Monitoring of telomere length regulation has become an important aspect of stem-cell research, studies on cellular senescence, and cancer research. A number of techniques have been developed to study telomere length in a quantitative manner. The standard procedure is Southern blot hybridization (1). The quantitative fluorescence in situ hybridization (Q-FISH) on metaphase chromosomes allows the quantification of the telomeric FISH signals on individual chromosomes (2–4). Since fluorescence intensity of the telomeric signals was found to be proportional to the size of telomeric repeats, Q-FISH is now widely used (2). For measuring telomere length in interphase nuclei, flow cytometry or fluorescence microscopy can be applied(5–7). Nonadhesive hematopoietic cells seem to be better suited for flow cytometric analyses than solid tumor cells (8, 9). Therefore, for the analysis of nonhematopoietic cells, interphase FISH is the method of choice. However, the fluorescence microscopical method to quantify telomere length in interphase nuclei has so far only been performed manually on a restricted number of cells (6, 7). To combine the positive aspects of flow cytometric measurements with the ability to quantify individual nuclei by fluorescence microscopical examination, we took advantage of fluorescence-based automatic microscope (FLAME). Thus, all the advantages of interphase measurements, i.e., the analysis of individual cells and the applicability to nonproliferating cells, can be combined with the analyses of statistically relevant cell populations. To develop a reliable method to automatically quantify telomere length in interphase nuclei, we applied a two-color hybridization assay and measured the fluorescence signals with FLAME. We then described the parameters essential for intra- and interexperimental comparisons. We performed intensity measurements in interphase nuclei and compared the results of single channel measurements of the target probe with the results obtained after introducing an internal reference and performing double channel measurements. We validated the quantitative-FISH on interphase nuclei (IQ-FISH) method by measuring telomere lengths of different adult individuals, by measuring neuroblastoma cell lines and by correlating the results of our method with the results obtained by other methods for telomere length measurements. Furthermore, telomere lengths were evaluated in senescent tumor cells and were compared to those of nonsenescent tumor cells. Finally, we investigated different osteosarcoma samples to test whether the ALT (alternative lengthening of telomeres) type can be detected by this automatic approach.
To benefit from the fluorescence-based automatic microscope (FLAME), we have adapted a PNA FISH technique to automatically determine telomere length in interphase nuclei. The method relies on the simultaneous acquisition of pan-telomeric signals and reference probe signals. We compared the quantitative figures to those for existing methods, i.e. Southern blot analysis and quantitative FISH (Q-FISH). Quantitative-FISH on interphase nuclei (IQ-FISH) allows the exact quantification of telomere length in interphase nuclei. Thus, this enables us to obtain not only exact information on the telomere length, but also morphological and topological details. The automatic measurement of large cell numbers allows the measurement of statistically relevant cell populations. © 2005 Wiley-Liss, Inc.
MATERIALS AND METHODS
Neuroblastoma cell lines STA-NB-3, -7, -9, -10 (10), -12, -15, and SH-SY-5Y (kindly provided by J. L. Biedler, NY) were grown in RPMI 1640 (Gibco-BRL, Grand Island, NY) supplemented with 10% fetal calf serum (FCS) at 36.6°C and 5.1% CO2.
Mononucleated cells of heparinized peripheral blood were isolated by lymphoprep (Nycomed, Roskilde, Denmark) gradient centrifugation and washed three times in PBS. Cytospin preparations of the different neuroblastoma cell lines, seven osteosarcoma tumors, and peripheral blood from three different individuals (age 28, 29, and 39) were prepared on glass slides by the use of the Hettich cytocentrifuge system (Rotofix 32, Universal 16 A, Hettich, Kirchlengern, Germany).
Fluorescence In Situ Hybridization
Q-FISH (2) was performed using a Cy3-labeled telomeric PNA (DAKOCytomation, Vienna, Austria) probe and, as internal reference, a chromosome 2 centromere-specific FITC-labeled PNA probe (kindly provided by DAKOCytomation, Denmark) by essentially following the protocol from Lansdorp et al. (2).
