J. Phys. Chem. B 2008, 112, 5795-5802
5795
Analysis and Optimization of Saturation Transfer Difference NMR Experiments Designed to Map Early Self-Association Events in Amyloidogenic Peptides Hao Huang,† Julijana Milojevic,† and Giuseppe Melacini*,†,‡ Departments of Chemistry, Biochemistry, and Biomedical Sciences, McMaster UniVersity, 1280 Main Street West, Hamilton, Ontario L8S 4M1, Canada ReceiVed: December 18, 2007; In Final Form: February 8, 2008
Saturation transfer difference (STD) methods recently have been proposed to be a promising tool for selfrecognition mapping at residue and atomic resolution in amyloidogenic peptides. Despite the significant potential of the STD approach for systems undergoing oligomer/monomer (O/M) equilibria, a systematic analysis of the possible artifacts arising in this novel application of STD experiments is still lacking. Here, we have analyzed the STD method as applied to O/M peptides, and we have identified three major sources of possible biases: offset effects, intramonomer cross-relaxation, and partial spin-diffusion within the oligomers. For the purpose of quantitatively assessing these artifacts, we employed a comparative approach that relies on 1-D and 2-D STD data acquired at different saturation frequencies on samples with different peptide concentrations and filtration states. This artifact evaluation protocol was applied to the Aβ(12-28) model system, and all three types of artifacts appear to affect the measured STD spectra. In addition, we propose a method to minimize the biases introduced by these artifacts in the HR STD distributions used to obtain peptide selfrecognition maps at residue resolution. This method relies on the averaging of STD data sets acquired at different saturation frequencies and provides results comparable to those independently obtained through other NMR pulse sequences that probe oligomerization, such as nonselective off-resonance relaxation experiments. The artifact evaluation protocol and the multiple frequencies averaging strategy proposed here are of general utility for the growing family of amyloidogenic peptides, as they provide a reliable analysis of STD spectra in terms of polypeptide self-recognition epitopes.
Introduction The saturation transfer difference (STD) NMR method was initially proposed to investigate the binding of small ligands to biological macromolecules, including proteins and nucleic acids.1-4 One of the strengths of the STD method is its ability to detect weak binding with great sensitivity.5 In addition, the binding epitopes can also be mapped based on the intensity of the ligand STD signal.3 Only recently has the application of the STD NMR methods been extended to self-recognition mapping in oligomerizing peptide systems.6,7 While this pioneering use of STD experiments opens up new opportunities for investigating the molecular determinants in the formation of toxic soluble oligomers of amyloidogenic peptides,6,7 such a novel use of the STD pulse sequence also poses new experimental challenges. Specifically, it is important to assess to what extent the STD experiments initially developed to map proteinligand interactions can be transferred without modifications to self-associating amyloidogenic peptides. One of the most significant differences between the STD applications to protein/ligand (P/L) complexes and to oligomer/ monomer (O/M) systems is in the method used to introduce selective saturation into the relatively slow-tumbling components of the system (Figure 1a,b). When the STD NMR experiments are applied to O/M interactions (Figure 1b), the selective RF field mainly saturates monomer signals (e.g., typically methyls * To whom correspondence should be addressed. E-mail: melacin@ mcmaster.ca; tel.: (905) 525-9140, ext. 26959; fax: (905) 522-2509. † Department of Chemistry. ‡ Departments of Biochemistry and of Biomedical Sciences.
Figure 1. Comparison between STD experiments as applied to P/L (a) and O/M (b) systems. In the case of a P/L complex, the saturation is introduced through a selective RF irradiation of the large molecular weight component (i.e., the protein or macromolecule), whereas in the case of an O/M system, the saturation is introduced through a selective RF irradiation of the low molecular weight component (i.e., the monomer).
