Research Article

Dendritic Mechanisms Underlying Rapid Synaptic Activation of Fast-Spiking Hippocampal Interneurons

See allHide authors and affiliations

Science  01 Jan 2010:
Vol. 327, Issue 5961, pp. 52-58
DOI: 10.1126/science.1177876

Abstract

Fast-spiking, parvalbumin-expressing basket cells (BCs) are important for feedforward and feedback inhibition. During network activity, BCs respond with short latency and high temporal precision. It is thought that the specific properties of input synapses are responsible for rapid recruitment. However, a potential contribution of active dendritic conductances has not been addressed. We combined confocal imaging and patch-clamp techniques to obtain simultaneous somatodendritic recordings from BCs. Action potentials were initiated in the BC axon and backpropagated into the dendrites with reduced amplitude and little activity dependence. These properties were explained by a high K+ to Na+ conductance ratio in BC dendrites. Computational analysis indicated that dendritic K+ channels convey unique integration properties to BCs, leading to the rapid and temporally precise activation by excitatory inputs.

Fast-spiking, parvalbumin-expressing, γ-aminobutyric acid (GABA)–releasing (GABAergic) interneurons (BCs) play a key role in the function of neuronal networks. These neurons set a narrow time window for temporal summation in principal neurons by fast feedforward and feedback inhibition, contribute to the generation of network oscillations, and are thought to be involved in higher brain function and dysfunction (17). After stimulation of excitatory input synapses in vitro, BCs respond with remarkable speed, exquisite temporal precision, and preferential activity in the onset phase of a stimulus train (8, 9). Similarly, during network activity in vivo, such as sharp-wave ripple or theta rhythms, BCs are activated by input from principal neurons with short latency and minimal jitter (1013). The mechanisms underlying this rapid and precise activation are unclear. It is generally thought that synaptic factors, such as the time course of the postsynaptic conductance and the extent of depression or facilitation, play an important role (8, 9, 1416). Alternatively or additionally, the electrical properties of interneuron dendrites may contribute. Whereas the dendrites of pyramidal neurons were extensively characterized (1722), those of fast-spiking, parvalbumin-expressing BCs have not been directly examined. However, Ca2+ imaging experiments suggest that Na+, K+, and Ca2+ channels may be present in the dendrites of neocortical fast-spiking interneurons (23).

To study the dendrites of BCs directly, we used confocally targeted patch-clamp recording in hippocampal slices (Fig. 1) (20). After filling BCs in the dentate gyrus with Alexa Fluor 488 via somatic recording, we traced dendrites into molecular layer or hilus by using confocal imaging (Fig. 1A) and obtained dendritic recordings by using correlated infrared-differential interference contrast videomicroscopy (Fig. 1B). BCs were identified on the basis of the location of the axon in the granule cell layer (24) (Fig. 1C), the fast-spiking AP phenotype (25) (Fig. 1D; mean maximal frequency 104.2 ± 2.2 Hz; n = 98 dual recordings), and the immunoreactivity to the Ca2+-binding protein parvalbumin tested in a subset of neurons (Fig. 1E; 23 of 25 cells). Detailed analysis of the axonal arbor revealed that our sample was mainly composed of classical basket cells with tangential collaterals (78 of 83 recovered cells) but also included a subpopulation of cells with radial collaterals, suggestive of axo-axonic cells (5 of 83 cells) (fig. S3) (26).

Fig. 1

Confocally targeted recording from BC dendrites. (A) Confocal image (pseudocolor representation) of a BC in the dentate gyrus filled with Alexa Fluor 488 taken during the experiment. (B) Infrared differential interference contrast videoimage of the apical dendrite of the same cell as in (A). (C) Light micrograph of a BC filled with biocytin during recording and labeled by using 3,3′-diaminobenzidine. 10-μm stack projection. Arrows indicate the axonal arbor, forming “baskets” around granule cell somata. (D) Train of APs evoked by a 1-s, 0.75-nA current pulse applied at the soma (top) and AP frequency (f) – current (I) relation (bottom). Same cell as in (C). Bottom right graph shows mean maximal AP frequency (bar) and data from individual cells (points). (E) (Left) Confocal micrograph of a BC filled with biocytin and stained with fluorescein isothiocyanate (FITC)–conjugated avidin; (center) parvalbumin immunoreactivity of the same BC; (right) overlay. 40-μm stack projection.

