Genomic DNA was isolated from HeLa cells as described (36). dsDNA (dsDNA) carrying a T7 promoter was prepared by overlap extension polymerase chain reaction (OE-PCR) (Table 1). Other dsDNA was prepared by direct amplification of specific region of genomic DNA containing G-quadruplex sequence using polymerase chain reaction (PCR). Primers (Table 2) for PCR were purchased from Invitrogen (Shanghai, China). For some experiments, one primer was labeled at the 5′-end with a fluorescein (FAM) for quantitation. PCR was conducted in a total volume of 50 µl containing 75 mM Tris–HCl, pH 8.8, 20 mM (NH₄)₂SO₄, 0.01% Tween-20, 250 µM dNTP, 0.4 µM primer, 1.5 mM MgCl₂, 5% (v/v) DMSO, 2 U Taq Polymerase (Fermantas, MBI) and 170 ng genomic DNA as template. Thermal cycling was carried out on a Biometro thermal cycler with initial denaturation at 94°C for 3 minutes, followed by addition of Taq Polymerase and subsequent 30 cycles of 94°C for 30 s, 58°C for 30 s and 72°C for 30 s. PCR product was purified using the TIANgel Midi Purification Kit (Tiangen, China). Table 1.

Sequence of dsDNAs used in Figures 1–4

Transcription was carried out using 0.6 pmol FAM-labeled dsDNA in a total volume of 29 µl (or its multiples) at 37°C for 1 h in transcription buffer containing 150 mM KCl, 20 U T7 RNA Polymerase (Fermentas, MBI), 0.5 mM GTP and TTP (for C-MYC) or 0.5 mM ATP and GTP (for NRAS and control), in the absence or presence of 40% (m/v) PEG 200. The reaction was stopped by addition of 1/29 vol of 0.5 M EDTA followed by treatment with 0.6 mg/ml Proteinase K (Fermentas, MBI) at 37°C for 1 h.

dsDNA was made in 10 mM Tris–HCl (pH 7.4) buffer containing 1 mM EDTA and the indicated concentration of KCl (or LiCl) and PEG 200, heated at 95°C for 5 min and then cooled down to room temperature at a rate of 0.02°C per second.

Ten picomoles of FAM-labeled dsDNA in 200 µl volume derived from samples that had undergone transcription or heat denaturation/renaturation were mixed with 4 µl of 10% (v/v) dimethyl sulfate (DMS) in ethanol and incubated for 6 min at room temperature. The reaction was stopped by addition of 200 µl stop buffer (0.6 M NaOAc, 0.1 M β-mercaptoethanol, 20 µg sperm DNA). After phenol/chloroform extraction and ethanol precipitation, the DNA was dissolved in 50 µl water. The NTPs in the samples that had undergone transcription were removed by the desalt column mini Quick Spin Oligo Column (Roche, Germany). Thereafter, 50 µl 20% (v/v) piperidine in water was added and the samples were heated at 90°C for 30 min, followed by phenol/chloroform extraction and ethanol precipitation. The precipitated DNA was dissolved in 50% (v/v) deionized formamide in water, denatured at 95°C for 5 min and resolved on a denaturing 12% polyacrylamide gel.

DNA samples were loaded on 8% polyacrylamide gel containing 150 mM KCl, 40% (w/v) PEG 200 and electrophoresed at 4°C, 8 V/cm, in 1X TBE buffer containing 150 mM KCl. For the DNA samples amplified with unlabeled primers, the gel was stained with ethidium bromide (EB) and recorded on a ChemiImager 5500 (Alpha Innotech, San Leandro, CA, USA). For those amplified using FAM-labeled primers, the gel was scanned on a Typhoon phosphor imager (Amersham Biosciences, Sweden) and quantitated with the software ImageQuant 5.2. In some experiments, the DNA samples were incubated with 15 µM T4 Gene 32 single-stranded DNA-binding protein (SSB) (NEB, USA) on ice for 1 h before electrophoresis.

