Berberine Molecular Recognition of the Parallel MYC G-Quadruplex in Solution

Abstract

The medicinal natural product berberine is one of the most actively studied and pursued G-quadruplex (G4)-ligands. The major G-quadruplex formed in the promoter region of the MYC oncogene (MycG4) is an attractive drug target and a prominent example and model structure for parallel G-quadruplexes. G4-targeted berberine derivatives have been actively developed; however, the analogue design was based on a previous crystal structure in which berberine binds as a dimer to a parallel G-quadruplex. Herein, we show that in solution, the binding mode and stoichiometry of berberine are substantially different from the crystal structure: berberine binds as a monomer to MycG4 using a base-recruitment mechanism with a reversed orientation in that the positively charged convex side is actually positioned above the tetrad center. Our structure provides a physiologically relevant basis for the future structure-based rational design of G4-targeted berberine derivatives, and this study demonstrates that it is crucial to validate the ligand–DNA interactions.

Graphical Abstract

INTRODUCTION

Nature is a trove of chemically diverse compounds that are potential drug candidates. A well-known example is the medicinal natural product berberine, a protoberberine alkaloid (Figure 1C). Berberine has a long history of use in traditional Chinese medicine, and its anticancer activity is related to its binding to DNA.¹ Berberine and derivatives bind G-quadruplex (G4) DNA structures with higher affinity than duplex DNA.^2–4 As such, the G4-binding of berberine has been extensively studied. Berberine has an asymmetric scaffold of four fused rings with a methylenedioxy group at one end (ring A) (Figure 1C). It contains a concave side and a convex side, with the convex side bearing a central quaternary nitrogen that introduces a (+1) positive charge mostly located on this side of the molecule.⁵ Berberine has been shown to bind human telomeric and promoter G-quadruplexes of human oncogenes including MYC.^6,7 Structural characterization of the G4 binding of berberine largely focused on human telomeric G-quadruplexes.^2,8 In 2013, a crystal structure of a berberine–G4 complex was reported, in which berberine binds as a dimer to a dimeric parallel telomeric G-quadruplex with a 6:2 binding stoichiometry.⁹ Berberines bind as a dimer with their concave sides in the center at both 5′- and 3′-ends. In 2017, a paper by Moraca et al. in PNAS using computational simulation also suggested a dimer binding mode.¹⁰ Subsequently, the crystal structure and the dimer binding mode have been actively used as a basis to design G4-targeted berberine derivatives.^11–14

Figure 1.

(A) Sequence of the major MYC promoter G4. (B) Schematic structure of the MycG4. (C) Structure of berberine. (D) Berberine stacking over the external G-tetrad with its convex side in the center based on the observed NOE interactions between Ber-H6 and the imino protons of all four external-tetrad-guanines.

DNA G-quadruplexes are four-stranded globular secondary structures formed in guanine-rich regions of the human genome with functional importance and are considered attractive drug targets. They consist of stacked G-tetrads of four guanines connected via Hoogsteen hydrogen bonds with coordination of monovalent cations, mostly potassium or sodium which are physiologically relevant.¹⁵ The stacked G-tetrads are connected by different types of loop and flanking motifs. Parallel G-quadruplexes are prevalent in the promoter regions of oncogenes as transcription regulators, and their targeting by small molecules is a promising strategy for cancer therapeutics.^16–18 A prominent example and model structure for parallel G-quadruplexes is the major G-quadruplex formed in the promoter region of the MYC oncogene (MycG4, Figure 1A/B).^19,20 The MYC protein is overexpressed in cancer cells as a central player in tumorigenesis, and the MYC promoter G-quadruplex is a transcription silencer that can be targeted by small-molecule drugs.²¹ Berberine and derivatives have been shown to bind MYC G4 and can downregulate the MYC expression.^6,13,22 However, the structural details of berberine recognition have not been determined. Herein, we combine nuclear magnetic resonance (NMR) spectroscopy and mass spectrometry (MS) to gain structural insight into the berberine recognition of the parallel MycG4 in solution, which is the physiologically relevant condition. Our results show that the binding mode of berberine to parallel G4s in solution is substantially different from the previous crystal structure and should be used to guide the future design of berberine derivatives.

