Regulation of Axon Initial Segment Diameter by COUP-TFI Fine-tunes Action Potential Generation

Abstract

The axon initial segment (AIS) is a specialized structure that controls neuronal excitability via action potential (AP) generation. Currently, AIS plasticity with regard to changes in length and location in response to neural activity has been extensively investigated, but how AIS diameter is regulated remains elusive. Here we report that COUP-TFI (chicken ovalbumin upstream promotor-transcription factor 1) is an essential regulator of AIS diameter in both developing and adult mouse neocortex. Either embryonic or adult ablation of COUP-TFI results in reduced AIS diameter and impaired AP generation. Although COUP-TFI ablations in sparse single neurons and in populations of neurons have similar impacts on AIS diameter and AP generation, they strengthen and weaken, respectively, the receiving spontaneous network in mutant neurons. In contrast, overexpression of COUP-TFI in sparse single neurons increases the AIS diameter and facilitates AP generation, but decreases the receiving spontaneous network. Our findings demonstrate that COUP-TFI is indispensable for both the expansion and maintenance of AIS diameter and that AIS diameter fine-tunes action potential generation and synaptic inputs in mammalian cortical neurons.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12264-021-00792-8.

Introduction

The axon initial segment (AIS) is a small specialized region, which contains a high density of voltage-gated sodium (Na_V) and potassium channels (K_V) to promote action potential (AP) generation in neurons [1]. These ion channels are clustered by the master organizing protein ankyrin G (AnkG), which is anchored to the actin cytoskeleton by βIV-spectrin [2]. Superresolution imaging has revealed a periodic structure of the AIS in which the cytoskeletal proteins form rings spaced ~190 nm apart [3, 4]. During embryonic development, AIS assembly occurs shortly after a neuron is generated [5]. AnkG is first restricted to the proximal part of the axon before recruiting other AIS proteins to this region [6–8]; in the absence of AnkG, ion channels and the major cytoskeletal proteins cannot assemble at the proximal axon [8–10].

The AIS length and location along the axon, which vary across different subtypes of neurons, are crucial for modulating excitability and ensuring the temporal precision of spike coding [11–14]. AIS plasticity has been well documented as changes in length and location as well as an alteration in diameter in response to neuronal activity in vitro and in vivo [12, 13, 15, 16]. Changes in the AIS length and location can in turn impact AP generation and hence neuronal excitability [12, 17]. AIS plasticity requires Ca²⁺-calmodulin-calcineurin signaling activated by Ca²⁺ influx through L-type Ca²⁺ channels [15]. How the AIS length and location affect AP generation is well known, but little is known about the factors that control the AIS diameter in mature neurons, beyond the fact that the diameter is correlated with soma size [18] and is enlarged in motor neurons of the spinal cord from human patients and from the G127X SOD1 mouse model of amyotrophic lateral sclerosis [18, 19].

In this study, we found that COUP-TFI (chicken ovalbumin upstream promoter-transcription factor I), also known as NR2F1, was capable of regulating AIS diameter in both immature and mature neurons. In cortical pyramidal cells, deletion of COUP-TFI in sparse single neurons or populations of neurons reduced AIS diameter and caused a shorter AP, a slower AP onset, and a depolarized AP threshold, whereas overexpression of COUP-TFI produced opposite effects. Interestingly, population-wide ablation decreased spontaneous excitatory postsynaptic current (sEPSC) frequency, but had no effect on miniature excitatory postsynaptic current (mEPSC) frequency or the density of spines. By contrast, sparse single-cell deletion increased sEPSC frequency, mEPSC frequency, and the density of spines. Thus, these data suggest that an alteration in AIS diameter can re-scale synaptic input through a negative feedback mechanism known as homeostatic plasticity.

Materials and Methods

Animals

COUP-TFI flox/flox (fl/fl) C57BL/6 mice (a kind gift from Dr. Ke Tang) were crossed with Emx1-Cre knock-in C57BL/6 mice (a kind gift from Dr. Ke Tang, Mouse Genome Informatics ID: 1928281) or Emx1-CreER knock-in C57BL/6 mice (a kind gift from Dr. Yong-Chun Yu, Fudan University) to generate an Emx1-Cre; COUP-TFI fl/fl mouse line (conditional KO; cKO) and an Emx1-creER; COUP-TFI fl/fl mouse line (tamoxifen-induced cKO). Littermates with the COUP-TFI fl/fl or COUP-TFI fl/+ genotypes served as wild-type (WT) controls. All mice were maintained under standard conditions of 22 ± 2°C, 50% ± 10% relative humidity, and a 12-h light/dark cycle, with food and water available. All experiments were carried out in accordance with animal experimental protocols approved by the Animal Care and Use Committee of ShanghaiTech University.

