Histochemistry and Cell Biology

, Volume 125, Issue 1, pp 103–117

The microarchitecture of DNA replication domains


  • Anatoly A. Philimonenko
    • Institute of Experimental MedicineAcademy of Sciences of the Czech Republic
  • Zdeněk Hodný
    • Institute of Experimental MedicineAcademy of Sciences of the Czech Republic
  • Dean A. Jackson
    • Faculty of Life Sciences, Jackson’s MillUniversity of Manchester
    • Institute of Experimental MedicineAcademy of Sciences of the Czech Republic
Original paper

DOI: 10.1007/s00418-005-0090-0

Cite this article as:
Philimonenko, A.A., Hodný, Z., Jackson, D.A. et al. Histochem Cell Biol (2006) 125: 103. doi:10.1007/s00418-005-0090-0


Most DNA synthesis in HeLa cell nucleus is concentrated in discrete foci. These synthetic sites can be identified by electron microscopy after allowing permeabilized cells to elongate nascent DNA in the presence of biotin-dUTP. Biotin incorporated into nascent DNA can be then immunolabeled with gold particles. Two types of DNA synthetic sites/replication factories can be distinguished at ultrastructural level: (1) electron-dense structures—replication bodies (RB), and (2) focal replication sites with no distinct underlying structure—replication foci (RF). The protein composition of these synthetic sites was studied using double immunogold labeling. We have found that both structures contain (a) proteins involved in DNA replication (DNA polymerase α, PCNA), (b) regulators of the cell cycle (cyclin A, cdk2), and (c) RNA processing components like Sm and SS-B/La auto antigens, p80-coilin, hnRNPs A1 and C1/C2. However, at least four regulatory and structural proteins (Cdk1, cyclin B1, PML and lamin B1) differ in their presence in RB and RF. Moreover, in contrast to RF, RB have structural organization. For example, while DNA polymerase α, PCNA and hnRNP A1 were diffusely spread throughout RB, hnRNP C1/C2 was found only at the very outside. Surprisingly, RB contained only small amounts of DNA. In conclusion, synthetic sites of both types contain similar but not the same sets of proteins. RB, however, have more developed microarchitecture, apparently with specific functional zones. This data suggest possible differences in genome regions replicated by these two types of replication factories.


DNA replicationCell nucleusUltrastructureNuclear bodiesReplication factoryReplication fociNuclear proteins



Proliferating cell nuclear antigen


Green fluorescent protein


Replication bodies


Replication foci


Sites of DNA replication are not diffusely spread throughout eukaryotic nuclei but concentrated in discrete foci. These sites can be visualized by light microscopy in intact or permeabilized cells using incorporation of bromodeoxyuridine or biotin-dUTP, and subsequent immunofluorescent labelling of the sites containing the incorporated analogues (Blow and Laskey 1986, 1988; Bravo and Macdonald-Bravo 1987; Hozak et al. 1993; Hutchison et al. 1987; Mills et al. 1989; Nakamura et al. 1986). Replication foci are not fixation artifacts because similar structures are seen after incorporation of fluorescein-dUTP into permeabilized, but unfixed cells (Hassan and Cook 1993) and in intact living mammalian cells expressing chimeric GFP-PCNA protein (Cardoso et al. 1997; Leonhardt et al. 2000a). They also remain even when most chromatin is removed (Hozak et al. 1993; Nakayasu and Berezney 1989), implying that they are attached to an underlying structure. Early during S-phase the replication sites are small and discrete. Later they become larger when centromeric and other heterochromatic regions are replicated (Fox et al. 1991; Hozak et al. 1994; Humbert and Usson 1992; Kill et al. 1991; Manders et al. 1992; Nakayasu and Berezney 1989; O’Keefe et al. 1992). Replication sites may also lie close to the sites of transcription, especially early during S-phase (Hassan et al. 1994).

At the ultrastructural level, two types of DNA synthetic sites can be distinguished. One type are electron-dense structures attached to an internal lamin-containing nucleoskeleton (Hozak et al. 1993, 1995). As each replication site of this type contains at least 40 active forks and associated leading- and lagging-strand polymerases (Hughes et al. 1995), they were called replication “factories”. Subsequently, we found that these factories were identical to a subset of nuclear bodies identified previously in studies on nuclear structure (Bouteille et al. 1967; Brasch and Ochs 1992). Both were labeled with biotin-dUTP and their number, shape and distribution during the cell cycle changed in much the same way (Hozak et al. 1994). Again, they cannot be fixation artifacts as they can also be labeled in non-permeabilized cells using antibodies directed against their typical components. In this study, they are referred to as “replication bodies” (RB). Another type of DNA synthetic sites (termed in this study “replication foci”, RF) corresponds to the sites of extra-RB labeling which have no distinct underlying structure. In resinless sections this label was found in focal regions on the diffuse part of the nucleoskeleton besides the dense RB. Such scattered extra-RB labeling becomes more prominent during mid S-phase, and later on the RF become concentrated under the lamina (Hozak et al. 1994). In the latest experiments we have also shown that these foci are the dynamic structures and possibly represent some fixed multiprotein complexes, through which nascent DNA reels (Philimonenko et al. 2004).

Replication sites are also known to contain proteins specifically involved in DNA synthesis (e.g., DNA polymerase α, PCNA and DNA methyltransferase; Hozak et al. 1993; Leonhardt et al. 1992; Prosperi et al. 1994) as well as others that might be involved in regulation of this process (e.g. cyclin A, CDK2 and RPA70, but not cyclin B1, CDC2 or RPA34; Adachi and Laemmli 1992; Cardoso et al. 1993; Sobczak-Thepot et al. 1993, for a review, see also Leonhardt et al. 2000b). However, the structural and functional differences between RF and RB are not known.

