Loss of TET Activity in the Postnatal Mouse Brain Perturbs Synaptic Gene Expression and Impairs Cognitive Function

Abstract

Conversion of 5-methylcytosine (5mC) to 5-hydroxymethylcytosine (5hmC) by ten-eleven translocation (TET) family proteins leads to the accumulation of 5hmC in the central nervous system; however, the role of 5hmC in the postnatal brain and how its levels and target genes are regulated by TETs remain elusive. We have generated mice that lack all three Tet genes specifically in postnatal excitatory neurons. These mice exhibit significantly reduced 5hmC levels, altered dendritic spine morphology within brain regions crucial for cognition, and substantially impaired spatial and associative memories. Transcriptome profiling combined with epigenetic mapping reveals that a subset of genes, which display changes in both 5hmC/5mC levels and expression patterns, are involved in synapse-related functions. Our findings provide insight into the role of postnatally accumulated 5hmC in the mouse brain and underscore the impact of 5hmC modification on the expression of genes essential for synapse development and function.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12264-024-01302-2.

Introduction

The precise control of gene expression is fundamental for the development and functioning of the central nervous system, with epigenetic modifications serving as a critical regulatory mechanism [1, 2]. DNA methylation at the fifth position of cytosine (5-methylcytosine, 5mC) is an important epigenetic modification that is dynamically regulated during brain development and can be influenced by neuronal activity [3]. The seminal discovery that 5mC can be oxidized to 5-hydroxymethylation (5hmC) by the ten-eleven translocation (TET) family enzymes has sparked significant interest in unraveling the function and mechanism of DNA demethylation across various cell types, including neurons [4–8]. Subsequent studies have revealed that active demethylation involves further oxidization of 5hmC to produce 5-formylcytosine and 5-carboxycytosine by TET proteins, as well as restoration of oxidized 5mC to unmodified cytosine through the thymine DNA glycosylase-dependent base excision repair pathway [9]. In addition to serving as an intermediate for demethylation, 5hmC also functions as a stable epigenetic marker that is abundant in the brain and exhibits specific distribution patterns across different tissues and cell types [5, 10, 11]. In humans, 5hmC has been implicated in both neurodevelopmental and neurodegenerative disorders, highlighting its important role in the brain [12–15].

Two distinct mechanisms are known by which 5hmC regulates gene expression. One is through TET-mediated active DNA demethylation, in which 5hmC acts as an intermediate [16, 17]. The other involves the direct influence of 5hmC on chromatin structure [10, 18]. Genome-wide profiling studies have demonstrated that 5hmC in the CG context is enriched in regulatory elements and gene bodies, and its abundance is positively correlated with gene expression levels [19–21]. In line with these findings, the modification of gene bodies by 5hmC diminishes MeCP2 binding and facilitates transcription [18]. Analysis of various human tissues has revealed a preferential buildup of 5hmC in chromatin regions associated with the tissue-specific regulation of transcription, particularly in the brain [20, 21]. During postnatal brain development, there is a rapid increase in 5hmC levels, which coincides with synaptogenesis [12, 22], suggesting a potential role for 5hmC in regulating the expression of genes crucial for synapse formation. This notion is further supported by the abundance of 5hmC in synaptic genes [23]. Despite these important findings, a causal relationship between 5hmC and gene expression has yet to be established in the postnatal forebrain. Furthermore, the comprehensive landscape of the postnatally accumulated 5hmC in the genome and the intricate gene regulatory networks associated with 5hmC accumulation, especially in the forebrain, remain largely unexplored.

The role of TETs (TET1, TET2, and TET3) has been investigated through the deletion of Tet genes in mice, as well as by downregulating Tet expression using RNA interference. These studies provide evidence that TET proteins are essential for adult neurogenesis, cerebellar development, and cognitive function [16, 24–30]. However, the loss of individual Tet genes has negligible or minimal impact on 5hmC levels [25, 30], and it has become evident that the phenotypes of these mutant mice exhibit significant variation depending on which Tet gene is deleted [31], raising a concern that the redundant/compensatory effects in these single Tet-knockout mice may confound the interpretation of the data. Therefore, simultaneous deletion of all three Tet genes in a genetically defined population of cells would be advantageous when investigating the role of TETs and 5hmC in the brain.

In this study, we conditionally deleted Tet1-3 in forebrain excitatory neurons shortly after birth and comprehensively examined the molecular and behavioral consequences to elucidate the role of TET family enzymes and 5hmC in governing the regulation of gene expression and cognitive function.

Materials and Methods

Animals

Tet1/2/3 floxed mice on a 129Sv background were generated in Dr. Guoliang Xu’s Lab [29, 32–34]. To conditionally inactivate Tet genes in excitatory neurons, we crossed the Tet1/2/3 floxed mice with Camk2a-iCre transgenic mice [35] to obtain Tet1^flox/flox; Tet2^flox/flox; Tet3^flox/flox; Camk2a-iCre mice. Littermates negative for Camk2a-iCre were used as controls. The cTKO mice were on a mixed 129Sv&C57BL/6J background. The primers for genotyping are listed in Supplementary Table S1A. Animals were housed under controlled conditions of temperature and humidity with a 12-h light/dark cycle. They were provided ad libitum access to food and water. All experimental procedures were approved by the Institutional Animal Care and Use Committee of the Center for Excellence in Brain Science and Intelligence Technology, Chinese Academy of Science.

