Ca²⁺‐independent transmission at the central synapse formed between dorsal root ganglion and dorsal horn neurons

Abstract

A central principle of synaptic transmission is that action potential‐induced presynaptic neurotransmitter release occurs exclusively via Ca²⁺‐dependent secretion (CDS). The discovery and mechanistic investigations of Ca²⁺‐independent but voltage‐dependent secretion (CiVDS) have demonstrated that the action potential per se is sufficient to trigger neurotransmission in the somata of primary sensory and sympathetic neurons in mammals. One key question remains, however, whether CiVDS contributes to central synaptic transmission. Here, we report, in the central transmission from presynaptic (dorsal root ganglion) to postsynaptic (spinal dorsal horn) neurons in vitro, (i) excitatory postsynaptic currents (EPSCs) are mediated by glutamate transmission through both CiVDS (up to 87%) and CDS; (ii) CiVDS‐mediated EPSCs are independent of extracellular and intracellular Ca²⁺; (iii) CiVDS is faster than CDS in vesicle recycling with much less short‐term depression; (iv) the fusion machinery of CiVDS includes Cav2.2 (voltage sensor) and SNARE (fusion pore). Together, an essential component of activity‐induced EPSCs is mediated by CiVDS in a central synapse.

Axonal action potential triggers synaptic transmission independent of Ca²⁺ in a mammalian central synapse.

Introduction

The tight coupling of synaptic neurotransmitter release to the action potential within a millisecond is critical for neuronal communication and brain function (Augustine et al, ¹⁹⁸⁷; Hopfield, ¹⁹⁹⁵; Sabatini & Regehr, ¹⁹⁹⁶; Neher & Sakaba, ²⁰⁰⁸). The classical view of synaptic transmission is based on the “Ca²⁺ hypothesis”: presynaptic membrane depolarization opens voltage‐gated Ca²⁺ channels (VGCCs), which leads to Ca²⁺ influx into the cytosol and triggers vesicular neurotransmitter release (Katz & Miledi, 1967; Augustine et al, ¹⁹⁸⁷; Geppert et al, ¹⁹⁹⁴; Jackson & Chapman, ²⁰⁰⁶; Neher & Sakaba, ²⁰⁰⁸; Sudhof, ²⁰¹²). In this half‐century dogma, synaptic transmission is directly triggered by Ca²⁺, synaptotagmin, and SNARE complex assembly (Sollner et al, ¹⁹⁹³; Jahn & Scheller, ²⁰⁰⁶), or membrane depolarization triggers synaptic transmission indirectly by mediating Ca²⁺ influx through VGCCs (Neher & Zucker, 1993; Felmy et al, ²⁰⁰³; Meinrenken et al, ²⁰⁰³; Jackson & Chapman, ²⁰⁰⁶; Parnas & Parnas, ²⁰¹⁰; Nanou & Catterall, ²⁰¹⁸). Since 2002, this Ca²⁺ hypothesis has encountered exceptions in a series of reports on Ca²⁺‐independent but voltage‐dependent secretion (CiVDS) in primary sensory dorsal root ganglion (DRG) neurons (Zhang & Zhou, 2002; Zhang et al, ²⁰⁰⁴; Zheng et al, ²⁰⁰⁹; Chai et al, ²⁰¹⁷; Huang et al, ²⁰¹⁹), where action potentials per se directly trigger somatic vesicular exocytosis.

In addition to the DRG (Zhang & Zhou, 2002), somatic CiVDS has been extended to trigeminal ganglion neurons (Sforna et al, ²⁰¹⁹), sympathetic superior cervical ganglion neurons, and neuroendocrine chromaffin cells (Huang et al, ²⁰¹⁹; Moya‐Diaz et al, ²⁰¹⁹). Further studies have revealed unique features of CiVDS compared with Ca²⁺‐dependent secretion (CDS): (i) rather than slow and dynamin‐dependent endocytosis following CDS (Artalejo et al, ²⁰⁰²; Ferguson et al, ²⁰⁰⁷; Wu et al, ²⁰¹⁹), CiVDS is coupled to a fast and dynamin‐independent but protein kinase A‐dependent endocytosis (Zhang et al, ²⁰⁰⁴); (ii) the vesicle pool replenishment of CiVDS is much faster than that of CDS (Zhang & Zhou, 2002; Huang et al, ²⁰¹⁹); (iii) at the single‐vesicle level, CiVDS has a smaller quantal size and faster release kinetics (Huang et al, ²⁰¹⁹); and (iv) VGCC activation plays dual roles—indirectly triggering CDS via Ca²⁺‐influx and directly triggering CiVDS through the “synprint” binding domain between VGCC and SNARE (Chai et al, ²⁰¹⁷; Nanou & Catterall, ²⁰¹⁸; Huang et al, ²⁰¹⁹). This enables a faster on/off‐gating for vesicle fusion via CiVDS in somata (Liu et al, ²⁰¹¹; Chai et al, ²⁰¹⁷; Huang et al, ²⁰¹⁹). The key question remains, however, whether CiVDS occurs in synaptic transmission.

In the present study, by combining electrophysiological recordings, live‐cell pHluorin imaging, optogenetic stimulation, and genetic manipulations, we report that CiVDS is present in the central synapses between presynaptic DRG and postsynaptic spinal dorsal horn (DH) neurons. The CiVDS‐mediated EPSC (EPSC_CiVDS) has much less short‐term depression than that of CDS‐mediated (EPSC_CDS) and predominates during sustained neuronal activity. Altogether, this work provides the first example of CiVDS‐mediated synaptic transmission and demonstrates complementary roles of CiVDS and CDS under physiological conditions.

Results

CiVDS mediates synaptic transmission in co‐cultured DRG and DH neurons

Our previous reports have shown robust CiVDS in the somata of freshly isolated DRG neurons (Zhang & Zhou, 2002; Zhang et al, ²⁰⁰⁴; Zheng et al, ²⁰⁰⁹; Chai et al, ²⁰¹⁷). To investigate whether CiVDS functions in synapses, we co‐cultured DRG neurons with postsynaptic DH neurons in which excitatory postsynaptic currents (EPSCs) were collected with whole‐cell patch‐clamp recordings (Fig 1A). Strikingly, local pulses of electrical field stimulation evoked EPSCs in postsynaptic DH neurons, even in 1 mM EGTA (a potent Ca²⁺ chelator)‐containing Ca²⁺‐free (0Ca) bath solution (Fig 1A). Specifically, ~39% of the total cells randomly encountered showed positive EPSC_CiVDS, and ~100% showed positive EPSC_CDS. Intracellular Ca²⁺ ([Ca²⁺]_i) recording validated that there was indeed no [Ca²⁺]_i rise during recordings in the 0Ca solution (Fig 1A and Appendix Fig S1A–C). In contrast, the EPSCs recorded in cultured hippocampal neurons were fully dependent on extracellular Ca²⁺ (Fig 1B), indicating the absence of CiVDS during synaptic transmission in hippocampal neurons and confirming that the 1 mM EGTA‐containing Ca²⁺‐free solution was sufficient to block Ca²⁺ transient at synapses.

