A specific nanobody prevents amyloidogenesis of D76N β2-microglobulin in vitro and modifies its tissue distribution in vivo

Systemic amyloidosis is caused by misfolding and aggregation of globular proteins in vivo for which effective treatments are urgently needed. Inhibition of protein self-aggregation represents an attractive therapeutic strategy. Studies on the amyloidogenic variant of β2-microglobulin, D76N, causing hereditary systemic amyloidosis, have become particularly relevant since fibrils are formed in vitro in physiologically relevant conditions. Here we compare the potency of two previously described inhibitors of wild type β2-microglobulin fibrillogenesis, doxycycline and single domain antibodies (nanobodies). The β2-microglobulin -binding nanobody, Nb24, more potently inhibits D76N β2-microglobulin fibrillogenesis than doxycycline with complete abrogation of fibril formation. In β2-microglobulin knock out mice, the D76N β2-microglobulin/ Nb24 pre-formed complex, is cleared from the circulation at the same rate as the uncomplexed protein; however, the analysis of tissue distribution reveals that the interaction with the antibody reduces the concentration of the variant protein in the heart but does not modify the tissue distribution of wild type β2-microglobulin. These findings strongly support the potential therapeutic use of this antibody in the treatment of systemic amyloidosis.

The mechanism of amyloid conversion of wild type β 2 m has been very extensively studied in the last two decades, however much of this work was performed under non-physiological conditions making it difficult to relate the findings to the pathological processes which occur in vivo. The possibility to reproduce in vitro the fibrillogenesis of β 2 m was exploited toward the identification and characterization of putative inhibitors suitable for drug development. To the best of our knowledge, characterization of inhibitors of β 2 m aggregation regarded only the full-length wild type β 2 m 3,4 and its truncated form lacking the first six residues, Δ N6β 2 m, under the specific conditions necessary for fibrillogenesis in vitro. Indeed, full-length wild type β 2 m cannot be converted into amyloid fibrils in physiological conditions and therefore we cannot be completely confident that the effect in vitro may be reproduced in vivo. Methods compatible with the physiological environment are highly desirable for drug discovery and when we reported the first observed genetic variant of β 2 m 2 , we also found that the mutation confers to the protein a very high propensity to make amyloid fibrils in a physiologically relevant buffer 5 . Extensive investigation of the mechanism of fibrillogenesis of this genetic variant has revealed that biomechanical forces, compatible with those present in vivo 6 can drive the amyloid conversion of this globular protein. We have therefore tested the efficacy of previously characterized inhibitors of the fibrillary conversion of wild type β 2 m 3,7 on the fibrillogenesis of the natural D76N β 2 m variant causing hereditary systemic amyloidosis and found that complete abrogation of amyloid conversion can be achieved only by a specific monoclonal nanobody raised against the full-length wild type β 2 m 7,8 .

Results
Inhibition of β 2 m fibrillogenesis. Fibrillogenesis of D76N β 2 m was carried out in physiologically relevant conditions and fluid agitation 5 in the presence and in the absence of ligands previously shown to inhibit wild type β 2 fibrillogeneisis 3,7,8 . In particular, we tested the effect of doxycycline 3 and nanobodies 7,8 which were shown to be good inhibitors of wild type β 2 m fibril formation using procedures based either on the addition of 20% trifluoroethanol (TFE) 3 or on low pH (pH 5) 7 .
First of all, three nanobodies, Nb23, Nb24 and Nb30, previously selected for their capacity to inhibit the fibrillar conversion of the full-length wild type, truncated Δ N6 β 2 m and the non-natural variant P32G β 2 m 8 , were screened to determine whether they retained inhibitory activity against D76N variant β 2 m fibril formation. After 72 h of aggregation in physiological relevant conditions, Nb24 was found to be the only antibody able to inhibit the variant fibrillogenesis monitored with the thioflavin T (ThT) assay 9 (Supplementary Fig. S1). This result prompted us to compare the effect of Nb24 and doxycycline on the fibrillogenesis of D76N β 2 m using a combination of ThT assay and quantification of the soluble fraction. ThT emission fluorescence decreased with increasing concentrations of doxycycline (Fig. 1a) showing that the drug inhibited amyloid formation by D76N β 2 m but did not completely abrogate its aggregation even at 300 μ M, the highest concentration used (Fig. 1a). In contrast, 70 μ M Nb24 inhibited amyloid formation completely (Fig. 1b). Assessment of soluble protein remaining after the end of the reaction (72 h) with either doxycycline or Nb24 was carried out using centrifugation followed by SDS-PAGE of the supernatants. Quantitative analysis of the SDS-PAGE bands corresponding to the monomeric β 2 m confirmed that Nb24 was a more potent inhibitor of D76N β 2 m fibrillogenesis (Fig. 1c,d). Negative stain transmission electron microscopy (TEM) of the pellet, harvested after centrifugation, showed that short fibrils could be still observed in the presence of 300 μ M doxycycline (Fig. 1e), the highest experimental concentration used whereas fibrils were not present in the presence of two fold molar excess (80 μ M) of antibody (Fig. 1f) compared to the classical fibrillar material formed after incubation of D76N β 2 m alone (Fig. 1g).
