Overcoming sorafenib evasion in hepatocellular carcinoma using CXCR4-targeted nanoparticles to co-deliver MEK-inhibitors

Sorafenib is a RAF inhibitor approved for several cancers, including hepatocellular carcinoma (HCC). Inhibition of RAF kinases can induce a dose-dependent “paradoxical” upregulation of the downstream mitogen-activated protein kinase (MAPK) pathway in cancer cells. It is unknown whether “paradoxical” ERK activation occurs after sorafenib therapy in HCC, and if so, if it impacts the therapeutic efficacy. Here, we demonstrate that RAF inhibition by sorafenib rapidly leads to RAF dimerization and ERK activation in HCCs, which contributes to treatment evasion. The transactivation of RAF dimers and ERK signaling promotes HCC cell survival, prevents apoptosis via downregulation of BIM and achieves immunosuppression by MAPK/NF-kB-dependent activation of PD-L1 gene expression. To overcome treatment evasion and reduce systemic effects, we developed CXCR4-targeted nanoparticles to co-deliver sorafenib with the MEK inhibitor AZD6244 in HCC. Using this approach, we preferentially and efficiently inactivated RAF/ERK, upregulated BIM and down-regulated PD-L1 expression in HCC, and facilitated intra-tumoral infiltration of cytotoxic CD8+ T cells. These effects resulted in a profound delay in tumor growth. Thus, this nano-delivery strategy to selectively target tumors and prevent the paradoxical ERK activation could increase the feasibility of dual RAF/MEK inhibition to overcome sorafenib treatment escape in HCC.


Sorafenib induces RAF heterodimerization and ERK activation in BRAF-wild type HCC cells in vitro and in vivo.
We found a moderate and variable cytotoxicity after sorafenib treatment in a panel of HCC cell lines, consistent with previous studies and with the transient and moderate responses typically seen in HCC patients 18,19 . The BRAF-wild type HCC cells showed differential sensitivity, with IC 50 values ranging from 2 μM (a clinically relevant concentration 18 ) to over 5 μM (Fig. 1a). Of note, BRAF V600 mutant HCC cells (SK-Hep-1) were more sensitive to sorafenib (IC 50 = 0.5 μM). We first evaluated the effects of sorafenib on RAF/ERK and MAP kinase p38 (p38MAPK) activation -two relevant targets of this drug in HCC cells 20,21 . As expected, sorafenib efficiently inhibited p38MAPK activity in all cell lines ( Fig. 1b and Table S1). However, all BRAF-wild type HCC cell lines tested showed undetectable to low levels of CRAF and ERK phosphorylation at baseline (Fig. 1b). Moreover, sorafenib treatment induced a rapid and sustained increase in ERK activation in BRAF-wild type but not in BRAF V600E mutant HCC cells ( Fig. 1b and Table S1). The robust ERK1/2 phosphorylation was more substantial in sorafenib-resistant HCC cell lines (Fig. 1a,b). Similar to the ERK activation seen in HCC cells in vitro, we found that sorafenib increased ERK activation in orthotopic Hep3B xenografts and in two murine models of spontaneous HCC (in Mst1 −/− Mst2 F/− mice and chemically-induced carcinogenesis) ( Fig. 1c-e) 22,23 .
To test whether this effect is due to inefficient BRAF inhibition by sorafenib, we repeated the experiment using supra-physiological concentrations of sorafenib (4-16 μM) 24 . At these high concentrations, sorafenib inhibited CRAF and ERK activity similarly to PLX4720-a potent and specific BRAF inhibitor-in BRAF-wild type HCC cells (Fig. 1f). We also examined if ERK activation was due to RAF heterodimerization in BRAF-wild type HCC cells-as previously shown with BRAF inhibitors in other cancers 4 . Indeed, after treating HCC cells with sorafenib and evaluating the presence of BRAF in CRAF immunoprecipitate, we found that BRAF heterodimerized with CRAF (Figs 1g and S1).
