Both the cis-trans equilibrium and isomerization dynamics of a single proline amide modulate β2-microglobulin amyloid assembly

Chemical Synthesis of Human β2m and Analogs by Native Chemical Ligation.

WT β2m was assembled by “one pot” native chemical ligation of three unprotected peptide fragments (Fig. 2 and SI Appendix, Figs. S1 and S2) (21). The peptide segments were synthesized using an “in situ neutralization” protocol for Boc-based solid phase peptide synthesis (22) and subsequently purified to homogeneity by reverse-phase HPLC (RP-HPLC). Native chemical ligation of the C terminal and middle fragments was followed by thiazolidine (Thz) deprotection, ligation with the N-terminal peptide, and formyl-deprotection of tryptophans. After RP-HPLC purification, formation of the native disulfide bond by air-oxidation, and further RP-HPLC purification, the protein was folded by dialysis against 25 mM sodium phosphate buffer (pH 7.5). High purity of synthetic β2m was verified by analytical RP-HPLC and high-resolution mass spectrometry (Fig. 2B). The protein adopts a native fold as judged by CD spectroscopy (Fig. 3A) and ¹H-NOESY NMR spectroscopy (SI Appendix, Fig. S8).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Chemical synthesis of WT β2m and its analogs. (A) The synthetic strategy involved one pot native chemical ligation of three polypeptide fragments. (B) Reverse-phase HPLC analysis and electrospray ionization (ESI) mass spectrum of chemically synthesized β2m (1–99) (SI Appendix, Fig. S2). Analogous analytical characterization of the fluoroproline-containing β2m variants is presented in SI Appendix, Figs. S3–S6.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Properties of chemically synthesized WT β2m and three chemical analogs. (A) Far-UV CD spectra at 37 °C. (B) Thermal denaturation monitored by CD. (C) Chemical denaturation with GdmCl. A two-state folding mechanism was assumed to compare the data for different analogs. The data are summarized in Table 1. (D) Fluorescence of ANS dye [concentration (c) 10 μM] upon addition of WT β2m and three analogs (c protein 10 μM). (E) Size-exclusion HPLC of samples (c 40 μM) incubated for 2 wk at pH 7.5, 37 °C. (F) Kinetics of oligomerization of the variants (c 40 μM) in the presence of Cu²⁺ (c 80 μM) at pH 6.8. (G) The pH-dependent solubility of the synthetic proteins upon incubation for 2 wk (initial c 40 μM in 0.2 M Na.phosphate, 0.1 M Na.citrate buffers). For each plot, data for WT β2m are shown in black, data for [4S-fpr32]β2m are shown in green, data for [4R-Fpr32]β2m are shown in blue, and data for [F₂Pro32]β2m are shown in red.

The three β2m analogs containing the noncanonical amino acids 4R-Fpr, 4S-fpr, and F₂Pro at position 32 were synthesized analogously using N-Boc-protected 4-fluoroprolines instead of proline (see SI Appendix and Figs. S3–S6 for analytical data). Near-UV CD and ¹H-NOESY spectra (SI Appendix, Figs. S7–S10) show that the variants adopt the same overall tertiary structure as WT β2m. As expected for such close structural analogs, only small differences localized near residue 32 are observed in the ¹H-NOESY spectra (SI Appendix, Figs. S8–S10).

Fluoroprolines Modulate β2m Stability.

Far-UV CD spectra and thermal denaturation curves for WT β2m and the three analogs are depicted in Fig. 3 A and B, respectively. Chemically synthesized β2m has a CD spectrum that is virtually identical to the published spectra of recombinantly expressed protein containing an additional N-terminal methionine (9, 23). Chief features are a strong maximum at 203 nm and a weak minimum at 221 nm. The CD spectra of [4S-fpr32]β2m and [4R-Fpr32]β2m are similar (Fig. 3A). The relative stability of these three proteins was assessed by thermal denaturation. Their apparent melting temperatures (T_m) correlate with the cis-amide conformational preferences of the respective proline derivative at position 32 (Fig. 1D), decreasing in the order [4S-fpr32]β2m > WT β2m > [4R-Fpr32]β2m (Table 1).

