WO2018167519A1

WO2018167519A1 - Biomarker for identifying responders to cancer treatment

Info

Publication number: WO2018167519A1
Application number: PCT/GB2018/050701
Authority: WO
Inventors: Marco RANZANI; David Adams; Daniel Simon Peeper; Oscar Krijgsman; Kristel KEMPER
Original assignee: Genome Research Limited; Stichting Het Nederlands Kanker Instituut - Antoni Van Leeuwenhoek Ziekenhuis
Priority date: 2017-03-17
Filing date: 2018-03-16
Publication date: 2018-09-20
Also published as: GB201704267D0

Abstract

The invention relates to the use of AXL as a biomarker for identifying responders to cancer treatment with a pharmaceutical composition comprising one or more tyrosine kinase inhibitors in combination with one or more MEK inhibitors. The invention also relates to said pharmaceutical compositions for use in the treatment of cancer and methods of treating cancer comprising said pharmaceutical compositions.

Description

BIOMARKER FOR IDENTIFYING RESPONDERS TO CANCER TREATMENT

FIELD OF THE INVENTION

The invention relates to the use of AXL as a biomarker for identifying responders to cancer treatment with a pharmaceutical composition comprising one or more tyrosine kinase inhibitors in combination with one or more MEK inhibitors. The invention also relates to said pharmaceutical compositions for use in the treatment of cancer and methods of treating cancer comprising said pharmaceutical compositions. BACKGROUND OF THE INVENTION

The combination of a BRAF and a MEK inhibitor has achieved response rates of up to 68% and median progression free survival of up to 12 months in BRAF^V600E-mutant melanoma patients^{1 ,2}. Unfortunately, however, recurrence is almost inevitable³. More recently immunotherapies have further revolutionized the treatment of melanoma with immune checkpoint inhibitors achieving response rates >60% and unprecedented long term durable responses^4,5. Notwithstanding these advances, most melanoma patients are not cured by available therapies. Further, second line therapies are required for patients who relapsed following immunotherapies. Recent sequencing efforts have distinguished two subsets of BRAF/NRAS wildtype (WT) melanoma, which represents 25-30% of all melanoma cases^{6 7}. These groups are NF1 mutant (NF1 m), which carry mutations inactivating NF1 and generally have a high mutational load, and triple wild type melanoma (TWT), characterized by a high frequency of copy number alterations, and the absence of mutations in BRAF, NRAS and NF1^{6 7}. While some BRAF/NRAS WT melanomas carry mutations in the driver gene KIT and may respond to KIT inhibitors^{8 9}, most patients do not benefit from targeted therapies highlighting the need for new therapeutic approaches.

There is therefore a great need for new approaches to treat these patients.

SUMMARY OF THE INVENTION

According to a first aspect of the invention, there is provided the use of AXL as a biomarker for identifying responders to cancer treatment with a pharmaceutical composition comprising one or more tyrosine kinase inhibitors in combination with one or more MEK inhibitors.

According to a second aspect of the invention, there is provided a pharmaceutical composition comprising one or more tyrosine kinase inhibitors in combination with one or more MEK inhibitors for use in the treatment of cancer in a subject identified as having AXL expression in a biological sample obtained from said subject.

According to a further aspect of the invention there is provided a method of treating cancer in a subject which comprises the steps of:

(a) detecting the presence of AXL expression in a biological sample obtained from said subject; and

(b) administering a pharmaceutical composition comprising one or more tyrosine kinase inhibitors in combination with one or more MEK inhibitors to said subject identified as having AXL expression in said biological sample.

BRIEF DESCRIPTION OF THE FIGURES

Figure 1. High-throughput drug screening of melanoma cell lines

a) Outline of data generated for the collection of 22 melanoma cell lines (see Methods), b) Correlation between the frequency of mutation in BRAF/NRAS WT melanoma drivers in the 20 BRAF/NRAS WT cell lines of our collection (X axis) and in 74 BRAF/NRAS WT tumors from the TCGA cohort⁷. P value, R² and r by Pearson correlation, c) Delta AUC (Y axis, see Methods) of the library drugs combined with the different concentrations of the anchor drugs (X axis). Each dot represents the delta AUC of a drug combination in a cell line; the black line shows the mean, the dashed line the 0.2 delta AUC value, d) Number of combination that achieved a delta AUC>0.2 (Y axis) in recurrent (X axis) cell lines.

Figure 2. Confirmation of the synergy between nilotinib and MEK inhibitors in

BRAFINRAS WT melanoma

a) Survival curves of two representative cell lines treated with temozolomide plus olaparib (top panels, representative of a biological duplicate), nilotinib plus PD-0325901 (middle panel, representative of a biological triplicate) and nilotinib plus trametinib (bottom panel, representative of a biological triplicate). The Y axis shows viability vs vehicle treated control, X axis the concentration (μΜ for olaparib, nM for PD-0325901 and trametinib) of the library drug. The red line shows the viability of the cells treated with the library drug alone, the blue line the viability with the anchor drug alone, the yellow line the viability with the drug combination, the grey line the predicted additivity (see Methods). Each point is the average value of a technical triplicate, b) Clonogenic assays confirming the synergy between nilotinib and MEK inhibitors. The concentrations of the library and anchor drugs are indicated on the top or on the right, respectively. These assays are representative of a biological duplicate, c) Survival curves of a sensitive BRAF^V600E-mutant cell line in the independent validation collection (representative of a biological triplicate). Legend as in b. d) Summary of the delta AUC (top panel) and AUC combination (bottom panel) of nilotinib/trametinib combination in the two collections of cell lines. The dashed lines represent values of AUC = 0.4 and delta AUC = 0.1 (see Methods). Red symbols indicate cell lines with delta AUC>0.1. Figure 3. AXL expression and p-ERK downregulation are associated with the synergy between nilotinib and MEK inhibitors

a) Top panel: western blot (WB) for AXL and MITF expression in 5 sensitive and 4 non- sensitive representative cell lines. Middle panel: western blot for AXL expression in 5 non- sensitive cell lines. WM321 1 represents an outlier that expresses AXL but is not sensitive to the drug combination. HSP90 or beta tubulin is displayed below as loading control. Bottom panel: contingency table in BRAF/NRAS WT cell lines for AXL expression and synergy (defined as nilotinib/trametinib combination achieving a delta AUC>0.1); P-value by two tailed Fisher's exact test, b-c) Western blot for p-ERK in sensitive (b) and non-sensitive (c) cell line upon treatment with DMSO vehicle, nilotinib (2μΜ), trametinib (1 nM) or combination for 6h (see Methods). Beta tubulin (β-tubulin) loading control is displayed below the blot. The blot is representative of experiments conducted in biological triplicate. * indicates NRAS^{Q61 R}-mutant, ^Λ indicates BRAF^V600E-mutant cell lines; all the other cell lines are BRAF/NRAS WT. d) Level of p-ERK quantified by WB in sensitive and non- sensitive cell lines. For each biological triplicate experiment, the levels of p-ERK were normalized for the loading control, and then averaged among the 3 values of the triplicate (each point in the dot plots). The dot plot shows the average and the standard error mean of the group. Significance is calculated by unpaired Student's t-test.

Figure 4. Identification of mechanism of drug resistance to the nilotinib/trametinib combination by CRISPR-Cas9 genome-wide library screening

a) Venn diagram of genes conferring resistance to the drug combination in CHL-1 , C077, MeWo cell lines; number of genes and % of total are indicated. We considered genes with FDR<0.1 (by MAGeCK, see Methods) in both replicates b) Venn diagram of the genes conferring resistance to the nilotinib/trametinib or trametinib in 2 or more cell lines (criteria as in a), c) Network of protein-protein interaction for the genes conferring resistance to nilotinib/trametinib combination (criteria as in a-b). Blue line indicates binding, black line reaction, grey line unspecified interaction; PPI enr is the protein-protein interaction enrichment P-value calculated by STRING (see Methods). The coloured circles highlight genes belonging to top enriched pathways, d) Log2 of the normalized sgRNA count (see Methods) for each gRNA in vehicle treated (X axis) and drug combination treated (Y axis) CHL-1 cells after 18 days of treatment. The different sgRNAs targeting each of the top 10 enriched genes are color coded as detailed in the legend, e) CHL-1 KO clones for TSC1, TSC2, CDKN1B displayed significant proliferative advantage over the WT parental cell line upon treatment with nilotinib 2μΜ plus trametinib 100nM for two weeks. The KO clones were infected with a GFP expressing lentivirus, FACS sorted for GFP positivity, and mixed at a 5% ratio in a population of WT cells (see Methods). Fold expansion over WT in Y axis (with average and standard error mean of the triplicate). **** P<0.0001 by one way Anova and Tukey's multiple comparisons test.

Figure 5. The combination of nilotinib plus trametinib is synergistic in two in vivo models

a) Volume (Y axis) of tumors from MeWo cell line inoculated in NOD.Cg-Pr/ccfc³^ //2rg^{tm1 yil}/SzJ (NSG) mice upon treatment with vehicle (green), nilotinib 75mg/kg/day (blue), trametinib 0.1 mg/kg/day (red) or their combination (yellow) (n=10 tumors/group, see Methods). The graph shows the mean and the standard error mean. The vertical dashed red line highlights the start of the treatment. P-value calculated by unpaired Student's t-test on the last time point; *P<0.05, **P<0.01 , ****P<0.0001. b) Western blot for AXL, MITF and vinculin loading control in a collection of BRAF/NRAS WT melanoma PDX. c) Quantification of AXL RNA expression by Q-PCR in BRAF/NRAS WT PDX. d) Volume (Y axis) of tumors from M003.X2 PDX inoculated in NSG mice upon treatment with vehicle (green), nilotinib 75mg/kg/day (blue), trametinib 0.3mg/kg/day (red) or their combination (yellow). Graph and analysis as in a). The suffix .X1-.X2 indicates the passage number of the PDX line, e) Top panel: western blot for p- ERK, total ERK and vinculin loading control in 4 representative M003.X2 tumors per group of treatment (indicated above the plots) collected at the experimental endpoint. Bottom panel: quantification in p-ERK levels (Y axis, normalized for total ERK) from the western blot displayed above. Box plot extends from the 25th to 75th percentiles, whiskers from min to max, the middle line indicates the median. P-value by one way anova and Tukey's multiple comparisons test; *P<0.05, **P<0.01 f) Representative hematoxylin and eosin stained section of a tumor per each treatment group collected at the experimental endpoint. Scale bar = 100μηι, the bottom left corner display a higher magnification.

DETAILED DESCRIPTION OF THE INVENTION

Tyrosine Kinase Inhibitors

References herein to the term "tyrosine kinase inhibitor" include any pharmaceutical agent which inhibits tyrosine kinase. Tyrosine kinases are enzymes responsible for the activation of many proteins by signal transduction cascades. The proteins are activated by adding a phosphate group to the protein (phosphorylation), a step that tyrosine kinase inhibitors inhibit. Tyrosine kinase inhibitors are typically used as anticancer drugs. Examples of suitable tyrosine kinase inhibitors include one or more of: axitinib, bosutinib, cediranib, dasatinib, eriotinib, gefitinib, imatinib, lapatinib, lestaurtinib, nilotinib, semaxanib, sunitinib, ponatininb, bafetinib, vandetanib, cabozantinib, BMS-777607, R428 (BGB324), Gilteritinib, LDC1267, TP-0903, BGB324 and S49076. In one embodiment, the tyrosine kinase inhibitor is selected from nilotinib.

In one embodiment, the pharmaceutical composition comprises a single tyrosine kinase inhibitor. In a further embodiment, the pharmaceutical composition comprises a single tyrosine kinase inhibitor and said tyrosine kinase inhibitor is selected from: axitinib, bosutinib, cediranib, dasatinib, eriotinib, gefitinib, imatinib, lapatinib, lestaurtinib, nilotinib, semaxanib, sunitinib, ponatininb, bafetinib, vandetanib, cabozantinib, BMS-777607, R428 (BGB324), Gilteritinib, LDC1267, TP-0903, BGB324 and S49076.

In a yet further embodiment, the pharmaceutical composition comprises a single tyrosine kinase inhibitor and said tyrosine kinase inhibitor is selected from nilotinib.

References herein to "nilotinib" also include AMN107 which has the trade name Tasigna™. Nilotinib is a small-molecule tyrosine kinase inhibitor approved for the treatment of imatinib- resistant chronic myelogenous leukemia. Structurally related to imatinib, it was developed based on the structure of the Abl-imatinib complex to address imatinib intolerance and resistance. Nilotinib is a tyrosine kinase inhibitor that inhibits Bcr-Abl, KIT, LCK, EPHA3, EPHA8, DDR1 , DDR2, PDGFRB, MAPK11 and ZAK. Nilotinib is 10-30 fold more potent than imatinib in inhibiting Bcr-Abl tyrosine kinase activity and proliferation of Bcr-Abl expressing cells.

Nilotinib has the chemical structure shown as formula (I):

(I)

In one embodiment, nilotinib is present within the pharmaceutical composition as either a free base or a pharmaceutically acceptable salt or solvate thereof.

Examples of suitable pharmaceutically acceptable salts include mono- or di-salts formed with an acid selected from the group consisting of acetic, 2,2-dichloroacetic, adipic, alginic, ascorbic (e.g. L-ascorbic), L-aspartic, benzenesulfonic, benzoic, 4-acetamidobenzoic, butanoic, (+) camphoric, camphor-sulfonic, (+)-(1 S)-camphor-10-sulfonic, capric, caproic, caprylic, cinnamic, citric, cyclamic, dodecylsulfuric, ethane-1 ,2-disulfonic, ethanesulfonic, 2- hydroxyethanesulfonic, formic, fumaric, galactaric, gentisic, glucoheptonic, D-gluconic, glucuronic (e.g. D-glucuronic), glutamic (e.g. L-glutamic), a-oxoglutaric, glycolic, hippuric, hydrohalic acids (e.g. hydrobromic, hydrochloric, hydriodic), isethionic, lactic (e.g. (+)-L- lactic, (±)-DL-lactic), lactobionic, maleic, malic, (-)-L-malic, malonic, (±)-DL-mandelic, methanesulfonic, naphthalene-2-sulfonic, naphthalene-1 ,5-disulfonic, 1-hydroxy-2-naphthoic, nicotinic, nitric, oleic, orotic, oxalic, palmitic, pamoic, phosphoric, propionic, pyruvic, L- pyroglutamic, salicylic, 4-amino-salicylic, sebacic, stearic, succinic, sulfuric, tannic, (+)-L- tartaric, thiocyanic, p-toluenesulfonic, undecylenic and valeric acids, as well as acylated amino acids and cation exchange resins.

In one embodiment, nilotinib is present within the pharmaceutical composition as the hydrochloride salt. References herein to "solvates" include references to hydrates, alcoholates and the like.

The compounds used within the pharmaceutical composition of the invention may form solvates, for example with water (i.e., hydrates) or common organic solvents. As used herein, the term "solvate" means a physical association of the compounds of the present invention with one or more solvent molecules. This physical association involves varying degrees of ionic and covalent bonding, including hydrogen bonding. In certain instances the solvate will be capable of isolation, for example when one or more solvent molecules are incorporated in the crystal lattice of the crystalline solid. The term "solvate" is intended to encompass both solution-phase and isolatable solvates. Non-limiting examples of suitable solvates include compounds of the invention in combination with water, isopropanol, ethanol, methanol, DMSO, ethyl acetate, acetic acid or ethanolamine and the like. The compounds of the invention may exert their biological effects whilst they are in solution.

Solvates are well known in pharmaceutical chemistry. They can be important to the processes for the preparation of a substance (e.g. in relation to their purification, the storage of the substance (e.g. its stability) and the ease of handling of the substance and are often formed as part of the isolation or purification stages of a chemical synthesis. A person skilled in the art can determine by means of standard and long used techniques whether a hydrate or other solvate has formed by the isolation conditions or purification conditions used to prepare a given compound. Examples of such techniques include thermogravimetric analysis (TGA), differential scanning calorimetry (DSC), X-ray crystallography (e.g. single crystal X-ray crystallography or X-ray powder diffraction) and Solid State NMR (SS-NMR, also known as Magic Angle Spinning NMR or MAS-NMR). Such techniques are as much a part of the standard analytical toolkit of the skilled chemist as NMR, IR, HPLC and MS. Alternatively the skilled person can deliberately form a solvate using crystallisation conditions that include an amount of the solvent required for the particular solvate.

Thereafter the standard methods described above, can be used to establish whether solvates had formed. In one embodiment, nilotinib is present within the pharmaceutical composition as the hydrochloride monohydrate salt.

MEK Inhibitors

References herein to the term "MEK inhibitor" include any pharmaceutical agent that inhibits the mitogen-activated protein kinase enzymes MEK1 and/or MEK2 (also known as MAP2K1 and MAP2K2, respectively). MEK inhibitors are known to affect the MAPK/ERK pathway which is often overactive in some cancers.

Examples of suitable MEK inhibitors include one or more of: trametinib, cobimetinib (XL518), binimetinib (MEK162), selumetinib, PD-325901 , CI-1040, PD035901 , pimasertib, RG7304, SHR7390, ATR001 , ATR004, ATR005, CCT196969, CCT241 161 , CCD450, EBI 1051 , E601 and TAK-733. In one embodiment, the MEK inhibitor is selected from trametinib.