or aromatics) because the exact resonance frequencies of the oligomers are often unknown. Then, the saturation is transferred through chemical exchange of polypeptide chains from the monomers to the oligomers, where it diffuses more effectively due to their slow tumbling. Finally, the saturation is transferred back to the monomers again through chemical exchange, and the saturated 1H spins of the monomer will result in the detected STD signals. In other words, the monomers serve the double purpose of effectively injecting saturation into the oligomers as well as for providing a sensitive and relatively well-resolved spectroscopic handle for signal detection. This situation is very
10.1021/jp7118718 CCC: $40.75 © 2008 American Chemical Society Published on Web 04/15/2008
5796 J. Phys. Chem. B, Vol. 112, No. 18, 2008 different from that encountered in traditional STD experiments on P/L systems, where saturation is introduced directly into the macromolecule through the selective RF irradiation of protein signals without affecting the ligand (Figure 1a). The fundamental differences between the STD approaches to P/L and O/M systems warrant a critical re-examination of the STD methods originally developed for P/L complexes before they are reliably employed in the context of O/M self-associating peptides. Here, we propose a protocol to critically assess and quantify possible artifacts arising from the application of classical STD methods to O/M systems. For the purpose of validating the proposed protocol, we employed the wellcharacterized Aβ(12-28) peptide, which has been previously shown to serve as a model system for the early steps of prenuclear oligomerization in the self-association equilibria that lead to amyloid plaque deposition in Alzheimer’s disease.8 Using dilution and/or filtration, it is possible to prepare Aβ(12-28) NMR samples with different oligomeric distributions,7,8 and therefore, the Aβ(12-28) peptide is ideally suited to test how the STD signal relates to polypeptide self-association. In addition, based on the characterization of the artifacts in STD experiments applied to O/M systems, we propose an alternative saturation transfer strategy for the minimization and/or elimination of these experimental limitations. These results were validated by comparing the STD-based self-recognition maps with those independently obtained through nonselective offresonance relaxation methods, where significant residue-specific artifacts were absent due to the use of an intense off-resonance spin-lock field.9,10 The protocols presented here for the STD artifact analysis and correction will be useful in general to obtain reliable STD data on oligomerizing systems. Materials and Methods Aβ(12-28) Sample Preparation. The Alzheimer’s peptide fragment Aβ(12-28) with a purity of 98% was purchased form Genscript Corp. The 1 mM Aβ(12-28) NMR samples were prepared by dissolving the lyophilized peptide powder in 50 mM acetate buffer-d4 at pH 4.7 in the presence of 10% D2O. After being dissolved in the acetate buffer, the peptide was filtered through Ultrafree-MC Millipore 30 kDa cutoff filters at 5 min intervals and 3000 rpm. The final peptide solution was centrifuged at 2000 rpm to remove any air bubbles. Dilute (i.e., 0.1 mM) Aβ(12-28) samples were prepared by 10-fold dilution of the unfiltered 1 mM Aβ(12-28) sample using 50 mM acetate buffer-d4 at pH 4.7. 1-D and 2-D STD NMR Experiments. All NMR data were collected using a Bruker Avance 700 MHz spectrometer equipped with a 5 mm TCI CyroProbe at 293 K. The pulse sequence used for the STD NMR experiments was implemented as previously described.3,7 In brief, a train of Gaussian-shaped pulses of 50 ms each and separated by a 1 ms interpulse delay was applied to introduce selective saturation. The train of Gaussian pulses was preceded by a 100 ms delay in all STD experiments. The strength of each saturating Gaussian pulse was 116.1 Hz with a 1% truncation and 1000 digitization points. The magnetization trajectories during such a Gaussian pulse were analyzed through Bloch equation simulations implemented in the NMRSim software. For these simulations, a one-spin system was employed, and it was assumed that before the Gaussian pulse, the magnetization was represented by a unit vector aligned along the +z-axis of the laboratory frame. All magnetization losses due to relaxation during the Gaussian pulse were neglected in the simulations. The optimal saturation time was estimated through 1-D STD experiments acquired at different saturation times and frequen-
