Toll-1-dependent immune evasion induced by fungal infection leads to cell loss in the Drosophila brain

It has previously been reported that exposure to fungal volatiles is sufficient to reduce life span and climbing and induce neurodegeneration in Drosophila [4,5]. To ask whether and how the fungus B. bassiana – which is well-known to activate Toll signalling – could affect the brain, we mimicked natural exposure of flies using an infection chamber (Fig 1A). Here, a B. bassiana (strain 80.2) fungal culture was established in a fly bottle, flies were inserted and shaken to get them directly exposed to spores, and thereafter kept there for the duration of the experiments. Wild-type non-infected control flies lived up to 70 days, but flies exposed to B. bassiana died within less than 20 days and by day seven more than half of the flies had died (Fig 1B). Negative geotaxis—also known as startle-induced negative geotaxis (SING) or the climbing assay—is commonly used to measure locomotor impairment caused by neurodegeneration [63,64]. No effect was seen after exposure for three days to B. bassiana, but seven days of exposure impaired climbing (Fig 1C). Altogether, these data showed that exposure to B. bassiana decreased locomotion and longevity. Multiple factors could directly and indirectly affect both locomotion and survival in diseased flies after infection. Nevertheless, these phenotypes are also common indicators of neurodegeneration in flies.

Fig 1. B. bassiana decreased longevity and impaired climbing.

(A) Infection chamber. (B) B. bassiana exposure reduced life span of wild-type (Oregon/CantonS) flies. Percentage surviving flies at each scored day Log-rank test p n = 104 non-infected flies, and infected flies n = 104. (C) B. bassiana exposure impaired climbing of wild-type (Oregon/CantonS) flies, after 7 days of exposure. Dot plots, with lines indicating the mean and standard deviation. Each dot represents one cohort of 7–10 flies. Two-way ANOVA group statistics: 3-days-after infection versus 7-days-after infection (Row factor) p = 0.0122; Noninfected versus infected (Column factor): p p = 0.0048. Asterisks in the figure show Turkey’s multiple comparison correction tests. Sample sizes: 3 days after infection: n = 60 for each infected and non-infected. 7 days after infection: n = 51 for each infected and non-infected. **p p S5 Table Source data.

https://doi.org/10.1371/journal.pbio.3003020.g001

We wondered whether B. bassiana could affect brain function by entering the brain. After three days exposure of wild-type flies to B. bassiana, fungal cells were detected within the brain using the FM4–64 dye together with nuclear DAPI (Figs 2A and S1A). A second dye, calcofluor white (CLW) also revealed fungal cells in the optic lobes and central brain (Fig 2B). Finally, adult wild-type flies were exposed to a GFP-transgenic B. bassiana strain (EABb 04/01-Tip GFP5) [65], and abundant GFP + cells were found within the optic lobes and central brain (Figs 2C–F, S1B and S1C). Interestingly, B. bassiana-GFP cells were found at the entry point of the proboscis into the brain (Fig 2C and 2E) and at the exit point of the esophagus (Fig 2E). B. bassiana-GFP cells were also found across the blood-brain barrier (BBB) (Fig 2D), which is formed of glial cells in Drosophila. Furthermore, germlings or germinating spores, with clear filaments or germtubes were also abundant in the brain (Fig 2F-F”). These findings show that B. bassiana infiltrated the adult brain. To enter the brain, B. bassiana would either have to damage the BBB or be carried across by macrophages [66]. The dye Dextran Red is commonly used to test the integrity of the BBB in Drosophila [67]. Following an injection into the thorax, Dextran Red outlines the retina’s edge if the BBB is intact, whereas if it is not, the dye leaks into the retina [67]. We found that in flies that had been exposed to B. bassiana for seven days, Dextran Red spread within the retina (Fig 2G), meaning that the BBB was damaged. Altogether, these findings showed that following exposure, B. bassiana penetrated the fruit-fly brain.

Fig 2. B. bassiana was detected within the fly brain.

B. bassiana infiltrated the Drosophila adult brain of wild-type flies (Oregon R) as visualised with (A) FM4–64 dye, revealing fungal cells as they co-localise with the nuclear dye DAPI+ (arrows). (B) Calcofluor white, revealing fungal cells adjacent to neurons within the central brain (arrows). (C) Transgenic B. bassiana-GFP found at the point of entry of the proboscis into the brain (arrows). Orthogonal views of one optical section. Nuclei labelled with DAPI. (D) DAPI labelled nuclei reveal the outer surface of the brain, including the BBB and B. bassiana-GFP crossing the BBB (sagittal and transverse views) and within the brain (transverse view). Orthogonal views of one optical section. (E) B. bassiana-GFP found at the entry point of the proboscis into the brain (arrow, anterior brain) and exit point of the oesofagus from the brain (arrow, posterior brain). (F-F’’) B. bassiana-GFP germlings or germinating spores colocalise with the neuronal marker DAPI (arrows, F’) and have filaments (arrows, F’’). (G) B. bassiana damaged the blood-brain barrier in the eye, as Dextran Red could diffuse within the retina in infected brains (arrows) but not in non-infected controls. Scale bars: (A, B, F) left: 100 mm; (A, B) right: 7 mm; (C, D) 50 μm, (E) 20 μm; (F’): 10 μm; (F’’): 5 μm; (G) 1,023 μm. Sample sizes: (1) FM4–64: n = 4/10 infected brains had fungal cells in the brain; non-infected n =  7. (2) CLW: n = 2/2 brains had fungal cells in the brain. (3) B. bassiana-GFP: n = 3/3 infected brains had spores in the brain; non-infected controls: n = 3. Non-infected controls for (A, C, D, E) in S1 Fig. Genotype of Drosophila melanogaster for all images: wild-type Oregon. Genotype of B. bassiana: 80.2 (A, B and F) and B. bassiana-GFP: EA-Bb-Tip 04/01 (C–E). See S1 Table for statistical details.

