DMSO affects Aβ_1–40's conformation and interactions with aggregation inhibitors as revealed by NMR

Abstract

We show via 3D-heteronuclear NMR spectroscopy that Aβ_1–40 adopts a disordered conformational ensemble with fluctuating turns in DMSO_d6. Using NMR, we map the binding sites of three water-insoluble aggregation inhibitors to Aβ_1–40 in DMSO_d6 and discover remarkable differences in Aβ_1–40 recognition by a fourth inhibitor in H₂O versus DMSO_d6.

---

Introduction

Alzheimer's disease is the most common neurodegenerative illness and a major cause of death in developed nations. Despite major efforts, no drugs that effectively prevent or cure the disease are currently available. Many biophysical and preclinical studies require the amyloid β peptide (Aβ) to be monomeric and unfolded at the start of the experiment. However, since Aβ has a strong tendency to oligomerize in aqueous solution, several protocols^1–3 based on inteins or click chemistry have been developed to ensure that it is unfolded and monomeric. However, these approaches require special materials and expertise. The simplest approach is to dissolve lyophilized Aβ in a small amount of organic solvent and then dilute it into aqueous buffer. Out of several organic solvents tested; namely, acetylnitrile, trifluoroethanol, HFIP, DMSO, dichloromethane, water + 0.1% TFA, only DMSO and HFIP maintain Aβ in a monomer state without β-structure.^4,5 Whereas using TFE and HFIP to prepare Aβ stock solutions is a relatively common practice, it was shown that small quantities of TFE⁶ or HFIP⁷ like those present after dilution into aqueous buffer, actually accelerate β-structure formation and aggregation. Unlike fluorinated alcohols, DMSO is a polar aprotic solvent and does not seem to promote the formation of Aβ aggregates or β-structure, either neat or in any dilution with water⁵ or alter Aβ aggregation kinetics.⁴ Other studies of Aβ_1–40 showed it is monomeric in DMSO.⁸

One aim here is to characterize the structure and dynamics of Aβ_1–40 DMSO using high field heteronuclear NMR spectroscopy to test its suitability for preparing stock solutions. One widely accepted hypotheses for Aβ neurotoxicity is that its small oligomers disrupt membrane function, leading to dendrite loss and altered neural signaling.⁹ Moreover, Aβ oligomerization is thought to be stimulated by membrane components such as gangliosides.¹⁰ Therefore, membrane-soluble Aβ inhibitors could have an advantage over water-soluble inhibitors. Recently, some of us¹¹ identified four compounds that block Aβ toxicity in cells. Three of the four are sparingly soluble in water, which thwarts attempts to characterize their binding to Aβ. However, they do dissolve well in DMSO, a polar aprotic solvent which partially mimics the membrane milieu. Like palmitoyloleoyl-phosphatidylcholine, a common neuron membrane component, DMSO can accept but not donate H-bonds. The second objective here is to characterize inhibitor binding to Aβ in DMSO using NMR. In the case of the water-soluble inhibitor, we shall compare how changing the solvent from water to DMSO affects its binding to Aβ.

Materials and methods

Aβ_1–40, ¹³C, ¹⁵N-Aβ_1–40 and DMSO_d6 (99.9% atom D) were purchased from rPeptide and Aldrich, respectively. All NMR spectra were acquired at 30 °C (in DMSO_d6) or 5 °C (aqueous solution) in solvent-matched Shigemi NMR tubes on a Bruker 800 MHz (¹H) spectrometer, equipped with a triple resonance cryoprobe and Z-gradients. The ¹H, ¹³C, ¹⁵N resonance assignments of ¹³C, ¹⁵N-Aβ_1–40 were obtained by analysis of a thorough series of (4,2)D, 3D and 2D NMR spectra as described more fully in the ESI.† For the titration experiments, small volumes of inhibitor compounds, predissolved in solvent to a known concentration determined by weight, were added to the ¹⁵N-Aβ_1–40 sample.

