Structure of an early native-like intermediate of β2-microglobulin amyloidogenesis

Abstract

To investigate early intermediates of β2-microglobulin (β2m) amyloidogenesis, we solved the structure of β2m containing the amyloidogenic Pro32Gly mutation by X-ray crystallography. One nanobody (Nb24) that efficiently blocks fibril elongation was used as a chaperone to co-crystallize the Pro32Gly β2m monomer under physiological conditions. The complex of P32G β2m with Nb24 reveals a trans peptide bond at position 32 of this amyloidogenic variant, whereas Pro32 adopts the cis conformation in the wild-type monomer, indicating that the cis to trans isomerization at Pro32 plays a critical role in the early onset of β2m amyloid formation.

Introduction

For many proteins, abnormal structure or metabolism results in self-aggregation, causing the formation of fibrillar structures that give rise to amyloid fibrils.^1,2 The full elucidation of the mechanism of this aggregation process requires the identification of all the transitional conformational states and oligomeric structures adopted by the polypeptide chain. The identification and characterization of oligomers preceding the formation of fibrils is of particular interest because of an increasing awareness that these species are likely to play a critical role in the pathogenesis of protein deposition diseases.^3,4 However, the characterization of early aggregation-prone monomeric species also remains a challenge.⁵

Over the years, human β2-microglobulin (β2m) has been used extensively as a model system to elucidate the molecular mechanism of amyloidosis. β2m is a small protein (99 amino acids) that adopts the typical seven-stranded β-sandwich immunoglobulin fold formed by two antiparallel β-sheets (strands ABED and strands GFC) that are connected via a single disulphide bond (Cys25–Cys80).^6,7 β2m is part of the major histocompatibility complex I (MHC I), a protein complex that is involved in the presentation of antigenic peptides.^6,8 In healthy individuals, the catabolism of β2m involves the shedding of the complex from the cell membrane, the dissociation of β2m from the complex and the excretion of the freed β2m by the kidney. However, patients suffering from renal failure cannot clear the free β2m from the serum causing the β2m concentration to rise 5–60 times above the normal level of 0.1 µM. In persistent dialysis patients, excess β2m undergoes self-association to form amyloid fibrils that deposit in the musculo-skeletal system of the patient, causing dialysis related amyloidosis (DRA).⁹ Although an elevated concentration is not sufficient for aggregation, it is probably a decisive factor in causing the disease.^1,2,7

It is well known that the cis–trans isomerization rates of peptidyl–prolyl amide bond are slow, even in random coil polypeptide chains¹⁰ and that cis–trans isomerization is the rate limiting step in many structural transitions.^11–13 By studying the folding kinetics at pH 7 and 37°C, Radford and coworkers demonstrated that β2m folds by two parallel routes involving two native-like intermediates. One folding intermediate contains the peptidyl-Pro32 amide bond in the cis conformation while the other has this bond in a trans conformation, with a rate limiting isomerization from trans to cis causing a minor fraction to fold slower. The folding intermediate containing this non-native trans-proline isomer was identified as a direct precursor of dimeric species and oligomers that accumulate before the development of amyloid fibrils.^14,15 Miranker's team independently confirmed the importance of this backbone isomerization in β2m fibrillogenesis, using divalent copper as oligomerization trigger. Addition of Cu²⁺ initiates the isomerization of a conserved cis proline at position 32, thereby facilitating the formation of amyloid fibrils.^16,17 Consistent with these models, Pro32 also adopts the trans conformation in the X-ray structure of a domain-swapped dimer of ΔN6 β2m—another amyloidogenic β2-microglobulin variant—confirming that the cis to trans isomerization at Pro32 plays a critical role in the onset of β2m amyloid formation.⁵

