A Chromosome 3-encoded Repressor of the Human Telomerase Reverse Transcriptase (hTERT) Gene Controls the State of hTERT Chromatin¹

One of the hallmarks of human cancer is the expression of telomerase, which is absent from the majority of normal somatic human tissues. At least 85% of human tumors have active telomerase (1), with only a small fraction relying on alternative telomerase-independent, recombination-based mechanisms for telomere maintenance (2). Human telomerase is composed of a RNA moiety, hTERC (3, 4) and a catalytic protein subunit, hTERT⁵ (5, 6). Telomerase activity correlates with the presence of hTERT mRNA, whereas hTERC and other components are present in both telomerase-positive and -negative cells. Ectopic expression of hTERT is sufficient to confer telomerase activity in several types of normal cells (7, 8, 9, 10). hTERT transcripts are detectable in the nuclei of telomerase-positive cells but not in telomerase-negative cells (11), suggesting that hTERT expression is regulated via changes in the rate of transcription. A number of trans-acting factors have been implicated as regulators of hTERT expression (reviewed in Ref. 11), but the mechanisms responsible for its expression in tumor cells are not understood.

For most tumors, it is not clear whether hTERT expression is caused by reactivation of the gene from a repressed state or attributable to selection and expansion of stem cells (or dividing progenitors) in which the hTERT gene remains active (11). hTERT expression behaves as a recessive trait and can be extinguished by microcell-mediated transfer of single chromosomes from normal cells (reviewed in Refs. 11 and 12). This finding has led to attempts to clone the genes encoding telomerase repressors. Introduction of a normal chromosome 3 has been shown to down-regulate hTERT expression in two renal carcinoma cell lines, as well as in a breast carcinoma and a cervical carcinoma line (13, 14, 15, 16). Using a renal carcinoma- or a breast cancer-derived line, two groups have mapped the region conferring repression to overlapping subchromosomal segments on 3p (16, 17). However, the putative repressor gene has not yet been identified, and the mechanisms by which it regulates hTERT expression are not yet understood.

Many cis-acting regulatory elements have been identified in the hTERT 5′ noncoding sequence by the use of reporter constructs (18, 19, 20, 21, 22, 23, 24). However, such constructs have not always mimicked the recessive expression of the endogenous hTERT gene and, in our previous study (24), did not respond to the chromosome 3-linked repressor.

We have used an alternative approach, based on limited nuclease digestion, to search for differences between the in vivo conformation of hTERT chromatin in hTERT-expressing and -nonexpressing cells (25). Many regulatory elements map near or within a region that alters its nuclease sensitivity concomitantly with changes in the expression of the associated gene (reviewed in Ref. 26). Here, we report the identification of segments in the second intron of the hTERT gene that are nuclease sensitive in telomerase-positive but not in telomerase-negative cells. The significance of this difference has been validated using the breast carcinoma cell line 21NT, in which transfer of a normal chromosome 3 is known to down-regulate hTERT expression. Our results indicate that chromosome 3-mediated repression “closes” the hTERT chromatin structure. Furthermore, we show that reverse chromatin remodeling to the “open” configuration occurs as a consequence of segregation of the normal chromosome 3. This event appears necessary but not sufficient to allow re-expression of silenced hTERT.

The origin of the cells used here and their culture regimen was described previously by Ducrest et al. (24). The breast carcinoma cell line 21NT and its derivatives 21NT pCineohTERT (21NT hTERT) and 21NT pCineohTERT HyTkchromosome 3 (Chr3.1 and Chr3.2) were maintained in culture as described previously (13).

The selection was done as described previously (17). Briefly, chromosome 3 hybrid clones (Chr3.1 and Chr3.2) were cultured in the absence of hygromycin for 2 weeks before GCV selection to accumulate random chromosome 3-loss segregants. Replicate cultures of 10⁵ hybrid cells were plated in a medium containing various concentrations of GCV (2, 20, and 50 mm; Cymevene; Roche), without hygromycin. All of the GCV concentrations proved sufficient for selection, and no toxicity was observed in control 21NT cells. Twenty to 100 GCV^R colonies were visible after 2 weeks. Segregants were propagated further in the absence of GCV for analysis.

