Abstract
Mutations in the activity-dependent transcription factor MEF2C have been associated with several neuropsychiatric disorders. Among these, autism spectrum disorder (ASD)-related behavioral deficits are manifested. Multiple animal models that harbor mutations in Mef2c have provided compelling evidence that Mef2c is indeed an ASD gene. However, studies in mice with germline or global brain knock-out of Mef2c are limited in their ability to identify the precise neural substrates and cell types that are required for the expression of Mef2c-mediated ASD behaviors. Given the role of hippocampal neurogenesis in cognitive and social behaviors, in this study we aimed to investigate the role of Mef2c in the structure and function of newly generated dentate granule cells (DGCs) in the postnatal hippocampus and to determine whether disrupted Mef2c function is responsible for manifesting ASD behaviors. Overexpression of Mef2c (Mef2cOE) arrested the transition of neurogenesis at progenitor stages, as indicated by sustained expression of Sox2+ in Mef2cOE DGCs. Conditional knock-out of Mef2c (Mef2ccko) allowed neuronal commitment of Mef2ccko cells; however, Mef2ccko impaired not only dendritic arborization and spine formation but also synaptic transmission onto Mef2ccko DGCs. Moreover, the abnormal structure and function of Mef2ccko DGCs led to deficits in social interaction and social novelty recognition, which are key characteristics of ASD behaviors. Thus, our study revealed a dose-dependent requirement of Mef2c in the control of distinct steps of neurogenesis, as well as a critical cell-autonomous function of Mef2c in newborn DGCs in the expression of proper social behavior in both sexes.
Significance Statement
Autism spectrum disorder (ASD) is a neurodevelopmental disorder causing significant social, communication, and behavioral deficits in children worldwide. Genetic complexity and heterogeneity associated with ASD have hindered the field from establishing any precise cellular substrate associated with ASD. Recent studies have implicated hippocampal neurogenesis as one of the key players in social behavior and ASD-like behaviors. Here, using conditional deletion of transcriptional factor Mef2c in hippocampal newborn neurons or adult-born DGCs (abDGCs), we have demonstrated how Mef2c impacts behavior by regulating the structural development, physiology, and function of abDGCs. Our results revealed the essential role of Mef2c in neurogenesis and identified hippocampal neurogenesis as a neural substrate necessary for social behaviors.
Introduction
Autism spectrum disorder (ASD) is a family of neurodevelopmental disabilities that is characterized by social deficits, repetitive or stereotyped movements, and communication impairment. According to the CDC, 1 in every 54 children in the United States is diagnosed with ASD, and the prevalence of ASD is 4.3 times higher in boys than in girls. Over the past three decades, the diagnosis of ASD has not significantly advanced, and no proven cures have been developed (Medavarapu et al., 2019), mainly because the precise mechanisms and neural substrates underlying ASD have not been fully established.
One major factor that has hindered the field from establishing the precise etiology of ASD is the genetic complexity and heterogeneity of autism. Hundreds of genes have been implicated in autism, and a wide range of severity has been associated with each genetic mutation (Ey et al., 2011). Recent genetic analyses have provided critical clues to unraveling a “unifying pathway” that links various autism genotypes to common autism phenotypes (Geschwind and Levitt, 2007; Parikshak et al., 2013); the majority of autism-associated genes control common aspects of neurobiology, namely, neuronal development, migration, and connectivity (Voineagu et al., 2011; Delorme et al., 2013). The implication is that impaired structural and functional development of neurons can lead to autism-specific cognitive and social impairments (Geschwind, 1965). Mef2c (myocyte enhancer factor 2c) falls into this category because children (Paciorkowski et al., 2013; Rocha et al., 2016) and animal models (Tu et al., 2017; Harrington et al., 2020) harboring Mef2c germline and/or de novo mutations (Le Meur et al., 2010; Wan et al., 2021) show autistic behaviors as well as altered neuronal development. Mef2c is a transcription factor that belongs to the MADS (MCM1, agamous, deficiens, and serum) response factor box family of transcription factors (Dietrich, 2013; Rashid et al., 2014; Chen et al., 2017a), and its function has been identified as an activator or repressor in a context-dependent manner (Rashid et al., 2014; Di Giorgio et al., 2017; Sebastian et al., 2018; Pereira et al., 2020). Four Mef2 isoforms, including Mef2a, Mef2b, Mef2c, and Mef2d have been identified, and they have distinct but partially overlapping temporal and spatial expression patterns in the developing and adult brain (Lyons et al., 1995; Dietrich, 2013). The functions of the Mef2 family have been identified in the context of neuronal development and several neuropsychiatric disorders, including ASD (Lipton et al., 2009; Harrington et al., 2016; Zhang and Zhao, 2022). Among these, MEF2C haploinsufficiency is associated with global developmental and motor delay, seizures, repetitive behaviors, as well as social deficits, some of which are manifested in ASD (Barbosa et al., 2008; Lipton et al., 2009; Harrington et al., 2016; Chen et al., 2017a; Assali et al., 2019; Borlot et al., 2019; Moyses-Oliveira et al., 2020; Zhang and Zhao, 2022). Indeed, an association of MEF2C mutations with some patients who are diagnosed with ASD has been identified (Bourgeron, 2015; Wang et al., 2018). Several genetically engineered mice harboring Mef2c mutations have recapitulated ASD-related behaviors, further supporting that Mef2c is an ASD gene (Barbosa et al., 2008; Harrington et al., 2016, 2020; Tu et al., 2017). However, global or brain-wide mutations in Mef2c limit the ability to link precise neural substrates that transmit Mef2 mutations to the presentation of ASD-related behaviors.
