Wnt Pathway Inhibition Via the Targeting of Frizzled Receptors Results in Decreased Growth and Tumorigenicity of Human Tumors

Home , Frizzled, FZD2, FZD8, WNT3

Wnt pathway inhibition via the targeting of Frizzled receptors results in decreased growth and tumorigenicity of human tumors

Austin Gurney1, Fumiko Axelrod, Christopher J. Bond, Jennifer Cain, Cecile Chartier, Lucas Donigan, Marcus Fischer, Aurélie Chaudhari, May Ji, Ann M. Kapoun, Andrew Lam, Sasha Lazetic, Shirley Ma, Satyajit Mitra, In-Kyung Park, Kellie Pickell, Aaron Sato, Sanjeev Satyal, Michelle Stroud, Hoang Tran, Wan-Ching Yen, John Lewicki, and Timothy Hoey

OncoMed Pharmaceuticals, Redwood City, CA 94063

Edited by Roeland Nusse, Stanford University School of Medicine, Stanford, CA, and approved June 6, 2012 (received for review December 6, 2011) The Wnt/β-catenin pathway, which signals through the Frizzled (Fzd) signaling. This selectivity relative to potential intracellular sig- receptor family and several coreceptors, has long been implicated in naling components may offer advantages in safety and thera- cancer. Here we demonstrate a therapeutic approach to targeting peutic index. Previous studies have indicated that use of soluble the Wnt pathway with a monoclonal antibody, OMP-18R5. This an- Wnt inhibitors and decoy receptor molecules can impact tumor tibody, initially identified by binding to Frizzled 7, interacts with five growth (11, 12). In this article we demonstrate the ability to Fzd receptors through a conserved epitope within the extracellular generate a potent Wnt inhibitory antibody that functions by domain and blocks canonical Wnt signaling induced by multiple Wnt binding to select Fzd receptors, and that this antibody is active family members. In xenograft studies with minimally passaged hu- across a range of human tumor types. man tumors, this antibody inhibits the growth of a range of tumor types, reduces tumor-initiating cell frequency, and exhibits synergis- Results tic activity with standard-of-care chemotherapeutic agents. Identification of an Antagonistic Wnt Pathway Antibody That Targets Multiple Fzd Receptors. Antibodies to the Fzd receptors were differentiation | cancer stem cell | pancreatic | breast | lung identified using phage display and were subsequently screened for the ability to inhibit Wnt3A signaling in a cell-based assay. nvestigation into the role of the Wnt pathway in human tumors Through this approach, an antibody that was initially isolated by Ihas been hampered by a lack of therapeutic agents able to ability to bind to FZD7, OMP-18R5, was found to block most inhibit Wnt signaling. The Wnt pathway is complex, with a large β-catenin signaling in response to Wnt3A (Fig. 1A). In sub- number of known ligands, receptors, coreceptors, and regulatory sequent experiments it was also found to inhibit β-catenin sig- components. Members of the Wnt family have been shown to naling in response to each of the Wnt family members that were induce several distinct signaling events, including activation of active in these reporter studies (Fig. 1B). OMP-18R5 also re- the β-catenin signaling (termed the “canonical pathway”) as well duced both Wnt3A-induced accumulation of β-catenin and as other signaling cascades, including the planar cell-polarity phosphorylation of LRP6 (Fig. 1C). OMP-18R5 did not inhibit pathway and the Ca+2 pathway (1–5). Lack of a detailed un- β-catenin signaling in response to intracellular pathway activa- derstanding of the receptor-ligand specificities of the various tion with the GSK3 inhibitor compound BIO (6-bromoindirubin- pathway components and their relationship to the Wnt-mediated 3′-oxime) (Fig. S1A) and also did not inhibit activation of signaling pathways has also hindered development of therapeutic a Notch pathway reporter in response to the Notch ligand DLL4 agents. However, mutations in the Wnt signaling pathway occur (Fig. S1B). The OMP-18R5 anti-Fzd antibody was also screened in most cases of human colon cancer, and active signaling has by flow cytometry with cells overexpressing each individual Fzd. been noted in multiple major tumor types, highlighting this OMP-18R5 was found to bind 5 of the 10 Fzd receptors: FZD1, pathway as a potentially promising target for the development of FZD2, FZD5, FZD7, and FZD8 (Fig. 2A and Fig. S2). In- new anticancer agents (6, 7). In recent years, several small terestingly, each of these receptors has been previously impli- molecules that impact the Wnt pathway have been identified. cated in canonical Wnt signaling (12–17). The epitope on the Fzd The nonsteroidal anti-inflammatory compound Sulindac and the proteins bound by OMP-18R5 was mapped through a survey of antibiotic salinomycin have been reported to possess ability to introduced amino acid substitutions within FZD8, several of down-regulate β-catenin (8, 9). Inhibitors of the poly-ADP which eliminated OMP-18R5 binding (Fig. 2B and Fig. S3). ribosylating enzymes tankyrase 1 and tankyrase 2 also have been OMP-18R5 binds to a discontinuous epitope that spans a “cleft” found to be inhibitors of Wnt signaling, by a mechanism that region that is apparent in the reported crystal structure (18) of results in stabilization of axin and the β-catenin destruction complex (10). However, these agents also impact other signaling events beyond the Wnt pathway. To date, highly potent small Author contributions: A.G., F.A., C.C., A.M.K., S.L., I.-K.P., A.S., W.-C.Y., J.L., and T.H. molecule agents that specifically inhibit Wnt signaling have not designed research; A.G., F.A., C.J.B., J.C., C.C., L.D., M.F., A.C., M.J., A.L., S.L., S. Ma, S. Mitra, I.-K.P., K.P., A.S., S.S., M.S., H.T., and W.-C.Y. performed research; A.G., F.A., C.J.B., J.C., C.C.,

