Introduction

Adult organs such as intestine rely on resident stem cells to replenish damaged tissues during homeostasis and regeneration (Biteau et al. 2011; Jiang and Edgar 2012). Drosophila midgut, an equivalent of mammalian small intestine, has emerged as a powerful system to study stem cell biology in adult tissue homeostasis and regeneration (Casali and Batlle 2009; Jiang and Edgar 2011; Jiang et al. 2016). Intestine stem cells (ISCs) in adult midguts are localized at the basal side of the gut epithelium where they can undergo asymmetric cell division to produce renewed ISCs and enteroblasts (EBs) that differentiate into enterocytes (ECs) (Micchelli and Perrimon 2006; Ohlstein and Spradling 2006). At low frequency, an ISC daughter is fated to preEE/EEP that differentiates into two enteroendocrine cells (EEs) after another round of cell division (Biteau and Jasper 2014; Beehler-Evans and Micchelli 2015; Zeng and Hou 2015; Chen et al. 2018). About 20% ISCs undergo symmetric cell division to produce two ISCs or two EBs (O’Brien et al. 2011; Goulas et al. 2012; Tian and Jiang 2014). The decision between ISC self-renewal and differentiation into EB lineage is controlled by Notch (N) signaling whereby N activation drives ISC differentiation into EB (Micchelli and Perrimon 2006; Ohlstein and Spradling 2006; Ohlstein and Spradling 2007; Bardin et al. 2010). The asymmetric N signaling between ISC and EB is influenced by Par/integrins-directed asymmetric cell division and differential BMP signaling (Goulas et al. 2012; Tian and Jiang 2014). EC-produced BMP ligands containing Decapentaplegic (Dpp) and Glass bottom boat (Gbb) heterodimers are secreted basally and concentrated on the basement membrane aligning the basal side of the gut epithelium (Tian and Jiang 2014). After asymmetric cell division of ISCs, basally localized daughter cells transduce higher levels of BMP signaling activity than the apically localized daughter cells and the differential BMP signaling promotes ISC self-renewal by antagonizing N pathway activity through an unknown mechanism (Tian and Jiang 2014; Tian and Jiang 2017).

The N pathway inhibitor Numb plays a decisive role in asymmetric cell fate determination in Drosophila peripheral and central nervous systems whereby Numb segregates asymmetrically into one daughter during division of a neuronal precursor cell and confers distinct fates to the two daughter cells (Uemura et al. 1989; Rhyu et al. 1994; Spana et al. 1995). The mammalian homologs of Numb are also critical for asymmetric fate determination during neurogenesis and myogenesis (Zhong et al. 1996; Conboy and Rando 2002; Petersen et al. 2002; Shen et al. 2002; Petersen et al. 2004). Previous studies showed that, during asymmetric division of an ISC in Drosophila adult midgut, Numb was preferentially segregated into the basally localized daughter that becomes future ISC (Goulas et al. 2012; Salle et al. 2017); however, many numb mutant clones retained ISC after many rounds of cell division although they failed to produce EE (Bardin et al. 2010; Salle et al. 2017). Given the well-established role of Numb in blocking N pathway activity and the observation that Numb is asymmetrically segregated into the future ISC, it is surprising and puzzling that loss of Numb does not lead to ectopic N pathway activation that drives ISC-to-EB differentiation. We speculate that BMP signaling may play a more dominant role in ISC self-renewal than Numb and that the BMP signaling gradient could generate differential N signaling between the apical and basal pair of ISC daughters to generate ISC/EB binary fates even when Numb is depleted (Tian and Jiang 2017). Accordingly, attenuating BMP signaling may unmask the role of Numb in ISC self-renewal.

To test this hypothesis, we employed RNA interference (RNAi) and genetic mutations to inactivate Numb in otherwise wild type background or in midguts defective in BMP signaling due to RNAi or genetic mutation of mothers against decapentaplegic (mad), which encodes a signal transducer in BMP signaling pathway (Sekelsky et al. 1995; Newfeld et al. 1997). Consistent with our previous findings (Tian and Jiang 2014), neither mad RNAi nor its hypomorphic mutation led to ISC loss. However, inactivation of Numb in these backgrounds resulted in stem cell loss due to precocious ISC-to-EB differentiation. By carefully examining mutant clones for multiple numb alleles, we also observed an increased number of numb clones that lack ISCs compared with wild type control clones. Interestingly, the stem cell loss phenotype was exacerbated by feeding flies with bleomycin, which resulted in EC damage and fluctuation of BMP signaling (Amcheslavsky et al. 2009; Tian et al. 2017), underscoring an essential role of Numb in ISC maintenance during gut regeneration.