Ten picomoles of FAM-labeled dsDNA in 90 µl 10 mM Tris–HCl (pH 7.4) buffer containing 150 mM K⁺ solution and 40% (w/v) PEG 200 were treated as described in the heat denaturation/renaturation section. Thereafter, 10 µl of 20 U mung bean nuclease (MBN) or 50 U S1 nuclease (Takara, China) were added. After incubation at 37°C for 5 min, the reaction was subjected to phenol/chloroform extraction and ethanol precipitation. The precipitated DNA was dissolved in 50% (v/v) deionized formamide in water, denatured at 95°C for 5 min and resolved on a denaturing 12% polyacrylamide gel.

Oligonucleotides, 5′-GGGAGGGGCGGGTCTGGG-3′ (F-NRAS-T), 5′-CCGCACGCACCGCTCG-CCTTTTGGCGAGCGGTGCGTGCGG-3′ (F-rNRAS-T), 5′-GGGTTAGGGTTAGGGTTAGGG-3′ (F-TEL-T), and 5′-CACACTCACCTCCACACTCCATTTTTGGAGTGTGGAGGTGAGT-GTG-3′ (F-rTEL-T), labeled at the 5′-end with a fluorescein (FAM) and the 3′-end with a tetramethylrhodamine (TAMRA), respectively, were purchased from Takara Biotech (Dalian, China). The loop region in F-rNRAS-T and F-rTEL-T is underlined. The Thermal melting analyses were carried out in 10 mM lithium cacodylate buffer (pH 7.4) containing either 150 mM KCl or 5 mM KCl/145 mM LiCl and the indicated concentration of PEG 200 as described (37,38) on a Rotor-Gene 2000 Real-time Cycler.

Using overlap PCR and genomic DNA from HeLa cells as template, we first constructed two dsDNAs carrying the core G-rich sequence from the C-MYC and NRAS gene, respectively. The G-rich core sequence was on the nontemplate strand and connected with a flanking promoter sequence for the T7 RNA polymerase at its 5′ side (Table 1). The two dsDNAs were subjected to heat denaturation/renaturation or transcription with T7 RNA polymerase. To mimic the intracellular environment, the treatments were carried out at neutral pH in 150 mM K⁺ solution containing 40% (w/v) PEG 200 or no PEG (Figure 1). G-quadruplex formation was then detected by the accessibility of the N7 of guanines to DMS. The N7 in DNA duplex is prone to methylation by DMS and subsequent cleavage with piperidine, but protected by the Hoogsteen bonding in the G-quartet of G-quadruplex structure. The two dsDNAs were labeled at the 5′-end of the G-rich strand with a fluorescent dye FAM. Cleavage fragments were resolved by denaturing gel electrophoresis. In Figure 1, distinct bands corresponding to the cleavage of the four runs of guanines marked with a circle in the core G-rich sequence can be clearly seen for the dsDNAs that were subjected to transcription or heat denaturation/renaturation in the absence of PEG. However, these bands become protected from cleavage when the treatment was carried out in the presence of 40% (w/v) PEG, indicating that in these DNAs G-quadruplex composed of three G-quartets formed during the process of RNA transcription and heat denaturation/renaturation. In contrast, the guanines in the flanking sequences were always similarly attacked no matter whether the samples were processed in the presence or absence of PEG indicating these sequences were in the duplex form. Figure 1.

G-quadruplex formation in long dsDNA carrying G-quadruplex-forming sequence from the C-MYC, NRAS gene examined by DMS footprinting. The N7 (arrowed in the structures on the left side) is not protected from chemical attack in the G-C pair but is in the G-quartet. DNAs labeled at the 5′-end of the G-rich strand with a FAM (asterisk) were subjected either to RNA transcription (T) followed by protease treatment or heat denaturation/denaturation (H) in 150 mM K⁺ solution in the absence (–) or presence (+) of 40% (w/v) PEG 200 before footprinting. Filled circles beside the sequences indicate the guanines that are fully protected, open circles indicate the guanines that are partially protected in the G-quartets; diamonds indicate the unprotected guanines in the duplex region. To increase the resolution, some small bands were allowed to run out of the gel.

