Current live-imaging protocols for the adult Drosophila midgut are limited to 16 hr imaging sessions (Deng et al., 2015; Xu et al., 2017; He et al., 2018; Martin et al., 2018; Hu and Jasper, 2019). To extend the survival of midguts ex vivo we developed a novel culture setup. The method is based on common available techniques for the culture of adult organs (Sato et al., 2009; Handke et al., 2014; Reilein et al., 2018; Hu and Jasper, 2019) but it includes a refinement of several steps: (1) the dissection procedure was optimized to reduce tissue damage; (2) explanted organs are cultured in a sandwiched structure of agarose, rather than in a dome; (3) midguts are placed at a liquid-air interface for improved oxygenation; (4) the culture media has been adjusted to better approximate adult hemolymph. Please see the Materials and methods section step-by-step descriptions of the procedure.

We found that the dissection technique is a key parameter in extending the viability of explanted midguts. Indeed, any stress (e.g. pulling and thus stretching the gut, pinching, etc...) introduced during dissection results in structural damage which lead to breaks of the epithelium during prolonged culture. Hence, we have optimized the dissection procedure to limit the handling of the midgut and thus the risk of damaging it (Figure 1A–N and Figure 1—video 1).

A key element of our system is enveloping explanted organs in a sandwich of low-melt agarose (Figure 1O). Immobilization in gels is a common solution for ex vivo culture of organs and tissues, and is especially important to provide stability for live-imaging applications (Sato et al., 2009; Handke et al., 2014; Reilein et al., 2018; Hu and Jasper, 2019). For the adult Drosophila midgut specifically, an agarose gel also minimizes the effects of peristaltic movements, which, if uncontrolled, will impair imaging and can lead to epithelial tearing. Our approach is a slight departure from previously published techniques for Drosophila tissues (Handke et al., 2014; Reilein et al., 2018; Hu and Jasper, 2019) in that explanted midguts are transferred to evenly spaced thin agarose pads and then covered with an additional layer of agarose (Figure 1O). This sandwiched structure allows the guts to be held in place for imaging, while also protecting them from damage that can result from their contact with the culture chamber walls if left freely floating. The agarose pads can be reproducibly cast and are thin enough (~100 µm) not to interfere with imaging. Moreover, each agarose pad can comfortably house up to 3 midguts, allowing the multiple imaging of several explanted intestines. The agarose pads have the additional function of elevating the midguts from the bottom of the imaging chamber so that the surface of the agarose structure is directly exposed to air, creating an air-liquid interface. This is a key design element of the setup, as proper oxygenation was found to be essential for the long-term survival of explanted midguts (data not shown), similarly to what had been previously observed for the culture of wing imaginal discs (Strassburger et al., 2017). Moreover, during dissection the trachea surrounding the intestine are severed and can no longer oxygenate the intestinal epithelium.

To increase the stability of the setup, agarose bridges connect each agarose sandwich (Figure 1O, middle panel), allowing the sample to endure 3 days of continuous imaging (Figure 2C and Figure 2—video 1). Moreover, using a microscope equipped with an incubation chamber to control evaporation removes the need to replenish the imaging vessel with fresh culture medium. As such, the culture system, despite its simple design, is highly efficient and well suited for long-term imaging experiments.

