At 5 μM peptide, 5% pE-Aβ aggregated faster than Aβ_3(pE)-42 or Aβ_1–42 alone based on thioflavin T fluorescence shifts¹⁵ (Supplementary Fig. 3). The OD₄₅₀/OD₄₉₀ ratio for Aβ_3(pE)-42 rose and peaked more rapidly than for Aβ_1–42, but peaked at an ~25% lower level. The fastest rise in the OD₄₅₀/OD₄₉₀ ratio was for 5% pE-Aβ, which peaked similarly to Aβ_3(pE)-42. Aβ_3(pE)-42, Aβ_1–42 and 5% pE-Aβ thus oligomerized by different pathways.

To test whether distinct biological activities were coupled to these oligomerization differences, we compared cytotoxicity of the peptides towards cultured neurons or glia using calcein-AM and fluorescence microscopy¹⁶. Twelve hours of Aβ_1–42 exposure had little effect on cell viability for wild type (WT) or tau knockout (KO) neurons, or WT glial cells (Fig. 1a). Contrastingly, most WT neurons died and detached from the substrate after exposure to Aβ_3(pE)-42 or 5% pE-Aβ. Tau KO neurons and WT glia, which express little tau, were resistant to Aβ_3(pE)-42 and 5% pE-Aβ.

Cytotoxicity dose-dependence was examined by incubating WT neurons for 24 hours in oligomers comprising 0.1, 0.5 or 1 μM peptides, and using the XTT reduction assay¹⁷ (Fig. 1b). Cells were unaffected by Aβ_1–42, but Aβ_3(pE)-42 and 5% pE-Aβ had substantial cytotoxicity at 0.5 μM and even more at 1.0 μM. Cytotoxicity of 5% pE-Aβ required Aβ_3(pE)-42 and Aβ_1–42 to incubate together for 24 hours before being added to cells. When they were incubated separately for 24 hours and mixed together at a 1:19 molar ratio immediately before being applied to cells, they were not cytotoxic. A small amount of Aβ_3(pE)-42 can thus dramatically enhance the cytotoxicity of a large excess of Aβ_1–42, provided the two peptides oligomerize together.

Evidence for hybrid oligomers came from immunoprecipitation (IP) of various forms of Aβ using aggregation-dependent M64, which does not recognize Aβ_3pE-42 (see Supplementary Fig, 4 for characterization of all anti-Aβ antibodies used, including M64). IPs were analyzed on dot blots using 4G8, which equally recognizes Aβ_3(pE)-42 and Aβ_1–42, and anti-pE-Aβ, which does not react with Aβ_1–42. M64 IP’d oligomers made from Aβ_1–42 or 5% pE-Aβ, but it did not IP Aβ_3(pE)-42 oligomers, nor monomers of either peptide (Fig. 2a). Because anti-pE-Aβ reacted with material IP’d out of 5% pE-Aβ, M64 pulled down hybrid peptide oligomers. Aβ_3(pE)-42 accounted for ~16% of the Aβ in gel filtered cytotoxic oligomers after 3 hours of oligomerization, and steadily dropped to ~8% by 24 hours (Fig. 2b). Aβ_3(pE)-42 thus acts as a template that initiates formation of cytotoxic oligomers.

Cytotoxicity was sensitive to oligomerization time (Fig. 2c). Baseline cytotoxicity was observed at alltime points for Aβ_1–42, and for 5% pE-Aβ solutions in which Aβ_3(pE)-42 and Aβ_1–42 oligomerized separately. Pure Aβ_3(pE)-42 killed ~50% of the cells after 24 hours of oligomerization, but was virtually non-toxic at 0 hours and after 96 hours of oligomerization. The most cytotoxic solutions were 5% pE-Aβ in which the constituent peptides co-oligomerized for 24 hours. These solutions killed ~60% of the cells within 24 hours, and lower, but robust cytotoxicity was observed at 96 hours. Even the 0 hour co-oligomers of 5% pE-Aβ exhibited low, significant cytotoxicity. Co-incubated mixtures of 5% Aβ_3(pE)-42 and 95% Aβ_1–42 can therefore form oligomers whose cytotoxicity is both greater and more enduring than oligomers formed by Aβ_3(pE)-42 alone.

