The ultimate step of retrotransposon activation is mobilization–the insertion of a new copy of itself into the host genome^3,5,8. Although retrotransposon activity has been examined at the transcription and translation levels, the spatiotemporal patterns of mobilization events during development are relatively unexplored²⁷. Encouraged by our previous work monitoring their mobilization during oogenesis²⁸, we sought to systematically characterize retrotransposon jumping events in somatic tissues of Drosophila, a powerful genetic model with well-curated retrotransposon annotation²⁸.

To achieve this, we examined nine retrotransposon families that maintain full-length copies in the genome and potentially can mobilize: 3S18, 412, Blood, Burdock, Copia, Mdg1, Doc, I-element, and mdg4^29–37. For each family, we engineered a corresponding eGFP mobilization reporter (Fig. 1a and Extended Data Fig. 1a). Similar to reporters designed previously^5,29,38, an eGFP cassette was inserted into the retrotransposon in the antisense direction. This cassette contains a disruptive intron that can be spliced during transcription of the retrotransposon, but not of eGFP. As such, the reporter can only express eGFP after intron-spliced retrotransposon mRNAs are used as templates to make cDNA (Fig. 1a and Extended Data Fig. 1a). To minimize the positional effects from chromatin on retrotransposon activity, we generated fly alleles that carry each eGFP reporter at two different genomic loci and obtained the consistent findings reported below.

We examined seven somatic tissues from 2–4-day-old flies to assess eGFP signals from the mobilization reporters: brain, salivary gland, proventriculus, midgut, hindgut, Malpighian tubule, and fat body. Among the nine retrotransposon families, eight displayed no detectable eGFP signals (Extended Data Fig. 1b). In contrast, the reporter for mdg4 produced eGFP signals in the salivary gland, proventriculus, midgut, and hindgut (Fig. 1b,c and Extended Data Fig. 1b,c). In summary, our data indicate that mdg4 still mobilizes in certain somatic tissues.

Based on the reporter design, both original cells harboring mobilization events and their progeny can produce eGFP. The mobilization events we detected could occur either in the adult cells we examined, or in their developmental precursors. As a holometabolous insect, Drosophila proceeds through four different life stages: embryo, larva, pupa and adult. To determine when mdg4 mobilizes, we first examined the most highly labelled tissue, the hindgut (Fig. 1c), at these stages. We discovered that mdg4 selectively becomes active at the pupal stage (Fig. 2 and Extended Data Fig. 2). Notably, metamorphosis is initiated in pupae, during which many larval tissues are first destroyed and then regenerated to produce adult structures, such as the salivary gland and the gut³⁹. For the first 20 hours after puparium formation (APF), when the hindgut degenerates, we barely detected any mdg4 mobilization (Fig. 2a,b and Extended Data Fig. 2a). However, at 24 hours APF, we observed eGFP-positive cells in newly formed hindguts (Fig. 2a,b). As the hindgut grew during metamorphosis, we detected more cells harboring mobilization events (Fig. 2a,b). Similar to the hindgut, mdg4 does not appear to mobilize in the degenerating salivary gland, proventriculus, and midgut at the early pupal stage, but mobilizes in these tissues during their developmental regeneration (Extended Data Fig. 2b,c).

As adult Drosophila hindguts possess no active stem cells and little cell turnover⁴⁰, an unceasing flow of mdg4 mobilization is predicted to steadily increase the number of eGFP-positive cells as pupae develop into adulthood. However, we detected that the average number of eGFP-positive cells per hindgut decreased from an average of 46 in pupae to 13 in adults (Fig. 2b). Similarly, we also observed more eGFP-positive cells in pupal salivary glands, proventriculi, and midguts than in the corresponding adult structures (Extended Data Fig. 2c). Therefore, our data indicate that a proportion of cells harboring mdg4 mobilization events are eliminated during development, and that mdg4 only mobilizes at the pupal stage.

Since reporter cassettes can be epigenetically silenced after retrotransposition⁴¹, we further examined intron removal from the reporter by PCR to validate our findings from eGFP expression (Extended Data Fig. 3). Consistent with eGFP signals, we detected a clear intron-removed PCR product from the tissue that harbors the most eGFP-positive cells–hindguts from the 48-hour pupae (Extended Data Fig. 3b). For tissues without eGFP-positive cells, including embryos, hindguts from larvae or 0-hour pupae, the intron-removed PCR product was not detected (Extended Data Fig. 3b). Notably, this PCR-based method could not detect the intron-removed DNA from the adult tissues that harbor a few GFP-positive cells within a large population (Extended Data Fig. 3c), indicating that this method is less sensitive than monitoring eGFP fluorescence when probing infrequent mobilization events. Overall, these findings further support that mdg4 mobilizes during tissue regeneration, but not at other stages.

