The Nobel-prize-winning discovery of using antibodies to block immune checkpoints, thus allowing immune cells to attack and kill cancer cells, has revolutionized the way we treat cancer. Curiously, immune checkpoint therapy (ICT) can complete regression of tumours in some patients, but no response in others. It is not well understood what biological processes contribute to an effective response to ICB, as the anti-tumour immune response is a complex system, with great variability between cancer types and each patient. Importantly, the immune response is dynamic, and constantly adjusting, suggesting time-dependent mechanisms may underpin an effective response. It has not been possible to identify dynamic events in human studies due to the difficulty to sample the same tumour at multiple time-points during treatment.
We developed dual-tumour mouse models where ICT either leads to a response or a failure to respond in both tumours, allowing us to take one tumour during ICT whilst determining the response from the remaining tumour. In order to identify the biological process which occur in tumours over the course of ICT, we mapped the gene expression of 144 responding and non-responding tumours from two mouse models at four time-points, using bulk and single-cell RNAseq.
We found that responding tumours displayed a dynamic on/fast-off pattern of type I interferon (IFN) signaling, whilst in non-responsive tumours IFN was slowly and persistently activated. By mimicking the on/off IFN signal using IFN agonists and neutralizing antibodies, we were able to markedly improve the response to ICT, but only when IFNβ or its receptor were blocked, not IFNα.
Our results show that when therapeutically targeting a dynamic process, a drug target may need to be modulated in a time-dependent biphasic manner in order to achieve optimal effect. With anti-IFNβ antibodies currently in development, these results can be rapidly translated into the clinic.