Direct Pharmacological Inhibition of β-Catenin by RNA Interference in Tumors of Diverse Origin

This article is featured in Highlights of This Issue, p. 2009

The Wnt/β-catenin signaling pathway promotes differentiation events during embryonic development and leads to tumor formation when aberrantly activated (1). Dysregulation of the Wnt/β-catenin pathway is observed in approximately 90% of colorectal cancers and >50% of liver cancers (2). The two most common oncogenic mutations causing unchecked Wnt/β-catenin signaling are gain-of-function point mutations or deletions in CTNNB1, the gene encoding β-catenin, or loss-of-function mutations in APC, a tumor suppressor gene that encodes the adenomatous polyposis coli protein, which negatively regulates β-catenin function (3). These and other known genetic lesions impair the cell's ability to degrade β-catenin via the ubiquitin-proteasome pathway, resulting in its nuclear accumulation and subsequent complexation with T-cell factor/lymphoid enhancing factor (TCF/LEF) to activate gene transcription (4). The result of such tumorigenic lesions is transcriptional upregulation of specific genes involved in proliferation, evasion of apoptosis, deregulated metabolism, and metastatic potential (2, 5, 6). Preclinical evidence in genetic tumor models suggests that inhibiting β-catenin function would yield therapeutic benefit to sizable subpopulations of cancer patients (7, 8).

Like several other well-characterized oncogenes, β-catenin is very challenging to target by conventional small molecule or monoclonal antibody approaches because it is an intracellular protein with no enzymatic activity (9). Therefore, no approved therapy targeting β-catenin has emerged despite decades of research. RNA interference (RNAi) technology has enabled the inhibition of previously undruggable targets at the mRNA level and has advanced to late clinical development for several indications (10). Most RNAi triggers that have been investigated as potential therapeutics are double stranded or partially double-stranded oligonucleotides of less than 30 base pairs and contain varying degrees of chemical modification to confer stability and reduce potential for immunostimulatory activation (11). The mechanism of action for all RNAi triggers that leads to target degradation is well established. After the therapeutic RNA duplex reaches the cell cytosol, the guide (antisense) strand is bound by the Ago2 endonuclease, which targets its cognate mRNA for cleavage and destruction at a well-defined position exactly ten bases from the 5′ end of the guide stand.

Dicer substrate small interfering RNAs (DsiRNA) are a unique class of RNAi triggers that leverage the microRNA machinery found in all mammalian cells. DsiRNAs are processed by the Dicer enzyme to 21 or 22mer RNA duplexes that are loaded into Ago2 to direct the specific degradation of target mRNAs with high potency and specificity (12). DsiRNAs have higher intrinsic potency than their cognate conventional siRNAs, which bypass Dicer for many (13, 14) but not all (15) target mRNAs. The often favorable profile of DsiRNAs as a structural class has been hypothesized to involve molecular interactions between the Dicer and Ago2 endonucleases.

To achieve RNA interference in a disease-relevant tissue after systemic administration, the oligonucleotide must have sufficient biodistribution, cell internalization, and cytosolic partitioning properties to enable activation of the RNA-induced silencing complex (RISC) and subsequent degradation of the target mRNA. Because RNAi duplexes are inherently unstable, rapidly cleared and cell-impermeable, engineering drug delivery solutions has been a major focus of translational RNAi research for over a decade. Recent efforts have primarily focused on one of two strategies in investigational RNAi therapeutics: ligand-mediated targeting of a liver-specific receptor, or encapsulation in a functionalized nanoparticle that drives biodistribution and cell trafficking (10). So far, systemic delivery of RNAi triggers has advanced to late clinical development only for targets in the liver, underscoring the difficulty of tumor delivery despite well-validated targets and unmet medical need. A tumor delivery strategy must overcome a unique physiology, which features a malformed vasculature responsible for heterogeneity, hypoxia, acidotic regions, and increased interstitial fluid pressure (16). These features of the microenvironment have been shown to act as challenging barriers to drug delivery (17–19). In addition to overcoming complex biology, a clinically viable delivery solution must meet pharmaceutical criteria: stability, scalability, and amenability to good manufacturing practices (GMP). Classes of RNAi delivery formulations that have been investigated as cancer therapeutics either preclinically or clinically include cyclodextrin-based polymers (20), chitosan particles (21), antibody-linked conjugates (22), carbon nanotubes (23), and lipid nanoparticles (LNP; refs. 13, 24–26).

