Cellular and Gene Expression Responses Involved in the Rapid Growth Inhibition of Human Cancer Cells by RNA Interference-mediated Depletion of Telomerase RNA*

Inhibition of the up-regulated telomerase activity in cancer cells has previously been shown to slow cell growth but only after prior telomere shortening. Previously, we have reported that, unexpectedly, a hairpin short interfering RNA specifically targeting human telomerase RNA rapidly inhibits the growth of human cancer cells independently of p53 or telomere length and without bulk telomere shortening (Li, S., Rosenberg, J. E., Donjacour, A. A., Botchkina, I. L., Hom, Y. K., Cunha, G. R., and Blackburn, E. H. (2004) Cancer Res. 64, 4833–4840). Here we have demonstrated that such telomerase RNA knockdown in cancer cells does not cause telomere uncapping but rather induces changes in the global gene expression profile indicative of a novel response pathway, which includes suppression of specific genes implicated in angiogenesis and metastasis, and is distinct from the expression profile changes induced by telomere-uncapping mutant template telomerase RNAs. These cellular responses to depleting telomerase in human cancer cells together suggest that cancer cells are “telomerase-addicted” and uncover functions of telomerase in tumor growth and progression in addition to telomere maintenance.

Telomeres are specialized complexes at chromosome tips that protect the chromosomal ends and preserve the stability of the genome (1). Telomeres are essential not only for chromosome stability but also for complete DNA replication (2). The telomeric DNA repeat tracts are maintained by the telomerase ribonucleoprotein (RNP) 1 complex, in which the integral RNA component contains a short sequence that acts as the template for synthesis of telomeric DNA repeats (3,4).
Lack of telomerase in yeast is sufficient to cause a low but measurable frequency of catastrophic telomere shortening, even when the bulk telomere population is still long (5). Only after a delay of tens of generations does the shortening of the bulk of the telomere population eventually trigger a stereotyp-ical cellular response in the majority of the cell population. This response includes a DNA damage response (6 -8). Genetic inactivation of telomerase in mice causes age-and generationdependent shortening of their normally very long telomeres accompanied by genetic instability. These phenotypes are only apparent after a significant delay, during which bulk telomere shortening occurs (9 -13). In cultured human primary cells with low endogenous telomerase, telomeric DNA can be lost from the ends of linear chromosomes at rates of up to 150 -200 bp/cell division (14,15), and eventually cellular senescence ensues, unless the chromosome termini can be replenished (16,17).
In humans, the core essential components necessary for the reconstitution of telomerase activity in vitro are the catalytic protein subunit hTert and the RNA subunit hTER (also called hTR) (18,19). Previous studies have shown that hTER is widely expressed in many normal cell types. The long term importance of the full gene dosage of telomerase RNA is demonstrated by human genetics studies showing that, if one telomerase RNA gene copy is non-functional, this is sufficient to cause an autosomal dominant form of non-X-linked dyskeratosis congenita. Patients with this inherited disease die in early adulthood to middle age primarily because of progressive bone marrow failure. In at least one such family pedigree, the cause is apparently a simple haploinsufficiency for telomerase RNA (13,20).
Telomerase activity, although detectable, is diminished in many adult somatic cells, except stem cells and germ cells (18,19,(21)(22)(23). In contrast, cancer cells commonly up-regulate telomerase; in 85-90% of human cancers and over 70% of immortalized human cell lines (21,24), telomerase activity is highly elevated by up to 100-fold over the normal counterpart cells (25). These observations are consistent with telomerase conferring a strong selective advantage for continued growth of malignant cells (26). However, in yeasts, a well characterized telomerase-independent Rad52-mediated DNA recombination mechanism can also maintain telomeres in the absence of telomerase (27)(28)(29)(30), and a small percentage of tumors and immortalized human cell lines can utilize an apparently similar mechanism known as "alternative lengthening of telomeres" (31). Although cells without telomerase can be tumorigenic in mouse xenografts, they have weak metastatic potential (32). These observations, together with the frequency and strength of telomerase up-regulation in human cancers, have suggested that telomere maintenance is essential for cancer cell immortalization and, further, that it may be possible to inhibit cancer growth by interfering with telomerase action.
