Conformational differences in the light chain constant domain of immunoglobulin G and free light chain may influence proteolysis in AL amyloidosis

Abstract

Immunoglobulin light chain amyloidosis (AL) is a life-threatening disease caused by the deposition of light chain (LC) and its fragments containing variable (V_L) and portions of constant (C_L) domains. AL patients feature either monoclonal free LCs (FLCs) circulating as covalent and noncovalent homodimers, or monoclonal immunoglobulin (Ig) wherein the LC and heavy chain (HC) form disulfide-linked heterodimers, or both. The role of full-length Ig in AL amyloidosis is unclear as prior studies focused on FLC or V_L domain. We used a mammalian cell-based expression system to generate four AL patient-derived full-length IgGs, two non-AL IgG controls, and six corresponding FLC proteins derived from an IGLV6–57 germline precursor. Comparison of proteins’ secondary structure, thermal stability, proteolytic susceptibility, and disulfide link reduction suggested the importance of local vs. global conformational stability. Analysis of IgGs vs. corresponding FLCs using hydrogen-deuterium exchange mass spectrometry revealed major differences in the local conformation/dynamics of the C_L domain. In all IgGs vs. FLCs, segments containing β-strand and α-helix βA_C-αA_CB_C were more dynamic/exposed while segment βD_C-βE_C was less dynamic/exposed. Notably, these segments overlapped proteolysis-prone regions whose in vivo cleavage generates LC fragments found in AL deposits. Altogether, the results suggest that preferential cleavage in segments βA_C-αA_CB_C of FLC or βD_C-βE_C of LC in IgG helps generate amyloid protein precursors. We propose that protecting these segments using small-molecule stabilizers, which bind to the interfacial cavities C_L-C_L in FLC and/or C_L-C_H1 in IgG, is a potential therapeutic strategy to complement current approaches targeting V_L-V_L or V_L-C_L stabilization of LC dimer.

Graphical Abstract

INTRODUCTION

In amyloid diseases, or amyloidoses, proteins and/or protein fragments deposit as insoluble fibrils in various organs, leading to organ damage and, potentially, death. Over 40 different proteins and peptides that can form pathological amyloids in humans have been identified to-date.^1,2 Immunoglobulin light chain amyloidosis (AL) is the most common form of human systemic amyloidosis with an estimated annual prevalence over 10 cases per million.^3,4 This incurable life-threatening disease is a plasma cell dyscrasia that results from the overproduction of a monoclonal immunoglobulin (Ig) light chain (LC) by an abnormally proliferating B-cell clone, followed by the deposition of LC and its N-terminal fragments as amyloid fibrils in the heart, kidney, nerves and other organs. ^3–5 The monoclonal LC circulates either in a free form or bound to the heavy chain (HC) in the monoclonal Ig, most commonly of IgG type.⁶ Untreated AL amyloidosis is a rapidly progressing lethal disease.^3,4 Current therapies target the aberrant B-cell clone using chemotherapy that can be combined with bone marrow transplantation or immunomodulatory agents.⁴ These therapies, along with supportive treatments, help extend the patients’ lives and improve their quality, but do not cure the disease and have serious side effects. Much-needed new treatments are being developed, including targeted gene therapies to block the production of the aberrant protein precursor of amyloid, ⁴ as well as stabilization of the native LC conformation to protect it from misfolding. ^6–10

Native IgG is a disulfide-linked hetero-tetramer of ~150 kDa containing two LC and two heavy chain (HC) copies; each HC is N-glycosylated at Asn297 (Fig. 1B). Each LC contains two domains over 110 residues each, an N-terminal variable (V_L) and a C-terminal constant (C_L) domain, connected by a flexible linker (Fig. 1A, B).¹¹ In intact IgG, V_L and C_L LC domains pack against their HC counterparts, V_H and C_H1, and LC is connected to HC via the disulfide link near the C-terminus of C_L domain. In a stand-alone (free) LC (FLC), two LC molecules pack end-to-end to form a canonical homodimer with substantially hydrophobic V_L-V_L and C_L-C_L interfaces packed against each other.¹² The sole free Cys in LC can form a disulfide bond linking the two C_L copies via their C-terminal ends to form a covalent homodimer.¹³ Dimer stabilization is thought to protect an LC from misfolding and subsequent amyloid formation (^13,14 and references therein).

Figure 1.

Structural properties of proteins explored in the current study.

A - Top to bottom: Primary structures of six λ6 LCs used in the current study. Amino acid sequences of V_L and C_L domains are shown for the four AL patient-derived proteins, (AL1-AL4), two non-AL controls (Ctr5, Ctr6), and the corresponding germline IGVL6–57*01 (Germ). LC residues that differ from the germline are in red. Amyloidogenic segments predicted using AmylPred⁴² are color-coded as shown in the legend according to the consensus number. Linear diagram (marked 2°) shows native secondary structure based on the x-ray crystal structure of a closely related λ6 JTO LC⁸ (PDB ID: 6MG4) shown in panel C); arrows – β-strands, cylinders – α-helices; lines – loop/ turn/ unordered regions. IMGT (International Immunogenetics System) regions are indicated: FR – framework; CDR – complementarity-determining regions. Bars in C_L show: i) protease-sensitive segments (orange) identified by MS analysis of patient-derived AL deposits for closely related LCs; ^35–37 ii) segments showing decreased (purple) or increased (green) dynamics/exposure in IgG vs. FLC identified by HDX MS analysis of six protein pairs in the current study. Figure S1 shows the HC sequences of the IgGs explored.

