(C) PLOS One [1]. This unaltered content originally appeared in journals.plosone.org.
Licensed under Creative Commons Attribution (CC BY) license.
url:https://journals.plos.org/plosone/s/licenses-and-copyright

------------



Traveling waves in the prefrontal cortex during working memory

['Sayak Bhattacharya', 'The Picower Institute For Learning', 'Memory', 'Department Of Brain', 'Cognitive Sciences', 'Massachusetts Institute Of Technology', 'Cambridge', 'Massachusetts', 'United States Of America', 'Scott L. Brincat']

Date: 2022-02 

Neural oscillations are evident across cortex but their spatial structure is not well- explored. Are oscillations stationary or do they form “traveling waves”, i.e., spatially organized patterns whose peaks and troughs move sequentially across cortex? Here, we show that oscillations in the prefrontal cortex (PFC) organized as traveling waves in the theta (4-8Hz), alpha (8-12Hz) and beta (12-30Hz) bands. Some traveling waves were planar but most rotated. The waves were modulated during performance of a working memory task. During baseline conditions, waves flowed bidirectionally along a specific axis of orientation. Waves in different frequency bands could travel in different directions. During task performance, there was an increase in waves in one direction over the other, especially in the beta band.

We found that oscillations in the prefrontal cortex form “traveling waves”. Traveling waves are spatially extended patterns in which aligned peaks of activity move sequentially across the cortical surface. Some traveling waves were planar but most rotated. The prefrontal cortex is important for working memory. The traveling waves changed when monkeys performed a working memory task. There was an increase in waves in one direction over the other, especially in the beta band. Traveling waves can serve specific functions. For example, they help maintain network status and help control timing relationships between spikes. Given their functional advantages, a greater understanding of traveling waves should lead to a greater understanding of cortical function.

Oscillatory activity in the prefrontal cortex has been linked with cognitive functions like working memory and attention but there has been little examination of whether they form spatio-temporal structures like traveling waves. Most studies have averaged oscillations across spatially distributed electrodes. This increases the “signal” of the oscillations but prohibits analysis of any spatial organization. Thus, we examined their spatial organization from microarray recordings in the PFC of monkeys performing a working memory task. This revealed that low-frequency (beta and lower) traveling waves are common in the PFC. We characterized their speed, direction and their patterns. They often rotated and changed direction during performance of a working memory task.

Traveling waves are of interest because they have a variety of useful properties for cognition, development, and behavior. They can create timing relationships that foster spike-timing-dependent plasticity and memory encoding [ 14 , 18 ]. They add information about recent history of activation of local networks [ 19 ]. They are thought to help “wire” the retina [ 20 ] and cortical microcircuits during development [ 21 ]. Their functional relevance in the adult brain is suggested by observations that traveling wave characteristics can be task-dependent and that they impact behavior. For example, EEG recordings have shown that alpha band waves reverse their resting state direction during sensory inputs [ 22 ]. Behavioral detection of weak visual targets improves when there are well organized low-frequency (5-40Hz) traveling waves in visual cortex vs when there is a weaker, “scattered” organization [ 23 ].

Thus far, most studies of neural oscillations have focused on what we can call “standing wave” properties (e.g., power of and coherence between oscillations at different cortical sites), ignoring any organization of where and when the peaks and troughs of activity appear. However, there is mounting evidence for such organization. Oscillations can take the form of “traveling waves”: Spatially extended patterns in which aligned peaks of activity move sequentially across the cortical surface [ 12 , 13 ]. This apparent movement of the amplitude peak is facilitated by the existence of a phase gradient along a particular direction, along which the movement occurs. Traveling waves have most often been reported in the lower-frequency bands (<30 Hz). Examples include beta-band (15–30 Hz) traveling waves in motor and visual cortices [ 14 , 15 ] and theta band (3–5 Hz) traveling waves in the hippocampus [ 16 , 17 ].

Oscillatory dynamics have been linked to a wide range of cortical functions. For example, higher frequency (gamma, >40 Hz) power (and spiking) increases during sensory inputs (and their maintenance) and during motor outputs [ 1 – 5 ]. Gamma power is anti-correlated with lower frequencies (alpha/beta, 8–30 Hz), whose power is often higher during conditions requiring top-down control (e.g., when attention is directed away, or an action is inhibited) [ 6 – 8 ]. Such observations have led to a theoretical framework in which oscillatory dynamics regulate neural communication [ 9 – 11 ].

