Endogenous Amyloid-β is Necessary for Hippocampal Synaptic Plasticity and Memory

Animals

C57BL/6, male, wild-type mice (3–4 months old) were obtained from a breeding colony housed in the animal facility at Columbia University. α7-nAChR knockout (KO) mice have been previously described.²⁹ APP KO mice were purchased from Jackson Laboratories (Bar Harbor, ME).

Anti-Aβ Antibodies

The antirodent Aβ monoclonal antibody, JRF/rAb2, was generated at Janssen Pharmaceutica (Beerse, Belgium). Balb/c mice were immunized with rodent Aβ-derived peptide (amino acids 1–15) coupled to keyhole limpet hemocyanin (KLH) according to the manufacturer's instructions (Pierce, Rockford, IL). Mice were boosted every 2 weeks with 100μg KLH-coupled peptide, first in complete Freund adjuvant, and subsequently in incomplete Freund adjuvant. Mouse spleen cells were fused with Sp2/0 cells by a modified procedure of Kohler and Milstein. The hybridomas were seeded in 96-well plates and screened after 10 days in a direct enzyme-linked immunosorbent assay (ELISA) on bovine serum albumin-coupled rodent Aβ peptide and confirmed on noncoupled rodent Aβ peptide. Positive cells on free rodent Aβ1–15 were immediately subcloned, and positive clones were frozen in liquid nitrogen. All hybridomas were grown in Dulbecco modified Eagle medium supplemented with 10% fetal calf serum (Hyclone Europe, Edegem, Belgium), 2.5% ESG Hybridoma supplement (Elscolab, Kruibeke, Belgium), 2% HT media supplement (Sigma-Aldrich, St Louis, MO), 1mM sodium pyruvate, and 2mM L-glutamine and penicillin (100U/ml) and streptomycin (50mg/ml) (Life-Technologies, Paisley, UK). Cells were incubated in a humidified 8% CO₂ air incubator. This antibody recognizes an epitope within residues 1–15 of the rodent Aβ sequence, a region that contains 3 amino acid differences compared to the human sequence. This antibody binds synthetic and endogenous, brain-derived rodent Aβ (both Aβ₄₀ and Aβ₄₂) with high affinity by ELISA³⁰ and immmunoprecipitation,^30,31 and completely fails to recognize human Aβ (Supplementary Fig 1). In addition, in an aged APP transgenic mouse with robust β-amyloid deposition, which allows for detection of the abundantly deposited Aβ, JRF/rAb2 monoclonal antibody (mAb) detects codeposited murine Aβ by immunohistochemistry.³⁰ No cytotoxicity has been seen when neuronal primary mouse cultures as well as mouse N2a cells have been grown in the presence of 1μg/ml JRF/rAb2 mAb-containing media for 24 hours, nor has intraperitoneal injection of 0.5 mg JRF/rAb2 mAb daily over 1 week shown any toxic effects in wild-type or human APP transgenic mice (data not shown). The working concentration of 4μg/ml used in the electrophysiology experiments was based upon the concentration of JRF/rAb2 mAb typically employed at Janssen Pharmaceutica for immunoprecipitation, and the amount injected into mice (4μg/hippocampus) was calculated from this. As a control, we used an IgG2a isotype-matched control antibody that does not bind to any rodent proteins (termed control mAb, Tepnel Life Sciences, Manchester, UK).

Aβ preparation, electrophysiological studies, infusion technique, and behavioral studies have been previously described²⁹ (see also Supplementary Material and Methods for a detailed description).

siRNA Preparation and Testing

siRNAs were designed according to published guidelines with dTdT 3′ overhangs³² (see Supplementary Material and Methods for a detailed description). The sequence for the sense strand of the central 19-nucleotide double-stranded region was GCAC-TAACTTGCACGACTATG. Short hairpin RNA (shRNA) with the same sequence has been previously shown to acutely knock down APP function.³³ The control siRNA sequence used was AAUGGUGUGGCCAUUGACCCC.

