An Analysis of Nucleotide–Amyloid Interactions Reveals Selective Binding to Codon-Sized RNA

Interactions between RNA and proteins are the cornerstone of many important biological processes from transcription and translation to gene regulation, yet little is known about the ancient origin of said interactions. We hypothesized that peptide amyloids played a role in the origin of life and that their repetitive structure lends itself to building interfaces with other polymers through avidity. Here, we report that short RNA with a minimum length of three nucleotides binds in a sequence-dependent manner to peptide amyloids. The 3′–5′ linked RNA backbone appears to be well-suited to support these interactions, with the phosphodiester backbone and nucleobases both contributing to the affinity. Sequence-specific RNA–peptide interactions of the kind identified here may provide a path to understanding one of the great mysteries rooted in the origin of life: the origin of the genetic code.

Oligonucleotides were deprotected from support using 1 bar gaseous methylamine at 65 °C for 1.5 h. For the pdGd, TBDMS protection group was removed by adding a freshly prepared mixture of 1-N-methyl-2-pyrrolidone:TEA:HF⋅3 TEA 6:3:4 (130 μL) and incubation at 70 °C for 90 min. The DNB protection group was removed by UV irradiation at 365 nm for 15 min on a transilluminator. In all cases, except pddd, the oligonucleotides, were obtained in desalted form and then purified by reverse phase HPLC on a Kinetex C18 5 µm 10x250 mm column (Phenomenex) in a CH3CN/H2O/TEAA solvent system and quantitated by their extinction coefficient at 260 nm. Pure oligos were then lyophilized and dissolved in water at a concentration of more than 1 mM and stored at -80 °C and then diluted to different buffers as required. For pddd the dinitrobenzhydryl (DNB) protecting group was left in place to facilitate HPLC purification and quantitation. It was removed by UV irradiation at 365 nm for 5 min on a transilluminator and then passing the sample over a C18 SPE column. pddd concentration was determined by the 1 H-NMR signals of the ribose compared to the corresponding signals in the uncleaved DNB-pddd sample.
Upon completion, the reaction was quenched using 3 ml of quenching buffer (GlenResearch, USA). The mixture was then applied to a HiPrep 26/10 desalting column (GE Healthcare, Austria) using an ÄKTA start system (GE Healthcare, Austria). The RNA was eluted using HPLC-grade water. The fractions containing the desired RNA (UV detection at 254 nm) were collected in a 50 ml round bottom flask and evaporated to dryness. The residue was dissolved in 1 ml HPLC-grade water and stored at -20 °C until needed. The quality of the RNA was checked via anion-exchange chromatography over a Dionex DNAPac PA-200 column (4x250 mm; Eluent A: 25 mM Tris.HCl, pH 8.0, 20 % (v/v) acetonitrile, 10 mM sodium perchlorate; Eluent B: 25 mM Tris.HCl, pH 8.0, 20 % (v/v) acetonitrile, 600 mM sodium perchlorate).

Fibrillization
Lyophilized aliquots of pure peptides were dissolved in water (peptides 4, 5, 7, 8, 9, 10 and 11 in Table 1, and peptide series 1-5 in Table S1 at pH ~ 8. [5][6][7][8][9] or DMSO (peptides 1, 2, 3 and 6 in Table 1 and peptide series 6-18 in Table S1) at a concentration of more than 5 mM. The stock was then diluted to typical concentration of 200 µM (unless otherwise stated) in the citrate-phosphate buffer of the desired pH and incubated overnight at room temperature in a thermomixer (Eppendorf) with agitation at 1000 r.p.m. The fibrillization was monitored by Fourier transform infrared spectroscopy and HPLC-based supernatant analysis.

Fourier transform infrared spectroscopy
The peptide aggregates in citrate-phosphate buffer were centrifuged at 25,000 g, and the insoluble material (pellet) was washed in 10 mM HCl and centrifuged again. The pellet was resuspended in 5 μl 10 mM HCl and applied to a diamond ATR cell on a Bruker Alpha FTIR spectrometer. The samples were air-dried before measuring their spectra with 32 scans and a resolution of 2 cm −1 . The buffer exchange was important as the strong citrate-phosphate buffer absorbance overlaps with the relevant portion of the peptide spectrum.

