Rapid telomere motions in live human cells analyzed by highly time-resolved microscopy
© Wang et al; licensee BioMed Central Ltd. 2008
Received: 28 May 2008
Accepted: 27 October 2008
Published: 27 October 2008
Skip to main content
© Wang et al; licensee BioMed Central Ltd. 2008
Received: 28 May 2008
Accepted: 27 October 2008
Published: 27 October 2008
Telomeres cap chromosome ends and protect the genome. We studied individual telomeres in live human cancer cells. In capturing telomere motions using quantitative imaging to acquire complete high-resolution three-dimensional datasets every second for 200 seconds, telomere dynamics were systematically analyzed.
The motility of individual telomeres within the same cancer cell nucleus was widely heterogeneous. One class of internal heterochromatic regions of chromosomes analyzed moved more uniformly and showed less motion and heterogeneity than telomeres. The single telomere analyses in cancer cells revealed that shorter telomeres showed more motion, and the more rapid telomere motions were energy dependent. Experimentally increasing bulk telomere length dampened telomere motion. In contrast, telomere uncapping, but not a DNA damaging agent, methyl methanesulfonate, significantly increased telomere motion.
New methods for seconds-scale, four-dimensional, live cell microscopic imaging and data analysis, allowing systematic tracking of individual telomeres in live cells, have defined a previously undescribed form of telomere behavior in human cells, in which the degree of telomere motion was dependent upon telomere length and functionality.
Telomeres, essential for protecting chromosome ends [1–3], consist of tandem telomeric DNA repeats bound by multiple proteins that collectively 'cap' the telomere (reviewed in ). A minimum length of telomeric repeats is necessary to support this protective function. Telomeric repeats are replenished by the cellular ribonucleoprotein enzyme, telomerase [2, 5]. Most human cancer cells have high telomerase activity. In contrast, normal human cells have naturally limited levels of telomerase that can lead to telomere shortening over time. Many aspects of telomeres remain incompletely understood, especially their dynamic properties and behavior over time. At one extreme, over the decades of human lifespan, telomeres generally shorten, with rates and extents that have been associated with disease progression and risk [6–9]. Here we report new findings on telomere behavior at the other end of the timescale spectrum: the dynamics of individual human telomeres in living cells analyzed at 1-second time resolution.
The dynamics of the nuclear contents are complex and determined by multiple contributing factors. First, the diffusion of non-interacting particles experimentally inserted into the nucleoplasm is much slower compared with their diffusion in water . The apparent diffusion coefficient of these free inserted particles has strong size-dependence, beyond that predicted by Stokes' Law [10, 11]. This deviation from the values expected from Stokes' Law has been attributed to the characteristic free space between structural elements. In the specific case of a polymer molecule, the diffusion within entangled concentrated polymers is described as snake-like reptation motion confined by a tube formed by the other polymer molecules . Reptation is orders of magnitude slower than diffusion in dilute solutions. Secondly, the dynamics of chromatin within the dense nucleus are even further slowed down, probably due to chromatin architecture and its specific interactions. Chromosomal dynamics include oscillatory-like fast motions as well as apparently diffusive motions. Labeled chromatin sites move with seconds timescale oscillations , as well as showing constrained diffusion [11, 13, 14]. This motion is highly variable for different regions within the same nucleus or even the same chromosome ([15, 16], and references cited therein). This variability has been related to interactions with nuclear structures (nuclear envelope , nuclear pores, nucleoli , promyelocytic leukemia (PML) and Cajal bodies [19–21], nuclear matrix, DNA-associated proteins, telomeres [14, 22]) and to the conformation of the chromatin itself (centromeres , heterochromatin, DNA unwinding at highly expressed genes, histone acetylation , DNA methylation and cell cycle-related chromosome condensation ). In this study, we focused on short timescale, apparently diffusive, motion of the telomeres. We did not attempt to explain theoretically the diffusion coefficient values measured. Rather, we concentrated on differences found between telomeres and their relation to other measured properties (notably position and size) and to controlled perturbations.
