Drosophila melanogaster lines used in this study are as follows. The NP3253-Gal4 and NP1-Gal4 lines were obtained from Kyoto Drosophila Stock Center. The UAS-Kir2.1::EGFP (20), tubP-Gal80ts, and UAS-mCD8::GFP lines were obtained from the Bloomington Drosophila Stock Center (BDSC). UAS-dTrpA1 was gifted by P. Garrity (21). elav-Gal80 was gifted by Y. Jan (19). Flies were reared at 18 or 25°C in plastic vials containing standard cornmeal-agar medium. To inhibit or activate neural activity, adult flies of NP3253-Gal4/tubP-Gal80ts; UAS-Kir2.1::EGFP/+ (NP3253>Kir2.1 flies) or NP3253-Gal4/UAS-dTrpA1 (NP3253>dTrpA1 flies) were reared at 29 or 30°C for 2 days before experiments. Adult flies of NP3253-Gal4/tubP-Gal80ts; UAS-mCD8::GFP/+ (NP3253>GFP flies) were used as controls. The following RNAi lines were obtained from the Vienna Drosophila Resource Center (VDRC): UAS-Cyp6d5 RNAi (#107641), UAS-Sodh-1 RNAi (#3761), and UAS-Cyp313b1 RNAi (#102986). UAS-GFP RNAi was obtained from BDSC (#9331) and used as the control. Adult flies of NP1-Gal4, UAS-XX RNAi (NP1 > XX RNAi flies) were used for the knockdown experiment. Female flies were used in all of the following experiments.

Pseudomonas entomophila (Pe) was a gift by S. Kawabata (22). The Pe bacteria were grown in LB broth (Nacalai Tesque) at 29 or 30°C overnight. The bacterial culture was centrifuged at 5,000 rpm for 10 min. The bacterial pellet was resuspended in 2.5 or 5% sucrose indicated in the experiment (bacteria solution). The cell concentration was adjusted to an OD600 of 20, 10 (for the survival assay), 5 (for the bacterial load assay), or 0.5 (for RNA-seq). For oral infection, flies were starved for 2 h at 29 or 30°C and placed in a culture vial with a filter paper containing bacterial solution for a day.

The assays were performed as previously described (23). To quantify feeding, five female flies were starved for 2 h at 30°C and fed with 5% sucrose solution containing 0.5% bromophenol blue (BPB) sodium (Sigma) for 30 min. The flies were placed in new vials containing normal cornmeal medium for 1 h to quantify the excretion rate. Subsequently, they were placed in a tube and homogenized in 50 µL Milli-Q water with a pestle. Fly extracts were centrifuged twice (14,600 x g, 2 min, 4 °C) to remove debris. The absorbance at 594 nm was measured using a NanoDrop spectrophotometer (Thermo Fisher Scientific).

To measure the Pe bacterial load, the flies were collected 24 h after Pe oral infection and were washed briefly with 70% ethanol. Subsequently, one fly was placed into a tube and homogenized using a pestle in 100 μL of LB broth. The fly extract was serially diluted, and 10 μL of the diluted extract was spotted onto LB agar plates. The plates were incubated at 30 °C overnight. The bacterial load in the original fly extract was calculated and expressed as colony-forming units per fly (CFU/fly). In the context of negative control, we observed a lack of colony formation in the flies without Pe infection. This confirmed the absence of colony formation by commensal bacteria under this condition.

To measure the commensal bacterial load, the conventional or axenic-reared flies were collected (see Axenic rearing section) and washed briefly with 70% ethanol. Five flies were placed in a tube and homogenized using a pestle in 500 μL of LB medium. One hundred microliters of the extract were spread on MRS (BD Difco) agar plates. The plates were incubated at 30 °C for 3 days, and colony numbers were counted. We usually observed numerous colonies in conventionally reared flies, but absence of colony in axenic reared flies.

