Post‐Translational Isoaspartate Promotes Amyloid Formation in β2‐Microglobulin Authors: Ryuji Kawakami, Toshiki Takei, Masatomo So, Toshifumi Takao, Yoshinori Taguchi, Hironobu Hojo Published: 2026 ABSTRACT ABSTRACTβ2‐Microglobulin (β2m) aggregation is central to dialysis‐related amyloidosis (DRA), yet the molecular triggers underlying its fibrillogenesis remain incompletely defined. Among proposed mechanisms, isoaspartate (isoAsp) formation at Asn residues has been implicated but never directly tested due to synthetic inaccessibility. Here, we report the total chemical synthesis of β2m variants containing isoAsp at known in vivo hotspots, Asn17 and Asn42, enabling precise structural and functional interrogation. Remarkably, the isoAsp17‐modified β2m exhibited greater fibril formation capacity than the pathogenic ΔN6‐β2m variant, suggesting a previously underappreciated contribution of isoAsp17. Structural characterization was complemented by spectroscopic and TEM analyses, which demonstrated that isoAsp17 promotes fibril formation. Our findings provide the first direct experimental evidence that a spontaneous post‐translational modification can initiate amyloidogenesis in β2m, redefining the molecular basis of DRA and highlighting isoAsp as a general driver of pathological protein aggregation. β2‐Microglobulin (β2m) is a component of major histocompatibility complex class I (MHC‐I) molecules, which are expressed on the surface of all nucleated cells and function in the presentation of endogenous antigens to cytotoxic T cells [1]. During MHC‐I turnover, β2m is released into circulation and normally cleared by the kidneys. In healthy individuals, its serum concentration remains around 1–2 mg/L. However, in patients with impaired renal function undergoing long‐term dialysis, β2m cannot be efficiently removed, even with advanced dialysis membranes, and accumulates to levels exceeding 40 mg/L [2]. Prolonged β2m accumulation leads to the formation of amyloid fibrils, particularly in joint tissues, causing dialysis‐related amyloidosis (DRA)—a debilitating condition that significantly impairs patient mobility. Although β2m's involvement in DRA has been firmly established, the precise molecular mechanisms underlying its amyloidogenic transformation remain incompletely understood. One known pathological variant is ΔN6, a β2m isoform lacking the six N‐terminal residues, which has been consistently identified in amyloid deposits from DRA patients [3, 4, 5, 6]. ΔN6 exhibits reduced structural stability and enhanced fibril formation compared to the native protein, providing key insights into how structural alterations can promote β2m misfolding. In addition to such proteolytic variants, previous studies have highlighted the role of spontaneous post‐translational modifications, particularly non‐enzymatic deamidation of asparagine (Asn) residues [7, 8]. β2m contains two Asn‐Gly sequences—Asn17‐Gly18 and Asn42‐Gly43—which are known hotspots for deamidation [9]. This non‐enzymatic process is a ubiquitous molecular aging reaction that affects long‐lived proteins and has been implicated in various pathological conditions, including neurodegenerative diseases and systemic amyloidoses. Deamidation converts Asn into either aspartate (Asp) or isoaspartate (isoAsp), with the latter introducing a β‐peptide linkage that shifts the protein backbone and potentially destabilizes the structure (Figure 1b). Our previous analysis showed that Asn17 undergoes deamidation more readily than Asn42, with a nearly equal distribution of Asp and isoAsp products (Asn:Asp:isoAsp = 28:40:32), whereas Asn42 predominantly forms isoAsp (Asn:Asp:isoAsp = 89:0.4:10.6) [10]. (a) Structural basis for the differential impact of Asn17 and Asn42 isomerization on β2m stability. Asn17 forms a hydrogen bond network with Arg97, connecting the N‐terminal AB loop and the C‐terminal FG loop. In contrast, Asn42 interacts locally with Asp76 and Glu77 within the same β‐sheet (PDB: 2axf, resolution: 1.80 Å). (b) The mechanism of deamidation at the ‐Asn‐Gly‐ sequence, (c) amino acid sequence of β2m. The arrows indicate the site of ligation. (d) Synthesized β2m variants. Although recombinant expression systems have facilitated analysis of Asp substitutions [11], the incorporation of isoAsp cannot be achieved by these methods. In contrast, recent advances in the chemical synthesis of proteins have enabled the preparation of native and post‐translationally modified proteins for structural and functional studies [12, 13]. Building on these advances, in this study, we aimed to elucidate the impact of site‐specific deamidation by chemically synthesizing β2m variants containing Asp or isoAsp at Asn17 and/or Asn42 (Figure 1d). Through comparative biophysical analyses with the ΔN6 variant, we demonstrate that the isoAsp17 modification markedly destabilizes β2m and accelerates fibril formation under physiological conditions. These findings provide direct evidence that isoAsp formation at Asn17 serves as a potent molecular trigger of β2m amyloidogenesis and may represent a novel therapeutic target in DRA. β2m is a 99‐residue protein that contains a single disulfide bond. The amino acid sequence is shown in Figure 1c. In this study, we planned to synthesize the full‐length sequence using the native chemical ligation (NCL) method [14] with two cysteine residues forming the disulfide bond serving as the condensation sites. The entire sequence was divided into three segments: N‐terminal (1–24), middle (25–79), and C‐terminal (80–99). To realize the synthesis of the β2m variants, 1a–1e, we prepared the N‐terminal with Asp17 2a, isoAsp17 2b, and Asn17 2c, middle with isoAsp42 3a and Asn42 3b, and C‐terminal segment 4 (Figure S1) by the solid‐phase method [15]. The NCL reaction was then performed. When synthesizing peptides containing an Asp‐Gly sequence using the prevalent 9‐fluorenylmethoxycarbonyl (Fmoc) strategy, it is known that the aspartimide formation during piperidine treatment usually occurs as a major side reaction. To suppress this, a dipeptide unit with the 2,4‐dimethoxybenzylated amide between Asp and Gly is frequently used [16]. On the other hand, in the t‐butoxycarbonyl (Boc) strategy, it is known that aspartimide formation can be suppressed by using a cyclohexyl (cHex) ester for Asp side chain protection [17]. In addition, the peptide thioester, essential for the ligation of the peptides by the NCL, can be directly synthesized by the Boc method. Therefore, the N‐terminal and middle segments, which retain the Asp‐Gly or isoAsp‐Gly sequence, were synthesized by the Boc strategy using the in situ neutralization method [18]. As an example, the synthesis of the N‐terminal segment with isoAsp17 2b is shown in Figure S5. To the p‐methylbenzhydrylamine (MBHA) resin, Boc‐Ala‐OH was introduced by the O‐(1H‐benzotriazol‐1‐yl)‐N,N,N′,N′‐tetramethyluronium hexafluorophosphate (HBTU)/N,N‐diisopropylethylamine (DIEA) method. This residue is essential to avoid the cleavage of the peptide chain from the resin during the Boc group removal by trifluoroacetic acid (TFA) treatment [19]. S‐Triphenylmethyl (Trt)‐protected 3‐mercaptopropionic acid (MPA) was introduced by the same procedure. After removal of the Trt group by TFA containing triethylsilane (TES), the thioester linkage was formed by introducing the C‐terminal amino acid by HBTU and DIEA. The remaining peptide chain was elongated by the Boc method. For the introduction of the 17th amino acid residue, Boc‐Asp‐OcHex, which has the free side chain carboxy group, was introduced to form the isoAsp bond. After the complete assembly of the peptide chain, the protected peptide resin was treated with HF containing anisole as a cation scavenger to obtain the crude peptide thioester, which underwent reversed‐phase high‐performance liquid chromatography (RPHPLC) to obtain the purified peptide 2b. The other N‐terminal and middle segments were synthesized in a