Research Article15 June 2015Open Access Source Data Targeting the LOX/hypoxia axis reverses many of the features that make pancreatic cancer deadly: inhibition of LOX abrogates metastasis and enhances drug efficacy Bryan W Miller Bryan W Miller Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Search for more papers by this author Jennifer P Morton Jennifer P Morton Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Search for more papers by this author Mark Pinese Mark Pinese The Garvan Institute of Medical Research, Sydney, NSW, Australia Search for more papers by this author Grazia Saturno Grazia Saturno Cancer Research UK Manchester Institute, Withington, Manchester, UK Search for more papers by this author Nigel B Jamieson Nigel B Jamieson West of Scotland Pancreatic Unit, Glasgow Royal Infirmary, Glasgow, UK Search for more papers by this author Ewan McGhee Ewan McGhee Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Search for more papers by this author Paul Timpson Paul Timpson The Garvan Institute of Medical Research, Sydney, NSW, Australia Search for more papers by this author Joshua Leach Joshua Leach Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Search for more papers by this author Lynn McGarry Lynn McGarry Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Search for more papers by this author Emma Shanks Emma Shanks Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Search for more papers by this author Peter Bailey Peter Bailey Institute of Cancer Sciences, University of Glasgow, Garscube Estate, Glasgow, UK Search for more papers by this author David Chang David Chang Institute of Cancer Sciences, University of Glasgow, Garscube Estate, Glasgow, UK Search for more papers by this author Karin Oien Karin Oien Institute of Cancer Sciences, University of Glasgow, Garscube Estate, Glasgow, UK Search for more papers by this author Saadia Karim Saadia Karim Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Search for more papers by this author Amy Au Amy Au Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Search for more papers by this author Colin Steele Colin Steele Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Search for more papers by this author Christopher Ross Carter Christopher Ross Carter West of Scotland Pancreatic Unit, Glasgow Royal Infirmary, Glasgow, UK Search for more papers by this author Colin McKay Colin McKay West of Scotland Pancreatic Unit, Glasgow Royal Infirmary, Glasgow, UK Search for more papers by this author Kurt Anderson Kurt Anderson Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Search for more papers by this author Thomas R Jeffry Evans Thomas R Jeffry Evans Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Institute of Cancer Sciences, University of Glasgow, Garscube Estate, Glasgow, UK Search for more papers by this author Richard Marais Richard Marais Cancer Research UK Manchester Institute, Withington, Manchester, UK Search for more papers by this author Caroline Springer Caroline Springer Institute of Cancer Research, London, UK Search for more papers by this author Andrew Biankin Andrew Biankin Institute of Cancer Sciences, University of Glasgow, Garscube Estate, Glasgow, UK Search for more papers by this author Janine T Erler Corresponding Author Janine T Erler Biotech Research & Innovation Centre (BRIC), University of Copenhagen, Copenhagen (UCPH), Denmark Search for more papers by this author Owen J Sansom Corresponding Author Owen J Sansom Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Search for more papers by this author Bryan W Miller Bryan W Miller Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Search for more papers by this author Jennifer P Morton Jennifer P Morton Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Search for more papers by this author Mark Pinese Mark Pinese The Garvan Institute of Medical Research, Sydney, NSW, Australia Search for more papers by this author Grazia Saturno Grazia Saturno Cancer Research UK Manchester Institute, Withington, Manchester, UK Search for more papers by this author Nigel B Jamieson Nigel B Jamieson West of Scotland Pancreatic Unit, Glasgow Royal Infirmary, Glasgow, UK Search for more papers by this author Ewan McGhee Ewan McGhee Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Search for more papers by this author Paul Timpson Paul Timpson The Garvan Institute of Medical Research, Sydney, NSW, Australia Search for more papers by this author Joshua Leach Joshua Leach Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Search for more papers by this author Lynn McGarry Lynn McGarry Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Search for more papers by this author Emma Shanks Emma Shanks Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Search for more papers by this author Peter Bailey Peter Bailey Institute of Cancer Sciences, University of Glasgow, Garscube