Dual face of axonal inhibitory inputs in the modulation of neuronal excitability in cortical pyramidal neurons.

Limited by the tiny structure of axons, the effects of these axonal hyperpolarizing inputs on neuronal activity have not been directly elucidated. Here, we imitated these processes by simultaneously recording the activities of the somas and proximal axons of cortical pyramidal neurons. We found that spikes and subthreshold potentials propagate between somas and axons with high fidelity. Furthermore, inhibitory inputs on axons have opposite effects on neuronal activity according to their temporal integration with upstream signals. Concurrent with somatic depolarization, inhibitory inputs on axons decrease neuronal excitability and impede spike generation. In addition, following action potentials, inhibitory inputs on an axon increase neuronal spike capacity and improve spike precision. These results indicate that inhibitory inputs on proximal axons have dual regulatory functions in neuronal activity (suppression or facilitation) according to neuronal network patterns.

Introduction

The axon is a long process extending from the cell body of a neuron to the terminal, and is the main structure responsible for neuronal signal conduction. Numerous studies have focused on axonal function (Chen et al., 2010; Debanne et al., 2011; Grubb et al., 2011; Harty et al., 2013; Sasaki, 2013; Gulledge and Bravo, 2016; Whalley, 2016). Axons transmit both spikes and analog signals with high fidelity (Cross and Robertson, 2016). With a high density of sodium channels, the axon initial segment is a common onset point for action potentials (Kole et al., 2008; Kole and Stuart, 2008; Yu et al., 2010; Cross and Robertson, 2016). However, a sustained depolarizing current induces fewer spikes in an axon than in the soma, while in contrast, oscillation waves induce a higher frequency of spikes in the axon (Ge et al., 2011, 2014; Apostolides et al., 2016). Therefore, axons appear to have different responses to varied input patterns.

Additionally, receptors have been found on axons, especially on the axon initial segment and axon terminal (Sasaki et al., 2011; Trussell and Bender, 2012; Wefelmeyer et al., 2015; Gao and Heldt, 2016; Kerti-Szigeti and Nusser, 2016; Yin et al., 2017). Activation of these receptors leads to orthodromic propagation of the neuronal signal to axon terminals, which then influence subsequent neurotransmitter release (Sasaki et al., 2011). Field potentials from axon terminals also antidromically propagate back toward the soma, which can change the threshold for initiating action potentials (Paradiso and Wu, 2009). However, the roles of these receptors on proximal axons are still not clear. Structural studies have indicated that chandelier neurons, a subtype of GABAergic interneuron, have abundant projections that end on the proximal axons of pyramidal neurons in the cortex (Somogyi, 1977; Christie and De Blas, 2003; Taniguchi et al., 2013; Inan and Anderson, 2014).

These inhibitory axo-axonic synapses may have different effects on the output of pyramidal neurons according to the patterns of connections between interneurons and pyramidal neurons. To test this hypothesis, we conducted paired whole-cell recordings on the somas and proximal axons of pyramidal neurons in the cortex. We found that both spikes and subthreshold potentials propagated between somas and axons with high fidelity. Hyperpolarizing potentials on axons had entirely different effects on neuronal outputs depending on their temporal integration with upstream depolarization. These characteristics may play important roles in feedforward and feedback inhibitory circuits.

Materials and Methods

Animals

Twelve specific-pathogen-free male FVB/N mice at postnatal day 15–20, irrespective of sex, were enrolled in this study. This study and all experiments were fully approved by the Institutional Committee of Animal Care Unit in the Beijing Administration Office of Laboratory Animals (approval No. B10831). Surgeries were performed under anesthesia, and all possible efforts were made to minimize suffering.

The study protocol was approved by the Institutional Committee of Animal Care Unit in the Beijing Administration Office of Laboratory Animals (approval No. B10831). The experimental procedure followed the United States National Institutes of Health Guide for the Care and Use of Laboratory Animal (NIH Publication No. 85-23, revised 1986), and “Consensus Author Guidelines on Animal Ethics and Welfare” produced by the International Association for Veterinary Editors (IAVE).

