CHMP7/ESCRT-III Is Localized at the Nuclear Envelope of Cortical Neurons and Required for Expression of Activity-Regulated Genes

Simple Summary

Neurons in the brain constantly change their activity in response to experience, and this activity must be translated into changes in gene expression to support learning, memory, and normal brain function. How this communication between neuronal activity and the cell nucleus is organized is still not fully understood. In this study, we investigated how proteins at the boundary of the nucleus help regulate activity-dependent gene expression in cortical neurons. We focused on proteins that form a physical connection between DNA and the nuclear envelope, the membrane that surrounds the nucleus. We found that increased neuronal activity is associated with the accumulation of specific membrane-remodeling proteins at the nuclear envelope. Reducing the levels of one of these proteins led to widespread decreases in genes involved in neuronal activity and synaptic communication. Importantly, these gene expression changes closely matched those seen when other components of the same nuclear scaffold were disrupted. Our findings suggest that a dynamic protein assembly at the nuclear envelope helps coordinate neuronal activity with gene regulation. Understanding this mechanism provides new insight into how brain cells adapt to activity and may be relevant for neurological disorders in which gene regulation is impaired.

Abstract

The epigenome and nuclear architectural mechanisms that regulate neuronal activity-induced transcriptional responses in cortical neurons remain incompletely understood. Previously, we have shown that the chromatin organizer SATB2 and the inner nuclear membrane protein LEMD2 form a chromatin tether at the nuclear lamina, and that activity-induced transcription is impaired in both Satb2 and Lemd2 loss-of-function models. Interaction of SATB2 and LEMD2 with subunits of the ESCRT-III complex indicates that the ESCRT-III complex could serve as an activity-dependent, dynamic component of this tether. Here, we study the activity-dependent subcellular localization and function of the ESCRT-III components CHMP7 and CHMP4B in primary cortical neurons. We find that increased neuronal activity correlates with the accumulation of co-localized CHMP7 and CHMP4B foci at the nuclear envelope. shRNA-mediated Chmp7 knockdown causes a reduction in the expression of activity-regulated genes and genes with highly specialized functions in synaptic organization and trans-synaptic signaling. Furthermore, the observed similarity in the global transcriptome responses in Satb2, Lemd2, and Chmp7 loss-of-function models points toward a previously unrecognized role of the SATB2–LEMD2–CHMP7 tether in linking chromatin architecture and nuclear envelope plasticity to activity-dependent gene regulation.

1. Introduction

Neuronal activity induces widespread and tightly controlled changes in gene expression that are essential for synaptic plasticity, circuit maturation, and cognitive function [1]. The activity-driven transcriptional response involves rapid induction of immediate early genes (IEGs), followed by several waves of late-response gene induction, ultimately causing long-lasting alterations in synaptic structure and function [2]. While cell-signaling pathways and transcription factors leading from increased synaptic input to gene expression have been extensively elucidated [3], the epigenome and nuclear architectural mechanisms that facilitate, sustain, and coordinate activity-dependent gene expression in neurons remain incompletely understood.

We have previously demonstrated that cortical neurons lacking the chromatin organizer SATB2 exhibit impaired IEG induction in response to synaptic NMDA receptor stimulation caused by bicuculline treatment [4,5]. Proteomic analysis has revealed that, in cortical lysates, SATB2 interacts with the inner nuclear membrane protein LEMD2 and unexpectedly with subunits of the ESCRT-III complex [6].

The ESCRT-III pathway executes membrane remodeling reactions throughout the cell [7]. Within the nucleus, ESCRT-III performs critical surveillance and repair functions at the nuclear envelope, such as removal of defective nuclear pore complexes [8], sealing of nuclear membrane ruptures [9], and nuclear envelope reassembly after mitosis [10]. These activities depend on CHMP7, a hybrid ESCRT protein that interacts with LEMD2 to recruit ESCRT-III subunits, including CHMP4B, to the inner nuclear membrane [11]. Currently, it is unclear whether these CHMP7-dependent recruitment events are conserved across organisms and cell types, including different neuronal subtypes, and whether the composition and function of the recruited individual ESCRT-III proteins at the nuclear envelope are conserved or cell type-specific [12]. The interaction between pyramidal neuron-specific chromatin scaffolding protein SATB2 and the inner nuclear membrane protein LEMD2 [4,6] indicates the existence of an activity-dependent tethering complex in cortical neurons, including SATB2, LEMD2, and ESCRT-III, that regulates gene expression by mediating interactions between chromatin and the inner nuclear membrane.

In this study, by using primary cortical cultures as an experimental model, we examined (i) whether increased neuronal activity correlates with CHMP7 accumulation at the nuclear envelope; (ii) whether CHMP7 localization at the nuclear periphery is accompanied by recruitment of the ESCRT-III subunit CHMP4B; and (iii) whether loss of CHMP7 impairs activity-regulated transcription. Our findings reveal a novel role of CHMP7/ESCRT-III in activity-dependent gene regulation.