One of the factors currently limiting the extended survival of explanted adult Drosophila tissues is the lack of culture media specifically designed for this task. To obviate this issue, we analyzed several parameters that distinguish the hemolymph of larvae, on which most current Drosophila cell culture media are based, to that of adult flies. We therefore compared the performance of different media formulations using the incorporation ex vivo of the cell-impermeable dye NucGreen (Thermofisher) as a measure of cell death (Figure 2A). The dye was added at the start of the culture, and its accumulation in the tissue measured over the course of three days. At the start of the culture, we usually observed NucGreen incorporation only in trachea, which are inevitably severed and damaged as a result of the dissection. We found that raising the concentration of trehalose, which is found at high levels in Drosophila circulation (Pasco and Léopold, 2012; Dus et al., 2013; Park et al., 2014; Tennessen et al., 2014; Matsushita and Nishimura, 2014), and mimicking the Na⁺/K⁺ ratio of adult hemolymph (Singleton and Woodruff, 1994; Naikkhwah and O’Donnell, 2014; MacMillan et al., 2015b; MacMillan et al., 2015a; Olsson et al., 2016) is sufficient to significantly reduce cell death after 24 hr of culture. As Schneider’s medium (Schneider, 1964; Schneider, 1966; Schneider, 1972; Schneider and Blumenthal, 1978) is a common solution for several published protocols for the live-imaging of adult Drosophila tissues (Fichelson et al., 2009; Reilein et al., 2018; Cheng and Hunt, 2009; Xu et al., 2017; Martin et al., 2018), we modified it to raise the trehalose concentration to 50 mM and Na⁺/K⁺ ratio to levels similar to those found in adult Drosophila hemolymph (Table 1; Dus et al., 2013; Park et al., 2014; Tennessen et al., 2014; Singleton and Woodruff, 1994; Naikkhwah and O’Donnell, 2014; MacMillan et al., 2015b; MacMillan et al., 2015a; Olsson et al., 2016). Our tests showed that supplementing the culture medium with 10% fetal bovine serum (FBS) did not have a significant beneficial effect on the midgut epithelium (Figure 2A), but we did notice that FBS addition resulted in the visceral muscle remaining intact and capable of peristalsis for longer (data not shown). Not surprisingly (Davis and Shearn, 1977; Britton and Edgar, 1998), co-culturing explanted midguts with ovaries and abdomens lined with fat body (adipocytes) could also decrease cell death ex vivo. This effect was not due to the sequestration of NucGreen by ovaries and adipocytes as the dye was supplemented at a saturating concentration and its incorporation in these organs/tissues was mostly limited to areas mechanically damaged during dissection. As ovaries and fat bodies may enhance the survival of explanted midguts by secreting growth factors and/or nutrients, we reasoned that fly extract might have a similar effect. Indeed, midguts cultured in 100% fly extract prepared using modified Schneider’s medium had greatly reduced rates of cell death over time (Figure 2A). The calcium blocker isradipine was also added to the medium to reduce peristaltic movements, improving midgut survival and imaging. This drug did not have a noticeable long-term effect on visceral muscle viability. Indeed, its effect appeared to wear off after 48–72 hr, at which point the visceral muscle resumed regular peristalsis (Video 1). Lastly, as we designed the medium for the purpose of extended live-imaging, we also added N-acetyl cysteine and sodium citrate as antioxidant agents to reduce phototoxicity (Icha et al., 2017). The former is an antioxidant widely used in cell culture (Ezeriņa et al., 2018). Citrate is also known to have antioxidant properties (Wu et al., 2019) and to be present in Drosophila hemolymph at detectable levels (Echalier et al., 2018). Combining all the findings mentioned above resulted in culture conditions (see Materials and methods) that minimized cell death and allowed the live-imaging of explanted midguts for up to 3 days (Figure 2C and Figure 2—video 1), a significant improvement over previously published culture protocols for the midgut in which imaging was limited to 16 hr (Martin et al., 2018; Hu and Jasper, 2019).

We successfully imaged cultured intestines with inverted microscopes of different kinds, including widefield (Nikon TiE), scanning confocal (Leica SP8), and lattice lightsheet (Zeiss LLS7). We did not notice signs of phototoxicity with any of these microscope types in samples imaged for 48 hr, but acquisition parameters had to be chosen with care. For example, by optimizing laser power and dwell time, single intestines could be imaged at high frame-rates (15 min) without noticeable phototoxicity for at least 48 hr even with a confocal microscope (Video 5). However, we found confocal imaging to be less optimal than widefield microscopy for high frame-rate imaging due to the longer acquisition time required for multiple midguts in a single sample.

In examining cultured midguts, we observed the accumulation of luminal contents (i.e. previously ingested food) in the posterior midgut (Figure 2B, white arrows). This appeared to be caused by peristaltic movements of the visceral muscle which persisted despite the use of the calcium blocker isradipine, although this drug suppressed them significantly. These areas were found to darken and expand over the course of culture, an effect we attribute to the growth of enteric bacteria, which eventually caused cell death and tissue damage. We found that feeding flies fresh food in the days prior to an experiment and a short (~3–6 hr) starvation prior to dissection reduced the negative effect of food accumulation and growth of enteric bacteria. Supplementing the culture medium with antibiotics also enhanced explanted midgut survival. Lastly, it is reasonable to assume that axenically reared flies should be immune to the issue of growing enteric bacteria.