LNPs and liposomes represent the most clinically advanced class of delivery vehicles for oncology therapeutics (27). Liposomal doxorubicin, daunorubicin, and cytarabine have been approved for indications ranging from metastatic ovarian and breast cancer to acute myelogenous leukemia (AML), and numerous other LNPs and liposomes are in late clinical development. For cytotoxic agents and other small molecules, advances in LNP-based delivery offer the potential for improved pharmacokinetic (PK) properties, tumor bioavailability, and therapeutic index. For cell-impermeable oligonucleotides, specialized LNPs enable the escape of siRNA payloads from the endosomal compartment of tumor cells to the cytosol, in part by undergoing pH-dependent structural transitions (28). LNP-based formulations for oncology have not yet advanced beyond early clinical evaluation, where the focus has been primarily on tumors with liver involvement (29). Consequently, there is a need for the systematic development of LNPs with the ability to deliver RNAi triggers efficiently to tumors of diverse origin, with acceptable tolerability profiles.

In this article, we describe the development and characterization of Encore, an LNP system for the delivery of DsiRNA to β-catenin-dependent tumors. We describe structure–activity relationship (SAR) analysis for this multicomponent, modular system. EnCore LNPs represent a unique structural class of nanoparticles and are currently in clinical development for tumor delivery of RNAi triggers (30). EnCore LNPs carrying CTNNB1-targeting DsiRNA delivered pharmacology-relevant intratumoral concentrations of the oligonucleotide, yielding mRNA silencing and antitumor efficacy in multiple mouse models. The models evaluated include subcutaneous cell line–derived xenografts (CLDX), orthotopic patient-derived xenografts (PDX), disseminated hematological tumors, colorectal cancer liver metastases, experimental lung metastases and spontaneous liver tumors induced by specific genetic drivers. Together, these results demonstrate the feasibility of DsiRNA therapeutics using EnCore LNPs for traditionally undruggable targets. These data advance our knowledge of tumor-centric LNPs as a platform for formulation of RNAi triggers and support clinical investigation of direct β-catenin inhibitors for cancer.

All DsiRNAs were synthesized by Integrated DNA Technologies, Inc. (IDT) or ChemGenes Inc. Primer and probe oligonucleotides used in quantitative real-time polymerase chain reaction (qRT-PCR) detection were synthesized by IDT or Life Sciences.

EnCore LNPs were prepared using a two-step mixing process. The first step involved mixing the DsiRNA payload with a cationic liposome to form a solid loaded Core. The cationic liposome was made by extrusion of a mixture of a cationic lipid (e.g., DL-048) and a polyethylene glycol (PEG)-lipid (e.g., DSG-PEG2K). The lipids were mixed in equal volume using either a batch process or an in-line process. The resulting loaded Core was then used in the second mixing step with the envelope lipids to form the final EnCore LNP. A typical mixture of envelope lipids comprised a cationic lipid (e.g., DL-103), DSPC, cholesterol, and DSPE-PEG2K. The ethanolic solution of the envelope lipids was mixed with the solution of the loaded Core using either a batch process or an in-line process. The resulting solution was then subjected to diafiltration by KrosFlo Tangential Flow Filtration System (Spectrum Labs). Phosphate-buffered saline (PBS) was exchanged into the formulation. The formulation was then concentrated, collected, filtered through 0.22-μm membrane and stored at 4°C before vialing. The particle size of the EnCore LNPs was determined by dynamic light scattering (Zetasizer Nano ZS) and the mean diameter was in the range of 70 to 100 nm with a polydispersity of ∼0.1 to 0.2. The concentration of the DsiRNA was measured using UV absorbance chromatography. Lipid quantification test was done by the UPLC-CAD assay. Encapsulation efficiency was measured by RiboGreen assay and ranged from 85% to 95%.

Human colorectal cancer cell lines LS411N, Ls174t, and SW403, murine melanoma cell line B16F10, hepatocellular carcinoma cell line Hep3B, and AML cell line KG-1 were obtained from ATCC. The acute lymphoblastic leukemia (ALL) cell lines 697 and NALM-6 were obtained from DSMZ. Colorectal cancer and the ALL cell lines were grown in RPMI medium supplemented with 10% fetal bovine serum (FBS). B16F10 and Hep3B cells were grown in DMEM supplemented with 10% FBS. The KG-1 cell line was grown in IMDM supplemented with 20% FBS. Mice were obtained from Harlan Laboratories. All the cell lines were purchased in 2014 and authenticated originally by ATCC and DSMZ using short tandem repeat analysis. Cells were expanded and frozen down at low passage within 1 month after the receipt of the original stocks. Cells were then thawed and used within 5 to 6 passages for this study. All the cell lines were subjected to mycoplasma testing (IMPACT 1 profile by IDEXX BioResearch) before being released for use in studies.

