Grin2b 3’UTR is necessary for synaptic plasticity and spatial learning

Significance

Many studies have dissected NMDA receptor proteins in plasticity and memory, yet the untranslated regions (UTRs) of their mRNAs remain largely unexplored. We present a mouse model that harbors a deleted Grin2b 3’UTR—one of the three longest in mammals—in a key synaptic gene implicated in autism, intellectual disability, epilepsy, and schizophrenia. Despite an intact coding sequence, the deletion impairs spatial learning and synaptic plasticity. Using in situ hybridization, synaptosome isolation, electrophysiology, proteomics, and behavior, we show that the 3’UTR is essential for synaptic mRNA localization. Our findings reveal a previously overlooked layer of gene regulation that may underlie brain disorders and open an avenue for understanding how 3’UTRs control synaptic function and cognition.

Abstract

Spatially precise protein synthesis is critical for synaptic plasticity and cognitive function. The GRIN2B transcript encodes the GluN2B subunit of N-methyl-D-aspartate (NMDA) receptors and has a 25 kb-long 3’-untranslated region (3’UTR) of unknown function. To investigate its role, we generated a ∆3’UTR mouse line by deleting the Grin2b 3’UTR while preserving the coding sequence. Despite unchanged Grin2b mRNA levels in ∆3’UTR mice, GluN2B protein was reduced 50%. In wildtype (WT) mice, Grin2b mRNA was enriched in synaptosomes, but this enrichment was impaired in ∆3’UTR mice along with reduced GluN2B phosphorylation. Notably, ∆3’UTR mice had no long-term potentiation (LTP) and exhibited hippocampal-dependent spatial learning impairments. Together, these findings demonstrate that the 3’UTR is critical for Grin2b mRNA function, ultimately influencing synaptic composition and plasticity.

Neuronal function is driven by the balance between excitation and inhibition of signals within the interactive cell network, which is maintained by multitiered gene regulation.

N-methyl-D-aspartate (NMDA) receptors are crucial players in synaptic signaling as a subclass of ionotropic glutamate receptors that react to the release of glutamate from the excitatory presynapse by opening of a cation channel. First, activation of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors allows influx of sodium, and if the membrane becomes sufficiently depolarized, the NMDA receptors may also open. NMDA receptors are calcium-permeable, and their activation therefore leads to a cascade of intracellular signaling events, which are critical in neuronal development (1), neuronal differentiation (2), synaptogenesis (3), and synaptic plasticity (4). NMDA receptor dysregulation is also observed in neurodegenerative disease (5), psychiatric disorders (6), and neurodevelopmental disorders (7). NMDA receptors are heterotetrameric receptors comprised of two obligatory GluN1 subunits (encoded by Grin1) and two other subunits, GluN2 or GluN3 (encoded by Grin2A-D, 3A-B). The GluN2 and GluN3 subunits have distinct kinetic properties and regional expression profiles (8). Grin2b is largely expressed in the cerebral cortex and hippocampus and is strongly implicated in synaptic plasticity and neuronal circuitry formation (9), which are understood to be the molecular basis of learning and memory (10).

Variants in the protein coding region of GRIN2B are observed in neurological diseases (11–13), and various genetic knock-in/knock-out studies of Grin2b in mice have been performed (9, 14–17). While knockouts of, e.g., Grin2a, c and d are viable (18, 19), homozygote coding region Grin2b^−/− mice die shortly after birth due to impaired suckling response (14). Conditional knockout, heterozygote, and mutation models have been used to investigate the effects of Grin2b removal. For example, a knock-in mutation model (Grin2b^{Y1472F/ Y1472F}) investigated the importance of Tyr-1472 phosphorylation, revealing mutant mice to exhibit impaired fear-learning and reduced LTP in the amygdala (15). A conditional knock-out of Grin2b^−/− in the forebrain or hippocampus shows impaired spatial learning and hippocampal LTP (16). Interestingly, a heterozygote mode (Grin2b^+/C456Y) showed 50% Grin2b expression with reduced hippocampal LTD but did not show altered LTP or spatial learning (9). Although these models reveal alterations in plasticity mechanisms, protein levels alone are not sufficient to explain the observed deficits, suggesting that a more nuanced view of NMDAR regulation is required to fully elucidate these pathways.

These Grin2b mouse studies have primarily focused on characterizing how GluN2B abundance or coding sequence mutations affect receptor activity and/or its protein interactions within the postsynaptic density (PSD) and the effects on neuronal physiology or behavior. Less attention has been given to the posttranscriptional regulation of Grin2b, such as mRNA transport, mRNA stabilization, and translational efficiency.

Growing evidence suggests that local translation occurs in neurons (20–22) and is crucial for activity-dependent neuronal protein synthesis as an adaption to the unique cellular structure of neurons (23–25). The localization to neuronal processes of transcripts has been observed to be preferential for those with longer 3’ untranslated regions (UTRs) (26). RNA-binding proteins (RBPs) interact with regulatory motifs, such as zip code regions (27), in the 3’UTR of several genes, such as Arc/Arg3.1 (24), Camk2α (28), and Map2 (29), to mediate mRNA localization (26), stabilization (26), and translational efficiency (30).

Human 3’UTRs are typically 1 kilobase (kb) long (31), while highly expressed neuronal genes contain an average 1.3 kb 3’UTR (31). That of GRIN2B is 25 kb (32)—19 kb in mice—which is among the longest of any mammalian transcript (33), but little is known about its function. Of the NMDA receptor genes, only that encoding GluN1 is known to be alternatively spliced, for the others, each gene has a single dominant transcript variant. Interestingly, although glutamate signaling affects local translation in the dendritic spines of Actb (25) and Grin2a (34), total Grin2b levels are unaffected (34).

