A Synthetic Pur-based Peptide Binds and Alters G-quadruplex Secondary Structure Present in the Expanded RNA Repeat of C9orf72 ALS/FTD

Abstract

Increased Pur-alpha (Pura) protein levels in animal models alleviate certain cellular symptoms of the disease spectrum amyotrophic lateral sclerosis/frontotemporal dementia (ALS/FTD). Pura is a member of the Pur family of evolutionarily conserved guanine-rich polynucleotide binding proteins containing a repeated signature PUR domain of 60 – 80 amino acids. Here we have employed a synthetic peptide, TZIP, similar to a Pur domain, but with sequence alterations based on a consensus of evolutionarily conserved Pur family binding domains and having an added transporter sequence. A major familial form of ALS/FTD, C9orf72 (C9), is due to a hexanucleotide repeat expansion (HRE) of (GGGGCC), a Pur binding element. We show by circular dichroism that RNA oligonucleotides containing this purine-rich sequence consist largely of parallel G-quadruplexes. TZIP peptide binds this repeat sequence in both DNA and RNA. It binds the RNA element, including the G-quadruplexes, with a high degree of specificity versus a random oligonucleotide. In addition, TZIP binds both linear and G-quadruplex repeat RNA to form higher order G-quadruplex secondary structures. This change in conformational form by Pur-based peptide represents a new mechanism for regulating G quadruplex secondary structure within the C9 repeat. TZIP modulation of C9 RNA structural configuration may alter interaction of the complex with other proteins. This Pur-based mechanism provides new targets for therapy, and it may help to explain Pura alleviation of certain cellular pathological aspects of ALS/FTD.

1. Introduction

Pura (Pur-alpha) is a polynucleotide binding protein with a preference for a purine-rich sequence element in DNA or RNA [1–4]. Pura is a member of the Pur protein family, each of which possesses one to three copies of a signature Pur domain of approximately 60 amino acids [5]. This domain is highly conserved throughout evolution, and it includes the nucleic acid binding portion of the proteins [5, 6]. Pura preferentially binds single strands of DNA or RNA at an element comprising repeated short sequences of G residues separated by 1–3 non-G residues [1–3, 7]. Pura prefers the binding element (G_2–4N_1–3)_n, N is not G (Daniel and Johnson, 2018). Pur binding elements are present at many origins of DNA replication, including one upstream of the c-MYC gene in humans [1], and those of polyomaviruses [8]. Pura binds the human telomeric DNA repeat, (TTAGGG)_n [7]. Pura also binds the fragile-X repeat, (CGG)_n [9], the expansion of which is linked to the human genetic brain disease, fragile-X syndrome (FXS) and the related fragile-X ataxia (FXTAS). Here we have characterized a newly synthesized peptide, TZIP, representing a generic Pur domain, regarding its ability to bind single-stranded DNA and RNA etiologically linked to a neurological disease spectrum. This disease spectrum encompasses the C9orf72 familial form of amyotrophic lateral sclerosis (ALS) and related frontotemporal dementia (FTD). TZIP closely mimics the third signature domain of Pura. This third domain is implicated in binding both polynucleotide [6, 7] and protein [5, 10–14]. We show that TZIP binds with moderately tight affinity to the RNA hexanucleotide repeat sequence of the C9orf72 form of amyotrophic lateral sclerosis and frontotemporal dementia (ALS/FTD). This hexanucleotide repeat expansion (HRE), (GGGGCC)_n, is prone to formation of G-quadruplexes [15–19]. We show by circular dichroism that these structures are parallel G-quadruplexes, and we document the effects of TZIP binding upon them.

The C9orf72 form (C9) of ALS/FTD is the most common familial form and also accounts for varying percentages of sporadic cases among populations. The C9 form presents a spectrum of cellular, cognitive and movement CNS pathologies with brain disorders in FTD prominent at one end of the spectrum and neuromuscular disorders in ALS prominent at the other [20–22]. The C9 form provides a window for study of processes leading to pathology in all types of ALS/FTD. Binding of Pura to RNA has been implicated in these processes [23–25]. An association of Pura with RNA transcripts in neurons had previously been noted [26–28]. There is increasing evidence that Pura activity counteracts neuronal degeneration in cellular models of ALS [25, 29]. Overexpression of Pura in mouse neuronal cells and in a Drosophila model mitigates C9 repeat RNA-mediated neurodegeneration. That study also showed that Pura was the major protein in mouse spinal cord lysates that bound to the C9 repeat DNA [25]. Beneficial effects of Pura in ALS/FTD have not been limited to the C9 form. In cells expressing an ALS-causing mutant of FUS, ectopic Pura expression blocked cytoplasmic mislocalization of mutant FUS protein and strongly reduced toxicity of the mutant protein in primary neurons [29]. Cellular shRNA knockdown of Pura inhibited formation of cytoplasmic RNA stress granules containing FUS and Pura. These reports indicate that interaction of Pura with transcribed RNA is a common beneficial theme in mitigating cellular pathology of different forms of ALS/FTD.

Despite the documented importance of Pura binding to the HRE RNA in C9 ALS/FTD, little is known about the binding mechanism or structural changes induced. Several consequences of protein binding to newly transcribed HRE RNA may lead to neurodegeneration. Mutations in certain ALS-linked RNA binding proteins lead to dysfunctional autophagy, which contributes to increases in RNA-protein aggregates, such as stress granules characteristic of ALS/FTD pathology [30–32]. Alternatively, the HRE RNA can be translated, in the direction of the positive sense RNA and its complement, by an unusual repeat-associated non-ATG (RAN) translation mechanism to yield dipeptide repeats (DPR) [33, 34]. Depending on their sequence, these DPR lead to varying degrees of neurotoxicity [35, 36]. Recent data now suggests that the C9 repeat RNA may itself be neurotoxic independent of DPR formation. Pura mitigated RNA toxicity, but not DPR toxicity in a zebrafish embryo model [37]. Although the mitigation by Pura is most likely due to its nucleic acid binding capabilities, little is known about the interaction between Pura and the expanded repeat RNA.

