Elongator subunit 3 (ELP3) modifies ALS through tRNA modification

Abstract

Amyotrophic lateral sclerosis (ALS) is a fatal degenerative motor neuron disorder of which the progression is influenced by several disease-modifying factors. Here, we investigated ELP3, a subunit of the elongator complex that modifies tRNA wobble uridines, as one of such ALS disease modifiers. ELP3 attenuated the axonopathy of a mutant SOD1, as well as of a mutant C9orf72 ALS zebrafish model. Furthermore, the expression of ELP3 in the SOD1G93A mouse extended the survival and attenuated the denervation in this model. Depletion of ELP3 in vitro reduced the modified tRNA wobble uridine mcm5s2U and increased abundance of insoluble mutant SOD1, which was reverted by exogenous ELP3 expression. Interestingly, the expression of ELP3 in the motor cortex of ALS patients was reduced and correlated with mcm5s2U levels. Our results demonstrate that ELP3 is a modifier of ALS and suggest a link between tRNA modification and neurodegeneration.

Figure 1.

ELP3 is neuroprotective in two zebrafish ALS models. Effect of human ELP3 in the axonopathy induced by C9orf72-associated repeat RNA (A and B) or SOD1A4V RNA (C and D), assessed by quantification of motor axon length (A and C) and affected embryos/axonal branching quantification (B and D). Co-injection of wild-type human ELP3 (ELP3) partially prevents the toxicity induced by either 90x(GGGGCC) RNA or SOD1A4V RNA. 3x(GGGGCC) RNA and SOD1WT were used as controls. Data represent mean ± S.E.M., one-way ANOVA (A and C), or mean ± 95% CI, logistic regression (B and D). n = 90. ***P < 0.001, **P < 0.01. *P < 0.05.

Figure 2.

Overexpression of human ELP3 attenuates the ALS-like phenotype in the SOD1^G93A mouse. (A) ELP3 mRNA relative expression in neurons and glia from the lumbar spinal cord of symptomatic (post-natal day 140) SOD1^G93A mice. Levels were reduced by 31% and 52% in neurons and glia, respectively, compared to non-transgenic mice. Moreover, ELP3 was less abundant in glia than in neurons: 37.5% in non-transgenic mice and 57.5% in SOD1^G93A mice. Data represent mean ± S.E.M., one-way ANOVA, n = 4. ***P < 0.001, **P < 0.01. (B–E) Effect of ubiquitous tamoxifen-induced expression of human ELP3 in SOD1^G93A mice. Tamoxifen was administered at post-natal day 60. (B) Mean disease onset was not affected by ELP3 expression: SOD1^G93A/ELP3 KI 92.3 ± 1.8 days, n = 22 versus SOD1^G93A/ELP3 KI^CAGG 97.1 ± 2.0 days, n = 19. Data are mean ± S.E.M., log-rank test, P = 0.1671. (C) Mean survival was significantly prolonged by ELP3 expression: SOD1^G93A/ELP3 KI 152.0 ± 1.8 days, n = 35 versus SOD1^G93A/ELP3 KI^CAGG 160.7 ± 1.9 days, n = 19. Data are mean ± S.E.M., log-rank test, P = 0.006. (D) Relative quantification of the innervation of NMJs of the gastrocnemius muscle in symptomatic mice (post-natal day 150). Denervated NMJs were decreased by ELP3 expression: SOD1^G93A/ELP3 KI 78.4 ± 1.0%, n = 3 versus SOD1^G93A/ELP3 KI^CAGG 59.9 ± 2.7%, n = 3. Data represent mean ± S.E.M., one-way ANOVA, P = 0.001. (E) Quantification of ventral horn lumbar motor neurons, in function of the cell body area, in symptomatic mice (post-natal day 150). No differences were found between genotypes. Data represent mean ± S.E.M., two-way ANOVA, n = 3, P > 0.5. (F, G) Effect of neuronal-specific tamoxifen-induced expression of ELP3 in SOD1^G93A mice. Tamoxifen was administered at post-natal day 60. (F) Mean disease onset was not affected by ELP3 expression: SOD1^G93A/ELP3 KI 89.1 ± 3.7 days, n = 15 versus SOD1^G93A/ELP3 KI^Thy1 94.2 ± 1.9 days, n = 28. Data are mean ± S.E.M., log-rank test, P = 0.383. (G) Mean survival was not affected by ELP3 expression: SOD1^G93A/ELP3 KI 149.0 ± 2.4 days, n = 13 versus SOD1^G93A/ELP3 KI^Thy1 149.1 ± 1.8 days, n = 18. Data are mean ± S.E.M., log-rank test, P = 0.982.

Figure 3.