The impact of several Pro32 mutations on the amyloidogenic properties of β2m has been analyzed to link peptidyl–prolyl cis–trans isomerization to fibril formation. In the presence of Cu²⁺, Pro32Ala β2m oligomers are formed within one minute, while the oligomerization of wild type β2m shows a kinetic profile of 1 h.¹⁶ Similarly, the Pro32Gly mutation causes a dramatic enhancement in the rate of amyloid fibril elongation.¹⁴ Remarkably, this mutant folds considerably faster than the wild-type β2m, involving only one native-like intermediate. It was suggested that Gly32 adopts a trans conformation in the folded protein as in the folding intermediate that was identified as a direct precursor of amyloidogenesis, explaining faster fibril elongation of the mutant because this trans-peptide species generates the amyloidogenic properties of β2m.¹⁴

Aiming to further characterize the aggregation-prone nature of the folding intermediate that serves as the precursor for amyloidosis, we used nanobody-assisted X-ray crystallography to solve the structure of Pro32Gly β2m (P32G β2m). One nanobody (Nb24) that was previously identified to block the aggregation of the ΔN6 amyloidogenic variant, was used in this study to co-crystallize the amyloidogenic P32G β2m variant as a monomer.

Results

The P32G mutation promotes fibril elongation of β2m under physiological conditions

The ability of recombinant wild-type β2m and two amyloidogenic β2-microglobulin variants, P32G β2m and ΔN6 β2m, to elongate β2m seeds under physiological conditions was monitored by measuring the fluorescence increase of the amyloid-specific dye Thioflavin T (ThT) or by visualizing the amyloids using negative stain electron microscopy (EM). Amyloid fibrils from P32G β2m and ΔN6 β2m could be detected after 14 days while wild-type β2m fibrils could only be observed after 30 days (Fig. 1). Although wild-type β2m was able to elongate β2m seeds into amyloid fibrils under physiological conditions, the fibrillogenesis process was completed significantly faster with the P32G β2m and ΔN6 β2m variants, confirming their amyloidogenic properties.^14,15,18

Figure 1.

Seeded fibrillogenesis of β2m, P32G β2m, and ΔN6 β2m. (A) ThT fluorescence increase was used to monitor the progress of fibril elongation of β2m (light gray diamonds), P32G β2m (dark gray circles), and ΔN6 β2m (black triangles). (B) EM images confirmed the presence of P32G β2m fibrils and ΔN6 β2m fibrils after 14 days incubation whereas β2m fibrils only appeared after 30 days incubation. The scale bar represents 200 nm.

Nanobodies efficiently block P32G β2m fibrillogenesis under physiological conditions

Nanobodies are single domain antibodies harboring the full antigen-binding capacity of naturally occurring heavy chain antibodies,^19,20 and have been used successfully as crystallization chaperones.^5,21,22 Several nanobodies with nM to µM range dissociation constants for β2m were tested as fibrillogenesis inhibitors by incubating wild type β2m, P32G β2m, and ΔN6 β2m with β2m fibril seeds in the presence or absence of each nanobody. A nanobody raised against an irrelevant antigen (Nb108) was included in these seeding experiments as a negative control.

Using ThT fluorescence increase to follow fibrillogenesis, several nanobodies (Nb22, Nb23, Nb24, Nb29, and Nb30) were identified to inhibit fibril elongation of β2m, P32G β2m, and ΔN6 β2m (data not shown). According to the relative ThT fluorescence increase, virtually no fibrils or other aggregates are formed from the P32G β2m•Nb24 complex indicating that this nanobody (50 µM) fully protects the amyloidogenic variant against aggregation (Fig. 2). These results were confirmed by negative stain EM. The fact that we were able to purify the P32G β2m•Nb24 complex as a stable heterodimer by size exclusion chromatography prompted us to test Nb24 as a cocrystallization chaperone for monomeric P32G β2m, aiming at characterizing its aggregation-prone properties.

Figure 2.

The effect of an inhibitory nanobody (Nb24) and an unrelated nanobody (Nb108) on the seeded fibrillogenesis of β2m (after 30 days incubation), P32G β2m (after 14 days incubation), and ΔN6 β2m (after 14 days incubation), monitored by ThT fluorescence and EM imaging. The scale bar represents 200 nm.