The aim of this study was to investigate ultrastructure and protein composition of these two types of DNA synthetic sites by immunoelectron microscopy using multiple immunogold labeling. As RF have no distinct underlying structure, we developed a new technique for statistical evaluation of immunogold colocalization patterns (Philimonenko et al. 2000; http://nucleus.biomed.cas.cz/gold). Our experiments have shown that in comparison to RF, RB have a distinct microarchitecture, with certain components concentrated in specific regions. This observation allows us to suggest the existence of specific functional zones inside the RB. On the other hand, both types of replication sites contain similar set of major synthetic and regulatory components, which indicates that RF are true replication factories. However, at least four principal regulatory and structural proteins (Cdk1, cyclin B1, PML and lamin B1) differ in their presence in RB and RF. This data suggest possible differences in genome regions replicated by these two types of replication factories.

Materials and methods

Cell culture and cell synchronization

Suspension culture of HeLa cells was grown in minimal essential Eagle medium (S-MEM, Sigma, St. Louis, Missouri, USA) supplemented with 10% foetal calf serum (Sigma, St. Louis, Missouri, USA) at 37°C on rotamix. Cells were synchronized in mitosis using thymidine and nitrous oxide/nocodazole block according to Rao (1968) and Jackson and Cook (1986). Briefly, cells blocked in S-phase by incubation in 2.5 mM thymidine for 22 h were then rinsed carefully, re-grown for 4 h in fresh medium, and arrested in mitosis by either nitrous oxide at high pressure or nocodazole (20 ng/ml, Sigma, St. Louis, Missouri, USA) for 8 h (>98% cells in mitosis). Cells were then washed carefully and re-grown. Cell samples (about 5×10cells per sample) were taken 14 h after the release from the mitosis to obtain cell populations in mid S-phase (Hozak et al. 1994).

DNA replication in permeabilized cells

Synchronized HeLa cells were collected by centrifugation, washed twice in phosphate-buffered saline (PBS), and encapsulated in agarose microbeads to protect them during subsequent manipulations (Jackson et al. 1988). Cells were then permeabilized with saponin (0.1 mg/ml; Sigma, St. Louis, Missouri, USA) in “physiological” buffer (PB; 22 mM Na+, 130 mM K+, 1 mM Mg2+, <0.3 μM free Ca2+, 67 mM Cl, 65 mM CH3COO, 11 mM phosphate, 1 mM ATP, 1 mM dithiothreitol and 0.1 mM phenylmethylsulphonyl fluoride, pH 7.4; Jackson et al. 1988) for 3 min at 0°C and then washed twice by PB. Ten-times concentrated replication mix prewarmed to 33°C was added to permeabilized cells to give final concentrations of 100 μM CTP, UTP and GTP, 250 μM dCTP, dATP and dGTP, plus 100 μM biotin-16-dUTP (Boehringer Mannheim, Mannheim, Germany) and 2.15 mM MgCl2. After 15 min incubation at 33°C the reaction was stopped with 10 volumes of ice-cold PB. Nascent DNA chains are extended by <750 nucleotides under these conditions to ensure that the incorporated analogues remained close to sites of polymerization (Hassan et al. 1994; Hozak et al. 1993).


Primary monoclonal mouse antibodies were directed against human DNA polymerase α (clone SJK 287–38; Bensch et al. 1982), proliferating cell nuclear antigen (PCNA; Santa Cruz Biotechnology), DNA (Boehringer Mannheim, Mannheim, Germany; the antibody binds to both single- and double-stranded DNA), rat liver 5E10 antigen, which corresponds to promyelocytic leukemia protein PML (Stuurman et al. 1992), A1 hnRNP (clone 9H10—a gift of Dr. G. Dreyfuss—which is similar to clone 4B10; Pinol-Roma et al. 1988), C1/C2 hnRNP (clone 4F4; Choi and Dreyfuss 1984), cdc2 kinase (now designated Cdk1, affinity purified IgG1 raised against a synthetic peptide corresponding to a fifteen amino acid sequence located near the carboxyl terminus of the intact human p34cdc2, Transduction Laboratories), cyclin-dependent kinase 2 (cdk2, affinity purified IgG2a raised against a 21.7 kDa protein fragment corresponding to residues 109–298 of cdk2; Transduction Laboratories), human CDC34 protein (affinity purified IgG2a raised against a 21.5 kDa protein fragment corresponding to amino acids 108–298 of the human CDC34; Transduction Laboratories), human cyclin B1 (Santa Cruz Biotechnology), lamins A/C and B1 (131C1 and 119D5-F1, respectively, both were gifts of Dr. Y. Raymond; Raymond and Gagnon 1988), p21 (Cip1/Waf1/CAP20/Sdi1/Pic1, an IgG2a raised against the amino-terminal residues 1–150 of the human Cip1 protein, Transduction Laboratories), and cyclin D1 (an IgM raised against a 23 kDa protein fragment corresponding to amino-terminal residues 1–200 of cyclin D1, Transduction Laboratories).

Primary rabbit polyclonal antibodies (affinity purified) were directed against human cyclin A (affinity purified serum C-22, an IgG raised against the amino-terminal residues 104–123; Santa Cruz Biotechnology), human cyclin E (affinity purified IgG raised against a peptide corresponding to amino acids 377–395 of cyclin E of human origin; Santa Cruz Biotechnology), rat liver nucleolar phosphoprotein Nopp 140 (Meier and Blobel 1992), p80-coilin (R228, a gift of Dr. R. L. Ochs; Andrade et al. 1993).

Primary human auto-antibodies recognize Sm and SS-B/La antigens (both ANA/CDC reference sera, Centers for Disease Control, Atlanta, USA). Secondary 5 nm gold-conjugated antibodies produced in goat, included anti-mouse IgG or IgM, anti-human IgG, and anti-rabbit IgG (British BioCell International Ltd., Cardiff, UK). Sites containing incorporated biotin were detected using goat anti-biotin antibodies conjugated with 10 nm gold particles (British BioCell International Ltd., Cardiff, UK).

