Z-Phe-Ala-diazomethylketone (PADK) Disrupts and Remodels Early Oligomer States of the Alzheimer Disease Aβ42 Protein*

Background: The early oligomer states of Aβ are the primary therapeutic target for AD. Results: PADK binds to Aβ directly and inhibits and reverses the formation of dodecamer of Aβ42. Conclusion: PADK disrupts and remodels the early oligomerization of Aβ42. Significance: The study of PADK and Aβ42 provides an example of small molecule therapeutic development for AD and other amyloid diseases. The oligomerization of the amyloid-β protein (Aβ) is an important event in Alzheimer disease (AD) pathology. Developing small molecules that disrupt formation of early oligomeric states of Aβ and thereby reduce the effective amount of toxic oligomers is a promising therapeutic strategy for AD. Here, mass spectrometry and ion mobility spectrometry were used to investigate the effects of a small molecule, Z-Phe-Ala-diazomethylketone (PADK), on the Aβ42 form of the protein. The mass spectrum of a mixture of PADK and Aβ42 clearly shows that PADK binds directly to Aβ42 monomers and small oligomers. Ion mobility results indicate that PADK not only inhibits the formation of Aβ42 dodecamers, but also removes preformed Aβ42 dodecamers from the solution. Electron microscopy images show that PADK inhibits Aβ42 fibril formation in the solution. These results are consistent with a previous study that found that PADK has protective effects in an AD transgenic mouse model. The study of PADK and Aβ42 provides an example of small molecule therapeutic development for AD and other amyloid diseases.

The oligomerization of the amyloid-␤ protein (A␤) is an important event in Alzheimer disease (AD) pathology. Developing small molecules that disrupt formation of early oligomeric states of A␤ and thereby reduce the effective amount of toxic oligomers is a promising therapeutic strategy for AD. Here, mass spectrometry and ion mobility spectrometry were used to investigate the effects of a small molecule, Z-Phe-Ala-diazomethylketone (PADK), on the A␤42 form of the protein. The mass spectrum of a mixture of PADK and A␤42 clearly shows that PADK binds directly to A␤42 monomers and small oligomers. Ion mobility results indicate that PADK not only inhibits the formation of A␤42 dodecamers, but also removes preformed A␤42 dodecamers from the solution. Electron microscopy images show that PADK inhibits A␤42 fibril formation in the solution. These results are consistent with a previous study that found that PADK has protective effects in an AD transgenic mouse model. The study of PADK and A␤42 provides an example of small molecule therapeutic development for AD and other amyloid diseases.
Alzheimer disease (AD) 2 is the leading cause of late life dementia and is characterized as a progressive brain disorder that damages synapses and eventually destroys brain cells (1,2). The aggregation of amyloid-␤ protein (A␤) into soluble oligomeric species has been implicated as a key step in AD pathogenesis (3). Among the A␤ peptides that exist in vivo, the 42amino acid A␤42 has been found to be the primary component of amyloid deposits that are a hallmark of AD (4). Recently, increasing evidence shows that the early oligomeric states rather than the later stage fibrillization of A␤42 are implicated in the onset of AD (5)(6)(7)(8). A␤42 monomers form small oligomers in solution, as well as paranuclei (pentamers and hexamers) that self-assemble to form decamers and dodecamers (9 -14). An in vivo study identified a 56-kDa A␤ species (corresponding to a dodecamer) as the cause of memory disruption in affected mice (15). Because of their role in neuronal pathology, early oligomers of A␤42 are a primary therapeutic target for AD.