Nanobody-assisted crystallization of the P32G amyloidogenic variant of β2m

Atomic-level structural investigation of the key conformational intermediates of amyloidosis remains a challenge due to the dynamic equilibrium between diverse structural species. Fibril formation of β2m in vivo usually takes several years and intermediate species are short lived and highly unstable. The use of specific antibodies offers promising strategies for probing the process of fibril formation by biophysical methods including X-ray crystallography.^5,23,24

Aiming to characterize the aggregation-prone nature of the folding intermediate that serves as the precursor for amyloidosis, we used nanobody-assisted X-ray crystallography to solve the structure of Pro32Gly β2m (P32G β2m). P32G β2m was mixed in a 1:1 molar ratio with Nb24 in 20 mM Tris, 150 mM NaCl at pH 7.5. After separating the complex by size exclusion chromatography, the purified complex was easily concentrated (8 mg/mL) as a soluble entity and subjected to different crystallization screens. Using the hanging drop vapor diffusion method, diffracting crystals were formed in 0.1M MES (pH 6.5) using 1.6M MgSO₄ as the precipitant.

X-ray diffraction data of the P32G β2m•Nb24 complex were processed until 2.6 Å resulting in a CC1/2 of 62.3% and an I/sigI of 1 was reached at 3.2 Å (Table I). The P32G β2m•Nb24 complex was crystallized in space group P4₂2₁2 and the asymmetric unit contains two molecules P32G β2m each bound to one Nb24 (RMSD between the two P32G β2m•Nb24 complexes = 0.42 Å calculated over 884 main chain atoms) [Fig. 3(A)]. Nb24 binds the last residues of β-strand C and the loops linking β-strands C to D and β-strands E to F [Fig. 3(C)], very similar to the interaction observed in the ΔN6 β2m•Nb24 complex.⁵ When compared to β2m in complex in the MHC I (PDB entry 1DUZ),²⁵ the binding of Nb24 has little effect on the main-chain conformation of its binding epitope. The loops that are part of the Nb24 epitope (C to D and E to F) have an RMSD of 0.33 Å calculated over 60 main chain atoms [Fig. 3(C)]. The higher overall RMSD of 1.7 Å calculated over the 380 main chain atoms is mainly due to structural differences concentrated in two regions. The first area is located at the loop connecting β-strands B to C and the second region comprises the β-strand D up to the loop linking β-strands D to E. These two segments of the protein do not interact with the nanobody. When we exclude these two regions, an RMSD value of 0.61 Å (calculated over 300 main chain atoms) could be calculated and the overall structure adopts native-like conformation.

Table I.

Data-Collection, Refinement and Validation Statistics of the Structure of the P32G β2m•Nb24 Complex

Data Set	P32G β2m•Nb24
Data-collection
X-ray source	Diamond IO3
X-ray wavelength (Å)	0.97950
Temperature (K)	100
Space group	P 42212
Unit-cell parameters
a, b, c (Å)	96.8, 96.8, 167.8
α, β, γ(°)	90.0, 90.0, 90.0
Resolution range (Å)	48.43–2.6 (2.67–2.60)
Total/Unique reflections	362392/25281
Rmerge (%)a	35.7 (392)
Rmeas (%)b	37.0 (406)
Data completeness (%)	99.9 (99.8)
Average I/σ	8.5 (1.0)
Redundancy	14.3 (14.7)
Wilson B factor (Å2)	56.1
CC(1/2)	99.7 (62.3)
Refinement
Correlation coefficients
Correlation coefficient Fo − Fc	0.948
Correlation coefficient Fo − Fc Free	0.914
Rwork/Rfreec	22.84/27.69
Total number
Amino acid residues	445
Water molecules	50
Ligand atoms	8
rmsd
Bond length (Å)	0.0150
Bond angles (°)	1.5567
Average atomic B-factor (Å2)
Protein atoms	58.67
Solvent atoms	35.0
Ramachandran plot (%)
Favored regions	97.70
Allowed regions	2.30
Disallowed regions	0
PDB entry	4KDT

^a

^b

^c, F(h)_o and F(h)_c are observed and calculated structure factor amplitudes, respectively. A random subset of data (5%) was used for the R_free calculation.