Figure 1. Ca²⁺‐independent but voltage‐dependent synaptic neurotransmission in co‐cultured DRG–DH neurons.

• A
  (a1) Whole‐cell current recording of evoked EPSC signals in response to local electrical field stimulation (Estim, arrows) from a postsynaptic dorsal horn (DH) neuron co‐cultured with presynaptic dorsal root ganglion (DRG) neurons in Ca²⁺‐free (0Ca²⁺, black) and 2.5 mM Ca²⁺ bath (2Ca²⁺, green); (a2), evoked intracellular Ca²⁺ signals (dF/F₀) in a DRG neuron; inset, micrograph showing the setup for EPSC recording from DH neurons co‐cultured with EGFP‐expressing DRG neurons (scale bar, 20 μm).
• B
  (b1) Evoked EPSCs from cultured hippocampal neurons in Ca²⁺‐free (black) and 2.5 mM Ca²⁺‐containing solution (green); (b2) evoked intracellular Ca²⁺ signals (dF/F₀) in hippocampal neurons; inset, micrograph showing the setup for EPSC recordings from a hippocampal neuron (scale bar, 20 μm).
• C
  As in (A), except that 50 μM BAPTA‐AM was pre‐loaded into the DRG and DH neurons before recording in Ca²⁺‐free (black) or 2.5 mM Ca²⁺‐containing solution (green). BAPTA‐AM was loaded for 30 min at 37°C.
• D
  Left, quantification of amplitude as in (A) (n = 26 cells for 0Ca²⁺ and 29 cells for 2Ca²⁺). Right, quantification of EPSC amplitudes as in (C) (n = 16 cells for 0Ca²⁺ and 14 cells for 2Ca²⁺).
• E
  Evoked EPSC_CiVDS (0Ca²⁺, black) and total EPSCs (CDS + CiVDS, 2Ca²⁺, green) induced by paired‐pulse stimulation with a 50‐ms interval from DH neurons co‐cultured with DRG neurons. The 0Ca²⁺ and 2Ca²⁺ EPSCs were recorded in two different DH neurons.
• F
  Summary plot of paired‐pulse ratios as in (E) with different intervals (in 0Ca²⁺, n = 7 cells for 10 ms, 13 for 50 ms, 14 for 500 ms, 10 for 5 s, and 8 for 15 s; in 2Ca²⁺, n = 14 cells for 10 ms, 17 for 50 ms, 22 for 500 ms, 19 for 5 s, and 14 for 15 s). Inset shows the initial plot at an expanded scale.
• G
  Representative EPSCs induced by a 10‐Hz stimulus train in DH neurons co‐cultured with DRG neurons in Ca²⁺‐free (black) or 2.5 mM Ca²⁺‐containing solution (green). The 0Ca²⁺ and 2Ca²⁺ EPSCs were recorded in two different DH neurons.
• H
  Summary plots of the EPSC amplitudes as in (G), including CiVDS (0Ca²⁺, black), CDS + CiVDS (2Ca²⁺, green), and CDS (2Ca²⁺ − 0Ca²⁺, blue) (n = 11 cells).
• I
  As in (H), statistics for amplitudes of EPSC_CiVDS (0Ca²⁺, black) and EPSC_CDS (“2Ca²⁺” − “0Ca²⁺,” blue) evoked by single pulse (first EPSC, left) or 10 pulses (cumulative 10 EPSCs, Σ_EPSC, right) during 10‐Hz train stimulation (n = 11 cells).

Data information: All but (B) were from DRG and DH co‐cultures. EPSC recording and Ca²⁺ imaging were separate experiments. Data are shown as the mean ± s.e.m.; Mann–Whitney test for (D) and (F); paired Student's t‐test for (I); *P < 0.05, **P < 0.01, ***P < 0.001, ns, not significant.

To further exclude the possible contribution of intracellular microdomain Ca²⁺, BAPTA‐AM, a potent Ca²⁺ chelator, was pre‐loaded into the co‐cultured neurons. Consistent with our previous reports on somatic secretion (Zhang & Zhou, 2002; Chai et al, ²⁰¹⁷; Huang et al, ²⁰¹⁹), BAPTA had no effect on the EPSCs recorded in the 0Ca solution, but reduced the EPSCs recorded in 2Ca to that in 0Ca (Fig 1C and D). Thus, under physiological conditions (2Ca), the evoked EPSC contains both a CiVDS‐mediated EPSC (EPSC_CiVDS) and a pure CDS‐mediated EPSC (EPSC_CDS), or EPSC (2Ca) = EPSC_CiVDS + EPSC_CDS. The amplitude of EPSCs in 0Ca solution was ~50% of that in the standard 2.5 mM Ca²⁺ (2Ca) solution (Fig 1D). Consistently, the phenomenon was confirmed in co‐cultured DRG and DH neurons at physiological temperature (Appendix Fig S2A–C). Furthermore, we recorded spontaneous EPSC (sEPSC) and miniature EPSC (mEPSC) in 0Ca solution in postsynaptic DH neurons co‐cultured with DRG neurons (Appendix Fig S3A and B). The average amplitude of the mEPSC in 0Ca and 2Ca solutions is 27.45 ± 1.55 and 28.31 ± 1.74 pA, respectively, which is much smaller than that of CiVDS‐mediated EPSC (~200 pA from Fig 1D), suggesting that CiVDS‐mediated EPSC most likely arises from multiple vesicular release.

