Mnk1/2 kinases regulate memory and autism-related behaviours via Syngap1

Abstract

MAPK interacting protein kinases 1 and 2 (Mnk1/2) regulate a plethora of functions, presumably via phosphorylation of their best characterized substrate, eukaryotic translation initiation factor 4E (eIF4E) on Ser209. Here, we show that, whereas deletion of Mnk1/2 (Mnk double knockout) impairs synaptic plasticity and memory in mice, ablation of phospho-eIF4E (Ser209) does not affect these processes, suggesting that Mnk1/2 possess additional downstream effectors in the brain. Translational profiling revealed only a small overlap between the Mnk1/2- and phospho-eIF4E(Ser209)-regulated translatome. We identified the synaptic Ras GTPase activating protein 1 (Syngap1), encoded by a syndromic autism gene, as a downstream target of Mnk1 because Syngap1 immunoprecipitated with Mnk1 and showed reduced phosphorylation (S788) in Mnk double knockout mice. Knockdown of Syngap1 reversed memory deficits in Mnk double knockout mice and pharmacological inhibition of Mnks rescued autism-related phenotypes in Syngap1^+/− mice. Thus, Syngap1 is a downstream effector of Mnk1, and the Mnks–Syngap1 axis regulates memory formation and autism-related behaviours.

Chalkiadaki et al. establish a previously unknown link between Mnk kinases and the autism risk gene Synap1 upstream of translational control. They reveal a key role for the Mnk–Syngap1 axis in memory and autism-related behaviours and show that the axis is amendable to pharmacological manipulation.

Introduction

Downstream of Ras/ERK (extracellular regulated kinase) and p38 (mitogen-activated protein kinase) MAPK pathways, MAPK interacting protein kinase 1 (Mnk1) and Mnk2 exert a plethora of biological functions in response to external stimuli (e.g. mitogens) and internal cues.^1,2 Only a few substrates are known for Mnk1/2 kinase activity, of which the best described is the Ser209 residue on the eukaryotic translation initiation factor 4E (eIF4E), which is involved in the regulation of cap-dependent translation.¹ Indeed, in most tissues and cell types examined, phospho-eIF4E (Ser209) is considered the key downstream effector of Mnks. Activation of Mnk1 enhances its binding to the eukaryotic initiation factor 4G (eIF4G) and promotes phosphorylation of eIF4E. In addition, eIF4E forms a complex with eIF4A and eIF4G (called eIF4F) to facilitate ribosome recruitment and promote translation initiation.²

Messenger RNA translation downstream of Mnks was shown to be important for tumorigenesis, inflammation, immunity, internal ribosome entry site-mediated translation initiation and resistance to anti-cancer drugs (e.g. inhibitors of mechanistic target of rapamycin, mTOR).³

In the brain, downstream of MAPKs, Mnks were hypothesized to regulate synaptic plasticity, learning, memory and other behaviours via eIF4E.^4,5 Accumulating evidence highlights a role for Mnks in Fragile X syndrome (FXS) and autism spectrum disorder (ASD). Mnk1 was shown to regulate translation of mRNAs involved in neurotransmission and synaptic plasticity, a significant subset of which overlaps with proteins regulated by fragile X messenger ribonucleoprotein (Fmrp).⁶ In addition, genetic deletion or pharmacological inhibition of Mnks in a mouse model of FXS rescues core FXS-related behaviours as well as numerous cellular and molecular phenotypes.⁷ A link between Mnks and ASD was recently strengthened by demonstrating that pharmacological inhibition of Mnk restores mRNA translation, oxytocin signalling and social novelty responses in a mouse model of the syndromic ASD gene Neuroligin 3 (Ngln3).⁸ Strikingly, we previously demonstrated that mice harbouring an unphosphorylatable Ser209Ala transgene (Eif4e^Ser209Ala) exhibit intact hippocampal long-term potentiation (LTP) and long-term memory,⁹ suggesting a role for additional Mnk downstream effectors, other than eIF4E, in these processes. Thus, despite the diverse roles and functions of Mnks, Mnks-mediated eIF4E phosphorylation cannot fully explain all processes under the control of Mnks in the brain, including learning and memory.

Here, we demonstrate that Mnks in the brain regulate learning, memory and ASD-like behaviours via regulation of the syndromic ASD gene, synaptic Ras GTPase activating protein 1 (Syngap1), which controls mTOR Complex 1 (mTORC1) activity. First, we show that Mnk1/2 deletion in mice (Mnk double knockout, DKO)¹⁰ impairs synaptic plasticity, learning and memory, in contrast to phenotypes previously shown in Eif4e^Ser209Ala mice.⁹ We reveal a small overlap between the differentially translated mRNAs in Eif4e^Ser209Ala and Mnk DKO mice brains, suggesting that substrates other than eIF4E dictate the effects of Mnks on brain functions. Second, we provide evidence that Mnks promote mTORC1 signalling and protein synthesis through interaction with and phosphorylation-dependent inhibition of Syngap1 by regulating its GAP (GTPase-accelerating protein) activity for Ras homologue enriched in brain (Rheb), a small GTPase upstream activator of mTORC1. Syngap1 acts as a repressor of protein synthesis and its deletion is known to lead to elevated mTORC1 signalling, increased global protein synthesis and autism-like behaviours.^11–13 Moreover, we show that manipulation of the Mnk–Syngap1 axis, using genetic or pharmacological approaches, normalizes altered mTORC1 signalling and corrects behavioural deficits in both Mnk DKO and Syngap1^+/− mice. Thus, these data establish a previously unknown link between Mnks and Syngap1 upstream of translational control and demonstrate the important role of Mnks–Syngap1 axis in memory and autism-related behaviours, which is amenable to pharmacological manipulation.

Materials and methods

Animals

All procedures were in accordance with UK Home Office and Canadian Council on Animal Care regulations and were approved by the University of Edinburgh and McGill University. Mnk1^−/−Mnk2^−/−, Mnk1^+/+Mnk2^+/+, Syngap1^+/+ and Syngap1^+/− animals were backcrossed for more than 10 generations to C57Bl/6J background. Food and water were provided ad libitum. Pups were kept with their dams until weaning at postnatal Day 21. After weaning, mice were group housed (maximum of five per cage) by sex. Cages were maintained in ventilated racks in temperature (20–21°C) and humidity (∼55%) controlled rooms, on a 12-h circadian cycle (7 a.m.–7 p.m. light period). Male and female mice were used (for Figs 1–5 only male mice were used; for Fig. 6, 50% male and 50% female mice were used for all groups).

Figure 1.

Mnk1/2 depletion impairs synaptic plasticity, learning and memory. (A) Morris water maze task. Left: Graphic depiction of experimental design; latency (s) to find hidden platform during experimental days. Right: Number of platform crosses and % quadrant occupancy during probe test on Day 6; two-way ANOVA, with Tukey’s post hoc test. (B) Contextual fear conditioning. Top: Experimental outline; bottom: percentage freezing 24 h after initial training, Student’s t-test. (C) Social approach and preference for social novelty (three-chamber test). Time spent sniffing the social (S1: Stranger 1, S2: Stranger 2) or non-social stimulus (E: Empty), two-way ANOVA with Bonferroni’s post hoc test. (D) Self-grooming. Total time grooming and number of grooming bouts are shown, Student’s t-test. (E) CA1 L-LTP recordings in response to TBS in Mnk DKO mice. Normalized fEPSP slope over 240 min. (F) Quantification of percentage potentiation during the last 10 min, Student’s t-test. For A–D, all data are shown as mean ± SEM; *P < 0.05, **P < 0.01, ***P < 0.001. Mnk DKO: Mnk1^−/−Mnk2^−/−; WT: Mnk1^+/+Mnk2^+/+. See also Supplementary Table 9 for details of statistical analysis.

Figure 2.

Altered translational landscape in Mnk DKO whole brain. (A) Experimental design for genome-wide profiling of mRNA translation with RNA sequencing in WT and Mnk DKO mouse whole brain. (B) Scatter plot showing log₂RPKM of translationally (DTG; FDR < 0.15) or transcriptionally (DEG; FDR < 0.1) upregulated and downregulated mRNAs in WT versus Mnk DKO libraries (n = 2 for footprints and mRNA). (C) Comparison of translational profiling in Mnk DKO (in B) with Eif4e^Ser209Ala ribosome profiling data from Amorim et al.⁹Left: Scatter plot comparing log₂RPKM; Mnk DKO and Eif4e^Ser209Ala datasets and DTGs; R²: Pearson correlation between depicted datasets. Right: GO analysis using DAVID for 20 common DTGs between the depicted RPF datasets. Mnk DKO: Mnk1^−/−Mnk2^−/−; eIF4E^Ser209Ala: Eif4e^{Ala209/Ala209}; WT: Mnk1^+/+Mnk2^+/+ or Eif4e^{Ser209/Ser209}. See also Supplementary Figs 1 and 2 and Supplementary Tables 1 and 2.

