The early endosomal protein Rab21 is critical for enterocyte functions and intestinal homeostasis

Membrane trafficking is defined as the vesicular transport of molecules into, out of, and throughout the cell. In intestinal enterocytes, defects in endocytic/recycling pathways result in impaired function and are linked to genetic diseases. However, how these trafficking pathways regulate intestinal tissue homeostasis is poorly understood. Using the Drosophila intestine as an in vivo model system, we investigated enterocyte-specific functions for the early endosomal trafficking machinery in gut homeostasis. We focused on the small GTPase Rab21, which regulates specific steps in early endosomal trafficking. Rab21-depleted guts showed severe abnormalities in intestinal morphology, with deregulated homeostasis associated with a gain in mitotic cells and increased cell death. Increases in both apoptosis and yorkie signaling were responsible for compensatory proliferation and tissue inflammation. Using a RNA interference screen, we identified specific regulators of autophagy and membrane trafficking that phenocopied Rab21 loss. We further showed that Rab21-induced hyperplasia was rescued by inhibition of epidermal growth factor receptor signaling, and identified improperly trafficked cargoes in Rab21-depleted enterocytes. Our data shed light on an important role for early endosomal trafficking, and particularly Rab21, in enterocyte-mediated intestinal homeostasis.


INTRODUCTION
Membrane trafficking is characterized by the vesicular transport of proteins and macromolecules throughout the various cellular compartments, as well as in and out of cells. This process requires exchanges between compartments that are essential to maintain cellular functions and help cells adapt to the ever-changing conditions in the extra-and intra-cellular environments. The intricate link between membrane trafficking and cellular signaling means that modulating trafficking has strong consequences on cellular homeostasis [1]. Mutations in membrane trafficking genes are associated with a large array of human diseases [2]. Importantly, some of these mutations affect several tissues, while others are restricted to one organ system, and can be used to shed light on the cell typespecific functions of vesicular trafficking genes.
The digestive tract performs several essential functions, including food digestion, nutrient/water absorption, and protection against external pathogens. These tasks are achieved by enterocytes, a differentiated cell population constituting approximately 80% of the intestinal epithelium. Given their absorptive function, enterocytes display high membrane trafficking flux, which must be tightly orchestrated to ensure appropriate organ activity [3]. Defects in trafficking processes affect enterocyte properties (e.g., their shape, absorptive function, and resistance to stress) and lead to intestinal diseases [3,4]. Intestinal pathologies, such as microvillus inclusion and chylomicron retention disease, are associated with mutations in the membrane trafficking related-genes myosin 5B (MYO5B) and secretion associated Ras related GTPase 1B (SAR1B), respectively [2,5,6]. MYO5B is required for proper recycling to the apical membrane [5] while SAR1B is necessary for chylomicron secretion [6].
The small Rab GTPases constitute an important class of membrane trafficking regulators [7]. Rabs localize to specific cellular compartments, where they recruit various effectors to mediate their functions. In enterocytes, a few studies have reported the involvement of Rab8 and Rab11, both components of the endosomal recycling machinery, in the regulation of apical brush border formation [3,4,8]. In Rab8-knockout mice, apical peptidases and transporters accumulate in lysosomes. This leads to decreased absorptive functions, which are associated with microvillus inclusion and defects in microvillar structure [4]. Similarly, loss of Rab11 in mouse intestinal epithelial cells results in defects in enterocyte polarity and abnormal microvilli organization [8]. Notably, the enterocyte polarity defect upon Rab11 knockdown is conserved in Drosophila [9]. In the Caenorhabditis elegans intestine, Rab11 also displays an apical distribution, which is required to establish proper polarity in intestinal epithelial cells [10]. Other studies have demonstrated roles for Rab11 in toll-like receptor regulation and microbe tolerance in mouse and Drosophila intestines [11]. Finally, recent work performed in mice and flies highlighted a conserved role for Rab11 in intestinal tumor progression, through the regulation of the Hippo pathway [9,12]. Although roles for the endosomal recycling machinery in enterocytes have been well documented, little is known about the functions of other membrane trafficking routes in these cells, and more generally, in intestinal tissue homeostasis.
Early endosomes act as sorting centers, in which endocytosed cargos must be correctly directed toward either recycling routes or lysosomal degradation [13]. Therefore, early endosomes are important for the proper regulation of internalized signals [1]. Rab5 is often associated with early endosomes, where it acts in endosomal homotypic fusion and in endocytosis [14]. Rab5 is also necessary for phosphatidylinositol-3-phosphate (PtdIns(3)P) synthesis in early endosomes [15]. In flies, Rab5 is required for the proper localization of apical proteins in polarized epithelia [16,17], while in mice, it is required for lysosome homeostasis in the liver [18]. Several other Rabs are also present in early endosomes [19], including Rab21, which is involved in endocytosis and early endosomal trafficking of integrins, epidermal growth factor receptor (EGFR), and certain clathrin-independent cargos [20][21][22][23]. Rab21 was first identified in human intestinal Caco-2 cells [24] and is expressed mainly in enterocytes in both human and mouse intestines [24,25]. Interestingly, recent studies demonstrated that Rab21 expression is decreased in enterocytes in solute carrier family 15 (oligopeptide transporter), member 1 (Slc15a1)-knockout mice [25].
These data led us to investigate the specific functions of Rab21 and other trafficking regulators in enterocytes in vivo using the Drosophila intestine as a model system.
Drosophila enables the use of powerful genetic tools, and is an excellent model to investigate intestinal biology, with intestinal signaling pathways and functions highly similar to those of humans [28]. Like the mammalian intestine, the Drosophila gut is composed of intestinal stem cells, which differentiate into progenitor cells that yield differentiated enteroendocrine cells and enterocytes [28][29][30]. The coordination of several signaling pathways orchestrates the balance between intestinal stem cell proliferation, progenitor differentiation, and differentiated cell turnover. Under normal conditions, Wnt signaling is required for intestinal stem cell self-renewal, similar to in mammals [31,32], while high levels of Notch signaling in progenitor cells promotes enterocyte differentiation [29,30]. Upon stress, the Janus kinase-signal transducer and activator of transcription (JAK-STAT), EGFR, c-Jun NH(2)-terminal kinase (JNK), and Hippo signaling pathways activate intestinal stem cell proliferation and progenitor differentiation in response to various injuries [33]. Furthermore, upon intestinal epithelial stress, enterocytes serve as stress sensors to initiate tissue regeneration, and rapidly compensate for cell loss by promoting the non-cell autonomous activation of intestinal stem cells. The JNK and Hippo/yorkie (yki) pathways are involved in this process in enterocytes, where they trigger expression of the proinflammatory cytokine unpaired 3 (upd3) [34][35][36][37][38][39]. Secretion of upd3 promotes intestinal stem cell proliferation by directly activating JAK-STAT and indirectly the EGFR pathway [34,39,40].
Here, we demonstrate that functional early endosomes are required to maintain proper tissue homeostasis in fly intestines. Loss of Rab21 in Drosophila enterocytes results in yki activation and apoptosis, which both induce secretion of the pro-inflammatory cytokine upd3, leading to compensatory proliferation in a non-cell autonomous manner. An enterocyte focused screen of membrane trafficking genes reveals that blocking autophagy, endosomal trafficking, and EGFR-mitogen-activated protein kinase (MAPK) signaling also induces tissue inflammation and compensatory proliferation. Using epistasis experiments, we show that Rab21 independently regulates enterocyte EGFR signaling and autophagy. Additionally, we highlight a previously unappreciated contribution of Rab21 in PtdIns(3)P and PtdIns(3,5)P2 regulation. Finally, using a tandem mass tag (TMT)-based quantitative proteomic approach we shed light on deregulation of SLC transporters, intestinal proteases and cytochrome P450 proteins in enterocytes depleted for Rab21.

