Segregation of Axonal and Somatic Activity During Fast Network Oscillations

See allHide authors and affiliations

Science  15 Jun 2012:
Vol. 336, Issue 6087, pp. 1458-1461
DOI: 10.1126/science.1222017


In central neurons, information flows from the dendritic surface toward the axon terminals. We found that during in vitro gamma oscillations, ectopic action potentials are generated at high frequency in the distal axon of pyramidal cells (PCs) but do not invade the soma. At the same time, axo-axonic cells (AACs) discharged at a high rate and tonically inhibited the axon initial segment, which can be instrumental in preventing ectopic action potential back-propagation. We found that activation of a single AAC substantially lowered soma invasion by antidromic action potential in postsynaptic PCs. In contrast, activation of soma-inhibiting basket cells had no significant impact. These results demonstrate that AACs can separate axonal from somatic activity and maintain the functional polarization of cortical PCs during network oscillations.

In response to synaptic inputs, action potentials (APs) are generated at the axon initial segment (AIS) and propagate along the axon to provide an output signal (1, 2). However, APs can also be initiated in the distal axon under certain conditions (37), but it is unknown how back-propagation of such ectopic APs (EAPs) to the somatodendritic compartment is controlled.

We performed patch-clamp recordings from the soma or axon of hippocampal CA3 pyramidal cells (PCs) during fast network oscillations. In parallel, we monitored the local field potential in the stratum pyramidale (Fig. 1 and fig. S1). During gamma-frequency activity, CA3 PCs discharged phase-locked with the oscillations, but only at a low frequency (810) of 3.5 ± 0.5 Hz (Fig. 1A). Unexpectedly, the frequency of action currents (ACs) in axons recorded >600 µm from the soma was higher by a factor of 4 to 5, with a mean frequency of 16.1 ± 1.4 Hz (Fig. 1, B and C). These results indicate that in the distal axon of PCs, EAPs are generated at high frequencies during gamma oscillations; however, most of these APs do not reach the somatodendritic compartment. To directly demonstrate that axonal spikes are ectopically generated and fail to invade the soma, we performed dual somatic and axonal cell-attached recordings from individual cells (Fig. 1, D to F). We again found a low discharge frequency in the soma (2.7 ± 0.5 Hz) but a considerably higher frequency in the axon (15.8 ± 0.7 Hz; n = 9 cells) (Fig. 1, E and F).

Fig. 1

High-frequency discharge of the axon, but not the soma, of hippocampal CA3 pyramidal cells during gamma-frequency oscillations in vitro. (A and B) Schematic representation of the recording configuration (top). Somatic (A) and axonal (B) cell-attached patch-clamp recordings were obtained from PCs during kainic acid (KA)–induced gamma oscillations in the local field potential (LFP). (C) Summary plot of AC frequency in the soma (8 cells) and axon (12 cells) reveals a significant difference between the two compartments (***P < 0.0001). (D) Scheme of dual somatic and axonal recording configuration. APs evoked in whole-cell configuration by brief depolarizing current injection into the soma (800 pA) (inset) reliably induced ACs in the axon, confirming that recordings are made from two compartments of the same cell. (E) Dual somatic and axonal cell-attached recordings directly demonstrate that high-frequency axonal spikes fail to invade the soma during gamma-frequency oscillations (LFP). (F) Summary plot shows the highly significant difference in the discharge frequency observed at the soma and proximal axon in dual recordings (n = 9 cells) during network oscillations.

The absence of EAP invasion of the soma may reflect strong γ-aminobutyric acid type A (GABAA) receptor (GABAAR)–mediated inhibition in the soma or at the AIS of PCs. We therefore applied GABAAR antagonist during dual axonal and somatic recordings from PCs. Bath application of GABAAR antagonist gabazine considerably increased the probability of the back-propagation of antidromic APs (from 19.3 ± 2.9% to 91.2 ± 1.2%, before and after gabazine application, respectively; n = 3) (fig. S2, A to C). Somatic invasion of APs in the presence of gabazine was maintained during hyperpolarization to –90 mV (92.7 ± 1.8%; n = 3), further indicating that APs are initiated antidromically in the distal axon. These results provide evidence that GABAAR-mediated inhibition effectively controls axosomatic coupling and prevents the back-propagation of EAPs to the soma.