The [F₂Pro32]β2m protein deviates from these trends. Its far-UV CD spectrum shows a diminished maximum at 203 nm and a slightly blue-shifted minimum at 218 nm (Fig. 3A). Interestingly, similar features were previously observed for β2m variants containing trans-configured Pro32, including the P5G point mutant and ΔN6, which lacks six N-terminal amino acids (23). Like these proteins, [F₂Pro32]β2m exhibits reduced thermal stability (T_m = 56 °C) and less cooperative thermal denaturation compared with WT β2m (Fig. 3B).

Chemical denaturation (Fig. 3C and SI Appendix, Fig. S13) and folding–unfolding kinetics (SI Appendix, Fig. S14) of the Pro32 variants further illuminate the interplay between cis-trans equilibrium propensities, the free-energy barrier for amide bond isomerization, and protein stability. As shown in Table 1, cis-trans equilibrium preferences correlate with the chemical denaturation midpoint ([D]_50%]) and unfolding rate (k_u), whereas the isomerization barrier is reflected in the cooperativity of denaturation (m value) and refolding kinetics. The Gibbs free energy for unfolding (ΔG_u) is lowest for [F₂Pro32]β2m and highest for [4S-fpr32]β2m.

Because F₂Pro and unmodified proline have comparable cis-trans equilibrium ratios, the destabilizing effects due to incorporation of residue F₂Pro32 are counterintuitive. Although the ¹H-NOESY NMR spectra (SI Appendix, Fig. S10) provide strong evidence that [F₂Pro32]β2m and WT β2m adopt similar folds, the far-UV CD spectra and folding data suggest that F₂Pro32 is more structurally disruptive than the monofluorinated prolines. To gain additional insight into the effects of this substitution, we prepared a derivative of [F₂Pro32]β2m in which seven residues in the hydrophobic core and an adjacent loop (Val9, Ile7, Val27, Gly29, Gly18, Ala15, Phe22) plus two residues flanking F₂Pro32 (Phe30 and Ile35) were ¹⁵N-labeled (SI Appendix, Fig. S11A). The ¹H-¹⁵N resonances in the ¹H-¹⁵N heteronuclear single quantum coherence (HSQC) spectrum of this variant exhibit average chemical shifts that differ by less than 0.1 ppm from the corresponding resonances in WT β2m (SI Appendix, Fig. S11). In addition, ¹⁹F NMR studies showed that the F₂Pro32 amide adopts a cis conformation in the folded protein but a trans conformation when denatured (SI Appendix, Fig. S12). Together, these observations strengthen the conclusion that the average structures of [F₂Pro32]β2m and WT β2m are quite similar.

The destabilization of [F₂Pro32]β2m can be rationalized by additionally considering the energy barrier for cis-trans isomerization (16). Inserting fluorine atoms into the proline five-membered ring reduces the double-bond character of the prolyl amide and lowers the free-energy barrier for cis-trans isomerization. For the model system Ac-Xaa-OMe, the energy barrier decreases in the order Pro > 4R-Fpr ∼ 4S-fpr > F₂Pro (Fig. 1D) (16). Introducing F₂Pro, the most conformationally labile residue in the series, at position 32 of β2m may destabilize the protein by “melting” the BC loop that maintains Pro32 in the cis conformation (Fig. 1B). Consistent with a more dynamic structure and increased exposure of hydrophobic surfaces, [F₂Pro32]β2m binds the environmentally sensitive dye 8-anilino-1-naphthalenesulfonic acid (ANS) to a greater extent than the other variants, giving rise to enhanced fluorescence and a characteristic blue shift of the emission maximum (Fig. 3D). As for the previously published P32G variant (13), minor structural perturbations in the vicinity of residue 32, enhanced by the rapid isomerization kinetics of the F₂Pro32 amide, may be propagated to other parts of the molecule without altering the core structure of the protein. The lower stability of [F₂Pro32]β2m can then be attributed either to this slightly altered, less compact structure and/or to changes in the denatured state.

Aggregation Properties of Chemically Synthesized β2m Analogs.

