Aggregation of Full-length Immunoglobulin Light Chains from Systemic Light Chain Amyloidosis (AL) Patients Is Remodeled by Epigallocatechin-3-gallate*

Intervention into amyloid deposition with anti-amyloid agents like the polyphenol epigallocatechin-3-gallate (EGCG) is emerging as an experimental secondary treatment strategy in systemic light chain amyloidosis (AL). In both AL and multiple myeloma (MM), soluble immunoglobulin light chains (LC) are produced by clonal plasma cells, but only in AL do they form amyloid deposits in vivo. We investigated the amyloid formation of patient-derived LC and their susceptibility to EGCG in vitro to probe commonalities and systematic differences in their assembly mechanisms. We isolated nine LC from the urine of AL and MM patients. We quantified their thermodynamic stabilities and monitored their aggregation under physiological conditions by thioflavin T fluorescence, light scattering, SDS stability, and atomic force microscopy. LC from all patients formed amyloid-like aggregates, albeit with individually different kinetics. LC existed as dimers, ∼50% of which were linked by disulfide bridges. Our results suggest that cleavage into LC monomers is required for efficient amyloid formation. The kinetics of AL LC displayed a transition point in concentration dependence, which MM LC lacked. The lack of concentration dependence of MM LC aggregation kinetics suggests that conformational change of the light chain is rate-limiting for these proteins. Aggregation kinetics displayed two distinct phases, which corresponded to the formation of oligomers and amyloid fibrils, respectively. EGCG specifically inhibited the second aggregation phase and induced the formation of SDS-stable, non-amyloid LC aggregates. Our data suggest that EGCG intervention does not depend on the individual LC sequence and is similar to the mechanism observed for amyloid-β and α-synuclein.


Intervention into amyloid deposition with anti-amyloid agents like the polyphenol epigallocatechin-3-gallate (EGCG) is emerging as an experimental secondary treatment strategy in systemic light chain amyloidosis (AL). In both AL and multiple myeloma (MM), soluble immunoglobulin light chains (LC) are produced by clonal plasma cells, but only in AL do they form amyloid deposits in vivo. We investigated the amyloid formation of patient-derived LC and their susceptibility to EGCG in vitro
to probe commonalities and systematic differences in their assembly mechanisms. We isolated nine LC from the urine of AL and MM patients. We quantified their thermodynamic stabilities and monitored their aggregation under physiological conditions by thioflavin T fluorescence, light scattering, SDS stability, and atomic force microscopy. LC from all patients formed amyloid-like aggregates, albeit with individually different kinetics. LC existed as dimers, ϳ50% of which were linked by disulfide bridges. Our results suggest that cleavage into LC monomers is required for efficient amyloid formation. The kinetics of AL LC displayed a transition point in concentration dependence, which MM LC lacked. The lack of concentration dependence of MM LC aggregation kinetics suggests that conformational change of the light chain is rate-limiting for these proteins. Aggregation kinetics displayed two distinct phases, which corresponded to the formation of oligomers and amyloid fibrils, respectively. EGCG specifically inhibited the second aggregation phase and induced the formation of SDS-stable, non-amyloid LC aggregates. Our data suggest that EGCG intervention does not depend on the individual LC sequence and is similar to the mechanism observed for amyloid-␤ and ␣-synuclein.
Systemic light chain amyloidosis (AL) 2 is the most common form of systemic amyloidosis (1), but it is still a rare disease with an incidence of 6 -7 in 1,000,000 people (2). Underlying AL pathology is a clonal plasma cell disorder, which produces an immunoglobulin light chain in the bone marrow with a unique sequence. These light chains are released into the blood (monoclonal gammopathy). When elevated LC levels in serum overwhelm renal absorption, LC is also found in the urine. Fulllength LC can be isolated from the urine of these patients (3)(4)(5).
Clonal plasma cell disorders can present different types of pathologies; light chains form amyloid deposits in AL patients. Here, LC fibrils adopt a conformation characterized by highly stable, intramolecular cross-␤-sheets that stain with Congo red (6) and thioflavin T (ThT) (7)(8)(9).
In contrast, amyloid deposits are not observed in patients suffering from multiple myeloma (MM), a malignant disease characterized by bone marrow failure and bone destruction. The two contrasting pathologies of LC gammopathies raise the question of which factors determine amyloid formation. Amyloid formation could be initiated (a) by the sequences of the individual light chains, (b) by high levels of light chain expression, (c) by posttranslational modification of the LC, (d) by fragmentation of the light chain, or (e) by a combination of these factors. A possible modification would be the formation or cleavage of intermolecular intradomain disulfide bonds.
Previous quantitative studies have focused on a small number of recombinant proteins spanning the variable domain of the LC proteins (18 -26). These systems cannot address the relevance of interdomain interactions in LC amyloid formation.
Our study, for the first time, quantifies stabilities and amyloid formation propensities of nine authentic, full-length LC proteins that we isolated from the urine of patients (4 with AL, 5 with MM). Our study tested whether AL and MM light chains had inherently different stabilities or amyloid formation propensities. We found that both AL and MM full-length light chains formed amyloid aggregates in vitro, but only under conditions that cleaved intermolecular disulfide bonds. Aggregation kinetics were unique for each LC protein and correlated only weakly with thermodynamic stability.
Treatment for the plasma cell clone pathology exists in the form of chemotherapy, but the direct treatment of amyloid pathology is still under investigation. Recently, we could show that the green tea polyphenol epigallocatechin-3-gallate (EGCG) redirects amyloid formation of amyloid-␤ and ␣-synuclein into inert amorphous aggregates (31)(32)(33). A case report of a patient who self-administered EGCG after several cycles of chemotherapy found that EGCG reduced LC deposits in the patient's heart and extended the patient's life over a period of several years (34). A reduction in cardiac pathology was observed in some but not all patients treated with EGCG (35). This observation raises the possibility that LC proteins from different patients may respond differently to EGCG.
