In this study, we identified a novel rat model with a defect in a cilia gene that mimics specific phenotypic changes which are likewise observed in diabetic retinopathy. In contrast to diabetic retinopathy, however, this model exhibits primary neuronal degeneration followed by vasoregression with a loss of capillary endothelial cells and pericytes during the second phase. Subsequent progressive impairment of retinal functions and concomitant glia activation and induction of neurotrophins in the retina of transgenic rat occur, which are at least in part similar to the pathologic evolution in diabetic retinopathy. While the expression of FGF2 and CNTF likely reflects a response to photoreceptor damage 
, NGF regulation may have a close link to retinal vasoregression.
An important finding in this study is the temporal attribution of neuronal damage with vasoregression. In TGR rats, photoreceptor cell death via apoptosis with a loss of cells in the inner nuclear layer develops during the first month prior to vasoregression which follows during the second month. Despite similarities of the neuronal changes in TGR rat with several animal models such as the rd/rd, rds mouse and Royal College of Surgeons (RCS) rat, and reported vascular alterations, specific differences exist regarding the onset and developmental dynamics of the photoreceptor death. For example, in the rd/rd mouse, rod cell death starts at the second week and is completed by the third week 
. In this model, the vasculature starts to regress while still developing 
. Vascular changes are unknown in the rds mouse model with early onset at 2 weeks and slow progression of retinal degeneration up to 12 months. Moreover, late vascular changes due to the progressive loss of photoreceptors are reported from RCS rats 
, but neither the temporal and spatial allocation nor an exact quantitation of vascular damage has been provided.
We observed both exponential vasoregression and pericyte loss in the TGR model. Vasoregression is also the hallmark of diabetic retinopathy in both, humans, and animal models 
. Changes in pericyte coverage of microvessels, in endothelial cells and an increasing number of acellular capillaries of TGR retinas are qualitatively similar to those in experimental diabetic retinopathy 
. Pericyte loss is an early and archetypical feature of diabetic retinopathy. In contrast to the diabetic model, in which pericyte loss appears to be causally linked to glucose-induced changes in angiopoietin-2 expression, the mechanism of pericyte loss in the TGR model remains yet to be determined. Neither glial activation which occurs prior to pericyte loss in the TGR model nor endothelial cell loss which starts in parallel with pericyte loss might be responsible, since key factors determining pericyte recruitment are not altered in TGR model. Angiopoietin-1 is produced by glial cells, and PDGF-B by endothelial cells. Neither of which are changed in the TGR model (unpublished data). Pericytes act as survival factors for established capillaries, and their loss may therefore be relevant in vasoregression 
. However, despite the similar degree of pericyte dropout in the TGR compared to the diabetic model, the degree of acellular capillaries differs substantially between the two models. Therefore, pericyte loss is unlikely to explain the enormous level of vasoregression in the TGR. The final cause for the exorbitant demise of vessels remains unclear, but it is obvious that the disturbed integrity of the neurovascular unit is responsible.
One factor relevant for proper retinal vessel function and neurovascular integrity, i.e. VEGF, does not respond to the neuronal damage in the TGR model. In contrast to other models of retinal degeneration such as the rho-/- mouse 
in which VEGF is reduced, we did not find hypoxia in the TGR model, and consistent with the lack of hypoxia, VEGF was not regulated with progressive vascular regression. The absence of VEGF regulation is one of the discrepancies between the TGR and the diabetic retinopathy model in which VEGF is induced by hyperglycemia and the resultant increase in oxidative stress and advanced glycation endproducts formation. However, we also did not find a downregulation of VEGF as has been observed in some models of neurodegeneration, in which VEGF deficiency can impair neuronal survival.
