Genetic intratumour heterogeneity in high-grade brain tumours is associated with telomere-dependent mitotic instability


  • D.G. and X.F. contributed equally to this work.

David Gisselsson, Department of Clinical Genetics, Lund University Hospital, SE-221 85 Lund, Sweden. Tel: +46 46 175867; Fax: +46 46 131062; E-mail:


Glioblastoma multiforme (GBM) and other high-grade brain tumours are typically characterized by complex chromosome abnormalities and extensive intratumour cytogenetic heterogeneity. The mechanisms behind this diversity have been little explored. In this study, we analysed the pattern of chromosome segregation at mitosis in 20 brain tumours. We found an abnormal segregation of chromatids at mitosis through anaphase bridging (10–25% of anaphase cells) in all 10 GBMs. Anaphase bridging was also found in two medulloblastomas (7–15%), one anaplastic astrocytoma (17%) and one oligodendroglioma (6%). These tumours showed a relatively high degree of cytogenetic complexity and heterogeneity. In contrast, cell division abnormalities were not found in low-grade brain tumours with less complex karyotypes, including two pilocytic astrocytomas and two ependymomas. Further analysis of two GBMs by fluorescence in situ hybridization with telomeric repeat probes revealed excessive shortening of TTAGGG repeats, indicating dysfunctional protection of chromosome ends. In xenografts established from these GBMs, there was a gradual reduction in cytogenetic heterogeneity through successive passages as the proportion of abnormally short telomeres was reduced and the frequency of anaphase bridges decreased from >25% to 0. However, bridging could be reintroduced in late-passage xenograft cells by pharmacological induction of telomere shortening, using a small-molecule telomerase inhibitor. Telomere-dependent abnormal segregation of chromosomes at mitosis is thus a common phenomenon in high-grade brain tumours and may be one important factor behind cytogenetic intratumour diversity in GBM.


Gliomas are the most common primary brain tumours in the adult central nervous system, with glioblastoma multiforme (GBM) being the most aggressive type among them [1–3]. GBMs are characterized by extensive inter-and intratumour heterogeneity manifested in morphology, gene expression and genetic abnormalities [4–7]. Several studies have demonstrated that cytogenetically related or unrelated clones coexist in different regions within the same GBM specimen [7–10] and this cytogenetic heterogeneity could correspond to phenotypically diverse cell populations. The cytogenetic heterogeneity is typically concurrent with a high degree of cytogenetic complexity, including both numerical and structural chromosome aberrations. The most common recurrent changes that have been reported in GBM include gain of chromosome 7, losses of 9p sequences and chromosome 10, and gene amplification, primarily of the epidermal growth factor receptor (EGFR) gene [11,12]. Among high-grade glioma, one set of the chromosome changes seen in a tumour is typically common to the vast majority of the tumour cells, whereas other sets of chromosome rearrangements are restricted to subpopulations of tumour cells [7]. However, several cytogenetic studies have shown that GBMs are monoclonal in origin [13–15]. Recently, a small subpopulation of CD133+ GBM cells have been identified as GBM stem cells [4–6,16,17]. Phenotypically, these cells express the stem cell marker CD133 and they functionally posses the ability to regenerate CD133+ cells and to initiate tumour growth when grafted into the central nervous systems of animals. Taken together, this indicates that cytogenetically and phenotypically diverse subclones are generated during tumour development, possibly resulting from a combination of selection and loss of genetic stability [8].

One mechanism behind genetic instability and karyotypic diversity in tumours is abnormal chromosome segregation at cell division [18]. The most common type of unbalanced chromatid segregation so far identified consists of failed chromatid segregation at anaphase due to the formation of a chromatin bridge. Such anaphase bridges may break at later stages in cell division and the broken ends can reunite into complex structural chromosome changes in the daughter cells [19,20]. The formation of anaphase bridges has been strongly associated with shortening of telomeric repeat sequences, leading to disruption of the telosomic nucleoprotein complex normally protecting the integrity of chromosome ends [21]. Nonhomologous end-joining of unprotected chromosome ends may, in turn, lead to formation of functionally dicentric chromosomes that can form bridges when the two centromeres in each chromatid are pulled in different directions at anaphase. This mechanism has been shown to contribute to cytogenetic intratumour heterogeneity in cancers of the colon [22], breast [23], and several other malignancies. It is not known whether similar processes are active in tumours of the central nervous system. Moreover, it has been little explored to what degree the genomic instability and cytogenetic diversity found in primary tumours are retained in brain tumour xenograft models. This is of interest as hetero- and orthotopic xenografts are often used to study brain tumour biology and response to candidate therapeutic agents [24–29].

