Immunoglobulin Light Chains Form an Extensive and Highly Ordered Fibril Involving the N- and C-Termini.

Light-chain (AL)-associated amyloidosis is a systemic disorder involving the formation and deposition of immunoglobulin AL fibrils in various bodily organs. One severe instance of AL disease is exhibited by the patient-derived variable domain (VL) of the light chain AL-09, a 108 amino acid residue protein containing seven mutations relative to the corresponding germline protein, κI O18/O8 VL. Previous work has demonstrated that the thermodynamic stability of native AL-09 VL is greatly lowered by two of these mutations, Y87H and N34I, whereas a third mutation, K42Q, further increases the kinetics of fibril formation. However, detailed knowledge regarding the residues that are responsible for stabilizing the misfolded fibril structure is lacking. In this study, using solid-state NMR spectroscopy, we show that the majority of the AL-09 VL sequence is immobilized in the fibrils and that the N- and C-terminal portions of the sequence are particularly well-structured. Thus, AL-09 VL forms an extensively ordered and β-strand-rich fibril structure. Furthermore, we demonstrate that the predominant β-sheet secondary structure and rigidity observed for in vitro prepared AL-09 VL fibrils are qualitatively similar to those observed for AL fibrils extracted from postmortem human spleen tissue, suggesting that this conformation may be representative of a common feature of AL fibrils.

Introduction

Light-chain (AL) amyloidosis is a severe form of systemic amyloidosis arising from the misfolding and deposition of immunoglobulin light chains as fibrils in the extracellular matrix of major organs throughout the human body, targeting sites such as the heart, kidneys, liver, spleen, and peripheral nerves.[1] The average prognosis for AL disease patients varies depending on which organ is affected, generally ranging from 2 to 3 years but dropping to below 1 year if the site of fibril deposition is the heart.[2−4] Furthermore, recent work has shown that the accumulation of AL fibrils around cardiomyocytes cultured in vitro may lead to cellular internalization of the fibrils and subsequent cell death by interrupting cell growth,[5] possibly through affecting normal metabolic function.[6] Devastatingly, the number of cardiac-afflicted AL patients accounts for approximately 50% of all AL disease cases, and despite ongoing efforts to more effectively hinder AL aggregation,[7−9] the availability of therapy options remains severely limited and primarily targets the plasma cell population secreting the pathogenic protein.[1−4]

AL misfolding in AL disease presents a particularly interesting case of amyloidosis in that the fibril-forming light chain generally differs by a few mutations among patients because of the natural introduction of somatic mutations during the process of generating a wide variety of antibodies.[10] One severe instance of cardiac AL disease is exhibited by the variable domain (VL) of light-chain protein AL-09 (derived from patient AL-09),[2] which differs from its germline counterpart (κI O18/O8 VL) by seven somatic mutations (Figure A).[11] Three of these mutations are nonconservative, Y87H, N34I, and K42Q, and are found to promote the propensity and rate of fibril formation at least in part by contributing a significant destabilizing effect on the native, soluble dimer form of AL-09 VL.[10] In addition, previous studies on the soluble dimer structures of AL-09 VL and κI O18/O8 VL have demonstrated that the interface of AL-09 VL is twisted by 90° relative to the germline dimer interface and is the site at which the three nonconservative mutations are located (Figure C,D).[1,11] Indeed, these observations strongly suggest that the presence of these nonconservative mutations in AL-09 VL contributes to the protein’s amyloidogenicity by distorting the VL–VL interface of the soluble dimer structure. Furthermore, solution-state NMR studies on closely related VL’s have demonstrated that the residues located at the dimer interface are predisposed to exhibiting a greater conformational flexibility relative to that of residues at most regions outside the interface, particularly under destabilizing conditions, thus implying that increased dynamics also contributes to the fibril-forming propensity.[12,13] However, there remains limited knowledge on the extent and manner of influence that the different regions of the light-chain VL have on stabilizing the final fibril form.

