G-quadruplexes Induce Apoptosis in Tumor Cells

Several G-rich oligodeoxynucleotides (ODNs), which are capable of forming G-quadruplexes, have been shown to exhibit antiproliferative activity against tumor cell lines and antitumor activity in nude mice carrying prostate and breast tumor xenografts. However, the molecular basis for their antitumor activity remains unclear. In the current study, we showed that a variety of telomeric G-tail oligodeoxynu-cleotides (TG-ODNs) exhibited antiproliferative activity against many tumor cells in culture. Systematic mutational analysis of the TG-ODNs suggests that the antiproliferative activity depends on the G-quadruplex conformation of these TG-ODNs. TG-ODNs were also shown to induce poly(ADP-ribose) polymerase-1 cleavage, phosphatidylserine flipping, and caspase activation, indicative of induction of apoptosis. TG-ODN–induced apoptosis was largely ataxia telangiectasia mutated (ATM) dependent. Furthermore, TG-ODN–induced apoptosis was inhibited by the c-Jun NH 2-terminal kinase (JNK) inhibitor SP600125. Indeed, TG-ODNs were shown to activate the JNK pathway in an ATM-dependent manner as evidenced by elevated phosphorylation of JNK and c-Jun. Interestingly, a number of G-quadruplex ODNs (GQ-ODN) derived from nontelomeric sequences also induced ATM/JNK-dependent apoptosis, suggesting a possible common mechanism of tumor cell killing by GQ-ODNs.


Introduction
G-quadruplexes are four-stranded DNA with stacks of Gquartets formed by four guanines in a planar structure through Hoogsteen base pairing (1).G-quadruplexes can be parallel or antiparallel, as well as intramolecular or intermolecular (2)(3)(4).Nucleic acid sequences with multiple stretches of guanine (G) have a tendency to form thermodynamically stable G-quadruplexes.The stability is derived from hydrogen bonding within each quartet, stacking of the hydrophobic quartets on one another, and coordinating of a monovalent cation in the central channel (reviewed in ref. 5).Both nuclear magnetic resonance and X-ray crystallographic structures of G-quadruplexes have been obtained at high resolution (2,3).Despite the extensive study on Gquadruplex structure, its function in cells is not well understood.
Several studies have shown that G-rich oligodeoxynucleotides (ODNs), which are capable of forming G-quadruplexes, exhibit antiproliferative activities (6,7).For example, the G-quadruplex ODN (GQ-ODN) T40214, which forms a G-quadruplex structure at intracellular K + ion concentrations, has been shown to inhibit interleukin-6-stimulated signal transducers and activators of transcription 3 (Stat3) and suppress the Stat3-mediated upregulation of Bcl-X L and Mcl-1 gene expression (6).Computersimulated docking studies have indicated that GQ-ODNs interact mainly with the SH2 domain of Stat3 and are capable of inserting itself between the SH2 domains of Stat3 dimers bound to DNA (6,8).Most importantly, recent studies have shown that two Stat3inhibiting GQ-ODNs exhibit antitumor activity in a nude mice model xenografted with human breast cancer, prostate cancer, and head and neck cancer cells (6,8,9).Stat3 is an oncogene that is activated in many human cancer cells.Because disruption of Stat3 activity is expected to inhibit cancer cell growth and enhance apoptosis, it has been suggested that GQ-ODNs represent a new class of anticancer drugs by targeting Stat3 within cancer cells (6,8).
Another group has shown that G-rich ODNs have antiproliferative activity against a number of cancer cell lines (7).Again, the antiproliferative activity of these G-rich ODNs (e.g., GRO29A) is associated with their ability to form stable G-quartet-containing structures and their potential binding to a specific cellular protein, nucleolin (7).Cell cycle analysis and the 5-bromodeoxyuridine incorporation study have suggested that these GQ-ODNs specifically inhibit DNA synthesis and arrest cells at S phase (10).These GQ-ODNs have also been shown to inhibit proliferation of tumor cells (7).