Enterocyte Rab21 is required to maintain intestinal epithelium morphology
Fluorescence-activated cell-sorted and single-cell RNA-sequencing databases indicate that the early endosome components Rab5 and Rab21 and the recycling endosome component Rab11 are expressed in various cell types throughout the gut. Rab21 is normally expressed at lower levels than Rab5; however, it is upregulated upon infection [41][42][43]. As the role of Rab11 in gut homeostasis is well-described [3], we decided to focus our initial analysis on Rab21, which plays roles in specific early endosomal trafficking events [20][21][22]44,45] and is modulated by stress in both flies and in mouse models of IBD [25,41]. We first determined the precise Rab21 expression pattern in the gut using a GAL4 driver line with GAL4 controlled by the Rab21 promoter [46] and a UASt-green fluorescent protein (GFP):Rab21 responder line. In accordance with transcriptomic data [41][42][43], we observed that Rab21 was present throughout the digestive tract ( Figure 1A Interestingly, Rab21 displayed stronger expression in the distal part of the posterior midgut (R5 region) [41] and in the copper cell region ( Figure 1B). In the R5 region, Rab21 expression was enriched in large polyploid cells with low levels of armadillo (arm) staining, characteristic of enterocytes ( Figure 1C-D, arrowheads), with no restricted basal or apical localization (Suppl. Fig. 1A). Using in situ hybridization, we confirmed that Rab21 was expressed in enterocytes in the posterior midgut, which were visualized with 4′,6-diamidino-2-phenylindole (DAPI) to highlight their polyploid nuclei (Suppl. Fig. 1B).
We also detected Rab21 in cells displaying high arm staining, which were likely intestinal stem cells ( Figure 1C-D, arrows). Immunofluorescence against Delta (Dl) confirmed that Rab21 was present in intestinal stem cells (Dl + ; Suppl. Fig. 1C). It was also detected in prospero (pros) + cells (Suppl. Fig. 1C), revealing that enteroendocrine cells also express Rab21. As previously observed in human and mouse intestines, these data show that Rab21 is expressed throughout the fly gut, with high expression in enterocytes, suggesting potential important functions in these cells.
Flies were housed at 18°C during embryonic development and at 29°C for 10 days after hatching, then analyzed ( Figure 1E). Loss of Rab21 function was associated with a significant increase in cell density (Suppl. Fig. 2A), indicating defects in tissue morphology. Consistent with this notion, compared to controls ( Figure 1F; Suppl. Fig. 2B), the organization of the cell junction markers coracle (cora; Figure 1F) and arm (Suppl. Fig.   2B) was disrupted with decreased Rab21 expression. Contrary to Rab11 defects [48], we did not observe noticeable changes in the microvillar structure or polarity of Rab21-depleted enterocytes (Suppl. Fig. 2C), and Smurf assays revealed no defects in tissue permeability in Rab21-depleted guts (Suppl. Fig. 2D). Furthermore, Rab21 knockdown in enterocytes led to a multilayered epithelium ( Figure 1H; Video 2), which was never observed in control flies ( Figure 1H, Video 1). Consistent with this observation, the intestinal lumen area tended to be reduced in flies depleted of Rab21 ( Figure 1G).  Fig. 3A) and the cell junction marker arm was decreased (Suppl. Fig. 3B). This suggests that Rab21 is necessary for proper gut morphology, but not sufficient to drive strong phenotypes. Altogether, these results highlight the contribution of enterocyte Rab21 in the maintenance of gut morphology, as well as its requirement during aging.