To further investigate the activity of mdg4 during development, we monitored its mRNA levels at different stages (Fig. 2c,d). In addition to transcribing full-length transcripts that encode Gag and Pol proteins, for its activation, mdg4 can produce spliced transcripts (Env mRNAs) encoding Env proteins (Fig. 2c)⁴². We found that full-length mdg4 mRNAs were transcribed throughout all life stages (Fig. 2d), arguing against transcriptional silencing. However, the Env mRNAs were primarily produced at the pupal stage (Fig. 2d). These results were confirmed by RNA sequencing (Extended Data Fig. 4). In summary, our data show that mdg4 selectively becomes active at the pupal stage to produce Env, potentially for infectious virus-like particle packaging and mobilization.

Given their repetitive nature, precisely quantifying transposon copy numbers has been challenging²⁷. Nanopore sequencing technology, which can sequence DNA up to megabases without PCR amplification⁴³, allows us to accurately examine the copy numbers of endogenous mdg4 in laboratory strains used in this study. From the laboratory strains we sequenced, we detected 30–33 copies of full-length endogenous mdg4 (Fig. 3). For most copies, we can unambiguously assign them to unique genomic regions (Type 1 from Fig. 3). Meanwhile, we also detected full-length mdg4 flanked by discordant sequences (Type 2 from Fig. 3) or unassembled genomic sequences (Type 3 from Fig. 3). These data indicate that the currently assembled fly genome still underestimates the full-length copies of mdg4. Furthermore, we analyzed published Nanopore sequencing data from two wild strains⁴⁴, and found that both contain 6–7 copies of full-length mdg4. Similar to laboratory strains, while most of these full-length copies in wild strains can be uniquely assigned to the assembled fly genome (Type 1 from Fig. 3), there are one or two copies of full-length mdg4 flanked by unassembled DNA sequences (Type 3 from Fig. 3). Given that the laboratory strains were isolated ~100 years ago^45,46, the heterogeneity we detected between laboratory and wild strains highlights that mdg4 has been actively mobilizing, potentially facilitating species evolution or adaptation.

To determine the function of developmental mdg4 activation, we used RNAi constructs to ubiquitously suppress mdg4 in the laboratory strains. By depleting over 80% of mdg4 mRNAs and abolishing mdg4 mobilization during development (Extended Data Fig. 5a,b), we concluded that our RNAi constructs efficiently silenced mdg4. With mdg4 being suppressed (sh-mdg4), we found that flies still developed normally and exhibited no overt phenotypes.

Given the retroviral origin of mdg4, we next tested whether flies can establish protective immunity from experiencing mdg4 activation for future viral defense. Since we observed most mdg4 mobilization events in digestive tissues (Fig. 1 and Extended Data Fig. 1), we utilized oral feeding to infect adult flies with multiple pathogenic viruses: Drosophila C virus (DCV), Cricket paralysis virus (CrPV), Flock House virus (FHV), and Invertebrate iridescent virus 6 (IIV-6). Among them, the first three are RNA viruses, and IIV-6 is a dsDNA virus. Upon housing adult flies on virus-containing food for 20 days, we observed that mdg4 activity optimally protects the hosts (Fig. 4a, Extended Data Fig. 6a,7a). For DCV infection, while only 30% of flies from the control group (sh-white) died (Fig. 4a), the lethality rate of flies with mdg4 silenced (sh-mdg4) significantly soared to 78% (P = 7.2 × 10⁻⁹, Cox proportional hazards regression; the same test applies to other P value calculations on lethality rates; Fig. 4a). For the other three viruses, the lethality rates were also significantly higher when mdg4 was silenced (Fig. 4a and Extended Data Fig. 7a): CrPV (from 24% in controls to 60% in sh-mdg4 flies, P = 2.5 × 10⁻⁶); FHV (from 21% to 43%, P = 0.0013); and IIV-6 (from 11% to 32%, P = 0.0022).

The aforementioned experiments were performed by activating the sh-mdg4 construct throughout the fly lifespan. While full-length transcripts are produced at all stages, Env mRNA production and mobilization only occur at the pupal stage (Fig. 2d). Thus, we next tested whether the protection from viral infection was gained from mdg4 activity specifically during the pupal stage. For the flies that only harbor the endogenous copies of mdg4, we suppressed mdg4 activity specifically at the pupal stage (Extended Data Fig. 6a–d). Compared to control flies (sh-white), adult flies with mdg4 silenced at the pupal stage showed significantly higher lethality rates upon viral infection (Fig. 4b). By treating flies with DCV or FHV for 20 days, the mortality rate significantly increased for sh-mdg4 flies compared to control flies (Fig. 4b). By contrast, specifically activating RNAi at the adult stage and raising the flies on food containing either DCV or FHV for 20 days led to similar survival rates between mdg4 depleted and control flies (Fig. 4c). Altogether, our data from stage specific suppression indicate that it is the mdg4 activity during the pupal stage, but not other stages, that bolsters the immune system for future antiviral defense.

