Because eyes were received from various eye banks around the United States, they had been fixed in a variety of fixatives, from paraformaldehyde-glutaraldehyde mixtures to 10% formalin. Portions from the postbulbar region of the optic nerve, 1 to 2 mm behind the globe, were embedded in Epon for electron microscopy, or in paraffin for immunohistochemistry. In this region, the central retinal artery and vein are still located in the center of the nerve. Semithin cross sections were stained with toluidine blue and fuchsin. Quantitative analysis of sections was performed similar to our previous reports,
8 10 based on the method of Quigley et al.
11 The area of the entire nerve, excluding the meninges, was measured. The area of the nerve fiber bundles and the area of connective tissue between the bundles and surrounding the central retinal vessels were measured separately. For determination of the number of axons, the cross section of the optic nerve was divided into eight sectors. In each sector, axons in five sample areas of 1000 μm
2 were counted. These sample areas extended from the central to the peripheral nerve at equal distances from each other. The mean of all measurements was multiplied with the nerve fiber area, to yield the total axon counts.
The postbulbar retrolaminar portion of the optic nerve was chosen for study because of the regularity of the tissue and ease of making quantitative measurements when compared with the bowing and cupping of the laminar region. This retrolaminar portion of the optic nerve receives most of its blood supply from the pial arterioles, which penetrate the connective tissue septa and branch into capillaries; only a few capillaries originate at the central retinal artery (CRA).
12 Evaluation of both the capillaries and arterioles was performed. Within the connective tissue septa, both the total number and the density of capillaries were determined. The total number of capillaries was determined by dividing the cross section of the nerve into quadrants. In the middle of each quadrant, three regions with an area of 0.155 mm
2 each were analyzed at 40× magnification. These regions were spaced at equal distances from the central to the peripheral portions of the nerve. The mean of these 12 regions was multiplied with the entire cross-sectional area of the nerve, to produce the capillary number per nerve. The potential tortuosity of capillaries did not allow the determination of their absolute number, and so a simple count of the number of visible capillary lumens (or capillary profiles) present in a cross-section was made
(Fig. 1) . The measurements were performed by two of the authors (JG, AK) independently, and the average of the data was taken. The
density of capillaries was defined as the total number of capillaries per square millimeter of total nerve area.
Arterioles within the pial sheath of the nerve and the CRA were evaluated for arteriosclerosis, and a grading system was devised. In stage 1, there was no identifiable arteriosclerosis. Stage 2: The basement membrane of the endothelium was thickened and at places showed small infiltrations of light-microscopic homogeneous material. The lumen of the arteriole was not narrowed; in the CRA, in addition, the internal elastic membrane was partly fragmented, and an ingrowth of muscle cells in the intimal layer was possible. Stage 3: Arterioles showed a concentric narrowing of the lumen due to a lift-off of the endothelium by additional infiltration of amorphous material; in the CRA arteriosclerotic plaques occurred
(Fig. 2) . Grading was performed independently by two pathologists (Amann K, Institute of Pathology, University of Erlangen-Nürnberg, Erlangen, Germany; and Blaser B, Institute of Pathology, City Hospital, Bamberg, Germany). Limited information on general medical diseases, such as systemic hypertension and myocardial infarction, was available, and we were unable to make an assessment of general arteriosclerosis.
For electron microscopic study, 11 of the best-preserved POAG nerves spanning the entire range of axon loss, from 790,000 to 42,000, were selected. Similarly, eight PEXG specimens (axon range, 170,000–872,000), and six age-matched normal specimens were selected. Ultrathin sections were viewed with an electron microscope (model EM902; Zeiss, Oberkochen, Germany).