similar manner by the Boc method (See Supporting Information). The C‐terminal segment 4 was prepared by the Fmoc method (Figure S9). Having all the segments in hand, the ligation of the segments by the NCL method was performed as exemplified in the case of variant 1e in Figure 2. The middle 3a and the C‐terminal segment 4 were dissolved in a phosphate buffer (pH 7.0) containing 4‐mercaptophenylacetic acid (MPAA), tris(2‐carboxyethyl)phosphine (TCEP) hydrochloride, and guanidinium hydrochloride (GuHCl), and the condensation reaction was carried out by stirring under an argon atmosphere at room temperature in the dark. The reaction was completed in approximately 10 h, and the desired product was obtained as the main product. Subsequently, aq. CH3ONH2·HCl was added, and the pH was adjusted to 4.0 using aq. HCl to open the N‐terminal thiazolidine ring. After 12 h, the reaction was completed to obtain peptide 5a, which was monitored by RPHPLC and electrospray ionization (ESI)‐mass spectrometry (MS). The product was purified by RPHPLC to obtain the desired product 5a in 47% yield based on peptide 3a. The ligation of the N‐terminal segment 2b and intermediate 5a was also performed in a similar manner, and the entire polypeptide of β2m was assembled. A mixed solution of piperidine and 2‐mercaptoethanol (βME) was then added and stirred to remove the formyl group on the Trp side chain. After 15 min, the product was purified by RPHPLC to obtain peptide 6e in 67% yield based on peptide 5a. The obtained peptide 6e was dissolved in a phosphate buffer (pH 7.1) containing GuHCl, and aq. hydrogen peroxide was added to form the SS bond. After 5 min, the reaction was completed, and the product was isolated by RPHPLC to obtain 1e in 59% yield based on peptide 6e. The other β2m variants, 1a to 1d, were also synthesized in a similar manner. Synthetic route of β2m variant 1e. *: Non‐peptide. To verify that the chemically synthesized β2m variants retained their native‐like structures, circular dichroism (CD) spectra were measured under physiological buffer conditions (20 mM phosphate, pH 7.0). All variants, including the wild‐type, exhibited a characteristic negative peak around 220 nm and a positive peak near 202 nm (Figure 3a), consistent with the typical immunoglobulin fold of β2m. Notably, the isoAsp17‐containing variants, 1b and 1e, showed a marked reduction in both peaks, suggesting an increase in random coil content and a disruption of the native secondary structure. In contrast, the isoAsp42 variants, 1c and 1d, exhibited only a slight attenuation of the spectral features compared to the wild‐type, indicating relatively minor structural alterations. These results imply that isomerization at Asn17 has a more pronounced destabilizing effect on β2m structure than at Asn42. Our structural interpretation is further supported by the crystal structure of β2m (PDB: 2axf, 1.80 Å resolution), which clearly visualizes a hydrogen bond network between the side chain of Asn17 and the guanidinium group of Arg97 (Figure 1a). This interaction connects the N‐terminal AB loop with the C‐terminal FG loop, suggesting that Asn17 plays a role in stabilizing the global β‐sandwich architecture through long‐range inter‐loop interactions. Isomerization of Asn17 to isoAsp is likely to disrupt this specific interaction, contributing to conformational destabilization. (a) Secondary structure of wild‐type and variant β2m confirmed by CD spectra. (b) Temperature dependence of CD intensities monitored at 205 nm (dots). The lines represent the fitting curves analyzed by the two‐state denaturation model. Data were normalized using fitting values. The physicochemical parameters obtained are shown in Table 1. In contrast, Asn42 is located in the CD loop and forms hydrogen bonds with residues Asp76 and Glu77, which are part of the loop connecting the E and F strands. While these interactions help stabilize the adjacent β‐sheet region, they are spatially more localized