A critical aspect of AD treatment and prevention is the clearance of A␤ proteins and protein aggregates from the brain (16). This clearance can be accomplished by directly disrupting the aggregation of A␤ proteins or by the degradation of A␤ proteins by proteases. Numerous natural proteins, peptides, and small molecules have been found to interfere with A␤ and disrupt the aggregation pathway (17). Small molecules are particularly attractive as a direct therapeutic strategy for the treatment of amyloidosis (18 -20). They include known bioactive molecules (for example, curcumin (21) and (Ϫ)-epigallocatechin gallate (22)) and polyphenols (23,24). On the other hand, A␤-degrading enzymes have been investigated to understand the proteolysis in AD and targeted for therapeutic intervention (25). In this respect, lysosomal enzymes play an important role in protein degradation and clearance (26,27). One of these enzymes, cathepsin B, has been shown to have anti-amyloidogenic and neuroprotective functions (28,29). Recently, a small molecule, Z-Phe-Ala-diazomethylketone (PADK; the chemical structure is shown in Fig. 1b), has been investigated for its up-regulation of the lysosomal system (30). In AD transgenic mouse models (29), PADK was shown to selectively increase the cathepsin B level in the central nervous system. This resulted in reduced A␤42 levels in the brain, which in turn offset the defects in synaptic composition and cognitive functions in the two transgenic models expressing different levels of A␤ pathology.
In the current study, we tested whether PADK directly interacts with A␤ itself. Also, we examined whether such interaction influences A␤ oligomerization and aggregation, in addition to the positive effect of PADK on the lysosomal system. Such a direct effect would provide evidence for the positive impact of PADK in AD pathology and provide an example of small molecule therapeutic development for AD and other amyloid diseases.
Recently, ion mobility spectrometry-mass spectrometry (IMS-MS) (31) has been successfully used to study the structure * This work was supported, in whole or in part, by National Institutes of Health Grants IPOIAG 027818-010003 (to M. T. B.) and R25 GM077634 (to B. A. B.). This work was also supported by the Institute for the Study of Aging (to B. A. B.). □ S This article contains supplemental text and Figs. S1-S4. 1 To whom correspondence should be addressed. and aggregation of amyloid systems (32)(33)(34)(35)(36), including amyloid-␤ protein (12,13,35,37,38). In this work, we use IMS-MS to elucidate the interaction of PADK and A␤42.

EXPERIMENTAL PROCEDURES
Sample Preparation-A␤42 was synthesized using 9-fluorenylmethyloxycarbonyl-based methods (39), and PADK was obtained from Bachem Americas, Inc. (N-1040; Torrance, CA). A␤42 and PADK were allowed to interact with each other by premixing in ammonium acetate buffer (7.5 mM, pH ϭ 7.4) with a 1:10 ratio. To slow down the aggregation process for better signal, the samples were kept on ice until the IMS experiment was performed.
Ion Mobility Spectrometry-Mass Spectrometry-The samples were analyzed on a home-built ion mobility mass spectrometer (40). IMS (31) is capable of separating species that have the same mass-to-charge ratio but different shapes and oligomer sizes (12,13). In the IMS measurements, the ions are pulsed into a drift cell filled with helium gas and passed through the cell under the influence of a weak electronic field. The species are separated in time according their sizes, and their arrival times at the detector are measured.
Electron Microscopy (EM)-All the A␤42 samples were exactly the same ones as used in the IMS experiments and kept in a refrigerator for 2 weeks after IMS experiments. 10 l of samples were drop-casted to the silicon chips and analyzed by scanning electron microscopy.

