Soma-Targeted Imaging of Neural Circuits by Ribosome Tethering

Abstract

Neuroscience relies on techniques for imaging the structure and dynamics of neural circuits, but the cell bodies of individual neurons are often obscured by overlapping fluorescence from axons and dendrites in surrounding neuropil. Here, we describe two strategies for using the ribosome to restrict the expression of fluorescent proteins to the neuronal soma. We show first that a ribosome-tethered nanobody can be used to trap GFP in the cell body, thereby enabling direct visualization of previously undetectable GFP fluorescence. We then design a ribosome-tethered GCaMP for imaging calcium dynamics. We show that this reporter faithfully tracks somatic calcium dynamics in the mouse brain while eliminating cross-talk between neurons caused by contaminating neuropil. In worms, this reporter enables whole-brain imaging with faster kinetics and brighter fluorescence than commonly used nuclear GCaMPs. These two approaches provide a general way to enhance the specificity of imaging in neurobiology.

Keywords: C. elegans; GCaMP; calcium imaging; ribo-GCaMP; ribosomal tagging; soma-targeting; whole-brain imaging.

Figure 1.. Ribosome Tethering Enables Ultrasensitive Visualization of GFP Reporters

(A) Synthetic proteins tagged with ribosomal unit L10a become tethered to ribosomes and are enriched in the soma. (B) The Nano-L10 transgenic mouse expresses a GFP-binding nanobody fused to ribosomal protein L10 in all neurons. When crossed to a second transgenic line expressing GFP from a cell-type-specific promoter, the GFP is captured by the nanobody, anchoring it in the soma. (C–H) Images from the brains and peripheral ganglia of GFP reporter mice, either alone (bottom row) or after crossing to the Nano-L10 mouse (top row). All of the images show direct GFP fluorescence, except for (D), which is immunostained for GFP. Panels show Agtr1a-GFP (C), GAD67-GFP (D), GCG-GFP (E), TRPV1-GFP (F), TRPM8-GFP (G), and TrkB-tauGFP (H). Scale bar sizes are indicated in each panel. See also Figure S1.

Figure 2.. Ribo-tagging Restricts GCaMP to the Neuronal Soma

(A and B) Linking GCaMP to ribosomal protein L10a (A) enables soma targeting of GCaMP fluorescence (B). (C) Ribo-tagging restricts GCaMP6f to the soma of PVH-SIM1 neurons. (D and E) Ribo-GCaMP6m, when expressed in medial prefrontal cortex (mPFC), dentate gyrus (DG), and superior colliculus (SC) (D), is soma restricted compared to regular GCaMP6m (E) (white box indicates magnified region). (F) Comparison of ribo-GCaMP6m and regular GCaMP6m when expressed in DG. (G) Fluorescence intensity from the start of the granular layer to the molecular layer (250 μm length) of DG in samples expressing ribo-GAMP6m (n = 11) or regular GCaMP6m (n = 9). Sample size indicates number of brain slices used in analysis from 3 animals for each GCaMP6m. Each sample trace is normalized to its maximal intensity. (H) Cumulative distribution function (CDF) of normalized fluorescence of ribo-GCaMP6m and GCaMP6m in the molecular layer, where DG dendrites are located. (I) Comparison of regular GCaMP6m and ribo-GCaMP6m in CA3, where the axons from DG are located. (J) Z projection of an individual striatal Ntrk1 neuron expressing mCherry and GCaMP6m (right) or ribo-GCaMP6m (left). (K) Intensity of green fluorescence along the axon of Ntrk1 neurons expressing either GCaMP6m (black, n = 8) or ribo-GCaMP6m (green, n = 8). Tracking the axon was guided by mCherry fluorescence. **p < 0.01, ***p < 0.001, ns, p > 0.05, 2-way ANOVA, Holm-Sidak post hoc test. All of the images are immunostained against GCaMP. Scale bar sizes are indicated in each panel.

Figure 3.. Ribo-tagging Does Not Change the Kinetics of GCaMP in Response to Evoked Neural Activity

(A) Schematic describing the measurement of fluorescence in response to electrically evoked neural activity in DG. Grayscale heatmap shows the brightness of DG neurons expressing ribo-GCaMP6m before and during pulsed field stimulation. (B) Response of ribo-GCaMP6m and regular GCaMP6m to different numbers of electrical field pulses (FPs). When calculating ΔF/F, the average fluorescence during the 300-ms time window before the start of the first FP was used as F0 (the denominator). (C–E) Peak signal (C; p 0.71, 0.66, 0.65, 0.60, 0.67, < 0.01, 0.67, and 0.16), decay time constant (D; p 0.95, 0.27, 0.96, 0.95, 0.65, 0.87, < 0.01, and < 0.0001), and rise time constant (E; p 0.84, 0.84, 0.25, 0.84, 0.78, 0.43, < 0.0001, and 0.06) for ribo-GCaMP6m (green, n = 304 from 4 brain slices of 2 mice) and GCaMP6m (black, n = 137 from 4 brain slices of 2 mice) in response to the increasing numbers of FPs.

