Image analysis techniques. The problem of the quantitative evaluation of thechromatin ultrastructure.

The application of image analysis methods to conventional thin sections for electron microscopy to analyze the chromatin arrangement are quite limited. We developed a method which utilizes freeze-fractured samples; the results indicate that the method is suitable for identifying the changes in the chromatin arrangement which occur in physiological, experimental and pathological conditions. The modern era of image analysis begins in 1964, when pictures of the moon transmitted by Ranger 7 were processed by a computer. This processing improved the original picture by enhancing and restoring the image affected by various types of distorsion. These performances have been allowed by the third-generation of computers having the speed and the storage capabilities required for practical use of image processing algorithms. Each image can be converted into a two-dimensional light intensity function: f (x, y), where x and y are the spatial coordinates and f value is proportional to the gray level of the image at that point. The digital image is therefore a matrix whose elements are the pixels (picture elements). A typical digital image can be obtained with a quality comparable to monochrome TV, with a 512×512 pixel array with 64 gray levels. The magnetic disks of commercial minicomputers are thus capable of storing some tenths of images which can be elaborated by the image processor, converting the signal into digital form. In biological images, obtained by light microscopy, the digitation converts the chromatic differences into gray level intensities, thus allowing to define the contours of the cytoplasm, of the nucleus and of the nucleoli. The use of a quantitative staining method for the DNA, the Feulgen reaction, permits to evaluate the ratio between condensed chromatin (stained) and euchromatin (unstained). The digitized images obtained by transmission electron microscopy are rich in details at high resolution. However, the application of image analysis techniques to these images and especially to those referring to nuclei, is limited by several drawbacks: i) the thin section represents only a small fraction of the nuclear volume entirely visible in optical microscope specimens; ii) the identification of nucleosomes, of the solenoid fibres and of the higher levels of compaction of the heterochromatin is not thinsectioned specimens; iii) the differences between heterochromatin and euchromatin are based only on their grey level but do not reveal possible variations of their structural organization. Therefore, the applications of image analysis to the nuclear content does not utilzes the high resolution power of e.m. images and simply quantify the areas occupied by electron-dense chromatin with respect to the more electron-transparent ones. This result is less significative of those obtainable by optical microscopy, since the electron staining is not quantitative as the Fulgen reaction. On the other hand, the following problems still remain unresolved and should be clarified only by the use of quantitative image analysis: ultrastructural organization of the different types of heterochromatin (1); relationships between gene activation, transcription and chromatin decondensation; chromatin arrangement transformation induced by exogenous agents. In order to face these problems, in the last years we applied image analysis to cell or tissue specimens frozen in liquid nitrogen and then fractured in order to expose the inner content of the nucleus (Fig. 1). The obtained metal replicas represent very suitable specimens for digitalized image elaboration, since the fibers which give rise to the chromatin domains are exposed by the fracturing and evidentiated by the shadowing as black dots with a clear white shadow (Fig. 2). Therefore, their size and shape can be quantitatively evaluated by a digital image processor; in this vay the structural elements of the chromatin fibres are also detectable inside a fractured nucleus and their relative percentage ca be determined in each nuclear area (Fig. 3). This type of analysis has been initially used for characterizing in quantitative terms the organization of the nucleolar, heterochromatin and euchromatin areas in isolated nuclei (2) since the isolation procedure increases the differences among the nuclear domains. By using freeze-fractured isolated nuclei and conventional image analysis procedures, we quantitatively described the changes induced in the chromatin superstructure by the intercalating dye ethidium bromide (3), by the polyanionic phospholipid phosphatidylserine (4) and by the chemical carcinogen diethylnitrosamine (5). In all these cases, the principal affected parameter has been the ratio between the nucleosome and dolenoid percentage in different nuclear domains. The selection of a given class of particles, based on their size and/or shape, allowed to determine the spatial localization of some components such as the nuclear matrix, not easily detectable with conventional staining methods (6).