Multiomics of early epileptogenesis in mice reveals phosphorylation and dephosphorylation-directed growth and synaptic weakening

Summary To investigate the phosphorylation-based signaling and protein changes occurring early in epileptogenesis, the hippocampi of mice treated with pilocarpine were examined by quantitative mass spectrometry at 4 and 24 h post-status epilepticus at vast depth. Hundreds of posttranscriptional regulatory proteins were the major early targets of increased phosphorylation. At 24 h, many protein level changes were detected and the phosphoproteome continued to be perturbed. The major targets of decreased phosphorylation at 4 and 24 h were a subset of postsynaptic density scaffold proteins, ion channels, and neurotransmitter receptors. Many proteins targeted by dephosphorylation at 4 h also had decreased protein abundance at 24 h, indicating a phosphatase-mediated weakening of synapses. Increased translation was indicated by protein changes at 24 h. These observations, and many additional indicators within this multiomic resource, suggest that early epileptogenesis is characterized by signaling that stimulates both growth and a homeostatic response that weakens excitability.


INTRODUCTION
Epileptogenesis constitutes the process by which brain tissue becomes capable of generating spontaneous recurrent seizures 1 and is triggered by genetic factors or a precipitating insult, followed by a latent period before the first observation of a clinical seizure.In experimental animal models of epileptogenesis, induction of status epilepticus (SE) results in chronic recurring seizures after a latent period. 2 The identification of factors involved in the development of chronic epilepsy may form the basis for the discovery of new anti-epileptogenesis drugs.
A persistent hypothesis is that epileptic activity arises from excessive positive electrophysiological feedback on excitatory synapses.Loss or damage to inhibitory synapses 3 and growth or sprouting that results in new synapses 4,5 may cause positive feedback.However, many neurobiological processes are implicated in epileptogenesis, and a focus on excitatory/inhibitory imbalance may be limiting. 2Inflammation and the immune response, cell growth, plasticity, oxidative stress, and epigenetic reprogramming are all implicated in epileptogenesis and lead to reorganization of circuits via synapse loss, neurite outgrowth, altered synaptic strength, reorganization of the extracellular matrix, damage to the blood-brain barrier, recruitment of inflammatory cells, and gliosis. 6,7Despite this knowledge, the molecular mechanisms and timing of events in epileptogenesis are not clear.Thus, a broad and unbiased view of the protein and signaling landscape is important for further understanding.
][10] Using patient samples, a 2020 study identified greater than 100 differentially abundant proteins in the dentate gyrus 11 and a recent study found hundreds of changes in the hippocampus and cortex of patients. 12Changes to the transcriptome in animal models [13][14][15][16][17][18] or in patient samples 19,20 have been examined, however, with limited agreement between studies. 1,6Many existing studies have demonstrated a limited correlation between transcript expression and protein abundance.Therefore, it is important to gain a better understanding of the early phase of epileptogenesis via multiple omics approaches.
Stimulus-dependent phospho-signaling occurs faster than transcriptomic and proteome changes, is dynamic, and can persist to produce a new phosphoproteome status, potentially perpetuating epileptogenic states.Surveying the phosphoproteome also enables identification of key protein kinases involved in signaling through modeling of observed changes to phosphorylation sites that conform to kinase motifs. 21

Gene ontology enrichment and general approach to data analysis
To shed light on how this vast level of phospho-signaling might be influencing cellular functions, we devised an approach to extract relevant information from gene ontology (GO) enrichment.Since SE is a very strong stimulation, we took an approach that deliberately reduced the significant pathways identified by (a) limiting the background to only proteins we could detect by mass spectrometry (Figure S2C) and (b) used the ranked list method within gProfiler, 24 which weights enrichment toward highly ranked genes.We separated up-and down-regulated phosphorylation, assigned the maximum positive and negative quantitative values for phospho-regulation to each protein, and ranked the proteins using both the quantitative value and significance of maximal change (Table S1).

Pilocarpine-induced phosphorylation of posttranscriptional regulatory proteins at 4 h
Here, we focus first on phosphorylation changes at 4 h, represented as a volcano plot in Figure 1C and as enriched GO terms in Figures 1D, 1E, S2D, and S2E.A group of biological process terms including ''mRNA processing,'' ''mRNA transport,'' and ''RNA export from nucleus'' and the WikiPathways term ''mRNA processing'' were associated with increased phosphorylation for Pilo4h (Figures 1D and S2E; Table S3).These Figure 1.Phosphoproteome changes at 4 h after SE (A) Status epilepticus was induced in mice by pilocarpine or were mock treatment.The hippocampus was dissected at 4 and 24 h.(B) The hippocampus was processed for proteomics.Tryptic peptides were labeled with TMT11plex reagents.Phosphopeptides were enriched and fractionated using the TiSH method prior to LC-MS/MS analysis.The phosphopeptide-depleted sample was also fractionated and analyzed by LC-MS/MS to determine changes to the proteome.The data were log 2 transformed and specific comparisons were applied in multiple analyses.(C) Volcano plot for the phosphopeptides at 4 h.Light gray dots are values below the threshold for significance.Dots representing the top 50 ranked phosphopeptides from proteins with the terms synapse organization or cell junction organization or related to posttranscriptional regulation are colored.Enriched (D) biological process and (E) molecular function terms.A scale bar for probability is shown.The size of the circle represents the number of genes enriched.(F) Phosphorylation sites of posttranscriptional regulatory proteins ranked in the top 50.The overlapping membership of gene ontology terms is shown.The color of circles represents the log 2 fold change.(G) Median fold change of selected postsynaptic density (PSD) proteins, transsynaptic proteins, and neurotransmitter receptors (scale bar shown).(H) Schematic representation of the same proteins (descriptively) colored according to the direction of their median phosphorylation change.Created with BioRender.com.See also Figures S1-S3, Tables S1 and S3.
terms describe sub-processes of the general process of posttranscriptional regulation and included RNA-binding proteins that target the mRNA of neurotransmitter receptors, encoded by Sfpq, Nova2, Celf4, Pum1, Pum2, Eif4enif1, and Cpeb4. 25Focusing on the top 50 phosphopeptides with enriched GO terms (Table S1), we found that 27 were from proteins with posttranscriptional regulation terms.While RNA splicing was not enriched at 4 h, many high magnitude phosphorylation increases were on spliceosome or RNA splicing proteins (Figure 1F).There were approximately 240 posttranscriptional regulatory proteins with significant phosphorylation changes.The top 50 ranked phosphopeptide signals involved in posttranscriptional regulation are highlighted in Figure 1C.

Pilocarpine increased phosphorylation of actin cytoskeleton-linked postsynaptic proteins at 4 h
Multiple biological process terms related to neuron projection or dendrite spine morphogenesis/organization were enriched, whether increased or decreased in phosphorylation (Figure 1D).This bi-directional phosphorylation (simultaneous phosphorylation and dephosphorylation of the same protein) was also observed for related molecular function terms (Figure 1E) and cellular component terms (Figure S2D).To cut through to the underlying meaning of the bi-directional phosphorylation, we determined the median significant log 2 fold change for individual proteins in the enriched neuronal terms (Table S3), as represented by heatmap in Figure 1G.This approach made clear the proteins that were weighted toward either an increase or decrease in phosphorylation.
A subset of highly ranked postsynaptic proteins with positive median log 2 fold changes included cortactin, kalirin, microtubule-associated protein 1A (MAP1A), MAP2, drebrin, homer 1/3 and SH3, synaptic Ras GTPase-activating protein (synGAP), and multiple ankyrin repeat domains protein 1-3 (SHANK1-3) (Figure 1G).While cortactin appeared approximately balanced in phosphorylation, closer inspection of the sites indicated that cortactin was converted to a more multi-phosphorylated state, that is, increasing in overall phosphorylation status (Table S1; Figure S3A).Increased phosphorylation of MAP1A at S1898 (Figure S3B) was ranked second of all proteins associated with enriched GO terms (Table S1).Sixty-eight MAP1A sites changed abundance at 4 h and 82% of these were increased (Figure S3B).
Postsynaptic density protein kalirin was one of at least 46 proteins with positive median log 2 fold changes (Figure 1G) that contributed to the Rho GTPase binding molecular function term enrichment (Figure 1E; Table S3).Many of these proteins promote actin binding, morphogenesis, and/or neurite outgrowth.Thus, pilocarpine-induced phosphorylation occurred on proteins with the potential to alter the cytoskeleton within synapses or sites of neurite outgrowth.