Slide Scanning and Cell Analysis by IQ-FISH
Slide scanning, cell identification, intensity measurement, and quantification of telomere length were performed by the use of the fluorescence-based microscopic scanning system, Metafer4 (MetaSystems, Germany). This scanning system is based on a fully motorized Axioplan 2 microscope (Zeiss, Germany), a motorized eight-slide scanning stage (Märzhäuser, Germany), and a high-resolution CCD camera. MetaCyte, a software for single-cell analysis, is integrated in the Metafer4 system. The MetaCyte software classifiers enable the setting of capture and exposure parameters as well as image and cell processing steps. These classifiers allow the determination of cell selection criteria and analysis of specific cell and signal characteristics, such as cell area, aspect ratio, concavities, or signal intensities. For the telomere length quantification, we measured the FITC and Cy3 intensities in single nuclei. The slide scanning and cell analysis procedures were performed by using the 63× objective. For details of the scanning procedure, see Narath et al. (11). Automatic focusing based on a local contrast criterion was performed in the DAPI counterstain channel, defining the nucleus center focus plane. Then a stack of nine images at z positions symmetrical to the center plane and with a spacing of 0.5 μm were captured in the FITC and Cy3 channels. From each stack, a 2D image with all signals in focus was created using a pixel by pixel selection algorithm based on an absolute intensity criterion. Finally, a local background correction was done by applying a standard Top Hat Filter. Cells containing no FITC signals were excluded from further analysis. The nine focus planes with a height of altogether 4.5 μm were sufficient to cover the size of the analyzed nuclei. The Cy3 (pan-telomeric probe) intensity was measured by an appropriate filter. A minimum of 100 nuclei were scanned for every sample. The fluorescence intensities of the hybridized pan-telomere probe (Cy3) and the chromosome 2 centromere probe (FITC) were displayed at the bottom of every gallery image (Fig. 1). In addition, the ratio between both intensities Cy3/FITC was presented in the upper right corner of every image. Cell numbers were given in the upper left corner. Relocation and visual inspection were performed when unusual Cy3 or FITC intensities were noticed. By a mouse click on the image-of-interest, the system relocates the selected cell in the microscope for visual inspection. Nuclei with poor FITC intensity were excluded from evaluation. In our experiments, we defined 30 IU (intensity units) as cut-off level, but every investigator has to set the cut-off levels individually according to the experimental conditions. A final number of at least 50 nuclei per sample was retained for evaluation. The mean value of the fluorescence ratios of all cells analyzed was automatically calculated. The scanning time depended on the cell density on the slide and varied from 6 to 10 nuclei/min. For the interactive evaluation, about 15 min per slide were needed.
Number of Analyzed Slides and Nuclei per Sample
Peripheral blood of three healthy adults
For sample A, we analyzed nine slides with 174/600 nuclei per slide (in summary 3,499 nuclei). For sample B, we analyzed four slides with 240/420 nuclei per slide (in summary 1,184 nuclei). For sample C, we analyzed seven slides with 127/242 nuclei per slide (in summary 1,070 nuclei). The number of analyzed nuclei was determined by the number of nuclei per slide and the hybridization quality.
Neuroblastoma cell lines
For STA-NB-3, we analyzed nine slides (in summary 152 nuclei). For STA-NB-7, we analyzed nine slides (in summary 204 nuclei). For STA-NB-9, we analyzed two slides (in summary 168 nuclei). For STA-NB-10N, we analyzed six slides (in summary 1,441 nuclei). For STA-NB-10F, we analyzed six slides (in summary 967 nuclei). For STA-NB-12, we analyzed four slides (in summary 574 nuclei). For STA-NB-15, we analyzed three slides (in summary 104 nuclei). For SH-SY-5Y, we analyzed two slides (in summary 350 nuclei).
Telomere Length Measurement by Q-FISH on Metaphase Spreads
Fluorescence signals were analyzed using an epifluorescence microscope (DMRA, Leica, Wezlar, Germany) equipped with appropriate filter sets. Images were recorded using a CCD camera (Quantix, Photometrics, Tucson, AZ) connected to an image analysis work station (Quips XL, Applied Imaging, Newcastle upon Tyne, UK). For quantitation, ˜15 metaphases were analyzed per slide. The quantitation of the fluorescence signals was performed using the Telomere Quantifier Software (DAKOCytomation).
HaCaT cells, which exhibit a stable mean telomere length of 4 kb (as determined repeatedly by Southern blot analysis), were included as a reference line in each experiment. The telomere length (in kb) of all other cells was calculated by mathematical normalization to the 4 kb of the HaCaT cells.
Genomic DNA was isolated by washing the cells twice with PBS and adding 3 ml lysis buffer, 25 μM proteinase K (20 mg/ml), and 125 μl SDS (20%). After incubation overnight at 56°C, 1 ml of 6 M NaCl was added. After centrifugation, the supernatant was transferred into a new tube and two times of its volume of 96 % ethanol was added. After inverting the tube several times, the DNA was precipitated by centrifugation and dissolved in 1× TE. The Southern blot analysis was performed according to standard protocols (12).