Cells from GCV^R populations and chromosome-3 hybrids were distributed into 96-well plates at a concentration corresponding to approximately one cell/well. Wells were screened immediately for single cells, and for colonies after 1 week; colonies were transferred to dishes before reaching confluence.

For nuclease sensitivity assays, 3–5 × 10⁷ cells were cultured to 70–90% confluence in T175 dishes. The entire preparation was carried out at 4°C, essentially as described previously (27) with the following modifications. After washing, cells were harvested into 15 ml of lysis solution {(1 × buffer A [15 mM Tris-HCl (pH 7.4), 0.2 mM spermine, 0.2 mM spermidine, 2 mM K-EDTA (pH 7.4), and 800 mM, KCl], 0.5 mm EDTA, 1% TDG, 1 mm DTT, and 0.05% NP40}. After dounce homogenization and centrifugation, the pellet was resuspended in 7.5 ml of lysis solution containing 20% glycerol and was homogenized (four strokes). Nuclei were resuspended in 0.5 × buffer A without EDTA, 5% glycerol, 1% TDG, 1 mm DTT, 5 mm MgCl₂, 1 mm CaCl₂ for MNase digestion, and in the same buffer without DTT for DNaseI digestion.

When MNase was used, nuclei (∼7 × 10⁶ nuclei in 300 μl solution per treatment) were digested with 0, 50, 100, and 200 units of MNaseI (Roche Diagnostics Ltd) for 5 min at 25°C. Samples containing ∼20 μg of purified human genomic DNA were simultaneously treated with 4, 6, and 10 units of MNase. When DNaseI was used, cells were treated in 300 μl of DB with 0, 20, 50, and 100 units of DNaseI (Roche Diagnostics Ltd) at 20°C for 4 min. The reactions were stopped by adding 6 μl of 0.5 m EDTA.

Cells were washed twice with ice-cold PBS (0.1 m KH₂PO₄ and 1.5 m NaCl). Nuclei (after nuclease treatment) or whole cells were suspended in lysis buffer [25 mm Tris (pH 8.0), 10 mm EDTA, 200 mm NaCl, and 0.4% SDS] and digested with 500 μg/ml proteinase K for at least 4 h at 37°C. Samples were extracted twice with phenol-chloroform (1:1). The aqueous phase of the second extraction was digested with 10 μg/ml RNaseA at 37°C for 15 min when isolated from whole cells. After a third phenol-chloroform extraction, the DNA was ethanol-precipitated and redissolved in 10 mm Tris (pH 8.0).

Purified genomic DNA was digested to completion with the appropriate restriction enzymes (about 3–4 units of enzyme/estimated μg of DNA) under conditions specified by the supplier. The DNA was phenol-chloroform extracted, ethanol-precipitated, redissolved in 10 mm Tris (pH 8.0), and electrophoresed overnight through a 1% agarose gel in 0.5× TBE buffer (89 mm Tris-borate, 89 mm boric acid, and 2 mm EDTA) at 50 V. The gel was stained with ethidium bromide and was treated with 0.2 m HCl for depurination, with 0.5 m NaOH/1.5 m NaCl for denaturation, and with 0.5 m Tris/1.5 m NaCl for neutralization. The DNA was capillary-transferred overnight in 20× SSC (3 m NaCl, 0.3 m trisodium-citrate) to a positively charged nylon membrane (Appligene-Oncor, Basel, Switzerland), and cross-linked by UV-treatment (UV-Stratalinker 2400, Stratagene). [α-³²P]dATP-labeled hTERT probes were prepared by random priming with Klenow polymerase (Pharmacia Biotech, Dübendorf, Switzerland) using PCR-amplified DNA fragments (shown in Fig. 1) as templates. Membranes were hybridized according to Church and Gilbert (28) overnight at 65°C. Blots were washed three times in 0.1× SSC/0.1%SDS for 15 min at 65°C or at 42°C (depending on the probe), dried, and scanned with a Fuji BAS 1500 scanner (Raytest, CH-8902 Urdorf, Switzerland). The signal intensity distribution in each lane was determined with Aida software (Raytest, Urdorf, Switzerland), and the resulting data were transformed into histograms using Excel (Microsoft).

