Aβ*56 is a stable oligomer that correlates with age-related memory loss in Tg2576 mice

Amyloid-β (Aβ) oligomers consist of fibrillar and non-fibrillar soluble assemblies of the Aβ peptide. Tg2576 human amyloid precursor protein (APP)-expressing transgenic mice modeling Alzheimer’s disease produce Aβ*56, a non-fibrillar Aβ assembly that has been shown by several groups to relate more closely to memory deficits than plaques. Previous studies did not decipher specific forms of Aβ present in Aβ*56. Here, we confirm and extend the biochemical characterization of Aβ*56. We used anti-Aβ(1-x), anti-Aβ(x-40), and A11 anti-oligomer antibodies in conjunction with western blotting, immunoaffinity purification, and size-exclusion chromatography to probe aqueous brain extracts from Tg2576 mice of different ages. We found that Aβ*56 is a ∼56-kDa, SDS-stable, A11-reactive, non-plaque-related, water-soluble, brain-derived oligomer containing canonical Aβ(1-40) that correlates with age-related memory loss. The unusual stability of this high molecular-weight oligomer renders it an attractive candidate for studying relationships between molecular structure and effects on brain function.

These results indicate that a ~56-kDa entity that is not artificially formed from monomers or lower molecular-weight Aβ species exists, and that its mass is similar whether measured by SEC or SDS-PAGE. The ~14-kDa band may represent a trimeric form of Aβ that weakly binds to other trimers to form a hexamer. The constituents in the D8Q7I-reactive, ~56-kDa doublets are unclear; it is possible that they reflect heterogeneity in the N-termini of the constituent Aβ species that assemble to form hetero-oligomers.
Non-denatured, captured proteins (exposed to TBS only) probed with 82E1 antibodies revealed not only a ~56-kDa band, but also bands at ~40 kDa, ~35 kDa, and ~12 kDa, in addition to monomers (Fig. 4). All four bands disappeared after the captured proteins were denatured using urea or GuHCl, indicating that the proteins in these bands are not covalently linked (Fig. 4). We cannot exclude the possibility that a macromolecule is non-covalently bound to Aβ  in a

