Bombardier Enables Delivery of Short-Form Bomanins in the Drosophila Toll Response

Toll mediates a robust and effective innate immune response across vertebrates and invertebrates. In Drosophila melanogaster, activation of Toll by systemic infection drives the accumulation of a rich repertoire of immune effectors in hemolymph, including the recently characterized Bomanins, as well as the classical antimicrobial peptides (AMPs). Here we report the functional characterization of a Toll-induced hemolymph protein encoded by the bombardier (CG18067) gene. Using the CRISPR/Cas9 system to generate a precise deletion of the bombardier transcriptional unit, we found that Bombardier is required for Toll-mediated defense against fungi and Gram-positive bacteria. Assaying cell-free hemolymph, we found that the Bomanin-dependent candidacidal activity is also dependent on Bombardier, but is independent of the antifungal AMPs Drosomycin and Metchnikowin. Using mass spectrometry, we demonstrated that deletion of bombardier results in the specific absence of short-form Bomanins from hemolymph. In addition, flies lacking Bombardier exhibited a defect in pathogen tolerance that we trace to an aberrant condition triggered by Toll activation. These results lead us to a model in which the presence of Bombardier in wild-type flies enables the proper folding, secretion, or intermolecular associations of short-form Bomanins, and the absence of Bombardier disrupts one or more of these steps, resulting in defects in both immune resistance and tolerance.


INTRODUCTION
Innate immune pathways are found in plants, fungi, and animals and provide a rapid defense against a broad range of pathogens (1)(2)(3). In the fruit fly Drosophila melanogaster, the two major innate immune pathways are Toll and Imd (4)(5)(6). The Toll pathway is activated by Gram-positive bacteria with Lys-type peptidoglycan and by fungi, and is required for defense against these microbes (7)(8)(9)(10). Conversely, the Imd pathway is activated by and plays a major role in survival against Gram-negative bacteria and Gram-positive bacteria with DAP-type peptidoglycan (11,12). These pathways, which are both mediated by NF-κB transcription factors, are broadly conserved as initiators of innate immune responses. Activation of either pathway induces robust production of an array of immune molecules, including antimicrobial peptides (AMPs) (13)(14)(15)(16)(17).
AMPs are found in all kingdoms of life (18)(19)(20)(21)(22). These peptides have long been thought to play the principal effector role in innate immune defense due to their demonstrated in vitro antimicrobial activity and their marked upregulation after infection. However, recent research in D. melanogaster suggests that AMPs play a major role in Imd-mediated defense, but a relatively minor role in Toll-mediated immunity (23).
In contrast to the AMPs, the Drosophila-specific Bomanin peptides (Boms), which are highly induced after infection, are indispensable for resistance against pathogens controlled by the Toll pathway (24). Bom 55C flies, which lack 10 of the 12 Bom genes, succumb to fungal and Gram-positive bacterial infections at rates indistinguishable from Toll-deficient flies (23,24), suggesting that Boms rather than AMPs are the primary Toll effectors.
Bom peptides, like AMPs, are secreted from the fat body, the Drosophila immune organ, into the hemolymph, the Drosophila circulatory fluid. The family is comprised of three groups. The short-form peptides are 16-17 residues long and contain only the Bom motif. The tailed forms contain the Bom motif followed by a C-terminal tail. Finally, the bicipital forms consist of two Bom motifs connected by a linker region (24). Bom 55C flies lack all six of the short-form Boms, two of the three tailed Boms, and two of the three bicipital Boms. High-level expression of short-form Boms is sufficient to rescue the sensitivity of Bom 55C flies to C. glabrata infection (25). Furthermore, the absence of Toll-induced candidacidal activity in Bom 55C hemolymph can be rescued by high-level expression of a short-form Bom (25). However, no in vitro antimicrobial activity has been observed with Bom peptides alone (25), suggesting that the Bomanins act in coordination with additional humoral effectors.
In this study, we demonstrate an essential role in Tollmediated humoral defense for a previously uncharacterized hemolymph protein, Bombardier (one that deploys Boms).