GCV^R cells were screened for loss of the extra copy of chromosome 3 with polymorphic markers D3S3717, D3S3719, D3S3722, and D3S3697, as described by Cuthbert et al. (17) except that PCR products were labeled with [γ-³²P]ATP.

RNA samples were treated with DNaseI before RT. Endogenous hTERT transcripts were measured as described previously (24). To detect CRR9 and Slc9a3 mRNA the RT reaction was carried out using 100 ng of random hexamers. The following PCR primers were used: CRR9, CRR-615F, 5′-TGCCCATCCTGTTCATCGA-3′, and CRR-679R, 5′-GGAGCGGTTTATGACCATCAG-3′; Slc9a3, SLC-1965F, 5′-TCTTCCACAGGACCATGCG-3′, and SLC-2031R, 5′-TTCTGGTTGAGCCCCAGCT-3′.

The most commonly used approach for investigating chromatin structure in vivo involves limited digestion with DNaseI. When we compared the chromatin of the hTERT gene in cells with (HeLa, HT1080, GM693, 21NT) and without (GM847, HLF) detectable hTERT mRNA, we found that a segment of the hTERT gene, encompassing several kilobase pairs at the beginning of the second intron, was more accessible to digestion by the enzyme in nuclei of hTERT-expressing cells compared with nuclei from telomerase-negative fibroblasts (Fig. 1 and data not shown). The differences in DNaseI sensitivity were quite weak, which made precise mapping of the borders of the affected region problematic. However, when we probed with MNase, a segment containing 10 kb of the 5′ flanking region and 16 kb of the transcribed region (Fig. 1) in tumor cells that contained hTERT transcripts, we observed clearly elevated MNase sensitivity of two discrete sites in the second intron of hTERT (referred to hereafter as HS1 and HS2; Fig. 2). In marked contrast, when normal human fibroblasts were analyzed in this way, the MNase sensitivity of this region resembled that of naked DNA. Surprisingly, we detected no differences in the pattern of MNase-sensitive sites in the 5′ flanking region. Increased sensitivity at HS2 was also detectable in hTERT-expressing SV40-transformed fibroblasts (GM639). These cells displayed an additional MNase-sensitive site about 100 nucleotides downstream from HS1 (see Fig. 2,B). In contrast, hTERT-negative SV40-transformed fibroblasts (GM847) that use an alternative (ALT) pathway of telomere maintenance (29) displayed a MNase sensitivity pattern similar to that of normal fibroblasts. We confirmed the reproducibility of these patterns by repeating the analyses of MNase sensitivity at least once for each cell line and by digesting the extracted DNA with different restriction enzymes (EcoRI, SphI, SphI-BclI; see Fig. 1). HS1 and HS2 mapped within the first 1 kb of the second hTERT intron and were found to be separated by ∼400 bp. The extent of MNase sensitivity varied among the hTERT-expressing cells (Fig. 2 B). Increased sensitivity at HS1 was observed in a colon carcinoma (HCT116) and a cervical carcinoma (HeLa) line but was less evident in a fibrosarcoma (HT1080) line. In all hTERT-positive cells, HS2 was more sensitive to MNase than it was in hTERT-negative fibroblasts. Among the cell lines investigated, HS2 sensitivity was highest in HeLa cells, and it was of interest that, of the lines studied here, these cells contained the highest amount of hTERT RNA (24).

