Distinguishing Closely Related Amyloid Precursors Using an RNA Aptamer

Background: Altering the co-polymerization of proteins into amyloid fibrils provides an opportunity for manipulating fibril assembly. Results: NMR and kinetic analysis showed that an RNA aptamer distinguishes between two highly similar co-aggregating proteins. Conclusion: RNA aptamers are specific and discriminatory probes able to modulate amyloid formation. Significance: Aptamers can be used as tools to differentiate amyloid precursors that are closely related and alter assembly.

Despite the array of different proteins and peptides with distinct amino acid sequences that are known to be able to assemble into amyloid fibrils in vitro and/or in vivo (1), the precise molecular mechanism(s) by which these different proteins/ peptides self-assemble into amyloid fibrils and how the assembly process results in disease remain unclear (2). Amyloid formation commences with the generation of aggregation-prone monomeric precursors. These species can be unfolded/disordered, partially structured or even native-like (3) and their structural properties, even though potentially similar to their non-amyloidogenic counterparts, dictate the fate of amyloid assembly (4). This is exemplified by the observation that the same amino acid sequence can form conformationally distinct amyloid structures in vitro by varying the temperature, altering the agitation conditions, adding co-solvents, metal ions or other molecules, or even changing the surface properties of the incubation vessel (reviewed in Ref. 5). An extra level of complexity is added by the ability of different protein/peptide precursors to co-polymerize, resulting in new fibril polymorphs with different amyloid architectures, stabilities, and/or different kinetics of assembly than those formed by each protein alone (4,6,7). Indeed, there are multiple examples of amyloidogenic proteins that are able to co-polymerize such as islet amyloid polypeptide and A␤ (6,8), Tau and ␣-synuclein (9), and insulin and transthyretin (10). Although the importance of identifying and characterizing rarely populated amyloidogenic precursors is widely appreciated (3), this remains a significant challenge because of the transient nature and heterogeneity of assembly intermediates (11). The development of reagents able to discriminate aggregation-prone species among a pool of structurally similar molecules is crucial to deciphering the mechanisms of protein assembly into amyloid and to inform the design of therapeutic/diagnostic strategies able to target individual amyloid precursors (12).
Human ␤ 2 -microglobulin (h␤ 2 m) 5 is a small protein that forms amyloid deposits in collagen-rich osteoarticular sites, resulting in the disorder dialysis-related amyloidosis (DRA) (13,14). Despite the propensity of h␤ 2 m to form amyloid fibrils in vivo, conditions that destabilize the native structure of h␤ 2 m such as low pH (15), the presence of SDS (16), or other co-solvents or metal ions (17,18), are required for fibril formation on an experimentally tractable time scale in vitro. Removal of the N-terminal six residues from h␤ 2 m (the sequence IQRTPK), creating the variant ⌬⌵6, disrupts the thermodynamic and kinetic stability of h␤ 2 m and, as a result, ⌬N6 can self-assemble into amyloid fibrils rapidly and spontaneously without the need to add detergents, metal ions, or other reagents (19,20). ⌬N6 retains a native-like structure, displaying a backbone r.m.s.d. of only ϳ1.5 Å compared with h␤ 2 m (19), and contains a nonnative trans X-Pro-32 (see Fig. 1, A and B), which has been shown to be vital for fibril formation (21,22). Isomerization of the X-Pro-32 bond results in structural reorganization of the side chains in the apical region of h␤ 2 m, resulting in a protein with different surface hydrophobicity and electrostatic properties (19). ⌬N6 can also promote the aggregation of h␤ 2 m even when added in trace amounts (19), resulting in co-polymerization of both proteins into hetero-polymeric amyloid fibrils (4). This interaction allows amyloid formation of h␤ 2 m to be investigated in the absence of additives at physiologically relevant pH values (4).
The design of molecules able to bind h␤ 2 m or its amyloidogenic counterpart, ⌬N6, would offer an opportunity to increase understanding of the interaction between these co-assembling monomers and explore the aggregation pathway that leads to their co-polymerization into amyloid fibrils. However, such a task is hindered by the high sequence and structural homology (see Fig. 1A) of the two proteins and their dynamic nature (19). In this study, we used in vitro selection to identify an RNA aptamer able to bind h␤ 2 m preferentially to ⌬N6 and to alter fibril co-assembly. Nucleic acid aptamer selection has been used previously to generate RNA aptamers able to discriminate monomeric PrP SC and recombinant PrP C (23,24) and to bind to A␤ monomers rather than fibrils (25)(26)(27). Oligomers of amyloidogenic proteins have also been used as targets: DNA/RNA aptamers have been raised against oligomers of ␣-synuclein (28) and 〈␤40 (29), respectively.
