Biopsy-resolved cryo-EM structures of amyloid fibrils provide molecular insights into AL amyloidosis

Significance

Determining amyloid fibril structures is essential for understanding disease mechanisms and guiding therapeutic development. However, most cryo-EM structures have been derived from postmortem tissues, leaving fibrils from living patients largely uncharacterized. Here, we present high-resolution structures of light chain (LC) amyloid fibrils extracted from cardiac and abdominal fat biopsies of three living patients with systemic light chain amyloidosis. We identify five distinct structures with patient- and tissue-specific features. Notably, fibrils from abdominal fat exhibit highly conserved global architectures compared to those from affected organs within the same patient. These findings offer insights into how LC sequence and tissue environment shape fibril conformation.

Abstract

Systemic light chain amyloidosis (AL) is characterized by amyloid fibril deposition in multiple organs, often severely affecting cardiac function. In this study, we extracted amyloid fibrils directly from abdominal fat and cardiac tissue biopsies obtained from three AL patients. Using cryo-electron microscopy, we determined five distinct structures of light chain (LC) amyloid fibrils. Our results demonstrate that LC fibrils from different patients adopt unique structural conformations, highlighting patient-specific fibril variations. Conversely, LC fibrils extracted from different tissues within the same patient share highly similar overall fibril structures, yet exhibit localized conformational variations, potentially shaped by distinct environmental cofactors. This study emphasizes the combined roles of patient-specific protein sequences and tissue-specific microenvironments in defining LC fibril conformation. The determination of LC fibril structures directly from easily accessible abdominal fat biopsy provides critical molecular insights into AL amyloidosis pathology, facilitating the development of therapeutic strategies.

Systemic light chain amyloidosis (AL) is the most common form of systemic amyloidosis (1–3), with an incidence of approximately 10 cases per million people (4, 5). It arises from the overproduction of amyloidogenic immunoglobulin light chains (LCs) by monoclonal B cells (6), leading to the formation of insoluble amyloid fibrils that cause progressive organ dysfunction (7, 8), primarily affecting the heart and kidney (9, 10). Cardiac dysfunction is believed to arise from amyloid deposits that extensively disrupt tissue architecture (11–13), resulting in a median survival of less than 1 y for patients with severe cardiac involvement after initial diagnosis (14–16).

Current diagnosis of AL requires both the detection of fibrillar deposits in tissue and evidence of plasma cell dyscrasia (4). Congo Red staining of tissue biopsies is essential for initial detection (17, 18), with direct biopsy of the affected organ serving as the gold standard (19–22). Less invasive methods, such as subcutaneous fat aspiration (23), bone marrow biopsy (24), or lip biopsy (25), are widely used in clinical practice. Among these, abdominal fat aspiration is particularly preferred due to its simplicity, low risk, and 84% sensitivity in cardiac AL amyloidosis (26, 27). Once amyloid deposition is identified, serum and urine immunofixation electrophoresis, serum free light chain assays are required to confirm the presence of a monoclonal protein (28, 29). Additional strategies include mass spectrometry (MS) for precise amyloid typing (30, 31), and cardiac imaging modalities (e.g., echocardiography, cardiac MRI) to evaluate disease burden (32, 33). Although the diagnostic workflow is well established, it remains unclear whether fibril structures from easily accessible biopsy sites accurately represent those in major affected organs, a key issue for understanding disease pathology and mechanism.

Recently, several studies have employed cryo-electron microscopy (cryo-EM) to determine the atomic structures of LC fibrils in AL amyloidosis, primarily from postmortem and explanted cardiac tissues (34–42). These studies reveal that each AL patient typically harbors a unique precursor LC sequence (43, 44), leading to distinct fibril conformations and underscoring the structural heterogeneity of AL pathology at the atomic level. In this study, we resolved five different atomic structures of LC fibrils extracted from biopsy samples of three living AL patients. Our findings not only confirm the presence of fibril polymorphs in AL amyloidosis but also demonstrate how environmental factors from different tissues may influence fibril architecture. By highlighting both tissue-specific polymorphism and patient-specific fibril structural variation, this work provides insights into the molecular determinants of LC amyloid formation and the role of cofactors in shaping fibril conformation.

Results

Characterization of LC Fibrillar Deposits in AL Patient Tissues.

Initially, we examined amyloid fibrillar deposits from biopsy-derived abdominal fat tissues of three AL patients (designated as AL cases 1, 2, and 3) and biopsy-derived heart tissue from AL case 1 using electron microscopy (Fig. 1A and SI Appendix, Table S1). In AL case 1, abundant fibrils were present around cardiac capillary vessel walls and throughout the interstitial spaces within fat tissue. In AL cases 2 and 3, fibrils were densely accumulated near blood vessels and in the interstitial spaces of fat tissues, appearing rigid and randomly oriented (Fig. 1A).

Fig. 1.

Characterization of amyloid fibrils in AL patients. (A) The ultrastructure of biopsy tissues damaged by amyloid fibrils. The fibril deposits are marked with yellow asterisks. Red arrows indicate blood vessels, and blue arrows indicate adipocytes. (B) Schematic workflow of amyloid fibril extraction from biopsy tissues and representative cryo-EM micrographs of eluted fibrils after extraction. Insets show fibrils with different morphologies, with white arrowheads indicating fibril branching. The cartoons were created with BioRender.com.

