Introduction Chromatin organization in the cell nucleus influences gene expression, DNA replication, damage, and repair. When the interphase nucleus forms, chromosomes partially decondense but still occupy distinct territories [ 1], which have nonrandom radial positions that are conserved through evolution [ 2– 5]. Current models suggest that chromosome territories (CTs) are separated by an interchromatin domain (ICD), rich in the nuclear machinery for nucleic acid metabolism. According to the ICD model, active genes are found in direct contact with the ICD [ 6], and occasionally fine chromosome fibers extend into this domain, where rare interchromosomal interactions may occur [ 1, 7– 9]. However, a physical separation between CTs is not supported by data on translocation frequencies and chromatin dynamics. Simulations of chromosome translocations based on models of chromosome organization have suggested the existence of a significant degree of intermingling between CTs [ 10– 12]. Furthermore, in vivo studies have shown that although chromatin domains are relatively stable [ 13], individual loci show diffusion dynamics constrained to approximately 0.4 μm [ 14– 16] and can exhibit movements as large as 1.5 μm [ 15]. This argues against a strict localization of chromatin within a CT that would prevent extensive intermingling. Recently, specific associations between loci on different chromosomes have been reported [ 17, 18], which are reminiscent of intrachromosomal clustering that is essential for correct gene expression [ 17, 19– 23]. It remains unclear whether these are just a few rare examples of interchromosomal associations that occur via chromatin fibers that extend from their own CTs or whether a greater potential exists for interactions through more extensive intermingling of chromosomes in interphase. Such interactions, if abundant, would be expected to determine chromosome organization and thereby influence the range of translocations that occur in each cell type. Previous data on chromosome morphology and organization have mainly originated from painting of whole chromosomes by fluorescence in situ hybridization (FISH) in three-dimensional (3D) nuclei. However, 3D-FISH is known to provide low spatial resolution and to compromise chromatin organization at the local level [ 24]. We have developed a novel FISH procedure for ultrathin cryosections (approximately 150 nm thick; cryo-FISH) of well-fixed [ 25], sucrose-embedded cells, that maximizes chromosome-painting efficiency, provides high resolution, and simultaneously preserves chromatin nanostructure. We show here that chromosomes intermingle significantly in interphase nuclei of human cells, arguing against the presence of an interchromosomal domain that separates CTs. The extent with which particular pairs of CTs intermingle correlates with the frequency of chromosome translocations in the same cell type [ 26]. Furthermore, we show that blocking of transcription changes the pattern of intermingling while preserving general chromosome properties, such as compaction and radial position, indicating that transcription-dependent associations between CTs are frequent enough to influence chromosome organization. In line with this view, we find that activation of the MHC class II gene cluster by interferon-gamma (IFN-γ) causes an increased colocalization of this locus with other chromosomes, concomitant with the relocation to a more external position in relation to its own CT [ 27]. Results/Discussion Chromosome Territories Intermingle Previous studies of chromosome organization during interphase have relied on the painting of chromosomes in whole nuclei, in conditions that compromise painting efficiency to preserve three-dimensionality. However, even in the best conditions, the nanostructure of chromatin at the level of single chromatin domains is lost [ 24]. To overcome this limitation, we developed a FISH procedure (cryo-FISH) using ultrathin cryosections of cells fixed under stringent conditions [ 25]. To test for chromosome intermingling we cohybridized pairs of whole chromosome paints to sections of phytohemagglutinin-activated human lymphocytes ( Figure 1). Binary masks were obtained for each CT and their intersections used to identify areas of colocalization ( Figure 1B– 1E). Fluorescence intensity profiles confirm that these areas contain DNA from two chromosomes ( Figure S1). Intermingling was detected for all chromosome pairs analyzed in these primary cells, but also in other human cell types (resting lymphocytes, HeLa cells, and primary fibroblasts; unpublished data). Due to the low resolution of the light microscope (LM; at best 200 nm in the x and y axes), we tested by electron microscopy (EM) whether DNA from different chromosomes is found in close proximity within areas of intermingling ( Figure 1G). After FISH, sections were first imaged on the LM to locate areas of intermingling ( Figure 1F), before indirectly immunolabeling the fluorochromes in the paints (FITC and rhodamine) with 5- and 10-nm gold particles, respectively ( Figure 1G). CTs labeled by immunogold particles strongly correlate with the corresponding LM image. Areas of intermingling identified by LM were found to contain colocalized gold particles labeling different chromosomes ( Figure 1G, inset, arrows; more