Combined High-Pressure and Multiquantum NMR and Molecular Simulation Propose a Role for N-Terminal Salt Bridges in Amyloid-Beta

Several lines of evidence point to the important role of the N-terminal region of amyloid-beta (Aβ) peptide in its toxic aggregation in Alzheimer’s disease (AD). It is known that charge-altering modifications such as Ser8 phosphorylation promote Aβ fibrillar aggregation. In this Letter, we combine high-pressure NMR, multiquantum chemical exchange saturation transfer (MQ-CEST) NMR, and microseconds-long molecular dynamics simulation and provide evidence of the presence of several salt bridges between Arg5 and its nearby negatively charged residues, in particular, Asp7 and Glu3. The presence of these salt bridges is correlated with less extended structures in the N-terminal region of Aβ. Through density functional theory calculations, we demonstrate how the introduction of negatively charged phosphoserine 8 influences the network of adjacent salt bridges in Aβ and favors more extended N-terminal structures. Our data propose a structural mechanism for the Ser8-phosphorylation-promoted Aβ aggregation and define the N-terminal salt bridges as potential targets for anti-AD drug design.

and first-(B1) and second-order (B2) pressure coefficients were determined separately for amide 1 H and 15 N nuclei of different residues. This approach enables correcting for the direct effect of pressure on the chemical shifts of each amino acid, especially in the titratable residues such as histidine in which the pKa of side-chains are pressure-dependent.

Multi-quantum Chemical Exchange Saturation Transfer (MQ-CEST) NMR experiments
The NMR samples contained 100 mM free [ 13 C6, 15 N4]-L-arginine (25 mM HEPES, pH 5.1, 10% D2O) or ca. 80 M 13 C, 15 N-labeled A40 (25 mM HEPES, 50 mM NaCl, pH 6.4, 10% D2O). The MQ-CEST experiments were measured on a Bruker Avance Neo 800 MHz NMR spectrometer equipped with a 3 mm TCI cryo probe at 274 K, following the NMR pulse sequence introduced by Karunanithy et al 9 . Briefly, this is a 13 C-detected NMR experiment starting with the steady-state longitudinal magnetization on the 1 H  spins of arginine sidechains and correlating the chemical shifts of the 13 C  spins of arginine side chains in the direct dimension to the chemical shifts of their scalar coupled 15 N  spins in the indirect dimension. At a specific point during the sequence a three-spin order density element proportional to 4Cz  Nz  Nz  is generated, which is then subjected to a weak 15 N B1 field during the CEST period, where the 15 N carrier frequency is varied over a range covering the chemical shifts of the two N  spins of arginine sidechains. As a result, the CEST intensities are obtained from the intensity of the N  -C  correlation peaks as a function of the 15 N carrier offset. In the present study, the 15 N CEST elements were 250 ms long, during which the 15 N B1 fields of 12.9, 21.4 and 29.9 Hz for the free arginine sample or 29.9 Hz for the A40 sample were applied. For the reference free arginine sample, 61 evenly spaced 15 N carrier offsets between 63.0 and 81.5 ppm at 25 Hz (ca. 0.31 ppm) intervals were used during the 15 N CEST element. For the much lower concentrated A40 sample 30 evenly spaced 15 N offsets between 63.6 and 81.5 ppm at 50 Hz (ca. 0.62 ppm) intervals were used. At each 15 N B1 field value a reference spectrum was recorded without the CEST element (TCEST = 0), but including the heatcompensating element of identical duration before recycle delay (d1), during which the same 15 N B1 field was applied far off-resonance (at 232 ppm). The 15 N offset-dependent CEST intensities were presented as the intensity ratios (I/I0) with respect to the N  -C  peak intensity in the reference spectrum (I0). The 15 N B1 field strengths were calibrated as described in reference 10 . It is worth mentioning that the sensitivity of MQ-CEST experiment is highly dependent on water-H exchange rate, therefore the pH of sample had to be reduced (from 7.4 of other experiments) to 6.4 for the sake of sensitivity.

