The Drosophila midgut primordium forms from the endoderm during gastrulation (Campos-Ortega and Hartenstein, 1985). Under the coordinated action of the GATA transcription factor Serpent and the winged-helix transcription factor Forkhead, the posterior midgut primordium (PMG) together with the ectodermally-derived hindgut primordium are internalised into the embryo (Weigel et al., 1989; Reuter, 1994; Nakagoshi, 2005). The PMG cells initially have the same apical-basal polarity as all ectodermal cells, which is established during the process of cellularisation (Tepass and Hartenstein, 1994b). Stranded-at-second (Sas), and the canonical apical polarity factors, Crb and Sdt, localise to the apical surface and Baz and ECad are localised to the apical AJ (Table 1) (Figure 2A) (Campbell et al., 2011). During stage 10 of embryogenesis, Serpent induces the PMG to undergo an EMT and become migratory by repressing the expression of Crb, Sdt, Sas, and pleated SJ genes (Tepass and Hartenstein, 1994b; Campbell et al., 2011). As a result, the apical AJs dissolve and ECad and Baz relocalise from the AJs to dynamic puncta at cell-cell contacts, which are presumably scattered spot AJs. At stage 11, the PMG has established contact with the visceral muscle primordium and uses it as a substrate for its migration (Tepass and Hartenstein, 1994a). Three different cell types can be distinguished transcriptionally and morphologically among the migrating midgut mass. Most cells are principal midgut epithelial cells (PMECs), which will give rise to the larval midgut ECs and always contact the muscle primordium. The other two populations of mesenchymal cells, interstitial cell precursors (ICPs) and AMPs, are attached to the apical surface of PMECs and are carried along by the latter. ICPs express Inscuteable and Asense from late stage 10 to mid-stage 11, and AMPs, which will give rise to the future adult midgut, are Asense-positive from early stage 11 to late stage 12 (Figure 2B) (Tepass and Hartenstein, 1994a, 1995; Campbell and Casanova, 2015). ICPs and AMPs delaminate sequentially from the outer layer of PMECs between stage 10–11 during their posterior migration (Tepass and Hartenstein, 1995). Some AMPs at this stage can also be marked with anti-Pros antibody staining, suggesting that they may be progenitor cells for future larval ee cells (pLee), although there is no lineage tracing data to support this. Like AMPs, pLees remain in the mesenchymal inner mass during migration, but become esg- and segregate from the AMPs by stage 14 (Jiang and Edgar, 2009; Takashima et al., 2011a). The cohesive and ordered migration of these three/four types of cells along the visceral mesoderm is coordinated through ECad-mediated cell adhesion and relies on the Integrin/Laminin and Frazzled/Netrin signalling pathways (Martin-Bermudo et al., 1999; Devenport and Brown, 2004; Campbell and Casanova, 2015; Pert et al., 2015; Pitsidianaki et al., 2021). Between late stage 11 and stage 12, shortly before and during germ band retraction, the PMECs reorganize and go through MET to form the midgut epithelium. By the end of germ band retraction at stage 13, the anterior and posterior midgut rudiments approach each other and finally fuse, the PMECs assume a columnar shape and the ICPs form two clusters in the middle of the developing midgut (Tepass and Hartenstein, 1994a). MET coincides with the downregulation of Fkh and Srp (Weigel et al., 1989; Campbell et al., 2011). However, Srp down-regulation is not sufficient to trigger MET, which instead depends on basal cues from Laminin and Netrins produced by the visceral mesoderm acting through Integrins and Fra respectively (Pert et al., 2015; Pitsidianaki et al., 2021) (Figure 2B and discussed later).

