G-quadruplex preferentially forms at the very 3' end of vertebrate telomeric DNA.

Human chromosome ends are protected with kilobases repeats of TTAGGG. Telomere DNA shortens at replication. This shortening in most tumor cells is compensated by telomerase that adds telomere repeats to the 3' end of the G-rich telomere strand. Four TTAGGG repeats can fold into G-quadruplex that is a poor substrate for telomerase. This property has been suggested to regulate telomerase activity in vivo and telomerase inhibition via G-quadruplex stabilization is considered a therapeutic strategy against cancer. Theoretically G-quadruplex can form anywhere along the long G-rich strand. Where G-quadruplex forms determines whether the 3' telomere end is accessible to telomerase and may have implications in other functions telomere plays. We investigated G-quadruplex formation at different positions by DMS footprinting and exonuclease hydrolysis. We show that G-quadruplex preferentially forms at the very 3' end than at internal positions. This property provides a molecular basis for telomerase inhibition by G-quadruplex formation. Moreover, it may also regulate those processes that depend on the structure of the very 3' telomere end, for instance, the alternative lengthening of telomere mechanism, telomere T-loop formation, telomere end protection and the replication of bulky telomere DNA. Therefore, targeting telomere G-quadruplex may influence more telomere functions than simply inhibiting telomerase.

INTRODUCTION

Chromosomes in human cells are capped at both ends with telomere DNA of tandem (TTAGGG)n repeats. Due to the end-replication problem, telomere cannot fully replicate its 3′ end and, as a result, shortens during each round of DNA replication (1). Telomere length homeostasis is essential for the generation and growth of cancer cells. More than 85% tumors use telomerase for telomere maintenance (2), others use the alternative lengthening of telomere (ALT) mechanism (3). Four TTAGGG repeats can fold into a four-stranded structure called G-quadruplex and inhibit telomere addition by telomerase (4). Therefore, G-quadruplex has been suggested to act as a negative regulator of telomere elongation by telomerase in vivo and is currently considered a potential target for cancer therapy (5–7). Intense interest has arisen recently in searching for small molecules that stabilize G-quadruplex to inhibit telomere elongation by telomerase.

It has been assumed that telomere G-quadruplex can form anywhere at an increment of 6 nucleotides along the G-rich strand that stretches for several kilobases long (8–11). Telomere DNA terminates beyond the double stranded region with a G-rich single-stranded 3′ overhang that can be as long as several hundred bases (12,13). Several G-quadruplex-stabilizing compounds have been shown to displace telomerase (14,15) and other single telomere strand-binding proteins (16–18), suggesting that G-quadruplex can efficiently form at the telomere overhang. Apparently, where G-quadruplex forms may have implications in the biological functions and pharmaceutical applications of telomere G-quadruplex. For instance, a telomere G-rich strand can possibly have at least four forms of terminal structures with respect to the number of single stranded TTAGGG repeats at the 3′ end (i.e. 0–3 repeats). Among those structures, the one with a G-quadruplex formed at the extreme 3′ end is expected to inhibit telomerase while the others may be less or no inhibitory (Figure 1A). Similar requirement of terminal structure may also apply to the ALT mechanism in which a single stranded G-rich 3′ end has to invade and basepair with a C-rich strand for its extension (Figure 1B). Formation of G-quadruplex has also been employed in constructing molecular devices (19–21) and the position where G-quadruplex forms may also have implication in such applications.

Figure 1.

Expected effect of G-quadruplex formation on telomere DNA elongation by (A) telomerase or (B) through the ALT (alternative lengthening of telomeres) mechanism involving strand exchange. A free single-stranded G-rich 3′ end of sufficient size is required for the elongation in both pathways. A terminal G-quadruplex may carry zero to three TTAGGG repeats before the 3′ end. The G-quadruplex formed at the very 3′ end is expected to inhibit telomere elongation (dashed arrow) while the others may be less or not inhibitory. Two examples of ALT involving strand exchange are given, other examples based on the same local interaction can be found in a review (3).

