Suppression of intestinal dysfunction in a Drosophila model of Parkinson’s disease is neuroprotective

The innate immune response mounts a defense against foreign invaders and declines with age. An inappropriate induction of this response can cause diseases. Previous studies showed that mitochondria can be repurposed to promote inflammatory signaling. Damaged mitochondria can also trigger inflammation and promote diseases. Mutations in pink1, a gene required for mitochondrial health, cause Parkinson’s disease, and Drosophila melanogaster pink1 mutants accumulate damaged mitochondria. Here, we show that defective mitochondria in pink1 mutants activate Relish targets and demonstrate that inflammatory signaling causes age-dependent intestinal dysfunction in pink1-mutant flies. These effects result in the death of intestinal cells, metabolic reprogramming and neurotoxicity. We found that Relish signaling is activated downstream of a pathway stimulated by cytosolic DNA. Suppression of Relish in the intestinal midgut of pink1-mutant flies restores mitochondrial function and is neuroprotective. We thus conclude that gut–brain communication modulates neurotoxicity in a fly model of Parkinson’s disease through a mechanism involving mitochondrial dysfunction. The authors show that pink1-mutant flies display intestinal dysfunction and that suppression of Relish, an innate immune response mediator, in the gut rescues mitochondrial dysfunction and cell death in the brain.

A nimals use the innate immune system as a defense against foreign invaders, which allows a rapid reaction to invading pathogens such as microbes. Microbes often release foreign nucleic acids into the cytosol of infected cells, which can trigger innate immunity.
Mitochondria are energy-generating organelles that evolved from endosymbionts related to bacteria. Mitochondria retain some of their genetic material in the form of mitochondrial DNA, which can trigger inflammatory responses similarly to bacterial DNA, and these responses are associated with the absence of methylated CpG sequences in both bacterial DNA and mtDNA 1 .
Cells safeguard the health of their mitochondria by operating several quality control (QC) mechanisms that ensure the disposal of faulty mitochondria. The QC of organelles involves the degradation of defective mitochondria via mitophagy, which is a form of autophagy 2 . Mutations in the gene encoding the mitochondrial kinase PINK1 lead to the accumulation of defective mitochondria and cause a form of familial Parkinson's disease (PD), an age-related neurodegenerative disease 2 . PINK1 functions in a molecular pathway that ensures the degradation of faulty mitochondria via mitophagy. In mice, blockage of mitophagy induces the escape of mtDNA into the cytosol where it activates innate immunity 3 .
Constitutive activation of innate immunity in the fruit fly (D. melanogaster) decreases lifespan and increases neurodegeneration 4 . The innate immune system in Drosophila consists of two branches 5 : the Toll signaling pathway and the immune deficiency (Imd) pathway. The Imd pathway activates the Rel/NF-κB transcription factor Relish, which controls the expression of several antimicrobial peptides and is indispensable for normal immunity in flies. Relish rewires the metabolism by attenuation of FOXO-mediated lipolysis 6 and can also activate programmed cell death 7 .
The main immune organs in Drosophila are the fat body, which is considered to be equivalent to both vertebrate adipocytes and liver, and the intestine. The intestine, in particular, plays a key role in activation of the Imd pathway 8 .
Mounting an immune response is energetically costly and requires tradeoffs with other important biological functions 9 and, in flies, this tradeoff is important during starvation. In such settings, Relish mutants exhibit increased survival under food deprivation 10 . The activation of Relish in response to cytosolic DNA can be mediated by interactions with Drosophila Eya, which is a molecule that maintains Relish in the cytoplasm 11 .
Here we found that, in pink1-mutant flies in which mitophagy is blocked, Relish was activated with neurotoxic consequences. The accumulation of defective mitochondria in these mutants led to increased Relish signaling, caused intestinal dysfunction and resulted in cell death in the midgut. Enhanced Relish signaling also resulted in metabolic alterations in pink1-mutant flies. These alterations are characterized by the accumulation of triglycerides (TAGs) due to the Relish-dependent regulation of lipid catabolism 6 and failure of beta oxidation 12 . Genetic suppression of Relish or Eya, a Relish-binding protein involved in the sensing of cytosolic DNA, suppressed neurodegeneration in pink1-mutant flies. To understand the links between intestinal dysfunction and neurodegeneration, we subsequently investigated whether intestinal dysfunction could prime neurodegeneration and found that the prevention of intestinal dysfunction through either the suppression of Relish expression or the blockage of cell death in the midgut of pink1-mutant flies is sufficient to suppress the central nervous system (CNS) phenotypes of these flies. We conclude that CNS defects in pink1-mutant flies are modulated by a non-cell autonomous signaling pathway induced by mitochondrial toxicity acting between the intestine and brain.

results
Identification of a Relish signature in pink1 mutants. We previously showed that mitochondrial defects in pink1-mutant flies led to the upregulation of both nucleotide metabolism and immune response genes 12 . Here, we first explored the mechanism underlying activation of the innate immune response pathways associated with mitochondrial dysfunction in pink1-mutant flies. We analyzed innate immunity-related transcripts and proteins in pink1mutant flies through an in silico approach (Fig. 1a) and detected 45 upregulated transcripts that matched a curated list of innate immunity-related genes in flies ( Fig. 1b and Supplementary Table  1). Because pink1-mutant flies exhibit a global shutdown of protein synthesis 13 , we also measured the levels of individual proteins in adult flies through quantitative proteomics. We detected upregulation of nine proteins belonging to the innate immunity pathways in pink1-mutant flies ( Fig. 1b and Supplementary Tables 2 and 3). Transcriptional control of the innate immune response in Drosophila is mediated by signaling cascades that regulate the NF-κB-like transcription factors Dif, Dorsal and Relish (Rel) 14 . To identify the upstream regulators of the innate immunity signature present in pink1-mutant flies we utilized iRegulon, a tool used for the reverse engineering of transcriptional networks 15 . The analysis of 45 innate immunity transcripts with iRegulon matched the majority of these (27) to Rel as the top upstream regulator of the innate immunity signature in pink1-mutant flies (Fig. 1b,

A Relish mutation suppresses neuronal defects in pink1-mutants.