A recently developed microscopy platform was employed. This platform (and variants of it) is applicable for a variety of cell-biological observations. Here we applied it to analyze rapid telomere dynamics, in order to compare the motions of telomeres of different known lengths and functionality. As the present study focused on short timescales, where the motion is described closely by diffusion laws, the motions are expected to reflect telomere interactions relatively directly. We report that in living cells, telomeres exhibit novel, heterogeneous rapid motions that directly reflect impaired telomere functionality.
High-resolution, three-dimensional (3-D, xyz) spatial image datasets were acquired within rapid timeframes (100 ms for a stack of 17 consecutive focal planes with 0.5 μm focal spacing, encompassing the entire 6–8 μm high nucleus, every second for 200 seconds) using an optical microscope platform (OMX) (Figures 1D and 1E, Additional file 2). By analyzing local motion over such short times, we avoided the complexities of reaching the constrained motion thresholds found in yeast, Drosophila and mammalian nuclei [11, 13, 18], and could apply diffusion equations to quantify the telomere motion tracks as a function of time.
Two separate software programs (Track4D and TrackIt4D) were custom written to analyze the image datasets for large numbers of cells. Typically, per UMUC3 nucleus, 80 or more spots were resolved, of which 40 to 60 unambiguous telomeres could be tracked for extended periods within each total 200-second data collection timeframe (see C-E in Additional file 1). The total numbers of identifiable telomeres were consistent with those expected if the great majority of fluorescent spot signals are from single telomeres. Hence in these cells, unlike telomeres in yeast [30, 31], it is unlikely that in general, multiple telomeres cluster into single spots. For the experiments reported here, the complete four-dimensional (4-D) datasets (3-D and time) for every nucleus were displayed and analyzed in six different ways and are shown as appropriate throughout this paper in one or more of the following ways: (1) as movies (Additional files 3, 4, 5); (2) as kymographs – plots showing the movement of each individual telomere as its position in the xy-plane as a function of time T (see A and B in Additional file 6, Additional files 7 and 8); (3) as a trajectory of each telomere, showing its position in 3-D space as a function of time for 200 consecutive seconds (see E in Additional file 1, C in Additional file 6, Additional files 7 and 8); (4) as the end-to-end (E2E) distance in 3-D space as a function of time (each line in the plot tracks the distance of an individual telomere at time T seconds away from its original starting point at time 0) (see D in Additional file 6, Additional files 7 and 8); (5) as the cumulative path distance traveled by a telomere between time 0 and time 200 seconds (see E in Additional file 6, Additional files 7 and 8); (6) as the average E2E distance for each telomere for the time interval traveled (see F in Additional file 6, Additional files 7, 8, 9). E2E measurements were used to calculate D, the effective diffusion coefficient, according to Einstein's diffusion equation, E2E = (6DT)1/2 (T is the time in seconds). Hence telomere motion was quantified for every telomere and the population of telomeres, both in a single cell and in multiple cells.
The mean and median of these UMUC3 cell telomeric D values fell in the general range for diffusion coefficients previously reported for chromatin segments in internal chromosome regions [11, 13, 32] (see A and B in Additional file 6, Additional files 7 and 8). Notably, the histogram based on around 400 telomeres exhibited a left-skewed distribution of D values with its mode at around 2.3 ± 0.6 × 10-4 μm2/second (Figure 2E). We estimated the error bars in such histograms using a non-parametric statistics bootstrap procedure (see B in Additional file 10). When the histograms were plotted with a log scale for the D values, the distribution still showed asymmetrical skewing (shown in C in Additional file 10 for the data in the histogram in Figure 2E). This indicates that the skewing seen in Figure 2E does not describe an underlying additive or multiplicative random process. Based on the modal value, we arbitrarily defined the sub-set of telomeres having an average D value of less than 2.3 × 10-4 μm2/second as 'calm' telomeres (moving less than 0.37 μm during 100 seconds tracking). Within the same nucleus, other telomeres, defined as 'jittering' telomeres, moved back and forth rapidly and irregularly (some having D > 10-3 μm2/second, corresponding to their moving on average more than 0.77 μm during 100 seconds) (Figure 2C and 2E, Additional file 3).