Results

BMP signaling inhibits ligand-independent N pathway activity

Our previous study showed that depleting the type II receptor Punt (Put) for BMP in progenitor cells (ISC/EB) resulted in precocious ISC-to-EB differentiation, leading to stem cell loss (Tian and Jiang 2014). In Put deficient progenitor cells, the N pathway was activated in the absence of detectable N ligand Delta (Dl). Progenitor cells deficient for both Put and N failed to differentiate to EBs and formed stem cell-like tumors (Tian and Jiang 2014). These observations imply that inactivation of Put may unleash a ligand-independent N pathway activity that drives precocious ISC-to-EB differentiation, leading to stem cell depletion. However, it remains possible that a trace amount of Dl that is beyond the detection by immunostaining might activate N in Put deficient progenitor cells.

To further explore the relationship between the BMP and N pathways, we carried out genetic epistatic experiments to determine the functional relationship between Put and Dl. We depleted Put and Dl either individually or in combination via RNAi in midgut progenitor cells using the esg-Gal4 tub-Gal80^ts (esg^ts) system, in which Gal4 is under the control of a temperature sensitive Gal80 (McGuire et al. 2004). UAS-GFP was included in the esg^ts system to mark all precursor cells whereas Su(H)-lacZ (also called Su(H)GBE-lacZ), a transcriptional reporter of N signaling, was used to monitor N pathway activity and mark the EBs. 3-to 5-day-old adult females expressing UAS-Put-RNAi and UAS-Dl-RNAi individually or in combination via esg^ts were shifted to 29 ⁰C for 10 days prior to dissection, followed by immunostained with the corresponding antibodies. In control guts, most pairs of esg>GFP⁺ precursor cells, which may represent two recently divided ISC daughters, contained only one Su(H)-lacZ positive cell, indicating that these ISCs divided asymmetrically to produce one ISC and one EB (Figure1A and B-B’’). In line with our previous findings, most esg>GFP⁺precursor pairs from the Put RNAi guts expressed Su(H)-lacZ in both ISC daughters, suggesting that ISCs underwent symmetric cell division to produce two EBs when Put was depleted (Figure1C-C’’). By contrast, esg>GFP⁺cells in Dl RNAi guts formed ISC-like tumor clusters that were negative for Su(H)-lacZ (Figure1D-D’’), similar to the ISC-like tumor clusters in guts containing Dl or N mutant clones (Micchelli and Perrimon 2006; Ohlstein and Spradling 2006; Ohlstein and Spradling 2007; Siudeja et al. 2015), suggesting that in Dl RNAi guts, N pathway activity was diminished. Put and Dl double RNAi suppressed the formation of ISC-like clusters (Figure1E). In these guts, esg-GFP⁺cells exhibited Su(H)-lacZ expression (Figure1E-E’’), suggest that they activate the N pathway and differentiate into EBs. Hence, loss of BMP signaling resulted in ectopic N pathway activity that drives ISC-to-EB differentiation even when Dl was depleted.

Figure 1.