RESULTS AND DISCUSSION

Berberine Binds MycG4 with a 2:1 Stoichiometry.

We first investigated the binding of berberine to MycG4 by 1D ¹H NMR titration, which showed notable changes of the imino proton resonances upon berberine addition (Figure 2A). The resonances were broadened at a 1:1 ligand–DNA ratio indicating a binding interaction in intermediate-to-fast exchange on the NMR timescale. Further addition up to a 3:1 berberine–MycG4 ratio led to a well-resolved spectrum with 12 sharp signals.

Figure 2.

(A) 1D NMR spectra of the MycG4 (1.2 mM) imino region titrated with berberine in the presence of 10 mM K⁺, pH 7, 25 °C. The NMR spectra were measured with an 800 MHz instrument. The imino protons of the 5′-, central, and 3′-tetrads are labeled in blue, black, and red, respectively. The imino protons of the berberine–MycG4 complex are labeled at 3:1 and 4:1 ligand–DNA ratios. (B) Negative ESI mass spectra of MycG4 (5 μM) after incubation with different ratios of berberine. The highlighted peaks represent the G4 complexes with 0, 1, 2, or 3 berberines.

We next determined the berberine binding stoichiometry to MycG4 by MS. MS observes single complexes and thereby provides insight into the exact binding stoichiometry.²³ Mass spectra at increasing berberine–MycG4 ratio revealed the 2:1 complex as major species, indicating high-affinity binding of two berberine molecules (Figure 2B). The apparent dissociation constant for this dominating binding is 9.9 ± 1.0 μM based on a fluorescence-based binding assay (Figure S1). A third binding site is visible at a 3:1 ligand–DNA ratio; however, its population is very small even at a 5:1 ligand–DNA ratio, indicating a much lower binding affinity. Collectively, the data show that there are two high-affinity specific binding sites of berberine to MycG4, that is, a 2:1 stoichiometry.

Structure Determination of the 2:1 Berberine–MycG4 Complex by NMR Spectroscopy Shows That Berberine Binds as a Monomer at the 5′- and 3′- Ends.

In light of the good spectral quality of the 1D NMR data (Figure 2A), 2D nuclear Overhauser effect spectroscopy (NOESY) and ¹H–¹³C heteronuclear single quantum coherence (HSQC) spectra were recorded to determine the berberine–MycG4 complex structure (Figure S2/S3). The DNA and drug resonances were completely assigned using standard strategies mostly following sequential contacts in the NOESY spectrum in combination with the HSQC experiment which confirmed the residue types and anti glycosidic conformation (Figure S2/S3 and Tables S1/S2).²⁴ Berberine binding does not change the folding of the MycG4, as shown by the similar nuclear Overhauser effect (NOE) cross peak pattern observed for the G-core of the free and bound MycG4.²⁰ Both the 5′- and 3′-flanking residues rearrange upon berberine binding according to the NOE cross peaks. Numerous intermolecular NOE cross peaks provide detailed information about the interactions of berberine with MycG4 (Figures 3 and S4).

Figure 3.

Intermolecular NOE contacts between MycG4 and berberine involving the DNA (A) imino and (B) aromatic protons. Measured with 1.2 mM DNA at a berberine-MycG4 ratio of 3:1 and a mixing time of 300 ms in the presence of 10 mM K⁺, pH 7, 35 °C. The NMR spectra were measured with an 800 MHz instrument. The intermolecular NOEs used for structure determination are boxed in red, blue, or black for conformation (A,B) or both, respectively.

