Targeting the WNT Signaling Pathway in Cancer Therapeutics

The WNT signaling cascade is integral in numerous biological processes, including cell cycle regulation and cancer. Alterations in WNT signaling have been identified in numerous tumor types, and in recent years, numerous WNT pathway modulators have been tested in preclinical studies. These agents are now being investigated in the clinical arena, and this review describes the WNT pathway and therapeutics currently in development.

Abstract

The WNT signaling cascade is integral in numerous biological processes including embryonic development, cell cycle regulation, inflammation, and cancer. Hyperactivation of WNT signaling secondary to alterations to varying nodes of the pathway have been identified in multiple tumor types. These alterations converge into increased tumorigenicity, sustained proliferation, and enhanced metastatic potential. This review seeks to evaluate the evidence supporting the WNT pathway in cancer, the therapeutic strategies in modulating this pathway, and potential challenges in drug development.

Implications for Practice:

The WNT signaling cascade is integral in numerous biological processes, including cell cycle regulation and cancer. Alterations in WNT signaling have been identified in numerous tumor types, and in recent years, numerous WNT pathway modulators have been tested in preclinical studies. These agents are now being investigated in the clinical arena, and this review describes the WNT pathway and therapeutics currently in development.

Introduction

The WNT signaling cascade is a central regulatory pathway in controlling key functions of normal and malignant epithelial cells and has become an important new target for cancer drug development. Despite its complexity, it is worth laying forth the pathway’s multiple components and detailing the points of interest in cancer therapeutics.

The pathway is composed of a group of signal transduction elements that regulate gene transcription, cytoskeletal changes, and calcium flux within epithelial cells. This cascade is integral to embryonic development, cell cycle regulation, inflammation, and cancer [1, 2]. The pathway is driven by WNT ligands (WNT), which are a family of 19 glycoproteins that signal through both canonical (β-catenin dependent) and noncanonical (β-catenin independent) mechanisms, as depicted in Figure 1 [3]. We now know that WNT and its downstream effectors, when deregulated, promote cancer initiation, growth, and metastasis. Alterations in this pathway have been implicated in the development of breast cancer, colorectal cancer, melanoma, prostate cancer, lung cancer, and other tumor types.

Figure 1.

Therapeutic targeting of the canonical WNT pathway. The WNT signaling pathway plays an essential role in regulating self-renewal of the cancer stem cell population and tumorigenesis by enhancing cellular proliferation and survival. When WNT ligands are not present (left), β-catenin is phosphorylated, ubiquitinated, and marked for degradation by the destruction complex that is composed of Axin, adenomatous polyposis coli (APC), and glycogen synthase kinase 3β (GSK3β). This results in the association of the repression complex HDAC, Groucho, and TCF/LEF, which prevents WNT-dependent transcription. The binding of WNT ligand to the frizzled receptor (right) activates dishevelled (DVL). This leads to the inactivation of the destruction complex and subsequent translocation of β-catenin into the nucleus and to WNT-dependent transcription. Many different compounds are currently under preclinical (blue) and clinical (red) development that target many different components of the WNT pathway to inhibit WNT-dependent tumor growth. A detailed description of the inhibitors under development is shown in Table 2.

Abbreviations: CBP, CREB binding protein; CK1, casein kinase 1; DKK, dickkopf-related protein; HDAC, histone deacetylase inhibitors; LEF, lymphoid enhancer factor; LRP, low-density lipoprotein receptor-related protein; TCF, T cell transcription factor; Ub, ubiquitin.

WNT signaling is required for regulation of growth, differentiation, and cell death in normal epithelial cells. The complex regulatory pathway begins when WNT ligands bind to multiple transmembrane receptors, including 10 members of the frizzled (FZD) family of G-protein-coupled receptors, receptor tyrosine kinases (RTKs) ROR1 and ROR2, and RTK-like protein kinases triggering downstream activation [4, 5]. In the β-catenin-dependent pathway, absence of WNT stimulation leads to phosphorylation and degradation of β-catenin by a destruction complex containing adenomatous polyposis coli (APC), glycogen synthase kinase 3β (GSK3β), and Axin. Destruction of β-catenin results in expression of β-catenin-repressed target genes c-Myc, cyclin D1, and survivin.

A critical regulator of β-catenin destruction is the protein Axin. Overexpression of Axin induces β-catenin degradation in cell lines expressing truncated APC [6, 7]. The levels of Axin are in turn controlled by tankyrases, members of the PARP-family of poly-ADP-ribosylation enzymes that destabilize Axin, resulting in activation of WNT signaling [8].