Results and discussion

First we found that adding small amounts of DMSO_d6 to a ¹⁵N-Aβ sample in aqueous buffer produced small, linear changes in the ¹H–¹⁵N HSQC spectra (ESI Fig. S1†), suggesting that small proportions of DMSO_d6 do not notably perturb Aβ's conformational ensemble. In contrast, the ¹H–¹⁵N HSQC spectrum of Aβ_1–40 is strongly altered in neat DMSO_d6 (ESI Fig. S1†). This reflects possible conformational changes or stripping of H₂O from Aβ_1–40 or both. To assign Aβ_1–40 in DMSO_d6, we utilized a standard 3D approach based on intraresidual and sequential ¹³Cα and ¹³Cβ connectivities (ESI Fig. S2†). Next, the 3D (H)CC(CO)NH and 3D H(CCCO)NH and (4,2)D HN(COCA)NH spectra (ESI Table S1†) were analyzed to corroborate the backbone assignments and to assign many side chain resonances. Overall, the assignment process was straightforward except for some nuclei in or adjacent to His residues which showed significant broadening or low intensity in some spectra. The resulting ¹H, ¹³C, ¹⁵N Aβ_1–40 chemical shift values (δ) are reported in ESI Table S2† and have been deposited in the BMRB under accession number 25450. The assigned ¹H–¹⁵N spectrum of Aβ_1–40 in DMSO_d6 is shown in Fig. 1. Integration of the ¹H–¹⁵N peaks revealed that H5, D7, S8, H13, H14 (not observed) and L17 have low intensity and/or broad peaks. These differences could be due to conformational heterogeneity or exchange with residual H₂O. The aliphatic –CH₂– and –CH₃ groups show little δ dispersion; this is consistent with a lack of preferred conformations (see the ¹H–¹³C HSQC spectrum, ESI Fig S3†). The δ of εC¹H₃ of Met 35 is 2.00 ppm, which indicates that the sulfur is reduced. This point is relevant as oxidation of Met35 is believed to inhibit Aβ aggregation.¹² The ¹HN and ¹Hα δ values determined here resemble those reported previously for Aβ_1–40 in 95% DMSO/5% dichloroacetic acid¹³ and Aβ_1–28 in neat DMSO¹⁴ (ESI Fig. S4†). Based on the obtained ¹⁵N, ¹Hα, ¹³CO, ¹³Cα and ¹³Cβ assignments, the suite TALOS-N was used to predict tendencies to adopt secondary structure. Most residues were predicted to lack preferences except residues I31, I32 and M35, which trend toward β-strand conformations. These predictions should be interpreted with caution as DMSO is known to produce large δ changes in ¹³CO, ¹³Cα, ¹³Cβ and ¹H nuclei of short unstructured peptides.¹⁵ In contrast, many inter-residual ¹HN–¹HN NOEs were observed; namely between residues: 2 and 3, 7 and 8, 11 and 12, 17 and 18, 21–24, 23 and 24, 24 and 25, 25–28, 27 and 28, 31 and 32, 31–33, 32 and 33, 36–40 and 37–39 which are evidence for turn or helix-like conformations (ESI Fig. S5†). These findings, obtained using a 800 MHz spectrometer and a 150 ms mixing time, resemble and extend the observations, by Sorimachi & Craik,¹⁴ of ¹HN–¹HN NOE cross peaks between residues 2 and 3, 3 and 4, 4 and 5, 11 and 12, 12 and 13, 22 and 23, 23 and 24, 24 and 25, 25 and 26 in the Aβ fragment (1–28) in experiments performed at 400 MHz with a 250 ms mixing time. The few, weak ¹H aliphatic–¹H aromatic interresidual NOEs observed likely arises from transient contacts and no stable hydrophobic cluster is present (data not shown).

Fig. 1 Superposition of ¹H–¹⁵N HSQC spectra of Aβ_1–40 in DMSO_d6 without (gray peaks) and with (red = negative/blue = positive peaks) application of the ¹H–¹⁵N heteronuclear NOE. ¹H and ¹⁵N chemical shifts are plotted on the x- and y-axes, respectively. The spectra are shown separately in ESI Fig. S6.†

Additional experiments were done to see if the turn or helix-like conformations detected by the NOEs are transient and weak or long-lasting and stable. The heteronuclear NOE ratios, which provides information on the ps–ns local backbone dynamics of Aβ_1–40 in neat DMSO_d6, are shown in Fig. 1 and ESI Fig S6 and S7.† Overall, the polypeptide appears to be flexible; all residues have values much lower than the theoretical limit of 0.85 for complete rigidity. The region composed of Arg5–Gln15 appears to be modestly rigid, residues Lys16 to Val24 are more flexible, and the first 3 and the last 12 C-terminal residues are highly flexible. Notably the heteronuclear NOE ratio values of Aβ_1–40 in DMSO_d6 closely resemble those reported previously¹⁶ for Aβ_1–40 freshly dissolved in aqueous buffer (ESI Fig. S7†). To test for stable secondary structure, a ¹⁵N-Aβ_1–40 sample in DMSO_d6 was diluted into D₂O buffer, the first fast ¹⁵N–¹H SOFAST HMQC showed that only 5 resonances, corresponding to F4, L34, I31, I32 and V40 resisted exchange. These resonances are predicted¹⁷ to have the slowest H/D exchange kinetics (ESI Table S3†). By monitoring peak intensity over time, the residues' exchange rates were determined and protection factors of 17, 7, 7, 4 and 2 for F4, L34, I31, I32 and V40, respectively, were calculated (ESI Fig. S8†). These very low factors (ESI Table S3†) are solid evidence that Aβ_1–40 in DMSO adopts no stable secondary structure. To test if Aβ_1–40 forms amyloid-like conformers in DMSO, we performed ThT fluorescence assays. No significant differences in ThT fluorescence are induced by Aβ_1–40 that had been incubated in DMSO (ESI Fig. S9A†). In contrast, Aβ_1–40 incubated in aqueous solution induces a hundred-fold increase in ThT fluorescence (ESI Fig. S9B†) indicating the formation of amyloid-like conformers.