been reported. CELL BIOLOGY L.D., M.F., A.M.K., S.L., S. Ma, S. Mitra, I.-K.P., A.S., S.S., H.T., W.-C.Y., J.L., and T.H. analyzed An alternative approach to developing agents that block Wnt data; and A.G. and T.H. wrote the paper. – signaling is to focus on the extracellular receptor ligand inter- Conflict of interest statement: All of the authors are employees of OncoMed Pharma- actions. There are 19 human Wnts and 10 Frizzled (Fzd) re- ceuticals, which provided research funding. R.N. is a member of the OncoMed Scientific ceptors. In addition, Wnt signaling through Fzds involves two Advisory Board and holds stock in the company. coreceptors, LRP5 and LRP6, as well as other coreceptors, such This article is a PNAS Direct Submission. as ROR and Ryk. Although this substantial complexity and Freely available online through the PNAS open access option. functional redundancy makes the prospect of developing effec- 1To whom correspondence should be addressed. E-mail: [email protected]. tive inhibitors more challenging, it also offers the potential for This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. the identification of agents that block only portions of Wnt 1073/pnas.1120068109/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1120068109 PNAS | July 17, 2012 | vol. 109 | no. 29 | 11717–11722 Downloaded by guest on September 28, 2021 Fig. 1. Anti-Fzd antibody OMP-18R5 inhibits Wnt signaling. (A) OMP-18R5 antibody inhibits recombinant Wnt3A signaling as assessed by β-catenin responsive TOP-FLASH luciferase reporter and a FOP-FLASH control reporter. (B) OMP-18R5 antibody (50 nM) inhibits the ability of several Wnts to induce canonical signaling. Shown is relative luciferase signal normalized to control Renilla luciferase and relative to signal in the absence of Wnt (no Wnt). Bars represent the mean ± SD. (C) OMP-18R5 blocks phosphorylation of LRP6 and the accumulation of active β-catenin. HEK-293 cells were treated with puriﬁed Wnt3A and OMP- 18R5, as indicated. Cells were lysed in the presence of phosphatase inhibitor and Western blot analysis was performed. OMP-18R5 inhibited the induction of phosphorylated (Ser1490) LRP6 by Wnt3A and attenuated the Wnt induced accumulation of β-catenin and unphosphorylated (Ser37, Thr41) β-catenin.

mouse Fzd8 (Fig. 2C). Interestingly, the residues lining the base OMP-18R5 directly blocked the ability of Wnt to interact with of this cleft are highly conserved across the Fzd family, suggesting Fzd, binding studies were conducted. OMP-18R5 blocked that this site is functionally important. To determine whether Wnt3A binding to Fzd5 ECD, suggesting that a mechanism by

Fig. 2. Anti-Fzd antibody OMP-18R5 binds to 5 of the 10 human Fzd receptors and inhibits Wnt binding. (A) Binding curves of OMP-18R5 to cells that overexpress the indicated Fzd. Binding was assessed by ﬂow cytometry analysis with cells transfected with cDNA encoding the indicated FZD protein. (B) Epitope mapping of OMP-18R5 to FZD8. Expression vectors comprising N-terminal FLAG-tagged FZD8 variants encoding the indicated amino acid substitution were transiently transfected along with GFP and OMP-18R5 binding was assessed by ﬂow cytometry. (C) Surface rendering of the cysteine-rich domain of FZD8 (18), highlighting important residues involved in OMP-18R5 binding in green and highly conserved residues in pink. (D) Binding of OMP-18R5 to FZD inhibits the interaction of Wnt with FZD. Wnt3A, FZD5-Fc and OMP-18R5 were coincubated as indicated, and then OMP-18R5 and FZD5-Fc were removed by protein A immunoprecipitation. Supernatant was then assayed for Wnt activity in an 8xTCF luciferase reporter assay.

