Projection-Specific Visual Feature Encoding by Layer 5 Cortical Subnetworks.

Animals Adolescent (6–8 week) wild type C57/bl6 mice (Charles River Laboratories) were used in accordance with the Yale Institutional Animal Care and Use Committee and federal guidelines. In vivo imaging GCaMP6s was expressed in V1 using an adenoassociated virus vector (AAV2- hSynapsin1 -GCaMP6s, serotype 5, University of Pennsylvania Vector Core). Projection-specific subtypes of L5 PNs were labeled using CTB-Alexa Fluor-555 injected into the SC, dStr, or cV2. Imaging was performed 25–30 days after injection under light isoflurane anesthesia through an acutely implanted glass cranial window. Imaging was performed using a resonant scanner-based two-photon microscope (MOM, Sutter Instruments) coupled to a Ti:Sapphire laser (MaiTai DeepSee, Spectra Physics) tuned to 940 nm for GCaMP6 and 1000 nm for CTB-Alexa Fluor 555. Images were acquired using ScanImage 4.2 (Vidrio Technologies) at ~30 Hz from a depth of ~450–600 µm relative to the brain surface. Visual stimuli consisted of full-screen sinusoidal drifting gratings with a temporal frequency of 1 Hz and with varied contrast, orientation, and spatial frequency. For all experiments, visual stimuli were 3 seconds in duration and separated by an inter-stimulus interval of 5 seconds.

Data analysis

Analysis was performed using custom-written routines in MATLAB (The Mathworks) and IgorPro (Wavemetrics). Regions of interest (ROIs) corresponding to single cells were selected as previously described ( Chen et al., 2013 ). Ca2+ signals in response to visual stimuli were averaged and expressed as ΔF/F. A cell was classified as visually responsive if the Ca2+ signals during stimulus presentation were statistically different from the signals during five blank periods (p0.4. Noise correlations were calculated as the partial correlation coefficient between pairs of cells.

Statistical analysis

For most analyses, we developed a method of using semi-weighted estimators to compare individual animals, rather than cells ( Chung et al., 2013 , DerSimonian and Laird, 1986 ). This approach minimizes false positives while maintaining statistical power ( Aarts et al., 2014 ). We used this semi-weighted estimator to calculate the statistical significance of the difference between cell populations using a standard Student’s t-test. The only exception to this was the noise correlation analysis in Figure 4 , where we used the weighted estimator to reflect the pair-wise nature of the comparisons.