Telomere-Driven Karyotypic Complexity Concurs with p16^INK4a Inactivation in TP53-Competent Immortal Endothelial Cells

The limited replicative life span of endothelial cells is largely attributed to the shortening of telomeres with each cell division (1). Telomeres are chromosomal-end structures that are composed of TTAGGG DNA repeats and folded into t-loops (2, 3). Telomeres function to protect against DNA degradation and prevent chromosomal fusions. However, critically short telomeres enable chromosome fusions, which contribute to the development of aneuploidy and chromosomal instability (4–6). Chromosome rearrangements resulting from telomere dysfunction drive progression of human malignancies (7–10). Excessive proliferation of endothelial cells is central to a number of premalignant and malignant disorders, including hemangioma, hemangioblastoma, and angiosarcoma (11). The molecular mechanisms underlying endothelial cell malignancies are not well defined, although a number of studies have shown complex karyotypes in angiosarcomas (12, 13).

In normal cells, telomere dysfunction and chromosomal abnormalities trigger TP53-mediated DNA damage signals that halt cell cycle progression and induce senescence (14–16). Inactivation of TP53 circumvents DNA damage signals, enables continued proliferation of cells with dysfunctional telomeres, and thereby promotes chromosomal instability and transformation (5, 17). The cyclin-dependent kinase inhibitor p16^INK4a is also up-regulated in senescent cells. Expression of p16^INK4a inhibits phosphorylation of the retinoblastoma protein (pRB) and prevents cell cycle progression (18). A recent study indicated that up-regulation of p16^INK4a in senescent cells may also be a consequence of telomere dysfunction (19).

Enforced expression of SV40 T antigens (SV40Tag) or the human papilloma virus E6 and E7 oncogenes, which inactivate pRB and TP53 functions, extends the replicative life span of human endothelial cells (20, 21). However, telomeres in viral antigen–transformed cells continue to shorten, resulting in extensive chromosomal instability and eventual genetic catastrophe, which terminates the culture in a growth crisis (4). As a rare event, viral antigen–transformed cells escape crisis and are immortalized as a consequence of spontaneous activation of a telomere maintenance mechanism (4, 22).

The enzyme telomerase maintains telomeres in 80% to 90% of cancer cells (23). Ectopic expression of telomerase reverse transcriptase (hTERT) was shown to be sufficient to reconstitute telomerase enzyme activity and immortalize specific fibroblast and epithelial cell strains (24). However, inactivation of p16^INK4a was implicated as a co-operating event in hTERT-mediated immortalization of other cell strains (25–30). Previous studies, including our own, showed that the life span of human endothelial cells was extended by overexpression of hTERT (31, 32). However, we found that two of three mass cultures of hTERT-transduced bone marrow endothelial cells (BMEC) underwent a growth crisis during immortalization (32). We concluded that telomerase was insufficient for immortalization and further showed that the combination of SV40Tag and hTERT very efficiently immortalized BMECs. Our results were consistent with investigations that showed mammary endothelial cells were immortalized by SV40Tag and hTERT, but not hTERT alone (33). In contrast, another report indicated that expression of hTERT was sufficient for immortalization of several other types of endothelial cells (31). These contrasting reports may partially reflect strain differences, as altered culture conditions did not reconcile our study with the latter (32). Despite the apparent differences, an important observation that was common to our study, and the latter was that telomeres continued to shorten in hTERT-transduced endothelial cells despite high levels of telomerase activity (31, 32).

Before the current investigation, the consequences of telomere shortening in hTERT-transduced endothelial cells have been unexplored. In the present study, we show that telomere shortening in hTERT-transduced BMECs results in chromosomal rearrangements, which mirror the types of aberrations found in human cancer cells, including angiosarcoma. Furthermore, complex karyotypes developed exclusively in a subset of cells that silenced p16^INK4a. These investigations show that reconstituted telomerase activity is not sufficient to prevent telomere dysfunction in human endothelial cells and are the first to show a relationship between p16^INK4a inactivation and the development of karyotypic complexity in continuously proliferating TP53-competent cells.