One of the most striking features of Drosophila melanogaster as a research model is the wide range of readily available genetic tools, allowing the cell-type-specific and temporally-controlled activation or suppression of expression of genes of interest. The possibility to use such genetic tools in a live-imaging setup is highly advantageous, but so far the temporal control of gene expression ex vivo has proven unfeasible (Martin et al., 2018). To further assay the behavior of midgut cells in explanted organs, we tested the induction of UAS-GFP by cell type-specific, temperature-sensitive Gal4 drivers (Figure 2D–E; Brand and Perrimon, 2018; McGuire et al., 2003). As expected, midguts explanted from flies grown at the restrictive temperature (18 °C) did not show any GFP expression (Figure 4—video 1 and Figure 2—figure supplement 1, left panels). However, when incubated at the permissive temperature (29 °C) at 0 hr after explant (Figure 4—video 1 and Figure 2—figure supplement 1, right panels), they started expressing the UAS-GFP transgene. Moreover, GFP expression could also be induced in intestines cultured at 18 °C for 24 hr and then shifted to the permissive temperature (29 °C), indicating that the epithelium remains viable and genetically functional long-term (Figure 2—figure supplement 1B). Indeed, the GFP intensity in midguts cultured at 29 °C for 24 hr from the time of dissection was similar between in vivo and ex vivo conditions for both progenitor cells (Figure 2D) and enterocytes (Figure 2E) using the escargot- (esg) and mex-Gal4 Gal80^TS driver lines, respectively. This indicates that transcription and protein synthesis are maintained at normal levels in our culture system and shows that this system can be used to assay the effects of transgene induction in real time. Interestingly, progenitor cells were found to be asynchronous in their expression of the reporter GFP (Figure 4Aand Figure 4—video 1). Usually, the first GFP + cells were detected at 8–10 hr after temperature shift, though some cells achieved detectable GFP levels only 20 hr after that. Overall, if the shift to 29 °C coincided with the start of imaging, all progenitor cells usually expressed detectable levels of GFP by the 36 hr time-point. This could be explained by variations in the activity of the esg promoter or by different global rates of transcription and translation, which in turn could be indicative of different cell states.

When imaging explanted adult Drosophila midguts we observed esg-expressing progenitor cells (ISCs and EBS) to be quiescent (Figure 3A and Figure 2—video 1, Figure 3—video 1, Figure 3—video 4). Progenitor cells in explanted healthy guts did not show major changes in their GFP levels, indicative of continuous esg expression, nor in their nuclear size (Figure 3D, F and G and Figure 3—video 4), which indicates that DNA content remains constant. Via cell tracking we also observed that GFP-labeled progenitor cells did not divide in healthy explanted intestines (Figure 4). It is important to note that in said healthy intestines we did not observe cell death or enterocyte (EC) extrusion events until after 48–72 hr of culture. As healthy enterocytes are known to suppress ISC proliferation (Liang et al., 2017), this may explain the lack of cell division in our explants. The suppression of peristaltic movements by isradipine may also prevent EC loss by reducing mechanical tissue stress (Li et al., 2018).

A key feature of Drosophila intestinal progenitor cells is their ability to rapidly respond to tissue damage (Ohlstein and Spradling, 2006; Micchelli and Perrimon, 2006; Jiang et al., 2009; Patel et al., 2019; Amcheslavsky et al., 2009). To confirm that this capability is maintained ex vivo, we fed female flies Sodium dodecyl sulfate (SDS) mixed with solid fly food at a final concentration of 0.2% (v/w) for 18 hr. Following this treatment, the flies were fed 0.05% sucrose in aqueous solution on a cotton pad for an additional 4–6 hr prior to dissection, in order to flush the SDS from the gut. This protocol caused only mild initial tissue damage, and was advantageous as it did not result in the retention of high levels of SDS in explanted guts. However, by mixing the SDS-laced food with a blue food-safe dye, we found that some low amounts of SDS-laced food do persist in the lumen of intestines at the time of dissection. The transient exposure to SDS and the low amounts retained in the lumen resulted in gradual EC death and/or extrusion, allowing the live-imaging of damage response. When imaged, intestines from SDS-fed flies showed progenitor cells that become highly motile, similar to what was recently observed in similarly damaged intestines (Hu et al., 2021) . ISC proliferation events could also be detected (Figure 3C). However, ISC mitoses could generally be observed only in cases where tissue damage was extensive, as evidenced by the appearance of pyknotic or fragmented nuclei and the extrusion of multiple (>3) contiguous ECs. As this tended to happen towards late time-points after dissection, only a few (1-2) mitotic events per imaged field could be observed with this damage protocol. In addition to ISC divisions, many progenitor cells could be observed growing in nuclear size while simultaneously losing GFP expression (Figure 3B and E–G and Figure 3—video 3, Figure 3—video 4) indicative of differentiation events towards the EC identity. Indeed, differentiating progenitor cells could be seen replacing dying enterocytes (Figure 3—video 3).