During endoderm migration, the ECM between the endoderm and the mesoderm is not yet fully organised, since early electron microscopy studies demonstrated that PMEC migration is mediated through direct mesoderm/endoderm contact without any detectable ECM or junctional specialisations (Tepass and Hartenstein, 1994a). However, Srp activates LanB1 and LanB2 RNA expression in stage 11 midgut primordium cells (Wolfstetter and Holz, 2012; Töpfer et al., 2019). Moreover, the laminin matrix secreted by the visceral muscle primordium contains Wb, which is thought to induce MET, whereas that secreted by endodermal cells contain LanA, and both LanA and Wb play crucial roles in controlling the speed of migration (Urbano et al., 2009; Wolfstetter and Holz, 2012; Pitsidianaki et al., 2021). At this stage, haemocytes (migrating macrophages) are the only source of secreted type IV collagen and Perlecan (Matsubayashi et al., 2017) and they do not reach the endoderm until after the migration is complete (Urbano et al., 2011; Pitsidianaki et al., 2021). Nidogen is reported to have similar expression pattern to LanB1 during embryogenesis but is not required for endoderm migration or formation (Urbano et al., 2009; Dai et al., 2018; Töpfer and Holz, 2020). At stage 16, Laminins, Collagens, Nidogen, and Perlecan, as well as other mature ECM components, such as MDP-1 and SPARC are all found in between the endoderm and mesoderm, forming a more complex ECM network (Wolfstetter and Holz, 2012).

Cells rely on ECM receptors to receive migratory/adhesive cues from the ECM, including Integrins, Fra, Dystroglycan (Dg), the Glycipans Dally and Dally-like and Syndecan (Sdc) (Table 1). Integrins function as heterodimers of α and β subunits and are required for both midgut migration and visceral muscle formation (Devenport and Brown, 2004). Flies have five α integrin subunits, αPS1-5 and two β subunits, Mys and βν (Table 1). Embryonic midgut migration requires the expression of both αPS1 in the endoderm and αPS2 in the visceral muscle, while αPS3 cooperates with αPS1 in the endoderm layer but is not required (Brown, 1994; Stark et al., 1997; Martin-Bermudo and Brown, 1999; Martin-Bermudo et al., 1999). Phylogenetic studies show that the αPS3-5 subunits are closely related and the result of gene duplication events (Hughes, 2001). αPS3 and αPS4 are expressed in adult midgut ECs, whereas αPS5 is not (Lin et al., 2013; Patel et al., 2015). Mys is widely expressed and is essential for viability, whereas βν is specifically expressed in the developing endoderm and the larval and adult midgut, but is not required for viability or fertility (Yee and Hynes, 1993). Integrins must form heterodimers in the endoplasmic reticulum to be trafficked to the cell surface and flies without both β subunits have no integrin function at all (Leptin et al., 1989; Devenport and Brown, 2004). Both αPS1/Mys and αPS3/βν pairs of integrins can be found in the migrating endoderm at late stage 11, with αPS1/Mys localising to basal side and αPS3/βν localising mainly apically at the end of migration (Figure 2B) (Devenport and Brown, 2004; Pitsidianaki et al., 2021). Two of the three Dystroglycan splicing isoforms are expressed in the midgut at stage 16, but their functions have not yet been characterised (Schneider and Baumgartner, 2008). Fra localises to the basal side of the PMECs at stage 12 and to the basal and junctional domain of the migrating midgut cells at stage 13 (Figures 2B,C). Interestingly, AMPs, which normally remain apical to the migrating PMECs at stage 12, are mis-localised and contact the visceral muscle in netrin mutant embryos. This phenotype has been attributed to the dis-organisation and loose adhesion of the PMG epithelium, rather than loss of direct signalling to AMPs, (Pert et al., 2015).

The PMECs are the first cell-type in the midgut primordia to go through MET, with AMPs, pLees, and ICPs remaining mesenchymal in the apical lumen until later. Although the exact time at which AMPs invade and translocate across the epithelium is not defined, they are located at the basal side of the gut in newly hatched larvae while the pLees have polarised and integrated into the epithelium (Hartenstein and Nung Jan 1992; Micchelli, 2012). This raises the question of how AMPs translocate to the basal side of the epithelium, since the apical junctions between the PMECs, which are marked by ECad, start to develop during migration in the outermost trailing region of the posterior midgut and sSJs start to develop in midgut from stage 15. Moreover, it is not clear whether the pLees become polarised and integrate into the epithelium during translocation or repolarise/integrate after translocating to the basal side (Takashima et al., 2011a). It has been hypothesized that the early delamination and late segregation and translocation of the AMPs and ICPs are due to differences in cell-cell affinity (Tepass and Hartenstein, 1995). However, there are no defects in the apical location of AMPs in Ecad/shg mutant embryos, it is therefore unclear whether their delamination and translocation is a passive cell-sorting event or an active migration process (Tepass and Hartenstein, 1994a). Furthermore, it will be important to determine the relationship between cell fate determination and the corresponding EMT-MET processes.