At present, however, no information is available regarding G-quadruplex formation at different positions along the long telomere G-rich strand. We addressed the question with a quantitative DNA footprint method specifically designed for this purpose and a 3′ exonuclease hydrolysis stop assay using (TTAGGG)7 in which one G-quadruplex can potentially form at four different positions. Our work revealed that G-quadruplex preferentially forms at both the 5′ and 3′ ends than at internal positions. The probability at the 3′ end is higher than at the 5′ end. When an open 5′ end is not present, which mimics the telomere G-rich strand in vivo, G-quadruplex tends to form at the very 3′ end and the probability of G-quadruplex formation rapidly decreases as the position moves inwards. Experiments with longer strands revealed that the very first G-quadruplex from the 3′ end preferentially forms at the 3′ terminus and the probability rapidly decreases as it moves inwards. The polarized G-quadruplex formation may be important for the function of telomere DNA. It not only provides a molecular basis for the regulation of telomere extension by both telomerase and the ALT mechanism, but may also have implications in other processes, such as telomere end protection, T-loop formation and replication of bulky telomere DNA. Therefore, drugs targeting telomere G-quadruplex may have effects on more telomere-related processes than simply inhibiting telomerase activity.

MATERIALS AND METHODS

Principle of identifying G-quadruplex formation at different positions by footprinting

Dimethyl sulfate (DMS) footprinting that specifically cleaves at guanines (22) was used in this work. This technique uses oligonucleotide labeled at the 5′ end with 32P. DMS methylates guanine at the N7 position and subsequent treatment with piperidine breaks the DNA backbone at the methylated sites. The DNA sample is then resolved by gel electrophoresis and the 32P-containing fragments are visualized by autoradiography. The size of a band identifies the first nucleotide from the 5′ end that has been attacked.

A G-quadruplex requires four TTAGGG repeats. The N7 of those guanines in the G-quartets of a quadruplex is involved in hydrogen bonding (Figure 2A) and thus protected from the attacks by DMS (23, 24). This property has been used to probe the formation of quadruplex (25, 26). To simplify the analysis, we used oligonucleotides of 4–7 TTAGGG repeats in which only one quadruplex can form. We initially assumed that a quadruplex involves only four consecutive repeats. The guanines not involved in a quadruplex will be exposed to attacks, thus the position where a quadruplex forms determines whether a G triplet (GGG) is protected from or exposed to the attacks (Figure 2B). As a result, the sizes of cleaved fragments reflect where the quadruplex forms. The mathematical equations describing the production of DNA fragments used for analyzing quadruplex formation at different positions were given in the Supplementary data. Qudraplex formation with longer loops was also considered and details are available in the Supplementary data.

Figure 2.

Schematic illustration of quadruplex and the principle of quantifying quadruplex formation at different positions by DMS footprint. (A) (TTAGGG)4 can fold back to form G-quadruplex structure hold together by three stacked G-quartets. The Hoogsteen hydrogen bonding between adjacent guanines prevents the N7 position (arrowed) from methylation by DMS. (B) The position at which quadruplex forms determines whether a G triplet (GGG) is exposed to or protected from attack by DMS and the size of cleavage product. In a five-repeat sequence, for example, quadruplex at the very 5′ end will produce 32P-labeled fragment four repeats larger than that produced by quadruplex at the very 3′ end.

Oligonucleotides

The oligonucleotides were purchased from TaKaRa Biotech (Dalian, China). When used in gel electrophoresis, oligonucleotides were labeled at the 5′ end with [γ-32P]ATP using T4 polynucleotide kinase (Fermentas, Lithuania).

Quantitative footprint of G-quadruplex

DNA treatment and gel electrophoresis was carried out essentially as previously described (25) using 32P-labeled oligonucleotide (∼20 000 cpm, 5 ng). The oligonucleotides were treated with DMS for 30, 60, 90 s, respectively, in the presence or absence of 150 mM K+. DNA cleavage products were resolved by gel electrophoresis, autoradiographed on a Typhoon phosphor imager (Amersham Biosciences, Sweden) and the intensity of the bands corresponding to attacks at each G triplet quantified with the software ImageQuant 5.2.