Figure 1

(A) Sequence alignment of AL-09 and germline κI O18/O8 VL’s. The highlighted residues indicate the mutation sites; those that are nonconservative are underlined. The residues labeled blue are site specifically assigned; those that are unambiguous are in bold.[14] (B) Averaged secondary 13C chemical shifts, ΔδC, plotted for each assigned residue in AL-09 VL fibrils. Values below or above the average random coil chemical shift values.[16] δCcoil values (i.e., ΔδC < 0 or ΔδC > 0 ppm) correspond to a β-strand or an α-helix secondary structure, respectively. (C, D) Crystallographic structures of the soluble dimer forms of the germline κI O18/O8 and AL-09 VL’s, respectively.[11] The mutation sites in AL-09 are drawn in red.

In pursuit of this interest, here we investigate the secondary structure and relative dynamics of AL-09 VL fibrils at the residue-specific level by magic-angle spinning (MAS) solid-state NMR (SSNMR) spectroscopy. Specifically, we demonstrate through chemical shift and dipolar coupling analyses that the majority of the residues in AL-09 VL fibrils are highly rigid and exhibit primarily a β-strand secondary structure. Moreover, our results reveal that most of the uniquely assigned rigid residues are located near the N- and the C-termini (i.e., before N30 and after L94) and maintain a high degree of structural order and homogeneity at a site-specific level. These observations also demonstrate notable structural differences between the fibril and soluble forms of AL-09 VL, such as residues Y32–Q38 and K45–D50, which are ordered as a β-sheet in the dimer crystal structure (Figure D) but are not observed site specifically in the fibril form (Figure A,B). By contrast, residues S12–V15 and L94–T97 are observed as β-strands in the fibril form but fall on the outside of β-strands in the soluble form (Figure ). Finally, we show that the recombinantly generated AL-09 VL fibrils studied here exhibit global structural features similar to those of AL fibrils directly extracted from human spleen tissue (referred to as HIG), supporting both the physiological relevance of our SSNMR studies and the possibility of identifying a structural motif that is common to all AL fibrils. Together, these results provide detailed insight into the extensive and unique structure of a toxic AL fibril, thus informing on the structure-based search for improved diagnostic and therapeutic strategies for AL disease.

Results and Discussion

AL-09 VL Fibrils Exhibit an Extensive and β-Strand-Rich Structured Region

On the basis of the resonance assignments for approximately half of the protein sequence for AL-09 VL fibrils (BMRB 26879) and observation of approximately 70 spin systems (the majority of which have been identified at the amino acid (aa)-specific level),[14] we pursued an analysis of the corresponding secondary structure and dynamics. In addition to obtaining initial insight into the secondary structure content of the fibril species through chemical-shift-based predictions of the backbone dihedral angles using TALOS-N[15] (Figure S1A),[14] we performed secondary 13C chemical shift (Δδ) calculations for residues in which the 13Cα, 13C′, and/or 13Cβ chemical shifts are assigned (Figure B).[16] Here, we report the averaged value of Δδ for each residue. Collectively, these 13C Δδ values inform on the type of secondary structure that a protein region exhibits—values close to zero indicate random coil (unstructured), whereas positive and negative values indicate α-helical and β-strand structures, respectively.[16] As illustrated in Figure B, the large majority of uniquely assigned residues are predicted to exhibit a β-sheet secondary structure, consistent with our refined TALOS-N predictions as well as findings typically observed for other fibrils.[14,17]

In addition to providing secondary structure information, the dipolar coupling-based SSNMR 13C–13C and 15N–13C correlation experiments report on the relative dynamics across the sequence of AL-09 VL in its fibril state, with cross-polarization (CP) emphasizing the rigid regions. Thus, a comparison of the relative CP intensities obtained from these experiments for different residues in AL-09 VL fibrils offers a qualitative measure of their corresponding rigidity. Here, we used the relative 13Cα–15N–13C′ (CANCO) peak heights of assigned AL-09 VL residues to inform on the site-to-site variation of rigidity across the sequence of the fibril species (Figure S1B). Our results reveal that the majority of the assigned regions of AL-09 VL exhibit CP intensities greater than 50% of the maximum peak height for residues with nonoverlapping signals. These residues are spread across the majority of the AL-09 VL sequence. Additionally, the general increase in intensity from residue Q6 to L11 and the subsequent decrease from I21 to D28 suggest that Q6 is near the beginning of its associated β-strand, whereas D28 is near the end and that the flanking residues may be on the periphery of the highly ordered region.