B – Schematics of domain packing in full-length IgG and FLC dimer. Orange lines show intra- and inter-domain S-S links; blue squares show N-glycosylation sites at Asn297 of IgG. White ovals show small interfacial cavities in HC-LC heterodimer (in IgG) or LC-LC homodimer (in FLC). Interdomain space between paired V_L and C_L domains in the heterodimer (in IgG) or homodimer (in FLC) is shown by an irregular shape.

C – 3D structure of λ6 FLC homodimer (PDB ID: 6MG4). Molecule 1 is in light blue and molecule 2 is in gray. Dotted line shows V_L-C_L domain boundary.

Overproduction of monoclonal LC, either free or bound to HC and incorporated into an Ig, is necessary but not sufficient for AL development, as over 90% of patients with multiple myeloma or monoclonal gammopathies featuring excess circulating monoclonal FLC or Ig do not develop AL amyloidosis.¹⁶ Therefore, not only the quantity but also the quality of a monoclonal LC critically influences amyloid deposition in vivo. Prior studies of AL-derived FLC proteins or their V_L fragments have revealed some of these qualities that correlate with AL, including decreased kinetic stability, altered local conformation and dynamics, increased exposure of amyloidogenic segments, enhanced FLC proteolysis, altered domain-domain interactions, post-translational modifications, etc. (^15–24 and references therein). Nevertheless, no unified picture of the shared LC properties responsible for amyloid formation has emerged because of the highly variable amino acid sequence of V_L domains. It is generally believed that, during B-cell diversification, some of the V_L mutations and their combinations destabilize the native FLC assembly and/or stabilize the amyloid fold, and thereby promote amyloidogenesis. However, in AL and other amyloidoses, decreased protein stability is neither necessary nor sufficient for protein misfolding in amyloid, suggesting the role of other important but incompletely understood factors (^{10,18,21,22,25–27,28} and references therein).

AL amyloidosis presents a unique challenge as each patient features a unique monoclonal LC that is a product of germline gene recombination and somatic hypermutations.^20,24 Moreover, unlike intact IgG and FLC proteins, which have similar native structures, amyloid fibrils derived from different AL patients show distinct structures. ^28–33 This structural polymorphism, which reflects sequence heterogeneity of LCs³³ and their distinct fragmentation patterns,³⁴ suggests that the misfolding pathways of these LCs also vary. This variability complicates identification of sequence-specific properties that make an LC amyloidogenic.

One critical step in the LC misfolding in vivo and in vitro is proteolysis that releases V_L-containing amyloidogenic fragments.^23,34,35 Such fragments are found in AL patient-derived fibrillar deposits. ^28–31 In addition to V_L, AL deposits also contain longer fragments including parts of the C_L domain as well as full-length LCs.^{10,32,35–37} Unlike full-length FLC, which is resilient to amyloid formation in vitro, stand-alone V_L and some longer fragments found in AL deposits readily form amyloid in vitro and potentially in vivo.³⁵ The roles of these heterogeneous fragments in AL development and the timing of their formation via the proteolysis are subject of debate.^35–37 A series of careful mass spectrometry (MS) studies of AL patient-derived protein deposits by Lavatelli and colleagues^35–37 has revealed several proteolysis-prone segments in the C_L domain (orange bars in Fig. 1A). The authors suggested that the proteolytic cut in these segments is mediated by multiple broad-specificity proteases, such as matrix metalloproteases (MMPs) and cathepsins, and can occur either in the misfolded and/or in the native protein states.³⁵ In the latter case, such a proteolytic cut can release the misfolding-prone LC fragments, which can be detrimental to the formation of amyloid deposits. ²³

Since AL deposits contain full-length FLC and its various fragments containing intact V_L, prior in vitro studies focused solely on FLCs and their V_L domains^7,8,10 Moreover, a highly sensitive FreeLite assay utilizing antibodies specific for FLC detects FLC in circulation of over 90% of all AL patients,⁶ and very low levels of FLC may potentially form amyloid, as suggested by the clinical studies reporting that therapies leading to lower absolute levels of FLC are associated with improved clinical outcomes.³⁸ However, full-size monoclonal IgG found in circulation in AL patients may also play a role. In fact, a widely used serum immunofixation electrophoresis (SIFE) detects exclusively monoclonal FLC in nearly one third of AL patients, while nearly one third features only intact monoclonal IgG, and one third features both monoclonal IgG and FLC.^12,39 The role of full-length IgG in amyloid formation has not been explored.

We hypothesize that in a subset of patients featuring high concentrations of full-length monoclonal IgG in blood, IgG fragmentation may generate V_L-containing LC fragments that deposit as amyloid. The aim of the current study is to identify the properties of full-length monoclonal IgG molecules, such as the global and local conformational stability and dynamics, that can potentially contribute to amyloid fibril formation.

