Soluble Aβ Inhibits Specific Signal Transduction Cascades Common to the Insulin Receptor Pathway*

Numerous studies have now shown that the amyloid β-protein (Aβ), the principal component of cerebral plaques in Alzheimer disease, rapidly and potently inhibits certain forms of synaptic plasticity. The amyloid (or Aβ) hypothesis proposes that the continuous disruption of normal synaptic physiology by Aβ contributes to the development of Alzheimer disease. However, there is little consensus about how Aβ mediates this inhibition at the molecular level. Using mouse primary hippocampal neurons, we observed that a brief treatment with cell-derived, soluble, human Aβ disrupted the activation of three kinases (Erk/MAPK, CaMKII, and the phosphatidylinositol 3-kinase-activated protein Akt/protein kinase B) that are required for long term potentiation, whereas two other kinases (protein kinase A and protein kinase C) were stimulated normally. An antagonist of the insulin receptor family of tyrosine kinases was found to mimic the pattern of Aβ-mediated kinase inhibition. We then found that soluble Aβ binds to the insulin receptor and interferes with its insulin-induced autophosphorylation. Taken together, these data demonstrate that physiologically relevant levels of naturally secreted Aβ interfere with insulin receptor function in hippocampal neurons and prevent the rapid activation of specific kinases required for long term potentiation.

Alzheimer disease (AD) 2 is characterized by a progressive loss of memory and the accumulation of amyloid plaques and neurofibrillary tangles in discrete brain regions. The amyloid ␤-protein (A␤), the principal constituent of amyloid plaques observed in AD and to a lesser extent during normal brain aging, is generated by sequential proteolytic cleavages of the ␤-amyloid precursor protein (APP) (1). The recognition that several forms of early onset AD are caused by mutations that alter A␤ production and that transgenic mice expressing these mutant alleles develop several features of AD, including A␤ plaques and memory deficits, have provided strong evidence that A␤ plays an important role in the disease process (2). In particular, small oligomers of soluble A␤ have been shown to be potent inhibitors of memory in rodents (3,4) and can be found in human cerebral spinal fluid and APP transgenic mice (3,5,6). However, how A␤ leads to a disruption in hippocampal synaptic function and deficits in memory remains unclear.
There is considerable evidence that A␤ can bind to several different molecules on the neuronal surface. It has been reported that a brief treatment of cultured neurons with trypsin reduces synthetic A␤ binding (7). Moreover, Klein and co-workers (8) have used a far Western technique to demonstrate the binding of synthetic A␤ oligomers called A␤-derived diffusible ligands to discrete proteins in cortical brain homogenates. Various laboratories have identified candidate receptors that may play a role in A␤ neurotoxicity, including the ␣7-type nicotinic acetylcholine receptor (9 -11), the metabotropic glutamate receptor 5 (mGluR5) (12), the ␤-adrenergic receptor (13), and the IR (14). It is unclear whether some or all of these cell surface proteins contribute to the rapid effects of A␤ on long term potentiation (LTP).
Significant progress has occurred over the last decade in elucidating the molecular basis of memory in the hippocampus (15). In particular, the synapses in the CA1 region undergo an NMDA receptor-mediated LTP that has been well studied from the molecular to the behavioral levels. Because these synapses are vulnerable to A␤-mediated toxicity, we examined five kinases (CaMKII, PKA, PKC, MAPK/Erk, and Akt/PKB (which is activated via phosphatidylinositol 3-kinase)) that are known to be necessary for early steps in the LTP cascade in order to determine how soluble A␤ affects their activation by neuronal stimulation. We observed that naturally secreted A␤ inhibited the activation of only a subset of these kinases in 14-dayold hippocampal cultures. Multiple pharmacological antagonists were then tested to determine whether any could mimic the selective pattern of kinase inhibition seen with soluble A␤ in older cultures, and only an antagonist of the IR family showed a closely similar pattern to that of soluble A␤ and disrupted LTP. Further investigation revealed that soluble A␤ co-immunoprecipitates with the IR and interferes with the autophosphorylation of the IR induced by insulin. Our results demonstrate that brief applications of physiological concentrations of soluble human A␤ selectively inhibit the IR and multiple downstream kinases that are required for hippocampal LTP.

