Brain Research, 585 (1992) 431-434 © 1992 Elsevier Science Publishers B,V. All rights reserv'cd 0006-8993/92/$05,00 431 BITES 25253 Axon types classified by morphometric and multivariate analysis in the rat optic nerve Joaquin De Juan, Nicolfis Cuenca, Carlos Ifiiguez and Eduardo Fernfindez Departamento de Histologfa e Institute de Neurociencias, Unit,ersidad de Alicante, Aiicante (Spain) (Accepted 31 March 1992) Key words: Nerve fiber; Cluster analysis; Visual pathway; Myelin sheath; Cytoskeleton; Optic nerve Calibers of the rat optic nerve axons distribute unimodally and it is difficult to distinguish groups among them. However, these fibers arose from 3 types of ganglion cells and showed 3 conduction velocities. Performing a cluster analysis over several uitrastructural parameters we found 3 main groups of fibers. These groups are present in a very similar proportion to the ganglion cells groups described in the rat retina. Most of the rat optic nerve axons originate in the retinal ganglion cells although a few of them run from the pretectal area to the retina It. Several studies indi- cate that both area and caliber of cross-sectioned axons distribute unimodally a's'6, which makes it difficult to differentiate groups between these axons I'1°. Neverthe- less, the fibers of this nerve are far from being a homogeneous population since at least 3 main groups of ganglion cells have boon recognized in rat retinas H'~5 and 3 different kinds of axons according to their con- duction velocity sa~. The aim of this work is to classify the optic nerve axons by examining a combination of ultrastructurai parameters using a cluster analysis, which is a statistical procedure for distributing objects into homogeneous groups based on inter-object simi- larities n. The fibers fell into 3 main groups. The rela- tions among these groups with the axon types de- scribed in electrophysiological studies and the different types of ganglion ceils are discussed. Three one-year-old albino Wistar rats were anes- thetized with 35% chloral hydrate (1 ml/kg, i.p) and transcardially perfused with 4% paraformaldehyde, 2% glutaraldehyde in 0.1 M phosphate buffer (pH 7.4). Prechiasmatic portions of the nerves were dissected and kept overnight in the same fixative at room tem- perature, post-fixed in 2% OsO4 for 1 h and embedded in Epon 812. Electronmicrographs were taken in a ZEISS C-10 electronmicroscope and printed at a final magnification of x 140,000. With the aid of a com- puter-linked planimeter and a morphometrical package designed by us4, the following axonal parameters were determined: area of cross-sectioned axons and fibers, number of microtubules and neurofilaments, myelin sheath thickness and G-ratio (axon diameter/fiber di- ameter). These parameters were measured on 100 axons per section randomly chosen from photomicrographs as previously described 2'a. Seven agglomerative clustering algorithms from the SPSS statistical package I~ were applied on the parameters obtained from one nerve. G-ratio was excluded from the analysis since it may bias the covariance structure. The algorithm that pro- duced the clearest cluster structure and the highest number of subgroups was Ward's method. Statistical validation was performed by the following procedures: (1) correlation between the groups yielded by the dif- ferent cluster methods by constructing contingency ta- bles and measuring the degree of correlation by Cramer's Vla, (2) assessment of cluster consistency by obtaining randomly selected subsamples of half the data from each optic nerve and comparing these sub- samples to the total sample cluster by contingency Correspondence: J. De Juan, Departamento de Histologia, Facuitad de Medicina, Apdo. Correos 374, Alicante 03080, Spain. Fax: (34) (6) 565-8511.