2. Materials and Methods

2.1. Primary Cortical Cultures

Primary cortical cultures were established from P0 mice, as previously described [5]. Briefly, cortices were dissected in HBSS buffer (1X HBSS, 2.5 mM HEPES pH 7.4, 30 mM D-Glucose, 1 mM CaCl₂, 1 mM MgSO₄, 4 mM NaHCO₃), the tissue was washed twice in HBSS buffer without salts (1X HBSS, 0.05 mM D-AP5, 30 mM D-Glucose, 1 mM Sodium Pyruvate, 10 mM HEPES pH 7.4) and incubated at 37 °C for 10 min in previously activated papain solution (Papain Worthington Bio., Lakewood, NJ, USA, Cat. No. LK003178), containing DNase I (Sigma-Aldrich, St. Louis, MO, USA, Cat. No. D5025). Subsequently, cortices were washed in HBSS buffer without salts, and attachment medium (MEM, Gibco, Grand Island, NY, USA, Cat. No. 11095080) was added, supplemented with 1% horse serum, 0.1% penicillin/streptomycin, and 1 mM sodium pyruvate. Tissue was triturated by using a siliconized glass pipette, and neurons were seeded on tissue culture plates, previously coated with 0.5 mg/mL of poly-L-ornithine hydrobromide (PORN) solution (Sigma-Aldrich, Cat. No. P4538, overnight at 4 °C) and 0.005 mg/mL laminin (Sigma-Aldrich, Cat. No. L2020, 2 h at 37 °C) as follows: 0.5 × 10⁶, 1 × 10⁶, and 0.124 × 10⁶ cells were plated on 12-well plates, 6-well plates, and 8-well IBIDI μ-slides with glass bottoms (Ibidi^®, Gräfelfing, Germany, Cat. No. 80827), respectively. After pre-culturing for 2 h, medium was replaced with feeding medium (NBM-A, Burlington, MA, USA, Gibco, Grand Island, NY, USA, Cat. No. 10888-022), supplemented with 0.5 mM GlutaMAX, 0.1% Penicillin/Streptomycin, 1XB27 (Thermo Fisher Scientific, Waltham, MA, USA). Neurons were cultured for 14 days at 5% CO₂ and 37 °C. Glial cell proliferation was inhibited by adding cytosine arabinoside (5 μM, Sigma-Aldrich, Cat. No. C6645) to the culture medium at DIV3. On DIV4, half of the medium was replaced with fresh medium. On DIV 14, cortical neurons were first pre-treated with NBQX (20 μM, Bio-Techne, Minneapolis, MN, USA, Cat. No. 1044) for 16 h, and then either used like that or treated with bicuculline (50 μM, Bio-Techne, Cat. No. 0131) for an additional 1 h.

2.2. Cloning of AAV Constructs

All plasmids used in this study were generated by using In-Fusion assembly (Takara-Bio USA Inc., San Jose, CA, USA, Cat. No. 102518) according to the manufacturer’s protocol. The sequences of the primers used for cloning and the sequence of Chmp7 shRNA are listed in Table S1.

To create a V5-tagged version of human CHMP7, the coding sequence (CDS) was amplified from pMGF182 (Addgene ID: #97006) using primers C7-V5_F and C7-V5_R and cloned into linearized pAAV-hSyn-LEMD2-V5, removing the LEMD2 CDS and, thus, generating pAAV-hSyn-CHMP7-V5. To produce an expression plasmid carrying V5-tagged CHMP7 under a doxycycline-inducible promoter, the CHMP7 CDS in-frame with V5 was amplified using primers TetOn_C7-V5_F and TetOn_C7-V5_R from the previously generated plasmid pAAV-hSyn-CHMP7-V5, and then cloned into pAAV-Ptet-RFP-shR-UbC-rtTA (Addgene-ID: #35625), replacing RFP-shR with CHMP7-V5 to generate pAAV-Ptet-CHMP7-V5-UbC-rtTA.

A cassette carrying an mCherry CDS in frame with human CHMP4B CDS was amplified with primers mCh-C4B_F and mCh-C4B_R from pLNCX-mCherry-CHMP4B (Addgene-ID: #116923) and assembled with a pAAV-hSyn backbone acquired by linearizing pAAV-hSyn-V5-SATB2. The resulting construct pAAV-hSyn-mCherry-CHMP4B served as a template to obtain the CHMP4B CDS using primers C4B-3xf_F and C4B-3xf_R, and the amplicon was subsequently cloned into a pAAV-hSyn-3xFLAG backbone (Addgene-ID: #127862) to generate pAAV-hSyn-CHMP4B-3xFLAG. For generation of an in-frame 3xFLAG-tagged version of CHMP4B under control of an rAAV-compatible inducible promoter system, as described above, pAAV-hSyn-CHMP4B-3xFLAG served as a PCR template to clone CHMP4B-3xFLAG into pAAV-Ptet-RFP-shR-UbC-rtTA using the primers TetOn_C4B-3xf_F and TetOn_C4B-3xf_R, removing RFP-shR and generating pAAV-Ptet-CHMP4B-3xFLAG-UbC-rtTA.

To produce an mCherry reporter plasmid serving as the negative control, mCherry CDS was obtained from the previously generated pAAV-hSyn-mCherry-CHMP4B by PCR amplification using TetOn_mCh_F and TetOn_mCh_R, and once again assembled with linearized pAAV-Ptet-RFP-shR-UbC-rtTA, replacing RFP-shR with mCherry, thus creating pAAV-Ptet-mCherry-UbC-rtTA.

