Two fly strains were used in this study. Preliminary screening was performed using Drosophila melanogaster of the genotype w; Sp/Cyo; Sb/TM6B carrying the endogenous wMel Wolbachia strain (Serbus and Sullivan, 2007; Christensen et al., 2016). Drosophila simulans (D. sim) infected with the endogenous Wolbachia riverside (wRi) strain were used for further analyses (Hoffmann et al., 1990; Serbus and Sullivan, 2007). GAL4 lines sourced from the Bloomington Drosophila Stock Center (BDSC) were also used to drive RNAi expression. Constitutive somatic expression was driven by the Actin5C-GAL4 driver w; P{Act5C-GAL4-w}E1/Cyo (BDSC# 25374) and the daughterless-GAL4 driver w; P{w+, GMR12B08-GAL4}attP2 (BDSC# 48489). Mifepristone-induced gene expression was driven by the GeneSwitch-GAL4 driver yw {hs-FLP}; {w+, UAS-GFP}; {w+, Act5C-GS-GAL4}/TM6B, Tb (BDSC #9431).

Fruit flies were maintained in plastic bottles/vials containing standard fly food media. The recipe was derived from Bloomington Drosophila Stock Center as described previously (Christensen et al., 2016). The flies were raised in an Invictus Drosophila incubator at 25°C, under standard 12:12 h light–dark cycle. For the experiments, “0-day old” flies were collected and kept on standard fly food medium for 2 days. The flies were then used for drug treatments in vials or within a plate assay format as described previously (Christensen et al., 2019). Only female flies were used for the plate-based screening experiments, to reduce possible variation in population behavior per well.

Two or more chemicals were used to alter the functionality of each of the candidate host processes pursued in this study. Where possible, compounds with opposite effects on the process of interest were included, such as the microtubule-depolymerizing drug, colchicine, and the microtubule-stabilizing drug, taxol, as well as the phospholipase C (PLC) inhibitor, U73122, and the PLC activator, 3-M3MFBS. The final list included a total of 37 candidate compounds (Supplementary Table S1). All the drugs were dissolved in DMSO. Most stock solutions, including rifampicin, were prepared in advance as 10 mM solutions, aliquoted, and stored at −20°C. Rapamycin was ordered as a 5 mM solution in DMSO, also stored at −20°C. Light-sensitive drugs were stored in the dark.

Immediately before use, chemical stocks were thawed and diluted 100X into fly food that had been re-melted, then cooled. Control vials, prepared in parallel with the chemical treatment vials, were treated with equivalent amounts of DMSO alone. In all cases, the final concentration of DMSO in food was capped at 1%. For the chemical screen, a minimum of 10 mL drug food was prepared per condition, to be further dispensed in approximately 1 mL amounts per treatment well. For drug lethality tests and GS-GAL4 induction experiments that were carried out in vials, food containing control DMSO and DMSO-solubilized compounds was prepared in larger volumes, to be dispensed into vials as 5 mL final amounts. After pouring, plates and vials were cooled and solidified in the fume hood, with foil wrappings used to protect light-sensitive compounds. Treatment vials were stored in Ziplock bags at 4°C as needed.

For the chemical screen,10 female flies were transferred to each treatment well. A DMSO-solubilized rifampicin control was also run on every qPCR plate to confirm the ongoing capacity of Wolbachia to respond to compound treatments. After 3 days of chemical feeding, pools of 5 female flies were removed from each treatment well and processed as a group for wsp quantification using real-time qPCR. For the vial-based experiments that assessed drug lethality, flies were incubated in groups of 12, 6 females and 6 males per vial, with viability scored every 3 days. Flies were transferred to new treatment vials at day 6, using vials from the 4°C fridge that had been re-warmed. For vial-based experiments using DMSO and mifepristone, flies were incubated on treatment food in groups of 15 females and 5 males. Flies were transferred to new treatment vials every 3 days, using DMSO and mifepristone vials from the 4°C fridge that had been re-warmed. After 14 days of feeding were completed (Haselton et al., 2010; Serbus et al., 2015), pools of 5 female flies were removed from each vial and processed as a group for wsp quantification using real-time qPCR.

To incorporate Wolbachia into GAL4 driver lines, the driver males were crossed to virgin females of the genotype w; Sp/Cyo; Sb/TM6B, carrying the wMel Wolbachia strain (Christensen et al., 2016), which in this study is referred to as DB wMel. F1 progeny were backcrossed to the parental lines to establish Wolbachia-infected driver stocks, with the same genotypes as the originally uninfected lines.

