Introduction

Amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD) are devastating neurological diseases. ALS is the most frequent motor neuron disease, caused by the loss of motor neurons in brain and spinal cord, resulting in progressive muscle weakness and paralysis, ultimately leading to death from respiratory failure1,2. FTD is the second most common form of dementia in individuals younger than 65 years3. It is characterized by changes in personality or language impairment, due to the progressive degeneration of the frontal and anterior temporal lobes of the brain4. ALS and FTD represent a clinicopathological spectrum of neurological diseases5, which are characterized by the cytoplasmic aggregates of a nuclear protein TAR DNA-binding protein (TDP-43)6,7.

Recently, large expansion of a GGGGCC (G4C2) hexanucleotide repeat in the first intron of the C9orf72 gene has been demonstrated to cause ALS and FTD8,9,10. The healthy control population carries fewer than 25 repeats of this hexanucleotide11, whereas the FTD and ALS patients carry very large expansions ranging between 700 and 1600 repeats8,9. The accumulation of transcripts containing the G4C2 hexanucleotide repeat expansion (HRE) as nuclear RNA foci is detectable in patient cells8,11,12,13,14, which may sequester and inactivate RNA binding proteins15. Additionally, the transcribed expanded repeats can undergo repeat-associated non-ATG initiated translation resulting in accumulation of a series of dipeptide repeat proteins16,17,18.

As recently reviewed, both C9orf72 HRE DNA and RNA may contribute to the pathogenesis of ALS/FTD diseases through mechanisms associated with their structure polymorphisms19. The C9orf72 HRE DNA and RNA enable the formation of complex structures including G-quadruplexes15,20,21. G-rich DNA and RNA sequences can form stable four-stranded structures known as G-quadruplexes, in which four guanine bases are connected through Hoogsteen hydrogen bonding to form a square planar structure or guanine tetrad. Two or more guanine tetrads stack to form a G-quadruplex, which is stabilized by monovalent cations22. G-quadruplexes form intra-molecular or inter-molecular structures from one, two or four strands of nucleic acids23. Defined by the orientation of the strand loops, G-quadruplexes adopt parallel, antiparallel, or mixed topology23,24,25. Genome-wide search has identified 376,000 potential quadruplex sequences in the human genome26. Recent advances in in vivo detection of G-quadruplex by structure-specific antibody and structure-binding ligands provide convincing evidence for the in vivo existence of G-quadruplex structures at telomeres and in human chromosomes27. DNA G-quadruplexes are enriched at telomeres and are also found in promoter regions of a number of oncogenes28,29. Due to the functional relationship between telomeres, oncogenes and cancer, great efforts have been devoted to find potential ligands that target G-quadruplexes to be applied in anticancer therapies30,31.

Recent research has been mainly focused on the structures of C9orf72 HRE RNA, which form highly thermodynamically stable parallel G-quadruplex. Depending on the repeat lengths, G4C2 RNA repeats can form unimolecular and multimolecular G-quadruplexes in vitro21. Nucleolin, the most abundant non-ribosomal proteins found in the nucleolus32, was observed in C9orf72 HRE RNA pull-downs and shown to be mislocalized in the nucleoli of patients carrying C9orf72 HRE15. Additionally, as was demonstrated in vitro, nucleolin preferentially recognized the RNA G-quadruplex, rather than RNA sense or antisense hairpins15. The C9orf72 HRE DNA G-quadruplex has also been investigated in several studies. By circular dichroism (CD) spectroscopy, the C9orf72 HRE DNA were shown to form parallel (e.g. d(G4C2)G4), antiparallel (e.g. d(G4C2)4) or mixed G-quadruplexes (e.g. d(G4C2)3 and d(G4C2)5)15,33.

Till now, no detailed structure information of the C9orf72 HRE G-quadruplex has been obtained, which is required for better understanding the formation of C9orf72 HRE G-quadruplexes and the molecular mechanism underlying C9orf72 HRE-caused ALS/FTD. Here we investigated the structures of C9orf72 HRE DNA G-quadruplexes by CD and NMR spectroscopy and determined the topology of d(G4C2)4, which forms a chair-type G-quadruplex with a four-layer antiparallel G-tetra core and three edgewise loops.

