Structural Insights into Aβ42 Oligomers Using Site-directed Spin Labeling*♦

Background: Aβ42 oligomers underlie neurotoxicity in Alzheimer disease, but their molecular structures are unknown. Results: Electron paramagnetic resonance studies reveal intermolecular distances at 11.5–12.5 Å for C-terminal region of Aβ42. Conclusion: Aβ42 oligomers consist of a tightly packed C-terminal region that adopts antiparallel structures. Significance: This work provides insights into the structures of Aβ42 oligomers and helps understand their oligomerization mechanism and toxicity. Oligomerization of the 42-residue peptide Aβ42 plays a key role in the pathogenesis of Alzheimer disease. Despite great academic and medical interest, the structures of these oligomers have not been well characterized. Site-directed spin labeling combined with electron paramagnetic resonance spectroscopy is a powerful approach for studying structurally ill-defined systems, but its application in amyloid oligomer structure study has not been systematically explored. Here we report a comprehensive structural study on a toxic Aβ42 oligomer, called globulomer, using site-directed spin labeling complemented by other techniques. Transmission electron microscopy shows that these oligomers are globular structures with diameters of ∼7–8 nm. Circular dichroism shows primarily β-structures. X-ray powder diffraction suggests a highly ordered intrasheet hydrogen-bonding network and a heterogeneous intersheet packing. Residue-level mobility analysis on spin labels introduced at 14 different positions shows a structured state and a disordered state at all labeling sites. Side chain mobility analysis suggests that structural order increases from N- to C-terminal regions. Intermolecular distance measurements at 14 residue positions suggest that C-terminal residues Gly-29–Val-40 form a tightly packed core with intermolecular distances in a narrow range of 11.5–12.5 Å. These intermolecular distances rule out the existence of fibril-like parallel in-register β-structures and strongly suggest an antiparallel β-sheet arrangement in Aβ42 globulomers.

Aggregation of amyloid ␤ (A␤) 2 peptide is a key event in the pathogenesis of Alzheimer disease (AD) (1). A␤ is the proteolytic product of amyloid precursor protein by the sequential cleavages of ␤and ␥-secretases (2). There are two major A␤ isoforms: A␤42 (42 residues long) and A␤40 (40 residues long). A␤42 differs from A␤40 by having two additional C-terminal residues. Although A␤40 is severalfold more abundant than A␤42 in the brain, A␤42 is the major component in AD plaques (3), suggesting that the aggregation of A␤42 precedes A␤40. Aggregation of A␤ results in the formation of insoluble amyloid fibrils, soluble oligomers, and protofibrils. In recent years, AD research has shifted from a fibril-centric to an oligomer-centric view, reflecting the good correlation between soluble A␤ oligomers and AD progression (4,5). A number of oligomeric A␤ assemblies have been identified in vivo and in vitro. Oligomer formation has also been found in other amyloid-related disorders, suggesting that oligomer-induced toxicity may be a common mechanism for a range of protein misfolding diseases (6).
Different A␤ oligomers such as dimers (7,8), trimers (9), and A␤*56 (9) have been identified in AD brains, transgenic AD model animals, or cultured cells. Various protocols have also been devised to stabilize and enrich A␤ oligomers in the test tube (10,11). Antibodies have been generated using several of these oligomer preparations including A␤-derived diffusible ligands (12), globulomers (13), prefibrillar oligomers (14), and amylospheroids (15). Positive staining has been found in AD brain slices using these antibodies, suggesting that these A␤ oligomers are related to endogenous A␤ assemblies. Without the atomic-level structural information of these A␤ oligomers, it is impossible to know how many unique structures are present in these oligomers. The general lack of cross-reactivity among different oligomer-specific antibodies suggests that the surface-exposed groups are distinct in different oligomer preparations. On the other hand, similar structural features have also been found, which include ␤-strand structures at the central hydrophobic cluster (residues [17][18][19][20] and the C-terminal hydrophobic region (residues 30 -40) and a turn/loop structure at residues 20 -30. Antiparallel arrangement between ␤-strands has also been suggested based on Fourier transform infrared (FTIR) studies of A␤ oligomers prepared using a range of protocols (16 -24).