Our study therefore tested the effect of EGCG on amyloid formation kinetics of our nine patient-derived LC proteins. We found that EGCG inhibited the conversion of oligomeric LC into ThT-positive fibrils and induced the formation of SDSstable aggregates in all LC proteins.

Results
Purification of Ig Light Chains from Patient Urine-Utilizing a simple protocol, we purified nine monoclonal Ig light chains from the urine of patients. Patients displayed either MM pathology or AL pathology. AL patients also displayed MM markers in addition to AL pathology (AL ϩ MM in Table 1). Instead of a standard dialysis approach (27,36,37), we utilized sequential membrane ultrafiltration, which removed small molecule components and concentrated the LC by a factor of 40 -80 ( Fig. 1A) (38). This protocol allowed us to parallelize the purification of a larger set of patient samples efficiently.
Purified LC selected for further biochemical and biophysical analysis contained less than 6% albumin and came from patients with a minimum of 265 mg of LC/24 h of secretion in urine (Table 1). These samples showed at most a weak 67 kDa band in silver staining (Fig. 1B, lane t) that was confirmed as albumin by immunostaining (Fig. 1B, lane a).
The isolated LC proteins each were of a single isotype that ran in SDS-PAGE as single bands with apparent molecular masses of 24 -28 kDa. Under non-reducing conditions, both monomers and S-S-linked dimers were present. Densitometric analysis of silver-stained proteins found 46 Ϯ 11% of LC dimers. However, analysis by blue native PAGE (Figs. 4B and 5A), size exclusion chromatography, and dynamic light scattering (data not shown) revealed that the light chain proteins did not, in fact, present as free monomer and dimers, as suggested by SDS-PAGE analysis. Rather, LC ran as distinct di-and oligomeric species between 50 and 600 kDa, suggesting the formation of non-SDS-stable multimers built from LC monomers and dimers. Purification of LC by an alternative dialysis protocol yielded di-and oligomers identical to those yielded by our membrane filtration protocol (Fig. 1C). These data confirm that oligomeric LC species were present in the patient urine and not a result of our purification protocol. Strikingly, light chains largely ran as monomers in native PAGE after reduction with DTT (100 mM; Fig. 1D), suggesting that reduction of intermolecular disulfide bonds destabilized LC oligomers.
Secondary Structure and Thermal Stability of Light Chains-When analyzed by CD spectroscopy, LC proteins exhibited the expected ␤-sheet structure (Fig. 2, A and B) (39,40). CD spectra could be grouped into -( Fig. 2A) and -isotype (Fig. 2B) LC (41). However, we observed no systematic differences between samples coming from MM or AL patients.
We then tested whether the stability of LC correlated with MM or AL pathology by recording thermal denaturation curves of the nine LC proteins (Fig. 2, C and D). Results are summarized in Table 2. Similar to the CD spectra, the melting points (T m ) grouped according to the LC isotypes (Table 1; -LC T m ϭ 48.6 -51.6°C, -LC T m ϭ 55.3-61.5°C) with the exception of the -MM-1 (T m ϭ 60.4°C, light green), which had a higher thermal stability than other -LC. These data suggest that the and -isotypes of constant domains have a stronger influence on the overall LC stability than the individual variable domains (42). Reduction of LC with DTT under native conditions had little effect on stability (Fig. 2E). In contrast, reduction under denaturing conditions reduced thermal stability by ϳ12°C (Fig. 2E). These results are consistent with literature data comparing the effect of interversus intramolecular disulfide bonds on LC stability (43). The results also indicate that only intermolecular disulfide bonds were reduced under the conditions of our aggregation assay. We observed a temperature gradient-dependent hysteresis of 2.6 -7.7°C between denaturation and renaturation curves (data not shown). This behavior has been described for recombinant V L domains and full-length LC (44). Although no obvious differences in thermal stability between AL and MM LC were observed during denaturation, -MM proteins did not renature into their native fold. These LC irreversibly formed aggregates after prolonged denaturation (not shown). Denaturation of all other LC proteins was at least partially (45-80%) reversible. This difference may reflect a difference in unfolding and aggregation pathways between light chains in AL and MM patients.
Thermodynamic Stabilities of Light Chains-We determined the free energies of unfolding via guanidine hydrochloride (GdnHCl) denaturation and monitored unfolding via tryptophan fluorescence (14,15,(45)(46)(47)(48) (Fig. 3A). The V L and C L domains contain a conserved tryptophan in position 41, the -isotype C L domain contains an additional conserved Trp-96, and 0 -4 non-conserved Trp residues may be present (16,17). Fluorescence spectra exhibited both changes in intensity and shifts of the emission maximum on unfolding (Fig. 3, A and B). We incubated all samples at 6 M GdnHCl overnight and then diluted the sample to 0.025 M GdnHCl to test for reversibility of unfolding (supplemental Fig. 1). CD spectra indicated that LC refolded into a native-like conformation ( Table 2 and supplemental Fig. 1).
We analyzed denaturation and renaturation data for the -AL-1 light chain recorded by Trp fluorescence and by CD spectroscopy in more detail (Fig. 3, C and D). Unfolding energies (10 Ϯ 1 kJ/mol) and refolding energies (Ϫ11 Ϯ 5 kJ/mol) measured by fluorescence and by CD (13 Ϯ 5 kJ/mol) were identical within experimental error. This indicated that unfolding was fully reversible and could be measured by both methods with similar accuracy.

LC Aggregation and EGCG
FEBRUARY 10, 2017 • VOLUME 292 • NUMBER 6 more stable than -AL light chains (-AL-1, 10 Ϯ 1 kJ/mol; -MM-2, 14 Ϯ 3 kJ/mol). In summary, MM light chains were marginally more stable than the AL light chains. Aggregation of Light Chains-LC proteins are deposited in AL patients into amyloid aggregates, which can be stained by amyloidophilic dyes, such as Congo red and ThT. We chose the -AL-1 light chain to establish conditions for amyloid formation in vitro.