Enterocyte Rab21 maintains intestinal cellular homeostasis
Enterocytes are non-mitotic, terminally differentiated cells arising from progenitors produced by intestinal stem cells ( Figure 2A). Surprisingly, while never observed in control guts ( Figure 2B), Rab21 depletion from enterocytes resulted in supernumerary cells underneath the Myo1A + cells ( Figure 2B). These cells did not express Myo1A, suggesting that they were not enterocytes. Furthermore, the percentage of Myo1A + cells was decreased in guts depleted of Rab21 compared to control guts ( Figure 2C). To identify the accumulating cells, we assessed the impact of Rab21 depletion in enterocytes on the other main midgut cell types. Upon Rab21 knockdown, we noticed increases in both the enteroendocrine cells and stem cell populations, using immunofluorescence against pros and Dl, respectively ( Figure 2D, E). The increase in the proportion of intestinal stem cells was further confirmed using a Dl:GFP reporter line ( Figure 2E). From these data, we conclude that enterocyte Rab21 is necessary for the maintenance of intestinal homeostasis.

Rab21 is necessary for proper cell turnover
Next, we investigated the signaling mechanisms impacting the various gut cell populations upon Rab21 depletion from enterocytes. In various contexts, increased cell density in the gut is compensated by increased apoptosis [37]. Hence, we first assessed cell death by immunostaining for cleaved caspase 3. In guts with Rab21-depleted enterocytes, the intensity of cleaved caspase 3 staining was significantly increased compared to controls ( Figure 3A). The increased cell death in Rab21-depleted intestines was further corroborated using SYTOX staining ( Figure 3B). Consistent with these results, the transcript level of head involution defective (hid), which promotes caspase activity, was upregulated upon Rab21 depletion (Suppl. Fig. 3C).
Since cell death in the midgut is often compensated by proliferation [34], we next investigated mitotic activity by immunostaining for phospho-histone H3 (pH3). Loss of Rab21 resulted in a significant increase in the percentage of proliferating cells compared to controls ( Figure 3C). Importantly, we were able to rescue the increased proliferation after Rab21 RNAi by overexpressing Rab21 deg in enterocytes ( Figure 3D).
Upon stress, damaged enterocytes help trigger the compensatory proliferation of intestinal stem cells by secreting the inflammatory cytokine upd3, which activates JAK-STAT signaling in the intestinal epithelium [40,49]. We thus hypothesized that depletion of enterocyte Rab21 leads to non-cell autonomous compensatory proliferation through the upd3-JAK-STAT signaling axis. Using a upd3 reporter line, we observed higher levels of the reporter upon Rab21 depletion from enterocytes ( Figure 3E). Consistent with this observation, the 10×STAT92E-GFP reporter revealed a dramatic increase in JAK-STAT signaling activity in Rab21-depleted guts ( Figure 3F). These data support the idea that increased proliferation in Rab21-depleted guts is due to a non-cell autonomous mechanism.