Golgi-Cox Staining

Adult Emx1-Cre; COUP-TFI fl/fl mice and WT mice were used for Golgi-Cox staining. Mice were anesthetized by chloral hydrate [5% in 0.1 mol/L phosphate-buffered saline (PBS), 10 μL/g], transcardially perfused with 0.1 mol/L PBS, and then quickly washed with double-distilled H₂O (ddH₂O) to remove blood from the surface. Golgi staining was performed according to the standard protocol provided with the FD Rapid GolgiStain^TM Kit (Cat #: PK401, FD NeuroTechnologies, Inc., Columbia, USA). Briefly, fresh brains were treated with a mixture of Solutions A and B (prepared immediately before sacrifice) in the dark at room temperature, left for 3 weeks, and then transferred into Solution C to be incubated for another 5 days. Afterwards, the brains were sectioned into 200 μm coronal slices using a vibratome. The sections were washed with ddH₂O and incubated in Solutions D and E mixed buffer for 10 min followed by a ddH₂O wash and dehydration with 75%, 95%, and 100% ethanol. The slices were treated with xylene (Cas #: 1330-20-7, Greagent, Shanghai, China) and cover-slipped with the DPX Mountant (Lot #: BVBH4393V, Sigma, St. Louis, USA).

Immunostaining

For brain tissue preparations, mice were anesthetized by chloral hydrate (5% in 0.1 mol/L PBS, 10 μL/g), transcardially perfused with paraformaldehyde (PFA) in 0.1 mol/L PBS at pH 7.4 (1% PFA for AnkG or COUP-TFI staining and 4% PFA for EGFP or βIV-spectrin). Brains were removed, post-fixed for 3 h at 4°C, and then transferred into 20% sucrose in 0.1 mol/L PBS at 4°C for >24 h. Coronal sections of the dehydrated brains were cut at 25 µm on a cryostat microtome (Leica CM1950, Wetlar, Germany). Immunostaining was performed as described previously [20]. The primary antibodies used in this study were as follows: mouse anti-COUP-TFI (1:500; PP-H8132-00, PPMX, Tokyo, Japan), rabbit anti-Pan-Na_V1 (1:500; SKU 75-405, Antibodies Inc. Neuromab, Davis, USA), rabbit anti-Na_V1.6 (1:500; ASC009, Alomone Labs, Jerusalem, Israel), mouse anti-AnkG (1:500; SC-137105, Santa Cruz Biotechnology, Santa Cruz, USA), chicken anti-GFP (1:1000; GFP-1020, Aves Lab, Tigard, USA), rabbit anti-NeuN (1:1000, ABCam, Cambridge, UK), and rabbit anti-βIV-spectrin (1:1000, a gift from Matthew N. Rasband).

Electrophysiological Recordings

Electrophysiological recordings were performed as previously described [20–22]. Adult mice were deeply anesthetized by 2% isoflurane with O₂ at 1 L/min supplied via animal anesthesia vaporizers (580S, RWD Life Science, San Diego, USA) at and transcardially perfused with ice-cold ACSF consisting of (in mmol/L): 126 NaCl, 4.9 KCl, 1.2 KH₂PO₄, 2.4 MgSO₄, 2.5 CaCl₂, 26 NaHCO₃, 20 glucose. Adult brain slices for electrophysiological recordings were prepared as previously described [21, 22]. Briefly, mice were anesthetized by isoflurane, quickly decapitated, and brains were transferred into ice-cold modified ASCF containing (in mmol/L): 93 N-methyl-d-glucamine (NMDG), 93 HCl, 2.5 KCl, 1.2 NaH₂PO₄, 30 NaHCO₃, 20 HEPES, 25 glucose, 5 sodium ascorbate, 2 thiourea, 3 sodium pyruvate, 10 MgSO₄, and 0.5 CaCl₂ at pH 7.35. The brain was coronally sliced at 350 μm on a vibratome at 0.03 mm/s. The slices were transferred to 37°C NMDG solution, incubated for 15 min, and then transferred to 37°C ASCF for another 1 h. Recordings were made from pyramidal cells within layers II/III of the mouse motor cortex at room temperature (~24°C). The patch-clamping rig was equipped with an upright microscope (BX51WI, Olympus, Tokyo, Japan) with a 60× objective lens (water-immersion, NA 1.00) and differential interference contrast. Recording pipettes (~12 MΩ) were manufactured on a P1000 micropipette puller (Sutter Instruments, Novato, USA) and filled with an internal solution containing (in mmol/L): 136 K-gluconate, 6 KCl, 1 EGTA, 2.5 Na₂ATP, 10 HEPES (280 mOsm, pH 7.2 with KOH). Data were collected with a low-pass filter at 2 kHz using Multiclamp 700B (Axon Instruments, Inc., Union City, USA) and a Digidata 1322 A/D converter (Axon Instruments, Inc.), and sampled at 10 kHz using pCLAMP software (Axon Instruments, Inc.). Recordings with series resistance >30 MΩ were excluded from further analyses. The first action potential elicited by step current injections was chosen for analyses with Clampfit (version 10.7, Axon Instruments, Inc.). The threshold of action potentials was documented as the membrane voltage (V) at the first point at which the dV/dt reached 10 V/s during the rising phase of the phase-plane plot of the membrane voltage against its change rate. The AP onset rapidness is documented as the slope of the phase plot at the threshold voltage [14]. Peak amplitude was calculated as the difference between the threshold and the peak point. mEPSCs were recorded in the presence of tetrodotoxin (1 µmol/L TTX, Tocris, Bristol, UK ) and bicuculline (10 µmol/L BMI, Tocris). mIPSCs were recorded in the presence of TTC (1 µmol/L), D-2-Amino-5-phosphonovalerate (50 µmol/L D-APV, Tocris) and 6-cyano-7-nitroquinoxaline-2,3-dione (10 µmol/L CNQX, Tocris). MiniAnalysis (version 6.0.7 Synaptosoft, Inc., Fort Lee, USA) was used to document synaptic currents. The threshold was set at 8 pA for event detection.