Reproducibility of the Scanning Data
To test whether the data are reproducible, we scanned slides hybridized in the same experiment at the same and at different timepoints. To test whether the data are comparable between experiments, we analyzed slides scanned within the same time interval after hybridization, but hybridized in different experiments. The number of analyzed cells depended on the cell number present and evaluable on the slide.
Single channel measurement of the Cy3 fluorescence intensity
When measuring single channel fluorescence intensities (Cy3-labeled pan-telomeric probe), we observed a good reproducibility when slides derived from the same hybridization experiments were measured at the same time (Fig. 2A). However, slides from identical hybridization experiments scanned at different timepoints showed a variation of up to 100% in the fluorescence intensities. Figure 2B shows an example of two slides hybridized in the same experiment and scanned one day and nine days after hybridization, respectively. The lamp was replaced in the course of the experiments. When scanning the slides of the same hybridization date before and after changing the light source, we found a 3.5-fold increase of the fluorescence intensities after the change of the light source (Fig. 2C). When comparing slides derived from different experiments but scanned within the same time interval after hybridization, we observed marked variations in the single channel fluorescence intensities (Fig. 2D). The measurements of the Cy3 fluorescence alone revealed differences of between 82% and 156%.
Double channel measurements and calculation of fluorescence ratio units
Normalization of the measurement data by calculation of fluorescence ratio units
The simultaneous measurement of the fluorescence signals of the pan-telomeric probe (Cy3-labeled) and of the internal reference probe (FITC-labeled probe specific for centromere 2) showed that the changes of the fluorescence intensity of the reference probe correlated with the intensity changes of the pan-telomeric probe (Fig. 2). Similarly, the FITC fluorescence intensities varied from 108% up to 171%. Thus, when calculating the ratio of both fluorescence intensity values (fluorescence ratio units (FRU)), the deviations of the fluorescence intensities could be reduced considerably. In the following, we refer to FRU values when we report on telomere length.
When using the two color channel measurements, the differences between the different experiments decreased drastically (Fig. 2D). Also, the measurement after a longer storage period at 4°C was comparable with the results obtained one day after the hybridization date (Fig. 2B), and no marked differences were observed after changing the light source (Fig. 2C).
When analyzing both color channels separately, we were able to detect cells with poor hybridization signals, as these cells displayed reduced fluorescence intensity in both color channels. In summary, the combination of the two color channel measurements decreased the differences in the results down to 9–34%.
Validation of Double Channel IQ-FISH
We validated the double-channel IQ-FISH method by telomere length measurements of normal lymphocytes and by comparing telomere length measurements in interphase nuclei on metaphase spreads and by Southern blot analyses.
Telomere length measurements of normal lymphocytes
When comparing fluorescence intensities of the telomeric signals of lymphocytes of similarly aged individuals of the same gender, only slight differences (maximal 13%) in the telomere length were obtained (data not shown).
Comparison of telomere length measurements in interphase nuclei, on metaphase spreads and by Southern blot analyses
Five samples of different cell types were measured in parallel by IQ-FISH and Southern blot analyses. Two of these samples were additionally analyzed by Q-FISH on metaphase spreads. To better compare the data, we set the longest mean telomere length value to 100% and calculated the shorter mean values relative to this figure. We obtained correlating results when comparing the three measurement methods. Three samples of the parallel measurements by IQ-FISH and Southern blot had similar mean telomere length values with both techniques. Southern blot analysis of these three samples resulted in telomere length of 40% relative to the longest telomere values. IQ-FISH analysis of these three samples resulted in 35%, 34%, and 35%, respectively. As no metaphases were available of these samples, Q-FISH analysis on metaphase spreads was not possible. The sample with the second longest telomeres was analyzed, with all three methods. However, for this sample, we observed discrepancies among the three techniques used. While Southern blot analysis resulted in 68%, IQ-FISH resulted in 48%, and Q-FISH resulted in 51% compared to the value of the sample with the longest telomeres.
Different Examples of Telomere Length Measurements
Telomere length measurements of neuroblastoma cell lines
We analyzed seven different neuroblastoma cell lines (STA-NB-3, -7, -9, -10, -12, -15, and SH-SY-5Y) to see whether marked differences occur among these cell lines.
All cell lines, except STA-NB-12 and SH-SY-5Y, displayed an amplification of the MYCN oncogene. The cell line STA-NB-3 was near triploid, whereas the other cell lines were near diploid. Interestingly, the seven analyzed neuroblastoma cell lines showed marked differences in the telomere length (Fig. 3).