Numb is essential for ISC maintenance when BMP pathway activity is attenuated

Our previous study showed that partial loss of BMP pathway activity in several genetic background including Mad RNAi and mad^1-2, a hypomorphic allele of mad, did not lead to ISC loss whereas more complete loss of BMP signaling in Put RNAi guts or put mutant ISC lineage clones resulted in ISC loss. It is possible that a backup mechanism for ISC self-renewal may exist, which could compensate for the partial loss of BMP signaling to prevent ectopic N pathway activation that drives differentiation. During an asymmetric ISC division, the N inhibitor Numb is segregated into the basally localized daughter that becomes the future ISC (Goulas et al. 2012; Salle et al. 2017). We hypothesized that the asymmetric distribution of Numb may provide such a backup mechanism to ensure that the basally localized ISC daughter has lower N pathway activity than the apically localized one when differential BMP signaling is compromised so that the differential N signaling between the apical and basal is still sufficient to drive asymmetric fate determination. To test this hypothesis, we inactivated Numb and Mad either individually or in combination using two independent approaches: 1) RNAi and 2) genetic mutations. For the RNAi experiments, 3∼5-day old females expressing UAS-Numb-RNAi, UAS-Mad-RNAi, or UAS-Numb-RNAi + UAS-Mad-RNAi under the control of esg^ts were transferred to 29°C for 14 days. The guts were then dissected out for immunostaining to detect the expression of esg>GFP, Dl-lacZ (ISC marker), E(spl)mβ-CD2 (EB marker) and Pros (EE marker). Because preEE expressed both Dl-lacZ and Pros and Dl-lacZ signals could be found in some EBs due to its perdurance, we counted Dl-lacZ⁺ mβ-CD2^- Pros^- cells as ISCs and mβ-CD2⁺ cells as EBs. Compared with control guts (Figure 2A-A’’), Mad (Figure 2B-B’’) or Numb (Figure 2C-C’’) single RNAi guts contained comparable number of Dl-lacZ⁺ mβ-CD2^- Pros^- cells and E(spl)mβ-CD2⁺ cells (Figure 2E-G). By contrast, in Numb and Mad double RNAi guts (Figure 2D-D’’), there was a significant decrease in the number of precursor cells (Figure 2E) and Dl-lacZ⁺ mβ-CD2^- Pros^- cells (Figure 2F), and a simultaneous increase in the number of E(spl)mβ-CD2⁺ cells (Figure 2G), suggesting that inactivation of both Mad and Numb results in stem cell loss, likely due to precocious ISC to EB differentiation.

Figure 2.

In the second approach, we generated guts that carried mad^1-2or numb⁴ single mutant clones or mad^1-2, numb⁴ double mutant clones using the MARCM system that positively mark the clones with GFP expression. 3∼5-days-old females of appropriate genotypes were heat-shocked for 1 hr for clonal induction and kept at 18°C for 14 days prior to dissection. ISCs were identified as Dl⁺ cells or mβ-CD2^- Pros^- cells containing small nuclei. ISC-containing clones (ISC⁺) and clones without ISCs (ISC^-) were quantified for each genotype. We also quantified the size of ISC lineage clones for each genotype by counting GFP⁺ cells in individual clones. Consistent with previous findings (Tian and Jiang 2014; Salle et al. 2017), the average size of mad^1-2 clones is significantly larger than the control clones (Figure 3A-A’, B-B’, E-E’, F-F’, K) whereas numb⁴ clones had similar clone size distribution compared with control clones (Figure 3C-C’, G-G’, K). In addition, most of mad^1-2 or numb⁴ clones contained at least one ISC similar to control clones (Figure 3K). However, the average size of mad^1-2 numb⁴ clones is significantly smaller than that of control clones (Figure 3D-D’, H-H’, K). More importantly, a much larger fraction of mad^1-2 numb⁴ clones (∼40%; n=252) did not contain ISC (Figure 3L), many of which only contained ECs with large nuclei and stained positive for Pdm1(Figure 3I-J’). Taken together, these results suggest mad^1-2 numb⁴ double mutation leads to ISC loss.

Figure 3.

Inactivation of Numb and Mad leads to precocious ISC-to-EB differentiation

We employed a two-color lineage tracing system called RGT (Tian and Jiang 2014; Tian et al. 2017) to determine whether simultaneous inactivation of Numb and Mad would change the outcome of an ISC division. In this system, FLP/FRT-mediated mitotic recombination in individual dividing ISCs will generate two distinctly labelled clones that express either RFP (red) or GFP (green) (Figure 4A). As shown schematically in Figure4B, asymmetric ISC division (ISC/EB) will generate one clone with multiple cells and a twin spot that contains only one EC. Symmetric self-renewing division (ISC/ISC) will produce two multiple-cell clones whereas symmetric differentiation division (EB/EB) will produce two clones each of which contains one EC. Control or RNAi expressing adult flies containing hs-FLP FRT19A ubi-GFPnls/FRT19A ubi-mRFPnls; esg^ts were grown at 29 °C for 8 days (for Mad-RNAi only) or 14 days (for control, Numb-RNAi, and Numb-RNAi + Mad-RNAi) before clone induction by heat shock at 37 °C for 1 hr. After clone induction, the flies were incubated at 18 °C for another 4 days before guts were dissected out for analysis (Figure 4C). The frequencies of ISC/EB, ISC/ISC and EB/EB divisions in control guts were 69%, 15% and 16% respectively (n=119) (Figure 4D-F, P). Mad RNAi guts had higher frequency of ISC/ISC (30%), and lower frequency of EB/EB (5%) division compared to control guts (n=73) (Figure 4G-I, P). The increase in symmetric self-renewing division in Mad RNAi guts is likely due to an increase in BMP ligand production in these guts because BMP signaling in EC inhibits BMP ligand expression (Tian et al. 2017). The frequencies of different ISC division classes in Numb RNAi guts are comparable to those of control guts (ISC/EB: 73%, ISC/ISC: 18%, EB/EB: 9%, n=127) (Figure 4J-L, P). By contrast, Mad and Numb double RNAi guts had lower frequency of ISC/ISC division (11%) and much higher EB/EB division (37%) than control guts (n=54) (Figure 4M-O, P). Thus, inactivation of Numb in backgrounds where BMP signaling was compromised altered the ISC division outcome that favors symmetric differentiation division leading to ISC loss.