Here we show that a Pur-based TZIP peptide, binds both complementary strands of C9 repeat RNA and that TZIP preferentially binds parallel G-quadruplexes formed by the G-rich strand. By interacting with these structures, TZIP promotes their incorporation into GGGGCC-repeat RNA higher order structures. These results help provide a basis for understanding the potential beneficial activities of Pura in mitigating ALS/FTD pathology due to the C9 HRE and highlight potential benefits of developing small C9-HRE-interacting peptides.

2. Material and Methods

2.1. Preparation of recombinant GST-Pura and custom peptide preparation

The generation of the human GST-Pura construct, pGPUR4, and its propagation has previously been described [11]. The GST-fusion protein was recovered on glutathione-agarose beads (Amersham Pharmacia Biotech, Inc.) following the method of Smith and Johnson [38]. TZIP peptide was custom synthesized and HPLC purified by Biomatik Corporation. HPLC was carried out in 100% water by the company preparing the peptide, and this solvent was recommended. After the lyophilized peptide was dissolved in water at 1 mg/ml, the concentration was confirmed using a NanoDrop spectrophotometer (Thermo Fisher Scientific). TZIP has is a net positive charge of +11, and dissolves in 100% water with brief mixing. The concentrations used in the experiments were dilutions in the micromolar range.

2.2. Oligonucleotides

Fluorescently labelled DNA and RNA oligonucleotides were synthesized by Integrated DNA Technologies, Inc. The oligonucleotides were 5′-tagged as indicated with IRDye 700 or IRDye 800 (LI-COR Biosciences, Inc.) and imaged with a LI-COR Odyssey Infrared imaging System in the 700 nm (red) and 800 nm (green) channels respectively. To protect against RNA degradation, all RNA sequences were synthesized with a 2′-O-methyl modification. They are as listed:

ALS/FTD-DNA-5′-IRD700-T(G₄C₂)₄T-3′

FXS-DNA-5′-IRD800-T(CGG)₈T-3′

ALS/FTD-RNA-5′-IRD700-AA(G₄C₂)₄-3′

ALS/FTD-RNA-5′-IRD700-CC(G₄C₂)₄-3′

ALS/FTD-RNA-5′-IRD800-CC(G₄C₂)₄-3′

ALS/FTD-RNA-5′-CC(G₄C₂)₄-3′

ALS/FTD-RNA-5′-IRD700-CC(G₄C₂)₆-3′

ALS/FTD-RNA-5′-IRD700-CC(G₄C₂)₁₀-3′

ALS/FTD-RNA-5′-IRD700-U(G₄C₂)₄U-3′

ALS/FTD-RNA-5′-IRD800-A(C₄G₂)₄A-3′

Random RNA-5′-IRD800-AGCGCUAAAAGUCCAUAGCACGUGCA-3′

2.3. Circular dichroism of C9 HRE oligonucleotides and TZIP peptide

Circular dichroism data were obtained with a Jasco J-815 CD spectrometer. Buffer for all samples was 20 mM KPO₄, 70 mM KCl aqueous buffer at pH 7. A buffer blank was subtracted from all spectra. For thermal stability studies, RNA oligonucleotides were at a concentration of 22 μM or 2.2 μM. Spectra were recorded from 340 to 220 nm. A single scan was recorded for the 22 μM sample, with a scan speed of 20 nm/min. For signal-averaging purposes, three scans were recorded and combined for the 2.2 μM sample, with a scan speed of 50 nm/min. Thermal stability was assessed by recording spectra from 25 to 95°C at 10°C increments using a Jasco peltier temperature controller. A three-minute equilibration time was used at each temperature. CD spectra of the TZIP peptide were obtained using a peptide concentration of 10 μM. For CD spectral analysis of the TZIP:RNA interactions, the spectrum of the RNA alone at a concentration of 2.5 μM was first obtained, with a scan speed of 50 nm/min at 30 °C with three scans recorded and averaged for each sample. Next, 2.5 μM RNA was incubated with either 2.5 μM TZIP (1:1 molar ratio) or 10 μM TZIP (1:4 molar ratio) at 30 °C for at least thirty minutes. The CD spectra were again recorded with a scan speed of 50 nm/min at 30 °C with three scans recorded and averaged for each sample.

2.4. Electromobility shift assay (EMSA)

EMSA was performed as previously published [1]. Briefly, binding reactions containing stated concentrations of fluorescently labelled oligonucleotide were incubated in 9 mM HEPES (pH 7.9), 45 mM KCl, 0.25 mM dithiothreitol, 0.5 μg poly(dI-dC), 4.5 μg BSA, and 4% sucrose with stated amounts of protein or peptide in a final volume of 20 μl, for 30 min at 30° C. Electrophoresis was conducted using 4% acrylamide/bis-acrylamide (29:1) gels in TBE buffer. Gel lanes were scanned and analyzed as described by Sambrook [39] using the Odyssey CLx Infrared Imaging System and Image Studio (LI-COR Biosciences). Apparent binding affinities were determined as equilibrium dissociation constants [39] using GraphPad Prism 4 equations for nonlinear regression. The concentrations of oligonucleotides and protein or peptide are provided in the figure legends.

