Different origins of gamma rhythm and high-gamma activity in macaque visual cortex

Abstract

During cognitive tasks electrical activity in the brain shows changes in power in specific frequency ranges, such as the alpha (8-12 Hz) or gamma (30-80 Hz) bands, as well as in a broad range above ∼80 Hz, called the high-gamma band. The role or significance of this broadband high-gamma activity is unclear. One hypothesis states that high-gamma oscillations serve just like gamma oscillations, operating at a higher frequency and consequently at a faster timescale. Another hypothesis states that high-gamma power is related to spiking activity. Because gamma power and spiking activity tend to co-vary during most stimulus manipulations (such as contrast modulations) or cognitive tasks (such as attentional modulation), it is difficult to dissociate these two hypotheses. We studied the relationship between high-gamma power, gamma rhythm, and spiking activity in the primary visual cortex (V1) of awake monkeys while varying the stimulus size, which increased the gamma power but decreased the firing rate, permitting a dissociation. We found that gamma power became anti-correlated with the high-gamma power, suggesting that the two phenomena are distinct and have different origins. On the other hand, high-gamma power remained tightly correlated with spiking activity under a wide range of stimulus manipulations. We studied this relationship using a signal processing technique called Matching Pursuit and found that action potentials are associated with sharp transients in the LFP with broadband power, which is visible at frequencies as low as ∼50 Hz. These results distinguish broadband high-gamma activity from gamma rhythms as an easily obtained and reliable electrophysiological index of neuronal firing near the microelectrode. Further, they highlight the importance of making a careful dissociation between gamma rhythms and spike-related transients that could be incorrectly decomposed as rhythms using traditional signal processing methods.

Conflict of interest statement

The authors have declared that no competing interests exist.

Figures

Figure 1. Dissociation of the gamma rhythm and high-gamma activity by manipulating stimulus size.

(A) Average multiunit recorded from a single site in Monkey 1 during the presentation of a static grating (0 to 400 ms) at six different sizes, shown in different colors. The inset shows the average firing rate between 200 and 400 ms, indicated by a thick black line on the abscissa. (B) Time-frequency energy difference plots (in dB) showing the difference in energy from baseline energy (−300 to 0 ms, 0 denotes the stimulus onset, difference computed separately for each frequency) for the smallest (radius of 0.3°, left panel), medium (1.14°, middle), and largest (2.4°, right) sizes. The gamma rhythm at ∼50 Hz increases with size, while the high-gamma activity above the gamma band decreases with size. (C) The LFP energy between 200 and 400 ms (denoted by a thick black line on the abscissa in B) as a function of frequency for the six sizes, whose radii are listed in the legend. The black line shows the LFP energy in the baseline period. (D–F) and (G–I) show corresponding population responses of 15 and 104 sites from Monkeys 1 and 2, respectively. For (D) and (G), the responses are normalized by dividing by the maximum firing rate for each site. Monkey 2 showed two distinct gamma bands at ∼50 and ∼90 Hz.

Figure 2. Correlations between power and firing rates have different signs in gamma versus high-gamma bands.

(A) Average relative change in power between 200 and 400 ms from baseline power (difference between the colored traces and the black trace in Figure 1F and 1I), for 15 and 104 sites in Monkeys 1 (left panel) and 2 (right panel). Radii are listed again in the legend for clarity. (B) Spearman rank correlation between the six power values (one for each size) at each frequency and the six firing rates values, computed individually for each site and then averaged. Black and gray traces show the mean and SEM of 15 and 104 sites in the two monkeys. The correlation values significantly different from zero are shown in green (p<0.01, uncorrected) and red (p<0.05 with Bonferroni correction).

Figure 3. Changes in power with stimulus size are observed even when firing rates are negligible.

(A) Average firing rate of 30 and 10 sites in Monkeys 1 (left column) and 2 (right column), for which less than 0.5 spikes/s were obtained between 200 and 400 ms. (B) Difference in power between 200 and 400 ms from baseline power (same format as Figure 2A) for these sites.

Figure 4. Correlations between the time-courses of firing rate and LFP power in different frequency bands.