Figure 3.

Crystal structure of P32G β2m•Nb24 complex. (A) The P32G β2m•Nb24 complex as a stable heterodimer with P32G β2m in red and Nb24 in gray. (B) The 2F_o – F_c omit density map is shown for residues Gly29, Phe30, His31, Ser33, and Phe62. The map is contoured at 1 σ within 1.6 Å of the residues. (C) Comparison between the P32G β2m•Nb24 structure (red) and the β2m structure in the MHC I complex (PDB entry 1DUZ²⁵;green) focusing on the high similarity of the main-chain conformation in the binding region of P32G β2m to Nb24 (in gray surface representation). (D) Comparison between P32G β2m•Nb24 structure (red) and the β2m structure (PDB entry 1DUZ²⁵; green) focusing on the key structural consequences of the cis–trans isomerization at position 32.

Whereas Pro32 consistently adopts the cis conformation in wild-type β2m (PBD entries: 1DUZ, 1JNJ, 1LDS)^25–27 Gly32 adopts the trans conformation in the P32G β2m monomer (current structure) [Fig. 3(B)]. This isomerization causes Phe30 to move out of the hydrophobic core to adopt a solvent exposed configuration. The freed space is filled by Phe62, causing compensating structural rearrangements of β-strand D up to the loop linking β-strands D to E (Glu50 to Phe62) [Fig. 3(D)].

Discussion

Amyloid fibril formation generally occurs via a nucleation-dependent oligomerization process characterized by a lag phase.^4,28,29 During this rate-limited phase, amyloidogenic intermediates are formed while the conversion into amyloid fibrils only occurs in the following elongation phase. The formation of these amyloidogenic intermediates involves the disruption of the native structure to a greater or lesser extent, in order to allow self-association and the formation of a cross-β-sheet structure that is the hallmark of amyloid fibrils. The lag phase can be shortened or ultimately abolished in vitro by adding fibrillar seeds, by changing the experimental conditions or by using fibrillogenic mutants.^4,24,28–32

Local fluctuations of one or more regions of the β2m monomer causing the generation of precursors that are prone to spontaneous self-assembly are also at the origin of DRA.⁷ It has been shown that a β2m conformer with a non-native peptidyl-Pro32 trans peptide bond serves as a direct precursor of dimeric species and oligomers that accumulate before the development of amyloid fibrils, linking cis–trans isomerization of the Pro32 imidic peptide bond to β2m fibrillogenesis.^14,15 From our structure, it appears that the trans conformer at position 32 is the most stable folded conformation of P32G β2m, explaining the intrinsic amyloidogenic nature of this mutant. Consistent with this notion, the trans conformation at position 32 has already been observed in other amyloidogenic variants of β2m including the domain-swapped dimer of ΔN6 β2m⁵ and Pro32Ala β2m.¹⁶ The trans peptide bond at position 32 as observed in Pro32Gly β2m causes several structural rearrangements that may increase to susceptibility of β2m to aggregate and form cross-β-sheet structures.