For fruit flies, natural viral infection also can occur from single-meal consumption. Instead of continuously housing flies on virus-containing food, we next tested the condition of one-time infection by feeding flies with viruses for only 16 hours (Extended Data Fig. 6a). Silencing mdg4 appears to have no impact on virus consumption (Extended Data Fig. 5c). Monitoring mortality rates thereafter revealed a similar function of mdg4 on protecting flies from the three RNA viruses (Fig. 5a and Extended Data Fig. 7b). For DCV, CrPV, and FHV, the lethality rates were all significantly higher when mdg4 was suppressed, compared with control animals (P = 4.3 × 10⁻⁷ for DCV, P = 0.0012 for CrPV, and P = 0.0002 for FHV; Fig. 5a and Extended Data Fig. 7b).

Noting a previous report that fruit flies are capable of clearing DCV upon one-time infection⁴⁷, we next tested the potential function of mdg4 activation on DCV clearance. Control flies (sh-white) that experienced normal mdg4 activation at the pupal stage cleared the viruses within one day (Fig. 5b and Extended Data Fig. 5d). In contrast, flies with mdg4 silenced were unable to clear viruses (Fig. 5b and Extended Data Fig. 5d). Based on RT-qPCR experiments, there were > 62,557-fold more DCV RNA in sh-mdg4 flies than control animals (sh-white) for the first 10 days after infection (Extended Data Fig. 7c). During these first 10 days, sh-mdg4 flies showed a significantly 4.4-fold higher mortality rate than sh-white controls: 28.8% for sh-mdg4 flies vs. 6.6% for sh-white animals (P = 4.3 × 10⁻⁷; Fig. 5a). Subsequently, the mortality rates of sh-mdg4 flies were decreased to a level that is indistinguishable from sh-white control animals (Fig. 5a). Consistently, there was minimal, if any, DCV detectable in the sh-mdg4 survivors at day 15 after infection (Fig. 5b), indicating that the survivors eventually clear out DCV. In summary, our findings indicate that mdg4 activation renders a robust ability to rapidly clear invading DCV at the adult stage.

We also tested the necessity of having mdg4 activation at the pupal stage on adult DCV clearance with the one-time feeding assay (Extended Data Fig. 6a). When only silencing mdg4 during metamorphosis, adult flies lacked the ability to clear orally infected DCV (Fig. 5c and Extended Data Fig. 5e). For example, there was 787,514-fold more DCV in sh-mdg4 flies than controls at day 6 after infection (sh-mdg4: DCV ΔCt = −4.83; sh-white: DCV ΔCt = 14.76; and ΔΔCt = 19.59, Extended Data Fig. 7d). By contrast, flies with full-length transcripts suppressed at adulthood, but had experienced mdg4 activation at the pupal stage, had no defects on DCV clearance (Fig. 5d and Extended Data Fig. 5e). With the DCV one-time feeding assay, these flies cleared the invading viruses as efficiently as control animals (Fig. 5d). Our data thus suggest that having mdg4 activation at the pupal stage is essential for rapid clearance of invading viruses in adults.

In Drosophila, multiple mechanisms have been reported to execute an antiviral function¹⁹. To screen key factors that potentially establish protective immunity at the pupal stage, we individually depleted the following four proteins only during metamorphosis and assayed DCV clearance in adult flies after one-time infection (Fig. 6a). These factors were: Dicer-2 and Argonaute-2 from the RNAi pathway, Pelle from the Toll pathway, and Relish from the IMD and dSTING pathways^19,48–51. While suppressing Dicer-2, Argonaute-2, or Pelle at the pupal stage had no impact on DCV clearance in adult flies, silencing Relish led to similar DCV accumulation as observed from sh-mdg4 flies (Fig. 6a).

Although recent studies established the antiviral function of Relish when viruses were administrated via injection¹⁹, whether Relish possesses similar role upon oral-feeding is still unclear. By housing flies on virus-containing food for 20 days, we found Relish is essential to protect flies from oral infection (Fig. 6b and Extended Data Fig. 8a). Similar to mdg4 suppression, silencing Relish resulted in significantly higher mortality rates when flies were orally challenged by DCV, FHV, CrPV, and IIV-6 (Fig. 6b and Extended Data Fig. 8a). Similarly, sh-relish flies failed to efficiently clear DCV from the one-time feeding assay (Fig. 6c). At day six after infection, sh-relish flies had 292,659-fold more DCV than controls (sh-relish: DCV ΔCt = −3.25; sh-white: DCV ΔCt = 14.91; and ΔΔCt = 18.16; Extended Data Fig. 8b).

