Exploration of the Structure and Recognition of a G-quadruplex in the her2 Proto-oncogene Promoter and Its Transcriptional Regulation

G-quadruplexes in oncogene promoters provide putative targets for transcriptional regulation. The structure of a putative G-quadruplex sequence (S1: GGAGAAGGAGGAGGTGGAGGAGGAGGG) in potassium solution in the her2 promoter has been resolved mainly through nuclear magnetic resonance (NMR) spectroscopy. By application of various NMR spectra, we proved the formation of a four-layer G-quadruplex composing of two G-tetrads and two G/A-mixed planes with a four-residues loop (A3-G4-A5-A6). Further evidence from a luciferase reporter assay, Q-RT-PCR and Western blotting indicates that S1 G-quadruplex formation can repress her2 promoter activity, and a selected G-quadruplex ligand cβ can enhance the repression by down regulating her2 transcription and expression. These findings provide a G-quadruplex target and perspective implications in her2 transcriptional regulation.


Results and Discussion
The her2 promoter S1 sequence formed an intramolecular parallel G-quadruplex. Circular dichroism (CD) spectroscopy is used to evaluate the formation and directions of strands of S1 G-quadruplex. Numerous studies have shown that parallel G-quadruplexes are characterized by a dominant positive CD absorption around 260 nm, whereas anti-parallel ones have a negative CD absorption at 260 nm and a positive absorption at 290 nm, and both quadruplex types show an additional characteristic positive peak at 210 nm [41][42][43][44] . The CD spectrum of S1 gives a strong positive absorption at 260 nm and another high-intensity positive band at 210 nm in 100 mM potassium solution (Fig. S1a), suggesting the formation of a parallel G-quadruplex induced by potassium ions. It is notable that unlike the CD spectra of canonical parallel G-quadruplexes which display a relatively shallow negative band at 240 nm 45 , Fig. S1a barely shows an absorption at 240 nm. This CD pattern is quite similar as that for the G-quadruplex formed by d(GGA) 8 37 , implying a similar folding topology of S1 with d(GGA) 8 . The formation of S1 G-quadruplex is also supported by ESI mass spectrometry. The mass spectrum of S1 (Mw = 8673.7) in 100 mM ammonium solution reveals a base peak (100%) of a complex ion formed by one molecule of S1 and three ammonium ions (Fig. S1b, [S1 + 3NH 4 + -8H + ] 5− at m/z 1742.9). As ammonium ions have a similar atomic radius to potassium ions and are located between adjacent G-tetrad planes 46,47 , the existence of the complex ion [S1 + 3NH 4 + -8H + ] 5− indicates that S1 forms an intramolecular G-quadruplex with four planes. Furthermore, polymerase stop assays were performed to investigate the S1 G-quadruplex formation in a DNA template strand, and the results demonstrate that S1 G-quadruplex is formed and arrests Taq polymerase in a potassium-dependent manner (Supporting information, Fig. S2).
Determination of guanines participating in G-tetrad core structure. To investigate which guanines participated in the G-tetrad core structure, we performed DMS-induced strand cleavage for S1. As Fig. 2a illustrates, all S1 guanines are distinctly cleaved in pure water or 100 mM lithium solution, suggesting that all guanines are methylated and G-quadruplex is unformed under these conditions. However, in 100 mM potassium solution (30 mM Tris-HCl, pH 7.4), all other sixteen guanines show strong protection from DMS except G4 and G27, indicating that four G-tetrads are formed by these sixteen bases. Further evidence for the G-tetrad composition comes from two mutation strand cleavages. When the fourth S1 guanine (G4) was mutated to thymine (G4T, Fig. 2b), the G-quadruplex structure remains in potassium solution, and all guanines except 3′-terminal G27 are protected against DMS. When the guanine participating in the G-tetrad core structure, such as G1, was mutated to thymine (G1T, Fig. 2c), the G-quadruplex structure collapsed even in potassium solution, and all of the guanines were distinctly cleaved. This result is supported by the NMR spectra of S1-S3 in 100 mM potassium solution (30 mM Tris-HCl, pH 7.4). As shown in Fig. S3, the her2 sequence S1 (WT) and S2 (G4T) show quite similar peaks at 10.2-12.0 ppm which are characteristics of G-tetrad imino protons, while S3 (G1T) shows no detectable peaks in this region. This confirms that the S1 fourth guanine (G4) is excluded from the G-tetrad core and mutation of it barely changes the quadruplex structure, while mutation of those participating in the G-tetrad core, like G1, will destroy the quadruplex structure. Proton assignments through low-level site-specific fully 15 N, 13 C-labelling method. The NMR spectroscopy method was used to resolve the G-quadruplex structure that was formed by her2 S1. Proton chemical shifts were unambiguously assigned through a site-specific low-enrichment (6-8%) 15 N, 13 C-isotope labelling approach 48,49 . The chemical shifts of exchangeable imino and amino protons were gained through 15 N-1 H-HSQC spectra (Fig. 3) and those of non-exchangeable base and sugar protons were gained through 13 C-1 H-HSQC spectra (Supporting information, Fig. S4). The proton assignments were also confirmed by correlations in the TOCSY, NOESY and 31 P-1 H-HSQC spectra. Figure 3 demonstrates the guanine imino proton assignments by 1D-15 N-1 H-HSQC spectra. Notably, the imino proton signals of G4 and G27 were undetectable in this region, indicating www.nature.com/scientificreports www.nature.com/scientificreports/ that they are excluded from the G-tetrad core hoogsteen hydrogen bonds, which is consistent with the DMS footprinting results.