In the present study we have attempted to explore these issues by analyses of cytogenetic heterogeneity, chromatid segregation pattern, telomere status, and phenotypic features in primary brain tumours and serial tumour xenografts.

Materials and methods


The use of patient material was approved by the Regional Ethics Review Board in Lund. For comparison of mitotic segregation patterns and chromosome changes, patients were selected from each of a consecutive series of GBMs and paediatric brain tumours, in which clonal chromosome aberrations had been found by chromosome banding. Tumours with normal karyotypes were excluded from the analysis. Of the remaining cases, the 10 most recent ones were selected from each series (Table 1).

Table 1.  Karyotypes and anaphase bridge frequencies in primary tumours. Clones are indexed in Lu-1 and Lu-2 for comparison with Table 2
CaseAge/sexHistology/gradeKaryotypeABF (%)
  1. AA, anaplastic astrocytoma; ABF, anaphase bridge frequency; EP, ependymoma; F, female; GBM, glioblastoma multiforme; M, male; MB, medulloblastoma; ODG, oligodendroglioma; PA, pilocytic astrocytoma.

Lu-146/FGBM/IV44–45,X,−X,+ider(7)(q10)del(7)(q35),−10,−13,del(14)(q22),del(16)(q21) (A)/
(q12;p13) (B)
Lu-267/MGBM/IV90,XX,−Y,−Y,−1,−1,+7,−8,−8,der(14)t(1;14)(q12;q22)×2,1–30dmin (A)/

Assessment of sister chromatid segregation patterns

Paraffin sections from primary brain tumour biopsies and resection specimens were stained by haematoxylin/erythrosine and mitotic segregation patterns were analysed in 30–268 cell divisions in each tumour. The anaphase bridge frequency (ABF) was calculated as the proportion of anaphase cells where the two poles were joined by one or several chromatin filaments as previously defined [22]. For analysis of chromosome dynamics at cell division in cultured cells, chamber slide cultures were harvested without Colcemid, washed in PBS, and fixed in 3:1 methanol : acetic acid. To distinguish murine from human mitotic cells in cultures from xenografts, chromosomes were denatured in 70% formamide at 72°C for 10 s, after which the cells were dehydrated in ethanol and stained with diamidinophenylindol. This procedure confers bright staining to heterochromatic chromosome regions, including the pericentromeric satellite sequences present in all murine chromosomes and the constitutive heterochromatin regions in human chromosomes 1, 9, 16 and Y. Mitotic cells with a punctuate, brightly staining segment present in the majority of chromosomes were classified as murine cells and excluded from the analysis. At least 30 human anaphase cells were analysed in each culture.

In vitro culture and cytogenetic analyses

In vitro cell lines from primary brain tumours and xenografts were cultured in Iscove's Modified Dulbecco's Medium (IMDM; Gibco, Stockholm, Sweden) with antibiotics and 17% foetal bovine serum (FBS). Prior to analysis, cells from the Lu-3 – Lu-10 adult tumours were subcultured in vitro one to five times, whereas Lu-1, Lu-2, and Lu-11 – Lu-20 were not subcultured in vitro. Dividing cells were arrested in metaphase by Colcemid (Gibco), harvested, and prepared for G-banding and multicolour fluorescence in situ hybridization (M-FISH) analyses according to standard procedures. Clonality criteria were according to the ISCN (1995) recommendations [30]. Telomeric TTAGGG repeats were visualized by FISH with fluorescein-conjugated (CCCTAA)3 peptide nucleic acid probes [31], and the number of negative chromosome termini was scored in metaphase cells of the lowest ploidy level. Although a negative terminus may still contain up to 500 bp of TTAGGG repeat sequences, this method has previously been shown to yield a valid assessment of the protective capacity of telomeres [32].