MATERIALS AND METHODS
Cells-Primary neuronal hippocampal cultures were generated from embryonic day 18 wild-type Swiss Webster mouse pups. The hippocampus was dissected out in Hanks' balanced salt solution buffered with HEPES and dissociated via trypsin/ EDTA treatment. Cells were plated at 1 ϫ 10 6 cells on 3.5-cm dishes precoated with poly-DL-lysine and laminin. After 2 days of culturing in neurobasal medium with B-27 supplement and glutamax, cytosine arabinofuranoside was added to reduce glial proliferation. Half the medium was exchanged every 3-4 days. After 13 days in vitro, the medium was replaced with neurobasal medium containing 2 mM L-glutamine and 5 M AP5 (to reduce excitotoxicity).
The 7PA2 cell line described previously (16,17) stably expresses the AD-causing human APP V717F mutation in a CHO cell line and secretes biologically active forms of A␤. Conditioned medium (CM) containing secreted A␤ was prepared by incubating the cells overnight in serum-free medium. The A␤ CM was centrifuged at 1000 ϫ g to remove cellular debris and then concentrated in a YM-3 Centricon filter to obtain a 15ϫ stock solution. These samples were diluted to 1ϫ in artificial cerebral spinal fluid (ACSF) just prior to stimulation.
Neuronal Stimulation-After 14 days in vitro, cells were washed in ACSF containing 124 mM NaCl, 2.8 mM KCl, 3.6 mM CaCl 2 , 2 mM MgSO 4 , 1.25 mM NaH 2 PO 4 , 26 mM NaHCO 3 , 10 mM D-glucose, 0.4 mM sodium ascorbate, pH 7.4 (osmolarity 306 mM). The cells were equilibrated in ACSF for 1 h. Hippocampal neurons were treated for 20 min with 1ϫ CM from either untransfected CHO cells (control CM) or 7PA2 cells (A␤-rich CM) or else with specified pharmacological antagonists and then switched into the stimulating ACSF solution containing 126 mM NaCl, 2.8 mM KCl, 3.6 mM CaCl 2 , 1.25 mM NaH 2 PO 4 , 26 mM NaHCO 3 , 10 mM D-glucose, 0.4 mM sodium ascorbate, 0.1 mM glycine, and 5 M picrotoxin and also containing the CM or the antagonists. Cells were chemically stimulated for 20 min and then homogenized in lysis buffer containing 50 mM HEPES, 250 mM NaCl, 5 mM EDTA, 1% Nonidet P-40, pH 7.7, as well as complete protease inhibitors (10 g/ml leupeptin, 10 g/ml pepstatin, 1 g/ml aprotinin, and 400 g/ml EDTA) and phosphatase inhibitor mixtures I and II (catalog numbers P285 and P5726; Sigma). Cells were scraped from the tissue culture dish and sonicated for 5 s at 4°C. Samples were stored at Ϫ20°C until they were mixed with 2ϫ sample buffer.
For the co-immunoprecipitation studies in Fig. 4A, two hippocampi per condition were homogenized in lysis buffer containing 0.32 M sucrose, 50 mM Tris-HCl, pH 7.4, with complete protease inhibitors. Nuclei were pelleted, and the supernatants were treated with either 10% deoxycholate or 1% SDS for 2 h at 4°C. The samples were dialyzed overnight with three solution exchanges in phosphate-buffered saline. The samples were then mixed with 16 ml of 7PA2 CM with complete protease inhibitors for 1 h at room temperature. Samples were immunoprecipitated with anti-IR antibodies with continuous rocking overnight. Beads were washed three times with 0.5 M NaCl STEN buffer (18) without detergent. Because the interaction between the insulin receptor and A␤ appeared to be sensitive to detergent, the samples in Fig. 4B were prepared by a distinct method; synaptosomes were prepared from hippocampal brain tissue and sonicated for 5 s. Lysates were treated with A␤ CM for 2 h before immunoprecipitating with anti-insulin receptor antibodies that had been covalently cross-linked to protein A-Sepharose beads. Beads were washed three times with 0.5 M NaCl STEN buffer without detergent.