Discussion
In this paper, we show that Aβ*56 is a ~56-kDa, A11-reactive, SDS-stable, water-soluble, non-plaque-related, brain-derived oligomer containing canonical Aβ(1-40) that correlates with age-related memory loss in Tg2576 mice. Aβ*56 is the only consistently present, high molecularweight, SDS-stable, brain-derived oligomer that we were able to detect using the current experimental settings. We found that Aβ*56 can be sufficiently stable such that it does not undergo noticeable changes in conformation or size when boiled in reducing agents and exposed to SDS.
To our knowledge, no other high molecular-weight Aβ oligomers exhibiting this degree of stability have been described.
The term Aβ* pays homage to Charles Weissmann's moniker for a hypothetical form of the prion protein (PrP) that is "an infectious entity [which] could be a subspecies of PrP Sc or a different modification of PrP altogether (which one might call PrP*)" 35 . Conceptually an essential component of the prion, PrP* is by definition neurotoxic and could potentially be involved in the replication of prions 36 . More recent studies indicate, however, that prion neurotoxicity and infectivity can be dissociated 37 . In keeping with PrP*, Aβ*56 appears to be neurotoxic and associates closely with impaired memory 1,17,18,20,21,24,26,28,29,38 , but whether it is involved in the replication, aggregation, or propagation of Aβ is unclear. Aβ*56 is an A11-reactive oligomer, but has not yet been shown to behave like synthetic, A11-reactive, "prefibrillar" or α-sheet-binding Aβ oligomers in solution, which evolve to become A11-negative, fibrillar oligomers and fibrils 39,40 . To explain the inverse relationships between insoluble Aβ and memory function that While many groups have successfully studied Aβ*56 in mice 1,17,20,24,26,28,29 , aging dogs 18 , and humans diagnosed with AD 21,38 , detecting Aβ*56 can be challenging because it is rare (~1 part per million by mass). We have found that high background and non-specific binding are the greatest impediments to accurate detection. In the current studies, we have modified the original protocol developed 20 years ago to reduce non-specific signals. Using this protocol, an independent laboratory detected Aβ*56 without difficulty (Fig. S8). It is worth mentioning a few examples of the kinds of optimized experimental parameters that enhance the sensitivity and specificity of signals sufficiently to detect Aβ*56 reliably. We adjusted the buffers used to extract proteins, because we found that extracts containing detergents form irreversible, insoluble precipitates and may contain membrane-associated proteins. We found that tricine in gels accentuates the compactness of the Aβ*56 band; the corresponding bands in gels lacking tricine are more diffuse and less intense. The use of monoclonal antibodies that have been purified using protein A to capture proteins obscures Aβ*56 due to a non-specific band at ~50 kDa caused by protein A contaminants in the precipitated proteins 41 . The low concentrations of Aβ*56, which comprise less than 1% of total soluble Aβ in 20 month Tg2576 mice 2 , necessitate taking advantage of the greater sensitivity of biotin-avidin coupled to chemiluminescence compared to colorimetric methods. It is important to beware that some batches of A11 do not detect Aβ*56 (Fig. S7).
We surmise, based on comparing banding patterns in our current western blots with those in the literature, that the Aβ*56 entity we describe here overlaps with but may not correspond exactly to the ~56-kDa entities that have been identified using other protocols. Technical refinements permit more accurate characterizations of Aβ assemblies. For example, the constituents of highly neurotoxic entities in 7PA2 culture medium originally believed to be dimers and trimers 42 were found upon later investigation also to contain N-terminally extended, noncanonical Aβ 43 . While our current protocol reliably and reproducibly detects Aβ*56, we anticipate that continued, methodological advancements will enable future investigations of Aβ*56 to be conducted more efficiently.
Not all lines of mice produce Aβ*56. The levels and species of Aβ generated in a particular line of mice affect the expression of Aβ*56. For example, Tg2576 mice expressing APP-Swe express Aβ*56, but rTg9191 mice expressing APP-Swe with the V717I London mutation which promotes fibril formation do not 2,30 . The balance between aggregation processes that do (onpathway) and do not (off-pathway) culminate in fibrils may influence the production of Aβ*56, as the two pathways may compete for monomers in a common pool.
The structure, spatial distribution, and temporal expression of Aβ oligomers are important determinants of their effects on the brain 2,44 . Two distinctive spatial and temporal features of Aβ*56 are that it forms before dense-core plaques and is present within but not confined to densecore plaques 1,2 . This is significant, because in mice the oligomers that are confined to dense-core plaques do not impair neurological function unless they are burst open and dispersed in the brain 2 .
Although the precise structure of Aβ*56 is unknown, some general features may be deduced from binding studies using the polyclonal, conformational antibodies, OC and A11 4,7 . OC antibodies, which recognize in-register, parallel β-sheets, and to a lesser extent antiparallel β-sheets, do not bind Aβ*56 2 . A11 antibodies, which recognize a variety of conformational epitopes, bind Aβ*56 in mice 1,20 (and this paper), dogs 18 , and humans 21,45 , but the specific epitope(s) recognized is unknown. Because A11 antibodies are polyclonal, more than one A11-reactive variant may be present, complicating the ability to decipher the structural properties of Aβ*56. Currently, there are no monoclonal, conformational antibodies that selectively bind Aβ*56.
In summary, in this paper, we confirmed and extended the characterization of a ~56-kDa, A11-binding, SDS-stable, water-soluble, non-plaque-related, brain-derived oligomer containing canonical Aβ , also known as Aβ*56. We showed that its presence correlates with age-related memory loss in Tg2576 mice. We provide detailed information about the techniques, tools, and specimens that enable Aβ*56 to be detected reliably and reproducibly.

Limitations of the Study
These studies are limited to the Tg2576 mouse model. It is possible that other mouse models express different, SDS-stable, Aβ oligomers, and that Aβ*56 in other mouse models may contain other forms of Aβ, including Aβ(1-42).
Although we did not observe any SDS-stable Aβ oligomers larger than 100 kDa, we cannot exclude their existence because our experimental protocol is optimized to detect oligomers smaller than 100 kDa. Optimized experimental parameters are needed to investigate the existence of larger, SDS-stable oligomers.
While the current studies are limited to mice, it is likely that Aβ*56 is present in humans.
One group previously reported detecting higher levels of A11-reactive, ~56-kDa entities in nasal fluid from humans with AD 21,38 . We previously described an isolated cluster of ~56-kDa, 82E1reactive bands on a western blot of immunoprecipitated Aβ(x-40)/Aβ(x-42) proteins from human cerebrospinal fluid (CSF) 46 . When the immunoprecipitated CSF proteins were probed using 6E10 antibodies, however, non-oligomeric, APP fragments were also detected 46 , illustrating the importance of employing antibodies that specifically detect canonical Aβ. It will be informative to use our current protocols to examine Aβ*56 levels in AD, since an earlier study using 6E10 antibodies reported paradoxically lower levels of Aβ*56 in brain tissue from AD patients 45 .

