An SH3PX1-dependent endocytosis/autophagy network restrains intestinal stem cell proliferation by counteracting EGFR/ERK signaling

SUMMARY

The effect of intracellular vesicle trafficking on stem cell behavior is largely unexplored. We screened the Drosophila sorting nexins (SNXs) and discovered that one, SH3PX1, profoundly affects gut homeostasis and lifespan. SH3PX1 restrains intestinal stem cell (ISC) division through an endocytosis/autophagy network that includes Dynamin, Rab5, Rab7, Atg1, 5, 6, 7, 8a, 9, 12, 16, and Syx17. Blockages in this network stabilize ligand-activated EGFRs, recycling them via Rab11-dependent endosomes to the plasma membrane. This hyperactivated ERK, calcium signaling, and ER stress, autonomously stimulating ISC proliferation. The excess divisions induced epithelial stress, Yki activity, and Upd3 and Rhomboid production in enterocytes, catalyzing feed-forward ISC hyperplasia. Similarly, blocking autophagy increased ERK activity in human cells. Many endocytosis/autophagy genes are mutated in cancers, most notably those enriched in microsatellite instable-high and KRAS-wildtype colorectal cancers. Disruptions in endocytosis and autophagy may provide an alternative route to RAS/ERK activation, resulting in EGFR-dependent cancers.

Graphical Abstract

The origins and functions of resident immune cells in barrier tissue are highly diverse. Lin et al. identify in the zebrafish epidermis an ectoderm-derived immune cell type – metaphocytes – that capture soluble antigens from external environment through transepithelial protrusions and convey these antigens to conventional Langerhans cells via an apoptosis-phagocytosis pathway.

INTRODUCTION

In the intestine, genetic and epigenetic changes that dysregulate epithelial homeostasis contribute to common maladies such as colorectal cancer (CRC) and inflammatory bowel diseases (IBD) (Pesic and Greten, 2016). Cancer genetics suggests that many of the genes and pathways that affect gut epithelial homeostasis remain either unknown, or poorly characterized in the context of real epithelia. However, such genes can be readily identified and functionally analyzed in the gut epithelium of Drosophila, which shares many similarities with its human counterpart. Epithelial renewal in the fly’s midgut is sustained by intestinal stem cells (ISCs) that divide to renew themselves and also produce transient cells (enteroblasts; EBs) that further differentiate into multiple types of enterocytes (ECs) or enteroendocrine (EE) cells (Micchelli and Perrimon, 2006; Ohlstein and Spradling, 2006). The intestinal epithelium turns over rapidly in both humans and insects, at rates that vary from days to weeks according to food composition, ingested toxins and pathogens (Jiang et al., 2009; Ohlstein and Spradling, 2006). Under both normal homeostatic and stressed conditions, ISC divisions are regulated by numerous signaling pathways that maintain the critical balance between cell generation and loss (Jiang and Edgar, 2011; Miguel-Aliaga et al., 2018). But how the cell biology of ISCs impacts signal transduction in these pathways and maintains the critical signaling balance necessary for epithelial homeostasis, remains largely unexplored.

Recent studies show that calcium signaling integrates diverse mitogenic signaling inputs that control ISC proliferation (Deng et al., 2015; Xu et al., 2017), and accumulating evidence indicates that calcium fluxes are involved in many constitutive membrane trafficking events (Hay, 2007). However, whether intracellular vesicle trafficking affects ISC division and gut homeostasis has not been checked. One family of genes that regulate intracellular membrane traffic are the sorting nexins (SNXs), evolutionarily conserved proteins that are classified by the presence of phox-homology (PX) domains (Cullen, 2008). The phosphoinositide-binding ability of the PX domain explains in part the role for SNX in multiple steps of membrane trafficking including endocytosis, endosomal sorting and signaling. Not surprisingly, SNX mutations have increasingly been associated with diseases (e.g., Alzheimer’s disease and various cancers) in which endosomal function is compromised (Teasdale and Collins, 2012). The human genome encodes 33 known SNXs (Cullen, 2008) that can be classified on the basis of common domain structures into different subfamilies: PX-only, PX-BAR, SH3-PX-BAR and PX-FERM (Teasdale and Collins, 2012). Nine Drosophila SNXs are conserved with the 33 human SNXs (Rodal et al., 2011; Zhang et al., 2011), but little is known of their in vivo functions. SNX3, a PX-only subfamily member, is reportedly a component of the Retromer complex, which regulates Wnt/Wingless secretion (Zhang et al., 2011). SNX16 complexes with Nervous Wreck (Nwk) to control synaptic growth signaling (Rodal et al., 2011). The fly ortholog of human SNX9, 18, and 33, called SH3PX1, is required for autophagosome formation in the fat body (Knaevelsrud et al., 2013).

Over the past decade, autophagy has been linked to diverse pathophysiological processes including metabolic and neurodegenerative disorders, cardiovascular diseases and cancer (Levine and Kroemer, 2008). Yet while autophagy has been reported to affect stem cell maintenance and differentiation (Vessoni et al., 2012), the precise mechanisms underlying these consequences are unclear. Several studies report that autophagy genes are required in intestinal epithelial cells (iECs) for ISC maintenance and CRC formation, presumably via cell non-autonomous mechanisms. Asano et al showed that deletion of Autophagy-related gene 5 (ATG5) in murine iECs using Villin-Cre (a pan-iEC recombinase), or in Paneth cells using Ah-Cre, led to a decrease in ISCs over time (Asano et al., 2017). Lévy et al showed that knockout of ATG7 in iECs inhibits adenoma formation in APC^+/− mice (Levy et al., 2015). Loss of function studies of ATG16L1 show that this autophagy gene, which is associated with bowel inflammation in Crohn’s disease, is necessary for Paneth cell survival and secretory function (Bel et al., 2017; Cadwell et al., 2008; Matsuzawa-Ishimoto et al., 2017). So far, however, the roles of autophagy and endosomal trafficking general have not been investigated specifically in ISCs using ISC-targeted deletions. Moreover, the effect of SNX-mediated endosomal trafficking on ISC functions is still unknown.

In this study, a screen of the entire family of Drosophila SNXs discovered that one, SH3PX1, is critical for restraining ISC proliferation and maintaining gut epithelial homeostasis. SH3PX1 accomplishes this via its role in an extended endocytosis/autophagy network that contains many other genes that similarly affect ISCs. Further, we found that EGFR/Ras/MAPK signaling is the primary trigger for ISC proliferation upon disruption of the endocytosis/autophagy network. Blocking autophagy and re-routing endosomal traffic in this network leads to accumulation of ligand-activated EGF receptors in Rab11-dependent endosomes, increased ERK and Ca²⁺ signaling, and ER stress in ISCs, thereby driving their division. We noted similar effects with autophagy inhibitors in human cells. In human cancers, SNX9, 18, and 33 and other endocytosis and autophagy genes are commonly altered. This occurs most significantly in CRC where these genes are tightly associated with microsatellite instable-high (MSI-H) and CpG island methylator phenotype-high (CIMP-H) molecular phenotypes. A negative association was also uncovered between loss-of-function mutations in SNX9/18/33 and activating KRAS mutations. This suggests, like our Drosophila work, that the disruption of the endocytosis/autophagy network provides an alternate route to EGFR signaling activation in human cancers. The endocytosis/autophagy network may offer new diagnostic and therapeutic strategies for CRC and other EGFR-dependent diseases that involve excessive stem cell proliferation and epithelial turnover.

RESULTS

Drosophila SH3PX1 restrains ISC proliferation

Nine sorting nexin genes (Snx1, Snx3, Snx6, SH3PX1, Snx16, Snx17/CG5734, Snx21/CG3077, snz, and Snx27/CG32758) have been identified in Drosophila (Rodal et al., 2011; Zhang et al., 2011). All of the Snx null mutants except Snx17 were homozygous viable and had no obvious developmental defects (Rodal et al., 2011; Zhang et al., 2011), thereby allowing examination of ISC status in mutant adults. Three SNXs were shown to regulate ISC mitoses: SH3PX1, SNX16 and Snz. SNX16 mutants had a modest upregulation of ISC mitoses, whereas homozygous Snz mutants exhibited a mild repression of ISC mitoses relative to heterozygous controls (Figure 1A). Homozygous mutants of SH3PX1^d1/d1 (a P-element excision null allele, Figures S1A–S1B) showed a strong increase in ISC mitoses (Figure 1A). As evaluated using miranda-GFP (mira-GFP), a marker for ISCs and EBs (Bardin et al., 2010), homozygous SH3PX1^d1/d1 mutants had marked increases in GFP⁺ cells compared with heterozygote controls (Figure 1B). MARCM clonal analysis showed that the SH3PX1^d1/d1 mutant cells grew faster than controls, generating larger than normal clones after 14 days (Figures S3A–S3C).