The telomere motion can be described as 'a constrained non-directional walk'. Unlike non-constrained free diffusion, which predicts a linear relationship between the square of the distance traveled by a particle, the mean square distance (MSD), and the time duration T [32, 36], the MSD curve in the internal regions of a chromosome generally reaches a plateau of T > 200 seconds in yeast and mammalian cells [11, 18]. This is interpreted as 'constrained diffusion'. However, our telomeric analyses, performed for only 200 seconds, were done before the E2E distances (around 0.37 μm) reached any evident chromosome territory confinement limit, estimated to be 0.6 μm in mammalian and other eukaryotic nuclei [11, 32]. Therefore, below 200 seconds, the assessment of telomere movement by the effective diffusion coefficient is justified (see A in Additional file 16). The movement of individual telomeres was also clearly not directional (Figure 3D). In yeast, telomeres and centromeres are tethered at or near the nuclear envelope  and reviewed in Gasser , with obvious implications for telomere motility. As mentioned above, a study on the motion of non-telomeric loci in mammalian cells (examined with a 1-minute time resolution) found that these motions were more constrained when the loci were close to the nuclear periphery . In contrast, no attachment of telomeres to the nuclear periphery has been reported in mammalian cells. Consistent with this, we observed that telomere motion was statistically no lower at the peripheral region of a nucleus than deep in the same nucleus (Figure 3E). This is shown by inspection of the nuclear localization of individual telomere trajectories, approximating the nucleus as an ellipsoid and using the Mahalanobis distance from the center to define proximity to the nuclear periphery (see B in Additional file 16). Hence, jittering and calm telomeres may occur randomly throughout the nucleus.
Two types of observations suggested that the more rapid telomere motion was related to telomere shortness. First, in three other human cancer cell lines with different mean telomere lengths, we noted that the longer the average bulk telomeric DNA length, the lower the telomere motions (see A and B in Additional file 13). This relationship between mean telomere spot signal intensity and telomeric DNA length within (see below) and across different cell types provided additional corroborative evidence that the signals were bona fide telomeres. Second, comparing telomere motions within each single UMUC3 nucleus, there was an inverse relationship between telomere fluorescence intensity and motion: short telomeres, identified as those with lower intensity values, contributed the majority of the jittering telomeres (see C and D in Additional file 13). Finally, plotting the entire data for a collection of 20 different parental UMUC3 cells in a clonal sub-line expressing GFP-tagged TRF1 shows an inverse linear relationship between the effective diffusion constant (D) of a telomere with signal intensity, across the spectrum of telomeric signal intensities (except at the extreme upper signal intensity boundaries) (see Additional file 17). These data showed that in these UMUC3 cells, a two-fold decrease in telomere length (as shown by the telomere signal intensity) corresponded to a 2.7-fold increase in effective diffusion coefficient.
We tested whether there was a causal relationship between telomere shortness and higher telomere motion. To change the average bulk telomere length with the minimal perturbation possible while keeping the same cell background, we stably expressed telomerase RNA (WT-hTER) above its endogenous level in UMUC3 cells . As shown, this extends telomeres moderately, by three to four kilobase pairs, in these cells (Figure 4B). Telomeres became less motile and more homogeneous in their movement than in the parental cells or cells that had received the control empty vector. The quantitative changes in motility were evident from two different parameters: both the modal D value and the number of highly motile telomeres decreased (see E-G in Additional file 13). The great majority of the telomeres (as shown by the position of the modal peak) became shifted to the lower D value when the great majority of the telomeres in the UMUC3 cells had been shifted to a longer telomeric DNA tract length (the mean telomere DNA length seen in the Southern blot in Figure 4B). The cell doubling time was the same in all these cells (parental, vector-control, and WT-telomerase RNA overexpressing cells), as reported previously . Together, these results ruled out the possibility that a shift in the cell population cell cycle could account for the difference seen. Hence, shifting the bulk telomere population length upwards dampened telomere motility.
