Dysfunctional BCAA degradation triggers neuronal damage through disrupted AMPK-mitochondrial axis due to enhanced PP2Ac interaction

Abstract

Metabolic and neurological disorders commonly display dysfunctional branched-chain amino acid (BCAA) metabolism, though it is poorly understood how this leads to neurological damage. We investigated this by generating Drosophila mutants lacking BCAA-catabolic activity, resulting in elevated BCAA levels and neurological dysfunction, mimicking disease-relevant symptoms. Our findings reveal a reduction in neuronal AMP-activated protein kinase (AMPK) activity, which disrupts autophagy in mutant brain tissues, linking BCAA imbalance to brain dysfunction. Mechanistically, we show that excess BCAA-induced mitochondrial reactive oxygen species (ROS) triggered the binding of protein phosphatase 2 A catalytic subunit (PP2Ac) to AMPK, suppressing AMPK activity. This initiated a dysregulated feedback loop of AMPK-mitochondrial interactions, exacerbating mitochondrial dysfunction and oxidative neuronal damage. Our study identifies BCAA imbalance as a critical driver of neuronal damage through AMPK suppression and autophagy dysfunction, offering insights into metabolic-neuronal interactions in neurological diseases and potential therapeutic targets for BCAA-related neurological conditions.

Studies on Drosophila BCAA-catabolism mutants reveal disrupted coordination between AMPK, PP2Ac, and mitochondrial function in brain tissues, uncovering a pivotal mechanism linking metabolic dysfunction to neuronal damage.

Introduction

The homeostasis of essential branched-chain amino acids (BCAAs) (e.g., leucine, isoleucine, and valine) is critical for organismal health, playing a key role in peripheral metabolic balance and immunity¹. Dysfunctional BCAA degradation is a potential biomarker of many diseases, such as insulin resistance, diabetes, cancer, and neurological disorders^1,2, including the rare, inherited maple syrup urine disease (MSUD). The branched-chain aminotransferase (BCAT) and the branched-chain α-keto acid dehydrogenase (BCKDH) complex, which drives the second step of BCAA breakdown, are central regulators in this metabolic pathway³ (Supplementary Fig. 1a). Interestingly, adult patients with MSUD caused by mutations to the BCKDH commonly present with symptoms of Parkinson’s disease⁴. There is further evidence of a potential causal link between a deficiency in another protein known to cause MSUD⁵, BCAT, and Parkinson’s in a C. elegans model⁶, Alzheimer’s in a mouse model⁷, and autistic features in human patients⁸. Overall, this supports the idea that dysregulated BCAAs might be involved in neuropathology, but the underlying interactions remain to be determined. Notably, the interesting similarities between MSUD and other neurological diseases indicate that studying the pathways known to be responsible for MSUD might shed light on these other more common diseases.

MSUD affects approximately 1 in 150,000 infants worldwide⁹. Neurological dysfunction is a key symptom of MSUD, which is characterized by developmental delays, encephalopathy, and neurodegeneration^10,11 as well as documented neuronal loss¹². Though an understanding of the underlying genetic causes of MSUD, clinical phenotypes of MSUD lack good genotype-phenotype correlation and are based on the residual BCKDH activity⁵. The most severe cases, characterized by 0–2% residual BCKDH activity, represent the majority of MSUD cases^5,13. Although animal models, including mice and zebrafish deficient in the dihydrolipoamide branched-chain transacylase (DBT) protein, a subunit of the BCKDH complex^14,15, have been developed to investigate MSUD, the mechanisms driving the neurological dysfunction associated with the disease remain poorly understood. Additionally, aberrant BCAA overload has been recognized as a detriment for MSUD, though the mechanisms underlying the neuronal malfunction and amino acid imbalance remain obscure.

To investigate the effect of BCAA dysregulation on organismal fitness, we previously disrupted BCAA-catabolizing activity by generating a Drosophila DBT null mutation (dDBT^∆) using CRISPR/Cas9 in the conserved gene encoding for the enzyme comprising the E2 subunit of BCKDH complexes¹⁶. This generated an excess BCAA accumulation model in Drosophila melanogaster that displayed aberrant BCAA accumulation, poor mobility, and developmental defects that recapitulated the neurological symptoms of MSUD. Overall, this previous work indicated that BCAA homeostasis in Drosophila is controlled by the evolutionarily conserved modulation of the activity of identical BCAA-catabolizing enzymes. From this study, we also identified several proteins of BCAT and other BCKDH enzyme complexes in Drosophila predicted to be relevant to MSUD. These are encoded by the genes CG1673 (henceforth dBCAT), CG8199 (E1α; henceforth dBCKDHA), CG17691 (E1β; henceforth dBCKDHB), and CG7430 (E3; henceforth dDLD). However, the detailed roles of these genes in BCAA catabolism in Drosophila and the consequent effects of mutations in these genes remained unknown.

Thus, we herein generated and extensively tested four Drosophila variants with mutations of enzyme-encoding genes that we previously predicted to have BCAA-catabolizing activity (i.e., dBCAT, dBCKDHA, dBCKDHB, and dDLD genes). Similar to the dDBT^∆ mutant, these mutants exhibited aberrant BCAA accumulation in the circulating cavity and neurological dysfunction. Using these BCAA–catabolism-disabled Drosophila mutants, we determined that neurological dysfunction was attributed to declined signaling of the AMPK-autophagy pathway in brain tissue due to aberrant circulation of excess BCAAs. While dissecting the mechanistic models of these signaling pathways, we found that mitochondrial redox homeostasis was also disrupted by an excess of dysregulated BCAAs, which created reactive oxygen species (ROS) stress that triggered enhanced binding of PP2Ac to AMPK. This enhancement attenuated AMPK activity and evoked a feedback loop between the declined AMPK activity and mitochondrial dysfunction to cause neuronal damage within brain tissues. Altogether, our findings uncover a previously unrecognized mechanism elucidating how dysregulated BCAAs lead to neurological injury. This discovery advances our understanding of metabolic-neuronal crosstalk and suggests that neuronal AMPK activation could provide a therapeutic strategy for BCAA dysregulation-associated MSUD and other more common neurological diseases, particularly given the ongoing debate regarding its efficacy across different contexts.

Results

Dysfunctional BCAA catabolism harms fitness in Drosophila

To explore the central effectors that concurrently arise from various scenarios of insufficient BCAA catabolism, we first generated Drosophila dBCAT, dBCKDHA, dBCKDHB, and dDLD-deficient strains (Supplementary Fig. 1b) based on predicted protein signatures and known CRISPR/Cas9 methodology from our previous work¹⁶. Physiologically similar to dDBT^Δ, the homozygous mutants were viable from embryo to larvae, except for the embryonically lethal dDLD^Δ mutant; the heterozygous dDLD^Δ mutant remained viable. To confirm these target genes were disrupted by genetic excision, we performed quantitative RT-PCR analysis. In homozygous mutants, the mRNA of the targeted BCAA-catabolizing enzyme was not expressed, though the heterozygous dDLD mutant retained moderately detectable gene expression that was lower than that of the control flies (Supplementary Fig. 1c).

To characterize the functions encoded by these genes, we determined whether their activities contributed to BCAA catabolism. To quantify circulating levels of BCAAs (e.g., leucine, isoleucine, and valine), we used liquid chromatography-mass spectrometry (LC-MS), which revealed a high accumulation of BCAAs in each of the mutants (Fig. 1a). This suggested that these genes in Drosophila participate in the modulation of BCAA catabolism. Since MSUD symptoms arise from aberrant BCAA accumulation, we studied these BCAA-catabolizing enzyme mutants for similar symptoms of neuronal damage, poor mobility, and developmental delays. In fact, these mutants did also present with similar neurological symptoms, as evidenced by dysfunctional development in larval pupation (Fig. 1b) and eclosion (Fig. 1c), poor crawling behavior in larvae (Fig. 1d), as well as neuronal damages in older heterozygous adults (Supplementary Fig. 2).

Fig. 1. Defects in fitness and brain autophagy response in BCAA metabolism-disabled Drosophila.

a LC-MS analysis of BCAA levels in Drosophila larval body fluid (30 flies per group/experiment; n ≥ 3). b, c Developmental assay of pupation (b) and eclosion (c) rates (≥ 100 flies per group/experiment; n = 3). d Crawling behavior assay (6 flies per group/experiment; n = 3). e–g Confocal images of Atg8 puncta shown in central brain region (3–5 larval brains per group/experiment; n = 5) (e), western blot analysis of Atg8 and Ref(2)p protein expression in whole larval brains (f), and confocal images of whole larval brains stained with Lysotracker dye (g). All experiments were individually conducted in five Drosophila strains containing mutations in BCAA-catabolizing enzymes compared to w¹¹¹⁸. Larval crawl was recorded by video-tape and presented with ImageJ. Each experiment was conducted with at least three biological replicates. Statistical significance defined as *p < 0.05; **p < 0.01; ***p < 0.001. Error bars indicate mean ± SEM. Scale bars: 25 μm (for e); 100 μm (for g).