Cell culture. The establishment and propagation of hTERT-transduced BMEC mass cultures (BMhTERT-1 and BMhTERT-2) was previously described (32). BMhTERT clones were established by limiting dilution of the BMhTERT-1 mass culture at 28 population doublings in a 96-well plate. Wells containing a single cell were identified under an inverted microscope.

Telomerase activity and telomere length measurements. Telomerase activity was quantified using the telomeric repeat amplification protocol (TRAP), as previously described (32). The SK-N-SH neuroblastoma cell line (NB) was used as a positive control, and an internal control (TSNT) was included for quantitations. Telomerase activity was calculated as: 100 × [(T − B) / TSNT] / [(NB − B) / TSNTNB], where T is sample, B is background, and TSNTNB is the internal control of the neuroblastoma. Mean telomere restriction fragment (TRF) length was measured using the TeloTAGGG telomere length assay kit (Roche, Indianapolis, IN) as previously described (30).

Cytogenetics. Cytogenetic analysis was done at Sydney Genetics, Sydney IVF (Sydney, Australia) using standard GTG banding techniques and examination of chromosomes at a resolution of 400 to 550 bands per haploid set. Chromosomal abnormalities were described according to the International System for Human Cytogenetic Nomenclature (1995).

May Grunwald-Giemsa staining. Cells (1 × 10⁵) were resuspended in 100 μL PBS, cytocentrifuged onto slides at 60 × g for 5 minutes, then stained with May-Grunwald stain (BDH Laboratory Supplies, Poole, England) for 1 minute, rinsed in H₂O, immersed in 10% Giemsa stain for 12 minutes, and rinsed again in H₂O.

Fluorescent in situ hybridization. Fluorescence in situ hybridization (FISH) analysis was done at Sydney Genetics, Sydney IVF using probes for chromosome 17 centromere (CEP17 spectrum aqua, VYSIS, Downers Grove, IL), the TP53 locus (LSI p53 spectrum orange, VYSIS) and the subtelomeric region of 17p (TelVysion 17p spectrum green, VYSIS). Images were digitally captured using CytoVision v2.7 software (Applied Imaging, Santa Clara, CA).

Reverse transcription-PCR. Total RNA was extracted using Trizol Reagent (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions. First-strand cDNA was synthesized from 0.5 μg total RNA using 10 units M-MLV Reverse Transcriptase (Life Technologies, Gaithersburg, MD). The PCR reaction, including 100 ng cDNA, 2× SYBR Green PCR master mix (Applied Biosystems, Foster City, CA) and previously published primer sets for p16^INK4a and β₂-microglobulin (34), was done on a ABI PRISM 7000 Sequence Detection System (Applied Biosystems) at 94°C for 10 minutes, then 40 cycles at 94°C for 30 seconds, 60°C for 30 seconds, and 72°C for 30 seconds, with a final step at 72°C for 10 minutes. PCR products were resolved through 1.5% agarose gels and visualized using Versadoc imaging system (Bio-Rad Laboratories, Hercules, CA).

Immunoblot analysis. Protein was isolated and immunoblot analysis done as previously described (30). Proteins were detected with the following antibodies: p53 (DO-1) monoclonal antibody (Santa Cruz Biotechnology, Santa Cruz, CA), pRB, p16^INK4a, and p21^Cip1 monoclonal antibodies (BD Biosciences PharMingen, San Diego, CA), and actin polyclonal antibody (Sigma, St. Louis, MO).

CDKN2A promoter methylation analysis. The bisulfite reaction was carried out for 16 hours at 55°C on 2 μg of digested DNA, then semi-nested PCR amplifications were done as previously described (35, 36). PCR fragments were analyzed in 2× SYBR Green 1 Master mix using real-time PCR melting temperature dissociation, cycling through 95°C for 15 seconds, 60°C for 20 seconds, with the temperature increasing from 60°C to 90°C. CpGenome Universal Methylated Control DNA (Chemicon International, Temecula, CA) and Human Genomic DNA (Roche) were used as controls for fully methylated DNA and unmethylated DNA, respectively. Melting dissociation of the trace was analyzed on an ABI prism 7700HT Sequence Detection System. Results from dissociation curves were confirmed by direct sequencing and clonal sequencing of PCR products (35).

