Dynamics of Telomeres and Promyelocytic Leukemia Nuclear Bodies in a Telomerase-negative Human Cell Line

Abstract

Telomerase-negative tumor cells maintain their telomeres via an alternative lengthening of telomeres (ALT) mechanism. This process involves the association of telomeres with promyelocytic leukemia nuclear bodies (PML-NBs). Here, the mobility of both telomeres and PML-NBs as well as their interactions were studied in human U2OS osteosarcoma cells, in which the ALT pathway is active. A U2OS cell line was constructed that had lac operator repeats stably integrated adjacent to the telomeres of chromosomes 6q, 11p, and 12q. By fluorescence microscopy of autofluorescent LacI repressor bound to the lacO arrays the telomere mobility during interphase was traced and correlated with the telomere repeat length. A confined diffusion model was derived that describes telomere dynamics in the nucleus on the time scale from seconds to hours. Two telomere groups were identified that differed with respect to the nuclear space accessible to them. Furthermore, translocations of PML-NBs relative to telomeres and their complexes with telomeres were evaluated. Based on these studies, a model is proposed in which the shortening of telomeres results in an increased mobility that could facilitate the formation of complexes between telomeres and PML-NBs.

INTRODUCTION

Telomeres are specialized chromatin structures at the end of linear chromosomes, in which repetitive DNA sequences (5′-TTAGGG-3′ in vertebrates) associate into a nucleoprotein complex (de Lange et al., 2006). This complex—the “telosome”—protects the chromosome ends from degradation and genomic rearrangement and includes the “shelterin” proteins TRF1, TRF2, and POT1 that directly recognize the TTAGGG repeat sequence (de Lange, 2005; Bertuch and Lundblad, 2006; Croy and Wuttke, 2006; Blasco, 2007). Because of incomplete DNA synthesis at the chromosome ends, 50–200 base pairs of telomeric DNA are lost during each replication cycle (Harley et al., 1990; Martens et al., 2000). After ∼60–80 cell divisions telomere repeats are shortened from a typical initial length of 10–15 kb in human cells to ∼5 kb and below, which triggers cell senescence or apoptosis (Harley et al., 1990; Martens et al., 2000; Blasco, 2007). Accordingly, tumor cells need to compensate the loss of their telomere repeats in order to sustain an unlimited proliferative potential. In most cases this is accomplished by reactivating telomerase, a reverse transcriptase that synthesizes telomeric repeats at the chromosome ends (Greider and Blackburn, 1985; Chan and Blackburn, 2004; Collado et al., 2007; Johnson and Broccoli, 2007). However, a fraction of ∼10–15% of tumors is able to maintain their telomeres in the absence of telomerase activity. This process has been designated as alternative lengthening of telomeres (ALT; Bryan et al., 1995; Neumann and Reddel, 2006; Johnson and Broccoli, 2007). In yeast and mammals, it has been shown that the ALT mechanism involves homologous recombination events between telomere repeats (Dunham et al., 2000; Kass-Eisler and Greider, 2000; Lundblad, 2002; Muntoni and Reddel, 2005). It is characterized by a number of features that include 1) a highly heterogeneous telomere length distribution ranging from <3 to >50 kb (Bryan et al., 1995; Perrem et al., 2001; Scheel et al., 2001; Neumann and Reddel, 2006), 2) the presence of circular and linear extrachromosomal telomeric repeat DNA (ECTR-DNA) of up to ∼60 kb in size (reviewed in Neumann and Reddel, 2006), and 3) a particular class of promyelocytic leukemia nuclear bodies (PML-NBs) that form specific complexes with telomeres termed ALT-associated PML bodies (APBs; Yeager et al., 1999).

PML-NBs are mobile nuclear organelles that have been implicated in multiple biological functions including transcription, DNA replication and repair, apoptosis, cellular proliferation, senescence, and tumor suppression (Dellaire and Bazett-Jones, 2004; Takahashi et al., 2004; Bernardi and Pandolfi, 2007). The DNA recombination and DNA repair–related activities localized in the PML body components of the APB complex are regarded as essential for telomere lengthening via the ALT mechanism (Yeager et al., 1999; Nabetani et al., 2004; Neumann and Reddel, 2006). Several genes potentially involved in the ALT pathway were recently identified using an RNAi-based approach (Jiang et al., 2007). These are related to PML body formation (PML) and DNA repair (MRE 11, Rad50) as well as telomere metabolism (TRF1 and TRF2). Furthermore, it has been demonstrated that the SMC5/6 (structural maintenance of chromosomes) complex is present in APBs and that the SUMO (small ubiquitin-like modifiers) ligase activity of the MMS21 component of SMC5/6 is required for APB formation (Potts and Yu, 2007).

Both telomeres and PML-NBs can translocate in the nucleus, and the findings mentioned above point to a crucial role of telomere–telomere and telomere–PML-NB interactions for the ALT mechanism. Thus, it is essential to determine the factors that govern their mobility in cell lines that maintain their telomeres via the ALT pathway. Previously, we have proposed that the dynamic chromatin organization of the interphase nucleus is responsible for the confined translocation of mobile nuclear subcompartments like PML bodies and their targeting to certain regions of the nucleus (Görisch et al., 2004). The apparent mobility of nuclear particles and chromatin loci as described by their diffusion coefficient and accessible space is highly dependent on the observation period because of the multiscale organization of chromatin as discussed in detail elsewhere (Wachsmuth et al., 2008). Accordingly, chromatin loci show multiscale movements as inferred from the analysis of different time regimes (Marshall et al., 1997; Vazquez et al., 2001; Chubb et al., 2002; Levi et al., 2005; Chuang and Belmont, 2007).