Estate, Glasgow, UK Search for more papers by this author David Chang David Chang Institute of Cancer Sciences, University of Glasgow, Garscube Estate, Glasgow, UK Search for more papers by this author Karin Oien Karin Oien Institute of Cancer Sciences, University of Glasgow, Garscube Estate, Glasgow, UK Search for more papers by this author Saadia Karim Saadia Karim Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Search for more papers by this author Amy Au Amy Au Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Search for more papers by this author Colin Steele Colin Steele Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Search for more papers by this author Christopher Ross Carter Christopher Ross Carter West of Scotland Pancreatic Unit, Glasgow Royal Infirmary, Glasgow, UK Search for more papers by this author Colin McKay Colin McKay West of Scotland Pancreatic Unit, Glasgow Royal Infirmary, Glasgow, UK Search for more papers by this author Kurt Anderson Kurt Anderson Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Search for more papers by this author Thomas R Jeffry Evans Thomas R Jeffry Evans Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Institute of Cancer Sciences, University of Glasgow, Garscube Estate, Glasgow, UK Search for more papers by this author Richard Marais Richard Marais Cancer Research UK Manchester Institute, Withington, Manchester, UK Search for more papers by this author Caroline Springer Caroline Springer Institute of Cancer Research, London, UK Search for more papers by this author Andrew Biankin Andrew Biankin Institute of Cancer Sciences, University of Glasgow, Garscube Estate, Glasgow, UK Search for more papers by this author Janine T Erler Corresponding Author Janine T Erler Biotech Research & Innovation Centre (BRIC), University of Copenhagen, Copenhagen (UCPH), Denmark Search for more papers by this author Owen J Sansom Corresponding Author Owen J Sansom Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK Search for more papers by this author Author Information Bryan W Miller1,‡, Jennifer P Morton1,‡, Mark Pinese2,‡, Grazia Saturno3,‡, Nigel B Jamieson4, Ewan McGhee1, Paul Timpson2, Joshua Leach1, Lynn McGarry1, Emma Shanks1, Peter Bailey5, David Chang5, Karin Oien5, Saadia Karim1, Amy Au1, Colin Steele1, Christopher Ross Carter4, Colin McKay4, Kurt Anderson1, Thomas R Jeffry Evans1,5, Richard Marais3, Caroline Springer6,‡, Andrew Biankin5,‡, Janine T Erler 7,‡ and Owen J Sansom 1,‡ 1Cancer Research UK Beatson Institute, Garscube Estate, Glasgow, UK 2The Garvan Institute of Medical Research, Sydney, NSW, Australia 3Cancer Research UK Manchester Institute, Withington, Manchester, UK 4West of Scotland Pancreatic Unit, Glasgow Royal Infirmary, Glasgow, UK 5Institute of Cancer Sciences, University of Glasgow, Garscube Estate, Glasgow, UK 6Institute of Cancer Research, London, UK 7Biotech Research & Innovation Centre (BRIC), University of Copenhagen, Copenhagen (UCPH), Denmark ‡These authors contributed equally to this work *Corresponding author. Tel: +45 3532 5666; Fax: +45 3532 5669; E-mail: [email protected] *Corresponding author. Tel: +44 141 330 3656; Fax: +44 141 942 6521; E-mail: [email protected] EMBO Mol Med (2015)7:1063-1076https://doi.org/10.15252/emmm.201404827 PDFDownload PDF of article text and main figures. Peer ReviewDownload a summary of the editorial decision process including editorial decision letters, reviewer comments and author responses to feedback. ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions Figures & Info Abstract Pancreatic ductal adenocarcinoma (PDAC) is one of the leading causes of cancer-related mortality. Despite significant advances made in the treatment of other cancers, current chemotherapies offer little survival benefit in this disease. Pancreaticoduodenectomy offers patients the possibility of a cure, but most will die of recurrent or metastatic disease. Hence, preventing metastatic disease in these patients would be of significant benefit. Using principal component analysis (PCA), we identified a LOX/hypoxia signature associated with poor patient survival in resectable patients. We found that LOX expression is upregulated in metastatic tumors from Pdx1-Cre KrasG12D/+ Trp53R172H/+ (KPC) mice and that inhibition of LOX in these mice suppressed metastasis. Mechanistically, LOX inhibition suppressed both migration and invasion of KPC cells. LOX inhibition also synergized with gemcitabine to kill tumors and significantly prolonged tumor-free survival in KPC mice with early-stage tumors. This was associated with stromal alterations, including increased vasculature and decreased fibrillar collagen, and increased infiltration of macrophages and neutrophils into tumors. Therefore, LOX inhibition is able to reverse many of the features that make PDAC inherently refractory to conventional therapies and targeting LOX could improve outcome