Since an excess of circulating BCAAs harms patient outcomes⁵, we thus assessed the effects of a high-protein diet on the BCAA-catabolizing mutants. Adult heterozygous mutants fed a high-protein diet exhibited a more significant accumulation of BCAAs compared to the control flies, with no significant disruption to threonine homeostasis (Supplementary Fig. 3a). This group also had shorter lifespans than the control (w¹¹¹⁸) (Supplementary Fig. 3b) and displayed behavioral deficits (Supplementary Fig. 3c), indicating that a dysregulated excess of BCAAs is harmful for Drosophila mutants that are incapable of BCAA catabolism. Altogether, we characterized that the previously unidentified Drosophila BCAA-catabolizing enzyme-encoding genes function in the metabolic homeostasis of BCAAs, as in mammals. Additionally, these mutants exhibited heat-induced seizure-like behavior (Supplementary Fig. 4), reminiscent of features commonly associated with amino acid metabolism disorders¹⁷. Overall, as these BCAA metabolism-disabled Drosophila mutants closely mimic the symptoms of MSUD, they make an ideal model for dissecting the underlying mechanism of neuropathology associated with BCAA dysregulation.

BCAA dysregulation perturbs autophagy response in Drosophila

As BCAA dysregulation may affect energy homeostasis, we next looked to the self-degradative autophagy system, which is important for balancing sources of energy^18–20 and is activated by nutrient starvation²¹. To investigate whether the autophagy response is affected by dysregulated BCAAs in these Drosophila mutants, we compared the effects of starvation on our generated mutants versus the control flies. To measure autophagy activation in Drosophila, we monitored several indicators expected to increase upon activation, such as autophagosome formation, cytoplasmic Atg8 punctuation, and expression of lipidated-Atg8 (termed Atg8-II, an ortholog of mammalian LC3-II). First, we examined autophagosome formation using LysoTracker, a fluorescent dye for labeling and tracking acidic organelles in live cells²², which showed poor formation of autolysosomes in larval fat bodies (Supplementary Fig. 5a), an organ responsible for energy storage and utilization²³. Consistent with this, Atg8-II expression within fat bodies was reduced in starved mutants (Supplementary Fig. 5b), indicating reduced autophagy function. To measure Atg8 punctation, we specifically expressed exogenous mCherry-Atg8a in fat bodies using genetic GAL4-UAS manipulation²⁴ driven by a fat–body-specific driver called r4-Gal4 (r4-Gal4 > UAS-mCherry-Atg8). Upon starvation, the fat bodies of mutants expressing this tagged Atg8 had less Atg8 puncta (Supplementary Fig. 5c).

Though these results show that BCAA-catabolizing mutants may fail to activate autophagy, nutrient responses are likely to be highly tissue-specific²⁵. Therefore, we assessed autophagy activity in nutrient-demanding brain tissue upon dysfunctional BCAA degradation. Here, there was a reduction in the steady autophagy response in brain tissues of all of the mutants as evidenced by scarce expressions of Atg8 puncta (Fig. 1e) and lipidated Atg8-II proteins (Fig. 1f) as well as reduced autolysosome formation (Fig. 1g). Additionally, impaired autophagy was marked by the accumulation of Ref(2)P, a homolog of the mammalian ubiquitin-binding protein p62, which is indicative of autophagic failure²⁶. Ref(2)p accumulation (Fig. 1f) supports that autophagy is defective in the mutant brains. These results imply that excess BCAAs might alter the machinery involved in brain energy homeostasis.

Decline in brain AMPK results in BCAA-associated neurological dysfunction

AMP-activated protein kinase (AMPK) acts as a master regulator in energy balance and autophagy^27–29, though its specific role in neurological function and amino acid homeostasis remains to be elucidated. To thus assess whether AMPK activity in the brain might be affected by BCAA dysregulation, we monitored Thr184 phosphorylation in an α subunit of Drosophila AMPK, which is known to be responsible for full activation²⁷. Here, AMPK activity declined significantly in brain tissues of all mutants (Fig. 2a), indicating that AMPK activity is indeed affected upon BCAA dysregulation. To clarify whether AMPK activity is a driving force of these effects, we genetically overexpressed exogenously active or inactive AMPKα in the brains, driven by the brain-specific driver elav-Gal4 via the GAL4-UAS genetic system²⁴. Neuronal active AMPKα overexpression (elav-Gal4 > UAS-mCherry-AMPKα^WT) improved developmental deficits as well as poor mobile behavior in several mutants, such as dBCAT^Δ, dDBT^Δ, and dDLD^Δ (Fig. 2b, d). In contrast, inactive AMPKα^KA overexpression (elav-Gal4 > UAS-AMPKα^K57A) increased the developmental and behavioral defects in these same mutants (Fig. 2c, d). We further found that reinforcing active AMPKα expression in dDBT^Δ brain tissues alleviated previously characterized neuronal apoptosis¹⁶, while AMPKα^KA overexpression exacerbated it (Fig. 2e). Additionally, neuronal reinforcement of AMPKα expression extended the shortened lifespan in adult dDBT^Δ/+ mutant (Fig. 2f), particularly under conditions of a high-protein diet (Fig. 2g). These findings suggest that compromised AMPK function in the brain contributes to neurological dysfunction and overall health impairment upon dysfunctional BCAA catabolism.

Fig. 2. The decline of the neuronal AMPK-autophagy axis contributes to neurological dysfunction in BCAA-catabolizing enzyme mutants.

a Western blot analysis of AMPKα phosphorylation in whole larval brains of BCAA-catabolizing mutants or w¹¹¹⁸. b, c Developmental analysis of pupation and eclosion rates in the mutants without (elav alone) or with neuronal overexpression of wildtype AMPKα (elav>mCherry-AMPKα) (b) or inactive AMPKα (elav > AMPKα^KA) (c). d Crawling behavior of w¹¹¹⁸ or mutants in the absence or presence of neuronal AMPKα or AMPKα^KA expression. The genetic mutants of dBCAT^Δ/y; UAS-mCherry-AMPK^αWT/+; elav-Gal4/+ or dBCAT^Δ/y; UAS-AMPKα^KA/+; elav-Gal4/+, dDBT^Δ/y; UAS-mCherry-AMPK^αWT/+; elav-Gal4/+ or dDBT^Δ/y; UAS-AMPKα^KA/+; elav-Gal4/+, and dDLD^Δ/elav-Gal4; UAS-mCherry-AMPKα^WT/+ or dDLD^Δ/elav-Gal4; UAS-AMPK^αKA/+ were used for (b), (c), and (d). The dBCAT^Δ/y; elav/+, dDBT^Δ/y; elav/+, and elav/⁺; dDLD^Δ/+ mutants were used individually as controls for each group. e Confocal images of immunostained active caspase-3 signals in whole-mount brains of control larvae (elav/+) or dDBT^Δ mutants without (dDBT^Δ/y; +/+; elav-Gal4/+) or with neuronal AMPKα (dDBT^Δ/y; UAS-mCherry-AMPK^αWT/+; elav-Gal4/+) or AMPKα^KA (dDBT^Δ/y; UAS- AMPKα^KA/+; elav-Gal4/+) overexpression. 3–5 larval brains were used per group/ experiment (n = 5) f, g Lifespan assay of adult female dDBT^Δ/+ flies with neuronal overexpression of active AMPKα^WT or inactive AMPKα^KA (n = ~ 100 flies; across 3 vials) were fed without or with regular (f) or high protein (g) food. For median survival, 30–35 flies were used per group/ experiment (n = 3). h Western blot analysis of phospho-AMPKα and Atg8 protein expressions in whole larval brains of w¹¹¹⁸ or dDBT^Δ mutant fed without or with the indicated concentration of metformin. i Western blot analysis of Atg8, endogenous AMPKα, mCherry-tagged AMPKα protein expressions in whole larval brains of neuronal exogenously AMPKα-overexpressed dDBT^Δ mutant (dDBT^Δ/y; UAS^-mCherry-AMPK^αWT/+; elav-Gal4/+), as compared to dDBT^Δ control (dDBT^Δ/y; +/⁺; elav-Gal4/+) or w¹¹¹⁸. j, k Analysis of pupation rate and eclosion rate (j) and crawling behavior (k) in dDBT^Δ mutant or neuronal Atg1-depleted dDBT^Δ mutant (dDBT^Δ/y; UAS^-Atg1-RNAi/+; elav-Gal4/+) after feeding with 10 mM metformin, as compared to dDBT^Δ control mutant (dDBT^Δ/y; +/+; elav-Gal4/+) or w¹¹¹⁸. l Western blot analysis of Atg8 protein expression in whole larval brains of w¹¹¹⁸ or dDBT^Δ mutant fed without or with the indicated concentration of rapamycin. m, n Analysis of pupation rate and eclosion rate (m) and crawling behavior (n) in dDBT^Δ mutants orally fed without or with 50 mM rapamycin, as compared to w¹¹¹⁸. The fluorescence signals of caspase-3 in confocal images were measured by ImageJ and presented as relative intensity. For developmental analysis, n ≧ 100 flies were used per group/experiment. For crawling behavior analysis, n = 6 flies were used per group/experiment. At least three biological replicates were performed for each experiment. Statistical significance defined as *p < 0.05; **p < 0.01; ***p < 0.001. Error bars indicate mean ± SEM. Scale bars: 100 μm.