Results

C9orf72 HRE DNAs form G-quadruplex structures

In order to study the structural heterogeneity of C9orf72 HRE DNA G-quadruplexes, we first screened DNA sequences with various d(G4C2) repeats for candidate DNAs that form G-quadruplexes and have sufficient quality for structural determination. We examined the secondary structures of these DNA samples using CD spectroscopy (Fig. 1a). The CD spectrum of d(G4C2)G4 in K+ solution displayed a negative and a positive peak at 240 nm and 260 nm, respectively, indicating a parallel G-quadruplex fold23. The d(G4C2)4 DNA in K+ solution showed a spectrum of predominantly antiparallel G-quadruplexes, which displayed a positive absorption peak at 295 nm and a negative peak at 260 nm23. The CD spectra of d(G4C2)2, d(G4C2)3 and d(G4C2)5 indicated a mixed form of parallel and anti-parallel G-quadruplex folds, which was consistent with the previous report15,33. The thermal stabilities of these five DNA samples were examined using CD melting experiments (Supplementary Fig. S1). The melting profiles of d(G4C2)G4, d(G4C2)2 and d(G4C2)3 showed no obvious transitions, suggesting that these DNA samples did not form stable G-quadruplex structures. In contrast, the melting curves of d(G4C2)4 and d(G4C2)5 exhibited apparent transitions, demonstrating that these two DNA samples formed stable G-quadruplex structures. The melting temperatures of d(G4C2)4 and d(G4C2)5 were around 90 °C and 75 °C, respectively, indicating that d(G4C2)4 formed the most stable G-quadruplex structure.

Figure 1
figure 1

C9orf72 HRE G4C2 DNAs form complex G-quadruplex structures.

(a) CD spectra of d(G4C2)G4, d(G4C2)2, d(G4C2)3, d(G4C2)4 and d(G4C2)5. (b) The imino region of 1D 1 H spectra of various DNA samples.

Full size image

To determine the feasibility for pursuing further structural studies, 1D 1H NMR spectra of these DNA samples were also examined. The d(G4C2)G4, d(G4C2)2, d(G4C2)3 and d(G4C2)5 DNA samples appeared to form a mixture of multiple G-quadruplex conformations (Fig. 1b). Encouragingly, the 1D 1H NMR spectrum of the d(G4C2)4 DNA exhibited 16 well-resolved imino proton resonances at 10–12 ppm with narrow line widths (Fig. 1b), clearly demonstrating the formation of a predominant unimolecular G-quadruplex fold. The NMR melting experiment showed that the melting temperature of d(G4C2)4 was higher than 50 °C, which was consistent with the result obtained in the CD melting experiment (Supplementary Fig. S2).

To probe the molecular size of these DNAs, we performed the native polyacrylamide gel electrophoresis (PAGE) and employed DNA oligonucleotides dT21 and dT42 as size indicators (Supplementary Fig. S3). The d(G4C2)4 DNA migrated as a single band which was slower than dT21, but faster than dT42, suggesting that it existed as a homogeneous monomer. Other DNAs migrated as several bands indicating inhomogeneity in the sample composition and conformation. Because a homogeneous sample is required for high-resolution NMR structural study, we focused on d(G4C2)4 for further structural investigation.

NMR assignment of guanines of d(G4C2)4 sequence

The presence of 16 imino peaks in the 1D 1H spectrum of d(G4C2)4 in K+ solution indicates that all 16 guanines were involved in the intramolecular G-quadruplex formation and this G-quadruplex structure contained four layers of G-tetrads. In order to unambiguously assign guanine imino (H1) and aromatic (H8) proton resonances, low-enrichment 15N, 13C site-specific labeled oligonucleotides were synthesized34. The 1D 15N-filtered heteronuclear single quantum coherence (HSQC) spectra were recorded for these oligonucleotides (Fig. 2). One signal was observed in each of the spectra for the imino protons of the isotopically-labeled guanine bases, which demonstrated the single conformation of the d(G4C2)4 G-quadruplex in solution and was consistent with the PAGE result of d(G4C2)4 (Supplementary Fig. S3). Identified imino signals of 1D 1H-NMR spectra were shown in Fig. 2. The signals around 10–11 ppm are likely corresponding to imino protons of guanine residues involved in Hoogsteen hydrogen bonds in a minor conformation of d(G4C2)4.

