The RNA-binding protein EIF4A3 promotes axon development by direct control of the cytoskeleton

Abstract

The exon junction complex (EJC), nucleated by EIF4A3, is indispensable for mRNA fate and function throughout eukaryotes. We discover that EIF4A3 directly controls microtubules, independent of RNA, which is critical for neural wiring. While neuronal survival in the developing mouse cerebral cortex depends upon an intact EJC, axonal tract development requires only Eif4a3. Using human cortical organoids, we show that EIF4A3 disease mutations also impair neuronal growth, highlighting conserved functions relevant for neurodevelopmental pathology. Live imaging of growing neurons shows that EIF4A3 is essential for microtubule dynamics. Employing biochemistry and competition experiments, we demonstrate that EIF4A3 directly binds to microtubules, mutually exclusive of the EJC. Finally, in vitro reconstitution assays and rescue experiments demonstrate that EIF4A3 is sufficient to promote microtubule polymerization and that EIF4A3-microtubule association is a major contributor to axon growth. This reveals a fundamental mechanism by which neurons re-utilize core gene expression machinery to directly control the cytoskeleton.

Keywords: CP: Cell biology; CP: Neuroscience; EIF4A3; Eif4a3; RNA; axon growth; axonal tracts; cortical development; cortical organoids; exon junction complex; microtubules; neural development; neuronal maturation.

Figure 1.. Loss of Eif4a3 but not Magoh and Rbm8a impairs early axonal development independent of apoptosis

(A) Left: canonical RNA regulation by the exon junction complex (EJC) composed of core components EIF4A3, RBM8A, and MAGOH. Right: the question addressed in this study. How do EJC components control neuronal development?. (B) Coronal section from a control (Eif4a3^lox/lox) E17.5 mouse cortex, stained with L1CAM and Hoechst. (C) Higher-magnification images from the region indicated by the dashed box in (B), stained with L1CAM and CC3. White bracket, cortical axonal tract thickness. Arrow, region with apoptotic nuclei. (D) Fold change of L1CAM+ axonal thickness in E17.5 Magoh, Rbm8a, and Eif4a3 cHet and cKO cortices relative to control (Magoh^lox/lox, Rbm8a^lox/lox and Eif4a3^lox/lox, respectively). **p = 0.0065, one-way ANOVA (Dunnett’s multiple-comparisons test), n = 3–6 embryos. (E) CC3+ cell density in E17.5 Magoh, Rbm8a, and Eif4a3 cHet and cKO relative to control (Magoh^lox/lox, Rbm8a^lox/lox, and Eif4a3^lox/lox, respectively). **p < 0.01, ***p < 0.001, one-way ANOVA (Dunnett’s multiple-comparisons test), n = 3 embryos. (F) Coronal sections of control (Eif4a3^lox/lox;p53^lox/lox) and Eif4a3; p53 dcKO E17.5 mouse cortices, stained with L1CAM, CC3, and Hoechst. White bracket, cortical axonal tract thickness. (G) Fold change of L1CAM+ cortical axonal thickness in E17.5 Eif4a3, p53 dcHet, and dcKO relative to control (Eif4a3^lox/lox;p53^lox/lox). **p = 0.0099, one-way ANOVA (Dunnett’s multiple-comparisons test), n = 4 embryos. (H) Fold change of L1CAM+ fibers/cortical thickness ratio from E17.5 Eif4a3 cKO relative to control (Eif4a3^lox/lox) and on a p53 background. *p = 0.0491, ***p < 0.001, one-way ANOVA (Dunnett’s multiple-comparison test), n = 6 embryos. All graphs, mean + SD; ns, not significant. Scale bars: 500 μm (B), 100 μm (C), and 50 μm (F). See also Figures S1 and S2.

