Prions activate a p38 MAPK synaptotoxic signaling pathway

Synaptic degeneration is one of the earliest pathological correlates of prion disease, and it is a major determinant of the progression of clinical symptoms. However, the cellular and molecular mechanisms underlying prion synaptotoxicity are poorly understood. Previously, we described an experimental system in which treatment of cultured hippocampal neurons with purified PrPSc, the infectious form of the prion protein, induces rapid retraction of dendritic spines, an effect that is entirely dependent on expression of endogenous PrPC by the target neurons. Here, we use this system to dissect pharmacologically the underlying cellular and molecular mechanisms. We show that PrPSc initiates a stepwise synaptotoxic signaling cascade that includes activation of NMDA receptors, calcium influx, stimulation of p38 MAPK and several downstream kinases, and collapse of the actin cytoskeleton within dendritic spines. Synaptic degeneration is restricted to excitatory synapses, spares presynaptic structures, and results in decrements in functional synaptic transmission. Pharmacological inhibition of any one of the steps in the signaling cascade, as well as expression of a dominant-negative form of p38 MAPK, block PrPSc-induced spine degeneration. Moreover, p38 MAPK inhibitors actually reverse the degenerative process after it has already begun. We also show that, while PrPC mediates the synaptotoxic effects of both PrPSc and the Alzheimer’s Aβ peptide in this system, the two species activate distinct signaling pathways. Taken together, our results provide powerful insights into the biology of prion neurotoxicity, they identify new, druggable therapeutic targets, and they allow comparison of prion synaptotoxic pathways with those involved in other neurodegenerative diseases.


Introduction
Prion diseases are a group of fatal, infectious neurodegenerative diseases affecting humans and animals. The infectious agent, or prion, is composed of PrP Sc , a conformationally altered form of a normal, cell-surface glycoprotein designated PrP C . Prions propagate themselves by a highly specific templating process in which PrP Sc molecules impose their unique, β-sheet-rich conformations on endogenous PrP C substrate molecules [1][2][3][4]. Consistent with this model, PrP knockout mice, in which PrP C expression is absent, are completely resistant to prion infection [5,6]. Moreover, these mice do not display symptoms of a prion disease [7], indicating that the disease phenotype is due primarily to a gain-of-function attributable to PrP Sc or a related toxic species, rather than to a loss of the normal function of PrP C . Therefore, it is important to understand the molecular mechanism of PrP Sc neurotoxicity.
Strikingly, the question of prion neurotoxicity has received relatively little attention in the field, in comparison to the extensive body of research that has been published on prion infectivity, propagation, and transmission. An important clue to the underlying mechanism is the observation that neurons that do not express endogenous PrP C are relatively resistant to the toxic effects of exogenously supplied PrP Sc [8,9]. This result suggests that a critical neurotoxic signal is generated as part of the process by which endogenous cell-surface PrP C is converted into PrP Sc , and in the absence of PrP C , this signal is not produced. Consistent with a role for PrP C as a neurotoxic mediator, there is evidence that prion disease progression in mice is characterized by two, mechanistically distinct phases: rapid accumulation of PrP Sc and infectivity, followed by slower development of neuropathology and clinical symptoms over a time course that is inversely related to expression levels of PrP C [10,11]. Although there are a number of studies suggesting signal-transducing activities for cell-surface PrP C [reviewed by 12], the pathways by which its interaction with PrP Sc produces neurotoxic signals remain mysterious.
Synaptic loss is a common theme in many neurodegenerative disorders [13,14]. In prion diseases, neuropathological and in vivo imaging studies in infected mice suggest that synaptic degeneration begins very early in the disease process, predating other pathological changes, and eventually contributing to the development of clinical symptoms [15][16][17][18][19][20][21][22]. However, there is very little mechanistic understanding of this process, due largely to the absence of suitable cell culture models amenable to experimental manipulation.
To address this gap, we previously established a novel neuronal culture model, using which we showed that PrP Sc induces rapid retraction of spines on the dendrites of hippocampal neurons [23]. Importantly, this effect is entirely dependent on expression of endogenous PrP C by the neurons, consistent with the previously demonstrated role of PrP C as an essential transducer of PrP Sc toxicity. Dendritic spines are the contact sites for most excitatory synapses in the brain, and they undergo constant morphological remodeling during development, learning, and memory formation [24,25]. Therefore, spines are an important locus for the pathogenesis of neurological diseases, particularly those involving symptoms of dementia.
Here, we have used cultured hippocampal neurons to dissect, using specific pharmacological inhibitors as well a dominant-negative kinase mutant, the mechanism of PrP Sc -induced synaptotoxicity. Our data establish a synaptotoxic signaling pathway involving, in stepwise sequence, activation of NMDA and AMPA receptors, calcium influx, stimulation of p38 mitogen-activated protein kinase (MAPK), and depolymerization of actin filaments in dendritic spines. Blocking any one of these steps prevented dendritic spine retraction in response to PrP Sc , and could, in some cases, even restore normal morphology to spines that had already degenerated. Taken together, our results provide powerful insights into the biology of prion neurotoxicity, they identify new, druggable therapeutic targets, and they allow comparison of the synaptotoxic pathways underlying prion diseases with those responsible for other neurodegenerative disorders like Alzheimer's disease.

