Time course of PrP^Sc-induced synaptotoxicity

Previously, we showed that treatment of cultured hippocampal neurons for 24 hours with purified PrP^Sc, but not control preparations, induced retraction of dendritic spines and impairment of synaptic transmission, effects that were entirely PrP^C-dependent [14,15]. To determine more precisely the time course of these synaptotoxic effects, thereby providing a guide to the time points at which to collect samples for phosphoproteomic and transcriptomic analysis, we treated hippocampal cultures for 0.5, 2, 4, and 24 hours with PrP^Sc purified from the brains of mice infected with the RML strain of murine-adapted scrapie. Mock-purified material from uninfected brains was used as a negative control. We found that dendritic spine retraction was first detectable after 4 hours of PrP^Sc treatment and progressed further by 24 hours (S1A and S1B Fig). Since PrP^Sc activation of phosphorylation-dependent signaling cascades and subsequent changes in gene expression were likely to lie upstream of, and occur earlier than, observable changes in spine morphology, we decided to include time points earlier than 4 hours in our subsequent experiments. This decision was also motivated by our previous observation that NMDAR-activated Ca²⁺ entry can be detected within 0.5-1 hour, and increased phosphorylation of p38 MAPK by 1 hour following PrP^Sc application [15]. Based on these factors, we elected to collect phosphoproteomic data at 1 hour and RNA-seq data at 0.5, 2, 4, and 24 hours after PrP^Sc application to hippocampal neuronal cultures (S1C Fig).

PrP^Sc causes rapid phosphoproteomic alterations associated with synaptic pathways

A key feature of our culture system is that hippocampal neurons are physically separated from the astrocyte feeder layer, allowing collection of proteins and RNA from highly neuron-enriched samples, as astrocytes remain confined to the feeder layer and are rarely observed in the neuronal layer. We undertook an experiment in which five replicate cultures were treated for 1 hour with purified PrP^Sc, and five cultures for 1 hour with mock-purified material. For proteomic analysis, neuronal protein lysates were processed by a standard workflow that included trypsin digestion, labeling of the resulting peptides with isobaric, tandem mass tags (TMT), isolation of phosphopeptides on Fe-NTA beads, and tandem mass spectrometry (MS/MS) analysis. A portion of each sample prior to phosphopeptide isolation was saved for total proteome analysis. TMT labeling enabled multiplexed, relative quantification of peptide abundances across samples (see PCA plots in S2A Fig).

In this experiment, we identified 8,217 proteins and 2,135 phosphoproteins with 1,455 overlaps (S2B Fig). Proteins with │log2fold-change│ > 0.25 and p-value <0.05 values were considered significant, which led to the identification of 1,212 differentially expressed proteins (62 upregulated and 1,150 downregulated), and 387 differentially phosphorylated proteins (71 increased and 316 decreased) (S2C and S2D Fig and S1 and S2 Tables). To identify pathways associated with the differentially phosphorylated proteins, we performed Gene Set Enrichment Analyses (GSEA) [20] using terms from Gene Ontology (GO) gene sets. Among the most prominently altered pathways (adjusted p < 0.05) were those related to cell junctions/projections, axons, synapses, calcium ion binding, and microtubule-based processes (Fig 1 and S3 Table). These terms are consistent with the dramatic changes to synaptic structure and function caused by PrP^Sc, as documented in our previous studies [14,15].

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Top pathways altered by PrP^Sc treatment of hippocampal neurons based on phosphoproteomic analysis.

Pathway enrichment using GO terms for Molecular Function, Biological Process, and Cellular Component was performed on phosphoproteomic data obtained after treatment of neurons with PrP^Sc for 1 h. The 20 most highly enriched pathways are ranked by normalized enrichment score (NES), all with adjusted p-values<0.05. See S3 Table for full descriptions of each pathway.

https://doi.org/10.1371/journal.ppat.1014314.g001

PrP^Sc causes modest transcriptional changes in pathways linked to mitochondria, neurotransmission, and extracellular matrix

Triplicate samples for RNA-seq analysis were collected from hippocampal neurons treated with PrP^Sc or mock-purified material for 0.5, 2, 4, and 24 hours. S4 Table lists differentially expressed genes and S3 Fig shows volcano plots of transcriptional changes at each time point. Using cut-off values for each gene of │log2fold change│ > 0.25 and p-value <0.05, the number of genes showing significant transcriptional change was modest: 7 upregulated/24 downregulated at 30 minutes; 10 upregulated/22 downregulated at 2 hr; 10 upregulated/12 downregulated at 4 hr; and 86 upregulated/47 downregulated at 24 hrs. Nevertheless, GSEA analysis using the entire transcriptional output yielded statistically significant pathway enrichments (adjusted p-values <0.05) (Fig 2A - 2D and S5 Table). The most prominent changes were observed at 0.5 hours (upregulation of ribosomal and mitochondrial pathways), likely reflecting an early response to PrP^Sc-induced oxidative stress, and at 24 hours (upregulation of collagen and extracellular matrix pathways), suggesting remodeling of the extracellular matrix following dendritic spine retraction. The larger number of gene expression changes at 24 hours compared to earlier time points is consistent with a slowly evolving transcriptional response that may occur over several days.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Top pathways altered by PrP^Sc treatment of hippocampal neurons based on RNA-seq analysis.

GSEA using GO terms for Molecular Function, Biological Process, and Cellular Component was performed on RNA-seq data obtained after treatment of neurons for (A) 30 min, (B) 2 hrs, (C) 4 hrs, and (D) 24 hrs. The 20 most highly enriched pathways are ranked by normalized enrichment score (NES). See S5 Table for full descriptions of each pathway.