Based on the observed intermolecular NOEs, berberine stacks over the external tetrads at both the 5′- and 3′-ends with its convex side in the center, in contrast to the crystal structure in which berberine’s concave side is in the center (Figure S5).⁹ Clear cross peaks (medium) were observed between the Ber-H6 protons, located at the center of berberine’s convex side, and all guanine imino protons of the two external tetrads, that is, G16, G7, G20, and G11 of the 5′-tetrad and G13, G9, G22, and G18 of the 3′-tetrad (Figures 1D and 3A). Notable cross peaks were also observed between the Ber H5/H8 protons on the berberine’s convex side and guanine imino protons of the two external tetrads. On the other hand, no or only weak cross peaks between G-tetrad imino protons and the berberine’s concave side, that is, the Ber-H11/H12/H13/H1 protons, were observed (Figure S4). Therefore, berberine’s convex side is in the center when binding the parallel MycG4 in solution, reversed from the orientation in the crystal structure of the parallel telomeric G-quadruplex.⁹

For both the 5′ and 3′ binding sites, continued stacking was observed for the flanking sequences, that is, A6/G5/T4, and G23/A24/A25, respectively (Figure S5 and Tables S3/S4). Sequential NOE interactions of G7H8 with A6H8 and the A6 H1′, H2′, H2″, and H3′ sugar protons support a continued stacking of A6 upon G7 and the 5′-tetrad. Similar sequential NOE interactions are observed between G23 and G22 in line with continued stacking of G23 upon the 3′-tetrad. A6H2 showed cross peaks to G20H1 and G7H1 of the 5′-tetrad, positioning A6 as stacked on the 5′-tetrad with its Watson–Crick edge facing the tetrad center (Figure S6D). Furthermore, intermolecular cross peaks were observed between A6H2 and BerH5, BerH6, BerMeA, and BerMeB (Figure 3B). These NOE interactions indicate the formation of a ligand–base joint plane as a major binding mode of berberine, in which its convex side faces the recruited flanking residue A6 or G23 at either the 5′-end or 3′-end, respectively.

Interestingly, additional intermolecular cross peaks between berberine and MycG4 were observed, suggesting th coexistence of more than one berberine orientation at each binding site (Figure 3B, Table S5). For example, clear cross peaks are observed for BerMeA with the G11 and G20 H8 protons of the 5′-tetrad, with the NOE of G20H8 being stronger. These cross peaks should be mutually exclusive due to their location on the opposite sides of the tetrad and thereby point at the coexistence of two different berberine orientations, with the D-ring pointing either toward G11 or the G20-G16 edge. Similar cross peaks of the coexisting orientations are observed at the 3′-tetrad for BerMeA with H8 protons of G9 and G13, with the NOE of G13H8 being stronger. Such coexistence of different ligand conformations is seldom described but not surprising for ligand binding to parallel G-quadruplexes with short loops.^20,25 The wide binding pockets are only formed by flanking sequences and external tetrads, so that small molecules can potentially stack in different orientations.

Numerous intramolecular NOEs were observed for the MycG4 in the berberine complex (Figure S2/S3, Tables S3/S4) that clearly define the overall G-quadruplex structure and the ligand binding sites. To take into account the two coexisting berberine orientations at each binding site, we grouped the intermolecular NOE cross peaks that clearly belonged to one or both berberine complexes (Figure 3 and S4 and Table S5). The structures of the berberine–MycG4 complex in K⁺-containing solution were calculated using NOE-restrained molecular dynamics simulations as described previously (Table 1; Figures 4 and S7).²⁰ The final 10 lowest energy structures of each complex are shown in Figure S7 with a well-converged G-core (Table 1). Note that there are four possible combinations of conformers A and B at each binding site (5′- and 3′-ends) for the berberine–MycG4 complex structure, although we only show two combinations in Figure 4. Intermolecular NOE restraints clearly define the position and orientation of the berberine at each binding site in both conformers, albeit the coexistence of two orientations leads to a lower number and less accurate NOE constraints for each complex. Consequently, the flanking residues at both ends are less well-defined and contribute to a larger root-mean-square deviation (rmsd) (Table 1).

Table 1.