Brain AMPK-autophagy axis modulates fitness during BCAA dysregulation

Evidence suggests that autophagy is a crucial signal linking AMPK effects to maintaining energy balance during energy stress³⁰, as well as improving neurological outcomes following ischemic brain injury³¹. To investigate whether the decline in AMPK conferred by BCAA–dysregulation-associated neurological dysfunction is related to autophagy attenuation in brain tissues. To do this, we fed the dDBT^Δ mutant with metformin, a pharmaceutical drug that can activate AMPK in various organisms, including Drosophila²⁸. We found that metformin not only activates brain AMPK activity but also reverses the declined autophagy (Fig. 2h and Supplementary Fig. 6a). Furthermore, the genetic overexpression of neuronal AMPKα^WT (elav-Gal4 > UAS-AMPKα^WT) mitigated the autophagy deficit in dDBT^Δ brains (Fig. 2i and Supplementary Fig. 6b). We speculate from these results that an AMPK deficiency might hamper autophagy activation in mutant brain tissues.

To further examine the impact of the brain AMPK-autophagy axis in metformin-fed dDBT^Δ, we blocked brain autophagy in these mutants via the genetic depletion of Atg1 (elav-Gal4 > UAS-Atg1-RNAi), an initial upstream factor in the signaling cascade of autophagy activation³². Here, we showed that the beneficial effects of metformin were diminished in terms of development (Fig. 2j) and crawling behavior (Fig. 2k). These phenotypic outcomes support that autophagy might act as a beneficial downstream effector of AMPK activation in dDBT^Δ brain tissues. To validate this result, we fed the dDBT^Δ mutant with rapamycin, a pharmaceutical activator of autophagy, and observed that it not only activates brain autophagy (Fig. 2l) but also ameliorates the defective development and crawling behavior (Fig. 2m, n). Combined, our data suggest that the brain AMPK-autophagy axis deficit is responsible for neurological dysfunction in the presence of excess BCAAs due to their reduced catabolism.

Dysfunctional BCAA degradation as a determinant of brain AMPK attenuation

Even though the role of AMPK has been relatively understood in cellular metabolism in terms of sensing intracellular levels of nucleotides (e.g., AMP, ADP, and ATP) and glucose^29,33, the response of neuronal AMPK to BCAAs still remains unclear. Therefore, we investigated the impact of dysregulated excess BCAAs on AMPK activity to determine which factors attenuate neuronal AMPK in dDBT^Δ mutants. When fed extra leucine, brain AMPK activity deteriorated in dDBT^Δ, but no such change was observed in control flies (Fig. 3a). We then assessed whether brain AMPK attenuation was orchestrated by a direct effect or secondary regulation resulting from dysregulated excess BCAAs by ex vivo treating dissected dDBT^Δ or control flies brains with leucine alone. Our result showed that leucine excess further decreased AMPK activity in dDBT^Δ mutants but not the control flies (Fig. 3b), thereby implying that an excess of dysregulated BCAAs emerged as a direct suppressor for attenuating brain AMPK in conditions of reduced BCAA-catabolizing activity.

Fig. 3. Excess of circulating leucine dampens AMPK activity in dDBT^Δ brain.

(a, b Western blot analysis of phospho-AMPKα and total AMPKα protein expressions in whole larval brains of w¹¹¹⁸ or dDBT^Δ mutant with or without leucine feeding (a) or in the leucine ex vivo-treated dissected larval brains from w¹¹¹⁸ or dDBT^Δ mutant (b). c–f Western blot analysis of phospho-AMPKα and total AMPKα proteins in whole larval brain tissues (c), analysis of pupation rate and eclosion rate (d), analysis of crawling behavior (e), confocal images of dissected larval brains co-stained with anti-cleaved caspase-3 (red) and anti-ELAV antibodies (green) (f), from dDBT^Δ control mutant (dDBT^Δ/y; +/+; tub-Gal4/+) or ubiquitous exogenously dDBT-overexpressed dDBT^Δ mutant (dDBT^Δ/y; dDBT-HA/+; tub-Gal4/+) with or without 50 mM leucine feeding, as compared to w¹¹¹⁸. 3–5 larval brains were used per group/ experiment (n = 5). For developmental analysis, n ≧ 100 flies were used per group/experiment. For crawling behavior analysis, n = 6 flies were used per group/experiment. The fluorescence signals of active caspase-3 in confocal images were measured by ImageJ and presented as relative intensity. At least three biological repeats were performed each experiment. Statistical significance defined as *p < 0.05; **p < 0.01; ***p < 0.001; n.s., not significant. Error bars indicate mean ± SEM. Scale bars: 100 μm.

For further evidence that BCAA-catabolizing activity is required for BCAA homeostasis and that this functional deficit creates a sensitized environment for attenuating AMPK activity, we genetically compensated dDBT function in dDBT^Δ mutants by ubiquitously expressing exogenous HA-tagged dDBT proteins (tub-Gal4 > UAS-dDBT-HA). We showed here that AMPK activity was restored in dDBT^Δ brains compared to the control flies. Significantly, a dDBT^Δ brain compensated with dDBT showed no decline in neuronal AMPK activity caused by leucine feeding (Fig. 3c). Similarly to the adult mutants harmed by a high-protein diet (Supplementary Fig. 3), leucine-fed dDBT^Δ larvae showed developmental impairments (Fig. 3d), reduced motility (Fig. 3e), and neuronal apoptosis (Fig. 3f). However, leucine-induced damages to dDBT^Δ mutants were significantly alleviated by exogenous dDBT compensation (Fig. 3d–f), supporting the theory that dysfunctional BCAA degradation acts as a physiological effector and creates a sensitized environment for attenuating AMPK activity within dDBT^Δ brain tissues. Altogether, our data identified that circulating BCAAs play a primary role in dampening brain AMPK, thereby resulting in neurological damage due to dysfunctional BCAA catabolism.

Mitochondrial reactive oxygen species induced by dysregulated BCAAs diminish AMPK activity

We previously demonstrated that a dDBT^Δ brain displays stress and oxidative damage caused by ROS¹⁶, a pathological feature common to a multitude of neurological diseases³⁴. However, as the trigger of the ROS stress burst in the dDBT^Δ brain remains unclear, we next assessed whether neuronal ROS stress is correlated with a dysregulated excess of BCAAs. Therefore, we monitored ROS stress in larval brain tissues subjected to excess BCAAs by generating dDBT^Δ mutants carrying the ROS stress-responsive gstD-GFP reporter³⁵. The gstD-GFP reporter assay revealed that brain ROS stress, indicated by positive GFP signals, was elevated in leucine-fed dDBT^Δ mutants (Fig. 4a, b) and in the dissected dDBT^Δ brains treated ex vivo with leucine (Supplementary Fig. 7). This suggests that the ROS burst in a dDBT^Δ brain is closely associated with circulating BCAA dysregulation.

Fig. 4. Mitochondrial ROS stress confers BCAA-mediated AMPK attenuation in dDBT^Δ brain.

a, b Confocal images of larval brains immunostained with anti-GFP (green) and anti-ELAV (red) antibodies (a) or western blot analysis of GFP protein expression in whole larval brains (b) dissected from w¹¹¹⁸ (+/Y; gstD-GFP/+) or dDBT^Δ mutant (dDBT^Δ/Y; gstD-GFP/+) that carried gstD-GFP reporter without or with leucine feeding. c, d Western blot analysis of phospho-AMPKα protein expressions in whole larval brains of dDBT^Δ mutant co-fed leucine with or without 10 mM NAC (c) or in 10 µM H₂O₂ ex vivo-treated dissected dDBT^Δ brains, as compared to dDBT^Δ mutant alone or w¹¹¹⁸ (d). e Confocal images of DHE dye-stained whole larval brains of dDBT^Δ mutant or w¹¹¹⁸. f Western blot analysis of phospho-AMPKα and Atg8 proteins expressions from SOD2-overexpressed whole larval brains of dDBT^Δ mutant (dDBT^Δ/Y; UAS-Sod2/+; elav-Gal4/+), as compared to dDBT^Δ control (dDBT^Δ/Y; UAS-gfp/+; elav-Gal4/+), dDBT^Δ mutant alone or w¹¹¹⁸. g–i Confocal images of larval brains with 4HNE (yellow) or anti-cleaved caspase3 (red) immunostaining (g), analysis of crawling behavior (h), pupation rate and eclosion rate (i) were performed from neuronal SOD2-overexpressed dDBT^Δ mutant (dDBT^Δ/Y; UAS-Sod2/+; elav-Gal4/+), as compared to dDBT^Δ control (dDBT^Δ/Y; +/+; elav-Gal4/+) or w¹¹¹⁸. For caspase 3-staining analysis, 3–5 larval brains were used per group/ experiment (n = 5). For developmental analysis, n ≧ 100 flies were used in each group/experiment. For crawling behavior analysis, n = 6 flies were used in each group/experiment. The fluorescence signals of caspase-3 or 4HNE in confocal images were measured by ImageJ and presented as relative intensity. At least three biological repeats were performed each experiment. Statistical significance defined as *p < 0.05; **p < 0.01; ***p < 0.001. Error bars indicate mean ± SEM. Scale bars: 100 μm.