The equations describing the production of DNA fragments were solved by the standard Steepest Descent Method (27, 28). The algorithm was implemented with Visual C++ in a stand-alone application. In this program, the sum of the intensity of the bands produced by cleavage at each G triplet can be input through a friendly graphical interface and the parameters, R, Qi and p, can be obtained simultaneously with an optional initial value calculation.

Exonuclease hydrolysis analysis

32P-labeled oligonucleotide was dissolved in 14 μl of 1× T4 polymerase buffer (33 mM Tris–acetate, pH 7.8, 66 mM CH3COOK, 10 mM (CH3COO)2Mg, 0.5 mM DTT) with the concentration of K+ made to 150 mM by adding KCl, heated at 95°C for 5 min and then cooled down to room temperature and kept for 30 min. Hydrolysis was carried out by adding 1 μl of T4 polymerase (5 U/μl, Takara Biotechnology, Dalian, China) and incubating at 37°C for 3 min. The reaction was stopped by adding equal volume of stop buffer (10 mM NaOH, 10 mM EDTA, 96% deionized formamide and 100 ng/μl salmon sperm DNA) and heated at 90°C for 5 min. Hydrolysis products were run on 12% sequencing gel for 2 h at 40 V/cm. The gel was autoradiographed on a Typhoon phosphor imager (Amersham Biosciences, Sweden) and band intensity was quantified with the software ImageQuant 5.2.

Incomplete hydrolysis was observed (see the first three reference lanes in Figure 5A). Calibration was made for this in each of the structural isoforms in T24(TTAGGG)7, which was calculated as (intensity ratio of the undigested/digested band in the relevant reference lane) × (intensity of the digest product of a corresponding structural isoform of T24(TTAGGG)7).

Figure 5.

Distribution of G-quadruplex formation at different positions in T24(TTAGGG)7 detected by the 3′ exonuclease activity of T4 polymerase. (A) Gel electrophoresis of hydrolysis products of T24(TTAGGG)7 and the four reference oligonucleotides designated to form G-quadruplex at desired positions leaving a tail of 3, 2, 1, 0 repeats of T6 at the 3′ end, respectively. (B) Densitometry scans of the gel shown in A). (C) Quantified distribution of G-quadruplex formation at different positions in T24(TTAGGG)7 with incomplete hydrolysis calibrated. Data are the mean of three independent experiments with standard deviation.

UV thermal melting analysis

Thermal denaturation was carried out as described previously (29,30) on a DU-640 UV–VIS spectrophotometer (Beckman, USA) equipped with a digital circulating water bath. The absorbance of oligonucleotide in 10 mM phosphate buffer (pH 7.4), 150 mM KCl was monitored at 295 nm and the temperature was simultaneously measured using a thermal probe immersed in sample cell when a sample was heated at a rate of about 1° C/min. A solution containing no oligonucleotide was used as reference.

RESULTS

DMS footprinting of 5′-32P-labeled (TTAGGG)4, (TTAGGG)5, (TTAGGG)6, (TTAGGG)7 and T24(TTAGGG)7 was carried out (25) in physiological concentration (150 mM) of K+ that is well known to stabilize G-quadruplex (31). The number of TTAGGG repeats was limited to be no more than seven to allow at most one G-quadruplex to form so that the analysis could be simplified. All these oligonucleotides formed intra-molecular G-quadruplexes as examined by gel electrophoresis (Figure S4, Supplementary data), which is consistent with previous results (32). DMS methylates guanine, when it is not involved in a quartet, at the N7 position and subsequent treatment breaks the DNA backbone at the methylated sites. We derived a set of mathematical equations for each of the oligonucleotides to describe the DNA cleavage using an assumption that G-quadruplex involves only four consecutive TTAGGG repeats. The equations allowed us to calculate the probability of G-quadruplex formation at different positions based on the autoradiographs of footprint.