The stacking between the G-tetrads and G/A-mixed planes are shown in Fig. 6d,e,f. The hydrogen-bond directionalities of the four planes are anti-clockwise, anti-clockwise, clockwise and clockwise, respectively. It is worth noticing that the bases of A9, A12, A18, A21 and A24 are stacked on the sugar rings of G8, G11, G17, G20 and G23, respectively, which induces the extreme upfield shift of sugar H4′ of these G residues (2.63-2.80 ppm). However, because A6 is attached with A3-G4-A5 flexible linker, the co-planarity of it with the G1•G7•G10•G13 plane is poor (Fig. 7a), and thus causes a moderate H4′ up-field shift of G2 (3.13 ppm). Also, because G1 does not form sheared base-pair with A6, its amino protons are active enough that they cannot be detected at room temperature (RT), while those of the other G residues (G7, G10, G16, G19 and G22) in the G:A pair can be www.nature.com/scientificreports www.nature.com/scientificreports/ detected even at RT. Figure 6f demonstrates the stacking between the two G/A-mixed planes. A9 and A12 in the upper plane stacks on A21 and A18 in the lower one, respectively, which is consistent with the NOE interactions between aromatic H2 and sugar H between them. A6 is arranged on top of A24, and thus causes the A24H2-A6H1′/H2′/H2′′ and A6H2-A24H1′ correlations (Fig. S7).
The schematic structures of G-quadruplexes formed by her2 S1 sequence and d(GGA) 8 37 were shown in Fig. 7. In d(GGA) 8 , all of the sequential G-G-A motifs participating in the G-tetrads and G/A-mixed planes are similar with a single A residue linking the tetrad and G/A-mixed plane; while in her2 S1 G-quadruplex, except for the single A residue, there is a four-residues (A3-G4-A5-A6) loop which links the first G-tetrad and the second G/A-mixed plane. This result provides further implications of formation of G-quadruplexes with G/A-mixed planes by non-sequential GGA repeats.
Recognition of the her2 S1 G-quadruplex by a flexible cyclic polyamide cβ. As a novel G-quadruplex structure formed by the original G-rich sequence in the her2 promoter, S1 G-quadruplex recognition implies potential importance in the regulation of her2 expression. Here, the ESI-MS method was used to evaluate the binding affinities of a series of cyclic polyamide (cα, cβ, cγ, cPT and cPTN, Supporting Information, Fig. S12) to the her2 S1 G-quadruplex. Figure 8a shows the ESI mass spectrum of a mixture solution of S1 G-quadruplex and cβ in which the base peak is a complex ion composed of one S1 G-quadruplex ion (abbreviated to Q1) and two cβ molecules ([Q1 + 2cβ] 5− , m/z = 1972.2), and the free Q1 ion cannot be detected. While the ESI mass spectra of the mixtures of the other small molecules and the G-quadruplex (Q1) reveal no complex ions but free Q1 (for cα, cγ and cPTN, Supporting Information, Figs S13a,b,d) or an intensity of complex ion of less than 20% (for cPT, Supporting Information, Fig. S13c). These results indicate that out of the five cyclic polyamides, only cβ has high binding affinity to the her2 S1 G-quadruplex. Because the selectivity between the G-quadruplex and duplex DNA is also an important aspect for a G-quadruplex ligand, we generated a mixture of G-quadruplex and its corresponding duplex DNA by mixing S1 and its complementary cytosine-rich strand C1 (Mw = 7890.3) at a molar ratio of 2:1. When four cβ equivalents were added to the mixture, the ESI mass spectrum (Fig. 8b) revealed a base peak of a complex Q1 ion and two cβ molecules ([Q1 + 2cβ] 5-), while no complex duplex DNA ion ([S1 + C1-7H + ] 7-, m/z = 2363.4, abbreviated to [D1] 7-) and cβ were observed, indicating that cβ selectively recognized the her2 G-quadruplex over the duplex DNA. Further evidence from a polymerase stop assay supported this result. When cβ was added to the system, the ratio of the fully-elongated band to the sum of arrested and fully-elongated band of primer extension using the template containing S1 G-quadruplex decreased to 80% (Fig. 8c,d), demonstrating that cβ bound to the her2 S1 G-quadruplex and inhibited the sliding of Taq polymerase along the template.

