Transglutaminase Induces Protofibril-like Amyloid β-Protein Assemblies That Are Protease-resistant and Inhibit Long-term Potentiation*

An increasing body of evidence suggests that soluble assemblies of amyloid β-protein (Aβ) play an important role in the initiation of Alzheimer disease (AD). In vitro studies have found that synthetic Aβ can form soluble aggregates through self-assembly, but this process requires Aβ concentrations 100- to 1000-fold greater than physiological levels. Tissue transglutaminase (TGase) has been implicated in neurodegeneration and can cross-link Aβ. Here we show that TGase induces rapid aggregation of Aβ within 0.5-30 min, which was not observed with chemical cross-linkers. Both Aβ40 and Aβ42 are good substrates for TGase but show different aggregation patterns. Guinea pig and human TGase induced similar Aβ aggregation patterns, and oligomerization was observed with Aβ40 concentrations as low as 50 nm. The formed Aβ40 species range from 5 to 6 nm spheres to curvilinear structures of the same width, but up to 100 nm in length, that resemble the previously described self-assembled Aβ protofibrils. TGase-induced Aβ40 assemblies are resistant to a 1-h incubation with either neprilysin or insulin degrading enzyme, whereas the monomer is rapidly degraded by both proteases. In support of these species being pathological, TGase-induced Aβ40 assemblies (100 nm) inhibited long term potentiation recorded in the CA1 region of mouse hippocampus slices. Our data suggest that TGase can contribute to AD by initiating Aβ oligomerization and aggregation at physiological levels, by reducing the clearance of Aβ due to the generation of protease-resistant Aβ species, and by forming Aβ assemblies that inhibit processes involved in memory and learning. Our data suggest that TGase might constitute a specific therapeutic target for slowing or blocking the progression of AD.

Oligomerization and aggregation of the amyloid ␤-protein (A␤) 2 are thought to comprise a central mechanism in the initiation and progression of Alzheimer disease (AD). A␤ was first implicated in AD when it was found to be the major protein in amyloid plaques, one of the histopathological hallmarks of AD, in which A␤ exists as fibrils. In support of the involvement of A␤ in AD, early studies investigating the neurotoxicity of A␤ found that longer forms of A␤ (e.g. A␤42) aggregated faster than shorter ones (1) and that aggregation was essential for neurotoxicity (2). For self-aggregation to occur in vitro, micromolar A␤ concentrations are required (3). These A␤ concentrations far exceed the physiological levels, even in the AD brain, which are in the low nanomolar range (4 -6). This discrepancy suggested that a pure self-assembly mechanism may not account for oligomerization and aggregation in vivo.
Tissue transglutaminase (TGase) is a complex protein with multiple functions, including serine kinase activity, G protein signaling, and the catalytic capability to cross-link proteins between lysine and glutamine residues, forming a covalent isodipeptide bond (7). TGase occurs abundantly in the brain and has been implicated in neurodegeneration (8,9). A␤ contains the necessary lysine and glutamine residues for TGase-catalyzed cross-linking and has previously been shown to be a substrate for TGase (10 -12). Thus, TGase may be a likely candidate to play a role in the pathophysiology of AD.
Here we found that TGase induces monomeric A␤ to rapidly form oligomers and aggregates in a time-and concentrationdependent manner. Most importantly, we show that TGase lowers the concentration for A␤ oligomerization so it can occur at physiological A␤ levels and that it induces A␤ species that are resistant to degradation by metalloproteases. Additionally, EM analysis showed that the TGase-induced A␤ assemblies have the same morphologies as the previously described protofibrils that result from self-aggregation (13,14). The importance of these findings is supported by the ability of the TGase-induced A␤ assemblies to attenuate long term potentiation (LTP), a process thought to be affected in the early stages of AD. Thus, the TGase-induced oligomerization and aggregation of A␤ provides a mechanism that could explain both the formation and the persistence of pathologically active A␤ species in vivo.

EXPERIMENTAL PROCEDURES
Chemicals-Unless noted otherwise, chemicals were purchased from Sigma and Invitrogen. Insulin-degrading enzyme was a gift from Dr. Malcolm Leissring. Human neprilysin was purchased from R&D Systems (Minneapolis, MN), 6E10 antibody from Signet Laboratories (Dedham, MA), and anti-TGase antibody ab2386 from Abcam (Cambridge, MA).
A␤ Peptide Preparation-Lyophilized A␤ peptides were obtained from multiple sources as a trifluoroacetic salt (Biopolymer Facility, Brigham and Women's Hospital, Anaspec, San Jose, CA, California Peptide, Napa, CA). A␤ was prepared for TGase-mediated cross-linking reactions by initially dissolving the peptide in 100% DMSO and then diluting the solution with water and HEPES buffer (final concentration, 232 M A␤ in 5% DMSO and 25 mM HEPES, pH 7.6).