To identify the co-oligomer size(s) that were cytotoxic, Aβ solutions were oligomerized for various times from 0–96 hours before fractionation by gel filtration. Total Aβ in all fractions was determined using 4G8 dot blots, which as shown in Fig. 3a (for 5% pE-Aβ) and Supplementary Fig. 5 (for Aβ_1–42 and Aβ_3(pE)-42), illustrate the full fractionation range of the column but exclude most void volume fractions. Presumptive monomeric Aβ_1–42 dominated initially and persisted at 3 hours, but was nearly undetectable after 12 hours. 3 hours also marked the appearance of Aβ_1–42 oligomers, which gradually increased in size over the next 93 hours. Aβ_3(pE)-42 and 5% pE-Aβ oligomerized differently. Putative monomers were present at 0 hours for both samples, when slightly larger species, LNOs that possibly corresponded to dimers/trimers (Supplementary Fig. 6), were also present. These persisted as the main species for 24 hours for Aβ_3(pE)-42 and for nearly 72 hours for 5% pE-Aβ, and later time points were dominated by larger aggregates that eluted in void volume fractions. Cytotoxicity was assayed for individual fractions of 5% pE-Aβ that oligomerized for 24 hours (Fig. 3b). Most cytotoxicity was associated with the possible dimers/trimers that eluted at 12.5 ml, which at 425 nM peptide killed more than 60% of the cells. Low cytotoxicity was also observed at 554 nM peptide for the larger oligomers that eluted at 8.5 ml.

The dramatic enhancement of Aβ_1–42 cytotoxicity by Aβ_3(pE)-42 suggested a prion-like templating mechanism of Aβ_1–42 misfolding initiated by Aβ_3(pE)-42. To test that hypothesis, 5% pE-Aβ that oligomerized for 24 hours was diluted into 19 volumes of monomeric Aβ_1–42. A 24 hour incubation of this mixture yielded “serial passage 1”, which was followed by two equivalent, sequential dilutions into monomeric Aβ_1–42 to yield serial passages 2 and 3. A gradual loss of cytotoxicity was observed with successive passages, but even passage 3, which contained only 0.000625% Aβ_3(pE)-42, killed ~50% of the neurons within 24 hours (Fig. 3c). Serially passaged gel filtration samples contained abundant material that eluted at 12.5 ml in passages 1–3, despite the progressive dilution of Aβ_3(pE)-42 (Fig. 3d). Aβ_3(pE)-42 can therefore template formation of metastable, cytotoxic LNOs from excess Aβ_1–42, yielding potent bioactivity that can be serially passaged multiple times into monomeric Aβ_1–42 without further addition of Aβ_3(pE)-42.

One possible explanation for why Aβ_1–42 LNOs were inert is they lacked sufficient properly sized oligomers. Accordingly, we altered the oligomerization protocol from 5 μM peptide for 24 hours at 37° C to 10 μM peptide for 30 minutes at 4° C to obtain abundant Aβ_1–42 oligomers that eluted at 12.5 ml (Fig. 3e). These LNOs were not cytotoxic (Fig. 3f), implying they were structurally distinct from the putative dimers/trimers initiated by Aβ_3(pE)-42. This was confirmed by dot blots using M87, a conformation-sensitive anti-Aβ antibody, to compare the putative dimers/trimers used for the cytotoxicity assays shown in Fig. 3f. We first lyophilized aliquots of all the Aβ solutions, resuspended them with hexafluoroisopropanol (HFIP) to restore them to monomers, and then analyzed them using 4G8. When parallel samples that were not lyophilized but were otherwise identical were analyzed using M87, immunoreactivity was ~2X as strong with LNOs made from Aβ_1–42 versus those made from 5% pE-Aβ (Supplementary Fig. 7). Cytotoxic LNOs of 5% pE-Aβ are thus structurally distinct from comparably sized LNOs of Aβ_1–42.