We then tested whether these observations could be made in an in vivo condition. Before dissection, flies were fed either a control diet or SDS-laced food as described above. Flies dissected at later time-points (i.e. 24 and 48 h after the initial SDS feeding) where kept on fresh food, thus allowing the intestine to clear any trace of SDS and recover. This is different to the ex vivo condition, where a small amount of SDS is retained in the lumen and continues to damage the intestine during imaging. Although we found it impossible and impractical to perfectly match the luminal SDS content in vivo and ex vivo, the in vivo condition did confirm our ex vivo observations (Figure 3—figure supplement 1). After SDS feeding, we observed several esg + cells with larger nuclei (Figure 3—figure supplement 1A-B). At later time-points many differentiating cells with large nuclei and low GFP expression were also present, similarl to what was observed ex vivo. Moreover, similarly to explanted midguts, proliferative cells could be found almost exclusively at 48 hr after the end of the SDS treatment (Figure 3—figure supplement 1C).

A limitation of the SDS-feeding protocol described above is that midguts were damaged prior to imaging. Therefore, it cannot be excluded that the observed differentiation events were EBs that were poised to differentiate prior to SDS feeding, and then differentiated due to the damage stimulus. To confirm that, at the time of damage, pre-existing EBs were indeed capable of responding to tissue stress, we used a thin tungsten needle to create a small lesion in explanted intestines. The lesions perforated both visceral muscle and epithelial layers, but left the peritrophic matrix intact (Figure 3—figure supplement 2A). Intestines were then enveloped in agarose and imaged within 20’ of the time of damage. Similarly to the SDS damage protocol, we were able to record progenitor cells both dividing and differentiating (Figure 3—figure supplement 2B, C and Figure 3—video 5). Moreover, wounding stimulated the robust expression of the cytokine Unpaired 3 (Upd3), whose function in midgut damage response is well documented (Figure 3—figure supplement 2D, E and Figure 3—video 6; Figure 3—video 7; Jiang et al., 2009; Patel et al., 2015; Buchon et al., 2009).

All in all, our observations suggest that both ISCs and EBs are quiescent in undamaged intestines. While previous studies did show varying rates of ISC proliferation in healthy midguts in vivo (Ohlstein and Spradling, 2006; Micchelli and Perrimon, 2006; Jiang et al., 2009; Reiff et al., 2019; de Navascués et al., 2012), this may be attributed to cell death (Liang et al., 2017) as well as the passage of food through the intestinal lumen (Li et al., 2018), both of which do not occur in our midgut cultures until later time-points (i.e. after 48–72 hr). Moreover, several previous studies used lineage-tracing tools requiring a 37 °C heat shock for their activation, a treatment known to induce ISC proliferation (Beebe et al., 2010; de Navascués et al., 2012; Lin et al., 2008; Shaw et al., 2010; Strand and Micchelli, 2011). Indeed, in the presence of tissue damage, progenitor cells could be robustly activated in explanted intestines, indicating that both ISCs and EBs are capable of responding to their environment after organ explant.

A key pathway necessary for progenitor cell differentiation is the Delta/Notch signaling that occurs between ISCs and adjacent EBs (Ohlstein and Spradling, 2006; Micchelli and Perrimon, 2006; Ohlstein and Spradling, 2007). The depletion of Notch (N) by RNA interference (N^RNAi) in esg⁺ cells has been shown to promote the formation of undifferentiated tumor-like masses in vivo (Ohlstein and Spradling, 2006; Micchelli and Perrimon, 2006; Patel et al., 2015; Ohlstein and Spradling, 2007). Notably, tumor initiation by N-depleted ISCs requires proliferation induced by tissue stress (Patel et al., 2015; Apidianakis et al., 2009). Consistent with this, when we explanted midguts from flies expressing N^RNAi in progenitor cells for 24 h prior to dissection and cultured them for 48 hr, we did not observe tumorigenesis in the absence of tissue damage (Figure 3—figure supplement 3A and Figure 3—video 8). However, when guts were accidentally damaged during dissection (Figure 3—figure supplement 3, yellow ellipses), esg⁺ cells started to proliferate (Figure 3—figure supplement 3B, yellow squares, and Figure 3—video 9). Notably, progenitor cells remained small and did not lose GFP expression (compare Figure 3B and Figure 3—figure supplement 3B as well as Figure 3—video 2 and Figure 3—video 9), suggesting a block of differentiation ex vivo in response to Notch knock-down.