By stage 15, the visceral mesoderm (VM) expands ventrally and dorsally to form the circular muscle fibres and the endodermal layer follows this movement to form a closed chamber. Although the early specification of the endoderm into distinct PMEC, ICP, pLee and AMP cell types does not depend on interaction with the mesoderm, VM induces the further specification and development of future larval midgut epithelium after the midgut rudiments fuse, including the formation of the three midgut constrictions during stages 14–16 and the specification of the middle midgut region and proventriculus (Nakagoshi, 2005).

Between stage 16 and 17, the future larval ECs change their morphology from short cuboidal cells to tall columnar cells and develop elaborate cellular junctions and an apical brush border (Morgan et al., 1995). Smooth SJ components start to express during stage 12 and become localised at stage 16, but mature sSJs only become visible at late stage 17 (Tepass and Hartenstein, 1994a; Izumi et al., 2012). Myo61F relocates from the basal-lateral region to the apical microvilli, coincident with the disappearance of the yolk mass which indicates the start of digestive function (Morgan et al., 1995).

Drosophila embryonic midgut formation takes less than 9 h, between stage 10 when PMG starts EMT and stage 15 when the midgut migration finishes. Both EMT and MET happen gradually, whereas cell polarity changes dramatically and rapidly during this process. The apical polarity factor Crb disappears early on, the original apical-lateral junctions dissolve, giving rise to a group of mesenchymal migratory cells connected by limited spot AJs. These cells later re-polarise forming smooth SJs rather than pleated SJs at the apical/lateral side of the cell-cell junctions. Embryonic midgut development also demonstrates the importance of the sustained basal signalling from the mesoderm much like the polarisation of the EBs in the adult midgut epithelium. It is still unclear, however, what lies downstream of basal ECM and their receptors to induce epithelial polarisation, and whether apical extracellular LanA and apically localised αPS3/βν integrins plays any role in polarising the embryonic midgut epithelium. Past work has focused on the morphological development of the midgut and the genetic control of endoderm formation and differentiation (Bilder and Scott, 1995; Harbecke and Lengyel, 1995). Much less is known about the genetic control of AMP and ICP delamination and translocation, which also involves the loss and gain of cell polarity. These processes are challenging to study, however, because they occur over short time periods in the centre of the embryo.

The larval midgut is composed of anterior, middle and posterior regions, each maintaining a different pH, and is anatomically similar to the adult midgut, although the constriction around the middle midgut is less obvious (Shanbhag and Tripathi, 2005; Overend et al., 2016). The larval midgut contains four gastric caeca, which are blind sacs that emerge from the anterior midgut just posterior to the proventriculus. They persist in the larva but are lost during pupation and are not present in the adult fly (Skaer, 1993). Larval ECs are polyploid and derive from PMECs, whereas larval ee (lee) cells are diploid and derive from pLees (Takashima et al., 2011a). Cell specification has been well-studied in the larval middle midgut (Hoppler and Bienz, 1994, 1995). Large cells in this region were first called calycocytes, and were later named cuprophilic or copper cells, since they accumulate copper and display orange fluorescence when the larvae are fed with copper-enriched food. This property is attributed to the binding of copper ions to metallothionein, which is constitutively expressed in the cytoplasm in the middle midgut region (Skaer, 1993; Durliat et al., 1995; McNulty et al., 2001). It has been proposed that copper cells derive from the ICPs, although this is at odds with the observation that the ICPs disseminate over the whole embryonic midgut after stage 15 (Poulson and Waterhouse, 1960; Skaer, 1993). The copper cells are cup-shaped, with an invaginated apical domain containing long microvilli. They are surrounded by columnar interstitial cells with a normal apical domain, short microvilli and a more extensive basal labyrinth (Filshie et al., 1971). It is thought that the copper cells are the acid secreting cells, based on the correlation between the number of residual copper cells in labial mutant larvae and the number of remaining acid-retaining cells (Hoppler and Bienz, 1994; Dubreuil et al., 1998; Dubreuil, 2004). Several V-ATPase and other ion transporters are required for the acidic pH generation (Overend et al., 2016; Tian et al., 2022). Interestingly, copper absorption from the food can inhibit acid secretion and the acid secretion defective α-spec mutant copper cells are not able to accumulate copper, which raises the question of how copper absorption and acid secretion are linked (Dubreuil et al., 1998; McNulty et al., 2001). Furthermore, we still do not know how and why the apical domain in the copper cells invaginates nor how the interdigitated arrangement of copper cells and interstitial cells arises. One clue comes from the stage 15 embryonic midgut, when the inner ICPs interdigitate between the outer labial-positive ICPs (Skaer, 1993), which means that the arrangement of copper cells and interstitial cells are probably also under the control of labial.