Though AMPK has long been primarily thought of as an energy-sensitive kinase, there is a growing body of evidence that it is also redox-sensitive and regulated by oxidative stress^36–38. This prompted us to examine whether a decline in brain AMPK activity is associated with elevated ROS stress in the presence of excess BCAAs. To thus examine whether brain AMPK activity that had been deteriorated by an excess of leucine could be affected by the administration of an antioxidant, leucine-fed dDBT^Δ mutants were co-fed with N-acetylcysteine (NAC), an antioxidant drug³⁹. Our results here showed improvement upon NAC administration (Fig. 4c), suggesting that ROS stress mediates or directs the attenuation of AMPK activity. To further clarify this point, we ex vivo treated the dissected dDBT^Δ brains with hydrogen peroxide (H₂O₂), which further deteriorated the already declined AMPK activity (Fig. 4d). Collectively, these results support that ROS stress participates in inhibiting AMPK activity in the presence of dysregulated BCAA.

We noticed that the loss of dDBT activity also led to the reduced expressions of mitochondrial complexes I, II and IV (Supplementary Fig. 8). Considering that mitochondria are the primary source of intracellular ROS due to electron transfer during ATP production^40,41 and that a deficiency in those mitochondrial complexes elevates mitochondrial ROS (mROS)^42–44, we subsequently examined the effect of mROS stress on AMPK activity. We first used a fluorescence dihydroethidium (DHE) dye as a probe for ROS, which revealed that the concentration of cellular superoxide was highly increased in dDBT^Δ brain tissues (Fig. 4e). We then decreased mROS stress in a dDBT^Δ brain via the genetic overexpression of the superoxide dismutase 2 (SOD2) protein (elav-Gal4 > UAS-dSod2), the principal scavenger of mitochondrial superoxide⁴⁵. This expression relieved the reduced AMPK activity and autophagy response (Fig. 4f) and reduced the associated the neuronal damage (Fig. 4g), as evidenced by reduced brain apoptosis and oxidative damage visualized using lipid peroxidation with 4-hydroxynonenal (4-HNE) immunostaining⁴⁶. Predictably, SOD2 expression also reduced the defective crawling activity and developmental phenotypes (Fig. 4h, i). Hence, our results indicate that intracellular redox status acts as a determinant for the inhibition of neuronal AMPK activity within a dDBT^Δ brain. Altogether, we can speculate that mitochondrial dysfunction may be one cause of bursting ROS stress that attenuates the cytosolic AMPK activity within brain tissues during dysregulated BCAA accumulation.

Dysfunctional AMPK-mitochondrial axis communicates within dDBT^Δ brain

We next used aggregation of the dye JC-1 to assess the mitochondrial membrane potential, which serves as a key indicator of mitochondrial activity and controls the respiratory rate, ATP synthesis, and the generation of ROS⁴⁷. In dDBT^Δ brains, we observed poor mitochondrial membrane potential (reduced aggregation of JC-1 dye) (Fig. 5a) and attenuated ATP production (Fig. 5b), indicating mitochondrial malfunction. As AMPK is also considered a crucial sensor and regulator of mitochondrial homeostasis²⁸, we examined whether feedback from AMPK activity could impact mitochondrial function in dDBT^Δ brain tissues. We reinforced AMPK activity within dDBT^Δ brains by metformin feeding or neuronal AMPKα overexpression, which improved the observed mitochondrial malfunction. Conversely, the overexpression of inactive AMPKα^KA further deteriorated mitochondrial function (Fig. 5a, b). Additionally, AMPK activity seemed to orchestrate mROS stress in dDBT^Δ brains, which was observed using MitoSOX dye, an indicator of mROS stress. Here, metformin or neuronal AMPKα overexpression decreased mROS stress in the dDBT^Δ brain tissues, whereas AMPKα^KA overexpression increased mROS (Supplementary Fig. 9).

Fig. 5. Insufficient AMPK exacerbates mitochondrial malfunction in dDBT^Δ brain.

a Confocal images of larval brains (3–5 brains per group/experiment; n = 5) stained with JC-1 dye, wherein the red/green (aggregates/monomers) fluorescent intensity was quantified to represent relative mitochondrial membrane potential. b The relative level of ATP production in larval brains (50 brains per group/ experiment; n = 3). c Confocal images of larval brains (3–5 brains per group/experiment; n = 5) co-immunostained with anti-4 hydroxynonenal (4HNE) and anti-ELAV antibodies. d Malondialdehyde (MDA) assay used for quantification of lipid peroxidation (150 brains per group/ experiment; n = 5). All above experiments used larval brain tissues dissected out from w¹¹¹⁸, dDBT^Δ mutant (dDBT^Δ/Y; +/+; elav-Gal4/+) orally administrated without or with 10 mM metformin, neuronal exogenously AMPKα-overexpressed (dDBT^Δ/Y; UAS-mCherry-AMPKα^WT/+; elav-Gal4/+), or AMPKα^KA-overexpressed dDBT^Δ mutant (dDBT^Δ/Y; UAS-AMPKα^KA/+; elav-Gal4/+). The fluorescence signals in confocal images were measured by ImageJ and presented as relative intensity. At least three biological replicates were performed. Statistical significance defined as *p < 0.05; **p < 0.01; ***p < 0.001. Error bars indicate mean ± SEM. Scale bars: 100 μm.

Furthermore, we also examined whether increasing AMPK activity functionally protected brain oxidative damage from mROS stress in dDBT^Δ mutants using a series of lipid peroxidation analyses, including staining by 4HNE (Fig. 5c) and malondialdehyde (MDA) (Fig. 5d). From these assays, it does appear that reinforcing brain AMPK activity can attenuate the oxidative damage to the brain and vice versa (Fig. 5c, d). Overall, because an increase in mitochondrial ROS stress suppressed AMPK activity (Fig. 5), these results suggest that there is a reciprocal interaction between mitochondrial function and AMPK activity in dDBT^Δ brains. Together, these data reveal that a malfunctional brain AMPK-mitochondrial axis in dDBT^Δ triggers feedback that causes the neurological damages associated with BCAA dysregulation.

Enhanced PP2Ac-AMPK protein interaction reduces AMPK activity

We next sought to elucidate the mechanisms contributing to the decline of AMPK within the dysregulated BCAA-mROS axis. While AMPK activity is known to be finely regulated by protein phosphatase 2 A (PP2A)^48–50, which has essential regulatory functions in brain development and function^51,52, the influence of PP2A on neurological damages from dysregulated BCAAs and its modulation of AMPK remain unclear. In Drosophila, the gene microtubule star (mts) encodes a catalytic subunit of PP2A⁵³ (referred to as dPP2Ac) that shares 94% sequence homology with its human counterpart (Supplementary Fig. 10a). To investigate the role of dPP2A in regulating AMPK activity in mutants lacking BCAA catabolism, we initially exposed dissected dDBT^∆ larvae brains to the PP2A inhibitor okadaic acid, which restored AMPK activity (Fig. 6a). AMPK activity was also improved via genetic knockdown of mts expression in dDBT^∆ brain tissues (elav-Gal4>mts-RNAi) (Fig. 6b).

Fig. 6. Enhanced PP2Ac binding reduces AMPK activity in dDBT^Δ brain via BCAA-mROS axis.

a–c Western blot analysis of phospho-AMPKα, dPP2Ac or Flag protein expressions in whole dissected dDBT^Δ larval brains exposed to okadaic acid ex vivo (a), in mts-depleted whole larval brains (dDBT^Δ/Y; elav-Gal4/+; UAS-mts-RNAi/+) (b), or in inactive mts-overexpressed (dDBT^Δ/Y; elav-Gal4/+; UAS-mts^H118N-Flag/+) or active mts-overexpressed (dDBT^Δ/Y; elav-Gal4/+; UAS-mts^Y307F-Flag /+) whole larval brains (c), compared to dDBT^Δ mutant alone or w¹¹¹⁸. d, e Co-immunoprecipitation assay of dPP2Ac and AMPK in mCherry-AMPKα-overexpressed whole larval brains of dDBT^Δ mutant (dDBT^Δ/Y; UAS-mCherry-AMPKα^WT/+; elav-Gal4/+) with or without leucine feeding (d), carrying the presence or abscence of neuronal SOD2 expression (e). After immunoprecipitation with an anti-mCherry antibody, western blot analysis was performed via immunoblotting with anti-PP2Ac antibody. Quantitative analysis of the signal intensity for dPP2Ac or mCherry from three independent co-immunoprecipitation experiments was performed using ImageJ software. f Confocal images of lipid peroxidation in central brain sections stained with C-11 BODIPY 581/591 dye. g MDA assay of whole larval brains (150 brains per group/experiment; n = 5). The larval brain samples were collected or extracted from control flies (w¹¹¹⁸) or dDBT^Δ mutant without or with neuronal inactive mts overexpression (dDBT^Δ/Y; elav-Gal4/+; UAS-mts^H118N-Flag/+) or active mts overexpression (dDBT^Δ/Y; elav-Gal4/+; UAS-mts^Y307F-Flag/+), or mts depletion (dDBT^Δ/Y; elav-Gal4/+; UAS-mts-RNAi/+). Scale bars: 10 μm. Confocal microscopy: 63x oil magnification. h Analysis of larval development, including pupation rate and eclosion rate (≧ 100 flies were used per group/experiment; n = 3). i, j Adult lifespan assay (n = ~ 100 flies; across 3 vials). For median survival, 30–35 flies were used per group/ experiment (n = 3). dDBT^Δ larvae or female dDBT^Δ/+ adult flies with neuronal overexpression of inactive mts^H118N or active mts^Y307F proteins were fed without or with regular (h, i) or high protein (j) food. At least three biological replicates were performed. Statistical significance defined as *p < 0.05; **p < 0.01; ***p < 0.001. Error bars indicate mean ± SEM.