Figure 2
figure 2

Imino proton (H1) assignments from 15N-filtered experiments.

The imino region of 1D 1H NMR spectra with the assignment of guanine bases in d(G4C2)4 being indicated over the reference spectrum on top. Guanine imino protons were assigned using 1D 15N-filtered HSQC spectra of samples containing site-specific low-enrichment (2%) 15N-labelled oligonucleotides at the indicated positions.

Full size image

The 2D 13C-1H HSQC spectra were also recorded for these oligonucleotides and aromatic proton assignments were obtained (Supplementary Fig. S4). These identified H1 and H8 signals were further confirmed by 13C-1H heteronuclear multiple bond correlation (HMBC) (Supplementary Fig. S5) and nuclear overhauser effect spectroscopy (NOESY) experiments25,35. The H1′ proton resonances of d(G4C2)4 were identified with the use of 2D NOESY spectra through correlations with H6/H8 protons of the same base. The 2D 13C-1H HSQC spectra for low-enrichment 15N, 13C site-specific labeled oligonucleotides also provided the H1′ proton resonances for cross-validation.

d(G4C2)4 adopts an antiparallel chair type G4 topology

Based on the proton assignments obtained as described above, we analyzed the NOESY spectrum (75 ms mixing time) and COSY spectrum. The peak intensities of G1, G3, G7, G9, G13, G15, G19 and G21 were measured through non-linear fitting with the use of nmrDraw36 and gauged by the H5-H6 NOE peak intensities of cytosines to confirm the syn conformation of these guanosines (Supplementary Fig. S6). The remaining eight G-quadruplex guanines, namely G2, G4, G8, G10, G14, G16, G20 and G22, adopted the anti-conformation.

In the current study, the characteristic synG(i)H1′/antiG(i + 1)H8 and synG(i)H8/antiG(i + 1)H1′ NOEs were observed, including synG1-antiG2, synG3-antiG4, synG7-antiG8, synG9-antiG10, synG13-antiG14, synG15-antiG16, synG19-antiG20 and synG21-antiG22 (Supplementary Fig. S7). The assignments of the imino H1 and base H8 protons of guanines in 2D NOESY spectrum (mixing time 300 ms) resulted in the direct determination of the folding topology of the G-quadruplex structure formed by d(G4C2)4 in K+ solution (Fig. 3a). In a G-tetrad plane with a Hoogsteen-type H-bond network, the imino proton H1 of a guanine was in close spatial vicinity to the base H8 of one of the adjacent guanines. Analysis of characteristic NOEs between H1 and H8 protons revealed the formation of an intramolecular G-quadruplex involving four G-tetrads: G1•G10•G13•G22, G2•G9•G14•G21, G3•G8•G15•G20 and G4•G7•G16•G19 (Fig. 3b–d). To resolve the ambiguity of the H8 chemical shifts of G4 and G16 and that of G7 and G19, we made a T substitution at C17 in the loop region, which was close to G16 and G19. The mutant C17T exhibited an excellent 1D 1H spectrum which closely resembled the wild type spectrum, indicating that the T substitution did not disrupt the wild type fold (Supplementary Fig. S8). Meanwhile, the peaks of G7 and G19 were clearly distinguished in the C17T spectrum. The 2D NOESY spectrum of C17T with 500 ms mixing time was also recorded. The characteristic NOE patterns of H1-H8 and H8-H1′ regions were consistent with the wild type while G4H8-H1′ and G16H8-H1′ were displayed as two individual peaks (Supplementary Fig. S9). Based on the characteristic NOEs between imino protons (G3H1-G7H1 and G15H1-G19H1) and the G(i)H2′/H2″-G(i + 1)H8 pattern, the G4H8 and G16H8 were unambiguously assigned (Supplementary Fig. S9). Due to the overlapped H8 chemical shifts of G7/G19 and G9/G21 and close H1′ chemical shifts of G8/G20 and G10/G22 in the H8-H1′ region of NOESY spectrum, synG(i)/antiG(i + 1) connectivities of synG7-antiG8, synG9-antiG10, synG19-antiG20 and synG21-antiG22 were further verified through the G(i)H2′/H2″-G(i + 1)H8 pattern based on the NOESY spectrum of C17T (Supplementary Fig. S9). Combined with the characteristic NOEs between H1 and H8 protons of C17T, we determined that the hydrogen-bond directionalities of the four G-tetrads were clockwise, anti-clockwise, clockwise and anti-clockwise, respectively. The glycosidic conformations of guanines around the tetrads were syn•anti•syn•anti. The G-tetrad core was of the antiparallel-type, in which each G-tetrad was oriented in opposite direction with respect to its neighboring ones (Fig. 3d).