Figure 1. Drosophila SH3PX1 restrains ISC proliferation.

(A, C, F, G) Midguts were stained with anti-pH3 antibody. ISC mitoses were quantified by pH3⁺ cells. Quantification data shown in A, C, F, and G represent the mean±SD (t-test, ^nsP>0.05, **P<0.01, ****P<0.0001). Each dot represents one sample. (B, E) Midguts were stained with anti-GFP and anti-pH3 antibodies, and DAPI. (D) 2–3 day-old esg-Gal4,UAS-GFP;tubGal80^ts (esg^ts) flies were shifted from 18°C to 29°C and raised for 2d on normal fly food before infection. Flies were then transferred to empty vials containing Whatman discs soaked with 5% sucrose and yeast paste +/− P.e. for 16hr at 29°C. Midguts were dissected and stained with anti-SH3PX1 antibody. (E-F) Two independent RNAi lines against SH3PX1 were driven using an ISC-specific driver esg-Gal4,UAS-2xEYFP;Su(H)GBE-Gal80,tub-Gal80^ts (esg^ts;Su(H)GBE-Gal80). 2–3 day-old adult females were shifted from 18°C to 29°C for 7d before dissection. The SH3PX1^RNAi lines’ knockdown efficiencies are shown in Supplementary Figures 2A–2C’. (G) UAS-GFP-SH3PX1 was overexpressed using esg-Gal4 at 25°C for 3d in the SH3PX1^d1/d1 mutant background. (H) Flies were raised at 25°C on normal fly food. Survival curves are shown for each genotype as indicated (n=30, per genotype). Scale bars in B and E, 40 μm; D, 10 μm.

To rule out potential effects caused by the SH3PX1^d1/d1 genetic background, another null allele, SH3PX1^HK62b, was utilized (Knaevelsrud et al., 2013) (Figure S1A). Trans-heterozygous SH3PX1^d1/HK62b mutants showed an ISC mitotic phenotype similar to SH3PX1^d1/d1 mutants (Figures 1C and S1C), confirming that SH3PX1 is specifically restrains ISC divisions. To further assess the intestinal role of SH3PX1, SH3PX1 protein was visualized using an anti-SH3PX1 antibody (Worby et al., 2001). SH3PX1 was expressed in all cell types in the midgut, membrane localized, and showed relatively high amounts in escargot-positive (esg⁺) progenitor cells (ISCs + EBs; Figure 1D). Interestingly, enteric infection with Pseudomonas entomophila (P.e.), which damages the gut epithelium and results in faster regenerative growth (Jiang et al., 2009), disrupted the membrane-associated pattern of SH3PX1 (Figure 1D, lower panels), suggesting that proper subcellular localization of SH3PX1 might be required for intestinal homeostasis and normal regeneration.

We next tested whether SH3PX1 functions in a cell autonomous or non-autonomous manner to control ISC proliferation, using different cell type-specific Gal4 “drivers” to express RNAi targeting SH3PX1. Knockdown of SH3PX1 using two independent RNAi strains driven by the ISC-specific driver esg^ts Su(H)GBE-Gal80 generated large increases in ISC mitoses similar to those seen in the null mutants (Figures 1E–1F). Knockdown of SH3PX1 using the EB-specific driver Su(H)GBE^ts slightly increased the pH3⁺ cell numbers (Figure S1D), consistent with a previous observation that a small population of EBs maintains proliferative capability (Jin et al., 2015). However, depleting SH3PX1 in ECs (via the Myo1A^ts driver), or in EEs (via the ProsV1^ts driver), had no effect on ISC mitoses (Figures S1E–S1F), indicating that SH3PX1 functions cell-autonomously to regulate ISC proliferation. As expected, expression of SH3PX1 in progenitor cells successfully rescued the ISC over-proliferation defect and lifespan deficit in homozygous SH3PX1^d1/d1 mutants (Figure 1G). Moreover, esgGal4-driven SH3PX1 expression also rescued the lifespan defect in these mutant flies (Figure 1H), indicating that the mortality of SH3PX1^d1/d1 mutants was due primarily to midgut progenitor cell hyperplasia.

Previous studies showed that blocking differentiation (e.g., by Notch mutation) causes ISC hyperproliferation. To rule out the possibility that SH3PX1^−/− promotes ISC mitosis by blocking differentiation, SH3PX1^d1 mutant MARCM clones were generated and stained for the differentiation markers Pdm1 (an EC marker) and Prospero (an EE marker). Both Pdm1 and Prospero (Figures S3D–S3D“) were present at normal frequencies inside SH3PX1^d1/d1 mutant clones, indicating that SH3PX1 loss does not block differentiation. Thus SH3PX1 specifically restrains ISC proliferation.

SH3PX1-dependent autophagy restricts ISC proliferation

SH3PX1 and its human ortholog SNX18, have been reported to be positive regulators of autophagosome formation (Knaevelsrud et al., 2013). To determine whether SH3PX1 controls ISC proliferation by affecting autophagy, an esgGal4>UASp-GFP-mCherry-Atg8a fly line was generated to visualize autophagosome formation in ISCs and EBs. After 6hr starvation, autophagosomes were clearly observed in the ISCs of heterozygous SH3PX1^d1/+ flies but not in homozygous SH3PX1^d1/d1 flies (Figure 2A). This indicated that SH3PX1 is required in ISCs for autophagosome formation, and suggested that other mediators of autophagy might also restrain ISC proliferation. Indeed, RNAi’s directed against the autophagy-related genes Atg1, 5, 6, 7, 8a, 9, 12 and 16 significantly increased ISC proliferation when expressed in progenitor cells (Figures 2B–2C). Depletion of Syntaxin 17 (Syx17), a regulator of autophagosome-lysosome fusion (Takats et al., 2013), also promoted ISC proliferation (Figure 2B). As these Atg genes control different steps in the autophagy pathway (Mulakkal et al., 2014), we conclude that SH3PX1 restricts ISC proliferation by promoting autophagy.

Figure 2. SH3PX1-dependent autophagy restricts ISC proliferation.

(A) Flies were raised at 18°C. 3-day-old esg-Gal4/UASp-GFP-mCherry-atg8a;SH3PX1^d1FRT^2A/TM6B or esg-Gal4/UASp-GFP-mCherry-atg8a;SH3PX1^d1FRT^2A/SH3PX1^d1FRT^2A flies were starved for 6hr in the vials containing Whatman paper wetted with 5% sucrose, then were stained with anti-GFP and anti-RFP antibodies, and DAPI. esgGal4>UASp-GFP-mCherry-atg8a marked autophagosomes in progenitor cells. The red boxes in the left panels were enlarged and visualized in the middle/right panels. Arrowheads indicate autophagosome formation. (B-G) Different genetic manipulations in progenitor cells driven by esg^ts. 2~3-day-old adult females were shifted from 18°C to 29°C for 5–6d (as indicated in panels) before dissection. Midguts were stained with anti-GFP and anti-pH3 antibodies, and DAPI. ISC mitoses were quantified by pH3⁺ cells. Quantification data shown in B, D-G represent the mean±SD (t-test, ^nsP>0.05, **P<0.01, ***P<0.001, ****P<0.0001). Each dot represents one sample. Scale bars in A, 5 μm; C, 40 μm.

Previous studies indicated that multiple stages of endocytic flux are also essential for efficient autophagy (Tooze et al., 2014), raising the possibility that endocytic vesicle trafficking also controls ISC proliferation. Two key regulators of endocytosis, Rab5 and Rab7, were tested. Overexpression of Rab5^SN, a dominant negative form of Rab5 (Strutt et al., 2011), via the esg^ts driver led to a striking increase of ISC mitoses (Figure 2D). Similarly, RNAi targeting Rab7 in progenitor cells also promoted proliferation (Figure 2D). Thus inhibition of endocytosis also leads to ISC hyperproliferation.

The proliferative phenotypes that arise from these disruptions of endocytic traffic could be due to excessive accumulation of mitogenic cargoes (e.g., activated transmembrane receptors) that are normally degraded via autophagy. If fluxes in the endocytosis/autophagy network were reversed due to blockages, mitogenic cargoes might accumulate in the active form in various endocytic compartments or be recycled to the cell surface. There are two well characterized endocytic recycling pathways: a Rab4-mediated fast route directly from early endosomes, and a slow Rab11-mediated route from perinuclear recycling endosomes (Guichard et al., 2014). ISC-specific knockdown of Rab11 strikingly repressed the hyperproliferation caused by SH3PX1 depletion (Figure 2E). However, overexpression of Rab4^SN, a dominant negative form of Rab4 (Strutt et al., 2011), did not (Figure 2E). Consistently, knockdown of Rab11 also repressed ISC mitoses caused by Rab5^SN, Rab7^RNAi, Atg1^RNAi and Syx17^RNAi (Figures 2D–2F). These data indicate that Rab11-mediated recycling endosomes play an important role in triggering ISC mitosis following endocytosis/autophagy disruption. This may be because Rab11 endosomes are required to return active, mitogenic cargoes to the plasma membrane, or because such cargoes only remain active in Rab11 endosomes.