The design of the Chmp7 shRNA sequence and the cloning into the empty pAAV-U6-hSyn::mCherry.3xFLAG-WPRE backbone was performed as previously described [13]. The final rAAV vector was labelled rAAV-U6-shChmp7-hSyn-mCherry. The plasmid encoding scrambled siRNA, pAAV-U6-scrambled-hSyn::mCherry.3xFLAG-WPRE, was purchased from Addgene (#120395).

2.3. AAV Production and Transduction

In-house production of recombinant AAV (rAAV) was performed following previously published protocol [14]. Titration of produced rAAV was performed by quantitative PCR (qPCR) using SYBR Green technology.

For overexpression studies, primary cortical cultures were transduced on DIV 4 with rAAV-Ptet-CHMP4B-3xFLAG-UbC-rtTA or rAAV-Ptet-CHMP7-V5-UbC-rtTA at multiplicity of infection (MOI) of 125,000 VG/cell. In Chmp7 knockdown experiments, primary cortical neurons were transduced on DIV6 with rAAV-U6-shChmp7-hSyn-mCherry or rAAV-U6-shScr-hSyn-mCherry at MOI of 100,000 VG/cell.

2.4. Antibodies

Primary and secondary antibodies used in this study are listed in Table S2. All anti-V5 tag and anti-FLAG tag primary antibodies were validated for their specificity by immunostaining of non-transduced cultures or cultures transduced with a different target protein.

2.5. Western Blotting

Cultured cortical neurons were collected on DIV 14. Cells were washed twice with 1 mL of PBS and collected in 2X Roti-Load buffer (Roth, Dautphetal, Germany, Cat. No. K929.1) using a cell scraper. Collected cells were then sonicated using a sonifier (Branson Ultrasonics™ sonifier 250 CE, Brookfield, CT, USA, output control 6, duty cycle: 30%, 2 pulses for 2 s) and centrifuged for 5 min at 13,000 rcf at 4 °C. Then, samples were boiled for 5 min at 95 °C and centrifuged briefly. The prepared samples and the protein marker (Precision Plus Protein Dual Color; Bio-Rad, Hercules, CA, USA, Cat. No. 1610374) were loaded onto a 10% SDS polyacrylamide gel, which was run initially at 80 V for approx. 3 h. Protein was transferred onto a PVDF membrane (Millipore, Burlington, MA, USA, Cat. No. IPVH00010) via semi-dry electroblotting for 30 min at 25 V. To validate successful transfer, membranes were stained with Ponceau S solution (Sigma-Aldrich, Cat. No. P7170) for 5 min at room temperature. The membrane was then rinsed with TBS-T (1 mL Tween 20, Roth, Cat. No. 9127, in 1 L TBS) and cut according to the molecular weight of the target proteins. Membrane sections were then incubated in blocking buffer (5% milk powder, Roth, Cat. No. T145.4, in TBS-T) for 1 h and subsequently incubated with primary antibody solution in blocking buffer overnight at 4 °C. The next day, membranes were washed three times with washing buffer (20 mL Blocking buffer in 80 mL TBS-T, 1:5) and then incubated with HRP-conjugated secondary antibody solution in washing buffer for 1 h at room temperature. After incubation, membranes were again washed three times in washing buffer. Membranes were briefly washed in TBS-T and developed using ECL reagent (GE Healthcare, Chicago, IL, USA, Cat. No. 28980926) and analyzed using the ChemiDoc chemiluminescence detection system (Bio-Rad). Protein expression was quantified using Image Lab analysis software version 6.1.0 build 7 (Bio-Rad, Hercules, CA, USA).

2.6. Immunocytochemistry

Primary cortical neurons plated on glass-bottom IBIDI μ-slides were fixed on DIV 14. Briefly, wells were gently washed twice with 300 μL of PBS (Gibco, Cat. No. 10010023) warmed to 37 °C and then incubated with 300 μL of 4% (w/v) paraformaldehyde (PFA, Roth, Cat. No. 0335.1) for 10 min at room temperature (RT). Cells were washed three times with PBS, and plates were stored at 4 °C wrapped in parafilm until further use. Prior to immunostaining, cells were permeabilized with 0.5% Triton-X (Roth, Cat. No. 3051.2) in PBS for 10 min at 5× rpm agitation at RT. Wells were washed three times with 300 μL of PBS for 5 min, and then incubated with 300 μL of PBS+ blocking solution for 1 h at room temperature (1X Casein, ThermoFisher #37528, 1% BSA, Roth #8076.5, 0.2% Fish Skin Gelatine, Sigma-Aldrich #G7041, 1 X PBS, Gibco #700011036) diluted in ddH₂O. Primary antibodies, diluted in PBS+, were added at the indicated dilutions (Table S2) and incubated for 1–2 h at RT. After incubation, cells were subjected to a series of washing steps, comprising three washes with 300 μL of PBS+ followed by three washes with 300 μL of PBS, each for 5 min, at 5× rpm agitation, at room temperature. Secondary antibodies, diluted in PBS+, were added at the indicated dilutions (Table S2) and incubated for 1 h at RT in the dark. Wells were then washed three times with 300 μL of PBS for 5 min at 5× rpm agitation at RT and stored in 7 drops per well of DAPI-containing mounting medium (Ibidi, Cat. No. 50011).