Recent studies have implicated hippocampal neurogenesis as one of the key players in social behavior and ASD-like behaviors (Rubin et al., 2014; Montagrin et al., 2018; Cope et al., 2020). The continuous production and integration of newborn neurons provides ample plasticity to the hippocampus, which is critical for cognitive function, emotional stability, addictive behaviors, and social behaviors (Eisch and Harburg, 2006; Sahay and Hen, 2007; Lagace et al., 2010; Castilla-Ortega et al., 2016; Anacker and Hen, 2017; Lee et al., 2019; Toda et al., 2019). Neurogenesis is a tightly regulated process (Suh et al., 2009), and a combination of both intrinsic programs and environmental factors determines not only the number of newborn neurons but also their structural and functional development and circuitry integration (Song et al., 2012, 2016; Gonçalves et al., 2016a; Oppenheim, 2019; Abbott and Nigussie, 2020). Disruption of any of these processes results in impaired cognition, emotion, and social behaviors, providing compelling evidence that hippocampal neurogenesis may be a key player in the manifestation of ASD-related behaviors (Guo et al., 2011; Stephenson et al., 2011; Amiri et al., 2012; Hussaini et al., 2014; Sun et al., 2019; Bicker et al., 2021; Haan et al., 2021). Given the expression of Mef2c in dentate granule cells (DGCs) and the role of Mef2c in synapse formation and elimination (Mao et al., 1999; Flavell et al., 2006; Harrington et al., 2016; Cosgrove et al., 2021), we hypothesized that hippocampal neurogenesis may be a critical neural substrate that transmits Mef2c deficiency into the expression of ASD-related behaviors.
To test this hypothesis, we employed a retrovirus-mediated gene transfer method and transgenic mice that allowed us to specifically manipulate the Mef2c gene dose in adult-born DGCs (abDGCs) and investigated the role of Mef2c in structure, function, and behavior. This approach identified the dose-dependent role of Mef2c in the development of abDGCs. While a higher gene dose of Mef2c than normal arrested neurogenesis at Sox2+ neuronal progenitors, Mef2c deficiency interfered with structural maturation and spine formation of abDGCs, which underlie decreased excitatory synaptic transmission to abDGCs. Most importantly, specific deletion of Mef2c in abDGCs resulted in sociability and social recognition deficits, which are hallmarks of ASD behavior. Thus, our results revealed the essential role of Mef2c in neurogenesis and identified hippocampal neurogenesis as a neural substrate necessary for social behaviors.
Materials and Methods
Animals
All experimental protocols were approved by the Institutional Animal Care and Use Committee of the Cleveland Clinic Foundation and Augusta University Medical College of Georgia. C57BL/6 (stock no. 000664), Mef2cfl/fl or Mef2ctm1Jjs/J (stock No. 025556), Ai3 or B6.Cg-Gt(ROSA)26Sortm3(CAG-EYFP)Hze/J (stock no. 007903), and Ascltm1.1(Cre/ETR2)Jejo/J (stock no. 012882) were purchased from Jackson Laboratory. These mice were used to breed in-house to generate double (Ascl1-CreERT2; Mef2cflox/flox and Ascl1-CreERT2; Ai3+/+) and triple transgenic mice (Ascl1-CreERT2; Mef2cflox/flox/Ai3+/+). The mice were kept in a 12 h light/dark cycle, with food and water available ad libitum. All experimental procedures occurred during the light phase using both sexes.
Experimental design
Retrovirus injection
To label hippocampal newborn neurons, mice were anesthetized with 100 mg/kg ketamine plus 10 mg/kg xylazine, and a reporter-expressing retrovirus [RV-CAG-RFP; 108 CFU/ml; RV-Mef2c-tPT2A-GFP (Addgene plasmid #111771) or RV-CAG -GFP-IRES-Cre 5–8 × 107 CFU/ml] was injected bilaterally into the hippocampus. The following coordinates relative to bregma were used: anterior to posterior (AP), −1.7 mm; medial to lateral (ML), ±1.4 mm; and dorsal to ventral (DV), −2.35 mm. One microliter of virus was injected into each site through a Hamilton syringe with a 33-gauge needle. The injection speed was 0.1 μl/min. The needle was retained in place for 5 min before retraction.
Tamoxifen injection
Tamoxifen (180 mg/kg, i.p., dissolved in a 1:10 mixture of ethanol and corn oil, MilliporeSigma) was administered to Ascl1-CreERT2; Mef2cfl/fl, Mef2cfl/fl mice (control littermates) and Ascl1-CreERT2; Ai3fl/fl or Ascl1-CreERT2; Mef2cflox/flox/Ai3fl/fl mice for 3 consecutive days in order to induce cre-mediated recombination and deletion of Mef2c only from newborn neurons in Ascl1-CreERT2; Mef2cfl/fl or Ascl1-CreERT2; Mef2cflox/flox/Ai3fl/fl mice when they were 6–8 weeks old.
Immunohistochemistry
The following antibodies were used in this study: GFAP (rabbit and/guinea pig, Dako and Advanced Immuno; 1:1,000); SOX2 (goat, Santa Cruz Biotechnology, and rabbit, Chemicon 1:500); DCX (goat, Santa Cruz Biotechnology, 1:500); NEUROD1 (goat, Santa Cruz Biotechnology, 1: 200); MEF2C (rabbit, Abcam, and cell signaling, 1:1,000); PROX1 (rabbit, Chemicon, 1:500); TBR2 (rabbit, Abcam, 1:200); and NEUN (mouse, Chemicon, 1:1,000); secondary antibodies (AF555/AF647-donkey anti-goat, AF488/AF555/AF647-donkey anti-rabbit, AF 488/AF 647-donkey anti-mouse, AF 488/AF 647-donkey anti-guinea pig) were purchased from Jackson ImmunoResearch. Briefly, animals were deeply anesthetized with a cocktail of ketamine (300 mg/kg) and xylazine (30 mg/kg) and transcardially perfused first with 0.9% saline and subsequently with 4% paraformaldehyde (PFA). Brains were dissected and postfixed overnight in PFA at 4°C. Brains were cryoprotected in 30% sucrose for 48 h at 4°C and sectioned at 40 μm with a sliding microtome (Leica SM2010 R). Floating tissue sections were washed 3 times in tris-buffered saline (TBS) and blocked in TBS-Tween (TBST) containing 3% donkey serum. Tissue was then incubated with primary antibodies (made in the same blocking solution) overnight at 4°C. Tissues were then washed 3 times for 15 min in TBST before incubation with secondary antibodies at 1:400 for 2 h. Tissues were then washed 3 times in TBST for 15 min. Finally, tissues were rinsed with PBS, counter-stained with DAPI (Sigma-Aldrich, D9542), and mounted.