https://doi.org/10.1371/journal.ppat.1014314.g002

Application of a chemo-omic pipeline

We leveraged two public databases, L1000 and P100 [16–19,21], which contain transcriptomic and phosphoproteomic signatures, respectively, from a panel of cell lines treated with a large number of compounds (referred to as perturbagens). The L1000 database contains over 1 million transcriptomic signatures derived from 9 different core cell lines that have been subjected to >30,000 small molecule perturbagens. The P100 database contains phosphoproteomic signatures from 6 cell lines treated with 90 small molecule perturbagens. Our approach (Fig 3, and Materials and Methods) was to compare the transcriptomic and phosphoproteomic signatures of PrP^Sc-treated neurons with the corresponding signatures in these databases, with the goal of identifying compounds that produced the inverse effect on the expression levels and phosphorylation states of each of the input genes or proteins; such compounds would be predicted to reverse the biological effect of PrP^Sc. The L1000 database was queried using as input the transcriptomic signatures of neurons treated with PrP^Sc for 0.5, 2, 4, and 24 hrs, and the P100 database using the phosphoproteomic signature of neurons treated with PrP^Sc for 1 hr. Both databases were queried using the CMap portal (https://clue.io/). In the output lists from these queries, we prioritized perturbagens with the most statistically significant anti-correlation with the input signatures, and those with identified molecular targets. In cases where the perturbagens had relatively low specificity for their targets, we identified inhibitors that are more specific, based on literature and commercial sources. The final list of 52 compounds is shown in Table 1.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Top-scoring compounds from the chemo-omics pipeline based on their ability to prevent prion synaptotoxicity. Compounds derived from the L1000 transcriptomic databased and the P100 proteomics database are indicated by checkmarks in the columns labeled RNA-seq and Phosphoproteomics, respectively. Compounds with no checkmarks in either column were selected based on similarity of mechanism of action to compounds derived from the two databases. Compounds are grouped according to their presumed molecular target, shown in the last column of the table, with the three prioritized kinases (CaMKII, PKC, and GSK3β) listed at the top. Dendritic spine retraction was measured with the standard assay [14]. Hippocampal neurons were treated with PrP^Sc for 24 hrs at a concentration of 250 nM for each compound, after which they were stained with AF488-phalloidin to visualize dendritic spines, as illustrated in S4 Fig. Spine numbers were quantitated, and inhibition of spine retraction was estimated by comparison to neurons treated with vehicle-only (S7 Table). For simplicity, inhibition of spine retraction is listed in the table using a semi-quantitative scale: +++, maximal effect; ++, moderate effect; + , minimal effect; -, no effect. Compounds that were toxic at 250 nM are indicated (see S5 Fig for an example of a toxic compound).

https://doi.org/10.1371/journal.ppat.1014314.t001

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Chemo-transcriptomic-phosphoproteomic pipeline to identify prion synaptotoxic signaling pathways and components.

Hippocampal neurons treated with PrP^Sc were processed for phosphoproteomics at 1 hr, and for RNA-seq at 30 min, 2 hr, 4 hr, and 24 hr. Synaptotoxic pathways enriched for differentially phosphorylated proteins and differentially expressed genes were identified. Then the lists of differentially phosphorylated proteins and differentially expressed genes were used to interrogate the L1000 and P100 databases, respectively, yielding a final list of 52 candidate compounds that reversed these signatures. These compounds were then tested for their ability to suppress dendritic spine retraction on hippocampal neurons, resulting in a final list of inhibitor targets that represent key components of prion synaptotoxic signaling pathways.

https://doi.org/10.1371/journal.ppat.1014314.g003

A subset of compounds suppresses PrP^Sc-induced dendritic spine retraction

These 52 compounds were then tested at a concentration of 250 nM for their ability to inhibit dendritic spine retraction over 24 hrs in our standard assay. While this single-point screening approach enabled efficient comparison across a large set of compounds, it may have overlooked compounds with distinct dose-response characteristics or temporal dynamics. The effects of the compounds were scored on a semi-quantitative scale (- to +++) in Table 1, with the underlying spine counts shown in S7 Table and representative fluorescence micrographs shown in S4 Fig. Twelve of the compounds were toxic at the tested dose and were eliminated from further consideration (see S5 Fig for an example of a toxic compound). Of the remaining compounds, we identified 17 that maximally suppressed dendritic spine retraction (+++) at a concentration of 250 nM with minimal toxicity. The most consistent effects were observed for inhibitors of three kinases (CaMKII, PKC, and GSK3β), which we decided to prioritize for further analysis by biochemical and cell biological techniques. Fig 4 shows quantitative measurements of the effects of two different, highly specific inhibitors of each of these three kinases on dendritic spine number, including appropriate negative controls.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Inhibitors of CaMKII, PKC, and GSK3β prevent spine retraction. (A,B) Primary hippocampal neuron cultures were treated with PrP^Sc or with mock-purified material (Mock) in the presence of the indicated compounds at 250 nM. Neurons were fixed after 24 hrs and stained with AF488-phalloidin to visualize dendritic spines. In the top row of each panel (no compound), white arrowheads in images of mock-treated neurons and red arrowheads in images of PrP^Sc-treated neurons indicate healthy and retracted spines, respectively. Scale bars are 5 μm. (B, C) Bar graphs show measurements of dendritic spine number per µm for the indicated treatments. Pooled measurements were collected from 5–7 neurons, including 15-25 dendritic regions from at least 2 independent experiments. Data are shown as the mean ± SEM. Statistical analysis was performed using unpaired t-tests; Significance is indicated by ns (not significant), *p < 0.05, **p < 0.01, ***p < 0.001, ****P < 0.0001.

https://doi.org/10.1371/journal.ppat.1014314.g004

CaMKII, PKC and GSK3β are components of well-known and richly connected signaling networks, consistent with the fact that they were all identified based on analysis of both phosphoproteomics and RNA-seq datasets. There is abundant literature documenting their importance in synaptic function (see Discussion). These kinases are themselves regulated by upstream kinases and phosphatases that alter their catalytic activities and subcellular distributions. To further document their contributions to prion synaptotoxic signaling, we therefore wished to investigate changes in their phosphorylation state and intracellular localization in hippocampal neurons exposed to PrP^Sc.

We initially employed western blotting with phospho-specific antibodies to look for alterations in the phosphorylation states of the three kinases in neuronal lysates but did not detect consistent changes in response to PrP^Sc treatment at time points ranging from 1-24 hours (S6 Fig). This result suggested that PrP^Sc might be altering the subcellular localization and/or phosphorylation state of only a subset of kinase molecules, thereby changing their distribution with respect to dendritic spines as part of the process of spine collapse, a phenomenon we observed previously for p38 MAPK [15]. We therefore undertook a detailed analysis of the localization of CaMKII, PKC and GSK3β by immunofluorescent staining using antibodies specific for phosphorylated forms of the kinases in conjunction with fluorescent phalloidin to visualize dendritic spines.

