Independently paced Ca²⁺ oscillations in progenitor and differentiated cells in an ex vivo epithelial organ

Ca²⁺ is a versatile signaling molecule that regulates vital cellular functions such as contraction and cellular excitability in all organ systems (Berridge et al., 2000; Carafoli, 2002; Clapham, 2007). Changes in Ca²⁺ signal dynamics have been linked to crucial cell behaviors, such as intercellular communication, cell cycle, proliferation and migration (Hofer et al., 2000; Humeau et al., 2018; Wei et al., 2009; Xu et al., 2017). Intracellular Ca²⁺ concentrations can also regulate cellular responses and physiology by modulating signal transduction pathways such as MAPK (Apáti et al., 2003; Kupzig et al., 2005) and inositol trisphosphate (IP₃) (Harootunian et al., 1991; Sjaastad et al., 1996; Taylor and Thorn, 2001). In excitable tissues, such as in electrically coupled cells of heart muscle, Ca²⁺ plays a central role in propagating the impulse that coordinates the pacing of contractions (Cheng et al., 1996; Dewenter et al., 2017). Large-scale Ca²⁺ waves have been observed in cultured astrocytes and in mouse hippocampus astrocyte networks (Innocenti et al., 2000; Kuga et al., 2011).

Although Ca²⁺ dynamics have been studied widely in cell culture, investigations in tissues and organs – particularly those that are non-excitable – have been more limited. Studies in Drosophila demonstrated intercellular Ca²⁺ waves that traverse large tissue domains and might depend on actomyosin organization (Balaji et al., 2017). These tissue-level Ca²⁺ dynamics, which occur in imaginal discs, the tightly coupled epithelial structures that give rise to the external structures of the adult fly, were implicated in organ growth and size modulation (Brodskiy et al., 2019; Soundarrajan et al., 2021 preprint). Furthermore, intercellular Ca²⁺ waves were shown to be induced mechanically in these developing epithelia (Narciso et al., 2017). Finally, blood progenitors in the Drosophila lymph gland were recently shown to form a gap-junction-mediated network that can regulate Ca²⁺ signaling (Ho et al., 2021).

By contrast, little work has been done to examine Ca²⁺ dynamics in non-excitable tissues composed of heterogenous cell types. The question of whether and, if so, how tissue-scale Ca²⁺ oscillations are coordinated between different cell types is particularly intriguing for stem-cell-based tissues that undergo constitutive cellular turnover. As the fates of individual stem-cell daughters change during differentiation, any fate-associated differences in Ca²⁺ oscillations must evolve as fate decisions are made.

Here, we establish the adult Drosophila intestine as an ex vivo model for studying diverse Ca²⁺ dynamics that occur simultaneously in different cell types in a self-renewing mature organ. The midgut, like most mature organs, undergoes continuous turnover in which tissue-specific stem-cell divisions produce progeny that differentiate into multiple cell types. Ca²⁺ signaling has been established as a key regulator of midgut stem cell activity (Deng et al., 2015; He et al., 2018). Ca²⁺ transients are mechanically induced via the mechanosensitive ion channel Piezo in a subpopulation of stem cells in the fly gut (He et al., 2018). However, how Ca²⁺ dynamics regulate and are regulated by cell differentiation and organ-scale inputs remains largely unknown.

Here, we simultaneously expressed spectrally distinguishable Ca²⁺ indicators in each midgut cell type and performed real-time analysis of single- and multi-cell oscillations to produce a fate-resolved, tissue-scale overview of Ca²⁺ dynamics. We investigated fate-specific changes as cells differentiated in their native tissue environment and describe rhythmic multicellular Ca²⁺ waves in enterocytes and cell-type-specific Ca²⁺ oscillations. These results demonstrate that as cells differentiate from stem-cell-like into distinct terminal fates, they adopt cell-type-specific Ca²⁺ oscillations and waves that are a hallmark of the mature organ.