We further assessed the relative dynamics of AL-09 VL fibrils through a series of 15N–13Cα (NCA) two-dimensional (2D) experiments at temperatures below, near, and above the phase-transition temperatures (−30, −10, and +20 °C) of the bulk water in the samples (Figures A and S2). It is well known from a variety of amyloid systems that upon freezing the bulk water the CP intensity increases for the residues in mobile regions and disordered portions of the sequence exhibit broader signals at the lower temperature.[18−20] The NCA 2D spectrum provides multiple uniquely resolved cross-peaks corresponding to a wide range of residues over the AL-09 VL sequence; thus, these experiments allow for the temperature dependence of fibril rigidity to be assessed by monitoring the changes in the signal-to-noise ratio and line width of each peak. Accordingly, the sensitivity and resolution of 11 resolved 15N–13Cα cross-peaks were measured at each temperature (T0, G16, D17, I21, A25, S26, S67, T85, Y96, G99, and T102). These results demonstrate a consistent number of resolvable peaks for all temperatures, with only slight increases in the average line width from +20 to −30 °C (13 ± 8 Hz, 15N dimension; 8 ± 19 Hz, 13Cα dimension) and decreases in the signal-to-noise ratio (26 ± 15% lower on average). This finding indicates that the AL-09 VL fibril structure remains highly ordered across a wide range of temperatures and that this order is consistently maintained for the approximately 70 spin systems observed in total.[14]

Figure 2

(A) Overlay of 15N–13Cα 2D spectra with short contact times recorded at +20 °C (red) and −30 °C (blue). Individual spectra are shown in Figure S2. (B, C) 1Hα–13Cα dipolar coupling line shapes obtained from 2D T-MREV experiments at +20 °C (red), −10 °C (black), and −30 °C (blue) are plotted along ω1/(2π) from −5.0 to +5.0 kHz for integrated 13Cα regions of (B) 60–61 ppm and (C) 54–55 ppm (experimental, solid; simulated, dashed). All fitting simulations for the line shapes shown demonstrated order parameters of S ≥ 0.95. Additional fittings are provided in Figure S3.

To extend upon these observations and obtain a quantitative measure of protein dynamics, we performed 1H–13Cα dipolar coupling measurements for 13Cα one-dimensional (1D) regions between 50 and 63 ppm using the T-MREV[21,22] pulse sequence (Figures B,C and S3). The scaled dipolar couplings (and order parameters) corresponding to 1 ppm wide integrals of the 13Cα region were determined from fitting simulations (and normalization to NAV). Across the 13Cα chemical shift range of 50–63 ppm, the order parameters consistently demonstrated values of ≥0.95, supporting the presence of an extensive and highly ordered fibril. Furthermore, the order parameters for the 13Cα spectral region exhibited no significant dependence on temperature (for all fitting simulations; see Figure S3). Together, these dipolar coupling data provide strong support for the extensive rigidity of AL-09 VL fibrils, encompassing the majority of the 108 aa protein sequence (∼70 residues),[14] as well as a low dependence on temperatures above and below the bulk water phase-transition temperatures. It is interesting to note that this rather extensive structured region observed for AL-09 VL fibrils is slightly larger than the sizes of several previously determined fibril structures, such as amyloid-β (21–40 residues), human prion protein variant huPrP23–144 (∼28 residues), and α-synuclein fibrils (∼56 residues), yet comparable in size to that observed for PI3-SH3 fibrils (75 residues) and the long straight fibrils of β2-microglobulin (64 residues).[18,23−27]

N- and C-Terminal Residues of AL-09 VL Are Highly Ordered in the Fibril Structure

As demonstrated above, one prominent feature of AL-09 VL fibrils is that a large proportion of the most highly ordered residues is located near the N- and C-termini of the protein sequence (Figures and S1).[14] The 15N–13Cα–13CX three-dimensional (3D) strips in Figure A,B demonstrate the high resolution and sensitivity for a subset of the N- and C-terminal residues, supporting both their high rigidity and structural order. Furthermore, the observation of strong and well-resolved side-chain cross-peaks for these residues provides substantial support for their adoption of a single, ordered structure in the fibril state, as the existence of many different conformations would cause side-chain signals arising from multiple magnetization transfer steps to be much weaker and broader. Additionally, these residues appear to exhibit a homogeneous conformation because only a single set of corresponding backbone and side-chain cross-peaks is observed in each 3D strip (Figure ). For example, for residue V15, two unique chemical shifts (22.6 and 20.3 ppm) are observed for the geminal side-chain methyl carbons (13Cγ1 and 13Cγ2), indicating that the residue exists in only one conformation (Figure A). Collectively, these observations of well-defined side-chain rotamers in the termini are particularly interesting in contrast to those in the previous studies of other fibrils for which the termini are highly dynamic.[19,28] Given both the high rigidity and order of the N- and C-terminal residues, it is very likely that these two regions are the major participants in maintaining the rigid structural integrity of the fibril state. These data further suggest that AL-09 VL fibrils demonstrate a long-range structural order, which may be a unique feature of AL proteins relative to other reported amyloid fibril structures.[29,30]

