Every breath you take...

A project originally driven by the strictures of lockdown has resulted in a major contribution to our understanding of how air pollution can cause lung cancer. Following a survey of epidemiological data—the only lab work possible at the start of the pandemic—Charlie Swanton’s lab has used a combination of mouse and human studies in a technical tour-de-force that suggests inflammation induced by polluted air is a tumour promoter providing the impetus for mutated lung cells to become cancers [1].

The burden of disease caused by air pollution is now thought to be on a par with or even exceed, other major global health risks such as high blood pressure and smoking. Fine particle pollution (PM2.5), defined as environmental particulate matter measuring less than 2.5mm, is a group 1 carcinogen [2] causing some 4 million early deaths annually. Around 6% of these premature deaths are from Lung Cancer In Never-Smokers (LCINS), a cancer distinct from that seen in smokers which is now the fifth most common cause of cancer-related deaths worldwide [3]. As smoking-related cancer deaths decrease, LCINS, with its different genetic and environmental underpinnings, is likely to assume even more importance in the future, as will the need for tailored screening, diagnostics and therapeutics.

LCINS are almost universally non-small-cell lung cancers, with the majority of cases being lung adenocarcinomas (LUAD). The disease preferentially occurs in women (although much of this bias may reflect smoking biases between males and females), is particularly prevalent in East Asian populations, and has a rather different genetic signature to smoking-related cancers: LCINS are highly enriched for oncogenic alterations such as EGFR mutations, but have low tumour mutational burdens—which likely accounts for their poor response to immune checkpoint inhibitors— and lack tobacco-associated mutational signatures. Epidemiological studies show there is a correlation between LCINS frequency and PM2.5 levels, but it is difficult to draw definitive conclusions from such work: there are too many variables involved, not least in how air pollution is measured, and in accurately quantifying levels of individual exposure. However, the importance of the topic is clear: latest figures estimate there are over 190,000 new air-pollution related lung cancer cases a year worldwide, and a clear upward trend, especially in women [4].

In 2020, lockdown having nixed all chances of doing experiments, the Swanton lab and a worldwide network of collaborators embarked on an epidemiology and informatics binge to look at the association between PM2.5 levels and the particular variety of LCINS involving EGFR mutation. Across nearly 30,000 cases in four within-country cohorts, there was a significant association, confirming the results of previous studies, but reconfirming that only correlation, rather than causation, could be shown by this method.

The correlation seen was sufficient to justify doing some wet-lab work using mouse animal lung cancer models with EGFR mutations to try to pin down more definitively what was going on. There is an array of lung adenocarcinoma (LUAD) mouse models, and Charlie and his colleagues used one of these, in which the mutated human EGFRL858R gene can be activated using intratracheal delivery of an adeno-Cre virus, to ask a very specific question: if mice in which EGFRL858R is switched on (ET mice) are exposed to fine particle air pollution, what is the effect on the developing cancers?

The results were impressive: when ET mice with activated EGFRL858R were exposed for three weeks to fine particulate matter (PM) dissolved in PBS, ten weeks later there was an increase in both the number and size of pre-invasive lung neoplasias relative to PBS controls, and the size of the increase was directly proportional to the concentration of PM. The result was also seen in two other LUAD models: one where EGFRL858R was induced only in lung alveolar type 2 (AT2) cells, and the other where KRasG12D was the oncogenic driver mutation.

These data suggested that PM can promote tumour progression in both oncogenic KRas and EGFR models of LUAD, and that it works both by increasing the number of mutant cells able to form a tumour, and by increasing the proliferation rate within the early lesions.

How might this be happening? Whole-genome-sequencing on tumours from ET mice exposed to either PM or PBS showed there was no significant increase in mutations, suggesting that this was not the route. Strikingly however, if the experiment was conducted in an immune-deficient background, PM had no significant effect, leading to the conclusion that an immune response was necessary.

Inhalation of toxic particles has long been known to induce an inflammatory response in lungs, mediated by macrophages and lung epithelial cells [4]. In both ET mice and controls, an increase in infiltrating alveolar macrophages was evident 24h after the PM inhalation period, and this increase persisted for weeks after exposure, strongly suggesting that transient PM exposure can cause enhanced and sustained lung macrophage infiltration well beyond the period of exposure.

