Mitochondrial dysfunction induces dendritic loss via eIF2α phosphorylation

Abstract

Mitochondria are key contributors to the etiology of diseases associated with neuromuscular defects or neurodegeneration. How changes in cellular metabolism specifically impact neuronal intracellular processes and cause neuropathological events is still unclear. We here dissect the molecular mechanism by which mitochondrial dysfunction induced by Prel aberrant function mediates selective dendritic loss in Drosophila melanogaster class IV dendritic arborization neurons. Using in vivo ATP imaging, we found that neuronal cellular ATP levels during development are not correlated with the progression of dendritic loss. We searched for mitochondrial stress signaling pathways that induce dendritic loss and found that mitochondrial dysfunction is associated with increased eIF2α phosphorylation, which is sufficient to induce dendritic pathology in class IV arborization neurons. We also observed that eIF2α phosphorylation mediates dendritic loss when mitochondrial dysfunction results from other genetic perturbations. Furthermore, mitochondrial dysfunction induces translation repression in class IV neurons in an eIF2α phosphorylation-dependent manner, suggesting that differential translation attenuation among neuron subtypes is a determinant of preferential vulnerability.

Figure 1.

Dendritic arbors of class IV neurons show vulnerability to altered prel function. (A–C) Representative dendritic and somal mitochondrial morphologies of wild-type (A), prel[1] mutant (B), and Prel-overexpressing (O/E; C) class IV neurons. Dendritic arbors and mitochondria were visualized with membrane-bound GFP and mitoGFP, respectively. Bars: (dendrite images) 50 µm; (mitochondrial images) 5 µm. (D–I) Quantification of the total dendritic length of prel[1] mutant (D, F, and H) and Prel-overexpressing (E, G, and I) da neurons. Total length values of each neuron were plotted, and boxplots represent median and interquartile ranges. Class IV neurons (D and E) showed more severe dendritic losses than class I (F and G) and class III (H and I) neurons. (J–L) Developmental changes in the total length of class IV neuron dendrites. Representative images of dendritic arbors of class IV neurons at early L1 (22–26 h AEL; J) and at early L2 (46–50 h AEL; K). Prel was expressed by using a pan da neuronal Gal4 driver Gal4^109(2)80, which expresses from early in embryonic development onward (Gao et al., 1999). Quantification of the total dendritic length of wild-type and Prel-overexpressing class IV neurons throughout larval development (L and L′). Plots of early L1 larvae (a black box in L) are magnified in L′ for clarity. (M–O) Time-lapse analysis of larval dendritic development. Representative images of wild-type (M) and Prel-overexpressing (N) Class IV neurons. Quantification of regression of preexisting terminals (O) shows decreased stability of terminals of class IV arbors overexpressing Prel (n = 860 terminals from four neurons for control and 701 terminals from six neurons for Prel O/E). Bars, 50 µm. *, P < 0.05; ns, not significant. The statistical tests used, exact p-values, 95% confidence intervals, and sample sizes in this and subsequent figures are summarized in Table S1, and genotypes are summarized in Table S2.

Figure 2.

The FRET signal of AT[NL] reports ATP levels in class IV neurons. (A) Prel O/E in S2 cells decreased the FRET signal (FRET/CFP emission ratio) of AT[NL], but not that of AT[RK]. Prel plus either AT[NL] or AT[RK] were transiently coexpressed in S2 cells using actin-Gal4. Plots indicate the FRET signal values of individual cells. Similar results were obtained in three or more independent experiments. (B and C) Real-time imaging of ATeams in class IV neurons of dissected larvae. Representative pseudocolored images of the FRET signals of class IV neuronal soma expressing AT[NL] or AT[RK] (B, left) in the presence of 100 µM antimycin (AM), an inhibitor of the respiratory complex III. Representative plots of the signal values in individual neurons (B, right). A thick black line indicates the duration of AM treatment. Arrowheads in B (right) indicate ATP-independent drops of the signals of ATeams (see also Materials and methods) in class IV neurons treated with AM. (C) Changes in mean signal values of AT[NL] in class IV neurons treated with inhibitors of glycolysis and/or OXPHOS. AM and 2-deoxyglucose (2-DG), an inhibitor of glycolysis, were used at a concentration of 50 µM and 50 mM, respectively. Values are mean ± SD. (D and E) FRET imaging of class IV neurons expressing ATeams plus Prel driven by Gal4^109(2)80 in whole live larvae. Pseudocolored images (D, left) and quantification (D, right) of the FRET signals of AT[NL]. Plots indicate the FRET signal values in each neuron. We coexpressed the mCherry-CAAX transgene as a control for UAS copy-number dependence and did not detect significant changes in the signal. (E) Quantification of the signals of AT[RK] in class IV neurons over larval development. (F) Quantification of the FRET signals of AT[NL] in the cell body of wild-type and prel[1] mutant class IV MARCM clones at late L3. (G) Quantification of the FRET signals of AT[NL] at the distal parts of dendritic arbors (>100 µm distant from the cell body along dendrites) of class IV neurons. Prel and AT[NL] were expressed with a class IV–specific ppk-Gal4 driver. Bars, 5 µm. *, P < 0.05; ns, not significant.