Figure 3
figure 3

Topology determination for the C9orf72 HRE DNA d(G4C2)4.

(a) The NOESY spectrum (300 ms mixing time) showing H1–H8 connectivity. (b) Characteristic guanine H1–H8 NOE connectivity patterns around a Gα•Gβ•Gγ•Gδ tetrad as indicated by arrows. (c) Guanine H1–H8 NOE connectivities observed for the G1•G10•G13•G22, G2•G9•G14•G21, G3•G8•G15•G20 and G4•G7•G16•G19 tetrads. (d) Topology of d(G4C2)4. Anti guanines are colored in cyan, while syn guanines are colored in magenta. The backbones of the core and loops are colored in black and red, respectively.

Full size image

Discussion

Recent studies have shown that the C9orf72 HRE DNA can form G-quadruplex structures, which was demonstrated mainly by CD spectroscopy15,33. C9orf72 HRE DNAs with different repeat lengths adopted parallel, anti-parallel or mixed conformation. Sket et al. also obtained the 1D 1H NMR spectrum of the C9orf72 HRE DNA, which indicated a G-quadruplex fold, but they failed to obtain homogeneous samples for 3D structure determination33.

In this study, we also employed CD and NMR spectroscopy to investigate the G-quadruplex formation by C9orf72 HRE DNA. First, C9orf72 HRE DNAs with various repeat lengths were screened for homogeneous samples that allow for structural determination by NMR. After the screening, d(G4C2)4 that adopted a single conformation was further pursued for structural investigations. By a combination of NMR approaches, we determined the topology of d(G4C2)4, which represents an antiparallel conformation adopted by the C9orf72 HRE DNA. This DNA forms a chair-type G-quadruplex with a four-layer antiparallel G-tetra core and three edgewise loops. Although the chair-type topology for G-quadruplexes has been identified for a Bombyx mori telomeric DNA37 and a thrombin aptamer DNA38, the loop arrangements of these DNAs and d(G4C2)4 differ from each other in the length and sequence. The G-quadruplex of the Bombyx mori telomeric sequence, d[TAGG(TTAGG)3], contains a two-layer antipaprallel G-tetrad core and three edgewise loops, in which one loop (TTA) spans across a narrow groove and the other two loops (TTA/TTA) span wide grooves. The G-quadruplex of the thrombin aptamer DNA, d(GGTTGGTGTGGTTGG), also contains a two-layer antipaprallel G-tetrad core and three edgewise loops, in which one loop (TGT) spans a wide groove and the two remaining loops (TT/TT) span narrow grooves. In contrast, in our reported topology of d(G4C2)4, three edgewise loops are all composed of CC, which span both wide and narrow grooves.