https://doi.org/10.1371/journal.pbio.3003020.g002

Fungi can degrade insects’ cuticle, enabling them to penetrate hosts’ bodies [68–71]. In this way, B. bassiana could invade the brain through the optic lobes, the head capsule and the proboscis. The proboscis is used for feeding [72], and it traverses the brain to join the oesophagus, thus providing a route for B. bassiana into the brain. However, B. bassiana could trigger avoidance in the fly, like in C. elegans worms, where pathogenic bacteria induce the worms to turn away [73,74]. Thus, we tested whether flies feed on B. bassiana spores. We offered them water or sucrose or spores mixed with blue dye and measured optical density (OD) in their abdomens after feeding. Flies fed on sucrose—which they are known to like—and similarly fed on spores, more than on water (Fig 3A). Thus, ingestion would provide B. bassiana with an effective route into the brain. To further test this, we next fed flies that had not been previously exposed to the fungus, with water containing B. bassiana-GFP spores. Following feeding, we detected B. bassiana-GFP in the brain (Fig 3B), in the proboscis (Fig 3B’), germlings (Fig 3B’’), B. bassiana-GFP at the point where the proboscis enters the brain (Fig 3B’’’ anterior), at the point where the esophagus and neuropiles exit the brain (Fig 3B’’’ posterior) and over glial membranes deep within the brain (Fig 3B’’’’). As these flies had not been exposed to B. bassiana-GFP other than through feeding, this means that ingested B. bassiana spores could penetrate the brain.

Fig 3. Exposure to B. bassiana activated Toll signalling within the adult brain.

(A) Drosophila wild-type (Oregon R) flies fed on B. bassiana spores that had been mixed with blue dye diluted in water. One Way ANOVA P value (B) Non-infected repo>myr-tomato flies were fed a solution of B. bassiana-GFP spores, after which GFP + cells were found: (B) in the central brain, (B’) in the proboscis, (B’’) germling, (B’’’) in the entry site of the proboscis (anterior) into the brain, and exit site of esophagus (posterior), and (B’’’’) on glia within the brain. (C) Toll-1 > FlyBow and sarm^NP0257 > FlyBow reveal expression in the proboscis, note axons and dendrites of sensory neurons in the labellum (arrows). n = 3–7 brains. (D) Activating Sarm+ neurons triggered the proboscis extension response, a proxy for feeding. (E) Quantification of (D). Chi-square test, ****p sarm^NP0257>TrpA1 n = 34 flies, Control was responder line crossed to wild-type Oregon: UAS-TrpA1/+ n=34 Test: sarm^NP7460>TrpA1 n = 40, Control: UAS-TrpA1/+ n = 40. 40. (F, G) B. bassianaB. bassiana infection of wild-type (Oregon R) flies for 7 days raised the expression levels of AMPs, wek and sarm within the brain. qRT-PCR data comparing fold change levels of (F) drs and mtk mRNA and (G) wek and sarm in infected brains, normalized to GAPDH as a housekeeping control. DeltaCT mean ± standard deviation. drs: Unpaired Student t test, *p mtk: p = 0.0634; n = 3 b.r.; wek: Unpaired Student t test ****p sarm: t test: *p ” stands for GAL4/UAS and “+” stands for wild-type chromosome. See S4 Table for further details. The data underlying panel A, E, F and G can be found in S5 Table Source data.

https://doi.org/10.1371/journal.pbio.3003020.g003

The findings that B. bassiana entered the brain and flies fed on it were surprising. Firstly, in Drosophila, the innate immune response driven by Toll-1 should clear the fungus [15]. Secondly, they contrast with the bacterial avoidance response of C. elegans, which depends on Tol-1 [73,74]. We reasoned that perhaps the proboscis might not express Toll-1. Thus, we asked whether Toll-1 and sarm might be expressed in the proboscis. Using the in vivo reporters Toll-1 > FlyBow and sarm^NP0257 > FlyBow, we found expression of Toll-1 and sarm in sensory neurons of the proboscis (Fig 3C). We tested the consequences activating Sarm+ neurons might have on the proboscis extension response (PER), which is required for feeding [72]. Using two distinct sarmGAL4 drivers, we found that activating Sarm neurons with TrpA1 increased the incidence of PER events compared to unstimulated controls (Fig 3D and 3E). This showed that sarm-expressing neurons are not involved in avoidance of B. bassiana, and instead could be involved in feeding.

Next, we wondered whether B. bassiana could activate Toll signalling inside the brain. All the Toll-1 signalling components – including GNBP3, the downstream SPE protease, spz-1, Toll-1 plus six other Toll receptors, MyD88, sarm, Dif/NF-κB and target antimicrobial peptide genes drs and mtk – are normally expressed in the adult brain (Scope Fly Atlas) [36,75]. We visualised the expression of Toll-1, MyD88 and sarm in the brain, showing they are widely expressed (S2A Fig). Using the DenMark reporter which labels cell membranes, MyD88 revealed the BBB ensheathing the brain (S2B and S2B’ Fig). At least some of the MyD88 + cells at the BBB were glia, as when labelled with nuclear Histone-YFP, they co-localised with the nuclear glial marker Repo (S2C and S2C’ Fig). Consistent with this, we had previously shown that MyD88+ cells include both neurons and glia [36]. Importantly, Toll-1, MyD88 and sarm are all expressed in cells that surround the entry (anterior brain) and exit (posterior brain) sites of the proboscis and oesophagus, respectively, through the brain (S2D Fig). At least some of these MyD88 + cells surrounding the tunnel across the brain are glia, as they colocalise MyD88 > his-YFP and Repo (S2E Fig). This means that as B. bassiana travelled from the proboscis into the brain, it would encounter abundant cells expressing Toll-1 and its signalling machinery.