Previously, we used SELEX to isolate RNA aptamers against fibrillar h␤ 2 m that were counter-selected against the low pH, partially unfolded, h␤ 2 m monomer from which these fibrils were formed (30). Here, we extend this approach using SELEX to isolate 2Јfluoro-modified RNA aptamers against native monomeric h␤ 2 m. The selected aptamer discriminates in its binding to h␤ 2 m or ⌬N6 at pH 6.2, conditions in which both proteins are folded, but only ⌬N6 is able to assemble spontaneously into amyloid fibrils (19). The h␤ 2 m specific aptamer was minimized to a 44-nucleotide-long fragment and its binding interface, affinity, and specificity for h␤ 2 m were determined. The aptamer binds tightly and specifically to the ␤-sheet of h␤ 2 m containing the A, B, E, and D ␤-strands, but only weakly and less specifically to ⌬N6. Addition of the aptamer to a mixture of h␤ 2 m and ⌬N6 under conditions (pH 6.2) that promote co-assembly (4) disfavors the interaction between the two proteins early in assembly, making h␤ 2 m remain soluble for longer. The results reveal the ability of RNA aptamers to discriminate and bind to a specific protein conformer within a complex mixture of structurally similar co-polymerizing species, altering the course of amyloid assembly.

EXPERIMENTAL PROCEDURES
Protein Preparation-h␤ 2 m and ⌬N6 were expressed and purified as described previously (19). For NMR experiments, 15 N-and 13 C-labeled h␤ 2 m and ⌬N6 were prepared as described in Ref. 31.
Biotinylation and Immobilization of h␤ 2 m-Monomeric h␤ 2 m (ϳ1 mg) was biotinylated (EZLink TM Sulfo-NHS-LC-LC-biotin, Pierce Biotechnologies) at pH 7 using a 20-fold molar excess of biotin over the total protein concentration, according to the manufacturer's instructions. The biotinylated monomer was then immobilized on 1-m streptavidin-coated microspheres (Dynabeads TM , Invitrogen) using the manufacturer's instructions.
In Vitro Selection-A Biomek 2000 laboratory automation work station (Beckman Coulter) was used to perform 12 rounds of in vitro selections with an N30 library of 2ЈF-modified pyrimidine RNA, encompassing ϳ10 15 potential sequences, and transcribed using the Y639F/H784A variant of T7 RNA polymerase (32), using minor modifications of the protocols described previously (30). Selections were carried out in 50 mM MES buffer containing 120 mM NaCl, pH 6.2. Negative selections were carried out at each round of SELEX using streptavidin Dynabeads coated with Tris-inactivated linker. Stringency was increased after round 5 by decreasing the number of beads containing monomeric h␤ 2 m by half and increasing the number of washes from 10 to 13. The reverse transcriptase-PCR products were analyzed by native PAGE after each group of five rounds of selection to confirm the isolation of products for the next round of selection. Individual aptamer clones were produced by in vitro transcription using 10 mM final concentrations of each nucleotide triphosphate using 2ЈF-CTP and 2ЈF-UTP for production of modified RNAs. RNA concentrations were determined using the following extinction coefficients: B6, 1026.3 mM Ϫ1 cm Ϫ1 ; B6 minimized (B6min), 553.2 mM Ϫ1 cm Ϫ1 ; B9, 1054.2 mM Ϫ1 cm Ϫ1 .
Synthesis of Minimized B6 -B6min (5Ј-GGG AAU UCU GAG CUA CUC CCU UUU GGG CCC GGC UAU GAU UCC CG-3Ј) was synthesized with and without 2ЈF-modified pyrimidine nucleotides (named 2ЈF B6min and 2ЈOH B6min, respectively) on an ABI 394 RNA synthesizer at a 1 M scale using the protocols described previously (33). The phosphoramidites used for synthesis of 2ЈF-B6min were as follows: N-benzoyl-protected adenosine, N-dimethylformamidinyl-protected guanosine, N-acetyl-protected-2Ј-fluoro deoxycytidine, and 2Ј-fluoro-deoxyuridine. For synthesis of 2Ј-OH B6min Nacetyl-protected-2Ј-fluoro-deoxycytidine and 2Ј-fluoro-deoxyuridine were replaced with N-acetyl-protected-cytidine and uridine phosphoramidites (Link Technologies, Ltd.). Cyanoethyl-(N,NЈ-diisopropyl) and t-butyldimethylsilyl groups were present on the 3Ј-and 2Ј-hydroxyl groups. Treatment with ammonia-saturated methanol at room temperature for 24 h was used to remove protecting groups and to cleave RNA from controlled pore glass resin. Methanol was removed under vacuum and the RNA pellet re-suspended in anhydrous dimethyl sulfoxide. One volume of triethylamine trihydrofluoride was added and incubated at room temperature to remove t-butyldimethylsilyl, the deprotected RNA was precipitated with butan-1-ol and resuspended in diethylpyrocarbonate-treated water (Severn Biotech) before being purified by reverse-phase HPLC at 55°C (34). RNA fractions were collected, lyophilized, and desalted into 18.2 megohm H 2 O. The RNA was analyzed on a 10% (w/v) denaturing polyacrylamide urea gel stained with ethidium bromide. The RNA was synthesized using N-dimethylformamidinyl-protected guanosine controlled pore glass to avoid incorporation of a pyrimidine with a ribose sugar at the 3Ј-end. This additional guanosine has no effect on the secondary structure of 2ЈF-B6min or 2Ј-OH B6min as predicted by Mfold (35).
