Effect of Tetracyclines on the Dynamics of Formation and Destructuration of β2-Microglobulin Amyloid Fibrils*♦

The discovery of methods suitable for the conversion in vitro of native proteins into amyloid fibrils has shed light on the molecular basis of amyloidosis and has provided fundamental tools for drug discovery. We have studied the capacity of a small library of tetracycline analogues to modulate the formation or destructuration of β2-microglobulin fibrils. The inhibition of fibrillogenesis of the wild type protein was first established in the presence of 20% trifluoroethanol and confirmed under a more physiologic environment including heparin and collagen. The latter conditions were also used to study the highly amyloidogenic variant, P32G. The NMR analysis showed that doxycycline inhibits β2-microglobulin self-association and stabilizes the native-like species through fast exchange interactions involving specific regions of the protein. Cell viability assays demonstrated that the drug abolishes the natural cytotoxic activity of soluble β2-microglobulin, further strengthening a possible in vivo therapeutic exploitation of this drug. Doxycycline can disassemble preformed fibrils, but the IC50 is 5-fold higher than that necessary for the inhibition of fibrillogenesis. Fibril destructuration is a dynamic and time-dependent process characterized by the early formation of cytotoxic protein aggregates that, in a few hours, convert into non-toxic insoluble material. The efficacy of doxycycline as a drug against dialysis-related amyloidosis would benefit from the ability of the drug to accumulate just in the skeletal system where amyloid is formed. In these tissues, the doxycycline concentration reaches values several folds higher than those resulting in inhibition of amyloidogenesis and amyloid destructuration in vitro.

The discovery of methods suitable for the conversion in vitro of native proteins into amyloid fibrils has shed light on the molecular basis of amyloidosis and has provided fundamental tools for drug discovery. We have studied the capacity of a small library of tetracycline analogues to modulate the formation or destructuration of ␤2-microglobulin fibrils. The inhibition of fibrillogenesis of the wild type protein was first established in the presence of 20% trifluoroethanol and confirmed under a more physiologic environment including heparin and collagen. The latter conditions were also used to study the highly amyloidogenic variant, P32G. The NMR analysis showed that doxycycline inhibits ␤2-microglobulin self-association and stabilizes the native-like species through fast exchange interactions involving specific regions of the protein.
Cell viability assays demonstrated that the drug abolishes the natural cytotoxic activity of soluble ␤2-microglobulin, further strengthening a possible in vivo therapeutic exploitation of this drug. Doxycycline can disassemble preformed fibrils, but the IC 50 is 5-fold higher than that necessary for the inhibition of fibrillogenesis. Fibril destructuration is a dynamic and timedependent process characterized by the early formation of cytotoxic protein aggregates that, in a few hours, convert into non-toxic insoluble material. The efficacy of doxycycline as a drug against dialysis-related amyloidosis would benefit from the ability of the drug to accumulate just in the skeletal system where amyloid is formed. In these tissues, the doxycycline concentration reaches values several folds higher than those resulting in inhibition of amyloidogenesis and amyloid destructuration in vitro.
Amyloidosis associated with long term hemodialysis results from the deposition of full-length ␤2-microglobulin (␤2-m) 2 and its N-terminal truncated species ⌬N6␤2m in target tissues (1,2). Although all peripheral organs (but not the brain) can be potentially affected (3), the muscle-skeletal tissues are the preferential target always involved in this type of amyloidosis. Despite significant progress achieved in the hemodialysis techniques, including the increased biocompatibility and the active removal of circulating ␤2-m, the onset of this amyloidosis can be delayed but not avoided in dialysis-related amyloidosis patients (4). New therapeutic approaches, targeting the process of protein aggregation and promoting fibril solubilization (5), are under investigation for the treatment of different types of amyloid diseases. Up until now, different classes of structurally unrelated compounds have been investigated for their ability to interfere with protein self-aggregation and to weaken the intermolecular interactions that stabilize the fibrillar structure of the aggregates (6). Over 10 years ago, iododoxorubicin was serendipitously discovered as the prototype of a class of compounds able to inhibit protein aggregation (7), but this compound was subsequently abandoned for its toxicity. The resemblance of the polycyclic conjugated structure of tetracyclines with iododoxorubicin prompted a group studying the pathologic aggregation of the prion protein to test the properties of those antibiotics in the propagation of prion-related diseases (8). The successful results thereof stimulated further challenges. So far it has been shown that tetracyclines are able to antagonize the infectivity of the PrP sc aggregates (9), to inhibit in vitro the aggregation of amyloidogenic variants of transthyretin (10), the amyloid-␤ peptide fibrillogenesis (11,12), and amylin fibrillogenesis both in vitro (13) and in vivo (14). The generic anti-aggregation property of tetracyclines has been confirmed in the fibrillogenesis of myoglobin (15), a protein that, although not amyloidogenic in humans, represents a very informative model of the mechanism of conversion of globular proteins into fibrillar assemblies.