We utilized our assignments of Aβ_1–40 in DMSO to study how solvent affects the peptide's interaction with four inhibitors. The chemical structures and the NMR spectra of the four inhibitors are shown in ESI Fig. S10.† The spectral resonances observed are consistent with their chemical structures and no significant amounts of impurities were detected.

We took advantage of C1's relatively high water solubility to compare its binding to Aβ_1–40 in different solvent conditions. In aqueous solution, the ¹H–¹⁵N HSQC spectra (ESI Fig. S11A†) show that regions most affected by C1 correspond to HNs belonging to the hydrophobic segments of Aβ_1–40 (Fig. 2A), which coincide more or less with the β-strands in mature Aβ amyloid fibrils.¹⁸ In contrast, the ¹H–¹⁵N HSQC spectra (ESI Fig. S11B†) reveal that in DMSO_d6, C1 binds to the HN groups of the 18 N-terminal residues of Aβ_1–40 (Fig. 2B), a region rich in polar and charged residues. This is an important result as it indicates, for the first time to our knowledge, that solvent can alter the binding mode of an inhibitor to Aβ. This finding has a crucial implication for developing therapeutics; namely, the Aβ binding sites of an inhibitor could well change according to the solvent milieu.

Fig. 2 The weighted average shift changes for Aβ's ¹H and ¹⁵N nuclei in the presence of 1 eq. of C1 (large open blue circles, dotted line), 3 eq. C1 (green diamonds, dashed lines) and 5 eq. C1 (small red circles) in water (A) and in DMSO_d6 (B). The data at sequence #42, 45/46 and 48/49 correspond to the side chain groups of R5, Q15 and N27, respectively.

The effect of C2, C3 and C4 on Aβ's ¹⁵N–¹H resonances were also determined (ESI Fig. S12†). C2 appears to have a modest effect on Aβ's signals; only modest chemical shift changes in the first 10 residues are observed. In contrast, C3-induced ¹⁵N–¹H chemical shift changes are large in magnitude and more extensive. Only residues 2, 3, 4, 12, 15, 18, 19, 29–34 and 37–40 and the sidechains of R5, Q15 and N27 are relatively unaffected by C3. The changes induced by C4 are somewhat larger than those provoked by C3. For all the samples, we attempted to further characterize the binding using NOESY spectroscopy, but no unambiguous NOEs could be assigned. This could be due to the low concentration of Aβ_1–40 (100 μM) and/or heterogeneity in the Aβ_1–40/inhibitor complex. The basis for the inhibitors' divergent modes of interaction with Aβ_1–40 is not immediately apparent from their chemical structures.¹¹ C3 and C4, but not C2 contain a phenyl moiety, as does the recently discovered Aβ inhibitor D373.¹⁹ This phenyl group might account for the more extensive interaction of C3 and C4 with Aβ_1–40 compared to C2 (ESI Fig. S10†).

Conclusions

We have obtained for the first time the essentially complete ¹H, ¹³C and ¹⁵N resonance assignments of Aβ_1–40 in DMSO_d6. Utilizing these results, we determine that Aβ_1–40 adopts a disordered conformational ensemble in DMSO_d6 with only weak turn or helix-like structures. Since Aβ is monomeric⁸ and chiefly unfolded without amyloid-like conformers (this work) in neat DMSO and because low concentrations of DMSO (which would be present after dilution into aqueous buffer) neither increase the rate of Aβ aggregation,⁵ nor significantly perturb the ¹H–¹⁵N HSQC NMR spectrum of Aβ (this work), we conclude that dissolving Aβ in DMSO is a simple yet effective way to prepare unstructured stock solutions of Aβ. There is intense interest in developing small molecule therapeutics for AD,²⁰ which is driven by the ever increasing number of patients and disappointing results from clinical trails for antibody-based drugs.²¹ A key discovery here is that the binding of inhibitor compounds to Aβ_1–40 can be strongly affected by the solvent milieu. C1 binds to Aβ_1–40's hydrophobic regions in aqueous buffer, but in DMSO_d6 it binds exclusively to the N-terminal residues. Based on this finding and considering that Aβ neurotoxicity likely occurs at the membrane,²² we emphasize the importance of testing potential Aβ_1–40 therapeutics in membrane-mimicking solvent media.