Figure 3.

Prel-overexpressing class IV neurons preserve ATP levels by alterations in energy supply and demand. (A) FRET imaging of Class IV neurons expressing AT[NL] plus Prel in the absence or presence of 2DG (50 mM, 30 min). Pseudocolored images (A, left) and quantification (A, right) of the FRET signals of AT[NL]. (B and C) Time courses of changes in the FRET signals of AT[NL] in class IV neurons with Prel O/E (B) or loss of function of prel (C) in the absence or presence of AM (50 µM; n = 5 for each genotype) or oligomycin (OM; an inhibitor of ATP synthase; 100 µM; n = 7 for each genotype). Values are mean ± SD (D) 2-NBDG imaging of class IV neurons of dissected larvae. Representative images of 2-NBDG signals around the cell bodies of wild-type (top row) and Prel O/E (second row) class IV neurons (left). Yellow arrowheads indicate cell bodies of class IV neurons, and red arrowheads indicate the accumulation of 2-NBDG in a wild-type larvae. To clearly reveal features, these images were nonlinearly adjusted. We observed six class IV neurons from six larvae for each genotype, and similar patterns of 2-NBDG fluorescent signals were observed. Representative two-dimensional graphs of fluorescent signals along a line crossing the cell body of control and Prel-overexpressing class IV neurons (D, right). Quantification is based on images that are not nonlinearly adjusted. Each x axis represents distance along the respective lines. Black and red double-headed arrows under the graphs indicate the regions of the cytoplasm and the nucleus, respectively. Fluorescent peaks of the control were not located within class IV neurons (visualized by myr-mRFP) but surrounded the class IV soma. These signals may represent 2-NBDG captured by glial cells that wrap around cell bodies and axons of da neurons (Yamamoto et al., 2006; see also Fig. S1, G–I). Consistent with this notion, glial cells in Drosophila partially degrade sugars and provide alanine and lactate as intermediate metabolites for neurons (Schirmeier et al., 2015). (E) Real-time FRET imaging revealed that Prel O/E reduced the rate of ATP consumption of class IV neurons. Representative pseudocolored images of the FRET signals of soma expressing AT[NL] in control (top row) and Prel-overexpressing (second row) class IV neurons in the presence of 100 µM AM and 50 mM 2-DG (E, left). Changes in mean values of the AT[NL] signal (E, middle). Error bars represent SD. Plots of the signal decline of each neuron (E, right). Bars, 5 µm. *, P < 0.05; ns, not significant.

Figure 4.