Figure 2.. EIF4A3-mediated axonal development is EJC and RNA independent

(A) Cartoon depicting the experimental paradigm. (B) Neuronal cultures stained against TUBB3. Arrowheads, axons of polarized neurons. Asterisks, unpolarized neurons. (C) Axonal length of low-confluency neuronal cultures from control and Eif4a3 cKO and Eif4a3;p53 dcKO littermates. **p < 0.01, ****p < 0.0001, unpaired t test (two tailed), n = 5–8 cultures/embryos. (D) E15.5 WT neurons from high-confluency DIV2 cultures, co-electroporated with CAG-GFP and scrambled shRNA, Rbm8a or Magoh at DIV0. Arrows, axons. (E) Axonal length of neuronal cultures; one-way ANOVA (Dunnett’s multiple-comparisons test), n = 3 cultures/embryos. (F) Schematic of rescue experiments in which E14.5 control (Eif4a3^+/+) or Eif4a3^lox/lox littermate embryos were electroporated in utero with CAG-GFP and the constructs indicated in (G)–(I). (G–I) Axonal length from DIV3 neuronal cultures of control (Eif4a3^+/+ or Eif4a3^lox/+) or Eif4a3^lox/lox littermate embryos electroporated in utero with the indicated constructs. *p < 0.05, unpaired t test (two tailed), n = 4–8 cultures/embryos. (J) Fold change of axonal length from data in (G)–(I) relative to control (Eif4a3^+/+). *p < 0.05, **p < 0.01, one-way ANOVA (Tukey’s multiple-comparisons test), n = 4–8 cultures/embryos. All graphs, mean + SD. Scale bars, 20 μm (B) and 50 μm (D). See also Figure S3.

Figure 3.. Impaired axonal development in RCPS patient-derived cortical organoids

(A) Left: generation of human brain cortical organoids from 3 control, 1 isogenic, and 3 Richieri-Costa-Pereira syndrome (RCPS) patient-derived iPSCs. Right: day 35 control cortical organoid stained against SOX2 and TUBB3. (B) RT-qPCR analysis of EIF4A3 mRNA levels in control and RCPS organoids. Magenta, isogenic control. All organoids day 25 except RCPS line F6099-1 (day 49). *p = 0.0487, unpaired t test (two tailed), n = 4 cell lines for control (including isogenic); n = 3 cell lines for RCPS. (C) Cryosections of day 35 cortical organoids, stained for the axon initial segment (AIS) marker AnkyrinG (AnkG) and NEUN. Arrows, AIS in longitudinal plane. Arrowheads, AIS puncta. (D) AnkG puncta per NEUN+ nuclei in control and RCPS patient-derived organoids. Magenta, isogenic control. **p = 0.0087, unpaired t test (two tailed), n = 4 control (including isogenic); n = 3 RCPS. (E) Schematic of cortical organoid dissociation to generate 2D neuronal cultures. (F) DIV7 neuronal cultures from day 71–73 control and RCPS cortical organoids, stained with AnkG, TUBB3, and Hoechst. Arrow, defined AIS. White dashed circle, diffuse somatic AnkG staining. (G) DIV4 neuron cultures from day 71–73 control and RCPS cortical organoids, stained with the axonal marker SMI-312 and TUBB3. Arrows, strong SMI-312 signal in the distal part of the axon. Arrowhead, shorter axon with dimmer SMI-312 staining. (H) Axonal length across 2 controls and 2 RCPS lines. n = 203 control neurons and n = 222 cKO neurons. ****p < 0.0001, unpaired t test. All graphs, mean + SD. Scale bars, 200 μm (A), 10 μm (C), 20 μm (F), and 50 μm (G). See also Figure S4.