APs propagate into BC dendrites with attenuated amplitude. We made simultaneous recordings from somata and apical dendrites of dentate gyrus BCs at distances up to 300 μm from the soma, close to the physical dendritic length (Fig. 2). Current injection at both the soma and the apical dendrite evoked high-frequency trains of APs (Fig. 2A). Although the AP frequency was identical at the soma and the dendrite, single APs at the two locations differed substantially (Fig. 2B). First, APs in the apical dendrite showed markedly attenuated amplitude in comparison with somatic APs. For the first AP in the train, the peak amplitude measured from threshold was 80.8 ± 2.0 mV at the soma and 23.4 ± 2.1 mV at the dendrite (at distances >100 μm from soma; P < 0.01; Fig. 2, B and D). Second, the maximal rate of rise declined as a function of distance (444.4 ± 29.5 mV ms−1 at the soma versus 81.3 ± 10.8 mV ms−1 at the dendrite at >100 μm; P < 0.01; Fig. 2, B and E). Lastly, the duration at half-maximal amplitude was slightly prolonged (0.50 ± 0.02 ms at the soma versus 0.84 ± 0.08 ms at the dendrite at >100 μm; P < 0.01; Fig. 2, B and F).

Fig. 2

APs in BC dendrites show marked amplitude attenuation, moderate broadening, and little activity dependence. (A) Train of APs evoked by somatic (left) and dendritic (right) current injection. Black traces, somatic voltage and corresponding current; red traces, dendritic voltage and corresponding current. (B and C) First AP (B) and last AP (C) in the 1-s train shown at expanded time scale. Dendritic recording site on apical dendrite 124 μm from soma. (D to F) Summary plot of AP peak amplitude measured from threshold (D), maximal slope of the rise (E), and duration at half-maximal amplitude (F) plotted against distance (positive distance, apical dendrite; negative distance, basal dendrite; both measured from the center of the soma). Data from 42 simultaneous somatodendritic and 8 somatic recordings (distance 0). Recording temperature ~23°C. Solid symbols, somatic current injection; open symbols, dendritic current injection. Red, first AP; blue, last AP in a 1-s train. For the half-duration of the first AP, only 39 out of 42 recordings could be analyzed in which voltage crossed the half-maximal amplitude level during repolarization. Lines represent exponential functions fit to the data points. For length constant values, see table S1.

To test whether dendritic AP propagation was activity-dependent, we compared the properties of the first and the last AP at distal dendrites in the high-frequency train (Fig. 2, B and C). The amplitude of the dendritic AP at the end of the train was similar to that at the onset (22.0 ± 1.8 mV versus 23.4 ± 2.1 mV; P > 0.5 for distal recordings >100 μm). Likewise, the maximal rate of rise and the duration at half-maximal amplitude were comparable between the first and the last AP (Fig. 2, D to F). To examine whether the properties of the dendritic AP depended on the site of current injection, we compared APs evoked by somatic and dendritic current injection in the same cell (Fig. 2, B to C). Dendritic APs evoked by somatic and dendritic current injection had similar amplitudes (22.7 ± 2.8 mV versus 24.8 ± 3.4 mV for the first AP; P > 0.4 for distal recordings). Likewise, the maximal rate of rise and the duration at half-maximal amplitude were independent of the site of current injection (Fig. 2, D to F).

Analysis of APs at near-physiological temperature gave comparable results (fig. S1). Furthermore, APs recorded in basal dendrites [receiving granule cell input (14)] were similar to those in apical dendrites (receiving mossy cell and entorhinal cortex input) (fig. S2). Lastly, similar results were obtained for neurons with radial axon collaterals, representing putative axo-axonic cells (fig. S3).

