R-Spondin1/LGR5 Activates Tgfb Signaling and Suppresses Colon

Published OnlineFirst September 22, 2017; DOI: 10.1158/0008-5472.CAN-17-0219

Cancer Tumor and Stem Cell Biology Research

R-Spondin1/LGR5 Activates TGFb Signaling and Suppresses Colon Cancer Metastasis Xiaolin Zhou1, Liying Geng1, Degeng Wang2, Haowei Yi1, Geoffrey Talmon3,and Jing Wang1,4,5

Abstract

Leucine-rich repeat containing G-protein–coupled receptor an orthotopic model of colon cancer in vivo.UponRSPO1 5 (LGR5), an intestinal stem cell marker, is known to exhibit stimulation, LGR5 formed complexes with TGFb receptors. tumor suppressor activity in colon cancer, the mechanism of Studies of patient specimens indicate that LGR5 expression which is not understood. Here we show that R-spondin 1 was reduced in advanced stages and positively correlated with (RSPO1)/LGR5 directly activates TGFb signaling cooperatively markers of TGFb activation in colon cancer. Our study with TGFb type II receptor in colon cancer cells, enhancing uncovers a novel cross-talk between LGR5 and TGFb signaling TGFb-mediated growth inhibition and stress-induced apopto- in colon cancer and identiﬁes LGR5 as a new modulator sis. Knockdown of LGR5 attenuated downstream TGFb signal- of TGFb signaling able to suppress colon cancer metastasis. ing and increased cell proliferation, survival, and metastasis in Cancer Res; 77(23); 6589–602. 2017 AACR.

Introduction providing evidence that intestinal tumors arise from LGR5-positive cells. However, ablation of LGR5-positive stem cells in the LGR5, a leucine-rich repeat-containing G-protein–coupled crypts had no effect on APC loss–induced crypt hyperplasia (12), receptor (GPCR; refs. 1, 2), was found to be a marker for prolif- indicating that the absence of LGR5 does not impair intestinal erating adult stem cells in small intestine (3). LGR5 and its tumor initiation. Furthermore, there are other conﬂicting reports homologs LGR4 and LGR6 are receptors of R-spondins (RSPO; regarding the functional role of LGR5 in colon cancer tumori- refs. 4, 5), which are secreted agonists of canonical Wnt signaling genesis. Some studies reported that RSPO/LGR5 signaling has a (6, 7). Unlike other GPCRs, LGR4-6 are not coupled to either G tumor suppressive activity in colon cancer (8, 13) whereas others proteins or b-arrestin after stimulation by RSPOs (4). Instead, they showed that LGR5 promotes tumorigenicity (14). Therefore, the have been shown to regulate Wnt canonical signaling, with both role of LGR5 in colon cancer tumorigenesis remains ambiguous, activation and inhibition reported in different cell contexts (5, 8). although in a transposon mutagenesis study RSPO2 scored sec- LGR5 itself is also a transcriptional target of Wnt signaling (9). ond only to APC as a colon cancer tumor suppressor locus (15). The physiological role of LGR5 during normal intestinal devel- Moreover, it is yet to be determined whether LGR5 contributes to opment and tumor initiation has been extensively studied. LGR5 colon cancer progression and metastasis. deﬁciency leads to premature Paneth cell differentiation, associ- TGFb regulates many aspects of cellular function. Upon ligand ated with Wnt activation, whereas it has no detectable effects on binding, TGFb type II receptor (RII) recruits and activates TGFb differentiation of other cell lineages, nor on epithelial cell pro- type I receptor (RI), which phosphorylates and activates Smad2/3. liferation or migration (10). The role of LGR5 in intestinal tumor Activated Smad2/3 form complexes with Smad4 and translocate development is controversial. Conditional deletion of APC in to the nucleus, where they regulate gene expression (16). The LGR5-positive but not LGR5-negative cells led to a rapid growth of TGFb axis has been shown to inhibit cell proliferation, induce large adenomas in small intestine in a mouse model (11), apoptosis, and suppress tumorigenicity in many colon cancer cell lines (17) and in intestinal adenomas (18). Abrogation of TGFb b 1Eppley Institute for Research in Cancer and Allied Diseases, Fred & Pamela signaling increases metastasis whereas enhanced TGF signaling Buffett Cancer Center, Omaha, Nebraska. 2Department of Environmental Tox- suppresses metastasis in an orthotopic model of colon cancer icology, The Institute of Environmental and Human Health, Texas Tech Univer- (19, 20). In genetic mouse models, Smad3 or Smad4 mutation sity, Lubbock, Texas. 3Department of Pathology and Microbiology, University of increases malignancy and invasiveness of intestinal tumors of 4 þ Nebraska Medical Center, Nebraska Medical Center, Omaha, Nebraska. Depart- Apcmin/ or knockout mice (21, 22), indicating a suppressive role ment of Genetics, Cell Biology and Anatomy, Fred & Pamela Buffett Cancer 5 of TGFb signaling in malignant progression of colon tumors. Center, Omaha, Nebraska. Department of Biochemistry and Molecular Biology, b Fred & Pamela Buffett Cancer Center, Omaha, Nebraska. Indeed, TGF signaling is defective in 30% to 40% of colon cancer patients due to defects in TGFb RII or Smads (23, 24), and loss or Note: Supplementary data for this article are available at Cancer Research b Online (http://cancerres.aacrjournals.org/). reduction of TGF signaling is tightly associated with metastasis development (25, 26). X. Zhou and L. Geng contributed equally to the article. To address the potential function of LGR5 in colon cancer Corresponding Author: Jing Wang, University of Nebraska Medical Center, metastasis, we examined LGR5 expression in clinical specimens, 985950 Nebraska Medical Center, Omaha, NE 68198. Phone: 402-559-5558; Fax: and found that LGR5 expression decreased in advanced stages of 402-559-4651. E-mail: [email protected] colon cancer as compared to early stages. Knockdown of LGR5 doi: 10.1158/0008-5472.CAN-17-0219 expression in colon cancer cells increased their clonogenicity, 2017 American Association for Cancer Research. survival capacity in vitro, and liver and lung metastasis in an

www.aacrjournals.org 6589

Zhou et al.