Regulation of her2 transcription by the G-quadruplex ligand cβ.
To investigate the function of the selected G-quadruplex ligand cβ in regulation of her2 transcription, we performed a luciferase reporter assay experiment according to the protocols described previously 39,52 . First, we evaluated the function of the S1 G-quadruplex in the her2 promoter by constructing two plasmids including wildtype (WT) and a mutant (MU) sequence of S1 (Fig. 9a). Based on our previous study, the WT sequence formed an S1 G-quadruplex structure, and the MU sequence destroys the G-quadruplex. As illustrated in Fig. 9b, the relative luciferase activity of the MU plasmid increased by nearly 40% compared with that of the WT plasmid, indicating that the formation of the S1 G-quadruplex can inhibit her2 promoter activity, while its destruction (MU) up-regulates the activity. We then evaluated the influence of cβ on her2 transcription by adding 0-800 nm cβ to the luciferase reporter assay system. As illustrated in Fig. 9c, relative luciferase activity of the her2 plasmid decreased gradually to less than 20% when the cβ concentration increased to 800 nM, indicating inhibition on her2 promoter activity by cβ. This result was supported by further RT-PCR and Western blotting experiments. The relative expression level of her2 mRNA gradually decreased to nearly 30% after MCF-7 cells were treated with 0-100 μM cβ (Fig. 9d), www.nature.com/scientificreports www.nature.com/scientificreports/ demonstrating that cβ can inhibit her2 transcription. Western blotting experiments indicated that the relative her2 protein expression level reduced to 70% or 20% when 10 or 50 μM cβ along with S1 DNA as a carrier was added to cells (sample in Fig. 9e), while carrier DNA without cβ had no influence on her2 expression (control in Fig. 9e). In summary, our study demonstrates a synthetic small molecule cβ, which can selectively bind to the her2 promoter S1 G-quadruplex and inhibit her2 proto-oncogene transcription and expression.

Conclusion
To conclude, our work revealed a G-quadruplex structure formed by an original G-rich sequence (S1: GGAGAAGGAGGAGGTGGAGGAGGAGGG) in the her2 proto-oncogene promoter. The A3-G4-A5-A6 loop in the S1 G-quadruplex provides further evidence of four-layer G-quadruplex formation with G/A-mixed planes by non-sequential GGA repeats. We also demonstrated that the S1 G-quadruplex can inhibit her2 promoter activity, and a selected S1 G-quadruplex ligand cβ can suppress her2 transcription and expression. Since the elevated expression of her2 has been shown to be strongly associated with the development and progression of aggressive breast cancers, our study provides a putative G-quadruplex target for breast cancer treatment. Further efforts may be made on the study of the her2 S1 G-quadruplex ligands, either bind to the tetrad core or the loops, to stabilize the G-quadruplex and hence repress the her2 oncogene expression for related pathologies regulation.

Material and Methods
Sample preparation. Unlabeled oligonucleotides were purchased from Sangon Biotech. (Beijing, China).