Based on these and other results obtained in vitro and in animal models, at least three clinical trials have been undertaken to assess possible clinical benefits of tetracyclines in the treatment of the prion 3 amyloid-␤ peptide (16), and transthyretin 4 amyloidosis. In this study, we sought to evaluate the effect of tetracyclines in modulating, in vitro, the formation and the stability of ␤2-m fibrils. It is worth noting that tetracyclines selectively accumulate in the skeletal system (17), where they maintain biological properties such as metalloprotease inhibition (18). Therefore, the local concentration of bioactive tetracyclines, at the site of amyloid formation, should largely exceed that needed to interfere with stability and growth of the ␤2-m amyloid deposits that accumulate preferentially onto bone joints and cartilage. We show here that doxycycline and a group of analogues inhibit amyloid aggregation and induce amyloid fibril disassembly with the initial appearance of toxic oligomers that subsequently reorganize into harmless assemblies.

EXPERIMENTAL PROCEDURES
The 13 tetracycline congeners, reported in supplemental Fig. 1 in the hydrochloride form, were obtained from Sigma-Aldrich. Their purity was always above 95%. Drugs were dissolved in deionized water at 3 mM (stock solutions) and kept at Ϫ20°C for no more than 2 months. Under these conditions, they were stable as determined by reverse phase HPLC. Aliquots of drugs were added to 50 or 70 mM phosphate buffer at pH 7.4 or 6.7, containing 100 mM NaCl, in the presence or absence of 20% (v/v) trifluoroethanol (TFE) or in pure water and immediately used to minimize their oxidation.
In Vitro Fibrillogenesis of ␤2-m and Effect of Doxycycline-Expression and purification of recombinant ␤2-m species were carried out as reported previously (19). In addition to wild type protein that makes fibrils under non-physiologic conditions, we have also investigated the P32G variant that can make fibrils in more physiological conditions (20), at neutral pH, and in the absence of any organic solvent. Recombinant wild type ␤2-m (100 M) was incubated at 37°C in 50 mM phosphate buffer, 100 mM NaCl, pH 7.4, containing 20% (v/v) TFE and preformed ␤2-m fibrils seeds at 2.5 g/ml (21). Recombinant ␤2-m presenting the P32G mutation, at a concentration of 40 M, was incubated at 37°C under agitation at 250 rpm in 25 mM sodium phosphate buffer, pH 7.0, in the presence of 100 g/ml heparin and preformed ␤2-m fibrils seeds at a concentration of 2.5 g/ml. Quantification of amyloid fibril formation was performed using thioflavin T (ThT) as described previously (22). ␤2-m fibrillogenesis was carried out in the presence of different concentrations of tetracyclines (50, 100, 200, 300 M) to evaluate their effect on the growth of fibrils. Dose-response fluorescence data were fitted by nonlinear regression using GraphPad Prism 5.03 (GraphPad Software, San Diego, CA), and IC 50 was determined from the best fitting of three experiments using a standard Hill slope doseresponse curve where a and b are the bottom and top plateau, respectively. 30 l of the fibrillogenesis mixture with 300 M different tetracycline analogues were centrifuged after 96 h of incubation at 10,000 ϫ g for 15 min. The supernatant and protein pellets were analyzed by SDS-PAGE under reducing conditions (23). Pellets were resuspended in 3 l of PBS buffer to be analyzed. Quantification of bands within each lane was carried out using the Quantity One software (Bio-Rad), and the percentages of ␤2-m in both supernatant and pellet were determined as compared with control ␤2-m. Destructuration of ␤2-m fibrils by doxycycline was evaluated by ThT assay and by electron microscopy by incubating ex vivo and synthetic ␤2-m fibrils in the presence of 300 M doxycycline for 12 days.
