RNA Docking and Local Translation Regulate Site-Specific Axon Remodeling In Vivo

Abstract

Nascent proteins can be positioned rapidly at precise subcellular locations by local protein synthesis (LPS) to facilitate localized growth responses. Axon arbor architecture, a major determinant of synaptic connectivity, is shaped by localized growth responses, but it is unknown whether LPS influences these responses in vivo. Using high-resolution live imaging, we examined the spatiotemporal dynamics of RNA and LPS in retinal axons during arborization in vivo. Endogenous RNA tracking reveals that RNA granules dock at sites of branch emergence and invade stabilized branches. Live translation reporter analysis reveals that de novo β-actin hotspots colocalize with docked RNA granules at the bases and tips of new branches. Inhibition of axonal β-actin mRNA translation disrupts arbor dynamics primarily by reducing new branch emergence and leads to impoverished terminal arbors. The results demonstrate a requirement for LPS in building arbor complexity and suggest a key role for pre-synaptic LPS in assembling neural circuits.

Keywords: FRAP; RNA labeling; RNA localization; RNA trafficking; axon branching; axon guidance; local protein synthesis; mitochondria; neural wiring; retinotectal projection; β-actin.

Figure 1

Dynamics of Endogenous RNA Granules Correlate with Distinct Aspects of Axon Branching In Vivo (A) RNA granule (white arrowheads) docking in RGC axons during branching. Top: a single RNA granule docks at multiple branch point sites before the formation of protrusions (cyan arrows). The single z plane inset demonstrates localization of the RNA granule at the base of the protrusion. Middle: multiple RNA granules move into branches and to branch tips during protrusion stabilization. Bottom: branch retraction (yellow arrow) occurs shortly after RNA granules exit the branch. (B) Left: proportion of protrusions with RNA docking at the base for >10 s preceding protrusion formation. Right: occurrence of RNA docking in protrusion-forming or random positions in the same axons (t₇ = 21.2, p < 0.0001, paired t test). Red diamonds represent the averages. (C) Time of RNA granules presence was longer in branches with longer lifetime (base of branch: U = 369, p = 0.004; within branch: U = 137.5, p < 0.0001; branch tip: U = 225, p < 0.0001). (D) Time of RNA granules presence was longer in branches with longer maximal branch length (base of branch: U = 240, p < 0.0001; within branch: U = 297, p < 0.0001; branch tip: U = 369, p < 0.0001). Error bars represent SEM. ^∗∗p < 0.01, ^∗∗∗p < 0.001 (Mann-Whitney test for C and D). (E and F) Pearson’s correlation between time of RNA presence and lifetime of branch (E) or maximal branch length (F). Scale bars, 5 μm. See also Figures S1–S3.

Figure 2

Acute Inhibition of Translation Disrupts Axonal Branching Dynamics In Vivo (A) Live imaging experiment on branching dynamics of somaless RGC axons in the tectum in vivo. Electroporated eye was removed to eliminate somatic contribution. (B–D) Axonal branching in control condition (B) and after incubation in translation inhibitors cycloheximide (C; CHX) and anisomycin (D; ANI). A merged overlay of three time points (0′, 5', and 10′ in blue, red, and green, respectively) is shown for each condition (far right). More protrusions were added than removed in control condition (filopodia: t₁₁ = 3.8, p = 0.003; branches: t₁₁ = 4.6, p = 0.0008) (B′ and B″). No significant differences were observed in the number of protrusions that were added and removed in CHX condition (filopodia: t₁₈ = 0.2, p = 0.82; branches: t₁₈ = 1.1, p = 0.29) (C′ and C″). No significant differences were observed in the number of protrusions that were added and removed in ANI condition (filopodia: t₂₁ = 0.5, p = 0.66; branches: t₂₁ = 1.4, p = 0.18) (D′ and D″). (E and F) The dynamics of filopodia (E; addition: F_2,50 = 18.7, p < 0.0001; removal: F_2,50 = 13.0, p < 0.0001) and branches (F; addition: F_2,50 = 20.2, p < 0.0001; removal: F_2,50 = 9.5, p = 0.0003) were inhibited by CHX or ANI treatment. Error bars represent SEM. ^∗∗p < 0.01, ^∗∗∗p < 0.001 (paired t test for B–D) versus Control ^∗∗∗p < 0.001 (one-way ANOVA with Tukey multiple comparisons test for E and F). Red diamonds represent the averages (B–D). Scale bars, 5 μm. See also Figure S4.