Paradoxical activation of ERK promotes resistance to sorafenib via degradation of Bim in BRAF-wild type HCC cells. Next, we determined whether the increased ERK phosphorylation after sorafenib treatment mediates HCC cell viability and tumor growth. We found that blocking ERK activation with AZD6244 significantly reduced HCC cell viability both in vitro and in vivo when combined with sorafenib treatment (Fig. 2a,b). Furthermore, we examined whether knocking down CRAF or ERK expression-and thus preventing sorafenib-induced RAF dimer transactivation and consequent ERK activation-could also affect viability of BRAF WT HCC. Downregulation of CRAF alone did not affect HCC cell viability or primary tumor growth (Fig. 2c,d). However, when siRNA-induced CRAF knockdown was combined with sorafenib, the knock down of CRAF reduced ERK activation, decreased HCC cell viability, triggered apoptosis and caused a significant delay in HCC growth (Figs 2c-j and S2).
We further dissected how paradoxical ERK activation modulates HCC cell apoptosis. ERK mediates ubiquitination and degradation of Bim-a pro-apoptosis molecule-leading to resistance to chemotherapeutic drugs 25,26 . We found that sorafenib treatment alone led to Bim phosphorylation and degradation (Fig. 3a,b). However, Bim degradation was prevented when the HCC cells were treated with sorafenib in combination with AZD6244 both in vitro and in vivo, leading to apoptosis in HCC cells (Fig. 3c-e). Collectively, these data show that ERK activation and Bim degradation may mediate the rapidly acquired resistance to sorafenib in BRAF-wild type HCC.

Co-delivery of sorafenib and the MEK inhibitor AZD6244 by the tumor-targeted nanoparticles prevents the paradoxical activation of ERK and PD-L1 expression and facilitates intra-tumoral infiltration of cytotoxic CD8+ T cells in HCC, resulting in enhanced anti-tumor efficacy.
To overcome this evasion mechanism and reduce systemic toxicities, we developed tumor-targeted nanoparticles (TTNPs), with the structure shown in Fig. 4a, to co-deliver sorafenib with a MEK inhibitor into HCC. To this end, we loaded sorafenib and the MEK inhibitor AZD6244 into NPs developed as previously described with several modifications 27,28 . The average diameters of drug-loaded TTNPs determined by DLS were 139.7 ± 9.7 nm, with poly-dispersity indexes (PDIs) of 0.425 ± 0.057. The efficacy of sorafenib and AZD6244 encapsulation in the NPs was approximately 70%. To more selectively target HCC, CTCE-9908, a peptide antagonist for CXCR4, was conjugated to NPs as a targeting ligand (CTCE-NPs) 29,30 . We examined the cellular uptake of CTCE-NPs containing a tracer molecule C6 using both murine HCC (HCA-1) and human HCC (Mahlavu and Hep3B) cell lines. As shown in Fig. 4b-d, the uptake of C6 was greater in HCC cells treated with targeted CTCE-NPs than in cells treated with non-targeted NPs modified with scramble peptides (SC-NPs) (Fig. S3). The uptake of CTCE-NPs was competitively inhibited by addition of free CTCE-9908 peptides in a dose-dependent manner (Fig. S3), indicating that the cellular uptake was ligand dependent. Furthermore, CTCE-NPs loaded with sorafenib and AZD6244 exerted more potent cytotoxic effects on HCC cells than the combination of the free agents, unloaded CTCE-NPs, loaded SC-NPs or CTCE-NPs loaded with each agent alone (Figs 4e and S4). In addition, co-delivery of sorafenib and AZD6244 by CTCE-NPs prevented the paradoxical activation of ERK and increased the expression of Bim (Figs 4f,g and S4). These results demonstrate a synergistic cell-killing effect when using CTCE-NPs loaded with sorafenib and AZD6244.
Next, we evaluated the in vivo effects of CTCE-NPs loaded with sorafenib and AZD6244 on ERK activation as well as on tumor environment in orthotopic murine HCC models. As shown in Fig. 5a, increased tumor uptake of CTCE-NPs or non-targeted SC-NPs was observed compared with that of free drugs due to the EPR (enhanced permeability and retention) effect. Moreover, the enhanced intracellular uptake of CTCE-NPs was observed, indicating that NPs presenting a CXCR4 antagonist CTCE9908 peptide can increase tumor uptake in the murine HCC model (Figs 5b and S5). Furthermore, co-delivery of sorafenib and AZD6244 using CTCE-NPs significantly prevented ERK activation induced by sorafenib (Figs 5c and S6). Of note, this effect was achieved while maintaining a potent anti-angiogenic effect of sorafenib, as demonstrated by the decrease in tumor microvascular density (MVD) (Fig. 5d,e).
