Local field potentials primarily reflect inhibitory neuron activity in human and monkey cortex

Abstract

The local field potential (LFP) is generated by large populations of neurons, but unitary contribution of spiking neurons to LFP is not well characterised. We investigated this contribution in multi-electrode array recordings from human and monkey neocortex by examining the spike-triggered LFP average (st-LFP). The resulting st-LFPs were dominated by broad spatio-temporal components due to ongoing activity, synaptic inputs and recurrent connectivity. To reduce the spatial reach of the st-LFP and observe the local field related to a single spike we applied a spatial filter, whose weights were adapted to the covariance of ongoing LFP. The filtered st-LFPs were limited to the perimeter of 800 μm around the neuron, and propagated at axonal speed, which is consistent with their unitary nature. In addition, we discriminated between putative inhibitory and excitatory neurons and found that the inhibitory st-LFP peaked at shorter latencies, consistently with previous findings in hippocampal slices. Thus, in human and monkey neocortex, the LFP reflects primarily inhibitory neuron activity.

Figure 1. Spikes of single neurons are associated with spatially diffuse and non-causal LFP patterns.

(A) Local field potentials and spikes were measured in the premotor cortex of a macaque monkey (top) and temporal cortex of human subjects (bottom) using the Utah arrays. (B) LFP (top, subset of LFPs recorded simultaneously from macaque premotor cortex) and spikes (bottom, subset of neurons) obtained from Utah array. Neurons were classified into regular spiking (bottom, blue) and fast spiking (red) types based on spike waveform. (C) Spatio-temporal spike-triggered LFP average (st-LFP) in human temporal cortex. Top: Average of the st-LFPs (band-pass filtered 15–300 Hz, average of 7558 spike-triggered segments) from the electrodes neighbouring with the trigger neuron. Middle: A color map of st-LFP amplitudes at selected time lags around the spike. The values for missing electrodes were replaced with the average of the neighbouring electrodes. Bottom: The st-LFP from all valid electrodes of the array plotted in time (plotting window adjusted to the gray-shaded segment in top panel). The st-LFP at the neuron position (black rectangle) was replaced with the spike waveform (amplitude normalised). Most st-LFPs express non-causal components preceding the spike (spike onset shown with vertical dotted line). The gray-shaded area represents 95% confidence intervals calculated from jittered spikes (1000 repetitions, gaussian jitter 100 ms). (D) st-LFPs triggered on spikes of a single neuron (same as shown in (C)) and averaged for all electrodes separated by the same distance from the neuron (3 distances: 0.4 mm, 0.8 mm, 1.2 mm shown from left to right, note the change of the amplitude scales). For the details of the calculation see Supplementary Fig. 9. The 95% confidence intervals (gray-shaded area) were calculated as ±1.96 × s.e.m. (E) The st-LFP trough amplitude as a function of the distance from the neuron to the LFP electrode. The data points were fitted with an exponential A exp(−x/λ) + C, where x is the distance and λ is the space constant (fitted value ± SD in the top-right corner).

Figure 2. Spatial and temporal st-LFP components across neurons and subjects.

(A and B) Human subjects. (C) Monkey. Three left-most panels: st-LFPs triggered by spikes of regular spiking (RS, blue line) or fast spiking (FS, red line) neurons at 0.4, 0.8 and 1.2 mm from the trigger neuron. The st-LFPs were averaged both over neurons and electrodes. The trough latencies of RS and FS neurons (values given in bottom-right corner) are significantly different for short distances in one human subject and monkey. The shaded areas represent the 95% confidence intervals of the respective st-LFPs (±1.96 × s.e.m.). Right-most panel: The decay of trough amplitudes with distance. The RS/FS space constants λ (coefficient ± SD given in top-right corner) as determined by fitting an exponential function (solid line) to the estimated amplitudes (circles) were significantly different (tested using t-test) for second human subject (B) and monkey (C). n.s.: not significant, *p < 0.05, **p < 0.01, ***p < 0.001.

Figure 3. The focal LFP contribution of fast spiking (FS) and regular spiking (RS) neurons recovered by spatial decorrelation (whitening) of st-LFP.

(A) Spatial filters designed to decorrelate (whiten) the LFP signals. Colors show filter weights associated with each electrode. Inset: Scaled-up heatmap of weights to whiten a single st-LFP (single row of whitening matrix). (B) Whitened st-LFPs (wst-LFPs) of a single neuron (green) compared with the non-whitend st-LFP (black). Right: close-up of the whitened st-LFPs enclosed in black rectangle. Most st-LFPs are suppressed after spatial whitening and only st-LFPs directly adjoining the neuron are conserved. (C) Spatial maps of whitened st-LFP (wst-LFP, compare with Fig. 1C). (D–F) The population-averaged wst-LFPs for human subject 1 (D), human subject 2 (E) and monkey (F). Three panels from left: wst-LFPs averaged across neurons and electrodes at three distances from the trigger neuron (0.4 mm, 0.8 mm, 1.2 mm; for details see legend of Fig. 2). Right-most panel: wst-LFP trough amplitude as a function of the distance between the neuron and the LFP electrode (cf. Fig. 1E). The star indicates significant differences in the trough amplitude between RS and FS neurons at the respective distance. n.s.: not significant, *p < 0.05, **p < 0.01, ***p < 0.001.

Figure 4. Propagation of st-LFP and wst-LFP in human and monkey.

(A) Propagation of the non-whitened st-LFP across the electrode array. From left to right: single-neuron st-LFPs averaged over all electrodes with the same distance from the trigger neuron; the latency of averaged st-LFP troughs (dots) and the linear fit (solid line); the latency of st-LFPs from each electrode as a heatmap (blank squares are due to missing electrodes). The latencies increase with the distance supporting the hypothesis of spike-evoked LFP propagation. (B) The propagation of population-averaged wst-LFPs in human. Latency vs. distance, separately for each neuron type, is plotted against the distance from the trigger neuron (circles) and fitted with a linear function (solid line). The speeds of wst-LFPs propagation for RS and FS neurons were calculated from the inverse of the slope (number at the top). The speed differences tested using bootstrap method could not be shown significantly different (bracket next to speed values). (C) Second human subject. (D) Monkey. The wst-LFP traces for which the latencies were calculated are shown in Fig. 3 and Supplementary Fig. 3. n.s.: not significant, *p < 0.05, **p < 0.01, ***p < 0.001.