Thus, next we asked whether exposure of wild-type flies to B. bassiana might activate innate immunity in the brain. At seven days post-infection, expression of AMPs drs mRNA was upregulated within the brain, and that of mtk also, albeit not significantly (Fig 3F). The mechanism of activation of Toll-1 is very well known. Instead, we wondered whether B. bassiana could also drive an immune evasion pathway downstream of Toll-1, via Wek with Sarm. Intriguingly, following infection, the expression of both wek and sarm was also upregulated in the brain (Fig 3G).

To conclude, B. bassiana could enter the brain also through feeding, and it induced Toll signalling within the adult brain, also engaging Wek and Sarm. Wek is not required for innate immunity [33], and it links Tolls to Sarm instead [34]. Sarm inhibits MyD88 and immune signalling. Toll signalling via Wek and Sarm induces apoptosis downstream [34], and Sarm also induces axonal destruction [46]. Thus, the infection-dependent up-regulation of wek and sarm downstream of Toll-1 could potentially drive neurodegeneration.

To ask whether B. bassiana could induce neurodegeneration via Toll-1, Wek and Sarm, we first tested whether exposure to B. bassiana caused cell loss in the brain. We used the nuclear reporter histone-YFP to visualize Sarm + cells (sarm^NP0257 > hisYFP) and quantified them automatically. Seven days exposure to B. bassiana decreased Sarm + cell number in the central brain (Fig 4A and 4B). As glial cells form the BBB in Drosophila, and at last, some of them are MyD88 + (S2B-E Fig) [36], we asked whether exposure to B. bassiana could affect glia. Glial cells were labelled with pan-glial anti-Repo antibodies, glial cell number in the brain was quantified automatically, and it also decreased with infection (Fig 4C and 4D). Importantly, loss of glial cells could lead to BBB breakdown and facilitate entry of B. bassiana into the brain. Exposure of fruit-flies to the fungal volatile 1-octen-3-ol caused loss of dopaminergic neurons [4]. Thus, we asked whether exposure to B. bassiana could elicit similar effects. We first tested whether seven-day exposure to B. bassiana might affect the expression of tyrosine hydroxylase (TH), which is required for dopamine synthesis [76]. In fact, B. bassiana exposure caused a decrease in TH mRNA levels within adult brains (Fig 4E), meaning that B. bassiana exposure decreased dopamine production. Furthermore, using anti-TH antibodies, we found that at seven days post-exposure, the number of PPL1, PPL2, PPM1/2, PPM3 (Fig 4F–I), and PAM (Fig 4I, 4J and 4K) dopaminergic neurons had decreased.

Fig 4. Exposure to B. bassiana caused cell loss in the fly brain.

(A,B) Seven-day exposure to the B. bassiana caused loss of sarm^NP0257 > HisYFP cells. Quantification in (B), Mann-Whitney U test, p = 0.002. (C,D) Seven-day exposure to B. bassiana caused loss of glia cells, visualised with anti-Repo antibodies. Quantification in (D), Unpaired Student t test, p = 0.002. (E) Seven-day exposure to B. bassiana reduced tyrosine hydroxylase (TH) expression. qRT-PCR showing fold-change relative to non-infected wild-type controls and normalised to GAPDH, mean ±  standard deviation. Unpaired Student t test on delta-Ct values p = 0.0011, n = 3 biological replicates. (F) Seven-day exposure to B. bassiana caused loss of dopaminergic neurons (DANs) visualised with anti-TH antibodies, in the posterior brain, quantification in (H): PAL, SP1, T1: Mann-Whitney U tests; PPL1, PPM1/2, PPL2, PPM3, SVP, VUM: Student t test; non-infected brains n = 12, infected brains n = 12. Arrows point to each DAN class. (G,I) Illustration of dopaminergic neurons in the adult brain. (J,K) Seven-day exposure to the B. bassiana caused loss of PAMs, quantification in (K): unpaired Student t test, p = 0.0043, non-infected brains n = 7, infected brains n = 6. Graphs in panels (B,D,K) show box-plots around the median, box with 50% of data points and whiskers with 25% of data points. Graphs in panels (E,H) show bar charts with mean ±  standard deviation. Dotted lines in (A,C) indicate ROI for automatic cell counting with DeadEasy software. **p p SarmNP0247 > hisYFP); (C-F, H, J): wild-type Oregon R. Controls: non-infected flies of the same genotypes. See S4 Table for further details. The data underlying panels B, D, E, H and K can be found in S5 Table for source data.