In general, it appears that both the translocations of chromosomal loci as well as those of nuclear bodies are confined to 200–300 nm, with apparent diffusion coefficients in the range of 10⁻⁴ to 10⁻⁵ μm² s⁻¹ on the minute-to-hour time scale (Wachsmuth et al., 2008). These relatively slow and confined diffusive movements would make interactions between telomeres and PML-NBs a relatively slow process. Accordingly, an increase of their ability to translocate in the nucleus could significantly promote interactions. Previously, a fluorescently tagged peptide nucleic acid (PNA) probe directed against telomere repeats was used to gain insight into the in vivo mobility of telomeres and their interaction with PML bodies in the osteosarcoma U2OS cell line that maintains telomeres via the ALT mechanisms (Molenaar et al., 2003). The vast majority of the visible telomeres exhibited constrained movements within nuclear regions of ∼0.5-μm radius, whereas a small fraction of telomeres exhibited an increased mobility. In addition, some telomeres displayed transient interactions indicative of the formation of dynamic telomere clustering that might be relevant for the recombination events associated with the ALT mechanism. Although the labeling of telomere repeats with a fluorescent PNA has the advantage of visualizing a large fraction of telomeres in the U2OS cell line, this approach also has some limitations. First, very short telomeres in ALT cells, which are particularly relevant for inducing cell senescence or apoptosis, can hardly be detected because the fluorescent signal is not sufficient for tracing them (Perrem et al., 2001; Scheel et al., 2001). Second, the fluorescent labels cannot be assigned to a specific chromosome because all repeats of sufficient length are visualized. Third, signals arising from telomeres cannot be distinguished from those that originate from hybridization of the probe to ECTR-DNA. The same restriction would apply to the in vivo labeling of telomere repeats via a bound autofluorescent construct of a shelterin protein such as TRF1 or TRF2 (de Lange, 2005). An alternative approach to tag chromosomal loci makes use of the high-affinity binding of bacterial LacI repressor protein fused with autofluorescent green fluorescent protein (GFP-LacI) to repeats of lac operator (lacO) sequences stably integrated into the genome (Robinett et al., 1996; Straight et al., 1996). This lacO/GFP-LacI tagging of genomic loci has been applied in a number of studies of chromatin structure and dynamics in living cells as reviewed recently (Chuang and Belmont, 2007). Here, the mobility of telomeres and their interaction with PML-NBs in U2OS cells was monitored by fluorescence microscopy after labeling of integrated lacO repeats with GFP-LacI/mRFP1-LacI. The experiments yielded a comprehensive description of telomere mobility on different time scales and identified differences in the accessible nuclear space. In conjunction with studies of PML-NBs a model is proposed in which the shortening of telomeres results in an increased mobility that could facilitate the formation of APBs.

MATERIALS AND METHODS

Cell Culture and Generation of Clones with Stable lacO Array Integrations

Cells of the human U2OS osteosarcoma cell line (HTB-96) were grown in DMEM (Invitrogen, Carlsbad, CA) containing 10% FCS (PAA Laboratories, Linz, Austria). Cell lines with stably integrated lacO arrays were constructed by cotransfection of the pLAU 43 vector containing ∼240 lacO repeats, provided by David Sherratt (University of Oxford, Oxford, United Kingdom) (Lau et al., 2003), together with the neomycin resistance-coding vector pcDNA3.1 (Invitrogen). Selection was conducted with 750 μg/ml G418, and stable cell clones were picked and cultured for 1 wk in 96-well plates. The surviving cell clones were then screened for lacO integrations by transfection with GFP-LacI, which was generated from an mRFP1-LacI construct provided by Cristina Cardoso (University of Darmstadt, Darmstadt, Germany). For in vivo imaging of U2OS clones that had lacO arrays stably integrated into the genome, cells were seeded on coverslips or in Lab-Tek chambers (Nunc, Roskilde, Denmark) that were covered with Matrigel (BD Biosciences, San Jose, CA) and cultured for 2 d to achieve a stable attachment to the surface. The cells were transiently transfected ∼17 h before the experiment with Fugene 6 (Roche Diagnostics, Mannheim, Germany) or Effectene (Qiagen, Hilden, Germany), and various plasmid vectors. These included N-terminally tagged GFP/mRFP1-LacI and GFP/mRFP1-PML III as used previously (Görisch et al., 2004) or provided by Peter Hemmerich (Leibniz Institute for Age Research, Jena, Germany) (Weidtkamp-Peters et al., 2008), proliferating cell nuclear antigen (PCNA)-GFP/cyan fluorescent protein (CFP) from Cristina Cardoso (Leonhardt et al., 2000), H2A-mRFP1 derived from H2A-YFP (yellow fluorescent protein; Fejes Tóth et al., 2004), and a GFP-TRF2 construct. The latter was obtained by amplifying the human TRF2 cDNA sequence by PCR with an upstream forward primer containing an HindIII restriction site and a downstream reverse primer containing a BamHI restriction site. The PCR product was then cloned in the pEGFP-C1 vector (Clontech, Palo Alto, CA). Fixation of cells was conducted with 4% paraformaldehyde (PFA) in PBS buffer and incubation for 12 min at room temperature. Chromatin was stained either by mounting cover slips in Vectashield containing DAPI (Vector Laboratories, Peterborough, United Kingdom) or by staining with solution of 1 μM TO-PRO 3 (Invitrogen).

Fluorescence Microscopy

Long-term observations were conducted with a Leica TCS SP2 confocal laser scanning microscope (CLSM; Deerfield, IL). Z-stacks of 12–20 frames with 0.3-μm spacing were collected every 5 min for 3–5 h. The Leica CLSM was used with an objective heater and an environment chamber for maintaining the temperature at 37°C and the CO₂ concentration at 5%. For high-speed imaging a Perkin Elmer-Cetus Ultra-View or a Perkin Elmer-Cetus ERS-FRET spinning disk system (Norwalk, CT) with a Tokai Hit chamber (Fujinomiya, Japan) at 37°C temperature were used. These experiments were conducted either with medium containing 10 mM HEPES at pH 7.4 in the absence of CO₂ or with L-15 (Leibovitz) medium (PAA). For minute time-scale experiments Z-stacks were collected every 10 s for 15 min. For experiments with millisecond time resolution, cells were selected that had the three lacO loci in the same focal plane, and images were acquired for 5 min with a time resolution between 50 and 100 ms.

Metaphase Spreads

Cells were incubated with 0.5 μg ml⁻¹ colcemid (CaryoMax, Invitrogen) for 90 min at 37°C, detached with trypsin, pelleted by centrifugation, and resuspended in 500 μl fresh medium. Then 10 ml of a prewarmed 75 mM KCl solution was added, and the cells were incubated for 30 min at 37°C, pelleted, and resuspended in 500 μl KCl solution. Fixation was conduced by addition of 10 ml ice cold solution of methanol and glacial acetic acid at a 3:1 vol/vol ratio, and cells were incubated 30 min at −20°C, pelleted, washed five times, resuspended in 500 μl fixation solution, and deposited onto the slides.