Morphological analysis
For quantification of MEF2C colocalization with different neurogenesis markers (Staging) and MEF2C overexpression (Mef2cOE) experiments, images were taken at 40× with a zoom factor 3 using a Zeiss confocal microscope in three separate animals per experiment. For the MEF2C expression pattern in newborn neurons using retrovirus RV-CAG-RFP, 20× and 60× images were taken from three animals (30–50 cells quantified from each animal) at 2, 4, 8, and 12 weeks after virus injection (which corresponded to the ages of the newborn neurons). Dendritic length and spine analysis methods pertaining to studying the morphology of newborn neurons were described in detail previously (Lee et al., 2019). Briefly, for morphological analysis, z-series images at 1 μm intervals were captured for GFP-expressing neurons under a 40× oil objective using a Zeiss confocal microscope. Quantification and analysis were performed using ImageJ and software with the NeuronJ plugin for measurements of dendritic length. Sholl analysis was performed with a semi-manual method, where the concentric circles correlating to each branch point intersection were generated using ImageJ and Sholl plugin and the total number of intersections per 5 μm was counted manually. For analysis of dendritic spine density, z-series images with 0.1 μm intervals were captured under a 63× oil objective with a digital zoom factor of 4. The distal, middle, and proximal segments of the dendrite were captured, and the linear spine density was determined. The maximum z-projection images were opened in ImageJ software to measure the length of each segment. The number of spines on each segment was counted manually and is presented as the number of spines per micrometer. Spines were categorized into mushroom, thin, or stubby spines on the basis of three parameters, including spine length, diameter of the neck, and diameter of the spine head. Spines were scored as mushrooms if the diameter of the head was 3 times larger than that of the neck, as stubby spines if the diameter of the spine head was greater than the spine length, or as thin spines when the spine length was greater than the diameter of the head.
Brain slice preparation and ex vivo electrophysiology
Brain slices containing the hippocampus were prepared as described previously (Hong et al., 2019; Kang et al., 2020). Briefly, mice were deeply anesthetized using ketamine/xylazine (100 mg/10 mg/kg, i.p.), and the brain was rapidly removed and placed in ice-cold sucrose-based artificial cerebrospinal fluid (aCSF) containing the following (in mM): 75 sucrose, 87 NaCl, 2.5 KCl, 11.25 NaH2PO4, 7 MgCl2, 0.5 CaCl2, 25 NaHCO3, 0.3 L-ascorbate, and 25 glucose, and oxygenated with 95% O2/5% CO2. Coronal (350 μm) slices were cut with a vibrating compresstome (VF-310-0Z, Precisionary Instruments), then transferred to a holding chamber, and incubated for 30 min at 34°C and kept for at least 1 h at room temperature (24–25°C) in carbonated (95% O2/5% CO2) standard artificial cerebrospinal fluid (aCSF) containing the following (in mM): 126 NaCl, 2.5 KCl, 1.25 NaH2PO4, 1 MgCl2, 2 CaCl2, 25 NaHCO3, and 11 glucose. After equilibration, a single slice was transferred to a submersion-type recording chamber and mechanically stabilized with an anchor (Warner Instruments, Hamden, CT). Electrical signals were recorded with an Axon 700B amplifier, a Digidata 1550B A/D converter, and Clampfit 11.0 software (Molecular Devices). Throughout the experiments, the bath was continually perfused with warm (32°C) carbonated aCSF (2.0–2.5 ml/min). Patch pipettes had a resistance of 4–6 MΩ when filled with a solution containing (in mM): 140 Cs-methanesulfonate, 5 KCl, 2 MgCl2, 10 HEPES, 2 MgATP, and 0.2 Na2GTP for voltage clamping. The pH was adjusted to 7.2 with Tris-base and the osmolality to 310 mOsmol/L with sucrose. The newborn neurons were identified by the expression of GFP fluorescence. mEPSCs and mIPSCs were recorded in the presence of tetrodotoxin (0.5 μM) at a holding potential of −70 and 0 mV, respectively. The EPSCs were confirmed pharmacologically with the AMPA receptor antagonist DNQX (20 μM) and the NMDA receptor antagonist AP5 (50 μM). The IPSCs were confirmed with the GABAA receptor antagonist Picrotoxin (PTX, 100 µM).
Behavioral protocols
All behavioral tests were performed in the Rodent Behavioral Core of the Cleveland Clinic during the animal's light cycle phase. Age-matched littermates, using both male and female sexes, at 14–18 weeks of age were used for these studies as controls. All behavioral tests were performed 8–12 weeks after tamoxifen injection for two separate cohorts, except for the open field and fear conditioning, which were done only at the 12-week time point. Each behavioral paradigm was separated by a day. All home cages containing mice were transported to the testing room at least 30 min prior to the start of testing. The tests were scored using Ethovision software.
Three-chamber social interaction
This test comprised a three-chamber white opaque box with openings for the mice to pass through each chamber. The center region length was 22.5 cm, the two sides measured 18.5 cm each, and the two side chambers had two empty pencil cup holders. Each test consisted of 30 min. The first 10 min was for habituation where each experimental mouse when placed inside the center explored all three chambers for 10 min. After 10 min, the mice were taken out and placed in a cage, during which one novel mouse that the experimental mouse had never interacted with before (stimulant mouse 1) was placed inside the pen holder on one side, while a Lego object was placed on the other side's pen holder. The experimental mouse was allowed to explore for 10 min, after which it was returned to the cage again. Then, the Lego was replaced by another novel mouse (stimulant mouse 2), and the experimental mouse was again placed in the center of the three-chamber box and explored the chambers for 10 more minutes. After 10 min, the experimental mice were taken out and placed back in their home cage. The chambers and the pen holders were thoroughly cleaned with trifectant between each experiment. The object and the mice were alternately switched between each experimental mouse so as to remove potential bias from behavior.
Fear conditioning
Mice were initially exposed to a neutral context with metal grid floors for a 3 min habituation period, followed by the presentation of a 30 s tone (conditioned stimulus CS; 90 dB white noise) that co-terminated with a 2 s 1.2 mA footshock (unconditioned stimulus; UCS). Following the shock, the rodent was given a 2 min inter-trial interval before receiving a second tone and footshock pairing, followed by a 2 min post-shock “recovery” period. Freezing was recorded using freeze frame software. Chambers were cleaned with trifectant between each experiment. For the contextual memory test, 24 h later the mice were exposed to the same context for 5 min, without the tone or shock. Three hours later, cued fear memory was tested by placing the mice in a completely different context than the training day (different floor; white plastic and vanilla odor) for 6 min. The first 3 min were for habituating to the novel environment, followed by 3 min of constant presentation of the control stimulus (white noise).