Inhibition to the soma and the AIS is mediated by basket cells (BCs) and axo-axonic cells (AACs), respectively. We next examined the relative contributions of these two sources of inhibition to the blockade of back-propagation. Consistent with previous reports (911), BCs discharged at gamma frequencies (33.6 ± 2.6 Hz; 6 BCs) phase-locked to the oscillation in the field potential (Fig. 2, B and D). In contrast, morphologically identified AACs fired at markedly higher frequencies (mean 105.3 ± 13.3 Hz; n = 4) and did not show correlation with the field oscillations (Fig. 2, A and C). The stark contrast between the behavior of AACs and the rhythmic activity of other inhibitory interneuron types (911) suggests that AACs play a unique role in gamma activity: They do not contribute to phasic inhibition underlying these oscillations but instead provide tonic inhibitory control to the AIS of PCs.

Fig. 2

AACs and BCs are differentially involved in gamma-frequency oscillations. (A and B) Reconstructions of a recorded and biocytin-filled AAC (A) and BC (B) in the CA3 area. The axon (in black) of the AAC is found at the border of the pyramidal layer (Pyr.) and stratum oriens (Or.), where AISs of pyramidal cells are localized. The light micrograph shows axo-axonic cartridges of labeled AAC along the AIS of two unstained PCs. Asterisks mark PC somata. In contrast, the BC shows a broader distribution of the axon in the pyramidal layer and adjacent region of the strata oriens and radiatum. The high-power photomicrograph shows two PC somata surrounded by biocytin-labeled axonal baskets. Inset traces illustrate the characteristic fast-spiking discharge of the interneurons in response to 0.3 nA depolarizing current pulses. (C) Whole-cell recording from an AAC reveals that the cell discharges at a constant high frequency (left, red trace, ic) during KA-induced network gamma oscillations (black trace, LFP). Spectral analysis of the AAC discharge shows a sharp peak at 104 Hz (right, red curve, right y axis; mV2/Hz), separate from the peak in the power spectrum of the extracellular oscillations (30 Hz, black curve, left y axis; µV2/Hz). (D) Corresponding data from a BC. In contrast to the AAC, the BC discharged at gamma frequencies (30 Hz) with clear phase-locking to the oscillations. (E to G) The firing frequencies of the two types are different during network activity (E), although there was no significant difference in the mean frequency (F) and power (G) of LFP between the two sets of recordings (n = 4 and 6, respectively).

To better understand the function of AAC-mediated inhibition during gamma oscillations, we next explored whether GABAAR activation at the AIS has a hyperpolarizing or depolarizing effect (fig. S3). Both depolarizing (12) and hyperpolarizing (13) AAC effects have been reported. To avoid interference with the intracellular ionic environment, in particular the Cl concentration, cell-attached current-clamp recordings (14) were obtained from somata of CA3 PCs. Under these conditions, local GABA application to the AIS of the PCs resulted in transient hyperpolarizing responses (fig. S3, A and B). Similar hyperpolarizing inhibitory effects were observed when GABA was applied to the soma (fig. S3, A and B). Bath application of the antagonist bicuculline (10 µM) reversibly blocked the hyperpolarizing responses (by 83.6 ± 4.8% and 86.0 ± 2.2% at the AIS and soma, respectively; n = 4 each), confirming that they were indeed mediated by fast ionotropic GABAARs (15).

These results convergently indicated that AACs firing at high frequency could produce strong inhibition in the AIS of CA3 PCs during network activity. Could this inhibition underlie the failure of EAPs to back-propagate to the soma? We thus examined the effect of local GABA application on antidromic APs elicited by brief trains of electrical pulses to the distal axon in nonoscillating slices. AP invasion of the soma was detected by recording ACs in cell-attached voltage-clamp configuration. Without GABA application, ACs were recorded with high probability in response to the antidromic stimulation, indicating the reliable back-propagation of axonal APs. Application of GABA to the AIS during the stimulation significantly reduced the rate of ACs recorded at the soma (Fig. 3, A and C, and fig. S4, A and B). We repeated our experiments in whole-cell configuration, voltage clamping the PCs at –100 mV to prevent the generation of APs in the soma and the AIS. Without GABA application, antidromic ACs were recorded at the soma with similar high probability, indicating that these events were initiated by the extracellular stimuli in the distal axon rather than in the AIS or the soma. Application of GABA to the AIS of the hyperpolarized PCs resulted in a substantial reduction in the probability of back-propagating ACs (fig. S4C), which suggests that the shunting effect of inhibition, rather than membrane potential hyperpolarization, blocks the back-propagation of EAPs. In contrast, when GABA was applied to the soma, no significant reduction in the rate of ACs was observed (Fig. 3, B and D).

