Introduction

The interphase nucleus has emerged as a dynamic organelle with rapid movements of many nuclear proteins and RNA molecules. In contrast, interphase chromatin appears relatively immobile (Abney et al. 1997; Bornfleth et al. 1999; Manders et al. 1999; Dundr and Misteli 2001; Gerlich et al. 2003; Walter et al. 2003). Nevertheless, some chromosome subregions demonstrate more extensive mobility. Giant chromatin loops containing gene clusters have been reported to extend outwards from the surface of corresponding chromosome territories upon gene(s) activation (Volpi et al. 2000; Mahy et al. 2002; Williams et al. 2002). In mouse lymphocytes, dynamic repositioning of genes towards the chromocenter clusters or away from them has been described in association with gene silencing and activation (Brown et al. 1997, 1999). Extensive rapid movements of certain chromosome loci have also been reported for nuclei of budding yeast and of Drosophila embryos (Csink and Henikoff 1998; Gasser 2002; Marshall 2002).

Centromeres provide another example where major movements have been observed in nuclei of both postmitotic cells (Manuelidis 1985; Alcobia et al. 2000; Martou and De Boni 2000; Solovei et al. 2004a) and cycling cells (Bartholdi 1991; Ferguson and Ward 1992; Weimer et al. 1992; Vourc’h et al. 1993; Hulspas et al. 1994). Previous studies of the distribution of centromeres/kinetochores in cycling and non-cycling cell types had important technical limitations, including the unequivocal discrimination of non-cycling cells (G0) from cells at different stages and substages of interphase (see Discussion). As already emphasized by Ferguson and Ward (1992) “the morphological preservation of specimens through the various techniques of isolation, fixation, and hybridization is of the utmost concern.” Preservation of the three-dimensional (3D) structure of cells and their nuclei during fixation is a prerequisite for reliable morphological studies using immunostaining and/or fluorescence in situ hybridization (FISH) techniques in combination with confocal microscopy (Bridger and Lichter 1999; Solovei et al. 2002a). In FISH studies, the detrimental effects of the heat denaturation step have to be taken into account (Solovei et al. 2002a,b). While this step apparently does not lead to a major distortion of higher order chromatin arrangements, it seems advantageous, whenever possible, to use protocols that avoid this step.

Considering these limitations we decided to perform studies of 3D centromere arrangements in five human cell types (normal diploid fibroblasts, normal lymphocytes and diploid lymphoblastoid cells, normal peripheral blood monocytes and neuroblastoma SH-EP cells) employing protocols for centromere/kinetochore visualization that keep the 3D nuclear architecture intact to the best possible extent. Precise identification of the quiescent or cycling stage of each individual cell was based on the nuclear patterns of pKi67 (Kill 1996; Starborg et al. 1996; Bridger et al. 1998) and replication labeling with 5-bromo-2′-deoxyuridine (BrdU) (O’Keefe et al. 1992; van Driel et al. 1998). Following fixation of cells that preserved the 3D structure, centromeres were visualized either by kinetochore immunostaining or by 3D-FISH with a pancentromeric probe. The 3D centromere/kinetochore arrangements were assessed in light optical image stacks from nuclei recorded by a confocal microscope.

Materials and methods

Cell types and 3D fixation

Human lymphocytes (G0) and monocytes from peripheral blood were isolated in a Ficoll gradient, and incubated for 3 days in RPMI 1640 supplemented with 20% FCS and 1% phytohemagglutinin (PHA). A diploid lymphoblastoid cell line was established in our laboratory from a healthy male donor; these cells were grown in RPMI 1640 supplemented with 15% FCS. For 3D-preserving fixation, lymphoblastoid cells and lymphocytes were harvested and resuspended in fresh medium with 50% FCS at a final concentration of ca. 1×106 cells/ml. Three hundred microliter aliquots of the suspension were placed on coverslips coated with polylysine (1 mg/ml). Cells were allowed to attach for about 1 h at 37°C, and then fixed in 4% paraformaldehyde in 0.3×PBS. To prevent shrinkage of spherical lymphocyte and lymphoblastoid nuclei, cells were briefly (1 min) incubated in 0.3×PBS before fixation. Human skin fibroblasts were grown directly on coverslips in DMEM supplemented with 10% FCS. Neuroblastoma SH-EP cells [47, XX, der (1;14) t (1;14)/der (7) t (7;8)/der (8) t (7;8)/der (22) t (17;22)] were kindly donated by Prof. W.W. Franke, DKFZ, Heidelberg) and subcultured on coverslips in RPMI supplemented with 10% FCS. The two adherently growing cell types were fixed in 4% paraformaldehyde in 1×PBS. In all cases, special coverslips (26×76 mm) with an even thickness of 0.17±0.01 mm (Assistent, Germany) were used for cell attachment and growth. These coverslips are recommended for optimized confocal microscopy measurements. After fixation, the cell membrane and the nuclear envelope were permeabilized by incubation in 0.5% Triton X-100. The 3D fixed and permeabilized cells were stored in PBS with 0.04% sodium azide at 4°C until use.

Synchronization of neuroblastoma SH-EP cells

SH-EP cells were subcultured on coverslips in full medium (RPMI 1640, 10% FCS) until 40–50% confluency. Effective synchronization of SH-EP cells was obtained with the following protocol. Coverslips with subconfluent SH-EP cells were transferred to serum-deprived medium (RPMI 1640 without FCS) for 14–24 h. Thereafter cells were transferred to full medium supplemented with aphidicolin (1 μg/ml) to block DNA polymerase (Ikegami et al. 1978). After 12 h cells were released from the block by washing three times and transferred to medium with 10% FCS. The degree of synchronization and duration of the cell cycle stages were determined using BrdU incorporation and pKi67 staining (see below). About 95% of cells were in early S phase in 30–60 min after release from the block; in mid-S phase after 4.5 h; in late S phase after 6.5 h; in G2 after 7 h; most of the cells entered mitosis 8–9 h after release. The majority of the cells reached the G1 stage of the next cell cycle 10–11 h after release. The degree of synchronization in the next cell cycle was still high: about 85% of the cells could be simultaneously labeled by BrdU in the next early S.