Figure 3

Axon Navigation in the Optic Tract Is Not Affected by Acute Inhibition of Translation (A) Live imaging experiment on somaless RGC axon navigation in the optic tract in vivo and translation assay on whole brains. Electroporated eye was removed to eliminate somatic contribution. (B) Anti-puromycin immunolabeling of whole-mount brains, shown as fluorescent intensity heatmaps, illustrates the incorporation of puromycin after 10 min, as readout of translation. Cycloheximide (CHX) and anisomycin (ANI) treatments greatly reduce puromycin immunolabeling. (C) The incorporation of puromycin was reduced in the ventral optic tract (VOT) (F_3,67 = 204.6, p < 0.0001), dorsal optic tract (DOT) (F_3,61 = 213.4, p < 0.0001), and whole brain (F_3,80 = 501.9, p < 0.0001) after CHX and ANI treatments. (D–F) Axon navigation through the VOT and DOT in control (D) and after incubation in translation inhibitors CHX (E) and ANI (F). (G) Axon behaviors were unaffected in axons after CHX or ANI incubation (death: p = 0.44; misprojected: p = 0.19; stalling: p = 0.80; normal: p = 0.47, chi-square test). (H and I) Axon elongation velocities were unaffected by CHX or ANI incubation (H, VOT: F_2,140 = 1.3, p = 0.29; I, DOT: F_2,140 = 1.3, p = 0.27). Error bars represent SEM versus Control ^∗∗∗p < 0.001 (one-way ANOVA with Tukey multiple comparison’s test for C, H, and I). Scale bars, 50 μm.

Figure 4

Knockdown of β-actin Reduces Axon Branching Dynamics and Arbor Complexity In Vivo (A) Lateral view of single RGC axons in the tectum. Line drawings are shown with the branch order color coded: white, axon shaft; branches: red, primary; blue, secondary; yellow, tertiary; purple, quaternary. (B) Reduction in number of branches in β-actin morphants (primary: F_2,81 = 8.9, p = 0.0003; secondary: F_2,81 = 17.6, p < 0.0001; tertiary: F_2,81 = 13.0, p < 0.0001; total: F_2,81 = 29.3, p < 0.0001). (C) Branch length decreased in the β-actin MO (β-aMO) condition (F_2,81 = 14.69, p < 0.0001). (D) The proportion of branches in the β-aMO condition shifts toward lower branch orders (primary: F_2,81 = 2.1, p < 0.0001; secondary: F_2,81 = 4.7, p = 0.0006; tertiary: F_2,81 = 4.2, p = 0.0002). (E) Formulation of axon complexity index (ACI). (F) The ACI was reduced in the β-aMO condition (F_2,81 = 12.0, p < 0.0001). (G) The percentage of complex arbor (ACI ≥ 1.4) was reduced in β-aMO condition (^∗∗∗p < 0.0001, ^###p < 0.0001, Fisher’s exact test). (H–J) Axon branching in Con MO- (H) and β-aMO-positive (I) (+/– rescue construct; J) axons in the tectum. More protrusions were added than removed in control morphants (filopodia: t₁₇ = 3.9, p = 0.0011; branches: t₁₇ = 3.2, p = 0.0049) (H′ and H″). No significant differences were observed in the number of protrusions that were added and removed in β-actin morphants (filopodia: t₂₃ = 0, p = 1; branch: t₁₇ = 0.8, p = 0.42) (I′ and I″). More protrusions were added than removed in β-actin morphants that were rescued with β-aMO resistant construct (filopodia: t₉ = 3.5, p = 0.007; branches: t₉ = 2.8, p = 0.022) (J′ and J″). (K and L) The dynamics of filopodia (K; addition: F_2,49 = 9.3, p = 0.0004; removal: F_2,49 = 6.6, p = 0.003) and branches (L; addition: F_2,49 = 16.1, p < 0.0001; removal: F_2,49 = 10.2, p = 0.0002) were inhibited in β-actin morphants. (M) Eye electroporation and live imaging of axonal branching. Error bars represent SEM. ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001, ^##p < 0.01, ^###p < 0.001 (one-way ANOVA with Tukey multiple comparisons test for B–F, K, and L) and paired t test for H–J). Red diamonds represent the averages (H–J). Scale bars, 20 μm for (A) and 5 μm for (H–J). See also Figure S5.

Figure 5

Local Synthesis of β-actin Is Required for Axon Branching In Vivo (A and B) Axon branching in the tectum after local delivery of MO at stages 41–43. More protrusions were added than removed in Control MO (Con MO) condition (filopodia: t₁₇ = 3.1, p = 0.007; branches: t₁₇ = 2.4, p = 0.03) (A′ and A″). No significant differences were observed in the number of protrusions that were added and removed in β-actin MO (β-aMO) condition (addition: t₂₅ = 1.4, p = 0.16; removal: t₂₅ = 1.9, p = 0.07) (B′ and B″). (C) Filopodia dynamics were inhibited in β-aMO condition (addition: t₄₂ = 3.9, p = 0.0004; removal: t₄₂ = 2.7, p = 0.01). (D) Branch dynamics were inhibited in β-aMO condition (addition: t₄₂ = 3.1, p = 0.004; removal: t₄₂ = 3.0, p = 0.005). (E) Local delivery of MO into RGC axons by tectum electroporation and imaged immediately thereafter. Electroporated eye was removed to eliminate somatic contribution. (F and G) Time-lapse images of axonal branching in the tectum at stages 41–43 after local delivery of MO at stages 35/36–37/38. More protrusions were added than removed in Con MO condition (filopodia: t₁₁ = 3.4, p = 0.006; branches: t₁₁ = 4.9, p = 0.0005) (F′ and F″). More protrusions were added than removed in β-aMO condition (addition: t₁₇ = 4.2, p = 0.0006; removal: t₁₇ = 4.0, p = 0.0008) (G′ and G″). (H and I) The dynamics of filopodia (H; addition: t₂₈ = 0.6, p = 0.58; removal: t₂₈ = 0.6, p = 0.55) and branches (I; addition: t₂₈ = 0.3, p = 0.78; removal: t₂₈ = 0.07, p = 0.95) were unaffected in β-aMO condition. (J) Local delivery of MO into the tectum before tectal entry of RGC axons (stages 35/36–37/38) and live imaging of axonal branching after tectal entry of RGC axons (stages 41–43). Scissors and dashed line denote that only the skin overlying the tectal area was removed to minimize damage to the brain. Error bars represent SEM. ^∗p < 0.01, ^∗∗p < 0.01, ^∗∗∗p < 0.001 (paired t test for A, B, F, and G and unpaired t test for C, D, H, and I). Scale bars, 5 μm. Red diamonds represent the averages (A, B, F, and G). See also Figures S4 and S6.