Next, we asked whether autophagy and endocytosis work together or in parallel to control ISC proliferation. To this end, an epistasis test was performed between SH3PX1 and the Drosophila dynamin homolog shibire (shi), which is required at the initial step of endocytosis. Overexpression of shi^TS, a temperature-sensitive dominant negative mutant form of shi (Kitamoto, 2001) using the esg^ts driver moderately increased ISC mitoses (Figure 2G). This is consistent with the idea that blocking endocytosis stabilizes growth factor receptors or other mitogenic factors at the plasma membrane, thereby promoting ISC proliferation. If shi-dependent endo-lysosomal degradation works in parallel with SH3PX1-dependent autophagy, overexpression of shi^TS in conjunction with SH3PX1^RNAi would be expected to have additive effects on ISC mitoses. Contrary to this expectation, expression of shi^TS suppressed midgut ISC mitoses caused by SH3PX1^RNAi (Figure 2G). This suggests that the mitogenic cargoes that accumulate after SH3PX1 dysfunction likely originate, at least in part, from early endocytosis at the plasma membrane. From this, we surmise that autophagy and endocytosis work together to restrain ISC proliferation.

Autophagy restrains ISC division by dampening EGFR/ERK activity

To elucidate which downstream effectors of SH3PX1-dependent autophagy are responsible for inducing ISC proliferation, we did epistasis tests between SH3PX1 and signaling pathways known to promote ISC division. These included the insulin/PI3K, JAK/STAT, JNK, wingless, and EGFR pathways. Silencing each of these pathways could suppress ISC mitosis caused by SH3PX1 depletion (Figure 3A). However, in most cases these inhibitory effects diminished over time (Figures 3A and S4A). Silencing of the EGFR pathway by MEK depletion, however, gave a singularly strong and persistent inhibitory effect (Figures 3A and S4A), suggesting that EGFR signaling may be dominant to the other pathways. To further test the primacy of EGFR signaling, the epistatic relationships of SH3PX1 with other components of the pathway (EGFR, Ras, pointed (pnt) and Ets21C) were tested (Jin et al., 2015). Suppressing any of these genes’ functions strongly and permanently repressed SH3PX1^RNAi-driven ISC mitoses and intestinal dysplasia (Figures 3B–3C). Furthermore, ISC mitoses induced by deficiencies in autophagy (Atg1^RNAi, Atg8a^RNAi and Syx17^RNAi) could also be blocked by the depletion of EGFR pathway components (Figures 3D and S5A). These data suggest that EGFR signaling is a major downstream target of SH3PX1-dependent autophagy.

Figure 3. SH3PX1-mediated autophagy restrains ISC mitoses by damping EGFR/ERK activity.

(A-D) Different genetic manipulations in progenitor cells driven by esg^ts. 2~3-day-old adult females were shifted from 18°C to 29°C for 6d or 12d (as indicated in panels) before dissection. Midguts were stained with anti-GFP and anti-pH3 antibodies, and DAPI. ISC mitoses were quantified by pH3⁺ cells. Quantification data shown in A, C and D represent the mean±SD (t-test, ^nsP>0.05, ****P<0.0001). Each dot represents one sample. (E-F) Flies with indicated genetic manipulations were shifted from 18°C to 29°C for 6d or 7d (as indicated in panels) before dissection. Midguts were stained with anti-GFP and anti-dpERK antibodies, and DAPI. Scale bars in B, E and F, 40 μm.

Because these results suggest that dysfunction in any step of the endocytosis/autophagy network may trigger hyperactivation of EGFR signaling, the activation of the principal EGFR effector, mitogen-activated protein kinase (MAPK), was examined by staining for active, di-phosphorylated ERK (dpERK). Indeed, dpERK signals were strikingly induced in SH3PX1^d1/d1 mutant midguts (Figure 3E). RNAi-mediated SH3PX1 depletion also increased the levels of dpERK (Figure S5B), predominantly in the targeted esg⁺ progenitor cells. Consistently, RNAi expression targeting either autophagic flux (esg^ts>Atg1^RNAi, Atg8a^RNAi, Syx17^RNAi) or endocytosis (esg^ts>Rab5^SN, Rab7^RNAi) in progenitor cells induced high levels of dpERK, similar to the effects of SH3PX1 loss (Figures 3F and S5D).

Further tests showed that two other receptor tyrosine kinases expressed in the midgut, fibroblast growth factor receptor (FGFR) and PDGF/VEGF receptor (PVR), were not important for MAPK activation following SH3PX1 loss (Figure S5C). Similarly, SH3PX1^RNAi was fully capable of activating dpERK in ISCs depleted of Insulin Receptor (InR), the Upd cytokine receptor (Dome), JNK, or the Wnt effector Pangolin (TCF) (Figure S4B). Altogether these data indicate that EGFR/ERK signaling is directly activated as a result of autophagy dysfunction.

EC-localized stress signaling is dispensable for proliferation of SH3PX1 mutant ISCs

Two possible mechanisms of MAPK activation following SH3PX1 loss were examined: In the first scenario, EGFR/Ras/MAPK signaling is a direct cell autonomous target of SH3PX1-dependent autophagy. In the second, dysfunction of SH3PX1-dependent autophagy causes ISC proliferation via a different mechanism, but proliferating ISCs stress the gut epithelium, which leads to the production of Unpaired cytokines, EGFR ligands, and Rhomboids (Jiang et al., 2011), thus triggering EGFR/Ras/MAPK signaling in indirectly. This second, non-cell autonomous mechanism is well documented in the fly gut (Jiang et al., 2011; Patel et al., 2015; Xiang et al., 2017). Previous studies have demonstrated that the JNK and Hippo pathways are major inducers of the cytokines Unpaired 2 and 3 (Upd2, 3) following stress (Jiang and Edgar, 2011), and that Upd/Jak/Stat signaling can secondarily promote EGFR signaling (Jiang et al., 2011). We first examined the activity of the Hippo/Yki pathway. In the R4 region of wild-type midguts, the Yki target, ex-lacZ, showed weak expression in ECs and was absent in ISCs (Figure 4A, upper panel). In contrast, loss of SH3PX1 in progenitor cells markedly increased ex-lacZ levels in ECs (Figure 4A, lower panel). To determine whether Yki activation was responsible for ISC proliferation in the context of SH3PX1 loss, we depleted yki in ECs using an RNAi. In this case, ISCs mitoses in the SH3PX1^d1/d1 mutants were significantly reduced (Figure 4B). Given that Yki can upregulate Upd2 and Upd3 in ECs (Shaw et al., 2010), we tested the induction of Upd3 using the upd3-lacZ reporter (Jiang et al., 2009). As expected, upd3-lacZ was exclusively upregulated in ECs surrounding SH3PX1-depleted progenitors (Figure 4C, lower panel). To determine whether Upd induction contributes to ISC proliferation in SH3PX1 mutant, we removed both upd2 and upd3 using a viable deletion mutant, and found that ISC mitoses in the SH3PX1^d1/d1 mutants were significantly reduced (Figure 4D). This repression was attenuated over time, however (Figure 4D) and moreover, in upd2^Δ,3^Δ; SH3PX1^d1/d1 mutant midguts, we still observed massively high dpERK signals in the hyperproliferative progenitor cells (Figure 4E). These results indicate that while the Upd2 and 3 cytokines are induced and contribute to SH3PX1⁻-driven hyperplasia, cytokine signaling is not essential for the hyperplasia or for EGFR activation.

Figure 4. EC-localized stress signaling is dispensable for proliferation of SH3PX1 mutant ISCs.

(A, C, F) UAS-SH3PX1^RNAi−2 was overexpressed using esg^ts at 29°C for 3d or 4d, as indicated in panels before dissection. Ex-lacZ (A, red) expression indicated Yki activity. Upd3 expression levels were indicated by the upd3-lacZ reporter (C, red). Rho induction was indicated by the rho-lacZ reporter (F, red). Progenitor cells were marked by GFP. (B, D, G and H) Midguts were stained with anti-pH3 antibody. ISC mitoses were quantified by pH3⁺ cells. (B) UAS-yki^RNAi was overexpressed using myo1A-Gal4 at 25°C for 8d in the SH3PX1^d1/d1 mutant background. (D) Adult males of +/Y;mira-GFP;SH3PX1^d1/d1 (control) and upd2^Δ,3^Δ /Y;mira-GFP;SH3PX1^d1/d1 were incubated at 25°C for 4 or 8d after eclosion. (E) The same genotype flies as described in panel D were stained with anti-GFP and anti-dpERK antibodies and DAPI at day 7. (G-H) UAS-rho^RNAi (G) or UAS-Krn^RNAi (H) was overexpressed using Myo1A-Gal4 at 25°C for 3d or 6d (as indicated in panels) in the SH3PX1^d1/d1 mutant background. Quantification data shown in C, D, G and H represent the mean±SD (t-test, ^nsP>0.05, ****P<0.0001). Each dot represents one sample. Scale bars in A, 30 μm; C, E and F, 20 μm.