We determined the sequence of the -AL-1 LC by MS peptide mapping combined with de novo MS sequencing (49). We found 10 point mutations compared with the parent germ line sequence (IGLV1-44*01; IGLC-3*03) (Fig. 4A). All mutations are in the V L domain. Four mutations are concentrated in CDR3 (complementarity-determining region 3). We analyzed the amyloidogenicity of the -AL-1 LC using the WALTZ algorithm (50). WALTZ predicted two amyloidogenic regions in the V L domain of the germ line parent at aa 32-38 and aa 45-51, respectively. Interestingly, mutations S36G and Q44H of the -AL-1 light chain flank the first amyloidogenic region. Surprisingly, mutations I54M and N65D in -AL-1 removed the amyloidogenicity in the second region. The analysis suggests that the relation between point mutations in clonal LC and amyloid formation is complex. The accumulation of amyloidpromoting mutations is probably not the only factor that determines amyloid formation in vivo.
We incubated -AL-1 LC for 1 day under permanent shaking in 25 mM sodium phosphate (pH 7.2), 120 mM NaCl at 37°C. The light chain remained soluble under non-reducing conditions. In contrast, it formed large aggregates when incubated under reducing conditions (8 mM DTT). In native PAGE, we observed LC aggregates (Ͼ720 kDa) and high molecular weight (HMW) aggregates incapable of migrating into the gel after 1 day of incubation (Fig. 4B). AFM imaging confirmed that large aggregates were formed in the presence of DTT, whereas in the absence of DTT, no aggregates were visible (Fig. 4C). The -AL-1 LC formed a mixture of amorphous and fibril-like structures at neutral pH (Fig. 4C). These aggregates bound the amyloidophilic dyes ThT and Congo red and displayed Congo red fluorescence and birefringence (Fig. 4, D and E). Aggregation kinetics followed a sigmoidal time course that is typical for amyloid formation (Fig. 4E). We analyzed the aggregation of all nine LC under the same conditions. AL LC formed aggregates with fibrillar substructure at neutral pH, whereas aggregate morphologies of MM LC were diverse. All light chains formed typical amyloid fibrils when incubated at pH 2.8 ( Fig. 4C and supplemental Fig. 2), displayed SDS resistance (Fig. 5), and bound ThT (Fig. 6). Hence, light chains isolated from patients' urine formed aggregates with amyloid characteristics under reducing conditions.
Characterization of Light Chain Aggregates by Native PAGE and SDS-PAGE-Resistance to denaturation by SDS is a characteristic property of amyloid. We monitored aggregation of light chains over 3 weeks to test whether aggregates would acquire resistance to denaturation in SDS. Samples were frozen after 0, 0.2, 1, 2, 5, 7, 14, and 21 days and stored at Ϫ80°C for further analysis. All nine LC proteins formed large aggregates within 1 day when analyzed by native PAGE (Fig. 5A). However, at this time point, LC aggregates were still disaggregated into monomers by 2% LDS at room temperature (semidenaturing conditions). Light chains formed aggregates that were resistant to denaturation at room temperature within 5-7 days (Fig. 5B). All nine LC samples displayed similar aggregation kinetics, although MM light chains rapidly formed very large SDS-resistant aggregates that did not enter the gel. In contrast, LC from AL patients produced aggregates that ran as dimers or oligomers (Fig. 5B), which may reflect differences in the aggregation mechanism.
To analyze differences in aggregate stability, we denatured LC samples by boiling in 2% LDS after a 21-day incubation. Under these denaturing conditions, LC aggregates dissociated into discrete ladders of monomeric, dimeric, and higher multimeric species (Figs. 5C (top) and 8). However, a substantial fraction of MM LC remained in the form of HMW aggregates that did not enter the gel. We then performed an ultracentrifugation experiment for 30 min at 200,000 ϫ g to test whether SDS-resistant oligomers were the result of a disaggregation of larger aggregates. Supernatants and washed pellets were denatured by boiling in 2% LDS and analyzed by denaturing SDS-PAGE (Fig. 5C). All LC protein was pelleted by ultracentrifugation, and no oligomers were visible in the supernatant fraction. This indicates that oligomeric species in SDS-PAGE resulted from partial denaturation of high molecular weight aggregates (Fig. 5C).
Aggregation Kinetics of Light Chains-We recorded aggregation kinetics of light chains by ThT fluorescence and by light scattering. Ig light chains formed ThT-positive aggregates at neutral pH under reducing conditions, which could be also observed turbidimetrically (Fig. 6A). We observed two types of aggregation kinetics that are represented by the four light chains shown in Fig. 6A. In most light chains, ThT aggregation kinetics showed a biphasic behavior with a very rapid initial rise in ThT fluorescence, as exemplified by LC -AL-1 and -MM-1 (Fig. 6, A and B, and supplemental Fig. 3, A and B). This rapid first phase of aggregation was invisible to turbidity measure-ments. This suggests that it corresponded to the formation of small oligomeric species (Fig. 6A). In contrast, the second phase, which was observable in both ThT and light scattering, corresponded to the formation of larger aggregates. However, the relative ThT fluorescence amplitudes of the first and second aggregation phases varied. Light chains -AL-1 and -MM-1 lacked detectable first aggregation phases.
We tested whether human serum albumin (HSA) had any influence on ThT kinetics of LC proteins by supplementing -AL-1 LC aggregation assays with HSA (supplemental Fig. 3  C). We found no effect of albumin on the first aggregation phase and a weak delay of the second aggregation phase at 10% of the LC concentration. This albumin concentration was higher than the HSA present in any of our purified LC samples. Therefore, we conclude that albumin had little or no effect on the LC aggregation kinetics under the conditions of our assay.