Non-cell autonomous proliferation in the Rab21-depleted gut is caused by cell death and activation of the transcriptional co-activator yki
To confirm that non-cell autonomous proliferation was responsible of the increased amount of pH3+ cells in guts with Rab21-depleted enterocytes, we co-depleted upd3 and Rab21 in enterocytes using inducible upd3 RNAi. Depleting enterocyte upd3 was sufficient to suppress the overproliferation phenotype observed after Rab21 knockdown ( Figure 4A).
Enterocyte-derived upd3 activates various pathways in the Drosophila adult midgut, and among these, the Hippo and JNK pathways are well documented [34,36,[50][51][52]. We performed epistasis experiments using inducible RNAi of basket (bsk) and yki to assess the requirements for the JNK and Hippo pathways, respectively, in mediating increased upd3 expression upon Rab21 depletion. Only yki inhibition was sufficient to attenuate the compensatory proliferation induced by Rab21 depletion ( Figure 4B).
As we observed an increase in apoptotic cells in guts depleted of Rab21, we also evaluated whether cell death participates to the increased mitotic activity and upd3-related inflammatory phenotypes. To do so, we blocked apoptosis by overexpressing the baculovirus protein P35 in Rab21-depleted enterocytes. In this context, proliferation ( Figure 4C) and JAK-STAT activation ( Figure 4D) were both largely diminished compared to controls, demonstrating that cell death also increased upd3 secretion to mediate JAK-STAT activation. These data show that enterocyte Rab21 is required to block yki activation and enhance enterocyte survival, which both participate in restricting intestinal stem cell proliferation in a non-cell autonomous manner ( Figure 4E).
We therefore also assessed whether the decreased PtdIns(3)P level was due to excessive generation of PtdIns(3,5)P2. Using the mCherry:ML1N2× probe, which labels PtdIns(3,5)P2, we noticed a significant decrease in PtdIns(3,5)P2 pools in Rab21-depleted guts compared to controls (Suppl. Fig. 4C). PtdIns(3,5)P2 localizes to late endosomal and lysosomal membranes; therefore, we investigated whether these compartments were affected by Rab21 depletion. We observed no significant differences in Rab7 + vesicles (Suppl. Fig. 4D), suggesting that late endosomes were not affected. In addition, we used the GFP-lysosomal-associated membrane protein (LAMP) construct [54] to assess lysosome localization upon Rab21 knockdown. As observed in other fly tissues and in mammalian cells [22,44], we did not notice any differences in the lysosomes of control and Rab21-depleted enterocytes (Suppl. Fig. 4E). These data show that Rab21 depletion perturbs specific phosphoinositide pools, potentially impacting enterocyte sorting pathways. Moreover, as membrane trafficking and cellular signaling are closely related, the effects of enterocytic Rab21 depletion on endosomal trafficking may impact signaling pathways responsible for the observed phenotypes.