11718 | www.pnas.org/cgi/doi/10.1073/pnas.1120068109 Gurney et al. Downloaded by guest on September 28, 2021 which OMP-18R5 inhibits Wnt signaling may involve the direct chemotherapy, whereas tumors treated with control antibody inhibition of Wnt binding (Fig. 2D and Fig. S4). demonstrated rapid regrowth. Several tumors showed synergistic response to OMP-18R5 and Wnt Pathway Blockade Inhibits the Growth of Human Tumor Xenografts. chemotherapy. Gene-expression analysis of the response of The impact of inhibiting Wnt/β-catenin signaling on tumor growth breast tumor PE-13 to OMP-18R5 alone or in combination with was assessed using human tumor xenografts in mice. These models taxol indicated that OMP-18R5 and taxol induced opposing were conducted with minimally passaged human tumors (19, 20) changes in the expression of numerous genes involved in stress (Table S1). OMP-18R5 treatment resulted in inhibition of growth responses, as well as ABC family transporters, suggesting Wnt in several types of human tumors, including breast, pancreatic, pathway inhibition may alter cellular responses to taxol treat- colon [wild-type adenomatous polyposis coli (APC) and β-catenin] ment (Fig. S5). mRNA for several genes previously associated and lung tumors (Fig. 3). Striking synergy was observed when with breast tumor tumorigenicity (20–22), including CD44, OMP-18R5 was combined with several standard-of-care chemo- ALDH1A1, SOX1, and SOX2, were decreased by OMP-18R5. therapeutic agents, including taxol in nonsmall-cell lung cancer Similarly, OMP-18R5 also displayed synergy with gemcitabine in and breast cancer models, irinotecan in colon cancer models, and pancreatic tumors. OMP-18R5 was noted to induce changes in gemcitabine in pancreatic cancer models (Fig. 3 B, C, E, and G–I). the histopathology of the pancreatic tumors. In PN4, a pancre- OMP-18R5 had activity in 6 of 11 pancreatic tumors, 3 of 6 breast atic tumor with a moderately differentiated ductal phenotype, tumors, and 7 of 8 nonsmall-cell lung tumors tested. OMP-18R5 the ducts were greatly enlarged and the cells lining the ducts did not display activity in colon tumors harboring APC or β-cat- were larger following OMP-18R5 treatment. Staining with enin mutations, but was highly active in a colon tumor with wild- Alcian blue, to reveal mucin expression, revealed large increases type APC and β-catenin. OMP-18R5 also inhibited the growth of in mucin content (Fig. 4 A and B). Immunohistochemical (IHC) the established cell line PA-1 (derived from a human teratocar- staining also revealed reduced nuclear β-catenin levels following cinoma), which had previously shown to be inhibited by FZD8-fc OMP-18R5 treatment. (Fig. 4 C and D). Proliferation, as mea- F G protein (12) (Fig. 3 ). In tumor recurrence studies (Fig. 3 and sured by Ki-67 staining, was also reduced by OMP-18R5 treat- H), OMP-18R5 treatment induced an extended delay in the ment (Fig. 4 E and F). Quantitative PCR of known targets of regrowth of tumors following a treatment with high-dose Wnt/β-catenin signaling Wnt target genes and genes associated CELL BIOLOGY

Fig. 3. Inhibition of canonical Wnt signaling inhibits tumor growth as monotherapy and in combination with chemotherapy. Results of nine tumor models: colon tumor C28 (A), breast tumor T3 (B), lung NSCLC Lu24 (C), pancreatic tumor PN8 (D), breast tumor PE13 (E), teratocarcinoma line PA1 (F), pancreatic tumor PN4 (G), breast tumor PE13 (H), and colon tumor C28 (I). Mean tumor volumes with SEs are presented (n =10).(G and H) Duration of chemotherapy and antibody treatment is indicated by brackets. Asterisks denote statistical signiﬁcance P < 0.02 (* vs. control Ab; ** vs. chemotherapy alone).