Size Exclusion Chromatography-Monomeric A␤ was purified, and A␤ assemblies were characterized by size exclusion chromatography (SEC) using either a Superdex 75 or Superose 6 column (10 ϫ 300 mm). The Superdex 75 column was initially used to separate low molecular weight A␤ from protofibrils as previously described (14). The Superose 6 column was used to separate a larger range of A␤ aggregates. For a subset of experiments, A␤40 (1 mg/ml) was denatured in 6 M guanidine HCl prior to SEC purification. Samples were run at a flow rate of 0.5 ml/min, and protein was monitored at wavelengths of 210 and 280 nm. Both columns were run isocratically using a buffer containing 70 mM NaCl and 5 mM Hepes, pH 7.5.
Protein Assay-The Bradford protein assay (Bio-Rad) was modified by adding a 10-min incubation of the samples in 4 M urea at room temperature, before dye addition.
Thioflavin T Binding-Thioflavin T dye (10 M) in 100 mM glycine buffer, pH 8.5, was mixed with an equal amount of sample (50 l of each for 96-well assays), incubated for 10 min at room temperature in the dark, and read out at 446 ex and 490 em (15).
Cross-linking Reactions-Tissue transglutaminase (TGase, or TGase 2, EC 2.3.2.13) reactions in 70 mM NaCl, 2.5 mM CaCl 2 , 40 mM HEPES, pH 7.5, 1 mM dithiothreitol were initiated by adding A␤40. The reactions were incubated for the times and at the temperatures specified for the various experiments. Reactions were stopped either by the addition of EDTA (50 mM final concentration) or the addition of a TGase inhibitor (LDDN-80042, courtesy of Laboratory of Neurodegeneration and Drug Discovery, Brigham and Women's Hospital, Boston) at 10 -20 M. In some experiments the A␤ peptide was concentrated using a 3-kDa cutoff spin filter in the presence of 50 mM EDTA and 0.11 mM SDS. Because A␤42 has a higher propensity to self-aggregate, the experiments described above were performed with A␤40 to keep self-aggregation at a minimum, because it might confound the kinetics of TGase-induced aggregation (i.e. cross-linking of monomer versus cross-linking of pre-formed oligomers or aggregates). Guinea pig TGase purified from liver was purchased from Sigma-Aldrich. Recombinant human TGase expressed in Escherichia coli (16, 17) was a gift from Dr. Ross Stein. To allow direct comparison of the ability of human and guinea pig TGase to induce oligomerization of A␤40, the concentrations of active sites (micromolar) of both enzyme preparations were determined by the Case and Stein (16) method and were found to be 13.5 and 40 M, respectively. Unless noted otherwise, all experiments were carried out with guinea pig TGase.
Electron Microscopy-A␤ was purified by SEC using the guanidine HCl and Superose 6 column method (see above). A common stock of fresh A␤ was prepared, and aliquots were sampled at various time points for EM analysis. The samples were prepared by conventional negative staining with 0.75% uranyl formate as described previously (19). Reactions were considered to be terminated at the time when excess peptide was washed from the grid, which was ϳ25 s into the EM grid preparation procedure. Grids were examined with a FEI Tecnai 12 electron microscope operated at an acceleration voltage of 120 kV. Images were recorded at a magnification of 52,000ϫ and a defocus of ϳϪ1.5 m using a Gatan 2k ϫ 2k charge-coupled device camera.
Protease Degradation-Aggregated A␤ was prepared by mixing A␤ with TGase (see above) and stopped with the TGase inhibitor LDDN-80042 (10 M). Degradation was evaluated with 50 -100 nM protease for different time periods at 37°C and halted with 1 mM 1,10-phenanthroline. Samples were analyzed by Western blotting using the 6E10 anti-A␤ antibody.
Electrophysiological Recordings-Mouse transverse hippocampal slices were prepared and field excitatory postsynaptic potentials (fEPSP) were recorded in the CA1 region of the hippocampus as described previously (20). To induce LTP, two consecutive trains (1 s) of stimuli at 100 Hz separated by 20 s were applied to the Schaffer collaterals.
Traces were analyzed using the LTP Program (21). The fEPSP magnitude was measured using the initial fEPSP slope, and three consecutive responses (1 min) were averaged and normalized to the mean value recorded 20 -30 min into the baseline recording, but before tetanus application. Data were pooled across animals of the same age and are presented as mean Ϯ S.E. LTP was evaluated by measuring fEPSPs 60 min after the tetanus application. The following statistical analysis was carried out: the same time window samples of the control and drug administrations were compared using paired, twotailed Student's t test. In the case of multiple comparisons, samples were analyzed by one-way analysis of variance. Student-Newman-Keuls post-hoc tests were used to examine the significance of multiple pairwise comparisons.