Figure 4.. Eif4a3 is required for axonal tract development in vivo

(A) Representation of CUBIC tissue clearing, with dashed circle depicting the cleared embryonic mouse head. (B) 3D reconstructions of tdTomato+ E14.5 mouse brains from Eif4a3 cHet (control) and Eif4a3 cKO (caudal view). Gray, tdTomato signal used to generate volumetric quantifications. Arrowheads, anterior commissures. (C) Anterior commissure volume in E14.5 Eif4a3 cHet (control) and Eif4a3 cKO brains. **p = 0.0011, unpaired t test (two tailed), n = 6 control brains (4 litters) and n = 7 Eif4a3 cKO brains. (D) 3D reconstructions of tdTomato+ E17.5 mouse brains from Eif4a3 cHet (control) and Eif4a3 cKO (caudal view). The white dashed box indicates the location of the corpus callosum (CC), hippocampal commissure (HC), and anterior commissure (AA) (indicated by arrows from top to bottom, respectively). (E and F) CC (E) and HC (F) volume in E17.5 Eif4a3 cHet (control) and Eif4a3 cKO brains. **p = 0.0057 (E), **p = 0.0047 (F), unpaired t test (two tailed), n = 4 control and Eif4a3 cKO brains. (G) Experimental design to sparsely label neurons and their projections in 3D (dashed circle, cleared brain). (H) 3D-reconstructed cortices from Eif4a3^+/+ (control) and Eif4a3^lox/lox embryos electroporated as shown in (G). Arrows, groups of axons extending to the midline after 3 days of electroporation. (I) Optical planes of 3D-reconstructed brains from Eif4a3^+/+ (control) and Eif4a3^lox/lox embryos labeled as in (H). Arrows, the longest groups of axons extending to the midline after 3 days of electroporation. (J) Axonal length of the average of the five longest electroporated tdTomato+ axons, measured from the cortical plate to the midline in Eif4a3^+/+ (control) and Eif4a3^lox/lox embryos. *p = 0.0302, unpaired t test (one tailed), n = 3 control and Eif4a3 cKO brains. (K) tdTomato+ axonal fiber volume measured from the electroporated region in the cortical plate to the midline in Eif4a3^+/+ (control) and Eif4a3^lox/lox embryos. *p = 0.0476, unpaired t test (one tailed), n = 3 control and Eif4a3 cKO brains. All graphs, mean + SD. Scale bars: 500 μm (B and I), 700 μm (D), and 250 μm (H). See also Figure S5.

Figure 5.. EIF4A3 localizes to neuronal projections and microtubules and is required for microtubule growth in developing neurons

(A) SIM super-resolution images of DIV2 cortical neurons from WT E15.5 neuronal cultures, stained for either EIF4A3, RBM8A or MAGOH, TUBB3, and Hoechst. Arrows, nucleus; arrowheads, neuronal projections. (B) DIV2 cortical neuron from WT E16.5 neuronal cultures, stained with two distinct antibodies against EIF4A3 that target different epitopes than that used in (A). (C) Mitotic HeLa cells co-stained for α-tubulin and EIF4A3 transfected with scramble or small interfering RNA against EIF4A3 and imaged after 48 h. Arrowheads indicate EIF4A3 colocalization with α-tubulin. (D) Schematic of microtubule live imaging of DIV2 neurons from E14.5 control (Eif4a3^+/+) or Eif4a3^lox/lox littermate embryos in utero electroporated with GFP-MACF43 and Dcx-mCherry-Cre. (E) Representative snapshots of live imaging of DIV2 cortical neurons from control (Eif4a3^+/+) or Eif4a3^lox/lox. GFP puncta indicate GFP-MACF43+ growing microtubules. Insets show higher-magnification image from regions indicated with white boxes. Arrowheads, growing microtubule comets visualized as GFP+ puncta. (F) Representative kymograph reconstructions from videos generated as shown in (D) and (E). Bottom: tracings of individual comet events. (G–I) Comet speed (G), comet distance (H), and comet lifetime (I) from neuronal cultures from control (Eif4a3^+/+) or Eif4a3^lox/lox littermates. **p = 0.0042 (H), **p = 0.0077 (I), unpaired t test (two tailed), n = 15 control electroporated neurons (3 embryos) and n = 20 Eif4a3^lox/lox electroporated neurons (4 embryos). Graphs represent truncated violin plots with median (thicker dashed-line) ± quartiles (thinner dashed lines). (J) Rescue experiment in which neurons of the indicated genotypes were treated with paclitaxel (Taxol) as indicated. Quantification reflects axonal length of the indicated genotypes and treatments. One-way ANOVA (Tukey’s multiple-comparisons test), n = 6 cultures/embryos, *p < 0.05, **p = 0.0026. All graphs, mean + SD. (K) DIV2 neurons of the indicated genotypes that were either mock treated or chronically treated with paclitaxel. Arrows denote axons. Scale bars: 10 μm (A–C, E, and K). See also Figure S6.