Finally, we evaluated the treatment-induced changes in programmed death-1-ligand 1 (PD-L1) expression, an immune checkpoint which could suppress cytotoxic CD8+ T cell proliferation and activation, resulting in immunosuppression and treatment resistance 31 . We examined surgical specimens of HCC from patients with recurring tumors after sorafenib treatment and observed profound ERK activation and increased PD-L1 expression in recurrent HCC (Figs S7 and S8). We further found that sorafenib treatment directly induced PD-L1 expression via ERK-mediated NF-kB activation in murine and human HCC cells in a dose dependent manner (at low doses), which was prevented by MEK or NF-kB inhibition (Figs 6a-c and S9). Moreover, co-delivery of sorafenib and AZD6244 using CTCE-NPs significantly reduced PD-L1 expression and increased cytotoxic T cell accumulation and activation in HCC in vivo ( Fig. 6d-f). Treatment with CTCE-NPs loaded with sorafenib and AZD6244 induced potent tumor regression and increased overall survival in orthotopic HCC model in immunocompetent mice (Fig. 6g,h). Furthermore, the preclinical safety study indicated that the serum levels of hepatotoxicity makers such as aspartate aminotransferase (AST), alanine aminotransferase (ALT), alkaline phosphatase (ALP) and γ-Glutamyltransferase (γ-GT) remained the same as the untreated control animals after treatment of sorafenib and MEK inhibitors in CTCE-NPs (Table S2). We also examined the histology of the various organs of mice treated with sorafenib and AZD6244 in CTCE-NP by H&E staining. We detected no histological changes in the lungs, liver, spleen, kidneys or heart of C3H mice 24 hr after treatment (Fig. S10).
In summary, we demonstrate that rapid ERK activation is a mediator of sorafenib resistance in HCC (Fig. 7). Co-delivery of sorafenib and MEK inhibitors via CXCR4-targeted NPs overcomes the cell autonomous mechanism of resistance to sorafenib in HCC as well as inhibits angiogenesis and converts the immunosuppressive microenvironment to an immunostimulatory microenvironment.

Discussion
Multiple targeted therapies that entered phase III clinical trials in HCC failed to demonstrate superiority or non-inferiority to sorafenib. Thus, sorafenib remains standard of care for patients with advanced stage HCC. The poor understanding of sorafenib's cell-autonomous mechanisms of action in HCC prompted us to examine whether the effect of sorafenib on HCC cell viability is dependent on RAF inhibition. Recently, the use of (c) Combination of sorafenib with CRAF or ERK siRNA leads to a synergistic cytotoxic effect in Hep3B cells. (d) Combination of sorafenib with CRAF or ERK siRNA increases apoptosis in Hep3B cells. Combination of sorafenib with CRAF siRNA encapsulated in liposome-based nanoparticles silenced CRAF expression (e), downregulated ERK activation (f,g) and increased cell apoptosis (h) in orthotopic Hep3B xenografts in nude mice, which resulted in synergistic tumor growth delay, as estimated by blood Gluc measurements (i) and tumor size (j). The data are the mean value ± S.E.M., *p < 0.05, **p < 0.01, ***p < 0.001. RAF inhibitors such as vemurafenib or sorafenib has been shown to paradoxically activate MAPK pathway in BRAF-wild type lung cancer or melanoma, due to induction of RAF (BRAF or CRAF) dimerization 2,4,32 . The clinical responses observed in patients with HCC patients (i.e., a transient delay in tumor growth followed by progression) is consistent with the moderate and highly variable cytotoxicity of sorafenib that we observed in various murine and human HCC cell lines. It suggests that inhibition of non-mutated cell-proliferation pathways (i.e., BRAF-wild type) may result in compensatory activation of downstream or other pathways that might become the rescue pathway for cell survival. Our data show RAF inhibition with knocking down BRAF/CRAF or using RAF inhibitors only moderately affected cell viability in BRAF-wild type HCCs. Moreover, partial inhibition of RAF using low doses of sorafenib or the specific BRAF inhibitor PLX4720 resulted in rapid MAPK activation due to BRAF heterodimerization as a cell-autonomous mechanism of resistance to sorafenib in BRAF-wild type HCCs.