Figure 4.

Numb mutant clones exhibit weak ISC loss phenotype

When we examined adult midguts containing numb⁴ clones, we noticed a slight increase in the frequency of ISC^- clones compared to the control guts even though the average clone size of numb⁴clones was comparable to that of the control clones (Figure 3K, L), suggesting that numb mutation may result in a mild stem cell loss phenotype. To verify this result, we examined another numb allele, numb¹⁵. By immunostaining for Dl expression that marks ISC, we found that both numb⁴ and numb¹⁵ clones contained similarly higher frequence of Dl^- clones than the control clones (Figure 5 A-C’, E). Consistent with previous findings (Bardin et al. 2010; Salle et al. 2017), most numb¹⁵ clones grew into large size similar to the control clones (Figure 5D), suggesting that many ISC^- numb clones lost ISC at late stages during their clonal growth. We also examined Pros expression and found that numb mutant clones did not contain Pros⁺ cells (Figure 5-figure supplement 1A-C’), which is consistent with a previous study showing that numb is required for EE fate regulation (Salle et al. 2017).

Figure 5.

Figure 5-figure supplement 1.

Numb is essential for ISC maintenance during regeneration

The weak ISC loss phenotype associated with numb mutant clones could be due to fluctuation in BMP pathway activity under normal homeostasis because the expression of two BMP ligands Dpp and Gbb is uneven in homeostatic guts (Tian and Jiang 2014). If so, the stem cell loss phenotype caused by numb mutations could be enhanced under conditions where tissue damage causes more dramatic and widespread fluctuation in BMP signaling activity in regenerative guts. To test this possibility, we fed adult female flies carrying either control or numb clones in the guts with sucrose (mock), bleomycin, or dextran sodium sulfate (DSS). In mock-treated control guts, approximately 12% (n=178) of the clones did not contain Dl⁺ cell, while 21% (n=216) of the numb⁴ and 24% (n=219) of the numb¹⁵clones were void of stem cells (Figure 6A-A’, D-D’, G-G’, J). Previous studies showed that bleomycin treatment caused EC damage and enhanced the fluctuation in BMP ligand production whereas DSS affected basement membrane organization but did not increase the fluctuation in BMP ligand production (Amcheslavsky et al. 2009; Tian et al. 2017). In guts treated with bleomycin, 12% (n=160) of control clones did not contain Dl⁺ cells (Figure 6B-B’, J). However, bleomycin feeding resulted in a dramatic increase of Dl^- ISC-lineage clones in guts containing numb mutant clones as Dl⁺ cells were absent in 43% (n=149) of numb⁴ and 45% (n=213) of numb¹⁵clones (Figure 6E-E’, H-H’, J). By contrast, DSS feeding did not increase the frequency of Dl^- clones in guts containing numb mutant clones, as the frequencies of Dl^- clones in control, numb⁴and numb¹⁵ clonal guts were 10% (n=167), 24% (n=165) and 21% (n=141), respectively (Figure 6C-C’, F-F’, I-I’, J). Bleomycin also resulted in a reduction in numb mutant clone size, as compared with the mock-treatment (Figure 6K). Taken together, these results suggest that Numb plays an essential role in ISC maintenance in regenerative guts in response to bleomycin-induced tissue damage.

Figure 6.