orthotopic model in vivo. Investigation of potential cross-talk vector. The retroviral expression vectors for C-terminal-tagged or between RSPO1/LGR5 and TGFb signaling revealed that non-tagged full-length human LGR5 were gifts from Dr. Keith RSPO1/LGR5 activated the canonical TGFb pathway and elicited Johnson (University of Nebraska Medical Center, Omaha, NE). its inhibitory effect on clonogenicity and cell survival of colon Dominant negative TCF expression vector (EdTC) was a gift from cancer cells through the activation of TGFb signaling. Mechanis- Roel Nusse (24310; Addgene). Transfection and infection were tically, LGR5 colocalized and formed complexes with TGFb performed as described previously (30). receptors upon RSPO1 stimulation. Importantly, LGR5 expression positively correlated with Smad2 activation in colon cancer RT-PCR assays specimens. Therefore, our studies identify LGR5 as a novel com- Total RNA was isolated from colon cancer cells, liver, or lung of ponent/modulator of the TGFb signaling pathway and demon- mice using TRIzol reagent (15596026) from Thermo Fisher Sci- strate that RSPO1/LGR5 functions as a suppressor of colon cancer entific. Two micrograms of total RNA was reverse-transcribed to metastasis. cDNA with M-MLV reverse transcriptase (1701; Promega) using a random primer 2 mL of cDNA product was used to amplify specific Materials and Methods genes. Actin was used as a loading control. Sequences of primers Colon cancer cell culture for each gene were listed in Supplementary Table S1. Human colon carcinoma cell lines (HCT116, RKO, FET, CBS, HCT116b, and TENN) were originally established in Dr. Brattain's Colony formation and soft agarose assays lab in 1981 (27). All cell lines were validated by short-tandem For colony formation assays, cells were seeded in six-well plates repeat (STR) analyses at the University of Nebraska Medical at 500 cells per well. After 2 weeks, colonies were stained with MTT Center Human DNA Identification Laboratory. STR profiles were and dissolved in DMSO. Relative cell numbers were determined cross-checked with the ATCC database. HCT116 and RKO cell by OD values read at 570 nm. For soft agarose assays, cells were lines with 80% match with the ATCC online STR database are seeded in culture medium containing low melting point agar considered valid (28). STR profiles of FET, CBS, HCT116b, and (BMA50101; Lonza) at a density of 3,000 cells per well in 6-well TENN cell lines have never been reported before. Their STR plates. Colonies were stained with 1% iodonitrotetrazolium profiles were confirmed in ATCC database to be of human origin violet (I7375; Sigma-Aldrich) after 3 weeks and the numbers of and contain no mammalian interspecies contamination (Supple- colonies were counted. mentary Fig. S1A). There were 5–8 passages between thawing and use in the described experiments for all the cell lines. Apoptosis assays Cells were maintained at 37 C in a humidified incubator with Cells were plated and allowed to grow to 80% confluence 5% CO2 and cultured in McCoy's 5A serum-free medium (M4892; before switching to medium in the absence of growth factor or Sigma) supplemented with 10 ng/mL EGF, 20 mg/mL insulin, and serum supplements for 2–5 days. Apoptosis was determined by 4 mg/mL transferrin (29). When the cells were subjected to growth measuring cleaved PARP or cleaved caspase-9 or by performing factor deprivation stress, they were cultured in the medium in the DNA fragmentation ELISA assays (11585045001; Roche Diag- absence of growth factor or serum supplements. nostics) following the manufacturer's instruction. Cleaved caspase-9 and PARP are indicators of induction of apoptosis, whereas Antibodies and reagents DNA fragmentation assays can relatively quantify the amount of Anti-LGR5 (ab75850) and anti-phospho-Smad2 (Ser465/467; histone-complexed DNA fragments during apoptosis. ab3849) antibodies were purchased from Abcam. Anti-cleaved PARP (5625), anti-cleaved caspase-9 (9502), anti-Smad2/3 Luciferase assays (8685), anti-Smad4 (9515), anti-TGFb RII (11888), and anti- SBE4-Luc (16495; Addgene) is a reporter plasmid containing actin (4970) antibodies were obtained from Cell Signaling Tech- multiple copies of Smad-binding site and encoding the firefly nology. Anti-TGFb RI (sc-398) was from Santa Cruz Biotechnol- reporter gene. TOPflash (12456; Addgene) reporter consists of ogy. Anti-Flag (F1804) and anti-HA (MMS-101P) antibodies were multiple TCF-binding sites and encodes the firefly reporter gene. purchased from Sigma and Biolegend, respectively. Recombinant When they are transfected into the cells, the firefly luciferase human TGFb1 and RSPO1 were purchased from R&D Systems activity reflects the activation of the TGFb or Wnt signaling and Origene. The TGFb RI kinase inhibitor (SB525334) and the pathway, respectively. The pRL-CMV plasmid encoding Renilla human RSPO1 neutralizing antibody (AF4645), which is antigen luciferase (E6621; Promega) was cotransfected as an internal affinity-purified and displays less than 1% cross-reactivity with control. Luciferase assays were performed using the Dual Lucif- other human RSPOs, were purchased from Sigma and R&D erase Assay System (E1910; Promega). The luminescent signal of Systems, respectively. firefly and Renilla was detected using the BioTek Synergy 2 Multi- Mode Reader according to the manufacturer's instruction. Values Plasmids, transfection, and infection of firefly luciferase were divided by those of Renilla luciferase for Two LGR5 shRNAs were cloned into the FSIPPW lentiviral the normalization. vector. Smad2 and Smad3 shRNAs were cloned into the pSU- PER.retro retroviral vector (Addgene). The pLKO.1 lentiviral TGFb RNA sequencing RI, RII, and Smad4 shRNA constructs were from Sigma-Aldrich. FET cells in triplicates were treated with RSPO1 or TGFb1 for 4 Two pLKO-b-catenin shRNA constructs were gifts from William hours. RNA was isolated using RNeasy Mini Kit (47104; Qiagen). Hahn (19761 and 19762; Addgene). Scrambled shRNAs were RNA sequencing was performed by the City of Hope Integrative used as controls. Full-length human TGFb RII cDNA was ampli- Genomics Core facility. Poly(A) RNA-seq libraries were con- fied from FET cells and cloned into the pBabe-based retroviral structed using TruSeq mRNA Library Preparation Kit (RS-122;

6590 Cancer Res; 77(23) December 1, 2017 Cancer Research

R-Spondin1/LGR5 Activates TGFb in Colon Cancer Metastasis

Illumina). Purified libraries were validated using Bioanalyzer formed as described previously (33). Briefly, exponentially grow- 2100 system with DNA High Sensitivity Chip (Agilent) and ing GFP-labeled CBS cells (5 106) were inoculated subcutane- quantified with Qubit Fluorometer (Invitrogen). All libraries ously into athymic nude mice (male, 5–6 weeks, 20–25 g; Harlan were sequenced on the Illumina Hiseq 2500 with single read Laboratories). Once xenografts were established, they were 51-bp reads following the manufacturer's recommendations. excised and minced into 1-mm3 pieces, two of which were then The 51-bp-long single-ended sequence reads were mapped to transplanted onto the subserosal layer of the cecum of other mice. human genome (hg19) using TopHat, and the frequency of Seventy days posttransplantation, mice were euthanized. Organs Refseq genes was counted with customized R scripts. The raw were explanted, imaged, and immediately frozen or placed in counts were then normalized using trimmed mean of M values buffered 10% formalin. Tissues were then processed and embed- method (31) and compared using Bioconductor package ded in paraffin. Four-micron thick sections were cut for IHC or "edgeR" (32). The average coverage for each gene was calcu- H&E staining. lated using the normalized read counts from "edgeR." Differ- entially expressed genes were identified if the average coverage Human tissue specimens 1 in at least one sample, log2-based fold change 1or1 The patient studies were conducted in accordance with and FDR <0.05.TheRNA-seqdatahavebeensubmittedtoGEO Belmont Report and US common rule ethical guidelines. repository (accession no. GSE103650). Informed written consent from the subjects (wherever neces- sary) were obtained. Following approval by the Institutional Immunoprecipitation assays Review Board of the University of Nebraska Medical Center HCT116 cells expressing Flag-tagged TGFb RII and HA-tagged (UNMC), a tissue microarray (TMA) was made by the Tissue 6 LGR5 (HCT116-RII/LGR5) were seeded at 1.0 10 cells/dish in Science Facility. The TMA contained triplicate samples obtained 100-mm dishes and allowed to attach for 4 hours. Cells were then from formalin-fixed paraffin-embedded (FFPE) blocks of nor- treated with 200 ng/mL RSPO1 or 4 ng/mL TGFb1 for 30 minutes. mal human colorectal mucosa from specimens removed for Whole-cell lysates were made in NP-40 lysis buffer (20 mmol/L reasons other than malignancy (i.e., diverticulosis) and colonic Tris-HCL, pH 8.0, 137 mmol/L NaCl, 10% glycerol, 1% nonidet adenocarcinomas at stages I, II, III, or IV retrieved from the files P-40, 2 mmol/L EDTA) with protease inhibitor cocktail (P8340, of Department of Pathology and Microbiology. The ages of all Sigma). Three hundred micrograms of the cell lysate was incu- patients (including both men and women) were between 55 bated with a primary antibody for overnight at 4 C on a rotator and 85 years. followed by incubation with secondary antibody at room temperature. Protein complexes were captured on magnetic beads IHC staining (LSKMAGA10; Millipore, EMD Millipore) for 30 minutes. As a FFPE blocks of human/mouse colon tumors were cut into 4- negative control, a normal IgG antibody was used. Immunopre- mm-thick tissue sections. IHC staining was performed to exam- cipitated proteins were resolved by SDS-PAGE followed by West- ine LGR5 and pSmad2 expression in TMA and MSI samples as ern blot analysis as described previously (30). well as primary tumors of CBS control or LGR5 knockdown cells following Novolink Min Polymer Detection System Kit fl Immuno uorescence and confocal microscopy protocol (RC7290-CE; Leica). Briefly, slides were subjected to HCT116-RII/LGR5 cells were plated onto glass coverslips and antigen retrieval using Novocastra Epitope Retrieval Solutions, b treated with RSPO1 or TGF 1 for 30 minutes. The cells were then pH 6 (RE7113; Leica), followed by incubation with primary fi washed with PBS and xed with 4% ice-cold paraformaldehyde antibodies (anti-LGR5 and anti-pSmad2) for overnight at 4C. fi for 10 minutes. The xed cells were blocked with 3% BSA-PBS for One the next day, slides were developed with DAB after incu- 1 hour at room temperature, followed by incubation with primary bation with Novolink polymer for 30 minutes. Finally, the antibodies (anti-Flag, anti-HA, anti-LGR5, anti-TGFb RI) for over- sections were counterstained with hematoxylin. The TMA slide night at 4 C. On the next day, the cells were washed and incubated was scanned at 40 using Ventana iScan Coreo Au Scanner. Ten with secondary antibodies conjugated with Alexa Fluor 488 or fields were randomly selected from each slide for quantifica- 594 (Invitrogen) for 1 hour at room temperature. Nuclei were tion. The staining intensity and density were quantified with counterstained with DAPI. Double stained images were obtained Imagescope Software (V12.1.0.5029; Leica). by sequential scanning for each channel to eliminate the cross- reactivity of chromophores. Images were captured and analyzed using confocal laser scanning microscope (LSM800; Carl Zeiss Statistical analyses Microscopy GmbH). Quantification of fluorescence signal was Statistics analysis was performed in GraphPad Prism 5 after fi t performed using an algorithm direct Gaussian fit in the ZEN blue gures were generated. Two-sided paired Student test or one-way version 2.3 software. ANOVA was used to analyze the differences among groups. fi For each sample, approximately 20 different cell images were Pearson correlation coef cient analysis was used to determine fi acquired and quantified. After cell cropping, background correc- the correlation between two sets of samples. Statistical signi - tion and intensity thresholds were applied, colocalization and cance of metastatic incidence was determined by Fisher exact test P < correlation were calculated using Pearson's coefficient. in the orthotopic transplantation experiments. A value 0.05 was considered statistically significant. In vivo orthotopic transplantation Experiments involving animals were approved by the Univer- Results sity of Nebraska Medical Center Institutional Animal Care and LGR5 expression decreases in advanced stages of colon tumors Use Committee (IACUC; protocol no. 1509711FC) and IACUC To determine the potential function of LGR5 in colon cancer, regulations were followed. Orthotopic transplantation was per- we examined LGR5 expression in clinical specimens. We first