RESULTS and DISCUSSION

Clinical characteristics of AL cases and gene sequencing

The demographic, clinical and laboratory features of four AL cases explored in the current study are summarized in supplemental Table S1. All cases presented with low bone marrow plasma cell burden and λ LC restriction, had monoclonal intact IgG detected by SIFE, and showed abnormal serum IgG heavy-to-light chain ratios. Cases AL1 and AL2, corresponding to AL-204 and AL-222 patients (Table S1), had normal FLC κ/λ ratios, consistent with the absence of FLCs detected by SIFE. Cases AL3 and AL4, corresponding to AL-167 and AL-224 patients (Table S1), showed circulating FLC on SIFE and abnormal FLC κ/λ ratios. Two cases featured renal involvement, with elevated creatinine and nephrotic range proteinuria in one case; none of the cases were dialysis-dependent at the time of initial evaluation. Two other cases presented with cardiac and soft tissue involvement. Other involved organs included liver, gastrointestinal tract, and autonomic and peripheral neuropathy; all cases were treatment naïve at presentation.

The IGL and IGH gene sequences were derived from the patients’ bone marrow. Supplemental Table S2 lists primers used for sequencing, and Table S3 presents germline gene usage by IGL and IGH gene sequences. IGL and IGH genes were co-expressed to generate full-length IgG proteins using a mammalian cell-based expression system.⁴⁰ Additionally, IGL genes were expressed to generate FLC proteins as described in Methods.

Protein generation and characterization

A total of 12 recombinant proteins were generated in this study, including six full-length IgG proteins from four AL patient-derived (AL1-AL4) and two healthy subject-derived controls (Ctr5 and Ctr6). Two patients (AL1 and AL2) featured only monoclonal IgG by SIFE; two patients (AL3 and AL4) had both monoclonal IgG and FLC by SIFE; and IgGs from two healthy subjects (Ctr5 and Ctr6) were used as controls. Supplemental Methods list the criteria for the protein sequence selection. Additionally, we generated six FLCs used in each of these IgGs. All LCs were from λ6 family, which is predominant in AL amyloidosis and accounts for approximately a quarter of all AL cases. ⁴¹ Figure 1A shows amino acid sequences and structural properties of the LCs used, and Figure S1 shows the corresponding HC sequences for the six IgGs explored in this study. The recombinant proteins were assessed by non-reducing and reducing SDS PAGE (Fig. S2) and MS as detailed in the Supplement.

Amino acid analysis of LCs using a sequence-based consensus prediction algorithm AmylPred2⁴² identified several amyloidogenic segments with high propensity to initiate protein misfolding (Figure 1A). Consistent with prior studies, ²¹ most of these segments were in the V_L domain, yet one major amyloidogenic segment was in β-strand B_C (βB_C) of the C_L domain in residues 140–146, LVCLISD (corresponding to residues 136–142 using Kabat antibody numbering system, which for C_L differs by −4 from the continuous numbering used in the current study, Figure 1A). In the native structure, βB_C is linked to βF_C via the intramolecular Cys142-Cys201 bond, which stabilizes the major amyloidogenic segment in C_L and protects it from misfolding and aggregation. This adhesive segment is found in amyloid deposits of AL-55 and other patients containing various-length C_L fragments as well as full-length LCs^17,37 and refs. therein); however, this segment it is not a part of the fibril cores in the currently available fibril structures of the patient-derived AL deposits. ^28–32 The potential role of this C_L segment in amyloid formation remains to be determined.

AL amyloidosis does not correlate with reduced IgG thermostability

To determine whether amyloidogenic LCs were associated with destabilized IgGs, protein secondary structure and thermal stability were probed by circular dichroism (CD) spectroscopy. Far-UV CD spectra of different IgG proteins showed variations, yet all spectra were consistent with an Ig fold and showed a negative peak at 218 nm indicating predominantly β-sheet conformation (Figure 2A). Protein structural stability was monitored by CD at either 206 nm or 218 nm for β-sheet unfolding during heating at a constant rate. All IgGs showed a single sigmoidal transition with midpoints ranging from T_m,app = 64.0 °C for AL1 to 73.3 °C for AL4 (Figure 2B). Concurrently, protein aggregation was monitored by turbidity (dynode voltage) measured in CD experiments as previously described.^17,43 All proteins showed an initial increase followed by a decrease in turbidity, reflecting aggregation of the heat-denatured IgG⁴⁴ followed by protein precipitation (Figure 2B). After heating, selected IgGs were pelleted by centrifugation and visualized by transmission electron microscopy; large aggregates were observed in the supernatant, while the pellet showed micron-size droplets suggesting liquid-liquid phase separation in heat-denatured IgG (Figure 2C, insert). Importantly, both CD and turbidity measurements showed similar rank order of the apparent thermostability (T_m,app): AL1 ~ AL2 < Ctr6, AL3 < Ctr5 << AL4 (Figure 2B, C). Hence, AL1 and AL2 IgGs were less stable than AL3 and AL4 IgGs, while controls showed intermediate stability. To minimize IgG aggregation and phase separation of the heat-denatured proteins while retaining the native protein structure at ambient temperatures, we used 1 M urea in the buffer, which did not alter the far-UV CD spectra at 25 °C. Protein heat denaturation was monitored at 218 nm to minimize the noise in the CD data due to UV absorption by urea. CD melting data in 1 M urea showed biphasic transitions for AL1, AL4, and Ctr6 IgGs, (Figure 2D, E) reflecting consecutive unfolding of Fab domains followed by Fc domains,⁴⁴ while AL3 and Ctr5 IgGs showed concurrent unfolding of these domains, and AL2 rapidly aggregated upon unfolding (Fig. 1E). Again, AL1 and AL2 showed lower apparent thermostability than AL3 and AL4. Comparison of AL with control IgGs showed that association with amyloidosis does not necessarily correlate with decreased thermostability of IgG.