Sequence analysis. Genomic DNA was extracted by standard phenol/chloroform procedure and precipitated with 0.3 mol/L sodium acetate and 2 volumes of ethanol. Exons 2 to 11 of TP53 gene were amplified by PCR using commercial primers from the Human p53 Amplimer Panel (Clontech, Palo Alto, CA) according to the manufacturer's protocol. PCR products were purified using BRESAspin PCR purification kit (GeneWorks, Adelaide, Australia) and sequenced by SUPAMAC (Camperdown, Australia).

hTERT-transduced BMECs undergo crisis during immortalization. BMhTERT-1 and BMhTERT-2 mass cultures were established by transduction of BMECs with a retroviral vector encoding hTERT (32). During immortalization, the BMhTERT-2 cell line underwent a prolonged growth crisis, whereas the BMhTERT-1 culture exhibited no clear crisis but had a reduced growth rate compared with early passage cells. For the present investigations, 32 clones were established by limiting dilution of the BMhTERT-1 mass culture at 28 population doublings (Fig. 1A). Out of 32 clones, 26 had an extended life span compared with parental BMEC cultures, which senesced at ∼35 population doublings. Twenty of these 26 clones underwent a growth crisis, characterized by an overall decrease in proliferation, increased cell death, the appearance of cells with a senescent morphology, and demonstrable senescence-associated β-galactosidase activity (data not shown). The onset of crisis occurred at ∼50 population doublings. The duration of crisis varied, lasting up to 200 days for some clones. Most clones (16 of 20) eventually escaped crisis and resumed proliferation. Overall, 22 of 26 (85%) of the BMhTERT clones that bypassed senescence proliferated beyond 100 population doublings (∼3-fold increase in life span) and were considered immortalized. These results confirm that BMECs undergo a crisis during immortalization and show that crisis may be masked in mass cultures by clonal heterogeneity.

Crisis was independent of telomere length and the level of telomerase activity. To determine whether crisis was triggered by short telomeres, the mean telomere length of BMhTERT clones was measured at multiple time points (Fig. 1B and C). The mean telomere length of BMhTERT-1 mass cultured cells declined from 8.5 kbp (38 population doublings) at a rate of 67 bp/population doubling, until it stabilized around 3.4 kbp (115-130 population doublings; Fig. 1C). These results were striking considering we previously found that parental BMECs senesced with an average telomere length of 6.0 kbp (32). Consistent with the mass culture, the mean telomere length of most clones eventually stabilized around 3.0 kbp (between 3.6 and 2.4 kbp; Fig. 1C). Indeed, clones BM2A11 and BM2A9 maintained telomere length in this range until at least 179 and 281 population doublings, respectively (additional data not shown). The apparent shortening of mean telomere length at earlier population doublings (<85) and eventual maintenance at a subsenescent length were likely a reflection of preferential maintenance of critically short telomeres (37, 38).

At the onset of crisis (around 50 population doublings), the telomere length of different BMhTERT clones ranged from 12.5 to 2.1 kbp (Fig. 1C). However, there was no significant difference in the average telomere length of clones that went into crisis (7.6 ± 0.7 kbp, n = 6) compared with clones that showed no distinct crisis (7.7 ± 0.8 kbp, n = 9; Student's t test, P = 0.8925). Furthermore, the average telomere length was significantly shorter at late passage (4.9 ± 0.7 kbp, n = 14, at >65 population doublings) compared with earlier time points (7.7 ± 0.5 kbp, n = 15 at <65 population doublings; Student's t test, P = 0.006). These results are consistent with the decline in mean telomere length observed in the mass culture but are converse to the results expected if crisis was due to telomere shortening. To investigate whether shortening of mean telomere length or crisis was due to low levels of telomerase activity, the TRAP assay was done on representative clones (Fig. 1D; additional data not shown). Quantitation of telomerase activity against an internal control and tumor cell line revealed no significant difference in telomerase activity assayed at pre- and post-crisis time points (25.2 ± 4.4, n = 13 compared with 27.1 ± 3.8, n = 9, respectively; Student's t test, P = 0.75). Regression analysis showed no significant correlation between the level of telomerase activity and mean telomere length of BMhTERT clones at crisis (n = 17, r² = 0.03, P = 0.53). Overall, results from TRF and TRAP analyses showed that (a) telomeres were maintained at a subsenescent average telomere length; (b) crisis was independent of the mean telomere length and the level of telomerase activity; and (c) shortening of mean telomere length in BMhTERT clones was not due to low levels of telomerase activity.