To investigate whether there exists a causal relationship between the presence of hTERT transcripts and the increased nuclease sensitivity of specific sites in the second intron of the hTERT gene, we used the breast carcinoma cell line 21NT. As indicated above, transfer of a normal chromosome 3 into these cells has been shown to abolish telomerase activity and induce replicative senescence (13, 24). However, senescence can be avoided through ectopic expression of hTERT cDNA, and Ducrest et al. (24) have shown that the transfer of normal chromosome 3 into these modified 21NT hTERT cells reduces the levels of endogenous (intron-containing) hTERT transcripts by at least 30-fold. We compared the MNase sensitivity of hTERT chromatin in 21NT cells, in 21NT hTERT cells, and in two chromosome 3-containing microcell hybrids (Chr3.1 and Chr3.2; Fig. 3 A). 21NT and 21NT hTERT were found to display identical patterns of MNase sensitivity, which were similar to those obtained with other telomerase-positive lines. In contrast, in the two chromosome 3-hybrids, the chromatin containing the second intron of the hTERT gene resembled that of normal, telomerase-negative fibroblasts. Whether the repressor that is encoded by sequences on chromosome 3 silences hTERT expression directly, e.g., by binding to regulatory elements in the hTERT gene, or whether it alters hTERT chromatin indirectly, is unknown.

There is good evidence implicating proteins of the c-Myc/Mad transcription factor network in hTERT regulation. For example, overexpression of c-Myc can directly induce the hTERT gene (30, 31), whereas its antagonist, Mad, down-regulates hTERT expression (32, 33). Published reports regarding the role of c-Myc and Mad in hTERT regulation implicate two conserved E-boxes upstream from the translation start site (30) but do not address the role of two E-boxes ∼50 and 380 nucleotides upstream from HS2 (see Fig. 1). To investigate whether c-Myc-activated hTERT transcription involves changes in the chromatin confirmation at HS2, we compared the pattern of MNase sensitivity of normal fibroblasts with that of cells transduced with c-Myc (which express hTERT at levels comparable with those of some telomerase-positive tumor cells; Ref, 24). We observed that the hTERTchromatin of the c-Myc-transduced cells was no more sensitive to MNase than that of normal fibroblasts (Fig. 3 B). This result is consistent with the finding by Ducrest et al. (24) that showed that chromosome 3-mediated hTERT down-regulation in 21NT cells does not involve the c-Myc regulatory network.

To obtain additional evidence for a causal relationship between increased MNase sensitivity and hTERT expression, we attempted to select segregants of 21NT chromosome 3-hybrids that had lost the introduced copy of normal chromosome 3. The chromosome used in the transfer study had been tagged with a marker gene that is a fusion between the hygromycin-resistance and the HSV thymidine kinase (HSV-tk) genes. Expression of HSV-tk makes cells sensitive to GCV, and segregation of the normal chromosome 3 was expected to be the most likely event giving rise to GCV^R variants. To obtain such segregants, we cultured the two chromosome 3-hybrids in GCV, in the absence of hygromycin. We isolated five independently derived, clonal GCV^R cell populations and genotyped them for informative polymorphic markers (D3S3717, D3S3719, D3S3722, D3S3697) that map in the interval containing the putative hTERT repressor gene (see Fig. 4,A and Ref. 13). As expected, both of the initial chromosome 3-hybrids contained all of the alleles present on the normal chromosome 3. Surprisingly, most of the GCV^R cell populations had not lost all of the (normal) chromosome 3 markers (Fig. 4 B). The most likely explanation for this is that the normal chromosome 3 in the GCV^R cells had undergone fragmentation, attributable perhaps to unstable regions or fragile sites in or near the region that contains the hTERT repressor (34). Therefore, to isolate segregants that had lost all of the alleles specific for the extra copy of chromosome 3, we subcloned two GCV^R populations (referred to as populations B and C, derived from Chr3.1 and Chr3.2 clones, respectively). For control purposes, the two original chromosome 3-hybrids were subcloned simultaneously in media containing hygromycin but no GCV; none of 16 subclones had lost any of the normal chromosome 3 alleles. In contrast, 14 GCV^R clones no longer retained any of the markers mapping to the chromosome 3 region containing the repressor gene. However, four of these clones still retained the D3S3697 marker, which is centromeric to the critical region.