Results

Two animals performed a delayed match-to-sample task (Fig 1A). They maintained central fixation while a sample object (one of eight used, novel for each session) was briefly shown. After a two-second blank memory delay (with maintained fixation), two different objects (test screen, randomly chosen from the eight) were simultaneously presented at two extrafoveal locations. Then the animals were rewarded for holding fixation on the object that matched the sample. Two 8x8 multi-electrode “Utah” arrays, one in each hemisphere, were used to record local field potentials (LFPs) from the dorsal pre-frontal cortex (dlPFC). All data is from correctly performed trials. We analyzed data from 14 experimental sessions, five from one animal (Animal/Subject 1) and nine from the other (Animal/Subject 2).

PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 1. (A) Delayed-match-to-sample (DMTS) working memory task. The subject fixated at the center for 0.5 s (leftmost panel) before an object (one of eight) were presented at the center for another 0.5 s. There was a 2 s blank memory delay after which a test screen was presented at extrafoveal locations. The subject had to saccade towards the remembered object and hold fixation there (arrow, rightmost panel). (B) Filtered LFP power trends for all electrodes (four arrays), trial-averaged and shown across time with the lines denoting the major trial epochs. Four frequency ranges were chosen–theta (4–8 Hz), alpha (8–12 Hz), beta (12–30 Hz) and gamma (40–120 Hz), with the dots above the curve denoting if the LFP power at that instant is significantly (p<0.01) different from baseline (0.5 before fixation started). (C) Theta LFP organization across the right hemisphere array of Subject 1 of a particular trial. Each tile denotes a recording site on the array (8x8 total) with the color denoting the LFP amplitude at that site. The array position with respect to anatomical brain landmarks is overlaid. Each panel denotes the organization at an instant in time. The black arrow in the second panel indicates the direction of movement of the high amplitude LFPs with time. (D) Voltage traces from the 8 adjacent electrodes (color graded by position) in a row of the array in (C–dashed line). The circles mark the peak positions of the oscillation cycles with the dashed line indicating how the peaks shift gradually in time and space. (E) Phase maps from the corresponding panels in (C) with the color on each tile denoting the phase of the LFP oscillation cycle on that recording site. https://doi.org/10.1371/journal.pcbi.1009827.g001

Waves travelled in preferred directions The waves did not travel in random directions. To check if there were preferred directions of travel along each array, we leveraged a property of the circular-circular coefficient. The coefficient could discriminate between waves in opposing directions (red and blue regions respectively, Fig 3A) along a particular axis. Waves directed towards the positive half (red arrows, Fig 3A) had ρ c >0 and vice versa (blue arrows). The orientation of the axes splitting the positive and negative regions depended on the net direction of the rotation map chosen and hence differed with the chosen point on the array (Fig 3A, left vs right, the chosen point shown in yellow). We chose points such that we could split the array into polar segments of around 10–15 degrees each. We confirmed these properties with simulated data (see Materials and Methods). To test for statistical significance of the correlation obtained, we calculated a ρ c threshold beyond which the phase organization exceeded chance (p<0.01, ρ c >0.3, random permutation test). This approach allowed us to measure wave directions but did not discriminate between planar and rotating waves. Fig 3E shows the placement of one of the arrays relative to the principal and arcuate sulci, with polar histograms of observed wave directions overlaid. Some wave directions were more common than others. This was consistent across the frequency bands. We reoriented the data from all four arrays such that their most preferred direction (for each frequency band) was along the horizontal axis (Fig 3F). The wave counts in each direction were then averaged across trials. Clear directional preferences were seen across all arrays and all frequency bands (theta to beta). Note that the directional preference remained consistent across trials epochs including the baseline (dashed lines). In other words, there were preferred “default” directions. Task performance increased or decreased the probability of waves traveling in those directions. A degree of bimodality was also observed for all frequencies. In addition to the preferred direction at 0° (by definition), there was a secondary preference for directions around 180°. Thus, there was typically a preferred axis of wave propagation (dashed arrow, Fig 3F) with an additional preference for one direction over the other along that axis.