Identification of cell cycle stages and kinetochore staining

Replication labeling with BrdU was used to identify S phase cells and to distinguish between cells in early S, mid-S, and late S phase (O’Keefe et al. 1992). For pulse labeling, cells were incubated in culture medium with 10 μg/ml BrdU (Sigma) for 30 min. To avoid morphological alterations caused by standard methods of BrdU detection due to the DNA denaturation step (van Driel et al. 1998), detection of BrdU in fixed cells was performed following the 3D-preserving protocol described by Tashiro et al. (2000). Cells were incubated with mouse anti-BrdU antibodies (Roche) in a solution composed of 0.5% BSA, 0.5×PBS, 30 mM TRIS, 0.3 mM MgCl2, 0.5 mM 2-mercaptoethanol, and 10 μg/ml DNase I (Roche); then sheep anti-mouse Cy3-conjugated (Jackson ImmunoResearch Laboratories) secondary antibody was applied. Cells at G0 and early G1 stages were identified by staining with antibodies against protein Ki67 (Dianova) as described by Bridger et al. (1998). For kinetochore staining, rabbit anti-CENP-B and rabbit anti-CENP-C antibodies (kindly donated by Prof. W.C. Earnshaw, University of Edinburgh, UK) were used either separately or as a 1:1 mixture with identical results. Alexa 488-conjugated goat anti-rabbit antibody (Molecular Probes) was applied as the secondary antibody. All antibodies (except anti-BrdU) were diluted in blocking solution containing 1×PBS, 0.05% Triton X-100, and 3% BSA; all washings were performed in 1×PBS with 0.05% Tween at 37°C. Nuclear DNA was counterstained with TO-PRO-3 (Molecular Probes) and 4’,6-diamidino-2-phenylindole (DAPI) (Sigma). Cells were mounted in Vectashield (Vector Laboratories) antifade medium.

Three-dimensional fluorescence in situ hybridization with a pancentromeric probe

Normal human diploid fibroblasts were fixed and prepared for three-dimensional fluorescence in situ hybridization (3D-FISH) according to standard protocols (Solovei et al. 2002a). Briefly, cells were fixed in 4% paraformaldehyde in 1×PBS, permeabilized with 0.5% Triton X-100, incubated in 20% glycerol, repeatedly frozen-thawed using liquid nitrogen, incubated in 0.1 N HCl for 5 min, and stored in 50% formamide, 2×SSC at 4°C until hybridization. A probe for α-satellite sequences contained in centromere regions of all human chromosomes (Choo 1997), referred to as a pancentromeric probe, was generated and labeled by the polymerase chain reaction using 5′-CAT CAC AAA GAA GTT TCT GAG GCT TC and 5′-TGC ATT CAA CTC ACA GAG TTG AAC CTT CC primers and human placenta DNA as a template. Following labeling with fluorescein isothiocyanate-12-dUTP (Roche Molecular Biochemicals) the pancentromeric probe was shortened to 100–300 bp by DNase digestion. The probe was dissolved in hybridization mixture (50% formamide, 10% dextran sulfate, 1×SSC), loaded on a coverslip with cells, covered with a smaller coverslip, and sealed with rubber cement. Cell and probe DNA were denatured simultaneously on a hot-block at 75°C for 2 min. Hybridization was performed for 2 days at 37°C in humid boxes. Post-hybridization washes were done in 2×SSC at 37°C and 0.1×SSC at 60°C. After BrdU and/or pKi67 detection, nuclear DNA was counterstained with TO-PRO-3 and DAPI, and cells were mounted in Vectashield. To preserve 3D nuclear morphology, air-drying of cells was carefully avoided at all steps from fixation to mounting in the antifade (Solovei et al. 2002a, 2002b).

Confocal microscopy and image processing

Series of confocal sections through whole nuclei were collected using a Leica TCS SP equipped with Plan Apo 100×/1.4 NA oil immersion objectives. For each optical section, images were collected sequentially for two or three fluorochromes. Fluorochromes were excited using an argon laser with excitation wavelengths of 488 nm (for Alexa 488) and 513 nm (for Cy3), or a helium–neon laser with excitation wavelength of 633 nm (for TO-PRO-3). Stacks of 8-bit gray-scale images were recorded with an axial distance between optical sections of 300 nm and a pixel size of 50 nm. Galleries of RGB confocal images were assembled using NIH Image and Adobe Photoshop software packages. Three-dimensional reconstructions of nuclei and kinetochore or centromere signals were performed by volume and surface rendering of image stacks using Amira 2.3 TGS software (http://www.amiravis.com).

Quantitative assessment of the 3D positioning and clustering of kinetochore signals

Positions of kinetochore signals within the nucleus were classified as shown in Fig. 1a. Kinetochore signals that abutted the nuclear border or were separated from it by a distance not exceeding the diameter of the kinetochore signal were classified as peripheral; all other signals were scored as internal. For spherical nuclei of lymphocytes and lymphoblastoid cells, RGB galleries of serial XY-optical sections were used for visual tracing and scoring of kinetochore signals. Owing to the limited Z resolution of the confocal microscope, this approach was not applicable to the flat nuclei of fibroblasts and SH-EP cells. For scoring kinetochore signals in these cells, image stacks were loaded into the Amira 2.3 program and data were viewed as ZY and/or ZX optical sections (Fig. 4a). In late G2, most kinetochore signals are duplicated and appear as doublets. However, since not all kinetochores are seen as doublets and not all close pairs of signals are true doublets, we counted all visible kinetochore signals separately.