PrP Sc causes functional changes in synaptic transmission
Previously, we showed that treatment of cultured hippocampal neurons for 24 hrs with purified PrP Sc , but not control preparations, induced a dramatic retraction of dendritic spines, an effect that was entirely dependent on expression of endogenous PrP C by the neurons [23].
Before embarking on pharmacological studies, we undertook several experiments to characterize the synaptotoxic effects of PrP Sc in neuronal culture, and to gain further insight into the underlying cellular mechanisms. First, we wondered whether the dramatic changes in dendritic spine morphology caused by PrP Sc were accompanied by alterations in the efficiency of synaptic transmission. To test this possibility, we used patch clamp recording to measure the amplitude and frequency of miniature excitatory postsynaptic currents (mEPSCs) in hippocampal neurons treated with PrP Sc . mEPSCs, which are recorded in the presence of TTX to block action potentials and picrotoxin to block GABA-evoked inhibitory currents, are a measure of spontaneous synaptic currents evoked by glutamate. We found that treatment of hippocampal neurons with purified PrP Sc , but not with mock-purified material from uninfected brains, caused a marked reduction in mEPSC frequency, and a less pronounced but statistically significant decrease in mEPSC amplitude (Fig 1A-1C). These effects were not observed in neurons derived from PrP knockout mice (Prn-p 0/0 ), demonstrating that the functional as well as the morphological effects of PrP Sc on synapses are entirely PrP C -dependent (Fig 1D-1F). Of note, PrP Sc did not significantly affect the frequency or amplitude of miniature inhibitory postsynaptic currents (mIPSCs) (Fig 1G-1I). This result indicates that PrP Sc is strikingly selective in its effects, targeting primarily excitatory and not inhibitory synapses.
For the electrophysiology experiments shown in Fig 1, we used a culture system that was slightly different from the one used to monitor dendritic spine retraction. In this system, neurons were grown at a higher density on the same substrate with supporting astrocytes, an arrangement which improves neuronal integrity during patch clamp recording. We confirmed that, when recordings were done on neurons cultured at low density over an astrocyte feeder layer, PrP Sc also caused a reduction in mEPSC amplitude and frequency (S1 Fig).

PrP Sc causes loss of excitatory postsynaptic markers without affecting presynaptic markers
The reduction in mEPSC amplitude and frequency caused by PrP Sc could be due to effects on either presynaptic processes (e.g. synaptic release) or postsynaptic characteristics (e.g., number and distribution of active zones). To distinguish presynaptic from postsynaptic effects, we immunostained neurons with antibodies to GluR1, an AMPA receptor subunit present in the postsynaptic density, and synaptophysin, a presynaptic marker. When neurons were treated with purified PrP Sc , but not mock-purified material, there was a dramatic loss of GluR1 staining, consistent with the retraction of dendritic spines revealed by fluorescent phalloidin staining (Fig 2A-2J and 2U). In contrast, there was no statistically significant loss of synaptophysin staining after PrP Sc treatment (Fig 2K-2T and 2U). These data demonstrate that PrP Sc exerts a were treated for 24 hrs with either mock-purified material or purified PrP Sc . Whole-cell patch clamping was used to record mEPSCs (A, D) and mIPSCs (G), the frequencies and amplitudes of which were quantitated (B,C, E, F, H, I). N = 15 cells from 3 independent experiments. ÃÃÃ p<0.001, ÃÃ p<0.01, by Student's t-test; N.S., not significantly different.
https://doi.org/10.1371/journal.ppat.1007283.g001 highly selective effect on postsynaptic elements, with no detectable effect on presynaptic structures, even in the face of massive morphological changes in dendritic spines.
We also found that PrP Sc treatment did not change the number of inhibitory synapses, as shown by staining for gephyrin (a postsynaptic anchoring component for glycine and GABA A receptors) (S2 Fig). This result is consistent with the inhibitory effect of PrP Sc on mEPSCs and not mIPSCs (Fig 1). We conclude that PrP Sc primarily targets the postsynaptic elements of excitatory synapses in the neuronal culture systems we are using.

PrP Sc synaptotoxicity requires glutamate receptor activation, and is accompanied by calcium influx
Glutamate receptor-dependent excitotoxicity contributes to the pathogenesis of many neurodegenerative diseases [26][27][28][29][30]. The selective effect of PrP Sc on excitatory synapses suggested the possibility that ionotropic glutamate receptors might play a role in the degeneration of dendritic spines seen in this system. To test this possibility, we treated hippocampal cultures with purified PrP Sc , or with mock-purified material form uninfected brains, in the presence or absence of NMDA or AMPA receptor blockers (see Table 1 for a list of pharmacological inhibitors used in this study). We observed that the competitive AMPA receptor antagonist CNQX, as well as the uncompetitive NMDA channel blocker memantine, prevented PrP Sc -induced retraction of dendritic spines (Fig 3). Spine number in PrP Sc -exposed cultures in the presence of these drugs was not statistically different from cultures exposed to mock material from uninfected brains.
Glutamate receptor-mediated excitotoxicity is often accompanied by an influx of Ca 2+ ions via NMDA receptors. To test whether this mechanism was operative during PrP Sc synaptotoxicity, we used the calcium-sensitive dye Fluo-3 to image intracellular calcium levels in neurons treated with PrP Sc . We found that purified PrP Sc , but not mock-purified material, caused a significant increase in intracellular calcium (Fig 4). The effect of PrP Sc on Ca 2+ levels was absent in neurons derived from Prn-p 0/0 mice lacking PrP C , and was completely blocked by memantine. Since voltage-gated calcium channels (VGCCs) are also major mediators of calcium influx into dendritic spines, we tested the effect of inhibitors of the major classes of VGCCs (R, T, N, P/Q, and L type). None of these inhibitors had a statistically significant effect on PrP Scinduced reduction in dendritic spine numbers, with the exception of lomerizine (an L-type VGCC inhibitor), which had a partial protective effect (S3 Fig).
We conclude from these data that activation of NMDA and AMPA receptors plays an essential role in dendritic spine retraction induced by PrP Sc , and that the effect of PrP Sc is accompanied by Ca 2+ influx, primarily via NMDA receptors. excitotoxicity [32]. In mammals, the MAPKs are grouped into three main families, ERKs (extracellular-signal-regulated kinases), JNKs (Jun amino-terminal kinases), and p38/SAPKs (stress-activated protein kinases) [31]. We tested whether any of these MAPK families are involved in PrP Sc synaptotoxicity. We observed that a p38 MAPK inhibitor (SB239063), which targets all four isoforms (α, β, γ, δ), effectively prevented spine retraction caused by PrP Sc , while pan-isoform inhibitors of ERK and JNK were without effect ( Fig 5). The inhibitors alone had no effect on spine number. In parallel with its ability to block the effects of PrP Sc on spine morphology, we found that the p38 MAPK inhibitor prevented the PrP Sc -induced reduction in mEPSC frequency and amplitude (Fig 6). Taken together, these data indicate that p38 MAPK plays an essential role in mediating the toxic effects of PrP Sc on synaptic structure and function in this system.
Mammalian p38 MAPK has four isoforms (α, β, γ, and δ) [33]. p38γ is most highly expressed in skeletal muscle, and p38δ in testis, pancreas, kidney and small intestine [34]. Thus, p38α and p38β are the isoforms most likely to be involved in neuronal signaling. The α and β isoforms are 75% identical, and both are inhibited by the compound SB239063 used in Figs 5 and 6. To determine which of the two p38 isoforms is involved in PrP Sc synaptotoxicity, we utilized the p38α -specific inhibitor VX745. We found that VX745 completely blocked the effects PrP Sc on dendritic spine number and mEPSC properties, suggesting that p38α is one of the isoforms involved (S4 Fig). The absence of a p38β-specific inhibitor precluded specific testing of this isoform.