NMR Experiments-The interaction between ␤2-m and doxycycline was analyzed by NMR spectroscopy under three different solvent conditions: (a) 70 mM sodium phosphate buffer, 100 mM NaCl, pH 6.7; (b) 70 mM sodium phosphate buffer, 100 mM NaCl, 18% (v/v) TFE, pH 6.7; (c) pure water, pH 5.5-6.6. A few microliters of 100 mM stock solution of doxycycline in pure water were incrementally added to the 0.125-0.7 mM ␤2-m samples to reach protein:doxycycline molar ratios of 1:0.3, 1:0.6, 1:1, 1:1.5, 1:2, 1:2.6, 1:3, and 1:6. All the NMR samples contained 5% D 2 O for frequency lock purpose. NMR experiments were recorded on a Bruker AVANCE spectrometer operating at 500 MHz ( 1 H) and equipped with x-y-z-gradient accessory. Besides one-dimensional spectra, conventional proton homonuclear two-dimensional total correlation spectroscopy (24) and NOESY (25) were recorded to assign doxycycline. Two-dimensional [ 1 H-15 N] HSQC (26) experiments were acquired with 128 -256 increments collected in t1 with 2,048 data points and 16 -32 scans/FID in t2, over a spectral width of 14 and 33.5 ppm in proton and nitrogen dimension, respectively. Data were processed with Topspin (Bruker Biospin) and analyzed with Sparky. The chemical shift variations of the [ 1 H-15 N] HSQC spectra were calculated according Mulder et al. (27) using the formula with the hydrogen (first term) and nitrogen (second term) ⌬␦ values expressed in ppm. The chemical shift values of ␤2-m are deposited at the Biological Magnetic Resonance Bank (BMRB) (accession number 15521). Diffusion coefficients were measured by using the convection-compensated twodimensional double stimulated echo-bipolar pulse (DSTE-BPP) sequence (28) to collect matrices of 2,048 (t2) by 80 points (t1). The z axis gradient strength was varied linearly from 2 to 95% of its maximum value (61.1 G/cm). Water suppression was achieved with the addition, in the specific sequence, of a sculpting module (29) or using a flip-back pulse in the HSQC experiments (30).
Inhibition of MTT Reduction-Human SH-SY5Y neuroblastoma cells were obtained from ATCC (Manassas, VA) and cultured in 1:1 Ham's F-10:DMEM medium supplemented with 10% fetal calf serum, 3.0 mM glutamine, 100 g/ml streptomycin, and 100 units/ml penicillin, in a 5.0% CO 2 humidified atmosphere at 37°C. Cell viability was assessed by the MTT reduction inhibition assay. The cells were plated on 96well plates at a density of 6,000 cells/well in 200 l of fresh medium. After 72 h, the cells were exposed for 24 h to ␤2-m previously resolubilized in PBS in the absence or in the presence of different concentrations of drug at 37°C for 1 h or with vehicle for control. The cells were also exposed to preformed fibrils previously treated at 37°C with different concentrations of doxycycline for different lengths of time. At the end of the incubation, the cell culture medium was removed, and the cells were incubated for 2.0 h with 100 l of a 0.5 mg/ml MTT solution in DMEM without phenol red. After incubation, cell lysis solution (100 l/well: 20% SDS, 50% N,N-dimethylformamide) was added to the samples that were incubated overnight at 37°C in a humidified incubator. Blue formazan absorbance was measured at 570 nm using an automatic plate reader (Bio-Rad).
Atomic Force Microscopy-Aggregation of ␤2-m in the presence of fibrillar collagen and heparin was performed as reported previously (31), in the absence and in the presence of 360 M doxycycline. For atomic force microscopy (AFM) inspection, a collagen aliquot was extracted from the protein solution, deposited onto freshly cleaved mica, gently washed with buffer, and dried under a nitrogen stream. ␤2-m fibrils formed in the presence of TFE as detailed above were employed to test the effects of exposure to doxycycline on fibril structure. A 10-l aliquot of ␤2-m fibrils treated with doxycycline was incubated on freshly cleaved mica for 5 min, rinsed with Milli-Q water to remove the materials not adhering to the substrate, and then dried under mild vacuum. Tapping mode AFM images were acquired in air using a Dimension 3100 scanning probe microscope equipped with a "G" scanning head (maximum scan size 100 m) and driven by a Nanoscope IIIa controller and a MultiMode scanning probe microscope equipped with an "E" scanning head (maximum scan size 10 m) and driven by a Nanoscope IV controller (Digital Instruments-Veeco, Santa Barbara, CA). Single beam uncoated silicon cantilevers (type OMCL-AC160TS, Olympus, Tokyo, Japan) were used. The drive frequency was between 320 and 340 kHz, and the scan rate was between 0.5 and 1.5 Hz. Aggregate sizes were measured from the height in cross-section of the topographic AFM images; due to the drying procedure applied, the measured sizes are reduced with respect to fully hydrated conditions.