Figure 3.

Mnk1/2 deletion remodels whole brain and synaptic phosphoproteome. (A) Experimental design for phosphoproteome analysis of whole brain samples in WT and Mnk DKO mouse. Volcano plot showing significantly altered phosphopeptides identified in Mnk DKO mouse compared with WT. The x-axis demonstrates the log-transformed fold change in abundance (WT/DKO) and the y-axis indicates the log-transformed P-values (t-test) associated with individual phosphopeptides. A cut-off of log₂ fold-change (dashed vertical lines) and P-value 0.05 (dashed horizontal line) was applied. (B) Top ranked identified consensus motif using the Motif-x package on phosphosites significantly downregulated in the Mnk DKO whole brain. (C). Functional annotation of proteins with reduced phosphorylation (P-value < 0.05) in Mnk DKO whole brain by DAVID. GO terms for Molecular Functions (MF), Biological Processes (BP) and Cellular Component (CC) are shown. (D) Upstream kinase prediction analysis using the ingenuity pathway analysis (IPA) package for unannotated phosphosites with 2-fold change in DKO whole brain compared with WT. (E) Experimental design for phosphoproteome analysis of synaptosome samples in WT and Mnk DKO mouse. Volcano plot showing significantly altered phosphoproteins identified in Mnk DKO mouse compared with WT. The x-axis demonstrates the log-transformed fold change in abundance (WT/DKO) and the y-axis indicates the log-transformed P-values (t-test) associated with individual phosphopeptides. A cut-off of ±2-fold change (dashed vertical lines) and P-value 0.05 (dashed horizontal line) was applied. (F) The top ranked identified consensus motif using the Motif-x package on phosphosites significantly downregulated in the Mnk DKO synaptosomes. (G) Functional annotation of proteins with reduced phosphorylation (P-value < 0.05) in Mnk DKO synaptosomes by DAVID. Overrepresented GO terms for MF, BP and CC are shown. (H) Upstream kinase prediction analysis using the IPA package for unannotated phosphosites with 2-fold change in DKO synaptosome compared with WT. (I) Annotation of top Diseases and Functions enriched among the proteins with reduced phosphorylation (P-value < 0.05) in Mnk DKO synaptosomes using the IPA package. Mnk DKO: Mnk1^−/−Mnk2^−/−; WT: Mnk1^+/+Mnk2^+/+. See also Supplementary Figs 2–5 and Supplementary Tables 3–6.

Figure 4.

Mnk1/2 deletion alters synaptic translation. (A) Experimental design for genome-wide profiling of mRNA translation with RNA sequencing in WT and Mnk DKO mouse synaptosomes. Scatter plot showing log₂RPKM of translationally (DTG) upregulated and downregulated mRNAs in WT versus Mnk DKO synaptosome libraries (FDR < 0.14; n = 3 for footprints and mRNA). Mnk DKO: Mnk1^−/−Mnk2^−/−; WT: Mnk1^+/+Mnk2^+/+. (B) Comparison of translational profiling between whole brain and synaptosomes (WT-Mnk DKO). Scatter plot showing log₂RPKM Mnk DKO/WT for translational efficiency. R²: Pearson correlation between depicted datasets. Mnk DKO: Mnk1^−/−Mnk2^−/−; WT: Mnk1^+/+Mnk2^+/+. See also Supplementary Fig. 2 and Supplementary Table 5.

Figure 5.

Mnk1 binds to and phosphorylates Syngap1. (A) Identification of Mnk1 protein interactome in whole brain by co-immunoprecipitation–mass spectrometry (IP-MS). Volcano plot shows proteins co-enriched with Mnk1 using anti-Mnk1 antibody (over IgG). A cut-off of ± 3.5 log₂ fold-change (dashed vertical lines) and FDR < 0.05 (dashed horizontal line) was applied. Venn diagram demonstrates the overlap between proteins co-enriched with Mnk1 and the proteins with reduced phosphosites in Mnk DKO synaptosomes. (B) Analysis of Syngap1 phosphorylation in WT and Mnk DKO whole brains. Proteins with phospho-Ser residues were immunoprecipitated by specific anti-phospho-serine antibody and the presence of Syngap1 was measured by western blot. The bar graph represents the relative enrichment of Syngap1 protein in each genotype in the input and phospho-Ser-enriched fractions. n = 4 for each genotype; Student’s t-test, *P < 0.05. (C) Top: Conservation of the two observed Mnk-sensitive phosphosites and the presence of the predicted enriched motif (Fig. 3B and F) on Syngap1 in the indicated species. Bottom: Schematic of known domains of rat Syngap1 protein, positions of the validated and predicted phosphosites, and validated phosphorylation sites mediated by CDK5 and CamKIIα. (D) Puromycin incorporation assay in HEK-293H cells transfected with wild-type or phospho-mutant (S788A) or phospho-mimetic (S788D) Syngap1. Representative immunoblot analysis of lysates probed with antisera against the indicated proteins; HSC70 is the loading control (n = 3 for each group). One-way ANOVA with Tukey’s post hoc test, *P < 0.05. (E) mTORC1 activity in synaptosomes from Mnk DKO mouse brain. Left: Representative images from immunoblotting of synaptosome lysates probed with antisera against the indicated proteins; β-actin is the loading control. Right: Quantification of relative protein expression in immunoblotting experiment. Normalized expression of two phosphosites on rpS6 (240/244, 235/236) is depicted; n = 3 for each genotype. (F) Luminescence-based Syngap1 GAP activity assay for recombinant Rheb using Syngap1 mutants; one-way ANOVA with Bonferroni’s post hoc test, ***P < 0.001. (G) Active Rheb (Rheb-GTP) immunoprecipitation-based assay. Syngap1 WT, mutants or empty vector groups were probed with antisera against the indicated proteins using immunoprecipitation and total lysates; HSC70: loading control. (H) Luminescence-based Syngap1 GAP activity assay (from immunoprecipitated Syngap1) in brain synaptosomes. Residual GTP (indicative of GTP hydrolysis) for wild-type and Mnk DKO synaptosomes. For G and H, Student’s t-test, *P < 0.05, **P < 0.01. Mnk DKO: Mnk1^−/−Mnk2^−/−; WT: Mnk1^+/+Mnk2^+/+. See also Supplementary Figs 6 and 7 and Supplementary Tables 7 and 9 for details of statistical analysis.

Figure 6.

AAV9-mediated Syngap1 knockdown reverses memory deficits in Mnk DKO mice and pharmacological inhibition of Mnk corrects behavioural deficits in Syngap1^+/−mice. (A) Left: Experimental design for AAV9-mediated knock-down of Syngap1 mRNA in mouse brain via intra-hippocampal stereotactic injection. Short-hairpin RNA (shRNA) constructs driven by the U6 promoter were expressed for 4 weeks prior to behavioural testing. Representative images from immunoblotting of hippocampus lysates from WT mice injected with scrambled or Syngap1 shRNA expressing AAV9 for 2 or 4 weeks, probed with antisera against the indicated proteins; Hsc70 is the loading control. Middle: Behavioural analysis for the indicated groups in Morris water maze (MWM). Latency to find the hidden platform(s) is shown for different days and treatments Two-way ANOVA, with Tukey’s post hoc test. Right: Behavioural analysis for the indicated groups in contextual fear conditioning (CFC): percentage freezing 24 h after initial training is shown. (B) Left: Experimental design for eFT508 pharmacological inhibition of Mnk kinase activity in mouse brain. eFT508 (1 mg/kg, daily for 5 days) was administered via intraperitoneal injection in mice prior to behavioural testing. Behavioural analysis for the indicated groups; self-grooming: total time spent grooming, open field test: time spent in the centre of the open field and three-chamber social interaction and preference for social novelty tests: Sociability and social novelty indices are shown. Two-way ANOVA with Bonferroni’s post hoc test. (C) Immunoblot analysis and quantification of hippocampal tissue isolated from animals analysed in A. (D) Proposed mechanism for behavioural rescue in Mnk DKO mice by Syngap1 shRNA. (E) Immunoblot analysis and quantification from hippocampal tissue isolated from animals analysed in B. (F) Proposed mechanism for behavioural rescue in Syngap1^+/− mice by eFT508 treatment. For C and E: representative immunoblots probed with antisera against the indicated proteins are shown; Hsc70 is the loading control. One-way ANOVA with Bonferroni’s post hoc test. For A–E: all data are shown as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001. Mnk DKO: Mnk1^−/−Mnk2^−/−; Syngap1^+/−: Syngap1^+/−; WT: Mnk1^+/+Mnk2^+/+ or Syngap1^+/+. See also Supplementary Fig. 8 and Supplementary Table 9 for details of statistical analysis.

Four days before behavioural experiments, mice were handled for 30 min each day. On the day of the experiment, animals were transported from the housing room to the procedure room and habituated for 1 h before starting the test.