We could fit aggregation data to two sigmoid growth terms that corresponded to fast oligomer formation and to slower FIGURE 5. A, native PAGE of LC aggregated for 1 day. LC proteins formed large aggregates (Ͼ720 kDa) and HMW aggregates that were unable to migrate into gels. AL light chains produced mainly large aggregates, and MM samples produced mainly HMW aggregates. B, semi-native SDS-PAGE (no DTT, no heat) of LC aggregates formed in long term aggregation assays for 21 days. C, ultracentrifugation experiment of 21-day aggregates observed via denaturing SDS-PAGE (DTT, heating to 95°C). Aggregates from all LC samples dissociated into distinct ladder-like monomeric, dimeric, and multimeric species and residual HMW aggregates (top). These species are not soluble protein (middle) but insoluble aggregates sensitive to denaturing conditions (bottom).
fibril formation (Fig. 6B). We performed concentration-dependent aggregation assays between 5 and 100 M initial LC concentration ( Fig. 6B and supplemental Fig. 3A) and determined time constants for both phases of the aggregation kinetics ( Fig.  6C and supplemental Fig. 3B). The second phase of aggregation kinetics depended on protein concentration, whereas the first phase was largely independent of LC concentration. This suggests that a structural transition rather than an assembly step was rate-limiting in the first phase of the aggregation kinetics. We found that the first phase of aggregation accelerated with increasing concentration of DTT (supplemental Fig. 3D). These data strongly suggest that oligomer formation was initiated by the reduction of disulfide bonds.
We plotted the time constants (t 50 ) of the second aggregation step logarithmically as a function of LC concentration to analyze the underlying aggregation mechanism (51). All proteins displayed very weak concentration dependences with slopes m between Ϫ0.15 and Ϫ0.25 at low protein concentrations. Strikingly, concentration dependence of MM LC was weak (m Ͻ Ϫ0.2) in the whole concentration range probed in our experiment. In contrast, the concentration dependence of most AL LC curved from a flat slope at low concentrations to a slope of ϳϪ0.5 at higher protein concentrations.
Concentration dependences at low LC concentrations are much smaller than what would be predicted for a nucleated polymerization mechanism, for which the concentration dependence of ln t 50 scales with the nucleus size n as m ϭ Ϫ(n ϩ 1)/2 (52). A slope of m ϭ Ϫ0.5 is expected for aggregation that is dominated by secondary nucleation or by prion-like replication of amyloid fibrils (51). Therefore, our data suggest that fibril formation is dominated by unimolecular processes at low LC concentrations. A plausible explanation would be that a conformational conversion step of the LC limits the formation of amyloid fibrils. However, the aggregation of AL LC at higher concentration is dominated by protein assembly. Our data would be compatible with a mechanism that is limited by dimer formation or dominated by secondary processes (51). In contrast, conformational change remained dominant in MM LC under all conditions. Therefore, our kinetic data suggest that the difference in LC from patients with an AL diagnosis and those from MM patients may lie in the ease with which the LC converts into an amyloid-competent conformation.
The Effect of Epigallocatechin-3-gallate on Light Chain Aggregation-Having established the aggregation characteristics of our light chain proteins, we then analyzed the influence of EGCG on aggregate formation. We incubated Ig light chains under the same conditions as before in presence of equimolar (1ϫ) or 5-fold excess (5ϫ) of EGCG (Fig. 7, A and B). In all cases, EGCG completely prevented the second aggregation step but did not affect the first step in ThT kinetics (Fig. 7, A and B, and supplemental Fig. 4A). Correspondingly, the addition of EGCG to the LC sample after the first aggregation step (t ϭ 4 h) had the same effect on ThT fluorescence as the addition at t ϭ 0 h (supplemental Fig. 4B).
AFM imaging showed that EGCG did not prevent aggregate formation. LC formed aggregates of varied morphologies in the presence of EGCG (supplemental Fig. 2). We therefore tested whether EGCG had changed aggregate stabilities. We and others had previously observed that EGCG can induce the formation of SDS-stable aggregates of other amyloidogenic proteins (31,(53)(54)(55)(56). HMW aggregates formed in the absence of EGCG were broken down by boiling in LDS (Fig. 8). Although EGCG had little effect on LC aggregate sizes analyzed by native and semidenaturing PAGE (data not shown), it induced the formation of SDS-stable aggregates (Fig. 8). These SDS-stable aggregates increased in size with prolonged incubation times, whereas LC aggregates remained labile to denaturation in the absence of EGCG.
The data from SDS-PAGE suggest that EGCG induced the formation of insoluble, SDS-resistant aggregates. We tested by ultracentrifugation whether LC treated with EGCG did indeed become insoluble. Aggregates formed in the presence of EGCG after 21 days of incubation were pelleted by ultracentrifugation and analyzed by denaturing SDS-PAGE (Fig. 9). As expected, the aggregates formed in the presence of EGCG were insoluble and pelleted nearly completely.
Therefore, we conclude that EGCG had a similar effect on all nine examined light chains independently of the LC isotype and sequence and the originating disease. EGCG did not prevent the aggregation of the light chains. Rather, it induced their aggregation into SDS-resistant structures. While doing so, EGCG did not fundamentally alter aggregation kinetics of the nine LC relative to each other. LC proteins that rapidly formed HMW aggregates also more rapidly formed SDS-stable aggregates in the presence of EGCG and vice versa.

Discussion
Inhibition of amyloid deposition is emerging as a potential secondary treatment option in AL amyloidosis. Compounds like EGCG can inhibit amyloid formation in a wide variety of misfolded proteins in vitro and in animal models (57). Preliminary evidence suggests that EGCG may counteract amyloid deposition in AL patients and improve their prognosis (34,35,58). However, the fact that light chains are uniquely different from patient to patient poses specific challenges to the study and to the treatment of AL amyloid deposition.