RESULTS
Titration of the Inhibition of ␤2-m Fibrillogenesis by Different Tetracyclines-We initially tested the concentration-dependent ability of 13 tetracycline analogues (supplemental Fig. 1) to inhibit fibrillogenesis of wild type ␤2-m in vitro. According to the ThT assay proposed by Yamamoto et al. (21), we found that only seven analogues were able to inhibit ␤2-m fibril growth at a drug concentration limit Ͻ300 M. For these seven compounds, we determined the IC 50 activity value that corresponds to the molar concentration of the drug at which the specific fluorescence of the ThT-amyloid fibril complex is reduced by 50% (Fig. 1). The IC 50 values for these seven compounds ranged from 47 Ϯ 5 M for doxycycline and 49 Ϯ 9 M for 4-epi-oxytetracycline (the most effective compounds) to 136 Ϯ 9 M for anhydrochlortetracycline and rolitetracycline. The inhibition of fibrillogenesis by tetracyclines is consistent with the inhibition of protein self-aggregation. The inhibition of aggregation was confirmed by quantifying the soluble fraction of ␤2-m after centrifugation. Electrophoresis of the supernatant and the precipitate (Fig. 2) showed that both doxycycline and 4-epi-oxytetracycline exhibit the highest effect of solubilization of ␤2-m fibrils. Other analogues, such as minocycline or 4-epianhydrotetracycline, left most of the protein in the precipitate. The ability of doxycycline to inhibit ␤2-m fibrillogenesis was confirmed using the P32G variant that can elongate fibril seeds in phosphate buffer at pH 7.0 (20). Fig. 3 reports the dose-response effect of doxycycline on fibril formation by this variant from which we derived an IC 50 of 95 Ϯ 22 M.
Doxycycline Inhibits ␤2-m Fibrillogenesis in the Presence of Collagen-Inhibitors of protein/peptide fibrillogenesis of potential pharmacological interest should be tested in assays mimicking the in vivo environment. For ␤2-m, we have recently set up a method of in vitro fibrillogenesis that generates ␤2-m fibrils under conditions reproducing a physiological milieu in the presence of two ubiquitous constituents of amyloid fibrils: collagen and heparin (31,32). Although the method exhibits some variability in the time of onset of fibril growth, probably because of the complexity of the aggregation conditions, AFM analysis is extremely sensitive and specific for monitoring fibril growth on the surface of collagen fibers. Therefore, we used these conditions to investigate the effect of doxycycline on ␤2-m fibrillization onto collagen. We compared the AFM pattern of a ␤2-m sample aggregated in the absence of doxycycline (Fig. 4a) or in the presence of 360 M doxycycline (Fig. 4b) after 3 months of incubation. In the absence of doxycycline, large bundles of fibrils (black arrows), intertwined or running parallel to each other, were present in close proximity to collagen fibers (white arrows). The latter exhibit a much larger size and appear completely covered by a homogeneous layer of aggregated proteins that masks, at least partially, their typical banding pattern. At variance, in the presence of doxycycline, a completely different pattern was observed (Fig. 4b). Strikingly, in this case, ␤2-m fibrils are totally missing, whereas both collagen fibers and background appear covered by tightly packed globular material (Fig. 4b, inset) that masks, even more effectively, the banding pattern of collagen fibers. Having assessed the inhibitory effect of doxycycline, we tested whether fibril growth was inhibited even at lower (50 -200 M) doxycycline concentrations. After 2 months of aggregation, fibrils were repeatedly observed in the sample incubated with 50 M doxycycline, whereas at 100 and 200 M doxycycline, the fibrils were absent (results not shown).
Effect of Tetracyclines on Preformed ␤2-m Fibrils-The tetracyclines previously found to display the best anti-fibrillogenic activity with soluble ␤2-m were tested for their ability to solubilize preformed ␤2-m fibrils. The specific results obtained with doxycycline are reported in Fig. 5a. Fibrils at a concentration of 100 M (based on monomeric ␤2-m) were exposed to increasing concentrations of doxycycline, and the ThT fluorescence was measured after 12 h of incubation. The IC 50 required for fibril decomposition was much higher (5fold) than that necessary to inhibit fibril growth. The material incubated with doxycycline at the concentration inducing 100% inhibition of ThT fluorescence was further analyzed by electron microscopy that confirmed the complete loss of the fibrillar structure (Fig. 5, b and c). Despite the destructuration of the ordered fibrillar assembly, the material coming from doxycycline-treated fibrils was still insoluble in physiologic buffer, suggesting that the fibrillar protein does not recover a native or native-like soluble structure upon incubation with doxycycline.
Tetracycline-␤2-m Interaction at Molecular Level-To provide information on the drug effects at the atomic level, the interaction between ␤2-m (0.12-0.7 mM) and doxycycline was analyzed by NMR spectroscopy under different solvent conditions, namely aqueous 70 mM phosphate buffer, 100 mM NaCl (pH 6.7), the same saline buffer containing 18% TFE (v/v), and pure water. These environments, although nonphysiological, were selected to monitor the effect of doxycycline under conditions supporting (saline buffer, and much less effectively, pure water) or challenging (TFE) the conformational stability of the protein in solution. As in most biophysical investigations, however, physiologically distant con-ditions enable highlighting trends and drawing conclusions that help in interpreting functional biology. In particular, the saline buffer containing 18% TFE, although preventing excessive protein precipitation, is a fibrillogenic condition in the presence of nucleating fibril seeds (33) that is expected to approach very closely the partial unfolding and spreading over conformational ensembles also including fibril-competent species. Despite being far from the physiological fibrillogenic conditions, a TFE content around 20% in aqueous suspensions of fibril nuclei (seeds) has been repeatedly shown to result in the growth of fibrils that exhibit the same properties as the natural ones (21), making such a condition an established method of in vitro fibrillogenesis of wild type ␤2-m (34). At the doxycycline concentrations that were used for NMR experiments (0.12-3.0 mM), as imposed by the required protein concentrations and ratios, the best solubility was observed in pure water.