These observations raised the possibility that other EC-derived mitogens might promote ISC mitoses. Several Rhomboid (Rho) proteases are induced in stressed ECs, and these promote the cleavage and secretion of EGFR ligands Spitz (Spi) and Keren (Krn) (Jiang et al., 2011; Liang et al., 2017). Consistently, we found that rho-lacZ, a rhomboid transcriptional reporter, was highly induced in ECs surrounding SH3PX1-depleted progenitors (Figure 4F). To test whether rho and its substrates were required for ISC activation, we depleted either rho or Krn in ECs using RNAi, in the SH3PX1^d1/d1 mutant background. This failed to repress the ISC divisions caused by SH3PX1 loss (Figures 4G–4H), suggesting that the paracrine induction of rho and secretion of Krn from ECs are not required for ISC hyperproliferation. Since the induction of dpERK activity in ECs was observed in most cases of endocytosis/autophagy dysfunction (Figures 3F, 5E, S4B, S5B and S5D), we tested whether RNAi-mediated suppression of MEK in ECs might affect ISC activation in SH3PX1 mutants. While MEK^RNAi effectively blocked ERK activity in ECs (Figure 5E), it did not suppress ISC-autonomous ERK activation or mitoses (Figure 5D, E). Overall these data indicate that stress signaling within and from ECs is dispensable for ISC mitoses triggered by SH3PX1 mutations.

Figure 5. Cell autonomous EGFR/MAPK activation triggers SH3PX1^--ISC division.

(A, D) Midguts were stained with anti-pH3 antibody. ISC mitoses were quantified by pH3⁺ cells. (A) SH3PX1 knockdown or SH3PX1/stg double knockdown was driven by esg^ts. esg^ts>w1118 was used as control. Flies were raised at 18°C and then shifted to 29°C for 5d before dissection. (B-C) SH3PX1/stg or SH3PX1/MEK double knockdown was driven by esg^ts at 29°C for 4d. Induction of upd3 or rho was indicated by upd3-lacZ (upper panels of B, and C) or rho-lacZ (lower panels of B, and C), respectively. Progenitor cells were marked by GFP. (D-E) UAS-MEK^RNAi was overexpressed using myo1A-Gal4 at 25°C for 3d in the SH3PX1^d1/d1 mutant background. Myo1A-Gal4/+;SH3PX1^d1/d1 flies were selected as positive control (D and left panels of E). Myo1A-Gal4/MEK^RNAi; SH3PX1^d1/+ flies were selected as negative control (E, right panels). (E-F) Midguts were stained with anti-GFP antibody and anti-dpERK antibodies, and DAPI. White asterisk in (E) indicates the EC cell. (F) SH3PX1 knockdown (control) or SH3PX1/stg double knockdown was driven by esg^ts at 29°C for 5d. Quantification data shown in A and D represent the mean ± SD (t-test, nsP>0.05, ***P<0.001, ****P<0.0001). Each dot represents one sample. Scale bars in B and C, 20 μm; E, 30 μm; F, 40 μm.

Cell autonomous EGFR/Ras/MAPK activation triggers SH3PX1⁻-ISC division

Given these findings, we inferred that ISC-autonomous EGFR/Ras/MAPK activation was the primary trigger of hyperplasia in the SH3PX1 mutants, and that the induction of Yki-, Upd3-, and Rho-reporters in ECs was likely a secondary consequence of ISC hyperplasia. To test this hypothesis, we blocked ISC mitoses using an RNAi against String (stg), a Cdc25 homolog required for all somatic cell mitoses in Drosophila (Edgar and O’Farrell, 1989). Co-expressed stg^RNAi completely blocked ISC mitoses caused by SH3PX1^RNAi (Figure 5A), and also suppressed the induction of upd3-lacZ and rho-lacZ (Figure 5B). Similarly, MEK^RNAi expressed in SH3PX1-mutant ISCs blocked mitoses and the induction of upd3-lacZ and rho-lacZ in ECs (Figure 5C). These results confirm that it is the hyperproliferation of SH3PX1-mutant ISCs that induces stress-dependent mitogenic signaling from ECs. Importantly however, even when epithelial stress was pre-empted by blocking ISC mitoses with stg^RNAi, activation of dpERK was still observed in the SH3PX1-mutant ISCs (Figure 5F). Thus, all of our data are consistent with the idea that ISC-autonomous activation of EGFR/Ras/MAPK signaling is the primary trigger for ISC mitoses in cases of endocytosis-autophagy dysfunction.

SH3PX1 regulates EGFR signaling via multiple mechanisms

We next considered how blockages in the endocytosis/autophagy network might activate EGFR/Ras/MAPK signaling in ISCs. To assess EGFR expression in the fly midgut we used esg^tsF/O, a lineage induction system that labels ISCs and all their newborn progeny after a temperature shift (Jiang et al., 2009). Consistent with our previous observations (Xiang et al., 2017), we noted that EGFR is highly expressed in ISCs and EBs and strongly reduced in ECs (Figure 6A, left), suggesting that it is down-regulated as progenitor progeny differentiate. Surprisingly, overexpression of EGFR with esg^tsF/O gave essentially the same expression pattern (Figure 6A, middle), even though this induction system drives the transcription of EGFR in ECs as well. This suggests that the stability of EGFR protein must be reduced as progenitor cell progeny differentiate. Remarkably, in esg^tsF/O>SH3PX1^RNAi,EGFR^wt midguts, EGFR protein became visible in newborn ECs (Figure 6A, right). This implies that EGFR is likely stabilized by SH3PX1 depletion, and thus that SH3PX1-dependent autophagy promotes EGFR degradation. Stabilization of activated EGF receptors, either at the plasma membrane or in endosomes, may be one mechanism via which ERK is activated by blocking flux through the endocytosis-autophagy network.

Figure 6. SH3PX1 controls the EGFR pathway via multiple routes.

(A) UAS-EGFR^wt or UAS-SH3PX1^RNAi−1+UAS-EGFR^wt was overexpressed by the esg^tsF/O system. Esg^tsF/O>w1118 was used as control. Flies were raised at 18°C and then shifted to 29°C for 3d before dissection. Midguts were stained with anti-EGFR and anti-GFP antibodies, and DAPI. White asterisks indicate ISC or EB cells. Yellow asterisks indicate EC cells. (B) w1118, UAS-SH3PX1^RNAi−1 and UAS-EGFR^wt were overexpressed by esg^ts. Flies were raised at 18°C and then shifted to 29°C for 1–6d before dissection. Samples were dissected every day from d1 to d6 and stained with anti-pH3 antibody. ISC mitoses were quantified by pH3⁺ cells. Values represent mean±SEM. (C-F, I) Midguts were stained with anti-pH3 antibody. ISC mitoses were quantified by pH3⁺ cells. (C-F) Different genetic manipulations in progenitor cells driven by esg^ts. 2~3-day-old adult females were shifted from 18°C to 29°C for 6d before dissection. (G-H) UAS-GCaMP6s or together with UAS-SH3PX1^RNAi−1 was overexpressed using esg^ts. (G) Representative Ca²⁺ plots of single esg⁺ midgut cells from wild-type (blue) and SH3PX1 knockdown (red) flies. Normalized fluorescence intensity is plotted over time, highlighting differences in intracellular Ca²⁺ peak durations. (H) Peak durations as determined from Ca²⁺ plots of wild-type (n=155 peaks) and SH3PX1 knockdown (n=80 peaks) esg⁺ midgut cells; P-value calculated by Mann-Whitney two-tailed U test. (I) SH3PX1 knockdown, SH3PX1/TrpA1 double knockdown, and SH3PX1/RyR double knockdown were driven by esg-Gal4 at 25°C for 6d. Quantification data shown in C-F and I represent the mean±SD (t-test, ^nsP>0.05, ****P<0.0001). Each dot represents one sample. Scale bars in A, 10 μm.