Autophagy and EGFR signaling are independent in enterocytes
To gain deeper insight into the trafficking processes that regulate enterocyte functions, we performed a genetic screen to identify genes phenocopying the increased compensatory proliferation and inflammation caused by Rab21 depletion, by assaying pH3 and upd3 levels, respectively. We focused on endosomal genes linked to different membrane trafficking steps or processes ( Figure 5A), as well as genes associated with functions ascribed to Rab21: autophagy regulation (Autophagy-related 1 (Atg)1, Atg4, Vesicle- Adaptor Protein complex 2, α subunit (AP-2) had no effect (Suppl. Fig. 5A-B). Finally, loss of the Retromer subunit Vps26 ( Figure 5B-E) also phenocopied Rab21 depletion; however, depleting two other Retromer subunits (Vps29 and Vps35) did not (Suppl. Fig.   5A-B). None of the other tested genes phenocopied both the over proliferative and inflammatory phenotypes observed upon Rab21 depletion (Suppl. Fig. 5A-B). We conclude that early endosomal functions (Vps26, Rab5, Vps8, and Vps18) and autophagy (shi, Atg4, Vamp8, and Syx17) are required in enterocytes for proper gut homeostasis.
Activated Ras phenocopied Rab21 depletion ( Figure 5B-E and [55]), and interestingly, recent work has highlighted a role for autophagy in regulating EGFR signaling in Drosophila intestinal stem cells [56,57]. Therefore, we assessed if EGFR signaling was impaired upon Rab21 knockdown by immunostaining for diphosphorylated extracellular signal-regulated kinase ((dp)ERK). Guts depleted of Rab21 displayed a significant increase in dpERK compared to controls ( Figure 6A). Importantly, overexpression of a DN form of Egfr (Egfr DN ) was sufficient to rescue the non-cell autonomous proliferation of intestinal stem cells induced by Rab21 knockdown ( Figure 6B). These data demonstrate that Rab21 contributes to the regulation of enterocyte EGFR signaling. We next investigated if this contribution could be mediated by regulating autophagy. When we immunostained guts for refractory to sigma P (ref(2)p), an autophagic cargo protein, we noticed an increase in ref (2)p + dots in Rab21-depleted guts compared to controls (Suppl. Fig. 6A). This suggested a deregulation of autophagic flux, as previously observed in other cell types [44,58]. These data were further supported by transmission electron microscopy (TEM), which revealed an increased number of autophagosomes, recognizable by their double membranes, upon Rab21 depletion (Suppl. Fig. 6B, arrowheads). To validate the involvement of Rab21's autophagic function in the regulation of EGFR signaling, we assessed dpERK signal intensity after depleting autophagy-related genes. Surprisingly, no significant differences were observed ( Figure 6C; Suppl Fig. 6C), with the exception of Syx17 RNAi ( Figure 6C).
In addition, inhibition of EGFR signaling was unable to rescue the increased mitotic activity observed in guts depleted of autophagic genes ( Figure 6D). These data show that Rab21 contributes to both EGFR signaling and autophagy regulation; however, these effects are independent of each other. These results also highlight that contrary to intestinal stem cells, autophagy and EGFR signaling are not epistatic in enterocytes.

Enterocyte Rab21 knockdown affects solute carrier transporter abundance
The above data genetically identified the cellular signaling pathways affected by Rab21 depletion from enterocytes. However, they do not provide an unbiased view of the proteins affected by Rab21 knockdown. Therefore, to supplement our genetic analyses, we systematically identified proteins affected by Rab21 depletion in enterocytes. To do so, we performed a TMT-based quantitative proteomic analysis of control fly guts and those with enterocyte-specific Rab21 depletion in biological triplicate. We identified 2,691 proteins in total, of which 101 were differentially modulated more than two folds ( Figure 7A), with 57 proteins that increased and 44 that decreased ( Figure 7B) in Rab21-depleted cells compared to controls. Interestingly, many of the increased proteins belonged to the solute carrier (SLC) transporter family ( Figure 7B, highlighted with a star). Previous studies have demonstrated functional links between Rab21 and certain SLC members [22,25], supporting our data. The proteomic approach also revealed decreased abundance of several proteins in the cytochrome P450 (CYP) family ( Figure 7B, square), as well as proteins related to proteolysis ( Figure 7B, circle). Finally, three proteins related to lipid metabolism were also deregulated upon Rab21 depletion from enterocytes ( Figure 7B, hash symbol).
We then investigated Reactome pathways enriched with differentially regulated proteins ( Figure 7C). This analysis revealed the deregulation of several processes related to sugar: "lysosomal oligosaccharide catabolism" was decreased, while "cellular hexose transport", "intestinal hexose absorption", and "glucogenesis" were increased. Consistent with the increase in SLC proteins, "SLC-mediated transmembrane transport" was also enriched.
Interestingly, some of the upregulated SLC proteins are involved in glucose transport, including proteins belonging to the SLC2, SLC5 and SLC16 families. From these data, we conclude that Rab21 is likely required for the proper regulation of SLC, CYP, and lipid metabolism proteins, and maintains proper absorption of sugar and potentially other nutrients by enterocytes.