Given the abundant fibrillar deposition, we next extracted amyloid fibrils from these four biopsy samples for detailed structural analyses via cryo-EM. Through water-based extraction procedure, we successfully isolated fibrils from both heart and fat tissues (Fig. 1B). Cryo-EM grid preparation involved depositing 3 μL aliquots of the extracted fibrils, followed by cryo-EM data collection. The resulting micrographs revealed that fibrils isolated from heart and fat biopsies of AL case 1 exhibited similar morphologies. In AL case 2, individual fibrils displayed considerable variation in diameter, with some appearing branched. By contrast, AL case 3 fibrils were thin and straight (Fig. 1B).

Tissue-Specific Structural Polymorphs of Fibrils in AL Case 1.

To further resolve atomic structures of fibrils extracted from AL case 1, we collected cryo-EM data, including 3,063 micrographs from the heart-derived fibrils and 2,000 micrographs from fat-derived fibrils. Reference-free two-dimensional (2D) classification revealed that approximately 73% of cardiac-derived fibril segments and 88% of fat-derived fibril segments consist of a single protofilament, with the remaining segments exhibiting multiprotofilament morphology (SI Appendix, Fig. S1). However, 2D class averages did not resolve a clear crossover distance of the multiprotofilament class, precluding further helical reconstruction. Three-dimensional (3D) reconstructions of single-protofilament class showed fibrils with highly similar helical parameters (Fig. 2A and SI Appendix, Fig. S1 and Table S2). Interestingly, extra densities adjacent to the fibril cores were observed in similar positions of both heart- and fat-derived fibrils (Fig. 2 A and B and SI Appendix, Fig. S1).

Fig. 2.

Cryo-EM structures of amyloid fibrils from heart and abdominal fat tissues of AL case 1. (A) The reconstructed cryo-EM density maps of heart-derived fibril (colored in pink) and fat-derived fibril (colored in deep red). Fibril parameters including fibril widths, lengths of half pitches (180° helical turn), helical rises, and twist angles are indicated. Extra densities, colored orange, are displayed with the same threshold (0.0095) as protein densities. (B) Cross-sectional views of structural models of heart-derived and fat-derived fibrils fitted in the density maps, respectively. The density maps are restricted to areas within 2-Å radius of the structural models and then combined with the extra densities in (A). Residues adjacent to the extra densities are labeled. (C) Superimposition of heart-derived and fat-derived fibrils (global alignment). Primarily different regions are shown in boxes below.

We performed de novo sequencing via PEAKS AB software (45), revealing the fibrils originated from the immunoglobulin light chain encoded by the IGLV1-44*01 gene segment (SI Appendix, Fig. S2), a germline gene preferentially involved in patients with cardiomyopathy (46, 47). To address potential ambiguities in the MS-derived sequence, atomic structural models of both heart-derived and fat-derived fibrils were built by combining the MS-determined sequence with high-quality density map and the germline sequence (SI Appendix, Figs. S3 and S4 A and B). Remarkably, two fibril types agree closely over 76% of their amyloid cores, with an overall all-atom root mean square deviation (r.m.s.d.) of only 0.5 Å across 78 C_α atoms (Fig. 2C). The fibril cores, spanning residues P8–V109, consist of four or five β-strands (β1-β4 in heart-derived fibrils, and β1-β5 in fat-derived fibrils) which all arranged into a serpentine fold (SI Appendix, Fig. S5).

Despite their global structural similarity, notable local structural differences were observed, particularly between residues S25 and P45. Specifically, in heart-derived fibrils, residue Y26 was oriented outward, and S27 inward. This orientation was reversed in fat-derived fibrils, causing subsequent 18 residues to shift. Importantly, we observed distinct tissue-specific extra densities: one density adjacent to residue Y26 exclusively in heart-derived fibrils and another density near residue I29 only in fat-derived fibrils (Fig. 2C). To explore the possible effect of tissue microenvironment on fibril conformation, we conducted mass spectrum analysis on purified fibril samples. Interestingly, we identified several tissue-unique copurified proteins. For instance, desmoplakin was among top-ranking proteins in the heart-derived sample (SI Appendix, Table S3), whereas proteins such as amine oxidase 3, apolipoprotein E were highly abundant in the fat-derived tissue (SI Appendix, Table S4).

Collectively, our findings demonstrate that LC fibrils exhibit fundamentally conserved structural architecture across different tissues within the same patient. Although minor local structural variations might arise due to microenvironment-specific cofactors, the overall fibril conformation remains consistent. This indicates that amyloid fibrils extracted from easily accessible biopsy sites, such as abdominal fat, can largely represent fibril structures in other affected organs, supporting the diagnostic utility of accessible biopsies in systemic AL amyloidosis. Notably, although the overall fold is conserved, subtle differences (such as side-chain orientations and cofactor binding sites) could be functionally relevant.

Structural Polymorphism of AL Case 2 Fibrils.

We next investigated the cryo-EM structures of amyloid fibrils isolated from abdominal fat biopsies of AL case 2. Reference-free 2D class averaging identified three distinct fibril morphologies: polymorph 1 (PM1), polymorph 2 (PM2), and polymorph 3 (PM3), accounting for 49%, 41%, and 10% of the LC fibrils, respectively (SI Appendix, Fig. S6A). However, due to the limited number of particles, we obtained a density map of PM3 at 4.4 Å resolution (SI Appendix, Fig. S3 and Table S2), which was insufficient for reliable model building and therefore not included in the subsequent structural analysis.

