Projection-Specific Visual Feature Encoding by Layer 5 Cortical Subnetworks

Abstract

Primary neocortical sensory areas act as central hubs, distributing afferent information to numerous cortical and subcortical structures. However, it remains unclear whether each downstream target receives a distinct version of sensory information. We used in vivo calcium imaging combined with retrograde tracing to monitor visual response properties of three distinct subpopulations of projection neurons in primary visual cortex. Although there is overlap across the groups, on average, corticotectal (CT) cells exhibit lower contrast thresholds and broader tuning for orientation and spatial frequency in comparison to corticostriatal (CS) cells, whereas corticocortical (CC) cells have intermediate properties. Noise correlational analyses support the hypothesis that CT cells integrate information across diverse layer 5 populations, whereas CS and CC cells form more selectively interconnected groups. Overall, our findings demonstrate the existence of functional subnetworks within layer 5 that may differentially route visual information to behaviorally relevant downstream targets.

Figure 1. CT cells exhibit lower visual detection threshold than CC and CS neurons

(A) Schematic of in vivo 2-photon Ca2+ imaging of labeled L5 PN populations. (B) Example field of view. Green somata express GCaMP6s. Magenta cells express GCaMP6s and are retrogradely labeled with red fluorescent CTB-Alexa Fluor-555. (C) Example raw traces recorded from cells indicated in (B) and corresponding EEG signal. (D) Example ΔF/F traces (black) and de-trended visual responses (blue) with best fit sine waves (red) to calculate modulation index (MI). (E) Bars represent mean ± SEM modulation index for CT (blue) CC (green) and CS (red) cells. (F) Example raw (gray) and average (black) ΔF/F traces recorded at varying contrast values. (G) Hyperbolic ratio function fit (red) to contrast response (black circles). Dashed lines highlight c50 and R_max points. (H) Bars represent mean ± SEM c50 values of CT (blue) CC (green) and CS (red) cells. (I) Bars represent mean ± SEM exponent values of CT (blue) CC (green) and CS (red) cells. *: p<0.05, Student’s t-test, semi-weighted statistics (see Methods).

Figure 2. CT neurons are more broadly tuned for orientation than CC and CS cells

(A) Example raw (gray) and average (black) traces of CT (top), CC (middle) and CS (bottom) neurons at varying orientations. (B) Polar plots indicating the orientation tuning of the cells in (A). (C) Distribution of OSI values for CT, CC, and CS populations. (D) Bars represent mean ± SEM OSI values of CT (blue) CC (green) and CS (red) cells. (E) Bars represent mean ± SEM orientation tuning width of CT (blue) CC (green) and CS (red) cells. *: p<0.05, Student’s t-test, semi-weighted statistics (see Methods).

Figure 3. CC and CT neurons filter spatial frequencies at a broader band than CS cells

(A) Example raw (gray) and average (black) traces of a CT (top), CC (middle) and CS (bottom) neurons at varying spatial frequencies. (B) Gaussian curves (red) fit over spatial frequency data (black circles) from (A) on a log10 scale. (C) Distributions of bandwidths and fractions of low-pass (LP) and high-pass (HP) for CT, CC, and CS cells. (D) Bars represent mean ± SEM fraction of band pass cells in CT (blue) CC (green) and CS (red) populations. (E) Bars represent mean ± SEM preferred spatial frequency of CT (blue) CC (green) and CS (red) cells. (F) Bars represent mean ± SEM spatial frequency bandwidth of CT (blue) CC (green) and CS (red) cells. *: p<0.05, Student’s t-test, semi-weighted statistics (see Methods).

Figure 4. CC and CS neurons form local subnetworks

(A) Heat map showing the strength of partial noise correlations between pairs of labeled CC neurons within an example field of view. (B) Heat map showing partial noise correlations between pairs of labeled CC and non-identified (NI) neurons in the same field of view as in (A). (C) 2-photon fluorescent image of the field of view in (A) and (B) highlighting visually responsive CC (white circles) and NI (gray circles) neurons. (D) Web graph showing the connections and correlation strength between CC neurons in the same field of view as in (A–C). (E) Web graph showing the connections and correlation strength between CC and NI neurons in the same field of view as in (A–D). (F) Bars representing mean ± SEM correlation strength between CT-CT (dark blue), CT-NI (light blue), CC-CC (dark green), CC-NI (light green), CS-CS (dark red) and CS-NI (light red) cell pairs. *: p<0.05, paired t-test. (G) Change in correlation strength with distance between CT-CT (dark blue), CT-NI (light blue), CC-CC (dark green), CC-NI (light green), CS-CS (dark red) and CS-NI (light red) cell pairs. *: p<0.05, paired t-test. (H) Change in correlation strength related to the degree of co-tuning for orientation between CT-CT (dark blue), CT-NI (light blue), CC-CC (dark green), CC-NI (light green), CS-CS (dark red) and CS-NI (light red) cell pairs. *: p<0.05, paired t-test.