As PP2A activity is negatively regulated by a conserved tyrosine at 307 (Y307)⁵³, we investigated the effects of post-translational modifications to PP2Ac^54,55 by generating two transgenic Drosophila mts mutants, namely the inactive (UAS-mts^H118N) or active (UAS-mts^Y307F) forms of dPP2Ac. Here, neuronal overexpression of inactive mts^H118N, driven by elav-Gal4, exhibited a dominant-negative effect and alleviated the reduced AMPK activity. Conversely, active mts^Y307F overexpression exacerbated the reduction in AMPK activity in dDBT^∆ brains (Fig. 6c). Combined, these results indicate a regulatory role for dPP2Ac in modulating AMPK activity in dDBT^∆ brains.

Next, we explored how dPP2Ac might modulate AMPK activity in dDBT^∆ by examining the dPP2Ac-AMPK interaction. Using a co-immunoprecipitation assay, genetic overexpression of mCherry-tagged AMPKα protein in larval brains revealed a direct interaction between dPP2Ac and AMPK in dDBT^∆ but not wildtype (Fig. 6d). However, dPP2Ac activity remained unchanged in these scenarios, as no observable alteration in Y307 phosphorylation of dPP2Ac was noted in any of the five BCAA–catabolism-dysfunctional mutants (Supplementary Fig. 10b). Thus, the direct interaction of dPP2Ac emerged as a pivotal process in the decline of neuronal AMPK activity. To further clarify the roles of excess BCAAs and mROS stress in the dPP2Ac-AMPK interaction, dDBT^∆ larvae fed a high leucine diet had a stronger dPP2Ac-AMPK interaction (Fig. 6d). Additionally, neuronal SOD2 overexpression mitigated the dPP2Ac-AMPK interaction in brain tissues of dDBT^∆ larvae (Fig. 6e). These findings suggest that a dysregulated excess of BCAAs and brain mROS stress act as physiological effectors promoting the dPP2Ac-AMPK interaction, thereby reducing AMPK activity. This highlights a potential role of PP2Ac in triggering neuronal damage and compromising overall health along the dDBT^∆-BCAA-mROS axis.

dPP2Ac ablation mitigates neurological damages and shortened lifespan in dDBT^∆

Recognizing the pivotal role of ROS stress-induced lipid peroxidation in dDBT^∆-induced neuronal damage¹⁶, we next sought to investigate the impact of PP2Ac. Overexpression of inactive mts^H118N or mts knockdown mitigated this effect, as evidenced by C11-BODIPY 581/591 staining (Fig. 6f) and a MDA assay (Fig. 6g), and these treatments alleviated developmental defects in dDBT^∆ (Fig. 6h). In contrast, the overexpression of active mts^Y307F increased the accumulation of lipid peroxidation and exacerbated developmental defects (Fig. 6f–h). This indicates that PP2Ac contributes to dDBT^∆-induced neuronal damage.

While homozygous dDBT^∆ mutants did not survive to adulthood, heterozygotes did (Supplementary Fig. 3), reinforcing the notion that the severity of MSUD is closely linked to the extent of disrupted BCAA homeostasis and residual BCAA catabolic activity. Consistent with observations at the larval stage, a heightened dPP2Ac-AMPK interaction was evident in the brain tissues of adult heterozygous dDBT^∆ mutants (dDBT^∆/+) (Supplementary Fig. 11). This prompted an investigation into the impact of the dPP2Ac-AMPK axis on the fitness of adult dDBT^∆/+ mutants under varying BCAA levels. While reinforcing neuronal AMPK activity via the overexpression of inactive mts^H118N extended the shortened lifespan in adult dDBT^∆/+ mutants, further disruption of neuronal AMPK functionality via the overexpression of active mts^Y307F further reduced the already shortened lifespan (Fig. 6i). The detrimental effect of a high protein diet on lifespan could be significantly mitigated by neuronal overexpression of inactive mts^H118N in adult dDBT^∆/+ mutants (Fig. 6j). These results suggest that neuronal signaling via the AMPK-PP2Ac axis is crucial for rescuing the health defects associated with reduced BCAA catabolism. Overall, the decline of AMPK activity due to enhanced PP2Ac binding thus emerges as a principal cause of the neurological dysfunction associated with BCAA dysregulation.

Discussion

In this study, we investigated several unknown mechanisms underlying the neuropathies induced by the dysregulation of BCAA-catabolizing pathways. We first showed that disrupting BCAA homeostasis in Drosophila via the genetic deletion of evolutionarily conserved genes encoding for key BCAA-catabolizing enzymes was highly correlated with neurological dysfunction. Further investigation in clinically relevant scenarios in Drosophila showed that reduced AMPK signaling facilitated by an enhanced PP2Ac interaction and a thus dysfunctional AMPK-mitochondrial axis played a key role in facilitating neurological damages.

Similar to clinical genetic variants for MSUD⁵, the five generated Drosophila mutants used in this study also lacked a good genotype-phenotype correlation in terms of the dysfunctional BCAA catabolism leading to developmental, behavioral, and neuropathological deficits. This means that we currently still lack explanations for the genetic link to MSUD neuropathology. Regardless, these Drosophila mutants did shed light on the widespread decline in AMPK activity in brain tissues, thereby prompting us to question whether the identical phenotypes are caused by a deficiency in brain AMPK as a central and shared event that occurs downstream of the mutations.

As such, this study showed AMPK acts as a downstream cellular effector negatively regulated by disturbances to BCAA catabolism, indicating that dysfunctional BCAA degradation may control AMPK inactivation. The crosstalk between general control nonderepressible 2 (GCN2) and mechanistic target of rapamycin (mTOR) signaling was known to be important in integrating amino acid sensing and metabolic homeostasis⁵⁶, though the GCN2 route was recently shown not to be an essential one for nutritional amino acids⁵⁷. Before our study, little was known about the ability of AMPK to sense alterations in amino acid levels^29,33. Our study highlights this aspect, supported by the notion that AMPK activation is thought to respond to cysteine deprivation independently via both canonical AMP/ATP ratios and GCN2 pathways⁵⁸. Additionally, AMPK could antithetically control mTOR in the regulation of cellular metabolism²⁹. Altogether, these pieces of evidence indicate that AMPK is a vital energy sensor modulated by a broad spectrum of energy statuses, with changes in the BCAA-catabolizing pathway likely modulating its activity.

Demonstrating the causative role of AMPK signaling in neuropathologies, our reactivation of AMPK signaling in the dDBT^Δ larval brain had noted neurological benefits and repaired developmental defects. This neuronal effect to dDBT deficiency is backed by evidence indicating that the developmental defects resulting from a lack of DBT activity in C. elegans could be most improved by DBT overexpression in neurons⁵⁹. This underscores the notion that repairing neuronal AMPK activity in brains, whether through genetic or pharmaceutical means, could potentially treat mitochondrial–dysfunction-associated neurological injuries resulting from BCAA dysregulation.

However, more investigation into its use as a potential therapeutic target is necessary, as treatments that affect AMPK signaling have shown mixed results due to temporal and spatial effects. For instance, some cases of AMPK activation exerted neuroprotective effects in early phases of Huntington’s disease, but overactivated AMPK increased apoptotic neuronal death in the striatum of mice with CAG repeats in a more advanced disease model⁶⁰. This suggests that the downstream effect of AMPK may depend on the stimulus, the extent, and the duration of its activation. Another possibility is that short-term and long-term activation of AMPK may act on different downstream targets, thus resulting in different effects. Additionally, increasing brain AMPK activity was observed to shorten the Drosophila lifespan under starvation conditions²⁷ and AMPK activation was found to support tumor growth during energy stress⁶¹, indicating that AMPK exerts a variety of roles in organismal health in response to diverse environments or situations. Combined, we can state that adjusting the dynamic homeostasis of AMPK activation may benefit brain fitness in cases of BCAA dysregulation, but significantly more work needs to be done in terms of uncovering the mechanisms of AMPK activation and signaling in general.