Confocal Imaging and Quantification

Immunostaining was imaged using Leica SP8 STED 3X equipped with Leica sCMOS DFC9000 at 40× (1.0 NA). The z-stack was set at 0.2–0.5 μm. The AISs that emerged from the soma were chosen for further analyses. The diameter of the AIS at the start, middle, and end points was measured based on AnkG and βIV-spectrin immunostaining using ImageJ (NIH)-Fuji (Rasband, W.S., ImageJ, National Institutes of Health, Bethesda, USA). The start and end of the AIS was defined as the point of AnkG or βIV-spectrin immunoreactivity that fell below 10% of the maximum fluorescence intensity [20, 23, 24]. The diameter of the AIS was estimated as the width of the three-dimensional (3D) reconstruction of the fluorescence of AnkG or βIV-spectrin staining from a series of z-stack confocal images with Imaris (version 9.1.2, Bitplane, Inc., Zurich, Switzerland) [25]. The length of a line drawn along an AIS from the start point to the end point was considered to be the length of the AIS. EGFP-expressing or biotin-filled neurons from the z-projected confocal images were outlined with ImageJ to estimate the maximal 2D soma area [26]. The distance between the AIS and soma was the shortest length of the linear path from the base of the soma (EGFP) to the start point of the AIS immunostained with antibodies against AnkG or βIV-spectrin [26]. To calculate the intensity of AIS fluorescence, after the confocal images were thresholded and the background was subtracted, the mean intensity of an AIS was measured with ImageJ and the integrated intensity was calculated as the mean intensity times the AIS area. In cKO mice (Emx1-Cre-COUP-TFI fl/fl mice), AISs from 5 locations containing 4 corners and 1 middle field (1/25 of the total area) in the motor cortex were chosen for analyses. A total of 20 AISs from one image were averaged for each brain slice. For hippocampal AISs, a total of 15 AISs that were discernible from 3 locations (2 sides and 1 middle) were analyzed for each brain slice.

To determine spine density, neurons were filled with neurobiotin through recording pipettes and Golgi staining was applied to neurons from cKO mice. The dendritic length and number of spines were measured with ImageJ (NIH)-Fuji.

Virus Delivery

The EGFP-Cre and EGFP pAAV2s were from Shanghai OBiO Technology Corp., Ltd (Shanghai, China). For virus delivery, mice were deeply anesthetized by chloral hydrate (5% in 0.1 mol/L PBS, 10 μL/g), the head was fixed in a stereotaxic apparatus, and one dose of AAV2/9-CRE or EGFP virus (1 µL, 1 × 10¹⁰ genome copies per mL, GC/mL) was injected into the motor cortex at the coordinates 1.10 mm anterior to bregma, 1.5 mm lateral to the midline, and 1.0 mm below the bregma. Two weeks after adeno-associated virus (AAV) injection, mice were sacrificed for immunostaining and electrophysiological recordings. iCre-EGFP and COUP-TFI-IRES-EGFP were cloned into the retrovirus vector. For embryonic retrovirus delivery, pregnant mice were deeply anesthetized by isoflurane by 2% isoflurane with O₂ at 1 L/min supplied via animal anesthesia vaporizers (580S, RWD Life Science). On embryonic day 14.5 (E14.5), mouse embryos were exposed to a clean environment and one dose of retrovirus (1 µL, 1 × 10⁷ transducing units per mL, TU/mL) mixed with dye was injected into the lateral ventricle through a glass micropipette. The pregnant mice were sutured and allowed to recover.

Data Analysis and Statistics

To ensure the reliability of our data, pyramidal cells were chosen based on the following criteria: (1) they were located at the lower half of layers II/III; (2) they possessed a typical dendrite; (3) their AISs emanated from the soma; and (4) their soma area varied within ±20%. Data analyses were conducted by investigators blinded to the experimental groupings. Statistical analyses and data plots were performed using GraphPad Prism 5.0 (GraphPad Software Inc., San Diego, USA). Data were tested for significant differences between groups using the unpaired t-test and one-way or two-way ANOVA with post hoc Bonferroni’s multiple comparisons test, and P < 0.05 was considered significant. All figures were composed using Illustrator software (Adobe Systems, Inc., San Jose, USA).

