Identified Sources and Targets of Slow Inhibition in the Neocortex

See allHide authors and affiliations

Science  21 Mar 2003:
Vol. 299, Issue 5614, pp. 1902-1905
DOI: 10.1126/science.1082053


There are two types of inhibitory postsynaptic potentials in the cerebral cortex. Fast inhibition is mediated by ionotropic γ-aminobutyric acid type A (GABAA) receptors, and slow inhibition is due to metabotropic GABAB receptors. Several neuron classes elicit inhibitory postsynaptic potentials through GABAA receptors, but possible distinct sources of slow inhibition remain unknown. We identified a class of GABAergic interneurons, the neurogliaform cells, that, in contrast to other GABA-releasing cells, elicited combined GABAA and GABAB receptor–mediated responses with single action potentials and that predominantly targeted the dendritic spines of pyramidal neurons. Slow inhibition evoked by a distinct interneuron in spatially restricted postsynaptic compartments could locally and selectively modulate cortical excitability.

Gamma-aminobutyric acid (GABA) is the major inhibitory transmitter in the cerebral cortex (1). Extracellular stimulation of afferent cortical fibers elicits biphasic inhibitory postsynaptic potentials (IPSPs) in cortical cells. The early phase is due to the activation of GABAAreceptors resulting in Cl conductance, and the late phase is mediated by K+ channels linked to GABABreceptors through heterotrimeric GTP-binding proteins (2–6). Although dual recordings revealed several classes of interneurons evoking fast GABAAreceptor–mediated responses in the postsynaptic cells, it is not clear whether distinct groups of inhibitory cells are responsible for activating GABAA and GABAB receptors. GABAergic neurons terminate on separate subcellular domains of target cells (7, 8), and several studies suggest that dendritic inhibition is mediated by GABAB receptors and possibly by a discrete group of interneurons (9,10) that can modulate dendritic excitability (11). IPSPs with similar kinetics to GABAB receptor–mediated responses are produced by interneurons possibly targeting the dendritic regions in the hippocampus (10), but other experiments provide evidence for pure GABAA responses evoked on dendrites (12–15). Moreover, repetitive firing of interneurons and/or cooperation of several interneurons is thought to be necessary for the activation of GABAB receptors (6, 14,16, 17), possibly by producing extracellular accumulation of GABA to levels sufficient to activate extrasynaptic receptors (4, 14, 16–19).

Whole-cell recordings with biocytin filling from synaptically coupled pairs of three types of presynaptic interneurons and postsynaptic pyramidal cells, combined with correlated light and electron microscopy, were performed (20, 21). GABAB receptor localization studies indicated a gradient-like immunoreactivity for GABAB receptors, with stronger labeling in the upper layers (22). We thus tried to identify the sources of slow inhibition in layers 2 to 3 of the rat somatosensory cortex. Neurogliaform cells (NGFCs, n = 78) were identified on the basis of a late spiking firing pattern and their axonal and dendritic morphology (23–28) (Fig. 1, A and D). Basket cells (n = 19) showed a fast spiking firing pattern, received depressing unitary excitatory postsynaptic potentials (EPSPs) arriving from pyramidal cells (n = 5), showed immunoreactivity for parvalbumin (n = 4 out of 4 tested), and preferentially innervated postsynaptic somata (31%), dendritic shafts (66%) and occasionally spines (3%). Bitufted cells (n = 15) responded to depolarizing current pulses with a so-called low-threshold spiking firing pattern (28,29), received facilitating EPSPs from neighboring pyramidal cells (n = 3), placed their synapses onto dendritic shafts and spines (74 and 26%, respectively; n = 45), and contained somatostatin (n = 4 out of 4 tested). Postsynaptic potentials in pyramidal neurons elicited by NGFCs showed slower (P < 0.001, Mann-Whitney test) 10 to 90% rise times (23.4 ± 9.8 ms, n = 54) when compared to IPSPs due to basket cell (5.8 ± 2.0 ms, n = 19) or bitufted cell (6.5 ± 1.7 ms, n = 15) activation (Fig. 1B). The decay of NGFC-to-pyramid IPSPs could not be fitted with single or double exponential functions. We thus measured the half-width of IPSPs for statistical comparison and found that NGFCs to pyramid IPSPs were significantly longer (P < 0.001; 183.9 ± 82.5 ms, 61.3 ± 16.3 ms, and 58.9 ± 17.9 ms for NGFC, basket, and bitufted-to-pyramid connections, respectively). Voltage clamp experiments confirmed the conclusions of these recordings (Fig. 1C).