During our investigation of cellular mechanisms in this study, we also uncovered that excess BCAAs led to mitochondrial-derived ROS stress in brains that diminished AMPK activity, resulting in neurological damage. This fits with known research, as mitochondrial dysfunction is widely accepted as a key driver of the pathogenesis of neurodegenerative diseases, including Alzheimer’s, Parkinson’s, and Huntington’s diseases⁶². Additionally, the presence of mitochondrial dysfunction was featured in patient fibroblasts⁶³ and Drosophila dDBT^∆ brain tissues¹⁶, while oxidative stress appears elevated in rat brains of MSUD model^64,65 and patient plasma⁶⁶. In this study, we also demonstrated an interplay between AMPK attenuation and mitochondrial malfunction in a dDBT^∆ brain. This suggests that attenuating AMPK signaling might be a common feature of mitochondrial dysfunction in neurological diseases, which makes sense, as AMPK has been identified as a central integrator of mitochondrial homeostasis and as important for mitochondrial health³⁰. AMPK activity also decreases in the brains of Alzheimer’s patients⁶⁷ and the brain of a mouse model of Parkinson’s disease⁶⁸, which may imply that an excess of BCAAs in these other neuropathologies may impact brain AMPK activity as a route to their pathogenesis, though this will need to be investigated further.

Addressing the ongoing controversy regarding the role of oxidative stress on AMPK function^36–38, we attribute the attenuation of AMPK activity to an intensified direct interaction with dPP2Ac, which is mediated by signals from the dysregulated BCAA-mROS stress axis in the brain. Although our study showed no alteration in dPP2Ac activity as determined by unaltered phosphorylation at Y307, there is evidence for other posttranslational modifications affecting PP2Ac, such as carboxymethylation⁶⁹ and various other phosphorylation sites^69,70. Therefore, it cannot be excluded that changes in dPP2Ac activity may occur via different modification sites. Furthermore, phosphorylation events at different sites on PP2Ac have been shown to impact its association with a regulatory B subunit, thereby dictating substrate selectivity^70,71. As a result, it may be speculated that in dDBT^∆ versus control scenario, dPP2Ac more intensively interacts with AMPK, though further exploration is necessary to validate how PP2Ac activity responds to the reduction in BCAA catabolism.

Our results thus also indicate that activating neuronal AMPK might be a potential strategy for MSUD treatment. Similar to age-related Alzheimer’s disease, MSUD lacks effective and curable treatments, with current options limited to dietary restrictions, hemodialysis, and liver transplantation, and even these are unsatisfactory or unattainable for many patients⁹. Therefore, new treatment strategies applicable to MSUD are urgently needed. Our results here suggest that AMPK signaling in the brain can be activated with an oral administration of metformin, and though metformin has been shown to cross the brain-blood-barrier in a mouse model after systemic administration⁷², the appropriate concentration for affecting the brain is still unknown. It is difficult to consider the potential benefits of metformin for human health when lactic acidosis and potential effects on cell death are caused by metformin-mediated inhibition of mitochondrial complex I⁷³ and evidence that Drosophila lifespan is reduced after metformin administration at high concentrations⁷⁴. Thus, studies into the distribution and dosage of metformin will be important for using it as a treatment for activating AMPK in the CNS.

In conclusion, this study contributes to the slow unraveling of the mysteries behind AMPK signaling by shedding light on its previously unclear impact on neuronal dysfunction in amino acid homeostasis. We, for the first time, demonstrated that PP2Ac-mediated attenuation of AMPK activity in the brain plays a primary role in sensing and signaling in neurological function via its regulation due to dysfunctional BCAA degradation and neuronal ROS stress derived from mitochondrial dysfunction (Fig. 7). Thus, our study suggests that dysregulated BCAA-AMPK signaling might have merit as a therapeutic strategy for a wide range of neurological diseases associated with BCAA dysregulation.

Fig. 7. Overview of BCAA catabolism deficits leading to neurological impairment via the mitochondrial dysfunction-AMPK axis.

Deficits in BCAA catabolism, such as those resulting from genetic loss-of-function in BCAT or the BCKDH complex, lead to BCAA accumulation and subsequent neuropathological effects. This accumulation induces neuronal apoptosis and inhibits autophagy, primarily by reducing neuronal AMPK activity. Mitochondria-derived ROS enhance the interaction between PP2Ac and AMPK, further dampening AMPK activity, exacerbating mitochondrial dysfunction, and contributing to neurological impairments.

Methods

Fly stocks

The following mutant fly lines targeting BCAA-catabolizing enzymes were generated using CRISPR/Cas9 genome editing and employed in this study: dBCAT^∆ (Branched-chain amino acid transaminase; CG1673), dBCKDHA^∆ (Branched chain keto acid dehydrogenase E1 subunit alpha; CG8199), dBCKDHB^∆ (Branched chain keto acid dehydrogenase E1 subunit beta; CG17691), and dDLD^∆/+ (Dihydrolipoamide dehydrogenase; CG7430). The previously described dDBT^Δ (Dihydrolipoamide branched chain transacylase E2; CG5599) mutant¹⁶ was also included in this study. To generate heterozygous mutant lines, the following genotypes were established: dBCAT^∆/Y, dBCKDHA^∆/TM6B, dBCKDHB^∆/CyO-GFP, dDBT^∆/Y, and dDLD^∆/TM6B. These lines were backcrossed into the genetic background of the wild-type strain w¹¹¹⁸ (BDSC#5905), yielding heterozygous mutants of dBCAT^∆/+, dBCKDHA^∆/+, dBCKDHB^∆/+, dDBT^∆/+ and dDLD^∆/+, and heterozygous female mutant flies from these lines were used in the experiments conducted in this study. gstD-GFP reporter⁷⁵ (gift from Dirk Bohmann) and the UAS-dDBT-HA flies (newly generated in this study) were also used in this study. elav-Gal4/Cyo (BDSC#8765), elav-Gal4 (BDSC#8760), tub-Gal4 (BDSC#5138), UAS-mCherry-AMPKα (BDSC#32109), UAS-AMPKα^K57A (BDSC#32112), UAS-EGFP (BDSC#5431), UAS-Sod2 (BDSC#24494), UAS-mts-RNAi (BDSC#27723) and pk¹ (prickle mutant; BDSC#367) were obtained from Bloomington Drosophila Stock Centers (BDSC). UAS-Atg1-RNAi (VDRC#16133) lines were obtained from Vienna Drosophila Resource Center (VDRC). For the autophagy immunostaining assay, the fly carrying UAS-mCherry-Atg8 (BDSC#37750) driven by fat body-specific r4-Gal4 was crossed into the genetic background of BACT^∆ or the other BCKDH^∆ mutants. Fly strains were reared on a standard medium and kept at 25°C.

Generation of mutant and transgenic flies

CRISPR-mediated mutagenesis was performed by WellGenetics Inc. following modified protocols from Kondo and Ueda⁷⁶. Briefly, gRNA sequences for each gene (see Supplementary information) were cloned into a U6 promoter plasmid. A donor repair template was constructed by cloning an attPX-RFP cassette—containing attPX, two STOP codons, and 3xP3 RFP—along with two 1-kb homology arms into a pBluescript SK(+) or pUC57-kan plasmid vector. Individual targeting gRNAs for the dBCAT, dBCKDHA, dBCKDHB, and dDLD genes, along with hs-Cas9, were delivered in separate DNA plasmids along with the donor plasmid for microinjection into embryos of the control strain w¹¹¹⁸. F1 flies carrying the 3xP3 RFP selection marker were validated via genomic PCR and sequencing. CRISPR mutagenesis generated mutant alleles of dBCKDHB and dDLD, which involved deletion of most of the coding sequence (CDS) region and replacement with the attPX-RFP cassette. Mutant alleles for dBCAT and dBCKDHA were generated by inserting the attPX-RFP cassette into the exon (see Supplementary Fig. 1b). These mutations were evaluated for on-target and off-target effects. Genomic DNA was extracted from a single fly from each mutant line, and the target site fragment was amplified via PCR (primers listed in Supplementary Information). The PCR products were sequenced using Sanger sequencing to confirm on-target mutagenesis. The assessment of off-target effects relied significantly on the genome database used for gRNA design. This study utilized the Drosophila melanogaster genome database (Drosophila melanogaster genome version 3) for gRNA design, and the CHOPCHOP web tool (http://chopchop.cbu.uib.no/) was employed for sgRNA design and off-target site prediction, confirming the absence of off-target effects (see Supplementary Information). The successful generation of mutants was further validated by RT-qPCR, which confirmed the absence of target gene transcripts (see Supplementary Fig. 1c). Functional assays using LC-MS demonstrated the accumulation of branched-chain amino acids (BCAAs) in the mutants compared to w¹¹¹⁸ control flies (see Fig. 1a). To generate UAS-dDBT-HA transgenic flies, molecular cloning was performed using the In-Fusion HD Cloning Kit (Clontech). The entire coding sequence of the dDBT gene (excluding the stop codon) was PCR-amplified and cloned into the pUAST-3xHA-attB vector using EcoRI and XhoI restriction sites. The resulting pUAST-dDBT-3xHA-attB plasmid was injected into w¹¹¹⁸ embryos to produce transgenic flies carrying the UAS-dDBT-HA construct. For generation of transgenic UAS-mts mutant fly, a mutant Drosophila mts gene, harboring either the Y307F (mts^Y307F) or H118N (mts^H118N) mutation, was generated through site-directed mutagenesis. These mutant genes, characterized by a 4 bp Drosophila Kozak Sequence (caaa) preceding the protein start codon, were fused with 90 bp DNA fragments containing a 3x FLAG tag at the C-terminal region. Subsequently, the mts^Y307F- or mts^H118N-3xFLAG constructs were cloned into the pUAST-attB vector at the XhoI site. The resulting constructs were then directed to integrate into the same attP2 landing site on chromosome 3 L (68A4) using PhiC31 integrase⁷⁷, facilitating uniformity across all transgenic insertions.