Even with adequate telomerase, as in the human cancer cells analyzed here, telomeres cycle between capped and uncapped states as a normal part of their maintenance . It was proposed that such temporarily uncapped, often shorter, telomeres are those acted on by telomerase [3, 41], thereby replenishing their length and restoring their ability to be capped. We hypothesized that the greater motility of shorter telomeres in the experimental settings described above could reflect such a partially uncapped status. We therefore tested whether uncapping of telomeres by an independent means could cause increased motion. We induced rapid and irreversible telomere uncapping by an independent means: expressing a mutant-template telomerase RNA (MT-hTer). Such MT-hTers cause telomerase to add terminal mutant repeats, which are designed to be unable to bind sequence-specific protective telomeric proteins. MT-hTers were shown to rapidly de-protect (uncap) telomeres and accumulate colocalizing DNA damage foci on telomeres [38, 39, 42]. Notably, MT-hTers uncap telomeres without detectable shortening of the rest of the telomeric DNA tract or depletion of bulk telomeric proteins TRF1 or TRF2 from the telomeres [38, 39, 42]. Such expression of an MT-hTer ('MT-hTer 47A') in UMUC3 cells greatly increased the fraction of jittering telomeres, as seen by kymograph analyses, their expanded trajectories and large E2E and D values (Figure 4C, A-F in Additional file 14). We eliminated the possibility that this exaggerated telomere movement induced upon telomere uncapping is a result of the general DNA damage response state of these cells. This was done by examining telomere motions in cells treated with a DNA alkylating agent, methyl methanesulfonate (MMS), which produces comparable numbers of DNA damage foci that do not colocalize with telomeres (see E in Additional file 14). Under MMS treatment, telomeres remained as restrained in their movement as the controls (Figure 4C). Measurement of nuclear volumes upon these treatments established that cell/nuclear volume shrinkage or expansion did not occur, excluding this contribution to changes in telomere motions. Hence, uncapping of telomeres by MT-hTer 47A increases telomere motion; in contrast, treatment with MMS did not.
The ability of the OMX to collect 4-D live datasets with high time resolution (every second for a short timeframe of 200 seconds) has allowed examinations of large numbers of data points for statistically robust analyses on telomere dynamic behavior. This enabled evaluation of the apparent telomere diffusion coefficient without the complications of long-term constraint diffusion imposed by the restriction of chromosomes to limited spatial nuclear domains. In this setting, we have discovered sub-sets of telomere dynamics within human cancer cell nuclei. Previous work has shown energy dependence of the motions of PML bodies in live human cancer cells, but such previous analyses have not addressed the rapid high-resolution motions of validated telomeres [14, 29].
A novel finding is that chromosome end motion is highly heterogeneous and inversely proportional to telomere length. Furthermore, experimentally uncapping telomeres, but not the general DNA damage caused by MMS treatment, greatly increased telomeric motions (Figure 4D, Additional file 18). It is possible that the telomeric uncapping resulting from incorporation of mutant repeats causes the increased telomeric motion and that DNA damage at other genomic locations in general does not. Alternatively, MMS damage may engage a different DNA response pathway from that elicited by telomere uncapping.
Even in untreated cells there was a wide range of telomere motility, with the shorter telomeres more motile on average than longer ones in the same nucleus. Experimentally increasing the telomerase activity level, with concomitant lengthening of the telomeres, damped down their movement and reduced the number of jittering telomeres in the population. Although it could be conceived that longer telomeres could allow more binding sites for (unspecified) protein attachments, such load addition to just a few kilobase pairs of telomere length would be insignificant considering the size of the chromosome. We therefore propose the model (Figure 4E) that short and, therefore, uncapped telomeres, or telomeres uncapped by mutant repeat addition, with resultant loss of protection, become more motile and move at a greater speed, which enables them potentially to explore and possibly sample a larger volume in the nucleus. The greater motility could occur during or upon uncapping as a result of telomeres detaching from a nuclear sub-structure, or active (and hence potentially energy-dependent) decompaction of a relatively large sub-telomeric (heterochromatic) region, hence altering the elasticity of this region and allowing greater freedom of telomere movement. The data also are consistent with the possibility that telomeres may not specifically attach to an immobile structure in the nuclear matrix, since one specific class of internal arm regions of chromosomes moved slower than telomeres.