Brain slices

Slices from sensory cortex (300 μm) were prepared from FVB mice. Postnatal day 15–20 mice were anesthetized with an injection of chloral hydrate (300 mg/kg) and decapitated with a guillotine. The cortical slices were cut with a vibratome in modified and oxygenized (95% O2/5% CO2) artificial cerebrospinal fluid (124 mM NaCl, 3 mM KCl, 1.2 mM NaH2PO4, 26 mM NaHCO3, 0.5 mM CaCl2, 5 mM MgSO4, 10 mM dextrose, and 5 mM hydroxyethyl piperazine ethanesulfonic acid; pH 7.35) (Sinopharm Chemical Reagent Co., Ltd., Beijing, China) at 4°C and were then maintained in normal oxygenated artificial cerebrospinal fluid (126 mM NaCl, 2.5 mM KCl, 1.25 mM NaH2PO4, 26 mM NaHCO3, 2.0 mM CaCl2, 2.0 mM MgSO4, and 25 mM glucose; pH 7.35) at 35°C for 1 hour before the experiments. A slice was transferred to a submersion chamber (Warner RC-26G, Molecular Devices, CA, USA) that was perfused with normal artificial cerebrospinal fluid for electrophysiological experiments (Ge et al., 2011, 2014).

Dual recording

We simultaneously recorded from somas and axonal blebs (Hu et al., 2009; Hu and Shu, 2012) of identical pyramidal cells in layers IV–V of cerebral cortex (MultiClapm-700B, Molecular Devices, CA, USA) under a fluorescent/DIC microscope (Nikon FN-E600). We differentiated between xonal blebs and dendritic blebs based on the process diameter and number of branches, as well as the polarity of neurons (Chen et al., 2010; Ge et al., 2014). Neuronal processes with fewer branches and fine diameters were categorized as axons. The electrical signals were input into a patch clamp (pClamp-10, Molecular Devices); the sampling rate was 50 kHz.

We confirmed that recordings were from two sites on the same neuron based on the presence of direct and corresponding electrical signals. Transient capacitance was compensated and the output bandwidth was 3 kHz. The pipette solution contained 150 mM K-gluconate, 5 mM NaCl, 0.4 mM ethylenebis (oxyethylenenitrilo) tetraacetic acid, 4 mM Mg-ATP, 0.5 mM Tris-GTP, 4 mM Na-phosphocreatine, and 5 hydroxyethyl piperazine ethanesulfonic acid (pH 7.4 adjusted by 2 M KOH) (Sinopharm Chemical Reagent Co., Ltd.). The osmolarity of the freshly made pipette solution was 295–305 mOsmol, and the pipette resistance was 10–15 MΩ.

Notably, pyramidal neuron axonal blebs formed from resealing the ends of cut axons during slice preparation. Although this preparation could be judged as an injured axon, several lines of evidence have indicated that axonal blebs are functionally intact. In axonal blebs, the resting membrane potential and action potential values are close to normal (Ge et al., 2011, 2014). Moreover, other immunohistochemical and electrophysiological study has suggested that the functions of axonal blebs are likely normal (Hu et al., 2009).

Capacity and timing precision (spike programming) are represented as the interspike interval and the standard deviation of spike timing, respectively (Guan et al., 2006; Wang et al., 2009). Data were analyzed if the recorded neurons had resting membrane potentials less than –60 mV. Additionally, data were excluded unless the changes in resting membrane potential, spike magnitude, and input resistance throughout each experiment were all less than 5%. Input resistance was monitored by measuring cell responses to hyperpolarization pulses at the same values as the depolarization that evoked spikes.

Statistical analysis

Data are presented as the mean ± SEM. All data were analyzed using SPSS 12.0 software (SPSS, Chicago, IL, USA). The parametric tests (unpaired t-test) were applied when normality (and homogeneity of variance) assumptions were satisfied. Otherwise, the equivalent non-parametric tests were used (Mann-Whitney U test). P values less than 0.05 were considered statistically significant.

Results

Signal transmission between somas and axons

We first determined whether subthreshold potentials or sequential spikes were transmitted from somas to axons with high fidelity. In whole-cell recordings of the soma and axonal bleb (), subthreshold analog current and long-term square pulses were injected into somas. The subthreshold analog current induced a subthreshold potential in the soma, and the same membrane oscillation was recorded in the axon (). The subthreshold analog signal was reliably transmitted from the soma to the axon (both in pattern and magnitude). Then, a long-term depolarizing pulse was injected into the soma to induce sequential spikes (upper panel in ). In axonal recordings, corresponding spikes were recorded at almost the same time (lower panel in ).

