Amyloid-β Peptide Aβ3pE-42 Induces Lipid Peroxidation, Membrane Permeabilization, and Calcium Influx in Neurons*

Pyroglutamate-modified amyloid-β (pE-Aβ) is a highly neurotoxic amyloid-β (Aβ) isoform and is enriched in the brains of individuals with Alzheimer disease compared with healthy aged controls. Pyroglutamate formation increases the rate of Aβ oligomerization and alters the interactions of Aβ with Cu2+ and lipids; however, a link between these properties and the toxicity of pE-Aβ peptides has not been established. We report here that Aβ3pE-42 has an enhanced capacity to cause lipid peroxidation in primary cortical mouse neurons compared with the full-length isoform (Aβ(1–42)). In contrast, Aβ(1–42) caused a significant elevation in cytosolic reactive oxygen species, whereas Aβ3pE-42 did not. We also report that Aβ3pE-42 preferentially associates with neuronal membranes and triggers Ca2+ influx that can be partially blocked by the N-methyl-d-aspartate receptor antagonist MK-801. Aβ3pE-42 further caused a loss of plasma membrane integrity and remained bound to neurons at significantly higher levels than Aβ(1–42) over extended incubations. Pyroglutamate formation was additionally found to increase the relative efficiency of Aβ-dityrosine oligomer formation mediated by copper-redox cycling.

Pyroglutamate formation significantly increases the hydrophobicity of A␤, causing the peptide to aggregate more rapidly and form oligomers at lower concentration thresholds (5,14,15). pE-A␤ peptides also demonstrate increased ␤-sheet (aggregate structure) stability (16,17), differences in fibril ultrastructure (18,19), and altered interactions with copper ions (20,21) and synthetic lipid membranes (22,23). Notably, trace quantities of A␤3pE-42 have been observed to dramatically enhance the aggregation and neurotoxicity of A␤   (24), prompting descriptions of pE-A␤ as "prionlike." Still, it remains unclear as to the cytotoxic potency of pE-A␤ peptides compared with their full-length A␤ counterparts. Some studies have demonstrated pE-A␤ peptides to have enhanced toxicity (24 -26), although others have reported no difference in toxicity between the isoforms (27)(28)(29)(30). Methodological differences may account somewhat for variability in the relative toxicities reported (Table 1), yet molecular mechanisms to explain changes in cytotoxicity have not been defined.
One mechanism through which A␤ peptides cause cytotoxicity is by production of reactive oxygen species (ROS) via facile copper-redox cycling (31)(32)(33), which can in turn effect oxidative damage to neuronal proteins and lipids (34). Imbalances in ROS production and detoxification are strongly implicated in AD neurodegeneration, reflected by cerebral elevations in oxidized DNA, lipids, and proteins (35)(36)(37). Pyroglutamate formation alters A␤-Cu 2ϩ coordination modes (20,21), although it is not known whether this affects the capacity of pE-A␤ peptides to undertake redox cycling and produce cytotoxic ROS. We therefore aimed to determine whether full-length A␤ and pE-A␤ possess differences in their capacity to alter ROS flux and cause oxidative damage to neurons in vitro. Additionally, A␤ isoforms were compared for their capacity to form oligomers and covalent tyrosine-tyrosine bonds (dityrosine) as a result of A␤-copper-redox cycling. The capacity for A␤ to form dityrosine has previously been correlated with neurotoxicity (38), although recent reports have found that A␤ fibrils within amyloid plaques contain intense dityrosine immunoreactivity (39), indicating that dityrosine formation may be associated with AD amyloidogenesis. Further comparisons were made between the peptides for their capacity to perturb neuronal membranes and induce changes in neuronal ion homeostasis (Ca 2ϩ flux).
Preparation of A␤ Solutions and Cu 2ϩ -oxidized A␤ Oligomers-A␤ stock solutions were prepared by dissolving lyophilized peptides to 5 mg/ml in NaOH (60 mM) and incubating at ambient temperature for 5 min to dissociate aggregated material. Solutions were then diluted to 1 mg/ml in MilliQ H 2 O and 10ϫ PBS (PBS is defined as 50 mM sodium phosphate, 2.7 mM KCl, 137 mM NaCl) at a v/v/v ratio of 2:7:1 (NaOH, H 2 O, 10ϫ PBS). The preparation was sonicated for 10 min in an iced water bath and then centrifuged at 16,500 ϫ g for 10 min at 4°C. Supernatants (upper 80% of solution) were removed to prechilled tubes on ice; pH was adjusted by addition of NaH 2 PO 4 (0.5 M) to 1.0% v/v and then kept on ice for immediate use. A␤ stock concentrations were determined by UV spectrometry by measuring absorbance at 214 nm (A 214 ) and applying the following extinction coefficients (M Ϫ1 cm Ϫ1 ) determined from UV scans and amino acid analysis: A␤(1-40) ϭ 91,462; A␤3pE-40 ϭ 89,705; A␤(1-42) ϭ 94,526; and A␤3pE-42 ϭ 90,925. A␤ oligomers were generated by reacting A␤ (10 M) with Cu 2ϩ (10 M, as CuCl 2 ⅐glycine 6 ) and ascorbate (100 M) at 37°C on a vertically rotating wheel at 20 rpm, in 2-ml roundbottom tubes (catalog no. 0030123.344, Eppendorf). Reactions were halted with the addition of EDTA (to 250 M) and placed on ice. Control incubations of A␤ only were incubated and sampled under identical conditions for comparison. A␤ fibrillization assays (thioflavin-T assays) were performed as described previously (41).
