Introduction

Metazoan organisms must constantly adapt to the surrounding environment. Environmental factors, including radiation, pathogens or toxins, constantly lead to the generation of reactive oxygen species (ROS)¹. Low levels of ROS support cellular pathways like immune cell differentiation and even immune function such as antimicrobial defense^{2, 3}. At high concentrations, ROS leads to oxidative stress, a deleterious process that negatively impacts protein synthesis and stability, lipids as well as overall genome stability. ROS mediated cellular damage can further result into degenerative diseases or cancer⁴. Therefore, organisms employ efficient repair and defense mechanisms to circumvent high levels of oxidative stress and subsequent organ damage. Hence, keeping the balance between the production and the antioxidant-dependent clearance of ROS is essential for the survival and well-being of the affected organism⁵. Oxidative stress induced tissue damage is most often followed by a strong local immune response. These molecular processes are conserved across vertebrates and invertebrates such as the fruit fly Drosophila melanogaster⁶.

Hemocytes, the Drosophila blood cells, are essential cellular components of the immune and stress response machinery. They consist of three different subtypes: crystal cells, lamellocytes and plasmatocytes⁷. Plasmatocytes are considered to be macrophage equivalents in Drosophila and comprise up to 95 % of all hemocytes in the adult fly⁸. They share many functions as well as the multilayered ontogeny with vertebrate tissue macrophages^{9, 10}. Adult plasmatocytes are actively involved in many immune functions such as pathogen recognition, defense and phagocytosis but also chronic immune activation upon metabolic stress by dietary lipids^{11, 12}. Recent studies highlighted the essential role of adult plasmatocytes as key players in regulation of tissue homeostasis, including the gut by regulating the intestinal stem cell niche or in muscles by controlling glucose homeostasis^{13, 14}. Similar to higher eukaryotes, levels of ROS are used as a central antimicrobial defense mechanisms of hemocytes against invading pathogens¹⁵. Moreover, larval hemocytes respond differentially to ROS, with two cell subsets mounting a biphasic immune reaction to invading bacteria by controlling hemocyte spreading and adhesion as well as crystal cell rupture¹⁵. Larval hemocytes have been also attributed to respond to increased oxidative stress of other tissues, as seen in response to oxidative stress in the eye-antenna imaginal disc¹⁶. Similarly, by regulating the intestinal stem cell (ISC) proliferation in the damaged midgut during infection-induced or chemical-induced oxidative stress¹⁴, adult hemocytes were described to release the bone morphogenetic protein (BMP) homologue decapentaplegic (dpp) to facilitate tissue repair¹⁴. Other studies pointed to an essential role of hemocyte-derived unpaired-3 (upd3), which is the invertebrate orthologue of Interleukin-6, in the regulation of ISC proliferation during septic injury¹⁷, as well as wound healing during sterile injury¹⁸. In line with these findings, hemocytes have been implicated in organ-to-organ immune signaling for example upon infection-induced injury in the gut and antimicrobial peptide (AMP) production in the fat body¹⁹. Hemocyte activation is thus a hallmark of oxidative stress in defined tissues, though their role in modulating and controlling oxidative stress response to facilitate tissue repair and survival of the organism remains undefined.

Here, we demonstrate an essential role of hemocytes to control susceptibility to oxidative stress, where hemocytes orchestrate systemic oxidative stress response by defined cellular states identified by unbiased transcriptomic profiling. Plasmatocytes, which make up 95% of all hemocytes, segregate in oxidative stress responder subclusters characterized by immune activation and cellular stress response and potential bystander subsets more associated with metabolic gene regulations. Mechanistically, we unravel that the direct immune response to oxidative stress by plasmatocytes seems to be essentially controlled by non-canonical DNA damage signaling which subsequently regulates Jun kinase (JNK) activation and induction of the pro-inflammatory cytokine upd3. We further describe that plasmatocytes are a major cellular source of upd3 and that regulation of upd3 levels in the adult fly determine the organism’s susceptibility to oxidative stress, most likely by orchestrating systemic stress responses in other tissues, such as energy mobilization in the fat body.

Results

Hemocytes control susceptibility to oxidative stress

In vertebrates and invertebrates, prolonged oxidative stress can cause severe tissue damage, hence controlled activation of stress signaling cascades are essential for the survival of the organism and the resolution of oxidative stress with subsequent tissue repair. Yet the precise role of macrophages in the initiation and control of oxidative stress signaling needs to be determined.

To analyze the role of macrophages in signaling cascades to oxidative stress, we took advantage of the well-established Paraquat (PQ)-feeding model in adult Drosophila melanogaster to induce systemic oxidative stress²⁰. Feeding of 7-days old w;Hml-Gal4,UAS-2xeGFP/+ (Hml/+) flies with PQ-supplemented sucrose solution led to an increased lethality of adult Drosophila after 18 hours in a dose dependent manner (0mM: 1.25 ± 1.25 %, 2mM: 2.5 ± 1.6 4%, 15mM: 27.78 ± 5.47 %, and 30mM: 38.75 ± 6.67 %) (Figure S1A). In agreement with previous studies²⁰, 15mM PQ was chosen for further experiments (Figure 1A). First, we determined which systemic changes are associated with susceptibility to oxidative stress. Infection induced damage in enterocytes can result in gut leakage and subsequent death¹⁷. We tested if the death of flies after 18 hours on 15mM PQ food was associated with PQ-induced damage of gut enterocytes and resulting gut leakage, but we found no indication of gut leakage on 15mM PQ food (Figure S1B). Only very few flies on control food and PQ indicated gut leakage of the food color to thorax and abdomen which was unrelated to the treatment and not altered between the groups (control: 3.75%, 2mM PQ: 2.53 %, 15mM PQ: 1.11 % and 30mM PQ: 3.75 %) (Figure S1B). We investigated if aberrant systemic induction of immune signaling cascades are seen upon PQ-mediated oxidative stress. For this gene expression levels of different known immune activated genes were analyzed by qPCR such as unpaired (upd) cytokines (upd1, upd2, upd3), Jak/STAT down-stream targets (Turandot-A (TotA), suppressor of cytokine signaling at 36E (Socs36E)), stress induced JNK activation (puckered (puc)) (Figure 1B), as well as expression of nuclear factor ‘kappa-light-chain-enhancer’ of activated B-cells (Nfkb)-induced antimicrobial peptides (AMP) and further cytokines such as eiger, dpp and dawdle (daw) (Figure S1C-D). These analyses did not reveal significant differences for these gene groups on a systemic level. In contrast, genes associated with carbohydrate metabolism such as insulin receptor (InR), phosphoenolpyruvate carboxykinase (Pepck), and Thor were significantly upregulated upon PQ-feeding (Figure 1C). These findings point to an activation of forkhead box, sub-group O (foxo)-target genes and a potentially reduced insulin signaling, despite the lack of significant alterations in the expression of the Insulin-like peptides (Ilps) Ilp2, Ilp3 and Ilp5 (Figure 1C).

Figure 1.