Telomeres are ends of chromosomes consisting of repeated DNA sequences with bound proteins.In humans, telomeric DNA consists of TTAGGG/CCCTAA DNA repeats of 5 to 20 kb in length (11,12).Single-stranded overhangs (telomeric G-tails) of f100 to 200 bp with TTAGGG repeats are present at the 3 ¶ ends of human telomeres (13,14).Telomeric G-tails can readily adopt the thermodynamically stable G-quadruplex conformation in vitro (2,3).G-quadruplexes have also been detected at telomeres in vivo (15)(16)(17)(18).Deprotection of telomeres has been shown to induce ataxia telangiectasia mutated (ATM)-dependent apoptosis (19).However, it is unclear whether telomeric G-tails in the Gquadruplex conformation participate in apoptotic signaling.In the current study, we tested the ability of a number of telomeric G-tail ODNs (TG-ODNs) in inducing apoptosis by transfection studies.We showed that TG-ODNs in their G-quadruplex conformation were highly effective at induction of apoptosis in tumor cells.TG-ODN-induced apoptosis was largely ATM dependent.In addition, an inhibitor of c-Jun NH 2 -terminal kinase (JNK) inhibited TG-ODN-induced apoptosis.Interestingly, nontelomeric GQ-ODNs, including the Stat3-targeting G-quadruplex T40214, were also shown to induce ATM-and JNK-dependent apoptosis, suggesting a possible common mechanism of tumor cell killing by GQ-ODNs.

Materials and Methods
Cells and oligonucleotides.Human primary lung fibroblast WI38 and its variant, 2RA (SV40 T antigen-transformed WI38), were obtained from American Type Culture Collection (Manassas, VA).Primary mouse embryonic fibroblasts (MEF) and MEF/E1A (E1A-transformed MEFs) were kindly provided by Dr. Alexey Ryazanov (University of Medicine and Dentistry of New Jersey-Robert Wood Johnson Medical School, Piscataway, NJ).YZ5 (ATM+) and pEBS7 (ATMÀ) cells, obtained from Dr. Gary Kuo (University of Pennsylvania, Philadelphia, PA), were derived from SV40immortalized AT fibroblast cells by transfection with ATM cDNA (pEBS7-YZ5) and the empty vector pEBS7, respectively (20).GW33 (ATR+) and GK41 (ATRÀ) cells, obtained from Dr. Stuart Schreiber (Harvard University, Cambridge, MA), were derived from human osteosarcoma U2OS cells transfected with plasmids carrying tetracycline-inducible ataxia telangiectasia-mutated and Rad3-related (ATR) wild-type cDNA and ATR kinasedead cDNA, respectively (21).p53+ (wild-type) and p53À MEFs were obtained from Dr. Shenkan Jin (University of Medicine and Dentistry of New Jersey-Robert Wood Johnson Medical School, Piscataway, NJ).Human colon cancer lines HCT-116 and HCT-116/p53À were obtained from Dr. Bert Vogelstein (Johns Hopkins Medical School, Baltimore, MD; ref. 22).KB (human oral epidermoid carcinoma) cells were obtained from the Cancer Institute of New Jersey (New Brunswick, NJ).All cells were cultured in DMEM supplemented with 10% fetal bovine serum at 37jC in a 5% CO 2 incubator.
Transfection and 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay.For transfection, cells were seeded in 96-well plates (f10 3 per well for transformed cells and 3 Â 10 3 per well for primary cells) and cultured overnight.Transfections were done by mixing ODNs or plasmids with Cellfectin (Invitrogen, Carlsbad, CA) for 1 hour at 25jC following the manufacturer's instructions.The mixtures were then diluted with DMEM and added to cells.Cells were incubated at 37jC for 18 hours, or as indicated.
For 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay, ODN-transfected cells in triplicates were replenished with fresh DMEM containing 10% fetal bovine serum and cultured for 3 days at 37jC.Cells were stained with MTT for 3 hours and then dissolved in 100 AL of DMSO.A 570 was measured using a microplate reader as the MTT value.The MTT value was then normalized against the A 570 of cells treated with Cellfectin alone.At least three independent experiments were done for each MTT assay.The results shown were from representative experiments, with error bars being the SD of the triplicates.
Caspase activation and phosphatidylserine flipping.About 2.5 Â 10 5 cells were transfected with various ODNs for 18 hours.Transfected cells were then stained with either FITC-VAD-FMK (10 Amol/L; Promega, Madison, WI) for 15 minutes for detecting caspase activation or with ApoDETECT Annexin V-FITC (Invitrogen) for detecting phosphatidylserine flipping.
For double staining with FITC-VAD-FMK and propidium iodide, FITC-VAD-FMK-stained cells were washed with PBS and fixed with 66% ethanol at 4jC for 30 minutes.Cells were then treated with RNase A (0.1 mg/mL) and stained with propidium iodide (25 Ag/mL) for 15 minutes before fluorescence activated cell sorting (FACS) analysis.