Surface Plasmon Resonance-A BIAcore3000 instrument was used with a streptavidin-coated gold sensorchip (BIAcore SA chip). A flow-rate of 10 l min Ϫ1 was used with a running buffer of 50 mM MES, 120 mM NaCl, pH 6.2. 50 l of 50 g ml Ϫ1 of biotinylated monomer was injected over separate flow cells so that ϳ200 response units of protein was immobilized. RNAs were dialyzed into running buffer before injection across the surface to minimize bulk refractive index effects. Flow cells were regenerated using a 20-l wash of 5 M NaCl. All sensorgrams were corrected by subtracting the signals of an equivalent injection across an underivatized flow cell. Data were analyzed using the manufacturer's software (BIAevaluation).
Intrinsic Fluorescence Quenching-The fluorescence of tryptophan residues in 1 M h␤ 2 m or ⌬N6 was excited at 290 nm, and fluorescence emission was measured between 300 and 390 nm in the presence of increasing concentrations of 2ЈF B6min or 2ЈOH B6min in 50 mM MES buffer containing 120 mM NaCl, pH 6.2, at 25°C. Due to the large extinction coefficient of the RNA aptamer at 260 nm (553.2 mM Ϫ1 cm Ϫ1 ), the absorbance of the h␤ 2 m/aptamer solution at 290 nm was measured after each addition of aptamer to ensure that the absorbance of the solution was below 0.05 at 290 nm so that inner filter effects do not contribute to the data (36). Fluorescence emission was measured using a Photon Technology International QM-1 spectrofluorimeter using 10-nm slit-widths. The data for binding of 2ЈF B6min to h␤ 2 m were normalized to a value of 0 in the absence of aptamer and a fluorescence signal of 1 obtained upon saturation. The data were then fitted to the following logistic equation to extract the half maximal effective concentration (EC 50 ) using in-house scripts, where max and min represent the maximum and minimum fluorescence signals, respectively, Hill is the Hill coefficient, f(x) is the fluorescence units, and x is the concentration of the aptamer in nM. For 2ЈF B6min to ⌬⌵6 and 2ЈOH B6min to h␤ 2 m, no change in fluorescence was observed over the concentration range studied.
NMR Spectroscopy-Samples of 13 C-15 N-labeled protein (60 M) in 50 mM MES buffer containing 120 mM NaCl, pH 6.2, or 2ЈOH B6min was added into the protein solution from a concentrated stock (typically 200 M). Working at a concentration of 60 M necessitated the use of a sensitivity optimized strategy for obtaining assignments. This was achieved using a reduced dimensionality approach based on Hadamard encoding (37). Sequential assignments were obtained from analysis of Hadamard encoded two-dimensional H(N-H 2 )CA and H(N-H 2 )(CO)CA experiments where a two-step Hadamard matrix is introduced on 15 N to subdivide the peaks into two subspectra where most signals can be addressed from their 1 H shift alone and the dimensionality can be reduced to 2 to maximize sensitivity. Spectra were recorded at 25°C on a Varian Inova 750 MHz spectrometer equipped with a cryogenic probe and were processed using NMRPipe and analyzed using Collaborative Computational Project for NMR analysis (38). To calculate the intensity profiles shown in Fig. 7, peak intensities were normalized to the number of scans and the protein concentration used for each experiment. Intensity profiles were calculated as the ratio of the normalized peak intensity of each resonance in the apo spectrum (I 0 ) versus the normalized intensity at the same position but in the aptamer-bound spectrum (I). Therefore, the loss of native signal plotted in Fig. 7 does not require full assignment of the aptamer-bound spectrum.
Assembly of Amyloid Fibrils-40 M h␤ 2 m and 40 M ⌬N6 in the presence or absence of two molar equivalents of 2ЈF B6min were co-incubated in 50 mM MES buffer containing 120 mM NaCl, pH 6.2, 0.02% (w/w) sodium azide at 600 rpm, 37°C in a Thriller Thermoshaker incubator (Peqlab). Each sample (100 l) was incubated in 0.5-ml plastic Eppendorf tubes. Aliquots of 8 l were removed at different time points during incubation and immediately centrifuged at 14,000 ϫ g for 20 min. The supernatant was separated from the pellet, and both supernatant and pellet were frozen at Ϫ20°C for subsequent analysis by SDS-PAGE.
SDS-PAGE-The effect of 2ЈF B6min on fibril formation was monitored using 15% polyacrylamide Tris-Tricine gels. Samples of the supernatant and pellet were thawed, and the pellet was resuspended in 8 l of 50 mM MES buffer containing 120 mM NaCl, pH 6.2. Both the supernatant and resuspended pellet were added 1:1 to loading buffer (50 mM Tris-HCl, pH 6.8, 100 mM DTT, 2% (w/v) SDS, 0.1% (w/v) bromphenol blue, and 10% (v/v) glycerol) and boiled for 5 min before loading 15 l into the gel. Gels were stained with Coomassie Instant Blue (Expedeon) and imaged by SnapGene software (Syngene).