Characterization of the interaction of Nb24 and D76N β 2 m. Since D76N β 2 m displays the fastest kinetics of aggregation so far described for natural β 2 m species and Nb24 is the only inhibitor able to inhibit the formation of fibrils, we have further characterized the binding properties of Nb24 with this variant using Biacore surface plasmon resonance, native mass spectrometry and NMR.
Biacore. Analysis of the sensorgrams of the interaction between increasing concentrations of Nb24 and D76N β 2 m ( Supplementary Fig. S2) immobilized on the sensor chips allowed us to determine both (mean ± SD) k on (1.04 ± 0.22 × 10 5 M −1 s −1 ) and k off (9 ± 0.001 × 10 −3 s −1 ) values resulting in a K D of 87 ± 0.33 × 10 −9 M. The affinity is therefore slightly lower than that measured with wild type β 2 m, K D = 58 × 10 −9 M 7 , and is consistent with the mutation D76N located within the epitope recognized by Nb24 7 .
Native mass spectrometry. The dynamics of β 2 m aggregation has been previously investigated by mass spectrometry 10 . However, due to the very slow kinetics of fibrillogenesis under physiologically relevant conditions in vitro, most of the studies have been performed in acidic conditions where the nucleation process is reasonably fast 11 . Since these conditions are not compatible with the functions of interactors designed for the inhibition of protein aggregation in physiological environment, we have monitored the incubation of both wild type and D76N variant β 2 m at pH 7.4 and 37 °C under stirring conditions. Untreated D76N β 2 m generated ions at m/z 2372.9 (5+ ), 1977.6 (6+ ) and 1695.2 (7+ ) corresponding to a MW of 11,859 ± 0.5 Da. These ions decreased in intensity over time and completely disappeared after 7 h, reflecting precipitation of the monomer (Fig. 2a). Although turbidity caused by fibrillar aggregates was macroscopically observed, no oligomers were visible by native MS (Fig. 2a), probably because the concentration of these transient species was too small compared to monomers and fibrils. As expected, the wild type did not form fibrils under those conditions; the sample remained clear after the overnight incubation and no decay was observed by native MS with ions at m/z 2372.8 (5+ ), 1977.5 (6+ ) and 1694.9 (7+ ) (MW 11,858 ± 0.9 Da) (Fig. 2b). To monitor the specificity of Nb24, both β 2 m isoforms were pre-incubated for 1 h with Nb24, or an unrelated nanobody, Nb108 7 which was unable to inhibit the D76N variant fibrillogenesis ( Supplementary Fig. S3), and then stirred under the same conditions. Scientific RepoRts | 7:46711 | DOI: 10.1038/srep46711 The D76N variant formed a heterodimer with Nb24 (MW 26827.7 ± 2.9 Da) generating ions at m/z 2981.9 (9+ ), 2683.8 (10+ ) and 2439.9 (11+ ) and the complex remained soluble with no change in the relative ion intensities (Fig. 2c). The wild type protein also formed a stable complex with Nb24 (MW 26,826.3 ± 0.6 Da) generating ions at m/z 2981.8 (9+ ), 2683.6 (10+ ) and 2439.7 (11+ ) (Fig. 2d). When incubated with the unrelated Nb108 neither D76N nor the wild type interacted with the antibody as shown by the lack of ions related to the complexes. By following the intensity of the β 2 m monomers (5+ to 7+ charge states) incubated with Nb108, we could clearly observe loss of the amyloidogenic variant from solution after 7 h (Fig. 2e) whereas the wild type protein remained soluble, as expected under those conditions, with no change in the relative ion intensities (Fig. 2f).  exchange regime should lead, at equilibrium, to the observation of separate resonance sets for the free and bound 15 N-labeled β 2 m isoforms. No contribution to the spectra signal from unlabeled Nb24 was expected. Figure 3 shows overlays of the backbone and side-chain NH signals of 15 N-1 H HSQC spectra as obtained from wild type and D76N β 2 m solutions without and with the addition of a twofold Nb24 molar excess. The spectra confirmed the presence of stable complexes for both β 2 m isoforms in slow exchange with the free species. The expanded region in Fig. 3 illustrates the typical features observed throughout the titrations for wild type β 2 m, namely peaks exhibiting substantial, minor or essentially no chemical shift change upon addition of Nb24.