APs are initiated in the axon. To determine the site of AP initiation, we measured latency differences of APs at different subcellular locations (Fig. 3). In the majority of apical dendritic recordings (19 of 28 BCs), somatic APs preceded dendritic APs during long current pulses, independently of the site of current injection (Fig. 3, A and B). When measured at the half-maximal amplitude, the somatic AP preceded the dendritic AP by 0.21 ± 0.05 ms (14 apical dendritic recordings at distances of >100 μm). Likewise, in the majority of basal dendritic recordings (12 of 14 BCs), the somatic AP preceded the dendritic AP (dendritic recordings 17 to 62 μm from the soma).

Fig. 3

The axon is the exclusive site of AP initiation in BCs under a variety of conditions. (A and B) Recording from a BC in which the somatic preceded the dendritic AP (31 of 42 BCs). First somatic (black) and dendritic (red) APs evoked by 1-s somatic (A) and dendritic (B) current pulses. Left graphs, somatic and dendritic APs at absolute voltage scale; right graphs, APs normalized to same peak amplitude. Dendritic recording site on apical dendrite 176 μm from soma. (C) Responses to brief depolarizing current injection in the dendrite. (Left) Dendritic responses and (right) voltage-current relation for the voltage immediately after the pulse (arrow) in a subset of BCs in which no somatic spikes could be elicited (four out of six cells). Line represents linear regression of data points for voltages ≤–30 mV. (D) Somatic (black) and dendritic (red) APs initiated by evoked postsynaptic potentials (PSPs). Axons of the lateral perforant path were stimulated in the outer molecular layer. Left traces, somatic and dendritic APs at absolute voltage scale; right traces, APs normalized to same peak amplitude. (E and F) Recording from a BC in which the dendritic preceded the somatic AP (10 of 42 BCs). First somatic (black) and dendritic (red) APs evoked by 1-s somatic (E) and dendritic current pulses (F). (G) Correlated light microscopic morphological analysis of the same cell shown in (E and F). The axon (arrows) originated near the dendritic recording site. (H) Plot of latency between somatic and dendritic APs against the corresponding difference of distances, using the axon origin as reference point. Solid circles, somatic current injection; open circles, dendritic current injection. Line represents the result of linear regression; average dendritic AP propagation velocity was 0.53 m s−1. Recording temperature ~23°C except for synaptic experiments in (D) (~31°C).

In pyramidal neurons, brief high-intensity dendritic current pulses or distributed activation of excitatory synapses trigger dendritic spikes more effectively than long current pulses (2729). We therefore tested these stimulation paradigms in BCs. However, 0.5-ms current pulses applied to the dendrite (137 ± 17 μm) failed to initiate dendritic spikes in six out of six BCs and even failed to trigger somatic APs in four out of six recordings, despite the injection of large current amplitudes of up to 3.5 nA (Fig. 3C). Failure of AP initiation was presumably related to the marked rectification of the voltage-current relation, showing that it becomes increasingly difficult to depolarize the dendritic membrane potential beyond –30 mV (Fig. 3C). Activation of distal synapses by stimulation of axons of the lateral perforant path initiated APs (Fig. 3D). However, as with the other stimulation paradigms tested, the somatic AP constantly preceded the dendritic AP. Similarly, stimulation of two synaptic inputs with stimulation electrodes ~30 μm apart initiated APs in which the somatic consistently preceded the dendritic response (four BCs; fig. S4).

In a subset of BCs, either the dendritic APs preceded the somatic APs (10 of 42 BCs) or dendritic and somatic APs occurred simultaneously (1 of 42 BCs; Fig. 3, E and F). This could be due to either dendritic AP initiation (2729) or dendritic origin of the axon (3032). We therefore performed a correlated analysis of the morphological properties of BCs by using post-hoc biocytin labeling. In all cases in which the AP occurred first in the dendrite, the axon originated from the dendrite and was closer to the dendritic than the somatic recording electrode (Fig. 3G). To test whether the results from all morphologically recovered BCs were quantitatively consistent with axonal AP initiation, we plotted AP latency against the difference of distances of the recording locations from the axon origin (Fig. 3H). The linear regression line intersected the abscissa near 0, indicating AP initiation in the axon (correlation coefficient r = 0.90; P < 0.001).

