Recycling Endosomes in Mature Epithelia Restrain Tumorigenic Signaling

The initiation and progression of the vast majority of epithelial neoplasia involve expansion of tumor cell mass and concurrent dysplastic organization, exemplified by loss of cell polarity, of the neoplastic glands. Typically, dysplasia precedes development of advanced invasive carcinoma in most epithelial organs, including the colon (1–3). Whether the structural changes in tissue morphology solely reflect disease progression initiated by oncogenic drivers remains a matter for debate. To date, there is a scarcity of available data examining whether a disruption of cell polarity cues, not linked to known oncogenic drivers, may also constitute an unrecognized pathway to neoplastic development. Mounting cancer research literature has focused on cellular processes directly related to cell proliferation and apoptosis. However, research linking impaired intracellular trafficking governing epithelial polarity to tumor initiation/progression remains underexplored (4).

Apical-basolateral cell polarity manifested by columnar epithelial organization is maintained by membrane trafficking regulators, in particular the Rab small GTPases. Among the diverse intracellular trafficking organelles, recycling endosomes represents a hallmark structure in mature glandular epithelial cells (5). The recycling endosome compartment is composed of a heterogeneous collection of tubule-vesicular membrane compartments that sort and deliver plasma membrane constituents. The evolutionarily conserved Rab11 family of small GTPases are dedicated recycling endosome resident proteins that execute the canonical sorting functions of this compartment. Drosophila has a single Rab11 gene and zygotic homozygous Rab11 mutants are lethal (6). In Drosophila egg chamber, Rab11 maintains junctions between germ and niche cap cells (7), and appears to impact cellular turnover and differentiation (8). Humans and mice have 3 Rab11 family members, Rab11a, Rab11b, and Rab25. In developing mouse gut, differentiated intestinal epithelial cells (IEC) express high levels of Rab11a, whereas epithelia with an impaired apical-basolateral polarity showed a reduction in Rab11a expression (9). Rab11a^−/− mice die in early fetal stage (10). In both Drosophila midgut and mouse intestine, epithelial cell-specific ablation of Rab11 (Rab11a in mice) caused epithelial hyperplasia accompanied by an elevated production of proinflammatory cytokines (11). Loss of Rab11a in IECs disturbs apical membrane morphogenesis and apical transport of cargo essential for mature enterocyte's structure and function (12, 13). In addition, when Rab11a was acutely removed from mouse IECs via CreER, a rapid induction of abnormal apical membrane structures was ectopically formed along the basolateral domain (13). Thus, genetic, developmental, and cell biological studies demonstrated the existence of an evolutionarily conserved function of Rab11 in maintaining intestinal epithelial integrity and polarity in vivo. The consequences of Rab11 endosome functionality on cellular growth and long-term gut homeostasis have not been defined.

Here, we show that RAB11 is reduced in a majority of colorectal cancer patient tissue samples, and Rab11a protein deficiency in mice causes early onset of epithelial dysplasia. Using genetic approaches across species, we demonstrate that the integrity of recycling endosome function in mature gut epithelia can counterbalance tumorigenic growth elicited by Ras or APC mutations. Alteration of recycling endosome function synergizes with oncogenic pathways and accelerates carcinogenesis induced by chemical carcinogen, genetic mutation, or aging. Additionally, loss of Rab11 immediately expands a stem cell pool and cells autonomously activate Yki/Yap. We found that Rab11 endosomes maintain the integrity of LATS kinase activity in part through delivery and assembly of a junctional complex. Rab11-deficient tumor cells produce high levels of inflammatory and growth signaling, whereas suppressing either Yki/Yap or upd3/IL6 reduces epithelial dysplasia and hyperplasia. We propose that an important and novel function of Rab11 endosomes in mature epithelia is to confine tumorigenic signaling to appropriate homeostatic levels in order to maintain normal tissue growth and turnover.

Rab11a^flox, APC^Min, Villin-Cre, and Villin-CreER mice have been described previously (please see key resource table in Supplementary Information). A new mouse Rab11b knockout allele was generated by single guide RNA and Cas9 mRNA coinjection into murine zygotes (detailed in Supplementary Methods). All procedures carried out have been approved by Rutgers Institutional Animal Care and Use Committee community.