Postembedding immunoelectron microscopy

Following DNA replication in permeabilized HeLa cells, the cells were fixed on ice for 30 min at 0°C in 3% paraformaldehyde and 0.1% glutaraldehyde in Sörensen buffer (SB; 0.1 M sodium/potassium phosphate buffer, pH 7.3), washed twice by SB (10 min each), and resuspended in 1% low melting point agarose (Sigma, St. Louis, Missouri, USA) in SB prewarmed to 37°C. Cells were then spun down and the pellet was cut into small pieces. The pieces were dehydrated in series of ethanol solutions with increasing concentration of ethanol. The ethanol was then replaced in two steps by LR White resin (Polysciences Inc., Warrington, USA), and the resin was polymerized for 5 days at −20°C under UV light. After cutting 80 nm sections, nonspecific labeling was blocked by preincubation with 10% normal goat serum (British BioCell International Ltd., Cardiff, UK), 1% BSA and 0.1% Tween 20 in PBS for 30 min at room temperature. For double immunogold labeling experiments, the sections were incubated with primary antibodies against appropriate antigens (2–10 μg/ml), washed three times in PBT (0.005% Tween 20 in PBS), incubated with appropriate 5 nm gold-conjugated antibodies and with 10 nm gold-conjugated goat antibodies to biotin, washed again twice in PBT, then twice in bidistilled water, and air-dried. Finally, sections were contrasted with a saturated solution of uranyl acetate in water (4 min) and observed in electron microscopes Philips CM100 (Philips, Eindhoven, The Netherlands) or Morgagni (FEI, Eindhoven, The Netherlands) operated at 80 kV. Control incubations without primary antibodies proved that the signal was highly specific and that there was no cross-reactivity in case of multiple labelling.

Cytochemical detection of DNA in replication factories

Synchronized cells that had been permeabilized and allowed to make DNA in the presence of biotin-dUTP were washed in SB, fixed and embedded in LR Wight resin as described above. Silver-gold 80 nm thick sections were mounted on gold grids and DNA was specifically stained with osmium-ammine according to Derenzini et al. (1982). Briefly, DNA was hydrolyzed in 5 N HCl (30 min, room temperature), sections washed in distilled water (1 h), incubated (10–15 min) in a freshly prepared osmium-ammine B solution (Polysciences; Olins et al. 1989), washed thoroughly in distilled water and observed in the electron microscope. In order to confirm that replication sites were being examined in such sections, the presence of incorporated biotin in RB in adjacent sections was monitored. Pretreatment with DNase (1 μg/ml, 30 min) abolished this osmium-ammine staining.

Identification of colocalization patterns in replication foci

For the statistical analysis of colocalization patterns, about 50 random digital electron microscope images of nuclear sections per each experimental group were taken. Each experimental group corresponds to an experiment in which biotin-labeled nascent DNA was colocalized with one of the proteins of interest. XY coordinates of all gold particles were recorded from the images using a macro developed for LUCIA image-processing software (Laboratory Imaging Ltd., Prague, Czech Republic). In order to analyze the colocalization of particles of different size, the pair cross-correlation function (PCCF) and the cross-K function were used (Philimonenko et al. 2000). PCCF is the ratio of the density of particles of one type at the given distance from a typical particle of the second type to the average density of the particles of the first type. The cross-K function is the ratio of the number of particles of one type at a distance shorter than given distance from a typical particle of the second type divided by the average density of the particles of the first type. The pair cross-correlation function was used for exploratory analysis, while the cross-K function for statistical testing. The above functions were calculated from pooled data from all images of nucleoplasm in each experimental group. For more detailed description of this method see also our www pages at http://nucleus.biomed.cas.cz/gold. Additionally, densities of gold labels were estimated in randomly chosen sections, and in replication bodies.

Statistical evaluation of labeling density in inner and outer zones of replication factories

LR White sections of mid S-phase HeLa cells were labeled using an anti-biotin antibody coupled with 5 nm gold particles as described above; only one side of the 80 nm section was labeled. Images of 300 RB labeled with a density >30 gold particles per μm2 were captured with a CCD camera and the number of gold particles was counted in the internal and external area of each RB (both occupying 50% of the total RB area). The results were then statistically evaluated using Student’s t test.


To uncover the ultrastructural organization of DNA synthetic sites in eukaryotic nuclei, mid S-phase synchronized HeLa cells were gently permeabilized and then allowed to synthesize DNA in the presence of biotin-16-dUTP. After fixation and sectioning, sites containing the incorporated biotin were immunolabeled with 10 nm gold particles. Simultaneously, various proteins of interest were immunolocalized with 5 nm gold particles (see Materials and methods).

Figures 1 and 2 represent typical electron micrographs of HeLa cell nuclei, in which sites containing nascent DNA with incorporated biotin are localized. Two morphological types of DNA synthetic sites/replication factories can be distinguished: (1) replication bodies (RB)—ellipsoid or globular dense structures of about 400 nm in diameter at this stage of S-phase (Fig. 1a) and (2) replication foci (RF)—DNA replication sites of about 100 nm in diameter without any typical underlying structure (Figs. 1b, 2a, b). In Fig. 1a, a densely stained RB (f) is labeled with the large (10 nm) gold particles (incorporated biotin). Most DNA polymerase α (which also serves in our study as a positive control, small 5-nm particles) is also concentrated within—and distributed uniformly throughout—the RB. Small foci of label are also found in the nucleoplasm (Fig. 1b; Table 1). In Figs. 1b, 2a and b, RF recognized by labeling nascent DNA are presented (clusters of large gold particles). Clusters of DNA polymerase α are detectable here separately (Fig. 1b, arrowhead) or intermingled with clusters of labeled nascent DNA (Fig. 1b, arrow). The functional difference between these two types of DNA synthetic sites is unknown.
Fig. 1

There are two ultrastructurally distinct types of replication sites in nuclei of HeLa cells. DNA polymerase α colocalizes with nascent DNA in both RB and RF. Permeabilized cells in mid S-phase were allowed to make DNA in the presence of biotin-dUTP, ultrathin sections prepared and immunolabeled. Five nanometer gold particles mark DNA polymerase α, whilst 10 nm particles label sites containing incorporated biotin (i.e., nascent DNA). a RB (f) is strongly labeled with both antibodies. b Clusters of DNA polymerase α are detectable separately (arrowhead) or intermingled with RF (arrow). Bar: 100 nm