Toll Activation, Drosophila Infection, and Survival Analysis
Flies were raised at 25 • C on cornmeal molasses agar media 1 . The w 1118 strain was used as the wild type. Microbial isolates, culture conditions, and conditions for infection for Enterococcus 1 http://blogs.cornell.edu/drosophila/dssc-cornmeal-recipe/ faecalis, Enterobacter cloacae, Fusarium oxysporum, and Candida glabrata were as described previously (24), except that C. glabrata was concentrated to OD 600 = 100. Flies were incubated at 25 • C after live bacterial infection and at 29 • C after fungal infection. For heat-killed challenge, bacterial cultures were autoclaved and resuspended in 20% glycerol to OD 600 = 10 for E. faecalis and OD 600 = 300 for M. luteus. For both survival assays and hemolymph preparation, flies challenged with heat-killed bacteria were incubated at 29 • C.

Hemolymph Antimicrobial Assays
Candidacidal activity of hemolymph was assayed as described previously (25), except that hemolymph was prepared from groups of 30 flies and all activity assays were carried out for 1 h at room temperature. The number of colonies representing zero percent killing was set as the value obtained by assaying uninduced w 1118 hemolymph.

MALDI-TOF Analysis of Hemolymph
The Toll pathway was activated in flies using heat-killed M. luteus, then incubated at 29 • C for 24 h. Hemolymph was extracted as in Lindsay et al. (25), with slight modifications. Hemolymph extracted with glass capillaries from five male flies was pooled and transferred into 0.1% trifluoroacetic acid (TFA)/50% acetonitrile (ACN). One µl of each mixture was spotted on a Bruker MSP 96 ground steel plate, mixed 1:1 with a saturated solution of Universal MALDI matrix (Sigma-Aldrich) in 0.1% TFA/78% ACN, and air-dried. MALDI-TOF spectra were acquired using a Bruker Autoflex mass spectrometer. Data were collected from 1,500 to 10,000 m/z in positive linear mode, and 1,000-5,000 m/z in positive reflectron mode. Peptide calibration standard II (Bruker) was mixed with Universal MALDI matrix and used as an external calibration standard. At least ten independent samples were collected for each genotype. For peptide identification, peaks were matched to those of corresponding peaks in prior studies (13,25). Representative spectra were visualized using R 3.3.2 and ggplot2 2.2.1 (27,28).

Gene Expression Quantitation
The Toll pathway was activated with heat-killed M. luteus. Using TRIzol (Ambion), total RNA was extracted 18 h after Toll activation from four to six adult flies (2-5 days old). Next, cDNA was synthesized from 500 ng total RNA using the SuperScript II Reverse Transcriptase kit (Invitrogen). Quantitative RT-PCR was performed on an iQ5 cycler (BioRad) using iQ SYBR Green Supermix (BioRad). Quantification of mRNA levels was calculated relative to levels of the ribosomal protein gene rp49 using the Pfaffl method (29). Three independent replicates were completed. See Supplemental Table 1 for qPCR primer sequences.

Hemolymph LC-MS
Flies were challenged with heat-killed M. luteus to activate the Toll pathway. Hemolymph was extracted from 100 to 110 each of w 1118 , bbd, and Bom 55C flies using the same method as in the hemolymph antimicrobial assays, with 50-60 flies processed per Zymo-Spin IC column (Zymo Research) and yielding a total of ∼10 µl hemolymph per genotype. Three independent biological replicates were processed for bbd and Bom 55C , and two independent biological replicates were processed for w 1118 . Extracted hemolymph was mixed 1:1 (vol/vol) with denaturing buffer (8 M Urea, 50 mM Tris, pH 7.8, 150 mM NaCl, protease and phosphatase inhibitors) and protein concentration was determined using a BCA assay. For each sample, 40 µg of hemolymph was diluted to 1 M urea using 50 mM ammonium bicarbonate and digested overnight with trypsin (Promega, V511A) at a 1:100 (trypsin:protein) ratio. After digestion, peptides were reduced with 1 mM dithiothreitol at room temperature for 30 min and then alkylated with 5 mM iodoacetamide at room temperature in the dark for 30 min. Formic acid was added to a 0.1% final concentration and peptides were desalted using the C18-Stage-Tip method and then vacuum dried. The dried peptides were reconstituted in 5% formic acid/5% acetonitrile and 1 µg of total peptide for each sample was loaded for MS analysis. Samples were run in technical triplicates on a Q-Exactive mass spectrometer with instrument and chromatography settings as described previously (30), except for the following modifications: the RAW files were analyzed using Andromeda/MaxQuant (version 1.6.7.0) (31) with default settings (32) except the match between the run and LFQ quantitation settings was enabled for label free quantification. Data were searched against a concatenated targetdecoy database comprised of forward and reversed sequences from the unreviewed UniprotKB/Swiss-Prot FASTA Drosophila database (2019). A mass accuracy of 20 ppm was assigned for the first search and 4.5 ppm for the main search. The statistical analysis was calculated using the DEP analysis R-package (33).

