Animal studies were performed in accordance with the protocols approved by the Institutional Animal Care and Use Committee at Duke University. The Tyr-CreER.Braf^Ca.Pten^fl/fl mice were provided by Martin McMahon of UCSF to Duke Cancer Institute, and induced as previously described (40, 42). Briefly, the back skins of 5-7 weeks old male and female animals were shaved and the center of the shaved region was treated with 3 topical applications of 1.5 µl of 5 mM 4-hydroxy-tamoxifen (4-OHT, dissolved in 99% ethanol, Sigma, St Louis, MO, USA) spaced at 1-day intervals (43, 44). Pigmented lesions became visible 9-13 days after the first 4-OHT application. At this point, the animals (n=4-8/group) were treated every other day for 6 weeks via intraperitoneal injections of solvent control (5% DMSO, 5% methylcellulose and 0.5% Tween-80 in water), trametinib (TRA, 3 mg/Kg (45) alone, and TRA together with chloroquine (CQ, 40 mg/kg) (35). Mice were weighted and monitored every week for tumor development and health conditions and euthanized at the end-point for necropsy and tissue collection of the skin lesions.

Tissue samples harvested at the end-point were embedded in optimal cutting temperature compound or fixed in 10% formalin and then embedded in paraffin blocks. The paraffin sections (6 µm thick) were de-waxed in 100% xylene for 5 minutes followed by sequential treatments of 100%, 90% and 70% ethanol for 5 minutes each and antigen unmasking by boiling in 10 mM citrate buffer (pH6.1) for 10 minutes. Endogenous peroxidase activity was blocked using 3% hydrogen peroxide in water. Non-specific binding was blocked using 10% serum. Sections were then incubated with primary antibodies against the transcription factor MITF (Abcam ab20663, Canada), pERK (Novus NBP1-78017, Littleton, CO USA), Ki-67 (RM-9106-S, ThermoFisher Scientific, Waltham, MA). Primary antibodies were used at 1:200 dilution and the secondary antibody reactions were obtained using the Vectastain ABC Elite Kit (Vector Laboratories, Burlingame, CA, USA). Slides were rinsed, stained with Mayer’s hematoxylin (Sigma), dehydrated, and mounted with Permount mounting medium (Fisher Scientific, New Haven, CT, USA). Positive controls were tested in tissues known to be positive to the antibodies. Negative controls for cross contamination were used without the primary antibodies. For detection of DNA fragmentation as a marker of apoptosis, tissues slides were stained with the ApopTag peroxidase assay kit for the TdT-mediated dUTP nick-end labeling (TUNEL) (Millipore, S7100, Billerica, MA, USA). For immunofluorescence staining, 6-7 µm cryosections were fixed in 100% methanol as described (46), and incubated with primary antibodies against F4/80 (#23115, Biolegend, San Diego, CA), IFNγ (#585, R&D systems, Minneapolis, MN) and LC3A/B (#4108, Cell Signaling Technology, Danvers, MA) used at 1:100 dilutions. Samples were counterstained with DAPI (ThermoFisher Scientific). Images were taken with the Olympus BX41 microscopic imaging system or the Olympus IX73155 imaging system (Center Valley, PA). Quantitative assessment of tumor areas, pERK1/2, Ki-67, TUNEL, MITF and IFN γ was performed using ImageJ software (version 1.53g, NIH). For this, about 15 photomicrographs per group were randomly selected. Ki-67+ cell count was carried out with the particle count module in ImageJ. Cellular staining of the other markers was assessed automatically by thresholding the area stained by the antibodies. Some sporadic melanin pigmentations were removed using the erase tool in Adobe Photoshop prior image analysis. F4/80+ cells were manually counted from 5-7 images per group. For CD4 and CD8, tissue sections were processed using the automated IHC assay Discovery Ultra (Ventana Medical Systems, Tucson, AZ). Slides were subject to antigen retrieval (CC1 buffer, 100°C, 56 minutes) and pretreatment with cell conditioner followed by incubation with primary antibodies against CD4 (4SM95, #14-9766-80) and CD8 (4SM15, #14-0808-82) (Thermofisher Scientific) for 60 minutes at 36°C. The detection reagents were the Omap-anti-Rat HRP RUO (760-4457) and Purple RUO Discovery kit (760–229) (Roche Diagnostic, Indianapolis, IN). Spleen was used as positive control and negative control was obtained following antibody omission. Image analysis was performed with a multispectral imaging system (Nuance, Perkin Elmer). The wavelengths of the antibodies, melanin and hematoxylin were calculated and the resulting spectra library was used to assess the percentage of tissue stained by CD4 and CD8. Sixteen optical density images were acquired per group at 40x. Results were expressed by the percent of positivity of the purple signal representing each antibody respectively within the image. For statistical analysis the values were expressed as the mean ± S.E. All statical analyses were performed using Wilcoxon / Kruskal-Wallis tests. Unless otherwise specified a p-value of 0.05 was used for statistically significant differences among groups.