A number of studies have scrutinized the factors governing LC amyloid formation, most of which have been performed on recombinant variable domains from five different LC proteins: Wil and Jto (18,19), AL-09 and AL-103 (20 -24), and SMA (25,26). Studies on V L domains (18, 59 -61) found that their amyloidogenicity roughly correlated with their thermodynamic stability. More recent studies have painted a more differentiated picture. Mutations that alter the interface in AL-LC dimers may alter amyloid formation without altering V L monomer stability (44,62).
We have generated detailed sequence information for one of our LC proteins (-AL-1) and found 10 substitutions from the germ line parent in the V L domain but no mutations in the C L domain. Half of the point mutations in this protein affected regions that are predicted to be amyloidogenic. Interestingly, these mutations did not uniformly increase amyloid propensity. Rather, two mutations flanked region 1 (aa 37-43), and two others strongly reduced the amyloid propensity of region 2 (aa 45-51). At the same time, we have no evidence that the mutations lowered the thermodynamic stability of the LC, because stabilities of all -LC were very similar to each other (Fig. 3F). This finding supports the hypothesis that mutations in AL LC alter the folding/unfolding pathway rather than destabilize global LC structure. Two mutations (T14A and Q44H) lie at or near the interface between the V L and C L domains, and five mutations lie at the LC dimer interface, suggesting that they may alter interdomain contacts. A future full analysis of all nine LC sequences will allow us to assess whether these structural features are unique to -AL-1 or whether they are common to amyloid-forming LC proteins.
The significance of full-length LC for amyloid formation was only recently addressed (27)(28)(29)(30). These studies find that interactions between the constant and variable domain of LC alter the unfolding/folding pathways and that the unfolding of native state LC is a complex, often irreversible, process. The independent unfolding of the two protein domains fails to explain the observed unfolding dynamics (28,29,63). Furthermore, thermodynamic instability is not sufficient for amyloid formation, suggesting the importance of kinetic factors (27,30). All studies suggest that the enhanced aggregation of the full-length protein results not only from the combined aggregating propensities of its variable and constant domain but also from the intramolecular interactions between these regions.
Our study supports the view that full-length LC differ not only in stability but also in reversibility of unfolding, suggesting different unfolding/folding pathways. Whereas the denaturation of most LC was reversible, all -MM LC denatured irreversibly after heating. We found thatand -isotype LC had different stabilities in thermal denaturation, with an average T m of 53 Ϯ 5 and 59 Ϯ 2°C, respectively, which corresponds to previously published data (28,30,44,64). We did not find strong systematic differences in stability between AL-and MM-LC. MM-LC had, on average, slightly higher unfolding energies than AL-LC. We also observed slightly higher melting points of -MM-LC compared with -AL-LC from thermal denaturation, which would correspond to the results from isolated V L domains (18). In general, our data support the model that amyloid propensity is unique to each LC and does not correlate strongly with thermodynamic stability.
Our study also supports data from isolated VL domains, which suggest that disruption of intermolecular disulfide bonds in LC dimers is a prerequisite of amyloid formation (61,65). This model is also supported by data from a recent drug screen (66). We found that ϳ50% of LC in urine existed as disulfidecross-linked dimers, and all LC existed as di-or multimers under native conditions. Efficient amyloid formation under our experimental conditions required the reduction of intermolecular disulfide bonds. Although disulfide bonds were not reduced in other aggregation studies of LC, these studies employed temperatures near the melting point, low pH, and/or very long incubation times to achieve amyloid formation (27,30). These results suggest that the stability of folded light chains needs to be compromised for amyloid formation. Although analysis of LC from serum by immunoaffinity LC-MS/MS (67) did not resolve to what degree disulfide-linked dimers are present in the serum of patients, our data suggest that a reducing milieu may initiate amyloid formation in vivo. Specifically, aggregation was initiated by the reduction of interchain and intermolecular disulfide bonds. In contrast, intradomain disulfide bond were only cleaved under denaturing conditions but stayed intact under native conditions (43). We found that LC were present in oligomeric form in the urine samples. After reduction of intermolecular disulfide bonds, LC ran as monomers in native PAGE, suggesting that disulfide bonds stabilized LC oligomers.
Both the V L and C L domains of LC display complex unfolding mechanisms (68,69). Our data suggest that reduction of intermolecular disulfide bonds did not alter LC domain structure. However, it made protein surfaces accessible that would otherwise be obstructed in the di-or oligomer. Notably, the amyloidogenic regions of the -AL-1 LC map to a single defined region on the surface of the V L domain (Fig. 4A). We hypothesize that monomerization enabled LC to enter a partial unfolding route that allows amyloid formation. This model would explain the two-phase aggregation kinetics with a first fast DTT-dependent oligomeric phase and a second phase where cross-␤-sheet structures are formed, which were inhibited by EGCG.
V L domains of light chains form amyloid fibrils under acidic and neutral conditions (18,70). Meshlike structures were observed after prolonged incubation of recombinant fulllength LC at neutral pH under permanent shaking (30). We found that patient-derived LC from AL patients formed similar fibril cluster structures after 3 weeks of constant agitation ( Fig.  4C and supplemental Fig. 2). These aggregates bound the amyloidophilic dyes ThT and Congo red and showed Congo red birefringence, suggesting that they contained cross-␤-sheet amyloid structures.
Aggregation studies of recombinant V L and LC find sigmoidal aggregation kinetics typical of amyloid formation. Our study found that native reducing conditions rapidly induce amyloid formation of authentic LC at physiological pH, salt, and temperature. Kinetics could be described by a common two-phase aggregation mechanism. The first phase was independent of the LC concentration and varied in relative ThT fluorescence amplitude. The second phase displayed sigmoid kinetics. Two-phase aggregation kinetics have been observed in A␤ aggregation, where oligomers formed in the first aggregation phase (71,72). Non-amyloid aggregation processes with oligomer formation also present similar kinetics (73).