Figure 2.

Global structure and stability properties of IgG proteins.

A – Far-UV CD spectra of intact IgGs in standard buffer at 25 °C. Protein color coding (indicated) is similar in all figures.

B - Melting data recorded at 206 nm by CD during protein heating at a rate of 1 °C/min.

C - Melting data monitored at 206 nm by turbidity (dynode voltage) in CD experiments.

Insert: Electron micrographs of AL2 IgG aggregates formed after heating to 80 °C and cooling to 25 °C, followed by centrifugation to isolate the supernatant (super) from the pellet (indicated).

D – Melting data in the presence of 1 M urea monitored by CD at 218 nm.

E – Melting data in the presence of 1 M urea monitored by turbidity at 218 nm.

F - Limited proteolysis by pancreatin monitored by SDS PAGE (12%) under mildly reducing conditions (0.35 βME). IgGs were incubated with pancreatin for 5 to 240 min as indicated on the lanes; 0 marks intact IgG. Approximate positions corresponding to the migration of full-length IgG, HC and LC proteins are indicated for a reference. Brackets indicate different proteolytic patterns in AL vs. Ctr LCs.

G - Limited proteolysis by proteinase-K monitored by reducing SDS PAGE (8–16% gradient) under mildly reducing conditions (0.35 βME). IgGs were incubated with proteinase-K for 10 to 240 min as indicated on the lanes; 0 marks intact IgG. Additional details are described in Methods.

IgG proteolytic susceptibility was explored using two proteases with complementary broad specificity, pancreatin and proteinase K. Pancreatin is an enzymatic complex of lipases and proteases such as trypsin (which cleaves at the carboxylic side of basic residues), chymotrypsin (which favors aromatic residues) and elastase (which favors small aliphatic residues); proteinase K preferentially cleaves at the carboxylic side of aliphatic and aromatic residues. First, IgG proteins were incubated with pancreatin for up to 4 hours; the products were assessed by SDS PAGE under mildly reducing conditions (0.35 βME, Figure 2F). The bands at apparent molecular weight from about 250 to 150 kDa, which reflect anomalous migration of IgG and the presence of populations with distinct susceptibility to reduction,⁴⁵ declined faster for AL1, AL2 and Ctr6 vs. AL3, AL4, and Ctr5, in general accord with protein thermostability. Additionally, proteolytic fragments appeared different for different proteins: the bands migrating in the 22–26 kDa range, which likely correspond to FLCs and their fragments but may potentially include other fragments, showed a triplet in AL proteins but a duplet in controls (Figure 2F, brackets).

Next, IgG was subjected to limited proteolysis by proteinase K and the products were assessed by gradient SDS PAGE (Figure 2G). Different IgGs showed comparable proteolytic stability. Although AL4 IgG was most thermostable and apparently more resilient to proteolysis, other proteins showed no correlation between the rate of proteolysis and thermostability. Small differences in the proteolytic patterns of the four AL vs. two control IgGs were observed in the fragment range of 15–30 kDa and in the double band near 50 kDa (Figure 2G, black and teal brackets). The former finding is consistent with the differences in the 22–26 kDa range observed using digestion by pancreatin (Fig. 2F). These differences are potentially relevant, as proteolysis to liberate the aggregation-prone N-terminal LC fragments has been proposed as an early step initiating amyloid formation. ³⁵

Next, to test whether the interdomain disulfide links⁴⁶ have altered stability in AL vs. control IgGs, the IgGs were incubated with 0.1–10 mM β-mercaptoethanol (βME). Non-reducing SDS PAGE indicated a release of HC and LC upon reduction of intermolecular disulfides, followed by altered migration upon reduction of intramolecular disulfide bonds at high βME concentrations (Figure S3A). The results, together with a more detailed analysis using smaller steps of 0.1 mM βME (Figure S3B), indicated similar reduction rates for all six IgGs explored.

Finally, the kinetics of IgG aggregation/amyloid formation was explored. The proteins were incubated for 100 hours under amyloid-promoting conditions with 0.5 mM SDS in the presence of 0 to 1.0 mM βME. The goal was to reduce the intermolecular disulfide bonds and release LCs with intact internal disulfides (Figure S3) to mimic the intact internal disulfide in V_L domain of patient-derived AL deposits.^30,33 Thioflavin-T (ThT) fluorescence was monitored in real time; increasing signal reports on formation of amyloid-like structure but is complicated by the light scattering from protein aggregates. IgG proteins differed in their aggregation kinetics, with Ctr5 showing the greatest and AL2 and AL3 the smallest signal changes during incubation in the presence of various βME concentrations (Figure S4), including 0–0.2 mM βME when IgG retains its full length (Figure S3). These signal changes correlated neither with thermostability of a protein nor with its association with AL amyloidosis.

Taken together, the results show that the six IgGs differed in their overall conformation, thermostability, susceptibility to proteolysis, and aggregation properties, yet they showed similar susceptibility to the disulfide link reduction (Figures 2, S3, S4). The only small but consistent AL-associated features were apparent differences in proteolytic patterns of AL vs. control IgGs detected using either pancreatin or proteinase K (Figure 2F, G). We conclude that, compared to healthy controls, AL-derived monoclonal IgGs do not show decreased stability of the C_L-C_H1 disulfide bond that could promote the release of FLC. However, subtle alterations in proteolytic patterns of AL vs. control IgGs raise a possibility of different initial proteolytic cuts to release LC fragments; some of which potentially contribute to amyloidogenesis.

