Radiation-induced Adaptive Response: New Potential for Cancer Treatment

The sophistication of radiotherapy technology has increased markedly in the last two decades, revolutionizing tumor targeting and normal tissue sparing. New dose and fraction sizes, along with the application of stereotaxis, allow safer delivery of a large single dose (SD) or a few larger fraction (hypofractionation) treatments. In addition, the use of charged particle therapy and high dose-rate brachytherapy limits normal tissue injury and potentially results in novel radiation biology. The basic tenets of radiation biology for tumor control are traditionally the 4 R's: repopulation, redistribution, repair, and reoxygenation (1). With an increased understanding of tumor response, additional “R's” have been suggested, including “radiosensitivity” and immunologic “rejection” (2–5). The mechanisms behind these concepts are known to be complicated and interact with one another during treatment (e.g., DNA damage and immune response, stem cell repopulation, and radioresistance), and understanding these responses, which are likely operational for all cancer therapeutics, offers novel precision-medicine treatments.

Recognizing that data from experimental doses used in the laboratory were often not clinically relevant, we examined the impact of clinically relevant radiotherapy doses and schedules along with NSAIDs on tumors and normal tissues (6–8). Very early in the current era of immuno-oncology and checkpoint inhibitors, we demonstrated that fractionated radiation triggers canonical immune response pathways in both tumors and normal tissues (9). These observations led to the novel hypothesis that multifraction (MF) radiation induces cellular adaptations in tumors that can be targeted with drugs to which the tumors had little to no sensitivity before radiotherapy. As described in this review, the data demonstrated that SD- and MF-induced changes depended on the underlying cell type (brain, breast, and prostate tumor and endothelial cells) and occurred even at doses with little cell killing (7). In addition, cancer stem cell repopulation can lead to acquired radioresistance (10), demonstrating another type of radioadaptive response. Thus, radiotherapy adaptation is relevant to the tumor as a whole, individual tumor cells, and surrounding normal tissue beyond the high-dose field margin.

Previous publications from our laboratories demonstrated that MF radiation (e.g., 1 Gy × 10 over 5 days) induced more differential mRNA and miRNA gene expression compared with SD radiation (e.g., 10 Gy × 1; refs. 11, 12). Herein, we discuss new findings that support the applicability of radiation-induced adaptations for new therapeutic combinations (13). These adaptations can accompany the advances in physical and spatial focusing that are part of “Accurate, Precision Radiation Medicine” (14). A new paradigm established in a recent NCI workshop (13) emphasizes the importance of describing radiation in both the physical dose, Gy, and in biological perturbations. This includes the adaptations discussed in this review to potentially exploit radiotherapy-inducible targets for cancer care including developing radiomitigators and protectants of normal tissue injury.

In essence, we consider radiation “as a drug” where the pharmacokinetics and pharmacodynamics (PK/PD) can be used in unique ways to impact both local therapy and distant metastases. The technical capability of radiotherapy allows the treatment to be focused in a wide range of targets, doses, and schedules, a concept we labeled “focused biology” (15). While we recognize the limitations of laboratory models and that there is much work to be done for these adaptations to be clinically applicable, this report provides insight into the current state of the science.

The range and depth of “omics” analyses continually expand with studies on coding and noncoding RNAs, metabolomics, and proteomics. Our initial work on fractionation was performed in vitro and in vivo in prostate, breast, and brain tumor cell lines, and demonstrated that the tumor microenvironment influences gene expression patterns after both SD and MF (7). The pattern of gene expression differed for the different cell lines with most changes occurring in the immune response pathways (6, 9, 16). Both the extent and stability of changes in gene expression were greater following MF compared with SD. In addition, findings from both our laboratory and others demonstrated that gene expression changes can vary extensively depending on whether the dose is delivered as SD or MF (7, 9, 17). These studies clearly demonstrated that MF exposure was shown to alter several genes, selectively providing the opportunity to explore molecular target-directed interventions to enhance the tumor response to radiation.