NMR in Aqueous Phosphate-In aqueous phosphate and NaCl, with or without TFE, doxycycline addition to protein solution was always accompanied by visible clouding and partial precipitation. When doxycycline was added to the ␤2-m solution in the absence of the organic co-solvent but in the presence of phosphate buffer and NaCl, we found very minor changes in the 1 H- 15   be detected only for the 15 N-1 H correlation of the Phe-22 residue, whereas a number of other backbone amide crosspeaks exhibited distinct and progressive amplitude changes (supplemental Fig. 2). The corresponding residues map to the A-B loop, the B and C strands together with the intervening loop, the CЈ strand and the following CЈ-D loop, the D-E loop, the F strand, and the C-terminal segment. Invariably, under any of the investigated solvent conditions, the doxycycline resonances shifted from the frequency observed without the protein and showed only minor line width increase (supplemental Fig. 3) due to a weak interaction leading to fast exchange. The observed doxycycline shifts were rather limited, and in phosphate buffer, with or without TFE, the most detectable effects occurred for the C4-N,N dimethyl group, the adjacent location (C4-H), and the aromatic hydrogens (data not shown), whereas in pure water, substantial shifts were unequivocally observed on the same side also for H4Ј, H5Ј, H6, and CH3(C6) (supplemental Fig. 3). The weakness of the fast exchange interaction was confirmed by the difficulty to detect direct NOEs between the protein and the doxycycline ligand, except a few saturation transfer difference correlations (35) (data not shown) that arise, by definition, from spin diffusion and are therefore nonspecific.

JOURNAL OF BIOLOGICAL CHEMISTRY 2125
NMR Results in Pure Water-The NMR spectrum of the isolated protein (0.12-0.59 mM) at 25°C in the absence of saline buffers revealed increased line widths, suggesting that the lack of an ionic environment increases the average extent of ␤2-m association. Upon titration with doxycycline at a titrant:protein ratio up to 3:1, several amide resonances of the 15 N-1 H HSQC spectrum showed small shifts and/or intensity changes (supplemental Figs. 4 and 5), namely at the N-terminal tail, the A-B, B-C, D-E, and F-G loops, the E strand, and the C-terminal tail. In addition, the HSQC spectra also revealed a general line sharpening that prompted us to run diffusion-ordered spectroscopy measurements to estimate the translational diffusion coefficient (D) of the protein and hence its effective molecular weight in solution. We calculated an apparent molecular mass of 20.2 kDa for ␤2-m in pure water, i.e. about twice the actual value (11.9 kDa), suggesting the occurrence of fast association equilibria with an average distribution around the dimeric adduct. Upon the addition of doxycycline, the diffusion-ordered spectroscopy spectra showed an increased D value of ␤2-m (Fig. 6a), consistent with a decrease of the protein apparent molecular mass to 12 kDa at 1:3 stoichiometric ratio with respect to doxycycline. Again, the whole NMR experimental pattern points to a fast exchange interaction between the tetracycline and the folded ␤2-m molecule. This interaction effectively stabilizes the protein and shifts its dynamic association equilibria in aqueous solvent toward the monomeric native state. The analysis of the chemical shift and intensity changes throughout doxycycline titration experiments (with and without saline buffer, supplemental Figs. 2 and 5), along with simultaneous diffu-  . a, two-dimensional diffusion-ordered spectroscopy map for ␤2-m in pure water in the presence of doxycycline at 1:0 molar ratio (black), at 1:1 molar ratio (blue), and at 1:3 molar ratio (red). The diffusion coefficients were estimated in the limit of a globular isotropic assumption for aqueous ␤2-m, using the trimeric EMILIN1 gC1q D value as a calibrator for a 51.2-kDa isotropic globule (41). b, ␤2-m interaction sites with doxycycline in pure water. These locations were obtained from the 15 N-1 H HSQC chemical shift (⌬␦) and intensity change analysis upon doxycycline titration. The compensated ⌬␦ values were calculated as described under "Experimental Procedures" after Mulder et al. (27). The color code is: yellow, the residues with ⌬␦ Յ ⌬␦ ϩ , where ⌬␦ is the mean chemical shift change and is the chemical shift difference standard deviation; orange, residues with ⌬␦ ϩ Ͻ ⌬␦ Յ ⌬␦ ϩ 2; red, residues with ⌬␦ Ͼ ⌬␦ ⌬␦ ϩ 2.