Morris water maze test

Training in the pool (100 cm pool diameter and 10 cm diameter platform; water temperature was 24°C, room at 20 lx) consisted of three trials per day (20 min inter-trial interval); each mouse was allowed to swim until it reached the hidden platform. Animals that did not find the platform after 60 s were gently guided to it and allowed to stay on the platform for 10 s prior to returning them to the cage. For the probe test on day 6, the platform was removed, and animals were allowed to swim for 60 s. The swimming trajectory and velocity were monitored with a video tracking system (HVS Image or ANY-maze, Stoelting).

Contextual fear conditioning test

Mice were, as described previously,^9,14 conditioned in the chamber: 2 min acclimatization to the context, followed by the unconditioned stimulus (US); one foot shock (0.5 mA, 2 s) followed by a 30 s interval, terminating with another identical foot shock. The mice remained in the chamber for an additional 1 min after the end of the last US, after which they were returned to their home cages. Contextual fear memory was assayed 24 h after training by re-exposing the animals to the conditioning context for a 5-min period. During this period, the incidence of freezing response (absence of movement except for respiration) was recorded (FreezeFrame, Coulbourn Instruments). Freezing behaviour was analysed by assigning animals at 5 s intervals as either freezing or not freezing. Data are expressed as the percentage of 5 s intervals scored as ‘freezing’ (freezing behaviour during the session).

Extracellular field electrophysiology

Transverse hippocampal slices (400 μm) were prepared from wild-type (WT) or Mnk DKO male mice (6–8 weeks old). Slices were allowed to recover submerged for at least 2 h at 32°C in oxygenated artificial cerebrospinal fluid (ACSF) containing 124 mM NaCl, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 26 mM NaHCO₃, 10 mM glucose, 1.3 mM MgCl₂ and 2.5 mM CaCl₂ before transferring to a recording chamber at 28–29°C, which was continuously perfused with ACSF at 2 ml/min. Field excitatory postsynaptic potentials (fEPSPs) were recorded in CA1 stratum radiatum with glass electrodes (2–3 MΩ) filled with ACSF. Schaffer collateral fEPSPs were evoked with a concentric bipolar tungsten stimulating electrode placed in stratum radiatum proximal to CA3 region. Baseline stimulation was applied every 30 s (0.033 Hz) by delivering 0.1 ms pulses, with intensity adjusted to evoke 35% of maximal fEPSPs. For the induction of late-LTP (L-LTP), theta-burst stimulation (TBS, 15 bursts of four pulses at 100 Hz separated by 200 ms intervals) was performed. For analysis, the slope of the fEPSPs was measured and values were normalized to the averaged baseline slope value for each recording. Percentage of potentiation was calculated as the difference between averaged values for a 30-min period before the tetanus (baseline) and the last 10 min of recording.

Self-grooming test

Mice were placed in a new Plexiglas cage with fresh bedding and no nest or cardboard material, as previously described.¹⁵ Self-grooming behaviour was recorded for 10 min after an initial 10 min habituation phase¹⁵ at room light of 5 lx.

Three-chamber social interaction test

Sociability and social novelty were measured in a rectangular apparatus divided into three interconnected chambers (each chamber: 36 × 28 × 30 cm, L × W × H). The test consisted of four sequential 10-min trials: (i) habituation to centre chamber; (ii) habituation to all chambers; (iii) sociability, measured as the time the mouse spent in proximity to a conspecific or an object; and (iv) social novelty preference as measured by the time the mouse spent with an unfamiliar conspecific or a familiar one. The test was conducted at 5 lx. Behaviours were video recorded and the time exploring the cylinder was measured by a blind observer. Preference in exploration shown by the test mouse was assessed in % according to the formulas: t [mouse]/(t [mouse] + t [object]) and t [novel mouse]/(t [novel mouse] + t [familiar mouse]).

Synaptosome preparation

Synaptosomes were prepared using the whole brain of adult (∼3-month-old) WT or Mnk DKO male mice of each genotype using the Syn-PER™ Synaptic Protein Extraction Reagent (Thermo Fisher) as per the manufacturer’s instructions. The lysis buffer was supplemented with 100 µg/ml cycloheximide for ribosome profiling (Sigma-Aldrich) and protease (Complete EDTA-free, Pierce) and phosphatase inhibitor (PhosSTOP, Sigma-Aldrich) tablets. Synaptosome lysates were stored at −80^οC until use.

Ribosome profiling of mouse whole brain and synaptosomes

We used the Epicentre TruSeq Ribo Profile (Mammalian) Kit (Illumina, RPYSC12116), as previously described,^9,14,16,17 with some modifications,^9,14,16,17 to generate sequencing libraries. In brief, polysomes were extracted from snap-frozen, whole brain (from 1 animal) or synaptosomes (from 2–3 animals) of each genotype in the presence of cycloheximide. A portion of the lysate was used for footprint generation using TruSeq Ribo Profile Nuclease (ribosome protected fragments, RPFs), while an equal portion of the lysate was kept as an internal transcription control (total mRNA). After digestion, RPFs corresponding to monosomes were size-purified on MicroSpin S-400 columns as described in the kit to enrich for small RNA fragments (28–32 nt). RPFs and total mRNA were depleted of ribosomal RNA using the Ribo-Zero Gold (Human/Mouse/Rat) Kit (Illumina, MRZG126). RPFs were further purified on a 15% TBE–urea polyacrylamide gel, selecting bands running between 28 and 32 nt. Total mRNA samples were heat fragmented. All samples were end-repaired using TruSeq Ribo Profile Polynucleotide kinase, followed by ligation of a TruSeq Ribo Profile 3′ Adapter. All samples were reverse transcribed into cDNA, followed by a further polyacrylamide gel electrophoresis (PAGE) purification on a 10% TBE–urea gel, to separate sample cDNA from excess adapter. Purified cDNA was circularized, and PCR amplified, following purification using the Agencourt AMPure XP kit (Beckman Coulter). PCR products were further purified on an 8% TBE PAGE, to yield sufficient quantity and quality for sequencing. All samples were analysed on an Agilent Bioanalyzer High Sensitivity DNA chip to confirm expected size range and quantity and sequenced on an Illumina HiSeq 2500 system.

Bioinformatics analysis

Data were analysed as previously described with modifications.^9,14,16,17 Synaptosome ribosome profiling analysis was carried out as described previously.¹⁷ Sequencing data were de-multiplexed by the sequencing facility (Edinburgh Genomics). Obtained sequences were analysed using a custom developed pipeline (following the methods used by Ingolia et al.¹⁸). In brief, reads were adapter-trimmed using the FASTX toolkit, contaminant sequences (rRNA, tRNA) removed using bowtie and reads aligned to a reference genome using STAR. Cufflinks was used to quantify reads and calculate reads per kilobase of transcript per million mapped reads (RPKM) values for each transcript. Translational efficiency for each transcript was calculated by dividing RPKM values of the RPF libraries by RPKM values of the Total mRNA libraries. Changes in transcription were analysed for pairwise comparisons, based on experimental design, using microarray normalization methods, as reviewed by Quackenbush.¹⁹ Changes in translation were assessed using the R package Xtail v1.1.5.²⁰

Gene Ontology analysis

Gene Ontology (GO) analysis was carried out, as previously,^9,14,17 with ingenuity pathway analysis (IPA; Qiagen, Inc.). Datasets were uploaded on IPA and submitted to Core Analysis with analysis parameters set to include Direct and Indirect Interactions and Experimentally Observed data only. Ingenuity canonical pathways were obtained for all datasets and processed according to P-value, or DAVID (Database for Annotation, Visualization and Integrated Discovery, version 6.8). Datasets were submitted to DAVID and GO annotation gathered for Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways and molecular function and cellular component GO annotations, or g:profiler software (functional enrichment analysis was carried using the g:Ghost package of g:profiler to assign GO categories to ribosome profiling lists of differentially translated genes).²¹ Hierarchical filtering was used—best per parent group-strong. The probability threshold for all functional categories was set at 0.05, using correction for multiple testing with the g:SCS algorithm.²¹ All output from GO analyses is summarized in the respective Supplementary material.