Author contributions
P.L. conceived the study, designed and performed biochemistry experiments, analyzed data, and wrote the manuscript. I.P.L. and S.L.S. helped design and perform biochemistry experiments, and analyzed data. L.J.K. analyzed archived behavioral data. K.H.A. conceived the study, analyzed data, and wrote the manuscript.

Acknowledgements
Behavioral studies were supported using funds from NIH R01-NS33249. Biochemistry studies were supported using funds from the N. Bud Grossman Center for Memory Research and Care. The authors thank Dr. Swathy Babu and Dr. Shauna Yuan for contributing Fig. S8.

Declaration of Interests
The authors declare no competing interests. The percentage of Tg2576 mice exhibiting impaired spatial reference memory increases with age.
The percentage of impaired 2-5 month Tg2576 mice (+) is similar to that of 2-16 month, nontransgenic littermates (-), representing baseline-deficits. This analysis was performed using previously described data acquired using a water maze. 32 Impaired spatial reference memory is defined here as a mean probe score (MPS) <35%. 32 The sample size for each age period is shown.

Antibodies
Basic information (Table S2) and detailed usage (Table S3) of all antibodies used in this paper were shown.

Brain protein extraction
Detergent-free, water-soluble brain protein extracts were prepared using a protocol adapted from a previous publication. 22 Briefly, each frozen hemi-forebrain free of olfactory bulb and cerebellum was weighed and then transferred to 5 volumes (i.e., 5 mL per g of wet brain tissue) of ice-cold To analyze biological samples, detergent-free, water-soluble brain protein extracts of Tg2576 mice at 10-11 months of age prior to the appearance of ThioS-reactive, dense-core plaques in the forebrain 31 were pooled from 5 mice (Tables S1 and S3), each contributing 20% protein mass. The prepared sample was injected at the dose of 2 mg in 250 µL. Proteins were eluted, and the elution chromatogram was determined as described above. Immediately after the elution was complete, the microplate that collected the fractions was stored on ice. Each of the 96 fractions was transferred into a 1.5-mL sterilized microcentrifuge tube that contained 2.6 mM PMSF, 5.2 mM phen, 2.6% (v/v) protease inhibitor cocktail, 2.6% (v/v) phosphatase inhibitor cocktail A, and 2.6% (v/v) phosphatase inhibitor cocktail 2 in 10 µL of PBS. After mixing with an SEC fraction, the final concentrations of the five types of inhibitors were 0.1 mM, 0.2 mM, 0.1% (v/v), 0.1% (v/v), and 0.1% (v/v), respectively. Each inhibitor-containing fraction was evenly split into two parts, snap-frozen on dry ice, and stored at -80°C.

Immunoprecipitation (IP)
IP was performed according to previously published procedures. 2 Briefly, brain extracts containing 200-400 µg of total proteins (Tables S1 and S3) were incubated with capturing reagents (Tables S1 and S3) and Protein G-coated matrix (Protein G Sepharose 4 Fast Flow resin (GE healthcare) or Dynabeads Protein G (Thermo Fisher Scientific); Tables S1 and S3) at 4°C for 14-16 hr.

