absorbance with several staining methods (Romen et al. 1980; Oud et al. 1986). The Feulgen technique (Oud 1986) is utilized for measuring the DNA content of ...
Histochemis
Histochemistry (1989) 91 : 47-50
9 Springer-Verlag 1'989t~
A double staining method for measuring DNA content of specific cell populations utilizing computerized microscopy E. Kerem, D. Schwartz-Arad, E. Bartfeld, N. Ron, I. Ariel, and G. Zajicek H.H. Humphrey Center for Experimental Medicine and Cancer Research and Department of Pathology, Hadassah University Hospital, Hebrew University-Hadassah Medical School Jerusalem, Israel Received March 29, 1988 / Accepted June 8, 1988
Summary. A method for identifying specific cell populations with the computerized microscope is described. This method utilizes double staining techniques using immunofluorescent antibodies and the Feulgen technique. It was applied to measure the area and D N A content of the nuclei of beta-cells in patients with Persistent Neonatal Hypoglycemia with Hyperinsulinism (PNHH). A significant increase in the area and D N A content was found in nuclei of patients compared to a control group. This method may be applied to sample cells from other tissues as well.
Introduction The technique of computerized microscopy can be applied in sampling cells to determine their nuclear size and optical absorbance with several staining methods (Romen et al. 1980; Oud et al. 1986). The Feulgen technique (Oud 1986) is utilized for measuring the D N A content of the nuclei. However, when all the cells or nuclei of the tissue appear histologically similar, it is not possible to identify specific cell populations. Therefore, when there is no confidence that all nuclei belong to the same cell population, the use of the computerized microscope is limited to measure D N A content of cells from homogeneous tissues or of cells with prominent appearance. To overcome this problem we wish to describe a method using the Feulgen technique and immunofluorescent antibodies, for selective cell population identification and computerized microscope scanning. We used this method to sample beta-cells from sections of the pancreas taken from patients with Persistent Neonatal Hypoglycemia and Hyperinsulinism ( P N H H ) (Jaffe et al. 1980, 1982; G o u l d et al. 1981 ; G o u d s w a a r d et al. 1986), and compared them to a control group.
Materials and methods Subtotal pancreatectomy specimens of five patients with PNHH were included in the current study. The control group consisted of five age-mached infants who died from a cause not associated with hypoglycemia, and who underwent autopsy. The pancreatic tissue was fixed in 10% buffered formaldehyde, processed, embedded in paraffin, and sectioned at 5 ~tm. The sections were stained Please send offprint requests to: Eitan Kerem, The Hospital for Sick Children, 555 University Avenue, Toronto, Ontario, Canada M5G 1X8
by the cold hydrolysis Feulgen technique (30 min 5 N HC1 at room temperature) (Oud 1986). These sections were not mounted, but were instead immersed in buffer and then stained by an indirect immunofluorescent technique using guinea pig antibodies to porcine insulin (diluted 1:20) and fluorescein conjugated thiocyanate rabbit anti-guinea pig IgG (diluted t :20) (Coons 1958; Johnson and Nogueira Araujo 1981). The sections were examined by an Olympus BH-2 light microscope, and scanned by a computer controlled T.V. camera (Hamamatsu - Japan). The light source consisted of a halogen lamp covered by a green filter (maximal transfer at 540 nM wave length). The tissue was magnified x 40 and digitalized in 512 x 512 pixel frame resolution. Ultimately one pixel was equivalent to 0.267 g2. Software packages which were developed in house were utilized. The first step was to depict the image on a color monitor, in pseudo-color display. The operator using a cursor allowed the computer to select a single cell. The image information from the nucleus was transferred to a VAX-11/780 where a boundary search was performed. After the operator selected a nucleus the computer program was able to find and display its border. Two main steps were taken : image pre-processing and boundary tracking. Pre-processing includes median filtering and edge enhancement (sobel operator) (Ballard and Brown 1982). In the edge enhanced image, pixels which lie on edge (nuclear boundary) get high values, while pixels in the interior and exterior of the nucleus get low values. The enhanced image was entered into a boundary finding routine. The boundary tracking search algorithm is based on location of an initial point which is the highest pixel in the filtered image, and whose distance from the origin (selected by the operator) approximates the assumed nuclear radius. Boundary tracking routine selects the next highest pixel, which is then chained to the previously selected pixel. This set of pixels eventually forms a closed connected circular line representing the nuclear boundary (Lester et al. 1978). Since nuclei of different cell types differ greatly in shape, no a-priori assumption was made about their expected shape (round, oval etc.). The only criterion used for selecting next edge pixel was high value in filtered (edge enhanced) image. After the nuclear boundary was found, the computer extracted features from each nucleus and performed statistical analysis (SPSSX 1986). The features were extracted from each nucleus: area, and total optical density. Total optical density was the integrated sum of pixel optical densities in a nucleus. Optical density (OD) was measured by dividing measured light intensity by blank light intensity (which was measured on an empty slide). The logarithm of the quotient was taken as net pixel OD. Total optical density which is proportional to DNA content (Bibbo et al. 1985), was taken as the sum of pixel optical densities in a nucleus. With our method of sampling, after calibrating the source light to a fixed value, ten fibroblasts from each section were sampled and used as an internal standard to correct stain differences between sections. Then the section was examined under fluorescent light with a dichroic mirror (B DM-500+0-515). Nesidioblasts
Fig. I a
Fig. i b
Fig. 2a
Fig. 2b Fig. 1. a Pancreatic section of P N N H patient stained by Feulgen technique, b The same section stained by the immunofluorescent technique. The fluorescent cells are nesidioblasts, x 100 Fig. 2. a Pancreatic section of P N N H patient stained by Feulgen technique, b The same section stained by the immunofluorescent technique. The fluorescent cells are beta cells in Islet of Langerhans. x 100
49 were identified as isolated fluorescent stained cells not associated with a pancreatic islet. After the identified cells were marked by cursor, the shutter was moved to cover the fluorescent light. The dichroic mirror was then removed and the nucleus was examined using the regular green light source as described. The closest neighboring duct cell was then sampled and utilized as an age-matched control for non-insulin producing cell. In each section 30 nesidioblasts and 30 duct cells were sampled. With the same technique 30 Langerhans Islet cells (LIC) were also examined.
Results
Figure 1 a shows the pancreatic section of a patient with P N H H stained by Feulgen technique. All the nuclei in the pancreatic section appear similar and it is not possible to identify the insulin producing cells. However, after introducing the immunofluorescent staining method to the same section, the betacells can be identified as Nesidioblasts (Fig. lb), or LICs (Fig. 2a and b). After sampling the cell by computerized microscopy each nucleus was extracted, and a boundary search was performed as described above. Table 1 summarizes the measured mean area and mean optical density of the nuclei stained by the Feulgen technique in the P N H H patients and control groups. The P N H H cells have statistically significant larger nuclear area and OD, compared to the control group, in all type of cells. This increase is more prominent in the insulin producing cells (Nesidioblasts and LIC). There is also increase in the size and variation between the cells in each group; in the control group, the cells with the largest nuclear area are the LICs, with a mean area of 31.7_+8 g2 compared to the nesidioblasts' mean area of 28.7 + 0.5 g2. This is a statistically significant difference with p < 0 . 0 0 1 . There is no significant difference between the area o f the nesidioblasts and the neighbouring duct cells. In the P N H H patients the LICs have the largest nuclear area with mean of 36.9 + 1.9 la2 compared to the nesidioblasts' area of 33.6_+0.5 gz (p