Bacterial Load Quantification
Bacterial load upon death (BLUD) was obtained as in Duneau et al. (34), with slight modifications. Briefly, flies were infected with E. faecalis and vials were monitored every 30 min for newly dead flies. These flies were then individually homogenized with a pestle in 400 µl LB media. Homogenates were also prepared from individual live w 1118 flies 120 h postinfection (hpi). Homogenates were diluted serially in LB and spread on LB agar plates for incubation at 37 • C overnight. Colonies were counted manually and the number of viable bacteria per fly was calculated. Data were obtained from three independent experiments.

Data Analysis
GraphPad Prism 5 was used for statistical tests. Survival data were plotted as Kaplan-Meier curves and were analyzed using the Gehan-Breslow-Wilcoxon test to determine statistical significance. Statistical differences in candidacidal activity were calculated using one-way ANOVA followed by Tukey's test. Multiple Mann-Whitney U tests were used to calculate differences between BLUD samples (p = 0.0085 after Šidák correction for multiple comparisons, α = 0.05, k = 6). Spearman rank correlation was used to assess the relationship between BLUD and time of death.

The bombardier Gene Is Specifically Required for Toll-Mediated Defense
The bombardier (bbd) gene contains a consensus Toll-responsive NF-κB binding site within its promoter region and is strongly expressed upon Toll activation by Gram-positive bacterial infection or other inducers (14,17,35,36). The encoded protein is predicted to be secreted and to generate a mature protein of 222 amino acids with a coiled coil near its C-terminus (37,38). Orthologs of Bombardier are found across the Drosophila genus, but in no other genera (39).
We began our analysis of the bombardier gene by generating a null mutant, using CRISPR/Cas9 to delete 1,906 bp encompassing the annotated transcriptional unit. Flies homozygous for this deletion (hereafter bbd) were viable and morphologically wild-type. Given that bombardier is Tollinducible, we assayed bbd flies for a potential loss-of-function phenotype in Toll-mediated immunity. Specifically, we infected adult bbd flies with various pathogens and then monitored survival. Two additional genotypes were used as controls: w 1118 flies, which served as the wild type, and Bom 55C flies, which lack Toll-mediated humoral defenses due to deletion of the 10 of the 12 Bom genes (24).
As shown in Figure 1, we observed a marked immunodeficiency when bbd flies were challenged with representative species for the three classes of microbes against which Toll provides defense. With the yeast Candida glabrata, more than 90% of w 1118 , but no bbd flies, survived 5 days after infection ( Figure 1A). In the case of the filamentous fungus Fusarium oxysporum, 70% of w 1118 adults, but fewer than 20% of bbd adults, were alive 5 days post-infection ( Figure 1B). Finally, with Enterococcus faecalis, a Gram-positive bacterium, 50% of wild-type flies, but no bbd flies, were alive 5 days after infection ( Figure 1C).
The impairment of Toll-mediated defenses by deletion of bombardier was significant for all three pathogens (p < 0.0001). In the case of C. glabrata, the immunodeficiency of bbd phenocopied that observed for Bom 55C flies (n.s., p > 0.05). In contrast, with either F. oxysporum or E. faecalis, the rate of death was greater for Bom 55C than for bbd ( * * * p < 0.0001 for both infections). The bbd mutant thus displays a substantial, but not complete, loss of Toll-mediated defense.
The expression of bombardier is strongly induced by Toll, but not Imd activation (14). We therefore hypothesized that Imdmediated defenses would not require bombardier function. To test this prediction, we infected bbd flies with Enterobacter cloacae, a Gram-negative bacterium. In this experiment, bbd flies were as immunocompetent as w 1118 flies: more than 90% of both genotypes survived at least 5 days post-infection ( Figure 1D). In contrast, 100% of Rel E20 flies, which are deficient in Imd signaling (40), succumbed to infection within 1 day. Thus, bombardier functions in defense against a range of pathogens for which Toll mediates defense-yeast, filamentous fungi, and Lys-type Gram-positive bacteria-but not against Gram-negative bacteria, against which the Imd pathway is active. The w 1118 strain was the wild-type control; Bom 55C and Rel E20 were the susceptible controls (24,40). Experiments were completed in triplicate with at least 25 flies per genotype in each replicate. Statistical significance was determined using the Gehan-Breslow-Wilcoxon test and bbd is shown relative to w 1118 in black, relative to Bom 55C in orange, and relative to Rel E20 in red ( *** p < 0.0001; n.s., not significant, p > 0.05).
The Candidacidal Activity of Hemolymph Requires Bombardier, but Neither Drosomycin Nor Metchnikowin Next, we investigated the potential humoral role of Bombardier by preparing and assaying cell-free hemolymph. We have previously shown that hemolymph from wild-type flies exhibits a Toll-dependent and Bomanin-dependent candidacidal activity (25). However, we were also curious as to the identity of the active antifungal component. In particular, we considered the potential role of Metchnikowin (Mtk) and Drosomycin (Drs), two antimicrobial peptides (AMPs) that have documented antifungal activity in vitro and are strongly Toll-induced in vivo (14,41,42). We therefore took advantage of the recently described AMPs strain, which is deficient for Mtk and Drs, as well as all other induced AMPs other than the Cecropins (23). Extracting and assaying Toll-induced hemolymph, we found that hemolymph from AMPs flies had a killing activity against C. glabrata comparable to that of wild-type hemolymph (Figure 2).
In contrast, we failed to detect any killing of C. glabrata by bbd hemolymph. We conclude that Boms and Bombardier, but neither Mtk nor Drs, are required for humoral defense against C. glabrata.

