Analysis of replication factories in human cells by super-resolution light microscopy
© Cseresnyes et al; licensee BioMed Central Ltd. 2009
Received: 5 November 2009
Accepted: 16 December 2009
Published: 16 December 2009
Skip to main content
© Cseresnyes et al; licensee BioMed Central Ltd. 2009
Received: 5 November 2009
Accepted: 16 December 2009
Published: 16 December 2009
DNA replication in human cells is performed in discrete sub-nuclear locations known as replication foci or factories. These factories form in the nucleus during S phase and are sites of DNA synthesis and high local concentrations of enzymes required for chromatin replication. Why these structures are required, and how they are organised internally has yet to be identified. It has been difficult to analyse the structure of these factories as they are small in size and thus below the resolution limit of the standard confocal microscope. We have used stimulated emission depletion (STED) microscopy, which improves on the resolving power of the confocal microscope, to probe the structure of these factories at sub-diffraction limit resolution.
Using immunofluorescent imaging of PCNA (proliferating cell nuclear antigen) and RPA (replication protein A) we show that factories are smaller in size (approximately 150 nm diameter), and greater in number (up to 1400 in an early S- phase nucleus), than is determined by confocal imaging. The replication inhibitor hydroxyurea caused an approximately 40% reduction in number and a 30% increase in diameter of replication factories, changes that were not clearly identified by standard confocal imaging.
These measurements for replication factory size now approach the dimensions suggested by electron microscopy. This agreement between these two methods, that use very different sample preparation and imaging conditions, suggests that we have arrived at a true measurement for the size of these structures. The number of individual factories present in a single nucleus that we measure using this system is greater than has been previously reported. This analysis therefore suggests that each replication factory contains fewer active replication forks than previously envisaged.
The biochemical steps required to accurately replicate the genome are well understood . In recent years we have also gained an understanding of how epigenetic information may be transmitted, including the copying of DNA methylation states and the histone code from parental to daughter DNA strands during replication [2, 3]. All of these processes are essential if chromosomal replication is to generate a faithful copy of the parental state, crucial for genetic and epigenetic stability and prevention of carcinogenesis [4, 5].
In mammalian cells chromosomal replication occurs at discrete nuclear sites, known as replication foci or replication factories [6, 7]. These structures form transiently in the nucleus and are the sites of DNA synthesis and high local concentrations of replication proteins [8–15]. Although not permanent, each factory persists in one location for a significant fraction of S phase, during which time replication proteins dynamically and independently move in and out [16–20]. The observed intra-nuclear organisation of replication factory patterns changes with progression through S phase, with early/mid S phase nuclei having a large number of small and evenly distributed factories that move to the nucleolar and cellular periphery as replication progresses to later stages [10, 21, 22, 16]. The internal organisation of these replication factories and the possible protein-protein interactions dictating and controlling their formation, persistence and disassembly remain unclear.
Standard light microscopy is limited in resolution by diffraction and thus, as described by E. Abbe in 1873, objects closer together than approximately one half the wavelength of the light used cannot be resolved . Even on a confocal microscope this prevents the resolution of objects closer than ~ 200 nm . Replication factories are reported to be between 100 nm and 1 μ m in diameter [11, 25, 16] and thus light microscopes are clearly limited in their capacity to probe fine structures within these foci. Electron microscopy has also been used to investigate replication factory structures at higher resolution [13, 26, 27], and these studies give estimates for replication factory size between 100 and 400 nm. Here, we have applied stimulated emission depletion (STED) microscopy to analyse replication factories in human fibroblasts. STED microscopy is a technical advance that effectively breaks the resolution limit of a confocal microscope by using a donut-shaped depletion beam to deplete fluorescence from all but the very centre of the excitation spot [28, 24, 29]. This generates an effective excitation region smaller than can be obtained by diffraction limited optics, resulting in better definition of objects closer than Abbe's theoretical resolution limit. The advantage of this technique is that it gives high resolution analysis of immunolabelled specimens using the relatively mild sample preparation conditions of the confocal microscope. This study provides a new estimate for the maximum size of a replication factory and also for the minimum number of these factories that form at any one point in an unperturbed S phase. We have also analysed the effect that inhibition of DNA replication by hydroxyurea (HU) has on the size and number of these factories.