RESULTS
Amyloid-␤ protein (A␤) undergoes self-aggregation in vitro above a critical concentration, which has been reported as 5-100 M depending on whether A␤40 or A␤42 was used (22,23). This process requires hours to days to occur, although aggregation can be accelerated with higher concentrations of A␤. Physiological concentrations of A␤ are significantly lower, in the pico-to nanomolar range, posing the conundrum as to what in vivo factors allow A␤ to oligomerize or aggregate at lower concentrations. Our experiments show that TGase causes rapid, time-dependent oligomerization of A␤40. TGase (6.25 ϫ 10 Ϫ3 units/ml) was added to an A␤40 solution (25 M) at room temperature, aliquots were taken at 0, 1, 5, 15, 30, and 60 min, and the reaction was stopped with EDTA. The aliquots were resolved on an SDS-PAGE gel, illustrating the rapid kinetics of the TGase-induced oligomerization of A␤ (Fig. 1A). This rapid assembly appears as a laddering of A␤ that represents the incremental addition of A␤ monomers into oligomers that develop into larger assemblies and eventually grow too large to enter the gel. This time-course experiment shows that TGase induces rapid oligomerization of A␤ under these specific conditions, reaching ϳ50% activity in 5 min (Fig. 1B). TGase activity was inhibited in the presence of the calcium chelator EDTA (5 mM) (Fig.  1, A and B). It is important to point out that self-assembled A␤ are SDSlabile, particularly at elevated temperatures. Boiling the reaction products of self-assembled A␤ briefly prior to electrophoresis dissociates most of the self-aggregated A␤ species (data not shown). By contrast, A␤ species formed in the presence of TGase remain intact upon boiling.
The oligomerization of A␤ by TGase can be observed by SEC as the substrate-to-product ratio. At time 0 most of A␤40 is in a low molecular weight form (predominantly monomeric A␤40) and elutes at 31 min, but after a 30-min incubation with TGase, A␤40 is converted into larger assemblies as observed by an increase in the void peak that elutes at 17 min (Fig. 1C, dotted line). In contrast, A␤40 incubated for 30 min without TGase shows little conversion of the low molecular weight A␤40 peak into larger assemblies (Fig.  1C, solid line). In support of TGase being responsible for the fast A␤ oligomerization kinetics, addition of the recently developed TGase inhibitor LDDN-80042 (24) at a concentration of 2.5 M resulted in almost complete inhibition of A␤ oligomerization (Fig. 1, D and E). Additional support for TGase driving this process is the lack of aggregation observed with a mutant A␤40 peptide. Wild-type A␤40 showed rapid assembly formation in the presence (Fig. 1F, lane 1) but not in the absence of TGase (Fig. 1F, lane 2). To show the specificity of this reaction, the two lysine residues in A␤40 were mutated to alanine, because TGase requires a lysine for the formation of an ⑀-(␥glutamyl)lysine isopeptide bond (7). Little to no oligomerization was observed with the mutant A␤40 irrespective of the presence (Fig. 1F, lane 3) or absence of TGase (Fig. 1F, lane 4). Interestingly, the single remaining cross-link site, a glutamine In the "EDTA" sample, the chelator was added before TGase. B, densitometry of Western blots such as the one in panel A shows that TGase induces rapid aggregation of A␤. For the quantification of the immunoblot, the total signal due to oligomerization (Ͼ4 kDa) in each lane was measured and normalized to the signal at T ϭ 0, which was given a value of 1. Each data point represents the mean of three separate experiments (ϮS.E.). C, a Superdex 75 size exclusion column was used to show the conversion from substrate (low molecular weight A␤) to product (aggregated A␤) induced by TGase addition. A␤40 alone showed little conversion of LWM A␤ (31-min peak) (solid line), whereas in the sample containing TGase the majority of A␤ was converted into aggregates (19-min peak) (dotted line). D, the TGase inhibitor LDDN-80042 was highly effective in attenuating the aggregation of A␤40, blocking aggregation at concentrations ranging from 0.1 to 10 M. The "No TGase" sample was run under identical conditions, but without TGase. E, densitometry of Western blots such as the one in panel D shows TGase inhibition with LDDN-80042. For this analysis, the signal from the bands at 8 kDa and higher were measured and normalized against the signal from the same area measured in the sample lacking TGase, which was given a value of 1 (n ϭ 3, ϮS.E.). F, a mutant A␤40 peptide (i.e. A␤40 with its two lysines mutated to alanine ϭ mut) did not form aggregates in the presence of TGase (lane 1, A␤40 ϩ TGase; lane 2, A␤40 only; lane 3, mutant A␤40 ϩ TGase; lane 4, mutant A␤40 only). The asterisk indicates A␤40 bound to TGase. residue in the mutant peptide, is available to react with TGase, as suggested by a weak 6E10 band near the molecular weight of TGase (76.6 kDa) (Fig. 1F, lane 3, asterisk). The presence of TGase in this band was confirmed by reprobing the same blot with anti-TGase antibodies (data not shown). These results suggest that TGase can specifically catalyze the oligomerization of A␤40 and supports the previous finding that A␤ is a substrate for the enzyme (10 -12).