PrP Sc -induced retraction of dendritic spines is reversible by application of a p38 MAPK inhibitor
Because abnormalities in dendritic spines and synaptic transmission are early effects of PrP Sc , which occur well before large-scale changes in neuronal morphology or loss of neuronal viability, we wondered whether the effects of PrP Sc might be reversible by treatment with a p38 MAPK inhibitor. To test this possibility, we treated neurons with PrP Sc for 24 hrs, at which point most of the dendritic spines were retracted ( Fig 7A). We then exposed the neurons to PrP Sc for additional 24 hrs in the presence of a p38 MAPK inhibitor (SB239063) or vehicle control, after which cultures were fixed and assessed for dendritic spine morphology with fluorescent phalloidin. Amazingly, we found that the p38 MAPK inhibitor was able to reverse the dendritic spine retraction that had accrued during the first 24 hrs of PrP Sc treatment (Fig 7C), compared to the cultures treated with vehicle ( Fig 7B). Quantitation of spine number under the three conditions is shown in Fig 7E. These data indicate that the extensive dendritic spine abnormalities induced by PrP Sc are reversible by p38 MAPK inhibition, at least within a 48 hr time window.

A dominant-negative mutant of p38 MAPK prevents PrP Sc synaptotoxicity
To confirm the results obtained with pharmacological inhibition of p38 MAPK, we employed a genetic method to suppress signaling through this pathway, which makes use of a dominantnegative form of p38α MAPK (T180A/Y182F, referred to as p38AF). This double-mutation in the activation loop of the kinase prevents phosphorylation by upstream kinases, and has a dominant-negative effect on the activity of co-expressed wild-type p38, thereby significantly attenuating signaling [35]. We prepared hippocampal neurons from mice that were heterozygous for the p38AF allele. This method of reducing p38 signaling avoids the embryonic lethal phenotype that results from complete germline inactivation of the p38 MAPK gene [36]. We found that neurons prepared from p38AF mice were morphologically comparable to WT  Prions activate a p38 MAPK synaptotoxic signaling pathway neurons, but were almost completely resistant to the dendritic spine retraction effect of PrP Sc (Fig 8). This experiment, which depends on constitutive down-regulation of p38 signaling, complements the previous experiments, which involved acute, pharmacological inhibition of p38 MAPK activity.

PrP Sc causes increased phosphorylation of p38 MAPK in dendritic spines
To provide biochemical evidence for activation of the p38 MAPK pathway in response to PrP Sc , we utilized an immunocytochemical approach in order to detect localized changes in p38 phosphorylation. We doubly stained PrP Sc -treated cultures with antibodies to phospho-p38 and total p38, and then imaged the ratio of the two signals in the region of dendritic spines, as marked by phalloidin staining. We found that the amount of phosphorylated p38 in dendritic spines was increased after 1 hr of PrP Sc treatment, and remained elevated after 24 hrs in the region of collapsed spines (Fig 9).

MAPK-activated protein kinases (MKs) are downstream targets of p38 MAPK in the PrP Sc synaptotoxic pathway
Known downstream targets of p38 MAPK include MAPK-activated protein kinase (MAPKAP or MK), MSK and MNK [reviewed in 37]. To determine which of these kinases is downstream of p38 MAPK in the PrP Sc synaptotoxic pathway, neurons were treated with PrP Sc in the presence or absence of specific inhibitors. We found that a pan MK inhibitor (CAS1186648), which targets all three isoforms (2, 3, and 5), effectively prevented PrP Sc -induced retraction of dendritic spines, while pan-isoform inhibitors of MSK and MNK had no significant effect ( Fig  10). None of the three inhibitors alone had a significant effect on spine number. We conclude from these results that MK2, 3, or 5 are potential targets of p38 MAPK in the pathway leading to dendrite retraction caused by PrP Sc . Of these, MK2 and MK3 are the most likely isoforms to be involved, since MK5 is expressed primarily in the heart, while MK2 and MK3 are ubiquitously expressed [38,39]. MK2/3 are both substrates for the α isoform of p38 MAPK.

p38 MAPK and MK inhibitors do not affect PrP Sc propagation in ScN2a cells
One possible mechanism by which inhibitors of MAPK pathways could prevent PrP Sc -induced synaptotoxicity would be by inhibiting conversion of endogenous neuronal PrP C to PrP Sc , on the assumption that newly formed PrP Sc is the primary trigger for synaptic degeneration. To address this possibility, we tested whether these inhibitors affected PrP Sc formation in scrapieinfected N2a (ScN2a) cells. We found that neither the p38 MAPK inhibitor SB239063, nor the MK2/3/5 inhibitor CAS1186648, had a significant effect on the levels of protease-resistant PrP Sc in ScN2a cells after 7 days of treatment (S5 Fig). As a positive control, the known antiprion agent Congo red [40,41], dramatically reduced PrP Sc levels under the same conditions. None of the compounds had a significant cytotoxic effect, as indicated by measurement of the total amount of cellular protein. These results support the notion that the p38 MAPK and MK inhibitors block PrP Sc synaptotoxicity by interfering with signaling pathways linked to synaptic integrity, rather than by reducing the amount of the toxic agent (PrP Sc ).