Attempts to attenuate telomerase function in cultured human cancer cells have been reported to lead to the expected telomere shortening (33). However, suppression of cell growth by telomerase inhibition or depletion had previously been predicted to require a long lag period, during which time their telomeres would be progressively shortening. Indeed, when cancer cells were selected to stably overexpress a catalytically inactive form of human telomerase reverse transcriptase (hTERT) protein, which overwhelms the resident wild-type hTERT and, hence, limits its assembly into active telomerase ribonucleoprotein, there was the expected delay before cell growth inhibition ensued (34,35), and the length of the delay was dependent on the initial telomere length (35). Similarly, a delay in slowing of cell growth was also seen when telomerase enzymatic activity was inhibited by an antisense oligonucleotide that binds to the telomerase RNA-templating region (36).
We have recently reported that a non-messenger RNA, the wild-type human telomerase RNA, hTER, can be efficiently and specifically targeted by a short interfering hairpin RNA construct (hairpin siRNA) (37) expressed from a lentivector. In contrast to the delay previously predicted for any effects of telomerase depletion, we noticed that, unexpectedly, such siRNA expression was sufficient to cause rapid and specific slowing of cancer cell proliferation. Strikingly, these effects occurred without bulk telomere shortening and were independent of wild-type p53 and telomere length. The rapidity of the effect of telomerase depletion caused by this siRNA was consistent with a previous report (38) in which depleting the telomerase RNP by an independent method (knockdown of hTERT with an antisense oligonucleotide targeting the mRNA for hTERT) caused rapid apoptosis of DU145 prostate cancer cells.
Here we have investigated key features of the surprisingly rapid effects of telomerase RNA knockdown in human cancer cells. First, we showed that the anti-telomerase RNA siRNA treatment does not cause any telomere uncapping or DNA damage response, in contrast to the effects of expressing mutant template telomerase RNAs (MT-hTer), which are predicted to cause mutant repeats to be added to telomeres (37) and which we showed directly here cause telomere uncapping. The lack of telomere uncapping by the siRNA treatment of cancer cells also contrasted with the telomere uncapping previously reported for primary human cells entering cellular senescence (39). Second, the telomerase RNA knockdown induced changes in the global gene expression profile indicative of a novel response pathway. This pathway is distinct from that induced by the expression of an MT-hTer that caused telomere uncapping. We observed and confirmed rapid down-regulation of expression of genes involved in cell cycle progression, including Cyclin G2 and Cdc27. These results uncover a rapid novel response of human cancer cells to depletion of the level of the telomerase ribonucleoprotein. Furthermore, the anti-telomerase RNA siRNA also rapidly lowered the expression of specific genes previously reported to play important roles in tumor growth, angiogenesis, and metastasis, such as integrin ␣V and Met oncogene protein. We have recently reported that metastasis of melanoma cells in an in vivo mouse model is specifically inhibited by knockdown of telomerase RNA using a ribozyme targeting telomerase RNA (40). Together with that finding, these results provide new evidence that telomerase affects gene expression to confer additional pivotal functions in tumor progression other then telomere length maintenance.

EXPERIMENTAL PROCEDURES
Plasmid Construction-The three-plasmid-based lentivector system was generously provided by Dr. Didier Trono (41). The construction of lentivectors that express MT-hTers and anti-hTER siRNA were described in a previous publication (37). The hTert expression vector pBabe-puro-hTert was engineered as described previously (42).
Cell Culture-The isogenic p53-WT and p53-null HCT116 cell lines were kindly provided by Dr. Bert Vogelstein, with the wild-type p53 gene being completely disrupted by homologue recombination (43). The LOX melanoma cells were maintained in RPMI 1640 medium supplemented with 10% fetal bovine serum. The primary fibroblast WI38 cells were obtained from the American Type Culture Collection (Manassas, VA). The lentivirus production and infection procedure were described previously (37). For cell growth measurements, ϳ5 ϫ 10 5 cells were infected with lentivirus as indicated at 10 -20 transduction units/cell to achieve Ͼ95% infection efficiency. Cells were either unselected or selected with 1 g/ml puromycin for 24 h before cell counting. ϳ2 ϫ 10 4 cells were reseeded in six-well plates at 48 h after virus infection, and the cell numbers were counted every day or every other day for 7 days (for the LOX cells) and 14 days (for the HCT116 cells). For cell growth measurement in HeLa cells, the cells were transiently transfected with various synthetic siRNAs at 10 nM using Lipofectamine 2000 transfection reagent. Twenty-four hours after transfection, the cells were trypsinized, 2 ϫ 10 4 cells were reseeded in six-well plates, and the cell numbers were counted every other day for 7 days.
