Learning-related congruent and incongruent changes of excitation and inhibition in distinct cortical areas

Abstract

Excitatory and inhibitory neurons in diverse cortical regions are likely to contribute differentially to the transformation of sensory information into goal-directed motor plans. Here, we investigate the relative changes across mouse sensorimotor cortex in the activity of putative excitatory and inhibitory neurons-categorized as regular spiking (RS) or fast spiking (FS) according to their action potential (AP) waveform-comparing before and after learning of a whisker detection task with delayed licking as perceptual report. Surprisingly, we found that the whisker-evoked activity of RS versus FS neurons changed in opposite directions after learning in primary and secondary whisker motor cortices, while it changed similarly in primary and secondary orofacial motor cortices. Our results suggest that changes in the balance of excitation and inhibition in local circuits concurrent with changes in the long-range synaptic inputs in distinct cortical regions might contribute to performance of delayed sensory-to-motor transformation.

Fig 1. Multiarea recordings during delayed whisker detection task and assignment of RS and FS units to cortical subdivisions.

(A) Schematic of the whisker detection task with delayed response and the targets of silicon probe recordings. (B) Training paradigm. Novice and Expert mice were first pretrained in a task, where licking after the auditory cue was rewarded. Expert mice were further trained to only lick in whisker trials. (C) Final task structure used during recording sessions (for both groups of mice) and behavioral outcomes. (D) Example coronal section of an Expert mouse brain with fluorescent track of a probe in wS1, registered to the Allen Mouse Brain Atlas, https://mouse.brain-map.org [47]. (E) Reconstructed laminar location of recording sites of the silicon probe shown in (D) according to the Allen Atlas (left); filtered recorded raw data of 7 sites around one detected spike; and average extracted spike waveform for this example neuron (right). After spike sorting, the position of each cluster (i.e., neuron) was assigned to the location of recording site with the largest spike amplitude (filled circle), and spike width was calculated on the average spike waveform from this site. (F) Raster plot and peri-stimulus time histogram for the example neuron shown in (E). Trials are grouped based on outcome. (G) Spike width distribution for neurons recorded in Expert mice. Neurons were categorized as FS (spike width <0.26 ms) or RS (spike width >0.34 ms). Neurons with intermediate spike width (gray bins) were excluded from further analyses. (H) Baseline AP rate in Expert mice. Spike width distribution versus baseline AP rate (left) and overlay of spike rate distribution for RS and FS units (right). Note the log-normal distribution of baseline firing rates for both RS and FS units. Normal distributions were fitted to the RS and FS histograms (solid lines). (I) Comparison of mean spike rate in RS versus FS neurons of Expert mice. Error bars: SEM. ***: p < 0.001, nonparametric permutation test. (J–O) Opto-tagging GABAergic neurons in VGAT-ChR2 mice. (J) Grand average firing rate of RS (orange, spike width >0.34 ms, 130 neurons from 4 mice) and FS (green, spike width <0.26 ms, 51 neurons from 4 mice) units upon 100-Hz blue light stimulation (shading shows SEM). Note the suppression of activity in RS and the strong increase of activity in FS population. Inset shows the overlay of average spike waveforms for all RS and FS neurons. (K) OMI versus spike width (left) and percentage of modulated neurons (right). Each circle represents one neuron, filled circles indicate neurons with significant OMI (p < 0.05, nonparametric permutation tests). Pie charts show the percentage of neurons in each group with nonsignificant modulation (NS), and significant positive (OMI > 0) or negative (OMI < 0) modulation upon blue light stimulation. (L) Blue light stimulation in VGAT-ChR2 mice increased the activity of narrow-spike neurons labeled as FS, while it suppressed the activity of broad-spike neurons labeled as RS; 100 to 500 ms after light onset. Error bars: SEM; **: p < 0.01; ***: p < 0.001. (M) Raster plot and peri-stimulus time histogram during the first 10 ms of the 100-Hz trains of blue light stimulation for an example opto-tagged neuron. (N) Waffle plots showing broad-spike (orange) and narrow-spike (green) neurons, and the opto-tagged neurons (blue) in each group. Numbers indicate the percentage of opto-tagged neurons in each group. (O) Weighted proportion of neurons with narrow (FS) or broad (RS) spike among opto-tagged neurons in (N). The underlying data for Fig 1 can be found in S1 Data. ALM, anterior lateral motor cortex; AP, action potential; FS, fast spiking; OMI, opto modulation index; RS, regular spiking; tjM1, tongue-jaw primary motor cortex; wM1, whisker primary motor cortex; wM2, whisker secondary motor cortex; wS1, whisker primary somatosensory cortex; wS2, whisker secondary somatosensory cortex.