Circulating BCAA measurement

Based on LC-based metabolomics, we collected larvae hemolymph to detect circulating BCAA levels through LC-MS analysis, following a previously described methodology¹⁶. In brief, 5 μl hemolymph samples collected from each BCAA-catabolizing mutant were separately mixed with 45 μl of 50% methanol prior to vigorous mixing with 150 μl chloroform for 1 min. To separate the aqueous layer after centrifuging for 10 mins at 4 °C, the supernatant was diluted for LC-MS/MS (XeVO TQ-MS, WATERS) analysis using an ACQUITY UPLC BEH C18 column (1.7 μm × 2.1 mm × 100 mm). For adult Drosophila, hemolymph was collected from groups of eighty 30-day-old female flies (dBCAT^∆/+, dBCKDHA^∆/+, dDBT^∆/+, and w¹¹¹⁸) fed either a high-protein or control diet. A small puncture was made in the thorax, and hemolymph was extracted after centrifugation and then subjected to LC-MS analysis.

Preparation of high protein diet

The high protein diet was prepared following a previously described recipe⁷⁸. In brief, the control food (1 L) consisted of 10 g of agar, 80 g of brewer’s yeast, 20 g of yeast extract, 20 g of peptone, and 51 g of sucrose. For the high protein diet, the control food (1 L) was augmented by 150 g of soy protein and 66 g of Crisco.

RT-qPCR analysis and primers

Total RNA from whole larvae was extracted using the Trizol method⁷⁹. The extracted RNA samples were treated with DNase I and purified by phenol/chloroform-based extraction and ethanol precipitation. For the synthesis of cDNAs, reverse transcription was performed using SuperScript RT-III kit (Invitrogen) followed by a SYBR®Green-based quantitative RT-PCR analysis using an AB ViiA-7 Real-Time PCR system (Applied Biosystems). Ct values were used to calculate the relative expression. The primers used for the study are given in Supplementary Information.

Histological analysis

From each group, the heads of five 30-day-old flies were dissected out, placed into Bouin’s solution (Polysciences, Inc.), and fixed and rotated for five days at room temperature (RT). The methodology for the brain histological assay was described previously¹⁶, and the protocols presented in that study were followed here. The brain samples were handled using the standard protocols for histological methods⁸⁰, and the brain sections were stained for visualization using hematoxylin and eosin (H&E). Serial sections of each brain were taken to count the number of vacuoles and calculate the average number of total lesions in each group.

Developmental examination

Male w¹¹¹⁸ and male BCAA-catabolizing mutants of the same age were used in the developmental assay. The developmental examinations of pupation and eclosion rates were described in our previous study¹⁶. To determine the pupation rate, 1^st instar larvae were picked and reared in density-controlled vials. The pupation rate was calculated as the ratio of pupae to the original number of larvae. The eclosion rate was calculated as the ratio of the number of adults to the original number of pupae.

Crawling behavior assay

Detection and analysis of crawling behavior in Drosophila larvae followed previous descriptions¹⁶. In brief, six genotype-tested larvae at the early 3^rd-instar stage were selected and placed on an agar plate (200 mm × 115 mm × 30 mm) containing a solution of 1% agar, 0.1 M sucrose, and brilliant blue dye. Kept in a constant light, temperature, and humidity, larvae were then allowed to move for four minutes and recorded using a video camera. Following video collection, the middle two minutes of each video were analyzed for crawling movements and the speed and length of movement using the ImageJ Plugin wrMTrck (ImageJ: http://imagej.nih.gov/ij/; http://www.phage.dk/plugins/wrmtrck.html). At least three biological replicates were made for each genotype.

Locomotion assay

The locomotion of adult flies was measured via a climbing assay performed using a counter-current apparatus equipped with six chambers, as described previously⁸¹. 30-day-old dBCAT^∆/+, dBCKDHA^∆/+, dDBT^∆/+, and w¹¹¹⁸ female flies were put into the climbing tubes for the assay. The climbing index was calculated from 150 flies per genotype tested using a weighted average and the standard formula described previously⁸¹. These tests were all performed in the daytime to prevent interference.

Lifespan assay

A group of about 35 female flies from the lines w¹¹¹⁸, dBCAT^∆/+, dBCKDHA^∆/+, dDBT^∆/+ or dDBT^∆/+; elav/+ carrying neuronal overexpression of mts^Y307F, mts^H118N, mCherry-AMPKα^WT, or AMPKα^KA were placed in a plastic tube containing normal or high protein food. The flies were transferred to new vials daily without anesthesia, and the number of dead flies was recorded. Each group was assessed in triplicate, and statistical comparisons were based on median survival. In total, 100 flies per group were monitored throughout the lifespan assay.

Adult heat-shock assay

To assess seizure threshold and duration, adult flies were subjected to a heat-shock assay⁸². Female flies aged 10 or 20 days were anesthetized with CO₂ and transferred into empty plastic vials, with 10 flies per vial. Following anesthesia, the flies were allowed to recover at room temperature for 1 h before the assay. For the heat-shock assay, the vials with flies were submerged in a water bath set to a constant temperature of 41 ± 0.5 °C. The flies were visually inspected every 15 s, and the number of paralyzed flies was recorded. Paralysis was identified as a loss of posture and the inability to move during heat assay. Observations were repeated at 15-second intervals until all flies were paralyzed, and seizure status was noted throughout the experiment.

Staining and imaging

For immunostaining, dissected larval brains or fat bodies were fixed in 4% paraformaldehyde for 25 mins prior to washing with 0.5% PBST (0.5% Triton X-100 in PBS buffer) three times. The dissected tissues were then treated with blocking buffer (2% BSA with 0.5% PBST) for 30 mins at room temperature (RT). After washing three times with 0.5% PBST, the dissected tissues were incubated with primary antibodies overnight at 4°C. The following antibodies were used: anti-Atg8 (1:200) for Atg8 staining; anti-mCherry (1:300) for mCherry-Atg8a staining used in the fat bodies of mCh-Atg8a-expressed larvae; anti-ELAV (1:50) for labeling the differentiated neurons; anti-cleaved-caspase3 (1:200) for indicating cell apoptosis; anti-4HNE (1:300) for monitoring oxidative damage; and anti-GFP (1:200) for the gstD-GFP reporter assay. Subsequently, 0.5% PBST-washed brains were incubated with a secondary antibody (1:400) of goat anti-mouse (Abcam) or goat anti-rabbit conjugated with Alexa 488 or Alexa 647. Adult brains for apoptosis staining with the anti-cleaved-caspase3 antibody were processed according to a previously published protocol⁸³ (for Supplementary Fig. 2c). For staining with chemical dyes, larvae brain tissues were dissected in Schneider’s medium (Sigma-Aldrich, S0146) and incubated with chemical dye added into Schneider’s medium. For the lysotracker assay, the dissected brains were fixed in 4% paraformaldehyde for 5 mins, and the brain tissues were stained with Lysotracker red DND-99 (1:1000, Molecular Probes, L7528) for 20 mins. The larval fat body tissues were dissected out from the third instar larvae of w¹¹¹⁸ or BCAA-catabolizing enzyme mutants after regular feeding or starvation for 4 h. The tissues were immediately stained with Lysotracker red DND-99 for 10 min and DAPI co-staining for 30 min. In other staining for related experiments, dissected brains were freshly stained with a final concentration of 30 μM DHE (Thermo Fisher Scientific, D11347) for 10 mins for the preferential detection of superoxide and other ROS, with JC1 (1:1000, Molecular Probes, T3186) for 30 mins for the mitochondrial membrane potential assay, or to co-stain with a final concentration of 5 µM of MitoSOX (Molecular Probes, M36008) and 1 µM MitoTracker (Molecular Probes, M7514) for 30 mins for detecting mitochondrial ROS. Brain samples were washed three times with cold PBS, were mounted with mounting media (Vector Laboratories, H-1500), and immediately imaged by Leica SP5 confocal microscopy.