Telomere dysfunction leads to the evolution of cytogenetic abnormalities. Karyotypic analyses were done to investigate potential telomere dysfunction in BMhTERTs. No cytogenetic abnormalities were detected in the parental BMECs and the BMhTERT-1 mass culture at early passage (35 population doublings; Table 1). However, an additional copy of chromosome 11 (+11) was detected in BMhTERT-1 at 65 population doublings and persisted until at least 135 population doublings. No other abnormalities were detected in this mass culture, despite telomere shortening to 3.5 kbp at 135 population doublings. Trisomy 11 was also detected in 9 of 12 BMhTERT clones (Table 1). It is important to note that the clones were established from the BMhTERT-1 mass culture at 28 population doublings, which was before the appearance of +11 in the mass culture. Other recurring cytogenetic abnormalities that were present in three or more cell lines included −Y, del(1)(q12), i(5)(p10), +5, −13, add(13)(p11.2), t(2;17)(p13;p13.3), +19, add(19)(q13.1), +20, and −21. There were also various abnormalities affecting chromosome arm 2p in four clones. To determine whether the recurring abnormalities were present at a low frequency in the mass culture at the time of cloning, an additional 400 BMhTERT-1 cells at 33 population doublings were scored for aberrations involving the chromosomes most frequently affected (i.e., chromosomes 2, 11, 13, 17, 19, 21, and 22). No abnormalities of these chromosomes were detected in 400 metaphases.

The mass cultures and clones could be clearly subgrouped according to karyotypic complexity. Simple karyotypes, defined as near diploid with only one to two abnormalities, were detected in 7 of 12 clones and the BMhTERT-1 mass culture. In contrast, the remainder of the clones (5 of 12) and the BMhTERT-2 mass culture had complex karyotypes, characterized by at least five chromosomal abnormalities and frequent polyploidy. The detection of cytogenetically related sublines within clones with complex karyotypes were indicative of karyotypic evolution. BM2A4 had the most complex karyotype and exhibited overt chromosomal instability, characterized by a high frequency of random aberrations and numerous non-clonal marker chromosomes. Cells with simple verses complex karyotypes could not be distinguished based on proliferative rate or whether they went through crisis. Telomere length dynamics also did not predict karyotypic status. For instance, both BM2A7 (simple karyotype) and BM2B12 (complex karyotype) had a mean telomere length of about 5.0 kbp at 50 population doublings and showed shortening to around 3 kbp by ∼110 population doublings.

Short dysfunctional telomeres promote chromosomal aberrations via breakage-fusion-bridge (BFB) cycles, which are initiated by the fusion of sister chromatids that have lost telomeres. BFB events promote specific patterns of chromosome abnormalities, including nonreciprocal translocations, deletions, amplifications, and whole chromosome losses (5, 8, 39). The possible involvement of BFB events in karyotypic evolution of BMhTERTs was investigated by screening for dicentric chromosomes, telomere associations, and ring chromosomes, which are the hallmarks of telomere dysfunction and BFB cycles. In BM2B8, which had a complex karyotype with two evolving sublines, 12 of 155 G-banded metaphase cells harbored dicentric chromosomes (5 cells) or telomere associations (7 cells; Fig. 2A). Six of 12 of these telomere associations and dicentrics involved chromosome arms that were detected in clonal rearrangements in this cell line at a later time point (additional data not shown). In the BM2A4 cell line, 12 of 20 (60%) cells showed evidence of BFB events, including ring chromosomes, telomere associations, and anaphase bridges. Multipolar mitoses and lagging chromatin, which are associated with a high rate of anaphase bridging, were also detected in BM2A4 (Fig. 2B). In contrast, no BFB events were detected in BMhTERT-1 (100 cells scored), which exhibited a simple karyotype and no evidence of heterogeneity or evolution. The nonreciprocal translocations and chromosome losses detected in BMhTERTs are consistent with BFB-generated karyotypic changes (39). Taken together, these data indicate telomere dysfunction and BFB events drove the evolution of karyotypic complexity in BMhTERT cells.