DMS footprinting is intuitive for identifying G-quadruplex structure, but difficult to quantify the amount of G-quadruplex formed. We anticipated that when the G-quadruplex forms, the structural change should alter the electrophoretic migration of a dsDNA. To test this possibility, the two dsDNAs and one control dsDNA without G-quadruplex-forming sequence was examined (Figure 2). The dsDNA with G-quadruplex-forming sequence from C-MYC or NRAS without prior heat treatment or heated in the absence of PEG appeared as a single band on the gel. A new major slow band emerged if they were heat treated in the presence of 40% (w/v) PEG 200. According to the DMS footprinting results, the new band should present the dsDNA containing G-quadruplex. In the two dsDNAs that had undergone transcription in the presence of PEG, two new slow bands appeared. The lower major band should be the G-quadruplex-containing dsDNA since it had a similar migration rate as the new band in the heat-treated dsDNA in PEG solution. The slowest band disappeared when the samples were treated with protease after transcription suggesting that it was a DNA/polymerase complex. The quantification by digital scan revealed that this band carried G-quadruplex since its disappearance intensified the G-quadruplex-containing band beneath it. When the transcription was carried out in the absence of PEG, a weak G-quadruplex-containing band was detected in the C-MYC DNA, but not in the NRAS DNA. This weak band did not appear in the heat-treated C-MYC DNA. A possible interpretation could be that the transcribed RNA bound to the C-rich template enhanced the competition of the C-MYC G-quadruplex against duplex. This enhancement was not seen in the NRAS DNA probably because the NRAS G-quadruplex was too weak to compete with duplex even with the transcribed RNA. For the control dsDNA, that did not contain G-quadruplex-forming sequence, very fainter bands were also detected in the samples treated with T7 polymerase. These bands all disappeared if the samples were treated with protease before electrophoresis, suggesting that they were DNA/polymerase complexes. Their amount was lower than that in the C-MYC and NRAS DNA. It is possible that formation of G-quadruplex facilitated the association of the polymerase with the DNAs. However, the quantity of such bands in the C-MYC and NRAS DNA did not correlate with the quantity of G-quadruplexes, suggesting the complex formation may also depend on sequence. By supplying only two NTPs, the transcription was allowed to proceed for 8 nt. This ensured homogeneous transcription product and sharp band for the DNA carrying G-quadruplex. When full-length transcription was allowed, the DNA band containing G-quadruplex became extremely smeary due to the heterogeneity in the progress of transcription (data not shown). Figure 2.

G-quadruplex formation in long dsDNA carrying G-quadruplex-forming sequence from the C-MYC, NRAS gene examined by native gel electrophoresis. DNAs, including the control that does not form G-quadruplex, were labeled at the 5′-end of the G-rich strand with a FAM and subjected either to RNA transcription or heat denaturation/denaturation as described in Figure 1. The samples marked as transcription negative were treated in the same way as the transcription positive samples except that no NTP was supplied. The drawing at the right side shows schematic illustration of the DNA structure associated with the indicated DNA band. The graph beneath the C-MYC and NRAS gel is the digital scan of the indicated lane to quantitate the percentage of each DNA band.

The results presented in the above section demonstrate that the gel electrophoresis can provide a simple method to identify and quantify G-quadruplex formation in long dsDNA. Using this method, we next studied the competition between G-quadruplex and duplex formation and how it is affected by K⁺ and PEG concentration in 15 long dsDNAs derived from human genome by PCR (Table 2). In these dsDNAs, the G-rich sequences were flanked by non-G-rich sequences at both ends. The core telomere sequence (T₂AG₃)₄ was also made into double-stranded form flanked by non-G-rich sequences at both ends by overlap PCR as a reference since it is the most extensively studied G-quadruplex sequence. G-quadruplex formation was induced by the heat denaturation/renaturation treatment since it is a pure physical process and is expected to more faithfully reflect the competition between the two DNA structures themselves without potential influence from protein binding in transcription. Figure 5 shows the gels of these dsDNAs stained by EB. G-quadruplex formation varied significantly in the 15 dsDNAs as judged from the ratio of the two DNA bands in the heat-treated sample in each gel. For example, the majority of the AAVS1 and BCL2 DNA formed G-quadruplex, while the GUK1, ASS1 and telomere DNA were largely in duplex form. Figure 5.

G-quadruplex formation in 16 long dsDNAs derived from human genome examined by native gel electrophoresis. DNAs in 150 mM K⁺, 40% (w/v) PEG 200 solution were subjected to heat denaturation/renaturation (right lane) or were not heat-treated (left lane) before electrophoresis. The gel was stained with ethidium bromide (EB). The drawing at the right side shows schematic illustration of the structure associated with the corresponding DNA band.