We next assessed the vesicle recycling rate by using paired‐pulse stimuli with different time intervals. Consistent with our previous findings in somatic secretion (Zhang & Zhou, 2002), the 80% recovery of the total (CiVDS + CDS) EPSC recorded in 2Ca solution required > 5‐s intervals (Fig 1E and F). In contrast, 80% recovery of the EPSC_CiVDS was achieved within 50 ms (Fig 1E), indicating a faster recycling rate/replenishment of vesicle pools in CiVDS than that in CDS in DRG synaptic terminals. In addition, we used physiological 10‐Hz train stimulation to evaluate the contribution of CiVDS during sustained synaptic transmission (Xu et al, ²⁰⁰⁰; Fang et al, ²⁰⁰⁵; Zheng et al, ²⁰⁰⁹). Intriguingly, EPSC_CiVDS showed much less short‐term depression than the total EPSCs in 2Ca solution (Fig 1G and H and Appendix Fig S4A–C). Importantly, the EPSC_CiVDS contributed 49% of the total EPSC (in amplitude) recorded in the 2Ca bath (EPSC_CiVDS + EPSC_CDS) during single‐pulse stimulation, and this increased to 87% during 10‐Hz train stimulation (Fig 1H and Appendix Fig S4C), suggesting that, in contrast to CDS, CiVDS makes a dominant contribution during sustained neural activity (Fig 1I and Appendix Fig S4C).

In addition, we performed live‐cell imaging of synaptophysin (Spy)‐pHluorin in the nerve terminals of DRG neurons co‐cultured with DH neurons. Consistent with EPSC recordings, an electrical field stimulation (20 Hz, 20 s) induced a notable increase in the Spy‐pHluorin signal in either 0Ca or 2Ca bath solution (Fig 2A–C; Movies EV1 and EV2). Similar to electrophysiological results during 10‐Hz stimulation (Appendix Fig S4C), the peak amplitude of ΔF/F₀ evoked in 0Ca solution was 77% of that evoked in 2Ca solution, further confirming the dominant contribution of CiVDS during sustained neural activity (Fig 2C). However, the same electrical stimulation triggered Spy‐pHluorin exocytosis in hippocampal nerve terminals only in 2Ca solution, but not 0Ca solution (Fig 2D–F; Movies EV3 and EV4). Thus, CiVDS mediates synaptic transmission in co‐cultured DRG and DH neurons.

Figure 2. Imaging of CiVDS‐mediated synaptic transmission in co‐cultured DRG and DH neurons.

• A
  A representative photograph showing the co‐cultured DRG and DH neurons. The DRG neurons were expressed with Spy‐pHluorin for imaging the synaptic transmission. Scale bar, 10 μm.
• B
  Images of a presynaptic bouton marked in (A) showing the Spy‐pHluorin fluorescence at 20 s before (−20 s, pre‐stimulus), 20, 40, and 120 s after electrical stimulation (post‐stimulus) in 0Ca²⁺ (upper panel) or 2Ca²⁺ (lower panel) solution.
• C
  Averaged fluorescence changes (ΔF/F₀) of Spy‐pHluorin in 0Ca²⁺ (left) or 2Ca²⁺ solution (right) in response to the same electrical stimulation (20 Hz, 20 s) (n = 45 puncta from six cells for 0Ca²⁺ and 57 puncta from six cells for 2Ca²⁺). The shadow in the traces represents the error bars (s.e.m) of each point.
• D–F
  The same as in (A–C), but the experiments were performed in cultured hippocampal neurons (n = 70 puncta from three cells for 0Ca²⁺ and 72 puncta from three cells for 2Ca²⁺).

EPSC_CiVDS is mediated by glutamate release from presynaptic DRG neurons

To determine whether the EPSC_CiVDS is mediated by presynaptic glutamate release, we tested the effects of ionic glutamate receptor blockers (D‐AP5 for NMDA receptor and CNQX for AMPA receptor; Trussell et al, ¹⁹⁹³; Wang et al, ²⁰¹⁶). Three minutes of exposure to AP5 (50 μM) and CNQX (10 μM) diminished the EPSCs in both the 0Ca and 2Ca solutions (Fig 3A and B). As a control, there was only minimal run‐down during repeated stimulation of EPSC_CiVDS (Appendix Fig S5A and B). Furthermore, 3‐min treatment with 100 μM cyclothiazide (CTZ), a blocker of AMPA receptor desensitization (Fucile et al, ²⁰⁰⁶), greatly slowed the decay and increased the charge of the EPSC in both the 0Ca and 2Ca solutions (Fig 3C and D). Thus, the EPSC_CiVDS is mediated by synaptic glutamate transmission in co‐cultured DRG and DH neurons.

Figure 3. EPSC_CiVDS is mediated by glutamate transmission.

• A
  Representative evoked EPSCs in DH neurons co‐cultured with DRG neurons before (black) and after (red) applying 50 μM AP5 and 10 μM CNQX in Ca²⁺‐free (left) or 2.5 mM Ca²⁺‐containing solution (right). The EPSCs of 0Ca²⁺ and 2Ca²⁺ were recorded in two different DH neurons.
• B
  Quantification of EPSC amplitudes as in (A) (n = 15 cells for 0Ca²⁺ and 10 cells for 2Ca²⁺).
• C
  Evoked EPSCs recorded in DH neurons co‐cultured with DRG neurons before (black) and after (red) applying 100 μM cyclothiazide (CTZ) in 0Ca²⁺ (left) or 2Ca²⁺ solution (right). The traces are fitted with a single exponential curve (blue). The EPSCs of 0Ca²⁺ and 2Ca²⁺ were recorded in two different DH neurons.
• D
  Quantification of the decay time (τ) and charge as in (C) (n = 10 cells for 0Ca²⁺ and 12 cells for 2Ca²⁺). EPSCs were evoked by local electrical stimulation (Estim) at arrows.

Data information: Data are shown as the mean ± s.e.m.; paired Student's t‐test; *P < 0.05, **P < 0.01, ***P < 0.001.

To determine whether DRG neurons are responsible for the presynaptic glutamate release in EPSC_CiVDS, we performed paired whole‐cell recordings to examine the specific synaptic transmission between the patched presynaptic DRG (current‐clamp) and postsynaptic DH neurons (voltage‐clamp). EPSC traces were recorded in the standard 2Ca bath within 1 min and 0Ca bath 5 min after whole‐cell dialysis, in which 10 mM BAPTA was whole‐cell dialyzed into the patched DRG neuron (Fig 4A) to ensure the intracellular Ca²⁺‐free condition. A single pulse of current injection (5 ms, 1,000 pA) after break‐in triggered an action potential in the presynaptic DRG neuron, followed by an EPSC in the postsynaptic DH neuron in 2Ca bath solution (Fig 4B). Strikingly, a notable EPSC signal (~40%) remained 5 min after whole‐cell dialysis even though the patched neurons were bathed in 1 mM EGTA‐containing Ca²⁺‐free solution (Fig 4B). In contrast, similar Ca²⁺‐free treatment completely blocked the EPSCs in hippocampal neurons (Fig 4C and D). Thus, CiVDS‐mediated transmitter release from the presynaptic DRG neurons contributes to EPSC signals in the postsynaptic DH neurons.