Proteins linked to neurotransmitter receptors and cell junctions were mainly dephosphorylated at 4 h
A group of biological process terms were significantly associated only with decreased phosphorylation (Figure 1D).The terms synapse organization and cell junction organization encompassed most of these proteins.The top 50 ranked phosphopeptide signals associated with these terms are highlighted in Figure 1C.Examining the higher ranked signals, we found that scaffolding proteins directly or indirectly linked to neurotransmitter receptors or cell junctions were prominent (Figure 1G).For example, disks large-associated protein 1 (DLGAP1) had a negative median log 2 fold change (Figure 1G).
Postsynaptic density protein 95 (PSD95), disks large family member PSD93, synapse-associated protein 97, and glutamate receptor-interacting protein 1 had a negative median log 2 fold change (Figure 1G).Most AMPA receptor subunits, N-methyl D-aspartate (NMDA) receptor subunits, and metabotropic glutamate receptors had a negative median log 2 fold change at 4 h after SE (Table S1; Figure 1G).Ultrastructural data suggest that the postsynaptic density (PSD) is composed of interconnected scaffold layers. 26In Figure 1H, we used the median log 2 fold change of PSD proteins to show that most proteins in the first two scaffold layers beneath the plasma membrane and glutamate receptors had a negative median log 2 fold change at 4 h, whereas the lower scaffold proteins and MAPs had a positive median log 2 fold change, indicating differing influence of protein kinases and phosphatases in the scaffold layers.

Known regulatory phosphorylation mechanisms supported increased translation and cell differentiation, but not growth at 4 h
Since the vast majority of phosphorylation sites have no known function, we cannot make generalizations about the effects of increased or decreased phosphorylation.However, we can use those sites with known functions to determine some trends.We identified the processes with evidence for being either induced or inhibited by phosphorylation (according to PhosphoSitePlus 27 nomenclature).Figures 2A and 2B show the log 2 fold change for individual phosphorylation sites at 4 h with known functions and the median log 2 fold change for each function.There was dephosphorylation of sites known to induce cell growth and phosphorylation of sites that inhibit growth, indicating overall inactivation of growth.Translation was induced by phosphorylation (Figure 2A).Greater than 40% of all detected sites that are known to induce translation were perturbed (Figure 2C) and none that inhibit translation were detected (Figure 2B).However, only a few sites regulating translation were present in our dataset (Figure S4) and they are rare in public databases. 28There were also few cell differentiation sites detected (Figure S4); however, half of those detected had increased phosphorylated (Figure 2C) and supported induced cell differentiation (Figure 2A).Many sites regulating transcription were detected (Figure S4) but both induction and inhibition were promoted by increased phosphorylation (Figures 2A and 2B).Cell motility and apoptosis also had mixed signals (Figures 2A and 2B).Overall, there were indications that cell growth was prevented while translation and cell differentiation were promoted.

Protein kinases and phosphatases activated at 4 h implicated MAPK pathways and neuronal plasticity
We used KinSwing, an algorithm we previously developed, 21 to predict protein kinases upstream of induced signaling.KinSwing uses a network-based algorithm that combines the relationship between quantitative phosphopeptide data and modeled substrate motifs.We may only infer net kinase activity since the contribution from protein phosphatases on phosphorylation site levels is unknown.At 4 h,  S4. components of the MAPK signaling pathway, MAPK1, MAPK3, P38A, P38B, c-Jun N-terminal kinase 1 (JNK1), JNK2, JNK3, and ribosomal S6 kinase 1 (RSK1) were predicted from the phosphopeptides to be significantly activated (Figure 2E; Table S4).In contrast, casein kinase 2 a1 (CK2a1) and CK2a2 had decreased inferred activity.Since the detected phosphopeptides included known substrates of protein kinases, 27 we compared the average significant log 2 fold change of these known substrates to KinSwing predictions (Figure 2F).The known substrates of MAPK1, MAPK3, P38A, and RSK1 were up-regulated and CK2a1 substrates were down-regulated, validating KinSwing predictions.Increased MTOR activity was also supported by known substrates in Figure 2F despite the activity increase predicted by KinSwing not being statistically significant at the predetermined probability threshold of 0.05 (Figure 2G, p = 0.059).
Direct phosphorylation of protein kinases may cause a change in activity.The known MAPK1 T183 + Y185 and MAPK3 T203 + Y205 activation sites were phosphorylated at 4 h (shown via volcano plot in Figure 2G), confirming the KinSwing prediction (Figure 2E).Multiple phosphosites for CAMK family members were detected (Table S1).The up-regulated phosphorylation of CAMK2a at likely inhibitory site, S314 (Figure 2G), and the autonomous activity site of CAMK2a, T286, was down-regulated at 4 h (Table S1).Even though the prediction for decreased activity of CAMK2a was approaching, but not significant (Figure 2F), validated CAMK2a substrates were, on average, significantly dephosphorylated (Figure 2G).Together, this demonstrated KinSwing predictions were performing well.
The phosphorylation status of several tyrosine protein kinases had relatively large log 2 fold changes at 4 h (Figure S5A).Both alpha and beta catalytic subunits of the important neuronal plasticity protein phosphatase 2B (PP2B)/calcineurin underwent significant phosphorylation changes at 4 h (Figure S5B).Significant changes were also detected for phosphatase regulatory proteins (Figure S5C) and tyrosine protein phosphatases (Figure S5D).Overall, there was evidence at 4 h for activation of MAPK pathway kinases as well as altered activity of kinases and phosphatases that regulate neuronal plasticity.

A new phosphoproteome status at 24 h post-SE
Only 36% of the phosphopeptides regulated at 4 h were also regulated at 24 h, indicating that there was a new phosphoproteome status at 24 h after SE.A volcano plot of the 3,472 significantly changing phosphopeptides for Pilo24h is shown in Figure 3A.GO enrichment was performed (Figures 3B-3E and S6A; Table S3) using our approach which limits enriched terms to only those with highly ranked phospho-signals.The WikiPathways ''TYROBP Causal Network'' term was associated with increased phosphorylation at 24 h (Figure 3E) and describes a microglia activation network, 29 including proteins encoded by the Spp1, Bin2, Ppp1r18, Samsn1, Fkbp15, and Apbb1ip genes (Figure 4F).There were 78 proteins with increased phosphorylation that resulted in the ''RNA splicing'' term enrichment (Figures 3A and 3B).Looking deeper at the highly ranked RNA splicing factors, we identified the Ser/Arg-rich protein family of splicing factors as a prominent sub-group (Figure 3G).Another group with increased phosphorylation at 24 h were numerous proteins with guanyl-nucleotide exchange factor activity (Figure 3H).

Inhibitory and excitatory neurotransmitter receptors were dephosphorylated at 24 h
Decreased phosphorylation at 24 h was associated with the molecular function term ''G protein-coupled neurotransmitter receptor activity'' and the Reactome terms ''glutamate binding, activation of AMPA receptors, and synaptic plasticity'' and ''trafficking of AMPA receptors'' (Figures 3C and 3D).These terms included multiple receptor and ion channel subunits (Figures 3A and 3I-3N).Glutamate receptor phosphorylation sites were dephosphorylated at 24 h, for example, the AMPA receptor subunit GluA1 (encoded by Gria1) dephosphorylation of S863 (Figure 3J).The transmembrane AMPA receptor regulatory proteins (TARPs) were included in the significantly enriched Reactome terms (Figure 3D) and were dephosphorylated at sites conserved in TARP g-2, 3, 4, and 8 (Figure 3K).NMDA receptors were not part of enriched Reactome terms but were highly ranked and occurred in most of the enriched cellular component terms (Figure S6A).The 2A subunit (GluN2A, encoded by Grin2a) had 13 dephosphorylated sites at 24 h.Muscarinic acetylcholine receptors are G protein-coupled receptors and followed the trend of decreased phosphorylation at 24 h (Figure 3L), for example, dephosphorylation of S384 of muscarinic acetylcholine receptor M3 (Figure 3L).Since there are many types of G protein-coupled receptors, we asked whether they were all predominantly dephosphorylated at 24 h in our data.We found that most receptor phosphorylation site changes involved decreased phosphorylation (Figure 3N).An example is the m-opioid receptor (Figure 3N).