In flies, mutations in pink1 affect neuronal function, which leads to disruption of circadian rhythms in young adults 16 by preventing the secretion of neuropeptides 17 , and the selective loss of dopaminergic (DA) neurons in the protocerebral posterior lateral 1 (PPL1) cluster in aged flies 18 . We subsequently tested whether mutation of the Relish gene in Rel E20 /+ flies 19 would affect the neuronal phenotype of pink1-mutant flies. Specifically, we monitored locomotor activity under a light/dark cycle (LD) for 7 days and confirmed that pink1-mutant flies show aberrant activity patterns (Fig. 2a). As previously reported 20 , pink1-mutant flies exhibited a significantly longer rest (or inactivity) duration, which was correlated with lower activity levels. Interestingly, comparison of pink1-mutant flies with pink1, RelE20/+ double-mutant flies revealed that the Relish mutation led to a significant rescue of the sleep-wake patterns observed in pink1-mutant flies compared to those of control flies (Fig. 2a, red and blue). Additionally, the presence of a Relish mutation in pink1-mutant flies was sufficient to rescue the loss of DA neurons (Fig. 2b,c).
The activity of Relish during infection is negatively regulated by calcineurins and protein phosphatases. Calcineurins can be targeted by immunosuppressants, such as tacrolimus, which is used in human organ transplantation to lower the risk of rejection. Calcineurins are present in Drosophila 21 , and we thus tested whether pharmacological intervention with tacrolimus could suppress both locomotor activity deficits and DA neurodegeneration in pink1mutant flies. The exposure of these mutants to a diet supplemented with tacrolimus significantly improved sleep impairment (Fig. 2d) and prevented the selective loss of DA neurons in the PPL1 cluster (Fig. 2e). Previously it has been shown that feeding pink1-mutant flies with rapamycin, an inhibitor of the target of rapamycin pathway, reduces the phosphorylation of 4E-BP1 and is neuroprotective 22 . Because tacrolimus is a rapamycin derivative (rapalog), we next measured the degree of 4E-BP1 phosphorylation in pink1mutant flies maintained on a diet supplemented with tacrolimus. We found that while pink1-mutant flies evidenced a decrease in nonphosphorylated 4E-BP1 as previously shown 22 , these were not altered in tacrolimus-fed flies when compared to those maintained on a normal diet (Fig. 2f). Collectively, these findings show that suppression of innate immunity pathways by manipulation of Relish signaling is neuroprotective in a model of pink1 mutation-induced mitochondrial dysfunction.
Intestinal dysfunction in pink1-mutant flies. Relish can induce immune responses to pathogens by induction of the transcription of antimicrobial peptides such as Attacin (Att) 23 . We subsequently monitored the activation of Relish by expressing a Rel/NF-κB reporter (Att-GFP) 24,25 and detected green fluorescent protein (GFP) fluorescence in several regions of pink1-mutant flies, including the abdominal region (Fig. 3a). A more detailed analysis of the gut showed an increase in GFP expression in the gut enterocytes of pink1-mutant flies (Fig. 3b), which indicated the activation of Relish in the gut. Relish activation involves cleavage of an inhibitory domain on the full-length protein and translocation of the cleaved domain (Rel-68) to the cell nucleus 5 . Decrease in cytosolic Relish correlates with its nuclear translocation 26 . We next monitored the levels of cytosolic Relish and found lower levels of the full-length protein (Fig. 3c) and increased levels of the transcript (Fig. 3d) in pink1-mutant flies. Taken together, these results suggest that mitochondrial dysfunction in pink1-mutant flies is associated with Relish activation.
Increase in the expression of immunity-related transcripts is closely linked to intestinal barrier dysfunction in Drosophila 27 . We subsequently performed a Smurf assay, which is a noninvasive method for determination of the intestinal integrity in adult flies. The Smurf assay showed that pink1-mutant flies exhibit a compromised intestinal barrier that can partially be rescued by a mutation in Relish (Fig. 3e).
In Drosophila, the fat body, which is a lipid storage reservoir, and the gut form the primary immune organs. To visualize the gut and fat body in pink1-mutant flies, we labeled these tissues with phalloidin (to mark the gut) and BODIPY, a fluorescent lipophilic dye that stains neutral lipids. Confocal microscopy analysis of pink1-mutant flies showed a ring of visceral fat (fat 'doughnut') surrounding the posterior end of the posterior midgut (anterior of the Malpighian tubule junction) (Fig. 4a). Fat body cells migrate to sites of tissue damage where they clear wound-related cell debris and upregulate antimicrobial peptides (AMPs) 24 . Given their role in cell repair and renewal, we subsequently assessed whether the recruitment of fat body cells to the midgut is associated with damage to the midgut. Mitochondrial dysfunction in pink1 mutants causes apoptosis, a form of cell death associated with fragmentation of the mitochondrial network 28 . Confocal analysis of mitochondria labeled with a fluorescent tag (mito-GFP) showed that gut enterocytes in pink1 mutants have decreased mitochondrial length (Fig. 4b,c). Therefore, we next assessed the degree of apoptotic cell death in the midgut of pink1-mutant flies and found increased levels of active Drosophila caspase Dcp-1, an apoptosis effector (Fig. 4d,f), which is partially suppressed in pink1, Rel E20 /+ double-mutant flies (Fig. 4f). This indicated that pink1-mutant flies exhibit increased intestinal damage. Damage to the intestine also results in a robust proliferative response by intestinal stem cells (ISCs) 29 . We also monitored the levels of ISC proliferation using a GFP reporter for Escargot (Esg), which is expressed in these cells 30 , and found an increase in the number of Esg-positive cells in pink1-mutant flies (Fig. 4e,g). Taken together, these data show that mitochondrial dysfunction in pink1mutant flies is associated with intestinal damage.