Fig. 2

Examples of electron micrographs of RF with strong or none colocalization of nascent DNA and proteins of interest. a Double immunogold labeling of PCNA (5 nm gold particles) and biotin-labeled nascent DNA (10 nm gold particles). Clusters of PCNA are detectable separately or intermingled with clusters of labeled nascent DNA (arrow). b Double immunogold labeling of Nopp140 (5 nm gold particles) and nascent DNA (10 nm gold particles). Nopp140 is detected throughout the nucleoplasm either separately (arrow), or sometimes close to labeled nascent DNA (arrowhead). While it is difficult to assess the colocalization visually, the statistical evaluation clearly demonstrated significant colocalization for PCNA and no colocalization for Nopp 140. Bar: 100 nm

Table 1

Antigens detected within DNA replication sites of both types





Significancy of colocalization with nascent DNA



Proteins involved in DNA replication

DNA polymerase α








Cell cycle regulators

Cdk1 (cdc2-kinase)






− to +

Small zones

p21 WAF1/Cip1


− to +

Small zones

Cyclin A


− to +

Small zones

Cyclin B1


− to +

Small zones

Cyclin E



Small zones enriched at chromatin contact sites with RB

Cyclin D1




Proteins involved in RNA metabolism

A1 hnRNP




C1/C2 hnRNP


− to +

Only in peripheral chromatin threads

Nopp 140














Other proteins



Lamin A/C


Lamin B1


Labeling densities (LD) in RB were expressed as “−” (<30 particles/μm2), “+” (30–59 particles/μm2), “++” (60–89 particles/μm2) and “+++” (>89 particles/μm2)

The significance of colocalization of nascent DNA and various nuclear proteins was tested using K-function in the interval of distances 30–250 nm at 1% (**) and 5% (*) critical values obtained by Monte-Carlo simulations of Poisson process. The results that differ for RB and RF are shown in bold italic

It is attractive to suppose that RB contain stores of the required enzymes and that DNA synthesis occurs as templates slide through polymerases fixed on the surface of these stores (Cook 1991; Hozak et al. 1993). Then little DNA should be found within the RB; alternatively, replicating DNA might be restricted to certain zones. Therefore, we investigated the DNA content of the RB using an anti-DNA antibody (Fig. 3a and b, small gold particles). Although surrounding chromatin was intensely labeled with small gold particles (arrows) running up to the RB, few, if any, particles are found over the RB. About 80% of RBs contained no DNA labeled inside (Fig. 3a), the remaining 20% contained typically a few gold particles (Fig. 3b). The density of labeling in the RBs with anti-DNA label was significantly higher (18.7 particles/μm2) than that of background (0.21 particles/μm2; P<0.01]. This suggests that RB contain less DNA than their surroundings. However, proteins can obscure the epitopes on DNA. The observation was therefore confirmed using osmium-ammine that specifically stains DNA (Fig. 3c). To ensure that replication sites were being examined in such sections, the presence of incorporated biotin in RB in adjacent sections was monitored (Fig. 3a, large gold particles). Whilst 30 nm chromatin fibrils (large arrows) apparently run up to the RB, only a few were found inside (small arrows). These internal fibrils were 10–15 nm thick.
Fig. 3.

RB contain little DNA. a, b 5 nm gold particles mark DNA and 10 nm particles label sites containing incorporated biotin (nascent DNA). Little DNA is detectable in replication bodies (f), while strings of gold particles apparently label chromatin fibrils running to them (arrows); most 5 nm particles are clustered outside factories over chromatin. In about 20% of factories, a weak labeling of DNA is present also inside of factories (b, arrows). c DNA visualized using the osmium-ammine technique. This section shows the unstained cytoplasm (c), densely stained condensed heterochromatin (ch) at the nuclear periphery and one gray RB (f) within the nucleus. The identification of the gray area as an RB was confirmed by immunogold detection of incorporated biotin in the corresponding region of the adjacent section (not shown). Densely stained fibrils (DNA) apparently run to the RB (large arrows) which contains only few fibrils ~10 nm thick (small arrows). Bar: 100 nm

The low concentration of DNA in the interior of the RB might suggest that DNA synthesis occurs at their surface; then, incorporated biotin should be concentrated there. To study the distribution of incorporated biotin label inside the RB, sections through 300 RB (labeled with at least 30 gold particles per μm2) were randomly selected, an area of each divided into an inner and outer part (with identical surface area) and the number of particles counted over each region. Fifty-two percent (SD±16) of all particles lay over the outer half. Student’s t test indicates that the value is not significantly different from a random distribution and thus suggests that nascent DNA is not preferentially concentrated in the inner or outer parts of RB.

To investigate protein composition of both types of DNA synthetic sites and localization of proteins in RB, double-labeling experiments like those illustrated in Figs. 1, 2 and 4 were then repeated for each of the antigens listed in Table 1. Table 1 summarizes the colocalization data for RF and labeling pattern and intensity for RB of all tested antigens. These are listed in groups in relation to their function. In each case, the tested antigen was labeled by small gold particles and incorporated biotin by large gold particles. Labeling densities (LD) in RB were measured and expressed in particles per μm2, while the significance of colocalization of nascent DNA and various corresponding antigens in RF was tested using K-function in the interval of distances 30–250 nm (for details see Materials and methods).
Fig. 4

Typical examples of localization of components involved in RNA metabolism and cell cycle regulators in mid S-phase RBs. Ten nanometer gold particles label nascent DNA, and 5 nm particles mark Sm antigen (a), SS-B/La antigen (b), p80-coilin (c), A1 hnRNP (d), C1/C2 hnRNP (e), and cyclin B1 (f) correspondingly. Bar: 100 nm

Proteins involved in DNA replication

As expected, DNA polymerase α (Fig. 1) and PCNA (Fig. 2a), colocalized strongly with both types of replication sites (see Table 1).