Electron Microscopy-At the end of fibril assembly, 10 l of sample were applied to a carbon-coated grid. The grid was then carefully dried with filter paper before it was negatively stained by the addition of 10 l of 4% (w/v) uranyl acetate as described in Ref. 39. Micrographs were recorded on a Philips CM10 or a JEOL JEM-1400 electron microscope.

RESULTS
Selection of h␤ 2 m-specific 2ЈF-RNA Aptamers-Co-incubation of ⌬N6 and h␤ 2 m results in the two proteins polymerizing into hetero-polymeric amyloid-like fibrils that are morphologically and thermodynamically distinct compared with fibrils formed by ⌬N6 or h␤ 2 m alone (4). To control the co-assembly of these proteins, we attempted to select RNA aptamers capable of discriminating between natively folded h␤ 2 m and ⌬N6 at pH 6.2 (Fig. 1A). H␤ 2 m was biotinylated (predominantly at the N terminus and on Lys-7 and/or Lys-92) and immobilized as a target on streptavidin-coated magnetic beads, as described previously (30). The initial SELEX protocol used an N30 2ЈF-pyrimidine substituted RNA library to create aptamers resistant to nucleases (25,32) and included counter-selection against ⌬N6 monomers immobilized as for h␤ 2 m, as well as long straight and worm-like amyloid fibrils formed from h␤ 2 m at acidic pH (40). This protocol resulted in the removal of most of the aptamers from the selected pool, consistent with the different protein conformers containing many epitopes in common. We therefore abandoned counter-selections, except against biotin linker-blocked streptavidin beads alone. In addition, in the final round of SELEX, aptamers bound to bead-immobilized h␤ 2 m were competed off the beads using non-biotinylated h␤ 2 m in solution to ensure that the selected aptamer pool contained ligands for native epitopes, SELEX was carried out at pH 6.2 as this is both optimal for h␤ 2 m/⌬N6 co-polymerization (h␤ 2 m does not self-assemble spontaneously on a relevant time scale at this pH, whereas ⌬N6 assembles rapidly) (4) and is physiologically relevant for amyloid deposition in patients with DRA (41,42). In total, 12 SELEX rounds were performed, with rounds 6 -12 having increased stringency (see "Experimental Procedures"). From the final pool, 11 RNA clones were sequenced and aligned using the program AliBee (43). These were then clustered using the phylogenetic software Clustal Omega ( Fig. 2A) (44).
Isolation of 2ЈF B6 and Characterization of the Binding Affinity to h␤ 2 m-Two aptamers, B6 and B9, contained the most frequently occurring sequence motifs within the sequenced clones and showed some motif similarities (Fig. 2, A and B). To identify which aptamer to utilize for further studies, an initial binding assay was employed using surface plasmon resonance. Biotinylated h␤ 2 m, ⌬N6, or the non-amyloidogenic murine ␤ 2 m (45) were immobilized on separate flow cells and aptamer binding monitored at pH 6.2. 2ЈF B6 binds to h␤ 2 m with an apparent affinity of ϳ500 nM (red trace in Fig. 2C), but did not bind ⌬N6 (dark green trace) or murine ␤ 2 m (light green trace). In contrast, binding of 2ЈF B9 was so weak that the K d could not be determined (data not shown). The secondary structure of B6, computed via Mfold to be a stable stem loop (⌬G°ϳ Ϫ16 kcal/ mol) (Fig. 3A) (35), was confirmed using enzymatic solution structure probing (Fig. 3B). This analysis suggests that the selected region consists of an extended base-paired stem loop interrupted by several single-stranded bulges with a terminal loop consisting of a poly-U tetraloop (highlighted in red in Figs. 2B and 3A). Note that both Mfold and enzymatic probing were of transcripts containing natural pyrimidines. B9 is predicted to have several equivalently stable structures that are all identical in the selected region, which forms a structure very similar to that of B6 around one of the bulges (Fig. 3C). B9 differs radically at the terminal loop, however, which is composed of six purine nucleotides. It appears that the loop is the motif that provides much of the binding energy for the interaction of B6 with h␤ 2 m. Further characterization was therefore restricted to B6 and its derivatives.