Fig 2. FS neurons had similar but larger task modulation compared to RS neurons in the same region.

(A) Baseline-subtracted (2 seconds prior to visual onset) population firing rates (mean ± SEM) of RS and FS neurons from different regions of Novice mice are superimposed for hit trials. wS1: 73 RS units in 7 mice, 103 FS units in 7 mice; wS2: 120 RS units in 8 mice, 68 FS units in 8 mice; wM1: 147 RS units in 7 mice, 66 FS units in 7 mice; wM2: 244 RS units in 7 mice, 57 FS units in 7 mice; ALM: 234 RS units in 6 mice, 37 FS units in 5 mice; tjM1: 271 RS units in 8 mice, 61 FS units in 8 mice. Average first lick histogram for all Novice mice is shown in the bottom. (B) Percentage of RS (left) and FS (right) neurons in different regions of Novice mice that are positively (top) or negatively (bottom) modulated compared to baseline (nonparametric permutation test, p < 0.025). (C) Similar to (A), but for Expert mice. wS1: 258 RS units in 15 mice, 237 FS units in 15 mice; wS2: 342 RS units in 12 mice, 161 FS units in 12 mice; wM1: 452 RS units in 11 mice, 134 FS units in 11 mice; wM2: 401 RS units in 10 mice, 107 FS units in 10 mice; ALM: 766 RS units in 12 mice, 109 FS units in 12 mice; tjM1: 505 RS units in 11 mice, 83 FS units in 11 mice. Average first lick histogram for all Expert mice is shown in the bottom. (D) Similar to (B), but for Expert mice. Note the difference in color scales for fraction of positively or negatively modulated neurons in b and d. The underlying data for Fig 2 can be found in S2 Data. ALM, anterior lateral motor cortex; FS, fast spiking; RS, regular spiking; tjM1, tongue-jaw primary motor cortex; wM1, whisker primary motor cortex; wM2, whisker secondary motor cortex; wS1, whisker primary somatosensory cortex; wS2, whisker secondary somatosensory cortex.

Fig 3. Fast propagation of sensory responses across cell classes and cortical regions.

(A) Change in firing rate (mean ± SEM) of different cortical regions in the first 100-ms window after whisker deflection for RS (top) and FS (bottom) neurons in Novice (left) and Expert (right) mice (numbers of units and mice are the same as in Fig 2). (B) Whisker-evoked response latency maps. For each silicon probe in Novice (left) and Expert (right) mice, average latency of whisker-evoked response is shown separately for RS and FS units. Circles represent silicon probes and are colored according to the average latency across all responsive neurons recorded on the probe. (C) Comparison of latency of RS versus FS neurons in Novice (left) and Expert (right) mice. (D) Comparison of latency of neurons from Novice versus Expert mice for RS (left) and FS (right) neurons. In (C) and (D), only neurons with a significant whisker response in the first 200 ms (compared to 200 ms before whisker onset, nonparametric permutation test, p < 0.05) were included. Midline represents the median, bottom and top edges show the interquartile range, and whiskers extend to 1.5 times the interquartile range. ***: p < 0.001, **: p < 0.01, *: p < 0.05, ns: p > = 0.05. The underlying data for Fig 3 can be found in S4 and S5 Data. ALM, anterior lateral motor cortex; FS, fast spiking; RS, regular spiking; tjM1, tongue-jaw primary motor cortex; wM1, whisker primary motor cortex; wM2, whisker secondary motor cortex; wS1, whisker primary somatosensory cortex; wS2, whisker secondary somatosensory cortex.