Huang et al. cies. Separate saturation transfer reference (STR) experiments also were acquired under the same experimental conditions. For each saturation frequency, two different spectral regions were used to determine the optimal saturation time. A total saturation time of 2 s was selected as the minimal saturation time that gives the maximum ISTD/ISTR ratio for all saturation frequencies, and therefore, it was used in all 2-D STD experiments. The number of scans and dummy scans in the 1-D STD experiments were 512 and 64, respectively. For the 1-D STR experiments, 64 scans and 64 dummy scans were acquired. In all experiments, the on-resonance irradiation was performed at 0.75 ppm (i.e., methyl saturation) or at 7.26 ppm (i.e., aromatic saturation). The off-resonance control irradiation for the reference spectra was performed at 30 ppm. The STD spectrum was obtained by phase cycling subtraction of the on-resonance and off-resonance data acquired in interleaved mode. The strength of the 2-D TOCSY spin-lock was 10 kHz, and water was suppressed prior to acquisition through a 3-9-19 Watergate gradient-echo.11 In the 2-D STD experiments, the spectral widths for both dimensions were 8389.26 Hz and were digitized by 200 t1 and 1024 t2 complex points. The number of scans and dummy scans in the 2-D STD experiments were 16 and 128, respectively. Separate 2-D STR experiments were acquired with 8 scans and 128 dummy scans. In the processing of all 2-D data sets, a 90° phase shifted squared sine bell window function was employed for both dimensions prior to zero filling. 2-D Nonselective Off-Resonance NMR Experiments. The nonselective off-resonance relaxation data were acquired using a 2-D TOCSY pulse sequence as the detection block as previously described.9,10 The off-resonance spin-lock with a trapezoidal shape including two adiabatic pulses of 4 ms duration was applied at an angle of 35.5° to ensure optimal NOE (Nuclear Overhauser Effect)/ROE (Rotational frame Overhauser Effect) compensation in the spin-diffusion limit. The total spin-lock durations were 13 and 88 ms. The strength of the off-resonance spin-lock was 8.25 kHz. The interscan delay between the end of the acquisition and the start of the first adiabatic pulse was 2 s. The spectral widths for both dimensions were 8389.26 Hz with 256 t1 and 1024 t2 complex points, respectively. Water suppression was achieved through a Watergate scheme implemented with the binomial 3-9-19 pulse train.11 For each experiment, 16 scans and 128 dummy scans were employed. For the 13 ms spin-lock time that resulted in intense signals, only one data set was obtained, while at 88 ms, two replicas were collected. All 2-D replica sets were co-added to increase the S/N ratios and processed with Xwinnmr (Bruker, Inc.). The 2-D cross-peak intensities were measured with Sparky 3.11112 through Gaussian line fitting and determination of the fit heights. The error in the intensity measurements for each data set was estimated as 5 times the noise standard deviation. After the addition of the replicate spectra, the error of the individual spectra was scaled up proportionally to the square root of the total number of scans. For all residues, except V12, HR,i-H N,i cross-peaks were used to monitor the HR relaxation, whereas for V12, a HR,12-HMe,12 cross-peak was used. Because of the overlap of its degenerate HR protons, G25 was omitted from the analysis of the off-resonance relaxation rates as previously explained.9,10 The nonselective off-resonance relaxation rates in s-1 were computed from the experimental fit heights though the equation
R35.5°,ns ) ln
height at 13 ms /(0.088s - 0.013s) (fitfit height at 88 ms)
(1)
STD Experiments for Amyloidogenic Peptides
J. Phys. Chem. B, Vol. 112, No. 18, 2008 5797
The nonselective off-resonance relaxation rates calculated based on eq 1 are within error of the off-resonance relaxation rates calculated using five different spin-lock times.9,10 The obtained rates and errors were normalized with respect to the largest measured rate. Results and Discussion Artifacts in STD Experiments Applied to O/M Systems. A common feature of STD experiments applied to P/L and O/M systems is that the low molecular weight states (i.e., the ligand or the monomer in Figure 1) are used to detect and quantify the extent of saturation transfer. However, in the P/L application of the STD approach, the saturation is introduced through the highest molecular weight component of the binding equilibrium (i.e., the macromolecule, Figure 1a), while in the O/M STD, it is the low molecular weight monomer that is directly saturated (Figure 1b). This fundamental difference between the STD approaches to P/L and O/M systems implies that when STD methods are applied to O/M equilibria, the offset effects of the saturating RF field as well as monomer intramolecular crossrelaxation must be assessed very carefully before any reliable conclusions about self-recognition can be drawn from the STD spectra. Because of the large excess of monomers with respect to oligomers in samples suitable for NMR investigations, these effects can possibly lead to significant artifacts, and their indepth characterization is warranted. Another major difference between STD applications to P/L and O/M systems arises from the continuous dynamic exchange of polypeptide chains between the oligomeric and the monomeric states. This means that the intraoligomer saturation may not be able to fully diffuse from the entry point to the rest of the whole oligomeric 1H pool because of competition with the off-exchange pathways. Such an off-exchange process effectively removes polypeptide