In order to visualise replication factories in human cells we selected antibodies against two key components, PCNA and RPA. RPA is a heterotrimeric single stranded DNA (ssDNA)-binding protein that associates with the template strands produced at replication forks by the action of the replicative helicase. It is important for strand stability and for recruiting other replication components to the advancing replication fork . PCNA is a ring-shaped sliding clamp protein which is loaded onto template DNA at replication forks, where it acts as a platform for the recruitment of multiple enzymes required for replication .
Replication factory numbers determined by confocal and STED microscopy.
1.2 - 2.1
2.1 - 5.4
1.0 - 4.1
3.0 - 7.6
1.4 - 1.8
1.5 - 3.5
1.2 - 2.6
1.9 - 3.9
This analysis was also used to determine the apparent size of the replication foci in each case. As above we performed this analysis using a garbage volume of 5 voxels, which resulted in clear object definition without the unwanted elimination of too many small objects. The STED images have an average replication focus size (determined as the maximum axial width of an object) of 150 nm for PCNA and 160 nm for RPA (figure 4C), an approximately 40% reduction when compared to the same cells imaged using confocal mode (object sizes 270 nm and 240 nm for PCNA and RPA). Thus STED imaging has altered our perceptions of the nature of replication factories, one factory is apparently much smaller than previously envisaged from confocal studies - indeed these sizes approach the measured size determined from electron microscopy studies [13, 26].
HU inhibits ribonucelotide reductase and depletes the pools of dNTPs required for DNA synthesis . Initially the replicative helicase is not inhibited by this, and ssDNA is generated as the helicase unwinds DNA ahead of stalled polymerases . It is therefore not unexpected that the size of the replication structures including RPA increases after HU treatment, as more RPA can be recruited to this excess ssDNA. For PCNA the increase in size is perhaps more unexpected, PCNA is loaded at primer template junctions on the leading and lagging strands, and in the absence of dNTPs no further primers will be synthesised. It may be that PCNA disengages from the primer template junction in the absence of processive synthesis and slides over the synthesised or template strands. This would cause enlargement of the PCNA bound region, either concomitant with loading of extra PCNA, or even in the absence of this. HU is unlikely to completely deplete dNTP pools and some DNA synthesis may well continue in the presence of this drug. Interestingly, both classic papers and recent studies have suggested that the replication inhibition caused by nucleotide depletion may result in the activation of origins that would otherwise remain dormant and be passively replicated [36–39]. This might be necessary in order to ensure that the genome can be completely replicated in the presence of a high rate of fork stalling events . If such origin initiation events occur within a pre-existing replication factory, this could account for the increase in size of the replication foci containing both RPA and PCNA after HU treatment in these experiments. Alternatively excess ssDNA may cause the replication factories to enlarge in some other unknown manner, perhaps by altering connections between replicative polymerases and helicases and an underlying nuclear matrix scaffold. The fact that the number of observed factories decreases can be explained if the checkpoint that senses depleted nucleotide pools is more sensitive to small fluctuations in the dNTP pool than the active polymerases. In this case the initiation of new origins might be prevented while previously fired origins continue to terminate, resulting in fewer factories after HU treatment.
We have shown that STED microscopy can be applied to improve the imaging of small nuclear replication factories. The resolution of the STED system used in this study was approximately 75 nm (data not shown), so objects closer together than this still cannot be resolved. Nevertheless, in this example STED technology does bridge the gap between the resolution of the confocal and electron microscopes. The size that we measured here for a replication factory (~150 nm) approaches the measured diameter of replication structures visualised using electron microscopy, and for objects that truly are in the 80 - 150 nm range, STED will give an accurate picture of subcellular organisation. Of course, objects smaller than this or those that are densely clustered within the cell will still require the resolving power of the electron microscope for true visualisation. The theoretical STED resolution limit is only dependent on the power of the depletion beam and thus the photostability of the labelling dye under depletion conditions. It is therefore likely that further xy resolution improvements will be available in the relatively near future as new dyes become available [41–43]. The next steps for STED microscopy will be the development of multi-colour imaging, and resolution improvement along the Z axis. Both of these aims are likely to need significant technology development before they can be implemented on a commercial system.