The cryo-EM density map of PM1, resolved at 3.0 Å resolution, showed a left-handed helical fibril approximately 11.7 nm in width, with a half-pitch of 75.0 nm (Fig. 3A and SI Appendix, Table S2). The fibril core consists of two protofilaments intertwining along an approximate 2₁ screw axis, with helical rise and twist values of 2.4 Å and 179.4°, respectively. De novo sequencing revealed that the fibril-forming protein of AL case 2 is derived from the immunoglobulin light chain encoded by IGLV1-44*01 gene segment (SI Appendix, Fig. S7). Based on the high-quality cryo-EM density map and MS-determined sequence (SI Appendix, Figs. S3 and S7), we were able to unambiguously build the atomic model, which revealed that each protofilament core consists of eight β-strands arranged into two ordered segments spanning residues Q16–Q54 and D61–L110 (Fig. 3B and SI Appendix, Fig. S5). The two protofilaments form a large cavity, and their interface is stabilized primarily through electrostatic interactions. Notably, two additional extra densities were identified within the fibril: One adjusts to residue Y37 (site I), while the other is located near residues N53, Q54, T70, and S71 (site II) (Fig. 3B).

Fig. 3.

Polymorphic fibrils are observed in fat tissue of AL case 2. (A) Cryo-EM density map and parameters of PM1 fibril in AL case 2. The two protofilaments are colored in green and gray, respectively. Extra densities are colored orange and displayed with the same threshold (0.015) as protein densities. (B) Cross-sectional view and interface region of structural model of PM1 fitted in the density map. The density map is restricted to areas within a 2-Å radius of the PM1 model and then combined with the extra densities in (A). (C) Cryo-EM density map and parameters of PM2 fibril in AL case 2. Extra densities are displayed with the same threshold (0.013) as protein densities. (D) Cross-sectional view and interface region of structural model of PM2 fitted in the density map. The density map is restricted to areas within a 2-Å radius of the PM2 model and then combined with the extra densities in (C). (E) Structural comparison of atomic models of a single protofilament from PM1 and PM2, respectively.

We also determined the cryo-EM structure of PM2 at 3.2 Å resolution. Similar to PM1, PM2 consists of two protofilaments exhibiting an approximate 2₁ screw symmetry. The fibril is also left-handed but shows an increased width of 17.3 nm and a longer half-pitch of 103.8 nm, with rise and twist values of 2.4 Å and 179.6°, respectively (Fig. 3C and SI Appendix, Table S2). Structural modeling of the PM2 fibril core reveals a six β-strands architecture consisting of two segments spanning residues Q16–Q54 and T70–L110 (Fig. 3D and SI Appendix, Fig. S5). Interestingly, a similar extra density was again found near Y37 (site I). However, unlike PM1, the protofilament interface in PM2 is stabilized predominantly by hydrophobic interactions involving residues T43 and A44 (Fig. 3D).

A structural comparison between protofilament of PM1 and PM2 revealed high similarity, with an r.m.s.d. of 0.341 Å across 72 C_α atoms (Fig. 3E). The most notable difference resulted from variations in the core length of the protofilament. Specifically, each protofilament of PM2 lacks residues D61–G69, which are found in protofilament of PM1. Correspondingly, we found that the absence of D61 to G69 in PM2 protofilament correlated with the lack of an extra density at site II, which may stabilize T70 and adjacent residues in PM1 protofilament. These findings suggest that the structural polymorphism of fat-derived fibrils in AL case 2 may result from specific cofactors, which contribute to shaping certain fibril conformation.

Cryo-EM Structure of Fibril Derived from AL Case 3.

To investigate the structure of fibrils extracted from AL case 3, we collected cryo-EM images and performed 2D classification. Approximately 7% of fibrils exhibited a helical twist, allowing for structure determination via helical reconstruction (SI Appendix, Fig. S6B). The reconstructed 3D density map of the twisted fibril shows several extra densities on the surface of the fibril core (SI Appendix, Fig. S6B). The fibril core consists of a single protofilament, featuring a width of 8.4 nm and a half pitch of 60.4 nm. The rise and twist of the reconstructed fibrils are 4.80 Å and −1.43°, respectively (Fig. 4A and SI Appendix, Table S2).

Fig. 4.

Cryo-EM structures of fibrils extracted from abdominal fat of AL case 3. (A) Cryo-EM density map and parameters of fibril from AL case 3. Extra densities are colored in orange and displayed with the same threshold (0.00245) as protein densities. (B) Cross-sectional view of the structural model fitted in the density map. The density map is restricted to areas within a 2-Å radius of the model and then combined with the extra densities in (A). (C) Ribbon representation of the side view of the fibril core. The structures are colored by different layers. The difference in height within a single molecule is labeled.

De novo sequencing identified the fibril-forming protein is derived from the germline segment IGLV1-47*02 (SI Appendix, Fig. S8) of LC variable domain (V_L domain). The atomic model, built based on cryo-EM density and germline sequence information (SI Appendix, Fig. S4C), allowed fitting of a continuous LC polypeptide segment from R17 to V110. This fibril core contains eight β-strands with a central hydrophilic cavity stabilized by electrostatic interactions (Fig. 4 B and C). Similar to a previously reported LC amyloid fibril structure (36), the fibril chain does not adopt a strictly planar arrangement but exhibits variations in chain height along the helical axis (Fig. 4C).

Comparison of Fat-Derived LC Fibrils of Three AL Patients.