Task-related changes in wave direction Though task demands did not change the preferred axis of wave motion, we found that it could alter the balance between opposite directions along that axis. To determine this, we again used the circular-circular correlation analysis. For this purpose, we examined correlation values (ρ c ) with the rotation map around the point that showed the maximum number of waves for each array, adjusting such that the task-enhanced direction had positive ρ c . This approach included both planar and rotating waves. Oppositely directed waves showed opposite signs. Fig 4A shows the distribution of ρ c values, averaged across all trials and all four arrays in both animals. Results are shown separately for three frequency bands, and for three task epochs (green lines), with each compared to the pre-trial baseline (red lines). The ρ c at each time instant was classified as a wave if it exceeded the value expected by chance (indicated by shaded areas of Fig 4A, see Materials and Methods). Alpha and theta waves showed unimodal ρ c histograms centered around zero during baseline and delay epochs. Presentation of the sample or test-array shifted the histogram peak out of the shaded area (indicating increase in wave incidence) but did so in one direction over the opposite. This could be seen in theta and alpha frequency bands (Fig 4A). PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 4. (A) Histograms quantifying the correlation values seen on average in each trial during the sample, delay, and test-onset intervals (all 0.5 s in length), combined across arrays. A correlation value to the right of the shaded region (positive) denotes waves in a particular direction, while the left means the opposite direction. The shaded region denotes the “chance zone” where no conclusion regarding wave direction can be made. For each frequency range (each row), the red histogram corresponds to the correlations observed in baseline conditions (0.5 s pre-fixation), while the green histogram corresponds to the correlations observed in that epoch (0.5 s). The blue dots denote if the two are significantly different from each other (p<0.01). (B) Quantification of the difference between the green and red curves in (A) for all 0.5 s intervals during each trial. The red line shows difference from baseline for the positive wave direction, while the blue line for the negative wave direction. https://doi.org/10.1371/journal.pcbi.1009827.g004 Beta waves showed prominent bidirectionality throughout the trial (Fig 4A, bottom row). During both the baseline (red line) and task performance (green line), the distribution of ρ c values for waves in the beta band were bimodal with “bumps” on the ends of the distributions (outside the “chance zone”). This indicates waves in opposite directions. Relative to baseline, during task performance the “bump” on one end of the distribution rose while the other lowered, indicating an increase in waves in one direction and a decrease in the opposite. Baseline ρ c values (red lines) for the beta band skewed toward the negative direction indicating a baseline default bias for waves to travel in that direction. During the sample and delay epochs (green lines), waves skewed more toward positive ρ c values). After test screen presentation and the monkey’s behavioral response, this reversed back to a skew toward negative ρ c values, the baseline direction opposite of that seen in the task. This is illustrated in more detail in Fig 4B. It shows changes in wave direction bias (relative to baseline) over time. For theta and alpha waves, negative ρ c values were significantly increased from baseline during the sample (blue line, Fig 4B) and especially after the test screen appeared. Positive ρ c values also increased intermittently for these two frequency bands. The beta band showed the most consistent changes in wave direction preference with task performance (Fig 4B, bottom row). During sample presentation and the memory delay, there was an increase of positive ρ c values and a decrease in negative ρ c values indicating a consistent shift toward waves flowing in one direction over the other. After test screen presentation and the animal’s behavioral response (i.e., post-test), this shift ended. There was a decrease in positive, and an increase in negative ρ c values resulting in the mix of the two directions seen during baseline. For this representation, we adjusted the coefficient points such that the enhanced wave direction had positive ρ c for all frequencies . However, it is important to note that although the waves in different frequency bands had similar preferred direction axes (Fig 3F), this did not mean that the waves in different bands traveled in the same directions at the same time, as part of a multiband wave. S3 Fig shows the correlation coefficient calculated around the central point (4,4) for the left dlPFC array of Animal 2. As can be seen in the histograms (S3A Fig), theta waves preferred the negative direction during baseline, sample presentation and delay (higher histogram bump towards negative ρ c values). By contrast, beta waves preferred the opposite direction, i.e., the higher histogram bump was towards positive ρ c values. This is also evident in the quantification of the directional increase during task performance from baseline (S3B Fig). While the negative direction was enhanced during the task for theta waves, positive values were enhanced for beta waves.

[END]

[1] Url: https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1009827

(C) Plos One. "Accelerating the publication of peer-reviewed science."
Licensed under Creative Commons Attribution (CC BY 4.0)
URL: https://creativecommons.org/licenses/by/4.0/


via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/