To investigate the role of the unusually long Grin2b 3’UTR, we generated a mouse model that harbors a 19 kb deletion of the Grin2b 3’UTR (∆3’UTR). Using this model, we examine how the 3’UTR regulates Grin2b mRNA localization and translation, evaluate synaptic function, and assess behavior to define the impact of Grin2b 3’UTR disruption.

Methods

Mouse Model and Animal Care.

The Grin2b^TM1EMFU allele was generated by the Danish Genetically Modified Animal Resource (DAGMAR) facility at Aarhus University using CRISPR/Cas9 and two different guide RNAs (gRNAs) designed with the ChopChop tool (35). Nine different gRNA combinations (three upstream and three downstream gRNAs) were first tested by lipofection into murine 129sv embryonic stem cells (36). Cutting efficiency was determined from harvested DNA by PCR, showing deletions for four of the nine combinations.

C57BL/6JRj female mice purchased from Janvier-Labs, France, were superovulated and mated with C57Bl/6JRj male mice. Zygotes were harvested at E0.5 at the pronucleus stage. For electroporation of mouse zygotes (37), we chose gRNA1 that directs cut between bases 1917 and 1918 of exon 14, and gRNA2 between 20543 and 20544 (NM_008171.4). Final gRNA concentration was 2.5 µM for each with a Cas9 protein concentration of 4.2 µM (Integrated DNA Technologies Coralville, IA).

Primer information for gRNAs and internal primers can be found in SI Appendix. Mice successfully harboring at least one allele containing the deletion were back crossed with wildtype (WT) mice, and subsequent heterozygote pups were selected for breeding pairs.

Generation, breeding, and experiments were approved by the Animal Testing Authority—Danish Ministry of Food, Agriculture, and Fisheries (License #2020-15-0201-00463, #2022-15-0201-01154, & #2024-15-0201-01686), and all experiments were conducted in accordance with the Danish Animal Experimentation Act and Danish Animal Welfare Act under EU Directive. Animals were housed up to 9 per cage under 12 h:12 h light:dark cycles at a temperature of 20 to 26 °C at 40 to 60% humidity with food and water ad libitum. For behavioral tests, all mice underwent testing between 12 to 18 wk old. Behavioral tests consisted of mixed genotype and sex with cage-mate WT controls.

RNAscope In Situ Hybridization and Immunofluorescence.

Fluorescent in situ hybridization (FISH) was performed using RNAscope™ Multiplex Fluorescent Detection Kit V2 (Cat. no. 323110, Advanced Cell Diagnostics), as per the manufacturer’s instructions. Primary hippocampal cells were rehydrated sequentially, blocked with 3% H₂O₂ for 10 min at RT, and permeabilized with protease III (diluted 1:15) for 10 min at RT. Samples were incubated at 40 °C with Mm-Grin2b RNAscope™ probe (Cat. no. 417391-C3, Advanced Cell Diagnostics), AMP1, AMP2, and AMP3 reagents to build the amplification complex, followed by incubation with horseradish peroxidase (HRP) conjugated to the channel-specific probe. Signal development was achieved using the Opal 690 fluorophore (Cat. no. SKU FP1497001KT, Akoya Biosciences), and the reaction was concluded with an HRP blocker to prevent cross-reactivity.

Following in situ hybridization, neurons were processed for immunofluorescence. Briefly, cultures were permeabilized with 0.1% Triton-100X in PBS and blocked with 10% normal goat serum (NGS) in 1xPBS for 1 h at RT, posthybridized neuron coverslips were incubated with chicken anti-Map2 (1:5,000, Abcam, Ab5392) in 2% NGS in 1xPBS overnight at 4 °C. After three washes with PBS, the slides were incubated with goat anti-chicken Alexa Fluor® 488 secondary antibody (1:1000; Invitrogen A-11039) for 2 h at RT. Coverslips were rinsed with PBS three times and mounted on SuperFrost^TM Plus microscope slides (Cat. no. 22-037-246, Fisher Scientific) with Prolong Diamond Antifade Mountant with DAPI (Cat. no. P36966, Invitrogen).

Electrophysiology.

Mice (LTD P13-23; LTP, input–output and paired pulse facilitation P30-40) were deeply anesthetized with 4% isoflurane and decapitated. The brain was rapidly extracted and transferred to ice-cold artificial cerebrospinal fluid (aCSF) continuously bubbled with 95% O₂ and 5% CO₂. Coronal brain slices (400 µm) were prepared using a vibratome (Vibratome 3000 Sectioning System). For Schaffer collateral-Cornu Ammonis 1 (SC-CA1) circuit recordings, slices were transferred to a storage chamber containing aCSF (126 mM NaCl, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 26 mM NaHCO₃, 2.5 mM CaCl₂, 1.3 mM MgCl₂, 10 mM D-glucose) at RT, where they recovered for at least 1.5 h before recording. Slices were transferred to an interface recording chamber perfused with aCSF. Perfusion rate was maintained at 2 ml/min, and the temperature was kept stable at 33 °C.