A375, A2058, and B16-F25 melanoma cells were obtained from (ATCC, Manassas, VA), and cultured in Dulbecco’s Modified Eagle Medium with 10% fetal bovine serum (Life Technologies, Grand Island, NY) at a 37°C incubator supplemented with 5% CO2. They were confirmed to express Melan-A. For growth analysis, cells were seeded onto 96-well dishes at 5,000 cells/well, and next day treated in quadruples with varying concentrations of trametinib and chloroquine (Selleckchem, Houston, TX). Two days later, cells were incubated with 5 µL 3-[4,5-dimethylthiazol-2-yl]-2,5 diphenyl tetrazolium bromide (20mg/mL, Sigma) for 2 hours and media were then replaced with DMSO. The optical density at 590 nm was measured using a plate reader (Synergy H1, BioTek Winooski, VT). For protein analysis, cells were cultured in 6-cm dishes, treated with chloroquine and trametinib for 24-48 hours, and then lysed with RIPA buffer.

Protein extracts (20-30 µg/sample) were separated by 10% acrylamide gel SDS–PAGE, and immunoblotted with antibodies against p62/SQSTM1 (5114S), Cyclin D1 (5114S), pErk (2978S), CDK4 (12790S), p21 (2946S), pSTAT1 (9167S) from (Cell Signaling Technology) and pc-Jun (PAS-17879) (ThermoFisher Scientific). Actin was used as control (SC-1516, Santa Cruz Biotechnology, Santa Cruz, CA). Membranes were blocked with 5% BSA (Sigma-Aldrich) in Tris-buffered saline, 0.1% Tween 20 and, after incubation with primary antibodies, signals were visualized with IR-dye-conjugated secondary antibodies and scanned using Odyssey imaging system (Li-COR, Lincoln, NE).

To assess the response to treatments, we induced melanoma growth in 5-7 weeks old Tyr-CreER. Braf^Ca.Pten^fl/fl mice via 3 topical applications of 1.5 µl of 5 mM 4-hydroxytamoxifen (4-OHT) on the back skin spaced at 1-day intervals. About 1-week after the induction when pigmented melanoma lesions became visible, animals (n=4-8/group) were treated via intraperitoneal injection of the solvent control, the MEK inhibitor trametinib (TRA, 3 mg/Kg) either alone or together with chloroquine (CQ, 40 mg/kg) ( Figure 1A ). In agreement with previous studies on the conditional Braf mutant melanoma model (40), the solvent control animals developed darkly pigmented spot lesions about 1 week after induction and they reached the humane end-point 4 weeks later due to an aggressive melanoma growth ( Figure 1B , Supplementary Figure S1A, B ). In addition, some control animals showed asynchronous tumor formation with dome-shaped papules of variable size that enlarged and ulcerated ( Supplementary Figure S1B ). In contrast, animals treated with TRA showed a reduced tumor growth, which was further enhanced by the combination treatment of CQ and TRA ( Figure 1B ). Quantification of the tumor areas showed a significant decrease (p<0.05) of the lesions in the TRA and combo treatments compared to the control subjects ( Figure 1C ).