and may have a more limited effect on the global β‐sandwich conformation. Compared to the long‐range interaction between Asn17 and Arg97, the structural impact of isomerization at Asn42 is likely to be more limited, which aligns with the relatively minor changes observed in the CD spectra of isoAsp42‐containing variants. The greater susceptibility of Asn17 to deamidation under physiological conditions [10] further supports its structural lability and functional relevance, whereas isoAsp42 formation may be a less common and less disruptive event in vivo. The stability of the structure of the β2m variants was analyzed by the change in the CD spectrum upon heating. Wild‐type‐ and ΔN6‐β2m [3, 4, 5, 6] stability was also assessed for comparison. All the β2m variants showed a decreased stability compared to the WT (Figure 3b and Table 1). The ΔN6‐, isoAsp17‐ and isoAsp42‐β2m variants, 1b‐1d, showed a comparable thermal stability, while isoAsp17,42‐β2m 1e showed a further decreased stability, indicating that the Asn to isoAsp conversion significantly decreased the stability of β2m. TABLE 1Physicochemical parameters of heat denaturation of β2ms. T m (°C) a ΔH U (×105 J/mol) a ΔS U (J/mol K) a ΔG U (×105 J/mol) a WT63.6 ± 0.053.03 ± 0.17901 ± 491.67 ± 1.23ΔN652.8 ± 0.152.20 ± 0.07674 ± 221.26 ± 1.00D17 1a 55.8 ± 0.122.35 ± 0.24715 ± 721.28 ± 0.98isoD17 1b 53.3 ± 0.142.24 ± 0.10685 ± 311.28 ± 1.01isoD42 1c 51.5 ± 0.202.40 ± 0.08740 ± 251.40 ± 1.13D17, isoD42 1d 50.6 ± 0.272.45 ± 0.18757 ± 541.46 ± 1.19isoD17, isoD42 1e 47.8 ± 0.331.99 ± 0.03619 ± 101.16 ± 0.95 a Tm, ΔH U, ΔS U, and ΔG U represent the transition midpoint, and the enthalpy, entropy, and Gibbs free energy of unfolding, respectively. Errors represent standard deviations (SD) from 3 independent experiments.John Wiley & Sons, Ltd. Physicochemical parameters of heat denaturation of β2ms. Tm, ΔH U, ΔS U, and ΔG U represent the transition midpoint, and the enthalpy, entropy, and Gibbs free energy of unfolding, respectively. Errors represent standard deviations (SD) from 3 independent experiments. To assess whether mutations at Asn17 or Asn42 promote amyloidogenesis by destabilizing the native β2m structure, we evaluated the fibril‐forming abilities of synthetic β2m variants using thioflavin T (ThT) fluorescence assays. Samples were prepared in phosphate buffer (pH 6.2) containing 115 mM NaCl and 5 µM ThT, with a final β2m concentration of 25 µM. The mixtures were incubated at 37°C with agitation, and fluorescence intensity was monitored at 485 nm (excitation at 445 nm). Figure S22 shows the normalized ThT fluorescence profiles. Variants 1b and 1e, both containing isoAsp17, exhibited accelerated fibril formation. Their half‐times (T 1/2) to reach 50% of maximum fluorescence are shown in Figure 4a. These isoAsp17 variants aggregated more rapidly than the ΔN6 variant, which is itself known for rapid fibril formation. Transmission electron microscopy (TEM) images (Figure 4b) confirmed the presence of long, unbranched fibrils in the samples containing variants 1b, 1e, and ΔN6. In contrast, variants 1c and 1d, which carry isoAsp42, showed no significant increase in ThT fluorescence (Figure 4a), suggesting a lack of typical amyloid fibril formation. However, TEM analysis revealed that these variants exhibited abnormal behavior compared to the wild‐type, forming amorphous or irregular aggregates rather than remaining as soluble monomers. This observation suggests that while the isoAsp42 modification destabilizes the native fold to some extent, it simultaneously reduces the amyloidogenic potential of the protein, instead promoting the formation of ThT‐negative amorphous species due to the overall decrease in structural stability. Amyloid formation of β2m variants. (a) The T 1/2 values were calculated using ThT kinetics (Figure S22) for ΔN6‐, isoAsp17‐, and isoAsp17, isoAsp42‐β2m, which showed a significant increase in ThT. N/A represents no significant increase in