Regular β-sheets are inherently aggregation-prone because the motive for β H-bonding with any other β-strand is available.³³ Natural β-sheet proteins are designed to avoid this interaction and make use of different blocking features to prevent edge-to-edge β aggregation by the formation of intersubunit β H-bonds.³³ For β-sandwich proteins, a very common strategy to avoid aggregation is the presence of a charged side chain of lysine, arginine, glutamic acid, aspartic acid, or histidine. In monomeric β-sandwich proteins, such charged residues are capable to simultaneously form hydrophobic sheet-packing interactions while exposing their charged side chains to the solvent, preventing edge-to-edge association through electrostatic repulsion between these exposed charges.^34–36

β2m (a typical β-sandwich protein) contains four edge strands: strands A and G form one pair at one side of the protein and strands C and D form a second pair at the opposite side (Fig. 4). Focusing on these vulnerable strands, the wild-type protein contains several gatekeepers including His51 to protect strand D and Lys91 to protect strand G, thus avoiding edge-to-edge aggregation.²⁷ In our structure of P32G β2m, Lys91 is also exposed and remains the gatekeeper for the first edge pair formed by strands A and G (Fig. 4). The trans peptide bond caused by the Pro32Gly mutation is linked to significant structural changes in edge strand D of the ABED antiparallel β-sheet and in the D to E loop. However, His51 points towards the hydrophobic core of P32G β2m, as it does in wild type β2m, where it still acts as a gatekeeper to avoid edge-to-edge aggregation of strands C and D (Fig. 4).

Figure 4.

Structural consequences of the cis–trans isomerization at position 32 on three gatekeepers for P32G β2m (red) in comparison with β2m (PDB entry 1DUZ²⁵ green). The gatekeeper for β-strands A and G is Lys91 (magenta for P32G β2m and light pink for β2m) and remains solvent exposed. For the second edge pair (β-strands C and D) there are two gatekeepers, but only His51 (cyan for P32G β2m and light cyan for β2m) remains in its original orientation while Asp53 (blue for P32G β2m and light blue for β2m) rotates and the β bulge gets lost for P32G β2m, forming a more aggregation-competent surface.

The presence of β bulges to kink β-strands is a second approach to avoid edge-to-edge aggregation by distorting the geometry and accentuate the twist of the associating strands.³⁵ Remarkably, Asp53—a key residue that is usually forming a β bulge by pointing outwards and distorting the geometry of the D-strand^26,27—is rotated inward in P32G β2m and the β bulge is lost, resulting in a more continuous D-strand that is probably more prone to intermolecular pairing (Fig. 4).

The trans peptide bond at position 32 also causes Phe30 to become solvent exposed and Phe62 to repack within the hydrophobic core to fill the freed space. This structural rearrangement of these large hydrophobic residues causes a significant increase in the surface hydrophobicity of the protein, apparently reducing its intrinsic solubility. Concomitantly, the inward rotation of the Asp53 side chain also contributes to the increase of the surface hydrophobicity (Fig. 5). Increased surface hydrophobicity has been linked to aggregation and fibril formation.^37,38

Figure 5.

Consequences of the cis–trans isomerization at position 32 on the surface electrostatic potential. (A) P32G β2m colored by electrostatic surface potential on the solvent accessible surface. (B) β2m (PDB entry 1DUZ²⁵) colored by electrostatic surface potential on the solvent accessible surface. Visualization is done by setting the values at which the surface colors are clamped at red (−5 kT/e) for negatively charged regions and blue (+5 kT/e) for positively charged regions.

In conclusion, we were able to trap the P32G amyloidogenic variant of β2m as a monomer and solve its structure by nanobody-assisted X-ray crystallography. The structure of monomeric P32G β2m in complex with Nb24 reveals a trans peptide bond at position 32, whereas Pro32 adopts the cis conformation in the wild-type monomer. Consistent with earlier work on the native protein¹⁴ and on the amyloidogenic ΔN6 ^5,39 and Pro32Ala¹⁶ variants, this study identifies the cis-to-trans isomerization at position 32 in the β2m monomer as an early event in amyloidogenesis that makes the protein surface more hydrophobic and probably renders it more susceptible to edge-to-edge β aggregation by the formation of intersubunit β H-bonds and confirms the key role of Pro32 to safeguard the aggregation prone β-sheets of β2m against edge-to-edge β association.