To test the consequences for energy mobilization and nutrient storage, we measured free glucose content, as well as glycogen and triglyceride levels in whole flies after 18 hrs on control or 15mM PQ. Although no changes in free glucose levels (control: 1.49±0.02µg and PQ: 1.50±0.07µg, ns p=0.954) (Figure 1D) were found, we observed a clear decrease in stored glycogen (control: 1.02±0.07µg and PQ: 0.33±0.08µg, **p=0.003) (Figure 1E) and stored triglycerides (Figure 1F) in PQ-treated flies compared to control. Next, we examined histological sections of flies to define which tissues might be affected. Hematoxylin and eosin (H&E) stainings showed a loss of stored lipids in the abdominal fat body compared to flies kept under control conditions (Figure S1E). Fat body development, energy mobilization and storage have been recently connected to the presence, cell number and subsequent cross-talk of hemocytes with fat body²¹, therefore we quantified GFP⁺ hemocytes in Hml/+ flies after 18 hrs on PQ or control food (Figure 1G-H). Our analysis did not reveal alterations in hemocyte numbers and especially no oxidative stress induced loss of hemocytes (control: 243.4±24.33 hemocytes per fly and PQ: 219.6±24.22 hemocytes per fly, ns p=0.508) (Figure 1G-H). To evaluate if hemocytes are essential contributors to the stress-mediated metabolic changes and susceptibility to oxidative stress, we used “hemocyte-deficient” crq-Gal80^ts>reaper (w;tub-Gal80ts/UAS-CD8-mCherry;crq-Gal4/UAS-reaper) flies and tested their survival rate compared to control flies on PQ food (Figure 1I). Hemocyte-deficient flies displayed a significantly higher mortality rate on 15mM PQ (43.33±6.30%) compared to control flies (13.08±2.63%). Our findings show detrimental changes in energy mobilization upon oxidative stress in the absence of profound over-activation of immune signaling cascades, however the presence or absence of hemocytes seem to be a key determinant of the susceptibility to oxidative stress in adult Drosophila. This strongly points to an essential role for hemocytes in controlling stress response and organismal survival upon oxidative stress.

Paraquat exposure generates diverse transcriptional states of adult hemocytes with specific transcriptional state associated with oxidative stress response

To gain further insights into the alterations induced in the adult fly upon oxidative stress and especially to the hemocyte compartment, we decided to take advantage of unbiased single nucleus RNA-sequencing (snRNA-seq). We isolated nuclei from control food and PQ-treated Hml-DsRed.nuc flies with a modified Frankenstein protocol²² and performed fluorescence activated nuclear sorting of DAPI⁺Draq7⁺ nuclei (Figure S2A). From two independent sorts we obtained a total of 29000 nuclei which were utilized for subsequent droplet-based barcoding of single nuclei, and generation of single nucleus cDNA libraries. We could successfully obtain gene expression data from 11839 nuclei (Figure S2B-D). Quality control analysis followed by unsupervised hierarchical clustering of all analyzed nuclei using Seurat package²³ revealed 20 distinct nuclei clusters (C) with differential gene expression (Figure 2A). Cell identity was annotated according to the cell type-specific signatures determined by the Fly Cell Atlas consortium (Figure 2A)²⁴. All annotated cell types were detected in PQ-treated and control flies (Figure 2B).

Figure 2.

Using cell type-specific signatures, three of the clusters were annotated as hemocytes (clusters: 2, 4 and 11) (Figure 2A-B). Re-scaling and clustering of all 1354 identified hemocyte nuclei revealed 8 distinct clusters (Figure 2C). We verified hemocyte-specific signature genes across these clusters including genes such as the transcription factor Serpent (Srp) and one or more of the following hemocyte signature genes: Hemolectin (Hml), croquemort (crq), Nimrod C1 (NimC1), eater, and Peroxidasin (Pxn) (Figure S3A-L)^{25, 26}. Cluster 8 expressed various crystal cell-specific markers including lozenge (lz), pebbled (peb), prophenoloxidase 1 (PPO1), and prophenoloxidase 2 (PPO2), whereas all other clusters seemed to be plasmatocytes (Figure S3A-L). In line with previous reports, the embryonic and larval hemocyte markers Hemese (He) (data not shown) and singed (sn) were not expressed across any of the identified clusters (Figure S3H). Interestingly, when we defined which hemocyte clusters were enriched in control flies or PQ-treated flies, the two plasmatocyte clusters C1 and C7 were primarily derived from control flies, whereas C2, C3, C4 and C6 are predominantly found in PQ-treated flies (Figure 2D and Figure S3M). The plasmatocyte cluster C5 as well as the crystal cell cluster C8 presented in both conditions to similar proportions (Figure 2D and Figure S3M).

Next, we analyzed the TOP20 differentially expressed (DE) genes across all hemocyte clusters. We found clear differences between the 7 distinct plasmatocyte clusters and the crystal cell cluster (Figure 2E and Table S1). The crystal cell cluster C8 showed a high expression of crystal cell specific genes including lz, peb, PPO1 and PPO2 (Figure 2E). Plasmatocyte cluster C1, which was mostly present in control conditions, had a remarkable expression of hemocyte specific marker genes such as Hml and Pxn, as well as migration-associated genes including the semaphorins Sema2a and Sema5c (Figure 2E). Yet this cluster also showed higher expression of genes contributing to metabolic regulations such as the L-gulonate 3-dehydrogenase Had2 and the gluconolactonase regucalcin (Figure 2E). The cluster C7 and C1, mostly found in control flies, substantially shared gene expression profiles. In contrast, cluster C2 and C3 were nearly exclusively found in PQ-treated flies and showed expression of genes involved in metabolic adaptations, with high expression of the glucose transporter 4 enhancer factor (Glut4EF) or the phosphatidate phosphatase Lipin (Lpin) in C2 or the induced expression of the lipase brummer (bmm), the autophagy-associated gene Atg17 or the lipophorin receptor LpR1, involved in uptake of neutral lipids from the hemolymph, in C3 (Figure 2E). Overall gene expression profiles in clusters C2 and C3 showed an induction of insulin signaling, increased autophagy and lipolysis pointing to adaptation of energy mobilization in these clusters upon PQ treatment (Figure 2E). Cluster C4 was characterized by a remarkable upregulation of genes involved in translation including many ribosomal proteins such as RpL8, RpL31 or RpS13 (Figure 2E). Furthermore, C4 included genes associated with unsaturated fatty acid metabolism such as Desaturase 1 (Desat1). Similarly, cluster C5 also showed induction of genes involved in fatty acid metabolism such as ELOVL fatty acid elongase 7 (ELOVL) (Figure 2E). Of note, we also found Tie-like receptor tyrosine kinase (Tie) specifically expressed in this cluster, which is described to bind PDGF- and VEGF-related factor (Pvf) ligands and contribute to cell survival and migration, as well as actin-associated binding proteins including polychaetoid (pyd) and formin 3 (form3). All these gene expression changes are pointing to a plasmatocyte cluster with a gene expression profile associated with mobility and migration. Among all plasmatocyte clusters, we found cluster C6 which presented a strong immune activation phenotype with induction of Jak/STAT target genes including Fasciclin 3 (Fas3) and Turandot-M (TotM) among its TOP20 DE genes, but also activation of oxidative stress response genes such as the heat shock proteins Hsp23, Hsp26 and Hsp27 or cytoskeletal genes including myosin Myo31DF or spire (spir) (Figure 2E).