Tetracycline-␤2-Microglobulin Fibril Interactions
sion-ordered spectroscopy controls and similar analysis as a function of the protein concentration in the absence of doxycycline (supplemental Fig. 6), enabled us to identify and distinguish unequivocally the protein sites involved in ligand interaction and in protein-protein association in pure water. In particular, doxycycline binds with the highest affinity to the region comprising the end of strand A, the initial part of strand B, and the central residues of the intervening loop (supplemental Fig. 7). A second doxycycline binding region occurs at the N-terminal stretch (fragments 2-4) and the facing B-C and F-G loops (supplemental Figs. 6b and 7). The protein-protein interaction surface comprises the N-terminal segment and the neighbor B-C and D-E loops (supplemental Fig. 6), in agreement with previous molecular dynamics evidence (36).
NMR Analysis in Water/TFE-The scenario changed drastically when ␤2-m was titrated with doxycycline in the presence of 18% (v/v) TFE that approaches a highly fibrillogenic condition (33). We deliberately avoided reaching 20% (v/v) TFE, a condition that impairs NMR observation; rather, we chose to deal with a native structure destabilization close to the fibrillogenic transition limit (33), while retaining at 25°C a soluble native-like species, as inferred from the chemical shift dispersion pattern very closely related to that observed in the absence of the organic co-solvent (supplemental Fig. 8). In the absence of doxycycline, an increase of the temperature to 37°C resulted in massive protein precipitation, whereas the NMR pattern for most of the protein fraction still in solution was consistent with a largely unfolded species, with a native-like species being visible only at very low contour levels (Fig. 7a). Increasing the temperature to 37°C in the presence of doxycycline (1.96 mM), at a nominal 3:1 ratio with respect to ␤2-m, resulted again in substantial protein precipitation, but in this case, the 1 H-15 N HSQC map revealed the equilibrium of the largely unfolded species and the native-like form at higher contour level (Fig. 7b). Therefore, the net effect of the tetracycline on the thermal destabilization of ␤2-m in 18% TFE is to shift the native/unfolded equilibrium toward the native-like species, although the largely unfolded species still predominates. The result was also confirmed by CD analysis (supplemental Fig. 9). Interestingly, the effect was also observed when the tetracycline was added after the denaturing exposure at 37°C, revealing a surprising reversibility (in 18% TFE) of the doxycycline/␤2-m system (supplemental Fig. 10). Whatever the timing of ligand addition, the NMR parameters of ␤2-m in aqueous TFE, i.e. 1 H-15 N peak chemical shifts and amplitudes, did not show substantial changes, nor did the NMR parameters of the doxycycline itself. Therefore, the observed difference in the distribution pattern of ␤2-m species in aqueous TFE (Fig. 7, a and b) should be attributed to a weak, transient, preferential interaction of the doxycycline with the native-like ␤2-m species. A quantitative assessment using pairs of resolved cross-peaks in the spectra of Fig. 7, a and b, led us to estimate a doxycycline-induced gain in ⌬G unfolding of 0.9 kJ/mol. This very limited stabilization is not surprising in 18% aqueous TFE at 37°C, i.e. a condition of substantial destabilization of the native protein (⌬G unfolding ϳϪ4 kJ/mol).
Doxycycline Modifies the Cytotoxicity of Soluble and Fibrillar ␤2-m-A fundamental issue regarding the mechanism of tissue damage caused by the amyloid deposits consists in defining the cytotoxic role of prefibrillar oligomeric aggregates. We recently reported that soluble ␤2-m is cytotoxic if oligomers are present (3), thus confirming and enriching previous reports on ␤2-m cytotoxicity (37). As some tetracyclines are apparently capable of decomposing fibrils into smaller aggregates, we studied the cytotoxicity of native and fibrillar ␤2-m exposed to doxycycline. Cytotoxicity was quantified by the MTT and lactate dehydrogenase release assays, widely used cell viability tests. We confirmed that the native ␤2-m is cytotoxic for human  (1.96 mM). The spectra were obtained using the same acquisition conditions from two samples prepared using the same ␤2-m mother solution, with a 10.0-l addition of 0.100 M doxycycline for b. The experimental data were treated with the same processing parameters and at the same absolute scaling (TOPSPIN, Bruker). The reported maps were drawn starting at the same lowest contour level and with the same level numbers and spacing. JANUARY 21, 2011 • VOLUME 286 • NUMBER 3 neuroblastoma SH-SY5Y cells when oligomers are not eliminated from the solution, whereas the fibrils are not toxic even at 20 M (monomer concentration), a value exceeding by 4-fold the IC 50 of soluble ␤2-m cytotoxicity (3). Fig. 8a shows that doxycycline, at a concentration equimolar to ␤2-m, abrogated the cytotoxicity of native soluble ␤2-m. On the contrary, and unexpectedly, fibril exposure to doxycycline triggered a cytotoxicity that was absent in untreated fibrils (Fig. 8b). The rescue of fibril cytotoxicity was directly related to the drug concentration. The results reported in Fig. 8b refer to an experiment in which ␤2-m fibrils were incubated for 72 h with doxycycline before their inclusion in the cell culture medium. However, to check whether a longer fibril incubation with doxycycline could modify cytotoxicity, we also analyzed ␤2-m fibrils incubated with the tetracycline