2.5. Renaturation of RNA oligonucleotides to form G-quadruplexes as discrete, slowly migrating bands and their separation using non-denaturing polyacrylamide gel electrophoresis

RNA G-quadruplexes were formed following the procedure of Zhou et al., 2015 [19]. Fluorescently tagged RNA oligonucleotides, in 20 mM KPO4 (pH 7.0), 70 mM KCl, at concentrations of 5 to 15 μM, were denatured at 95^o for 5 min and slowly cooled overnight under conditions previously shown to promote formation of G-quadruplexes [19]. Products were separated on 10% non-denaturing PAGE (acrylamide/bis-acrylamide, 29:1) gels in 0.5X TBE buffer supplemented with 20 mM KCl and scanned with the Odyssey Infrared Imaging System, as described above.

3. Results

3.1. Cellular protein Pura binds the guanine-rich hexanucleotide repeat sequence DNA and RNA of the C9orf72 form of ALS/FTD as well as the FXS/FXTAS DNA repeat sequence

Recent reports indicate that the cellular protein, Pura, exerts a beneficial effect on ALS-associated neurotoxicity in animal models and cells [9, 25, 29, 37]. We observed binding of Pura (322 aa) to C9orf72 repeat sequence single-stranded DNA and RNA using polyacrylamide gel electrophoretic mobility shift assay (EMSA). In Fig. 1 we show that Pura binds the C9 DNA repeat sequence represented by T(G₄C₂)₄T and the C9 RNA repeat sequence represented by AA(G₄C₂)₄. Likewise, Pura also binds the FXS DNA repeat sequence represented as T(CG₂)₈T. The binding of Pura to the basic CG₂ repeat has been shown using other models [9, 40]. These guanine-rich repeat sequences are canonical Pur binding elements [5, 7]. The fastest-migrating band of the G-rich repeat DNA and RNA in Fig. 1 is the unbound form of linear single-stranded oligonucleotide as indicated in lanes 5 and 5′ (heat denatured DNA or RNA). The band migrating above the linear band is the basic G-quadruplex unit. Pura binds and shifts the repeat sequence DNA and RNA to slower-migrating bands, some of which extend to the top of the gel (lanes 2, 3, 2′,3′). These bands are protein-bound oligonucleotides, and may be other structural forms derived from the oligonucleotides. In contrast, addition of GST protein to the oligonucleotide solution does not lead to these additional forms.

Fig 1. Pura binds guanine-rich DNA and RNA oligonucleotides that represent repeat sequences expanded in ALS/FTD C9orf72 as well as the DNA triplet repeat expanded in FXS.

Fluorescently tagged oligonucleotides as indicated above the images were incubated with recombinant GST-Pura or GST protein and subjected to polyacrylamide gel EMSA. (A) ALS (IRD700) and FXS (IRD-800) repeat sequence DNA or (B) ALS repeat sequence RNA oligonucleotides were incubated in binding mixtures containing: lanes (1, 1′) no additional protein; (2, 2′) 200 nM Pura; (3, 3′) 500 nM Pura; (4, 4′) 500 nM GST; (5, 5′) oligonucleotides with no added protein were heat denatured prior to loading. Oligonucleotide concentrations were 1 nM throughout. The two different labels were imaged in the red (700 nm) and green (800 nm) channels respectively.

3.2. Altered secondary structures of the G-rich C9orf72 repeat sequence involve Hoogsteen base pairing in guanine quadruplexes

It has been repeatedly demonstrated that the C9 ALS/FTD hexanucleotide repeat in either single-stranded DNA or RNA can form stable G-quadruplexes [15–19, 41, 42]. These are stacked structures of G-quartets formed by G-G Hoogsteen hydrogen bonding [43]. The Hoogsteen bond, an alternative to Watson-Crick base pairing, involves the N-7 hydrogen as a donor and the C-4 amino group as an acceptor. Other than G-quadruplexes, the other well-characterized structure within the C9 repeat would be a DNA triple helix. It would result in amorphous structures rather than discrete quadruplex quanta. Although the G-quadruplex is the most well-characterized structure formed by the C9 repeat sequence, and although the basic unit requires 4 hexanucleotide repeats, higher numbers of repeats can form multiple G-quadruplexes [17]. It is not yet known to what extent G-quadruplexes in vivo can alter the structure of the expanded G-rich repeats that cause neurological diseases. A protein-DNA complex can harbor Hoogsteen base pairing and specific protein binding can shift the configuration of the resulting structure [44].

3.21. The C9orf72 repeat sequence, represented by CC(GGGGCC)₄ RNA, forms parallel-oriented G-quadruplexes with high thermal stability and higher orders of repeat multiplicity

Circular dichroism (CD) spectroscopy can be used to identify G-quadruplex structures in DNA and RNA [45–50]. As discussed in section 3.2, properties of G₄C₂ quadruplexes have been presented by others [15–19, 41, 42] To assess whether CC(GGGGCC)₄ RNA that we are using to represent the C9orf72 hexanucleotide repeat forms G-quadruplex structure, CD spectra were obtained at 25°C in 20 mM KPO₄ (pH7.0), 70 mM KCl buffer at 22 μM RNA (Fig. 2A, purple line). The observed positive peak at 265 nm and negative peak at 235 nm are characteristic of parallel-oriented G-quadruplex formation [45–47, 49]. To assess the thermal stability of the G-quadruplex, spectra were acquired from 25 to 95°C at 10°C increments (Fig. 2A). Melting experiments indicate thermal stability, with insignificant change in ellipticity at 265 nm, at least as high as 45°C. Melting commenced at 55°C and accelerated at higher temperatures [51]. This process was completely reversible, as seen in Fig. 2B, which shows the refolding process carried out by decreasing temperature. Next the concentration-dependence of ellipticity was used to assess whether the G-quadruplex forms via intermolecular or intramolecular association. The RNA oligonucleotide was diluted 10-fold to 2.2 μM and the melting and refolding experiments were repeated. Fig. 2C summarizes the result. While the molar ellipticity of the two samples was equivalent at 95°C, the ellipticity of the 22 μM sample exceeded that of the 2.2 μM sample at all other temperatures. This concentration dependence indicates that at least some G-quadruplexes form intermolecularly, most likely responsible for higher order (HO) EMSA bands (Fig. 2). Intramolecular formation is not excluded, and apparently predominates at higher temperatures [15, 48].