Behavioral Testing

Five cohorts of adult cTKO mice and control littermates aged 3–6 months were used for the behavioral assessment. The behavioral tests were conducted at the Behavioral Testing Core Facility of the Center for Excellence in Brain Science and Intelligence Technology. The tests included open field (OF), elevated plus maze (EPM), light/dark transition (LTD), tail suspension (TST), forced swimming (FST), fear conditioning (FC), visual cliff, novel subject recognition (NOR), and the Barnes maze (BM). Another cohort of animals was used for the coordination test in the following order: gait analysis, beam balance, coat hanger test, and rotarod. All experimental procedures were approved by the Institutional Animal Care and Use Committee of the Center for Excellence in Brain Science and Intelligence Technology, Chinese Academy of Science. Details of the behavioral testing methods can be found in the Supplementary Material.

Bisulfite (BS)- and Oxidative Bisulfite (oxBS)-sequencing

Total genomic DNA was extracted from the hippocampus of adult male control and cTKO mice aged 3 months (n = 3 per group) using the AllPrep DNA/RNA Mini kit. The isolated DNA was then mixed with unmethylated lambda DNA and fragmented to an average size of ~ 250 bp. Subsequently, the purified fragmented DNA underwent a series of treatments including repair, blunting, and phosphorylation of ends using T4 DNA polymerase, Klenow Fragment, and T4 polynucleotide kinase, respectively. The resulting blunt-ended DNA fragments were supplemented with spike-in control DNA, adenylated at their 3′ ends using Klenow Fragment (3′-5′ exo-), and ligated to adaptors synthesized with 5’-mC instead of cytosine utilizing T4 DNA Ligase. Following purification, the fragmented DNA samples were divided into two aliquots for bisulfite conversion or oxidation followed by bisulfite conversion using the TrueMethyl Seq Kit (Cambridge Epigenetics, CEGX, Essex, UK). The final oxBS and BS libraries were generated by PCR amplification with adapter-compatible barcode primers, quantified, and sequenced on the Illumina HiSeq platform. Low-quality bases were trimmed using Trim Galore v0.5.0 [36]. Trimmed reads were aligned to the mouse genome using BSMAP software v2.90 [37]. Methylation ratios were extracted from the BSMAP output using the methratio.py (from the BSMAP package) script. Only cytosines in a CpG context with a minimum coverage of 5× were retained for further analysis.

RNA-seq and Data Analysis

The RNeasy Plus Mini Kit (Roche, USA) was used to extract total RNA from the hippocampus of adult control and cTKO mice aged 3 months (n = 4 per group). The extracted RNA was then purified using polyA selection, converted into cDNA, and sequenced on an Illumina HiSeq 2000 sequencing system with a paired-end protocol generating reads 150 bp in length. Quality control analysis was applied using FastQC v0.11.5. Sequencing reads were trimmed utilizing Skewer v0.2.2 [38]. The trimmed reads were aligned to the reference genome using STAR 2.5.3a [39]. Gene expression levels were estimated by StringTie v1.3.1c [40] and differential expression analysis was applied using DESeq2 v1.16.1 [41]. Genes were considered differentially expressed between groups if FDR < 0.05, the absolute fold change > 1.5, and RPKM > 0.5 in at least two samples.

Detection of Differentially Methylated Regions (DMRs) and Differentially Hydroxymethylated Regions (DHMRs)

DMRs and DHMRs were identified using Metilene v0.2-6 [42] in de-novo mode among CpG sites with at least 3× coverage. The detected DMRs and DHMRs were subjected to stringent selection criteria: (1) corrected Mann-Whitney U test P-value <0.05; (2) 2D Kolmogorov-Smirnov test P-value <0.05; (3) relative methylation level difference >0.1; (4) CpG number contained in DMR >3; and (5) length of the DMR >50 bp.

Golgi Staining

Control and cTKO littermates at postnatal day 30 (P30) and P60 were sacrificed under anesthesia with pentobarbital (70 mg/kg, i.p.). Brains were quickly collected, and Golgi staining was applied using the FD Rapid GolgiStain™ Kit (FD NeuroTechnologies, Columbia, USA) as previously reported [43]. The basal dendrites of pyramidal neurons from hippocampal CA1 and the barrel field of the primary somatosensory cortex (S1BF) region were imaged on a Nikon A1 confocal microscope (Nikon, Tokyo, Japan) with a Plan Apo VC 60 × Oil DIC N2 objective (N.A. = 1.4), at 3 × optical zoom and 1 μm Z intervals. Spine density, length, and head diameter were evaluated using Fiji/ImageJ software. Spine subtypes were classified into mushroom, stubby, thin, and branched, as previously described [44].