The oligomerization state of β2m variants was analyzed by size-exclusion chromatography. Upon incubation at pH 7.5 for a period of 2 wk, WT β2m and the monofluoro-proline analogs (40 μM each) behaved as stable monomeric proteins, whereas [F₂Pro32]β2m afforded a population of oligomers (Fig. 3E). Although previous studies on amyloidogenic proteins have shown that partially folded species serve as precursors to aggregates and, ultimately, amyloids (24⇓⇓–27), analysis of these samples by transmission electron microscopy (TEM) showed no evidence for amyloid fibrils even for the least stable [F₂Pro32]β2m protein.

Oligomerization of β2m at neutral pH is known to be promoted by Cu²⁺ ions (6, 7). Although this process is characterized by a several hour lag phase in the case of WT β2m, the P32A mutant produces protein oligomers instantaneously in the presence of metal ions (6). The trans amide conformation adopted by Ala32 apparently preorganizes the protein for tighter Cu²⁺ binding, triggering formation of oligomers. The synthetic proteins also undergo Cu²⁺-induced oligomerization at rates that correlate with their copper affinity and preference for the trans amide conformation (Fig. 3F). Thus, the [4R-Fpr32]β2m analog, which favors the trans-amide and has the highest affinity for Cu²⁺ (K_d = 7.31 ± 0.53 μM), shows the most pronounced fluorescence response in a ThT-binding assay, followed by WT β2m (K_d = 10.3 ± 0.49 μM) and [4S-fpr32]β2m (K_d = 20.2 ± 0.84 μM). Interestingly, no oligomerization was detected with [F₂Pro32]β2m, which has the lowest affinity for Cu²⁺ (K_d = 70.8 ± 4.59 μM) (SI Appendix, Fig. S15). Enhanced flexibility of the BC loop in this variant, as discussed above, may hinder formation of an appropriately preorganized environment for Cu²⁺ coordination even though F₂Pro and unsubstituted proline have similar conformational preferences.

Protein solubility was further examined as a function of pH. The synthetic proteins (40 μM) were incubated at 37 °C without stirring for a period of 2 wk. The concentration of soluble protein in filtered (0.2 μm) aliquots was quantified by absorbance at 280 nm. The resultant solubility curves are shown in Fig. 3G. The most stable analog, [4S-fpr32]β2m, only precipitated below pH 3.5, whereas the other proteins began to form insoluble aggregates at higher pH values. The least stable protein, [F₂Pro32]β2m, displayed the least cooperative behavior. However, in no sample were amyloid fibrils detected above pH 4, either by ThT fluorescence or TEM.

Polymorphism of β2m Amyloids Regulated by cis-trans Pro32 Isomerization.

Incubation at acidic pH (<3) is the most reliable way to produce β2m amyloids (9). Indeed, all synthetic variants spontaneously form amyloid fibrils at pH 2.5 within a few days to a week (Fig. 4A). Monitoring amyloid growth at this low pH value showed that WT β2m, [4S-fpr32]β2m, and [4R-Fpr32]β2m form fibrils at qualitatively similar rates, whereas [F₂Pro32]β2m aggregates more slowly. Nevertheless, intrinsic tryptophan fluorescence suggests that the different Pro32 analogs give rise to distinct fibril polymorphs (SI Appendix, Fig. S16). For example, WT and [4S-fpr32]β2m amyloids afford similar fluorescence spectra, but the fluorescence emission of [4R-Fpr32]β2m amyloids is red-shifted and strongly diminished; amyloids derived from [F₂Pro32]β2m show intermediate values. These results suggest that Trp60, which is buried in WT amyloids (28), is located in very different environments in the other polymorphs as a consequence of altered self-assembly of the respective cross-β structures.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Different polymorphic forms of the amyloids formed by WT β2m and its variants. (A) Substitution of Pro32 by 4S-fpr, 4R-Fpr, or F₂Pro led to formation of fibrils of different morphology as seen in the TEM images. Growth conditions: 50 μM protein in 50 mM citrate buffer, 100 mM NaCl, pH 2.5 for 2 wk at 37 °C with shaking (250 rpm). Relative growth rates: WT β2m ∼ [4S-fpr32]β2m ∼ [4R-Fpr32]β2m > [F₂Pro32]β2m. (Scale bar, 200 nm.) (B) Kinetics of fibril growth of WT β2m and the chemical analogs upon seeding with preformed WT, [4S-fpr32]β2m, [4R-Fpr32]β2m, or [F₂Pro32]β2m amyloid seeds monitored by the increase in ThT fluorescence at pH 2.5. Differences in maximum ThT fluorescence levels for the different protein analogs are presumably due to structurally distinctive fibril surfaces that bind ThT dye to different extents. Control absorption measurements (at 280 nm) of protein remaining in solution confirmed that most of the sample (>90%) had precipitated at the end of each experiment.