Gurney et al. PNAS | July 17, 2012 | vol. 109 | no. 29 | 11719 Downloaded by guest on September 28, 2021 Targeting Fzd Receptors Reduces Tumorigenicity. Several studies have suggested that Wnt signaling plays a key role in mediating self-renewal essential for tumor initiating cells (25–27). To assess the impact of OMP-18R5 on tumor-initiating cell frequency, limiting dilution studies were conducted wherein deﬁned num- bers of human cells from treated or control tumors were assessed for their ability to initiate tumor growth after secondary transplant into a new cohort of mice in the absence of further treatment. OMP-18R5 treatment of the pancreatic tumor PN4 (Fig. 6A) and the breast tumor PE13 (Fig. 6B) reduced the tumor- initiating cell frequency of the tumor cells by threefold. Neither gemcitabine treatment of the pancreatic tumor nor taxol treatment of the breast tumor reduced tumorigenic cell frequency (in fact, taxol treatment increased tumorigenicity, P < 0.05). In contrast, the combination of OMP-18R5 and chemotherapeutic agents reduced tumorigenicity by 10-fold in both pancreas and breast tumors.

Intracellular Blockade of the Canonical Wnt Pathway Inhibits Tumor Growth. The OMP-18R5 antibody blocks Wnt signaling by binding to multiple Fzd receptors, suggesting a requirement for activation of the canonical Wnt pathway during tumor growth. Although OMP-18R5 is a human IgG2 isotype—an isotype that mediates little effector function, such as complement mediated cytotoxicity or antibody dependent cell killing—and further that these studies were performed in immunocompromised mice, it is conceivable that immune function contributes to OMP-18R5 efficacy. To probe Wnt pathway involvement further, we tested the impact of an alternative approach to Wnt pathway inhibition. Overexpres- sion of the intracellular Wnt pathway component Axin leads to Fig. 4. Inhibition of Wnt signaling in pancreas tumor induces mucininous β differentiation and decreased nuclear β-catenin and cellular proliferation. destabilization of -catenin and inhibition of canonical signaling Shown is staining of sections of pancreatic tumor PN4 with Alcian blue to (28). We find that introduction of Axin into tumors by lentivirus- highlight mucin (A and B), and with anti–β-catenin (C and D) and with anti- mediated transduction substantially abrogated tumor growth (Fig. Ki-67 to indicate proliferating cells (E and F). Sections are from tumors fol- S6), suggesting importance of canonical Wnt pathway signaling lowing treatment (41 d) with control antibody (A, C, and E) or OMP-18R5 (B, within the tumor cells. D, and F). Discussion with pancreatic lineages was performed (Fig. 5). Down-regula- The role of the Wnt-Fzd pathway in cancer has been the subject tion of AXIN2 and osteopontin (SPP1) was observed in tumors of investigation for nearly three decades following the initial treated with OMP-18R5, consistent with inhibition of Wnt sig- observation that mouse mammary tumor virus retroviral in- naling (23, 24). In addition, in these pancreatic xenografts, tegration events resulting in inappropriate Wnt1 expression multiple genes within the mucin family and certain cytokeratins could result in murine mammary gland tumors (29). Recognition β were induced by OMP-18R5. Gemcitabine treatment resulted in of the importance of APC and -catenin mutations in colon elevated mRNA for several genes associated with epithelieal- cancer has provided strong evidence that the pathway plays mesenchymal transition, including fibronectin, vimentin, slug, a major role in the genesis of at least one major type of human and snail. These effects were blocked by combined treatment cancer (30, 31). Although not as frequently mutated in other tumor types, increased Wnt pathway activation and epigenetic with OMP-18R5. silencing of Wnt antagonists has been noted in several tumor types (7, 32–37). In the present study, we provide evidence that it is possible to generate potent inhibitors of canonical Wnt signaling by simultaneous targeting of several Fzd receptors. The antibody described in this study binds to five distinct Fzd receptors through a conserved epitope. Interestingly, this epitope is distinct from the region of FZD8, previously implicated in Wnt binding by alanine scanning mutation analysis (18). The structure of Wnt has not yet been reported, but both Wnt and the antibody are substantially larger proteins than the relatively small cysteine-rich domain of Fzd receptors, so it is possible that steric hindrance prevents simultaneous binding of Wnt and antibody. Our data Fig. 5. Expression changes induced by Wnt pathway inhibition. Shown is indicate that blockade of Wnt signaling can inhibit growth in the relative mRNA expression (normalized to control antibody and to human a variety of human tumor types. Inhibition of tumor growth was control mRNA GAPDH) for the indicated human genes from pancreatic PN4 noted in lung, breast, colon, and pancreatic tumors. Interestingly, tumors treated with OMP-18R5, gemcitabine, or the combination. Data there was strong synergy observed with several chemotherapy shown is average from four tumors in each treatment group analyzed in- fi ’ < agents, particularly taxol. Gene-expression analysis suggests that dividually. Shown are genes with statistically signi cant (Student s t test P “ ” 0.05) expression difference in response to OMP-18R5 or combination treat- rather than enhancing the magnitude of taxol-induced effects, ment relative to either control or gemcitabine treatment arms) with the the combination with OMP-18R5 actually inhibits some taxol- exception of CDH1 which is included for completeness. induced gene changes and may therefore be altering cellular