Summary of Restraints Used in the Structure Calculation and Structural Statistics for the Ensembles of the 10 Lowest-Energy 2:1 Berberine–MycG4 Complex Structures

	conformer A	conformer B
NOE-Based Distance Restraints
intra-residue	306	306
inter-residue sequential	128	128
long-range	25	25
Ber-G4	22	19
Other Restraints
hydrogen bonds	48	48
torsion angles	22	22
G-tetrad planarity	36	36
Structural Statistics Pairwise Heavy Atom rmsd (Å)
G-tetrad core	0.52 ± 0.16	0.44 ± 0.10
core + Ber	0.54 ± 0.15	0.50 ± 0.10
5′-site (flanking, Ber, 5′-tetrad)	0.63 ± 0.25	1.28 ± 0.71
3′-site (flanking, Ber, 3′-tetrad)	1.39 ± 0.61	1.28 ± 0.61
all violations	1.23 ± 0.34	1.46 ± 0.34
mean NOE restraint violation (Å)	0.001 ± 0.012	0.001 ± 0.008
max NOE restraint violation (Å)	0.133	0.233

Figure 4.

Conformer A (PDB ID: 7N7D) and conformer B (PDB ID: 7N7E) of the berberine–MycG4 complex and respective 5′- and 3′-binding sites in solution.

G-Quadruplex Binding Mode of Berberine in Solution is Substantially Different from the Crystalline State.

In contrast to the dimer binding mode in the berberine–G-quadruplex crystal structure,⁹ the solution structure of the berberine–MycG4 complex shows that berberine binds as a monomer at both 5′- and 3′-ends (Figure 4). Furthermore, berberine stacks on the external tetrad with its positively charged convex side above the center, instead of at the outer edge of the tetrad as seen in the crystal structure (Figure S5). Berberine recruits the first flanking residue A6 or G23 to form a ligand-base joint plane covering the 5′- or 3′-external tetrad, respectively. Two different orientations (A and B) of berberine are found at each binding site. In both binding sites, the positively charged convex side of berberine is central and faces the recruited flanking base, whereas the concave side points outside (Figure 4). For the 5′-binding site, berberine stacks on G11 and G16, with its D-ring on top of G16 in the major conformation (A), as suggested by the stronger cross peak observed for BerMeA with G20H8 (conformer A) than with G11H8 (conformer B) (Figure 3B). For the 3′ binding site, berberine stacks mostly on G9 and G13. The D-ring is on top of G13 in conformation A, whereas it stacks on top of G9 in conformation B, as suggested by the stronger cross peak with G13H8 than G9H8 observed for BerMeA.

While berberine predominantly binds as a monomer to the MycG4 in solution, additional berberine binding was suggested by MS and NMR spectroscopic data at higher ligand–DNA ratios, however, only at the 3′-end of the MycG4. In the NMR titration data, continued addition of berberine further shifted the imino proton signals with a selective broadening of the resonances found in the 3′-external tetrad, suggesting an additional binding at the 3′-end (Figure 2A). Weak berberine–berberine cross peaks were observed in the NOESY spectra with a short 80 ms mixing time for H1–H11 and MeB-HC that are at the opposite ends of berberine (Figure S8), implying inter-berberine interactions. These inter-berberine NOEs suggest that a minor population of berberine dimer binds the 3′-end at higher ligand ratios, with berberine’s concave sides facing each other in a head-to-tail orientation (Figure S5). Importantly, the MS spectra show that even at 5 equivs of the ligand, this minor species of 3:1 berberine–MycG4 complex is barely populated (Figure 2B).

Insights into the Rational Design of G4-Targeting Berberine Derivatives.

Our new solution structure of berberine in complex with MycG4 provides a physiologically relevant molecular basis for the structure-based rational design of G4-targeted berberine derivatives. Many berberine derivatives have been designed and synthesized based on the previous crystal structure which does not represent the ligand recognition under physiologically relevant solution conditions. Our results put previous development efforts in a new perspective and suggest a change in strategy. For example, in the crystal structure, C13 on berberine’s concave site is positioned centrally above the tetrad. Thus, position 13 has commonly been substituted with bulky aromatic moieties, for example, phenylalkyl, diphenylalkyl, dimethylaminophenyl, or pyridine containing groups, to improve stacking interactions or avoid the berberine dimerization.^11,12,14,26 However, in the solution structure of the berberine–G-quadruplex complex, C13 is located at the edge of the tetrad close to the negatively charged quadruplex groove. Therefore, the more flexible and often cationic alkyl side chains^27–29 could be better suited for the berberine concave side, such as position 13 as well as positions 1, 11, and 12, to facilitate backbone interactions and improve binding affinity.

