Differential Shunting of EPSPs by Action Potentials

See allHide authors and affiliations

Science  05 Jan 2001:
Vol. 291, Issue 5501, pp. 138-141
DOI: 10.1126/science.291.5501.138


Neurons encode information and communicate via action potentials, which are generated following the summation of synaptic events. It is commonly assumed that action potentials reset the membrane potential completely, allowing another round of synaptic integration to begin. We show here that the conductances underlying the action potential act instead as a variable reset of synaptic integration. The strength of this reset is cell type–specific and depends on the kinetics, location, and timing of the synaptic input. As a consequence, distal synapses, as well as inputs mediated byN-methyl-d-aspartate receptor activation, can contribute disproportionately to synaptic integration during action potential firing.

Most neurons fire action potentials (APs) continuously in vivo. As a consequence, APs and synaptic potentials are constantly interacting during normal brain function. The pioneering work of Eccles and colleagues in spinal motor neurons (1, 2) suggested that APs can reset the membrane potential during an excitatory postsynaptic potential (EPSP), consistent with the large conductances underlying the AP (3). This “shunting” behavior of APs is incorporated into integrate-and-fire models of neurons, where APs typically generate a complete reset of synaptic integration (4). Most neurons, however, have extended dendritic trees and cannot be described by a single electrical compartment. Furthermore, recent experiments have demonstrated that different neurons express different voltage-gated channels (5), which are not uniformly distributed over the neuronal membrane (6). This raises the possibility that in different neurons and at different synapses, APs will shunt EPSPs to different extents.

To examine this possibility, APs were paired with EPSPs during whole-cell patch-clamp recordings from neocortical layer 5 pyramidal and cerebellar Purkinje neurons in rat brain slices (7). Subtraction of APs elicited in isolation from paired responses revealed that the EPSP remaining after the AP (the “shunted EPSP”) was greatly reduced both in peak amplitude and integral (Fig. 1A). The magnitude of this EPSP shunting depended on cell type. In pyramidal neurons, APs reduced EPSP amplitude to 30 ± 3% (n = 10) of control EPSP amplitude at the same time point (8), whereas EPSPs in Purkinje neurons were reduced to only 73 ± 3% of control (Fig. 1B; n = 7; P < 0.05). In both cell types, large (>6 mV) and small EPSPs (<2 mV) were shunted to a similar extent (P > 0.05).

Figure 1

Shunting of EPSPs by APs. (A) Activation of an EPSP (top), an AP by a brief current pulse (2 ms, 1.4 nA; second from top), and concurrent activation of both EPSP and AP (second from bottom). The bottom panel shows the shunted EPSP (red), obtained by subtracting the AP from the trace where the EPSP and AP were evoked concurrently, together with the control EPSP evoked in isolation (blue). The dotted line indicates the time at which the amplitudes of the shunted and control EPSPs were compared. Averages of 10 to 15 interleaved trials recorded at the soma of a layer 5 pyramidal neuron (V m = –60 mV). (B) Control and shunted EPSPs in a layer 5 pyramidal neuron (left) and a Purkinje neuron (right). The EPSP preceded the antidromic AP by 10 ms. (C) Time window for shunting in layer 5 pyramidal and Purkinje neurons. Percentage reduction in shunted EPSP amplitude plotted against the time difference between EPSP and AP onset; positive times: AP precedes EPSP. Average from five to seven neurons.

EPSP shunting was highly sensitive to the relative timing of the AP and the EPSP. Shunting was maximal when the AP was initiated after EPSP onset and absent when the AP occurred more than 15 ms before the EPSP (Fig. 1C) (9). These findings suggest that EPSP shunting is generated predominantly by the AP itself, rather than conductances activated during the after-hyperpolarization (AHP). Consistent with this, EPSP shunting was similar in pyramidal neurons recorded with a pipette solution containing 20 mM 1,2-bis(2-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid (BAPTA) to block the calcium-dependent AHP (10) (26 ± 12% of control, n = 7; P> 0.05). To investigate whether shunting of EPSPs by APs is mediated by a change in driving force of the synaptic conductance, we mimicked EPSPs with somatic current injection (11). At times when the synaptic conductance should be active (AP onset 3 ms after EPSP onset), shunting of EPSPs generated by somatic current injection (30 ± 5% of control) was similar to that of evoked EPSPs (29 ± 6% of control, n = 4; P > 0.05). This indicates that the conductances underlying the AP, rather than changes in synaptic driving force, generate EPSP shunting.