Open field test
During this test, mice were first placed (from one corner) inside a white opaque square box, 45 cm in width and height, and allowed to explore for 10 min. For the analysis, the white square box was divided into 26 zones, each approximately 9 cm in diameter. The exploration time in the outer 16 zones was considered as the periphery, while the inner 10 zones were considered as the time spent in the center.
Novel object recognition test
This test was comprised of the same white opaque box used in the open field test, where two similar objects were first placed at two corners (glass bottle with bedding), and the mice were allowed to explore this arena with the objects freely for 5 min; this was the habituation phase (each time the experimental mice were placed inside the box from a particular corner away from the two objects). After 1 h, they were again placed back into the same arena with the same objects for 5 min; this was the familiarization phase. Four hours after the habituation and familiarization phases, one of the objects was replaced by a novel object (Lego), and the mice were allowed to explore the arena with the familiar and the new object for 10 min. Each time, the side for the familiar object (bottle with glass bedding) and the novel object (Lego) were exchanged alternately so as to remove potential bias.
Statistics
All statistical analyses were performed using commercial software (GraphPad Prism). Data were analyzed either by Student's t test or by two-way ANOVAs. Bonferroni's multiple comparison post hoc test was applied to compare differences among groups. The significance threshold was placed at α = 0.05, and corrections for multiple comparisons are reflected in p values. Detailed descriptive statistics are presented in Table 1.
Descriptive statistics
Results
MEF2C is expressed in DGCs, but not in neural stem cells
Neural stem/progenitor cells (NSCs) continuously produce new cells. Newly generated cells transit multiple stages and become DGCs that are integrated into existing hippocampal neural circuits (Suh et al., 2009; Song et al., 2016; Gage, 2019). To understand the temporal expression pattern of MEF2C, we used cell type-specific markers to define the developmental stages that express MEF2C during neurogenesis (Fig. 1A). Mef2c was deleted in DGCs by injecting a CRE-expressing retrovirus into the dentate gyrus (DG) of the hippocampus of Mef2cfl/fl mice, and the specificity of MEF2C antibody was confirmed by showing a significant reduction in MEF2C expression in Mef2c-deficient DGCs (Fig. 1B,C). Type 1 NSCs show a characteristic radial morphology and express GFAP or NESTIN in their radial processes (Suh et al., 2007, 2009; Song et al., 2016). MEF2C was not expressed in radial glial-like GFAP+ cells (Fig. 1D) or GFP+ cells in Nestin-GFP transgenic mice in which GFP is expressed under the control of NSC-specific promoter Nestin (Mignone et al., 2004) (Fig. 1E). However, the expression of MEF2C was initially detected in a subset of SOX2+ Type 2 cells (Suh et al., 2007; Fig. 1F), as well as 40–50% of NEUROD1+ or DCX+ cells that represent neuroblasts or immature neurons (Fig. 1G,H). In the DG, 100% of NEUN+ cells co-expressed MEF2C (Fig. 1I). Collectively, these results show that MEF2C expression is initiated at the later stage of SOX2+ Type 2 progenitors, gradually increases in neuroblasts (NEUROD1+) and immature neurons (DCX+), and is sustained in fully differentiated DGCs (Fig. 1J).
MEF2C is mainly expressed in DGCs, but not in neural stem cells. A, Cartoon illustrating the progression of neurogenesis and the temporal expression of corresponding stage-specific markers. B,C, RV-CAG-GFP-IRES-Cre injection into Mef2cflox/flox mice deleted Mef2c in ∼80% of infected (GFP+) abDGCs. D,E, Immunohistochemistry with neural stem cell (NSC) markers showed that MEF2C is not expressed in GFAP+ or Nestin+ radial glial-like cells. F–H, A subset of SOX2+, NEUROD1+, and DCX+ cells showed colocalization with MEF2C. I, MEF2C was expressed in 100% of NEUN+ DGCs. D–I, MEF2C, stage-specific markers, and DAPI counterstaining are shown in green, red, and blue, respectively. Arrows indicate the location of stage-specific markers. J, Percentage of MEF2C+ cells among stage-specific markers shown in A–F. For each cell stage marker, 30 cells from three mice were examined. K, Experimental paradigm to determine the temporal expression of MEF2C during neurogenesis. L,M, Retrovirus-mediated cell birth date labeling showed that MEF2C expression (green) was detected in approximately 50% of 2-week-old abDGCs (red) and was sustained in 100% of abDGCs older than 4 weeks. RFP+ cells (n = 30–50) from three mice were examined for colocalization with MEF2C. Numbers in parentheses represent cell numbers. The scale bar is 5 μm.
To assess the expression of MEF2C during the differentiation of newborn cells into DGCs, we labeled newborn cells using a replication-incompetent retrovirus (RV) expressing a red fluorescent protein (RFP) and examined the expression of MEF2C during a time course of neuronal development of newborn cells (Fig. 1K). Because an RV only infects proliferating cells, we defined the time between the RV injection and examination as the maximal age of newborn cells. This age-synchronized labeling method revealed that almost 50% of 2-week-old abDGCs (RFP+) were colocalized with MEF2C; however, 100% of mature abDGCs older than 4 weeks expressed MEF2C (Fig. 1L,M). Thus, MEF2C was expressed in neuronally committed cells and mature DGCs, consistent with the results from MEF2C immunohistochemistry with cell type-specific markers.
Mef2C overexpression arrests neuronal differentiation
To understand the gain-of-function effects of MEF2C on the development of newborn neurons, we overexpressed Mef2c (Mef2cOE) in newborn neurons using RV-mediated gene delivery. The RV used was RV-Mef2c-tPT2A-GFP, which expresses both MEF2C and a green fluorescent protein (GFP) reporter gene (Liu et al., 2017). This RV was injected into the DG to overexpress MEF2C in abDGCs, and the effects of Mef2cOE were compared to control mice that received a RFP- or GFP-expressing RV (RV-GFP or RV-RFP, pseudo-colored to green in the representative image) at 2, 4, and 8 weeks postinjection (WPI) (Fig. 2A). While only a small population of Mef2cOE newborn cells maintained the typical morphology of DGCs with radially developed dendrites, the majority of Mef2cOE newborn cells displayed abnormal structural development (Fig. 2B,C). Mef2cOE newborn cells were clustered and failed to project their processes into the molecular layer, instead projecting processes either parallel to the granular layer or into the hilus, which is reminiscent of the morphology of immature stages of newborn cells (referred to as “projection” impairment; Fig. 2B,C). Moreover, some Mef2cOE newborn cells migrated to the hilus instead of the granular layer (referred to as “migration” impairment; Fig. 2B,C). The onset of both projection and migration impairments in Mef2cOE newborn cells was at 2 WPI, and such morphological deficits were sustained up to 8 weeks from birth, which is the longest time period that we examined (Fig. 2B,C).