Binding of WNT (e.g., WNT 1 and WNT 3) to its known receptors or coreceptors (LRP5 or LRP6) results in activation of dishevelled (DVL) protein. Activated DVL inhibits the destruction complex, resulting in accumulation of cytoplasmic β-catenin that then enters the nucleus. Intranuclear β-catenin acts as a coactivator of T cell transcription factor (TCF) and lymphoid enhancer factor-mediated gene transcription, causing changes to cell proliferation, survival, and differentiation. Beyond the destruction complex, negative regulators of the WNT pathway include secreted dickkopf (Dkk), soluble frizzled-related proteins, and WNT inhibitory factor [9, 10]. Porcupine (PORCN), a membrane bound O-acyltransferase that is required for WNT palmitoylation, secretion, and biological activity, acts as a positive regulator of the pathway.

WNT also modulates its downstream effects via noncanonical WNT pathways: the planar cell polarity signaling pathway and calcium-dependent and small-GTPase-dependent signaling pathways [11]. Both the β-catenin-dependent and -independent pathways are implicated in tumorigenesis and cancer progression, albeit with some pleiotropy. This review will focus primarily on the canonical pathway.

WNT Pathway Dysregulation in Cancer

The importance of aberrant WNT signaling in tumorigenesis was first discovered during tumor induction in a mouse model following proviral insertion at the WNT locus [12]. Activated WNT signaling has also been implicated in proliferation, survival, and ability to metastasize in various cancers cell types. In addition, numerous studies indicate that this pathway is crucial for maintenance of highly tumorigenic cancer-initiating cells.

Activation of WNT signaling can be achieved through either mutational or nonmutational alterations. Thwarting this signaling cascade has led to inhibition of tumor growth in preclinical models, prompting clinical development of WNT inhibitors.

Mutational Alterations in WNT Pathway

Activating mutations involve three major components of the WNT cascade: degradation complex, β-catenin, and TCF. Loss-of-function mutations characterize the negative regulators of the WNT pathway (degradation complex comprising APC, AXIN, and WTX), whereas gain-of-function mutations are observed for β-catenin and TCF transcription factor (positive regulators). WNT pathway-inactivating mutations remain to be seen in cancer.

Colorectal cancer (CRC) epitomizes the mutational aberrations of hyperactivated WNT signaling. Truncating mutations of the APC gene have been linked to both hereditary (familial adenomatous polyposis [FAP]) and sporadic CRC. Downregulation of transcription activation mediated by β-catenin and TCF is crucial to APC’s suppressive effect, and this regulation can be abrogated by mutations involving either APC or β-catenin [13]. Patients with FAP typically develop innumerable polyps in the colon at a young age, and progression to CRC is the rule [14].

Alterations in Axin, a negative regulator of the WNT pathway, have also been described in CRC. Most AXIN1 mutations in CRC occur between exons 1 and 5, which code for protein regions in which the APC, GSK3β, and β-catenin binding domains are located. Mutations in AXIN2 have been associated with colorectal cancer with defective mismatch repair by activating β-catenin/TCF signaling and typically involve one base deletion or insertion in exon 7 [15]. These mutations lead to elimination of the DIX domain, at which DVL binds and negatively regulates Axin activity [16]. Axin has been found to regulate the efficiency of the β-catenin destruction complex in a concentration-dependent manner [17]. Intriguingly, overexpression of Axin1 causes β-catenin degradation, even in cell lines with truncated APC [7].

Nonmutational Alterations

Constitutive activation of WNT/β-catenin signaling in cancer can occur via epigenetic silencing of extracellular WNT antagonists. Both canonical and noncanonical WNT have been shown to be dysregulated in numerous cancers [18–21], and comprehensive reviews have addressed WNT dysregulation in cancer in detail [1, 22, 23].

Table 1 gives a detailed description of WNT pathway alterations involving various tumor types.

Table 1.

WNT pathway alterations in various tumor types

WNT Pathway Modulation

Pharmacological modulation can be divided into compounds that modulate the ligand/receptor interface, stabilize the degradation complex, or interfere with β-catenin-dependent gene transcription. Figure 1 depicts the WNT signaling pathway, annotated with therapeutic compounds at various nodes.