in surgically resectable disease. Synopsis Lysyl oxidase (LOX) is identified as a therapeutic target in pancreatic ductal adenocarcinoma (PDAC). Inhibition of LOX resulted in increased drug efficacy and stromal changes and reduction in metastasis. A signature of hazardous and protective genes in PDAC was defined. High expression of hypoxia-associated genes, including LOX, was associated with poor patient prognosis. Using transgenic mouse models of PDAC, LOX was found to be overexpressed in metastatic disease and its expression was required for PDAC cell invasion. Inhibition of LOX in transgenic mice inhibited metastasis, while combination therapy with LOX inhibition and gemcitabine induced stromal alterations, immune cell infiltration and tumor necrosis and improved survival. Introduction Pancreatic cancer is one of the leading causes of cancer-related death in the UK with around 7,000 cases being diagnosed every year (Mukherjee et al, 2008). Pancreatic ductal adenocarcinoma (PDAC) is almost universally lethal. Aggressive invasion and early metastases are characteristic of the disease, such that 80–90% of patients have surgically unresectable disease at the time of diagnosis (Giovinazzo et al, 2012). The majority of patients who are selected to undergo potentially curative resection for small, localized lesions almost inevitably develop recurrent or metastatic disease (Yeo et al, 2002), presumably due to the presence of undetected micro-metastases at initial diagnosis. Adjuvant (post-operative) chemotherapy can improve outcome despite the modest anti-tumor efficacy of these agents (Neoptolemos et al, 2004; Stocken et al, 2005). Nevertheless, overall survival remains disappointing with most patients developing local recurrence or extra-pancreatic metastasis within 2 years (Giovinazzo et al, 2012). Hence, there is a requirement for biomarkers to predict/prognosticate those patients that do not succumb to early recurrence post-resection and identify potentially targetable pathways that may be associated with poor prognosis. This has been made even more important by recent sequencing studies of pancreatic cancer, which have found PDAC to be very complex and contain multiple low-frequency mutations of unknown functional significance. Moreover, none of the major mutations (KRAS, TP53, SMAD4 and INK4A) currently confer treatment opportunities. Importantly, these surgically resectable patients who die of recurrent metastatic disease represent a set of patients where suppression of metastasis (either prevention or suppression of growth of micro-metastasis) may be a realistic therapeutic option and trials are underway to test whether SRC inhibitors may be beneficial in this patient set. A characteristic feature of PDAC is the highly desmoplastic stromal microenvironment. This stroma consists of immune cells, collagen and fibronectin laid down by fibroblasts and stellate cells (Chu et al, 2007). In addition to contributing to disease progression, and promoting tumor growth and invasion, recent studies suggest that the stroma also limits drug delivery to tumor cells, partly explaining the profound resistance of pancreatic tumors to systemic chemotherapy agents (Olive et al, 2009). This situation is made worse by a very poor tumor vasculature which may also limit drug access to tumors (Olive et al, 2009). Consistent with this, a number of studies have shown that potential stromal markers such as S100A2 and A4 have prognostic value in patients with resected PDAC (Jamieson et al, 2011). More recently, there has been great interest in the role of the enzyme lysyl oxidase (LOX) in driving metastasis in epithelial cancers such as breast and colorectal cancer (Payne et al, 2005; Baker et al, 2011; Cox & Erler, 2013). LOX is a copper-dependent enzyme which cross-links collagen and elastins to drive tissue stiffness (Barker et al, 2012). It has previously been shown to be induced by hypoxia-inducible factor 1α (HIF1α) and can function upstream of the SRC/FAK tyrosine kinases (Payne et al, 2005; Erler et al, 2006; Baker et al, 2011, 2013). These studies suggest that LOX inhibition will have the potential to suppress metastasis rather than causing regression of established tumors (Cox et al, 2013; Cox & Erler, 2014), consistent with observations using SRC inhibitors in cancer cell lines and mouse models (Morton et al, 2010a). Additionally, LOX has been reported to have direct effects on cancer cells themselves through regulation of senescence (Wiel et al, 2013). There is a plethora of evidence that inflammation is tumor promoting in pancreatic cancer. Pancreatitis leads to increased risk of PDAC, while cerulein promotes acinar-to-ductal metaplasia and disease progression (Guerra et al, 2007; Rhim et al, 2012). However, it is also clear that the presence of leukocytes such as neutrophils within tumors is often associated with a good prognosis