The larval midgut is remarkably similar to the adult midgut at the level of cellular structure, with an apical brush border facing the gut lumen, a basal side in contact with the visceral muscle and a basal labyrinth of invaginations from the basal membrane (Figure 3) (Shanbhag and Tripathi, 2005, 2009). The larval midgut also forms sSJs (Izumi et al., 2012, 2016, 2021) and the apical domain is enriched for actin and β_H-spectrin/α-spectrin, while β-spectrin/α-spectrin heterotetramers label the basolateral domain (Dubreuil et al., 1998). Spectrins are not required for copper cell polarity, but loss of β_H-spectrin leads to loss of the apical proton pump, the H⁺V-ATPase which probably causes the defect in acid secretion seen in α-spec mutant larvae (Phillips and Thomas, 2006). The two class I myosin family proteins, Myo31DF and Myo61DF can also be found in the apical terminal web and brush border microvilli in the larval ECs, but neither is required for cell polarity or brush border organisation (Morgan et al., 1995; Okumura et al., 2015). Interestingly, the AJs marked by ECad and Baz localise apical to the sSJ before the embryo hatches, whereas, AJs localise to the basal side of the sSJ in the first instar larva and adult ECs (Tepass and Hartenstein, 1994b; Chen et al., 2018). It is not clear how and when apical AJs disappear and basal AJs form during larval midgut formation.

One important feature of the larval midgut is the presence of AMPs. They first appear as single cells residing at the basal side of the larval midgut epithelium (L1, Figure 3A). They divide 7-10 times during the larval midgut development. The daughter cells of the first three divisions migrate and spread along the basal surface of the epithelium. The AMPs continue to divide during the third instar stage, but the daughter cells stay attached to each other to form AMP nests, which contain 8–30 cells by the onset of metamorphosis (Figure 3B) (Mathur et al., 2010; Takashima et al., 2011a; Jiang et al., 2011). 1-3 of the cells in an AMP nest differentiate into STAT92E > GFP (JAK-STAT pathway reporter) and Su(H)GBE > GFP/lacZ (Notch signalling reporter)-positive peripheral cells, which elongate and surround the inner mass of small, round central cells. Peripheral cells are post-mitotic with bigger nuclei than the central cells. They function as a niche to maintain the stem-cell state of the central AMP cells until metamorphosis (Mathur et al., 2010). However, the central cells can differentiate and become partially regenerative when the larval midgut is challenged with infection (Houtz et al., 2019). Some AMP cells also differentiate into Pros + cells and become an integral part of the future transient pupal midgut (Takashima et al., 2011b). The AMP nests can reach 2/3 of the height of the larval ECs, but do not reach the apical lumen, presumably because they cannot pass the sSJs between the larval ECs (Figure 3B). Both peripheral cells and central cells appear to maintain the contact with ECM and the peripheral cells contact the larval ECs. It is not known how the peripheral cells adopt a sheath-like shape and encase the central cells, nor how they provide a niche for the central AMP cells, except that the Dpp signalling is required (Mathur et al., 2010).