Cell cycle regulators

We next screened the presence of regulators of DNA synthesis and of the cell cycle in replication sites of both types. Cyclin A was present in variable amounts in mid S-phase RB in small zones and significantly colocalized with nascent DNA in the RF (Table 1). Cyclin B1 was also present in variable amounts in mid S-phase RB (Fig. 4f) in pattern similar to cyclin A, however, it did not colocalize with nascent DNA in RF (Table 1). Cyclin E was found in the sites of replication of both types, enriched at edges in RB. In contrast, no cyclin D1 was detected in either RB or RF. Cdc2 kinase was enriched in RB, while it was virtually absent from RF. CDC34 was not found in both types of DNA synthetic sites (see Table 1).

Nuclear lamins

The nuclear lamins (A/C and B) play an essential role in maintaining nuclear integrity. Besides formation of nuclear lamina, they are also present in the interior of the nucleus as distinct nucleoplasmic foci and possibly as a network throughout the nucleus. The formation of nuclear lamina is required for the initiation of DNA replication (Meier et al. 1991; Newport et al. 1990). However, no significant amounts of lamins A/C and B1 were detected in RB (Table 1). Lamin A/C was also absent from RF; however, lamin B1 significantly colocalized with nascent DNA in RF.


The PML protein is a human tumor suppressor concentrated in 10 to 20 nuclear bodies per nucleus (PML bodies) potentially involved in DNA metabolism (Jul-Larsen et al. 2004). There was no PML detected in RB; however, PML colocalization with nascent DNA in RF was observed (see Table 1).

Factors involved in RNA metabolism

As replication sites are often found close to the sites of transcription (Hassan et al. 1994), we checked whether some components involved in RNA processing were associated with RB and/or RF. The Sm auto antigen, which is a component of various small nuclear RNPs involved in RNA splicing, is localized in nuclear ‘speckles’ (Spector 1993). We found it concentrated within the RB as well as in RF (Fig. 4a; Table 1). Another nuclear auto-antigen SS-B/La (La protein protects the 3′ ends of several small RNAs from exonucleases; Wolin and Cedervall 2002) was also found both in RB and RF (Fig. 4b; Table 1). p80-coilin, a coiled bodies marker (Andrade et al. 1991), was concentrated both in RB (Fig. 4c; Table 1) and RF (Table 1). Within the RB, it had a characteristic zonal distribution (Fig. 4c). A1 hnRNP, a “core” hnRNP protein (Dreyfuss et al. 1993) and also a well known modulator of alternative pre-mRNA splicing (Blanchette and Chabot 1999), was also found in RB (Fig. 4d; Table 1). Another “core” hnRNP C1/C2, which is believed to be involved in pre-mRNA packaging, spliceosome assembly and in nuclear retention of unspliced hnRNA (Dreyfuss et al. 1993), is localized in RB and apparently runs in threads up to their edges (Fig. 4e; Table 1). These two hnRNPs also strongly colocalized with nascent DNA in the RF (Table 1).

Proteins not expected to localize in replication sites

Various proteins were undetectable in any type of DNA synthetic sites, such as Nopp140 (Fig. 2b; Table 1), nuclear phosphoprotein AHNAK (Shtivelman and Bishop 1993), PIKA antigens (Saunders et al. 1991) and centromeric antigens recognized by human autoimmune sera (not shown). The absence of significant labeling in replication sites provides one set of controls for specificity of antibody binding and for the statistical evaluation of colocalization patterns. Furthermore, no immunogold labeling was seen in regions of high protein and nucleic acid density (e.g. in the granular component of nucleoli) with all the antigens listed in Table 1. Moreover, labeling was negligible when the first antibody was omitted or when cross-reactivity with improper secondary antibodies was tested. These data confirm the specificity of immunolabelling procedure used in our experiments.


There are two types of DNA synthetic sites in the cell nucleus

In our experiments we have demonstrated that two ultrastructurally differing types of replication sites can be observed in the nuclei of HeLa cells. One type is electron-dense structures (Hozak et al. 1993, 1995), termed in this paper as RB (replication bodies). Another type corresponds to the sites of extra-bodies labeling that have no typical dense underlying structure, termed here as RF (replication foci). Functional differences between these two replication structures are still unknown. However, while both types of replication sites contain expected set of synthetic and regulatory nuclear proteins, we have also found some significant differences between them.

RB contain little DNA

We first checked whether replication takes place preferentially on the surface (in the outer part) of RB or just uniformly distributed within RB. As already noted, one may suppose that RB are stores of the essential factors required at the appropriate parts of the cell cycle and that DNA synthesis occurs on the surface of these stores (Cook 1991; Hozak et al. 1993). Therefore, RB would contain rather less DNA inside. Our results of DNA immunolabeling and osmium-ammine staining support this suggestion, as both methods revealed little DNA inside RB in comparison to their surrounding. Moreover, if DNA synthesis occurs at the surface, then the incorporated biotin should be concentrated there. We tested this hypothesis and found that nascent DNA (as well as DNA polymerase α and PCNA) is distributed uniformly throughout RB. However, nascent chains are elongated by up to 750 nucleotides during the incubation with biotin-dUTP, thus under the conditions that we used incorporated biotin could move up to 260 nm away from the synthetic site (260 nm is the extended length of 750 bp B-DNA). This picture is in a good agreement with the “replication factory” idea developed earlier (Cook 1991).

The diameter of internal DNA fibrils stained with osmium-ammine was not wider than 10–15 nm, which corresponds to the diameter of a basic chromatin fiber (Baudy and Bram 1978; Rattner and Hamkalo 1979). This data indicate that DNA found inside factories is completely decondensed during replication.

The presence of tested antigens differs in RB and RF

To reveal functional differences between RB and RF, we have screened various proteins for their presence in the mid S-phase RB and RF.