Results

Ablation of COUP-TFI Reduces AIS Diameter and Impairs AP Generation During Development

To investigate the mechanism that regulates AIS diameter, mouse brain sections were immunostained for AnkG, a general AIS marker, at different developmental stages. We analyzed the diameter of the AIS at the start (close to the soma) and middle points in cortical pyramidal neurons and found that at both sites, the AIS diameters progressively increased during normal postnatal development. Consistent with the findings from a recent study [27], AIS diameter was significantly reduced in cortical pyramidal cells from conditional COUP-TFI-null mice (Emx1-Cre;COUP-TFI fl/fl mice, referred to COUP-TFI cKO), compared with those of controls (COUP-TFI fl/fl or COUP-TFI fl/+ mice) (Fig. 1A–C). We then restricted our study to layers II/III of the motor neocortex because this region was well preserved in the COUP-TFI cKO mice. The AIS was similar in diameter in WT and cKO mice at postnatal day 7 (P7), when the AIS completes its assembly but still undergoes developmental expansion of its diameter [9]. During subsequent development, the diameter of the AIS continued to expand in WT cortical pyramidal cells, but not in COUP-TFI-deficient neurons (Fig. 1B, C), so AIS diameter was significantly smaller in cKO neurons at P30. Immunostaining for another AIS marker, βIV-spectrin, confirmed this reduction (Fig. S1A–C). In contrast, ablation of COUP-TFI caused the bending and misdirection of the AIS, but had no effect on the diameter or length in hippocampal CA1 pyramidal cells (Fig. S2), reminiscent of the bending axon of COUP-TFI-mutant neurons [28].

Fig. 1.

Embryonic deletion of COUP-TFI in cKO mice prevents the developmental expansion of AIS diameter and AP maturation. A Left panels, representative images of AnkG staining in layers II/III of the neocortices from P30 COUP-TFI fl/fl (WT control; Ctrl) and COUP-TFI fl/fl^Emx1-Cre (cKO) mice. Right panels, 3D reconstructed AISs from the corresponding boxed regions. B, C Quantification of AIS diameter at the start point (B) and the middle (C) point for both control and cKO mice at P7, P10, and P30 (control, n = slices from 5 mice; cKO, n = brain slices from 5 mice). D Overlapping sample traces of the first APs in cortical pyramidal neurons from control and cKO mice (V_m, initial membrane potential before current injection; note that V_m is similar in WT and cKO). E Phase-plane plots of membrane voltage and its rate of change in response to current injection as in D [slope (dashed lines, θ) of the phase plot at the threshold voltage is defined as AP onset rapidness]. F–H Quantification of AP threshold (F), onset rapidness (G), and peak amplitude (H) for both control and cKO mice at P7, P10, and P30 (P7: control, n = neurons from 3 mice, cKO, n = neurons from 3 mice; P10: control, 3 mice, cKO, 3 mice; P30: control, 6 mice, cKO 5 mice). *P < 0.05; **P < 0.01; ***P < 0.001; two-way ANOVA with post hoc Bonferroni tests. Data are presented as the mean ± SEM.

To assess the functional consequences of reduced AIS diameter, we made whole-cell electrophysiological recordings of APs in the soma of cortical pyramidal neurons, finding no difference in the AP amplitude, onset rapidness, or threshold potential between cKO and WT at P7, but APs were shorter and displayed a slower onset rapidness and a more positive threshold in cKO neurons at P30 (Fig. 1D–H), perfectly correlating with the changes in AIS diameter. To determine whether impaired AP generation is due to a reduction in the expression of Na_V channels at the AIS, we immunostained brain sections with antibodies against Na_V1 and Na_V1.6, the major types of Na_V channel distributed in the AIS. Unexpectedly, while the integrated fluorescent intensity of either Na_V1 or Na_V1.6 in the AIS remained unchanged, their mean intensities were significantly up-regulated in cKO mice compared with WT controls (Fig. 2A–F), at least suggesting that changes in Na_V channels cannot account for the altered AP properties. Likely, the mean intensity but not the integrated intensity of AnkG fluorescence in the AIS was somehow increased in cKO mice (Fig. 2G–I).

Fig. 2.

Loss of COUP-TFI increases the density of Na_V in the AIS. Representative images of Na_V1.6 (A), PanNa_V1 (D), and AnkG (G) staining in layers II/III of the neocortices from P30 WT controls and cKO mice. Quantification of the normalized mean intensity and the normalized integrated intensity of Na_V1.6 (B, C), PanNa_V1 (E, F), and AnkG (H, I) (WT, n = slices from 4 mice; cKO, n = slices from 5 mice). The cKO intensity is normalized to the control. *P < 0.05; **P < 0.01; ***P < 0.001; Student’s t-test. Data are presented as the mean ± SEM.

Numerous studies have reported that COUP-TFI plays important roles in cortical arealization/lamination, neuronal migration, and neural circuit formation as well as the specification of neuronal identity [29]. These findings raise the possibility that neurons from the same regions of WT and cKO cortices might belong to different subtypes of pyramidal cells, which could confound data interpretation concerning AIS diameter and function. To address this caveat, we delivered low-titer retroviruses expressing iCre fused with EGFP into the lateral ventricle of E14.5 embryos from COUP-TFI fl/fl dams (Fig. 3A), which allowed us to delete COUP-TFI in sparse single cortical neurons in the layers II/III. Retroviruses expressing EGFP only served as a control. Immunostaining confirmed efficient deletion of COUP-TFI by iCre expression (93%; 27 out of 29 iCre-expressing neurons, Fig. S3). In line with the results from cKO mice, we observed a significant reduction in AIS diameter in sparse COUP-TFI^−/− mutant neurons compared with neurons expressing EGFP only or with neighboring non-iCre-expressing neurons (Fig. 3B, C). In contrast, the diameter of AISs at the distal end point was not altered (Fig. 3C), indicating that axonal diameter is not affected after the deletion of COUP-TFI. Note that AIS diameter and length were similar in EGFP-only and non-iCre neurons. Moreover, we found no difference in AIS length, the distance between the AIS and the soma, or the maximal cell area between WT and cKO neurons (Fig. 3D–F), suggesting that the regulation of the AIS by COUP-TFI is specific to the diameter. Again, sparse single-cell COUP-TFI deletion resulted in a more positive threshold, decreased AP onset rapidness, and smaller AP amplitude (Fig. 3G–K).