Fig. 2. Temperature-dependent CD spectra of C₂(G₄C₂)₄ RNA oligonucleotide.

Positive dichroism near 265 nm and negative dichroism near 235 nm is characteristic of parallel G-quadruplex formation. (A) Increasing temperature promotes partial unfolding of the G-quadruplex, which begins at 55°C. RNA concentration is 22 μM. (B) Decreasing temperature promotes refolding of the G-quadruplex, indicating reversibility. RNA concentration is 22 μM. (C) Plot of dichroism at 260 nm for 22 μM RNA (black) and 2.2 μM RNA (blue) vs. increasing temperature. The dichroism is equivalent at 95°C, but diverges at lower temperatures.

3.3. Comparison of the TZIP peptide sequence to the signature Pur domain

The goal of the design of TZIP is to retain those highly conserved sequences responsible for the essential, most basic activities of Pura (e.g., those involved in polynucleotide binding and RNA transport) while allowing us to optimize those less-conserved activities potentially important for functions (e.g., transcription/translation exclusive to certain higher organisms). It is among this latter group that functions beneficial to Pura treatment of ALS/FTD would potentially be found. The design did not just consider sequence amino acids, but also highly conserved sequence motifs that could determine structural aspects of Pura regions (e.g., amino acid motif GVFMR).

Fig. 3 shows an alignment of Pur domain sequences from several different organism. What is shown is alignment for Pur domain 3 of Pura. It can be seen that there are six amino acids identical in all the sequences examined (Black arrows). These include (F) phenylalanine, (N) asparagine, (G) glycine, and (E) glutamate, (R) arginine and finally another phenylalanine. Of these six amino acids, there are four that are also identical within the 89 amino acids of TZIP. These are found among the stars (*) under the TZIP sequence, which indicate conserved amino acids. Another two amino acids in TZIP comprising a (Q) glutamine and a (M) methionine are very conservative substitutions (colon symbol under corresponding aa). Another phenyalanine is strongly conserved in over 85% of the examined sequences and is also in TZIP. In all, looking at the starred sequences under TZIP, it can be seen that TZIP is considerably homologous to repeat 3, based not on identical, but on the conserved nature of these similar amino acids at many positions. We have made changes in several sequences designed to begin to examine the effects of these mutated versions of Pura repeat 3 domain on the efficacy of binding and ability to alter structure of G-quadruplexes.

Fig. 3. Comparison of the TZIP peptide sequence to the signature Pur domain.

Top panel: Pur domain sequences of similar proteins of eight of the most diverse species (spirochetes to humans) are shown aligned to genomic databases via CDD/SPARCLE protocol [67]. Species listed: (1.) Caenorhabditis elegans; (2.) Drosophila melanogaster; (3.) Arabadopsis thaliana; (4.) Caenorhabditis elegans, repeat 2; (5.) Homo sapiens Pura, repeat 3; (6.) Schistosoma mansoni; (7.) Treponema denticola; (8.) Treponema pallidum. Whereas bacterial sequences shown have only one Pur domain, C. elegans, D. melanogaster and A. thaliana have three repeats. The Pur domains presented all align with human repeat 3, number (5.). Six out of the eight amino acids identically conserved or nearly so also align with TZIP. Bottom panel: Alignment of the Homo sapiens Pura, repeat 3 sequence (number 5 in the top panel) with the TZIP peptide sequence. Star: fully conserved residue. Colon: strongly similar aa property conservation. Period: weakly similar aa property conservation.

The Pur domains aligned include some evolutionarily conserved as far back as bacteria. Therefore, the strongly conserved features of these amino acids must be related to cellular processes functioning in all of these species. While it is unlikely that these amino acids are essential for transcription of genes not present in each of these organisms, it is highly likely that these amino acids regulate a process universally essential for survival. This would include nucleic acid binding. We designed TZIP with one goal in mind:

The goal is to make a peptide based on certain highly conserved Pura amino acids, which will bind RNA and can alter features of the nucleic acid essential for participating in a highly conserved regulatory process.

Such a process could include RNA transport or aspects of RAN translation both common to C9orf72 ALS/FTD. Whether the peptide binding stimulates or inhibits function of a nucleic acid region is of interest because either one can be used for further peptide development. Based on our current TZIP sequence, we may mutate individual codons involved in its activity. G-quadruplexes are one such nucleic acid structural feature that may be altered by peptide binding. TZIP contains a 14 aa transporter sequence, beginning at aa 64 in Fig. 3, that is designed to be an improvement over a similar one in Pura.

3.4. A Pur-based peptide, TZIP, binds both G-rich and C-rich strands of the C9orf72 repeat sequence RNA, but not a random sequence of the same length

Binding of TZIP peptide (89 aa) to single stranded RNA was observed using polyacrylamide gel EMSA (Fig. 4). We sought to determine the specificity of binding different RNA hexanucleotide repeats by TZIP peptide. Particularly, can TZIP bind repeats of G₂C₄, the antisense complement of the C9 G₄C₂ repeat? Although the mechanism by which such complementary RNA is generated is unknown, dipeptide repeats derived from unusual, RAN, translation of the G₂C₄ RNA repeat are detected in neurons of ALS and FTD patients expressing the C9 HRE [52, 53], and some of these RAN proteins are neurotoxic [35, 36]. In the EMSA experiment of Fig. 4, the G₄C₂ RNA repeat sequence (red color) is compared to its G₂C₄ RNA complement (green color) with regard to TZIP binding. Both repeats are compared to a control oligoribonucleotide of random sequence (R, also green color). Both RNAs consist of four repeats (24 nt), which could form exactly one G-quadruplex. As seen in Fig. 4, there is just one gel band migrating slower than the linear forms of the oligoribonucleotides. This is in the lane for the G-rich RNA strand (G₄C₂)₄ unbound to peptide. Furthermore, this band is not observed with the (G₂C₄)₄ complementary RNA strand. This result strongly indicates that the discrete bands at or below the gel well seen on these EMSA gels are G-quadruplexes (labelled as GQ), but other possibilities have not been excluded. TZIP binds to (G₄C₂)₄ and to (G₂C₄)₄ RNA strands and shifts the RNA to higher order structures (HO) near the top of the gel.