EGFR-Ras-Erk signaling is activated upon gut tissue damage, and this pathway is required for stem cell activation in the adult Drosophila intestine (Buchon et al., 2010; Biteau and Jasper, 2011; Jiang et al., 2011). Previous experiments showed that expression of a constitutively active form of Ras (Ras^G12V) strongly promotes stem cell proliferation in adult midguts (Jiang et al., 2011; Jin et al., 2015). As our culture protocol allows the expression of transgenes ex vivo, we induced Ras^G12V in explanted midguts using the esg-Gal4 Gal80^TS progenitor-specific driver gene combination (esg^TS). When explanted midguts were shifted to the permissive temperature (29 °C) at the start of imaging, progenitor cells, marked by GFP co-expressed with Ras^G12V, started to rapidly proliferate (Figure 4B). About 20% of progenitor cells tracked from the moment they expressed visible amounts of GFP till the end of the imaging session, were seen proliferating (Figure 4D). The progeny of observed mitoses were not counted for this analysis. However, since many cells were not trackable for the duration of the imaging session due to major tissue rearrangements, we may be underscoring the percentage of proliferative GFP⁺ cells. As GFP⁺ cells accumulated in the tissue, enterocytes were displaced and extruded from the epithelium (Figure 4B, yellow arrowhead, and Figure 4—video 2). Interestingly, some GFP⁺ progenitor cells did not proliferate, but their nuclei rapidly grew in size, which aligns with the previously reported role of EGFR signaling in promoting EB growth (Jiang et al., 2011; Xiang et al., 2017). Moreover, these rapidly growing progenitors also lost GFP expression during the course of imaging, which suggests their differentiation towards the EC lineage.

The extended live-imaging that our protocol allows also permitted us to follow cells through multiple rounds of mitosis. Using manual 3-dimensional (3D) cell tracking, we reconstructed the lineages of 17 dividing cells that expressed Ras^G12V (see Figure 5 and Figure 5—video 1 for an example). As ISCs are the only cell type in the Drosophila intestine that normally divide multiple times (Ohlstein and Spradling, 2006; Micchelli and Perrimon, 2006; Chen et al., 2018), the founding cells in these lineages were most likely ISCs. However, a recent work suggested that Ras^G12V can push a small number of EBs to de-differentiate to an ISC state (Tian et al., 2021). We directly measured the duration of progenitors’ cell cycle in these lineages to be 8±2.76 hr (Figure 4F). Interestingly, of 17 Ras^G12V-expressing ISC lineages characterized, one had divisions where both daughter cells from the first recorded division were seen dividing further (Figure 4E). For simplicity, pairs of dividing daughter cells will be henceforth be referred to as ‘co-dividing siblings’ (see Figure 6A–B and Figure 6—video 1 for an example). These are most likely symmetric divisions that give rise to two new ISCs. However, since we did not track differentiation markers in our lineages, other possibilities cannot be excluded. For example, after being generated from an ISC division, enteroendocrine (EE) progenitors are also known to divide once to give rise to two mature EEs (Chen et al., 2018). Moreover, we cannot exclude that cells that were not observed to divide during our imaging session would not divide again, given enough time.

Since Ras^G12V stimulation results in multiple cellular changes, we also tested whether we could stimulate ISC proliferation more directly. For this we used the esg^TS driver to co-express string (stg), a Cdc25C homolog, and Cyclin E (CycE) (Figure 4C and Figure 4—video 1). String directly activates Cdk1 to promote mitosis, whereas CycE directly activates Cdk2 to promote DNA replication and S-phase progression. The combined expression of these two gene products is sufficient to strongly induce ISC proliferation (Kohlmaier et al., 2015). Notably, Stg and CycE have also been shown to promote EB mitoses (Kohlmaier et al., 2015), although much less strongly than in ISCs. The fraction of progenitor cells that divided in response to stg and CycE co-expression (~25%; Figure 4D) was not significantly different from that observed after forced expression of Ras^G12V, suggesting that all receptive progenitors are activated in both cases. However, co-dividing siblings appeared in 6 of 12 ISC lineages that overexpressed stg and CycE (Figure 4E), a significantly higher frequency to the 1 of 17 observed after Ras^G12V overexpression (p=0.0106, Fisher’s exact test). We detected no significant difference in cell cycle duration in progenitors’ cell cycles driven by stg and CycE (9.4±4.6 hr) and cell cycles driven by Ras^G12V (8±2.76 hr; Figure 4F), suggesting that their differing abilities to produce co-dividing sibling cells may reflect different effects on the differentiation process.