Loss of the subtelomeric region of chromosome arm 17p in all complex karyotypes. The nonreciprocal translocation t(2;17)(p13;p13.3), which involved translocation of the entire short arm of chromosome 2 to 17p13.3, was detected in the BMhTERT-2 mass culture and four of five clones with complex karyotypes. BM2A4, which also had a complex karyotype, did not have t(2;17)(p13;p13.3) but had alternative rearrangements of 17p [der(17)t(3;17)(q10;p10) × 2 and add(17)(p10) × 2] that resulted in loss of the entire 17p arm. Because the TP53 gene is located on 17p, FISH analysis was done for TP53, as well as 17p subtelomeric sequence and chromosome 17 centromere in BM2A4 and BM2B8 [an example of a cell line with t(2;17)(p13;p13.3); Fig. 2C]. The results showed that all rearranged 17p arms in both cell lines had lost subtelomeric DNA, whereas the TP53 locus was retained on all copies of chromosome 17 in the BM2B8 cells, including chromosomes with the t(2;17)(p13;p13.3) rearrangement. Although TP53 was also retained on two apparently normal copies of chromosome 17 in BM2A4, it was deleted from chromosomes with der(17)t(3;17)(q10,p10) and add(17)(p10). Thus, whereas loss of 17p subtelomeric sequence was common to all clones with complex karyotypes, the TP53 locus was only lost in association with der(17)t(3;17)(q10;p10) and add(17)(p10), which only occurred in BM2A4.

Repression of p16^INK4a by methylation occurs in a subset of clones. Immunoblot analysis of p16^INK4a showed readily detectable protein in the BMhTERT-1 mass culture at late passage (131 population doublings), as well as in BM2A7, BM2A11, BM2B2, BM2A6, BM2A10, BM2A2, and BM10A3 at post-crisis time points (Fig. 3A). In contrast, p16^INK4a was undetectable in the post-crisis BMhTERT-2 mass culture and BM2A9, BM2B8, BM2B12, and BM2A4 clones. Down-regulation of p16^INK4a in the latter cell lines seemed to be complete, as there was no mRNA detected by real-time PCR comparison of these cell lines with the MCF-7 cell line, which has a homozygous deletion of the CDKN2A locus (Fig. 3B; additional data not shown). p16^INK4a protein was detected at a very low level in BM10B1 at 107 population doublings (Fig. 3A), whereas real-time PCR analysis of clone BM10B1 revealed robust mRNA expression at 87 population doublings but no detectable p16^INK4a mRNA by 112 population doublings (Fig. 3B). Notably, the timing of p16^INK4a repression in BM10B1 at about 107 population doublings did not correspond with the main growth crisis, which began at 52 population doublings in this cell line. Overall, p16^INK4a protein expression was repressed in 5 of 12 post-crisis clones.

We investigated the possibility that p16^INK4a was silenced by methylation of CDKN2A CpG island promoter regions by performing bisulfite methylation followed by PCR amplification (36). Analysis of the heat dissociation curves of the PCR products, as well as direct and clonal PCR sequence analysis showed a direct correlation between promoter hypermethylation and p16^INK4a silencing (Fig. 3C; Supplementary 1; Table 2). For example, CDKN2A promoter was methylated and inactivated in clones BM2A9, BM2B8, and BM2B12, as well as BMhTERT-2 mass culture at late passage. Furthermore, the CDKN2A promoter was methylated in BM10B1 at 115 population doublings, but not 90 or 75 population doublings, which is consistent with expression data (Fig. 3A and B; Supplementary 1; Table 2). Hypermethylation is, therefore, the likely cause for gene silencing in the BMhTERT-2 mass culture and in four of five clones. The CDKN2A promoter sequence could not be amplified from clone BM2A4, which suggests that p16^INK4a expression was silenced in this cell line by gene deletion or rearrangement. Overall, these data show that p16^INK4a expression was extinguished before 120 population doublings in 5 of 12 BMhTERT clones and one of two mass cultures. Further, inactivation of p16^INK4a may occur but is not essential for hTERT-mediated immortalization of BMECs.