Pilocarpine induced dephosphorylation of potassium channels at 24 h
Potassium channels were highly ranked in five of the cellular component terms associated with decreased phosphorylation (Figure S6A).The delayed rectifiers Kv2.1 and Kv2.2 (encoded by Kcnb1 and Kcnb2) had multiple down-regulated phosphorylation sites (Figure 3M) and two sites were decreased for epilepsy-associated Kv2.1 (encoded by Kcnb1) (Figure 3M).S1 and S3.iScience Article Functional phosphorylation sites support cell growth, cell motility, and inhibition of apoptosis at 24 h Using curated information of known functional phosphorylation sites, it became clear that cell motility was likely induced by phospho-regulation at 24 h after SE, without significant inhibition (Figures 4A-4D).Cell growth was induced, as was the associated process of carcinogenesis (Figure 4A), although both were simultaneously inhibited by a lesser number of functional phosphorylation sites (Figure 4B).Phospho-regulated inhibition of apoptosis was also clearly supported at 24 h (Figure 4B).Functional phosphorylation sites for apoptosis were well represented in the data (Figures 4C and 4D), although the overall level of perturbed functional phosphorylation sites was decreased at 24 h relative to those at 4 h (Figures 2C and 2D).

Protein kinase and phosphatase activity at 24 h implicates signaling involved with cell growth and chronic synaptic activity
At 24 h, multiple CDKs had decreased inferred activity, using KinSwing (Figure 4E; Table S4).CDK1 and CDK5 substrates had the most evidence for decreased phosphorylation using the average fold change for known substrates (Figure 4F).Mitogen and stress response kinase 1 (RSK1) and protein kinase B (AKT1) were predicted to be up-regulated (Figure 4E) and had known substrates with increased phosphorylation (Figure 4F), implicating signaling pathways associated with the stress response and cell growth.
Direct phosphorylation of protein kinases at 24 h did not implicate many phosphorylation sites with known function.However, we observed CAMK2a inhibitory S314 phosphorylation at 24 h (Figure 4G), which was diminished relative to the phosphorylation level at 4 h (Figure 2H), and phosphorylation of ephrin type-A receptor 4 phosphorylation at Y779 (Figure S6B).There were fewer changes detected for phospho-regulated phosphatases at 24 h (Figures S6C-S6E) than at 4 h.

Persistent signaling from 4 to 24 h after SE supports main findings
There were 1,248 phosphopeptides significantly changed at both 4 and 24 h post-SE (Figure 4H), which was too small to obtain sufficient enriched GO terms with the procedure used for the single time points.After relaxing the background to include the entire mouse proteome, we identified enriched terms (Figures 4I-4K; Table S3) consistent with our trends.The persistently phospho-regulated proteins were mainly those with guanyl-nucleotide exchange factor activity, involved in RNA processing or located at the postsynapse.

Diazepam induced phosphorylation of cytoskeletal and plasticity proteins while GABA receptor subunits were dephosphorylated
We used a group of mice that received no treatments or injections (''untreated'') as a control to investigate the specific effects of diazepam injection (Figure 5A), a first-line treatment for SE.We identified 503 and 541 significant phosphorylation-level changes, respectively, at 4 and 24 h post-mock injection that were induced by diazepam (Figures 5B and 5C; Table S1).There were 754 phosphopeptide changes significant at one or more time points.Forty-one changes were significant at both time points (Figure 5D).The term ''negative regulation of protein depolymerization'' was enriched and associated with increased phosphorylation at 4 h after diazepam treatment (Figures 5E and 5F).This term included the cytoskeletal proteins MAP1A, MAP1B, MAP2, b-adducin, and spectrin-a non-erythrocytic 1 (Table S3).Diazepam altered the phosphorylation of its target GABA receptor at 4 h, mainly with dephosphorylation, which contributed to enrichment of the ''postsynapse'' cellular component term (Figure 5D).However, most synaptic proteins had up-regulated phosphorylation (Figure 5F), including AMPA and NMDA receptor subunits, such as S863 on GluA1 (Figure 5D).Increased phosphorylation was also associated with synaptic terms at 24 h (Figure 5G).Neurofilament heavy chain had increased phosphorylation at multiple sites, which was significant at both 4 and 24 h after diazepam (Figure 5D).
When we used KinSwing to investigate inferred protein kinase activity, we found that CK1a1 had the greatest inferred activity at 4 h (Figure 5H).CDK5 increased inferred activity at 24 h after diazepam (Figure 5I).Further, a known substrate of CDK5, b-2-syntrophin S75, 30 increased phosphorylation at 24 h after diazepam, again supporting the prediction that CDK5 is affected by diazepam treatment.

Proteome changes at 4 h after SE indicated MAPK pathway stimulated immediate-early gene expression
The proteomic screen detected 6,277 protein groups (Table S2).Protein groups are made up of one or more proteins that have evidence for existence from the detected peptides.The 6,277 protein groups mapped to 6,051 non-redundant gene names.When phosphopeptide-based .Functional trends and protein kinase prediction for the phosphoproteome analysis at 4 h and persistently regulated phosphorylation Log 2 fold changes are shown for phosphorylation sites with known ability to (A) induce or (B) inhibit various functions at 24 h.The median value is shown for each function as a line.The functional phosphorylation sites detected as perturbed in the phosphoproteome data as a percentage of all detected phosphorylation sites is shown for (C) induced functions and (D) inhibited functions.(E) Volcano plot of KinSwing predicted activity of protein kinases induced by pilocarpine at 24 h.The size of each dot is scaled to the number of known substrates used to build the protein kinase consensus motif.(F) The average substrate log 2 fold change for known substrates of protein kinases perturbed by pilocarpine at 24 h versus the KinSwing predicted protein kinase activity.(G) Volcano plot of the log 2 fold changes for phosphopeptides occurring on Ser/Thr protein kinases at 24 h.(H) Heatmap of phosphopeptides significantly and persistently perturbed at both 4 and 24 h.Enriched (I) molecular function, (J) biological process, and (K) cellular component terms for persistently perturbed phosphoproteins.A scale bar for probability is shown.The size of the circle represents the number of genes enriched.See also Figure S6, Tables S3 and S4. protein identifications were allowed in the count (requiring at least two phosphopeptides), the number of non-redundant genes increased to 7,553 (Figure S2C).After discarding protein mapping to the same gene, there were 5,765 and 5,550 proteins in the 4 and 24 h TMT11plexes, respectively, which intersected on 5,264 proteins (Figure 6A; Table S2).1,328 proteins were significantly perturbed at one or more time points after SE (Figure 6B).
Sixty-one proteins had significant level changes 4 h post-pilocarpine treatment and most of these had decreased abundance (Figure 5C).The regulated proteins were analyzed for GO enrichment after separation into sets of up-and down-regulated proteins.Only the WikiPathways term ''p38 MAPK signaling pathway'' was significantly enriched at 4 h (Table S5).The levels of two immediate-early genes, Jun and Arc  S2 and S5.
(Figure 5C), were significantly up-regulated at 4 h with Jun being the most significantly up-regulated protein, which is a hallmark of early SE (Table S6).Further proteins linked to neuronal activity also had increased abundance, protocadherin 8 (encoded by Pcdh8) and neuronal pentraxin 2 (encoded by Nptx2) (Figure 6C).Jun, Arc, protocadherin 8, and neuronal pentraxin 2 were also among 10 persistently up-regulated proteins (Figure 6E).There were more persistently down-regulated proteins at 4 h, but our analyses did not associate any specific GO terms.

The proteosome, ribosome proteins, and translation factors had increased expression at 24 h
Approximately 25% of the detected proteome had significantly changed abundance 24 h after pilocarpine injection (Figures 6B and 6D).Enriched GO terms indicated that the main proteins with increased abundance at 24 h were involved in translation (Figures 6F-6H and S7A-S7C; Table S5).More than 70 ribosomal proteins increased abundance, covering more than half of the small and large ribosomal subunits.At least 20 eukaryotic initiation factor subunits increased abundance (Table S5).
Associated with increased abundance were ''protein folding'' and ''unfolded protein binding'' terms (Figures 6F and 6G), which further indicated an increased use of protein synthesis mechanisms.In contrast, there was increased abundance of 20 components of the proteosome complex (Table S5) resulting in enrichment of the ''Proteasome'' Kyoto Encyclopedia of Genes and Genomes term (Figure 6H) and ''proteosome degradation'' WikiPathways term (Figure S7C).