To examine Ca²⁺ dynamics in the adult fruit fly midgut, we performed live imaging of whole organs employing the genetically encoded Ca²⁺ sensors, GCaMP6s (Chen et al., 2013) and jRCaMP1b (Dana et al., 2016), hereafter referred to as GCaMP and RCaMP, respectively. To examine Ca²⁺ oscillations in each midgut cell type (Fig. 1A–C), we expressed a genetically encoded Ca²⁺ sensor using the GAL4/UAS system (Brand and Perrimon, 1993) under the control of cell-type-specific drivers. We dissected midguts from mated female fruit flies aged 4–7 days and mounted them by adapting an ex vivo organ culture protocol (Marchetti et al., 2021 preprint) (see Materials and Methods). To characterize cell-type-specific Ca²⁺ activity in terms of temporal dynamics, we traced the mean fluorescence intensity of individual cells as a function of time (see Materials and Methods) and quantified the frequency of oscillations.

The R3 region of the midgut, which is responsible for acid secretion, attracted our attention due to the consistent appearance of rapid, multicellular Ca²⁺ waves that travel through R3 enterocytes. We also observed Ca²⁺ dynamics in enterocytes in the posterior region of the midgut, but not consistently. We did not observe any Ca²⁺ dynamics in the anterior (R2) region, which might reflect a physiological difference in this region of the gut, or which might be a consequence of dissection or ex vivo culture. We used the GAL4 driver midgut expression 1 or mex1 (mex-GAL4) to label enterocytes and focused the remainder of our experiments on R3.

As a starting point, we measured the frequency of Ca²⁺ oscillations in single cells from each cell type over time. Enterocytes in the R3 region are subdivided into acid-secreting copper cells (CCs) and interstitial cells (ICs) (Dubreuil, 2004; Poulson and Bowen, 1952; Strasburger, 1932), which we distinguished by the unique shape of copper cells (Fig. 1B).

Interstitial cells exhibited wave-like Ca²⁺ dynamics with a mean oscillation frequency of 31±8.9 mHz (s.e.m.) per gut. By comparison, copper cells primarily displayed Ca²⁺ spikes at a frequency of 5.2±0.9 mHz per gut (Fig. 1D).

Interestingly, we also observed relatively rapid fluorescence oscillations in enteroendocrine cells [prospero (pros)-expressing] of 15±4.3 mHz per gut, approximately twice that of progenitors [stem cells and enteroblasts, which both express escargot (esg)], 7.1±2.6 mHz per gut. Finally, the robust Ca²⁺ oscillations we observed in progenitors are consistent with prior reports (Deng et al., 2015; He et al., 2018). The complete dataset of the average oscillation frequencies per midgut, including and excluding non-oscillating cells for all cell types, is included in Tables S2 and S3.

Unexpectedly, we found long-range Ca²⁺ waves that propagate across large fields of enterocytes in the R3 region of the midgut (Fig. 2A). We observed high GCaMP signals that shifted rapidly across several cell lengths in multiple directions (Movie 1), with the longest wave we observed covering approximately five cells. These Ca²⁺ waves, which exhibited a multitude of dynamic patterns, such as propagating along a single trajectory, splitting and colliding, did not have an obvious point of origin. Multi-enterocyte waves traveled almost exclusively through interstitial cells and rarely through copper cells. GCaMP signals appeared repeatedly in the same cell over time as part of multi-directional waves that travelled through the tissue (Fig. 2B). In this example (Movie 1), individual interstitial cells spiked, on average, every 23 s, leading to an average frequency of 44±1.4 mHz.

To assess whether individual waves traveled along recurring paths, we followed GCaMP signals in selected regions of a single midgut (Fig. 3A–D), which encompassed eight to 13 cells in an area of ∼3000 µm². Based on the fluorescence intensity, we estimated cell outlines and identified copper cells based on their unique cell shapes and signals (Fig. 1B); interstitial cells were identified as the space between the copper cells (Dubreuil, 2004; Shanbhag and Tripathi, 2009). The waves did not, to the extent of our observation, exhibit a predictable pattern, even when they recurred in the same region. This is exemplified in Fig. 3E, where we traced the mean fluorescence intensity as a function of time in four adjacent regions of interest (ROIs), across approximately 50 µm in width from anterior to posterior. Over a 100 s interval, there were three large increases in the Ca²⁺ signal (∼3- to 5-fold above the baseline) and the signal traveled in both proximal-distal and distal-proximal directions.