Figure 4.. Subcellular Distribution and Kinetics of Ribo-GCaMP

(A and B) 2P z stacks of layer 5 pyramidal neurons expressing GCaMP6m (A) or ribo-GCaMP6m (B). Left, overlay of Alexa 594 cell fill and and non-calcium-bound GCaMP fluorescence (810 nm imaging wavelength). Right, overlap of Alexa 594 and GCaMP fluorescence. Scale bars are 10 μm. (C and D) Example 2P line scans of (C) GCaMP6m and (D) ribo-GCaMP6m fluorescence in response to trains of 5 APs (imaging wavelength 920 nm). Example transients are averages of 10 trials. Scale bars are 50 μm. (E) Response to 5 APs measured at the soma and 50 μm along the apical dendrite for GCaMP6m (n = 15) and ribo-GCaMP6m (n = 14). Bars indicate means ± SEMs, ****p < 0.0001, 1-way ANOVA, Holm-Sidak correction. (F) Normalized peak fluorescence imaged at varying distances along the apical dendrite in response to 5 APs (10, 20, 30, 50, 100, 150, and 200 μm; normalized to signal at 0 μm) for GCaMP6m (n = 7–15) and ribo-GCaMP6m (n = 6–14). Lines and bars indicate means ± SEMs. ****p < 0.0001 for all comparisons between GCaMP6m and ribo-GCaMP6m, 2-way ANOVA, Holm-Sidak correction. (G) Response at the soma to 1, 3, 5, and 10 APs of GCaMPm (n = 8–18) and ribo-GCaMP6m (n = 7–14). p = 0.73, 0.06, 0.73, and 0.10 for 1, 3, 5, and 10 APs), 2-way ANOVA, Holm-Sidak correction. (H) Standard deviation of the baseline fluorescence signal for GCaMP6m (n = 16) and ribo-GCaMP6m (n = 14). p = 0.67, 2-tailed unpaired t test. (I) Fold increase of somatic fluorescence signal over the standard deviation of the baseline in response to 5 APs for GCaMP6m (n = 16) and ribo-GCaMP6m (n = 14). p = 0.59, 2-tailed unpaired t test. (J) Time to peak after stimulus offset for GCaMP6m (n = 8–18) and ribo-GCaMP (n = 6–14). p = 0.56, 0.82, and 0.56 for 3, 5, and 10 APs, 2-way ANOVA, Holm-Sidak correction. (K) Half-maximal decay time for GCaMP6m (n = 4–13) and ribo-GCaMP6m (n = 4–12). p = 0.14, 0.01, and 0.13 for 3, 5, and 10 APs, 2-way ANOVA, Holm-Sidak correction. See also Figure S2.

Figure 5.. Ribo-GCaMP Reduces Neuropil Contamination during 2P Imaging

(A) Example fields of view of V1 neurons expressing GCaMP6m (top) or ribo-GCaMP6m (bottom). (B) Calcium dynamics of neurons recorded from a single video. Calcium dynamics are either uncorrected (left) or with surrounding neuropil signal subtracted (right). Cells are sorted based on orientation tuning. (C) Color coded ROIs based on orientation selectivity in V1. (D) Peak response of angle-selective V1 neurons to drifting bars with different angles. (E) Distribution of orientation selectivity. ROIs are either neurons or neuropils, which is defined as the area surrounding each identified neuron. (F–I) Correlation coefficient analysis of neural signal for either GCaMP6m or ribo-GCaMP6m. The neuropil signal was either subtracted or not, as indicated. During recording, mice were exposed to either drifting bars (F and G) or white noise (H and I). (F) Correlation coefficient of calcium dynamics between pairs of neurons plotted against the centroid distance. R² = 0.92, 0.19, 0.84, and 0.10; p < 0.0001, 0.0208, < 0.0001, and 0.0970 for regular uncorrected, ribo-uncorrected, regular, and ribo, respectively, linear regression of averaged values. (G) Correlation coefficient of calcium dynamics between pairs of neurons that are within 100 pixels. Regular corrected: 0.105 ± 0.010; ribo-corrected: 0.081 ± 0.010; regular uncorrected: 0.227 ± 0.010; ribo-uncorrected: 0.085 ± 0.008; values are means ± SEMs. (H) Correlation coefficient of calcium dynamics between pairs of neurons plotted against centroid distance. R² = 0.80, 0.003, 0.62, and 0.003; p < 0.0001, 0.79, < 0.0001, and 0.79 for regular uncorrected, ribo-uncorrected, regular, and ribo, respectively, linear regression of averaged values. (I) Correlation coefficient of calcium dynamics between pairs of neurons that are within 100 pixels. Regular corrected: 0.034 ± 0.002; ribo-corrected: 0.021 ± 0.002; regular uncorrected: 0.109 ± 0.002; ribo-uncorrected: 0.024 ± 0.002; values are means ± SEMs. In (C)–(E), arrows indicate the drifting bar orientation that the corresponding color (C and E) or column (D) represents. In (C) and (E), N denotes “nonselective,” which means that these neurons did not pass the criteria to be defined as orientation tuned (see Method Details). In (D), B denotes “blank,” in which no stimulus was given. In (G) and (I), **p < 0.01, ****p < 0.0001, ns, p > 0.05, 1-way ANOVA, Holm-Sidak post hoc test. Total numbers of cells were used to estimate degrees of freedom when calculating SEM; 1 pixel = ~1.9 μm. See also Figure S3 and Videos S1, S2, S3, and S4.