Site-specific labelled ( 15 N and 13 C) oligonucleotides were synthesized by Takara Biotech. (Dalian, China) using nucleotides labelled with 15 N and 13 C from Cambridge Isotope Laboratories, Inc. (MA, USA). Small molecules (cα, cβ, cγ, cPT and cPTN) were synthesized in our lab 53,54 . DNA samples for NMR spectroscopy were purified by successive dialysis against 0.5 M KCl solution and then against water [55][56][57][58] . For a stable G-quadruplex solution generation, purified S1 was diluted in K + buffer containing 75 mM KCl, 25 mM K 2 HPO 4 -KH 2 PO 4 (pH 7.0) at a final concentration of 1.0 mM strands for NMR spectra scanning. NMR Spectroscopy. NMR spectra were recorded on Bruker Advance 800 MHz, and Varian VNMRS 600 MHz spectrometers equipped with cryogenic probes at 25 °C 57,59-64 . Resonances were assigned unambiguously by 1D-15 N-1 H-HSQC and 1D-13 C-1 H-HSQC experiments using a site-specific low-enrichment (6-8%) fully 15 N, 13 C-labeling approach 48,49 . Assignments were confirmed through standard 2D NMR experiments including TOCSY (τ m of 120 ms) and NOESY (τ m of 50, 200 and 250 ms). Spectra in water solutions were performed Figure 8. ESI mass spectra of 40 μM cβ with (a) 10 μM S1 G-quadruplex (marked Q1) and (b) 10 μM Q1 with equal doses of its corresponding duplex DNA (marked D1), which demonstrate high affinity and selectivity of cβ to Q1 with respect to D1. (c) Polymerase stop assay showing the primer extension band in a system with or without cβ. The marker P indicates primer, and S indicates arrested stop band. M1 and M2 represent two marker bands at 35 nt and 57 nt, respectively. (d) The relative intensity ratio of fully-elongated to the sum of fully-elongated and arrested bands decreased to 80% in the presence of cβ. Data represents the average of two replicates. The full-length gel is presented in Supplementary Fig. S14. www.nature.com/scientificreports www.nature.com/scientificreports/ with water suppression using watergate W5 pulse sequence with gradients. Relaxation delays were set to 2 s. The acquired sizes were set to 1024 × 320, and the final spectral sizes were 1024 × 1024. 2,2-dimethylsilapentane-5-s ulfonic acid (DSS) was used as an internal standard for the 1 H chemical shifts assignments.
The NMR structure of the her2 promoter S1 sequence was calculated using the Simulated Annealing protocol of the Crystallography & NMR system (CNS) version 1.3 program 65,66 . Briefly, an initial random conformation of S1 sequence is generated and imposed to simulated annealing using experimentally measured interproton distance estimates, hydrogen bonds and dihedral angle restraints. In total, 100 trial structures were calculated with 100 cycles of final minimization, and 74 structures were accepted based on violations. A planarity restraint of 10 kcal•mole −1 •Å −2 was imposed on the G-tetrads during the annealing. Ten final structures were selected out of the 74 accepted ones based on their minimal energy terms and number of NOE violations. The structures were viewed using the UCSF Chimera program 67 .

ESI Mass Spectrometry (ESI-MS). ESI-MS experiments were performed using the Finnigan LCQ Deca
XP Plus ion-trap mass spectrometer (CA, USA). S1 (10 μM) was dissolved in 100 mM NH 4 OAc solution (pH 6.9) and pre-incubated at 37 °C for 8 h to generate the G-quadruplex. Each of the synthetic small molecule ligands were dissolved in 1:9 (v/v) DMSO/CH3OH solution to a stock concentration of 500 μM and added to the DNA samples for binding affinity investigation. Duplex DNA were generated by mixing 10 μM S1 and the equivalent of its complementary cytosine-rich strand (designated C1). For each sample, 25% (v/v) methanol was added to improve electrospray efficiency. Negative ion mode was used with a spray voltage of 2.7 kv and a capillary temperature of 140 °C. The infusion rate was 2 μl/min, and scans lasting 5 min were summed for each spectrum. www.nature.com/scientificreports www.nature.com/scientificreports/ Circular Dichroism (CD) Spectroscopy. CD experiments were performed with a JASCO J-815 CD spectrometer (Tokyo, Japan). S1 (10 μM) was dissolved in 30 mM Tris-HCl (pH 7.4) buffer with 100 mM or without KCl. Samples were pre-annealed at 90 °C for 10 min and then slowly cooled down to 25 °C (more than 8 h) before being tested in a 0.1 cm path-length cuvette. The scanning wavelength range was from 200 to 400 nm, and three times scans were averaged for each spectrum.
3′-6-FAM labelled oligonucleotides (S1-S3, Supporting Information Table S1) were diluted to 0.1 μM in 100 μl 30 mM Tris-HCl (pH 7.4) in the presence of H 2 O, 100 mM LiCl or KCl. DNA samples were then annealed at 90 °C for 10 min and slowly cooled down to 4 °C (more than 8 h) before being treated with 10% (v/v) DMS for 1.5 min at 25 °C. The reaction was quenched by adding 100 μl stop solution (20 μl β-mercaptoethanol, 40 μl 3 M sodium acetate, 50 μg sperm DNA). The DMS-treated DNA samples were extracted with a Tris-phenol-chloroform solution (pH 8.0) and then precipitated using ethanol at −80 °C before being cleaved by 10% (v/v) piperidine at 90 °C for 30 min. The cleaved DNA samples were separated on a 20% sequencing gel at 1500 v for 4 h in a 4 °C cold room and imaged using a GE Healthcare Typhoon9400 gel scanner (CT, USA).