These different effects on chemical shifts were accompanied by a general broadening of the signals with a complex pattern reflecting the extent of the inter-and intra-molecular dipolar interactions ( Supplementary Fig. S4). In the absence of a precise structural model, it is very risky to attempt a rational quantitation of the individual signal intensity variations. On the contrary, changes in chemical shift of the signals, due to either average . In both panels the green contours are relative to the spectra with Nb24. All spectra were recorded at 600 MHz ( 1 H frequency), 25 °C in 25 mM phosphate buffer (pH* 7.2). Nb24 forms a stable complex with both β 2 m species and the typical pattern observed along titration is depicted in the zoomed portion. The expanded region in panel (a) reports relevant contour overlay of Nb24 titration, namely at wild type β 2 m/Nb24 ratio of 1:0 (red), 1:0.7 (black), 1:2 (green), to illustrate the typical features of complexation induced perturbation. The single-letter code assignments refer to the 15 N-1 H correlation of the backbone amides, except for W95 that concerns side-chain (indole N ε1 -H ε1 ). For the sake of clarity, labels are omitted for A15 and L23 that overlap Y66 and Q8 peaks, respectively.
Scientific RepoRts | 7:46711 | DOI: 10.1038/srep46711 perturbation from ligand fast exchange or ligand complexation, may support a general, more reliable, qualitative inference for the interaction with Nb24. This approach leads to epitope mapping by chemical shift perturbation or complexation induced shifts 12 . Figure 4 illustrates the individual NH chemical shift deviations Δ δ 's obtained from spectra of β 2 m species in the presence and in the absence of Nb24 and, classified according to the number of standard deviation (σ ) from the average Δ δ value. Hence displacements larger than σ , 2σ , 3σ or more indicate locations of the specific β 2 m species with progressively closer contacts with the interacting nanobody or/and larger deviations from the geometry of the isolated protein 12 .
Therefore the experimental data on the extent of chemical shift deviations for β 2 m and its D76N variant treated with Nb24 allowed us to reconstruct the epitopes involved.
Wild type β 2 m showed significant chemical shift perturbation at residues in the apical loops AB and EF as well as at the C-terminal segment, which together identified a possible conformational epitope (Fig. 4). In addition, a more remarkable perturbation involved several consecutive residues in loop C-D that might define a sequential epitope as another possible surface of contact with Nb24. Furthermore, large Δ δ s were also observed at the end of strand B, in strand C and strand F. These data are in part consistent with the crystallographic study carried out on the complex between Nb24 and P32G variant 8 .
Similarly, D76N β 2 m exhibited δ displacements at apical residues in loops AB and EF, at G strand and C-terminal that were consistent with a conformational epitope on the apex of the molecule. Again, a conspicuous perturbation occurred for C-D loop residues featuring a sequential epitope. Finally, other large displacements from the average deviation were observed at strand F (Fig. 4). The closely related patterns of Δ δ s observed for wild type and D76N β 2 m on binding to Nb24 (Supplementary Fig. S4) suggest a substantial analogy of the interaction mode of the nanobody with both species.