To further develop the efficacy of molecular-targeted therapy following radiotherapy adaptation, we studied miRNA expression patterns after SD ranging from 5 to 10 Gy and MF of 10 fractions ranging from 0.5 to 1 Gy (Fig. 1A), using three human prostate tumor cell lines with different p53 status: LNCaP (p53 wildtype), PC3 (p53-null), and DU145 (p53-mutated). The radiotherapy-induced changes in miRNA expression pattern depend on dose size and fractionation (Fig. 1B). Despite significant changes in miRNA expression, the surviving fraction following 0.5 Gy × 10 was approximately 85%. Significant differences in the miRNA profiles of breast cancer cell line MDA-MB-361 after SD versus MF exposure have also been reported (18), signifying a dependency on dose fractionation and the presence of a radioadaptive miRNA response. We also studied the differential mRNA and miRNA expression pattern after SD (10 Gy) and MF (2Gy × 5) in normal human coronary artery endothelial cells (HCAEC; ref. 19). The miRNA expression pattern in HCAEC was significantly altered between SD and MF at both 6 and 24 hours after the final radiotherapy dose (Fig. 1C). There were only 17 miRNAs altered after SD, in contrast to 103 differentially altered miRNAs detectable after MF. Among the altered miRNAs, only five were common to SD and MF, pointing toward the importance of dose delivery in post radiotherapy-druggable targets. Another study using a mouse model with Lewis lung carcinoma cells (LLC1) reported that gene and miRNA expression profiles are dependent on whether the cells received SD or MF radiation (17), further demonstrating the importance of the type of dose delivery in a preclinical model. These dose delivery-dependent miRNA adaptations are important for continuation of radiotherapy and cancer treatment but could be strengthened with development of predictive miRNA biomarkers. A pretreatment signature that measures radiosensitivity of head and neck squamous cell carcinoma to SD exposure shows the potential of predictive miRNA biomarkers (20); however, more studies are needed to determine the feasibility of using such biomarkers for predicting response to MF radiation.

Using prostate cancer and HCAEC cell lines, we demonstrated that cells subjected to MF promote a proimmunogenic molecular signature, among other changes (6, 9, 21). To investigate the time course of the adaptation, DU145 and PC3 cells were irradiated with 1 Gy every 6 hours, 2–3 times per day, for total doses from 2 to 10 Gy. The data demonstrate an inflection point starting from the 6th 1 Gy fraction (Fig. 1D) such that the timing and number of fractions is relevant for the induction of a proimmunogenic molecular signature.

Early radiation-induced adaptations can be exploited for immunotherapy and molecular-targeted therapy (16, 22–24). Dewan and colleagues and Vanpouille-Box and colleagues demonstrated that certain radiation doses and fractions, in combination with immune checkpoint inhibitors, were capable of inducing an immune response that produced an abscopal effect (23, 25). Cytosolic DNA that accumulates as a consequence of radiation activates the cGAS/STING pathway with downstream induction of IFN type I (IFN-I) and IFN-stimulated genes. This response can be antagonized by the DNA exonuclease Trex1. IFN-I is produced during acute viral infections and plays a key role in the activation of cytotoxic CD8 T-cells that eliminate infected cells. In tumors, the acute IFN-I production triggered by radiation elicits antitumor CD8 T-cells (23, 26); however, chronic IFN-I stimulation has been associated with therapeutic resistance (27, 28), underscoring the importance of the adaptation as an evolving response that influences the type of tumor microenvironment that develops.

Given that radiation can generate antitumor T-cells, another type of early adaptation observed in response to fractionated radiation is the upregulation by tumor cells of immune checkpoint ligand PDL-1, which protected tumor cells from immune-mediated rejection. In this case, concomitant blockade of PDL-1 improved responses by enabling T-cells to reject the tumor, providing an example of a targetable adaptation that sensitizes the tumor to immunotherapy. PDL-1 upregulation was also implicated in radiation-induced tumor equilibrium as a chronic adaptation that led to a standstill between the tumor and the antitumor T-cells (29). We have recently shown that upregulation of the ectonucleotidase CD73 on breast cancer cells represents another example of early adaptation to radiation (30). CD73 generates adenosine, a pleotropic immunosuppressive mediator, preventing the infiltration of the tumor by antigen-presenting cells while increasing regulatory T-cells. The details of CD73 signaling pathways and metabolism in tumors have been previously reviewed in depth (31, 32). In our recent study, antibody-mediated blockade of CD73 improved tumor response to radiation significantly.