Temperature-dependent absorption spectroscopy.Temperaturedependent absorption experiments were conducted using an AVIV Model 14DS Spectrophotometer (Aviv Biomedical, Lakewood, NJ) equipped with a thermoelectrically controlled cell holder.Quartz cells with a path length of either 0.1 or 1.0 cm were used for all the absorbance studies.Temperaturedependent absorption profiles were acquired at 295 nm with a 5-second averaging time.The temperature was raised in 0.5jC increments and the samples were allowed to equilibrate for 1.5 minutes at each temperature setting.In these UV melting studies, the DNA solutions ranged in concentration from 4 to 120 Amol/L in strand.Buffer solutions contained Figure 1.TG-ODNs exhibit antiproliferative activity.ODNs were transfected into cells using Cellfectin.After 18 hours, ODN-transfected cells were replenished with fresh DMEM and incubated for 3 days at 37jC with 5% CO 2 .Cell survival was measured by the MTT assay.The MTT values (average of triplicates) for cells treated with ODNs were normalized against the MTT value for cells treated with Cellfectin alone.At least three independent experiments were done for each ODN.Data shown are from a representative experiment.A, 24-mer TG-ODN exhibits antiproliferative against KB cells.ODNs used in this experiment were TG24 (four TTAGGG repeats), CA24 (the complementary sequence of TG24), dsTEL (the annealed duplex product of TG24 and CA24), and N24 (totally degenerate 24-mer single-stranded ODN).B, G-rich ODNs, but not G-less ODNs, exhibit antiproliferative activity against KB cell lines in culture.C, TG-ODN TG24 exhibits antiproliferative activity against a number of tumor cells.D, G15T3 exhibits selective antiproliferative activity against transformed cells.WI38 , primary human fetal lung fibroblasts.2RA , SV40 T antigentransformed WI38.
10 mmol/L EPPS (pH 7.5), 150 mmol/L KCl, and 0.1 mmol/L EDTA.Before their use in UV melting experiments, all DNA solutions were preheated at 90jC for 5 minutes and then slowly cooled to room temperature over a period of 4 hours.

Results
TG-ODNs exhibit antiproliferative activity against tumor cells in culture.Various TG-ODNs were transfected into KB cells using Cellfectin.Growth inhibition was monitored by the MTT assay.As shown in Fig. 1A, the single-stranded TG-ODN TG24, consisting of four TTAGGG repeats (Fig. 1A), was highly effective in inhibiting the growth of KB cells with an IC 50 of f50 nmol/L.Neither its complementary C-rich ODN (CA24) nor its doublestranded ODN (dsTEL) had any significant effect on cell growth under our assay conditions.A 24-mer single-stranded ODN with totally degenerate sequences (N24) was also shown to be essentially inactive under our assay conditions (Fig. 1A).A mixture of four deoxynucleotide triphosphates was used as a control and found to be inactive (data not shown).Similar results were obtained using TG18 (three TTAGGG repeats), CA18 (three CCCTAA repeats), N18 (totally degenerate 18-mer), and dsTEL18 (annealed duplex of TG18 and CA18; Supplementary Fig. S1A).G-less ODNs, T15C3, A15C3, and C15T3, were essentially inactive in inhibiting the growth of KB cells, whereas M5 and G15T3, which contain more Gs than TG18, exhibited higher antiproliferative activity (Fig. 1B).TG-ODNs were shown to exhibit antiproliferative activity against, in addition to KB cells, many other tumor cell lines, such as HeLa cells, breast cancer MDA231, prostate cancer PC3-1, and melanoma A875 (Fig. 1C).Moreover, TG-ODNs were more effective at inhibiting the growth of transformed cells than their nontransformed counterparts.As shown in Fig. 1D and Supplementary Fig. S1B, G15T3 was much more effective in inhibiting the growth of transformed cells, 2RA and MEF/E1A, than their respective nontransformed counterparts, WI38 and MEF/WT.
The antiproliferative activity of TG-ODNs depends on their G-quadruplex conformation.The importance of the G content for the activity of TG-ODNs suggests a potential involvement of G-quadruplexes in cell killing (Fig. 1B).To test this possibility, TG18 was heated to 100jC for 10 minutes and cooled down on ice quickly.The denatured TG18 was then assayed immediately for its antiproliferative activity using the MTT assay.As shown in Fig. 2A (left), the growth inhibitory activity was completely lost upon heating.To rule out the possibility that TG18 was destroyed during heating (e.g., by depurinated or degraded), heat-inactivated TG18 was renatured in K buffer before transfection.As shown in Fig. 2A (left), renatured TG18 completely regained its antiproliferative activity.
