β-Catenin mRNA Silencing and MEK Inhibition Display Synergistic Efficacy in Preclinical Tumor Models Free

Colorectal cancer is the third most common cancer and the second leading cause of cancer-related deaths in the United States, despite advances in early detection and targeted therapeutics (1). Targeted therapies that have been incorporated into current standard-of-care regimens for colorectal cancer include anti-EGFR mAbs. Although EGFR is expressed in approximately 85% of colorectal cancers (2–4), anti-EGFR antibodies were found to only confer a survival benefit to patients with wild-type KRAS and BRAF (5–7). This leaves a large population of colorectal cancer patients with KRAS or BRAF mutations, at least 50%, ineligible for the anti-EGFR therapies cetuximab and panitumumab (8–10). For this and other reasons (11–15), refractory colorectal cancer remains a severe unmet medical need and the focus on dozens of current clinical trials.

One approach to treating this poor-prognosis population is to target effectors that are functionally downstream of KRAS and BRAF. The best characterized targets in this category are the MEK1 and MEK2 kinases, which regulate MAPKs via direct phosphorylation. Small-molecule MEK inhibitors are already approved for BRAF-mutant melanoma and are being evaluated for colorectal cancer as part of various combination regimens (16). In one recent phase I/II trial, triple inhibition of EGFR, BRAF, and MEK in BRAF-mutant colorectal cancer with panitumumab, dabrafenib, and trametinib, respectively, showed improved progression-free survival compared with double inhibition of EGFR and BRAF (17). However, even this complex drug regimen yielded only modest gains, underscoring many observations suggesting that such potent drug cocktails are insufficient to address refractory colorectal cancer. In addition, clinical experience in melanoma and other tumors has shown that development of resistance to MEK inhibitors remains a key challenge, even when these agents are used in combination (18).

Strikingly, constitutive activation of Wnt/β-catenin signaling, an evolutionarily conserved network found in nearly every tissue type with broad roles in development and homeostasis, is implicated in approximately 90% of colorectal cancers (19). Although this activation sometimes occurs through direct gain-of-function mutation of CTNNB1, the gene that encodes β-catenin, the more common genetic lesion in this pathway is loss-of-function mutation of upstream negative regulator APC (20). The role of unchecked Wnt/β-catenin signaling in tumor initiation and tumor maintenance is suggested in many preclinical rodent models, including the recent demonstration that restoring APC function in advanced stage colorectal cancer was sufficient to induce differentiation of the colonic epithelium back to the normal physiologic state (21). In addition to the well-characterized effects of canonical Wnt pathway activation, new mechanisms by which β-catenin contributes to tumor progression are emerging, including a role in evasion of the immune system (22–24). It has been proposed that cross-talk between the Wnt/β-catenin and KRAS/BRAF–driven MAPK pathways was shown to cooperate in tumor initiation and tumor progression (25). Importantly, several studies have identified the aberrant activation of Wnt signaling as a primary cause of colorectal cancers and as a resistance mechanism in MAPK pathway–activated colorectal cancers (26, 27). Despite this knowledge, none of the therapeutic agents specifically targeting the WNT pathway has yet been approved to date (28).

Despite the central role of Wnt/β-catenin in colorectal cancer, pharmacologic intervention of this critical pathway has proven to be challenging, and consequently, there is a dearth of drug development programs that have proceeded to advanced clinical trials (29, 30). With few exceptions, the majority of prior efforts to inhibit Wnt signaling do not target β-catenin itself, but rather target specific Wnt ligands or ligand-secretory pathways, Wnt receptors or transcriptional coactivators where pathway redundancies may limit efficacy (31). In addition to difficulty associated with direct targeting of β-catenin by conventional pharmaceutical modalities, there is also a concern that nonselectively blocking its function in normal tissues will interfere with essential homeostasis functions, particularly in the gastrointestinal tract (31).

RNAi technology is a promising alterative to small molecules and biologics for cancer therapy, as it offers the potential to potently silence any expressed gene at the mRNA level. As a drug modality, RNAi triggers are currently progressing through advanced clinical trials, where they have demonstrated sufficient therapeutic index, potency, duration, and safety to enable expansion beyond their early successes targeting genetic diseases of the liver (31). We have recently shown robust preclinical activity in Wnt-dependent tumors of diverse origin with an RNAi agent targeting CTNNB1 (32). This agent, hereby termed DCR-BCAT, is an optimized Dicer substrate siRNA (DsiRNA) formulated in a tumor-selective nanoparticle and demonstrates dose-dependent silencing of β-catenin after systemic administration in preclinical tumor models (32). A DCR-BCAT precursor featuring an earlier generation lipid nanoparticle, DCR-MYC, achieved RNAi target engagement in diverse tumor types at well-tolerated doses in a phase I clinical trial (33).