These above findings shed light on how radiation might be used in combination therapy by modifying treatment based on biological adaptation. Effective treatment should strategically exploit adaptations rather than empirically following an initial regimen and examining changes only at the time of disease progression. Our previous work demonstrated that change in expression of a gene is not predictable based on its initial expression (9).

There are a range of changes that occur to enable molecular target definition (33). In proof-of-principle experiments, we measured phosphoprotein changes that occurred 2 hours after the last radiation dose. Using a more physiologic 3-dimensional cell culture system and with the goal of targeting inducible changes postradiation, we demonstrated targetable activation of the AKT/mTOR (Fig. 2; ref. 34). To test the hypothesis of dose and fractionation dependence of adaptation, a regimen of three daily 2 Gy fractions was compared with a SD of 6 Gy. Phosphoproteomic upregulation of AKT and mTOR and increased protein–protein interaction were observed in DU145 cells at 2 hours after MF radiotherapy (MF, 3 × 2 Gy) but not after a SD of 6 Gy (Fig. 2A). When drug treatment was given before and during fractionated radiotherapy, there was no enhancement in cell killing (Fig. 2B); however, when given 2 hours following completion of MF radiation—at which point the activated AKT/mTOR signaling was observed—the efficacy of the AKT inhibitor was significantly increased.

Early work by Aryankalayil, indicated that adaptation can persist up to 72 hours, the latest time point tested (9). To examine the duration of the adaptation, Eke treated PC3 cells with MF (1 Gy 10) and SD (10 Gy × 1) and cultured cells for 2 months (34). Initial cell growth was significantly slower but resumed at preradiation growth rate after approximately 6 weeks. On the basis of a long-standing interest in integrin biology and radiation enhancement with integrin-targeted therapy (35–37), the expression of integrins and the ability to target them was studied in the cells that survived SD and MF at 2 months following their last radiation treatment (Fig. 2C and D). Integrin β1 and β4 were upregulated after SD and MF compared with the mock treated controls. The ability of antibodies against β1 (AIIB2) to kill cells was enhanced following long-term adaptation after radiation when compared with unirradiated control cultures. These experiments demonstrate that postradiotherapy adaptation persists in the surviving cells, and can be targeted long after radiation treatment (38).

Our ongoing observations from the inducible changes demonstrate that there are more changes early after radiation from MF compared with SD (12, 16, 21, 38); however, at 2 months, the pattern is reversed with cells that survived SD showing more changes. These data will be the subject of future reports. As noted below, understanding and exploiting this adaptation is a key emphasis of improved molecular-targeted therapy.

Alterations in tumor metabolism, with a focus on glucose utilization, have been studied after radiation injury with the goal of more effectively destroying cancer cells (39–41). However, these studies do not necessarily take into account the speed with which these changes may occur, differences in SD versus MF schedules and to what extent these changes continue postradiation.

Preliminary metabolomic data from our laboratory indicates dynamic metabolic changes assessed by both gene expression and metabolite content at early time points between 6, 24 and 48 hours after both MF (1 Gy × 10) and SD (10 Gy × 1) radiation in PC3, DU145, and LNCaP cells (Supplementary Table S1). Of interest were the changes in fatty acid oxidation (FAO) after radiation treatment. Long-chain fatty acids enter the cell primarily through a protein-mediated system (42). Once in the cell, they are bound to coenzyme A via acyl-CoA synthases (43). Fatty acids then bind to l-carnitine via carnitine palmitoyltransferase 1 (CPT-1), producing an acylcarnitine which is then ferried into the inner mitochondrial membrane for fuel in FAO (Fig. 3A; ref. 44). Microarray data from 24 hours post 10 Gy or 1 Gy × 10 radiation demonstrated downregulation of genes which regulate the long-chain FAO pathway (Fig. 3B). IPA analysis (data not shown) indicates that these changes in FAO pathway gene expression lead to significant perturbations of the FAO pathway (Fig. 3C). This was consistent with observed alterations in certain acylcarnitines in PC3 cells after MF and SD radiation compared with controls. Increases in acylcarnitines are associated with impaired FAO (45, 46). Information on acylcarnitine expression is routinely obtained using blood and serum samples (47) which might serve as a biomarker of effect. Differential carnitine and acylcarnitine expression has recently been proposed as a biomarker in hepatocellular carcinoma (48).