Fig. 4. TZIP binds to RNA containing the G₄C₂ ALS/FTD hexanucleotide repeat and to its C-rich complement, but it does not bind a random sequence control of the same size.

Oligonucleotides were synthesized and subjected to electrophoresis in the presence or absence of 0.2 μM TZIP as described in Materials and Methods. For imaging on the LI-COR Odyssey Infrared Imaging System, the G-rich sequence, 5′-(G₄C₂)₄, was 5′-labelled with IRDye 700. The C-rich, 3′-(C₄G₂)₄, and random sequences were labelled with 5′-IRDye 800. The two different labels were imaged in the red (700 nm) and green (800 nm) channels respectively. GQ represents the most basic G-quadruplex of the sequence (Fig. 2), which is present as one major, slower-migrating band with the (G₄C₂)₄ oligonucleotide. It is only formed with the G-rich strand and not the C-rich complementary strand. The slowest migrating band contains higher-order (HO) structures of the repeat sequence. Oligonucleotide concentration is 1 nM in all lanes.

In Fig. 4, TZIP binds both the linear form of (G₄C₂)₄ and its basic GQ form and shifts them both to the same HO position. TZIP has no affinity for the negative control random sequence, R, an oligoribonucleotide of the same length.

3.5. Evidence for potential unwinding of G-quadruplexes by TZIP to form higher-order RNA structures

We sought to determine whether the binding of Pura to the C9 RNA repeat sequence is the basis for its protective effects. We, therefore, asked whether effects of TZIP on electrophoresis of the (GGGGCC)_n oligonucleotides are due to structural changes in the C9 repeat RNA molecules or due to the continued presence of TZIP. In the experiment of Fig. 5, a 2.5 μM solution of 5′-IRD-700-C₂(G₄C₂)₄ was either untreated or treated with increasing concentrations of TZIP and subjected to EMSA gel electrophoresis. TZIP shifts fast-migrating bands representing linear and G-quadruplex basic units (GQs) in lane a to bands positioned at or below the gel well. We refer to these bands with retarded mobility as (HO) (lanes b, c). This shift could be due to a change in G-quadruplex structures or to the presence of bound TZIP. The greater concentrations of oligonucleotide and peptide used in this experiment compared to those used in Fig. 4 contribute to a greater number of bands on the gel, which could be due to an increased formation of intermolecular G-quadruplexes (also see Fig. 2C) [18].

Fig. 5. TZIP binds to the ALS/FTD repeat sequence RNA forming higher order (HO) structures.

Following EMSA, oligonucleotide 5′-IRD700-R-C₂(G₄C₂)₄ forms gel bands of differing mobilities, indicating multiple structural forms, even in the absence of TZIP peptide binding (lane a). The G-quadruplex (labelled as GQ) and linear RNA forms are indicated. TZIP binds the RNA repeat sequence, forming gel bands of slower mobility as indicated (lanes b, c). TZIP concentration: lanes (a) 0 M, (b) 1.6 μM, (c) 12.4 μM. 5′ IRD700-R- CC(G₄C₂)₄ is 2.5 μM in all lanes.

3.51. TZIP directs G-quadruplexes to higher order (HO) structures both in the absence and presence of detergent

To determine whether or not bound TZIP is present in HO structures, we carried out the EMSA experiment shown in Fig. 6A. We employed 1% Tween 20, a non-ionic detergent, to remove some but not all protein from the RNA synthetic oligonucleotides. Lanes a, a′ show that in the absence of TZIP, the C9 repeat sequence in the form of six hexanucleotide repeats, R-C₂(G₄C₂)₆, forms a ladder of discrete structures characteristic of multiple G-quadruplexes in addition to the fast migrating linear form; a slight band is seen at the top of the ‘no TZIP’ control lane because the gel was overloaded in order to emphasize bands at the HO position. In the presence of TZIP, both the linear and G-quadruplex forms of R-C₂(G₄C₂)₆ are shifted to slower migrating bands, an effect which is not entirely reversed by the addition of detergent.

Fig. 6. TZIP binds to ALS/FTD repeat sequence RNA forming higher order (HO) structures in a manner not reversible by the removal of the peptide.

(A) The removal of peptide from ALS/FTD repeat sequence RNA does not dissociate the higher order structures. The 5′-IRD700-C₂(G₄C₂)₆-3′ oligoribonucleotide was incubated with TZIP and either followed or not followed by 1.0% Tween 20 at room temperature for 10 minutes and then subjected to EMSA. Lanes a,b,c and a′,b′,c′ indicate increasing concentrations of TZIP. As a non-ionic detergent, Tween 20 removes the peptide without altering the RNA secondary structure. ‘Peptide bound’ indicates bands that are seen only in the presence of TZIP. Higher-order (HO) structures contain bound TZIP as seen in lanes b and c on the left as well as multiples of the hexaribonucleotide, as seen in lanes b′ and c′ on right, where TZIP is displaced by Tween 20. The linear form of the oligoribonucleotide is the fastest migrating band. Concentrations of TZIP are lanes (a, a′) 0 M, (b, b′) 200 nM, (c, c′) 500 nM. 5′-IRD700- CC(G₄C₂)₆-3′ is 1.0 nM in all lanes. (B) TZIP binds the linear and G-quadruplex forms of the C9 repeat sequence RNA with different affinities. EMSA (not shown) was performed in triplicate to determine the binding affinity of TZIP and G-quadraplex forms of the 4-mer repeat sequence RNA. Kd-apparent were determined by fitting data to GraphPad Prism equations for plotting nonlinear regression. These analyses yielded plots conforming to simple non-cooperative binding. Concentration of TZIP ranged from 25 nM to 500 nM. The concentration of 5′ IRD700-R-C₂(G₄C₂)₄ was 1.0 nM.