When the clones were grouped according to karyotypic complexity, it was striking that p16^INK4a expression was repressed exclusively in the five clones and one mass culture that had complex karyotypes (Table 2). The relationship between p16^INK4a repression and karyotypic complexity was highlighted by the time course of BM10B1 (Fig. 3; Tables 1 and 2). This clone expressed p16^INK4a and had a simple karyotype at 75 population doublings, then developed a complex karyotype at 108 population doublings when p16^INK4a was down-regulated and showed clonal heterogeneity by 164 population doublings. These data provide evidence that inactivation of p16^INK4a permits accumulation of cytogenetic aberrations, and, conversely, that p16^INK4a functions as a gatekeeper to prevent karyotypic evolution.

Dysfunction of pRB phosphorylation in a subset of post-crisis clones. To investigate whether repression of p16^INK4a affected pRB function, immunoblot analysis of phosphorylated and hypophosphorylated pRB was done on BMhTERTs treated with 25 μmol/L of the DNA damage agent etoposide (VP-16; Fig. 4A). As expected, pRB was detected in untreated parental BMECs as a broad band, representing both phosphorylated and hypophosphorylated forms, and was down-regulated and detected as a single low molecular weight band, representing the hypophosphorylated form, in VP-16-treated BMECs. The same normal response to VP16 treatment was observed in four pre-crisis cultures (BMhTERT-2, BMhTERT-1, BM2A2, and BM2A10), as well as all post-crisis clones that retained p16^INK4a expression. Among the five clones and one mass culture that silenced p16^INK4a, two clones (BM2A9 and BM10B1) and the mass culture (BMhTERT-2) also showed a normal pRB response, whereas pRB remained up-regulated and phosphorylated following treatment of BM2B8, BM2B12, and BM2A4, which were also p16^INK4a negative. BM10B1 at 101 population doubling showed a weak response, which was consistent with incomplete down-regulation of p16^INK4a at this time point. Overall, the data show that dysfunction of pRB phosphorylation corresponded with inactivation of p16^INK4a in some but not all BMhTERT cultures. Notably, the three clones that exhibited pRB dysfunction all developed polyploid sublines (Table 1). In contrast, the three cell lines that were p16^INK4a negative but retained normal pRB function were near diploid at the time of the assay.

Loss of TP53 function is infrequent during immortalization of BMhTERT cells. The TP53 DNA damage response was investigated by treating cells with VP-16, then assessing expression of p53 and its transcriptional target p21^Cip1 by immunoblot analysis (Fig. 4B). In the parental BMECs, expression of p53 and p21^Cip1 was elevated 6 hours after drug treatment. Expression of p21^Cip1 remained high for at least 24 hours. Induction of p53 and p21^Cip1 followed a similar pattern in both pre- and post-crisis mass cultures, as well as in 11 of 12 post-crisis clones. Sequence analysis of exons 2 to 11 of the TP53 gene confirmed wild-type sequence in all clones with complex karyotypes (data not shown). BM2A4 was the only clone to show TP53 dysfunction, as it expressed no steady state or induced p53 and p21^Cip1 proteins (Fig. 4B). There was also no p53 mRNA detected in this clone by reverse transcription-PCR (data not shown). Thus, TP53 was transcriptionally silenced on the apparently structurally normal copies of chromosome 17 detected in this cell line (Fig. 2C). It was notable that the p53/p21^Cip1 pathway was only defective in BM2A4, as this was the clone that showed the highest degree of karyotypic complexity and overt chromosomal instability (Tables 1 and 2). Together, the data indicate that disruption of TP53 function was an infrequent event during immortalization of BMhTERTs. Furthermore, these data indicate that TP53 dysfunction was not necessary for the evolution of complex karyotypes, although it may have conferred more overt genomic instability in BM2A4.