We investigated the state of the hTERTchromatin in the GCV^R cells using MNase sensitivity assays. Four GCV^R clones were analyzed; three of them had lost all of the normal chromosome 3 markers, whereas one had retained the D3S3697 allele. In all four of the GCV^R clones, we observed an MNase sensitivity pattern identical to that of the parental 21NT cells (Fig. 4,C). In the control chromosome 3-containing subclones, the chromatin pattern remained in the telomerase negative-type configuration (Fig. 4,D). The question of whether chromatin remodeling in the chromosome 3-loss segregants correlated with up-regulation of hTERT expression was addressed by quantifying immature endogenous hTERT RNA with real-time RT-PCR (Fig. 5). The level of endogenous hTERT expression in 21NT is very low and gives signals that are close to background (24). Therefore, to obtain reliable RT-PCR results, we repeated the PCR at least five times. GAPDH expression was constant among the samples (not shown). The results (Fig. 5) show a clear difference between the hTERT transcript levels of the parental 21NT hTERT cells and their chromosome 3-containing hybrid subclones. The parental cells contained at least 10-fold more hTERT transcripts per cell than did the chromosome 3-hybrids. Surprisingly, only a proportion of GCV^R segregants that had lost the region of the normal chromosome 3 containing the hTERT repressor gene re-expressed the hTERT gene. This was the case in the C-type subclones, derived from chromosome 3-hybrid 2 (Fig. 5). In most of the B-type clones derived from chromosome 3-hybrid 1, the hTERT gene remained repressed. There was a slight shift toward re-expression in clone 1.B.9. The presence of the D3S3897 marker in six clones was not associated with a significant difference in hTERT expression.

The hTERT gene is close to the telomere on the short arm of chromosome 5 (Fig. 6,A), whereas the mouse gene, mTERT, maps much closer to the centromere. This raises the question as to whether TERT repression in human somatic cells, not observed in the mouse, is caused by a telomere position effect (TPE), which has recently been shown to exist in human cells (35). If this were the case, tumor-specific hTERT expression might be attributable to the alleviation of the TPE. To address this question, we compared expression of two genes close to the 5p telomere in normal and tumor cells by real-time RT-PCR. According to the June 2002 update of the UCSC Human Genome Project Working Draft⁶ CRR9(cisplatin resistance-related protein), and Slc9a3, (solute carrier family 9 protein isoform-3) map, respectively, 0.1 Mb and 1.1 Mb on the telomeric side of hTERT(Ref. 36; Fig. 6,A). Both of the genes are transcribed from the same DNA strand as hTERT. The PCR primers for both genes were designed to avoid the amplification of homologues or consensus motifs. There were no significant differences in the expression levels of these genes among normal fibroblasts (HLFs, Myc-transduced fibroblasts, HeLa, 21NT, 21NT hTERT, two chromosome 3-containing subclones (1.1, 2.4), three hTERT-low GCV^R segregants (1.B.6, 1.B.7, 1.B.9), and two hTERT-high GCV^R segregants (2.C.7, 2.C.11; see Figs. 5 and 6 B).

To identify the critical mechanisms controlling hTERT expression in normal human cells and in cancer, we have studied the nuclease sensitivity of the hTERTlocus. DNaseI treatment of chromatin from telomerase-positive human cells (derived from tumors and virally immortalized cell lines) revealed a broad, diffuse region of increased sensitivity in the second intron of the hTERT gene that was not present in the chromatin of telomerase-negative fibroblasts. Moreover, using MNase, we obtained a clearly defined pattern that permitted reliable detection of differences between the chromatin of telomerase-positive and -negative cells. Despite strong indications for the involvement of specific elements in the 5′ flanking region in hTERTregulation (37, 38) we did not observe any differences between the chromatin of hTERT-positive and -negative cells in 10 kb of 5′ flanking region upstream from the translation start site.