Fig. 3

GABA-mediated inhibition at the AIS reduces the back-propagation of ACs from the axon to the soma. (A) Schematic representation of the experimental configuration (top). Antidromic AC recorded from the PC in cell-attached configuration in response to alveus stimulation (five pulses at 50 Hz) in control conditions (bottom left, Alv, triangle) and in combination with puff-application of GABA to the AIS (100 µM; pressure 10 to 15 pounds per square inch; five pulses at 100 Hz) (bottom right, Puff(AIS)+Alv, triangle). (B) Corresponding data obtained in experiments with GABA application to the soma. (C) Bar chart shows that antidromic ACs can be recorded with high probability from the soma under control conditions (62.4 ± 4.1%), whereas the probability is reduced significantly when GABA is applied to the AIS (34.5 ± 7.3%, P = 0.006, data from 5 cells). (D) When GABA was puff-applied to the soma, the probability of antidromic AC at the soma was not significantly different between control conditions and GABA application in cell-attached configuration (62.0 ± 3.7% and 60.0 ± 3.6%, respectively; n = 7, P = 0.7). All these experiments were carried out in the presence of antagonists of fast glutamatergic transmission.

We then investigated whether a single AAC could provide sufficient GABAergic input in the AIS to suppress antidromically evoked APs. Simultaneous patch-clamp recordings were taken from synaptically coupled AAC and CA3 PC pairs (Fig. 4). The identity of the interneurons was confirmed by subsequent visualization and morphological analysis (Fig. 4, A to C). When the PC axon was stimulated without activating the presynaptic interneurons, ACs were observed with high probability at the soma (Fig. 4, D and E). Conversely, when a train of APs was elicited in the AACs during the stimulation of the PC axon, the probability of somatically recorded ACs was significantly lowered (Fig. 4, D and E). In stark contrast, activation of single BCs did not significantly decrease the rate of antidromic ACs in synaptically coupled CA3 PCs (Fig. 4E).

Fig. 4

Activation of a single AAC suppresses antidromic AP invasion of the soma in synaptically coupled PCs. (A) Neurolucida reconstruction of a synaptically coupled AAC and PC pair in the CA3 area (soma and dendrites of the AAC are in blue, axon in black; soma and dendrites of the PC are in pink, axon in green). The inset, bottom, illustrates the characteristic fast-spiking discharge of the AAC in response to a depolarizing current pulse. (B) Projection of a confocal image stack of the same PC AIS. Arrows indicate the location of putative synaptic contacts formed by AAC axon varicosities in close apposition with the AIS. The AIS (blue), with immunostaining for the marker phospho-IκBa superimposed (green), is shown. (C) The axon of the interneuron formed a series of boutons (blue, arrows) in close apposition with other PC AISs immunolabeled using antibodies against phospho-IκBa (green), confirming the identity of the AAC. (D) Antidromic ACs were recorded from the whole-cell voltage-clamped PCs in response to alveus stimulation (5 pulses at 50 Hz) under control conditions (left, Alv, triangle) and after the activation of the synaptically coupled AAC (right, AAC+Alv; train of 5 APs elicited at 100 Hz by brief somatic current injection during the stimulus train to the axon). (E) Schemes (top) illustrate the experimental configurations. Summary bar charts on the effect of AAC (bottom left) and BC (bottom right) activation on the probability of antidromic AC invasion of the soma of postsynaptic PCs. Whereas activation of AACs (five paired recordings) results in a significant reduction in the probability of AC invasion (AAC+Alv) in comparison with the control condition (Alv) (**P = 0.002), activation of BCs (four paired recordings) had no significant impact (P = 0.2).

To further analyze the mechanisms underlying the control of somatic invasion by EAPs, we have used computational models of CA3 pyramidal cells (16), AACs, and BCs (figs. S5 to S7). In agreement with our experimental findings, brief high-frequency activation of a presynaptic AAC reliably blocked the invasion of AIS and the somatodendritic domain by antidromic APs (fig. S5C). For AACs, peak conductance amplitudes ≥0.01 µS reliably abolished the back-propagation of antidromic APs, whereas for BCs, a conductance larger by a factor of 3, 0.03 µS, was required (fig. S5, D and E). Qualitatively similar results were obtained when various parameters, including the reversal potential for GABAAR-mediated inhibition, were changed (fig. S5, F and G), further indicating that the shunting effect of the inhibitory conductance at the AIS plays the major role in blocking the back-propagation of EAPs from the distal axon.

