The m6 A Readers YTHDF1 and YTHDF2 Synergistically Control Cerebellar Parallel Fiber Growth by Regulating Local Translation of the Key Wnt5a Signaling Components in Axons

Abstract

Messenger RNA m⁶ A modification is shown to regulate local translation in axons. However, how the m⁶ A codes in axonal mRNAs are read and decoded by the m⁶ A reader proteins is still unknown. Here, it is found that the m⁶ A readers YTHDF1 and YTHDF2 are both expressed in cerebellar granule cells (GCs) and their axons. Knockdown (KD) of YTHDF1 or YTHDF2 significantly increases GC axon growth rates in vitro. By integrating anti-YTHDF1&2 RIP-Seq with the quantitative proteomic analysis or RNA-seq after KD of YTHDF1 or YTHDF2, a group of transcripts which may mediate the regulation of GC axon growth by YTHDFs is identified. Among them, Dvl1 and Wnt5a, encoding the key components of Wnt pathway, are further found to be locally translated in axons, which are controlled by YTHDF1 and YTHDF2, respectively. Specific ablation of Ythdf1 or Ythdf2 in GCs increases parallel fiber growth, promotes synapse formation in cerebellum in vivo, and improves motor coordination ability. Together, this study identifies a mechanism by which the m⁶ A readers YTHDF1 and YTHDF2 work synergistically on the Wnt5a pathway through regulating local translation in GC axons to control cerebellar parallel fiber development.

Keywords: YTHDF1; YTHDF2; cerebellar parallel fibers; local translation; m6A.

Figure 1

KD of YTHDF1 or YTHDF2 significantly promoted GC axon growth in vitro. A) Representative confocal images showing YTHDF1 and YTHDF2 are expressed in the growth cones and axons of cultured P6–P8 GCs. B) Western blotting (WB) validating the KD efficiency of shYthdf1 in cultured GCs. Data of quantification are mean ± SEM and represented as dot plots (n = 3): shYthdf1#2 versus shCtrl, ****p = 3.91E‐08; shYthdf1#3 versus shCtrl, ****p = 8.62E‐09; by one‐way ANOVA followed by Tukey's multiple comparison test. C) WB validating the KD efficiency of shYthdf2 in cultured GCs. Data of quantification are mean ± SEM and represented as dot plots (n = 3): shYthdf2#1 versus shCtrl, ****p = 9.15E‐06; shYthdf2#3 versus shCtrl, ****p = 2.53E‐06; by one‐way ANOVA followed by Tukey's multiple comparison test. D) Representative images showing that axon growth rates of GCs are significantly increased after KD of YTHDF1. Black and blue arrowheads indicate the terminals of the same axons imaged at 0 and 15 h, respectively. E) Quantification of axon growth rates in (D). Data are represented as box and whisker plots: shYthdf1#2 versus shCtrl, **p = 0.0030; shYthdf1#3 versus shCtrl, ***p = 1.39E‐04; n = 24 axons for each group; by one‐way ANOVA followed by Tukey's multiple comparison test. F) Representative images showing that axon growth rates of GCs are significantly increased after KD of YTHDF2. Black and blue arrowheads indicate the terminals of the same axons imaged at 0 and 15 h, respectively. G) Quantification of axon growth rate in (F). Data are represented as box and whisker plots: shYthdf2#1 versus shCtrl, ****p = 1.40E‐07; shYthdf2#3 versus shCtrl, ****p = 6.38E‐08; n = 24 axons for each group; by one‐way ANOVA followed by Tukey's multiple comparison test. A,D,F) Scale bars represent 10 µm.

Figure 2

The putative mRNA targets were identified by anti‐YTHDF1 and anti‐YTHDF2 RIP‐seq. A) Venn diagram showing numbers of mRNA targets identified by anti‐YTHDF1 and anti‐YTHDF2 RIP‐seq. B,C) GO analysis of target mRNAs identified by B) anti‐YTHDF1 and C) anti‐YTHDF2 RIP‐seq. The GO terms in Biological Process are shown. The most relevant terms are highlighted in red texts. D,E) KEGG analysis of target mRNAs identified by D) anti‐YTHDF1 and E) anti‐YTHDF2 RIP‐seq. Axon guidance, mTOR signaling pathway, and Wnt signaling pathway are highlighted in red texts.