Figure 4. Presynaptic DRG neurons elicit EPSC_CiVDS in postsynaptic DH neurons.

• A
  Setup for paired patch‐clamp recording of dorsal root ganglion (DRG) and dorsal horn (DH) neurons in co‐culture. The presynaptic DRG neuron was whole‐cell dialyzed with 10 mM BAPTA under current‐clamp mode, while the postsynaptic DH neuron was in voltage‐clamp mode.
• B
  Left, representative dual recordings of presynaptic action potentials (upper) and postsynaptic EPSCs (lower) following current‐step injection (1,000 pA, 5 ms) in a DRG neuron dialyzed with 10 mM BAPTA. Following whole‐cell break‐in and intracellular BAPTA dialysis, double recordings were performed at 0 min/2.5 mM Ca²⁺ bath (green), and 5 min/0Ca²⁺ bath (black), Right, statistics of EPSC amplitudes (n = 9 cells).
• C, D
  As in (A) and (B), except that EPSCs were recorded from cultured hippocampal neurons (n = 11 cells).
• E
  Image showing a DRG neuron infected by ChR2‐mCherry AAV2/9 virus and a co‐cultured DH neuron.
• F
  Upper, diagram showing the protocol for infection of DRG neurons by ChR2‐mCherry virus and co‐culture with DH neurons; lower, cartoon of EPSC recording from DH neurons co‐cultured with ChR2‐expressing DRG neurons.
• G
  Left, typical EPSCs from DH neurons induced by light stimulation (475 nm, 5 ms, at arrows) of co‐cultured ChR2‐expressing DRG neurons in Ca²⁺‐free or 2.5 mM Ca²⁺‐containing solution before (black) and after (red) exposure to 50 μM AP5 and 10 μM CNQX. The EPSCs of 0Ca²⁺ and 2Ca²⁺ were recorded in two different DH neurons; right, statistics of EPSC amplitude (n = 5 cells for 0Ca²⁺ and 14 for 2Ca²⁺).

Data information: Data are shown as the mean ± s.e.m.; Wilcoxon test for (B); paired Student's t‐test for (D) and (G); **P < 0.01, ***P < 0.001. Scale bars, 20 μm.

We next used an optogenetic approach to specifically activate DRG neurons by expressing channelrhodopsin‐2 (ChR2) in DRG neurons with AAV2/9 virus before the co‐culture with DH neurons (Fig 4E and F). The transient application of 475‐nm light evoked action potentials in ChR2‐expressing DRG neurons (Appendix Fig S6). Importantly, the light activation of DRG neurons also induced EPSCs in DH neurons bathed in either 0Ca or 2Ca solution, and these were fully abolished by the blockade of ionic glutamate receptors with D‐AP5 and CNQX (Fig 4G). Together, these findings demonstrate that an action potential per se is able to directly trigger presynaptic glutamate release independent of Ca²⁺ influx during the synaptic transmission from presynaptic DRG terminals to postsynaptic central DH neurons.

EPSC_CiVDS is mediated by SNARE complex and N‐type Ca²⁺‐channels

To determine whether the EPSC_CiVDS is mediated by SNARE‐dependent vesicular exocytosis, we performed DRG neuron‐restricted knockout of synaptobrevin 2 (Syb2/VAMP2) by infecting DRG neurons from homozygous floxed Syb2‐null mice with Cre recombinase‐carrying AAV2/5 virus (Fig 5A). Western blots showed the complete loss of Syb2 in DRG neurons infected with Cre virus (Fig 5B). Strikingly, the EPSC_CiVDS in DRG–DH transmission was substantially reduced by the Syb2‐knockout (Fig 5C). Thus, similar to CiVDS in somatic secretion (Chai et al, ²⁰¹⁷), CiVDS‐mediated presynaptic glutamate release occurs through SNARE‐dependent vesicular exocytosis.

Figure 5. EPSC_CiVDS are mediated by the SNARE complex and N‐type Ca²⁺‐channels.

• A
  Left, diagram showing the protocol for EPSC recording from DH neurons co‐cultured with Syb2‐knockout (cKO) DRG neurons; right, cartoon of EPSC recording from DH neurons co‐cultured with Syb2‐cKO DRG neurons.
• B
  Representative western blots (upper) and analysis (lower) of the expression levels of Syb2 in control (Ctrl) and Syb2‐cKO DRG neurons (n = 3 per group). Control DRG neurons are from homozygous floxed Syb2‐null mice infected with AAV5‐CAG‐EGFP virus.
• C
  Evoked EPSCs and statistics from DH neurons co‐cultured with control (Ctrl) or Syb2‐cKO DRG cells (cKO) in 0Ca²⁺ solution (n = 13 cells for Ctrl and 11 for cKO).
• D
  Evoked EPSCs and statistics from DH neurons co‐cultured with DRG neurons before (black) and after (red) applying 1 μM ω‐conotoxin‐GVIA (GVIA) in 0Ca²⁺ bath (n = 9 cells).
• E
  Upper, diagram showing the protocol for EPSC recordings from DH neurons co‐cultured with Cav2.2 knockdown (KD) DRG neurons; lower, cartoon of EPSC recording from DH neurons co‐cultured with Cav2.2‐KD DRG neurons.
• F
  Left panels, evoked EPSCs from DH neurons co‐cultured with scrambled shRNA control (Ctrl) or Cav2.2‐KD (sh‐1/sh‐2) DRG neurons in 0Ca²⁺ solution. Right panel, quantification of evoked EPSCs (n = 17 cells for Ctrl, 12 for sh‐1, and 19 for sh‐2). EPSCs were evoked by local electrical stimulation (Estim, at arrows).

Data information: Data are shown as the mean ± s.e.m.; paired Student's t‐test for (B) and (D), unpaired Student's t‐test for (C), and Kruskal–Wallis test followed by Dunn's multiple comparisons test for (F); *P < 0.05, **P < 0.01, ***P < 0.001.

Source data are available online for this figure.

