In the context of studying the nuclear architecture, the 3D location of chromosomal territories and their genes, especially the question of the localization of actively transcribed versus silent genes, were the subject of multiple investigations (Dietzel et al., 1999; Kosak et al., 2002; Mahy et al., 2002b). The majority of these studies were performed using human or mouse cells. It is generally assumed, that the nuclear architecture and the principles that govern gene regulation and transcription are conserved in all mammalians (Mahy et al., 2002a; Mahy et al., 2002b; Tanabe et al., 2002). It would therefore be desirable to be able to
visualize each individual territory in the interphase nucleus. With only 3 chromosome pairs in the female muntjac but about the same DNA content and number of genes as humans (Levy et al., 1993), M. muntjak seems to be ideally suited for a comparative analysis. However, the very limited number of mapped genes makes it a very difficult task. As part of this work, a number of important tools were generated to study M. muntjak. Forty-one genes, the localization of which were known in human or cattle, were mapped in M .
muntjak. In addition, whole chromosome paints for M. muntjak were generated, which
specifically stained individual metaphase chromosomes. However, centromeres were labeled by cross-hybridizations, most likely due to the presence of highly repetitive sequences or to the over amplification of repetitive sequences located at these regions during the DOP-PCR procedure (Cheung and Nelson, 1996). This problem has been observed before (Yang et al., 1995). This kind of crossreactivity of paints with centromeric regions has also been observed in studying human chromosomes and it could be blocked by adding Cot-1 fractionated DNA to the FISH-paint. Human Cot-1 DNA contains a lot of the repetitive sequences of the human genome (Strachan and Read, 1999). Unfortunately, during the course of the thesis, it was impossible to obtain enough genomic DNA from M. muntjak, to generate Cot-1 DNA. Cot-1 DNA produced using genomic DNA from cattle, roe deer, or even from Chinese muntjac, did not suppress the nonspecific signals sufficiently. Especially problematic was the constricted neck region on chromosome 3X which was described to contain repetitive sequences like interstitial telomeric DNA, satellite I (C5) and satellite II (MMV0.7) sequences (Lee et al., 1993; Li et al., 2000; Yang et al., 1997b). Whole chromosomal paints specific for each muntjac chromosome have been used before to study chromosomal evolution (Yang et al., 1995). However, to study the 3D localization of chromosomal territories in interphase nuclei, the level of crossreactivity would have to be substantially reduced. With adequate M. muntjak Cot-1 DNA, this problem could potentially be solved. The availability of specific whole chromosome paints for M. muntjak would allow the localization of genes within their chromosomal territories in the context of a 3D nucleus. In the following research, M. muntjak cells were used to analyze the localization of genes in the 3D nucleus.
6.3. 3D analysis of mammalian nuclei
The overwhelming majority of the published analyses of the spatial relationship of genes and chromosomal territories to each other and within the nucleus was done using the method of 2D FISH (Roix et al., 2003; Taslerova et al., 2003; Volpi et al., 2000). The general fixation method for 2D FISH involves a hypotonic treatment, methanol-acetic acid dehydrating fixation, and ultimately dropping of the fixed cell material on a slide resulting in flattened nuclei suitable for 2D analysis. This hypotonic treatment causes an enlargement of cells and their nuclei and the chromatin to be more loosely packed (Yokota et al., 1997; Yokota et al., 1995) making distance measurements between genes inaccurate. Another disadvantage of the 2D analysis is the possibility of misinterpretation of images. Genes that would be located on top of each other in a spherical nucleus (e.g. of a hematopoietic cell) can appear to be in close proximity. To compensate for these problems, Roix et al. analyzed several thousand cells using a semi-automated high-throughput image acquisition system to find significant differences in gene positions (Roix et al., 2003). Kozubek and coworkers developed a prediction model to interpret 2D images with the calculation of the theoretical distribution (Kozubek et al., 2001). However, these analyses can still only provide an approximation of the real 3D organization in a spherical nucleus.
In contrast to most studies, which performed the bulk of their analyses in 2D fixed cells with only a few 3D experiments to confirm the data (Mahy et al., 2002a; Volpi et al., 2000), almost all analyses in this thesis were performed on 3D fixed nuclei. To achieve a more accurate representation of a cell's 3D structure, 4% paraformaldehyde fixation was used in the present study. This established method (Kurz et al., 1996) with few a modifications to optimize analysis in different cell lines, was the method that gave the most reproducible and accurate results. It is the fixation method that has repeatedly been shown to conserve the chromatin architecture during FISH best (Kurz et al., 1996; Solovei et al., 2002). Performing all analyses of hematopoietic cells in 3D presented a number of challenges. For the nearly spherical morphology of these cell, an analytic method had to be found to interpret 3D confocal image stacks. This was accomplished with the help of the Amira 3.1 software and the computational tools developed by Juntao Gao. This form of
analysis allowed the accurate determination of the position genes in the 3D volume of interphase nuclei.
Most analyses expressed the location of genes (Bartova et al., 2002) or chromosomes (Cremer et al., 2003) as a percent of the nuclear radius. To describe positions of various genes in 3D fixed nuclei more accurately, I analyzed data as a percent of nuclear radius and in addition, used two different methods to assign genes to concentric shells. A similar system for the analysis of 2D images was published earlier (Bridger et al., 2000; Croft et al., 1999). The 2D area of the nucleus was divided into 5 concentric shells of equal area, and was used to describe positions of chromosome territories in 2D fixed nuclei. For the analyses in this thesis, the 3D volume of the nuclei was divided into 5 concentric shells, to describe the preferred location of a gene in 3D fixed nuclei. In one system (see Figure 18A) the nucleus was divided into 5 concentric shells of equal volume and in another (see Figure 18B) into 5 equidistant shells (with equal fractions of the nuclear radius). The results obtained with the method of rendering confocal image stacks followed by analysis of gene distances using the mathematical tools developed by Juntao Gao proved to be highly reproducible in determining the 3D position of genes in nuclei of hematopoietic cells. For instance, in all 10 tested cells and cell lines, the order of the localization of all genes and loci tested was preserved.
The improved accuracy of the methods used in this study came at the cost of the time consuming analysis involved to produce the IV-files required to evaluate the data. Proposed improvements for the future would therefore include various automated steps of the rendering process involved in generating "clean" IV-files. Even a small background surface located far outside the signal of interest to be rendered can affect its center of gravity calculation required for distance measurements. Furthermore, in many of the reports describing localization of chromosomes using 2D analysis the location of the chromosomes was expressed as the distance of its intensity weight center to the surface of the nucleus (Boyle et al., 2001; Croft et al., 1999; Taslerova et al., 2003). Calculation of the intensity weight center of rendered objects is not yet possible in 3D with the method developed in this thesis. They will have to be developed to make the method even more accurate.