A truncated 44-nucleotide version of the 110-nucleotide fulllength B6 was produced encompassing nucleotides 22 to 59 with 2ЈOH (termed 2ЈOH B6min) (Fig. 3D) or 2ЈF pyrimidines (termed 2ЈF B6min) (Fig. 3E), i.e. all of the selected region defining the stem loop with some stabilizing additional base pairs. We examined the solution binding of 2ЈF B6min to native h␤ 2 m and ⌬N6 using fluorescence spectroscopy. h␤ 2 m has two tryptophan residues: Trp-60, which lies in the DE loop (Fig. 1A) and is solvent exposed, and Trp-95, which lies toward the C terminus of the 100-residue protein and is buried. Tryptophan fluorescence of h␤ 2 m can be used to probe changes in conformation or chemical environment upon aptamer binding, with Trp-95 reporting on alterations within the hydrophobic core (46), whereas Trp-60 is sensitive to ligand binding (at least in proximity to this residue) at the protein surface. The fluorescence emission spectrum of monomeric h␤ 2 m (1 M) was monitored upon titration with 2ЈF B6min. The results showed a decrease in tryptophan emission intensity (with little change in max ), consistent with binding of 2ЈF B6min to the protein surface adjacent to Trp-60. Fitting the normalized intensity of Trp fluorescence versus the concentration of 2ЈF B6min added (Fig.  4A) (see "Experimental Procedures") yielded a Hill slope of 0.99 Ϯ 0.06, suggesting a specific one-site binding event, with an EC 50 of 223 Ϯ 10 nM. Similar assays using 2ЈOH B6min showed no binding to h␤ 2 m (Fig. 4B), indicating that the 2ЈF modifications to the pyrimidines are required for tight binding, consistent with the contribution of the polyU tetraloop to affinity. The fluorescence assay also showed no binding of 2ЈF B6min to ⌬N6 monomers (Fig. 4C), consistent with the surface plasmon resonance data with full-length aptamers. These results indicate, therefore, that 2ЈF B6min is capable of discriminating between h␤ 2 m and ⌬N6. 2Ј-Fluoro-ribose is known to prefer different sugar pucker conformations compared with unmodified residues (O4Ј-endo versus C3Ј-endo, respectively (47)). This could alter the conformation of the tetraloop and hence its interaction with the protein.
Determining the Binding Interface of B6min with Natively Folded h␤ 2 m Using NMR-To determine whether binding of 2ЈF B6 to h␤ 2 m induces conformational changes in the protein and to map the binding site in residue-specific detail, 2ЈF B6min was titrated into 15 N, 13 C-labeled h␤ 2 m at 0, 0.25, 0.5, 1.0, and 2.0 molar equivalents at pH 6.2, and 1 H-15 N HSQC spectra were recorded. The 1 H-15 N HSQC spectrum of the 2:1 mixture of 2ЈF B6min and h␤ 2 m is shown in Fig. 5, A and B. Addition of 2ЈF B6min results in the appearance of new peaks in the spectrum and the loss of resonances assigned to native apo-h␤ 2 m, indicating that the complex is in slow exchange with the apoprotein, as expected for a high affinity complex. The chemical shift changes involve some, but not all, resonances, indicative of binding of the aptamer to a specific surface. The 1 H-15 N HSQC spectrum of the h␤ 2 m-2ЈF B6min complex was assigned using a combination of two-dimensional and three-dimensional NMR techniques (see "Experimental Procedures") (Fig. 6A). The low sample concentration (60 M) and relatively large size of the complex (25.6 kDa) made assignment challenging. Of the 88 main chain resonances in the 1 H-15 N HSQC spectrum of h␤ 2 m, 55 were successfully assigned. The assigned spectrum of the 2ЈF B6min-h␤ 2 m complex was then used to map the binding site for 2ЈF B6min on the surface of the protein. Residues with the largest chemical shift differences upon aptamer binding are located on the face of h␤ 2 m that contains the A, B, E, and D ␤-strands (Fig. 6, B and C). A significant number of residues in this region could not be assigned unambiguously in the spectrum of the complex, suggesting that they experience large chemical shift differences upon aptamer binding or are not detected due to exchange line broadening (Fig. 6, B and C). The titration was also performed using 2ЈOH B6min (Fig. 5, C and  D). No changes in the chemical shifts of h␤ 2 m were observed, even at the 2:1 aptamer/h␤ 2 m molar ratio, confirming that the presence of 2ЈF-modified pyrimidines is vital for high affinity binding. To investigate whether 2ЈF B6min is able to recognize ⌬⌵6, 2 molar equivalents of the aptamer were added to 60 M 15 N-labeled ⌬⌵6, and binding was again assessed by monitoring changes in chemical shifts (Fig. 5, E and F). In this sample, the large changes in chemical shifts observed previously in the h␤ 2 m-2ЈF B6min complex (Fig. 5, A and B) were not detected (e.g. compare residues Lys-41 and Ala-79 in Fig. 5, B  and F). For some resonances, small changes in chemical shift FIGURE 3. Secondary structures of the B6 and B9 aptamers. A, the Mfold secondary structure prediction of the full-length B6 aptamer with the nucleotides within the green box showing the region truncated to create the B6min aptamer sequence. Nucleotides circled in red define the random region. B, enzymatic solution structure probing of the full-length B6 transcript with the random region highlighted in red. Cleavage sites by the G-specific RNase T1 (green arrows), U and C-specific RNase A (blue arrows), and single-stranded RNA specific S1 nuclease (purple arrows) are shown. C, the Mfold of the full-length B9 aptamer with the selected region highlighted as described in A. The dotted red boxes in A and C showed the conserved sequences and secondary structure elements of both aptamers. D, secondary structure of 2Ј OH B6min. E, 2ЈF B6min stem loops. These have additional 5Ј-GGG and 3Ј-CCCG sequences added to increase their folded stability. 2ЈF pyrimidines are circled in green in E.
were observed; however, in those cases, the chemical shifts did not saturate, even in the presence of a 2-fold molar excess of 2ЈF B6 (e.g. residues Ser-20 and Cys-80 (Fig. 5F)). The results thus confirm a significantly lower affinity of this aptamer for ⌬N6.