Actin stabilization counteracts the synaptotoxic effects of PrP Sc
Actin is abundant in dendritic spines and has been shown to regulate spine morphology [42].
In addition, there is evidence that most signaling pathways linking synaptic activity to spine morphology influence local actin dynamics [43,44]. To address the role of actin in PrP Scinduced alterations in dendritic spine morphology, we used SiR-actin, a fluorogenic, cell-permeable peptide derived from jasplakinolide that both stabilizes and labels F-actin [45]. Neurons were treated with SiR-actin alone or in combination with PrP Sc , after which spines were visualized using SiR-actin fluorescence. For comparison, we exposed cultures to taxol, a microtubule-stabilizing agent, with or without PrP Sc , after which spines were visualized using fluorescent phalloidin staining. We found that SiR-actin, but not taxol, prevented PrP Sc -induced spine retraction (Fig 11A-11E).
In a second experiment, we treated cultures with SiR-actin in the presence of PrP Sc or mock-purified material, and then stained them with an antibody to the AMPA receptor subunit GluR1. We observed that SiR-actin prevented loss of GluR1 staining in response to PrP Sc (Fig 11F-11H), parallel to the effect of this compound on spine retraction monitored using SiR-actin fluorescence (Fig 11A, 11B and 11E).
Consistent with these microscopic imaging results, electrophysiological recordings showed that SiR-actin prevented the decreases in mEPSC amplitude and frequency caused by PrP Sc treatment (Fig 11I-11K).
Taken together, these data demonstrate that actin dynamics play an important role in the morphological changes in dendritic spines induced by PrP Sc , and that stabilizing actin filaments can prevent these changes.

The unfolded protein response (UPR) does not play a prominent role in PrP Sc synaptotoxicity
Activation of the UPR has been suggested to play a role in the pathogenesis of prion diseases, and blocking this pathway has been shown to produce a therapeutic benefit [46][47][48]. To address whether the UPR plays a role in PrP Sc -induced synaptotoxicity in our system, we treated cultured neurons with an inhibitor of PERK kinase (GSK2606414) or an activator of eIF2B (ISRIB [49,50]), both of which reverse the eIF2α-mediated translational repression that occurs during the UPR. Neither compound had a significant effect on spine retraction induced by PrP Sc (S6 Fig), suggesting that the UPR, in particular the translational repression arm mediated by PERK and eIF2α, does not play a role in PrP Sc synaptotoxicity in this system.

PrP Sc and Aβ oligomers activate different molecular pathways
It has been proposed that PrP C is a cell-surface receptor for Aβ oligomers, which mediates some of the neurotoxic effects of these assemblies, including loss of dendritic spines [51][52][53][54][55]. There is evidence that binding of Aβ oligomers to PrP C triggers a signaling pathway involving mGluR5 and Fyn kinase, and that preventing activation of these molecules using specific inhibitors prevents Aβ neurotoxicity and ameliorates neurological symptoms in mice [56][57][58][59]. Since the synaptotoxic effects of both PrP Sc and Aβ are dependent on the expression of PrP C as a cell surface receptor in target neurons, we asked whether both toxic aggregates activated the same cellular pathways downstream of PrP C .
To investigate this question, we treated neurons with either Aβ or PrP Sc in the presence or absence of the mGluR5 inhibitor, MPEP, or the p38 MAPK inhibitor, SB239063. Consistent with previously published studies [57], Aβ oligomers alone caused a significant reduction in dendritic spine number (Fig 12A and 12B), an effect that was dependent on expression of PrP C (S7 Fig). MPEP completely blocked this effect (Fig 12C), also in agreement with published results [56]. In contrast, MPEP had no influence on PrP Sc -induced retraction of dendritic spines (Fig 12E and 12F). Moreover, the p38 MAPK inhibitor, which completely blocked PrP Sc synaptotoxicity (Fig 5), had no significant effect on Aβ oligomer-induced dendritic spine loss (Fig 12D). Taken together, these data suggest that Aβ oligomers and PrP Sc trigger different neurotoxic signaling pathways downstream of a common cell-surface receptor, PrP C .