Nb24 rescues partially unfolded D76N β 2 m. Consistently with evidence from SPR and mass spectrometry, the binding efficiency of Nb24 to both β 2 m isoforms was quite remarkable under the conditions used for the current NMR study and had an important effect on the stability of the amyloidogenic variant solutions. Indeed,  15 N-1 H NMR peak frequencies in wild type β 2 m (pale gray) and D76N β 2 m (pale green) upon interaction with Nb24 nanobody. For wild type β 2 m and D76N β 2 m, the average Δ δ ± σ values of 0.08 ± 0.06 and 0.08 ± 0.06 ppm were obtained, respectively. NH locations with Δ δ larger than one standard deviation from the average shift are indicated by spheres, with larger deviations by increasing color darkness. The two classes of chemical shift perturbations here presented indicate surface residues (orange-brown grading), and internal residues (cyan grading). For both β 2 m species, the effects of Nb24 binding are observed at the apical loops AB and EF and the C-terminal segment that represent a conformational epitope moiety. A further perturbation involves several consecutive residues of CD loop identifying a sequential epitope portion. Allosteric conformational effects involve internal residues at the end of strand B and, within strands C and F. (b) Elements of secondary structure in β 2 m and D76N β 2 m where residue locations are shown in yellow, orange and red when the corresponding backbone amide, upon Nb24 binding, exhibits a Δ δ value larger than (Δ δ av + σ ), (Δ δ av + 2σ ) and (Δ δ av + 3σ ), respectively.
we found that the conformational heterogeneity, which often characterizes preparations of D76N β 2 m could be abolished upon addition of Nb24. The onset of one or more additional conformers in freshly prepared or aged solutions of D76N β 2 m has been observed nearly always in our NMR determinations 5 preluding to the shift of the natively folded conformation towards extensively unfolded species that eventually aggregate and precipitate. In our hands, with 0.05-0.3 mM D76N β 2 m solutions in phosphate buffer (20-50 mM) around neutrality (pH range 6.9-7.4), the first step of such a transformation pathway, i.e. the onset of the conformational heterogeneity, may develope over 5-10 days, a time lapse sensibly shortened at temperatures above 25-30 °C leading to the loss of the sample. Figure 5a reports the overlay of two HSQC spectra obtained from D76N β 2 m samples at similar concentrations, respectively with and without traces of conformational heterogeneity. As illustrated, peak doublings were most clearly detectable for the C-terminal segment of the protein and the preceding residues of strand G. In addition, signals changed also in the adjacent N-terminal end of strand F and preceding loop residues, as well as in neighboring residues of strand C' , CD loop and C-terminal end of strand E (Fig. 5), i.e. a critical region that closely matches the epitope region recognized by Nb24. Further involvement of the initial fragment of strand D that in solution was largely lost 13 , appeared just a propagation of the conformational perturbation at the CD loop. By addition of Nb24, peak doublings were totally removed because the conformational equilibrium was quantitatively shifted by the binding species and the two HSQC maps became perfectly overlapping (Fig. 5b). The effect of Nb24 on the thermostability of D76N β 2 m was quite remarkable. HSQC spectra acquired in a range of increasing temperatures, from 25 °C to 57 °C, showed the loss of secondary and tertiary structures in the free monomeric protein compared to the protein in the complex which maintained its folded state (Supplementary Fig. S5).

Stability, clearance and tissue distribution of the complex Nb24/β 2 m. The remarkable efficacy of
Nb24 to protect the D76N variant from its amyloid conversion and its superiority in comparison to doxycycline prompted us to explore the stability in plasma and the clearance and tissue distribution of the β 2 m/Nb24 complex in vivo since these aspects are essential for the therapeutic exploitation for this type of nanobody. To assess whether Nb24 was able to bind D76N β 2 m in plasma, the recombinant protein was incubated in human plasma (50 μ g/ml) and, incubated at 37 °C in the presence or in the absence of twofold molar excess of Nb24. After centrifugation, supernatants were separated on a gel filtration column and fractions analyzed by western blotting following SDS-PAGE electrophoretic analysis. Control β 2 m eluted from the gel filtration at Ve ~ 15.1 ml and was separated from a higher molecular weight species (Ve ~ 13.6 ml) corresponding for the Nb24/β 2 m complex ( Supplementary Fig. S6), showing that the complex had been formed in plasma.