Fluorescence In Situ Hybridization, Multicolor FISH, and PNA FISH

Chromosome spreads were treated with a solution containing 250 μg ml⁻¹ pepsin and 10 mM HCl for 2–3 min, washed with PBS buffer, and refixed with PFA. After an additional wash with PBS buffer, slides were dehydrated by series of ethanol washes (70, 90, and 100% ethanol) each with 5-min incubation time. DNA was denatured by incubating the slides for 5 min at 72°C in a solution containing 70% deionized formamide and 2× SSC buffer at pH 7. Cells were immediately dehydrated by another series of ethanol washes as described above. After drying of the slide, 200 ng of a fluorescein-labeled oligonucleotide probe coding for the lacO sequence was added onto the sample in a 30 μl solution containing 1.5 M NaCl and 0.25 M Na-acetate at pH 7. A coverslip was added and sealed with fixogum, and slides were incubated overnight at 37°C in a wet chamber, washed in 4× SSC, 0.2% Tween 20 for 5 min, and mounted in Vectashield. For simultaneous staining of lacO arrays and telomere repeats a FISH (fluorescence in situ hybridization) label of the lacO sequence was first conducted as described above. Then a telomere PNA FISH was conducted with a Cy3-labeled (CCCTAA)₃ PNA probe according to the protocol of the manufacturer (Dako, Glostrup, Denmark). Chromosomes spreads from the preceding lacO FISH experiment were relocated and imaged. The multicolor FISH (mFISH) analysis was performed using the SpectarVysion 24 color chromosome paint assay and Quips Spectra Vysion software (Vysis, Downers Grove, IL). Images were acquired using an Axioplan 2 microscope (Zeiss, Jena, Germany) together with a Sensys CCD camera (Photometrix, Tucson, AZ). After the mFISH analysis, slides were incubated two times for 10 min in a 4× SSC solution containing 0.1% Tween 20, washed two times with 2× SSC buffer, and dehydrated by a series of ethanol washes. Then labeling of the lacO arrays by FISH was conducted as described above, and chromosome spreads were relocated and imaged.

Mobility Measurements

For single-particle tracking maximum intensity projections of three-dimensional (3D) stacks along the Z-axis were used. It is noted that for an isotropic movement of the particle in 3D the value determined for the diffusion coefficient D from projections to 2D will be the same as that obtained from the analysis of 3D data sets as described previously (Görisch et al., 2004). The coordinates for the movements of the GFP-LacI–labeled lacO arrays were determined with ImageJ 1.36 (NIH, http://rsb.info.nih.gov/ij/) using the SpotTracker software for tracking particle translocations with subpixel resolution (Sage et al., 2005). Alignment of the cell nucleus was made with the StackReg plug-in (Thévenaz et al., 1998) or with a self-written MatLab (MathWorks, Natick, MA) program that uses an image correlation algorithm to maximize the image overlap between successive images. From the x-y coordinates of the tracked spots either the mean squared displacement (MSD) or the mean squared distance change (MSΔD or <Δd2>) between two labeled telomeres was determined. The latter approach was developed here to improve the accuracy and is described in detail in Supplemental Materials. For the mobility analysis the three telomeres in clone F6B2 were tracked on different time scales in 22 cells (second time scale, Δt = 50–100 ms), 18 cells (minute time scale, Δt = 8–15 s), and 44 cells (hour time scale, Δt = 5 min). For the mobility analysis only cells were evaluated that were stably attached to the surface and showed little deformation of their shape.

RESULTS

Characterization of U2OS Cell Lines with Integrated lacO Arrays

A series of 30 cell clones was generated that had lacO repeats stably integrated into the genome. The cell clones were distinct with respect to the chromosomal integration loci and the number of operator integrations. The lacO arrays were visualized in vivo by transient transfection with GFP-LacI or mRFP1-LacI constructs. Examples are shown in Figure 1. Because also the unbound GFP-LacI/mRFP1-LacI repressor remained in the nucleus, it provided a nuclear counterstaining during the in vivo observations (Figure 1A), whereas for fixed cells a DNA staining with DAPI or TO-PRO 3 was conducted (Figure 1B). The sites of the lacO integrations were determined by FISH (Figure 1C). Remarkably, all except one of the clones had an integration at the telomeres (Figure 1D). This observation indicates a high DNA recombination activity at the telomeres in U2OS cells that use the ALT mechanism for maintaining their chromosomal ends. The experiments reported in the following used clone F6B2 with lacO integrations sites at three different telomeres (Figure 1B). Because the U2OS cell line displayed a complex karyotype, the genomic integration sites were identified from metaphase FISH (Figure 1C) and mFISH experiments (Supplemental Figure S1) followed by a rehybridization with the fluorescently labeled lacO probe. The integration sites were found to be at the telomere regions of chromosomes 6q, 11p, and 12q. Because of the different repeat numbers of the lacO integrations, the three chromosomes could be distinguished in vivo. The biggest spot b is on chromosome 11p, the medium sized spot m is on chromosome 6q, and the smallest spot s is on chromosome 12q. The three loci had apparent diameters of 0.74 ± 0.09 (b), 0.54 ± 0.05 (m), and 0.42 ± 0.05 μm (s), respectively. Assuming an average nucleosome concentration of 0.14 mM and a 200-base pair nucleosome repeat length (Weidemann et al., 2003), this would correspond estimates of 3.6, 1.4, and 0.7 Mb of DNA. Because of the diffraction-limited resolution of the optical system used here, the true dimensions of the array are smaller. This is inferred from comparitive measurements of array sizes by spatially modulated illumination microscopy (Reymann et al., 2008). On the other hand the chromatin density of the lacO arrays appears to be above average.

Figure 1.

U2OS cell clones with a stably integrated lacO array. Scale bar, (A) 10 μm; (B–D) 5 μm. (A) Cell clone EC4 with a single lacO integration observed in vivo after transient transfection with GFP-LacI. (B) The clone F6B2 showed three distinct integration sites that could be identified due to their differences in size. Cells were transiently transfected with GFP-LacI and fixed and DNA-staining was with TO-PRO 3. (C) The chromosomal integration site of lacO arrays was determined by metaphase FISH using a 6-FAM fluorescently labeled oligonucleotide probe coding for the lacO sequence and DNA counterstaining with DAPI. The lacO arrays were found to integrate at the telomeric regions of chromosomes 11p, 6q, and 12q. See also Suppelmental Figure S1 for mFISH analysis. (D) Metaphase chromosomes from different cell clones with a stably integrated lacO array showed a highly preferred integration at the telomeres (29/30 clones).

Because the metaphase chromosomes are in a highly condensed chromatin state, the spatial overlap of the lacO integration site with the telomere repeat does not exclude the presence of a significant amount of DNA between the two loci. We therefore investigated the localization of lacO arrays with respect to the telomeres in decondensed chromosomes during interphase. The lacO arrays were labeled with GFP-LacI or mRFP1-LacI, whereas the telomeres were visualized by three different approaches: 1) FISH using a Cy3-labeled PNA probe directed against the telomere repeats (Figure 2A). 2) Immunostaining of endogenous TRF2 protein that binds with high specificity to the (TTAGGG)_n sequence (de Lange, 2005; Figure 2B). 3) In vivo labeling with GFP-TRF2, which was introduced by cotransfection with an mRFP1-LacI vector (Figure 2C). All three methods showed essentially the same results. The three lacO array integrations were located directly adjacent to the telomere repeat sequence and no differences between arrays b, m, and s were apparent. Live imaging with cells cotransfected with mRFP1-LacI together with GFP-TRF2 demonstrated that the colocalization was stable through a period of several hours and that the lacO arrays showed a highly correlated movement with the GFP-TRF2 label (Supplemental Movie 1). Thus, tracing the lacO array in vivo is an appropriate representation of the mobility of the adjacent telomere. It is also noted that the intensity of both the PNA and the GFP-TRF2 fluorescence signal correlates with the telomere repeat length (Lansdorp et al., 1996). Thus, the absence of a PNA/GFP-TRF2 signal close to a given lacO label is indicative of a very short telomere repeat at this chromosome (Perrem et al., 2001; Scheel et al., 2001).