Discussion
Although the molecular templating process underlying the infectivity of prions is now well understood, the mechanisms by which prions cause neurodegeneration, in particular, damage to synapses, remain poorly understood. In a previously published study [23], we established a neuronal culture system that recapitulates one of the earliest events in prion synaptotoxicity, PrP Sc -induced retraction of dendritic spines. In the present work, we have exploited the simplicity of this system to dissect the cellular pathways underlying the toxic effects of PrP Sc on synapses. Our results uncover a multi-step signaling cascade that begins with binding of PrP Sc to PrP C on the cell surface, and is followed by activation of NMDA and AMPA receptors, calcium influx, stimulation of the stress-inducible MAPK, p38, and finally collapse of the actin cytoskeleton, retraction of dendritic spines, and a decrease in excitatory neurotransmission (Fig 13). This work provides new insights into the mechanisms of synaptic degeneration in prion diseases, it identifies novel molecular targets for treatment of these disorders, and it allows comparison with pathologic mechanisms operative in other neurodegenerative disorders such as Alzheimer's disease.
While our previous study focused on retraction of dendritic spines as the main morphological consequence of PrP Sc toxicity [23], the results presented here provide a more complete picture of the underling cellular events. We have shown that PrP Sc exhibits striking functional and morphological specificity in its synaptotoxic effects: it damages excitatory and not inhibitory synapses, and it targets postsynaptic and not presynaptic sites. This selectivity may reflect the presence on dendritic spines of glutamate receptors capable of initiating excitotoxic processes (see below), or the presence of a higher concentration of specific intracellular signaling molecules in these locations. We have also shown that spine loss is accompanied by a reduction in postsynaptic AMPA receptors, and by a decrement in mEPSC frequency and amplitude, suggesting that functional changes in synaptic transmission are early indicators of prion neurotoxicity. Although many studies have emphasized neuronal death as an important feature of prion and other neurodegenerative diseases, our results emphasize the importance of looking specifically at synaptic dysfunction and loss in order to understand the earliest events in the pathological process.
Our evidence suggests that synaptotoxic signaling induced by PrP Sc is initiated by its interaction with PrP C on the cell surface (Fig 13). All of the synaptotoxic effects of PrP Sc observed in our system require expression of endogenous PrP C by the target neurons [this work and 23]. Moreover, neurons from transgenic mice expressing N-terminally deleted forms of PrP C (Δ23-31 and Δ23-111) are resistant to PrP Sc toxicity, pinpointing an essential role for these  Fig 2A-2J and 2U). A third set of cultures was treated for 24 hrs with mock-purified material or purified PrP Sc in the absence or presence of SiR-actin, and was then analyzed by patch clamping to measure mEPSC frequency and amplitude (I-K). N = 20 cells from 4 independent experiments. In E, H J, K: ÃÃÃ p<0.001 and ÃÃ p<0.01 by Student's t-test; N.S., not significantly different. Scale bar in panel G = 20 μm (also applicable to panels A-D and F).
https://doi.org/10.1371/journal.ppat.1007283.g011 residues in the conversion and/or signal-transduction processes [23]. The requirement for cell-surface PrP C in our system is consistent with the documented role of PrP C in mediating the neurotoxic effects of PrP Sc in vivo [8,9]. We have hypothesized that neurotoxic signals may be generated either by the initial binding of PrP Sc to PrP C , or by the subsequent conversion of PrP C into nascent PrP Sc on the cell surface [23]. Since both PrP C and PrP Sc are attached to the plasma membrane by a GPI anchor, production of intracellular signals would presumably require their interaction with partners that are transmembrane proteins, for example ion channels [see below and 56, 60], adhesion molecules [61], or receptors [62].
We have shown that activation of NMDA and AMPA receptors is required for PrP Scinduced synaptotoxicity in our system, and that specific blockers of these channels prevent dendritic spine retraction. Moreover, PrP Sc induces a rapid (within 30 min) increase in intracellular calcium, mediated by NMDA receptors. These results suggest that glutamate receptor activation accompanied by calcium entry is an early step in the signaling cascade initiated by PrP Sc , and that glutamate-induced excitotoxicity may play a role in the ensuing synaptic damage. PrP C has been suggested to be a modulator of NMDA receptor function [63] and posttranslational modification [64], and to interact physically with NMDA receptor subunits [65], processes that could be altered by binding of PrP Sc to cell-surface PrP C during the initial phase of the conversion process. Also, activation of NMDA-dependent transcriptional pathways is an early event during the pre-clinical phase of prion infection in mice, as determined by microarray profiling [66]. More generally, NMDA receptors play an important role in learning and memory [25], and they are known to be involved in many neurodegenerative diseases in the context of excitotoxic activation [14]. Of note, glutamate stimulation of cultured hippocampal neurons causes dendritic abnormalities (spine retraction) and calcium influx that are similar to those seen in our system with PrP Sc treatment [67].
Our results implicate p38 MAPK, specifically the α isoform, in PrP Sc -induced synaptotoxicity, based on blockade of dendritic spine retraction and mEPSC reduction by specific inhibitors of this kinase, and by expression of a dominant-negative form of the kinase. Moreover, using immunofluorescence staining, we demonstrated that PrP Sc induces rapid (within 1 hr) phosphorylation of p38 MAPK in dendritic spines, a region where this kinase was previously shown to reside [68]. Thus, PrP Sc activates a localized, p38-mediated signal transduction cascade that precedes dendritic spine retraction. p38 MAPK is, like its counterpart JNK, a stressactivated protein kinase, and it was originally found to be stimulated by a variety of environmental stresses and cytokines that induce inflammation [31,37]. It has now been implicated in a wide range of functions, including regulation of the cell cycle, induction of cell death, differentiation, and senescence. In the nervous system, p38 MAPK has been found to play a role in neuronal damage and survival, as well as in synaptic plasticity, and it has been linked to a number of neurodegenerative diseases [32,69]. Consistent with the results presented here, p38 MAPK, and its downstream substrates MK2/3, have been shown to regulate AMPA receptor trafficking, dendritic spine morphology, and synaptic transmission [70]. We have identified MK2/3 as likely substrates for p38 MAPK in the PrP Sc synaptotoxic pathway, based on the use of specific inhibitors, but the identity of the upstream kinases (MAPKKs and MAPKKKs) responsible for p38 MAPK activation remain to be determined. Activation of MAPK pathways typically leads to programmed changes in gene expression [31], and it will be of great interest to characterize these in order to understand PrP Sc synaptotoxicity at a genomic level.
Our results demonstrate that re-organization of the actin cytoskeleton within dendritic spines represents a key cellular correlate of PrP Sc synaptotoxicity, based on the ability of actin stabilizers (SiR-actin) to block the morphological and electrophysiological consequences of PrP Sc treatment. Actin plays an important role in the structure and function of dendritic spines [42]. The equilibrium between G-and F-actin in spines is regulated in an activity-dependent manner, and actin dynamics influences spine development and synaptic plasticity [43,44]. Postsynaptic receptors, including NMDA and AMPA receptors [71], regulate the actin cytoskeleton via their effect on cytoplasmic actin-binding proteins, either by directly interacting with these proteins or, in the case of NMDA receptors, by increasing intracellular calcium [72,73]. This mechanism provides a plausible link between PrP Sc activation of NMDA receptors, with rapid influx of calcium, and the resulting abnormalities in dendritic spine morphology and function. There is evidence that p38 MAPK can also influence actin cytoskeletal dynamics [74], providing another potential pathway by which PrP Sc could alter spine morphology via activation of this kinase.
Several previous studies have implicated specific signal transduction pathways in the pathogenesis of prion diseases, including the UPR [46], oxidative stress [75], MAPKs [76][77][78], phosphoinositide-dependent kinase-1 (PDK1) [79], metabotropic glutamate receptors [80], NMDA receptors [81], and voltage-gated calcium channels [82]. Given published evidence that the PERK/eIF2α arm of the UPR is activated during prion infection, and that pharmacological blockage of this response has therapeutic benefit [46][47][48], we tested inhibitors of this pathway in our system. However, we found that these inhibitors had no effect on PrP Scinduced loss of dendritic spines. This discrepancy is likely to reflect fundamental differences between the experimental systems employed. All of the previously quoted studies utilized brain slices or tissue samples from human patients or transgenic mice undergoing neurodegenerative changes after prion infection, making it difficult to isolate the primary synaptotoxic mechanisms engaged by PrP Sc . In contrast, we have employed a neuronal culture system that allows us to characterize acute responses to exogenous PrP Sc exposure. We have shown that, in this system, treatment with PrP Sc causes a detectable increase in intracellular calcium within 30 min, followed by alterations in electrophysiological properties, with complete collapse of dendritic spines by 24 hrs. Moreover, our experiments have demonstrated the involvement of MAPK signaling cascades, which are typically activated within 60 min of exposure to an extracellular stimulus. Our results do not rule out the engagement of additional pathways, which contribute to pathological changes occurring at later times, once PrP Sc begins to accumulate.
Our results identify novel molecular targets for therapeutic intervention. Previous efforts to develop anti-prion therapies have focused on strategies for inhibiting formation of PrP Sc or enhancing its clearance [83]. In contrast, our studies suggest the possibility of interfering with neurotoxic pathways that lie downstream of PrP Sc . We have shown that several kinds of pharmacologic agents prevent PrP Sc -induced synaptotoxic effects in our system, including NMDA and AMPA receptor antagonists, and p38 MAPK inhibitors. Importantly, we have found that treatment with a p38 MAPK inhibitor is able to reverse dendritic spine retraction that has already occurred during an initial exposure to PrP Sc . This result, which presumably reflects the dynamic nature of dendritic spines, suggests the existence of a therapeutic window for treatment of patients who have already been infected with prions and who might even have sustained a certain level of synaptic damage. The advent of methods for pre-mortem diagnosis of prion diseases [84,85] makes this type of treatment modality especially compelling. The NMDA receptor antagonist memantine is now a widely used treatment for Alzheimer's disease [86], and p38 MAP kinase inhibitors have been developed for therapy of inflammatory diseases [87] and CNS disorders, including Alzheimer's disease [88][89][90]. It may be feasible to re-purpose these agents for treatment of prion diseases.
It has been proposed that PrP C acts as a cell-surface receptor for Aβ oligomers in Alzheimer's disease, and that it mediates the neurotoxic effects of these oligomers [51][52][53][54][55][56][57][58][59]. We have been able to use our neuronal culture system to test whether PrP Sc and Aβ oligomers activate a common synaptotoxic pathway upon binding to PrP C . Our results indicate that the two pathways diverge, based on our observation that the synaptotoxic effects of PrP Sc are blocked by p38 MAPK inhibitors but not mGluR5 inhibitors, while the reverse is true for the synaptotoxic effects of Aβ oligomers. The precise differences between the PrP Sc and Aβ synaptotoxic pathways remain to be determined. Our system could be used to probe the synaptotoxic mechanisms activated by other oligomeric aggregates, including tau and α-synuclein, which are thought to participate in an extracellular transmission phase [91], and which may also utilize PrP C as a cell-surface receptor [92].
Chronic neurodegenerative disorders, like prion, Alzheimer's and Parkinson's diseases, are likely to involve multiple pathogenic mechanisms, each of which may be operative at different stages of the disease process. The experimental system described here has allowed us to isolate one very early event (synaptic degeneration) in the pathological cascade, and study its cellular and molecular underpinnings. Future experiments will be aimed at translating these findings into animal models and testing the efficacy of therapeutic interventions directed at specific steps in the signaling pathway we have identified.