Material and Methods

Expression and purification of the proteins

β2m, ΔN6 β2m, and P32G β2m were expressed in Escherichia coli and purified out of inclusion bodies as described.^40,41 All nanobodies were expressed in E. coli and purified from the periplasm as described.⁵

Fibrillogenesis under physiological conditions

Fibrils of β2m, P32G β2m, or ΔN6 β2m were grown at 37°C in the presence of β2m seeds with agitation in microtiter plates essentially as described by Radford and coworkers.¹⁸ Lyophilized protein was dissolved in 25 mM sodium phosphate, 25 mM sodium acetate buffer containing 0.5% sodium azide at pH 7.0 at a concentration of 0.5 mg/mL to which 10% of heparin-stabilized β2m seeds were added. Samples of 200 µL were agitated at 250 rmp at 37°C. Fibril formation was followed in time by measuring the ThT fluorescence increase (excitation 440 nm, emission 480 nm). In a typical experiment, the average of 10 replicates was normalized to the signal from buffer only containing β2m seeds. The presence of long straight fibrils was confirmed by negative-stain electron microscopy. For microscopy, protein samples (5 µL) were applied dropwise on carbon-coated grids (Formvar/carbon on 400 Mesh Copper). The grids were washed with water, stained with 1% uranyl acetate and the samples were analyzed on a Jeol JEM-1400 electron microscope at 100 kV.

The same experimental conditions (37°C, 250 rpm) were used to measure the inhibitory effect of nanobodies on the seeded fibrillogenesis of β2m, P32G β2m, and ΔN6 β2m. Lyophilized nanobodies (0.8 mg/mL) were dissolved in buffer (25 mM sodium phosphate, 25 mM sodium acetate buffer containing 0.5% sodium azide at pH 7.0) containing 10% β2m heparin-stabilized seeds before β2m, P32G β2m, or ΔN6 β2m was added to the reaction mixture (0.5 mg/mL). ThT fluorescence measurements were done as described above. The inhibitory effect of a nanobody (in %) was estimated from the ratio of the ThT fluorescence of samples containing a particular nanobody and samples without nanobody, incubated under the same conditions in the same microtiter plate.

Crystallization, data-collection, and structure determination of the P32G β2m•Nb24 complex

P32G β2m was mixed with equimolar amounts of Nb24 in 20 mM Tris, 150 mM NaCl, pH 7.5. After 2 h of incubation at 4°C, a size exclusion chromatography was performed in the same buffer to purify the complex. Fractions containing the complex were collected and concentrated to 8 mg/mL by ultrafiltration. Crystals were grown at 20°C using the hanging drop vapor diffusion method by mixing equal volumes of the complex with a reservoir solution containing 0.1M MES (pH 6.5) and 1.6M MgSO₄. Before data collection, crystals were flash frozen in liquid nitrogen using 15% glycerol as the cryo-protectant. Diffraction data were collected at the IO3 beamline (Diamond, UK) at 100 K. Data indexing, integration and scaling were done using the XDS suite.⁴² Data were processed to 2.6 Å, resulting in a CC_1/2 of 62.3% as cut-off and an I/sigma I of 1 was reached at 3.2 Å resolution.⁴³ The crystal structure of the P32G β2m•Nb24 complex was solved by molecular replacement with PhaserMR⁴⁴ using the separate coordinates of the ΔN6 β2m monomer and Nb24 (both from PDB entry 2X89)⁵ as the models. Initial model building was done automatically using ARP/wARP^45,46 and the model was refined manually using COOT.⁴⁷ Structure refinement was carried out using Refmac5.⁴⁸ TLS refinement was implemented in the refinement protocol using four individual TLS groups determined by TLSMD.^49,50 The MolProbity server was used to stereochemically validate the model.⁵¹ Data collection and refinement statistics are summarized in Table I. Electrostatic surface calculations were prepared with PDB2PQR⁵² using the PARSE force field and APBS.⁵³ Figures were prepared using Yasara⁵⁴ or Pymol.⁵⁵