We have shown that the N‐type Ca²⁺ channel Cav2.2 serves as a voltage sensor for CiVDS in the somata of DRG and superior cervical ganglion neurons, while the L‐type Ca²⁺ channel mediates CiVDS in adrenal slice chromaffin cells (Chai et al, ²⁰¹⁷; Huang et al, ²⁰¹⁹). To determine the voltage sensor for CiVDS in synaptic transmission of DRG–DH neurons, we blocked Cav2.2 with ω‐conotoxin‐GVIA (GVIA, 1 μM), which directly blocks the voltage‐sensing ability of Cav2.2 (Ellinor et al, ¹⁹⁹⁴; Yarotskyy & Elmslie, ^{2009, 2010}; Chai et al, ²⁰¹⁷), and found that CiVDS‐mediated EPSCs were remarkably decreased (Fig 5D). On the contrary, CiVDS‐mediated EPSCs were insensitive to Cd²⁺, a well‐known pore blocker of VGCCs (Chow, 1991; Tang et al, ²⁰¹⁴; Huang et al, ²⁰¹⁹; Appendix Fig S7A and B), confirming that the gating charge but not pore permeability is crucial for EPSC_CiVDS. To further confirm the essential role of Cav2.2 in EPSC_CiVDS, we adopted an shRNA‐based knockdown (KD) approach. As validated previously (Chai et al, ²⁰¹⁷), two independent Cav2.2‐targeting shRNAs (sh‐1 and sh‐2) were delivered into DRG neurons by the AAV2/5 virus before co‐culture with DH neurons (Fig 5E). Both the CiVDS‐ and CDS‐mediated EPSCs in DH neurons were substantially decreased by Cav2.2‐KD in presynaptic DRG neurons compared with those in control group (Fig 5F and Appendix Fig S8A and B). Together, these findings indicate that Cav2.2 serves as a voltage sensor for CiVDS‐mediated glutamate release during the synaptic transmission from DRG to DH neurons.

Discussion

The classic view of neurotransmission is that impulse‐induced presynaptic release (both somatic and terminal) occurs exclusively via CDS (Katz & Miledi, 1967; Neher & Zucker, ¹⁹⁹³; Geppert et al, ¹⁹⁹⁴; Felmy et al, ²⁰⁰³; Meinrenken et al, ²⁰⁰³; Jackson & Chapman, ²⁰⁰⁶; Sudhof, ²⁰¹²). In somatic secretion, the CDS‐only concept has recently been revised by a series of studies on CiVDS in peripheral sensory and sympathetic neurons (Zhang & Zhou, 2002; Zhang et al, ²⁰⁰⁴; Zheng et al, ²⁰⁰⁹; Liu et al, ²⁰¹¹; Huang et al, ²⁰¹⁹; Moya‐Diaz et al, ²⁰¹⁹; Sforna et al, ²⁰¹⁹), including its functions and mechanisms (Chai et al, ²⁰¹⁷). In the present study, we show that CiVDS contributes to synaptic transmission in the central nervous system: at synapses between presynaptic DRG terminals and postsynaptic central DH neurons. The CiVDS‐mediated EPSCs have a faster recycling rate and less short‐term depression than CDS‐mediated EPSCs. Similar to somatic CiVDS, the machinery governing CiVDS‐mediated synaptic transmission includes Cav2.2 (voltage sensor) and Syb2‐SNARE (fusion pore) (Fig 6).

Figure 6. Model of the CiVDS‐mediated EPSC in synaptic transmission.

In Ca²⁺‐free bath, an action potential activates the voltage‐gated Ca²⁺ channel (VGCC/Cav2.2) and triggers presynaptic glutamate release through Ca²⁺‐independent but voltage‐dependent secretion (CiVDS) and activates an excitatory postsynaptic current (EPSC_CiVDS). In physiological solution containing Ca²⁺, however, both CiVDS‐ and Ca²⁺‐dependent secretion (CDS)‐mediated glutamate release contribute to a larger evoked postsynaptic EPSC (not shown).

The major finding of the present work is that CiVDS contributes to the central synaptic transmission between co‐cultured presynaptic DRG and postsynaptic DH neurons. This is supported by the following evidence: (i) EPSCs were recorded in DRG–DH (Fig 1A and C) but not hippocampal neurons (Fig 1B) in 0Ca solution; (ii) field stimulation triggered a [Ca²⁺]_i rise only in Ca²⁺‐containing solution (Fig 1A and Appendix Fig S1A–C); (iii) both spontaneous (sEPSC) and miniature EPSC (mEPSC) could be recorded in 0Ca solution (Appendix Fig S3A and B); (iv) in 2Ca solution, BAPTA reduced the EPSC amplitude to that in 0Ca (Fig 1C and D); (v) the evoked EPSC_CiVDS were abolished by antagonists against AMPA receptors (CNQX) and NMDA receptors (AP5) (Trussell et al, ¹⁹⁹³; Wang et al, ²⁰¹⁶; Fig 3A and B); (vi) the decay of evoked EPSC_CiVDS was greatly slowed by the AMPA receptor enhancer CTZ (Fucile et al, ²⁰⁰⁶; Fig 3C and D); (vii) Spy‐pHluorin live‐imaging confirmed CiVDS from the presynaptic terminals of DRG neurons but not hippocampal neurons (Fig 2; Movies [Link], [Link]); (viii) paired patch‐clamp recordings revealed that EPSCs between DRG and DH neurons were independent of extracellular and intracellular Ca²⁺, while the EPSCs in hippocampal neurons were completely blocked in the 0Ca condition (Fig 4A–D); (ix) optogenetic stimulation of DRG neurons triggered EPSCs in either 0Ca or 2Ca solution, confirming that presynaptic glutamate release from DRG neurons is responsible for both the EPSC_CiVDS and EPSC_CDS recorded in DH neurons (Fig 4E–G); (x) SNARE is the CiVDS fusion machinery of the EPSCs, which were attenuated by knockout of the SNARE protein Syb2 in DRG neurons (Fig 5A and C); and (xi) Cav2.2 is the voltage sensor of EPSC_CiVDS, as EPSC_CiVDS was attenuated by either RNAi knockdown or an antagonist GVIA, which inhibits the voltage sensor of Cav2.2 (Ellinor et al, ¹⁹⁹⁴; Yarotskyy & Elmslie, ^{2009, 2010}; Chai et al, ²⁰¹⁷), but was insensitive to Ca²⁺ channel pore blocker Cd²⁺ (Chow, 1991; Tang et al, ²⁰¹⁴; Huang et al, ²⁰¹⁹; Fig 5D and F; Appendix Fig S7A and B). Together, CiVDS contributes to the EPSCs between DRG and DH neurons in physiological solution and in response to physiological stimulation.