Subthreshold potentials and spikes propagate from somas to axons with high fidelity.

(A) Diagram of paired whole-cell recording at the soma (triangle) and the axon (lower black line) of the same pyramidal neuron in the cortex. The axonal recording was carried out on the axon bleb (black dot), which is a structure formed during slice incubation. (B) Subthreshold analog currents were injected into the soma, and corresponding potentials were recorded at the soma (upper panel, the waves come from another in vivo recording). At almost the same time, similar potentials were recorded at the axon (lower panel). (C) Long-term square-wave current was injected into the soma to induce spikes. A regular pattern of spikes was recorded at the soma (upper panel). Meanwhile, spikes with the same pattern were recorded at the axon (lower panel). These results were repeated in all 10 neurons tested.

Critically, subthreshold waves and sequential spikes were also faithfully propagated from the axon to the soma. With the same recording pattern as that in , input pulses were injected into the axonal bleb and somatic responses were recorded (). We observed that subthreshold analog waves induced membrane potential oscillations in axons (lower panel in ). Simultaneously, a similar wave was recorded in the soma (upper panel in ). Sequential spikes were induced by injecting long-term depolarizing pulse into axonal blebs (lower panel in ). Corresponding sequential spikes were recorded simultaneously in the soma (upper panel in ). Interestingly, although the same current was injected into the soma and axons, the spiking patterns were different (compare Figures and ).

Subthreshold potentials and spikes propagate from axons to somas with high fidelity.

(A) Diagram of the same recording setup as in Experiment 1. (B) Subthreshold analog currents were injected into the axon, and corresponding potentials were recorded in the axon (upper panel, the waves come from another in vivo recording). Similar potentials were recorded in the soma at almost the same time. (C) Long-term square-wave current was injected into the axon to induce sequential spikes, which were fired in an irregular pattern (lower panel). Spikes with the same pattern were simultaneously recorded at the soma (upper panel). These results were repeated in all 10 neurons tested.

Axonal hyperpolarization impeded spike generation when the neuron is depolarized

With soma-axon paired recordings (), we investigated the effects of axonal activity on neuronal output. A short depolarizing current was injected into the soma to evoke a spike (upper panel in ), without any axonal input, and the induced spike was transmitted to the axon (lower panel in ). Then, we repeated the same procedure, but added a hyperpolarizing pulse (with the same intensity of the depolarizing current) applied to the axon. We found that for this combination, the pulse given to the soma failed to evoke an action potential (). Thus, when inhibitory potentials arrive at axons at the same time as upstream depolarization, neuronal excitability is reduced.

Colliding with upstream depolarization, axonal hyperpolarization impeded spike generation.

(A) Diagram of paired whole-cell recording at the soma (triangle) and the axon (lower black line) of the same pyramidal neuron in the cortex. (B) A short current (square pulse in middle panel) was injected into the soma to induce a spike without any axonal input (lower panel). The intensity of the somatic input current was adjusted to induce a spike. A spike was recorded both in somatic and axonal recordings. (C) The same depolarizing current in B was injected into the soma. Simultaneously, a hyperpolarizing current (with the same strength of the depolarizing current) was injected into the axon (lower panel). In this case, no spike was generated. These results were repeated in all 6 neurons tested.

Subsequent hyperpolarization increased spike capacity and timing precision

Interspike interval (an index of spike capacity) and the standard deviation of spike timing (an index of spike timing precision) were measured by evoking sequential spikes with depolarizing currents (Wang et al., 2009). illustrates a repeat recording of sequential spikes in the soma. Spike intervals increase during sequential spiking, which represents greater spike capacity. However, spikes also drifted, especially during the latter part, which represents reduced spike precision.

Subsequent hyperpolarization increases spike capacity and improves spike precision on the soma.

(A) Recording of sequential spikes at the soma when long-term depolarizing current was delivered to the soma. (B) Recording of sequential spikes at the soma when depolarizing current was delivered to the soma and hyperpolarizing currents were delivered following each spike. (C) Summary of the standard deviation of spike timing (SDST, which represents spike precision) for spike 1 and spike 2 induced by long-term depolarizing current in A and mixed currents in B. (D) Summary of the interspike intervals (ISI, which represents spike capacity) for ISI1–2 and ISI2–3 induced by LT and HP currents. Data are presented as the mean ± SEM (n = 6, unpaired t-test; *P < 0.05). LT: Long-term depolarization; HP: hyperpolarization.