TZIP appears to display different affinities in binding the basic G-quadruplexes (GQs) compared to the linear R-C₂(G₄C₂)₆ form. With increasing concentration of TZIP, the G-quadruplex band in Fig. 6A (lane a) is shifted to the peptide bound or HO positions (lanes b, c), while the linear band at the lower 200 nM TZIP concentration is largely unbound. However, following the addition of detergent to TZIP-bound R-C₂(G₄C₂)₆ (lane b′), the same G-quadruplex form is present as a fast migrating band above the linear band similar to corresponding bands in lanes without TZIP and detergent. The effect of increased concentration of TZIP (500 nM) on binding the C9 RNA is seen in lanes c, c′. In lane c, all of the linear as well as the G-quadruplex bands are bound and shifted to either the peptide bound or HO levels. In the presence of Tween 20, lane c′, the intensity of the linear band is increased compared with lane c, but the G-quadruplex bands are substantially diminished compared to the same band in lanes a′ and b′. This indicates that the G-quadruplex form can be bound by TZIP at a 200 nM concentration without an irreversible change in structure (lanes a, b, and a′, b′). On the other hand, at higher TZIP:RNA ratios, seen in lane c′, there may be irreversible changes in polynucleotide structure.

Both the linear polynucleotide and the basic G-quadruplexes are strongly shifted to the HO position by TZIP (lanes b, c). The presence of detergent deconstructs the HO band. The linear form reappears and a less intense band of G-quadruplex appears (lanes b′, c′), but much of the HO band in these lanes continues to migrate as a smear of bands with decreased mobility and some possible G-quadruplex multimeric structures. A minor amount of protein may be bound in the presence of detergent prior to protein inactivation or removal. Once the nucleic acid bound protein is removed by the detergent, induced nucleic acid secondary structure may remain. Although more evidence is needed for potential unwinding of G-quadruplexes by TZIP, unwinding could help explain generation of the multiple HO structures generated by TZIP.

The apparent differing affinities of TZIP in the binding of the two structural forms of the (GGGGCC)_n RNA, the linear and the G-quadruplex, as seen in Fig. 6A, led to an analysis of their relative binding affinities. The 4-mer repeat sequence RNA, C₂(G₄C₂)₄, and TZIP peptide were both used at nM concentrations with the TZIP in large excess. In the replicate experiments, TZIP was found to have moderately tight and specific binding to the G-quadruplex with a dissociation constant (K_d) of 77 nM. The affinity for linear RNA is somewhat lower with a K_d of 174 nM (Fig. 6B). Our binding curves of TZIP and R-C₂(G₄C₂)₄ can be fit by a simple binding mechanism (see Fig. 6 legend) to obtain a binding constant, K_d, which indicates that the binding is not cooperative. The data does not, however, distinguish between the binding of a monomer and a multimeric unit. The effect on K_d, if TZIP were a dimer, would be to halve the K_d values.

It is significant that TZIP binds both structural forms of the RNA. There is considerable evidence that the HRE in ALS/FTD C9orf72 affected cells will form G-quadruplex structures as well as exist as linear RNA [15–17, 54]. TZIP clearly binds both forms.

3.6. Molecules of different C9orf72 (GGGGCC)_n repeats combine to form mixed structures containing discrete units

We asked whether the potential G-quadruplex structures seen as discrete slow-migrating bands on EMSA gels are formed between two C9 repeat sequence molecules of different repeat lengths [55]. Such formation of mixed discrete structures would strongly indicate that they are G-quadruplexes and that such gels may be used as an assay for them (Fig. 7). To combine the different repeat molecules, samples of each were treated together under conditions previously shown to generate G-quadruplexes [19]. In this procedure, samples were dissociated together, including RNA denaturation, under conditions ensuring that the starting configuration was the same for each GGGGCC-repeat molecule. All oligonucleotides representative of the C9 repeat sequences are G-rich with no significant C tracts. Any structures formed within mixed GGGGCC-repeat molecules would thus be based on Hoogsteen bonds. Fig. 7 shows the results of formation of mixed structures obtained by combining two (GGGGCC)_n molecules of different repeat lengths (a 10-mer with a 4-mer). The left and center panels show samples of the oligoribonucleotides as commercially synthesized: 10-mer (A, red color), 4-mer (B, green color) while panel C shows a mixture of the two oligoribonulceotides at the same concentrations. All oligoribonulceotides were subjected to the same denaturing and slow cooling steps to form G-quadruplexes as described in Materials and Methods. Panels A - C (lanes e, j, o) are aliquots of the annealed products that were heat denatured immediately before loading onto the gel. Lanes a-d, f-i and k-n are decreasing concentrations of the mixture loaded on the gel. Combined forms containing both red and green fluorescence would be yellow. Panel C (lane k) shows that upon combining the HRE 4-mer with the 10-mer, bands appear that contain structures consisting of both oligoribonucleotide molecules. There is yellow signal where a background of molecules containing red and green overlaps on the gel. To avoid any mislabelling resulting from overlapping, we focus on discrete labelled structures in areas of the gel with minimal diffuse background. One such discrete, labelled band (GQ, arrows between center and right panels) is revealing in several ways. Most importantly, this band is not present with either the 4-mer (green color) or the 10-mer (red color) alone. It is only seen when the two repeat molecules are added together, dissociated together and allowed to reform together. We show in Figs. 5, 6 that TZIP facilitates both the dissociation of these RNA structures and their reformation. This would be critical for potential regulation of C9 G-quadruplex effects in the C9 HRE found in ALS/FTD.