PrP^Sc induces rapid translocation of CaMKII to the postsynaptic density

We immunostained hippocampal neurons treated with PrP^Sc for 1 hr using antibodies for total CaMKII, two phosphorylated forms of CaMKII (pT286 and pT305/6), and PSD95. Phosphorylation of T286 occurs upon activation of CaMKII, while phosphorylation of T305/6 is inhibitory [22,23]. We analyzed neurons at 1 hr, well before dendritic spine retraction becomes evident (S1 Fig), in order to capture the earliest events triggered by NMDA-mediated Ca²⁺ influx, which we know occurs within minutes of PrP^Sc exposure [15]. We observed a dramatic translocation of total CaMKII to dendritic spines following application of PrP^Sc (Fig 5A, 5B, 5E). Interestingly, CaMKII translocation to spines was accompanied by increased staining for PSD95, indicating an initial enlargement of the postsynaptic density (PSD) prior to subsequent spine retraction (Fig 5C,5D, and 5H). When we individually quantitated the amounts of the phosphorylated forms of CaMKII within the PSD, there was a significant decrease in the amount of CaMKII phosphorylated on the inhibitory T305/6 site compared to mock-treated cultures, consistent with preferential translocation of the de-inhibited form of the enzyme (Fig 5A, 5B, and 5F). Surprisingly, the proportion of CaMKII-pT286 (activated form) decreased (Fig 5C, 5D, and 5G), possibly due to rapid dephosphorylation by PSD-associated phosphatases, while CaMKII translocation to spines occurred concurrently. Inhibition of CaMKII with KN-93 prevented translocation of CaMKII to spines, and normalized its ratio with NR2B, the NMDA subunit to which it binds at the PSD when activated (S7A, and S7B Fig).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. PrP^Sc causes rapid translocation of CaMKII to dendritic spines.

Hippocampal neurons were treated for 1 hr with mock-purified material (A, C) or purified PrP^Sc(B, D). After fixation, neurons were stained with fluorescent phalloidin (gray), along with antibodies to CaMKII (red) and CaMKII-pT305/306 (green) (A, B); or with antibodies to CaMKII-pT286 (red) and PSD-95 (green) (C, D). Boxed regions in the square panels are shown at higher magnification in the rectangular panels to the right or below (A1-2, B1-2, C1-3, D1-3). Dotted lines in the higher magnification panels outline the positions of intact spines, based on phalloidin staining. Scale bars on all panels are 5 μm. Violin plots show quantitation of Total CaMKII (E), p-Thr305-6/Total CaMKII (F), p-Thr286/Total CaMKII (G), and Total PSD95 (H) in spine regions. Measurements were collected from 5-7 neurons, 12-15 dendritic regions from at least 2 independent experiments. Each data point for the PrP^Sc-treated samples was normalized to the average Mock value from the same experiment. Dotted red lines in each violin plot indicate the median. Statistical analysis was performed on SEM values using unpaired t-tests. Significance is indicated as *p < 0.05, **p < 0.01, ****p < 0.0001.

https://doi.org/10.1371/journal.ppat.1014314.g005

PrP^Sc causes accumulation of primed PKC in endosomes within the soma and then in dendritic spines

We immunostained hippocampal neurons treated with PrP^Sc or mock purified material for 0.5, 2, 4, and 24 hr using antibodies for total PKC and PKC phosphorylated on S660 (Fig 6). The latter is a priming phosphorylation that is necessary to stabilize newly synthesized molecules of PKC prior to their subsequent activation by signal transduction processes [24]. We found that, at early time points during PrP^Sc treatment (0.5 and 2 hrs), PKC-pS660 accumulated prominently in a juxtanuclear compartment within the soma (Fig 6B, 6C, and 6G), a pattern similar to what is observed after simulation of neurons with the strong PKC activator phorbol ester (Fig 6F). This juxtanuclear compartment likely represents a previously described subset of recycling endosomes (the pericentrion), which are thought to amplify the PKC signal during sustained activation [25,26]. During this same period, there was a decrease in the amount of PKC-pS660 within dendritic spines (Fig 6B, 6C, and 6H). By 4 hours, the amount of PKC-pS660 in spines began to increase (Fig 6D and 6H), coinciding with significant PrP^Sc-induced spine retraction observed at this time point (S1 Fig), and suggesting that PKC activation may mediate early synaptic loss. At 24 hours, PKC-pS660 accumulated in both the soma and dendritic spines (Fig 6E, 6G, and 6H). Pharmacological inhibition of PKC with Ro 32–0432 prevented these changes in the soma but not in the dendritic spines at 4 and 24 hours (S8 Fig). This timing implicates PrP^Sc-triggered PKC signaling as a critical mediator of early synaptotoxicity via spine retraction, highlighting 4 hours as a key window for synaptic degeneration driven by PKC-dependent pathways.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. PrP^Sc causes accumulation of primed PKC in endosomes within the soma and then in dendritic spines.

Hippocampal neurons were treated either with mock-purified material (A), or with purified PrP^Sc for 30 min (B), 2 hrs (C), 4 hrs (D) or 24 hrs (E). After fixation, neurons were stained with fluorescent phalloidin (gray), along with antibodies to total PKC (green) and PKC-pS660 (red). Boxed regions in the square panels are shown at higher magnification in the rectangular panels below (A1-4, B1-4, C1-4, D1-4). Dotted lines in the higher magnification panels outline the positions of intact spines, based on phalloidin staining. Note the accumulation of vesicles in the soma (indicated by asterisks) that stain for PKC-pS660 after 30 minutes of PrP^Sc treatment. This pattern is similar to that seen in neurons treated for 30 minutes with the strong PKC activator phorbol ester (F). Scale bars in all panels = 5 μm. Violin plots show quantitation of PKC-pS660/total PKC ratios within soma regions (G) and spine regions (H). Measurements were collected from 5-7 neurons, 30-45 dendritic and somatic regions from at least 2 independent experiments. Each data point for the PrP^Sc-treated samples was normalized to the average Mock value from the same experiment. Dotted red lines in each violin plot indicate the median. Statistical analysis was performed on SEM values using unpaired t-tests. Significance is indicated as ns (not significant), *p < 0.05, **p < 0.01, ****p < 0.0001.

https://doi.org/10.1371/journal.ppat.1014314.g006

PrP^Sc causes accumulation of activated GSK3β within dendritic spines

Based on the known activation of GSK3β by auto phosphorylation at Y216 [27], we examined whether PrP^Sc treatment alters phosphorylation at this site. Since four different GSK3β inhibitors (SB216763, TWS119, Tideglusib, and AR-A014418) prevented dendritic spine retraction after 24 hrs of PrP^Sc treatment (Figs 4 and 7C–7E - 7E), we hypothesized that PrP^Sc would cause accumulation of enzymatically active GSK3β-pY216 in dendritic spines, contributing to spine collapse. To test this prediction, we stained hippocampal cultures treated with PrP^Sc for 1 hr with antibodies to GSK3β-pY216 and total GSK3β and then imaged the ratio of the two signals in the region of dendritic spines, as marked by phalloidin staining. As predicted, we found that the amount of GSK3β-pY216 in dendritic spines was significantly increased after 1 hour of PrP^Sc treatment (Fig 7A and 7B).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 7. PrP^Sc increases active pY216-GSK3β in dendritic spines, and this effect is prevented dose-dependently by GSK3β inhibitors.