To test for the effects of A␤-rich CM on CREB phosphorylation (supplemental Fig. 1), primary neuron cultures were stimulated as before for 20 min but returned to normal ACSF for an additional 20 min before homogenization.
For Western blots, antibodies were used according to the manufacturer's recommendations. Blots were scanned using a Licor Odyssey system in two channels, enabling quantification of phosphorylated and control bands in the same gel.
Electrophysiology-Field potential recordings were made from coronal sections of postnatal day 16 -28 male and female Swiss Webster mice in compliance with Harvard University's animal resources and comparative medicine policies for use of laboratory animals. All recordings were made at room temperature. Electrodes were specifically placed just below the surface of the slice to maximize the exposure to circulating soluble A␤. The intensity of the stimulus was set to 20 -30% of the maximum evoked EPSP or until a population spike was elicited. Slices were perfused for 20 min in ACSF to establish a steady base line. A 1-ml aliquot of 15ϫ concentrated CM was thawed, added to 14 ml of circulating medium, and recirculated over the slice at 2.5-3 ml/min while being continuously aerated with 95% oxygen. LTP was induced 20 min later by delivering four 100-Hz stimuli every 5 min to the Schaeffer collaterals. The recording electrode was placed in the stratum radiatum, and the slope of the EPSP was monitored for 1 h after the last high frequency stimulation.
Whole cell recordings were made from hippocampal neuron cultures (13-15 days in vitro) grown on Bellco German glass coverslips. Electrodes were pulled to 3.5-5.5 megaohms and were filled with recording solution containing 110 mM potassium gluconate, 10 mM HEPES, 1 mM EGTA, 20 mM KCl, 4 mM NaCl, 2 mM MgATP, 0.25 mM Na 3 GTP, 10 mM phosphocreatine, pH 7.3. Excitatory postsynaptic currents (EPSCs) were sampled at 10 kHz and filtered at 5 kHz using an Axon Instruments 200B amplifier. Healthy neurons were selected using infrared/differential interference contrast. Series resistance was 10 -16 megaohms, was not compensated, and was checked periodically throughout the recording. Experiments where the series resistance deviated by more than 10% were discarded. Input resistance was between 270 and 390 megaohms.
Cultures were equilibrated in ACSF for 20 min. After breaking into the neuron, 10 M picrotoxin was added to the perfusion to inhibit GABAergic and glycinergic currents, along with the control or A␤-rich CM. Recordings of spontaneous EPSCs were made after 5 min to obtain base-line activity. Cells were then switched into current clamp mode and stimulated by perfusing the cultures with ACSF containing 0 mM Mg 2ϩ , picrotoxin, and glycine for 20 min, as described above. Normal ACSF containing picrotoxin was then washed in, and cells were switched back to voltage clamp mode. After 30 min, EPSCs were again recorded to obtain a poststimulation base line.
Statistics-For multiple comparisons, analyses of variance were performed with Tukey-Kramer post hoc tests. Dual comparisons were analyzed by Student's t tests. Error bars report the S.E.

