Cancers often are the result of ongoing genomic instability (1), to which telomere dysfunction contributes (2). Ongoing telomere attrition due to incomplete replication of the lagging strand during DNA synthesis results in critically short telomeres and uncapping, a hallmark of premalignant and tumor cells. The loss of telomeric DNA promotes activation of Ataxia telangiectasia mutated (ATM)-kinase and nonhomologous end joining (NHEJ) (3, 4).
Telomeres were used in the past as a prognostic marker in cancer, including breast cancer (5–7) prostate cancer (8–10), neuroblastoma and glioblastoma (11, 12), or esophageal squamous cell carcinoma (13).
Normal and tumor cell nuclei significantly differ from each other with respect to their nuclear architecture including telomere numbers and sizes and the formation of telomeric aggregates. Telomeric aggregates are clusters of telomeres that at an optical resolution of 200 nm, cannot be separated further by conventional 3D fluorescent microscopy (2). Telomeric aggregates represent telomeres that display end-to-end fusions or telomeres in very close proximity. Telomeric aggregates that represent fusions will initiate breakage-bridge-fusion cycles resulting in dynamic changes of the genetic material found in each daughter cells (2). The changes, alterations of chromosome number, and 3D nuclear architecture of telomeres, have been used to define the transition of mono-nucleated H-cells to multi-nucleated Reed-Sternberg cells in Hodgkin's Lymphoma (14–17), to classify subgroups of glioblastoma patients (11) to detect premalignant cells in mouse plasmacytoma and early lesions in cervical cancer (18).
Based on these findings we now developed a reliable test that is easy to use and that will be applicable to a broad range of such cancers. For cancer diagnosis it is common to use the four-stage TNM system, a classification system developed by Denoix (19), which is based on the size of the tumor and how far it has spread from its original location in the body.
Staging methods for leukemia are based on RAI or Binet. The widely used RAI classification system (20) is based on the risk of developing lymphocytosis and the degree of involvement of lymphoid organs. The Binet staging (21) classifies chronic lymphocytic leukemia (CLL) according to the number of lymphoid tissues that are involved, i.e., spleen and lymph nodes, as well as the presence of low red blood cell count (anemia) or low number of blood platelets (thrombocytopenia). Staging assists in the planning of a person's treatment, and is useful in the estimation of a person's prognosis (likely outcome or course of the disease). This classic staging system needs to be refined by a new array of biomarkers those address the heterogeneity of tumors.
Biomarkers such as estrogen receptor (ER), HER-2/neu in breast cancer are used for optimization of therapies and prediction of the outcome of treatments. Other predictors are PSA (prostate cancer) or EGFR (colon cancer). In recent years the focus has shifted toward molecular markers in genetics, epigenetics, the analysis of gene-expression patterns (22), and proteomics (23–25). The molecular cancer diagnostic is looking at changes on genetic and epigenetic levels (26), translocations (SKY), and variation in copy numbers (array CGH). Overall there is a need for a diagnostic tool that allows for the detection of a broad array of tumors.
The ideal biomarker for tumors would not only be applicable for a wide range of tumors but it would also provide a basis for the tailoring of optimal treatments, the prediction of success, and monitoring of treatment response (27–30).
Our choice for tumor cell detection was the development of an automated 3D detection system. The automated 3D scanning we present here allows for a high throughput of samples, while the data output includes numbers of telomeric aggregates, telomere count, and telomere signal intensity and size. The purpose of this study was the assessment of sensitivity of tumor cell detection based on the presence of 3D alterations in telomeres using the 3D scan of interphase nuclei. Sensitivity tests of this new tool show that we are able to detect one aberrant cell in 1,000 normal cells.