DISCUSSION
The importance of the membrane trafficking machinery in cell homeostasis is well established [2]; however, questions remain regarding the cell-specific functions of the majority of its components. In enterocytes, previous studies uncovered a role for endosomal recycling in coordinating apical-basal axis formation [4,8,9] and brush border formation [4,8]. Here, we found that the early endosomal protein Rab21 is required in enterocytes to maintain tissue homeostasis through the regulation of multiple signaling events crucial for their survival and function.
We showed that Rab21 knockdown in enterocytes leads to the formation of a multilayered epithelium with aberrant localization of the adherens and septate junction markers arm and cora. Intriguingly, while flies depleted of enterocyte Rab21 have shorter lifespans, epithelium integrity is conserved, indicating that the reduced lifespan is not due to gut leakiness. Similarly, flies depleted of intestinal septate and adherens junction components (Tetraspanin 2A and E-Cadherin, respectively) do not show tissue permeability defects [59,60]. Therefore, cora and arm mislocalization in Rab21-depleted guts may reflect defects in tissue homeostasis and related signaling events rather than impairment of junction formation and function.
Our data clearly illustrate that cellular homeostasis is dramatically perturbed upon Rab21 depletion from enterocytes. Indeed, the enterocyte population diminished while the intestinal stem cell and enteroendocrine cell populations expanded. The results indicate that the reduction in enterocytes is due to induction of apoptosis, and that compensatory proliferation is responsible for the rise in intestinal stem cells. However, it is unclear why we observed expansion of the enteroendocrine cell population. Previous studies highlighted a role for the JAK-STAT pathway in enteroendocrine cell specification [61,62].
A high level of JAK-STAT signaling is required for enteroendocrine fate specification, demonstrated by the fact that the intestinal epithelia of hypomorphic Stat92 E06346 mutant flies are composed mainly of enterocytes [61]. Interestingly, we showed that guts depleted of Rab21 display a massive increase in JAK-STAT signaling activity. Therefore, it is conceivable that the increase in enteroendocrine cells is caused by hyperactivity of the JAK-STAT pathway.
In addition, we observed the induction of inflammation upon Rab21 depletion. This inflammation was mediated by upd3 secretion and led to compensatory proliferation.
Enterocytes secrete inflammatory cytokines in response to a large range of stresses (e.g., apoptosis, infection, reactive oxygen species, JNK activation, injury) [34][35][36]38,39,63,64], with upd3 being the most abundant [34]. Secretion of upd3 triggers compensatory proliferation by activating JAK-STAT signaling in intestinal stem cells and visceral muscle cells [34,39,61]. Visceral muscle cells, in turn, secrete the Egfr ligand vein, which promotes intestinal stem cell proliferation through Egfr [39,65]. Various signaling pathways act as stress sensors in the intestinal epithelium, including the Hippo-yki, JNK, and p38 pathways [34][35][36]38,63]. In enterocytes, activation of any one of these pathways stimulates upd3 secretion [34][35][36]38,63]. Interestingly, constitutive activation of the JNK pathway induces cell death [34]. However, enterocyte apoptosis is not responsible for JNK-mediated compensatory proliferation, since JNK-induced compensatory proliferation cannot be rescued by P35 expression [34]. In addition to upd3 secretion, stressed enterocytes participate in promoting intestinal stem cell proliferation through the transcriptional activation of the Egfr ligand maturation factor rhomboid, which promotes the secretion of mitogenic Egfr ligands [66]. Consistent with these data, in Rab21-depleted guts, we were not able to rescue increased proliferation by inhibiting JNK signaling, while blocking apoptosis was sufficient to significantly reduce compensatory proliferation. Furthermore, we showed that yki was required for the compensatory proliferation of intestinal stem cells after Rab21 depletion. The Hippo-yki pathway plays well-characterized roles in promoting cell growth/proliferation and inhibiting cell death [67]; therefore, the relationship between yki signaling and apoptotic signaling in triggering compensatory proliferation in Rab21depleted guts remains unclear. It is likely that these pathways act synergistically to mediate the non-cell autonomous proliferation of intestinal stem cells upon Rab21 knockdown, and may not be individually sufficient for this purpose. In accordance with this hypothesis, recent work showed that in unchallenged conditions, yki depletion from enterocytes tended to diminish enterocyte apoptosis, while hippo inhibition tended to increase it [37]. Such data suggest context-dependent requirements for yki functions, with the Hippo-yki pathway involved in enterocyte apoptosis in normal conditions. The compensatory proliferation and inflammation phenotypes associated with Rab21 depletion from enterocytes are similar to those observed in Rab11-depleted guts [9,12].