We also did the same denaturation/renaturation experiment using TG24 instead of TG18.As shown in Fig. 2A (right), denatured TG24, unlike denatured TG18, only partially lost its antiproliferative activity.Renatured TG24 completely regained its antiproliferative activity.These results suggest that the growth inhibitory activity of TG-ODNs is conformation dependent.The difference in antiproliferative activity between denatured TG18 and denatured TG24 may be due to their ability to form intermolecular and intramolecular G-quadruplexes, respectively, during transfection (see below).
In contrast to the hyperchromic shift at 260 nm associated with melting of duplex DNA, the melting of G-quadruplex DNA is associated with a hypochromic shift at 295 nm (23)(24)(25)(26).Consequently, the thermally induced structural transitions of G-quadruplexes into single-stranded forms can be readily detected by examining their characteristic UV melting profiles at 295 nm. Figure 3 shows the UV melting profiles (depicted in their firstderivative forms) of TG12 (A), TG18 (B), TG24 (C), and TG48 (D) acquired at two different concentrations of each ODN.Note that all four TG-ODNs exhibited detectable thermal transitions at 295 nm, an observation consistent with each ODN forming one or more quadruplex structures.Similar UV melting experiments on TG6 and CA18 (Supplementary Fig. S3) did not produce any detectable thermal transitions at 295 nm, indicating that these ODNs did not form quadruplex structures under the conditions employed.
Further inspection of Fig. 3 reveals that the transition temperatures (T tran ) corresponding to the minima of the first-derivative melting profiles of TG24 and TG48 do not vary with DNA concentration (Fig. 3C and D).This result is indicative of both ODNs forming monomolecular (i.e., intramolecular) quadruplex structures.In contrast to the T tran values of TG24 and TG48, the corresponding T tran values of TG12 and TG18 vary with DNA concentration, with larger DNA concentrations yielding higher T tran values (Fig. 3A and B).This concentration dependence of T tran indicates that the G-quadruplex structures formed by TG12 and TG18 have molecularities >1.In other words, TG12 and TG18 form intermolecular rather than intramolecular G-quadruplex structures.Note that the UV melting profiles of TG12 are biphasic (Fig. 3A).Thus, TG12 seems to form a mixture of intermolecular quadruplex structures.
The conformation of TG-ODNs was also examined by nondenaturing PAGE.As shown in Fig. 2B, both TG18 and TG24 migrated in the gel as a single species (see the band labeled G in lanes 1 and 5).However, upon heat denaturation, TG18 and TG24 migrated much slower (see the band labeled L in lanes 2 and 6).Upon further renaturation in K buffer, TG18 and TG24 again migrated in the gel as a single band with the same mobility as their untreated counterparts (see the band labeled G in lanes 3 and 7).By contrast, the mobility of CA18 and CA24 was independent of heat denaturation/renaturation (Supplementary Fig. S2).Denatured TG24 migrated at a position slightly slower than CA24 (Supplementary Fig. S2).Based on results from the UV melting analysis (see above), the bands labeled G most likely represent ODNs in their G-quadruplex conformation, whereas the bands labeled L represent ODNs in their denatured states.
It is interesting to note that, despite its lower molecular weight, untreated or renatured TG18 migrated slower than untreated or renatured TG24.This result may reflect the formation of an intermolecular G-quadruplex by TG18 and an intramolecular G-quadruplex by TG24, consistent with results from UV melting analysis.
To confirm that the G-quadruplex conformation of TG-ODNs is responsible for their growth inhibitory activity, the repeating unit TTAGGG in TG18 was systematically mutated to change its G content.As shown in Fig. 2C, the antiproliferative activities of M1, M2, M3, and M4 were dramatically reduced as compared with that of TG18 (Fig. 2C).Neither M1, M2, M3, nor M4 exhibited a thermal transition at 295 nm (data not shown), suggesting that no ODN formed a G-quadruplex structure.By contrast, M5 exhibited a UV melting profile consistent with the formation of at least three intermolecular G-quadruplex structures (see Supplementary Fig. S3).Note that the T tran values of M5 (65.5jC, 81.5jC, and 87.5jC) are greater than that of TG18 (44.0jC) at an identical DNA strand concentration of 120 Amol/L.Further, note that the antiproliferative activity of M5 also exceeds that of TG18 (see Fig. 1B).We have repeated the same mutagenesis and UV melting experiments with TG24 and obtained essentially the same results (not shown).These results further support the notion that the G-quadruplex conformation is responsible for the antiproliferative activity of TG-ODNs.