than ten sections with intermingled CTs analyzed), showing that they are sufficiently close to interact at the molecular level. Stereoviews of regions of intermingling show that gold particles of different sizes are found at the same z-planes ( Figure 1H and 1I; eight regions of intermingling in four nuclear profiles analyzed), such that the intersection between CTs cannot be simply explained by distant territories that overlap within the thickness (approximately 150 nm) of the section. We next tested whether intermingling could result from artefactual chromatin disruption due to the harsh cryo-ISH procedure, in spite of the stringent fixation used. We compared the distribution of histone H2B, DNA, and sites of transcription labeled with Br-UTP, before and after ISH, and found that intermingling or the close proximity of gold particles labeling different chromosomes could not be explained by loss of fine chromatin structure during the procedure ( Figures 1J, 1K, and S2). Strikingly, the position of gold particles labeling histone H2B remains constant before and after FISH ( Figure 1J and 1K). To determine whether cryo-ISH had simply revealed the rare interactions between looped chromatin or showed more extensive intermingling of CTs, we measured how much of one CT intermingles with all others. We labeled sections with a Chromosome 3 paint, together with a probe that hybridizes with all other chromosomes ( Figure S3), and found that 41% of the volume of Chromosome 3 intermingles with the remaining genome. Although this argues against the existence of an interchromatin space that separates CTs, it remained possible that intermingling involved only loops of less condensed chromatin [ 1, 7, 8]. Therefore, we asked whether chromatin concentration within areas of intermingling is lower to that within a CT. We compared the fluorescence intensity of general DNA dyes (DAPI or TOTO-3 after RNase treatment) in intermingled regions of a CT with nonintermingled regions, or with the whole nucleoplasm, and found no significant differences (ratios of 1.12 ± 0.20 and 1.02 ± 0.52, respectively; n = 32). This shows that similar average DNA concentrations are present in intermingled and nonintermingled regions, indicating that chromatin has similar average properties in both areas. In fact, we also observed mixing of chromatin fibers within a CT (“intramingling”) between both arms of a chromosome (approximately 10% of Chromosome 3 volume; see also [ 28]). Therefore, regions of higher accessibility to transcription and pre-mRNA processing factors do not preferentially locate between CTs but are more uniformly distributed throughout the nucleoplasm, as shown previously [ 29, 30]. Different Extents of Chromosome Intermingling between CTs Correlate with Translocation Frequencies Chromosome intermingling has been suggested by modeling of translocation frequencies [ 10– 12], but not previously visualized, except for rare interactions [ 9]. A prediction of such models is that the extent of intermingling between each pair of chromosomes should be reflected in their translocation potential. We therefore measured the intermingling volumes for 24 pairs of chromosomes in activated human lymphocytes using a simple stereological principle (see Materials and Methods). Chromosome pairs were selected to reflect a wide range of translocation frequencies as measured in the same cell type by Arsuaga et al. [ 26] ( Table S1). The fraction of one chromosome (both homologs) that intermingles with any of the other 22 chromosomes is, on average, 2.1 ± 1.1%. This would correspond to 46% of each chromosome being intermingled with the rest of the genome (2.1% × 22 chromosomes), which is in agreement with the experimental value of 41% obtained for Chromosome 3. To obtain absolute values that are independent of CT volume, we expressed intermingling as a percentage of the nuclear volume ( Figure 2A). These values are representative of the average across the cell population, thus taking into account the frequency of CT association and the extent of intermingling when they are associated. Intermingling volumes between individual chromosome pairs vary by 20-fold ( Figure 2A), and show statistically significant differences ( p 3 h), rinsed in PBS, fixed in 0.5% glutaraldehyde in PBS (10 min), washed in distilled water, and incubated in 2% methylcellulose (10 min). Excess liquid was blotted and grids were left to dry. For the biotin-labeled paint, FITC-conjugated streptavidin (1/500; Sigma) or AlexaFluor350 conjugated Neutravidin (1/100; Molecular Probes) were used. The biotin-labeled BAC probe for the MHC II locus was detected using rhodamine-conjugated neutravidin (1/500; Molecular Probes), followed by a biotin-conjugated goat anti-avidin antibody (1/500; Vector) and rhodamine-conjugated neutravidin. PML was detected with anti-PML rabbit IgG clone H238 (1/10; Santa Cruz Biotechnology), followed by an AlexaFluor 488–conjugated goat anti-mouse antibody (1/1,000; Molecular Probes). Histone H2B was detected with a rabbit anti-histone H2B polyclonal antibody (1/100; Chemicon), followed by a goat anti-rabbit antibody conjugated with 5-nm gold particles (1/50; British BioCell). Serine2-phosphorylated PolII was indirectly immunolabeled with H5 (1/1,000; Covance, Berkeley, California, United