AL amyloidosis does not correlate with reduced FLC thermostability

Structural, stability and aggregation properties of FLCs were explored next (Figure S5). FLCs circulate as a mixture of monomers, which self-associate into non-covalent homodimers, and covalent homodimers disulfide-linked via Cys219¹³ (Figure 1B). In the current study, non-reducing SDS PAGE of all FLCs showed mainly covalent homodimers, along with a population of monomers (greatest in AL1 and Ctr1) and FLC fragments (minimal in AL1 and Ctr5 and maximal in AL2 and Ctr6) (Figs. S4, S5C). Far-UV CD spectra of AL1, AL3, AL4 and Ctr5 FLCs largely overlapped, suggesting similar overall protein conformations, while CD spectra of AL2 and Ctr6 FLC suggested increased helical content (Figure S5A). Melting data of all FLCs showed sigmoidal transitions varying in their midpoint temperature, T_m,app, and steepness, i.e. apparent cooperativity. Similar to IgG proteins, AL1 and AL2 FLCs showed lower thermostability compared to AL3 and AL4, while controls showed intermediate (Ctr5) or low stability (Ctr6) (Figure S5B). Consistent with this finding, limited proteolysis using proteinase-K showed increased proteolytic stability of AL4, AL3 and Ctr5 vs. AL1, AL2 and Ctr6 (Figure S5C). Finally, FLC aggregation kinetics in the presence of 0.5 mM SDS, which is used by us and others to promote LC misfolding and aggregation,²¹ was explored using ThT fluorescence as a readout. The aggregation was fastest for AL4, slowest for AL3 and Ctr5, and did not correlate with proteins’ association with AL amyloidosis. Electron microscopy of FLCs after incubation with SDS showed mostly amorphous and linear aggregates (Figure S5E).

Overall, the results in Figures 2 and S3-S5 suggest that IgG and FLC proteins have lower thermal and proteolytic stability for AL2 and AL1 vs. AL4 and AL3, while controls have intermediate (Ctr5) to low stability (Ctr6). Similar to prior studies of FLC and V_L domains,^{10,18,21,22,25–27,28} no clear correlation has emerged between the overall structural and stability properties of IgGs or FLCs and their association with AL amyloidosis. This compelled us to explore local conformational dynamics in these proteins.

Hydrogen deuterium exchange mass spectrometry reveals differences in IgG vs. FLC

To explore local protein backbone conformation/accessibility and dynamics, we used hydrogen deuterium exchange (HDX) mass spectrometry (MS) following published protocols²¹ that were modified to maximize the peptide coverage for both IgG and FLC proteins. Figure 3 shows a representative peptide coverage map illustrating common peptides in AL1 IgG and AL1 FLC (residues 1–220), which were used for analysis. Similar data for other AL proteins and controls are shown in the supplemental data file. Comparative HDX analysis did not reveal any trends that differentiated AL from control proteins, or AL proteins featuring intact IgG (AL1, AL2) from those featuring both IgG and FLC in circulation (AL3, AL4). However, the results revealed large, localized differences in deuterium incorporation in the C_L domain for intact IgG vs. the corresponding FLC. Importantly, these results were consistent for all six protein pairs and are reported below.

Figure 3.

Peptide coverage maps and deuterium uptake curves for a representative protein pair, AL1 IgG and AL1 FLC. Peptide coverage map for the LC variable domain, V_L (A, top panel), and constant domain, C_L (B, top panel). Together, individual gray and colored bars indicate those selected peptides that were identified from replicate HDMS^E analyses of undeuterated control samples (as detailed in supplementary data file). Colored bars show selected peptides representing near linear sequence coverage of different parts of V_L and C_L. The relative percent deuterium uptake curves for these peptides are shown beneath the coverage plots (A, B). Each uptake plot shows the relative deuterium level in Da versus labeling time for FLC (gray circles and dotted lines), IgG (red circles and solid lines). The gray dashed line indicates the measured maximum deuterium incorporation for the peptide (the maximum deuteration for peptide 49–63 was not determined). Each data point is the average of duplicate labeling experiments with error bars showing the aggregate errors calculated in DynamX. All HDX MS data used to make these selected graphs as well as the complete HDX MS data for all proteins are found in the supplementary data file.

Figure 3 shows deuterium incorporation curves (relative deuterium incorporation vs. labeling time, from 10 seconds to 4 hours) for representative peptides linearly across different parts of the AL1 LC. In both the V_L and C_L domains, the similarity of several curves for IgG and FLC suggests a similar local conformation and backbone dynamics for these regions in the context of either full-length IgG or FLC. However, incorporation curves for other regions of the C_L domain showed large differences in HDX suggesting different local backbone dynamics/exposure in IgG vs. FLC. For example, compared to FLC, IgG showed much lower deuterium incorporation (less exposed/dynamic conformation, increased structural stability) for the peptide covering residues 128–137 of AL1 LC, while HDX for the peptide covering residues 166–182 showed the opposite trend (Fig. 3B, bottom left).