To directly test whether Ca²⁺ oscillations in distinct cell types are coupled, we used the orthogonal driver systems LexA/LexAop (Lai and Lee, 2006; Szüts and Bienz, 2000) and GAL4/UAS (Brand and Perrimon, 1993) to express spectrally differentiable Ca²⁺ sensors in two cell types simultaneously, then we performed multi-channel imaging to record the signal from both. We labeled progenitor cells using the insertion StanEx^SJH1, from the StanEx collection of LexA-based enhancer trap drivers (Kockel et al., 2016), in which the LexA transcription factor is inserted upstream of the transcription start site for escargot. For convenience, we refer to this insertion as esg-LexA (Fig. 4A,B; Figs S1 and S2). In combination with mex-GAL4, we could visualize Ca²⁺ dynamics directly in enterocytes and progenitors, simultaneously (Fig. 4C,D; Movie 4).

We selected ROIs to cover a region a little larger than a single progenitor (Fig. 4A), so that the cell of interest remained within the ROI even as the gut shifted due to the contracting muscle cells. Mean fluorescence intensities of the ROIs were traced as a function of time for both GCaMP (enterocyte) and RCaMP (progenitor) channels (Fig. 4E). We illustrate the feasibility of using the orthogonal driver systems to simultaneously image Ca²⁺ dynamics in interstitial cells and progenitors. Characteristic oscillations of ∼24 mHz were observed for interstitial cells in the GCaMP channel and wide signal elevations (20 s) in the RCaMP channel; these were comparable to observations when using a single-driver system (Fig. 1D).

To quantitatively assess whether Ca²⁺ oscillations of progenitors and enterocytes were correlated, we analyzed the four ROIs (Fig. 4E) that covered progenitor cells 1–4 and their juxtaposed enterocytes (direct enterocyte-progenitor pairs), by applying a formula for cross correlations between two time series with time lag h (see Table S4, cross correlation with time lag for finite time series). With time lag 0, which corresponds to the Pearson correlation, the enterocyte-to-progenitor cross correlations for the four direct enterocyte-progenitor pairs in the ROIs (enterocyte 1 to progenitor 1, enterocyte 2 to progenitor 2, etc.) have absolute values <0.2 for all four direct pairs (these four correlation values are approximately 0.064, 0.032, −0.10 and 0.19), indicating weak correlation. When all non-zero time lags are considered, the cross correlations for direct enterocyte-progenitor pairs never exceed an absolute value of 0.32 (Fig. S4). These values do not support the hypothesis that progenitor and enterocyte oscillations are correlated.

We also considered cross correlations between enterocyte-progenitor pairs in which the enterocyte and the progenitor are in different ROIs. Table S4 reports all the correlation values for all direct and non-direct pairs for time lags up to 12 s.

We also used the combination of the GAL4/UAS and LexA/LexAop systems to simultaneously visualize Ca²⁺ dynamics in enteroendocrine cells (pros-GAL4>UAS-GCaMP) and progenitors (esg-LexA>LexAOp-jRCaMP). Similar to enterocytes and progenitors, we observed that Ca²⁺ oscillations in enteroendocrine cells and progenitor cells were independent of each other (Movie 5).

Gap junctions are intercellular channels that allow direct transfer of small molecules and ions. Thus, cells sharing gap junctions are electrically coupled. For example, in the regenerative basal layer of the skin epithelium, directed Ca²⁺ signaling is regulated by a major gap junction protein (Moore et al., 2021 preprint). As the Ca²⁺ waves we observed appeared to traverse several cell lengths, we hypothesized that Ca²⁺ ions propagate across cells via gap junctions. To examine this hypothesis, we blocked gap junctions by adding the small-molecule inhibitor carbenoxolone (CBX) (Balaji et al., 2017; Spéder and Brand, 2014) to the imaging medium for 15 min prior to and during imaging midguts with cell-type-specific expression of Ca²⁺ indicators. We analyzed the fluorescence intensity of Ca²⁺ indicators in individual cells as a function of time for each cell type in control and CBX-treated guts (Fig. 5D,E; Movies 6–8).

We found that gap junction inhibition sharply inhibited Ca²⁺ oscillations for all cell types, but the consequence to cytosolic Ca²⁺ levels differed depending on the cell type. Although interstitial cells exhibited weaker GCaMP signals with CBX treatment, copper cells exhibited substantially higher GCaMP signals (Fig. 5A). One possible explanation for this unexpected result is that copper cells might accumulate cytosolic Ca²⁺, perhaps from extracellular sources, when Ca²⁺ flux through interstitial cells is blocked.