Notwithstanding these findings, overexpressing EGFR^wt in progenitors produced a consistently weaker mitotic phenotype than SH3PX1^RNAi (Figure 6B), suggesting that SH3PX1 loss might potentiate ERK signaling via additional mechanisms. Previous studies show that two autocrine EGFR ligands, Spi and Krn, are expressed by ISCs (Buchon et al., 2010; Jiang et al., 2011; Patel et al., 2015). Cleavage and secretion of Spi and Krn require the intramembrane protease, Rhomboid (Rho). We thus examined whether Rho, Spi, and Krn were required for SH3PX1^RNAi-dependent EGFR/Ras/MAPK activation. Whereas ISC-targeted RNAi against spi did not inhibit SH3PX1^RNAi-dependent mitoses, suppressing either rho or Krn did (Figures 6C–6E). These results suggest that ISC-produced Rho and Krn potentiate the activity of EGFR, which is stabilized by dysfunctions in autophagy. Considering the importance of Krn and Rho, we note that the activity of these factors might also be affected by impairing autophagic flux. Notably, the over-expression of rho in ISCs was sufficient to trigger mitoses (Figure 6F), suggesting the cleavage of Krn may be a limiting step for EGFR activation in ISCs, and the proliferative effect of overexpressed rho was suppressed by knockdown of Rab11 (Figure 6F), which has been shown to be required for the recycling of EGFR (Cullis et al., 2002). Taken together, these data suggest that in SH3PX1 mutant ISCs, Rho-activated Krn but not Spi may activate EGFR in Rab11 endosomes, thereby initiating ISC mitoses.

Since autophagy functions as a non-canonical Endoplasmic Reticulum-Associated Degradation (ERAD) pathway (Senft and Ronai, 2015), the perturbation of SH3PX1-dependent autophagy might cause retention of proteins in the ER, triggering ER stress. We assessed this possibility by monitoring ER stress in SH3PX1-depleted midguts using the UAS-xbp1-EGFP reporter, in which EGFP is expressed in frame only upon ER stress (Ryoo et al., 2007). This test indicated that SH3PX1^RNAi apparently triggered ER stress (Figure S6). Further, as ER stress can trigger Calcium release from ER stores (Deniaud et al., 2008), we used UAS-GcaMP6s, a Ca²⁺ reporter (Xu et al., 2017), to record intracellular Ca²⁺ flux in midgut progenitor cells. Relative to controls, SH3PX1^RNAi expressing progenitors displayed reduced Ca²⁺ oscillation frequencies but longer peaks of high Ca²⁺ activity (Figures 6G–6H), suggestive of increased overall Ca²⁺ signaling. A recent study showed that high cytosolic Ca²⁺ is necessary and sufficient to activate ERK and mitosis in ISCs, and that EGFR was required for this effect (Xu et al., 2017). Consistent with this mechanism, we found that esg^ts-driven RNAi targeting either transient receptor potential A1 (TRPA1), a plasma membrane Ca²⁺-permeable cation influx channel (Xu et al., 2017), or ryanodine receptor (RyR), an ER-localized cation channel that releases Ca²⁺ into the cytosol (Xu et al., 2017), strongly suppressed ISC mitoses caused by SH3PX1 depletion (Figure 6I). These data suggest that SH3PX1 may deregulate EGFR/Ras/MAPK signaling by perturbing intracellular calcium fluxes. Altogether, our data indicate that dysfunctions in endocytosis/autophagy network activate EGFR/Ras/MAPK signaling via multiple inputs.

Autophagy mutations are associated with MSI-H and CIMP-H colorectal cancers

To test whether the restriction of EGFR/Ras/MAPK activation via the SH3PX1-dependent endocytosis/autophagy network is a conserved process from insects to humans, we generated transgenic flies carrying human SNX9, 18 and 33 genes, and tested their ability to rescue SH3PX1 mutant phenotypes in the Drosophila midguts. The tests showed that human SNX9, 18, or 33 could effectively rescue Drosophila SH3PX1’s loss-of-function phenotype in ISCs (Figure 7A), confirming that Drosophila SH3PX1 is functionally conserved with human SNX9, 18 and 33 genes.

Figure 7. Autophagy mutations are associated with MSI-H and CIMP-H colorectal cancers.

(A) UAS-hSNX9, hSNX18 or hSNX33 were overexpressed using esg-Gal4 at 25°C for 3d in the SH3PX1^d1/d1 mutant background. (B-E) RPE-1 cells were serum starved for 18hr, then treated with the autophagy inhibitor, 3-MA (4mM), for the times indicated. Cell lysates were subjected to western blot analysis with the indicated antibodies. (C) The induction (fold change) of dpERK samples in (B) was quantified after normalization against total-ERK. Values represent mean±SEM (n=4 independent experiments, t-test, *P<0.05). (E) The relative levels of EGFR shown in (D) were quantified after normalization against Vinculin. Values represent mean±SEM (n=2 independent experiments, t-test, *P<0.05). (F) Summary of SH3PX1-dependent endocytosis/autophagy mutations and deep deletions by cancer type in human TCGA datasets (n=14 datasets with ≥10% mutation and deep deletion alteration frequency and n³30 cases). The network gene list and corresponding human orthologs are shown in Table S1. (G-J) Analysis of somatic mutations in human colorectal adenocarcinoma (DFCI, n=619 cases). (G) Summary of somatic mutations by gene. (H-J) Proportion of wild-type (black) versus mutated (red) samples with the MSI-H genomic phenotype (H) and the CIMP-H epigenomic phenotype (I) (n=5 genes with ≥1% altered cases). (J) Proportion of SH3PX1 human homologs SNX9/18/33 wild-type versus mutated samples by KRAS wild-type status.

Next, we examined the ability of autophagy to restrain EGFR/ERK signaling in cultured human Retinal Pigment Epithelial (RPE-1) cells (Spalluto et al., 2013) and CaCo-2 (KRAS/BRAF wild-type) colon cancer cells (Ahmed et al., 2013). Consistent with our results from Drosophila, depressing autophagy in either cell line using the autophagy inhibitors, 3-MA (Figures 7B–7C) or Thapsigargin (TG) (Figures S7A–S7B), rapidly increased dpERK levels. Further, 3-MA treatment caused a rapid accumulation of EGFR in RPE-1 cells (Figures 7D–7E). This indicates that, as in Drosophila, autophagy can restrain EGFR/ERK signaling in human cells.

Given that genetic and epigenetic aberrancies that dysregulate epithelial homeostasis contribute to malignancy (Pesic and Greten, 2016), we examined mutation data of 15 human endocytosis/autophagy network genes orthologous to the genes we characterized in Drosophila (Table S1). Analysis of The Cancer Genome Atlas (TCGA) datasets by cancer type indicated that the colorectal adenocarcinoma and other cancers harbor a high frequency of mutations and deep deletions in these genes (Figure 7F, Table S2). From a colorectal adenocarcinoma (DFCI) gene set we found that, ULK1, the human ortholog of Atg1, and SNX18 and SNX33 were the most frequently mutated endocytosis/autophagy genes in CRC (Figure 7G, Table S3). This is consistent with the proposition that CRCs harbor frameshift mutations in autophagy genes (Choi et al., 2017), and suggests that the endocytosis/autophagy network may play a critical role in human colorectal carcinogenesis. We next assessed the degree to which somatic mutations in the endocytosis/autophagy genes are related to other genomic and epigenomic changes in human CRC. Classification according to microsatellite instability demonstrated that mutations in the endocytosis/autophagy network were significantly enriched among MSI-H CRC samples (Figure 7H, Table S4). These samples also showed a strong association between endocytosis/autophagy pathway mutations and CIMP-H status, an epigenomic CRC phenotype that underlies sporadic microsatellite instability (Figure 7I, Table S5) (Issa, 2004; Weisenberger et al., 2006). Finally, we found that mutations in the three human orthologs of Drosophila SH3PX1 (SNX9/18/33) have a negative association with KRAS activating mutations in CRCs (Figure 7J, Table S6). This suggests an alternative route to RAS-ERK activation that may underlie novel CRC diagnostic and therapeutic strategies. We suggest that dysfunction in the endocytosis/autophagy network, acting at least partially through SNX-dependent autophagy and activated KRAS signaling, may promote human colorectal carcinogenesis.

DISCUSSION

Exocytic and endocytic pathways are known to play crucial roles in signal transduction during normal and pathophysiological processes (Gissen and Maher, 2007), but how these pathways impact the behavior of stem cells or epithelial homeostasis has not been extensively explored. In Drosophila, just one relevant study shows that the vesicle-mediated COPI-Arf79F complex is required for intestinal stem cell (ISC) survival (Singh et al., 2016). Here, we identify a novel mechanism by which SH3PX1-dependent vesicle trafficking regulates EGFR/Ras/MAPK activity and ISC behavior. A large set of endocytosis and autophagy genes act to contain EGFR signaling in multiple ways. These include de-stabilizing the EGFR, limiting ligand (Krn) function, and dampening cytosolic Ca²⁺ signaling, all of which act to restrain ISC proliferation. These mechanisms are of such importance in ISCs that blockages in endocytic flux drastically upregulate basal EGFR/Ras/MAPK activity, promoting ISC proliferation, gut epithelia hyperplasia, and eventual death of the organism. While published data show that the endocytosis factors Dynamin (Vieira et al., 1996), Rab GTPases (Ceresa, 2006), and SNX1, 9 and 16 (Choi et al., 2004; Kurten et al., 1996; Lin et al., 2002) mediate EGFR degradation, only one paper showed enhanced signaling activity (ERK phosphorylation) following endocytic flux inhibition (Vieira et al., 1996), and none of the relevant papers report effects on cell proliferation or stem cell activation, centrally important functional consequences. Further, to our knowledge, no previous reports attribute endocytosis-mediated degradation of EGFR to autophagy.