Effect of heterozygous deletion of the ELP3 gene in the SOD1G93A mouse. (A) Mean disease onset was reduced by ELP3 heterozygous deletion: SOD1^G93A/ELP3^+/+ 105.8 ± 2.1 days, n = 45 versus SOD1^G93A/ELP3^+/− 98.7 ± 2.2 days, n = 40. Data are mean ± S.E.M., log-rank test, P = 0.0141. (B) Mean survival was not affected: SOD1^G93A/ELP3^+/+ 157.5 ± 1.6 days, n = 38 versus SOD1^G93A/ELP3^+/− 155.2 ± 1.5 days, n = 30. Data are mean ± S.E.M., log-rank test, P = 0.2307. (C) Relative quantification of the innervation of NMJs of the gastrocnemius muscle in symptomatic mice (post-natal day 104). Denervated NMJs were not affected: SOD1^G93A/ELP3^+/+ 43.7 ±3.5%, n = 6 versus SOD1^G93A/ELP3^+/− 40.7 ± 10.4%, n = 3. Data represent mean ± S.E.M., two-way ANOVA, P = 0.9173. (D) Quantification of ventral horn lumbar motor neurons, in function of the cell body area, in symptomatic mice (post-natal day 104). No differences were found between genotypes. Data represent mean ± S.E.M., two-way ANOVA, n = 3, P > 0.5.

Figure 4.

SAM domain, not HAT domain, mediates ELP3 protective effect in the zebrafish. Effect of ELP3 mutants in the axonopathy induced by C9orf72-associated repeat RNA (A and B) or SOD1A4V RNA (C and D), assessed by quantification of motor axon length (A and C) and affected embryos/axonal branching quantification (B and D). Deletion of the HAT domain (ELP3^ΔHAT) does not affect the protective effect of ELP3 (dashed lines), whereas mutation of the SAM domain (ELP3^SAM) abolishes the protective effect of ELP3. 3x(GGGGCC) RNA and SOD1WT were used as controls. Data represent mean ± S.E.M., one-way ANOVA (A and C) or mean ± 95% CI, logistic regression (B and D). n ≥ 45. ***P < 0.001, **P < 0.01, *P < 0.05.

Figure 5.

Effect of ELP3 expression levels on wobble mcm⁵s²U tRNA modification and protein aggregation. (A) ELP3 mRNA relative expression in NSC34 cells, 72 h post-transfection. Levels of ELP3 were reduced by 85.0 ± 2.2% in cells transfected with ELP3-siRNA, compared to control cells. Data represent mean ± S.E.M., unpaired t-test, n = 5. ***P < 0.001. (B) mcm⁵s²U levels in NSC34 cells were reduced by 60.1 ± 2.1% in cells transfected with ELP3-siRNA, compared to control cells. t6A was used as internal control. Data represent mean ± S.E.M., unpaired ttest, n = 4. ***P < 0.001. (B′) Representative mcm⁵s²U chromatograms of the conditions analyzed in B. (C) ELP3 mRNA relative expression in spinal cord after ubiquitous tamoxifen-induced deletion of endogenous ELP3 (ELP3^lox/lox/CAGG^CreER mouse, designated ELP3 KO). Levels were reduced by 91.5 ± 3.9%. Data represent mean ± S.E.M., unpaired t-test, n = 3. ***P < 0.001. (D) mcm⁵s²U levels in spinal cord after ubiquitous tamoxifen-induced deletion of endogenous ELP3. Levels were reduced by 72.2 ± 1.7%. t6A was used as internal control. Data represent mean ± S.E.M., unpaired t-test, n = 3. ***P < 0.001. (D′) Representative mcm⁵s²U chromatograms of the conditions analyzed in (D). (E–J) Western blot analyses and quantification of SOD1 protein in detergent-insoluble and detergent-soluble fractions of NSC34 cells. (E–G) Aggregates of mutant SOD1 (G93A and A4V) accumulate upon siRNA-mediated depletion of ELP3, compared to control (control-siRNA, –). GAPDH was used as loading control. Data represent mean ± S.E.M., one-way ANOVA, n ≥ 4. **P < 0.01, *P < 0.05. (H–J) ELP3 overexpression reduces the accumulation of SOD1^G93A aggregates induced by ELP3 depletion. FLAG was used to control ELP3 overexpression, GAPDH was used as loading control. Data represent mean ± S.E.M., one-way ANOVA, n = 3. ***P < 0.001, **P < 0.01, *P < 0.05. (K) Effect of ELP3 overexpression on the levels of mcm⁵s²U in NSC34 cells expressing SOD1^G93A and either ELP3-siRNA or control-siRNA. Data represent mean ± S.E.M., one-way ANOVA, n = 4. *P < 0.05. (K′) Representative mcm⁵s²U chromatograms of the conditions analyzed in (K). ***P < 0.001, **P < 0.01, *P < 0.05.

Figure 6.

ELP3 expression correlates with wobble mcm⁵s²U tRNA modification in sALS patients. (A) ELP3 mRNA relative expression in the motor cortex from sALS patients. Data represent mean ± S.E.M., unpaired t-test, n = 8. *P < 0.05. (B) mcm⁵s²U levels in the motor cortex from sALS patients. Data represent mean ± S.E.M., unpaired t-test, n ≥6. P > 0.05. (C) Levels of ELP3 mRNA correlate with mcm⁵s²U levels. Linear regression, n = 9, P = 0.0142.