Pilocarpine induced a decrease in abundance of synaptic proteins at 24 h
Synaptic-related terms were associated with reduced protein abundance at 24 h (Figures 6F and 6G), including glutamate receptor subunits GluN2B, GluA4, and metabotropic glutamate receptor 5. Postsynaptic density proteins PSD93, PSD95, DLG1/SAP97, DLGAP1/2/4, SH3, SHANKs 1-3, and homer 2 were decreased.Also reduced were presynaptic proteins, transsynaptic proteins, and cytoskeletal proteins.Overall, decreased postsynaptic density protein levels were mainly observed for the upper and middle layers of the PSD scaffold (Figure 6I).Proteins specific to GABAergic neurons, including gephyrin, neuroligin 2, five GABA receptor subunits, glutamate decarboxylase 1 and 2, and the potassium-chloride cotransporter were also decreased at 24 h.

Diazepam induced a decrease in mitochondrial respiratory chain proteins
Diazepam injection had no detectable effect on the proteome at 4 h compared to untreated controls.Approximately 3% of the detected proteome was affected 24 h after diazepam injection (Figures S8A and S8B).The main class of proteins producing significantly enriched terms were mitochondrial respiratory chain proteins (Figures S8C-S8G; Table S5), which were decreased.

Independent experimental validation confirms the main findings are reproducible
We produced a new set of experimental samples using the same pilocarpine mouse model and analyzed those samples using the targeted MS method of parallel reaction monitoring (PRM). 31Over 80 phosphorylation sites (Figure 7A) and 32 proteins (Figure 7B) were targeted by PRM (Table S7).These targets represented different classes of proteins found to be enriched in the GO analysis derived from our primary TMT data.These PRM data supported our TMT data.For example, posttranscriptional regulators had mainly increased phosphorylation and protein kinases such as MAPK1/3 increased in phosphorylation at activating sites.Furthermore, PSD proteins were segregated into groups that had increased and decreased phosphorylation.The GluA1 S863 phosphorylation site was decreased in both TMT and PRM experiments at 4 h.Validation of the dephosphorylation by immunolabeling of hippocampal slices also indicated a reduction in the CA1 region at 4 h, (Figures 7C and 7D, p = 0.032).Notably, some phosphorylation sites did not reach significance by PRM and a minority showed changes in phosphorylation that were opposite to the TMT measurement.However, the overall finding was that both the 4 and 24 h phosphorylation site level measurements by PRM were highly correlated to the TMT data (Figures 7E and 7F).
Protein changes were also supported by PRM.PSD proteins that had mostly decreased abundance at 24 h by TMT were also decreased in abundance by PRM (Figure 7B).Astrocytic proteins, glial fibrillary acidic protein (encoded by Gfap), and clusterin (encoded by Clu) had increased abundance.Two presynaptic protein targets (encoded by Syt2 and Slc17a6) gave the opposite result (Figure 7B).The ribosomal proteins, that had increased abundance using TMT, were not significantly increased via PRM.In general, PRM measurements were greater in magnitude (Figures 7E-7H, y axis) than the TMT measurements but less significant (Figures 7A and 7B).This is likely due to the different methods and the different scale of data, leading to different processing and statistical treatment.The log 2 fold changes for protein at 4 h were expected to be small and they were (Figure 7G).The TMT and PRM protein data at 24 h were highly correlated (Figure 7H).S7.

DISCUSSION
SE triggers a large number of changes in the structural organization and cellular composition of the hippocampus as well as in the electrophysiological properties of the neuronal network which ultimately result in the manifestation of spontaneous self-recurring chronic seizures.Since neuronal activity impacts the cellular properties of brain cells, in particular of neurons, the pathological strong neuronal activity induced by SE is likely to trigger cellular signaling cascades that underlie epileptogenesis.However, these signaling cascades have not been detailed at the protein level, until this work.Here we find that early phospho-signaling is directed toward controlling posttranscriptional regulation and postsynaptic functions.Also, the earliest protein changes are supportive of translation but result in decreased postsynaptic proteins.Our initial analysis using TMT was validated by a targeted analysis of representative proteins and phosphorylation sites using PRM on independent samples.Although not always reaching statistical significance, the PRM data were highly correlated to the TMT data, supporting the reproducibility of our findings.

Reduced phosphorylation and protein expression of proteins involved in neuronal plasticity after SE
It is well established that neuronal networks aim to maintain their firing rate within a physiological target range by homeostatic regulation of synaptic strength. 32Neurons experience strong pathophysiological activity during SE.The limited electrophysiological studies in rats 33,34 and mice 35 suggest that long-term potentiation (LTP) is impaired as an early compensatory mechanism opposing strong activity.Our data strongly support a homeostatic synaptic response.
The interplay of CAMK2a and PP2B controls the phosphorylation status of postsynaptic proteins and neurotransmitter receptors, thereby regulating synaptic strength during plasticity. 36Our results show that SE results in changes in the activity of these two proteins.Changes in the phosphorylation status of the CAMK2a autophosphorylation sites indicated that the kinase was deactivated at 4 and 24 h after SE and, fittingly, most predicted and known CAMK2a substrates were dephosphorylated at 4 h after SE.However, there were also examples indicative of prior hyperphosphorylation by CAMK2a, such as the increased phosphorylation of homers (homer1 S117 and homer3 S159) which are known to cause the dissolution of homer's interaction with metabotropic glutamate receptor 5 and circuit hyperexcitability. 37In accordance with an overall decrease of CAMK2a activity, our data suggest that calcium-dependent phosphatases are active at 4 h after SE.Both low and high frequency action potentials activate PP2B, which has a higher affinity for calmodulin than CAMK2. 38Supportive of this, we found that phosphorylation sites on PP2B which are known to inhibit PP2B activity were dephosphorylated in our model. 39,40Furthermore, we observed dephosphorylation of known PP2B substrates.Dephosphorylation of S863 on GluA1, which was validated by PRM and immunolabeling, is known to be mediated by PP2B and to cause receptor internalization and long-term depression (LTD). 41We observed TARP dephosphorylation, which is downstream of PP2B and is associated with AMPA receptor trafficking and LTD. 42Presynaptically, synaptic strength can be modulated by the interplay of CDK5 and PP2B. 43Postsynaptic CDK5 function is less understood, but has some important substrates.For example, neurotrophic receptor tyrosine kinase 2 was dephosphorylated at a CDK5 substrate site which is required for activity-dependent structural plasticity associated with LTP. 44There was evidence of synaptic downscaling at 24 h, which included dephosphorylation of EEF2K substrates, 45 dephosphorylation of DLGAP1 S499, 46,47 phosphorylation of synGAP S842, 48 and phosphorylation of ephrin type A receptor 4 Y779. 49The RNA splicing protein Nova2 was phosphorylated at T27 at 4 h after SE and is involved in the regulation of LTP and homeostatic plasticity via alternative splicing of synaptic proteins. 50,51Taken together, our data suggest that an early response to SE is the activation of predominantly phosphatase-mediated regulation of postsynaptic receptor complexes, and other proteins broadly involved in neuronal plasticity, to alter synaptic strength, which suggests an early neuroprotective homeostatic response.Subsequent reduced abundance of the postsynaptic receptor complex is suggestive of weakened synaptic strength, synapse loss, 52 or neuronal cell death. 53

SE triggers changes in the phosphorylation status of proteins associated with synaptic and neuronal reorganization
Pilocarpine-induced SE has been shown to result in dendritic spine loss. 54Changes in the phosphorylation status of PSD scaffolding proteins and cytoskeletal proteins present in dendritic spines within our data might contribute to dendritic structural plasticity.For example, all SHANKs had increased phosphorylation at 4 h after SE.SHANK3 activity has been shown to regulate spine number and size. 55Also, the phosphorylation status of multiple proteins associated with the cytoskeleton, for example MAP2 and drebrin, was mainly increased early after SE.These postsynaptic scaffold components exist in a layer furthest from the membrane, which is a route of access to the PSD for CAMK2 26 and it is striking that they have similar increased phosphorylation.
In addition, cell junction proteins such as paralemmin-1 and hevin were heavily dephosphorylated; however, the functions of the dephosphorylated sites are unknown.Hevin is implicated in epilepsy since it accumulates at excitatory synapses at 24 h in a rat pilocarpine seizure model 56 and phospho-regulation of paralemmin may influence its role in membrane remodeling and synapse formation. 57Dysregulated cell junction proteins such as paralemmin-1 and hevin could contribute to synapse destabilization.