Fly stocks were maintained at room temperature (approximately 22°C) in yeast extract/cornmeal/molasses/agar food medium. w¹¹¹⁸, Myo1A-Gal4^ts >UAS-GFP, or Myo1A-Gal4^ts >UAS-GFP X w¹¹¹⁸ flies were used as wild-type control in different experiments. Transgenic and RNAi fly stocks were obtained from Vienna Drosophila Research Center (VDRC, labeled as V) or Bloomington Drosophila Stock Center (labeled as B), or otherwise as cited (14–18). The stocks include Rab11^S25N dominant negative (B9792), Rab11 RNAi (RNAi1: V108382, RNAi2: B27730), UAS-EGFR^{Act (A887T)} (B9533), UAS-Ras^v12 (B64195), msn TRiP RNAi (B28791), UAS-Wg-HA (B108487), APC RNAi (RNAi1:V110290, RNAi2: V51468), Yki RNAi (B31965), YkiN+C RNAi (gift from J. Jiang, UT Southwestern, Dallas, TX), upd3 mutant (17), UAS-Yki-act (18). For transgenic expression of UAS-dsRNA or UAS-cDNA constructs, newly enclosed flies were kept in vials with regular food medium at room temperature until 5 to 8 days old, and were shifted to 29°C for 1 to 5 days as indicated in the individual figures for optimal results to inactivate the tubulin-Gal80^ts repressor and to enable Gal4 dependent UAS-dsRNA or transgene expression. Female flies were used for tissue dissection for various assays including immunofluorescence staining and RNA isolation for qPCR.

The following data of all colorectal cancer cases in The Cancer Genome Atlas (TCGA) using cBioportal were extracted: RAB11A mRNA levels, intervals between the times of last follow-up and diagnosis, statuses of all-cause death (for overall survival) and disease-free (for disease-free survival), tumor stage, and patient sample ID. The patients who were deemed lost to the follow-up was censored in the survival analyses, which were conducted using the Cox regression model of the Stata (version 14, Stata Corp.) and adjusted for tumor stage. We conducted a step-wise search for the best cutoff of RAB11A mRNA level ((RNA Seq V2 RSEM), using 5 and 10 percentiles. The 5-percentile was chosen for its smallest P value (the most statistical significance). We compared cases of the low mRNA expression level (the lower 5 percentile, N = 16) to those (between 5 and 95 percentiles) as reference (N = 298), and those of the highest 5 percentile (N = 17). The regression data shown were those of multivariate Cox regression, including age, tumor stage, and sex as the covariates. A P < 0.05 was considered statistically significant.

Human tissue microarray for colon cancer BC051110a comprised of 110 human colon cancer tissue core samples (stage I: 4, stage II: 71, stage III: 28, and stage IV: 7; grade I: 41, grade II: 46, and grade III: 23) and 10 human normal colon core samples was purchased from US BioMax, Inc. The tissue microarray was processed for IHC analysis using a primary rabbit anti-Rab11a antibody (1:500; US Biologicals, R0009). IHC signals of RAB11A were quantified using the NIH ImageJ with an IHC toolbox package plugin designed for DAB-chromogenic IHC. Three colonic epithelial glands (normal or tumorous) from each of the 120-sample array were manually circled as Region of Interest and individually measured for DAB signals. An average value for each sample was then obtained. After ranking, the sample with the medium value was identified and used to calculate the fold changes for the remaining samples. A graph plotting the fold changes according the medium value (log₁₀) was generated. A Mann–Whitney U test showed significant difference (P < 0.001) between the normal and colon cancer samples. For RAB11A and YAP IHC on colon cancer sections, serial sections (thickness of 4 micrometers) of 13 colon tumor samples (3 tubulovillous adenoma, one stage I, 2 stage II, 5 stage III, and 2 stage IV colorectal cancer) with adjacent normal colon mucosa were selected from the tissue bank of University Hospital at Princeton, New Jersey.

At 3 months of age, Rab11a^FL/FL (no CreER), Rab11a^FL/FL; Villin-CreER, Rab11a^FL/FL; Rab11b^−/−; Villin-CreER, and Rab11b^−/− mice (n = 3 for each genotype) were intraperitoneally injected with 200 μL of tamoxifen (20 mg/mL) dissolved in corn oil. Two days after tamoxifen injection, mouse jejunum was collected and immediately snap frozen in liquid nitrogen. 2 to 3 mm of frozen jejunum was then processed for RNA isolation. Tissues were immersed in 800 μL of Buffer RLT (provided by Qiangen RNeasy Plus Mini Kit; 74134) with 10 μmol/L β-mercaptoethanol, homogenized by passing tissue/lysate through a 20-gauge needle (BD Biosciences; 305176) for 20 times on ice. The remaining isolation procedure was completed as outlined in the Qiagen RNeasy mini protocol. RNA was eluted from spin column using 30 μL of RNase-free water. Quality of RNA was assessed by a bioanalyzer.

RNA was sequenced using Illumina's TruSeq RNA Library Prep Kit v2. RNA reads were processed using HISAT2 (v2.1.0) using the Rutgers University Amarel High Performance Computing Cluster. Aligned files were run together using StringTie (v1.3.4) and analyzed using DESeq2 (Release 3.7) to normalize fragments per million (FPM) values using default settings and expression was visualized using heatmaps. For differential gene expression analysis, genes were preranked based on signed P value, where sign is based on logFC from DESeq2. Preranked gene lists were then analyzed against select gene lists of interest to determine significance of differential expression between conditions using gene set enrichment analysis (GSEA). One thousand permutations were performed for each gene list tested, and normalized enrichment scores (NES) and P values are reported for Lgr5 intestinal stem cell genes, and differentiation genes signatures based on Gene Ontology Terms.

