Summary.-Plasma membranes from 6 spontaneously metastasizing and 4 nonmetastasizing rat mammary carcinomata were isolated by discontinuous sucrose density gradient centrifugation of microsomal pellets. The starting microsomal fraction contained 40-50%o plasma membranes as determined by the levels of 5'-nucleotidase activity, with a negligible amount of nuclear (1%), mitochondrial (5%o) and lysomal (7%o) contamination. Five distinct fractions (F1-F5) were banded at densities 1 09, 1 13, 1 15, 1-17 and 1-21 at 25°C, in addition to a pellet (F6) obtained by centrifuging at 76,000 g for 17 h. The fractions Fl through F5, all contained various concentrations of membranous structures, while the pellet (F6) contained only amorphous materials as evidenced by electron microscopy. The F3 fraction at the gradient 1 15 had the highest specific as well as total activity of the plasma membrane marker enzyme, with aggregates of the least contaminated plasma membranes in vesicular forms. This fraction also had the lowest specific activity for glucose-6-phosphatase (smooth ER marker) and for ,B-D-glucuronidase (lysomal marker), and therefore was considered to be the " cleanest " plasma membrane fraction. When the activity of 4 additional plasma membrane marker enzymes, i.e., alkaline phosphatase, phosphodiesterase I, nucleotide pyrophosphatase and alkaline ribonuclease was determined in the same F3 fraction, their levels were significantly lower in every metastasizing tumour than in the non-metastasizing ones, with the enzyme activity decreasing in direct proportion to the metastasizing capacity. On the other hand, the marker enzymes were high in all non-metastasizing tumours, with the activity seemingly increasing with the immunogenicity of tumour cells. There was no significant difference between the 2 groups of mammary tumours in the levels of sialic acid, hexosamine, phospholipid or cholesterol in the plasma membranes. Thus, the level of plasma membrane marker enzymes is considered an accurate indicator for metastasizing capacity in the rat mammary tumour system. IN EARLIER studies (Kim and Pickren, 1974;Kim et al., 1975), it was demonstrated that the level of various plasma membrane marker enzymes correlated roughly with the amount of membrane bound glycoproteins or glycocalyx and the immunogenicity of rat mammary tumour cells, and, inversely, with their antigen shedding property and metastasizing capacity. Since subcellular contaminants tend to interfere with an accurate quantitation of tumour cell plasma membranes, we have directed our efforts towards developing a method whereby the plasma membranes of rat mammary tumour cells, free of contaminations, can be prepared reproducibly. This paper describes an improved method for isolation of plasma membranes and also confirms and extends our earlier