for 2, 3, 4, 7, and 15 days. Fig. 8c shows that the cytotoxicity of this material reached a maximal activity after 3 days of incubation and then progressively declined to eventually disappear after 7 days. It is noteworthy that the cells incubated with ␤2-m fibrils not treated with doxycycline showed a surprisingly greater viability, as compared with control cells incubated only with vehicle. This aspect will be further investigated as it might suggest a positive role of the fibrillar net on cell growth. To this end, we notice that the presence of non-toxic fibrils in the cell medium can somehow promote cell adhesion to the support, as is known in the case of the curli amyloid fibrils favoring bacterial attachment to the substrate (38). Consistently, although ThT decreases along a monophasic curve, the cytotoxicity trend reflects a multiphasic process. Such a different dependence of ThT fluorescence and cytotoxicity on the time of fibril incubation with doxycycline ( Fig. 8d) suggests that remodeling of the aggregated protein goes far beyond the modifications detected by the ThT assay that reaches a plateau after 48 h of fibril incubation with the drug. Fig. 9 shows the relation between aggregate toxicity and the evolution of fibril morphology imaged by AFM as a function of the incubation time in the presence of doxycycline. Fig. 9a shows ␤2-m fibrils immediately after drug addition. Exposure to doxycycline gives rise to a partial fibril disaggregation into smaller units, as shown in Fig. 9b, acquired after 24 h of incubation. An extensive release of such oligomeric units took place in the subsequent hours. After 72 h of incubation with doxycycline (Fig. 9c), the accumulation of oligomers reached its maximum level; this result correlated very well with the maximum cytotoxicity found in the same conditions. The size distribution of these oligomers is rather broad, with height values ranging from less than 1 nm to ϳ10 nm. At longer incubation times, oligomers are reassembled into more complex structures. In particular, after 7 days, the sample mainly consisted of ϳ1-nm-high aggregates of variable length (Fig. 9d). Again, this result correlates well with the decrease of cytotoxicity found in the same conditions.

DISCUSSION
The inhibition of the fibrillogenesis of wild type ␤2-m by tetracyclines displays the features of a dose-response reaction, and the effective concentration is graded from 47 and 49 M of the more effective compounds (doxycycline and 4-epi-oxytetracycline, respectively) to 136 M (anhydrochlortetracycline-rolitetracycline) and to Ͼ300 M as measured for minocycline. The poor activity of minocycline was quite surprising because, among the tetracyclines, this drug was the most active against the aggregation of another amyloidogenic protein, PrP sc (9). This finding represents unequivocal evidence that in amyloidogenesis inhibition, a specific structure-function relation does exist between the single amyloidogenic proteins and each tetracycline species. On the basis of the first screening, we decided to focus our investigation on the anti-amyloidogenic effects of doxycycline because a vast pharmacological and toxicological experience already exists for this drug. The interference of doxycycline with ␤2-m aggregation was characterized by NMR spectroscopy that confirms the global persistence of the connectivities of the native protein even when it was exposed to conditions that favor unfolding and can trigger fibrillogenesis by seeding.
The anti-aggregating effect of doxycycline was confirmed by NMR analysis for its ability to rescue soluble native-like ␤2-m from the aggregates grown in ϳ20% TFE and to decrease ␤2-m association in pure water. The stronger ␤2-m self-association in pure water reflects the effect of exposed hydrophobicity; such an effect is attenuated by the presence of salts that assist molecule solvation. The strong activity of doxycycline and 4-epi-oxytetracycline, in comparison with other tetracyclines, highlights the relevance of the C5 oxydryl. Despite the highly dynamic character of the interaction, the contact surfaces in the ␤2-m/tetracycline adduct have been successfully identified by specific NMR observations. Based on the current evidence, it is possible to establish that the N,N-dimethyl group of doxycycline, and probably, to a larger extent, the adjacent locations (C4Ј, C5, C5Ј, and C6), are involved in the transient adduct with the protein, whereas the protein appears to employ preferentially the region at the end of strand A and the initial part of strand B together with a part of the AB loop area. An additional lower affinity site for doxycycline binding appears to overlap the region that ␤2-m transiently uses for dynamic oligomerization, i.e. the N-terminal end with the facing B-C, D-E, and F-G loops. The structural details of this behavior will be discussed in a forthcoming study. Here, however, two aspects of the doxycycline-␤2-m interaction are worth stressing, namely (i) the efficiency of the transient contact between doxycycline and the target protein, reaching a proper turnover rate because of the relative lability of the interaction, and (ii) the preferential doxycycline affinity for the natively folded conformer of ␤2-m that also accounts for the results obtained in the presence of TFE. The ability of doxycycline to inhibit the amyloidogenesis of ␤2-m was validated not only in 20% TFE but also in physiologic-like conditions, by using the variant P32G or in the presence of pro-amyloidogenic cofactors, i.e. collagen and heparin.