Figure 3

Strip plot from the 3D NCACX spectrum of [13C, 15N]-labeled AL-09 VL fibrils, demonstrating the backbone and side-chain atom assignments for (A) S10–D17 (N-terminal residues) and (B) L94–T102 (C-terminal residues). The sequential backbone atom connectivity of these and all other assigned residues is described previously.[14] Data were acquired at 600 MHz with a MAS rate of 13.333 kHz and dipolar-assisted rotational resonance (DARR) mixing time of 25 ms.

To quantify the rigidity of the termini at a site-specific level, we performed the 3D R48318 version of R-symmetry sequences[31] with supercycled POST-C7 (SPC-7) 13C–13C mixing[32] to measure the scaled 1H–13Cα dipolar couplings of 14 residues in total, located at either the N- or the C-terminus (resolved in a third dimension via 13Cβ). Specifically, these residues include S7, S12–V15, V19–I21, and A25 (N-terminus) and A84, T85, L94, P95, and T97 (C-terminus) and were chosen because they exhibited strong 13Cβ–13Cα cross-peaks in the first CC 2D spectral plane (i.e., those above 15 times the noise floor) from which the dipolar dephasing trajectories could be readily extracted (Figure A). For all cases, the simulated 1H–13Cα scaling factors corresponded to order parameters of S ≥ 0.95 relative to the residue exhibiting the highest scaling factor, I21 (Figures B,C and S4). Collectively, the large magnitude of these backbone 13Cα order parameters reveals the extensive and high rigidity of the residues within both the N- and C-termini.

Figure 4

Site-specific 1Hα–13Cα dipolar coupling measurements for AL-09 VL fibrils. (A) First 13C–13C 2D plane of the R48318-symmetry 3D experiment, collected with SPC-7 q42 13C–13C mixing at 20 °C (the black and red peaks indicate the positive and negative signal intensities, respectively). Residues for which the 1H–13Cα dipolar coupling line shapes were extracted and simulated are labeled. (B, C) A subset of the experimental (solid) and simulated (dashed) line shapes illustrating the site-specific rigidity of several residues near the N- and C-termini (B and C, respectively) of the AL-09 VL sequence. The corresponding order parameter, S, for each 13Cα site is indicated next to the line shape (normalized to the site with the largest measured 1H–13Cα scaling factor). Additional fitting simulations are shown in Figure S4.

Interestingly, as described previously on amyloidogenic/AL VL’s, a “cryptic epitope” is proposed to occur within the first 18 residues of the N-terminus, in which the presence of a proline at position 8 in combination with residues 1–4 and 13–18 supports the formation of an aberrant bend within β-strand A of the native VL.[33] This conformational rearrangement of the N-terminus around P8 upon VL misfolding would be consistent with our observation of its involvement in the fibril structure, as the CANCO CP efficiency generally increases in either direction for residues before and after P8 (Figure S1B). Furthermore, the presence of a β-bulge in strand G at the C-terminus of VL’s (residues 97–108) has been implicated in serving as a protective structural motif that discourages fibril formation,[34] with a particularly prominent β-bulge in kappa VL’s over lambda structures.[35] It is possible that during fibril formation this motif undergoes a conformational change that supports an aberrant, yet highly ordered, conformation. Given that both the N- and C-termini of the AL-09 VL sequence contain an approximately 12 aa stretch of generally intervening hydrophilic–hydrophobic residues (i.e., S10–T22 and Y96–K107), one entropically favorable rearrangement that may occur could be through the interaction of these regions via their hydrophobic side chains. Certainly, understanding the specific orientation and interactions associated with the N- and C-terminal residues of AL-09 VL fibrils is fundamental to determining the tertiary fibril structure and may be pursued in future SSNMR studies by the measurement of intramolecular distance restraints[36,37] in combination with sparsely or selectively 13C-labeled samples.[17,36,38−40]