Fig. 3.

Embryonic deletion of COUP-TFI in sparse single cells results in reduced AIS diameter and impaired AP generation. A Schematic of the experimental paradigm. B Left panels, representative images of AnkG immunostaining for EGFP- or iCre-EGFP-infected neurons in layers II/III of mouse neocortices (arrowheads, start points of AISs; dashed lines, maximal area of cells). Right panels, 3D reconstructed AISs in the corresponding left panels. C Quantification of AIS diameter at the start, middle, and end points in EGFP- and iCre-EGFP-expressing neurons or non-EGFP surrounding iCre-EGFP neurons (neighboring non-iCre) (EGFP, n = neurons from 4 mice; iCre-EGFP and non-iCre, n = neurons from 4 mice). D–F Quantification of the length of the AIS (D), maximal cell area (E), and distance between AIS and the soma (F) (D–F: EGFP, n = neurons from 4 mice; iCre-EGFP and non-iCre, n = neurons from 4 mice). G Overlapped sample traces of the first APs. H Phase-plane plots of membrane voltage and its rate of change in response to current injections. I–K Quantification of AP threshold (I), onset rapidness (J), and peak amplitude (K) (EGFP, n = neurons from 6 mice; iCre-EGFP, n = neurons from 4 mice). Data are presented as the mean ± SEM. **P < 0.01; ***P < 0.001 in C, two-way ANOVA with post hoc Bonferroni tests. Red and blue stars in C indicate significant differences between iCre-EGFP and EGFP and between iCre-EGFP and neighboring non-iCre, respectively. Other P-values were calculated using Student’s t-test.

Overexpression of COUP-TFI Increases AIS Diameter and Promotes AP Generation

To determine whether overexpression of COUP-TFI increases AIS diameter, we similarly delivered retroviruses expressing EGFP-only or COUP-TFI-IRES-EGFP into the lateral ventricle at E14.5. AIS diameter was significantly increased following COUP-TFI overexpression (Fig. 4A–C), resulting in a more negative AP threshold and increased AP onset rapidness and amplitude (Fig. 4D–H). Thus, COUP-TFI controls the developmental expansion of AIS diameter and regulates AP generation during postnatal development.

Fig. 4.

Embryonic overexpression of COUP-TFI in sparse single neurons causes an increase in AIS diameter and facilitates AP generation. Mice that received retroviruses expressing EGFP or COUP-TFI-IRES-EGFP at E14.5 were sacrificed at P30 for immunostaining and electrophysiological recordings. A Left panels, representative confocal images of co-immunostaining of EGFP (green) and AnkG (red). Right panels, 3D reconstructed AISs from the dashed box regions. B, C Quantification of AIS diameter at the start (B) and middle (C) points in neurons expressing EGFP, COUP-TFI-IRES-EGFP (overexpression, OE), or neighboring non-OE (EGFP, n = neurons from 4 mice; COUP-TFI OE and neighboring non-OE, n = neurons from 4 mice). D Overlapped sample traces of the first APs (V_m, initial membrane potential). E Phase-plane plots of membrane voltage and its rate of change in response to current injection. F–H Quantification of AP threshold (F), onset rapidness (G), and peak amplitude (H) for the first AP evoked (EGFP, n = neurons from 6 mice; COUP-TFI OE, n = neurons from 8 mice). Data are presented as the mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001 in B and C, one-way ANOVA with post hoc Bonferroni tests. Other P-values were calculated using Student’s t-test.

COUP-TFI is Required for Maintenance of AIS Diameter in the Adult Neocortex

Can AIS diameter be regulated in mature cortical neurons? To address this question, we injected AAVs expressing EGFP-only or iCre-P2A-EGFP into layers II/III of the motor cortex of COUP-TFI fl/fl mice at P30 (Fig. 5A). Brain sections (at ~P45) were immunostained for AnkG 15 days after virus injection. The quantitative comparison showed a significant reduction in AIS diameter, a more positive AP threshold, and decreased AP onset rapidness and amplitude in iCre-P2A-EGFP-expressing neurons in contrast to those expressing EGFP-only or neighboring non-iCre-expressing neurons (Fig. 5B–I), suggesting that COUP-TFI is required for the maintenance of AIS diameter in adult cortical neurons.

Fig. 5.