We further studied how ERK activation mediates HCC cell survival by dissecting the signaling downstream of MAPK pathway. Bim is a member of the Bcl-2 homology 3 (BH3)-only subgroup of the Bcl-2 family. Bim induces apoptosis by blocking the activity of the anti-apoptotic members of the Bcl-2 family. It has been previously shown that ERK mediates ubiquitination and degradation of Bim through the Bim phosphorylation, which leads to resistance to chemotherapeutic drugs 25,26 . We show that the rapid activation of the MAPK pathway -induced by sorafenib treatment -leads to Bim degradation and protection from apoptosis in HCC cells. When CRAF knockdown or MEK inhibition-mediated by CRAF siRNAs or MEK blockade-was combined with sorafenib, we found reduced ERK activation, increased Bim and apoptosis induction and thus decreased HCC cell viability.
Next, we evaluated whether the sorafenib-induced paradoxical activation of ERK contributed to the changes of PD-L1 expression in HCC. We have previously shown that sorafenib increases PD-L1 expression in HCC in vivo and others have linked PD-L1 upregulation with MAPK activation in melanomas resistant to BRAF inhibition 8,33 . In this study, we observed the increased PD-L1 expression in HCC from patients with recurring tumors after sorafenib treatment. BRAF-wild type HCC cells showed enhanced ERK activation and PD-L1 expression while treated with sorafenib. The increased PD-L1 expression is transcriptionally modulated by NF-kB. MEK inhibition counteracts sorafenib resistance associated with MAPK activation and PD-L1-mediated immunosuppression. Combination of sorafenib and MEK inhibitors revealed dual therapeutic effects with apoptosis induction and reduction in PD-L1 expression, leading to increased accumulation and activation of tumor-infiltrating cytotoxic T cells. The targeted therapy in combination with immunotherapy not only overcomes resistance, but also provides long-term clinical benefit in HCC patients.
To overcome this evasion mechanism and reduce systemic toxicities, we developed CXCR4-targeted nanoparticles to co-deliver sorafenib with the MEK inhibitor AZD6244 into HCC. We have previously demonstrated that CXCR4 antagonists on the surface of targeted NPs display dual functions, serving both as tumor-targeting ligands and as inhibitors of CXCR4/SDF1α to suppress the accumulation and activation of stromal cells with tumor-promoting properties 28 . In this study, we utilized a peptide inhibitor of CXCR4-CTCE-9908 peptide to modified NPs. The multifunctional NPs loaded with sorafenib and MEK inhibitor achieve potent MAPK inhibition, induced significantly increased cell apoptosis, suppressed PD-L1 expression and inhibited immunosuppression in HCC. These findings were consistent among various human and murine HCC cells in vitro, and were reproduced in vivo using both orthotopically implanted and chemically-induced murine HCC tumors. Molecular profiling of human HCCs has shown that that most tumors harbor no RAF-activating mutations and are indeed BRAF-wild type 34 . Moreover, consistent with our in vitro data obtained from human HCC cell lines, sorafenib treatment activates the RAF/ERK pathway and increases PD-L1 expression in patients with HCC.
Collectively, these data demonstrate that sorafenib can cause rapid RAF dimerization and ERK activation, leading to resistance to sorafenib in BRAF-wild type HCC cells. Adding a MEK inhibitor to sorafenib is clinically feasible 35 , but this strategy may be limited by the specific pharmacokinetics of these agents, and equally important by adverse systemic effects. We addressed this challenge by using a CXCR4-targeted nano-delivery strategy to target HCCs and co-deliver sorafenib with a MEK inhibitor. We showed that treatment with sorafenib and MEK inhibitors co-delivered by tumor-targeted NPs PEG-PLGA revealed more potent tumor growth inhibition effect when compared to the free-form drug cocktail without causing unwanted toxicity, indicating the CTCE-NP formulation could increase the therapeutic effects and therapeutic window of combined sorafenib and MEK inhibitors. This approach may be useful in safely preventing a key cell autonomous mechanism of resistance to sorafenib in HCC as well as inhibiting angiogenesis and reprogramming the immune microenvironment to promote anti-tumor immunity.   of DOPC, 0.04 mg cholesterol and 0.08 mg DSPE-PEG(2000)-maleimide in 40 μl of DMSO were mixed as the oil phase and each time 10 μl of oil phase was added to 280 μl of deionized water dropwise under gentle stirring. The NPs were self-assembled with continuous stirring for 30 minutes at room temperature. The ratio of the volumes of the oil and water phases was fixed at 1/7 (v/v) throughout all of the experiments. For peptide conjugation, peptide was reduced using immobilized TCEP disulfide reducing gel (Thermo Scientific) according to manufacturer's recommendations. Twenty-five μl of 1 mg/ml of CTCE-9908 peptides or scramble peptides were added into the emulsion where it reacted with DSPE-PEG(2000)-maleimide. Four hours later, the unreacted maleimide groups were quenched by adding l-cysteine. The solution was centrifuged at 16,220 rpm for 30 min at 25 °C to collect the NPs, and the NPs were then resuspended in water in the volume equal to the emulsion for characterization. The same procedure was used to synthesize C6-loaded NPs.