However, these phenotypes are related to distinct signaling dysfunctions. We showed that Rab21, in enterocytes, negatively regulates EGFR signaling, while Rab11 functions independently of this pathway [9,12]. Overexpression of Egfr DN in Rab21-depleted guts suppressed compensatory proliferation. Consistent with these findings, in HeLa and HEK293T cells, RAB21 is required for proper EGFR internalization and degradation [45], although the specific molecular mechanism by which it regulates EGFR remains to be defined. Our genetic screen revealed that depletion of autophagy-and endosome-linked genes from enterocytes phenocopied knockdown of Rab21. Furthermore, constitutive activation of EGFR-MAPK signaling led to similar results, as previously observed [55].
Recently, an autophagy-endocytosis network was linked to the negative regulation of EGFR signaling in intestinal stem cells [56,57]. However, EGFR pathway regulation appears to be different in enterocytes, as inhibiting EGFR signaling did not rescue autophagy-induced increases in proliferation. Therefore, our data show that autophagy and EGFR signaling are independent in enterocytes, yet are both regulated by the early endosomal protein Rab21.
We showed that Rab21 depletion negatively affects PtdIns(3)P and PtdIns (3,5)P2 pools, as well as Hrs + early endosomes. We hypothesize that these defects result in impaired endosomal trafficking and inappropriate Egfr trafficking. Therefore, the increased Egfr activity could result from decreased Hrs endosomal recruitment or inefficient endosomal maturation. We did observe normal late endosomes and lysosomes, suggesting defects in earlier trafficking stages. This was consistent with the phenotypes of guts containing enterocytes depleted of Rab5 and CORVET complex components.
Proper regulation of EGFR signaling in early enterocytes is crucial for their morphology and maturation [65,68]. Previous studies identified that EGFR-MAPK activation in enterocytes induces their endoreplication [68] and that overexpression of Egfr triggers enterocyte delamination [65]. Interestingly, joint activation of Egfr and mir-8 stem loop in the wing disc epithelia leads to polyploid cells that induce apoptosis in neighboring cells [69]. From these observations, we hypothesize that Rab21 depletion from enterocytes might lead to EGFR-MAPK activation in these cells, which would in turn trigger the death of surrounding cells, with both effects required for the observed Rab21-related phenotypes.
Finally, to uncover the physiological functions altered by defective early endosomal trafficking, we performed a TMT-based quantitative proteomic analysis. The elevated levels of multiple SLC families are consistent with previously reported roles for Rab21 [22,23], and suggests that SLC transporters accumulate either intracellularly or at the plasma membrane. Rab21 has recently reported to be involved in clathrin-and dynaminindependent endocytosis [22,23]. Thus, both scenarios are consistent with Rab21 functions. However, we favor the possibility that SLCs accumulate in intracellular vesicles that are improperly targeted for degradation, or that they are not recycled efficiently to the plasma membrane and are present in aberrant/inefficient recycling endosomes, given that Rab21 depletion sensitizes enterocytes to cell death. We posit that the opposite would occur in conditions where more transporters would be at the cell surface. Intriguingly, the SCL2 (CG8837), and SLC5 (CG2187, salt, CG8966, CG31090) families were the most highly represented, and both are involved in sugar transport [28,70,71], consistent with enrichment in sugar-related Reactome pathways. In the future, it will be interesting to determine if loss of enterocyte sugar import contributes to their death, and to identify the responsible SLC(s). Our data also reveal decreases in several proteins related to proteolysis, which are likely digestive enzymes. Prior to absorption, digestive enzymes are responsible for breaking down ingested macromolecules before their absorption by enterocytes [3]. Therefore, decreased digestive enzymes might result in improper absorption, which could also lead to dysfunctional enterocytes and decreased lifespan.
Finally, we demonstrated decreased abundancies of CYP family members [72,73]. CYPs are a superfamily of enzymes with heme and iron binding functions as well as oxidoreductase activity [74]. Decreased amounts of these enzymes in enterocytes could lead to decreased detoxification capacities in these cells, which could also account for their death. Untangling these observations will be of great interest in our future studies.
To conclude, our data shed light on the importance of the enterocyte early endosomal machinery in maintaining proper tissue homeostasis. Our results demonstrate that in enterocytes, Rab21 acts differently than Rabs previously investigated in this cell type. We show that Rab21 regulates the EGFR pathway and autophagy, although independently of each other. Moreover, we identify deregulation of specific SLC families, digestive enzymes, and CYP proteins, indicating physiological defects in specific cellular processes.
Further investigations of the cell-specific functions of membrane trafficking regulators will highlight their underappreciated roles in tissue and organismal homeostasis.