Fig. 1a, b
figure 1

Scheme for scoring the centromere signals. a Spherical nuclei of lymphocytes and lymphoblastoid cells and nuclei of monocytes. b Flat nuclei of fibroblasts and SH-EP cells. The position of a kinetochore signal in the nucleus was considered as peripheral when it was touching the border of the nucleus defined by the counterstain (1) or was separated from the border by a distance not exceeding the size of the signal itself (2). Other signals, including those adjacent to the nucleoli (n), were scored as internal (3, 4). Kinetochore signals on the surfaces of the flat nuclei facing the substrate and free surface of the cell were classified as bottom (1b) and top (1t), respectively; signals adjacent to the lateral surface of the nucleus were classified as side signals (1s)

Full size image

To analyse the distribution of kinetochores with regard to different surfaces of the nucleus, peripheral signals were subclassified as top, bottom, and lateral (Fig. 1b). Each cell could therefore be characterized by proportions of top (P t) and bottom (P b) signals to total number of signals in this cell. Since scatter diagrams with P b and P t axes indicated the presence of two groups of cells in some stages of the cell cycle, relative numbers of top and bottom signals in individual nuclei were analyzed in more detail. Regression equations for P t and P b were calculated and regression lines were used as new x-axes with zero in the point P b=P t (R-axes). Positive and negative values on the R-axis represent nuclei with kinetochores preferentially located at the top and bottom side, respectively (see Fig. 6).

In addition to visual tracing and scoring of kinetochore signals in spherical nuclei of lymphocytes, a special three-dimensional relative radius distribution (3D-RRD) computer program was used (see Cremer et al. 2001 for detailed description of the program). This program finds borders of nuclei based on the DNA counterstain and calculates the gravity centers of the nuclei. Then the length of the nuclear radius in any direction is assigned to 100% and the nuclear space is divided into 25 shells of equal relative thickness (4%). In this way the average distribution of kinetochore signals can be presented for a set of nuclei as a function of the their average relative distances from the center of a normalized nucleus. Signals distributed randomly in the nuclear DNA should have the same distribution as the DNA counterstain; deviations from the counterstain curve would indicate a non-random distribution.

Results

Cell cycle stage identification

The definition of cell cycle stages for individual cells was based on three criteria: the distribution of BrdU replication label (O’Keefe et al. 1992), the pattern of pKi67 immunostaining (Bridger et al. 1998), and the morphology of the kinetochore signal (Table 1). Figure 2 presents examples of lymphocyte nuclei with typical BrdU or pKi67 staining patterns used for this purpose (see also Solovei et al. 2004b). The same patterns were also observed in the other three cell types included in our study. In early G1, pKi67 is accumulated in granules of various sizes distributed throughout the nuclear volume, but mainly associated with heterochromatin regions (Fig. 2b). In late G1, S, and G2, pKi67 is localized in the nucleoli (Fig. 2c). Nuclei of postmitotic and quiescent (G0) cells lack a pKi67 signal (Fig. 2a). Nuclear incorporation of BrdU marks S phase cells. Early S phase is characterized by replication of chromatin in most of the nuclear volume except for chromatin domains surrounding the nucleoli and adjacent to the nuclear periphery (Fig. 2d). In mid-S phase, replication is mostly restricted to the domains located around the nucleolus and at the nuclear periphery (Fig. 2e). In late S phase, a few large chromatin clusters are labeled with both peripheral and internal nuclear locations (Fig. 2f). The morphology of kinetochore signals, as observed after immunostaining with antibodies against CENP-B/CENP-C, helped to distinguish between late G1 (single, small signals) and late G2 (larger signals, some still appearing as single, others as doublets) (Table 1).

Fig. 2a–f
figure 2

Simultaneous detection of kinetochores and cell cycle stage in nuclei of normal lymphocytes from peripheral blood stimulated with phytohemagglutinin (PHA). a Four representative confocal sections (1–4) through the nucleus of a non-stimulated lymphocyte. Note the large size of the kinetochore signals (green) produced by clusters of kinetochores on the periphery of the nucleus. The TO-PRO-3 counterstain is shown in blue. b–f Mid-sections through nuclei of stimulated lymphocytes at different cell cycle stages: early G1 (b), G2 (c), early S (d), mid-S (e), late S (f). Overlay of kinetochore immunostaining (green), immunostaining with pKi67 (b, c) or 5-bromo-2′-deoxyuridine (BrdU) (d–f) (red), counterstaining with TO-PRO-3 (blue), and corresponding gray scale images are shown for each stage. Bar represents 5 μm

Full size image
Table 1 Cell cycle identification based on intranuclear localization of incorporated 5-bromo-2′-deoxyuridine (BrdU), pKi67 pattern, and kinetochore signal morphology
Full size table

Clustering of kinetochores varies in cycling and non-cycling cells

In all cell types, the number of kinetochore signals per cell changed during interphase and upon exit from the cell cycle (Fig. 3). In G1 cells, the number of signals was always close to the expected 46 for diploid cells and 47 for SH-EP cells, i.e., most of the kinetochores were separate. The number of signals in S phase cells was lower than in G1, indicating kinetochore clustering. The difference between G1 and S phase was more pronounced in fibroblasts and stimulated lymphocytes, than in SH-EP and lymphoblastoid cells. In late G2, most of the kinetochores were represented by doublets. Each doublet was counted as two signals, explaining the high number of signals scored at this cycle stage. The number of signals observed in G0 nuclei (Fig. 3) was generally smaller than in nuclei from cycling cells and most G0 signals were considerably larger than signals observed in cycling cells (Fig. 2a). This finding indicates a strong clustering of kinetochores upon exit from the cell cycle.