Telomerase Assays and Telomere and RNA Analyses-Telomerase activity from cell extracts was analyzed using the PCR-based telomeric repeat amplification protocol assay. Telomere length was measured by hybridization of a 32 P-labeled (CCCTAA) 4 probe to genomic DNA digested with HinfI and RsaI. RNA was extracted using TRIzol reagent (Invitrogen). For Northern blotting analysis, the RNA samples were separated in 1.5% agarose gel and transferred to Hybond-N ϩ membrane. The blot was hybridized with 32 P-labeled cDNA probes, as indicated, by random primer labeling (Amersham Biosciences). Ribonuclease protection assays were performed with 32 P-labeled hTER, 2Ј5Ј-oligoadenylate synthetase 1, glyceraldehydes-3-phosphate dehydrogenase, Cyclin G2, and integrin ␣V riboprobe as described previously (Ambion Inc., Austin, TX).
Histochemistry-Combinational immunostaining and telomere fluorescence in situ hybridization was carried out as previously described (44) using monoclonal anti-p53BP1 antibody (BD Biosciences) and fluorescein isothiocyanate-labeled peptide nucleic acid telomere repeat probe (Applied Biosystems, Foster City, CA).
DNA Microarray-Microarray analysis was performed as described previously (45). In brief, ϳ1 g of total RNA was linearly amplified through one round of modified in vitro transcription and coupled to N-hydroxysuccinimidyl esters of Cy3 or Cy5 (Amersham Biosciences). The 42,680 cDNAs used in these studies were from the Unigene project (Research Genetics, Huntsville, AL). On the basis of the Unigene data base build155, these clones represent 29,778 independent genes. Primary data were analyzed using GENEPIX version 3.0 software (Axon Instrument, Union City, CA). Gene expression was linearly normalized in NOMAD (derisilab.ucsf.edu) and analyzed with Cluster (46), filtering the dataset for genes changed by at least 2-fold using the average linkage metric and visualized with Java Treeview. Genes were grouped based on their expression pattern. The GENEPIX ratio of median value for each gene was log-transformed for the statistical analysis for microarrays using custom macros that were the kind gift of Anthony Dobson, Univerity of California, San Francisco. In Fig. 4B, each column represents data from one analysis; duplicate analyses for each condition are shown in pairs across the panel. Each row represents the expression pattern of a single gene at all time points analyzed. Gene expression ratios shown in red are up-regulated, and those in green are downregulated. Gray areas indicate missing data.

RESULTS
Anti-telomerase RNA siRNA Rapidly Causes Growth Inhibition with No Bulk Telomere Shortening-As previously reported (37), for efficient delivery and expression of the hairpin siRNA, we utilized a safety-engineered, human immunodeficiency virus-1-based lentiviral system (41). The hairpin siRNA was designed to target specifically the sequence encompassing the hTER 11-nucleotide template region, as shown in supplemental Fig. S1. By analysis of the human genome and cDNA databases, we found no other exact matches to the 19-base telomerase RNA-matching siRNA sequence used. To confirm the desired target specificity of the anti-wild-type hTER siRNA, we used a human cell line, VA13, which normally lacks any endogenous hTER and hTert expression (47). VA13 cells were first infected with lentivirus expressing either wild-type hTER (WT-hTER) or two different telomerase RNAs bearing mutations engineered into the template region (MT-hTers). Two days later, we superinfected with lentivirus expressing the siRNA targeting wild-type telomerase RNA. Expression of this siRNA effectively and selectively knocked down the level of WT-hTER, with no significant effect on the levels of the mutant template hTers, which only differ from WT-hTER by mutations in the templating region (supplemental Fig. S1B, left); one mutant template telomerase RNA sequence differs from the wild-type by only two bases. This specific reduction of WT-hTER was not due to differential expression of siRNA in cells expressing WT-hTER versus MT-hTer, as Northern hybridization detected comparable amounts of siRNA in all of these cells (supplemental Fig. S1B, right). This specificity of the siRNA for the wild-type template telomerase RNA seen in VA13 cells confirmed, in a different cell line, the results reported previously in endogenously as well as ectopically expressed wild-type template and mutant template telomerase RNAs in HCT116 human colon cancer cells (37). We conclude that the hairpin siRNA is specific for the wild-type telomerase RNA subunit.