Mef2c overexpression halts neurogenesis. A, Experimental scheme to examine the effects of Mef2c overexpression on the structure and state of abDGCs. B,C, The majority of Mef2cOE newborn cells (green) displayed impaired projection of processes and aberrant migration up to 8 weeks after birth. Cells (n = 50–100) from three mice were examined. D,E, While a subset of Mef2cOE cells differentiated into PROX1+ DGCs, the majority of Mef2cOE cells that showed impaired projection and migration expressed SOX2, but not neuronal or astrocyte lineage-specific markers, such as TBR2, NEUN, PROX1, or GFAP, respectively. In comparison, control or RV-GFP/RFP-injected mice, 2, 4, and 8 WPI DGCs properly differentiated into PROX1+ and NEUN+ DGCs. Representative images of eight WP WT DGCs co-localizing with just PROX1 and NEUN and not with other lineage-specific markers. Per marker, n = 15 cells from three mice were examined. This arrest of neurogenesis of Mef2cOE cells at progenitor stages was sustained for at least 8 weeks after the birth of newborn cells. Mef2cOE cells and WT cells are shown in green, and lineage-specific markers are shown in magenta or red (SOX2). Numbers in the parentheses represent cell numbers.
To define the cell types of Mef2cOE newborn cells, neuronal developmental stage-specific markers were used. A small population of Mef2cOE cells that appeared to develop normally showed colocalization with PROX1, a marker for DGCs (Fig. 2D). Surprisingly, Mef2cOE cells with structural and migratory impairments were predominantly colocalized with SOX2, but not with any other cell type-specific markers, including GFAP, Tbr2, NEUN, or PROX1 (which represent astrocytes, immature and mature neurons, respectively; Fig. 2D,E). SOX2 co-expression in Mef2cOE cells was sustained up to 8 WPI, indicating that Mef2cOE sustained the expression of the neuronal progenitor marker, SOX2, irrespective of the ages of newborn cells (Fig. 2D,E). These results indicate that Mef2cOE did not simply delay neurogenesis, but it arrested newborn cells at the SOX2+ neuronal progenitor stage and blocked neuronal differentiation of newborn cells.
MEF2C is required for the structural development of newborn neurons
To understand the essential role of MEF2C in the development of abDGCs, Mef2c was specifically deleted in abDGCs. We injected a RV that expresses both a CRE recombinase and GFP (RV-GFP-IRES-Cre) into the DG of Mef2cflox/flox and wild-type (WT) control mice and analyzed them at 4, 8, and 12 WPI (Fig. 3A). In mice in which Mef2c was conditionally knocked out (Mef2ccKO), newborn cells successfully differentiated into abDGCs; however, the total dendritic length (Fig. 3B–G) and dendritic arborization (Fig. 3C,E,G) of Mef2ccKO abDGCs were significantly reduced in 4-, 8-, and 12-week-old abDGCs (total length, unpaired t tests, p = 0.014 at 4 WPI; p = 0.0008 at 8 WPI; and p < 0.0001 at 12 WPI; dendritic arborization, two-way ANOVA; main effect, genotype, p < 0.0001, p = 0.0001, and p < 0.0001 for 4, 8, and 12 WPI, respectively).
MEF2C is required for the structural and synaptic development of newborn neurons. A, Experimental design to delete Mef2c in hippocampal abDGCs and timeline of analysis. B–G, Specific deletion of Mef2c in abDGCs reduced the total dendritic lengths and arborization of 4-week-old (B,C, total dendritic length, n = 13; Sholl analysis, n = 9), 8-week-old (D,E, total dendritic length, n = 16; Sholl analysis, n = 13), and 12-week-old (F,G, total dendritic length, n = 16; Sholl analysis, n = 9) abDGCs. H–M, Specific deletion of Mef2c in abDGCs did not change dendritic spine densities in 4-week-old abDGCs (I, n = 28/group); however, dendritic spine densities of 8-week-old (K, n = 24/group) and 12-week-old (M, n = 24/group) Mef2ccKO abDGCs were significantly reduced regardless of the locations of the spines along the dendrite. N, Representative images showing different types of spines that include mushroom, thin, and stubby spines. O, Mef2c deficiency in abDGCs decreased and increased the densities of mushroom and thin spines of 4-week-old abDGCs, respectively, while those of stubby spines were unchanged (n = 8/group). P,Q, Both 8-week-old (n = 12/group) and 12-week-old (n = 8/group) Mef2ccKO abDGCs showed increased densities of thin spines along the dendrite. For Sholl analysis, two-way ANOVA with Bonferroni's multiple comparison tests. For the other experiments, Student's t test, *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. Numbers in the parentheses represent cell numbers. Data are represented as mean ± SEM.