An eya mutation suppresses neuronal defects in pink1 mutants. In Drosophila, lipid droplets act as mediators of host-pathogen interactions 31 and thus their recruitment around the gut in pink1mutant flies (Fig. 4a) Fold change ≥2 Transcripts upregulated by more than twofold were matched to a list of all genes potentially involved in responses to microbial infection, which are available from the Lemaitre group (lemaitrelab.epfl.ch). After annotation, these genes were matched to the corresponding proteins through global quantitative proteomics analysis. The transcripts were also matched to upstream regulators using the iRegulon algorithm in Cytoscape. b, Relish targets in pink1-mutant flies. Transcripts that are potentially involved in the Drosophila immune response, as curated by the Lemaitre group, are shown. The iRegulon algorithm predicted that the majority (60%) of annotated transcripts are regulated by Relish. The detected proteins matched to each of the filtered transcripts are also shown. Transcripts and proteins upregulated by at least twofold are shown in red. ND, not detected; NS, not significant. c, Network visualization of Relish (Rel) targets (red) identified by iRegulon. Of a total of 45 query nodes, 18 were not identified as Relish targets (gray). This figure is related to Supplementary Tables 1-4. stress activates the innate immune response 32 . Because mitochondrial stress can trigger the release of mitochondrial DNA into the cytosol, we assessed the presence of extramitochondrial DNA in intestinal enterocytes of pink1-mutant flies. A confocal analysis of cells labeled with mito-GFP and stained with anti-DNA showed increases in extramitochondrial DNA levels in pink1-mutant flies (Fig. 5a,b).
In Drosophila, the Eya protein acts in a cascade that senses undigested cytosolic DNA and activates the immune response. Eya binds to the inhibitor of nuclear factor kappa-B kinase subunit beta (IKKβ) 11 and full-length Relish 11,33 (Fig. 5c). Because neuronal dysfunction in pink1-mutant flies is suppressed by a Relish mutation, we subsequently tested whether eya can act upstream of Relish in response to naked mtDNA. We noted that the messenger RNA levels of Relish target genes were decreased and that extramitochondrial DNA levels were unaltered in pink1 and eya double-mutant flies ( Fig. 5d and 5b, respectively). The eya mutation also rescued the defects in activity observed in pink1-mutant flies (Fig. 5e, red and blue) and the loss of DA neurons (Fig. 5f). We conclude that the neurotoxic consequences of the activation of the immune response in pink1-mutant flies are linked to the presence of extramitochondrial DNA in the cytosol.
Relish signaling induces a starvation signature in pink1 mutants. Drosophila pink1 mutants exhibit an impaired circadian clock 16 , and here we found that these flies also present with defects in their gastrointestinal tract. Takeout is a Drosophila clock-controlled hormone that is primarily involved in feeding behavior 34 and is strongly expressed in the crop, which results in dilatation of the esophagus that serves as a food reservoir (Fig. 6a), the fat body and the antennae of male flies. Takeout expression is also induced upon starvation 34 . While no differences in food intake between control and pink1-mutant flies (Fig. 6b) were found, we detected a significant increase in the level of Takeout in pink1-mutant flies ( Fig. 6c and Supplementary Table 2). Furthermore, analysis of Takeout transcript levels showed that this circadian output hormone was upregulated in pink1-mutant flies and that this upregulation was partially suppressed in pink1, Rel E20 /+ double-mutant flies (Fig. 6d). The Drosophila genome encodes eight different insulin-like peptides (DILPs), with four of these (DILP1, DILP2, DILP3 and DILP5) functionally similar to human insulin and produced by neuronal insulin-producing cells (IPCs) in the fly brain (Fig. 6a, blue). Starvation, or low nutrient levels, will decrease the release of DILPs by IPCs 35 . The increase in Takeout, a starvation marker, together with the report that another clock-regulated hormone, DILP2, is arrested in the cell bodies of IPCs in pink1-mutant flies 17 , led us to investigate whether these flies suffer from disruptions in metabolic and energy homeostasis. First we focused on insulin, which is known to regulate dietary metabolism 36 . A confocal analysis of an epitope-tagged DILP2 (ref. 37 ) showed that pink1-mutant flies have higher levels of DILP2 in IPCs (Fig. 6e), which confirmed previous observations 17 . We then measured the total content of DILP2 peptide through an ELISA-based assay 37 and found that the levels of DILP2 in aged pink1-mutant flies were lower than those in controls (Fig. 6f). Systemic repression of insulin signaling leads to a FOXOdependent activation of NFκB-Relish signaling, which results in the induction of TAG metabolism upon starvation 6 . Fasting induces the mobilization of stored TAGs from lipid droplets, which are used by other tissues and organs as a source of energy. We subsequently measured TAGs in control and pink1mutant flies and found significant accumulation in both young and aged pink1B9 flies (Fig. 6g) while Relish mutation restored TAG levels in old flies to normal levels ( Fig. 6h), in accordance with previous findings 6,38 , but did not restore DILP2 levels (Fig. 6i). Taken together, these results link intestinal dysfunction in pink1-mutant flies to an imbalance in hormones related to feeding and changes in energy storage in adults.

Suppression of intestinal toxicity is neuroprotective.