Clearance and tissue distribution of 125 I-D76N β 2 m in vivo was studied in groups of four knock-out mice receiving either the monomeric protein or the equimolar complex 125 I-D76N β 2 m/Nb24. Gel filtration of plasma collected at the end of the study (180 min) from mice receiving the complex showed that the radioactive species eluted had the same elution volume of 125 I-D76Nβ 2 m/Nb24 complex (Supplementary Fig. S7). The clearance of D76N β 2 m and the β 2 m/Nb24 complex were similar (Fig. 6a). When organs were counted at the end of the study (180 min), radioactivity was mostly found in the kidneys, which is the main pathway of clearance of circulating β 2 m, followed by spleen, heart and liver. Radioactivity in the heart of mice receiving 125 I-D76N β 2 m alone was significantly higher than those given the pre-formed complex with Nb24 (Fig. 6b) in which total counts were reduced by approximately 60%. Comparative experiments carried out with 125 I-wild type β 2 m showed that the Nb24 had appeared to slow β 2 m clearance; again, the kidneys were, as expected, the main tissue compartment for 125 I-wild type β 2 m (Fig. 6c,d). Histological examination of Congo red stained sections of the heart from mice receiving the protein alone or in complex with Nb24 did not show any evidence of aggregates as expected in the time frame of the experiment (Supplementary Fig. S8).

Discussion
Despite remarkable progress achieved in the elucidation of the pathogenesis of systemic amyloidosis, the therapy of this very severe disease remains a challenging and unmet medical problem. Almost all the therapeutic strategies target the amyloidogenic protein in the attempt to minimize its expression 14 , or to stabilize its native folded state 15 or even to impede the aggregation by chaperone-like synthetic or by biologic compounds 16 . For certain amyloidogenic proteins such as β 2 m the suppression of synthesis or stabilization by specific ligands is unfeasible and inhibitors of aggregation should be considered. The use of low molecular weight ligands such as doxycycline to slow and prevent wild type β 2 m fibrillogenesis has been demonstrated both in vitro 3 and in vivo 17 . Nanobodies against β 2 m have been shown to be effective in blocking the amyloid transition of both wild type and the truncated β 2 -m isoform in vitro. We have evaluated the effect of one of these nanobodies, Nb24, on the amyloidogenesis of D76N β 2 m, a variant that causes a rare form of systemic amyloidosis which is not associated with hemodialysis, and for which no therapy is currently available 2 .
The affinity constant of D76N β 2 m for Nb24 was measured using surface plasmon reasonance and appears to be only slightly lower than that of the wild type protein 7 . Based on previous models of Nb24 in complex with P32G β 2 m and Δ N6 β 2 m 8 , the epitope contacting surface always involves the region around residue 76, i.e. the location of the variant mutation. Consistently, on complex formation, our NMR data show significant chemical shift changes in the backbone NHs of the same region including residue 76 NH for both β 2 m species (Supplementary Fig. S4). Therefore the lack of the negative charge of Asp76, replaced by Asn, may represent an important element responsible for the slightly reduced affinity of the variant species for the nanobody.
Our data clearly show that a therapeutic exploitation of doxycycline as therapeutic inhibitor of D76N β 2 m fibril formation is very unlikely since it cannot completely inhibit its amyloid transition in vitro under conditions of amyloidogenesis compatible with the biological environment. It is worth noting that even at high concentration of doxycycline, corresponding to the maximal inhibition of ThT signal, typical amyloid fibrils can be still observed. More promising and noticeably better results are achieved by using Nb24. In particular, in the presence of twofold molar excess of nanobody, a complete abrogation of amyloid formation can be achieved thus eliminating even a minimal residual formation of amyloid that is difficult to quantify but that is microscopically detectable at the end of the fibrillogenesis in the presence of the highest concentrations of doxycycline. The remarkable shift in the inhibition of D76N β 2 m fibrillogenesis from 60 to 70 μ M Nb24 does reinforce the importance of minimizing the concentration of free β 2 m required for the formation of amyloid nuclei by maintaining a molar excess of antibody. The potent effect of fibrils in recruiting monomeric β 2 m may depend on the favorable thermodynamic state of fibrils versus both free D76N β 2 m 18 and D76N β 2 m within the immuno-complex with Nb24 ( Supplementary Fig. S9).
Furthermore, our NMR data suggest that Nb24 may prevent the conformational transitions leading to amyloidogenic species, and, furthermore, it may revert early conformational changes of the β 2 m variant occurring along the fibrillogenic pathway.
The complex D76N β 2 m/Nb24 is stable in human and mouse plasma and persists in circulation over the experimental time frame (Supplementary Fig. S7). The plasma clearance of the complex studied in β 2 m-knock-out mice is similar to that of free β 2 m, but tissue distributions reveals a difference in the amount of variant β 2 m that persists in the heart.