To obtain more detailed information about the position of the 2ЈF B6min binding site on the surface of h␤ 2 m, the intensity of each resonance was determined in the presence of a 2-fold molar excess of aptamer and compared with the intensity of its apo counterpart. The results of this analysis are shown in Fig.  7A. Resonances arising from residues in the A, B, E, and D ␤-strands, the AB and DE loops, residues 3-6 in the N-terminal region, and the C-terminal 6 residues (red in Fig. 7A) lose Ͼ80% of their intensity in the spectrum of the complex. These residues form a contiguous surface on h␤ 2 m (Fig. 7A) and include the N-terminal 6 residues of h␤ 2 m that are lacking in ⌬N6 and confer increased affinity, consistent with these residues forming part of the interface between the RNA and the protein. Consistent with this, there is little or no change in intensity for residues that lie in the CCЈ loop and F and G ␤-strands on the opposite face of h␤ 2 m (gray in Fig. 7A). By contrast with these results, addition of a 2-fold molar excess of 2ЈOH B6min to h␤ 2 m has no significant effect on the intensities of the resonances of native h␤ 2 m (Fig. 7B), consistent with its lack of binding.
A similar analysis was performed to assess the possible interaction between 2ЈF B6min and ⌬⌵6. As expected based on the fluorescence titration results shown in Fig. 4C, little change in intensity was observed for the vast majority of residues in this sample (compare Fig. 7, A and C), consistent with weak binding to ⌬N6. Furthermore, the residues that do show a difference in resonance intensity differ from those involved in the 2ЈF B6min-h␤ 2 m interface. For example, although resonances belonging to residues in the AB loop, the E strand, and the C-terminal 6 residues of native h␤ 2 m diminish in intensity by Ͼ80% upon interaction with 2ЈF B6min, these resonances are largely unaffected (retaining Ͼ60% average intensity) when ⌬⌵6 is incubated with the aptamer. Moreover, the residues in ⌬N6 showing the largest decrease in intensity upon addition of 2ЈF B6min (red in Fig. 7C) are spread throughout the structure of the protein, suggesting that binding of 2ЈF B6min to ⌬⌵6 is less specific than the 2ЈF B6-h␤ 2 m interaction. These differences in binding presumably explain the insensitivity of tryptophan fluorescence observed upon addition of 2ЈF B6min to ⌬⌵6.
The 2ЈF B6min-h␤ 2 m interface defined by these experiments (Fig. 8, A-C) includes a large number of aromatic side chains (Tyr-10, Phe-22, Tyr-26, Phe-56, Tyr-63, Tyr-66, and Tyr-67, green in Fig. 8C), as often found in protein-RNA complexes (48). The residues involved in the binding interface might also be expected to be positively charged, but there appears to be an equal balance of positively charged residues (Arg-3, Lys-6, His-13, Lys-19, Lys-48, His-51, and Lys-94 (blue in Fig. 8C)) and negatively charged side chains (Glu-16, Asp-38, Glu-50, Asp-53, Asp-59, Glu-69, and Asp-98 (pink in Fig. 8C)). Analysis of the NMR data shows that residues 3-6 are clearly part of the binding site. This sequence is absent in ⌬⌵6, which binds very poorly, and contains two positive charges (Arg-3 and Lys-6), but no negative charges. This region is therefore a candidate for a favorable electrostatic interaction with the aptamer. Indeed, murine ␤ 2 m, which does not bind this aptamer (Fig. 2C), has a Gln substituted for Lys at residue 6 (N-terminal sequence of These data were normalized between 0 (no aptamer) and 1 (the fluorescence signal when 1.7 ⌴ of 2ЈF B6min was added to h␤ 2 m. All experiments were performed in 50 mM MES buffer, 120 mM NaCl, pH 6.2. AFU, arbitrary fluoresence units. murine ␤ 2 m is IQKTPQ), implying that Lys-6 is a likely key recognition element for h␤ 2 m.