Then, in addition to initial depolarizing current, short hyperpolarizing currents were injected into the soma following each spike. The result was a greater number of spikes within the same duration and more accurate spike firing (). Therefore, this subsequent hyperpolarization decreased interspike interval and improved spike timing precision. The first interspike interval values for corresponding spikes in these two groups did not differ significantly (unpaired t-test, n = 6, P = 0.37, t = 0.94, df = 10), but the second ones did (unpaired t-test, n1 = 6, n2 = 6, P = 0.02, t = 2.83, df = 10). Similarly, the first standard deviation of spike-timing values for corresponding spikes did not differ significantly between these two groups (unpaired t-test, n1 = 6, n2 = 6, P = 0.45, t = 0.79, df = 10), but the second ones did (unpaired t-test, n = 6, P = 0.01, t = 3.01, df = 10).

Subsequent hyperpolarization significantly increased spike capacity and spike precision in the axon. The same long-term depolarizing pulse as depicted in failed to induce sequential spikes in the axon (), and the later spikes drifted across a wide range. These results indicate that axons have poor firing rates and low accuracy following long-term depolarizing pulses.

Subsequent hyperpolarization significantly increases spike capacity and improves spike precision on the axon.

(A) Recording of sequential spikes at the soma when long-term depolarizing current was delivered at the axon. (B) Recording of sequential spikes at the axon when hyperpolarizing currents were also delivered following each spike. (C) Summary of the standard deviation of spike timing (SDST, which represents spike precision) for spike 1 and spike 2 induced by LT and HP currents at the axon (n1 = 6, n2 = 7, unpaired t-test for spike1 and Mann-Whitney U test for spike2). (D) Summary of the interspike interval (ISI) for ISI1–2 and ISI2–3 induced by LT and HP currents at the axon. Data are presented as the mean ± SEM (n1 = 6, n2 = 7, unpaired t-test for ISI1–2 and Mann-Whitney U test for ISI2-3; *P < 0.05). LT: Long-term depolarization; HP: hyperpolarization.

shows that subsequent hyperpolarization following each spike markedly increased axonal spike capacity and spike precision. The interspike interval values for spike1 differed significantly between these two groups (unpaired t-test, n1 = 6, n2 = 7, P = 0.02, t = 2.81, df = 9), as did the values for spike2 (Mann-Whitney U test, P = 0.03, u = 3). Although the first standard deviation of spike-timing values for corresponding spikes did not differ significantly between these two groups (unpaired t-test, P = 0.70, t = 0.40, df = 11), the second ones did (Mann-Whitney U test, P = 0.04, u = 7).

Discussion

Numerous studies on the subcellular components of neurons have gradually identified elaborate structures and functions (Grubb and Burrone, 2010; Sheffield et al., 2011; Dugladze et al., 2012; Trussell and Bender, 2012). A growing number of studies have found that axons are not limited to propagating information downstream. Synaptic structures, which are mostly inhibitory, have been found in axons (Inan and Anderson, 2014; Kerti-Szigeti and Nusser, 2016; Kubota et al., 2016). In addition to dendritic input, synaptic input directly onto axons likely has a significant influence on neuronal output. The present study was designed to determine the effects of these axo-axonic inhibitory synapses on neuronal output. Regarding propagation, we found that action potentials and analog waves are faithfully propagated on axons (both orthodromically and antidromically). The most interesting finding was that inhibitory input onto axons has opposite effects depending on how it is temporally integrated with upstream signals. When they coincide with upstream depolarization, inhibitory inputs on axons might impede the generation of action potentials. Following spikes, inhibitory inputs on axons can increase neuronal spike capacity and spike precision.

Subcellular propagation of subthreshold potentials and spikes

Subthreshold potentials can propagate significant distances along axons. Modest upstream changes propagate orthodromically toward the axon terminal to modulate neurotransmission (Cox et al., 2000; Shu et al., 2006). Additionally, the small potentials in nerve terminals back-propagate up the axon to influence spike generation (Paradiso and Wu, 2009). We observed both orthodromic and antidromic propagation of subthreshold voltage between somas and axons in our research. Furthermore, sequential spikes induced by long-term square pulses were also faithfully propagated between somas and axons. However, the sequential spikes induced at the soma have different patterns than those at axons. Compared with spikes induced on somas, spikes induced on axons are less frequent and more irregular. As mentioned in our previous study, sodium channels in axons are likely inactive and difficult to reactivate during sustained depolarization (Ge et al., 2014). Because of this, sustained depolarization on axons failed to induce regular sequential spikes.