For EXP1 (see Results section), NP3253>GFP (control), NP3253>dTrpA1 (neuron-activated), NP3253>Kir2.1 (neuron-inactivated), and elav-Gal80, NP3253>Kir2.1 (neuron-rescued) flies were collected. To inactivate or activate NP3253 neurons, adult female flies were reared on cornmeal medium at 29 °C for 2 days and were starved for 3 h. Then, flies were fed Pe or sucrose solution for 1 day, and the guts were dissected. As NP3253>Kir2.1 flies are susceptible to Pe infection, we used a low concentration of Pe (OD = 0.5) for RNA-seq experiment.

For EXP2 (see Results section), NP3253>GFP (control) and NP3253>Kir2.1 (neuron-inactivated) flies were collected under conventional (CV) and germ free (GF) conditions (see Axenic rearing section). To inactivate NP3253 neurons, the conventional and axenic flies were reared on normal and sterile cornmeal medium, respectively, at 30 °C for 2 days, and the guts were dissected.

For both EXP1 and EXP2, three biological replicates were prepared for each experimental condition. Total RNAs were prepared from 10-15 guts of each sample using the TRIzol reagent (Thermo Fisher Scientific). The yield and purity of the RNAs were evaluated using the NanoDrop (Thermo Fisher Scientific), Qubit (Invitrogen), and Bioanalyzer (Agilent). Sequencing libraries were prepared using a strand-specific RNA library prep kit (Agilent Technologies). The Illumina HiSeq 2000 system was used for sequencing (100 bases, single-end). Raw data of sequences were deposited on DDBJ, DRA (accession number DRA008209 and DRA012434).

Transcriptome analyses were performed using Linux or Macintosh operating systems as described previously (24). Analyses were performed using the NIG supercomputer at the National Institute of Genetics (ROIS). Adaptor sequences (Illumina TruSeq Adaptors) were removed from the read sequences (fastq files) using Cutadapt. The cleaned read sequences were mapped to the D. melanogaster reference genome (ver. 6.04) using the Hisat2 software. The number of reads was counted for each gene from the gene annotation data (ver. BDGP6.79) using HtSeq analysis. Genes showing an extremely low or high expression level (average read count was < 10 or > 100,000 for EXP1; < 3 or > 100,000 for EXP2) were eliminated from the subsequent analyses.

Read count data were analyzed using the R environment (version 3.6.0). A multidimensional scaling (MDS) analysis was performed using the edgeR package. Differentially expressed genes (DEGs) were identified using a statistical criterion (FDR-adjusted p-value was < 0.05, and fold change was > 2 or < 0.5). Z-scores of the DEGs were calculated using the Scrime package. Cluster analysis of DEGs was performed using the gplots package. The web-based databases DAVID (25) and FlyBase (https://flybase.org/) were used to analyze Gene Ontology (GO).

The expression levels of Diptericin A (DptA) and rp49 were determined using RT-qPCR as described previously (24). Adult females were reared at 29 °C for 2 days, and the guts were dissected. Total RNA was extracted from 10 guts for each sample, and three samples were analyzed as biological replicates for each experimental condition. All samples were derived from a single batch experiment. Complementary DNA (cDNA) was synthesized from total RNA using ReverTra Ace (Toyobo), according to the manufacturer’s instructions. RT-qPCR was performed using FastStart DNA Master SYBR Green I (Roche) on the LightCycler ST300 (Roche). The primers used for the analysis were DptA Fw (5′- GTTCACCATTGCCGTCGCCTTAC-3′), DptA Rv (5′CCCAAGTGCTGTCCATATCCTCC-3′), rp49 Fw (5′- AGATCGTGAAGAAGCGCACCAAG-3′), and rp49 Rv (5′- CACCAGGAACTTCTTGAATCCGG-3′). Copy numbers were calculated using the data obtained from standard plasmids carrying PCR products.