Fig 4. Fast whisker responses in FS neurons of sensory areas.

(A) Baseline-subtracted (50 ms prior to whisker onset) population firing rate (mean ± SEM) of RS (left) and FS (right) neurons in wS1 and wS2 of Novice mice. wS1: 73 RS units in 7 mice, 103 FS units in 7 mice; wS2: 120 RS units in 8 mice, 68 FS units in 8 mice. (B) Same as (A) but for Expert mice. wS1: 258 RS units in 15 mice, 237 FS units in 15 mice; wS2: 342 RS units in 12 mice, 161 FS units in 12 mice. (C) Whisker-evoked change in spike rate in the first 50 ms (mean ± SEM) in wS1 and wS2 for RS and FS units and in Novice and Expert mice. ***: p < 0.001. Gray lines show nonsignificant comparisons. (D) Latency of the whisker-evoked response in wS1 and wS2. Only neurons with a significant whisker response in the first 100 ms (compared to 100 ms before whisker onset, nonparametric permutation test, p < 0.05) were included (Novice wS1: 56/73 RS units, 96/103 FS units, 8 mice; Novice wS2: 97/120 RS units, 57/68 FS units, 8 mice; Expert wS1: 190/258 RS units, 210/237 FS units, 18 mice; Expert wS2: 262/342 RS units, 148/161 FS units, 18 mice). Boxplots represent the distribution of the latency defined as the time to reach to half-maximum response. Midline represents the median, bottom and top edges show the interquartile range, and whiskers extend to 1.5 times the interquartile range. ***: p < 0.001, **: p < 0.01. Gray lines show nonsignificant comparisons. (E) Inactivation of wS1 and wS2. Left: Schematic showing the inactivation of wS1 and wS2 areas during whisker stimulus presentation, in VGAT-ChR2 mice [22,51]. Light trials were interleaved with no-light control trials and comprised 1/3 of total trials. Right: Change in hit and FA rate—comparing light and no-light trials—upon optogenetic inactivation of wS1 and wS2. Light colors show individual mice (9 mice), thick lines represent averages, and error bars show SEM. **: p < 0.01, ns: p > = 0.05. The underlying data for Fig 4 can be found in S6 Data. ChR2, channelrhodopsin-2; FA, false alarm; FS, fast spiking; RS, regular spiking; VGAT, vesicular GABA transporter; wS1, whisker primary somatosensory cortex; wS2, whisker secondary somatosensory cortex.

Fig 5. Distinct frontal projections of wS1 and wS2.

(A) Schematic of anterograde axonal tracing of wS1 and wS2 projections in frontal cortex. Fluorescent proteins of different colors were expressed in wS1 and wS2 regions and frontal projection patterns were identified using anatomical reconstructions and registration to Allen Brain Atlas. (B–D) Coronal sections showing example 2-color injections in wS1 (magenta) and wS2 (green) and their frontal projection centers. Viral expression in wS1 and wS2 (B) and frontal sections showing the center of frontal projections in wM1 (C) and wM2 (D). All brains were registered to the Allen Mouse Brain Atlas, https://mouse.brain-map.org. (E) Grand average cortical fluorescent map of wS1 projections (4 mice). (F) Same as (E) but for wS2 projections (4 mice). (G) Overlay of grand average fluorescent map of wS1 (magenta) and wS2 (green) projections in frontal cortex. (H) Center of projections from wS1 and wS2 in frontal cortex. Contour plots at 95% and 75% maximum of the grand average fluorescent intensity from wS1 (magenta) and wS2 (green) projections, showing the location of wM1 and wM2, respectively. Markers show the center of projections for different mice. Projections in the same mice are indicated with similar markers. The underlying data for Fig 5 can be found in S7 Data. wM1, whisker primary motor cortex; wM2, whisker secondary motor cortex; wS1, whisker primary somatosensory cortex; wS2, whisker secondary somatosensory cortex.