DNA polymerase α is one of the principal replicative DNA polymerases in actively multiplying eukaryotic cells (Salas et al. 1999), while PCNA is a homotrimeric protein that forms a closed ring structure around duplex DNA (Krishna et al. 1994). It serves as a sliding clamp for DNA polymerases, allowing these enzymes to move along their template, while PCNA remains topologically linked to the DNA (Waga and Stillman 1998). The results of our experiments fit very well to the expected patterns: both DNA polymerase α and PCNA demonstrate a strong colocalization with nascent DNA in RF and were found in large amounts in RB thus serving as positive control in our experiments.

The cyclins and associated kinases are controllers of cell-cycle progression (Hunter and Pines 1994; Miller and Cross 2001; Minshull et al. 1990; Sanchez and Dynlacht 2005; Sherr 1994). We performed an ultrastructural analysis to reveal which of them colocalize with DNA replication in mid S-phase. Cyclin A is one of the main S-phase cyclins (Sanchez and Dynlacht 2005). It becomes predominantly nuclear during S-phase and was shown to colocalize with nascent DNA in immunofluorescent experiments (Cardoso et al. 1993). Recently, a cyclin A responsive protein Ciz1 was identified that localizes to replication sites and functions to promote DNA replication after replication complex formation (Coverley et al. 2005). Consistent with all this data, we confirmed ultrastructurally that cyclin A was present in variable amounts in mid S-phase RB and RF.

Cyclin B1 was shown to be accumulated in the cytoplasm during interphase (Bailly et al. 1992; Pines and Hunter 1991; Sherwood et al. 1994) and excluded from the nucleus by a nuclear export signal (Yang et al. 1998). Light microscopy also indicates that it is not colocalized with DNA synthetic sites (Cardoso et al. 1993). In agreement with these data, we demonstrate here that RF are free of cyclin B1. However, in RB it was found in small zones. This is quite an intriguing finding, as recently an S-phase promoting potential of cyclin B1 was demonstrated in Xenopus in vitro system (Moore et al. 2003). Thus one can suggest that RB serve as storing device for cyclin B1 in S phase. Another possibility is that cyclin B1 plays some yet unknown role in regulation of replication specifically in RB. Interestingly, Cdk1, which in complex with cyclin B1 triggers mitosis, is also localized in RB but not in RF. Thus, we cannot exclude a role of RB in regulation of activity of cyclin B/Cdk1 complex during replication. In addition, our finding of cyclin B1 in RB shows the importance of ultrastrucural approach for analysis of protein composition of nuclear structures.

Cyclin D1 appears to be required for the progression into S-phase, as microinjection of anti-cyclin D1 antibodies into normal fibroblasts during G1 prevents cells from entering S-phase (Baldin et al. 1993; Quelle et al. 1993, for review see Sanchez and Dynlacht 2005). Cyclin D1 localizes to the nucleus in the G1 phase and exits the nucleus as cells proceed into S-phase (Alt et al. 2000; Baldin et al. 1993). In agreement with these data we did not detect cyclin D1 neither in RF nor in RB. Cyclin E, an essential regulatory subunit of Cdk2 (Dulic et al. 1992; Koff et al. 1992), plays a central role in coordinating both the onset of S phase and centrosome duplication in multicellular eukaryotes (Reed 1997; Sherr 1996). It has cyclic pattern of mRNA expression, with maximal levels being detected near the G1/S boundary (Koff et al. 1991; Lees et al. 1992; Lew et al. 1991).

Cyclin E has been found to be associated with the transcription factor E2F in a temporally regulated manner (Lees et al. 1992). The cyclin E/E2F complex is detected in the cell nucleus primarily during the G1 phase of the cell cycle and its content decreases (but not disappear) as cells enter S-phase. E2F activity is crucial for the G1/S transition and DNA replication in mammalian cells (for review see Cam and Dynlacht 2003). E2F activity may be directly regulated by cyclin E-Cdk2 (Morris et al. 2000). Recently, it was demonstrated that endoduplication in all multicellular organisms depends on cyclin E-mediated minichromosome maintenance (MCM) proteins loading onto pre-replicative chromatin (reviewed in Sanchez and Dynlacht 2005). At the same time since a mammalian Mcm 4/6/7 complex exhibits DNA helicase activity (Ishimi 1997), MCM proteins may be involved in DNA unwinding at replication forks. Our finding of cyclin E localization both in mid S-phase RB and RF is thus in agreement with previously reported data and additionally suggest a role of cyclin E in regulation of DNA replication itself.

Cdk1 (cdc2 kinase) is the principal mammalian G2-M cyclin-dependent kinase (Sanchez and Dynlacht 2005). In immunofluorescent labeling experiments it was not detected within replication sites (Cardoso et al. 1993). We also did not detect cdk1 in RF, however, it shows very high labeling densities in RB. Recently it was reported that Cdk1/cyclin A regulates mammalian ORC (origin recognition complex) activity by preventing it’s largest subunit (Orc1) from binding to chromatin during mitosis (Li et al. 2004). As ORC activity is inhibited during the G1 to S-phase transition until mitosis is complete (DePamphilis 2003), we can suggest that Cdk1/cyclin A complex is involved in ORC inhibition not only in mitosis but also in S phase. In this paper we show the presence of both cyclin A and Cdk1 in RB. This supports the hypothesis that RB serve as storage sites for regulatory factors.

Cdk 2 is one of the central gene activators that functions during G1 and S phase (for review see Sanchez and Dynlacht 2005). In this paper we report that cdk 2 is present in variable amounts in mid S-phase RB and RF. However, a direct involvement of Cdk2 in replication has not been established and it may be that localization to sites of replication is related to Cdk2 role in transcriptional activation (Hassan et al. 1994).