Six- to eight-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), Hep3B (5 × 10⁶ cells + matrigel), or B16F10 (2 × 10⁶ cells) under the right shoulder. Tumor volume was measured twice a week to monitor tumor growth/suppression. Dosing was initiated when the tumors reached 150 mm³. For tumor growth inhibition studies, animals were randomized and assigned to one of four cohorts and subjected to dosing cycles as described in Results [n = 8/cohort, plus 3 additional animals for pharmacokinetic/pharmacodynamics (PK/PD) analysis after the first cycle]. B16F10 experimental lung metastases were generated by injecting 7.5 × 10⁵ cells intravenously (i.v.) in nude mice. The disseminated leukemia models were developed by injecting 5 × 10⁶ 697, KG-1, and NALM-6 cells i.v. into NOD.CB17-Prkdc^scid/NCrHsd (NOD.SCID) mice. Colorectal cancer liver metastasis models were generated by surgically implanting 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 i.v. dosing was performed via lateral tail vein at a total volume of 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).

Left liver lobes of 6- to 8-week-old male Balb/c nude mice were surgically implanted with primary human liver cancer model (LI0050) fragments (2–3 mm in diameter) for tumor development. When the orthotopic tumor size reached about 300–500 mm³ [extrapolated by serum alpha-fetoprotein (AFP) levels and satellite mice], mice were randomized and assigned to two cohorts. The protocol and any amendments or procedures involving the care and use of animals in this study were reviewed and approved by the IACUC of CrownBio, and the care and use of animals were conducted in accordance with the regulations of the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC).

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 (RIN) score by Agilent 2100 Bioanalyzer (31). Total RNA (100 ng) was used to make complementary DNA (cDNA) using high-capacity cDNA Reverse Transcription Kit (Applied Biosystems). The cDNA was then diluted 4× 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), and AXIN2 (Hs00610344_m1). In some experiments, SuperScript III quantitative one-step RT-PCR kit (Applied Biosystems, #11732-088) was used as well. Stem loop qRT-PCR was performed as described previously, whereby cDNA was prepared from whole-tissue homogenates using a microRNA amplification protocol (Applied Biosystems #4366597; ref. 32). Primers and probes used for detection of CTNNB1 DsiRNA included: CTNNB1 Stem Loop RT primer (5-GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACGCAGAATA-3), CTNNB1 forward primer (5′-CGCGCGTTCTACATCATTTGTATTG-3′), universal reverse primer (5′-AGTGCAGGGTCCGAG-3′) and CTNNB1 probe (5′-FAM-TTCGCACTGGATACGACGCAG – MGB-3′). Wnt pathway profiling was performed using TaqMan Low Density Array (LDA, Applied Biosystems) technology using custom gene expression panels according to the manufacturer's protocols.

ViewRNA technology (Affymetrix) was used for all in situ hybridization experiments, following the vendor's protocol with the following exceptions. Formalin-fixed/paraffin-embedded tissue sections were cut to 5 μm, baked at 60°C for 60 minutes, and then stored at −20°C until use. Slides were washed using the Little Dipper automated system (SciGene, #1080-65-1). Histo-Clear (Electron Microscopy Sciences, #64110-04) was used for clearing the slide after melting the paraffin. Pretreatment of tumor tissue was for 20 minutes at 95°C. Protease digestion of tumor tissue was 15 to 20 minutes at 40°C. Slides were mounted in Vectashield Mounting Media with DAPI (Vector Labs, #H-1200). Analysis of slides was performed using a Nikon Microscope with a 20× objective with an NA of 0.75. Signal intensity and area of signal were measured using Nikon Elements software. β-Catenin immunohistochemistry was performed as described previously (33), using a β-catenin antibody (Cell Signaling, 8480) at a concentration of 1:500 and Signal Stain DAB substrate kit (Cell Signaling, #8059). Ki67 immunochemistry was performed using an anti-Ki67 antibody (Abcam, #ab16667) at a concentration of 1:100 and DAB under standard conditions.