To address whether inhibition of both pathways simultaneously yields benefit, we sought to evaluate a combination of the novel β-catenin RNAi-based inhibitor, DCR-BCAT and the MEK inhibitor (MEKi) trametinib (34) in multiple preclinical colorectal cancer models. Importantly, in addition to primary tumor models, metastasis models were included as this setting represents the source of most genetic complexity (35), mortality, and medical need for this disease (36–39). This combination demonstrated synergistic efficacy in three colorectal cancer models, all with Wnt and MAPK activation. DCR-BCAT was also efficacious in models that are inherently resistant to MEKi, or that acquire resistance to MEKi after several cycles of therapy. We also investigate the potential of dual therapy in other tumor types with relevant genetics, particularly melanoma and hepatocellular carcinoma (HCC). Together, these findings suggest that direct RNAi-mediated inhibition of β-catenin could be an effective strategy to overcome the current limitations of combination therapy for refractory colorectal cancer and other cancers.

All DsiRNAs were synthesized by Integrated DNA Technologies, Inc. (IDT). Primer and probe oligonucleotides used in qRT-PCR detection were synthesized by IDT or Life Sciences. Trametinib was purchased from Selleckchem with >99% purity.

DCR-BCAT and Placebo lipid nanoparticle (LNP) were prepared as described previously (32). Trametinib was dissolved in DMSO to make 5 mg/mL solution. This solution was further diluted to the appropriate concentration for oral dosing with the solvent mixture containing 10% ethanol, 10% Kolliphor EL, and 80% water. Vehicle used in the studies was 10% DMSO in the solvent mixture.

Human (colorectal cancer) cell lines LS411N, Ls174t, and SW403, HCC cell line Hep3B, and human melanoma cell line A2058 were obtained from ATCC. Colorectal cancer cell lines were grown in RPMI medium supplemented with 10% FBS. Hep3B and A2058 cells were grown in DMEM medium supplemented with 10% FBS. Mice were obtained from Harlan Laboratories. All cell lines used in this study were authenticated originally by ATCC using short tandem repeat analysis. Cells were expanded and frozen at low passage within one month after the receipt of the original stocks and used within 6 passages after thawing. All cell lines were subjected to mycoplasma testing (IMPACT 1 profile by IDEXX BioResearch) before being released for use in animal studies.

Six- to 8-week-old Hsd:Athymic Nude-Foxn1^nu mice (hereby referred to as nude mice) were injected subcutaneously with LS411N (5 × 10⁶ cells), SW403 (5 × 10⁶ cells), Ls174t (5 × 10⁶ cells), and Hep3B (5 × 10⁶ cells + Matrigel) under the right shoulder. Tumor volume was measured twice a week to monitor tumor growth/suppression. Dosing was initiated when the tumors reached 200 mm³. For tumor growth inhibition (TGI) studies, animals were randomized and assigned to one of four cohorts and subjected to dosing cycles as described in Results (n = 6/cohort, plus 3 additional animals for pharmacokinetic/pharmacodynamics analysis after the first cycle). Colorectal cancer liver metastasis models were generated by surgically implanting 1–2 × 10⁶ cells in the spleen of nude mice after midline abdominal incision. After surgery, the abdominal incision was closed with 5-0 to 6-0 absorbable, nonbraided suture, and the skin was closed with a single wound clip. Mice were anesthetized with isoflurane before initiating the surgery and during surgery. Buprenorphine was given preoperatively and postoperatively at 0.1 mg/kg subcutaneously for pain relief. All intravenous dosing was performed via lateral tail vein at a total volume of 10 mL/kg. Trametinib was given orally at 10 mL/kg. Mice were held in a pathogen-free environment, and all procedures involving animals were performed according to protocols approved by Dicerna Pharmaceuticals' Institutional Animal Care and Use Committee (Dicerna-IACUC).

