Barrages of synaptic activity control the gain and sensitivity of cortical neurons

Abstract

Ongoing synaptic activity, ever present in cortical neurons, may vary widely in its amplitude and characteristics, potentially having a strong influence on neuronal processing. Intracellular recordings in layer 5 pyramidal cells in prefrontal and visual cortical slices maintained in vitro revealed spontaneous periods of synaptic bombardment. Testing the responsiveness of these cortical cells to synaptic inputs or the injection of artificial excitatory postsynaptic conductances of various amplitudes revealed that background synaptic activity dramatically increased the probability of response to small inputs, decreased the slope of the input-output curve, and decreased both the latency and jitter of action potential activation. Examining the effects of different components of synaptic barrages (namely, depolarization, increase in membrane conductance, and increase in membrane potential variance) revealed that the effects observed were dominated by the membrane depolarization and increase in variance. Depolarization increased the peak cross-correlation between injected complex in vivo-like waveforms through enhancement of responsiveness to small inputs, whereas increases in variance did so through a shift in firing mode from one of threshold detection to probabilistic discharge. These results indicate that rapid increases in neuronal responsiveness, as well as increases in spike timing precision, can be achieved through balanced barrages of excitatory and inhibitory synaptic activity.

Figure 1.

Transition to the UP state results in an increase in responsiveness in layer 5 neurons to synaptic inputs. A, Simultaneous extracellular multiple unit (MU) in layer 5 and intracellular recording from a layer 5 cell. The response of the cell to a single local electrical stimulus (Stim. location 2) was examined while the network was in the DOWN state (evoked PSP) or while it was in an UP state that had been triggered by a separate electrical stimulus (Stim. location 1) applied to the white matter (wm; evoked UP and PSP). B, Probability of initiation of a spike in response to the evoked PSPs during the UP and DOWN states. The UP state facilitates the probability of generating a spike to the smaller-amplitude inputs, whereas the large-amplitude synaptic responses initiate spikes with a high probability in both the DOWN and UP states. C, The UP state is associated with a decrease in the both the latency and jitter (SD of latency) of spikes that are initiated by the PSPs. Error bars indicate SEM; no bars indicates that they were smaller than the size of the symbol.

Figure 2.

Characteristics of the UP-DOWN states and the protocol for injection of artificial EPSPs. A, Simultaneous extracellular and intracellular recording of the UP and DOWN states of activity in layer 5 of the prefrontal cortex. The UP state is associated with the arrival of a barrage of EPSPs that depolarize the membrane potential, increase the membrane conductance, and increase the membrane potential variance. MU, Multiple unit. B, Artificial EPSPs of random amplitude conductance (limited by steps of 2 nS; see Materials and Methods) were injected at a rate of 10 Hz and then separated into those that occurred during the UP and DOWN states. C, Expansion of the transition from DOWN to UP for detail. D, Overlay of responses of the cell to the 30 nS EPSP. Three measures were derived: probability that the EPSP will generate a spike, latency to onset of the first action potential, and finally the spike jitter, measured as the SD of the spike latency.

Figure 3.

Spontaneously occurring UP states result in an increase in responsiveness to mimicked synaptic conductances as well as a change in slope of the input-output relation. A, Examination of the probability of different-amplitude mimicked synaptic conductances to evoke an action potential during the DOWN and UP states. The UP state is associated with an increase in responsiveness to inputs and a decrease in slope of the input-output relation. B, Normalized data from seven cells illustrating the reproducibility of these effects. The synaptic conductance for each cell was normalized such that the conductance that resulted in a 0.5 probability of action potential discharge during the DOWN state was given a value of 1. The amplitude of each synaptic conductance was randomly chosen from a distribution of 2 to 60 nS in steps of 2 nS (see Materials and Methods). Extrapolation of the graphs to 0 conductance yields an estimate of the probability of spontaneous activity generating a response (denoted by an X).

Figure 4.

The UP state is associated with an increase in neuronal responsiveness and a decrease in spike latency and jitter. A, Example of intracellular recording from a layer 5 pyramidal cell with the injection of random-amplitude artificial EPSPs. B, The UP state is associated with a marked increase in neuronal responsiveness. C, Raster plots and peristimulus histograms for two different-amplitude EPSPs (26 and 40 nS) showing that, during the UP state, the latency and jitter are reduced. The late spikes that occur during the UP state are not included in the latency and jitter calculations because no comparable spikes occurred during the DOWN state (see Materials and Methods). D, The spike latency and jitter (SD of latency) decrease with increases in the amplitude of the injected EPSP. Both the spike latency and jitter are significantly smaller during the UP state in comparison with the DOWN state.