Given the high sequence diversity and structural variation of LCs, we analyzed the aggregation-prone segments in the fibril structures (SI Appendix, Tables S5–S7). The predicted aggregation-prone regions were all located within the fibril cores. Notably, a highly aggregation-prone segment (aggregation score 3 or 4) spanning residues 47 to 51 (⁴⁷V/L-LIY-S/N⁵¹) was identified, which also exhibits a strong aggregation tendency in other LC fibrils derived from IGLV1 (34). Besides, a conserved segment comprising residues 73 to 77 (⁷³ASLAI⁷⁷) also shows high aggregation propensity across these fibrils (SI Appendix, Fig. S9).

An intriguing finding of our fibril structures is the presence of extra densities adjacent to the fibril cores. To investigate their potential identity, we performed MS analysis on the fibril extracts. Consistent with previous reports (39, 41, 42), collagen VI chain alpha-1, alpha-2, and alpha-3 were identified in all cases (SI Appendix, Tables S4, S8, and S9). Collagen helices have been predicted to interact with exposed aromatic sidechains (41), and in our structures, extra densities near tyrosine are frequently observed (Y26, Y37, and Y50 in heart-derived fibril of case 1; Y37 and Y50 in fat-derived fibril of case 1 (Fig. 2B); Y37, Y50 in case 2 PM1 and Y37 in case 2 PM2 (Fig. 3 B and D); and Y50 in case 3 fibril (Fig. 4B)), suggesting a broader involvement of collagen VI in LC amyloid fibril formation.

Next, we evaluated the posttranslational modifications (PTMs) of LC fibrils. De novo sequencing identified a disulfide bond and deamidation in all cases (SI Appendix, Figs. S2, S7, and S8). The disulfide bond is highly conserved in all LC fibrils and is likely retained from the native LC V_L domain (35). Notably, deamidation has been implicated in amyloid formation across multiple diseases (48–50). In addition, patient-unique PTMs were also identified, including pyroglutamate modification of AL case 2 fibril (SI Appendix, Fig. S7) and oxidation in AL case 3 fibril (SI Appendix, Fig. S8). Though pyroglutamate modification and oxidation have also been reported in some LC fibrils (34, 51–53), their direct involvement in LC fibrillation requires further exploration.

To further characterize the biochemical properties of LCs from different patients, proteinase K (PK) digestion assay was conducted. AL case 1 fibril was more resistant to proteolysis, whereas AL case 2 and case 3 fibrils were more susceptible (SI Appendix, Fig. S10 A and B). To further investigate the seeding capability of these fibrils, we purified recombinant germline protein and conducted ThT kinetic assay in the presence and absence of patient-derived LC seeds. All patient-derived fibrils exhibited potent seeding ability, yet fibrils from different patients exhibit distinct kinetic profiles (SI Appendix, Fig. S10C).

Altogether, these results indicate that cellular cofactors, PTM heterogeneity collectively contribute to the structural and functional diversity of LC fibrils, potentially underlying differences in fibril stability and seeding behavior among patients.

Discussion

Systemic light chain amyloidosis is characterized by the deposition of insoluble amyloid fibrils in multiple organs (54), particularly the heart (21) and kidney (55), leading to severe organ dysfunction (30, 56). Understanding the atomic-level structures (57, 58) and mechanisms underlying fibril formation (59–61) is essential for accurate diagnosis, effective monitoring, and targeted therapeutic strategies development (62, 63). In this study, we utilized cryo-EM to determine the atomic structures of LC amyloid fibrils directly isolated from cardiac and abdominal fat biopsy samples from three living AL patients. Our findings underscore the importance of patient-specific LC sequences in determining fibril morphology, as each patient demonstrated unique fibril structures correlating with distinct LC sequences (Fig. 5A). Furthermore, we identified local conformational variations and tissue-specific extra densities associated with fibril cores, suggesting a significant influence of the local microenvironment on fibril architecture and stability (64, 65). Thus, our structural analyses revealed both tissue-specific polymorphism and patient-specific fibril structural variation, highlighting the roles of intrinsic LC protein sequences and extrinsic tissue environmental factors in determining the atomic structures of LC fibrils.

Fig. 5.

Schematic diagram of structural determinants underlying LC amyloid fibril formation. (A) Patient-specific precursor LC sequences define the unique LC fibril structures. (B) Tissue-specific microenvironments modulate the generation of distinct fibril polymorphs. Created with BioRender.com.

A key observation was the conservation of overall fibril architecture between different tissues within the same patient. Despite minor local conformational variations that may be associated with tissue-specific environmental factors, fibrils isolated from abdominal fat biopsies closely mirrored those found in critically affected organs such as the heart (Fig. 5B). Given that each fibril structure represents an average of approximately 10% of particles from the initial cryo-EM data, it is possible that both the heart-derived and fat-derived structures coexist within the same tissue, lessening the influence of the tissue microenvironment. This finding supports the clinical utility of using less invasive biopsy methods, such as abdominal fat aspiration, to represent the amyloid fibril structures present in major affected organs.

Our study markedly expands the repertoire of LC fibril structures. By comparing our structures with the other previously reported AL fibrils, we found that LC fibrils derived from different germlines display highly structural diversity. Moreover, even within the same germline, LC fibril structures from different patients remain heterogeneous, implying the potential role of patient–specific mutations in defining fibril conformation. Despite these variations, several conserved features were observed, including the consistent involvement of V_L domain in the fibril core with an intramolecular disulfide bond (SI Appendix, Fig. S11). Notably, the LC fibrils characterized in our study all originate from IGLV1, with two LC structures derived from IGLV1-44*01 (Case 1 and 2) and one from IGLV1-47*02 (Case 3). These two germlines have been significantly associated with AL amyloidosis, with patients harboring IGLV1-44 more likely to suffer cardiac involvement (31, 66). Structural comparison of fibril from the same germline (IGLV1-44*01) showed ~80% sequence conservation and revealed several common structure motifs (SI Appendix, Fig. S12), which may help explain the organotropism observed in certain germlines.