Combining all the lineages described above, we were able to identify eight divisions that yielded co-dividing siblings (see Figure 6A–B and Figure 5—video 1 for an example). For seven of these, sibling cells divided within 2 hr of each other, while for one pair the interval was 5 hr. Based on this, we looked for divisions where one sibling cell was seen dividing further, while the other remained quiescent for the remainder of the experiment, which had to last at least 6 hr after the sister’s mitotic event (i.e. the second division in the lineage). For simplicity, pairs of siblings with this behavior will be henceforth be referred to as ‘non-co-dividing siblings’. These are most likely asymmetric divisions that give rise to an ISC/EB pair. However, since specific cell fate markers were not assayed in our experiments, this cannot be verified. Moreover, we cannot exclude that both cells would not eventually divide, given enough time. Regardless, the behavior of these daughter pairs was indeed distinct from that of co-dividing siblings described above. Following this definition, we classified 17 sibling pairs as non-co-dividing (see Figure 6D–F and Figure 6—video 2 for an example), 9 from Ras^G12V- and 8 from stg and CycE-expressing guts. Interestingly, for 8 of these 17 pairs, the non-dividing cells displayed increases in nuclear size over time, and also lost GFP expression, which is indicative of differentiation towards the EC cell fate (Figure 6C and G–H). This behavior was not observed for co-dividing siblings, where both cells increased their nuclear size before dividing further, but maintained GFP expression (Figure 6A and E–F). Hence, future studies using our culture and imaging techniques in conjunction with differentiation markers should be able to discern new details about ISC differentiation.

It was previously reported that spindle orientation is indicative of whether an ISC mitosis is asymmetric or symmetric, with symmetric divisions being planar relative to the visceral muscle layer, and asymmetric divisions displacing one daughter cell apically, away from the visceral muscle (Hu and Jasper, 2019; Ohlstein and Spradling, 2007; Goulas et al., 2012). As we restricted the interval between subsequent imaging frames to 1 h to reduce phototoxicity, we were not able to measure spindle orientations. However, it has been proposed that the orientation of the mitotic spindle results in daughter cells residing at different levels within the pseudostratified intestinal epithelium, with new-born EBs being more apical (Ohlstein and Spradling, 2007). We therefore selected mitotic events that occurred in regions of epithelium that were flat with reference to the imaging plane (based on cell nuclei positions). The 3D profile of sister cells after mitosis was then reconstructed and we measured the angles between sister cells. Measured angles were found to be <15° in 29/33 cases for Ras^G12V- and in 30/32 cases for stg +CycE-induced mitoses. Moreover, no significant differences in sister cell angles were observed between co-dividing and non-co-dividing siblings (Figure 7A–C). These observations do not support that view that apical displacement of daughter cells is associated with EB fate specification. However, as we did not image the mitotic spindle during mitosis, it is possible that spindles were oriented differently in divisions giving rise to the two sibling pair types, but that this difference in orientation was lost after cytokinesis.

A significant difference between co-dividing and non-co-dividing siblings was found, however, when we tracked newborn cells over time. Due to some cells having cell cycles lasting less than 8 hr, we only considered the first six 1 hr time-points following the appearance of a sister pair (defined as time-point 0). Measuring the distance between sister cells at each time interval, we found that cells in non-co-dividing pairs remained close to one another until the following mitotic event (Figure 7D, red line). Cells in co-dividing pairs, however, moved apart from one another (Figure 7D, green line). These behaviors could be observed even when considering unequal and co-dividing siblings in the same lineage (Figure 7E–F). If co-division and non-co-division are indeed indicative of symmetric and asymmetric division, respectively, this observation suggests that commitment to differentiation, as occurs after asymmetric divisions, requires that sister cells remain in contact for ~3–5 hr (Figure 7D and F, red line). This is in line with previous measurements of Notch activation dynamics in differentiating EBs (Martin et al., 2018). Conversely, rapid separation of sister cells following a division may be necessary to generate a symmetric division that duplicates ISCs.

Differences in the motility of co-dividing and non-co-dividing siblings could explain these observed effects. We therefore measured cell motility as the distance in 3D space that a cell travelled between one time-point and the next. We found that the genotype used to induce proliferation did not have an effect on the cell motility of either sibling type (Figure 7—figure supplement 1A). Likewise, no difference in cell motility was found when considering the effect of time on cell movement of non-co-dividing (Figure 7—figure supplement 1B,C) or co-dividing (Figure 7—figure supplement 1D) siblings. Therefore, we directly compared the motility of both pair types irrespective of genotype or time-point analyzed (Figure 7—figure supplement 1E), but also found no significant difference. This indicates that cells of both pair types migrate within the epithelium at similar speeds. Since non-co-dividing siblings tend to remain close to one another over time, their movement is most likely random. On the other hand, since the distance between co-dividing siblings increases over time, their movement is likely more directional, such that cells in a sibling pair move away from each other. It would therefore be expected that, when considering co-dividing pairs, the relative movement of one cell to its sister should be greater than that observed for non-co-dividing pairs. We therefore considered the motion of sister pairs, decomposing their movement in X- and Y-axis components and summed the resulting vectors, thus calculating the movement of a cell relative to its sister along the X- and Y-axis. As expected, the magnitudes of the reconstructed relative movements between co-dividing pairs was significantly greater than that for non-co-dividing pairs (Figure 7—figure supplement 1F). This confirms that the movement of co-dividing cell pairs was directional, with the two cells moving away from one another right after division.

