For more than a century, radiotherapy (RT) has been the cornerstone for the treatment of cancer. The classical radiobiological mechanisms underlying tumor cell death are well known, mainly involving deoxyribonucleic acid chain (DNA) damage, either directly or via water radiolysis and the production of free radicals and reactive oxygen species (ROS). The relative biologic effectiveness of radiation is influenced by several mechanisms known as the ‘5Rs’: repair of sublethal damage, repopulation, redistribution within the cell cycle, reoxygenation and intrinsic radiosensitivity, which mostly explains variations in radiosensitity/radioresistance for a given tissue/tumor (1).

Conventional RT consists in delivering once daily fractions of 1.8-2.2 Gy for 5-8 weeks, as this empirical approach turned out to achieve a differential effect between tumor cells and normal tissue. With the advances in dose delivery, patient immobilization and repositioning and tumor motion management, stereotactic RT has emerged, enabling the delivery of higher biological effective doses (BED) in fewer fractions and with a sharp dose fall-off.

Immunotherapy (IO) to restore and/or to boost anti-tumor immunity, especially with immune check-point inhibitors (ICI), has changed the standard of care in many fields of oncology for a decade. For example, the PACIFIC trial led to an unprecedented gain in progression-free survival (PFS) and overall survival (OS) for the management of non-resectable stage III non-small-cell lung cancer by adding durvalumab as maintenance therapy following chemoradiotherapy (2).

Along with the advances in anti-tumor immunity research, the deciphering of radiation-induced immune effects has led to the concept of a dual immune effect of RT with both immunostimulatory and immunosuppressive effects (3). Briefly, RT can release tumor-antigens (TA) along with the translocation of calreticulin to the tumor cell membranes, leading to tumor cell phagocytosis (4) and the activation of the cytosolic DNA sensing cGAS/STING pathway, with in turn induction of interferon β (IFN-β) (5), and the release of damage-associated molecular patterns (DAMPs) (such as heat shock proteins, high mobility group box 1 molecules (HMGB1) or adenosine triphosphate (ATP)). These DAMPs are recognized by toll-like receptors (TLRs) expressed at the surface of dendritic cells (DCs), and can promote processing and cross-presentation of TA by IFN-β-induced mature DCs (6). Following this immunogenic cell death, DCs then migrate to the tumor-draining lymph nodes and prime CD8⁺ T cells (7), with in turn leukocyte extravasation and recruitment to the tumor site through chemokine secretion by tumor cells and other cell types in the tumor micro-environment (CXC motif chemokine ligand (CXCL)9, CXCL10, and CXCL16) (8, 9). Once T cells have infiltrated the tumor tissue, they encounter tumor cells with the radiation-induced expression of several surface molecules and receptors, such as MHC-I molecules (10), the TNF-R superfamily (11, 12) and ligands for the NKG2D receptor (13), leading to enhanced tumor cell killing by CD8⁺ T cells and NK cells.

Together with this radiation-induced in situ “vaccination”, RT can induce immunosuppressive effects via several mechanisms: upregulation of PD-L1 levels on tumor cells via IFN-γ released by CD8⁺ T cells and of PD-1 levels on CD8⁺ tumor infiltrating lymphocytes (TILs), contributing to T cell exhaustion (14, 15); direct depletion of circulating lymphocytes and lymphoid progenitors in primary and secondary lymphoid organs (16, 17); enhancement of immune suppressive pathways (mostly: HIF1α upregulation, increased colony-stimulating factor 1 (CSF1) levels, induction of TGF-β and generation of adenosine from ATP), which in turn lead to a suppressive tumor micro-environment (TME) with induction of CD4⁺CD25⁺ regulatory T cells (T-reg) proliferation, M2 polarization of tumor-associated macrophages (TAMs), and myeloid-derived suppressor cells (MDSC) activation (18).

Overall, this dual effect can explain why RT alone is not able to drive a strong anti-tumor immune response with a so-called “abscopal” effect in most cases and underlies the rationale for combining RT with IO, not only to amplify the in situ vaccination effect but also to overrule immunosuppressive effects. This rationale has generated considerable interest in this field, and around 700 trials are currently ongoing assessing different regimens of such associations in oncology. However, while some trials have provided unprecedented results and shown much promise (2, 19–21), others have led to disappointment (22–24). These discrepancies leave many open questions regarding the optimal combinations of these treatment modalities.