RESULTS
Mass spectra were recorded to determine whether PADK binds to A␤42. The mass spectrum of a 1:10 mixture of A␤42 and PADK is shown in Fig. 1b (see supplemental Figs. S2 and S3 for the results of the 1:1 ratio). In the mass spectrum, there are three peaks corresponding to A␤42 z/n ϭ Ϫ4, Ϫ3, and Ϫ5/2 (z ϭ charge, n ϭ oligomer size), which is similar to the pure A␤42 mass spectrum (Fig. 1a). The peak at m/z ϭ 394 represents pure PADK. Peaks trailing the z/n ϭ Ϫ4 and Ϫ3 peaks of A␤42 labeled with stars represent A␤42-PADK complexes. For z/n ϭ Ϫ4, peaks representing A␤42 monomer with one, two, and three PADK molecules bound are observed, whereas for the z/n ϭ Ϫ3, there are peaks corresponding to A␤42 monomer with one, two, three, and four PADK molecules bound. For z/n ϭ Ϫ5/2, there are two A␤42-PADK complex peaks (m/z ϭ 1884 and 1963), corresponding to A␤42 dimer with one and two PADK molecules bound, respectively. Overall, the mass spectrum clearly shows that the PADK molecule binds directly to A␤42. Moreover, the ratio of the Ϫ3 monomer peak to the Ϫ5/2 FIGURE 1. a, mass spectrum of pure A␤42. b, mass spectrum of a 1:10 mixture of A␤42 and PADK (PADK molecule structure included). A␤42 peaks are represented by triangles, PADK peaks are represented by the circle, and the peaks due to the complex of A␤42 and PADK are represented by stars. z/n is noted for each peak, where z is the charge and n is the oligomeric number. c and d, ATDs for z/n ϭ Ϫ5/2 of A␤42. c, pure A␤42. d, 1:10 mixture of A␤42 and PADK. Oligomeric order (n) is noted for each feature.
peak in the A␤42 sample with PADK, 6.4 (Fig. 1b), is higher than that in the A␤42 sample without PADK, 3.3 (Fig. 1a), which shows quantitatively that the A␤42 monomer concentration is increasing in the solution relative to the aggregate concentration as the PADK relative concentration increases.
The arrival time distributions (ATDs) for the z/n ϭ Ϫ5/2 peaks are given in Fig. 1, c and d. The A␤42 sample without PADK (Fig. 1c) shows four features with arrival times at ϳ720, 680, 600, and 540 s. No other peaks at lower arrival times are observed. These peaks have previously been assigned as the Ϫ5 dimer, Ϫ10 tetramer, Ϫ15 hexamer, and Ϫ30 dodecamer (See Refs. 12 and 13 for a detailed discussion of Ϫ5/2 peak assignment). In Fig. 1d, the ATD of the Ϫ5/2 charge state of A␤42 in the mixture of A␤42 and PADK shows three features, which are assigned as dimer, tetramer, and hexamer based on their measured collision cross-section. Notice that the feature corresponding to the A␤42 dodecamer is absent in the presence of PADK, which indicates that the formation of the dodecamer is inhibited by PADK. No other peaks appear at shorter arrival times, suggesting that A␤42 forms only through hexamer in the presence of PADK. After 2 days, the A␤42/PADK mixture still shows strong signal as compared with pure A␤42 sample, and the ATDs show that there are no large aggregates forming in the solution with PADK. (See supplemental data and supplemental Fig. S1).
To investigate whether PADK also has an effect on removal of preformed dodecamer from solution, PADK was added to a preaggregated A␤42 sample. The ATD of the Ϫ5/2 peak was monitored over time. After several hours of aggregation, the ATD of the Ϫ5/2 peak of A␤42 (Fig. 2a) shows dodecamer, hexamer, tetramer, and dimer features. Concentrated PADK was added to this same sample with 1:10 ratio and monitored over time. After the addition of PADK, the dodecamer feature in the ATD disappears, whereas hexamer, tetramer, and dimer features are maintained (Fig. 2b). At later times, the most abun-FIGURE 2. a-c, ATDs of z/n ϭ Ϫ5/2 of A␤42 in a time-course study of recovery of aggregated A␤42 by PADK. a, A␤42 without PADK on ice for ϳ5 h. b, 5 min after PADK was added to the same aggregated A␤42 sample. c, 3 h after PADK was added to the aggregated A␤42 sample. d, ATD of z/n ϭ Ϫ5/2 of the A␤42-PADK complex (2:1) peak at m/z ϭ 1884. EM images of fibrils of A␤42 samples that were kept in a refrigerator for 2 weeks are shown. e, A␤42 sample without PADK. The dark regions are A␤ fibrils and plaques. f, 1:10 mixture of A␤42 and PADK. Essentially no fibrils are observed. dant feature in the ATD shifts from hexamer to tetramer (Fig.  2c). The disappearance of dodecamer in this recovery experiment indicates that the PADK not only inhibits the formation of dodecamer but also removes preformed dodecamer in the solution. These results also suggest that PADK inhibits hexamer or paranucleus (11,14) formation as well (Scheme 1).