Short-Form Bom Peptides Are Specifically Absent From bbd Hemolymph
MALDI-TOF provides a robust tool for characterizing small (<5,000 MW) peptides present in hemolymph after Toll activation. As shown in Figures 3A,B, such a readout includes the aforementioned AMPs (Mtk and Drs), several short-form Boms (BomS1, S2, S3, and S6; see Supplemental Table 2 for updated Bomanin nomenclature), and other induced peptides (e.g., IM4). We have previously shown that deleting the 55C Bom gene cluster removes the peaks attributable to the shortform Boms, while leaving the remaining signals unaffected (25). Remarkably, analysis of bbd hemolymph yielded a similar pattern. As shown in Figures 3C,D, the short-form Boms that were readily detectable in the wild type-S1, S2, S3, and S6-were absent in bbd hemolymph, whereas the remaining peptides, including Mtk, Drs, and IM4, displayed a wild-type profile.
Although bbd disrupts the accumulation of short-form Bom peptides in hemolymph, this effect does not reflect a disruption in transcription or stability of the corresponding Bom mRNAs: robust induction of Toll-regulated genes, including genes of short-form Boms, was readily detectable with qRT-PCR (Supplemental Figure 1). FIGURE 2 | The Toll-induced candidacidal activity of hemolymph requires Bombardier, but neither Drosomycin nor Metchnikowin. Heat-killed M. luteus was used to activate the Toll pathway in flies. Hemolymph was extracted from flies 24 h after Toll induction, mixed with C. glabrata and incubated for 1 h to allow for killing. The surviving yeast cells were plated, and colonies were counted to determine the level of candidacidal activity in the extracted hemolymph. Colony counts from uninduced w 1118 hemolymph were used as the control for no (0%) killing. Experiments were completed four times, with each point representing one replicate. One-way ANOVA was calculated followed by Tukey's test. Significance is shown relative to the null hypothesis of 0% killing (***p < 0.0001; n.s., not significant, p > 0.05). Error bars represent the 95% confidence interval.
Because proteins such as Bombardier and bicipital Boms are too large to be detected by our MALDI-TOF protocol, we used LC-MS to further characterize the relationship between Bombardier and the Boms in hemolymph. For these studies, we prepared Toll-induced hemolymph from three genotypes: w 1118 , bbd, and Bom 55C . In wild-type hemolymph, we readily detected Bombardier protein (Figure 4), consistent with the presence of a canonical secretion signal sequence in the Bombardier coding sequence. Bombardier, like the Boms, is thus secreted into hemolymph upon Toll induction. We also detected all three bicipital Boms-BomBc1, BomBc2, and BomBc3. The LC-MS studies thus complemented the MALDI-TOF studies, with bicipital Boms detected by the former and short-form Boms by the latter (tailed Boms are not detected by either protocol). Next, we assayed bbd hemolymph. As expected, Bombardier was not detected. However, the three bicipital Boms were present at comparable levels in wild-type and bbd hemolymph (see Figure 4). Combined with the MALDI-TOF studies, these results demonstrate that bbd blocks accumulation in hemolymph of short-form, but not bicipital, Boms. Lastly, we analyzed hemolymph from Bom 55C flies, which lack 10 of the 12 Bom genes. As expected, the products of the two deleted bicipital genes (BomBc1 and BomBc2) were absent, whereas the product of the remaining bicipital gene (BomBc3) was present at wild-type levels (see Figure 4). Turning our attention to Bombardier, we observed no effect of the 55C Bom deletion. Thus, Bombardier is required for the presence of short-form Boms in hemolymph, but the 55C Boms are not required for the presence of Bombardier.