Animal tissues were preserved by either snap-freezing or RNA-later fixative (Life Technologies) and were homogenized using a bead mill (TissueLyzer, Qiagen). After total RNA isolation, representative RNA samples were subjected to QC and determination of the RNA Integrity Number score by Agilent 2100 Bioanalyzer. Total RNA (100 ng) was used to make cDNA using High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). The cDNA was then diluted 4 times for qRT-PCR using TaqMan Fast Advanced Master Mix (Applied Biosystems) and gene-specific primer–probe sets. Primer/probe sets were obtained from Life Technologies and included CTNNB1 (Hs00170025_m1) PP1A (4326316E), MYC (Hs00153408_m1) AXIN2 (Hs00610344_m1). In some experiments, SuperScript III Quantitative One-Step RT-PCR Kit (Applied Biosystems, #11732-088) was used as well. FISH (Supplementary Fig. S1) for CTNNB1 mRNA was performed exactly as described in ref. 32.

β-Catenin IHC was performed as described, using a β-catenin antibody (Cell Signaling Technology, 8480) at a concentration of 1:500 and Signal Stain DAB Substrate Kit (Cell Signaling Technology, #8059). Ki67 IHC was performed using an anti-Ki67 antibody (Abcam, #ab16667) at a concentration of 1:100 and DAB under standard conditions. Image intensity quantitation was performed using Nikon Elements Software.

Tumor tissues (30–50 mg/mouse) were resuspended in lysis buffer containing protease and phosphatase inhibitors, and homogenized using a bead mill (Qiagen TissueLyser. The lysates were then clarified by centrifugation at 14,000 × g for 15 minutes at 4°C. Protein concentrations of the lysates were determined using the bicinchoninic acid protein assay. Equal amounts of total proteins were loaded on SDS-PAGE, transferred onto nitrocellulose membrane, and probed with primary antibodies overnight at 4°C. The primary antibodies against ERK, pERK, and β-catenin were obtained from Cell Signaling Technology. Blots were visualized using the Odyssey Infrared Imaging System (LI-COR Biosciences).

For the synergy analysis, initial assessments of group comparisons were made using ANOVA testing. Prior to ANOVA, normality was determined using the Shapiro-Wilk Test. Combination cohorts showing evidence of a difference between them and their single-agent counterparts were then further tested for synergy. Combination cohorts determined to not have synergy were tested for additivity using Tukey's test for additivity. All models for additivity included a statement for drug interaction. Further testing for additivity and synergy was performed using the Loewe method of determining a combination index (40). Evidence of synergy was determined if the combination index was found to be less than 1.

β-Catenin as a target for colorectal cancer and other tumors is strongly supported by both human genetic and functional preclinical data. DCR-BCAT, based on the formulation previously referred to as EnCore-R/CTNNB1 DsiRNA (32), is an intravenously administered RNAi drug product. DCR-BCAT causes specific, dose-dependent silencing of CTNNB1 mRNA in preclinical murine and human tumors in vivo (32). In xenografted human colorectal cancer tumors, robust mRNA silencing occurs throughout the tumor parenchyma, indicating efficient tumor penetration of the LNP (Supplementary Fig. S1; ref. 32). Importantly, DCR-BCAT is well tolerated in wild-type mice at doses as high as 10 mg/kg/week, with no elevation of liver transaminases, cumulative tissue exposure upon repeat dosing, or accelerated blood clearance that would suggest an anti-drug antibody response (Supplementary Fig. S2). Because the target site shares complete identity between the human and murine genomes, mechanism-based and nonspecific causes of toxicity can both be considered. This tolerability profile enables a sufficient therapeutic window in preclinical models to evaluate efficacy as a monotherapy and in combination with other targeted therapeutics.