2.7. Image Acquisition and Analysis

Confocal images were obtained using a Zeiss LSM700 Axio Observer microscope (Oberkochen, Germany) with a Zeiss Plan-Apochromat 63x/1.4 NA Oil Ph3 M27 objective and a Zeiss Axiocam 503 Mono camera. For confocal images, the following lasers were used: 405 nm (DAPI), 488 nm (Alexa Fluor 488, Waltham, MA, USA), 555 nm (Alexa Fluor 555), and 639 nm (Alexa Fluor 633). ZEN 2.3 SP1 FP3 (black) software version 14.0.24.201 (Carl Zeiss, Oberkochen, Germany) was used for image acquisition. Optical sections were acquired at a resolution of 0.11 μm (x) × 0.11 μm (y) × 0.32 μm (z). All acquisition parameters were kept constant across all images within an experiment.

Confocal images were analyzed using FIJI (ImageJ) v2.9.0/1.53t. Custom ImageJ macro scripts were developed to automate processing steps and are available upon request. Raw z-stacks were visually inspected to confirm protein expression and cropped into single-cell images. Background intensity was estimated from randomly placed ROIs across focal planes, and the highest mean value was used for subsequent background subtraction (CHMP4B-3xFLAG, cFOS) or as an intensity cut-off for classifying positive nuclei (mCherry, CHMP7-V5).

Segmentation of the nuclear region was based on Lamin B1 staining. Due to the resolution limits of confocal microscopy, inner versus outer nuclear membrane localization cannot be distinguished. For quantification of CHMP7–CHMP4B co-localization, nuclear CHMP7-V5 foci were identified in 3D using DAPI-based segmentation and quantified for the CHMP4B-3xFLAG signal. For each focus, number, volume, and mean fluorescence intensity (MFI) in the CHMP4B channel were recorded. Foci with CHMP4B MFI exceeding three times the background levels were classified as CHMP4B-positive.

For classification of CHMP4B nuclear localization, the ratio of CHMP4B-3xFLAG total fluorescence in the nuclear lamina versus intranuclear compartment was calculated. Integrated Density was measured across the stack and averaged. Based on the lamina/intranuclear ratio, cells were classified as predominantly laminar (≥60%), mixed lamina–intranuclear (27–59%), or predominantly intranuclear (≤26%). Cytoplasmic localization was assessed visually due to absence of a cytoplasmic marker.

For whole-nucleus CHMP4B and CHMP7 quantification, Lamin B1-based segmentation was used to define a whole-nucleus ROI by synthetically filling intranuclear space. MFI values for CHMP4B were averaged across z-sections. cFOS levels in CHMP4B-3xFLAG-expressing cells were quantified using the same nuclear segmentation. To avoid bleed-through from the CHMP4B (Alexa Fluor 488) channel into the cFOS (Alexa Fluor 555) channel, regions with high CHMP4B signal were excluded. Z-slices outside the Lamin B1-positive region were removed manually, and whole-nucleus cFOS MFI was averaged across the stack.

To quantify laminar CHMP7 signal and intranuclear cFOS immunoreactivity, masks for nuclear lamina were generated from LaminB1 signal for each optical section. This produced a ring-like structure comprising an intralaminar and extralaminar margin enclosing the Lamin B1-positive region. The script measured CHMP7-V5 signal intensity within NL ROIs across all z-planes, outputting Raw Integrated Density (RawIntDen) and ROI area for each slice. For intranuclear cFOS quantification, the ROI corresponded to the region enclosed by the outer (extralaminar) boundary of the Lamin mask. After excluding the extralaminar surrounding region, RawIntDen and ROI area for cFOS signal were recorded for each slice. For both analyses, RawIntDen and area values from all slices of a nucleus were summed, and MFI was calculated by normalizing the total RawIntDen to the total area. Unless stated otherwise, figures show the mean nuclear MFI per replicate.

2.8. RNA Isolation

Cultured mouse primary cortical neurons were washed twice with PBS, lysed in TRIzol (Invitrogen, Carlsbad, CA, USA, Cat. No. 12044977), scraped, and collected in RNase-free Microfuge Tubes (Invitrogen, Cat. No. AM12400). Samples were incubated for 5 min at RT, chloroform (200 μL per 1 mL TRIzol, Sigma-Aldrich, Cat. No. C2432) was added, and the tubes were mixed by shaking. After 3 min of incubation at RT, samples were centrifuged for 15 min at 12,000× g at 4 °C. The upper aqueous phase was transferred into a new RNase-free tube, an equal volume of 70% ethanol (Sigma-Aldrich, Cat. No. 51976) was added, and the samples were loaded on Invitrogen PureLink RNA Micro Scale Kit (Cat. No. 12183016) columns. Samples were processed following the manufacturer’s instructions. Finally, RNA was eluted twice with 11 μL of nuclease-free water, resulting in a final elution volume of 20 μL per condition.