Low-density neuronal cultures (used for dendritic spine measurements)
Timed-pregnant C57BL/6 mice (referred to as wild-type, WT) were purchased from the Jackson Laboratory (Bar Harbor, ME). Prnp 0/0 mice [7] on a C57BL6 background were obtained from the European Mouse Mutant Archive (EMMA; Rome, Italy), and were maintained in a homozygous state by interbreeding.
Mice carrying the p38AF dominant-negative mutation [35] on a C57BL6 background were obtained from the Jackson Laboratory (B6.Cg-Mapk14 tm1.1Dvb /J; stock #012736). The mutant allele was maintained in a heterozygous state by breeding with C57BL6 inbred mice. PCR genotyping of tail DNA was performed as per information and protocols are provided by Jackson Laboratory using the following primers: 5'-TAG AGC CAG CCC CAC TTT AGT C-3' and 5'-GAA GAT GGA TTT TAA GCA TCC GT-3'. The expected PCR products included a 328 bp band representing the dominant-negative allele, and a 195 bp band representing the WT allele.
All procedures involving animals were conducted according to the United States Department of Agriculture Animal Welfare Act and the National Institutes of Health Policy on Humane Care and Use of Laboratory Animals.
Hippocampal neurons were cultured from P0 pups as described [93]. All experiments shown, except those indicated in Figs 1 and 4, were performed with neurons from WT mice. Neurons were seeded at 75 cells/mm 2 on poly-L-lysine-treated coverslips, and after several hrs the coverslips were inverted onto an astrocyte feeder layer and maintained in NB/B27 medium until used. The astrocyte feeder layer was generated using murine neural stem cells, as described [94]. Neurons were kept in culture for 18-21 days prior to PrP Sc or ADDL treatment.