Next, we examined the function of Mef2c on dendritic spine formation. Because neurons from the lateral and medial entorhinal cortex (EC) and hilar mossy cells project onto the outer (distal), middle (medial), and inner (proximal) regions of dendrites of abDGCs (Amaral et al., 2007), spine densities of abDGCs were examined at the different segments of dendrites (Golub et al., 2015; Lee et al., 2019). Four-week-old Mef2ccKO abDGCs did not show any significant differences in spine density in any of the three dendritic segments (unpaired t tests; p > 0.05 at inner, middle, and outer segments; Fig. 3H,I); however, the spine densities in all three dendritic segments of abDGCs were significantly increased in 8- and 12-week-old Mef2ccKO abDGCs (Fig. 3J–M; 8 WPI, unpaired t tests, p < 0.00001 for all segments; 12 WPI, unpaired t test, p = 0.004 in inner, p = 0.006 in middle, and p = 0.006 in outer dendrites), suggesting a role of Mef2c in spine development at the later stages of neuronal development of abDGCs. We further analyzed the types of dendritic spines in Mef2ccKO abDGCs. The loss of Mef2c decreased the densities of mushroom spines in all three segments at 4 WPI, but the differences were not significant at later stages of development except in the proximal dendrites at 12 WPI (Fig. 3O; p = 0.002 for inner, p = 0.00003 for middle, and p = 0.01 for outer segments at 4 WPI). Interestingly, the densities of filopodial or thin spines instead increased at 8 WPI, and the increased densities of filopodia were sustained at 12 WPI of Mef2ccKO abDGCs. The increased densities of these thin spines were evident regardless of dendritic segment of abDGCs (Fig. 3N–Q; unpaired t tests; p = 0.000002, inner and middle segments; p = 0.0036, outer segments at 8 WPI; p = 0.000001 inner segments; p = 0.009, middle segments; p = 0.01, outer segments at 12 WPI).
Mef2c deficiency leads to reduced synaptic transmission to abDGCs
To understand the impact of impaired spine formation on the synaptic transmission onto Mef2ccKO abDGCs, we performed ex vivo electrophysiology. WT and Mef2cflox/flox mice were injected with RV-CAG-GFP-IRES-Cre, and a whole-cell patch clamp was performed on abDGCs between 8 and 12 WPI (Fig. 4A–C). Compared to age-matched control abDGCs, the frequency, but not amplitude, of miniature excitatory postsynaptic currents (mEPSCs) was significantly reduced in Mef2ccKO abDGCs (unpaired t tests; p = 0.0182 for frequency and p = 0.3532 for amplitude) [Fig. 4D,E; n = 9 cells/group (6 mice/group)]; however, neither the frequency nor amplitude of miniature inhibitory postsynaptic currents (mIPSCs) were altered in Mef2ccKO abDGCs [unpaired t tests; p = 0.6460 for frequency and p = 0.9341 for amplitude; Fig. 4F,G; n = 9 cells/group (6 mice/group)]. These results show that Mef2c deletion specifically reduced excitatory synaptic transmission onto Mef2ccKO abDGCs.
Mef2c deficiency leads to reduced synaptic transmission onto abDGCs. A, Experimental design to delete Mef2c in abDGCs and perform whole-cell patch clamping. B,C, CCD Camera captured infrared (B,C left) and fluorescence (C middle with ET-GFP filter and C right with ET-DAPI filter) images showing the recorded abDGCs in the whole-cell recording. Scale bar = 200 μm (B) and 20 μm (C). D,E, mEPSC frequencies, but not amplitudes, of Mef2ccKO abDGCs were significantly decreased in Mef2ccKO abDGCs (nl = 9 cells/group [6 mice/group]). F,G, The absence of Mef2c did not impact the frequencies or amplitudes of mIPSCs of abDGCs [n = 9–10 cells/group (6 mice/group)]. Numbers in the parentheses represent cell numbers. Data are represented as mean ± SEM. Student's t test, *p < 0.05.
The absence of Mef2c in abDGCs is sufficient for the expression of ASD behaviors
Consistent with an association between MEF2C and ASD, Mef2c heterozygote mice display ASD behaviors (Harrington et al., 2016; Anacker and Hen, 2017). Given the implication of MEF2C in ASD and the role of hippocampal neurogenesis in ASD behaviors, we investigated how an absence of Mef2c in abDGCs would affect cognitive and social behaviors, the latter being a prominent characteristic of ASD.
We generated double transgenic mice that harbor Ascl1-CreER and Mef2cflox/flox alleles, Ascl1-CreERT2;Mef2cflox/flox, which allowed us to delete Mef2c in abDGCs in a temporal-specific manner (Kim et al., 2007). Mef2cflox/flox littermates were used as controls (Fig. 5A). Consistently with a previous report (Kim et al., 2007), we confirmed that tamoxifen-mediated activation of CreER induced the expressions of EYFP in abDGCs when we tested with a separate cohort of Ascl1-CreERT2;ROSA26 reporter mice (Fig. 5B). Tamoxifen was injected into 6- to 8-week-old Ascl1-CreERT2;Mef2cflox/flox (Mef2ccKO) and control mice, and then mice were subjected to three-chamber social interaction and social novelty tests 8–12 weeks after tamoxifen injection. Both control and Mef2ccKO did not show any side preference when objects were presented in both chambers (Fig. 5C,D). While control mice spent more time with live mice than with objects, Mef2ccKO mice displayed no preference for live mice over objects, showing comparable interaction times (Fig. 5E,F; two-way ANOVA, main effect: social objects, p = 0.004). Subsequently, the object was replaced with a novel mouse, and then mice were subject to interact with a novel mouse (that replaced the object) or a familiar mouse (the same that was used for the sociability test). As expected, control mice spent significantly more time with the novel mouse; however, Mef2ccKO mice did not show any preference for the novel mouse, and the interaction times of Mef2ccKO mice with familiar and novel mice were not significantly different (two-way ANOVA, main effect: social novelty, p = 0.005; Fig. 5G,H), indicating both sociability and social novelty discrimination deficits in Mef2ccKO mice.
The absence of Mef2c in abDGCs is sufficient for the expression of ASD-like behaviors. A, Experimental design to produce Mef2ccKO and to perform behavioral tests. B, Expression of EYFP in Ascl1-CreER; Mef2cflox/flox; Ai3flox/flox showing that EYFP is expressed only in the hippocampal DGCs. C,D, Both control and Mef2ccKO had no bias towards any one side in the three-chamber sociability test. E,F, While control mice spent more time with live mice than objects, Mef2ccKO mice failed to display social interaction (control, n = 22; Mef2ccKO, n = 21). F,G, Control mice spent more time with novel mice than familiar mice; however, Mef2ccKO mice displayed an impaired preference in social novelty recognition (control, n = 22; Mef2ccKO, n = 21). I–K, Mef2ccKO mice showed deficits in contextual memory, showing decreased freezing time when they were re-exposed to the same context where they had received foot shocks (control, n = 17; Mef2ccKO, n = 16). However, both Mef2ccKO and control mice showed comparable amounts of freezing time when the tone that had been associated with foot shocks was applied in a context that was not associated with the foot shocks (control, n = 17; Mef2ccKO, n = 16). L,M, Mef2ccKO mice did not show any differences in recognition memory in the novel object recognition test compared to control mice (control, n = 20; Mef2ccKO, n = 20). O,P, Mef2ccKO mice did not show any differences in preference between the center and peripheral zones compared to control mice (control, n = 17; Mef2ccKO, n = 16). N,Q, Both control and Mef2ccKO moved comparable distances in the novel object recognition test and the open field test. Two-way ANOVA with Bonferroni's multiple comparison tests (C,E,H,J,M), Student's t test (G,K). *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. Numbers in the parentheses represent animal numbers. Data are represented as mean ± SEM.