Immunohistochemistry

Mice were anesthetized with pentobarbital (70 mg/kg, i.p.) and transcardially perfused with freshly prepared 4% paraformaldehyde in 0.1 mol/L phosphate buffer (pH 7.4) followed by post-fixation in the same solution overnight at 4 ℃. Brains were dehydrated in 30% sucrose in 0.1 mol/L PBS overnight before being embedded in OCT and cut into 40-μm-thick coronal sections on a cryostat (Leica, Wetzlar, Germany). Sections were incubated in the blocking buffer for 60 min at room temperature and incubated with primary antibodies overnight at 4 °C. After washing, the sections were incubated with appropriate secondary antibodies from the Alexa Fluor series (Invitrogen, Carlsbad, USA) for 2 h at room temperature and then counterstained with Hoechst 33342 (Beyotime, Shanghai, China) for 15 min at room temperature to identify cellular nuclei. After mounting the sections, images were captured with a VS200 fluorescence microscope (Olympus, Tokyo, Japan) or a Nikon A1 confocal microscope (Nikon, Tokyo, Japan). The following primary antibodies were used: anti-NeuN (MAB377, Merck Millipore, Darmstadt, Germany), anti-Prox1 (AB5475, Merck Millipore), anti-Cux1 (11733-1-AP, Proteintech, Wuhan, China), anti-Ctip2 (ab18465, Abcam, Massachusetts, USA), and anti-Tbr1 (Ab319740, Abcam, England).

Fluorescent in situ Hybridization and Immunostaining

Coronal brain sections from adult cTKO and control mice aged 4 months were used for fluorescent in situ hybridization using a vGlut1 probe (501101, Advanced Cell Diagnostics) according to the manufacturer’s instructions. After visualization of in situ hybridization, sections were processed for immunohistochemistry. The sections were incubated in the blocking buffer for 60 min at room temperature and incubated with an anti-5hmC antibody (39769, Active Motif, Carlsbad, USA) overnight at 4 °C. After washing, the sections were incubated with appropriate secondary antibodies from the Alexa Fluor series (Invitrogen, Carlsbad, USA) for 2 h at room temperature and then counterstained with Hoechst 33342 (Beyotime, Shanghai, China) for 15 min at room temperature to identify cellular nuclei. After mounting the sections, images were captured with a VS200 fluorescence microscope (Olympus, Tokyo, Japan) or a Nikon A1 confocal microscope (Nikon, Tokyo, Japan).

Cell Counting in the Cortical Wall

Coronal sections from P4 cTKO and control mice were collected from the medial cortical region and stained with Cux1, Ctip2, and Tbr1 antibodies. 3–4 sections at different levels from each mouse were used for cell counting in a fixed area of 496 μm × 768 μm.

Recording Spontaneous Excitatory Postsynaptic Currents (sEPSCs)

cTKO and control mice aged either 1 month or 5 months were used for sEPSC recording. Whole-cell recordings from pyramidal neurons in the hippocampal CA1 and pyramidal neurons in layer II–III of the S1BF were performed with a MultiClamp 700B amplifier (Axon CNS, Molecular Devices, Sunnyvale, CA, USA) as previously reported [45].

Statistics

Data were analyzed using GraphPad Prism 8 (GraphPad Software, La Jolla, CA), SPSS Statistics 26 (IBM, Chicago, IL, USA), R v4.1.2, and R studio v2022.02.0+443. Data are presented as the mean or the mean ± standard error of the mean (SEM), and "n" refers to either the number of animals, neurons, or the number of spine segments (for spine density/length/width). Two means were compared by the unpaired Student’s t-test, Welch’s t-test, or Mann-Whitney U test based on the variances and distribution of the data. One-way ANOVA followed by Tukey’s multiple comparisons test was used to determine the differences between the means of 3 or more groups. For the conditions with two independent variables, two-way ANOVA followed by Sidak’s multiple comparisons test was used. Differences were considered significant when P <0.05 (*P <0.05, **P <0.01, ***P <0.001, ****P <0.0001). Statistical reports are shown in Table S2.

Results

Postnatal Deletion of all Three Tet Genes in Forebrain Excitatory Neurons Significantly Reduces 5hmC Levels Without Affecting Overall Brain Morphology

To achieve simultaneous deletion of Tet1, Tet2, and Tet3 in the forebrain, we generated Tet triple knockout mice (Tet1^fl/fl; Tet2^fl/fl; Tet3^fl/fl; Camk2a-iCre, hereafter referred to as cTKO) by crossing Tet1/2/3 floxed mice with Camk2a-iCre mice [35]. This breeding strategy ensured the specific deletion of all three Tet genes in forebrain excitatory neurons shortly after birth [35]. Quantitative real-time PCR (qPCR) analysis revealed a significant reduction in the levels of Tet transcripts in the hippocampus and cortex, but not in the cerebellum, of cTKO mice compared to control mice aged 3 months (Fig. S1), confirming the selective loss of Tet mRNA expression in the mutant mice.