Evidence for MAPK pathway signaling potentially leading to posttranscriptional regulation and cell growth
We found evidence of MAPK pathway signaling, including the extracellular signal-regulated kinase, P38, RSK, and JNK families, at 4 h after SE.MAPK pathways are known to target transcription activators in response to stress, growth factors, and neuronal activity. 58MAPKs mainly activate transcription, but also the MTOR pathway, which is a known activator of translation. 59There is growing evidence that MAPKs and MTOR are implicated in epilepsy by phosphorylating and regulating RNA-binding proteins, 25,59 for which our unbiased screen gives further support.][61] Strikingly, posttranscriptional regulatory proteins were the major targets of phosphorylation, at both 4 and 24 h after SE.The brain uses alternative splicing more than any other organ and has an expanded set of posttranscriptional regulatory mechanisms to match. 25,62However, there have been few mechanistic studies of posttranscriptional regulatory proteins. 28Our study revealed that posttranscriptional regulatory proteins that span the life cycle of an RNA molecule, including splicing, export, transport, translation, and degradation, 25 were major substrates of early signaling.
The RNA-binding proteins encoded by Celf4 and Sfpq were prominently phospho-regulated.Sfpq is important for axon survival and Celf4 regulates neurotransmitter receptor abundance. 25Since RNA-binding proteins have cell-type-specific functions that control neurotransmitter receptor abundance, their perturbation is likely to affect excitatory/inhibitory balance.The RNA-binding protein, pumilio 2, was phosphoregulated in our data and has been implicated in epilepsy, 63 but so far there are few such examples of RNA-binding proteins related to epilepsy.It is not known if phosphorylation of a broad group of posttranscriptional regulatory proteins leads to mainly increased or decreased translation and which cell types are most affected.The major factors will need to be investigated further to reveal their role in epileptogenesis.
At 24 h, ribosomal proteins, translation initiation factors, and proteins regulating folding had increased expression.Ribosome proteins can be locally synthesized and selectively bind to 3 0 -untranslated regions creating specialized ribosomes for regulating the specificity and efficiency of local protein synthesis. 25The detection of approximately 20 proteosome components with increased abundance at 24 h post-SE suggests that protein degradation was highly activated at the same time that protein synthesis was promoted.The proteosome is required for both apoptosis, to remove unwanted protein, and as a protein quality control mechanism in cell growth. 64It is notable that dysregulated proteostasis is a common feature of neurological diseases.For example, in models of hyperexcitability in fragile X syndrome. 65The signals for both ribosome and proteasome protein increases were relatively weak and will require further investigation.
There were multiple additional indicators of cell growth.Many Rho GTPase proteins had up-regulated phosphorylation at 4 h and guanylnucleotide exchange factors had persistently increased phosphorylation at both 4 and 24 h.Many of these proteins are known to mediate cytoskeletal functions which lead to morphogenesis or cell growth. 66Also, there was increased phosphorylation of sites known to promote cell migration.The mechanisms for cell growth and migration are highly relevant to understanding epileptogenesis since growth or sprouting that might be necessary for repair may also lead to more excitatory synapses, causing positive feedback and more frequent seizures. 4,5Protein kinases associated with growth signaling, such as MTOR and AKT1, had positive indicators for increased activity.MTOR is also known to contribute to an immune response.There was only minor evidence of microglia activation, beginning with a few relevant protein changes at 24 h.Thus, signals indicative of inflammation were present but relatively weak compared to indicators of posttranscriptional regulation and growth.

Diazepam induced synaptic plasticity mechanisms and decreased expression of mitochondrial respiratory chain proteins
In the pilocarpine animal model of epileptogenesis, diazepam is used to allow mice to recover from SE due to activation of GABA receptors that inhibit excitatory synapse depolarization.Humans routinely use diazepam or homologous drugs to induce sleep.However, to our knowledge, there is limited understanding about the mechanistic effect of diazepam in the 24 h after administration.Surprisingly, we found that diazepam application on its own caused significant alterations in the hippocampal phosphoproteome.For example, proteins involved in GABAergic signaling were phospho-regulated, GluA1 was phosphorylated at LTP-associated S863 41,67 and there were phosphorylation changes on methyl-CpG-binding protein 2, a protein which regulates synaptic scaling. 68The direction of phosphorylation change on GluA1, in particular, supports previous findings at the functional level that diazepam can induce synaptic upscaling functionally similar to that induced by tetrodotoxin. 69,70Reduced abundance of proteins involved in the mitochondrial respiratory chain was the major protein change following diazepam at 24 h.Since diazepam suppresses brain activity, likely, there was reduced energy metabolism.Our data call for future studies investigating whether there are even longer term effects (>24 h) following a single diazepam application and whether they might be related to the development of addictions.

Limitations of the study
We acknowledge that a major limitation of this work is that measurements from whole hippocampus cannot be used to resolve changes that are specific for certain cell types or subcellular compartments and proteins can have a different response to stimuli in different locations.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Mark Graham (mgraham@cmri.org.au).

Materials availability
This study did not generate new unique reagents.

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS Animals
All animal procedures were planned and performed to minimize pain and suffering and to reduce the number of used animals in accordance with the guidelines of the University Hospital Bonn, Animal-Care-Committee as well as the guidelines approved by the European Directive (2010/63/EU) on the protection of animals used for experimental purposes and ARRIVE guidelines.All mice were housed in a humidity (55 G 10 %) and temperature (22 G 2 C) controlled environment under a 12-h light-dark-cycle (light cycle 7 am to 7 pm) with water and food ad libitum and nesting material (Nestlets, Ancare, USA).All mice were male C57Bl6/N strain (Charles River; 60 days old, weight R 20 g) and were allowed to adapt to the animal facility at least for seven days before any treatment.

METHOD DETAILS Induction of chronic epilepsy
The mice were injected subcutaneously with 335 mg/kg pilocarpine hydrochloride (Sigma) to induce status epilepticus, 20 minutes after pre-treatment with subcutaneous injection of 1 mg/kg scopolamine methyl nitrate (Sigma). 22,23Animals received an injection of diazepam (4 mg/kg, s.c.; Ratiopharm) forty minutes after SE onset.Control animals were treated identically but received saline instead of pilocarpine (mock treated).Behavioural SE was clearly identified using a modified seizure scheme: sustained convulsions with postural loss designated as SE. 22,23Among pilocarpine injected animals, only those developed SE (SE-experienced) were further used for analysis.

Sample processing for proteomics analysis
The freeze-dried samples were resuspended in a 100 mL solution of 10 mM tris(2-carboxyethyl) phosphine, reduced for 10 minutes at 85 C and subsequently alkylated in 25 mM iodoacetamide at 22 C in the dark.Proteins were precipitated from the samples using chloroform-methanol extraction 76 and air dried.The precipitates were dissolved in 20 mL solution containing 7.8 M urea, 50 mM triethylammonium bicarbonate and 5 mg of Lys-C (FUJIFILM Wako Pure Chemical Corporation) for an 8 h digestion at 25 C.The samples were diluted with a solution containing 170 mL 100 mM 4-[2-hydroxyethyl]-1-piperazineethanesulfonic acid (HEPES) (pH 8) and 5 mg TrypZean trypsin (Sigma Life Sciences) for subsequent digestion at 37 C for 4 h.The trypsin digestion was twice repeated with an additional 5 mg for 4 h at 37 C.The peptide concentration was estimated by absorption at 280 nm (Implen Nanophotometer, Labgear).

