Anatomical considerations for registration of mesoscale imaging data.

(A) Intact-skull wide-field preparation with registration pattern superimposed. (B) Schematic of wide-field imaging system. (C) Coronal slices of GCaMP6f-expressing triple-transgenic mouse lines specific to L2/3, L5, and L6. (D) Three-dimensional schematic of the layer-specific reference maps based on the Allen CCF for L2/3, L5, and L6. (E) Transversal top view of panel D superimposed on the L2/3 reference map (black lines).

Layer-specific registration of sensory-evoked maps.

(A) Schematic of wide-field imaging in anesthetized mice (left) and trial structure for the different sensory stimuli presented (right) (B) Average stimulus-evoked calcium activity maps (n= 10 trials), registered with layer-specific reference maps for L2/3, L5, and L6 (top to bottom row). Columns from left to right show the evoked activity maps for a 0.5-s window after stimulus onset for the five different sensory stimuli (min-max values of displayed fluorescence changes ΔF /F are indicated by color bars). Below each map the corresponding calcium traces are shown, with stimulation periods indicated as grey shade (100 ms for visual, and 1.0 s for the rest of the stimuli). (C) Benefit of registration using layer-specific reference maps exemplified for L5 and L6, respectively. Registering sensory-evoked calcium activity maps with the L2/3 reference map is suboptimal (e.g., for A1 and HL) but improves when the appropriate layer-specific reference maps are applied.

Control functional mapping with hemodynamic correction.

(A) Example average functional maps upon hindlimb stimulation in a lightly anesthetized L2/3 mouse (n = 20 trials). Wide-field imaging was performed using dual-channel imaging with 2 excitation wavelength to correct for hemodynamic artifacts. Left: stimulus-averaged ΔF /F map with 470-nm excitation, maximizing GCaMP6f calcium sensitivity. Middle: stimulus-averaged ΔF /F map with 405-nm excitation, near the isosbestic (calcium-independent) wavelength of GCaMP6f. Right: corrected map ΔF /F obtained from the hemodynamic correction (Methods and Materials). Maps are calculated from the image frames within the stimulus window. (B) Same as (A) but with forelimb stimulation. (C) Comparison of ΔF /F maps for 470-nm excitation (upper row) and corrected ΔF /F maps (lower row) for all different types of stimulation (average from n = 2 L2/3 mice). Maps are aligned to the L2/3 reference map and the composite maps with all functional maps overlaid are shown on the right. No differences are apparent between 470-nm and corrected maps.

Wide-field imaging of light scattering across cortical depths.

(A) Schematic of wide-field imaging of a point source of 530-nm light from an optical fiber with a 45° angle mirror at the tip, introduced from the side at different cortical depths in an anesthetized WT mouse. A histological coronal section showing the fiber insertion track (red dye) and DAPI stain is shown on top left. (B) Example bright field image of the skull surface (left) and fluorescence image of fiber-emission (right). Fiber tip at 0.2 mm depth. (C) Radial profiles of fluorescence intensity for an example mouse (outside and at 0.2-1.0 mm inserted depths). (D) Decrease in mean peak fluorescence intensity (peak 10% intensity) with depth. n = 3 mice; error bars are SEM. (E) Wide-field fluorescence images for two mice with the fiber source placed at different depths. Images are min-max normalized. (F) Radial profiles of the images shown in E, normalized to the central peak amplitude. (G) Top: FWHM of the experimentally determined radial profiles as a function of depth (n = 3 mice; error bars are SEM). Bottom: Corresponding FWHM values obtained from simulated data.

Comparison of depth-dependent surface blurring due to scattering for experiment and simulations.

(A) Simulated scattering of 100 photons launched from a 50-μm core diameter multi-modal fiber positioned 0.4 mm beneath the skull (dashed green line). (B) Distribution of photon detection across the surface based on simulations shown in (A) for 500,000 photons with a Lorentzian distribution fit (green line). (C) Distribution of experimentally measured light intensity across the surface of mouse 2 with a Lorentzian fit (green line). (D) Same as (A) but for 100 photons launched from a fiber 1.0 mm beneath the skull. (E) Same as (B) but for 1.0 mm fiber depth. (F) Same as (C) but for 1.0 mm fiber depth.

Deconvolution of single-whisker evoked maps using depth-dependent spatial kernels.