Furthermore, fragmentation of LCs has been widely reported, while the timing of this process remains controversial (67, 68). The homogeneity of LC fibril structures across different tissues raises the possibility that full-length LCs may undergo proteolysis in circulation prior to deposition in specific organs. This hypothesis is also supported by SDS-PAGE and de novo sequencing data, which revealed that the constant domain of patient LC fibril is always absent or in low abundance (SI Appendix, Figs. S2, S7, S8, and S13). It is possible that pathogenic LCs, mainly V_L domains, nucleate in the circulation and seed the growth of LC fibrils, which may subsequently infiltrate through vessels into affected organs and adapt to the local tissue microenvironment.

In conclusion, this study provides important insights into the structural determinants governing LC amyloid fibril formation, emphasizing the interplay between patient-specific sequences and tissue-specific cofactors. By resolving cryo-EM structures of fibrils directly obtained from clinically accessible abdominal fat biopsy, our study advances the molecular understanding of AL amyloidosis and establishes a structural basis for future therapeutic interventions.

Methods

Source of the Fibril-Containing Tissue and Ethical Statement.

The cardiac and abdominal fat tissues were obtained from AL patients during biopsies performed at Zhongshan Hospital. Detailed clinical characteristics of the patients are provided in SI Appendix, Table S1. Ethical approval of this study was granted by the Medical Ethics Committee of Zhongshan Hospital, Fudan University (ethics number: B2025-302). Informed consent was obtained from the donors.

Electron Microscopy Specimen Processing and Observation.

The electron microscope specimens were fixed in 2.0% polytoluene and 2.5% neutral glutaraldehyde for more than 4 h, washed with 0.1 M phosphate buffer. The samples were then fixed with 1% OsO4 for 2 h, dehydrated with acetone, permeated, embedded with Epon resin, and polymerized into resin blocks at elevated temperature. After ultrathin sectioning at 60 to 80 nm and staining with 2% uranyl acetate and 3% lead citrate, the sections were observed under Jeol electron transmission electron microscope JEM1400 (80 kv) and electron microscopy images were collected by EMSIS camera MORADA.

Fibril Extraction from Biopsy Samples.

The extraction of light chain fibrils was conducted using a previously established water-extraction protocol (38, 69). In brief, ~100 mg to 220 mg biopsy samples were cut into small pieces with a scalpel. The diced tissues were homogenized and washed five times in 1 mL Tris-calcium buffer (20 mM Tris, 138 mM NaCl, 2 mM CaCl₂, and 0.1% (w/v) NaN₃, pH 8.0) containing phenylmethylsulfonyl fluoride (PMSF) and cocktail (Roche), followed by centrifugation for 5 min at 3,100× g and 4 °C after each wash. The pellet and the fat layer were carefully retained during the first four washes, while after final wash, the supernatant and fat layer were removed. The pellet was resuspended in 1 mL of 5 mg/mL collagenase solution (Sigma, C0130, dissolved in Tris-calcium buffer). After incubating at 37 °C overnight, the tissue material was centrifugated for 30 min at 3,100× g and 4 °C. The supernatant was discarded, and the pellet was washed three times in 1 mL Tris-EDTA buffer (20 mM Tris, 140 mM NaCl, 10 mM EDTA, and 0.1% (w/v) NaN₃, pH 8.0), followed by centrifugation for 5 min at 3,100× g and 4 °C after each wash. The remaining pellet was resuspended in ice-cold deionized distilled water (ddH₂O). The suspension was centrifuged for 5 min at 3,100× g and 4 °C, and the washing step was repeated four additional times. All fibril-rich water elution was collected for further analysis when needed.

Mass Spectrometry (MS) Sample Preparation, Data Acquisition, and Analysis.

The MS sample preprocessing was optimized following an established protocol (70). The fibril-containing water elution was first dissolved in 2 M urea, 2% sodium dodecyl sulfate (SDS), and 2 mM dithiothreitol (DTT) and then boiled at 100 °C for 3 h. Then, the SDS-PAGE loading dye was added to the sample, followed by boiling at 100 °C for 15 min. For antibody sequencing, the prepared sample was loaded on 4 to 20% Bis–Tris gels (GenScript), followed by electrophoresis at 200 V for 30 min. The gels were stained by Coomassie brilliant blue. Consistent with previous report (36, 38, 42), we did not observe the full-length LC band at 25 kDa but instead detected major bands between 14.4 kDa and 18.4 kDa, which were collected by a clean scalpel into cubes for mass spectrometry analysis. For copurified protein analysis, the sample was loaded on 15% SDS-PAGE. Electrophoresis was performed at 80 V to allow the proteins to concentrate within the stacking gel. Once the samples migrated ~1 cm into the resolving gel, the electrophoresis was stopped. The gel was then stained with Coomassie brilliant blue, and all protein bands in the resolving gel were excised for downstream analysis.

The excised gel bands were reduced with 5 mM of DTT and alkylated with 11 mM iodoacetamide which was followed by in-gel digestion. For antibody sequencing, sequencing grade modified trypsin (Promega, V5111), chymotrypsin (Promega, V1061), Asp-N (Promega, V1621), proteinase K (Promega, V3021), a-lytic (New England Biolabs, P8113) were added. For analysis of copurified protein, sequencing grade modified trypsin was used for digestion at 37 °C overnight. The peptides were extracted twice with 0.1% trifluoroacetic acid in 50% acetonitrile aqueous solution for 30 min and then dried in a speedvac. Peptides were redissolved in 20 μL 0.1% trifluoroacetic acid and 6 μL of extracted peptides were analyzed by Thermo Scientific Orbitrap Exploris 480.