https://doi.org/10.1371/journal.pbio.3003020.g004

Toll-1 signalling can induce cell death via the Sarm/JNK pathway [34], so we next asked whether B. bassiana-induced cell loss depended on Toll-1. Toll-1 activates innate immunity through the well-known MyD88 pathway, which leads to the activation of Dif/NFκB and the expression of anti-microbial peptides (e.g., drs, mtk) [77]. Toll-1, MyD88 and sarm are widely expressed in the brain, and at least MyD88 also including in the BBB [36] (S2 Fig). Sarm is the inhibitor of MyD88 [34,39]. So in the brain, in MyD88+ cells, Toll-1 could potentially drive signalling via the two alternative MyD88 and Sarm pathways, in co-expressing cells. Thus, to account for both signalling routes, we visualised adult MyD88+ cells with MyD88 > HisYFP and counted cell number automatically. Seven-day exposure to B. bassiana caused MyD88+ cell loss (Fig 5A and 5C). To knock-down Toll-1 expression specifically in adult flies, we used the temperature-sensitive Gal4 repressor tubulin-Gal80^ts (Fig 5B, tubGAL80^ts, MyD88 > hisYFP). In non-infected controls, adult Toll-1 RNAi knockdown using line UAS-Toll-1RNAi^kk/100078 (shown to work effectively in [36]) caused an increase in MyD88-YFP + cells within the central brain (Fig 5A–C). This increase in cell number could correspond to induced proliferation or increased cell survival. In fact, Toll-1 can induce apoptosis in at least larvae and pupae [34]. There, Toll-1 loss of function increased neuronal number, and constitutively active Toll-1^10b decreased neuron number and increased apoptosis [34]. This suggests that the increase in cell number caused by Toll-1-RNAi knock-down could result from decreased cell death. Importantly, in infected adult brains, Toll-1 RNAi knockdown prevented loss of MyD88-YFP + cells that would have been caused by B. bassiana exposure (Fig 5A–C). These data mean that B. bassiana-induced cell loss depends on Toll-1 signalling.

Fig 5. B. bassiana-induced cell loss depends on Toll-1.

(A) Seven-day exposure to B. bassiana caused loss of MyD88+ cells, visualised with tubGAL80^ts; MyD88 > HisYFP, and this was rescued with adult-specific Toll-1-RNAi ^KK/100078 knockdown. (B) Diagram explaining the experimental temperature regime. (C) Quantification of data in (A). Two-way ANOVA: Infected versus not-infected: p = 0.7557; Genotypes: p p n = 10, infected control brains n = 9, non-infected Toll-1-RNAi brains n = 9, infected Toll-1RNAi brains n = 9. (D, E) Seven-day exposure to B. bassiana caused loss of glial cells visualised with anti-Repo antibodies (D), which was rescued with adult-specific Toll-1-RNAi ^KK/100078 knockdown, automatic quantification with DeadEasy in (E). Two-way ANOVA: Infected versus not-infected: p = 0.1003; Genotypes: p = 0.1285; interaction: p = 0.0666. Tukey’s multiple comparisons correction test. Sample sizes: non-infected control brains n = 11, infected control brains n = 8, non-infected Toll-1-RNAi brains n = 6, infected Toll-1-RNAi brains n = 6. (F,G) Seven-day exposure to B. bassiana caused loss of PPM3 DANs (arrows), visualised with anti-TH antibodies and this phenotype was rescued with adult-specific Toll-1-RNAi ^KK/100,078 knockdown. Arrows point to PPM3 neurons. Quantification in (G). Two-way ANOVA: Infected versus not-infected: p = 0.0042; Control versus Toll-1 RNAi: p = 0.8548; interaction: p = 0.1760, and Tukey’s multiple comparisons correction test. Sample sizes: non-infected control brains n = 8, infected control brains n = 9, non-infected Toll-1-RNAi brains n = 7, infected Toll-1-RNAi brains n = 6. (H) Illustration showing how adult-specific knock-down of Toll-1 expression prevented cell death after infection. (I, J) Adult restricted knock-down of Toll-1 in MyD88 cells improved (I) survival, (J) and climbing, at 7 days post-infection (n = 82−194 flies). (I) Two-way ANOVA survival: interaction: p = 0.0458; Genotype comparisons: p = 0.0495; Infected versus not-infected: p (J) Two-way Anova climbing: interaction: p = 0.0017; Genotype comparisons: p = 0.0177; Infected versus not-infected: p p p p MyD88 GAL4 hisYFP/ + tubGAL80^ts/ + flies, whereby the GAL4 driver line was outcrossed to wild-type. Test flies: MyD88 GAL4 hisYFP/UAS-Toll-1RNAi ^kk/100,078; tubGAL80^ts/+. Infected versus not-infected compares flies of the same genotypes. See S4 Table for further details. The data underlying C, E, G, I, and J can be found in S5 Table Source data.

https://doi.org/10.1371/journal.pbio.3003020.g005

Neuronal loss can cause glial loss [78], MyD88 is also expressed in Repo+ glial cells (S2B-D Fig) [36] and Toll-1 is also expressed in glial cells [75]. Thus, we tested the effect on glial cell number. Toll-1-RNAi knock-down did not affect the number of Repo+ glial cells in non-infected controls (Fig 5D and 5E). However, Toll-1 RNAi knock-down in MyD88+ cells prevented the decrease in glial cell number caused by B. bassiana infection (Fig 5D and 5E). Thus, B. bassiana-induced glial cell loss requires Toll-1 signalling in MyD88+ cells. As MyD88 > hisYFP+ cells include glial cells of the BBB (S2B-E Fig), by inducing Toll-signalling dependent glial cell loss, B. bassiana could make holes in the BBB, making its way into the brain.

We also tested the effect of Toll-1 RNAi knock-down in dopaminergic neurons (DANs). Toll-1 is expressed in DANs, as there was colocalization of Toll-1 > HistoneYFP and anti-TH within multiple DAN clusters, including PAL, PPM3, PPL2, and PPL1 (S3A Fig), but in very few neurons within the PAM clusters (S3A Fig). Following Toll-1 RNAi knock-down in MyD88 + cells, in non-infected flies, Toll-1 knockdown did not alter DAN cell number (Fig 5F and 5G). By contrast, Toll-1 RNAi knockdown prevented infection-induced neuronal loss within the TH + PPM3 and PPL1 DAN clusters (Figs 5F, 5G and S3B). Thus, DAN degeneration upon B. bassiana infection required Toll-1 (Fig 5H).