Detailed structural studies of these A␤ oligomers are urgently needed for several reasons. First, A␤ oligomers are toxic species and thus represent drug targets in AD. Second, structural knowledge is critical for understanding the structural basis of A␤ toxicity and the mechanism of aggregation. Third, detailed structural studies will help to determine whether each in vitro oligomer preparation can be assigned a unique structural identity. High-resolution structural studies on peptide fragments and macrocyclic ␤-sheet mimics have provided insights into the potential organization of amyloid oligomers (25,26). However, detailed structures of full-length A␤ oligomers still remain elusive.
To this end, we performed a comprehensive structural study on a toxic A␤42 oligomer termed globulomer (13) using sitedirected spin labeling and electron paramagnetic resonance (EPR) spectroscopy (27)(28)(29)(30)(31). Globulomers are a relatively stable and homogeneous preparation of A␤ oligomers, and they do not convert to fibrils (32). Endogenous A␤ globulomers are detected in the brain by globulomer-specific antibodies (13,33). Previous studies suggest that globulomers are dodecamers and show synaptic toxicity (13,33,34). Solution NMR studies on preglobulomers show that they contain both antiparallel and parallel structures (35), but the detailed structures of globulomers are still unknown.
In this work, we demonstrate that A␤42 forms globular oligomers revealed by transmission electron microscopy (TEM), which are toxic to mammalian cell lines. Circular dichroism (CD) and x-ray powder diffraction show that these oligomers consist of cross-␤-structures. We then performed site-directed spin labeling EPR studies on the A␤42 globulomers with spin labels introduced, one at a time, at 14 different residue positions throughout the A␤42 sequence. Spin label mobility analysis shows increasing structural order from N-to C-terminal regions. Distance measurements reveal that intermolecular distances for most labeling sites are in the range of 11.5-13.5 Å in the oligomers. These measured distances indicate a tightly packed C-terminal region and a loosely packed N-terminal region. These distances also suggest that A␤42 globulomers adopt antiparallel ␤-structures.

EXPERIMENTAL PROCEDURES
Preparation of A␤42 Peptides and Spin Labeling-The DNA construct of GroES-ubiquitin-A␤42 (36) and the deubiquitylating enzyme Usp2cc (37) were kindly provided by Dr. Rohan T. Baker at Australian National University (Australia) and Dr. Il-Seon Park at Chosun University (South Korea). Single cysteine mutations at various sites were introduced into A␤42 sequence using a QuikChange kit (Agilent). Mutations were confirmed with DNA sequencing.
Expression of GroES-ubiquitin-A␤42 and Usp2cc proteins in Escherichia coli and their purification were performed as described previously (28,38). Full-length A␤42 was cleaved from the fusion protein with Usp2cc. The enzymatic digestion was performed in a buffer containing 19 mM phosphate, 3 M urea, 2 mM tris(2-carboxyethyl)phosphine, pH 10.0. Usp2cc was added to the fusion protein at a molar ratio of 1:100. The digestion reaction was allowed to proceed at 37°C for 15 min. The reaction mixture was then immediately filtered with a 0.2-m filter (Whatman) and loaded on a 5-ml HisTrap column (GE Healthcare) equilibrated with PSU buffer (50 mM phosphate, 0.3 M NaCl, 8 M urea, pH 10.0). A␤42 protein was eluted using a gradient of 25 mM imidazole. Purified A␤42 was checked with SDS-PAGE, and no uncleaved proteins were detected.
For spin labeling, dithiothreitol was added to purified protein fraction to a final concentration of 10 mM and was allowed to incubate at room temperature for 20 min to break any disulfide bonds. Then the A␤42 sample was buffer-exchanged to labeling buffer (20 mM MOPS, 7 M guanidine hydrochloride, 50 mM NaCl, pH 6.8) using a 5-ml HiTrap desalting column (GE Healthcare). The spin labeling reagent 1-oxyl-2,2,5,5-tetramethylpyrroline-3-methyl methanethiosulfonate (Enzo Life Sciences) was added at 10 times molar excess and then incubated at room temperature for 1 h. The spin-labeled A␤42 was further buffer-exchanged to 30 mM ammonium acetate, pH 10.0, lyophilized, and stored at Ϫ80°C. MALDI-TOF mass spectrometry was performed to ensure that the mass of A␤42 is correct and that the extent of labeling is Ͼ95%.