Acknowledgements

This work was supported by grants from the Governments of Spain (Project numbers CTQ2010-21567-C02-02, CTQ2011-22514, BFU2010-16297, SAF2013-49179-C2-1-R) and Aragon (B89-2011). We thank Profs. Marta Bruix and Avijit Chakrabartty for critical comments on the manuscript and Prof. Jorge Santoro for advice on NMR pulse sequences.

Notes and references

1. J. J. Bockhorn, K. L. Lazar, A. J. Gasser, L. M. Luther, I. M. Qahwash, N. Chopra and S. C. Meredith, Pept. Sci., 2009, 94, 511–520.
2. A. Taniguchi, Y. Sohma, M. Kimura, T. Okada, K. Ikeda, Y. Hayashi, T. Kimura, S. Hirota, K. Matsuzaki and Y. Kiso, J. Am. Chem. Soc., 2006, 128, 696–697.
3. Y. Sohma, H. Wang, A. Taniguchi, Y. Hirayama, T. Kakizawa, M. Yamaski, H. Mukai and Y. Kiso, Bioorg. Med. Chem., 2011, 19, 3787–37921.
4. S. Snyder, U. S. Ladror, W. S. Wade, G. T. Wang, L. W. Barret, E. D. Matayoshi, H. J. Huffaker, G. A. Krafft and T. F. Holzman, Biophys. J., 1994, 67, 1216–1228.
5. C. L. Shen and R. M. Murphy, Biophys. J., 1995, 69, 640–651.
6. Y. Fezoui and D. B. Teplow, J. Biol. Chem., 2002, 277, 36948–36954.
7. M. R. Nichols, M. A. Moss, D. K. Reed, S. Cratic-McDaniel, J. H. Hoh and T. L. Rosenberry, J. Biol. Chem., 2005, 280, 2471–2480.
8. W. Garzón-Rodríguez, M. Sepulveda-Becerra, S. Milton and C. Glabe, J. Biol. Chem., 1997, 272, 21037–21044.
9. J. J. Palop and L. Mucke, Nat. Neurosci., 2010, 13, 812–818.
10. K. Yanagisawa, J. Neurochem., 2011, 116, 806–812.
11. L. C. López, S. Dos-Reis, A. Espargaró, J. A. Carrodeguas, M. L. Maddelein, S. Ventura and J. Sancho, J. Med. Chem., 2012, 55, 9521–9530.
12. L. Hou, H. Shao, Y. Zhang, H. Li, N. K. Memon, E. G. Neuhaus, J. M. Brewer, I.-J. L. Byeon, D. G. Ray and M. P. Vitek, et al., J. Am. Chem. Soc., 2004, 126, 1992–2005.
13. N. A. Whittemore, R. Mishra, I. Kheterpal, A. D. Williams, R. Wetzel and E. H. Serpersu, Biochemistry, 2005, 44, 4434–4441.
14. K. Sorimachi and D. J. Craik, Eur. J. Biochem., 1994, 219, 237–251.
15. K. Wüthrich, NMR in Biological Research: Peptides and Proteins, Amsterdam North Holland, 1975.
16. R. Riek, P. Güntert, H. Döbeli, B. Wipf and K. Wüthrich, Eur. J. Biochem., 2001, 268, 5930–5936.
17. Y. Bai, J. S. Milne, L. Mayne and S. W. Englander, Proteins Struct. Funct. Gent., 1993, 17, 75–86.
18. A. T. Petkova, Y. Ishii, J. J. Balbach, O. N. Antzutkin, R. D. Leapman, F. Delaglio and R. Tycko, Proc. Natl. Acad. Sci. U. S. A., 2002, 99, 16742–16747.
19. A. Fortner-McKoy, J. Chen, T. Schupbach and M. H. Hecht, J. Biol. Chem., 2012, 287, 38992–39000.
20. A. J. Doig and P. Derreumaux, Curr. Opin. Struct. Biol., 2015, 30, 50–56.
21. S. Salloway, R. Sperling, N. C. Fox and K. Blennow, et al., N. Engl. J. Med., 2014, 370, 322–333.
22. G. M. Shankar and D. N. Walsh, Mol. Neurodegener., 2009, 4, 48–60.

---

Footnote

† Electronic supplementary information (ESI) available: Detailed experimental procedures, tables listing NMR experiments, Aβ_1–40 ¹H, ¹³C, ¹⁵N chemical shifts in DMSO_d6 and H/D exchanges rates and twelve figures. See DOI: 10.1039/c5ra12100k

---