Bombardier Mediates Both Infection Resistance and Tolerance
The bbd survival phenotype could be due to an inability to control pathogen growth-a defect in resistance-or an inability to endure infection-a defect in tolerance. Because flies lacking Bombardier demonstrate an increased susceptibility to infection and decreased levels of known resistance factors, the shortform Boms, it seemed likely that bbd flies, like Bom 55C flies, have a defect in infection resistance. In exploring this hypothesis, we found that the model recently developed by Duneau et al. provided a useful framework (34). Following infection of an individual fly, there are two stereotypic outcomes: either the pathogen replicates, reaches a lethal burden, and the fly dies; or the pathogen is controlled at a level below the lethal burden and the fly survives with a persistent infection. Variation in survival curves for different pathogens and fly genotypes reflects variation in both the time required to reach lethal burden and in the fraction of flies that are able to control the infection before it reaches such a threshold. In cases where a fraction of flies control infection, group survival typically drops after infection and then reaches a plateau (23).
The survival curve for bbd flies infected with E. faecalis does not plateau (see Figure 1C). Instead, it exhibits a profile that we hypothesize reflects two phases of death. In the first phase, extending roughly 2 days post-infection, some bbd flies reach a lethal burden of E. faecalis and die, as reflected in a sharp decline in survival; the remainder control the infection. In the second phase, from 2.5 days onward, those flies with a persistent infection die at a reduced but steady rate, due to a defect in tolerance. If this hypothesis is correct, flies dying in the first phase should have a bacterial load upon death (BLUD) comparable to that of wild-type flies dying from infection. Furthermore, those dying in the second phase should have a much lower pathogen burden, comparable to that of wild-type survivors with a persistent infection.
To test our predictions regarding pathogen burden, we measured the BLUD of individual flies after infection with live E. faecalis and divided the data into two time intervals (Figure 5). For the earlier interval (dead flies obtained between 17 and 51.5 hpi), both Bom 55C and bbd bacterial loads upon death were not significantly different from w 1118 (Figure 5, red, p > 0.05). For the later time interval (flies obtained between 68 and 120.5 hpi), bbd flies perished at significantly lower bacterial loads compared to that of bbd flies which died earlier (Figure 5, bbd early compared to bbd late, p < 0.0001), indicating that these two groups die from distinct causes. Importantly, late-death bbd flies perished at significantly lower bacterial loads than those of w 1118 suffering early deaths (p < 0.0001), demonstrating that bbd flies have a defect in tolerance.
Together, the survival curve and BLUD data offer strong support for our two-phase-model: bbd flies died early in infection with high bacterial loads, due to a defect in resistance, and died later with lower bacterial loads, reflecting a deficiency in tolerance. However, we note that the bacterial loads of . For peptide identification, peaks were matched to those of corresponding peaks in prior studies (13,25). Spectra were obtained from at least ten independent biological replicates and representative spectra are shown (a.u., arbitrary units; m/z, mass/charge).  Table 3 shows the full dataset.
bbd flies dying in the later phase were still significantly greater than those of w 1118 flies alive 120 hpi (Figure 5, bbd late compared to live w 1118 , p < 0.0001). This indicates that the later-death bbd group has not completely controlled infection compared to the live w 1118 flies, and suggests that both resistance and tolerance contribute to the later bbd fly deaths. Although we cannot rule out a minor resurgence in bacterial proliferation preceding late death of bbd flies, we note that BLUD and time of death were not significantly correlated for these flies (Supplemental Figure 2, Spearman correlation test, r = −0.2654, p = 0.1564).

