Changes in dendritic spine number and morphology were detectable within one hour of incubation with 7PA2-CM (Fig. 2A–F). Spine number was significantly reduced by incubation with 7PA2-CM containing 40pM or 80pM Aβ (Fig. 2), average length of remaining protrusions increased 35%, and spine head width decreased 22–28%. Spine numbers did not decrease with exposure to 7PA2-CM containing 4pM Aβ, but their morphology was significantly affected, with protrusions becoming longer and thinner on average (Fig. 2C, G–I).

Given the dramatic effects on synaptic morphology induced by soluble Aβ described above, we next evaluated whether these translated into functional deficits at hippocampal synapses. Miniature excitatory postsynaptic currents (mEPSCs) were recorded 2–3 hours after incubation in the presence of CHO-CM or 7PA2-CM. As predicted from the observed decrease in the numbers of both presynaptic synaptophysin clusters and postsynaptic dendritic spines, we observed that both the frequency and amplitude of mEPSCs was significantly reduced in 7PA2 treated cells (Fig. 7A, B). Despite these decreases, the time constants for excitatory currents were not significantly changed (Fig. 7C), indicating that the properties of the remaining excitatory receptors were unaltered. Together, our data demonstrate that soluble Aβ induces within hours both morphological and functional changes at excitatory synapses.

To assess whether the rapid effects on postsynaptic structure represent early stages of neurotoxicity, we incubated rat hippocampal neurons with 7PA2-CM continuously for one or two days. Such prolonged incubations did not induce detectable neuronal cell death (data not shown). After one day morphological changes in dendrites were still present (see below, Fig. 4). However, to our surprise, by two days dendritic spines spontaneously recovered their shape and partially their number (Fig. 3). To exclude a possible reduction in Aβ efficacy or concentration we incubated “virgin” hippocampal neurons, which had never seen 7PA2-CM, for two hours with medium from neurons that had been incubated with 7PA2-CM for 2 days. Such “neuronally conditioned” 7PA2-CM was still able to affect the stability and morphology of dendritic spines (Fig. 3), indicating that Aβ efficacy was preserved.

To assess whether the rapid effects on postsynaptic structure represent early stages of neurotoxicity, we incubated rat hippocampal neurons with 7PA2-CM continuously for one or two days. Such prolonged incubations did not induce detectable neuronal cell death (data not shown). After one day morphological changes in dendrites were still present (see below, Fig. 4). However, to our surprise, by two days dendritic spines spontaneously recovered their shape and partially their number (Fig. 3). To exclude a possible reduction in Aβ efficacy or concentration we incubated “virgin” hippocampal neurons, which had never seen 7PA2-CM, for two hours with medium from neurons that had been incubated with 7PA2-CM for 2 days. Such “neuronally conditioned” 7PA2-CM was still able to affect the stability and morphology of dendritic spines (Fig. 3), indicating that Aβ efficacy was preserved.

Nicotinic acetylcholine receptors (nAchRs) and NMDA receptors (NMDARs) have been implicated in AD pathology, (D’Andrea and Nagele, 2006;Oddo and LaFerla, 2006;Sonkusare, et al., 2005). To test whether these receptors mediate the rapid effects of soluble Aβ on dendritic spines, we incubated hippocampal neurons for 2 hours with 7PA2-CM in the presence or absence of the noncompetitive NMDAR antagonists memantine (1μM) or MK801 (20μM), or the nAChR antagonist hexamethonium (C6; 100 μM). All three compounds significantly attenuated the effects of Aβ on spine morphology (Fig. 8), although, at the concentrations used, only C3 completely prevented changes in all three morphological parameters (spine density, spine length, and spine width).

In control neurons few spines displayed either rapid elongation or overall shrinkage (Fig. 5A₁; Supplemental Movie S1) although most exhibited the typical ‘morphing’ behavior of the spine head (Calabrese and Halpain, 2005;Dunaevsky, et al., 1999;Fischer, et al., 1998). In contrast, within 13 hours of incubation with 7PA2-CM, 34% of existing spines elongated into thin filopodial-like protrusions (Table 1). However, another 30% shrunk in size, or disappeared altogether (Table 1). We conclude, therefore, that soluble Aβ induces at least two types of morphological changes in spines, resulting in either spine elongation or spine shortening. Although protrusion elongation and shortening are seemingly opposite structural modifications, they both were accompanied by a reduction in spine head diameter (Fig. 5A₂,₃). Both effects could be observed on the same dendrite (Fig. 5A₂,₃; see Supplemental Fig. S3 for additional examples), and occasionally multiple changes occurred sequentially in individual spines, wherein outgrowth was followed later by collapse (Fig. S3B, open arrow). Overall the dendritic protrusions appear to become more dynamic in the 7PA2 treated neurons as compared to control (Supplemental Movies S2 and S3).

Of the 31 fluorescently labeled spine/nerve terminal pairs we identified in Aβ-treated cultures (111μm total length of dendrites analyzed from 3 postsynaptic neurons), 3 out of 31 spines (10%) were observed to spatially disconnect from their adjacent presynaptic varicosity prior to or concomitant with spine collapse (Fig. 6A, arrow; see also Fig. S5 for additional examples and Supplemental Movies S8, S9, and S10). In three additional pairs we observed either transient spine detachment or it appeared that a spine detached from only one of the two terminals to which it was connected. The remaining spines appeared to remain connected to their presynaptic varicosity during this time frame. Such spines either shrunk in head width only (15 of 31; 48%; Fig. 6A), elongated (6 of 31; 19%; Fig. 6B), or showed no significant change (4 out of 31; 12 %). In contrast, during our recordings of neurons exposed to control CHO-CM (n = 25 spine/nerve terminal pairs from 3 postsynaptic neurons), which were imaged at 10 min intervals for up to 48 hours, we never observed spatial separation of presynaptic varicosities from adjacent spines. These results suggest that a physical uncoupling of synapses may be one consequence of exposure to soluble Aβ.

Decreased synapse density, as measured by reduced numbers of puncta for the vesicle marker synaptophysin, is well documented in AD (Masliah, 2000;Selkoe, 2002). To assess whether soluble Aβ might have direct effects on the distribution of presynaptic markers, we incubated three-week old hippocampal cultures with conditioned medium from CHO cells stably expressing the APP V717F mutation (7PA2-CM) containing 80pM Aβ and quantified the size and density of puncta stained for either synaptophysin, to label all nerve terminals, Vesicular Glutamate Transporter (VGLUT), to specifically label excitatory glutamatergic terminals, or Vesicular GABA Transporter (VGAT), to specifically label inhibitory GABAergic terminals. We found that within one hour the size and number of synaptophysin clusters decreased significantly after incubation with 7PA2-CM, as compared to incubation with control medium from untransfected CHO cells (CHO-CM; Fig. 1). Interestingly, glutamatergic but not GABAergic synapses were affected: VGAT clusters were unaltered, but VGLUT clusters showed a nearly 40% decrease in number, and those that remained showed a 10% decrease in area compared to control cultures (Fig. 1). These data indicate that excitatory nerve terminals are selectively vulnerable to rapid changes induced by the Aβ conditioned medium (7PA2-CM). Since dendritic spines are the main postsynaptic sites for excitatory synapses in the mammalian brain, we next investigated whether we could detect a similar drop in their numbers following exposure to 7PA2-CM.