Figure 2.

Integrated lacO arrays colocalize with telomeres during interphase. Maximum intensity projections of confocal picture stacks of the F6B2 clone are shown. The white arrows indicate lacO arrays that colocalize with the telomere signal in a single focal plane. The analysis was conducted with three different labeling methods. In all three methods, a fraction of the lacO arrays were without a detectable telomere signal indicative of very short telomere repeats. Scale bar, 5 μm. (A) The lacO repeats were detected by transfecting the cells with GFP-LacI, and telomeres were stained by FISH with a Cy3-labeled (CCCTAA)₃ PNA probe. (B) Cells were transfected with mRFP1-LacI and endogenous TRF2 was detected by immunostaining with an Alexa-488–coupled secondary antibody. (C) Cells were cotransfected with mRFP1-LacI together with GFP-TRF2 for labeling of living cells.

The mobility analysis presented here was conducted with cells in the middle of G1 to early S phase. From experiments with an H2A-YFP label it was apparent that shortly after cell division in early G1 phase the cells were more rounded and showed relatively large translocations on the surface. These cells in early G1 phase were excluded because we selected only flat cells with fully decondensed chromosomes that were stably attached to the surface. Using the replication associated protein PCNA as a cell cycle marker, we identified an arrest in early S phase before replication of the lacO arrays in cotransfection experiments with mRFP1-LacI and PCNA-CFP (Supplemental Movie 2). In the absence of LacI repressor no effect on cell cycle progression was observed with respect to the parental U2OS cell line without the lacO integrations (data not shown). Thus, we concluded that LacI bound to the lacO repeats blocked DNA polymerase at the late replicating telomeric regions (Gilson and Geli, 2007). This is consistent with the finding that LacI or TetR repressors tightly bound to arrays of their operator sites efficiently stall the replication fork in Escherichia coli (Possoz et al., 2006). As determined from the characteristic PCNA distribution pattern, 62 ± 5% of the cells with three lacO array labels were in G1 phase, whereas the remaining 38 ± 4% displayed PCNA in replication foci. When evaluating the lacO mobility in G1 cells with those in early S phase (either in separate cells or in cells that progressed from G1 to S phase, Supplemental Movie 2), no correlation of mobility and cell cycle state was observed.

Theoretical Framework for Evaluating the Telomere Mobility

The accurate measurement of the MSD in the nucleus is difficult because the cell itself can move during the measurement and the alignment to correct for this is fraught with difficulties. Accordingly, an alternative method was developed here that evaluated changes in the distance between two loci (Figure 3A). This approach was applied previously to analyze the mobility of lacO-tagged genomic loci but not in the context of a quantitative confined diffusion model (Marshall et al., 1997; Vazquez et al., 2001; Chubb et al., 2002). A detailed derivation of the formalism used here is given in Supplemental Materials. If multiple particles/loci are present, the change in distance d⃗(t) = p⃗₁(t) − p⃗₂(t) between the position p⃗ of two particles with indices 1 and 2 is given by Equation 1. In this regard, the mean squared difference in distance, MSΔD, referred to here also as [Δd²], is the sum of the mean squared displacement by diffusion, and MSD_diff, of the two particles under the condition that rotation of the cell can be neglected:

This was found to be a valid assumption under the experimental conditions used here as inferred from the mean linear displacement of the difference of two spots (Supplemental Materials Equation S8). To describe the restricted mobility of particles in the nucleus we have previously introduced a so-called “moving corral” (MC) model (Görisch et al., 2004). An analysis of the data on the mobility of lacO loci indicated that a similar approach was also appropriate to represent the mobility of telomeres on different length and time scales. In the MC model the overall mobility in n dimensions is determined by three parameters: The particle experiences fast but confined diffusion with a coefficient D_fast in a corral with radius r_c, and the corral can also translocate by free diffusion with a coefficient D_slow. This model was extended to derive the above parameters from the change in the distance between two particles that experience corralled diffusion. It was found that telomere mobility on the hour time scale could be described by a simplified model according to Equation 2 that contained only the diffusion term within the corral (corresponding to D_slow = 0 in Equation 3):

The additional offset c represents the faster mobility contributions detected in the analysis of telomere mobility on the second and minute time scale. The data obtained from measurements obtained with these shorter observation times were described by Equation 3. Here the translocation of the corrals by free diffusion with coefficient D_slow was included:

The analysis based on Equations 2 and 3 was found to give more accurate and robust values for the mobility of the telomeres than those obtained from the analysis of a single telomere after correction for cell movements. Furthermore, a direct and faithful comparison of the telomere mobilities within the same cells could be made.

Figure 3.

Average mobility of telomeres on the second, minute, and hour time scale. Images were recorded with a spinning disk or a point scanning CLSM. The fluorescent spots were tracked and the mean squared changes <Δd²> in the distance between two telomeres were plotted against time. Averages of 20–40 trajectories are shown in each panel. (A) The mobility analysis of clone F6B2 was conducted in vivo by measuring changes in the distances between the three loci. In this clone the three loci were distinct with respect to the number of integrated lacO repeats and are referred to as b (for big), m (for medium), and s (for small). The distances b-m, b-s, and m-s are indicated. Scale bar, 10 μm. (B) Images were taken with a time resolution between 50 and 100 ms for a total of 5 min (see Supplemental Movie 3). The corresponding tracing of fixed cells is also plotted. The curve is a fit of the data to the MC model (Equation 3). (C) Image stacks were taken every 8–15 s for 20 min (see Supplemental Movie 4). The curve is a fit of the data to the MC model (Equation 3). (D) Image stacks were acquired in 5-min time intervals for a total of 3–5 h min (see Supplemental Movies 5 and 6). The solid line is a fit of the data to a confined diffusion model (Equation 2).

Telomere Mobility on Different Time Scales

The mobility of the three telomeres in clone F6B2 on the second, minute, and hour time scale was recorded with corresponding time resolutions of milliseconds, seconds, and minutes. The quantitative analysis of telomere mobility (MSΔD vs. time) was conducted as described in the preceding paragraph from measurements of the distance changes between the three labeled loci in the F6B2 clone (Figure 3A). Experiments with fixed cells showed that the position determination with subpixel resolution had an accuracy of ∼60 nm.