High K+ to Na+ conductance ratio in BC dendrites. Both the robust axonal AP initiation and the marked dendritic AP attenuation could suggest that BC dendrites behave as passive cables. We therefore determined the density of voltage-gated Na+ and K+ channels in outside-out patches isolated at various locations (Fig. 4). In somatic outside-out patches obtained with Cs+-internal solution, voltage pulses from –120 mV to 0 mV evoked inward currents in the majority of patches (Fig. 4A). These currents showed fast activation and inactivation and were blocked by 1 μM tetrodotoxin in the external solution, demonstrating that they were mediated by voltage-gated Na+ channels (33) (Fig. 4B). In contrast, in dendritic outside-out patches, the amplitude of inward currents was substantially smaller (Fig. 4, A and C). Quantitative analysis revealed that the Na+ current density was 13.3 ± 2.1 pA μm−2 at the soma but steeply declined as a function of distance, with estimated length constants of 87 μm in apical dendrites and 25 μm in basal dendrites (Fig. 4D and table S1).

Fig. 4

High K+ to Na+ conductance density ratio in BC dendrites. (A) Na+ currents evoked in outside-out patches from soma, apical dendrite, and basal dendrite. Test pulse potential was 0 mV. Na+ currents were recorded with Cs+-internal solution. (B) Tetrodotoxin (TTX) sensitivity of Na+ channels. Top graphs, Na+ currents in control conditions and in the presence of 1 μM TTX (somatic outside-out patch). Bottom, plot of Na+ peak current against time during TTX application (horizontal bar); each point represents the mean of 10 consecutive measurements. (C) Difference between somatic and dendritic Na+ channel density in the same cell. (Left) Location of somatic and dendritic recording pipette, superimposed with morphological reconstruction of the somatodendritic domain of the BC. (Right) Na+ currents recorded in outside-out patches isolated from these locations by using two patch pipettes pulled from same glass capillary. (D) Na+ current density as a function of distance from the soma. Data from 17 somatic (black), 15 apical dendritic (red, positive distances), and 9 basal dendritic (red, negative distances) outside-out patches. (E) K+ currents evoked in outside-out patches from soma, apical dendrites, and basal dendrites. Test pulse potential was 70 mV. K+ currents were recorded with K+-internal solution. (F) K+ current density as a function of distance from the soma. Data from 7 somatic (black), 17 apical dendritic (red, positive distances), and 14 basal dendritic (red, negative distances) outside-out patches. Red lines represent exponential functions fit to the data points. For length constant values, see table S1. Na+ currents are the average of 38 to 50 sweeps; K+ currents are either single traces or the average of three sweeps. Leakage and capacitive currents were digitally subtracted.

Expression of Kv3-type K+ channels is a hallmark property of fast-spiking GABAergic interneurons (34, 35). We therefore explored the possible presence of Kv3 or other types of K+ channels (19, 36) in BC dendrites. In somatic outside-out patches isolated with K+-internal solution, voltage pulses from –120 mV to 70 mV evoked large voltage-dependent outward currents (Fig. 4E). Quantitative analysis revealed a K+ current density of 91.5 ± 21.1 pA μm−2 at the soma. In apical dendrites, the K+ current density decayed moderately as a function of distance, with an estimated length constant of 763 μm (Fig. 4F). In contrast, in basal dendrites, the decay was steeper, with a length constant of 57 μm.