Next, we tested the contextual and cued fear memory of Mef2ccKO mice. Mice were first trained to associate white noise with foot shocks in a specific context, and then we tested whether the context or tone could induce freezing behavior without the actual presentation of foot shocks (Fig. 5I). When mice were re-exposed to the same context where they had received the foot shock, Mef2ccKO mice showed significantly less freezing time compared to control mice (unpaired t test, p = 0.009; Fig. 5J). However, tone-induced freezing behavior did not show any significant difference compared to that of control mice (two-way ANOVA, main effect: genotype, p < 0.0001 for both control and Mef2ccKO; Fig. 5K), indicating that Mef2ccKO in abDGCs led to deficits in the ability to associate to context. In the novel object recognition test, both control and Mef2ccKO mice spent more time with the novel object than with the familiar object, revealing comparable levels of discrimination ability (two-way ANOVA, main effect: genotype, p = 0.0003 for control and p = 0.0031 for Mef2ccKO; Fig. 5L,M). The open field test showed that control and Mef2ccKO mice did not show any differences in time spent in the periphery and central zones (two-way ANOVA, main effect: genotype, p < 0.0001 for both control and Mef2ccKO; Fig. 5O,P). During the behavior tests, Mef2ccKO mice did not show any differences in motility compared to control mice (Fig. 5N,Q).
Discussion
In this study, we showed how a transcriptional factor, Mef2c, regulates the development and synaptic connectivity of hippocampal newborn neurons, and how the altered structure of abDGCs, in turn, disrupts hippocampus-dependent physiology, function, and behavior. The Mef2 gene family has distinct temporal and spatial expression patterns in the developing and adult brains and plays a critical role in neuronal development (Dietrich, 2013). Mef2c is highly expressed in the cortex, amygdala, and hippocampus, and it controls gene expression in an activity-dependent manner (Lyons et al., 2012; Dietrich, 2013; Assali et al., 2019; Chen et al., 2020a). In particular, Mef2c has been recognized to play an important role in the context of synapse elimination, since Mef2c deletion increases synaptic densities (Pfeiffer et al., 2010; Tsai et al., 2012; Hussaini et al., 2014). Interestingly, alterations in the chromosome containing Mef2c loci and genetic deletion of Mef2c have been associated with ASD, suggesting that synaptic and neuronal connectivity are key elements that determine autism pathobiology. Given that hippocampal neurogenesis is tightly regulated by activity and that disrupted hippocampal neurogenesis has often been associated with autism (Guo et al., 2011; Stephenson et al., 2011; Amiri et al., 2012; Hussaini et al., 2014; Haan et al., 2021; Gioia et al., 2022), it is reasonable to hypothesize that Mef2c may play an important role in ASD-related behaviors by regulating hippocampal neurogenesis. Although the contribution of the entire Mef2 gene family to hippocampal neurogenesis has been addressed (Latchney et al., 2015), the specific role of Mef2c in hippocampal neurogenesis, and its functional significance in physiology and behavior, has yet remained unanswered. Here, using conditional deletion of Mef2c in hippocampal newborn neurons or abDGCs, we have demonstrated how Mef2c impacts behavior by regulating the structural development, physiology, and function of abDGCs.
Hippocampal neurogenesis is sensitive to Mef2c gene dosage
During hippocampal neurogenesis, Mef2c expression was initially detected in a small population of SOX2+ NSCs, gradually increased in DCX+ neuroblasts/immature neurons, and was eventually maintained in DGCs. The specific overexpression (Mef2cOE) and deletion (Mef2ccKO) of Mef2c in the lineage of newborn cells revealed the gene dosage sensitivity of hippocampal neurogenesis to Mef2c. Our virus-mediated gene delivery system predominantly infects and transduces Mef2c in proliferating cells that subsequently give rise to abDGCs (Lee et al., 2019; Zhou et al., 2019; Zhang et al., 2020). Mef2cOE cells did not express neuronal lineage-specific markers but sustained the expression of SOX2, a marker for neural stem/progenitors (Suh et al., 2007), even at 8 weeks after birth, indicating that Mef2cOE maintained cells undifferentiated. This is reminiscent of the Mef2c function as a cardiac stem cell factor, which has been used to produce undifferentiated cardiac stem cells (Ieda et al., 2010), suggesting that Mef2cOE keeps cells from differentiating. Moreover, Mef2cOE induced abnormal development of processes and aberrant migration, indicating that a higher level of Mef2c not only stalled the progression of neurogenesis but also impaired structural development and migration of newborn cells. On the contrary, our retrovirus-mediated specific deletion of Mef2c did not interfere with the neuronal commitment of NSCs. Mef2ccKO cells differentiated into abDGCs, expressing neuronal lineage-specific markers such as DCX, NEUN, and PROX1 (data not shown); however, Mef2ccKO impaired the arborization of abDGCs and abnormally increased spine density. Specifically, densities of thin spines or filopodia were significantly increased, while those of mushroom spines were relatively unaltered in Mef2ccKO abDGCs. The increased densities of filopodia may be attributed to the function of Mef2c in synapse elimination as proposed previously (Pfeiffer et al., 2010; Tsai et al., 2012; Hussaini et al., 2014). This is also consistent with a previous observation that spine maturation is an activity-dependent process (Yasuda et al., 2021) and that lack of Mef2c may interfere with spine maturation or mushroom spine development. Collectively, our results revealed the requirement of a precise Mef2c dosage for tight control of multiple aspects of neurogenesis. The identification of molecular pathways that link Mef2c to activity-dependent synapse development of abDGCs and also a multiphoton-based in vivo microscopy approach to distinguish Mef2c function in spine elimination and formation or maturation from highly motile filopodia to stable spines are subjects of future research (Gonçalves et al., 2016b).