For LC–MS/MS analysis, the peptides were separated by a gradient elution (85 min for antibody sequencing, 120 min for copurified protein analysis) at a flow rate 0.30 µL/min with a Thermo-Dionex Ultimate 3000 HPLC system, which was directly interfaced with a Thermo Scientific Orbitrap Exploris 480 mass spectrometer. The analytical column was a homemade fused silica capillary column (75 µm ID, 350 mm length) packed with C-18 resin (1.9 µm, Dr.Maisch GmbH). Mobile phase consisted of 0.1% formic acid, and mobile phase B consisted of 80% acetonitrile and 0.1% formic acid. Orbitrap Exploris 480 mass spectrometer was operated in the data-dependent acquisition mode using Xcalibur 4.5.445.18 software and there was a single full-scan mass spectrum in the orbitrap (350 to 1,600 m/z, 60,000 resolution) followed by 2 s data-dependent MS/MS scans in an Ion Routing Multipole at 30 normalized collision energy (HCD).

Antibody sequence was analyzed by PEAKS AB (45). The search criteria were as follows: Each sample chose a specific enzyme; two missed cleavage was allowed; carbamidomethylation (C) were set as the fixed modifications; the oxidation (M) was set as the variable modification; precursor ion mass tolerances were set at 20 ppm for all MS acquired in an orbitrap mass analyzer; and the fragment ion mass tolerance was set at 0.02 Da for all MS2 spectra. Confidence levels were set to 1% FDR (high confidence).

For copurified protein analysis, the MS/MS spectra from each LC–MS/MS run were searched against the human.fasta using an in-house Proteome Discoverer (Version PD3.0, Thermo-Fisher Scientific, USA) combined with patient LC sequence determined in this study. The search criteria were as follows: Trypsin was chosen as specific enzyme; two missed cleavage was allowed; carbamidomethylation (C) were set as the fixed modifications; the oxidation (M) was set as the variable modification; precursor ion mass tolerances were set at 20 ppm for all MS acquired in an orbitrap mass analyzer; and the fragment ion mass tolerance was set at 0.02 Da for all MS2 spectra. Confidence levels were set to 1% FDR (high confidence).

Determination of V_L Germline Segments.

The amino acid sequence retrieved from mass spectrometry was used to search in IgBLAST database (71) [Germline V gene Database: IMGT human V gene (F + ORF+in-frame P)], which returned the top germline V gene hit.

Cryo-EM Sample Preparation and Data Collection.

3 μL of fibril-containing water elution was applied to glow-discharged holey carbon-coated grid (C-Flat, 300 mesh, 1.2/1.3, cat. 71159), blotted with filter paper, and plunge-frozen in liquid ethane using Vitrobot Mark IV (FEI, Thermo).

Cryo-EM images were collected on a Thermo Fisher Titan Krios G4 cryo transmission electron microscope at 300 kV, equipped with BioContinuum K3 direct detector (Gatan, Inc.) in counting mode. Inelastically scattered electrons were removed using a GIF Quantum energy filter (Gatan, Inc.) with a slit width of 20 eV based on the previous setup (72). The superresolution movies were recorded at × 105,000 magnification with a pixel size 0.83 Å pixel⁻¹ and the total dose was ~55 e⁻/Ų with exposure time of 2.0 s. Automated cryo-EM data collection was performed by using EPU software (Thermo Scientific™) with defocus values from −1.0 to −2.4 μm.

Cryo-EM Image Preprocessing and Helical Reconstruction.

For image preprocessing, 40 movie frames per micrograph were corrected for beam induced motion, aligned, dose-weighted, and further binned with a physical pixel size of 0.83 Å using MotionCorr2 (73). The contrast transfer function was estimated from motion-corrected images by CTFFIND-4.1.8 (74). Helical reconstruction was performed in RELION (75–77).

AL Case 1 Fibril Datasets.

Heart-derived fibril dataset.

Using RELION 3.1.0 software (76), fibrils were manually picked and extracted to segments with a box size of 288 pixels and an interbox distance of 23.9 Å. All particles were then re-extracted with a box size of 960 pixels and downscaled to 512 pixels. Two rounds of reference-free 2D classifications were performed with decreasing in-plane angular sampling rates (4° and 2°) and a T = 2 regularization parameter. 2D classification revealed two polymorphs, and their proportions were calculated based on particle numbers after 2D classifications.

Segments from PM1 exhibited a clear half pitch and were selected and used for 3D classification with k = 1, using an initial model constructed using the relion_helix_inimodel2d program (76). The best 3D class was selected as a reference model for the next round of classification (k = 3). After that, the best 3D class were selected and re-extracted with smaller box size (768 pixels and downscaled to 384 pixels) for subsequent 3D classifications. After multiple rounds of 3D classification (two rounds of k = 1 and two rounds of k = 3), the clearest class was selected and re-extracted with a 360-pixel box size. All 360-pixel particles were used for high-resolution 3D refinement. Local search of symmetry was carried out to optimize the helical twist and rise. Bayesian polishing and contrast transfer function refinement were used to further improve the resolution of 3D reconstruction maps. The final reconstruction map was sharpened using “Postprocessing” program with a soft-edge solvent mask. Overall resolution estimate was calculated based on the gold-standard 0.143 Fourier shell correlation (FSC) between the two independently refined half-maps. Then, the local resolution was estimated using the “Local resolution” procedure in RELION 5.0 with the same mask and B-factor in postprocessing.