chains from the oligomers before they are fully saturated and brings them to the monomeric state, where cross-relaxation is significantly less efficient. This scenario is very different from that encountered in P/L systems, where the saturated macromolecule typically remains wellfolded within the time scale of the saturation period. In addition, in the O/M systems, once the polypeptide chain is in the monomeric state, hydrogen exchange with the bulk solvent provides another efficient path to dissipate part of the saturation accumulated in the oligomeric state. Because of these inefficiencies in the spin-diffusion of saturation within the oligomers, it is therefore highly likely that the STD results for O/M systems are biased by the choice of the saturation frequency. For instance, STD experiments with methyl selective saturation may result in a different STD profile from that obtained through aromatic selective saturation since in each case spin-diffusion extends only to the vicinity of the saturated groups. General Strategy for Artifact Characterization in STD Experiments Applied to O/M Systems. The previous analysis reveals that the potential artifacts in STD measurement of O/M systems arise either from the monomer (i.e., offset-effects and monomer intramolecular cross-relaxation) or from the oligomers (i.e., partial spin-diffusion). The offset-effects are effectively assessed through simulations based on the Bloch equations, while the remaining monomer-related biases are evaluated by acquiring STD data under conditions that favor the monomeric state (i.e., diluted or filtered samples).7,8 The extent of oligomerrelated biases (i.e., partial spin-diffusion) is gauged by acquiring STD spectra at different RF saturation frequencies (e.g., methyl and aromatic regions), and for each frequency, the STD experiment is repeated both before and after filtration so that
Figure 2. Bloch simulations of the effect of the Gaussian RF pulse typically used for introducing selective saturation in STD experiments, as described in the text. The phase of this pulse is +x. In all simulations, it is assumed that before the Gaussian pulse is applied, the magnetization is described by a unit vector aligned along the +z-axis. (a) The solid line shows the inversion profile (i.e., arbitrary units vs resonance offset from the pulse carrier frequency). This inversion profile takes into account only the longitudinal magnetization after the completion of the Gaussian pulse. The dashed line corresponds to the plot of the minimum absolute value of the z-component of the magnetization reached during the Gaussian pulse under the assumption of an adiabatic trajectory (i.e., the magnetization remains aligned with the effective field throughout the duration of the Gaussian pulse). (b) The solid line shows the time evolution during the 50 ms Gaussian pulse of the x-component of the magnetization. The Mx computation was repeated at five different frequency offsets at a field of 700 MHz. The dashed lines refer to the x-magnetization components obtained under the assumption of an adiabatic trajectory at the same five frequency offsets. (c) Same as panel b except for the z-component of the magnetization.
monomer contributions to the STD effect are subtracted out. The frequency dependence of the oligomer-specific STD effects reveals then the extent of the partial spin-diffusion artifacts. Evaluation of Offset Effects. The selectivity of the saturating RF field is reliably modeled based on the Bloch equations. Figure 2 shows the results of simulations carried out at a field of 700.23 MHz for a 50 ms Gaussian pulse with a 1% truncation and a maximum strength of 116.1 Hz, as commonly used in the selectively saturating pulse train of STD experiments.3,7 The initial magnetization is assumed to be aligned along the +zaxis with a magnitude of one arbitrary unit. Figure 2a (solid line) shows that the inversion profile describing the z-component of the magnetization after the Gaussian pulse is limited to a relatively narrow frequency range of approximately (70 Hz (i.e., approximately (0.1 ppm) centered at the carrier frequency
5798 J. Phys. Chem. B, Vol. 112, No. 18, 2008
Huang et al.
Figure 3. Each panel shows a comparison of 1-D STD spectra at 1 mM (top trace) and 0.1 mM Aβ(12-28) (bottom trace) in 50 mM acetate buffer-d4 (pH 4.7) at 293 K and 10% D2O. For all spectra, a total saturation time of 3 s and a 3 Hz exponential multiplication window function were employed. (a and b) Spectra obtained with methyl saturation at 0.75 ppm (high field methyl of V18). Panel a shows the expansions of the methyl region, while panel b illustrates an expansion of the remaining aliphatic region. (c and d) Spectra obtained with aromatic saturation at 7.26 ppm (Hδ1,2 of F20). Panel c shows the expansions of the aromatic and amide regions, while panel d illustrates an expansion of the aliphatic region. In all panels (a-d), the intensity of the STD spectrum at 0.1 mM Aβ(12-28) is normalized according to the ratio of the saturated signals in the two samples (i.e., 1 mM and 0.1 mM Aβ(12-28)).