Adult deletion of COUP-TFI in sparse single cells leads to reduced AIS diameter and impaired AP generation. A Schematic of the experimental paradigm. B Left panels, representative confocal images of AnkG immunostaining for these groups. Right panels, 3D reconstructed AISs. C, D Quantification of AIS diameter at the start (C) and middle (D) points in neurons expressing EGFP-only, iCre-P2A-EGFP, or neighboring non-iCre. (EGFP, n = neurons from 4 mice; iCre-P2A-EGFP and neighboring non-iCre, n = neurons from 6 mice). E Overlapped sample traces of the first APs. F Phase-plane plots of membrane voltage and its rate of change in response to current injections. G–I Quantification of AP threshold (G), onset rapidness (H), and peak amplitude (I) for the first AP (EGFP, n = neurons from 4 mice; iCre-P2A-EGFP, n = neurons from 5 mice). Data are presented as the mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001 in C and D, one-way ANOVA with post hoc Bonferroni tests. Other P-values were calculated using Student’s t-test.

Positive Correlation Between AIS Diameter and AP Generation

So far, manipulation of COUP-TFI expression implicates a correlation between AIS diameter and AP generation. To further determine whether this is applicable to the mouse neocortex under normal physiological conditions, we made whole-cell electrophysiological recordings from pyramidal neurons in layers II/III of the motor cortex. A total of 84 AISs from 84 pyramidal cells were measured by immunostaining of recorded neurons with antibodies against βIV-spectrin (Fig. 6A). To minimize the effects of AIS length and location on APs, 76 neurons (out of 84) that fell within the 95% confidence intervals predicted by normal probability plots of AIS length and the distance between the AIS and the soma were selected for further analyses (Fig. 6B). These analyzed neurons displayed AIS lengths ranging from 15 µm to 40 µm and the distance was < 7 µm. To determine the linear correlation between AIS diameter and AP parameters, the diameter of the AIS was plotted against AP peak amplitude, onset rapidness, and threshold. Apparently, a neuron with a larger AIS diameter generated a larger amplitude and a faster AP onset (Fig. 6C–F). The larger the AIS diameter, the more negative the AP threshold (Fig. 6G). These results suggest a positive and strong correlation between AIS diameter and AP generation.

Fig. 6.

AIS diameter is strongly correlated with AP parameters. A Left panels, representative images of βIV-spectrin staining in three recorded cortical pyramidal cells filled with neurobiotin. Recorded neurons were post hoc immunostained for βIV-spectrin after the completion of whole-cell recordings. Right panels, 3D reconstructions of the AISs in dashed box regions. B Normal probability plots of AIS length and the distance between AIS and soma. C Overlapped sample traces of the first APs corresponding to the three neurons shown in A. D Phase-plane plots of membrane voltage and its rate of change in response to current injection from the three neurons in C. E–G Linear probability plots of AIS middle diameter against AP peak amplitude (E), onset rapidness (F), and threshold (F) in cortical pyramidal cells of the mouse neocortex; n = neurons from 9 mice; correlation significance defined by P < 0.05 (t-test).

Tuning of Synaptic Plasticity by AIS Diameter

Neuronal excitability is fine-tuned through altering AIS location and length in response to changes in synaptic input or neuronal activity [12, 13]. Conversely, synaptic input can be re-scaled by chronically altering neuronal excitability for compensatory neuronal output [30, 31]. The above-described association between the AIS and neuronal excitability suggests that synaptic input can be fine-tuned by AIS diameter. To address this issue, we examined the spontaneous network of COUP-TFI^−/− mutant neurons from cKO mice using whole-cell electrophysiological recordings of sPSCs and mPSCs in pyramidal cells with a pipette solution containing 6 mmol/L Cl⁻ (the equilibrium potential of Cl⁻ is −77.6 mV) at a holding potential of −70 mV. Since the holding potential was close to the equilibrium potential of Cl⁻, PSCs were roughly considered to be EPSCs. The frequency but not the amplitude of AP-dependent sEPSCs was found to be significantly reduced in neurons from cKO mice compared with those from WT controls (Fig. S4A–D). But the frequency and amplitude of AP-independent mEPSCs, as well as the density of spines in cKO mice, were comparable to those in WT controls (Fig. S4E–I). Given that population-wide deletion of COUP-TFI in cKO mice caused both the shrinkage of AIS diameter in presynaptic neurons and many developmental abnormalities, the reduced sEPSC amplitude could be attributable to dampened excitability in presynaptic neurons resulting from AIS diameter shrinkage or to other developmental changes. To address these questions, we injected tamoxifen (250 mg/kg) (TMX) intraperitoneally into COUP-TFI fl/fl^Emx1-CreER mice at P15 (Fig. S5A), when neuronal migration, axonal projection, and cortical patterning were completed. Immunostaining confirmed the ablation of COUP-TFI in COUP-TFI fl/fl^Emx1-CreER mice (Fig. S5B). In agreement with the results from cKO mice, the AIS diameter was evidently shrunk and AP generation was impaired in cortical neurons from COUP-TFI fl/fl^Emx1-CreER mice compared with COUP-TFI fl/fl mice (Fig. S5C–I). Electrophysiological recordings revealed a decrease in sEPSC frequency, but no significant changes in sEPSC amplitude, mEPSC frequency, or amplitude, demonstrating that the spontaneous network was impaired by the shrinkage of AIS diameter by way of reduced neuronal output (Fig. S5J–O).

We next recorded sEPSCs in sparse single COUP-TFI^−/− mutant neurons from mice that received low-titer retrovirus injections at the embryonic stage. Contrary to the results of population-wide deletion, we found that sparse single-cell COUP-TFI ablation caused a significant increase in sEPSC frequency, but no change in amplitude (Fig. 7A–C). Interestingly, COUP-TFI overexpression in sparse single neurons resulted in reduced sEPSC frequency and amplitude (Fig. 7A–C). Thus, these results suggest that an altered spontaneous network at the single-cell level results from homeostatic plasticity rather than direct genetic regulation of synaptic inputs by COUP-TFI; that is to say, the gain of synaptic inputs is scaled to compensate for a reduction in neuronal excitability resulting from decreased AIS diameter.

Fig. 7.

COUP-TFI overexpression and deletion at the single-cell level have opposed effects on the receiving spontaneous network. A–C Sample traces (A), quantification of the frequency (B), and the amplitude (C) of sEPSCs recorded in single EGFP-positive cortical pyramidal neurons from mice that received retroviruses expressing EGFP only, iCre-EGFP, or COUP-TFI-IRES-EGFP at the embryonic stage. (B and C: EGFP, n = neurons from 6 mice; iCre-EGFP, n = neurons from 4 mice; COUP-TFI-IRES-EGFP, n = neurons from 8 mice). Data are presented as the mean ± SEM. **P < 0.01; ***P < 0.001, one-way ANOVA with post hoc Bonferroni tests.

Since COUP-TFI is required for maintaining AIS diameter in the adult cortex, we determined whether a change in AIS diameter could induce synaptic plasticity in adult cortical neurons. sEPSC frequency was significantly increased in sparse single COUP-TFI^−/− mutant neurons (~P45) in mice that received an AAV injection at P30, compared with EGFP-only or neighboring non-iCre neurons (Fig. 8A–C). The significant difference between iCre neurons (mutant neurons) and neighboring non-iCre neurons excluded the possibility that iCre-expressing AAVs altered the excitability of presynaptic neurons. This increase likely resulted from increased mEPSC frequency (Fig. 8D–F) and spine density (Fig. 8J, K), but not from a change in inhibitory synaptic transmission (Fig. 8G–I). Taken together, these data suggest that AIS diameter fine-tunes synaptic inputs in adult cortical neurons, thereby enabling the maintenance of neuronal homeostasis.

Fig. 8.

Adult deletion of COUP-TFI in sparse single cortical neurons promotes excitatory synaptic transmission. A–I Sample traces and quantification of sEPSCs (A–C), mEPSCs (D–F), and miniature inhibitory postsynaptic currents (mIPSCs) (G–I) from mice that received AAV injection at P30. J, K Representative images (J) and quantification (K) of spines in pyramidal cells recorded with the pipette solution containing biotin. (B–K: EGFP, n = neurons from 4 mice; neighboring non-iCre and iCre-P2A-EGFP, n = neurons from 5 mice). EGFP only and neighboring non-iCre serve as controls. Data are presented as the mean ± SEM. **P < 0.01; ***P < 0.001, one-way ANOVA with post hoc Bonferroni tests.

Discussion

In this study, we found that AIS diameter in layer II/III pyramidal cells of the motor cortex was smaller in cKO mice than in WT controls, owing to a defect in the developmental expansion of AIS diameter after removal of COUP-TFI. Adult deletion of COUP-TFI similarly reduced the AIS diameter, but had no effect on the length or the distance between AIS and soma in layer II/III pyramidal cells of the mouse motor cortex. This reduction of AIS diameter was not observed in the hippocampus of cKO mice, possibly due to the fact that both COUP-TFI and COUP-TFII are expressed in the hippocampus and that loss of COUP-TFI could be compensated by COUP-TFII [32, 33]. Taken together, these results suggest that COUP-TFI drives developmental expansion of the AIS diameter, and is also required for the maintenance of AIS diameter in adult cortical neurons. Our results are partially consistent with the findings from a recent study showing significant reductions in AIS diameter and length in layer V pyramidal cells in COUP-TFI cKO mouse somatosensory cortex [27].

To clarify this inconsistency regarding a change in AIS length, we performed quantitative analyses of AIS diameter and length in layer V of the motor cortex and in layer V of the somatosensory cortex. Consistent with the findings from this study in layer II/III neurons, the AIS diameter but not the length of AISs was decreased in layer V of the cKO motor cortex in contrast to WT controls (Fig. S6A–D). Both AIS diameter and length were significantly reduced in layer V of the somatosensory cortex of cKO mice compared with WT controls (Fig. S6E–H), in agreement with the results of the Del Pino study. These results together suggest that the regulation of AIS length by COUP-TFI may be specific to cortical regions, reminiscent of the involvement of COUP-TFI in region specification, possibly due to its gradient expression along the caudal-rostral axis of the cortex [28, 29].