The first phase of LC aggregation kinetics depended on the DTT concentration but not on LC concentration, suggesting that reduction of disulfide bonds is the rate-limiting process resulting in the formation of LC oligomers. The second aggregation phase was concentration-dependent for most light chains. However, analysis of the reaction order in a double logarithmic plot as a function of LC concentration suggests that conformational change rather than aggregate assembly or propagation may be limiting amyloid formation at low LC concentrations.
At high protein concentrations, aggregation of light chains from AL patients switched from kinetics that were dominated by structural conversion to kinetics that are compatible with a nucleated polymerization mechanism. These data suggest that AL and MM LC may differ in the kinetics of early conformational change, which makes the protein competent for amyloid formation. Our data also suggest that molecular AL and MM pathologies could exist on a continuum and that additional factors, such as protein homeostasis, may determine the ultimate fate of LC in vivo (74).
Although our use of authentic LC from urine avoided potential bias from lack of posttranslational modifications, it may have introduced a degree of sampling bias. We found that the albumin in the urine of AL patients with albuminuria tightly bound to LC and co-eluted with the light chains in size exclusion chromatography (data not shown). It was reported that albumin can bind proteins covalently via disulfide bond formation (75,76), but our biochemical analysis under non-reducing conditions indicated a non-covalent binding mode between the two proteins (Fig. 1B). The strong interaction between albumin and LC made it impossible to purify LC from the urine of patients with albuminuria under native conditions. We therefore excluded urine samples with high albumin content from the analysis.
We performed patient selection, sample purification, and biophysical studies blind to the detailed clinical diagnosis. After unblinding the patient samples, we found that our selection criteria (high LC, low albumin) directed AL patient selection toward AL deposits in soft tissue and gastrointestinal tract ( Table 1). As a consequence, the underlying plasma cell clone of all of these AL patients had expanded to symptomatic MM with bone lesions. The fact that our experimental approach selected for AL patients with a "weak" amyloid deposition phenotype may account for the modest differences in aggregation kinetics between AL and MM patient samples. It is reasonable to assume that LC proteins from AL patients with strong heart or kidney pathology would display a stronger difference in aggregation kinetics from MM LC, as had been observed in previous studies (27).
We have studied the effect of EGCG on nine authentic full-length LC proteins and found that EGCG had a generic, sequence-independent effect on light chain aggregation. It prevented the second aggregation phase and induced the formation of SDS-stable aggregates. These observations are similar to those in A␤ and ␣-Syn aggregation, suggesting a similar mechanism, in which EGCG preferentially interacts with aggregation intermediates of LC and inactivates them via off-pathway aggregate formation. This is seemingly in contrast to recent data finding that EGCG binds to the V L domain of LC with 1:1 stoichiometry and a dissociation constant of ϳ75 M (80).
However, binding studies on A␤ revealed that EGCG can interact both with the monomeric peptide (84,85) and with amyloid fibrils (56,82). When systematically reviewing EGCG mechanisms in various proteins, we found that EGCG binding to aggregation intermediates may inhibit amyloid formation very effectively in cases where these species are rate-limiting to amyloid formation (57). In other cases, amyloid formation is dominated by a structural transition or dissociation of the native protein. Here, EGCG may bind to native proteins and have less of an effect on kinetics but may nevertheless divert the protein into stable non-amyloid aggregates (57).
We find that EGCG effectively inhibits amyloid formation of a variety of light chains in vitro. This suggests that treatment with anti-amyloid drugs that either stabilize the native LC dimer or inhibit amyloid formation may be a promising supplemental therapeutic strategy to ameliorate the amyloid pathology in AL patients. It would therefore be valuable to assess the potential of drugs, such as EGCG, methylene blue, and Orcein/ O4, which have proven effective in inhibiting amyloid formation in other disease-related proteins, for the treatment of AL amyloidosis (32,66,86).

Experimental Procedures
Diagnosis and Patient Selection for This Study-Patients were diagnosed for monoclonal gammopathy by FLC from serum and immunofixation electrophoresis from serum as well as from urine (87). The diagnosis of AL amyloidosis was established via Congo red staining of fat aspirates and/or tissue biopsies (6) at the Amyloidosis Center Heidelberg according to established clinical protocols and in compliance with the ethical guidelines for treatment and patient consent. Patients were selected who presented with Ͼ265 mg/24 h clonal light chain in urine but without significant albuminuria (Ͻ6% albumin). 24-h urine samples were collected, and aliquots of 4 ϫ 15 ml were frozen afterward at Ϫ20°C for storage in the Heidelberg Amyloidosis Center biobank. Nine immunoglobulin light chains were purified from patient urine and subsequently analyzed: 2ϫ -AL, 2ϫ -AL, 2ϫ -MM, and 3ϫ -MM ( Table 1). The deposition patterns in the AL patients were blinded until after all experimental data were acquired.
Light Chain Protein Purification from Human Urine-Immunoglobulin light chains were isolated from patients suffering from AL or MM as follows. Particulate matter was removed from 20 -40 ml of patient urine by centrifugation for 15 min at 4,500 ϫ g at 4°C, followed by sequential filtration through first a 5-m and second a 0.22-m syringe membrane filter (Millipore). Afterward, the protein was concentrated in ultrafiltration units (Amicon Ultra 15) to 0.5 ml and buffer-exchanged in the same by six sequential dilution and reconcentration steps in 50 mM sodium phosphate (NaP) buffer, pH 7.2 (6 ϫ 15 ml).
Total protein concentrations were determined via BCA assay (Pierce). The albumin content was estimated densitometrically from silver-stained SDS-PAGE versus an HSA standard. Samples were adjusted to final LC concentrations of 1 mM, flashfrozen, and stored at Ϫ80°C.