In one embodiment, the pharmaceutical composition comprises a single MEK inhibitor.

In one embodiment, the MEK inhibitor inhibits MEK1 and/or MEK2. In a further embodiment, the MEK inhibitor inhibits both MEK1 and MEK2.

In a further embodiment, the pharmaceutical composition comprises a single MEK inhibitor and said MEK inhibitor is selected from: trametinib, cobimetinib (XL518), binimetinib

(MEK162), selumetinib, PD-325901 , CI-1040, PD035901 , pimasertib, RG7304, SHR7390, ATR001 , ATR004, ATR005, CCT196969, CCT241161 , CCD450, EBI 1051 , E601 and TAK- 733. In a yet further embodiment, the pharmaceutical composition comprises a single MEK inhibitor and said MEK inhibitor is selected from trametinib.

References herein to "trametinib" also include reference to GSK1120212 and the trade name Mekinist™ which is a known cancer drug. Trametinib is known to inhibit both MEK1 and MEK2.

Trametinib had good results for metastatic melanoma carrying the BRAF V600E mutation in a phase III clinical trial. In this mutation, the amino acid valine (V) at position 600 within the BRAF gene has become replaced by glutamic acid (E) making the mutant BRAF gene constitutively active.

Trametinib has the chemical structure shown as formula (II):

(II)

In one embodiment, trametinib is present within the pharmaceutical composition as either a free base or a pharmaceutically acceptable salt or solvate thereof. In a further embodiment, trametinib is present within the pharmaceutical composition as a free base.

AXL Expression

References herein to "AXL" include references to AXL, AZF, AZFA, SP3, AZF1 , ARK, JTK11 , Tyro7, UFO or AXL receptor tyrosine kinase. AXL is a tyrosine kinase enzyme that in humans is encoded by the AXL gene. The AXL protein is a cell surface receptor. The AXL gene is evolutionarily conserved between vertebrate species. This gene has two different alternatively spliced transcript variants. The protein encoded by this gene is a member of the receptor tyrosine kinase subfamily. Although it is similar to other receptor tyrosine kinases, the AXL protein represents a unique structure of the extracellular region that juxtaposes IgL and FNIII repeats. The AXL receptor transduces signals from the extracellular matrix into the cytoplasm by binding growth factors like vitamin K-dependent protein growth-arrest-specific gene 6 (GAS6). It is involved in the stimulation of cell proliferation. This receptor can also mediate cell aggregation by homophilic binding. AXL is an essential epithelial-to- mesenchymal transition-induced regulator of breast cancer metastasis and patient survival.

It will be appreciated that detection of AXL expression may be achieved either by detecting protein or RNA expression, the detection of which will be readily apparent to the skilled person.

It will also be appreciated that references herein to AXL protein expression comprise detection of expression of the AXL protein wherein the AXL protein comprises the sequence identified by Ensembl ID No. ENSG00000167601 and Uniprot No. In particular, the AXL protein comprises the following 894 amino acid sequence: MAWRCPRMGR VPLAWCLALC GWACMAPRGT QAEESPFVGN PGNITGARGL TGTLRCQLQV QGEPPEVHWL RDGQILELAD STQTQVPLGE DEQDDWIVVS QLRITSLQLS DTGQYQCLVF LGHQTFVSQP GYVGLEGLPY FLEEPEDRTV AANTPFNLSC QAQGPPEPVD LLWLQDAVPL ATAPGHGPQR SLHVPGLNKT SSFSCEAHNA KGVTTSRTAT ITVLPQQPRN LHLVSRQPTE LEVAWTPGLS GIYPLTHCTL QAVLSNDGMG IQAGEPDPPE EPLTSQASVP PHQLRLGSLH PHTPYHIRVA CTSSQGPSSW THWLPVETPE GVPLGPPENI SATRNGSQAF VHWQEPRAPL QGTLLGYRLA YQGQDTPEVL MDIGLRQEVT LELQGDGSVS NLTVCVAAYT AAGDGPWSLP VPLEAWRPGQ AQPVHQLVKE PSTPAFSWPW WYVLLGAWA AACVLILALF LVHRRKKETR YGEVFEPTVE RGELWRYRV RKSYSRRTTE ATLNSLGISE ELKEKLRDVM VDRHKVALGK TLGEGEFGAV MEGQLNQDDS ILKVAVKTMK IAICTRSELE DFLSEAVCMK EFDHPNVMRL IGVCFQGSER ESFPAPWIL PFMKHGDLHS FLLYSRLGDQ PVYLPTQMLV KFMADIASGM EYLSTKRFIH RDLAARNCML NENMSVCVAD FGLSKKIYNG DYYRQGRIAK MPVKWIAIES LADRVYTSKS DVWSFGVTMW EIATRGQTPY PGVENSEIYD YLRQGNRLKQ PADCLDGLYA LMSRCWELNP QDRPSFTELR EDLENTLKAL PPAQEPDEIL YVNMDEGGGY PEPPGAAGGA DPPTQPDPKD SCSCLTAAEV HPAGRYVLCP STTPSPAQPA DRGSPAAPGQ EDGA (referred to as Uniprot No. P30530-1 ; SEQ ID NO: 1).

Alternatively, the AXL protein comprises an 885 amino acid sequence which is the short form derivative of SEQ ID NO: 1 wherein amino acids 429-437 are missing (referred to as Uniprot No.: P30530-2).

The inventors have surprisingly identified a synergistic interaction between a tyrosine kinase inhibitor (such as nilotinib) and a MEK inhibitor (such as trametinib or PD035901 ) when AXL protein expression has been detected (see data presented in Results section and Figures 1-5).

It will be appreciated that measurement of AXL protein or RNA expression may be conducted in accordance with procedures known in the art. In particular, a biological sample may be taken from the tumor mass of a subject suffering with cancer, such as the melanomas described herein, or any derivative of such tumor (including cell line, short term culture, xenotransplanted samples, organoid and the like) and if AXL protein expression is detected then said subject is identified as being suitable for treatment with the synergistic pharmaceutical composition of the invention. It will also be appreciated that any levels of detection of the AXL protein or RNA would justify the use of the composition of the invention. Thus, it will be appreciated that once a "responder" (i.e. a patient having AXL protein or RNA expression) has been identified, the patient will then suitably be administered a composition of the invention.

AXL protein detection can be conducted by a number of known techniques. In one embodiment, detection of AXL protein expression may be conducted by a biosensor capable of detecting the presence of the AXL protein or a portion thereof. Examples of biosensors are described herein. AXL RNA detection can also be conducted by a number of known techniques. In one embodiment, detection of AXL RNA expression may be conducted by using endpoint RT-PCR, quantitative real time RT-PCR (Taqman and related techniques), microarray hybridization, RNA sequencing, Northern Blotting, in situ hybridization, nuclease protection assay and SmartFlare probes.

Suitable biosensors may comprise a ligand or ligands, as described herein, capable of specific binding to the AXL protein. Such biosensors are useful in detecting the AXL protein.

The AXL protein may be directly detected, e.g. by SELDI or MALDI-TOF. Alternatively, the AXL protein may be detected directly or indirectly via interaction with a ligand or ligands such as an antibody or an AXL protein-binding fragment thereof, or other peptide, or ligand, e.g. aptamer, or oligonucleotide, capable of specifically binding the AXL protein. The ligand may possess a detectable label, such as a luminescent, fluorescent or radioactive label, and/or an affinity tag.

For example, detecting can be performed by one or more method(s) selected from the group consisting of: SELDI (-TOF), MALDI (-TOF), a 1-D gel-based analysis, a 2-D gel-based analysis, mass spectroscopy (MS) such as selected reaction monitoring (SRM), reverse phase (RP) LC, size permeation (gel filtration), ion exchange, affinity, HPLC, UPLC and other LC or LC MS-based techniques. Appropriate LC MS techniques include ICAT® (Applied Biosystems, CA, USA), or iTRAQ® (Applied Biosystems, CA, USA). Liquid chromatography (e.g. high pressure liquid chromatography (HPLC) or low pressure liquid chromatography (LPLC)), thin-layer chromatography, NMR (nuclear magnetic resonance) spectroscopy could also be used.

In one embodiment, the detecting is performed using mass spectroscopy (MS). In a further embodiment, the detecting is performed using selected reaction monitoring (SRM). SRM is a method used in tandem mass spectrometry in which an ion of a particular mass is selected in the first stage of a tandem mass spectrometer and an ion product of a fragmentation reaction of the precursor ion is selected in the second mass spectrometer stage for detection. Specific analyte panels can be developed for SRM matching the analytes on the biomarker panel. The analyte panels can quantitatively measure the protein analytes with high precision. This methodology has the advantage of allowing raw blood to be used instead of blood serum which minimizes the number intermediate processing steps.

Methods according to the invention may comprise analysing a sample of blood serum by SELDI-TOF or MALDI-TOF to detect the presence of the AXL protein. Detecting the AXL protein may be performed using an immunological method, involving an antibody, or a fragment thereof capable of specific binding to the AXL protein. Suitable immunological methods include sandwich immunoassays, such as sandwich ELISA, in which the detection of the AXL protein is performed using two antibodies which recognize different epitopes on the AXL protein; radioimmunoassays (RIA), direct, indirect or competitive enzyme linked immunosorbent assays (ELISA), enzyme immunoassays (EIA), Fluorescence immunoassays (FIA), western blotting, immunoprecipitation and any particle-based immunoassay (e.g. using gold, silver, or latex particles, magnetic particles, or Q-dots). Immunological methods may be performed, for example, in microtitre plate or strip format.

Immunological methods in accordance with the invention may be based, for example, on any of the following methods. Immunoprecipitation is the simplest immunoassay method; this measures the quantity of precipitate, which forms after the reagent antibody has incubated with the sample and reacted with the target antigen present therein to form an insoluble aggregate. Immunoprecipitation reactions may be qualitative or quantitative. In particle immunoassays, several antibodies are linked to the particle, and the particle is able to bind many antigen molecules simultaneously. This greatly accelerates the speed of the visible reaction. This allows rapid and sensitive detection of the AXL protein.

In immunonephelometry, the interaction of an antibody and target antigen on the AXL protein results in the formation of immune complexes that are too small to precipitate. However, these complexes will scatter incident light and this can be measured using a nephelometer. The antigen, i.e. AXL protein, concentration can be determined within minutes of the reaction.

Radioimmunoassay (RIA) methods employ radioactive isotopes such as I¹²⁵ to label either the antigen or antibody. The isotope used emits gamma rays, which are usually measured following removal of unbound (free) radiolabel. The major advantages of RIA, compared with other immunoassays, are higher sensitivity, easy signal detection, and well-established, rapid assays. The major disadvantages are the health and safety risks posed by the use of radiation and the time and expense associated with maintaining a licensed radiation safety and disposal program. For this reason, RIA has been largely replaced in routine clinical laboratory practice by enzyme immunoassays. Enzyme (EIA) immunoassays were developed as an alternative to radioimmunoassays (RIA). These methods use an enzyme to label either the antibody or target antigen. The sensitivity of EIA approaches that of RIA, without the danger posed by radioactive isotopes. One of the most widely used EIA methods for detection is the enzyme-linked immunosorbent assay (ELISA). ELISA methods may use two antibodies one of which is specific for the target antigen and the other of which is coupled to an enzyme, addition of the substrate for the enzyme results in production of a chemiluminescent or fluorescent signal.

Fluorescent immunoassay (FIA) refers to immunoassays which utilize a fluorescent label or an enzyme label which acts on the substrate to form a fluorescent product. Fluorescent measurements are inherently more sensitive than colorimetric (spectrophotometric) measurements. Therefore, FIA methods have greater analytical sensitivity than EIA methods, which employ absorbance (optical density) measurement. Chemiluminescent immunoassays utilize a chemiluminescent label, which produces light when excited by chemical energy; the emissions are measured using a light detector.

Immunological methods can thus be performed using well-known methods. Any direct (e.g., using a sensor chip) or indirect procedure may be used in the detection of the AXL protein.

The Biotin-Avidin or Biotin-Streptavidin systems are generic labelling systems that can be adapted for use in immunological methods. One binding partner (hapten, antigen, ligand, aptamer, antibody, enzyme etc) is labelled with biotin and the other partner (surface, e.g. well, bead, sensor etc) is labelled with avidin or streptavidin. This is conventional technology for immunoassays, gene probe assays and (bio)sensors, but is an indirect immobilisation route rather than a direct one. For example, a biotinylated ligand (e.g. antibody or aptamer) specific for the AXL protein may be immobilised on an avidin or streptavidin surface, the immobilised ligand may then be exposed to a sample containing or suspected of containing the AXL protein in order to detect the AXL protein. Detection of the immobilised antigen may then be performed by an immunological method as described herein.

The term "antibody" as used herein includes, but is not limited to: polyclonal, monoclonal, bispecific, humanised or chimeric antibodies, single chain antibodies, Fab fragments and F(ab')2 fragments, fragments produced by a Fab expression library, anti-idiotypic (anti-Id) antibodies and epitope-binding fragments of any of the above. The term "antibody" as used herein also refers to immunoglobulin molecules and immunologically-active portions of immunoglobulin molecules, i.e., molecules that contain an antigen binding site that specifically binds an antigen. The immunoglobulin molecules of the invention can be of any class {e.g., IgG, IgE, IgM, IgD and IgA) or subclass of immunoglobulin molecule.

The AXL protein can be detected using biosensor incorporating technologies based on "smart" holograms, or high frequency acoustic systems, such systems are particularly amenable to "bar code" or array configurations.

In smart hologram sensors (Smart Holograms Ltd, Cambridge, UK), a holographic image is stored in a thin polymer film that is sensitised to react specifically with the AXL protein. On exposure, the AXL protein reacts with the polymer leading to an alteration in the image displayed by the hologram. The test result read-out can be a change in the optical brightness, image, colour and/or position of the image. For qualitative and semi-quantitative applications, a sensor hologram can be read by eye, thus removing the need for detection equipment. A simple colour sensor can be used to read the signal when quantitative measurements are required. Opacity or colour of the sample does not interfere with operation of the sensor. The format of the sensor allows multiplexing for simultaneous detection of several substances. Reversible and irreversible sensors can be designed to meet different requirements, and continuous monitoring of the AXL protein is feasible. Suitably, biosensors for detection of the AXL protein combine biomolecular recognition with appropriate means to convert detection of the presence of the AXL protein in the sample into a signal. Biosensors can be adapted for "alternate site" diagnostic testing, e.g. in the ward, outpatients' department, surgery, home, field and workplace. Biosensors to detect the AXL protein include acoustic, plasmon resonance, holographic and microengineered sensors. Imprinted recognition elements, thin film transistor technology, magnetic acoustic resonator devices and other novel acousto-electrical systems may be employed in biosensors for detection of the AXL protein. Methods involving detection of the AXL protein can be performed on bench-top instruments, or can be incorporated onto disposable, diagnostic or monitoring platforms that can be used in a non-laboratory environment, e.g. in the physician's office or at the patient's bedside. Suitable biosensors for performing methods of the invention include "credit" cards with optical or acoustic readers. Biosensors can be configured to allow the data collected to be electronically transmitted to the physician for interpretation and thus can form the basis for e- neuromedicine. In one embodiment, AXL expression may be replaced by expression of a molecule, or a measurable fragment of said molecule, found upstream or downstream of AXL in a biological pathway. Examples of such molecules found upstream or downstream of AXL in a biological pathway include the pathways: MAPK-ERK, PI3K-AKT, Phospholipase C and N FKB inter alia.

Cancers

Examples of cancers (and their benign counterparts) which may be treated (or inhibited) include, but are not limited to tumours of epithelial origin (adenomas and carcinomas of various types including adenocarcinomas, squamous carcinomas, transitional cell carcinomas and other carcinomas) such as carcinomas of the bladder and urinary tract, breast, gastrointestinal tract (including the esophagus, stomach (gastric), small intestine, colon, rectum and anus), liver (hepatocellular carcinoma), gall bladder and biliary system, exocrine pancreas, kidney, lung (for example adenocarcinomas, small cell lung carcinomas, non-small cell lung carcinomas, bronchioalveolar carcinomas and mesotheliomas), head and neck (for example cancers of the tongue, buccal cavity, larynx, pharynx, nasopharynx, tonsil, salivary glands, nasal cavity and paranasal sinuses), ovary, fallopian tubes, peritoneum, vagina, vulva, penis, cervix, myometrium, endometrium, thyroid (for example thyroid follicular carcinoma), adrenal, prostate, skin and adnexae (for example melanoma, basal cell carcinoma, squamous cell carcinoma, keratoacanthoma, dysplastic naevus); haematological malignancies (i.e. leukemias, lymphomas) and premalignant haematological disorders and disorders of borderline malignancy including haematological malignancies and related conditions of lymphoid lineage (for example acute lymphocytic leukemia [ALL], chronic lymphocytic leukemia [CLL], B-cell lymphomas such as diffuse large B-cell lymphoma

[DLBCL], follicular lymphoma, Burkitt's lymphoma, mantle cell lymphoma, T-cell lymphomas and leukaemias, natural killer [NK] cell lymphomas, Hodgkin's lymphomas, hairy cell leukaemia, monoclonal gammopathy of uncertain significance, plasmacytoma, multiple myeloma, and post-transplant lymphoproliferative disorders), and haematological malignancies and related conditions of myeloid lineage (for example acute

myelogenousleukemia [AML], chronic myelogenousleukemia [CML], chronic

myelomonocyticleukemia [CMML], hypereosinophilic syndrome, myeloproliferative disorders such as polycythaemia vera, essential thrombocythaemia and primary myelofibrosis, myeloproliferative syndrome, myelodysplasia syndrome, and promyelocyticleukemia);

tumours of mesenchymal origin, for example sarcomas of soft tissue, bone or cartilage such as osteosarcomas, fibrosarcomas, chondrosarcomas, rhabdomyosarcomas,

leiomyosarcomas, liposarcomas, angiosarcomas, Kaposi's sarcoma, Ewing's sarcoma, synovial sarcomas, epithelioid sarcomas, gastrointestinal stromal tumours, benign and malignant histiocytomas, and dermatofibrosarcomaprotuberans; tumours of the central or peripheral nervous system (for example astrocytomas, gliomas and glioblastomas, meningiomas, ependymomas, pineal tumours and schwannomas); endocrine tumours (for example pituitary tumours, adrenal tumours, islet cell tumours, parathyroid tumours, carcinoid tumours and medullary carcinoma of the thyroid); ocular and adnexal tumours (for example retinoblastoma); germ cell and trophoblastic tumours (for example teratomas, seminomas, dysgerminomas, hydatidiform moles and choriocarcinomas); and paediatric and embryonal tumours (for example medulloblastoma, neuroblastoma, Wilms tumour, and primitive neuroectodermal tumours); or syndromes, congenital or otherwise, which leave the patient susceptible to malignancy (for example Xeroderma Pigmentosum).

In one embodiment, the cancer is selected from a tumour characterised by AXL expression. In a further embodiment, the cancer is selected from malignant melanoma. In a yet further embodiment, the melanoma is selected from BRAF/NRAS wildtype (WT) melanoma. In a still yet further embodiment, the BRAF/NRAS wildtype (WT) melanoma is selected from the NF1 mutant (NF1 m) or the triple wild type melanoma (TWT) (i.e. BRAF, NRAS and NF1⁶-⁷).

Pharmaceutical Compositions

In one embodiment the pharmaceutical composition is a sterile pharmaceutical composition.

Thus, the present invention further provides pharmaceutical compositions, as defined above, and methods of making a pharmaceutical composition comprising (e.g. admixing) at least one of the components of the formulation as defined herein together with one or more pharmaceutically acceptable excipients and optionally other therapeutic or prophylactic agents, as described herein.

The pharmaceutically acceptable excipient(s) can be selected from, for example, carriers (e.g. a solid, liquid or semi-solid carrier), adjuvants, diluents, fillers or bulking agents, granulating agents, coating agents, release-controlling agents, binding agents, disintegrants, lubricating agents, preservatives, antioxidants, buffering agents, suspending agents, thickening agents, flavouring agents, sweeteners, taste masking agents, stabilisers or any other excipients conventionally used in pharmaceutical compositions. Examples of excipients for various types of pharmaceutical compositions are set out in more detail below.

The term "pharmaceutically acceptable" as used herein pertains to compounds, materials, compositions, and/or dosage forms which are, within the scope of sound medical judgment, suitable for use in contact with the tissues of a subject (e.g. human) without excessive toxicity, irritation, allergic response, or other problem or complication, commensurate with a reasonable benefit/risk ratio. Each carrier, excipient, etc. must also be "acceptable" in the sense of being compatible with the other ingredients of the formulation.

Pharmaceutical compositions can be formulated in accordance with known techniques, see for example, Remington's Pharmaceutical Sciences, Mack Publishing Company, Easton, PA, USA. The pharmaceutical compositions can be in any form suitable for oral, parenteral, topical, intranasal, intrabronchial, sublingual, ophthalmic, otic, rectal, intra-vaginal, or transdermal administration. Where the compositions are intended for parenteral administration, they can be formulated for intravenous, intramuscular, intraperitoneal, subcutaneous administration or for direct delivery into a target organ or tissue by injection, infusion or other means of delivery. The delivery can be by bolus injection, short term infusion or longer term infusion and can be via passive delivery or through the utilisation of a suitable infusion pump or syringe driver.

Pharmaceutical formulations adapted for parenteral administration include aqueous and non- aqueous sterile injection solutions which may contain anti-oxidants, buffers, bacteriostats, co-solvents, surface active agents, organic solvent mixtures, cyclodextrin complexation agents, emulsifying agents (for forming and stabilizing emulsion formulations), liposome components for forming liposomes, gellable polymers for forming polymeric gels, lyophilisation protectants and combinations of agents for, inter alia, stabilising the active ingredient in a soluble form and rendering the formulation isotonic with the blood of the intended recipient. Pharmaceutical formulations for parenteral administration may also take the form of aqueous and nonaqueous sterile suspensions which may include suspending agents and thickening agents (R. G. Strickly, Solubilizing Excipients in oral and injectable formulations, Pharmaceutical Research, Vol 21 (2) 2004, p 201-230).

The formulations may be presented in unit-dose or multi-dose containers, for example sealed ampoules, vials and prefilled syringes, and may be stored in a freeze-dried (lyophilised) condition requiring only the addition of the sterile liquid carrier, for example water for injections, immediately prior to use. In one embodiment, the formulation is provided as an active pharmaceutical ingredient in a bottle for subsequent reconstitution using an appropriate diluent.

The pharmaceutical formulation can be prepared by lyophilising components of the formulation as defined herein or sub-groups thereof. Lyophilisation refers to the procedure of freeze-drying a composition. Freeze-drying and lyophilisation are therefore used herein as synonyms.

Extemporaneous injection solutions and suspensions may be prepared from sterile powders, granules and tablets.

Pharmaceutical compositions of the present invention for parenteral injection can also comprise pharmaceutically acceptable sterile aqueous or non-aqueous solutions, dispersions, suspensions or emulsions as well as sterile powders for reconstitution into sterile injectable solutions or dispersions just prior to use.

Examples of suitable aqueous and nonaqueous carriers, diluents, solvents or vehicles include water, ethanol, polyols (such as glycerol, propylene glycol, polyethylene glycol, and the like), carboxymethylcellulose and suitable mixtures thereof, vegetable oils (such as sunflower oil, safflower oil, corn oil or olive oil), and injectable organic esters such as ethyl oleate. Proper fluidity can be maintained, for example, by the use of thickening or coating materials such as lecithin, by the maintenance of the required particle size in the case of dispersions, and by the use of surfactants.

The compositions of the present invention may also contain adjuvants such as

preservatives, wetting agents, emulsifying agents, and dispersing agents. Prevention of the action of microorganisms may be ensured by the inclusion of various antibacterial and antifungal agents, for example, paraben, chlorobutanol, phenol, sorbic acid, and the like. It may also be desirable to include agents to adjust tonicity such as sugars, sodium chloride, and the like. Prolonged absorption of the injectable pharmaceutical form may be brought about by the inclusion of agents which delay absorption such as aluminum monostearate and gelatin.

In one particular embodiment of the invention, the pharmaceutical composition is in a form suitable for i.v. administration, for example by injection or infusion. For intravenous administration, the solution can be dosed as is, or can be injected into an infusion bag (containing a pharmaceutically acceptable excipient, such as 0.9% saline or 5% dextrose), before administration. In another particular embodiment, the pharmaceutical composition is in a form suitable for sub-cutaneous (s.c.) administration.

Pharmaceutical dosage forms suitable for oral administration include tablets (coated or uncoated), capsules (hard or soft shell), caplets, pills, lozenges, syrups, solutions, powders, granules, elixirs and suspensions, sublingual tablets, wafers or patches such as buccal patches.

Thus, tablet compositions can contain a unit dosage of active compound together with an inert diluent or carrier such as a sugar or sugar alcohol, eg; lactose, sucrose, sorbitol or mannitol; and/or a non-sugar derived diluent such as sodium carbonate, calcium phosphate, calcium carbonate, or a cellulose or derivative thereof such as microcrystalline cellulose (MCC), methyl cellulose, ethyl cellulose, hydroxypropyl methyl cellulose, and starches such as corn starch. Tablets may also contain such standard ingredients as binding and granulating agents such as polyvinylpyrrolidone, disintegrants (e.g. swellable crosslinked polymers such as crosslinked carboxymethylcellulose), lubricating agents (e.g. stearates), preservatives (e.g. parabens), antioxidants (e.g. BHT), buffering agents (for example phosphate or citrate buffers), and effervescent agents such as citrate/bicarbonate mixtures.

Tablets may be designed to release the drug either upon contact with stomach fluids (immediate release tablets) or to release in a controlled manner (controlled release tablets) over a prolonged period of time or with a specific region of the Gl tract.

Capsule formulations may be of the hard gelatin or soft gelatin variety and can contain the active component in solid, semi-solid, or liquid form. Gelatin capsules can be formed from animal gelatin or synthetic or plant derived equivalents thereof.

The solid dosage forms (eg; tablets, capsules etc.) can be coated or un-coated. Coatings may act either as a protective film (e.g. a polymer, wax or varnish) or as a mechanism for controlling drug release or for aesthetic or identification purposes. The coating (e.g. a Eudragit™ type polymer) can be designed to release the active component at a desired location within the gastro-intestinal tract. Thus, the coating can be selected so as to degrade under certain pH conditions within the gastrointestinal tract, thereby selectively release the compound in the stomach or in the ileum, duodenum, jejenum or colon. Instead of, or in addition to, a coating, the drug can be presented in a solid matrix comprising a release controlling agent, for example a release delaying agent which may be adapted to release the compound in a controlled manner in the gastrointestinal tract. Alternatively the drug can be presented in a polymer coating e.g. a polymethacrylate polymer coating, which may be adapted to selectively release the compound under conditions of varying acidity or alkalinity in the gastrointestinal tract. Alternatively, the matrix material or release retarding coating can take the form of an erodible polymer (e.g. a maleic anhydride polymer) which is substantially continuously eroded as the dosage form passes through the gastrointestinal tract. In another alternative, the coating can be designed to disintegrate under microbial action in the gut. As a further alternative, the active compound can be formulated in a delivery system that provides osmotic control of the release of the compound. Osmotic release and other delayed release or sustained release formulations (for example formulations based on ion exchange resins) may be prepared in accordance with methods well known to those skilled in the art.

The components of the formulation as defined herein may be formulated with a carrier and administered in the form of nanoparticles, the increased surface area of the nanoparticles assisting their absorption. In addition, nanoparticles offer the possibility of direct penetration into the cell. Nanoparticle drug delivery systems are described in "Nanoparticle Technology for Drug Delivery", edited by Ram B Gupta and Uday B. Kompella, Informa Healthcare, ISBN 9781574448573, published 13^th March 2006. Nanoparticles for drug delivery are also described in J. Control. Release, 2003, 91 (1-2), 167-172, and in Sinha et al., Mol. Cancer Ther. August 1 , (2006) 5, 1909.

The pharmaceutical compositions typically comprise from approximately 1 % (w/w) to approximately 95% (w/w) active ingredient and from 99% (w/w) to 5% (w/w) of a

pharmaceutically acceptable excipient or combination of excipients. Particularly, the compositions comprise from approximately 20% (w/w) to approximately 90%,% (w/w) active ingredient and from 80% (w/w) to 10% of a pharmaceutically acceptable excipient or combination of excipients. The pharmaceutical compositions comprise from approximately 1 % to approximately 95%, particularly from approximately 20% to approximately 90%, active ingredient. Pharmaceutical compositions according to the invention may be, for example, in unit dose form, such as in the form of ampoules, vials, suppositories, pre-filled syringes, dragees, tablets or capsules. The pharmaceutically acceptable excipient(s) can be selected according to the desired physical form of the formulation and can, for example, be selected from diluents (e.g solid diluents such as fillers or bulking agents; and liquid diluents such as solvents and co- solvents), disintegrants, buffering agents, lubricants, flow aids, release controlling (e.g. release retarding or delaying polymers or waxes) agents, binders, granulating agents, pigments, plasticizers, antioxidants, preservatives, flavouring agents, taste masking agents, tonicity adjusting agents and coating agents.

The skilled person will have the expertise to select the appropriate amounts of ingredients for use in the formulations. For example, tablets and capsules typically contain 0-20% disintegrants, 0-5% lubricants, 0-5% flow aids and/or 0-99% (w/w) fillers/ or bulking agents (depending on drug dose). They may also contain 0-10% (w/w) polymer binders, 0-5% (w/w) antioxidants, 0-5% (w/w) pigments. Slow release tablets would in addition contain 0-99% (w/w) release-controlling (e.g. delaying) polymers (depending on dose). The film coats of the tablet or capsule typically contain 0-10% (w/w) polymers, 0-3% (w/w) pigments, and/or 0-2% (w/w) plasticizers.

Parenteral formulations typically contain 0-20% (w/w) buffers, 0-50% (w/w) cosolvents, and/or 0-99% (w/w) Water for Injection (WFI) (depending on dose and if freeze dried).

Formulations for intramuscular depots may also contain 0-99% (w/w) oils.

Pharmaceutical compositions for oral administration can be obtained by combining the active ingredient with solid carriers, if desired granulating a resulting mixture, and processing the mixture, if desired or necessary, after the addition of appropriate excipients, into tablets, dragee cores or capsules. It is also possible for them to be incorporated into a polymer or waxy matrix that allow the active ingredients to diffuse or be released in measured amounts.

The compounds of the invention can also be formulated as solid dispersions. Solid dispersions are homogeneous extremely fine disperse phases of two or more solids. Solid solutions (molecularly disperse systems), one type of solid dispersion, are well known for use in pharmaceutical technology (see (Chiou and Riegelman, J. Pharm. Sci., 60, 1281- 1300 (1971)) and are useful in increasing dissolution rates and increasing the bioavailability of poorly water-soluble drugs. This invention also provides solid dosage forms comprising the solid solution described above. Solid dosage forms include tablets, capsules, chewable tablets and dispersible or effervescent tablets. Known excipients can be blended with the solid solution to provide the desired dosage form. For example, a capsule can contain the solid solution blended with (a) a disintegrant and a lubricant, or (b) a disintegrant, a lubricant and a surfactant. In addition a capsule can contain a bulking agent, such as lactose or microcrystalline cellulose. A tablet can contain the solid solution blended with at least one disintegrant, a lubricant, a surfactant, a bulking agent and a glidant. A chewable tablet can contain the solid solution blended with a bulking agent, a lubricant, and if desired an additional sweetening agent (such as an artificial sweetener), and suitable flavours. Solid solutions may also be formed by spraying solutions of drug and a suitable polymer onto the surface of inert carriers such as sugar beads ('non-pareils'). These beads can subsequently be filled into capsules or compressed into tablets.

The pharmaceutical formulations may be presented to a patient in "patient packs" containing an entire course of treatment in a single package, usually a blister pack. Patient packs have an advantage over traditional prescriptions, where a pharmacist divides a patient's supply of a pharmaceutical from a bulk supply, in that the patient always has access to the package insert contained in the patient pack, normally missing in patient prescriptions. The inclusion of a package insert has been shown to improve patient compliance with the physician's instructions. Compositions for topical use and nasal delivery include ointments, creams, sprays, patches, gels, liquid drops and inserts (for example intraocular inserts). Such compositions can be formulated in accordance with known methods.

Examples of formulations for rectal or intra-vaginal administration include pessaries and suppositories which may be, for example, formed from a shaped moldable or waxy material containing the active compound. Solutions of the active compound may also be used for rectal administration.

Compositions for administration by inhalation may take the form of inhalable powder compositions or liquid or powder sprays, and can be administrated in standard form using powder inhaler devices or aerosol dispensing devices. Such devices are well known. For administration by inhalation, the powdered formulations typically comprise the active compound together with an inert solid powdered diluent such as lactose. The components of the formulation as defined herein will generally be presented in unit dosage form and, as such, will typically contain sufficient compound to provide a desired level of biological activity. For example, a formulation may contain from 1 nanogram to 2 grams of active ingredient, e.g. from 1 nanogram to 2 milligrams of active ingredient. Within these ranges, particular sub-ranges of compound are 0.1 milligrams to 2 grams of active ingredient (more usually from 10 milligrams to 1 gram, e.g. 50 milligrams to 500 milligrams), or 1 microgram to 20 milligrams (for example 1 microgram to 10 milligrams, e.g. 0.1 milligrams to 2 milligrams of active ingredient).

For oral compositions, a unit dosage form may contain from 1 milligram to 2 grams, more typically 10 milligrams to 1 gram, for example 50 milligrams to 1 gram, e.g. 100 miligrams to 1 gram, of each of said active compound.

The active compound will be administered to a patient in need thereof (for example a human or animal patient) in an amount sufficient to achieve the desired therapeutic effect.

Methods of Treatment

The composition of the invention is generally administered to a subject in need of such administration, for example a human or animal patient, particularly a human.

The composition will typically be administered in amounts that are therapeutically or prophylactically useful and which generally are non-toxic. However, in certain situations (for example in the case of life threatening diseases), the benefits of administering the composition may outweigh the disadvantages of any toxic effects or side effects, in which case it may be considered desirable to administer the composition in amounts that are associated with a degree of toxicity.

The composition may be administered over a prolonged term to maintain beneficial therapeutic effects or may be administered for a short period only. Alternatively they may be administered in a continuous manner or in a manner that provides intermittent dosing (e.g. a pulsatile manner). A typical daily dose of the composition can be in the range from 100 picograms to 100 milligrams per kilogram of body weight, more typically 5 nanograms to 25 milligrams per kilogram of bodyweight, and more usually 10 nanograms to 15 milligrams per kilogram (e.g. 10 nanograms to 10 milligrams, and more typically 1 microgram per kilogram to 20 milligrams per kilogram, for example 1 microgram to 10 milligrams per kilogram) per kilogram of bodyweight although higher or lower doses may be administered where required. The composition can be administered on a daily basis or on a repeat basis every 2, or 3, or 4, or 5, or 6, or 7, or 10 or 14, or 21 , or 28 days for example. The composition may be administered orally in a range of doses, for example 1 to 1500 mg, 2 to 800 mg, or 5 to 500 mg, e.g. 2 to 200 mg or 10 to 1000 mg, particular examples of doses including 10, 20, 50 and 80 mg. The composition may be administered once or more than once each day. The composition can be administered continuously (i.e. taken every day without a break for the duration of the treatment regimen). Alternatively, the composition can be administered intermittently (i.e. taken continuously for a given period such as a week, then discontinued for a period such as a week and then taken continuously for another period such as a week and so on throughout the duration of the treatment regimen).

Examples of treatment regimens involving intermittent administration include regimens wherein administration is in cycles of one week on, one week off; or two weeks on, one week off; or three weeks on, one week off; or two weeks on, two weeks off; or four weeks on two weeks off; or one week on three weeks off - for one or more cycles, e.g. 2, 3, 4, 5, 6, 7, 8, 9 or 10 or more cycles.

In one particular dosing schedule, a patient will be given an infusion of the composition for periods of one hour daily for up to ten days in particular up to five days for one week, and the treatment repeated at a desired interval such as two to four weeks, in particular every three weeks.

More particularly, a patient may be given an infusion of the composition for periods of one hour daily for 5 days and the treatment repeated every three weeks.

In another particular dosing schedule, a patient is given an infusion over 30 minutes to 1 hour followed by maintenance infusions of variable duration, for example 1 to 5 hours, e.g. 3 hours.

In a further particular dosing schedule, a patient is given a continuous infusion for a period of 12 hours to 5 days, an in particular a continuous infusion of 24 hours to 72 hours. In another particular dosing schedule, a patient is given the composition orally once a week.

In another particular dosing schedule, a patient is given the composition orally once-daily for between 7 and 28 days such as 7, 14 or 28 days.

In another particular dosing schedule, a patient is given the composition orally once-daily for

1 day, 2 days, 3 days, 5 days or 1 week followed by the required amount of days off to complete a one or two week cycle.

2 weeks followed by 2 weeks off.

In another particular dosing schedule, a patient is given the composition orally once-daily for 2 weeks followed by 1 week off.

In another particular dosing schedule, a patient is given the composition orally once-daily for 1 week followed by 1 week off. Ultimately, however, the quantity of composition administered and the type of composition used will be commensurate with the nature of the disease or physiological condition being treated and will be at the discretion of the physician.

The following study illustrates the invention:

High-throughput drug screens of cancer cell lines represent an effective approach to identify candidate compounds with high activity in specific subtypes of human cancer^10,11. Given the compelling need for new targeted therapies for patients with BRAF/NRAS WT melanoma, we performed a high-throughput drug combination screen in 20 BRAF/NRAS WT melanoma cell lines generating over 17,000 viability curves with 180 drugs combinations. We subsequently defined the mechanism of synergy of the lead combination and used genome-wide CRISPR screening to identify mechanisms of drug resistance. We also validated the efficacy of the combination in vivo. METHODS

Cell line The cell lines were obtained from different sources: the Sanger Institute Cancer Cell Line Panel, Nicholas Hayward lab collection, Meenhard Herlyn lab collection and Daniel Peeper lab collection (see Table 1). The cell lines were grown in RPMI 1640, DMEM/F12 or DMEM media (Lonza, see Table 1) supplemented with 10% Fetal Bovine Serum (Gibco), Penycillin (final 100U/ml), Streptomycine (final 100U/ml) and L-glutamine (final 292μg/ml, all from Gibco in 100X mix). All cell lines were maintained at 37° and 5% CO2. All the cell lines were tested to exclude mycoplasma contamination.

Table 1

WM1963 Herlyn lab RPMI 1640 validation

WM3311 Herlyn lab RPMI 1640 validation

WM3211 Herlyn lab RPMI 1640 validation

WM3246 Herlyn lab RPMI 1640 validation

SK-MEL-5 Sanger DMEM/F12 validation

WM-115 Sanger DMEM/F12 validation

A2058 Sanger DMEM/F12 validation

LB373-MEL-D Sanger DMEM/F12 validation

MZ2-MEL Sanger DMEM/F12 validation

Hs-940-T Sanger DMEM/F12 validation

SK-MEL-31 Sanger DMEM/F12 validation

WM793B Sanger RPMI 1640 validation

RVH-421 Sanger RPMI 1640 validation

IST-MEL1 Sanger RPMI 1640 validation

WM278 Sanger RPMI 1640 validation

WM1552C Sanger RPMI 1640 validation

The nomenclature of PDX line (and derived protein and RNA extracts analysed herein) include a suffix (.X1 , .X2) that indicate the passage in vivo of the PDX and the .CL suffix that indicate the cell line growing in vitro, as previously described¹.

Analysis of mutation

Exome sequencing was performed for the 21 cell lines used for the screening and for the available matched germline. For C037 and the matched germline, whole genome sequencing was performed. Briefly, DNA libraries were prepared from genomic DNA, the exonic regions were captured with the Agilent SureSelect Target Enrichment System, 50 Mb Human All Exon kit or with lllumina's TruSeq Exome kit. Whole-genome libraries (for C037) were prepared using the standard lllumina library preparation protocol. Paired-end reads of between 70 and 100 bp were generated on the HiSeq 2000 lllumina platform. Whole-exome sequencing reads (or whole genome sequencing reads for C037 cell line and correspondent germline) were aligned to the reference genome GRCh37 using the Burrows- Wheeler Aligner software (version 0.7.5a-r406)². MuTect (v1.1.4)³, with default parameters, was used to identify somatic point mutations from read alignments by comparing cell lines with matched germlines. For the 5 cell lines without a normal germline reference (MeWo, Colo- 792, CHL-1 , M002.X1.CL, D10), we used the whole genome sequencing data from C037 germline as reference. The effects of mutations on protein sequences was predicted using the Variant Effect Predictor⁴ and gene models from EnsembI release 75⁵. Each mutation was also annotated with data from the ExAC database (ExAC allele frequency from version 0.3)⁴, and the COSMIC database (mutation ID and number of human tumor samples in COSMIC carrying that mutation, database version 71 )⁶. To reduce the number of false somatic mutations in the 5 cell lines without matched normal (germline), we removed: 1) single nucleotide variants reported by SAMtools mpileup against the human reference genome (GRCh37) in any of the 16 germline samples in our collection; 2) all the mutations with an ExAC allele frequency >0.5%.

We considered all the mutations in splice site or coding regions and used the Variant Effect Predictor annotation to define missense and loss of function (LOF) mutation according to Table 2: Table 2

These mutations are the set of somatic mutations referred to herein.

The NRAS^Q6 mutation in D38s was identified from the RNA Sequencing data (see below), validated by Sanger sequencing performed on PCR-amplified DNA collected from aliquots of the cell line at the same time (± 3 passages) of the drug screening. In order to have uniform counts of somatic mutations among samples processed with 2 different bait sets used for the exome capture and the cell line for which we had whole genome sequencing, we considered only the lesion within the overlap between Agilent SureSelect Target Enrichment System 50 Mb Human All Exon kit and lllumina's TruSeq Exome kit.

The mutations of the cell lines from Sanger Institute Cancer Cell Line Panel within the second collection of cell lines were previously described⁷. The BRAF and NRAS status for the cell lines from Herlyn's lab collection has been previously described

(https://wwvy.wistar.org/lab/meenhard-herlyn-dvm-dsc/page/melanoma-cey-lines-0).

According to the criteria defined by TCGA for human tumors⁸, we defined as BRAF/NRASwM type melanomas those cell lines that do not carry any mutation at these amino acid positions: BRAF^⁶⁰⁰, BRA^ , NRAS^G , NRAS? , NRAS^Q6 None of our cell lines carry any mutation in HRAS or KRAS. We then classified as NF1 mutant (NF1 m) those BRAF/NRAS wild type melanoma cell lines that carry any non-synonymous (including splice site) mutation in NF1 gene, and as triple wild type (TWT) all the others.

Mutation validation with RNA-seq

RNA-sequencing reads (see below) from melanoma cell lines were aligned to the reference genome GRCh37 using the STAR aligner (version 2.5.0)⁹. A 2-pass STAR alignment was performed, and BAM files from replicates were merged. PCR duplicates were flagged using Picard (version 1.135; http://broadinstitute.github.io/picard/) and base quality score recalibration (BQSR) from the Genome Analysis Toolkit (GATK; version 3.5)¹⁰ was performed prior to running the GATK HaplotypeCaller. Sites covered with a minimum of 20 reads were considered for comparison with mutations called from WES. We only considered missense mutations since LOF mutations would be likely associated to unstable mRNA thus leading to an underrepresentation over the wild type allele in the RNASeq reads. Most of the cell lines displayed a very high match between RNA-seq and exome sequencing called mutations (median 91.9%). C092 and D38s displayed a lower frequency of validated mutations that is likely due to low number of mutations they carried. Nonetheless, for those cell lines we confirmed a genotype match with germline and RNA-seq data (see below).

Verification of cell line identity

STR analysis confirmed the match of the cell line profile in our hands with the profile determined from the original repository. To further test the match of the cell line in use with the patient from which it was generated, the exome sequencing data of all the cell lines and of the available matched germlines were subjected to a genotype comparison which consisted of variant calling against the reference genome GRCh37 using SAMtools (version 1.3.1)¹¹ 'mpileup' and BCFtoools (version 1.3.1 ; http://samtools.github.io/bcftools/) 'call', followed by filtering of variants using BCFtools 'filter', and calculation of all pairwise sample discordant genotypes and discordance scores using BCFtools 'gtcheck'. This analysis compared all the single nucleotide polymorphism (SNP) of each samples (range from 30048 to 132579 in the samples analysed), and was run on the exome sequencing of the melanoma cell lines, of the patient's matched germline (when available) and also on the RNA-Seq data. For each cell line, the reciprocal best match (as determined by the lowest discordance) was found to be the paired germline, as expected. The concordance between exome sequencing and RNA-Seq SNPs was >87% for all samples (including C092 and D38s cell lines). Analysis of copy number variation

Genome wide copy number was determined using the Affymetrix Genome-Wide Human SNP Array 6.0. The data analysis was performed with PICNIC¹² enabling simultaneous identification of actual allelic copy number and genotype data. This genome wide analysis was used to determine the copy number information for each gene presented as: maximum and minimum copy number (of any genomic segment containing coding sequence of the gene); zygosity (scored Ό', 'L' or Ή' if any genomic segment is homozygously deleted, has loss of heterozygosity or the whole region is heterozygous, respectively) and disruption status (D, if the gene resides on more than one genomic segment). Gene expression analysis

The collection of 22 cell lines used for the drug screening were grown in biological triplicate (>3 independent passages among replicates) and collected when 60-80% confluence was reached (in a T75 or T150cm² flasks) and 24h after the last media change. Total RNA, including small RNA, was extracted with microRNAeasy mini kit (Qiagen). We prepared 2μg of RNA for each sample and spiked in with ERCCv92 Mix 1 (Ambion, Thermo Fisher) to measure the dynamic range of detection. Stranded RNA-Sequencing library were prepared with the standard lllumina cDNA protocol with a library fragment size between 200 and 300bp. Three multiplexed libraries were prepared each with 22 samples and containing a biological replicate for each cell line. For each sample we obtained ~ 55 x 10⁶ paired end reads of 100bp with HiSeq2000. The RNA sequencing reads were mapped against the human genome (GRCh37d5) using Tophat2¹³ (v2.0.10) and an annotation file containing ENSEMBL v75 with the following parameters (-library-type fr-firststrand -g 1 -G). Subsequently, read pairs were counted using htseq-count from HTSeq¹⁴, based on ENSEMBL v75 annotation, with parameters (-m intersection-nonempty -a 10 -i gene_id -s reverse). Using the counts obtained and the average transcript length per gene, we calculated the number of Fragments Per Kilobase per Million reads mapped (FPKM) per gene to assess expression. RNASeq data from cell-lines C022 and D35 were too low quality to reliably assess gene expression (low mapping quality and poor correlation among the biological replicates, all due to low RNA quality) hence, they were not considered for any further analyses.

Definition of the differentially expressed genes between sensitive and non-sensitive cell lines

We contrasted the expression of the 5 cell lines sensitive to nilotinib/trametinib combination (see below for definition) with the expression of the 6 non-sensitive cell lines (out of 7 non- sensitive cell lines since gene expression is not available for D35). The statistical approach took as input the RNA sequencing reads data from the biological triplicate of each cell line through the following steps. Firstly we considered all those genes that are expressed with FPKM>1 in >2 cell lines of the collection (n=20 cell lines with RNASeq data). We used Limma¹⁵ and Voom¹⁶ to identify differentially expressed genes. Voom was used to normalize the read counts for the library size, log-transform the read counts such that the distribution becomes Gaussian-like and estimate precision weights to account for variation in precision between observations¹⁶. Limma's duplicateCorrelation function was used to incorporate the information from replicates using a mixed modelling framework. Voom was used both before and after duplicateCorrelation, to normalize the input and to subsequently take into account the replicates in the normalization respectively. Finally, Limma was used to identify differentially expressed genes. We considered as differentially expressed those genes that had a FDR corrected P-value <0.05 and fold change >2 or <0.5. Interrogation of the genes differentially expressed in sensitive and non-sensitive cell line in human tumor transcriptome data

The expression of the genes differentially expressed between sensitive and non-sensitive cell lines was used to probe the transcriptome of melanoma tumors from TCGA and Leeds Melanoma Cohort (LMC)¹⁷.

We used gene expression data from the 5 sensitive and 6 non-sensitive cell lines (see below) to generate a list of differentially expressed genes, which we then applied to tumors of the 2 cohorts (TCGA and LMC) using the nearest centroid method^17,18. Briefly, we averaged the 320 genes of the synergy signature within each cell type class (sensitive and non-sensitive), creating a 'synergy' and a 'non-synergy' centroid vector, each gene having been standardised (mean 0 and variance 1) beforehand. To classify each tumor as synergy-like or non-synergy- like, its standardised expression values (mean 0 and variance 1) were correlated with each centroid. Then the tumor was assigned to the group showing the highest correlation, with at least a difference of 0.1 in Spearman correlation coefficients between the 2 groups. The tumor was deemed unclassified if the difference in correlation coefficients was lower than 0.1. A similar approach was used to classify tumors in one of the 4 molecular classes defined by Jonsson et al. signature (proliferative, pigmentation, high-immune and normal-like^18,19). For this analysis, a tumor was deemed classifiable if its Spearman correlation coefficient with one of the 4 classes was greater than 0.1 , with the highest correlation coefficient determining the Jonsson's class to which the sample was allocated. Quantitative RT-PCR for AXL in PDX samples

RNA of BRAF/NRASWT PDX was isolated using Trizol, according to manufacturers' protocol. cDNA was generated using the Maxima First Strand cDNA Synthesis Kit (Thermo) according to manufacturers' protocol. Real-time PCR was performed using the following primers:

HPRT-F:5'-CGGCTCCGTTATGGCG-3' (SEQ ID NO: 2);

HPRT-R: 5'- GGTCATAACCTGGTTCATCATCAC-3' (SEQ ID NO: 3);

AXL-F: 5'-GGTGGCTGTGAAGACGATGA-3' (SEQ ID NO: 4);

AXL-R: 5'- CTCAGATACTCCATGCCACT-3' (SEQ ID NO: 5);

using the SYBR-Hi ROX kit (Roche) according to manufacturers' protocol. The real-time PCR was run on the Step One Plus Real Time PCR System (Applied Biosystems). The AXL expression levels were normalized to the HPRT housekeeping control.

MicroRNA expression analysis

The same total RNA including the small RNA collected from each cell line in biological replicate (see Gene expression analysis) was used for microRNA sequencing. The samples libraries were prepared with the lllumina Small RNA library kit to generate fragments between 20 and 30bp size. For each sample we obtained ~ 9 x 10⁶ single end reads of 50bp with HiSeq2000. The sequencing reads were mapped by Chimera²⁰ and blasted against the microRNA precursors sequence obtained from miRBase version 21 (http://www.mirbase.org/). Counts were subsequently normalised using DESeq2²¹.

To define the cell line specific up or downregulated microRNAs, we then compared each cell line vs all the other cell lines within the collection and calculated statistical significance of the difference with Voom¹⁶. We considered as significantly up or downregulated those microRNAs with a Voom t-statistic value >10 or <-10, and with an absolute value of the log2 fold change >V2. If more than 50 microRNAs resulted, we selected only the top 50 microRNAs (by T value ranking). If fewer than 5 microRNAs resulted, we selected the top 5 regardless of the threshold criteria.

Definition of BRAF/NRAS WT melanoma drivers

In order to identify cancer driver specific for BRAF/NRAS WT melanoma, we analysed the previously published mutation calls from 74 BRAF/NRAS WT (i.e. NF1 mutant plus triple wild type) from the TCGA collection⁸ using the IntOGEn pipeline²² with the 3 algorithms Mutsig, OncodriveClust and OncodriveFM. We considered as mutation drivers those genes that have at least a significant signal (Q-value<0.05) from one of the 3 algorithms and that are known drivers in other tumor types. Given the difficulty to identify recurrently amplified and deleted genes from a small collection of samples such as the BRAF/NRAS WT melanomas, we considered as CNV melanoma drivers the genes that map within chromosomal regions previously defined as significantly amplified or deleted in a TCGA collection of melanoma (n= 333)⁸.

High throughput drug screening

We tested a library of 60 drugs targeting the main pathways deregulated in cancers. The range of the drug concentration was defined according to the activity of the drug in a large panel of cell lines⁷'²³. The 3 anchor drugs temozolomide, nilotinib and roscovitine were used at 2 different concentrations, 4 fold dilution one from the other. Cell lines were seeded in 384-well microplates at -15% confluency in culture medium with 10% FBS and Penicillin/Streptomycin. The optimal cell number for each cell line was determined to ensure that each was in growth phase at the end of the assay (-85% confluency). After overnight incubation cells were treated with 5 concentrations of each compound (4-fold dilution series, covering a 256-fold drug concentration range), using liquid handling robotics (Beckman Coulter), and then returned to the incubator for 6 days. After 6 days in drugs, the cells were then fixed in 4% formaldehyde for 30 minutes and then stained with 1 μΜ of the fluorescent nucleic acid stain Syto60 red fluorescent nucleic acid stain (Molecular Probes, Thermo Fisher) for 1 hour. Quantitation of fluorescent signal intensity is performed using a fluorescent plate reader at excitation and emission wavelengths of 630/695nm. All screening plates were subjected to stringent quality control measures and to assess the quality of our screening a Z-factor score comparing negative and positive control wells was calculated across all screening plates. Analysis of high-throughput drug data

We derived the Area Under the Curve (AUC) parameter from the 6-day cell line viability data to identify cell lines that are sensitive to a specific compound, with decreasing AUC associated with increasing sensitivity. The data input for the AUC calculation were the viability measured by Syto60 staining and normalized for vehicle treated control. Empirical values above 1 were capped to value of 1 , values below 0 were capped to value of 0. The AUCs were computed using a trapezoid integration below the 5 measured viability of the dose-response curve. For each cell line treated with a drug combination, we calculated the AUC for the library drug, for the anchor drug and for the drug combination.

In order to measure drug synergy, we derived the delta AUC. For two drugs, each at a given concentration, we used the Bliss independence model²⁴ to compute the expected viability of the cell line when exposed to the drug pair (product of the viability measured with the library drug alone and the viability measured with the anchor drug alone). This defines the expected dose-response curve on the 5 measured concentrations of the library drug used in combination with the anchor drug. Herein we called the predicted AUC of the combination defined by the Bliss model as predicted additivity, since it represents the cell viability that you would measure if the effect of the 2 drugs are added one to the other with additive effect and no synergy occurring. The delta AUC is defined as the difference between the AUC below the predicted additivity dose-response curve and the AUC below the observed dose-response curve as experimentally measured in the presence of the two drugs (briefly called AUC combination).

Low throughput drug assays

To validate the results obtained with the high-throughput drug screening, we screened the same drug combinations in representative cell lines (see below Criteria for the validation of synergistic combination). In these assays each experimental point was performed in technical triplicate and we used the same dose of anchor drug concentration and the same 256 fold range of library drug concentration of the high-throughput drug screening, but with 9 points with a 2-fold dilution series for the library drugs.

The cell lines were seeded in 180μΙ of complete media in 96 well microplates at a non- saturating density (confluency 60-90% after the 7 days of the assay in vehicle treated control; see Table 3). 24h after seeding, 20μΙ of the appropriate drug(s) dilution was added to the microplate which were then incubated in standard growth conditions for 6 days. At the experimental endpoint of 6 days the media was removed, and the cells were incubated for 20 minutes at room temperature (RT) with 75μΙ formaldehyde diluted at 4% in PBS. Then the fixative was removed, 3 washes in dhbO were performed, and staining with 75μΙ of 1 μΜ Syto 60 red fluorescent nucleic acid stain (Molecular Probes, Thermo Fisher) was performed for 1 h at RT protected from light. After 2 washes in dhbO, the MW96 plates were read by Biomek FX^P Liquid Handling Automation Workstation (Beckman Coulter) according to Syto60 manufacturer protocol.

The analysis of the AUCs and delta AUCs was performed as described above for the high- throughput drug screening.

Table 3

WM3702 2500

WM1963 3000

WM3311 7000

WM3211 9000

WM3246 6000

SK-MEL-5 3000

WM-115 3000

A2058 600

LB373-MEL-

D 3000

MZ2-MEL 4000

Hs-940-T 5000

SK-MEL-31 2000

WM793B 1000

RVH-421 2000

IST-MEL1 1000

WM278 1700

WM1552C 5000

Clonogenic assays

We selected representative cell lines among the ones that displayed the synergy between the drug combination of interest and that successfully grew clones when seeded at low density. We seeded in each well the same number of cells described in Table 3 but in 6 well microplates in 2ml of media and 24h after seeding we added 2ml of media containing the dilution of the drug(s). We tested one dose of the anchor drug (the same of the previous assays) and two representative doses of the library drug. After 10-15 days of drug treatment, when clones became evident, the cells were fixed with methanol for 1 h, then stained for 30 seconds with 0.5% crystal violet (Sigma-Aldrich) dissolved in 25% methanol, washed twice in water.

Definition of the delta AUC threshold for synergy

In the triage of the drug combinations for the low throughput validation assays, firstly we selected the five top drug combinations with an average delta AUC in the high-throughput screening that met the criteria of having a delta AUC synergy score >0.2 in three or more cell lines. Priority was given to the highest dose of anchor drug as this resulted in increased activity of the combination (lower average AUC combination). The threshold of delta AUC>0.2 was selected as it corresponds to the top 1.5% delta AUC of all the screened drug combinations. Then, in an effort to extend the pool of validated drug combinations, we tested the three combinations with delta AUC>0.2 in two cell lines only that displayed the highest activity (AUC<0.4, a value corresponding to the top 5% AUC combination in the high-throughput drug data). None of those 3 combinations was successfully validated in any of the cell lines, suggesting that decreasing the thresholds is unlikely to identify reliable hits.

Since the low throughput validation experiments are performed in technical triplicate with more library drug concentration points than the high-throughput viability curves (9 and 5 in low and high-throughput, respectively), we re-considered the synergy threshold based on the increased reliability of the assay. Biological replication of the assay (n=8) in C077 cell lines showed high reproducibility of the results, with a standard deviation of 0.0475267. Therefore we considered as synergistic those drug combinations that displayed a delta AUC>0.1 in the low throughput assays, a value that is above 2 fold the standard deviation of the assay. For a broad panel (n=21) of melanoma cell lines we tested nilotinib combined with 2 MEK inhibitors (trametinib and PD-0325901). We classified as sensitive to the drug combination those cell lines that displayed a delta AUC>0.1 for both nilotinib plus trametinib and nilotinib plus PD-0325901. We classified as non-sensitive those cell lines that displayed a delta AUC<0.1 for both nilotinib plus trametinib and nilotinib plus PD-0325901. The cell lines which displayed delta AUC>0.1 for one of the MEK inhibitors combined with nilotinib and a delta AUC<0.1 for the other MEK inhibitor were classified as "intermediate". The cell lines that displayed an AUC<0.3 for the anchor drug alone or the library drug alone were classified as not detected ("ND"), since the high activity of a single drug alone hamper the reliable detection synergy. The biological replicates of the assays confirmed the reliability of the sensitive/non- sensitive cell line classification.

For comparative gene expression analyses, we considered only the cell lines classified as sensitive or non-sensitive to the drug combination, excluding the ones with a intermediate or not detected drug response.

For the definition of synergy and for the association with mutation, CNV, gene and microRNA expression data, we used the data from the low throughput validation described herein, as these data were generated in parallel experiments with the same batches of drugs. All the synergies were tested at least in biological duplicate (range of biological replicates 2-8); at least 2 biological replicates were performed by 2 independent operators.

Definition of the cell lesions used for the association with the drug sensitivity data. To prioritize the mutation, copy number variations, gene and microRNA up and downregulation to be used for the association with the delta AUC score, we followed a multistep approach as detailed below. We used the delta AUC score for nilotinib plus MEK inhibitors, excluding the delta AUC of the cell line classified as "ND" (see above). We collected a list of high confidence cancer driver lesions defined in 6 previous works^25"30. To generate the list of driver lesions in melanoma driver genes, we collated the somatic LOF mutations in the 24 BRAF/NRAS WT melanoma drivers and only the somatic missense mutations that match by position and alternative amino acid the previously defined list of high confidence cancer driver lesions. The list of all lesions in melanoma drivers was compiled considering all the LOF and missense somatic mutations in any position within the 24 BRAF/NRAS WT melanoma drivers (see above). We then defined the list of genes affected by copy number variation considering the PICNIC output for the 39 genes in region significantly amplified/deleted in melanoma and the 24 BRAF/NRAS WT melanoma drivers and defined them as amplified (AMP) if the gene is in a segment with >5 copies or deleted (DEL) if the gene is in a segment with <2 copies (from PICNIC output). We defined the list of up and downregulated genes for these 24 BRAF/NRAS WT drivers and the 39 genes in region significantly amplified/deleted in melanoma by 1) averaging the biological triplicate per cell line; 2) removing the genes that have FPKM>1 in less than 3 cell lines (poorly expressed genes); 3) dividing the cell line specific FPKM expression value of each gene for the median of expression of that gene in the whole collection (20 cell lines analysed by RNASeq); 4) defining as upregulated (UP) those genes with a fold change over the median >4 and as downregulated those genes with a fold change over the median <0.25. We defined the up/downregulated microRNAs as described above. We then combined the two versions of the mutation data, copy number variation and up/downregulation data and summarized them by gene to generate the list of cell features to be used for association with drug synergy data. For both sets (one with driver lesion in driver genes, one with all lesion in driver genes) we considered the different type of lesions combined together (considering for instance a gene as altered if it has an amplification or a downregulation or a mutation, "ALL" below), as single input, or grouped according to putative functional impact (for instance, in "GAIN" a gene is considered as altered if it has a missense mutation, an amplification or an upregulation). The combination of the lesions that we tested are detailed below:

• Poorly expressed melanoma genes "DOWN"

• Overexpressed melanoma genes = "UP"

• Amplifications "AMP"

· Deletion alone "DEL"

• Loss of function mutation "LOF" • Missense mutations"MISSENSE"

• All the perturbation/lesion "ALL"

o ALL <- MISSENSE + LOF + AMP +DEL + UP + DOWN

• All non-silent mutations + all amplification and deletions (no expression) "ANY"

o ANY <- MISSENSE + LOF + AMP +DEL

• LOF + deletion "LOSS"

o LOSS <- LOF +DEL

• Missense mutations + amplifications "GOF"

o GOF <- MISSENSE + AMP

· Missense plus loss of function mutation plus deletion "MUTLOSS"

o MUTLOSS <- MISSENSE + LOSS

• Poorly expressed and loss of function mutation "MINUS"

o MINUS < DOWN + LOF

• Poorly expressed and loss of function and deletion "LESS"

o LESS <- DOWN +LOSS

• Poorly expressed and missense mutations "MINUSMUT"

o MINUSMUT <-DOWN +MISSENSE

• Overexpressed and missense mutations "PLUS"

o PLUS <- UP + MISSENSE

· Overexpressed and amplified "AMPUP"

o AMPUP <-UP +AMP

• Overexpressed, amplified and upregulated "GAIN"

o GAIN<-UP + MISSENSE + AMP For the identification of the statistical association with the drug synergy score, identical alteration profiles were merged: if gene 1 and gene 2 have the same alteration profile (e.g. are mutated in the same cell lines), we consider a single feature denominated as the concatenation gene1-gene2. Only genes with a lesion in >2 cell lines were considered for the analysis. For each cell lesion defined as detailed above, we compared the delta AUC score between the cell lines showing the feature or not using a t-test. The p-values obtained were corrected for multiple testing across all the features in the specific set (the 16 sets of lesions detailed above) tested with the Benjamini-Hochberg method. Statistical analyses were performed with R/Bioconductor³¹. Western blot To measure the level of P-ERK activity upon drug treatment, we seeded the cell lines at a density 2 fold higher than the one used for the low throughput validation assay (see Table 3) in a Petri dish of the proper area to contain 300,000 to 1 ,000,000 of cells. The day after seeding, the drugs at the indicated concentrations were added to the media. The cells were collected 6h after the drug treatment as it follows: we performed a wash with 4°C cold PBS, then cells were lysed in NP40 lysis buffer (Thermo Fisher Scientific) with Protease/Phosphatase Inhibitor Cocktail (Cell signalling) adding 200-1000μΙ of buffer to each petri dish (with a constant cell number/volume of buffer); the dish was incubated in ice for 30 minutes and tilted every 10' to allow an homogenous coverage of the surface; cell lysates were collected by scraping the cells and collecting the lysates into 1.5ml tubes, centrifuged at 13,000 rpm for 10 minutes at 4°C in benchtop centrifuges, and then protein supernatant was collected and frozen at -80C. For AXL and MITF analyses across different cell lines, protein lysates were quantified with Pierce BCA Protein Assay kit (Thermo Fisher Scientific) following the manufacturer protocol.

Protein was denaturated by adding 25% of NuPAGE LDS Sample Buffer (Thermo Fisher Scientific) and 5% of dithiothreitol 1 M (Sigma) and incubating 15 minutes at 75C. We loaded 5-10μg of protein on NuPAGE™ Novex™ 4-12% Bis-Tris Protein Gels, 1.5 mm, 15-well (Thermo Fisher Scientific) and performed electrophoresis at 120V in NuPAGE® MOPS SDS Running Buffer with NuPage Antioxidant (both from Thermo Fisher Scientific) in a Xcell Surelock elecrtoforesis cell. The protein were then transferred to Amersham Hybond N+ nylon membrane (GE Healthcare) by overnight blotting at 4C at 10V in XCell II blot machine (Lifetech) in NuPAGE Transfer Buffer with NuPage Antioxidant (both from Thermo Fisher Scientific). The membrane blocking was performed in 5% non-fat milk (Cell Signalling) or 5% BSA (Acros Organics) dissolved in Tris buffered saline with 0.25% of Tween 20 (TBS-Tween, Sigma-Aldrich, see Table 4).

We used the primary antibody described below in Table 4 at the indicated conditions. Table 4

total ERK #9102 Cell Signalling 1 :2000 BSA rabbit

P-ERK #4370 Cell Signalling 1 :6000 BSA rabbit beta tubulin

#5346 Cell Signalling 1 :3000 BSA rabbit vinculin V9131 Sigma 1 :1000 milk mouse

Santa Cruz

AXL for PDX sc-20741 Biotechnology 1 :1000 milk rabbit

TSC1 #4906 Cell Signalling 1 :2000 milk rabbit

TSC2 #4308 Cell Signalling 1 :2000 milk rabbit

CDKN1 B #2552 Cell Signalling 1 :1000 milk rabbit

Membranes were incubated overnight at 4°C with the primary antibody. After 3 washes in TBS-Tween, an incubation with anti-rabbit or anti-mouse IgG HRP-linked (1 :6000 and 1 :3000, #7074 and #7076, respectively, both from Cell Signalling) was performed at RT for 1 h. The membrane was washed 3 times for 5 minutes each in TBS-Tween and the signal was detected with Amersham ECL Select Western blotting detection reagent (GE Healthcare) using Image Quant Las4000 to acquire the pictures.

Immunoblotting for PDX samples was performed following the protocol previously described³². The signal on the western blots images were quantified using ImageJ (Rasband, W.S., ImageJ, U. S. National Institutes of Health, Bethesda, Maryland, USA, http://imagej.nih.gov/ij/, 1997-2016.). Briefly, each gel lane was outlined and the densitometry plotted. The peak of interest was then defined and quantified. Each value obtained was normalised to a loading control. Each experiment for the detection of P-ERK levels was performed in biological triplicate.

CRISPR/Cas9 screening

We performed a CRISPR/Cas9 screening as described³³ to identify mechanisms of resistance to the combination of nilotinib with MEK inhibitors. We used a previously described genome- wide library of synthetic guide RNAs (sgRNA) targeting the human genome (library V1 , containing 90,709 sgRNA targeting 18,010 human genes³⁴) for CHL-1 cell line, and an updated version of the same library (library V1.1) in C077 and MeWo cell lines. The lentiviral vector containing the library was titered on each cell line by BFP detection by FACS in order to precisely define the Multiplicity of Infection (MOI)³⁴. We generated stably expressing Cas9 cell lines, infecting them with a lentitival vector (LV) encoding for Cas9-2A-Blasticidin resistance³⁴ and selecting the cell line in blasticidin 15 μg/ml for 10 days. We then measured the activity of Cas9 with a reporter system delivered by a LV that express BFP, GFP and a sgRNA that target GFP³⁴. All the cell lines had a Cas9 activity >75%. For each of the 3 cell lines we infected 45-60 x 10⁶ cells in duplicate at MOI 0.3 (150-200X library representation in each replicate) to maintain 1 copy of sgRNA per cell. Two days after infection we added puromycin 2 μg/ml and selected the infected cells for 4 days. Then each of the library-infected population was expanded for 2 weeks to allow the clearance of the protein after the sgRNA/Cas9 mediated knockout. Two weeks after the infection 45-60 x 10⁶ cells per each replicate were seeded in drug regimens: trametinib (100nM for CHL-1 ; 12nM for C077 and MeWo), nilotinib (2μΜ for all the lines) plus trametinib (100nM for CHL-1 ; 12nM for C077 and MeWo) and matched DMSO vehicle. Given the limited activity of nilotinib alone which would have resulted in the absence of selective pressure, we avoided to include an experimental arm of cells treated with nilotinib only. An aliquot of 60 x 10⁶ was collected at the start of drug selection as a reference population (PRE population, whose resulting sgRNA counts were used to calculate the ROC curve, see below). Media with drugs were replaced 3 times a week; cells were split when confluent maintaining in culture >45-60 x 10⁶ cells per population. After 18 days of drug selection, >45-60 x 10⁶ cells were collected from each cell line population. DNA was extracted from > 50 x 10⁶ cells pellets with Qiagen Blood & Cell Culture DNA Maxi Kit (Qiagen). We then amplified and sequenced sgRNA as previously described³⁴. Briefly, a first PCR was performed on ~10⁷ cells (72 μg of DNA, 100X library representation) to amplify the sgRNA with previously described oligos³⁴ in 36 PCR reactions (with 2 μg of DNA each) 50μΙ each with Q5 Hot Start High-Fidelity 2* Master Mix. PCR products were then purified by QIAquick PCR Purification Kit and 200pg of the purified product was nested-amplified to add sample indexes and adapters with KAPA HiFi HotStart ReadyMix. The 2^nd PCR products were purified by SPRIselect reagent kit (Beckman Coulter), then pooled and high-throughput sequenced with HiSeq (lllumina) to obtain ~ 50 x 10⁶ reads per sample, as previously described³⁴.

The sequencing reads were then mapped to the sgRNA library, and the sgRNA counts were used as input for MAGeCK analysis³⁵. To measure the extent of reproducibility of the screening, we compared the sgRNA normalized counts between the replicates of infection and selection before the drug administration (PRE population) and found a high level of correlation (R = 0.74-0.93). As a quality control we estimated the ability of each CRISPR/Cas9 screen to discriminate between genes belonging to one between two a priori known sets of essential and non-essential genes³⁶, E and N respectively. To do this we aggregated sgRNA depletion p-values through MAGeCK, yielding a gene level summary of essentiality. The genes were then sorted according to their gene-level depletion p-values. At each gene rank position in the sorted list we compiled the true positive rate (fraction of genes belonging to E) and false positive rate (fraction of genes belonging to N), and so we created a receiver operating characteristic (ROC) plotting Sensitivity vs. (1 -Specificity) at each rank. The area under the ROC curve for the 3 cell lines was >0.9, indicative of high sensitivity and specificity of the screening. Finally, we found that none of the significantly enriched genes with FDR<0.1 in both replicates of each cell lines (i.e. what we defined as hits) was a poorly expressed gene (FPKM<1), further indicative of the specificity of the screening.

Each replicate of infection and selection (with trametinib or drug combination) was compared with the sister DMSO treated control. MAGeCK was used to evaluate the significance of each sgRNA enrichment or depletion, and to identify genes whose sgRNA targeting pool was significantly enriched or depleted compared to the control. We considered as significant hits those genes with a MAGeCK FDR corrected p-value <0.1 in both the replicates of infection and selection per each cell line. We then considered as resistance genes those hits that were found in 2 or more cell lines.

Enrichment and protein-protein network analysis.

The Enrichment analysis of the list of genes that are upregulated or downregulated in the gene expression signature of synergy, as well as the list of gene representing hits of the Cas9 screening, were performed by MsigDB (http://software.broadinstitute.org/gsea/msigdb/annotate.jsp) considering Canonical pathways and Hallmark gene sets and the top 100 pathways with FDR<0.05. The database was accessed in December 2016.

The Network of protein-protein interactions were defined using STRING³⁷ (http://string- db.org). We used the function to interrogate multiple proteins from Homo Sapiens gene symbols and downloaded the picture of the protein network obtained in the protein network display using the options "molecular action" and "disable structure previews inside network bubbles". We also used the output of the Gene Ontology enrichment analysis from the "analyses" sheet in STRING and integrated with the MsigDB output described above to display the top enriched protein complex or pathways in Figure 4c. The STRING database was accessed in January 2017.

Interrogation of the status of the drug resistance genes identified from CRISPR/Cas9 screening in melanoma.

We interrogated the cBioportal database (http : //www. cb i o o rta L org/) to investigate the status of the 18 nilotinib/trametinib resistance genes found as significant hits of the CRISPR/Cas9 screening in 2 or more cell lines. Four datasets of skin melanoma were available: Broad Cell 2012 (121 samples) ; Broad/Dana Faber, Nature 2012 (25 samples); TCGA, provisional (287 samples); Yale, Nat Genet 2012 (91 samples). CCDC101 gene symbol was present in the database with the alternative symbol SGF29. We reported the frequency of samples with an alteration in one of those genes in the four datasets, and provided detailed mutation type for the 2 larger dataset (TCGA and Yale). Only the TCGA dataset included copy number variation data. The cBioportal database was interrogated in March 2017.

Interrogation of previously published screen in melanoma drug resistance.

We interrogated GenomeCRISPR website (http://genomecrispr.dkfz.de/) for Cas9 screening performed in melanoma to induce drug resistance. We could find 2 studies that performed genome-wide Cas9 screenings in A375 (BRAF^^600E-mutant) melanoma cell line to find vermurafenib (BRAF inhibitor) or selumetinib (MEK inhibitor) resistance^38,39. We compiled a list of vemurafenib resistance genes considering the hits identified by Shalem et a/³⁸ that were found as in the top 100 ranking genes by RIGER score in both the replicates; to this list we added 2 new hits found by Li et a/.³⁵ by a re-analysis with MAGeCK of the results from the same screening. We then included the list of vemurafenib resistance genes identified by Doench et a/³⁹ with the GeckoLV2 library that have a FDR corrected P-value from STARS algorithm<0.1. We compiled a list of selumetinib resistance genes with the genes identified by Doench et a/³⁹ with the GeckoLV2 library that have a FDR corrected P-value from STARS algorithmO.1.

Generation of TSC1^KO, TSC2^KO , CDKN1B^KO clones for CHL-1 and C077 cell lines

We designed sgRNA sequences (independent from the ones in the high-throughput sgRNA library) for the knockout of TSC1, TSC2 and CDKN1B in CHL-1 and C077, to hit essential functional domain in the respective proteins: exon 4 for TSC1, exon 17 for TSC2 and the central part of exon 1 for CDKN1B. We used Genespy (http://genespy.genescripts.org/) to select the sequence of the target genomic region and then the link to the CRISPRdirect tool (http://crispr.dbcls.jp/) to select sgRNAs for each gene that had no other 20-mer target in the human genome. We then checked the sequence of selected sgRNA by CasDesigner (http://www.rgenome.net/cas-designer/) for off-target hits and then selected 2-3 sgRNA sequences per gene that have the lowest number of off-target hits. We then added the sequence for complementarity with the tracrRNA (GUUUUAGAGCUAUGCUGUUUUG (SEQ ID NO: 6)) to each of the 20bp sgRNA and ordered the synthetic RNAs from Integrated DNA Technology (see Table 5). Table 5

We applied the standard protocol from IDT for "CRISPR/Cas9 editing: transfection of synthetic RNA oligos" (https://www.idtdna.com/pages/docs/default-source/catalog-product- documentation/crispr-2-part-rna-transfection.pdf?sfyrsn^:=11). Briefly, we prepared an equimolar mix of sgRNA and tracRNA in PBS, and annealed them by heating at 95°C for 5 minutes and then letting the solution reach room temperature. We prepared a pool of 3 μΜ of each 2-3 sgRNAs per gene with RNAiMAX (Thermo Fisher) (0.75μΙ in total 50μΙ of sgRNAs solution) and incubated for 20 minutes at room temperature. Then we performed a reverse transfection in 24-wells microplates by adding 100,000 Cas9 expressing CHL-1 or C077 cells to each well containing the transfection complex. The media were changed 24h after transfection. The transduced cells were expanded for one week, then seeded at low density in 10cm dishes and cultured for 2-4 weeks to isolate single clones. The clones were then individually picked (-100 clones per gene per cell line) and expanded in 96-well microplates. Sister plates were viably frozen in media with 10% DMSO and 20% FBS. DNA was extracted from confluent 96-well microplates by lysing the cells at 55°C for 4h in Yolk Sac Lysis Buffer prepared with 50mM KCI, 10mM TrisHCI pH8.3, 2mM MgCI₂, 0.45% NP40, 0.45% Tween 20 and 1 ^g/ml proteinase K (Sigma). Proteinase K was then inactivated by heating the DNA to 95°C for 10 minutes. The DNA was diluted 1 : 10 in TrisHCI pH 8 and 2μΙ were used for amplification with the oligos indicated in the table and PCR Master Mix 2X (Sigma). The PCR products were submitted in 96 wells-microplates for Sanger sequencing (Eurofins) from both sides of the amplicon using the same oligos used for the PCR. The resulting sequences were analysed by aligning to the reference sequence of the gene and the homozygous knockout were confirmed by further analysis with Tide (https://tide.nki.nl/). We sequenced -50-100 clones per gene per cell line and identified 1-5 clones with homozygous deletion (reads from both sides showing the deletion). The clones were then expanded and Western blot was performed on protein extracts to confirm the knockout.

Competitive assay to test drug sensitivity in of TSC1^KO, TSC2^KO , CDKN1B^KO clones for CHL-1 and C077 cell lines

The above-described KO clones for CHL-1 and C077 cell lines were then infected with a lentiviral vector expressing GFP under the controls of the ubiquitous phosphoglycerate kinase enhancer/promoter sequence (pCCLSIN.cPPT.hPGK.EGFP.wPRE⁴⁰). Two weeks after the infection the clones were sorted using a BD FACSAria II (BD Biosciences), to select a population 100% positive that expressed high and homogenous level of GFP. The GFP⁺ cells were then mixed at 5% with a population of parental WT Cas9 expressing cells that were GFP negative. The mixed population were prepared in biological triplicate per each clone. The day after seeding, the cells mixed population were drugged with DMSO, or trametinib 100nM plus nilotinib 2μΜ combination. The drug selection was performed for 2 weeks and cells were split and FACS analysed to determine the % of GFP every 3-7 days (BD LSRII, using application settings to standardise the instrument). The % of GFP⁺ cells in the population was determined and expressed relative to the untreated sample at the same timepoint. Determination of cell cycle by flow cytometry

Cells were seeded 24h prior to the treatment with vehicle or the combination of trametinib 100nM and nilotinib 2μΜ for 24 hours. To stain cells in S phase, EdU (Life Technologies) was added at a final concentration of 10 μΜ. Two hours after, the cells were harvested and washed with PBS prior to viability staining (LIVE/DEAD fixable red, Life Technologies according to manufacturers instructions). Cells were then fixed and permeabilised prior to Click-it detection of EdU using Alexa Fluor 647 according to the manufacturer instructions (Click-It plus EdU, Life Technologies). DNA was labelled with 2 μg/ml Hoechst 33452 (Life Technologies) and cells washed prior to acquisition on a BD LSRFortessa (BD Biosciences). Doublets were excluded on the basis of pulse parameters in the Hoechst channel (height versus width), followed by dead cell exclusion and identical gates applied to identify G0/G1 (EdU negative, 2N), S (EdU positive between 2N and 4N) and G2/M (EdU negative, 4N) phases of a cell type. We measured the reduction of the cell cycle mediated by the drug combination by subtracting the % of the S phase of the drugged cells to the percentage of the S phase in the undrugged cells and normalized it for the percentage of S phase in the undrugged cell. All analysis was performed in a blinded manner apart from knowledge of the matching sample pairs for setting the same gates for cell cycle.

In vivo experiments

Animal experiments were approved by the animal experimental committee of the Netherlands Cancer institute and performed according to Dutch law. PDX were generated as described⁴¹. We injected 70,000 (M003.X2) - 300,000 (MeWo) cells in both flanks of NOD.Cg-Pr/ccfc³^ //2rg^tm1Wjl/SzJ (NSG) mice. After the tumors reached ~50-90mm³ volume, mice were randomized in 4 treatment groups (n=10 tumors/group) that were administered by oral gavage: a) vehicle (0.5% hydroxypropylmethyl cellulose (HPMC, Sigma) aqueous solution containing 0.05% Tween 80); b) nilotinib 75mg/kg/day (administered 37.5mg/kg twice daily), c) trametinib 0.1-0.3 mg/kg/day for MeWo and M003.X2, respectively; d) nilotinib plus trametinib combination. Mouse weight was monitored weekly; tumor size was measured by caliper 3 times per week. Mice were euthanized either when the tumor volume reached 1000mm³ or when the weight loss of the mice was more than 30%, or at the experimental endpoint.

Statistical analyses Graphs and statistics were generated using the GraphPad Prism software. The statistical test applied is indicated in the respective Figure legend. Briefly, when two groups were analysed, the P-value was calculated by unpaired Student's t-test; when three or more groups were analysed, the P-value was calculated by one way Anova and Tukey's multiple comparisons test. For the tumor growth curve in vivo, the group treated with the drug combination was compared to each of the other 3 groups by unpaired Student's t-test considering the tumor volumes measured at the last available experimental timepoint for each group. A two tailed Fisher's exact test was used to compare the number of tumors classified as synergy like and non-synergy like between one of the 4 Jonsson's expression classes and the remaining 3 classes.

The significance of gene enrichment/depletion in the CRISPR/Cas9 genome-wide library experiment was calculated with MAGeCK³⁵. References for Methods Section Only

1. Kemper, K., et al. Intra- and inter-tumor heterogeneity in a vemurafenib-resistant melanoma patient and derived xenografts. EMBO molecular medicine 7, 1 104-1118

(2015) .

2. Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25, 1754-1760 (2009).

3. Cibulskis, K., et al. Sensitive detection of somatic point mutations in impure and heterogeneous cancer samples. Nature biotechnology 31 , 213-219 (2013).

4. McLaren, W., et al. The Ensembl Variant Effect Predictor. Genome biology 17, 122

(2016) .

5. Aken, B.L, et al. Ensembl 2017. Nucleic acids research 45, D635-D642 (2017).

6. Forbes, S.A., et al. COSMIC: exploring the world's knowledge of somatic mutations in human cancer. Nucleic acids research 43, D805-811 (2015).

7. lorio, F., et al. A Landscape of Pharmacogenomic Interactions in Cancer. Cell 166,

740-754 (2016).

8. Cancer Genome Atlas, N. Genomic Classification of Cutaneous Melanoma. Cell 161 ,

1681-1696 (2015).

9. Dobin, A., et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 15-21 (2013).

10. McKenna, A., et al. The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data. Genome research 20, 1297-1303

(2010). Li, H., et al. The Sequence Alignment/Map format and SAMtools. Bioinformatics 25, 2078-2079 (2009).

Greenman, CD., et al. PICNIC: an algorithm to predict absolute allelic copy number variation with microarray cancer data. Biostatistics 11 , 164-175 (2010).

Kim, D., et al. TopHat2: accurate alignment of transcriptomes in the presence of insertions, deletions and gene fusions. Genome biology 14, R36 (2013).

Anders, S., Pyl, P.T. & Huber, W. HTSeq-a Python framework to work with high- throughput sequencing data. Bioinformatics 31 , 166-169 (2015).

Ritchie, M.E., et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic acids research 43, e47 (2015).

Law, C.W., Chen, Y., Shi, W. & Smyth, G.K. voom: Precision weights unlock linear model analysis tools for RNA-seq read counts. Genome biology Λ5, R29 (2014). Nsengimana, J., et al. Independent replication of a melanoma subtype gene signature and evaluation of its prognostic value and biological correlates in a population cohort. Oncotarget 6, 1 1683-11693 (2015).

Harbst, K., et al. Molecular profiling reveals low- and high-grade forms of primary melanoma. Clinical cancer research : an official journal of the American Association for Cancer Research 18, 4026-4036 (2012).

Jonsson, G., et al. Gene expression profiling-based identification of molecular subtypes in stage IV melanomas with different clinical outcome. Clinical cancer research : an official journal of the American Association for Cancer Research 16, 3356-3367 (2010).

Vitsios, D.M. & Enright, A.J. Chimira: analysis of small RNA sequencing data and microRNA modifications. Bioinformatics 31 , 3365-3367 (2015).

Love, M.I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome biology 15, 550 (2014).

Rubio-Perez, C, et al. In silico prescription of anticancer drugs to cohorts of 28 tumor types reveals targeting opportunities. Cancer cell 27, 382-396 (2015).

Garnett, M.J., et al. Systematic identification of genomic markers of drug sensitivity in cancer cells. Nature 483, 570-575 (2012).

Bliss, C.I. The calculation of microbial assays. Bacteriological reviews 20, 243-258 (1956).

Nik-Zainal, S., et al. Landscape of somatic mutations in 560 breast cancer whole- genome sequences. Nature 534, 47-54 (2016).

Kandoth, C, et al. Mutational landscape and significance across 12 major cancer types. Nature 502, 333-339 (2013). 27. Tarpey, P.S., et al. Frequent mutation of the major cartilage collagen gene COL2A1 in chondrosarcoma. Nature genetics 45, 923-926 (2013).

28. Behjati, S., et al. Distinct H3F3A and H3F3B driver mutations define chondroblastoma and giant cell tumor of bone. Nature genetics 45, 1479-1482 (2013).

29. Stephens, P. J., et al. The landscape of cancer genes and mutational processes in breast cancer. Nature 486, 400-404 (2012).

30. Behjati, S., et al. Recurrent PTPRB and PLCG1 mutations in angiosarcoma. Nature genetics 46, 376-379 (2014).

31. Gentleman, R.C., et al. Bioconductor: open software development for computational biology and bioinformatics. Genome biology 5, R80 (2004).

32. Possik, P.A., et al. Parallel in vivo and in vitro melanoma RNAi dropout screens reveal synthetic lethality between hypoxia and DNA damage response inhibition. Cell reports 9, 1375-1386 (2014).

33. Koike-Yusa, H., Li, Y., Tan, E.P., Velasco-Herrera Mdel, C. & Yusa, K. Genome-wide recessive genetic screening in mammalian cells with a lentiviral CRISPR-guide RNA library. Nature biotechnology 32, 267-273 (2014).

34. Tzelepis, K., et al. A CRISPR Dropout Screen Identifies Genetic Vulnerabilities and Therapeutic Targets in Acute Myeloid Leukemia. Cell reports 17, 1193-1205 (2016).

35. Li, W., et al. MAGeCK enables robust identification of essential genes from genome- scale CRISPR/Cas9 knockout screens. Genome biology 15, 554 (2014).

36. Hart, T., etal. High-Resolution CRISPR Screens Reveal Fitness Genes and Genotype- Specific Cancer Liabilities. Cell 163, 1515-1526 (2015).

37. Szklarczyk, D., etal. STRING v10: protein-protein interaction networks, integrated over the tree of life. Nucleic acids research 43, D447-452 (2015).

38. Shalem, O., et al. Genome-scale CRISPR-Cas9 knockout screening in human cells.

Science 343, 84-87 (2014).

39. Doench, J.G., et al. Optimized sgRNA design to maximize activity and minimize off- target effects of CRISPR-Cas9. Nature biotechnology 34, 184-191 (2016).

40. Montini, E., etal. Hematopoietic stem cell gene transfer in a tumor-prone mouse model uncovers low genotoxicity of lentiviral vector integration. Nature biotechnology 24, 687-

696 (2006).

41. Kemper, K., et al. BRAF(V600E) Kinase Domain Duplication Identified in Therapy- Refractory Melanoma Patient-Derived Xenografts. Cell reports 16, 263-277 (2016). RESULTS

Genomic and transchptomic analysis of BRAFINRAS wildtype melanoma cell lines We assembled a collection of 22 melanoma cell lines, including 20 BRAF/NRAS WT, one

cell line for screening. We catalogued somatic single nucleotide variation (SNV) in these lines by exome sequencing, copy number variation by use of SNP6 arrays, and gene and microRNA expression by RNA sequencing (outline in Figure 1 a). In agreement with the analysis of BRAF/NRAS WT human tumors from The Cancer Genome Atlas (TCGA) collection⁷, the BRAF/NRAS WT melanoma cell lines had a high mutational load (median 59.01 SNV/Mb, range 1.34-512.96) dominated by C>T mutations at dipyrimidines. The NF1 m cell lines displayed a significantly higher mutation frequency than the TWT cell lines (P<0.0001 ; One way Anova and Tukey's multiple comparison test), recapitulating the pattern described in tumors from the TCGA and Yale Melanoma Genome Projects⁶'⁷. To identify putative melanoma driver genes we ran IntOGen¹² using SNV data from 74 BRAF/NRAS WT tumors⁷ and found 24 statistically significant mutation driver genes (see Methods). Similarly, we collated melanoma drivers in regions defined as recurrently amplified or deleted in 333 melanomas from the TCGA collection⁷ spanning all cutaneous melanoma subtypes. All 24 BRAF/NRAS WT melanoma mutation drivers were mutated in at least one cell line, and 32 out of 39 driver genes in amplified or deleted regions were captured by genomic alterations in at least one cell line in our collection. The mutation frequency of the 24 mutation drivers in our BRAF/NRAS WT cell line collection correlated with the frequency found in BRAF/NRAS WT tumors (Pearson correlation P<0.0001 , R² = 0.6198; Figure 1 b). Collectively, these data show that our cell line collection is representative of the driver lesions found in BRAF/NRAS WT melanoma.

Nilotinib synergizes with MEK inhibitors in BRAFINRAS wUd type melanoma cell lines

Our collection of cell lines was used to perform a high-throughput combination drug screen with 60 library drugs, and three anchor drugs: temozolomide (alkylating agent), nilotinib (tyrosine kinase inhibitor) and roscovitine (broad CDK inhibitor). These anchors were selected because of their broad, yet distinct, modes of action thus allowing us to cover a wide biological space. Anchor drugs were tested at two concentrations and combined with each of the 60 library drugs tested at five concentrations over a 256-fold concentration range. Cellular viability was measured 6 days after drug treatment and normalized against DMSO-treated controls. Overall, we tested 180 drug combinations and generated 18,810 survival curves (3 curves per cell line per combination). Survival curves were analysed using the Area Under the Curve method as described previously¹⁰. The viability of cell lines treated with the single library drug alone was generally higher than cell lines treated with drug combinations (average AUC 0.8684 and 0.7925 for library drugs and drug combinations, respectively; P<0.0001 by unpaired Student's t-test). To prioritize the most effective drug combinations, we measured drug synergy as the difference between the AUC of the predicted additive effect¹³ and the AUC of the drug combination (delta AUC) (see Methods). With a threshold of delta AUC>0.2 (selected to represent the top 1.5% of 6270 delta AUCs, see Methods), we identified 94 occurrences of synergy from 53 drug combinations (Figure 1 c). Most combinations showed synergy in a single cell line; 6 combinations displayed synergy in 3 or more cell lines, 3.3% (6/180) of all combinations tested (Figure 1 d). We triaged these combinations as follows. Firstly, we assayed the synergy observed in the high-throughput screen in the three cell lines with the highest delta AUC using a low-throughput viability assay (see Methods). We focused on the 5 drug combinations with the highest average delta AUC where synergy was observed in >2 cell lines. Dose response curves were performed in triplicates using the same 256-fold doses range used in the high-throughput screen. In this way we confirmed synergy between temozolomide and olaparib (PARP inhibitor) and between nilotinib and PD-0325901 (MEK inhibitor) in three cell lines (Figure 2a). Testing of the three synergistic combinations just below the defined threshold (delta AUC >0.2 in 2 cell lines only, see Methods) did not replicate the results of the high-throughput screen. We next repeated the low-throughput viability assays on 21 of the previously screened cell lines (including 19 BRAF/NRAS WT, one SRAF^^-mutant and one A/ ¾AS^{Q61 R}-mutant cell lines) using the drug combinations temozolomide with olaparib, nilotinib with PD-0325901 and nilotinib with trametinib (a second MEK inhibitor in clinical use). The synergies between temozolomide and olaparib and between nilotinib and both MEK inhibitors were confirmed in 12 and 4 out of 19 BRAF/NRAS WT cell lines, respectively (Figure 2a, see Methods). We further confirmed the potency of these drug combinations by performing clonogenic assays (Figure 2b). Intriguingly synergy was also observed in an NRAS mutant line analysed in parallel.

Given the limited activity of alkylating agents combined with PARP inhibitors in clinical trials^14,15, we focussed on the nilotinib/trametinib combination for testing in a second independent collection of 19 melanoma cell lines, including 10 B AF^-mutant, 4 A/RAS⁰⁶¹- mutant and 5 BRAF/NRAS WT lines. We observed synergy (delta AUC<0.1 , see Methods for synergy threshold) in 10 out of 19 cell lines (Fig. 2c). Synergies were confirmed in clonogenic assays. Overall, we tested the nilotinib/trametinib combination in two collections of melanoma cell lines and found that it is synergistic in 17 out of 40 lines (42.5% Fig. 2d, top panel), including 6/24 BRAF/NRAS WT cell lines. In 26 out of 40 (65%) melanoma cell lines we observed high activity of the drug combination (AUC <0.4, see Methods) as a result of synergy, additivity and single agent activity (Figure 2d, bottom panel). AXL expression is associated with synergy between nilotinib and MEK inhibitors in BRAF/NRAS WT melanoma To identify markers that may predict synergy between nilotinib and MEK inhibitors, we looked for an association between the drug synergy score (delta AUC), coding mutations, copy number alterations and/or gene/microRNA expression (see Methods). To reduce multiple testing, we only considered lesions that were previously characterized as cancer drivers following an approach described previously¹⁰. We classified each lesion as a gain-of-function or loss-of-function alteration partitioning them into functional groups (see Methods). Following this approach, we failed to identify any statistically significant gene/drug associations. We then extended our analysis to all lesions in melanoma drivers, but again no associations were found. Analysis of differentially expressed microRNAs (see Methods) also failed to identify significant associations.

We then looked for pathways differentially expressed between cell lines that were sensitive or non-sensitive to the nilotinib/trametinib combination (see Methods for cell line classification into sensitive and non-sensitive). The sensitive cell lines displayed higher levels of cell cycle genes and lower levels of genes associated with pigmentation (see Methods). This gene expression pattern was observed in -30% of tumors from two melanoma cohorts and was more frequent in tumors classified as "Proliferative" by the Jonsson's gene expression classifier¹⁶. This approach did not highlight any specific pathway associated with nilotinib/trametinib synergy. Therefore, we looked for genes expressed (FPKM>1 by RNA Sequencing) exclusively in the sensitive cell lines and found 4 transcripts, among which AXL displayed the highest differential expression. Since AXL is involved in resistance to targeted therapies in melanoma¹⁷'¹⁸, we further investigated its role as a putative marker of response to nilotinib/trametinib treatment. Firstly, we confirmed the previously described inverse correlation between AXL and MITF RNA expression in our BRAF/NRAS WT cell lines (P= 0.0013, R squared = 0.4842, by Pearson correlation). While MITF expression was not significantly different between sensitive and non-sensitive cell lines, AXL expression levels were significantly higher in sensitive cell lines (P=0.01 17 by unpaired Student's t-test), as expected. We next measured protein expression in sensitive cell lines and representative non- sensitive cell lines revealing that all BRAF/NRAS WT sensitive cell lines expressed AXL, while it was undetectable in non-sensitive lines (Figure 3a, top panel). Extending the analysis to the 5 BRAF/NRAS WT non-sensitive cell lines of the second collection we showed that just 1/5 expressed AXL protein (Figure 3a, middle panel). Overall, AXL+ cell lines are significantly enriched for the occurrence of synergy between nilotinib and trametinib (Figure 3a, bottom panel). Remarkably, although AXL expressing cell lines displayed higher sensitivity/synergy for the nilotinib/trametinib combination, they showed higher resistance to MEK inhibitors alone, in agreement with previous studies^17,18. Notably, we did not observe a clear association between AXL expression and synergy in BRAF^⁶⁰⁰ or A/RAS^-mutant cell lines. The nilotinib/trametinib combination suppresses the MAPK pathway in sensitive cell lines

Given the prominent role of the MAPK pathway in melanoma proliferation and in melanomagenesis, we investigated its activation via phosphorylated ERK (p-ERK) in sensitive and non-sensitive cell lines upon treatment. As expected, trametinib reduced p-ERK in most lines (Figure 3b-c). Nilotinib increased p-ERK levels in some cell lines, a phenomenon previously explained by its activity as a mild RAF inhibitor which paradoxically activates ERK¹⁹ (Figure 3b-c). Upon treatment with the drug combination, the sensitive cell lines displayed a reduction of p-ERK that was significantly more pronounced when compared to non-sensitive cell lines, both when compared to vehicle treated controls or trametinib alone (P=0.0154 and P=0.0069, respectively; unpaired Student's t-test) (Figure 3d). Total ERK did not change with treatment. Resistance to nilotinib/trametinib occurs via regulators of MAPK signalling

We next performed CRISPR/Cas9²⁰ screens to identify genes conferring resistance to the nilotinib/trametinib combination. To perform these screens we generated three Cas9 expressing cell lines (CHL-1 , C077 and MeWo, selected because of their sensitivity to the combination) and transduced them with a genome-wide sgRNA library²⁰. Cells were cultured for 18 days in trametinib, nilotinib/trametinib, or DMSO vehicle (see Methods). We observed only limited overlap of genes whose loss conferred resistance to the nilotinib/trametinib combination among the three cell lines (Figure 4a, see Methods). We also found limited overlap among genes conferring resistance to trametinib. This suggests that many different genes are potentially operative in mediating resistance.

Given the heterogeneity of resistance mechanisms observed we focussed on the genes that conferred drug resistance in at least two of the three cell lines. In line with the converging inhibitory activity on p-ERK, we observe a large overlap between nilotinib/trametinib resistance genes and trametinib-only resistance genes (Figure 4b). Accordingly, 7/18 nilotinib/trametinib resistance genes have previously been identified as vemurafenib (BRAF inhibitor) resistance genes, and 9/18 genes have previously been identified as selumetinib (MEK inhibitor) resistance genes^{21 ,22} (Figure 4c). Notably, fewer genes (18 vs 29) appeared to confer resistance to the combination compared to trametinib alone (Figure 4b), suggesting that the combination may overcome some mechanisms of resistance observed with trametinib. Interestingly, we observed that many combination resistance genes interact (P=1.57 10^"9 by STRING²³) and that these genes are significantly enriched for members of the SAGA-type complex, estrogen receptor beta network and for chromatin regulators (Figure 4c by STRING²³).

Divergent role of the TSC complex in resistance to nilotinib/trametinib

To further test representative hits from the CRISPR/Cas9 screen, we performed experiments to validate three of the top drug resistance genes in CHL-1 , all of which are associated with the tuberous sclerosis complex (TSC) (Figure 4d). Paradoxically, rather than conferring resistance, in the C077 cell line these genes appeared to sensitize to the combination. Using CRISPR/Cas9, we generated KO clones for TSC1, TSC2, CDKN1B in CHL-1 and C077 (see Methods) and marked them with GFP by lentiviral vector infection. In a competitive growth assay with parental WT unmarked cell lines, the CHL-1 TSC1^KO, TSC2^KO, CDKN1BI^KO clones significantly expanded over time upon treatment with the drug combination (vs untreated control, P-value <0.0001 by unpaired Student's t-test, Figure 4e). Cell cycle analysis showed that CHL-1 KO clones were partially protected from the reduction of cell cycle induced by the nilotinib/trametinib combination compared to WT controls (-74% and -46% average cell cycle reduction in WT and KO clones, respectively, P = 0.0315 by unpaired Student's t-test), suggesting that TSC ablation partially abolished the effect of the drug inhibition. Conversely, C077 TSC1^KO and TSC2^KO clones were counter-selected upon drug treatment and showed altered cell cycle profiles. Presumably differences in the somatic mutation profile of CHL-1 and C077 dictate these divergent roles of the TSC pathway in growth response following treatment with the combination.

Nilotinib and trametinib synergise in two in vivo melanoma models

We next tested the nilotinib/trametinib combination in vivo. Firstly, we inoculated NOD.Cg- PrkdcF** //2rg^tm1Wjl/SzJ (NSG) mice with the sensitive cell line MeWo and after tumor establishment treated these mice with vehicle, trametinib, nilotinib or the nilotinib/trametinib combination by gavage (n=5 mice, n=10 tumors per each of the 4 treatment groups, see Methods). The combination induced a significant reduction of tumor growth compared to mock, nilotinib only and trametinib only treatments (P value = 0.0004, 0.0005, 0.004, respectively by unpaired Student's t-test on the last time point, Figure 5a). To validate the drug combination in a model that more closely represents human tumors^24,25, we interrogated a collection of BRAF/NRAS WT melanoma patient derived xenografts (PDX)²⁶ for expression of AXL and MITF (Figure 5b). Since AXL⁺ cell lines are enriched for synergy between nilotinib and trametinib (Figure 4a), we selected the PDX line M003 which expressed the highest level of AXL for in vivo studies (Figure 5c). In the in vivo experiments (performed as described above for MeWo), nilotinib induced a mild reduction of tumor growth, trametinib induced a more pronounced tumor growth reduction and the combination induced a partial regression of the tumors with a reduction in tumour volume that was maintained for the duration of the experiment (39 days of treatment; P= 0.0002, 0.0009, 0.0023 vs mock-treated by unpaired Student's t-test on the last time point, respectively; Figure 5d). Analyses of representative tumors (n=4) at the experimental endpoint confirmed that the drug combination induced a significant reduction of p-ERK (Figure 5e), as previously observed in cell lines (Figure 3d), and caused an alteration of the tumor cellular morphology (Figure 5f). These results confirmed the synergy between trametinib and nilotinib in two in vivo models of BRAF/NRAS WT melanoma. DISCUSSION

We assembled a collection of 20 BRAF/NRAS WT melanoma cell lines characterized in-depth for somatic mutations, copy number variation, gene and microRNA expression and viability following treatment with 180 drug combinations. Analysis of these data revealed that synergy between anti-cancer drugs is rare, and in many cases private to a specific cell line. By applying a stringent and multi-step validation approach, we confirmed robust synergistic interactions between temozolomide and olaparib, and also a combination of nilotinib and MEK inhibitors. Since the combination of alkylating agents and PARP inhibitors has failed to elicit clinical benefit in melanoma clinical trials¹⁴'^15,27, we focussed our efforts on the nilotinib/trametinib combination. Notably, nilotinib and trametinib are approved for the treatment of leukemia and melanoma²⁸ and we detected synergy at concentrations far below the peak of plasma concentration achieved in patients^28"30. Additionally, our results in mouse models show that the nilotinib/trametinib combination can be tolerated in vivo with a regimen that induced regression in a PDX model of BRAF/NRAS WT human melanoma. In sum the nilotinib/trametinib combination showed synergy (AUC >0.1) in 42.5% (17/40) of all melanoma cell lines including 6/24 BRAF/NRAS wildtype lines. Further, we also observed strong activity of the drug combination (AUC<0.4) in 65% (26 out of 40) of lines, including 62.5% (15 out of 24) of our BRAF/NRAS WT lines. Collectively this effect on melanoma cell growth is the result of high single agent activity, additivity, and synergy and suggests that the combination could benefit a broad range of patients.

Notably, we failed to identify a genomic biomarker linked to drug synergy but observed that AXL expression was associated with synergy between nilotinib and trametinib in BRAF/NRAS WT cell lines. AXL expression was frequently found in BRAF/NRAS WT PDX (5 out of 6, Figure 5B) and is reported in a significant fraction of melanomas^18,31 , thus suggesting that a sizeable fraction of patients with BRAF/NRAS WT melanoma could benefit from the nilotinib/trametinib combination. Previous studies have suggested that AXL expression is associated with a phenotype switch of melanoma cells towards a transcriptional status associated with drug resistance^17,18. In support, we observed that BRAFI N 'RAS WT AXL⁺ cell lines display resistance to MEK inhibitors alone, in agreement with previous reports of the resistance of AXL⁺ melanomas to MAPK inhibitors^17,18, yet were sensitive to the nilotinib/trametinib combination, suggesting a way of using the combination in the clinic. Given the promising results obtained by combining targeted and immune therapy^32"34, we also envision the possible use of the nilotinib/trametinib combination with immune checkpoint inhibitors to improve patient outcome and disease control, or as a second line treatment following relapse after immune checkpoint therapy.

Mechanistically we revealed that that the synergy between nilotinib and trametinib was associated with enhanced p-ERK inhibition compared to trametinib or nilotinib alone in all melanoma subtypes, suggesting that inhibition of the MAPK pathway is linked to their synergistic effect. This finding was confirmed in vivo in a PDX model and further supported by CRISPR/Cas9 screening in sensitive lines which revealed that nilotinib/trametinib resistance can be mediated by the same genes responsible for resistance to MAPK pathway inhibitors^{21 ,22}. The previous findings demonstrating nilotinib activity on RAFs¹⁹ further corroborate that nilotinib and trametinib could synergize by inhibiting MAPK signalling and blunting ERK activation.

Among the common nilotinib/trametinib resistance genes, we found 4 previously validated vemurafenib resistance genes²¹ , including two members of the SAGA/STAGA complexes, MED12 and NF2 , and also regulators of estrogen beta pathway which have anti-proliferative activity in melanoma^35,36. Some of those genes are mutated in a fraction of melanoma, thus representing putative prospective markers. Intriguingly, we validated that loss of TSC activity induces drug resistance in CHL-1 but causes sensitization of C077 cells, a dichotomy that could be explained by KRAS amplification in CHL-1 synergising with TSC loss and consequent mTORCI activation in promoting cell growth and drug resistance^37,38. These results highlight how cancer somatic mutations impact drug response ³·³⁹·⁴⁰.

Although we primarily focused on BRAF/NRAS WT melanoma models, we also observed the synergy in 6 out of 9 BRAF^⁶⁰⁰- and in 2 out of 5 NRAS⁰⁶^ -mutant cell lines. It is intriguing to speculate that the combination could be effective also in a subgroup of patients affected by BRAF^⁶⁰⁰ or A/ ¾AS^Q61 -mutant melanoma. Currently we lack a biomarker for the patients' selection in these subgroups of melanoma. Further studies are warranted to confirm our findings and to test if nilotinib/trametinib combination can outperform other available treatments. In summary, we performed a high-throughput drug screening in melanoma cell lines and identified that the combination of nilotinib with trametinib is synergistic in BRAF/NRAS WT melanoma. We showed that AXL expression is associated with drug synergy and validated our in vitro data with two in vivo models. Our results provide a rationale for the development of the nilotinib/trametinib combination in BRAF/NRAS WT melanomas.

As a postscript, our paper, published after the priority date of the present application as https://www.biorxiv.org/content/early/2017/09/28/195354, indicates that a nilotinib/M EK inhibitor combination may represent an effective therapy in BRAF/NRAS wild type melanoma patients. Nilotinib and MEK inhibitors synergise in killing a significant fraction of melanoma cell lines. Both drugs work in concert to suppress pERK; a finding supported by genome- wide CRISPR screening which revealed that resistance mechanisms converge on regulators of the MAPK pathway.

In this manuscript we found that AXL expression is significantly associated with the synergistic effect displayed by a nilotinib/M EK inhibitor combination in BRAF/NRAS wild type melanoma cell lines. Additionally, we performed experiments of overexpression and knockdown of AXL, through which we showed that AXL knockdown reduced the effect of nilotinib/trametinib in sensitive cell lines, thus providing a mechanistic link between AXL expression and the drug combination sensitivity.

As discussed in the manuscript, our findings indicate that AXL expression as a marker of sensitivity to this combination is likely to be linked to its association with a transcriptional cell state involved in drug resistance rather than being the only functional regulator of resistance.

References for Background, Results and Discussion

1. Larkin, J., et al. Combined vemurafenib and cobimetinib in BRAF-mutated melanoma.

The New England journal of medicine 371 , 1867-1876 (2014).

2. Robert, C, et al. Improved overall survival in melanoma with combined dabrafenib and trametinib. The New England journal of medicine 372, 30-39 (2015).

3. Welsh, S.J., Rizos, H., Scolyer, R.A. & Long, G.V. Resistance to combination BRAF and MEK inhibition in metastatic melanoma: Where to next? European journal of cancer 62, 76-85 (2016).

4. Hodi, F.S., et al. Combined nivolumab and ipilimumab versus ipilimumab alone in patients with advanced melanoma: 2-year overall survival outcomes in a multicentre, randomised, controlled, phase 2 trial. The Lancet. Oncology M , 1558-1568 (2016). Larkin, J., Hodi, F.S. & Wolchok, J.D. Combined Nivolumab and Ipilimumab or Monotherapy in Untreated Melanoma. The New England journal of medicine 373, 1270-1271 (2015).

Krauthammer, M., et al. Exome sequencing identifies recurrent mutations in NF1 and RASopathy genes in sun-exposed melanomas. Nature genetics 47 , 996-1002 (2015). Cancer Genome Atlas, N. Genomic Classification of Cutaneous Melanoma. Cell 161 , 1681-1696 (2015).

Carvajal, R.D., et al. Phase II Study of Nilotinib in Melanoma Harboring KIT Alterations Following Progression to Prior KIT Inhibition. Clinical cancer research : an official journal of the American Association for Cancer Research 21 , 2289-2296 (2015). Guo, J., et al. Phase II, open-label, single-arm trial of imatinib mesylate in patients with metastatic melanoma harboring c-Kit mutation or amplification. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 29, 2904-2909 (201 1).

lorio, F., et al. A Landscape of Pharmacogenomic Interactions in Cancer. Cell 166, 740-754 (2016).

Barretina, J., et al. The Cancer Cell Line Encyclopedia enables predictive modelling of anticancer drug sensitivity. Nature 483, 603-607 (2012).

Gonzalez-Perez, A., et al. IntOGen-mutations identifies cancer drivers across tumor types. Nature methods 10, 1081-1082 (2013).

Khan, O.A., et al. A phase I study of the safety and tolerability of olaparib (AZD2281 , KU0059436) and dacarbazine in patients with advanced solid tumours. British journal of cancer 104, 750-755 (2011).

Middleton, M.R., et al. Randomized phase II study evaluating veliparib (ABT-888) with temozolomide in patients with metastatic melanoma. Annals of oncology : official journal of the European Society for Medical Oncology 26, 2173-2179 (2015).

Konieczkowski, D.J., et al. A melanoma cell state distinction influences sensitivity to MAPK pathway inhibitors. Cancer discovery 4, 816-827 (2014).

Muller, J., et al. Low MITF/AXL ratio predicts early resistance to multiple targeted drugs in melanoma. Nature communications 5, 5712 (2014). Packer, L.M., et al. Nilotinib and MEK inhibitors induce synthetic lethality through paradoxical activation of RAF in drug-resistant chronic myeloid leukemia. Cancer cell 20, 715-727 (201 1).

Tzelepis, K., et al. A CRISPR Dropout Screen Identifies Genetic Vulnerabilities and Therapeutic Targets in Acute Myeloid Leukemia. Cell reports 17, 1193-1205 (2016). Shalem, O., et al. Genome-scale CRISPR-Cas9 knockout screening in human cells. Science 343, 84-87 (2014).

Doench, J.G., et al. Optimized sgRNA design to maximize activity and minimize off- target effects of CRISPR-Cas9. Nature biotechnology 34, 184-191 (2016).

Szklarczyk, D., etal. STRING v10: protein-protein interaction networks, integrated over the tree of life. Nucleic acids research 43, D447-452 (2015).

Girotti, M.R., et al. Application of Sequencing, Liquid Biopsies, and Patient-Derived Xenografts for Personalized Medicine in Melanoma. Cancer discovery 6, 286-299 (2016).

Kemper, K., et al. Intra- and inter-tumor heterogeneity in a vemurafenib-resistant melanoma patient and derived xenografts. EMBO molecular medicine 7, 1 104-1118 (2015).

Kemper, K., et al. BRAF(V600E) Kinase Domain Duplication Identified in Therapy- Refractory Melanoma Patient-Derived Xenografts. Cell reports 16, 263-277 (2016). Hopkins, T.A., et al. Mechanistic Dissection of PARP1 Trapping and the Impact on In Vivo Tolerability and Efficacy of PARP Inhibitors. Molecular cancer research : MCR 13, 1465-1477 (2015).

National-Cancer-lnstitute. Drugs Approved for Different Types of Cancer. (https://www.cancer.gov/about-cancer/treatment/drugs/cancer-type, 2015).

Weisberg, E., Manley, P.W., Cowan-Jacob, S.W., Hochhaus, A. & Griffin, J.D. Second generation inhibitors of BCR-ABL for the treatment of imatinib-resistant chronic myeloid leukaemia. Nature reviews. Cancer!, 345-356 (2007).

Infante, J.R., et al. Safety, pharmacokinetic, pharmacodynamic, and efficacy data for the oral MEK inhibitor trametinib: a phase 1 dose-escalation trial. The Lancet. Oncology 13, 773-781 (2012).

Tirosh, I., et al. Dissecting the multicellular ecosystem of metastatic melanoma by single-cell RNA-seq. Science 352, 189-196 (2016).

Hornet Moreno, B., Mok, S., Comin-Anduix, B., Hu-Lieskovan, S. & Ribas, A. Combined treatment with dabrafenib and trametinib with immune-stimulating antibodies for BRAF mutant melanoma. Oncoimmunology 5, e1052212 (2016).

Hu-Lieskovan, S., Robert, L., Hornet Moreno, B. & Ribas, A. Combining targeted therapy with immunotherapy in BRAF-mutant melanoma: promise and challenges. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 32, 2248-2254 (2014).

Eroglu, Z. & Ribas, A. Combination therapy with BRAF and MEK inhibitors for melanoma: latest evidence and place in therapy. Therapeutic advances in medical oncology 8, 48-56 (2016).

James, M.F., et al. NF2/merlin is a novel negative regulator of mTOR complex 1 , and activation of mTORCI is associated with meningioma and schwannoma growth. Molecular and cellular biology 29, 4250-4261 (2009).

Marzagalli, M., et al. Estrogen Receptor beta in Melanoma: From Molecular Insights to Potential Clinical Utility. Frontiers in endocrinology 7, 140 (2016).

Liang, M.C., et al. TSC1 loss synergizes with KRAS activation in lung cancer development in the mouse and confers rapamycin sensitivity. Oncogene 29, 1588- 1597 (2010).

Komatsu, N., Fujita, Y., Matsuda, M. & Aoki, K. mTORCI upregulation via ERK- dependent gene expression change confers intrinsic resistance to MEK inhibitors in oncogenic KRas-mutant cancer cells. Oncogene 34, 5607-5616 (2015).

Johnson, G.L., Stuhlmiller, T.J., Angus, S.P., Zawistowski, J.S. & Graves, L.M. Molecular pathways: adaptive kinome reprogramming in response to targeted inhibition of the BRAF-MEK-ERK pathway in cancer. Clinical cancer research : an official journal of the American Association for Cancer Research 20, 2516-2522 (2014). Roesch, A. Tumor heterogeneity and plasticity as elusive drivers for resistance to MAPK pathway inhibition in melanoma. Oncogene 34, 2951-2957 (2015).

Claims

1. Use of AXL as a biomarker for identifying responders to cancer treatment with a pharmaceutical composition comprising one or more tyrosine kinase inhibitors in combination with one or more MEK inhibitors.

2. A pharmaceutical composition comprising one or more tyrosine kinase inhibitors in combination with one or more MEK inhibitors for use in the treatment of cancer in a subject identified as having AXL expression in a biological sample obtained from said subject.

3. A method of treating cancer in a subject which comprises the steps of:

4. The use, pharmaceutical composition for use or method as defined in any one of claims 1 to 3, wherein the tyrosine kinase inhibitor is selected from one or more of: axitinib, bosutinib, cediranib, dasatinib, erlotinib, gefitinib, imatinib, lapatinib, lestaurtinib, nilotinib, semaxanib, sunitinib, ponatininb, bafetinib, vandetanib, cabozantinib, BMS-777607, R428 (BGB324), Gilteritinib, LDC1267, TP-0903, BGB324 and S49076.

5. The use, pharmaceutical composition for use or method as defined in claim 4, wherein the tyrosine kinase inhibitor is selected from nilotinib, in particular nilotinib as the sole tyrosine kinase inhibitor.

6. The use, pharmaceutical composition for use or method as defined in claim 4 or claim 5, wherein nilotinib is present within the pharmaceutical composition as the

hydrochloride monohydrate salt.

7. The use, pharmaceutical composition for use or method as defined in any one of claims 1 to 6, wherein the MEK inhibitor is selected from one or more of: trametinib, cobimetinib (XL518), binimetinib (MEK162), selumetinib, PD-325901 , CI-1040, PD035901 , pimasertib, RG7304, SHR7390, ATR001 , ATR004, ATR005, CCT196969, CCT241 161 , CCD450, EBI 1051 , E601 and TAK-733.

8. The use, pharmaceutical composition for use or method as defined in any one of claims 1 to 7, wherein the MEK inhibitor is selected from trametinib, in particular trametinib as the sole MEK inhibitor.

9. The use, pharmaceutical composition for use or method as defined in any one of claims 1 to 8, wherein the cancer is selected from a tumour characterised by AXL expression.

10. The use, pharmaceutical composition for use or method as defined in any one of claims 1 to 9, wherein the cancer is selected from malignant melanoma, such as BRAF/NRAS wildtype (WT) melanoma.

1 1. The use, pharmaceutical composition for use or method as defined in claim 10, wherein the BRAF/NRAS wildtype (WT) melanoma is selected from the NF1 mutant (NF1 m) or the triple wild type melanoma (TWT) (i.e. BRAF, NRAS and NF1).

12. The use, pharmaceutical composition for use or method as defined in any one of claims 1 to 1 1 , wherein AXL expression is protein or RNA expression.

13. The use, pharmaceutical composition for use or method as defined in claim 12, wherein AXL expression is protein or RNA expression.

14. The use, pharmaceutical composition for use or method as defined in any one of claims 1 to 13, wherein AXL expression is replaced by expression of a molecule, or a measurable fragment of said molecule, found upstream or downstream of AXL in a biological pathway.

15. The use, pharmaceutical composition for use or method as defined in claim 14, wherein said molecule is selected from the pathways: MAPK-ERK, PI3K-AKT,

Phospholipase C and N FKB.