Figure 3

The differentially expressed genes were identified by proteome and transcriptome analysis after KD of YTHDF1 and YTHDF2, respectively. A) GO analysis of downregulated proteins revealed by quantitative proteomic analysis after YTHDF1 KD in GCs. B) KEGG analysis of downregulated proteins revealed by quantitative proteomic analysis. The Wnt signaling pathway is highlighted in red texts. C) Heatmap showing the differential expression profiling of genes by RNA‐seq after YTHDF2 KD in GCs. D) GO analysis of differentially expressed genes revealed by RNA‐seq after YTHDF2 KD in GCs. E) Axon‐related GO terms of differentially expressed genes revealed by RNA‐seq after YTHDF2 KD in GCs. The GO terms in Biological Process are shown. BP, biological process; MF, molecular function; CC, cellular component.

Figure 4

YTHDF1 and YTHDF2 regulate local translation of Dvl1 and Wnt5a, respectively, to control the GC axon growth. A) Relative Dvl1 protein level detected by TMT‐labeled proteomic analysis after YTHDF1 KD. Data are mean ± SEM: *p = 0.047; n = 3 replicates; by unpaired Student's t test. B) RT‐qPCR confirming the Dvl1 mRNA level was unchanged after KD of YTHDF1 in GCs. Data are mean ± SEM: p = 0.78; n = 3; ns, not significant; by unpaired Student's t test. C) Relative Wnt5a mRNA level measured by RNA‐seq after YTHDF2 KD. Data are mean ± SEM: ****p = 4.92E‐05; n = 3 replicates; by unpaired Student's t test. D) Axon growth rate significantly increased after KD of Dvl1. Quantification of axon growth rates after KD of Dvl1 using siRNAs. Data are represented as box and whisker plots: n = 21 axons for each group; siDvl1#4 versus siCtrl, ****p = 2.09E‐05; siDvl1#5 versus siCtrl, ****p = 3.15E‐10. All by one‐way ANOVA followed by Tukey's multiple comparison test. E) Axon growth rate significantly decreased after KD of Wnt5a. Quantification of axon growth rates after KD of Wnt5a using siRNAs. Data are represented as box and whisker plots: n = 20 axons for each group; siWnt5a#1 versus siCtrl, ****p = 4.79E‐05; siWnt5a#3 versus siCtrl, ****p = 4.96E‐06. All by one‐way ANOVA followed by Tukey's multiple comparison test. F) Dvl1 and Wnt5a mRNAs were detected in axons by RT‐PCR using total RNA from pure axons or soma, respectively. Similar to β‐actin mRNA which is a positive control for axonal mRNAs, Dvl1 and Wnt5a mRNAs were detected in both axons and soma. The absence of H1f0 mRNA from axons indicated that the axonal material was pure with no soma incorporation. G,H) Detection of Dvl1 and Wnt5a mRNA localization in growth cones of GC neurons by FISH. Dissociated GCs were cultured for 2 DIV and then FISH was performed using RNAscope riboprobes. Dvl1 and Wnt5a mRNAs were detected in growth cones of GC neurons as red punctate patterns. β‐actin and Dapb serve as positive and negative controls, respectively. Tuj1 immunostaining was used to visualize axons. Quantification of puncta density was shown in (H). I,J) Compartmentalized KD of Dvl1 in GC axons. GCs were cultured in microfluidic chambers and siDvl1 was specifically transfected to axons only. Compared with siCtrl, siDvl1#4 and siDvl1#5 led to significant decrease of Dvl1 IF signals. Quantification data are represented as box and whisker plots (J). siDvl1#4 (n = 15 axons) versus siCtrl (n = 18 axons), ****p = 1.11E‐11; siDvl1#5 (n = 17 axons) versus siCtrl, ****p = 3.56E‐12; by one‐way ANOVA followed by Tukey's multiple comparison test. K) Axon growth rates significantly increased after axon‐specific KD of Dvl1. Data are represented as box and whisker plots. siDvl1#4 (n = 18 axons) versus siCtrl (n = 21 axons), **p = 0.0024; siDvl1#5 (n = 17 axons) versus siCtrl, ***p = 0.00049; by one‐way ANOVA followed by Tukey's multiple comparison test. L,M) Compartmentalized KD of Wnt5a in axons. Compared with siCtrl, siWnt5a#1 and siWnt5a#3 led to significant decrease of Wnt5a IF signals. Quantification data are represented as box and whisker plots (M). siWnt5a#1 (n = 38 axons) versus siCtrl (n = 32 axons), ****p = 6.80E‐14; siWnt5a#3 (n = 39 axons) versus siCtrl, ****p = 5.20E‐14; by one‐way ANOVA followed by Tukey's multiple comparison test. N) Axon growth rates significantly decreased after axon‐specific KD of Wnt5a which can be rescued by application of recombinant Wnt5a protein into axonal compartments. Data are represented as box and whisker plots. siWnt5a#1 (n = 15 axons) versus siCtrl (n = 16 axons), ****p = 3.86E‐06; siWnt5a#3 (n = 19 axons) versus siCtrl, ****p = 1.59E‐07; siWnt5a#1+rWnt5a (n = 18 axons) versus siCtrl, p = 0.34; siWnt5a#3+rWnt5a (n = 16 axons) versus siCtrl, p = 0.85; siWnt5a#1+rWnt5a versus siWnt5a#1, *p = 0.036; siWnt5a#3+rWnt5a versus siWnt5a#3, ****p = 3.40E‐05; ns, not significant; by one‐way ANOVA followed by Tukey's multiple comparison test. O) Overexpression of YTHDF1 increased axonal Dvl1 protein level in cultured GCs and axon‐specific siDvl1 KD eliminated this increase. Data are represented as box and whisker plots. Ythdf1‐IRES‐GFP + siCtrl versus IRES‐GFP + siCtrl, ****p = 1.34E‐05; Ythdf1‐IRES‐GFP + siDvl1#4 versus IRES‐GFP + siDvl1#4, p = 0.99; Ythdf1‐IRES‐GFP + siDvl1#5 versus IRES‐GFP + siDvl1#5, p = 0.84; IRES‐GFP + siDvl1#4 versus IRES‐GFP + siCtrl, p = 5.60E‐14; IRES‐GFP + siDvl1#5 versus IRES‐GFP + siCtrl, p = 5.80E‐14; Ythdf1‐IRES‐GFP + siDvl1#4 versus Ythdf1‐IRES‐GFP + siCtrl, ****p = 1.01E‐15; Ythdf1‐IRES‐GFP + siDvl1#5 versus Ythdf1‐IRES‐GFP + siCtrl, ****p = 1.02E‐15; ns, not significant; n = 27 axons for each group; by one‐way ANOVA followed by Tukey's multiple comparison test. P) KD of YTHDF2 increased axonal Wnt5a protein level in GCs and axon‐specific siWnt5a KD eliminated this increase. Data are represented as box and whisker plots. shYthdf2#3 + siCtrl versus shCtrl + siCtrl, ****p = 4.30E‐15; shYthdf2#3 + siWnt5a#1 versus shCtrl + siWnt5a#1, p = 0.99; shYthdf2#3 + siWnt5a#3 versus shCtrl + siWnt5a#3, p = 0.89; shCtrl + siWnt5a#1 versus shCtrl + siCtrl, ****p = 1.34E‐11; shCtrl + siWnt5a#3 versus shCtrl + siCtrl, p = 1.01E‐13; shYthdf2#3 + siWnt5a#1 versus shYthdf2#3 + siCtrl, ****p = 1.01E‐15; shYthdf2#3 + siWnt5a#3 versus shYthdf2#3 + siCtrl, ****p = 1.02E‐15; ns, not significant; n = 27 axons for each group; by one‐way ANOVA followed by Tukey's multiple comparison test. Scale bars represent G) 10 µm and I,L) 5 µm.