We previously showed that degeneration of the indirect flight muscles of pink1-mutant flies occurs in a non-cell-autonomous manner through signaling between neurons and muscle cells 12 . Recent findings in mouse models of PD have shown that α-synuclein pathology occurs via interorgan communication between the gastrointestinal tract and brain 39 . Like those of mice, the CNS and gut of flies are interconnected. The Drosophila brain is connected to the midgut by neurons that produce insulin-like peptide 7 (ILP7) (Fig. 6a). The cell bodies of these neurons are in the abdominal ganglion, and these cells innervate the mid-/hindgut junction and rectal ampulla 40 .
To test whether gut-brain communication can cause neurotoxicity in pink1-mutant flies, we first suppressed Relish expression in the gut by RNA interference (RNAi) using the NP3084 gut-specific 41,42 Gal4 driver. This downregulation of Relish in the gut decreased overall levels of TAGs in pink1-mutant flies (Fig. 7a) and increased the level of fatty acid oxidation (Fig. 7b). However, suppression of Relish in the gut failed to block the increase in cytosolic DNA in gut enterocytes (Fig. 7c). Nevertheless, Relish downregulation improved mitochondrial function in the brains of pink1-mutant flies (Fig. 7d), suppressed inactivity defects (Fig. 7e) and rescued the loss of DA neurons (Fig. 7f). We next used another independent Gal4 driver, NP1 (ref. 42 ), to confirm that suppression of Relish in the gut blocks neurodegeneration (Fig. 7f), as well as the specificity of the NP3084 driver (Fig. 7g). To eliminate the possibility that Relish expression causes neurotoxicity through a cell-intrinsic mechanism, we also suppressed its expression directly in DA neurons using the tyrosine hydroxylase (TH) driver. This failed to suppress the loss of DA neurons in pink1-mutant flies (Fig. 7f). We reason that the  neurotoxic consequences of the activation of the immune response in pink1-mutant flies are linked to a gut-brain communication mechanism.
We subsequently tested whether blockage of cell death in the intestine of pink1-mutant flies is sufficient for the suppression of neurotoxicity. We targeted the expression of either Buffy, a Drosophila Bcl-2-like protein with antiapoptotic activity 43 , or reexpressed Pink1 in the gut of pink1-mutant flies using the NP3084 driver. We found that expression of either Buffy or Pink1 decreased the number of DCP-1-positive cells (Fig. 8a,b) and prevented loss of DA neurons in pink1-mutant flies (Fig. 8c). Next, we tested whether suppression of pink1 expression in the gut was sufficient to cause degeneration of neurons with healthy mitochondria. RNAimediated suppression of pink1 using the NP3084 driver decreased  levels of the pink1 transcript in the intestine but failed to induce loss of PPL1 neurons (Fig. 8d,e). We thus conclude that suppression of intestinal dysfunction in pink1-mutant flies is sufficient to rescue neurotoxicity.

Discussion
Mitochondria tune their performance to adjust their energy output to the requirements of individual cells. To achieve this tuning, mitochondria communicate with the cell nucleus using the retrograde response 44 , which modifies the flow of information in cells by altering the transcriptional control of cellular functions. Here, we found that Relish signaling serves as a retrograde response pathway that is activated by mitochondria and has neurotoxic consequences. Relish, the Drosophila ortholog of NF-κB in mammals, controls the expression of several immunity genes. Recent studies in humans and mice have shown that PINK1 plays a role in restraint of innate immunity and that inflammation plays a positive role in PD 45 . Here we extend these observations and demonstrate that, in flies, this inflammatory response is triggered in the intestine and communicated to the CNS. These conclusions build on the increasing body  of evidence obtained from models of α-synuclein pathology 39 that show a role for the gut-brain axis in the etiology of PD.
We detected increased levels of enterocyte cell death in the intestines of pink1-mutant flies (Fig. 4). In Drosophila, Relish drives the removal of cells that are perceived as being unfit via apoptotic cell death 46 , and we reason that the release of DNA from defective mitochondria in pink1-mutant flies labels cells as unfit and prone to Relish-mediated apoptosis. It is also possible that a defective intestinal barrier in pink1-mutant flies allows the invasion of gut bacteria in the intestine and activates an inflammatory response.
In Pink1-knockout mice, mitochondrial dysfunction induces an inflammatory response via the cGAS-STING pathway that senses cytosolic DNA and acts upstream of the Relish ortholog NF-κB 45 . In Drosophila, dmSTING functions downstream of Relish to control viral infection 47 , and a previous study showed that its loss of function does not rescue the pink1-mutant phenotype 48 . Our data (Fig. 5) suggest that eya can act upstream of Relish to sense cytosolic DNA released by defective mitochondria in pink1-mutant flies.
Taken together, data from both mice 45 and flies (our study) suggest that increased activation of NF-κB might be a risk factor for the initiation or progression of PD, associated with mitochondrial dysfunction.
Overexpression of Relish in glial cells, but not in neurons, causes neurodegeneration in the fly CNS 49 . This suggests that the neurotoxicity induced by Relish is non-cell autonomous. It is therefore possible that activation of inflammatory pathways by the overexpression of constitutively active Relish in the gut of pink1 mutant flies would exacerbate their phenotype. This would further support the concept that immune dysfunction in the gut is a risk factor for PD.
Epidemiological evidence suggests that the risk of PD is increased in inflammatory bowel disease, a chronic condition usually diagnosed in young individuals 50 . Further work is required to understand how early gut inflammation contributes to the vulnerability of neurons with impaired mitochondrial function. This will improve our understanding of disease pathogenesis and has the potential to provide new therapeutic targets.