Telomeres and their replenishment machinery including telomerase have elicited interest as possible targets for use in anti-cancer therapies . Therefore, understanding the basic behavior of single telomeres in living cells will be an important underpinning of any use of such potential therapeutics. Furthermore, the critically short telomeres in non-cancerous cells without sufficient telomerase eventually become uncapped, triggering cellular senescence and apoptosis, which are thought to contribute to aging and mortality (reviewed in Shay and Wright ). Tracking the very rapid, seconds-scale time events at telomeres in cells in real time will thus provide more understanding of the heterogeneity of the dynamics of this complex biological system.
The examination and systematic analysis of telomere dynamics at 1-second time intervals for 200 consecutive seconds reported here has shown that chromosome ends show distinct movements that are different from chromosome-internal HP1α heterochromatin regions. Specifically, individual telomeres within a single live cell nucleus displayed wide heterogeneity in their degrees of movement over short time intervals, and quantitative analyses show that such telomere motion consists of a 'constrained non-directional walk'. The class of chromosome-internal heterochromatin regions examined in this study showed less, more homogeneous motion. Short telomeres and experimentally 'uncapped' telomeres had higher motility than longer telomeres. Conversely, experimentally extending telomere length lowered telomere motility. Hence, integrating physical/mathematical and biological approaches has shown new properties of telomeres reflective of telomere functionality. The systems developed here to accurately and statistically track telomeres should be generally applicable to any punctate chromatin (or cellular) foci in live cells.
The N-terminus of the TRF1 and TRF2 cDNAs were fused in-frame with enhanced GFP (Clontech, Mountain View, CA, US). The C-terminus of the TRF1 and TRF2 cDNAs were fused with internal ribosome entry site-hygromycin (IRES-hygromycin; Clontech, Mountain View, CA, US). The resultant fusion constructs were sub-cloned into the pHRCMVGFPWSin18 vector under the CMV promoter (courtesy of Dr Didier Trono) . The CFP-HP1α was a kind gift from Dr Tom Misteli . WT-hTER and 47A MT-hTer in lentiviral vectors were generated as in Li et al . Briefly, WT-hTER was polymerase chain reaction-cloned from human genomic DNA and sub-cloned into the Blg II/Sal I site in pIU1-T7 plasmid (generously provided by Dr Edouard Bertrand). MT-hTer was then generated by site-directed mutagenesis to create an hTer with the RNA template sequence 3'-CAAA CCCAAA C-5' 47A-MT-hTer. Both the WT-hTER and 47A MT-hTer were cloned into the pHRCMVGFPWSin18 lentiviral vector under the IU1 promoter (see A in Additional file 14).
Lentivirus was generated as described previously . Briefly, 5 μg of pMD.G plasmid (virus envelope plasmid), 10 μg of pCMVDR8.91 (virus packaging plasmid), and 15 μg of the lentivector (GFP-TRF1, GFP-TRF2, lentiviral empty vector, WT-hTER, or 47A MT-hTer) were cotransfected into amphotropic 293T cells using the Fugene 6 transfection method (Roche, Indianapolis, IN, US). Virus was harvested 48 and 72 hours post-transfection and filtered through 0.45-μm filters. Virus titers were calculated in 293T cells by counting the number of GFP-expressing foci divided by the dilution factor (106-107 transduction units/ml viral titer is achieved). For virus infection, culture cells were incubated with culture medium-diluted virus supernatant supplemented with 8 μg/ml of polybrene (Sigma, St Louis, MO, US) for 8 hours. A virus titer of 20 transduction unit/cells were used for a > 95% infection efficiency. GFP color starts to appear after 13 hours of infection and low expression monoclonal cell population stables were purified with hygromycin (200 μg/ml) (Invitrogen, Carlsbad, CA, US) or puromycin (2 μg/ml) (Sigma, St Louis, MO, US) selection for 5 days. Weak expression stables were chosen for Western analysis and imaging analysis. For cell growth measurement and cell cycle analysis, 2 × 104 cells were reseeded in triplicates in six-well plates 48 hours after virus infection, and the cell numbers were counted every day or cells harvested for propidium iodide (Molecular Probes, Carlsbad, CA, US) staining and fluorescence activated cell sorting analysis.