Figure 6.. Microendoscope Recording of mPFC Dynamics Using Ribo-GCaMP

(A) Ribo-GCaMP6m expression was targeted to mPFC, and a GRIN lens was implanted for microendoscope recordings. (B) Representative field of view of ribo-GCaMP6m imaged through a microendoscope, with ROIs (pink contours) indicated. (C) Representative extracted calcium dynamics (black traces) time synchronized with licking events (pink shading) of liquid food in the fasted state. (D) Response of mPFC neurons in fasted mice to the consumption of Ensure. Red bar indicates when food was made available. (E) Peristimulus time histogram (PSTH) aligned to either the initiation or termination of licking bouts. Calcium dynamics were measured using either ribo-GCaMP6m (left) or GCaMP6m (right). (F and G) Pearson correlation coefficient between the signal from different pairs of ROIs. mPFC neural dynamics were recorded in fed mice expressing either regular GCaMP6m (n = 154 from 5 mice) or ribo-GCaMP6m (n = 69 from 4 mice) during cage exploration. In (F), calcium dynamics were extracted using either PCA-ICA or CNMF_E methods, and the correlation coefficient is plotted against the distance between corresponding ROIs. In (G), calcium dynamics were extracted by PCA-ICA and the correlation coefficient between pairs within 100 pixels is compared for GCaMP6m and ribo-GCaMP6m. *p < 0.05, unpaired t test; cell number was used as degree of freedom when calculating SEMs.

Figure 7.. Stimulated In Vivo Calcium Imaging of C. elegans ASH and AFD Neuronal Activity with Ribo-GCaMP

(A) Schematic representation of ASH and AFD neurons in the worm head. Boxes indicate ASH cell body, AFD cell body, and ASH dendrite, respectively. (B) Animals were imaged in a microfluidic device that allows switchable fluid flow past the nose. (C) Response amplitude ΔF/F0 for imaging traces shown in (D)–(F); GCaMP6m localization for soma in gray, dendrite in blue, and nucleus in orange. (D–G, top) Diagram of constructs used to make transgenic worm strains. sra-6 promoter drives expression in ASH; gcy-8 promoter drives expression in AFD. Representative images showing fluorescent responses in ASH and AFD; dotted white lines denote worm head; scale bars, 10 μm. (D–G, center) Mean fluorescence transients in ASH neurons expressing GCaMP6m (D), ribo-GCaMP6m (E), or nuclear-localized (nls)-GCaMP6m (F), in response to a 10-s 0.5 M NaCl pulse (green area denotes stimulus; means ± SEMs). Black and gray traces denote responses in cell soma; blue traces denote responses in dendrites. (G) Mean fluorescence transients in AFD neurons expressing ribo-GCaMP6m in response to a 10-s pulse of warm (~30°C) buffer (red area denotes stimulus). (D–G, bottom) Heat plots of individual soma responses, 1 neuron per row. Change in fluorescence from baseline over time (ΔF/F0) is represented in the color bars at left. n = 11 worms responding to 2 consecutive salt pulses for ASH ribo-GCaMP6m; 10 for ASH soluble GCaMP6m; 5 for ASH nls-GCaMP6m; 11 worms responding for 3 consecutive heat pulses for AFD ribo-GCaMP6m. **p < 0.01, ***p < 0.001, ns, not significant using unpaired t test with Welch’s correction. See also Figure S5.

Figure 8.. Whole-Brain Calcium Imaging of C. elegans Basal Neuronal Activity with Ribo-GCaMP

(A and B) Maximum intensity projection of representative worms recorded under constant conditions expressing pan-neuronal nls-GCaMP6m (A) or ribo-GCaMP6m (B). Dotted white lines denote worm head. Scale bars, 10 μm. (C and D) Representative heat plots of fluorescence (ΔF/F0) 5-min time series for pan-neuronal nls-GCaMP6m (C) or ribo-GCaMP6m (D), 1 neuron per row. Labeled neurons indicate putative cell IDs, PC1⁺ neurons in pink and PC1⁻ neurons in cyan. (E) Histogram of frequency distributions for all rise times of peaks in calcium transients measured for 47 ribo-GCaMP6m traces (from 5 time series recordings) and 60 nls-GCaMP6m traces (from 8 time series recordings). (F) Rise times of peaks in calcium transients measured from the same neuron across recordings expressing either nls-GCaMP6m or ribo-GCaMP6m. ***p < 0.001, ns, not significant using Kolmogorov-Smirnov test.