Polymerase Stop Assays.
Polymerase stop assays were performed as follows 39,52,69 . Briefly, the her2 promoter S1 sequence was inserted into a general template and annealed with a 5′-6-FAM labelled primer (sequences listed in supporting information Table S1) in a PCR buffer with 0-150 mM KCl at 95 °C for 5 min before being slowly cooled down to 25 °C over 8 h. The general template without S1 sequence was used as a control. To verify the binding properties of the small molecule ligand, 0 or 10 doses of cβ were added to the template-primer system in the presence of 150 mM KCl. Afterward, Taq polymerase (BBI, MA, USA) was added to conduct primer extension at 37 °C for 15 min. The reaction was quenched on ice, and products were separated on a 12% denaturing gel before imaging using a GE Healthcare Typhoon9400 gel scanner (CT, USA).
Luciferase Assay. The luciferase assay experiments were performed as follows 39,52 . An approximately 500 bp sized fragment of the human her2 promoter (from −414 to +78) including the S1 sequence was generated via PCR from Hela cells and inserted into a pGL4.10-basic vector (Promega, WI, USA) for construction of the wild-type (WT) plasmid. The mutant plasmid (MU) containing S3 instead of S1 was also constructed as described, and both plasmids were verified via automated sequencing.
The cultured MCF-7 cells were plated at a density of 1 × 10 5 cells/well into a 24-well plate and incubated at 37 °C overnight. To verify the influence of the mutation on promoter activity, 400 ng WT or MU plasmids were transfected into MCF-7 cells using Lipofectamine 2000 (Invitrogen, CA, USA) as a transfection reagent. p-RL-TK vector (40 ng) was co-transfected as an internal control. To verify the effect of the small molecule ligand on her2 promoter activity, 0-800 nM cβ was mixed with the WT plasmid and transfected into MCF-7 cells. Luciferase activities were measured using a dual luciferase reporter assay system (Promega, WI, USA) according to the manufacturer′s protocol. Relative luciferase activities were acquired by normalizing the ratio of firefly luciferase activity to renilla luciferase activity of the mutant construct with that of the wild-type construct.
Q-RT-PCR. Q-RT-PCR experiments were performed as described previously 52 . Cultured MCF-7 cells were split into a 6-well plate at a density of 8 × 10 5 cells/well and incubated at 37 °C overnight. Powdered cβ was dissolved in DMSO to generate a 10 mM stock solution. A mixture of this stock solution and PEI (Mw = 25000, linear) at a mass ratio of 1:0.8 was generated and diluted with complete growth medium to form a working solution (0-100 μM), which was used to treat MCF-7 cells at 37 °C overnight. Cells were harvested, and total RNA was extracted using Trireagent (Invitrogen, CA, USA) according to the manufacturer′s instructions. TransScript II All-in-One-First-Strand cDNA SuperMix kits (TransGen, Beijing, China) were used to generate cDNA. Forward and reverse primers (Supporting information, Table S1) were added to the cDNA solution for the real-time PCR reaction using SYBR Green PCR Master Mix (Applied Biosystems, CA, USA), and GAPDH was used as a control. The reaction was performed in an Eppendorf Realplex Real-time PCR system (Hamburg, Germany).

Western Blotting.
Western blotting experiments were performed as described previously 52 . Cultured Hela cells were split into a 6-well plate (6 × 10 5 cells/well) and incubated at 37 °C overnight. Powdered cβ was dissolved in DMSO to generate a 10 mM stock solution and mixed with her2 S1 DNA solution (0.5 mM) at a volume ratio of 4:1 to generate a cβ-DNA mixture before annealing at 90 °C. A serial volume of this cβ-DNA mixture was diluted using complete growth medium to a working concentration of 0-50 μM for cβ. Lipofectamine 2000 (Invitrogen, CA, USA) was used as a transfection reagent, and DMSO was the control. The samples were harvested using a general procedure and then loaded on a 10% SDS gel and transferred to a nitrocellulose membrane at 45 v for 18 h in a 4 °C cold room. The samples were then blocked with 5% non-fat milk at room temperature for 1 h and probed with specific primary antibodies at 4 °C overnight. Target proteins were visualized via treatment with HRP-linked secondary antibodies.