As an independent control, an siRNA mutant (siRNA-mut), engineered to target a sequence unrelated to hTER, was used to confirm whether it is the down-regulation of telomerase activity by hTER targeting that inhibits cell growth specifically. Empty lentiviral vector was infected in parallel to control for any toxicity caused by the lentiviral infection itself. The lentivirus-expressing siRNA targeting endogenous hTER was used to infect a pair of human colon cancer lines, derived from line HCT116, which either have or lack functional p53 but otherwise are isogenic (43). The high infection efficiency of the lentivirus (Ͼ95%) allowed the rapid evaluation of the cellular effects of expression of the anti-hTER siRNA in the pooled cell population, without the need to delay analysis because of the requirement for selection for a drug resistance gene. This whole-population analysis also eliminated any clone-to-clone variation that might have resulted from analysis of clonal cell cultures. The results showed that any cellular toxicity resulting from lentivirus infection per se was negligible, but the anti-hTER siRNA (not the siRNA-mut) caused rapid cell growth inhibition (Fig. 1, A and B). Notably, as described previously (37), in these isogenic HCT116 cell lines with or without p53, the levels of rapid cell growth inhibition and apoptosis were similar.
We used different cancer cell lines to test whether the cancer cell growth inhibition caused by the hTER-targeting siRNA depends on bulk telomere shortening. The human melanoma cell line LOX has very long (Ͼ40 kb) telomeres. We reasoned that if the bulk of telomeres need to shorten before the siRNA has an effect on cell growth, we would expect a delay in the effect of the siRNA in LOX cells compared with a cell line with short telomeres, such as the HCT116 cells, which have relatively short (ϳ3-4 kb) telomeres, as reported previously (37). LOX cells were infected with the lentiviruses expressing hTERtargeting siRNA and siRNA-mut or the control empty lentivector. Rapid inhibition of cell growth was observed in LOX cells expressing the hTER-targeting siRNA (Fig. 1C). The rapidity of the cell growth inhibition was similar to that in HCT116 cells (Fig. 1, A and B) (37). There was no growth inhibition in LOX cells infected with the control lentivirus or with the lentivirusexpressing siRNA-mut (Fig. 1C). A second control, unrelated siRNA in the same lentiviral vector tested in the LOX cells, also failed to show any effect on cell growth. 2 Similar expression of both the hTER-targeting siRNA and siRNA-mut in the LOX cells was confirmed by Northern blotting (Fig. 1D, bottom  panel). The knockdown of endogenous hTER was confirmed by Northern blotting analysis (Fig. 1D, top panel). A decrease in telomerase activity was shown using in vitro telomeric repeat amplification protocol assays (Fig. 1E). Finally, consistent with the results reported for the HCT116 cells expressing the siRNA (37), no bulk telomere shortening of the long LOX cell telomere population was seen (Fig. 1F).
We also used Northern blotting to examine the levels of siRNA expression in duplicate cell populations that grew after 14 days post-infection. Those cell populations showed a large overall reduction in anti-hTER siRNA levels in comparison with the cells analyzed 3 days post-infection (data not shown). In contrast, in the control siRNA-mut cells, the siRNA-mut expression showed no reduction after 14 days. For all samples, GAPDH levels were measured in the Northern blots to normalize the RNA loading and detection. This result confirmed that selection occurred against continued growth of cells that expressed the anti-hTER siRNA but not the control siRNA-mut.
These data showed that decreasing the telomerase activity by an siRNA expressed from the viral vector that specifically targets hTER rapidly affects growth of the telomerase-positive cancer LOX and HCT116 cell lines, within just a few cell divisions. Furthermore, the cell growth inhibition occurs without detectable bulk telomere shortening ( Fig. 1F and data not shown) (37). Inhibition of cell growth was also seen with the telomerase-positive human bladder cancer cell line UM-UC-3; in a representative experiment, 17 days after infection with the siRNA-expressing lentivirus, the cell number count was 3.4 ϫ 10 7 cells compared with 5.1 ϫ 10 7 cells for the control lentiviral infected cells and 0.6 ϫ 10 7 cells for cells infected with the MT-hTer mutant AU5. In contrast, no significant decrease in cell proliferation was seen upon expression of the anti-hTER siRNA in the telomerase-negative alternative lengthening of telomere cell line VA13, as expected (data not shown). We conclude that the knockdown of telomerase in cancer cells rapidly inhibits cancer cell growth independently of any requirement to shorten the bulk telomere population.