MATERIALS AND METHODS
Mice and B Cell Isolation
These experiments were approved by the ethical committee, protocol number 07-002/1/2/3. The plasmacytoma cell line MOPC460D, a gift of J. Mushinski (National Institutes of Health, Bethesda), were cultured as described (31). T38H mice (32) were obtained from Harwell, UK (33). The mice were kept under specific pathogen-free (SPF) conditions. At 8 weeks of age, three male mice were euthanized, and the spleens were harvested. B and T lymphocytes were flushed out of the spleen with 3 ml RPMI medium. The cells were sedimented for 5 min at 270g. The pellets were resuspended with 5 ml ACK-buffer (150 mM NH4Cl, 10 mM KHCO3, 100 mM Na-EDTA, pH 7.4), then centrifuged for 5 min at 120g. The pellet was washed once with RPMI-medium.
Mouse plasmacytoma cells (MOPC) and primary mouse cells were fixed in a way that preserves the shape of 3D nuclei (34–36) using the following protocol: ∼10 million cells were washed in PBS then centrifuged at 120g for 5 min at room temperature (RT). Cell pellets were resuspended in 5 ml of 75 mM KCl for 10 min at room temperature. After adding 1 ml fixative (3:1; methanol/acetic acid) the tubes were carefully inverted three to four times to gently mix the cells with the fixative. Cells were centrifuged again at room temperature for 10 min at 120g. The cell pellet was washed with 3 ml fixative and centrifuged for 10 min at 120g. The final cell pellet was resuspended in 1 ml fixative and stored at −20°C.
Determination of Tumor Cell Frequencies in Mixtures with Normal Lymphocytes
For the dilution of MOPC with normal mouse lymphocytes the cell number of each suspension was determined by counting using a phase hemacytometer, (Hausser Scientific, VWR International, Mississauga, Ontario, Canada). A 2 × 105 cells were applied onto each microscope slide and left to air dry prior to hybridization with telomere specific paints. For spiking experiments we diluted samples of 2 × 105 normal cells with 1 MOPC cell in 1,000 normal lymphocytes, 1:100, 1:20, and 1:10. In parallel, we prepared slides with mouse plasmacytoma and normal cells, only.
Telomeres were hybridized with Cy3-labelled peptide nucleic acid probes (DAKO, Denmark) according to our published protocols (37).
The 3D-fixed cells were washed in freshly prepared methanol/acetic acid (3:1) fixative and positioned on the slides. After air-drying the slides, the cells were fixed in 3.7% formaldehyde/phosphate-buffered saline (PBS) for 20 min, washed three times for 5 min in PBS. After an incubating in TPBS (0.5%, Triton X-100 in PBS) for 10 min the slides were incubated in 20% glycerol for 1 h followed by four freeze-thaw cycles in liquid nitrogen and three washes with PBS. After a 5-min incubation in 0.1N HCl the slides were washed for 5 min in PBS, twice. Prior to the hybridization the samples were equilibrated for 1 h in 70% formamide (Fluka-Sigma Aldrich, St Louis, MO), 2× SSC at room temperature. The slides were hybridized with Cy3-labeled telomere-specific PNA probe (DAKO) and washed as previously published (34–36).
DAPI (4′,6-diamidino-2-phenylindole) was purchased from Sigma Aldrich (Oakville, ON) and used at 0.1 μg ml−1 to counterstain the nuclei on the slides (31, 35). For the mounting medium we used ProLong® antifade Gold mounting medium (Molecular Probes™, Invitrogen detection technologies, Carlsbad, CA). The slides were allowed to dry over night at 4°C under light protected conditions. The slides were stored at −20°C until use.
Automated Image Acquisition and Processing
The automated Image acquisition of interphase nuclei was performed using the ScanView system [Applied Spectral Imaging (ASI)], using an Olympus BX61 microscope with a VDS CCD camera, model 1300DS. For scanning purposes the microscope was equipped with a motorized eight-slide stage (Märzhäuser, Germany). The 3D-images were acquired with dry 40× objective and a 0.63× c-mount (Olympus) taking 11 focal planes per cell. The axial sampling distance between planes, Δz, was 500 nm. Exposure times were constant at 200 ms (DAPI) and 1,000 ms (Cy3) throughout the experiments. The tissue sample mode with aggregate detection level of 15 was used to enable segmentation of touching cells and optimized aggregate detection. Cells with <21 detected signal were excluded as nonclassified (NC). Approximately 10,000 to 15,000 cells were scanned and analyzed within 60 min.