All telomeres displayed a fast movement in a region of ∼80-nm radius (after subtraction of 60 nm) with D_sec = 2·10⁻³ μm² s⁻¹ (Figure 3B, Supplemental Movie 3), and further slower translocations in a radius of 150 nm with D_min = 4·10⁻⁴ μm² s⁻¹ (Figure 3C, Supplemental Movie 4). For the analysis of movements on the hour time scale telomeres were traced via the lacO label for 3–5 h. From these trajectories the MSΔD versus time of the big, medium, and small lacO arrays was determined. The confinement radii r_c, diffusion coefficients, and baseline offset for the three individual telomeres in each cell were determined from a global fit to Equation 2 (Figures 3D and 4, C and D) with a typical time-lapse series shown in Figure 4A. Some telomeres displayed an extended mobility with one particularly striking example shown in Figure 4B (see also Supplemental Movies 5 and 6). The existence of a bimodal mobility distribution for telomere translocations measured on the hour time scale was clearly apparent in the distribution of the radius of the accessible region (Figure 5A, n = 132). A lower mobility fraction of ∼85% of the telomeres moved within a radius of 0.36 ± 0.16 μm average, whereas the mobility of the remaining ∼15% was extended to a region with an average radius of 0.8 ± 0.1 μm. These values include the contributions of the faster movements on the second and minute time scale that are represented by a baseline offset in Equation 2 that is added to the value of the parameter r_c to describe the extend of the total accessible region. For both populations the same average diffusion coefficient of D_hour = 1·10⁻⁵ μm² s⁻¹ was retrieved from the fit to Equation 2. No statistically significant differences between the telomere labels at 6q, 11p, and 12q were observed, indicating that the number of the lacO repeats integrated did not affect the mobility. Furthermore, the increased mobility on the hour time scale was not restricted to a certain telomere but was observed with a fraction of all three labeled telomeres (b, m, s) of clone F6B2. The results of the analysis of telomere translocations on different time scales are summarized in Figure 5B.

Figure 4.

Differences in the mobility of telomeres on the hour time scale. (A) Cell with relatively slow-moving telomeres, which represents the majority of the observations (Supplemental Movie 5). (B) Cell showing two telomeres, b and s, that display extended translocations, whereas telomere m is only moderately mobile (Supplemental Movie 6). The quantitative analysis of distance variations between the three spots is shown in C for the cell in A and in D for the cell in B. The individual telomere mobility was calculated as described in the text from the distance changes between b-s, b-m, and m-s. The solid lines represent a fit to Equation 2, in which the translocations on the second and minute time scale are included as an offset.

Figure 5.

Mobility of telomeres in U2OS cells on different time scales. (A) Distribution of telomere translocations on the hour time scale as given by the radius of the total accessible region. The solid line represents a fit to a sum of two Gaussian functions. A ∼85% fraction of the telomeres showed a mobility that was restricted to average translocations of 0.36 ± 0.1 μm, whereas the remaining fraction of telomeres displayed an extended mobility described by a Gauss distribution centered around 0.8 ± 0.1 μm. For three telomeres r_c values of 1.35, 1.44, and 1.94 μm were determined. These data are not included in the histogram. (B) Summary of combined results from the telomere mobility analysis on the second, minute, and hour time scale, which yielded the indicated diffusion coefficients and translocation ranges (Figures 3 and 4).

Analysis of Telomere Repeat Lengths

The telomere repeats were stained by FISH with a Cy3-labeled PNA probe specific for the (TTAGGG)_n sequence (Figure 6). In a comparison of human lymphocytes (Figure 6A) with U2OS cells (Figure 6B) it is evident that the telomeres in the U2OS cells showed large variations in their repeat lengths as reflected by the intensity differences of the fluorescent label (see also Figure 2). In addition, the overall signal of the U2OS label was significantly higher than that of the lymphocytes, consistent with the increased average telomere length reported for this cell line (Scheel et al., 2001). By conducting FISH experiments with a lacO probe (Figure 6C) followed by a rehybridization with the PNA probe against the telomeric repeats (Figure 6D), the telomere signal intensity was measured for the three lacO-labeled telomeres in the F6B2 clone. In Figure 6E examples for this type of analysis are presented. The s, m, or b lacO inserts displayed similar distributions of fluorescence intensities, indicating that the populations were not distinct with respect to the telomere repeat length. A large variability of signal intensity was observed at each of the three loci. For 20–30% of the lacO-labeled sites, the telomeres were undetectable by the PNA probe hybridization with a signal intensity less than twofold above the background value. A comparison of U2OS cells and the non-ALT HeLa cell line suggested that the presence of very short telomeres is pronounced in U2OS cells (Figure 6F). The distribution of telomere lengths in U2OS and HeLa cells was derived from the intensity of the GFP-TRF2 fluorescence signal (Figure 2C). For each GFP-TRF2 spot the ratio of its integrated intensity to the average value of all GFP-TRF2 telomere signals in a given cell was computed to normalize for differences between experiments. The distributions plotted in Figure 6F showed a much higher fraction of very short telomeres (GFP-TRF2 signal ≤0.5) in U2OS cells. It was then examined whether the telomere length correlated with the fraction of telomeres that showed an extended mobility on the hour time scale. Telomeres were traced for 3–4 h in living cells that had a double-label of the telomeric lacO arrays with mRFP1-LacI and the telomere repeats via GFP-TRF2. From these trajectories the radius of the average accessible space was determined as described above and plotted against the corresponding telomere length as given by the normalized GFP-TRF2 signal. This analysis revealed an inverse correlation of the accessible space and the telomere repeat length/GFP-TRF2 signal, indicating that short telomeres indeed displayed an extended mobility (Figure 6G).

Figure 6.

FISH of telomere repeats. Metaphase chromosome spreads were hybridized with a Cy3-conjugated PNA probe against the telomere repeat sequence. The DNA was counterstained with DAPI. (A) Human lymphocytes showing a homogenous distribution of telomeres with repeats of 10–15 kb in length. (B) U2OS cell line, in which the alternative telomere lengthening (ALT) mechanism leads to large variations in telomere length. (C) FISH experiment conducted on metaphase chromosomes of the F6B2 clone using a fluorescently labeled probe coding for the lacO sequence. (D) Rehybridization of the metaphase spread shown in C with a Cy3-(CCCTAA)₃ PNA probe used for labeling the telomere repeats. (E) Examples for the comparison of telomere length as given by the intensity of the Cy3-PNA signal. The three different integration sites of lacO arrays were identified by FISH with a 6-FAM fluorescently labeled oligonucleotide probe against the lacO sequence. (F) Quantification of telomere length in U2OS (■) and HeLa cells (□) as derived from the intensity of the GFP-TRF2 fluorescence intensity signal. For normalization the integrated intensity of a given GFP-TRF2 spot was determined as the ratio to the average value of all GFP-TRF2 telomere signals in a given cell. The U2OS cells have a significantly higher amount of very short telomeres as reflected by the GFP-TRF2 signal ≤0.5. The two normalized distributions compare to absolute telomere repeat lengths in the range from <2 to ∼23 kb reported for U2OS (Scheel et al., 2001) and of 2–10 kb for HeLa cell lines (Bryan et al., 1998). (G) Correlation of telomere mobility on the hour time scale as represented by the radius of accessible space versus the telomere repeat length as represented by the normalized GFP-TRF2 signal (see F). A total of 78 telomeres traced for ∼3 h were evaluated. From the plot an inverse correlation of mobility with the telomere repeat length/GFP-TRF2 signal is apparent.