Figure 6.. EIF4A3 directly binds to microtubules in a mutually exclusive and EJC-independent manner

(A) Cartoon depicting the experimental paradigm of the co-immunoprecipitation assays. (B) Immunoblots (IBs) depicting co-immunoprecipitation (coIP) of TUBB3 or FLAG (control) with core EJC components (EIF4A3, MAGOH, and RBM8A), re-probing of the same blot with TUBB3 antibody, and input lysates. n = 2 experiments. (C) CoIP of EIF4A3 or RFP (control) with TUBB3, re-probing of the same blot with EIF4A3 antibody, and input lysates. n = 4 experiments. (D) Cartoon depicting the experimental paradigm of the co-sedimentation assays. (E) Immunoblot depicting mock- and RNase-treated lysates subjected to microtubule polymerization and co-sedimentation, probed for EIF4A3 and α-tubulin. Right: agarose gel with rRNA from either mock- or RNase A-treated samples. n = 5 experiments. (F) Cartoon depicting the experimental paradigm of the GST pull-down assays. (G) Immunoblots of experiments as in (F), where microtubules were precipitated with GST alone, GST-MAGOH, or GST-EIF4A3 (and probed with anti-GST). Right: GST pull-down assay using the RNA- and EJC-binding mutant EIF4A3^T163D. A blot against TUBB3 is shown above. (H) TUBB3 levels relative to total fusion protein for independent experiments. *p < 0.05, one-way ANOVA (Tukey’s multiple-comparisons test), n = 4–6 experiments. (I) Cartoon depicting the experimental paradigm of the targeted mass spectrometry assays. (J) Relative protein levels quantified from E15.5 cortices and DIV2 neurons. ****p < 0.0001, two-way ANOVA (Tukey’s multiple-comparisons test), n = 3 embryos/cultures. (K) Competition experiment in which endogenous TUBB3 from N2a cells was co-immunoprecipitated with FLAG-EIF4A3 and increasing amounts of equimolar V5-MAGOH. TUBB3/FLAG densitometry ratios (normalized to FLAG-EIF4A3 alone): 1 (lane 2), 0.89 (lane 3), and 0.42 (lane 4). Arrows, V5-RBM8A (top) and V5-MAGOH (bottom). (L) Quantification of the competition experiments. *p = 0.0261, unpaired t test (two tailed), n = 4 experiments. All graphs, mean + SD. See also Figure S6.

Figure 7.. EIF4A3 directly controls microtubule polymerization, and EIF4A3 association with microtubules contributes to axon growth

(A) Cartoon depicting the experimental paradigm of the total internal reflection fluorescence microtubule growth assay. (B) Example kymographs of experiments as described in (A) in the presence or absence of 10 nM StrepII-EIF4A3. (C) Polymerization rates for microtubules in the absence or presence of increasing amounts of StrepII-EIF4A3. Truncated violin plots show median (thicker line) ± quartiles (thinner dashed lines). ****p < 0.0001, one-way ANOVA (Dunnetts’ multiple-comparisons test) from 3 independent experiments. (D) CoIP of TUBB3 with either FLAG-EIF4A3 or FLAG-EIF4A3-ΔN100 in N2a cells and re-probed against FLAG. (E) Fold change of TUBB3 levels co-immunoprecipitated with either EIF4A3 WT or EIF4A3-ΔN100 (see Figures S7I and S7J for full reference). The graph represents mean + SD, ****p < 0.0001, two-tailed unpaired t test, n = 13–18 independent samples. (F) Left: axonal length quantification from neuronal cultures of control (Eif4a3^+/+) or Eif4a3^lox/lox littermate embryos electroporated in utero with CAG-GFP, Dcx-mCherry-Cre, and CAG-3×-FLAG-EIF4A3-ΔN100 (same experimental paradigm as in Figures 2F-2I). **p = 0.0016, unpaired t test (two tailed), n = 6 cultures/embryos per condition. Right: fold change of axonal length relative to control (Eif4a3^+/+) from Eif4a3^lox/lox embryos electroporated in utero with the indicated constructs (combined with data from Figure 2J for direct comparison). One-way ANOVA (Tukey’s multiple-comparison test), n = 4–8 cultures/embryos. (G) Graphical summary depicting canonical and non-canonical roles of EIF4A3 in neuronal development. All graphs, SD. See also Figure S7.