Genetic disruptions were achieved using VALIUM20 transgenic RNAi lines which depend on short hairpin RNA, also known as artificial microRNAs, to trigger gene silencing in both somatic and germline cells (Ni et al., 2011). Each host pathway was tested by two different UAS-shRNA responder lines (Supplementary Table S2), selected in accord with pathways targeted by “hit” compounds from the chemical screen. To generate RNAi-expressing flies, wMel-infected virgin females were selected from freshly emerging bottles of each GAL4 driver stock. These females were crossed to males that carried responder UAS (upstream activating sequence) elements adjacent to a promoter that drives shRNA production (Ni et al., 2011). The parent flies were removed from the vials after 3–4 days of mating. Emerging F1 flies were collected in daily cohorts and aged for 5 days. The F1s that carried the GAL4 driver and the UAS responder were identified by phenotypic markers and separated within each cohort. Control siblings that contained either the GAL4 or UAS responder, but not both, were collected when available. In some cases, a separate control set was also generated in parallel by outcrossing Wolbachia-infected driver females to Oregon R (OreR) males. In all cases, control and treatment groups were generated and processed in parallel for wsp quantification.

Real-time qPCR was used to assess whole-body Wolbachia abundance, using the candidate gene wsp as a proxy for Wolbachia genomes per sample. Because Wolbachia reportedly carry one genome per bacterial cell (McGarry et al., 2004), resulting genome counts are expected to represent Wolbachia abundance per sample. DNA was extracted from pools of 5 female flies as per established methods (Christensen et al., 2019). Absolute measurements of the wsp gene from the extracted DNA samples were compared against reference plasmid standards, specifically a PGEM-T vector carrying a 160 bp PCR-amplified fragment of the wsp gene (Christensen et al., 2016). Real-time qPCR was carried out on a Bio-Rad CFX96 Connect Optics Module Real-Time System. Absolute wsp copy numbers were obtained by comparing cycle threshold (Ct) values to the standard curve generated from the plasmid standard. The wsp amplification primers were: Fwd 5’ CATTGGTGTTGGTGTTGGTG 3′ and Rev. 5’ ACCGAAATAACGAGCTCCAG 3′, used at 5 μM (Christensen et al., 2016).

Graphical displays showing “normalized” wsp counts as a scatter plot were created for display purposes only. To generate such graphs, median wsp counts for the DMSO controls per replicate were identified, then compared to the median wsp count of the entire dataset. A scaling factor was then identified and applied to each replicate, to normalize the median wsp value for the DMSO control and all associated experimental data. The raw absolute count data are available for review as needed (Supplementary material S1). Statistical analyses were conducted on raw (non-normalized) data within each experimental replicate for all experiments. Statistics appropriate to data normality and homogeneity were identified and applied as previously (Christensen et al., 2019). Power analysis was performed with an alpha set at 0.05 using a MATLAB-based data sub-sampling program, designed by Dr. Philip K. Stoddard. This program has the advantage that analyses can be customized to the statistical test appropriate to each dataset (Christensen et al., 2019). All statistical analysis worksheets for each experiment performed are also available (Supplementary material S1).

A literature search was first conducted to assess how intracellular bacterial abundance is regulated in commonly studied bacterial infections. After assessing 52 species from 17 genera, 26 bacterial species were identified, for which host gene/pathway effects on density regulation had been discussed (Supplementary Table S3). Of these, the literature highlighted 14 host mechanisms that altered the intracellular abundance of multiple bacterial classes (Supplementary Table S4; Supplementary material S2). Because these mechanisms were identified as more commonly involved in host–microbe interactions, they were prioritized for testing in the Wolbachia-Drosophila endosymbiosis model. Candidate compounds known to target each process were selected, with two or more compounds identified for testing each of the 14 classes of host targets. This culminated in the selection of 37 total candidate compounds to test for effects on whole-body Wolbachia titer (Table 1).