We observed a general correlation between the potency with which the TG-ODNs inhibited tumor cell growth and the number of TTAGGG repeats.As shown in Fig. 2D, the order of the potency was TG48 % TG24 > TG18 > TG12.Note the general concordance between the antiproliferative activities of TG12, TG18, TG24, and TG48 and the corresponding thermal stabilities of the quadruplex structures they form (see T tran values in Fig. 3).This concordance lends further support to the notion that the G-quadruplex conformation is responsible for the antiproliferative activities of TG-ODNs.Indeed, TG6 (one repeat of TTAGGG), which does not form a quadruplex structure (see Supplementary Fig. S3), exhibits no antiproliferative activity (Fig. 2D).The slightly higher IC 50 of TG48 as compared with that of TG24 could be due to more Gquadruplex contents of TG48 when compared at equal molarity, in combination with lower T tran .
GQ-ODNs induce apoptosis.To test whether the antiproliferative activity of GQ-ODNs is due to cell growth arrest, cell death, or both, GQ-ODN-transfected KB cells (18-hour transfection) were double stained with propidium iodide for cell cycle analysis and FITC-conjugated VAD-FMK for caspase activation.As shown in were transfected with Cellfectin alone (A), or 0.4 Amol/L TG18 (B) or G15T3 (C ).After 18 hours of transfection, cells were harvested and stained with 10 Amol/L FITC-VAD-FMK, a FITC-conjugated caspase inhibitor.Cells were then fixed and stained with propidium iodide before FACS analysis.Top, FACS analysis of double-stained cells: Y axis, FITC-VAD-FMK signal as indication of apoptosis.X axis, propidium iodide signal for DNA content.Bottom, DNA content only.D, TG-ODNs induce ATM-dependent apoptosis.YZ5 (ATM+) and pEBS7 (ATMÀ) cells were transfected with TG-ODNs for 5 hours.Left, cleavage of the caspase-3 substrate PARP1 was measured by Western blotting with anti-PARP1 antibody.Quantification of results was done by using densitometric scanning and the Quantity One software from Bio-Rad (Hercules, CA).The intensities of gel bands were normalized against the internal loading control tubulin.Right, phosphatidylserine flipping measured by FACS using FITC-conjugated Annexin V.The intensity of Annexin V staining of ODN-treated cells was normalized to cells treated with Cellfectin alone (relative Annexin V staining).Columns, average of triplicate experiments.
Fig. 4B and C, KB cells transfected with either TG18 or G15T3 for 18 hours exhibited a large increase in apoptotic cell population (>50%) as evidenced by caspase activation (top, Y axis).It is also evident from Fig. 4B and C that apoptosis occurred in all phases of the cell cycle including G 1 , S, and G 2 -M.A similar observation was made for MDA435 breast cancer cell line transfected with G15T3 and T40214 (a putative Stat3-targeting GQ-ODN; refs.6, 8; Supplementary Fig. S4A).Because no significant change in cell cycle profile (e.g., G 1 accumulation) was observed following transfection with GQ-ODNs, the primary cause of the antiproliferative activity of GQ-ODNs may not be growth arrest but rather cell death (e.g., apoptosis).
To further confirm that GQ-ODNs induce apoptosis, additional apoptosis end points were monitored.As shown in Fig. 4D (right), TG24 and TG24M5 induced phosphatidylserine flipping from the inner to outer cell membrane as evidenced by FACS analysis of FITC-conjugated Annexin V-stained cells.Similarly, G15T3 and T40214 also induced phosphatidylserine flipping as shown in Fig. 6C.Moreover, as shown in Fig. 4D (left), TG18 and G15T3 induced PARP1 cleavage, indicative of apoptosis.Also shown in Supplementary Fig. S4B, TG18, TG24, and TG24M5, but not CA24 or their denatured counterparts (see d-TG18, d-TG24, and d-TG24M5 in Supplementary Fig. S4B), induced cleavage of PARP1, which was abolished by cotreatment with the pan-caspase inhibitor z-VAD-FMK during transfection (see TG18 + CI, TG24 + CI, and TG24M5 + CI in Supplementary Fig. S4B).Together, these results suggest that GQ-ODNs induce apoptotic death.