Total RNA was isolated using the Qiagen RNA miniprep kit (Qiagen) according to the manufacturer's instructions. 5′RACE analysis was performed on 5 μg of total RNA using the GeneRacer kit (Invitrogen) according to the manufacturer's instructions with the exception that RNAs were directly ligated to the kit 5′ Racer Adapter. cDNA generation and RACE amplification were performed using the kit 5′ Racer Primer and CTNNB1-specific primer (5′-TGCAGCCCACCAGCTAACGCACTG-3′). A touchdown qPCR protocol was used to improve amplicon specificity. The cDNA library was constructed using the NEBNext mRNA library prep kit (NEB) according to the manufacturer's instructions. Briefly, PCR products were purified using AMPure beads and adaptors were ligated followed by pair-end library enrichment by amplification. The cDNA library was purified and validated for size, quality, and concentration using an Agilent Bioanalyzer (Agilent Technologies). The cDNA libraries were multiplexed and sequenced on the Illumina MiSeq platform. Data were filtered, and sequences with 15 nucleotides of the 3′ end of the RACE adapter were included in downstream analysis. The RACE adapter sequences were trimmed from the reads, and the remaining portion was mapped to the CTNNB1 cDNA sequence. Reads that matched to CTNNB1 were reported in the final analysis.

LNPs have been applied in clinical testing to achieve favorable pharmacological properties for complex biomolecules, including oligonucleotides. We investigated a novel tumor delivery platform for DsiRNA payloads termed EnCore LNPs. This modular system contains a solid premanufactured Core to which the oligonucleotide is complexed, and an Envelope layer that contributes to the transfection and biodistribution properties of the particle. We recently reported the activity of a CTNNB1-targeting DsiRNA formulated in a first-generation LNP, hereby termed EnCore-A, in Hep3B orthotopic hepatocellular tumors (13). The Core and Envelope of EnCore-A each contain a cationic lipid (sometimes referred to as an ionizable lipid because it is expected to be neutral at physiological pH) and a PEGylated lipid (Supplementary Fig. S1). The EnCore manufacturing process is fully scalable, amenable to current good manufacturing process (cGMP) standards, and subjected to analytical release criteria prior to biological evaluation. These release criteria typically include >90% RNA encapsulation efficiency and a mean diameter ranging from 70 to 100 nm, depending on formulation components. The EnCore LNP system has been validated for manufacturability, efficacy, and safety through formal preclinical IND-enabling studies. DCR-MYC, a first-in-class oncology therapeutic targeting the MYC oncogene, is an EnCore LNP/MYC DsiRNA-based investigational drug currently in clinical testing (30).

To further advance the EnCore LNP technology, we first explored the chemical space around the Core lipid components. Starting from EnCore-A, we generated EnCore-B by replacing DODMA with the novel lipid DL-48 (Fig. 1A) to study potential differences in bioperformance between a charged dimethylglycine and a neutral piperizine lipid head group. Next, we generated EnCore-C by extending the acyl chain length of DMPE-PEG of the EnCore-B core from 14 to 18 carbons using DSPE-PEG. The rationale behind this investigation is that longer acyl chains on PEG-lipids can prevent diffusion of the PEG from the particle and therefore affect its PK properties (34). EnCore-A, B, and C carrying CTNNB1 DsiRNAs (ref. 13; Fig. 1B) were then evaluated in mice bearing subcutaneously implanted human colorectal LS411N tumors by i.v. administration. Like many colorectal tumors, LS411N cells contain an activated Wnt/β-catenin pathway because they are heterozygous for a loss-of-function mutation in the APC gene (35). A CTNNB1-targeting DsiRNA payload, which targets a region of 100% identity between mouse and human, was chosen so that its activity could be monitored in normal mouse tissues as well as in the human tumor.

After i.v. administration of each of the three EnCore formulations to LS411N tumor-bearing mice, we measured the intratumoral concentration of the DsiRNA payload as well as the CTNNB1 mRNA target. While EnCore-A and EnCore-B delivered comparable DsiRNA copy numbers and yielded similar tumor CTNNB1 mRNA knockdown (KD) levels, EnCore-C delivered >30 times more DsiRNA to tumors (Fig. 1C) and yielded improved mRNA KD (Fig. 1D). Duration of effect was dependent on the dosing regimen, but robust mRNA KD was observed for at least 5 days after treatment in both single-dose and multiple-dose regimens (Supplementary Fig. S2). Finally, improved duration of KD was observed when the DsiRNA payload was further modified by the incorporation of stabilizing 2'-fluoro nucleotides (36), suggesting a correlation between RNAi payload stability and tumor PD (Supplementary Fig. S3).

To compare the activity of Encore-A/B/C in tumors relative to normal tissue, we also measured murine Ctnnb1 (mCtnnb1) mRNA levels in the livers of the LS411N tumor-bearing mice. Notably, while EnCore-C showed improved mRNA KD activity in tumors relative to its precursor, it resulted in decreased activity in the normal liver compared with the LNPs containing the shorter C14 PEG-lipid (Fig. 1E). While EnCore-A and -B treatment resulted in a tumor/liver KD ratio of 0.6 to 0.7, indicating higher potency in liver than in tumor, EnCore-C treatment resulted in a tumor/liver KD ratio of 1.3, indicating tumor selectivity. These data contributed to an improved understanding of the SAR for the Core lipid components.