TMT labelling and phosphopeptide enrichment
Approximately 400 mg of peptide was aliquoted into new tubes of equal concentration, using 100 mM HEPES (pH 8) to adjust the volume.TMT11plex reagent sets (Thermo Fisher Scientific) were dissolved in anhydrous acetonitrile and applied to each of the twenty-two samples in a 1 h and 20 min incubation at 25 C, using one reagent vial per 200 mg.To ensure completeness of labelling 0.6 mL from each sample was tested and were found to be > 98% labelled.Excess TMT reagent was quenched with 16 mL of a 5% hydroxylamine solution for 15 minutes at 25 C.The 4 h and 24 h treatment samples were pooled separately according to their respective timepoints.The six control untreated replicates were then pooled together and thoroughly mixed before splitting the combined sample in half, to mix with the two timepoint treatments.The 4 h TMT sample set consisted of: control untreated bio-replicate ).The two pooled samples were acidified with 100% formic acid to pH < 3 and the volume was reduced to 150 mL under vacuum centrifugation.The two samples were diluted to 2 mL with 0.1% trifluoracetic acid and desalted using solid phase extraction (Sep-Pak tC18 3cc Vac cartridge, Waters).
To obtain non-, mono-and multi-phosphorylated peptides from each timepoint, the samples were applied to the ''TiSH'' method of phosphopeptide enrichment. 77All recovered multi-and mono-phosphorylated enriched peptides were desalted, with multi-phosphorylated enriched samples ready for liquid chromatography tandem mass spectrometry (LC-MS/MS) analysis.The enriched mono-phosphorylated peptides were subjected to hydrophilic interaction chromatography (HILIC) for further fractionation prior to LC-MS/MS analysis.The mainly non-phosphorylated peptides which do not bind to titanium dioxide were collected from the first TiSH wash step.Approximately 3.5% of the 4 h and 24 h mainly non-phosphorylated peptides were desalted and fractionated by HILIC prior to LC-MS/MS analysis.
HILIC sample fractionation was completed using a Dionex UltiMate 3000 HPLC system (Thermo Fisher Scientific) controlled by Chromeleon 6.8 software with a TSKgel Amide-80 1 mm inside diameter x 250 mm long column (Tosoh Bioscience).Samples were injected in 150 mL in 90% acetonitrile, 9.9% water, 0.1% trifluoracetic acid (HILIC Buffer A) at a flow rate of 60 mL/min with buffer A at 100% for 10 min.The gradient was from 100% HILIC buffer A 100% to 40 % HILIC buffer B (99.9% water and 0.1% trifluoracetic acid) for 35 minutes at a flow rate of 50 mL/min, then to 80% buffer B for 3 minutes then back to 100% buffer A for 12 minutes.Sample fractions were collected in 60 s intervals into a 96 well plate using a Probot (LC Packings) controlled by mCarrier 2.0.Selected fractions were pooled based on the intensity of the ultraviolet absorbance at 214 nm resulting in seventeen phosphopeptide enriched fractions and twelve mainly non-phosphopeptide fractions.The mainly non-phosphorylated peptide HILIC fractionation was repeated and collected in 30 s fractions to make an additional set of twenty-seven fractions.Both 30 s and 60 s mainly non-phosphorylated peptide fractions were analysed.The fractions were dried under vacuum centrifugation and dissolved in 5.5 mL 0.1% formic acid for LC-MS/MS analysis.

TMT mass spectrometry
LC-MS/MS for all samples was completed with a Dionex UltiMate 3000 RSLC nano system and Q Exactive Plus hybrid quadrupole-orbitrap mass spectrometer (Thermo Fisher Scientific).The enriched multi-phosphorylated samples and HILIC fractions of the enriched non-and mono-phosphorylated samples were loaded directly onto a 300 3 0.075 mm column packed with ReproSil Pur C18 AQ 1.9 mm resin (Dr Maisch, Germany).A column oven (PRSO-V1, Sonation lab solutions, Germany) integrated with the nano flex ion source (Thermo Fisher Scientific) maintained the column at 50 C with an electrospray operating at 2.3 kV.The S lens radio frequency level was 50 and capillary temperature was 275 C. For MS analysis, 5 mL from all samples was injected.The reversed phase buffers were: A (0.1% formic acid in water) and B (0.1% formic acid, 90% acetonitrile, 9.9% water).There were two gradient variations used for MS acquisition.All enriched non-and mono phosphorylated fractions were subjected to a 110 min gradient beginning with sample loading at 99% buffer A and 1% buffer B separation for 25 min at 300 mL/min.The flow rate was then reduced to 250 mL/min in 30 s and the gradient was to 95% buffer A in 1 min, to 75% buffer A in 59 min, to 65% buffer A in 8 min, to 1% buffer A in 3 min and held at 1% buffer A for 2 mins before returning to 99% buffer A in 1 min and held for 10 min.For the multi phosphorylated samples the gradient was from 99% A to 95% buffer A in 6 min, to 72% buffer A in 116 min, to 65% buffer A in 10 min, to 1% buffer A in 3 min and held at 1% buffer A for 3 mins before returning to 99% buffer A in 2 min and held for 10 min.MS acquisition was performed throughout the 110 min and 180 min analysis.
Data-dependant acquisition LC-MS/MS was performed on all samples.In both the 110 min and 180 min methods, MS scans were at a resolution of 70,000 with an automatic gain control target of 1,000,000 for a maximum ion time of 100 ms from m/z 375 to 1500.Both the 110 min and 180 min methods had MS/MS scan resolution at 35,000 with an automatic gain control target of 200,000.The maximum ion time was 110 ms for the 180 min method and 115 ms for the 110 min method.The loop count was 12 for the 110 min method and 13 for the 180 min method.For both methods the isolation window was 1.2 m/z, the first mass was fixed at m/z 120 and the normalized collision energy was 34.Singly charged ions and those with charge >8 were excluded from MS/MS and dynamic exclusion was for 25 s in the 110 min method and 30 s in the 180 min method.
The raw LC-MS/MS data was processed with MaxQuant 1.6.7.0 71 using the following settings.Variable modifications were oxidation of Met, acetylation of protein N-terminus, deamidation of Asn/Gln and phosphorylation of Ser/Thr/Tyr.Carbamidomethyl of Cys was a fixed modification.Digestion was set to Trypsin/P with a maximum of 3 missed cleavages.The inbuilt contaminants file was used as well as the Mus musculus reference proteome with canonical and isoform sequences downloaded Feb 21 2020.Also, a fast file containing the sequence of immunoglobulin G-binding protein A (UniProt accession: A0A0H3K686) was included.TMT-labelled protein A was added to selected samples to assess TMT ratio compression.The accession exists in the MaxQuant output but was removed in the post-MaxQuant analysis of the data.Minimum peptide length was 6 and maximum peptide mass was 6000 Da.Second peptides and dependent peptides searches were enabled.Peptide spectrum matching and protein false discovery rates were set to 0.01.Minimum score for modified peptides was 40.Modified peptides and counterpart non-modified peptides were not used in the calculation of protein level changes.Modified peptides or their non-modified counterparts were not used in the protein quantification.All other parameters were default.One razor plus unique peptide for proteins < 150 kDa and two razor plus unique peptides was required for proteins > 150 kDa for protein detection.The raw mass spectrometry files and MaxQuant results are available from the PRIDE repository and have the dataset identifier PXD038241.

Phosphopeptide enrichment for PRM samples
New pilocarpine and mock treated samples (four each) were produced and prepared for proteomics, as for the TMT analysis.Approximately 1 mg of peptide for each sample was desalted by solid phase extraction (Sep-Pak tC18 3cc Vac cartridge, Waters), dried and resuspended in water.10% of the sample was reserved for PRM to determine protein levels.The remainder was used to enrich for phosphopeptides in each sample using titanium dioxide according to the first enrichment step of the TiSH method. 77Each sample was desalted and one third of the total was applied to PRM mass spectrometry.