For analyzing the data, the following software modules of the ScanView system (ASI) were used: SpotScan with TeloScan for the detection of nuclei, signals, and aggregates. Such a large size of data must be managed correctly, and it was performed by the ScanView database module—case data manager (CDM). Up to 30,000 classified single cells per mixture were analyzed. The numbers of classified cells analyzed per mouse sample were as follows; for the 1:1,000 dilution about 30,000 cells, for the 1:100 dilution about 10,000 cells, for the 1:20 and 1:10 dilution about 5,000 cells, for the pure normal- and tumor samples about 1,000 cells were analyzed.
3D Image Analysis for Telomeres
Telomere measurements were performed using TeloView for the manual 3D-acquisition (35, 38), TeloScan for the automated 3D-acquisition (11). The integrated intensity of each telomere was calculated based on the linear correlation between telomere length and signal intensity.
Telomeric aggregates are defined as clusters of telomeres that cannot be resolved as separate signals at the optical resolution limit of 200 nm (63× oil) and 350 nm (40×) (18, 35, 39).
The statistical significance of the differences was determined using the ANOVA test.
We analyzed normal mouse lymphocytes and mouse plasmacytoma cells, MOPC460, for their specific telomere signatures using the automated 3D-scanner. The telomere signature comprises parameters such as signal numbers per nucleus, signal intensity, and aggregate formation. The software provides a large set of parameters on the analyzed signals, which include information about the cell's identification in the image gallery, the position of each signal within the nucleus, signal intensity, and the presence of aggregates. The second part of the data sheet summarizes the information specifically to each cell; a third part gives detailed information about the number of telomeres within the cells.
The scanned cells are shown in a gallery on the right hand side for view and selection purposes (Figs. 1a and 1b). The selected cell of the gallery, displayed in the upper left hand corner of the screen, has detailed information in its lower left hand corner such as the number of detected signals (red number) and the class it has been designated to (white number). The classification of cells is based on the number of telomere signals within each scanned nucleus, as defined by the user. Accordingly, a histogram (bar graph, shown in the lower left hand corner of Figs. 1a and 1b) is generated displaying the distribution of analyzed cells. The histograms of normal cells (Fig. 1a) and tumor cells (Fig. 1b) differ significantly from each other. The majority of normal cells is classified in the group with >20 signals/cell (Fig. 1a). As indicated in Figure 1b, tumor cells (MOPC460D) display large aggregates indicated with arrows, and higher numbers of telomere signals per cell (histogram).
The difference between normal mouse lymphocytes and MOPC is illustrated in Figure 2, directly comparing the distribution of cells in classes based on the signal numbers. The majority of normal cells were classified as >20 signals. The mouse plasmacytoma cells show a shift to higher signal numbers due to repeated breakage-bridge-fusion cycles and aberrant mitosis.
One important step in the initiation of genomic instability, thus in tumor formation, is the formation of telomere aggregates (18). Therefore we examined the presence and the frequency of such aggregates in normal mouse lymphocytes and in mouse plasmacytoma cells (Table 1). The data obtained with TeloScan shows that normal mouse lymphocytes have a background of telomeric aggregates of ∼4.3% ± 0.9%. This is in agreement with earlier data obtained with AxioVision (Carl Zeiss Canada Ltd, Toronto, ON) and TeloView (38) (unpublished data). However, about 87% of the mouse plasmacytoma cells showed the presence of telomeric aggregates (Table 1).
Table 1. Signal intensity and aggregate detection in normal mouse lymphocytes and MOPC
Average signal intensity (a.u.)
The signal intensity of all signals and the presence of telomeric aggregates within three independent scans were analyzed. The signal intensity is given in arbitrary units.