Finally, we tested whether acute adult-restricted Toll-1 knock-down could rescue the behavioural impairments caused by B. bassiana infection. We found that Toll-1 knock-down in MyD88 cells rescued fly survival after 7 days of infection compared to non-infected controls, albeit not significantly when compared to infected RNAi controls (Fig 5I). By contrast, Toll-1 knock-down rescued the climbing impairment caused by B. bassiana infection, compared to infected genetic controls, but did not achieve the normal climbing performance of non-infected control flies (Fig 5J). Altogether, acute adult-restricted Toll-1 RNAi knock-down improved fly survival and locomotion after infection. The rescues are consistent with the findings that Toll-1 knock-down in adult flies prevents B. bassiana-induced neurodegeneration, whilst the incomplete nature of the rescues can be explained as immunity would be compromised in these flies.

Apoptotic Toll-1 signalling requires the adaptor Wek, which binds Sarm, which induces cell death [34]. As B. bassiana-induced the upregulation of wek and sarm in the fly brain (Fig 3G), and B. bassiana-induced cell loss depended on Toll-1, we asked whether wek and sarm might also be required for infection-dependent cell loss.

First, we tested wek. In non-infected controls, adult-specific wek knock-down using line UAS-wek ^MH046534-RNAi (shown to work effectively in [34,36]) had no effect on MyD88+ cells (Fig 6A and 6C). However, wek-RNAi knock-down rescued B. bassiana-induced MyD88-HisYFP + cell loss (Fig 6A and 6C). In non-infected control flies, wek-RNAi knockdown did not alter the number of Repo+ glial cells in the brain (Fig 6D and 6E). However, no difference was found in Repo+ glial cell number between infected and non-infected brains upon wek-RNAi knock-down (Fig 6D and 6E). These data mean that B. bassiana could not induce MyD88+ and Repo+ cell loss without Wek.

Fig 6. B. bassiana-induced cell loss requires Wek, but Wek has pleiotropic functions.

(A) Loss of MyD88 > HisYFP+ cells caused by seven-day exposure to B. bassiana was rescued with adult-specific wek-RNAi^MHC046534 knockdown, and cell number increased further. (B) Diagram explaining the experimental regime. (C) Automatic quantification of data in (A). Two-way ANOVA: Infected versus not-infected p = 0.2327; Genotypes: p p n = 14, infected control brains n = 15, non-infected wek-RNAi brains n = 8, infected wek-RNAi brains n = 9. (D, E) Loss of Repo + glial cells caused by seven-day exposure to B. bassiana was rescued with adult-specific wek-RNAi^MHC046534 knockdown, automatic quantification with DeadEasy in (E). Two-way ANOVA: Infected versus not-infected p = 0.0236; Genotypes: p = 0.7775; Interaction: p = 0.0231, and Tukey’s multiple comparisons correction test. Sample sizes: non-infected control brains n = 11, infected control brains n = 9, non-infected wek-RNAi brains n = 10, infected wek-RNAi brains n = 10. (F, G) Adult-specific wek-RNAi^JF01681 knockdown decreased the number of TH + PPM3 DANs (arrows in F), suggesting that wek may be required for their differentiation. In fact, wek-RNAii^JF01681 knockdown did not rescue the cell loss caused by B. bassiana infection, but the infection did not reduce cell number further either. Two-way ANOVA: Infected versus not-infected: p = 0.3098; Genotypes: p = 0.0016; Interaction p = 0.0005, followed by Tukey’s multiple comparisons correction test. Sample sizes: non-infected control brains n = 8, infected control brains n = 9, non-infected wek-RNAi brains n = 7, infected wek-RNAi brains n = 9. (H) Diagram of Toll signalling pathway, whereby Wek links Toll-1 to Sarm, enabling Toll signalling to cause cell death and cell loss via Sarm, and this is rescued in some cells with wek RNAi knock-down. Wek also has functions independently of Sarm and MyD88, promoting adult neurogenesis. (I,J) wek–RNAi knock-down in MyD88 cells had no effect on survival at 7 days post-infection (I), nor in climbing (J) (n = 73–194). Two Way ANOVA survival: interaction: p = 0.2017; Genotype comparisons: p = 0.2437; Infected versus not-infected: p (J) Two Way Anova climbing: interaction: p = 0.7414; Genotype comparisons: p = 0.2014; Infected versus not-infected: p p p p p MyD88 GAL4 hisYFP/ + tubGAL80^ts/ + flies, whereby the GAL4 driver line was outcrossed to wild-type. Test flies: MyD88 GAL4 hisYFP/UAS-wek-RNAi^{TRIPHMC045534}; tubGAL80^ts/ + . Infected versus not-infected compares flies of the same genotypes. See S4 Table for further details. The data underlying panels C, E, G, I and J can be found in S5 Table Source data.

https://doi.org/10.1371/journal.pbio.3003020.g006

By contrast, wek knock-down in non-infected adult flies caused a significant decrease in the number of DANs of the PPL1, PPL2 and PPM3 clusters (Figs 6F, 6G and S4). This suggested that Wek is required to maintain DAN cell survival or promote neurogenesis or differentiation (e.g., TH expression) in the adult brain. These data are consistent with the known pleiotropic functions of wek, including in the adult brain [33,34,36]. Importantly, in wek-RNAi knock-down fruit-flies, B. bassiana infection failed to induce further neuronal loss (Figs 6F, 6G, S4A and S4B). These data revealed that in the absence of Wek, some DANs do not develop normally and only a few DANs remain, but these are not susceptible to further damage by B. bassiana infection. Finally, we tested whether wek RNAi knock-down could alter the behavioural impairments caused by exposure to B. bassiana. We found that wek knock-down in MyD88+ cells improved fly survival after seven days of infection, albeit not significantly (Fig 6I), and it did not rescue climbing (Fig 6J), which could be explained by DAN loss. Altogether, these data show that Wek is required for B. bassiana-induced loss of Myd88+ , Repo+ and TH+ cells (Fig 6H).