To test the feasibility of our culture setup to sustain other adult Drosophila organs, we first focused on the Malpighian (renal) tubules. Being physically connected to the intestine, we reasoned the two organs may share similar requirements for their survival ex vivo. A shared characteristic between Malpighian tubules and the adult midgut is the presence of a population of progenitor cells marked by esg expression (Singh et al., 2007; Wang and Spradling, 2020). Using the esg^TS driver system, GFP expression could be induced in tubules cultured at 18 °C for 24 hr and then shifted to the permissive temperature (29 °C), indicating their long-term survival (Figure 8A–B). Moreover, we found that Malpighian tubules cultured for 3 days could still contract regularly (Video 2), albeit only if still attached to intestines.

A key component of our culture system that prolongs the viability of midguts is the co-culture with dissected abdomens and ovaries. On closer inspection, we found that adult hearts (the dorsal vessel), which lay along the abdominal cuticle, could be seen still beating regularly after up to 10 days in culture (Video 3). Similarly, the muscle sheet that envelopes the ovaries still contracted after 3 days in culture (Video 4). We then took a closer look at ovaries by dissecting individual ovarioles for live-imaging. A distinctive characteristic of stage 1–8 follicles is their rotation within their follicle cell sheath, along their long axis (Cetera et al., 2014). In our ex vivo cultures, we routinely observed rotating stage 4 follicles that grew in size and started to elongate (Figure 8C–C’ and Figure 8—video 1), indicative of progression to stage 5 (Cetera et al., 2014). Notably, this phenomenon continued for up to 48 hr, suggesting the long-term survival of follicles in our explants.

Therefore, we believe our culture protocol could be applied to other adult Drosophila tissues and will be useful in investigating a wide range of biologically relevant phenomena.

Flies were raised on standard cornmeal and molasses fly food. Prior to dissection, flies were flipped to fresh vials without live yeast daily for 3 days at 18 °C to reduce the load of commensal bacteria. On the morning of the dissection, flies were fed a sucrose 0.05% aqueous solution on a cotton pad at room temperature (25 °C) for 4–6 hr to clear most luminal contents. This also helped to reduce accumulation of food in the posterior section of the gut, which could lead to the mechanical stress of the epithelium. For SDS feeding experiments, flies were fed overnight either standard food or food mixed with SDS to a final concentration of 0.2% v/w. Both foods were also mixed with a blue food-safe dye to control for feeding. The following morning, flies were fed a sucrose 0.05% aqueous solution on a cotton pad to clear most of the luminal SDS.

Schneider’s medium (Genesee Scientific, 25–515) was modified by adding the following reagents to the stated final concentrations: 1 mM trisodium citrate dihydrate (ThermoFisher Scientific, BP327), 91.2 mM sodium chloride (Sigma-Aldrich, S9888), 55.8 mM D-trehalose anhydrous (Sigma-Aldrich, T0167), 10 mM glutamine (Gibco, 25030), and 2 mM N-acetyl cysteine (Sigma-Aldrich, A7250) (Table 1). Glutamine needs to be added only if the batch of Schneider’s medium used is glutamine-free. Medium was then filtered using 0.22 µm syringe filters (VWR, 28145) and stored at 4 °C. See Supplementary file 1 for recipe. This medium was used without additives during dissection, agarose gel stock preparation, and as a base for fly extract.

For fly extract, well-fed female flies were anesthetized on ice. Using mortar and pestle, flies were homogenized on ice in the presence of 10µλ per mg of flies of modified Schneider’s medium (as described above) with bovine serum albumin (BSA) added to 1%. The homogenate was centrifuged at 0.6 G and 4 °C for 10'. Supernatant was saved and fly carcasses discarded. The centrifugation step was repeated 3 times until all solid fly residues were eliminated. Extract was heat inactivated by heating at 60 °C for 5’, then centrifuged at 0.6 G and 4 °C for 10'. Supernatant was saved and filtered using 0.22 µm syringe filters. Extract was aliquoted and stored at –20 °C before use.