www.aacrjournals.org Cancer Res; 77(23) December 1, 2017 6591

Zhou et al.

analyzed the Oncomine database obtained from Skrzypczak LGR5 inhibits cell survival and clonogenicity of colon cancer colorectal dataset (34), which showed that LGR5 mRNA expres- cells in vitro sion is higher in colon adenomas than in normal colon, but is These observations promoted us to explore whether LGR5 lower in carcinomas than in adenomas (Supplementary Fig. S1B). might play a suppressive role in tissue culture models of colon Furthermore, we assayed LGR5 protein expression in human cancer progression and metastasis. We first examined LGR5 colonic adenocarcinomas using IHC staining. We first verified expression in a panel of human colon cancer cell lines with the specificity of the anti-LGR5 antibody used successfully for IHC differing metastatic capability. Highly metastatic cells (TENN, staining (Supplementary Fig. S1C and S1D; refs. 35, 36). Paraffin RKO, and HCT116) express lower amounts of LGR5 than mod- blocks from 12 normal colon and 73 colon cancer patients at erately, weakly, or nonmetastatic CBS, HCT116b, and FET cells, different stages were stained. LGR5 staining varied significantly in respectively, at both mRNA and protein levels (Fig. 2A), indicating different groups of samples (Fig. 1A). Quantification of the an inverse correlation between LGR5 expression and metastatic staining showed that, although there were overlaps in LGR5 potential. Accordingly, HA-tagged LGR5 (LGR5/HA) was intro- expression between groups, the average intensity and density of duced into CBS cells expressing intermediate level of LGR5 (Fig. LGR5 staining were higher in stage I/II tumors than in normal 2B, top). Because aberrant survival capacity of tumor cells is an crypts and lower in stage III/IV tumors than in stage I/II tumors important determinant of metastatic potential (37, 38), we first (Fig. 1B and C). These results indicate that expression of LGR5 evaluated the ability of colon cancer cells to withstand stress- decreases in advanced stages of colon cancer as compared with induced apoptosis in vitro. Under growth factor and nutrient early-stage disease, suggesting that there may be selection for deprivation stress (GFDS), LGR5-expressing cells became more active LGR5 expression early in colon tumorigenesis, or it simply sensitive to GFDS-induced apoptosis than control cells, reflected reflects the origin of the tumors as derived by clonal expansion of by increased caspase-9 cleavage (Fig. 2B, bottom) and enhanced LGR5-positive stem cells, but suppression thereafter during tumor apoptosis in DNA fragmentation assays (Fig. 2C). In addition, progression. increased expression of LGR5 reduced clonogenicity in colony

Figure 1. Evaluation of LGR5 expression in colon cancer patient specimens. A, IHC staining with an anti- LGR5 antibody was performed in normal colon and colon cancer specimens at different stages. Representative images are shown. Scale bar, 100 mm. B and C, Quantiﬁcation of percentage of LGR5-positive cells (B)and LGR5 staining intensity (C) was performed. The values of individual samples are shown. Error bars, SEM of the values in each group. , P < 0.05; , P < 0.01; , P < 0.001.