Figure 1

Correlated electrophysiology and anatomy of neurogliaform cell–to–pyramidal cell connections. (A) Response of a neurogliaform cell to hyperpolarizing (top) and depolarizing (middle and bottom) current steps. (Band C) A single action potential (blue) elicited in a neurogliaform cell evokes IPSPs (black) in the postsynaptic pyramidal cells (average of n = 54 pairs). Superimposed traces compare fast and slow IPSPs (B) and IPSCs (C) from basket cell–to-pyramid connections (green, average of n = 19 pairs) and bitufted cell–to-pyramid connections (red, average ofn = 15 pairs) with the neurogliaform-to-pyramid connections at –50 mV membrane potential. The expanded time scale of the bottom panels reveals the differences in activation kinetics. (D) Two examples of random electron microscopic samples of targets postsynaptic to NGFCs. Labeled neurogliaform axons (a) form synapses (arrows) on a dendrite (d) at the base of a spine (s, left) and on a spine head (s, right). Asterisks mark asymmetrical synapses established by unidentified terminals (t). (E) Reconstruction of a neurogliaform cell (soma and dendrites, blue; axon, red)–to-pyramid (soma and dendrites, black; axon, green) connection. The inset at upper left shows the number and position of electron microscopically–verified synapses mediating the connection. Cortical layers are indicated on the right. (F) Serial electron microscopic sections of a synaptic junction [arrow, corresponding to number 1 in (E)] established by the axon of the neurogliaform cell (a) targeting the base of a spine (s) emerging from a dendritic shaft (d) of the pyramidal cell.

Random electron microscopic sampling of postsynaptic targets showed that NGFCs predominantly innervated dendritic spine necks (30%), spine heads (41%), and dendritic shafts (29%, n = 65 target profiles) (Fig. 1D) (27, 28). Three-dimensional light microscopic mapping of NGFC-to-pyramid connections (n = 8) confirmed these results and predicted 10 ± 6 synapses on dendritic spines and shafts of pyramidal cells at distances 62 ± 28 μm from the somata. Full electron microscopic analysis of all synapses mapped by light microscopy was performed on a randomly selected pair, and it revealed one synapse on a dendritic spine neck, three on spine heads, and one on a dendritic shaft 63 ± 27 μm (range, 25 to 92 μm) from the soma (Fig. 1, E and F).

NGFC-to-pyramid IPSPs were composed of two components (n = 21, Fig. 2, A and B). The early component could be blocked by bicuculline (10 μM,n = 10) or gabazine (20 μM, n = 3), indicating the involvement of GABAA receptors (Fig. 2A). Bicuculline or gabazine blockade alone never abolished the response completely and revealed a residual slow component of neurogliaform IPSPs with onset latencies of 60.6 ± 17.3 ms. This late component contributed to the integral of the control IPSPs by 32.1 ± 19.8% and could be blocked by further addition of the GABABreceptor antagonist CGP35348 (60 μM). The presence of a postsynaptic GABAB receptor–mediated slow component was confirmed by experiments in which the decay of NGFC-to-pyramid IPSPs was reversibly shortened by CGP35348 (n = 8, Fig. 2B). Although CGP35348 decreased the amplitude of the early component in three out of eight connections, the difference was not significant for the whole data set. The early component was absent at –72 ± 1 mV (n = 8), which was the expected reversal potential for mixed chloride and hydrocarbonate conductance (Fig. 2C); therefore, anion passage through GABAA receptors was responsible for the early phase, in agreement with the bicuculline blockade. Hyperpolarization of the postsynaptic cells near the equilibrium potential for potassium ions (–87 ± 2 mV) largely eliminated the late component, consistent with GABAB receptor involvement.

