Broadband shifts in local field potential power spectra are correlated with single-neuron spiking in humans

Abstract

A fundamental question in neuroscience concerns the relation between the spiking of individual neurons and the aggregate electrical activity of neuronal ensembles as seen in local field potentials (LFPs). Because LFPs reflect both spiking activity and subthreshold events, this question is not simply one of data aggregation. Recording from 20 neurosurgical patients, we directly examined the relation between LFPs and neuronal spiking. Examining 2030 neurons in widespread brain regions, we found that firing rates were positively correlated with broadband (2-150 Hz) shifts in the LFP power spectrum. In contrast, narrowband oscillations correlated both positively and negatively with firing rates at different recording sites. Broadband power shifts were a more reliable predictor of neuronal spiking than narrowband power shifts. These findings suggest that broadband LFP power provides valuable information concerning neuronal activity beyond that contained in narrowband oscillations.

Figure 1.

LFP power and neuronal firing time series. Each box details the activity in one 500 ms epoch. a, This panel illustrates how various features of the LFP change over time. In each epoch, the black lines indicate the overall LFP power spectrum, brown lines indicate robust-fit lines, and the horizontal red lines indicate mean broadband powers. b, This panel illustrates changes in neuronal firing rate concurrent with changes in the LFP power spectrum. Black vertical ticks represent the times when individual spikes occurred, dark blue lines indicates the smoothed firing rate (see Materials and Methods), and horizontal blue lines indicate mean firing rates in each epoch. Mean broadband power is shown in b (horizontal red lines) on a different scale (indicated at right).

Figure 2.

A representative neuron exhibiting a positive correlation between firing rate and broadband LFP power. a, Broadband power and firing rate for the neuron analyzed in Figure 1. Each 500 ms epoch of the recording session is represented by one colored dot. The color of each dot represents its relative firing rate. Warm colors depict epochs with high firing rates, and cool colors indicate epochs with low firing rates. The dashed black line shows an ordinary least-squares regression to these data. b, Average LFP power spectra for epochs with different firing rates. The same color scheme is used in both panels. As firing rate increases, the power spectrum exhibits a positive shift at all observed frequencies. The thickness of each line represents ±1 SEM.

Figure 3.

Categories of observed neurons. Each box corresponds to one of the nine possible categories into which our regression framework assigned each neuron in our dataset. The right and left columns correspond to significant positive and negative broadband shifts, respectively. The top and bottom rows correspond to significant positive and negative narrowband shifts, respectively. The middle column and middle row correspond to neurons that did not show significant broadband and narrowband shifts, respectively. The proportion of neurons placed into each category is indicated in the upper-right corner of each box. Proportions that were significantly greater than the estimated false-positive rate for each category (see Materials and Methods) are filled in black. Each box contains power spectra (in the same format as Fig. 2b) of an example neuron from the corresponding category. The brain region of each example neuron is indicated in the lower left corner of each box.

Figure 4.

Broadband-shift neurons throughout the brain. Each row shows two example neurons from a single brain region. Within each row, each plot illustrates the power spectra from an example neuron, in the same format as Figure 2b. (Each of these neurons were categorized as positive broadband-shift neurons, corresponding to the middle-right box of Fig. 3.)

Figure 5.

Percentage of positive broadband-shift neurons observed in each brain region. Dark gray bars correspond to neocortical regions: frontal cortex (Fr) and posterior cortex (Cx). Light gray bars correspond to medial temporal lobe structures: amygdala (Amyg), hippocampus (Hippo), and parahippocampal region (Par). Positive broadband-shift neurons were more concentrated in the medial temporal lobe than in the neocortex. The dotted horizontal black line marks the false-detection rate for positive broadband-shift neurons. Error bars indicate 95% confidence intervals based on the binomial distribution.

Figure 6.

LFP components that predict firing rate. Dark gray bars indicate the percentage of neurons in each region that exhibited positive correlations between firing rate and a particular LFP feature; light gray bars show the percentage of neurons in each region that exhibited negative correlations. The bars on the left indicate the proportions of neurons whose firing rates were correlated with power in each narrow frequency band: delta (2–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), beta (12–30 Hz), and gamma (30–150 Hz). Each neuron may be counted in at most one direction (i.e., either positive or negative) per narrow frequency band. The bars on the right indicate the proportions of neurons whose firing rates were correlated with broadband power (i.e., broadband-shift neurons).

Figure 7.

Average power spectra for different subsets of neurons. a, Average power spectra for all 2030 neurons in our dataset. The power spectra are normalized by subtracting the mean of the lowest firing rate power spectrum (dark blue) from all five curves in each panel. The thickness of each line represents ±1 SEM. As in Figure 2, cooler colors correspond to lower firing rates, while warmer colors correspond to higher relative firing rates. Each neuron contributes exactly once to each curve, at each frequency. b, Average power spectra for the 759 broadband-shift neurons. c, Average power spectra for the 1271 neurons that did not exhibit the broadband-shift effect.