This narrative review aims at revisiting the old, well-established concepts of RT relating to dose, fractionation, target volumes and organs at risk in the era of IO, in order to propose potential innovative approaches to be further assessed when considering an RT + IO association, especially in the field of non-small-cell lung cancer (NSCLC).

Searches for original and review articles in the PubMed and Google Scholar databases were conducted until September 2020. General search terms (including both Medical Subject Headings (MeSH) and free text words) included the following: “radiotherapy”, “immunotherapy”, “immune checkpoint inhibitor”, “anti-PD(L)1”, “abscopal effect”, “lung cancer”, “non-small-cell lung cancer”, “lymphopenia”. Individual bibliographies were reviewed for additional relevant references.

To establish the best RT scheme in the context of an RT + IO association, one should first define the main objective of such an approach (25) (Figure 1).

Schematically, the first objective is to promote the in situ vaccination effect of RT, either by adding IO to a short course of ablative (i.e. tumoricidal) RT towards the whole tumor sites (in early-stage disease or oligometastatic disease; in this case, IO in itsef also addresses the micrometastatic disease), or by adding RT at one or several metastatic sites to IO (polymetastatic disease). Several IO agents potentially trigger such an effect: activation of DCs via TLR agonists (26) or CD40 agonists (27); enhancement of T-cell priming via CTLA-4 antagonists (28, 29), OX40 agonists (30) or PD-1/PD-L1 antagonists (as PD-1 acts by inhibiting signaling downstream of the CD28 costimulatory receptor following B7-ligation) (31); enhancement of killing by effector T cells, mostly via PD-1/PD-L1 antagonists (14, 15).

Another approach consists in counteracting immunosuppressive signals induced by conventional daily definitive (chemo-)radiotherapy schedules for locally advanced disease, due to the enhancement of immunosuppressive pathways as described above. In this case, IO and especially ICI can be used preferentially as a consolidative agent immediately following standard of care (chemo-)radiotherapy, then addressing the micrometastatic disease. Indeed, the restoration of the effective functions of TILs with ICI given concomitantly with tumor irradiation can be counterproductive owing to the profound suppression of TILs induced by daily RT. This is the rationale underpinning the PACIFIC trial, in which the addition of durvalumab following chemo-radiotherapy for stage III NSCLC led to the reduction of distant metastases and improved PFS and OS (2). The question as to whether IO can also act as a local radiosensitizer through a synergistic effect in this setting remains a matter of debate.

Finally, irradiation may serve as a strategy to modify the response to IO, in order to increase the immunogenicity of “cold tumors” through the homing of TILs or the reprogramming of the TME, inducing macrophage M1 polarization, for example (32).

