In vivo measurement of afferent activity with axon-specific calcium imaging

Abstract

In vivo calcium imaging from axons provides direct interrogation of afferent neural activity, informing the neural representations that a local circuit receives. Unlike in somata and dendrites, axonal recording of neural activity-both electrically and optically-has been difficult to achieve, thus preventing comprehensive understanding of neuronal circuit function. Here we developed an active transportation strategy to enrich GCaMP6, a genetically encoded calcium indicator, uniformly in axons with sufficient brightness, signal-to-noise ratio, and photostability to allow robust, structure-specific imaging of presynaptic activity in awake mice. Axon-targeted GCaMP6 enables frame-to-frame correlation for motion correction in axons and permits subcellular-resolution recording of axonal activity in previously inaccessible deep-brain areas. We used axon-targeted GCaMP6 to record layer-specific local afferents without contamination from somata or from intermingled dendrites in the cortex. We expect that axon-targeted GCaMP6 will facilitate new applications in investigating afferent signals relayed by genetically defined neuronal populations within and across specific brain regions.

Figure 1:. Axonal enrichement of GCaMP sensor driven by GAP43 targeting motif in dissociated cells and in vivo.

(a) Schematic representation of untargeted GCaMP6 (u-GCaMP6), vesicle protein synaptophysin-fused GCaMP6 (syGCaMP6) and GAP43-targeted GCaMP6 (axon-GCaMP6). (b) Representative images showing enhanced axon localization of axon-GCaMP6m compared to u-GCaMP6m in dissociated neuronal culture. Results were found to be consistent across 3 different cultures per construct. Scale bars: insets 20μm, cell traces 100μm. (c) Normalized Axon to Dendrite Ratio (nADR, see Methods for description). Average nADRs of axon-GCaMP6m (4.83±0.78) represent an approximate 5-fold increase compared to u-GCaMP6m (0.76±0.08) or syGCaMP6m (1.77±0.11). Data as mean ± s.e.m. F(2,27) = 26.52, P = 8.36E-7, n.s. = 0.23, ****P = 3.09E-5, One-way ANOVA with Tukey-Kramer multiple comparisons test; n = 8–10 cells from three cultures for each construct. (d-f) Characterization of ADR of thalamocortical axons in V1. (d) Schematic representation of viral injection site of adeno associated virus encoding biscistronic constructs in mouse dLGN. (e) Representative images showing densely labeled L4 axons projected from dLGN labeled with axon-GCaMP6s, whereas cell somata were nearly devoid of labeling. Results were consistent across four animals per construct. Scale bars: 100μm (cortex), 500μm (thalamus). (f) Histogram of nADR values derived for both constructs with medians indicated by vertical line. Axon-GCaMP6s, 2.54±0.28; u-GCaMP6s, 0.50±0.06; data as median ± s.e.m.; P = 4.37E-16, Wilcoxon’s rank-sum test; n = 56 slices analyzed for axon-GCaMP6s and n = 37 for u-GCaMP6s from 3 animals per construct.

Figure 2:. Axon-GCaMP maintains photostability and displays enhanced SNR.

(a-e) Characterization of photostability and diffusibility. Schematic representation of the FRAP (a) and FLIP (d) experiments. (b) Representative pseudo-linescans from u-GCaMP6m, axon-GCaMP6m, and syGCaMP6m FRAP demonstrated rapid recovery of fluorescence for uGCaMP6m and axon-GCaMP6m, but not in the case of the syGCaMP6m. Experiment was repeated on two independent cultures per construct with similar results. (c) Mobile fraction of uGCaMP6m, syGCaMP6m, and axon-GCaMP6m as assessed by FRAP. Data plotted as mean ± s.e.m. X²(2,11) = 9.85, P = 0.0073; *n.s. = 0.26, ***P = 0.0048, Kruskal-Wallis test with Dunn’s test for multiple comparisons; n = 4 cells from 2 plates per construct. (e) Time course of normalized fluorescence intensity changes during FLIP experiments. Data plotted as mean (solid line) ± s.e.m. (shaded regions); n = 3 cells per construct. (f-g) Characterization of photostability in vivo. (f) Schematic representation of viral injection in dLGN followed by in vivo imaging axons projecting to V1. (g) Time course of fluorescence intensity normalized to the first 50 seconds of acquisition during imaging of dLGN projection. Data plotted as mean (solid line) ± s.e.m. (shaded regions); n = 377, 70, and 236 for u-GCaMP6s, syGCaMP6s, and axon-GCaMP6s, respectively. Individual traces were smoothed by a 100 time-point boxcar filter to emphasize low frequency components of data. (h) (top) Schematic representation of bicistronic constructs expressed in dissociated neuronal culture. (bottom) Representative images showing increased green fluorescence in axons when labeled with axon-GCaMP6m, whereas u-GCaMP6m primarily labels somato-dedritic compartments of neurons. Experimental data was derived from 3 independent cultures per construct with similar results. Scale bar: somatic image 10μm, axonal image 3μm. (i-j) Average ΔF/F (i) and SNR (j) in response to indicated number of field potential stimuli. Inset shows reduced baseline noise in single-trial time-lapse traces of axon-GCaMP6s in response to 20FP stimuli. Data plotted as mean peak values ± s.d.; n = 9 imaging sessions from 3 transductions per construct.

Figure 3:. Axon-GCaMP6s improves SNR and image-wise correlations in long-distance axons.