Additional insight into the structure and formation of AL-09 VL fibrils is provided by consideration of the effects of mutations on the protein stability. Globular proteins undergo partial unfolding to initiate misfolding events. This is unlike unstructured peptides (Aβ peptide) and intrinsically disordered proteins (α-synuclein) that initiate misfolding events from an unstructured or unfolded state. Mutations associated with amyloid diseases, in general, decrease the conformational stability of the globular native fold and promote aggregation in vitro. These mutations usually influence the thermodynamics of the process and may not participate in direct interactions within the fibril structure.[41] Our mutational analyses of AL-09 VL have shown that the nonconservative mutations located within the dimer interface of the soluble protein structure destabilize the protein and accelerate amyloid formation.[10] As nonconservative mutations in AL proteins may lower the energy barrier for the initial misfolding events to occur, we hypothesize that these effects are involved in the case of AL-09 VL during the amyloid nucleation event. Accordingly, in combination with our observation of possibly reduced CP signal intensity and/or resolution for peaks corresponding to mutation sites, this description suggests that the mutations have critical influence in the early misfolding event but may not be key contributors to maintaining the rigid final fibril structure. Such an explanation would be consistent with the extensive rigidity and order that we report to primarily be localized in the N- and C-terminal regions of AL-09 VL fibrils.

AL-09 VL Fibrils Prepared in Vitro Exhibit Structural Similarities to AL Extracts from Human Tissue

To assess the physiological relevance of the AL-09 VL fibril samples studied herein, we investigated the structural similarities to ex vivo amyloid by comparing the (1H)–13C CP 1D spectrum of AL-09 VL fibrils (U-[15N,13C]-labeled) with that of AL fibril deposits extracted directly from postmortem human spleen tissue (HIG; natural abundance) (Figure A). These ex vivo fibrils share the same κI germline gene product with AL-09 VL, which minimizes the number of amino acid differences between the two samples (Figure B). As illustrated in Figure A, the HIG amyloid is well-structured and exhibits similarities in chemical shift patterns as AL-09 VL fibrils, such as with respect to the downfield Thr- and Ser-13Cβ chemical shifts indicative of predominantly a β-sheet secondary structure. Furthermore, unique glycosaminoglycan (GAG) signals are observed near 75 and 183 ppm in the HIG spectrum (arising from the hydroxyl-containing and carboxylate carbons of GAGs, respectively), consistent with mass spectrometric analysis, which revealed the presence of heparan sulfate proteoglycan (HSPG2) in the spleen tissue extract. Together, these results are in agreement with the known association of GAGs with AL fibrils in vivo.[42,43] Other spectral differences can be accounted for by variations in the protein sequences of AL-09 VL and HIG (Figure B).

Figure 5

Comparison of AL-09 VL fibrils with AL fibrils from the spleen tissue extract (HIG). (A) Overlay of 13C 1D SSNMR spectra (normalized to the 13Cα region) for the 13C natural abundance amyloid extract from postmortem splenic tissue (HIG; green) and AL-09 VL (black) fibrils. (B) Sequence alignment for HIG, AL-09, and κI O18/O8 (germline; gray) VL proteins (residues −1 and 0 are purification artifacts). Highlighted residues indicate mutation sites; those that are nonconservative are underlined.

In addition to providing foundational insight into the structural similarities of the two AL fibrils, these results represent the first demonstration of the feasibility of studying the molecular structures of AL fibrils that are pathologically relevant. Moreover, although the spectral sensitivity of ex vivo fibrils is limited by the lack of 13C and 15N enrichment, future structural studies of tissue fibril deposits will be facilitated by using the fibril material as a seed for in vitro preparations of isotopically labeled fibril samples. This approach has been shown to be successful in previous SSNMR studies for Aβ1–40 fibrils extracted from Alzheimer’s disease-afflicted brain tissue.[44] These studies will be further supported by the ability to immediately assess the physiological relevance and structural similarity of the labeled fibrils using the (1H)–13C CP 1D spectrum obtained for HIG fibrils here as “fingerprint” comparison (Figure A).