Several lines of evidence demonstrate in vivo relevance for the data described so far. First, we identified LNOs containing Aβ_3(pE)-42 in 3 out of 3 AD samples, based on gel filtration of human brain extracts followed by dot blots of resulting fractions with anti-pE-Aβ and M87. In contrast, only 1 of 3 age-matched samples with normal neuropathological diagnoses was positive for Aβ_3(pE)-42. (Fig. 4a and Supplementary Fig. 8). Secondly, we crossed TBA2.1 mice¹⁸ into a tau KO background¹⁹. By 3 months, TBA2.1 mice accumulated small amounts (40–100 ng/g brain weight) of Aβ_3(pE)-42, which formed primarily intraneuronal aggregates, and was associated with massive hippocampal neuron loss and gliosis¹⁸. Knocking out tau provided almost complete protection against neuron loss and glial activation (Fig. 4b). Additional in vivo data are shown in Supplementary Fig. 9. Long term potentiation (LTP) of mouse hippocampal neurons in slice cultures was potently and equally inhibited by oligomers made from 5% Aβ_3(pE)-42 or 100% Aβ_3(pE)-42, whereas Aβ_1–42 oligomers had no effect on LTP. 1% Aβ_3(pE)-42 provoked mild, but statistically insignificant LTP impairment (Supplementary Fig. 9a). To evaluate effects of increased Aβ_3(pE)-42 in animal models, we crossed mice with neuron-specific expression of human APP harboring Swedish and London mutations (hAPP_SL)²⁰, with mice expressing human QC²¹. Nine month old double (hAPP_SL/hQC) and single (hAPP_SL) transgenic mice were indistinguishable in terms of insoluble and soluble Aβ_x-42 levels, but the double transgenics had ~2-fold more insoluble Aβ_3(pE)-42 and ~9-fold more soluble Aβ_3(pE)-42 than single transgenics (Supplementary Fig. 9b). Further analysis of the soluble Aβ_x-42 by the A4 assay²² revealed an ~8-fold excess of oligomers in the double, versus single transgenics (Supplementary Fig. 9c). Double transgenics performed more poorly in Morris water maze tests (Supplementary Fig. 9d) and had reduced hippocampal immunoreactivity for the synapse marker, synaptophysin (Supplementary Fig. 9e). Finally, peri-hippocampal injection of 5% pE-Aβ at 5 μM into APP_SwDI/NOS2^−/− AD model mice²³ led 3–5 months later to the presence of plaques containing both pE-Aβ and conventional Aβ. Comparable plaques were rarely seen in sham injected AD mice or in WT mice injected with 5% pE-Aβ (Supplementary Fig. 9f). These collective in vivo results emphasize the physiological significance of the companion biochemical and cultured cell results.

Our studies provide new insights into AD pathogenesis by demonstrating that hypertoxic Aβ oligomers can be triggered by small quantities of a specifically truncated and post-translationally modified version of Aβ. Although some previous studies demonstrated that pE modification of Aβ significantly enhances its aggregation kinetics^13,14,24, toxicity^12,25 and resistance to degradation¹², a mechanistic explanation for the unique properties of pE-Aβ has been lacking until now. Prior studies suggest coincident appearance of Aβ_3(pE)-42 with development or progression of human AD^26,27. Co-localization of QC and Aβ_3(pE)-42 was found in cored plaques of vulnerable regions in AD, and evidence was provided for axonal transport of Aβ_3(pE)-x from QC-rich neuronal populations of the entorhinal cortex and locus coeruleus²⁸. Since LNOs containing Aβ_3(pE)-42 are reasonably stable (Fig. 3a), they might initiate tau-dependent cytotoxicity intracellularly during axonal transport²⁹ or extracellularly following release at remote hippocampal synapses³⁰ of projection neurons²⁸. The Aβ_3(pE)-42 induced formation of toxic mixed oligomers provides a rationale for these previous observations, and the tau-dependent cytotoxicity of 5% pE-Aβ establishes a new functional connection between Aβ and tau in AD pathogenesis.

1