Preparation of A␤42 Globulomer-A␤42 oligomers were prepared using the globulomer protocol (35) with modifications. A␤42 peptide was suspended in 100% 1,1,1,3,3,3hexafluoro-2-propanol (HFIP) at 1 mM and then bath-sonicated for 5 min. Then the sample was incubated at room temperature for 30 min. HFIP was removed by evaporation overnight in the fume hood and then under vacuum for 1 h. HFIP-treated A␤42 was resuspended in dimethyl sulfoxide at 5 mM and sonicated for 5 min. Then A␤42 was diluted in phosphate-buffered saline (PBS) (20 mM phosphate, 140 mM NaCl, pH 7.4) to 400 M, and sodium dodecyl sulfate (SDS) was added to a final concentration of 0.2%. After incubation for 6 h at 37°C, A␤42 samples were further diluted with 3 volumes of H 2 O to a final concentration of 100 M and incubated for another 18 h at 37°C. After centrifugation at 14,000 ϫ g for 20 min, the supernatant was concentrated by ultrafiltration using 30-kDa molecular mass cut-off filters.
Transmission Electron Microscopy-The A␤42 oligomer sample (5 l) was placed on glow-discharged copper grids covered with 400-mesh Formvar/carbon film (Ted Pella). The sample was negatively stained with 2% uranyl acetate. Samples were examined using a JEOL JEM-1200EX transmission electron microscope at 80 kV, and images were recorded using a topmounted charge-coupled device camera (Gatan).
MTT-based Cell Viability Assays-3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT)-based CellTiter 96 nonradioactive cell proliferation assay (Promega) was performed for a cell viability test. HeLa and PC-12 cell lines were used in the assay. Prior to the MTT test, PC-12 cells were cultured in ATCC-formulated RPMI 1640 medium (ATCC catalog number  with 10% heat-inactivated horse serum and 5% fetal bovine serum at 37°C in 5% CO 2 , and HeLa cells were cultured in DMEM medium with 10% fetal bovine serum at 37°C in 5% CO 2 . HeLa and PC-12 cells were plated out at 10,000 cells/well in 96-well plates (Costar catalog number 3596) and cultured for 20 h at 37°C in 5% CO 2 . To initiate the test, 10 l of sample was added into each well containing 90 l of medium. After 24 h of incubation, 15 l of dye solution (Promega catalog number G4000) was added into each well, and the mixture was further incubated for 4 h at 37°C in 5% CO 2 . Then 100 l of solubilization solution/stop mix (Promega catalog number G4000) was added to each well with 12 h of incubation at room temperature for complete solubilization of the dye crystals. The absorbance was measured at 570 nm. The background absorbance was recorded at 700 nm. The readout from buffer blank treated cell is regarded as 100% viability, and that from 0.2% treated cell is treated as 0% viability. Four replicates per sample were measured for calculating standard deviation.
Circular Dichroism-Secondary structures of A␤42 samples were analyzed by CD spectroscopy. Samples (ϳ40 l) were placed into a 0.01-cm path length quartz cell (Starna). A JASCO J-715 CD spectrometer was employed. The measurements were carried out in a wavelength range of 190 -260 nm at a rate of 20 nm/min with a step resolution of 0.5 nm, a time constant of 4 s, and a bandwidth of 1 nm. The CD spectra were obtained by averaging 16 scans. The temperature was set at 25°C. All the spectra were corrected by subtracting the buffer background.
X-ray Powder Diffraction-A␤42 oligomer was washed with H 2 O using the 30-kDa molecular mass cut-off filter and lyophilized. The powder was mounted on the tip of a broken glass rod. Then, the specimen was placed on the goniometer of an in-house x-ray machine and shot using a Rigaku FR-D x-ray generator equipped with a Rigaku HTC imaging plate detector.