Telomerase RNA Knockdown and Cell Growth Inhibition Induced by in Vitro Synthesized Double-stranded siRNAs-Recent studies (48,49) have shown that overexpression of 21-mer siRNA in mammalian cells can induce a nonspecific interferon response under certain circumstances. HCT116 cells have a defective interferon response pathway (50), suggesting that it is unlikely the cell growth inhibition induced by anti-hTER siRNA is due to nonspecific interferon response. However, to further confirm the specificity of the rapid cell growth inhibition by anti-hTER siRNA, we tested two sets of in vitro synthesized double-stranded siRNAs in HeLa cells. The first set of synthetic siRNAs were unmodified double-stranded RNA with a 2-nucleotide 3Ј overhang, as shown in Fig. 2A. The anti-hTER siRNA-2 targets the same templating region of telomerase RNA as the siRNA expressed by our lentiviral vector, whereas the anti-hTER-siRNA-3 targets the telomerase RNA pseudoknot domain P3 region (51). Dramatic knockdown of telomerase RNA expression was observed at 24 h after tran-sient transfection of synthetic siRNA into HeLa cells. Rapid cell growth inhibition occurred in these HeLa cells transiently transfected with anti-hTER siRNA-2 and anti-hTER siRNA-3 but not with the control siRNA in our 7-day cell growth measurement (Fig. 2C). Telomerase RNA knockdown was sustained up to 7 days by single transient transfection, as shown in supplemental Fig. S2A. A separate set of synthetic siRNAs (stealth RNAi) were also engineered to target the same regions in human telomerase RNA, except that they are chemically modified double-stranded 25-mer RNA oligonucleotide duplexes, without a single-stranded overhang. Such modification reportedly reduces nonspecific, off-target effects (Invitrogen). Similar to the unmodified siRNA, dramatic knockdown of telomerase RNA expression was observed at 24 h post-transfection of the chemically modified siRNA into HeLa cells (Fig. 2B). Again, rapid cell growth inhibition resulted (Fig. 2D). Transient transfection of synthetic siRNA into HeLa cells using our experimental condition did not induce significant change of the expression level of 2Ј5Ј-oligoadenylate synthetase 1 (Fig. S2, B and C), which has been shown to be up-regulated by the nonspecific interferon response. These data further confirmed the conclusion that the rapid cell growth inhibition induced by anti-hTER siRNA is specific and dependent on the down-regulation of telomerase activity.
Lack of a DNA Damage Response or Telomere Uncapping upon Expression of Anti-wild-type hTER siRNA-We have reported previously that the short term cancer cell growth inhibition caused by anti-hTER siRNA expression is accompanied by a rapid increase in the ratio of cells with G 2 DNA content, compared with G 1 content, and a small but significant increase in apoptosis; fluorescence-activated cell sorter analyses showed that 1-3% of the anti-hTER siRNA-expressing cells had sub-G 1 DNA content compared with 0.5% of control cells (37). To dissect the mechanism of this unexpected and rapid cellular response to perturbing the level of telomerase, first we tested whether telomere uncapping occurs. It has previously been found (5) that, in yeast, depletion of telomerase causes a low fraction of even long telomeres to undergo catastrophic telomere shortening events. Such shortened telomeres rapidly fuse, indicating that they become uncapped (5). Immunohistochemistry was performed on LOX cells using an anti-p53BP1 antibody. Previous studies (52) have shown that p53BP1 is involved in the early response to cellular DNA damage. Distinct p53BP1 foci, presumably at sites of DNA damage, can be observed in cells challenged with DNA-damaging agents (52). Telomere uncapping caused by a different type of telomere perturbation is also detectable as DNA damage foci at telomeres seen by immunofluorescence (53). Such foci forming at telomeres are also seen in long term cultured human fibroblasts once they enter senescence (54). However, there was no increase in foci above the background in LOX cells infected with the lentivirus expressing hTER-targeting siRNA, or WT-hTER, or the control lentivirus (Fig. 3A). As the positive control for the detection of telomerically located DNA damage foci, we used a method of telomere uncapping in which a mutant template telomerase RNA (MT-hTer) directs the addition of mutant telomeric DNA to telomeric termini (37,42). Foci were readily seen in cells expressing two different MT-hTers (see supplemental Fig. S1) (37) but not in cells expressing the hTER-targeting siRNA (Fig. 3A). We validated that a significant fraction of these foci were at telomeres by co-staining with an in situ hybridization probe, a peptide nucleic acid probe specific for telomeres (55) (Fig. 3B). Thus, in siRNA-treated LOX cells, we found no evidence suggestive of uncapped or shortened telomeres by immunofluorescence. This finding contrasts with the telomere uncapping evidenced by telomeric DNA damage foci that form in fibroblasts as they approach senescence (54). Together, our results showed that the response to siRNA against telomerase RNA does not resemble a senescence phenotype.