We next assessed the effects of Tet deletion on 5hmC accumulation in the brain by measuring global 5hmC levels in genomic DNA extracted from various brain regions of adult control and cTKO mice aged 3 months. The dot blot assay revealed a robust reduction of 5hmC levels by 62% and 41% in the hippocampus and cortex of cTKO mice, respectively, compared to control mice (Fig. 1A). The levels of 5hmC in the cerebellum remained unchanged, as expected (Fig. 1A). In addition, we measured the 5mC levels and found no significant differences between control and cTKO mice (Fig. 1B), likely due to much higher levels of basal methylation compared to hydroxymethylation [18, 43]. To assess the effects of Tet deficiency on 5hmC accumulation in excitatory neurons, we applied fluorescent in situ hybridization using RNAscope technology combined with 5hmC immunostaining. The results demonstrated a significant reduction in 5hmC immunoreactivity within vGlut1-positive cells in the cortex and hippocampus of cTKO mice (Figs. 1C, D, and S2).

Fig. 1.

Postnatal deletion of Tet1/2/3 in forebrain excitatory neurons significantly reduces 5hmC levels without altering overall brain morphology. A Dot blots showing genomic 5hmC levels in control and cTKO mice aged 3 months. B Dot blots showing genomic 5mC levels of control and cTKO mice aged 3 months. C, D Representative images of vGlut1 mRNA and 5hmC detected through in situ hybridization and immunostaining, respectively, in the cerebral cortex. E A cTKO mouse and its control littermate at the age of 2 months. F Growth curves of control and cTKO mice fed a chow diet (n = 2–8). G The brain weight of adult control and cTKO mice aged 3 months. H Representative images of NeuN immunostaining in the brains of adult control and cTKO mice aged 3 months. Scale bar, 1 mm. I Representative images of Cux1, Ctip2, and Tbr1 immunostaining in the primary somatosensory cortex from P4 control and cTKO mice. Scale bars, 100 μm. J Representative images of Ctip2 and Prox1 immunostaining in the hippocampus from control and cTKO mice aged 3 months. Scale bars, 200 μm and 20 μm (inset). Hip, hippocampus; Ctx, cerebral cortex; Cb, cerebellum. *P < 0.05, **P < 0.01, ****P < 0.0001; Student’s t-test (A, B, G); Welch’s t-test (C, D); Student’s t-test and Welch’s t-test (E); Two-way ANOVA (F)

The cTKO mice were viable and survived to adulthood, albeit with significantly lower body weight (Fig. 1E, F). However, their brain weight was comparable to control mice (Fig. 1G). Tet deletion did not cause apparent abnormalities in the overall morphology of the brain, as demonstrated by immunostaining coronal sections for the pan-neuronal marker NeuN (Fig. 1H). We also examined the lamination by utilizing antibodies specific to the cortical layer markers Cux1, Ctip2, and Tbr1, as well as the DG granule cell marker Prox1. The laminar organization of the cortex and hippocampus exhibited no discernible differences between control and cTKO mice (Fig. 1I, J). Together, these results demonstrate that the combined loss of Tet1-3 in excitatory neurons substantially suppresses the accumulation of 5hmC during postnatal development, while preserving overall brain structures.

Tet cTKO Mice Exhibit Increased Locomotor Activity

To gain insights into the role of TET-mediated 5hmC accumulation in the brain at the behavioral level, we assessed the performance of cTKO mice aged 3–6 months across a battery of behavioral tests. The general activity of these mice was initially evaluated using the OF test. During a 30-min testing session, cTKO mice exhibited significantly higher locomotor activity than control mice, as evidenced by their greater travel distance (Fig. 2A). The percentage of time spent in the center area did not differ significantly between control and cTKO mice (Fig. 2B), suggesting that loss of TET activity either does not impact anxiety-related behaviors or that the OF test lacks sufficient sensitivity to detect such deficits. We next conducted a series of assessments including a balance beam, coat hanger, pole test, rotarod, and gait analysis to further evaluate the coordination of cTKO mice. The groups of animals did not exhibit any significant differences in the coat hanger, pole test, rotarod, and gait test (Fig. S3B–H). However, in the beam test, an increase in the latency was found in cTKO mice compared to control mice (Fig. S3A), indicating mild impairments in their coordination. The light/dark transition (LDT) and EPM tests were subsequently applied, as they are more specific measures for evaluating anxiety traits. cTKO mice showed no significant alterations in the number of transitions and time spent in the light compartment in the LDT test (Fig. 2C, D). However, in the EPM test, cTKO mice spent significantly more time in the open arms (OA) (Fig. 2E, F), made a higher percentage of OA entries (Fig. 2G), and had a shorter latency of first entry to the OA (Fig. 2H) than control mice. We also found a tendency towards higher velocity values (Fig. S4A) and more total center crosses in cTKO mice (Fig. S4B). The behavioral phenotypes are unlikely to be attributed to impaired vision, as these mice had normal behaviors in the visual cliff experiment (Fig. S4C). Considering the unchanged anxiety-like behaviors in OFT and LDT experiments, these behavioral abnormalities may arise from the hyperactive locomotion of cTKO mice rather than being solely attributed to decreased anxiety levels.

Fig. 2.