The antiproliferative and apoptosis-inducing activities of GQ-ODNs are ATM dependent.Telomere deprotection is known to induce ATM-dependent apoptosis (19,27).To test whether GQ-ODNs induce ATM-dependent cell death, YZ5 (ATM+) and pEBS7 (ATMÀ) cells (28) were used for transfection studies.We first determined whether the antiproliferative activity of GQ-ODNs is ATM dependent using MTT assay.As shown in Fig. 5A, GQ-ODNs TG24 and G15T3 were shown to exhibit stronger antiproliferative activity against YZ5 (ATM+) than pEBS7 (ATMÀ) cells.T40214, a putative Stat3-targeting GQ-ODN known to inhibit tumor growth in mice (6,8), also exhibited ATM-dependent antiproliferative activity (Fig. 5A).Interestingly, ATM-dependent antiproliferative activity was not observed with camptothecin (a topoisomerase I-specific drug), etoposide (a topoisomerase II-specific drug; refs.29,30), or staurosporine (a tyrosine kinase inhibitor; Fig. 5B).The ATM dependence of the antiproliferative activity of TG24 was also confirmed with the use of another pair of ATM cell lines in which ATM was silenced by small interfering RNA (Supplementary Fig. S5).
We also evaluated the potential role of ATR and p53 in the antiproliferative activity of GQ-ODNs.Cells overexpressing a dominant negative kinase-dead mutant ATR (ATRÀ) and its A, both YZ5 (ATM+) and pEBS7 (ATMÀ) cells were transfected with TG24, G15T3, or T40214 for 18 hours.B, both YZ5 (ATM+) and pEBS7 (ATMÀ) cells were treated with camptothecin (CPT ), etoposide, or staurosporine (ST ) for 18 hours.C, the antiproliferative activity of GQ-ODNs is independent of ATR and p53.Twenty-four hours before transfection with G15T3, tetracycline (1 Ag/mL) was added to GW33 (ATR+) and GK41 (ATRÀ) cells (left) to induce expression of wild-type and kinase-dead ATR, respectively.To determine the role of p53 in TG-ODN-induced apoptosis, MEF (p53+) and MEF (p53À) were transfected with TG24 (right ).For experiments in (A) to (C ), transfected cells were cultured in fresh medium and survival was measured by MTT assay after 3 days.
wild-type ATR (ATR+) were employed (21).As shown in Fig. 5C, G15T3 was shown to exhibit antiproliferative activity against both ATR+ and ATRÀ cells to a similar extent (Fig. 5C).Wild-type MEF/ p53+ and mutant MEF/p53À were also used to evaluate the role of p53.As shown in Fig. 5C, TG24 exhibited similar antiproliferative activity against MEF/p53+ and MEF/p53À.The same result was obtained using HCT-116 and HCT-116/p53 cells (data not shown).
GQ-ODNs activate the JNK apoptosis pathway.The JNK pathway is known to play an important role in stress-induced apoptosis (31)(32)(33)(34)(35), and ATM is known to regulate the JNKdependent apoptotic pathway (36)(37)(38)(39).Consequently, we tested the possibility that GQ-ODNs may activate JNK and induce JNKdependent apoptosis.As shown in Fig. 6A ( first row) and Fig. 6B, JNK was activated on transfection of YZ5 (ATM+) cells with GQ-ODNs such as TG24 and T40214, but not with CA24, as evidenced by the increased phosphorylation of p46 and p54 JNKs using phospho-specific JNK antibody.Furthermore, c-Jun, a JNK substrate, was also shown to be phosphorylated in YZ5 (ATM+) cells transfected with GQ-ODNs, as detected by using antibody specific for Ser 63 phosphorylated c-Jun (Fig. 6A and B), indicative of JNK activation.Interestingly, activation of the JNK pathway was ATM dependent because phosphorylation of both JNK and c-Jun by TG24 and T40214 was much reduced in pEBS7 (ATMÀ) cells compared with that in YZ5 (ATM+) cells [Fig.6A (compare left and right) and B].Furthermore, caspase inhibition by a pancaspase inhibitor, z-VAD-FMK, did not block activation of JNK by TG-ODNs (Supplementary Fig. S6), suggesting that activation of JNK is upstream rather than downstream of apoptosis.

Discussion
In the current study, we have shown that many TG-ODNs exhibit potent growth inhibitory activity against tumor cell lines in culture.The ability of TG-ODNs to inhibit tumor cell growth seems to depend on their G-quadruplex conformation rather than their specific telomeric DNA sequences.This conclusion is based on three observations.First, the cell killing activity of TG-ODNs depends on the presence of a string of Gs rather than the G content.For example, TG18 [(TTAGGG) 3 ] and M4 [(TGTGAG) 3 ] have the same G content.However, only TG18 exhibited growth inhibitory activity.The presence of strings of Gs is known to favor G-quadruplex formation.Second, there is a strong correlation between the presence of the G-quadruplex conformation (based on their characteristic UV melting profiles) and cytotoxicity among a panel of TG-ODNs.For example, mutant and short TG-ODNs that fail to display G-quadruplex-specific UV melting profiles do not exhibit cytotoxicity.Third, the growth inhibitory activity of TG-ODNs is conformation dependent.A brief heating of TG-ODNs to 100jC abolished their cell killing activity.Most strikingly, renaturation of the heated TG-ODNs restored their cell killing activity.