Image quantification

Fluorescence intensity analyses were performed using ImageJ software. Average intensity was calculated by measuring intensity values in at least three equal-sized boxes in the central brain and ventral nerve cord region (excluding peripheral nerve fibers) of larval brains. Average intensities of at least 3 larval brains were combined to calculate the final average intensity for each experiment. Data collected from five independent experiments were plotted in the graphs using GraphPad Prism 9.0 software.

Western blot analysis

Dissected whole brain or ex vivo-treated whole brain tissues were dissected out from w¹¹¹⁸, BCAA-catabolizing mutants, neuronal Sod2-overexpresed, mts–variant-overexpressed, or mts-depleted dDBT^∆ larvae administrated without or with leucine or metformin. The brain tissues were collected and extracted for the preparation of whole cell lysates. The sample lysates were separated by SDS-PAGE and the gel was then transferred to a PVDF membrane, probed with anti-Atg8 (1:5000), rabbit antiphospho-S6K(Thr398) (1:1000), anti-Ref2p (1:2000), anti-phosphor-AMPKα (1:2000), anti-AMPKα (1:500), anti-mCherry (1:5000), anti-PP2Ac (1:5000), anit-phospho-PP2Ac (1:1000), anti-HA (1:5000), anti-GFP (1:2000), anti-Total OXPHOS Rodent WB Antibody Cocktail (1:2500), anti-actin (Developmental Studies Hybridoma Bank [DSHB], 1:50), or anti-actin (GeneTex, 1:5000) primary antibody overnight at 4°C, and incubated with the secondary antibody of anti-rabbit or anti-mouse IgG horseradish peroxidase (Jackson ImmunoResearch, 1:10000) for 1 h at RT. Immunoblot signals were developed by enhanced chemiluminescence (ECL solution, Millipore, WBKLS0500) followed by exposure to X-ray film.

Mitochondria Complex Analysis

Mitochondria were isolated from thirty-third instar larvae using the Mitochondria Isolation Kit for Tissue (Thermo Fisher Scientific, Waltham, MA, USA). The isolated mitochondria were then lysed in RIPA buffer and centrifuged at 13,200 rpm for 10 min at 4 °C. The resulting supernatant was collected and mixed with the sample buffer. The samples were subsequently boiled for 10 min and subjected to western blot analysis. Mitochondrial complex expression was detected using the Total OXPHOS Rodent WB Antibody Cocktail.

Co-immunoprecipitation assay

For the co-immunoprecipitation assay, approximately 500 larval whole brains or 100 fly heads were dissected from dDBT^∆ larvae or dDBT^∆/+ female adult flies expressing neuronal mCherry-AMPKα^WT in each experimental group. Subsequently, larvae were either fed 50 mM leucine or subjected to genetic neuronal SOD2 overexpression. Brain tissues were dissected and lysed using RIPA buffer supplemented with protease inhibitor (Roche, #04693132001) and phosphatase inhibitor (Roche, #04906837001). A total of 500 μg protein lysates were incubated overnight at 4 °C with the primary antibodies anti-mCherry or anti-GFP (1:200). Magnetic beads (Protein A Mag Sepharose, Cytiva, 28951378) at a ratio of 1:25 were then added to the lysates for a 2 h incubation at 4 °C to precipitate the protein complexes. The beads were washed four times with 500 μl lysis buffer before adding sampling dye and boiling at 95 °C for 10 min. The boiled samples were used for immunoblot analysis. PVDF membranes were probed with anti-mCherry and anti-PP2Ac antibodies followed by an HRP-conjugated secondary antibody, and the signal was subsequently detected.

Primary antibodies

The following antibodies were used for immunoblotting or immunoprecipitation assay: rabbit anti-Atg8 (Sigma-Aldrich, ABC974-I), rabbit anti-Ref2p (Abcam, ab178440), rabbit antiphospho-S6K(Thr398) (Cell Signaling, #9209), rabbit anti-phosphor-AMPKα (Cell Signaling, #2535), mouse anti-AMPKα (Millipore, MABS1232), rabbit anti-PP2Ac (Cell Signaling, #2038S), rabbit antiphospho-PP2Ac (Merck, SAB4503975), rabbit anti-Actin (GeneTex, GTX109639), mouse anti-Actin (DSHB, anti-JLA20), rabbit anti-mCherry (Abcam, ab213511), mouse anti-HA (Cell Signaling, #3724), mouse anti-Flag (Merck, F1804), rabbit anti-GFP (GeneTex, GTX113617). The following antibodies were used for immunostaining: rabbit anti-cleaved-caspase3 (Cell Signaling, #9661), rabbit anti-4HNE, (Abcam, ab46545), mouse anti-Total OXPHOS Rodent WB Antibody Cocktail (Abcam, ab110413), mouse anti-Elav (DSHB, 9F8A9) and rabbit anti-GFP (GeneTex, GTX113617).

Lipid peroxidation assay

The malondialdehyde (MDA) assay was described previously¹⁶. In brief, larval brain tissues were dissected in PBS with 0.05% butylated hydroxytoluene (BHT) prior to homogenization and an OxiSelect™ TBARS Assay Kit-MDA Quantitation kit (Cell Biolabs Inc., STA-330) was used to determine the concentration of MDA. Spectrometric absorbance was read at 532 nm. Triplicate experiments were performed. For C11-BODIPY 581/591 staining¹⁶, the brains of 3^rd-instar larvae were incubated in Schneider’s medium containing 10% fetal bovine serum and 2 μM C11-BODIPY 581/591 (Invitrogen, D3861), a fluorescent probe used to highlight lipid peroxidation⁸⁴, for 30 min before confocal microscopy. An image was captured from the non-oxidized (595 nm) and oxidized (520 nm) spectra.

ATP assay

Fifty larval brains per group per experiment were collected and used to measure mitochondrial ATP levels with an ATP Bioluminescence Assay Kit CLS II (Roche, cat.no.11699695001). The collected brain tissues were homogenized in cell lysis reagent (100 mM Tri and 4 mM EDTA, pH 7.75) and incubated for 5 min at 100 °C. A measurement of ATP level was then conducted following the instructions, and the results were compared to standards. The relative ATP level was calculated by dividing the luminescence by the total protein concentration, which was determined by protein assay dye reagent (Bio-Rad, #5000006).

Chemical administration

For oral chemical administration, L-leucine (Sigma-Aldrich, I8000), rapamycin (Sigma Aldrich, CAS# 53123-88-9), NAC (Sigma-Aldrich, A7250), and metformin (Sigma-Aldrich, D150959) were prepared separately or together to the indicated concentration and were added into the regular food of Drosophila. The 1st instar larvae were picked and reared in the chemical-containing food until reaching the stage of 3^rd instar larvae. Subsequently, the larvae or dissected brains were used to perform the following related experiments.

Ex vivo treatment

Thirty larval brains were dissected in Schneider’s medium. The dissected brain tissues were incubated with 2 mM leucine at 25°C for the indicated time point, or treated with 10 μM H₂O₂ (Honeywell, 31642) for 30 s, or 10 nM okadaic acid (Cell Signaling, #5934) for 10 and 30 mins. After washing with PBS, the brain samples were used for western blot analysis.

Statistics and reproducibility

Samples were collected randomly for all experiments. The sample sizes varied depending on the specific requirements of each experiment, and the exact sizes were provided in the figure legends. Results were analyzed and presented using GraphPad Prism 9.0. For eclosion rate experiments, post-hoc pairwise chi-squared tests with a Bonferroni correction were performed in R to compare the eclosion rates of WT and individual mutant strains. For other experiments, two-tailed unpaired Student’s t-tests were used for statistical comparisons. Data are presented as mean ± SEM. Statistical significance was set at p < 0.05.

Author contributions

Shih-Cheng Wu: Conceptualization, Methodology, Validation, Formal Analysis, Investigation, Resources, Data Curation, Writing—Original Draft, Writing—Review & Editing, Supervision, Project Administration, Funding Acquisition. Yan-Jhen Chen: Methodology, Validation, Formal Analysis, Investigation. Shih-Han Su: Methodology, Validation, Investigation. Pai-Hsiang Fang: Methodology, Investigation. Rei-Wen Liu: Methodology, Investigation. Hui-Ying Tsai: Methodology. Yen-Jui Chang: Methodology. Hsing-Han Li: Methodology. Jian-Chiuan Li: Methodology. Chun-Hong Chen: Conceptualization, Validation, Formal Analysis, Resources, Data Curation, Writing—Original Draft, Writing—Review & Editing, Supervision, Project Administration, Funding Acquisition.

Peer review

Peer review information

Communications Biology thanks Joseph Bateman and the other, anonymous, reviewers for their contribution to the peer review of this work. Primary Handling Editors: Rosie Bunton-Stasyshyn & Joao de Sousa Valente. A peer review file is available.

Data availability

All data supporting the findings of this study are included in the main text and Supplementary Materials. The accompanying Supplementary Data file provides source data for generating the plots in the main figures. Original uncropped western blot images are included in the Supplementary Information. Further details and reagents are available from the corresponding author upon reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-025-07457-6.