11720 | www.pnas.org/cgi/doi/10.1073/pnas.1120068109 Gurney et al. Downloaded by guest on September 28, 2021 Fig. 6. Inhibition of Wnt reduces tumorigenicity and induces the presence of nontumorigenic cells. Limiting dilution analysis of the ability of human cells isolated from treated pancreatic PN4 tumor (A) and breast tumor PE13 (B) to initiate tumor growth in the absence of further treatment following secondary transplant to a new cohort of mice. The primary tumor was treated with control antibody or OMP-18R5 or chemotherapy, as indicated. The tumor initiating frequency (TIC) refers to the average number of cells determined to be required to cause tumor growth in the recipient cohort.

response to taxol treatment. The robust ability of overexpressed Materials and Methods axin within tumors to also inhibit tumor growth is evidence that All animal work was performed according to OncoMed Pharmaceutical’s the tumor cells themselves are dependent upon this pathway. Institutional Animal Care and Use Committee guidelines. Generally, single-agent activity was observed to be significant, but modest relative to the efficacy observed in combination with Antibody Generation and Reporter Studies. OMP-18R5 was generated by chemotherapy. This finding may reflect the fact that OMP-18R5 panning the HuCAL GOLD phage-display library (MorphoSys) with recombinant does not target all Wnt pathway receptors. However, it has FZD7 protein (42). DNA fragments encoding the fAb generated from the previously been demonstrated that robust Wnt inhibition by phage display library were subcloned into a full-length human IgG2 expression fi adenovirus-mediated delivery of the pan-Wnt pathway inhibitor vector. The antibody was expressed in CHO cells and puri ed. The ability of Wnt to activate T-cell factor (TCF)-dependent transcription was assessed by dickkopf 1 (DKK1) results in profound inhibition of gastroin- transient transfection of HEK293 responder cells with the TOP-FLASH/FOP- testinal tract homeostasis with disruption of the crypt-villus ep- FLASH luciferase reporter assay and a Renilla luciferase transfection control ithelial cell layer (38). Because of the high degree of evolutionary reporter, followed by treatment with purified Wnt3A (R&D Systems) or 24-h conservation in this pathway, the OMP-18R5 binds to both hu- coculture (1:1) exposure to HEK293 cells transiently transfected with expres- man and mouse Fzd receptors. In fact, within the cysteine-rich sion vectors encoding human Wnt proteins, and then assayed using the Dual- domain FZD1, -2, -7, and -8 are identical between human and Glo luciferase assay reporter system (Promega). The TOP-FLASH reporter was mouse, and FZD5 has only two amino acid changes that do not synthesized and includes eight copies of a TCF binding site (AGATCAAAGG) impact OMP-18R5 binding (Fig. S2). There was no apparent upstream of a minimal promoter and firefly luciferase. BIO was obtained from impact of OMP-18R5 on the gastrointestinal tract at the doses Sigma-Aldrich. The ORF cDNA encoding human Fzd, Wnt, Smoothened, and used in these studies. At very large doses (100 mg/kg) we did Notch2 proteins were isolated by PCR or synthesized (DNA2.0). The Notch luciferase reporter assay was performed essentially as previously described (19). observe an intestinal colitis that resembled, although was less Briefly, PC3 cells were transfected with an expression vector encoding full- severe, the results reported with DKK1. OMP-18R5 did however length human Notch2 and an 8xCBF1-luciferase reporter vector and a Renilla substantially inhibit the expression of several known Wnt target luciferase vector as a transfection control. Cells were plated in wells coated genes in the liver (Fig. S7) (39). with 100 ng of human DLL4 (R&D Systems) and exposed to the indicated OMP-18R5 reduces tumor cell proliferation and tumor-initi- antibodies at 10 μg/mL and then assayed 18 h later using the Dual-Glo lucif- ating cell frequency. The reduction in tumor-initiating cell fre- erase system. Fzd-Fc fusion proteins were produced in baculovirus and con- quency was quantified using limiting-dilution tumorigenicity tained the N-terminal extracellular domain of human Fzd receptors fused to assays. This functional assay measures in vivo tumorigenicity and the CH2-CH3 domains of human IgG1. Lentiviral vectors containing CMV- makes no assumption about the frequency, FACS marker profile, mouse AXIN-IRES-GFP or CMV-IRES-GFP cassettes were generated by