These trends are further illustrated by the vertical deuterium incorporation heat maps (or chiclets) showing percent deuteration vs. labeling time for selected common peptides in AL1 IgG and FLC, %D IgG_AL1 and %D FLC_AL1, along with their relative difference in deuterium incorporation, ΔD=(IgG-FLC)_AL1, mapped on the 3D structure of an IgG Fab (Figure 4C, left panel; PDB ID: 7JX3) or FLC (Figure 4C, right panel, and Figure 4D; PDB ID: 6MG4); these structures contained a λ6 LC. The largest differences in protection were detected at all labeling times in two segments of AL1 C_L domain: IgG incorporated less deuterium (was more protected from HDX) than the FLC in residues 123–137 (purple Figs. 3A, 4 and supplemental data file), and IgG incorporated more deuterium (was less protected from HDX) than FLC in residues 166–183 (green, Figs. 3A, 4 and supplemental data file). Importantly, this trend was observed for all six pairs of proteins explored, as illustrated in their uptake curves and maps (supplemental data file), suggesting that other members of the λ6 LC family show a similar trend.

Figure 4.

Deuterium incorporation for IgG and FLC of AL1. (A) Percent deuterium incorporation shown as a vertical heat map for selected peptides from AL1 LC in IgG (left) for FLC (right) from N-to-C (top to bottom) and from 10 seconds to 4 hours (left-to-right). Percent deuterium incorporation values are colored according to the scale shown below the heat maps. (B) The relative difference in deuterium incorporation between IgG and FLC shown as a difference heat map for those peptides shown in panel A. The relative differences are colored according to the scale shown. The percent deuterium uptake (C) and the relative differences (D) shown in panels A and B are mapped on the ribbon diagrams showing structures of a Fab (PDB ID: 7JX3) and FLC (PDB ID: 6MG4) for the 10-minute deuterium labeling time point. CDR1 and CDR3 in V_L and peptide regions 129–137 (blue) and 166–183 (green) are identified in C_L. All HDX MS data used to make these graphs as well as the complete HDX MS data for AL1 proteins are found in the Supplementary Data File.

Across all time points, the AL1 residues 123–137 (which are less protected and show increased HDX in the FLC) encompass β-strand βA_C followed by an α-helix αA_CB_C; residues 166–183 (which are less protected in IgG and show increased HDX) extend from the end of βD_C to the end of βE_C. In the 3D structure of a closely related FLC (PDB ID: 6MG4)⁸, these segments are juxtaposed to their counterparts from the second dimer-forming molecule across the C_L-C_L interface (Figure 4D bottom right, purple and green). Therefore, the major differences observed by HDX MS in protection of FLC vs. IgG reflect the different domain-domain packing in the C_L-C_L homodimer in FLC vs. the C_L-C_H1 heterodimer in IgG.

Comparison of IgG with FLC shows that differences in protection were consistently greater in the C_L vs. the V_L domain in all protein pairs explored (Figure 4, supplemental data file). These differences may stem, in part, from the sole intermolecular disulfide involving Cys219 in the C_L domain, which links C_L to C_H in IgG but is either absent from FLC or replaced with the C_L-C_L disulfide. Interestingly, the differences in V_L of IgG vs. FLC clustered in complementarity-determining regions (CDRs) CDR1, CDR2 and CDR3 on the molecular surface of V_L furthest from C_L (Figure 4). These regions were generally better ordered in IgG vs. FLC in AL1 and in other protein pairs, despite some protein-to-protein variations (Figures 4, 5, and the supplemental data file). This observation suggests that increased rigidity of CDRs in an IgG vs. FLC decreases entropic penalty for antigen binding by an IgG.

Figure 5.

Relative difference butterfly plots for the six IgG and FLC pairs, including four amyloid proteins (AL1 – AL4) and two controls (Ctr5, Ctr6). Plotted values are calculated as the difference between the LC in IgG vs FLC, ΔD(IgG-FLC). Sequence information for the peptides selected are shown along the X-axis from N- to C-terminus. Each data point represents a different peptide; each color represents a differing labeling time as shown in the legend: 10 sec (orange), 1 min (red), 10 min (teal), 1 hour (blue), 4 hours (black). Negative relative difference values indicate that IgG is more protected from deuterium incorporation when compared to the respective FLC. All HDX MS data used to make these graphs as well as the complete HDX MS difference data are found in the supplementary data file.

Differences in deuterium incorporation of IgG vs. FLC for all proteins at all exchange times are summarized in Figure 5 that shows ΔD values plotted for selected peptides from N- to C-terminus (“difference butterfly plots”). Importantly, the major differences shared by all proteins were altered packing in C_L domain approximately in residues 126–140, which were less protected in FLC, and in residues 169–186, which were less protected in IgG. (These numbers are approximate as we are using a continuous amino acid numbering scheme based on a sequence alignment with the germline sequence for the C_L domain, Figure 1A). Additionally, significant protein-to-protein variations were observed, especially in residues 90–111 that overlap CDR3, βF_V, and a part of the joined (J) region. Since CDR3 and βF_V have highly variable amino acid composition (Figure 1), it is not surprising that HDX MS results suggest that this segment could adopt various conformations in different proteins. Notably, the major amyloidogenic segment in C_L domain residues 140–146 (Figure 1) showed no changes in HDX comparing IgG and FLC proteins.