The second finding is that both CiVDS and CDS make essential and complementary contributions to the total EPSC (CDS + CiVDS) of DRG–DH synaptic transmission under physiological conditions: 2Ca bath solution and ≤ 20 Hz stimulation (Xu et al, ²⁰⁰⁰; Fang et al, ²⁰⁰⁵; Zheng et al, ²⁰⁰⁹). This is supported by the following evidence: (i) compared with the EPSC_CDS, the EPSC_CiVDS had a faster vesicle recycling rate (Fig 1E and F) and less short‐term depression (Fig 1G and H); (ii) EPSC_CDS and EPSC_CiVDS contributed similarly (~50% each) to the total EPSC during synaptic transmission following single‐pulse/low‐frequency stimulation (≤ 0.2 Hz) (Fig 1D and F); (iii) CiVDS became dominant (~87%) in the total EPSC following “painful” high‐frequency stimulation (10–20 Hz) (Xu et al, ²⁰⁰⁰; Fang et al, ²⁰⁰⁵; Zheng et al, ²⁰⁰⁹), as CDS was attenuated more than CiVDS in the total EPSC (Fig 1G and I) due to much slower vesicle recycling of the EPSC_CDS (Neher & Sakaba, 2008; Wang et al, ²⁰⁰⁸; Zhang et al, ²⁰¹¹); (iv) single impulses or low frequency (≤ 0.2 Hz) triggered both CDS and CiVDS in synaptic release (Fig 1), while only CiVDS (Zhang & Zhou, 2002; Zheng et al, ²⁰⁰⁹) but not CDS (Zheng et al, ²⁰⁰⁹; Liu et al, ²⁰¹¹; Huang et al, ²⁰¹⁹) was triggered during somatic secretion; and (v) a “painful” burst of pulses (10–20 Hz) (Xu et al, ²⁰⁰⁰; Fang et al, ²⁰⁰⁵; Zheng et al, ²⁰⁰⁹) triggered both CiVDS and CDS in synaptic transmission (Fig 1), as well as somatic exocytosis (Zhang & Zhou, 2002; Zheng et al, ²⁰⁰⁹). Together, CiVDS and CDS make essential and complementary contributions to both synaptic EPSCs (Fig 1) and somatic transmitter release (Chai et al, ²⁰¹⁷) in the DRG–DH sensory system. However, one limitation of the present study is that we seperated EPSC_CDS from EPSC (2Ca) based on the assumption that EPSC_CDS and EPSC_CiVDS is additive, which has been validated in somatic secretion (Liu et al, ²⁰¹¹). Thus, EPSC (2Ca) = EPSC_CiVDS + EPSC_CDS. Because EPSC_CiVDS = EPSC (0Ca), so EPSC_CDS = EPSC (2Ca) −EPSC (0Ca).

Regarding the inconsistence for the existence of CiVDS at the synapses between DRG and spinal cord neurons in previous publication (Zheng et al, ²⁰⁰⁹), the improvement of the present study is that we added a few neurotrophic factors into the culture medium for CiVDS recording, including 10 ng/ml NGF, 50 ng/ml BDNF, and 50 ng/ml GDNF. Combined neurotrophic factors substantially increased the CiVDS‐ and CDS‐mediated EPSC compared to that without neurotrophic factors (Appendix Fig S9A–D). Consistently, combined neurotrophic factors restored DRG neuronal somatic CiVDS, which disappears after 3 days in culture (Chai et al, ²⁰¹⁷), and maintained the CiVDS up to 2 weeks (Appendix Fig S10A and B). Importantly, the combined neurotrophic factors significantly increased the expression level and functional current of Cav2.2 (Appendix Figs S11A and B and S12A and B). Thus, the expression level of Cav2.2 seems to be critical for CiVDS. Consistent with this hypothesis, the expression of Cav2.2 is much lower in hippocampal neurons than that in cultured DRG neurons (Appendix Fig S13A and B). Nonetheless, whether other effectors are also involved in EPSC_CiVDS regulation deserves systematic investigation in future.

The present study reported CiVDS‐mediated synaptic transmission in a central synapse between DRG and DH neurons. In contrast, release from the presynaptic terminal of the calyx of Held is abolished after blocking Ca²⁺ influx (Felmy et al, ²⁰⁰³). Future work is needed to determine: (i) whether EPSC_CiVDS occurs in the synaptic transmission of other neural circuits; (ii) whether IPSC_CiVDS exists; (iii) whether EPSC_CiVDS occurs in brain slices and in vivo; and (iv) the physiological relevance of EPSC_CiVDS.

In summary, the present work extends the occurrence of CiVDS from peripheral neuronal somata and neuroendocrine cells to a central presynaptic terminal and demonstrates EPSC_CiVDS in a rat/mouse model of cultured DRG–DH neurons, implying potential physiological roles of CiVDS in synaptic transmission in other mammalian neural circuits.

Materials and Methods

Animals

The use and care of animals was approved and directed by the Institutional Animal Care and Use Committee of Peking University and the Association for Assessment and Accreditation of Laboratory Animal Care. We used CRISPR/Cas9‐mediated one‐step genomic engineering to generate Synaptobrevin‐2 floxed mice (Yang et al, ²⁰¹³). A mixture of guide RNAs (CTCTGGTGATAGGCGGATCCAGG, AGGGTTCCTAGACGAACACCAGG), the corresponding donor DNA, and the Cas9 protein was micro‐injected into the fertilized eggs, followed by embryonic transplantation. Two LoxP sites were inserted into the 5′ intron of exon 3 and 3′ intron of exon 4 in mouse synaptobrevin‐2 gene donor fragment (Gene ID: 22318). All animals were housed in an animal facility under a 12‐h light/dark cycle at 22 ± 2°C with food and water available ad libitum. Sprague–Dawley rats were used for all experiments except that floxed Syb2‐null mice were used for Syb2‐KO experiments.

Cell culture

Three‐ to four‐day‐old postnatal Sprague–Dawley rats or floxed Syb2‐null mice were used for the culture of dorsal root ganglion (DRG) neurons, and 15–16‐day embryos of Sprague–Dawley rats were used for the culture of dorsal horn (DH) neurons. Postnatal day 0 Sprague–Dawley rats were used for the culture of hippocampal neurons.

