In this paper we describe a theory of DNA histogram debris generation and compensation that can be applied to paraffin-embedded frozen tissue preparations. The theory predicts the distribution of fragments generated from single and multiple random sectioning of threedimensional ellipsoids representing nuclei. The fragment distribution is assumed to be a major component of the underlying debris in DNA histograms. A comparison of S-phase fractions (SPF) from matched tissue prepared by frozen and formalin-fixed paraffin-embedded DNA methods demonstrates the usefulness of the theory.Key terms: DNA histogram analysis, cell cycle analysis, debris subtraction, debris compensation, S-phase analysis, paraffin processing Although over 100 retrospective clinical flow cytometry studies have examined DNA index and prognosis ( l l ) , less than 15% of these studies attempted to examine S-phase fraction (SPF). This low percentage is presumably due to investigators' lack of confidence in SPF estimates because of paraffin-embedded derived debris contamination in the S-phase region. Haag et al. (9) demonstrated that compensating for debris with either a n algebraic or exponential function yielded SPF estimates that were more comparable with 3H-thymidinelabeling indices. A few studies have compensated for debris with an exponential and have found SPF to be a more important prognostic factor than DNA index (8,2,3).In our own retrospective studies (Bagwell, unpublished results) we have found exponential debris compensation has a strong tendency to overcompensate for debris in the S-phase region in many samples resulting in abnormally low SPF estimates (see Fig. 1). The debris distribution for formalin-fixed paraffin-embedded tissue generally does not decrease as an exponential just prior to the GO-G1 peak. The mismatch in shape between the observed debris distribution and the exponential model component can result in poor computer regression fits and highly variable estimates of SPF.In this paper we present a theory that describes the distribution of debris resulting from random sectioning through nuclei. Although the presented theory is similar to that published by Bins et al. (7), the assumed geometry, calculated probability distributions, and method of compensation is fundamentally different. The shape of the debris curve presented in this paper is calculated from the observed histogram and is subsequently used as a model component in a Marquardt nonlinear least-squares algorithm ( 5 ) . We refer to this type of model component, with its shape derived from the histogram, as a histogram-dependent model component. The initial idea behind this technique was suggested to us by Dr. Dennis Way and Dr. Benjamin Love at the Annual Application Courses held in 1986 and 1987 and more recently a similar compensation algorithm has been implemented by Rabinovitch (13). We describe the theory in detail and test it with matched samples that have been prepared by frozen and paraffin-embedded DNA methods.
MATERIALS AND METHODS Sample Selection, ...