RESULTS
We have previously reported that naturally secreted, soluble A␤ disrupts the potentiation of synapses in hippocampal slices using a pharmacological stimulation protocol to increase NMDA receptor activation (18). Similar pharmacological methods have been widely used to potentiate and synchronize synaptic activity for biochemical and electrophysiologic studies (18 -24). To confirm that soluble A␤ has a similar effect in neuron cultures as in brain slices, primary hippocampal neurons were isolated from wild-type mice, cultured for 14 days in vitro, and studied using whole-cell recording or biochemical techniques. The stimulation protocol induced an approximate doubling of the amplitude of spontaneous EPSCs in control conditions, as expected (supplemental Fig. 1B). In contrast, A␤ CM strongly inhibited this enhancement, such that the mean EPSC amplitude was not significantly different from base line (supplemental Fig. 1B).
The transcription factor CREB has been shown to be integral to the expression of the late stages of LTP. Consistent with previous findings (25)(26)(27)(28), A␤-rich CM interfered with the phosphorylation of CREB induced by pharmacological stimulation (supplemental Fig. 1C) (analysis of variance, Tukey-Kramer post hoc; stimulated and stimulated ϩ control CM are significantly different from unstimulated, stimulated ϩ AP5, and stimulated ϩ A␤ CM; p Ͻ 0.05, n ϭ 5). Thus, A␤-rich CM, but not control CM, inhibited the potentiation of EPSCs and the phosphorylation of CREB in primary hippocampal cultures, as expected.
Since CREB can be activated by several upstream kinase cascades, including CaMKII, PKA, and Erk/MAPK, we used this pharmacological stimulation paradigm to test whether soluble A␤ caused a universal inhibition of kinases or whether it was more selective. Among the potential kinases to examine, we chose PKA, PKC, CaMKII, Erk/MAPK, and Akt/PKB (which is downstream of phosphatidylinositol 3-kinase), all of which play a principal role in LTP and have been widely studied (reviewed in Ref. 29). The Src family of tyrosine kinases is also of interest but was beyond the scope of the present study. PKC, CaMKII, and Erk/MAPK all require phosphorylation for proper activation. Ser-660 is conserved on most PKC isoforms and is the last autophosphorylation step that enables PKC to become cytosolic and interact with membrane lipids (30). Thr-286 is an autophosphorylation site on CaMKII that is required for the enzyme to become Ca 2ϩ /CaM-independent (31). Activation of Erk/ MAPK is achieved by phosphorylation at Thr-202/Tyr-204 by mitogen-activated protein kinase/extracellular signal-regulated kinase kinase (MEK) (32). The phosphorylation of inositol 1,4,5-bisphosphate by phosphatidylinositol 3-kinase allows for activation of Akt/PKB, which can be detected by its phosphorylation at Ser-473 (33). Although PKA itself is activated by cAMP, not by phosphorylation, we used phospho-specific antibodies that recognize the consensus PKA phosphorylation motif (RXXT* or RRXS*) on its substrates. Additional controls were run to demonstrate that this antibody is a reliable indicator of activated PKA (supplemental Fig. 2A).
Homogenates of 14-day-old hippocampal neuron cultures were collected 20 min after initiating the pharmacological stimulation. As shown in Fig. 1A, all five kinases were activated in the control conditions (lane 2, positive control), whereas all were inhibited from activation by the NMDA receptor antagonist AP5 (lane 3, negative control). None of the kinases was significantly activated by the addition of either control or A␤-rich CM alone (i.e. without stimulation) (lanes 4 and 5), and all five kinases were activated by stimulation in the presence of control CM (lane 6). Intriguingly, A␤ CM had no effect on the stimulation-induced activation of PKA and PKC, yet Erk/ MAPK, Akt, and CaMKII were significantly inhibited (lanes 7 and 8). Importantly, the inhibition could be attributed to A␤ species larger than monomer, as the inhibition of Erk/MAPK phosphorylation was reproduced using size exclusion chromatography fractions containing A␤ dimers and trimers that migrate at 8 -12 kDa (34) but not with size exclusion chromatography fractions from the same run that contain only 4-kDa A␤ monomers (supplemental Fig. 2B). Compared with controls, the neurons treated with A␤-rich CM showed significant impairment in the stimulus-induced activation of Erk/MAPK, Akt, and CaMKII (Fig. 1B). In contrast, there was no significant difference between control and A␤-treated neurons in levels of phospho-PKC or phospho-PKA substrates (Fig. 1B). Thus, soluble human A␤ specifically prevents the activity-dependent phosphorylation of Erk/MAPK, CaMKII, and Akt while sparing the activation of PKA and PKC.