Autophagy has been viewed as a “double-edged sword” in cancer because it can be either tumor suppressive or oncogenic, depending upon cancer type and stage (White and DiPaola, 2009). In early carcinogenesis, as in our study, autophagy can be tumor suppressive by degrading oncogenic proteins and limiting genotoxic stresses that promote cancer progression (Kondo et al., 2005). In the later stages of carcinogenesis, however, autophagy can act oncogenically by providing recycled nutrients to sustain cancer metabolism (Kondo et al., 2005). Recent studies in the mouse intestine report that autophagy genes (ATG5, ATG7 and ATG16L1) are necessary for ISC maintenance and CRC progression (Asano et al., 2017; Levy et al., 2015; Matsuzawa-Ishimoto et al., 2017), but these studies used the pan-iEC recombinase, Villin-Cre, making it impossible to distinguish ISC-autonomous requirements from functions in other epithelial cell types such as enterocytes. In contrast, our study in the fly gut shows that ISC-autonomous defects in autophagy can release stem cell proliferation without regulating differentiation, potentially increasing cancer risk. Further, as in other studies of the fly intestine, we found that hyperproliferative stem cells can stress the surrounding epithelium and induce widespread mitogenic signaling, which in turn catalyzes a feed-forward loop that drives massive hyperplasia of the mutant stem cells. The model we propose here places the SH3PX1-dependent endocytosis/autophagy network at the nexus of this feed-forward loop, a situation that may also play out in human cancers.

Previous literature addressing autophagy and Drosophila’s ISCs is sparse, and offers divergent conclusions. One study (Nagy et al., 2016) reported that UVRAG, a known regulator of autophagy, restricts ISC proliferation independent of autophagy. They also published a follow-up study showing that knockdown of other Atg genes adversely affected ISC maintenance (Nagy et al., 2018). That study, however, did not test the majority of Atg genes investigated here, and moreover our data are consistent with another report (Zeng et al., 2015) showing that loss of Atg2 or Atg6 promotes ISC hyperproliferation. Consistently, Thapsigargin, a drug that arrests autophagy by blocking autophagosome-lysosome fusions (Ganley et al., 2011), also stimulates ISC mitosis (Deng et al., 2015). Taken together, Nagy et al’s findings and our contrasting results suggest that the specific Atg gene that is inhibited may differentially affect proliferation vs. stem cell maintenance. Indeed autophagy is known to employ different genetic components in different cell types (Chang et al., 2013; Wen et al., 2017), and some Atg genes have autophagy-independent functions (Tang et al., 2013).

Elucidating the specific cargos of the SH3PX1-dependent network that mediate its effects on ISCs proved to be challenging. Using genetic tests we were able to rule out several mitogenic signaling pathways as secondary and involved non-cell autonomously, and to identify EGFR/ERK signaling as the primary, cell autonomous trigger for SH3PX1^--dependent ISC proliferation. The importance of this pathway is expected, as previous studies show that EGFR/ERK signaling is both universally required and sufficient to activate Drosophila ISCs for growth (Jiang et al., 2011; Jin et al., 2015; Xiang et al., 2017; Xu et al., 2017). Of note, our genetic tests showed that RNAi against Rab11, a small GTPase critical for recycling endosome function, dominantly suppressed ISC mitoses induced by any other endocytosis/autophagy network defect. Further, we found that SH3PX1⁻-dependent ISC proliferation requires the EGFR ligand, Krn, and the Rho intramembrane protease that cleaves and activates it. This suggests that SH3PX1 loss drives the accumulation of activated, Krn-bound EGFR complexes in Rab11 recycling endosomes and/or at the plasma membrane, thereby constitutively activating downstream signaling. This conclusion is supported by reports that clathrin-mediated endocytosis and the endosomal localization of EGFR sustain pathway activation (Sigismund et al., 2008; Sorkin and Von Zastrow, 2002). Unfortunately, due to a lack of reagents, we were unable to test for increased co-localization of Krn and EGFR in Rab11 endosomes. Future studies detailing how defects in endocytic flux affect the distribution and accumulation of EGFR pathway components in the various endosomal compartments should extend the general model we propose here.

SNX, RAB and ATG proteins are highly conserved, yet the genomic status of these genes in human disease is largely unexplored. Using TCGA data, we uncovered a strong association between colorectal tumors with somatic mutations in ATG and SNX genes, and MSI-H and CIMP-H status. Consistent with this, the DNA damage-activated kinase Chk2 (Checkpoint kinase 2) is upregulated by Atg6 knockdown (Nagy et al., 2018), suggesting that dysfunctions of autophagy might promote genomic instability. In addition, one previous study found that frameshift mutations in nucleotide repeats in human ATG5, ATB9B, and ATG12 were highly enriched in MSI-H CRCs (Kang et al., 2009), and another came to a similar conclusion regarding ULK1 (ATG1A) and ULK2 (ATG1B) (Choi et al., 2017). These reports suggest that aberrancies in ATG and SNX genes are linked to the distinct MSI-H/CIMP-H molecular phenotype found in a subset of CRCs. The association with MSI-H may simply reflect the fact that all of the ATG and SNX genes that show the association (Figure 7H) contain long (>7 nucleotide) mono-nucleotide repeats in their open reading frames and become very mutable when defects in the DNA mismatch repair arise, as is the case in MSI-H CRC. The association could also reflect more complex unknown functional interactions between endocytosis/autophagy and mismatch repair genes. Either way, the differential response of MSI-H CRCs to chemotherapeutic regimes illuminates an issue that could form the basis for novel cancer diagnostics or therapies involving autophagy (Vilar and Gruber, 2010).

The molecular phenotyping of CRCs for CIMP status is also an integral prognostic classifier in patients (Toyota et al., 1999). Transcriptional inactivation by cytosine methylation at promoter CpG islands in tumor suppressor genes is believed to be an important mechanism in human colorectal carcinogenesis (Ogino and Goel, 2008), and CIMP-H tumors represent a distinct etiologic and clinical CRC phenotype (Issa, 2004). Although no studies to date have examined CpG island methylation at promoter regions of the endocytosis and autophagy genes in CRC, the downregulation of ULK1 and ULK2 via promoter hypermethylation is critical to inhibit autophagy and promote tumor development in CIMP-H glioblastoma (Shukla et al., 2014). Our findings suggest that this association may apply in CRC and other cancers.

EGFR is overexpressed in 49–82% of CRCs (Antonacopoulou et al., 2008), and in cancers with EGFR activation, EGFR-specific antibodies such as cetuximab can suppress cell growth by blocking ligand-induced receptor signaling (Jonker et al., 2007). However, this efficacy is limited (Pozzi et al., 2016) and the activation mechanisms of EGFR signaling are only understood in a subset of cancers, most notably those with activating KRAS or BRAF mutations. Our findings suggest that EGFR/ERK signaling may also be activated by loss-of-function in SNX, RAB and ATG genes. We uncovered a negative association between mutations in the human SH3PX1 homologues SNX9/18/33 and activating mutations in KRAS in CRC patients, suggesting that this alternative route to ERK activation may be at play in some tumors. This and other studies addressing how EGFR signaling is activated in cancer stem cells that lack KRAS and BRAF mutations may help develop new strategies for suppressing ERK signaling in CRC and other EGFR-dependent cancers.

STAR METHODS

CONTACT FOR REAGENT AND RESOURCE SHARING

Requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Bruce A. Edgar (bruce.edgar@hci.utah.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

All fly strains were kept on standard fly medium at 25°C. For temperature shift experiments, flies were raised at 18°C prior to shifting to the temperature conditions described in the corresponding figure legends and method details. Midguts were dissected from adult females, except in Figure 4D.

Human SNX transgenic flies

The cDNA of hSNX9, hSNX18 and hSNX33 were cloned into the pUAST-attB-V5 vector, respectively. UAS-hSNX9, UAS-hSNX18 and UAS-hSNX33 were generated using PhiC31 integrase-mediated site-specific transgenesis system.

Fly Diet

For 1L: 18g active dry yeast, 61g corn meal, 15ml syrup, 81.5g malt and 11.2g agar. Ingredients were mixed well with water and brought to boil in kettle. After cooling, 7.5ml of 99.5% propionic acid (Sigma-Aldrich) and 10ml Tegosept (methyl 4-hydroxybenzoate in 95% ethanol at 100g/L, Sigma-Aldrich) were added to the mixture and the solution was dispensed into vials/bottles.