2.9. qPCR

cDNA was synthesized from extracted RNA with the iScript cDNA Synthesis Kit (Bio-Rad, Cat. No. 1708890), according to the manufacturer’s recommendations. For RT-qPCR using the SYBR Green technology, a 96-well plate was set up, and 5 μL of each cDNA sample was loaded in duplicates, as well as non-template controls (water). In total, 15 μL of the RT-qPCR master mix containing a respective primer pair was added, and the plate was sealed with a transparent film. The plate was briefly centrifuged, and the thermal cycling program was run. Data analysis was performed using the CFX Maestro 1.0 software (Bio-Rad, Version 4) and the 2-ddCt method.

2.10. RNA-Seq

RNA integrity was determined with the automated electrophoresis platform 4150 TapeStation (Agilent, Santa Clara, CA, USA, Cat. No. G2992AA). Samples with a RIN score >8.4 were used for library preparation and sequencing. PolyA-enriched libraries were generated by Novogene UK and sequenced on the Illumina Novaseq platform at 150 bp PE, with a sequencing depth of 30 million reads per replicate.

Raw reads were trimmed using trimmomatic (v0.39) [15] with the following command: java -jar trimmomatic-0.39.jar PE $reads1 $reads2 $out3.trimmed.gz $out4.gz $out5.trimmed.gz $out6.gz LEADING:3 TRAILING:3 MINLEN:36 SLIDINGWINDOW:4:15 -threads 2 ILLUMINACLIP:Trimmomatic-0.39/adapters/TruSeq3-PE.fa:2:30:10:2:True -summary $out3.stat.summary. Trimmed reads were mapped to mm10 using the STAR aligner (2.7.4a) [16]. The output bam files were sorted, indexed and assessed using samtools flagstat. The featureCounts [17] package of subreads (v2.0.6) (subread/2.0.6/bin/featureCounts) was used to count reads to obtain read counts along genes for each sample. The count matrices were generated and corrected for batch effect using the RUV (k = 3) function of the R package RUVseq (v1.44) [18]. Differential gene expression analysis was performed in DESeq2 [19]. For differential gene expression analysis, a threshold cutoff of adjusted (Benjamini–Hochberg) p-value < 0.05 and |Log2FoldChange| > 1 was applied. Vst-normalized expression values (VST, DESeq2) were used for clustering and heatmap analysis. Volcano plots were generated using the R package EnhancedVolcano (v1.28.2). RRHO analysis was performed as previously described [20,21].

Metascape [22] was used to identify overrepresented GO terms. As ontology sources, GO Molecular Functions, GO Biological Processes, and GO Cellular Components were used. Terms with a p value < 0.01, a minimum count of 3, and an enrichment factor >1.5 (the ratio between the observed counts and the counts expected by chance) were collected and grouped into clusters based on their membership similarities. p values were calculated based on a cumulative hypergeometric distribution, and q-values were calculated using the Benjamini–Hochberg procedure for multiple testing. A custom background was used, consisting of mouse brain-expressed genes. Fisher’s exact test was employed to test for significant overlaps between gene sets.

2.11. Statistical Analysis

All datasets were tested for Gaussian distribution by Kolmogorov–Smirnov test and for equal variance by F-test. Descriptions of data collection, data quantification, and statistical methods used to analyze each experiment are provided in the respective sections in the method details or in figure legends. Exact p values and number of replicates (n) can be found in the main text and/or figure legends.

3. Results

3.1. ESCRT-III Complex Subunits Co-Localize at the Nuclear Periphery in Cortical Neurons

The lack of validated, immunocytochemistry-compatible antibodies against ESCRT-III subunits generally precludes a reliable analysis of their subcellular localization. To circumvent this limitation and explore the subcellular co-localization of CHMP7 and CHMP4B—which indicates ESCRT-III assembly—in murine primary cortical neurons, we employed AAV-mediated inducible expression of tagged versions of both proteins (Figure 1A). The expression of V5-tagged CHMP7 and FLAG-tagged CHMP4B was induced by doxycycline treatment for 26 h only, to avoid excessive accumulation of recombinant proteins (Figure S1). Western blot analysis of CHMP7-V5-transduced versus non-transduced cultures confirmed that CHMP7-V5 overexpression under these conditions did not strongly increase overall CHMP7 levels (Figure 1B, Figures S7 and S8). Cortical cultures expressing tagged versions of CHMP7 and CHMP4B displayed normal neuronal morphology without evidence of overt cellular stress or degeneration (Figure S2).

Cortical neurons were then immunostained with antibodies against V5 to reveal subcellular localization of CHMP7. Lamin B1 staining was used to label the nuclear envelope. As expected, CHMP7-V5 immunoreactivity (IR) was detected in the cytoplasm and/or endoplasmic reticulum (ER), consistent with previously reported localization of CHMP7 in other cell types [23]. However, in primary cortical neurons, in addition to cytoplasmatic/ER localization, an accumulation of CHMP7 foci at the nuclear envelope was also observed (Figure 1C). Notably, nuclei exhibiting CHMP7 accumulation at the nuclear envelope displayed a continuous Lamin B1 rim (Figure S3), arguing against overt nuclear envelope rupture under the conditions examined. However, we cannot exclude more subtle alterations in nuclear envelope function or nuclear pore complex organization, which will require dedicated analyses in future studies.