It is interesting to note that denatured TG18 completely lost its antiproliferative activity whereas denatured TG24 only partially lost its antiproliferative activity (Fig. 2A).As shown in Fig. 3B, TG18 forms an intermolecular G-quadruplex due to the presence of only three TTAGGG repeats whereas TG24 forms an intramolecular G-quadruplex due to the presence of four TTAGGG repeats.Consequently, denatured TG18 is expected to form G-quadruplex slowly in a concentration-dependent manner whereas TG24 is expected to reform G-quadruplex rapidly in a concentrationindependent manner.It is possible that TG24, but not TG18, may Figure 6.GQ-ODNs induce JNK activation.A, GQ-ODNs induce phosphorylation of JNK and c-Jun.YZ5 (ATM+) and pEBS7 (ATMÀ) cells were transfected with GQ-ODNs (1 Amol/L) or treated with camptothecin (0.1 Amol/L) for 18 hours.Western blotting was done with antibodies against phospho-JNK (p54 and p46) and phospho-Ser 63 c-Jun, as well as antibody against tubulin for assessing loading (see labels on the left).Various treatment conditions (untreated, Cellfectin, CA24, TG24, T40214, and camptothecin) were shown on top of each lane.B, quantification of results from (A ).The intensities of gel bands in (A), normalized against the internal control tubulin, were determined by densitometric scanning and quantified using the Quantity One software from Bio-Rad.C, GQ-ODN-induced apoptosis is blocked by the JNK inhibitor SP600125.KB cells were transfected with T40214 (1 Amol/L) or G15T3 (1 Amol/L) for 18 hours using Cellfectin in the presence or absence of the JNK inhibitor SP600125 (3.5 Amol/L; Biomol, Plymouth Meeting, PA).Transfected cells were then double stained with FITC-conjugated Annexin V and propidium iodide, followed by FACS analysis.Annexin V-positive/propidium iodide-negative population was plotted.Columns, average of two independent experiments.
have reformed G-quadruplex during transfection.It should be noted that the loss of antiproliferative activity of denatured TG18 was not due to reduced transfection efficiency (Supplementary Table S1).
Whereas our studies focused on the use of TG-ODNs (e.g., TG18, TG24, and TG48), nontelomeric G-quadruplex-forming ODNs (e.g.M5, G15T3, and T40214) exhibited similar activity.Moreover, all these G-quadruplex-forming ODNs (GQ-ODNs) induced ATM-dependent JNK activation and apoptosis.These results further support that the G-quadruplex conformation, rather than the specific DNA sequences, is primarily responsible for the antiproliferative and apoptosis-inducing activities of TG-ODNs and other GQ-ODNs.
It is possible that the cytotoxicity of GQ-ODNs is due to their binding to a critical G-quadruplex binding protein(s) in tumor cells, which results in activation of apoptosis.Many G-quadruplex binding proteins, such as human topoisomerase I (41), human Stat3, transcription factor MyoD (42), nucleolin (7), and others (43,44), have been identified.Many of these proteins are essential (such as topoisomerase I and Stat3) for tumor cell proliferation.It seems possible that different GQ-ODNs may exhibit certain degrees of selectivity against their target proteins to induce cell death.For example, T40214, a putative Stat3-selective GQ-ODN, may induce apoptosis through its interference with Stat3 binding to DNA (6,8), and GRO29A, another GQ-ODN, may inhibit cell proliferation through binding to nucleolin (10).