Field excitatory postsynaptic potentials (fEPSPs) were recorded in the SC-CA1 circuit of the hippocampus using a MultiClamp 700B amplifier (Molecular Devices) and a glass microelectrode (7 to 14 MΩ) filled with aCSF. fEPSPs were evoked by stimulating the Schaffer collaterals (SC) with a tungsten concentric bipolar electrode (FHC) using 0.1 ms pulses delivered every 15 s. Stimulation intensity was adjusted to evoke fEPSPs at 40 to 50% of the maximum response based on the input–output curve. Paired-pulse facilitation (PPF) was assessed at interstimulus intervals of 20 to 500 ms, and the paired-pulse ratio (PPR) was calculated as the second-to-first fEPSP slope ratio. Input–output curves were generated by increasing stimulus intensity from 5 to 40% of 1 mA in 5% steps, and synaptic strength was quantified by fEPSP slope. NMDA receptor-mediated responses were isolated using aCSF with 0.1 mM MgCl₂ and step-wise stimulation (40% of 1 mA); after a 10-min baseline, DNQX (10 µM) was applied to block AMPA receptors. The NMDA component was measured as the slope of the isolated response.

LTP was induced after stable baseline recordings for at least 20 min. Two protocols were used: 1 × 100 Hz stimulation for 1 s and 4 × 100 Hz stimulation for 1 s each. LTP magnitude was measured as the percentage increase in fEPSP slope relative to the 20-min baseline. fEPSPs were analyzed using Clampfit 10 (Molecular Devices).

For LTD, a stable baseline was obtained, then LTD was induced using low-frequency pulses applied at a rate of 1 Hz for 18 min as previously described (38).

Barnes Maze and Morris Water Maze.

To test spatial learning and memory, we used the Morris Water maze (39) and Barnes maze (40).The Morris Water maze was conducted using a circular pool 120 cm in diameter. A hidden platform (10 cm in diameter) was submerged 1 cm beneath the opaque water surface, and four distal visual clues were used to enhance visuospatial learning and memory. The spatial acquisition phase consisted of two 90 s trials per day for 4 consecutive days. A probe trial was conducted at 72 h following the acquisition phase.

Barnes maze was performed using a white circular apparatus 122 cm in diameter and 24 evenly spaced holes around its perimeter. The escape hole was randomly selected, and four distal visual clues were used to enhance visuospatial learning and memory. The spatial acquisition phase consisted of two 180 s trials per day for 4 consecutive days. A probe trial was conducted 72 h following the acquisition phase. A bright spotlight was placed overhead as an aversive stimulus, encouraging mice to locate the escape box as quickly as possible. In both assays, mice were tracked using AnyMaze tracking system (Stoelting Co. version 7.2).

Results

3’UTR deletion decreases GluN2B protein expression with no change in Grin2b mRNA expression.

In both humans and mice, the Grin2b gene contains a long 3’-UTR of unknown function (Fig. 1A). To explore its role, we used CRISPR/Cas9 technology to generate a mouse line (hereon referred to as ∆3’UTR mice) that harbors a ~19 kb deletion of the Grin2b 3’UTR, leaving the protein-coding region intact (Fig. 1B). Successful deletion of the 3’UTR was confirmed by RNA-seq, with coverage maps indicating no mapped reads to the deleted 3’UTR segment in ∆3’UTR mice (Fig. 1C), as further supported by PCR (SI Appendix, Fig. S1A).

Fig. 1.

CRISPR-mediated deletion of Grin2b 3’UTR reduces GluN2B protein and impairs Grin2b translational efficiency. (A) Grin2b mRNA exon map from Mus musculus and Homo Sapiens highlighting the conservation of exon 14 containing the full 3’UTR length. (B) Schematic of CRISPR targeting strategy to delete the Grin2b 3’UTR within exon 14. Homozygous deletion (∆3’UTR) results in mRNA lacking the full 3’UTR. (C) RNA-seq coverage map from adult hippocampus showing absence of reads across the deleted 3’UTR region in ∆3’UTR mice. (D) RT-qPCR analysis showing no change in Grin2b mRNA levels in 14-wk-old adult male hippocampus (n = 6/group; unpaired two-tailed t test). (E and F) Western blot (E) and semiquant densitometry (F) reveal reduced GluN2B protein expression in ∆3’UTR hippocampus (n = 4 male/group, n = 3 female/group; ^***P = 0.0001, unpaired two-tailed t test). (G) RT-qPCR analysis from polysome fractionation showing no change in total homogenate Grin2b mRNA with significantly reduced Grin2b in the ∆3’UTR monosome and polysome fractions (n = 5/group; ^*P > 0.0356, Two-way ANOVA with Tukey post hoc).

RT-qPCR analysis revealed no significant change in adult hippocampal Grin2b mRNA expression (Fig. 1D), and no transcriptional changes at intron–exon spanning regions, indicating no gross effects in transcription of the transcript (SI Appendix, Fig. S1B). Nonetheless, western blot analysis of adult hippocampal tissue revealed a ~50% decrease in GluN2B protein expression (Fig. 1 E and F; P = 0.0001) with no change in hippocampal GluN1 expression (SI Appendix, Fig. S1C). To further investigate the effect of ∆3’UTR on translational efficiency, we utilized polysome fractionation followed by RT-qPCR to assess the levels of Grin2b in different unbound- and ribosome-bound lysate fractions (Fig. 1G). This revealed a significant effect across both ribosome fraction (P < 0.0001) and genotype (P = 0.0402). We observed a significant decrease in Grin2b mRNA in both monosome (P = 0.0356) and polysome lysate fractions (P = 0.0226). These data suggest the 3’UTR is important for ribosome binding in efficient translation.

3’UTR Deletion Disrupts mRNA Localization and Synaptic Protein Expression in the Hippocampus.

To investigate molecular factors that underlie the hippocampal-driven behavioral and physiological impairment observed in our ∆3’UTR mice, we focused on how this deletion affects posttranscriptional regulation of Grin2b. We first reanalyzed transcriptomic data from hippocampal synaptosomes of 4-wk-old mice (41) (Fig. 2A) and found high Grin2b expression (2.12 log₁₀FPKM; top 13%), similar to other known dendritically localized transcripts (23, 24, 28) (1.73 to 3.03 log₁₀FPKM), and above the median expression (1.61 log₁₀FPKM) (Fig. 2B).

Fig. 2.

Loss of 3’UTR reduces Grin2b localization and synaptic P-GluN2B expression. (A) Schematic depicting synaptosome isolation workflow. (B) Normalized gene counts from RNA-sequenced Percoll gradient-derived synaptosomes (42) show Grin2b among previously identified localized transcripts. (C) RT-qPCR of synaptosomes shows Grin2b enrichment in WT synaptosomes, while this enrichment is lost in ∆3’UTR synaptosomes (n = 5/group; ^**P < 0.0054, Two-way ANOVA with Fisher’s LSD test). (D) Representative maximum intensity projection images of RNAscope in situ hybridization performed on DIV 14 primary cultured hippocampal neurons (n = 30 WT neurons, n = 28 ∆3’UTR neurons). Grin2b single-channel images in gray scale and merged images in color. Scale bars denote 50 µm and 10 µm in overall images and zoomed-in images, respectively. (E) Distribution of Euclidean distance from the nucleus of all puncta detected across all imaged neurons for each group shows a significant reduction in the overall and maximum distance from soma in ∆3’UTR neurons (^****P < 0.0001, Mann–Whitney two-tailed t test). (F) Average Euclidean distance of all Grin2b puncta within each imaged neuron shows significant reduction in average Grin2b distance from soma in ∆3’UTR neurons (n = 30 WT, n = 28 ∆3’UTR; ^*P = 0.028, Mann–Whitney two-tailed t test). (G) Western blot detection of GluN2B and P-GluN2B (Y1472) from synaptosomes and total tissue homogenate tissue (n = 3 males/group, n = 3 females/group). (H) Semiquant densitometry of GluN2B western blot detection densitometry reveals significant enrichment of GluN2B within WT synaptosomes, while this trend is not observed in ∆3’UTR synaptosomes (^**P = 0.0041, Two-way ANOVA with Fisher’s LSD test). (I) Semiquant densitometry of P-GluN2B (Y1472) western blot detection densitometry displays increased phosphorylated GluN2B within WT synaptosomes, while 3’UTR deletion abolishes this trend (^**P = 0.0058, Two-way ANOVA with Fisher’s LSD test).

We next isolated synaptosomes from hippocampal tissue and observed a significant increase in Grin2b mRNA expression in WT synaptosomes compared to total homogenate (Fig. 2C; P = 0.0054; SI Appendix, Fig. S2; P < 0.0001). This trend was not retained in ∆3’UTR synaptosomes, showing a significant effect of genotype (P = 0.0294) and subcellular fraction (P = 0.0078).

We next visualized Grin2b using fluorescent in situ hybridization in cultured hippocampal neurons (14 DIV). WT neurons showed strong somatic signal and puncta in distal processes (Fig. 2D), while ∆3’UTR neurons displayed fewer, more proximal puncta (Fig. 2 E and F). Puncta volume was significantly increased (P < 0.0001) in ∆3’UTR neurons (SI Appendix, Fig. S3). These findings indicate that the long 3’UTR promotes distal transport of Grin2b transcripts within neurons.

In evaluating local protein expression, western blot of synaptosomes from hippocampal tissue of WT and ∆3’UTR mice revealed an increased relative GluN2B protein level in WT synaptosomes compared to that in homogenates (P = 0.0041), a pattern not observed in synaptosomes isolated from ∆3’UTR mice (Fig. 2 G and H). Similarly, the synaptic phosphorylation marker P-GluN2B-Y1472 (43) was enriched in WT synaptosomes (P = 0.0058) but unchanged in ∆3’UTR samples (Fig. 2 G and I). Broader insights from RNA-seq of total hippocampus suggest downregulated genes are associated with translation in the cytoplasm and at the synapse (SI Appendix, Fig. S5). Together, these data suggest that the Grin2b 3’UTR can promote synaptic mRNA localization and support synaptic accumulation of GluN2B protein.

To identify proteins that bind the Grin2b 3’UTR, we performed an RNA–protein pull-down, using two regions of interest: Stop codon-adjacent region (SAR) and PolyA-adjacent region (PAR), covering in total ~20% of the 3’UTR (SI Appendix, Fig. S6). To compare our dataset with a prediction tool, we observed 14 proteins within the RBPmap trained model in our pulldown hits (SI Appendix), including well-known RBP top hits TARDBP, ELAVL1, CELF1, hnRNPH2. Predicted motif analysis revealed high-confidence binding sites for these pulldown hits. Hits in the SAR were more densely clustered between positions 6,250 to 6,800, approximately 1 kb downstream of the stop codon, while PAR hits were more densely clustered between 23,500 to 24,123 around the polyA site (SI Appendix, Fig. S7). Mass spectrometry revealed distinct protein clusters in each region and subcellular fraction (SI Appendix, Fig. S8-S9). Gene ontology confirmed enrichment for relevant categories, including “RNA binding” and “purine-rich negative regulatory element binding.” InterPro (44) protein domain family composition was broadly similar across all groups, dominated by RNA-binding, RRM, and ribosomal protein families (SI Appendix). Protein expression, obtained from the Allen Brain Atlas (45), across all four groups was consistently enriched in cortical subplate, hippocampal formation, and isocortex (SI Appendix). Mean protein abundance was highest in the cytosolic PAR pulldown group, which was enriched with RNA-stabilizing factors identified by GO analysis. Contrarily, the synaptosomal PAR proteins had the lowest mean abundance among the four groups. In the cytosol, SAR was enriched for proteins involved in RNA processing, while synaptosomal SAR proteins were associated with translation (SI Appendix, Fig. S9). Overall, the synaptosome proteins were more associated with translation, while cytosolic proteins were associated with mRNA processing and splicing (SI Appendix, Fig. S9). These findings suggest region- and -specific interactions direct Grin2b mRNA toward distinct posttranscriptional fates.

Together, these results highlight a critical role for the long Grin2b 3’UTR in supporting synaptic mRNA localization and synaptic protein composition, providing molecular insight into the role of the Grin2b 3’UTR.

Impaired Synaptic Plasticity and Increased Glutamate Release Probability in the SC-CA1 Circuit of ∆3’UTR mice.

To assess synaptic plasticity in the Schaffer collateral-CA1 circuit, we performed electrophysiological recordings in acute hippocampal slices (Fig. 3A). PPF was significantly reduced in ∆3’UTR mice (Fig. 3B; P = 0.0206), showing a significant effect in both interstimulus interval (P < 0.0001) and genotype (P < 0.0001), indicating an increased probability of glutamate release from presynaptic terminals. Correspondingly, ∆3’UTR mice had increased basal synaptic transmission (Fig. 3C; P = 0.0002), which showed significant effects in current (P < 0.0001) and genotype (P < 0.0001). Post hoc analysis revealed heightened excitatory postsynaptic potential (EPSP) in ∆3’UTR mice at stimulation intensities 30 to 40 µA compared to WT (P < 0.0001), suggesting increased baseline synaptic activity. We observed no gross changes in total NMDA receptor currents (Fig. 3D), which, given the increase in glutamate release in the ∆3’UTR mice, may indicate that they have fewer NMDA receptors postsynaptically. Strikingly, an LTP protocol of 1 x 100 Hz stimulation gave LTP in WT but caused no strengthening of the synaptic connections in ∆3’UTR mice (Fig. 3E), showing significantly increased fEPSP in WT compared to WT baseline (Fig. 3H; P = 0.0004) and to ∆3’UTR LTP (P = 0.0010). A stronger protocol of 4 x 100 Hz induced robust LTP in WT slices, but no LTP in the ∆3’UTR, although a transient potentiation was induced in the first minutes (Fig. 3F), ultimately displaying a significant increase in fEPSP in WT LTP compared to WT baseline (Fig. 3I; P = 0.0028) and to ∆3’UTR LTP (P = 0.0008). LTD was also reduced in ∆3’UTR slices compared to WT controls (Fig. 3G). While we did observe a significant sustained effect of LTD in ∆3’UTR compared to ∆3’UTR baseline (Fig. 3J; −9% fEPSP slope; P = 0.0143), it was significantly less of an effect compared to that in WT LTD (−27%; P = 0.0001). This indicates that, while LTD is still marginally present in ∆3’UTR mice, it is significantly impaired compared to WT controls.

Fig. 3.

3’UTR deletion impairs bidirectional hippocampal plasticity and alters synaptic homeostasis. (A) Recording workflow schematic of coronal mouse hippocampal slice. Stimulating electrode placed in the Schaffer Collaterals (SC) transmitting to the recording electrode in the Cornu Ammonis 1 (CA1). (B) Paired-pulse ratios recorded by delivering two stimuli at interstimulus intervals of 20, 50, 100, 200, and 500 ms. The paired-pulse ratio was calculated as the amplitude ratio of the second to the first fEPSP. Data show decreased PPR in ∆3’UTR slices (n = 6 WT, n = 5 ∆3’UTR, *P = 0.0329, ***P = 0.0003, Two-way ANOVA with Šídák’s post hoc). (C) Input–output curves representing basal AMPA receptor-mediated synaptic transmission show a ~50% increase in ∆3’UTR mice (n = 11 WT, n = 12 ∆3’UTR, *P = 0.0412, **P = 0.0023, ****P < 0.0001, Two-way ANOVA with Šídák’s post hoc). (D) Total NMDA receptor-mediated transmission displays no significant change between genotypes. (n = 11 WT, n = 11 ∆3’UTR, Two-way ANOVA with Šídák’s post hoc). (E) LTP recordings in hippocampal slice elicited from 1 × 100 Hz 1 s stimulation show no LTP induction in ∆3’UTR mice with significant retention of synaptic strength in WT mice. (n = 5 WT, n = 6 ∆3’UTR). (F) LTP recordings in hippocampal slice elicited from 4 × 100 Hz 1 s stimulation present successful LTP induction in both groups but show markedly reduced synaptic strength retained in ∆3’UTR mice. (n = 2 WT, n = 2 ∆3’UTR). (G) LTD recordings in hippocampal slice elicited from 1 Hz stimulation. Successful LTD induction in both genotypes, with WT slices showing a greater magnitude of sustained depression compared to ∆3’UTR (n = 4 WT, n = 6 ∆3’UTR). (H) Average fEPSP of final 5 min of baseline and 1 × 100 Hz LTP recordings, displaying a significantly sustained LTP in WT compared to ∆3’UTR (n = 5 WT, n = 6 ∆3’UTR, ***P < 0.0010, Two-way ANOVA with Fisher’s LSD test). (I) Average fEPSP of final 5 min of baseline and 4 × 100 Hz LTP recordings, displaying a significantly sustained LTP in WT compared to ∆3’UTR (n = 2 WT, n = 2 ∆3’UTR, **P < 0.0028, ***P < 0.0008, Two-way ANOVA with Fisher’s LSD test). (J) Average fEPSP of final 5 min of baseline and 1 Hz LTD recordings, displaying a significantly decreased fEPSP slope in both genotypes with a larger magnitude of depression in WT compared to ∆3’UTR (n = 4 WT, n = 6 ∆3’UTR, *P < 0.0143, ****P < 0.0001, Two-way ANOVA with Fisher’s LSD test).

Together, these findings reveal that deletion of the Grin2b 3′UTR disrupts both LTP and LTD induction in the hippocampus, indicating that the 3′UTR is essential for maintaining synaptic integrity and plasticity. This likely occurs through loss of properly localized GluN2B-dependent signaling, which narrows the dynamic range of synaptic modification and provides a mechanistic link between disrupted molecular targeting and impaired synaptic physiology.

Adult ∆3’UTR Mice Display Deficits in Spatial Learning and Memory Tasks.

To evaluate hippocampal-dependent spatial learning and memory, we tested adult mice in the Barnes maze and Morris Water Maze, both of which rely on involvement of the basal forebrain, hippocampus, and prefrontal cortex (16, 46, 47). In the Morris Water Maze, ∆3’UTR mice displayed deficits during the acquisition phase (Fig. 4A), with significant effect of genotype (P < 0.0001) and time (P < 0.0001). Post hoc analysis showed an increased latency for ∆3’UTR mice to find the hidden platform on each day of the acquisition phase (P < 0.0351). To test spatial memory following the acquisition phase, a probe test where the platform was removed 72 h after the final training was performed. ∆3’UTR mice spent significantly less time in the target quadrant (P < 0.0001) and platform location (P = 0.0059) compared to WT (Fig. 4 B and C). There were no differences between the groups in total distance traveled in the Morris Water Maze during the probe trial (SI Appendix, Fig. S9A).

Fig. 4.

∆3’UTR mice exhibit spatial learning deficits. (A) Latency to platform is increased in ∆3’UTR mice during the acquisition phase of the Morris Water Maze (MWM). Data from two trials per day on 4 consecutive days. Data are shown as mean ± SEM (n = 6 WT Male, n = 6 ∆3’UTR Male, n = 6 WT Female, n = 5 ∆3’UTR Female; ^*P = 0.0351, ^****P < 0.0001, Two-way ANOVA with Tukey post hoc). (B) Percent time spent in target quadrant is decreased in ∆3’UTR mice in 72 h probe trial (^*P = 0.0332, ^**P = 0.0021, ^****P < 0.0001, Two-way ANOVA with Tukey post hoc). (C) Percent time spent in platform location during 72-h probe trial (^**P = 0.0059, Mann–Whitney two-tailed t test). (D) Latency to escape hole is increased in ∆3’UTR mice during the acquisition phase of the Barnes Maze. Data from two trials per day on 4 consecutive days. Data are shown as mean ± SEM (n = 9 WT Male, n = 8 ∆3’UTR Male, n = 7 WT Female, n = 5 ∆3’UTR Female; ^****P < 0.0001, Two-way ANOVA with Tukey post hoc). (E) Percent time spent in target quadrant is decreased in ∆3’UTR mice in 72 h probe trial (^**P = 0.0012, ^***P = 0.0006, Two-way ANOVA with Tukey post hoc). (F) Percent time spent in escape hole location during 72-h probe trial (^***P = 0.0009, Mann–Whitney two-tailed t test). Dashed line in all plots represents the percentage of area crossings expected by random chance, based on the ratio of the target area to the total maze size.

In the Barnes Maze, ∆3’UTR mice showed significant deficits during the acquisition phase, with increased latency to the escape hole on training days 2 to 4 compared to WT animals (Fig. 4D; P < 0.0001), exhibiting significant effects in both genotype (P < 0.0001) and time (P < 0.0001). 72-hours after the final acquisition day, the escape hole was removed, and a trial was conducted to test spatial memory. During the probe trial, ∆3’UTR mice displayed a significantly reduced time spent in the escape hole zone (P = 0.0009) and significantly less time in the target quadrant area (P = 0.0004) compared to WT mice (Fig. 4 E and F). ∆3’UTR mice did exhibit reduced total distance traveled during the probe day (SI Appendix, Fig. S9C; P < 0.0001). To assess the potential for altered motor coordination, we evaluated mice using an accelerating rotarod protocol. Both WT and ∆3’UTR mice consolidated their performance during 4 d of continuous training, with no observable difference between the two groups. In addition, no difference was seen on the accelerating trial day (4 to 40 rpm), indicating that the motor abilities are not affected in ∆3’UTR mice (SI Appendix, Fig. S9 E and F). These behavioral impairments underscore the functional significance of the Grin2b 3’UTR in supporting hippocampal-dependent learning and memory, translating the observed molecular and synaptic disruptions into measurable cognitive deficits.

Discussion

Using a Grin2b 3’UTR deletion model, we find disrupted mRNA localization, reduced phosphorylated GluN2B synaptic expression, impaired LTP, and deficits in hippocampal-dependent learning. These findings highlight the impactful role of the Grin2b 3’UTR for synaptic plasticity.

Loss of Grin2b 3’UTR Reproduces GluN2B Knockout Features Without Coding Sequence Disruption.

Deleting the Grin2b 3’UTR, while preserving the coding sequence, reproduces key features of GluN2B knockout (KO) models. ∆3’UTR mice exhibit impaired spatial learning (14, 16), disrupted hippocampal LTP (48), and decreased levels of synaptic phosphorylated GluN2B at Y1472 (49)—a site important for stabilizing NMDA receptors and preventing endocytosis (50).

Unlike homozygous GluN2B KOs, which are neonatally lethal (14), our homozygous ∆3’UTR mice are viable, though they display a ~50% reduction in GluN2B protein in hippocampal homogenates (Fig. 1), along with reduced P-GluN2B in synaptosomes (Fig. 2). Given that phosphorylated GluN2B has been linked to synaptic strength, learning, and memory (51, 52), these molecular changes provide a strong foundation for the behavioral and molecular impairments observed in our model.

