A new checkpoint in tumour immunity



  • Tumour-secreted prostaglandin E2 enables tumour growth by preventing NK cells from recruiting the cDC1 subset of dendritic cells to a growing tumour to activate an immune response

  • COX inhibition, and hence prostaglandin E2 ablation, enables immune control to be regained

  • In mouse cancers, COX inhibition synergises with checkpoint blockade therapy to shrink tumours, and this therapy is now in clinical trials

  • In human cancers, a COX inflammatory signature is a highly accurate prognostic indicator, predicting poor outcome in all malignancies tested

The success of immunotherapies, such as those based on immune checkpoint blockade, demonstrates that the immune system is capable of efficiently recognising and eliminating tumours. However, cancers can grow aggressively and invasively despite the presence of a functional immune system, and many patients fail to respond to immune-based therapies. Understanding the basis for this paradox is fundamental for cancer treatment.

The immunoediting theory suggests that the ability of cancers to escape immune surveillance is the result of the immune system selecting for less immunogenic cancer cells. Low immunogenicity as a consequence of lack or loss of tumour antigens, or by the tumour cloaking itself to hide from the adaptive immune system, are well-established mechanisms of immunoediting. However, how a cancer might prevent the triggering of an immune response by influencing the cells of the innate immune system has been less well studied. Work from Caetano Reis e Sousa’s laboratory has shown that dendritic cell activation leading to immune control can be subverted by prostaglandin E2 (PGE2) produced by tumour cells (Zelenay et al, 2015). PGE2 also suppresses accumulation of a dendritic cell subset, cDC1, in tumours, which led the lab to discover that cDC1 infiltration of cancers is driven by chemokines produced by intratumoural innate lymphocytes such as NK cells (Böttcher et al, 2018). Cancer cell traits that suppress or subvert DC activation are likely to contribute to decreased cancer immunogenicity and might constitute attractive therapeutic targets.

To study how the immune system can be subverted by a tumour, Reis e Sousa and colleagues, led by postdoc Santiago Zelenay, chose a cell line established from an autochthonous melanoma mouse model driven by the most common recurring mutation found in melanoma patients, a glutamic acid substitution for valine at position 600 of the B-RAF kinase (BrafV600E). To exist at all, BrafV600E tumours have clearly escaped immune surveillance in the original mouse model, and they also grow well when transplanted into immunocompetent hosts, meaning that they must have key attributes that allow them to develop in the presence of a functioning immune system.

Zelenay and co-workers found that when added to a mixture of dendritic cells and monocyte-derived macrophages, conditioned medium from cultures of BrafV600E cells could alter the inflammatory response, so that instead of suppressing tumour growth, it promoted it. This was due to the melanoma cells secreting PGE2, which is synthesised through a canonical pathway dependent on the activity of the enzymes cyclooxygenase (COX)-1 and -2 and the prostaglandin E synthases (PGES).

PGE2 has been implicated in carcinogenesis before in connection with angiogenesis, cancer cell survival, proliferation and invasion, but its production by the tumour cells themselves seems to be primarily to allow them to escape adaptive immunity. In vivo and in vitro studies showed that cancer cell-derived PGE2 was critical for the growth of BrafV600E tumours, and also for NrasG12D- driven melanoma, colorectal and breast cancer cell lines. Cells unable to produce PGE2 due to CRISPR-mediated ablation of either COX or PGES activity grew very poorly in immunocompetent mice, and the mice that rejected these PGE2-deficient cancer cells were largely resistant to subsequent challenge with parental tumours, showing that they had mounted an effective tumour-specific immune response.

Could the drastic effects of genetically ablating PGE2 production by tumours be mimicked by pharmacological intervention, using COX inhibitors such as aspirin? Mice given aspirin in their drinking water or a second COX inhibitor, celecoxib, were still susceptible to BrafV600E or NrasG12D -driven tumours, but strikingly, there was rapid tumour regression and eradication if COX inhibition was combined with an anti-PD-1 blocking antibody.

This synergistic effect of immune checkpoint blockade and COX inhibition is highly relevant to human cancer, as COX-2 is overexpressed in many tumours, and is associated with poor prognosis and treatment outcome in melanoma. Further, analysis of human and mouse melanomas showed remarkable parallels between high levels of COX-2 and a subversion of the inflammatory response to inhibit effective immunity and create a pro-tumour microenvironment.

Starting with basic questions of how DCs are able to infiltrate tumours to orchestrate immunity, our findings have suggested therapeutic avenues, some of which are currently being tested in clinical trials. The collaborations we established with other labs at the Crick and elsewhere were critical to the success of our investigations.

