Intermolecular alignment in β2-microglobulin amyloid fibrils.

The deposition of amyloid-like fibrils, composed primarily of the 99-residue protein β2-microglobulin (β2m), is one of the characteristic symptoms of dialysis-related amyloidosis. Fibrils formed in vitro at low pH and low salt concentration share many properties with the disease related fibrils and have been extensively studied by a number of biochemical and biophysical methods. These fibrils contain a significant β-sheet core and have a complex cryoEM electron density profile. Here, we investigate the intrasheet arrangement of the fibrils by means of (15)N-(13)C MAS NMR correlation spectroscopy. We utilize a fibril sample grown from a 50:50 mixture of (15)N,(12)C- and (14)N,(13)C-labeled β2m monomers, the latter prepared using 2-(13)C glycerol as the carbon source. Together with the use of ZF-TEDOR mixing, this sample allowed us to observe intermolecular (15)N-(13)C backbone-to-backbone contacts with excellent resolution and good sensitivity. The results are consistent with a parallel, in-register arrangement of the protein subunits in the fibrils and suggest that a significant structural reorganization occurs from the native to the fibril state.

β2-Microglobulin (β2m) is a 99-residue protein that forms amyloid fibril deposits associated with dialysis-related amyloidosis (DRA).(1) Under acidic conditions (pH = 2.5) and low salt concentration, the protein can also form amyloid fibrils in vitro through a nucleation-dependent mechanism.[2,3] These fibrils are long, straight, and unbranched in appearance (Figure S1) and share many properties with the fibrils isolated from tissues of DRA patients, including the same characteristic amide I′ band in FTIR spectra.(4) It has been shown that the fibrils themselves, and not the prefibrillar oligomeric species formed in the lag phase of assembly, can disrupt model membranes and are toxic to cells.(5) While an atomic structural model for these fibrils is not yet available, structural details emerged first through methods like limited proteolysis,[6,7] hydrogen exchange,[8,9] and more recently by magic angle spinning (MAS) NMR,(10) electron paramagnetic resonance (EPR),(11) and cryo-electron microscopy (cryoEM).(12) In particular, analysis of the chemical shifts of 64 assigned residues of β2m fibrils has shown that the protein contains a rigid fibril core with substantially more β-sheet character than the native protein.(10) CryoEM maps revealed a complex picture of the fibrils, where non-native globular β2m monomers pack in “dimer-of-dimers” building blocks that associate asymmetrically into crescent-shaped units.(12) In addition, site-directed EPR spin labeling suggested that the major building block consists of six β2m polypeptide chains, arranged in a parallel, in-register manner.(11)

In the experiments described here, we investigate the tertiary structure of β2m amyloid fibrils with 15N−13C MAS NMR correlation spectroscopy. MAS NMR has been successfully used to obtain information about the inter- and intramolecular interactions that form the β-sheet core of amyloid fibrils, including Aβ(1−40),(13) a 22-residue fragment of β2m,(14) Het-s(218−289),(15) and curli amyloid.(16) Various sample preparation techniques and experiments have been employed to achieve that end, including methods that rely on the incorporation of single labels[17,18] or proton-mediated transfer.(19) Here, we use ZF-TEDOR (z-filtered transferred echo double resonance) mixing[20,21] to obtain intermolecular 15N−13C correlations that establish that the protein subunits in long, straight β2m fibrils formed at pH 2.5 are arranged as parallel, in-register β-sheets.

Our experiments utilize fibrils formed from a 50:50 mixture of 15N,12C- and 14N,13C- labeled β2m monomers, the latter half being prepared using [2-13C]-glycerol as the carbon source. This sample, referred to as “mixed 2-β2m”, offers improved resolution in the 13C dimension[22−24] (Figure 1a) as well as potential gains in experimental transfer efficiency due to the significantly reduced number of directly bonded 13C atoms.(25) The absence of 13C J-couplings and the elimination of strong (intramolecular) dipolar 15N−13C couplings as a result of the mixed nature of the sample improve the efficiency of ZF-TEDOR.[20,26,27]

Figure 1

(a) 13C CP spectrum of mixed 2-β2m fibrils, 512 scans; (b) ZF-TEDOR spectrum obtained with τmix = 1.76 ms, 512 scans;, (c) ZF-TEDOR with τmix = 18 ms, 5120 scans.