Clonal lines with the lowest average expression from Western blotting analysis were chosen for comparison with the endogenous level of TRF1 (60 to 65 kilobase pairs) or TRF2 (66 kilobase pairs) on 10% or 7.5% sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis gels for telomere dynamic analysis. For the low-expressing stable lines used in these analyses, the cell proliferation rates, morphology and cell cycle distribution (around 80% of the cells were in G1 in these cultures) were verified as showing no significant differences from the parental line (see A in Additional file 1). CFP-HP1α plasmid was transiently transfected into UMUC3 cells using Fugene 6 transfection and cells were imaged 36 hours later.
Mammalian cancer cell lines were used in these analyses, as cell lines are known generally to retain the genomic characteristics of the primary tumor . For all analyses, we used cell lines that do not contain PML bodies to observe systematically telomere motion in live cells. PML bodies are free nuclear particles that are not part of the chromosomes, and contain telomeric proteins and other telomeric components [27, 28].
The endothelial bladder carcinoma cancer cell line UMUC3 was purchased from the American Type Culture Collection (ATCC), (#CRL-1749). We observed that the UMUC3 karyotype was stable over the course of this and other studies in our laboratory [39, 47], consistent with the near-triploid karyotype and modal chromosome number of 80 described for this line by the supplier (ATCC). The HeLa cell line (cervical adenocarcinoma cancer cell line) was purchased from ATCC (#CCL-2). H1299, an epithelial lung carcinoma cancer cell line, was purchased from ATCC (#CRL-5803). Cells were cultured under standard cell culture conditions (37°C, 100% relative humidity, 5% CO2) in Dulbecco's modified Eagle's medium (Gibco, Langley, OK, US) supplemented with 10% fetal bovine serum (Gibco, Langley, OK, US), 2 mM L-glutamine (Gibco, Langley, OK, US), 100 U/ml penicillin and streptomycin (Gibco, Langley, OK, US). LOX cells (a human amelanotic melanoma cancer cell line) were cultured in RPMI medium (Gibco, Langley, OK, US) with the same supplements as above. Clonal cells expressing either GFP-TRF1 or GFP-TRF2 from each cancer cell line were generated by picking a single cell colony. HEPES (N-2-hydroxyethylpiperazone-n-2-ethanesulfonic acid) buffer (pH 7.4) was added to cell culture medium at a final concentration of 20 mM to avoid the need for carbon dioxide gas before imaging.
ATP depleting medium was made in culture medium, supplemented with 50 mM 2-deoxy-D-glucose (Sigma, St Louis, MO, US) and 6 mM sodium azide (Sigma, St Louis, MO, US). After washing away azide, cells were allowed to rest for 30 minutes before imaging again (azide wash). The methylating agent MMS (Sigma, St Louis, MO, US) was added at a concentration of 0.01% for 1 hour before imaging and immunofluorescence. WT-hTER lentivirus was used to infect the UMUC3 cells for 20 days, and analyzed for telomere length before imaging. 47A MT-hTer lentivirus was used to infect the cells for 3 to 6 days before imaging and immunofluorescence.