Transmission electron microscopy
The samples prepared as for the IR measurements were applied directly to the negatively glowdischarged carbon-coated copper grids, washed with water and stained with phosphotungstic acid. The imaging was done on a FEI Morgagni 268 electron microscope. The data were analyzed in the Dynamics Center software (Bruker) by fitting the peak areas for several groups of resonances to the function:

Solid-state NMR structure determination of the complex between the VAQAQINI-NH2 peptide amyloid and the RNA pGUCAp
Amyloids of VAQAQINI-NH2 were prepared in citrate-phosphate buffer pH 3 with uniformly labeled peptide (sample I), mixed unlabeled and uniformly labeled peptides in a ratio of 2.8:1 (sample II, "diluted" sample), mixed-labeled peptides V 15 N-AQAQI 13 C-I-NH2 and V 13 C-AQAQI 15 N-I-NH2 in a ratio 1:1 (sample III), uniformly labeled peptide amyloids in presence of unlabeled RNA pGUCAp (sample IV) and uniformly labeled peptide amyloids in presence of specifically 13 C-labeled RNA pGUCAp (sample V, as described above) in an approximate ratio of 10:1. Sample I-IV were filled by established procedures using home-built tools 3 in 3.2 mm rotors, while due to limited amounts of sample V, it was filled in a 1.9 mm rotor, as was a second sample of sample I.
The solid-state NMR experiments collected for these samples for both the sequential assignments and structure determination are summarized in Tables S3-S5 and in part shown in Figures S12-S15. For the peptide, solid-state NMR sequential backbone and side-chain resonance assignments were obtained by state-of-the-art 2D NMR experiments, first for the peptide amyloid sample alone and then for the complex with RNA. The assignment of the specifically labeled 13 C of the RNA was established both by chemical-shift statistics (from the BMRB database) and the presence of inter-residue cross peaks between C8 of guanine and C6 of uracil, C6 of uracil and C6 of cytidine and an intra-residue cross peak C2-C8 of adenosine observed in the 450 ms DARR spectrum recorded on sample V. Three distinct 31 P chemical shifts were identified in the [ 13 C, 31 P] CHHP 2D spectrum. [4][5] However, they are of ambiguous nature, since due to a limited signal-to-noise ratio for sample V no 13 C-31 P cross peak within the RNA was detected in a corresponding experiment (not listed in the Tables S3-S5). All of the amino acid residues were assigned. The chemical-shift values have been deposited in the BMRB data bank, accession number 34838.
The collection of restraints for the structure calculation started from the sequential assignment.
Based on the sequential assignment the identification of the β-sheet secondary structure of residues 2-7 of the peptide amyloid could be identified and yielded angular restraints using the TALOS+ databank 6 within CYANA. Next, a [ 13 C, 15 N]-Proton Assisted Insensitive Nuclei Cross Polarization (PAIN) experiment 7 was measured on the mixed sample III. The identification of 13 C-15 N peaks for Ala2 and Ile8 indicates the presence of an in-register parallel β-sheet yielding valuable inter-molecular hydrogen-bond restraints for the structure calculation. Next, distance restraints were collected from a 20 ms DARR spectrum of sample I as well as for the RNA peptide complex sample IV showing very similar spectra as the apo form. All these distance restraints were treated in the following as ambiguous in respect to whether they are of intra or inter-molecular nature. A 2D PAR of the mixed sample III allowed the unambiguous identification of the inter-sheet distances CB Ala2 -CA Ile8, CB Ala2 -CB Ile8, and CB Ala2 -CG1 Ile8. RNA peptide distance restraints were collected from 13 C-31 P and 13 C-13 C cross peaks in the corresponding 2D spectra from sample IV and from 2D 150 ms and 450 ms DARR of sample V, respectively. 13 C-31 P distance restraints were treated as ambiguous with respect to 31 P excluding the 3'-end phosphate group, since no distance restraints from adenine were detected which also shows very weak signals for its 13 C moiety.
An unambiguous 13 C amyloid peptide assignment of both 13 C-31 P and 13 C-13 C cross peaks was limited by both the 1:7 binding stoichiometry between RNA and peptide ( Figure S7) as well as the low level of bound RNA ( Figure S11) leading to a low signal-to-noise ratio of these inter-molecular cross peaks when compared with the intra-amyloid cross peaks. Additionally, only a tool to help the structure calculation to converge towards a minimum that satisfies the experimental data. The second artificial restraint is a short-range distance between the positively charged N-terminus of the peptide and an ambiguous 31 P of the RNA. A 3 Å distance limit was selected in the final calculation, but similar structures are obtained also with a value of 6 Å (not shown). This restraint reflects the binding studies discussed above with many distinct peptide amyloids and RNA entities suggesting a charge-charge interaction between the two moieties.