In a 100% uniformly 15N,13C labeled β2m sample, the experimental one-bond 15N−13C transfer efficiency after 1.76 ms of ZF-TEDOR mixing is typically ∼20% of the 13C CP signal. In the mixed 2-β2m sample, after such a short mixing time, no significant buildup of 13C polarization is observed, as shown in Figure 1b. This is due to the absence of 13C nuclei in the 15N,12C-labeled monomers, which were prepared using 13C-depleted glucose (99.9% purity) to eliminate contributions from natural abundance. In particular, signals from one-bond 15N−13C interactions are not detected. On the other hand, longer ZF-TEDOR mixing times lead to the buildup of 13C intensity, which reaches a maximum at 18 ms (Figure 1c and Figure S2) and is consistent with 15N−13C distances of ∼5.0−5.5 Å. The maximum bulk transfer efficiency for the Cα region is ∼3%, which is better than the experimental transfer efficiencies observed for uniformly 13C labeled samples (<1% for similar distances).(20)

In order to obtain site-specific information regarding the origin of the 15N−13C intermolecular contacts in mixed 2-β2m, we recorded a 2D ZF-TEDOR experiment with τmix = 16 ms (Figure S3). This spectrum presents excellent resolution (13C line widths ∼50 Hz) and sufficient sensitivity after a long acquisition period, which was facilitated by the robustness of the TEDOR sequence. Overall, the positions of the observed cross-peaks in this mixed 2-β2m spectrum correspond exactly with the positions of cross-peaks in one-bond (Figure 2a) or two-bond TEDOR spectra (data not shown) of a β2m fibril sample prepared from 100% 15N, 2-13C glycerol labeled material (2-β2m). The majority of the cross-peaks in the mixed 2-β2m sample could be readily assigned based on known chemical shifts of long, straight β2m fibrils,(10) and they correspond exclusively to intermolecular Ni−Cαi, Ni−Cαi−1, Ni−COi, or Ni−COi−1 transfer (Figures 2b and S3). In particular, the following residues giving rise to intermolecular contacts in the mixed 2-β2m sample were assigned: H31−S33, N42, G43, R45, I46, V49, H51−F62, P72, T73, and Y78−V82. While P32, S33, G43, F56, S57, K58, and F62 are part of well-ordered loops in the fibrils, the remainder of the residues represent all of the currently assigned fibril β-strands.

Figure 2

Comparison of the 15N−13Cα region of correlation spectra obtained with ZF-TEDOR mixing for two differently labeled β2m fibril samples. (a) 2-β2m, τmix = 1.6 ms, 12 mg of sample, 2 days of experimental time. Labels correspond to intramolecular Ni−Cαi transfer, unless otherwise noted, while labels in gray denote cross-peaks that appear only in the 2-β2m spectrum. (b) Mixed 2-β2m, τmix = 16 ms, 16 mg of sample, 9 days of experimental time. Labels correspond to intermolecular Ni−Cαi or Ni−Cαi−1 transfer, unless otherwise noted.

Some cross-peaks in Figure 2b (mixed 2-β2m) do not presently have assignments. Conversely, not all of the strong cross-peaks shown in Figure 2a (2-β2m) appear in the mixed 2-β2m spectrum. This includes G18, G29, E44, H84, and V85 (shown in gray in Figure 2a) among others. This is most likely due to differences in local dynamics and relaxation whose effects are exacerbated at long mixing times, resulting in large variations in the cross-peak intensities.(30)

The data presented above suggest that long, straight β2m fibrils grown at pH 2.5 and low salt concentration form parallel, in-register β-sheets. In such a case the average distances for intermolecular Ni−Cαi and Ni−Cαi−1 contacts are ∼5 and ∼5.5 Å respectively (Figure S4), which is consistent with the bulk ZF-TEDOR buildup (Figure S2). In order to accommodate such an arrangement, substantial reorganization of the native antiparallel β-sheet structure[31−33] is required, indicating that the structure of the monomers within the fibrils must be highly non-native. Figure 3 highlights two clear pieces of evidence for the non-native structure of β2m within fibrils: first, residues involved in loops/turns in native β2m (Figure 3a) reorganize to form ordered β-strands in the fibrils (Figures 3b and S5), and second, while all β-strands form antiparallel β-sheet contacts with residues distant in sequence in native β2m, β-strands in the fibrils are parallel and in register.

Figure 3

(a) Crystal structure of native monomeric β2m (PDB ID: 1DUZ)(28) showing the antiparallel β-sheet arrangement of the strands (labeled A to G). (b) Residues that form β-strands in fibrillar β2m painted onto the native fold. β-Strands in the fibrils(10) are shown as thick tubes, and the residues giving rise to assigned intermolecular Ni−Cαi cross-peaks are shown in black. The structures were prepared using the Chimera software.(29)

The parallel arrangement of the β-strands in β2m fibrils was predicted initially by FTIR experiments[34,35] and is in agreement with data obtained by site-directed spin labeling and EPR.(11) The results described here verify and expand upon the latter, which indicates that spin labels attached to cysteine-substituted residues S33, S55, S61, and T73 among others give EPR spectra indicative of immobile, parallel, and in-register stacked spin labels (Figure S5). Stacks of six β2m monomers arranged in that manner are then required to fulfill the electron density maps obtained by cryoEM.(12) The site-specific information regarding the intermolecular arrangement of β2m fibrils presented here provides an important step toward a full molecular model of the fibrils. Additional experiments, particularly aimed at determining the quaternary fold of the fibrils, are in progress and should shed light on how this tertiary fibril arrangement fits into such a complex cryoEM electron density profile.