The lack of a detectable DNA damage response by the siRNA targeting hTER was further confirmed by the finding that p21 (another DNA damage response gene product) was not upregulated by the siRNA, as shown by Northern blotting anal-  4. Regulation of gene expression by siRNA targeting hTER in HCT116 (p53-WT) cells. A, microarray data were analyzed by significance analysis of microarrays (SAM). Solid line represents genes for which the observed and expected relative differences from the reference are identical and that would therefore be found to be "regulated" by chance. At ⌬ ϭ 0.24 (limited by dashed lines), 73 genes in HCT116 (p53-WT) cells are found to be significantly repressed. B, the log-transformed ratio of median values for these genes (inside solid line) in HCT116 (p53-WT) ysis (Fig. 3C). In contrast, p21 expression was increased in the cells expressing the MT-hTers (Fig. 3C). Together, the results suggest that siRNA directed against the wild-type endogenous telomerase RNA inhibits cell growth by activating a response pathway different from the DNA damage response pathway or a telomere-uncapping response.
siRNA Targeting Wild-type Telomerase RNA Causes a Rapid, Novel Global Gene Expression Profile Change in Cancer Cells-Because the anti-telomerase RNA siRNA effects were not mediated through telomere uncapping, a standard DNA damage response, or through telomere shortening and were not dependent upon p53, we tested whether the depletion of telomerase in cancer cells by the siRNA causes the observed rapid effects on cell growth by influencing gene expression. A cDNA expression microarray analysis was performed on HCT116 cells at day 5 after infection with either lentivirus that expressed siRNA targeting hTER or with control empty vector lentivirus. For comparison, we expressed the mutant template telomerase RNA-AU5 in these cells, with or without co-expression of the siRNA, which specifically targets only the wild-type hTER, as described previously (37) and above. Expression of the hTER-targeting siRNA caused a specific set of significant changes in the gene expression profile (Fig. 4). These changes differed from those induced by MT-hTer expression alone. These anti-hTER siRNA-induced changes included decreased expression of a set of 73 genes that were specifically down-regulated by the expression of the siRNA (but not of the MT-hTer). Some of these down-regulated genes are involved in cell cycle progression, such as Cyclin G2 and Cdc27. Specific genes that were also down-regulated have been found to play important roles in tumor growth, angiogenesis, and metastasis, such as integrin ␣V and Met proto-oncogene. No significant up-regulation of interferon-induced genes was observed, consistent with a previous report suggesting that HCT116 cells have a defective interferon response (50). No expression changes indicative of a DNA damage response were observed. The downregulation of Cyclin G2 and integrin ␣V in HCT116 cells expressing siRNA targeting hTER was confirmed by ribonuclease protection assays (Fig. 5A).
Cyclin G2 and Cell Growth Rate-We noted that, although the siRNA targeting the wild-type telomerase RNA decreased the expression of Cyclin G2 and integrin ␣V, overexpression of wild-type hTER did not affect the levels of these mRNAs (Fig.  5A). Therefore we tested whether overexpression of telomerase catalytic subunit-hTERT also affected the expression of these genes in HCT116 cells; again, no significant change in their expression was seen (Fig. 5B, left panel, and data not shown). These results suggest that these Cyclin G2, Cdc27, and integrin ␣V expression responses may already be saturated by the high level of telomerase present in this telomerase-positive cancer cell line.