CBX treatment eliminated multicellular Ca²⁺ waves that normally propagate through interstitial cells. CBX treatment also abrogated Ca²⁺ oscillations in enteroendocrine cells (Fig. 5B) and progenitors (Fig. 5C). As the oscillations of progenitors were not coupled to enterocytes (Fig. 4), this result was unexpected; potentially, whole-organ inhibition of gap junctions by CBX could disrupt organ-scale Ca²⁺ homeostasis with consequent inhibition of single-cell oscillations. Consistent with this notion, Ca²⁺ oscillations in the intestinal stem cells have been shown to depend on both the influx of Ca²⁺ ions through the plasma membrane and on the internal release of Ca²⁺ ions that have been actively sequestered in the intracellular stores by sarco-endoplasmic reticulum Ca²⁺-ATPase (SERCA) (Deng et al., 2015).

Examining Ca²⁺ dynamics in the midgut of adult Drosophila, we found that differently fated cells in the midgut R3 region exhibited characteristic patterns of Ca²⁺ oscillations. Performing live imaging of midguts ex vivo, we identified propagation of Ca²⁺ waves through networks of interstitial cells and characterized Ca²⁺ oscillations in enteroendocrine cells and progenitors. Employing orthogonal expression of red and green Ca²⁺ sensors, we demonstrated that the Ca²⁺ dynamics of enterocytes and progenitors are paced independently of one another. We also found that the Ca²⁺ dynamics of progenitors and enteroendocrine cells are paced independently. Our work demonstrated that whole-organ inhibition of gap junctions eliminated dynamic Ca²⁺ responses in all three cell types and led to an accumulation of Ca²⁺ ions in copper cells. Thus, gap junction inhibition potentially disrupts organ-scale Ca²⁺ homeostasis.

An intriguing aspect of our results is that differentiated cell types in the same epithelium, despite being derived from the same progenitor cell population, exhibit Ca²⁺ dynamics that differ from their mother cells and from each other. This finding implies that as stem-cell progeny undergo differentiation, they adopt fate-specific modes of Ca²⁺ pacing. Ca²⁺ signal integration remains an intriguing open question due to its pleiotropic and ubiquitous nature. One possibility is that unique oscillatory Ca²⁺ patterns are interpreted by downstream effectors to subsequently activate different cellular processes; another is that the Ca²⁺ dynamics support the physiological functions of the specific cell types.

What are these cell-type-specific physiological functions? Ca²⁺ oscillations in stem cells can influence division rate (Deng et al., 2015), whereas in terminal progenitors they regulate enteroendocrine differentiation (He et al., 2018). The functions of Ca²⁺ signaling in mature enteroendocrine cells and enterocytes remain to be elucidated, but several studies in developing tissues and in cultured cells raise attractive possibilities. For example, in developing epithelial tissues, Ca²⁺ waves and spikes have been implicated in organizing actomyosin networks (Balaji et al., 2017; Hoyer et al., 2020 preprint; Ready and Chang, 2021), remodeling and repair of damaged occluding junctions (Varadarajan et al., 2022), responding to exogenous mechanical loads (Narciso et al., 2017) and coordinating growth at the tissue scale (Brodskiy et al., 2019). It is conceivable that Ca²⁺ waves in midgut enterocytes have similar roles, as enterocytes must maintain cytoskeletal networks and occluding-junction-based barrier function during continual mechanical compression due to peristalsis. Another fascinating possibility comes from studies of the Caenorhabditis elegans intestine, which identified gap-junction-mediated Ca²⁺ waves in intestinal epithelial cells as essential for coordinating rhythmic defecation (Espelt et al., 2005; Peters et al., 2007; Teramoto and Iwasaki, 2006). Finally, in cultured cells, the frequency of Ca²⁺ oscillations and spikes has been shown to control differential activation of transcription factors (Dolmetsch et al., 1997; 1998; Li et al., 1998). It is conceivable that the distinct frequencies of Ca²⁺ oscillations in midgut enterocytes, enteroendocrine cells and/or progenitors are tuned to regulate the activity of cell-type-specific transcription factors.