Total RNA was extracted from 20 guts of flies from Myo1A^ts >GFP control and >Rab11 RNAi (V108382), respectively. The flies were shifted to 29°C for 5 days before dissection. Three independent biological samples were used for total RNA isolation. The RNA was sequenced using either the Applied Biosystems' Ion Torrent platform (PrimBio). Fastqc (version 0.10.1; http://www.bioinformatics.babraham.ac.uk/projects/fastqc) was used on the fastq sequences for all 6 samples (3 control and 3 Rab11 RNAi) to generate sequence quality reports. The raw sequences were mapped to the reference genome Drosophila melanogaster Ensembl BDGP5 using the aligner STAR (version 2.4.2a). HTseq-count (version 0.6.1) was used to count uniquely mapped reads at the gene level. DESEQ2 was used to perform statistical analysis to determine differentially expressed genes between the control samples and the knockdown samples. Expression of Rab11 and Upd3 were validated using the IGV tool (version 2.3.60). Differentially expressed genes are those with P_adj value < 0.05 (N = 2018). The differentially expressed genes were used for the pathway analysis using Enrichr. The RNA-sequencing (RNA-seq) data can be accessed from GEO (accession number GSE92705). The GSEA analysis on fly gut RNA-seq data was conducted using GSEA (v3.0: build 0160; Broad Institute, Massachusetts Institute of Technology, Cambridge, MA) as described before. The flybase.org was used to convert YAP-related mammalian genes to fly orthologs with a matching score greater than 3. The list of fly genes used for the analysis are below: unpg, ImpL2, vg, upd3, Cdc6, mys, tld, Actbeta, Gadd45, kibra, Det, cv-c, tj, tok, Cdk4, sd, Fs, maf-S, fz, CycD, Ephrin, phtf, dpp, TRAM, pk, chico, Lim3, ci, Ccn, esn. The following parameters were set for the GSEA, including 1,000 for number of permutations, phenotype for permutation type, weighted for enrichment statistic, Signal2Noise for Metric for ranking genes, real for gene list sorting mode, descending for gene list ordering mode, meandiv for normalization mode, no_balance for randomization mode, and timestamp for seed for permutation.

Statistical and graphic data analyses were conducted using Prism GraphPad 7.04 (https://www.graphpad.com) and Microsoft Excel 2016. The experimental results were reported from 3 to 6 mice of each genotype unless a specific number was described in the main text and figure legends. Bar graphs are presented as mean ± SEM. Box-and-whisker plots show mean and 10 to 90 percentile with outliners represented as dots. P values are labeled on individual graphs with P < 0.05 considered as statistically significant. See additional methods in Supplementary Information.

From birth, specific knockout of Rab11a in mouse IECs (Rab11a^ΔIEC) resulted in crypt hyperproliferation (Supplementary Fig. S1A) and low-grade dysplasia (Supplementary Fig. S1B and S1C) throughout the small intestine (also, see ref. 11). At 150 days, all Rab11a^ΔIEC mice exhibited irregular cribriform proliferation in villi (Supplementary Fig. S1D). At 1 year, the distal intestines of ∼30% of Rab11a^ΔIEC mice developed 2 to 3 well-demarcated tubular adenomas (box 1, Fig. 1A) and high-grade dysplasia (box 2, Fig. 1A) that exhibited complex glandular (cribriforming) structures, marked nuclear pleiomorphism, necrosis, and increased mitotic activities (Fig. 1A). Such dysplastic lesions were not observed in wild-type littermates, suggesting that Rab11a-deficient animals were predisposed to tumorigenesis during aging.

Rab11a-deficient IECs produced elevated levels of inflammatory cytokines (11) opening the possibility that an inflammatory microenvironment may participate in driving tumor progression. To address this question, the azoxymethane–dextran sulfate sodium (AOM–DSS) model for tumor induction by AOM after inflammatory response elicited by DSS (19) was adopted to study Rab11a^ΔIEC mice. As Rab11a^ΔIEC mice have higher baseline level of inflammatory cytokines compared with wild-type mice, a single dose of AOM (10 mg/kg body weight) was administered without exposing to DSS (Fig. 1B). After 8 weeks, AOM-treated Rab11a^ΔIEC mice showed a higher mortality rate (Fig. 1C), and surviving animals formed 2 to 6 poorly differentiated c-Myc⁺ and BrdUrd⁺ tubular adenomas in the ileum (Fig. 1D and E; Supplementary Fig. S2A and S2B). These tumors exhibited complex glandular (cribriforming) structures, marked nuclear pleiomorphism, necrosis, increased mitotic activities, and single infiltrating cells in mucosa; the tumors are best classified as intramucosal carcinoma (Supplementary Fig. S2B). By contrast, AOM treatment did not elicit any intestinal tumors in wild-type littermate mice (Fig. 1D). Thus, loss of Rab11a functionality and the associated elevated inflammatory response appeared to increase susceptibility to AOM-induced tumorigenesis.