Analytical Size Exclusion Chromatography-The size of A␤42 globulomer was analyzed by analytical SEC using a Superdex 200 10/300 GL column (GE Healthcare) attached to an ÄKTA explorer system (GE Healthcare). Prior to A␤42 oligomer sample injection (50 l), the column was washed with 2 column volumes of filtered, degassed globulomer buffer (5 mM phosphate, pH 7.4, 35 mM NaCl, 2% DMSO, and 0.05% SDS). The molecular mass of A␤42 globulomer was estimated based on the SEC profiles of protein standards (Bio-Rad).
EPR Spectroscopy-EPR measurements were performed at X-band frequency on a Bruker EMX spectrometer equipped with the ER 4102ST cavity. A modulation frequency of 100 kHz was used. For measurements performed at room temperature, 20 milliwatts of microwave power was employed. EPR spectra of both the retentate and the filtrate after ultrafiltration with 30-kDa filter were collected. The filtrate component, accounting for Ͻ1% of total EPR signal, was subtracted from the retentate spectra to obtain the spectra of "pure" globulomers. For measurements performed at 170 K, 631 milliwatts of microwave power was used. Modulation amplitude was optimized to each individual spectrum. For each sample, 20 l of solution was loaded into glass capillaries (VitroCom) sealed at one end. For measurements performed in frozen solution at 170 K, 20% (v/v) glycerol was added to the sample solution followed by flash-freezing the sample in liquid nitrogen. EPR spectra in each figure panel were normalized to the same number of spins.
Spectral Simulations and Distance Analysis-Spectral simulations were performed using the program MultiComponent (developed by Dr. Christian Altenbach, UCLA), which provides a LabVIEW interface for the program NLSL developed by Freed and co-workers (39,40). A microscopic order/macroscopic disorder model was used for all simulations (40). A nonlinear least squares fit of the user-defined spectral parameters was performed using the Levenberg-Marquardt algorithm. For all fits, the values for the magnetic tensors A and g were fixed as A xx ϭ 6.2, A yy ϭ 5.9, A zz ϭ 37.0, and g xx ϭ 2.0078, g yy ϭ 2.0058, g zz ϭ 2.0022, which were determined previously for R1 (59). Each spectrum was fitted with two spectral components. The isotropic rotational diffusion constant (D) was allowed to vary for each component. The rotational correlation time () was calculated from ϭ 1/(6D). The values of and populations of each spectral component are plotted in Fig. 3. Local structural stability at each labeling site was calculated using ⌬G ϭ ϪRTlnK, where K is the ratio between the slow component population and the fast component population, while assuming the fast component represents the locally unfolded state.
Distance analysis was performed using the program Short-Distances (developed by Dr. Christian Altenbach, UCLA). The detailed fitting procedure to obtain distances has been previously described (41). The 25% labeled spectra were used as the spectra without dipolar interactions. The width of the distance distribution was fixed at 3 Å. The distance, percentage of the spin labels at the fitted distance, the percentage of noninteracting spin labels were allowed to vary. The best fits of these parameters are plotted in Fig. 5.