Establishment of serially transplantable xenograft lines

The use of animals was approved by the Regional Animal Ethics Committee. Material from two primary GBMs (Lu-1 and Lu-2) was minced and digested in serum-free IMDM supplemented with 0.5 mg/ml collagenase (Sigma, Stockholm, Sweden) and 25 μg/ml DNase (Sigma) at 37°C for 40 min. Following meshing through a 70 μm cell strainer (BD Biosciences, Bedford, MA, USA), erythrocytes were lysed with ammonium chloride (Stem Cell Technology Inc., Vancouver, BC, Canada). Tumour cells washed and resuspended in IMDM were subcutaneously injected into the flanks of 6- to 8-week-old NOD/SCIDBeta2m–/– or SCID-Beige mice at 2–3 × 106 cells/animal. Xenograft GBM growth was observed within 2 months and the resulting xenograft GBM specimens (Lux-1 and Lux-2) were dissected when tumour size reached approximately 9 × 9 mm. Single cell suspensions were prepared from these tumours as described above and injected into new mice at about 2 × 106 cells/animal for serial in vivo passaging.

Histopathology and immunohistochemical staining

Primary GBM and xenograft tumour specimens were fixed in formaldehyde solution and embedded in paraffin. Five-micrometre-thick sections were stained with haematoxylin/erythrosine and van Gieson for conventional diagnostic procedures. For immunohistochemical analysis, 5-μm-thick sections were mounted on glass slides (DAKO ChemMate Capillary Gap Microscope Slides, 75 mm, Dako A/S, Glostrup, Denmark) and dried at 60°C for 1 h. All sections were microwave pretreated at 750 W for 19 min in 10 mM citrate buffer (pH 6.0) for antigen retrieval. An automated immunostainer (TechMateTM 500 Plus, Dako) was used for the staining procedure with DAKO ChemMate Kit peroxidase/3–3′diaminobenzidine. Antibodies used were anti-nestin (clone 10C2, Chemicon, dilution 1:500), anti-Ki-67 (clone MIB-1, DAKO, dilution 1:800) and anti-glial fibrillary acid protein (GFAP; polyclonal, Dako, dilution 1:5000). To ensure staining specificity, tissue specimens with known expression of the primary antibody (foetal periventricular brain tissue, tonsil and astrocytoma respectively) were used as positive controls. The nestin and GFAP immunostaining positivity was independently scored (by A.P. and E.E.) as weak and/or limited, moderate, as well as marked and/or extensive positivity. The proliferation marker Ki-67 was assessed as percentage of positive cells in 10 high-power microscopical fields.

Immunophenotypic analyses

Single cells prepared from primary and xenograft GBM specimens were resuspended in PBS containing 2% FBS and incubated with nonspecific mouse IgG1 (Clone MOPC 21, Sigma) at 5 μg/ml. For xenograft GBM cells, 2 μg/ml of purified rat anti-mouse CD16/CD32 (FcγIII/II R) was additionally applied to the cells to inhibit FcγIII/II R-mediated antibody binding to murine cells. Cells were subsequently stained with phycoerythrin (PE)-conjugated monoclonal antibodies (mAbs) for CD44, CD133, EGFR and platelet-derived growth factor receptor α (PDGFRα) in combination with allophycocyanin-conjugated anti-CD45 or anti-human MHC-I mAb at saturating concentrations at 4°C for 15 min in the presence of 1.0 μg/ml 7-aminoactinomycin D (7-AAD; Sigma). Anti-CD45 staining was used to identify leucocytes in primary GBM cell preparations and the anti-human MHC-I staining to identify human cells in xenograft GBM preparations. To assess A2B5 expression, the cells were first stained with unconjugated A2B5 mAb, followed by staining with PE-conjugated anti-mouse IgM antibody. The staining with the secondary PE-conjugated anti-mouse IgM antibodies alone did not give signals above the background staining. Cells were also stained with isotype-matched control mAbs for background staining. Stained cells were measured using a FACSCalibur and analysed in 7-AAD negatively stained fractions. The anti-CD133 mAb was purchased from Miltenyi Biotec (Bergisch Gladbach, Germany) and the A2B5 mAb from Chemicon Europe, Ltd (Billerica, MA, USA). All other mAbs were from BD Biosciences.