Quantitative proteomics and phospho-proteomics

Whole brain and synaptosome lysates were prepared as previously.^14,17 Four times the sample volume of cold (−20°C) acetone was added to the lysates in 1.5-ml Eppendorf tubes followed by vortexing for 10 s and overnight incubation at −20°C. Next the samples were centrifuged at 15 000 g at 4°C for 10 min. The supernatant was removed and the uncapped tubes were kept at room temperature for 30 min to allow evaporation of remaining acetone. The resulting pellets were stored at −80°C until analysis. For each sample, 500 µg of pellets of whole brain or synaptosome lysate (measured by Bradford assay) were reconstituted in 50 mM ammonium bicarbonate with 10 mM Tris(2-carboxyethyl) phosphine hydrochloride (Thermo Fisher Scientific) and vortexed for 1 h at 37°C. Chloroacetamide (Sigma-Aldrich) was added for alkylation to a final concentration of 55 mM. Samples were vortexed for another hour at 37°C. Ten micrograms of trypsin were added and digestion was performed for 8 h at 37°C. Samples were dried down in a speed-vac. For the TiO₂ enrichment procedure, sample loading, washing and elution were performed by spinning the microcolumn at 8000 rpm at 4°C in a regular Eppendorf microcentrifuge. The spinning time and speed were adjusted as a function of the elution rate. Phosphoproteome enrichment was performed with TiO₂ columns from GL Sciences. Digests were dissolved in 400 μl of 250 mM lactic acid (3% trifluroacetic acid, TFA/70% acetonitrile, ACN) and centrifuged for 5 min at 13 000 rpm, and the soluble supernatant was loaded on the TiO₂ microcolumn previously equilibrated with 100 μl of 3% TFA/70% ACN. Each microcolumn was washed with 100 μl of lactic acid solution followed by 200 μl of 3% TFA/70% ACN to remove non-specific binding peptides. Phosphopeptides were eluted with 200 μl of 1% NH₄OH pH 10 in water and acidified with 7 μl of TFA. Eluates from TiO₂ microcolumns were desalted using Oasis HLB cartridges by spinning at 1200 rpm at 4°C. After conditioning with 1 ml of 100% ACN/0.1% TFA and washing with 0.1% TFA in water, the sample was loaded, washed with 0.1% TFA in water, then eluted with 1 ml of 70% ACN (0.1% TFA) prior to evaporation on a SpeedVac. The extracted peptide samples were dried down and solubilized in 5% ACN–0.2% formic acid (FA). The samples were loaded on an Optimize Technologies C₁₈ precolumn [0.3-mm inside diameter (i.d.) × 5 mm] connected directly to the switching valve. They were separated on a home-made reversed-phase column (150-μm i.d. × 150 mm) with a 240-min gradient from 10 to 30% ACN–0.2% FA and a 600-nl/min flow rate on a Nano-LC-Ultra-2D (Eksigent) connected to a Q-Exactive Plus (Thermo Fisher Scientific). Each full mass spectrometry (MS) spectrum acquired at a resolution of 70 000 was followed by 12 tandem-MS (MS-MS) spectra on the most abundant multiply charged precursor ions. Tandem-MS experiments were performed using higher-energy collisional dissociation at a collision energy of 25%. The data were processed using PEAKS 8.5 (Bioinformatics Solutions) and a Uniprot mouse database. Mass tolerances on precursor and fragment ions were 10 ppm and 0.01 Da, respectively. Variable selected posttranslational modifications were carbamidomethyl (C), oxidation (M), deamidation (NQ), acetyl (N-term) and phosphorylation (STY). The data were visualized with Scaffold 4.3.0 [protein threshold, 99%, with at least two peptides identified and false discovery rate (FDR) < 0.1% for peptides].

Co-immunoprecipitation

Freshly dissected whole mouse brain samples were homogenized in lysis buffer (50 mM HEPES pH 7.4, 150 mM NaCl, 1% CHAPS + protease inhibitor + phosphatase inhibitor) on ice using a Dounce glass homogenizer. The homogenates were incubated for 20 min at 4°C with rotation, followed by centrifugation at 15 000 rpm for 10 min at 4°C. The collected supernatants were precleared with 50 µl of prewashed protein A agarose slurry beads on rotator for 30 min at 4°C followed by centrifugation at 3500 rpm for 1 min at 4°C to collect the precleared supernatants. The precleared supernatants were incubated with either 3 µg of normal rabbit IgG (CST #2729S) or 3 µg of Mnk1 antibody (CST #2195S) for 30 min at 4°C, followed by incubation with 50 μl of prewashed protein A agarose slurry beads overnight at 4°C. The mixtures were centrifuged at 3500 rpm for 1 min at 4°C to remove the supernatants as unbound fractions. Beads were washed three times with lysis buffer for 10 min at 4°C and the bound proteins were eluted in 2 × sodium dodecyl sulphate (SDS) sample buffer.

Analysis of immunoprecipitation proteomics

Mass spectrometry data were analysed using MaxQuant v1.5.8.3 using the label free quantification settings and interactors were filtered in Perseus v1.5.8.5. Proteins were selected that contained at least one unique peptide and three valid values in triplicates of Mnk or IgG pull-down groups. Values were log₂ transformed and missing values were replaced by random values that were drawn from normal distribution using the default settings (Width = 0.3, Down shift = 1.8) and as described previously.²² Significance of fold changes was calculated using a two-sided t-test and P-values were corrected for multiple testing using 250 permutation-based FDR correction. Interactors were defined significant when fold change was greater than the 99% CI of IgG fold change (log₂FC > 3.5) and FDR < 0.05 (−Log FDR was greater than 1.3). Corresponding volcano plots were generated using R.

Phosphoserine immunoprecipitation

Whole brains were freshly dissected from WT mice (Mnk1^+/+Mnk2^+/+). The synaptosomal fraction was prepared from one hemisphere as described above, while the other hemisphere was homogenized in lysis buffer (50 mM HEPES pH 7.4, 150 mM NaCl, 1% CHAPS + protease inhibitor + phosphatase inhibitor) on ice using a Dounce glass homogenizer. The homogenates (whole brain or synaptosome) were rotated for 20 min at 4°C and centrifuged at 15 000 rpm for 10 min at 4°C, followed by collection of the supernatant, which was subsequently precleared with 50 μl of protein A agarose slurry beads for 30 min at 4°C. The mixtures were centrifuged at 3500 rpm for 1 min at 4°C and the precleared supernatants were collected. The precleared supernatants (1.8 mg of whole brains, 600 μg of synaptosome fractions) were incubated with phosphoserine antibody (Millipore, AB1603) for 30 min at 4°C, followed by addition of and incubation with 50 μl of protein A agarose slurry beads at 4°C, overnight. The mixtures were subsequently centrifuged at 3500 rpm for 1 min at 4°C. The first supernatants were collected as unbound fractions. Beads were washed with lysis buffer 3 × 10 min, at 4°C and the bound proteins were eluted in 2 × SDS sample buffer.

Plasmid generation (mutant Syngap1)

Mouse Syngap1 coding sequence (CDS; WT or S788A or S788D) were synthesized using Custom Gene Synthesis (Biomatik) and were subcloned to the pBSK(+) vector and subsequently transferred using blunt cloning to a pcDNA3.1(+) backbone (Addgene) to generate the final plasmids. Syngap1 sequence was confirmed with Sanger sequencing.

Mammalian cell culture and transfection

All cell culture and transfection reagents were from ThermoFisher Scientific. Human Embryonic Kidney cells (HEK-293H ATCC® CRL-1573) were cultured (37°C, 5% CO₂) in Dulbecco’s modified Eagle’s medium (11995065) supplemented with 10% foetal bovine serum (10500064) and 1% Pen/Strep (15140148). Transient transfection was performed with Lipofectamine 3000 (L3000008) in Opti-MEM (31985070) following the manufacturer’s protocol for 48 h.

Surface sensing of translation assay

After 48 h of transfection, cells were pulsed with 5 µg/ml puromycin hydrochloride for 1 h in culture medium. Cells were washed three times with ice-cold phosphate-buffered saline (PBS) and lysed in RIPA buffer (10 mM Tris–HCl, pH 8.0, 1 mM EDTA, 0.5 mM EGTA, 1% Triton X-100, 0.1% sodium deoxycholate, 0.1% SDS, 140 mM NaCl; all from Sigma, supplemented with phosphatase and protease inhibitors; Roche) for immunoblotting with anti-puromycin antibody (MABE343; 12D10 Sigma). Puromycin incorporation was quantified with ImageJ (immunoblot signal intensity) and was normalized to control (no puromycin or no antibody) and membrane background signals.

Rheb GTPase and Syngap1 GAP activity assays

Rheb GTPase assay and Syngap1 GAP activity assays were performed according to the GTPase-Glo™ Assay manufacturer’s instructions (Promega, V7681). Recombinant Rheb protein (0–2.5 μg, Abcam, ab78768) was used in reactions containing 10 µM GTP in GTPase/GAP buffer and 1 mM DTT. Reactions were incubated for 90 min at room temperature. For the Syngap1 GAP activity assay, immunoprecipitated Syngap1 from HEK-293H cells transfected with empty vector, WT, S788D or S788A Syngap1 was added to a Rheb GTPase reaction in GTPase/GAP buffer containing 10 µM GTP and 1 μg Rheb (Abcam, ab78768). GTPase reactions for Syngap1 GAP activity were incubated for 2 h. For both assays, GTPase-Glo™ Reagent was then added. After a further incubation of 30 min at room temperature Detection Reagent was added and luminescence was recorded on a PerkinElmer LS55 luminometer. Luminescence is measured as relative light units and corresponds to residual GTP amount after the GTPase reaction.