Given that mdg4 possibly induces the antiviral function of Relish at the pupal stage, we next tested whether Relish activation at this specific stage is essential for combating viruses in adults. Suppressing Relish solely at the pupal stage made adult flies significantly more vulnerable to viral infection (Fig. 6d,e). When raising these flies on DCV-containing food for 20 days, the mortality rate significantly increased from 33% in control flies to 82% in sh-relish group (P = 2.8 × 10⁻¹³; Fig. 6d). FHV infection led to similar findings: while control flies had a 39% mortality rate, the rate from sh-relish animals soared to 76% (P = 1.9 × 10⁻⁷; Fig. 6d). With the one-time feeding assay, flies with Relish silenced at the pupal stage failed to rapidly clear DCV (Fig. 6e and Extended Data Fig. 8c), resembling the findings from sh-mdg4 flies (Fig. 5c and Extended Data Fig. 7d). In contrast, the Relish activity at the adult stage is less prominent during viral warfare (Extended Data Fig. 8d). Raised on DCV-containing food, flies with Relish suppressed solely during adulthood increased the mortality from 24% in control animals to 44% (P = 0.0096; Extended Data Fig. 8d). For FHV infection, the changes in mortality rates between control and sh-relish animals were not statistically significant (from 24% in controls to 35% in sh-relish animals, P= 0.087; Extended Data Fig. 8d). It is worth noting that besides the protective response induced by mdg4/Relish, the RNAi pathway also contributes to antiviral responses upon oral infection in adults (Extended Data Fig. 8e)⁴⁷. This pathway could directly degrade the viral RNA upon infection^19,52, likely providing a compensatory mechanism on virus-silencing for the flies’ lack of adult Relish activity.

Altogether, our data indicate that inducing the antiviral function of Relish at the pupal stage provides flies optimal safeguard from future viral infection.

As an NF-κB factor, Relish activation requires an endoproteolytic step in the cytoplasm to release an N-terminal fragment (Relish-N), which contains the DNA-binding domain⁵³. Upon translocating to the nucleus, Relish-N acts as a transcription factor to drive the expression of genes for antiviral responses⁵³. Since mdg4 activation appears to have no impact on Relish expression (Extended Data Fig. 8f), we next sought to test whether mdg4 pupal activation affects Relish-N localization. Given that most tissues undergo degradation at the pupal stage, it is technically challenging to monitor Relish-N localization. Therefore, we focused our examination at the early pupal stage preceding tissue degeneration.

The antiviral response can be activated locally at epithelial barriers that directly encounter pathogens or in the fat body, which mediates a systemic response¹⁹. We first examined Relish localization in the gut, where we detected mobilization events and orally-acquired viruses likely invade the animals. Utilizing Relish-N antibodies that detect both full-length and N-terminal fragments of Relish, we observed very low, if any, signals in normal pupal midgut and the anterior hindgut (Extended Data Fig. 9a). However, at the posterior hindgut from control flies, Relish-N antibodies produced prominent signals in both cytoplasm and nucleus, with a slight enrichment at the nuclear periphery (Fig. 7a). Notably, upon mdg4 suppression, the nuclear Relish-N signal significantly decreased 1.6-fold (P = 1.2 × 10⁻²³; Fig. 7a). These data suggest that mdg4 activation drives Relish translocation in the posterior part of the hindgut.

Besides initiating a local antiviral response, fruit flies can also induce a systemic response by triggering the humoral immunity, which is mainly mediated by the fat body¹⁹. Although mdg4 mobilization events were nearly undetectable in fat body cells, mdg4 products produced within these cells or from the virus-like particles could potentially activate Relish in the fat body. To test this, we first measured mdg4 mRNA levels in fat body from early pupae and found both full-length and Env mRNAs (Extended Data Fig. 9b). Notably, in fat body cells from the control pupae, we detected strong nuclear localization of Relish-N (Fig. 7b and Extended Data Fig. 9c), indicating activation of the systemic antiviral response. However, in mdg4 silenced pupae, the amount of nuclear Relish-N significantly decreased 1.8-fold (P = 5.5 × 10⁻²⁶; Fig. 7b and Extended Data Fig. 9c). Our data hence indicate that even without mobilization, mdg4 products can potentially license Relish activation in the fat body to systemically induce antiviral responses.