Dendritic spine quantitation
Hippocampal neurons cultured as described above were treated with purified PrP Sc , ADDLs, or control preparations for 24 hrs, followed by fixation in 4% paraformaldehyde and staining with either Alexa 488-phalloidin or rhodamine-phalloidin (ThermoFischer Scientific, Waltham, MA) to visualize dendritic spines, and anti-tubulin antibodies (Sigma-Aldrich, St. Louis, MO) to visualize axons and dendrites. Images were acquired using a Zeiss 880 (Figs 8 and 9) Zeiss 700 (all other figures) confocal microscope with a 63x objective (N.A. = 1.4). The number of dendritic spines was determined using ImageJ software. Briefly, 2-3 isolated dendritic segments were chosen from each image, and the images adjusted using a threshold that had been optimized to include the outline of the spines but not non-specific signals [95]. The number of spines was normalized to the measured length of the dendritic segment to give the number of spines/μm. For each experiment, 15-24 neurons from 3-4 individual experiments were imaged and quantitated.

Calcium imaging
Hippocampal neurons cultured as described above were washed once with PBS before treatment with Fluo-3 (ThermoFisher Scientific, Waltham, MA) at final concentration of 5 μM. Cells were incubated for 20-30 min at 37˚C, followed by a several washes to remove extracellular dye. Neurons were imaged in the green fluorescence channel of an Olympus wide field microscope with 20X objective. Each neuron was imaged for 5 min before either drug or solvent control was added, after which imaging was continued for an additional 30 min. The average signal during the 5 min pre-treatment was considered to represent the baseline calcium level for each cell. The proportion of 30 min recording period during which the calcium signal was ! 2X the baseline level was calculated and used as a measure of net calcium accumulation.

Measurement of p38 MAPK phosphorylation in dendritic spines
Low-density hippocampal neuronal cultures were treated with purified PrP Sc or mock preparations for 1 or 24 hrs, followed by fixation in 4% paraformaldehyde containing phosphatase inhibitors (Roche-Sigma, 04 906 845 001) and permeabilization in 1% Triton X-100 for 5 min. Dual immunostaining was performed with anti-phospho-p38 antibody (Thr180/Tyr182) (Cell Signaling, mAb #4511S, 1:100) and anti-total p38 antibody (Abcam, ab31828, 1:300), followed by Alexa 546 and Alexa 633 secondary antibodies, respectively. Cultures were also stained with Alexa 488-phalloidin to visualize dendritic spines. Multi-stack images were acquired using a Zeiss 880 confocal microscope with a 63x objective (N.A. = 1.4). The fluorescence intensities of the phospho-p38 and total p38 signals were quantitated within a region of interest (ROI) using ImageJ software. Images of phalloidin staining were used to determine the location of intact or collapsed dendritic spines. Two or three isolated ROIs on each neuron, coinciding with locations of individual dendritic spines, were captured at 10X zoom, and then the fluorescence intensity and area were determined within the ROI. The fluorescence intensities were normalized to the area of the ROI, and the ratio of phospho-p38/total p38 staining was calculated. For each experiment, 10-12 neurons from 3-4 individual experiments were analyzed.

Electrophysiological analysis using mixed hippocampal/glial cultures
With the exception of the experiment shown in S1 Fig, hippocampal cultures used for electrophysiological recording were prepared using a procedure that differs from the one used to prepare cultures for dendritic spine imaging. Briefly, hippocampi from newborn pups of the indicated genotypes were dissected and treated with 0.25% trypsin at 37˚C for 12 min [96]. Cells were plated at a density of 65,000 cells/cm 2 on poly-D-lysine-coated coverslips in DMEM medium with 10% F12 and 10% FBS.