Lung epithelial cells were also changed: RNA-seq analysis of gene expression showed that relative to controls, ET lung epithelium exposed to PM had upregulated the IL-6-JAK-STAT pathway, inflammatory responses and allograft rejection markers, and had specifically switched on genes involved in macrophage recruitment, including the gene for the cytokine interleukin-1b (IL-1b). In tandem, and highly relevant to the neoplastic data, genes expressed in AT2 progenitor cells, the probable cell of origin of lung adenocarcinoma, were also upregulated in the population, suggesting that transcriptional reprogramming to revert cells to an AT2 progenitor-like state could be induced by PM exposure.

To check if the changes in gene expression led to functional changes, lung epithelial cells were isolated from induced versus uninduced ET mice following exposure to PM, and co-cultured with normal lung fibroblasts in a 3D lung-organoid formation assay. Cells in which EGFRL858R had been induced were more efficient at forming organoids than uninduced controls. The result also held if uninduced mice were exposed to PM, and EGFRL858R was only triggered post-mortem in the organoid cultures. Finally, if only AT2 progenitor cells were used to seed the organoid cultures, the same increase in organoid formation was seen in induced versus uninduced controls. Taken together, these data showed that PM exposure causes increased progenitor function, but only when EGFRL858R is switched on in either the lung epithelium as a whole, or specifically in AT2 cells.

Turning to the role of the immune system, the lab focussed on macrophages, as it was known that IL-1b secreted from lung macrophages could stimulate AT2 progenitor cell formation following injury. They started by showing that following PM exposure, IL-1b was indeed switched on in mouse and human lung tissue, predominantly in lung macrophages. They then used the lung organoid formation assay to show that AT2 cells taken from ET mice and induced ex vivo to express EGFRL858R formed larger organoids when cultured with macrophages taken from normal mice exposed to PM. Importantly, when just IL-1b was added, they got the same result. Finally, and most excitingly, back in the PM-induced LUAD model, EGFRL858R was impaired in its induction of tumours in the presence of an anti IL-1b antibody, suggesting a crucial role for this cytokine in the process. 

Mouse models of cancer are all very well, but in human lung cancers, driver mutations occur sporadically, rather than simultaneously in many cells. For these data to be relevant to humans, one would have to propose that in healthy lung epithelium, a few of the ostensibly normal cells carried oncogenic driver mutations. It would be these cells that would be sensitised to the inflammation caused by fine particulate matter and from which a cancer could develop.

This turned out to be the case. Sequencing of non-cancerous lung biopsy tissue showed that in a tiny proportion of histologically normal cells, EGFR or KRAS driver mutations could be detected. There was no association between the presence of these mutations and smoking status or cancer diagnosis; the only significant correlation was with the age of the individuals.

This work provides strong evidence that inflammation plays an early and central role in lung cancer caused by air pollution. This is better news than might be expected. In the alternative model of how air pollution might cause cancer, mutagens in polluted air were proposed to cause sequential accumulation of oncogenic mutations until a tipping point was reached and a malignant lung cancer began to develop. Such a scenario leaves little room for preventive measures beyond constant invasive screening of the lungs of healthy individuals. However, if inflammation is key to jump-starting the transition from normality to disease, this opens the door for preventive anti-inflammatory regimes.

That this is feasible is evidenced in the unexpected results of a 2017 clinical trial [5, 6], in which the anti-IL-1b antibody canakinumab was shown to reduce cardiovascular disease, but also caused a drop in the number of lung cancer primary tumours. Canakinumab does not work on established lung cancers, so this finding suggests that IL-1b is important for initiation and/or evolution of lung cancer in humans, exactly as predicted in the mouse LUAD model. It may actually be possible to prevent two age-related pathologies—cardiovascular disease and lung cancer—by blocking a common upstream inflammatory pathway.

We are not quite there yet, as the sometimes severe immunosuppressive side effects induced by currently available anti-inflammatories limit their use long-term as cancer-preventive drugs, but this sort of public health intervention could be put in place until the long-term problem of exposure to PM caused by external air pollution is tackled. It may also be useful for preventing and treating other cancers whose root cause is inflammation.