States). After immunolabeling and washing (3×) in PBS, antibodies were fixed (1 h) with 8% paraformaldehyde in 250 mM HEPES (pH 7.6), before mock-ISH or chromosome painting. Microscopy For confocal laser scanning microscopy, images were collected sequentially on a Leica TCS SP2 (×100 PL APO 1.40 oil objective) equipped with argon (488 nm) and HeNe (543 nm; 633 nm) lasers or a Leica TCS SP1 (×100 PL APO 1.35 oil objective) equipped with UV (351/364 nm), argon (488 nm), krypton (568 nm), and HeNe (633 nm) lasers. For wide-field LM, images were collected sequentially on a Delta-Vision Spectris system (Applied Precision, Issaquah, Washington, United States) equipped with an Olympus IX70 wide-field microscope (×100 UPlanFl 1.3 oil objective), a charge-coupled device camera, and the following filters: DAPI, FITC, RD-TR-PE, CY-5, CFP, YFP. No bleed-through was detected in these conditions. The use of ultrathin cryosections allows for the use of wide-field microscopy with no reduction in axial (z) resolution and only a small reduction in lateral resolution [ 46]. For EM, images were collected on a JEOL 1011 transmission electron microscope (JEOL UK, Welwyn Garden City, Herts, United Kingdom) equipped with a cooled slow-scan KeenView charge-coupled device camera (1,392 × 1,024 pixels; Soft Imaging System, Münster, Germany). Image analysis and measurements For LM experiments, images (TIFF files) were automatically merged using a MatLab script (kindly provided by Tiago Branco, University College London, London, United Kingdom), saved as new TIFF files, and manually thresholded in Adobe Photoshop (Adobe Systems, Edinburgh, United Kingdom) to define masks for nuclei or CTs. Threshold values were chosen empirically so that the entire CT was selected but no widespread nuclear background was included. Independent drawing of masks by four different people on 10 images were compared to test the reliability of this empirical method. Variability in CT volume was found to be 15%, and in intermingling volumes was 30%, in the same order of magnitude as the variability obtained across independent experiments. The values of the areas of these masks and the intersection between the masks for both CTs were extracted using another MatLab script (Tiago Branco). CT and intermingling volumes were calculated according to stereological methods [ 47] after collecting random images of sections irrespective of their area and whether they contained CT signals (i.e., sections analyzed represented the whole nucleus). CT or intermingling areas were averaged across all sections and divided by the average of the nuclear areas. This ratio (R) is equivalent to the ratio of the respective average volumes, as shown here: where A ROI is the average CT or intermingling area, A NUC is the average nuclear area, V ROI and V NUC are the corresponding average volumes, and t is the section thickness. Using average section volumes for R gives the same result as using average whole nuclei volumes if enough random sections from different cells are included in the calculation. To obtain several values for R within one hybridization experiment (to allow statistical analysis), images were randomly grouped and R was calculated for each group. Standard deviations remained constant with increasing number of groups until a group size was reached at which R did not contain enough information and the standard deviation increased abruptly. The highest number of groups before this increase was used. Group size varied between different chromosome pairs, averaging 55 sections per group, and up to four groups were used in an experiment (a total of 57 to 211 sections were analyzed in individual experiments). Standard deviations obtained by this method were consistent with standard deviations between independent hybridization experiments. The R values were used for statistical tests, and considered to have a normal distribution, as normality plots for the analysis of residuals were positive. Two-sample comparisons were performed by two-tailed t-test and multisample comparisons by ANOVA. Regression analyses using an F-test were performed to test the significance of variable correlations. For the analysis of the MHC II locus data, we used Fisher's exact test for 2 × 2 contingency tables and chi-squared test for larger tables. Supporting Information Figure S1 Fluorescence Intensity Profiles of the Images in Figure 1B– 1E (469 KB JPG) Click here for additional data file. Figure S2 Additional Control Experiments Showing Preservation of Nuclear Structure during cryo-FISH (954 KB JPG) Click here for additional data file. Figure S3 Quantification of the Total Intermingling of Chromosome 3 with the Remaining Genome (214 KB JPG) Click here for additional data file. Figure S4 Graphical Analyses of the Distribution of Active RNA Polymerase II Sites within CTs and Areas of Intermingling (13 KB PDF) Click here for additional data file. Figure S5 Volumes for All CTs in Human Female Lymphocytes and Correlation with DNA Content (12 KB PDF) Click here for additional data file. Protocol S1 Quantification of Nuclear Volume That Contains Intermingled CTs (24 KB DOC) Click here for additional data file. Table S1 Interchange Yields in Phytohemagglutinin-Activated Human Lymphocytes (47 KB PDF) Click here for additional data file.