From checkpoint to checkpoint: DNA damage ATR/Chk1 checkpoint signalling elicits PD-L1 immune checkpoint activation

Multiple clinical studies have revealed a link between genomic instability and response to anti-PD-1/PD-L1 therapy in cancer management. A recent study has revealed an important role for the ATR/Chk1 DNA damage checkpoint in regulating PD-L1 expression, raising important clinical and translational questions for therapy selection and study design.

This raises the possibility that HR-deficient tumours may be sensitive to PD-1/PD-L1 inhibitors; however, their response to these agents appears less robust compared with MMR-deficient or other hypermutated cancers, such as melanomas and non-small cell lung cancers. 5 In addition, tumours with mutations in the exonuclease (proofreading) domain of polymerase epsilon (POLE) or polymerase delta (POLD1) have among the highest mutation burdens identified to date, and POLE/POLD1-mutated tumours exhibit high levels of immune activation and PD-1/PD-L1 expression, and are sensitive to immune checkpoint blockade. 6,7 Despite these robust clinical associations, the mechanisms linking DNA damage to increased PD-L1 expression and immunotherapy response are incompletely understood. To functionally characterise the impact of DNA damage on PD-L1 expression, Sato et al. subjected several DNA repair-proficient cancer cell lines to ionizing radiation, double strand break (DSB)inducing drugs such as etoposide, or a poly(ADP)-ribose polymerase inhibitor (PARPi). Treatment resulted in a timedependent increase in PD-L1 expression at both the transcriptional and protein levels. To investigate the role of the DNA damage response in this process, they added specific inhibitors of ATM, ATR, or Chk1, and demonstrated that inhibition of any of these DNA damage checkpoint kinases could suppress PD-L1 upregulation. Next, the authors screened an siRNA library targeting DNA repair genes; they found that depletion of BRCA2, PALB2 and XRCC5 [encoding Ku80, involved in non-homologous end joining (NHEJ)] resulted in the largest increase in radiationinduced PD-L1 expression. Again, they observed that pharmacologic inhibition of ATM, ATR, or Chk1 was sufficient to abrogate PD-L1 upregulation. Interestingly, PD-L1 upregulation in Ku80depleted cells could also be abrogated by depletion of either EXO1 (exonuclease) or BLM (helicase); two genes that are required for DSB processing and end resection. Similarly, in their screen, loss of BRCA1 did not show a significant increase in ionizing radiationinduced PD-L1 expression, compared with control cells. This highlights the importance of Chk1 activation via end resection as a key process in DSB-induced PD-L1 expression. Although both BRCA1 and BRCA2 are required for HR, they have distinct roles in this process; BRCA1 promotes DNA end resection by relieving the barrier posed by 53BP1 in HR, a process where BRCA2 is not involved. Therefore, according to their model, DNA end resection is impaired in BRCA1-defective cells, meaning ATR/Chk1 signalling www.nature.com/bjc is not effectively activated and PD-L1 activation does not occur. This is in stark contrast to BRCA2-defective cells where DNA end resection is not impaired as BRCA2 is not involved in this process. Finally, the authors showed that damage-induced ATM/ATR/Chk1mediated PD-L1 upregulation is dependent on IRF1 signalling through phosphorylated STAT1/3.
The work by Sato et al. provides important mechanistic insights linking DNA damage with immune activation, by identifying a critical role for the DNA damage checkpoint in regulating PD-L1 expression. Although these data have provided a window into one aspect of the signalling network that links genomic instability with immune signalling, the molecular details of much of the network remain to be elucidated. Particularly, the events downstream of ATR activation, including transcriptional changes and/or direct activation (or repression) of other signalling factors, may uncover additional mechanisms through which the immune response can be modulated by DNA damage.
Temporal aspects of DNA damage (or repair deficiency) may also impact the immune response: in the experiments by Sato et al., changes in PD-L1 levels were measured following acute damage exposure or siRNA-mediated depletion of a repair factor, and the observed PD-L1 increase did not persist beyond 14 days. Therefore, the dynamics of PD-L1 expression following chronic damage exposure or in a constitutively DNA repair-deficient background (such as BRCA2 -/cells) are unclear. In addition, many of these experiments will ultimately have to be conducted in immune-competent model systems in order to fully understand the breadth of the immune response to DNA damage.
This work also has several important clinical implications. The optimal timing and combinations of DNA damaging agents with immunotherapy are not known. Furthermore, both immune stimulating and suppressive effects of combined regimens have been noted. 8 Numerous combinations of DSB-inducing agents with immune checkpoint inhibitors have entered clinical trial evaluation in various disease settings, including combinations of immune checkpoint inhibitors with targeted agents (such as PARPi), radiation therapy, or DSB-inducing chemotherapy agents. 1 The data by Sato et al. raise the possibility that, at least in the setting of exposure to DSB-inducing agents, DNA damage checkpoint inhibitors (targeting ATR, ATM, or Chk1) may decrease tumour response to immune checkpoint inhibition by suppressing PD-L1 expression, thereby arguing against a triplet of DSBinducing therapy/ATR-Chk1 blockade/PD-1-PD-L1 blockade. Nonetheless, the optimal sequence of conventional or targeted agents with anti-PD1/PD-L1 therapy may be highly context-dependent, and this work highlights the important role for pre-clinical studies in identifying potential mechanisms of synergy or antagonism.
Another clinical implication of this work is the selection of anticancer therapy after PARPi progression. PARPi are synthetically lethal with HR deficiency and are now FDA approved for clinical use in ovarian cancer, while also being evaluated in several other HR-deficient tumour settings. However, a substantial fraction of patients eventually develop resistance to PARP inhibition and several mechanisms of PARPi resistance are now emerging. In BRCA1-mutated tumours, resistance to PARPi may occur due to a rescue of DNA end resection ability via loss of 53BP1, REV7, or Ku70/80, which increase HR capacity. [9][10][11] In this setting, reestablishing end resection may promote PD-L1 upregulation via EXO1/BLM (and ATM/ ATR/ Chk1) activity, and thus sensitise tumours to PD-1/PD-L1 blockade with or without DSB-inducing agents. In BRCA2-mutated tumours, resistance to PARPi may occur via protection of the replication fork, 12,13 which is dependent on ATR activity. 14 This ATR activity may lead to upregulation of PD-L1 and promote response to PD-1/PD-L1 blockade. In both scenarios of PARPi resistance in BRCA1/2-mutated tumours, the study by Sato et al. suggests that PD-1/PD-L1 inhibitors may be a useful therapeutic strategy (with or without concurrent DNA damaging agents) for tumours that have progressed on PARPi.
The work by Sato et al. highlights an important role for the DNA damage ATR/Chk1 checkpoint in regulating PD-L1 expression, thus linking DNA DSB signalling with regulation of the immune response. These observations have important clinical implications for therapy selection, particularly following progression on PARPi and other DNA damaging agents. They also have translational implications for the design of appropriate correlative studies in ongoing and future clinical trials of DNA damaging agents in combination with immunotherapy.  Fig. 1 Immune activating and suppressive effects of DNA doublestrand break (DSB) signalling. DSBs created by damaging agents such as ionizing radiation activate a network of signalling pathways that balance the host immune response. If repaired using an errorprone pathway such as non-homologous end joining (NHEJ) or alternative end joining, DSBs can result in somatic mutations that act as neoantigens. DNA damage can also activate the innate immune system via the STING pathway. 15