Figure 6

Visualization of De Novo β-actin Synthesis in Axon Terminals In Vivo (A) Venus-β-actin construct as a reporter for β-actin translation. (B) Fluorescence recovery after photobleaching experiment of Venus constructs in vivo. Electroporated eye was removed to eliminate somatic contribution. Tec, tectum. (C) Fluorescence heatmaps illustrating that limited recovery was detected with the Venus control construct. In contrast, Venus-β-actin signal recovered soon after photobleaching and was inhibited by the translation inhibitor cycloheximide (CHX), indicating de novo synthesis of β-actin in RGC axon terminals in the tectum. (D) FRAP over the course of 10 min. Dotted lines represent least-squares fits to a single-exponential function. (Venus control versus Venus-β-actin: F_3,191 = 36.0, p < 0.0001; Venus-β-actin versus Venus-β-actin + CHX: F_3,236 = 21.8, p < 0.0001; extra sum-of-squares F test). Error bars represent SEM. Scale bars, 5 μm. See also Figure S7.

Figure 7

Focal Hotspots of Nascent β-actin at Axonal Branch Points and within Branches In Vivo (A) Fluorescence recovery after photobleaching experiment of Venus constructs in vivo. (B) Fluorescence heatmaps illustrating a ubiquitous recovery pattern for the Venus control construct. In contrast, Venus-β-actin signal recovered in hotspots at branch points and within branches. The formation of these hotspots was inhibited by cycloheximide (CHX), indicating de novo synthesis and accumulation of β-actin in highly specific focal points in RGC axons. (C) An example of multiple Venus-β-actin hotspots forming at different sub-compartments of a branch. Kymograph displays the 300 s FRAP from the branch tip to axon shaft as indicated by the magenta arrow. At least four distinct hotspots can be identified in this single branch. (D) The fluorescence variation index (FVI) is defined by normalizing the standard deviation (SD) of fluorescence in branches (F_branch) to the SD of fluorescence in axon shaft (F_shaft). (E) FVI after FRAP over the course of 5 min. Dotted lines represent least-squares fits to a quadratic function. (Venus control versus Venus-β-actin: F_3,6594 = 396, p < 0.0001; Venus-β-actin versus Venus-β-actin + CHX: F_3,5994 = 466.7, p < 0.0001). Inset displays the differences between the conditions. Error bars represent SEM. ^∗∗∗p < 0.001, ###p < 0.001 (extra sum-of-squares F test for E). Scale bars, 5 μm.

Figure 8

Nascent β-actin Microdomains Form in Close Proximity to Docked RNA Granules In Vivo (A) Dual-channel simultaneous time-lapse images of Cy5-RNA and Venus-β-actin FRAP. The axon morphology was estimated by capturing a mRFP image before photobleaching and after time-lapse acquisition, which are overlaid on the Cy5-RNA images (gray scale, top) at time points 0″ and 300″, respectively. Axon outline of the mRFP image captured after time-lapse acquisition was used as an approximation for time points 10″ to 300″. Venus-β-actin signal recovery after photobleaching is illustrated by the fluorescence heatmaps (middle). The bottom row presents the overlays of Cy5-RNA (cyan) and Venus-β-actin FRAP (magenta). (B) Enlarged images of area signified by the arrowhead in (A). The images of Cy5-RNA (cyan) and Venus-β-actin FRAP (magenta) are individually presented on the left and in the middle columns. Images on the right display image overlays. The FRAP signal positions at 30 s closely resemble the localization of RNA at 10 s. (C) Cy5-RNA time series were compiled into z stacks and computed for the median signal intensities as a representation of RNA dwell time. The resulting images were then used to compute Pearson’s correlation coefficient (R) with the Venus-β-actin cumulative recovery signal. The averages of R(observed) were significantly higher than averages of R(random) yielded from 1,000 random images scrambled from each original axon image (t₈ = 11.55, p < 0.0001, paired t test). Scale bars, 5 μm for (A) and 1 μm for (B).