Next, we sought to investigate the SAR of the Envelope cationic lipid, which previous studies suggested is likely to be the driver of transfection efficiency (37). EnCore A-C contains DL-036, which contains a dimethylglycine head group and two symmetrical acyl chains (Fig. 2A). It has been previously suggested that incorporating asymmetric acyl chains into amino lipids can facilitate pH-dependent lipid phase transitions, thereby increasing endosomal release of the payload (38–40). In a test set of otherwise identical LNPs, changing the lengths or the number of double bonds of one or both of the acyl chains significantly affected the CTNNB1 mRNA silencing activity in tumors (Fig. 2A). After three i.v. doses of 3 mg/kg in LS411N tumor-bearing mice, the KD of the target mRNA varied widely within the test set, from 0% to 76% (Fig. 2A). Among the LNPs in the test set, the subset containing a short acyl chain length of 12 carbons yielded the highest potency. One of several examples of SAR discovered in this test set was that the activity of the most highly asymmetric lipids was dependent on their degree of unsaturation (for example, EnCore-D vs. EnCore-E and EnCore-G vs. EnCore-H; Fig. 2A). When the asymmetry was reduced by increasing the length of the short acyl chain, the degree of unsaturation was less impactful. These data suggest that transfection activity is highly dependent on the geometry of the EnCore particle surface. We selected the most potent LNP in the test set, EnCore-N, for further refinement. Because electrostatic effects can influence higher-order lipid structures (41), we further modified EnCore-N by changing its negatively charged PEG-lipid, DSPE, with the neutral PEG-lipid DSG, either in the Envelope (EnCore-Q), the Core (EnCore-R), or in both (EnCore-S). Based on the higher potency of EnCore-N, we reasoned that a single i.v. dose, instead of the 3 daily doses as used in the preliminary studies, would enable differentiation between the final formulation candidates at nonsaturating levels of mRNA knockdown. After a single 5 mg/kg dose in LS411N tumor-bearing mice, there was a trend toward higher potency for LNPs that contain DSPE in the Envelope, but this did not reach statistical significance (Fig. 2B). Based on these data, EnCore-R was selected as the lead formulation.

To further investigate intratumoral mRNA silencing for the most potent LNPs in this test set, we performed fluorescence in situ hybridization on LS411N tumor sections (Fig. 2C) 24 hours after administering a single 5 mg/kg dose, tumor sections were stained for CTNNB1 mRNA (magenta), as well as mouse Pecam mRNA (also known as CD31, cyan), an endothelial marker to outline the host vasculature. The EnCore-C, -N, and -R test articles all showed a marked decrease of CTNNB1 mRNA relative to the untreated control group, with EnCore-R yielding the most robust silencing according to the image intensity quantitation (Fig. 2D). Importantly, mRNA silencing was generally homogeneous throughout the tumor parenchyma and not concentrated around the blood vessels, suggesting efficient extravasation of the particle. A single-dose administration in LS411N tumors as well as a second xenograft model, Hep3B hepatocellular carcinoma (HCC), suggested that the approximate in vivo tumor ED₅₀ for EnCore-R/CTNNB1 DsiRNA was between 0.3 and 1 mg/kg in subcutaneous tumors (Fig. 2E).

While subcutaneously implanted tumors are useful for evaluating novel antitumor agents, the complexity of nanoparticle-mediated delivery necessitates investigation of delivery to tumors in different organ sites. We sought to explore whether the EnCore LNP can provide functional delivery of DsiRNA payloads to tumors in different organ sites, and to those of epithelial or hematopoietic origin. To this end, LS411N and LS174t tumors were surgically implanted into the spleens of nude mice (representative images of tumors from untreated animals are included in Fig. 3A and B), and evaluated for the ability of EnCore-R/CTNNB1 to silence mRNA in both the primary splenic tumor and liver metastases which develop over time (Fig. 3A and B). Next, to determine the performance of EnCore-R in disseminated solid tumors in immunocompetent animals, we used a murine syngeneic model. Applying a commonly used model of experimental metastasis, mouse B16F10 melanoma cells were implanted i.v. into wild-type mice to generate lung tumors, or injected hydrodynamically (42) to generate large liver tumors (Fig. 3C). For both the lung and liver B16F10 tumors, a single dose of EnCore-R/CTNNB1 DsiRNA effectively reduced CTNNB1 mRNA levels. Finally, experimental models of ALL and AML are also useful for evaluating the range of tumors permissive for LNP delivery, as they can also colonize the spleen and/or liver of immunocompromised mice after i.v. injection of tumor cells (Fig. 3D). Again, EnCore-R/CTNNB1 DsiRNA treatment caused significant CTNNB1 mRNA KD in these leukemia-derived lung and liver tumors, showing that activity of this formulation is not limited to the subcutaneous compartment or to tumors of epithelial origin. It should be noted that, for this series of experiments, we elected to administer EnCore-R/CTNNB1 at a comparatively higher dose level (3–10 mg/kg) to obtain a preliminary read in tumors with increased heterogeneity than conventional subcutaneous xenografts. Taken together, these data demonstrate the ability of EnCore-R to deliver pharmacological levels of DsiRNA to tumors of diverse tissue origin.