To determine whether CHMP7 localization at the nuclear envelope is associated with ESCRT-III assembly, we co-transduced cortical neurons with AAVs encoding V5-tagged CHMP7 and FLAG-tagged CHMP4B. In parallel, with doxycycline application, AAV-transduced cultures were treated for 16 h with NBQX—an AMPA receptor antagonist—to moderately silence neuronal activity, or additionally for 1 h with bicuculline (BIC)—a GABA-A receptor antagonist—to enhance synaptic activity [24]. CHMP4B-3xFLAG staining revealed an accumulation of CHMP4B foci in the nucleus of transduced cells under both inactive and active conditions. CHMP4B foci were detected in the nucleoplasm, perinuclearly—within or immediately beneath the nuclear envelope—or at both locations (Figure 1D). Importantly, in cells double-positive for CHMP7-V5 and CHMP4B-3xFLAG, perinuclear CHMP4B foci exhibited a very high degree of co-localization with CHMP7 foci (99.09% in BIC-treated cultures, n = 331 foci; 99.75% in NBQX-treated cultures, n = 397 foci; n = 4 biological replicates for both treatment groups) (Figure 1D and Figure S4A), indicating CHMP7-directed ESCRT-III polymerization at the nuclear envelope of cortical neurons.

3.2. CHMP4B and CHMP7 Localization at the Nuclear Envelope Correlate with Increased cFOS Expression

To determine whether CHMP7 localization at the nuclear envelope is linked to neuronal activity, we quantified two parameters in CHMP7-V5-transduced neurons: the mean fluorescence intensity of CHMP7-V5 signal at the nuclear lamina (identified by Lamin B1 staining), and the mean nuclear intensity of the activity marker protein cFOS. Under both active and inactive conditions, we found a weak-to-moderate but highly significant positive correlation between the two staining intensities (Figure 2A). Stratification of neurons based on high versus low CHMP7-V5 signal at the nuclear lamina revealed significantly higher intranuclear cFOS immunoreactivity in neurons with high CHMP7-V5 levels compared to those with low levels (Figure 2B). Therefore, CHMP7 accumulation at the nuclear envelope is correlated with the neuronal activity state of individual cortical neurons.

We next asked whether CHMP4B-3XFLAG subcellular localization patterns correlate with the activity state of individual neurons. We again used cFOS IR as an indicator of neuronal activation and employed a semi-automated image analysis pipeline to quantify the distribution of CHMP4B foci relative to nuclear lamina (Figure S4B). Based on the ratio of the CHMP4B signal at the nuclear lamina and in the nucleoplasm, we categorized neurons into three groups: (1) neurons showing CHMP4B predominantly at the nuclear lamina (NL; ≥60% of the CHMP4B signal at the nuclear lamina); (2) neurons exhibiting CHMP4B foci both at the nuclear lamina and inside the nucleoplasm (NL + IN; 27–59% of the signal at the nuclear lamina), and (3) neurons showing CHMP4B foci mainly in the nucleoplasm (IN; ≤26% of the signal at the nuclear lamina) (Figure 2C). Neurons showing CHMP4B exclusively outside the nucleus represented the fourth category (CP), defined by a lack of overlap of the CHMP4B signal with both DAPI and Lamin B1 signals.

Analysis of cFOS IR in these four categories of transduced neurons (Figure S4C) under both active and inactive conditions revealed significantly higher cFOS protein levels in NL + IN neurons compared to CP and IN neurons (Figure 2D, right). In NBQX-treated cultures, cFOS levels were also significantly elevated in NL neurons compared to CP and IN neurons (Figure 2D, left). As controls, cFOS levels were quantified in mCherry-transduced versus non-transduced neurons to account for potential effects of viral transduction alone. There was no statistically significant difference in cFOS staining between mCherry-transduced and non-transduced neurons under both silent and active conditions, excluding non-specific transduction effects on cFOS expression (Figure S5).

We next tested whether cFOS levels correlate with the total nuclear CHMP4B signal, regardless of the subnuclear localization. Using a computational pipeline to quantify whole-nucleus CHMP4B IR (Figure S4D), we observed a weak-to-moderate significant correlation between CHMP4B and cFOS nuclear signals under both active and silent conditions (Figure 2E).

Together, these findings indicate that accumulation of CHMP7 and CHMP4B foci at or near the nuclear envelope correlates with the activation state of cortical neurons.

3.3. CHMP7 Is Required for Maintenance of Synaptic and Activity-Regulated Gene Expression

Having established a link between CHMP7/ESCRT-III localization at the nuclear envelope and the activity status of individual neurons, we next ask if CHMP7 plays a role in activity-regulated transcription. To address this question, we established a shRNA-mediated knockdown of Chmp7 in primary cortical cultures (Figure 3A). Neurons were transduced on DIV6 with AAVs encoding either Chmp7-targeting shRNA or a scrambled shRNA control, and CHMP7 protein and mRNA levels were quantified by Western blotting and qPCR at DIV14. This analysis confirmed a reduction of at least 55% in protein levels and 92% in mRNA levels following AAV–Chmp7-shRNA transduction compared to the scrambled control (Figure 3B and Figure S9).