Fig. 3
figure 3

Changes in the mean number of kinetochore signals per nucleus (numbers above the histogram columns) with cell cycle stage in the nuclei of fibroblasts, SH-EP neuroblastoma cells, lymphocytes, and lymphoblastoid cells. Numbers in parentheses are the numbers of nuclei studied

Full size image

Kinetochore distribution and dynamics in flat nuclei of adherently growing cells

Nuclei of normal diploid fibroblasts and SH-EP cells were relatively flat. Their dimensions changed to some degree depending on the cell cycle stage, but their ellipsoidal (fibroblast nuclei) or oval shape (SH-EP cell nuclei) persisted (Figs. 4a–c, 8a,d,g). The average X, Y, Z diameters for nuclei of fibroblasts in early S were about 17×12×4 μm. The same parameters for nuclei of early S phase SH-EP were 15×12×5 μm. For both these adherently growing cell types we studied six cell cycle stages (G1 early, G1 late, S early, S mid, S late and mid-late G2). To locate the positions of kinetochore signals, 3D reconstructed nuclei were virtually sectioned in the XY, ZY and ZX directions (for example see Fig. 4a). Each nucleus was screened section by section in at least two (ZX, ZY) directions and signals were classified as peripheral or internal according to the scheme shown in Fig. 1a. Kinetochore signals that abutted the nuclear border or were separated from it by a distance not exceeding the kinetochore signal size in the xy-plane (or half the kinetochore size in the z-direction, because, due to the low resolution of the confocal microscope in the z-direction, the observed z-size of the signals was about three times greater than the size in the xy-plane), were classified as peripheral, all other signals were scored as internal (Fig. 1).

Fig. 4a–g
figure 4

Evaluation of the spatial distribution of kinetochore signals using Amira software (a–f) and a three-dimensional relative radius distribution (3D-RRD) computer program (g). a SH-EP cell nucleus in early S phase: XY, XZ, and YZ maximum intensity projections combined with an XZ ortho-section. Scrolling through XZ or YZ ortho-sections allowed counting of the peripheral and internal kinetochore signals. b–f Three-dimensional reconstructions (by volume rendering) combined with XZ ortho-section through nuclei of fibroblast (b) and SH-EP cell in early S (c), lymphocyte in G0 (d), lymphoblastoid cell in G0 (e), and postmitotic monocyte (f). Note predominantly peripheral location of the kinetochore signals (green) in counterstained (red) nuclei. Bars represent 5 μm. g Quantitative 3D evaluation of radial kinetochore distribution in nuclei of lymphocytes in G0 and at different cell cycle stages. n number of evaluated nuclei. Error bars indicate standard deviation of the mean for each shell

Full size image

The height of nuclei (size along the z-axis) is, depending on cell cycle stage, 3–5 μm in fibroblasts and 5–7 μm in SH-EP cells. The observed z-size of kinetochore signals was about 0.6 μm. Random distribution of kinetochores along the z-axis means in this case an even distribution. Using the counting rule described above, in the case of random distribution only 12–20% of kinetochore signals would be classified as peripheral in fibroblasts and 10–15% in SH-EP cells. In actuality at least 65% of kinetochores have peripheral localization at any stage in both cell types (Fig. 5). Hence, the distribution is markedly non-random.

Fig. 5
figure 5

Changes in the percentage of peripheral (light gray) and internal kinetochore signals (dark gray) depending on the cell cycle stage in nuclei of fibroblasts, neuroblastoma SH-EP cells, lymphocytes, and lymphoblastoid cells. Numbers in parentheses are the numbers of nuclei studied

Full size image

Figure 5 provides the percentages of internal and peripheral kinetochore signals in fibroblast and SH-EP cell nuclei. Both cell types showed the same trend in kinetochore distribution: at all interphase stages most kinetochores were situated peripherally. The proportions of peripheral kinetochores, however, increased from early G1 to early S phase but decreased in late S phase and G2. G0 fibroblasts contained very few internal kinetochores (SH-EP cells were not studied at this stage). Most of the internal kinetochores observed in S phase and G0 cells abutted the nucleolus, which was identified either by its more intense TO-PRO-3 staining (note: this fluorochrome stains RNA as well as DNA) compared with the surrounding chromatin or by staining with anti-pKi67 antibodies (see above). In early G1 and late G2 cells, a significant proportion of the internal kinetochores were not associated with the nucleolus.

Peripheral signals were additionally classified with respect to their location close to the nuclear envelope at the bottom of the nucleus (the side facing the glass slide on which cells were grown) or top (the other side of the nucleus). Signals located at the nuclear edge were scored as lateral signals (Fig. 1b). The latter were rather few and excluded from further consideration. With regard to top and bottom signals, the question arises as to whether kinetochores are evenly distributed between the top and bottom sides of the nucleus or not. The first aspect of this problem is comparison of average proportions of kinetochore signals at the top and bottom sides of the nucleus at different stages of the cell cycle. In fibroblast nuclei at G1, S and G2, the mean proportion of kinetochores on the bottom side was higher than on the top side, and all differences were significant (0.43/0.29, p<0.001 for G1; 0.45/0.33, p<0.006 for S; 0.42/0.31, p<0.007 for G2). In G0 fibroblasts, the difference was smaller, but still significant (0.46/0.39, p<0.005). Surprisingly, in SH-EP cells the proportions of kinetochores at the top and bottom sides of the nucleus were equal (0.31/0.32 for G1; 0.33/0.35 for S; 0.36/0.40 for G2).