In the midgut, as in other multicellular systems, gap junctions are crucial for Ca²⁺ oscillations (Leybaert and Sanderson, 2012). These cell-cell junctions are oligomers of connexins (vertebrates) or innexins (invertebrates), with eight innexin genes identified in Drosophila that encode at least ten transmembrane proteins (Adams et al., 2000; Stebbings et al., 2002). Bulk transcriptomic analysis of the R3 region indicates that enterocytes, enteroendocrine cells and progenitors express innexin7 (Inx7) and, to a lesser extent, innexin2 (Inx2) (Dutta et al., 2015); single-cell transcriptomic data also reveal the expression of innexin7 and innexin2 in enterocytes, but suggest that innexin3 (Inx3) is the predominant innexin in progenitor and enteroendocrine cells (Hung et al., 2020; Li et al., 2022). Future work to establish the molecular composition and connectivity of gap junctions in the midgut epithelium will be invaluable for understanding the regulation and functions of Ca²⁺ dynamics.

In summary, our findings establish a novel model for studying cell-type-specific Ca²⁺ dynamics within a heterogeneous population of stem and differentiated cells in an adult tissue. Our observations lay the groundwork for using this highly tractable genetic model to investigate the roles of spatiotemporal Ca²⁺ changes within signal transduction, organ renewal and stem cell differentiation.

We obtained 20XUAS-IVS-NES-jRCaMP1b-p10 (BL63793), 20XUAS-IVS-GCaMP6s (BL42746), 13XLexAop2-IVS-NES-jRCaMP1b-p10 (BL64428) and P{ST.lexA::HG}SJH-1 (Kockel et al., 2016) (BL66632, referred to as esg-LexA in this paper, Figs S1 and S2) from the Bloomington Stock Center, Indiana University, IN, USA. esg-GAL4 was obtained from the Department of Drosophila Genomics and Genetic Resources (DGGR), Kyoto Institute of Technology, Japan. The following stocks were gifts: mex-GAL4 (Carl Thummel, Howard Hughes Medical Institute University of Utah Salt Lake City, UT, USA), esg-GFP[KI]/CyO (Norbert Perrimon, Department of Genetics, Harvard Medical School, Boston, MA, USA) and pros-GAL4 (Sarah Siegrist, Department of Biology, University of Virginia, Charlottesville, VA, USA) (Matsuzaki et al., 1992). The complete list of key resources can be found in Table S1.

Flies were fed a diet of standard cornmeal molasses food at 25°C or room temperature. Flies were collected 0–24 h post eclosion, placed in vials with males and shifted to 25°C with 12 h light on and 12 h light off. The flies were fed a diet of standard cornmeal molasses food supplemented with yeast paste (Red Star, Active Dry Yeast) and the food vials were changed every 1–3 days. Experiments were performed on female flies, 4–7 days post eclosion.

To understand the expression of esg-LexA, esg-LexA>LexAop-jRCaMP1b was co-expressed with esg-GAL4>UAS-his2b::CFP and esg-GFP in two separate experiments. Comparison between esg-LexA>LexAop-jRCaMP1b and esg-GAL4>UAS-his2b::CFP has shown that esg-LexA>LexAop-jRCaMP1b almost always colocalizes with esg-GAL4>UAS-his2b::CFP (Matsuzaki et al., 1992). There are few cells that express esg-GAL4>UAS-his2b::CFP, but not esg-LexA>LexAop-jRCaMP1b, possibly because LexAop-jRCaMP1b is a transient signal. On occasion, we observed cells that express esg-LexA>LexAop-jRCaMP1b but not esg-GAL4>UAS-his2b::CFP (Fig. S1B). In general, esg-LexA>LexAop-jRCaMP1b exhibits a weaker signal. esg-LexA>LexAop-jRCaMP1b also colocalized with esg-GFP[KI] (esg-GFP) and similarly, the signal is stronger in esg-GFP (Matsuzaki et al., 1992). A large majority of esg-LexA>LexAop-jRCaMP1b cells also expressed esg-GFP. There are several cells that express esg-GFP but not esg-LexA>LexAop-jRCaMP1b.

Female flies were briefly anaesthetized on carbon dioxide and then placed on ice in Eppendorf tubes to induce a chill coma. Guts were dissected in room temperature in adult hemolymph-like (AHL) medium (108 mM NaCl, 5 mM KCl, 2 mM CaCl₂, 8.2 mM MgCl₂, 4 mM NaHCO₃, 1 mM NaH₂PO₄, 5 mM trehalose, 10 mM sucrose, 5 mM HEPES, pH 7.5, prepared by Electron Microscopy Sciences) and then bathed in whole-organ ex vivo culture medium with 10 µg/ml isradipine (Selleck Chemicals) to decrease muscle contractions. The guts were then transferred in a droplet to #1.5 coverslips coated with poly-L-lysine (P4832-50ML, Sigma-Aldrich) with 120 µm spacers (620,001, Grace Bio-Labs) and sealed with 250 µm-thick polydimethylsiloxane sheets (see Fig. S3). The assembled view of the ex vivo midgut mount for inverted microscopy is illustrated in Fig. S3.