Finally, to investigate whether the AACs are also more efficient in controlling orthodromic initiation of APs, we compared the impact of AAC- and BC-mediated inhibition on the generation of APs in response to somatic current injections in a computational model. High-frequency activation of a single AAC or BC did not change the site of AP initiation but shifted the current threshold to higher values (fig. S8A). However, activation of BCs resulted in a larger increase of the threshold at high (>0.04 µS) inhibitory conductances. Thus, although AACs are more efficient in preventing the back-propagation of antidromic APs, BCs can better control the initiation of orthodromic APs in response to strong excitatory inputs. This conclusion was supported by experimental evidence: Application of GABA to the AIS transiently attenuated the initiation of orthodromic ACs induced by weak, but not strong, somatic depolarization, which suggests that activation of GABAARs at the AIS is less efficient in preventing the generation of orthodromic ACs (fig. S8, B and C).

Our results reveal that cortical PCs are divided into two electrogenic compartments during gamma activity: The soma discharges at low frequency, whereas in the axon, EAPs are generated at higher frequencies. This functional separation is maintained by strong inhibition that is mediated by highly active AACs. This powerful inhibition to the AIS prevents the back-propagation of EAPs to somatodendritic compartments but allows for the generation of orthodromic APs when the excitatory drive is high. Our results are in agreement with proposed models of persistent gamma oscillations, whereby the plexus of PC axons generates fast oscillations, which coexist with local field gamma oscillations (1719). The net synaptic output produced by the PC axonal plexus provides the major excitatory drive onto inhibitory interneurons. These high-frequency axonal APs could orthodromically excite interneurons without necessarily causing PC somata to fire (17, 18), thereby providing the drive for local oscillations even during sparse orthodromic activation of the network from upstream regions (19). EAPs have been suggested to arise under certain conditions as a result of the activation of presynaptic GABAARs, with a depolarizing effect and transient elevations in (K+)o during synchronous GABA-mediated potentials (20). However, the potential mechanisms of EAP generation during the network oscillations remain unknown and will need future experimental evaluation.

The high-frequency, non–phase-locked discharge of AACs during gamma activity differs dramatically from that of BCs and other GABAergic interneurons. However, this discharge pattern is consistent with recent in vivo data indicating that cortical AACs, in contrast to other interneuron types, receive and respond to tonic rather than phasic excitation (21). AACs express significantly less extrasynaptic GABAAR α1 subunits in the plasma membrane than BCs do (22). This subunit contributes heavily to tonic inhibition and facilitates coincidence detection and temporally precise discharge in BCs (2224). The low level of tonic inhibition in AACs, in contrast, favors temporal integration of synaptic inputs and the firing of tonic APs. Indeed, these interneurons fire vigorously when excitation levels are heightened in the network in vivo (21).

Our experimental data demonstrate that AACs have an inhibitory effect in the AIS of CA3 PCs in the hippocampus. Thus, during gamma activity, high-frequency discharge of AACs leads to buildup of a tonic level of inhibition in the AIS of postsynaptic neurons. The hyperpolarization and the shunting influence by the inhibitory conductance act together to efficiently block the passage of APs from the axon to the soma. The small diameter of the distal axon limits the depolarizing current flowing into the AIS produced by the antidromic AP (25, 26). In contrast, when the neurons receive strong excitation onto their somatodendritic surface, sufficient charge can accumulate at the AIS to overcome inhibition and initiate APs, which then propagate to the axon and can also back-propagate to the dendrites.

Our study provides insight into the function of one of the most powerful cortical inhibitory interneuron types, AACs in cortical networks. In contrast to BCs, which provide strong inhibition to control orthodromic initiation of APs in response to excitatory synaptic inputs, AACs are more efficient at preventing the propagation of EAPs from the axon to somatodendritic compartments. Thus, AACs play an important role in functional dissociation of axonal/somatic compartments and orthodromic/antidromic activity, and they maintain the dynamic polarization of PCs in active cortical networks.

Supplementary Materials

Materials and Methods

Figs. S1 to S8

References (2739)

References and Notes

  1. Acknowledgments: We thank R. Traub, M. Brecht, and J. Geiger for critically reading the manuscript; F. W. Johenning for help with two-photon imaging; A. McMurtrie for her contribution to morphological reconstruction; C. Schultz for generously sharing the IκBa antibody; and H. Monyer for providing the parvalbumin-positive enhanced green fluorescent protein mice. This work was supported by the Deutsche Forschungsgemeinschaft (SFB-TR 3/B5 and GL 254/5-1 to T.G.; EXC 257 to D.S. and I.V.) and the Bundesministerium für Bildung und Forschung (BCCN II/A3 to T.G.).
View Abstract

Navigate This Article