The ATD of the Ϫ5/2 peak corresponding to an A␤42-PADK complex (m/z ϭ 1884) was also measured and is shown in Fig.  2d. (The signal of the m/z ϭ 1963 peak corresponding to z/n ϭ Ϫ5/2 with two PADKs attached was too low to record a reliable ATD; therefore the data are not shown here.) The ATD shows three features corresponding to dimer, tetramer, and hexamer, with one, two, and three PADKs bound, respectively. No dodecamer is present, which is similar to the result of z/n ϭ Ϫ5/2 for A␤42 (Fig. 1d). The result suggests that PADK not only to binds to A␤42 monomer in the solution, but also binds to A␤42 dimer, tetramer, and hexamer and thereby disrupts and reverses the formation of dodecamer and to a lesser degree the hexamer (Scheme 1).
EM was used to elucidate the effect of PADK on fibril formation. The EM images of both samples are given in Fig. 2, e and f. The A␤42 sample shows substantial fibril formation and mass aggregates resembling plaque-like structures (they appear as darker areas in Fig. 2e). For the A␤42 sample with PADK, no large aggregates are found, showing a much cleaner image in Fig. 2f. This indicates that there is less (if any) fibril formation in the solution with PADK, thus supporting our ion mobility data.

DISCUSSION
Mass spectra show that PADK binds to A␤42 monomer and small oligomers directly. IMS-MS reveals that PADK not only inhibits the formation of the A␤42 dodecamer but also removes preformed A␤42 dodecamer in solution (Scheme 1). The ratio of the Ϫ3 monomer peak to the Ϫ5/2 peak increases in the A␤42 sample with PADK, suggesting that the dodecamer is most likely converted into smaller species and eventually monomer. EM images indicate that PADK can also prevent the formation of A␤42 fibrils. Our study is consistent with the study showing that PADK can help reduce the A␤42 levels in the brain of AD transgenic mice and in turn improve their synaptic composition and cognitive function (29). In that AD transgenic mice study, PADK was observed to be a positive modulator that up-regulated the level of proteases, which in turn enhanced the clearance of A␤42 species. On the other hand, our current study probed the direct relationship between PADK and A␤42 and found that there is also a direct interaction involved in the positive effect of PADK on AD pathology. Perhaps the ability of PADK to disaggregate extracellular A␤ peptide leads to efficient uptake of monomers and small oligomers into neurons and microglia, thereby allowing trafficking to lysosomes for degradative detoxification by cysteine proteases (29,30,41). Theoretical studies on the interaction of PADK and A␤42 are underway that will provide more structural information to better understand structure-neurotoxicity correlations of this system. Due to the aromatic nature of PADK, we suspect that perhaps PADK prefers to interact with the N-terminal region of A␤ that has a high content of aromatic residues (fragment 13-20 contains 2 His and 2 Phe residues). Such a binding mode has been suggested in a previous study (42). On the other hand, the hydrophobic nature of PADK might result in binding to the hydrophobic C-terminal region of A␤. Finally, to gain a better understanding of the structure specificity of PADK on A␤42 toxicity, a family of molecules with structures similar to PADK is under study. Preliminary results indicate a wide variety of responses of A␤42 to these molecules and will be reported elsewhere when the study is finished. One EM image is included in supplemental Fig. S4c to indicate that the specific PADK structure is important, not just generic ring systems.