Previous results have shown that mGluR5 and c-Jun N-terminal kinase antagonists could rescue high frequency stimulation-induced LTP from the inhibitory effects of synthetic A␤ peptide (12). We tested whether these inhibitors would also rescue the effects of soluble, secreted A␤ on Erk/MAPK, Akt, and CaMKII. Under our conditions, c-Jun N-terminal kinase I/II antagonists did not rescue the activation of these three kinases that are inhibited by the soluble A␤-rich CM (data not shown). However, it is likely that c-Jun N-terminal kinase I/II function in parallel to the signal transduction cascades described above and could have beneficial effects downstream of these signaling events. In contrast, an mGluR5 antagonist (SIB1757) did partially restore the activation of these three kinases in the presence of the A␤ CM (stimulated ϩ A␤ CM ϩ SIB1757 was no longer significantly different from stimulated ϩ control CM, p Ͼ 0.05) (data not shown), a result that is consistent with an mGluR5 antagonist ameliorating the effects of synthetic A␤ on LTP. However, because the mGluR5 antagonist also elevated base-line PKA activity (i.e. without pharmacological stimulation), we concluded that our assay was not amenable to further study of the interaction between mGluR5 and the A␤-mediated interference with synaptic plasticity.
Because several candidate receptor proteins have been suggested to mediate the toxic effects of synthetic A␤, we surveyed multiple pharmacologic agents that inhibit these transmembrane receptors to determine if they would replicate the selective pattern of kinase inhibition observed above with A␤-rich CM ( Fig. 2A). Some compounds, such as KT5720, forskolin, SIN-1, and LY294002, were chosen to confirm the specificity of our phospho-specific antibodies. Others, such as ␣-bungarotoxin (11), pronethalol (13), and AG1024 (14), were selected to test specific hypotheses about potential A␤ receptor targets. Our initial screen of 12 different pharmacological agents focused on the response of Erk/ MAPK and PKA, since these kinases had shown a differential response to soluble A␤ (above) and since the relevant antibodies proved to be highly specific. Among all 12 compounds tested, only AG1024, a tyrophostin that is selective for the insulin receptor family of tyrosine kinases, inhibited Erk/MAPK phosphorylation to a similar extent as the A␤-rich CM while preserving normal PKA activation (Fig. 2B). We did not find evidence that ␣7 nicotinic acetylcholine receptor antagonists reproduced the same selective kinase inhibition pattern as soluble A␤ under these experimental conditions. The results in Fig. 2B led us to determine whether Akt, CaMKII, and PKC were similarly affected by both soluble A␤ and the insulin receptor antagonist (AG1024).
An additional set of experiments was therefore performed to test the effects of modulating insulin receptor activity on all five kinases. As shown in Fig. 2C and quantified in interfere with LTP. To test this idea, we made field potential recordings in the CA1 region of mouse hippocampal slices from p16-p28 animals (Fig. 3A), and a dose response was performed with AG1024. Even after 30 min postinduction, all doses of AG1024 tested had significant, graded effects on potentiation. Therefore, like A␤, inhibition of the insulin receptor family of tyrosine kinases adversely affects LTP. Based on the results shown in Fig. 2, C and D, we next examined whether adding exogenous insulin could reverse the A␤-mediated inhibition of LTP. Slices were perfused with 1 M insulin for 20 min prior to the addition of A␤-rich CM. As expected, LTP was inhibited in slices treated with A␤-rich CM alone, as reported previously (18) (Fig. 3B). However, the preapplication of 1 M insulin significantly increased the amount of potentiation recorded 60 min after LTP induction. Therefore, exogenous insulin can at least partially reverse the effects of soluble A␤ on Erk/MAPK, CaMKII, and Akt as well as the inhibitory effects of soluble A␤ on LTP in hippocampal slices. These results are consistent with the partial rescue of kinase activation in neuronal cultures treated with A␤ and insulin (Fig. 2, C and D).