To determine the molecular identity of the K+ channels expressed in BC dendrites, we measured functional parameters that discriminate among K+ channel subtypes (25, 34, 35) (fig. S5). Dendritic K+ channels in BCs showed several characteristic properties. First, they had a high activation threshold of ~–40 mV (fig. S5, a and b). Second, their activation time course was fast and highly voltage-dependent (fig. S5c). Third, they were blocked by low concentrations of extracellular tetraethylammonium (TEA), with only minimal distance dependence (fig. S5, d and e). Application of 1 mM TEA reduced the K+ current amplitude in dendritic patches to 62.7 ± 5.7%, similar to somatic patches (P > 0.05; fig. S5e), whereas 200 nM α-dendrotoxin (α-DTX) had almost no effect (amplitude 97 ± 0.6% of control; n = 3). Lastly, K+ channels showed only minimal inactivation, again with little distance dependence (fig. S5f). Because only Kv3 channels combine the properties of high activation threshold, rapid activation, high TEA sensitivity, and insensitivity to α-DTX, our results indicate that Kv3-type channels prevail in BC dendrites, consistent with immunocytochemical data [(34), but see (23)].

Dendritic K+ channels shape excitatory postsynaptic potential (EPSP) time course and coincidence detection in BCs. One implication of our results is that dendritic K+ channels may be efficiently activated by excitatory synaptic input (8, 9, 1416). We therefore injected artificial excitatory postsynaptic­ current–like (EPSC-like) waveforms (aEPSCs) into the dendrite (231 ± 12 μm) while recording the corresponding artificial EPSPs (aEPSPs) at the soma (Fig. 5). The parameters of the aEPSCs were chosen to mimic the amplitude and time course of unitary EPSCs in BCs (14). Bath application of 5 mM TEA reversibly prolonged the half-duration of single aEPSPs from 12.5 ± 0.7 ms to 14.8 ± 0.9 ms (P < 0.01; Fig. 5A). Because TEA had no significant effect on the membrane time constant (τ = 10.8 ± 0.8 ms in control versus 12.0 ± 1.8 ms in TEA; P > 0.2), these results indicate that voltage-gated K+ channels accelerate the decay of EPSPs (37).

Fig. 5

Dendritic K+ channels shape EPSPs and coincidence detection properties of BCs. (A) (Left) Artificial EPSPs (aEPSPs) evoked by injection of an EPSC-like current into the dendrite and recorded at the soma in control conditions (black) and in the presence of 5 mM TEA in the extracellular solution (blue). Peak amplitude of the aEPSC was 0.5 nA. (Right) Summary graph of half-duration of the aEPSP. Data from the same experiment are connected by lines. Peak amplitude of aEPSC was 0.5 – 3 nA; 10 μM 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) and 2 μM SR95531 were added in four out of nine experiments. (B) Simulation of K+ channel activation in a BC 2 ms after onset of an EPSP in the distal apical dendrite. Color code (right) shows density of activated K+ conductance. (C) Dendritic K+ channels lead to synapse-specific acceleration of the EPSP decay time course. EPSPs with passive dendrites (blue), K+ channels in the apical dendrites (black), and K+ channels in the basal dendrites (gray). Lower right graph shows normalized superposition. Synapse was located on the distal apical dendrite in all cases. Synaptic peak conductance was 1 to 10 nS. (D) Corresponding plot of EPSP half-duration against synaptic peak conductance. (E and F) Dendritic K+ channels enable synapse-specific coincidence detector properties. Plot of paired pulse summation ratio (EPSPmax/EPSP1) for dual activation of the same synapse [(E), continuous lines, distal apical dendrite; and dashed lines, basal dendrite] and activation of different spatially separated synapses (F). Corresponding traces are shown on top. Blue, passive dendrites; black, dendrites enriched with K+ channels.

To corroborate that the K+ channels activated by EPSPs were located dendritically, we simulated EPSPs in a previously established passive BC cable model (Fig. 5B) (38) with and without dendritic voltage-gated K+ channels. Insertion of K+ channels into the apical dendrites near the activated synapses accelerated EPSP kinetics (Fig. 5, C and D), reproducing the experimental observations (Fig. 5A). In contrast, insertion of K+ channels into the basal dendrites remote from the synapses had only minimal effects on EPSP kinetics.