Axenic flies were prepared as previously described (26). Briefly, embryos were collected and washed with Milli-Q water and 70% EtOH. Subsequently, they were dechorionated with 2.7% sodium hypochlorite for 1 min and washed with Milli-Q water, 70% EtOH, and sterile Milli-Q water. Then, embryos were transferred to a sterilized cornmeal medium and reared until eclosion. Axenic preparation according to this method was confirmed by the absence of bacteria in the bacterial load assay, as described above.

For the survival assay of the axenic flies, adult females reared at 18°C were transferred to a sterile cornmeal medium at 30°C; surviving flies were counted for 5 days. Concurrently, conventional flies were reared on a nonsterile cornmeal medium. For RNA-seq analysis, the guts were dissected from these flies as described above.

For the survival assay, hydrogen peroxide (Kanto Chemical) was diluted to 0.6% with a 5% sucrose solution. To inhibit neural activity, 5 to 8-day-old flies were reared at 29°C for 2 days and were starved for 2 h. Subsequently, the flies were placed in a culture vial with a filter paper containing 500 μL of 0.6% H₂O₂ or 5% sucrose solution for 3 days (exchanging fresh vials daily). After feeding, flies were transferred to a normal or sterile cornmeal medium, and the surviving flies were counted for the following 3 days.

R program (http://www.r-project.org/) (version 3.6.0) was used for all statistical analyses. The bacterial loads were statistically compared using the Kruskal–Wallis analysis of variance (ANOVA), followed by the Wilcoxon rank-sum test. Gene expression analyses for RNA-seq data and RT-qPCR data were performed using ANOVA, followed by Tukey’s HSD post-hoc test.

We previously showed that NP3253> Kir2.1 flies are susceptible to the bacterium Ecc15 by oral infection. We investigated whether this susceptibility could be extended to other bacteria. Although Ecc15 is not pathogenic to the wild-type Drosophila line, we examined susceptibility to the highly pathogenic bacterium, Pe (27). In our experimental system using Gal80ts (see Materials and Methods), NP3253 neurons were inactivated by Kir2.1 expression after a temperature shift ( Figure 1A ). Two days after neural inactivation, flies were fed the bacterial solution (or sucrose solution) for 1 day, and dead flies were counted daily for 10 days. As expected, Pe was highly toxic even in NP3253 >GFP flies (hereafter referred to as control flies). In addition, the survival rate under Pe infection was significantly lower in NP3253>Kir2.1 flies than in control flies ( Figure 1B ). These results suggest that NP3253>Kir2.1 flies are susceptible to oral bacterial infections, probably regardless of the bacterial species.

Next, we examined the phenotype of NP3253>dTrpA1 flies, in which NP3253 neurons were hyperactivated. Under Pe infection, the survival rate of NP3253>dTrpA1 flies was significantly higher than that of the control flies ( Figure 1C ). This suggests that manipulating NP3253 neurons may positively or negatively affect susceptibility to oral infection. To examine feeding behavior, we measured food intake and excretion of BPB-containing foods. We detected an increase and decrease in the amount of BPB in fly bodies after feeding and excretion, respectively. There was no difference in the amount of BPB between NP3253>dTrpA1 and control flies ( Figure 1D ), indicating that hyperactivation of NP3253 neurons did not alter feeding behavior. We also measured the bacterial load in fly bodies 24 h after oral infection with Pe. The Pe load was higher in NP3253>Kir2.1 flies than in control flies, suggesting that the activity of NP3253 neurons is required to suppress the growth of Pe in flies. In contrast, NP3253>dTrpA1 flies showed a similar Pe load as that of control flies, suggesting that hyperactivation of NP3253 neurons does not further suppress the growth of Pe. We hypothesized that NP3253 neurons contribute to tolerance (protecting the host from pathogen-induced damage) and resistance (suppressing bacterial growth) to oral infection.

