With the intention of illuminating the in vivo role of astrocytes, we have made use of the highly cell-and region-selective toxicity of α-chlorohydrin (S-3-chloro-propane-1,2-diol) to study the response of the brain vasculature to selective loss of this cell type. α-Chlorohydrin is a glycolytic inhibitor that is toxic to astrocytes, but did not increase permeability of monocultured endothelial cells (Romero et al. 1997). Systemic administration of α-chlorohydrin (140 mg kg^-1 I.P.) to 180-220 g male F344 rats produced astrocytic death limited to specific deep nuclei (most notably the inferior colliculi, and the red, vestibular, trigeminal motor and occulomotor nuclei). This occurred over the first 2-24 h, and was followed by neuronal loss from 36 h. From 2 to at least 8 days no astrocytes could be detected within damaged areas by confocal immunohistochemistry (glial fibrillary acidic protein) or electron microscopy. Due to the limited and strictly symmetrical nature of the damage, rats showed only a mild ataxia, and little weight loss, and all experiments were conducted within UK Home Office guidelines. Despite the marked early astrocytic pathology, blood flow in the inferior colliculus (measured by hydrogen polarography in conscious animals previously implanted under anaesthesia) showed no significant change until 24 h, but by 48 h had risen to 2.68 ± 0.50 times predose (mean ± S.E.M., n = 9), falling thereafter. In contrast, flow in the resistant cerebellar cortex increased only 1.25 ± 0.08-fold at 48 h. In parallel with this late flow increase, the vasculature in damaged areas became permeable to 10 kDa fluorescent labelled dextran tracers, and severely damaged areas developed petechial haemorrhages, both effects peaking at 2-3 days. Extravasation of 0.5 kDa gadolinium-DTPA was visualised by T1 weighted magnetic resonance imaging under isofluorane anaesthesia, and at 2 days in the inferior colliculus post-gadolinium enhancement reached 61.3 ± 6.6 % (n = 10) of that seen in the open barrier pineal. Despite the continuing lack of astrocytic contact, by 9 days damaged areas were no longer permeable to 10 kDa dextran, and gadolinium enhancement had fallen to 42.5 ± 2.7 % of its 2 day value (n = 10). These results show that recovery of barrier properties is molecular size dependent.

We hypothesise that the brain endothelium undergoes a reversible dedifferentiation in response to loss of astrocyte contact, but is then capable of a surprising degree of recovery of barrier properties, possibly due to distant diffusible astrocyte-derived factors.

This work was funded by the MRC.