In Drosophila, while the IMD pathway utilizes Relish-N for anti-bacterial/fungal responses, the antiviral function of Relish-N is activated through dSTING^48–51. dSTING is the fly ortholog of mammalian STING, which can be activated through innate sensors that detect invading pathogens. We next asked whether dSTING is also required for licensing the Relish-mediated immunity at the pupal stage. Similar to mdg4 suppression, silencing dSTING significantly decreases nuclear Relish-N signals in both hindgut and fat body cells from early pupae (Fig. 8a,b). These data suggest a function of dSTING on triggering Relish activation when flies establish the protective immunity. Given that these flies were raised on germ-free conditions, dSTING is likely activated from recognition of endogenous products, like mRNA or cDNA from mdg4.

If the Relish activity triggered by dSTING establishes protective immunity, silencing dSTING at pupal stage should leave adult flies compromised. Indeed, upon persistent DCV infection, the adult flies that experienced normal dSTING expression at the pupal stage had a mortality rate of 33%. However, the adults with dSTING suppressed at the pupal stage showed a significantly higher mortality rate: 73%, P = 7.8 × 10⁻¹³ (Fig. 8c). Similarly, we observed a significant increase of mortality rate when files were raised on FHV for 20 days: from 39% for control flies to 83.5% in sh-dSTING animals (P = 1.9 × 10⁻¹⁴, Fig. 8c). With the one-time feeding assay, we found that adult flies lacking dSTING at pupal stage failed to clear orally acquired DCV (Fig. 8d), reminiscent of the findings when either mdg4 or Relish was silenced at the same stage (Fig. 5c and Fig. 6e).

Collectively, our data indicate that dSTING is an essential component to induce Relish-mediated immunity at the pupal stage to provide adult flies maximal protection against viral infection.

Generally, all flies were maintained at 25 °C and grown on standard agar-corn medium. Flies used to achieve gene silencing at all life stages carried one copy of the engineered mdg4 reporter. Flies used to achieve stage-specific gene silencing only had endogenous copies of mdg4. To silence white, mdg4, Relish, or dSTING at the pupal stage, embryos and larvae were raised at 18 °C. Wandering larvae were transferred to 29 °C and newly eclosed flies were transferred to 25 °C. To silence white, mdg4, or Relish in adult flies, embryos, larvae and pupae were raised in 18 °C and the newly eclosed flies were transferred to 29 °C. To minimize effects caused by genetic background, all of the sh-RNA flies were isogenized to w¹¹¹⁸ isogenic background (w¹¹¹⁸; iso2; Dr/TM6B,Tb,Hu). The fly alleles used in this study are listed in Supplementary Table 1. Supplementary Table 2 summarizes the fly genotypes for each figure. Animals were assigned to each biological groups in a randomized manner. Given that flies had to be sorted based on genotypes, data collection and analysis were not performed blind to the conditions of the experiments.

Total RNA from 10 flies (5 females and 5 males) was extracted by using mirVana^™ miRNA Isolation Kit (Thermo Fisher Scientific, Cat #AM1560). Ten μg of total RNA was treated with 2 μl of TURBO DNase (Thermo Fisher Scientific, Cat #AM2238) at 37 °C for 30 minutes. After DNase treatment, RNA was purified by RNA Clean & Concentrator-5 (Genesee Scientific, Cat #11-326), and the concentration was measured by NanoDrop. One μg of purified RNA was used for reverse transcription in 20 μl of reaction by using 1 μl of SuperScript^™ III Reverse Transcriptase (Thermo Fisher Scientific, Cat #18080044) and 1 μl of random primers (Invitrogen, Cat #58875). To test the amount of DCV present after oral feeding, 1 μg of purified RNA was used in a 20 μl reaction with 4 μl of iScript reverse transcriptase supermix (BIO-RAD, Cat# 1708841). For each reaction, 1 μl of cDNA was used for either PCR by using GoTaq Green Master Mix or qPCR by using SsoFast EvaGreen Supermixes (BIO-RAD, Cat# 172-5204). PCR products were run on 1.5% agarose gel and scanned by BIO-RAD Gel Doc Universal Hood II. P values were calculated from at least three independent biological replicates using a two-tailed, two-sample unequal variance t-test. The error bars on the graphs report standard deviation for three biological replicates. qPCR for Extended Data Figure 8f was calculated from two biological replicates. Extended Data Figures 5c and 8f use the DART-PCR method of analysis⁶¹. Primers used for RT-PCR and RT-qPCR are listed in Supplementary Table 3a. P values were calculated by the Student’s t-test. The data met the assumptions of the statistical tests used. No animals or data points were excluded from these analyses.

