tin in both cell types. CTGF expression was decreased with TGF-b inhibition in BRPCs only. Fibronectin protein was present in higher levels in BRPCs. These results show that TGF-b has differential effects on ECM-related gene expression in BRECs and BRPCs. Pericytes are more responsive to TGF-b, and CTGF expression seemed to be regulated by TGF-b in pericytes and not in endothelial cells.

In summary, this study showed that retinal pericytes in particular have the essential characteristics to allow for a role of TGF-b in BL thickening in PCDR. Pericytes are of mesenchymal origin like fibroblasts, which may explain their TGF-b-dependent CTGF regulation. These results suggest that in retinal endothelial cells, CTGF expression is regulated by other pathways and factors, acting independently of TGF-b, such as VEGF, AGEs, and/or high glucose levels [26].

BL Thickening in Diabetic CTGF-Knockout Mice

As indicated above, STZ-induced diabetes in rodents is associated with a twofold increase in CTGF gene expression in total retina, which can be attenuated by treatment with the ACE-inhibitor perindopril or aminoguanidine, respectively [49, 135]. We studied the effects of STZ-induced diabetes on retinal capillary BL thickness in transgenic CTGF+/− mice [136] and wild-type mice (CTGF+/+) [124]. BL thickness was calculated by quantitative analysis of electron microscopic (EM) images of transversally sectioned capillaries in and around the inner nuclear layer of the retina. In the retinal capillaries, a significant increase in particularly the endothelial cell BL was detected in diabetic CTGF+/+ mice as compared to control CTGF+/+ mice, using two independent quantitative methods in EM images (Fig. 5). This preferential thickening of the endothelial BL and pericyte BL in diabetic mice had been observed previously [137].

In this study, the CTGF+/− and CTGF+/+ mice were in a similar diabetic state with respect to blood glucose levels. However, there was a clear genotype effect on CTGF expression in the CTGF+/− mice. Approximately 50% lower CTGF protein expression levels in plasma and urine were found in control animals lacking one CTGF allele. Retinal CTGF levels were not analyzed in this study. However, renal CTGF mRNA levels in diabetic CTGF+/− mice were 50% of those in diabetic CTGF+/+ mice. This suggests that retinal CTGF protein levels may also have been lower and prevented the diabetesinduced BL thickening of the retinal capillaries. Renal TGF-b1 mRNA levels were significantly increased due to diabetes, irrespective of the CTGF genotype. Similarly to the retinal vessels, a genotypic effect on the BL of glomeruli was found in diabetic mouse kidney [115].

Taken together, the data of this study indicate that CTGF is necessary for BL thickening in diabetes. This provides important direct evidence for an essential role of CTGF in diabetic retinal BL thickening. In concert with the supportive indirect evidence for such a role as described above, these data identify CTGF as a possible therapeutic target to prevent early changes in PCDR. This may be clinically relevant, as experimental animal studies have shown that prevention of BL thickening can ameliorate the subsequent development of other preclinical changes in DR [38].

Fig. 5. Examples of retinal capillaries analyzed for BL thickness. Distinct regions of the BL are identified as endothelial BL (eBL), pericyte BL (pBL), and joint endothelial cell and pericyte BL (jBL). Note the diabetes-induced BL thickening in diabetic CTGF+/+ mice (B) as compared with control CTGF+/+ mice (A) and the absence of this effect in diabetic CTGF+/− mice (D) compared with control CTGF+/− mice (C). Bar = 1 mm. (Reproduced from: Journal of Histochemistry and Cytochemistry. Online by Kuiper EJ et al. Copyright 2008 by Histochemical Society Inc. Reproduced with permission of Histochemical Society Inc in the format Trade book via Copyright Clearance Center).

CTGF in PDR

In PDR, CTGF was found in fibrovascular membranes, predominantly localized in myofibroblasts [104, 107], with a significant correlation between the number of a-SMA-positive myofibroblasts and the number of myofibroblasts expressing CTGF [104]. Myofibroblasts are activated matrix-producing fibroblasts, associated with (persistent) fibrosis [59]. Furthermore, CTGF was detected in endothelial cells in these membranes [104]. In the vitreous of a small series of patients with active PDR, levels of the N-terminal CTGF fragment were increased as compared to nondiabetic patients and patients with quiescent PDR [107]. Vitreous levels of full-length CTGF were similar in all groups, whereas the C-terminal fragment was not detectable. N-terminal CTGF levels were also higher in diabetic patients with vitreous hemorrhage than in nondiabetic patients with vitreous hemorrhage, who had similar N-CTGF levels as nondiabetic controls. This finding suggests that local synthesis of CTGF plays a role in PDR. On the basis of the association between CTGF levels and PDR, these authors concluded that CTGF has a role in angiogenesis. However, we showed that elevated CTGF levels are associated with degree of fibrosis and not with angiogenic activity in vitreoretinal conditions, including PDR, in a series of vitreous samples of 119 patients (Fig. 6) [101].