Representative autoradiographs are given in Figure 3A–E and calculated probability of G-quadruplex formation at different positions in Figure 3F–J. The chemical cleavage selectively occurred at the guanines as expected and weak cleavage was observed at adenines (Figure 3A–E). The extremely faint footprint signals of (TTAGGG)4 in K+ solution indicate protection of the guanines in G-quadruplex, which is in sharp contrast to that of the same sample without added salt (Figure 3A). For (TTAGGG)5, G-quadruplex either forms at the 3′ or 5′ end, leaving one repeat exposed at the 5′ or 3′ end, respectively. The footprint of (TTAGGG)5 shows that the intensity of the bands produced by attacks at the first G triplet (GGG) from the 5′ end was higher than that of the rest bands (Figure 3B). This suggests that G-quadruplex preferred to form at the 3′ than at the 5′ end resulting in protection of the four G triplets at the 3′ end and exposure of one GGG triplet at the 5′ end. Indeed, calculation shows that among the G-quadruplexes, more were formed at the 3′ end (Figure 3G). For (TTAGGG)6 and (TTAGGG)7, the preference could not be intuitively judged from the gel (Figure 3C–D). However, calculations revealed that G-quadruplex formation in the middle of the sequences had a lower probability than those at both the 3′ and 5′ ends. Similarly, the probability at the 3′ end was also higher than at the 5′ end (Figure 3H–I).

Figure 3.

(A–E) Representative gels showing DMS footprint of 5′-32P-labeled (TTAGGG)4, (TTAGGG)5, (TTAGGG)6, (TTAGGG)7 and T24(TTAGGG)7. DMS treatments were carried out for 0, 30, 60 and 90 s, respectively in the presence (+) or absence (–) of 150 mM K+. (F–J) Quantification of the footprints showing the probability of G-quadruplex formation at different positions (Q, where the subscript indicates the number of TTAGGG repeats at the 3′ side of the G-quadruplex), the probability of being in the relaxed form (R) of the oligonucleotides and the probability of a GGG triplet being attacked (p). Results represent the mean of two experiments for (TTAGGG)4 and three experiments for the rest oligonucleotides with standard derivation.

Since the G-rich strand is much longer than seven repeats and does not have an open 5′ end, we attached a 24mer poly-T to the 5′ end of (TTAGGG)7 to better mimic this situation. Calculation showed that the probability at the 5′ end became the lowest and the G-quadruplex at the very 3′ end became the dominant species (Figure 3J) among the four possible positions. Similar to (TTAGGG)5, the suppression at the 5′ and dominance at the 3′ end of G-quadruplex formation in the telomeric section can be intuitively judged from the much stronger bands for smaller fragments than for larger ones (Figure 3E). In Figure 4, the G-quadruplex at each specific position as percentage of the total G-quadruplexes in T24(TTAGGG)7 was presented. It shows that more than half of the G-quadruplexes formed at the immediate 3′ end and the probability of G-quadruplex formation rapidly decreased as it moved inwards from the 3′ end.

Figure 4.

Distribution of G-quadruplexes at different positions in T24(TTAGGG)7 calculated from the data in Figure 3J.

The above calculations were based on the assumption that a G-quadruplex only involves four consecutive TTAGGG repeats. Although G-quadruplexes with longer loops consisting of a TTA and one or more GGGTTA repeats (Figure S1, Supplementary data) is possible, they are relatively unstable (Table S1, Supplementary data) and constitute only a small fraction of the total G-quadruplexes (Table S2, Supplementary data). Our calculations show that omitting these structures had little effect on our conclusion (Figures S2 and S3, Supplementary data).

