A Non-canonical Feedback Circuit for Rapid Interactions between Somatosensory Cortices

Abstract

Sensory perception depends on interactions among cortical areas. These interactions are mediated by canonical patterns of connectivity in which higher areas send feedback projections to lower areas via neurons in superficial and deep layers. Here, we probed the circuit basis of interactions among two areas critical for touch perception in mice, whisker primary (wS1) and secondary (wS2) somatosensory cortices. Neurons in layer 4 of wS2 (S2_L4) formed a major feedback pathway to wS1. Feedback from wS2 to wS1 was organized somatotopically. Spikes evoked by whisker deflections occurred nearly as rapidly in wS2 as in wS1, including among putative S2_L4 → S1 feedback neurons. Axons from S2_L4 → S1 neurons sent stimulus orientation-specific activity to wS1. Optogenetic excitation of S2_L4 neurons modulated activity across both wS2 and wS1, while inhibition of S2_L4 reduced orientation tuning among wS1 neurons. Thus, a non-canonical feedback circuit, originating in layer 4 of S2, rapidly modulates early tactile processing.

Keywords: S1; S2; barrel cortex; feedback; layer 4; optogenetics; orientation selectivity; projection-specific; secondary somatosensory cortex; whisker system.

Figure 1. Layer 4 of S2 Provides a Major Output to S1

(A) Example images showing laminar distribution of labeling within wS2 after anterograde and retrograde tracer injections into wS1. Left: S2 → S1 neurons retrogradely labeled by CTB-Alexa injected into wS1. Layer boundaries were determined by GAD67 staining. Right: S1 → S2 axonal projections in wS2, anterogradely labeled by AAV tracer injected into wS1. (B) Similar to (A), but showing labeling within wS1 after anterograde and retrograde tracer injections into wS2. (C) Left: number of retrogradely labeled S2 →S1 neurons in wS2, plotted as a function of normalized depth within cortex (n = 2,830 neurons from 12 mice). Right: normalized S1 →S2 axonal fluorescence measured in wS2, shown for individual mice (gray curves, n = 8) and the mean (black). (D) Left: number of retrogradely labeled S1 →S2 neurons in wS1 (n = 2,624 neurons from 9 mice). L4 labeling reflects neurons located in septa (Figure S2). Right: normalized S2 →S1 axonal fluorescence measured in wS1 (n = 8 mice). (E) Coronal section from an Scnn1a-Tg3-Cre;Ai9 mouse injected in the thalamic posterior medial nucleus (POm) with AAV-EGFP. Right: zoom showing EGFP (cyan) in POm. The ventral posterior medial nucleus (VPM) is also indicated. (F) Thalamocortical axons from POm shown (EGFP fluorescence) in a coronal section spanning wS1 and wS2. A band of axons is visible in L5A of wS1 (arrow at far right), and can be seen to continue into wS2 (left and middle panels). In wS2, a second band of POm axons occurs superficial to the L5A band. Same animal as in (E). (G) tdTomato fluorescence from section in (F), showing L4 labeling in wS1 (arrow at far right), and the continuation of this band into wS2 (left and middle panels). (H) Overlay of EGFP and tdTomato fluorescence. Labeled neurons (red) in wS2 of the Scnn1a-Tg3-Cre;Ai9 mouse are located above L5A, in the band of POm axons corresponding to L4. See also Figures S1–S4.

Figure 2. Feedback from S2 to S1 Is Somatotopic within the Whisker Regions

(A) Two colors of CTB-Alexa were injected in different barrel columns of wS1, to retrogradely label neurons in wS2. The brain was sectioned tangentially and stained for GAD67 to show barrels in wS1. Inset: retrogradely labeled neurons in wS2 shown with higher image brightness. (B) Top: additional example of experiment in (A) from a different animal and with overlay of barrel map estimated from GAD67 staining. Bottom: outlines showing locations of CTB-Alexa injections in wS1 and extent of retrograde labeling in wS2. Outlines were used to determine the center of injection sites and retrogradely labeled sites across different animals. (C) Aggregate map across injections and mice (n = 13 injections in 7 mice) showing centers of injection sites in wS1 and retrogradely labeled locations in wS2 (corresponding sites in wS1 and wS2 indicated by colors). (D) Top left: a cocktail of retrograde (CTB-Alexa555) and anterograde (AAV-EGFP) tracers was injected at one site in wS1 and a different color of retrograde tracer (CTB-Alexa647) at another site in wS1. Top right: overlay of CTB and AAV-EGFP labeling in tangential sections shows reciprocal connections between wS1 and wS2. Bottom left: EGFP signal alone. Bottom right: CTB-Alexa signals. See also Figure S5.