C9orf72 HRE DNA/RNA can fold into G-quadruplex structures and form transcriptionally induced R-loops15,21,33. These structural polymorphisms lead to abortive transcripts, which may act as fundamental determinants of the pathogenesis of C9orf72 HRE-linked ALS/FTD. Therefore, detailed structure information of C9orf72 HRE is in high demand for understanding their structural diversity and the molecular mechanism underlying C9orf72 HRE-caused pathologies. Here we studied the structures of C9orf72 HRE DNA with short repeat lengths and demonstrated the structural variations of these G-quadruplex DNAs in vitro. The long expanded DNA repeats remain to be investigated for the structural features of these nucleic acids in vivo. Additionally, while d(G4C2)4 exhibited an antiparallel conformation, G4C2 HRE DNAs of other repeat lengths were found to adopt parallel or mixed conformation. Due to the inhomogeneity of the DNA samples, they were not applicable for further NMR investigations. Although not a focus in this study, some modification tools, like 8-bromo guanine substitution, may be used to improve the homogeneity of the samples to facilitate 3D structure determination of the parallel conformation adopted by the G4C2 HRE DNA.

Various small molecules have been employed to target G-quadruplex-forming sequences, to modulate telomere maintenance or gene activity31,39. As multimerization of C9orf72 HRE G-quadruplexes may exacerbate the formation of RNA foci and cause neuron-toxicity in ALS/FTD15, disruption of these foci by targeting C9orf72 HRE G-quadruplexes may have beneficial outcomes in the treatment of these diseases. Therefore, this study may also build the necessary basis for designing small molecules which can therapeutically target C9orf72 HRE to modulate ALS/FTD pathogenesis.

Methods

Sample preparation

Unlabeled and site-specific low-enrichment (2% 15N labeled or 15N, 13C-labeled) DNA oligonucleotides were chemically synthesized and HPLC purified (Takara, China). DNA was annealed at the 0.1 mM single stranded concentration by heating to 95 °C for 15 min, followed with slow cooling to room temperature overnight in the annealing buffer containing 70 mM KCl and 20 mM potassium phosphate (pH 7.0). The NMR sample was concentrated by ultrafiltration with the use of a Centricon 3kD (Millipore, MA) in 20 mM potassium phosphate buffer (pH 7.0) and 70 mM KCl. The final concentration of the DNA sample was 2.5 mM.

Circular dichroism

Circular dichroism spectra were recorded at 25 °C on a JASCO-815 CD spectropolarimeter using a 1 mm path length quartz cuvette with a reaction volume of 200 μl. DNA concentration was 20 μM. The DNA oligonucleotides were prepared in a 20 mM potassium phosphate buffer (pH 7.0) containing 70 mM KCl. The CD melting experiments were performed by measuring the absorbance at a single wavelength (260 nm for parallel G-quadruplexes and 295 nm for antiparallel G-quadruplexes) with a temperature range of 25 °C to 95 °C.

Polyacrylamide gel electrophoresis (PAGE)

Non-denaturing PAGE were carried out on a 18% polyacrylamide gel (acrylamide:bis-acrylamide 29:1), supplemented with 20 mM KCl in the gel and running buffer (0.5X TBE). The samples were prepared at a strand concentration of 100 μM. Bands were revealed by red-safe staining.

NMR spectroscopy

Experiments were performed on 500 MHz, 750 MHz and 800 MHz Varian spectrometers. Resonances were assigned unambiguously using samples with site-specific low-enrichment 15N or 15N, 13C labeling and through-bond correlations at natural abundance. Standard 2D NMR experimental spectra including NOESY, TOCSY and COSY, were collected at 25 °C to obtain the complete proton resonance assignments. The NMR experiments for samples in water solution were performed with Watergate or Jump-and-Return water suppression techniques. The variable temperature spectra were recorded on 500 MHz Varian spectrometer from 5 °C to 50 °C. All spectra were processed by the program NMRPipe. Spectral assignments were also carried out and supported by COSY, TOCSY and NOESY spectra. Peak assignments were achieved using the software Sparky.

Additional Information

How to cite this article: Zhou, B. et al. Topology of a G-quadruplex DNA formed by C9orf72 hexanucleotide repeats associated with ALS and FTD. Sci. Rep. 5, 16673; doi: 10.1038/srep16673 (2015).