Syngap1 immunoprecipitation

Five hundred micrograms of transfected HEK-293H lysate or mouse brain synaptosomes resuspended in RIPA buffer (20 mM Tris–HCl pH 7.5, 150 mM NaCl, 1 mM Na₂EDTA, 1 mM EGTA, 1% NP-40, 1% sodium deoxycholate, 2.5 mM sodium pyrophosphate, 1 mM beta-glycerophosphate, 1 mM Na₃VO₄, 1 µg/ml leupeptin and 1 mM PMSF) were incubated with 20 µl of protein G beads (Invitrogen, 10003D) preincubated with 1 μg rabbit IgG (Sigma, 12–370) or Syngap1 antibody (ThermoFisher Scientific, rabbit polyclonal, PA1-046) for 1 h at 4°C. Samples were washed thrice with RIPA buffer and either processed with immunoblotting or stored at −80°C.

Active Rheb (Rheb-GTP) immunoprecipitation assay

Rheb-GTP was detected with immunoblotting using the Rheb Pull-Down Activation Assay Kit (New East Biosciences, 81201). Transfected HEK-293H cells were washed thrice with PBS and lysed in glass Dounce homogenizers. Lysates were left on ice for 15 min with occasional vortexing and then centrifuged at 4°C at 16, 000 g for 20 min. A total of 1 mg of the supernatant was used to immunoprecipitate active-Rheb, to which 1 µg anti-Rheb-GTP antibody (New East Biosciences, Cat. No. 26910) and 20 µl of A/G agarose bead slurry (New East Biosciences, Cat. No. 30301) were added. Samples were incubated for 1 h at 4°C with gentle agitation. Beads were collected through centrifugation at 5000 g for 1 min, washed thrice with Assay/Lysis Buffer (New East Biosciences) and resuspended in SDS-PAGE buffer for immunoblotting.

Immunoblotting

Dissected brain tissue, HEK-293H lysates, immunoprecipitates, cap-column or synaptosomes were homogenized in RIPA (see above) or buffer B (50 mM MOPS/KOH pH 7.4, 100 mM NaCl, 50 mM NaF, 2 mM EDTA, 2 mM EGTA, 1% NP-40, 7 mM β-mercaptoethanol), both supplemented with protease and phosphatase inhibitors (Roche), using a glass Dounce homogenizer (∼30 strokes). Samples were incubated on ice for 15 min, with occasional vortexing, and cleared by centrifugation for 20 min at 16 000 g at 4°C. The supernatant was used for western blotting after the protein concentration of each sample was determined by measuring A₂₈₀ absorbance on a NanoDrop (ThermoFisher Scientific). Fifty micrograms of protein per lane were prepared in SDS sample buffer (50 mM Tris pH 6.8, 100 mM DTT, 2% SDS, 10% glycerol, 0.1% bromophenol blue), heated to 98°C for 5 min and resolved on 10–16% polyacrylamide gels. Proteins were transferred to 0.2 µm nitrocellulose membranes (Bio-Rad), blocked in 5% bovine serum albumin (BSA) in TBS-T (10 mM Tris pH 7.6, 150 mM NaCl, 0.1% Tween-20) for 1 h at room temperature, incubated with primary antibodies overnight at 4°C and with secondary antibodies for 1 h at room temperature. Primary antibodies were diluted in 1% BSA in TBS-T containing 0.02% Na azide, and between incubations membranes were washed extensively in TBS-T. Blots were imaged using an Odyssey Imaging System (Li-COR Biosciences) at a resolution of 169 μm and quantified using the ImageStudio Software (Li-COR Biosciences). For quantitative Western Blotting, the intensity of each protein band was measured in triplicate to minimize measuring variability. HSC70 or β-actin were used as a loading control. Data are shown as arbitrary units (AU) as a proxy for protein expression, after normalization to control (for protein phosphorylation: phospho-protein values were measured with ImageJ or Image Studio™ Lite and divided to total protein and to loading control: HSC70, β-actin or GAPDH band intensity values, after subtracting immunoblot background intensity).

AAV9-shRNA cloning and preparation

AAV9-shRNA viral particles were prepared by Vector Biolabs. The validated sequence targeting Syngap1 was: 5′-CACC-GCTCTATCAAACGTACAAAGTCTC-GAGACTTTGTACGTTTGATAGAGC-TTTTT-3′ and the scrambled sequence used as control was: 5′-CACC-GAACAAGATGAAGAGCACCA-CTCGAG-TGGTGCTCTTCATCTTGTTC-TTTTTT-3′.

Intrahippocampal injection of AAV9

Four-week-old mice were anaesthetized using isoflurane and secured in the stereotaxic apparatus (Kopf). A midline incision was made to expose the skull and two holes were drilled above the CA1 region of the hippocampi (AP: −1.90 mm, ML: ± 1.0 mm and DV: −1.50 mm). Next, 1 µl of either AAV9-GFP-U6-scramble-shRNA (3.1 × 10E13 GC/ml) or AAV9-GFP-U6-mSyngap1-shRNA (1.3 × 10E13 GC/ml) was bilaterally injected into the CA1 using a 10-µl Hamilton syringe connected to a 23-gauge needle mounted on a perfusion pump (Harvard Apparatus, pump 11 Elite). The injection rate was set at 500 nl per min and the needle was left for an additional 3 min before it was slowly withdrawn.

eFT508 treatment

Tomivosertib (eFT508) (MCE), which was dissolved as concentration stocks (3.22 mg/ml) in Tween 80 (20% v/v), DMSO (32% v/v) and PBS (48% v/v), was freshly diluted in PBS to appropriate working concentrations (1 mg/ml). Intraperitoneal injections (1 mg/kg) were carried out daily for 5 days prior to behavioural tests and throughout the experiment. Behavioural tests were performed 24 h after the last injection.

Statistical analysis

Experimenters were blinded to the genotype during testing and scoring for all behavioural tests in this study and for all molecular and bioinformatics analysis. All data are presented as mean ± SEM (error bars) and individual experimental points are depicted in column or bar graphs. No nested data were obtained in this study, as we only collected one observation per research object. Exclusion criteria for animal experimentation included only poor health as assessed by trained veterinarians and a routine overall health assessment during colony monitoring. Statistical significance was set a priori at 0.05 (non-significant, n.s.). No randomization was performed in the design and implementation of this study. Sample size was determined using power analysis only for rodent behavioural experiments and was based either on published or pilot data using G*Power (80% power and 0.05 alpha). Details for statistical tests used were provided within figure legends or the relative methods description and summarized in Supplementary Table 9. Supplementary Fig. 9 contains raw immunoblot data. Statistical analysis was carried out using GraphPad Prism 9.

Data availability

RNA-seq and proteomics data (∼70 GB and description) are available from the authors upon reasonable request. Proteomics data are available at Mendeley Data, V1, doi: 10.17632/6mssjkcjz2.1.

Results

Mnk1/2 depletion impairs synaptic plasticity, learning and memory

We previously showed that Eif4e^Ser209Ala mice display intact hippocampal synaptic plasticity, learning and memory.⁹ Unlike Eif4e^Ser209Ala mice, Mnk DKO mice exhibited impaired spatial learning and memory in the Morris water maze test (Fig. 1A) and showed significantly decreased long-term memory in the contextual fear conditioning test 24 h after training (∼31% reduction in freezing behaviour in Mnk DKO; Fig. 1B). In addition, we examined social behaviour (social approach and preference for social novelty) and we did not detect any significant changes between Mnk DKO and WT mice (Fig. 1C). Self-grooming was significantly increased in Mnk DKO mice (total time grooming, but not the number of grooming bouts; Fig. 1D). Moreover, the late phase of LTP (L-LTP) in CA1 hippocampal area, a form of plasticity which is MAPK- and protein synthesis-dependent,^23,24 was impaired in Mnk DKO mice. Theta burst stimulation of the Schaffer collateral-CA1 synapses elicited long-lasting potentiation of fEPSPs in WT, but not in Mnk DKO hippocampal slices (Fig. 1E and F). Taken together, these data show that Mnk1/2 deletion impairs L-LTP and hippocampus-dependent learning and long-term memory.