To determine if CTNNB1 silencing can result in antitumor efficacy in Wnt/β-catenin-activated tumors, LS411N tumor-bearing mice (n = 8/cohort) were given either an irrelevant control DsiRNA (“Placebo”) or CTNNB1 DsiRNA formulated in EnCore-LNPs (3 daily doses per weekly cycle). After two weekly dosing cycles, EnCore-C (3 mg/kg, Fig. 4A) caused tumor growth inhibition of 75% relative to vehicle-treated animals, whereas EnCore-R (1 mg/kg, Fig. 4B) caused tumor growth inhibition of 82%. In a cohort where therapy was discontinued after the second cycle, tumor growth resumed to a level roughly comparable to untreated subjects (Fig. 4B), suggesting that long-term suppression of β-catenin is required for maintenance of efficacy in this model. Comparable tumor growth inhibition was observed in a second Wnt-activated colorectal tumor model, Ls174t, which harbors direct CTNNB1 mutations (ref. 43; Fig. 4C). Efficacy in Ls174t tumors correlated to a sharp reduction of tumor β-catenin protein levels by immunohistochemistry after a single dosing cycle. Importantly, the reduction in tumor β-catenin occurs in all subcellular compartments, and most strongly in the nucleus where the activated oncoprotein resides (Fig. 4D). Finally, the decrease in β-catenin protein is associated with a significant decrease in Ki67 protein, indicating an effect on the proliferative capacity of the tumor cells (Fig. 4E).

To determine if EnCore-R/CTNNB1 also yields in vivo efficacy in other subcutaneous models that harbor defects in Wnt signaling or downstream effectors, we tested similar dosing regimens in SW403 (APC mutant) colorectal tumors (35) and in Hep3B hepatocellular tumors (Supplementary Fig. S4A and S4B). Interestingly, Hep3B tumors are not known to harbor classical mutations in Wnt/β-catenin signaling nodes; however, they are highly sensitive to inhibition of MYC, a well-characterized effector and a direct transcriptional target of β-catenin in tumorigenesis (13, 30, 44). Robust tumor growth inhibition was observed in both models. Efficacy in SW403 tumors also correlated to a near-complete depletion of tumor β-catenin protein levels by immunohistochemistry after two dosing cycles (Supplementary Fig. S4C). Overall, these studies suggest broad applicability for Wnt/β-catenin-dependent tumors regardless of their tissue origin.

As a control for target specificity, we also tested EnCore-R/CTNNB1 in RKO tumors (35). RKO tumors are wild-type for APC and CTNNB1, and therefore do not contain activated Wnt signaling, unlike >90% of human colorectal tumors. Importantly, no antitumor efficacy was observed in RKO tumors (Supplementary Fig. S5), suggesting that, as predicted, only Wnt-dependent tumors will respond to β-catenin inhibition.

The B16F10 syngeneic melanoma model (Fig. 3C) offers the opportunity to investigate LNP-mediated DsiRNA delivery in germline knockout and transgenic animals, as it eliminates the requirement to establish tumors in immunocompromised mice. To explore the mechanism of EnCore-mediated delivery to tumors, we established B16F10 tumors in both wild-type mice and in apolipoprotein E (ApoE)-deficient mice. LNP-mediated delivery to normal hepatocytes has previously been shown to be highly dependent on the opsonization of ApoE onto the particle surface, which facilitates endocytosis of the nanoparticle via the low-density lipoprotein receptor (LDLR; ref. 45). While silencing of the mouse Ctnnb1 gene in the liver was completely abrogated in the ApoE^−/− mice, loss of ApoE did not affect the DsiRNA activity in B16F10 subcutaneous tumors (Supplementary Fig. S6A), nor in B16F10 experimental lung metastases (Supplementary Fig. S6B). To our knowledge, this is the first demonstration of ApoE-independent delivery for an LNP and demonstrates that the cell internalization mechanism of LNPs is context dependent.