PRM mass spectrometry
The samples for PRM analysis used the same LC-MS/MS equipment, column and reversed phase buffer system as for TMT.The enriched phosphopeptide samples, injected in 5 mL, were loaded in 99% buffer A for 25 min at 300 nL/min.The gradient, at 250 nL/min, was to 5% buffer B in 1 min, then to 25% buffer B in 97 min, to 35% buffer B in 3 min, to 99% buffer B in 1 min, held at 99% buffer B for 2 min, then to 99% buffer A in 1 min and held at 99% buffer A for 10 min as the flow rate increased to 275 nL/min.The peptide samples with no enrichment, injected in 3.5 mL, were loaded in 99% buffer A for 17.5 min at 300 nL/min.The gradient, at 250 nL/min, was to 6% buffer B in 1 min, then to 33% buffer B in 101.5 min, to 40% buffer B in 8 min, to 99% buffer B in 1 min, held at 99% buffer B for 2 min, then to 99% buffer A in 1 min and held at 99% buffer A for 10 min as the flow rate increased to 275 nL/min.Data-dependent acquisition runs were performed to determine phosphopeptide and non-phosphopeptide retention times.For the phosphopeptide PRM runs, the MS settings were as follows.For 140 min, full MS scans were performed at a resolution of 70,000 with automatic gain control target of 1,000,000, maximum scan time of 100 ms for the range 350 to 1500 m/z.The PRM scans were performed at a resolution of 35,000 with automatic gain control target of 3,000,000, maximum scan time of 175 ms, isolation window 0.9 m/z, normalised collision energy of 29 and loop count of 10.For non-phosphopeptides the MS settings were the same except the run was for 142 min, PRM resolution was 70,000, maximum ion time was 240 ms, isolation window was 0.7 m/z and normalised collision energy was 27.There were sixty-five phosphopeptide m/z targets, including four additional targets that were controls (not expected to change level according to TMT analysis).Due to multiple phosphorylation sites within each phosphopeptide, eluting at different retention times, the phosphorylation sites detected and quantified was greater than sixty-five.For non-phosphopeptides, two separate runs were performed.One run targeted seventy-four peptides belonging to eighteen proteins identified by TMT to have increased abundance and the other run targeted forty-nine peptides belonging to fourteen proteins identified by TMT to have decreased abundance.Both runs also targeted fourteen control peptides (not expected to change according to TMT results).
The targeted peptides were required to be unique in the mouse proteome at the gene level.The non-phosphopeptides were required to have no missed cleavages (trypsin specificity).Methionine was an allowed residue in only two of the one hundred and twenty-three target nonphosphopeptides.Target phosphopeptides had no restrictions on missed cleavages or amino acid residue content.Met-oxidised peptides were not targeted.Some Asn/Gln-deamidated phosphopeptides were targeted.One protein, collagen alpha-1(I), was represented by a single peptide.The remining proteins were represented by two to five peptides (average of 3.8 peptides/protein).The m/z target lists are available in Table S8.

Processing of PRM mass spectrometry data
The PRM runs were searched using Mascot 2.7.0 using the following settings for non-phosphopeptides.Enzyme specificity was trypsin and one missed cleavage was allowed.Carbamidomethylation of cysteine was a fixed modification.Peptide mass tolerance was 10 ppm.Fragment mass tolerance was 0.05 Da.The mouse proteome (download from UniProtKB Jan 2023 with canonical sequences) was searched.Phosphopeptides were searched in the same way except that two missed cleavages were allowed.Variable modifications were deamidation (Asn/Gln) and phosphorylation (Ser/Thr/Tyr).Additionally, the phosphopeptide PRM data was searched using MaxQuant 1.6.7.0 using the same settings as for the TMT data, with the following exceptions.No TMT labels were applied.The mouse proteome was downloaded from UniProtKB Oct 2023 with canonical and isoform sequences.
Quantitative data was obtained using Skyline 23.1.0.268.Each peptide was required to be detected with five transitions not including y 1 -y 2 or b 1 -b 2 ions, unless the peptide was shorter than 8 amino acid residues, in which case four transitions were sufficient and either y 2 or b 2 was allowed.The y 3 and b 3 ions were only allowed on peptides shorter than ten residues.Transitions had to be detected with less than 2.5 ppm accuracy.The phosphorylation site in peptides had to be localised by the Mascot or MaxQuant searches or manually by specific ions that rule out alternative sites.If no such ions were available, the site localisation was noted as ambiguous.The raw mass spectrometry files, search results and quantitative analysis files for the PRM analysis are available from the PRIDE repository and have the dataset identifier PXD047870.

Additional TMT MS data processing
The MaxQuant TMT data was further processed using methods and a 'best annotation' process described previously 21 with a modification to the phosphopeptides retained for further analysis that are multi-phosphorylated.Briefly, all phosphorylation sites identified are first re-annotated and amino acid sequences obtained from the fasta file used for MaxQuant searching.Where multiple phosphorylation sites were detected on unique peptides or charge states, the median of the log 2 transformed intensities was retained.This result was then considered a unique phosphopeptide for counting purposes.A difference to the previously described method is that all multi-phosphorylated peptides with a class 1 probability (i.e., a primary phosphorylated site) above 0.5 were retained if there was additional evidence of another class 1 site with a probability above 0.75 in at least one site from any experimental group.This allows retention of multi-site phosphorylated peptides and their intensities where evidence for the site combination existed amongst all experiments.Consistent with our previously described methods, all replicates were then globally normalised using quantile normalisation 78 followed by imputation of missing data (each sample group was permitted 25% of missing data) using the knn nearest neighbour averaging method 79 and corrected for unwanted technical sources of variation by application of surrogate variable analysis. 80After an initial analysis, the ''MockInj_24_2'' sample was removed as a significant outlier, assessed using the first 2 principle components of the normalised and corrected phosphopeptide intensities (mahalanobis distance p-value = 0.0001). 81Data were then remapped as described above (without the ''MockInj_24_2'' sample) before further analysis.For differential phosphorylation and protein analysis, data were fit with a generalized linear model with Bayes shrinkage implemented in limma version 3.36.5 73 comparing the pilocarpine and mock injections for each matched timepoint and for each mock injection versus the pooled controls at 0 minutes.

Additional PRM MS data processing
For both protein and phosphorylation PRM results, intensities were log2 transformed, globally normalised so that each sample had the same median value and fit with a generalized linear model with Bayes shrinkage implemented in limma version 3.36.5 73 comparing the pilocarpine and mock injections for each matched timepoint.

KinSwing
Protein kinase activity was predicted with KinSwingR version 1.1.4(described here: 21 and implemented/available on R/BioConductor with additional implementation of significance scoring of kinase activity: http://bioconductor.org/packages/KinSwingR/). Briefly, KinSwing predicts kinase activity by first modelling kinase matches (position weight matrices of known mammalian kinase:substrate sequences) to each phosphorylated peptide.Kinase activity is then inferred using a network theoretic approach, where nodes represent kinases and edges represent modelled kinase:substrate interactions.Local network connectivity for each kinase is determined as the absolute proportion of positive and negative peptides that have significant fold changes (P % 0.05), weighted by number of edges and number of substrate sequences using to build the kinase model.Finally, the weighted kinase score is z-transformed to enable comparison between different conditions.
The KinSwingR predictions of Ser/Thr activity were compared to the changes in phosphorylation status of known protein kinase substrates in the data by first downloading the known substrates from PhosphoSitePlus 27 (downloaded November 2 2020).For each protein kinase identified by KinSwingR, the average significant phosphopeptide intensity for all known substrates in our data was calculated.A minimum of four substrates was required.Kinase-substrate relationships from all mammals were allowed if the gene name and the sequence within four residues either side of the phosphorylation site was identical.All duplicate combinations of genes and sequence values from PhosphoSitePlus using these criteria were discarded.Multi-site phosphorylated peptides were not simplified or discarded but kept throughout all analyses.The first sequence window for the first phosphorylation site was assigned the quantitative value for the phosphopeptide in sequence analyses.In displaying data for Ser and Thr or Tyr protein kinases, those protein kinases with specificity for all three residues were repeated in each display.Kinase specificity was determined using the ''Ser/Thr protein kinase activity'' or ''Tyr protein kinase activity'' term from UniProt.Pseudo-kinases were allowed when displaying protein kinase log 2 fold change data.Membership of various other enriched gene ontology terms was determined using QuickGO.

Gene ontology enrichment
The significantly regulated phosphopeptides for the four comparisons designated P/U(4 h), P/U(24 h), D/U(4 h) and D/U(24 h) were divided into those that increased or decreased abundance, resulting in eight sets.The phosphopeptides were ranked from most significant to least significant and then from largest to smallest quantitative value.The significance ranking value and quantitative ranking value was summed for each phosphopeptide to produce a final combined rank.The gene for each phosphopeptide was extracted to produce a ranked gene list.Redundant gene names were discarded, retaining the highest ranked gene name for each phosphopeptide.The ranked gene lists were submitted to gProfiler 24 using the 'multi-query' and 'ordered query' options.Note that an ordered query is an iterative process that may not use the entire ranked gene list to obtain significance, 24 weighting the output toward the higher ranking genes.The organism was set to Mus musculus (version: GRCm38.p6).A custom background was used which consisted of all the genes represented by the proteins detected in both the proteomics and phosphoproteomics.Genes without annotation were included in the background.The following data sources and versions were used: GO Molecular Function release June 1 2020, GO cellular component release June 1 2020, GO Biological Process release June 1 2020, KEGG release July 6 2020, Reactome annotations ensemble classes July 7 2020 and WikiPathways version June 10 2020.
Significant protein values were also divided into groups of increased or decreased abundance and ranked by both significance and by the magnitude of the quantitative values.For protein groups mapping to multiple genes, the first gene (ranked by MaxQuant rules) was kept and the remaining genes discarded.Redundant genes were also removed before submission to gProfiler using the same settings as for phosphopeptides.The list of significant terms was simplified using REVIGO. 75The full output from gProfiler is supplied Table S3.
In a separate analysis of persistently regulated phosphopeptides and proteins, the log 2 fold change was required to be significant for both 4 and 24 h for inclusion.The same procedure for GO enrichment was followed except that the background was the subset of annotated genes in the Mus musculus genome (version GRCm39), the threshold for significance was set to 0.00001, the term size was limited to 400 genes and the data sources were newer.We used GO Molecular Function, GO cellular component and GO Biological Process released March 22 2022.