Tet cTKO mice exhibit enhanced locomotor activity. A, B Total travel distance (A) and percentage of time spent in the center area (B) for both control and cTKO mice in the open filed test. C, D Number of transitions (C) and time spent (D) in the light compartment for control and cTKO mice in the light/dark transition test. E–H The behavior patterns of control and cTKO mice in the elevated plus maze test. Average heat maps indicate mean dwell times using a color scale (E); percentage of time spent in the open arms (F); percentage of entries into the open arms (G); and latency to enter open arms (H). I Immobility times in the forced swimming test. J Immobility times in the tail suspension test. **P < 0.01, ***P < 0.001, ****P < 0.0001; Welch’s t-test (A, H); Student’s t-test (B–D, F, G, I, J).

To investigate the impact of Tet deficiency on depression-related behaviors, we applied two behavioral assays widely used in rodents, namely the FST and TST. The immobility time in both tests remained unchanged in cTKO mice compared to control mice (Fig. 2I, J), suggesting that disruption of TET function does not affect depression-related behaviors.

Tet cTKO Mice Exhibit Impaired Long-term Associative and Spatial Memory

Since accumulating evidence suggests that dynamic DNA methylation is implicated in synaptic plasticity and memory formation [44, 45], we determined whether Tet deficiency affects cognitive performance in mice using well-established memory tests. We first applied an FC test to evaluate associative learning and memory [46]. In the contextual FC test, cTKO mice showed a significantly reduced freezing response during memory retrieval compared to control mice (Fig. 3A). In addition, we applied the cued FC test to evaluate hippocampus-independent fear memory and obtained similar results (Fig. 3B). These findings suggest that long-term associative memory is compromised by Tet deficiency.

Fig. 3.

Tet cTKO mice exhibit impaired long-term associative and spatial memory. A, B Fear responses of control and cTKO mice in contextual (A) and cued (B) fear conditioning tests. C–G The behaviors of control and cTKO mice in the training trial (C–E) or probe trial (F, G) of the Barnes maze test. The target hole is indicated by a dotted circle and arrow. Average heat maps for each group are displayed. *P < 0.05, **P < 0.01; Welch’s t-test (A); Student’s t-test (B, G); Two-way repeated ANOVA (C, D).

We next assessed the effects of Tet deficiency on spatial learning and memory using the BM test [47]. During the five-day training phase, both control and cTKO mice were able to learn the task, as demonstrated by a reduction in errors and latency to reach the target hole at the end of the training (Fig. 3C, D). However, cTKO mice exhibited significantly higher error rates and longer escape latencies throughout the training period than control mice (Fig. 3C, D). Rodents typically applied three search strategies in solving the BM: random, serial, and spatial approaches [48], of which the spatial strategy is the most efficient and depends on spatial learning. To investigate the potential causes of the poor performance of cTKO mice during training, we analyzed the search strategies used by these mice in locating the target hole. The control mice quickly adopted the spatial strategy to search the hole by Day 5 (Fig. 3E). In contrast, cTKO mice predominantly applied non-spatial search strategies (Fig. 3E), reflecting impaired spatial learning or an inability to apply the spatial strategy potentially due to their hyperactive locomotion. We further assessed long-term spatial memory in the probe trial and found that cTKO mice spent significantly less time in the target quadrant than control mice (Fig. 3F, G) while maintaining a comparable velocity (Fig. S4D). These findings suggest an impairment in long-term spatial memory in Tet-deficient mice.

Finally, to test whether TET-mediated 5hmC accumulation is important for other forms of memory, we subjected the mice to the NOR task to evaluate their capacity for memorizing novel objects. We found that both control and cTKO mice spent more time exploring the novel object than the familiar one, and the discrimination index showed no significant difference between the two groups of mice (Fig. S4E, F), indicating normal NOR memory in cTKO mice.

The Effects of Tet Deficiency on Dendritic Spine Morphology and Excitatory Synaptic Transmission in Brain Regions Crucial for Cognitive Function.

The postnatal accumulation of 5hmC is concomitant with active synaptogenesis [12], implying that the behavioral deficits in cTKO mice may be attributed to aberrant synaptic connectivity. To investigate the effects of Tet deficiency on synapse development, we focused on dendritic spines, the postsynaptic sites that receive most of the excitatory synaptic inputs [49]. Two crucial regions involved in cognitive function were chosen for analysis: the hippocampus and cortex. We first analyzed spine density on the basal dendrites of pyramidal neurons in both the hippocampal CA1 region and the barrel field of the S1BF. No significant differences in spine density were found between control and cTKO mice at either P30 or P60 within these regions (Fig. 4A, B and F, G).

Fig. 4.

Tet deficiency affects the morphogenesis of dendritic spines in brain regions critical for cognitive function. A, F Representative images of dendrites revealed by Golgi staining in CA1 (A) and the S1BF (F). B, G Quantification of spine density in the pyramidal neurons of CA1 (B) and the S1BF (G). C, H The fraction of spine subtypes in CA1 (C) and the S1BF (H). D, I The length of mushroom and stubby spines in CA1 (D) and the S1BF (I). E, J The head diameter of mushroom and stubby spines in CA1 (E) and the S1BF (J). K Representative recording of sEPSCs. Quantification of frequency and amplitude of sEPSCs in the pyramidal neurons of CA1 (L, M) and the S1BF (N, O). *P < 0.05, **P < 0.01; two-way ANOVA (B, C, G, H, L, M, N, O); Mann-Whitney U test (D, E, I, J). S1BF, the barrel field of the primary somatosensory cortex.