The expression of Cyclin G2 has been shown to fluctuate during the cell cycle, peaking in late S phase (56) and thus potentially coinciding with telomere synthesis in the cell cycle and possibly linking these processes. We then tested the association of Cyclin G2 levels and cell proliferation rates in an independent cell type, human fibroblasts, which normally have very low or no telomerase activity. We determined the effects on Cyclin G2 expression of hTERT overexpression in WI38 human fibroblasts, early and late in passaging (Fig. 5B). A previous study (26) of the effects of hTERT overexpression on mRNA profiles in human fibroblasts has not reported any effects on Cyclin G2 expression. Consistent with this, hTERT overexpression had no effect on the Cyclin G2 mRNA levels in the early passage cells, which already were proliferating rapidly, even without hTERT overexpression. However, using Northern blotting, we showed that, compared with the control late passage cultures that were entering senescence, over-expressing hTERT in the late passage cells increased the expression of Cyclin G2 (Fig. 5B) and also, as expected, increased the cell population doubling rate (57). Interestingly, overexpression of the catalytically dead "dominant negative" mutant of hTERT failed to produce any effect on Cyclin G2 mRNA levels in either late or early passage cells (Fig. 5B). Thus, increasing the cell growth rate by hTERT overexpression was associated with a higher level of Cyclin G2 expression.
These results showed that, in two different cell type settings, HCT116 colon cancer cells and normal fibroblasts, Cyclin G2 levels show an association with proliferation rates. We therefore propose that the down-regulation of Cyclin G2 may play a role in the cell growth inhibition induced by the siRNA targeting hTER. Together, these novel findings suggest that telomerase activity in cancer cells is not only necessary for long term telomere maintenance but also promotes cancer cell proliferation in the short term. DISCUSSION Here we have shown that depleting cells of telomerase RNA, thus knocking down the level of the telomerase ribonucleoprocells infected by various lentiviruses, as indicated above, and their GenBank TM accession numbers are shown. Cyclin G2, integrin ␣V, Met proto-oncogene, and cdc27 are indicated by arrows. Each column represents data from one analysis; duplicate analyses for each condition are shown in pairs across the panel. Each row represents the expression pattern of a single gene analyzed. Gene expression ratios shown in red are up-regulated, and those in green are down-regulated. Gray areas indicate missing data. tein, is effective in rapidly inhibiting cancer cell growth. Similar overall results were obtained with every telomerase-positive human cell line we have tested, including LOX, HeLa, and HCT116 cells, as described above, and also T24 bladder cancer cells, MCF-7 and BT474 breast cancer cells, and LNCaP prostate cancer cells (data not shown). No effects are seen in the immortalized, telomerase-negative VA13 cell line. Our results revealed that human cancer cells have a rapid and defined growth-inhibitory response to depleting telomerase, which occurs even without telomere shortening (Fig. 1F and data not  shown). A similarly rapid cell inhibitory and apoptotic response was also reported for knockdown of the telomerase protein hTERT in DU145 prostate cancer cells by antisense directed against the mRNA encoding hTERT (38). Thus we conclude that it is the knockdown of the often high level of telomerase ribonucleoprotein complex in cancer cells that is important in eliciting this unexpected rapid cell response. The requirement for up-regulated telomerase (a characteristic of most human cancers), the rapidity of the killing, and the lack of dependence on p53, initial telomere length, or progressive telomere shortening point to siRNA-targeting hTER as an attractive new anti-cancer therapeutic approach.
A notable feature of the rapid growth inhibitory effects of lowering telomerase RNP levels was that bulk telomere shortening was not required. In contrast, in previous reports in which telomerase activity was decreased by various means, a delay occurred as telomeres shortened before any cell growth inhibition. We propose that, in these previous reports, the failure to detect rapid effects are attributable to important differences between those experiments and the work described here and by Folini et al. (38). Previous reports aimed at attenuating telomerase in cells usually used selection of cells stably expressing, for example, dominant negative hTert protein (DN-hTert). In experiments in which this catalytically dead but assembly-competent protein was overexpressed in cancer cells, cells that stably expressed this protein were selected, often as clonal lines, and long periods of cell doublings and concomitant bulk telomere shortening were required before the growth of these stably transfected cells was inhibited (34,35). Both of these studies have in common a selection step for the drug resistance marker carried on the vector used to introduce the dominant negative hTERT. In that drug selection step, massive cell death of untransfected cells will occur, which would mask any rapid cell growth effects. Those cell subpopulations able to evade a short term all-growth inhibition response will grow out of the drug selection step and take over the cell population, and, hence, any cells that underwent a rapid growth inhibitory response would not be represented in the stably transfected population in these experiments. This contrasts with the experiments we report here, in which first, hTER level was targeted, and second, growth effects were monitored immediately and no prior selection of cells was involved.