To investigate how NP3253 neurons control gut physiology during oral infection, we performed RNA-seq analysis of the gut. Adult female flies were fed Pe or sucrose solution for 1 day. The guts were dissected from the following seven fly conditions (feeding conditions + genotypes): Pe- or sucrose-fed NP3253>GFP (control) flies, Pe- or sucrose-fed NP3253>dTrpA1 (neuron-activated) flies, Pe- or sucrose-fed NP3253>Kir2.1 (neuron-inactivated) flies, and sucrose-fed elav-Gal80, NP3253>Kir2.1 (neuron-rescued) flies. Total RNAs were extracted from guts and subjected to RNA-Seq analysis ( Table S1 , Data S1 ). This analysis was termed EXP1 because we performed an additional RNA-seq analysis, as described later.

MDS analysis was performed to examine the similarity of overall gene expression between samples ( Figure 2A ). Replicate samples under the same fly conditions were plotted closely, indicating the validity of our RNA-Seq. The Pe- and sucrose-fed samples were plotted closely for each fly genotype, suggesting that the effect of Pe infection was relatively low. When comparing fly genotypes, NP3253>Kir2.1 flies showed more significant changes from control flies than NP3253>dTrpA1 flies. This finding suggests that inhibition of NP3253 neurons may have a more significant effect on the gut than neuronal activation.

We hypothesized that elav-Gal80, NP3253>Kir2.1 flies would show similar gene expression to the control flies because neuronal inactivation is rescued by elav-Gal80. However, flies of this genotype showed more altered gene expression than those of the other genotypes ( Figure 2A ). We then consider the possibility that the elav-Gal80 line might have a genetic background that strongly affects gene expression. Therefore, we did not use the data from elav-Gal80, NP3253>Kir2.1 flies for overall gene expression analysis (e.g., Figure 2B ) but used them to analyze individual genes or gene groups (e.g., Figures 2C, D ).

Based on pairwise comparisons between fly conditions, DEGs were identified using the criteria: an adjusted p-value < 0.05 and the fold change > 2 or < 1/2 ( Table S2 ). In NP3253>Kir2.1 flies, expression levels of 408 and 345 genes (under sucrose- and Pe-feeding conditions, respectively) were upregulated and 217 and 408 were downregulated, compared to those in the control. In NP3253>dTrpA1 flies, expression levels of 83 and 103 genes (under sucrose- and Pe-feeding conditions, respectively) were upregulated and 149 and 130 were downregulated, compared to those in the control. Consistent with the MDS analysis, inactivation of NP3253 neurons altered the expression of more genes than neuronal activation.

Upon comparing the sucrose- and Pe-feeding conditions, DEGs were detected as 3, 97, and 100 genes in the control, NP3253>Kir2.1, and NP3253>dTrpA1 flies, respectively. These results suggest that Pe-feeding altered the expression levels of a larger number of genes under the activation or inactivation of NP3253 neurons compared to those in the control flies. Finally, we identified 1139 DEGs in total from all pairwise comparisons.

To characterize gene expression patterns, we calculated the normalized variations (Z-scores) between fly conditions for the 1139 DEGs. Based on the heatmap of Z-scores, the DEGs were categorized into 12 cluster groups ( Figure 2B , color bars with group numbers). Group 1 included 353 genes with upregulated expressions in NP3253>Kir2.1 flies compared to those in the other genotypes ( Figure S1 ; Figure 2C shows a subset of these genes). Group 7 included 98 genes with downregulated expression in NP3253>Kir2.1 flies compared to those in the other genotypes ( Figure S1 ; Figure 2D shows a subset of them).

The other 10 groups showed different expression patterns ( Figure S1 ; see Discussion). Among the 12 cluster groups, we focused on groups 1 and 7 for the following analyses because of significant changes in gene expression in NP3253>Kir2.1 flies.