We next performed transcriptome sequencing to identify differentially expressed genes (DEGs) between Chmp7 knockdown and control cultures under both inactive and active conditions. This analysis revealed a strong effect of Chmp7 knockdown on gene expression (Figure 3C). In total, 3222 genes were defined as differentially expressed (|log2 fold change| > 1, adj. p-value < 0.05, baseMean > 30) after NBQX treatment, and 3206 genes after BIC treatment between Chmp7 knockdown and control cultures. Because poly(A)-enriched RNA-seq reports steady-state mRNA abundance, these data do not distinguish between changes in transcription, nuclear export, or mRNA stability. A set of 2688 DEGs overlapped between the active and inactive conditions. Functional gene ontology (GO) enrichment analysis of the DEGs common to both conditions revealed that upregulated genes following Chmp7 knockdown were significantly enriched for only a small number of GO terms related to the extracellular matrix (external encapsulating structure, Log(p-value) = −6.05, extracellular matrix, Log(p-value) = −5.71, complex of collagen trimers, Log(q-value) = −5.28). By contrast, downregulated genes were strongly and significantly enriched for GO terms associated with synaptic transmission, synapse organization, and cholesterol biosynthetic process (Figure 3D).

Focusing specifically on synaptic gene ontologies by using the SynGO knowledgebase [25], we found a highly significant overrepresentation of seven “Cellular Component” and eleven “Biological Process” terms at 1% FDR among the downregulated genes (Figure 3E). This indicates that in cortical neurons, CHMP7 is essential for the expression of genes with highly specialized synaptic functions, such as those involved in trans-synaptic signaling (Nrnxn1, Nrnxn3, Plcb1, Alk, Gucy1a, Ilrap, and Bdnf), regulation of synaptic vesicle exocytosis (Unc13c, Unc13b, Stx1a, Mctp2, Rims1, Erc2, Prkcg, and Sv2c), and postsynaptic specialization (Gphn, Nptx2, Homer1, Homer2, and Dlgap1), among others. Moreover, downregulated genes upon Chmp7 knockdown were also highly significantly enriched for all three major classes of activity-regulated genes, as defined by Tyssowski et al., 2018 [26]—rapid primary response genes (PRGs), delayed PRGs, and secondary response genes (odds ratio = 10.04, p < 1 × 10⁻¹⁵, two-sided Fisher’s exact test).

In control scrambled-shRNA-transduced cultures, BIC treatment robustly induced almost all rapid PRGs (16 out of 19), as well as a subset of delayed PRGs (15 out of 112) (Figure 3F). In Chmp7-shRNA-transduced cultures, the induction of these genes was preserved. However, the expression levels of all rapid PRGs, delayed PRGs, and SRGs were significantly reduced under both silent and stimulated conditions (Figure 3G and Figure S6), indicating a requirement for CHMP7 in maintaining basal levels of activity-regulated gene transcription.

To test whether gene expression changes caused by Chmp7 knockdown share similarities with Lemd2 [4] and Satb2 [5] loss-of-function models, we next applied Rank–Rank Hypergeometric Overlap (RRHO) analysis [20,21]. This threshold-free algorithm tests for statistically significant overlap between two gene lists ranked by the degree of differential expression between two contrasting conditions. RRHO analysis revealed a highly significant overlap between the downregulated genes upon Chmp7 and Lemd2 knockdown, as well as upon Chmp7 knockdown and Satb2 deletion (Figure 3H). Genes downregulated upon Satb2 deletion were also mostly downregulated upon Chmp7 knockdown (Figure 3I). Likewise, genes downregulated upon Lemd2 knockdown were predominantly downregulated following Chmp7 knockdown (Figure 3I). Together, these findings indicate that CHMP7, LEMD2, and SATB2 exert convergent effects on gene expression programs in cortical neurons.

4. Discussion

We analyzed the subcellular localization and function of the ESCRT-III complex components CHMP7 and CHMP4B in murine primary cortical neurons under both silenced and moderately active conditions. Our data revealed an accumulation of CHMP7/CHMP4B foci at the nuclear envelope that correlates with the neuronal activity state of individual neurons. We further demonstrate that CHMP7 is required to sustain the expression of genes regulated by increased synaptic activity and genes encoding highly specialized synaptic proteins. These observations expand the ESCRT-III functional repertoire at the nuclear envelope beyond nuclear membrane sealing/repair [27] and nuclear pore complex quality control [12] to a previously unrecognized role in neuronal transcription.

Accumulation of CHMP4B in the interphase nucleus, often in the form of protein foci, has primarily been observed under pathological conditions such as micronucleus formation, nuclear membrane deformation, or during herpesvirus infection, where ESCRT-III is recruited to the nuclear envelope to facilitate viral egress [23,28,29,30,31]. Our findings demonstrate that in cortical neurons, nuclear or perinuclear CHMP4B localization is compatible with normal physiological function, since it was detected in healthy neurons under both low- and high-synaptic-activity conditions. Similarly, we find that in primary cortical neurons, CHMP7 localization at the nuclear envelope is a physiologically relevant and graded response to neuronal activity, in contrast to C9orf72- and sporadic ALS-induced pluripotent stem cell-derived spinal neurons, where increased nuclear accumulation of CHMP7 has been linked to pathological nuclear pore complex damage [32,33].

Our findings suggest that dynamic regulation of ESCRT-III recruitment to the nuclear envelope, as part of an LEMD2-SATB2-centered chromatin tether [4], serves as a molecular interface between changes in neuronal firing and activity-driven gene expression. The mechanisms that couple synaptic activity to CHMP7 nuclear lamina localization remain to be determined. One potential scenario includes post-translational modification(s) by specific kinases and/or phosphatases activated in response to neuronal activity-triggered Ca²⁺ influx, resulting in CHMP7 nuclear export signal masking [10,23] and consequently increased interaction with LEMD2.