For experiments with the gap junction inhibitor, carbenoxolone (C4790-1G, Sigma-Aldrich) was added to the ex vivo culture medium (100 µM CBX) and the tissues were incubated in this medium for 15 min prior to imaging.

The medium for ex vivo midgut culture was used as previously described (Marchetti et al., 2021 preprint). We supplemented a base of Schneider's Medium (21,720,024, Thermo Fisher Scientific) with 55 mM L-glutamic acid monosodium salt (AAJ6342409 Alfa Aesar, Thermo Fisher Scientific, diluted from 1 M stock prepared in Schneider's medium), 50 mM Trehalose (T5251-10G, Sigma-Aldrich, diluted from 1 M stock prepared in Schneider's medium), 2 mM N-acetyl cysteine (antioxidant that delays phototoxicity) (A9165-5G, Sigma-Aldrich, diluted from a 200 mM stock prepared in sterile water), 1 mM tri-sodium citrate (antioxidant that delays phototoxicity) (PHR1416-1G, Sigma-Aldrich, diluted from a 1 M stock prepared in Schneider's medium) and 5 mM HEPES (from 1 M solution, H0887-20 Ml, Sigma-Aldrich).

An inverted Leica SP8 resonant scanning confocal microscope with a 40×/1.1 water-immersion objective was used to acquire movies that were analyzed in this study. Movies were captured at room temperature (20–23°C). Confocal stacks were acquired with a z-step of 2 or 4 µm and typically contained three to nine slices. For the complete list of movies and their information, including the genotypes, see Table S5. The movies and fluorescence intensity data are available on the European Molecular Biology Laboratory's European Bioinformatics Institute (EMBL-EBI) BioImage Archive under the accession number S-BSST849. Movies were captured with cycle time of 4.7 s or faster. Movies were included in data analysis (Fig. 5E) only if they were at least 290 s long. A Zeiss Z.1 Lightsheet was used with a 20×/1.0 objective by flowing a midgut into tubing (FEP009-031-B, Western Analytical) to obtain a high-resolution z-stack of the R3 region for Fig. 1B.

Movie stacks were imported into Fiji (Schindelin et al., 2012) using the Bio Formats (Linkert et al., 2010) plugin. In the acquired movie stacks, we visually identified a plane that corresponded primarily to copper cells or to interstitial cells. In the selected plane, we drew ROIs manually to correspond to the cell type of interest, then we obtained mean fluorescence intensity values for each time point in Fiji. In the case of progenitors and enteroendocrine cells, we first converted z-stacks into maximum projection movies in Fiji. We identified cells of interest manually and then tracked the mean fluorescence intensity of cells using the Active Contours (Dufour et al., 2011) plugin in Icy (de Chaumont et al., 2012). We plotted mean fluorescence intensity values (F) relative to the minimum fluorescence values (F_min) and peaks were identified using custom codes in MATLAB 2019b (available upon request) with the aid of the Signal Processing Toolbox. Peaks were manually verified.

The average oscillations in mHz for the different cell types are presented in Tables S2 and S3, per midgut. The average oscillations are also presented if only cells exhibiting at least one oscillation are included in the analysis.

We acknowledge the Neuroscience Research Institute and the Department of Molecular, Cellular and Developmental Biology (NRI-MCDB) Microscopy Facility, use of the Resonant Scanning Confocal microscope, supported by the National Science Foundation (NSF) Major Research Instrumentation Program (MRI) grant DBI-1625770, and use of the Microfluidics Laboratory within the California NanoSystems Institute, supported by the University of California, Santa Barbara and the University of California, Office of the President. We thank Anthony Galenza for bioinformatics assistance; the Bloomington Drosophila Stock Center (National Institutes of Health, P40OD018537), the Kyoto Drosophila Genomics Resource Center (DGRC), Carl Thummel, Norbert Perrimon and Sarah Siegrist for fly stocks; and Jon-Michael Knapp for writing assistance.

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.260249.