As noted in Introduction, NP3253-Gal4 flies expresses Gal4 in neural and non-neural cells. Therefore, the DEGs identified above may include genes that are unrelated to NP3253 neuronal functions. To exclude such genes, we selected DEGs from each cluster group whose expression in NP3253>Kir2.1 flies was rescued by elav-Gal80. We extracted 203 and 28 genes from groups 1 and 7 and termed them groups 1-2 and 7-2, respectively ( Figures 2C, D , S2 ). Thus, the expression of these genes is regulated by NP3253-Gal4 positive neural cells.

Next, we performed GO analysis to examine the functions of the genes in each group. The 203 genes with upregulated expression levels in NP3253>Kir2.1 flies (group 1-2: Data S2 ) were enriched in the GO terms of innate immunity (in the BP category) ( Figure 2E ) and include the genes encoding AMPs and Toll signaling factors. We observed that the expression levels of these genes (e.g., AttB, AttA, Dro, and CecA1 genes) were upregulated in NP3253>Kir2.1, with or without Pe infection ( Figure S3 ). This suggests that the NP3253>Kir2.1 flies have a persistent inflammatory state in the intestine.

The 28 genes with downregulated expression levels in NP3253>Kir2.1 flies (group 7-2: Data S3 ) were enriched in the oxidoreductase and monooxygenase GO terms (in the MF category) ( Figure 2F ). These genes include the cytochrome P450 (Cyp) and glutathione S-transferase (GST) families, which are related to redox responses in cells. Previous studies have shown that expression levels of redox-related genes were upregulated during Pe oral infection in adult Drosophila (28). We observed that expression levels of these genes (Cyp313b1, Cyp6d5, and Sodh1) were downregulated in NP3253>Kir2.1 flies under sucrose- and Pe-feeding conditions ( Figure S3 ), suggesting that the redox response of the intestine may be impaired in NP3253>Kir2.1 flies.

Transcriptome analysis revealed that the immune reaction in the gut was activated in NP3253>Kir2.1 flies even without pathogen infection (group 1-2). We hypothesized that immune activation in NP3253>Kir2.1 flies might be induced by commensal bacteria in the gut, which are usually harmless in flies. To evaluate this hypothesis, we prepared GF axenic flies. First, we measured the expression levels of DptA, which encodes an AMP, under CV and GF conditions. Our RT-qPCR analysis showed that the expression of the DptA gene under CV conditions was increased in NP3253>Kir2.1 flies compared to that in the control flies; this increased expression was significantly suppressed by GF conditions ( Figure 3A ). In the control flies, there was no significant difference in DptA expression between the CV and GF conditions. Next, we performed survival assays and found that the survival rate of NP3253>Kir2.1 flies decreased to approximately 50% for 5 days, even without pathogen infection under CV conditions; however, it was restored under GF conditions ( Figures 3B, C ). These results suggest that NP3253>Kir2.1 flies are susceptible to microbiota and induce immune activation in the gut in a microbiota-dependent manner.

Furthermore, transcriptome analysis revealed that expression levels of certain genes related to redox reactions were downregulated in NP3253>Kir2.1 flies (group 7-2). We hypothesized that ROS sensitivity may be elevated in NP3253>Kir2.1 flies. To test this possibility, we measured the survival rate of NP3253>Kir2.1 flies under ROS stress conditions. NP3253>Kir2.1 flies treated with H₂O₂ showed a lower survival rate than the control flies under CV conditions ( Figure 3D ). Subsequently, we performed the same assay under GF conditions and found that NP3253>Kir2.1 flies were still vulnerable to H₂O₂ ( Figure 3E ). Thus, the ROS sensitivity of NP3253>Kir2.1 flies was independent of the gut commensal bacteria. In addition, we observed that ROS sensitivity in NP3253>Kir2.1 flies was rescued by elav-Gal80 ( Figure 3F ). These results indicate that NP3253 neurons regulate ROS sensitivity in the gut in a microbiota-independent manner.