From a drug delivery perspective, PDX tumors are widely accepted to better recapitulate the heterogeneous architecture of human cancers than cell line–derived tumors (46). To determine how EnCore-mediated delivery to a PDX tumor compares in efficiency to CLDX tumors, tumor fragments from a patient with HCC patient (model LI0050) were orthotopically implanted into the livers of immunocompromised mice. This particular tumor is a grade 4 HCC derived from a 49-year-old female and was previously determined to be unresponsive to the cytotoxic agents doxorubicin and cisplatin (data not shown).

After a single 5 mg/kg dose of EnCore-R/CTNNB1 DsiRNA, multiplexed gene expression analysis showed a dramatic effect on known Wnt/β-catenin-responsive mRNAs (2, 47) relative to treatment with a Placebo DsiRNA (Fig. 5A, top). Overall, 12 out of 20 transcripts included in the panel were reduced by greater than 30%, including the well-characterized effectors LEF1, MYC, AXIN2, and GLUL (glutamine synthetase). Additionally, we included the murine Ctnnb1 gene on the panel (Fig. 5A, “mCtnnb1”). While the mean KD of human CTNNB1 in the treated cohort was 85%, knockdown of mCtnnb1 was only 46%, suggesting that EnCore-R can deliver the DsiRNA payload to host nonparenchymal cells within the tumor (e.g., stromal cells, macrophages, and T cells), albeit with reduced efficiency. A mouse vascular endothelial marker, Pecam (CD31), was unaffected (Fig. 5A, “mPecam”). As an important control, CTNNB1 DsiRNA did not significantly affect a panel of unrelated ubiquitously expressed mRNAs (Fig. 5A, bottom). In situ hybridization for both CTNNB1 mRNA and downstream effector MYC mRNA in fixed tumor sections suggested homogeneous silencing and tumor penetration (Fig. 5B). Importantly, the RNA interference-based mechanism of action was confirmed by the 5'RACE method (20, 32) followed by next-generation sequencing (NGS), which detected significant levels of the predicted Ago2-mediated cleavage product of CTNNB1 mRNA at single-nucleotide resolution (Supplementary Fig. S7).

To investigate the ability of EnCore-R/CTNNB1 DsiRNA to selectively target the PDX tumor over normal adjacent liver, we measured mRNA levels for CTNNB1 and downstream effectors AXIN2 and MYC in the normal adjacent mouse liver as well as in the orthotopic tumor (Fig. 5C). While mCtnnb1 KD was observed in the normal liver, mAxin2 and mMyc expression were unaffected. These data suggest that the DsiRNA potency was insufficient to have a functional consequence in normal tissue, thereby providing a potential tolerability advantage over LNPs that were developed originally for liver indications (29).

Based on the PDX data (Fig. 5), we decided to further model the potential of EnCore-R/CTNNB1 DsiRNA for HCC. Plasmids encoding constitutively activated CTNNB1 (S45Y mutation) and KRAS (G12D mutation) were hydrodynamically injected into the livers of wild-type mice, yielding aggressive hepatocellular tumors (Fig. 6). In this model, random genomic integration of the transgenes is induced by the Sleeping Beauty transposon system, which has been used to study both hepatoblastoma and HCC (33, 48). Dosing was initiated 5 weeks after hydrodynamic injection (HDI), corresponding to the time that tumor nodes become clearly visible (data not shown). The dosing regimen is shown in Fig. 6A. At the time of study termination, extensive tumor burden was observed in the placebo-treated mice, as indicated by grossly enlarged livers and hundreds of individual tumor nodes (Fig. 6B). In contrast, EnCore-R/CTNNB1 DsiRNA-treated animals displayed significantly reduced tumor burden (Fig. 6B). This phenotype coincided with robust mRNA KD of the mutant human CTNNB1 mRNA transgene (hCTNNB1), the endogenous murine Ctnnb1 mRNA, and downstream effector mAxin2 (Fig. 6C). Curiously, cMyc expression was unaffected in this setting, consistent with previous observations that MYC-driven HCC and Wnt-driven HCC may represent two distinct molecular subclasses (49). In situ hybridization for CTNNB1 mRNA in fixed tumor sections also confirmed the homogeneous silencing and tumor penetration in these tumors (Fig. 6D). Comparably efficient delivery was also observed in a second Sleeping Beauty transposon-derived model driven by activated CTNNB1 (ΔN90/β-catenin) and Yap (YapS127A), which histologically resembles human hepatoblastoma (Supplementary Fig. S8; ref. 33). In addition to demonstrating potency and efficacy, preliminary studies show that EnCore-R LNP demonstrated an acceptable tolerability profile in nonhuman primates upon repeat administration (Supplementary Fig. S9), suggesting a path forward for clinical development.