Due to the partially shared expression profile of genes across hemocyte clusters, we wondered if there were genes exclusively expressed in individual clusters. However, cluster C1 showed that Sema2a seems to be exceptionally high expressed in these homeostatic plasmatocytes and genes associated with metabolic regulation (e.g. InR, Lpin or bmm) or Jak/STAT activation (e.g. Fas3), as seen in PQ-associated clusters, are downregulated (Figure 2F). Cluster C5 showed enrichment for genes such as the transcription factor grainy head (grh) but also cytoskeletal-associated genes such as jitterbug (jbug) and form3, suggesting a migratory potential of plasmatocytes from this cluster within their host tissues (Figure 2F). Interestingly, plasmatocyte cluster C6 presented with the most profound gene expression changes compared to all other clusters. Similar to our previous observations, we identified that genes specifically expressed in this cluster were associated with Jak/STAT pathway activation (e.g. Fas3, TotM), cytoskeletal rearrangements and migration (e.g. Myo31DF, spir or cheerio (cher)) and response to oxidative stress (e.g. Nicotinamide amidase (Naam) or Hsp23) (Figure 2F). We also included the crystal cell cluster C8 in this analysis and verified unique expression of crystal cell markers compared to the seven plasmatocyte clusters (Figure 2F). Our snRNA-seq analysis revealed that adult hemocytes respond differentially to oxidative stress with a specific cluster of plasmatocytes associated with immune activation and stress response. Despite our findings that on systemic level we did not detect a strong induction of immune activation upon oxidative stress, we now provide evidence that plasmatocyte cluster C6 showed a substantial immune response towards oxidative stress.

Direct response to oxidative stress in vitro mimics the transcriptional changes seen in cluster 6 plasmatocytes in vivo

Plasmatocytes responded to PQ-mediated oxidative stress in different cellular states. We defined cluster C6 as a cell state which is associated with oxidative stress response and immune activation. Since cluster C6 showed gene expression associated with oxidative stress and immune activation we focused our further analysis on this cluster. To further substantiate our findings and address the signaling pathways that could lead to diverse transcriptional states, we inferred the transcription factor (TF) network based on gene expression data (Figure 3A). This analysis revealed that genes with binding sites for Xbp1 were enriched. Xbp1 is a transcription factor associated with stress response. In addition to this, the Nf-κB-related transcription factor Relish (Rel) and kayak (kay), a component of the activator protein-1 (AP-1) transcription factor complex induced by JNK signaling, were also enriched in this cluster (Figure 3A). To determine if the gene expression signature of cluster C6 plasmatocytes reflects a functional state of plasmatocytes directly induced by oxidative stress, we treated S2 cells, a Drosophila plasmatocyte cell line, with different concentrations of PQ (15mM and 30mM) (Figure S4). First, we performed flow cytometric analysis of ROS levels via CellROX staining after 6 and 24 hours PQ treatment compared to controls (Figure S4A-B). We observed a dose-dependent increase in intracellular ROS at all time points analyzed with higher ROS levels at 6 hrs (0mM: 1903±64 median fluorescence intensity (MFI), 15mM: 6735±1096 MFI, and 30mM: 7435±237 MFI) compared to 24 hrs (0mM: 1236±135 MFI, 15mM: 4456±228 MFI, and 30mM: 5557±303 MFI) (Figure 4C). Increase in ROS and oxidative stress can lead to cellular damage such as damage in nuclear and mitochondrial DNA.

Figure 3:

Figure 4:

We next tested whether, we now aimed to test if plasmatocytes directly exposed to oxidative stress might also experience these type of stress response by performing comet assays of treated and untreated S2 cells. Upon PQ treatment, S2 cells showed a remarkable and dose-dependent increase in DNA damage compared to untreated cells (0mM: 101.0±2.1, 5mM: 117.6±3.3, 10mM: 136.6±3.8, and 15mM: 147.8±5.5, data shown as olive tail moment) (Figure S4D). We were then wondering if PQ-treated S2 cells would resemble cluster C6 plasmatocytes in vivo and performed gene expression analysis of S2 cells treated with 15mM PQ after 6 and 24 hrs. Comparable to our results for cluster C6 in vivo, we found a profound induction of the Jak/STAT target genes by RTq-PCR. In S2 cells, we also detected a significant increase of the Jak/STAT cytokines upd1, upd2 and upd3 (Figure S4E), which we did not see before in vivo. Furthermore, we detected an increased expression of the JNK-target gene puckered (puc), but barely any induction of metabolic genes (Figure S4E). All these results further support our hypothesis that direct exposure of S2 cells with PQ in vitro mimics gene expression changes seen in C6.

Because we found enrichment of genes associated with the Nf-κB-related transcription factor Rel in cluster C6, we further verified if any other cytokines or AMPs are induced in PQ-treated S2 cells. We identified a slight but significant induction of the cytokine daw, but no other alterations in dpp or the AMPs Drs, Dro or Mtk (Figure S4F). The profound induction of Jak/STAT target genes together with the activation of JNK target genes and oxidative stress induced DNA damage in S2 cells in vitro highly support our hypothesis that cluster C6 plasmatocytes reflect a cellular state associated with the direct response to oxidative stress in vivo. Along this line, we went back to our snRNA-seq data and specifically searched for Jak/STAT target genes (upd3, Socs36E, TotA, TotM and Fas3) across all clusters (Figure 3B). Here we identified an enriched expression of these genes in cluster C6, and a unique expression of the pro-inflammatory cytokine upd3 in cluster C6 which was not detected in any other hemocyte cluster (Figure 3B). Interestingly, we also identified that Growth arrest and DNA damage-inducible 45 (Gadd45), a gene linking DNA damage signaling and stress induced JNK signaling, was enriched in cluster C6 (Figure 3B). This could indicate that cluster C6 also undergoes oxidative stress mediated DNA damage with subsequent JNK activation, similar to what we have seen in S2 cells in vitro. Of note, the JNK target gene puc was found to be highly expressed across all hemocyte clusters analyzed in vivo (Figure 3B). We looked again into the expression of metabolic genes including Thor and InR as well as other cytokines such as dpp, daw or egr but could not detect a specific enrichment in cluster C6 similar to our results from PQ-treated S2 cells in vitro (Figure 3B). These findings from the snRNA-seq analysis in vivo and S2 cells in vitro support the hypothesis that cluster C6 plasmatocytes represent a functional state of activated plasmatocytes directly responding to oxidative stress by JNK signaling and Jak/STAT pathway activation. Furthermore, we found evidence that cluster C6 specifically expresses the cytokine upd3, which could reflect an important mediator for oxidative stress response.

Loss of canonical DNA damage signaling in adult hemocytes results in elevated upd3 levels and a decreased survival on oxidative stress

We observed that hemocytes respond with a distinct gene expression pattern specific to oxidative stress (Figure 1-3). Previously it has been observed that oxidative stress-induced DNA damage supports immune activation in macrophages via a non-canonical DNA damage response²⁷. Based on our findings, we predict that also in adult hemocytes oxidative stress not only drives genome instability and activates repair, but is rather essential to induce immune activation via specific gene expression changes in cluster C6 plasmatocytes. To test if DNA damage signaling is involved in the response of adult hemocytes to oxidative stress, we analyzed flies with a hemocyte-specific knock-down for different genes of the DNA damage repair (DDR) machinery including the DNA damage sensing kinases telomere fusion (tefu) and meiotic 41 (mei-41) and the MRN complex protein nbs. Of note, by targeting adult hemocytes, mostly plasmatocytes are targeted which make up 95% of all hemocytes in the adult fly.