of the selective pulse. However, the width of the inversion profile of a single Gaussian pulse provides only a lower limit for the selectivity window of the saturating train of pulses. This is because in typical STD pulse sequences,3,7 50 ms Gaussian pulses are repeated with an interleaving delay of 1 ms, and therefore, the magnetization is subjected for ∼98% of the time to the effective field generated by the repeated selective pulses. As a result, the selectivity of such pulse trains is affected also by the trajectory of the magnetization during the repeated Gaussian pulses (Figure 2b,c) and not just by its end point after each single pulse, as represented by the inversion profile (Figure 2a). For the purpose of evaluating the magnetization trajectories during the Gaussian pulses, it is useful to consider that at offsets outside the inversion bandwidth of the Gaussian pulse (i.e., >0.1 ppm and < -0.1 ppm), the evolution of the magnetization is essentially adiabatic because the magnetization remains closely aligned with the effective field throughout the duration of the Gaussian pulse as shown in Figure 2b,c. Under these conditions, the selectivity of the train of Gaussian pulses can be appreciated by reporting the minimum z-component of the magnetization reached during the Gaussian pulse (Figure 2a, dashed line), revealing that offset-effects extend up to (0.5 ppm from the carrier frequency of the saturating pulse. This prediction is fully supported by the experimental STD data. For instance, Figure 3 displays the 1-D STD spectra acquired for Aβ(12-28) samples at high (i.e., 1 mM, top traces) and low (i.e., 0.1 mM, bottom
traces) concentrations. The 10-fold dilution shifts the oligomerization equilibiria toward the low molecular weight species so that at 0.1 mM, the Aβ(12-28) peptide is mainly in the monomeric state under the conditions used, while at 1 mM, the Aβ(12-28) solution contains detectable amounts of soluble oligomers.9,8,13 Peaks arising from offset effects are expected to be to a large extent dilution independent as they are not related to self-association. In Figure 3, several intense dilution independent peaks are observed well beyond the (0.1 ppm saturation bandwidth predicted by the simple Gaussian inversion profiles. This observation applies to both data obtained using methyl saturation at 0.75 ppm (Figure 3a) and aromatic saturation at 7.26 ppm (Figure 3c). For instance, the signals at 0.97 and 7.65 ppm in Figure 3a,c, respectively, are fully consistent with the expanded saturation bandwidth predicted by the adiabatic trajectories occurring beyond the (0.1 ppm offsets (Figure 2a, dashed line). Evaluation of Intramonomer Cross-Relaxation. Possible contributions from intramonomer cross-relaxation were initially evaluated by the comparative analysis of 1-D STD spectra acquired at different peptide dilutions (Figure 3). At 1 mM, the Aβ(12-28) solution results in significant STD signals (Figure 3b-d, top traces) as expected based on the chemical exchange between monomers and oligomeric pool. When the same experiment was repeated for the 0.1 mM Aβ(12-28) solution, some weak STD signals still were observed (Figure 3b-d, bottom traces). These residual peaks observed at 0.1 mM Aβ-
STD Experiments for Amyloidogenic Peptides (12-28) were more than 0.5 ppm apart from the saturation frequencies, and hence, they cannot be accounted for by offset effects even when considering adiabatic trajectories (Figure 2). Such peaks are therefore an indication of intramonomer crossrelaxation. Indeed, several STD peaks that remain after dilution correspond to spins in the vicinity of the saturated groups. For instance, the V18 HR (∼4 ppm) and V18 Hβ (∼1.9 ppm) peaks in Figure 3b are caused by the monomer cross-relaxation arising from the selective saturation of the vicinal V18 methyl hydrogens. Similar considerations apply also to the case of the aromatic saturation, where irradiation of the F20 Hδ1,2 protons results in STD signals for the nearby F20 and F19 Hβ1,2 spins even at 0.1 mM Aβ(12-28) (Figure 3d). The intramonomer cross-relaxation contributions interfere with the measurement of the STD effects arising from selfrecognition and should therefore be taken into account when analyzing STD NMR data for O/M equilibria in terms of selfassociation distributions, especially under conditions that increase the average correlation time of the monomer (i.e., long peptide sequences, low temperatures, and high viscosity). For instance, 2-D STD-TOCSY experiments are potentially useful to map self-recognition at residue resolution by measuring the STD effect for different HR protons, as shown in Figure 4a for a sample of 1 mM Aβ(12-28).7 However, when the HR STDs are obtained through methyl saturation, the highest STD values map to the hydrophobic residues L17 and V18, whereas in the case of aromatic saturation, the maximum STDs correspond to the two aromatic residues F19 and F20 (Figure 4a). Such a saturation frequency dependence of the HR STD residue distribution points to the possible presence of significant intramonomer cross-relaxation contributions, which are independently quantified by measuring HR STD values for monomeric Aβ(12-28) samples. While dilution serves this purpose for simple 1-D STD spectra, the 2-D STD-TOCSY