hTERT expression analysis and telomerase inhibition

The expression of the human telomerase reverse transcriptase gene (hTERT) in xenograft GBM cells was quantified by using the housekeeping gene β-actin as a control in real-time RT-PCR as described previously [33]. Telomerase activity was inhibited by growing cells for 30 days in medium containing the MST-312 hTERT inhibitor (Calbiochem EMDBiosciences, Madison, WI, USA) at noncytotoxic concentrations of 1.0 and 0.5 μM respectively [34]. Medium was changed every third day to maintain stable concentrations of MST-312. Cultures were split 1:2 at subconfluence. Control cells were grown in parallel flasks in the absence of MST-312. Colony formation capacity was evaluated by plating single-cell suspension aliquots containing 1 × 104 cells on chamber slides. The number of colonies (>30 cells) was scored after 10 days. After telomerase inhibition, haematoxylin/erythrosine staining was used for ABF scoring because no dividing murine cells were found in culture by cytogenetic analysis. Apoptotic and necrotic cells were detected by the Vybrant assay according to the manufacturer's instructions (Invitrogen, Stockholm, Sweden).


Cytogenetic changes and mitotic segregation abnormalities in primary brain tumours

For assessment of mitotic segregation patterns in GBM, mitotic configurations were analysed in tissue sections from 10 consecutive GBM cases, in which clonal chromosome abnormalities had been found by G-banding (Table 1). All karyotypes were complex, with several numerical changes and unbalanced structural rearrangements. Abnormal mitotic figures were found in the tumour parenchyma of all 10 tumours, with the highest frequencies found in regions with high cellularity and grave nuclear atypia. The most common type of anomaly was the presence of chromatin bridges between the poles of anaphase cells (Figure 1a). The frequency of such anaphase bridges ranged from 10% to 25%. Other types of chromatid abnormalities found at lower frequencies (<4%) were multipolar mitoses and chromosome lagging. As a reference group, a series of 10 brain tumours not classified as GBM were also analysed. Among the four gliomas in this group, anaphase bridging was found in one anaplastic astrocytoma and one oligodendroglioma with complex karyotypes, whereas no bridges were found in two pilocytic astrocytomas, both of which had few cytogenetic changes. Neither of two ependymomas showed anaphase bridging. Cytogenetically, one of these only showed losses of whole chromosomes. The other case showed two clones, one with a whole-arm translocation and the other with another unbalanced translocation. Among the four medulloblastomas, two showed anaphase bridges (Figure 1b) and complex karyotypes, whereas two had no bridges and karyotypes consisting only of numerical changes. Thus, anaphase bridging in tumour biopsies was found primarily in tumours exhibiting complex karyotypes with both numerical and structural changes, whereas tumours not showing bridges had either few structural changes or numerical changes only.

Figure 1.

Anaphase bridging and telomere shortening in primary tumours and glioblastoma multiforme (GBM) xenografts. (a,b) Anaphase bridges (arrows) in a GBM (Lu-8) and a medulloblastoma (Lu-17) respectively; the marked areas in the top row images (×300) are shown at higher magnification (×1000) in the lower row. (c) Gradually decreasing anaphase bridge frequencies in cultured cells from Lu-1/Lux-1 (closed diamonds) and Lu-2/Lux-2 (open diamonds) during passages in mice; the higher frequencies in cultured Lu-1 and Lu-2 tumour cells compared with histology sections from these tumours (Table 1), probably correspond to a higher sensitivity for anaphase bridge detection in cells not subjected to sectioning. (d) Reduction of cells with high numbers of TTAGGG-negative chromosome termini at late xenograft passages (p); each circle corresponds to the number of TTAGGG-negative ends in a single cell. (e) Multiple chromosomal termini without signals for TTAGGG repeats (examples marked by arrows), indicating short, dysfunctional telomeres in the Lux-2 xenograft passage 1. (f ) Telomeric repeat signal are visible in all termini in a cell from Lux-2 xenograft passage 5.

Cytogenetic characterization of xenografted GBM cells

To explore further the relationship between cytogenetic abnormalities and the mitotic segregation pattern in GBM, xenografts were established from the two cases Lu-1 and Lu-2. The serially transplanted xenograft GBMs Lux-1 and Lux-2 had at the end of this study been transplanted for 2.5 and 1.5 years respectively.