The preferential formation of G-quadruplex at the very 3′ end was further confirmed by a hydrolysis assay of T24(TTAGGG)7 using the 3′ exonuclease activity of T4 polymerase (33). Four reference oligonucleotides were employed, each forms G-quadruplex at a position corresponding to a specific G-quadruplex isomer in T24(TTAGGG)7 (Figure 5A). G-quadruplex has been shown to be resistant to nuclease hydrolysis (34). T4 polymerase trimmed off nucleotides from the 3′ end until the G-quadruplex was met. The hydrolysis of T24(TTAGGG)7 produced four distinct bands corresponding to the blockage of hydrolysis by G-quadruplexes formed at the four different positions (Figure 5A). The preferential formation of G-quadruplex at the very 3′ end was indicated by the fact that the majority of the T24(TTAGGG)7 was protected from digestion. The scanned intensity distribution of these bands (Figure 5B) and further calculations (Figure 5C) resulted in a distribution similar to that revealed by the footprint (Figure 4). In human cells, the ending nucleotides at the 3′ end of human telomeric DNA varies from GGGT to GGGTTAG with a preference for the later (35). We also examined G-quadruplex formation in T24(TTAGGG)7TTAG by the T4 polymerase hydrolysis. The result (Figure 6) shows that the introduction of TTAG to the 3′ end of T24(TTAGGG)7 did not affect the distribution of G-quadruplex formation.

Figure 6.

Distribution of G-quadruplex formation at different positions in T24(TTAGGG)7TTAG detected by T4 polymerase digestion. Top panel: gel electrophoresis of hydrolysis products. Lane 1: T24(TTAGGG)7TTAG without digestion; 2: hydrolysis products of T24(TTAGGG)7TTAG; 3: T24(TTAGGG)7TTAG with the (TTAGGG)7 randomized without digestion; 4: hydrolysis products of T24(TTAGGG)7TTAG with the (TTAGGG)7 randomized. Bottom panel: Densitometry scans of lane 2 in the top panel.

We also analyzed an oligonucleotide with 12 repeats of TTAGGG by the T4 polymerase hydrolysis and the results are given in Figure 7. This oligonucleotide has nine potential positions for G-quadruplex to form and can accommodate a maximum of three G-quadruplexes. The bands of hydrolysis arrest indicate the positions of the very first G-quadruplex from the 3′ end. Similar to the oligonucleotides of seven repeats, the hydrolysis arrests in this longer oligonucleotide also indicate that such G-quadruplex preferentially formed at the 3′ end and the probability of leaving an unstructured tail at the 3′ side rapidly decreased with tail length. Hydrolysis of oligonucleotide with nine repeats also produced similar results (data not shown). These assays did not reveal the distribution of G-quadruplexes along the strand. However, the information on the first G-quadruplex from the 3′ end should reflect the probabilities of G-quadruplex formation along the DNA strand.

Figure 7.

Distribution of the first G-quadruplex from the 3′ end in long G-rich strand with 12 TTAGGG repeats detected by T4 polymerase digestion. Top panel: representative gel electrophoresis of hydrolysis products. Lane 1: hydrolysis products of HP-(TTAGGG)12; 2: HP-(TTAGGG)12 without digestion; 3: hydrolysis products of HP-(T6)12; 4: HP-(T6)12 without digestion. Bottom panel: Densitometry scans of lane 1 in the top panel. HP indicates a hairpin structure shown in the lower panel to prevent full digestion. Numbers 1–9 at the top of gel indicate hydrolysis arrests by G-quadruplexes at nine positions and HP that at the hairpin.

G-quadruplex formed in sequences with more than four repeats is flanked by repeats at one or both the 3′ and 5′ side. The presence of flanking sequence has been shown to destabilize G-quadruplex formed by G3(T2AG3)n (n = 3–16) (36) and TnG4 (n =1–8) (37). To examine the stabilities of G-quadruplexes at different positions, we performed thermal melting analysis with the four reference oligonucleotides used in the exonuclease hydrolysis assay. The result revealed that G-quadruplex at the very 3′ end is thermally more stable with a higher melting temperature (Tm) than those at internal positions (Table 1). This may partially explain the preferential formation of G-quadruplex at the very 3′ end. Formation of G-quadruplex in these sequences had to overcome the physical restraint and static electric repulsion experienced when moving along the flanking repeats. The G-quadruplex at the very 3′ end was restrained only at the 5′ end rather than at both the 3′ and 5′ ends for an internal G-quadruplex.

Table 1.