Discussion

Despite the prominent role of Numb in asymmetric cell fate decision in the nervous system and the observation that Numb is asymmetrically segregated in dividing ISCs, whether Numb plays any role in ISC fate determination has remained a mystery. Here we demonstrated that Numb is essential for ISC self-renewal during regeneration. We found that Numb is largely dispensable for homeostatic ISC self-renewal due to the predominant role of BMP signaling in this context. Indeed, we found that Numb becomes critical for ISC self-renewal under conditions where BMP signaling is compromised.

Previous studies did not score a stem cell loss phenotype associated with numb mutant clones because the majority of numb mutant ISC lineage clones could grow into large size comparable to control clones (Figs. 3 and 5)(Bardin et al. 2010; Goulas et al. 2012; Salle et al. 2017). Instead, Salle et al showed that numb mutant clones lacked EEs, suggesting that Numb is essential for EE fate determination (Salle et al. 2017), which we confirmed in this study (Figure S1). However, by carefully examining ISC/EB markers associated with numb mutant clones, we noticed an increase in the fraction of numb mutant clones that lack ISCs compared with the control clones (Figures 3 and 5). By introducing mad mutation (mad^1-2) into the numb mutant background, we found that numb⁴ mad^1-2double mutant clones had a much higher frequence to lose ISCs than numb⁴clones even though mad^1-2 single mutant clones showed no ISC loss phenotype compared with control clones (Figure 3).

Our previous study showed that immediately after bleomycin treatment, there was an increase in the ISC population size due to a transient surge in BMP ligand production (Tian et al. 2017). However, during regeneration, BMP ligand production was downregulated due to the autoinhibition of BMP ligand expression by BMP signaling in ECs (Tian et al. 2017). The reduction in BMP ligand production promoted ISC-to-EB differentiation to reset the ISC population size back to the homeostatic level after regeneration (Tian et al. 2017). It is likely that asymmetric distribution of Numb in dividing ISCs may prevent excessive ISC-to-EB differentiation that could otherwise lead to a decreased ISC population during regeneration. Indeed, we observed an increase in the frequency of numb mutant clones lacking ISCs in response to bleomycin treatment, suggesting that Numb becomes critical for ISC maintenance during gut regeneration in response to tissue damage.

Based on our findings in current and previous studies, we propose the following working model to account for the cooperation between Numb and BMP signaling in the regulation of ISC self-renewal under homeostatic and tissue regeneration (Figure 7). Under homeostatic conditions, most ISCs divide basally so that the basally localized ISC daughters inherit Numb and transducing higher levels of BMP signaling activity than the apically situated daughter cells, the combined differential activities in BMP signaling and Numb drive robust asymmetric division outcomes to produce ISC/EB pairs (Figure 7A). In the numb mutant background, differential BMP signaling activities between the apical and basal daughter cells suffice to drive asymmetric division outcomes in most cases (Figure 7B). In the mad mutant background, the BMP signaling activity gradient becomes shallower but differential Numb activity between the apical and basal daughter cells can compensate for the compromised BMP signaling gradient to drive asymmetric division outcomes (Figure 7C). However, in numb mad double mutant background or in numb mutant guts damaged by bleomycin feeding, the compromised BMP signaling activity gradient alone is often insufficient to drive asymmetric division outcomes, leading to ISC loss due to symmetric EB/EB division outcomes (Figure 7D). One interesting question is why asymmetric Numb activity is unable to drive asymmetric ISC division outcome in the absence of BMP signaling as seen in put mutant background. One possibility is that the Numb level is too low in midgut ISCs so that the asymmetric Numb inheritance during ISC division is not robust enough to ensure asymmetric N signaling in the absence of BMP signaling. Indeed, previous studies indicate that endogenous Numb was not undetectable by Numb antibodies that could detect Numb expression in the nervous system (Goulas et al. 2012; Couturier et al. 2013; Salle et al. 2017). Another non-mutually exclusive possibility is that Numb may not be able to counter the ligand-independent N pathway activity unleashed in put mutant backgrounds. Future study is needed to test these possibilities and to determine the precise mechanism by which BMP signaling inhibits N pathway activity.

Figure 7.