Dorsal root ganglion isolation and neuronal culture were performed as previously described (Chai et al, ²⁰¹⁷). DRG ganglia were dissected from the spine and placed in cold L15 medium (Gibco). After removing attached tissue, the ganglia were cut into several pieces and incubated in Dulbecco's modified Eagle's medium (DMEM)/F12 (Gibco) containing trypsin (0.2–0.3 mg/ml) and collagenase (1 mg/ml) for 40 min at 37°C under 5% CO₂. After that, the pieces were washed twice with 2 ml DMEM/F12 and dissociated into single cells by 8–10 bouts of trituration. For confocal imaging, cells were collected and transfected with Synaptophysin (Spy)‐pHluorin plasmid (a kind gift from Dr. G. Miesenböck, University of Oxford) by using the NeonTM (100‐μl) electroporation system (MPK10096, Invitrogen). Then, the cell suspension was placed on 0.1% poly‐L‐lysine pretreated coverslips and maintained in Neurobasal supplemented with 2% B27 and 0.5 mM L‐glutamine (all from Gibco).

Dorsal horn neurons were isolated and cultured as previously described (Zheng et al, ²⁰⁰⁹). The spinal cord was dissected from 15‐ to 16‐day embryos and placed in cold D‐Hanks solution. After removing attached tissue, the dorsal spinal cord was incubated in 0.25% trypsin solution for 15 min at 37°C under 5% CO₂. After that, the dorsal cord was washed twice with 2 ml Neurobasal supplemented with 2% B27 and 0.5 mM L‐glutamine (all from Gibco) and dissociated into single cells by 8–10 bouts of trituration. To co‐culture DRG and DH neurons, the suspension of DH neurons was evenly placed onto pre‐cultured DRG neurons. All cells were maintained in Neurobasal supplemented with 2% B27, 0.5 mM L‐glutamine (all from Gibco), 10 ng/ml nerve growth factor (NGF), 50 ng/ml brain‐derived neurotrophic factor (BDNF), 50 ng/ml glial cell line‐derived neurotrophic factor (GDNF), and 5 μM cytosine arabinoside. The mixture of NGF, BDNF, and GDNF helped to maintain CiVDS for > 2 weeks. The neurons were used between days 10 and 12 after the start of co‐culture.

For hippocampal cultures, hippocampal neurons were prepared as previously described (Wang et al, ²⁰¹⁸). Briefly, hippocampi were dissected from neonatal rats and placed in cold D‐Hanks solution. After removing attached tissue, the hippocampi were digested in 0.25% trypsin for 15 min at 37°C under 5% CO₂. After that, the tissue was washed twice with 2 ml Neurobasal supplemented with 2% B27 and 0.5 mM L‐glutamine (all from Gibco) and dissociated by 8–10 bouts of trituration. Then, the cells were placed on 0.1% poly‐L‐lysine‐coated coverslips and maintained in DMEM (Gibco) supplemented with 10% FBS for 3 h, which was then replaced by Neurobasal supplemented with 2% B27 and 0.5 mM L‐glutamine. Experiments were performed between days 14 and 16 after the start of culture. For confocal imaging, cultures at day 5 were transfected with Spy‐pHluorin plasmid by using a calcium‐phosphate transfection method. Briefly, plasmids in a 250 mM CaCl₂ solution were slowly added to Hank's balanced salt solution (HBSS) and incubated at room temperature for 25 min. The mixture was then added to the culture and incubated for 15 min. The cells were washed with MgCl₂‐containing medium and then maintained in the original medium. Confocal imaging was performed at day 13–14.

Virus infection

For virus infection, the virus was added to the culture solution at the beginning of DRG neuronal culture. After 3 days, the virus‐containing culture medium was fully replaced and then the DH neurons were co‐cultured with the DRG neurons. The adeno‐associated virus vector carrying CAG‐hChR2(H134R)‐mCherry (AAV2/9) was used for optogenetics and CAG‐EGFP (Ctrl, AAV2/5) or CAG‐EGFP‐2A‐Cre (AAV2/5) for Syb2‐knockout. For Cav2.2 knockdown, the nucleotide target sequences GG ACA TTT CTG CAA GCC TTA A (shCav2.2‐1) and GC TAC TTC CGG TCT TCC TTC A (shCav2.2‐2) were integrated into adeno‐associated virus (AAV2/5) to silence the expression of Cav2.2. A random sequence (TTC TCC GAA CGT GTC ACG T) that was predicted to target no genes in human, rat, and mouse cells was chosen as a negative control (Guangzhou RiboBio Co., Ltd). All the viruses were from Shanghai Heyuan Biotech (China).

Electrophysiology

Excitatory postsynaptic currents (EPSCs) were recorded under the whole‐cell configuration using an EPC10/2 amplifier controlled by Patchmaster software (HEKA Elektronik) as described previously (Zheng et al, ²⁰⁰⁹; Zhang et al, ²⁰¹¹). The membrane potential was clamped at −70 mV, and pipette resistance was controlled to ~10 MΩ for EPSC recordings. The standard external bath solution contained (in mM): 150 NaCl, 5 KCl, 2.5 CaCl₂, 1 MgCl₂, 10 H‐HEPES, and 10 D‐glucose, pH 7.4. The Ca²⁺‐free solution for CiVDS recording was the same, except that 2.5 mM Ca²⁺ was replaced by 1 mM EGTA. For EPSC recording, 100 μM picrotoxin was added to the bath to block IPSCs (inhibitory postsynaptic currents). The standard intracellular pipette solution contained (in mM): 145 K‐gluconate, 5 KCl, 4 MgCl₂‐6H₂O, 10 H‐HEPES, 5 EGTA, and 2 QX314, pH 7.2. Standard intracellular solution with CsCl containing (in mM) 153 CsCl, 1 MgCl₂, 10 H‐HEPES, 4 MgATP, 5 QX314, pH 7.2 was used for a stable recording in Syb2‐KO and optogenetic experiments. For EPSC recordings, local electrical field or light stimulation was applied. Field electrical stimulation (Estim) was used to evoke action potentials via a laboratory‐made bipolar microelectrode (150 μm in diameter) connected to an electronic stimulator (Nihon Kohden, SEN‐3201; Zhang et al, ²⁰¹¹).