Shunting of EPSPs generated by current injection depended strongly on the kinetics of the injected waveform, with EPSPs generated by using slow kinetics being shunted less than those generated by using fast kinetics (Fig. 2, A and B). Shunting of EPSPs generated by identical somatic current waveforms was always less in Purkinje neurons (Fig. 2B). This indicates that differences in EPSP shunting between pyramidal and Purkinje neurons (Fig. 1, B and C) are mediated by cell-specific differences in the AP rather than differences in excitatory postsynaptic current (EPSC) time course. In support of this idea, APs in neocortical pyramidal neurons are significantly broader than in Purkinje neurons (half-width 0.39 ± 0.02 ms (n = 5) versus 0.15 ± 0.01 ms (n= 4); P < 0.001).

Figure 2

EPSP shunting depends on synaptic input kinetics. (A) Shunting of EPSPs generated by current waveform injection with τrise = 0.2 ms and τdecay = 2 ms (left) and τrise = 5 ms and τdecay = 50 ms (right). Somatic recording from a layer 5 pyramidal neuron. (B) Relation between the half-width of the injected current waveform and the percentage reduction in shunted EPSP amplitude. Average from four to seven neurons. The AP was timed to occur near or after the peak of the EPSP. The lines are single exponential fits to the data. (C) Shunting of EPSPs in control conditions (“AMPA” EPSP, left) and in bicuculline, CNQX and 0.1 mM Mg2+ (NMDA EPSP, right). (D) Relation between the percentage reduction in shunted EPSP amplitude and the relative timing of the AP and EPSP for control AMPA EPSPs and NMDA receptor–mediated EPSPs in 0.1 mM or 1 mM Mg2+. Average from four pyramidal neurons.

The dependence of shunting on EPSC kinetics (Fig. 2B) suggests that EPSPs generated by transmitter-gated channels with slower kinetics will experience less shunting. Consistent with this idea,N-methyl-d-aspartate (NMDA) receptor–mediated EPSPs were shunted far less by APs timed at their peak (90 ± 8% of control; n = 8) than EPSPs mediated by activation of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionate (AMPA)–type glutamate receptors (Fig. 2C; 22 ± 4% of control;n = 6). The relation between EPSP shunting and AP timing was also different for the two types of EPSPs. NMDA EPSPs evoked simultaneously with or after the AP experienced little or no shunting, whereas shunting similar to that for AMPA EPSPs was attained when APs were evoked more than 20 ms after NMDA EPSP onset (Fig. 2D). The difference between shunting of EPSPs mediated by AMPA and NMDA receptors was further enhanced for NMDA EPSPs recorded in physiological concentrations of Mg2+ (1 mM), where relief of Mg2+ block (12, 13) could boost appropriately timed NMDA EPSPs (Fig. 2D).

Our finding that EPSP shunting is mediated primarily by the AP itself suggests that shunting may be greatest close to the soma, near the site of AP generation in the axon (14). We therefore generated EPSPs by current injection at different dendritic locations and examined shunting both at the site of EPSP origin and at the soma. The EPSP arriving at the soma was always shunted more than the EPSP at its site of generation in the dendrites (Fig. 3A), with a clear relation between the magnitude of shunting at the site of EPSP generation and distance from the soma (Fig. 3B). This result indicates that shunting is relatively ineffective at distances greater than ∼300 μm from the soma (Fig. 3, A and B). These experiments also allowed us to assess whether shunting of EPSPs at the soma depends on synaptic location (15). Indeed, somatically recorded EPSPs generated by dendritic current injection (408 ± 17 μm from the soma,n = 5) were shunted significantly less than those generated by current injection at the soma (Fig. 3C; somatic current injection 33 ± 4% of control versus 68 ± 4% for dendritic current injection; P < 0.005) (16). To examine the contribution of dendritic conductances activated by backpropagating APs to EPSP shunting, we blocked backpropagation by dendritic application of tetrodotoxin (TTX) (Fig. 3D). At the soma, shunting of EPSPs generated by dendritic current injection (290 ± 16 μm from the soma) was unchanged after application of TTX (control: 24 ± 5%; dendritic TTX: 18 ± 3%;n = 5; P > 0.05), despite a 57 ± 8% reduction in backpropagating AP amplitude at the site of EPSP generation (Fig. 3D) (17). These results indicate that conductances in the apical dendrite activated during backpropagating APs do not contribute significantly to EPSP shunting at the soma.

Figure 3

Dependence of EPSP shunting on synaptic location and AP backpropagation. (A) Shunting of EPSPs generated by dendritic current injection at the dendritic recording site (top) and soma (bottom). Simultaneous recording from the apical dendrite (300 μm from the soma) and the soma of a layer 5 pyramidal neuron. (B) The percentage reduction in shunted EPSP amplitude at the synapse is plotted against the location where EPSPs were generated in layer 5 pyramidal neurons. (C) Shunting at the soma for EPSPs generated by current injection into the apical dendrite (410 μm from the soma; left) or the soma (right) of a layer 5 pyramidal neuron in the presence of ZD7288. (D) Effect of blocking AP backpropagation on EPSP shunting in layer 5 pyramidal neurons. (Top) Backpropagating dendritic AP (250 μm from the soma) recorded in response to brief somatic current injection in control (left) and following dendritic application of TTX (1 μM; right). (Bottom) EPSP shunting recorded at the soma of the same neuron in control (left) and following block of backpropagation (right). EPSPs were generated by dendritic current injection.