in many different cancer types including pancreatic cancer (Caruso et al, 2002; Schaider et al, 2003; Jamieson et al, 2012). Thus far, there are no mechanistic data to explain this phenomenon, nor is it known how the desmoplastic "stiff" stroma would affect the ability of leukocytes to penetrate tumor tissue. Given the complex interplay between the tumor and the stroma, modeling therapies in these systems require immunocompetent mice that develop tumors that closely recapitulate human disease. Genetically engineered mice carrying the common mutations that occur in PDAC rapidly generate invasive and metastatic disease (Morton et al, 2010b). Therefore, these can serve as excellent models in which to test therapies aimed at targeting the stroma and assess the impact of the mutations that occur in pancreatic cancer upon the stroma, invasion and metastasis. Our and others' previous studies have shown that the accumulating mutant p53 (p53R172H) has gain-of-function properties over loss-of-function mutations (Olive et al, 2004; Jackson et al, 2005; Adorno et al, 2009; Morton et al, 2010b). Consistent with this, we and others have previously found that mice expressing p53R172H display a far higher frequency of liver metastasis than mice with loss of p53 function (Morton et al, 2010b; Weissmueller et al, 2014). However, it has been reported that mice carrying loss of function of p53 and INK4 loss alongside KrasG12D targeted to the pancreas develop adenocarcinoma that metastasizes to the liver (Bardeesy et al, 2006). As yet, most work has focused on how mutant p53 facilitates cell autonomous migration and invasion and has not examined whether it affects tumor–stromal interactions. While the majority of studies had suggested that targeting the stroma may be a promising therapeutic option in this disease, recent studies have shown dramatically different results. Here, targeted sonic hedgehog depletion or myofibroblast disruption has reported decreased survival in mouse models (Lee et al, 2014; Ozdemir et al, 2014; Rhim et al, 2014).These studies suggest that thinking of the tumor stroma simply as a tumor-promoting entity needs to be re-evaluated and it will be very important to understand which elements of the stroma drive the chemoresistance and early progression of pancreatic cancer, and which act to constrain the tumor. It should be noted that these studies also showed that subsequent co-targeting with either anti-angiogenesis or immunotherapy could then lead to significant tumor regression. In this study, we identify LOX, driven by mutant p53, as an important therapeutic target in pancreatic cancer, inhibition of which causes tumor necrosis in combination with gemcitabine. Importantly, the LOX–hypoxia axis defines the poor prognosis of surgically resectable cancers and can explain many of the features that contribute to making PDAC inherently refractory to conventional cytotoxic chemotherapy. Most importantly, inhibition of LOX can reverse most of these features, suggesting that this may be a promising therapeutic approach. Results LOX/hypoxia marks poor prognosis in patients with resected PDAC To elucidate the key components that contribute to poor prognosis in patients with PDAC, a gene signature analysis was performed on the transcriptome of 73 PDACs (Biankin et al, 2012). From this, a single signature was identified (called PC-1) that following cross-validation could predict survival in the discovery cohort (Fig 1A). PC-1 could be split into two gene sets: a hazardous set of 321 transcripts, for which increased expression was associated with poor prognosis, and a protective set of 238 transcripts, for which increased expression was beneficial. To identify potential biological processes underlying the survival signature, we tested the hazardous and protective sets of PC-1 for overlap against the MSigDB gene set database (Subramanian et al, 2005). We observed that the hazardous subset of PC-1 showed significant overlap with numerous hypoxia signatures, and the protective subset was linked to lymphocyte and antigen presentation signatures (Supplementary Table S1). We confirmed that high expression of a hypoxia signature correlated with poor prognosis (Fig 1B). From this hypoxia signature, we selected LOX for further analysis given we wished to investigate factors that could modulate the stroma that are targetable, its previous association with the mesenchymal subtype and reports of its involvement in metastasis in breast and colon cancer (Payne et al, 2005; Baker et al, 2011). We examined overall survival in terms of the expression of LOX using an expanded APGI cohort of 266 patients (Chou et al, 2013). Expression of LOX correlated with patient survival (Fig 1C), and we confirmed this correlation using multivariate analysis (Supplementary Table S2). We next looked in our Glasgow cohort of patients where 47 patients have been