(A) Radial profiles (left) and corresponding deconvolution kernels (right) for L2/3 and L5 generated from Figure 4 (n = 3 mice per layer). (B) Schematic of single-whisker stimulation targeting B1 and E1 whiskers (right) and their corresponding barrel columns (left). (C) Example maps of B1 (left) and E1 (middle) stimulation in a L2/3 mouse (n= 20 trials); composite map (right) shows overlay of activation peaks. (D) Same as C for a L5 mouse. Zoomed-in examples of B1 (top) and E1 (bottom) stimulation in a L2/3 mouse: original images (left) and deconvolved images using the L2/3 kernel from A (right). Color scale: min–max normalized ΔF /F. (F) Radial profiles of original (grey; B1: FWHM = 0.72 mm, E1: 0.49 mm) vs. deconvolved (blue; B1: 0.34 mm, E1: 0.15 mm) images from E. Shaded box indicates approximate barrel width in L2/3. (G) Average radial profiles of original vs. deconvolved images (B1 & E1 pooled; n = 510 trials, 17 sessions, 2 L2/3 mice; error bars = SEM). (H) Mean FWHM values of profiles in G (original vs. deconvolved; error bars = SEM). (I–J) Same as E–F for a L5 mouse: original (grey; B1: 0.84 mm, E1: 0.87 mm) and deconvolved (blue; B1: 0.23 mm, E1: 0.30 mm) images, using the L5 kernel. (K–L) Same as G–H for L5 (B1 & E1 pooled; n = 360 trials, 2 mice). ***p < 0.001; Wilcoxon signed-rank test.

Layer-specific imaging of resting-state activity and cross-regional correlations.

(A) Calcium activity maps (top) and raw ΔF /F traces in 26 cortical regions (below) from resting state measurements in a L2/3 (left), L5 (middle), and L6 (right) mouse. Example data from 10-s recording periods for each mouse. Gray-shaded areas indicate 0.3-s averaging periods, from which calcium activity maps were calculated. Bottom traces depict simultaneously recorded body movements. (B) Schematic of labeled regions in the left hemisphere with coloring of regional groups as in A. (C) Average region-to-region cross-correlation matrices for L2/3 (top; n = 6 mice), L5 (middle; n = 6), and L6 (bottom; n = 6). Regional grouping as in A and B; corresponding matrix modules are indicated by gray squares.

(A) Example cross-regional correlation matrices from an L2/3 mouse. (A) Forelimb stimulation under light anesthesia (n = 20 trials). Imaging was performed using dual-channel data (470 nm, 405 nm) to correct for hemodynamic artifacts. Left: correlation matrix from 470-nm channel without correction. Middle: correlation matrix after hemodynamic correction using the 405-nm channel. Right: difference matrix (uncorrected – corrected). Correlation matrices with and without correction were highly similar during anesthesia, as confirmed by Mantel–Spearman tests (r = 0.99, ***p < 0.001; 500 permutations). (B) Resting-state data from an awake L2/3 mouse. Left: correlation matrix from 470-nm channel without correction. Middle: correlation matrix after hemodynamic correction. Right: difference matrix. Correlation matrices with and without correction were also highly similar during the awake resting state (r = 0.77, ***p < 0.001; 500 permutations).

Layer-specific functional connectivity and anatomical comparison.

(A) Average functional connectivity matrices for L2/3 (left; n = 6), L5 (middle; n = 6), and L6 (right; n = 6), showing cross-correlations of GMR-corrected ΔF /F traces across 26 cortical regions during rest. Matrices were highly similar (Mantel–Spearman: L2/3–L5, r = 0.96; L2/3–L6, r = 0.93; L5–L6, r = 0.93; 500 permutations; all ***p < 0.001). (B) Anatomical connectivity matrices for the same 26 regions in three mouse lines with layer-specific expression in the source regions. Each row contains the connection strength from the source region to all target regions (columns). Connection strength is expressed as log10-transformed normalized projection volumes (NPV). Axonal projections to all layers were considered in the target regions. Rows for which experimental data were missing (often because of low Cre expression) are indicated in grey. Data from (Harris et al., 2019).

Functional connectivity matrices obtained with GMR using dual-channel imaging data (470 nm, 405 nm) from the same L2/3 mouse as in (Figure 5, figure Supplement 1).

(A) Average functional connectivity matrices for forelimb stimulation under light anesthesia (n = 20 trials) without (left) and with (middle) hemodynamic correction using the 405-nm channel. Right: difference matrix obtained by subtracting the corrected from the uncorrected matrix. Connectivity matrices with and without hemodynamic correction were highly similar during anesthesia, as confirmed by Mantel–Spearman tests (r = 0.99, ***p < 0.001; 500 permutations). (B) Functional connectivity matrices from resting-state periods without (left) and with (middle) hemodynamic correction using the 405-nm channel. Right: difference matrix obtained by subtraction. Connectivity matrices with and without hemodynamic correction were also highly similar during awake resting state, as confirmed by Mantel–Spearman tests (r = 0.90, ***p < 0.001; 500 permutations).

Layer differences in resting state functional connectivity.

(A) Differences between the GMR-corrected functional connectivity matrices for L2/3 and L5 (left), L2/3 and L6 (middle), and L5 and L6 (right). Only areas with adjusted p < 0.001 are shown (FDR-corrected, 10,000 permutations). Regional grouping as in Figure 5. (B) Left: Core regions of the default mode network as described in (Whitesell et al., 2021). Right: Topographical distribution of the number of region pairs with significant differences in functional connectivity (positive or negative) for each area comparing L2/3 to L5, L2/3 to L6, and L5 to L6 (from left to right).