In addition, covalently connected berberine dimers have been designed to cover the full tetrad to mimic the dimeric binding observed in the crystal structure.¹³ Our solution structure may help to improve selectivity. Berberine binds by recruiting the adjacent flanking residue with its convex side facing the DNA base to form a joint ligand–base plane covering the external tetrad. This base-recruitment recognition mechanism has been observed for many G4–ligands.^20,30–32 Based on the new solution structure, a potential strategy to improve binding selectivity could be functionalizing berberine’s convex site to enable hydrogen-bond interactions that are specific for the recruited DNA base. For example, the recruited A6 of the MycG4 has its 6-amino group as a hydrogen-bond donor and N1, N3, or N7 as hydrogen-bond acceptors. Judicial substituents of berberine C4–C8 or O9, which have been derivatized before,^14,26,33 may facilitate hydrogen-bond interactions with the recruited adenine. Similar strategies can be used to target G23 at the 3′ binding site of MycG4 or other bases if alternative G-quadruplexes are targeted. It is noted that in a recent study with a RET promoter G-quadruplex that berberine only binds at the 3′-end without ligand-flanking interactions,³⁴ likely because the flanking regions important for ligand interactions^20,35 were truncated.

CONCLUSIONS

In conclusion, our solution study of the MycG4–berberine complex structure provides a new and physiologically relevant recognition mode of the medicinal natural product berberine to the prevalent parallel G-quadruplexes. Significantly, the recognition mode of berberine in solution is substantially different from that observed in the previous crystal structure. Using a base-recruiting recognition mechanism, berberine binds at both ends of the MycG4 as a monomer instead of a dimer with its positively charged convex side over the center of the stacked tetrad instead of at the outer edge. Many previous berberine analogues have been based on the crystal structure and focused on the berberine concave side for substituents to either enhance the tetrad stacking or reduce dimer interaction, which may need to be reconsidered under solution conditions. Instead, appropriate modifications of berberine’s concave side based on the solution structure could be used for DNA backbone interactions, whereas modifications on the convex side might facilitate specific interactions with the recruited base and lead to better sequence selectivity. Therefore, our study provides an important structural basis for the future rational design of G4-targeted berberine derivatives.

Structural information about the recognition of a receptor by a small molecule is important to understand structure–activity relationships and for the rational design of improved derivatives. It is crucial to validate the results under physiologically relevant solution conditions. X-ray crystallography is an indispensable method in structure biology; however, G-quadruplexes and G4–ligand complexes are challenging and can be significantly influenced by crystal packing due to highly flexible motifs such as capping structures or binding pockets which are formed by dynamic flanking or loop sequences. Moreover, high ligand concentrations are often necessary for crystallization and might lead to occupation of otherwise less-populated binding sites.³⁶

EXPERIMENTAL SECTION

Sample Preparation.

The oligonucleotides were synthesized and purified as described previously.³⁷ The DNA samples were annealed by heating to 95 °C for 5 min and subsequent slow cooling to room temperature. The DNA concentration was quantified based on UV/vis absorption at 260 nm. A 60 mM solution of berberine (Shanghai Standard Technology Co., Ltd., ≥98%) was prepared with DMSO-d₆ (Cambridge Isotope Laboratories, Inc.).

Fluorescence Spectroscopy.

The fluorescence binding assay²⁹ was performed at 20 °C using a Jasco FP-8300 spectrofluorometer (JASCO, Inc.) equipped with a temperature controller. The 3′-Cy5-labeled MycG4 (Sigma-Aldrich) DNA was dissolved in 25 mM potassium phosphate buffer and 75 mM KCl, pH 7 to obtain a 20 nM solution. Berberine was titrated to the solution, and fluorescence spectra between 650 and 675 nm were recorded after every step with an excitation wavelength of 640 nm. For the measurements, a response time of 1 s, a data interval of 0.5 nm, and an excitation as well as emission bandwidth of 5 nm were used. The change in Cy5 fluorescence was plotted over the berberine concentration and fitted to determine the apparent binding affinity K_d,app using the equation ΔF = ΔF_max*[Ber]/(K_d,app + [Ber]), in which ΔF is the actual change in Cy5-fluorescence and ΔF_max is the maximum change in fluorescence.