We generated Hml>mei-41-IR (w;Hml-Gal4,UAS-2xeGFP/UAS-mei-41-IR), Hml>tefu-IR (w;Hml-Gal4,UAS-2xeGFP/UAS-tefu-IR), Hml>mei41-IR;tefu-IR (w;Hml-Gal4,UAS-2xeGFP/UAS-mei-41-IR;UAS-tefu-IR) and Hml>nbs-IR (w;Hml-Gal4,UAS-2xeGFP/UAS-nbs-IR) flies (Figure S5A). First, we compared their susceptibility to PQ treatment after 18 hrs and found significantly increased susceptibility to the treatment across all DDR-deficient lines compared to the control genotype (Hml/+: 10.7±1.2%, Hml>mei-41-IR: 40.7±4.8%, Hml>tefu-IR: 20.7±4.3%, Hml>mei41-IR;tefu-IR: 37.5±4.8% and Hml>nbs-IR: 48.2±6.7%) (Figure 4A). In these flies, decreased survival after PQ did not correlate to an overall reduction in lifespan on normal food and a decrease in fitness (Figure S5B).

Next, we inquired if DDR-deficiency in hemocytes might increase susceptibility of the fly to other stresses. For this we challenged all genotypes with starvation. Importantly, we did not observe a decreased survival of DDR-deficient flies on starvation compared to the control genotype, indicating that the observed susceptibility to oxidative stress of flies with a DDR-deficiency in hemocytes is specifically caused by oxidative stress due to PQ treatment (Figure S5C). As defects in the DDR machinery in hemocytes might already affect the development during embryonic and larval stages, we used a hemocyte-specific temperature-inducible knock-down of mei-41, tefu and nbs1 which was induced by a temperature shift to 29°C directly after eclosion from the pupae to induce a knockdown in adult hemocytes only. We analyzed the lifespan of Hml-Gal80^ts>mei-41-IR (w;Hml-Gal4,UAS-2xeGFP/UAS-mei-41-IR;tub-Gal80), Hml-Gal80^ts>tefu-IR (w;Hml-Gal4,UAS-2xeGFP/UAS-tefu-IR;tub-Gal80), Hml-Gal80^ts>mei-41-IR;tefu-IR (w;Hml-Gal4,UAS-2xeGFP/UAS-mei-41-IR;tub-Gal80/UAS-tefu-IR) and Hml-Gal80^ts>nbs-IR (w;Hml-Gal4,UAS-2xeGFP/UAS-nbs-IR;tub-Gal80) compared to control flies Hml-Gal80^ts/+ (w;Hml-Gal4,UAS-2xeGFP/+;tub-Gal80) and again found no alterations in overall lifespan of all genotypes analyzed on control food except a slight decrease in lifespan of Hml-Gal80^ts>mei-41-IR;tefu-IR. (Figure S5D). In contrast, we again found a higher susceptibility to oxidative stress via PQ-feeding in the DDR-deficient lines compared to control (Hml-Gal80^ts/+: 30.3±5.2%, Hml-Gal80^ts>mei-41-IR: 36.2±4.3%, Hml-Gal80^ts>tefu-IR: 52.4±6.9%, Hml-Gal80^ts>mei-41-IR;tefu-IR: 72.3±6.3% and Hml Hml-Gal80^ts>nbs-IR: 36.2±4.3%) (Figure S5E), which directly confirms our previous results and conclusions (Figure 4A).

As we have seen before that loss of hemocytes is detrimental for the survival on oxidative stress, we wanted to exclude a reduction in hemocytes in the DDR deficient fly lines as a cause of the higher susceptibility to oxidative stress. We quantified GFP⁺ hemocytes in Hml>mei-41-IR, Hml>tefu-IR, Hml>mei-41-IR;tefu-IR and Hml>nbs-IR flies (Figure 4B, Figure S5A). We did not find alteration in hemocyte number across all fly lines on control food (with the exception of an increase in hemocytes in the Hml>tefu-IR line), nor did we observe any effect of PQ treatment on the number of hemocytes in DDR-deficient flies (Figure 4B, Figure S5A).

Next we analyzed if the loss of the DDR signaling machinery via knockdown of tefu and mei41 results in increased DNA damage in hemocytes. Comet assays of isolated hemocytes from DDR deficient flies and the control genotype on PQ and control food only indicated a slight increase in DNA damage in these flies, which demonstrates that the higher susceptibility of DDR deficient flies is not due to a loss of hemocyte numbers or a substantial increase in DNA damage, but rather due to changes in the downstream DDR-signaling events modulating immune activation of hemocytes (Figure 4C). Other studies demonstrated that upon tissue damage JNK-mediated secretion of the JAK/STAT cytokines upd1, upd2 and upd3 is a critical step for innate immune signaling^{28, 29}. Furthermore, DNA damage signaling is essential to induce such an immune response and to regulate insulin signaling³⁰. To elucidate the implication of DDR signaling on hemocyte activation and oxidative stress resistance, we measured the gene expression levels of upd cytokines as well as JNK and Jak/STAT targets in DDR deficient flies and controls on PQ and control food. Loss of DNA damage signaling in hemocytes induced a significant induction of upd3 expression upon PQ treatment in Hml>tefu-IR and Hml>mei-41-IR;tefu-IR flies (Figure 4D). In contrary, upd1 and upd2 were not significantly changed in DDR-deficient flies (Figure 4D). We measured the Jak/STAT target genes Socs36E and TotA on systemic level. A slight induction of Socs36E expression in all DDR-deficient genotypes on PQ but only Hml>mei-41-IR;tefu-IR flies showed a significant induction of Socs36e upon PQ treatment compared to control food and Hml/+ flies on PQ (Figure 4D). TotA expression tended to be reduced upon PQ treatment across all genotypes analyzed. However, when we checked the levels of puc as a JNK signaling target gene, we found a significant induction on PQ treatment in Hml>tefu-IR, Hml>mei-41-IR;tefu-IR and Hml>nbs-IR flies, which was not seen in control flies upon PQ treatment (Figure 4D). When we analyzed the gene expression of genes involved in metabolism and energy mobilization, we detected a significant induction and dysregulation of Thor and InR in all DDR-deficient genotypes upon PQ treatment compared to control flies (Figure 4E), as well as a slight induction of Pepck1 (Figure 4E). To exclude that these changes in energy mobilization are due to altered expression of Ilps, we analyzed the levels of Ilp2, Ilp3 and Ilp5 in all genotypes on control and PQ-food, but could not detect remarkable alterations here (Figure 4E). To further exclude that the higher susceptibility of DDR-deficient flies and seen alterations in stress response are due to inefficient mobilization of triglycerides, we measured stored triglycerides across all genotypes and found efficient mobilization of triglyceride storage upon PQ treatment in all genotypes compared to control food (Figure 4F).

From our results we can conclude that DDR signaling in adult hemocytes, mostly consisting of plasmatocytes, essentially determines susceptibility to oxidative stress. Loss of canonical DNA damage signaling in hemocytes, which normally would induce DDR or cell death, triggers enhanced non-canonical DNA damage signaling such as upregulation of JNK signaling and elevated expression of the pro-inflammatory cytokine upd3, as well as transcriptional alterations in energy mobilization and insulin signaling. Based on our previous results on the role of hemocytes during oxidative stress, we can therefore assume that this JNK activation and upd3 induction is essentially mediated by hemocytes.