As described earlier, one inherent challenge associated with the study of AL disease is the large sequence variation that can occur among patients, consequently limiting the ability to predict and identify early on the occurrence of AL sequences with high fibril-forming propensity. However, interestingly, previous work has identified a murine monoclonal antibody that can be used to specifically bind AL fibrils (as opposed to the native soluble species) irrespective of slight sequence variations, even between germline isotypes, so long as a proline is present at position 8 of the protein sequence.[33,45] These observations indicate that a unique structural motif forms around P8, previously proposed to be a β-turn, that binds specifically to the binding domain of the murine antibody. Given that P8 is 100% conserved among kappa AL sequences such as AL-09 and HIG (Figure B) and that it exhibits structural order in AL-09 VL fibrils, it is possible that the corresponding structure of the N-terminal region represents this unique recognition motif. Although additional structural studies will be necessary to confirm this, agreement of these two structures would suggest that all AL fibrils containing this proline-8 would adopt the same structural motif and possibly other similar aspects of the structural framework of the fibril. Certainly, knowledge of whether a prevailing structural conformation or motif exists between different AL fibrils would provide enormous support for the structure-based design of potential diagnostic or therapeutic strategies, thus further emphasizing the importance of future structural studies on pathologically relevant AL fibrils.

Conclusions

In this study, we investigated the secondary structure and relative dynamics across the AL-09 VL sequence in its fibril form using MAS SSNMR spectroscopy and subsequently assessed the structural similarity of ex vivo AL amyloid obtained from spleen tissue with in vitro-generated AL-09 VL fibrils used in the present study. Overall, our data reveal that extensive regions of the AL-09 VL sequence are highly rigid and β-strand-rich in the fibril state, encompassing the majority of the residues in the protein. Furthermore, our results indicate that the residues near the N- and C-termini are especially well-ordered and exhibit only a single conformation in the fibril form. These measurements of fibril rigidity and relative order were performed at both the global sample and site-specific levels. In light of the success of previous work on the determination of other fibril structures, such as amyloid-β in Alzheimer’s disease and α-synuclein in Parkinson’s disease,[23,25] the results presented here on AL-09 VL fibrils are very promising with respect to the feasibility of determining the corresponding 3D structure. Furthermore, the unique structural features already observed for AL-09 VL fibrils, including the high degree of order for N- and C-terminus residues, relative to other fibril structures suggest that AL-09 exhibits a different and possibly novel type of fold or packing. Moreover, the serine-rich sequence of AL-09 VL relative to the generally hydrophobic-rich sequences of other amyloid fibrils[46,47] may lead to additional complexities in the fibril structure.

On comparing the 1D 13C spectrum of in vitro prepared AL-09 VL fibrils with that of ex vivo AL fibrils obtained from human spleen tissue, we found that both fibrils exhibit a high degree of rigidity and β-strand content. Importantly, these data also set the framework on which additional pathologically relevant AL samples may be studied, including the in vitro preparation of isotopically labeled AL fibrils by the introduction of a fibril seed from the tissue extract to propagate the relevant structure. Indeed, given the recent identification of AL fibrils as the species toxic to cardiomyocytes in vitro, causing cell growth arrest,[5] there is tremendous value in the delineation of the molecular fibril structure for benefiting the pursuit of AL therapies that are directed at the fibril species. Accordingly, we envision that these structural insights into AL fibrils will inform the ongoing search for effective treatments and biomarkers of AL disease, including small-molecule inhibitors of fibril formation as well as monoclonal antibodies specific for insoluble AL fibrils.[4,8,9,33,48]

Experimental Section

Protein Expression and Purification

The uniformly [13C, 15N] labeled AL-09 VL protein (an AL VL derived from a cardiac AL patient)[2] was expressed and purified utilizing a modification of earlier methods[4,49] to accommodate minimal media with 13C and 15N isotopes, as described.[14]Escherichia coli BL21 (DE3) Gold competent cells (Stratagene, La Jolla, CA) were used to express the AL-09 VL plasmid (pET12a) in 2xYT media (Sigma-Aldrich, Saint Louis, MO) and grown until an A600 nm of 0.6. The culture was subsequently transferred into minimal M9 media[50] (with labeled [13C]-glucose and [15N]-ammonium chloride; 3.0 and 1.0 g/L media, respectively) at a 4-fold higher cell concentration and incubated for 1 h at 37 °C before induction with isopropyl β-d-1-thiogalactopyranoside (0.8 mM). The expression was allowed to persist overnight at 25 °C and 100 rpm prior to harvesting. Protein purification was performed using osmotic shock (as described),[14] followed by injection onto a HiLoad 16/60 Superdex 75 size-exclusion column in a 10 mM Tris–HCl, pH 7.4, buffer using an AKTA fast protein liquid chromatography system (GE Healthcare, Piscataway, NJ).