Next we tested sarm. In the absence of infection, when we downregulated sarm expression with RNAi in adult MyD88 cells (using line UAS-sarm-RNAi^JF01681, shown to work effectively in [34]), there was no effect (Fig 7A–C). However, RNAi knock-down of sarm rescued B. bassiana-induced MyD88 > hisYFP+ cell loss (Fig 7A–C). Similarly, sarm RNAi knock-down had no effect on glial cell number in the adult brain, but it rescued the Repo+ cell loss caused by B. bassiana infection (Fig 7D and 7E). And sarm RNAi knock-down did not affect PPM3 cell number in non-infected controls, but it prevented PPM3 DAN cell loss caused by B. bassiana infection (Fig 7F and 7G). Finally, sarm RNAi knock-down caused a mild albeit not significant increase in fly survival after seven days of infection (Fig 7I), and it improved climbing, albeit not significantly (Fig 7J). Altogether, these data showed that sarm is required for B. bassiana-induced loss of MyD88-HisYFP+ , Repo+ and TH+ PPM3 cells (Fig 7H).

Fig 7. B. bassiana-induced cell loss requires sarm.

(A) Seven-day exposure to the B. bassiana caused loss of MyD88 > HisYFP + cells, and this was rescued with adult-specific sarm-RNAi^JF01681 knockdown, automatic quantification in (C). (B) Diagram explaining the experimental regime. (A, C) Two-way ANOVA: Infected versus not-infected p = 0.0077; Genotypes: p p n = 14, infected control brains n = 15, non-infected sarm-RNAi brains n = 8, infected sarm-RNAi brains n = 9. (D, E) Seven-day exposure to B. bassiana caused loss of Repo + glial cells (D), and this effect was rescued with adult-specific sarm-RNAi^JF01681 knockdown, automatic quantification with DeadEasy in I. Two-way ANOVA: Infected versus not-infected p = 0.0205; Genotypes: p = 0.3649; Interaction: p = 0.0295, followed by Tukey’s multiple comparisons correction test. Sample sizes: non-infected control brains n = 11, infected control brains n = 9, non-infected sarm-RNAi brains n = 10, infected sarm-RNAi brains n = 7. (F, G) Seven-day exposure to B. bassiana caused loss of TH + PPM3 DANs (arrows), which was rescued with adult-specific sarm-RNAi^JF01681 knockdown, quantification in (G). Two-way ANOVA: Infected versus not-infected p p = 0.0071; Interaction: p = 0.0071, followed by Tukey’s multiple comparisons correction test. Sample sizes: non-infected control brains n = 8, infected control brains n = 9, non-infected sarm-RNAi brains n = 11, infected sarm-RNAi brains n = 8. (H) Diagram of Toll signalling pathway, whereby Sarm leads to cell death and cell loss. B. bassiana infection caused cell-loss is rescued with sarm RNAi knock-down. (I, J) sarm-RNAi knock-down in MyD88 cells slightly improved survival (I) and (J) climbing albeit not significantly (n = 76–194 flies). Two-way ANOVA survival: interaction: p = 0.5092; Genotype comparisons: p = 0.5591; Infected versus not-infected: p = 0.0001. (J) Two Way Anova climbing: interaction: p = 0.2992; Genotype comparisons: p = 0.0290; Infected versus not-infected: p p p p MyD88 GAL4 hisYFP/ + tubGAL80^ts/+ flies, whereby the GAL4 driver line was outcrossed to wild-type. Test flies: MyD88 GAL4 hisYFP/+ tubGAL80^ts/UAS-sarm-RNAi^JF01681 Infected versus not-infected compares flies of the same genotypes. See S4 Table for further details. The data underlying panels C, E, G, I and J can be found in S5 Table Source data.

https://doi.org/10.1371/journal.pbio.3003020.g007

To conclude, these data show that signalling via Wek and Sarm is required for B. bassiana-induced cell loss in the brain. They suggest that the upregulation of Wek and Sarm upon fungal infection (see Fig 3G) could cause neurodegeneration in the host (Fig 7H).

We next asked whether over-expression of activated Toll-1, wek or sarm could be sufficient to induce cell loss in the absence of infection. Wek has pleiotropic functions downstream of Tolls: it binds Tolls and MyD88 to enable canonical signalling via NFκB/Dorsal/Dif downstream, which in the CNS promotes cell quiescence and cell survival; it can bind Sarm, to promote cell death; or function independently of both, to promote neurogenesis in the adult brain from quiescent MyD88 + progenitor cells [33,34,36]. Thus, the pleiotropic functions of wek could lead to compound phenotypes. To simplify this, we tested the effect of activated Toll-1^10b, wek or sarm over-expression on PAM neurons, which are differentiated neurons and whose number can vary [79]. We used a DAN-specific GAL4 driver (THGAL4; R58E02GAL4), visualised PAMs with histone-YFP and counted them automatically. Toll-1 is normally expressed only in a fraction of PAM neurons (S3A Fig). Over-expressed activated Toll-1^10b in DANs caused a mild and not significant decrease in PAM cell number (Fig 8A and 8B), suggesting that not many PAMs may express wek, sarm or both in the un-infected brain. To note, B. bassiana infection causes an up-regulation in wek and sarm expression (see Fig 3G). And over-expression of either wek (Fig 8A and 8B) or sarm (Fig 8C and 8D) was sufficient to decrease PAM neuron number in the absence of infection. Such targeted and partial DAN cell loss was not sufficient to impair climbing (S5 Fig). Altogether, these data show that increased Wek and Sarm levels can induce PAM cell loss. These data suggest that the upregulation of wek and sarm caused by B. bassiana infection leads to cell loss in the host brain.