The preceding findings demonstrate that soluble A␤ and AG1024 share similar effects on kinase signaling and synaptic plasticity, but is the effect of A␤ on neurons mediated, at least in part, through the insulin receptor pathway? A recent report demonstrated that synthetic A␤ can disrupt insulin receptors extracted from peripheral tissue (14), but the consequences of this in neurons have not been examined. To test for a direct interaction between A␤ and neuronal insulin receptors, hippocampal neuronal homogenates were solubilized in either deoxycholate or SDS, dialyzed against phosphate-buffered saline (to remove the detergent), and then incubated with A␤-rich CM or else with heat-denatured A␤-rich CM. Small amounts of oligomeric A␤ were co-immunoprecipitated by an antibody specific to the insulin receptor using the deoxycholate-treated (but not SDS-treated) samples (Fig. 4A). Importantly, boiled A␤ CM (which lacks biological activity) (18) no longer bound to insulin receptors.
The low recovery of A␤ in the presence of detergent prompted us to test alternative methods of detecting a biological interaction. In Fig. 4B, synaptosomes were isolated from hippocampal brain tissue and sonicated prior to incubating with A␤-rich CM for 2 h   ( Fig. 4A). Co-immunoprecipitations were performed as before. Under these conditions, several forms of A␤-specific bands were co-precipitated by the insulin receptor antibody. Importantly, this interaction was entirely competed away by 1 M insulin. The presence of even trace amounts of detergent during the co-incubation of the neuronal lysates and the A␤-rich CM or during washes following immunoprecipitation disrupted this interaction. These results suggest that synaptotoxic forms of soluble A␤ can bind to neuronal insulin receptor complexes and that the interaction is highly sensitive to detergent. Finally, to determine directly whether A␤ disrupts insulin receptor signaling in neurons, we performed a dose response with insulin and assayed insulin receptor autophosphorylation. Hippocampal cultures were pretreated with control CM or A␤-rich CM for 20 min and then stimulated with varying concentrations of insulin ( Fig. 4C; quantified in Fig. 4D). In the presence of soluble A␤, low concentrations of insulin failed to stimulate autophosphorylation of the insulin receptor, as they did in the presence of the control CHO-CM. However, as the concentration of insulin increased above 5 nM, the insulin receptor was effectively activated, overcoming the effect of soluble A␤ (Fig. 4, C and D). These data are consistent with the results shown in Figs. 2 and 3, which demonstrate that nanomolar levels of insulin can at least partially reverse the effects of A␤ on selective kinase activation and LTP in the hippocampus.

DISCUSSION
The inhibitory effect of human A␤ on LTP in the hippocampus has been well documented by several laboratories. The objective of the present study was to address how A␤ interferes with some of the signal transduction cascades that mediate activity-dependent synaptic plasticity. By applying soluble A␤ to cultured hippocampal neurons for short intervals and stimulating immediately thereafter, our experiments were designed to test early events in the inhibition of synaptic function by naturally secreted human A␤, before significant compensation could occur. Our observation that of the five signaling kinases studied, only Erk/MAPK, CaMKII, and Akt/PKB were significantly inhibited, not PKA and PKC, demonstrates a selectivity of A␤-mediated effects on synapses. Consistent with this finding, Zhao et al. (35) have reported that 200 nM synthetic A␤ reduced phosphorylation of the AMPA-type glutamate receptors at the CaMKII site (Ser-831) but not at the PKA site (Ser-845). Taken together, these results demonstrate that an initial effect of A␤ on synapses is a selective impairment of specific elements of the signal transduction machinery rather than a global dysregulation of synaptic function.