The human ubiquitin-conjugating enzyme Cdc34 controls cellular proliferation through regulation of p27Kip1 protein levels (Butz et al. 2005). p27Kip1 is an inhibitor of cyclin E and A-dependent kinases (Sanchez and Dynlacht 2005). Our data show no colocalization of Cdc34 with both types of DNA replication sites.

p21 was originally identified as an inhibitor of cdk activity in a quaternary complex that also included cyclin D, cdk4 and PCNA (Xiong et al. 1993). Later it was also demonstrated that p21 can inhibit DNA replication in vitro by directly binding to PCNA (Chen et al. 1995). Further it was shown that p21 might inhibit cell cycle progression to G2 by inhibition of PCNA function in vivo (Cayrol et al. 1998). Our data show, that p21 is present both in replication factories and replication foci in agreement with the role of p21 in regulation of DNA replication in vivo.

Intranuclear lamins have also been correlated with sites of DNA synthesis. B-type lamins (but not A/C) localize to mid-late S-phase replication sites in mouse 3T3 cells (Moir et al. 1994), whereas lamin A/C was found at the earliest sites of DNA synthesis in human WI38 cells (primary diploid fibroblasts; Kennedy et al. 2000). Biochemical support for lamin function comes from the immunodepletion of lamin B3 (XLB3) from interphase Xenopus egg extracts. This immunodepletion impairs replication in reconstituted nuclei (Meier et al. 1991; Newport et al. 1990), although the effects could be indirect, reflecting destabilization of the nuclear envelope or nuclear transport mechanisms. The ability to replicate then can be rescued by restoring of lamin B3 to the depleted extract (Goldberg et al. 1995). Other compelling experiments are those in which dominant negative forms of lamin A (ΔN lamin A, which lacks the first 33 aminoacids of human lamin A) were added after nuclear assembly. These did not impair the initiation of DNA replication, but instead blocked elongation, while nuclear envelope integrity and protein import were unaffected (Ellis et al. 1997; Moir et al. 2000). When DNA synthesis was blocked by microinjection of ΔN lamin A, elongation factors such as PCNA and RFC became dispersed, suggesting that lamins may help stabilize replication foci (Spann et al. 1997). Structural role of lamins in organization of replication is also supported by our earlier work, where we demonstrated that “replication factories” corresponding to RB are attached to an internal lamin A-containing nucleoskeleton (Hozak et al. 1993, 1995; Hozak and Cook 1994). In this paper we show the presence of lamin B1 in RF of mid S-phase HeLa cells, while there is no lamin A/C in both types of replication sites. This data is in agreement with the results of Moir and coworkers (1994) and support the involvement of B-type lamins in replication itself. Presumably, lamin B1 participates in organization of replication machinery. Lamin A/C, however, may take part in stabilization of replication structures at higher levels and thus be not localized directly within replication sites. Interestingly, the complete absence of lamin A/C clearly does not impair replication during early chicken and mouse development (Lehner et al. 1987; Rober et al. 1989) suggesting that the requirement for lamin A/C for replication maintenance may be cell type or differentiation stage specific. Lamin B1 was not detected in RB in our experiments. We cannot exclude that it was inaccessible to antibodies within RB. Another possibility is that in RB, the structural organization of replication is maintained by a different set of proteins.

The PML protein is a human growth suppressor concentrated in 10 to 20 nuclear bodies per nucleus (PML nuclear bodies). Seven groups of PML protein isoforms were already described (for a review see Jensen et al. 2001). All of the PML isoforms contain similar N-terminal region comprising the RBCC/TRIM motif (which is essential for PML nuclear body assembly in vivo as well as for PML growth suppressor, apoptotic and antiviral activities), but differ either in the central region or in the C-terminal region, due to alternative splicing. The functions of the PML splice variants are not known, however, alternative splicing could add new functional domains to the protein and/or may be an important mechanism for generating diverse PML-binding interface for a variety of factors. For example interaction of PML bodies with viral DNA may be necessary for optimal viral replication (Maul 1998). Recently a role of PML and PML nuclear bodies in post-replication DNA processing, and linkage of PML nuclear bodies to sites of viral DNA synthesis due to a role of these structures in DNA metabolism was demonstrated (Jul-Larsen et al. 2004). Moreover, a subset of PML bodies was found in association with mechanism of alternative lengthening of telomeres in mammalian cells (Henson et al. 2002). Finally it was demonstrated that a large fraction of PML bodies (50–80%) is closely associated with DNA replication domains but exclusively during middle-late S-phase (Grande et al. 1996). PML bodies were then suggested to facilitate these functions by sequestering and releasing proteins, localizing proteins to sites of action and facilitating interactions between other proteins. Our experiments demonstrated that only RF colocalize with the PML protein. We suggest that the sequestering and facilitating mechanisms in RB are provided by their own structure, while RF may need association with PML. Furthermore, direct and indirect evidence supports the hypothesis that PML bodies interact with specific genes or genomic loci (reviewed in Ching et al. 2005). Thus, replication of such genes could also take place in connection with PML bodies and we can suggest that they are replicated in RF but not in RB.

Heterogeneous nuclear ribonucleoproteins (hnRNPs) are predominantly nuclear RNA-binding proteins that form complexes with RNA polymerase II transcripts (for review see Krecic and Swanson 1999). Immunopurification of hnRNP complexes and two-dimensional electrophoresis demonstrated that hnRNA exists in the nucleus in association with more than 20 proteins designated A1 through U. The most abundant proteins A1, A2, B1, B2, C1 and C2 are referred to as “core” hnRNP proteins (Dreyfuss et al. 1993). hnRNP A1 participates in splicing as exonic repressor and functions in mRNA transport and telomere biogenesis, while hnRNP C1/C2, in transcript packaging, splicing, nuclear retention and mRNA stability (Krecic and Swanson 1999). Recently it was also demonstrated that hnRNP A1 and UP1 protect mammalian telomeric repeats and modulate telomere replication in vitro (Dallaire et al. 2000).