Altered translational landscape in Mnk1/2 DKO brain

Given the cardinal role of Mnk kinases in translational control and the significant effects of their deletion on behaviour (Fig. 1), we measured the genome-wide changes in mRNA translation in Mnk DKO mice using ribosome profiling.²⁵ Using whole brain tissue, we generated libraries for RNA sequencing from fragmented poly(A)-enriched total RNA (to measure mRNA abundance) and from ribosome-protected footprints following RNase I digestion (to measure translation), to assess the genome-wide translation efficiency (TE) of mRNAs (Fig. 2A). Ribosome profiling yielded high-quality reads, as shown by the high correlation in RPKM between biological replicates (R² < 0.95), the canonical distribution of footprint size (28–32 nt), the read distribution within the three mRNA reading frames and by the canonical periodicity of ribosomal footprints across mRNA’s coding and non-coding regions (Supplementary Fig. 1A–D). We observed a modest reduction in global translation (∼17.2%) and no significant change in global mRNA levels in Mnk DKO whole brain (Fig. 2B; R² = 0.828 and 0.993, respectively). However, we detected significant changes in translation efficiency and abundance of specific mRNAs. Differentially translated genes (DTGs) in Mnk DKO included 51 upregulated and 57 downregulated mRNAs, while differentially expressed genes (DEGs) included 55 upregulated and 134 downregulated genes (FDR < 15% for DTG and < 10% for DEG; Fig. 2B). GO analysis revealed extracellular matrix (ECM) as a major category significantly enriched in Mnk DKO DTGs and DEGs, while other categories include calcium ion binding, circadian rhythm and links to neurological disorders, cancer and embryonic development (Supplementary Fig. 1E and F and Supplementary Tables 1 and 2). Interestingly, comparison of Mnk DKO whole brain ribosome profiling to Eif4e^Ser209Ala (carried out previously by Amorim et al.⁹) revealed only 20 common DTGs (Fig. 2C), 19 common DEGs (Supplementary Fig. 2A) and a very low correlation between these datasets (R² = 0.048). The top GO categories for the 20 common DTGs are related to ECM (Fig. 2C). While this result is in accordance with previous studies^7,26 and further highlights a key role for translational control via eIF4E Ser209 phosphorylation downstream of Mnks in the regulation of ECM, it also suggests that there are other yet unidentified Mnk1/2 downstream targets underlying Mnk1/2-mediated translational regulation in the brain.

Mnk1/2 deletion remodels whole brain and synaptic phosphoproteome and alters synaptic translation

To study the molecular substrates of Mnk kinases in the brain, we carried out label-free phospho-proteomics MS analysis of WT and Mnk DKO mice whole brain lysates using titanium dioxide (TiO₂) mediated selective enrichment of phosphorylated peptides, coupled with liquid chromatography-mass spectrometry (LC-MS). Whole brain phosphoproteomic analysis revealed significant changes in a sizable portion of phosphopeptides in Mnk DKO brains compared with WT (fold change > 2, P-value < 0.05; 145 downregulated and 164 upregulated unique phosphopeptides; Fig. 3A and Supplementary Table 3). Motif analysis of the downregulated phosphopeptides in Mnk DKO whole brain, using the motif-x biological sequence motif discovery tool,²⁷ revealed a ∼24-fold enrichment for the motif RxxSxSP (Fig. 3B). Interestingly, more than 70% of the differentially phosphorylated proteins in Mnk DKO whole brains are annotated as synaptic proteins or linked to synaptic function and the GO analysis revealed several significantly regulated categories, with an enrichment for synaptic compartments (postsynaptic density, synapse, dendrite), which are also linked to local protein synthesis (Fig. 3C and Supplementary Table 4). Notably, predicted upstream regulators include Ca²⁺ and the cyclin-dependent kinase 5 (CDK5) and its activator enzyme CDK5R1 (Fig. 3D and Supplementary Table 4).

Because of the enrichment of synapse-related GO categories in whole brain Mnk DKO lysates, suggesting a potential mechanistic link between the changes in synaptic phosphoproteome and the role of Mnks in brain, we performed phosphoproteomics on synaptosome fractions (Fig. 3E). We extracted synaptosomal fractions with high purity, as evidenced by enrichment for pre- and post-synaptic proteins and depletion of nuclear markers and glial-specific proteins (Supplementary Fig. 3). Mnk1/2 deletion induced pervasive changes in the synaptic phosphoproteome, even more so when compared with whole brain (Fig. 3E). We found 469 downregulated and 423 upregulated unique phosphopeptides in Mnk DKO brains compared with wild-type (fold change > 2, P-value < 0.05; Fig. 3E and Supplementary Table 3). Compared to whole brain, motif-x analysis identified a similar, albeit better-defined phosphorylation motif: SP*E*KSP*EAK (hereafter referred to as ‘SPEAK’ motif), with 166.97-fold enrichment among downregulated synaptic phosphopeptides in Mnk DKO (Fig. 3F). The SPEAK motif is similar to phosphorylation motifs of other kinases: glycogen synthase kinase-3 (GSK3),²⁸ CDK5²⁹ and ERK³⁰ (Supplementary Fig. 4A). Akin to whole brain phosphoproteomics, postsynaptic density and Ca²⁺ transport were among the top GO categories (Fig. 3G and Supplementary Table 4), while Ca²⁺, CDK5 and CDK5R1 were the top upstream regulators (Fig. 3H and Supplementary Table 4). In particular, for the diseases and functions category of the IPA analysis, the synaptosomes phosphoproteomics dataset contained significantly more proteins within the Nervous System Development and Function and Behaviour categories, compared with the whole brain dataset (Fig. 3I and Supplementary Table 4). Given the spatial memory phenotypes in Mnk DKO mice (Fig. 1) and the emerging role of Mnks in ASD, we compared the phosphoproteomics datasets to Simons Foundation Autism Research Initiative (SFARI) syndromic genes and FMRP CLIP mRNAs^31,32 and identified several common targets (Supplementary Fig. 5A), while there is also a significant overlap between our datasets and a previous SILAC proteomics dataset from Mnk1 mouse knockout neurons⁶ (Supplementary Fig. 5B). Finally, it is plausible that the SPEAK motif is not a bona fide Mnk1/2 phosphorylation motif, but its enrichment is the result of indirect phosphorylation by other kinases. Thus, we identified known kinases, which were differentially phosphorylated in the whole brain and synaptosome datasets, among which is calcium/calmodulin dependent protein kinase II alpha (CamKIIα; Supplementary Fig. 4B and C). Several phosphorylation sites corresponding to CamKIIα are among the top differentially phosphorylated (upregulated and downregulated) in Mnk DKO whole brain and synaptosomes (Supplementary Table 3). Pathway analysis of known kinases identified in Mnk DKO whole brain and synaptosomes phosphoproteomics datasets highlighted several significantly altered categories, such as inositol triphosphate (IP3) and diacylglycerol (DAG) pathways, inflammation (TRP channels), LTP, insulin resistance, calcium and calmodulin related pathways and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionate (AMPA) receptor trafficking (Supplementary Fig. 4B and C). We also detected oxytocin signalling among GO categories (Supplementary Fig. 4C) in accordance with a recent report.⁸

Because of the pronounced effects of Mnk1/2 deletion on the synaptic phosphoproteome, we reasoned that local mRNA translation could also be significantly affected. To this end, we employed ribosome profiling in synaptosomes, a technique recently developed in our laboratory¹⁷ (Fig. 4A). We generated high-quality ribosome profiling libraries from synaptosomes (Supplementary Fig. 2B–F) and detected 122 DTG in Mnk DKO (Fig. 4A and Supplementary Table 5). We detected a modest correlation between Mnk DO whole brain and synaptosome translational efficiency (R² = 0.689; Fig. 4B), suggesting that the translation of a distinct subset of synaptic mRNAs is regulated by Mnk1/2. ∼10% of mRNAs that are downregulated in Mnk1/2 DKO synaptosome were also downregulated in Mnk1/2 DKO brain (Supplementary Fig. 5C and D). GO analysis of Mnk1/2 translationally controlled mRNAs in synaptosomes revealed enrichment for mitochondria-related categories, but on ECM-related genes (Supplementary Table 6). Furthermore, several ribosomal protein coding mRNAs were also translationally downregulated in Mnk DKO (Supplementary Table 5), consistent with an mTORC1 downregulated signature in Mnk DKO brain (see below; see also Thoreen et al.³³). Together, these data reveal a pivotal role for Mnk1/2 kinases in the regulation of the synaptic phosphoproteome and synaptic mRNA translation, with wider implications for synaptic plasticity, learning, memory and autism-related behaviours.

Mnk binds to and modulates Syngap1 phosphorylation

Mnk1 is the predominant Mnk in the brain.⁶ To identify targets that may directly bind to and are phosphorylated by Mnks, we performed an immunoprecipitation assay using the Mnk1 antibody in whole brain lysates (Supplementary Fig. 6A). We detected in immunoprecipitates a single band of ∼48 kDa corresponding to Mnk1, evident by Coomassie blue stain and confirmed by immunoblotting with Mnk1-specific antisera, which was absent from immunoglobulin G (IgG) control (Supplementary Fig. 6B). Quantitative MS analysis detected Mnk1 along with 10 other proteins as the top targets among the immunoprecipitated proteins (log₂ fold change > 3.5, FDR < 0.05; Fig. 5A and Supplementary Table 7). Interestingly, the only common target between the Mnk1 interactors and the differentially phosphorylated proteins in the Mnk DKO mice was Syngap1 (Fig. 5A), which is linked to intellectual disability and is a syndromic ASD gene.³⁴ We confirmed the Mnk-regulated phosphorylation of Syngap1 by immunoprecipitation with anti-phosphoserine antibody in WT and Mnk DKO brain lysates and immunoblotting with the Syngap1 antibody, which revealed significantly reduced levels of phosphorylated Syngap1 in Mnk DKO compared with WT brain lysates (∼65% decrease; Fig. 5B). In our phosphoproteomics dataset, we detected two phosphorylation sites on Syngap1: S788 (downregulated in Mnk DKO) and S1165 (upregulated in Mnk DKO), which are highly conserved between mouse, rat and human (Fig. 5C). Intriguingly, these sites are proximal to known CDK5 and CamKIIα phosphorylation sites on rat Syngap1³⁵ (Fig. 5C, bottom). Recent analysis of several human Syngap1 variants³⁶ and previous work³⁵ showed that CDK5 phosphorylates Syngap1 at S788.