Another aspect of this problem is the distribution of kinetochores in individual nuclei. Images and kinetochore counts clearly showed that in some nuclei the majority of kinetochores were located on the bottom side, while in other nuclei most kinetochores were gathered at the top side. To address this matter, we made scatter diagrams as shown in Figs. 6a, 7a,b,c. Each dot on these diagrams represents an individual nucleus. The coordinates of a dot show the proportions of kinetochores situated at top (abscissa) and bottom (ordinate) sides of the nucleus (the sum of abscissa and ordinate is less than 1, because there are some internal and a few lateral kinetochores). If kinetochores had been distributed evenly between the top and bottom sides, the dots would have formed one cluster on these scatter diagrams. However, at least in some cases two clear clusters with more kinetochores at the top or bottom sides, respectively, were observed (Fig. 7a,b). Simple histograms do not allow these two clusters to be demonstrated because their projections on the x-axis overlap (Fig. 6a,b). To cope with this difficulty, regression equations for proportions of kinetochores at the top and bottom sides (P t, P b) were calculated and used as a new x-axis (R-axis). The zero of the R-axis was always set to the point P t=P b. This allowed minimization of overlapping of clusters, and observation of them as two maxima of a bimodal distribution (Fig. 6c). Positive and negative values show predominant localization of kinetochore signals in a nucleus at the top and bottom sides, respectively. For example, a nucleus with 59% kinetochore signals adjacent to the bottom surface and 11% of kinetochore signals adjacent to the top surface (Fig. 6a, arrow) has a value of −0.33 on the histogram on Fig. 6c.

Fig. 6a–c
figure 6

Approach to the analysis of distribution of kinetochores in individual nuclei. a A scatter diagram showing individual nuclei as dots with coordinates corresponding to proportions of signals at the top (P t) and bottom (P b) sides of the nucleus (SH-EP cells, S phase). Note that dots form two clusters. b A histogram showing number of nuclei (n) depending on P t. Such straightforward histograms do not allow demonstratation of the presence of two clusters because their projection on the abscissa overlaps. c A regression line is used as a new abscissa (R-axis); zero is set to the point P t=P b. Nuclei with prevalence of kinetochores on the bottom side (e.g. the one marked by an arrow) have negative values on the R-axis and vice versa. The histogram on c is clearly bimodal: nuclei with similar proportions of kinetochores at the top and bottom sides are least common

Full size image
Fig. 7a–r
figure 7

Relative numbers of kinetochores at the top and bottom surfaces of individual nuclei in fibroblasts and SH-EP cells at different stages of the cell cycle. a–c Scatter diagrams representing individual nuclei as dots with coordinates corresponding to the proportion of kinetochore signals at the top (abscissa) and bottom (ordinate) surfaces of the nucleus (c.f. Fig. 6a). d–j Histograms showing number of nuclei (d–f SH-EP cells, g–j fibroblasts) with regard to preferential top or preferential bottom location of kinetochores. The ordinate shows the number of nuclei, the abscissa the R-axis (c.f. Fig. 6c). k–n Partial projections of XZ ortho-sections through characteristic fibroblast nuclei at the stages corresponding to those for histograms g–j. k, l Nuclei with predominantly bottom signals; m, n nuclei with more or less equally abundant signals at the top and bottom surfaces. o Standard deviations of distributions shown in d–j. p Successive cell cycle stages of SH-EP cells (projections of confocal stacks) after kinetochore immunostaining (green) and chromatin counterstain (red). Note concentration of kinetochores at one pole of anaphase rosettes, which is inherited by the following telophase. r Projections of XZ ortho-sections through the nuclei of early G1 SH-EP cells showing preferential top (t) and preferential bottom (b) localization of kinetochores. Bars represent 5 μm

Full size image

Figure 7a–j shows the distribution of kinetochores in both cell types studied at different stages of the cell cycle using the same histograms as in Fig. 6a,c, respectively. Figure 7d,e shows a clearly bimodal distribution for G1 and S phase SH-EP cells with the minimum not far from zero. This means that nuclei with similar proportions of kinetochores at the top and bottom sides are the least frequent, although one would expect them to prevail. Instead, there are two groups of cells: (1) with the majority of kinetochores gathered at the bottom side and (2) with the majority of kinetochores at the top side. On the histograms for G1 phase and S phase fibroblasts (Fig. 7g,h), a similar bimodality is seen, although it is less obvious. The left peak is much higher than the right one, which means that the majority of cells have most peripheral kinetochores on the bottom side of the nucleus (correspondingly, there are more bottom centromeres, than top centromeres; see above). In G0 fibroblasts (Fig. 7j), we observed essentially one group of cells with a moderate prevalence of kinetochore signals localized at the nuclear bottom compared with the nuclear top. The distribution for G2 fibroblasts (Fig. 7i) had an intermediate character between those for G1-S and G0. In G2 SH-EP cells the bimodality of the kinetochore signal distribution was also much less pronounced, than at G1 and S phase (Fig. 7f). This means that the proportion of cells with kinetochores more or less equally divided between the top and bottom surfaces increased notably in G2 and G0 (Fig. 7k–n). Correspondingly, the standard deviation of the distributions described above decreased from G1 and S to G2 and G0 (Fig. 7o).

The kinetochore provides the core structure of the centromere to which microtubules attach during mitosis; however, it is only a small part of the whole human centromere, which comprises a relatively large (1–5%) chromosome region consisting of α-satellite DNA (Choo 1997). To trace the behavior of centromeres as a whole during the cell cycle, we performed 3D-FISH with a pancentromeric α-satellite probe in nuclei of normal diploid fibroblasts in G0, early G1, and S phase. Compatibility of FISH and BrdU/pKi67 detection (Bridger and Lichter 1999; Solovei et al. 2002a) allowed us to determine the stage of the cell cycle of individual cells. The shape of the centromere signals was strikingly different between cycling and non-cycling cells. Early G1 centromeres were mostly spherical compact bodies; S phase centromeres looked similar but were notably larger than those in G1 (Fig. 8b,c,e,f). Centromere signals in G0 fibroblasts had elongated and highly irregular shapes (Fig. 8h,i).