We explored the consequence of loss of Rab11a on tumor progression in genetically susceptible mice expressing the APC^Min driver mutation by comparing Rab11a^ΔIEC; APC^Min mice and their APC^Min littermates having intact Rab11a alleles. Two-month-old Rab11a^ΔIEC; APC^Min mice developed twice the number of adenomas that were 5 times larger and occupied larger epithelial fields as compared with tumor fields in APC^Min (Fig. 1F–J; Supplementary Fig. S2D). Increased tumor burden was also reflected by a rapid decline of animal health that required termination of the experiments. Compared with larger APC^Min adenomas (Fig. 1K), pan cytokeratin staining revealed that Rab11a^ΔIEC; APC^Min mice had a higher frequency of, horizontally-expansile intramucosal adenomas with microinvasive features (arrow in Fig. 1L; Supplementary Fig. S2E).

Analysis of the Oncomine database uncovered a significant reduction of RAB11A mRNA level in human colon cancers across 19 independent databases (P = 2.7E−11, Fig. 2A). Similarly, normal colonic epithelia had strong and specific Rab11A apical and subapical cytoplasmic immunostained regions (Fig. 2B, lefts; Supplementary Fig. S3A), whereas the majority of colon cancers exhibited lower Rab11A levels (Fig. 2B, right). These data represent Rab11A levels across 110 human colon cancer tissue samples along with 10 normal colon tissues (Supplementary Fig. S3B). Fifteen percent of the carcinomas exhibited over 2-fold reduction in Rab11A level compared with the median sample (Fig. 2C) and 11 samples with the most reduced Rab11A levels were found to have stage II or more advanced carcinomas (Fig. 2C and D). Within the same tissue samples, areas of colonic adenocarcinoma (Supplementary Fig. S3D, see red dotted region) showed reduced apical Rab11A staining compared with the adjacent normal tissue with strong apical Rab11A localization (Supplementary Fig. S3D, see black lined region). Reduction of apical Rab11A was well correlated with epithelial dysplasia (Supplementary Fig. S3E), as even low-grade tumors displayed significant Rab11A reductions (n = 41, P < 0.01; Supplementary Fig. S3F). Ten adenocarcinomas (9%) showed higher Rab11A levels than average normal colonic epithelia (Fig. 2C).

The trend for reduction of RAB11A expression in colon cancers was also independently observed in 70 human colorectal cancers (Nie et al., UMass, personal correspondence).

Kaplan–Meier survival analysis of 331 patients with colon cancer using the TCGA database showed that patients with the most reduced RAB11A mRNA levels (lowest 5 percentile, mean value of 1894.3 within a range of 1626.21–2071.2) had a significantly reduced survival rate (P < 0.001) than the mid-level group (mean mRNA value of 3404.28 within a range of 2102.26–5036.07). By contrast, patients with the most upregulated RAB11A levels (top 5 percentile) do not show a different survival rate (P = 0.25; Supplementary Fig. S3G). In addition, examination of a human colon cancer-resembling mouse colonic adenocarcinoma induced by 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP; ref. 20) revealed a reduction of Rab11 apical localization as compared with adjacent normal glands (Supplementary Fig. S3H). The above results indicate that reduced levels of Rab11 expression is linked to both advanced-stage carcinoma and poorer survival rates.

In Drosophila, zygotic homozygous Rab11 mutants are lethal (21), consequently a molecular system to spatially and temporally control expression of Rab11 RNAi in the adult Drosophila midgut enterocytes enabled development of 3 independent Rab11 RNAi lines (11). Briefly, a Myo1A promoter-Gal4 (Myo1A>) construct in only active in enterocytes but not in intestinal stem cells (ISC) or enteroblasts (16). A Myo1Ats> (tubulin promoter-Gal80ts) construct acts as a temperature sensitive repressor of Gal4 that in turn provides temporal control of UAS-Rab11 RNAi constructs, as well as the expression of the UAS-dependent mCD8GFP reporter (15, 16). Midguts stained for phosphorylated histone 3 (p-H3; Fig. 3A) revealed condensed chromatin in ISCs, which is consistent with ISCs being the major mitotic cells in adult Drosophila midgut (14, 22). UAS-driven Rab11 RNAi resulted in a pronounced increase in the number of mitotic enterocytes (Fig. 3B; see Fig. 3H and I, for quantitation). Independent experiments using a previously characterized dominant-negative (DN) point mutant Rab11 S25N (23) also led to a similar increase in enterocyte mitotic count (see Fig. 3G, for quantitation).