LC-MS/MS-based Sequencing-Lyophilized -AL-1 sample (5 g) was reconstituted with 20 l of ultrapure water. Gel bands from reducing SDS-PAGE were excised, destained, reduced, and alkylated as described previously (49). The protein was digested in-gel at 37°C overnight with 0.025 g/l trypsin (Promega) or chymotrypsin (bovine; Promega) and for 3 h at room temperature with 0.025 g/l GluC (Staphylococcus aureus, Roche Applied Science) in 20 mM Tris, pH 8.2, 0.0002% Zwittergent 3-16. Peptide extraction was performed in two steps with 25 l of 2% trifluoroacetic acid and 50 l of acetonitrile. Samples were lyophilized and stored at Ϫ80°C before MS analysis.
Nano-LC-electrospray ionization-MS/MS measurements were performed as described previously (49) with an Orbitrap Elite detector and the following LC gradient of solvent B: 5 to 45% in 39 min, 45 to 90% in 2 min, 90% for 5 min, 90 to 5% in 2 min. The mass spectrometer performed a full scan MS scan from 340 -1,500 m/z, with a resolution of 140,000, followed by MS/MS scans at a resolution of 17,500. The Top20 [M ϩ 2H ϩ ] 2ϩ, [M ϩ 3H ϩ ] 3ϩ, and [M ϩ 4H ϩ ] 4ϩ ions were analyzed in a 1.5 m/z isolation window.
MS data were processed in PEAKS Studio software (Bioinformatics Solutions, Waterloo, Canada). For peptide mapping, a homemade IMGT-derived LC sequence database and the Uniprot human protein database were combined. The IMGTderived database contained V L -J L -C L transition peptide sequences, and complete V L and C L domain sequences. Database searches and de novo sequencing were performed assuming cysteine carbamidomethylation as a fixed modification, and methionine oxidations and asparagine/glutamine deamination as variable modifications. Additionally, the PEAKS PTM and the SPIDER search algorithm were used to detect PTMs and mutations. A precursor mass tolerance of 4.5 ppm and a fragment tolerance of 0.02 Da (49) were selected.
Relevant identified peptide sequences were determined by base peak interrogation as described previously (49) and aligned to the IMGT numbering system (16,17,45) with the IMGT/DomainGapAlign tool (88). Missing sequence patches were identified via inverse interrogation of the PEAKS results and LC-MS/MS data. The sequence representation shown here was created using the IMGT/Collier de Perles tool (89,90). Amino acids are numbered according to the IMGT numbering system. Amyloidogenicity of LC sequence was predicted using the WALTZ online tool, as described (50).
Gel Electrophoresis, Western Blotting Analysis, SDS Stability Assays, and Molecular Weight Determination-For SDS-PAGE and blue native PAGE, NuPAGE and NativePAGE precast gel systems (Invitrogen) were used with precast gradient gels (3-12% and 4 -12%, respectively). For SDS-PAGE, samples were diluted in LDS sample buffer (Invitrogen) either reduced with DTT (187.5 mM) and heated for 10 min to 95°C (denaturing conditions) or directly applied to the gel (semidenaturing). Gel bands were visualized either by Coomassie staining (EZ blue) or silver staining (ProteoSilver kit, both from Sigma-Aldrich) according to the manufacturer's protocols with minor changes. In short, the volumes of all kit components were reduced by a factor of 4, whereas the volumes of fixing solution, ethanol, and water washes were reduced by a factor of 2. The gels were stained in single-use plastic dishes, and the incubation with the silver solution was elongated to 15 min, following three short consecutive rinsing steps with ultrapure water.
For Western blotting analysis, proteins were transferred to nitrocellulose membranes. Membranes were blocked for 30 min in 3% milk in PBS and then were either immunostained for -isotype LC and HSA or -isotype LC and HSA in parallel. As primary antibodies, we used polyclonal goat anti-human Ig chain, polyclonal goat anti-human Ig chain (both 1:2,000; Millipore, AP506 and AP502), and polyclonal rabbit anti-human HSA (1:2,000; Thermo, R0101-1D). As secondary antibodies IRDye 800CW donkey anti-goat IgG and LI-COR IRDye 680RD donkey anti-rabbit IgG were used (both 1:15,000; LI-COR, 926-32214 and 926-68073). All antibodies were diluted in 3% milk in PBS. Secondary antibodies were detected in a fluorescent detection system (LI-COR) at 800-nm (LC) or 700-nm wavelength (HSA). We determined the apparent molecular weight of the light chains compared with a protein marker (Precision Dual Plus, Bio-Rad) from three independent SDS-PAGE samples.
CD Spectroscopy, Thermal and Chemical Denaturation-CD spectroscopy was performed in a JASCO J-810 CD spectrometer. CD spectra were accumulated between 265 and 190 nm with bandwidth 1 nm, response time 8 s, standard sensitivity, data pitch 0.5 nm, scanning speed 50 nm/min, 0.1-cm cell length at 20°C, and 3 accumulations at a concentration of 10 -20 M LC in NaP buffer. Thermal de-and renaturations were performed in triplicate at a temperature slope of 0.5°C/ min. Spectra before and after denaturation and after renaturation were acquired to test for refolding of the light chains. For thermal denaturation under reducing conditions, LC (200 M) was reduced with DTT (10 mM) for 2 h at room temperature either in NaP buffer or in 2 M GdnHCl. LC samples were renatured by dilution to 10 M in NaP.