The impact of the candidate compounds on whole-body Wolbachia titer was screened by absolute quantification of wsp by real-time qPCR (Kim et al., 2008; Serbus et al., 2012; Markstein et al., 2014; Christensen et al., 2019). D. melanogaster flies, carrying the endogenous wMel Wolbachia strain, termed DB wMel, were exposed for 3 days to food supplemented with DMSO-solubilized drugs or DMSO alone as a control. Treatments were initially tested for impact on whole-body wsp across two independent plate replicates. Those treatments which significantly changed wsp abundance in both plates were identified as preliminary hits. Of 37 chemicals tested, the primary screen identified 16 compounds were identified as meeting this criterion. These preliminary hit compounds were re-tested for reproducibility in a third plate replicate. 11 compounds were reconfirmed as hits, implicating a total of 9 host pathways and processes (Figure 1A). Most “hit” compounds elicited an increase in whole-body wsp abundance, with median values ranging 6–57% higher than the control (p < 0.001–0.036, n = 6 per plate replicate). The only exception was bortezomib, which reduced wsp to 48–71% of the DMSO control (p < 0.001, n = 6 amplifications per plate replicate) (Figure 1A; Table 2).

To investigate a role for candidate processes across systems, the hits from DB wMel were retested against the D. simulans (Dsim) model, which naturally carries the wRi Wolbachia strain. The Dsim wRi re-screen identified a subset of 6 compounds that significantly affected whole-body wsp counts across 3 plate replicates (Figure 1B). 5 compounds increased whole-body wsp abundance to 15–52% higher than the DMSO control (p-value range: <0.001–0.041, n = 6 per plate replicate). These hits were associated with host Imd signaling, Calcium signaling, Ras/mTOR signaling, and Wnt signaling functions. By contrast, the proteasome inhibitor bortezomib continued to reduce wsp counts to 56–69% of the DMSO control (p < 0.001, n = 6 amplifications per plate replicate) (Figure 1B; Table 2).

To further investigate why some chemical treatments increase wsp counts, but others suppress wsp, a lethality assay was conducted. Flies were exposed to each of the “hit” compounds for a 12-day period. Bortezomib induced high lethality by the 6-day exposure timepoint for both DB wMel and Dsim wRi (Figures 2A,C). Thus, it is possible that wsp reductions by bortezomib reflect a Wolbachia response to toxic host conditions. However, flies exposed to all other “hit” compounds exhibited viability profiles comparable to DMSO controls, as per the example of IWR-1 (Figures 2B,D) (Supplementary Figure S1). Because the non-lethal “hit” compounds were all shown to elevate wsp counts, these results suggest that functions of multiple host pathways normally reduce whole-body Wolbachia loads.

To confirm the basis for host cellular effects on whole-body Wolbachia titer, genetic disruption experiments were performed, focusing on the pathways dually implicated by chemical screening of DB wMel and Dsim wRi. We used the GAL4::UAS system in D. melanogaster, which enables directed manipulation of gene expression (Brand and Perrimon, 1993; Duffy, 2002). In this case, the GAL4::UAS system was set to drive expression of short hairpin RNAi suppress the corresponding gene product (Perrimon et al., 2010; Ni et al., 2011; Perkins et al., 2015). The wMel strain was crossed into well-established GAL4 driver lines that drive constitutive whole-body expression, including the reputedly “strong” Actin5C-GAL4 driver (Act-5C), and the “milder” daughterless-GAL4 driver (da-GAL4) (Supplementary Table S5). Few to no F1 progeny were recovered that carried Act5C-GAL4 as well as UAS-shRNA chromosomes, indicating lethality for such genetic combinations. However, crossing the UAS-shRNA lines to da-GAL4 (Serbus et al., 2015) yielded ample RNAi-expressing F1 progeny for analysis.

No changes in wsp abundance were detected in response to constitutive shRNA disruption of Calcium signaling by knockdown of L-type calcium channels encoded by Ca-alpha1D and Cac. Inconsistent effects on wsp abundance were associated with disruptions to the Imd pathway by knockdown of NF-kappa-B/Rel, and the Rel activator, Tak1. Similar inconsistencies were observed for knockdown of the Wnt pathway gene, shaggy (sgg) gene (Supplementary material S1).