To investigate the specificity of TGase-mediated cross-linking, we compared enzymatic with chemical cross-linking of A␤. In Fig. 2A, cross-linking of three different substrates, glyceral-dehyde-3-phosphate dehydrogenase (G3PD), aldolase, and A␤, were compared. All proteins were separated using SDS-PAGE and visualized using SYPRO Red dye. G3PD and aldolase have previously been shown to multimerize in the presence of DMS and were used as positive controls (18). The addition of DMS to either protein induced higher molecular weight assemblies ( Fig. 2A, lanes 2 and 4). In contrast, little to no oligomerization was observed with A␤ incubated with DMS ( Fig. 2A, lane 6). Similar results were observed with all three proteins using the chemical crosslinker dimethyl adipimidate (data not shown). The finding that TGase induces robust oligomerization of A␤ ( Fig. 2A, lane 12) but fails to cross-link G3PD or aldolase ( Fig.  2A, lanes 7-10) suggests that crosslinking of A␤ by TGase is a specific reaction. Similar results were obtained when A␤ cross-linking was analyzed by SEC. A␤ treated with DMS and separated in 2 M guanidine HCl (to disassemble non-cross-linked A␤) showed predominately monomeric A␤ (Fig. 2B). However, A␤ treated with TGase showed numerous peaks between 15 and 24 min (Fig. 2C), that presumably represent the dimeric, trimeric, and tetrameric A␤ assemblies observed on the Western blots. These data thus support the notion that TGase interacts with A␤ in a specific manner to promote oligomerization.
Because AD is specific to humans, a comparison was conducted between guinea pig and human TGase to validate the possibility of a role for this enzyme in the disease. Analysis of mouse and human TGase with calculated molecular weights of 76,699 and 77,253, respectively, showed an 84% homology at the protein level (25). Moreover, the active catalytic site is highly conserved in TGases from invertebrates to humans (26). Accordingly, a series of reactions conducted using guinea pig and human TGase showed similar oligomerization patterns by Western blotting (data not shown).
The molarity of the active sites was determined for the human and the guinea pig preparations according to Case and Stein (16) and was found to be 13.5 M and 40 M, respectively. The amount of oligomerization observed by Western blotting (data not shown) was plotted against the amount of catalytic activity for both guinea pig (circles) and human (triangles) TGase (Fig. 3). The results show that A␤ is an equally good substrate for both the human and guinea pig transglutaminase. This conclusion is further supported by the almost identical enzymatic activity previously observed for these two enzymes (16,27).
To further study the potential role of TGase in AD, experiments were conducted to determine if another important, disease-related A␤ species, A␤42, is also a substrate for TGase.  To directly compare the enzymatic activity of guinea pig and human TGase, the amount of catalytic sites in the two preparations was plotted against the substrate conversion as measured by densitometry of the Western blots (data not shown). The results showed that the two enzymes catalyze reactions with very similar kinetics (n ϭ 3, ϮS.E.).
The possibility of A␤42 being more pertinent to the pathophysiology of AD has been discussed, both because of its greater production relative to A␤40 associated with familial AD mutations and because of its greater propensity to aggregate (28). The monomer is the predominant form of A␤ peptides observed without TGase (Fig. 4, lanes 1 and 3), with A␤42 showing a slight tendency for self-assembly (Fig. 4, lane 3). In contrast, SDS-PAGE and immunoblotting showed oligomeric laddering for both A␤40 and A␤42 in the presence of TGase (Fig. 4, lanes 2 and 4). Even though TGase induced both peptides to rapidly oligomerize, there is a qualitative difference in the banding patterns representing the smaller oligomeric species (Fig. 4, lane 2 versus 4), whereas the high molecular weight bands appeared identical. This difference may relate to A␤40 and A␤42 being assembled through different pathways (29). Nonetheless, these results demonstrate that TGase recognizes both A␤42 and A␤40 as substrates.