(A) The mean time-frequency energy difference plot (in dB) of 15 sites from Monkey 1, when the largest stimulus is presented. Same as the right panel of Figure 1E, except that the displayed frequency range is up to 500 Hz. The vertical colored lines in the right mark the four frequency bands used for analysis—alpha (8–12 Hz; magenta), gamma (30–80 Hz; dark green), high-gamma (102–238 Hz, excluding 118–122 Hz; light green), and 250–500 Hz (brown). (B) Panels in the left column show the relative change in LFP power in the four frequency bands (colored traces) for the largest stimulus, along with the relative change in firing rate (black trace, same for all panels). The Spearman rank correlation between the two traces is denoted in the upper-left corner. Panels in the right show the relative change in LFP power for different stimulus size (same color code as Figures 1 and 2, the orange trace is the same as the colored trace in the left column). (C, D) Same as (A, B) for 104 sites in Monkey 2.

Figure 5. Correlations between firing rate and LFP power in different frequency bands for stimuli with different temporal frequency profiles.

(A) The left panel shows the average time-frequency energy difference spectrum of 19 sites in Monkey 1 when the stimulus was presented with a counter-phasing temporal frequency of 2.5 Hz. The contrast profile is shown in red on top of the right panels. The right panels show the relative change in power in different frequency bands as well as in the firing rates, as a function of time. Same format as in Figure 4. The Spearman correlation values between the firing rate and power traces are shown in the top left corner. (B) Same as panel (A), for a temporal frequency of 5 Hz. (C, D) Same as (A, B) for 66 sites in Monkey 2.

Figure 6. Trial-by-trial Spearman correlation between firing rates and LFP power at different frequencies when stimulus conditions are identical.

(A) The median Spearman rank correlation between LFP power at different frequency bins (size of 25 Hz, computed in steps of 10 Hz) and firing rates, both computed between 200 and 400 ms after stimulus onset, for 15 sites in Monkey 1. The correlations were computed separately for each size, site, and orientation, so that the stimulus conditions were identical. The first column shows the median correlations during the pre-stimulus period (denoted “BL” for baseline). The remaining six columns represent the six stimulus sizes (denoted by the respective color below the x-axis). (B) Median Spearman correlation, computed for the four frequency bands used in Figures 4 and 5. Correlations significantly different from zero (p<0.05, Bonferroni corrected) as shown with asterisks. (C,D) Same as (A, B) for 104 sites in Monkey 2.

Figure 7. Spike-triggered average in time-frequency domain during baseline period.

(A) The mean spike-triggered average from spikes taken between 268 and 132 ms before stimulus onset, from the 14 sites for which at least 25 spikes were obtained. (B) Left panel shows the spike-triggered time-frequency average (STTFA), computed by averaging short 2-D segments of the time-frequency energy spectrum centered on the spikes. The middle panel shows the STTFA computed after first randomizing the spike times (called rSTTFA). The panel in the right shows the relative change in the time-frequency spectrum locked to the spike, computed by taking the difference between log(STTFA) and log(rSTTFA) (called the normalized STTFA, or nSTTFA). (C) The mean energy between −1 to 3 ms of the STTFA (black) and the rSTTFA (gray), as a function of frequency (upper plot). The difference between the two is shown in the lower panel (mean in black, SEM in gray). The values significantly different from zero are shown in green (p<0.01, uncorrected) and red (p<0.05 with Bonferroni correction). (D–F) Same as (A–C), for 103 sites in Monkey 2.

Figure 8. Spike-triggered average in time-frequency domain during stimulus presentations.

(A) Mean spike-triggered average of 15 sites for Monkey 1 for which at least 25 spikes were available between 232 and 368 ms after stimulus onset, for the six stimulus sizes. (B) The normalized STTFA (see text and Figure 7 for details) when a small (left), medium (middle), and large (right) stimulus was presented. (C) The difference between the mean energy between −1 and 3 ms of the STTFA and rSTTFA (same as the lower panel of Figure 7C), for the six stimulus sizes. The horizontal lines at the bottom indicate the “cutoff frequency” for each stimulus size (see text for definition). (D–F) Same as (A–C), for 94–103 sites in Monkey 2 for which at least 25 spikes could be obtained. The number of sites decreases from 103 to 94 because the firing rates decrease with increasing stimulus size.