We performed seeding experiments to characterize the kinetics of amyloid growth and structural differences in amyloids derived from the different protein variants. The preformed amyloid seeds serve as growth templates, effectively eliminating the nucleation requirement and hence the kinetic lag phase (29). This templating effect is structure-dependent: amyloid seeds of a particular protein polymorph are generally most efficient at inducing fibrils of the same protein polymorph (30, 31). Fibril formation was monitored by a standard ThT fluorescence assay using 40 μM protein in the presence of seeds derived from preformed amyloids of all four variants (16 combinations in total) (Fig. 4B).

The results obtained using cognate seeds, which likely resemble the nucleating species that arise spontaneously during unseeded aggregation, show the same qualitative trend for aggregation rates observed in the absence of seeds, namely WT β2m ∼ [4S-fpr32]β2m ∼ [4R-Fpr32]β2m > [F₂Pro32]β2m. Because [4R-Fpr32]β2m does not aggregate faster than other variants, this trend does not support the simple hypothesis that aggregation is driven by the cis-trans amide bond ratio of residue 32. Other factors, including isomerization kinetics and the altered conformational preferences of the proline ring itself, probably play a role.

In cross-seeding experiments, both growth onset and assembly rate depend on the origin of the seeds. In Fig. 4B, the kinetics of fibril growth of the four studied proteins is quantitatively similar in the case of WT and [F₂Pro32]β2m seeds, whereas [4S-fpr32]β2m and [4R-Fpr32]β2m seeds produced very different growth patterns. Furthermore, analysis of the samples upon completion of the reaction by TEM and intrinsic fluorescence indicates that seeding with WT or [F₂Pro32]β2m amyloid seeds resulted in fibrils with well-defined morphology in all cases (SI Appendix, Fig. S17); they also exhibit a strong intrinsic tryptophan emission signal (SI Appendix, Fig. S18). In contrast, experiments with [4S-fpr32]β2m and [4R-Fpr32]β2m seeds yielded well-defined fibrils for [4S-fpr32]β2m and [4R-Fpr32]β2m, respectively, but gave mixtures of long, straight amyloids and large amounts of amorphous precipitate for the WT β2m and [F₂Pro32]β2m proteins (SI Appendix, Fig. S17).

Although [4S-fpr32]β2m and [4R-Fpr32]β2m seeds have, individually, distinctive templating properties, hybrid seeds derived from amyloid fibrils grown from a mixture of [4S-fpr32]β2m and [4R-Fpr32]β2m proteins recapitulate the properties of WT β2m seeds (SI Appendix, Figs. S19 and S20). Thus, growth kinetics for the four β2m variants seeded with WT and hybrid seeds was equivalent (compare Fig. 4B, Top Left, and SI Appendix, Fig. S20A). Intrinsic tryptophan fluorescence of hybrid amyloids generated from a 1:1 mixture of [4S-fpr32]β2m and [4R-Fpr32]β2m proteins (SI Appendix, Fig. S20B) also suggests that their structural properties approximate those of WT fibrils. At pH 2.5, β2m exists as a mixture of partially structured and extensively unfolded conformers (32). Because the BC loop that maintains Pro32 in the cis conformation will be (partially) denatured under these conditions, the average trans-cis amide bond ratio for an equimolar mixture of 4R-Fpr (6.7:1) and 4S-fpr (2.5:1) effectively mimics that of unmodified proline (4.6:1) (20). Because hybrid fibrils formed by two β2m analogs containing isomeric 4-fluoroprolines behave like WT β2m amyloids, it is unlikely that replacement of hydrogen with a larger fluorine atom modulates the structure and stability of the amyloid fibrils for steric reasons. Instead, polymorphism would appear to have a conformational origin.