Fig 8. Increased Toll-1, wek and sarm levels can induce cell loss in the absence of infection.

(A,B) In the absence of infection, over-expression of activated Toll-1^10b in DANs (with THGAL4; R58E02GAL4) caused a rather mild and not significant decrease in PAMs. By contrast, over-expression of wek was sufficient to induce cell loss in 7-day-old flies. Sample sizes: control n = 9 brains, UAS-Toll-1^10b n = 12; UAS-wek-HA n = 13. One-way ANOVA p = 0.0003, multiple comparisons corrections Dunnett test to a fixed control. (C,D) Over-expression of sarm was sufficient to induce PAM cell loss in the absence of infection in 2-day-old flies. Sample sizes: control n = 13 brains, UAS-sarm n = 12. Student t test. Asterisks on graphs indicate multiple comparisons corrections: ****p (E) Diagram illustrating the multiple signalling pathways downstream of Toll-1 that can be activated by B. bassiana. Signalling via MyD88 and Dif/NFκB results in fungal elimination, whereas signalling via Wek and Sarm inhibits MyD88-dependent innate immunity and drives host cell death instead. Genotypes: Controls: THGAL4; R58E02GAL4/+ flies, whereby the GAL4 driver was outcroseed to wild-type. Test flies: THGAL4/UAS-Toll-1^10b; R58E02GAL4/+ ; THGAL4; R58E02GAL4/UAS-wek-HA; THGAL4; R58E02GAL4/UAS-dsarm. See S4 Table for further details. The data underlying panels B and D can be found in S5 Table source data.

https://doi.org/10.1371/journal.pbio.3003020.g008

Please see S1 Table for the list of the stocks used. Conditional over-expression and knock-down in adult flies were carried out using ubiquitously expressed tub-GAL80^ts. GAL80^ts is a temperature-sensitive GAL4 repressor that prevents GAL4 expression at 18°C and enables it at 30°C. The temperature regimes we used are indicated in the figures and were set to enable GAL4 adult-onset and Beauveria bassiana growth together. Controls used were standard F1 outcrosses of GAL4 or UAS lines to wild-type (Oregon R in our work), except for Fig 1, where controls were the F1 progeny of wild-type Oregon crossed to wild-type Canton-S. Throughout the paper, the symbol “>” stands for F1 flies bearing both a GAL4 driver and a UAS responder. The symbol “>+” stands for F1 from GAL4 driver crossed to wild-type. Details of genotypes and sample sizes are provided in S4 Table. Both females and males were used for experiments in Figs 1–4 (exactly the same number of each); only females were used for all other data.

Longevity was measured as described in [99], using F1 wild-type Oregon/CantonS flies raised on standard cornmeal food. They were placed in an infection chamber, the fly food was replaced every four days in both control and experimental setups, and the number of dead flies and censor events were manually recorded after every transfer. Both infected flies and non-infected controls were kept at 25°C. Dead flies were scored as 1 and flies that got stuck to the food or escaped were scored as 0. Data were analysed using GraphPad prism and Log rank (Mantel-Cox test) to analyse the data and generate a Kaplan-Meier survival curve. The experiment was conducted with n = 104 flies for each condition, from two biological replicates of n = 58 and n = 46 flies each.

B. bassiana cells were visualised after 3 days of exposure, using: (1) FM4–64, a red fluorescent dye (excitation/emission maxima ∼ 515/640 nm) used to stain yeast vacuolar membranes [103]. FM4–64 dye has been reported to stain B. bassiana spores in-vivo in the haemolymph of tobacco budworm [104], suggesting its potential use in staining spores in the adult Drosophila brain. (2) Calcofluor white (CLW, which stains cell walls of algae, plants and fungi), a fluorescent dye that binds to 1–3 beta and 1–4 beta polysaccharides of chitin and cellulose present in fungal cell walls [105]. (3) Transgenic B. bassiana-GFP, used after 1 day of exposure, as flies died earlier with this strain. Adult Drosophila brains were dissected and fixed in 4% paraformaldehyde (PFA) in Phosphate buffer saline (PBS) at room temperature for 20 min, taking great care to remove fat body and all other surrounding cells and tissues. Following washes in 0.5% Triton in PBS, brains were incubated in FM4–64 dye at 1:200 dilution, or calcofluor white (50 mg/ml) at 1:1,000 dilution for 1h at room temperature. The brains were washed 20 mins in 0.5% PBT five times plus a final wash in PBS and mounted in DAPI containing Vectashield, prior to imaging.