Fig. 8a–i
figure 8

Shape of fibroblast nuclei and centromere signals in G1 stage (a–c), mid-S stage (d–f), and in G0 (g–i). a, d, g Volume rendering of the nuclei (red) and centromere signals (green); b, e, h maximum intensity projections of image stacks; c, f, i reconstruction of centromere signals by surface rendering. Note elongated irregular shape of the centromere signals in G0 cells in contrast to their more or less spherical shape in cycling cells. Bars represent 10 μm

Full size image

Kinetochore distribution and dynamics in spherical nuclei of cells growing in suspension

In contrast to the flat shapes of fibroblast and SH-EP cell nuclei, lymphocytes and lymphoblastoid cells possess nuclei with a spherical shape (Fig. 4d,e). Kinetochore signals were scored on RGB galleries of serial optical sections, as well as on XZ/YZ sections through 3D-reconstructed nuclei using Amira 2.3. Positions of kinetochore signals within the nucleus were classified as peripheral or internal according to the scheme in Fig. 1a.

As in the case of flat cells, the distribution of kinetochores is clearly not random. The diameter of nuclei in lymphocytes and lymphoblastoid cells is 7–11 μm depending on the cell cycle stage. The kinetochore signal was 0.1–0.3 μm in size. In the case of random distribution, only 13–20% of kinetochores would be situated in the outer 0.5 μm thick shell and would be classified as peripheral. In reality, at least ca. 40% of kinetochores occupy a peripheral position (Fig. 5). Scores proved to be very similar for early, mid, and late S phase, and therefore these data were summarized for the whole S phase (Fig. 5). In G0 lymphocytes about 90% of the kinetochores had a peripheral location. In early G1, the frequency of kinetochores with internal localization reached 50% (Fig. 5). These kinetochores were distributed throughout the whole nuclear volume, often being in contact with pKi67-positive granules (Fig. 2b) represented by small forming nucleoli and heterochromatin granules. In S phase, about 70% of signals were found peripherally, while about 30% were internally located and predominantly adjacent to the nucleoli (identified by TO-PRO-3 counterstain or anti-Ki67 immunostaining). In late G2, the proportion of internal kinetochores was about 50% as in G1 nuclei. Interestingly, the majority of internal signals at this cell cycle stage were located adjacent to the nucleoli (Fig. 2c). The intranuclear distribution of kinetochores in cycling and non-cycling lymphoblastoid cells was similar to those described above for the lymphocytes (Fig. 5). The proportion of internal signals was slightly higher but their localization with regard to pKi67 immunofluorescence followed the same pattern.

In addition to the visual examination, the radial 3D positioning of kinetochores at different cell cycle stages and in non-cycling lymphocytes was evaluated using the 3D-RRD computer program (see Materials and methods). Figure 4g shows the radial distribution of fluorescence signal intensities as a function of the relative nuclear radius for G0, early S, mid-S, late S, late G2, and early G1 lymphocytes. The radial intensity distribution in G0 cell nuclei shows a single, relatively narrow peak with a modal value at about 85% of the relative nuclear radius and thus confirms the mostly peripheral location of kinetochore signals. All curves for cycling cells also show a major peak at about 85%, but in addition a “plateau” at a relative radius of ca. 45–65%. This small second peak apparently represents the internal signals associated with nucleoli. The fraction of internally located signals increases in this series from S early → S mid → S late → late G2 → early G1. The curve for early G1 nuclei is rather similar to that for the DNA counterstain, reflecting the particularly pronounced variability of kinetochore locations at this cell cycle stage.

Monocytes from peripheral blood

One of our blood samples was taken from a person with acute inflammatory disease. This sample was strongly enriched in monocytes. Monocytes are the largest of the circulating leukocytes and possess a characteristic kidney-shaped nucleus (Fig. 4f). Anti-pKi67 staining showed that all monocytes were in G0. Immunostaining of CENP-B/CENP-C revealed on average 27 kinetochore signals per nucleus (n=14) with strict peripheral intranuclear positioning (90% of peripheral signals). The few internally located signals (ca. three per nucleus) were all adjacent to the two nucleoli found in these cells.

Discussion

Kinetochore arrangements were studied in four diploid human cell types (diploid fibroblasts, lymphocytes, lymphoblastoid cells, monocytes) and in a tumor cell line (SH-EP N14 neuroblastoma cells).

Clustering of kinetochores

Our data demonstrate that clustering of kinetochore regions in all cell types changes with the cell cycle stage. Very little or no clustering was observed immediately after mitosis (early G1) as indicated by signal numbers close to the expected chromosome number. The number of kinetochore signals decreased in late G1/S and even more so upon exit from the cell cycle (G0). This change could be explained either as a result of kinetochore antigen masking or as the result of kinetochore clustering. Several arguments support the latter conclusion. Signals in late G1/S and G0 were generally larger and more intense than in early G1. Furthermore, the average number of 13 kinetochore signals in G0 lymphocytes observed in the present study corresponds to the number of 11 and 13 centromere signals reported in two FISH studies performed with a pancentromeric α-satellite probe (Alcobia et al. 2000; Weierich et al. 2003, respectively). Notably, kinetochore clustering during S phase was less pronounced in flat nuclei (fibroblasts, SH-EP cells) than in spherical nuclei (lymphocytes, lymphoblastoid cells). The reason for this difference is unknown, but topological constraints depending on the different relationships between nuclear surface and nuclear volume in flat and spherical nuclei may be considered. Strong centromere clustering was observed in particular in terminally differentiated cells, such as monocytes (present study) and neurons (Manuelidis 1984; Martou and De Boni 2000; Solovei et al. 2004a).