Cell culture

Human RPE-1 (female) and CaCo-2 (male) cell lines were cultured in DMEM/F12 (Gibco) supplemented with 10% FBS (Gibco) and 100U/ml penicillin/streptomycin (Gibco) in a sterile humidified 37°C incubator with 5% CO₂ atmosphere.

Bacteria culture

Pseudomonas entomophila (P.e., gift from Nicolas Buchon), a gram-negative bacterium that naturally infects Drosophila (Vodovar et al., 2005), was grown in LB medium (400μL P.e. glycerol stock in 200mL LB) containing Rifampicin (100μg/mL) at 29°C with shaking at 130rpm for 48hr. The culture was centrifuged at 2500g for 25min at 4°C and the resultant pellet was resuspended in 8mL 5% sucrose for fly oral infection.

METHOD DETAILS

Bacterial infection

For oral bacterial infection, 2–3 day-old adult flies were shifted from 18°C to 29°C and raised for 2d on normal fly food before infection. Flies were then transferred to empty vials containing Whatman paper discs soaked with an infection solution (P.e. pellet resuspended with 5% sucrose) or a control solution (5% sucrose). Yeast paste made with the above control or infection solution was also provided on the wall of the vials. Flies were orally infected for 16hr at 29°C and then subjected for dissection.

Immunostaining

After dissection, samples were fixed in PBS with 4% paraformaldehyde for 30min, washed in PBS with 0.1% Triton X-100, and blocked in PBS with 0.1% Triton X-100 and 10% NGS for at least 30 min at room temperature. For the dpERK and EGFR staining, to ensure consistent staining, the midgut samples were first fixed in 8% paraformaldehyde, followed by a critical methanol (−20°C, 10min) fixation step. All samples were then stained with primary antibodies at 4°C overnight with the following dilutions: chicken α-GFP (Thermo Fisher Scientific, 1:1000), rabbit α-DsRed (Clontech Laboratories, 1:500), rabbit α-PH3 (Millipore, 1:1000), mouse α-PH3 (Cell Signaling Technology, 1:300), rabbit α-phospho-p44/42 MAPK (Erk1/2) (Cell Signaling Technology, 1:200), mouse α-β-gal (Promega, 1:500), mouse α-EGFR (Sigma-Aldrich, 1:100), rabbit α-SH3PX1 (gift from Jack Dixon, 1:300), mouse α-Prospero (DHSB, 1:50), rabbit α-Pdm1 (gift from Xiaohang Yang, 1:200).

DAPI (Thermo Fisher Scientific, 1:1000) was used to label nuclei. Staining was detected by Alexa Fluor 488, 568, or 633 conjugated species appropriate secondary antibodies (Thermo Fisher Scientific, 1:1000).

Western blot

Cell lysates were prepared in 2× SDS Loading Buffer (80 mM Tris pH6.8, 2% SDS, 12% glycerol, 10% β-mercaptoethanol, bromophenol blue). For the western blot shown in Figure S1B, three adult flies per sample were homogenized by grinding in 200 ul 2× SDS Loading Buffer. Lysates were boiled and clarified by centrifugation before analysis by western blotting. Primary antibodies used for western blotting were: rabbit anti-SH3PX1 (gift from Jack Dixon, 1:1500), mouse anti-α-tubulin (DSHB 12G10, 1:10,000), mouse anti-p44/42 MAPK (Erk1/2) (Cell Signaling Technology, 1:1500), rabbit anti-phosphop44/42 MAPK (Erk1/2) (Cell Signaling Technology, 1:3000), mouse anti-vinculin (Sigma-Aldrich, 1:5000), rabbit anti-EGFR (Cell Signaling Technology, 1:2000).

Live-imaging and cytosolic Ca²⁺ measurements

Live-imaging experiments were performed using 3–5d old flies with the following genotypes: esg^ts,UAS-GCaMP6s>w1118 and esg^ts,UAS-GCaMP6s>SH3PX1^RNAi-1. Flies were raised at 18°C. Adult females were shifted to 29°C after eclosion for 4d before dissection in adult hemolymph-like (AHL) saline (108 mM NaCl, 25 mM KCl, 2 mM CaCl₂, 8.2 mM MgCl₂, 4 mM NaHCO₃, 1 mM NaH₂PO₄, 5 mM trehalose, 10 mM sucrose, and 5 mM HEPES pH 7.5). Three explanted intact midguts were placed in a clear bottom plastic well from a 4-well chambered slide (IBIDI, Cat#80426), carefully oriented with forceps, and gently covered with a piece of mixed cellulose esters membrane (Millipore, AABP02500) to keep them in place. The well was then filled with 0.5 ml of AHL containing 10 μg/mL Isradipine so as to minimize visceral muscle movement. For each condition, 2 replicates were produced, each consisting in the measurement of 3 midguts. Each midgut was imaged separately using a Nikon (Eclipse Ti) widefield microscope or a Leica SP8 confocal microscope with comparable results. A Z-stack encompassing a 60–100μm section of the posterior midgut with a Z-step of 10 μm was captured every 10” for a total live-imaging duration of 10’. Images were processed in ImageJ using a custom macro. Briefly, movies were stabilized using the StackReg plugin so as to compensate for residual midgut movements, and a maximum intensity projection was produced which was used for the subsequent analysis. Cells expressing GCaMP6s that were clearly visible for the whole imaging duration were then selected, and their mean fluorescence intensity was manually calculated. As differences in depth within the tissue were found to influence the observed fluorescence intensity for each cell, their mean values were normalized by setting the minimum and maximum fluorescence values to 0 and 10, respectively, while any value in between was approximated to an integer. This allowed us to define a fixed set of rules for easier peak identification from GCaMP6s mean fluorescence intensity data plotted against time. Peak durations were then estimated from the width of each peak at its base, and the comparison of the two sample types was performed in R with the Mann–Whitney two-tailed U test.

Ca²⁺ peak definitions for live-imaging experiments

Normalized GCaMP6s mean fluorescence intensity levels were plotted against time and used for peak identification. Original live-imaging data and maximum intensity projections were referenced to define the following rules and to randomly check the fidelity of the peaks identification. Fluorescence values 0 and 1 were defined as baseline, while 4 was set as the minimum value a peak has to reach to be defined as such. Peak starts were defined by a steady increase in fluorescence intensity from a baseline value. Alternatively, if a peak starts during the declining phase of a previous one, then the local minimum between the two peaks is set as the end of the first and the beginning of the second. Moreover, in order to be clearly visible when separating two distinct peaks, such local minimum must have a fluorescence value at least 3 points below the maximum point reached by the first peak. Lastly, peak ends were set as the lowest point reached during the declining phase of a peak (e.g., either a baseline value or a local minimum). Since each time-point represents a 10” interval, peak durations were calculated as the difference between the time-points marking each peak’s end and start.

Bioinformatics analysis of cancer patient data

Mutation profile datasets for all human cancer studies used in this research were queried using the cBio Cancer Genomics Portal (cBioPortal version 2.0.1; accessed February 15, 2019). Genomic (whole exome sequencing) and clinicopathologic datasets of 619 colorectal adenocarcinoma samples (DFCI) (Giannakis et al., 2016) from The Cancer Genome Atlas (TCGA) (Cerami et al., 2012; Gao et al., 2013) were downloaded from cBioPortal. Differences in clinicopathologic characteristics (e.g., microsatellite instability, CIMP status, KRAS) by mutation status were examined by chi-square tests. Analyses of TCGA data were conducted in SAS version 9.4 statistical software (SAS Institute; Cary, NC). All statistical tests were two-sided, with P<0.05 considered to be statistically significant.