RESULTS
Characterization of Wild-type A␤42 Globulomers-We prepared A␤42 oligomers with a globulomer protocol (13), which uses low concentration of SDS (0.05%) to mimic the lipid environment in the cell (42). TEM shows that the oligomer preparation contains mostly globular structures (Fig. 1A). Most globular oligomers have diameters ranging from 5 to 10 nm, with the majority of them around 7-8 nm (Fig. 1B). Elongated structures with a beaded string morphology can also be observed. The morphologies are similar to previous atomic force microscopy studies on globulomers, which show heights of 4 -5 nm (35). MTT-based cell survival assays show that these A␤42 oligomers are toxic to both PC-12 neuronal cells and HeLa cells in a dose-dependent manner (Fig. 1C). Approximately 20% of the cells died in the presence of 0.2 M (monomer equivalent) A␤42 globulomers, and ϳ60% of cells died in the presence of 2 M A␤42 globulomers. The CD spectrum shows a major negative peak at 216 -218 nm, indicative of mostly ␤-structures (Fig.  1D), and it is similar to A␤42 fibrils (Fig. 1D). In comparison, the A␤42 monomer shows a single large negative peak at around 198 nm, characteristic of random coil structures (Fig.  1D). X-ray powder diffraction reveals three major diffraction peaks at 4.2, 4.7, and 10 Å (Fig. 1E). The 4.7 Å reflection arises from the separation of strands within the same ␤-sheet. The peak at ϳ10 Å corresponds to separations between ␤-sheets. The diffuse nature of the 10 Å peak suggests high heterogeneity of the intersheet packing. The origin of the 4.2 Å reflection is not well understood. Previously, an x-ray powder diffraction study of porcine stomach mucin protein showed a 4.2 Å reflection corresponding to distorted antiparallel ␤-structures (43), suggesting that similar structures may be present in A␤42 globulomers. A SEC profile shows that the globulomers contain a major peak with a molecular mass of ϳ100 -150 kDa (Fig. 1F). A minor peak at Ͼ700 kDa likely corresponds to the species with beaded string morphology under TEM (Fig. 1A). SEC also shows that A␤42 globulomers do not contain a significant amount of monomer or small oligomers (Fig. 1F).
Preparation of Spin-labeled A␤42 Globulomers for EPR Study-To study the structure of A␤42 globulomers by EPR, we intro-duced spin labels, one at a time, at 14 residue positions throughout the A␤42 sequence ( Fig. 2A). The spin label side chain used in this study is named R1 (Fig. 2B). TEM studies on selected spin-labeled A␤42 globulomer samples show mostly globular structures without fibrils or protofibrils (Fig. 2C), suggesting that spin labeling does not significantly perturb the formation of A␤42 globulomers.
Spin Label Mobility Analysis in A␤42 Globulomers-The EPR spectra of spin-labeled A␤42 globulomers are shown in Fig. 3A (black traces). For spin label mobility analysis, we prepared the oligomers using a mixture of spin-labeled A␤42 with wild-type A␤42 at 1:3 molar ratio. Therefore, in the oligomer sample, only 25% of the A␤42 molecules were labeled, and this will greatly reduce dipolar interactions between spin labels.
Spin-spin interactions may broaden the EPR spectra and complicate the mobility analysis.
The site-specific spin label mobility is an indicator of residue-level structural order (44). The EPR spectra of spin-labeled A␤42 globulomers encode information about the mobility of the spin label at each labeling site. The spin label mobility can be measured using center line width (45) or the ratio of high field and center line amplitude (46). For EPR spectra with multiple components reflecting multiple motional states of the spin label, however, these measurements report only the combined mobility from different spin label states. A quantitative method to extract mobility information is spectral simulation. Multiple spectral components can be simulated simultaneously, and the population of each component can be obtained from simulation (47). Therefore, we performed nonlinear least squares fitting to the EPR spectra using spectral simulation (see "Experimental Procedures"). The best fits are shown in Fig. 3A (red  traces). Spectral fitting revealed two spectral components at every spin labeling position: a fast component and a slow component (Fig. 3A, magenta and blue traces).
The presence of two spectral components suggests that globulomers contain at least two structural states whose interconversion is slower than the nanosecond timescale. The slow component has a correlation time of 2.0 -3.0 ns in the N-terminal region and 3.0 -4.5 ns in the C-terminal region, corresponding to a structured state. Spin labels at positions 1 and 4 have correlation time of ϳ2.1 ns. Although they are the most flexible region in the A␤42 sequence, the spin label mobility at these residues is comparable with ordered helix surface sites (44) or solvent-exposed ␤-sheet sites (48,49). Therefore, the EPR data suggest that the N-terminal region adopts some ordered structures. The structural order increases from N terminus to C terminus. The most ordered region spans residues 32-40, with a correlation time between 3.5 and 4.5 ns. These residues have spin label mobility comparable with spin labels located in the hydrophobic core of globular proteins (50).