Normal mouse lymphocytes
4.3 ± 0.9
86.9 ± 2.8
Next we tested the detection limit of the cancer cell detection method based on the telomere signature. Therefore we spiked a suspension of normal cells with tumor cells at various concentrations and compared the scanning results with respect to signal numbers in the cells, presence of telomeric aggregates, and signal intensity.
The presence of tumor cells within a population of normal cells showed a shift of cells classified as >20 signals per cell, to cells classified as >40 signals per cell. With increasing concentrations of tumor cells we observed an increased percentile of cells with higher numbers of telomere signals (Fig. 3). The presence of one tumor cell within 1,000 normal cells causes a shift of the classified cells to classes with higher numbers of telomere signals.
With increasing numbers of tumor cells we would expect an increasing number of cells being positive for telomeric aggregates. We therefore spiked the normal cells with MOPC and determined the number of cells harboring telomeric aggregates. As expected the percentile of cells positive for telomeric aggregates increased with the amount of tumor cells present (Fig. 4). The analysis showed that the presence of one tumor cell within a population of 1,000 normal cells already display a significant increase of aggregates. All experiments were performed in triplicates. Applying ANOVA the significance was P < 0.001 (Fig. 4).
Next we examined the mixture of the cells for signal intensity in correlation to the number of telomeres (Figs. 5a– 5c). As shown in Table 1 the MOPC showed a 4.8 fold higher signal intensity than the normal mouse lymphocytes. We therefore analyzed the changes of intensity in the presence of tumor cells. One tumor cell within 1,000 analyzed cells causes a change in the signature compared with the control (Figs. 5a–5c blue), characterized by the appearance of two peaks (Figs. 5a–5c red). The peaks increase with increasing number of tumor cells present in the test mixtures. This change is reproducible comparing the graphs of three independent experiments (Figs. 5a–5c).
Our study documents that 3D automated scanning of telomeres is feasible. We show that the sensitivity of detection is one cancer cell in 1,000 normal cells using mouse plasmacytoma cells and normal mouse lymphocytes. We thus conclude that this approach might be relevant for the future analysis of clinical samples. We anticipate the use of this scanning technology in the detection of rare cancer cells, in the screening of risk groups, and in the assessment of treatment success. Moreover, it has already shown to have the ability of sub-classifying patient subgroups that could not previously be identified (11). We also expect its success in guiding patient treatment decisions based on the recognition of recurrent/aggressive telomere profiles observed at diagnosis (40). More analyses will be needed to confirm the applicability of 3D telomere scanning to many cancers.
Previous studies with normal cells have demonstrated that telomeres are organized in a nonoverlapping fashion (35, 38) in discrete microterritories (41, 42). Telomere attrition and uncapping, lead to the formation of telomeric aggregates, a hallmark of premalignant and tumor cells. These telomeric aggregates within the interphase nucleus lead to breakage-bridge-fusion cycles (43, 44), resulting in the generation of aberrant cells that display an altered 3D organization telomere length.
Cells with deficiencies in telomere maintenance are susceptible to enhanced telomere loss during cell proliferation, resulting in telomere dysfunction and genomic instability. Various cancers have been associated with short telomeres like esophageal squamous cell carcinoma (13). Other data demonstrate that telomere attrition is a common early alteration in many human cancers, including gastric cancer (45, 46), colon cancer (47), lung cancer (48), and breast cancer (6, 49). Telomere dysfunction is also associated with bone marrow failure (50), specifically MGUS and multiple myeloma (51, 52). Very short telomeres are also associated with CLL (53), dyskeratosis congenita (54), pancreatic cancer (55), prostate cancer (9), and Barrett's esophagus, which is associated with an increased risk of esophageal adenocarcinoma (56). These data suggest that the alteration of telomeres is potentially an important tool for cancer diagnostics for a broad range of tumors.