Our results showing that complex karyotypes evolved exclusively in BMhTERT cultures that silenced p16^INK4a provide compelling evidence that p16^INK4a plays a role in the maintenance of genomic integrity. Complex karyotypes in BMhTERT cells seemed to be the consequence of telomere dysfunction, as dramatic telomere shortening, telomere associations, and dicentric chromosomes were detected in cells with complex karyotypes. The nonreciprocal translocations, large deletions, and chromosome losses that were detected in BMhTERT karyotypes are typical outcomes of BFB cycles and classic features of human carcinomas and premalignant conditions (7–10).

It is well accepted that TP53-mediated DNA damage signals are elicited by telomere dysfunction. However, there is still considerable debate over whether telomere attrition also induces p16^INK4a, or if telomere-independent stresses, such as oxidative stress and culture stress, are responsible for induction of p16^INK4a during senescence (16, 19). There are no prior reports of p16^INK4a inactivation in human endothelial cells. However, previous studies have shown chromosomal aberrations and evidence of BFB cycles in mortal mammary epithelial cells that silenced p16^INK4a by hypermethylation and proliferated beyond senescence (40, 41). However, in contrast to the immortalized endothelial cells described here, the epithelial cells in the latter studies did not express telomerase and eventually ceased proliferating. The most likely explanation for the correlation among p16^INK4a repression, BFB events, and the evolution of complex karyotypes in BMhTERTs is that p16^INK4a inactivation circumvented antiproliferative signals elicited by short telomeres and thereby enabled chromosomal abnormalities to accumulate. Because the TP53 DNA damage pathway was defective in only one of six BMhTERT cell lines with complex karyotypes, our results are consistent with the possibility that p16^INK4a is involved in TP53-independent antiproliferative effects elicited by telomere dysfunction (19). TP53 dysfunction promotes chromosomal instability by enabling proliferation of premalignant cells with dysfunctional telomeres (5, 17). Our results suggest that repression of p16^INK4a plays an analogous role in karyotypic evolution. Furthermore, our investigations show that a competent TP53 DNA damage response pathway does not prevent telomere-driven karyotypic evolution. Nevertheless, TP53 function did seem to confer some genetic stability in BMhTERTs, as the clone that expressed neither p16^INK4a nor p53 displayed the most overtly unstable karyotype. The apparently additive effect of TP53 dysfunction may reflect the role of TP53 at both G₁ and G₂ cell cycle checkpoints, whereas p16^INK4a is thought to function only at G₁ (18, 42).

Strikingly, many chromosome aberrations were common to different BMhTERT clones. Recurrent chromosome abnormalities occur in certain human cancers and were previously shown in association with telomere dysfunction in late-generation telomerase-negative mice (8). The recurrence of multiple chromosomal abnormalities may be attributed to an inherent susceptibility of the chromosomes with the shortest telomeres to form bridges and drive karyotypic evolution (43). The t(2;17)(p13;p13.3) translocation was the most frequent structural rearrangement detected in immortal BMhTERT cells (4 of 12 clones and 1 mass culture). Loss of subtelomeric DNA from chromosome arm 17p implicates telomere dysfunction in the generation of this abnormality. However, the single most common karyotypic alteration in BMhTERT cells was trisomy 11, which occurred in 9 of 12 clones and 1 mass culture. Other frequent whole chromosome changes included loss of chromosomes 13 (6 of 12 clones) and Y (5 of 12 clones), as well as gains of chromosomes 19 (6 of 12 clones) and 20 (5 of 12 clones). Trisomy 11 and monosomy 13 frequently arise in cultures of normal endothelial cells and have an increased incidence in cultures approaching senescence (44, 45). However, it is unclear how BFB events would give rise to whole chromosome gains, such as +11, in the absence of other rearrangements. Furthermore, because no telomere fusions were detected in BMhTERT-1 cells, it seems likely that trisomy 11 was generated by an alternative mechanism.