Figure 5

Parallel fiber growth was enhanced in both Ythdf1 and Ythdf2 cKO mice. A,B) Representative images of A) YTHDF1 and B) YTHDF2 immunostaining in P15 cerebellum of A) Ythdf1 and B) Ythdf2 cKO, respectively. YTHDF1 or YTHDF2 was successfully eliminated in GCs while their expression in PCs was not affected. Scale bars represent 500 µm. C,D) Lengths of parallel fibers labeled by DiI were significantly increased in Ythdf1 and Ythdf2 cKO mice. The white arrowheads indicate the terminals of DiI‐labeled PFs. Scale bars represent 100 µm. E,F) Quantification of parallel fiber (PF) lengths in (C) and (D). Data are expressed as box and whisker plots. In (E), ****p = 1.38E‐05; for Ythdf1^fl/fl mice, n = 36 confocal fields from 11 pups, for Ythdf1 cKO mice, n = 42 confocal fields from 11 pups. In (F), ****p = 2.29E‐05; for Ythdf2^fl/fl mice, n = 46 confocal fields from 13 pups, for Ythdf2 cKO mice, n = 43 confocal fields from 12 pups. All by unpaired Student's t test. G,H) Significantly higher Tag1 IF in the deep layer of cerebellar EGL of Ythdf1 and Ythdf2 cKO mice was detected. Scale bars represent 40 µm. I,J) Quantification of Tag1 IF intensity signals in (G) and (H). Data are expressed as box and whisker plots. In (I), ****p = 2.80E‐06; n = 18 confocal fields for Ythdf1^fl/fl mice, n = 15 confocal fields for Ythdf1 cKO mice. In (J), ****p = 2.02E‐12; n = 26 confocal fields for Ythdf2^fl/fl mice, n = 28 confocal fields for Ythdf2 cKO mice. All by unpaired Student's t test.