Chromosome banding and multicolour FISH analyses of the Lu-1 primary tumour revealed multiple structural and numerical abnormalities. Although there was extensive cytogenetic heterogeneity, reflected by nonclonal changes present in every analysed cell, two clones (A and B) with hypodiploid chromosome complements could be defined (Table 1). Clone B was found again at the first xenograft passage (P) in vivo (Lux-1 P1; Table 2), including two polyploid subclones (B1 and B2), whereas clone A could not be found. At P5 and P6, the cell population had become monoclonal, showing a karyotype (B3) consistent with further development from subclone B. A similar scenario was observed in Lu-2/Lux-2. The analysable cells from the primary tumour all showed a complex karyotype with multiple numerical and structural changes, including double minute chromosomes (dmin), and a heterogeneous pattern of nonclonal changes. The first in vivo xenograft passage (Lux-2 P1) showed extensive cytogenetic heterogeneity with at least four clones (A1-4), all of which were related to the complex karyotype in the primary tumour. Only one clone (A2a) could be found in the second passage xenograft (Lux-2 P2), highly similar to the A2 clone in P1. In the three subsequent xenograft passages (Lux-2 P3-P5) the tumour population remained monoclonal with a stable karyotype (A2b), derived from A2a by loss of several chromosomes. Thus, in both Lux-1 and Lux-2, cytogenetic heterogeneity decreased during successive passages in mice, ultimately leading to one monoclonal population in each xenografted GBM.

Table 2.  Clonal evolution during successive xenograft passages
P5, P6B343,X,−X,der(1)t(1;14)(p34;q22),−4,der(6)t(6;17)(q23;p11),+ider(7)(q10)del(7)(q35),

Mitotic segregation patterns and telomere lengths in xenografted cells

For both Lu-1/Lux-1 and Lu-2/Lux-2, the frequencies of anaphase bridges were high in primary tumour cells and then gradually decreased towards zero when the cell populations had become monoclonal (Figure 1c). At the first Lux-1 and Lux-2 in vivo passages, FISH analysis of telomeric TTAGGG repeats in metaphase chromosomes from cultured cells showed extensive variability among cells in the number of TTAGGG-negative chromosome ends (Figure 1d). In both Lux-1 and Lux-2 P1 cells, there was a subpopulation of cells with >10 TTAGGG-negative chromosome ends, indicating a high number of abnormally short telomere repeats (Figure 1e). At P5 of both lines, these subpopulations could no longer be detected and the analysed cells all showed less than 10 TTAGGG-negative telomeres (Figure 1f ). This indicates that the lower ABF values in later in vivo passages were most probably caused by a successive elimination of cells with a high number of defective, short telomeres.

Phenotypic characterization of xenografted cells

To explore the relationship between cytogenetic heterogeneity and phenotypic diversity, successive passages of xenograft tumours were also subjected to morphological and immunophenotypic analyses.

Lu-1 and Lu-2 both showed the typical morphological features of glioblastoma (grade IV according to the WHO classification) [35]: marked cellular polymorphism with nuclear atypia, scattered mitoses, vascular endothelial proliferation (Figure 2a) and necrosis. Staining with antibodies against GFAP (Figure 2c) and nestin (Figure 2e) resulted in marked positivity within most of the tumour parenchyma. The proliferation activity, by Ki-67 immunostaining, was 15% on average (Figure 2g). The Lu-1 and Lu-2 tissue sections exhibited similar staining properties within the areas of solid tumour. Xenograft GBM specimens (Lux-1 and Lux-2, P1-5) all showed a lower degree of cellular polymorphism and less atypia, with more homogenous nuclei and only minimal vascular proliferation (Figure 2b). Sections stained with GFAP exhibited only limited positivity (Figure 2d), whereas nestin staining showed homogenously moderate or marked positivity (Figure 2f). The proliferation index reached 80% in the xenograft GBM cells (Figure 2h), indicating strong growth activity.

Figure 2.

Morphological and immunohistochemical features of representative primary and xenograft glioblastoma multiforme (GBM) tissue specimens. (a,b) Haematoxylin/erythrosine staining of Lu-2 and Lux-2 respectively. (c,d) Staining with anti-glial fibrillary acid protein of Lu-2 and Lux-2 respectively. (e,f ) Staining with anti-nestin of Lu-2 and Lux-2 respectively. (g,h ) Staining with anti-Ki-67 of Lu-2 and Lux-2 respectively. Scale: bar denotes 0.1 mm.

