LAUSR

The role of the lysyl oxidase (LOX) family of enzymes, consisting of LOX and LOXlike (LOXL) proteins 1–4, is to catalyse the crosslinking of collagen and elastin in the extracellular matrix (ECM). Members of the family have been implicated in tumour progression in a range of cancers, with multiple studies suggesting that LOX/ LOXLmediated ECM remodelling facilitates tumour cell invasion and metastasis. Increased expression, particularly of LOX and LOXL2, has been noted in many aggressive cancers and has been linked to reduced survival. However, the mechanisms of action are still unclear, and there is an ongoing debate over whether expression is tumour supporting or tumour suppressive. 2 This appears, to some degree at least, to be dependent on tumour site, stage and even the source of expression, highlighting the need to understand the cell and stagespecific consequences of LOX/LOXL activity. Pancreatic ductal adenocarcinoma (PDAC) is notable for the abundance of collagen in the tumour microenvironment (TME), and as such has been the focus of several studies examining LOX or LOXL2 activity. However, the conflicting results reported have raised as many questions as answers. Indeed, more generally, there has been controversy over the stroma and ECM in PDAC, and although they clearly influence disease initiation and aggressiveness, they can exhibit tumoursuppressive activity in some contexts. In this issue, AlonsoNocelo et al sought to address some of these questions, investigating the role of LOXL2 in pancreatic tumourigenesis using model systems. The authors exploited genetically engineered mouse models of loss or gain of LOXL2 expression specifically within the pancreatic epithelium to demonstrate that overexpression of LOXL2 promotes tumour initiation and metastatic capacity, potentially through increased epithelial–mesenchymal transition (EMT) and stemness. Consistent with this, they found that loss of LOXL2 decreased metastasis and increased survival, primarily due to significantly less organised ECM, and reduced stiffness and mechanosignalling. Importantly, human datasets support these findings, with higher LOXL2 expression associated with decreased survival and enrichment for an EMT signature, suggesting that evaluation of LOXL2 targeting in the metastatic setting may be warranted. Given the ongoing debate around LOX and LOXL activity and the role of ECM in pancreatic tumour progression, these are important findings. Recent studies have reported conflicting data to those presented here and suggested that both collagen 1 and LOXL2 suppress PDAC progression. In one study, Jiang et al found that low stromal content correlated with poor prognosis in human PDAC, and using an orthotopic syngeneic transplant model, they showed that ECM ablation using an antiLOXL2 antibody promoted PDAC progression. In this setting, the tumour cells themselves had higher proliferative capacity, but there were no significant changes to cancerassociated fibroblast (CAF) or immune cell populations, or tumour vasculature. In an analogous study, Chen et al demonstrated that deletion of collagen 1 from alpha smooth muscle actin cells in an autochthonous model led to a reduction in stromal collagen 1 content and decreased stiffness, but a more poorly differentiated and invasive phenotype resulting in accelerated disease progression. They observed myeloid cell accumulation in tumours and an ultimately immunesuppressive environment, which was important for the aggressive tumour phenotype and reduced survival since myeloid targeting therapies reversed this effect. There are a few differences to highlight here. In the studies by AlonsoNocelo et al and Chen et al, LOXL2 or collagen 1 was deleted from epithelial cells or myofibroblasts, respectively, in contrast to the systemic therapeutic approach taken by Jiang et al. Interestingly, proteomic analyses have shown that while the majority of ECM proteins originate from stromal cells, tumour cellderived ECM proteins are more clearly associated with poor prognosis. Thus, the cell type targeted may influence outcomes in preclinical studies. Both AlonsoNocelo et al and Chen et al also noted changes to the immune TME, which may impact on the phenotypes observed. These studies both used autochthonous models in which tumour progression and TME establishment occur spontaneously and over a longer time, while Jiang et al used a transplant model. Clearly, the choice of model system may influence study outcomes, a point underscored by the lack of LOXL2 expression in many PDAC cell lines. Finally, in both the AlonsoNocelo et al’s and the Chen et al’s study, deletion of LOXL2 or collagen 1 occurred from birth, prior to tumour development, whereas, in the antiLOXL2 therapeutic study conducted by Jiang et al, treatment commenced posttransplant. Again, the timing of treatment could significantly influence outcomes, and the effects of LOXL2 inhibition in established tumours in the autochthonous setting remain to be seen. Indeed, in the clinic, a trial of the LOXL2 blocking antibody, simtuzumab, in combination with gemcitabine, failed to improve survival in patients with metastatic PDAC, although it is worth noting that this may not be the most appropriate disease setting. Over the past few years, several studies have proposed targeting the stroma as a potential therapeutic strategy in pancreatic cancer; however, despite many of these reaching the clinic, results have been disappointing. More recently, evidence has emerged to suggest that the stroma can have tumourrestricting potential, with complete ablation resulting in tumour promotion. Thus, the consensus is now that more nuanced targeting of signalling within the stroma is more likely to yield clinical benefit. The prospects for more selective approaches have been enhanced by the description of different cellular subtypes within the PDAC microenvironment, notably subpopulations of CAFs, which can exhibit both tumourpromoting versus restrictive behaviour. Interestingly, it is not only the CAFs that exhibit functional heterogeneity in the PDAC microenvironment. Many studies have highlighted a tumourpromoting role for tumour associated macrophages (TAMs) in PDAC, but these too display heterogeneity and can have distinct profiles. In PDAC, TAMs can originate from both the yolk sac and the bone marrow; however, the yolk sacderived Cancer Research UK Beatson Institute, Glasgow, UK Institute of Cancer Sciences, University of Glasgow, Glasgow, UK