The morphology of spines is closely associated with the strength and efficacy of synaptic transmission [50]. To determine whether loss of TET activity affects spine structure, we categorized spines into four groups based on their shape: thin, stubby, mushroom, and branched [50] and quantified the percentage of spines in each group. We found a significantly lower proportion of mushroom spines and a higher proportion of thin spines in the CA1 area of cTKO mice at P30 compared to control mice (Fig. 4C). The distribution of each spine type in the S1BF region exhibited similar trends, although a statistically significant difference was not achieved (Fig. 4H). The changes seemed to be transient during development, as they were not detected at P60 when mice reached maturity (Fig. 4C, H). These findings suggest a delay in spine maturation during postnatal development in cTKO mice, as mushroom spines typically form synapses that are considered “mature”, while thin spines often carry small or immature synapses [50]. We next analyzed the morphology of individual spines in detail, with a specific focus on two morphologically stable spine types: mushroom and stubby. In the hippocampal CA1 region, there were no discernible differences in spine length or head diameter between control and cTKO mice (Fig. 4D, E). However, in the S1BF region, while head diameter remained unchanged (Fig. 4J), spine length was significantly longer in cTKO mice at both P30 and P60 (Fig. 4I), resembling the morphology of immature spines during early neocortical development [51]. To functionally evaluate the impact of Tet deficiency on excitatory synaptic transmission, we applied the whole-cell patch clamp technique to record sEPSCs in ex vivo CA1 and S1BF slices from P30 control and cTKO mice. While no significant differences in sEPSC frequencies and amplitudes were found between groups in CA1, there was a decrease in EPSC amplitude in the S1BF region of cTKO mice (Fig. 4K–O). Notably, these changes were absent from 5-month-old mice, consistent with the transient but not permanent alterations in spine morphology (Fig. S5).

Together, these findings suggest that TETs are involved in the maturation of excitatory synapses during the late stages of development with a certain degree of regional specificity and imply that the absence of TETs may affect the formation and refinement of neural circuits underlying learning and memory.

Tet Deficiency Alters the Genome-wide Distribution of 5hmC and 5mC in the Postnatal Brain

The genomic distribution of 5hmC has been extensively investigated in various tissues, including the brain [16, 18, 20, 21]. These studies have unveiled the predominant occurrence of 5hmC in euchromatin regions and its enrichment within actively transcribed gene bodies. To explore the impact of Tet deficiency on the epigenomic landscapes of the postnatal forebrain, we generated high-resolution maps of 5hmC and 5mC across the entire genome using bisulfite sequencing (BS-seq) and oxidative bisulfite sequencing (oxBS-Seq) techniques in the hippocampus of both control and cTKO mice. In our analysis, we specifically focused on 5hmC and 5mC in the CG context due to their high abundance in the neuronal genome. The BS-seq and oxBS-seq data were collected from three independent biological replicates. The processes of oxidation and bisulfite conversion were highly efficient and reproducible (Fig. S6). The number of reads per sample ranged from 264 to 372 million (Table S3). Consistent with previous studies [4, 18, 52], we found a much higher abundance of 5mC (~ 60%) than 5hmC (~ 16%) in the hippocampus of control mice (Fig. 5A, B). Notably, Tet deficiency resulted in significantly reduced levels of 5hmC accompanied by elevated levels of 5mC in the hippocampal genome (Fig. 5A, B).

Fig. 5.

Tet deficiency leads to genome-wide alterations in 5mC and 5hmC distributions. A, B Violin plots showing the percentages of 5hmC (A) and 5mC (B) distributed in the CG contexts across the genome. C A Circos plot representing the distribution of DHMRs and DMRs across the genome. D The distribution patterns of hypo-DHMRs and hyper-DMRs across various genomic regions. E Distribution of annotated hypo-DHMRs and hyper-DMRs overlapping with gene loci in CpG islands and shores/shelves. Mann-Whitney U test (A, B).

Next, we proceeded to characterize specific loci that displayed altered distribution of 5hmC and 5mC in cTKO mice compared to control mice by annotating differentially hydroxymethylated regions (DHMRs) and differentially methylated regions (DMRs). In total, we identified 187,746 DHMRs and 103,748 DMRs across the genome. These DHMRs and DMRs were widely distributed throughout all autosomes but were scarcely detectable in sex chromosomes (Fig. 5C), showing that sex chromosomes are mostly depleted of 5hmC, which is consistent with previous findings [12]. Nearly all DHMRs (99.54%) were hypo-hydroxymethylated, while almost all DMRs (99.94%) were hyper-methylated. Both hypo-DHMRs (~ 65%) and hyper-DMRs (~ 62%) were enriched in gene bodies (Fig. 5D) and depleted from CpG islands (Fig. 5E), aligning with previous studies [20, 31, 53].