Aβ Measurement

Following euthanasia, mouse hippocampi were quickly removed and frozen at given time points (along with the control regions of the cerebellum). The specimen was homogenized in 1ml of HEPES-sucrose buffer (20mM HEPES/NaOH pH 7.4, 1mM ethylenediaminetetraacetic acid, 1mM ethyleneglycoltetraacetic acid, 250mM sucrose) supplemented with protease inhibitors. The tissue homogenates were treated with diethanolamine to extract soluble Aβ₄₂. Endogenous murine Aβ levels were determined by sandwich ELISAs using the mouse Aβ_x-42 kit from Covance (Princeton, NJ) or the Aβ_1–42 kit from Invitrogen (Carlsbad, CA), according to the manufacturer's protocol. For rescue experiments of siRNA effect on LTP, 8 hippocampal slices were pooled together and homogenized prior to performing ELISA measurements.

Statistical Analyses

Experiments were performed in blind. Results were expressed including standard error of the mean with level of significance set at p < 0.05. Results were analyzed by Student t test (pairwise comparisons) or analysis of variance for repeated measures (multiple comparisons) with treatment condition as main effect. Planned comparisons were used for post hoc analysis. For analysis of behavior tests, experiments were designed in a balanced fashion, and mice were trained and tested at each of the different conditions in 3 or 4 separate experiments.

Depletion of Endogenous Aβ Impairs Hippocampal LTP

To investigate the role of Aβ in synaptic plasticity, we blocked endogenously produced Aβ with the monoclonal antibody JRF/rAb2. This antibody recognizes a rodent-specific epitope within the first 15 amino acids of rodent Aβ₄₀ and Aβ₄₂ with high affinity under multiple conditions (nondenaturing, denaturing, monomeric, oligomeric/fibril)^30,31 (see Supplementary Fig 1). When hippocampal slices were perfused with JRF/rAb2 mAb (4μg/ml, for 20 minutes prior to tetanizing Schaffer collateral fibers), LTP was dramatically reduced (Fig 1A). By contrast, the IgG2a isotype-matched negative control antibody (4μg/ml; termed control mAb) did not affect LTP (see Fig 1A). Nontetanized slices treated with JRF/rAb2 mAb alone or with control mAb alone showed no change in basal synaptic transmission (data not shown). To clarify whether Aβ acted on the LTP maintenance phase or induction, we applied the antibody after tetanization. However, perfusion with JRF/rAb2 mAb 20 minutes post-tetanus did not affect LTP (Supplementary Fig 2), suggesting that LTP maintenance is not affected by depletion of endogenous Aβ.

Because the JRF/rAb2 mAb might have effects independent of Aβ depletion, such as binding to APP, βCTF, sAPPα, or other nonspecific targets, our next goal was to determine whether the impairment of LTP was specifically mediated by Aβ. We performed rescue experiments with human Aβ₄₂, which is not recognized by the antibody (see Supplementary Fig 1). Addition of an aged preparation of human Aβ₄₂ (200pM) (Supplementary Fig 3) concurrent with the JRF/rAb2 mAb rescued the LTP deficit (see Fig 1B). Furthermore, a higher concentration of Aβ₄₂ (300pM) in addition to JRF/rAb2 increased LTP by ~50% (see Fig 1B), resembling the enhancement of LTP that we had previously demonstrated in slices perfused with 200pM Aβ₄₂ alone.²⁹ JRF/rAb2 mAb paired with Aβ₄₂ without tetanization did not affect baseline transmission (data not shown). Scrambled Aβ₄₂ peptide was used as a control for any nonspecific effects of Aβ₄₂ due to potential changes in osmolarity of the bath solution and failed to rescue the reduction in LTP by JRF/rAb2 mAb (~30% less of tetanized vehicle-treated slices; data not shown). Finally, a preparation enriched in monomers was not capable of rescuing LTP in slices that were concomitantly treated with mAb (Supplementary Fig 4), suggesting that oligomeric forms of Aβ are likely to be critical for normal synaptic plasticity.