The cytogenetic aberrations that were most frequently detected in post-crisis clones were not detected in 400 metaphase cells prepared from the BMhTERT-1 mass culture at early passage. However, we cannot rule out the possibility that these abnormalities were present at a very low frequency (<0.25%) and conferred a proliferative advantage or co-operated with hTERT in the immortalization process. Consistent with this possibility, many of the recurring aberrations have also been implicated in leukemia and carcinogenesis (46). For instance, trisomies 11 and 19 have been detected in myeloid leukemias. Monosomy 13 is indicative of a poor prognosis in multiple myeloma and is also observed in retinoblastoma and angiosarcoma (12, 46). Other abnormalities that angiosarcomas have in common with BMhTERT cells include cytogenetic complexity, aneuploidy, trisomies 20 and 11, loss of chromosome Y, and 17p deletions (12, 13). Thus, further characterization of genetic alterations in BMhTERT cells may provide insight to the molecular biology of these highly malignant endothelial cell–derived tumors, which remain poorly characterized due to their rarity. Subtelomeric deletion of 17p in association with t(2;17)(p13;p13.3) may have affected a number of putative tumor suppressor genes located at 17p13.3, such as HIC1, OVCA1, OVCA2, and LOST1, which are implicated in various human malignancies (46). Changes in gene copy number, resulting from the 17p rearrangement and other chromosomal aberrations, will be more precisely resolved in future studies by comparative genomic hybridization array analysis.

Our results, showing that p16^INK4a was repressed in a subset (5 of 12) of immortal clones, and that the TP53 pathway was dysfunctional in just one clone, indicate that defects in these pathways were not essential for immortalization of BMECs. Varied expression of p16^INK4a was also noted when two hTERT-immortalized retinal pigmented epithelial clones were analyzed in a previous study (47). It is possible that these variations reflect clonal differences in p16^INK4a expression in the parental cell strains. The growth crisis observed in 20 of 26 BMhTERT clones implicates additional molecular events in hTERT-mediated immortalization of BMECs. Time course of the BM10B1 cell line showed that crisis was unrelated to p16^INK4a and TP53 status. Crisis also did not seem to be related to specific cytogenetic aberrations or telomere length. The molecular events involved in crisis in BMhTERTs, therefore, remain to be determined. Clonal variations in the timing of crisis resulted in a masking effect in the BMhTERT-1 mass culture, which did not show a clear crisis. Similarly, we have previously shown proliferative variability among clones derived from a mass culture of hTERT-transduced MRC5 lung fibroblasts (30). Clonal variations, such as these, may contribute to some of the discordance among independent investigations into the effects of hTERT in various types of human cells.

The lack of mean telomere lengthening in BMhTERT cells was not due to low levels of telomerase activity, as the same hTERT vector induced very effective telomere lengthening in MRC5 lung fibroblasts (48). Furthermore, quantitation by real-time PCR showed that telomerase activity in BMhTERT clones was comparable with a number of tumor cell lines (additional data not shown). One likely explanation for shortening of mean telomere length in BMhTERTs is that oxidative stress overrode the lengthening capacity of telomerase (49). Mounting evidence indicates that the length of a subset of the shortest telomeres, rather than the average length of the telomeres, predicts proliferative arrest (50). Previous studies also suggest that telomerase may preferentially maintain the shortest telomeres (37, 38). The possibility that the subsenescent average telomere length of immortal BMhTERTs reflects selective maintenance of short telomeres is consistent with the eventual stabilization of average telomere length at ∼3 kbp. Skewed densitometric TRF profiles of late-passage cells, which showed accumulation of shorter telomeres (data not shown), were also consistent with preferential maintenance of critically short telomeres. However, the telomere associations and consequential chromosomal rearrangements observed in the BMhTERT cells indicate that there was inadequate maintenance of at least a subset of telomeres at some stage of the immortalization process.

Taken together, the data from these investigations show that telomerase does not prevent telomere dysfunction during hTERT-mediated immortalization of BMECs. Moreover, these results show a specific association between p16^INK4a inactivation and the evolution of karyotypic complexity in TP53-competent cells. These results suggest that p16^INK4a may function as a gatekeeper to prevent telomere-driven cytogenetic evolution and provide a valuable model of the early stages of malignant transformation.

Grant support: Cancer Institute of New South Wales (V.W. Wen and K.L. MacKenzie), New South Wales Cancer Council (V.W. Wen and K.L. MacKenzie), and National Health and Medical Research grant 293810 (R.A. Hinshelwood and S.J. Clark).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dale Wright (Sydney Genetics, Sydney IVF) for extensive consultation on cytogenetics and for reading this article.