To determine the effect of DCR-BCAT monotherapy in aggressive models of colorectal cancer liver metastasis, Ls174t and LS411N cells were surgically implanted into spleen of nude mice as described previously (32). These Wnt-activated human colorectal cancer cell lines colonize the liver and form multiple metastatic lesions (Fig. 1A), resulting in progressive disease and mortality within approximately one to three months (32). Systemic administration of DCR-BCAT results in significant reduction (50%) of CTNNB1 mRNA in these liver metastases one day after a single dose (Fig. 1B). Ls174t tumors progress more rapidly than LS411N tumors, enabling a comparison drug response between models with different growth kinetics. For both models, tumor-bearing mice were randomly cohorted into 3 groups (n = 8) and treated with PBS, DCR-Placebo, or DCR-BCAT at 3 mg/kg/dose intravenously for a short regimen lasting 17 days (Fig. 1C and D). DCR-Placebo contains identical nucleic acid content and chemistry in a scrambled sequence, formulated in the same LNP composition as DCR-BCAT. Animals were monitored daily for lethality or health deterioration to the point of moribund status, at which point they were euthanized according to IACUC protocol. In both the Ls174t model (Fig. 1C) and the LS411N model (Fig. 1D), DCR-BCAT treatment conferred a significant survival advantage over either DCR-Placebo or the PBS vehicle. For example, median survival of LS411N tumor-bearing mice increased by >11 weeks relative to placebo, extended for >5 months after the last dose was administered (Fig. 1D). These data demonstrate that silencing of CTNNB1 mRNA in an acute regimen slows the progression of both primary and metastatic lesions.

Mutations that affect the Wnt/β-catenin and/or MAPK pathways are the most common genetic lesions in colorectal cancer and are identified in the overwhelming majority of patient biopsies collected from primary or metastatic tumors. Ls174t tumors harbor activating mutations in both the CTNNB1 and KRAS genes (41) and therefore are expected to be sensitive to inhibition of the Wnt/β-catenin and MAPK pathways, respectively. We sought to explore whether inhibition of both pathways simultaneously is advantageous in the preclinical metastatic colorectal cancer setting. Trametinib, a MEK inhibitor, was chosen as the MAPK pathway modulator based on its specificity profile and extensive preclinical and clinical data availability (34). First, trametinib monotherapy was tested in this model and exhibited dose-dependent efficacy (Fig. 2A). On the basis of these single-agent trametinib data and the single-agent DCR-BCAT data (Fig. 1C), we selected a dose level of 2 mg/kg/dose of each agent for combination therapy in the Ls174t metastasis model. We also used our experience in the single-agent settings to tune the kinetics of disease progression and treatment. Namely, we used a high cell implantation number (2 × 10⁶) and waited 3 weeks after implantation before initiation of therapy. These experimental conditions serve to accelerate tumor progression and ensure that significant metastatic disease is present in the liver at the time of treatment, as shown in Fig. 1A. Strikingly, the DCR-BCAT and trametinib combination provided a dramatic survival benefit compared with the single agents tested at the same dose level, improving the median survival time by 30 days versus each single agent. The median survival was 44 to 45 days for monotherapy, and 74 days for dual therapy (Fig. 2B). Several animals in the combination cohort survived for longer than 100 days, even though the final dose was administered on day 37. Importantly, the nanoparticle-formulated placebo DsiRNA (DCR-Placebo) offered no benefit over MEKi monotherapy alone, demonstrating that the lipid excipients or off-target activity did not contribute to the observed efficacy. These data offer proof of concept that direct inhibition of β-catenin is a viable strategy to sensitize colorectal cancer to MEK inhibition.

The subcutaneous tumor xenograft setting enables rapid evaluation of therapeutic regimens for TGI. To determine whether the survival benefit in the metastatic setting can translate into TGI when implanted subcutaneously, we performed an efficacy study in mice harboring subcutaneous Ls174t tumors (Fig. 3A). The dose levels for this study were chosen based on the single-dose TGI curves for DCR-BCAT (32) and trametinib (Supplementary Fig. S3). At a trametinib dose level of 0.3 mg/kg in the monotherapy setting, we observed submaximal TGI and also submaximal effects on MYC mRNA, a recently reported indicator of MEK inhibitor activity due to the stabilizing effect of ERK phosphorylation on MYC activity (Supplementary Fig. S3; refs. 42–44). Although treating the subcutaneous Ls174t tumors with either agent alone yielded approximately 60% TGI after one dosing cycle, >90% TGI was observed in the combination-treated cohort (P < 0.001, Loewe combination index <1). As an important control, we also readily observe MEK inhibition by phospho-ERK1/2 Western blot in both the trametinib monotherapy and combination settings, but not with DCR-BCAT alone (Fig. 3A, bottom). These data not only corroborate the findings from the survival study (Fig. 2), but also exemplify the ability of the optimized LNP formulation to deliver the RNAi trigger to extrahepatic tumor sites.