Altered structural protection in C_L and generation of misfolding-prone LC fragments

Importantly, residue segments 126–140 and 169–186, which showed the largest differences in HDX protection of IgG vs. FLC, overlap LC segments that are particularly susceptible to proteolysis in vivo, as determined by MS analyses of LC fragments in multiple patient-derived AL deposits from λ6 and λ1 families^35–37 (Fig. 1, orange bars). These analyses are highly relevant to our work since, like the current study, they also explored LC proteins derived from the IGVL6–57*01 germline gene. The proteolytic cuts in these segments are thought to be mediated by various plasma proteases acting on either the misfolded or native proteins. In the latter case, such a proteolytic cut is proposed to trigger amyloid formation in vivo by releasing the misfolding-prone LC fragments containing V_L and parts of C_L domains.^32,35 Our limited proteolysis studies using pancreatin and proteinase K, which suggest distinct proteolytic patterns for AL vs. Ctr proteins (Figure 2F, 2G, S5C), are consistent with the key role of proteolysis in protein misfolding in vivo. However, neither the exact nature of the amyloidogenic fragments that are detrimental to the amyloid formation nor the proteases that generate these fragments in vivo have been unambiguously identified. It is also not known whether these proteases preferentially cleave FLC and/or IgG, and whether cleavage of FLC and IgG molecules by the same protease generates different LC fragments.

Increased proteolysis reflects decreased structural protection of the polypeptide backbone at the cleavage sites. This compels us to hypothesize that the proteolytic cut between residues 126 and 140 will preferentially occur in FLC, which shows decreased HDX protection in this residue segment as compared to IgG. Conversely, we hypothesize that a proteolytic cut between residues 169 and 186 will preferentially occur in IgG that is less protected from HDX at these sites (Figs. 3–5). If so, AL deposits in cases featuring monoclonal FLC in circulation are expected to contain increased populations of shorter fragments cleaved after residue 123, while those featuring circulating monoclonal IgG are expected to contain increased population of longer LC fragments cleaved after residue 166.

The native structure of the C_L domain is stabilized by an internal Cys142-Cys201 disulfide, which protects the protein from misfolding by restricting the solvent exposure and dynamics of the major amyloidogenic segment in residues 140–146 of βB_C (Figure 1). LC proteolytic cut in segment 126–140 removes both Cys142 and Cys201, while cut in segment 166–183 removes Cys201, so either cut eliminates the internal disulfide in C_L domain. This removal is expected to destabilize the resulting LC fragment and promote its misfolding/aggregation. Moreover, proteolysis in residues 169–186 will retain the highly amyloidogenic segment βB_C but eliminate the S-S bond that stabilizes βB_C; as a result, this major amyloidogenic segment will likely lose structural protection, which is expected to contribute to its misfolding/aggregation.

Stabilization of C_L-C_L and C_L-C_H1 interfaces as a potential therapeutic target

If proteolytic cuts in specific LC segments help initiate amyloid formation, then protecting the protein from such cuts provides a potential therapeutic strategy for AL amyloidosis. Notably, segments 126–140 and 169–186, which show large changes in backbone dynamics/exposure in IgG vs. FLC and overlap with the proteolytically labile regions identified in patients’ amyloid deposits,^35–37 form the lining of a water-filled cavity at the C_L-C_L interface in FLC homodimer (Figure 3). These segments also provide the lining for one side of the C_L-C_H1 cavity in IgG molecule (Figure 1). We hypothesize that binding of small-molecule stabilizers to these cavities in IgG and FCL will help protect these segments from cleavage while enhancing the overall protein stability. This approach complements current strategies developing small molecules to stabilize V_L-V_L cavity^8,10 or proteolytic targeting of the C_L-V_L interface for diagnostic and therapeutic purposes.^47,48 Compared to these strategies, our proposed approach has two advantages. First, due to the constant nature of C_L and C_H1, their therapeutic targeting is particularly attractive, since the same small molecules will stabilize different AL FLCs from the same family (e.g. λ6) in different patients. Second, since AL patients can feature excess circulating monoclonal IgG, FLC, or both, stabilizing both IgG and LC structures in critical regions will provide a comprehensive strategy to delay the progression of AL amyloidosis in diverse populations of patients.

Summary, hypotheses and future studies

The goal of the current study was to test whether full-length IgGs containing λ6 LCs, an LC isotype predominant in AL, may give rise to the V_L-containing protein precursors of amyloid. Comparison of six recombinant IgGs, including four AL-associated and two control proteins, with their corresponding FLCs revealed previously unknown differences in the local conformational dynamics of the C_L domain. The segments showing these differences between IgGs and FLCs overlapped the proteolysis-prone C_L segments identified by MS analyses of AL deposits.^35–37 Furthermore, AL vs. control IgGs showed subtle but significant differences in their proteolytic patterns produced by two broad-specificity proteases. Taken together, the results compel us to postulate that AL deposits in patients featuring high blood levels of monoclonal FLCs contain slightly shorter LC fragments as compared to those featuring high levels of full-length IgGs.

This hypothesis can be tested in future MS analyses of circulating FLCs, IgGs and AL deposits from such patients, as well as the circulating IgGs and FLCs from patients with multiple myeloma who also feature high blood levels of monoclonal proteins but do not develop AL amyloidosis. Whether or not this hypothesis extends to other LC isotypes should be tested experimentally, since prior HDX MS studies reveal substantial differences in local conformational dynamics of λ and κ-family LCs.^21,22 Another hypothesis that needs direct verification is that if proteolytic cut in C_L to generate LC fragments found in AL deposits occurs in the native structure of IgG or FLC, this cut may initiate the misfolding of the LC fragments by removing the stabilizing inter- and intramolecular interactions involving C_L and exposing the major amyloidogenic segment in βB_C.