AL-09 VL purity and concentration were assessed by sodium dodecyl sulfate polyacrylamide gel electrophoresis and UV absorption (extinction coefficient, ε = 13 610 cm–1 M–1), respectively. Far UV circular dichroism spectroscopy was used to verify native protein folding prior to the fibril formation reaction, acquiring from 200 to 260 nm at 4 °C with Jasco spectropolarimeter 810 (JASCO, Inc., Easton, MD) and a 0.2 cm path-length quartz cuvette. Percent [13C, 15N]-isotope incorporation in purified AL-09 VL was determined by matrix-assisted laser desorption ionization–time of flight to be approximately 99%.

In Vitro Fibril Formation

Uniformly [13C, 15N]-labeled AL-09 VL fibril samples were prepared from purified protein monomers as described.[14,51] The AL-09 VL monomer was separated from any preformed aggregates by filtering the purified product through a 0.45 μm membrane and subjecting the filtrate to ultracentrifugation (90 000 rpm, 3.3 h) at 4 °C using an NVT-90 rotor and Optima L-100 XP centrifuge (Beckman Coulter, Brea, CA). Fibril-forming samples were subsequently prepared on ice from the monomer-containing supernatant at 20 μM in the acetate borate citrate buffer (10 mM each, pH 2.0), containing 150 mM NaCl and 0.02% NaN3 in 1.5 mL low-binding microcentrifuge tubes (1.0 mL total volume). An additional sample was prepared for each replicate under the same conditions but with 10 μM thioflavin T (ThT) added for monitoring fibrillation progress. Fibril formation reactions were performed in triplicate in sealed (Nunc, Roskilde, Denmark), black 96-well polystyrene plates (Greiner, Monroe, NC) at 37 °C and 300 rpm (in a New Brunswick Scientific Innova40 incubator shaker), using 260 μL aliquots per well until fibril formation was complete, based on the change in the enhanced fluorescence intensity of ThT. The presence of fibrils was subsequently confirmed by transmission electron microscopy, according to previous protocols.[14] The average time for the AL-09 VL fibril formation was approximately 2 days.

Preparation of Fibrils for SSNMR Studies

As previously described,[14] completed fibril reaction mixtures were subjected to ultracentrifugation in polypropylene microfuge tubes using a table-top ultracentrifuge with a TLA-100.3 (Beckman Coulter) rotor and microfuge tube adapters for 1 h at 4 °C and 55 000 rpm (∼100 000g). The pelleted fibril material was separated from the supernatant and washed twice with deionized water using a hand homogenizer and repeated ultracentrifugation. Washed fibril pellets were dried completely under nitrogen gas until the sample mass remained unchanged. Dried AL-09 VL fibril samples were packed into either a 3.2 mm (outer diameter) standard-wall or thin-wall SSNMR MAS rotor (Agilent Technologies, Santa Clara, CA) and hydrated to a level of 50% water by mass. Hydration of packed samples was retained using Kel-F and rubber spacers on both ends of the sample in the rotor.

AL Amyloid Extraction

Patient HIG was a male Caucasian diagnosed with multiple myeloma and subsequently hepatosplenic ALκ amyloidosis. The AL component comprising the amyloid fibrils was shown by aa sequencing to be a κI protein expressed by the O18–O8 (IGKV-1-33) germline gene with a glycosylation site introduced at N30–N31–S32.