Approximately 50% of human colorectal cancers harbor genetic lesions that cause dysregulation of both Wnt/β-catenin and MAPK signaling. We had previously shown that only Wnt-activated tumors are sensitive to DCR-BCAT monotherapy (32). The mutations that affect Wnt/β-catenin signaling are most commonly found in the CTNNB1 and APC genes, whereas the mutations that affect MAPK signaling are most commonly found in KRAS and BRAF. To investigate whether these specific genetic backgrounds affect the response to dual therapy, we evaluated DCR-BCAT and trametinib in two additional models (Fig. 3B and C). Although Ls174t has mutations in CTNNB1 and KRAS, LS411N carries APC/BRAF mutations and SW403 carries APC/KRAS mutations. To determine whether we can achieve even greater mathematical synergy than in the Ls174t experiment (Fig. 3A), we decreased the dose of DCR-BCAT to levels that were not expected to show efficacy in the monotherapy setting (0.1–0.3 mg/kg/dose; corresponding to approximately one tenth of the doses required for single-agent efficacy). In both LS411N and SW403 tumors, near-complete tumor stasis was achieved under conditions where single-agent efficacy was negligible. Synergy was demonstrated using Loewe drug combination index (Figs. 3B and C; ref. 40). Furthermore, IHC staining of the SW403 tumor sections for Ki67 demonstrates significantly reduced cell proliferation in the combination cohort, therefore identifying inhibition of cell proliferation as a primary mechanism (Fig. 3D and E). These data offer preliminary mechanistic insight into dual therapy and suggest the possibility of achieving a high therapeutic window and tolerability profile with exceeding low doses of the CTNNB1-targeting RNAi trigger.

Acquired resistance to trametinib and other MEK inhibitors is well documented, owing to compensatory secondary genetic lesions in receptor tyrosine kinases, RAS/RAF family members, MEK1/2 itself, or activation of ERK-independent pathways (45). This ultimately limits the clinical potential of this promising therapeutic class (46). We sought to determine whether we could model resistance to trametinib in a preclinical colorectal cancer model, and whether the RNAi combination therapy would be efficacious in this setting. Figure 4A shows that xenografted SW403 tumors are initially sensitive to trametinib, but fail to respond robustly after progressive dosing cycles. Indeed, the third dosing cycle shows little or no TGI. However, adding DCR-BCAT (3 mg/kg/dose) to the regimen yielded rapid tumor regression after the tumors have acquired the MEKi resistance (Fig. 4A). The tumors fully recover within 2 weeks of dosing, but an additional cycle of combination therapy is again effective at inducing tumor regression. These data demonstrate that DCR-BCAT activity is not affected by the acquired resistance mechanisms that affect MEKi. In addition, these data demonstrate the potential for this combination to treat large established tumors.

To further investigate the mechanism by which DCR-BCAT overcomes trametinib resistance, we generated tumor samples at different time points corresponding to the experiment in Fig. 4A. Tumor RNA was prepared from the time of pretreatment (day 15), trametinib sensitivity (day 26), trametinib resistance (days 36–42), and reversal of resistance with DCR-BCAT (day 45). Interestingly, onset of resistance was associated with increases in CTNNB1, AXIN2, and MYC mRNAs, all of which were subsequently decreased upon addition of DCR-BCAT to resensitize the tumors (Fig. 4B). These data suggest that the tumor employs modulation of Wnt/β-catenin signaling as a strategy to overcome MAPK pathway inhibition, consistent with recent data generated using a BRAF inhibitor in melanoma (47) Finally, another mechanism that has been reported to enable acquired resistance to MAPK pathway inhibitors is maintenance of the eIF4F translation initiation complex, in part through transcriptional suppression of proapoptotic Bcl-2 family member BMF (48) To that end, we have also observed a large increase in BMF mRNA during reversal of resistance, indicating that at least two distinct mechanisms appear to be involved in sensitization to the drug combination (Fig. 4B, right).