For immunophenotypic profiling, single cells were prepared from fresh tumour tissue and analysed for detection of the cell surface markers CD133, PDGFRα, A2B5, CD44 and EGFR (Figure 3a,c). Notably, 75% of the Lu-1 and 50% of the Lu-2 GBM cells were CD133+, indicating that a substantial fraction of the primary GBM cells in these two patients were endowed with GBM stem cell features [4,5]. In addition, PDGFRα and A2B5, markers for oligodendrocyte progenies [36–39] were detected in a substantial fraction of the cells. CD44, an adhesion molecule suggested to be a marker for astrocyte precursors [40] was expressed in the majority of primary GBM cells. Similarly, EGFR which is frequently overexpressed in GBMs due to gene amplification [41] was detected in the majority of Lu-1 and Lu-2 GBM cells. Thus, both primary GBMs contained heterogeneous cell populations, including cells with cancer stem/progenitor cell features. All analysed passages (P1-P6) of Lux-1 and the early passages (P1-P2) of Lux-2 xenografts showed a great similarity in the expression of the tested cell surface markers to that of the primary Lu-1 and Lu-2 (Figure 3b,d). However, expression of A2B5 was not detected in the first passage xenograft GBM cells in either line (data not shown). Furthermore, in P3-P5 of Lux-2, the expression of CD44 and EGFR dropped to very low levels; in contrast, CD133 was consistently expressed in 30–70% of the cells. Thus, although a shift towards cytogenetic monoclonality and a more monotonous morphology occurred in late-passage xenografts, substantial immunophenotypic diversity was retained.

Figure 3.

Phenotypic characterization of primary and xenograft glioblastoma multiforme (GBM) cells by flow cytometry. (a,b) Lu-1 primary tumour and multiple Lux-2 xenograft passages respectively. (c,d) Lu-2 primary tumour and multiple Lux-2 xenograft passages respectively. Cells were stained with the indicated monoclonal antibodies (mAb). Staining with allophycocyanin (APC)-conjugated anti-CD45 mAb was used to identify the contaminating leucocytes in the primary GBM specimens. Similarly, staining with APC-conjugated anti-human MHC-I mAb was used to distinguish between xenograft GBM cells and murine cells in xenograft GBM specimens. The percentage numbers in each panel represent the proportion of cells positively stained with the indicated markers among the CD45-negative primary GBM cells or the human MHC-I positive xenograft GBM cells. The ST1 to ST11 in panels b and d represent individual xenograft tumours in SCID mice from the passages indicated.

Telomerase inhibition in xenograft cells

Assessment of telomerase expression in Lux-2 cells showed hTERT overexpression in relation to the housekeeping gene β-actin at all tested xenograft passages, with no clear differences between early and late passages (data not shown). To further explore the association between abnormally short telomeres and mitotic disturbances, we subjected the cytogenetically stable P5 Lux-2 cells, having intact telomeres, to pharmacological telomerase inhibition in order to induce telomere shortening. At evaluation after 30 days of in vitro exposure to the telomerase inhibitor MST-312, the exposed Lux-2 cell showed an elevated number of TTAGGG-negative chromosome ends (3–4 for 0.5 μM, 4–10 for 1.0 μM) compared with unexposed Lux-2 control cells (0–1 negative ends). This was coupled to elevated frequencies of telomeric associations in metaphase cells (3%, 18% and 0 for 0.5 μM, 1.0 μM and controls respectively) and also to a higher frequency of bridges in anaphase cells (10%, 15% and 1% for 0.5 μM, 1.0 μM and controls respectively). The colony formation capacity of the exposed cells was markedly reduced compared with unexposed cells (Figure 4a). However, the proportion of apoptotic and necrotic cells did not differ between exposed and unexposed cells (4–7% in all). This suggested that disturbances in mitotic proliferation, rather than an increased rate of cell death, could explain the reduced colony formation capacity. Indeed, besides anaphase bridging (Figure 4b), the exposed cells showed displacement of whole chromosomes at metaphase and anaphase at high frequencies (Figure 4c; 12%, 48% and 2% for 0.5 μM, 1.0 μM and controls respectively). Furthermore, the frequency of polyploid cells was elevated in exposed cells (17%, 43% and 3% for 0.5 μM, 1.0 μM and controls respectively). Thus, pharmacologically induced telomere shortening in late-passage, mitotically stable GBM xenograft cells reproduced the elevated frequencies of anaphase bridging in early passage tumours and also led to other types of cell division errors, supporting a mechanistic link between telomere shortening and mitotic instability.