The correlation time of the fast component is in the range of 0.7-1.7 ns, with decreasing side chain mobility from N-to C-terminal regions (Fig. 3B). Spin labels in completely unfolded proteins have a correlation time of ϳ0.5 ns (46,47), suggesting that this fast component corresponds to a partly disordered state. The relative population of the fast component is ϳ25% for N-terminal region (residues 1-10) and 10 -15% for most other residues (Fig. 3C), suggesting that the fast component represents a locally disordered state rather than a disordered monomer. This notion is further supported by the absence of a significant amount of monomer or small oligomer peaks in the SEC profile (Fig. 1F).
The residue-level local stability can be calculated with the assumption that the structured state (corresponding to the slow component) and the locally disordered state (corresponding to the fast component) are in equilibrium. The free energy of local unfolding in A␤42 globulomers is calculated using the percentages of slow and fast components from simulation (Fig. 3D). This local stability plot reveals three low stability regions: Asp-1-Tyr-10, Ser-26 -Gly-29, and Gly-37. The low stability at residue Gly-37 supports a likely turn structure at this position.
We investigated the temporal stability of A␤42 globulomers using the spin label as a structural probe. The EPR spectral line shape of A␤42 globulomers spin-labeled at residue 34 remains extremely similar upon incubation at room temperature (ϳ21°C) for 0, 4, 8, 24, and 48 h (Fig. 4A), suggesting no structural changes during this incubation period. The results also show that there are no structural changes during EPR measurements, which typically take less than 20 min.
Next we studied the temporal stability of A␤42 globulomers upon dilution into cell culture medium. Following a 10-fold dilution to RPMI 1640 medium, the EPR spectral line shape of A␤42 globulomer remains largely the same, consisting of two spectral components (Fig. 4B). The line shape of both the ordered and the disordered components remains unchanged upon incubation at room temperature for 0, 4, 8, 24, and 48 h. The only change is the relative population of the two components, with the disordered component increasing slightly from ϳ3.7% at 0 h to ϳ5.5% at 48 h.
Intermolecular Distance Measurements in A␤42 Globulomers-To measure interspin distances using spin-spin interactions, we prepared globulomers using only spin-labeled A␤42 (i.e. 100% labeled). Fig. 5A shows that the spectral amplitude of the 100% labeled sample (red traces) is reduced when compared with the 25% labeled sample (black traces), suggesting the existence of dipolar interactions. It should be noted that the EPR spectral line shape of the 100% labeled A␤ oligomer is dramatically different from the 100% labeled A␤ fibrils (28). In the parallel in-register ␤-structure of A␤ fibrils, spin labels are stacked on top of each other, giving rise to a single-line spectrum characteristic of strong spin exchange interactions (28). Clearly, the EPR spectra of the 100% labeled oligomers are indicative of dipolar interactions and do not have the single-line feature. Therefore, EPR data suggest that A␤42 globulomers adopt structures different from fibrils. We measured intermolecular distances in A␤42 globulomers using continuous wave EPR, which is sensitive to distances in the range of 8 -20 Å (41). For distance measurements, EPR spectra were collected at 170 K. The EPR spectra are shown in Fig. 5B, and the measured distances are shown in Fig. 5C (top  panel). The segment with the shortest distances is located at residues Gly-29 -Val-40, with distances of 11.5-12.5 Å. Residues Phe-4 -Ser-26 have slightly longer distances in the range of 12.5-13.5 Å. The two terminal residues Asp-1 and Ala-42 give distances Ͼ13.5 Å.
The intermolecular distances from EPR provide structural restraints for the intermolecular organization of A␤42 subunits . Residue-specific EPR mobility analysis for spin-labeled A␤42 globulomers. A, the experimental EPR spectra of 25% labeled A␤42 globulomers, which consist of spin-labeled and wild-type A␤42 at 1:3 molar ratio, are shown in black. The best nonlinear least squares fits from spectral simulations are shown in red. All the fits contain two spectral components: a slow component (magenta) and a fast component (blue). B, plot of correlation time for the slow and fast components obtained from spectral simulations. C, plot of the populations of the fast and slow components from spectral simulations. D, residue-level local structural stability calculated using the relative populations of the slow (for structured state) and fast (for disordered state) components. Error bars in B and C indicate fitting errors.