spectra are not sensitive enough for measurements at a peptide concentration of ∼0.1 mM, even with the latest generation of high-field cryoprobes. We therefore opted to generate monomeric Aβ(12-28) samples through 30 kDa cutoff filtration. This method has been previously employed to prepare Aβ(12-28) samples devoid of large molecular weight oligomers7,8 and results only in partial dilution, while stabilizing the monomeric peptide through the elimination of oligomeric seeds that may nucleate and promote further oligomerization.7 Furthermore, the highly sensitive nonselective off-resonance relaxation experiments,9,10 which are also designed to probe peptide self-association, demonstrate that the residue distribution of the self-association maps is very similar in the filtered and in the diluted peptides (Figure 5). Filtration is therefore suitable to estimate the monomer contribution to the HR STD measurements, and Figure 4b shows the residue distribution of normalized HR STD values for a filtered Aβ(12-28) sample using both methyl and aromatic saturation. Consistently with the 1-D STD results (Figure 3), Figure 4b shows residual HR STD effects for the saturated residues (i.e., V18 for methyl saturation and F19 and F20 for aromatic saturation). Such residual HR STD values obtained after filtration are a direct result of intramonomer cross-relaxation and are useful to correct the STD measurements in the unfiltered and concentrated samples (Figure 4a). Figure 6 shows the HR STD residue distribution obtained from Figure 4a after subtracting the STD contributions measured for the filtered sample (Figure 4b). Interestingly, even after this correction, the saturation frequency dependence of the HR STD distribution is not fully eliminated (Figure 6). For instance, significant differences
J. Phys. Chem. B, Vol. 112, No. 18, 2008 5799
Figure 4. Plot of the HR (ISTD/ISTR)/max(ISTD/ISTR) ratios vs residue number in Aβ(12-28). Panel a reports the (ISTD/ISTR)/max(ISTD/ISTR) ratios for aggregated 1 mM Aβ(12-28) at different saturation frequencies. The red curve depicts the (ISTD/ISTR)/max(ISTD/ISTR) ratios obtained by saturation at 0.75 ppm, while the black curve depicts the (ISTD/ISTR)/ max (ISTD/ISTR) ratios obtained by saturating the aromatic region at 7.26 ppm. Panel b reports the (ISTD/ISTR)/max(ISTD/ISTR) ratios for a filtered 1 mM Aβ(12-28) sample at different saturation frequencies. The color cording is as in panel a. The (ISTD/ISTR) ratios were normalized according to the maximum HR ISTD/ISTR ratio measured for the aggregated 1 mM Aβ(12-28) sample. All measurements were performed at 700 MHz using a TCI CyroProbe at 293 K. Further details are available in the text.
remain between the V17 STD ratios measured with methyl and aromatic saturations (Figure 6). This residual saturation frequency dependence points to another possible artifact in the STD experiments applied to O/M systems (i.e., partial spindiffusion within the oligomers). Evaluation of Partial Intraoligomer Spin Diffusion. A possible simple cause for the incomplete spin-diffusion is an insufficiently long saturation period. We therefore measured the buildup of saturation as a function of the saturation time for both methyl and aromatic saturation frequencies (Figure 7). In both cases, after 2 s of saturation, as employed for the 2-D STD-TOCSY spectra shown in Figures 4 and 6, a plateau is reached. This observation indicates that the partial spin-diffusion within the oligomers leading to the saturation frequency dependence illustrated in Figure 6 cannot be simply explained by the duration of the saturation period. Another more likely cause of the saturation frequency dependence observed in Figure 6 is the exchange of polypeptide chains between high and low molecular weight pools occurring in a time scale that is fast with respect to the saturation period. As a result, polypeptide
5800 J. Phys. Chem. B, Vol. 112, No. 18, 2008
Huang et al.
Figure 5. Linear correlation between the effects of filtration and dilution on the HR nonselective off-resonance relaxation rates. The horizontal axis refers to the normalized differences between the HR R35.5°,ns relaxation rates measured for a 1 mM Aβ(12-28) sample before and after filtration, while the vertical axis refers to the normalized differences between the HR R35.5°,ns relaxation rates before and after 10-fold dilution of a 1 mM Aβ(12-28) sample. Each point corresponds to a different residue of Aβ(12-28). The slope of the correlation line is 0.92 ( 0.05, and the correlation coefficient is 0.982. All rates were measured at 700 MHz using a TCI CyroProbe at 293 K. Further details are available in the text.
Figure 7. Normalized ISTD/ISTR ratios as a function of the total saturation time for methyl (a) and aromatic (b) saturation frequencies. The selective saturation frequencies are as in Figure 3a-d. ISTD/ISTR ratios were measured for selected peaks at 6.97, 3.77, and 0.75 ppm, which are representative of aromatic, aliphatic, and methyl protons, respectively. The ISTD/ISTR ratios were quantified only for peaks in frequency regions different from the saturated area. The ISTD/ISTR ratios of each selected peak were normalized relative to the highest ISTD/ISTR value obtained for that peak at a given saturation frequency.