Phosphorylation site function analysis
The phosphorylation sites that are known to induce or inhibit a particular function were obtained from PhosphoSitePlus 27 (downloaded August 28 2020).We sequence matched the PhosphoSitePlus data to our data requiring an identical gene name and sequence within four residues either side of the phosphorylation site.Sequences from all mammals were allowed.Multi-site phosphopeptides were matched using the sequence window of the first phosphorylation site.If this caused a redundant match, the quantitative value with the largest magnitude was kept and the remainder discarded.

QUANTIFICATION AND STATISTICAL ANALYSIS
For differential phosphorylation and protein analysis by TMT or PRM mass spectrometry, statistical significance was determined using a moderated t-test within limma version 3.36.5. 73Then, multiple comparison correction was performed (adjustment of P values) using the Benjamini and Hochberg method was applied.An adjusted P value of <0.05 was considered significant.Phosphopeptides that exhibited underlying changes in protein were excluded from the figures and further analyses.Phosphopeptide data with no corresponding protein level data was kept, representing approximately one third of the detected phosphopeptides (Figure S2C).Averaged mass spectrometry data was clustered using Morpheus (https://software.broadinstitute.org/morpheus/).Protein or phosphopeptide data was hierarchically clustered using Euclidian distance with a complete linkage.
For KinSwing, 21 significance of the kinase activity score swing k was determined by permutating kinase node labels, K, to substrates, S, of the total network, M ks .Thereby the probability of observing swing k is conditional on this permuted reference distribution, of size, N (1,000 permutations), and is computed for each tail of the distribution, positive and negative swing k scores.For gene ontology enrichment, the significance threshold was P < 0.05 and the gProfiler algorithm for adjustment of multiple hypotheses was applied. 82The Pearson correlation coefficient for the comparison of TMT and PRM data was determined using GraphPad Prism 9.02.

Figure 2 .
Figure 2. Functional trends and protein kinase prediction for the phosphoproteome analysis at 4 h Log 2 fold changes are shown for phosphorylation sites with known ability to (A) induce or (B) inhibit various functions at 4 h.The median value is shown for each function as a line.The functional phosphorylation sites detected as perturbed in the phosphoproteome data as a percentage of all detected phosphorylation sites is shown for (C) induced functions and (D) inhibited functions.(E) Volcano plot of KinSwing predicted activity of protein kinases induced by pilocarpine.The size of each dot is scaled to the number of known substrates used to build the protein kinase consensus motif.(F) The average substrate log 2 fold change for known substrates of protein kinases versus the KinSwing predicted protein kinase activity.(G) Volcano plot of the log 2 fold changes for phosphopeptides occurring on Ser/Thr protein kinases.See also Figures S4 and S5 and TableS4.

Figure 3 .
Figure 3. Phosphoproteome changes at 24 h after SE (A) Volcano plot for the significantly regulated phosphopeptides at 24 h.Light gray dots are values below the threshold for significance.Dots representing the top 50 ranked phosphopeptides from proteins with the biological process term RNA splicing and the Reactome terms related to AMPA receptor activation and trafficking are colored.Enriched (B) biological process, (C) molecular function, (D) Reactome, and (E) WikiPathways terms.A scale bar for probability is shown.The size of the circle represents the number of genes enriched.Volcano plots are shown for phosphopeptides with membership of (F) the TYROBP Causal Network, (G) the UniProt curated splicing factor Ser/Arg-rich protein family, (H) guanyl-nucleotide exchange factor activity term, (I) GABA-A receptor activity term, (J) glutamate receptor activity term, (K) voltage-gated calcium channel activity term, (L) acetylcholine receptor activity term, (M) potassium channel activity term, and (N) G protein-coupled receptor activity term.See also FigureS6A, TablesS1 and S3.

Figure 4
Figure 4. Functional trends and protein kinase prediction for the phosphoproteome analysis at 4 h and persistently regulated phosphorylation Log 2 fold changes are shown for phosphorylation sites with known ability to (A) induce or (B) inhibit various functions at 24 h.The median value is shown for each function as a line.The functional phosphorylation sites detected as perturbed in the phosphoproteome data as a percentage of all detected phosphorylation sites is shown for (C) induced functions and (D) inhibited functions.(E) Volcano plot of KinSwing predicted activity of protein kinases induced by pilocarpine at 24 h.The size of each dot is scaled to the number of known substrates used to build the protein kinase consensus motif.(F) The average substrate log 2 fold change for known substrates of protein kinases perturbed by pilocarpine at 24 h versus the KinSwing predicted protein kinase activity.(G) Volcano plot of the log 2 fold changes for phosphopeptides occurring on Ser/Thr protein kinases at 24 h.(H) Heatmap of phosphopeptides significantly and persistently perturbed at both 4 and 24 h.Enriched (I) molecular function, (J) biological process, and (K) cellular component terms for persistently perturbed phosphoproteins.A scale bar for probability is shown.The size of the circle represents the number of genes enriched.See also FigureS6, TablesS3 and S4.

Figure 5 .
Figure 5. Phosphoproteome changes induced by diazepam at 4 and 24 h (A) Mice receiving a mock injection followed by diazepam injection at 40 min were compared to untreated mice.Volcano plots for the diazepam-induced phosphorylation level changes at (B) 4 h and (C) 24 h.Light gray dots are below the threshold for significance.Enriched (E) biological process terms at 4 h, (F) cellular component terms at 4 h, and (G) cellular component terms at 24 h after mock injection.A scale bar for probability is shown.The size of the circle represents the number of genes enriched.(D) Heatmap of persistently perturbed phosphorylation sites after diazepam.Volcano plots of KinSwing predicted activity of protein kinases induced by pilocarpine at (H) 4 h and (I) 24 h.The size of each dot is scaled to the number of known substrates used to build the protein kinase consensus motif.See also Tables S1, S3, and S4.

Figure 6 .
Figure 6.Proteome changes at 4 and 24 h after SE Venn diagram of (A) proteins detected and (B) perturbed at each time point, resolved to non-redundant gene names.Volcano plot of protein changes at (C) 4 h and (D) 24 h after SE.Light gray dots are below the threshold for significance.The top 50 ranked proteins with the biological process terms translation and synapse organization are shown as colored dots.(E) Protein changes persistently detected at 4 and 24 h after SE.Enriched (F) biological process, (G) molecular function, and (H) KEGG terms at 24 h.A scale bar for probability is shown.The size of the circle represents the number of genes enriched.(I) Schematic representation of the postsynaptic compartment with (descriptive) coloring according to direction of protein regulation.Created with BioRender.com.See also Figures S7 and S8, TablesS2 and S5.

Figure 7 .
Figure 7. Independent analysis of targets by parallel reaction monitoring Heatmap of (A) phosphorylation site and (B) protein changes observed by PRM at 4 and 24 h alongside the findings by TMT and the probabilities for each.Scale bars are shown.(C) Immunolabeling of hippocampal slices from mock or pilocarpine-treated mice.The slices were stained with Hoechst to mark nuclei or anti-GluA1 pS863.A region of CA1 neurites was defined (example in upper right panel).The intensity of staining was quantified and compared between the control and pilocarpinetreated samples (*, p = 0.032, one-sided t test for a decrease).Data are represented as mean G SEM. (D) Representative images.The PRM measurements for phosphorylation sites at (E) 4 h and (F) 24 h as well as the PRM measurements for protein at (G) 4 h and (H) 24 h plotted against the corresponding TMT measurements are shown.The Pearson correlation coefficient and linear regression line for the measurements is shown on each graph.Scale bars represents 200 mm.See also TableS7.