Mobility of PML Bodies in Relation to Telomeres

Cells of the F6B2 cell clone were cotransfected with GFP-PML III and mRFP1-LacI to examine the relation between telomere and PML-NB mobility. Immunostaining of endogenous PML protein showed that transient transfection of the cells with mRFP1-LacI did not modify number, size, or shape of PML-NBs compared with the cells used in this study. As expected, GFP-PMLIII colocalized with Sp100 protein, confirming that the corresponding structures were indeed PML-NBs (data not shown). Distinct telomere–PML-NB complexes, the APBs, were detected at 5–10% of the lacO-tagged telomeres (Figure 7, A–C). In three-color labeling experiments of lacO arrays, TRF2, and PML bodies, the APB structures were found with and without a TRF2 signal, indicating that APBs are present both at short and long telomeric repeats (data not shown). As evident from the enlarged view in Figure 7C, the APB complex was bigger than a normal PML-NB, and the PML protein adopted a cap-like structure around the telomere. APBs formed via binding of an existing PML-NB to a telomere as visualized by the image series shown (Figure 7D, Supplemental Movie 8). In addition, a close inspection of the time-lapse series indicated that some free PML protein was also accumulating from the nucleoplasm around the telomere until the typical APB structure was established.

Figure 7.

Structure of PML body-telomere complexes (APBs). (A) Cells were transiently transfected with mRFP1-LacI and fixed with PFA. PML bodies were stained with an antibody against PML protein together with an Alexa 488–coupled secondary antibody. Chromatin was stained with TO-PRO 3. (B) U2OS cell line F6B2 transiently cotransfected with mRFP1-LacI together with GFP-PML III. Cells were fixed with PFA, and chromatin was stained with TO-PRO 3. Scale bars, 5 μm. (C) Enlarged view of the PML complex at the telomere from the cell shown in B. The PML body formed a cap-like structure around the lacO/telomere label. (D) In the dual-color labeling experiments the formation of an APB complex could be visualized (Supplemental Movie 8).

The PML-NB mobility with respect to the three telomeric lacO inserts was investigated by monitoring the change of distance between a PML-NB to the closest lacO array over time (Figure 8). The PML-NBs were referred to as PML_b, PML_m, and PML_s according to the proximal lacO locus. Images of a typical experiment (Figure 8A, Supplemental Movie 9) and the corresponding plot of the squared distance change, Δd², for each three PML-telomere pairs are displayed in Figure 8B. Based on the analysis of 40 different trajectories, three types of PML mobility in addition to that of the APB complexes were observed. These are represented by the three curves of the averaged Δd² versus time relation in Figure 8C. Thus, four distinct types of PML body translocations were identified here: 1) PML bodies showing movements that were completely independent with respect to the proximal telomere-like PML_m in Figure 8A. 2) PML bodies with a mobility fully correlated with that of the telomere but without direct colocalization to the telomere (PML_b in Figure 8A). 3) PML bodies that displayed a movement, which partly correlated with that of the telomere (PML_s in Figure 8A). 4) PML bodies forming an APB. As expected, this stable PML–telomere complex (Figure 7) also showed a completely correlated movement with the telomere (Supplemental Movie 10).

Figure 8.

Relative mobility of PML bodies and telomeres. The U2OS cell line F6B2 was transiently cotransfected with GFP-PML III (green) and mRFP1-LacI (red). (A) The mobility of the three lacO inserts (b, m, and s) and proximal PML bodies was studied by recording changes in the relative distance between the spots (Supplemental Movie 9) and fitting these data to a confined diffusion model. (B) The mean squared changes <Δd²> in the distance between PML bodies and its proximal telomeres observed in the time lapse series of the cell shown in A were plotted against time. (C) Average mobility of PML-NBs with respect to a telomere with lacO label in close proximity, i.e., at a distance ≤1 μm. Each curve represents averages of 10–15 measurements. About 20 cells were evaluated, and 30 PML-NBs plus ∼10 APBs were traced with respect to the changes in the mean squared distance between telomere and PML signal. The results indicated a clustering of the trajectories into the three main groups shown. (D) Mobility of PML bodies close to telomeres (d ≤ 1 μm). The experiments indicated the existence of four types of PML body movements: One group showed a mobility that was not correlated to that of the telomere (“mobile PML body”), whereas another group was clearly separated from the telomere but nevertheless followed the movements of the telomere. This suggests a confinement within or binding to the same chromatin region (caged or bound PML body). A third group displayed an intermediate behavior, i.e., a partial correlation of the translocations. The fourth class of PML bodies is represented by the APB complex with a stable association of PML body and telomere.

DISCUSSION

Mobility of Telomeres in U2OS Cells

The mobility of chromatin loci as well as nuclear bodies has been examined in a number of studies as reviewed previously (Chuang and Belmont, 2007; Lanctôt et al., 2007; Wachsmuth et al., 2008). Here, a clone of the human U2OS cell line was investigated that had three lacO array integrations adjacent to the telomere of chromosomes 6q, 11p, and 12q. These could be distinguished in the in vivo experiments via bound GFP-LacI/mRFP1-LacI because of their different number of lacO repeats. A comprehensive model for telomere mobility was determined by combining measurements at the second, minute, and hour time scale that were all associated with different confinement radii for the telomere translocations (Figure 5B). By measuring the change of the distance between two labeled telomeres, instead of tracking a single telomere, the mobility measurements could be conducted with improved accuracy. On the minute time scale the lacO-tagged telomeres moved in a diffusive manner with D_min = 4·10⁻⁴ μm² s⁻¹, similar to other chromatin loci, fluorescent centromeric protein clusters, or nuclear bodies, for which respective diffusion coefficients between 1.1·10⁻⁴ and 1.8·10⁻⁴ μm² s⁻¹ were reported (Chubb et al., 2002; Molenaar et al., 2003; Görisch et al., 2004). It is noted that the time regime chosen for the experiments directly affects the resulting apparent diffusion coefficients, if for example, only the initial linear slope of the MSD versus time plot is evaluated (Wachsmuth et al., 2008). In a high-resolution analysis the mobility of lacO arrays in Chinese hamster ovary (CHO) cells was measured on the second time scale (Levi et al., 2005). Multiple mobility layers were detected, and the data were fitted to a confined diffusion model similar to the “moving corral” description used here (Equation 3). This analysis yielded D_fast = 3.1·10⁻³ μm² s⁻¹ and D_slow = 2.4·10⁻⁴ μm² s⁻¹, which is in good agreement with values of D_sec = 2·10⁻³ μm² s⁻¹ and D_min = 4·10⁻⁴ μm² s⁻¹ determined here for telomeres in U2OS cells.