A cell-autonomous function of Mef2c in structure, physiology, and function of abDGCs
The function of Mef2c has been identified in multiple aspects such as neural stem cell differentiation (Lipton et al., 2009), activity-dependent survival (Mao et al., 1999), and maturation of newborn neurons, as well as synaptic transmission (Barbosa et al., 2008; Lipton et al., 2009; Harrington et al., 2016, 2020; Rajkovich et al., 2017). Several initial studies suggested a function of Mef2c in synapse elimination because an absence of Mef2c increased spine densities, and mEPSC frequency and intensities accordingly increased. Similarly, mEPSC frequencies of Mef2c-deficient DGCs in hGFAP-Cre;Mef2fl/fl mice have also been shown to increase (Barbosa et al., 2008). Consistent with this result, our study showed increased dendritic spine densities in Mef2ccKO abDGCs, revealing a critical role of Mef2c in structural development, as well as in determining the numbers and types of dendritic spines in abDGCs. However, Mef2ccKO also decreased frequencies of mEPSCs of abDGCs. This difference in electrophysiological properties of Mef2ccKO abDGCs may be due to the different contexts or experimental manipulations to delete Mef2c in DGCs. Since hGFAP-Cre induces the deletion of Mef2c in neural stem cells at early embryonic stages (Barbosa et al., 2008), it is expected that both pre- and postsynaptic neurons lack Mef2c. On the contrary, our viral or genetic manipulation resulted in the deletion of Mef2c exclusively in postsynaptic neurons. Indeed, the probability of neurotransmitter release increased at the pre-synaptic site in hGFAP-Cre;Mef2fl/fl mice, indicating that both increased spine densities at the postsynaptic site and short-term synaptic activities at the pre-synaptic site contributed to increased synaptic transmission in hGFAP-Cre;Mef2fl/fl mice. However, since our manipulation resulted in the deletion of Mef2c only in postsynaptic neurons, it is likely that intrinsic changes caused by the deletion of Mef2c led to reduced synaptic transmission onto abDGCs. Alternatively, this difference in mEPSC frequencies may reflect an environmental effect on Mef2c-deficient abDGCs. While all DGCs are expected to lack Mef2c in hGFAP-Cre;Mef2cfl/fl mice, thereby Mef2c-deficient abDGCs are surrounded by Mef2c-deficient DGCs, our manipulations produced Mef2c-deficient abDGCs surrounded by WT DGCs. It is therefore possible that cell-non-autonomous effects may be reflected in the function of Mef2c-deficient cells in hGFAP-Cre;Mef2cfl/fl mice. A previous study also suggested that environmental factors may influence the function of Mef2c-deficient cells. Genetic deletion of Mef2c in the majority of cortical excitatory neurons in Emx-Cre;Mef2cfl/fl mice slightly decreased the frequencies and amplitudes of mEPSCs; however, mosaic deletion of Mef2c by infection with Cre-expressing adeno-associated virus in the same cortical excitatory neurons significantly increased both frequencies and amplitudes of mEPSCs (Harrington et al., 2016; Rajkovich et al., 2017). These results further suggest that both cell-autonomous and cell-non-autonomous effects need to be further investigated to gain a more complete understanding of the Mef2c function.
A role of abDGCs in social behavior
Patients harboring mutations in MEF2C have been linked to ASD. In addition, multiple animal models in which Mef2c is deleted have commonly displayed ASD-related behaviors. However, Mef2c mutations in the entire body (i.e., heterozygote for Mef2c) or vast areas of the brain (i.e., Emx1-Cre driven Mef2c deletion) hinder the field from identifying the precise brain area(s) that transmit Mef2c mutations to ASD-related behaviors. In this study, we identified hippocampal neurogenesis as a critical neural substrate that is necessary for Mef2c-dependent function and behavior. Hippocampal neurogenesis refers to the process by which new neurons are generated and incorporated into the pre-existing neural circuits. The precise regulation of hippocampal neurogenesis ensures tight control over the number of new neurons, as well as the structural development and neural circuit formation of newborn neurons. This controlled hippocampal neurogenesis provides ample plasticity, allowing the brain to exert cognitive functions and maintain emotional stability. Our study demonstrated that hippocampal neurogenesis has additional functions in social behavior. The specific deletion of Mef2c in abDGCs impaired social interactions, as well as the ability to recognize social novelty. This deficit in social behavior is one of the hallmarks manifested in ASD. This is interesting because a recent study proposed that the DG may be responsible for contextual novelty but that social novelty is attributed to CA2 (Hitti and Siegelbaum, 2014; Donegan et al., 2020; Chen et al., 2020b). However, ablation of hippocampal neurogenesis revealed its critical role in the maintenance of social behavior (Opendak et al., 2016). Moreover, deletions of many other ASD genes specifically in abDGCs commonly produced ASD-like social behaviors (Amiri et al., 2012; Guo et al., 2012; Chen et al., 2017b; Bicker et al., 2021; Kim et al., 2022), supporting a direct contribution of abDGCs to social behavior. It is intriguing that the recent discovery of direct projections of abDGCs to CA2 pyramidal neurons may explain the potential cooperative function of the DG to CA2 networks in social behaviors (Llorens-Martin et al., 2015; Whitebirch et al., 2022). Moreover, our observation that the conditional deletion of Mef2c in abDGCs disrupted contextual fear memory is in line with the recent reports on the role of Mef2c in cognitive function and resilience (Cole et al., 2012; Barker et al., 2021). Thus, these results collectively suggest that hippocampal neurogenesis is a critical neural substrate necessary for Mef2c-dependent social and cognitive behaviors.
Footnotes
This work was supported by the National Institute of Alcohol Abuse and Alcoholism (R01AA022377 and R01AA01027766 to H.S.; K01AA027773 to S.K.). We thank Dr. Thomas Jaramillo, the Director of Rodent Behavior Core in Cleveland Clinic Lerner Research Institute, for his guidance in all of our behavioral work. We thank Dr. Christopher L. Nelson for editorial assistance and Gregori Enikolpov for Nestin-GFP mice. We would also like to thank undergraduate students Mr. Brian Seo and Ms. Soumyaa Das for their help and support.
The authors declare no competing financial interests.
- Correspondence should be addressed to Hoonkyo Suh at suhh2{at}ccf.org.