ThT. Error bars represent the standard deviation (SD) of four (for [isoAsp17, isoAsp42]‐β2m) or five (for others) independent replicates. Statistical significance was determined by one‐way ANOVA followed by the Tukey–Kramer post‐hoc test. ***p < 0.001. (b) TEM images of the fibrils formed by the β2m variants. Scale bars represent 200 nm. The 3D structure of the ΔN6‐derived fibril (PDB: 8a7o) revealed that both Asn17 and Asn42 are located near inter‐ or intra‐molecular interfaces within the oligomeric assembly (Figure S23), suggesting that isomerization from Asn to isoAsp at these positions substantially alters the molecular architecture to favor fibril formation. The distinct fibril morphologies between 1b/1e/ΔN6 and 1c/1d may be attributed to differences in the structural roles of Asn17 and Asn42. While both residues are located near inter‐ or intra‐molecular interfaces (Figure S23), Asn42 resides near a flexible β‐turn region, potentially modulating protofilament packing and lateral assembly. In contrast, Asn17 is situated at a more rigid interface crucial for stabilizing the fibrillar core. In this context, it is noteworthy that another localized conformational change—the cis–trans isomerization of the Pro32 peptide bond, located structurally between positions 17 and 42—has also been identified as a trigger for β2m aggregation. Previous work by Torbeev and Hilvert demonstrated that a synthetic β2m analog locked in the trans‐Pro32 conformation promotes the exposure of hydrophobic patches and drives amorphous aggregation, similar to the behavior we observed for our isoAsp42‐containing variants (1c/1d) [20]. However, because the trans‐Pro32 analog alone did not form typical amyloid fibrils, they proposed that a partial or transient cis‐to‐trans conversion may act synergistically to initiate oligomerization and subsequent fibril formation. Thus, while Pro32 isomerization is a distinct event, the respective roles of Pro32 conformational dynamics and isoAsp‐induced destabilization—particularly the potent amyloidogenic effect we observed for isoAsp17—likely represent independent yet complementary factors within the complex β2m fibrillization pathway. Therefore, isomerization at Asn17 likely disrupts the core architecture more directly than Asn42, leading to enhanced amyloidogenicity and more uniform fibril formation. Furthermore, fibrils derived from isoAsp17‐containing variants effectively acted as seeds to promote fibrillization of wild‐type β2m monomer (Figure S24), highlighting their nucleating potential. Taken together and considering the longer in vivo half‐life of Asn42 deamidation relative to Asn17 [10], our results indicate that isoAsp formation at Asn17 is likely a more critical driver of β2m amyloidogenesis. In conclusion, this study demonstrates that site‐specific deamidation at the Asn‐Gly site, particularly the formation of isoAsp at position 17, markedly destabilizes β2m and enhances its propensity for amyloid fibril formation. The aggregation rate of the isoAsp17 variant exceeds that of the ΔN6 variant and facilitates fibril seeding of the wild‐type protein. These findings implicate isoAsp formation at Asn17 as a critical early event in DRA pathogenesis. Moreover, this work highlights the utility of chemical synthesis in precisely modeling post‐translational modifications, offering valuable tools for understanding amyloid mechanisms and exploring therapeutic strategies. Author Contributions Ryuji Kawakami: investigation. Toshiki Takei: data curation, supervision, investigation. Masatomo So: investigation, data curation. Toshifumi Takao: conceptualization, data curation, writing – review and editing. Yoshinori Taguchi: investigation. Hironobu Hojo: writing – review and editing, writing – original draft, project administration, data curation, supervision, investigation. Conflicts of Interest The authors declare no conflicts of interest. Supporting information Supporting File: anie73088‐sup‐0001‐SuppMat.pdf.