This protocol was adapted from [67]. A glass-pulled capillary needle was inserted into a glass pasteur pipette and sealed with parafilm at the point of insertion, and needle was loaded with 0.1 µl of 25 mg/µl fluorescent Dextran red dye. The needle was carefully inserted into the thorax beneath the wing of anaesthetised flies. The flies were then transferred to a vial containing fly food and placed in a 25°C incubator for 24 hrs to recover. After the 24 hours recovery period, the flies were fixed onto a glass slide using glue, and their retinas were imaged using Leica SP8 confocal fluorescent microscope, n = 10 flies.

qRT-PCR was carried out from 20 wild-type (Oregon) adult brains per sample, dissected from wild type Oregon non-infected and infected flies. The dissected adult brains were thoroughly cleaned removing all fat body, other surrounding cells and tissues, and then immediately transferred into the Eppendorf containing TRI reagent (Ambion #AM9,738). RNA extraction was carried out according to the TRI reagent protocol using isopropanol. Extracted RNA was treated with DNase to get rid of genomic DNA contamination using DNA-free kit (Ambion #AM1906). 200 ng of RNA was reverse transcribed into cDNA by using random primer by following the manuscript of GoScript^TM Reverse Transcription System (Promega #A50001) (S2 Table). A PCR was performed to check the contamination in the RNA and Reverse Transcribed (RT) cDNA (S2 Table). This is followed by a qRT-PCR. All qRT-PCR experiments were done in triplicate (i.e., 3 well condition). A DNA-binding dye SYBR was used to observe the dynamics of the PCR (SensiFast^TM SYBR) and used ABI Prism 7,000SDS machine for the qPCR (S2 Table). Following qRT-PCR, quantification was performed using the CT value generated by the qPCR machine, as described in [106]. For the internal control, GAPDH (housekeeping\reference gene) was used. The expression of the target gene (ΔCT) was normalised in infected and non-infected flies by subtracting the CT value of the GAPDH (reference gene) from the CT value of the target gene. ΔΔCT was calculated by subtracting the ΔCT (Target gene) of the infected flies from the ΔCT (Target gene) of the non-infected flies. The relative gene expression of the target gene was then calculated in infected and non-infected flies and expressed as fold change (2^-ΔΔCT). The statistical analysis was conducted on the ΔCT values. Three biological replicates were used each biological replicate consisted of 20 flies. ΔCT =  CT (a target gene) – CT (a reference gene\GAPDH); ΔΔCT =  ΔCT (non-infected sample) – ΔCT (infected sample); Fold change =  2^{– ΔΔCT}

Toll-1-GAL4 transgenic flies were generated using CRISPR/Cas9 enhanced homologous recombination. The 5′ homology arm was PCRed using primers 5’Fwd: ATGCGACCGGTAAAATCTCGTATTATGCAGCACTCGA and 5’Rev: GGAACTGAGCGGCCGCTGCAAATGGAGAAATTGAAAGGAAT; the 3′ homology arm was PCRed using primers 3’Fwd: GATGGCGCGCCGTGAACCCATTTGGACAACA and 3′ Rev: CGTACTAGTGCAGTTCAGCTCTCAGCCGT. The donor vector was pT-GEM-T2A-Gal4, and homology arms were cloned into AgeI and NotI (5′ homology arm) and AscI and SpeI (3′ homology arm) enzyme sites (as in [107]). The guide RNA targeted the start of the coding sequence: sense oligo: GTCGCCCATTTGGACAACATGAGTCGA; antisense oligo: AAACTCGACTCATGTTGTCCAAATGGG. The gRNA was cloned into the BbsI site of vector pU6.3, as described in [108]. Transgenesis was carried out by the BestGene. 3xP3-DsRed was subsequently removed with CRE-recombinase using conventional genetics. Primers are given in S2 Table.

To prepare flies for immunostaining, infected flies were first transferred to vials from infection chamber and flipped twice to remove excess spores. The flies were then transferred to 70% ethanol and washed in 1XPBS for 10 min. Adult Brains were dissected in 20 min and fixed in 650ul fixative (650ul 4% paraformaldehyde (PFA) in 1× Phosphate buffer saline (PBS) at room temperature for 20 min in a 1.5 ml Eppendorf. The dissected adult brains were thoroughly cleaned removing all fat body, other surrounding cells and tissues. Subsequently, the brains were washed for 20 mins in 750ul of 0.5% PBT (47.5 ml 1xPBS +  2.5 ml 10% Triton) 5 times. Following this, the brains were incubated in anti-TH (at 1:20) or anti-Repo (at 1:250) primary antibodies for 2 days. Following this, brains were washed for 20 min 5 times in 0.5% PBT, then incubated in fluorescently tagged secondary antibodies (anti-Mouse for Repo and anti-Rabbit for TH) overnight, as and if appropriate. After this, brains were washed again in PBT as above. Finally, the brains were mounted onto slides with either Vectashiled or DAPI containing Vectashield depending on the experiments and kept at 4°C overnight ready to be scanned. Antibodies used, their source and their working dilutions are given in S3 Table. We used both females and males for experiments in Figs 1–4, but we used exactly the same number of each gender for the experiments. We did not observe any different effects between the sexes. Thus, for all other experiments, we used female flies only, because they are bigger, so their brains are also bigger and easier to dissect.

Cells were counted automatically using DeadEasy software, developed as ImageJ plug-ins, as reported in [36]. Repo+ cells in the adult brain were counted automatically with the DeadEasy Glia Adult; Sarm>HisYFP+ and MyD88 > HisYFP+ cells were quantified with DeadEasy adult central brain; and THGAL4 R58E02GAL4 > HisYFP+ cells were counted with DeadEasy adult central brain. Dopaminergic neurons labelled with anti-TH were counted manually, assisted by the Fiji cell counter. DeadEasy plugins published and previously used in [36,79] are publically available through UBIRA https://edata.bham.ac.uk/; https://doi.org/10.25500/edata.bham.00001213.