We show that dietary supplementation with the immunosuppressant tacrolimus is neuroprotective in pink1-mutant flies. Tacrolimus is a rapalog but, contrary to tacrolimus, rapamycin is described as having immunostimulatory activity 51 . We therefore reason that the neuroprotective roles of these two drugs might act by activation of distinct signaling pathways.
The defects in insulin signaling observed in pink1 mutants result from abnormal trafficking of lipids between mitochondria and the endoplasmic reticulum, which in turn affects the formation of lipid vesicles and the release of neuropeptides such as DILPs 17 . These defects in lipid trafficking arise from increased contacts between mitochondria and the endoplasmic reticulum observed in pink1mutant flies 13,17 . Although partial suppression of Relish rescued lipid defects in pink1 B9 flies (Fig. 6h) it did not rescue the release of DILP2 (Fig. 6I), confirming previous reports that insulin signaling acts upstream of Relish 25 .
Low levels of insulin are often associated with nutrient deficits and starvation, as indicated by increased expression of Takeout (Fig. 6c,d) 34 , which results in reduced inhibition of FOXO by Akt and thus promotion of a stress response 52 . However, Relish can shape metabolic adaptation by attenuation of FOXO-mediated lipolysis 6 , which leads to the accumulation of lipids 53 . Therefore, the sustained activation of Relish in pink1-mutant flies is likely to lead to the accumulation of lipids by blockage of FOXO-mediated lipolysis.
We observed an accumulation of TAGs in pink1-mutant flies, and mitochondrial impairment has been shown to result in TAG accumulation 54 . We previously reported that dicarboxylate fatty acids are increased in pink1 mutants, and this increase is probably linked to defects in fatty acid beta oxidation 12 . Therefore, accumulation of TAGs in these mutants might result from the failure of fatty acid beta oxidation due to mitochondrial impairment. This is further supported by the observation of normal feeding in pink1 mutants compared to controls-that is, the starvation phenotype does not seem to arise from low food intake, but rather from impairment of nutrient absorption by the gastrointestinal tract.
In mice, the gut-brain transmission of PD pathology by α-synuclein occurs via the vagus nerve 39 . Both retrograde and anterograde vagal transport act to promote bidirectional propagation of α-synuclein toxicity to the brain and other organs in rats 55 .
Our results (Figs. 7 and 8) show that a gut-brain pathway is involved in neurotoxicity in a Drosophila model of PD. However, unlike the results obtained in α-synuclein models, our model showed that this pathway does not involve the transport of a toxic protein but rather a signal induced by defective mitochondria.
In Drosophila, neurons with cell bodies in the posterior segments of the abdominal ganglion of the ventral nerve cord (VNC) send axons that reach the posterior portion of the midgut 40 , which is the region where we detected intestinal dysfunction in pink1-mutant flies. Because this link between the intestine and CNS via the VNC is established by ILP7 neurons (Fig. 6a), these neurons, like the vagus nerve in rodents, might mediate toxicity between the midgut and brain in pink1-mutant flies. It would be interesting to test whether silencing the activity of these neurons would affect CNS toxicity in pink1-mutant flies.
Alternatively, the communication of toxicity from the intestine to the CNS in pink1 mutants could involve an indirect (humoral) mechanism(s); pink1 mutants have impaired energy metabolism, and metabolic alterations due to starvation are known to cause mRNA downregulation of the hormone bursicon (burs-a) in the gut. Recent studies have shown that bursicon is involved in a neuronal relay to both adipose tissue 56 and the CNS 57 and can regulate sleep via indirect modulation of the DA system 57 . Given that our data indicate that silencing of Relish in the gut leads to metabolic improvement and rescue of PPL1 loss and sleep deficits, we investigated a possible link between Relish signaling and burs-a expression. We measured the levels of bursicon mRNA in pink1-mutant flies and found that suppression of Relish increased mRNA levels of burs-a in the gut (Fig. 8f). It is therefore plausible that impaired bursicon signaling acts as a humoral mechanism to relay toxicity from the intestine to the CNS. Although this requires further work, we propose that suppression of the IMD pathway rescues normal gut homeostasis, restoring bursicon levels. The hormone in the hemolymph will then relay the nutrient information via its neuronal links both to adipose tissue (via adipokinetic hormones) and potentially to the CNS where the bursicon receptor, rickets, is expressed in a small population of neurons. Moreover, pink1 mutants show increased sleep, which can be explained by a combination of the IMD-activated system 58 with reduced DA signaling (loss of PPL1) and possibly reduced bursicon levels. This potential mechanism is under investigation by our team.
An alternative, but not mutually exclusive, mechanism of humorally induced neurotoxicity is provided by a previous study showing that AMPs induced by Relish activation directly induce mitochondrial depolarization, in particular AMPs targeting Gram-negative bacteria 59 . AMPs in the intestine of pink1-mutant flies might act as a humoral signal to induce the loss of mitochondrial membrane potential in the CNS of these flies.
A third scenario, based on observations by Clark and colleagues 60 , is where damage to the intestinal barrier observed in pink1 mutants would cause systemic inflammation due to leakage of gut microbes into surrounding tissues. This last scenario would be an indirect mechanism of mitochondrial impairment and lead to an alternative parallel pathway for induction of inflammation.
Our data add to an increasing body of evidence showing that signals from the gut can modulate brain activity. More specifically, our results suggest that mitochondrial toxicity in the gut can contribute to adverse effects in the brain and that therapeutic interventions aiming to decrease gut toxicity might be a viable approach for restoration of brain health in diseases caused by mitochondrial compromise.

Methods
Genetics and Drosophila strains. Fly stocks and crosses were maintained on standard cornmeal agar media at 25 °C. The strains used were the following: pink1 B9 , white 1118 and w; elav > GAL4, which are previously described 12  To distinguish pink1 B9 flies from those with X-nondisjunction, we crossed pink1 B9 females with the Canton-S wild-type strain backcrossed to w 1118 ; thus, flies with nondisjunction have red eyes. This strategy was used for all crosses with mutants (that is, Rel E20 and eya 2 ). For all other crosses, flies were crossed to males carrying the FM7 balancer on the first chromosome (that is, pink1 B9 ; NP3084Gal4 × Fm7a; UAS-Rel KK). All experiments involving adult flies were performed with males aged 3-5 days unless otherwise stated.
Microarray acquisition and analysis. RNA was obtained from male adult flies (six samples in total, three replicates of each genotype). RNA quality was confirmed using an Agilent 2100 Bioanalyzer (Agilent Technologies). Detailed descriptions of experimental protocols and raw data have been deposited in ArrayExpress under accession no. E-MTAB-6210. Differential expression was analyzed using Partek Genomics Suite (Partek Inc.) with an analysis of variance (ANOVA) model.
Tandem mass tag (TMT) labeling was performed according to the manufacturer's recommended protocol (https://www.thermofisher.com/order/ catalogue/product/90110). One hundred micrograms of each digested protein sample was labeled individually with each of the ten TMT tags. After labeling, samples were combined, cleaned on a Sep-Pak C18 cartridge, dried and dissolved in 20 mM ammonium formate (pH 10). TMT peptide fractionation was performed using an Acquity ethylene-bridged hybrid C18 UPLC column (Waters; 2.1 × 150 mm 2 internal diameter, particle size 1.7 µm). Dried fractions were separated using the liquid chromatography-tandem mass spectrometry (LC-MS/MS) method detailed below. The fractions were combined into pairs (that is, the first fraction with the middle fraction) and analyzed by LC-MS/MS using a Dionex Ultimate 3000 RSLC nanoUPLC system (Thermo Fisher Scientific) and a Lumos Orbitrap mass spectrometer (Thermo Fisher Scientific). Mass spectrometry proteomics data have been deposited with the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD030979.

Data analysis.
To explore the transcriptional changes in genes involved in the Drosophila immune response, transcripts with fold change (FC) ≥2 in pink1mutant flies were crosschecked with a list of Drosophila genes potentially involved in the immune response, which was made available as a resource by the Lemaitre group (lemaitrelab.epfl.ch). Transcripts with FC ≥2 in pink1-mutant flies annotated to this list were then analyzed using the iRegulon algorithm 15 in Cytoscape (v.3.5.1). A total of 45 upregulated transcripts were classed as nodes for iRegulon analysis. We used the default parameters in iRegulon that search for transcription binding motifs from 5,000 bases upstream of the full transcript. The network for nodes connected to the top enriched transcription factor (Relish) was visualized in Cytoscape (Fig. 1c).
For proteomics analysis, the raw data files were processed using Proteome Discoverer v.2.1 (Thermo Fisher Scientific) and Mascot (Matrix Science) v.2.6. Data were aligned with the UniProt data from Pseudomonas aeruginosa (5,584 sequences), which is the common repository of adventitious proteins (cRAP, v.1.0). All comparative analyses were performed with R statistical language. The R package MSnbase 62 was used for processing of proteomics data. Briefly, this process entailed the removal of missing values (instances where a protein was identified but not quantified in all channels were rejected from further analysis), log 2 -transformation of raw data and subsequent sample normalization utilizing the 'diff.median' method in MSnbase (this translates all samples columns such that they match the grand median). Differential abundances of proteins were evaluated using the limma package, and differences in protein abundances were statistically analyzed using Student's t-test with their variances moderated by the empirical Bayes method in limma. P values were adjusted for multiple testing using the Benjamini-Hochberg method 63 .
Locomotor assays. Three-to four-day-old males were individually loaded into Drosophila Activity Monitors (DAM5) within 8 × 65-mm 2 glass Pyrex tubes (Trikinetics, Waltham) containing normal fly food. The flies were maintained at 25 °C under a 12/12-h LD cycle for at least 8 days. Sleep and activity data were analyzed using the Sleep and Circadian Analysis MATLAB Program (SCAMP) developed by the Griffith laboratory 64 . Analyses were performed for 7 days starting at the first Zeitgeber time (ZT0) to allow acclimation. At least 16 flies of each genotype were used. Flies with rhythmic index score <1 were removed from analyses.

Assay of integrity of the intestinal barrier.
We measured the integrity/ functionality of the intestinal barrier by detecting the presence of a nonabsorbable blue food dye outside the digestive tract after feeding 65 . The flies were transferred to food containing 2.5% w/v Erioglaucine (FD&C Blue dye1-SIGMA) overnight 66 . Next, the flies were allowed to eat normal food for an additional 24 h and then retransferred to dye-containing food for an additional 24 h. Flies were scored after being fed normal food for 48 h. Notably, in the analysis of Smurf flies fed normal food for 48 h, a stained abdomen served as an indicator of gastrointestinal dysfunction and we did not score different degrees of "smurfness". A pairwise test, followed by a stack of P values with a false discovery rate (FDR) of 10%, was used to detect the significance of differences among genotypes.
Drug treatments. Tacrolimus (FK-506 monohydrate, Sigma, no. F4679) was incorporated into fly food to a final concentration of 5 mM. Flies treated with tacrolimus were transferred to drug-containing food up to 24 h after hatching.