Modulation of the Ligand/Receptor Interface

OMP-18R5

OMP-18R5 (vantictumab; OncoMed Pharmaceuticals, Redwood City, CA, http://www.oncomed.com) is a monoclonal antibody that interacts with 5 frizzled receptors, thus blocking the induction of WNT cascade by its ligand. This antibody exhibited synergism with standard of care chemotherapy causing tumor growth inhibition of xenografts derived from breast, colon, pancreas, and lung cell lines. Gene expression and quantitative polymerase chain reaction analysis demonstrated reduction of known β-catenin target genes consistent with WNT pathway blockade [24]. Experiments conducted on Ptf1a-Cre, LSL-KrasG12D, β –catenin^f/f (KBC) mice, in which both alleles of β-catenin are floxed in the context of mutant KRAS, revealed that treatment with OMP-18R5 inhibited PanIN formation.

In a phase I trial of refractory solid tumor patients, 29 patients were treated and maximum tolerated dose had not been reached. OMP-18R5 was well tolerated up to the current dose of 15 mg/kg every 3 weeks. Side effects included grade 1/2 fatigue, vomiting, abdominal pain, constipation, diarrhea, and nausea. One instance of dose-limiting grade 3 nausea and vomiting was reported at the second dose level. Marked increases in β C-terminal telopeptide (β-CTX), serum marker of bone turnover observed in a proportion of patients, coupled with grade 2 compression fractures after a fall in one patient, were notable observations. Prophylactic calcium carbonate/vitamin D and administration of zolendronic acid as indicated were instituted. OMP-18R5 demonstrated decreased expression of WNT pathway target genes in pharmacodynamics studies involving hair follicles. Three patients with neuroendocrine tumors, all of whom had radiologic progressive disease prior to enrollment, experienced prolonged stable disease (>300 days) while on treatment [25] (ClinicalTrials.gov identifier NCT01345201).

Combination studies incorporating OMP-18R5 with taxanes are currently under way in two separate phase Ib studies in patients with non-small cell lung carcinoma (ClinicalTrials.gov identifier NCT01973309) and non-Her2 advanced breast carcinoma (ClinicalTrials.gov identifier NCT01957007). A study evaluating combination nab-paclitaxel and gemcitabine with vantictumab in patients with untreated advanced pancreatic cancer is also actively enrolling (ClinicalTrials.gov identifier NCT02005315).

OMP-54F28

OMP-54F28 (OncoMed Pharmaceuticals) is a recombinant fusion protein combining human FZD8 receptor with the ligand binding domain and human IgG1 Fc fragment. OMP-54F28 acts as a decoy receptor by sequestering WNT ligands. Testing in xenograft models showed that OMP-54F28 impeded the growth of numerous solid tumor types and selectively reduced cancer stem cells when administered alone, or in combination with chemotherapeutic agents [26]. Notably, in pancreas tumors, OMP-54F28 promotes a marked differentiation of tumor cells that is coupled with profound reduction in tumorigenic potential.

A phase I study is ongoing and enrolling patients with advanced solid tumors (ClinicalTrials.gov identifier NCT01608867). As of June 2014, 25 patients had been treated in 7 dose-escalation cohorts (0.5, 1, 2.5, 5, 10, 15, and 20 mg/kg every 3 weeks). No dose-limiting toxicities (DLTs) were observed at the highest dose level. Grade 1/2 toxicities included anorexia, fatigue, hypocalcemia, nausea, hypertension, peripheral edema, and vomiting. A single grade 3 anemia was noted. Of the 25 patients treated thus far, 6 patients experienced a doubling of β-CTX, a phenomenon also seen with OMP-18R5 [25]. Five of 5 patients reversed the increase in β-CTX with zoledronic acid.

Three phase Ib clinical trials of OMP-54F28 are ongoing: one in pancreatic ductal adenocarcinoma (gemcitabine/nab-paclitaxel with OMP-54F28, ClinicalTrials.gov identifier NCT02050178), one in hepatocellular carcinoma (sorafenib with OMP-54F28, ClinicalTrials.gov identifier NCT02069145), and one in ovarian cancer (carboplatin/paclitaxel with OMP-54F28, ClinicalTrials.gov identifier NCT02092363).