For immunofluorescence, cells were fixed in chilled 4% paraformaldehyde, permeabilized with 0.5% NP-40 in phosphate buffered saline (PBS), blocked in 5% bovine serum albumin (BSA), incubated with the primary antibody in 5% BSA in PBS, washed, and then incubated with either Alexa fluor 568 or Alexa fluor 488-conjugated secondary antibody (either anti-mouse or anti-rabbit against the primary antibody) (Molecular Probes, Carlsbad, CA, US) in 5% BSA in PBS, washed, and finally in 0.15 μg/ml 4',6-diamidino-2-phenylindole (DAPI) before mounting. Indirect immunofluorescence showed the expected complete colocalization of telomeric signals from GFP-TRF1 with anti-TRF2 antibody or GFP-TRF2 with anti-hRap1 antibody (Figure 1C). For HP1α plasmid immunofluorescence, no DAPI staining was performed.
For immunoblotting, cells were lysed in radioimmunoprecipitation assay buffer (50 mM Tris/HCl (pH 7.4), 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate and 0.1% SDS, supplemented with a complete protease inhibitor cocktail from Roche, Indianapolis, IN, US). Lysates were quantitated using Bradford assay (BioRad) and 15 μg of protein were mixed with SDS sample buffer (0.08 M Tris (pH 6.8), 2.0% SDS, 10% glycerol, 0.1 M dithiothreitol, 0.2% bromophenol blue) before boiling for 5 minutes and then subjected to immunoblotting with the antibodies.
The following antibodies were used for indirect immunofluorescence analyses and Western blotting: mouse monoclonal antibodies TRF2 (#611200, BD, Franklin Lakes, NJ, US), β-actin (#A5441, Sigma, St Louis, MO, US), rabbit polyclonal antibodies 53BP1 (#A-300-273A-3, Bethyl, Montgomery, TX, US), ATM pS-1981 (#AF1655, R&D systems), Rap1 (#A300-306A, Bethyl, Montgomery, TX, US), γ-H2AX pS-139 (#NB100-384, Novus Biologicals, Littleton, CO, US) and TRF1 (#Ab1423, Abcam, Cambridge, MA, US).
A 32P-labeled telomeric (CCCTAA)3 probe was purified using the microsystem spin column (Bio-Rad) and hybridized to HinfI and RsaI-digested human genomic DNA at 37°C overnight. DNA samples were extracted from cells using a genomic DNA purification kit (Gentra, Minneapolis, MN, US) and separated on a 0.65% agarose gel. Southern results were exposed with a phosphor-imager.
Cells were seeded on to a glass-bottomed cell culture dish (Bioptechs, Butler, PA, US) for OMX microscopy. This microscope is described in Schermelleh et al . A ×60/1.25 water-immersion objective (Olympus) was used to collect 3-D stacks of 17 focal sections (an entire nucleus of 6–8 μm) with 0.5 μm focal step, once a second for 200 seconds. OMX requires 100 milliseconds (the fastest reported imaging speed to date) to acquire one complete 3-D live cell nucleus stack. Cells were exposed for 5 milliseconds to excitation light from a 500 mW 488 nm laser attenuated to 10% transmission with a neutral density filter. The total duration of timelapse imaging is limited by photobleaching (at the end of such data collection, fluorescence intensity is roughly half) (see D in Additional file 1). The high resolution 4-D data analysis allows the development of the system to accurately and statistically track any punctate chromatin foci including telomeres reported here.
Telomere tracking requires a robust image processing algorithm to localize the fluorescent spots in noise-limited minimal exposure images. We applied a 3-D Wiener filter with telomere modeled by a Gaussian with full-width half-height of 2.5 pixels. The filtered image was then subjected to a non-linear thresholding procedure based on intensity and local contrast threshold criteria for each pixel and its six 3-D neighbors. This algorithm reliably segmented telomere spots throughout the duration of the experiment despite bleaching.