Immune- and redox-related genes were identified as DEGs between NP3253>Kir2.1 and control flies in the RNA-seq analysis (EXP1). Next, we investigated whether the expression of these genes was microbiota-dependent and performed another RNA-seq analysis of the gut using axenic flies (referred to as EXP2). Samples were prepared under the following four fly conditions (genotypes + rearing conditions): control (NP3253>GFP) and neuron-inactivated (NP3253>Kir2.1) flies reared under CV and GF conditions. Total RNAs were extracted from guts and subjected to RNA-Seq analysis ( Table S3 , Data S4 ).

The MDS plot of EXP2 showed that the overall gene expression patterns were similar between replicate samples under the same fly conditions. The GF-reared samples showed less variation between replicates than the CV-reared samples for both genotypes ( Figure 4A ). We performed pairwise comparisons between genotypes and rearing conditions and identified DEGs using the criteria same as those used for EXP1. Comparisons between genotypes (NP3253>Kir2.1 vs. control flies) identified 1,727 and 979 genes as DEGs under CV and GF conditions, respectively ( Figure 4B ). Comparisons between the CV and GF conditions identified 255 and 603 genes as DEGs in NP3253>Kir2.1 and control flies, respectively. These results suggest that gene expression in the gut is considerably altered in response to commensal gut bacteria. This change is lesser in NP3253>Kir2.1 flies than in control flies, suggesting that NP3253 neurons may contribute to responses of the gut to commensal bacteria.

To determine whether the DEGs identified in EXP1 were responsive to commensal bacteria, we analyzed the EXP2 data for the genes of each cluster group of EXP1 ( Figure 4C ). As seen in EXP1 ( Figure 2C ), 203 genes of group 1-2 showed higher expression in NP3253>Kir2.1 flies than in control flies, under CV conditions in EXP2 ( Figure 4D ). However, under GF conditions, their expression in NP3253>Kir2.1 flies was not significantly different from that in control flies. As immune-related genes were mostly enriched in this group, we observed the expression patterns of some AMP genes. Consequently, expression levels of these genes (e.g., AttB, AttA, Dro, and CecA1) were upregulated in CV-reared NP3253>Kir2.1 flies, and this upregulation was suppressed in GF-reared flies ( Figures 4E-H ). These results suggest that the inactivation of NP3253 neurons enhances the immune responses to gut commensal bacteria.

In EXP1, 28 genes in group 7-2 showed lower expression in NP3253>Kir2.1 flies than in the control ( Figure 2D ). In EXP2, similar expression patterns were observed for these genes under CV and GF conditions ( Figure 5A ). This result suggests that this group of genes related to redox reactions is regulated by NP3253 neurons independent of the gut microbiota. For example, Cyp6d5 and Cyp313b1 expression was downregulated in NP3253>Kir2.1 flies compared to that in the control flies, under CV and GF conditions ( Figures 5B, C ). Sodh-1 was listed as a DEG in EXP1 ( Figure S3 ) but not in EXP2 ( Figure 5D ), probably because of large variance in the data.

We hypothesized that some redox-related genes contribute to gut immunity during pathogenic infections. To test this hypothesis, we measured the survival rates of knockdown flies after Pe infection. For the knockdown experiments, we used the NP1-Gal4 line, which induces RNAi for each gene in intestinal epithelial cells (29). The results showed that knockdown of Cyp6d5 and Cyp313b1 reduced the survival rate after Pe infection ( Figures 5E, F ), while under sucrose-feeding conditions, these flies were almost as alive as the control flies (NP1>GFP RNAi; Figure S4 ). In contrast, knockdown of the Sodh-1 did not affect survival rates after Pe infection ( Figure 5G ), suggesting that a subset of redox-related genes contributes to Pe susceptibility. We suggest that NP3253 neurons directly regulate the expression of redox-related genes in the gut, essential for defense against oral pathogenic infections.