A comparative analysis of clearance and tissue distribution of wild type β 2 m in the absence and in the presence of the same antibody (Nb24) suggests that the effect on the cardiac localization of the variant may arise from its particular misfolding. We hypothesize that the variant β 2 m, once exposed to cardiac shear forces, might visit a partially folded state which is more prone to a strong interaction with the hydrophobic component of the extracellular matrix thus slowing down the protein efflux from the heart.
In conclusion Nb24 is more effective than doxycycline in inhibiting the amyloidogenesis of D76N β 2 m variant and the stability acquired by the amyloidogenic protein within the immune-complex reduces its permanence in the heart.
Our latest data suggest that Nb24 strongly prevents the induced cytotoxicity of D76N β 2 m aggregates described on the human SH-SY5Y cell line 19 (Supplementary Fig. S10). In addition, previous studies showed that Nb24 cannot bind β 2 m within the MHC-I 7 thus eliminating another potential adverse effect on the cells. Antibodies presenting the functional properties of Nb24 may be considered as candidate for an immunological therapy of familial amyloidosis caused by the D76N β 2 m variant.

Material and Methods
Protein preparation. Recombinant wild type and D76N β 2 m were expressed in transformed E. coli BL21DE3 strains and purified to homogeneity by sequential gel filtration and anion exchange chromatography 2,20 .
Uniformly 15 N-labeled wild type or D76N β 2 m isoforms were also prepared using Spectra 9 minimal medium for NMR analysis.
Both wild type and D76N β 2 m were labelled with 125 I using N-bromosuccinimide and sodium [ 125 I] iodide (Perkin Elmer, Seer Green, UK) in PBS for 10-15 s and purified on a PD10 desalting column (Bio-Rad, Hemel Hempstead, UK) 21 ; 125 I-β 2 m was prepared at a specific activity of 16.0 MBq/mg. Approximately 1 μ g (0.017 MBq) of 125 I β 2 m was diluted into 100 μ g of native recombinant β 2 m to give a final specific activity of ~0.17 MBq/mg for each mouse for in vivo clearance and tissue localization experiments.
The anti β 2 m nanobody, Nb24, originally selected from a phage display library constructed from a β 2 m immunized-camel, was expressed in E. coli as C-terminal His 6 -tagged protein using a pHEN6 cloning vector. The nanobody was purified to homogeneity by immobilized-metal affinity chromatography and gel filtration 7 . without ligands, were incubated at 37 °C in Costar 96-well black-wall plates sealed with clear sealing film and subjected to 900 rpm double orbital shaking. Bottom fluorescence was recorded at 15-min interval (FLUOstar Omega, BMG LABTECH). Since both doxycycline and Nb24 do not interfere with ThT spectrofluorimetric assay, amyloid formation was monitored by ThT emission 9 at 480 nm after excitation at 445 nm in three or more replicates. The data were normalized to the signal plateau of the protein alone at 72 h. The protein remained soluble after the fibrillogenesis was analyzed by SDS homogenous 15% PAGE. The soluble fractions were separated by centrifugation at 10,600 × g for 20 min for the electrophoretic analysis; bands were quantified with Quantity One software (Biorad) and compared to the band of the protein before aggregation. The pellet harvested were resuspended with a minimal volume of water and analyzed by transmission electron microscopy (TEM) using a CM120 microscope at 80 keV after 2% w/v uranyl acetate staining. Nanobody Nb108, raised and selected against an unrelated antigen, was used at twofold molar excess as negative control in experiments of fibrillogenesis of D76N β 2 m using ThT fluorescence assay as described above.

Inhibition of D76N
Biacore. The experiments were performed at 25°C on a BIAcoreX instrument (GE Healthcare, Piscataway, NJ, USA). D76N β 2 m was covalently immobilized on the dextran matrix sensorchip surface (CM5 chip) using a standard amine coupling protocol (Supplementary Information).
