For images and movies of visually evoked GCaMP3 calcium Sotrastaurin manufacturer activity in neurons or axons in awake mouse visual cortex through a prism (Figures 3, 4, and 5), data were collected at 1 Hz (256 × 512 pixels spanning 600 μm in depth × 900 μm across) using ScanImage (Pologruto et al., 2003). For high-speed imaging using a resonance scanning system (Figure 6), images were collected at ∼32 Hz (256 × 240 pixels) using custom software (Bonin et al., 2011) written in MATLAB (MathWorks). In all cases, each image was coregistered offline to a common target from the same session, using custom software

in MATLAB (rigid-body translations; Andermann et al., 2011 and Glickfeld et al., 2013). The resonance scanning system was also used for volume imaging Everolimus solubility dmso of layer 2/3 neurons through the cranial window (Figure 2) using procedures described previously (Glickfeld et al., 2013; briefly, 25× objective (Olympus) and a piezo z scanner (Physik), scanning 31 planes separated by 3 μm (1 volume per second). During prism imaging, laser power at the objective was <70 mW (typical: ∼45 mW at 205 μm from the prism face), with the exception of axon imaging at 1 day postimplant (Figures 5D–5F: 130 mW was used due to residual blood at prism and brain surfaces). Visual stimuli consisting of local 40° Gabor-like circular patches (sigmoidal 10%–90% falloff in 10°) containing drifting

square waves of varying orientation, spatial, and temporal frequency (see Figures 3 and 4 and associated text; 5 s duration, with 5 s of mean luminance between second stimuli, 8–12 repetitions, pseudorandom order) were presented on a 120 Hz LCD monitor (Samsung) that was calibrated using a spectrophotometer (Photoresearch PR-650). The total recording time per session was 3–5 hr. The evoked responses for each of the 96 stimulus types was defined for each pixel in the imaging volume as the fractional change in fluorescence (ΔF/F) between [−2 s, 0 s] and [0 s, 5 s] following onset of the stimulus, averaged across trials (Andermann et al., 2011). Because baseline GCaMP3 fluorescence

is sometimes dim (Andermann et al., 2011), cell masks were extracted from the maximum intensity projection of average response images (ΔF/F) across all stimulus types, using custom semiautomated segmentation algorithms. Extraction of three-dimensional cell masks in Figure 2, and analyses of active axonal boutons in Figures 5D–5F are described in detail in Glickfeld et al. (2013). For analyses in Figure 6 of V1 activity as mice stood or ran on a linear trackball (Andermann et al., 2011) in near-complete darkness, cell masks were extracted from the projection across all endogenous fluorescence transients within the 20 min recording (projection of (F(t) − Fmedian)/Fmedian across the image stack).