Neurotrophic factors expressed in neuroglial cells of the retina, such FGF2, CNTF, and NGF, may play an important role in retinal neurodegeneration. NGF injection rescues photoreceptor degeneration in the RCS rat model of retinitis pigmentosa, involving secondary effects by other neurotrophins such as FGF2 and VEGF 
. NGF also inhibits retinal degeneration in the C3H mouse 
. In a hindlimb ischemia model, NGF induced angiogenesis, suggesting that NGF upregulation is protective against both, neurodegeneration and vascular regression 
. Our previous data suggest that NGF treatment of diabetic rats prevents both, early neuroglial damage and the development of pericyte loss and vasoregression 
. Our present data indicate that NGF is only upregulated after the onset of vasoregression, and after substantial neuronal cell loss has occurred.
In contrast, the two other neurotrophic factors studied, CNTF and FGF2, were upregulated prior to vasoregression and are found to be in close relationship with neurodegeneration. CNTF can delay photoreceptor degeneration in several models of genetic degeneration and ischemic injury. It is known that endogenous CNTF is upregulated in response to retinal injury, but the effect might be indirect (glia-mediated) rather than a direct effect, since the presence of CNTF-receptors on photoreceptors has not been unequivocally demonstrated. CNTF belongs to the CNTF/LIF group of cytokines. These have been extensively studied for their role in photoreceptor development. However, much less is known about the impact on vascular function. Recently, Kubota et al have demonstrated that LIF is involved in regulating microvessel density by regulating VEGF expression in mice 
. LIF-/- mice had a denser capillary network with sustained tip cell activity, and despite resistance to hyperoxic vasoregression, they developed more neovascular tufts. These data suggest that there may be a link between the vasoregressive phenotype and increased expression of the CNTF/LIF family of cytokines in the retina of TGR. CNTF expression parallels that of FGF2 in our TGR model. FGF2 knockout mice develop photoreceptor degeneration suggesting that FGF2 plays an important role in photoreceptor development and survival. Multiple degenerative and injurious retina models yield upregulation of FGF2 suggesting the pleiotropic and essential role for FGF2 in survival of retinal cells. The lack of functional FGF2 could be a factor that causes the impaired protection of the diabetic retina from progressive vasoregression during the non-proliferative phase. Thus, in contrast to CNTF upregulation, FGF2 appears to be regulated as a response to injury type of tissue reaction in the TGR rat.
The Müller cell and astrocyte, which interconnect vessels and neurons through their end-foot processes, participate in neuroprotection and neuronal repair after injury 
. Thus, increased GFAP expression as early as after 1 month suggests that glial activation occurs in response to neuronal degeneration in TGR rats. However, upregulation of GFAP is unspecific, as it occurs in various retinal injury models such as axotomy, retinal ischemia, retinal detachment and diabetic retinopathy. In contrast to the diabetic model, glial activation in the TGR model was not sufficient to induce VEGF, while bFGF upregulation is possibly the result of glial activation.
Mutations in cilia genes or defective cilia genes are associated with renal abnormalities, retinal degeneration, liver and respiratory diseases in patients 
. In a previous study using the TGR rat, defective polycystin-2 gene was found in the region of connection cilia of outer and inner segments of rod and cone photoreceptors 
. However, electron microscopy studies did not reveal any morphological cilia defects except photoreceptor degeneration in the TGR (unpublished data). The precise function of polycystin-2 in the retinal cilia remains unclear. Changes of polycystin-2 gene expression in the retinal cilia may lead to defective transport of functional proteins through this apparatus, although no change in the morphology of cilia in the TGR retina has been detected so far. The link between the genetic defect of polycystin-2 and the apoptosis of photoreceptors requires further investigations. Since the genetic makeup of the TGR rat has no human correlate, the model must not be considered as an animal model reflecting a specific human disease.
In summary, this study provides insight into the relationship between neurodegeneration, glial activation and vessel regression. Further evaluation of the molecular mechanisms involved in vasoregression in the TGR rat and comparative studies with models of vasoregression of diabetic origin supposedly yield novel targets for intervention of retinal vasoregression.