To assess whether TGase catalyzes A␤ oligomerization at physiological A␤ concentrations, experiments were conducted with A␤40 concentrations from the low nanomolar to the low micromolar range (Fig. 5). It was possible to detect TGase-induced oligomers of A␤ by our standard Western blotting procedure at A␤ concentrations of 500 and 1000 nM (Fig. 5A, lane  3 and 4, respectively), ϳ10to 20-fold lower than the concentrations needed for A␤ self-assembly to normally occur (Fig.  5A, lane 1 1, 3, and 4 were specific to A␤ and not due to the presence of TGase. TGase-induced oligomerization of A␤ may occur at even lower concentrations, but at A␤ concentrations below 500 nM protein bands were difficult to detect by Western blotting. To determine whether TGase-induced oligomerization of A␤ could occur near physiological conditions, the A␤40 concentration was lowered to 50 nM, but the reaction volume was increased to accommodate subsequent concentration of the sample to a level that was detectable by Western blotting. A␤40 was incubated for 24 h with and without TGase (1.1 ϫ 10 Ϫ3 units/ml). The reactions were stopped with a final concentration of 50 mM EDTA and 0.11 mM SDS. SDS was critical for recovering low concentrations of A␤ by reducing A␤ loss due to binding to surfaces. The samples were then concentrated 1000fold by spin filtration using a 3-kDa cutoff filter, limiting the loss of monomeric A␤. After concentrating, samples were boiled to dissociate non-cross-linked oligomers formed by self-aggregation during the concentration process. Western blotting revealed that A␤ assemblies occurred only in the presence (Fig.  5B, lane 2) but not in the absence of TGase (Fig. 5B, lane 1). Bands could be detected in the sample containing TGase (see arrows in Fig. 2B, ϳ8 and 12 kDa) that were not present in the sample without TGase (i.e. 12-kDa band) or only in lower amounts (i.e. 8-kDa band) (Fig. 2B, lane 1). An additional large smear centered near 76 kDa (asterisk) was present in the concentrated sample containing TGase (Fig. 5B, lane 2), which was identified as TGase by anti-TGase antibodies in samples containing TGase and A␤ (Fig. 5C, lane 2) or TGase alone (Fig. 5C,  lane 3). In these reactions TGase was not lowered proportionally to the decrease in A␤, producing a higher molar ratio of TGase to A␤ than in standard reactions, thus explaining the prominent TGase bands on the Western blot after concentrating the samples. These high molecular weight bands also appeared in lanes containing A␤ and TGase developed with the anti-A␤ antibody (Fig. 5B, lane 2), but not in lanes containing only A␤ or only TGase (Fig. 5B, lanes 1 and 3, respectively), indicating that A␤ was bound to TGase. The formation of the 8-kDa, 12-kDa, and high molecular weight bands was compared in samples containing only A␤ and A␤ treated with TGase and quantified in Fig. 5D. These experiments indicate that TGase can induce A␤ oligomerization at physiologically relevant A␤ concentrations.
We used EM to visualize the TGase-induced A␤ assemblies, allowing for a comparison with previous EM studies of selfaggregating synthetic A␤ (14,23,30,31). A␤40 was denatured and purified by SEC prior to the addition of TGase. No structures were observed with fresh, SEC purified A␤ (data not shown). In Fig. 6, SEC purified A␤ (17 M) was incubated with TGase (4 ϫ 10 Ϫ3 units/ml) at 30°C for the times specified in the panels (2.5 min to 68 h). Very few structures were observed before 10 -20 min, except for an occasional spherical particle of ϳ12 nm in diameter that was also seen in the sample containing only TGase (Fig. 6K). Western blot analysis suggested that smaller A␤ assemblies are present in the sample at time points before 10 min, but they seem to be too small to be visualized by EM (Fig. 6, A-C). After 10 min, structures appeared to be forming but were still too small to be clearly identified in EM images. Approximately 20 min after addition of TGase to A␤, 5 nm spherical structures appeared (Fig. 6D). The exact time when these structures began to occur varied from experiment to experiment, but the observed morphologies of the A␤ species and their progression (Fig. 6, A-E and G-I) were consistent. By 30 min, the particles began to elongate, showing A␤ species that were ϳ6 -8 nm in width, and up to 28 nm in length (Fig. 6E). The elongation continued with time (Fig. 6, G-I), whereas the width remained at ϳ6 -8 nm. The structures that formed in the presence of TGase were almost identical to the previously described protofibrils that formed during self-assembly of synthetic A␤40 incubated for 68 h at room temperature but at a concentration ϳ25 times greater (500 M) (14,23,30,31). The inset in Fig. 6L shows the heterogeneous size distribution of the formed protofibrils. The formation of these A␤ protofibrils was not due to self-assembly, because omission of calcium from the A␤ ϩ TGase mixture prevented the aggregation (Fig. 6F), and A␤40 incubated for 68 h in the absence of TGase also showed no aggregates (Fig. 6J). These experiments indicate that TGase can induce A␤40 to form assemblies that are similar, if not identical, to structures formed through self-assembly when A␤ reaches a critical concentration.
In support of the EM data, profiles of self-assembled versus TGase-induced A␤ species were compared by SEC in a non-denaturing buffer to determine if they had similar chromatographic profiles. Incubation of A␤42 without TGase for 0, 6, or 16 h showed an SEC aggregation pattern almost identical to A␤40 incubated with TGase for 0, 1, or 2 h (supplemental Fig. S1, A  and B). The similarity of the formed aggregates was corroborated by the degree of ␤-sheet content of both preparations as measured by thioflavin T binding (supplemental Fig.  S1C). A␤40 incubated for 2 h without TGase showed no oligomerization and was identical to the 0-h profile (data not shown). The EM, SEC, and thioflavin T analyses support that TGase can induce A␤ oligomerization and aggregation in a manner similar to that observed with A␤ self-assembly.