To verify the efficiency of drug delivery, we performed immunostaining for pERK, downstream substrate of MEK, and LC3, an autophagy adaptor protein destined for degradation in autophaygosome (47, 48). We found that pERK-positive cells were significantly reduced in tumors treated with TRA alone or together with CQ ( Figure 2A ). The intensity of LC3 was markedly elevated in tissues treated with CQ ( Supplementary Figure S2 ). These data indicate that TRA and CQ achieved inhibition of MEK and autophagy, respectively. To examine the effects on cell proliferation and cell death, we performed immunostaining for the cell proliferation marker Ki-67 and TUNNEL assay, respectively. We then quantified Ki-67+ and TUNNEL+ cells from 10-15 images of each group. We found that the number of Ki-67+ cells were significantly decreased in tissues of the TRA and CQ combo group compared to the control and the TRA single agent treatment groups (p<0.05) ( Figure 2B ). In agreement with the in vivo data, treatment of TRA alone or together with CQ decreased pERK and Cyclin D1 expression in B16, A2058 and A375 melanoma cells ( Supplementary Figure S3 ). Interestingly, the number of TUNNEL+ cells was increased in tumors treated with trametinib either alone or combo with CQ, but to our surprise this increase was less pronounced in the combo group (p<0.05) than that of trametinib alone (p<0.01) ( Figure 2C ). These data indicate that TRA and CQ combo delayed melanoma growth primarily through an enhanced inhibition of cell proliferation.

All melanoma lesions of the animals treated with TRA either alone or together with CQ appeared intensely dark pigmented, whereas the control group appeared more heterogeneous with most being dark and some unpigmented nodules ( Supplementary Figure S1 ). Consistent with the clinical presentation, histological analysis showed that all treated tumors and, most but not all, control tumors contained pigmented cells ( Figure 3A ). MITF is a critical regulator of melanocyte growth and differentiation (49). Immunostaining showed that TRA treatment induced a 30% increase of MITF+ cells and the effect of combo treatment did not reach a significance ( Figure 3B ). In agreement with the in vivo data, treatment of B16 melanoma cells with TRA and CQ induced increased pigmentation and cell death ( Supplementary Figures S4A–C ). By qRT-PCR, we found that in vitro treatment of CQ increased MITF mRNA levels in human melanoma cell lines ( Supplementary Figure S5A, B ). These data indicate that treatments of TRA and CQ inhibited melanoma cell dedifferentiation.

Cancer cell death can be caused by cancer cell-intrinsic mechanisms and by immune cells. Host immunity is crucial for the anti-melanoma activity of BRAF and MEK inhibitors (50, 51). Previous reports and our in vitro studies have shown that high concentrations of CQ increases melanoma cell apoptosis (36). It is therefore surprising that TUNEL+ apoptotic cell numbers were less pronounced in the CQ and trametinib combo group than that of trametinib alone ( Figure 2C ). This result led us to suggest that immune cell activity might be reduced by CQ in the tumor microenvironment. In this regard, it was previously suggested that CQ treatment of immunological diseases such as arthritis (52), can have a direct effect on immune cells (53–55). By immunostaining, we found that CD4 and CD+8 T-lymphocytes were significantly decreased in tissues treated with CQ and TRA ( Figures 4A, B ). F4/80+ macrophages showed a trend decrease by the treatments, but did not reach a significance ( Figure 4C ). These results indicate that CQ decreased immune cell infiltration into the tumor microenvironment.

IFNγ is commonly used to assess effector immune cell activities (56). While we were not able to distinguish immune cell expression status of IFNγ in the tissue sections by immunostaining, we found that IFNγ was readily detected in hair follicles and was increased in epidermal and dermal cells of trametinib and CQ. However, the combo treatment did not reach significance compared to TRA alone ( Supplementary Figure S6A ). This is rather intriguing as MEK inhibitors including trametinib is associated with acneiform dermatitis, owing to off-target effects on KLF4/NF-κB-dependent transcription of inflammatory cytokines (57). We examined whether trametinib directly affected interferon signaling by immunoblotting of melanoma cells treated in vitro. We found that pSTAT1 was decreased in human melanoma cells treated with trametinib at 0.1 µM concentration either alone or with CQ ( Supplementary Figure S6B ). These data indicate that the observed IFNγ changes may be a result of hair cycling or require cell-cell communications in an in vivo setting.