Given previous research linking Syngap1 to regulation of protein synthesis,^11,13 we reasoned that phosphorylated Syngap1 S788 could constitute a newly identified effector of Mnk1 in brain, regulating translation and ultimately synaptic plasticity, memory and autism-like behaviours. To assess the role of the Syngap1 S788 phospho-site in regulation of global protein synthesis, we generated Syngap1 WT and phospho-mutant (Syngap1^S788A; phospho-mutant or Syngap1^S788D phospho-mimetic) expression constructs and performed transient transfection in HEK-293H cells (Fig. 5D). We then pulsed HEK-293H cells with puromycin and detected its incorporation into nascent peptides with immunoblotting (Surface Sensing of Translation method; Fig. 5D). Expression of WT Syngap1 led to a ∼30% decrease in puromycin incorporation (a proxy for global protein synthesis), compared with empty vector (Fig. 5D). Phospho-mimetic Syngap1^S788D expression displayed loss of function in inhibiting protein synthesis, while phospho-mutant Syngap1^S788A expression inhibited protein synthesis akin to WT Syngap1 (Fig. 5D). These data, in conjunction with the reduction in Syngap1 S788 phosphorylation in Mnk DKO brain (Fig. 3 and Supplementary Table 3), suggest that Mnk phosphorylation of Syngap1 on S788 promotes protein synthesis.

To further understand the link between Mnk1 and Syngap1, and because Syngap1 is known to regulate signalling pathways upstream of translation (e.g. upregulation of mTORC1 signalling in Syngap1^+/− mouse brain¹³), we examined three key pathways linked to translational control (mTORC1, Akt and MAPK) in synaptosomes from Mnk DKO and WT mice by immunoblotting for phospho-ribosomal protein S6 (rpS6 240/244 and 235/236 sites), phospho-Akt(S473) and phospho-ERK (T202/Y204), respectively (Supplementary Fig. 6C and D). We detected significantly decreased phosphorylation of rpS6 on both sites (240/244, 235/236) in Mnk DKO compared with WT, indicating decreased mTORC1 activity (Fig. 5E). We did not detect any significant changes in Akt or MAPK signalling (Supplementary Fig. 6C and D). Akt and ERK are downstream to Ras, which is regulated by Syngap1.³⁵ Thus, we reasoned that plausibly Syngap1 S788 phosphorylation may regulate mTORC1 via increased GAP-activity for the Rheb GTPase, independent of p-ERK or p-Akt. To examine if S788 phosphorylation alters Syngap1 GAP activity towards Rheb GTPase (Supplementary Fig. 7A) we performed GAP activity assay using immunoprecipitated Syngap1 from HEK-293H cells (Supplementary Fig. 7B) transfected with the different mutants of Syngap1, using a recombinant Rheb GTPase assay based on luminescence (Fig. 5F). WT and S788A Syngap1 showed a significant increase in GAP activity for Rheb compared with empty vector, as evidenced by the decreased (∼41%) residual GTP using recombinant Rheb in the GAP activity assay, compared with S788D (Fig. 5F). IgG immunoprecipitation control samples did not significantly affect Syngap1 GAP activity for Rheb (Supplementary Fig. 7C). To further elaborate on these findings, we performed active-Rheb (Rheb-GTP) immunoprecipitation from HEK-293H cells transfected with the different mutants of Syngap1 (Fig. 5G). WT and S788A expression led to reduced active Rheb (Rheb-GTP) recovered in the immunoprecipitate (Fig. 5G). We also detected decreased (∼30.6%) residual GTP in Syngap1 immunoprecipitated from Mnk1/2 DKO brain synaptosomes compared with WT (Fig. 5H), indicating that immunoprecipitated Syngap1 has increased GAP activity for Rheb and that Rheb is significantly less active in Mnk DKO compared with WT mouse brain synaptosomes. IgG immunoprecipitation control samples did not significantly affect Syngap1 GAP activity for Rheb (Supplementary Fig. 7D).

These data suggest that Mnk-mediated phosphorylation of Syngap1 promotes mTORC1 activity and thereby protein synthesis, possibly via reduced Syngap1 GAP-activity for Rheb-GTPase.

Syngap1 inhibition reverses memory deficits in Mnk DKO mice

To further characterize the physiological significance of the identified link between Mnk1 and Syngap1, we investigated whether increased Syngap1 activity in Mnk DKO mice underlies memory deficits in these animals. To this end, we generated adeno-associated virus 9 (AAV9) expressing short-hairpin RNAs (shRNAs) against mouse Syngap1 mRNA, driven by the U6 promoter (AAV9-Syngap1-shRNA, Fig. 6A). We injected AAV9-Syngap1-shRNA or an AAV9-expressing scrambled sequence into the hippocampus of WT and Mnk DKO mice (Fig. 6A). AAV9-Syngap1-shRNA significantly decreased Syngap1 expression (∼50%; Fig. 6A) 2 and 4 weeks post-injection. We then subjected 8-week-old WT and Mnk DKO to the Morris water maze and contextual fear conditioning tests. Remarkably, decreasing Syngap1 expression with AAV9-Syngap1-shRNA completely reversed the spatial memory impairment (Fig. 6A and Supplementary Fig. 8A and B) and contextual fear memory deficits in Mnk DKO mice, without affecting WT animal behaviour (Fig. 6A). This result indicates that dysregulated (increased) Syngap1 expression, and thus activity downstream of Mnk, underlies memory deficits in Mnk DKO mice.

Pharmacological inhibition of Mnk with eFT508 corrects behavioural phenotypes in Syngap1^+/− mice

Pharmacological modulation of Mnks has recently emerged as a promising therapeutic avenue in cancer treatment,³⁷ ASD,^7,8 depression⁹ and neuropathic pain.³⁸ eFT508 is a brain-permeable, highly specific inhibitor of Mnks.³⁹ Because Mnk-mediated phosphorylation suppresses Syngap1 activity, we hypothesized that inhibition of Mnk might increase Syngap1 activity, restore mTORC1 signalling and thus rescue behavioural deficits in Syngap1^+/− mice. Thus, we injected Syngap1^+/− mice intraperitoneally with 1 mg/kg eFT08 daily for 5 consecutive days, a regimen that effectively reduces Mnk activity in vivo³⁸ (Fig. 6B and Supplementary Fig. 7E). Syngap1^+/− mice exhibit key autism-like phenotypes such as stereotypic behaviour, hyperactivity and social behaviour deficits.⁴⁰ First, we subjected Syngap1^+/− mice to a self-grooming test, where they displayed increased overall grooming time (Fig. 6B). Chronic eFT508 treatment completely reversed enhanced grooming in Syngap1^+/− mice, while it did not affect grooming behaviour in WT animals (Fig. 6B). Second, in the open-field test, vehicle-treated Syngap1^+/− mice spent significantly more time in the centre compared with WT, which is likely due to their hyperactivity (Fig. 6B). eFT508 treatment reversed the open-field phenotype of Syngap1^+/− mice, without affecting WT animals (Fig. 6B). Third, we carried out social behaviour analysis in eFT508-treated Syngap1^+/− mice using the three-chamber social interaction and preference for social novelty tests (Fig. 6B). Vehicle-treated Syngap1^+/− mice display impaired preference for social novelty but not sociability (Fig. 6B). eFT508 chronic treatment selectively rescued the preference for social novelty in Syngap1^+/− mice, without affecting WT mice behaviour (Fig. 6B).

Importantly, both AAV9-shRNA-mediated Syngap1 inhibition and chronic eFT508 treatment normalized mTORC1 signalling in both mouse models. In Mnk DKO mice, shRNA against Syngap1 but not scrambled-shRNA restored reduced phosphorylation of rpS6 (240/244), suggesting that mTORC1 activity was normalized to WT levels (Fig. 6C). In Syngap1^+/− mice, Mnk inhibition via chronic eFT508 treatment reduced enhanced phosphorylation of rpS6 (240/244) to WT levels (Fig. 6E). Altogether, pharmacological inhibition of Mnk kinase activity in Syngap1^+/− mice with eFT508 normalized exaggerated mTORC1 signalling in Syngap1^+/− mice and reversed autism-related behaviours. These data further validate the mechanistic link between Mnks, Syngap1 and mTORC1 and highlight its importance at the behavioural level (Fig. 6D and F).