The possible mechanisms and reasons for centromere clustering remain unclear. Chromocenters formed by clustered centromeres are typical of many cell types (Haaf and Schmid 1991; Manuelidis 1984). Chromocenters consist mainly of highly repetitive heterochromatin (pericentric hetrochromatin) and are known to have certain epigenetic “marks”, such as DNA methylation, histone H3 methylation at lysine 9 (H3-K9), and enrichment in heterochromatin protein-1 (HP1) (reviewed in Bird 2002; Richards and Elgin 2002; Lachner and Jenuwein 2002; Lachner et al. 2003; Maison and Almouzni 2004). One of the putative functions of chromocenters is linked to regulation of transcriptional activity by silencing genes situated in the vicinity of a chromocenter (Brown et al. 1997, 1999; Baxter et al. 2002). Therefore, the clustering of centromeres that takes place during G1 and early S probably manifests the establishment of epigenetically controlled “silencing” zones in the nucleus. This assumption corresponds to the fact that centromere clustering is especially strong in quiescent and terminally differentiated cells.

Kinetochore movements in cycling cells and upon exit from the cell cycle

The present study demonstrates for the first time that kinetochore topology in all investigated cell types differs markedly in G0 and G1 cells and changes from early to late G1, as well as from early to mid-late S and G2. In quiescent cells (G0), most kinetochores showed a peripheral localization (80–95% depending on the cell type), while the few internally located kinetochores were found mostly adjacent to the nucleoli. In all cell types, except for monocytes, which were only studied in a terminally differentiated state, we found distinct changes of kinetochore arrangements during interphase. In early G1, a considerable fraction of kinetochores (about 30–60% depending on cell type) were located in the nuclear interior, clearly remote from the nuclear envelope. From early to late G1, the internal fraction of signals decreased (to about 20–50%) with the result that most kinetochores now attained a peripheral localization (about 50–80%). In early S phase the fraction of peripherally located kinetochores became even higher (about 50–90%). In fibroblasts and SH-EP cells, the fraction of internal kinetochore signals started to increase again (about 20–25%) from mid to late S phase, while it remained roughly constant in lymphocytes and lymphoblastoid cells. During G2, the internal and peripheral kinetochore fractions did not change further in fibroblasts and SH-EP cells, while the internal fraction increased significantly in lymphocytes and lymphoblastoid cells. Fibroblasts and PHA-stimulated lymphocytes showed a higher frequency of peripherally located kinetochores compared with lymphoblastoid cells and neuroblastoma cells; also the fraction of centromeres that switched from a peripheral to an internal localization during mid-S/G2 was smaller in the latter two cell types. In conclusion, our data confirm cell type and cell cycle dependent differences in kinetochore arrangements. Possible functional implications and mechanisms involved in the dynamic centromere patterns have not yet been elucidated.

Changes in kinetochore arrangements during early G1 are consistent with a live cell imaging study performed in HeLa cells (Walter et al. 2003). This study indicated considerably higher chromatin mobility during early G1 compared with subsequent interphase stages, from mid G1 to late G2, when movements of chromosome territories/chromatin domains were strongly restricted. Furthermore, replication foci during S phase were found to be rather immobile except for some constrained Brownian movements (Leonhardt et al. 2000). These findings suggest that centromere regions may undergo more pronounced movements from mid G1 to late G2 than other chromosomal subregions. We do not know, however, whether all centromeres may occasionally perform large-scale movements (i.e., movements over a range of several micrometers), or whether the capability for such movements is restricted to the centromeres of certain chromosome territories.

Results from previous studies are in partial or complete disagreement with our present findings. Bartholdi (1991) investigated 3D centromere arrangements in a growing population of human fibroblasts, employing laser confocal scanning microscopy after staining with anti-kinetochore antibodies. In this study cell cycle staging was based on fluorescence measurements of propidium iodide in situ. For G1 nuclei, Bartholdi reported that many centromeres were located in association with nucleoli or fused in chromocenters that expanded from nuclear top to nuclear bottom. No specific association of kinetochores was observed with the nuclear envelope. During S phase, chromocenters dispersed often forming patterns of rings or lines. In contrast, our present study revealed that most centromeres were located at the 3D nuclear periphery both in G0 and cycling fibroblasts, while only a minor fraction of centromeres moved from the nuclear periphery to the nuclear interior and vice versa.

Weimer et al. (1992) also employed anti-kinetochore antibodies to study centromere arrangements in non-stimulated (G0) and stimulated human lymphocytes (G1, S, G2). They performed a 2D evaluation based on conventional epifluorescence microscopy and found a distinct tendency for a peripheral position of centromeres during G0 and G1, which became weaker during S phase. While we agree with the general conclusions of Weimer et al. concerning the dynamics of centromere topology during interphase, the difference in kinetochore topology, in particular between G0 and early G1 nuclei demonstrated in the present study, escaped their notice. This may be due to a less precise staging of cells. Weimer et al. (1992) synchronized stimulated lymphocytes by a thymidine block and then fixed cells at different times after release from the block. The time of harvest was used for a rough discrimination of G0, G1, S and G2, as well as G1 and early S of the subsequent cell cycle. In an attempt to confirm the cycle stage of the cells studied in situ after anti-kinetochore antibody staining, a fraction of cells was stained in parallel with Hoechst 33258 and subjected to flow cytometry. In our present study we took great care to establish a protocol that enabled us to discriminate unequivocally between G0 and early and late substages of G1, as well as early and late S and G2 phase at the level of single cells in situ.

Ferguson and Ward (1992) and Vourc’h et al. (1993) performed FISH and confocal microscopy experiments with stimulated human and mouse lymphocytes, respectively. For cell cycle staging, cells were flow sorted after DNA/RNA staining with propidium iodide. Probes for major and minor satellite DNA were used for murine cells. Ten chromosome-specific centromere probes were employed for human cells but detailed data were only presented for centromeres of chromosomes 1, 7, 11, and 17. Consistent with our present findings, the authors observed a centromere repositioning from the nuclear periphery to the nuclear interior during S/G2. For G1 lymphocyte nuclei they state that most centromeres were located in the nuclear periphery, although they were not able to discriminate cells in G1 from cells in G0. In the present study we found that about 90% of centromeres were located in the periphery of human lymphocyte nuclei in G0, while about half of centromeres were located in the periphery and half in the nuclear interior, including adjacent to the nucleoli, of lymphocyte nuclei in early G1.