LGK974

PORCN, a membrane bound O-acyltransferase, is a key enzyme in WNT biosynthesis. PORCN adds palmitate (acylates) to human WNT3A at serine 209. Acylation of S209 is essential for the both WNT secretion and extracellular interaction with frizzled receptors. LGK974 (Novartis, Basel, Switzerland, https://www.novartis.com), a potent and specific small molecule PORCN inhibitor, has been shown to reduce WNT-dependent LRP6 phosphorylation and expression of WNT target genes in preclinical models. LGK974 demonstrated antitumor response in in vivo breast, pancreas, and head and neck cancers [27]. Intriguingly, all LGK974-sensitive pancreatic cell lines carried a loss-of-function mutation in the ubiquitin E3 ligase ring finger 43 (RNF43) gene. Functionally, RNF43 is a negative regulator of WNT and inhibits the WNT signaling by reducing levels of membrane frizzled receptors [28], and whole-exome sequencing of colorectal and endometrial cancers identified somatic mutations of RNF43 in more than 18% of colorectal and endometrial carcinomas [29]. Phase I evaluation of LGK974 is under way (ClinicalTrials.gov identifier NCT01351103), enrolling patients with melanoma, breast cancer (lobular or triple negative), and pancreatic cancer.

Notably, in pancreas tumors, OMP-54F28 promotes a marked differentiation of tumor cells that is coupled with profound reduction in tumorigenic potential.

Antibodies Directed Against Dickkopf-1 (DKK1)

DKN-01

DKN-01 (HealthCare Pharmaceuticals, Cambridge, MA, http://healthcarepharmaceuticals.com), a monoclonal antibody against dickkopf WNT signaling pathway inhibitor 1 (DKK1), was evaluated in patients with multiple myeloma and advanced solid tumors in phase I testing (ClinicalTrials.gov identifier NCT01457417). Overexpression of protein DKK1 has been associated with multiple myeloma (MM). Production of DKK1 by myeloma cells leads to development of osteolytic lesions through direct suppression of osteoblast differentiation. Preclinical studies suggested an indirect antimyeloma effect secondary to inhibition of osteoclastogenesis by DKN-01 [30]. In addition, a phase I/II study combining DKN-01 with lenalidomide/dexamethasone is being evaluated in patients with relapsed or refractory multiple myeloma (ClinicalTrials.gov identifier NCT01711671).

BHQ880

BHQ880 (Novartis) is a phage-derived monoclonal antibody against DKK1. Although BHQ880 had no direct effect on MM cell growth, it significantly inhibited growth of MM cells in the presence of bone marrow stromal cells in vitro [31]. Preliminary results from a phase II study (ClinicalTrials.gov identifier NCT01302886) have been reported [32]. Patients (n = 25) with treatment-naïve intermediate- and high-risk smoldering MM were enrolled. Patient received 10-mg/kg infusions of BHQ880 every 28 days. No grade 3 or 4 adverse events were reported. Significant adverse events in the study included arthralgia, fatigue, extremity pain, and pyrexia. No antitumor activity was noted in these preliminary results.

Stabilization of Degradation Complex

Tankyrase Inhibitors: G244-LM, G007-LK, XAV939, IWR-1, and JW55

Tankyrases act on poly-ADP-ribosylate AXIN proteins, the concentration-limiting component of the β-catenin destruction complex. Inhibitors of tankyrase stabilize AXIN, leading to enhancement of β-catenin destruction. Multiple tankyrase inhibitors have been developed including G244-LM, G007-LK, XAV939, IWR-1, and JW55. G244-LM, G007-LK, and XAV939 showed a decrease in WNT/β-catenin signaling in APC-mutant CRC cell lines [33, 34]. XAV939 and IWR-1 showed growth suppression in both human and murine lung cancer cell lines [35]. These agents are currently in preclinical testing, and phase I trials are being planned.

Interference With β-Catenin-Dependent Gene Transcription

PRI-724

To generate a transcriptionally active complex, β-catenin recruits transcriptional coactivators, cAMP response element binding (CREB) binding protein (CBP), or p300 after translocation into cell nucleus. PRI-724 (PRISM BioLab, Yokohama Kanagawa, Japan, http://www.prismbiolab.com), a small molecule inhibitor of the binding between β-catenin and CBP, leads to downregulation of genes responsible for symmetric nondifferentiated division. This also leads to a shift to greater β-catenin/p300 cooperativity that results in initiation of differentiation [36].

In preliminary results from phase I testing (ClinicalTrials.gov identifier NCT01302405) [36], 18 patients were given PRI-724 over 7 days every 2 weeks, ranging from 40 to 1,280 mg/m² per day. No maximum tolerated dose was identified in dose escalation. A single DLT in the form of grade 3 hyperbilirubinemia was encountered in a patient given 1,280 mg/m² per day. Grade 1/2 side effects include diarrhea, hypophosphatemia, reversible elevated bilirubin, nausea, fatigue, anorexia, and thrombocytopenia. No objective responses were seen in this study. Reduction in survivin expression, a β-catenin target gene in circulating tumor cells after treatment with PRI-724, is consistent with a pharmacodynamic effect.

