The somatosensory cortex receives information about motor output

Abstract

During voluntary movement, the somatosensory system not only passively receives signals from the external world but also actively processes them via interactions with the motor system. However, it is still unclear how and what information the somatosensory system receives during movement. Using simultaneous recordings of activities of the primary somatosensory cortex (S1), the motor cortex (MCx), and an ensemble of afferent neurons in behaving monkeys combined with a decoding algorithm, we reveal the temporal profiles of signal integration in S1. While S1 activity before movement initiation is accounted for by MCx activity alone, activity during movement is accounted for by both MCx and afferent activities. Furthermore, premovement S1 activity encodes information about imminent activity of forelimb muscles slightly after MCx does. Thus, S1 receives information about motor output before the arrival of sensory feedback signals, suggesting that S1 executes online processing of somatosensory signals via interactions with the anticipatory information.

Fig. 1. Simultaneous recording of cortical and peripheral activities showed movement-related modulation of these activities.

(A) An example of simultaneous recording of three trials by Monkey T. Top row and second row: Power spectrograms of the respective electrodes in M1 and S1. Third row: EMG signal from the shoulder muscle. Fourth row: The instantaneous firing rate of an ensemble of peripheral afferents. Bottom row: Three forelimb joint angles along the extension-flexion axis (shoulder, solid line; elbow, dashed line; wrist, dotted line). (B) Modulations of cortical and peripheral activity in Monkey T aligned to movement onset. Top and second: Modulations of high-γ activity in M1 and S1. Third: EMG signals of 12 forelimb muscles. Fourth: Instantaneous firing rate of a neural ensemble of peripheral afferents. Bottom: Six forelimb joint angles (shoulder, solid lines; elbow, dashed lines; wrist, dotted lines). Thin lines represent the activity in each electrode (M1 and S1), units (afferent), and muscles, and thick lines represent their respective averages. The vertical solid, dotted, and dashed lines represent times of the onset of movement, pulling the lever, and the end of the movement, respectively. Arrowheads represent the peak times of respective activities. AU, arbitrary units.

Fig. 2. Combined activities in MCx and afferents account for S1 activity.

(A) Reconstruction of S1 activity using combined activities in MCx and peripheral afferents of Monkey T. Examples of two successive movements. Black line, the observed activity in S1; red line, the reconstructed activity; R, correlation coefficient between the observed and reconstructed activities; vertical solid lines, movement onset; vertical dashed lines, end of movement. (B) Mean decoding accuracy pooled across ECoG electrodes in S1. The correlation coefficient between the observed and reconstructed traces from the recorded data compared with the correlation coefficient between the observed traces and the traces reconstructed from random shuffling of activity (Monkey T: mean, n = 16 signals; Monkey C: mean, n = 12 signals; *P < 0.001). Superimposed bars, mean. (C) Average modulation of the observed S1 activity in Monkey T (S1, gray) and the reconstruction using combined activities in MCx and peripheral afferents (MCx + Afferent, purple) aligned to movement onset. Vertical solid lines, movement onset; vertical dashed lines, average time of end of movement; solid lines, means; shading, SD. (D) Variance accounted for (VAF) between the observed and reconstructed traces throughout the duration of S1 modulation (−100 to 1400 ms around movement onset for Monkey T, −100 to 1500 ms for Monkey C).

Fig. 3. MCx and afferent activities account for S1 activity in temporally different ways.