Cells and Materials.
Characterization of nanoparticles. The size and morphology of the NPs were characterized by transmission electron microscopy (TEM; H-7500, Hitachi High-Tech, Tokyo, Japan). The NPs were stained on dried formvar-coated 100-mesh copper grids at room temperature. All grids were further dried again for two days before imaging. The particle size and surface charge were examined using a Zetasizer (3000HS, Malvern Instruments Ltd., Worcestershire, UK) at room temperature.
In vivo HCC models. For orthotopic human HCC model, Gluc-Hep3B cells were orthotopically implanted in livers of 7-8-week-old male nude mice in a 10 μl Matrigel solution. HCC tumors in Mst1 -/-Mst2 F/transgenic mice were induced by i.v. injection of Cre-adenoviruses. Male C3H mice were administration with carbon tetrachloride (CCl4) (16% [v/v] in olive oil, 100-ml gavage, 3 times per week) in combination with 5% ethanol in drinking water for 24 weeks to create a chemically induced murine HCC model. For murine orthotopic HCC model, HCA-1 cells were orthotopically implanted in the liver of 7-to 8-week-old male mice in a 100 μl Matrigel solution.
All experiments were performed in accordance with relevant guidelines and regulations. All animals received humane care, in compliance with the "Guide for the Care and Use of Laboratory Animals" published by the National Academy of Sciences, and all study procedures and protocols were approved by the Animal Research Committee of Massachusetts General Hospital and National Tsing-Hua University. Free-form sorafenib was NPs modified with CTCE-9908 co-deliver sorafenib with the MEK inhibitor AZD6244 in HCC, downregulate RAF/ERK and PD-L1 expression, and facilitate intra-tumoral infiltration of cytotoxic CD8+ T cells, resulting in a profound delay in tumor growth. administered daily by gavage at a dose of 50 mg/kg body weight to mice with established tumors. Free-form AZD6244 was administered twice daily by gavage at a dose of 25 mg/kg body weight. For in vivo gene silencing, siRNA in LPD formulations were intravenously injected in orthotopic HCC-bearing mice daily at a dose of 1.2 mg of siRNA/kg for 4 days. Tumors were collected 2 hr after the last injection.
In vitro cellular uptake. C6 was formulated in the NPs as a tracer molecule, with a final weight ratio of C6 to PLGA as 1/150. HCA-1, Mahlavu or Hep3B cells (10,000 cells per well) were seeded in a 12-well plate (Costar, IL, USA) and incubated for 12 hours. The cells were then treated with different formulations containing C6 at 37 °C for 4 hours. The cells were washed with PBS, fixed with 4% paraformaldehyde (PFA) for 10 minutes and counterstained with mounting solution (4′,6-diamidino-2-phenylindole or DAPI, Vector Laboratories, Burlingame, CA). The cellular uptake of C6-loaded NPs was examined and quantified using a confocal microscope (LSM-780, Carl Zeiss, Germany).
To perform a CTCE-9908 peptide competitive assay, cells were prepared as previously described. Before treatment with C6-loaded CTCE-NPs, the cells were treated with free CTCE-9908 peptide at a final concentration of 0 μg/ml, 20 μg/ml, or 50 μg/ml for 10 minutes. The cellular uptake of C6-loaded NPs was examined and quantified using a confocal microscope.