To confirm our findings obtained using JRF/rAb2 mAb, we designed a siRNA specific for murine APP linked to Penetratin 1 (Pen1-APP-siRNA) peptide for intraneuronal delivery based on an shRNA sequence previously shown to acutely knock down APP expression.³³ In preliminary experiments, we demonstrated by quantitative Western blot analysis that full-length APP was reduced by approximately 60% in primary hippocampal cultures that were exposed to Pen1-APP-siRNA (80nM) for 24 hours prior to cell collection (Supplementary Fig 5), whereas a control siRNA did not affect APP expression. Next, we implanted cannulas bilaterally into mouse dorsal hippocampi and after 6 to 8 days injected siRNA (80nM in a final volume of 2μl over 1 minute, 24 hours prior to investigating LTP). APP suppression dramatically reduced LTP, whereas control siRNA failed to reduce LTP (see Fig 1C). Also, the effect of APP suppression by siRNA on synaptic plasticity was due, at least in part, to Aβ₄₂ deficiency, because the addition of 200pM Aβ₄₂ ameliorated LTP (see Fig 1C). Finally, we excluded the possibility that siRNA- or antibody-induced reduction in LTP was offset by the LTP-enhancing effects of Aβ via independent pathways. When we performed experiments in 2-month-old APP null mice exposed to JRF/rAb2 mAb or APP-siRNA, we found that neither the antibody nor APP-siRNA affected LTP (Supplementary Fig 6). Thus, our combined biochemical and genetic approaches indicate that Aβ₄₂ is required for the induction of LTP in the hippocampus.

Our next goal was to determine the threshold of Aβ₄₂ needed for normal synaptic plasticity. In these experiments, APP-siRNA was injected into hippocampi, after 24 hours mice were sacrificed, and slices were cut from the hippocampi. Following treatment with vehicle or human synthetic Aβ₄₂ at different concentrations (100, 200, 300pM), LTP was measured. Levels of endogenous Aβ₄₂ after siRNA treatment were also measured and found to be equal to 79.35 ± 3.11pM. Given that a complete rescue of potentiation was observed with 300pM Aβ₄₂ (Supplementary Fig 7), it is likely that the threshold of Aβ₄₂ needed for normal synaptic plasticity is ~380pM.

Endogenously Produced Aβ Is Required for the Induction of Memory

Given that LTP is thought to be associated with learning and memory,³⁴ we next assessed the effects of endogenous Aβ on hippocampal-dependent memory. JRF/rAb2 mAb, control mAb (4μg/μl, in a final volume of 1μl over 1 minute) or vehicle was bilaterally infused through cannulas, and after 15 minutes animals were tested for reference and contextual fear memory. We first tested reference memory by using the Morris water maze (MWM), a widely used spatial learning test known to require hippocampal function and assessing reference memory.³⁵ Mice treated with JRF/rAb2 mAb prior to each of the 6 sessions revealed a longer latency to find the hidden platform compared to vehicle-infused mice (Fig 2). The probe trial performed after removal of the platform demonstrated that JRF/rAb2 mAb-treated mice spent less time than vehicle-treated mice in the target quadrant (TQ), where the platform had been located during training. Infusion with control mAb or JRF/rAb2 mAb immediately after training did not affect mouse performance during the hidden platform and the probe trials. A visible platform trial performed after the probe trial did not reveal any significant difference in the time to reach the platform among the 4 groups tested (data not shown). Memory impairment was specifically mediated by Aβ, because application of 200pM Aβ₄₂ in addition to JRF/rAb2 mAb rescued the behavioral deficits during the hidden and probe tasks. Infusion with a higher concentration of Aβ₄₂ (250pM but not 300pM) and JRF/rAb2 mAb revealed faster learning compared to vehicle-infused mice, and planned comparisons showed a significant difference from the second to the fourth sessions. Given that the probe trial performed after the sixth session did not reveal further memory enhancement in the rescue experiments with 250pM Aβ₄₂, we postulate that excessive training may have obscured any difference in the time spent in the TQ. Thus, we performed additional experiments in which mice infused with 250pM Aβ₄₂ and JRF/rAb2 mAb, with JRF/rAb2 mAb alone, or with vehicle received only 3 training sessions with the hidden platform and then were tested in the probe trial. This experimental paradigm revealed a difference between mice treated with 250pM Aβ₄₂ and JRF/rAb2 mAb and those treated with vehicle and between mice treated with JRF/rAb2 mAb alone and those treated with vehicle during the hidden platform task (Supplementary Fig 8a) and the probe trial performed after the third hidden-platform session (see Supplementary Fig 8b). These findings were confirmed by infusing mice with Pen1-APP-siRNA 24 hours prior to performing the MWM test. APP suppression severely impaired reference memory, whereas control siRNA did not affect cognitive performance. Moreover, addition of 200pM Aβ₄₂ ameliorated the memory deficits. Thus, our combined biochemical and genetic approaches indicate that Aβ₄₂ is necessary for reference memory in hippocampus.