Figure 6. Plot of the difference in the HR ISTD/ISTR ratios between the aggregated and the filtered Aβ(12-28) samples. The red curve refers to the difference in the (ISTD/ISTR)/max(ISTD/ISTR) ratios obtained by saturation at the methyl resonances, while the black curve depicts the difference in the (ISTD/ISTR)/max(ISTD/ISTR) ratios obtained by saturating the aromatic region. All ratio differences were normalized relative to the highest value obtained at each saturation frequency.
chains exchange from the oligomeric to the monomeric state before the saturation is fully propagated throughout the oligomers far from the spins, which are directly saturated by the Gaussian train of pulses. In addition, once the polypeptide chains are in the monomeric state, they are mostly unstructured, and therefore, fast proton exchange between labile groups and water may contribute an additional relaxation pathway that competes with the saturation buildup occurred when the polypeptide chains were part of the oligomeric pool. Simple Experimental Strategy to Minimize STD Artifacts. A possible solution to the partial spin-diffusion problem is to combine STD information at different saturation frequencies. For instance, the normalized STD data of Figure 6 acquired at the methyl and aromatic saturation frequencies can be averaged
and then normalized again to the highest average STD value (Figure 8, solid circles). The highest values of such methylaromatic average STDs (abbreviated as maa-STDs) map well to the central hydrophobic core of Aβ (i.e., L17VFFA21), which has been shown independently by mutational analysis to be a key determinant of self-recognition.8,9,14,15 In addition, the averaging process also minimizes the biases arising from the intramonomer cross-relaxation, and indeed, the normalized distribution obtained by averaging the uncorrected STD plots of Figure 4a is within error very similar to that obtained by correcting for the filtered STD values (Figure 8, open circles). In other words, the methyl-aromatic averaging (maa) also obviates the need to acquire STD spectra for monomeric samples (i.e., diluted or filtered). Furthermore, the maa HR STD distribution is corroborated by its correlation with the HR nonselective off-resonance relaxation rates, which provide an independent self-recognition map9,10 (Figure 9). The linear correlation shown in Figure 9 has a slope of 1.17 ( 0.11 and a correlation coefficient of 0.94, confirming that the saturation frequency averaged HR STD values (i.e., maa-STD) and the nonselective off-resonance relaxation (ns-ORR) experiments result in very similar self-association maps. maa-STD vs ns-ORR Experiments in Peptide SelfRecognition Mapping. While the self-recognition maps provided by the maa-STD and the ns-ORR approaches are
STD Experiments for Amyloidogenic Peptides
J. Phys. Chem. B, Vol. 112, No. 18, 2008 5801 Conclusion
Figure 8. Effect of maa on the ISTD/ISTR ratios. Solid circles depict the HR (ISTD/ISTR)/max(ISTD/ISTR) ratios obtained by averaging the filtration corrected values measured with methyl and aromatic saturation (Figure 6). Open circles refer to the HR (ISTD/ISTR)/max(ISTD/ISTR) ratios obtained by averaging the filtration uncorrected values measured with methyl and aromatic saturation (Figure 4a). All plotted values were normalized to the maximum obtained for a specific data set.
Figure 9. Correlation between the normalized ns-ORR rates obtained as previously explained9,10 and the maa-STDs (i.e., the HR (ISTD/ISTR)/ max (ISTD/ISTR) ratios calculated by averaging the aromatic and methyl saturated HR (ISTD/ISTR)/max(ISTD/ISTR) values measured for the aggregated 1 mM Aβ(12-28) sample) (Figure 8, open circles).
comparable, these two NMR methods designed to probe peptide oligomerization at residue resolution are complementary to each other in several other ways. First, ORR experiments, unlike STD experiments, require relatively strong (∼8 kHz) spin-lock fields that are still implementable in cryoprobes but may result in sample heating, especially at high salt concentrations. However, it is the strength of the spin-lock field in the ns-ORR pulse sequences that avoids biases in the residue-specific mapping of self-recognition, unlike STD spectra, where the arbitrary choice of saturation frequency may bias the self-recognition map, thus requiring the acquisition of multiple STD data sets with saturation at different offsets. Nevertheless, maa-STD experiments, unlike ns-ORR experiments, have the advantage of providing reliable data also for glycine residues.9 In conclusion, the choice of maa-STD versus ns-ORR may depend on the sample conditions (e.g., salt concentration and number of Gly residues in the peptide) and on the NMR hardware (e.g., tolerance to strong spin-lock fields with a duration of several milliseconds).