In our experiments the PCNA-containing replication factories in an unperturbed early S phase have a mean size of less than 160 nm, significantly smaller than previously suggested from studies using light microscopes. We also suggest that there can be at least 1200 such factories coexisting in a normal early S phase MRC5 nucleus. Previous estimates that there are ~150 factories at any one time, and ~3000 active forks, gave rise to the supposition that one factory must contain ~20 forks . A more recent study challenged these numbers , and our data is in support of a larger number of replication factories than was even suggested in that detailed study. If in fact the number of factories is closer to 1200 as we determine here, there would only be 2-3 forks per factory. The idea that some factories might contain so few replication forks has been previously discussed , and our data support this.
Is the concept of a replication factory still valid? Even with the enhanced resolution of STED, localised concentrations of replication factors are clearly visualised, so currently our answer is yes. However our data suggests that each factory in early S phase is likely to contain fewer replication forks than is generally accepted. This would mean that the amount of DNA synthesised in a coordinated fashion within a single factory may be limited to that from only one or two origins. The observed factory enlargement after hydroxyurea treatment may be due to the formation of extensive ssDNA at stalled replication forks , or due to the activation and recruitment into factories of otherwise dormant nearby origins .
There are of course many questions still to be answered before we fully understand how DNA and chromatin is accurately copied in the replication factories. Are the factories attached to structures in the nucleus? Does the DNA move? How are diverse replication events coordinated? The answering of these questions will require the formation of testable hypotheses, and the chances of developing such are increased once the description of the problem is accurate. In this report we have utilised STED microscopy to give a more precise characterisation of the nature of replication factories in human cells. This technology can also be applied to other small nuclear structures such as PML and Cajal bodies, transcription factories and focal structures formed at break sites during DNA repair. Future advances in STED, and other super-resolution microscopy technologies are likely to improve resolution still further. This will take us closer to the ultimate goal of visualising molecular machines at work in a living cell.
STED microscopy dramatically improves the visualisation of replication factories under the relatively mild sample preparation conditions of the confocal microscope. This enables us to provide a firm estimate of ~150 nm for the size of RPA- and PCNA-containing replication factories in early S phase cells. This is much smaller than has been previously reported from light microscopic analysis and is very similar to the size reported from visualisation of these structures in the electron microscope. The agreement between these two very different techniques suggests that we have arrived at a correct measurement. The number of these structures is also much greater than has previously been determined (up to 1400 in a single nucleus). This suggests that there may be as few as 2 replication forks per factory. Replication inhibition by the depletion of nucleotide pools causes an increase in size and a decrease in number of replication factories.
The cells used were MRC5 SV40 transformed human fibroblasts (a gift from A. Lehmann, University of Sussex), grown at 37 °C with 5% CO2 in MEM with 10% fetal calf serum (Gibco) and L-glutamine and penicillin/streptomycin (PAA). For indirect immunofluorescence cells were grown on 170 μ m thick glass coverslips until 50% visual confluence. Cells were washed in CSK (10 mM PIPES pH 7.0, 100 mM NaCl, 300 mM Sucrose, 3 mM MgCl2) then the soluble proteins were removed by incubation in CSK with 0.5% triton X-100 for 3 minutes followed by a wash in CSK and fixation in 2% freshly dissolved formaldehyde in PBS. Non-specific antibody binding sites were blocked by incubation in 5% BSA in PBS with 0.1% tween20 (blocking buffer) for 30 minutes, primary antibody incubation was for 1 hour at room temperature with antibodies (anti-RPA34 monoclonal Abcam ab16855, anti-PCNA polyclonal Abcam ab18197) diluted 1/500 in blocking buffer. After three washes secondary antibodies coupled either to Alexa Fluor 488 (Molecular Probes) or Atto 647N (Sigma) (all at 1/500 dilution) were added for 1 hour in blocking buffer. After 5 washes of 5 minutes each in PBS with 0.1% tween20 the coverslips were mounted in aquapolymount.