Adult hemocytes control susceptibility to oxidative stress by JNK signaling and upd3 release

To this point, our data suggests that oxidative stress-induced DDR in hemocytes limits immune activation and stress response upon oxidative stress. In other words, enhanced DNA damage boosts immune activation and therefore susceptibility to oxidative stress. As we found that increased upd3 levels correlate with a higher susceptibility to oxidative stress of flies with DDR-deficient hemocytes, we aimed to test if hemocyte-derived upd3 is a major source to regulate susceptibility to PQ-induced oxidative stress. To do so, we either over-expressed upd3 in hemocytes in Hml>upd3 (w;Hml-Gal4,UAS-2xeGFP/UAS-upd3) flies or performed a hemocyte-specific knockdown of upd3 in Hml>upd3-IR (w;Hml-Gal4,UAS-2xeGFP/UAS-upd3-IR) flies. Both genotypes showed a normal overall lifespan compared to the Hml/+ control (Figure S5F), but when we compared their susceptibility to PQ-induced oxidative stress, Hml>upd3 flies showed a drastically decreased survival on PQ, whereas Hml>upd3-IR flies did not show a changed survival on PQ treatment after 18 hours (Figure 5A). We tested upd3 expression levels on a whole fly level and found a slight but not significant induction in Hml/+ flies after PQ treatment compared to control food, in line with our previous results (Figure 1B and 5B). As expected, Hml>upd3 flies showed increased levels of upd3 in both treatment groups compared to the other genotypes, but also showed a significant increase of upd3 upon PQ compared to control food (Figure 5B). Surprisingly, Hml>upd3-IR flies still showed a slight induction of upd3 upon PQ treatment, which could potentially be derived from another cellular source after eliminating hemocyte-derived upd3. To identify if the higher susceptibility of Hml>upd3 flies to oxidative stress was due to reduced hemocyte numbers, we quantified hemocytes across all three genotypes, but did not see any significant changes of hemocyte number in both genotypes compared to Hml/+ control flies on control food and PQ (Figure 5C and Figure S5G). Only in the Hml>upd3-IR flies we found a reduction of hemocytes between flies on control food and PQ, but without consequences for their susceptibility to oxidative stress (Figure 5C and Figure S5G). Our data indicate that elevation of hemocyte-derived upd3 is sufficient to increase the mortality of adult flies on PQ-mediated oxidative stress. Yet we could not exclude other cellular sources for upd3 upon oxidative stress other than hemocytes. Therefore, we wanted to probe if hemocyte-derived upd3 is sufficient to increase the susceptibility to PQ treatment in the absence of other cellular sources. We backcrossed Hml>upd3 flies to upd3-deficient flies (upd3^null) to generate upd3^null;Hml>upd3 flies (upd3^null;Hml-Gal4,UAS-2xeGFP/UAS-upd3) and compared their survival on PQ to different control genotypes (w¹¹¹⁸ and Hml/+) and upd3^null flies (Figure 5D). We did not identify any differences in the susceptibility of w¹¹¹⁸, Hml/+ and upd3^nullcontrol flies to PQ treatment, however we found a significantly decreased survival in upd3^null;Hml>upd3 flies when hemocytes are overexpressing upd3 (Figure 5D).

Figure 5:

Previous studies highlighted that JNK signaling can induce upd3 expression in hemocytes¹². In line with these reports, we found enrichment of genes with kayak/AP-1 binding motifs in the plasmatocyte subset C6 which showed further upregulation of Jak/STAT target genes and also upd3 expression (Figure 3). Therefore, we wanted to verify if JNK activation in hemocytes acts as a direct downstream signal activated by non-canonical DNA damage signaling upon oxidative stress and subsequently induces the release of the pro-inflammatory cytokine upd3. We either overexpressed a constitutive active form of JNKK hemipterous (hep) in hemocytes using Hml>hep[act] (w;Hml-Gal4,UAS-2xeGFP/UAS-hep[act]) flies or overexpressed a dominant-negative version of the Jun kinase basket (bsk) in hemocytes using Hml>bsk-DN (w;Hml-Gal4,UAS-2xeGFP/UAS-bsk-DN). We compared their survival on PQ after 18 hrs and in line with our previous results, found an increased susceptibility of Hml>hep[act] flies to PQ treatment, but not of Hml>bsk-DN flies compared to the control genotype (Figure 5E). We again analyzed the overall lifespan of these flies on normal food and could not find differences between the genotypes (Figure S5H). Furthermore, when we analyzed upd3 expression in these flies upon PQ treatment, we found a significant induction of upd3 expression after PQ treatment in Hml>hep[act] compared to control food, but only a slight increase in Hml/+ and Hml>bsk-DN flies upon PQ feeding. These findings support a substantial role of JNK-mediated induction of upd3 in hemocytes upon oxidative stress. In summary, our data discovers a defined immune-responsive subset of plasmatocytes during oxidative stress. We show that this oxidative stress-mediated immune activation seems to be controlled by non-canonical DNA damage signaling resulting in JNK activation and subsequent upd3 expression, which can render the adult fly more susceptible to oxidative stress when it is over-activated.

Discussion

Similar to vertebrate macrophages, oxidative stress triggers immune activation in Drosophila hemocytes¹⁸. However this immune activation was not obvious on a systemic level upon PQ-mediated oxidative stress. In contrast, localized oxidative stress-induced immune activation in damaged tissues is an essential step of the Drosophila immune response to infection or injury^{19, 31}. However, the precise contribution of Drosophila macrophages in this immune activation, in the organ-to-organ communication during stress response and the implication in the organism’s survival upon increased oxidative stress remained so far ill-defined. Overall, we describe responses of adult hemocytes, namely plasmatocytes, to oxidative stress with a particular subset of plasmatocytes (cluster C6) which directly responds to oxidative stress by immune activation. We demonstrated that this immune activation seems to be at least partially controlled by non-canonical DNA damage signaling in hemocytes. Loss of nbs, tefu or mei-41 in hemocytes renders the organism more susceptible to oxidative stress without signficantly affecting oxidative stress mediated DNA damage and its repair, but with an enhanced induction of JAK/STAT and JNK signaling, accompanied by changes in genes associated with energy mobilization and storage. Finally, we showed that the control of hemocyte-derived upd3 is key for the survival of the organism upon oxidative stress and that induction of the JNK/upd3 signaling axis in hemocytes is sufficient to increase the susceptibility to oxidative stress. Hence, we postulate that hemocyte-derived upd3, most likely released by the activated plasmatocyte cluster C6 during oxidative stress in vivo and subsequently controlling energy mobilization and subsequent tissue wasting upon oxidative stress.

Drosophila hemocytes have been described before to react to PQ-mediated or infection-mediated intestinal damage via release of cytokines to promote ISC proliferation in the gut^{14, 17}. PQ is a widely used model in Drosophila to induce oxidative stress via oral uptake³². In many reports the gut has been described to be the most affected organ and ISC proliferation is an essential step to regenerate the injured intestinal epithelium^{33, 34}. PQ treatment has been described to cause gut leakage and therefore this could be a potential cause of death for the fly¹⁴, though we did not find signs of gut leakage within the 18 hours PQ treatment with a concentration of 15mM. Aside from the gut, oral PQ treatment was shown to induce oxidative stress in many organs including hemocytes³⁵, oenocytes³⁶ and the brain³⁷, indicating that the induced oxidative stress is not only localized to the gut tissue. PQ-induced oxidative stress was described before to cause neurodegeneration with parkinsonism-like symptoms, accompanied by strong immune activation and upregulation of JNK signaling in the CNS³⁷, an activation profile similar to the identified activated plasmatocyte cluster C6 in our snRNA-seq analysis. The snRNA-seq analysis further revealed distinct hemocyte clusters with seven plasmatocyte clusters and one crystal cell cluster. Whereas the plasmatocyte clusters C1 and C7 seem to represent cellular states found in steady state conditions only, we identified several other clusters which seemed to be exclusive or at least enriched under PQ treatment. Interestingly, we only identified one cluster, C6, which specifically responded to oxidative stress via immune activation and upregulation of oxidative stress response genes but also Jak/STAT and JNK target genes. We postulate that these cells might be directly exposed to the oxidative stress because their signature resembles the one of S2 cells directly treated with PQ in vitro. Even though oxidative DNA damage in hemocytes was followed by immune activation in vitro and in vivo, it remains to be determined if this activation is induced by cytokines derived from damaged organs during oxidative stress or by the oxidative stress directly. The hypothesis that this cluster is directly exposed to oxidative stress is further supported by an upregulation of genes such as Hsps in this cluster which are known to be directly induced by oxidative stress³⁸. Furthermore, we found evidence that oxidative DNA damage and induced non-canonical DNA damage signaling induce several of these immune activation cascades in cluster C6 plasmatocytes. This was supported by the increased DNA damage in S2 cells by PQ treatment but also the enrichment of DDR-associated genes such as Gadd45 in cluster C6, which is associated with JNK activation and is further implicated in oxidative stress induced DNA damage³⁷. Constant DNA damage signaling and repair is an essential step in all tissues to maintain tissue homeostasis³⁹.