Promotion of eIF2α dephosphorylation prevents the dendritic loss induced by prel O/E. (A) The effect of repression of candidate pathways of mitochondria-derived stress signaling on the dendritic loss induced by Prel O/E in class IV neurons. bsk[DN], a dominant-negative form of JNK; p53[DN], a dominant-negative form of p53; SNF1A[K57A], a kinase inactive form of AMPK catalytic subunit (Adachi-Yamada et al., 1999; Ollmann et al., 2000; Johnson et al., 2010). Expression of p35, the baculovirus-derived inhibitory protein of effector caspases, slightly restored dendritic length of Prel-overexpressing class IV neurons (Hay et al., 1994). However, the degree of recovery was not striking, indicating a minor role of caspase activation in the dendritic loss. Analysis of variance (ANOVA) followed by post-hoc Dunnett’s test (versus mCherry-CAAX). (B) Immunoblot analysis revealed promotion of eIF2α dephosphorylation by dPPP1R15 O/E. dPPP1R15 was overexpressed using hs-Gal4, which drives Gal4 expression by a heat-inducible Hsp70 promoter. Whole larvae treated with a heat shock (37°C for 1 h) were lysed 10 h after the treatment. Lysates were immunoblotted with anti-phosphorylated eIF2α (P-eIF2α) and anti-eIF2α antibodies. Results of three biological replicates are shown. (C–G) Genetic interactions between O/E or loss-of-function of prel and eIF2α phosphorylation-related genes. Quantification of the effects of coexpression of dPPP1R15, S6k[STDETE], and mCherry-CAAX (UAS copy number control) with Prel on the dendritic length of class IV neurons (C). Note that the values of the total dendritic length of neurons that expressed dPPP1R15 and Prel were apparently distributed in a bimodal fashion (see also E). S6k[STDETE]: an active form of S6k (Barcelo and Stewart, 2002). ANOVA followed by post-hoc Dunnett’s test (versus mCherry-CAAX). Representative images of the dendritic arbors of Class IV neurons expressing Prel plus mCherry-CAAX (D) or dPPP1R15 (E; see also Fig. S4 A). An example of dendritic arbors (F, left) and quantification of the total dendritic length (F, right) of prel mutant class IV MARCM clones overexpressing dPPP1R15. Representative images (G, left) and quantification of the total dendritic length (G, right) of class IV neurons with Prel O/E plus KD of Perk, Gcn2, or Thor/4EBP. The amplicon for RNAi of Perk[HMJ] line does not overlap with those of other RNAi lines. ANOVA followed by post-hoc Dunnett’s test (versus mCherry[KD]). (H) Representative images (left) and quantification of the total dendritic length (right) of dPerk-overexpressing class IV neurons. *, P < 0.05. Bars, 50 µm.

Figure 5.

Differential eIF2α phosphorylation–mediated translational repression in da neurons. (A) Prel O/E increased P-eIF2α in vivo. Prel was overexpressed by using hs-Gal4. Brain lysates of control and Prel-overexpressing adult female flies treated with heat shocks were immunoblotted using antibodies against P-eIF2α or total eIF2α. Immunoblots of two biological replicates are shown and relative values of band intensities are indicated (left). Quantification of the band densities of five biological replicates (right). (B) The increase in the P-eIF2α amount by Prel O/E was partially suppressed by KD of Perk (Perk[GL]). A representative immunoblot of brain lysates and relative values of band intensities (left). Quantification of the band densities of 19 biological replicates (right). (C) Kaede imaging in da neurons detected translational repression induced by cycloheximide (CHX), an inhibitor of protein translation. Early L3 larvae expressing Kaede in da neurons were prefed with a fly food containing CHX for 4 h before photoconversion (PC). Whole larvae were irradiated with UV light. Intensity of unconverted, green Kaede fluorescence in the soma was quantified. Half of the photoconverted larvae were observed immediately after PC (0 h), and the other half were aged in the fly food containing CHX for 6 h, then observed (6 h). AU, arbitrary units. (D–H) Prel O/E preferentially suppressed representative protein synthesis in class IV neurons. Grayscale images of unconverted Kaede fluorescence in control (D and E), Prel O/E (F), and Prel plus dPPP1R15 coexpression (G). Whole early L3 larvae were irradiated with UV light and aged for 15 h. Arrows indicate the soma of class IV (white), class I (magenta), and class III (yellow) neurons. Bars, 50 µm. Quantification of unconverted Kaede signals in the soma of da neurons (H). Unconverted Kaede signals immediately after and 15 h after PC were quantified. ANOVA with the post-hoc Tukey-Kramer method. (I) Quantification of the intensity of photoconverted Kaede just after conversion (0 h) and 6 h after conversion (6 h) in class IV neurons of either control or Prel O/E. *, P < 0.05; ns, not significant.

Figure 6.