Figure 4.

MST-312 treatment of Lux-2 cells. (a) Reduced colony formation capacity (striped bars = mean number of colonies ±95% confidence interval) and elevated frequencies of displaced chromosomes at metaphase (white bars), and bridges and displaced chromosomes at anaphase (black bars) in treated cells compared with untreated cells. (b) Anaphase cell with chromatin bridge (long arrow) and failed sister chromatid separation (short arrow). (c) Anaphase cell with a displaced chromosome (arrow) with unseparated sister chromatids.


In contrast to low-grade brain tumours, extensive genetic heterogeneity has been demonstrated in high-grade neoplasms such as GBM [7,14]. The mechanisms behind this cytogenetic heterogeneity have been little explored. In the present study, we found mitotic instability in the form of anaphase bridging in all GBMs analysed, as well as in two other high-grade gliomas. Anaphase bridging was also found in medulloblastomas, but not in low-grade gliomas or in ependymomas. Tumours with anaphase bridges typically had complex karyotypes with both numerical and structural changes. In contrast, tumours without anaphase bridges had either few (<3) cytogenetic changes or numerical changes only. These findings are consistent with previous studies, showing a strong link between anaphase bridging and the generation of unbalanced structural changes [19,32]. Most of the tumours showing anaphase bridges in the present study were high-grade lesions. However, our limited series of patients is not sufficient for drawing any final conclusions regarding the relationship between mitotic instability and the degree of aggressiveness in brain tumours. A number of studies have demonstrated telomeric associations in low-grade as well as high-grade astrocytic tumours [42–46]. Furthermore, in paediatric low-grade gliomas, short telomeres have been associated with a lower risk of recurrence [47]. It was not investigated in these studies whether the telomeric fusions or the telomere shortening were associated with anaphase bridging. However, telomeric associations have been shown to result in anaphase bridges in low/borderline malignant soft tissue tumours [18]. One can therefore not exclude that occasional low-grade brain tumours will also show features of mitotic instability. The present study should thus only be regarded as a first survey, demonstrating one possible mechanism behind cytogenetic heterogeneity in tumours of the central nervous system.

Anaphase bridging can be caused by abrogated protection of chromosome ends due to abnormal shortening of telomeric repeat sequences [19,20]. Abnormal telomere shortening was indeed found in early passage xenograft GBM cells exhibiting cytogenetic diversity and a high frequency of anaphase bridging. In accordance with previous studies of GBM xenografts, the overall karyotypic complexity in terms of the number of clonal aberrations was retained in the xenografts through successive passages [26,27]. However, both the cytogenetic heterogeneity and the telomere-mediated instability decreased radically during serial in vivo passaging. This concurrent reduction of cytogenetic heterogeneity, anaphase bridging, and the proportion of cells with abnormally short telomeres indicates that telomere dysfunction contributes to genetic heterogeneity in GBM cells. This is further supported by the fact that re-introduction of telomere shortening in late passage xenograft cells by telomerase inhibition led to a dramatic elevation of fused chromosome ends and anaphase bridges. Whether this also led to the re-emergence of new clones could not be evaluated as the telomerase inhibition led to a severe reduction in tumour cell growth. It should ultimately be pointed out that there are also several other potential sources of anaphase bridging, including ionizing radiation, clastogenic agents, and other factors producing double-stranded DNA breaks. None of the patients in the present study had been subjected to preoperative radio- or chemotherapy. However, this does not completely rule out that also telomere-independent causes had a role in promoting mitotic instability in their tumours.