Immune Activation, Specifically Bom Expression, Is Deleterious in the Absence of Bombardier
What is the nature of the tolerance defect we observed in bbd flies? More specifically, is their health impaired by an excessive or toxic immune response, or is death due to another class of impaired tolerance (43)? To distinguish between these explanations, we assayed the effect of activating the immune response in bbd flies in the absence of infection.
When bbd flies were challenged with heat-killed E. faecalis, we observed a decrease in survival that first was apparent 3 days post-challenge followed by a steady decline in the number of live flies in the following days (Figure 6A), consistent with the timing of the late-phase deaths (see Figure 5). Overall, the death rate was slower than that of live infection, but the extent of killing was similar between heat-killed and live E. faecalis: fewer than 20% of flies survived (compare Figures 1C, 6A). In contrast, no effect on survival was observed upon challenge of either w 1118 or Bom 55C flies with heat-killed E. faecalis: >95% flies survived seven or more days post-challenge.
The effect of immune stimulation on bbd survival was not specific to E. faecalis. When we repeated the challenge experiments with heat-killed Micrococcus luteus, which activates the Toll response [see Supplemental Figure 1, as well as (25,44)], the effect on bbd survival was again marked: 5 days after challenge, fewer than 5% of bbd flies were alive, compared to The pathogen loads of early deaths for Bom 55C and bbd were not significantly different from w 1118 (p > 0.05). The pathogen load of late bbd fly deaths is significantly different from that of the early-death bbd and w 1118 groups (***p < 0.0001) and also significantly different from that of live w 1118 flies 120 hpi (***p < 0.0001). Finally, the early-death w 1118 pathogen load was significantly different from that of live w 1118 flies 120 hpi (***p < 0.0001) (hpi, hours post-infection).
survival of >95% of w 1118 and 85% of Bom 55C flies over the same period of time ( Figure 6B).
As both M. luteus and E. faecalis induce the Toll pathway, Toll activation could be the key factor in bbd mortality. To address this hypothesis, bbd flies were crossed with MyD88 kra1 (Tolldeficient) flies to generate the MyD88 kra1 bbd double mutant, and the resulting flies were challenged with heat-killed E. faecalis and M. luteus. Unlike bbd flies, MyD88 kra1 bbd flies survived challenge with Toll activators (Figures 7A,B). Because blocking the Toll pathway with MyD88 kra1 rescues the bbd phenotype triggered by heat-killed bacteria (p < 0.0001 compared to bbd, p > 0.05 compared to MyD88 kra1 for both heat-killed bacteria), we conclude that Toll activation underlies the death of bbd flies in the absence of infection.
As described above, Bom genes are transcribed in bbd flies (Supplemental Figure 1), but short-form Bom peptides do not appear in hemolymph (Figure 3). This suggests a mislocalization of these peptides, perhaps in an unprocessed or misfolded state. Given that short-form Bom genes are among the most abundantly transcribed genes after infection (17,36), such mislocalized or misfolded Boms could rapidly accumulate to high levels in bbd flies. Could this explain the death of bbd flies upon immune stimulation? To address this question, we generated Bom 55C bbd double mutants and assayed the effect of immune induction alongside both Bom 55C and bbd flies (Figures 7A,B). The result was unequivocal: introducing Bom 55C , which deletes all of the short-form Boms, eliminated the effect of bbd on survival following immune stimulation (p < 0.0001 compared to bbd, p > 0.05 compared to Bom 55C for both heat-killed bacteria). The fact that Bom 55C is epistatic to bbd demonstrates that Toll-driven expression of Bom genes is specifically responsible for the death of immune-stimulated bbd flies.

