPE cells (2.4-fold) and in the neuroretina (2.6-fold) from diabetic donors in comparison with nondiabetic donors.

In the present study, we have found that intravitreal erythropoietin concentrations were markedly higher in patients with DME than in control subjects. Furthermore, intravitreous erythropoietin levels in DME were in the same range as that obtained in PDR. After correcting for vitreal proteins, the differences were not so evident but remained at significant levels. In this regard, it should be mentioned that the rationale for performing the correction for intravitreous proteins is to explore the source (intraocular synthesis versus blood borne) of the specific peptide that we want to analyze rather than to obtain a better index of its activity. In the present study, erythropoietin serum concentration was also determined and it was strikingly lower than that obtained in the vitreous, not only in diabetic patients with DME and PDR (∼30 times less) but also in control subjects (∼3.5 times less). Therefore, it seems clear that the elevated concentrations of erythropoietin obtained in the vitreous fluid of diabetic patients are not serum derived, and in consequence, the crude concentration of erythropoietin rather than the ratio of erythropoietin to vitral proteins better reflects the production of erythropoietin by the retina. In addition, our results strongly suggest not only that erythropoietin is synthesized by the retina but also that its production is upregulated in diabetic subjects. Confirming this point, in the present study, erythropoietin RNA expression was detected for the first time in the adult human retina, and its expression was significantly higher in diabetic than in nondiabetic donors. In addition, in both diabetic and nondiabetic donors, erythropoietin expression was found to be more abundant in RPE than in the neuroretina.

Unlike PDR, DME is a condition in which the ischemia is not a predominant event. In this regard, it should be noted that fluorescein angiography confirmed the lack of significant retinal ischemia in all patients with DME included in the study. Therefore, our results strongly suggest that stimulating agents other than ischemia are involved in the increased intravitreous levels of erythropoietin found in DME. Watanabe et al. (18) observed an increase in the vitreous erythropoietin levels in patients with inflammatory eye diseases. Several data indicate that inflammation is involved in the pathogenesis of both PDR (22) and DME (6–9,23). Therefore, inflammation might contribute to the high levels of erythropoietin observed in patients with DME. Hyperglycemia could be another factor that induces erythropoietin production. Although there are no studies evaluating the effect of glucose on erythropoietin expression in retinal cells, a direct relationship has been shown between glucose and erythropoietin concentrations in a Chinese hamster ovary cell line (24). In addition, we have found higher expression of erythropoietin mRNA in the retinas from diabetic donors without diabetic retinopathy in comparison with retinas from nondiabetic donors. Therefore, it seems that erythropoietin overexpression is an early event in the retina of diabetic patients, and it is unrelated to the ischemic process. Furthermore, this finding supports the concept that ischemia is not necessary to induce erythropoietin expression in the diabetic eye.

It has been suggested that erythropoietin plays an important role in PDR development, and its blockade could be beneficial for the treatment of PDR (18). However, whether the increase of erythropoietin concentration observed in DME has a pathogenic action or, by contrast, might be part of a self-regulated physiological protection mechanism to prevent retinal damage needs to be clarified. From the data available so far, the second hypothesis seems the most likely (13). Martínez-Estrada et al. (25), using an in vitro model of bovine blood-brain barrier (BBB), provided evidence that erythropoietin protects against the VEGF-induced permeability of the BBB and restores the tight junction proteins. Erythropoietin treatment also prevents an increase in BBB permeability in a rat model of induced seizures (26). Since BRB is structurally and functionally similar to the BBB (27), it is possible that erythropoietin could act as an antipermeability factor in the retina. In this regard, a substantive improvement of DME after treating azotemia-induced anemia with erythropoietin has been reported (28). In addition, there is growing evidence that erythropoietin is a neurotrophic factor (11,12). In this regard, it has been shown that erythropoietin protects cultured neurons from hypoxia and glutamate toxicity (29–31), and its systemic administration reduces neuronal injury in animal models of focal ischemic stroke and inflammation (32–34). Retinal neurodegeneration is an early event in diabetic retinopathy, and therefore, it might be possible that higher production of erythropoietin is needed as a neuroprotective factor. In fact, Junk et al. (32) demonstrated that erythropoietin administration is associated with both histopathological and functional protection of retinal neurons in a model of transient global retinal ischemia. Grimm et al. (15) reported that erythropoietin expression induced in the hypoxic mouse retina protects against light-induced retinal degeneration. Tsai et al. (35) demonstrated that intravitreal administration of erythropoietin has a protective effect on the viability of retinal ganglion cells in a rat model of glaucoma. Layton et al. (36) demonstrated in primary retinal cultures that erythropoietin’s neurothrophic function is attenuated at glucose concentrations similar to those which occur in diabetes. Apart from its antipermeability and neuroprotective actions, erythropoietin exerts an anti-inflammatory effect on the brain (37,38), and this action might also be extrapolated to the diabetic retina. Finally, it has been demonstrated that erythropoietin protects against high-glucose–induced apoptosis as well as the deleterious effect of free radicals (39,40). For all these reasons, it seems that erythropoietin is crucial for retinal homeostasis and its enhancement in DME could be a physiological consequence rather than a pathogenic contributor.

The strikingly high intravitreous erythropoietin levels detected in patients with DME raises the question of why these patients do not develop neovascularization as occurs in PDR patients. At present, we do not have any reliable explanation for this. However, it should be considered that the development of neovessels, rather than depending on a single angiogenic factor, is mediated by the balance between angiogenic and antiangiogenic factors. Therefore, in patients with DME, other factors counteracting the angiogenic effect of erythropoietin and favoring the permeability of BRB must exist.

In conclusion, we have demonstrated that intravitreal erythropoietin levels are increased in patients with DME at the same range as that observed in PDR subjects. In addition, we have demonstrated mRNA erythropoietin expression in the human retina and that its upregulation in diabetic subjects is an early event in the development of diabetic retinopathy. These findings suggest that other factors unrelated to the ischemia might be crucial in the upregulation of erythropoietin. Further studies are needed to investigate the precise role of erythropoietin, not only in the setting of DME but also in the normal human retina.

This study was supported by grants from Novo Nordisk Pharma (01/0066), the Instituto de Salud Carlos III (G03/212 and C03/08), and the Ministerio de Ciencia y Tecnología (SAF2003-00550).