(A) Average modulation of the observed S1 activity in Monkey T (S1; gray), the reconstruction using combined activities in MCx and peripheral afferents (MCx + Afferent; purple), and each component in the reconstruction (MCx component, blue; Afferent component, green) aligned to movement onset (from the same session in Fig. 2C). Vertical solid line, movement onset; vertical dashed line, average time of end of movement; solid lines, mean; shading, SD. (B) Magnification of (A). Horizontal dashed line, threshold for the onset of activity; arrows, onset times. (C) Onset times of the observed S1 activity (gray), of the reconstruction using combined activities in MCx and peripheral afferents (purple), and of each component (MCx component, blue; Afferent component, green) (Monkey T: mean, n = 16 signals; Monkey C: mean, n = 12 signals: *P < 0.01). n.s., not significant. Superimposed bars, mean. (D) The areas above the baseline of the observed S1 activity (S1, gray), the reconstruction using combined activities of MCx and peripheral afferents (MCx + Afferent, purple), and each component (MCx component, blue; Afferent component, green) in 100-ms sliding time windows. Dashed line, average time of end of movement; error bar, SE.

Fig. 4. MCx and afferent activities account for S1 activity in spatially different ways.

(A) Average modulation of the observed activity in S1 electrodes of Monkey T (gray) and respective MCx (blue) and Afferent (green) components aligned to movement onset. Vertical lines, movement onset; solid lines, mean. (B) Principal component analysis (PCA) of the MCx and Afferent components in all S1 electrodes. Cumulative variance explained by the PCA components plotted against the number of feature dimensions used. Error bars, SE. (C) Coefficient of variation (CV) of areas above the baseline of the MCx and Afferent components in all S1 electrodes throughout the duration of S1 modulation (Monkey T: n = 21 sessions; Monkey C: n = 6 sessions; *P < 0.05). A pair including an outlier (CV of the Afferent component, 5.56; CV of the MCx component, 0.37 in Monkey C) is not used in the analysis. Superimposed bars, mean.

Fig. 5. Premovement activities in M1 and S1 encode muscle activity.

(A) Reconstruction of EMG profile (aligned to movement onset) of a hand muscle during the premovement period (−500 to 0 ms around movement onset) in Monkey C using the activity in M1 or S1 before movement onset (time = 0). Black and red traces show the observed and reconstructed EMG using the activity in M1 (left) or S1 (right), respectively. R, correlation coefficient between the observed and reconstructed EMG; solid lines, mean; shading, SD. (B) Mean decoding accuracy pooled across muscles (Monkey T: n = 12 muscles; Monkey C: n = 10 muscles). Plotted are correlation coefficients between the observed and reconstructed traces from the data (M1, cyan; S1, red) compared with those between the observed traces and the traces reconstructed from random shuffling of activity (Monkey T: n = 12 muscles; Monkey C: n = 10 muscles; *P < 0.01). Superimposed bars, mean. (C) Reconstruction of EMG profile throughout the premovement and movement periods (−500 to 2000 ms around movement onset). The diagrams use a similar format for (A). (D) Mean decoding accuracy of the reconstruction throughout the premovement and movement periods. The diagrams use a similar format for (B).

Fig. 6. Premovement activity in S1 encodes the initial burst of muscle activity slightly after M1 does.

(A) Scatterplots of observed peak EMG amplitude of the hand muscle versus reconstructed EMG amplitude using the premovement activity in M1 (cyan) or S1 (red) of Monkey T. Dot, a single trial. Equation of the fitting is shown in the lower right corner. R, correlation coefficient between the observed and reconstructed EMG. (B and C) Average slopes of regression lines (B) and correlation coefficients (C) pooled across muscles (Monkey T: n = 12 muscles; Monkey C: n = 10 muscles; *P < 0.05). Superimposed bars, mean. (D and F) Average slopes of regression lines (D) and correlation coefficients (F) were plotted against the end of 50-ms sliding windows. Solid lines, means; shading, SD; dashed black line, threshold for onset of activity; arrows, onset. (E and G) Onset times of M1 (cyan) and S1 (red) activities that encode peak EMG amplitude pooled across electrodes were less than the onset time of EMG (E, slopes; G, correlation coefficients; Monkey T: n = 12 muscles; Monkey C: n = 9 muscles; *P < 0.01). Onset time of the activity in M1 was earlier than that in S1 (Monkey T: n = 12 muscles; Monkey C: n = 9 muscles; **P < 0.01). Superimposed bar graphs, mean.