Water-soluble amyloid extracts were prepared following prior protocols with some modifications.[45,52] In short, 30–40 g of freshly frozen (−80 °C) spleen obtained postmortem from patient HIG was homogenized in about 300 mL prechilled saline solution (Virtis-Tempest; Virtis, Gardiner, NY). The resulting homogenate was centrifuged for 30 min at 6 °C and 15 000g, and the residual saline-soluble material was removed by repeated cycles of homogenization and washing until the optical density (at 280 nm) of the supernatant was less than 0.10. The pellet was then subjected to repeated rounds of homogenization, washing (cold, deionized water), and centrifugation; collecting and lyophilizing the amyloid-containing supernatant after each successive round. The amount of protein recovered represented ∼25% the weight of the starting material. Approximately 10 mg of the lyophilized material was rehydrated to ∼50% water by mass and packed into a 3.2 mm standard-wall SSNMR rotor (Agilent Technologies).

SSNMR Spectroscopy

MAS SSNMR data of [13C, 15N]-labeled AL-09 VL were collected on the following spectrometers: a 600 MHz InfinityPlus wide-bore or narrow-bore spectrometer equipped with a 3.2 mm Balun 1H–13C–15N or T3 1H–13C–31P probe, respectively; a 500 MHz VNMRS (Varian, Walnut Creek, CA) spectrometer equipped with a 3.2 mm Balun 1H–13C–15N probe; and a 750 MHz VNMRS spectrometer equipped with a 3.2 mm Balun or BioMAS 1H–13C–15N probe (Varian, Fort Collins, CO). Multidimensional 15N and 13C correlation experiments were performed using MAS rates of 11.111 kHz (500 MHz spectrometer), 13.333 kHz (600 MHz), and 12.500 kHz (750 MHz) and variable-temperature (VT) stack air flow set points of −10, 0, and 10 °C (for sample temperatures of −5, 4, and 12 °C ± 3 °C, respectively, as determined by calibration with ethylene glycol).[53] A SPINAL-64 1H-decoupling power of 65–80 kHz was applied during evolution and acquisition periods.[54] One-bond 13C–13C correlations were achieved using either DARR (τmix = 25–90 ms)[55] or SPC-7 (τmix = 0.9–1.8 ms)[32] for magnetization transfer, and one-bond 15N–13C or 13C–15N correlations were accomplished using SPECIFIC-CP.[56] Additional experimental details for multidimensional data acquisition (e.g., digitization settings and specific radio-frequency powers) were summarized previously.[14]

Temperature-dependent 15N–13Cα (NCA) 2D experiments were performed at −30, −10, and 20 °C (VT set points, corresponding to sample temperatures of −23, −5, and 21 ± 3 °C, respectively), using a MAS rate of 13.333 kHz (600 MHz) and SPINAL-64 1H-decoupling power[13] of 72 kHz. Short CP contact times of 0.1 and 2.0 ms were used for 1H–15N and 15N–13Cα transfers, respectively, to promote only intraresidue magnetization transfer (relative to other 1H–15N–13C correlation experiments, which used contact times of 1.5 and 5.0 ms, respectively). T-MREV 1H–13C dipolar recoupling experiments[22] were performed at the same three temperatures (VT) but using a MAS rate of 7.576 kHz (600 MHz) and two-pulse phase modulation 1H-decoupling power18 of 72 kHz. Scaling factors for 1H–13C dipolar couplings of AL-09 VL fibrils were normalized to the 1H–13Cα scaling factor measured for crystalline [13C, 15N]-labeled N-acetyl-valine. R48318-symmetry 1H–13C–13C–1H 3D dipolar recoupling experiments were performed using a MAS rate of 13.333 kHz (600 MHz) and VT set point of 20 °C, with a SPINAL-64 1H-decoupling power of 74 kHz. MAS SSNMR data for the spleen tissue AL extract sample (HIG) were collected on the 600 MHz narrow-bore spectrometer using a MAS rate of 10.000 kHz, a VT set point of 10 °C, and a SPINAL-64 1H-decoupling power of 65 kHz. The (1H)–13C CP 1D spectrum represents a total of 61.5 h of data collection time (108 544 scans).

SSNMR Data Processing and Analysis

SSNMR data were processed using NMRPipe.[57] Multidimensional 15N and/or 13C correlation spectra were analyzed using Sparky,[58] and chemical shift-based predictions on protein secondary structure were performed using TALOS-N[15] as well as by comparing with the random coil chemical shift values presented by Zhang et al.[16] T-MREV and R48318-symmetry 1H–13C dipolar coupling dephasing curves were analyzed using an in-house fitting program (written in FORTRAN).[22] The zero-frequency intensity present in R-symmetry spectra was excluded from the fitting simulations.[59]