6592 Cancer Res; 77(23) December 1, 2017 Cancer Research

R-Spondin1/LGR5 Activates TGFb in Colon Cancer Metastasis

formation (Fig. 2D) and soft agarose (Fig. 2E) assays. Of note, relevant pathways, particularly TGFb signaling, a well-estab- expression of non-tagged LGR5 showed a similar effect as HA- lished tumor suppressor pathway in colon cancer, but not tagged LGR5 (data not shown), indicating that the HA tag does previously interconnected with LGR5. not affect LGR5 function. Furthermore, treatment of FET cells with Accordingly, we deployed a luciferase plasmid driven by RSPO1 likewise increased GFDS-induced apoptosis (Fig. 2G) and multiple Smad binding elements (SBE4-Luc) as a surrogate reduced clonogenicity in soft agarose assays (Fig. 2H). These marker of TGFb-dependent promoter activity. FET and CBS cells results indicate that RSPO1/LGR5 inhibits cell survival and clo- were treated with RSPO1. As they both also express endogenous nogenicity in vitro. TGFb RII, TGFb1 was used as a reference ligand. RSPO1 exposure Complementarily, LGR5 expression was significantly reduced increased Smad reporter activity in both cell types in a dose- with each of two independent shRNAs (#5 and #6) in FET and CBS dependent manner (Fig. 5A and Supplementary Fig. S4A, result cells (Fig. 3A, top). LGR5 knockdown cells showed increased in Fig. 4D will be discussed below). A time course study showed resistance to GFDS-induced apoptosis, reflected by decreased PARP that RSPO1 and TGFb1 activated Smad2 with similar dynamics cleavage (Fig. 3A, bottom) and reduced apoptosis in DNA frag- (Fig. 5B). Besides FET and CBS cells, RSPO1 also activated Smad mentation assays (Fig. 3B). In addition, clonogenicity assays indi- reporter and increased Smad2 phosphorylation in HT29 cells, cated that knockdown of LGR5 expression led to almost two-fold of which express endogenous LGR5 (Supplementary Fig. S4B). increase in colony forming capacity when plated at sparse densities These studies indicate that RSPO1/LGR5 activates TGFb signal- (Fig. 3C) and 2.5- to 4.5-fold of increase in anchorage-independent ing in colon cancer cells. growth (Fig. 3D). These results demonstrate that reduced LGR5 To determine whether RSPO1 and TGFb1 activate similar expression increases cell survival and clonogenicity in vitro. transcription, RNA-seq analysis was performed to compare gene profiles after treatment by RSPO1 or TGFb1. FET cells were treated LGR5 suppresses metastasis of colon cancer cells in vivo with RSPO1 or TGFb1 in triplicates for 4 hours and RNAs extracted We next examined the role of LGR5 in colon cancer metastasis from these cells were used for RNA-seq analysis. Genes with log2- using an orthotopic mouse model that substantially recapitulates based fold change 1 and FDR <0.05 were considered upregu- human colon cancer metastasis to the liver and lungs (39, 40). lated genes and those with log2-based fold change 1 and FDR CBS cells, moderately metastatic in vivo (20), were used. CBS <0.05 were considered downregulated genes. There were 269 control and LGR5 knockdown cells were stably transfected with genes upregulated and 126 genes downregulated by RSPO1 and GFP. Mice implanted with either control or LGR5 knockdown 120 genes upregulated and 42 genes downregulated by TGFb1 cells (LGR5 KD: combined results of two shRNAs) showed 100% (Supplementary Table S2 and S3). The log2 data of genes regu- primary tumor growth. Knockdown of LGR5 expression resulted lated by RSPO1 or TGFb1 are shown as a hierarchical cluster in a modest 23% increase in primary tumor weight (Supplemen- diagram (Fig. 5C). When gene expression profiles were compared, tary Fig. S2A). However, it significantly increased the incidence of it was found that 153 genes were commonly regulated by either liver or lung metastasis from 29% to 77%, assessed by histologic RSPO1 or TGFb1. Interestingly, the majority of genes regulated by analyses (Fig. 4A and B). Fluorescence imaging of explanted liver TGFb1 were also regulated by RSPO1. A Venn diagram depicting or lungs showed increased tumor burden of metastases in mice the numbers of genes commonly or exclusively regulated by implanted with LGR5 KD cells (Fig. 4C). To confirm it, RNA was RSPO1 or TGFb1 is shown in Fig. 5D. The RNA-seq data have extracted from lungs and liver of each mouse and semiquantita- been validated by semiquantitative RT-PCR of p21 and c-myc, tive RT-PCR was performed using human-specific GAPDH pri- both regulated by RSPO1 or TGFb1 to similar degrees (Fig. 5E). mers. The level of human GAPDH mRNA expression represents Taken together, these results indicate that RSPO1 activates TGFb1- the amount of human RNA, which is reflective of tumor burden in mediated gene transcription. the lungs and liver of mice. RT-PCR results showed that human- In a complementary approach, we examined the converse effect specific GAPDH mRNA level was much higher in LGR5 KD group of reducing LGR5 expression on Smad activation. LGR5 knock- than in the control group (Supplementary Fig. S2B and S2C). down in FET and CBS cells (Fig. 3A) decreased the Smad reporter These results indicate LGR5 inhibits incidence and tumor burden activity by approximately 60% to 80% (Fig. 5F, left). Consistently, of colon cancer metastasis in vivo. induction of Smad2 phosphorylation by TGFb was attenuated in LGR5 knockdown cells as compared to control cells (Fig. 5F, LGR5 activates TGFb signaling in colon cancer cells right). Furthermore, IHC staining showed that nuclear Smad2 To explore the mechanism(s) of LGR5 suppression of metas- phosphorylation was reduced in primary tumors derived from tasis, we first investigated the effect of LGR5 on procarcinogenic CBS LGR5 KD cells by 79% as compared to those derived from Wnt signaling, given that RSPO1/LGR5 is a well-established control cells (Fig. 4D), suggesting that knockdown of LGR5 activator of this pathway in many cell types including HEK293T decreased Smad activation in vivo. (4). We were able to recapitulate the finding that RSPO1 Next, we determined which recognized components of TGFb increased Wnt activity in HEK293T cells as reflected by signaling might play a role in pathway activation by RSPO1/ increased phosphorylation of LRP6,aco-receptorofWnt,and LGR5. Expression of RI, RII, Smad2, Smad3, and Smad4, was b-catenin expression (Supplementary Fig. S3A; ref. 4). Howev- reduced individually by shRNAs in FET cells (Fig. 5G and H, top). er, no similar effect was observed in FET cells (Supplementary Expression of LGR5 was also separately knocked down (Fig. 5G, Fig. S3B), which are responsive to the RSPO1/LGR5 biological top). Reduced expression of any of these individual genes atten- effects shown earlier (Fig. 2G and H). The luciferase activity of uated or abrogated RSPO1-induced Smad reporter activation (Fig. TOPflash reporter consisting of multiple TCF-binding sites 5G and H, bottom), indicating that each plays a role in RSPO1- confirmed these observations (Supplementary Fig. S3C). These mediated TGFb activation. results were unexpected and remain under investigation, but in Because TGFb RI kinase is essential for activating Smads, a the meantime encouraged the exploration of other potentially potent and selective inhibitor of TGFb RI kinase, SB525334, was

www.aacrjournals.org Cancer Res; 77(23) December 1, 2017 6593

Zhou et al.

6594 Cancer Res; 77(23) December 1, 2017 Cancer Research

R-Spondin1/LGR5 Activates TGFb in Colon Cancer Metastasis

Figure 3. Knockdown of LGR5 expression increases cell survival and clonogenicity in vitro. A, LGR5 expression was significantly reduced by two LGR5 shRNAs in FET and CBS cells under normal condition (top). Cleaved PARP was determined under GFDS for 48 hours (bottom). B, Apoptosis was measured by DNA fragmentation assays under GFDS for 48 hours. C, Colony formation assays were performed in control and LGR5 knockdown cells with representative images (left) and quantification (right) shown. D, Soft agarose assays were performed in control and LGR5 knockdown cells, with representative images (left) and quantification (right) shown. The data are presented as the mean SEM of three replications. , P < 0.001.

used. This compound almost completely blocked Smad reporter shRNAs in FET cells (Supplementary Fig. S5A). In addition, a activation induced by RSPO1 in FET and CBS cells (Fig. 5A), dominant-negative TCF (dnTCF) was transfected into FET cells. indicating that TGFb RI kinase activity is required for RSPO1/ Knockdown of b-catenin almost completely abolished Wnt LGR5-mediated TGFb activation. signaling and dnTCF reduced Wnt activity by 50% as indicated Functionally, knockdown of RI expression mitigates RSPO1- by the luciferase activity of TOPﬂash reporter (Supplementary induced apoptosis under GFDS and inhibition of clonogenicity in Fig. S5B). However, inhibition of Wnt activity did not affect soft agar in FET cells (Fig. 2F–H). In addition, the RI kinase Smad reporter activation and Smad2 phosphorylation induced inhibitor reversed the inhibitory effect of RSPO1/LGR5 on cell by RSPO1 or TGFb1 (Supplementary Fig. S5C and S5D). These survival and clonogenicity in FET cells (Fig. 2I and J). These data results indicate that RSPO1/LGR5 activates TGFb/Smad signal- collectively indicate that RSPO1/LGR5 elicits tumor suppressor ing independent of Wnt activation. function through the activation of TGFb signaling. Because most colon cancer cells display activated Wnt sig- LGR5 forms complexes with TGFb receptors naling, we next determined whether canonical Wnt signaling Interestingly, knockdown of RII expression reduced Smad plays a role in RSPO1/LGR5-mediated TGFb/Smad signaling. reporter activity induced not only by TGFb1 but also by RSPO1. Expression of b-catenin was knocked down by two independent Similarly, knockdown of LGR5 expression inhibited RSPO1- or