3-D contiguous components analysis provided for each telomere a set of quantitative parameters, including local background, center of 'mass' (weighted by the background-subtracted fluorescence intensities), position of maximum intensity, volume, the number of boundary pixels, and the background-subtracted integrated intensity. Due to the high and inhomogeneous background, the local background was determined from repeated dilation of each telomere segment to define the asymptotic value for the averaged intensity out of the segments as the background value. The parameters are then used to reject non-telomere spots (volume, intensity and surface-to-volume outliers) and time-track telomere motion. For dense time-sampling, tracking is simply achieved by searching for the closest telomere spot in the preceding and following time points.
We corrected for nucleus drift or rotation during the live imaging, which might have confounded interpreting telomere mobility (see E in Additional file 1).
For each telomere track (each line in the Track4D plot tracks an individual telomere, but colors of telomere tracks are random between plots), we calculated the 3-D distance to position at time zero (E2E) and the accumulated root-mean-square distance between consecutive times. The effective diffusion coefficient D was calculated from E2E using Einstein's diffusion equation, E2E = √(6DT) (T is the time in seconds). D is based on E2E distances > 0.3 μm, and sampling at 150 nm steps (1/6 of the microscope resolution in z).
The TrackIt4D (tracks analysis) program calculated the accumulated path distance (PATH). It also generates a histogram plot for the D values of each single nucleus. An accumulated histogram was generated based on around 400 telomeres or heterochromatin spots. The distance that a cell nucleus has drifted during imaging is indicated by a line near the bottom of the x-axis (timescale). Drifts are corrected when quantifying telomere motion.
Average E2E was calculated independently. Telomere foci were first sampled above the background noise using the custom-written Python software and telomere mobility was tracked, calculated (Additional file 9) and plotted. By averaging E2E for all pairs with a given time difference, there is less noise. Briefly, telomere motion statistics were calculated from average E2E distance over sub-intervals. The rate of timelapse imaging was 1 second in all cases. All distances traveled by the same telomere over time intervals of the same length were averaged together and plotted. Telomeres tracked for N seconds were plotted for intervals up to N/2 seconds (N is any number). Global nucleus motion was estimated from the drift in the average spots position and subtracted from all spots. Rotational motion was found negligible for the tracking times used.
Trajectory images were created and plotted on top of original nucleus images after TrackIT4D tracking. The association of each telomere to the nuclear periphery was approximately assessed by viewing the nucleus as an ellipsoid and using the Mahalanobis distance from the center to define proximity to the nuclear periphery. Nuclear volumes were estimated by drawing the nuclear periphery contour with polygons and evaluation of the area and volume of the nucleus in all optical sections using the Priism software suite.
Error was estimated by statistical bootstrapping. Bootstrapping of histograms of D values for non-parametric statistical comparison was done by repeatedly and randomly splitting the histograms of 400 telomeres into two sub-groups of 200 telomeres each, and averaging mean and standard deviation for the mode or the Komolgorov-Smirnov (K-S) comparison test between them. Diffusion coefficient histograms were accumulated for about 20 nuclei (>400 telomere tracks) for each experimental condition. For skewed, certainly non-normal distributions, K-S statistics have to be used. The difference between histograms was scored by the K-S comparison test. Error bars were estimated from statistical bootstrap. A K-S score of more than three times the error (standard deviations) from bootstrapping procedures is considered significant.
alternative lengthening of telomeres
American Type Culture Collection
bovine serum albumin
green fluorescent protein
optical microscope platform
phosphate buffered saline
sodium dodecyl sulfate.
We are grateful to Drs Yuri Strukov, Wallace Marshall, Jennifer Fung, Baojie Li and Jack Griffith for helpful and stimulating discussions. We thank Eric Branlund, Sebastian Haase and Maura Clancy for technical assistance. We also thank Drs Tom Misteli and Shang Li for sharing plasmids and technical advice. This work was supported by grants from the NIH to JWS (GM25101) and EHB (CA96840). Support to EHB was also provided by the Bernard Osher Foundation and to JWS by the Sandler Program in Basic Sciences. XYW was supported by a postdoctoral fellowship from the Susan G Komen Foundation.
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.