The β-catenin/Wnt pathway is among the most functionally validated targets for HCC and colorectal cancers. Using an elegant genetic model, Dow and colleagues (50) recently demonstrated that inhibition of activated Wnt signaling (through restoration of APC tumor suppressor gene expression) causes tumor regression and terminal differentiation in well-established colorectal tumors. Additionally, inducible CTNNB1 shRNA experiments demonstrated that a normal crypt progenitor phenotype is restored upon suppression of CTNNB1 mRNA in established colorectal cancer (35). In HCC, even incomplete suppression of CTNNB1 through use of a research-grade antisense oligonucleotide yielded a complete antitumor response in Wnt-driven tumors (51). It is also worth noting that CTNNB1 inhibition may have applications in other tumor types, including immune checkpoint therapy-resistant melanoma, where activated Wnt signaling appears to suppress T-cell infiltration to promote resistance (52). While numerous attempts have been made to target the Wnt pathway and even β-catenin directly (53), most have encountered potency or tolerability concerns, due in part to their lack of tumor selectivity. Transcriptional modulators are historically difficult to target pharmacologically, and their pleotropic effects sometimes make targeting upstream signals or downstream effectors insufficient. Therefore, targeting CTNNB1 at the mRNA level with a tumor-selective RNAi agent is an appealing option for this significant unmet medical need.

LNPs represent the most advanced class of RNAi delivery vehicles in clinical oncology (29), but their tendency to be pharmacologically active in normal liver tissue could be problematic for certain targets. For example, in healthy liver, β-catenin regulates the expression of genes involved in many metabolic pathways, zonation, and homeostasis (33). To our knowledge, the optimized EnCore formulations described in this report represent the first LNP for which functional delivery is more efficient in the tumor than in normal adjacent tissue. While the molecular basis for this tumor selectivity is still being elucidated, we have determined that, unlike SNALP-class LNPs (45), functional delivery to tumors does not require apolipoprotein E optimization. This suggests that the mechanism of cell internalization for EnCore LNPs is different in tumor versus normal liver, and that an exogenous targeting ligand is not necessary to drive homogeneous delivery in the difficult tumor microenvironment (19). Continued progress in understanding the basis of delivery, along with our systematic medicinal chemistry approach, will yield LNPs of higher potency and selectivity. Other nonlipid-based RNAi delivery vehicles, including cyclodextrin polymers (54), have also shown promise in preclinical tumor models, although early clinical development appears to have stalled (20).

From a pharmaceutical perspective, the modular nature of RNAi and LNP technology makes it appealing for personalized medicine applications in oncology. DsiRNAs targeting combinations of historically undruggable oncogenes, including MYC, KRAS, and CTNNB1, in a tumor-selective manner are currently being investigated in additional preclinical models, including genetically defined PDX and advanced metastasis models.

S.P. Monga reports receiving a commercial research grant from and is a consultant/advisory board member for Dicerna. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S. Ganesh, G.R. Chopda, B. Ying, H. Dudek, W. Wang, B.D. Brown, M.T. Abrams

Development of methodology: S. Ganesh, G.R. Chopda, J. Tao, X. Shui, B. Ying, D. Chen, E. Chipumuro, Z. Siddiquee, W. Wang, B.D. Brown, M.T. Abrams

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Ganesh, M.L. Koser, W.A. Cyr, G.R. Chopda, J. Tao, X. Shui, B. Ying, P. Pandya, E. Chipumuro, Z. Siddiquee, K. Craig, S.P. Monga, M.T. Abrams

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Ganesh, W.A. Cyr, G.R. Chopda, X. Shui, P. Pandya, E. Chipumuro, Z. Siddiquee, S.P. Monga, M.T. Abrams

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

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Ganesh, G.R. Chopda, D. Chen, P. Pandya, M.T. Abrams

Study supervision: S. Ganesh, H. Dudek, S.P. Monga, W. Wang, B.D. Brown, M.T. Abrams

Other (Encore LNP preparation, characterization, and quality control): D. Chen

The authors would like to thank Hannah Bogich for additional technical assistance. The authors would also like to thank David Miller and James Weissman for helpful suggestions during the review process.

This work was funded in part by the Center for Strategic Scientific Initiatives, NCI (1R43CA186410-01A1) to B. 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.