Ishida et al. reply:

We thank Ferguson and Stuart for their comments and appreciate the opportunity to respond. Our data show that leukocytes mediate physiological retinal vascular pruning and pathological vaso-obliteration. These effects were demonstrated in both CD18-deficient mice and in rats treated with antibody to CD18 (ref. 1). The results, therefore, transcend a single species.

Our data also identify a role for FasL as an apoptosis-inducing molecule in vascular pruning and vaso-obliteration. We generated the data by antibody-based blockade1 and did not rely on gld mice. In the discussion of our paper1, we noted the possibility of the existence of redundant molecular or cellular mechanisms in the processes of pruning and vaso-obliteration.

Although our data and the data obtained from gld mice2,3 are seemingly contradictory, they may in fact be complementary. Additional studies are required to confirm this. The phenomena that Ferguson and Stuart describe in gld mice are similar to those observed in our experiments. In our CD18-deficient mice, leukocyte adhesion and vascular pruning were delayed by approximately two days, but were not indefinitely suppressed. Thus, the retinal vasculature of the adult CD18-deficient mice was essentially normal1, as was the neural retina. We have learned that the timing of the analyses is critical, as vascular modeling occurs quickly. When placed in a hyperoxic environment for 72 hours, the CD18-deficient mice showed vaso-obliteration similar to that seen in wild-type mice (S.I., K.Y., T.U. and A.P.A., unpublished data), but significantly less vaso-obliteration was seen in CD18-deficient mice at 48 hours1.

Both vascular pruning and vaso-obliteration may represent defense mechanisms aimed at protecting the sensory retina from oxidative stress. As noted in our paper, compensatory systems may be used in vivo to preserve this important defense mechanism. We hypothesize that the gld mice data2,3 reflect this biological redundancy. FasL blockade did not suppress leukocyte adhesion in our experiments1. Under prolonged oxidative stress, we speculate that known leukocyte products, such as perforin and TRAIL, have a similar role in endothelial cell apoptosis in gld mice. We also noted1 that FasL-bearing blood-borne cells of nonleukocytic origin may have a role in pruning and vaso-obliteration4,5. It is therefore possible that leukocyte cytotoxicity does not depend on a single molecular pathway, and that FasL expression in the retina is not from a single cell type. A temporal examination of the role of blood-borne cells in pruning and vaso-obliteration is required in the genetically altered mice that Ferguson and Stuart describe.

The fact that gld mice show increased pathological neovascularization after hyperoxia-induced retinal ischemia2,3 is also consistent with our published data6. In our model of oxygen-induced retinopathy, T lymphocytes served as negative regulators of pathological neovascularization.

See FasL, leukocytes and vascular modeling by Ferguson and Stuart.