Figure 2. RSPO1/LGR5 inhibits cell survival and clonogenicity of colon cancer cells in vitro. A, LGR5 expression was determined by RT-PCR (top) and Western blot (bottom) analyses in colon cancer cell lines.B,Western blot analyses show that HA-tagged LGR5 was stably expressed in CBS cells under normal condition (top). B and C, Vector- and LGR5/HA-expressing CBS cells were subjected to GFDS for 72 hours. Apoptosis was determined by cleaved caspase-9 (B, bottom) and DNA fragmentation assays (C). D, Colony formation assays were performed in vector- and LGR5/HA-expressing CBS cells, with representative images (left) and quantification (right) shown. E, Soft agarose assays were performed in vector- and LGR5/HA-expressing CBS cells, with representative images (left) and quantification (right) shown. F, TGFb RI expression was knocked down in FET cells. G, FET control and RI knockdown cells were subjected to GFDS while treated with RSPO1 or TGFb1 for 16 hours. Apoptosis was determined by DNA fragmentation assays. H, FET control and RI knockdown cells were treated with RSPO1 or TGFb1in soft agarose assays. Stained plates (top) and quantification (bottom) are shown. I, FET cells were treated with RSPO1 or TGFb1 and the RI kinase inhibitor under GFDS. Apoptosis was determined by DNA fragmentation assays. J, Soft agarose assays were performed in FET cells treated with RSPO1 or TGBb1 and the RI kinase inhibitor. Stained plates (left) and quantification (right) are shown. The data are presented as the mean SEM of three replications. , P < 0.01; , P < 0.001.

www.aacrjournals.org Cancer Res; 77(23) December 1, 2017 6595

Zhou et al.

TGFb1-mediated promoter activation (Fig. 5G). These ﬁndings complex. We therefore investigated whether LGR5 forms com- suggest that TGFb1/RII and RSPO1/LGR5 are intimately inter- plexes with TGFb RI and/or RII upon RSPO1 stimulation. To connected in activating TGFb signaling. One potential mecha- facilitate co-IP assays and IF staining to detect complex formation nism might comprise signaling through a common physical and colocalization respectively, we engineered HCT116 cells to

Figure 4. LGR5 suppresses colon cancer metastasis in vivo. A, Control and LGR5 knockdown CBS cells were orthotopically implanted into athymic nude mice. The incidence of liver and/or lung metastasis is shown. , P < 0.05; , P < 0.01 with power >80%. B, Representative images of H&E staining in primary tumors, liver, and lung metastases are shown. Scale bars, 200 mm. C, Representative GFP images of liver and lung metastases are shown. D, IHC staining of phosphorylated Smad2 was performed in primary tumors. Representative images are shown (left). The intensity of pSmad2 staining was quantiﬁed (right). Scale bars, 100 mm. The data are presented as the mean SEM. , P < 0.05.

6596 Cancer Res; 77(23) December 1, 2017 Cancer Research

R-Spondin1/LGR5 Activates TGFb in Colon Cancer Metastasis

express both Flag-tagged TGFb RII and HA-tagged LGR5 were observed in FET cells expressing endogenous LGR5 and RII. (HCT116-RII/LGR5; Fig. 6A). HCT116 cells express abundant When immunoprecipitated with the anti-RI antibody, RSPO1 or endogenous TGFb RI (Fig. 6A). Treatment of HCT116-RII/LGR5 TGFb1 increased the amount of LGR5 and RII associated with RI cells with either RSPO1 or TGFb1 increased Smad2 phosphory- (Fig. 6C). Of note, increased complex formation between TGFb RI lation (Fig. 6B–I), indicating the activation of the TGFb pathway. and RII induced by TGFb was expected from previous studies Protein lysates from nontreated control (C), RSPO1 (R)- or TGFb1 (41, 42). We have shown earlier that knockdown of TGFb RI, RII, (T)-treated cells were immunoprecipitated with an anti-LGR5, or LGR5 reduced RSPO1- or TGFb1-induced Smad reporter acti- anti-Flag (RII), or anti-RI antibody followed by Western blot vation (Fig. 5G). Taken together, these results suggest that the analysis of expression of LGR5, RII, and RI in the immunopre- complex formation of RI/RII/LGR5 may be important for RSPO1- cipitated complexes. Co-IP assays showed that, when immuno- or TGFb1-mediated TGFb signaling. precipitated with the anti-LGR5 antibody, RSPO1 or TGFb1 To further explore the hypothesis that LGR5 complexes with increased the amount of RII and RI associated with LGR5 as TGFb receptors, immunoﬂuorescence microscopy was used to compared to control samples (Fig. 6B–II). Similarly, when immu- determine whether LGR5 colocalized with TGFb RI and RII. noprecipitated with the anti-Flag antibody, which pulled down HCT116-RII/LGR5 cells treated with RSPO1 or TGFb1 were RII, the amount of LGR5 and RI associated with RII was increased labeled with an anti-HA (for LGR5), anti-RI, or anti-Flag (for RII) by RSPO1 or TGFb1 (Fig. 6B–III). When immunoprecipitated antibody followed by incubation with a secondary antibody with the anti-RI antibody, RSPO1 or TGFb1 increased the amount conjugated with either Alexa Fluor 488 (green) or Alexa Fluor of LGR5 and RII associated with RI (Fig. 6B–IV). Similar results 594 (red). Confocal microscopy analysis showed that in control

Figure 5. RSPO1/LGR5 activates TGFb signaling in colon cancer cells. A, FET and CBS cells were treated with 200 ng/mL RSPO1, 4 ng/mL TGFb1, or 200 nmol/L TGFb RI kinase inhibitor for 48 hours. Smad reporter activity was determined by dual luciferase assays. B, FET cells were treated with RSPO1 or TGFb1 for the indicated time. Phosphorylated Smad2 were determined. C, RNA-seq analysis was performed in triplicates on RNAs isolated from FET cells treated with RSPO1 or TGFb1 for 4 hours. Hierarchical clustering of regulated genes is shown. Red and green indicate genes upregulated and downregulated by RSPO1 or TGFb1, respectively. D, Venn diagram shows the numbers of genes commonly or exclusively regulated by RSPO1 or TGFb1. E, RT-PCR analysis of p21 and c-myc expression in FET cells treated with RSPO1 or TGFb1 for 4 hours. F, Smad reporter activity was determined in control and LGR5 knockdown cells (left). Smad2 phosphorylation was determined in FET control and LGR5 knockdown cells after TGFb1 treatment (right). G, Expression of TGFb RI, RII, or LGR5 was knocked down individually in FET cells (top). Scrambled control (Ctr) and knockdown cells were treated with RSPO1 or TGFb1 for 48 hours. Fold change of Smad activity was calculated as the ratio of TGFb1- or RSPO1-induced luciferase vs. that of nontreated controls (bottom). H, Expression of Smad2, Smad3, or Smad4 was knocked down individually in FET cells (top). Control and knockdown cells were treated with RSPO1 for 48 hours. Fold change of Smad activity was calculated as the ratio of RSPO1-induced luciferase versus that of nontreated controls (bottom). The data are presented as the mean SEM of three replications. , P < 0.01; , P < 0.001.

www.aacrjournals.org Cancer Res; 77(23) December 1, 2017 6597

Zhou et al.

I II III IP: Flag IV Input IP: LGR5 (RII) IP: RI

A B C R T C R T C R T C R T (kDa) V RII/ (kDa) LGR5 LGR5 100 LGR5 100 RII 75 RII 75 55 RI 55 C R T RI C Input IP: RI pSmad2 55 C R T C R T (kDa) Actin 43 Smad2 55 LGR5 100 Actin 43 D RII DAPI LGR5 RI Merge 75 RI 55