(A) Hippocampal neurons were treated for 1 hr with mock-purified material (Mock, left-hand group of panels) or purified PrP^Sc (PrP^Sc, right-hand group of panels). After fixation, neurons were stained with fluorescent phalloidin, along with antibodies to total GSK3β and GSK3β-pY216. Upper panels show images of phalloidin staining (gray). Lower panels show merged images of total GSK3β (green) and GSK3β-pY216 (red) staining. The boxed regions in the large, square panels are shown at higher magnification in smaller, rectangular panels to the right (A1-4). Dotted lines in the smaller panels outline the positions of intact spines, based on phalloidin staining. Spines in PrP^Sc-treated neurons (lower right) show enhanced GSK3β-pY216 /total GSK3b staining (yellow) compared to mock (lower left). (B) Violin plot showing quantification of GSK3β-pY216/total GSK3β within spine regions, normalized as described in Figs 5 and 6. Significance is indicated as ****p < 0.0001. (C) Representative fluorescence micrographs showing inhibition of spine retraction by Tideglusib and AR-A014418 (both at 30 nM). Neurons were pretreated for 2 hrs with the compounds, after which they were treated in the continued presence of the drugs for 24 hrs with either mock-purified material or purified PrP^Sc. After fixation, neurons were stained with Alexa 488-labeled phalloidin for visualization and counting of dendritic spines. The top pair of panels show neurons that were not pretreated with inhibitors. Scale bars = 5 μm. (D, E) Dose response curves for inhibition of PrP^Sc-induced dendritic spine retraction by the GSK3β inhibitors Tideglusib and AR-A014418. Conditions were the same as in panel C. Pooled measurements were collected from 30-45 dendritic regions on 5–7 neurons from at least 2 independent experiments. The y-axis shows spine density (number/µm) in PrP^Sc-treated neurons exposed to drug as a percentage of spine density in mock-treated neurons exposed to drug. Both drugs inhibited spine retraction with K_i values of ~30 nM.

https://doi.org/10.1371/journal.ppat.1014314.g007

Pathway analysis

Prior to applying the chemo-omic pipeline, we performed gene set enrichment analysis (GSEA) of the phosphoproteomic and transcriptomic data to identify cellular pathways altered by PrP^Sc treatment of cultured hippocampal neurons. Significant changes in the phosphoproteome were detected after 1 hr of PrP^Sc treatment, consistent with the hypothesis that acute exposure of neurons to prions activates a network of phosphorylation-dependent signaling pathways that are triggered by the initial interaction of PrP^Sc with cell-surface PrP^C [13,15]. Gene set enrichment analysis of the phosphoproteomic data returned GO terms related to cell junctions/projections, axons, synapses, calcium ion binding, and microtubules among the top 20 most highly regulated pathways. A role for phosphorylation-dependent processes in prion synaptotoxicity is not surprising, since specific kinases and phosphatases are known to be involved in remodeling of dendritic spine morphology and changes in synaptic transmission [28].

In contrast to the phosphoproteomic changes, the transcriptomic changes we observed in PrP^Sc-treated neurons were relatively modest across all time points (<50 genes up or down regulated). Nevertheless, GSEA analysis using the entire transcriptome yielded statistically significant pathway enrichments (adjusted p-values<0.05). The most prominent enrichments were observed at 0.5 hours (upregulation of ribosomal and mitochondrial pathways), likely reflecting an early response to PrP^Sc-induced oxidative stress, and at 24 hours (upregulation of collagen and extracellular matrix pathways), indicative of synaptic degeneration and consequent remodeling of the ECM. The modest effect of PrP^Sc on neuronal gene expression likely reflects the fact that, even after 24 hours of PrP^Sc treatment, when substantial numbers of dendritic spines have retracted and synaptic transmission has declined, neuronal viability is unaffected and dendritic shafts remain intact. The pattern of gene expression that we observe here is consistent with RNA-seq studies of the brains of prion-infected mice, in which neurons display relatively few transcriptional changes until the end-stage of the disease, with the earliest detectable changes occurring in astrocytes and microglia during the presymptomatic phase [29–31]. Taken together, our results suggest that the earliest response of neurons to prion exposure occurs at the level of protein phosphorylation-dependent signaling, prior to major compensatory changes in gene expression.

It is important to note that our in vitro hippocampal neuron model exhibits a rapid and more synchronized progression of synaptotoxicity compared to the chronic and asynchronous course of prion disease in vivo. However, this condensed time course has allowed us to capture both early signaling events and subsequent dendritic spine retraction, providing mechanistic insights that are relevant despite these temporal differences.

Chemo-omic pipeline

A novel aspect of this study is the use of a bioinformatic pipeline in which we used our transcriptomic and phosphoproteomic data to interrogate a set of public databases (L1000, P100) containing transcriptomic and phosphoproteomic signatures of a standard set of cell lines treated with large numbers of compounds at different concentrations for varying lengths of time [16–19]. The objective of this procedure was to identify compounds that produced signatures that were the inverse of the input signatures derived from PrP^Sc-treated hippocampal neurons. We predicted that such compounds would reverse the biological effect of PrP^Sc, as assayed by measurements of dendritic spine retraction.

Use of the L1000 database to connect experimentally derived transcriptomic changes with the effects of small molecules was first described by Lamb et al. [17], and there are now several published precedents for its successful application in disease contexts, including the identification of drugs for treatment of AD [32,33], atherosclerosis [34], SARS-CoV-2 infection [35], and cancer [36]. All of these cited studies used transcriptomic data to query the L1000 database; to our knowledge, ours is the first to interrogate the L1000 and P100 databases using both transcriptomic and phosphoproteomic data, respectively, and to integrate the output from both sources. While either transcriptomic or phosphoproteomic data alone can inform connectivity mapping, combining both provides complementary insights. Phosphoproteomics captures rapid PrP^Sc-triggered signaling, whereas transcriptomics reflects downstream responses. In our studies, prioritizing compounds supported by both datasets increased confidence in candidate selection and allowed us to identify three key kinases (CaMKII, PKC, and GSK3β) underlying prion-induced synaptic degeneration. This integrative approach enhanced robustness and suggested a strategy for multi-target interventions in prion disease.