Taking advantage of the Drosophila midgut genetics model, we examined the impact of Rab11 endosome function on tumorigenesis driven by oncogenic activity via EGFR-activating (EGFR^A887T) mutation, Ras-activating (Ras^V12) mutation, APC loss-of-function (LOF, APC RNAi), and Misshapen loss of function (msn RNAi; ref. 24). In Drosophila, the RasV12 mutant is widely used as a cancer model in various fly tissue (25, 26). Suppression of Rab11 expression in combination with the Ras^V12 mutation (Rab11RNAi; RasV12) led to a significant increase in the number of mitotic cells as compared with Ras^V12 mutation in Rab11 expressing cells (Fig. 3C and D; see Fig. 3H and I, for quantitation). Similarly, Rab11 inhibition with elevated Wnt signaling, induced by APC RNAi (27), led to a significant increase in midgut proliferation as compared with APC RNAi in Rab11 expressing cells (Fig. 3E and F; see Fig. 3I, for quantitation). Additive effects of loss of Rab11 with overexpressed Wingless or activated EGFR was not significant (Fig. 3H and I).

To understand the mechanisms contributing to the pro-oncogenic phenotypes initiated by loss of Rab11, we performed an unbiased transcriptomic analysis of Rab11a-deficient mouse intestines (GEO accession number in process) in parallel with an analysis of Rab11 RNAi Drosophila midguts (GEO accession number GSM2435587–GSM2435592). For RNA-seq analysis, intestinal epithelia were collected from Rab11a^FL/FL;Villin-CreER mice 2 days after tamoxifen administration. GSEA and heat map analyses uncovered a significant induction of ISC signature genes, for example Axin2 and Sox9 upon Rab11a deletion (P < 0.001, N = 3 each genotype; Fig. 4A and B). The increased ISC signature in the transcriptome was reflected by a mild expansion of Olfm4⁺ crypt based columnar cells within 2 days after Rab11a disruption (Fig. 4C). The above results demonstrate that disruption of Rab11a expression leads to entry of enterocytes into mitosis with an accompanying promotion of an ISC-like transcriptome as well as a potential linkage to certain oncogenic pathways.

GSEA analysis of mouse gut 2 days after induced Rab11a deletion did not detect a significant enrichment of Yap transcriptional targets (P = 0.13; Supplementary Fig. S4A). Interestingly by IHC, the total and nuclear levels of Yap were significantly elevated in adult Rab11a-deficient IECs (Fig. 4D). Nuclear elevation of Yap was observed in crypt cells and some villus epithelial cells, whereas Paneth cells that are reportedly Yap-negative remain weakly stained in Rab11a-deficient IECs (open arrowheads, Fig. 4D). Acute Rab11a ablation in adult mouse IECs may not completely remove the protein after 2 days (Supplementary Fig. S4B) and the effect of Rab11a-deficiency may also be compensated by Rab11b, a Rab11a homolog that is expressed in mouse intestine. Using CRISPR/Cas9 genome editing, a Rab11b knockout mouse allele was derived and used this to achieve Rab11a and Rab11b double knockout (DKO) mice (Supplementary Fig. S4B). Two days after Rab11a removal from Rab11b-/- background, a significantly enriched Yap pathway signature was uncovered (P < 0.001, Fig. 4E). Following tissue specific DKO (Rab11a from Rab11b^−/− IECs), there was upregulation of the total Yap and Taz in IECs within 2 days (Fig. 4F). Total β-catenin levels were not increased, suggesting a selective impact of Rab11 loss on Yap (Fig. 4F). By day 3, all DKO mice died.

Given the complexity of the mouse gut, enteroid cultures were prepared to determine if changes in Yap protein levels occurred in a cell-autonomous manner. Enteroids from wild type, Rab11a^FL/FL;Villin-CreER, and Rab11a^FL/FL;Rab11b^−/−;Villin-CreER mice were grown ex vivo and Rab11a deletion was induced by addition of 4-OHT (Fig. 4G). Quantification of nuclear Yap levels in the eteroids suggest that Rab11 (a and b) protein expression regulates Yap nuclear accumulation in a Rab11 protein dose-dependent fashion (Fig. 4H). In 4-OHT–treated wild-type enteroids, Yap was primarily localized to puncta at the apical surface and sub-apical cytoplasm with no detectable nuclear staining (Fig. 4I). Two days after 4-OHT treatment, Yap stained puncta were spread throughout the cytoplasm and nuclear Yap was readily detected in subsets of Rab11a^FL/FL;Villin-CreER cells (Fig. 4J; Supplementary Fig. S4C). In DKO enteroids, Yap location was primarily nuclear (also see arrowheads, Fig. 4K; Supplementary Fig. S4B).