Regarding the in situ vaccination effect, high dose per fraction irradiation (HDFI), usually through stereotactic radiotherapy (SRT), that delivers a few fractions with a high dose of radiation per fraction (generally above 6-8 Gy) is usually preferred, either as a tumoricidal schedule (e.g. 5 x 8 Gy) or as a non-tumoricidal schedule (e.g. 3 x 6 Gy). It has been shown that the release of intra-cellular peptides following single-fraction radiation took place in a dose-dependent manner via three main mechanisms with early and late effects: an increase in old protein degradation, upregulation of defined proteins through the response repair, and an increase in protein synthesis through the mTOR pathway activation. As peptides are the limiting factor, the increased intra-cellular peptide pool led to a dose-dependent increase in MHC class I presentation (10). Besides, Golden et al. showed that each component of immune-cell death following single-fraction radiation (calreticulin cell surface exposure, release of high mobility group box 1 (HMGB1) protein and release of ATP) was also induced in a dose-dependent manner from 2 to 20 Gy (33). Finally, Morisada et al. suggested a dose-dependent effect of radiation on both TA release and T-cell priming, with 8 Gy in a single fraction enhancing these pathways compared to 2 Gy in a single fraction, resulting in increased tumor cell susceptibility to T-cell-mediated killing (34). Importantly, Dewan et al. showed that a 3 x 8 Gy regimen was superior to 5 x 6 Gy in the induction of the abscopal effect and of tumor-specific T-cells, suggesting that this dose-dependent pro-immunogenic effect refers to the dose per fraction more than the total dose (35). However, the same group showed that HDFI (3 x 8 Gy 5 x 6 Gy), but not “ultra”-high single-dose RT (20 Gy x 1), was able to induce an abscopal effect when combined with anti-CTLA-4 (35). Vanpouille-Box et al. showed that the DNA exonuclease Trex1 is induced by radiation doses above a threshold ranging from 12-18 Gy (36). During phagocytosis by myeloid cells, DNA fragments hidden in irradiated tumor cells are released from tumor-derived exosomes to the cytoplasm of myeloid cells (37), and cytosolic DNA stimulates the secretion of IFN-β through the activation of the DNA sensor cGAS and its downstream effector STING, in turn promoting the cross-priming of CD8⁺ T cells (5). Above the threshold for Trex1 activation, DNA fragments are cleared from the cytosol, then precluding the secretion of IFN-β and T cell priming (36). Finally, while classical approaches tend to favor doses per fraction that are as high as possible in the context of SRT, these data suggest that the best SRT schedule for maximizing in situ vaccination in combination with IO is the delivery of 8-10 Gy fractions. In the context of early-stage NSCLC, several trials are currently assessing the benefit of IO in addition to standard of care SRT (NCT03110978; NCT03446547; NCT 03050554; NCT03383302). Some of them have implemented doses per fraction of around 10-12 Gy while another approach consists in a traditional fractionation of 3 x 18 Gy with addition of IO acting more as an adjuvant treatment than as a synergistic association to decrease the risk of regional and distant failures following SRT for high-risk stage I disease. In the context of NSCLC oligometastatic disease (generally fewer than five metastases), where SRT to all targets is now classically proposed as an ablative treatment (38, 39), the benefit of IO adjunction to SRT is being assessed in several trials with dose per fraction around 6-10 Gy (NCT03275597). In this perspective, when SRT to brain oligometastases is proposed in a context of IO, a hypofractionated schedule (e.g. 3 fractions of 8-10 Gy) could be better than a classical single fraction of 16-20 Gy. Such a schedule is being tested in patients with recurrent glioblastoma, in association with durvalumab (40). A provocative question is whether tumoricidal irradiation is absolutely required for localized disease when HDFI and IO are combined. This could pave the way for dose de-escalation with the definition of new therapeutic windows exploiting the synergy between RT and IO while decreasing radiation-induced toxicity. Finally, in the context of polymetastatic NSCLC disease, the benefit of the in situ vaccination effect of HDFI (tumoricidal or not) to one or several targets using doses per fraction of 6-10 Gy in addition to standard of care IO has been suggested (19–21) and is being assessed in the phase III NIRVANA-Lung trial NCT03774732.

When ICIs have been assessed as consolidation agents following standard of care definitive (chemo-)radiotherapy for locally advanced disease, RT has been delivered mostly in a conventional dose-fractionation schedule (1.8-2 Gy per fraction, one fraction per day, five days per week, to a total of 60-66 Gy for NSCLC) (2, 41, 42). The rationale behind this schedule is based on the linear quadratic model, whereby the optimal dose-fractionation regimen in order to kill cancer cells while sparing surrounding normal tissues may be established. However, the linear quadratic model accounts only for radiation cell killing and does not take the role of the immune system in antitumor responses into account (1). Therefore, the optimization of dose-fractionation chemoradiotherapy regimens for locally advanced disease in the context of IO combination requires careful consideration, as conventional fractionated regimens have been associated with lymphopenia and immune suppression in several types of cancers (16, 17). In this perspective, moderately hypofractionated (2.5-4 Gy per fraction) schedules could be of interest because the acceleration of treatment allowed by hypofractionated schedules could reduce the amount of blood passing through the beam and thus the duration and the severity of radiation-induced T-cell suppression and lymphopenia (43, 44). Indeed, in their study of 115 patients with unserectable stage III NSCLC treated by definitive RT, Zhao et al. found that overall treatment time within 4 weeks was significantly associated with a decreased risk of developing severe lymphopenia in multivariate analysis (44). Notably, in this setting of locally advanced disease, the question whether the addition of IO to (chemo-)radiotherapy can act as a radiosensitizer through a synergistic effect remains open for two main reasons. First, the majority of data regarding in situ vaccination have been obtained using a high radiation dose per fraction, corroborating the fact that the pro-immunogenic effects of radiation probably occur in a dose per fraction-dependent manner, provided that the 10-12 Gy threshold is not surpassed (10, 33, 34). Yet, the large fields required in the treatment of locally-advanced disease generally preclude the use of doses per fraction higher than 4 Gy. Second, data regarding the synergistic effects of moderately hypofractionated RT in association with IO are not consistent: while several preclinical studies suggest a benefit of hypofractionated over conventionally fractionated regimens, due to better CD8⁺ T cell dependent primary and abscopal tumor control (45) and reduced recruitment of MDSCs into tumors through the downregulation of vascular endothelial growth factor (VEGF), a recent clinical series of 47 metastatic melanoma patients treated with ipilimumab and RT showed that fraction size ≤ 3 Gy vs > 3 Gy was associated with an improved rate of index lesion response outside the radiation field after adjusting for total radiation dose, site irradiated, timing of ipilimumab, and time from diagnosis to radiation treatment (46). Overall, hypothesizing a potential synergy of moderate hypofractionated RT in combination with IO, and considering the concern of early data of the toxicity of such an association (47) as well as the inconsistent data regarding the dose response effect following chemoradiotherapy in locally advanced NSCLC (48–50), dose de-escalated hypofractionated RT in combination with IO (and especially durvalumab consolidation) in stage III NSCLC is probably an approach to be investigated.