We note that, while the L1000 dataset enables large-scale connectivity mapping, it directly measures only a limited set of ~978 landmark genes and relies on computational inference for the remainder of the transcriptome, which may reduce sensitivity for gene-specific effects. Subsequent bioinformatic studies have used machine-learning to transform the L1000 profiles to RNA-seq-like, whole-transcriptomic signatures [37,38], but it is uncertain whether use of this pipeline, which still relies on an imputation procedure, would have improved our ability to detect inhibitory compounds. Regardless, the limitation of the L1000 gene sets is mitigated here by our use of full RNA-seq and phosphoproteomic datasets as inputs, together with experimental validation of predicted compounds in primary neurons.

CaMKII

CaMKII has long been associated with synaptic changes underlying learning, memory, and neurodegenerative conditions [22,23]. The participation of CaMKII in prion synaptotoxicity is consistent with the rapid influx of Ca²⁺ into cultured hippocampal neurons that we observe within minutes of exposure to PrP^Sc, which would increase binding of Ca²⁺ to calmodulin (CaM), causing activation of CaMKII [15]. This Ca²⁺ influx is mediated primarily by NMDARs, since it is abolished by NDMAR blockers, which also prevent dendritic spine retraction. Canonically, during the process of long-term potentiation (LTP), Ca²⁺-CaM binding to CaMKII induces phosphorylation of T286, resulting in autonomous kinase activation, and translocation of CaMKII to the PSD, where it binds to the C-terminal cytoplasmic tail of NR2B-containing NMDA receptors [22,23]. Consistent with this model, we observed a large shift of CaMKII to the dendritic spines of hippocampal neurons exposed to PrP^Sc (Fig 5). Interestingly, however, the PSD-localized CaMKII is not phosphorylated at T286, perhaps reflecting the action of phosphatases that rapidly dephosphorylate the active enzyme upon arrival at the PSD, although this assertion requires experimental verification. In contrast, spine-localized CaMKII showed reduced phosphorylation at T305/6, an inhibitory site, consistent with prior activation of the kinase. Although the massive accumulation in spines of CaMKII that is not phosphorylated at either the activating T286 site or the inhibitory T305/6 site is paradoxical, and complicates a definitive interpretation, we note that there is little precedence for translocation of the enzyme without prior activation [22,23]. The initial burst of CaMKII translocation to dendritic spines, coupled with the transient increase in PSD95 staining we observed (Fig 5), is consistent with a transient enlargement of spines, which is then followed by spine collapse at later times. Activated CaMKII has been shown to regulate the actin cytoskeleton in spines via Ca²⁺-induced phosphorylation of cytoplasmic actin-binding proteins [39], a process that likely contributes to the enlargement and then retraction of spines in response to PrP^Sc. Treatment with the CaMKII inhibitor KN-93 reversed the initial translocation of CaMKII to dendritic spines, normalized the CaMKII/NR2B ratio, and ultimately prevented spine collapse following PrP^Sc treatment (Figs 4 and S6). It remains to be determined whether PrP^Sc-dependent activation of NMDARs is due to direct binding of PrP^Sc to NMDAR subunits, binding other membrane proteins, or perturbation of the lipid bilayer [13,40].

GSK3β

We have shown that GSK3β inhibitors block PrP^Sc-induced dendritic spine loss, and that PrP^Sc treatment causes accumulation within spines of the constitutively active form of GSK3β that is phosphorylated on residue Y216. GSK3β is a multi-functional Ser/Thr kinase that is enriched in dendritic spines, and is involved in both physiological and disease processes in the brain via phosphorylation of a number of key substrates, including PSD95, KLC3 and tau [49,50]. Unlike many other kinases, GSK3β is constitutively active due to autophosphorylation at Y216, and upstream regulators typically act to reduce GSK3β enzymatic activity via an inhibitory phosphorylation at residue S9 [27]. Relevant to the results presented here, GSK3β plays an important role in the induction of NMDA receptor-dependent long-term depression (LTD) during which GSK3β translocates to the PSD and phosphorylates PSD95, leading to AMPAR internalization [51]. Moreover, phosphorylation of phosphatidylinositol 4 kinase type IIα (PI4KIIα) by GSK3β stabilizes a constitutive pool of NMDARs at the synapse [49,52]. Over-activation of the latter pathway by PrP^Sc could therefore increase the number of synaptic NMDARs, thereby augmenting Ca²⁺ influx and activation of CaMKII.

Interestingly, GSK3β also plays a key role in Alzheimer’s disease, primarily by its ability to phosphorylate the microtubule-associated protein, tau [50]. In this context, Aβ oligomers activate GSK3β, possibly via the kinase Pyk, leading in turn to enhanced phosphorylation of tau and its aggregation into amyloid fibrils [53–55]. GSK3β inhibitors reverse electrophysiological and behavioral deficits in AD mouse models by targeting this pathway, and they have been tested in clinical trials for treatment of AD [56]. Although upstream components of PrP^Sc and Aβ synaptotoxic pathways are distinct (see above), it is possible that they share downstream, GSK3β-dependent processes.

A kinase-dependent signaling network for prion synaptotoxicity

Collectively, our results provide evidence for a signal transduction network stimulated by PrP^Sc that contains four major nodes: NMDARs, CaMKII, PKC, and GSK3β (Fig 8). In this scheme, PrP^Sc generated on the cell surface activates NMDARs, leading to massive Ca²⁺ influx, phosphorylation of CaMKII and PKC, and translocation of these kinases to the PSD, where they further potentiate NMDAR channel activity and trafficking. GSK3β may augment these responses by stabilizing NMDARs at the synapse. Importantly, this signaling network utilizes the same molecular components that are responsible for LTP and LTD, suggesting that prions co-opt normal, physiological processes underlying learning and memory to produce neurotoxic effects. We have recently shown that newly converted, membrane-anchored PrP^Sc, rather than exogenous, extracellular PrP^Sc, is the trigger for prion synaptotoxicity, suggesting that conversion of PrP^C to PrP^Sc on the plasma membrane flips the downstream signaling network into an over-drive mode [13]. It remains to be determined whether PrP^Sc-dependent activation of NMDARs is due to direct binding of membrane-anchored PrP^Sc to NMDAR subunits, indirect interaction via other membrane proteins, or perturbation of the lipid bilayer [40].