(a, f) Schematic representation of viral injection in dLGN (a) and LP (f), followed by in vivo imaging in V1 of projected axons. (b, g) Representative images of L1 axons projected from dLGN (b) and LP (g) demonstrating enhanced brightness of axon-GCaMP6s expressing axons. Experiments were performed in at least 3 animals per construct injection site pair with similar results. Scale bars:10 μm. (c,h) (left) Fluorescence of a representative ROI in response to the indicated grating directions (10 trials per direction) of axon-GCaMP6s and u-GCaMP6s in dLGN (c) and LP (h). Data presented with averages in dark colors and individual trials in light colors. Scale bars 3 seconds, 100% ΔF/F. (right) Polar plots of tuning properties of the selected ROIs. (d, i) Trial-averaged traces of SNR time course from all responsive ROIs to preferred stimulus direction. Point at which moving stimulus presented indicated by vertical dotted black line. (e, j) (top) Cumulative distribution of normalized fluorescence intensity of axon-GCaMP6s and u-GCaMP6s in dLGN (e) and LP (j) axons projecting to V1. (dLGN: axon-GCaMP6s, 5.85; u-GCaMP6s, 0.85; LP: axon-GCaMP6s, 19.47; u-GCaMP6s, 5.08; P = 2.15E-259 and 1.24e-319, respectively by the Kolmogorov-Smirnov test). (bottom) Cumulative distribution of SNR of individual ROIs with median values indicated by vertical lines as well as color-coded numeric values (dLGN: axon-GCaMP6s, 2.73; u-GCaMP6s, 1.54; LP: axon-GCaMP6s, 1.07; u-GCaMP6s, 0.64; P = 9.56E-130 and 1.13E-95, respectively by the Kolmogorov-Smirnov test). For (e) n=712 ROIs from 4 animals for axon-GCaMP6s and n=14478 ROIs for u-GCaMP6s from 19 animals. For (j) n=1393 ROIs from 3 animals for axon-GCaMP6s and n=917 ROIs from 4 animals for u-GCaMP6s. (k-m) axon-GCaMP6s improves frame to frame correlations. (k) Representative frames from one imaging session per construct in L1 dLGN experiments showing enhancements to image structure for axon-GCaMP6s. The text above and below the images demonstrates the procedure for calculating the frame-to-frame correlations. Scale bar: 10μm. (l, m) Axon-GCaMP6s permits significantly improved frame-to-frame correlations for image registration. Time course of image-wise correlation values for dLGN (l) and LP (m) boutons. Data plotted as mean (solid line) ± s.e.m. (shaded regions); n = 6 imaging sessions from two animals for each construct.

Figure 4:. Imaging thalamocortical afferents deep in cortical tissue with axon-GCaMP6s

(a) Experimental schematic demonstrating that brighter axon-GCaMP6s axons can be imaged at deeper layers within tissue than u-GCaMP6s axons. (b) Comparison of axon-GCaMP6s and u-GCaMP6s fluorescence normalized to the square of the excitation laser power as a function of imaging depth (Data plotted as individual ROIs with binned values (light color) and binned means ± s.e.m (dark colors), and log-linear fit across the depth). F(1,1396) = 1242.35, P < 4.94E-324, ANCOVA. For u-GCaMP6s we analyzed 500 ROIs from 2 imaging sessions per depth from each of 4 animals, for axon-GCaMP6s 900 ROIs from 2 imaging sessions per depth from 3 animals. (c) Representative images of dLGN axons projecting to V1 with overlays indicating analyzed ROI location at 580μm (top) and 600μm (bottom) below pia. Results were similar across three tested animals. Scale bars 3 μm. (d-e) Response properties of individual ROIs deep in tissue. (d) Average ΔF/F traces aligned to stimulus direction of ROIs indicated in (c) with baseline indicated by dotted black line. Data plotted as mean (black line) ± s.e.m. (gray shading). n = 10 trials per direction. (e) Tuning curves for individual ROIs.

Figure 5:. Axon-GCaMP6s permits layer specific recording of axons projected from local neurons without contamination from somatodendritic signals.

(a) Schematic representation of labeling L4 neurons with Cre-dependent adeno associated virus encoding axon-GCaMP6s or u-GCaMP6s in Scnn1a-Cre-Tg3 mice. (b) Expected pattern of transduction: axon-GCaMP6s strongly labeling axons with weak labeling in the somatodendritic compartment and the reverse pattern for u-GCaMP6s. (c) Representative images showing expression patterns of axon-GCaMP6s (top) or u-GCaMP6s (bottom) fused to P2A-mRuby3. Axon-GCaMP6s is enriched in axons, whereas u-GCaMP6s more strongly labels somatodendritic compartments. Inset shows representative zoomed-in images in L1 and L4. This result was consistently found in two animals per construct. Scale bar: somatic image 100μm, inset images 10μm. (d-e) Distributions of green-to-red ratio at axons (d) or dendrites (e). Distribution medians indicated by vertical lines. Axon-GCaMP6s displayed significantly enhanced green-to-red ratio in axons, but dramatically decreased green-to-red ratio in dendrites, compared to u-GCaMP6s. P < 4.94E-324 and P = 1.57E-26, Wilcoxon’s rank sum test; data from n = 1200 ROIs from 2 animals for both constructs. (f-i) Axon-GCaMP6s enabled layer-specific tuning properties of axons across cortical layers in V1. Data were derived from 2 imaging sessions per depth in 2 animals. (f) Representative images of axons at indicated depth with overlays indicating analyzed ROIs. (g) Color-coded tuning map demonstrating pixel-wise tuning across analyzed images. (h) Average ΔF/F traces aligned to stimulus direction of individual ROIs with baseline indicated by dotted black line. Data plotted as mean (black line) ± s.e.m. (gray shading). n = 10 trials per direction. (i) Tuning curves for individual ROIs.