Interestingly, at the hour time scale, the diffusion coefficient D_hour = 1·10⁻⁵ μm² s⁻¹ derived from our analysis is very close to the one observed for 1-Mb chromatin domains labeled with Cy3-UTP on a similar time scale (Bornfleth et al., 1999). This is consistent with the interpretation that this value represents the dynamics of large chromatin domains in the nucleus. The relatively low mobility measured for the majority of telomeres is also in agreement with a recent study where the mobility of DNA ends created from a double-strand break between a lacO/LacI-CFP and a tetO-TetR-YFP label was evaluated (Soutoglou et al., 2007). The results indicate that DNA ends per se do not display a higher mobility than other chromatin loci. This supports the model of a linked chromatin fiber network, in which the local chromatin mobility is confined to a radius of a 150–250 nm (Wachsmuth et al., 2008). In the analysis of the telomere mobility over 3–5 h two distinct populations were identified here (Figure 5A). Although ∼85% of the telomeres moved within an average confinement radius of 0.36 ± 0.16 μm, the remaining fraction showed more extended translocations of 0.8 ± 0.1 μm. Furthermore, the size of the accessible region was inversely correlated with the length of the telomere repeat sequence (Figure 6). In a previous study the telomere mobility in U2OS cells was analyzed over a time period of 30–60 min with a 1-min time resolution after labeling the telomere repeats by hybridization with a PNA probe (Molenaar et al., 2003). The majority of telomeres moved in a confined region with an apparent radius of ∼0.5 μm and a short-range diffusion coefficient of 1.8·10⁻⁴ μm² s⁻¹, as inferred from the plateau and initial slope of the MSD versus time plot. A ∼10% fraction of telomeres displayed a higher mobility for which translocations up to 1.2 μm and an apparent diffusion coefficient of 5.8·10⁻⁴ μm² s⁻¹ were reported. The multiple time-scale analysis conducted here described the telomere mobility by a model that provides a clear distinction between structural features of the nucleus environment (represented by the three different confinement radii) and the intrinsic telomere mobility (described by the corresponding three diffusion coefficients; Figure 5B). The previous conclusions on the existence of a fraction of telomeres with higher mobility were confirmed, albeit with some quantitative differences that are related both to the data analysis and the labeling method. It is noted that the in vivo PNA labeling approach allowed for the simultaneous detection of ∼50% of the telomeres in the U2OS (Molenaar et al., 2003). However, it is not suited for the analysis of the short telomeres that are characteristic for telomerase-negative cells, because these do not provide a signal of sufficient intensity for tracing them in vivo (Figures 2 and 6). Here, we were able to tag the chromosome ends via the lacO label so that the mobility of specific telomeres could be monitored independently of the telomere repeat length, albeit with the limitation to the simultaneous detection of only three telomeres.

Possible Origins of Increased Telomere Mobility

In yeast, telomeres and subtelomeric regions are attached to the nuclear envelope (Gotta et al., 1996; Laroche et al., 1998, 2000; Tham et al., 2001; Hediger et al., 2002). The analysis of in vivo mobility of chromosome VI-R, XIV-L, and VIII-L telomeres revealed relatively fast oscillating movements of 150–300 nm that could be assigned to an anchoring to the nuclear envelope via the Ku protein complex (Hediger et al., 2002). In mammals telomeres appear randomly positioned in the nucleus during the G1 phase of the cell cycle (Luderus et al., 1996; Chuang et al., 2004; Ermler et al., 2004), with some telomeres being specifically localized in heterochromatin at the nuclear and nucleolar periphery (Tam et al., 2004). An association of telomeres with a nuclear matrix was suggested on the basis of biochemical approaches (Luderus et al., 1996; Weipoltshammer et al., 1999). However, the existence of the nuclear matrix remains controversial, and several lines of evidence support the view that a stable chromatin fiber network is established by different types of chromatin linkers so that no additional support structure is required (Rippe, 2007). Thus, a remaining more likely possibility is the association of telomeres with other chromatin domains.

Changes in the nuclear mobility of chromatin loci have also been reported in relation to the gene expression state. On activation of GAL genes in yeast a confinement of the GAL locus to a 2D sliding motion along the nuclear envelope was observed in live cell–imaging experiments (Cabal et al., 2006). Furthermore, the results obtained here for telomere mobility bear some similarities with measurements of a VP16-inducible gene locus in CHO cells (Chuang et al., 2006). The activation of the lacO-labeled gene region triggered a dramatic increase of its mobility, presumably by inducing its dissociation from a stable heterochromatin region. Interestingly, telomeres as well as subtelomeric regions display constitutive heterochromatin markers that tend to be lost upon shortening of telomere repeats (Benetti et al., 2007). This type of anchoring could also include the self-association of different telomeres reported previously (Nagele et al., 2001; Molenaar et al., 2003; Chuang et al., 2004). Here, we suggest that the higher mobility fraction of telomeres is characterized by a reduced ability to interact with other (hetero)chromatin regions to explain the inverse relation of telomere repeat length and mobility observed in our experiments. Interestingly, a recent study reports very similar behavior. The disruption of the telosome by deletion of TRF2 was found to increase telomere mobility in 53BP1-deficient cells by allowing them to sample larger nuclear territories (Dimitrova et al., 2008).

PML Body Mobility in Relation to Telomeres

The comparison of telomere and PML-NB mobility is of interest in several aspects. First, it is an open question how mobile nuclear bodies and their associated activities could be targeted to a specific region of the nucleus (Muratani et al., 2002; Platani et al., 2002; Görisch et al., 2004). We have reported previously that the PML-NB mobility in HeLa cells is indistinguishable from that of a bona fide inert particle, i.e., directly reflects the accessibility and dynamics of its chromatin environment (Görisch et al., 2004). For the U2OS cell line this description is clearly inappropriate as specific complexes of PML-NBs with telomeres, the APBs, are formed that are not present in telomerase-positive cell lines like HeLa (Yeager et al., 1999). Thus, a chromatin-bound state of PML-NBs exists in U2OS cells that appears to play an essential role in the ALT pathway (Yeager et al., 1999; Nabetani et al., 2004; Potts and Yu, 2007). With respect to the mobility of PML-NBs, four types were identified in our analysis (Figure 8D): 1) PML-NBs with translocations that were independent of a close telomere label. These were confined to areas of an average radius of ∼0.3 μm where they can diffuse freely with a mobility that was identical to those reported in HeLa cells (Görisch et al., 2004). 2) PML-NBs that appeared to be stably associated with the same chromatin domain as the neighboring telomere, either by binding or by being trapped by the surrounding chromatin. These PML-NBs were observed at a distance of up to ∼1 μm from the lacO label. 3) PML-NBs that had a correlated mobility with respect to the proximal telomere but, in addition, showed some independent translocations. These PML-NBs appear to represent an intermediate state between the completely mobile and the bound/caged group of PML-NBs, indicating that the transition between the two states is reversible on the minute time scale. 4) PML-NBs that form APBs as discussed below. The APB complexes did not dissociate on the time scale of the in vivo imaging experiments, and their formation was initiated by the interaction of a telomere with a preexisting PML-NB.