Sm proteins have been shown to be key components of the ribonucleoprotein (RNP) assemblies that are required for high-fidelity cellular RNA processing, including rRNA and tRNA processing, mRNA decapping and decay, and intron splicing in pre-mRNA (for review see Yu et al. 1999). Some of them are main components of the spliceosomal small nuclear ribonucleoproteins (snRNPs) and form cyclic hetero- or homo-oligomers, which preferentially bind uridine-rich, single stranded snRNA.

Coilin is highly enriched in Cajal bodies (CBs; Andrade et al. 1991; Raska et al. 1991) and thus can be used as their marker. Earlier studies suggested that coilin is involved in some step in the transport of small nuclear ribonucleoproteins (snRNPs) to the CBs in the nucleus (Bauer and Gall 1997; Bellini and Gall 1998). More recent data from coilin knockout mice support this view (Jady et al. 2003; Tucker et al. 2001), as does biochemical evidence that coilin can associate with the survival of motor neurons protein (Hebert et al. 2001, 2002), which is part of the machinery for assembly of snRNPs (Fischer et al. 1997; Meister et al. 2002).

La protein binds to all nascent transcripts made by RNA polymerase III as well as to certain small RNAs synthesized by other RNA polymerases (for review see Wolin and Cedervall 2002). Binding by the La protein protects the 3′ ends of these RNAs from exonucleases. This La-mediated stabilization is required for the normal pathway of pre-tRNA maturation, facilitates assembly of small RNAs into functional RNA-protein complexes, and contributes to nuclear retention of certain small RNAs. La was also suggested as a transcription factor for RNA polymerase III (Wolin and Cedervall 2002).

Although no direct function in DNA replication was demonstrated for the above described proteins involved in RNA metabolism, we have shown high colocalization for all these proteins with replication sites of both types. However, as it was demonstrated that replication and transcription colocalize (Hassan et al. 1994) and that splicing starts before finishing the transcription of nascent RNA (Wuarin and Schibler 1994), this would explain the significant colocalization of all these splicing factors with replication sites. This would suggest the existence of functional domains in the cell nucleus where different processes connected by substrate or participating factors can be spatially arranged in a very economical way.

Different antibodies to proteins that would not be expected to localize to DNA replication sites (such as the nucleolar protein Nopp 140) were used as controls. No labeling in either RB or RF was obtained with these antibodies, indicating the validity of the immuno-labeling approach used.

Relationship between RBs and RFs

We show here that two types of DNA synthetic sites can be distinguished at the ultrastructural level: (1) electron-dense structures attached to nucleoskeleton—replication bodies, and (2) focal regions of incorporated precursors with no distinct underlying structure—replication foci. We have found that they are complex protein structures containing (a), proteins involved in DNA replication (DNA polymerase α, PCNA), (b), regulators of the cell cycle (cyclin A, Cdk2), and (c), RNA processing components like Sm and SS-B/La auto antigens, p80-coilin, hnRNPs A1 and C1/C2. In contrast to RF, RB have a distinct microarchitecture, with certain components concentrated in specific regions. For example, cyclin A was present in variable amounts in mid S-phase factories in small zones as well as p80-coilin, which also had a characteristic zonal distribution. This observation allows us to suggest the existence of specific functional zones inside the RB. On the other hand, both types of replication sites contain a similar set of major synthetic and regulatory components, which indicates that RF are true replication factories. However, at least four principal regulatory and structural proteins (Cdk1, cyclin B1, PML and lamin B1) differ in their presence in RB and RF.

This data suggest possible differences in genome regions replicated by these two types of replication factories and leads us to consider how the two classes of structure can be related. Three obvious possibilities might be considered:
  1. 1.

    RF are precursors of RB: This is probably not the case as our initial study (Hozak et al. 1993, 1994) showed that RF are only seen in sections that already contain labeled RB. If RF represented an early stage in RB assembly cells with many RF but no active RB should be seen early in S-phase.

  2. 2.

    RB are precursors of RF: RB are complex structures that are known to replicate clusters of replication units; they are only active for about 1 h, after which time some replication forks will fuse with forks from adjacent replicons. During termination, the replication machinery will be lost from the template, leading to decay of the replication factory (Hozak et al. 1994). RF could reflect the residual active sites that persist during the decay phase of the RB life cycle. However, the kinetics of appearance of RF and decay of RB do not fully support this view (Hozak et al. 1994).

  3. 3.

    RB and RF are unrelated, except for the common purpose of DNA synthesis. This could reflect differences in the structure of the genome regions that are replicated by the two classes of replication factory. Notably, the structure of RB is consistent with their role in replicating large, structurally related regions of the genome whereas RF might commonly appear to replicate smaller regions, which for whatever reason are surrounded by DNA that is replicated either earlier or later during S phase.

We conclude that replication in HeLa cells is performed in replication factories, with two forms that can be distinguished by electron microscopy. The relationship between these two forms has not been demonstrated unambiguously, though evidence presented here, together with earlier work (Hozak et al. 1993, 1994), has given a detailed view of the properties and microarchitecture of these replication sites. Based on the structure of these sites we propose that both classes are genuine replication factories. RB represent the major replication structures in this cell type and are responsible for replication of the bulk of DNA, which is organized in such a way that clusters of replicons are replicated together in each RB. RF, in contrast, are smaller, less complex structures, which appear to perform replication from regions of the genome that are organized in such a way that their replication involves isolated replicons or replicon clusters with very few linked replicons.


We thank Prof. Derenzini for advice on staining with osmium-ammine, Drs. M. Carmo-Fonseca, R. van Driel, I. Todorov, G. Dreyfuss, G. Blobel, Y. Raymond, and R.L. Ochs for kindly supplying antibodies, and Dr. J. Janáček for his help in the statistical evaluation of the results of colocalization experiments, and Dr. V. V. Philimonenko for critical reading of the manuscript. We thank BBSRC (Grant No. 36/G15592), the Grant Agency of the Academy of Sciences of the Czech Republic (Grant No. IAA5039202) and Ministry of Education, Youth and Sports (Grant No. LC 545). This work was also supported by the Institutional Grant No. AV0Z5039906.

Copyright information

© Springer-Verlag 2005