Figure 3. Putative S2_L4→S1 Neurons Are Rapidly Excited by Touch

(A) Tetrodes were used to record wS2 single-unit responses to whisker deflections in Scnn1a-Tg3-Cre mice injected with AAV-DIO-hChR2-EYFP into wS2. At the end of each recording session, trains of blue light pulses (2 ms pulses at 20 Hz) were applied over wS1 in order to antidromically stimulate putative S2_L4 → S1 neurons. (B) Expression of ChR2-EYFP in S2_L4. Blue arrowhead: estimated location of the optic fiber tip over wS1. White arrowhead: electrolytic lesion used to locate the tetrode tract. (C) Example spike raster and spike time histogram showing activity evoked by light pulses (indicated at top) for a putative S2_L4 → S1 unit. (D) Left: raster and PSTH for unit in (C) showing responses to all laser pulses within a shorter peri-stimulus time window. Blue arrowhead: spike latency (6.7 ms) measured from light onset to PSTH peak. Right: spike waveform means from the four tetrode channels, for spikes occurring in the absence (“Baseline”) of or in response to (“Photo”) light pulses. (E) Top: schematic of the sinusoidal whisker deflection waveform (40 Hz, 0.5 s). Middle: spike raster for a putative S2_L4 → S1 unit (first 100 trials). Bottom: PSTH (95% CI) for the example unit. (F) Population average PSTH from putative S2_L4 → S1 units (±95% CI; n = 58 units from 6 mice). (G) Top: overlay of population average PSTHs (±95% CIs) for those wS1 (gray; n = 585), wS2 (blue; n = 1,557), and putative S2_L4 →S1 (green; n = 39) units excited within the first 50 ms after stimulus onset. Middle: plot at top is shown after normalizing each PSTH to its peak. Bottom: zoom of time window near stimulus onset. (H) Left: spike raster (first 100 trials) and PSTH for an example wS2 unit with reliable response to 40 Hz whisker deflection waveform. Right: stimulus cycle-locked raster plot and PSTH for same unit. Orange curve: kernel density estimate used to quantify the range (F1) and mean (F0) of the average response to one cycle of the whisker stimulus. The F1/F0 ratio quantifies the degree of cycle-by-cycle spike rate modulation for each unit. (I) Same as (H) but for a putative S2_L4 → S1 unit without clear cycle-by-cycle modulation. (J) Cumulative histograms showing the F1/F0 ratio for all wS1 (n = 1,841), wS2 (n = 5,557), and putative S2_L4 → S1 (n = 58) units. S2_L4 → S1 units showed lower values compared with wS1 (**p = 4.4 × 10⁻⁹; Wilcoxon rank-sum) and wS2 (**p = 1.4 × 10⁻⁷; Wilcoxon rank-sum) units.

Figure 4. S2_L4 → S1 Neurons Send Orientation-Specific Activity to S1

(A) Two-photon calcium imaging of S2_L4 → S1 axons in superficial S1 was performed during sinusoidal deflections (40 Hz, 0.5 s) of a single whisker along either horizontal or vertical orientations. (B) Example field-of-view (FOV) showing GCaMP6s-labeled S2_L4 → S1 axons in S1 (mean fluorescence across one trial). (C) Heatmaps of Z scored responses of each axonal bouton (along rows) to either horizontal (left) or vertical (right) whisker deflections, for boutons that showed similar responses to the two orientations (n = 1,265 boutons from 12 FOVs in 4 mice). Dashed vertical lines: onset of whisker stimulus. (D) Same as (C) but for boutons with significantly larger responses for horizontal compared with vertical deflections (n = 42 from 11 FOVs in 4 mice). (E) Same as (C) but for boutons with significantly larger responses for vertical deflections (n = 70 from 12 FOVs in 4 mice).

Figure 5. Excitation of S2_L4 Neurons Modulates Activity across Layers in S2 and S1

(A) Schematic of wS2 recording with S2_L4 stimulation. Silicon probes recorded single units across layers in wS2. Responses were recorded to sinusoidal deflections of a single whisker along either horizontal or vertical orientations, and to optogenetic excitation (via ChR2) of S2_L4 neurons that began prior to and spanned the duration of the whisker stimulus. (B) Top: schematic of the whisker stimulus and 470 nm illumination (dark cyan). Middle: spike rasters for an S2_L4 unit with (green) and without (black) optogenetic stimulation of S2_L4 (first 100 trials for each). Bottom: PSTHs (95% CI) for the example unit with (green) and without (black) optogenetic stimulation. (C) Population average PSTHs from S2_L4 units with (green) and without (black) optogenetic stimulation (±95% CI; n = 51 units from 6 mice). (D) Example units with significantly higher (unit 1, at top) or lower (unit 2, at bottom) spike rates during a “transient” phase of the optogenetic stimulus (indicated by purple horizontal bar at top; first 100 ms of LED illumination, from −200 ms to −100 ms relative to whisker stimulus onset). Conventions as in (B). (E) Example units with significantly higher (unit 3, at top) or lower (unit 4, at bottom) spike rates during a “sustained” phase of the optogenetic stimulus (indicated by brown horizontal bar at top; 500 ms covering identical period as the whisker stimulus). Conventions as in (B). (F) Optogenetic modulation index for the transient phase of the optogenetic stimulus, plotted for each unit as a function of normalized depth within cortex. Estimated boundaries of cortical layer 4 are indicated with horizontal dashed lines. Filled circles indicate significantly excited (red) or inhibited (blue) units. Example units from (D) are indicated (dark red and dark blue filled circles). Pie charts show percentages of units with significant optogenetic modulation for L4 and layers above (“superficial”) and below (“deep”) L4 (n = 393 units from 6 mice). (G) Same as (F) but for the sustained phase of optogenetic stimulation. (H) Schematic of wS1 recording with S2_L4 stimulation. Experiments were identical to those of (A)–(G) except recordings were made in wS1. (I) Coronal section showing DiI-marked silicon probe recording tracts (magenta) in wS2 (left tract) and wS1 (right tract), and ChR2-EYFP fluorescence in S2_L4. (J) Zoom of (I) showing wS2 tract and estimated pia, L4, and white matter boundaries (white horizontal lines). (K–N) Same as (D)–(G) but for wS1 instead of wS2 recordings (n = 347 units from 6 mice). See also Figure S6.