Download:

• PNG
  larger image
• TIFF
  original image

Fig 8. Schematic showing effect of PrP^Sc on CaMKII, PKC, and GSK3β in dendritic spines. In the initial stage of prion infection, PrP^Sc activates NMDARs in the postsynaptic membrane, either directly as shown here, or indirectly via interaction with other membrane proteins or the lipid bilayer (not depicted).

The resulting influx of Ca²⁺ ions activates CaMKII by phosphorylation of T286 and dephosphorylation of T305/306, accompanied by CaMKII translocation to the PSD where it binds to NMDARs via the cytoplasmic tails of NR2B subunits (not depicted). PrP^Sc also causes primed PKC phosphorylated on S660, as well as activated GSK3β phosphorylated on Y216, to translocate to the PSD, where they can associate with PSD-95. Together, these signaling events lead to collapse of the actin cytoskeleton and retraction of the dendritic spine. Inhibitors of CaMKII, PKC, and GSK3β (KN-62, Ro 32-0432, and TWS119, respectively) prevent spine retraction. Amino acid sites in the three kinases that are phosphorylated are shown in red circles, and those that are dephosphorylated are shown in green circles. Created in BioRender. https://BioRender.com/01u1qp9.

https://doi.org/10.1371/journal.ppat.1014314.g008

In a previous study [15] using the same neuronal culture system as employed here, we identified p38 MAPK as a key player in prion-induced synaptotoxicity, based on both pharmacologic and genetic evidence. PrP^Sc treatment of hippocampal neurons increased the amount of phosphorylated p38 MAPK in dendritic spines, and specific p38 MAPK inhibitors, as well as a dominant negative mutant of p38 MAPK, prevented PrP^Sc-induced retraction of spines and associated decrements in synaptic transmission. Although p38 MAPK was not prioritized in the chemo-omic pipeline used here, there is ample evidence connecting it to the kinase-dependent signaling network proposed above. p38 MAPK directly phosphorylates GSK3β on serine 389/390, possibly modulating its susceptibility to other regulatory phosphorylations [57]. p38 MAPK is also activated during NMDAR-dependent LTD and contributes to NMDAR/Ca²⁺-mediated excitotoxic damage [58,59].

Several other studies have highlighted a role for Ca²⁺-mediated excitotoxic processes in prion neurotoxicity. For example, it has been reported that prion infection results in upregulation of Arc/Arg3.1, an immediate early response gene activated by neuronal activity [60,61]. Our observation that PrP^Sc stimulates rapid, NMDAR-dependent Ca²⁺ influx [15] followed by activation of Ca²⁺-dependent signaling processes (this work) is consistent with this idea. Although there have not been previous global, phosphoproteomic analyses of prion infected mice or cultured cells, a relevant recent study showed that expression of a neurotoxic mutant PrP in mouse brain or cultured neurons enhances NMDAR activity, activates CaMKII and PKC, and triggers excitotoxic damage [62]. A similar mechanism may be operative in mice and cells we created that express a different neurotoxic PrP mutant, which is associated with aberrant ion channel activity [63–65]. These mutant PrP molecules may be triggering some of the excitotoxic signaling processes that are induced by prion infection.

Our results highlight possible therapeutic avenues for treatment of prion diseases. Indeed, the chemo-transcriptomic-phosphoproteomic pipeline we have employed here has typically been leveraged to identify and repurpose therapeutic compounds [32,34–36]. Inhibitors of NMDARs, CaMKII, PKC, GSK3β, and p38 MAPK have all been tested for treatment of AD and neurological disorders [66], but only memantine, an NMDAR open-channel blocker, has been tested in prion-infected mice, where it was shown to extend survival times [67]. Our results suggest that further investigation of these inhibitors as treatments for prion diseases, particularly in the presymptomatic phase, is warranted. These compounds, along with others we have discovered using cell-based screening assays [68–71], may represent useful adjuncts to treatments based on reducing PrP^C expression [72]. The mechanistically informed inhibitor screening strategy we have employed, derived from both phosphoproteomic and transcriptomic data, naturally lends itself to discovery of combination therapies with drugs targeting multiple neurotoxic signaling nodes simultaneously.

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 Chobanian & Avedisian School of Medicine Institutional Animal Care and Use Committee (IACUC).

PrP^Sc purification

PrP^Sc purification from the brains of terminally ill, RML prion-inoculated C57BL/6J mice was performed as described previously utilizing pronase digestion and sodium phosphotungstate precipitation [73]. Parallel, mock purifications were performed using age-matched, uninfected brains. Purity was assessed by SDS-PAGE with silver staining and western blotting. In all experiments, purified PrP^Sc was added to neuronal cultures at a final concentration of 4.4 μg/ml to be consistent with our previous studies [14,15]. Mock-treated samples received an equivalent amount of material purified from uninfected brain tissue.

Primary hippocampal neuronal cultures

Timed-pregnant C57BL/6J mice (wild-type, WT) were obtained from The Jackson Laboratory (Bar Harbor, ME). Hippocampal neurons were isolated from postnatal day 0 (P0) pups as described previously [74]. Two types of neuronal-astrocyte culture systems were used.

For dendritic spine analysis, neurons were plated at low density (75 cells/mm²) on poly-L-lysine-coated coverslips. After several hours, coverslips were inverted over an astrocyte feeder layer, with each coverslip held just above the astrocyte layer by several wax dots. The astrocyte feeder layer consisted of astrocytes cultured from postnatal day 0 (P0) pups. Cultures were maintained in Neurobasal medium supplemented with B27 (NB/B27) for 18–21 days before treatment with PrP^Sc.

For omics-based analyses (transcriptomics, proteomics, and phosphoproteomics), which required larger number of neurons, the neuronal and astrocyte feeder layers were reversed. Hippocampal neurons were cultured at high density on poly-L-lysine-coated plates. After several hours, astrocytes plated on coverslips were inverted onto the neuron cultures. These cultures were maintained under the same medium conditions (NB/B27).

Proteomic and phosphoproteomic analyses

Neurons were lysed in 8M Urea 50mM Triethylammonium bicarbonate (TEAB) buffer containing protease inhibitors (Sigma) and phosphatase inhibitors (Roche). After brief sonication on ice, the samples were reduced by addition of dithiothreitol (5 mM final concentration) for 60 min at room temperature and alkylated by the addition of iodoacetamide (5 mM final concentration) and incubation at room temperature for 30 min in the dark. Proteins were diluted with 50 mM TEAB to bring the urea concentration below 1 M, and digested with MS-grade trypsin (1:50 enzyme to protein ratio) at 37°C overnight followed by the addition of formic acid to a final concentration of 1%. The resulting peptides were desalted using a C18 Tips (Thermo Scientific) as per the manufacturer’s instructions.