Melting temperature (Tm) of quadruplexes with 0, 1, 2 and 3 repeats of T6 at the 3′ end

Sequence	Tm (°C)
T24T18(TTAGGG)4	57.1
T24T12(TTAGGG)4T6	50.8
T24T6(TTAGGG)4T12	50.5
T24(TTAGGG)4T18	49.1

DISCUSSION

We revealed the preferential formation of G-quadruplex at the very 3′ end of the G-rich strand of telomeric DNA by two independent methods. Although the precise analysis was made with DNA in which only one G-quadruplex can form but at several potential positions, such polarity also holds for the very first G-quadruplex from the 3′ end in longer DNA that can accommodate multiple G-quadruplexes. This property should have implications in the function of telomere DNA and the pharmaceutical applications targeting the telomere G-quadruplex.

Recent reports suggest G-quadruplex exists in telomeres of human (38–41) and other species (42–46). The position where the first G-quadruplex from the 3′ end forms determines the size of the single stranded G-rich tail before the 3′ end. In a work where double-stranded telomere DNAs carrying various lengths of single-stranded G-rich overhangs were examined, a minimum tail of approximately six nucleotides was found to be required for extension by telomerase (47). Since G-quadruplex is more bulky than duplex, it might impose a more stringent length requirement than duplex does for telomerase. The polarity of G-quadruplex formation at the 3′ end keeps the size of the 3′ tail minimal to warrant the inhibition on telomerase (Figure 1A). While more than 85% tumors use telomerase (2), a subset adopts the ALT mechanism for telomere maintenance that can be achieved by telomere strand exchange (3) or telomere sister chromatid exchange (t-SCE) (48). When using telomerase inhibitors as anticancer treatment, there is a concern that inhibition of telomerase will have no effect on cancer cells using the ALT mechanism or, to be even worse, select for cells that activate the ALT mechanism for telomere maintenance (49). We expect that the preferential G-quadruplex formation at the very 3′ end should also be inhibitory to the ALT involving telomere strand exchange because in this mechanism the single stranded G-rich 3′ end has to invade and anneal with a C-rich strand to initiate the templated extension (3). The limited tail size and the presence of G-quadruplex nearby should discourage both the invasion and annealing (Figure 1B). In agreement with this, a few studies have reported antiproliferative effect in cells using the ALT mechanism for telomere maintenance (50–53).

The polarity of G-quadruplex formation may also possibly play roles in other processes or events that involve the very 3′ end of the G-rich strand. Telomere DNA terminates with a single stranded G-rich overhang beyond the double stranded region (12,13). Its 3′ end can invade the double-stranded telomere region and anneal with the C-rich strand to form circle-shaped T-loop structure (Figure 1B, second drawing without extension) (54) that is assumed to provide protection to the telomere end (55). G-quadruplex formation at the very 3′ end may potentially interfere with the formation of T-loop. G-quadruplex is resistant to nuclease hydrolysis (34). G-quadruplex formation at the 3′ end may provide protection against such hydrolysis by, for instance, exonuclease. G-quadruplex formed at the 3′ end should be less prone to the unwinding by helicases that require single-stranded DNA of sufficient size to load at the 3′ side of the structure (56–58). On the other hand, formation of G-quadruplex interferes with DNA replication in vitro (59). The lower probability of G-quadruplex formation at internal positions reduces such potential problem in the replication of the bulky telomere DNA. For the above reasons, small molecules targeting at telomere G-quadruplex may potentially affect more processes than telomerase catalyzed telomere lengthening.

Several proteins in human cells have been found capable of unfolding telomere G-quadruplex (60–62). The hPOT1 protein in human cells that binds the single-stranded telomere overhangs has been reported to disrupt G-quadruplexes allowing telomerase extension in vitro (61). Deletion of the pot1+ gene in fission yeast resulted in the loss of telomere DNA (63). These facts imply that the lack of activity to unfold G-quadruplex impaired telomere maintenance suggesting that the formation of G-quadruplex plays a role in such processes. By coordinating with such proteins, the opening and formation of G-quadruplex at the very 3′ end seems to provide a reasonable mechanism for regulating those processes and events where the 3′ end of the G-rich telomere strand is involved.

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online.