For live-imaging, 100% fly extract prepared in modified Schneider’s medium was used as a base for the complete culture medium. Fly extract was slowly thawed at 4 °C, then 10% fetal calf serum (Gibco, 26140079), 1:100 Antibiotic-Antimycotic (ThermoFisher Scientific, 152400062), 100 µg/ml Ampicillin (Fisher Scientific, AC611770250), and 25 µg/ml Chloramphenicol (Fischer Scientific, BP904-100) were added. To suppress peristaltic movements, 10 µg/ml isradipine (SigmaAldrich, I6658) was added to the complete medium immediately before imaging.

For imaging of midgut, hearts, and Malpighian tubules we used an inverted widefield Nikon Ti Eclipse microscope equipped with an incubation chamber (Okolab) for moisture and temperature control, a CoolSnap HQ2 camera (Photometrics), a SOLA LED light engine (Lumencore), and a Prior motorized stage. To limit phototoxicity, albeit at the cost of reduced signal/noise ratio, we used a 4 x neutral density filter, limited light intensity to 5%, and limited exposure times to 100–150ms and 200–300ms for GFP and RFP signals, respectively. Follicles were imaged using a Leica SP8 confocal microscope equipped with a white laser light source and an incubation chamber (Okolab). To limit dwell time, samples were imaged at a frequency of 700 Hz. Intestines could also be successfully imaged with the same setup. For both widefield and confocal setups, videos were captured at room temperature (25 °C) or at 29 °C for temperature-sensitive gene induction experiments using a 20 X air objective (APO 20 X NA 0.75 WD 1). Multipoint acquisition was used to image the R4-5 posterior midgut section of 12 intestines during each imaging section. For each midgut, a 45 µm Z-stack was captured using a 3 µm Z-step. Focus was manually checked and corrected during the course of the imaging session with the Nikon Ti Eclipse, or using a contrast-based autofocus routine for the Leica SP8. Frame rate was typically one full multi-channel Z-stack/midgut/hour. For the imaging of Malpighian tubules and ovaries contractions and beating hearts, the frame rate was 1 slice every 0.08ms instead.

For cell survival experiments, guts from W¹¹¹⁸ flies were cultured in media containing NucGreen (ThermoFisher Scientific, R37109) diluted 1:20. Midguts were imaged once a day for 3 days, using a 10 X air objective (Plan Fluor 10 X NA 0.3 WD 16) to capture a stitched z-stack of the whole organ with z-steps of 20 µm. Images were normalized to their median pixel value to account for changes in background, then midgut areas were manually selected from maximum intensity projections. The NucGreen signal was computed as the sum of pixel values in the selected areas, normalized to the first time-point in the series.

Midguts were explanted from 5 to 15 days old flies reared at 18 °C and cultured at 29 °C for 24 hr. Midguts were then carefully removed from the culture setup and fixed in para-formaldehyde 6% in PBS for 30 min at room temperature. Similarly, in vivo control flies were shifted to 29 °C for 24 hr, then dissected and their midguts fixed. Fixed intestines were stained for DAPI 0.1 mg/ml (Sigma-Aldrich, 10236276001) in PBS with 0.1% Triton-X100, then mounted with VECTASHIELD antifade mounting medium (Vector Laboratories, H-1000–10). The posterior region of the intestines was imaged and the GFP and DAPI signals were thresholded using Otsu’s method (Otsu, 1979). Regions of overlap between the two tresholded channels were used as a mask to calculate the mean GFP expression of each cell of interest in the imaged field. For each intestine, the values of each measured cell were averaged to express a mean fluorescence for the whole intestine.

For GFP induction tests, malpighian tubules were detached from midguts using micro-scissors and cultured in sandwiched agarose structures as described above. Tubules were cultured at 18 °C for 24 hr then at 29 °C. Images were captured daily. Individual GFP⁺ cells from 48 hr images were identified in previous time-points and their GFP intensity measured.

Image analysis was performed using either ImageJ or Python (v3.7.10). Using a custom Python script (Source code 1), each time-lapse movie was divided at the first time point into overlapping regions of interest that were then used for automatic local registration. This helps account for local movements and deformations of the intestinal epithelium which complicate the following image analysis step. Registration was based on cross-correlation of a region of interest with the next frame of the time-series. Individual ISC and their progeny were then manually tracked using a custom ImageJ macro (Source code 2), selecting each cell by drawing their outline at their most in-focus z slice. By identifying all cells between successive time-points, we avoided lineage assignment errors even with long time-intervals between frames.

Cells were then analyzed as follows: