Fig 1.
EPSP amplitudes and paired-pulse ratios of excitatory synaptic connections formed with L2/3 pyramidal neurons in barrel cortex.
A Example of recorded regular-spiking L2/3 neuron in mouse barrel cortex visualized through post-hoc biocytin histology. Blue dots indicate locations of successful extracellular stimulation, blue pipette signifies extracellular stimulation electrode. The neuron’s responses to stimulation at three different positions (labeled 1–3) are shown in B. B Somatic voltage recordings following 20 ms paired-pulse stimulation in the locations indicated by numbers. Grey traces, individual trials; black traces, average response; paired-pulse ratios (PPR) indicated. For timing of extracellular stimulation pulses (dashed lines), note the electrical stimulation artifact in somatic voltage responses. C Scatter plot showing, for all synaptic connections, the response to the second pulse versus the response to the first pulse (corresponding to the EPSP amplitude) of the paired-pulse stimulation paradigm. Large dots, average for each connection (n = 74); small dots, six randomly selected single-trial responses for each connection (n = 444). Data points below diagonal indicate depressing synaptic connections, dots above diagonal indicate facilitating connections. Voltage traces, same as traces 2 and 3 in B with identical scale bars. D Distribution of average EPSP amplitudes, histograms were fit with lognormal functions (R2, goodness of fit). E Distribution of average paired-pulse ratios, histograms were fit with Gaussian functions (R2, goodness of fit). F Left, scatter plot showing relationship between EPSP amplitude and paired-pulse ratio; large dots, average for each connection (n = 74; light green, facilitating connections; dark green, depressing connections); small dots, six randomly selected single-trial responses for each connection. The line was fit using linear regression, Pearson correlation results for single-trial data indicated. Right, comparison of paired-pulse ratios that were binned into ‘weak’ (EPSP < 2 mV) and ‘strong’ (EPSP > 2 mV) synaptic connections (parametric Welch’s t test).
Fig 2.
Excitatory synaptic connections formed with L2/3 neurons do not exhibit a systematic clustering of EPSP amplitude and short-term plasticity.
A Top, distribution of EPSP amplitudes recorded across all regular-spiking neurons (population distribution). Bottom, distributions of EPSP amplitudes measured across the 8 neurons, for which at least 5 synapses were found (cell distributions). N, number of synapses recorded per cell; p, non-parametric Kolmogorov-Smirnov test between each cell distribution and the population distribution, medians are indicated. Note that the synapses formed with the postsynaptic cell in any given experiment were removed from the respective population distribution that the cell was compared to. B Same analyses for short-term plasticity data, panel layout as in A; light green, facilitating synaptic connections; dark green, depressing connections, means are indicated. Note that the synapses formed with the postsynaptic cell in any given experiment were removed from the respective population distribution that the cell was compared to. C Estimation of the effect sizes that were detectible across the experimental series at the 5% significance level.
Fig 3.
Synaptic transmission between identified L2/3 pyramidal neurons measured with paired whole-cell recordings.
A Example of a synaptically connected pair of L2/3 pyramidal neurons in mouse barrel cortex visualized through post-hoc biocytin histology. B Scatter plot showing, for all synaptic connections, the response to the second pulse versus the response to the first pulse (corresponding to the EPSP amplitude) of the paired-pulse stimulation paradigm. Large dots, average for each connection (n = 22); small dots, six randomly selected single-trial responses for each connection (n = 132). Data points below diagonal indicate depressing synaptic connections, dots above diagonal indicate facilitating connections. C Top, distribution of average EPSP amplitudes recorded between connected L2/3 neurons, the histogram was fit with a lognormal function. Bottom, distribution of average paired-pulse ratios recorded between L2/3 neurons, the histogram was fit with a Gaussian function; R2, goodness of fit. D Left, comparison of average EPSP amplitudes recorded with minimal stimulation (n = 74; same as in Figs 1D and 2A) and with paired recordings (n = 22). Right, comparison of average paired-pulse ratios recorded with minimal stimulation (n = 74; same as in Figs 1E and 2B) and with paired recordings (n = 22). Non-parametric Kolmogorov-Smirnov p-values are indicated. E Scatter plot showing relationship of EPSP amplitude and paired-pulse ratio for synaptic connections between L2/3 pyramidal neurons obtained with paired recordings. Large dots, average for each connection (n = 22; light green, facilitating connections; dark green, depressing connections); small dots, six randomly selected single-trial responses for each connection. The line was fit using linear regression, Pearson correlation results for single-trial data indicated.
Fig 4.
Default setup of the L2/3 leaky integrate-and-fire neuron model.