Three clinical trials evaluating combinations of PRI-724 are ongoing: (a) phase I/II trial of PRI-724 with dasatinib in advanced myeloid malignancies (ClinicalTrials.gov identifier NCT01606579), (b) phase Ib trial of PRI-724 with gemcitabine for advanced pancreatic adenocarcinoma (ClinicalTrials.gov identifier NCT01764477) and, (c) phase Ib trial of PRI-724 with the mFOLFOX regimen in patients with advanced colon cancer (ClinicalTrials.gov identifier NCT01302405).

CWP232291

CWP232291 (JW Pharmaceutical, Seoul, Republic of Korea, http://www.jw-pharma.co.kr/pharma/en/intro/pharma.jsp) is a small molecule that binds Src associated with mitosis 68K protein (Sam68) and promotes apoptosis through inhibition of the antiapoptotic WNT driven gene survivin. CWP232291 has shown both in vitro and in vivo antitumor activity in multiple myeloma [37]. Phase I enrollment is under way for patients with acute myeloid leukemia, chronic myelomonocytic leukemia, and myelodysplastic syndrome (ClinicalTrials.gov identifier NCT0139846).

Table 2 summarizes the compounds evaluated in early phase trials and preclinical settings.

Table 2.

Agents in development targeting the WNT pathway

Challenges Ahead

Although it is clear that dysregulated WNT signaling plays an integral role in cancer pathogenesis, the ubiquitous nature of WNT signaling and its numerous effects significantly complicates WNT signaling blockade. Although studies agree that that β-catenin mediates cell proliferation and growth in melanoma cell lines [38], its role in metastasis remains unresolved [39–41]. Some studies, for instance, have demonstrated that suppression of β-catenin is associated with disease progression [40, 42]; in contrast, other studies have reported that stabilization of the β-catenin may facilitate a metastatic phenotype in melanoma [41, 43]. In particular, Grossmann et al. showed that WNT5A-dependent β-catenin signaling facilitated metastasis in melanoma [43]. One reason for this discrepancy is that receptor expression may have different effects on signaling [44]. In cases in which ROR2 is the predominant coreceptor, WNT5A signals through FZD2 or FZD5 to degrade β-catenin. When the LRP6 coreceptor is dominant, WNT5A pairs with FZD4/LRP6 and stabilizes β-catenin. Given that opposing results have been described to yield a more metastatic phenotype, further studies are needed to understand the role of β-catenin at enhancing disease progressing in melanoma.

The noncanonical WNT pathway has been implicated in metastasis in both pancreatic and castrate-resistant prostate cancers [45–47]; however, noncanonical WNT signaling pathways in cancer are still limited, and many studies have yet to be independently confirmed. It is unclear if aberrations in the WNT pathway are causative of poor clinical outcomes or simply correlative. Future studies will need to better delineate the context-dependent roles of WNT signaling in specific malignancies.

Activating mutations, epigenetic events, and autocrine activation frequently coexist within the same tumor type, albeit with varying dependence. It is conceivable that pathway activation from different means would result in varying outcomes. The differing effects on mammary glands, for example, were evaluated using transgenic mice with stable expression of WNT1, MMTV-WNT1 and ΔN89 β-catenin, respectively. Interestingly, ΔN89 β-catenin mice induced a lobuloalveolar differentiation with limited ductal branching along the luminal cells, whereas MMTV-WNT1 mice caused mammary tumors with extensive ductal branching primarily affecting basally located cells [48].

Taken together, cellular heterogeneity, dynamic interactions with different nodes of the pathway, and cellular plasticity explain the divergent response and phenotype seen. It is imperative to have a greater understanding of individual tumor biology and to identify predictive biomarkers of efficacy for future clinical trials.

Necessary but Not Sufficient

Although encouraging single-agent antitumor activity is seen with WNT pathway inhibition in preclinical models, the same has not been recapitulated thus far in early phase clinical trials. Efficacy data from two phase I studies evaluating vantictumab and PRI-724 have been reported [25, 36]. No significant tumor regression was observed in a combined patient population of 41. A plausible explanation is the extensive crosstalk between WNT and other signaling pathways.