Fig. 7. G-quadruplex formation between two oligoribonucleotides bearing different numbers of ALS/FTD hexanucleotide repeat units.

(A) RNA containing 10 repeats was 5′ end labelled with IRD700 (red) linked by 2 C. (B) RNA containing 4 repeats was 5′ end labelled with IRD800 (green) linked by 2 C. (C) RNAs containing 10 repeats and 4 repeats were mixed for this assay. RNA in lanes of all three panels was subjected to a heat denaturation and cooling procedure shown to yield G-quadruplexes [19] as described in Materials and Methods. Aliquots from each reaction volume were loaded at 20 μl per lane: Lane a: 1 pmole of RNA (10 repeats), Lane f: 3 pmole of RNA (4 repeats), Lane k: 1 pmole of RNA (10 repeats) with 3 pmoles of RNA (4 repeats). Lanes b, c, d: 2-fold serial dilutions of lane a. Lanes g, h, i: 2-fold serial dilutions of lane f. Lanes l, m, n: 2-fold serial dilutions of lane k. Lanes e, j, o: RNA from each reaction heat denatured just prior to loading.

3.7. Circular Dichroism demonstrates that TZIP unfolds G-quadruplexes in (GGGGCC)_n RNA and converts this RNA to higher order structures

TZIP may alter the structure of the existing G-quadruplexes of the CC(GGGGCC)₄ RNA. For TZIP to bind the G-rich sequences, the G-quadruplex structure would need to unfold, so a change in folding may occur in the transition from lower to HO structures. We sought to strengthen evidence for alterations wrought by TZIP. Fig. 8 describes an experiment in which circular dichroism was used to examine structures of the R-C₂(G₄C₂)₄ oligoribonucleotide in the presence of different concentrations of TZIP. A control was this same oligoribonucleotide in the absence of TZIP (indicated RNA alone). TZIP induces its effect on the C9 repeat sequence through its ability to interact with nucleotides in different secondary structural forms. Fig. 8 (left panel) shows the CD of TZIP alone in the absence of oligoribonucleotide. TZIP displays no appreciable dichroism in the range of 240–300 mM, indicating that TZIP does not directly contribute to the dichroism characteristic of the parallel G-quadruplex in C9 GGGGCC-repeat RNA. As can be seen in Fig. 8 (right panel), the RNA dichroism intensity decreases significantly upon addition of TZIP at a 1:1 ratio, and decreases further when TZIP is added at a 4:1 molar ratio to RNA. Because TZIP exerts its effect at such a high ratio to RNA, it is likely that the effect is due to direct binding of TZIP to individual G-quadruplexes, rather than a catalytic effect of TZIP on the overall G-quadruplex structure. Data not shown indicate that TZIP does not digest RNA. Furthermore, the RNA is 2′-O-methyl RNA, which is resistant to RNase activity. Precipitation is not observed. Therefore, the effect of TZIP is not to eliminate RNA, but to cause changes that induce a loss of standard G-quadruplex CD spectrum. The changes seen are dependent upon TZIP concentration and brought about by TZIP. The data of Fig. 6 indicate that in the presence of TZIP, G-quadruplex RNA is unfolded and then converted to higher order structures resulting in a change in the CD spectra. At TZIP:RNA=4:1, not only is there an effect on dichroism intensity, but also an affect on the position of the maximum. This is a clear indication of a change in RNA conformation (there could also be some change in particle size). This shift in dichroism maximum mirrors the shift that occurs upon melting (Figure 2A), which further corroborates a change in RNA structure upon interaction with TZIP at 4:1 in Fig. 8 (right panel).

Fig. 8. Circular dichroism (CD) analysis of effect of TZIP on the G-rich oligoribonucleotide C₂(G₄C₂)₄.

Left panel: CD spectrum of TZIP alone, showing little dichroism in the range needed for analysis of the G-quad RNA (240–300 nm). Right panel: Solid line: control CD spectrum of the RNA in the absence of TZIP. Dashed line: CD of a 1:1 molar ratio of TZIP:RNA. Dotted line: CD of a 4:1 molar ratio of TZIP:RNA.

Though beyond the scope of the present work, further information regarding the TZIP peptide structure and its changes upon interaction with the G-quadruplex could be derived from experiments analogous to those presented in Fig. 7. A complete examination of interactions between protein domains in TZIP with C9 GGGGCC-repeat RNA has been initiated. These data will allow an assessment of functional protein-RNA interactions that can lead to the design of peptides for therapeutic evaluation with potential to influence disease pathologies such as those resulting from G-quadruplex-forming ALS/FTD repeat expansions.