Compared with the primary GBMs, xenograft tumours showed homogeneous morphology with less nuclear atypia than the primary tumours. Considering that a high number of cells (2 × 106) were transferred at each mouse–mouse passage, the loss of cytogenetic and morphological heterogeneity was not likely a result of selection at the grafting procedure. It is more probable that it resulted from selection during in vivo growth. It has been shown that abnormal nuclear morphology could be a consequence of abnormal chromatid separation at cell division [48]. At least for the late passage xenografts, the reduction in nuclear polymorphism and atypia could thus be a direct consequence of a reduction in the number of cells undergoing anaphase bridging. In contrast, flow cytometry demonstrated that phenotypically heterogeneous cell populations including CD133+ cells were regenerated to some degree at all passages. Thus, the intratumour immunophenotypic diversity observed in GBM is not necessarily correlated to cytogenetic heterogeneity. The capacity for generation of phenotypically heterogeneous cell populations of the late-passage cytogenetically monoclonal populations, including CD133+ cells, could indicate that these, genetically stable clones partly consisted of long-term GBM stem cells. If so, cytogenetic aberrations common to Lux-1 and Lux-2 could point to genetic changes of importance to the long-term proliferation of GBM tumour cells. The only breakpoint in common between the two xenograft lines was 14q22. This chromosome band has not previously been denoted as a recurrent breakpoint in GBM to our knowledge, and its significance is presently unclear.

There were no significant differences in the rate of telomerase expression in primary GBM cells compared with the xenograft cells. However, inhibition of telomerase activity in xenograft cells in vitro resulted in a dramatically reduced proliferation capacity. This implies that the growth of GBM cells in the xenograft model was telomerase-dependent. Previous studies have demonstrated telomerase activity in the majority of GBMs [49] and it has been shown that telomerase activity correlates with higher grade, a shorter survival time, and the presence of TP53 mutations in gliomas [50,51]. The telomerase inhibitor GRN163 has been shown to reduce the in vivo growth of GBM xenografts [24]. Moreover, inhibition of telomerase with an antisense telomerase expression vector has been reported to potentiate cisplatin-induced apoptosis in a GBM cell line [52]. In the present study, telomerase inhibition in Lux-2 cells caused a marked reduction in colony formation capacity after 1 month of treatment. Considering that the late-passage Lux-2 cells had relatively stable, long telomeres prior to treatment, this degree of latency is expected if the effects of MST-312 were mediated by telomere-length reduction and subsequent telosome destabilization [34]. There was no significant increase in the proportions of apoptotic and necrotic cells after treatment. It is possible that nonviable, detached cells were lost from the sampled population. Therefore, our study does not entirely exclude apoptosis and necrosis as the ultimate consequence of telomerase inhibition. Still, the low rates of cell death is in accordance with previous studies showing that high-grade malignant cells may sustain a high degree of telomere-dependent genomic instability without impaired cellular viability [32]. The MST-312-treated Lux-2 cells did show a higher frequency of chromatid segregation anomalies than the untreated control cells, indicating that telomerase inhibition led to impairment of the mitotic process. This was supported by an elevated frequency of polyploid cells in the MST-312-treated cultures, accompanied by the accumulation of large, irregular nuclei, possibly reflecting abortion of the mitotic process prior to cytokinesis. Thus, our data indicate that GBM xenograft cells are dependent on telomerase activity to sustain normal chromatid segregation.

In conclusion, the present study shows that telomere-dependent anaphase bridging may be a common mechanism for the generation of cytogenetic heterogeneity in GBM as well as in other high-grade brain tumours. Although immunophenotypic heterogeneity is maintained to some degree in GBM xenografts, the level of cytogenetic heterogeneity and the rate of mitotic instability decrease during passaging in vivo towards a cytogenetically monoclonal cell population, the genetic stability of which is telomerase-dependent. The attenuation of genetic diversity in long-term xenograft tumours should be taken into consideration when xenograft models are used to study GBM biology and response to candidate therapeutic agents. This is particularly important because genetic heterogeneity is generally believed to be a key factor in the development of resistance to chemotherapeutic drugs [53].


This study was supported by grants from the Hans and Märit Rausing Charitable Foundation, the Swedish Cancer Society, the Swedish Children's Cancer Fund, the Royal Physiographic Society in Lund, the Siv-Inger & Per-Erik Anderssons Minnesfond, the Funds of Lund University Hospital, the Gunnar Nilsson Cancer Foundation, the Erik Philip-Sörensen Foundation, the Swärd Foundation, and the Lund University Medical Faculty. The authors thank Seema Rosqvist for critical reading of the manuscript. Xiaolong Fan is a Li Ka Shing Scholar.