Prel O/E activates the Ire1 branch of the UPR. (A–G) The XBP1-EGFP sensor accumulated in the nucleus of da neurons. Representative images of XBP1-EGFP and mCherry-CAAX in the cell bodies of control class IV neurons (A–C) and Prel-overexpressing class IV neurons (D–F) at early L2 (48–52 h AEL). XBP1-EGFP and mCherry-CAAX signals are presented in green and magenta, respectively (A–F, left). Yellow arrows indicate XBP1-EGFP signals accumulated in the nucleus (A–F, right). (G) Quantification of the proportion of neurons with XBP1-EGFP accumulation in the nucleus during larval development. Quantification of fluorescence intensity of XBP1-EGFP in the nucleus at early L2 yielded similar results (see Table S1). *, P < 0.05.

Figure 7.

Increased eIF2α phosphorylation is a common mechanism of dendritic loss induced by various mitochondrial lesions. (A) Representative images of mitochondrial morphologies in the cell body of class IV neurons expressing Opa1, Opa1[K273A], or Ttm50. Mitochondria were visualized with mitoGFP. (B–I) Expression of Opa1, Opa1[K273A], or Ttm50 in class IV neurons induced dendritic losses, and dPPP1R15 O/E partially prevented dendritic losses. Representative images of dendritic morphologies of class IV neurons expressing Opa1(B), Opa1[K273A] (C), Ttm50 (D), Opa1 plus dPPP1R15 (F), Opa1[K273A] plus dPPP1R15 (G), or Ttm50 plus dPPP1R15 (H). Quantification of the total dendritic length of class IV neurons expressing Opa1, Opa1[K273A], or Ttm50 (E) and expressing dPPP1R15 in conjunction with Opa1, Opa1[K273A], or Ttm50 (I). (J and K) Representative images and quantification of dendritic morphologies of CoVa[tenured] mutant class IV neurons (J) and dPPP1R15 overexpressed in CoVa[tenured] mutant class IV neurons (K). We generated CoVa mutant class IV neurons in the heteromutant background by the MARCM method, which utilizes flippase-based somatic recombination. Loss of function of CoVa caused a modest decrease in dendritic length (J). This mild phenotype may be explained by the fact that mutant neurons inherit the enodogenous CoVa protein that was produced before the mitotic recombination event (Liu et al., 2000) and also that mitochondrial OXPHOS complexes display relatively slower turnover rates in the nervous system (Price et al., 2010; Vincow et al., 2013). dPPP1R15 OE in CoVa[tenured] mutant class IV neurons restored dendritic patterning (K). (L) Quantification of the FRET signals of AT[NL] in class IV neurons expressing Opa1, Opa1[K273A], or Ttm50 during larval development. *, P < 0.05.

Figure 8.

Protein translation consumes a small fraction of the total ATP budget in class IV neurons. (A) Real-time imaging of AT[NL] revealed ATP used for protein translation was not high in class IV neurons. Fillet larvae were pretreated with CHX or ouabain for 30 min before the addition of 100 µM AM plus 50 mM 2-DG. The time course of changes in the FRET signals in class IV neurons in the presence of cycloheximide (CHX) or ouabain, an inhibitor of Na⁺/K⁺ ATPases (left). Quantification of the decline of FRET signals in the soma of class IV neurons expressing AT[NL] treated with DMSO (control), 25 mM CHX, or 50 mM ouabain (right). See also Fig. S3 E. (B) Real-time imaging of AT[NL] revealed ATP used for protein translation was high in Drosophila BG2-c2 cells. The cells transiently expressing AT[NL] were treated with 20 µM AM and 20 mM 2-DG in conjunction with 20 µM CHX or 50 mM ouabain. DMSO, CHX, or ouabain was added 30 min before the addition of AM plus 2-DG. Time courses of changes in the FRET signals in BG2-c2 cells in the presence of CHX or ouabain are shown (left). Quantification of decline of FRET signals in BG2-c2 cells expressing AT[NL] treated with DMSO (control), 20 µM CHX, or 50 mM ouabain (right). See also Fig. S3 F. (C) Quantification of the FRET signals of AT[NL] in the soma of wild-type class IV neurons and class IV neurons with prel plus dPPP1R15 coexpression. *, P < 0.05; ns, not significant.