Figure 5.

Effects of increase in membrane potential, membrane conductance, and noise on neuronal responsiveness. A, Depolarization of the membrane potential with the intracellular injection of current results in a leftward shift in the input-output response, a decrease in latency, and a decrease in jitter of spike responses. B, Increasing the background membrane conductance (with the dynamic clamp system) from 0 to 40 nS results in a shift of the input-output relation to the right, an increase in latency, and an increase in spike jitter. C, Increases in the variance of the membrane potential (noise) result in a change in the slope and a “smoothing” of the input-output relation. There is also a decrease in spike latency and jitter for lower-amplitude artificial EPSPs. noise 0, No additional noise (although the baseline conductance for the noise is active); noise 1, increases in the SD of both g_e and g_i to 1 nS; noise 2, increase of these SDs to 2 nS.

Figure 6.

A, Mimicking the UP state with combined depolarization (6.6 mV depolarization), an increase in conductance (11 nS combined excitatory and inhibitory conductance), and an increase in noise (4.2 mV SD) results in effects similar to the real UP state. The level of the bottom trace is a visual indicator of the level of conductance of the concurrent artificial EPSP. B, The natural UP state resulted in a large shift to the left and a change in the slope of the spike probability curve in this cell. Mimicking the UP state with a similar level of depolarization (depol) and membrane potential variance resulted in a similar effect.

Figure 7.

The influence of noise on the action potential response depends on the duration of the input. A, Action potential response to the intracellular injection of a 400-msec-duration current pulse of varying amplitudes with and without the addition of background noise (4 nS for both g_e and g_i). Increasing noise enhances the response to the small current pulse (0.4 nA) more than to the large one (1.2 nA). B, Raster plot of the response to a 1.4 nA current pulse without and with the addition of 4 nS of noise. C, Plot of the number of spikes generated per pulse for the first 5 msec of the current pulse versus the amplitude of the current pulse as well as the amplitude of the added noise (Medium Noise, 2 nS; High Noise, 4 nS, for both g_e and g_i). Note that the noise increases the response to small short-duration events but decreases the response probability to larger short-duration events. D, Number of spikes per pulse for the first 320 msec of the current pulse. Note that the addition of noise enhances the response to small-amplitude pulses, with relatively little effect on the amplitude of large pulses, thereby resulting in a decrease in slope.

Figure 8.

Depolarization increases the correlation between a complex waveform and neuronal discharge. A, Current waveform, derived from an intracellular in vivo recording (Nowak et al., 1997), that was injected with varying levels of depolarization. The vertical dashed lines are visual guides to help align peaks in activity and the injected current. The histogram illustrates the response of the neuron at a membrane potential of -83 mV. B, Depolarization of the resting membrane potential of the cell to -76 mV with the intracellular injection of current results in a marked enhancement of neuronal responsiveness to the injected waveform. C, Group data (6 cells) demonstrating that depolarization results in an enhancement of the peak cross-correlation between the amplitude of the input and the neuronal firing rate.

Figure 9.

Increases in membrane variance can enhance the correlation between the spike response and the injected current waveform. A, Examples of injected current (derived from a visual response recorded in vivo) and the resulting membrane potential deviation in one neuron recorded in vitro. B, Raster plots of spike times over 375 trials in response to the injection of the current waveform in A during the DOWN period and without the addition of injected noise. Note the regularity of the spike response and the six peaks of activity within the box. Red dots are the first spikes to occur after the beginning of the examined period for each trial, as indicated by the dashed box. C, Same neuron as in B but with the addition of noise (2 nS SD). Note the appearance of spikes throughout the injected waveform. The scale is different for the histogram in C versus B. Again, red dots are the first spikes to occur on each trial after beginning of the indicated time period (dashed box). D, Group statistics showing that there is a significant increase in peak cross-correlation between the injected current and the spike histogram with the addition of 1 and 2 nS of added noise, but this effect is less with 4 nS of noise. E, Overlay of the injected current in A and the histograms in B and C plotted to the same scale. Note the increase in correlation between the waveform and the spike output after the addition of noise. F, Reordering of the spike data in B and C according to the time of the first spike (red dots) within the period examined (B, C, boxes). Note that without noise, the cell only discharges at peak 2 if it has not discharged to peak 1. Similarly, the cell only discharges to peak 3 if it has not discharged to peaks 1 or 2. If the cell discharges to peak 3, then the response to peak 4 is delayed. These results indicate that the generation of an action potential has a significant effect on the probability of generation of following spikes for ∼50-80 msec. The addition of background noise results in a reduction of this shadow effect, allowing the cell to discharge throughout the injected waveform. G, Examples of cross-correlations in the no noise and 2 nS noise conditions.