Finally, when considering RT as a response modifier of IO, low-dose irradiation (LDI) with one or a few fractions of 0.5 to 2 Gy has been shown to potentially increase the immunogenicity of “cold tumors” through several mechanisms: preferential induction of T-reg apoptosis compared with effector T cell cells (40); skewing macrophages from an M2 phenotype (promoting tumor growth) towards an inducible nitric oxide synthase-positive (iNOS+) M1 phenotype. These M1 macrophages in turn produce a range of chemokines which facilitate T-cell recruitment and normalize tumor vasculature, inducing T-cell tumor-infiltration (32). Furthermore, an original approach has been proposed combining both HDFI and LDI in the context of the RT + IO association in order to generate in situ vaccination together with T-cell homing towards tumor sites (25). This hypothesis was corroborated in a preclinical study with bilateral mouse tumor models in which the authors suggested that HDFI of the primary tumor combined with LDI of the abscopal tumor and anti-PD-1 therapy achieved the best abscopal response, compared to HDFI + anti-PD-1, HDFI + LDI or LDI + anti-PD-1. The enhanced abscopal response was correlated with increased infiltration of CD8⁺ effector T cells and upregulated expression of T-cell attracting chemokines (51). Clinical evidence of such LDI in association with HDFI has also been suggested in several reports. In a post-hoc analysis of three immunoradiation trials monitoring SRT with HDFI to a limited number of targets in association with IO, the out-of-field response of non-target lesions among 26 patients was statistically improved among low-dose irradiated lesions (mainly due to scatter dose related to anatomic proximity to another targeted lesion) compared to no-dose (<1 Gy) lesions (52, 53). Similarly, when LDI (4.9 Gy, range 2-8 Gy in 2 Gy fractions) was given intentionally to one large lesion together with ICI and HDFI, the low-dose treated lesion shrunk by 28.2% on average in 6 out of 9 patients with metastatic NSCLC (51). The RACIN trial is currently assessing the benefit of LDI to several lesions among advanced TIL-negative tumors in association with nivolumab and other agents (NCT03728179).

The modality of delivery of the radiation dose should be taken into account when investigating the immune effects of RT.

Intensity-modulated radiotherapy (IMRT) emerged in clinical practice around two decades ago as a technique for delivering a more accurate dose distribution than conventional 2-dimensional RT or 3-dimensional conformal RT (3D-CRT), with limited exposure of adjacent OARs, including structures located within a concave area of the PTV. To do so, an “inverse planning” is performed, where the treatment planner first determines the dose distribution for the target tumor and OARs, and then the optimization method determines the intensity of the irradiation beam (125). In stage III NSCLC, IMRT as compared to 3D-CRT has been associated in dosimetric studies with improved PTV coverage, and with a decrease in the volume of whole lung receiving more than 20 Gy and in cardiac doses (126). This translated into a reduction in severe pulmonary toxicity and even improved OS in several large retrospective series (127–130). IMRT was also associated with decreased severe radiation pneumonitis as well as improved quality of life in secondary analyses of the RTOG0617 trial (131, 132). However, in comparison to 3D-CRT, IMRT increases the low-dose bath as a greater number of beams and monitor units are used, and several studies have shown an increase in lung V5Gy (126, 133). Concerns about V5Gy/V10Gy and fatal pneumonitis have been raised with IMRT in the context of post-operative RT for mesothelioma (134), however, no clear correlation has been established for IMRT in NSCLC, and it is commonly thought that the potential benefit of IMRT outweighs this risk in NSCLC. Nevertheless, as lung V5Gy was significantly associated with post-RT lymphocyte nadir among patients who received definitive RT for NSCLC (43), and given the negative impact of lymphopenia on outcome, attention should be paid to the low-dose radiation lung volume in the era of immunotherapy. This potential increased risk of lymphopenia with IMRT could be counterbalanced by a decrease in treatment time (beam on time) by using flattening filter free (FFF) radiation beams, which can provide high-dose rate beams (135).