In addition to acquired resistance, some tumors are inherently resistant to MEK inhibition, even if they harbor BRAF or KRAS mutations. The BRAF/PTEN-mutant human melanoma cell line A2058 is insensitive to BRAF and MEK inhibition due to a stabilized elF4F translation initiation complex (48). Unlike the acquired resistance mechanisms, A2058 innate resistance involves a point of convergence of several oncogenic pathways by dysregulating translation of signal-responsive mRNAs (48). Using a dose level of trametinib sufficient to yield complete stasis in the colorectal cancer models (Supplementary Fig. S3), 3 mg/kg/dose, only 36% TGI was achieved in subcutaneously xenografted A2058 tumors. DCR-BCAT monotherapy (3 mg/kg/dose) also yielded a partial response, whereas the combination led to complete TGI (105%, Fig. 4B). This result is confirmed by Ki67 IHC for cell proliferation, which correlates well with the TGI data (Fig. 4D and E). In the combination cohort, complete TGI is associated with a 70% decrease in Ki67 staining intensity, compared with an approximately 50% decrease in the RNAi monotherapy cohort (n = 3–5/group). Therefore, RNAi therapy is not affected by multiple clinically relevant resistance mechanisms, increasing its probability of success in combination with MEKi.

The MYC oncogene is amplified or overexpressed in the majority of human tumors and is particularly well characterized as a critical driver of HCC (49). Wnt/β-catenin is one of several signaling networks that regulate its expression and activity (50), and therefore, RNAi-mediated silencing of CTNNB1 is highly effective in HCC preclinical models (51, 52). Interestingly, MEK inhibition has been shown to destabilize MYC protein by regulating its ERK-dependent phosphorylation (53). We reasoned that indirectly targeting MYC transcription and MYC stability simultaneously using DCR-BCAT and trametinib, respectively, may yield significant benefit over monotherapy. Indeed, this is the case in subcutaneously xenografted MYC-dependent Hep3B human HCC tumors, where 90% TGI was achieved with the combination (Fig. 5A). To determine the relationship between MYC suppression and efficacy, MYC mRNA was measured after one dosing cycle. Although DCR-BCAT and trametinib (2 mg/kg/dose of either agent) suppressed expression by 37% and 59%, respectively, the combination treatment yielded an 85% decrease in MYC compared with vehicle-treated tumors (Fig. 5B). Finally, we sought to determine whether the potentiation of RNAi-mediated efficacy by trametinib could be reproduced by targeting MYC directly with an RNAi trigger. MYC-targeting DsiRNA (33) was formulated EnCore-R, the same LNP composition used for DCR-BCAT. Strikingly, synergistic efficacy was observed with the MYC/CTNNB1 RNAi combination in Hep3B tumors, even though the dose of each RNAi trigger was reduced by half (from 2 mg/kg/dose to 1 mg/kg/dose) compared with the single-agent cohorts (Fig. 5C). These data suggest strong potential for RNAi trigger combinations in diverse human tumors.

RNAi is an emerging drug modality that has progressed to advanced clinical trials. Specific target engagement in patients has been demonstrated for oligonucleotide therapeutics in cardiovascular diseases (54), viral infection (55), rare genetic diseases (56), and cancer (57). For the latter, safe and efficient tumor-selective drug delivery technology remains a key limitation. LNP-based approaches have been explored to overcome the harsh tumor microenvironment, poor pharmacokinetics, and cell-trafficking properties of nucleic acid–containing drugs. Previous RNAi candidates tested in the clinic for HCC contained derivatives of LNP formulations that were initially clinically developed for normal liver applications, not for oncology. EnCore LNP technology (32, 52) was optimized for tumor selectivity, enabling mRNA knockdown of well-validated oncogenic targets. β-Catenin is an example of a cancer target with an abundance of human genetic evidence, but which has been difficult to drug using conventional small molecules or biologics. The translatability of preclinical efficacy results to human cancers has yet to be fully investigated for RNAi.

Numerous preclinical colorectal cancer models are sensitive to Wnt pathway blockers of varying quality and specificity (32, 58, 59). However, several lines of evidence point to the potential benefits of blocking MAPK and Wnt signaling simultaneously. In transgenic mice, activating KRAS in a mutant APC background increases nuclear localization of β-catenin and accelerates colorectal cancer tumorigenesis (60). Several reports demonstrated synergistic activity in vitro after targeting both pathways using combinations of small-molecule research grade compounds or marketed drugs (41, 61), or combinations of shRNAs and small molecules (27). In vivo proof of concept was achieved using colorectal cancer subcutaneous xenograft models carrying inducible shRNAs targeting CTNNB1 and KRAS (62). To our knowledge, this concept has never been translated into a pharmacologic approach using a specific β-catenin inhibitor prior to the current report and has not previously been evaluated in the genetically distinct and clinically important setting of colorectal cancer metastasis.