Figure 10.

Examination of the reduced probability of spiking after the generation of an action potential in cortical neurons. A, Example waveforms of the response of cortical cells to activation of a real EPSP (from Fig. 1), a dynamic clamp EPSP (from Fig. 2), one of the larger “events” in the complex waveform (from Fig. 9A), and injection of the depolarizing phase only of a 40 Hz sine wave (see B). The peak amplitudes of the waveforms were normalized to the same value to facilitate the comparison of time courses. All of the wave forms are of a similar time course. B, Protocol for examining the effect of spike generation on the responsiveness of the neuron. The depolarizing half of a 40 Hz sine wave was injected, followed at various intervals by injection of the same waveform but at different amplitudes. The first current injection was of two different amplitudes and adjusted so that the larger amplitude caused an action potential, whereas the smaller failed to do so. C, Example of raw data. The amplitude and time of occurrence of the second sine wave were pseudorandomized (between discrete time steps of 9-10 msec). Overlay, Membrane potential response of the cell to the sine wave injection that caused a spike (black) and one that did not cause a spike (red). Note that the occurrence of a spike shortens the falling phase of the response. D, Plot of the change in probability of eliciting an action potential by inputs of various amplitudes and at various intervals after the generation of a spike. The change in probability is calculated by subtracting the probability curve with a spike from the curve without a spike for each amplitude-time point (e.g., see E-H as indicated by arrows). There are four distinct regions. Region I represents a separation of the two stimuli by ≥70 msec. In this region, the occurrence of a spike to the first input has little or no effect on the probability of generating a spike to the second. Region II represents amplitudes and time intervals in which the probability of generating an action potential to either input was 0. Region III represents those amplitude and time intervals in which the occurrence of a spike to the first stimulus significantly reduced the probability of spiking to the second stimulus. In region IV, the test stimulus was so large and at such a short interval that the occurrence of a spike to the first had no effect. E, Example of the spike probability curve for different-amplitude inputs with and without a preceding spike at an interval of 27 msec. F, At an interval of 80 msec, the generation of a spike has no effect. G, Effect for different intervals at a fixed amplitude of the test stimulus. At short intervals, there is significant temporal summation, which is reduced by the occurrence of an action potential. H, Example curves for 1.4 nA input at different intervals. There are two main effects. At short intervals, there is temporal summation, which overcomes the suppressive effect of a preceding spike. At long intervals, the preceding spike has no effect. Between these extremes, the occurrence of a spike reduces the probability of generating another action potential.

Figure 11.

The addition of background noise reduces the spike shadow effect. A, Effect of the generation of an action potential on the subsequent probability of the generation of an additional action potential in response to the second injection of the depolarizing phase of a 40 Hz sine wave at the indicated amplitude and latency (see Fig. 10 D). Graphs are generated by subtracting the probability of generating an action potential with a preceding spike (to a stimulus injected at time 0) from the probability curves when there was no preceding spike. B-D, Increasing the level of added membrane variance (noise) results in a marked decrease in the spike shadow effect, thereby allowing the cell to recover quickly from having generated a preceding action potential. Noise 1, 2, and 4 correspond to 1, 2, and 4 nS of added noise. The same added background conductance was present in all four conditions (A-D) with only the variance changing between conditions. Thus, the average membrane potential and conductance are not different in the four cases shown.

Figure 12.

Noise can facilitate the cross-correlation between the injected 40 Hz sine wave current and the resulting average spike rate histogram. A, Spike histogram over 350 trials of the action potential discharge of a cortical pyramidal cell in response to the intracellular injection of the depolarizing phase of a 40 Hz sine wave of various amplitudes. The spikes show little variation from trial to trial in their timing. B, The addition of background noise (4 nS SD of both excitatory and inhibitory components) results in a smoothing of the discharge rate of the pyramidal cell, as well as increased responsiveness to previously subthreshold stimuli (e.g., 0.8 nA). Note that the addition of noise also lessens the phase lag of spike times to threshold stimuli (e.g., 1 nA). C, Plot of the number of spikes generated per trial versus the current amplitude in the presence and absence of added noise. D, The addition of membrane noise results in a significant increase in spike jitter (SD of spike times) and an increase in the peak cross-correlation between the average spike rate and current amplitude.