Western blotting (WB)
WB was performed based on previously published procedures. 2,28 Detailed usage of brain protein extracts and detecting reagents are listed in Table S3.
Next, each protein-blotted membrane was transferred to 50 mL of PBS of room temperature, and the membrane was subjected to microwave heating under full power first for 25 sec and then for 15 sec with a 4-min cooling period after each episode of heating.
Next, primary antibodies were then added directly to blocking buffer (see Tables S2 and   S3 for the use of detecting antibodies in detail), and membranes were then incubated at 70 rpm, 4°C for 14-16 hr (For Figs. S5 and S6, membranes were incubated directly with blocking buffer at 70 rpm, 4°C for 14-16 hr.).
Membranes were then incubated with the SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific) at 200 rpm, room temperature for 5 min. Chemiluminescence signals were developed using the Kodak Scientific Imaging film X-OMAT Blue XB (PerkinElmer Life Sciences, Waltham, MA). Intensities of immunoreactive protein bands were determined densitometrically using Optiquant version 03.00 (Packard Instrument, Fallbrook, CA).
Alternatively, signals were revealed using the ChemiDoc MP Imaging System (Bio-Rad). Image Lab version 6.1 (Bio-Rad) was used to determine protein band intensities.
To determine whether the ~56-kDa, Aβ-containing entity was present in each studied animal, the band total intensity (i.e., the product of mean intensity per pixel and numbers of pixels in the designated band area) of the ~56-kDa, Aβ-containing entity was first measured. Next, the mean and standard deviation (SD) of background intensities were determined from the measured total intensities of four areas, each of which had an equal size to the size of the designated band area of the ~56-kDa, Aβ-containing entity. The four areas were sampled at 1) the ~56-kDa in the lane containing immunoprecipitated entities from non-transgenic mice, 2) the ~4.5-kDa (corresponding to the migrating area of monomeric Aβ) in the lane containing immunoprecipitated entities from non-transgenic mice, 3) the ~56-kDa in the lane containing only capturing antibodies but no immunoprecipitated entities from mice, and 4) the ~4.5-kDa in the lane containing only capturing antibodies but no immunoprecipitated entities from mice. The ~56-kDa, Aβ-containing entity was considered detected (i.e., present) in the brain of an animal if its band total intensity is equal to or higher than three SDs above the mean background intensity.
The experimental settings for each (IP)/WB reaction are shown in Table S3.

Denaturant treatment
Three hundred and seventy-five µg of mouse brain extracts (Tables S1 and S3) were completely dried but not over-dried at room temperature using a speed vacuum concentrator (Eppendorf,

Immunoaffinity purification
Brain protein extracts (Tables S1 and S3) of 15-21 month Tg2576 or age-matched non-transgenic mice were incubated with D8Q7I-bound Dynabeads Protein G matrix (Table S3) under a gently rotating mode (15 rpm) at 4°C for 20 hr. The immunocomplex-matrix was then washed twice with 1 mL of wash buffer 5 (TBS plus 1% (w/v) n-Octyl β-D-thioglucopyranoside (OTG)) under a gently rotating mode (15 rpm) at 4°C for 5 min. Immunoprecipitated entities were eluted three times using 50 μL of elution buffer 2 (Pierce IgG Elution Buffer, pH ~3 (Thermo Fisher Scientific); plus 1% (w/v) OTG) by agitating at 1,200 rpm, 25°C for 5 min. Each time immediately after the elution, the pH of the eluted material was neutralized to pH between 7 and 8 using a neutralization solution (1 M Tris-base (pH ~10.5)). The eluted materials were stored at -80°C.

Behavioral testing
We analyzed archived data generated as described in Westerman et al. 32 Briefly, we tested spatial reference learning and memory using a version of the conventional Morris water maze. 48 The water maze was a circular 1-or 1.2-m pool filled with water at 25-27°C and made opaque by the addition of nontoxic white paint. The pool was placed amid fixed spatial cues consisting of boldly patterned curtains and shelves containing distinct objects. Mice first underwent visible platform training for 3 consecutive days (eight trials per day), swimming to a raised platform (a square surface 12 x 12 cm 2 ) marked with a black and white striped pole. Hidden platform training was conducted over 9 consecutive days (four trials per day), wherein mice were allowed to search for a platform submerged 1.5 cm beneath the surface of the water. At the beginning of the 4 th , 7 th , and 10 th day of hidden platform training, a probe trial was conducted in which the platform was removed from the pool, and mice were allowed to search for the platform for 60 sec. All trials were monitored by a camera mounted directly above the pool, and were recorded and analyzed using a computerized tracking system (HVS image, Hampton UK). Further analysis was done using Wintrack (kindly provided by Dr. David Wolfer, University of Zurich, Switzerland). The mean probe score (MPS) is the arithmetic average of the %-time spent in the target quadrant in the three probe trials.

Key resource tables may be found in the Supplemental Information
Table S1 Mice used to detect the ~56-kDa, Aβ-containing entities