To extend our conclusions to an additional type of memory, we examined whether depletion of Aβ using JRF/rAb2 mAb (15 minutes prior to the foot shock) affects freezing in the contextual fear paradigm. Freezing was found to be unaffected following treatment with mAb during the training phase of the fear conditioning test (Fig 3). However, when contextual fear learning was assessed 24 hours later, JRF/rAb2 mAb-treated mice showed less freezing than vehicle-treated mice. By contrast, mice infused with control mAb or JRF/rAb2 mAb post-training showed normal freezing. In interleaved experiments, 200 or 300pM Aβ₄₂ concurrent with JRF/rAb2 mAb re-established normal freezing; however, 250pM Aβ₄₂ further enhanced freezing compared to vehicle-injected mice, resembling the enhancement of fear memory that we had previously demonstrated in mice infused with 200pM Aβ₄₂ alone.²⁹ The genetic approach confirmed that Aβ₄₂ is required for induction of associative memory. Indeed, when we infused mice with Pen1-APP-siRNA (24 hours prior to performing behavioral analysis for contextual memory), contextual fear memory was impaired, whereas perfusion with control siRNA did not affect the cognitive performance. Moreover, infusion with human synthetic Aβ₄₂ (200pM) rescued the memory deficit. Thus, these findings strongly support the hypothesis that, similar to LTP and reference memory, Aβ₄₂ is required for induction of associative memory.

Hippocampal Aβ42 Production Is Enhanced during Memory Induction

We measured hippocampal levels of Aβ_x-42 in mice that were trained for contextual fear learning and sacrificed at different intervals after the electric shock. Mice sacrificed at 1 minute (but not at 5 and 30 minutes) showed a significant increase in peptide levels compared to a naive control group (Fig 4). We obtained similar results when we measured Aβ₄₂ levels using a different antibody that recognizes Aβ_1-42. By contrast, cerebellar Aβ_x-42 levels were not modified at 1 minute after training. Interleaved experiments testing different conditions, including a group exposed to the context without tone and foot shock, a group exposed to the tone alone, and a group exposed to foot shock without context association showed no changes in Aβ_x-42 levels, with the exception of the cohort exposed to the tone plus foot shock. These findings strongly support the hypothesis that hippocampal Aβ₄₂ production is enhanced during memory induction.

Depletion of Endogenous Aβ Impairs Hippocampal PTP

Given that Aβ levels are regulated by synaptic activity^25,26 and that Aβ₄₂ enhances transmitter release during tetanus,²⁹ we examined post-tetanic potentiation (PTP), a type of synaptic plasticity that reflects a tetanus-induced enhancement of transmitter release. As previously demonstrated,²⁹ Aβ₄₂ (200pM) enhanced PTP, whereas JRF/rAb2 mAb (4μg/ml, 20 minutes prior to tetanus) reduced it (Fig 5A). Most importantly, the impairment of PTP due to the effect of the mAb was rescued by concomitant application of 200pM Aβ₄₂, with 300pM Aβ₄₂, producing a more pronounced enhancement similar to 200pM Aβ₄₂ alone (see Fig 5B). Taken together, these findings strongly support the hypothesis that endogenously produced Aβ is critical for normal transmitter release during tetanus.

The Effect of Endogenous Aβ on Synaptic Plasticity Involves α7-Nicotinic Acetylcholine Receptors

We have previously shown that α7-nicotinic acetylcholine receptors (nAChRs) are involved in synaptic plasticity and memory and play a role in enhancement upon the addition of exogenous Aβ₄₂.²⁹ Here, we determined whether α7-nAChRs are involved in the reduction of LTP by depletion of Aβ₄₂ levels. When hippocampal slices from α7-nAChR KO mice were perfused with JRF/rAb2 mAb (4μg/ml, 20 minutes prior to tetanus), LTP was no longer reduced (Fig 6). By contrast, JRF/rAb2 mAb reduced LTP in slices from wild-type littermates. Consistent with these findings, PTP was not affected by JRF/rAb2 mAb in slices from α7-nAChR KO mice, whereas the mAb was capable of impairing PTP in slices from wild-type littermates. Thus, involvement of α7-nAChRs is likely to be necessary for the effect of endogenous Aβ₄₂ on transmitter release.