In our study, we point to a potential non-canonical role of DNA damage sensing to trigger immune activation and response to oxidative stress in hemocytes. We did not find indications of canonical DNA damage response, such as induced apoptosis in hemocytes due to the oxidative DNA damage. Non-canonical DNA damage sensing has been implicated in immune responses in vertebrates and invertebrates^{30, 37, 40, 41}. It was demonstrated that vertebrate macrophages in vitro respond similarly to immune stimuli such as infection by synergistic activation of DDR and type I interferon signaling to induce downstream immune signaling cascades²⁷. Here, we provide new evidence in Drosophila that non-canonical DNA damage is used by immune cells in vivo to modulate and especially control their immune response including pro-inflammatory cytokine release upon oxidative stress. Localized DNA damage in larval epithelial cells has been shown to induce non-canonical DNA damage activity leading to the secretion of upds and the regulation of hemocyte expansion and activation, as well as subsequent hemocyte-mediated regulation of energy storage in the fat body³⁰. In further support of our results, one report previously demonstrated that loss of DNA damage signaling machinery induced the release of upd3 in hemocytes¹⁴. Most likely, the identified plasmatocyte cluster C6 is a central cellular source of this signal, which needs to be further determined in future studies. Furthermore, anatomical localization of this cluster and spatial distribution of the different plasmatocyte clusters would be of interest. We describe a strong connection between the control of upd3 expression by hemocytes and the susceptibility of the adult fly to oxidative stress. Upon oxidative stress, we observe decreased triglycerides and glycogen stores and altered gene expression such as induced expression of lipolytic enzymes and other foxo target genes indicating energy mobilization from the adult fat body. Accumulating evidence throughout our study points to an implication of hemocyte-derived upd3 in the control of energy mobilization and tissue wasting during oxidative stress. These findings are in line with our previous studies where we showed that upon high fat diet hemocytes produce high levels of upd3 which induces insulin insensitivity in many tissues by activating Jak/STAT signaling in these organs and inhibiting the insulin signaling pathway¹². Furthermore, we and others described a central role for hemocyte derived upds in the control of glucose metabolism in muscles during steady state and upon infection^{13, 42}. As shown before, hemocytes produce high levels of upd3 upon infection with Mycobacterium marinum which also caused tissue wasting and extensive energy mobilization in a foxo-dependent manner^{11, 43}.

Our study provides further evidence that hemocytes serve as essential signaling hubs in organ-to organ communication during oxidative stress, which is in line with previous immune challenge as well as steady state^{19, 42, 44}. Our findings in Drosophila highlight the role of macrophages for the survival on oxidative stress with an urgent need for the tight control of their immune activation. We further unravel the diverse responses of hemocytes towards oxidative stress which can now serve as a baseline to further explore the spatiotemporal macrophage/hemocyte responses to oxidative stress. Future in depth analysis of the regulation of these cellular states and their coupling to tissue repair will allow to gain new insights for the treatment of diseases including cancer or neurodegenerative diseases.

Acknowledgements

Acknowledgements and Funding

The authors are very grateful to the members of the Kierdorf lab for their help and support. The authors would especially like to thank M. Oberle for her exceptional help with the maintenance of the Drosophila stocks and the management of the fly room. We would like to acknowledge the Lighthouse Core Facility and its staff, especially J. Bodinek-Wersing and U. Jagadeshwaran, for their assistance with the sorting of nuclei. The authors are especially grateful to D. Ganser and H. Fischer for the continuous support throughout the study. Graphical abstract was created with BioRender.com. KK was supported by project grants of the Fritz Thyssen Foundation and of the German Research Foundation (DFG). MSD was supported by Wellcome Trust (207467/Z/17/Z) and Biotechnology and Biological Sciences Research Council (BBSRC) (BB/W/001004/1). KK, MP, and OG were supported by the DFG through project grants within SFB/TRR167 (Project ID 259373024) and by project grants within CRC1479 (Project ID 441891347). KK, MP, OG, and AC were supported by the DFG under Germany’s Excellence Strategy (grant no. CIBSS—EXC-2189, Project ID 390939984). AC was further supported by the Boehringer Ingelheim Foundation (Plus3 Programme). OG was further supported by the European Research Council (ERC) through Starting Grant 337689 and Proof-of-Concept Grant 966687, and by the DFG through project grants in GRK2606 (Project ID 423813989), SFB1425 (Project ID 422681845), and SFB1160 (Project ID 256073931), as well as by the EU-H2020-MSCA-COFUND EURIdoc programme (No. 101034170). MP was further supported by the Novo Nordisk Foundation, the Ernst Jung Foundation, the DFG through project grants in SFB992 (Project ID 192904750), and SFB1160 (Project ID 256073931), a Reinhart Koselleck grant and the Gottfried Wilhelm Leibniz prize, the Ministry of Science, Research and Arts of the state of Baden-Wuerttemberg (Sonderlinie ‘Neuroinflammation’) and by the Alzheimer Forschung Initiative e.V. (AFI). GM was supported by a Marie Skłodowska-Curie postdoctoral fellowship.

Author Contributions

FH contributed to project planning, performed all experiments, analyzed data, edited the manuscript and prepared the figures. TM and PW performed experiments, helped with the data analysis, and edited the manuscript. CC performed the snRNA-seq experiment and bioinformatics analysis. GM performed data preprocessing and supported bioinformatics analysis. AC, MSD, KP, OG, and MP provided equipment and reagents, helped to analyze and interpret data, and edited the manuscript. KK planned and supervised the project, conceptualized the study and wrote the manuscript.

Competing Interests Statement

The authors declare no competing financial or scientific interests.

Methods Section

Drosophila melanogaster stocks

Flies were reared on a high yeast food containing 10% brewer’s yeast, 8% fructose, 2% polenta and 0.8% Agar. Propionic acid and nipagin were added to prevent bacterial or fungal growth. All crosses (except those containing the tub-Gal80^ts construct or otherwise noted) were performed at 25°C with a 12 hours dark/light cycle. The crosses with tub-Gal80^tswere performed at 18°C to ensure the inhibition of the Gal4-protein during developmental stages of the fly. The experimental male F1 flies were transferred to 29°C as soon as they hatched and were aged for six days. All transgenic lines used in this study are listed in the Table S2.

Lifespan/Survival assays

Male flies were collected after eclosion and groups of 20 age-matched flies per genotype were housed in a food vial. The survival experiments were performed at 29°C. The vials were placed horizontal to avoid that flies fall into the food and become stuck. Dead flies were counted in a daily manner. The flies were transferred into a fresh food vial twice per week without CO₂ anesthesia.