The mobility analysis conducted here does not reflect the absolute mobility of PML-NBs but their mobility relative to a proximal chromatin locus, i.e., the chromatin-independent part of their translocations. Accordingly, the apparent diffusion coefficients are relatively small and cannot be compared directly with the previously determined value of D_min,PML = 2·10⁻⁴ μm² s⁻¹ from measurements on the minute time scale (Görisch et al., 2004). The latter value is close to the corresponding measurement of D_min = 4·10⁻⁴ μm² s⁻¹ determined here, suggesting that the translocations of PML-NBs mostly follow that of chromatin as proposed previously. In addition, the confinement radii for both PML-NBs and telomeres indicate that the typical size of the accessible region is limited to a radius of ∼0.3 μm as reported previously for nuclear bodies and chromatin domains (Wachsmuth et al., 2008). Thus, the data are consistent with a model in which chromatin dynamics, accessibility, and binding to (sub)telomeric chromatin loci determine the mobility of PML-NBs in ALT cells.

Structure and Formation of APBs

The PML protein is organized in a cap-like structure at the telomere that extends to the lacO array so that the chromosome end is covered with PML protein (Figure 7). In the 3D image reconstruction of APBs formed at the locus b of clone F6B2 (Supplemental Movie 7) the PML cap at the chromosome end appeared to extend to the neighboring fraction of the adjacent lacO repeats but did not encompass the whole array. Furthermore, the clone F4 2B8 with a lacO array integration close to the centromere (Figure 1D) displayed only an insignificant number of PML-NBs colocalizing with the lacO repeat (data not shown). These two observations indicate that PML-NBs do not have an affinity for the lacO repeat per se in U2OS cells as suggested previously in studies of a BHK cell line (Tsukamoto et al., 2000). The PML cap-like structure around the telomere found in APBs, albeit smaller in size, is reminiscent of the unusually large PML-NBs that form in lymphocytes of patients affected with immunodeficiency, centromeric instability, and facial dysmorphy (ICF) syndrome (Luciani et al., 2006). Interestingly, the hollow-spherical PML-NBs associated with the ICF syndrome contain a large amount of satellite DNA that displays the same repetitive nature and constitutive heterochromatin markers as the telomeric DNA in APBs. The lacO telomere in vivo labeling offered the unique opportunity to elucidate the formation process of an APB (Figure 7D, Supplemental Movie 8). A close examination of these data indicated that the APBs are formed in two steps: First, a preexisting PML-NB binds to the telomere and, in a second step, additional PML protein from the nucleoplasm is accumulated at this locus until the typical APB structure is formed. This two-stage process appears to be required because the typical PML content of the relatively large APBs significantly exceeds that of an isolated PML-NB.

A Model for the Mechanism of Alternative Telomere Lengthening

A valuable feature of the lacO/GFP-LacI telomere labeling in studies of the ALT system is the possibility to follow a specific telomere over time, independent of its repeat length. This type of analysis revealed an extended mobility of ∼15% of the telomeres, whereas a similar fraction of 20–30% of the telomeres had telomere repeats that were too short to be detectable by FISH with the PNA probe (Figures 4–6). From a comparison with HeLa cells it is concluded that the presence of a high number of short telomeres is a characteristic feature of U2OS cells (Figure 6F). Furthermore, an inverse correlation of the confinement radius r_c and the telomere length was revealed (Figure 6G). Based on these findings and the results on telomere mobility and interaction with PML-NBs discussed above, the model in Figure 9 is proposed. The telosome protein complex at the intact telomere promotes interactions with other chromatin domains and mediates its anchoring within the nucleus. In this state the telomere mobility is low and similar to that of other chromatin loci. However, upon shortening of the telomere repeat below a certain threshold the telosome is disrupted, which could then release the telomere anchoring to allow for extended telomere translocations. Telomeres with short repeats and unprotected by the shelterin proteins TRF1, TRF2, and POT1 that directly recognize (TTAGGG)_n sequence are likely to resemble the structure of DNA damage sites as discussed previously (de Lange, 2005). Interestingly, many proteins involved in the detection and the repair of double-strand breaks are found to accumulate, at least transiently, in PML bodies in general and also specifically in APBs (Dellaire and Bazett-Jones, 2004; Nabetani et al., 2004; Bernardi and Pandolfi, 2007; Fasching et al., 2007). Thus, the increased telomere mobility in conjunction with a double-strand break-like conformation of the chromosome end could favor both the formation of a stable APB complex as well as telomere–telomere interactions. It is apparent from the results obtained here that telomeres are recombination “hot spots” in U2OS cells. The vast majority clones constructed during this study (29 of 30) had the lacO arrays integrated at the telomeres. Moreover, the relatively large size of the lacO inserts obtained suggests the expansion of the integrated DNA. This could result from a pronounced DNA recombination and repair activity at the telomeres associated with the ALT pathway in the U2OS cell line (Dunham et al., 2000; Kass-Eisler and Greider, 2000; Lundblad, 2002; Nabetani et al., 2004; Muntoni and Reddel, 2005; Jiang et al., 2007). Inasmuch as these activities are associated with the APB complex, it appears likely that its structure persists during the telomere lengthening process for some time as indicated in the scheme in Figure 9. When the number of repeats becomes sufficiently high the shelterin proteins can bind again to reestablish the telosome complex at the newly synthesized telomere repeat region. This would then promote interactions with other chromatin domains to induce a state of low mobility as observed for the majority of telomeres.

Figure 9.

Model for telomere mobility and APB formation in the context of the ALT mechanism. In telomeres with sufficient repeat length the telosome protein complex mediates the anchoring to another chromatin domain. This results in a confined mobility of the telomere that is similar to that of other chromatin loci. When the telomere repeat length is reduced below a critical threshold, the telosome is disrupted, and the anchoring is lost. In this state the telomere displays a significantly extended mobility. The telomere associates with a PML body to form an APB complex, in which the telomere length is increased in a process that involves DNA recombination and repair. The telosome can reform on the extended telomere repeat sequence to reestablish the telomere anchoring associated with the reduced mobility state.

Abbreviations used:

APB

ALT-associated PML bodies

ALT

alternative lengthening of telomeres

FISH

fluorescence in situ hybridization

MSD

mean squared displacement

PML-NB

promyelocytic leukemia nuclear bodies

PNA

peptide nucleic acid.