Fat-derived fibril dataset.

Cryo-EM image preprocessing and helical reconstruction of the datasets were performed using RELION 3.1.0 software (76). Manually picked fibrils were extracted to segments with a box size of 288 pixels and an interbox distance of 23.9 Å. Then, particles were re-extracted with 960 pixels scaled down to a 512-pixel box size. 2D classification with a decreasing in-plane angular sampling rate from 2 to 0.5 and a T = 2 regularization parameter were performed to purified particles. All identifiable classes could be grouped into two fibril polymorphs, and the proportions were calculated using the particle numbers after 2D classifications. Purified PM1 segments were selected to splice the half pitch and generate an initial model using the relion_helix_inimodel2d program. These particles were then used for 3D classifications (two rounds of k = 3). The best classes were selected for further 3D classifications with a box size of 648 pixels. After optimizing the helical twist and rise with local search of symmetry, the clearest class was selected for 3D refinement with a smaller box size (360 pixels). CTF refinement and Bayesian polishing were applied, and the final reconstructions were generated by multiple rounds of additional golden-standard refinement. Sharpened map overall resolution estimate and local resolution estimate were processed in the same way as for heart-derived fibril dataset.

AL case 2 fibril dataset.

A total of 6,951 manually picked fibrils from 7,290 micrographs were extracted to segments with a 288-pixel box size and an interbox distance of 23.9 Å. The particles were then re-extracted with 1,050-pixel box size, downscaled to 448 pixels. Then, reference-free 2D classification steps with a sampling rate of 0.5 and a T = 2 regularization parameter were performed to classify segments. After discarding unidentifiable segments, the remaining particles were grouped into three polymorphs. Each polymorph was then selected separately for further reconstruction.

For PM1 and PM2, picked segments from 2D classification were used to calculate cross-over distances and construct an initial model. All particles were re-extracted to a box size of 512 pixels. Then, 3D classification (one round with k = 1 followed by one round with k = 3 for PM1, or one round with k = 3 for PM3) were performed. The best classes were selected. For PM1, an additional round of 3D classification was performed with a 360-pixel box size. Multiple rounds of 3D refinements combined with CTF refinement and Bayesian polishing were used for each polymorph to generate high-resolution density maps. The overall and local resolution estimates of the sharpened map were processed as previously described.

For PM3, selected segments from 2D classification were re-extracted with a 1,200-pixel box size and downscaled to 512 pixels and used for additional round of 2D classification to determine helical parameters. More purified particles were then selected for 3D classifications with a decreasing box size (1,200, 960, and 512 pixels). Subsequently, 3D refinements combined with CTF refinement and Bayesian polishing were performed. Then, the overall and local resolution estimates of the sharpened map were processed.

AL case 3 fibril dataset.

Manually picked fibrils were extracted using an interbox distance of 23.9 Å and a box size of 288 pixels. The particles were then re-extracted with a box size of 960 pixels scaled down to a 512-pixel box size for reference-free 2D classification (with a decreasing in-plane angular sampling rate from 2 to 0.5 and a T = 2 regularization parameter). Among the identifiable 2D classes, about 7% of the segments display clear helical twist and were selected for subsequent 3D classification. Several rounds of 3D classification were performed with a box size of 360 pixels to obtain a clear density map. Then, the best class was selected for high-resolution gold-standard refinement. CTF refinement and Bayesian polishing were performed, and overall and local resolution estimates were processed as previously described.

Atomic model building.

For AL case 1 and AL case 2, we manually build one-layer models with COOT (78). For AL case 3, we used ModelAngelo in RELION 5.0 to build the one-layer model (79). Then, one-layer models were manually docked into the central region of the sharpened density map in Chimera and ChimeraX to generate three-layer models (80, 81). Subsequently, the three-layer protein model was manually adjusted in COOT, followed by refinement against the corresponding map by phenix.real_space_refine program in PHENIX (82), with the restraints of rotamer, Ramachandran and geometry. Additional details for helical reconstruction and model building are shown in SI Appendix, Table S2.

Aggregation-prone region (APR) analyses.

Aggregation scores were computed using four different prediction tools: FoldAmyloid (83), Aggrescan (84), TANGO (85–87), and WALTZ (88). The criteria for identifying aggregation-prone residues were as follows: FoldAmyloid: five consecutive amino acids with a triple-hybrid threshold above 0.062; Aggrescan: at least five consecutive amino acids with an aggregation score greater than −0.02; TANGO: residue aggregation value over 0; WALTZ: residue with a score greater than 0. For each residue, the aggregation score was defined as the number of tools that predict it as aggregation-prone.

Purification of recombinant germline protein (GL protein).