Discussion

While there is extensive and definitive evidence that Mnk kinases regulate the phosphorylation and activity of proteins involved in diverse cellular functions, only a few substrates of Mnks have been identified in the brain, where their roles remain elusive. Accordingly, phosphorylation of eIF4E on Ser209 by Mnks can partially account for the plethora of Mnks-regulated functions in the brain. Our work reveals a hitherto unknown link between Mnk kinases and the genetic ASD risk factor Syngap1 in regulating mTORC1 signalling and protein synthesis. Pharmacological and AAV-mediated knockdown experiments further reveal that the Mnk–Syngap1 axis is crucial for ASD-linked behaviours (such as social interaction), learning and memory.

Translational control by the MAPK pathway is required for hippocampal synaptic plasticity, learning and memory⁴; thus, it was surprising that Eif4e^Ser209Ala mice exhibit intact hippocampal learning and memory, as well as L-LTP.⁹ Contrary to these findings, Mnk1/2 deletion in mice impairs learning and memory and synaptic plasticity (Fig. 1). Furthermore, pharmacological inhibition of Mnk1/2 reversed social behaviour deficits in Syngap1^+/− mice (Fig. 6) and restored social novelty in Nlgn3 KO mice.⁸ Taken together, our study highlights the role of Mnks in neurodevelopmental disorders such as ASD.

The common denominator between Mnks and syndromic ASD (Syngap1 or Nlgn3) and FXS is translational control. Herein, we showed that Mnk1/2 deletion in mice remodels whole brain and synaptic translatome and phosphoproteome (Figs 2–4). Importantly, there is a small overlap between Mnk DKO and Eif4e^Ser209Ala brain translatomes, chiefly composed of ECM-related genes. While this is in accordance with the importance of translational control of ECM genes in neurodevelopmental disorders such as FXS,^7,41 it further supports the notion that phosphorylation of eIF4E can only explain some aspects of Mnks in regulating brain functions. Interestingly, we detected pervasive changes in the synaptic translatome in Mnk DKO mice (Fig. 4), revealing a role for Mnks in regulating local translation, with wider implications for various synaptopathies. This finding may also be relevant to the emerging role for Mnks and eIF4E in depression via translational control of subsets of mRNAs.^9,42

We also observed significant remodelling of the synaptic phosphoproteome, whereby 892 phosphopeptides were differentially expressed in Mnk DKO (Fig. 3). In addition, we identified a consensus motif (‘SPEAK’) for Mnks, which is highly enriched in synaptic phosphoproteomics (Fig. 3). Because this motif is similar to motifs of other kinases (ERK, CDK5), it is plausible that Mnks may work in synergy with other kinases to phosphorylate a given subset of proteins. Such candidate kinases include CamKIIα and Cdk5, which are known to phosphorylate Syngap1.³⁵ Conceivably, Mnks could phosphorylate Syngap1 directly or in complex with CamKIIα/Cdk5, recruiting and phosphorylating Syngap1 on S788 (CamKIIα site) and S1165 (Cdk5 site). This is reminiscent of the mechanism by which Mnks phosphorylate eIF4E on Ser209, requiring first binding to eIF4G.⁴³ Mnk1 binding to Syngap1 (Fig. 5) at synaptic sites could lead directly or indirectly (via Cdk5/CamKIIα) to S788 phosphorylation, thus regulating local translation. Furthermore, expression of phosphor-mimetic Syngap1^S788D displays loss of function in inhibiting protein synthesis compared with WT Syngap1 (Fig. 5D). This suggests that Syngap1 S788 phosphorylation by Mnks is required to stimulate protein synthesis. In addition, we observed that mTORC1 is downregulated in Mnk DKO mice (Figs 5, 6 and see Barnes et al.¹¹ and Wang et al.¹³), while Akt and Erk phosphorylation was not altered (Supplementary Fig. 6). Syngap1 displays GAP activity for several small GTPases upstream of Akt, ERK, mTOR (e.g. Ras). In S788 phosphomutant lysates (but not phosphomimetic) and in Mnk DKO synaptosomes (where S788 phosphorylation was reduced) we measured increased Syngap1 GAP activity for Rheb, leading to inhibition of active Rheb (Fig. 5F–H). As Rheb is an upstream mTORC1 activator,⁴⁴ collectively these findings may explain reduced mTORC1 signalling in Mnk DKO, concomitant with reduced Syngap1 S788 phosphorylation. p-Akt and p-ERK could remain unaltered in Mnk DKO due to homeostatic regulation of different phosphosites on Syngap1. Undoubtedly, Syngap1 phosphosites work in synergy, engendering complex output to its downstream effectors (e.g. ERK and mTORC1 kinases)³⁵; thus, it will be interesting to systematically study the Syngap1 S788 phosphosite, in conjunction with Mnk1 in this signalling pathway.

mTORC1 signalling was previously linked to translational control of mitochondrial and ribosomal protein-coding mRNAs,^33,45 which was recapitulated in our ribosome profiling in Mnk DKO (Figs 2 and 4). mTORC1 and Mnks were implicated in neuropathic pain, via ras-related GTP-binding protein A (RagA)³³ and in cancer cells, via phosphatidyl inositol 3′ kinase-related kinase (PIKK) stabilizer telomere maintenance 2 (TELO2).⁴⁶ Notably, dual abrogation of Mnk/mTOR emerges as a promising therapeutic avenue for aggressive cancers.⁴⁷

Furthermore, we reveal that in addition to eIF4E, Mnks have additional downstream effectors in a ‘context-dependent manner’, which regulate mRNA translation. Plausibly, in synaptic sites, Mnk interaction with Syngap1 may exert local translational control via the Syngap1/Rheb/mTORC1 pathway, whereas in other sites Mnk1-mediated phosphorylation of eIF4E on Ser209 could have a more prominent yet mTORC1-independent effect on the translatome. Moreover, Mnk1 via eIF4E, Syngap1 or other as yet unknown mediators may have divergent effects on mRNA translation in different brain regions and thus regulate different behaviours (e.g. memory, affective behaviours).

In Syngap1 animal and cellular models, mTORC1 was implicated in regulating protein synthesis in excitatory cortical neurons via AMPA receptors.^11,13 This previous work, in conjunction with our rescue experiments performed in Mnk DKO and Syngap1^+/− mice (AAV knockdown and pharmacological inhibition; Fig. 6) further strengthens the Mnk–Syngap1 link to mTORC1 and protein synthesis. Our finding that eFT508 rescued autism-related behaviours in Syngap1^+/− mice, harbouring only one functional allele of Syngap1, presents an appealing therapeutic avenue for this genetic condition. All SYNGAP1 mutations (e.g. de novo, rare variants) linked to ASD are loss-of-function and cause SYNGAP1 haploinsufficiency leading to a defined intellectual disability and epilepsy phenotype, termed mental retardation-type 5 (MRD5, OMIM #612621).⁴⁸ Additionally, Mnk inhibitors could be used to reduce SYNGAP1 S788 phosphorylation in syndromic (e.g. FXS, tuberous sclerosis)⁴⁹ or sporadic ASD⁵⁰ associated with enhanced mTORC1 activity or protein synthesis.

In conclusion, by defining the Mnk phosphoproteome and translatome in whole brain and synaptosomes of Mnk DKO transgenic mice, we revealed a mechanistic link between Mnk and the syndromic autism risk protein Syngap1 (Fig. 7). This constitutes a previously unidentified mechanism of translational control in brain downstream of MAPK, with wider implications for synaptic plasticity, learning, memory and autism-related behaviours.

Figure 7.

Proposed mechanisms for the Mnk–Syngap1 interplay. Syngap1 phosphorylation by Mnk is a new translational control pathway downstream of MAPK, regulating memory and autism-related behaviours, while eIF4E phosphorylation is important for translation of mRNAs coding for extracellular matrix (ECM) proteins. Left: Syngap1 is an inhibitor of mTORC1, controlling protein synthesis and memory and autism-related behaviours. Mnks phosphorylate Syngap1 on S788 to reduce its GAP activity for Rheb and thus promote mTORC1 activity and protein synthesis, which are required for memory and autism-related behaviours. Right: Depletion of Mnks reduces Syngap1 S788 phosphorylation, increasing Syngap1 inhibitory GAP activity, reducing active Rheb and thus decreasing mTORC1 activity and protein synthesis, leading to memory impairment and autism-related behaviours.

Funding

This work was supported by grants to C.G.G.: start-up funds from the Foundation for Research and Technology-Hellas, the Hellenic Foundation for Research and Innovation (H.F.R.I.) under the ‘2nd Call for H.F.R.I. Research Projects to support Faculty Members & Researchers’ (Project Number: 2556), Sir Henry Dale Fellowship from the Wellcome Trust and Royal Society (107687/Z/15/Z), a NARSAD Young Investigator grant from the Brain & Behavior Research Foundation (24968). S.M.J. is supported by a Patrick G. Johnston Research Fellowship at Queen’s University Belfast.

Competing interests

The authors report no competing interests.

Supplementary material

Supplementary material is available at Brain online.