RNA-binding induced
The final CYANA structure calculation resulted in a well converged structure with an average target function of 6.6 Å 2 for the final bundle comprising the 10 best conformers (i.e. with lowest TF) indicating the presence of a self-consistent data set. The coordinates have been deposited at the PDB under ID 8PXS. The RMSD of the peptide amyloid is 0.1 Å for the backbone and 0.8 Å for all non-hydrogen atoms. The RNA is much less well defined compared to the peptide amyloid yielding an overall heavy atom RMSD of 1.5 Å (Table S2). We noted that during the early stages of refinement, the RNA molecule was sometimes oriented upside down bound to the peptide amyloid but that these alternative structures are absent in the final structure calculation result. Figure S1. β-sheet aggregation of the peptides confirmed by FTIR spectroscopy. Each of the 11 peptides in this study (see Table 1) were characterized by FTIR. All of them exhibit the characteristic amide absorption near 1620-1630 cm -1 confirming the b-sheet aggregation and all have a smaller band near 1690 cm -1 indicating that the peptides form anti-parallel sheets in the fibers (as in the Figure 1 schematic). The peptide aggregates made at a concentration of 200 µM in citrate-phosphate buffer at pH 3 were centrifuged at 25,000 g, and the insoluble material (pellet) was washed in 10 mM HCl for the FTIR measurements.  Table 1) were characterized by negatively stained transmission electron microscopy. The micrographs all contain fibrillar or amyloid-like structures. The same samples prepared for the FTIR measurements were used for the microscopy. 10 Figure S3. Peptide solubility at pH 3 and 7. Peptides used in this study (see Table 1) were aggregated at a concentration of 100 µM and the % aggregation was determined by HPLCbased supernatant analysis.             Suppl. Fig. S16     To visualize the relative binding between all 6 pairs of RNA trinucleotides to a particular peptide, the log of the ratio of % bound RNA for each pair is displayed as a heatmap (right) with only the variable nucleotide displayed on the axes of heatmap plots. To minimize the bias from errors, all measured values of less than 4% are set to 2% for the log-ratio calculation, because for these weak binders, small errors are large relative to the measured values. For reference, the measured absolute %-bound is given by numbers on each side of the diagonal line. (B)-(O) The log ratio of %-bound RNA to a particular peptide for each of the pGNG or pGNC trinucleotide sets is plotted for a selection of the tested peptide sets (Table S1) demonstrating the variable specificity in the amyloid-RNA interactions. Each panel presents results from one (or two related) peptide sets at different assay conditions. Peptide and RNA sequences and the concentrations used are noted on each plot as are the assay conditions which include variations in temperature, pH and salt.   Table S1. Amyloidogenic peptide sets used in the sequence-selective RNA trinucleotideamyloid interaction study a Amino acid residues are represented in the standard single letter code; NH2 is for amidated C-terminus; X = G, A, V or D (*or E in S17 and S18) b Dominant ionization states of the ionizable groups of the peptide at neutral pH are listed as the charges on the N-terminus, sidechains and C-terminus. The listed ionization states are based on individual pKas of the groups and are expected to vary depending on buffer pH and the aggregation state of the peptide. c Side chains with X = G, A and V d Side chains with X = D or E

Set
Composition a Ionization at neutral pH b N Sidechain c Sidechain d C Sum c Sum d S1 FXFEFQFX Backbone of residues (Å) 0.1 ± 0.03 All heavy atoms of residues (Å) 0.79 ± 0.1 RMSD to mean for the RNA -peptide amyloid complex All heavy atoms (Å) 1.5 ± 0.4 a Each group of symmetrically equivalent distance restraints is counted as a single restraint. Distance restraints with multiple assignments are classified by the assignment spanning the shortest residue range. b Each group of symmetrically equivalent distance restraints is counted as a single restraint. Distance restraints with multiple assignments are classified by the assignment spanning the shortest residue range. The distance restraints include one artificial distance restraint helping to find a good energy minimum (see text). c Each group of symmetrically equivalent distance restraints is counted as a single restraint. Distance restraints with multiple assignments are classified by the assignment spanning the shortest residue range. The distance restraints include one additional distance restraint derived indirectly from interaction studies using RNA variants (see text). d Each hydrogen bond was restrained by two upper and two lower distance bounds. e Where applicable, the average value and the standard deviation over the 10 conformers that represent the NMR structure are given.