However, our current studies have also suggested the possible existence of a common mechanism of tumor cell killing by GQ-ODNs.We have shown that, in addition to TG-ODNs, the Stat3targeting G-quadruplex-forming ODN T40214, as well as other G-quadruplex-forming ODNs derived from nontelomeric sequences, also induces ATM-and JNK-dependent cell killing.One possible explanation is that GQ-ODNs may bind to a common target, which triggers cell death.For example, GQ-ODNs may cause telomere deprotection by titrating a telomere binding protein and thus activate ATM-dependent apoptosis (19,27).As telomeric G-tails can readily adopt G-quadruplex conformation in vitro (2, 3) and G-quadruplexes have also been detected at telomeres in vivo (15)(16)(17)(18), it is tempting to speculate that G-quadruplex-induced cell death may occur during telomere deprotection and GQ-ODNs may mimic such a signal and hijack this death pathway.Our studies have shown that GQ-ODN-induced apoptosis is p53 independent.It has been reported that telomere dysfunction by overexpression of the dominant TRF2 can lead to p53-independent apoptosis in liver cells (45), and mutations in human telomerase RNA also trigger p53-independent apoptosis in tumor cells (46).It remains to be determined whether GQ-ODN-induced apoptosis mimics a pathway of apoptosis resulting from telomere dysfunction.
Our studies have also shown that GQ-ODNs activate the JNK pathway in an ATM-dependent manner.In addition, the JNKspecific inhibitor SP600125 partially inhibited GQ-ODN-induced apoptosis.These results suggest that GQ-ODN-induced apoptosis is mediated, at least in part, through activation of the JNK pathway.The JNK pathway is known to be activated by ATM in response to UVA or IR (38,39).c-Abl and Brca1, which are known ATM substrates, are upstream regulators of the JNK apoptosis pathway in response to IR (35,(47)(48)(49).It is possible that ATM activates JNK through one of these proteins.
It is known that oncogene Ras is activated in many tumor cells, which in turn can elevate the JNK pathway (to a sublethal level) through a signal cascade event (50).Thus, targeting the JNK pathway in tumor cells by G-quadruplexes may lead to tumorspecific cell death.Clearly, further studies are necessary to elucidate the molecular mechanism of GQ-ODN-induced tumor cell death.

Figure 2 .
Figure 2. The antiproliferative activity of TG-ODNs is conformation dependent.A, TG18 and TG24 were heat denatured at 100jC for 10 minutes and then used immediately to transfect KB cells.Aliquots of the heat-denatured ODNs were renatured in K buffer as described in Materials and Methods and used for transfection.Cell survival was monitored by MTT assay.B, TG18 and TG24 in their native, denatured, and renatured states were analyzed by 16% nondenaturing PAGE in TBE buffer.C, antiproliferative activity of TG18 variants (M1, M2, M3, M4, and M5) against KB cells.D, length-dependent antiproliferative activity of TG-ODNs (TG6, TG12, TG18, TG24, and TG48) compared at equal molarity (left) and equal weight (0.9 Ag/mL; right ).

Figure 3 .
Figure 3. First derivatives of the UV melting profiles of TG12 (A), TG18 (B), TG24 (C ), and TG48 (D ) at the indicated concentrations.The transition temperatures (T tran ) corresponding to the minima of the first-derivative profiles are indicated.These T tran values have an associated uncertainty of F0.5jC.The UV melting profiles of TG24 at both 4 and 40 Amol/L strand have the identical T tran value of 63.7jC.A 295 , absorbance at 295 nm.The solution conditions were 10 mmol/L EPPS (pH 7.5), 150 mmol/L KCl, and 0.1 mmol/L EDTA.

Figure 4 .
Figure 4. TG-ODNs induce apoptosis.A to C, TG-ODNs induce caspase activation in all phases of the cell cycle.KB cells (1.5 Â 10 5 )were transfected with Cellfectin alone (A), or 0.4 Amol/L TG18 (B) or G15T3 (C ).After 18 hours of transfection, cells were harvested and stained with 10 Amol/L FITC-VAD-FMK, a FITC-conjugated caspase inhibitor.Cells were then fixed and stained with propidium iodide before FACS analysis.Top, FACS analysis of double-stained cells: Y axis, FITC-VAD-FMK signal as indication of apoptosis.X axis, propidium iodide signal for DNA content.Bottom, DNA content only.D, TG-ODNs induce ATM-dependent apoptosis.YZ5 (ATM+) and pEBS7 (ATMÀ) cells were transfected with TG-ODNs for 5 hours.Left, cleavage of the caspase-3 substrate PARP1 was measured by Western blotting with anti-PARP1 antibody.Quantification of results was done by using densitometric scanning and the Quantity One software from Bio-Rad (Hercules, CA).The intensities of gel bands were normalized against the internal loading control tubulin.Right, phosphatidylserine flipping measured by FACS using FITC-conjugated Annexin V.The intensity of Annexin V staining of ODN-treated cells was normalized to cells treated with Cellfectin alone (relative Annexin V staining).Columns, average of triplicate experiments.