Two-dimensional homonuclear NMR experiments
The 1 H, 1 H TOCSY experiments were performed on a Bruker (Germany) Avance 800 MHz spectrometer equipped with a cryogenic probe, as in 11 . Briefly, the NMR samples contained 0.4 mg/mL Aβ in 20 mM sodium phosphate buffer, pH 7.2. The NMR measurements were performed at 278 K. The time domain data contained 2,048 and 600 complex data points in t2 and t1, respectively. The TOCSY mixing time was 60 ms.

Molecular Dynamics (MD) simulation
The MD trajectory of ref. 12 was analysed. There, the ca. 30-s long MD simulation of A40 was performed using the a99SB-disp force field with the optimized TIP4P-D water model at 1 bar and 300 K. The 30,000 MD frames saved at 1 ns intervals were analysed in this study. The salt bridges were identified using a cut-off of 4 Å between N-O atom pairs of basic (Arg and Lys) and acidic (Asp and Glu) residues. 13

Density Functional Theory (DFT) calculations
All DFT energy calculations of a 9-residue ( 2 AEFRHDSGY 10 ) N-terminal sequence of A40 were carried out using Gaussian 09. 14 The A40 conformers obtained from the MD trajectory were used for DFT calculation of energy without further geometry optimization. N-and C-termini of a 9-residue ( 2 AEFRHDSGY 10 ) fragment of A40 were capped with methyl groups. The energy calculations were performed at DFT level using hybrid meta exchange-correlation functional M06-2X [15][16][17] and Def2TZVPP 18-19 basis set. Integral equation formalism polarizable continuum model (IEFPCM) 20 was employed with water as a solvent. For each conformer two energy calculations were performed with and without the presence of doubly negative charged phosphate group of serine-8 residue.
A model pentapeptide Glu-Gly-Arg-Gly-Asp conformers with ( Figure S8a) and without ( Figure  S8b) salt bridge between Arg and Glu/Asp side chains were used for the NMR shielding tensor calculation of the guanidinium group of Arg using Gaussian 09. Geometry optimization of both the conformers was carried out at DFT/B3LYP 21-24 /6-31G(d) 25 level of theory and by employing IEFPCM solvent model with water as a solvent (for Cartesian coordinates of the energy-minimized model peptide in two conformations, please see below). The energy minimized conformers were used as input geometries for the NMR shielding tensor calculations using gauge-independent atomic orbital (GIAO) 26 method at DFT/mPW1PW91 27-28 /6-311+G(2d,p) level of theory and by employing IEFPCM solvent model with water as a solvent. The calculated 1 H and 13 C nuclear shielding tensors were converted into chemical shifts using the scaling factors obtained from the CHESHIRE (chemical shift repository) 29 web site. The scaled chemical shifts were derived by substituting the scaling factors for 1 H (slope: -1.0651; intercept: 31.8547) and 13 C (slope: -1.0275; intercept: 185.7787) into the following equation; where δ is the scaled chemical shift value relative to TMS and σ is the computed isotropic value. The computed 15 N nuclear shielding tensors were converted into chemical shifts relative to the calculated 15 N chemical shift of ammonia. Figure S1. Pressure-dependence of NMR spectra of A40. The HCO plane of HNCO spectra obtained at 1, 500, 1000, 1500 and 2000 bar are shown. Note the direction of peak displacement, which are highlighted by dotted arrows.        Cartesian coordinates of the energy minimized conformation of model pentapeptide Glu-Gly-Arg-Gly-Asp without the salt bridge between Arg and Glu/Asp, used for the Arg side chain nuclear shielding tensor calculations.  Table S2. Frequency of A40 conformers containing Arg5-based salt bridges, as obtained from analysis of a 30-s long MD trajectory of A40. 12 The frequencies are reported separately for various modes of interaction between the guanidinium and carboxylate groups.