Simulations using neuronal models were used to gain more insight into the mechanisms of EPSP shunting by APs. Some key features of EPSP shunting could be reproduced in a single-compartment model (18) by using brief rectangular conductance pulses in the nanosiemens range to represent the AP (Fig. 4, A and B). Increasing the duration—and thus also the integral—of the conductance pulse increased the extent of EPSP shunting (Fig. 4B), consistent with the idea that differences in AP width underlie differences in EPSP shunting in pyramidal and Purkinje neurons (Fig. 1, B and C). We also observed shunting of EPSPs using realistic AP conductances in compartmental models with morphologically realistic geometries (19).

Figure 4

Simulations of EPSP shunting in a single compartment model. (A) Shunting of EPSPs using a rectangular conductance pulse (1 ms in duration) with the reversal potential set at the resting potential (–65 mV). The conductance amplitude was varied from 1 to 50 nS in 1-nS steps. (B) The relation between the percentage reduction in shunted EPSP amplitude and conductance amplitude for three conductance pulse durations: 0.5, 1, and 2 ms.

Finally, we examined the effect of EPSP shunting by APs on temporal summation at the soma of neocortical pyramidal neurons. EPSPs were generated by somatic current injection, and temporal summation of two EPSPs was quantified by measuring the peak current required for the second EPSP to reach AP threshold. These experiments revealed that when an AP occurred during the first EPSP, the efficacy of the second input was similar to when the AP was evoked on its own (Fig. 5, A to D). This indicates that APs can significantly reduce temporal summation of EPSPs. Summation generated by EPSPs with slow current kinetics was less affected by an intervening AP (P < 0.05), consistent with reduced shunting of slower EPSPs (Fig. 2). These results are summarized in Fig. 5E, and demonstrate that the effect of APs on temporal summation will depend on the extent to which they shunt coincident synaptic input.

Figure 5

The effect of shunting on temporal summation of EPSPs in neocortical pyramidal neurons. (A) EPSPs were generated by somatic current injections (lower traces) and the amplitude required to reach threshold for AP generation (1.09 nA) determined. Several overlapping sweeps are shown (suprathreshold current traces in red). (B) Summation of this EPSP with a preceding EPSP (1.0 nA peak amplitude; 20 ms before the control EPSP) lowered the current threshold for AP generation (0.59 nA). (C) A preceding antidromic AP timed to occur 5 ms after the first EPSP did not significantly affect the current threshold for AP generation (1.10 nA). (D) When the preceding antidromic AP and EPSP were combined, the current threshold for AP generation (1.01 nA) was similar to that in the absence of the preceding EPSP, suggesting strong shunting of the preceding EPSP. (E) Summary of results from five neurons. The ratio of the current threshold for AP generation with a preceding EPSP, AP or EPSP/AP combination compared to the current threshold without any preceding event. “Fast” EPSPs were generated by current waveforms with τrise = 0.3 ms and τdecay = 3 ms [as in (A) to (D)]; “slow” EPSPs were generated by current waveforms with τrise = 2 ms and τdecay = 20 ms. The amplitude of the slow EPSP was scaled to produce a similar reduction in AP current threshold as the fast EPSP in the same cell.

Our findings provide a fresh perspective on the role of APs in synaptic integration. Rather than simply riding on top of EPSPs, the large conductances activated during the AP influence integration by rapidly shunting synaptic charge from the membrane. The reset of synaptic integration is not complete, however, as in classical integrate-and-fire models (4). Rather, a variable amount of synaptic charge survives each AP, depending on neuronal type, the relative timing of the AP and the synaptic input, and the kinetics and location of the synaptic conductance. As many neurons fire APs at high rates in vivo, integration will only take place over a narrow time window unless mechanisms are in place to mitigate shunting. It is noteworthy that Purkinje neurons fire at unusually high rates in vivo [>50 Hz (20)] and show much less shunting than pyramidal neurons, possibly because they have optimized their AP conductances to minimize shunting. Other neurons may achieve the same result by using slower synaptic conductances, such as those mediated by NMDA receptors (which Purkinje neurons lack). Finally, as distal inputs survive shunting, we predict that they will contribute disproportionately to synaptic integration during high levels of AP firing. By defining multiple spatial compartments and temporal windows for summation of synaptic inputs, shunting of synaptic potentials by APs represents a powerful computational mechanism influencing the way neurons integrate the many thousands of synaptic inputs they receive.

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: m.hausser{at}

  • Present address: Bell Laboratories, Lucent Technologies, Murray Hill, NJ 07974, USA.


View Abstract

Stay Connected to Science

Navigate This Article