The constructs of mdg4 transposition reporter (mdg4-TR) and negative control (NC), 412-TR, Blood-TR, Burdock-TR, Copia-TR and Mdg1-TR were generated by using Counter-Selection BAC Modification Kit (GENE BRIDGES, Cat# K002). The BAC clone used in this study to serve as template, which contains an intact copy of mdg4, is p[acman]-CH322-167N01. For other BAC clones: 412 is p[acman]-CH322-181C5, Blood is p[acman]-CH322-127N18, Burdock is p[acman]-CH322-187A21, Copia is p[acman]-CH322-186G13, and Mdg1 is p[acman]-CH322-86D10. Positive control of mdg4-TR (PC) and 3S18-TR was made by using Gibson Assembly and cloned into the BamHI site of attB-P[acman]-CmR-BW vector. mdg4 is 7,469 bp, Gag encoding sequence is from 1,080 to 2,435, Pol encoding sequence is from 2,438 to 5,470, and Env encoding sequence includes 567, 568 and from 5,551 to 7,000. The positive control, transposition reporter and negative control were inserted into mdg4 that is between 5,478 and 5,479, which will not disrupt any coding sequence. The transposition reporter for 3S18, 412, Mdg1, Blood, Copia and Burdock were constructed by landing the same GFP cassette (Extended Data Fig. 1a) into the following position respectively: 5,756–5,757, 6,914–6,915, 6,675–6,676, 6,823–6,824, 4,750–4,751, and 5,800–5,801.

To construct the plasmid for mdg4 silencing, the DNA fragments of sh-RNAs for depleting mdg4 were synthesized and cloned into the NheI and EcoRI sites of VALIUM20. All of the constructs were verified by colony PCR and sanger sequencing. All plasmids were site-specifically landed into fly genome at attP2 and attP40 sites. The 7 designed sh- mdg4 constructs suppressed mdg4 expression with different efficiencies (Extended Data Fig. 5a). However, all of them can silence mdg4 to a level that leads to consistent antiviral defects for the host flies (Extended Data Fig. 5). Notably, for the alleles that carry sh- mdg4-4 and −7 constructs that lead to lower RNAi efficiency, flies showed weaker virus-clearance defects. These data further suggest the antiviral response is specifically from mdg4. Primers used for constructing plasmids are listed in Supplementary Table 3b.

To quantify eGFP-positive cells from mobilization reporter, pupae or adult flies were heat-shocked at 37 °C for 30 minutes, then recovered at room temperature for at least 3 hours. Corresponding tissues were dissected in cold PBS and fixed with 4% PFA. Samples were then washed 3 times with PBST, stained with DAPI and mounted with VECTASHIELD MOUNTING MEDIUM (VWR Cat #:101098-042). The cells that had GFP and DAPI co-localization were counted. Both adult male and female flies harbored similar number of GFP positive cells. For quantification, GFP fluorescence from female flies was directly visualized and the GFP-positive cells were counted by using ZEISS ApoTome fluorescence microscope. Representative images were taken by the Leica TCS SP5 confocal microscope.

Drosophila S2 cells (Duke cell culture facility, Cat# CRL-1963) were cultured in T175 flask with 30 ml of complete medium (Schneider’s Drosophila Medium, Thermo Fisher Scientific, Cat# 21720024), 10% heat-inactivated FBS (Thermo Fisher Scientific, Cat# 10082139) and 1% penicillin & streptomycin (Thermo Fisher Scientific, Cat# 15140122)). For viral infection, S2 cells were expanded to 6 T175 flasks that contain complete medium plus 0.1% of Pluronic^® F-68 (Thermo Fisher Scientific, Cat# 24040032) with a 1:6 split. Three days later, when cells reach about 75% confluence, 60 μl of viruses was thawed (DCV, CrPV, FHV and IIV-6) (3 × 10¹¹ TCID₅₀/ml) in 6 ml of Schneider’s Drosophila Medium and filtered with a 0.2 μm syringe filter (VWR, Cat# 28145-477). S2 cells were infected in each T175 flask with 1 ml of viruses diluted in medium.

For DCV, CrPV, and FHV purification: when more than 75% of cells were dead, cells with viruses were harvested and frozen at −20 °C for 1 hour to facilitate cell lysis. Then, viruses were thawed on the bench, transferred to ultracentrifuge tubes (BRCKMAN, Cat# 344058) and spun at 27,000 g in SW32Ti for 40 minutes at 4 °C to pellet cell debris. Supernatant was transferred to new ultracentrifuge tubes and spun at 110,000 g in SW32Ti for 3 hours at 4 °C to obtain viral pellet. Viral pellet was re-suspended in 1 ml of 50 mM Tris buffer (pH 7.4) and shaken in a cold room overnight. Suspended viruses were overlaid onto 10% sucrose, and centrifuged at 110,000 g in SW32Ti for 3 hours at 4 °C. Viruses were re-suspended in 1 ml of 50 mM Tris buffer (pH 7.4) and shaken at 4 °C for 20 hours. Viruses were aliquoted and stored at −70 °C.