Cell viability assays. Cell viability was assessed using the MTT assay. Cells (10,000 cells per well) were seeded into 96-well plates, allowed to adhere overnight, and exposed to a range of drug concentrations. After 48 or 72 hr, 10 μl of 5 mg/ml MTT dissolved in PBS was added to each well and incubated for 3 hr in 37 °C. The medium was aspirated, and 100 μl of DMSO was added to each well. Absorbance was read at 570 nm.
To analyze the effect of downregulation of CRAF or ERK on in vitro cytotoxic effects of sorafenib, cells were transfected with siRNA using Lipofectamine 2000 reagent (Invitrogen). Twelve hours after transfection of siRNA (50 nM), media was replaced with fresh media. The trasfected cells were treated with sorafenib 24 hr after transfection. After 24 hr of sorafenib (2 μM) exposure, apoptosis was detected by staining the cells with Annexin V and propidium iodide solution followed by flow cytometry analysis. In addition, cell viability was assessed using the MTT assay 48 hours after sorafenib treatment.
Western blot analysis. Cells were lysed in lysis buffer RIPA for 30 min on ice and the supernatant was collected after centrifugation at 12,000 rpm. Cell lysate were separated on a 10% acrylamide gel and transferred to a PVDF membrane. Membranes were blocked for 1 hr in 5% skim milk and then incubated overnight with polyclonal antibodies against p-CRAF, CRAF, p-ERK, ERK, p-p38, p38, p-AKT, AKT, p-IκBα, Bim, p-Bim (Cell Signaling, Danvers, MA) or β-actin (Sigma-Aldrich, St. Louis, MO).
Immunoprecipitation. HCC1 and Hep3B cell immunoprecipitation was performed using extracts prepared in lysis buffer [0.5% Nonidet P-40, 100 mM NaCl, 10 mM EDTA, 20 mM Tris-HCl (pH 8.0), containing 1 g/ml leu-peptin, 1 g/ml aprotinin, 1 g/ml pepstatin, 0.5 mM dithio-threitol, 0.2 mM sodium vanadate, 100 nM microcysteine]. The extracts were sonicated on ice and clarified by centrifugation at 10,000 rpm for 15 min. BRAF and CRAF was co-immunoprecipitated using the anti-BRAF monoclonal antibody prebound to Protein G beads for Hep3B. CRAF was co-immunoprecipitated using the anti-BRAF monoclonal antibody (Cell Signaling, Danvers, MA) prebound to sepharose A beads for HCC1 cells. Immobilized immune complexes were washed three times, eluted in sample buffer, resolved by SDS-PAGE (10% gels), and transferred to nitrocellulose for immunoblot analysis. Membranes were blocked in 5% milk diluted in PBS for 1 hr, after which they were incubated with BRAF or CRAF primary antibodies diluted in 5% milk/PBS overnight at 4 °C. Membranes were washed in PBST (PBS with 0.1% Tween-20) three times and then incubated for 1 hr with a secondary antibody. Membranes were washed four times and then developed by an enhanced chemiluminescence system according to the manufacturer's instructions (PerkinElmer). Immunofluorescence analyses of HCC tissue. Frozen tumor sections (7-8 μm thick) were immunostained with primary antibodies against CRAF, p-ERK, vWF (Dako, Denmark), PD-L1 (Abcam Inc., Cambridge, MA) Granzyme B (Abcam Inc., Cambridge, MA) and CD8a (BD Biosciences, San Jose, CA) and further stained with fluorescent secondary antibodies. The sections were counterstained with DAPI (Vector Laboratories, Burlingame, CA). Samples were imaged by using Zeiss LSM 780 confocal microscope and quantified using 5-10 randomly selected fields per sample.

Assessment of Apoptosis by TUNEL
Preparation of PEGylated LPD nanoparticles. Cationic liposomes composed of DOTAP and cholesterol (1:1 molar ratio) were prepared by thin film hydration followed by membrane extrusion to reduce the particle size. To prepare LPD, 18 μl of protamine (2 mg/mL), 140 μl of deionized water, and 24 μl of a mixture of siRNA and calf thymus DNA (2 mg/mL) were mixed and kept at room temperature for 10 min before adding 120 μl of cationic liposome (10 mM). LPD stand at room temperature for 10 min before the addition of DSPE-PEG. LPD was then mixed with 40 μl of DSPE-PEG (17 mg/mL) and kept at 50-60 °C for 10 min.