4. Discussion

Several reports have documented the ability of Pur-alpha (Pura) to alleviate certain cellular responses to the C9 G-rich hexanucleotide repeat sequence expanded in ALS/FTD. Pura can reverse effects of the resulting RNA toxicity, whereas certain other RNA repeat-binding proteins do not affect toxicity [25, 37, 56]. A major mechanism causing neurodegeneration due to the C9 expanded repeat is the generation of dipeptide repeat proteins (DPRs) from sense and antisense RNA transcripts of the repeat sequence [35, 52, 53, 56, 57]. It may be possible that Pura, by binding to the C9 repeat sequence RNA, could negatively influence RAN translation, thereby decreasing DPR production and helping to ameliorate neurodegeneration. A zebrafish model reveals cytoplasmic RNA toxicity as a pathogenic mechanism. Pura reverses this mechanism [37]. Because the repeat unit, (G₄C₂), is an established Pura binding element [1, 7], it is possible that the expanded repeat in DNA or RNA sequesters Pura away from its normal cellular functions [9, 25, 58]. Little is known about the binding of Pur proteins to their G-rich elements. We find that the binding constants of Pur proteins and peptide derivatives are all within the nM sensitivity range. This is tight binding, but not irreversible, thus subject to competitive interference. We reasoned that a synthetic peptide with equal or better binding and better means of exogenous cell entry might be an improvement in ability to relieve HRE sequestering of Pura. TZIP peptide was designed with this in mind (Fig. 3). TZIP is not a Pura homolog or derivative. TZIP was synthesized to retain amino acids shown to be necessary for polynucleotide binding [7] and conserved throughout evolution [5], Fig. 3. An improved transporter sequence was added [59].

Here we have asked whether a synthetic peptide based on, but not identical to, sequences of a signature Pur domain can perform Pura-like binding and unwinding activities on the C9 GGGGCC-repeat RNA. We find that TZIP binds the C9 repeat sequence in single-stranded DNA and RNA. TZIP binds the C9 RNA and DNA hexanucleotide repeats with high affinities. DNA B-helical structure must be unwound to allow formation of G-quadruplexes. The B helix is negatively supercoiled in most chromatin. Although negative supercoiling is favorable to DNA unwinding, such supercoiling alone is not sufficient to drive G-quadruplex formation [60]. Pura locally unwinds DNA at its G-rich binding elements [7] and might be expected to facilitate G-quadruplex formation. We find that TZIP binds both the G-quadruplex form and the linear form of the CC(GGGGCC)₄ RNA repeat with similar affinities and mediates quadruplex unwinding and secondary structural changes among the RNA forms (Fig. 6). Structural changes occur at approximately the same level of binding affinity.

There is now considerable evidence for the existence and functional relevance of G-quadruplexes in vivo. Alternative explanations, however, contribute to the controversial nature of existing data. As reviewed [61], data for in vivo existence of G-quadruplexes fall into four groups: 1) existence of helicases specific for G-quadruplexes; 2) in vivo NMR results; 3) antibodies specific for G-quadruplexes; 4) indirect functional and epigenetic evidence. It has been reported that local epigenetic changes can be triggered by impeding DNA replication through small molecule stabilization of G-quadruplexes [62]. Perhaps the most compelling data are the existence of enzymes, i.e., helicases, specific for function with G-quadruplexes. For example, mutation of the DOG-1 gene of C. elegans leads to deletions at the 3′ end of long poly-G tracts, and DOG-1 is proposed to encode a G-quadruplex-specific helicase that facilitates lagging-strand DNA synthesis [63]. The human counterpart of DOG-1 is FANCJ/BRIP1 [64]. A further example, the Pif1 family of helicases, is conserved from bacteria through humans [65, 66]. These helicases bind to G-quadruplexes, protecting adjacent genomic DNA from damage during replication.

5. Conclusions

Several conclusions can be drawn from these data. Notably, a synthetic polypeptide, TZIP, mimicking a Pur domain, binds with high affinity to the G-rich hexanucleotide sequence present in the C9 form of ALS/FTD. This HRE is specifically expanded in the C9orf72 disease form, which is the most common genetic abnormality of ALS and FTD. The observation that Pura binds the G-rich repeat sequence of disparate diseases, and the fact that these G-rich repeat expansions are frequently present in chromosomal regulatory regions [5], suggest that there is a chromosome structure and functional feature driving evolution of the interaction between the Pura signature domain and the repeated Pura nucleic acid binding element. We identify the TZIP nucleic acid binding configuration in the C9 HRE, including the G-quadruplexes, as a promising target for ALS/FTD therapy. TZIP can help form or unform G-quadruplexes (Figs. 5, 6). One may ask how an 89 amino acid peptide, TZIP, could be a promising candidate for an interventional therapeutic agent. One example answer involves its delivery. TZIP can be stereotactically injected into the cerebrospinal fluid where its added transporter sequence will allow it to enter ALS/FTD affected neurons. The inclusion of a transporter sequence may allow it to be delivered to avoid entering the circulation and ensuing liver degradation. It is intriguing that TZIP unwinds the parallel G-quadruplex form of the C9 sequence and that no other form is visible in the spectrum of Fig. 8. It is likely that the C9 sequence in its milieu is prone to formation of the parallel G-quadruplex form exclusively. Pur-peptide modulation of C9 repeat RNA through conformational shifts represents a new mechanism for regulation of G-quadruplex functional activity. TZIP may change the configuration of the C9 repeat RNA in such a way that it can alter its interaction with other proteins. Specific modifications of TZIP may alter its efficacy by altering its interaction with G-quadruplexes. Studies of the TZIP-C9 repeat sequence interaction will help reveal activities of Pur-like domain protein motifs protective against ALS or FTD and help optimize peptide design.

Supplementary Material

1

Figure S1. CD melting (left) and refolding (right) profile of 2.2 μM representative C9 repeat sequence RNA CC(GGGGCC)4. Left and right profiles are identical.

HIGHLIGHTS.

1. Pur-based peptide, TZIP, binds the hexanucleotide repeat expanded in C9orf72 ALS/FTD.

2. Circular dichroism analysis of C9orf72 repeat RNA reveals parallel G-quadruplexes.

3. TZIP binds both linear and G-quadruplex forms of the C9orf72 RNA repeat sequence.

4. TZIP binding alters the configuration of the G-quadruplex RNA secondary structure.

5. Our results may help reveal Pur-based motifs protective in repeat expansion diseases.