Loss of TETs Leads to Aberrant Expression of Genes Crucial for Synaptic Function and Memory Formation

A growing body of research has suggested that the accumulation of 5hmC in gene bodies promotes gene expression [16, 18–21]. However, it remains unclear how neuronal genes are influenced by 5hmC accumulation during postnatal development and their causal roles in brain development and function. We annotated genes associated with the identified hypo-DHMRs, hyper-DMRs, or both. A total of 15,947 and 14,768 genes were found to be associated with hypo-DHMRs and hyper-DMRs, respectively (Fig. 6A). Among them, 12,626 genes (~ 70%) were associated with both hypo-DHMRs and hyper-DMRs, while 3321 genes (~ 18%) were specifically associated with hypo-DHMRs and 2,142 genes (~ 12%) were specifically associated with hyper-DMRs (Fig. 6A).

Fig. 6.

Tet deficiency leads to aberrant expression of genes essential for synaptic function and memory formation. A Venn diagram of DEGs and genes associated with hypo-DHMRs and hyper-DMRs. B Volcano plot showing genes differentially expressed between control and cTKO hippocampal samples. C Heat maps displaying mRNA expression patterns of DEGs associated with both hypo-DHMRs and hyper-DMRs. D GO enrichment analysis of DEGs associated with both hypo-DHMRs and hyper-DMRs. E Pathway interaction analysis by evaluating KEGG processes of the DEGs associated with both hypo-DHMRs and hyper-DMRs. F GSEA analysis of the whole transcriptome for the indicated gene sets. G qPCR analysis of mRNA expression levels of DEGs implicated in learning and memory. *P < 0.05, **P < 0.01; Student’s t-test and Mann-Whitney U test (G).

To identify genes whose expression levels are altered by Tet deficiency, we applied transcriptomic analysis using the RNA sequencing (RNA-seq) technique in the hippocampus of control and cTKO mice. The data exhibited strong correlations between biological replicates, confirming their high quality (Fig. S7A and Table S4). Furthermore, principal component analysis based on the expression profiles demonstrated distinct clustering of samples from control or cTKO mice (Fig. S7B). Applying the predefined criteria (FDR < 0.05, absolute fold change > 1.5, RPKM > 0.5 in at least 2 samples), we identified a total of 196 differentially-expressed genes (DEGs) in cTKO mice compared to control mice. Among these DEGs, 111 were downregulated (Table S5) and 85 were upregulated (Table S6). To validate the RNA-seq data, qPCR analysis was applied to a subset of genes selected from the list (Table S7); this validation showed a strong correlation between the two datasets (Fig. S7C).

We compared the list of genes associated with hypo-DHMR, hyper-DMR, or both to the list of DEGs obtained through transcriptome profiling to identify genes whose expression might be regulated by 5hmC accumulation. We identified a total of 167 overlapping genes, the majority of which (127/167) were associated with both hypo-DHMRs and hyper-DMRs (Fig. 6A). More than 70% of the DEGs (90/127) linked to both hypo-DHMRs and hyper-DMRs exhibited downregulation in their expression levels (Fig. 6B, C), in agreement with the role of 5hmC in promoting gene expression [18–21]. We applied Gene Ontology (GO) enrichment analysis to annotate the 127 DEGs and found significant enrichment of these genes in molecular functions associated with receptor-mediated signaling, such as growth factor binding and peptide receptor activity (Fig. S8). Notably, the activin receptor signaling pathway, known for its role in regulating synapse formation [54], was found to be one of the most enriched biological processes (Fig. 6D). Consistent with this, dendritic spine emerged as one of the highly enriched cellular components (Fig. 6D). To identify the key pathways associated with these DEGs, we applied a pathway interaction analysis using the Kyoto Encyclopedia of Genes and Genomes (KEGG). This analysis revealed that the transforming growth factor-β (TGF-β) signaling, Wnt signaling, and neuroactive ligand-receptor interaction pathways were significantly enriched (Fig. 6E). Considering that functional consequences are generally determined by the expression levels of an entire gene set, we subsequently applied gene set enrichment analysis (GSEA) to the transcriptome of control and cTKO mice. These analyses revealed systematic alterations in gene sets related to synaptic organization and neuronal differentiation in the cTKO hippocampus (Fig. 6F). Interestingly, several DEGs, including Glp2r [55], Rgs14 [56, 57], Rin1 [58], Ntsr1 [59, 60], and Cpne6 [61], have been implicated in synaptic function and memory formation, suggesting that their dysregulation in Tet-deficient mice may contribute to the cognitive deficits. In addition, RGS14 [62] and NTSR1 [63] have also been linked to human memory disorders. Importantly, the aberrant expression of these genes identified through RNA-seq was validated by qPCR (Fig. 6G). Collectively, these findings suggest that TET-mediated accumulation of 5hmC regulates crucial gene networks involved in synapse function and memory.