profiled and once again high LOX expression correlated with a poor prognosis (Fig 1D). Moreover, LOX was 1 out of only 5 probe sets that overlapped between the Glasgow (Jamieson et al, 2011) cohort and the Collisson (Collisson et al, 2011) signatures of poor prognosis in pancreatic cancer (the others being S100A2, TWIST, NT5E and PAPPA) (Supplementary Fig S1). Additionally, high expression of further LOX family members (LOXL1, 2, 3 and 4) were found to be associated with poor prognosis in the Glasgow data set (Supplementary Fig S1B–E). Taken together, and given the reproducibility across multiple patient cohorts, these findings argued for a significant role for LOX expression as a determinant of poor survival in pancreatic cancer. Figure 1. LOX/hypoxia marks poor prognosis in resectable PDAC Kaplan–Meier analysis showing cases from the Australian cohort (n = 73) delineated on the basis of expression of a set of hazardous genes, termed PC-1. Patients with the highest expression had a significantly poorer prognosis (red line, median survival: 11.6 months) compared with those patients with the lowest expression (blue line, median survival: 34.4 months, P = 0.01). The black line shows those with medium expression. Kaplan–Meier analysis showing that cases in the Australian cohort that fell in the highest quartile of a hypoxia signature (red line, median survival: 11.4 months) have significantly decreased survival compared with those in the lowest quartile (blue line, median survival: 25.2 months; P = 0.03). Kaplan–Meier analysis showing that cases in the APGI cohort (n = 266) with high LOX expression above the 3rd quantile (red line) have significantly decreased survival compared to those with low expression below the 1st quantile (blue line). Kaplan–Meier analysis showing that cases in the Glasgow cohort with high LOX expression (red line, n = 24) have significantly decreased survival compared with those with low expression (blue line, n = 23; P = 0.005). Kaplan–Meier analysis showing that fibrillar collagen is significantly associated with reduced survival in human PDAC (20 months vs. 28.3 months, P = 0.033). Mean decay distance of the second harmonic generation (SHG) signal emitted by human PDAC-associated collagen. Mean decay distance is represented by boxplots showing the second and third quartile of the data with the whiskers indicating the maximum and minimum data points. Outliers are indicated by individual markers. Stage T3 tumors have a higher collagen mean decay distance score compared with Stage T2 tumors (P = 0.02). Lymph node-positive tumors have a higher collagen mean decay distance score vs. lymph node-negative tumors (P = 0.047). Tumors showing vascular invasion have a higher collagen mean decay distance score than non-vascular invasive tumors (P = 0.037). Download figure Download PowerPoint As LOX is associated with regulation of collagen cross-linking, we assessed the status of fibrillar collagen in human PDAC. Using multiphoton microscopy, we analyzed the second harmonic resonance signal of collagen in a human pancreatic tissue microarray (TMA) consisting of 80 patients. Second harmonic generation (SHG) imaging of collagen has been used previously to identify collagen "signatures" in mammary tumors, which may impact on the ability of cancer cell migration (Provenzano et al, 2006, 2009). In our study, we found that fibrillar collagen was significantly associated with reduced survival (20 months vs. 28.3 months, P = 0.033) (Fig 1E). Further, we showed that fibrillar collagen was significantly associated with increased tumor stage, lymph node spread and vascular invasion (Fig 1F). This suggested that the generation of fibrillar collagen may be important in PDAC disease progression and that the inhibition of cross-linking needed to generate collagen fibers via LOX inhibition may be important therapeutically. In order to further investigate this, we used immunocompetent, autochthonous genetically engineered mouse models of PDAC, which are initiated by targeting Kras mutation to the murine pancreas using the Pdx1-Cre transgene. Without the addition of cooperating mutations, only one-third of Pdx1-Cre KrasG12D/+ mice develop PDAC by 500 days (Hingorani et al, 2003). However, when p53 is additionally targeted by deletion or mutation, mice rapidly develop invasive adenocarcinoma with a median latency of between 120 and 180 days (Hingorani et al, 2005) and we have previously found that p53 mutation, but not loss, could drive metastasis in this model (Morton et al, 2010b). In order to characterize the gene expression changes that underlie metastatic disease, we carried out microarray analysis of tumors driven by Kras mutation and mutation of one copy of p53 or deletion of p53. To address whether transcriptional changes observed in pancreatic cancer patients were reflected in our mouse models, we applied loading values