Both neprilysin (NEP) and insulin-degrading enzyme (IDE) are thought to be important proteases involved in the degradation of A␤ in vivo (32)(33)(34). Clearly, these proteases degrade monomeric A␤ (35)(36)(37), but it is controversial whether these enzymes also degrade oligomeric A␤ species (38 -41). Because AD pathology is thought not only to be due to an increase in A␤ aggregation but also to a failure of the body to degrade A␤ (42), the degradation of the TGase-induced A␤ polymers by these proteases was assessed. Incubation of SEC-purified, monomeric A␤40 with TGase induced significant oligomerization (Fig. 7A, lane 2), which was inhibited by 10 M TGase inhibitor LDDN-80042 (Fig. 7A, lane  1). In a third condition, the A␤ ϩ TGase reaction was stopped after 0.5 h with 10 M LDDN-80042 and then treated with 100 nM IDE at 37°C for 30 min (Fig. 7A, lane 3). Although a large fraction of the monomer was degraded in the presence of IDE, no significant degradation was observed for the upper oligomeric A␤ bands (compare Fig. 7A, lanes 2 and 3). These changes were quantified by densitometry of the individual bands on a Western blot (Fig. 7B). Individual bands were normalized to their corresponding partner in the A␤ ϩ TGase samples containing no IDE (i.e. no degradation was given a value of 1). A 60% loss of the monomer (4 kDa) was measured, whereas no degradation was observed for the bands containing  Similar results were obtained for NEP, another protease thought to play a role in A␤ degradation. After halting the A␤ ϩ TGase reaction by adding 10 M TGase inhibitor LDDN-80042, the resulting A␤ oligomers were subjected to 38 and 76 nM NEP for 1 h at 37°C (Fig. 7C, lanes 3 and 4, respectively). The ability of LDDN-80042 to inhibit TGase activity is seen in lane 1 of Fig.  7C (compare Fig. 7C, lane 1 (with inhibitor) with lane 2 (without inhibitor)). Densitometry of the Western blot shows that monomeric A␤40 was readily degraded (43% loss) (Fig. 7D, "mono"). Degradation of monomeric A␤ by NEP was completely blocked by 1 mM 1,10-phenanthroline (data not shown). Again, as with IDE, the A␤ assemblies generated by TGase were not significantly affected by NEP (8-, 12-, and 16-kDa bands in Fig. 7D). Thus, in both the NEP and IDE degradation assays, TGase-induced A␤ oligomers persisted despite significant degradation of monomeric A␤ by these proteases.
Various assemblies of both synthetic and biologically derived A␤ have been shown to inhibit LTP (43,44), a cellular correlate to memory and learning (45). Using a well studied N-methyl-Daspartate receptor-dependent paradigm (20), we tested whether TGase-induced A␤ assemblies would alter hippocampal LTP at the Schaffer collaterals-CA1 synapses. We produced A␤ assemblies with TGase, inhibiting the TGase reaction with LDDN-80042 (the resulting A␤40 assemblies are shown in Fig.  8D) and then diluting the solution to 100 nM. LTP was induced in hippocampal slices from 26-to 35-day-old mice by two high frequency stimulations. LTP was significantly reduced by exposure of the hippocampal slices to 100 nM TGase-treated A␤ compared with ACSF or 100 nM A␤ receiving no TGase treatment (Fig. 8, A-C) (fEPSP slopes: A␤ ϩ TGase ϭ 107.7 Ϯ 3.3%, n ϭ 14, versus ACSF ϭ 140.1 Ϯ 3.7%, n ϭ 19 or A␤ alone ϭ  . C, NEP also degraded monomeric, but not oligomeric A␤. No aggregation was observed if the inhibitor was added prior to TGase being added to A␤ (lane 1). NEP (38 and 76 nM) was added to TGase-treated A␤ (lanes 3 and 4, respectively) and showed degradation of monomeric, but not the oligomeric A␤ species (6 Ͼ kDa). D, by densitometry, a decrease in monomeric A␤ (ϳ4 kDa) was observed in the presence of 76 nM NEP, whereas the 8-, 12-, and 16-kDa bands were not affected. Bars represent the mean of three individual experiments (ϮS.E.).

DISCUSSION
Growing evidence implicates A␤ oligomerization and fibrillogenesis in the etiology of AD. However, in vitro models of fibrillogenesis by self-aggregation (3) require A␤ concentrations significantly higher than physiological concentrations (4 -6). To date, there has been no successful attempt to reconcile this fact with the known presence of aggregated A␤ in patients with AD. The results presented here suggest that TGase catalyzes the oligomerization of A␤ at physiological concentrations and produces assemblies with morphologies similar to previously described protofibrils (14,23), which have potential pathophysiological properties.