Many studies have focused on the analysis of telomere length (overall telomere length or chromosome-, and chromosome-arm-specific telomere length). There are various methods established to determine telomere length like telomere restriction fragment (TRF) analysis, quantitative PCR, and the single telomere length analysis (STELA) [for review: (57)]. More sensitive methods include fluorescence in situ hybridization (FISH) (54) and the primed in situ (PRINS) labeling technique (58, 59) allowing measurement of telomere lengths.
Not surprisingly, various attempts were made to automate the analysis of telomere length, cell in suspension or fixed on solid surfaces. An example for telomere analysis utilizing cell suspensions is the use of flow-FISH (60–62) or high throughput quantitative FISH (HT Q-FISH) (63), allowing for a high throughput of samples. However, these methods are less sensitive (kB-range) than Q-FISH.
Previously, Narath et al. (64) have automated the telomere length measurements in interphase nuclei. Their scanning method allowed for cell identification, spot counting, and intensity measurement, aided by a fluorescence-based microscopic scanning system. This scanning system is based on a fully motorized microscope using the same motorized stage as we used in this study. For the data acquisition nine focus planes were captured at a sampling distance of 0.5 μm. The processing rate was 6–10 nuclei min−1 compared to 160–250 cells min−1 with TeloScan described here in our study. Data obtained with this system summarize telomere length.
Based on telomerase activity as a tumor marker (65,68), an automated platform was developed to measure telomerase activity in live circulating tumor cells (CTC). Purified peripheral blood mononuclear cells (PBMC) would be captured on parylene-C microfilters. The telomerase activity is measured using the TRAP (telomere repeat amplification protocol). This method allows for the isolation and characterization of CTCs, isolating viable cells at a 1500-fold enrichment, but requires radioactive labeling of the samples.
To the best of our knowledge, no data have been published on the automated 3D scanning of nuclei that permit the assessment of spatial telomere organization in interphase, telomere length, the presence of telomeric aggregates, and telomere numbers. These are the four criteria that define differences between normal and tumor cells. Our approach was to integrate all known milestones of telomeric changes in nuclear architecture during cancer development into one automated assay, extending the parameters from telomere length/intensity by the additional parameters such as the automated cell identification, positioning of the telomeres within the 3D-nucleus, the presence of telomeric aggregates, and the number of telomere signals in the cells. Thus, the automated 3D-genome scanning we carry out quantifies telomere length in interphase nuclei and provides morphological and topological details.
The automation allows for a high throughput of about 10,000–15,000 cells within 1 h using the 40× objective. Mounting the slides with ProLong antifade gold, we also overcame the process of photobleaching, a common problem in fluorescence microscopy, which would otherwise lead to inaccurate measurement of fluorescence intensity (66).
The detection of tumor cells in a mixture of normal and tumor cells, presented here indicates that our 3D scanning approach is able to detect a significant difference between these two cell types. Our experiments clearly demonstrate that 3D telomeric scanning is capable of detecting one tumor cell within a population of at least 1,000 normal cells. This sensitivity is comparable with earlier data showing a detection limit of one tumor with amplified MYCN cell within 1,000 nonamplified cells (67).
The fact that changes in the 3D telomeric signatures are universally applicable to lymphoid- and nonlymphoid cancers, to single cell suspensions and tissue (fresh, frozen, or paraffin-embedded) could make this automated 3D telomere scanning tool interesting for the clinical setting. In future experiments we will have to validate our findings with human samples and prove that this new diagnostic tool may be applicable for early cancer detection, the detection of circulating tumor cells, differential diagnosis, as well as a prognostic marker and for therapy selection. Therefore, it could be anticipated that the automated three-dimensional (3D) genome scanning based on the nuclear architecture of telomeres could be a powerful tool to expand our knowledge on the role of telomere length in human disease and potentially a powerful diagnostic as well as prognostic tool in cancer diagnostics (40). We also envision that our new screening platform would be helpful in the assistance of tailoring therapeutic strategies in a personalized manner.
The authors thank Mary Cheang for statistical analyses.