Hulspas et al. (1994) carried out FISH with a human chromosome 11 centromere-specific probe in both non-stimulated lymphocytes harvested directly from peripheral blood and in cultured, PHA-stimulated human lymphocytes. Laser confocal microscopy using the nuclear center as a reference point revealed that the distribution of the chromosome 11 centromeres appeared to be random during the G0 stage, while most, if not all, centromeres were found at the nuclear periphery in G1. The peripheral topology of kinetochores described in the present study for G0 lymphocytes strongly differs from the apparently random topology described by Hulspas et al. for chromosome 11 centromeres. This result needs to be confirmed by independent experiments in order to demonstrate that chromosome 11 centromere topology is an exceptional case. It should be noted that PHA stimulates only T lymphocytes and, accordingly, non-cycling nuclei from B lymphocytes and T lymphocytes were assessed together with a fraction of stimulated T cells. In the present study we were able to distinguish cycling T cells unequivocally from non-cycling T cells and B cells.

Asymmetry of kinetochore distribution in nuclei of adherently growing cycling cells may reflect the behavior of the preceding anaphase rosettes

More than a century ago Carl Rabl described a polar orientation of chromosomes in the interphase nuclei of Salamandra maculata with centromeres and telomeres positioned at opposite sides (Polfeld and Gegen-Polfeld) of the nucleus (Rabl 1885). According to Rabl, this polar orientation results from anaphase chromosome movements (see for example, Abranches et al. 1998). In mammalian cells such a Rabl orientation is rarely observed (Weierich et al. 2003; reviewed in Parada and Mistelli 2002) likely due to additional chromatin movements that occur during telophase/early G1 (Walter et al. 2003). In the present study we observed an asymmetric distribution of kinetochores in the rather flat nuclei of adherently growing fibroblast and SH-EP cells. Kinetochores were located predominantly at the nuclear bottom in fibroblast nuclei, while in SH-EP cells the fraction of nuclei with centromeres predominantly located at the nuclear top was similar to the fraction of nuclei with centromeres preferentially located at the bottom (Fig. 7).

For an explanation of this obvious nuclear asymmetry of centromere arrangements we propose the following hypothesis, which we call the “fallen rosette” scenario (Fig. 9). In fibroblasts both the metaphase plate and the two resulting anaphase rosettes are arranged perpendicular to the substratum, while the spindle is arranged parallel to the substratum (our unpublished observations). During anaphase the centromeres are located on the external side of the chromatin mass facing their respective centriole (Fig. 7p for SH-EP cells; see also Habermann et al. 2001, Fig. 5e for human fibroblasts). At the end of anaphase the two anaphase rosettes fall over, resulting in a parallel arrangement with respect to the growth surface. The preferential location of centromeres at the bottom side of G1 fibroblast nuclei suggests that each anaphase rosette should fall over preferentially with their centromeres downwards. Correspondingly, the two anaphase rosettes should usually fall over in opposite directions yielding symmetrical arrangements of chromosome territories in the two daughter nuclei. We noted that the spindles in anaphase rosettes of SH-EP cells were located at various angles to the growth surface (our unpublished observations) and the direction to which a given anaphase rosette falls over should depend on this angle. Since the angle is the same for both anaphase rosettes of a given cell, while their centromeres are located at opposite sites of the two rosettes, we should expect that one anaphase rosette falls over with the centromeres facing the substratum, while the other falls over with an upward centromere positioning. In G0 and G2 fibroblasts most nuclei still showed their centromeres preferentially at the nuclear bottom, but the proportion of nuclei with kinetochores at the nuclear top was higher compared with nuclei in G1 and S phase. Similarly, in G2 SH-EP cells we observed a decreased fraction of nuclei with centromeres predominantly concentrated at either the nuclear top or bottom compared with G1 and S cells, while the fraction of nuclei showing similar numbers of centromeres in top and bottom positions increased. By and large, the further cells had departed from mitosis, the more uniformly their centromeres were distributed at the lower and upper nuclear periphery (Fig. 9). To explain these changes we must invoke some additional centromere migration during interphase and in postmitotic cells (Manuelidis 1985; Martou and De Boni 2000; Solovei et al. 2004a). The importance of the fallen rosettes mechanism for determination of the nuclear chromosomal arrangement has earlier been suggested by us based on the data on relative positions of small and large chromosomes in interphase nuclei (Habermann et al. 2001). The fact that data of a very different kind, the distribution of centromeres, also correspond to the fallen rosettes scenario strongly supports this hypothesis.

Fig. 9
figure 9

Changes in the localization of kinetochores in cell cycle: the “fallen rosettes” scenario. At the late anaphase stage kinetochores are concentrated on one side of the rosette. The rosette falls on one side, as a result of which nuclei with kinetochores gathered at the top side (left) or at the bottom side (right) arise. Judging from the relative amounts of top and bottom kinetochores in the two cell types studied, the bottom (right) variant strongly prevails in fibroblasts, while in SH-EP cells both variants are equally probable. After late G1, certain rearrangements of kinetochores take place owing to which the numbers of kinetochores at the top and bottom sides become more even in G2 or G0

Full size image

The dynamics and highly reproducible changes of centromere topology during the cell cycle, upon exit from the cell cycle and during terminal differentiation cannot be explained as a result of Brownian movements. The question as to whether these changes play a role in epigenetic gene regulation and require energy dependent mechanism(s) is a matter for future studies.