Shortly after puparium formation (APF), the larval midgut shortens, bringing the scattered AMP nests together. The outer peripheral cells contact each other first and by 6h APF become squamous and join together to form a multi-layered sheet called the transient pupal midgut (tPMG). Central cells also change their shape, flattening longitudinally and expanding laterally, to form a continuous layer of presumptive adult midgut (AMG) in a process that is thought to be MET (Takashima et al., 2011b). At this stage, the larval midgut, tPMG and AMG are all connected via spot AJs. At 8 h APF, the AMG starts to show polarised features, with aPKC localising apically, Fas3 at the apical and lateral domains and Arm along the lateral and basal domains. Precursors of the adult ISCs, the presumptive intestinal stem cells (pISCs), remain at the basal side of the AMG layer (Figure 4A). At the same time, the tPMG also differentiates to a certain degree, forming microvilli and containing ee cells that have a spindle shape and remain detectable until 24 h APF (Figure 4A, 3B) (Takashima et al., 2011b; Takashima et al., 2016b). At 6 h APF, the ECM layer surrounding the midgut and visceral muscle starts to break down and disappears by 24 h APF. By 36 h APF, myofibrils disappear since the visceral muscle fibres surrounding midgut de-differentiate into secondary myoblasts (Aghajanian et al., 2016). During this time, both the tPMG and the AMG keep differentiating. The tPMG develops pleated SJs, whereas the AMG develops apical microvilli and smooth SJs at the apical side of the lateral domain. The AMG is in direct contact with the myoblasts since no ECM is observed in between (Aghajanian et al., 2016). Interestingly, an electron dense liquid has been observed separating the larval and tPMG from the AMG at this stage (Figure 4B) (Takashima et al., 2011b). Both the myofibrils and ECM reorganise and reappear by 48 h APF (Aghajanian et al., 2016). Between 48 and 72 h APF, some esg-positive pISCs express Pros and divide asymmetrically to give rise to the adult ee cells (Guo and Ohlstein, 2015). The re-emergence of ECM is thought to be important for the pISC division and specification at this stage (Aghajanian et al., 2016). During later stages of metamorphosis, the larval and transient pupal midguts remain closely associated and further contract and become the “yellow body” in the lumen of the developing adult midgut. They are eventually discharged from the intestinal tract after eclosion.

The separation between larval midgut/tPMG and the AMG is essentially the delamination of larval midgut epithelial cells and the detachment between peripheral cells and central cells of the AMP nests. This results in the reorganisation of the tissue into three layers with spot AJ still present among them (Takashima et al., 2011b). The reorganisation happens within the first 12 h during pupal development, while the visceral muscle and ECM are still present. Both the tPMG and AMG keep differentiating, but only the AMG remains attached to the ECM, which means the separation cannot be simply explained by apoptosis-induced cell extrusion. It would be interesting to find out whether basal integrin adhesion is weakened in the larval epithelium and tPMG but retained in the AMG. Many other questions still remain about the adult midgut formation during pupal metamorphosis. First of all, before metamorphosis begins, there is direct signalling between the peripheral cells and central cells in the seemingly compact AMP nests, but almost nothing is known about the molecules that mediate adhesion between them or the molecular mechanisms that control the separation and reorganisation of the tPMG and AMG. Secondly, although the tPMG loses contact with the ECM and muscle layer, it still manages to differentiate to form pleated SJ. The functional significance of this junction and how it is formed are unclear. Thirdly, the AMG cells are believed to go through MET as they polarise, while the pISCs remain basal and in contact with the re-formed ECM and muscle layer. Based on what we know about the formation of the embryonic and adult midgut, it will be interesting to determine whether pISC specification requires a similar translocation process to AMPs in the embryo and if the AMG cells polarise in the same way as adult EBs and form a PAC as they integrate into the epithelium, and if their polarisation requires basal integrin signalling and SJ components.