Detection of Cell-free Hydroxyl Radical Production-Hydroxyl radical production in A␤-copper-redox cycling reactions was assessed with the fluorescent probe coumarin-3carboxylic acid (3-CCA) using methodology adapted from Manevich et al. (42). Reactions containing A␤ peptides (20 M) and 3-CCA (100 M) in the presence or absence of Cu 2ϩ (20 M, as CuCl 2 ⅐glycine 6 ) and ascorbate (300 M) in PBS, pH 7.4, were incubated in black-walled clear bottom microplates (PerkinElmer Life Sciences) at 37°C in a heated Flexstation III spectrophotometer (Molecular Devices). Fluorescence signals were read from the bottom of the microplate every 30 s using excitation 395 nm and measuring emission at 450 nm. Rate constants for the 3-CCA reactions were determined from one-phase association regressions of logY transformed data (data manipulations performed in GraphPad Prism version 6.0).

TABLE 1 Overview of publications comparing the cytotoxicity of pE-A␤ and full-length A␤ peptides in vitro
The following abbreviations are used in table: CSF, cerebrospinal fluid; DMSO, dimethyl sulphoxide; HFIP, 1,1,1,3,3,3-hexafluoro-2-propanol; LDH, lactate dehydrogenase;  PBS, phosphate buffered saline. A␤-Dityrosine Determination and Hydrophobic Index Calculation-The dityrosine content of Cu 2ϩ -reacted A␤ samples was determined by fluorescence spectrophotometry ( 320 nm excitation and 420 nm emission), as described previously (43). Reaction half-times and rate constants (K ϭ min Ϫ1 ) were calculated from one-phase association regressions of fluorescence data (GraphPad Prism version 6.0). Calculation of A␤ hydrophobic scores were determined from the sum of the amino acid residue hydropathic indexes (44). Pyroglutamyl residues were assigned a hydropathy value of Ϫ1.0.
A␤ Detection by Western Blot-Tissue extracts and synthetic peptides were separated by SDS-PAGE using 4 -12% XT Bis-Tris gels (Criterion, Bio-Rad) according to the manufacturer's instructions. Samples were transferred to pre-assembled PVDF membrane stacks using a Trans-Blot semi-dry transfer apparatus (Bio-Rad). Blots were blocked in TBS-T (10 mM Tris-HCl, 50 mM NaCl, 0.1% v/v Tween 20, pH 8.0) containing 5% w/v skim milk. Primary antibodies were incubated on blots for at least 1 h at room temperature or overnight at 4°C. HRP-conjugated rabbit anti-mouse or goat anti-rabbit immunoglobulins (Dako) were diluted 1:10,000 in TBST and incubated for 1 h at room temperature. All antibodies were diluted in TBS-T containing 5% skim milk and 0.05% w/v sodium azide. Blots were washed four times for 10 min in TBS-T after each primary and secondary antibody binding step. Chemiluminescence signals were captured after application of ECL (Immobilon, Millipore) with an LAS3000 detector and analyzed using MultiGauge software (Fujifilm). A␤ peptides were detected using the monoclonal mouse antibodies 4G8 (Covance) or 6E10 (Signet Laboratories) diluted to 1 g/ml. The dityrosine modification was detected using the 1C3 antibody raised against synthetic dityrosine (catalog no. NWA-DIY020, Northwest Life Science Specialties), at a 1:1000 dilution.
Size-exclusion Chromatography (SEC) and Atomic Force Microscopy (AFM)-A␤ oligomers in HBSS buffer were prepared as above but at twice the concentration (20 M A␤, 20 M Cu 2ϩ , 200 M ascorbate) to provide adequate signals for measurement.
SEC analysis was performed using a BioLogic DuoFlow QuadTec 40 system (Bio-Rad) fitted with a Superdex 75 10/300 column (catalog no. 17-5174-01, GE Healthcare). Both equilibration and operation of the column were in Tris-buffered saline (20 mM Tris, 200 mM NaCl, pH 8.0, 0.2 m filtered and de-gassed) at a flow rate of 0.5 ml/min and ambient temperature. The absorbance at 214, 260, and 280 nm was monitored, collecting 5 data points/s. Samples were injected onto the column (0.5 ml per run, ϳ45 g of A␤) immediately after collection at indicated time points.
AFM analyses were performed on A␤ reactions prepared in Neurobasal medium (catalog no. 21103-049, Gibco) at 10 M. Samples (10 l) were collected and spotted on freshly cleaved mica, dried at room temperature for 5 min in a laminar flow hood, and rinsed with 2 ml of de-ionized water (Milli-Q, Millipore). The sample was blown dry with nitrogen (Coregas Nitrogen 4.0) before being transferred to the AFM sample stage (Asylum Research Cypher AFM). Images were acquired in alternating current (tapping) mode in air using Tap300-G silicon AFM probes (Budget Sensors) with scan rates of 1.5-2.5 Hz; drive amplitude was kept to a minimum with amplitude set-points of 60 -80%. The mask threshold was set to 250 pm for image analysis.
Primary Neuronal Cell Culture and A␤ Clearance Assays-All experiments involving animals were conducted in accordance with the Australian Code of Practice for the Use of Laboratory Animals and were approved by the Institutional Animal Experimentation Ethics Committee.
Mouse cortical neuronal cultures isolated from C57Bl/6 E14 embryos were prepared as described previously (45). Cells were plated in poly-D-lysine-coated 48-or 96-well plates at a density of 150,000 cells/cm 2 . All cell culture materials were purchased from Gibco/Thermo Fisher unless otherwise stated. Cells were grown in Neurobasal medium (NB) containing B27 supplements, gentamicin, and 0.5 mM GlutaMAX TM . Fresh NB medium containing B27 minus antioxidants (B27-AO) and cytosine-␤-D-arabinofuranoside (2 M; catalog no. C1768, Sigma) were applied after 6 days in vitro (DIV). Neurons were further incubated until DIV 8 or 9 prior to applying treatments.
Levels of cell-bound A␤ were measured following exposure to A␤ peptides applied in NB medium containing B27-AO and cytosine-␤-D-arabinofuranoside. Cells (DIV 8 or 9) were treated for 48 h with A␤ mixtures (10 M total). The media were removed, and cells were washed twice with Dulbecco's PBS (catalog no. 14190-144, Gibco), and then the cells were extracted with M-PER reagent (catalog no. 78501, Thermo-Scientific). For fractionation studies, cells were scraped into TBS, pH 7.5, containing protease inhibitors (cOmplete, catalog no. 11873580001, Roche Applied Science) and probe-sonicated by two rounds of 10 bursts (40% power, 0.5 s each) with a Sonifier S-250D (Branson). Lysates were centrifuged at 100,000 ϫ g, and supernatants (TBS phase) were collected to fresh tubes. The pellets were extracted with an equal volume of Na 2 CO 3 (100 mM, pH 12) for 1 h of incubation on ice and then briefly vortexed and centrifuged again to collect supernatants (carbonate phase). Pellets were resuspended in an equal volume of urea buffer (7 M urea, 2 M thiourea, 4% CHAPS, 1% DTT, 50 mM Tris, pH 8.0). Western blots were performed on 5 g of total protein based on BCA assays (catalog no. 23225, Thermo-Scientific) of TBS phases, loading equal volumes for carbonate and urea phases.
Cell Viability Measurement-M17 neuroblastomas were cultured in DMEM containing fetal bovine serum (FBS; 10% v/v) and penicillin/streptomycin (catalog no. 15140, Gibco). Cells were plated in 6-well culture plates (catalog no. 140675, Nunc) at 40,000 cells/cm 2 and incubated overnight (37°C, 5% CO 2 ). Fresh medium supplemented with retinoic acid (10 M) was applied, and the cells were allowed to differentiate for 48 h. Freshly prepared A␤ solutions (10 M) or synthetic lipid peroxide (1 M) were applied to cells for 4 h and then the cells were collected using trypsin and resuspended in 1 ml of PBS containing 2% FBS. Propidium iodide was added to a final concentration of 1 g/ml. Flow cytometric analysis was performed on a Beckman Coulter CyAn ADP analyzer with Summit 4.0.3 acquisition software. Debris and cell aggregates were excluded by gating. Live and dead cells were then identified based on exclusion and inclusion, respectively, of propidium iodide. Analysis of data were performed using FCS Express 4 analysis software.
Lipid peroxidation was measured using the lipid hydroperoxide probe diphenyl-1-pyrenylphosphine (DPPP; catalog no. 62237, Cayman Chemicals), with methodology adapted from Ref. 47. DPPP (stock dissolved in N 2 -purged DMSO to 25.88 mM) was diluted to 50 M in NB media and applied to cells (DIV 8 -9) in 96-well black-walled microplates (Greiner) and then incubated for 60 min at 37°C, 5% CO 2 in the dark. The DPPPcontaining media were removed; the cells were rinsed twice with HBSS and then equilibrated in HBSS at 37°C and 5% CO 2 for 30 min before treatment. Freshly prepared A␤ peptides were applied to cells for 4 h until reading fluorescence ( 340 nm excitation and 380 nm emission) with an EnSpire multimode spectrophotometer (PerkinElmer Life Sciences). Positive control wells were treated with a synthetic lipid hydroperoxide standard (catalog no. 705014; Cayman Chemicals) for comparison. Care was taken at all steps to minimize artifactual oxidation of DPPP by reducing exposure to light.
Neuronal Calcium Flux Measurement-Cortical neurons plated at 230,000 cells/cm 2 in 96-well black-walled microplates (Greiner) were loaded with the fluorescent Ca 2ϩ sensor Fluo4-AM (catalog no. F14201, Molecular Probes) at DIV9 for 30 min at 37°C, 5% CO 2 , and then equilibrated to room temperature for 30 min. Fluorescence (excitation 490 nm and emission 520 nm) was measured at base line for 10 reads at 27-s intervals followed by a further 10 reads after injection of glutamate or A␤ (final concentrations of 1 and 10 M, respectively) using a Fluostar plate reader (BMG Labtech). Ca 2ϩ flux values (⌬F/F 0 ) were expressed as the difference between mean baseline and immediately following glutamate or A␤ application. Experiments determining the effect of metal depletion utilized neuronal cultures pre-treated for 1 h with Diamsar (10 M). To assess the contribution of NMDAR signaling in A␤-induced Ca 2ϩ flux, cultures were pre-treated for 15 min with the NMDAR antagonist MK-801 (1 mM). In separate experiments the supplied assay buffer (HBSS) was replaced with Ca 2ϩ /Mg 2ϩ -free HBSS to determine the source of Ca 2ϩ entering the cytosol.

Pyroglutamate Formation Alters the Production of ROS and
Dityrosine by A␤-pE-A␤ peptides demonstrated greatly increased fibrillization rates compared with the respective fulllength isoforms (Fig. 1A), consistent with previous reports (14,18). A␤(1-40), A␤(1-42), A␤3pE-40, and A␤3pE-42 were also compared for their capacity to produce ROS upon reaction with Cu 2ϩ and ascorbate. The rate of ⅐ OH production was significantly higher for A␤(1-42) compared with A␤3pE-42 (Fig. 1B). In contrast, A␤3pE-40 produced more ⅐ OH than A␤  and indeed all other A␤ isoforms. The kinetics of ⅐ OH production showed a similar sigmoidal pattern for all peptides with signals reaching plateau after ϳ20 min.
Differences in redox cycling and ⅐ OH production between A␤ peptides were further studied to compare formation of dityrosine bonds. Dityrosine formation rates partially reflected ⅐ OH production rates whereby A␤3pE-40 formed dityrosine more rapidly than A␤(1-40) (half-times of 5.44 and 33.64 min, respectively), whereas A␤(1-42) formed dityrosine more rapidly than A␤3pE-42 (half-times of 5.38 and 7.31 min, respectively) (Fig. 1, C and D). Total A␤-dityrosine content, however, did not mirror the levels of ⅐ OH produced by each peptide; A␤(1-42) formed more than double the amount of dityrosine than all other peptides. A direct relationship was not observed between the dityrosine formation rate and relative A␤ hydrophobicity (Fig. 1E); however, the efficiency of dityrosine formation (calculated as a ratio of dityrosine formed per unit of ⅐ OH generated) was increased for both pE-A␤ peptides relative to their full-length counterparts (Fig. 1F).

Comparison of the Oligomerization of Pyroglutamate-A␤ and Full-length A␤ in the Presence and Absence of Cu 2ϩ -Previous
studies have shown differences between A␤(1-40) and A␤  in oligomerization and dityrosine formation when reacted with Cu 2ϩ and biological reductants (48). We compared the profiles of full-length A␤ and pE-A␤ oligomers generated in the presence or absence of Cu 2ϩ and ascorbate. A␤(1-42) and A␤3pE-42 rapidly formed SDS-stable oligomers within 5 min of reaction with Cu 2ϩ , whereas increases in A␤(1-40) and A␤3pE-40 oligomers were observed at 15 min ( Fig. 2A). Significant formation of SDS-stable oligomers was not observed for A␤ incubated in the absence of Cu 2ϩ and ascorbate over the same time period.
We additionally examined the oligomer states of A␤(1-42) and A␤3pE-42 in non-denaturing conditions by SEC and AFM. After a 4-h incubation in PBS, the A␤(1-42) oligomer profile did not change considerably from the freshly prepared solution, resolving predominantly as a single low-mass peak (Ͻ14 kDa) by size exclusion, with minor signals corresponding to oligomers of ϳ40 -75 kDa (Fig. 3A). Addition of Cu 2ϩ and ascorbate to A␤(1-42) caused a sizable increase in the abundance of oligomers but did not appear to alter their mass. By comparison, A␤3pE-42 ternary states changed considerably during the 4-h incubation, observed as a dramatic loss of low-mass peak (Ͻ14 kDa) in either the presence or absence of Cu 2ϩ and ascorbate (Fig. 3B). After a 4-h incubation, residual signals remaining in the A␤3pE-42 reactions without Cu 2ϩ resolved predominantly as an oligomeric peak at ϳ30 -45 kDa. In the A␤3pE-42 reaction containing Cu 2ϩ there was some preservation of the monomeric signal after the 4-h incubation, with an additional polydisperse peak across the 30 -75-kDa range.
To simulate A␤ oligomerization in cell culture conditions, reactions were conducted in neurobasal media over an extended incubation period, and the peptide structures were assessed by AFM (Fig. 3C). Little difference in peptide structure was observed during a 48-h incubation of A␤  in unsupplemented neurobasal media (maximum z Ͻ1 nm). In the A␤(1-42) reactions supplemented with Cu 2ϩ /ascorbate, there was similarly no noticeable change after 4-h incubation, although a large increase in the size and abundance of oligomers was observed after 48 h (maximum z ϭ 4.9 nm). Likewise, A␤3pE-42 structures did not demonstrate a noticeable size difference over a 48-h incubation in neurobasal media in the absence of supplemental Cu 2ϩ /ascorbate (maximum z Ͻ1 nm), although the reactions supplemented with Cu 2ϩ /ascorbate showed a stepwise size increase over the incubation period (maximum z ϭ 0.4, 1.0, and 1.8 nm at 0, 4, and 48 h, respectively).
A␤3pE-42 Predominantly Associates with Neuronal Membranes and Is Resistant to Clearance-pE-A␤ has been found to resist proteolytic clearance in astrocyte cultures (49,50). We therefore compared the residual levels of A␤(1-42) and A␤3pE-42 following extended exposures to cortical neuron cultures, additionally assessing the capacity for minor quantities of A␤3pE-42 to affect the clearance of A␤ . Neuronal A␤ levels were strikingly different after 48 h of treatment with the two peptides; relative levels of A␤3pE-42 were 4.5-fold higher than cells treated with A␤(1-42) (Fig. 4A). A trend toward higher mean levels of residual A␤ was observed when A␤(1-42) solutions were supplemented with either 50 or 25% , measured by fluorescence spectrometry of dityrosine bond (320 nm excitation and 420 nm emission). E, relationship between A␤ hydrophobicity and dityrosine formation rate. F, relative efficiency of A␤-dityrosine formation was further calculated as a ratio of the amount of dityrosine formed per unit of ⅐ OH produced, plotted against A␤ hydrophobicity. All measurements were repeated a minimum of three times over separate days using freshly prepared A␤ solutions. Data shown are mean Ϯ S.E.
From experiments utilizing sequential extraction of cells following A␤ exposure, both the A␤(1-42) and A␤3pE-42 peptides were predominantly found in membrane fractions, with relatively little A␤ in the soluble (TBS) phase observed qualitatively by Western blot (Fig. 4B). A␤ in cell extracts resolved primarily as monomers on SDS-PAGE; however, bands corresponding to low-mass oligomers (ϳ8 -20 kDa) were observed in longer blot exposures.
Toxicity of Pyroglutamate-A␤ Is Associated with Membrane Damage but Not Cytosolic ROS Production-The capacities of each peptide to produce elevations in cytoplasmic and membrane ROS were studied using primary cortical mouse neurons. The fluorescent ROS probe DCF was used to measure cytoplasmic ROS flux over a 1-h treatment with freshly prepared A␤ peptides. A␤(1-42) induced an approximate 10-fold increase in DCF fluorescence compared with control cells, whereas A␤(1-40), A␤3pE-40, or A␤3pE-42 treatments were not found to be statistically different from controls (Fig. 5, A and B).
To assess A␤ induction of ROS at neuronal membranes, we directly measured lipid peroxidation in living cortical mouse neurons using the fluorescent probe DPPP over a 4-h treatment. In direct contrast to the DCF assays, A␤(1-42) (10 M) did not increase neuronal DPPP fluorescence, whereas cells treated with A␤3pE-42 (10 M) demonstrated a statistically sig-   MARCH 18, 2016 • VOLUME 291 • NUMBER 12 JOURNAL OF BIOLOGICAL CHEMISTRY 6139 nificant 35% increase in lipid peroxidation, comparable with levels induced by 1 M synthetic lipid hydroperoxide (Fig. 5C). Further contrasting the DCF assays, minor quantities of A␤3pE-42 had no effect on the lipoperoxide-inducing capacity of A␤ .

Pyroglutamate-A␤ Induces Neuronal Membrane Damage
To determine whether lipid peroxidation was associated with a loss of plasma membrane integrity, differentiated M17 neuroblastomas were measured for propidium iodide uptake by flow cytometry following a 4-h A␤ treatment. Consistent with the lipid peroxidation experiments, A␤3pE-42 induced significantly higher levels of membrane damage than A␤(1-42) (Fig.  5D). A␤(1-42) solutions seeded with 5% (mol/mol) A␤3pE-42 induced practically identical levels of plasma membrane damage as A␤  alone.

A␤3pE-42 Induces Rapid Ca 2ϩ Influx in Primary Cortical
Neurons-Recent reports of pE-A␤ interactions with synthetic lipid membranes have suggested that pE-A␤ neurotoxicity is exerted via formation of ion-permeable membrane pores (22,23). We compared the capacity of A␤  and A␤3pE-42 to alter cellular ion homeostasis by measuring Ca 2ϩ flux in cortical neurons immediately following application of freshly prepared A␤. A␤3pE-42 (10 M) effected a rapid and significant elevation in neuronal Ca 2ϩ levels as measured with the Fluo-4 fluorescent Ca 2ϩ sensor, comparable with levels induced by 1 M glutamate (Fig. 6, A and B). By comparison, A␤(1-42) did not cause significant elevation in Ca 2ϩ levels above those of the vehicle control. When neurons were pre-treated with the NMDAR antagonist MK-801, there was an approximate 50% decrease in Ca 2ϩ flux induced by both glutamate and A␤3pE-42, indicating that A␤3pE-42-induced Ca 2ϩ flux is at least partially attributed to NMDAR activation (Fig. 6B). To determine whether A␤3pE-42-induced Ca 2ϩ flux was dependent on A␤metal interactions, the neuronal culture media were depleted of bioavailable first row transition metals with the chelator Diamsar. Rapid induction of Ca 2ϩ flux by A␤3pE-42 was not affected by Diamsar pre-treatment, suggesting that the membrane perturbation was independent of A␤-copper-redox cycling (Fig.  6C). A␤3pE-42 did not induce significant elevations in cytosolic Ca 2ϩ when provided in Ca 2ϩ /Mg 2ϩ -free media, indicating that A␤3pE-42 causes Ca 2ϩ to enter from the extracellular space and not via release from intracellular stores (Fig. 6D).

Discussion
A␤ in the human brain is represented by a heterogeneous and dynamic mixture of isoforms, with significant compositional variation between individuals (1-3, 5, 51). The predominance of pE-A␤ in the central core of plaques suggests an early involvement in amyloid deposition in the AD brain (52), whereas correlation between pE-A␤ and a decline in Mini Mental State Examination scores implicate pE-A␤ cytotoxicity in AD neurodegeneration (53,54). In parallel, markers of oxidative stress are among the earliest detectable pathological changes in transgenic AD mouse models (55) and the human AD brain (36,56,57), with numerous lines of evidence implicating A␤ as a central contributor to oxidative stress in AD (58,59).
Our data indicate that full-length A␤ and pE-A␤ exert different oxidative insults upon neurons, representing different mechanisms of neurotoxicity. A␤(1-42) was found to significantly increase cytosolic ROS levels, whereas A␤3pE-42 induced lipid peroxidation in the absence of cytosolic ROS flux. Additionally, we found that that the neuronal membrane damage caused by A␤3pE-42 results in a functional loss of plasma membrane integrity. These findings are consistent with recent publications demonstrating the capacity for A␤3pE-42 to form membrane-disrupting pores in synthetic lipid bilayers (22,23). Similarly, previous studies have shown that pE-A␤ oligomers disrupt lysosome membrane integrity in cultured neurons (50) and cause lactate dehydrogenase leakage from cultured astrocytes (49).  42) alone, as A␤ signals were not observed in vehicle control wells in these conditions. B, cell extracts 48 h after application of freshly prepared A␤ peptides to cortical neurons were fractionated to separate soluble A␤ (TBS extracted), peripheral membrane-bound A␤ (carbonate phase), and integral membrane-bound A␤ (urea phase). A␤ monomers are apparent at the approximate molecular mass of A␤ (4 -5 kDa), and A␤ oligomeric signals are identified by arrowheads (contrast enhanced panel). 4G8-reactive signals above 35 kDa were present in all groups (including vehicle control) suggesting that these signals are not A␤ but may correspond to the amyloid precursor protein and its cleavage products in the cortical neurons. All densitometry signals were corrected for protein loading by re-probing blots for actin. Samples shown are representative of three separate fractionations per group; panels have been cropped from the same blot image to show only relevant lanes. Data are mean Ϯ S.E., statistical significance is relative to A␤(1-42)treated group.** represents p Յ 0.01.
The earliest detectable effect of A␤3pE-42 on neuronal homeostasis that we observed was the capacity to cause rapid Ca 2ϩ influx, an effect that was partially ameliorated by pretreatment of neurons with the NMDAR antagonist MK-801. The Ca 2ϩ flux induced by A␤3pE-42 appears to be separate from its capacity to undergo copper-redox cycling as the effect persisted when media were depleted of row 1 transition metals. Changes in cellular Ca 2ϩ homeostasis induced by A␤ have previously been implicated in A␤ toxicity and are thought to occur via multiple mechanisms such as pore formation and NMDAR activation (60,61). Importantly, however, the aggregation state of the A␤ preparation significantly affects the capacity to induce Ca 2ϩ flux; monomeric (freshly prepared) and fibrillar A␤(1-42) is not found to induce Ca 2ϩ flux in SH-SY5Y neuroblastomas, whereas oligomeric preparations agonize NMDAR and trigger rapid Ca 2ϩ flux (61,62). Similarly, we found freshly prepared A␤  to cause only modest elevation in cortical neuron Ca 2ϩ levels above controls, which is contrasted by the much larger Ca 2ϩ flux induced by A␤3pE-42. Collectively, these data indicate that the unique neurotoxicity of pE-A␤ peptides is exerted through multiple interactions at the cell surface, including activation of NMDAR pathways, subsequently followed by peroxidation of membrane lipids and a loss of membrane integrity.
The potential for amino-truncated A␤ and pE-A␤ to alter the oligomerization and toxicity of full-length A␤ has been an area of recent debate and speculation. A␤3pE-42 has a demonstrated capacity to dramatically enhance the aggregation of fulllength A␤ (14), as does A␤3-42 (41). A␤3pE-42 has further been reported to enhance the toxicity of A␤  in a prionlike seeding mechanism in cortical neuron cultures (24). Due to the strong membrane association of A␤3pE-42, we initially predicted A␤3pE-42 seeds to shift the ROS generation of A␤(1-42) to a lipid compartment, thereby increasing A␤(1-42)-induced lipoperoxidation; however, A␤3pE-42 was not found to increase A␤(1-42) ROS production in either the cytosolic or membrane fractions. In contrast, the capacity for A␤3pE-42 seeds to significantly reduce A␤(1-42) cytosolic ROS production indicates that A␤3pE-42 does not enhance A␤(1-42) toxicity via ROS production, yet it suggests significant interaction between the peptides in neuronal cultures. This effect may be a result of accelerated A␤  aggregation in the presence of A␤3pE-42, as previous studies have found that A␤-ROS production decreases with aggregation (32,63). In the study by Nussbaum et al. (24), trace quantities of A␤3pE-42 were found to enhance the toxicity of A␤(1-42) when co-aggregated prior to cell treatment; however, when A␤3pE-42 was seeded into A␤  solutions immediately before applying to cells (as in our experiments), the toxicity of the seeded mixture was identical to the individual A␤(1-42) treatment. It is therefore likely that A␤ aggregates possess different cytotoxic properties depending on both the composition of the A␤ mixture and the timing of the A␤ isoform interactions. The possibility also exists that A␤3pE-42 seeded mixtures cause oxidative damage that escape detection by the DCF and DPPP fluorescence ROS probes or possess different redox cycling capacities in other cerebral cell types (e.g. astrocytes and microglia). Other prevalent amino-truncated and pE-A␤ peptides found in AD brain tissues, such as A␤ , A␤ (4 -42), and A␤11pE-42 (2,52), require further investigation as they also demonstrate enhanced amyloid-seeding capacity and could potentially alter A␤-ROS dynamics.
AFM and SEC analysis revealed that A␤(1-42) and A␤3pE-42 oligomers differed not only in the rate of formation but also in size and structure. Previous studies have similarly reported differences between A␤(1-42) and A␤3pE-42 in the profile of oligomers and fibril ultrastructures formed in aqueous buffers (18,30). Lee et al. (22) report that A␤3pE-42 forms larger oligomers in synthetic lipid membranes than A␤  and with faster kinetics of assembly, although it remains to be determined how this relates to the relative level of toxicity. A␤3pE-42 neurotoxicity has also been demonstrated with a broad range of size fractions (monomers to Ͼ100 kDa), whereas the toxicity of A␤  fractions was isolated to oligomers with an observed mass larger than 14 kDa (25). This is consistent with our SEC findings demonstrating the relative stability of low-mass A␤(1-42) species (Ͻ14 kDa) in aqueous solution and the observation that freshly prepared A␤(1-42) is less neurotoxic than A␤3pE-42. The metastable nature of A␤ peptides, however, presents many technical challenges in delineating "toxic" fractions from "benign" fractions as purification processes undoubtedly alter oligomerization kinetics. Likewise, A␤ peptides undergo significant structural changes in extended cell culture incubations; hence, it is pertinent to correlate toxicity markers with time-matched biophysical characterizations.
The A␤3pE-40 and A␤3pE-42 peptides have been found to resist proteolysis in astrocyte cultures (49) and accumulate in the lysosomes of astrocytes in cell culture and the AD temporal cortex (50). Consistent with these reports, we observed A␤3pE-42 to resist clearance in neuronal cultures, remaining at significantly higher levels than A␤(1-42) over extended incubations. A␤3pE-42 was, however, not found to prevent the clearance of A␤  from neural cultures when present at minor quantities (5% mol/mol), suggesting that A␤3pE-42 does not transfer protease resistance to A␤(1-42) when applied to cells as fresh preparations. This does not exclude the possibility that pE-A␤ may affect the clearance of full-length A␤ when the peptides are aggregated, which will require further investigation. The capacity for A␤3pE-42 to resist proteolytic degradation in both neurons and astrocytes is highly relevant given that pE-A␤ peptides are found in the cores of amyloid plaques in the AD brain (52), suggesting that pE-A␤ peptides, once formed, are long-lived neurotoxins. Dityrosine is another post-translational protein modification that confers resistance to cellular catabolism. Total dityrosine levels are elevated in the AD hippocampus, neocortex, and ventricular cerebrospinal fluid compared with cognitively healthy individuals (64), yet the contribution of A␤ to cerebral dityrosine formation is not well understood. Dityrosine crosslinked A␤ fibrils are resistant to formic acid digestion, and sections of AD brains display intense dityrosine immunoreactivity within plaques (39). The pE-A␤ isoforms demonstrated increased efficiency for dityrosine oligomer formation compared with full-length A␤. This observation can likely be attributed to increased hydrophobicity and the propensity to oligomerize; dityrosine formation is dependent on both the production of A␤-tyrosyl radicals and the close proximity of A␤ molecules to allow tyrosine-tyrosine coupling (33). It is reasonable to speculate that the capacity for amyloid plaques to resist solubilization and clearance is due to the contribution of both the pE and dityrosine modifications to A␤, which may account for the abundance of pE-A␤ in plaque cores (52).
Our data demonstrate clear differences in the neurotoxic mechanisms of pE-A␤ and full-length A␤. The toxicity of pE-A␤ has recently been highlighted in mouse models overexpressing the soluble isoform of QC, demonstrating exacerbated neurodegeneration and behavioral deficits when crossed with mouse lines overexpressing A␤ or amyloid precursor protein transgenes (65,66). The specific targeting of pE-A␤ via inhibition of QC-catalyzed pyroglutamate synthesis has demonstrated promising results; transgenic AD mice treated with a QC inhibitor show reduced plaque load, reduced gliosis, and an improvement in context memory and spatial learning (67). QC knock-out does not completely inhibit pE-A␤ formation in transgenic AD mouse models, however (68), suggesting that multiple pathways of pE-A␤ formation may exist. Antibodies currently under development as passive vaccine therapies for AD show significant differences in their capacity to bind amino-truncated A␤ species (69) and thus may fail to remove pE-A␤ and its precursors (A␤3-40/42 and A␤11-40/42). Individuals with AD may therefore require different therapeutic interventions to target distinct A␤ isoforms and their specific mechanisms of toxicity. A␤ ROS production, oligomerization, and dityrosine formation are potential therapeutic targets for AD; metal-protein attenuating compounds that inhibit these reactions are found to effect a marked decrease in amyloid deposition and improvement in cognitive deficits in transgenic AD mice (70,71). Our observations indicate that the separate pathways of oxidative damage and neurotoxicity exerted by A␤3pE-42 potentially have a unique contribution to AD pathology. These findings elicit further consideration of pE-A␤ peptides as targets in the pursuit of biomarkers and disease-modifying therapeutics for AD.