obtained from the PC-1 human tumor signature to our mouse microarray data sets (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE67358). This allowed us to derive risk scores for each individual mouse transcriptome. While the majority of mice that had non-metastatic tumors had negative risk scores, 100% of KPC mice that developed metastatic disease had positive risk scores (Fig 2A, top). Of the gene expression changes observed between metastatic and non-metastatic tumors, multiple members of the LOX family were overexpressed in metastatic disease (Fig 2A, bottom). Figure 2. LOX expression is required for invasion in a mutant p53-driven model of PDAC PC-1 signature is predictive of metastatic disease in mouse models of pancreatic cancer. Log-transformed expressions of signature transcripts from mouse tumor microarrays were mean-centered across samples and scaled to unit variance. These values were then multiplied by the matching loading values from the PC-1 signature and summarized for each sample across all transcripts to yield the risk score for that sample. Inverted invasion assays were performed with PDAC tumor cell lines from KPC and KPflC mice. Tumor cell lines bearing mutant p53R172H (KPC) invade significantly further than tumor cells with deletion of 1 copy of p53 (KPflC) (P ≤ 0.01). Data are shown as the average of four wells + SEM. Introduction of shRNA targeting Lox into KPC tumor cells significantly inhibits invasion (P ≤ 0.01 by unpaired Student's t-test). Data are shown as the average of four wells + SEM. Introduction of exogenous LOX into KPflC tumor cells significantly promotes invasion (left panel, P ≤ 0.01 by unpaired Student's t-test). LOX expression was assessed by immunoblotting (right panel). Columns indicate the mean of four well and error bars indicate SEM. Integration of heterogeneous data sets identifies LOX as a therapeutic target in pancreatic cancer. Components of the hazardous PC-1 signature are overlaid with genes found to be overexpressed in a microarray and hits that cause a reduction in viable cell number in an RNAi functional screen. Overlap is shown as a proportional Venn diagram. Source data are available online for this figure. Source Data for Figure 2 [emmm201404827-sup-0012-SDataFig2.pdf] Download figure Download PowerPoint LOX is required for mutant p53-driven invasion Given that our data suggest a role for LOX in mutant p53-driven metastatic PDAC, we decided to address the consequences for cell migration and invasion, of manipulating LOX expression in PDAC cell models. For this, we used both KPC cell lines and cells derived from Pdx1-Cre KrasG12D/+ p53flox/+ (KPflC) tumors. Initially, to assess whether LOX expression was required for invasion, we used siRNA to knockdown LOX expression and measured migration of KPC cells through a Matrigel matrix (Supplementary Fig S2A). Loss of LOX expression significantly impaired KPC cell migration, and this could be rescued by recombinant LOX protein. Inhibition of LOX in KPC cells led to a reduction in SRC phosphorylation (Supplementary Fig S2B), and we confirmed that low doses of the SRC inhibitor dasatinib resulted in a slowing of wound healing (Supplementary Fig S2C). Given our previous data showing a requirement for SRC in KPC cell invasion (Morton et al, 2010a), this suggests that the pro-migratory effects of LOX could be mediated at least in part by SRC activation. We confirmed that LOX knockdown in human Panc-1 cells had similar effects on migration (Supplementary Fig S2D). We have previously shown that the ability of mutant p53 cells (and not p53-deleted cells) to invade in vitro correlates well with the ability of these cells to metastasize in vivo (Fig 2B). Analysis of the expression of LOX family members in these cell lines showed a clear increase in the expression of LOX and LOX family members in cell lines carrying mutant but not loss-of-function p53, consistent with our microarray analyses (Supplementary Fig S2E). Thus, we carried out stable knockdown of LOX in KPC cells using shRNA and found this significantly reduced invasion of mutant p53 cells (Fig 2C; Supplementary Fig S2F). Conversely, overexpression of LOX in p53 loss-of-function tumor cells promoted invasion of these cells (Fig 2D). In order to determine the effects of the modulation of LOX expression in these cell models in vivo, we carried out a series of allograft experiments. LOX knockdown slowed allograft growth of KPC cells (Supplementary Fig S2G), while LOX overexpression resulted in faster growth of KPflC cells (Supplementary Fig S2H). In parallel, we also performed a functional screen on primary cells from Pdx1-Cre KrasG12D p53R172H (KPC) metastatic tumors. Overall, we found that 920 siRNAs reduced cell viability using a Z-score threshold of -3 (Supplementary Table S3). A number of these screen hits were extracellular matrix components, including LOX, fi