Figure 6. Inhibition of S2_L4 Neurons Suppresses Activity in S2 but Only Weakly So in S1

(A) Schematic of wS2 recording with S2_L4 inhibition. Silicon probes recorded single units across layers in wS2. Responses were recorded to sinusoidal deflections of a single whisker along either horizontal or vertical orientations, with and without optogenetic inhibition (via eArch3.0) of S2_L4 that began prior to and spanned the duration of the whisker stimulus. (B) Coronal section showing DiI-marked silicon probe tract (green) in wS2, and eArch3.0-EYFP fluorescence in S2_L4. Inset: zoom of region in white box. (C) Top: schematic of the whisker stimulus and 565 nm illumination (light green; preceding ramp-up not depicted). Middle: spike rasters for an S2_L4 unit with (green) and without (black) optogenetic inhibition of S2_L4 (first 100 trials for each). Bottom: PSTHs (95% CI) for the example unit with (green) and without (black) inhibition. (D) Optogenetic modulation index for the sustained phase (same time period as the whisker stimulus) of the optogenetic inhibition, plotted for each unit as a function of normalized depth within wS2 (n = 663 units from 6 mice). Conventions as in Figure 5F. Example L4 unit from (C) indicated by dark blue filled circle. (E) Schematic of wS1 recording with S2_L4 inhibition. Experiments were identical to those depicted in (A)–(D) except silicon probe recordings were made in wS1. (F) Coronal section showing DiI-marked silicon probe recording tract (green) in wS1 and eArch3.0-EYFP fluorescence in S2_L4. (G) Same as (D) but for wS1 recordings (n = 621 units from 7 mice). (H) Boxplots for wS2 (left) and wS1 (right) units depicting 25th, 50th, 75th percentiles and range of the optogenetic modulation index after removal of outliers (STAR Methods; **, wS2: superficial, p = 2.3 × 10⁻¹⁴; L4: p = 5.0 × 10⁻⁸, deep: p = 1.4 × 10⁻³; wS1: superficial: p = 5.8 × 10⁻⁵, L4: p = 1, deep: p = 3.7 × 10⁻⁴; sign tests).

Figure 7. Inhibition of S2_L4 Neurons Reduces Orientation Sensitivity in S1

(A) Schematic of experiment: same as in Figure 6A, but with analysis (B)–(D) limited to whisker-responsive units. (B) Spike rasters (first 100 trials) and PSTHs (95% CIs) showing responses to horizontal (left column) or vertical (right) whisker stimulations for a wS2 unit with (green) and without (black) inhibition of S2_L4. (C) Top: spike rate during horizontal versus vertical whisker deflections (means ± 95% CIs) for the example unit in (B). Bottom: change in orientation sensitivity index (OSI; ±95% CI) due to optogenetic inhibition for same unit. (D) Change in OSI plotted for each whisker-responsive wS2 unit as a function of normalized depth within cortex. Conventions as in Figure 5F. Example unit from (B) indicated by dark blue filled circle. Pie charts show percentages of wS2 units with significant change in OSI during S2_L4 optogenetic inhibition (n = 467 whisker-responsive units from 6 mice). (E) Schematic of experiment: same as in Figure 6E, but with analysis (F)–(H) limited to whisker-responsive units. (F and G) Same as (B) and (C) but for a wS1 unit. (H) Same as (D) but for wS1 units (n = 398 whisker-responsive units from 7 mice). Example unit from (F) indicated by dark blue filled circle. (I) Left: change in mean spike rate during inhibition of S2_L4 for preferred and non-preferred whisker deflection orientations, for those wS2 units (filled circles from D) showing significantly increased (red; n = 12) or decreased (blue; n = 18) OSI. Right: same data as in left panel but expressed as percent change in firing rate. (J) Same as (I) but for wS1 units (n = 3 and 34 units for increased and decreased OSI, respectively). See also Figure S7.