Protein extraction and immunoblotting. Fly gut samples: ten to 15 guts were dissected in PBS and maintained on ice until required for processing. Protein extraction was performed using NE-PER Nuclear and Cytoplasmic Extraction Reagents (Thermo Scientific, no. 78833) following the manufacturer's instructions. Samples were separated using Bio-Rad Precast gels and wet blotted onto nitrocellulose membrane. Membranes were blocked in TBS (0.15 M NaCl and 10 mM Tris-HCl, pH 7.5) containing 10% (w/v) dried nonfat milk for 1 h at room temperature, then probed with primary antibodies before incubation with the appropriate horseradish peroxidase (HRP)-conjugated secondary antibody. Antibody complexes were visualized by Pierce enhanced chemiluminescence.
Whole-fly samples: protein extracts from whole flies were prepared by grinding in RIPA buffer (0.15 M NaCl, 1% (v/v) Triton X-100, 0.5% (m/v) sodium deoxycholate, 0.1% (v/v) SDS and 50 mM Tris (pH 7.5)) supplemented with 1× phosphatase and protease inhibitor. Tissue and cell debris were spun down at 14.000g for 10 min and supernatant collected. Next, 30 µg of protein was loaded in Novex Tris-Glycine precast gels (Thermo Fisher Scientific) and wet blotted onto nitrocellulose membranes. Membranes were blocked in TBS (0.15 M NaCl and 10 mM Tris-HCl, pH 7.5) containing 5% (w/v) dried nonfat milk for 1 h at room temperature, then probed with primary antibodies before incubation with the appropriate IRDye-conjugated secondary antibody. Antibody complexes were visualized using Odyssey (LI-COR) and quantifications were performed using Image Studio Lite v.5.2.5 (LI-COR), with normalization to the respective loading control (actin).

RNA extraction and quantitative real-time PCR with reverse transcription.
Total RNA was extracted from 10-15 freshly dissected guts using TRIzol (Ambion), and quantified by spectrophotometric analysis (Nanodrop, Thermo Scientific). Quantitative real-time PCR with reverse transcription (RT-qPCR) was performed with a real-time cycler (Applied Biosystems 7500, Fast Real-Time PCR Systems) using the SensiFAST SYBR Lo-ROX One-Step Kit (Bioline).
Immunofluorescence and confocal microscopy. For brain imaging, flies were fixed overnight at 4 °C in 4% PFA/1% Triton/PBS. Brains were subsequently dissected in ice-cold PBS and blocked overnight in 10% normal goat serum PBS/0.5% Triton. Primary staining was performed at 4 °C for 3 days, while samples were incubated with secondary antibodies for 2 h at room temperature.
For intestinal imaging, guts were dissected from live flies in PBS and fixed in 4% PFA for 1 h on ice. The staining protocol was identical to that used for brain samples.
Colocalization of anti-dsDNA (Abcam, no. ab27156, 1:100) and mito-GFP was analyzed with the colocalization tools in ZEN Blue v.2.6. The Pearson coefficient calculates pixel-based overlap of the fluorescence channels under analysis. A Pearson correlation coefficient of 1 indicates perfect correlation between two image channels. Images were obtained from five 10-µm ROIs per gut, and at least nine guts were analyzed per genotype.

Microscopy-based assessment of mitochondrial function and morphology.
Measurements of Δψm in fly brains were performed as previously described 12 . Briefly, fly brains were loaded for 40 min at room temperature with 40 nM tetramethylrhodamine methyl ester (TMRM) in loading buffer (10 mM HEPES pH 7.35, 156 mM NaCl, 3 mM KCl, 2 mM MgSO 4 , 1.25 mM KH 2 PO 4 , 2 mM CaCl 2 and 10 mM glucose), with the dye present during the experiment. In these experiments, TMRM was used in redistribution mode for assessment of Δψm and therefore a reduction in TMRM fluorescence represents mitochondrial depolarization. Confocal images were obtained using a Zeiss LSM 880 confocal microscope equipped with a ×40 oil immersion objective. Illumination intensity was kept to a minimum (0.1-0.2% of laser output) to avoid phototoxicity, and the pinhole was set to yield an optical slice of 2 μm. Fluorescence was quantified by excitation of TMRM using the 565-nm laser and measured >580 nm. Z-stacks of five 300-μm 2 fields were acquired from each brain, and the mean maximal fluorescence intensity of each group was measured.
Mitochondrial length was quantitated in enterocytes expressing mito-GFP in the posterior midgut region. Samples were fixed, mounted on Vectashield (Vector Laboratories) and imaged on a Zeiss LSM880 confocal microscope. Mitochondrial length was calculated using the ruler and measurement tools in Photoshop, to measure the length of mito-GFP-positive mitochondria across their largest dimension.
Analysis of DA neurons. Brains from 25-day-old flies were dissected and stained with anti-TH antibody (Immunostar, no. 22941, 1:50) as previously described 67 . Brains were positioned in PBS + 0.1% Triton in a coverslip clamp chamber (ALA Scientific Instruments) using a harp composed of platinum wire and nylon string, and imaged by confocal microscopy. The numbers of TH-positive neurons in the PPL1 cluster were determined for each brain hemisphere. The data acquired for assessment of each genotype were obtained as a single experimental set before statistical analysis.
Analysis of food intake. Ten adult male flies for each genotype were aged for up to 20 days on normal food, which was changed every 2-3 days. After 20 days the flies were transferred to food supplemented with 150 µl of 2.5% bromophenol blue diluted in water, inserted into the food by piercing the surface of the food several times. Bromophenol blue was allowed to diffuse into the food for 24 h at room temperature. Flies were placed in bromophenol blue-containing food for 16 h. Following feeding with dye-containing food, the flies were transferred to a preweighted empty tube. The tubes with flies were reweighted, flash-frozen in liquid nitrogen and stored at -80 °C until processing. Flies were processed by washing three times in PBS to remove traces of food on the cuticle (three × 5 min on a rotating wheel at 4 °C). They were then homogenized in 100 µl of PBS using a motor pestle. Homogenates were then centrifuged for 10 min at 16,000g at 4 °C and the supernatant was transferred to a new tube. The absorbance of the supernatant (80 µl) was read at 590 nm using 96-well microtiter plates and an Infinite M200Pro multifunction reader (TECAN). Absorbance values were normalized to the weight of the flies. Flies fed food without bromophenol blue were used as assay blanks.