Examination of Yap by IHC in AOM-induced tumors in Rab11a^ΔIEC mice revealed uniform nuclear Yap localization in most, if not all, tumor cells (Supplementary Fig. S5A). Tumor cells in 8-week-old APC^Min mouse intestines (with intact Rab11a) had low Yap levels (Fig. 5A, left). In contrast, the vast majority of tumors in Rab11a^ΔIEC; APC^Min mice of the same age showed significantly higher Yap levels (P < 0.001, Fig. 5A and B; Supplementary Fig. S5B). Immunoblots confirmed increased Yap levels in both proximal and distal intestines of Rab11a^ΔIEC; APC^Min mice compared with APC^Min littermates (Fig. 5C). A reduced level of phosphorylated Yap (S112) was detected in Rab11a^ΔIEC; APC^Min samples (Fig. 5C). Yap phosphorylation by LATS regulates Yap cytoplasmic retention and degradation whereas reduced phosphorylation of Yap results in protein stability and elevated nuclear localization (28). Compared with APC^Min adenoma, Rab11a^ΔIEC; APC^Min tumors also exhibited increased levels of Amphiregulin (Areg; Fig. 5C), an EGF superfamily member reported to be under Yap's regulation (29) and, consistent with Areg upregulation, there were an increased number of tumor cells exhibiting elevated MAPK signaling activity (Fig. 5D). Further, in larger epithelial fields of tumor cells lacking Rab11a there was increased nuclear pStat3 immunostaining (Fig. 5E), likely reflecting an expansion of tumor cell population in response to the elevated inflammatory cytokines, such as IL6 (11).

To investigate if Rab11A and YAP demonstrated inverse expressional correlation within human colorectal cancers, IHC studies were conducted on serial sections of 13 colon tumor samples. The biopsy samples possessed normal and disease tissue ranging from tubulovillous adenoma with high-grade dysplasia to stage IV colon cancers. As with mouse tissue, there was an equivalent inverse correlation between Rab11A and YAP expression from normal (high Rab11A, low YAP) versus dysplasia or adenocarcinoma biopsy material (low Rab11A, high YAP; Fig. 5F and G). Examining the transition from normal epithelia (black dotted lines) to dysplasia (blue dotted lines) to low-grade adenoma (red dotted lines) in stage I cancer tissue revealed an immediately reversed expression pattern of the 2 proteins (Fig. 5F).

Loss of Drosophila Hippo and Misshapen, 2 upstream Sterile 20/germinal center kinases that suppress the growth-promoting transcriptional coactivator Yki to regulate growth factor production, has been shown to increase ISC proliferation (30, 31). Given the discovered Drosophila and mouse linkage of Yki/Yap activation in Rab11-deficient guts (Fig. 4E and A), we examined the genetic interaction between Rab11 and Hippo-Yki pathway in Drosophila midgut. Two of the most significantly elevated targets in Rab11 RNAi fly midgut were upd3 (homolog of mammalian IL6) and ImpL2 (homolog of mammalian IGFBP7; Fig. 6B–D). IL6 is a proinflammatory cytokine shown to promote colon cancer progression (32), whereas ImpL2 secreted from cancer tissues are shown to inhibit whole animal metabolism in various Drosophila cancer models involving Yki or Scribble (33). Loss of Rab11 synergized with Ras^V12 mutation, leading to a near 780-fold and 70-fold induction of upd3/IL6 and ImpL2, respectively (Fig. 6C). Furthermore, inclusion of a loss-of-function mutant allele of upd3 (17) as homozygotes led to a significant suppression of mitotic counts in Rab11 RNAi midgut (Fig. 6E). These results functionally confirmed that upd3 played an important role in promoting elevated proliferation and tissue growth observed in Rab11-deficient gut.

Because Yki activation of upd3 expression is normally restricted by the upstream Ste20 kinases Hippo and Misshapen (34), we tested whether expression of constitutively active Yki was sufficient to induce phenotypes and transcriptomic changes observed in Rab11 RNAi; Ras^V12 mice. Using the Myo1A-Gal4 driver, a constitutively active Yki construct (Yki^Act), containing 3 mutated phosphorylation sites that eliminated regulation by upstream kinases (18), was expressed in fly midgut enterocytes. The experimental design mimics in fly the reduced pYAP observed in mouse gut (Fig. 5C). Overexpressing Yki^Act led to highly increased mitotic count (Fig. 6F) and elevated upd3 and ImpL2 transcription (Fig. 6G). To determine whether suppressing Yki could alleviate Rab11 RNAi midgut phenotypes, experiments were conducted combining Yki RNAi and Rab11 RNAi. Double RNAi led to partial suppression of Rab11 RNAi midgut phenotypes (Fig. 6H–O). Most notably, Yki RNAi suppressed the dysplastic multilayered epithelial structure induced by Rab11 RNAi and the midgut now possessed a more typical monolayer structure (Fig. 6L–O). These genetic experiments collectively suggest that elevated Yki activity in fly accounts for some phenotypes seen in Rab11-deficient gut and suppressing Yki alleviates the dysplastic phenotype.