Stereotactic RT is usually proposed for early-stage NSCLC in medically inoperable patients. Owing to the technical properties and characteristics of dose gradient, SRT is associated with low-dose bath to ensure a high conformal dose distribution around the target; however considering the small size of the lesions treated with SRT, this low-dose spread is usually limited. This could prompt to develop approaches of SRT-based sub-volume radiation boost following a conventionally fractionated course of RT for the treatment of locally advanced disease; this approach is currently being explored in stage III NSCLC (136).

Proton therapy is also gaining interest in locally advanced NSCLC owing to its dose distribution capabilities related to the release of proton energy, mostly at the end of the path, a phenomenon known as the Bragg peak phenomenon. Proton therapy has shown promise in reducing normal lung tissues receiving low-dose ranges, while maintaining dose constraints to other critical structures such as the heart, esophagus and spinal cord (137). This could explain the superiority of proton therapy over photon-based IMRT in terms of severe lymphopenia in patients treated with (chemo-)radiotherapy for glioblastoma (138), esophageal cancer (139, 140) or medulloblastoma (141). However, a phase II randomized trial comparing proton therapy and IMRT in the treatment of stage III NSCLC failed to show any advantage of proton therapy on toxicity or on local failures (142). In addition to this dosimetric advantage, proton therapy could have intrinsic immunomodulatory properties. A recent study suggests that proton therapy induces upregulation of surface molecules involved in immune recognition (HLA, ICAM-1 and tumor associated antigens), and translocation of calreticulin, in a manner similar to photon irradiation. The authors extended their observations to cancer stem cells, which are classically resistant to radiation (143, 144). These results also support the association of proton therapy with T-cell mediated immunotherapy.

Finally, owing to their particular features, emerging unconventional approaches of RT may provide additional benefits when combined with IO (145). FLASH RT is one of these promising approaches. It is able to deliver radiation at ultra-high dose rate, which is thought to induce massive oxygen consumption; while tumors are generally already hypoxic, FLASH RT can induce transient protective hypoxia in normal tissues. Therefore, FLASH RT could enhance the differential effect between tumors and normal tissues as compared to conventional RT (146). The modulation of immune response with FLASH RT is not well established; however, together with a decrease in treatment time, some particular features associated with FLASH RT such as massive TA release or decrease in immunosuppressive TGF-β cascade activation may provide additional mechanisms of the synergistic effect of the RT – IO association (147, 148).

The combination of RT and IO at any stage of cancer disease, i.e. from early stage to both oligo- and poly-metastatic disease, is offering new hope for the treatment of patients with malignancies. However, given the dual effect of RT upon the host immune system the RT schedule must be optimized whenever a synergistic effect of the combination of RT and IO is expected. To reach this objective, several traditional dogmas about RT might need to be revisited and challenged in this new therapeutic era, regarding dose, fractionation, target volumes, dose to organs at risk and dose delivery techniques. The main issues are summarized in Figure 2. Thus, both translational and clinical studies are necessary to better understand the mechanisms underlying the immune effects of RT and to provide a strong rationale for this combination. Along with the optimization of radiation dose delivery, biomarkers need to be validated to predict a synergistic effect of the RT – IO combination, based upon tissue analysis, circulating biomarkers, and quantitative imaging with radiomics (149, 150).

JK designed the review, analyzed the data and drafted the manuscript. JM, CG-R, EC-J and MA analyzed the data and critically revised the manuscript. EC-J and MA supervised the writing. All authors contributed to the article and approved the submitted version.

Open access publication fees funded by Institut Claudius Regaud/Institut Universitaire du Cancer de Toulouse - Oncopole.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.