trans- or heterogeneity of the tumor-initiating cell population. A strength complementation in HEK 293 cells transiently cotransfected with the plasmid containing the vector genome along with the gag-pol, rev, and VSV-G env of this approach is that it can be applied to any tumor. In contrast, packaging constructs. For lentiviral transduction studies, tumor cells were cell-surface markers may vary in different tumors and their ex- isolated as single cells and cultured (43) in the presence of lentivirus for 24 h. pression may be influenced by environment, which changes in re- Transduced cells were then isolated by flow cytometry gating on positive sponse to drug treatment. OMP-18R5 reduced tumor-initiating GFP expression. cell frequency both as a single agent and in combination with Histology and IHC, and Western Blottings. Formalin-fixed and paraffin-em- chemotherapy. In contrast neither gemcitabine nor taxol reduced CELL BIOLOGY tumorigenicity as single agents. Taxol treatment actually resulted bedded tumors were sectioned at 4-μm thickness. Sections were stained with in an increase in tumor-initiating cell frequency, consistent with Alcian blue as recommended by the manufacturer (Poly Scientific). The IHC –β – previous reports that tumor-initiating cells are selectively resistant analysis used anti -catenin (mAb2081, Millipore; 1:100) and anti Ki-67 (sp6, to various chemotherapeutics and radiation (40, 41). Vector Laboratories; 1:200) with optimal cutting temperature embedded frozen-tumor sections. For phospho-LRP6 and β-catenin Western blot anal- In conclusion, we have developed a unique antibody that ysis, HEK-293 cells were cultured in 2% (vol/vol) FBS containing DMEM for 4 antagonizes the Wnt signaling pathway. Targeting the Wnt h in the presence of 400 ng/mL recombinant Wnt3A (R&D Systems) and 50 pathway through by blockade of selected members of the Fzd μg/mL OMP-18R5, as indicated. Cells were lysed in lysis buffer (Invitrogen) receptor family may provide an efficacious approach for the containing 1 tablet/10 mL PhosSTOP phosphatase inhibitor (Roche). Western treatment of a broad range of tumors. blots were probed with α-LRP6 (Cell Signaling; C47E12), α-phospho LRP6

Gurney et al. PNAS | July 17, 2012 | vol. 109 | no. 29 | 11721 Downloaded by guest on September 28, 2021 (Ser1490; Cell Signaling), α-β-catenin (BD Transduction Laboratories; 610154), Version 1.1 software program (StemCell Technologies). OMP18R5 was dosed α−“active”-β-catenin (unphosphorylated Ser37, Thr-41; clone 8E7; Millipore). at 10 mg/kg twice weekly, except as indicated, and was dosed at 20 mg/kg once weekly in Fig. 4G and 15 mg/kg twice weekly in Fig. 4C. Taxol (pacli- Xenograft Models. The establishment of minimally passaged human tumor taxel) was dosed at 10 mg/kg once weekly except for Fig. 4H, which was xenograft models and determination of tumor initiating cell frequency was dosed at a maximum tolerated dose of 20 mg/kg twice weekly. Gemcitabine performed as previously described (19, 44). Brieﬂy, human tumors xenograft was dosed at 50 mg/kg twice weekly. Irinotecan was dosed at 7.5 mg/kg models were established by subcutaneous implantation of patient-derived twice weekly. All agents were administered intraperitoneally. solid tissue fragments in NOD/SCID mice. Established tumors were subsequently disassociated and single-cell suspensions of tumor cells were fro- Data Analysis. Data are expressed as mean ± SEM. Differences in mean values zen at −80C. Tumors for subsequent experiments were established by serial between groups were analyzed by nonparametric t test. Tumor-initiating implantation of frozen cell stocks. For tumorigenicity studies, single-cell cell frequency was determined using Poisson distribution statistics and suspensions from control and treated tumors were incubated with bio- L-Calc software. tinylated α-mouse CD45-biotin and α-mouse H2Kd on ice for 30 min followed by addition of streptavidin-labeled magnetic beads to remove murine stro- ACKNOWLEDGMENTS. We thank many people at OncoMed for their mal cells. Human tumor cells were collected, counted, and diluted, and then contributions to this work, including Jorge Monteon, Paul Sauer, Peter Stathis, injected subcutaneously in NOD/SCID mice. Tumor growth was monitored Ian Scott, Min Wang, Daniel Croom, Jim Evans, Xiaomei Song, and Michael for up to 3 mo. Cancer stem cell frequency was determined using L-Calc Mulkerrin. Research funding was provided by OncoMed Pharmaceuticals.

1. Chien AJ, Conrad WH, Moon RT (2009) A Wnt survival guide: From flies to human 23. El-Tanani MK, et al. (2006) The regulation and role of osteopontin in malignant disease. J Invest Dermatol 129:1614–1627. transformation and cancer. Cytokine Growth Factor Rev 17:463–474. 2. van Amerongen R, Nusse R (2009) Towards an integrated view of Wnt signaling in 24. Yan D, et al. (2001) Elevated expression of axin2 and hnkd mRNA provides evidence development. Development 136:3205–3214. that Wnt/beta-catenin signaling is activated in human colon tumors. Proc Natl Acad 3. Wang Y (2009) Wnt/Planar cell polarity signaling: A new paradigm for cancer therapy. Sci USA 98:14973–14978. Mol Cancer Ther 8:2103–2109. 25. Malanchi I, et al. (2008) Cutaneous cancer stem cell maintenance is dependent on 4. James RG, Conrad WH, Moon RT (2008) Beta-catenin-independent Wnt pathways: beta-catenin signalling. Nature 452:650–653. Signals, core proteins, and effectors. Methods Mol Biol 468:131–144. 26. Radtke F, Clevers H (2005) Self-renewal and cancer of the gut: Two sides of a coin. 5. Kohn AD, Moon RT (2005) Wnt and calcium signaling: Beta-catenin-independent Science 307:1904–1909. pathways. Cell Calcium 38:439–446. 27. Reya T, Clevers H (2005) Wnt signalling in stem cells and cancer. Nature 434:843–850. 6. Takebe N, Harris PJ, Warren RQ, Ivy SP (2011) Targeting cancer stem cells by inhibiting 28. Hart MJ, de los Santos R, Albert IN, Rubinfeld B, Polakis P (1998) Downregulation of Wnt, Notch, and Hedgehog pathways. Nat Rev Clin Oncol 8:97–106. beta-catenin by human Axin and its association with the APC tumor suppressor, beta- – 7. MacDonald BT, Tamai K, He X (2009) Wnt/beta-catenin signaling: Components, catenin and GSK3 beta. Curr Biol 8:573 581. mechanisms, and diseases. Dev Cell 17:9–26. 29. Nusse R, Varmus HE (1982) Many tumors induced by the mouse mammary tumor virus – 8. Steinert G, et al. (2011) Sulindac sulfide reverses aberrant self-renewal of progenitor contain a provirus integrated in the same region of the host genome. Cell 31:99 109. cells induced by the AML-associated fusion proteins PML/RARα and PLZF/RARα. PLoS 30. Munemitsu S, Albert I, Souza B, Rubinfeld B, Polakis P (1995) Regulation of in- ONE 6:e22540. tracellular beta-catenin levels by the adenomatous polyposis coli (APC) tumor-sup- – 9. Lu D, et al. (2011) Salinomycin inhibits Wnt signaling and selectively induces apoptosis pressor protein. Proc Natl Acad Sci USA 92:3046 3050. in chronic lymphocytic leukemia cells. Proc Natl Acad Sci USA 108:13253–13257. 31. Rubinfeld B, et al. (1996) Binding of GSK3beta to the APC-beta-catenin complex and – 10. Huang SM, et al. (2009) Tankyrase inhibition stabilizes axin and antagonizes Wnt regulation of complex assembly. Science 272:1023 1026. 32. DiMeo TA, et al. (2009) A novel lung metastasis signature links Wnt signaling with signalling. Nature 461:614–620. cancer cell self-renewal and epithelial-mesenchymal transition in basal-like breast 11. Hu J, et al. (2009) Blockade of Wnt signaling inhibits angiogenesis and tumor growth cancer. Cancer Res 69:5364–5373. in hepatocellular carcinoma. Cancer Res 69:6951–6959. 33. Fodde R, Brabletz T (2007) Wnt/beta-catenin signaling in cancer stemness and ma- 12. DeAlmeida VI, et al. (2007) The soluble wnt receptor Frizzled8CRD-hFc inhibits the lignant behavior. Curr Opin Cell Biol 19:150–158. growth of teratocarcinomas in vivo. Cancer Res 67:5371–5379. 34. Kanzaki H, et al. (2006) Single nucleotide polymorphism of the AXIN2 gene is pref- 13. Gazit A, et al. (1999) Human Frizzled 1 interacts with transforming Wnts to transduce erentially associated with human lung cancer risk in a Japanese population. Int J Mol a TCF dependent transcriptional response. Oncogene 18:5959–5966. Med 18:279–284. 14. Bhat RA, Stauffer B, Della Pietra A, Bodine PV (2010) Wnt3-Frizzled 1 chimera as 35. Klarmann GJ, Decker A, Farrar WL (2008) Epigenetic gene silencing in the Wnt a model to study canonical Wnt signaling. J Cell Biochem 109:876–884. pathway in breast cancer. Epigenetics 3:59–63. 15. Verkaar F, van Rosmalen JW, Smits JF, Blankesteijn WM, Zaman GJ (2009) Stably 36. Mazieres J, et al. (2004) Wnt inhibitory factor-1 is silenced by promoter hyper- overexpressed human Frizzled-2 signals through the beta-catenin pathway and does methylation in human lung cancer. Cancer Res 64:4717–4720. not activate Ca2+-mobilization in Human Embryonic Kidney 293 cells. Cell Signal 21: 37. Polakis P (2007) The many ways of Wnt in cancer. Curr Opin Genet Dev 17:45–51. – 22 33. 38. Kuhnert F, et al. (2004) Essential requirement for Wnt signaling in proliferation of 16. Holmen SL, Robertson SA, Zylstra CR, Williams BO (2005) Wnt-independent activation adult small intestine and colon revealed by adenoviral expression of Dickkopf-1. Proc of beta-catenin mediated by a Dkk1-Fz5 fusion protein. Biochem Biophys Res Com- Natl Acad Sci USA 101:266–271. – mun 328:533 539. 39. Benhamouche S, et al. (2006) Apc tumor suppressor gene is the “zonation-keeper” of 17. Ueno K, et al. (2008) Frizzled-7 as a potential therapeutic target in colorectal cancer. mouse liver. Dev Cell 10:759–770. – Neoplasia 10:697 705. 40. Diehn M, et al. (2009) Association of reactive oxygen species levels and radio- 18. Dann CE, et al. (2001) Insights into Wnt binding and signalling from the structures of resistance in cancer stem cells. Nature 458:780–783. two Frizzled cysteine-rich domains. Nature 412:86–90. 41. Dean M, Fojo T, Bates S (2005) Tumour stem cells and drug resistance. Nat Rev Cancer 19. Hoey T, et al. (2009) DLL4 blockade inhibits tumor growth and reduces tumor-initi- 5:275–284. ating cell frequency. Cell Stem Cell 5:168–177. 42. Rothe C, et al. (2008) The human combinatorial antibody library HuCAL GOLD com- 20. Al-Hajj M, Wicha MS, Benito-Hernandez A, Morrison SJ, Clarke MF (2003) Prospective bines diversification of all six CDRs according to the natural immune system with identification of tumorigenic breast cancer cells. Proc Natl Acad Sci USA 100: a novel display method for efficient selection of high-affinity antibodies. J Mol Biol 3983–3988. 376:1182–1200. 21. Wright MH, et al. (2008) Brca1 breast tumors contain distinct CD44+/CD24- and 43. Dylla SJ, et al. (2008) Colorectal cancer stem cells are enriched in xenogeneic tumors CD133+ cells with cancer stem cell characteristics. Breast Cancer Res 10:R10. following chemotherapy. PLoS ONE 3:e2428. 22. Ginestier C, et al. (2007) ALDH1 is a marker of normal and malignant human mam- 44. Dalerba P, et al. (2007) Phenotypic characterization of human colorectal cancer stem mary stem cells and a predictor of poor clinical outcome. Cell Stem Cell 1:555–567. cells. Proc Natl Acad Sci USA 104:10158–10163.

11722 | www.pnas.org/cgi/doi/10.1073/pnas.1120068109 Gurney et al. Downloaded by guest on September 28, 2021