In paired patch‐clamp recordings, the presynaptic DRG neuron was in whole‐cell current‐clamp mode and the postsynaptic dorsal horn neuron was in whole‐cell voltage‐clamp mode. For the current‐clamp of DRG neurons, the intracellular solution contained (in mM) 135 K‐gluconate (115 when adding 10 mM BAPTA), 5 KCl, 4 MgCl₂, 3 MgATP, 0.3 Na₂GTP, and 10 Na₂‐phosphocreatine, 10 HEPES, pH 7.2. For the voltage‐clamp DH neuron, standard intracellular solution (but CsCl replacing KCl) was used. For light stimulation, 475‐nm light was generated by a laser (VD‐IIIA, Beijing Viasho Technology, China).

Drugs were puffed locally to the cell under recording via RCP‐2B multi‐channels micro perfusion system (INBIO Inc., Wuhan, China), which had a fast exchange time (< 100 ms) with electronic switching between seven solution channels. The perfusion tubing and tip were modified as previously reported (Wu et al, ²⁰⁰⁵). The puffing pipette of 100 μm tip diameter was located approximately 120 μm from the cell. The perfusion rate was ~0.3 ml/min, and bath volume was ~2 ml.

All recordings were made at room temperature (25°C). Igor software (Wavemetrics, Lake Oswego, OR) was used for all offline data analysis. Series conductance and membrane conductance were used to monitor the seal condition during patch‐clamp recordings.

Ca²⁺ imaging

Ca²⁺ imaging was conducted as described previously (Shang et al, ²⁰¹⁶; Huang et al, ²⁰¹⁹). Cytosolic Ca²⁺ was measured with the Ca²⁺ indicator Cal‐520 (21130, AAT Bioquest). Co‐cultured DRG and DH neurons were loaded with 0.5 μM Cal‐520 AM for 30 min at 37°C and then washed three times with 2Ca solution at room temperature. Then, the cells were electrically simulated and imaged on an inverted confocal microscope (LSM710, Zeiss). The fluorescent signal was captured by excitation with a 488‐nm laser and emission at 500–560 nm. Fluorescence intensity values from the images were calculated and analyzed using ImageJ. The cutoff for selection (5 a.u.) was 1.5‐fold that of the full‐width at half‐maximum of the Gaussian distribution of fluorescence noise.

Confocal live‐imaging

Time‐lapse images were captured at 0.5‐s interval through the inverted confocal microscope (LSM710, Zeiss) with a 63× oil‐immersion objective lens (Zeiss). The fluorescent signal was captured by excitation with a 488‐nm laser and emission at 500–560 nm at room temperature bathed in Ca²⁺‐free or standard external solution. Electrical field stimulation was applied via an electrical stimulator (Nihon Kohden, SEN‐3201). Fluorescence changes at individual boutons were monitored over time and calculated as ΔF/F₀. Images were acquired for more than 200 s in total. The first 30 s before stimulation was used to establish a baseline. Data were analyzed offline with ImageJ.

Protein preparation and western blotting

Cells were washed with phosphate‐balanced saline (PBS) and homogenized on ice with lysate buffer (RIPA; C1503, Beijing Applygen Technologies Inc.), 1 mM PMSF, and 2% proteinase inhibitor (539134; Calbiochem). The homogenates were centrifuged at 13,000 rpm for 10 min at 4°C, and the supernatants were collected and boiled in SDS–PAGE buffer. Proteins were electrophoresed and transferred to nitrocellulose filter membranes (Pall Life Sciences). The membranes were blocked for 1 h with PBS containing 0.1% Tween‐20 (v/v) and 5% fat‐free milk (w/v) and then incubated with primary antibodies at 4°C overnight in PBST containing 2% bovine serum albumin. After washing with 0.1% Tween‐20 containing PBS (PBST), they were then incubated with secondary antibodies at room temperature for 1 h. Blots were scanned with an Odyssey infrared imaging system (LI‐COR Biosciences) and quantified with ImageJ (National Institutes of Health, Bethesda, MD). The primary antibodies used were anti‐Syb2 (104202, SYSY) and β‐actin (A5316, Sigma). The secondary antibodies were IRDye 800CW goat anti‐rabbit IgG (LIC‐926‐32211) and IRDye 680CW goat anti‐mouse IgG (LIC‐926‐32220), both from LI‐COR Biosciences.

Statistical analysis

All experiments were performed with controls side‐by‐side and in random order and were replicated at least three times. All data were collected at least every 2 weeks within 3‐month window to maintain the stability of a data set. Sample sizes are consistent with those reported in similar studies. Multiple recordings from one cell with the identical stimulus protocol were considered as technical replications, which were averaged to generate a single biological replication representing value/data from one cell. No samples or recordings that provided successful measurements were excluded from analysis. Data are shown as the mean ± s.e.m. All the data were tested for normality by Shapiro–Wilk test before the statistical analysis. If the data passed the normality test, two‐tailed Student's t‐test was applied for comparison between two unpaired groups and the paired Student's t‐test was used for comparison between two matched groups. If the data did not pass normality, Mann–Whitney test was applied for comparison between two unpaired groups, Wilcoxon matched‐pairs signed‐rank test (Wilcoxon test) was applied for two matched groups, and the Kruskal–Wallis test followed by Dunn's multiple comparisons test was used when multiple groups were compared with one variable. All tests were conducted using Prism V7.0 (GraphPad Software, Inc.) and SPSS 20.0 (Statistical Package for the Social Sciences). Significant differences were accepted at P < 0.05.

Author contributions

Yuan Wang: Data curation; formal analysis; methodology; writing – original draft; writing – review and editing. Rong Huang: Data curation; formal analysis; funding acquisition; methodology; writing – original draft; writing – review and editing. Zuying Chai: Data curation; formal analysis; methodology; writing – original draft; writing – review and editing. Changhe Wang: Data curation; formal analysis; funding acquisition; methodology; writing – original draft; writing – review and editing. Xingyu Du: Data curation; methodology. Yuqi Hang: Data curation; methodology. Yongxin Xu: Methodology. Jie Li: Methodology. Xiaohan Jiang: Methodology. Xi Wu: Methodology. Zhongjun Qiao: Methodology. Yinglin Li: Methodology. Bing Liu: Methodology. Xianying Zhang: Methodology. Peng Cao: Methodology. Feipeng Zhu: Formal analysis; methodology. Zhuan Zhou: Conceptualization; data curation; supervision; funding acquisition; methodology; project administration; writing – review and editing.

Disclosure and competing interests statement

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Movie EV1

Movie EV2

Movie EV3

Movie EV4

Source Data for Figure 5

Source Data for Appendix

Review Process File

EMBO reports (2022) 23: e54507

Data availability

No data were deposited in a public database.