Statistical analysis

Statistical analyses were performed using the Graphpad Prism 7 software package. Statistical significance (P values) of experiments were calculated by unpaired two-tailed Student’s t-test. Statistical significance was denoted as follows: non-significant (ns) P>0.05, *P<0.05, **P<0.01, ***P<0.001 and ****P<0.0001.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Chicken α-GFP	Thermo Fisher Scientific	Cat# A10262; RRID: AB_2534023
Rabbit α-DsRed	Clontech Laboratories, Inc.	Cat# 632496; RRID: AB_10013483
Rabbit α-phospho-Histone 3	Millipore	Cat# 06–570; RRID: AB_310177
Mouse α-phospho-Histone 3	Cell Signaling Technology	Cat# 9706; RRID: AB_331748
Rabbit α-phospho-p44/42 MAPK (Erk1/2)	Cell Signaling Technology	Cat# 4370; RRID: AB_2315112
Mouse α-p44/42 MAPK (Erk1/2)	Cell Signaling Technology	Cat# 9107; RRID: AB_10695739
Mouse α-β-galactosidase	Promega	Cat# Z3781; RRID: AB_430877
Mouse α-dEGFR	Sigma-Aldrich	Cat# E2906; RRID: AB_609900
Rabbit α-EGFR	Cell Signaling Technology	Cat# 4267; RRID: AB_2246311
Rabbit α-SH3PX1	Jack Dixon (UCSD, USA)	N/A
Mouse α-Prospero	DHSB	Cat# Prospero (MR1A); RRID: AB_528440
Rabbit α-Pdm1	Xiaohang Yang (ZJU, China)	N/A
Mouse α-alpha-tubulin	DHSB	Cat# 12G10; RRID: AB_1157911
Mouse α-vinculin	Sigma-Aldrich	Cat# V9131; RRID: AB_477629
Bacterial and Virus Strains
Pseudomonas entomophila (P.e.)	Nicolas Buchon (Cornell University)	N/A
Biological Samples
N/A	N/A	N/A
Chemicals, Peptides, and Recombinant Proteins
DAPI	Thermo Fisher Scientific	Cat# D1306; RRID: AB_2629482
Critical Commercial Assays
N/A	N/A	N/A
Deposited Data
N/A	N/A	N/A
Experimental Models: Cell Lines
N/A	N/A	N/A
Experimental Models: Organisms/Strains
Drosophila: snx1d1	(Zhang et al., 2011)	CG2774
Drosophila: snx3d1	(Zhang et al., 2011	CG6359
Drosophila: snx6d97	(Zhang et al., 2011)	CG8282
Drosophila: SH3PX1d1	(Zhang et al., 2011)	CG6757
Drosophila: SH3PX1HK62b	(Knaevelsrud et al., 2013)	CG6757
Drosophila: snx16PB	(Rodal et al., 2011)	CG6410
Drosophila: snx16Δ2	(Rodal et al., 2011)	CG6410
Drosophila: snx17d1	(Zhang et al., 2011)	CG5734
Drosophila: snx21d1	(Zhang et al., 2011)	CG3077
Drosophila: snzd1	(Zhang et al., 2011)	CG1514
Drosophila: snx27d2	(Zhang et al., 2011)	CG32758
Drosophila: esg-Gal4, UAS-2xEYFP/CyO; Su(H)GBE-Gal80, tubGal80ts/TM3	Heinrich Jasper (Genentech)	N/A
Drosophila: esg-Gal4, UAS-GFP/CyO; tubGal80ts/TM6B	(Jiang et al., 2009)	N/A
Drosophila: esg-Gal4, UAS-GFP, tubGal80ts/CyO	Bruce A. Edgar (HCI, USA)	N/A
Drosophila: tubGal80ts/FM7; esg-Gal4/CyO	Bruce A. Edgar (HCI, USA)	N/A
Drosophila: Su(H)GBE-Gal4, UAS-GFP/CyO; Ubi-Gal80ts/TM6B	(Xiang et al., 2017)	N/A
Drosophila: Myo1A-Gal4, tubGal80ts, UAS-GFP/CyO	(Jiang et al., 2009)	N/A
Drosophila: tub-Gal80ts, UAS-GFP/CyO; prosV1-Gal4/TM6B	This paper	N/A
Drosophila: esg-Gal4, tub-Gal80ts, UAS-GFP/CyO; UAS-flp, act>CD2>Gal4/TM6B	(Jiang et al., 2009)	N/A
Drosophila: hs-flp; act-Gal4, UAS-mRFP; tub-Gal80 FRT2A/TM6B	Bruce A. Edgar (HCI, USA)	N/A
Drosophila: hs-flp, UAS-GFP, act-Gal4; ; FRT2A UAS-Gal80/TM6	Bruce A. Edgar (HCI, USA)	N/A
Drosophila: enGal4	(Zhang et al., 2017)	N/A
Drosophila: mira-GFP	(Bardin et al., 2010)	N/A
Drosophila: Myo1A-Gal4, mira-GFP/CyO; MKRS/TM6B	This paper	N/A
Drosophila: ex-lacZ	(Zhang et al., 2017)	N/A
Drosophila: upd3.1-lacZ	(Jiang et al., 2011)	N/A
Drosophila: rho-lacZ	(Jiang et al., 2011)	N/A
Drosophila: UASp-GFP-mCherry-atg8a	BDSC	37750
Drosophila: UAS-SH3PX1RNAi-1	BDSC	54833
Drosophila: UAS-SH3PX1RNAi-2	BDSC	27653
Drosophila: UAS-SH3PX1EY08084	BDSC	17428
Drosophila: UAS-GFP-SH3PX1	Helene Knævelsrud (Oslo University Hospital)	N/A
Drosophila: UAS-Atg1RNAi	VDRC	16133
Drosophila: UAS-Atg5RNAi	BDSC	34899
Drosophila: UAS-Atg6RNAi	BDSC	35741
Drosophila: UAS-Atg7RNAi	BDSC	34369
Drosophila: UAS-Atg8aRNAi	BDSC	34340
Drosophila: UAS-Atg9RNAi	BDSC	34901
Drosophila: UAS-Atg12RNAi	BDSC	34675
Drosophila: UAS-Atg16RNAi	BDSC	34358
Drosophila: UAS-Syx17RNAi	BDSC	25896
Drosophila: UAS-Rab4SN	(Strutt et al., 2011)	N/A
Drosophila: UAS-Rab5SN	(Strutt et al., 2011)	N/A
Drosophila: UAS-Rab7RNAi	VDRC	40337
Drosophila: UAS-Rab11RNAi	BDSC	27730
Drosophila: UAS-shiTS	BDSC	44222
Drosophila: UAS-InRK1409A	BDSC	8253
Drosophila: UAS-DomeRNAi	BDSC	34618
Drosophila: UAS-BskRNAi	VDRC	34138
Drosophila: UAS-PanDN	BDSC	4785
Drosophila: UAS-EgfrRNAi	BDSC	25781
Drosophila: UAS-RasRNAi	BDSC	29319
Drosophila: UAS-MEKRNAi	VDRC	107276
Drosophila: UAS-pntRNAi	Bruce A. Edgar (HCI, USA)	N/A
Drosophila: UAS-Ets21CRNAi	VDRC	106153
Drosophila: UAS-PvrRNAi	Utpal Banerjee (UCLA, USA)	N/A
Drosophila: UAS-btlRNAi	Bruce A. Edgar (HCI, USA)	N/A
Drosophila: UAS-upd3RNAi	VDRC	106869
Drosophila: upd2Δ, upd3Δ	BDSC	55729
Drosophila: UAS-rhoRNAi-1	BDSC	41699
Drosophila: UAS-rhoRNAi-2	BDSC	28690
Drosophila: UAS-KrnRNAi	NIG	8056R-3
Drosophila: Krn27–7-4	(Jiang et al., 2011)	N/A
Drosophila: UAS-spiRNAi	BDSC	28387
Drosophila: UAS-stgRNAi	NIG	1395R-1
Drosophila: UAS-Egfrwt	(Xiang et al., 2017)	N/A
Drosophila: UAS-xbp1-EGFP	BDSC	39720
Drosophila: UAS-GCaMP6s	BDSC	42749
Drosophila: UAS-TrpA1RNAi	BDSC	31504
Drosophila: UAS-RyRRNAi	BDSC	29445
Drosophila: UAS-V5-hSNX9	This paper	N/A
Drosophila: UAS-V5-hSNX18	This paper	N/A
Drosophila: UAS-V5-hSNX33	This paper	N/A
Oligonucleotides
N/A	N/A	N/A
Recombinant DNA
cDNA: Human SNX9	GenScript USA Inc.	Clone ID: OHu12143; RefSeq Accession: NM_016224.4
cDNA: Human SNX18	GenScript USA Inc.	Clone ID: OHu08909; RefSeq Accession: NM_052870.2
cDNA: Human SNX33	GenScript USA Inc.	Clone ID: OHu30854; RefSeq Accession: NM_153271.1
Plasmid: pUASTattB-V5-hSNX9	This paper	N/A
Plasmid: pUASTattB-V5-hSNX18	This paper	N/A
Plasmid: pUASTattB-V5-hSNX33	This paper	N/A
Software and Algorithms
Prism 7	GraphPad Software	RRID: SCR_002798
ImageJ	https://fiji.sc	N/A
R package	https://www.r-project.org	N/A
SAS version 9.4 statistical software	SAS Institute; Cary, NC	N/A
cBioPortal	http://www.cbioportal.org	N/A
The Cancer Genome Atlas (TCGA)	https://cancergenome.nih.gov	N/A
Other
N/A	N/A	N/A

Highlights.

• Ectoderm-derived metaphocytes in zebrafish epidermis are myeloid-like cells

• Unlike Langerhans cells, highly mobile metaphocytes lack phagocytosis ability

• Metaphocytes capture external soluble antigens through transepithelial protrusions

• Antigen transfer from metaphocytes to Langerhans cells via apoptosis-phagocytosis