Paraquat treatment

Flies were maintained at 29°C for six days prior to treatment. On day six they were starved for 6 hours. A filter paper soaked in 5% sucrose solution with or without 15mM Paraquat (PQ, Methyl viologen hydrate, Acros Organics) was added after starvation. Flies were fed for 18 hours on the PQ containing food in groups of ten. The PQ treatment was performed at 29°C in the dark²⁰. To assess the survival rate the dead flies per vial were counted and the percentage was calculated. The living flies were further analyzed in other assays used in this study.

Starvation experiments

10-20 age-matched male flies were kept in a vial containing 1% agar supplemented with 2% 1xPBS. The starvation experiments were performed at 25°C. Dead flies were counted every hour. Several individual experiments were performed and started on different daytimes to exclude diurnal derived artifacts. The data of the individual experiments were pooled and analyzed with GraphPad Prism 9.2.0.

Paraffin sections and stainings

Anaesthetized flies were washed in 75% ethanol for 5-10 seconds and transferred into 4% PFA for 30 minutes at room temperature. The flies were washed for 1 hour in PBS on a shaker. Subsequently the flies were transferred into tissue cassettes, dehydrated and embedded in paraffin. The flies were cut in 7µm thick sections de-paraffinated and stained on slides with Hematoxylin-Gill (II) for five minutes. Subsequently the sections were blued in running water for ten minutes. It followed a 0.5% Eosin treatment for five minutes and a final ethanol treatment with increasing concentrations until 100%. The sections were transferred into Xylol and finally covered with synthetic resin and a cover slip.

Semi-quantitative Real-Time PCR

Three flies were pooled and smashed in 100µl TRIzol to stabilize and isolate the RNA. Chloroform was used to extract the RNA followed by an isopropanol precipitation step. The RNA was cleaned with 70% ethanol and solubilized in water. A DNAse treatment was performed to digest potential genomic DNA contaminations. The purified and DNAse treated RNA was written into cDNA using RevertAid Reverse Transcriptase (Thermo Scientific) at 37°C for one hour. The reaction was stopped by incubating for 10 minutes at 70°C. The subsequent RTqPCR was done in SensiMix SYBR Green no-ROX (Meridian Bioscience) and was performed on a LightCycler 480 (Roche). The qPCR cycling started with a 10-minute 95°C step followed by 50 cycles with the following times and temperatures: 15s at 95°C, 15s at 60°C and 15s at 72°C. The gene expression levels were normalized to the value of the measured loading control gene Rpl1.

Confocal microscopy

The flies were anesthetized with CO2, glued on a cover slip and imaged immediately. Confocal microscopy was done with a Leica SP8 microscope (Leica) and a 10x/NA0.4 Leica air objective. The images were acquired in a resolution of 512x512 with a scan speed of 600 Hz. Z-Stacks with a step size of 5µm and tile scans (2x2 or 2x3 images per fly) were acquired to quantify the hemocytes in whole flies. The tiled images were merged with the LAS-X software (Leica) and maximum projections were created using Fiji/ImageJ. The hemocytes were counted in the maximum projection images of whole flies.

Smurf Assay

Smurf assays were performed to check the feeding behavior as well as the gut integrity of 7 day old, male flies upon PQ treatment. Brilliant Blue dye (FCF) was added into 5% sucrose solution in a concentration of 1% (w/v). PQ was added in concentrations of 2mM, 15mM and 30mM. Control food was 5% sucrose and 1% Brilliant Blue without PQ. The flies were fed with blue food according the PQ-treatment protocol, as described above. Flies were analyzed after 18h of feeding on brilliant-blue food. Three kinds of flies were distinguished. Flies that did not eat at all were excluded. Flies with a blue gut or crop were classified as “non-smurf”. Flies which turned completely blue or showed distribution of blue dye outside the gut were classified as “smurf”.

Thin Layer Chromatography (TLC)

Each sample in the TLC analysis contains 10 flies of the respective genotype. The flies were starved for one hour before they were anesthetized with CO2 and transferred into 100µl chloroform:methanol (3:1) on ice. The flies were centrifuged for 5 minutes at 15,000g at 4°C and subsequently homogenized with a pestle. The sample was centrifuged again with the same settings. Lard was dissolved in chloroform:methanol (3:1) and served as triglyceride control (standard 1). Standard 1 was used to produce a standard curve with decreasing triglyceride concentrations. The samples and the standard curve were loaded on a glass silica gel plate (Millipore). The mobile phase in the TLC chamber was made with hexane:diethylether (4:1). The plate was placed in the chamber until the solvent front reached the upper end of the plate. The plate was taken out, air dried and stained with ceric ammonium heptamolybdate (CAM). Subsequently the plate was incubated at 80°C for 2 hours to visualize the stained triglyceride bands. Images were taken with a gel documentation system (gelONE). Triglyceride density was measured and the concentrations were calculated according the pre-determined concentrations of the standard curve. Image analysis and calculations were done using ImageJ.

Glucose Assay

Seven day old male flies were used for the analysis. They were treated for 18h with PQ or sucrose respectively. Subsequently the flies were starved for one hour to bring all flies into the same metabolic state. Three flies were pooled as one sample and manually homogenized in 75µl TE + 0.1% Triton X-100 (Sigma Aldrich). The samples were incubated for 20min at 75°C to inactivate any intrinsic enzymatic activity. 5µl of the samples were loaded into a flat-bottom 96-well plate. Each sample was measured four times to measure glucose, trehalose, glycogen and the fly background respectively. The fly background was determined by diluting the fly sample with water. Glucose reagent (Sentinel Diagnostics) was added to the fly sample to measure the glucose levels. Trehalase (Sigma Aldrich) or amyloglucosidase (Sigma Aldrich) was added to the glucose reagent to measure trehalose or glycogen respectively. Plates were incubated for 24h at 37°C before the colorimetric measurement was performed at a wavelength of 492nm on a Tecan® Spark plate reader. The concentrations were calculated according to standards loaded on the same plate.

Nuclei FACS sorting

100 male Hml-dsRed.nuc flies per sample were homogenized in 500µl ice-cold EZ nuclei lysis buffer (Sigma). Another 500µl EZ lysis buffer was added and the nuclei were incubated for five minutes on ice. Following a pre-filtration step with a 70µm cell strainer (Miltenyi filters) the nuclei were centrifuged at 500g for six minutes at 4°C. The pellet was resuspended and incubated in 1ml of EZ lysis buffer for five minutes on ice following a second centrifugation step with previous used settings. After discarding the supernatant, 500µl of nuclei wash and resuspension buffer (1x PBS; 1% BSA; 0.2U/µl RNase inhibitor, NEB) was added without resuspending the pellet and incubated for 5 minutes on ice. Subsequently the pellet was resuspended with additional 500µl nuclei wash and resuspension buffer. The nuclei were centrifuged using previous settings and subsequently washed with 1ml of nuclei wash and resuspension buffer. After centrifugation the pellet was resuspended and incubated for 15 minutes with nuclei staining solution containing DAPI (1:1000) and DRAQ7 (1:100). The nuclei were washed once with 1ml nuclei wash and resuspension buffer. After centrifugation the pellet was dissolved in 200µl wash and resuspension buffer, filtered through a 40µm filter and sorted on a BD FACS ARIAIII. A duplet exclusion was performed by gating on singlets in FSC-A vs. FSC-W plot. Subsequently DRAQ7+DAPI+ events were sorted into nuclei wash and resuspension buffer.