Discussion

Emerging evidence has implicated 5hmC in the pathogenesis of neurological disorders. However, the biological function of TETs and 5hmC in the brain remains largely unknown. To investigate the role of postnatally accumulated 5hmC, we generated a mouse line in which all three Tet genes were conditionally deleted in a specific type of cell. The combined loss of TETs in a defined cell type has only been achieved recently in cerebellar Purkinje cells [16]. The loss of all TETs resulted in a marked reduction in 5hmC levels, which did not occur when Tet genes were individually deleted, circumventing potential confounding compensatory effects among TET paralogs. We further showed that loss of TETs led to behavioral deficits, including severe impairments in learning and memory. Finally, the genome-wide epigenomic analysis and transcriptome profiling pointed to a critical role for TETs and 5hmC in regulating gene networks essential for synapse development and function.

The role of TETs in mouse behaviors has been examined through the deletion of individual Tet genes. The genetic studies using different mouse lines have identified largely non-overlapping behavioral deficits. Our finding that mice lacking all TETs exhibited more severe memory impairments compared to single Tet KOs suggests a significant contribution of redundant/compensatory effects to the behavioral phenotypes, which underscores the importance of using cell type-specific triple Tet KOs for elucidating the function of TETs and 5hmC in the brain. One of the most notable phenotypes in cTKO mice is the profound impairment of long-term associative and spatial memory. It is worth noting that the deletion of either Tet1 or Tet3 does not have any impact on the formation of long-term memory, while conditional deletion of Tet2 in adult neural progenitor cells only results in a deficit in hippocampus-dependent memory formation [25, 27, 29, 30]. These results support the notion that functional compensation among TET paralogs can mask key phenotypes. The conditional deletion of Tet2 in adult neural progenitor cells results in a deficit in hippocampus-dependent memory formation [64]. However, the role of TET2 in postmitotic neurons remains undetermined and requires further investigation.

How might Tet deficiency impair cognitive function? One potential explanation is that the absence of TET activity disrupts the neural circuits underlying cognition. This idea is supported by our findings that Tet-deficient mice exhibited altered dendritic spine morphology and decreased sEPSC amplitude in brain regions crucial for higher-order information processing. Although these effects seem to be most prominent during late developmental stages, it is still likely that brain wiring was permanently changed in adult cTKO mice, which eventually influenced behaviors. Other mechanisms such as impaired synaptic plasticity may also contribute to the memory deficits in cTKO mice. For example, it is possible that long-term potentiation or long-term depression, two forms of plasticity that are thought to be the cellular mechanism for learning and memory, is compromised in cTKO mice. It might be of interest to test these possibilities in future studies.

Comparing the profiles of 5hmC and 5mC between control and Tet-deficient mice enabled us to gain several important insights. First, Tet deficiency not only reduced 5hmC levels across the genome but also led to elevated levels of 5hmC in a subset of genes involved in spine organization (Fig. S9A). This finding immediately suggests that TET-mediated active demethylation occurs in the postnatal forebrain, as disruption of this pathway can lead to increased levels of 5hmC due to its further oxidation into 5-acetylcytosine and 5-formylcytosine by TETs, which extends previous findings in the cerebellum [16]. Second, through conditional deletion of Tet genes shortly after birth, we have demonstrated that the postnatal accumulation of 5hmC primarily occurs in excitatory neurons, with significant enrichment in gene bodies. Last but not least, genes with 5hmC accumulation during postnatal development are significantly enriched in biological processes such as the regulation of membrane potential, organization of the endomembrane system, and signal transduction mediated by small GTPase, indicating a preferential accumulation of 5hmC in gene sets with specific functional roles (Fig. S9B).

Characterizing the genes influenced by the accumulation of 5hmC is important for understanding the function of this epigenetic modification. We have identified > 12,000 genes associated with both DHMRs and DMRs. However, only 127 of these genes exhibited differential expression in Tet-deficient mice. Although we cannot exclude the possibility that our bulk RNA-seq analysis lacks sensitivity in identifying all DEGs in samples with high heterogeneity, it is plausible to speculate that a significant portion of 5hmC accumulation does not affect the constitutive expression of its target genes, but rather primes these genes for context-dependent and stimulus-induced expression through specific readers of 5hmC. If this hypothesis holds true, it would be highly intriguing to understand the role of 5hmC in the experience-dependent plasticity that plays a central role in neural development, plasticity, and cognitive function. We characterized the 127 DEGs associated with both hypo-DHMRs and hyper-DMRs using GO enrichment and pathway analyses. The results revealed an enrichment of these genes in synaptic function, which is consistent with a previous study demonstrating the abundant presence of 5hmC in synaptic genes [23]. Notably, some of these DEGs are implicated in memory formation in rodents and human memory disorders. Although it remains unclear how the loss of TETs impairs synapse development and alters animal behaviors, these DEGs are strong candidates that warrant further investigation. It should be noted that our sequencing data was obtained from a mixture of cells with high heterogeneity, which may have introduced potential confounding factors. This should be taken into consideration when interpreting the data. Future studies applying single-cell sequencing techniques ought to address these concerns and provide more insights into the neuronal functions of 5hmC under both physiological and pathological conditions.

Supplementary Information

Below is the link to the electronic supplementary material.

Conflict of interest

The authors declare no competing financial interests.