Constitutive shRNA disruption of the Wnt pathway gene armadillo (arm) yielded positive effects, increasing median whole-body wsp counts to 9–15% above the OreR-outcrossed control (p-value range: <0.001–0.034, n = 6 per plate replicate) (Figure 3). A significant wsp increase was also detected for the Ras/mTOR pathway, with tor disruption flies exhibiting higher whole-body wsp counts at 23–31% above both sibling controls and OreR-outcrossed controls (p < 0.001, n = 6 per plate replicate) (Figure 3). Ras/mTOR signaling was also retested by knockdown of the Epidermal Growth Factor Receptor (EGFR). Compared to OreR-outcrossed controls, EGFR knockdowns did not consistently affect wsp abundance measurements. However, in comparison to sibling controls, EGFR disruption yielded a 35–44% increase in median wsp counts (p-value range: <0.001–0.002, n = 6 per plate replicate) (Figure 3). Thus, detection of wsp responses to EGFR disruption may be context-dependent.

To confirm an effect of Wnt and Ras/mTOR pathways on Wolbachia, the most consistent outcomes from the da-GAL4::UAS-shRNA experiments were retested. Arm RNAi elicited a 16–50% increase in median wsp abundance over the OreR-outcrossed control (p < 0.001, n = 18) (Figure 4A). Power analysis indicated the arm RNAi outcome to be robust (β < 0.003 at n ≥ 12; total n = 18) (Figure 4B). tor RNAi triggered a 38–39% increase in wsp abundance as compared to OreR-outcrossed controls (p < 0.001, n = 18) (Figure 4C), a finding also well-supported by power analysis (β < 0.002 at n ≥ 4; total n = 18) (Figure 4D). Taken together, these data indicate that constitutive RNAi disruption of Wnt and Ras/mTOR signaling increases whole-body Wolbachia titer.

An intrinsic limitation of certain genetic disruption approaches, like constitutive RNAi induction, is that cumulative disruption effects could occur. To confirm ongoing Wnt and mTOR pathway effects on whole-body wsp abundance, “Gene-switch” GAL4 driver flies were used to induce GAL4 activity in adult flies. The Gene-switch version of GAL4 carries an inhibitory domain that blocks GAL4 function, until a de-repressor compound, mifepristone, is added (Roman et al., 2001). Because trial experiments on mifepristone identified low-power but potentially significant effects on wsp (Supplementary Figure S2), all GS-GAL4 experiments were carried out with multiple controls. Flies carrying GS-GAL4::UAS-shRNA genotypes were always compared to non-expressing siblings, with half of the flies DMSO-treated, and the other half exposed to DMSO-solubilized mifepristone.

In tests of arm and tor RNAi knockdowns, no significant wsp abundance differences were observed between the DMSO-treated flies and mifepristone-treated flies that were incapable of RNAi expression (Figures 5A,B). However, mifepristone-fed flies that were capable of shRNA expression did show significant differences in their wsp counts. In the case of GS-GAL4::arm-shRNA, the mifepristone-treated flies exhibited reduced wsp counts, down to 45–71% of all other conditions tested in parallel (p-value range: <0.001–0.033, n = 9) (Figure 5A). By contrast, mifepristone-treated GS-GAL4::tor-shRNA flies carried 81–184% more wsp than all other conditions run in parallel (p-value range: <0.001–0.031, n = 9) (Figure 5B). These results confirm ongoing Wolbachia sensitivity to Wnt and mTOR disruption in adult hosts. Unlike the da-GAL4 data, the GS-GAL4 results notably show that Wnt and mTOR exert opposing effects on Wolbachia titer. This highlights a functional difference between constitutive and adult-specific disruptions of the Wnt pathway with respect to regulation of Wolbachia titer in adult insects.

One way to reconcile effects of arm and tor disruptions on wsp abundance is to consider the possibility that both may affect a consensus target relevant to Wolbachia. Literature indicates that Wnt signaling can suppress autophagy onset via multiple routes (Pérez-Plasencia et al., 2020), including through down-regulation of Beclin-1, also known as ATG6 (Tao et al., 2017). mTORC1 is also known to inhibit ATG6 by suppressing the ATG6 activator, ULK1 (Hill et al., 2019). To test the effect of ATG6 on whole-body Wolbachia titer, the GS-GAL4::ATG6-shRNA flies were generated. The mifepristone-fed, RNAi-expressing condition exhibit 67–173% higher wsp levels than non-expressing mifepristone-fed siblings and DMSO-fed flies of equivalent genotype (p-value range: <0.001–0.045, n = 9) (Figure 5C). These data suggest that ATG6 normally suppresses whole-body wsp abundance, consistent with autophagy as a general suppressor of Wolbachia titer (Voronin et al., 2012; Strunov et al., 2022). Implications for Wnt and mTORC1 pathway interaction with autophagy are discussed below.