Tryptic peptides were labeled with isobaric, tandem mass tags (TMT), followed by isolation of phosphopeptides on Fe-NTA beads (CubeBiotech). A portion (10%) of each sample prior to phosphopeptide isolation was saved for total proteome analysis. Briefly, peptides (90%) were resuspended in 800 µl of sample buffer consisting of 0.5% TFA and 80% ACN, and shaken at 1,000 rpm on a temperature-controlled shaker at 20°C for 15 min. Beads (10 µg per 100 µg of peptides) were washed with washing buffer consisting of 0.1% TFA and 80% ACN, then added to the resuspended peptides. Samples and beads were co-incubated for 30 min (shaken at 1000 rpm at 20°C). Bound phosphopeptides were washed twice with 800 µl of washing buffer and then serially eluted twice with 200 µl of 7.5% ammonium hydroxide and 50% ACN. The two eluates were combined and dried in a SpeedVac prior to LC/MS.

Quantitative LC-MS analysis of peptides

LC-MS analysis was performed using a Q-Exactive HFX mass spectrometer(Thermo Scientific), interfaced with and EASY nLC 1200 ultra-high pressure pump system. Peptides were loaded onto a C18 pre-column (75 μm i.d. × 2 cm, 3 μm, 100 Å, ThermoFisher Scientific) and then separated on a capillary column (75 μm i.d. × 50 cm, 2 μm,100 Å, ThermoFisher Scientific) using gradient elution. The nLC flow rate was 250 nl/min. Peptides were injected and separated over a 90 min gradient. The mobile phase A was 0.1% FA-2% ACN-Water, and mobile phase B was 0.1% FA-80% ACN-Water. The gradient consisted of 2%-6% of mobile phase B for 5 min, was increased to 20% mobile phase B over 39 min, then increased to 35% over 23 min, and reached 98% over 3 min and maintained at 98% for 3 min before decreasing to 5% over 3 min. The MS instrument was operated in positive ion mode over a full mass scan range of m/z 350 − 1500 at a resolution of 120,000 with a normalized AGC target of 1e⁵. The source ion transfer tube temperature was set at 275°C and a spray voltage was set to 2.5kv. Data was acquired on data-dependent acquisition (DDA) mode. MS2 scans were performed at 45,000 resolution with a normalized collision energy of 33. The maximum injection time was set to 400ms. Dynamic exclusion was enabled using a time window of 45 s.

Phosphoproteomic data analysis

MS2 spectra were processed and searched by MaxQuant (version 1.6.7) against a database containing native (forward) mouse protein sequences (UniProt) and reversed (decoy) sequences for protein identification. The search allowed for two missed trypsin cleavage sites, and phosphorylation (Ser/Thr/Tyr) was included as variable modification. Both carbamidomethylation of cysteine residues and TMT (peptide N-termini, K residues) were set as fixed modifications. Precursor ions were searched with a maximum mass tolerance of 4.5 ppm and fragment ions with a maximum mass tolerance of 20 ppm. The candidate peptide identifications were filtered, assuming a 1% FDR threshold based on searching the reverse sequence database. Quantification was performed using the TMT reporter on MS2. Proteins with less than 50% missing value across all samples were kept for quantification. The reporter ion intensities were log-transferred normalized. Bioinformatic analysis was performed in the R statistical computing environment (version 3.6) using Omics Notebook [75]. The p-value was obtained using moderated Student t-test (Limma) and multiple testing was controlled using Benjamini-Hochberg FDR method. Pathway enrichment was performed using GSEA [20] mapped to GO terms for Molecular Function, Biological Process, and Cellular Component Ontology.

RNA-seq analysis

Total RNA was isolated from cultured neurons using the RNeasy Plus Mini Kit (Qiagen) following the manufacturer’s protocol. Briefly, samples were lysed in the provided RLT buffer containing 1% β-mercaptoethanol and homogenized using a sonicator. The lysate was then passed through an RNA-cleaning column to remove contaminants, and the RNA was eluted in RNase-free water. The integrity and concentration of the isolated RNA were assessed using the High Sensitivity RNA Pico 6000 assay run on an Agilent 2100 Bioanalyzer (Agilent Technologies, USA). Only samples with RNA Integrity Numbers (RIN) greater than 8 were used for sequencing. RNA samples were then subjected to DNase I treatment (Thermo Fisher Scientific) to remove any contaminating genomic DNA, followed by another round of purification using the RNeasy column.

RNA-seq was performed in-house at the Boston University Microarray & Sequencing Resource Core. Libraries were prepared using Illumina’s TruSeq Stranded mRNA Library Preparation kit according to the manufacturer’s protocol (Illumina, USA). To enrich for mRNA, 200 ng of total RNA for each sample was incubated with magnetic Oligo-(dT) beads. Enriched mRNA was fragmented at 94°C for 8 minutes. First strand cDNA was generated using random primers, and incorporation of dUTP occurred during synthesis of second strand cDNA to indicate strandedness. Double-stranded cDNA then underwent A-tailing and adaptor ligation. Sample-specific multiplex indices were incorporated during PCR amplification (15 cycles) according to the manufacturer’s protocol (Illumina, USA). Size distribution and molarity of amplified cDNA libraries were assessed via the Bioanalyzer High Sensitivity DNA Assay (Agilent Technologies, USA). All cDNA libraries were sequenced on an Illumina NextSeq 500 instrument (Illumina, USA) targeting 30–40 million reads per sample (75 x 75 paired-end reads). The resulting sequencing data were processed and analyzed using standard bioinformatic pipelines. Pathway enrichment was performed using GSEA [20] mapped to GO terms for Molecular Function, Biological Process, and Cellular Component Ontology.

Transcriptomic data availability

RNA-seq data are deposited in GEO under accession number GSE325339.

Chemo-omic analysis using L1000 and P100 databases

Transcriptomic and proteomic data were utilized as inputs for a systematic multi-platform connectivity analysis pipeline to identify candidate therapeutic compounds modulating PrP^Sc-induced perturbations. RNA-seq raw counts were processed to generate normalized expression values, followed by differential gene expression analysis at each time point (0.5, 2, 4, and 24 hours). Unfiltered DGE values, separated into up- and down-regulated genes (as listed in S4 Table), were used as inputs into the L1000 database. Proteomic and phosphoproteomic datasets were analyzed to determine proteins exhibiting significant abundance or phosphorylation changes (p < 0.05), which were compiled as filtered input feature lists. These gene and protein feature sets were harmonized to recognized gene symbols and annotations compatible with downstream connectivity tools. Values for differentially phosphorylated proteins (p < 0.05, separated into up- and down changes; see S2 Table) were used as inputs into the P100 database.