Purification of PrP Sc
Purification was carried out as previously described [23,97]. In a typical preparation (used for all experiments, except those shown Figs 8 and 9), 18 RML-infected C57BL6 brains were homogenized in 3 ml of 10% sarkosyl in TEND (10 mM Tris-HCl [pH 8], 1 mM EDTA, 130 mM NaCl, and 1 mM dithiothreitol) containing Complete Protease Inhibitor Cocktail (Roche Diagnostics, cat. no. 11836153001) using a glass bead homogenizer. Brain homogenates were incubated on ice for 1 hr and centrifuged at 22,000 x g for 30 min at 4˚C. The supernatant was kept on ice, while the pellet was resuspended in 1 ml of 10% sarkosyl in TEND, incubated for 1 hr on ice, and then centrifuged at 22,000 x g for 30 min at 4˚C. The pellet was discarded while the supernatants were pooled and centrifuged at 150,000 x g for 2.5 h at 4˚C. The new supernatants were discarded, while the pellets were rinsed with 50 ml of 100 mM NaCl, 1% sulfobetaine (SB) 3-14 in TEND plus protease inhibitors, and then pooled by resuspending them in 1 ml of the wash buffer, and centrifuging at 180,000 x g for 2 hr at 20˚C. The supernatant was discarded, and the pellet was rinsed with 50 ml of TMS (10 mM Tris-HCl [pH 7.0], 5 mM MgCl 2 , and 100 mM NaCl) plus protease inhibitors, resuspended in 600 μl of the same buffer containing 100 mg/ml RNase A and incubated for 2 hr at 37˚C. The sample was then incubated with 5 mM CaCl 2 , 20 mg/ml DNase I for 2 hr at 37˚C. To stop the enzymatic digestion, EDTA was added to a final concentration of 20 mM, and the sample was mixed with an equal volume of TMS containing 1% SB 3-14. The sample was gently deposited on a 100 μl cushion of 1 M sucrose, 100 mM NaCl, 0.5% SB 3-14, and 10 mM Tris-HCl (pH 7.4), and centrifuged at 180,000 x g for 2 hr at 4˚C. The supernatant was discarded and the pellet was rinsed with 50 μl of 0.5% SB 3-14 in PBS, resuspended in 1 ml of the same buffer, subjected to 5 X 5 sec pulses of bath sonication with a Bandelin Sonopuls Ultrasonicator (Amtrex Technologies, Montreal, Canada) at 90% power, and centrifuged at 180,000 x g for 15 min at 4˚C. The final supernatant was discarded and the final pellet was resuspended in 900 μl of PBS (50 μl for each starting brain) and sonicated 5 times for 5 sec. Aliquots were stored at -80˚C. Mock purifications were also carried out from age-match, uninfected brains. The purified preparations were evaluated by SDS-PAGE followed by silver staining and Western blotting.
For the experiments shown in Figs 8 and 9, PrP Sc was purified using the pronase E method [23].
In all experiments, purified PrP Sc was added to neuronal cultures at a final concentration of 4.4 μg/ml, identical to what was used in our previous study [23]. An equivalent amount of mock material was used, based on purification from the same proportion of brain tissue.

ScN2a cell assay
The uninfected Neuro-2a cells (N2a) were from the ATCC (Cat. #: ATCC CCL-131). A scrapie-susceptible sub-clone (N2a.3) infected with RML prions [98] were plated in 6-well plates. Cells were treated for 3 days with test compounds (p38 MAPK and MK2/3/5 inhibitors), or with DMSO vehicle as a negative control and Congo red as a positive control. Cultures were then split at a 1:5 ratio and treatment with compounds in fresh medium continued for 4 more days. At the end of the 7-day treatment, cells were lysed in 300 μl of lysis buffer (0.5% NP-40, 0.5% deoxycholate, 10mM Tris-HCl, pH8 and 100mM NaCl) and protein concentration was measured using BCA assay (ThermoFisher Scientific, Waltham, MA). Samples were then treated with proteinase K (40 μg/ml) at 37˚C for 1 hr. Digestion was stopped by addition of 10x Complete Protease Inhibitor Cocktail (Roche, Indianapolis, IN). Samples were centrifuged at 180,000 x g for 1 hr at 4˚C. Pellets were dissolved in 20 μl of 1x Laemmli loading buffer (Bio-Rad, Hercules, CA) and were boiled for 3 min before loading on pre-cast 12% SDS-PAGE Criterion gels (Bio-Rad, Hercules, CA). Western blotting was performed according to standard procedures. PrP Sc was detected using the anti-prion antibody D18 [99] and HRP-coupled, anti-human secondary antibody (Jackson ImmnoResearch, West Grove, PA). Quantitation was performed using the ImageJ gel quantification function.

ADDL preparation
ADDLs were prepared from synthetic Aβ 1-42 peptide as previously described [55,100]. Aβ peptide (ERI Amyloid Laboratory, Oxford, CT) was dissolved in HFIP at a concentration of 1 mM and sonicated for 10 min in an ice-water bath. The sample was allowed to incubate at room temperature for 1 hr before it was transferred to a centrifuge tube and spun at 15,800 x g for 1 min. The supernatant was transferred into a new glass vial and was dried under a fume hood with nitrogen gas. The film of dried peptide was dissolved in DMSO, and then diluted into Ham's F12 phenol-red-free medium (Gibco/ThermoFisher scientific, Waltham, MA) to a final concentration of 100 μM. The sample was incubated at room temperature for 16 hrs and was then centrifuged for 15 min at 15,800 x g. The supernatant was aliquoted, flash-frozen in liquid nitrogen, and stored at -80˚C.

Ethics statement
All procedures involving animals were conducted according to the United States Department of Agriculture Animal Welfare Act and the National Institutes of Health Policy on Humane Care and Use of Laboratory Animals. Ethical approval (AN-14997) was obtained from Boston University medical center institutional animal care and use committee. ScN2a cells were treated for 3 days with DMSO vehicle, Congo red (5 μm), p38 MAPK inhibitor (SB239063, 10 μM), or MK2/3/5 inhibitor (CAS1186648, 500 nM), after which cells were split at a 1:5 ratio and fresh inhibitors were added for 4 more days. At the end of the 7-day treatment, cells were harvested and lysed. BCA protein assays of lysates were performed as a measure of drug cytotoxicity (A). Cell lysates were also subjected to proteinase K digestion followed by Western blotting to reveal proteinase K-resistant PrP Sc (B). ÃÃÃ p<0.001 by Student's t-test; N.S., not significantly different. Data were derived from triplicate cultures.  6 μM) (B, D). Neurons were then fixed and stained with fluorescent phalloidin (green) to visualize F-actin in dendritic spines, and with anti-tubulin (red) to visualize overall dendritic morphology. Pooled measurements of dendritic spine number were collected from 15-20 cells from 3 indepdendent experiments (E). ÃÃÃ p<0.001 by Student's t-test; N.S., not significantly different. Scale bar in panel D = 20 μm (also applicable to panels A-C). (TIF)