We report here that in mice expressing elevated levels of PDGF-B in nestin positive cells of the retina during development, astrocytes fail to populate the retina and vascular progenitors do not colonize this tissue to produce a network of vessels. Interestingly, with the exception of their eyes, nes/tk-PdgfB-lacZ mice do not exhibit any profound alteration in phenotype 
. The transgenic eyes were smaller than normal with an iris of a reduced diameter and with frequent occlusions. The transgenic retinas had folds and adhered to the back of the lens, forming a retrolental mass that made dissection difficult. This morphology was accompanied by frequent bleeding and by the age of one year the transgenic retina had deteriorated completely. Many of these defects in the eye are likely to be associated with abnormal traction forces within the structures that eventually also leads to retinal detachment (). Patients with persistent fetal vasculature syndrome (PFVS) 
display similar problems 
with a large individual variability that was also seen among the nes/tk-PdgfB-lacZ mice. Many of the transgenic mice developed uni-lateral microphtalmia (around 40–50% pups per litter) and/or a cataract. These abnormalities in retinal structure of the transgenic mice are similar to PFVS, which results from failure to degrade the hyaloid vasculature.
The crucial role of PDGF signaling in connection with the normal formation of retinal blood vessels is well-established 
. The functions of PDGFRα-signaling include regulation of the migration and proliferation of astrocyte precursor cells, while signaling by PDGFR-β recruits mural cells, which allows for their proper attachment to the endothelial cells of the blood vessels 
. Seo and coworkers 
have proposed that gain-of-function mutations in PDGF-B exert more profound effects on retinal development than mutations in PDGF-A, due to the ability of the former to activate both α and β receptors, present on different types of cells. This suggestion was confirmed in studies involving overexpression of different isoforms of PDGF in photoreceptors 
and further supported by the findings of Vinores et al 
. In our previous study 
involving transgenic expression of PDGF-B under control of the myelin basic protein promoter a disorganized neural retina with an under-developed capillary network was observed. The transgenic mice employed here display more severe alterations in eye phenotype, probably because the transgene is expressed more widely, both spatially and during the period of development examined. By binding to both the α and β receptors 
the B isoform of PDGF could affect the hyaloid as well as astrocytes 
and pericytes 
, cells that form and provide support for the vasculature during a critical period of retinal development.
The defects in lamination discernible shortly after birth are consistent with the onset of transgene expression on E17.5 with peak expression on P1. Since no major effects on the representation of retinal cells, on their relative spatial orientation or on their proliferation were seen, despite the overall disorderly structure, we conclude that the transgene expression did not affect the general formation, differentiation or migration of retinal progenitors.
The elevation of the number of caspase positive cells observed on P5 occurred at a time when this tissue normally undergoes developmental cell death 
. Cell death is part of the normal development of a functional nervous system and half of all postmitotic retinal ganglion cells die during the 2 first post-natal weeks in mice 
. This period of cell death is associated with trophic interactions within the retina and with the central targets for the retinal ganglion cells, and the extent of death is determined in part by what connections the cells can establish. The fact that we see more apoptosis in the transgenic mice during this period suggests that the increased cell death is associated to the structural changes in the retina and is not a direct effect of the transgene. Moreover, the deterioration of the retina, as seen after one year, is probably not related to this increased apoptosis but rather related to the malfunctioned blood supply to the eye.
The inner and outer segments of the photoreceptor layer in the transgenic animals did not develop normally. The rhodopsin localized in patches or inside the rosette formations and the horizontal cells, which normally form synapses with photoreceptors 
, developed ectopic neurites that extended outside the outer plexiform layer and into rosettes. This pattern persisted in the adult retina for several months indicating that these cells may have formed synapses with the PR, despite their anomalous location.
The up-regulation of nestin and GFAP in Müller glia cells indicated that the transgenic retina experienced a gliotic response that is likely to be triggered by ischemia due to the missing retinal vasculature. Mechanical traction may be an alternative explanation since Müller glia is known to exert tractional forces when stimulated by PDGF in vitro 
. It seems likely that traction also occurred in the transgenic retinas. During normal retinal vascularization, astrocyte precursor cells migrate from the optic nerve to populate the inner portion of the retina, where they form a scaffold on which vascular progenitor cells can migrate and form vessels 
. In the nes/tk-PdgfB-lacZ mice most of the Pax2/GFAP positive cells remained associated with the vitreal side of the retina, and consequently, no astrocyte network on which vessels could expand was laid down. The lack of proper formation of new vessels was also apparent from the staining of CD31 and NG2 in both flat-mounts and cross-sections of the nes/tk-PdgfB-lacZ retina. This staining demonstrated defective vascularization, with few capillary-like structures, in a seemingly random fashion and devoid of large trunk vessels. Our results suggest that over-expression of PDGF-B delayed regression of the hyaloid, and prevented APCs from migrating and spreading across the retina, which in turn prevented endothelial and mural cells from populating the retina to form a normal vascular network. Endothelial cells recruit pericytes by secreting PDGF-B 
and it is well established that pericytes require PDGF-B to remain attached to the vessel wall 
. The opposite situation, i.e. an excess of PDGF-B is apparently not as detrimental, since many transgenic pericytes remained attached to the endothelial cells. Therefore, the abnormal vessel properties cannot mainly be related to altered coverage by mural cells.
Intraocular pressure (IOP) in mice varies between strains 
, underscoring the importance of using non-transgenic littermates as controls when congenic strains are not available. The mice examined here had been back-crossed with C57Bl/6 mice for 5 generations, but the remaining contribution of the CBA genome is unknown. Nonetheless, the wild-type control mice, which were always siblings, exhibited an IOP close to the reference value for C57Bl/6 
. The attenuated IOP in our transgenic mice is consistent with our hypothesis that the perfusion of the eye is severely affected in the transgenic animals. The lack of proper retinal vessels prevented these mice from controlling the ocular circulation. Furthermore, traction on the ciliary body can lead to acute and chronic hypotony, followed by retinal detachment, as has been reported in patients with PFVS 
The effect of STI571 on the retinal phenotype of the transgenic mice depended on the time when this inhibitor was administrated. Administration prior to E17.5 caused abortion, demonstrating that PDGF signaling, and/or other tyrosine kinases sensitive to STI571 are critical for embryo survival. From E17.5 to the time of birth, exposure to STI571 did not affect the birth of live offspring. Since this time period coincided with the onset and early phase of transgene expression in the eye, the earliest effects of transgenic PDGF could be monitored. The timing of inhibition was of importance for APC maturation and spread 
. APCs were less compacted upon treatment with STI751, which apparently allowed CD31-positive cells to follow, since these cells were also less tightly packed in the presence of the inhibitor. However, the retinal vascularization was not greatly improved despite the better APC colonization. The observation that vessel distribution was not influenced at all by administration of STI571 from birth to P4 suggests that at this point it is too late to reverse the abnormal development of the vascular network. In contrast, if STI571 was administered to the mice during the second week of postnatal life, the hyaloid regressed in part. Thus, even though STI571 blocks all PDGF signaling, not only that originating from the transgene, we could discern partial rescue of the phenotype with this inhibitor.
The present investigation reveals that timely regression of the hyaloid vasculature and development of the adult retinal vascularization are prevented by over-expression of PDGF-B in nestin expressing cells during development (Fig. S9
). Failure of astrocyte precursors to form a scaffold precluded retinal vessel formation and gave rise to a defective retina that deteriorated with time. We also showed these mice to be useful for testing pharmacological intervention by the ability of a small-molecular inhibitor to partially restore retinal vascularization.