Immunoblot analysis across a panel of seven human colon cancer cell lines showed that Caco2 and LIM2551 expressed the highest levels of Rab11A and relatively lowest levels of YAP/TAZ. SW480, LIM1215, and LIM2550 cells had low Rab11A levels and exhibited relatively higher YAP/TAZ levels (Fig. 7A). Following this lead, 3 stable RAB11A-knockdown (KD) cell lines were prepared using Caco2, SW480, and LIM2551 and lentiviral siRNA infection (Fig. 7B). These cell lines have distinct oncogenic mutations in APC and K-RAS (35). Rab11A-KD Caco2 and SW480 cells showed increased total YAP/TAZ (Fig. 7B) and all 3 lines have reduced pYAP at S127 (Fig. 7B and C). Phosphorylation of human YAP at S127 serves as a cytoplasmic retention signal (36) and reduced levels of S127 phosphorylation resulted in elevated levels of YAP in biochemically isolated nuclei (Fig. 7D and E).

Rab11A-KD Caco2 and LIM2551 cells also exhibited decreased phosphorylation of YAP at serine 397 (Fig. 7F) that would result in decreased proteasome-mediated YAP degradation (36). Increased nuclear YAP was correlated with elevated levels of AREG and IL6 (Fig. 7F and G); SW480 cells did not express detectable levels of AREG. Using a TEAD reporter (37), we detected significantly elevated canonical YAP activity in RAB11A-KD SW480 cells (Supplementary Fig. S6A). Chromatin immunoprecipitation (ChIP) for YAP1 demonstrated that in both Caco2 and SW480 cells, RAB11A-deficiency increased YAP1 chromatin association with its targets: LATS2, TEAD1, and CCND1 (Supplementary Fig. S6B and S6C). All of the cumulative data presented in the above section indicate that elevated YAP activity can cell-autonomously regulate the expression of YAP targets.

Phosphorylation of YAP at S127 in humans or S112 in mice is mediated by the LATS1 kinase that itself is activated by phosphorylation through the upstream Hippo kinase cascade. Rab11A-KD Caco2 cells showed no significant difference in the total level of Hippo complex components that control YAP activity including MST1, LATS1, MOB1, SAV1, or exportin-1 or α-catenin as compared with controls (Fig. 7H; ref. 38). Phosphorylation of LATS1 is reported to occur following trafficking to the apical junctions to suppress YAP (39). Interestingly, there were observable reductions in the levels of phosphorylated LATS1 at T1079, S909, or S872 in Rab11A-KD cells (Fig. 7I). Additionally, by immunofluorescence staining, there was readily detectable localization of YAP to cell–cell junctions (arrows, Fig. 7J). In Rab11A-KD cells, YAP was noticeably enriched in nuclei (Fig. 7K) whereas junctional localization was effectively lost in these cells (Fig. 7L). These data collectively suggest that Rab11a may likely play a key role in junctional delivery and/or assembly of key Hippo kinase-dependent molecules that are essential for regulation of YAP activity.

Taking advantage of several genetic models, this report demonstrated that Rab11 protein has the capability of restraining uncontrolled cell growth and progression of colon cancers. The results suggest that perturbation of recycling endosome traffic, a hallmark function in epithelia, can influence differentiated epithelial cell status by unbalancing signaling pathways leading to reversion into a progenitor-like, proliferative cell state. In addition, altered expression and localization of RAB11A in colon cancer tumors suggest that altered recycling endosomal functionality may participate in accelerating cancer progression.

Our results support the view that perturbation of recycling endosome assembly/trafficking is sufficient to induce early onset epithelial dysplasia leading to an oncogenic cue for uncontrolled growth. Loss of mouse Rab25, a member of Rab11 superfamily, was shown previously to increase intestinal tumor burden in mice carrying APC mutation. Rab25-deficient mouse intestinal tumor cells also exhibited reduced integrin expression, thereby enhancing invasive behavior in a Smad3 haploid insufficiency background (40). Our genetic experiments on Rab11 in fly and mouse gut showed that perturbation of this small GTPase robustly synergizes with various oncogenic driver mutations, including APC loss-of-function and Ras^V12 to promote not only the production of pro-growth inflammatory cytokines, for example, upd3/IL6 and Areg, but also the expansion of a tumorigenic stem cell pool. Rescue experiments suppressing Yap/Yki, or its target IL6/Upd3, in Rab11-deficient gut reduced the dysplastic and hyperplastic growth, supporting a conclusion that the observed rapid tumor expansion was in part due to the elevated inflammatory proteins produced in the Rab11-deficient tissue microenvironment.

Further, unbiased meta data analysis revealed reduced RAB11A mRNA level in the majority of documented colon cancers. In our analysis, most of the colon cancers showed altered Rab11a apical localization and a strong inverse correlation between Rab11a expression and total YAP cellular levels. Reduced and altered apical Rab11a localization has been reported in human congenital enteropathy (41). RAB11 was also reported to be upregulated in several nonintestinal tumors including non–small cell lung carcinoma (42), skin tumors (43), and pancreatic cancers (44). Also, nearly 9% of colon cancers showed elevated RAB11A level when compared with normal colonic epithelia; however, survival analysis of patients with colon cancer using the TCGA database demonstrated that only those patients with reduced RAB11A displayed significantly poorer rates of survival. Thus, the RAB11 endosomal compartment may have context- or cell-type-dependent functions in tumorigenesis. Caution should be taken for any proposed therapeutic targeting of this molecular machinery (45).