Furthermore, we hypothesize that if on-pathway proteolysis with release of the amyloid protein precursors initiates AL amyloidosis, this initial step could perhaps be targeted by small molecules that bind and stabilize the C_H1-C_L interface in IgG or the C_L-C_L interface in FLC. This idea can be tested in future studies using in silico and in vitro small-molecule screening methods.

Although the comparison of global and local structural stability in four AL and two control IgGs and their corresponding LCs used in the current work has not revealed any AL-associated molecular features stemming from V_L mutations, this is not surprising. In fact, many previous studies using FLCs and their stand-alone V_L domains, including our own HDX MS analyses of AL vs. multiple myeloma and germline LC and V_L proteins,^21,22 failed to reveal a common misfolding mechanism stemming from V_L mutations. Rather, multiple alternative patient-specific misfolding pathways stemming from diverse destabilizing effects of V_L mutations on the FLC assembly have emerged (^15,17–28 and references therein). The current study supports this notion and extends it to full-length IgGs, highlighting the extreme complexity of potential amyloidogenic pathways in AL disease.

METHODS

Gene sequencing and sample selection for AL cases and healthy controls

IGL and IGH gene sequences were derived from the patients’ bone marrow. Bone marrow aspirates, clinical information and laboratory data were obtained from the sample repository and patient database maintained by the Boston University Amyloidosis Center. Informed consent for sample and data collection was obtained from all patients at presentation with the approval of the Boston Medical Center Institutional Review Board. Sequence selection criteria for monoclonal (AL) and non-AL control paired IGL and IGH sequences are provided in the Supplement, along with the GeneBank access numbers for all sequences.

Protein expression and characterization

All proteins were expressed in Expi293F mammalian cell system (Thermo Fisher Scientific). IgGs were purified by protein-A affinity chromatography as previously described.⁴⁰ LCs were purified by ion exchange and size exclusion chromatography as detailed in the Supplement. Protein purity for all full-length IgG proteins exceeded 95%, and for LCs it ranged from 78% (for AL2, which was least stable) to over 95% (for AL1 and Ctr5), as estimated from the reducing SDS PAGE analysis (Figs S2). MS analysis following peptic digestion confirmed the identity of all proteins. Protein folding was verified by far-UV CD spectroscopy. Additional details are provided in the Supplement. Recombinant intact IgGs migrated on non-reducing SDS PAGE as a single band with apparent molecular weight near 250 kDa (Figure 1B). Non-reducing SDS PAGE of FLCs showed that covalent dimer was predominant in all FLCs explored, with the dimer-to-monomer ratio ranging from 1.8 for AL1 to 18 for AL4 (Fig. S2). Previously we showed that the intermolecular S-S bond via Cys219, which distinguishes covalent dimer from the monomer in FLC, has minimal effect on HDX MS results, indicating similar packing in covalent vs. non-covalent LC dimers ²¹. Therefore, variations in the covalent dimer-to-monomer ratio for LCs did not significantly influence HDX MS analysis of the current study.

Biophysical analyses of proteins

Protein thermostability was assessed by heating/cooling at a rate of 1 °C/min monitored simultaneously by far-UV CD and turbidity measured in CD experiments as previously described.^17,43 Protein susceptibility to proteolysis was probed by limited digestion with pancreatin or proteinase-K followed by reducing SDS PAGE analysis in the presence of 0.35 mM βME. Disulfide link protection was probed by incubation with increasing concentrations of βME, from 0.1 to 10 mM, followed by non-reducing SDS PAGE. Protein aggregation/amyloid formation kinetics was monitored by ThT binding/fluorescence, and the aggregates in selected samples were visualized by transmission electron microscopy as described in supplemental Methods and in our prior work.^21,22 Local protein conformation and dynamics were explored by HDX MS as described in the Supplemental Data File and in our prior work.^21,22 For LCs we used continuous residue numbering in the current study. For more details, please see Supplementary Methods for protein selection, generation and analyses using biochemical and spectroscopic methods, Supplemental Data File for HDX MS analysis, and references therein.

Research Highlights.

• Recombinant immunoglobulins and free λ6 light chains from patients with AL amyloidosis and healthy subjects were compared

• Constant domain segments βA_C-αA_CB_C and βD_C-βE_C showed altered dynamics/exposure in all IgGs vs. free light chains

• These segments overlapped proteolysis-prone segments previously identified in patients’amyloid deposits

• Low protection in these segments may promote in vivo generation of amyloidogenic light chain fragments

• Stabilizing interfacial cavities C_L-C_L and C_L-C_H1 may hamper amyloid formation

Abbreviations:

AL

light chain amyloidosis

βME

β-mercaptoethanol

CD

circular dichroism

C_H

constant domain of the heavy chain

C_L

constant domain of the light chain

FLC

free light chain

HC

heavy chain

HDX

hydrogen deuterium exchange

IgG

immunoglobulin G

LC

light chain

MS

mass spectrometry

SIFE

serum immunofixation electrophoresis

ThT

thioflavin T

V_L

variable domain of the light chain

DATA AVAILABILITY

HDX MS data have been deposited to the ProteomeXchange Consortium via the PRIDE [Perez-Riverol-2020] partner repository with dataset identifier: PXD055570