The functional consequences of ESCRT-III recruitment to the nuclear envelope in neurons remain incompletely understood. Our data demonstrate that CHMP7 and the downstream ESCRT-III subunit CHMP4B accumulate at the nuclear periphery in a manner that correlates with cFOS expression, suggesting that nuclear ESCRT-III assembly influences activity-regulated gene expression. Several non-mutually exclusive working models can be envisioned. First, ESCRT-III recruitment may influence nuclear envelope architecture, including membrane curvature, infoldings, or local remodeling of the inner nuclear membrane. Such changes could alter chromatin–lamina interactions [34] or nuclear mechanics in ways that promote rapid transcriptional responses. Second, ESCRT-III may regulate the composition or turnover of nuclear pore complexes, as suggested by prior work, implicating ESCRT components in NPC surveillance and removal [8,35]. Altered NPC dynamics could influence nucleocytoplasmic trafficking of transcription factors or RNA-binding proteins critical for immediate–early gene induction. Future studies will be required to discriminate among these possibilities and determine whether nuclear ESCRT-III acts primarily through effects on chromatin organization, nuclear envelope structure, or nucleocytoplasmic transport pathways in neurons.

ESCRT-III components have previously been implicated in neuronal development and synaptic remodeling, including neuronal pruning, through their roles in endosomal trafficking and membrane remodeling in the cytoplasm [36]. In contrast, our study identifies a nuclear envelope-associated role for CHMP7 and ESCRT-III that impacts activity-dependent gene expression. These findings suggest that ESCRT-III may be deployed in spatially and functionally distinct contexts within neurons. While we do not directly link nuclear ESCRT function to synaptic pruning, it is conceivable that altered gene expression programs could indirectly influence neuronal remodeling processes. Future work will be required to determine whether and how these nuclear and cytoplasmic functions of ESCRT-III are coordinated.

In addition to the link between CHMP7/ESCRT-III localization and cFOS levels, our loss-of-function experiments revealed an unexpected and profound effect of Chmp7 knockdown on neuronal gene expression, even under active and inactive conditions. Steady-state levels of a large group of synaptic genes and multiple categories of activity-regulated genes were markedly reduced, suggesting that CHMP7 contributes to the stable transcriptional homeostasis of genes related to synaptic function in cortical neurons. The selective downregulation of these specific sets of genes in CHMP7-deficient neurons raises the question of what confers gene specificity. The strong similarity between the transcriptomic changes induced by Satb2 conditional deletion, Lemd2 knockdown, and Chmp7 knockdown points to a shared regulatory framework involving the LEMD2-SATB2 chromatin tether. SATB2 is known to bind selectively to promoters of synaptic genes and organize higher-order chromatin structure [5]. While we do not directly assess SATB2 chromatin occupancy following Chmp7 knockdown, we do not necessarily expect ESCRT-III loss to disrupt SATB2 DNA binding per se. Instead, we propose that CHMP7-LEMD2-dependent ESCRT-III recruitment at the nuclear envelope modulates the nuclear context of SATB2-bound loci—such as their spatial positioning or chromatin topology—thereby influencing transcriptional output without directly altering promoter occupancy. Considering that SATB2 directly interacts with LEMD2 [4,6], and that CHMP7 binds LEMD2 as well [11], the potential trimeric complex—CHMP7–LEMD2–SATB2—may couple spatial genome architecture to transcriptional control, thereby selectively affecting synaptic gene programs.

In support of these findings, a recent study has identified CHMP7 as a gene conferring risk to ADHD [37]. Individuals, homozygous for the risk SNP mapped to the 5^|UTR of CHMP7, showed lower neurocognitive function, higher ADHD symptom traits, and reduced CHMP7 transcript levels. Furthermore, functional validation of CHMP7 as an ADHD risk gene using a zebrafish model revealed smaller brain volumes and hyperactivity of chmp7 heterozygotes, pointing toward neurodevelopmental and synaptic signaling deficits [38]. Our data revealed that the profound impairment in synaptic gene transcription in Chmp7-shRNA-transduced cortical neurons is consistent with these phenotypes, providing a potential underlying molecular mechanism.

5. Conclusions

Here, we show that in primary cortical neurons—CMPP7, a hybrid protein that interacts with LEMD2 to recruit ESCRT-III subunits at the inner nuclear membrane, and CHMP4B—co-localize at the nuclear envelope under normal physiological conditions. Furthermore, the accumulation of CHMP7 and CHMP4B foci at the nuclear lamina correlates with the activity state of cortical neurons, as represented by cFOS immunoreactivity. In addition, our findings reveal a novel role for CHMP7/ESCRT-III in maintaining the expression of activity-regulated genes and genes with highly specialized functions at the synapse. The observed strong similarity between the global transcriptome responses upon Satb2, Lemd2, and Chmp7 loss-of-function points toward a previously unrecognized role of the SATB2-LEMD2-CHMP7 tether in linking chromatin architecture and nuclear envelope plasticity to activity-dependent gene regulation.