Native mass spectrometry. β 2 m, Nb24 and Nb108 were resuspended in 200 mM ammonium acetate pH 7.4, at 40 μ M concentration. Each β 2 m isoform was then incubated in the absence and in the presence of equimolar Nb24 or N108 solutions (final 2 ml volume), at 37 °C under stirring conditions (1,500 rpm, IKA magnetic stirrer). After different incubation times, 2 μ l aliquots from the samples were loaded onto in-house pulled and coated silica needles 22 . The spectra were acquired on a LCT-TOF (Waters) modified for a better transmission of high m/z species. Parameters were as follows: capillary, sample cone and extraction cone voltages were set to 1.5 kV, 50 V and 5 V, respectively, to maintain non-covalent interactions. The pressure in the source was set to 5.5 mBar in order to increase the collisional cooling of the ions by reducing their internal energy. NMR spectra were collected at 14.0 T, on the Bruker Avance III NMR facility of the Core Technology Platform at New York University Abu Dhabi. The spectrometer, equipped with cryoprobe and z-axis gradient unit, operated at 600.13 and 60.85 MHz to observe 1 H and 15 N, respectively. In addition to 1D control spectra, 2D 15 N-1 H HSQC 23 were recorded over spectral widths of 40 ppm ( 15 N, t 1 ) and 15 ppm ( 1 H, t 2 ), and digitized over 192 and 2,048 points, respectively. For each t 1 dimension point, 16 or 32 scans were accumulated and quadrature in the same dimension was accomplished by gradient-assisted coherence selection (echo-antiecho) 24 . Processing with t 1 linear prediction, apodization and zero-filling prior to Fourier transformation led to 2K × 1K real spectra. Diffusion coefficients were determined by means of 2D 1 H DSTEBPP (Double STimulated Echo BiPolar Pulse) experiments that besides correcting for eddy current artifacts also compensate the induced convection contributions 25 . The z-axis gradient strength was varied linearly from 2 to 98% of its maximum value (~60 G/cm) and matrices of 2,048 by 80 points were collected by accumulating 64 scans per gradient increment. The acquired data were processed using the Bruker software Dynamics Center (version 2.2) to extract the diffusion coefficients. Both curve fitting and inverse Laplace transform 26 routines were employed, but only the results from the latter were retained because of the improved fitting reliability. Water suppression was achieved by appending to the DSTEBPP sequence a pair of WATERGATE 27 elements in the excitation-sculpting mode 28 , as for 1D experiments, or using a flip-back pulse in the HSQC experiments 29 . All measurements were performed at 25 °C. The chemical shift perturbation analysis is given in term of combined chemical shift (δ ) deviation, i.e. Δ δ 's, calculated by cumulating both 1 H and 15 N Δ δ values with proper scaling factors 30 .

Stability of D76N β 2 m/Nb 24 complex in plasma.
Human plasma equilibrated at 37 °C was sequentially filtered with 0.45 and 0.22 μ m filters before the addition of either 50 μ g/ml of D76N β 2 m alone or in complex with twofold molar excess of Nb24. After a 30 min incubation, aliquots of 50 μ l from each sample were separated using a Superdex 75 on an ÄKTA Explorer apparatus (GE Healthcare). The column was equilibrated and eluted in PBS pH 7.4 at 0.3 ml/min. Fractions of 300 μ l were collected and analyzed by SDS 8-18% Excel PAGE (GE Healthcare) and then blotted to PVDF membrane for identification with primary rabbit polyclonal anti β 2 m antibody (4.8 μ g/ ml Dako, Denmark) and secondary goat anti-rabbit IgG peroxidase conjugate (0.05 × 10 −3 μ g/ml. Western blotting was developed with SIGMA FAST 3,3′ -diamino benzidin tablets (Sigma Aldrich).
Clearance and tissue localization in β 2 m knock out mice. Groups of four mice (strain B6.129P2-B2m tm1Unc /J) received intravenously 100 μ g of protein either alone (D76N or wild type β 2 m) or in a 1:1 complex with Nb24 respectively. Each dose contained trace amounts of the corresponding 125 I-labeled β 2 m isoform. Plasma samples were collected at 30, 60 and 180 min for clearance studies. Gel filtration of plasma collected at 180 min was performed to assess the persistence of β 2 m in the complex (Supplementary Information). After 180 min, mice were killed and organs collected. Radioactivity was counted with a Perkin Elmer 2470 Automatic gamma counter. Before total 125 I measurement all the organs were rinsed in PBS, blotted-dried and weighted. Data expressed as cpm/gr of blood or tissue represent mean ± SD of four mice per group. Animal studies were Scientific RepoRts | 7:46711 | DOI: 10.1038/srep46711 ethically reviewed and approved by the UCL Royal Free Campus Ethics and Welfare Committee and the UK Home Office, and complied fully with European Directive 86/609/EEC. Statistical analysis. Data analyses used GraphPad Prism version 5. Applied analyses are indicated in corresponding legends. Differences with P < 0.05 were considered statistically significant.