Distinct Contributions of Whisker Sensory Cortex and Tongue-Jaw Motor Cortex in a Goal-Directed Sensorimotor Transformation

Abstract

The neural circuits underlying goal-directed sensorimotor transformations in the mammalian brain are incompletely understood. Here, we compared the role of primary tongue-jaw motor cortex (tjM1) and primary whisker sensory cortex (wS1) in head-restrained mice trained to lick a reward spout in response to whisker deflection. Two-photon microscopy combined with microprisms allowed imaging of neuronal network activity across cortical layers in transgenic mice expressing a genetically encoded calcium indicator. Early-phase activity in wS1 encoded the whisker sensory stimulus and was necessary for detection of whisker stimuli. Activity in tjM1 encoded licking direction during task execution and was necessary for contralateral licking. Pre-stimulus activity in tjM1, but not wS1, was predictive of lick direction and contributed causally to small preparatory jaw movements. Our data reveal a shift in coding scheme from wS1 to tjM1, consistent with the hypothesis that these areas represent cortical start and end points for this goal-directed sensorimotor transformation.

Keywords: goal-directed sensorimotor transformation; licking; motor cortex; reward-based learning; somatosensory cortex; whisker sensory processing.

Figure 1

Identification of Tongue-Jaw Primary Motor Cortex (A) For optogenetic motor mapping, a blue laser beam was directed in a grid-like manner over the dorsal cortex (blue dots) of a Thy1-ChR2 mouse. Whisker and jaw movements (side view by 45° mirror) were filmed simultaneously by a high-speed camera. Grand average (n = 6 mice) motor maps for the C2 whisker and the jaw were aligned on the intrinsic optical signal for the C2 whisker. Black crosses indicate the center of the frontal cortical region that evoked movements for individual mice. Green crosses indicate centers of the intrinsic optical signal evoked by C2 whisker and tongue-jaw sensory stimulation for each mouse. Red crosses represent average bregma position. (B) For wide-field functional calcium imaging, either the C2 whisker or the tongue was stimulated with a piezoelectric actuator in Thy1-GCaMP6f mice. Grand average (n = 5 mice) sensory-evoked responses during a 100–200 ms post-stimulus time window for C2 whisker (left) and tongue-jaw (right) stimulation were aligned to the wS1 calcium signal evoked by C2 whisker stimulation. Green crosses indicate centers of the first activation spot during the early phase of the response (0–100 ms). Black crosses are the centers of the frontal secondary spot (100–200 ms) for individual mice. Red crosses represent average bregma position. (C) For anterograde axonal tracing, AAV-hSyn-turboRFP was injected in the cortical region where the first activation spot of tongue-jaw stimulation (tjS1) was detected by calcium imaging. Axonal projections were found in a localized column in motor cortex (tjM1). Grand average (n = 5 mice) of cortical fluorescence aligned to the injection site (right). Black crosses indicate centers of the anterior spot for each mouse. Green cross corresponds to the center of the injection site. Red cross represents average bregma position. See also Figure S1, Table S1, and Video S1.

Figure 2

Different Roles of tjM1 and wS1 in Sensorimotor Behaviors (A) In a multisensory detection task, mice were rewarded by licking a spout within a 1.5 s time window following a brief whisker deflection (1 ms) or a short auditory tone (10 ms). Catch trials with no stimulus and no reward were interleaved. Table below describes the categories of possible trial types and behavioral outcomes. (B) Lick probability (above) and median first lick latency (below) for auditory (blue), whisker (green), and catch (gray) trials for expert mice (n = 15 mice). (C) Lick probability for the auditory (blue), whisker (green), and catch (gray) trials during optogenetic inactivation of wS1 (n = 7 mice), tjM1 (n = 17 mice), and ALM (n = 8 mice). (D) In a multimotor detection task, mice were trained to lick a spout on the right or left in response to whisker deflection. Every ∼50 trials the reward location was changed between right and left spouts. (E) Lick probability for right spout (green) and left spout (red) aligned to the spout switch event (n = 14 mice). The lick probability for the catch trials is shown in gray (right spout, solid; left spout, dashed). (F) tjM1 and ALM inactivation during the multimotor task. Upper graphs show the effect of tjM1 inactivation (n = 15 mice) on left and right spout lick probability during right block trials, as well as miss rates. Lower graphs show the same, but during ALM inactivation (n = 8 mice). Data are represented as mean ± SD, except for 95% confidence intervals in (E). Wilcoxon signed-rank test is shown in (B), (C), and (F). See also Figure S2.