Figure 2

Pharmacology of neurogliaform-to-pyramidal cell connections. Traces show averages ± SEM (gray) of several pairs. (A) The initial component of control IPSPs (n = 10) elicited by single presynaptic action potentials (top) was blocked by bicuculline (10 μM), and the late phase of IPSP was abolished by the subsequent addition of CGP35348 (60 μM). The IPSPs showed recovery after 30 min of washout. (B) The decay of the IPSPs (n = 8) evoked by single spikes in NGFCs (top) could be shortened by application of CGP35348 (60 μM), and this effect could be partially reversed by returning to the control solution. Superimposed traces are shown normalized to the amplitude of control IPSPs (bottom). (C) Voltage dependency of the unitary neurogliaform-to-pyramid IPSPs (n = 6) recorded at membrane potentials of –50, –72, and –87 mV. The early phase shows a reversal potential of –72 mV; the late phase is eliminated at –87 mV. (D) Rapid use-dependent exhaustion of NGFC-to-pyramid connections demonstrated by a triple recording with a single presynaptic and two postsynaptic cells. Top, the first five consecutive postsynaptic responses (single sweeps) to presynaptic spike trains elicited at 40 Hz once in 4 min. Bottom, single postsynaptic responses to a presynaptic spike after an inactive period of 30 min show partial recovery.

The compound IPSPs were highly sensitive to the firing rate of the presynaptic neurons. This could explain why the sources of slow inhibition have remained obscure up to now. We activated the presynaptic NGFCs with single action potentials delivered at various intervals, and a stable amplitude of postsynaptic responses could only be achieved if the interval between presynaptic spikes was more than 1.5 min. Accordingly, all single action potential–evoked responses for the kinetics, pharmacology, and reversal potentials detailed above were collected at especially low presynaptic firing rates (one spike in 100 to 120 s). When the presynaptic NGFCs were activated with trains of action potentials at 40 Hz, the amplitude of postsynaptic responses decreased rapidly (Fig. 2D). Even at a train interval of 4 min (n = 7), postsynaptic responses showed a rapid decrease in amplitude resulting in a complete loss of response after five to eight presynaptic spike trains. After total exhaustion, the recovery of IPSP amplitude was tested with a single presynaptic spike in every 15 min and recovery occurred in all cases. The recovery was initially detectable after 15 to 45 min and reached 31 to 79% of control amplitude as measured 90 min after exhaustion, indicating that the synapses remained functional. The application of high-frequency stimulation or presynaptic interspike intervals above 1.5 min did not have an effect on the kinetics of single spike–evoked events in the same pair.

Our results provide evidence that slow GABABreceptor–mediated IPSPs arrive from unitary sources in cortical networks. We identify the first cell type, NGFCs, which consistently recruit postsynaptic GABAB receptors in addition to GABAA channels. Synapses of NGFCs appear to be specialized for sparse temporal operation tuned for long-lasting metabotropic effects. Although it has been suggested that in some interneuron-to-pyramidal cell connections repeated presynaptic activation might be necessary to recruit slow inhibition (14, 16), single action potentials at very low firing rates are sufficient to elicit the metabotropic GABAB component. We cannot rule out, however, the possibility that other type(s) of GABAergic cells might also activate postsynaptic GABAB receptors. GABA uptake mechanisms powerfully remove the transmitter from the extracellular space within a distance restricted to about 1 μm from the release sites (30). Our results thus suggest that postsynaptic GABAB receptors could be spatially associated with the synapses formed by NGFCs. Electron microscopic studies revealed extrasynaptically placed GABAB receptors on dendritic spines and shafts (22, 31–33). A possible synaptic enrichment of these receptors remains to be determined. We showed that the action of NGFCs is predominantly targeted to dendritic spines. The slow rise times of NGFC-to-pyramidal cell IPSPs and inhibitory postsynaptic currents might support the filtering effect of the spine necks. Alternatively, the composition of GABAA receptor subunits might influence activation kinetics (34, 35). Although we cannot rule out the possibility that neurogliaform synapses on dendritic shafts and spines act through different receptors, data from the cerebellum suggest that GABAB receptors are placed on spines (22,31). Spines receive the majority of excitatory input, and simulations showed that if inhibitory synapses found on cortical spines are effective, then they should be mediated through GABABreceptors providing powerful hyperpolarizing inhibition that reduces the excitatory synaptic potentials on the same spine (36). In addition to hyperpolarizing inhibitory effects, the diffusion barrier provided by the targeted postsynaptic spines can locally enhance metabotropic changes after GABABreceptor activation. Therefore, even sparse temporal operation of NGFCs could result in sustained modulation of excitability.

Supporting Online Material

Materials and Methods

  • * To whom correspondence should be addressed. E-mail: gtamas{at}


Stay Connected to Science

Navigate This Article