Activation of BCs requires the coincident activation of multiple excitatory synaptic inputs (8, 14). To test how the K+ channel–mediated acceleration of the EPSP decay time course affects coincidence detection in BCs, we simulated pairs of EPSPs separated by variable time intervals (Fig. 5, E and F). If the same synapse was activated repetitively at variable time intervals, dendritic K+ channels reduced the extent and shortened the time window of temporal summation (Fig. 5E). In contrast, if two spatially separated synapses were activated at different times, the maximal extent of summation was unchanged, whereas the summation window was shortened by dendritic K+ channels (Fig. 5F). Thus, dendritic K+ channels enable BCs to selectively detect nearly coincident, spatially separated activity.

Dendritic K+ channels control EPSP-AP coupling and AP phenotype of BCs. To test how dendritic K+ channels influence EPSP-AP coupling, we inserted Na+ and K+ channels in densities consistent with our experimental results and compared scenarios with passive and active dendrites (Fig. 6). In the BC model with passive dendrites, APs evoked by both brief current pulses and synaptic activation were followed by a marked afterdepolarization. In contrast, in the model with active dendrites, the afterdepolarization was largely suppressed (Fig. 6B). Analysis of the underlying mechanisms revealed that the activation of dendritic K+ channels, which was particularly efficient for APs evoked by EPSPs, reduced the dendrosomatic current flow after APs (Fig. 6B).

Fig. 6

Dendritic K+ channels control EPSP-AP coupling and AP phenotype of BCs. (A) Simulation of K+ channel activation in a BC 2 ms after the onset of a somatic current stimulus triggering an AP. Color code (right) shows density of activated K+ conductance. (B) APs (bottom left), total activated dendritic K+ conductance (top right), and axial current in all primary dendrites (bottom right; negative peak truncated) versus time. APs were initiated either by brief somatic current pulses (2 ms, 1.2 nA, “1”) or by simultaneous activation of 10 synapses randomly placed on apical dendrites (“2”). Note that dendritic K+ channels abolished the afterdepolarization by reducing the dendrosomatic current flow (positive values of axial current, Iax). (C) Dendritic K+ channels ensure 1:1 EPSP-AP coupling over a wide range of synaptic strength. (Top) APs evoked by simultaneous activation of 50 synapses. (Bottom) Plots of number of evoked APs against number of activated synapses. (D) Dendritic K+ channels normalize the AP threshold for paired synaptic activation. (Top) APs evoked by consecutive activation of two sets of synapses on different dendrites (10 and 1 to 3 active synapses, respectively). (Bottom) Plots of threshold for AP initiation (in units of number of synapses) against time interval between EPSPs. Continuous lines, dual activation of the same synapse; dashed lines, activation of synapses on different dendrites. (E) Trains of APs evoked by 250-ms 2-nA depolarizing current pulses for passive and active dendrites. (F) Plot of AP amplitude, measured from the peak of AP to the trough of the subsequent afterpotential, versus time for different stimulus intensities (0.5 nA to 2.5 nA in 0.5-nA increments). Blue, passive dendrites; black, active dendrites. In all simulations, voltage given refers to the soma.

Correlated with this inhibition of the afterdepolarization, dendritic K+ channels changed the input-output characteristics of BCs. In the model with passive dendrites, strong synaptic stimuli (activation of >25 synapses) triggered bursts of APs. In contrast, in the model with active dendrites, single APs were generated over a wide range of synaptic stimulus intensities (Fig. 6C). Furthermore, with a repetitive synaptic stimulation paradigm the model with passive dendrites showed a greatly reduced AP initiation threshold for the second synaptic stimulus, whereas the model with active dendrites had an almost constant AP initiation threshold (Fig. 6D). Thus, dendritic K+ channels generate strength- and timing-independent activation properties of BCs.

Lastly, dendritic K+ channels contribute to the classical fast-spiking AP phenotype of BCs (25). In the model with passive dendrites, long depolarizing current pulses evoked trains of APs with markedly declining amplitude, and APs were occasionally initiated after the termination of the pulse (Fig. 6, E and F). In contrast, in the model with active dendrites, the AP amplitude was relatively constant during the train.