The second intron of hTERT-expressing cells contained chromatin-dependent cleavage sites that were much more pronounced in cells containing hTERT transcripts. Intronic nuclease-sensitive sites that correlate with gene expression have been reported in other loci (39, 40, 41, 42). The recent finding that the transfer of a normal chromosome 3 into 21NT cells represses hTERT expression and, in the present study, that this restores hTERTchromatin to the state characteristic of normal (telomerase-negative) cells suggests that chromatin opening, reflected by increased accessibility to nucleases, is required for hTERTtranscription. That forced loss of the normal chromosome 3 by reverse selection leads to the reappearance of the increased MNase sensitivity, but not inevitably to restoration of hTERT expression, would appear to indicate that chromatin opening is not sufficient to induce hTERTtranscription. There are other reported instances in which chromatin opening does not by itself result in transcriptional activation. For example, c-Myc induces chromatin remodeling of target genes that is necessary, but not always sufficient, to elicit a full transcriptional response (43). Furthermore, formation of DNaseI HSs at the human β-globin locus alone is not sufficient for high-level globin gene transcription (44).

Our finding that overexpression of c-Myc, sufficient to induce hTERT expression, was not accompanied by the appearance of the MNase-sensitive sites in the hTERT gene provides evidence against the specific hypothesis that c-Myc binding to the E-boxes in the second intron of the gene leads directly to chromatin remodeling. The absence, in c-Myc-transduced fibroblasts, of the sites characteristic of telomerase-positive tumor cells is consistent with the indications that, in 21NT cells, de-repression of the hTERT gene is not mediated by c-Myc overexpression (24).

When telomerase-expressing segregants of chromosome 3-microcell hybrids, derived from renal and breast carcinoma cells, were subjected to fine-structure microsatellite deletion mapping, the repression of hTERT expression was shown to cosegregate with the interval 3p12–21.1 (13, 15). According to recent mapping data using 21NT hybrids,⁷ the hTERT repressor sequences map to a 350-kb interval at 3p21.2. The results of the study reported here suggest that the repressor imposes a closing of hTERTchromatin, but that re-expression of hTERT requires additional changes apart from the loss of the repressor gene. Segregant mapping and chromosome fragment transfer have not implicated additional regions of chromosome 3 in hTERT re-expression.⁷ However, a role for additional epigenetic events (such as random changes in DNA-methylation of genes) in the reactivation of hTERT during human cell immortalization remains a possibility.

The report that human genes may be subject to TPEs (35) has suggested that expression of telomere-proximal genes, such as hTERT, could be controlled by telomere length (reviewed in Ref. 45). Our results show that two genes near the 5p telomere, one (CRR9) very close to hTERT, are not repressed in normal, hTERT-negative cells or after chromosome 3-transfer into the 21NT cell line. Thus, our data provide no evidence for a role of TPE in the control of hTERT expression. Rather, they indicate that repression in normal cells and in experimentally induced chromosome 3-mediated silencing of the hTERTgene, is specific and does not extend to neighboring loci.

Our observations do not define the specific role of increased nuclease sensitivity in the regulation of hTERT expression. Currently, the simplest hypothesis is that remodeling of intron 2 chromatin is the consequence of the binding of a transcription factor, required for hTERT expression in tumor cells, to a regulatory element within or near the gene. In this model, binding would be prevented, directly or indirectly, by the repressor (encoded by a gene in a 350-kb segment at 3p21.2). The molecular cloning of this gene, and its functional characterization in normal cells and cancer cells, may provide more insight into the relationship of the tumor-specific chromatin alteration in intron 2 to hTERT transcription and the importance of the silencing of the repressor gene in human cancer development.

We are grateful to Anne-Lyse Ducrest; ISREC for HLF-Myc cells, her help with cell culture of 21NT and its derivatives, protocols and material for the quantitative RT-PCR, and for discussions. We thank Roger Reddel, Children’s Medical Research Institute, Westmead, Australia, for GM847 and GM693 cells, and Corinne Rusterholz, Maria Chiara Bassi, and Patricia Corthésy for their help in the nuclease sensitivity assays.