Fig. 6. Geometric mean of CTGF levels in relation to degree of fibrosis. Fibrosis was graded as 0 when no fibrosis was present, 1 with only a few preretinal membranes present, 2 with some proliferative membranes/PVR grade a/b, or 3 with abundant proliferative membranes/PVR grade c/d. Error bars represent the 95% confidence intervals. (Reproduced from [101] Copyright © (2006) American Medical Association. All rights reserved).

In addition, the degree of fibrosis was best predicted by CTGF levels. Possibly, TGF-b has a role in regulating CTGF levels intravitreally and thereby fibrosis in DR. An earlier study has shown that TGF-b2 was associated with fibrotic proliferation in the vitreous of patients with PDR [138]. Furthermore, vitreous levels of both TGF-b2 and CTGF in patients with PDR were significantly higher than in those with nonproliferative diseases, with a correlation between the levels of TGF-b2 and CTGF [103].

Role of CTGF and VEGF in the “Angiofibrotic Switch” in PDR

In PDR, neovascularization progresses to a fibrotic phase. VEGF is considered to be the primary angiogenesis factor in this process [2, 6]. In vitreoretinal disorders (including PDR), N-terminal CTGF levels in the vitreous are elevated [107] and are strongly correlated with the degree of fibrosis [101]. Therefore, it was proposed that CTGF is a causal factor of fibrosis and scarring in PDR.

In vitreous of PDR and PVR patients, Kita et al. [139] found no significant correlation between the levels of CTGF and VEGF, even though concentrations of CTGF and VEGF were both significantly higher compared to those in vitreous from patients with nonproliferative diseases. With regard to a possible role of CTGF in retinal neovascularization, it was concluded that CTGF may have no direct effect on retinal neovascularization, but possibly works indirectly by modulation of VEGF levels.

We investigated the correlation between VEGF and CTGF levels and the degree of fibrosis and neovascularization in the vitreous of a series of 68 patients with PDR and other vitreoretinal disorders (macular hole or macular pucker) [111]. Neovascularization

Fig. 7. Mean levels of CTFG (A, D), geometric mean levels of VEGF (B, D), and mean ratio CTGF/log10(VEGF) (c, f) in relation with degree of neovascularization (A–C) and degree of fibrosis (D–F) in the vitreous of 32 PDR patients. Vertical bars represent 95% confidence intervals. Significant differences between groups are indicated. (From [101]).

and fibrosis in various degrees occurred almost exclusively in PDR patients, in which vitreous CTGF levels were significantly associated with the degree of fibrosis and with VEGF levels, but not with neovascularization. On the other hand, VEGF levels were associated only with neovascularization, in agreement with the widely accepted role of VEGF as the major angiogenic factor in PDR (Fig. 7). As the ratio of CTGF and VEGF levels was the strongest predictor of the degree of fibrosis, the results suggested that the balance of VEGF and CTGF levels in the vitreous determines progression of fibrovascular proliferation in PDR.

These findings led to the following concept of regulation of angiogenesis and fibrosis in ocular disease and in wound healing in general: angiogenesis in the vitreous is driven by VEGF, which upregulates the profibrotic factor CTGF in various cell types in the newly formed neovascular membranes. The elevated CTGF levels do not significantly

Fig. 8. Hypothesis of the angiofibrotic switch in PDR. Angiogenesis in the vitreous is driven by VEGF, which upregulates the profibrotic factor CTGF. Increasing levels of CTGF inactivate VEGF, and when the balance between these two factors shifts to a certain threshold ratio, the angiofibrotic switch occurs: angiogenesis ceases, and fibrosis driven by excess of CTGF leads to scarring and blindness.

Fig. 9. Fundus photographs of a patient with PDR and new vessels along the lower vascular arcade, before (A) and at 8 months after (B) an injection with bevacizumab followed by pan-retinal photocoagulation. Note the increase in fibrosis after combined anti-VEGF and laser treatment (B).

contribute to ocular angiogenesis. In contrast, increased levels of CTGF sequester VEGF, and when the balance between these two factors shifts to a certain threshold ratio, the angiofibrotic switch occurs: angiogenesis ceases, and fibrosis driven by excess CTGF leads to scarring and blindness (Fig. 8).

This concept predicts that a sharp decline in VEGF levels in a patient with active neovascularization due to PDR inhibits angiogenesis, causes the angiofibrotic switch, and temporarily increases fibrosis. This is supported by clinical observations in patients with active neovascularization treated with intravitreal inhibitors of VEGF, such as bevacizumab and ranibizumab, and/or pan-retinal laser photocoagulation, which destroys large areas of retina and markedly reduces intraocular VEGF levels [5]. Regression of neovascularization and the predicted temporary increase in fibrosis was observed in a nonsystematic survey of a small series of patients (Fig. 9) [111]. Others have also reported