2ЈF B6min Alters the Co-assembly of ⌬N6 and h␤ 2 m-The NMR and fluorescence data presented above indicate that 2ЈF B6min binds tightly to h␤ 2 m, but only weakly and non-specifically to ⌬N6. At pH 6.2 h␤ 2 m does not self-assemble into amyloid fibrils in vitro over a time scale of several weeks at a concentration of 40 M, even using significant agitation (13, 49, 50).  In contrast, ⌬N6 rapidly and quantitatively forms fibrils under these conditions (19,49). When the two proteins are incubated together at this pH, they co-polymerize, forming hetero-polymeric fibrils with distinct structural properties compared with either of their homo-polymeric counterparts (4). To determine whether 2ЈF B6min is able to affect the co-aggregation of h␤ 2 m and ⌬N6 (due to preferential binding of the aptamer to one of the fibrillating monomers), the two proteins were mixed (each at a concentration of 40 M) in the presence or absence of a 2-fold molar excess of 2ЈF B6min (160 M) at pH 6.2. Assembly was monitored by separating soluble and insoluble material by centrifugation and subsequent analysis of each fraction by SDS-PAGE (see "Experimental Procedures") ( Fig. 9). In parallel, a sample of the assembly products were monitored using transmission EM (TEM) to confirm whether amyloid fibrils were produced. The results of these experiments showed that in the absence of 2ЈF B6min, each protein remains in the soluble fraction up to the 24-h time point, after which time insoluble material containing both proteins forms (Fig. 9A). After 166 h of incubation, both proteins are also found in the pellet presumably due to their co-polymerization into fibrils (4). By contrast, in the presence of 2ЈF B6min aggregation occurs more rapidly, with Ͼ90% of ⌬N6 and ϳ40% of h␤ 2 m forming fibrillar material after 24 h. TEM images of the samples after 166 h confirmed that the insoluble material in the pellets contains amyloid fibrils (Fig. 9, A and B), although the precise location of each protein within each fibril (i.e. the extent to which co-polymerization occurred) could not be ascertained from these experiments. Presumably, the interaction between soluble ⌬N6 and h␤ 2 m is inhibited by 2ЈF B6min, leading to rapid polymerization of ⌬N6, which in part co-polymerizes with h␤ 2 m.

DISCUSSION
To derive a structural mechanism of amyloid formation, the identity and structure of all assembling components must be defined, and how these species interact and form the cross-␤structure of amyloid should be determined. Here, RNA SELEX has been used to generate a specific, high-affinity aptamer (2ЈF B6) against monomeric h␤ 2 m. Importantly, despite only subtle differences in the structures of monomeric h␤ 2 m and its N-terminal truncation variant ⌬N6 at pH 6.2 (Fig. 1, A and B), 2ЈF B6 is able to discriminate between these structures, showing tight and highly specific binding to the ␤-sheet containing the A, B, E, and D strands of h␤ 2 m. By contrast, weak, nonspecific binding is observed to ⌬N6 that is detectable only at the high protein and RNA concentrations used for NMR (60 M protein). The discrimination between h␤ 2 m and ⌬N6 by 2ЈF B6 can be explained, at least in part, by the presence of Lys-6 in the bind- ing interface. However, given that the binding interface appears to involve an extended region spanning the A, B, E, and D ␤-strands, other residues must also contribute to affinity. Indeed, differences in the organization of residues on the surfaces of h␤ 2 m and ⌬N6 that result from the isomerization of the X-Pro-32 peptide bond from the native cis isomer in h␤ 2 m to the trans isomer in ⌬N6 (19), and/or the decreased stability (13) and increased conformational dynamics of ⌬N6 compared with h␤ 2 m (19), may also contribute to 2ЈF B6 discriminating between these otherwise similar structures. For example, although the structure of the backbone is highly conserved between h␤ 2 m and ⌬N6 (Fig. 1B), the orientation of the side chains of aromatic residues involved in the aptamer binding interface differs significantly (Fig. 10A). Furthermore, the organization of hydrophobic and charged residues on the surface formed by the A, B, E, and D ␤-strands in h␤ 2 m differs significantly from ⌬N6 (Fig. 10B). Accordingly, the apical region of this surface in h␤ 2 m is more highly positively charged than its equivalent in ⌬N6 (this region contains the N-terminal six amino acids, including Lys-6) (Fig. 10B). In addition, the organization of negatively charged residues (involving the AB loop, the EF loop, and the C terminus) also differs between the two FIGURE 8. Mapping the 2F B6min-h␤ 2 m binding site. A, the residues in h␤ 2 m that show the largest decrease in intensity upon interaction with 2ЈF B6min are shown in red on the structure of h␤ 2 m (gray schematic) and predominantly involve residues in the A, B, E, and D ␤-strands of h␤ 2 m. By contrast, the C, F, and G ␤-strands show relatively little change in intensity (bottom). B, surface representation of h␤ 2 m highlighting the interface residues (red). C, the 2ЈF B6min-h␤ 2 m binding interface involves seven aromatic residues (light green), seven positively charged residues (blue), and seven negatively charged residues (pink). proteins (Fig. 10B). In total, therefore, the balance between electrostatic and hydrophobic residues, crucial for nucleic acid binding (51), is distinct in h␤ 2 m and ⌬⌵6, partly due to the removal of the N-terminal six amino acids, and due in part to differences in solvent exposure of hydrophobic residues in the DE and BC loops in the two proteins that occur as a consequence of X-Pro-32 isomerization.
The role of ⌬N6 in DRA is not currently understood. Although ⌬N6 is present in the amyloid deposits found in patients with DRA (52), it remains unknown whether the N-terminal truncation of h␤ 2 m occurs pre-or post-fibril formation. Additionally, the interaction between h␤ 2 m and ⌬N6 in vitro is complex, with ⌬N6 possessing the ability to convert monomeric h␤ 2 m into an amyloidogenic conformation (4,19,53) and to act as a fibrillar seed able to be elongated with h␤ 2 m monomers (19,49). The aptamer selected here may be useful as an analytical probe to derive greater clarity in understanding the early stages of h␤ 2 m and ⌬N6 co-assembly into amyloid. Given the complexity of amyloid formation, where self-assembly can be initiated by one or more rare conformers that may differ subtly in structure, and that different oligomeric species may exhibit profoundly different cytotoxicity (54,55), RNA aptamers offer unique potentials as reagents for the analysis of, and interference with, amyloid formation.
The specific and tight binding of 2ЈF B6 to h␤ 2 m alters the course of amyloid assembly in mixtures of h␤ 2 m and ⌬N6 at pH 6.2. Thus, aptamer binding to h␤ 2 m disfavors the incorporation of h␤ 2 m into amyloid fibrils during co-assembly with ⌬N6 and results in more rapid fibril formation. In the presence of the aptamer, h␤ 2 m molecules will become incorporated into fibrils only after aptamer dissociation, possibly by cross-seeding with preformed ⌬N6 fibrils (4,49). Alternatively, ⌬N6 may promote conversion of h␤ 2 m to an amyloidogenic conformation once 2ЈF B6min dissociates (4,19,53), pulling the equilibrium toward co-assembly into fibrils. Given that amyloid formation is under kinetic control, the development of aptamers able to bind their targets with slow off-rates (even for the same apparent K d ) would provide an effective strategy to control assembly. Such aptamers could be isolated by increasing the length of time of the elution steps in SELEX as stringency is increased. Alternatively, coupling of the RNA aptamer to a molecule with known affinity to the target could provide a route to achieving this goal by exploiting avidity effects. Doxycycline, a small molecule tetracycline analog, has been shown to modulate the formation of h␤ 2 m fibrils in vitro (56), to reduce articular pain and improve movement in DRA patients (57), and to correct a locomotory defect in Caenorhabditis elegans expressing h␤ 2 m (58). Analysis of the h␤ 2 m-doxycycline complex using NMR suggests that the highest affinity binding site (IC 50 ϳ 50 M (56)) involves residues that lie in the C-terminal region of strand A, the N-terminal region of strand B, and the central residues of the AB loop (56). A second, lower affinity binding site involves the N-terminal region and residues in the BC and DE loops. An intriguing possibility, therefore, would be to create an aptamer linked to doxycycline such that the relatively tight and specific binding of 2ЈF B6 can be exploited to enhance binding of doxycycline to its target interface. Creation of such bipartite molecules has been shown to be a highly effective strategy, not just for enhancing the effectiveness of RNA aptamers as delivery vehicles (59 -61), but in many other applications (62)(63)(64).
In conclusion, the biophysical and biochemical studies presented here demonstrate that RNA aptamers can be highly specific and discriminatory probes, modulating co-polymerization reactions and controlling the course of amyloid assembly. How the 2ЈF B6-h␤ 2 m complex changes as fibril formation proceeds and the effect of the aptamer on hetero-polymorphic fibril structure and stability will require further studies, for example, by exploiting the powers of solid-state NMR to analyze fibril structures (65,66). Further characterization and modification of 2ЈF B6 will potentially allow the affinity of the aptamer for h␤ 2 m to be increased, and selection of aptamers specific for ⌬N6 will also allow detailed biophysical analysis of the role of ⌬N6 in h␤ 2 m-⌬N6 co-polymerization. Understanding this process further may shed light on the molecular mechanisms of fibril formation and how the protein precursors of hetero-polymeric assemblies can be modulated to tailor the extent, rate, and structure of amyloid fibrils.
Acknowledgments-We thank members of our research groups for helpful discussion and comments. We also acknowledge The Wellcome Trust for funding the NMR facility (094232). FIGURE 10. Structural differences between h␤ 2 m and ⌬⌵6 in the aptamer binding surface. A, the aromatic residues located in the interface between 2ЈF B6min and h␤ 2 m (see Fig. 8) are highlighted as sticks on h␤ 2 m (black ribbon) and ⌬⌵6 (red ribbon). Close ups of four residues are shown alongside. B, the structure of h␤ 2 m (left) and ⌬⌵6 (right) shown as a surface representation colored by its electrostatic potential (blue, positive; red, negative). The N-terminal region is highlighted in a circle, and the DE loop region is annotated with a black arc.