This is yet another example of how carrying out blue skies research in a multidisciplinary environment can potentially lead to applications to human health.”

Caetano Reis e Sousa

In 2018, Reis e Sousa and colleagues extended this work to show how PGE2 production by tumours elicits such profound effects on cancer immune control (Böttcher et al, 2018). The tumour microenvironment contains stromal and immune cells that shape cancer development and also affect therapeutic responses. Intratumoural immune cells comprise T, B and NK lymphocytes, and several types of myeloid cells, including dendritic cells. Dendritic cells, especially the cDC1 subfamily, are essential for anti-cancer responses, and their abundance within tumours is associated with immune-mediated rejection and the success of immunotherapy.

The group began by confirming an observation from their previous paper, that in mice, loss of COX activity in BrafV600E melanomas, and hence loss of secreted PGE2, resulted in an early influx into the tumours of activated cDC1, which were absent from COX-competent tumours. This was also the case when COX activity was ablated in mice injected with two other PGE2- producing tumours, the colorectal cancer line CT26 and the breast cancer line 4T1. cDC1 recruitment was important, as mice lacking cDC1 were unable to mount an immune response to prevent COX-deficient tumour growth.

How might tumour-derived PGE2 impair the recruitment of cDC1 to the tumour microenvironment? In COX-deficient BrafV600E melanomas, there was a prominent early rise in conventional NK cells, clustered within the tumour in close apposition to cDC1. Depletion of the NK cells showed they were required for cDC1 accumulation. Therefore, for tumour growth to be inhibited, NK cells must enter the tumour and attract and position cDC1 within it, and PGE2 is able to prevent this happening.

NK cells were already known to play a key role in anti-tumour immunity, as they can directly kill tumour cells and produce anti-tumour cytokines, but this ability to recruit cDC1 was hitherto unknown. By analysis of existing gene expression data, NK cells were shown to express six chemokines; of these, XCL1 and CCL5 were highly overexpressed in the tumour microenvironment of tumours lacking PGE2. In vitro, in the presence of PGE2, expression of both cytokines was inhibited, and NK survival was markedly reduced.

Implantation of COX-deficient tumours into mice treated with neutralising antibodies against XCL1 and CCL5 resulted in far fewer cDC1 colonising the tumours, suggesting that XCL1 and CCL5 were necessary for cDC1 recruitment. They were also sufficient: if COX-deficient tumours were engineered to express CCL5 or XCL1, there was accelerated rejection compared to controls, except in mice lacking cDC1, where all tumour types grew strongly.

In a belt-and-braces approach, PGE2 impairs CCL5 and XCL1 signalling not just by affecting NK cells: it also forces cDC1 cells to downregulate their expression of the chemokines’ cognate receptors CCR5 and XCR1, resulting in an impairment of the ability of cDC1 to migrate towards a source of CCL5 and XCL1.

In the most recent paper on the topic, Santiago Zelenay, now a lab head at the CRUK Manchester Institute, collaborated with Reis e Sousa and others to slot the final piece of the puzzle in place: PGE2 exerts its effects on NK cells via their expression of its receptors EP2 and EP4, and in mice in which the cells lack both EP2 and EP4, even COX-competent tumours are unable to grow (Bonavita et al, 2020).

More content

It is clear from this work that in mice, recruitment of cDC1 by NK cells to tumours results in tumour regression, presumably due to efficient activation of the adaptive immune system. Importantly, these findings are also likely to be true in humans. Analysis of human tumour gene expression datasets shows that there is a high positive correlation between the NK, cDC1, cytotoxic T cell and XCL1, XCL2 and CCL5 chemokine gene signatures, and that across a wide array of human tumours, those with the highest NK and cDC1 content have the best prognosis (Böttcher et al, 2018). Conversely, those tumours with a high COX inflammatory signature had the worse outcome in all malignancies tested (Bonavita et al, 2020).

Comparison of COX-deficient and COX-sufficient tumours


These results uncover a new checkpoint in immunity that could form the basis for novel cancer immunotherapies, and indeed, since the publication of the 2015 study, eight clinical trials have begun, looking at the effectiveness of a combination of checkpoint inhibitors and aspirin in a spectrum of advanced cancers. Locally stimulating intratumoural NK cells or developing ligands to attract cDC1 into the tumour microenvironment could be an attractive means of eliciting anti-tumour immunity and increasing the response rate to immunotherapy. Data from human studies suggest that a low frequency of cDC1 in tumours may be one reason for the low response rate of cancer patients to immune checkpoint blockade. At the very least, therefore, cDC1 accumulation may be useful as a predictive biomarker for the outcome of such treatments.