pSmad2 55

Control Smad2 55 Actin 43 E RSPO1 0.4

*** 0.3 β 1 ***

TGF 0.2

DAPI RII LGR5 Merge 0.1

Pearson coefficient 0.0 Control RSPO1 TGFβ1 Control 0.3 ***

0.2 RSPO1

β 1 0.1 TGF Pearson coefficient DAPI RII RI Merge 0.0 Control RSPO1 TGFβ1

0.4 *** Control 0.3 ***

0.2 RSPO1 0.1 Pearson coefficient β 1 0.0 Control RSPO1 TGFβ1 TGF

Figure 6. LGR5 forms complexes with TGFb receptors. A, HCT116 cells were stably transfected with Flag-tagged TGFb RII and HA-tagged LGR5 (HCT116-RII/LGR5). Expression of LGR5, RII, and RI was detected. B, HCT116-RII/LGR5 cells were treated with 200 ng/mL RSPO1 (R) or 4 ng/mL TGFb1 (T) for 30 minutes. Input (top) and pSmad2 expression (bottom) are shown (B, I). Complex formation between LGR5, RII, and RI was detected by immunoprecipitation, followed by Western blot analyses (B, II–IV). C, FET cells were treated with 200 ng/mL RSPO1 (R) or 4 ng/mL TGFb1 (T) for 30 minutes. Input (left top) and pSmad2 expression (left bottom) are shown. Complex formation between LGR5, RII, and RI was detected by immunoprecipitation of RI followed by Western blot analyses (right). D, HCT116-RII/LGR5 cells were treated with RSPO1 (R) or TGFb1 (T) for 30 minutes. Single-channel (green or red) and merged (yellow) microscopic images of distribution of LGR5 (green) and RI (red; top), RII (green) and LGR5 (red; middle), and RII (green) and RI (red; bottom) are shown. Scale bars, 50 mm. E, Statistical analyses of colocalization of LGR5/RI (top), RII/LGR5 (middle), and RII/RI (bottom) using Pearson correlation coefﬁcient were performed. , P < 0.001.

6598 Cancer Res; 77(23) December 1, 2017 Cancer Research

R-Spondin1/LGR5 Activates TGFb in Colon Cancer Metastasis

cells, LGR5, RI, and RII were each distributed mainly on the cell cells by activation of TGFb signalingmediatedbyWnt5a membrane (Fig. 6D). Treatment with either RSPO1 or TGFb1 receptor Ror2 and TGFb RI kinase activity (43). However, induced internalization of these receptors and increased coloca- Wnt5a does not mimic RSPO1 effect in activating TGFb sig- lization of LGR5/RI (top), RII/LGR5 (middle), and RII/RI (bot- naling in colon cancer cells (Supplementary Fig. S6B). There- tom) as shown by yellow color in merged images (Fig. 6D). fore, in different contexts, there may be cross-talks between Wnt Quantification of fluorescent signals indicated that RSPO1 or and TGFb signaling through different ligand/receptor-mediated TGFb1 increased aggregation of LGR5 and TGFb RI (Supplemen- mechanisms. tary Fig. S6A). In addition, Pearson's correlation coefficient used Both LGR4 and LGR5 are co-expressed in the stem cells of for describing the correlation of intensity distributions between intestine, colon, stomach, and hair follicle (44–46). LGR6 marks channels was calculated, which showed that the extent of colo- multipotent stem cells in the epidermis (47). We found that LGR5 calization between green and red channels of LGR5/RI (top) and is differentially expressed in colon cancer cell lines (Fig. 2A), RII/RI (bottom) was significantly enhanced by RSPO1 or TGFb1 whereas LGR4 is ubiquitously expressed and LGR6 is almost and extent of colocalization of RII/LGR5 (middle) was markedly undetectable in those cells (Supplementary Fig. S6C). When LGR5 enhanced by TGFb1 (Fig. 6E). expression was knocked down in CBS and FET cells, expression of LGR4 was not affected (Supplementary Fig. S6D), suggesting that LGR4 does not compensate for LGR5 loss. The function of LGR4 LGR5 expression positively correlates with TGFb signaling in in colon cancer remains to be determined. human patient specimens RNA-seq analysis of gene expression mediated by RSPO1 or To explore whether there is any potential clinical relevance of TGFb1 indicates that majority of genes regulated by TGFb1are the activation of the TGFb pathwaybyRSPO1/LGR5,we shared by RSPO1. However, approximately 60% of RSPO1- determined the relationship between LGR5 expression and mediated genes are regulated exclusively by RSPO1 (Fig. 5D). Smad2 phosphorylation (pSmad2) in colon cancer patient These results indicate that, despite of an overlap between samples using IHC staining. We found that tumors with high RSPO1- and TGFb1-mediated transcriptional program, RSPO1 pSmad2 also expressed high LGR5 (Fig. 7A). The study revealed has its own specific set of target genes that are not mediated by a strong correlation between LGR5 and pSmad2 positivity TGFb1. It suggests that, in addition to the activation of TGFb (expressed as percentage of positive cells; Fig. 7B, left) and a signaling, RSPO1 activates other signaling pathways. RSPO1/ moderate correlation between LGR5 and pSmad2 intensity LGR5 has been shown to modulate Wnt signaling (5, 8). (expressed as intensity of staining;Fig.7B,right).Therefore, Analysis of a ChIP-Seq dataset, which identified genes whose our data demonstrate the clinical relevance and biological genomic upstream regulatory regions exhibit significant b-cate- importance of RSPO1/LGR5-mediated activation of TGFb sig- nin and TCF binding in intestinal crypt epithelial cells (48), naling in colon cancer. shows that only 19 of 242 RSPO1-exclusive genes (less than Taken together, our studies suggest a novel model of cross-talk 8%) overlap with b-catenin/TCF target genes. These results between RSPO/LGR5 and TGFb/RII signaling (Fig. 7C). In this indicate that the majority of RSPO1-exclusive genes are not model, RSPO or TGFb increases association of LGR5, RI and RII, subject to Wnt-mediated transcriptional regulation. Further enhancing TGFb downstream signaling. Reduction of LGR5 or RII studies show that knockdown of b-catenin or expression of a expression will decrease TGFb activation, enabling cancer cells to dnTCF had little effect on RSPO1-mediated upregulation of escape TGFb tumor suppressor function, leading to cancer pro- RASA4 (RAS p21 protein activator 4) and TSPAN2 (tetraspanin gression, poor prognosis, and/or poor survival of colon cancer 2), two genes exclusively regulated by RSPO1 (Supplementary patients. Fig. S6E). Taken together, these results indicate that RSPO1 regulates downstream gene expression largely independent of Discussion Wnt signaling in colon cancer cells. In this study, we have shown that RSPO1/LGR5 directly We also showed that either TGFb1orRSPO1inducesthe activates TGFb signaling, inhibits cell survival and clonogeni- complex formation of RI/RII/LGR5 (Fig. 6) and that knock- city of colon cancer cells in vitro,andsuppressescoloncancer down of either RII or LGR5 reduced TGFb activation induced by metastasis in vivo. Mechanistically, upon stimulation by either TGFb1orRSPO1(Fig.5G).Takentogether,weproposea RSPO1, LGR5 colocalizes and forms complexes with TGFb RI model of TGFb signaling as shown in Figure 7C. Upon binding and RII, suggesting that LGR5 may be a novel component of the their respective ligands, TGFb RII and LGR5 form a complex TGFb signaling complex (Fig. 7C). Importantly, in support of with TGFb RI and activate it, leading to phosphorylation and the results in colon cancer cell lines, studies of clinical speci- activation of Smads downstream. Therefore, TGFb/RII and mens show that LGR5 expression positively correlates with the RSPO/LGR5 can transduce TGFb signaling collaboratively activation of TGFb signaling in colon cancer and that LGR5 through the activation of the downstream TGFb RI–Smad expression decreases in advanced stage of colon cancer. In pathway (Fig. 7C). Reduction of RII or LGR5 would decrease summary, this study discovers a tumor suppressor function of TGFb signaling, enabling colon cancer cells to escape TGFb LGR5 in colon cancer and establishes a novel cross-talk tumor suppressor function. In addition to colon cancer, studies between an intestinal stem cell marker and a well-established in squamous cell carcinoma (SCC) show that inactivation of tumor/metastasis suppressor pathway, which provides a new TGFb receptors coupled with hyperactivation of RAS/RAF/ avenue for understanding how intestinal stem cells contribute MAPK signaling or Tp53 mutation in LGR5-positive stem cells to tumor development and progression. drives rapid development of cutaneous SCC (49), suggests that A recent study discovered that expression of Wnt5a, the TGFb signaling inactivation could be a tumor-initiating event. noncanonical Wnt ligand, is upregulated during crypt regener- These studies provide compelling support for the tumor sup- ation, which inhibits proliferation of intestinal epithelial stem pressive role of TGFb signaling in carcinogenesis.

www.aacrjournals.org Cancer Res; 77(23) December 1, 2017 6599

Zhou et al.