The MYCN nonamplified cell line STA-NB-12 displayed the longest telomeres. The second longest telomeres were observed when analyzing the nuclei of STA-NB-3, a near triploid neuroblastoma cell line, with a loss of the short arm of chromosome 1 and MYCN amplification. The third longest telomeres were found in the MYCN nonamplified cell line SH-SY-5Y. The MYCN-amplified cell lines STA-NB-7, -9,-10, and -15 showed decreasing telomere length, starting with STA-NB-7, followed by STA-NB-9 and -10, and finally, the shortest telomere length for cell line STA-NB-15.
Telomere length measurements in senescent tumor cells
Previous reports showed that tumor cells can re-enter the senescence pathway (13).
We wanted to test whether changes in the telomere length in senescent tumor cells can be detected by the use of this technique. Therefore, we used a neuroblastoma cell line that contains nonsenescent cells (N-cells) but also cells that re-entered the senescence pathway (F-cells) (10, I.M. Ambros, unpublished data).
We compared the neuronal- (N-) and flat- (F-) cell fraction of neuroblastoma cell line STA-NB-10: The nuclei of the parental neuroblastoma cell line STA-NB-10, containing predominantly N-cells, displayed telomere length comparable to STA-NB-15 (Fig. 3). The F-type cells, derived from the STA-NB-10 cell line, were hybridized and analyzed in parallel to the slides of the parental cell line containing predominantly N-type cells. However, the nuclei of the F-type cells showed a two-fold reduction of the telomere length compared to the N-type cells (Fig. 3).
Telomere length in osteosarcomas
We wanted to test whether this technique would also be suited for the detection of the ALT mechanism in tumor samples.
ALT cells display typically heterogeneous telomere length (14) and are characterized by the occurrence of brightly shining nuclear bodies, detectable by TTAGGG repeat probes in interphase nuclei (ALT-associated promyelocytic leukemia bodies (APBs)) (15). To find out whether the automatic telomere length measurement is able to discriminate between ALT and non-ALT cells, we analyzed samples of seven osteosarcoma patients, using the pan-telomere and centromere-2 PNA FISH probes.
Four out of seven analyzed tumor samples showed a Gaussian distribution of the telomere length. However, three of the seven samples showed one major peak of telomere length, and additionally, we observed nuclei with longer telomeres. An example of both patterns of fluorescence distribution is shown in Figure 4.
Analyses of cells grown on slides
We tested the applicability of the IQ-FISH method for the analysis of cells directly grown on slides. Therefore, cells of a neuroblastoma cell line were grown on slides and data analyses were performed in the usual manner. As the MetaCyte software saves all morphological and topological information of every analyzed cell, each nucleus can be viewed on the screen. Figure 5 shows the user interface with information on the measured features, the cell morphology, and the position of the analyzed cells at the scanned area on the slide. By clicking on the cell of interest, the cell is automatically relocated and can be inspected by eye.
To enable telomere length measurement in a large number of cells, we adapted a method for automatic telomere length quantification in interphase nuclei. The quantification of the telomere length in interphase nuclei was facilitated by fluorescence light measurements within well-separated nuclei. The IQ-FISH technique provides information on individual cells and allows the morphological analysis of interphase nuclei (11). Also, the topographic details are retained, as demonstrated by the analyses of cells grown on slides. Additionally, there is no need for the time-consuming manual measurements of nuclei, as is the case with current methods for analysis of interphase nuclei (6). The automatization of the measurement process allows the measurement of more cells, and thus, enables the detection of statistically relevant populations and even subpopulations. Furthermore, our method allows us to produce unbiased results, which is not always the case when measuring only proliferating cells.
The single channel analysis of slides derived from the same hybridization experiments resulted in stable results. However, single channel analysis of the pan-telomeric probe signals from samples derived from different experiments resulted in strong variation in the measurement results because of several reasons: variation in intensity between different experiments; differences in the times of the measurement after hybridization; changes of the luminosity of the light source; inhomogeneous illumination of the field and heterogeneous FISH results on the slide. As the pan-telomere sequences and the repeat sequences specific to chromosome 2 replicate at similar time points of the cell cycle, the fluorescence intensities of both probes correspond to each other. Consequently, the fluorescence ratios of both fluorochromes follow a Gaussian distribution and no G2/M peak was observed (Fig. 4A). Also, polyploid nuclei do not influence the fluorescence ratios as long as the chromosome 2 is gained in-line with the rest of the genome. Furthermore, the heteromorphisms of the reference sequence did not seem to influence the results, as heteromorphisms, when present, are consistent within one individual. In fact, parallel analysis of the internal standard demonstrated that the intensity of the internal reference correlated with the changes of the target color channel. The normalization of the fluorescence intensity values for single cells calculated by dividing the fluorescence value of the pan-telomeric signals by the fluorescence intensity of the reference probe increased the reproducibility of the results. This approach allows the comparison of different experiments, the measurement of slides at different times (from one up to nine days) after hybridization, and the compensation for variations of the light source. Additionally, the use of an internal standard enabled the detection of nuclei with poor hybridization. A reduced fluorescence intensity in both color channels indicated poor hybridization. When analyzing only the Cy3 channel, the reduced fluorescence intensity could be misinterpreted as indication of short telomeres.