In particular, the effect of the drug in conditions mimicking the environment in which wild type ␤2-m makes fibrils in vivo has been qualitatively confirmed in tests carried out in the presence of collagen and heparin, when fibril growth is intimately attached to the collagen surface (31). We are aware that the concentrations of doxycycline required for the abrogation of fibrillogenesis in all the conditions we assayed largely exceed that reached in vivo by the drug in classical therapeutic regimes, i.e. typically 1-5 M. However, tetracyclines are highly concentrated (up to 2,000-fold over the bulk tissue value) in the skeletal system (17), the elective target of this amyloidosis. Therefore, the successful abrogation in vitro of amyloid formation, at the drug concentration we used, supports the possibility that doxycycline can also be effective in vivo. It is worthy of note that upon prolonged exposure to the drug, the structure of ␤2-m fibrils is deeply remodeled, as shown by the loss of the extended polymeric pattern in the electron microscopy images and, consistently, by the remarkable reduction of ThT reactivity, although the protein remains irreversibly insoluble.
Dilemmas regarding the identification of the cytotoxic entity of amyloid proteins apply to dialysis-related amyloidosis similarly to the other systemic and localized amyloidoses. We have recently established that in the cell culture, oligomeric ␤2-m is cytotoxic, thus confirming previous data on the biological activity of soluble ␤2-m that is cytotoxic for certain cells (3) and mitogenic for other strains such as the osteoclasts (39). Because doxycycline deeply interferes with ␤2-m aggregation, we evaluated the effect of the drug on the biolog-

Tetracycline-␤2-Microglobulin Fibril Interactions
ical activity of soluble and fibrillar ␤2-m. At equimolar concentrations, the presence of doxycycline fully abrogates the toxicity of soluble ␤2-m oligomers. This is consistent with the NMR data highlighting the suppression induced by doxycycline of the dynamic ␤2-m association in solution. On the other hand, the effect of the drug on the biological activity of mature ␤2-m fibrils is significantly more complex. The fibrils, per se, are not cytotoxic; however, fibril exposure to doxycycline elicits a transient cytotoxic activity that reaches the maximum effect after 3 days of incubation and vanishes after 7 days of incubation. The inspection by AFM of the evolution of the morphology of ␤2-m fibrils incubated with doxycycline at time intervals corresponding to those at which cytotoxicity was measured allowed us to identify the assemblies responsible for toxicity. In fact, in the toxic specimens, we imaged globular oligomeric aggregates that were substantially absent in the fibrillar samples not exposed to the drug. The same oligomers were also absent in the fibrillar samples incubated with doxycycline for 7 days, i.e. a time lapse over which the presence of the drug results in extensive structural reorganization. As it has been recently demonstrated for another amyloidogenic protein (40), it is likely that cytotoxicity results from the exposure of hydrophobic groups for ␤2-m oligomers as well; such exposure is very likely to drive the rearrangement of oligomers into more stable structures with reduced toxicity.
The dissection of the molecular effects of the interaction between doxycycline and the amyloidogenic protein ␤2-m has therefore confirmed and better clarified the intimate mechanism by which tetracyclines counteract the tendency of amyloidogenic proteins to acquire an ordered fibrillar structure and, in their prefibrillar state, a specific toxic structure. Such information could be rapidly translated into a new therapeutic opportunity for this amyloidosis. In our view, a selected population of patients like those under prolonged hemodialysis at high risk of amyloid deposition or with early amyloid deposits are the natural candidates for this therapy. The onset of dialysis-related amyloidosis can be easily recognized by the occurrence of carpal tunnel syndrome. Therefore, these patients could be treated before entering a stage of the disease characterized by irreversible tissue damage and in which the excessive amount of amyloid could defeat the therapeutic efficacy of the drug. Furthermore, as the deposition of this amyloid takes place in a tissue environment where tetracyclines naturally reach the highest possible concentration, we suggest that the in vivo validation of the doxycycline therapy in dialysis-related amyloidosis could represent an extraordinary proof of principle of the effectiveness of the class of anti-amyloid compounds acting by inhibiting amyloid fibril growth.