Figure 7. LGR5 expression positively correlates with TGFb activity in colon cancer specimens. IHC staining of LGR5 and pSmad2 was performed in 49 colon cancer samples. A, Images of IHC staining of LGR5 and pSmad2 in three different colon tumors are shown. Scale bars, 100 mm. B, Quantiﬁcation of percentage of LGR5- and pSmad2-positive cells (left) and their staining intensity (right) was performed. Correlation of expression of LGR5 and pSmad2 was determined using Pearson test (n ¼ 49). Lines with 95% conﬁdence bands in dashed lines are the result of linear regression analysis [left, Y ¼ (38.29 9.13) þ (0.58 0.16)X; right, Y ¼ (37.27 5.71) þ (0.43 0.172)X]. C, Proposed model of TGFb signaling mediated by TGFb/RII and RSPO/LGR5 collaboratively. Upon binding, TGFb or RSPO, TGFb RII, and LGR5 form complexes with TGFb RI and activate it, leading to phosphorylation and activation of Smads downstream.

In summary, our studies have significant implications. Stud- metastasis. Therefore, LGR5 is another important tumor sup- ies of colon cancer specimens show that LGR5 expression pressor of colon cancer. increases in stage I/II tumors and decreases in stage III/IV tumors. These results suggest that upregulated LGR5 expres- fl sion, driven by Wnt activation at early stages of colon cancer, Disclosure of Potential Con icts of Interest No potential conflicts of interest were disclosed. enhances TGFb signaling as a fail-safe mechanism to prevent further tumor development. When tumors progress to late stages, LGR5 expression is suppressed by promoter methylation Authors' Contributions (50) or other unknown mechanisms, which leads to reduced Conception and design: X. Zhou, L. Geng, J. Wang TGFb signaling and facilitates colon cancer progression and Development of methodology: X. Zhou, L. Geng, J. Wang

6600 Cancer Res; 77(23) December 1, 2017 Cancer Research

R-Spondin1/LGR5 Activates TGFb in Colon Cancer Metastasis

Acquisition of data (provided animals, acquired and managed patients, funding (R01CA212241). H. Yi is supported by NIH/NCI funding provided facilities, etc.): X. Zhou, L. Geng, D. Wang, H. Yi, G. Talmon (R01CA215389). D. Wang is supported by NIH/NCI funding Analysis and interpretation of data (e.g., statistical analysis, biostatistics, (R15GM122006). G. Talmon is supported by Fred and Pamela Buffett computational analysis): X. Zhou, L. Geng, D. Wang, J. Wang Cancer Center (P30 CA036727). UNMC Advanced Microscopy Core Facility Writing, review, and/or revision of the manuscript: X. Zhou, L. Geng, J. Wang is supported by the Nebraska Research Initiative, the Fred and Pamela Buffett Study supervision: J. Wang Cancer Center (P30 CA036727), and an Institutional Development Award of NIH/NIGMSP30 GM106397. Acknowledgments We thank Janice A. Taylor and James R. Talaska of UNMC Advanced The costs of publication of this article were defrayed in part by the Microscopy Core Facility for providing assistance with microscopy. payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate Grant Support this fact. This work was supported by NIH/NCI funding (R01CA208063, R01CA215389, and R01CA212241 to J. Wang). X. Zhou is supported by Received January 27, 2017; revised June 16, 2017; accepted September 19, NIH/NCI funding (R01CA208063). L. Geng is supported by NIH/NCI 2017; published OnlineFirst September 22, 2017.

References 1. Hsu SY, Liang SG, Hsueh AJ. Characterization of two LGR genes homol- phosphoinositide 3-kinase/AKT/survivin pathway in colon cancer cells. ogous to gonadotropin and thyrotropin receptors with extracellular leu- Cancer Res 2008;68:3152–60. cine-rich repeats and a G protein-coupled, seven-transmembrane region. 18. Wiener Z, Band AM, Kallio P, Hogstrom J, Hyvonen V, Kaijalainen S, et al. Mol Endocrinol 1998;12:1830–45. Oncogenic mutations in intestinal adenomas regulate Bim-mediated apo- 2. McDonald T, Wang R, Bailey W, Xie G, Chen F, Caskey CT, et al. Identi- ptosis induced by TGF-beta. Proc Natl Acad Sci U S A 2014;111:E2229– fication and cloning of an orphan G protein-coupled receptor of the E2236. glycoprotein hormone receptor subfamily. Biochem Biophys Res Commun 19. Chowdhury S, Howell GM, Rajput A, Teggart CA, Brattain LE, Weber HR, 1998;247:266–70. et al. Identification of a novel TGFbeta/PKA signaling transduceome in 3. Barker N, Clevers H. Lineage tracing in the intestinal epithelium. Curr mediating control of cell survival and metastasis in colon cancer. PLoS Protoc Stem Cell Biol 2010.Chapter 5:Unit5A. ONE 2011;6:e19335. 4.CarmonKS,GongX,LinQ,ThomasA,LiuQ.R-spondins 20. Simms N, Rajput A, Sharratt EA, Ongchin M, Teggart CA, Wang J, et al. function as ligands of the orphan receptors LGR4 and LGR5 to Transforming growth factor-ss suppresses metastasis in a subset of human regulate Wnt/beta-catenin signaling. Proc Natl Acad Sci U S A 2011; colon carcinoma cells. BMC Cancer 2012;12:221. 108:11452–7. 21. Sodir NM, Chen X, Park R, Nickel AE, Conti PS, Moats R, et al. Smad3 5. de Lau W, Barker N, Low TY, Koo BK, Li VS, Teunissen H, et al. Lgr5 deficiency promotes tumorigenesis in the distal colon of ApcMin/þ mice. homologues associate with Wnt receptors and mediate R-spondin signal- Cancer Res 2006;66:8430–8. ling. Nature 2011;476:293–7. 22. Takaku K, Oshima M, Miyoshi H, Matsui M, Seldin MF, Taketo MM. 6. Binnerts ME, Kim KA, Bright JM, Patel SM, Tran K, Zhou M, et al. R- Intestinal tumorigenesis in compound mutant mice of both Dpc4 (Smad4) Spondin1 regulates Wnt signaling by inhibiting internalization of LRP6. and Apc genes. Cell 1998;92:645–56. Proc Natl Acad Sci U S A 2007;104:14700–5. 23. Grady WM, Myeroff LL, Swinler SE, Rajput A, Thiagalingam S, Lutterbaugh 7. Kim KA, Wagle M, Tran K, Zhan X, Dixon MA, Liu S, et al. R-Spondin family JD, et al. Mutational inactivation of transforming growth factor beta members regulate the Wnt pathway by a common mechanism. Mol Biol receptor type II in microsatellite stable colon cancers. Cancer Res 1999; Cell 2008;19:2588–96. 59:320–4. 8. Walker F, Zhang HH, Odorizzi A, Burgess AW. LGR5 is a negative 24. Salovaara R, Roth S, Loukola A, Launonen V, Sistonen P, Avizienyte E, et al. regulator of tumourigenicity, antagonizes Wnt signalling and regulates Frequent loss of SMAD4/DPC4 protein in colorectal cancers. Gut 2002; cell adhesion in colorectal cancer cell lines. PLoS ONE 2011;6: 51:56–9. e22733. 25. Veenendaal LM, Kranenburg O, Smakman N, Klomp A, Borel R, I, van Diest 9. Haegebarth A, Clevers H. Wnt signaling, lgr5, and stem cells in the intestine PJ. Differential Notch and TGFbeta signaling in primary colorectal tumors and skin. Am J Pathol 2009;174:715–21. and their corresponding metastases. Cell Oncol 2008;30:1–11. 10. Garcia MI, Ghiani M, Lefort A, Libert F, Strollo S, Vassart G. LGR5 deficiency 26. Bacman D, Merkel S, Croner R, Papadopoulos T, Brueckl W, Dimmler A. deregulates Wnt signaling and leads to precocious Paneth cell differenti- TGF-beta receptor 2 downregulation in tumour-associated stroma worsens ation in the fetal intestine. Dev Biol 2009;331:58–67. prognosis and high-grade tumours show more tumour-associated macro- 11. Barker N, Ridgway RA, van Es JH, van de WM, Begthel H, van den BM, et al. phages and lower TGF-beta1 expression in colon carcinoma: a retrospective Crypt stem cells as the cells-of-origin of intestinal cancer. Nature 2009; study. BMC Cancer 2007;7:156. 457:608–11. 27. Brattain MG, Brattain DE, Fine WD, Khaled FM, Marks ME, Kimball PM, 12. Metcalfe C, Kljavin NM, Ybarra R, de Sauvage FJ. Lgr5þ stem cells are et al. Initiation and characterization of cultures of human colonic carci- indispensable for radiation-induced intestinal regeneration. Cell Stem Cell noma with different biological characteristics utilizing feeder layers of 2014;14:149–59. confluent fibroblasts. Oncodev Biol Med 1981;2:355–66. 13. Wu C, Qiu S, Lu L, Zou J, Li WF, Wang O, et al. RSPO2-LGR5 signaling has 28. Eltonsy N, Gabisi V, Li X, Russe KB, Mills GB, Stemke-Hale K. Detection tumour-suppressive activity in colorectal cancer. Nat Commun 2014; algorithm for the validation of human cell lines. Int J Cancer 2012;131: 5:3149. E1024–E1030. 14. Tsuji S, Kawasaki Y, Furukawa S, Taniue K, Hayashi T, Okuno M, et al. The 29. Boyd DD, Levine AE, Brattain DE, McKnight MK, Brattain MG. Com- miR-363-GATA6-Lgr5 pathway is critical for colorectal tumourigenesis. parison of growth requirements of two human intratumoral colon Nat Commun 2014;5:3150. carcinoma cell lines in monolayer and soft agarose. Cancer Res 15. Seshagiri S, Stawiski EW, Durinck S, Modrusan Z, Storm EE, Conboy CB, 1988;48:2469–74. et al. Recurrent R-spondin fusions in colon cancer. Nature 2012; 30. Geng L, Chaudhuri A, Talmon G, Wisecarver JL, Are C, Brattain M, et al. 488:660–4. MicroRNA-192 suppresses liver metastasis of colon cancer. Oncogene 16. Massague J, Seoane J, Wotton D. Smad transcription factors. Genes Dev 2013;13;33:5332–40. 2005;19:2783–810. 31. Robinson MD, Oshlack A. A scaling normalization method for dif- 17. Wang J, Yang L, Yang J, Kuropatwinski K, Wang W, Liu XQ, et al. Trans- ferential expression analysis of RNA-seq data. Genome Biol 2010;11: forming growth factor beta induces apoptosis through repressing the R25.

www.aacrjournals.org Cancer Res; 77(23) December 1, 2017 6601

Zhou et al.

32. Robinson MD, McCarthy DJ, Smyth GK. edgeR: a Bioconductor package for 41. Wrana JL, Attisano L, Carcamo J, Zentella A, Doody J, Laiho M, et al. TGF differential expression analysis of digital gene expression data. Bioinfor- beta signals through a heteromeric protein kinase receptor complex. Cell matics 2010;26:139–40. 1992;71:1003–14. 33. Yokoi K, Thaker PH, Yazici S, Rebhun RR, Nam DH, He J, et al. Dual 42. Wrana JL, Attisano L, Wieser R, Ventura F, Massague J. Mechanism of inhibition of epidermal growth factor receptor and vascular endothelial activation of the TGF-beta receptor. Nature 1994;370:341–7. growth factor receptor phosphorylation by AEE788 reduces growth and 43. Miyoshi H, Ajima R, Luo CT, Yamaguchi TP, Stappenbeck TS. Wnt5a metastasis of human colon carcinoma in an orthotopic nude mouse potentiates TGF-beta signaling to promote colonic crypt regeneration after model. Cancer Res 2005;65:3716–25. tissue injury. Science 2012;338:108–13. 34. Skrzypczak M, Goryca K, Rubel T, Paziewska A, Mikula M, Jarosz D, et al. 44. Barker N, van Es JH, Kuipers J, Kujala P, van den BM, Cozijnsen M, et al. Modeling oncogenic signaling in colon tumors by multidirectional anal- Identiﬁcation of stem cells in small intestine and colon by marker gene yses of microarray data directed for maximization of analytical reliability. Lgr5. Nature 2007;449:1003–7. PLoS ONE 2010;5. 45. Barker N, Huch M, Kujala P, van de WM, Snippert HJ, van Es JH, et al. Lgr5 35. Femia AP, Dolara P, Salvadori M, Caderni G. Expression of LGR-5, MSI-1 (þve) stem cells drive self-renewal in the stomach and build long-lived and DCAMKL-1, putative stem cell markers, in the early phases of 1,2- gastric units in vitro. Cell Stem Cell 2010;6:25–36. dimethylhydrazine-induced rat colon carcinogenesis: correlation with 46. Jaks V, Barker N, Kasper M, van Es JH, Snippert HJ, Clevers H, et al. Lgr5 nuclear beta-catenin. BMC Cancer 2013;13:48. marks cycling, yet long-lived, hair follicle stem cells. Nat Genet 2008; 36. Zhou WJ, Geng ZH, Spence JR, Geng JG. Induction of intestinal stem cells 40:1291–9. by R-spondin 1 and Slit2 augments chemoradioprotection. Nature 47. Snippert HJ, Haegebarth A, Kasper M, Jaks V, van Es JH, Barker N, et al. Lgr6 2013;501:107–11. marks stem cells in the hair follicle that generate all cell lineages of the skin. 37. Inbal B, Cohen O, Polak-Charcon S, Kopolovic J, Vadai E, Eisenbach L, et al. Science 2010;327:1385–9. DAP kinase links the control of apoptosis to metastasis. Nature 1997; 48. Schuijers J, Mokry M, Hatzis P, Cuppen E, Clevers H. Wnt-induced 390:180–4. transcriptional activation is exclusively mediated by TCF/LEF. EMBO J 38. Mehrotra S, Languino LR, Raskett CM, Mercurio AM, Dohi T, Altieri DC. 2014;33:146–56. IAP regulation of metastasis. Cancer Cell 2010;17:53–64. 49. Cammareri P, Rose AM, Vincent DF, Wang J, Nagano A, Libertini S, et al. 39. Guo XN, Rajput A, Rose R, Hauser J, Beko A, Kuropatwinski K, et al. Mutant Inactivation of TGFbeta receptors in stem cells drives cutaneous squamous PIK3CA-bearing colon cancer cells display increased metastasis in an cell carcinoma. Nat Commun 2016;7:12493. orthotopic model. Cancer Res 2007;67:5851–8. 50. de Sousa E Melo, Colak S, Buikhuisen J, Koster J, Cameron K, de Jong 40. Jiang Y, Liu XQ, Rajput A, Geng L, Ongchin M, Zeng Q, et al. Phosphatase JH, et al. Methylation of cancer-stem-cell-associated Wnt target genes PRL-3 is a direct regulatory target of TGFbeta in colon cancer metastasis. predicts poor prognosis in colorectal cancer patients. Cell Stem Cell Cancer Res 2011;71:234–44. 2011;9:476–85.

6602 Cancer Res; 77(23) December 1, 2017 Cancer Research

R-Spondin1/LGR5 Activates TGFβ Signaling and Suppresses Colon Cancer Metastasis

Xiaolin Zhou, Liying Geng, Degeng Wang, et al.

Cancer Res 2017;77:6589-6602. Published OnlineFirst September 22, 2017.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-17-0219

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2017/09/22/0008-5472.CAN-17-0219.DC1

Cited articles This article cites 48 articles, 15 of which you can access for free at: http://cancerres.aacrjournals.org/content/77/23/6589.full#ref-list-1

Citing articles This article has been cited by 7 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/77/23/6589.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at Subscriptions [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/77/23/6589. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.