For IIV-6 purification: the viruses were cultured with S2 cells for 7 days. After incubation, cells with viruses were harvested and frozen at −70 °C. The cells with IIV-6 were thawed and re-frozen for 3 cycles to facilitate cell lysis. Then the cells were transferred to 50 ml tubes and spun at 600 g for 15 minutes to remove the cell debris. This step was repeated once more. Supernatant was transferred to ultracentrifuge tubes and spun at 10,000 g in SW32Ti for 30 minutes at 4 °C to obtain viral pellet. IIV-6 pellet was re-suspended in 1 ml of ultrapure water and was laid onto 30% sucrose, and centrifuged at 30,000 g in SW32Ti for 1 hour at 4 °C. Viruses were then re-suspended in 1 ml of 50 mM Tris buffer (pH 7.4), aliquoted and stored at −70°C.

Viral titer was measured by endpoint dilution assay. Briefly, viral stocks were thawed on ice and made 10-fold serial dilution from 10⁻¹ to 10⁻¹¹. In 96-well plate, each well was filled with 100 μl of medium, 40,000 S2 cells, and 20 μl of medium with diluted viruses. Cells were incubated at 25 °C for 7 days. TCID₅₀ was calculated by Spearman-Karber Method.

All flies used for DCV oral infection were dechorionated at the embryonic stage and treated with tetracycline for two generations. They were tested to be Wolbachia-negative by PCR (F: 5’-ACAGGGTCAGTAAAATATATTGCAG-3’, R: 5’-AGTTTTTCTGCTTTGATCACTACAC-3’). All viral infection assays were performed with at least two biological replicates. For each viral infection experiment: flies within 1 day old were collected (20 males and 20 females in each tube) and flipped to fresh food every day. These flies were used for the following experiments.

For one-time infection: five-day old flies were starved for 8 hours in an empty vial (horizontal). After starvation, flies were transferred to a vial containing a piece of sliced cotton plug in the bottom, which was soaked with 1 ml PBS with DCV at 3 × 10⁹ TCID₅₀/ml (DCV stock and 30% sucrose). After 16 hours, flies were transferred to normal food without DCV (day 0). Flies were transferred to clean food every other day. To extract RNA for RT-PCR and qPCR from these groups, 5 males and 5 females from each group were collected at day1, day3, day6, day10 and day15. To test the survival rate after one-time feeding, after starvation, flies were infected with DCV, CrPV, FHV and IIV-6 at 10 × 10⁹ TCID₅₀/ml. After infection, flies were transferred to fresh food every other day, and the number of dead flies were counted daily.

For persistent infection: 5 × 10⁹ TCID₅₀ of DCV, CrPV, FHV and IIV-6 were put on top of normal food and placed on bench until all liquid evaporated. Then 5-day old flies were transferred to the virus-containing food and counted as day 0. After that, we transferred flies to newly prepared virus-containing food every other day, and the number of dead flies were counted daily.

For survival analysis, we counted the number of dead flies from each group daily over 20 days. Comparison of the survival curves was completed in R 3.6.3 using Cox’s proportional hazards model.

The Relish antibodies (Raybiotech, Cat# 130-10080) were first tested in S2 cells. S2 cells were treated with Tris buffer (pH 7.4) or infected with DCV (1 × 10⁷ TCID₅₀/ml) for 24 hours. After treatment, cells were fixed with 4% PFA for 15 minutes at room temperature. S2 cells with DCV infection showed strong Relish signals in nuclei (data not shown). For pupal tissue staining: to minimize Relish activation caused by bacteria, we made germ-free flies for immunostaining. Gut and fat body from white pupae were dissected in cold PBS and fixed with 4% PFA for 15 minutes at room temperature. After fixation, samples were washed 3 times with PBST (5 minutes each time). After washing, samples were blocked in blocking buffer (10% goat serum (G9023-10ML) in PBST) for 1 hour. Relish antibody (1:300 dilution) was diluted in blocking buffer and incubated with samples at 4 °C overnight. Next day, samples were washed with PBST for 3 times and then incubated with secondary antibody (Molecular Probes, Cat# A11011, 1:300) in blocking buffer for 1.5 hours at room temperature. Samples were then washed 3 times with PBST, stained with DAPI and mounted with VECTASHIELD MOUNTING MEDIUM.