Drosophila strains
Fly stocks were maintained at 25°C on a standard diet composed of 7 g/L agar, 60 g/L cornmeal, 60 g/L molasses, 23.5 g/L yeast extract, 4.

Generation of the UAS-Rab21 deg transgenic line
The Drosophila Rab21 coding sequence was modified to prevent RNAi binding without affecting the amino acid sequence by modifying the third nucleotide of each codon present in the sequences targeted by Rab21 RNAi-1 and 2. The Rab21 degenerated coding sequence was synthetized by Integrated DNA Technology and cloned into the pUASt-AttB plasmid to generate pUASt-dRab21 deg . Transgenic flies were generated by Genome Prolab through phiC31 transgenesis at the attP2 landing site.

Immunofluorescence
Drosophila guts were dissected in 1× phosphate-buffered saline (PBS) over 20 minutes, fixed for 2 hours in 4% paraformaldehyde, and rinsed three times in PBS. To allow food to exit the lumen, the guts were incubated for 20 minutes in 50% glycerol/PBS and 10 minutes in PBS-0.1% Triton X-100 (PBT). Intestines were then blocked for 1 hour in 20%

Confocal microscopy
Confocal images of guts were acquired on a Zeiss LSM 880 confocal microscope, using either a 20× Plan-APOCHROMAT/0.8 numerical aperture (NA) or a 40× oil Plan APOCHROMAT/1.4 NA objective. Confocal images represent maximum projections.
Settings on the microscope were first adjusted on a control gut and maintained for the acquisition of the different conditions in each experiment. Images were analyzed in Fiji [76] or CellProfiler [77], then linearly thresholded and assembled into figure panels in Photoshop Version 21.1.1 (Adobe Systems, Inc., San Jose, CA, USA). All adjustments to contrast and other aspects of the images were performed similarly for all conditions in each experiment.

TEM
Guts were dissected in 1× PBS, fixed in 2% glutaraldehyde in 0.1 M cacodylate buffer (pH 7.4), and stored at 4°C until processing [58]. Images were acquired on a Hitachi H-7500 transmission electron microscope. Images were analyzed in Fiji and thresholded in Photoshop.

Survival curves
Flies were collected 1-3 days after eclosion and aged until death. Dead flies were scored every 2-3 days, and food was exchanged every other day.

Smurf assays
Flies (1-3 days old) were collected in a tube and incubated at 29°C for 20 days. They were then transferred overnight to food containing FD&C blue dye #1 (1.75 g/100 mL). The next day, blue flies were scored as Smurf + .

Statistical analysis
Unpaired two-tailed Student's t-tests and Mann-Whitney U tests were used to analyze significant differences between pairs of conditions. The proper statistical test for each experiment was determined by assessing the normal distributions of each condition in the experiment. Conditions with normal distributions were analyzed using unpaired two-tailed Student's t-test, while conditions without normal distributions were analyzed using Mann-Whitney U tests. All statistical comparisons were performed using data collected from at least three biological replicates, unless otherwise specified. GraphPad Prism (GraphPad Software, La Jolla, CA, USA) was used for statistical analyses and graph generation.

Protein processing for TMT-based proteomic analysis
Gut proteins were extracted as described above, using the same lysis buffer. Following gut dissection and protein extraction, biological replicates were flash frozen and simultaneously processed for TMT labelling. Proteins (50 µg spectrometer (Thermo Fisher Scientific). Settings used were as previously described [22], except that the full-scan mass spectrometry (MS) survey spectra acquisition (m/z 375-1,400) was realized using a resolution of 140,000 with 3,000,000 ions and a maximum injection time of 120 ms.

MS data analysis
For the TMT-based quantitative proteomic experiment, three independent biological replicates were used. Proteins were identified by MaxQuant using the UniProt Drosophila melanogaster proteome UP000000803. The "proteinGroup.txt" output file from the MaxQuant analysis [79] (Supplementary table 1) was used to collect corrected reporter intensities per sample for every detected protein, as a measure of their quantities in each sample. Since series of replicates were run on different days, we checked for the presence of batch effects by generating a multidimensional scaling plot using the limma v3.42.2 package [80] in the R environment. Once confirmed, batch effects were handled in R using the internal reference scaling (IRS) methodology [81], which is capable of correcting the random MS2 sampling that occurs between TMT experiments. Since the dataset contained proteins which were not quantified in all replicates, we first filtered for proteins that were identified in 2/3 replicates of at least one condition, and then we checked the spectra for missing data using the DEP v1.8.0 R package [82]. Once we verified that missing data were occurring at random, we imputed them using the k-nearest neighbor approach in DEP v1.8.0. Following this analysis, only proteins with 2 unique peptides were conserved for data representation ( Figure 7B) and reactome enrichments ( Figure 7C). Differential expression analysis was performed in R using DEP v1.8.0, with a false discovery rate (FDR) of 0.05 and a log2 fold change (FC) of 1.5. Enriched Reactome pathways were independently searched for proteins with >2-fold increases or decreases in abundance using the "STRING enrichment" plugin in Cytoscape [83]. The mass spectrometry data have been deposited to the ProteomeXchange Consortium via the PRIDE repository with the data set identifier PXD022413.

ACKNOWLEDGMENTS
We thank M. Avino for helpful statistical analysis of the TMT experiment. We thank Bruce Edgard for kindly providing the transgenic fly lines Myo1-Gal4; Tub-Gal80 ts ,GFP and Myo1-Gal4,Tub-Gal80 ts ; and upd3-LacZ. We thank Nicole St. Denis for stylistic and copyediting services. We thank the proteomics platform at the Université de Sherbrooke for proteomic services, the histology and electronic microscopy platform for sample preparation, and the Photonic microscopy platform for use of a confocal microscope. We