The cooperativity between WNT and Notch has been implicated in influencing proliferative effects on early intestinal precursors. Importantly, synergistic effects between both signaling pathways resulted in development of colonic adenoma [49]. Mechanistically, membrane-bound Notch binds to active β-catenin in colon cancer cells and negatively regulates post-translational accumulation of active β-catenin. Treatment with a γ-secretase inhibitor prevents cleavage of membrane-bound Notch, leading to abrogation of active β-catenin activity [50].

In the adult mouse bronchiolar epithelium, WNT/β-catenin signaling alone does not lead to lung oncogenesis; however, concurrent activation of WNT/β-catenin signaling and expression of constitutively active KRAS (p.G12D) led to a significant increase in overall tumor number and size and a more aggressive tumor phenotype compared with KRAS p.G12D alone [51]. In addition, a hyperactivated WNT pathway has been noted to confer resistance to PI3K/Akt inhibition in preclinical models of colon cancer. In contrast to xenografts derived from colon cancer cells with low concentration of β-catenin, xenografts with high concentration of β-catenin were resistant to PI3K/Akt inhibition. This resistance was reversed by XAV-939, a tankyrase inhibitor, when added to sphere cultures of patient-derived colon cancer cells that were insensitive to PI3K/Akt inhibition [52].

It is unclear if aberrations in the WNT pathway are causative of poor clinical outcomes or simply correlative.

Management of Toxicities

Hematopoietic, bone, and gastrointestinal toxicities are anticipated with abrogation of WNT signaling due to the importance of this pathway for self-renewal of normal stem/progenitor cells in these systems.

A single dose-limiting toxicity event of nausea and vomiting was reported in the phase I trial of OMP-18R5. Other grade 1/2 adverse events of significance included constitutional and gastrointestinal side effects. Notably, marked increase in β-CTX, serum marker of bone turnover, was observed in a proportion of patients, prompting institution of prophylactic measures. A similar adverse event profile was reported in the phase I trial of OMP-54F28.

Although no maximum tolerated dose was identified in the dose-escalation study of PRI-724, a single DLT in the form of grade 3 hyperbilirubinemia was encountered. Other grade 1/2 adverse effects included constitutional, gastrointestinal, reversible cholestasis and thrombocytopenia [36]. Although the WNT-targeted compounds explored thus far have relatively acceptable side effect profiles as single agents, development of WNT inhibitors are likely done in concert with other agents. Adverse events will undoubtedly be more pronounced with combination therapy.

Conclusion

WNT signaling is frequently dysregulated in cancer and is a valid target in antitumor therapy. Multiple agents are currently being explored in preclinical and early phase clinical settings. Some of the agents have thus far demonstrated abrogation of the pathway with acceptable toxicities but minimal preliminary antitumor activity as single agents; combination studies are ongoing. Pharmacodynamic analysis of serial tumor biopsies and surrogate normal tissue is ongoing in several studies. Multiple challenges remain including better understanding of this pathway and its relevance to individual tumor types, identification of predictive biomarkers to optimize the risk-benefit ratio, and derivation of rational partners for combination.

Author Contributions

Conception/Design: David Tai, Keith Wells, John Arcaroli, Chad Vanderbilt, Dara L. Aisner, Wells A. Messersmith, Christopher H. Lieu

Provision of study material or patients: David Tai, Keith Wells, John Arcaroli, Chad Vanderbilt, Dara L. Aisner, Wells A. Messersmith, Christopher H. Lieu

Collection and/or assembly of data: David Tai, Keith Wells, John Arcaroli, Chad Vanderbilt, Dara L. Aisner, Wells A. Messersmith, Christopher H. Lieu

Data analysis and interpretation: David Tai, Keith Wells, John Arcaroli, Chad Vanderbilt, Dara L. Aisner, Wells A. Messersmith, Christopher H. Lieu

Manuscript writing: David Tai, Keith Wells, John Arcaroli, Chad Vanderbilt, Dara L. Aisner, Wells A. Messersmith, Christopher H. Lieu

Final approval of manuscript: David Tai, Keith Wells, John Arcaroli, Chad Vanderbilt, Dara L. Aisner, Wells A. Messersmith, Christopher H. Lieu

Disclosures

Dara L. Aisner: Oxford Oncology, Casdin Capital (C/A), Clovis Oncology (H); Wells A. Messersmith: OncoMed (RF). The other authors indicated no financial relationships.

(C/A) Consulting/advisory relationship; (RF) Research funding; (E) Employment; (ET) Expert testimony; (H) Honoraria received; (OI) Ownership interests; (IP) Intellectual property rights/inventor/patent holder; (SAB) Scientific advisory board