For Relish signal quantification in pupal hindgut and fat body, single Z-stack images of individual sample were taken by the Leica TCS SP5 confocal microscope. The fluorescence signal from individual nucleus was determined by ImageJ. For each cell in fat body, the Relish signal in the nucleus was calculated by subtracting the brightness of the signal in cytoplasm from the brightness of the signal in nucleus determined by ImageJ. For each cell in hindgut, the Relish signal in nucleus was calculated by subtracting the brightness of the background (blank region within the same image) from the brightness of the signal in nucleus determined by ImageJ. Mean ImageJ value from the control animals was set as one and used for the calculation of relative signal sh- mdg4 and sh-Relish flies. P values were calculated by the Student’s t-test. The data met the assumptions of the statistical tests used. No animals or data points were excluded from these analyses.

Non-polyA selection strand-specific mRNA libraries were prepared by using TruSeq Stranded Total RNA Library Prep Gold (Illumina, Cat# 20020598). Briefly, 1 μg of total RNA from the 5 animals was diluted in nuclease-free ultrapure distilled water (Thermo Fisher Scientific, Cat #10977015) to a final volume of 10 μl, then rRNAs were removed by following the standard procedure from the kit. mRNAs were treated with 4 μl of Turbo DNase in 37 °C for 30 minutes and purified by RNA Clean & Concentrator-5. First strand cDNA was synthesized by using SuperScript^™ III Reverse Transcriptase and cDNA was purified by using Ampure XP beads (BECKMAN COULTER, Cat# A63987). dUTP was incorporated after RNA template was removed and second strand cDNA was synthesized (RNase H, Thermo Fisher Scientific, Cat# 18021014; DNA Polymerase I, NEB, Cat# M0209L). dUTP incorporation was done in 16 °C for 2.5 hours and DNA was purified by using Ampure XP beads. Blunting of fragmented DNA and phosphorylation of DNA ends were achieved at 20 °C for 30 minutes. The enzymes used for end repair were T4 DNA polymerase (NEB, Cat# M0203L), Klenow DNA polymerase (NEB, Cat# M0210L) and T4 PNK (NEB, Cat# M0201L). After end repair, a single ‘A’ nucleotide was added to the 3’ ends of the blunt fragments by using Klenow Fragment (3’→5’ exo-) (NEB, Cat# M0212L). Adapter was then ligated to DNA fragments by using T4 DNA ligase (Enzymatics, Cat# L603-HC-L). After UDG (NEB, Cat# M0280S) treatment, DNA fragments were amplified by using Phusion HF DNA polymerase (NEB, Cat# M0530L). Libraries were sent to Novogene for quality control and sequencing.

Genomic DNA was isolated from adult flies using the Zymo Quick-DNA Microprep Kit (Zymo, Cat#D3021). Around 2 μg of DNA per sample was used as input for the Nanopore Ligation Sequencing Kit protocol (Oxford Nanopore Technologies, Cat# SQK-LSK109). Libraries were barcoded with the Nanopore Native Barcoding Expansion Kit (Oxford Nanopore Technologies, Cat# EXP-NBD104). Libraries were sequenced with R9.4 flow cells using the GridION system.

The fast5 files generated by the Nanopore GridION machine were used as input in MinKNOW version 21.05.25 (MinKNOW core 4.3.12). Guppy 5.0.16 is integrated into the MinKNOW. Raw nanopore reads in fastq format were pre-processed by the software porechop with parameters: --extra_end_trim 0 --discard_middle. Clean reads were mapped to the mdg4 consensus sequence first. The mapped reads containing a full-length mdg4 were selected. To determine the flanking sequences, selected reads were mapped to the reference genome of Drosophila melanogaster version dm6 (GCA_000001215.4). Mapping was done by using minimap2 with the parameters: -ax map-ont -Y -t 16. Full-length mdg4 reads with continuous flanking sequences were classified as Type 1. Full-length mdg4 reads with flanking sequences that were mapped to two different chromosomes were classified as Type 2. Full-length mdg4 reads with flanking sequences that were not present in the major chromosomes of dm6 were classified as Type 3. For each type, individual reads were clustered based on the coordinates of the mdg4 and genome junctions. Drosophila chromosome ideogram is made by RIdeogram package. The heat map on the chromosomes represents gene density, which was calculated by 1 Mb windows.

Raw Illumina reads were processed by the software trim_galore, with the parameter –paired –fastqc. The output files of trim_galore were mapped to both the reference genome of Drosophila melanogaster version dm6 (GCA_000001215.4) and the mdg4 consensus sequence using hisat2 with the default parameters except -p 16. To quantify the full-length mdg4 transcripts and the spliced Env transcripts, Illumina reads were assembled by stringtie with the default parameters except for -f 0. The full-length mdg4 mRNA was presented in the stringtie output as a transcript (1..7469) with a single exon (1..7469), while the Env transcript was presented in the stringtie output as an isoform with two exons (1..568; 5551..7469). The normalized transcript abundance values in FPKM were used for plotting Extended Data Fig. 4. Alignment visualization was achieved by using IGV version 2.11.1.