in globulomers. First, A␤42 globulomers consist of a tightly packed C-terminal region (Gly-29 -Val-40) and a loosely packed N-terminal region. Second, the shortest distances observed at residues Gly-29 -Val-40 are in the range of 11.5-12.5 Å, suggesting the absence of parallel in-register ␤-structures. Third, the intermolecular distances for most labeling positions are in a narrow range of 11.5-13.5 Å, suggesting an overall parallel arrangement for the A␤42 subunits at this spacing. In Fig. 6, we depict four schematic models for the potential arrangement for the C-terminal region (Gly-29 -Val-40). In Fig. 6, A and C, the measured interspin distances correspond to spacing between alternating ␤-strands within the same ␤-sheet. In Fig. 6, B and D, the interspin distances correspond to spacing between two face-to-back packed ␤-sheets.
From distance analysis, we obtained not only distances in the range of 8 -20 Å, but also the population of spin labels at the measured distances. We also obtained population of spin labels at distances Ͼ20 Å, which do not contribute to spectral broadening. Fig. 5C (bottom panel) shows that only ϳ25-50% of spin labels give rise to the measured distances in the range of 11.5-14.5 Å. The majority of the spin labels are at distances Ͼ20 Å (Fig. 5D). This information may also be useful in detailed structural modeling studies. For example, in Fig. 6A, spin labels on the green ␤-strand would fall into the category of Ͼ20 Å.
EPR Studies of A␤42 Preglobulomers-Previous studies suggest that globulomer formation involves the formation of preglobulomers in 0.2% SDS (13). After a 4-fold dilution of preglobulomers with water, preglobulomers convert to globulomers. Yu et al. (35) characterized the preglobulomers and proposed an intermolecular parallel in-register ␤-sheet model for residues 34 -41. Here we also characterized the preglobulomer structures with A␤42 L34R1. The EPR spectrum of preglobulomers prepared in 0.2% SDS also consists of two spectral components, similar to globulomers (Fig. 7A). Superimposition of the preglobulomer and globulomer spectra shows that the major difference between the two spectra is the more pronounced population of the disordered component. The line shape for the ordered component is similar between preglobu-lomer and globulomer samples, suggesting structural similarity. Because Yu et al. (35) characterized the preglobulomers prepared in 1.5% SDS with solution NMR, we also performed EPR measurements at 1.5% SDS. The EPR spectrum of preglobulomers in 1.5% SDS (Fig. 7B) is slightly sharper than that in 0.2% SDS, suggesting higher structural flexibility in 1.5% SDS. However, the overall EPR spectral line shape is similar between 0.2 and 1.5% SDS, suggesting overall similar structure for A␤42 preglobulomers at the two SDS concentrations.

DISCUSSION
Formation of A␤42 oligomers has been viewed as a key event in the pathogenesis of AD. The structures of A␤42 oligomers, although essential for rational design of therapeutic agents that target these oligomers, remain elusive. In this work, we provide a systematic investigation into the structure of a toxic A␤42 oligomer termed globulomer. CD and x-ray powder diffraction data show that A␤42 globulomers are rich in ␤-structures (Fig.  1). Side chain mobility studies show increasing structural order from N-to C-terminal regions, with residues Ile-32-Val-40 being the most ordered region (Fig. 3). Intermolecular distance measurements show that side chains at residues Gly-29 -Val-40 have an intermolecular spacing of ϳ11.5-12.5 Å (Fig. 5). Because EPR measures distances between the nitroxide groups, the linking arm between the nitroxide group and the C␣ carbon introduces an additional variant when translating distances into structural models. Despite these uncertainties, the intermolecular distances for the C-terminal residues Gly-29 -Val-40 are in a range of 11.5-12.5 Å, close to the spacing between alternating ␤-strands within the same sheet or between adjacent ␤-sheets.
Although a detailed structural model for A␤42 globulomers is still beyond reach, the distance measurements provide restraints to assess published computational models. First, the measured distances rule out parallel in-register arrangement in globulomers. Recent studies using FTIR spectroscopy have suggested the existence of antiparallel ␤-structures in A␤ oligomers prepared using a wide range of protocols (16 -24). Some . Temporal stability of A␤42 globulomers. A, EPR spectra of A␤42 L34R1 globulomers at room temperature in preparation buffer (5 mM phosphate, pH 7.4, 35 mM NaCl, 2% DMSO, and 0.05% SDS). B, EPR spectra of A␤42 L34R1 globulomers at room temperature after 10-fold dilution to cell culture medium, which is ATCC-formulated RPMI 1640 medium with 10% heat-inactivated horse serum and 5% fetal bovine serum. Dotted line was drawn to aid comparison of different spectra. FIGURE 5. Intermolecular distance measurements for spin-labeled A␤42 globulomers. A, EPR spectra of 100% labeled oligomers at room temperature show significantly reduced spectral amplitude than 25% labeled samples, suggesting intermolecular dipolar interactions. B, EPR spectra of 25 and 100% labeled A␤42 globulomers at 170 K. Interspin distances are obtained by simulating the 100% labeled spectra. The residual is the difference between simulated spectra and 100% labeled spectra. C, plot of intermolecular distance between spin labels and the percentage of spin labels at measured distances. D, plot of the populations of spin labels at Ͼ20 Å. Error bars in C and D indicate fitting errors.
investigators have questioned whether FTIR is able to definitively distinguish parallel from antiparallel structures in amyloid oligomers (51,52). The distance measurements strongly support the antiparallel arrangement in A␤42 globulomers and may put this concern to rest. In computational modeling (53)(54)(55), A␤42 subunits (residues 17-42) are generally modeled as a ␤-hairpin (56,57) or a U-shaped strand-turn-strand motif as observed in A␤ fibrils (58). The measured distances are generally in support of the antiparallel structural models in these modeling studies (53)(54)(55).
Previously, Yu et al. (35) characterized the structures of preglobulomers using solution NMR. Preglobulomers are formed in a higher concentration of SDS (0.2% versus 0.05% for globulomers). NMR data on A␤42 preglobulomers show that residues 34 -41 adopt intermolecular parallel in-register ␤-sheet structures, similar to those in fibrils. Solution NMR studies were not performed on A␤42 globulomers due to the large size, but hydrogen exchange studies show that the only highly protected segment in both preglobulomers and globulomers is the C-terminal region (residues 30 -42) (35), consistent with EPR data that show higher structural order and closer intermolecular distances in the C-terminal region. The EPR spectra of A␤42 preglobulomers are similar to those of globulomers (Fig. 7), suggesting similar structural features. The reason for this apparent contradiction between EPR results and the previous NMR studies is not clear and may result from structural polymorphism commonly observed for A␤ oligomers.
The findings here demonstrate that EPR studies can resolve structurally heterogeneous states. One property of A␤ oligomer preparations that makes it challenging to do high-resolution structural studies is the sample heterogeneity. This work has demonstrated several advantages of EPR in obtaining structural information from structurally heterogeneous samples. First, spectral subtraction can be used to obtain the EPR spectrum of pure oligomers by subtracting a spectral component with known line shape (Fig. 8A). Using a 30-kDa molecular mass cut-off filter, we separated A␤42 globulomers into two fractions. The 30-kDa filter filtrate contains disordered A␤ monomers based on the sharp EPR lines. The EPR spectra of the pure high molecular mass species can be obtained by subtracting the 30-kDa filter filtrate spectra from the retentate spectra (Fig.  8A). Second, multiple spectral components can be resolved using spectral simulations, which give dynamic parameters and populations of each component (Fig. 8B). Third, in distance measurements, the distance and its population can be determined simultaneously using spectral fitting (Fig. 8C). Furthermore, EPR has the ability to measure a wide range of distances (8 -70 Å). These advantages make site-directed spin labeling   . EPR analysis can resolve structural heterogeneity in A␤ oligomers. A, spectral subtraction to isolate the spectral component of interest. B, spectral simulations to reveal multiple structural states. C, distance analysis to reveal both the distance and the relative population of spin labels at the measured distance.