Figure 6

Synapse formation was promoted in Ythdf1 and Ythdf2 cKO cerebella. A–E) Representative immunoblots showing that the protein levels of synaptic markers GluRδ2, Nrxn1, and PSD95 were increased in A) Ythdf1 cKO cerebellum at P30. Quantification of B) YTHDF1, C) GluRδ2, D) Nrxn1, and E) PSD95. For (B), ***p = 0.00064; for (C), *p = 0.016; for (D), *p = 0.028; for (E), *p = 0.013; n = 3 replicates; by unpaired Student's t test. F–J) Representative immunoblots showing that GluRδ2, Nrxn1, and PSD95 protein levels were increased in F) Ythdf2 cKO cerebellum at P30. Quantification of G) YTHDF2, H) GluRδ2, I) Nrxn1, and J) PSD95. For (G), **p = 0.0037; for (H), ****p = 2.65E‐05; for (I), **p = 0.0062; for (J), *p = 0.028; n = 3 replicates; by unpaired Student's t test. K–N) Representative images of VGLUT1 and PSD95 co‐immunostaining in the ML of P30 cerebellum of K) Ythdf1 and M) Ythdf2 cKO. VGLUT1⁺/PSD95⁺ puncta were counted to measure the number of synapses and quantifications were shown in (L) and (N). Data are expressed as box and whisker plots. In (L), ***p = 1.10E‐04; in (N), ****p = 2.16E‐07; n = 20 confocal fields for each group; by unpaired Student's t test. Scale bars represent K,M) 5 µm.

Figure 7

The motor coordination ability is enhanced in Ythdf1 and Ythdf2 cKO mice. A) Ythdf1 and B) Ythdf2 cKO showed normal animal size and cerebellar development at P40. Scale bars in the upper panels represent 1 cm and scale bars in the lower panels represent 0.25 cm. C,D) Normal body weight in Ythdf1 and Ythdf2 cKO mice. In (C), p = 0.99; n = 15 for Ythdf1^fl/fl mice; n = 14 for Ythdf1 cKO mice. In (D), p = 0.37; n = 9 for Ythdf2^fl/fl mice; n = 10 for Ythdf2 cKO mice; ns, not significant. All by unpaired Student's t test. E,F) The latency to fall measurements for E) each and F) total trial in rotarod test of Ythdf1 cKO mice. In (E), for Day1‐Run #3, *p = 0.037; for Day2‐Run #1, **p = 0.0049; for Day3‐Run #1, *p = 0.039; for Day3‐Run #2, *p = 0.034; for Day3‐Run #3, *p = 0.037. In (F), *p = 0.027; n = 15 for Ythdf1^fl/fl mice; n = 14 for Ythdf1 cKO mice. All by unpaired Student's t test. G,H) The latency to fall measurements for G) each and H) total trial in rotarod test of Ythdf2 cKO mice. In (G), for Day1‐Run #3, **p = 0.0020; for Day2‐Run #3, **p = 0.0089; for Day3‐Run #1, *p = 0.038; for Day3‐Run #2, *p = 0.013. In (H), *p = 0.015; n = 15 for Ythdf2^fl/fl mice; n = 14 for Ythdf2 cKO mice. All by unpaired Student's t test.

Figure 8

A working model shows that YTHDF1 and YTHDF2 work synergistically to regulate Wnt5a‐PCP signaling pathway and cerebellar granule cell axon growth. A) Under normal conditions, YTHDF1 promotes the translation of m⁶A‐modified Dvl1 mRNA in GC axons. Dvl1 can block Wnt5a‐Fzd3‐activated PCP signaling. Meanwhile, YTHDF2 facilitates Wnt5a mRNA degradation to downregulate Wnt5a protein level in GC axons. So both YTHDF1 and YTHDF2 negatively regulate Wnt5a‐PCP signaling pathway and GC axon growth. B) In Ythdf1 and Ythdf2 cKO mice, local translation of Dvl1 mRNA and decay of Wnt5a mRNA in GC axons are inhibited, respectively. The resulting downregulation of Dvl1 and upregulation of Wnt5a protein levels in axons potentiate Wnt5a‐PCP signaling and promote GC axon growth.