Metabolic assays. Assessment of DILP2 was performed by ELISA, while levels of TAGs and beta oxidation were assessed through colorimetric assays using 96-well microtiter plates and an Infinite M200Pro multifunction reader (TECAN).
For metabolic assays, groups consisting of a maximum of 20 male flies were placed in normal food-containing vials and the food was changed every 2 days until the desired age was reached-that is, either 3 or 30 days-and maintained under 12/12 LD at 25 °C. Day 0 was considered the day when flies emerged from the pupal case. At the defined aging stage, flies were transferred to vials containing cotton soaked with water for 2 h to minimize food contamination, from ZT2 to ZT4. CO 2 -anaesthetized flies were then divided into groups of five and flash-frozen for subsequent analysis. TAG assays were essentially performed as previously described 68 . Briefly, for TAG analysis, five adult flies were homogenized in 100 µl of PBS + 0.05% Tween 20 (PBST) for 60 s on ice and immediately incubated at 70 °C for 10 min for inactivation of endogenous enzymatic activity. Forty microliters of fly sample and glycerol standard (Sigma, no. G7793) were incubated together with either 40 µl of PBST (for free glycerol measurements) or 40 µl of TAG reagent (Sigma, no. T2449, for TAG measurements) at 37 °C for 60 min.
After 3 min of centrifugation at full speed, 30 µl of each sample (two technical replicates per biological sample) was transferred to a clear-bottom plate together with 100 µl of free glycerol reagent (Sigma, no. F6428) at 37 °C for 5 min.
For insulin ELISA, flies were maintained and aged as for the TAG assay with the modification that they were not starved for 2 h but rather collected at ZT2. ELISA was performed as previously described 37 with the modification that 5 µl of anti-HA-HRP antibody diluted 1:350 in PBS was added to 50 µl of protein sample 69 . Briefly, wells (Thermo Scientific, no. 46867) were coated with 100 µl of anti-FLAG antibody (Sigma, no. F1804) diluted in 0.2 M sodium carbonate/bicarbonate buffer (pH 9.4) to a final concentration of 2.5 µg ml -1 , and incubated overnight at 4 °C. The plate was washed twice with PBS containing 0.2% Tween 20 (PBTw0.2) for 5 mins each, then coated with 350 µl of PBS containing 2% bovine serum albumin (BSA) for 16 h at 4 °C. The plate was then washed three times with PBTw0.2. Single-fly extract was prepared by placing individual flies in a 1.5-ml Eppendorf tube containing 100 µl of PBS with 1% Triton X-100. Samples were ground using a motor pestle on ice and lysed for 30 min at room temperature on a rotary shaker. Samples were then spun at 21,000g for 2 min at room temperature and supernatant transferred to fresh tubes. Next, 50 µl of lysate was mixed with 5 µl of anti-HA-HRP antibody (Roche Applied, no. 12013819001) diluted 1:350 in PBS, vortexed, centrifuged briefly and then added to precoated, preblocked ELISA wells overnight at 4 °C with gentle agitation. Samples were then removed by aspiration and wells washed six times with PBTw0.2; 100 µl of TMB ELISA substrate (Thermo Scientific, no. 34029) was added to each well with incubation on a rotary sharker for 30 min in the dark at room temperature. The reaction was stopped by the addition of 100 µl of 2 M sulfuric acid, and absorbance at 450 nm was immediately measured on the plate reader. Subsequently, TAG and DILP2 absorbances were divided by the protein concentration of the respective sample, which was measured by Bradford assay. The results were normalized to the control group by dividing the average values 70 .
Similarly, for beta oxidation analysis, ten flies of the desired age were collected and maintained on ice following the manufacturer's instructions.
For beta oxidation analysis, flies were maintained and aged as described above with no starvation step and collected at ZT2. Briefly, 15 flies (for each biological replicate) were anesthetized, transferred to Eppendorf tubes, weighed and maintained on ice. Approximately 10 mg of tissue was used for each replicate. The flies were subsequently homogenized in 200 µl of provided lysis buffer, and the lysate was clarified by centrifugation at 4 °C. The Bradford assay was used for quantification of protein concentration, and all samples were normalized to 2 mg ml -1 . The assay was then performed based on the protocol suggested by the manufacturer (BMR, no. E-141). Incubation at 37 °C was performed for 1 h.
For all Bradford assays, a fresh standard curve of BSA (Thermo Scientific, no. 23210) was prepared for each experiment.
Statistical analyses. Statistical analyses were performed using GraphPad Prism (www.graphpad.com). Data are presented as mean values, with error bars indicating ±s.d. In violin plots the solid line represents the median while dashed lines represent the quartiles. Individual data points in the figures correspond to biological replicates. The number of biological replicates per experimental variable (n) is indicated in either the respective figure or figure legend. Significance is indicated as follows: *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001; NS indicates P ≥ 0.05.
For all statistical analyses, the Shapiro-Wilk test was used for for low n values (<7) and the D' Agostino-Pearson test for higher values (>7). Based on normality test results, parametric and nonparametric analyses were chosen accordingly.
Digital image processing. Fluorescence and immunoblot images were acquired as uncompressed bitmapped digital data (tiff format), and processed using Adobe Photoshop with established scientific imaging workflows 71 . To visualize pixel intensity, confocal images acquired with identical settings were processed using a five-tone heat map (f.64 Academy) in Adobe Photoshop.