TGase has been suggested as playing a role in certain neurodegenerative diseases (8,9). Multiple disease-associated substrates have been shown to be cross-linked by TGase: hunting-tin in Huntington disease (46), ataxin-1 in spinocerebellar ataxia-1 (47), and Tau and A␤ in AD (48). There is additional support for the involvement of TGase in AD. First, TGase occurs abundantly in the brain, especially in areas thought to be affected by AD (49). Second, increased concentrations and higher activity of TGase have been found in AD patients relative to agematched controls (49 -52). Third, a splice variant found only in AD tissue produces a truncated form of the enzyme, which lacks the calcium regulatory domain of the full-length enzyme (53). Fourth, the identification of the ⑀-N-(␥-glutamyl)lysine isopeptide bond specific to TGase activity in the cerebrospinal fluid has high sensitivity and selectivity for the diagnosis of AD (54). Finally, A␤ contains the necessary lysine and glutamine residues and has been shown to be a substrate for TGase (10 -12). This particular feature of A␤, combined with the many associations between TGase and AD, suggests that TGase could be an effective facilitator of A␤ oligomerization in vivo.
An important aspect of the results presented here is the observation that TGase can lower the oligomerization threshold from the high concentrations needed for in vitro self-assembly to nanomolar concentrations, approaching physiological concentrations found in AD brain tissue (4 -6). Western blot analysis of TGase reactions conducted with A␤ concentrations as low as 50 nM shows the formation of A␤ oligomers. These structures survive boiling in SDS, suggesting that they are not due to concentrating the sample, but are stabilized by covalent bonds. The immunoblot results showed the formation of 6E10-immunoreactive material in a band located between ϳ40 and 180 kDa. This "smeared" banding pattern contains TGase (76 kDa), suggesting that A␤ oligomers of varying lengths are crosslinked to TGase. The prominence of this smear is due to the concentrating of the TGase, which is retained by the 3-kDa spin filter used to concentrate the A␤ assemblies. If A␤-TGase bonds were being exclusively catalyzed, the immunoblot results would show only distinct bands around the molecular weight of TGase or above. However, specific oligomeric A␤-A␤ bands (lower than the molecular weight of TGase) were also observed, specifically the formation of 8-and 12-kDa bands, which were not present or in lower concentrations than in the concentrated samples containing only A␤. Higher order intermediates may be forming under these conditions, but may not be detected due to their low concentrations and possible loss during the concentration step. Additionally, it is possible that the TGase-A␤ interaction is affecting the catalytic activity of TGase (i.e. is poisoning the enzyme) or is removing A␤ from the reaction by its binding to TGase. Nonetheless, our data suggest that the formation of these A␤-specific reaction products at low nanomolar concentrations is catalyzed by TGase, thus lowering the threshold for oligomerization by a factor of 100 or more.
Our studies show that TGase not only stabilizes these structures but allows A␤ to assemble into structures that are similar to those observed upon self-aggregation. A number of soluble A␤ morphologies have been imaged and investigated: A␤ oligomers, A␤-derived diffusible ligands, amylospheroids, pores, and protofibrils (38,44,(55)(56)(57)(58)(59)(60)(61). Protofibrils form through an orderly process, visually starting as small, 5-6 nm globular structures, which appear to coalesce into longer and longer curvilinear structures that ultimately form fibrils (14,29,30,62) (also see Ref. 63). From the EM images presented here, A␤ aggregates induced by TGase transition through almost identical intermediates as those observed with protofibrils. In support of this observation, binding of Thioflavin T to TGaseinduced A␤ aggregates was very similar to its binding to selfaggregated A␤42, indicating similar formation of ␤-sheet structures (see supplemental data). These data would suggest that TGase is inducing oligomerization and aggregation in A␤ in a specific manner, catalyzing pathogenic folding in a manner similar to self-assembly.
Similar to the data presented here for A␤, previous studies have found that TGase can induce the oligomerization and aggregation of various proteins, including A␤, ␣-synuclein, huntingtin, and Tau (10, 11, 64 -66). However, other studies have shown that TGase can also inhibit protein aggregation, which was illustrated with some of the same proteins, including Tau, ␣-synuclein, and huntingtin, but also for truncated yeast prion Sup35 and ␤-lactoglobulin (67)(68)(69). The effect of TGase cross-linking, promotion versus inhibition of protein aggregation, may be determined by whether TGase catalyzes the formation of cross-links within a protein, which could inhibit aggregation, or between two proteins, which could enhance aggregation. Because cross-linking by TGase requires a glutamine and a lysine residue, the degree of intraversus intermolecular cross-linking may depend on the position of these two amino acids relative to each other in the folded protein. A mutational analysis of the two lysines in A␤ showed that only substitution of Lys-16 inhibited aggregation, whereas mutating Lys-28 had no effect (10), suggesting that only Lys-16 plays a role in A␤ aggregation. The only Gln residue in A␤ is Gln-15, directly preceding Lys-16. There are currently no data that address the question whether intramolecular cross-links can form between the adjacent amino acids Gln-15 and Lys-16. Our observation that TGase induces oligomerization of A␤ would, however, favor the notion that TGase catalyzes the formation of intermolecular cross-links between Gln-15 in one A␤ monomer with Lys-16 in a second A␤ monomer. Although intramolecular cross-linking between Gln-15 and Lys-16 would also compete with the formation of intermolecular cross-links, the fact that we see TGase-induced oligomerization of A␤ argues against significant formation of intramolecular cross-links.