NMR Spectroscopy.

A 1.2 mM solution of MycG4 DNA was prepared with 10 mM potassium phosphate (Fisher Scientific) buffer, pH 7, 90% H₂O/10% D₂O. The NMR spectra were measured on a Bruker AV-800 MHz spectrometer equipped with a QCI cryoprobe at 25 °C using excitation sculpting for water suppression. All 2D experiments were performed at a berberine–MycG4 ratio of 3:1. The NOESY mixing times were between 80 ms and 300 ms, and additional NOESY experiments were done at 35 °C. ¹H–¹³C HSQ experiments was recorded optimized for a ¹J(C,H) of 190 Hz. Chemical shift referencing was done directly for ¹H based on the water signal relative to 4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS) and indirectly for ¹³C relative to DSS. The spectra were processed with Topspin 3.5 (Bruker) and analyzed with CcpNmr Analysis 2.4.³⁸

Sample Preparation for Native MS.

MycG4 samples were reconstituted in Optima LC/MS grade water (Fisher Scientific), creating a stock solution of 100 μM. The sample underwent desalting three times via centrifugation with Optima LC/MS grade water using a 3 kDa molecular weight cutoff Amicon Ultra 0.5 mL filter (Millipore Sigma). The recovered sample was then diluted back to 100 μM. The solution conditions used for MS analysis comprised of 5 μM MycG4, 0–25 μM berberine (Shanghai Standard Technology Co., Ltd.), 0.25 mM KCl (Fisher Scientific), 37.5 mM trimethylammonium acetate buffer (Santa Cruz Biotechnology), 0.2% dimethyl sulfoxide (DMSO) (Fisher Scientific), and 10% Optima LC/MS grade methanol (Fisher Scientific). Samples for different ratio concentrations of berberine–MycG4 were made with a range of 0–25 μM berberine taken from a stock solution of 1 mM berberine in 4% DMSO due to berberine’s low solubility in water.

Native MS.

All MS experiments shown were performed using a QTOF 5600 (Sciex) spectrometer. Analyte anions were generated via a nano-electrospray ionization (ESI) source by applying a voltage of approximately −1200 V to a platinum wire in contact with the solution in a borosilicate glass emitter. A DDC voltage of 40 V was applied to q₀ to remove excess solvent and salts, and a quadrupolar RF frequency was applied to Q1 to pass a minimum m/z of roughly 1200. The most abundant charge state for the G4 and berberine adduction peaks with the given solution conditions was [G4 – 5H]⁵⁻.

Structure Determination.

The structures of the berberine–MycG4 complexes were determined as described in detail previously.^20,39 Berberine was parameterized for the Amber force field using the R.E.D. server.⁴⁰ Starting structures were calculated with Xplor-NIH 2.48^41,42 using the NOE-based distance restraints, hydrogen bonds restraints for the G-core, dihedral restraints for the glycosidic bonds, and planarity restraints for the G-tetrads. The Amber 16 package with the OL15 version of the Amber force field for DNA was used for the following steps.^43–46 First, a simulated annealing of 100 structures in implicit water was performed, and afterward, the best 20 structures were equilibrated for 4 ns in explicit water. Finally, a short minimization was performed in vacuum. The structures were visualized using the PyMOL software.⁴⁷

ABBREVIATIONS

G4

G-quadruplex

DMSO

dimethyl sulfoxide

DSS

4,4-dimethyl-4-silapentane-1-sulfonic acid

HSQC

heteronuclear single quantum coherence

MS

mass spectrometry

NMR

nuclear magnetic resonance

NOE

nuclear Overhauser effect

NOESY

nuclear Overhauser effect spectroscopy

rmsd

root-mean-square deviation