However, reduced GluN2B levels alone do not fully account for the observed phenotypes. Previous studies using Grin2b haploinsufficiency models—which similarly show ~50% protein reduction—do not report spatial learning or LTP deficits (9). This distinction suggests that reduced protein may be important but not sufficient alone to drive dysfunction. Instead, loss of 3′UTR-mediated posttranscriptional regulation likely produces broader disturbances in synaptic signaling.

In addition to reduced synaptic GluN2B and P-GluN2B, ∆3′UTR slices showed evidence of increased presynaptic glutamate release (Fig. 3), which is not necessarily expected given that NMDARs are typically recognized at the postsynapse. Several mechanisms may account for this presynaptic alteration. First, NMDA receptors have also been observed at the presynapse and can directly facilitate release probability by enhancing Ca²⁺ influx, and altered localization of these receptors has been shown to modify presynaptic strength (53, 54). Second, reduced postsynaptic GluN2B function may instigate a circuit-level feedback loop. This has been observed in interneurons, where NMDA hypofunction disrupts activity, causing disinhibition of excitatory synapses and increased glutamate release (55). Similar increases in ambient glutamate have been observed in models with impaired glutamate uptake, such as CNS TGF-β1 knockout mice, which show elevated spillover, enhanced activation of extrasynaptic GluN2B receptors, and impaired synaptic plasticity (56). The ∆3′UTR mutation likely perturbs Grin2b regulation in a way that extends beyond simple dosage reduction, producing a coordinated shift in glutamate release, synaptic receptor composition, and ultimately plasticity. This supports the view that our ∆3′UTR model represents a nuanced disruption of posttranscriptional regulation rather than a mere decrease in protein expression. While we did not directly assess AMPAR-mediated evoked postsynaptic currents (EPSCs) in this study, such measurements could provide additional insight into potential changes in synapse number or refinement associated with altered Grin2b 3′UTR regulation. Future studies could address this question to further explore the synapse-level mechanisms underlying the observed LTP and LTD deficits.

Further, behavioral analysis showed that there were spatial learning deficits evident in both the Morris Water Maze and Barnes Maze (Fig. 4). ∆3’UTR mice showed reduced exploration during the 72-h Barnes Maze probe (SI Appendix, Fig. S10), which could indicate diminished motivation. While prior studies have linked GluN2B expression to exploratory behavior (9, 42), the lack of movement reduction in the Morris Water Maze suggests that ∆3’UTR mice retain sufficient motivation under aversive conditions, pointing instead to a cognitive basis for the spatial learning deficits. Consistent with this interpretation, these behavioral impairments were accompanied by impaired LTP and LTD (Fig. 3). Because GluN2B-containing NMDARs are critical in shaping amplitude and duration of Ca²⁺ signaling required for bidirectional plasticity (10, 18), the combined reduction in LTP and LTD suggests a broader deficit in synaptic flexibility. These impairments align with altered glutamate dynamics and disrupted NMDAR signaling described above and provide a physiological corroboration of the cognitive deficits observed in this model.

Grin2b Is Regulated Posttranscriptionally to Mediate Translational Efficiency and Transport Distal to the Soma.

Our data suggest that Grin2b is present in WT hippocampal synaptosomes, which correlates with many other previously described synaptic transcripts, including Arc/Arg3.1 (24), Camk2α (28), and Map2 (29) (Fig. 2). Grin2b presence in distal neuronal compartments, which fits the general observation that longer transcripts are observed to preferentially localize to distal compartments, especially in neurons (30).

Deletion of the Grin2b 3′UTR led to a significant decrease in Grin2b mRNA abundance within synaptosomes and reduced average transcript distance from the soma (Fig. 2), suggesting that the 3′UTR contributes to proper dendritic localization. Similar imaging-based analyses of mRNA distribution have shown that reduced dendritic localization of transcripts like Map2 and Camk2α often corresponds to loss of localization signals in the 3’UTR (28, 29). Prior studies have demonstrated that 3’UTRs direct mRNA transport to dendrites and synapses by acting as scaffolds for RBP interaction and motor-protein recruitment (27, 57).

Consistent with a role in translational regulation, polysome profiling revealed significantly reduced Grin2b mRNA in monosome and polysome fractions of ∆3′UTR mice, despite unchanged total homogenate levels (Fig. 1). This pattern indicates diminished translational efficiency rather than altered transcript stability, suggesting that the Grin2b 3′UTR facilitates ribosome engagement or translational initiation at synaptic sites. Speculating how the 3’UTR deletion affects translational efficiency, we reviewed previous studies that proposed a common mammalian mechanism for efficient translation, where a closed-loop mRNA forms via a complex between polyA binding proteins and eukaryotic initiation factors circularize the mRNA to promote ribosome recycling and reinitiation (58). Deletion of the 3’UTR can disrupt this through loss of regulatory elements or altered UTR secondary structure, leading to impaired ribosome recruitment and initiation efficiency. Previous analyses of polysome-shifting UTR variants support this, showing that nucleotide substitutions in the 3′UTR can shift ribosome occupancy by altering trans-acting factor binding or local RNA structure (59).

In addition to translation initiation, 3’UTRs have also been implicated in regulating translation termination and ribosome recycling (60). Studies have also found ribosomes populating the 3’UTR to speculate on a potential function of this for efficient translational termination or recycling. We therefore consider a possible mechanism by which our model affects Grin2b translational efficiency, in that the ∆3’UTR removes necessary elements or structural motifs for recruiting translation factors and establishing the closed-loop conformation, thereby reducing ribosome loading and efficient ribosome turnover. Further studies must be dedicated to confirming the specific nature in how this 3’UTR deletion affects Grin2b translational efficiency and ribosome recruitment and turnover.