In accord with previous reports, we observed that the phosphorylation of the transcription factor CREB at Ser-133, which is necessary for the expression of the late phases of LTP, was inhibited by A␤. Multiple kinase pathways are known to converge on CREB to initiate its activation (e.g. see Refs. 36 and 37). Because PKA (which can phosphorylate CREB) (38) was activated normally in our A␤-exposed neuronal cultures, we conclude that the contribution of the PKA pathway to CREB phosphorylation must be relatively small under these conditions. Nevertheless, augmenting the PKA response with rolipram may surmount some of the effects of A␤ and explain its beneficial effects in APP Tg mice performing memory tasks (39).
Our observation that Erk/MAPK is inhibited by soluble A␤ appears to contradict some earlier reports in the literature (9,40), which suggested that synthetic A␤ causes an activation of Erk/MAPK in hippocampal slices or in slice cultures. However, there are three key differences among these studies. First, we examined the activation Erk/MAPK following stimulation, whereas both earlier groups documented the direct biochemical effects of synthetic A␤ on Erk/MAPK phosphorylation. Nevertheless, in our control conditions, we did not see a significant change in Erk/MAPK phosphorylation with A␤-rich CM alone. Second, our experiments employ subnanomolar quantities of cell-derived soluble A␤, whereas both Chong et al. (40) and Dineley et al. (9) used much higher concentrations of synthetic A␤. Third, Dineley et al. (9) reported a rapid activation of Erk within 5 min that then became inactivated by 10 min, which is well before our 20 min time point. On the other hand, our findings are consistent with Daniels et al. (41), who reported a decrease in Erk/MAPK phosphorylation in N2a mouse neuroblastoma cells treated with synthetic A␤.
We also observed that soluble A␤ did not inhibit the normal phosphorylation of PKC (Ser-660) after stimulating hippocampal neurons. This is the final autophosphorylation step in PKC activation, enabling the enzyme to respond to Ca ϩ2 and DAG. Our results demonstrate that PKC is activated normally in the presence of soluble A␤, but they cannot rule out the possibility that PKC would be incompletely stimulated if, for instance, Ca 2ϩ transients or phospholipase C activity were disrupted by A␤. In particular, the observation that CaMKII activity is impaired by soluble A␤ (perhaps because of insufficient Ca 2ϩ influx during synaptic activity) means that PKC may not achieve full activation. Although the antibody we used to detect phospho-PKC recognizes multiple isoforms of PKC, it does not detect the ␥ or form, the activation of which could still be compromised by A␤. The inhibition of Thr-286 phosphorylation on CaMKII suggests that soluble A␤ can interfere with a very early step in the LTP process, because CaMKII is localized in the postsynaptic density, is activated by Ca ϩ2 /CaM, and is directly downstream of the NMDA-type glutamate receptor and voltage-gated calcium channels. Our results would therefore be consistent with a model in which A␤ modulates calcium influx either directly or indirectly. It has recently been shown that synthetic A␤ causes an internalization of NMDA receptors from the cell surface, which could account for such a disruption in CaMKII activa-tion (10). However, the persistence of PKA activity in the presence of soluble A␤ in our experiments and the observation that AP5 inhibits all five kinases suggest that a nominal amount of calcium still enters the neurons, perhaps through NMDA receptors or calcium channels (42), to activate adenylate cyclases. The differential effects of soluble A␤ on PKA and CaMKII could be explained by differences in the dissociation constants. It has been reported that Ca 2ϩ /CaM-activated adenylate cyclase has a higher affinity for Ca 2ϩ than CaMKII (43,44) and that PKA is at least partially active under resting conditions (45).