GdnHCl denaturation was performed in biological triplicates by incubating the samples overnight at the respective concentrations of GdnHCl in NaP buffer before measurement. For renaturation samples, the protein was unfolded overnight at 100 M LC and 6 M GdnHCl, diluted to 10 M LC at the desired GdnHCl concentration, and incubated for at least 8 h. For Gdn-HCl denaturation, three CD spectra with three accumulations were measured between 213 and 211 nm using the spectrometer's interval measurement mode with bandwidth 1 nm, response time 16 s, standard sensitivity, data pitch 0.1 nm, scanning speed 20 nm/min, and 0.1-cm cell length, at a concentration of 10 M at 20°C. Protein unfolding curves were deter-mined at 211 nm. Spectra before and after denaturation and after renaturation were acquired as described above.
Tryptophan Fluorescence Spectroscopy, Guanidine Denaturation-We incubated LC samples overnight in biological triplicates, and Trp fluorescence was recorded between 300 and 470 nm in a spectrofluorimeter (Photon Technology International) at room temperature at 0.5 M LC in NaP, excitation at 280 nm. Samples for GdnHCl denaturation were prepared as described for CD but in disposable borosilicate glass culture tubes (10 ϫ 75 mm) (Fisher) to avoid the leaching of fluorescent molecules from plastic tubes into the sample.
To calculate the degree of unfolding, we determined the ratio between fluorescence intensities of the emission maximum before denaturation (e.g. 339 nm) and after denaturation (e.g. 351 nm). Different samples exhibited different peak shifts, which are summarized in Table 2.
Aggregation Assays-Short term aggregation assays were performed with the following parameters: 5-100 M LC, NaP buffer (pH 7.2), 150 mM sodium chloride, 100 mM DTT, 20 M ThT, 0.05% sodium azide, with or without EGCG (equimolar, 5ϫ equimolar; Taiyo) in a 100-l volume in non-binding surface 96-well plates, (Corning, catalog no. 3991). For measurements, we used a Tecan Infinite F200 filter based plate reader, 436/10-nm excitation band pass filter, and 482/25-nm emission filter. Measurements were performed at 37°C after 5 s of linear shaking with 2.5-mm amplitude in 15-min cycles. For turbidity measurements, absorption was measured at 560 nm. In concentration-dependent aggregation assays, we ran background traces containing the respective albumin concentration and subtracted the albumin signal before analyzing the data further.
Long term aggregation assays were performed in 200-l 8-tube PCR strips (Simport Scientific, Saint-Mathieu-de-Beloei, Canada) with 20 M LC, 25 mM sodium phosphate (NaP 2 , pH 7.2), 120 mM sodium chloride, 8 mM DTT, 16 M ThT, 0.05% sodium azide, and 50-l assay volume. Quadruplicates were incubated with and without 40 M EGCG for different time periods up to 21 days under permanent shaking with 200 rpm at 37°C in an incubator shaker (New Brunswick). Samples were stored at Ϫ80°C at each time point, afterward pipetted in non-binding surface 348-well plates (Corning, catalog no. 3544). Absorption and/or ThT fluorescence were measured as described above, including prior shaking. Samples were recovered from 348-well plates for further biochemical analysis. For EGCG stock solutions, EGCG was dissolved and diluted in ultrapure water directly before use.
Ultracentrifugation Assays-Ultracentrifugation assays were performed in a Beckman TL-100 ultracentrifuge, with a TLA-100 fixed angle rotor in thick walled polycarbonate 200-l tubes (Beckman Coulter) at 75,000 rpm (189,000 -245,000 ϫ g) and 4°C. The pellets were washed with 200 l of NaP buffer, again centrifuged for 30 min, and resuspended in SDS-PAGE LDS sample buffer, 187.5 mM DTT, and boiled for 5 min and analyzed by SDS-PAGE.
Atomic Force Microscopy Imaging-For AFM measurements, 10 l of 5 M sample was applied to grade V-1 Mica (catalog no. 01792-AB, Structure Probe, Inc.), followed by six wash steps with 15 l of ultrapure water. Samples were dried overnight, and AFM images were acquired on a Veeco Dimension 3100 machine with Bruker FESP tips. Scans were accumulated in tapping mode with 512 points/line for 10 ϫ 10-or 3 ϫ 3-m scan range, with a scan rate of 0.6 Hz.
Congo Red Staining and Imaging-Congo red (50 M) was added to -AL-1 LC (20 M) that had been aggregated for 1 week under standard conditions. The aggregate solution was pipetted on a coverslip and imaged by fluorescence microscopy and by birefringence in a Zeiss Axio imager at ϫ5 magnification. Fluorescence was detected at 546 nm excitation and 575-640 nm emission. Birefringence was measured under white light using two polarizers angled at 90°.
Evaluation of Unfolding Data-To determine unfolding energies for light chain proteins by CD and/or fluorescence spectroscopy, different properties had to be evaluated for different samples due to structural characteristics: (a) the ellipticity at 211 nm for the fluorescence intensity for samples -AL-1, -MM-2, and -MM-2; (b) the fluorescence intensity at 350 nm for the -AL-2; and (c) the ratio between fluorescence intensities at the lowest and the highest wavelength at the peak maximum observed for the other -LC samples. Unfolding curves were calculated from raw data as a function of the GdnHCl concentration by first correcting for pre-and post-transition baseline. Second, we assumed a two-state transition for the unfolding, an equilibrium state at each data point, and a completely folded and unfolded state at the beginning and end GdnHCl concentration, respectively. For each sample, the transition unfolding equilibrium constants K and ln K were determined as a function of the GdnHCl concentration. The Gibbs free energy (⌬G 0 ) was determined via linear regression from Equation 1 (91).
ln K ϭ ϪmЈ ϫ c denaturant ϩ ͩ ⌬G 0 ϪRT ͪ (Eq. 1) Fitting of ThT Aggregation Curves-Concentration-dependent aggregation data were fitted with a two-phase Boltzmann growth function with a linear correction term (Equation 2). Time constants of Boltzmann growth phases were plotted double logarithmically as functions of initial LC concentration and data points fitted with linear regression.