In some cases cells were pulse labelled for ten minutes with 40 μ M EdU (5-ethynyl-2-deoxyuridine) (Invitrogen), and after triton extraction and fixation as described above, sites of EdU incorporation into DNA were visualised by copper catalysed-click chemistry  with an amide derivative of Alexa Fluor 488 according to the manufacturer's protocol. Subsequent immunofluorescence was as above.
Where used hydroxyurea treatment was for 12 hours at 2 mM followed by immediate extraction and fixation as above.
Throughout this study we selected cells that showed the characteristic staining found in early S-phase, with many evenly sized nuclear replication factory distributed evenly throughout the nucleoplasm. All images were acquired on a Leica TCS STED equipped with an inverted microscope (DMI 6000, Leica) and a 100 × STED objective (HCX PL APO 100 × 1.4 oil STED, Leica). Confocal images were acquired using the 488 nm line for the excitation of Alexa 488 and a pulsed 635 nm laser diode (PicoQuant) to excite Atto 647N. Alexa 488 was detected using PMT 2 of the spectral detection unit with the detection range set to 495 - 550 nm and Atto 647N was detected on APD 2 equipped with a 685/40 (Semrock Bright Line, Semrock) nm bandpass filter. Imaging speed was at 400 Hz using 7 × line averaging and the pinhole was set to 0.5 Airy units. At zoom 11 and a format of 512 × 512 the resulting pixel size was 27.6 nm. Z-stacks were run at a step-size of 80 nm using the galvanometric-driven fine focussing stage of the system.
For STED microscopy all conditions were identical, but additionally the depletion laser was activated. For the stimulated emission depletion of Atto 647N the pulsed Ti:Sa IR laser (Mai Tai HP, Spectra Physics) was tuned to 750 nm and the AOM set to 100%. We calculated the xy resolution limit of our STED system to be 75 nm by imaging 20 nm crimson beads and measuring the full width half maximum of the obtained images using Leica's LAS AF quantification tool.
Confocal and STED image stacks were deconvolved using SVI's Huygens Professional package. We used the CMLE (Classic Maximum Likelihood Estimation) method of this software. The CMLE method was applied using SVI's "Confocal" optical parameters set, whereas the sampling intervals were set manually to the actual experimental values (27.6 nm for X and Y pixel sizes, 80 nm for Z step size), together with the refractive indices (1.51 for oil immersion objective and medium) and excitation/emission wavelengths. The complete parameter set was saved as a template and applied for each data set. We also performed deconvolution using a calculated PSF for the STED system (provided by Leica Microsystems). The data analysed in this way did not vary significantly from the deconvolution presented here.
Object counting was performed on the 3-D image stacks using Huygens Object Analyzer. Each data set was analysed with garbage volumes set at 1, 3, 5, 10 and 50. For each garbage volume, the fluorescence threshold was set to the value that resulted in the maximum number of objects identified. The seed value was always set at 0%, thus not setting a ceiling on voxel intensity values. Objects were counted within a pancake-shaped region of interest (ROI), where the top and bottom planes of the ROI were set to the highest and lowest Z positions of non-zero voxels of all identified objects, whereas the XY outline of the ROI was drawn manually in order to exclude stray objects from the analysis. The number of objects was divided by the volume of the ROI before calculating data average and scatter. The objects were also characterised by the axial width, as defined in Huygens: the largest width in an axial direction perpendicular to the length axis. Axial width values were only included for objects inside the ROI. The resulting datasets were analysed in Excel (Microsoft Office 2007): average values were calculated with the AVERAGE function, whereas the data scatter was characterised with the AVEDEV function.
Colocalisation analysis of doubly stained cells was performed after background subtraction using the "intensity correlation analysis" plugin for Image J from the MacMaster Biophotonics facility (full details and download available from http://www.macbiophotonics.ca/index.htm).
We thank Ross Chapman, Torsten Krude and members of the Green lab for critical reading of the manuscript. We also thank Leica microsystems, for providing access to the STED microscope, Scientific Volume Imaging (SVI), and personally Ms. Gitta Hamel, managing director of SVI, for providing us with a fully functional test license of Huygens Pro. This work was supported by Cancer Research UK, Grant number C24125/A8307 to CMG. ZC is supported by the Wellcome Trust (079204/Z/06/Z).
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.