DISCUSSION
The results presented in this study identify a key factor that regulates humoral and Bom-mediated defense in Drosophila.
We demonstrate that bbd flies are defective in resistance to pathogens controlled by the Toll pathway. The results support the hypothesis that this defect results from the absence of shortform Boms in bbd hemolymph. Absence of Boms is sufficient to cause a defect in resistance (24) and bbd hemolymph appears to be lacking the short-form Boms but no other component, save Bombardier itself. Furthermore, bbd phenocopies Bom 55C with regard to survival after C. glabrata infection, and resistance to C. glabrata can be restored in Bom 55C flies by expression of short-form Boms (25). Finally, bbd hemolymph lacks candidacidal activity, which is dependent on short-form Bom peptides (25) and which we show here does not require Drs or Mtk. For pathogens other than C. glabrata, the effect of deleting Bombardier is less severe than that of deleting the ten Bom genes clustered at 55C. Our mass spectrometry data suggest an explanation. Whereas, short-form Boms are absent from bbd hemolymph, bicipital Boms are present. (Tailed Boms were not detected with either mass spectrometry method.) Therefore, we postulate that the bicipital Boms, which are not required for resistance to C. glabrata (25), are functional against other pathogens. This would explain why bbd flies are more resistant than Bom 55C flies upon infection with E. faecalis or F. oxysporum (Figure 1). In this regard, we note that Bombardier and all three forms of Bom proteins-short, tailed, and bicipitalare found across the Drosophila genus, supporting the notion that all three classes of Boms are immunoprotective and therefore maintained across the Drosophila genus.
It might appear that our discovery of Bombardier was serendipitous, given our role in defining the Bomanin gene family (24,25). In hindsight, however, the link was forged in our approach. We selected CG18067 from the most strongly inducible Toll-regulated loci, a group that also includes eight of the Bomanin genes. Next, we engineered a CRISPR/Cas9 deletion of CG18067 and assayed this knockout with the identical set of pathogens that we had used for the Bom 55C deletion, screening for loss of survival upon infection. Having examined a gene that is as strongly induced as the Bomanins, present in the same range of species as the Bomanins, and with a spectrum of loss-of-function phenotypes similar to that of the Bomanins, it is not particularly surprising that we would find ourselves studying a gene that affects the Bomanins.

Bombardier Function and Structure
What is the function of Bombardier? Deleting the gene results in the absence of short-form Boms from hemolymph, an effect we find is at the level of protein. Other mature immune peptides are present at normal levels in the hemolymph, and there is thus no general defect in translation, secretion, or processing. Based on these findings, we propose that Bombardier normally functions either to chaperone short Boms as they are secreted from the fat body into the hemolymph or, alternatively, to protect the Boms from misfolding or aggregation while in the hemolymph. We further hypothesize that it is the ectopic localization or aberrant form of short-form Boms in bbd flies that generates morbidity upon Toll pathway activation. In support of this idea, we showed that Bom expression underlies the lethality observed in bbd flies (Figure 7). Whether the short-form Boms physically interact with Bombardier, perhaps in the context of a larger antimicrobial complex, is currently unknown.
Activation of Toll-like receptor (TLR) signaling is important for innate immunity, but induction of the pathway can lead to autoimmune disorders and chronic inflammatory disease (45)(46)(47)(48). Here we report an autoimmune activity driven by Toll-induced Bom expression in flies lacking a downstream pathway component, Bombardier. To what extent this parallel can be exploited in the context of understanding autoimmune disorders promises to be a significant focus for future investigation.

DATA AVAILABILITY STATEMENT
The LC-MS datasets generated for this study can be accessed on the MassIVE data repository using the accession identifier: MSV000084509.

AUTHOR CONTRIBUTIONS
SL and SW: conceptualization, writing-original draft, and visualization. SL, AF, and EB: methodology. SL and AF: formal analysis. SL, AF, and LC: investigation. SL, AF, LC, EB, and SW: writing-review and editing. EB and SW: supervision and funding acquisition.

FUNDING
This work was supported by National Institutes of Health (NIH) grant R01 GM050545 (to SW), NIH grant GM119132 (to EB), and NIH predoctoral training grant T32 GM008666 (to SL). The funders had no role in study design, data collection and analysis, the decision to publish, or preparation of the manuscript.