The analysis of the STD method as applied to O/M peptide systems has revealed three major sources of potential biases: offset effects, intramonomer cross-relaxation, and partial spindiffusion within the oligomers. These artifacts were quantitatively evaluated for the Aβ(12-28) model peptide through a comparative protocol that relies on 1-D and 2-D STD data acquired at different saturation frequencies on samples with different peptide concentrations and filtration states. We found that all three types of artifacts should be taken into consideration to obtain reliable STD data. Offset-effects extend well beyond the selectivity expected for a single Gaussian pulse and typically span a window of up to (350 Hz from the irradiation carrier frequency due to the presence of adiabatic magnetization trajectories during a train of saturating Gaussian pulses. Within this (350 Hz window, no STD data can be measured. Outside this frequency range affected by offset effects, our analysis reveals that even when the saturation occurs at frequencies in the methyl or aromatic regions, intramonomer cross-relaxation artifacts extend to the HR STDs, which are typically used to map self-recognition at residue resolution. However, these biases are significantly minimized by averaging the STD ratios measured at different saturation frequencies (i.e., maa-STD). The maa-STD approach also has the advantage of minimizing the impact of the artifacts arising from partial spin-diffusion effects within the oligomers, providing results comparable to those independently obtained through nonselective off-resonance relaxation methods. The bias assessment protocol and the multiple frequency averaging strategy proposed here will facilitate the reliable analysis of STD spectra in terms of polypeptide self-recognition maps. These optimized STD approaches open new opportunities for probing at residue resolution the oligomerization epitopes of Aβ and other amyloidogenic peptides, such as those belonging to the diabetes-related amylin family.16,17 In addition, STDbased self-recognition mapping methods provide a new tool to characterize oligomerization for other peptide classes, such as that of antimicrobial polypeptides for which self-association is a critical aspect of their function.18 Last but not least, the protocols outlined here are potentially relevant also for characterizing the interactions between peptides and membranes, mimicked by micelles or vesicles.16,19 However, this application is limited to systems that bind membranes only weakly (i.e., KD g µM range) and that are amenable to the presence of excess free peptide in solution. Acknowledgment. This work was supported by an NSERC Discovery Grant to G.M. Dr. Alex Bain, Dr. Paul Berti, and A. V. Raditsis are thanked for helpful discussions. G.M. is grateful to the HSFC for a Maureen Andrew award. References and Notes (1) Mayer, M.; Meyer, B. Angew. Chem., Int. Ed. 1999, 38, 1784. (2) Klein, J.; Meinecke, R.; Mayer, M.; Meyer, B. J. Am. Chem. Soc. 1999, 121, 5336. (3) Mayer, M.; Meyer, B. J. Am. Chem. Soc. 2001, 123, 6108. (4) Mayer, M.; James, T. L. J. Am. Chem. Soc. 2002, 124, 13376. (5) Stockman, B. J.; Dalvit, C. Prog. NMR Spectrosc. 2002, 41, 187. (6) Narayanan, S.; Reif, B. Biochemistry 2005, 44, 1444. (7) Milojevic, J.; Esposito, V.; Das, R.; Melacini, G. J. Am. Chem. Soc. 2007, 129, 4282. (8) Jarvet, J.; Damberg, P.; Bodell, K.; Go1ran Eriksson, L. E.; Garslund, A. J. Am. Chem. Soc. 2000, 122, 4261. (9) Esposito, V.; Das, R.; Melacini, G. J. Am. Chem. Soc. 2005, 127, 9358.
5802 J. Phys. Chem. B, Vol. 112, No. 18, 2008 (10) Milojevic, J.; Esposito, V.; Das, R.; Melacini, G. J. Phys. Chem. B 2006, 110, 20664. (11) Sklenar, V.; Piotto, M.; Leppik, R.; Saudek, V. J. Magn. Reson., Ser. A 1993, 102, 241. (12) Goddard, T. D.; Kneller, D. G. Sparky 3.111; University of California, San Francisco: San Francisco, 2005. (13) Mansfield, S. L.; Jayawickrama, D. A.; Timmons, J. S.; Larive, C. K. Biochim. Biophys. Acta 1998, 1382, 257. (14) Tjernberg, L. O.; Naslund, J.; Lindqvist, F.; Karlstrom, A. R.; Thyberg, J.; Terenius, J.; Nordstedt, C. J. Biol. Chem. 1996, 271, 8545.
Huang et al. (15) Wurth, C.; Guimard, N. K.; Hecht, M. H. J. Mol. Biol. 2002, 319, 1279. (16) Mascioni, A.; Porcelli, F.; Ilangovan, U.; Ramamoorthy, A.; Veglia, G. Biopolymers 2003, 69, 29. (17) Brender, J. R.; Ulrich, H. N.; Du¨rr, U. H.; Heyl, D.; Budarapu, M. B.; Ramamoorthy, A. Biochim. Biophys. Acta 2007, 1768, 2026. (18) Dhople, V.; Krukemeyer, A.; Ramamoorthy, A. Biochim. Biophys. Acta 2006, 1758, 1499. (19) Meinecke, R.; Meyer, B. J. Med. Chem. 2001, 44, 3059.