According to a model of the A␤ structure (70), Gln-15 and Lys-16 would be immediately adjacent to a hydrophobic region critical for aggregation of A␤ (71). The effect of TGase-catalyzed formation of intermolecular cross-links between A␤ subunits might be that the hydrophobic regions are brought into close proximity, thus enhancing oligomerization, especially at low A␤ concentrations. Konno et al. (67,68) suggested that the concentration of the substrate may determine the prevalence of intraversus intermolecular cross-linking, with low concentrations favoring intramolecular cross-links, which in turn are thought to inhibit oligomerization and/or aggregation. Our studies show, however, that TGase can induce oligomerization even at low, nanomolar concentrations of A␤. Furthermore, pre-existing self-assembled A␤ oligomers would be a much better substrate for TGase, because the Gln and Lys residues from different monomers would already be pre-oriented for crosslinking. However, even our monomeric A␤ preparation (i.e. purified by SEC in 6 M guanidine HCl) was a good substrate for TGase to induce the formation of oligomers protofibrils. Our data thus suggest that TGase favors to catalyze intermolecular cross-links, inducing the formation of A␤ oligomers, but we cannot rule out the possibility that intramolecular cross-linking is also occurring. It is possible that intramolecular cross-linking is competing with intermolecular cross-linking and slows down the oligomerization process we observe. How the abnormal folding of A␤ is initiated in AD is unknown, but possibly TGase may determine the degree to which A␤ folding is directed toward a pathogenic rather than non-pathogenic pathway. The transition from monomer to fibril is a complex process, and recent data showed that cross-linking with 4-hydroxynoneal enhanced oligomerization but blocked fibrillogenesis (72). Of further interest is the recent finding that various chemical compounds can separate the oligomerization and fibrillogenesis pathways (73). These findings will be important in determining which pathway is critical for the pathogenesis of AD and will help shape the therapeutic strategy that should be used to block the pathogenic A␤ assembly process.
In addition to A␤ aggregation, the time-dependent accumulation of A␤ in the brain is another invariable component observed in AD. Proposed mechanisms for the pathological accumulation of A␤ include not only the formation of A␤ aggregates but also an inability of the body to degrade A␤. Thus, mechanisms describing the failure of A␤ being degraded are also vital to understanding the initial pathogenic accumulation of A␤. Here, we show that the TGase-induced oligomeric A␤ species are resistant to both NEP and IDE, two proteases thought to be important for A␤ degradation in vivo (42), whereas the monomer was readily degraded by both metalloproteases. The importance of these results is corroborated by the manipulation of these proteases in animal models showing that increasing or decreasing their activity can alter A␤ accumulation inversely (74) as well as change synaptic function and behavior (75,76). Therefore, initiating A␤ aggregation could cause the incapacitation of these proteases by allowing A␤ to aggregate before the monomer can be proteolytically degraded. Our findings suggest that, once A␤ assemblies have formed due to TGase activity, they should persist in vivo, resulting in a progressive buildup of A␤. Should these A␤ assemblies be physiologically active, this persistence could increase their contribu-tion to the progressive neurodegeneration process observed in AD.
In associating A␤ species with pathology, soluble forms of aggregated A␤ correlate best with synaptic loss (77) and histopathological changes (78). These synaptic changes have been predictive of cognitive decline (79). Further support is provided by the findings that soluble forms of A␤ can specifically bind and disrupt synapses (61,80,81,83). Additionally, previous studies have shown that A␤ assemblies can mimic some of the symptomatology of AD, including alterations in synaptic function, memory and learning, and behavior, suggesting their early involvement in the progression of AD (38, 44, 55, 56, 84 -89). In support of TGase catalyzing the formation of potentially pathological species of A␤, we observed that A␤40 treated with TGase inhibited LTP at 100 nM A␤40, but not when A␤ or TGase were applied individually. These data are consistent with previous studies showing that various preparations of aggregated A␤ can inhibit LTP (38,44,55,82,84,(87)(88)(89).
The data presented here have physiological importance, because our results show that TGase may contribute to important attributes observed in AD: 1) TGase can initiate aggregation at physiological A␤ concentrations. TGase may not have to cross-link a large percentage of A␤ molecules to play a role in aggregation but may be important in driving the oligomerization or nucleating the aggregation process, allowing the selfaggregation to proceed more efficiently (3,11); 2) TGase can cause the accumulation of soluble A␤ species by producing protease-resistant A␤ assemblies; and 3) the A␤ species induced by TGase can inhibit processes involved in learning and memory, one of the first clinical symptoms observed in AD. For these reasons TGase may play an important role in the initiation and progression of AD. Therefore, the pharmacological manipulation of TGase could be a therapeutic strategy for slowing or blocking the progress of AD.