In addition to primary and metastatic colorectal cancer, we also explored the combination approach in models of melanoma and HCC. The diversity of preclinical settings for which dual therapy yields benefit suggests the potential for broad applicability in Wnt/β-catenin and MAPK-driven tumors, as well as MYC-dysregulated tumors (Figs. 3, 4, and 5). Although BRAF inhibitors are frequently employed in melanoma therapy, there is some controversy regarding their potential in combination with Wnt pathway modulators. Inducible CTNNB1 shRNA expression increases sensitivity to BRAFi (47) and inhibits metastasis in melanoma (63), but surprisingly, positive Wnt/β-catenin signaling was actually required for BRAFi-induced apoptosis (64) or associated with decreased proliferation (65) in specific contexts. These observations suggest that there may be some exceptions to the benefits of dual therapy. Indeed, one recent report strongly suggests that PTEN status in BRAF-activated melanoma affects whether Wnt signaling promotes or suppresses metastasis (66). Additional rigorous preclinical and clinical experimentation will be necessary to determine which genetic backgrounds are most sensitive to dual therapy in melanoma.

The precise molecular interactions that enable the effects of this dual therapy approach are still emerging. Our data suggest a role for transcriptional upregulation of Wnt effector genes by MEKi, as well as regulation of the eIF4a transcription initiation complex (Fig. 4B). However, we cannot exclude additional mechanisms. Cross-talk between the Wnt/β-catenin and MAPK networks appears to be extremely context dependent. For example, β-catenin has been reported to mediate resistance to BRAFi in a manner dependent on a physical interaction with transcription factor Stat3 (47). In another report, a key negative regulator of β-catenin activity, glycogen synthase kinase 3 (GSK3), has been implicated in stabilizing Ras family members and therefore offering a point of convergence between the two networks (67). Other potential intermediates include dual-specificity phosphatases (DUSPs), which negatively regulate the Ras–Raf–MAPK axis and have β-catenin–responsive promoter elements (68). Finally, ERK can directly phosphorylate and regulate LRP6, a coactivator of the Wnt receptor Frizzled (69). Taken together, multiple bidirectional intermediates facilitate MAPK/Wnt cross-talk and could offer mechanistic insight into the effects of this and other drug combinations.

The concept of targeting β-catenin to overcome MEKi resistance is also consistent with Bardelli's “trunk and branch” model of the cancer evolutionary tree (70). This model states that the tumor remains addicted to the “trunk,” or primary mutations (e.g., APC or CTNNB1 in colorectal cancer), even as secondary genetic lesions (“branches”) arise. Therapies that target the branch mutations (e.g., KRAS, BRAF, P13K) can produce dramatic responses at first, but then the resistance develops because of the selection of preexisting clones with trunk mutations (71). The model supports multiple parallel mechanisms of resistance and of dual-agent synergy. Trunk mutations are often undruggable by conventional modalities, but may be among the most critical points of intervention for many tumor types.

Potent combinations of novel targeted therapeutics are undoubtedly going to be incorporated into standard-of-care treatment for numerous cancer types in the coming years and will lead to improved outcomes. However, a key clinical limitation of these approaches is the presence of overlapping drug-induced toxicities. This is particularly apparent when combining multiple kinase inhibitors, such as BRAF and MEK inhibitors (72), MEK and MTOR inhibitors, and MEK and EGFR inhibitors (73). We believe that combining conventional kinase inhibitors with innovative drug modalities, including RNAi, has unexplored potential to improve outcomes while minimizing class-related adverse effects.

No potential conflicts of interest were disclosed by the authors

Conception and design: S. Ganesh, B.D. Brown, M.T. Abrams

Development of methodology: S. Ganesh, G.R. Chopda, W.A. Cyr, B.D. Brown

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Ganesh, K.P. Craig, G.R. Chopda, W.A. Cyr

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Ganesh, K.P. Craig, M.L. Koser, G.R. Chopda, W.A. Cyr, H. Dudek, M.T. Abrams

Writing, review, and/or revision of the manuscript: S. Ganesh, G.R. Chopda, W.A. Cyr, C. Lai, W. Wang, M.T. Abrams

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Ganesh, X. Shui

Study supervision: S. Ganesh, W. Wang, B.D. Brown, M.T. Abrams

This work was funded in part by the Center for Strategic Scientific Initiatives, NCI (IR43CA186410-01A1 and IR43CA186410-02) to Bob D. Brown and colleagues at Dicerna Pharmaceuticals, Inc.

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