Fig 6. Learning differently modulated sensory responses of RS and FS neurons in wM1 and wM2 areas.

(A) Decrease of whisker response in wM1 RS neurons across learning. Top: baseline-subtracted (50 ms prior to whisker onset) population firing rate (mean ± SEM) overlaid for Novice mice (147 neurons in 7 mice) and Expert mice (452 neurons in 11 mice). Bottom: Comparison of whisker-evoked response in Novice and Expert mice. Bar plots showing average population rate in 10- to 90-ms window (mean ± SEM) after whisker onset and statistical comparison using nonparametric permutation test (left) (**: p < 0.01; *: p < 0.05). The fraction of positively (filled bars) or negatively (empty bars) modulated neurons in the same window (right). Modulation of individual neurons compared to a similar window size prior to whisker onset, was identified using nonparametric permutation test (p < 0.005). The fractions of modulated neurons in Novice and Expert were compared using a chi-squared proportion test (*: p < 0.05; ns: p > = 0.05). (B) Increase of whisker response in wM1 FS neurons across learning. Panels are similar to (A) but for wM1 FS neurons in Novice (66 neurons in 7 mice) and Expert mice (134 neurons in 11 mice) (***: p < 0.001). (C) Increase of whisker response in wM2 RS neurons across learning. Panels are similar to (A) but for wM2 RS neurons in Novice (244 neurons in 7 mice) and Expert mice (401 neurons in 10 mice). (D) Decrease of whisker response in wM2 FS neurons across learning. Panels are similar to (A) but for wM2 FS neurons in Novice (57 neurons in 7 mice) and Expert mice (107 neurons in 10 mice). (E) Pair-wise correlation between sensory and motor cortices in Novice and Expert mice. Left: Scatter plot showing the trial-by-trial correlation between the whisker-evoked response of an example pair of neurons in wS2 and wM2. Each circle represents the response of the neuronal pair in one trial. Circles were jittered slightly for the purpose of visualization. Gray line: least-squares regression. Middle: Average pair-wise Pearson correlation of wS1-RS units with wM1-RS (110 neuron pairs in 1 Novice mouse, and 68 neuron pairs in 2 Expert mice) and wS1-RS units with wM1-FS units (44 neuron pairs in 1 Novice mouse, and 89 neuron pairs in 2 Expert mice) separately. Right: Average pair-wise Pearson correlation of wS2-RS units with wM2-RS (876 neuron pairs in 6 Novice mouse, and 583 neuron pairs in 3 Expert mice) and wS2-RS units with wM2-FS units (343 neuron pairs in 6 Novice mouse, and 209 neuron pairs in 3 Expert mice). Error bars: SEM. Statistical comparison between Novice and Expert was performed using Wilcoxon rank-sum test (ns: p > = 0.05; *: p < 0.05; ***: p < 0.001). (F) Interareal functional connectivity identified based on cross-correlograms. Left: Example cross-correlogram between a pair of simultaneously recorded neurons from wS2 and wM2. Red dotted line shows the threshold for detecting sharp peaks. A directional connection from wS2 to wM2 was detected as there is a threshold crossing within the time lags between 0 and 10 ms. Middle: Percentage of detected directional connections from wS1-RS units to wM1-RS and wM1-FS units in 1 Novice and 2 Expert mice. Right: Percentage of detected directional connections from wS2-RS units to wM2-RS and wM2-FS units in 6 Novice and 3 Expert mice. The numbers on each bar represent the number of identified connections and the total number of recorded pairs. The fractions of connections in Novice and Expert were compared using a chi-squared proportion test (ns: p > = 0.05; **: p < 0.01). The underlying data for Fig 6 can be found in S8 Data. FS, fast spiking; RS, regular spiking; wM1, whisker primary motor cortex; wM2, whisker secondary motor cortex.