Plasmids of the pET28a(+) vector encoding the IGLV1 germline sequence (Amino acid sequence: QSVLTQPPSASGTPGQRVTISCSGSSSNIGSNTVNWYQQLPGTAPKLL

IYSNNQRPSGVPDRFSGSKSGTSASLAISGLQSEDEADYYCAAWDDSLNGPVFGGGT) were synthesized by Azenta (Shanghai, China) and transformed into Escherichia coli BL21 (DE3) cells (TransGen Biotech, cat# CD601-02). The bacterial cells were cultured at 37 °C until reaching an OD₆₀₀ of 0.6, followed by induction by 1 mM isopropyl-1-thio-D-galactopyranoside at 25 °C for 16 h. After that, cells were harvested by centrifugation at 4,000 rpm for 20 min, and the pellet was lysed in 50 mM Tris, 500 mM NaCl, pH 8.0, with 1 mM phenylmethylsulfonyl fluoride (PMSF) using a high-pressure homogenizer (Union, UH-06). The cell lysate was centrifuged 14,000 rpm for 30 min at 4 °C. The resulting pellet was sequentially washed and sonicated with a buffer solution (50 mM Tris, 5% Triton X-100, pH 8.0), followed by centrifugation at 14,000 rpm, 30 min. The pellet was then washed and sonicated with an additional buffer (50 mM Tris, 1 M NaCl, pH 8.0), followed by centrifugation at 14,000 rpm, 30 min. The pellet was then dissolved in a denaturing buffer (6 M guanidine hydrochloride, 50 mM Tris, 50 mM NaCl, pH 8.0). The GL protein was purified using a Ni-NTA column (GE Healthcare) and eluted with eluted buffer containing 6 M guanidine hydrochloride, 50 mM PB, 50 mM NaCl, pH 4.0.

The eluted GL protein was diluted to a concentration of 0.5 mg/mL and subjected to a three-step dialysis process at 4 °C for the complete refolding. Buffers used in dialysis include 1) 2 M Guanidine hydrochloride, 50 mM PB, 50 mM NaCl, pH 6.7, 5% glycerin; 2): 1 M Guanidine hydrochloride, 50 mM PB, 50 mM NaCl, pH 6.7, 2.5% glycerin; and 3): 50 mM PB, 50 mM NaCl, pH 6.5, 1% glycerin. The refold proteins were concentrated and further purified using a Superdex 75 column (GE Healthcare, 28-9893-33) in 50 mM PB, 50 mM NaCl, pH 7.0 buffer.

Proteinase K digestion of LC fibrils.

The concentration of the fibril proteins was determined by dot blot using recombinant germline protein as a standard. For proteinase K digestion, fibril proteins were diluted in Tris-calcium buffer to a final concentration of 1 mg/mL. An initial aliquot was removed prior to PK treatment. Proteinase K (final concentration 0.05 mg/mL. Thermo Scientific™, EO0491) was added, and samples were collected after incubating at 37 °C for 0, 5, 10, 30, and 60 min. Upon collection, each aliquot was immediately supplemented with the same volume of buffer containing 8 M urea and 10 mM DTT, followed by rapid freezing in liquid nitrogen to terminate the reaction.

After all aliquots were collected, they were thawed at room temperature and prepared for denaturing gel electrophoresis. After adding the SDS-PAGE loading buffer, samples were denatured at 100 °C for 10 min and subsequently separated on 4 to 20% Bis–Tris gels (GenScript) by electrophoresis at 200 V for 30 min. Based on a previously established procedure (72), protein bands were transferred onto 0.2 µm PVDF membranes (Millipore). Following a 15-min blocking at room temperature with Protein Free Rapid Blocking Buffer (Epizyme) in Tris-buffered saline containing 0.5% (v/v) Tween20 (TBST), the membranes were incubated overnight at 4 °C with an Anti-Lambda light chain polyclonal antibody (Proteintech, Cat No. 11541-1-AP) diluted 1:1,000 in 5% BSA in TBST. Then, membranes were washed three times in TBST, followed by incubation for 1 h with Goat anti-Rabbit IgG HRP (Invitrogen, Catalog No.31460) diluted 1:10,000 in 5% milk in TBST. Membranes were washed three times in TBST. Super ECL Detection Reagent (Yeasen) was applied to membranes, and membranes were imaged using an iBright 1500 (Invitrogen).

ThT kinetic assay.

Patient-derived fibrils were diluted to 0.25 mg/mL in buffer containing 50 mM PB, 50 mM NaCl, pH 7.0, and then sonicated on ice using a JY92-IIN sonicator at 20% power for 60 cycles (1 s on/off per cycle) to prepare LC seeds. Subsequently, different LC seeds were added at concentrations of 5% w/w GL protein to 50 μM germline protein monomer with 50 μM ThT to monitor the aggregation kinetics. Following the previous procedure (72), the reactions were carried out at 37 °C with constant orbital shaking at 900 rpm in a 384-well optical plate (ThermoFisher Scientific, 142761). The fluorescent intensities were recorded hourly at 440 nm (excitation) and 480 nm (emission) wavelength using a BMG FLUOstar Omega plate reader (SI Appendix, Table S10). Kinetic curves were plotted and generated in GraphPad Prism 9.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Cryo-EM maps were deposited to the EM Data Bank under accession numbers EMD-63125 for AL Case 1 (Heart) PM1 (89), EMD-63124 for AL Case 1 (Fat) PM1 (90), EMD-63126 for AL Case 2 PM1 (91), EMD-63127 for AL Case 2 PM2 (92), EMD-66676 for AL Case 2 PM3 (93), EMD-63129 for AL Case 3 (94). The corresponding atomic models were deposited to the PDB under accession numbers 9LIW for AL Case 1 (Heart) PM1 (95), 9LIV for AL Case 1 (Fat) PM1 (96), 9LIX for AL Case 2 PM1 (97), 9LIY for AL Case 2 PM2 (98), 9LJ0 for AL Case 3 (99). The de novo sequencing data of AL case 1, case 2 and case 3 are available via ProteomeXchange with identifier PXD065430 (100), PXD065431 (101), and PXD065432 (102), respectively. Source data are provided with this paper.