Beyond its role in translation initiation and ribosome recycling, the 3′UTR also influence mRNA compartmentalization within ribonucleoprotein granules, the primary form of transport for RNA (57, 61). Based on this, we analyzed puncta volume as an indirect proxy for size of RBP-mRNA granules, as has been used in prior studies (62). We observed increased puncta volume in ∆3’UTR neurons, suggesting that Grin2b granule assembly or packaging may be influenced by the 3’UTR. This is consistent with work showing that certain 3’UTRs influence granule transport dynamics and assembly (63). However, we note that this measure is an indirect readout as differences in puncta volume could also arise from reduced dendritic localization, causing mRNA to accumulate in the soma and form densely clustered signals that are more difficult to resolve due to the diffraction limit of light. Further studies will be necessary to determine whether these changes reflect genuine alterations in granule structure or distribution.

Previously Reported RBPs bind Grin2b 3’UTR as Potential Mechanism of RNA Transport and Local Protein Synthesis.

We identified several previously characterized RNA-binding proteins (RBPs) as top hits in our dataset, with significant GO terms related to “RNA binding” and “purine-rich negative regulatory element binding,” indicating relevant interactions with the Grin2b 3’UTR (SI Appendix, Fig. S8). Notable RBPs included TARDBP (64), hnRNPH2 (65), and CELF1 (66), and ELAVL1 (67), which are well established regulate mRNA processing, stabilization, and translation (SI Appendix, Fig. S9). The presence of these factors supports the idea that the Grin2b-3′UTR engages canonical posttranscriptional regulatory machinery (26) to regulate mRNA fate in neurons.

Comparison with RBPmap predictions revealed 14 overlapping proteins (SI Appendix) between the RBPmap database and our hits, including the notable top hits. High-confidence binding motifs were detected within both Grin2b-3′UTR subregions, notably for example, ELAVL1, which recognizes U-rich elements and promotes transcript processing and stabilization (67) (SI Appendix). However, as RBPmap predictions are based solely on primary sequence, we recognize this may not fully capture interactions mediated by RNA secondary or tertiary structure, nor interactions involving regions beyond the two tested subsegments, which cover ~20% of the total 3′UTR. Contrarily, our pulldown was performed in vitro and therefore does not capture the complex intricacies of these interactions in situ. These results therefore likely represent only a subset of the complete Grin2b RBP interactome and require additional investigation to fully clarify the complex interactome of this key neuronal gene.

Protein domain analysis using InterPro (44) revealed that all pulldown groups were dominated by classical RNA-binding and ribosomal protein families (SI Appendix), consistent with the abundance of translational and RNP complex components in neurons. The functional gene ontology profiles, however, were distinct across sequence regions and subcellular compartments: Cytosolic pulldowns were enriched for RNA processing and stabilization, whereas synaptosomal fractions showed strong enrichment for translation-related proteins (SI Appendix, Fig. S9). This pattern is consistent with previous literature finding translational machinery present in synaptosomes (68), while others find the importance of RNA stabilizing factors and transport machinery in the cytosol (69). Our data suggest that while similar molecular scaffolds engage with both Grin2b 3′UTR regions, their functional outcomes depend on local subcellular context. These results align with previous findings showing distinct RBP compositions across neuronal compartments (70, 71), including the soma, dendrites, and synapses, where proteins like FMRP and Pumilio regulate local translation (20, 72). In the cytosol, proteins contribute to processing and stabilization (26), while in the synaptosome, they are more associated with translational control, consistent with known roles of 3’UTRs in regulating synaptic protein synthesis (21, 33).

Expression data from the Allen Brain Atlas (45) showed that the RBP hits are most highly expressed in cortical subplate, hippocampal formation, and isocortex (Supplementary Information). This is consistent with the fact that we used cortical and hippocampal tissue for lysate preparation, as they are crucial for synaptic plasticity and learning (16, 46, 47).

Together, these results underscore the functional relevance of the Grin2b 3’UTR in posttranscriptional regulation across cytosolic and synaptic compartments (26, 27). Ongoing work will further characterize RBP-3’UTR interactions throughout the Grin2b mRNA life cycle.

Conclusion

Our findings demonstrate that the Grin2b 3’UTR plays a crucial role in mRNA localization and translation at synapses, directly influencing synaptic plasticity and hippocampal-dependent learning. By removing the 3’UTR, we observed altered mRNA distribution, reduced synaptic phosphorylated GluN2B expression, impaired plasticity, and disrupted learning task performance. These results align with previous studies observing GluN2B heterozygotes with similar behavioral and physiological profiles. Importantly, our study highlights the broader significance of 3’UTRs in regulating neuronal transcripts through RNA-binding protein interactions, emphasizing their role in localized translation and, subsequently, neuronal function. Given the increasing recognition of 3’UTR roles in mediating important cellular functions, targeting RNA-based regulatory mechanisms offers a promising avenue for therapeutic intervention. Future studies in this research should explore how specific RBP interactions and structural dynamics dictate Grin2b mRNA fate, ultimately shaping our understanding of neuronal function.

Supplementary Material

Appendix 01 (PDF)

Appendix 02 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Data, Materials, and Software Availability

RNA-seq data supporting the findings of this study can be found on the Sequence Read Archive (SRA) under accession PRJNA1311532 (73). WT hippocampal synaptosome RNA-seq data were extracted from SRA accession SRR1772422 (GEO accession GSE61991) (74). Source data are provided with this paper.