Rab11 endosomes traffic a diverse spectrum of membrane and secretory proteins, and our transcriptomic study discovered a previously unrecognized link to important growth pathways. In intestinal epithelia, across various species, Rab11 perturbation appeared to intersect with the Hippo pathway. Experiments with mouse enteroids support a cell-autonomous effect of loss-of-Rab11 on Yap nuclear translocation. In prior work, we reported that induced deletion of Rab11a from intestinal epithelia led to rapid aberrant formation of apical membrane structures, such as the microvilli, at the basolateral cellular domains (12, 13, 46) and now RNA-seq analysis detected genome-wide activation of stem cell transcriptomic program. In addition, Rab11a removal from Rab11b^−/− IECs led to immediate alteration of the YAP pathway and death of the animals within 3 days. Although this lethality phenotype was reminiscent of transgenic mice that died after a week of induced overexpression of YAP(S127A) mutant (47), the loss-of-function of Rab11a/b appeared to have a more robust phenotype implicating a broad molecular impact that warrants future study. Our current work provides substantial in vivo evidence that perturbing Rab11 endosome functionality can rapidly and cell-autonomously alter Hippo signaling and accelerate colon cancer development (48–52). At this point, we speculate that RAB11 endosomes transport key Hippo pathway components to sites of YAP regulation or endosomal sorting may provide a scaffolding platform for Hippo complex assembly and YAP suppression.

J.R. Goldenring reports receiving commercial research grant from ViiV Inc. and is a consultant/advisory board member of Takeda. No potential conflicts of interest were disclosed by the other authors.

Conception and design: L. D'Agostino, Y. Nie, S. Goswami, S. Yu, S. Bandyopadhyay, I. Balasubramanian, C.S. Yang, E.M. Bonder, J.R. Goldenring, Y.T. Ip, N. Gao

Development of methodology: S. Goswami, K. Tong, S. Bandyopadhyay, I. Balasubramanian, R. Sakamori, B. Gao, E.M. Bonder, Y.T. Ip, N. Gao

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. D'Agostino, Y. Nie, S. Goswami, S. Yu, S. Bandyopadhyay, J. Flores, I. Balasubramanian, I. Joseph, Q. Li, B. Gao, G. Yehia, L. Zhang, Y.T. Ip, N. Gao

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Nie, S. Goswami, K. Tong, S. Yu, S. Bandyopadhyay, X. Zhang, I. Balasubramanian, Q. Li, R.P. Ferraris, E.M. Bonder, J.R. Goldenring, J.R. Goldenring, M.P. Verzi, L. Zhang, Y.T. Ip, N. Gao

Writing, review, and/or revision of the manuscript: S. Goswami, K. Tong, I. Balasubramanian, R.P. Ferraris, E.M. Bonder, J.R. Goldenring, J.R. Goldenring, M.P. Verzi, L. Zhang, Y.T. Ip, N. Gao

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. D'Agostino, S. Goswami, V. Farrell, R.P. Ferraris, J.R. Goldenring, M.P. Verzi, Y.T. Ip, N. Gao

Study supervision: S. Goswami, J.R. Goldenring, M.P. Verzi, Y.T. Ip, N. Gao

Other (This is part of a long-term collaboration between Dr. Nan Gao and me.): C.S. Yang

We thank the Facility of Genome Editing of Rutgers Cancer Institute of New Jersey (P30CA072720) for developing Rab11b-null mice and the Perceval HPC cluster (NIH grant #1S10OD012346-01A1) for computational analysis. Work was supported by the NIH (CA178599, DK102934, AT010243, DK119198), ACS Scholar Award (RSG-15-060-01-TBE), NSF/BIO/IDBR (1353890), and a Rutgers IMRT award to N. Gao; NIH/NCI R01CA190558 to M.P. Verzi; R01 DK70856 to J.R. Goldenring; a NSF grant (IOS-1456673, IOS-1754783) to R.P. Ferraris; a CCFA career development award (406794) to S. Yu; R01DK102934-S1 to I. Joseph; NIH grants (GM107457, DK083450), UMass DERC (DK32520), UMass Center for Clinical and Translational Science (UL1TR000161), Guangdong Innovative Research Team Program (No. 201001Y0104789252), and the Science and Technology Program of Guangzhou (No. 201704030044) to Y.T. Ip; NJCCR fellowships (DFHS16PPC038, DHFS16PPC036, DFHS17PPC036, DCHS19PPC038) to S. Goswami, K. Tong, S. Bandyopadhyay, J. Flores, respectively; and NIH R25 IMSD grant (2R25GM096161-08) to J. Flores.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.