A Example of input spike trains fed to the model cell. Strong inputs (top) fired with higher frequencies and temporal correlation (color coded) compared to weak inputs (bottom). Vertical grey bands indicate the resulting spike timing in the model cell (same as in C). Some weak inputs did not spike in the depicted 200 ms time window because of their low firing rates. B Strong inputs were set to have larger EPSP amplitudes and corresponding short-term depression, while weak inputs were set to evoke smaller EPSPs with weak net short-term plasticity, in accordance with our in vitro recordings. C Simulated membrane potential of model neuron following activation with the input spike trains shown in A. D Left, EPSP amplitudes across the 270 input spike trains. Center, comparison of EPSP amplitudes of strong and weak inputs (median, 25–75% percentile, and ranges are indicated). Right, same data plotted as histogram. E Left, 20 ms paired-pulse ratios across the 270 input spike trains. Center, comparison of paired-pulse ratios of strong and weak inputs (median, 25–75% percentile, and ranges are indicated). Right, same data plotted as histogram. F Left, Pearson correlation coefficient between the 270 input spike trains and the template spike train that was used to generate the pairwise correlation structure (see Methods); color code as in A, B. Center, comparison of correlation of strong and weak inputs with template spike train (median, 25–75% percentile, and ranges are indicated). Right, same data plotted as histogram. G Left, firing rates of the 270 input spike trains. Center, comparison of firing rates of strong and weak inputs (median, 25–75% percentile, and ranges are indicated). Right, same data plotted as histogram.
Fig 5.
Uncorrelated activity of weak inputs enhances information transfer of strong synaptic inputs.
A Schematic of model setup with weak inputs removed. B Example spike train of the model cell in its default setup (grey), when weak inputs are entirely removed (orange), and when weak inputs are replaced by a more depolarized Vrest (red). C Schematic of model setup with strong inputs removed. D Example spike train of the model cell in its default setup (grey) and when strong inputs are removed (purple). E Pearson correlation coefficients of the 270 input spike trains with the output spike train of the model cell. Results of three model setups are shown: default simulation (grey; all inputs, as in Fig 4) and setups introduced in A (orange, weak inputs removed; red, weak inputs replaced with depolarized Vrest). Dots indicate means, shaded regions indicate standard deviation of correlation coefficients for 100 runs of the simulation. F Top, Pearson correlation coefficients between the strong synaptic inputs and the output spike train of the model neuron for the default simulation and setups introduced in A. Bottom, output firing rate of model cell for the default simulation and the setups introduced in A-C. (Data are averages across 100 simulation runs; median and 25–75% percentile indicated; non-parametric Kolmogorov-Smirnov test, * p < 0.05.) G Probability of output spiking as a function of the number of coincident spikes across all input spike trains (grey, default simulation; red, weak inputs replaced with depolarized Vrest).
Fig 6.
Temporal correlation and firing rates primarily determine output spiking, synapse strength enhances responsiveness.
A Schematic of model setup with shuffled EPSP amplitudes; the relationship of EPSP amplitude and short-term plasticity was maintained. B Example spike train of the model cell in its default setup (grey) and with shuffled EPSP amplitudes (blue). C Pearson correlation coefficients of the 270 input spike trains with the output spike train of the model cell. Results of two model setups are shown: default simulation (as in Fig 4) and setup introduced in A. Dots indicate means, shaded regions indicate standard deviation of correlation coefficients for 100 runs of the simulation. D Top, Pearson correlation coefficients between the strong synaptic inputs and the output spike train of the model neuron for the default simulation and setup introduced in A. Bottom, output firing rate of model cell for the default simulation and the setup introduced in A. (Data are averages across 100 simulation runs; median and 25–75% percentile indicated; non-parametric Kolmogorov-Smirnov test, * p < 0.05.) E Probability of output spiking as a function of the number of coincident spikes across all input spike trains (grey, default simulation; blue, model setup with shuffled EPSPs).
Fig 7.
Short-term plasticity balances the computational effects of strong and weak inputs.
A Schematic of model setup with short-term plasticity mechanisms removed, i.e., all spike trains exhibit a paired-pulse ratio of 1. B Example spike train of the model cell in its default setup (grey) and when short-term plasticity mechanisms are removed (light green). C Schematic of model setup with short-term plasticity mechanisms removed and the weak inputs removed in addition. D Example spike train of the model cell in its default setup (grey) and when short-term plasticity mechanisms and weak inputs are removed (dark green). E Pearson correlation coefficients of the 270 input spike trains with the output spike train of the model cell. Results of three model setups are shown: default simulation (as in Fig 4) and setups introduced in A, C. Dots indicate means, shaded regions indicate standard deviation of correlation coefficients for 100 runs of the simulation. F Top, Pearson correlation coefficients between the strong synaptic inputs and the output spike train of the model neuron for the default simulation and setups introduced in A-C. Bottom, output firing rate of model cell for the default simulation and the setups introduced in A-C. (Data are averages across 100 simulation runs; median and 25–75% percentile indicated; non-parametric Kolmogorov-Smirnov test, * p < 0.05.) G Probability of output spiking as a function of the number of coincident spikes across all input spike trains for the default simulation (grey) and the setups introduced in A-C (light green and dark green, respectively).
Table 1.
Parameter sets Θ for strongly depressing and strongly facilitation synapses, adopted from [34].