Materials and Methods

Drosophila genetics and transgenes

Flies were maintained on cornmeal at 25°C. Transgenic lines and mutants include: UAS-Put-RNAi (VDRC #107071); UAS-Dl-RNAi (BL#28032); UAS-Mad-RNAi (VDRC #12635); UAS-Numb-RNAi (BL #35045); UAS-mCherry-RNAi (BL #35785); tub-Gal80^ts, esg-Gal4, Su(H) Gbe-lacZ (Su(H)-lacZ); E(spl)mβ-CD2 (BL#83353); numb⁴, numb¹⁵, and mad^1-2 (Flybase). yw, hs-FLP, UAS-GFP; tub-Gal80, FRT40A was used for MARCM clonal analysis. yw, hs-FLP, FRT19A, ubi-GFPnls and yw, FRT 19A, ubi-mRFPnls were used for twin spot clone analysis. For experiments involving tubGal80^ts, crosses were set up and cultured at 18°C to restrict Gal4 activity. 2 to 3-day-old progenies were shifted to 29°C for the indicated periods of time to inactivate Gal80^ts, allowing Gal4 to activate UAS transgenes in all experiments, only the female posterior midguts were analyzed.

MARCM clone analysis

For MARCM clone induction, crosses were set up and cultured at 18°C to avoid spontaneous clones. 2-to-3-day-old females were subjected to heat shock at 37°C for 1 hr and then kept at 18°C for another 14 days before dissection. Flies were transferred to new vials with fresh food every 2 days. The sizes of the clones were quantified from at least 10 midguts for each genotype.

Twin spot clone analysis

For twin spot clone generation, 2-to-3-day-old flies were kept at 29°C for 14 days and heat-shocked at 37°C for 1 hr and then raised at 29°C for another 4 days before dissection. Flies were transferred to new vials with fresh food every 2 days.

Feeding experiments

Flies were cultured in an empty vial containing a piece of 2.5 × 3.75-cm chromatography paper (Fisher) wet with 5% sucrose (MP Biomedicals) solution as feeding medium (mock treatment) or with 25 μg/mL bleomycin (Sigma-Aldrich) or 5% DSS (40 kDa; MP Biomedicals) for one day at 30°C. After treatment, flies were recovered on normal food at 18°C for another 4 days before dissection.

Immunostaining

Female flies were used for gut immunostaining in all experiments. The entire gastrointestinal tract was taken and fixed in 1 X PBS plus 8% EM grade formaldehyde (Polysciences) for 2 hours. Samples were washed and incubated with primary and secondary antibodies in a solution containing 1 X PBS, 0.5% goat serum (Thermos Fisher), and 0.1% Triton X-100 (Bio-rad). The following primary antibodies were used: mouse anti-Delta (DSHB), 1:10; rabbit anti-LacZ (MP Biomedicals), 1:1,000; mouse anti-CD2 (Thermos Fisher), 1:1000; chicken anti-GFP (Abcam), 1:1000; mouse anti-Pros (DSHB), 1:10; rabbit anti-Pdm1 (from Dr. Xiaohang Yang), 1:1000; Alexa Fluor-conjugated secondary antibodies were used at 1:1000 (Invitrogen). DAPI (4’,6-Diamidino-2-Phenylindole) is a nuclear dye (Thermos Fisher). Guts were mounted in 70% glycerol and imaged with a Zeiss confocal microscope (Zeiss LSM 710 inverted confocal) using 40X oil objectives (imaging medium: Zeiss Immersol 518F). The acquisition and processing software was Zeiss LSM Image Browser, and image processing was done in Adobe Photoshop.

Quantification and Statistical analysis

In Figure 2, cell number of the indicated cell types were counted per ROI (region of interest) on images taken using LEICA DFC345 FX camera on a LEICA DMI 400 B microscope, equipped with a 40 × objective lens. For each genotype, 8∼12 guts were analyzed. In each gut, three ROI were randomly selected in midguts for quantification. One-way ANOVA was performed for statistical comparisons. In Figure 3 - 6, All GFP⁺ clone cells (≥2) in midguts were counted individually. For each genotype, at least 10 guts were calculated. ᵡ² test was performed for statistical comparisons. All statistical significances were calculated in Prism 10 (GraphPad Software, Inc). *, p <0.05, **, p <0.01, ***, p <0.001, ****, p <0.0001; n.s., not significant.

Acknowledgements

We thank Bing Wang for technical assistance, Vienna Drosophila Resource Center and Bloomington Drosophila Stock Center for fly stocks, and Developmental Studies Hybridoma Bank for antibodies. This work is supported National Institutes of Health Grant GM118063 and Welch Foundation Grant I-1603 to J.J.

Additional information

Author Contributions

J.J. designed research; M.L. and A.T. performed research; M.L., A.T. and J.J. analyzed data; and M.L. and J.J. wrote the paper.