Statistical analysis of labelling patterns of mammary carcinoma cell nuclei on histological sections

Abstract

It is of central interest for tumour biology to explore the mechanisms of tumour cell proliferation. In this study, methods of spatial statistics were used to study the spatial distribution of proliferating cells within tumour tissue quantitatively and objectively. Mammary cancer tissue was studied as an example. It was attempted to clarify whether cell division occurs entirely at random (random labelling), i.e. the process of division occurs at random, independently from the state of the neighbouring nuclei, or whether the spatial distribution of proliferation is more complex, e.g. in the form of actively proliferating clusters alternating with relatively silent zones. In the case of random labelling, the reduced second moment functions K(r) of the labelled and the unlabelled nuclei would be identical. The same would hold for the pair correlation functions g(r). The alternative hypothesis is that the second-order properties of the processes of the labelled and the unlabelled nuclei are different. Twenty cases of invasive ductal mammary carcinomas were studied. The nuclei of proliferating cells were stained immunohistochemically with the monoclonal antibody MIB-1, which detects specifically the proliferation-associated nuclear antigen Ki 67. The planar coordinates of the tumor cell nucleus profiles from two rectangular visual fields per case were recorded. For each visual field, the following investigations were performed: estimation of the explorative summary characteristics K(r) and g(r), fitting of the parameters of a stationary Strauss hard-core model to the observed point patterns, estimation of two distance-dependent Simpson indices and Monte Carlo tests of all individual patterns on the null hypothesis of random labelling. Significant differences between the mean K-functions and the mean g-functions of the labelled and the unlabelled nuclei were found. Moreover, the mean interaction parameter gamma of the stationary Strauss hard-core model was significantly higher for the labelled nuclei than for the unlabelled nuclei. The estimates of the two distance-dependent Simpson indices showed a tendency of points with the same label towards a positive spatial correlation. In the Monte Carlo tests, the null hypothesis of random labelling was rejected for the majority of the visual fields. These four lines of investigation led to the concordant conclusion that the labelling of mammary carcinoma nuclei by MIB-1 is not simply random. The data suggest that the second-order properties of the point process of the labelled nuclei are significantly different from those of the unlabelled nuclei. In particular, the process of the labelled nuclei shows a higher degree of clustering (increased strength of interaction) than the process of the unlabelled points.