Fig 7. FS neuronal responses in tjM1 and ALM changed similarly to RS neurons.

(A) Suppression of tjM1 RS neurons in Expert mice. Top: baseline-subtracted (50 ms before whisker onset) firing rate (mean ± SEM) overlaid for Novice (271 RS units in 8 mice) and Expert mice (505 RS units in 11 mice). Bottom: Comparison of whisker-evoked response in Novice and Expert mice. Bar plots showing population rate in 40- to 90-ms window (mean ± SEM) after whisker onset and statistical comparison using nonparametric permutation test (left, ***: p < 0.001); fraction of positively (filled bars) or negatively (empty bars) modulated neurons in the same window (right). Modulation of individual neurons compared to a similar window size prior to whisker onset, was identified using nonparametric permutation test (p < 0.005). Fraction of modulated neurons in Novice and Expert were compared using a chi-squared proportion test (ns: p > = 0.05). (B) Suppression of tjM1 FS neurons in Expert mice. Panels are similar to (A) but for tjM1 FS neurons in Novice (61 neurons in 8 mice) and Expert mice (83 neurons in 11 mice). (C) Delay activity of RS neurons in Expert mice. Top: baseline-subtracted (1 second before whisker onset) firing rate (mean ± SEM) overlaid for Novice (234 RS units in 6 mice) and Expert mice (766 RS units in 12 mice). Bottom: Comparison of whisker-evoked response in Novice and Expert mice. Bar plots showing population rate in 200- to 1,000-ms window (mean ± SEM) after whisker onset and statistical comparison using nonparametric permutation test (left, ***: p < 0.001); fraction of positively (filled bars) or negatively (empty bars) modulated neurons in the same window (right). Modulation of individual neurons compared to a similar window size prior to whisker onset was identified using nonparametric permutation test (p < 0.005). Chi-squared proportion test: ***: p < 0.001, ns: p > = 0.05. (D) Delay activity of ALM FS neurons in Expert mice. Panels are similar to (C) but for ALM FS neurons in Novice (37 FS units in 5 mice) and Expert mice (109 FS units in 12 mice). The underlying data for Fig 7 can be found in S9 Data. ALM, anterior lateral motor cortex; FS, fast spiking; RS, regular spiking; tjM1, tongue-jaw primary motor cortex.

Fig 8. Diverse changes of putative excitation–inhibition balance in different cortical regions across learning.

(A) FS (putative PV GABAergic inhibitory) neurons and RS (putative glutamatergic excitatory pyramidal) neurons are typically considered to be strongly and reciprocally connected in local cortical microcircuits providing fast balance of excitation and inhibition. (B) Schematic showing the location of different cortical regions. (C) LMI in different cortical regions for RS and FS units, representing learning-induced change in putative excitation and inhibition, respectively. LMI was quantified as the normalized difference between whisker-evoked firing rate in Novice and Expert mice. (D) Map of the putative excitation–inhibition change across learning, shown as LMI across cortical regions for RS and FS neurons. (E) Change in the putative excitation–inhibition balance across learning, quantified as difference between LMI of RS and FS neurons in different cortical regions. (F) Cortical map of the putative excitation–inhibition balance change across learning, calculated as LMI difference between RS and FS neurons. The underlying data for Fig 8 can be found in S10 Data. ALM, anterior lateral motor cortex; FS, fast spiking; LMI, learning modulation index; PV, parvalbumin-expressing; RS, regular spiking; tjM1, tongue-jaw primary motor cortex; wM1, whisker primary motor cortex; wM2, whisker secondary motor cortex; wS1, whisker primary somatosensory cortex; wS2, whisker secondary somatosensory cortex.