For transcriptomic perturbational connectivity, two independent platforms were employed to comprehensively capture candidate compound signatures: (1) the CLUE platform (Connectivity Map Linked User Environment; https://clue.io/) accessed through Metascape [76], which calculates similarity metrics based on both the magnitude and direction of expression changes across the landmark L1000 gene set, thus capturing nuanced expression pattern alignments; and (2) L1000CDS² [77], which uses directional gene set enrichment to identify compounds predicted to reverse the up/down regulation pattern independent of fold change magnitude, emphasizing functional reversal potential. Both tools query the Library of Integrated Network-Based Cellular Signatures (LINCS) L1000 dataset [16–19] comprising thousands of small-molecule perturbations.

In parallel, the phosphoproteomic signatures were mapped to the LINCS P100 phosphoproteomic database using CLUE, enabling compound connectivity inference based on phosphorylation state modulation rather than transcriptomic profiles. This complementary analysis aimed to identify compounds capable of reversing phosphorylation changes induced by PrP^Sc.

Connectivity results from all platforms were integrated by identifying compounds recurrently predicted with high confidence (top connectivity scores or enrichment metrics) across multiple time points, genotypes, and data modalities. Candidates shared by both transcriptomic and phosphoproteomic analyses were prioritized for enhanced confidence.

Finally, the resulting shortlist of compounds underwent manual curation to assess biological plausibility in the context of neurodegeneration and prion pathology, confirmation by prior literature, and practical considerations such as commercial availability and suitability for experimental follow-up. This rigorous pipeline ensured that candidate therapeutic agents stemmed from robust, statistically grounded molecular signatures and cross-validated connectivity predictions, providing a rational basis for downstream testing.

Immunocytochemistry and microscopic imaging

Cells grown on glass coverslips were washed with phosphate-buffered saline (PBS) and fixed in 4% paraformaldehyde for 12 minutes at room temperature. For detection of intracellular antigens, cells were permeabilized with 0.1% Triton X-100 in PBS for 5 minutes prior to blocking. Non-specific binding was blocked with 1% bovine serum albumin (BSA) in PBS for 30 minutes at room temperature.

Neurons were stained with the primary antibodies listed in S6 Table and with fluorescent phalloidin (labeled with Alexa Fluor 488 or rhodamine) to label F-actin in dendritic spines. Antibodies and phalloidin were diluted in 1% BSA/PBS and applied to cells for 2 hours at room temperature. After washing with PBS, fluorophore-conjugated secondary antibodies diluted in 1% BSA/PBS were applied for 1 hour at room temperature in the dark. Cells were then washed five times (5 minutes each) in PBS and counterstained with DAPI (Invitrogen) to visualize nuclei. Coverslips were mounted using Vectashield Mounting Medium (Vector Laboratories) and stored at 4°C until imaging.

Fluorescence imaging was performed using Zeiss laser-scanning confocal microscopes with identical acquisition settings across conditions. For experiments to quantitate dendritic spines, images were acquired using a Zeiss LSM 700 microscope. For localization of proteins within dendritic spines and neuronal somata, images were acquired using a Zeiss LSM 880 microscope. A 63x oil immersion objective (numerical aperture, NA = 1.4) was used for high-resolution image capture.

Dendritic spine quantitation

Dendritic spine density was assessed to evaluate synaptotoxic effects, as previously described [14,15]. For quantification, 4–5 isolated dendritic segments were selected per image. Fields were selected randomly to minimize bias. Images were processed in ImageJ (FIJI) using a threshold optimized to include dendritic spine outlines while excluding background fluorescence [78]. Spine numbers were normalized to the corresponding dendritic length and reported as spines per micrometer (spines/μm). For each condition, 15–24 neurons were analyzed across 3–4 independent experiments. Analysis was performed in a blinded fashion.

Quantitation of immunofluorescence intensity

Fluorescence staining intensity of neuronal proteins (CaMKII, PKC, GSK3β, PSD95, NR2B) within defined regions of interest (ROIs) was quantified using ImageJ (FIJI). For each sample, ROIs were manually selected based on the area of interest, such as dendritic spines, dendrites, neuronal somata, or intracellular compartments. Fluorescence measurements were obtained by setting the analysis parameters to include Area and Integrated Density to calculate the total signal within the selected ROI.

Each image was processed to ensure uniformity, including background subtraction and optimization of thresholding to exclude non-specific fluorescence signals. The mean fluorescence intensity per pixel was calculated by dividing the integrated density by the ROI area. To control for variability between experimental conditions, all fluorescence intensities were normalized to the average intensity of the control group. Data from at least 10 randomly selected fields per condition were quantified to ensure robust sampling. For each neuron, 2–3 isolated ROIs were selected. For each condition, 15–24 neurons from 3-4 independent biological replicates were analyzed.

Western blotting

Neuronal lysates were prepared in RIPA buffer. Total protein concentration was quantified using a bicinchoninic acid (BCA) assay (Pierce). Equal amounts of protein (typically 20–30 μg) were mixed with an appropriate volume of SDS-PAGE loading buffer and heated for 5 minutes at 99°C prior to separation by SDS-PAGE using Tris-Glycine gels. Following electrophoresis, proteins were transferred onto Immobilon PVDF membranes (Millipore, Immobilon-P). Membranes were blocked for 30 minutes with 5% BSA in TBS buffer and subsequently incubated overnight at 4°C with primary antibodies. Antigen-antibody complexes were detected using HRP-conjugated secondary antibodies followed by development with ECL (Millipore) and visualization with a Bio-Rad Chemidoc XRS Molecular Imager; or using fluorescent secondary antibodies from LI-COR Biosciences (IRDye 800 (green) or IRDye 680 (red) for goat anti-mouse; IRDye 680 (red) for goat anti-rabbit), followed by visualization using the LI-COR Odyssey Classic Infrared Imaging System.

Statistical analysis

All quantification was performed in at least three independent biological replicates. Statistical comparisons were conducted using two-tailed, unpaired, Student’s t-tests (GraphPad Prism). Data are presented as bar graphs showing mean  ±  standard error of the mean (SEM), or as violin plots with the median indicated by a dotted red line. Details regarding the number of replicates (n), type of statistical test used, and significance levels are provided in the corresponding figure legends.