Single nucleus library preparation and sequencing

For single nucleus library preparation, nuclei were loaded onto a Chromium Single Cell 3′ G Chip (10X Genomics) to generate single-nucleus gel beads in emulsion (GEM) according to the manufacturer’s instructions without modifications using Chromium Next GEM Single Cell 3’ Reagents Kit v3.1 (10X Genomics). In order to multiplex the samples for sequencing, Single Index Kit T Set A (10x Genomics) was used for library preparation according to the manufacturer’s instructions. The cDNA content and size of post-sample index PCR samples was analyzed using a 2100 BioAnalyzer (Agilent). Library quantification was done using NEB Next® Library Quant Kit for Illumina® (New England Biolabs) following manufacturer’s instructions. Sequencing libraries were loaded on an Illumina Nextseq 550 flow cell, with sequencing settings according to the recommendations of 10× Genomics. Sample de-multiplexing was done using built-in BCL2FASTQ. Cell Ranger v6 software was implemented for gene alignment to the fly genome. The Drosophila melanogaster genome assembly and annotation file were downloaded from Ensembl (Release 104). The annotation files were filtered as suggested by 10x Genomics, to keep only the categories of interest (i.e. protein coding genes, long intergenic noncoding RNAs, antisense, pseudogenes). Cell Ranger v6 with the “include-intron” parameter was used for the generation of the count data.

snRNA-seq data analysis

Downstream analysis implemented the SoupX v1 and Seurat v4 R-based packages. The matrix for each sample was loaded using the SoupX package⁴⁵ and ambient RNA contamination was calculated using the default setting. The corrected matrix together with metadata on treatment were used to create an object in Seurat v4²³. Doublet detection was done using the DoubletFinder package v2⁴⁶. After doublet exclusion, a combined object was created from the list by using the “merge” function. Nuclei with < 5% of mitochondrial contamination and between 400 and 1,700 genes expressed were retained for further analysis. The filtered raw count matrix was log-normalized within each nucleus. The top 3,000 variable genes were calculated by Seurat using the variance stabilizing transformation selection method and data were scaled. The variable genes were used to perform principal component analysis (PCA) and the top 20 principal components were used for the unsupervised clustering. Seurat applies a graph-based clustering approach: the “FindNeighbors” function uses the number of principal components (dims = 20) to construct the k-nearest neighbors graph based on the Euclidean distance in PCA space, and the “FindClusters” function applies the Louvain algorithm to iteratively group nuclei (resolution = 1). We then used the PCA embedding to generate an UMAP for cluster visualization. For each cluster, differentially expressed genes (DEGs) were calculated using the “FindMarkers” and “FindAllMarkers” function with default parameters with adjusted P < 0.05 using Bonferroni correction. The DEGs were used to assign cell type identity to clusters based on known cell lineage markers described in previously published study²⁵. To understand the impact of Paraquat treatment on each cluster, “FindMarkers” function was used to compare “control” vs “Paraquat” treated cells per cluster. Based on the DEGs obtained from this analysis, volcano plots were created using the EnhancedVolcano package v1⁴⁷. UMAP visualizations, Feature plots and dot plots for snRNA-seq data were generated using in-built plotting functionality of Seurat.

Subclustering of hemocytes and further analysis

Based on the cell identity and known hemocyte marker genes from a previous study²⁵, sub-clusterings were performed using the same unsupervised algorithm and downstream analysis. After removing non-hemocyte clusters, final sub-clustering was done with a modification that “FindNeighbors” function was run with dims = 10 and “FindClusters” function was run with resolution 0.8. Based on the DEGs found by the “FindAllMarkers” function (parameters: only.pos = T, test.use= “MAST”) a heat map was plotted for TOP 20 DE genes per clusters.

Regulation network analysis of hemocyte clusters

In order to determine the transcription factor activity, regulon analysis was performed using the R package SCENIC (v1.3.1)⁴⁸. To remove noise, genes with low expression levels or low positive rates were filtered using the “geneFiltering” function with default settings to obtain a filtered matrix which was used to build co-expression network using the “runCorrelation” and “runGENIE3” functions. Gene regulatory networks (GRNs) were built and scored using the default parameters. Potential regulons based on DNA-motif analysis were selected by using RcisTarget and active gene networks were identified by AUCell. Regulon activity for each cell was calculated as the average normalized expression of putative target genes. Regulon activity matrix was exported which was used to create a new ‘AUC’ assay using “CreateAssayObject” function of Seurat. “DoHeatmap” function from Seurat was used to plot scaled TF activity in individual hemocytes.

S2 Cell Culture, CellROX staining and flow cytometry

Drosophila S2 cells (Invitrogen™) were cultured in Schneider’s Drosophila medium (Gibco™) at 26°C and atmospheric oxygen and carbon dioxide conditions. The medium contained 10% FCS and 1% penicillin/streptomycin. PQ was added into the medium in concentrations of 15mM or 30mM to investigate the influence on the production of reactive oxygen species (ROS). Control cells remained untreated. To visualize ROS and oxidative stress, the cells were stained with CellROX™ Deep Red Reagent (Invitrogen) for 30 minutes directly in the medium in a concentration of 1:500 at 26°C. The cells were washed twice with 1xPBS containing 2mM EDTA. In order to stain dead cells DAPI was added to the samples in a concentration of 1:1000. Samples were acquired using a BD LSRFortessa™ (BD Biosciences) and analyzed with FlowJo analysis software.

COMET assays

Comet assays were performed to analyze the amount of DNA damage in S2 cells treated with PQ. S2 cells were treated with 15mM or 30mM PQ for 24h. The cells were brought into a concentration of 1.8x10⁵ cells/ml. Low melting agarose (LMA) was boiled at 90°C and subsequently cooled down to 37°C. The cells were diluted in low melting agarose (LMA) at a ratio of 1:10 (cells:LMA) and transferred onto a CometSlide™ (Trevigen). To ensure proper attachment of the LMA, the slides were cooled for 30 minutes at 4°C. The lysis was performed for one hour at 4°C. Therefore the slides were submerged in CometAssay® Lysis Solution (Trevigen). Subsequently the cells were submerged in alkaline unwinding solution (200mM NaOH, 1mM EDTA, pH 13) for 1h at 4°C. The electrophoresis was performed under alkaline conditions (pH13) for 30 minutes with a current of 20V (1V/cm). The slides were neutralized with ddH₂O, dehydrated with 37% ethanol and dried properly at 37°C before they were stained with SYBR™Gold (Invitrogen) for 30 minutes. Comets were imaged with an Olympus BX61 fluorescence microscope. Comet data was analyzed via TriTrek CometScore 2.0.0.38 software.

Statistical analysis and data handling

Statistical significance of real-time qPCR, Paraquat survival, hemocyte quantification, TLC, Glucose assay, CellROX and comet assay data was calculated with an unpaired t-test or one-way ANOVA as indicated in the figure legends. Lifespan/survival assays were analyzed using Log-Rank and Wilcoxon test. Significance digits indicate the following significance levels: *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. Significance tests were performed using GraphPad Prism 9 software. Data handling of snRNAseq data is described above.

Data availability

snRNA-seq data will be publicly available as of the date of publication. Microscopy, flow cytometry and survival data reported in this paper will be shared by the lead contact upon request. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Supplementary Figure Legends

Figure S1:

Figure S2:

Figure S3:

Figure S4:

Figure S5:

Table S1:

Cluster Identity

Table S2: