A preclinical pipeline to evaluate migrastatics as therapeutic agents in metastatic melanoma

ARTICLE Translational Therapeutics A preclinical pipeline to evaluate migrastatics as therapeutic agents in metastatic melanoma Oscar Maiques 1,2 , Bruce Fanshawe 2,3 , Eva Crosas-Molist 1,2 , Irene Rodriguez-Hernandez 1,2 , Alessia Volpe 3,6 , Gaia Cantelli 2 , Lena Boehme 2 , Jose L. Orgaz 1,2,7 , Faraz K. Mardakheh 4 , Victoria Sanz-Moreno 1,2 and Gilbert O. Fruhwirth 3,5 BACKGROUND: Metastasis is a hallmark of cancer and responsible for most cancer deaths. Migrastatics were defined as drugs interfering with all modes of cancer cell invasion and thus cancers’ability to metastasise. First anti-metastatic treatments have recently been approved. METHODS: We used bioinformatic analyses of publicly available melanoma databases. Experimentally, we performed in vitro target validation (including 2.5D cell morphology analysis and mass spectrometric analysis of RhoA binding partners), developed a new traceable spontaneously metastasising murine melanoma model for in vivo validation, and employed histology (haematoxylin/ eosin and phospho-myosin II staining) to confirm drug action in harvested tumour tissues. RESULTS: Unbiased and targeted bioinformatic analyses identified the Rho kinase (ROCK)-myosin II pathway and its various components as potentially relevant targets in melanoma. In vitro validation demonstrated redundancy of several RhoGEFs upstream of RhoA and confirmed ROCK as a druggable target downstream of RhoA. The anti-metastatic effects of two ROCK inhibitors were demonstrated through in vivo melanoma metastasis tracking and inhibitor effects also confirmed ex vivo by digital pathology. CONCLUSIONS: We proposed a migrastatic drug development pipeline. As part of the pipeline, we provide a new traceable spontaneous melanoma metastasis model for in vivo quantification of metastasis and anti-metastatic effects by non-invasive imaging. British Journal of Cancer (2021) 125:699–713; https://doi.org/10.1038/s41416-021-01442-6 BACKGROUND Metastasis is a hallmark of cancer 1 but has been difficult to target due to the complexity of its underlying mechanisms and difficulties defining appropriate clinical study end points. Notably, an anti-metastatic prostate cancer drug, apalutamide, was recently approved by the FDA involving the new endpoint of ‘metastasisfree survival’, measuring the length of time that tumours did not spread to other parts of the body or that death occurred after starting treatment. 2 Recently, we defined a new class of drugs that we coined ‘migrastatics’, and which would comprise drugs interfering with all modes of cancer cell invasion and therefore with their ability to metastasise. 3,4 Migrastatics have the potential to be of particular value for the treatment of cancers with the propensity to metastasise early. While in most cases metastasis is associated with large primary neoplasms, it can occur in some cancers when primary tumours are still small. Melanoma, which is a particularly aggressive form of skin cancer, belongs to the group of cancers that spread early, with metastasis being reported for lesion sizes below two millimetres. 5 Moreover, patients suffering from malignant advanced melanoma have a very poor prognosis. 6,7 Rho kinase (ROCK) downstream of RhoA has been shown to play an important role in tumour growth, cell migration, invasion and survival. 8–10 The ROCK-myosin II pathway regulates actin polymerisation and cell contractility 11–14 and is a key regulator of melanoma invasion and metastasis. 9,13,15–17 Notably, cytoskeletal remodelling and changes in the expression and activity of the ROCK-myosin II pathway components were also recently discovered to play a major role in the acquisition of melanoma resistance to MAPK inhibitors. 18 Several key regulators of RhoA activity have been described including guanine nucleotide exchange factors (GEFs) and GTPase-activating proteins www.nature.com/bjc Received: 24 March 2020 Revised: 23 March 2021 Accepted: 13 May 2021 Published online: 25 June 2021 1 Centre for Tumour Microenvironment at Barts Cancer Institute, Queen Mary University of London, Charterhouse Square Campus, John Vane Science Centre, London, UK; 2 Randall Division of Cell and Molecular Biophysics, King’s College London, New Hunt’s House, Guy’s Campus, London, UK; 3 Imaging Therapies and Cancer Group, Comprehensive Cancer Centre, School of Cancer and Pharmaceutical Studies, King’s College London, Guy’s Campus, London, UK; 4 Centre for Cancer Cell & Molecular Biology at Barts Cancer Institute, Queen Mary University of London, Charterhouse Square Campus, John Vane Science Centre, London, UK and 5 School of Biomedical Engineering and Imaging Sciences, King’s College London, St. Thomas’Hospital, London, UK Correspondence: Victoria Sanz-Moreno ([email protected]) or Gilbert O. Fruhwirth (gilbert.fruh[email protected]) 6 Present address: Molecular Imaging Group, Department of Radiology, Memorial Sloan Kettering Cancer Center, New York, NY, USA 7 Present address: Instituto de Investigaciones Biomedicas ‘Alberto Sols’, CSIC-UAM, Madrid, Spain These authors contributed equally: Oscar Maiques, Bruce Fanshawe, Eva Crosas-Molist. These authors jointly supervised this work: Victoria Sanz-Moreno, Gilbert O. Fruhwirth. ©The Author(s) 2021 Published by Springer Nature on behalf of Cancer Research UK (GAPs). 19,20 RhoGEFs have been under extensive investigation as therapeutic targets, for example, various inhibitors of Rho GTPase signalling. 21–23 Together with ROCK, they represent potential antimetastatic targets. 3,4 What is less well understood is which of these molecules would be an effective target. More generally, development of migrastatic drugs is dependent on refinement of approval regulations as well as widespread screening for compounds with low general toxicity but high efficacy against cancer cell spread as determined in 3D cell models and pre-clinical animal models. The latter are required to reproduce the metastatic process as closely as possible to the human scenario. Preclinical metastasis models would greatly benefit from strategies to visualise cancer cells in vivo and quantitatively track their spread over time to assess the efficacy of migrastatics on the preclinical level. Long-term observation and quantification of cancer and its spread is achievable in pre-clinical models by non-invasive reporter gene-based radionuclide imaging. We have previously demonstrated that whole-body medical imaging technologies such as single-photon emission computed tomography (SPECT) or positron emission tomography (PET) in combination with reporter gene technology and appropriate imaging probes enabled quantitative in vivo tracking of murine cancer and its metastasis. 24–27 Therefore, we employed the sodium iodide symporter (NIS) as a radionuclide reporter. To streamline the generation of in vivo traceable cancer cell lines and their ex vivo detection (e.g. by histology or flow cytometry) we combined the radionuclide reporter NIS with a fluorescent protein such as GFP or RFP. 26,27 Here, we demonstrate a systematic ‘pipeline’approach for the preclinical development of migrastatics, which includes unbiased and targeted bioinformatic analyses for target identification, relevant examples for in vitro target validation, a specifically and newly developed traceable and spontaneously spreading melanoma model to quantify metastasis in vivo as well as the tissuelevel validation of candidate migrastatic efficacy. METHODS Information regarding cell proliferation and cellular radiotracer uptake assays, flow cytometry and microscopy of cells, immunoblotting and radioactivity analysis in tissues is detailed in Supplementary Information (SI). Reagents All standard chemicals and molecular biology reagents were either from Fisher-Scientific, NEB, Millipore, Sigma or VWR. Tissue culture materials were from Corning, Sarstedt or TPP. Tc-99mpertechnetate ([ 99m Tc]TcO 4 − ) in saline was eluted from a Drytec generator (GE Healthcare, UK) from the Radiopharmacy at Guy’s and St. Thomas’Hospital Nuclear Medicine Department. Wheat germ agglutinin (WGA) conjugated to AlexaFluor488 was from Invitrogen. Mowiol was from ICN (Costa Mesa, CA, US). ROCK inhibitors: Y27632 (Tocris Bioscience, UK), GSK269962 (Axon Medchem, Netherlands), H1152 (Merck, Germany), AT13148 (Selleckchem, USA). SiRNAs targeting ARHGEF1, ARHGEF11 and ARHGEF2 were from Dharmacon (siGENOME SMARTpool). Primary antibodies: polyclonal rabbit anti-mCherry (Abcam; #ab167543; 1.0 µg/mL for immunoblotting, 3.3 µg/mL for immunohistochemistry), polyclonal rabbit anti-phospho myosin light chain specific for phosphorylation at residue Ser19 (pMLC2) (Cell Signalling Technologies (CST); #3671; 10 µg/mL for immunohistochemistry), monoclonal mouse anti-GAPDH (Genetex; #GTX239; 0.3 µg/mL for immunoblotting), monoclonal mouse anti-MLC (MYL9/MYL12A/B; clone E-4; Santa Cruz Biotechnology; 0.4 µg/mL for immunoblotting), monoclonal rabbit anti-ARHGEF1 (clone D25D2; CST; #D3669, 1:1000 dilution for immunoblotting) and anti-ARHGEF2 (clone 55B6; CST; #4076, 1:1000 dilution for immunoblotting), mouse monoclonal anti-ARHGEF11 (Santacruz Biotechnology; clone D-9; 0.4 µg/mL for immunoblotting). Secondary antibodies: for immunofluorescence staining a Cy5-conjugated goat antirabbit antibody (Jackson Immunoresearch; #111-175-144; 2.0 µg/ mL) was used, while for immunoblotting ECL Plus or Prime ECL detection systems (GE Healthcare) with horseradish peroxidase (HRP)-conjugated secondary antibodies (GE Healthcare) were used for detection. For immunohistochemistry, relevant secondary antibodies were conjugated to HRP and included polyclonal IgGHRP specific for either rabbit or mouse antigens (Dako; #P0448 and #P0447, respectively, each used at 1:100 dilution). Cells 4599 murine BRAF V600E melanoma cells were a kind gift from Dr. Amine Sadok (Institute of Cancer Research) and Prof Richard Marais (Cancer Research-UK Manchester Institute). A375M2 human BRAF V600E melanoma cells were a kind gift from R. Hynes (Howard Hughes Medical Institute, MIT, Cambridge, MA, USA 11 ). MTLn3E. Δ34 cells were previously described. 26 293T cells were purchased from ATCC. All melanoma cell lines were confirmed by STR profiling (DNA Diagnostics Center, Fairfield/OH, USA; May 2016). Melanoma and 293T cells were cultured in DMEM containing 10% (v/v) fetal bovine serum (FBS; Biosera, UK), pyruvate (1 mM) and 4.5 g/L D-glucose. MTLn3E.Δ34 cells were grown in alpha-MEM supplemented with 5% (v/v) FBS. All media were also supplemented with L-glutamine (2 mM), penicillin (100 IU/mL) and streptomycin (50 μg/mL). All cell lines were confirmed negative for mycoplasma (LookOut Mycoplasma PCR Detection Kit, Sigma) quarterly throughout the study. Animals In this study, young adult (5–6 weeks old, 24 ± 1 g) male NOD.CgPrkdc scid Il2rg tm1Wjl /SzJ mice (NSG; purchased from Charles River UK) were used for all animal experiments. All mice were maintained within the King’s College London Biological Services Unit under specific pathogen-free conditions in a dedicated and licensed air-conditioned animal room (at 23 ± 2 °C and 40–60% relative humidity) under light/dark cycles lasting 12 h every day. They were kept in individually ventilated standard plastic cages (501 cm 2 floor space; from Tecniplast) including environmental enrichment and bedding material in the form of sterilised wood chips, paper stripes and one cardboard roll per cage. Maximum cage occupancy was five animals, and animals were moved to fresh cages with fresh environmental enrichment and bedding material twice per week. Sterilised tap water and food were available ad libitum; food was PicoLab Rodent Diet 20 (LabDiet) in the form of 2.5 × 1.6 × 1.0 cm oval pellets that were supplied at the top of the cages. For imaging, animals were anaesthetised using isoflurane (1.5% (v/v) in pure O 2 ). After imaging, mice were either left to recover from anaesthesia (by withdrawal of anaesthetic) in a pre-warmed chamber or sacrificed under anaesthesia by cervical dislocation. Tissues were harvested as indicated in ‘Methods’ sections below and Figure captions. In previous experiments with different traceable cell types expressing NIS as an imaging reporter we showed a clear dependence of the acquired radionuclide imaging signal from the cell number. The standard deviation (SD) of cell pellet signals at identical radiotracer labelling was previously determined to be <30% (at relevant low numbers of <10,000 NIS expressing cells in a pellet of 1 million cells; SD is much lower at higher cell numbers). The detection sensitivity was ~1000 labelled cells/million cells. 26 This data was used together with an αof 0.05 and a power of ≥90% to determine minimum cohort sizes. For longitudinal experiments, cohort sizes were oversubscribed to hedge against potential adverse effects of metastasis and resultant animal sacrifice, which if premature would endanger the whole study. For initial tumour model validation, 4599.NC and 4599 tumours were grown in cohorts of four animals each. For drug treatment studies, cohort sizes were N =6 with four animals per cohort being subjected to serial A preclinical pipeline to evaluate migrastatics as therapeutic agents in. . . O Maiques et al. 700 1234567890();,: imaging. The total number of animals used was 32. No adverse events were associated with the procedures performed in this study and animals put on weight in line with strain expectations (data from Charles River UK) throughout. Sentinel animals were kept on the same IVC racks as experimental animals and confirmed to be healthy after completion of the studies. Melanoma cell line and patient mRNA database analysis Expression data of primary melanocytes and melanoma cell lines were downloaded from public database websites and normalised as previously described. 28,29 Briefly, we used four melanocyte datasets from ref. 28 (GSE4570, GSE4840), refs. 30,31 , and data from melanoma cell lines (Mannheim cohort GSE4843 and Philadelphia cohort GSE4841) were obtained from ref. 28 Expression data from human melanoma patient studies; GSE8401, 32 GSE7553, 33 GSE3189, 34 and GSE46517 35 were extracted from Gene Expression Omnibus (GEO) and normalised using Gene Pattern software (http://www.broadinstitute.org/cancer/software/genepattern/). Normalised mRNA gene expression of 389 human melanoma samples and mutational data of 311 human melanoma samples from TCGA (The Cancer Genome Atlas) database were downloaded from cBioportal. 36,37 We only considered TCGA samples with greater than 70% tumour cell content from patients who had not received neo-adjuvant treatment prior to tumour resection. REACTOME pathway analysis Genes significantly upregulated and downregulated in metastatic versus primary melanomas from the TCGA database were selected using a log2FC ≥0.5 and adjusted p-value < 0.01. To identify overrepresented functional pathways in metastatic versus primary melanomas, these differentially expressed genes were analysed using the WEB-based GEne SeT AnaLysis Toolkit (WebGestalt). 38 Pathway enrichment was performed using the functional database REACTOME. The parameters for the enrichment analysis included categories with 5–2000 involved genes, the multiple test adjustment used was the Benjamini–Hochberg (BH) method, and the significant enriched pathways were selected using a false discovery rate (FDR) < 0.05. Weighted set cover method was run to reduce redundancy of the gene sets in the enrichment result. Differential mass spectrometric analysis of RhoA pulldown GST-RhoA pulldown of proteins and subsequent mass spectrometric analysis were performed as described previously, 39 with minor modifications. Briefly, heavy and light SILAC (Stable Isotope Labelling with Amino Acids in Cell Culture)-labelled A375M2 cells were lysed (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% NP-40, 10 mM EDTA, plus phosphatase and protease inhibitor cocktails from Roche), and lysates cleared by centrifugation at 8000 × gfor 20 min. GST pulldowns were performed using purified bacteriallyexpressed GST (light) or GST-RhoA (heavy) immobilised on glutathione–sepharose beads. 50 µL of the bed volume of beads was added to ∼2 mg of lysates for 1 h, before three 2 mL washes of the beads in lysis buffer, and elution in 50 µL of boiling 2-fold SDS-PAGE sample buffer. The eluates were then mixed and resolved on a 10% Bis-Tris-Midi gel using MOPS buffer (Life Technologies). In-gel trypsin digestions, peptide extractions and liquid-chromatography-coupled tandem mass spectrometry (LCMS/MS) analysis were performed as described previously. 40 Mass spectrometry search and quantifications were done by using Maxquant 41 using the IPI v3.68 database. Hits were selected based on their SILAC ratio value exhibiting a 4-fold or more enrichment in the GST-RhoA vs. GST sample. Generation of reporter gene-expressing melanoma cells 4599 melanoma cells were virally transduced to express NISmCherry. Lentiviral particles were generated in 293T cells using pLNT SFFV NIS-mCherry, pΔ8.91 and pVSV-G plasmids as described previously. 27 Cells were expanded and selected based on mCherry fluorescence by FACS sorting using a FACS Aria III (BD Biosciences, UK; 100 μm nozzle, 3 kV, 20kPa). Cells were expanded and grown for one week before reanalysis by flow cytometry. The cell line was considered stable as no change in purity was detected after subsequent culture of four weeks (see Supplementary Materials). Animal tumour model 4599.NC cells were trypsinised, washed with pre-warmed Hank’s buffered saline without Ca 2+ and Mg 2+ (HBSS), re-suspended in HBSS and counted. Aliquots of 2 × 10 5 cells in 20 µL HBSS were injected intradermally on the left flank of the mice. Once palpable, tumour volumes were measured with callipers using the formula V =π/6·L·W·D, wherein L is length, W is width and D is depth of the palpable tumour. Tumour volumes were determined by qualified staff using callipers at least every third day throughout the study. Drug treatment of tumour-bearing animals Tumours were established as described above and grown to sizes of ~20 mm 3 , a size reached around 10 days post inoculation. Animals were then randomised and assigned to either vehicle or treatment cohorts before a baseline SPECT/CT scan was performed, and ROCK inhibitor treatment started on the following day. Y27632 was dissolved in phosphate-buffered saline (without Ca 2+ /Mg 2+ ; PBS) and delivered intraperitoneally (100 µL) at a dose of 156 µmol/kg every other day to treatment groups, control cohorts received PBS only. GSK269962 (Axon Medchem, Netherlands) was freshly prepared on the day in PBS containing 10% (v/ v) Tween-80, 6.5% (v/v) ethanol and 7% (v/v) DMSO as previously described by ref. 42 GSK269962 was administered by oral gavage (200 µL) at a dose of 43.5 µmol/kg daily to treatment groups with control cohorts receiving vehicle instead. In vivo radionuclide imaging Mice were anaesthetised using isoflurane (1.5% (v/v) in O 2 ) and 20 MBq [ 99m Tc]TcO 4 − in 100 μL HBSS was administered intravenously under anaesthesia. Protective eye gel was applied, and animals were left sedated and placed onto the imaging platform of a nanoSPECT/CT Silver upgrade scanner (Mediso, Hungary). 30 min post radiotracer injection, CT images were acquired (55 kVp tube voltage, 1200 ms exposure time, 360 projections) and 40 min after initial radiotracer administration a static SPECT scan was performed using 1.0 mm collimators (scan duration 30 min). For specificity tests of the NIS radiotracer, animals were first imaged and then rested awake until the radioactivity had decayed sufficiently to be regarded as negligible, i.e., 48 h (0.4% residual 99m Tc radioactivity). Subsequently, the competitive substrate perchlorate was administered at a dose of 250 mg/kg and 40 min later animals were re-imaged as described above. In vivo image analysis All SPECT/CT datasets were reconstructed using a Monte Carlo based full 3D iterative algorithm (Tera-Tomo, Mediso, Hungary). Decay correction to time of injection was applied. All images were analysed using VivoQuant software (inviCRO, USA), which enabled the definition of regions of interest (ROIs) in co-registered SPECT/ CT images for quantification of radioactivity (SPECT) in tumours and metastases. The total activity in the whole animal (excluding the tail) at time zero was defined as the injected dose (ID). ROIs for different organs were defined to express uptake in each organ as a percentage of injected dose per volume (%ID/mL). The live tumour volume (LTV) was defined as the volume occupied by live tumour cells as NIS expressing tumour cells can only take up radiotracer when viable, because anionic radiotracer uptake requires symport of sodium cations. This process is dependent on an intact Na + /K + gradient across the cellular plasma membrane, which is driven by Na + /K + -ATPase function. 43 We exploited this NIS feature to determine LTV based on thresholded A preclinical pipeline to evaluate migrastatics as therapeutic agents in. . . O Maiques et al. 701 and background-corrected SPECT signals (using VivoQuant software and its implementation of Otsu’s thresholding 44 ). Tissue staining and histologic tissue analysis Formaldehyde-fixed paraffin-embedded (FFPE) tissues were prepared using standard methods and stored for a minimum of 48 h to let radioactivity decay. 5 µm tissue sections were cut using a microtome and adhered to poly-L-lysine slides, dried overnight at 40 °C, de-waxed and subjected to antigen retrieval in a pressure cooker at pH 9.0. Sections were blocked (Dual Endogenous Enzyme Blocking Reagent; Dako, UK) in 1% (w/v) BSA for 60 min at room temperature, incubated with indicated primary antibodies at 4 °C overnight before being stained with a horseradish peroxidase-conjugated secondary antibody (2 µg/mL in TBS) for 60 min at room temperature. Samples were developed using the Liquid DAB +Substrate Chromogen System (Dako, UK) and counterstained with haematoxylin before mounting. Slides were scanned using a Nanozoomer (Hamamatsu, Japan) with images being analysed using QuPath1.02. 45 Morphologic analysis of tumour tissues was performed on haematoxylinand eosin-stained sections, with the invasive front (IF) defined as melanoma cells with at least 50% contact with the matrix as previously described. 13,17,46 Using a 20-fold magnification, cell shape was graded from 0 to 3 (0 =round, 1 =ovoid, 2 = elongated and 3 =spindly), and a cell shape score assigned to IF and tumour body (TB) regions: Cell Shape Score =[(% cells with score ‘0’·0) +(% cells with score ‘1’·1) +(% cells with score ‘2’·2) +(% cells with score ‘3’·3)] as previously described. 13 The distal invasive front (DIF) was defined as the region separate from the tumour mass that contained individual or small groups of invading tumour cells. For quantification of pMLC2 levels, using QuPath1.02 positive cell detection was performed, and three different thresholds were applied according to intensity scores (0, 1, 2 and 3). Software was then trained using random trees classification algorithm and combined with intensity information to differentiate tumour from stroma. Based on intensity scores given to cells, an H-score value was extracted from each representative group. Statistical analysis GraphPad Prism v7 (La Jolla, USA) was used to calculate all statistical parameters as indicated. Generally, p-values were calculated using significance levels of α=0.05. In-text numbers indicate means of pooled data ± standard deviation (SD) unless otherwise stated. RESULTS We developed a preclinical pipeline to in vitro, in vivo and ex vivo validate candidate anti-metastatic drugs in a spontaneously metastasising non-melanogenic melanoma model (Scheme 1). Individual steps and corresponding results are reported under subsequent sub-headings. Database mining for target identification—patient databases revealed upregulation of actomyosin regulators during melanoma progression First, we enquired whether specific cellular signalling pathways were altered when comparing metastatic melanoma lesions with primary melanomas. Therefore, we used gene expression data from The Cancer Genome Atlas (TCGA) and performed pathway enrichment analysis using the open-source curated and peer-reviewed Reactome database implemented in the WEB-based GEne SeT AnaLysis Toolkit (WebGestalt). 38 This unbiased analysis revealed the significant enrichment of several pathways associated with the upand downregulated genes when comparing metastatic and primary lesions in melanoma samples (Fig. 1a). The Rho GTPase cycle was among the upregulated signalling pathways, and it has previously been implicated in tumour growth, cell migration and survival. 8–10 Patient database mining 1Target identification In vitro validation In vivo validation Ex vivo validation In vivo traceable cancer cells Orthotopic tumour model Serial whole-body imaging Treatments Time Unbiased pathway analyses (e.g. REACTOME) Gene expression analyses 23 54 Scheme 1 A preclinical pipeline for the development of novel migrastatic drugs. Schematic representation of the main workflow steps performed in this study (numbered to depict the order; for details see text). A preclinical pipeline to evaluate migrastatics as therapeutic agents in. . . O Maiques et al. 702 Notably, melanin biosynthesis genes represented the most pronounced downregulated pathway (Fig. 1a). We next investigated mRNA expression levels of various components belonging to the Rho GTPase cycle and downstream signalling pathways using (i) melanoma patient data from the Gene Expression Omnibus (GEO) studies Talantov, Kabbarah, Riker and Xu, and (ii) TCGA as well as (iii) data stemming from melanocytes and melanoma cell lines including Philadelphia and Mannheim cohorts. We present hierarchical clustering according to expression level changes of 58 GEFs, the major Rho GTPases, 0 CohesinLoading onto Chromatin Insulin processing Signalling by SCF-KIT Signalling by VEGF Class A/1 (Rhodopsin-like receptors) Rho GTPase cycle G alpha (i) signalling events Immune system Hemostasis GPCR ligand binding Neutrophil degranulation Developmental Biology Synthesis of GPI-anchored proteins Antimicrobial peptides Cell junction organization Degradation of the extracellular matrix WNT ligand biogenesis and trafficking Gap junction assembly Interleukin-36 pathway Melanin biosynthesis 12 2 3 4 4 10 5 681214 0 Up regulated genes MET vs. PRIMARYDown regulated genes MET vs. PRIMARY REACTOME pathway enrichment analysis (log2FC ≥ 0.5 and adjusted p-value < 0.01) a b Effectors Controls Melanin biosynthesis GEFs Melanoma vs Melanocytes Nevus vs Normal TTRRTKK K KX ARHGEF38 DOCK1 DNMBP PLEKHG6 ARHGEF4 BCR TIAM1 FGD6 ARHGEF7 ARHGEF10 DOCK4 ITSN2 RGNEF KALRN ARHGEF9 ARHGEF3 ARHGEF15 SOS1 SOS2 DEF6 ITSN1 DOCK9 DOCK3 DOCK6 DOCK10 DOCK5 OBSCN VAV3 VAV2 NET1 ARHGEF5 ARHGEF18 ARHGEF16 ARHGEF11 ARHGEF2 ARHGEF17 ARHGEF1 FARP2 ARHGEF40 ARHGEF6 AKAP13 VAV1 DOCK2 FGD2 TIAM2 SWAP70 FGD1 FARP1 MCF2L MCF2L2 RASGRF1 PLEKHG3 ABR MCF2 TRIO ARHGEF12 ARHGEF26 ECT2 CG Primary vs Normal Primary vs Nevus Metastasis vs Primary PM RhoGTPases Melanoma vs Melanocytes PM X = Xu CG = TCGA T = Talantov K = Kabbarah R = Riker P = Philadelphia M = Mannheim c p-value < 0.05 p-value > 0.05 0.0 1.5 TCGA Fold change MET vs. PRIMARY Nevus vs Normal Primary vs Normal Primary vs Nevus Metastasis vs Primary RK XCGTTRTKK K 2.0 1.5 1.0 0.5 0.0 PAK6 PAK4 LIMK2 LIMK1 PDPK1 PTK2 PAK1 PAK2 PAK3 ROCK1 ROCK2 MYLK3 MYLK PTK2B CIT ILK RHOB RHOC RHOA CIT LIMK1 ILK PDPK1 MYLK LIMK2 PAK6 PAK1 PTK2B MYLK3 TYRP1 SLC45A2 OCA2 GAPDH ATP5F1 CDH2 CDH1 PGK1 TYR DCT PTK2 ROCK2 ROCK1 PAK3 PAK2 PAK4 CDC42 RAC2 RAC1 RAC3 A preclinical pipeline to evaluate migrastatics as therapeutic agents in. . . O Maiques et al. 703 and corresponding effector kinases as well as melanin biosynthesis genes (cf. above) and five different control genes (Fig. 1b; for tabulated data see Supplementary Tables S1–S5). While there are over 70 human GEFs, 47,48 our analyses were restricted to 58 GEFs caused by data availability limitations in the microarray-based databases. Expression levels of many Rho-actomyosin contractility regulators were upregulated at various stages of melanoma progression. Notably, increases were more pronounced in metastatic melanoma lesions compared to melanomas at the pre-metastatic stage (Fig. 1b). In contrast, we found melanin biosynthesis genes to be significantly downregulated. We also detected upregulation of CDH2 (neural cadherin) while CDH1 (epithelial cadherin) was downregulated during progression, as was previously reported, 49 hence both CDH1 and CDH2 data agreed with prior literature reports (Fig. 1b/controls). Moreover, we analysed the previously validated ‘house-keeping’genes ATP5F1, PGK1 and GAPDH, 50 which in most comparisons showed little change across datasets. Furthermore, we used the TCGA database and compared RhoGTPases and their effector kinases of metastases and primary tumours. Interestingly, we found that the most upregulated kinases were ROCK1 and ROCK2 (Fig. 1c). We found a significant association between ROCK1/2 expression and NRAS mutations; Fig. S1). These data supported our previous finding that the RhoGTPase cycle was upregulated while melanin biosynthesis genes were downregulated (cf. unbiased REACTOME analysis in Fig. 1a). Target selection—RhoGEFs act in a redundant manner in metastatic melanoma As we found several RhoGEFs and ROCK1/2 to be increased during melanoma progression (Fig. 1), we first investigated which proteins were binding RhoA in melanoma cells. GTP-bound RhoA binds its key effector ROCK, which in turn regulates myosin II activity via phosphorylation of MLC2. To identify potential activators of RhoA, we employed Stable Isotope Labelling with Amino acids in Cell culture (SILAC) combined with subsequent liquid-chromatography-coupled tandem mass spectrometry analysis. We used the metastatic human melanoma cell line A375M2 that has been used as a model for high ROCK-myosin II pathway activity. 12,13 GST-RhoA or GST-only were used as baits in pull-down experiments under SILAC conditions with subsequent quantitative proteomic analysis to identify RhoA binders (Fig. 2a). Among the proteins with the highest Heavy (GST-RhoA): Light (GST-only) ratios (Fig. 2b, Supplementary Table S6) were ROCK2 and serine/ threonine-protein kinase N2 (PKN2), both well-known effectors of RhoA. 51 This validated the approach, demonstrating the detection of known RhoA binding partners both upstream and downstream to RhoA. Importantly, we found three different RhoGEFs, which included ARHGEF1, ARHGEF2 and ARHGEF11 (Fig. 2b, Supplementary Table S6). Independent validation of RhoA interactions using separate GST-RhoA pulldown experiments confirmed the results (Fig. 2c). We next tested whether these RhoGEFs had unique functions impacting actomyosin contractility as a marker for invasive capacity as was previously reported. 13,16,17,46,52 Using A375M2 amoeboid metastatic melanoma cells seeded on collagen we confirmed that individual RNAi depletion of these three GEFs had no effects on morphology (Fig. 2d, e; cells grown on collagen) and myosin activity (pMLC2; Fig. 2f). In contrast, RNAi depletion of myosin light chain 2 (MYL12B; Fig. 2g–i) and pharmacological inhibition of ROCK1/2 with the small molecule inhibitor H1152 (Fig. 2j–l) resulted in marked reduction of myosin II activity and the loss of the amoeboid ‘rounded’invasive phenotype on collagen matrices. This data suggested that both regulator of G-protein signalling (RGS) domain-containing RhoGEFs and ARHGEF2 could act redundantly, thereby rendering neither of them a suitable target for full inhibition of actomyosin contractility. Furthermore, these GEFs were not associated with specific melanoma mutations (Fig. S1c-d). In contrast, ROCK1/2 did represent an attractive target in this pathway (cf. Fig. 2j–l). These exemplary in vitro validation experiments reduced the number of targets worth investigating in subsequent preclinical in vivo experiments aiming for quantification of potential anti-metastatic effects of drug candidates. Generation and characterisation of a new in vivo traceable model of spontaneous melanoma metastasis To validate inhibitory effects on cancer metastasis in vivo, a reliable spontaneous metastasis model was required. We chose the murine melanoma cell line 4599, which was derived from tumours arising in the BRAF V600E mouse model 53 and shown to spontaneously metastasise when transplanted intradermally. 15 To render 4599 cells traceable in vivo, we engineered them using lentiviral technology to constitutively express the radionuclide reporter NIS fused C-terminally to the red fluorescent protein mCherry (Fig. 3a). We purified reporter-expressing 4599.NISmCherry cells (4599.NC) by FACS and confirmed expression of the fusion reporter by immunoblotting (Fig. 3a) and flow cytometry (Fig. S2a). NIS-mCherry co-localised with the plasma membrane marker wheat germ agglutinin (WGA; conjugated to the fluorophore Alexa488), thereby suggesting correct cellular localisation of the reporter (Fig. 3b). NIS function was confirmed by uptake of the radioactive NIS substrate [ 99m Tc]TcO 4 − (Fig. 3c). NIS had not been reported to be expressed in melanoma cells and in line with this, parental 4599 cells did not take up [ 99m Tc]TcO 4 − . Specificity of NIS uptake was demonstrated through reduction of [ 99m Tc]TcO 4 − uptake in the presence of the competitive NIS substrate ClO 4 − (Fig. 3c). 4599.NC tumours were established in immunodeficient male NOD.Cg-Prkdc scid Il2rg tm1Wjl /SzJ (NSG) mice. Notably, we found no significant difference in tumour growth between tumours established from parental 4599 and reporter expressing 4599.NC cells (Fig. S2b). In vivo SPECT/CT imaging detected 4599.NC tumours and distant metastases in lung and lymph nodes (Fig. 3d). Additional signals were detected from organs expressing murine NIS, i.e. thyroid and salivary glands, stomach, lachrymal glands and testes. Signals stemming from the endogenous expression of a host reporter gene could interfere with the detection of cancer cells expressing the reporter. Importantly, such endogenous signals were neither detected in the skin (orthotopic site for melanoma) nor in typical melanoma metastasis target tissues (i.e. Fig. 1 Cancer patient database mining to identify genes involved in melanoma progression. a Unbiased REACTOME pathway enrichment analysis of differentially expressed genes in metastatic versus primary melanoma samples from the TCGA database. Analyses were performed using WebGestalt tool, significant pathways were selected using a FDR < 0.05 and bar chart shows enrichment ratios of significant pathways. b Using the indicated databases from GEO and TCGA, GEF expression was analysed. Metastatic melanoma, primary melanoma, nevi and normal tissues were compared as indicated. Comparison of melanoma cell lines from both the Philadelphia and Mannheim databases with melanocytes was also performed (right columns). Blue represents downregulation while red represents overexpression between compared groups. Additional expression analyses were performed on RhoGTPases, effector kinases, melanin biosynthesis pathway genes and controls, including ‘house-keeping’genes and two cadherins. Tabulated numeric data underpinning the shown analyses including p-values for each comparison can be found in the Supplementary Tables S1–S5. cFold change expression between metastasis and primary tumours of RhoGTPase effector kinases from TCGA melanoma patients. ROCK1 and ROCK2 highlighted in red, as the most upregulated genes. P-values were calculated using unpaired t-tests and triangles show statistically significant data (p< 0.05). A preclinical pipeline to evaluate migrastatics as therapeutic agents in. . . O Maiques et al. 704 lung, lymph nodes, liver, brain, bone were free of endogenous signals). This means the combination of the chosen reporter gene and its use in the context of skin cancer is well suited to study spontaneous metastasis in vivo by non-invasive imaging. We also determined whether the observed signals in cancerous tissues were due to specific NIS uptake or were possibly generated by the enhanced permeability and retention effect (EPR), which is known to play a major role in tumour uptake of various agents delivered via the blood stream. 54 Animals were therefore re-imaged 2 days after the first imaging session, but with prior administration of the Light labelled cells Heavy labelled cells A375M2 human melanoma cells Cell lysis glutathione GST glutathione GST RhoA Pulldown Elution and mixing of eluates - + SDS-PAGE Cutting and tryptic digest LC-MS/MS and differential analysis − − + + − − + input pulldown ARHGEF1 ARHGEF11 ARHGEF2 GAPDH GST: GST-RhoA: + c ab H1152 Control g f d siARHGEF1 siARHGEF11 Control siARHGEF2 0.0 0.5 1.0 Rounding index (0–1) siARHGEF1 siARHGEF2 siARHGEF11 + + + − − −− −− −− − ns ns ns Rounding index (0–1) 0.0 0.5 1.0 **** +−H1152 +− pMLC2 MLC2 GAPDH H1152: pMLC2 MLC2 GAPDH ARHGEF1 +−siARHGEF1: pMLC2 MLC2 GAPDH ARHGEF2 siARHGEF2: +− pMLC2 MLC2 GAPDH ARHGEF11 siARHGEF11: +− e jk l h pMLC2 MLC2 GAPDH siMYL12B: +− **** Rounding index (0–1) 0.0 0.5 1.0 +−siMYL12B i siMYL12B Control Protein Descriptions Rho/rac guanine nucleotide exchange factor 2 Rho guanine nucleotide exchange factor 1 Rho guanine nucleotide exchange factor 11 isoform 2 RAP1, GTP-GDP dissociation stimulator 1 isoform 1 RAP1GDS1 RPS3 ROCK2 PAK1;PKN;PKN1 ARAP3;CENTD3 ACAT PKN2 ANT2 GEMIN5 ARHGEF2 ARHGEF1 ARHGEF11 3 11 7 15 5 3 17 9 9 24 6 14 27.4 15.2 14.6 13.0 4.9 11.8 10.6 6.2 16.2 20.4 8.5 4.3 Gem-associated protein 5 40S ribosomal protein S3 ADP/ATP translocase 2 Arf-GAP with Rho-GAP domain Acetyl-CoA acetyltransferase Rho-associated protein kinase 2 Isoform 2 of Serine/threonine-protein kinase N1 Serine/threonine-protein kinase N2 Gene Names Peptide number Normalized H:L ratio A preclinical pipeline to evaluate migrastatics as therapeutic agents in. . . O Maiques et al. 705 Fig. 2 Target identification and validation on the protein level. a Experimental schema of the SILAC experiment performed to identify RhoA interacting proteins. Isotope-labelled melanoma cells were lysed followed by incubation with RhoA conjugated to GST-beads or GST-beads alone. Bead samples were subsequently subjected to proteomic analysis by LC-MS/MS. bTop hits from the SILAC experiment in (a)sortedbythedetected heavy:light ratio (H:L) and also listing the number of peptides identified per protein. H:L indicates fold increase in the pulldown of RhoA beads as compared to control beads. cPull-down assay to validate GEFs binding to RhoA detected with SILAC experiment. dRepresentative images for siRNA knock-down of three Rho-GEFs (ARHGEF1, ARHGEF11 and ARHGEF2) in the A375M2 cell line seeded on collagen and ethe corresponding quantification using the Rounding index (0–1). fImmunoblot analyses of pMLC2 and MLC2 levels A375M2 for siRNA knock-down of ARHGEF1, ARHGEF2 and ARHGEF11, respectively. gRepresentative images for siRNA of MYL12B and hthe corresponding quantification using the Rounding index (0–1). ipMLC2 and MLC2 levels in A375M2 cells for siRNA knock-down of MYL12B analysed by immunoblots. jRepresentative images of A375M2 seeded on collagen with or without ROCK inhibition by the inhibitor H1152 and kits respective quantification using the Rounding index (0–1). lImmunblot analysis of pMLC2 corresponding to experiments in (j–k). Scale bars in all micrographs are 15 µm. Error bars are SEM and p-values were calculated using unpaired t-tests; N=3 per cohort. For all panels p-values are <0.05 (*), <0.01 (**), <0.001 (***), or <0.0001 (****) and nonsignificant (ns) as indicated on relevant comparisons. NIS mCherry PSFFV 37 100 kDa anti-mouse GAPDH anti-rabbit mCherry NIS-mCherry –+ f T/S ID/g(%) L S B 0 1 2 3 NIS-mCherry ClO4 − g NIS-mCherry + nuclei ab c d Normalized 99mTcO4 – uptake Scan #1 99mTcO4 – Scan #2 (+48 h) 99mTcO4 – + NaCIO4 WGA-Alexa488 Merge + nucleiNIS-mCherry − − − + + + + − − − 4599 4599.NC Ref e Te I T Lung AxLN 90 0 pMLC2 Haematoxylin/eosin Fig. 3 Characterisation of the new in vivo traceable non-melanogenic murine melanoma cell line 4599.NC. a/top Sketch of the lentiviral radionuclide-fluorescence fusion reporter gene construct. a/bottom Immunoblot analysis of lentivirally transduced and sorted 4599.NC cells compared to parental 4599 cells. bConfocal micrographs showing fusion reporter expression and overlap with the plasma membrane stain wheat germ agglutinin (WGA) conjugated to Alexa488. Representative cells are shown; scale bar =10 µm. cRadionuclide reporter function as quantified by uptake of the radioactive NIS substrate [ 99m Tc]TcO 4 − .‘Ref’indicates a fusion reporter reference cell line as previously described. 26 Specificity of uptake was demonstrated by abolished radiotracer uptake in the presence of the competitive substrate perchlorate; error bars are SD, N=3. d4599.NC cells were intradermally administered to 5-week-old male NSG mice to establish orthotopic tumours (N=4 animals). Three weeks post administration animals were imaged by [ 99m Tc]TcO 4 − -afforded NIS-SPECT/CT clearly indicating cancerous tissues (primary tumour: solid arrow; metastases: dashed arrows) alongside signals stemming from organs expressing NIS endogenously (thyroid and salivary glands (T/G), lachrymal glands (L), stomach (S), and lower in intestine (I) and testes (Te)); none of the latter interfered with the primary tumour or metastases in this model. To assess NIS specificity in vivo, animals received the NIS co-substrate perchlorate intraperitoneally 40 min before animals were re-imaged (48 h after the first imaging session); remaining signals in kidney and bladder (B) reflect radiotracer excretion routes. For corresponding tumour growth curves and ex vivo γ-counting results see Fig. S2. For growth comparison with tumours established from parental 4599 cells see Fig. S2b. eHarvested tumour tissues presented with red fluorescence stemming from reporter expression, which not only guided dissection, but enabled histological assessment of tumour tissues. A typical confocal micrograph of one animal from a cohort of N=4 is shown; scale bar =25 µm. fHematoxilin and eosin staining and (g) phospho-MLC2 immunohistochemistry of adjacent tumour sections from the same tumour as in (e/f); in (f/g) the red dashed line indicated the tumour front while large blood vessels are encircled with purple dots; scale bars =200 µm. A preclinical pipeline to evaluate migrastatics as therapeutic agents in. . . O Maiques et al. 706 competitive NIS substrate ClO 4 − before [ 99m Tc]TcO 4 − injection to visualise cancer cells. ClO 4 − blocked all NIS-associated signals including all tumour signals demonstrating NIS specificity of [ 99m Tc]TcO 4 − signals in vivo (Fig. 3d: compare left/right panels). Ex vivo analyses of radioactivity in harvested tissues corroborated in vivo imaging data (Fig. S2c). Harvested primary tumour tissues were analysed by histology, whereby 4599.NC cells were readily identified based on reporter presence (Fig. 3e); notably, the reporter also showed the expected plasma membrane localisation. Fig. 3f showed expected tumour morphology (by haematoxylin/ eosin staining) and Fig. 3g demonstrated high pMLC2 staining in the same tumour cells. This data demonstrated the suitability of the new 4599.NC melanoma cell line for preclinical in vivo cell tracking studies. a AT13148 GSK269962H1152 Y27632 12 nM 140 nM 300 nM 1.6 nM 4 nM 6 nM 4 nM Inhibitor H HO CI NH2 N N N NH N O HN NN NN N O O O O H2N H2N H NN H O N OO S IC50(ROCK1) IC50(ROCK2) c e b d Concentration Concentration [ μ M] Concentration [ μ M] [ μ M] [ μ M] 0 1.0 0.5 Normalized cell survival 0.0 –3 –2 –1 0 Inhibitor concentration [ μ M] 123 –3 –2 –1 0 Inhibitor concentration [ μ M] 123 –3 –2 –1 0 Inhibitor concentration [ μ M] 123 –3 –2 –1 0 Inhibitor concentration [ μ M] 123 1.0 0.5 Normalized cell survival 0.0 1.0 0.5 Normalized cell survival 0.0 1.0 0.5 Normalized cell survival 0.0 0.5 151050100500 0 0.05 0.1 0.5 1 5 10 50 0 1 5 10 50 100 250 500 Concentration 0.01 0.05 0.1 0.5 1 5 100 AT13148 GSK269962 H1152 Y27632 EC50 = 3.49 μ M [2.72;4.10] 4599.NIS-mCherry 4599 4599.NIS-mCherry 4599 4599.NIS-mCherry 4599 4599.NIS-mCherry 4599 EC50 = 2.92 μ M [2.22;3.61] EC50 = 1.24 μ M [0.97;1.96] EC50 = 0.94 μ M [0.82;1.10] EC50 = 32.3 μ M [24.5;43.1] EC50 = 34.4 μ M [27.3;43.5] EC50 = 0.93 μ M [0.30;0.48] EC50 = 0.39 μ M [0.33;0.46] n.d. Fig. 4 Effects of various ROCK inhibitors on 4599 and 4599.NC melanoma cell survival. a Chemical structures and reported IC 50 values for the listed compounds. Tabled IC 50 values (nM) were previously reported: H1152 63 ; AT13148 64 ; Y27632 65 ; and GSK269962. 66 b–eComparative cell survival assays in the presence of different concentrations of the indicated ROCK inhibitors revealed no differences between 4599 and 4599.NC cells; N=3, error bars represent SD. EC 50 including 95% confidence intervals [lower;higher end of range] were calculated using the variable slope dose-response model using Graphpad Prism v7. Grey shades depict 95% confidence intervals for 4599 cells (black circles/grey non-linear fit line) and red dotted lines depict 95% confidence intervals for 4599.NC cells (red squares/red non-linear fit line). A preclinical pipeline to evaluate migrastatics as therapeutic agents in. . . O Maiques et al. 707