Discussion. The dendrites of BCs differ from those of pyramidal neurons in several ways. First, in BCs, the axon is the invariant AP initiation site. In contrast, in pyramidal neurons APs are preferentially triggered in the axon (17, 18) but can be also initiated in dendrites under specific conditions (2729). Second, in BCs, APs backpropagate into dendrites with marked amplitude attenuation, but little AP broadening and minimal activity dependence. In contrast, in pyramidal neurons backpropagation is less decremental and more activity-dependent (17, 18). Third, BC dendrites are endowed with a low Na+ channel density but a high Kv3 channel density, whereas pyramidal neuron dendrites show a high density of both Na+ channels and A-type (Kv4) K+ channels (19, 36). Thus, Kv3 channels are expressed in dendrites, somata, and axons of fast-spiking interneurons (35, 39). Lastly, K+ channels in BC dendrites shorten the EPSP time course, whereas Na+ channels in pyramidal neuron dendrites boost the amplitude and prolong EPSPs (37, 40).

Dendritic K+ channels shape the electrical properties of fast-spiking GABAergic interneurons at the level of input, EPSP-AP conversion, and output. At the input level, dendritic K+ channels accelerate the decay time course of unitary EPSPs, reduce the time window for temporal summation, and help BCs to detect the synchronous activity of converging, spatially separated excitatory inputs. Dendritic location puts K+ channels into a strategic position to be efficiently activated by EPSPs and enables them to implement synapse-specific processing rules. At the level of EPSP-AP conversion, dendritic K+ channels ensure precise 1:1 coupling between EPSP and AP, suppressing the generation of AP bursts by large synaptic inputs. Furthermore, dendritic K+ channels ensure a constant AP initiation threshold during repetitive stimulation. These properties are relevant during in vivo network activity, for example, during sequential activation of feedforward and feedback inputs on BCs (41) or sequential activation of presynaptic granule cells with adjacent place fields converging on the same BC during movement (42, 43). At the output level, dendritic K+ channels contribute to the fast-spiking AP phenotype (25), enhancing AP repolarization and thereby promoting recovery of Na+ channels from inactivation. Dendritic channels represent an almost infinite pool, which can be efficiently recruited during physiological and pathophysiological AP activity. After APs, dendritic channels will be activated with longer delays than somatic channels with identical gating, enhancing their delayed rectifier properties. Whether and how dendritic K+ channels also contribute to the complex (sometimes anti-Hebbian) induction rules for long-term potentiation at glutamatergic principal neuron-interneuron synapses remains to be determined (16, 44).

Because the dendrites of BCs differ significantly from those of somatostatin-expressing interneurons (30, 31), our results demonstrate that the diversity of GABAergic interneurons extends to the dendritic level. Previous results suggested that synaptic facilitation or depression and passive cable properties underlie the “routing” that leads to a switch from perisomatic to dendritic inhibition during repetitive activity (9). Our findings suggest that dendritic properties may participate in this dynamic switch. In fast-spiking, parvalbumin-expressing BCs, limited synaptic dynamics (14, 16) and dendritic K+ channel activation will facilitate stimulus-locked AP generation in the early phase of repetitive input synapse stimulation. In contrast, in somatostatin-positive interneurons, synaptic facilitation (4547) and a high Na+ channel density (30) will promote asynchronous AP generation late in the train. Thus, dendritic properties may contribute to setting the rules for routing of activity in inhibitory microcircuits.

Supporting Online Material

www.sciencemag.org/cgi/content/full/science.1177876/DC1

Materials and Methods

Figs S1 to S5

Tables S1 and S2

References

References and Notes

  1. We thank G. Buzsáki and A. Roth for critically reading the manuscript; P. Somogyi for help with cell identification; A. Nörenberg for providing the passive cable model; and S. Becherer, I. Koeva, M. Northemann, U. Thirimanna, and K. Winterhalter for technical assistance. Supported by the Deutsche Forschungsgemeinschaft (SFB 780/A5, SFB-TR 3/B10, and Leibniz program), the Bundesministerium für Bildung und Forschung (01 GQ 0420), the Norwegian Research Council (178670/V40), and the Epilepsy Foundation.
View Abstract

Navigate This Article