Of particular interest is our observation that an insulin receptor family antagonist (AG1024) showed a similar profile of kinase inhibition as did soluble A␤. Although relatively little is known about the function of insulin receptors in neurons, insulin receptors and TrkB receptors have convergent signaling pathways (46). Therefore, the signaling cascades activated by the trophic factor BDNF may similarly be activated by insulin. In Fig. 2, we observed that Erk/MAPK phosphorylation was impaired by AG1024 (Fig. 2B), yet insulin itself had little effect on activation of Erk/MAPK (Fig. 2C). These results would suggest that the insulin receptor plays only a small roll in total Erk/ MAPK activation. In contrast, insulin receptor activity plays an important role in gating Erk/MAPK, CaMKII, and Akt activation, perhaps by regulating early events in glutamate receptor neurotransmission.
Additional reagents for insulin receptor biology will be required to further explore its importance in neuronal function and A␤-mediated synaptic effects. Currently, AG1024 is the only commercially available pharmacological inhibitor of the insulin receptor that is considered to be selective. Blocking antibodies have generally been shown to work well on peripheral, but not neuronal, insulin receptors (48). However, the antibody GR07 is reported to have blocking effects, and the addition of dialyzed GR07 to primary neurons partially blocked Erk phosphorylation (Fig. 2B), supporting a specific role of the insulin receptor in gating Erk/MAPK activation. Our observation that co-application of 1 M insulin with soluble A␤ largely restored Erk/MAPK, Akt, and CaMKII activations as well as LTP supports the hypotheses that A␤ and insulin share common signal transduction pathways and that at least one aspect of A␤-mediated synaptotoxicity may be through disruption of insulin signaling, either directly or

. A␤ co-immunoprecipitates with the IR and inhibits insulin-induced activation of the IR.
A, hippocampal lysates were solubilized in 10% deoxycholate (DC) or 1% SDS. Lysates were dialyzed prior to mixing with A␤ CM. Insulin receptors were immunoprecipitated, and blots were probed with the anti-A␤ specific antibody 6E10. Two oligomeric A␤ species showed a weak association with the insulin receptor (lane 3). B, without adding detergents, synaptosomes were sonicated prior to treatment with A␤ CM for 2 h. As in A, IRs were immunoprecipitated, and blots were probed with the anti-A␤ antibody 6E10. Under these conditions, the insulin receptor co-immunoprecipitated with a several A␤ forms found in the CM. This co-immunoprecipitation was disrupted by the addition of exogenous insulin. Thus, the interaction between the IR and A␤ was disrupted by detergents (SDS), by heat-denaturing the A␤ prior to nutating with lysates (boil), and by exogenous insulin (n ϭ 3). C, hippocampal neurons were pretreated with A␤ CM or control CM, stimulated with varying concentrations of insulin, lysed, and then subjected to immunoprecipitation with anti-IR antibodies. Treatment with A␤-rich CM, but not control CM, inhibited IR autophosphorylation at low concentrations but not at high concentrations of insulin. indirectly (49 -52). We provide evidence for a potentially direct competition between insulin and A␤ on insulin receptor function (Fig. 4C). However, because targeted deletion of insulin receptors in neurons is not known to recapitulate all of the hallmarks of AD (53), additional mechanisms of A␤-mediated toxicity are likely to play a role, along with other factors contributing to AD pathogenesis, such as tau and apolipoprotein E.
In conclusion, physiological concentrations of soluble oligomers of human A␤ have potent effects on select signal transduction cascades that mediate LTP. Acute applications of subnanomolar concentrations of soluble A␤ are sufficient to impair Erk/MAPK, CaMKII, and Akt/PKB activation. The persistent inhibition of these pathways in the brain during the development of AD could deprive synapses of this form of plasticity and actively suppress spine maintenance and/or formation (42,54). The demonstration that an antagonist of the insulin receptor family mimics the selective effects of soluble A␤ on kinase activation raises the possibility that A␤ synaptotoxicity may be mediated in part through this important class of receptors.