Figure 3

Neuronal Correlates of Multisensory and Multimotor Decision Making (A) A microprism assembly was inserted into the cortex for chronic two-photon calcium imaging across layers. (B) Top view of an implanted microprism assembly targeted to wS1. The red contour indicates the peak of the intrinsic optical signal evoked by C2 whisker stimulation. (C) Laminar view of cortical neurons in a Thy1-GCaMP6f mouse imaged through the microprism assembly shown in (B). (D) Traces of behavioral variables and calcium signals from five example neurons (labeled in C) during the multisensory detection task. (E) Average Z score calcium responses aligned to stimulus onset. (F) Left: example field of view during the multisensory task in wS1. Red image channel: average stimulus-triggered responses (0–200 ms) for auditory hit trials. Green image channel: the same but for whisker hit trials. Middle left: scatterplot over all neurons comparing average response during whisker and auditory hit trials (0–200 ms post stimulus). Black points indicate significantly modulated neurons. Gray points represent neurons that did not show any significant modulation. R indicates the Pearson correlation coefficient. Middle right: PCA of the time-varying neuronal population vector during whisker and auditory hit trials. Thick lines indicate the first 200 ms after stimulus onset. Arrow heads show direction of the trajectories. Right: grand average whisker and auditory Z score evoked responses in hit trials for deep (thick) and superficial (thin) neurons. (G) Left: example field of view during the multimotor task in wS1 (left). Red image channel: average stimulus-triggered responses (0–200 ms) for left hit trials. Green image channel: the same but for right hit trials. Middle left: scatterplot over all neurons comparing average response during left and right hit trials (0–200 ms post stimulus). Black points indicate significantly modulated neurons. Gray points represent neurons that did not show any significant modulation. R indicates the Pearson correlation coefficient. Middle right: PCA of the time-varying neuronal population vector during left and right hit trials. Thick lines indicate the first 200 ms after stimulus onset. Arrows show direction of the trajectories. Right: grand average left and right Z score evoked responses in hit trials for deep (thick) and superficial (thin) neurons. (H and I) Same as (F) and (G) but for two-photon imaging in tjM1. Average responses were calculated in the time window −100–100 ms relative to first lick time. Pearson correlation coefficients in (F)–(I): p < 0.001. See also Figure S3.

Figure 4

Decoding of Lick Direction and Small Anticipatory Movements (A) Computation of the maximum log-likelihood for decoding motor output. Left: a population activity matrix of error and hit trials for a tjM1 example session. Middle: the log-tuning curves for left and right licks are shown for all cells in one field of view. Right: matrix multiplication of the population activity matrix with the log-tuning curves (plus the same bias correction for each row, data not shown; see STAR Methods for details) leads to the log-likelihood for left and right lick in each trial. Choosing the side (L or R) with the highest value for each row (trial) gives a prediction of lick direction (red dots). (B) Decoder performance for a post-stimulus (0–200 ms) time window in tjM1 (n = 13 fields of view in 5 mice) and wS1 (n = 11 fields of view in 6 mice). Each gray point indicates the decoder performance of a single field of view and session. Black points represent grand averages. Dashed line shows chance level. (C) Same as (B) but for pre-stimulus (−1,000–0 ms) time window. (D) Left: calcium activity of two tjM1 example neurons during left and right block trials and corresponding jaw movements extracted from facial filming. Black dots, onset of isolated small jaw movements from the facial filming. Blue dots, spout contacts. Orange line, whisker stimulus timing. Right: average response of the neurons aligned to spout contact (blue) or on isolated small jaw movements (black). (E) Population analysis of (D) for tjM1. Differences between right and left spout contact responses were correlated to differences between right and left block responses during small jaw movements. (F) Same as (E) but for wS1. (G) Analysis of small jaw movements in catch trials during tjM1 optogenetic inactivation experiments. Each black point corresponds to the average ratio of jaw movements in a baseline window to jaw movements in the response window with light OFF or light ON (n = 7 mice). Data are represented as mean ± SD. Wilcoxon rank-sum test: tjM1 versus wS1 in (B) and (C); Wilcoxon signed-rank test: tjM1 and wS1 against chance in (B) and (C). Pearson correlation coefficient is shown in (E) and (F). Wilcoxon signed-rank test is shown in (G). See also Figure S4.