By determining the functional consequences of changes in AIS diameter, we revealed a positive correlation between AIS diameter and AP generation. While APs recorded in the soma of COUP-TFI-deficient neurons displayed a more positive threshold potential and smaller amplitude, somatic APs in COUP-TFI-overexpressing neurons showed a more negative threshold potential and larger amplitude. It is worth noting that these changes in APs parallel the shrinkage and expansion of AIS diameter, respectively. The peak amplitude of somatic APs includes miscellaneous components of Na⁺ currents, mainly resulting from the activation of somatic Na⁺ channels [34]. As APs are initiated in the AIS, the threshold potential and the initial onset rapidness of somatic APs at least reflect the capacity of the AIS to generate APs [14, 26, 34]. Previous reports have shown that the density of Na_V1 channels in the AIS, especially Na_V1.2 and Na_V1.6, is crucial for the generation of APs [1], so that the level of expression of Na_V1 channels could be a contributing factor for altered AP threshold. Our results showed that the density (mean intensity) was unexpectedly increased after deletion of COUP-TFI in cKO mice though the total amount (integrated intensity) of Na_V1 or Na_V1.6 in an AIS was not changed. Del Pino et al. also found no change in the global expression of Na⁺ channels in cKO mice. Increased Na_V1 density in the AIS actually facilitates AP generation, which is contrary to our finding showing a more positive threshold potential in COUP-TFI^−/− neurons. Thus, these results suggest that altered AP generation likely results from changed AIS diameter. However, these changes in AP threshold and amplitude were not found in COUP-TFI cKO mice of the Del Pino study. This discrepancy remains to be resolved.

Our work shows that the association between AIS diameter and AP generation is a physiological phenomenon in WT mouse neocortical pyramidal cells. By specifically studying a subtype of layer II/III cortical pyramidal cells, we found that neurons with a large diameter tended to fire APs at a more negative threshold potential. Recently, results from a study of layer V cortical pyramidal cells using similar methods indicated that AP generation does not correlate with AIS position, suggesting that AIS diameter rather than position is important for AP initiation.

Although COUP-TFI plays multi-faceted roles in brain development [29], its involvement in mediating AIS diameter provides us an opportunity to study how a change in AIS diameter affects synaptic plasticity. We recorded a significant reduction in the frequency of AP-dependent sEPSCs but no change in the frequency of AP-independent mEPSCs in cKO mice. Similar results were found in TMX-induced cKO mice. Considering that TMX was applied at P15, after the completion of the majority of neuronal development, changes in cortical development might merely have subtle effects on the spontaneous network. Since population-wide COUP-TFI deletion also led to impaired AP generation in the presynaptic neurons of cKO mice, this reduction in sEPSC frequency is likely due to dampened excitability in presynaptic neuronal excitability associated with reduced AIS diameter. These results are in agreement with previous findings showing decreases in sPSC frequency and in the complexity of dendritic morphology in cKO mice [35]. By contrast, deletion of COUP-TFI in sparse neurons not only resulted in a drastic increase in the frequency of sEPSCs, but also induced the scaling up of mEPSC frequency and the density of spines. Conversely, overexpression of COUP-TFI in sparse neurons caused a significant reduction in the sEPSC frequency. It is important to note that these differential effects on synaptic inputs in these sparse neurons were inversely correlated with changes in AIS diameter and AP generation. Given that COUP-TFI expression was only altered in sparse single cortical neurons, neither intrinsic excitability nor AIS diameter in their presynaptic neurons was likely affected. Thus, excitatory synaptic inputs were scaled up to compensate for reduced neuronal excitability that resulted from the shrinkage of AIS diameter, therefore leading to the maintenance of neural homeostasis. Taken together, our findings suggest that AIS diameter tunes neuronal excitability, supporting the idea that neuronal excitability is tightly linked to AIS morphology such as the length and location [12, 13].

Currently, to our knowledge, little is known about how a neuron controls the diameter of its AIS. It has been hypothesized that an increase in AIS diameter under pathological conditions might be associated with the accumulation of neurofilaments (NFs) [18]. Deletion of COUP-TFI relieves its inhibition of fibroblast growth factor (FGF) signaling [29] and then leads to the phosphorylation of NFs through activation of mitogen-activated protein kinase signaling [36–39]. Phosphorylation of NFs promotes NF accumulation, which actually increases AIS diameter. This is inconsistent with our finding showing reduced AIS diameter in COUP-TFI^−/− neurons, suggesting that the mechanism for regulating AIS diameter under physiological conditions might differ from that under pathological conditions. Another possibility is that adducin, an actin-capping protein, might be involved in the regulation of AIS diameter by COUP-TFI because deletion of α-adducin, an isoform that is necessary for the assembly of functional tetrameric adducin, increases the diameter of both actin rings and axons [40]. Although α-adducin is largely absent in the AIS [41], we cannot exclude the likelihood that α-adducin is ectopically expressed in the AIS after deletion of COUP-TFI. Thus, further investigations are needed to elucidate whether these potential mechanisms underlie the control of AIS diameter in the neocortex.

In summary, results from this study provide important insights into the understanding of how AIS diameter is regulated in cortical neurons and its physiological functions at the cellular level. Human patients carrying haploinsufficient COUP-TFI exhibit intellectual disability, making it interesting to test whether AIS diameter contributes to cognitive functions.

Supplementary Information

Below is the link to the electronic supplementary material.

Conflict of interest

The authors declare no competing financial interests.