Fluorescence intensity of the telomeric signals was found to be proportional to the telomere length (2). Therefore, we measured the fluorescence intensities of a pan-telomeric probe and normalized the values, using a second color channel and calculating the FRU values. In this manuscript, we refer to FRU values when reporting on telomere length.
When comparing the results obtained by IQ-FISH, Q-FISH, and Southern blot, the three methods showed a good correlation.
However, good IQ-FISH results can be obtained only when a strict standardization of the hybridization protocol is performed. Moreover, we found out that the timepoint of the analyses was essential. Measurement of fluorescence intensities on the day when the hybridization was performed led to unreproducible data.
We applied the IQ-FISH method for telomere length measurements to different tasks. When analyzing the telomere length of three healthy and similarly aged adult individuals of the same gender, only slight differences of the telomere length were observed. This may represent the natural range of telomere length in individuals (16). Yet, when analyzing several neuroblastoma cell lines, marked differences in the telomere length were observed. The MYCN nonamplified cell lines and a near-triploid cell line with MYCN amplification displayed the longest telomeres. All other MYCN-amplified cell lines had shorter telomeres. Whether a correlation exists between telomere length and MYCN amplification or possibly ploidy needs further investigations.
The comparison of N-cells with F-cells derived from a MYCN-amplified neuroblastoma cell line demonstrated pronounced differences in the telomere length. As previously shown, F-cells can be generated from N-cells by oncogene expulsion characterized by micronuclei formation (10). Recent findings have shown that these F-cells entered the senescence pathway (I.M. Ambros, unpublished data). Because of a reduced proliferation rate of the F-cell fraction, no metaphase chromosomes are available for analysis. Therefore, we applied the IQ-FISH technique to measure the telomere length in both cell types. We observed that the F-cells showed a two-fold decrease in telomere length compared to their N-cell counterpart. As telomere shortening is a pathway for tumor cells to enter the senescence pathway, (17, 18) these data support the finding that F-cells are senescent cells (I.M. Ambros, unpublished data).
As a further application of IQ-FISH, we investigated telomere length in osteosarcomas. Aggressive osteosarcoma tumors frequently show ALT (19) and it has been shown that, in osteosarcoma, ALT is a marker for poor outcome (20). Therefore, the reliable and rapid detection of ALT in osteosarcomas would be of clinical importance. We used IQ-FISH for analyzing osteosarcoma tumor samples. So far, all analyzed cell lines and tumors have displayed a Gaussian distribution of fluorescence intensities. Therefore, the Gaussian distribution of the fluorescence intensities found in the majority of the osteosarcoma samples was interpreted as normal pattern. However, in some samples—besides a major peak for relatively short telomeres—a fraction of cells also displayed high fluorescence values. Microscopical examination of these cells revealed on average one large telomere FISH spot in a background of weak telomeric signals. On the one hand, the presence of these large hybridization spots together with very weak telomeric signals in the same cell are in line with previous reports on hyper-variable telomeres and heterogeneity of telomere length in ALT cells (14, 19). On the other hand, this observation could support the reports of APBs, which showed that APBs appear as brightly shining nuclear bodies detectable by TTAGGG repeat probes in interphase nuclei (15). When taking all this into consideration, we assume that the non-Gaussian distribution of fluorescence intensities represents the ALT phenotype. We must, however, report on the fact that the conspicuously high intensities of the large hybridization spots influenced the results by shining through the reference channel. Thereby, the reference channel was increased and the fluorescence ratio values decreased. But, the differences were striking enough to allow the detection of cells with large hybridization spots. IQ-FISH could be used as a routine method for the detection of tumors with the ALT mechanism as well as for retrospective studies on collected tumor samples.
In conclusion, automatic telomere length measurement in interphase nuclei by FLAME allows the study of a high number of resting cells. Besides the analysis of a statistically relevant cell population, information on each individual cell can be obtained and evaluated by relocation and visual inspection. Thus, an accurate alternative to current techniques is presented.