Our results are also consistent with the results of Folini et al. (38). In that work, the level of telomerase hTERT protein in DU145 prostate cancer cells is decreased by targeting the hTERT mRNA with an antisense peptide nucleic acid oligonucleotide, and no drug selection step is involved. Hence, as in the present work, Folini et al. were able to inspect the entire bulk population of cells virtually immediately. Because both of the approaches reported by Folini et al. and us result in a lowering of the level of the telomerase RNP, we suggest that, if the level of functional telomerase RNP is reduced, either by lowering the hTert protein level, as done by Folini et al., or by lowering the hTER level, as in the present work, the response is a rapid cell growth inhibition and apoptosis. As discussed above, this can be detected only if appropriate cell growth assays are used that do not require selection of stably transfected cell lines. This interpretation is supported by the results obtained using the inhibition of telomerase enzymatic activity in vivo by antisense nucleotides (36). Unlike in the present work, this method is not expected to lower the level of telomerase RNP. Rather, only the enzymatic activity of the telomerase enzyme complex is decreased. It was shown that experimentally decreasing the telomerase activity level by an oligonucleotide complementary to the telomerase RNA template region inhibits telomerase activity but not telomerase RNA levels (58). In those reports, before any effects on cell growth is seen, there is the expected delay of many cell doublings, during which the bulk telomere population shortens. Hence, we propose that it may be the level of telomerase ribonucleoprotein, rather than the level of telomerase enzymatic activity, that results in the rapid cellular response.
We have demonstrated that the rapid cell growth inhibition and apoptosis elicited by siRNA directed against wild-type telomerase RNA did not depend on the cellular status of p53, which is frequently inactivated in human cancer. The property of at least some p53 independence is also seen for the telomere loss and apoptosis induced by overexpression of a catalytically dead telomerase catalytic subunit (DN-hTert) (35). Interestingly, this lack of dependence on p53 contrasts with observations made using a different type of disruption of telomere function that results in rapid telomere uncapping, expression of TRF2 ⌬B⌬M , a dominant negative form of the telomere-protective protein TRF2 (59,60). In that case, the induced apoptosis and growth inhibition were p53-dependent (59,60). We saw no telomere uncapping with the hTER-targeting siRNA, and therefore we propose that cells respond by different pathways to telomerase RNP knockdown versus TRF2 ⌬B⌬M .
What might account for the rapid cell growth effect of telomerase depletion in cancer cells? We suggest that cancer cells may have become adapted to their often high levels of upregulated telomerase. It has been proposed that cancer cells have significantly altered regulatory circuits, rendering them susceptible to a change in level of a protein to which they have become abnormally adapted or "addicted" (61). Specifically, we found that Cyclin G2 levels were decreased upon telomerase RNA depletion. We propose that cancer cells are telomeraseaddicted and, hence, susceptible to a fast growth-inhibitory (including apoptotic) response when the telomerase level is abruptly decreased.
Integrins are heterodimeric integral membrane proteins consisting of noncovalently linked ␣ and ␤ chains. They are fundamental regulators of cell growth, migration, survival, and differentiation (62). Previous studies suggest that integrin ␣V is essential not only for cancer cell proliferation and survival (63) but also for cancer angiogenesis and metastasis (64,65). Inhibition of integrin ␣V reduces tumor cell proliferation and cancer angiogenesis and metastasis (66,67). The observation that hTER-targeting siRNA inhibits integrin ␣V expression provides a possible mechanism for our recent finding that knocking down telomerase with a systemically delivered antitelomerase RNA ribozyme decreases melanoma metastasis in an in vivo mouse model (40). Together, these findings implicate telomerase status as a new player in the regulation of integrins in cancer progression.
Two recent reports (32,68) suggested there is a functional difference in tumor progression and metastasis between tumor cells with activated telomerase versus alternative lengthening of telomeres. Our results on gene expression changes, elicited by knocking down the telomerase RNA level, reveal effects of telomerase expression that may cast new light on the role of telomerase in cancer cells. Little was known about the roles of any cellular factors in connecting telomere maintenance to cell cycle progression and tumor progression in cancerous or other cells. Our findings indicate that elevated telomerase activity itself promotes cancer-proliferative properties through regulation of genes involved in tumor growth and progression.