Sources of axonal input and the effects on neuronal excitability

Receptors have been found along axons, especially in the proximal segment and the terminal end. Receptors on axonal terminals modulate local membrane excitability and neurotransmitter release (Starke et al., 1989; Herrero et al., 1992; Smirnova et al., 1993; Ohura and Kamiya, 2016). Spraying glutamate on an axon increases the width of an axon spike (Sasaki et al., 2011). Additionally, synapses have been found on axons, especially on the axon initial segment. These axo-axonic synapse are always GABAergic and innervated by chandelier neurons (Somogyi, 1977; Xia et al., 2014). Whether these GABAergic synapses on the axon initial segment have inhibitory or excitatory functions remains controversial (Szabadics et al., 2006; Woodruff and Yuste, 2008; Xia et al., 2014; Spampanato et al., 2016; Saha et al., 2017a, b).

Chandelier neurons are abundant in the cortex and form feedforward and feedback circuits with pyramidal neurons (Jang et al., 2005; Woodruff and Yuste, 2008; Inan and Anderson, 2014; Wang et al., 2016) (). In the feedforward circuit, chandelier neurons are excited simultaneously with pyramidal neurons. At the same time, pyramidal neurons receive excitatory inputs via dendrites/soma and inhibitory inputs via the axon initial segment. Inhibitory inputs on the axon initial segment decrease the excitability of pyramidal neurons. This hyperpolarization decreases total injected current. Whether this hyperpolarized current changes the spiking threshold needs to be verified in future studies. In the feedback circuit, pyramidal neurons excite chandelier neurons. Consequently, pyramidal neurons receive inhibitory inputs from chandelier neurons on the axon initial segment. As we observed here, this subsequent hyperpolarization increases spike capacity and spike precision of pyramidal neurons. With these feedforward and feedback models, we found that inhibitory inputs on the axon initial segment have significantly different effects on neuronal excitability. Concurrent inhibitory inputs on the axon initial segment offset upstream excitatory inputs and impede spike generation. Spikes following inhibitory inputs promote subsequent spike generation. Previous work has shown that hyperpolarization after spiking can accelerate the reactivation of sodium channels (Chen et al., 2006).

Diagrams of the sources for pyramidal axonal hyperpolarizing input.

(A) Feedforward inhibitory circuit formed by a pyramidal neuron and an interneuron (gray oval), which both receive upstream synaptic inputs. The axon of the interneuron projects to the proximal axons of several pyramidal neurons. Several synapses form on each pyramidal axon. In this circuit, the interneuron is excited at the same time as the pyramidal neurons (by the horizontal axon at the top). Thus, pyramidal neurons simultaneously receive upstream excitatory inputs and axonal inhibitory inputs. (B) Feedback inhibitory circuit formed by the pyramidal neuron and the downstream interneuron. The interneuron has back projections to the pyramidal neuron. In this circuit, the interneuron can be excited by pyramidal output and induce hyperpolarizing currents on the axon of the pyramidal neuron.

Coordinated communication between neurons is essential for functional neuronal networks (Zhao et al., 2012; Zhao and Wang, 2014; Maris et al., 2016). Conversely, network activity modulates both the structure and activity of single neurons during development and regeneration. During neuronal regeneration, recovering of neural activity and signal propagation are important. This study provides an efficient method and informative results for the measurement of axonal function.

Conclusion

With paired recordings at somas and axons, we directly observed the effects of axonal inhibitory inputs on neuronal excitability. Varied temporal integration with upstream waves and inhibitory inputs on axons regulate neuronal excitability in different ways. These adjustments enrich the activities of neuronal networks.

Additional file: .

Open peer review report 1 on “The dual face of axonal inhibitory inputs in the modulation of neuronal excitability in cortical pyramidal neurons”.

Click here for additional data file.

Additional file: .

Open peer review report 2 on “The dual face of axonal inhibitory inputs in the modulation of neuronal excitability in cortical pyramidal neurons”.

Click here for additional data file.