Fatty acid metabolism is notably altered in cancer cells (49). This has recently spurred an interest in inhibitors of lipid metabolism, particularly FAO as novel treatments. This work is complex due to the multifaceted cell survival roles played by constituents of lipid metabolism. Some studies suggest l-carnitine itself has antitumorigenic effects due to its function as a histone deacetylase inhibitor (50). Etomoxir, a CPT-1 inhibitor, has been tested as an anticancer agent (51). Another compound of interest, Mildronate, decreases l-carnitine entry into the cell by blocking organic cation transporter 2 (OCTN2) and inhibits endogenous l-carnitine production (52). Mildronate has been shown to decrease tumor growth in a rodent model of glioblastoma (53). Avocatin B, derived from avocados, has also been tested as a novel therapeutic in leukemia cells (54). It prevents FAO, potentially through competitive inhibition of fatty acids, accumulates in the mitochondria and increases reactive oxygen species accumulation which triggers apoptosis (55). A more thorough understanding of the roles of short- and medium-chain fatty acids, acylcarnitines and FAO on cancer survival is necessary to develop effective combination therapies.

Ongoing work to characterize and utilize the potential tumor adaptation includes detailed studies of phosphoproteomics, metabolomics, DNA repair, and epigenetic changes, as well as further in vivo studies (16, 34, 12, 56). Long-term studies of tumors postirradiation conducted with Citrin at NCI demonstrate long-term upregulation of integrins in PC3 tumors (38). Mitchell recently demonstrated an improvement in tumor growth delay when using a relatively standard combined chemoradiation treatment (57). Extending drug treatment for 2 weeks postradiation had a significantly greater effect than using the drug for 1 week only. The second week of drug treatment alone inhibited radiation-induced tumor vasculogenesis and thus delayed recurrent tumor growth. This finding is consistent with studies done by Brown and colleagues where induction of SDF-1 is used to monitor radiation-induced vasculogenesis (58). Detailed analysis is now in progress; notably at week 1, many more transcriptional changes were observed with combined drug and radiation compared with drug or radiation alone suggesting perhaps the possibility of identifying novel targets for treatment.

While the adaptive paradigm needs more in vivo confirmation, the data demonstrate that (i) radiotherapy can induce an adaptive response that depends on dose and fractionation; (ii) adaptive changes occur early in treatment (days) and at the end of a week of treatment, and can persist for months, with the time interval to be studied; (iii) exploiting the adaptation by either taking advantage of it or interfering with it have potential for novel treatment opportunities; and (iv) the dose fraction size can impact immunotherapy with checkpoint inhibitors.

Acute and delayed normal tissue radiation damage is a dose-limiting factor for radiotherapy and a major concern of accidental radiation exposure. Our previous studies with HCAEC demonstrated that normal tissues also undergo radiation-induced molecular adaptations (6, 7; Fig. 1C) that could be exploited as biomarkers to predict and avoid or mitigate normal tissue injury, thereby improving patient outcomes. The potential of using RNA as a biomarker of tissue-specific injury and of general radiation exposure is increasing with the description of noncoding RNAs that are relatively stable in the blood, are organized in complex regulatory networks, and provide information on tissue-specific changes and identify pathways for injury mitigation.

The implementation of biomarkers for radiation exposure and damaged normal tissue requires an analytic solution similar to the time- and dose-oriented changes for cancer treatment. Such a solution will require complex analyses using various models (e.g., Random Decision Forest, Support Vector Machine) and ultimately a time- and dose-oriented Markov decision tree (59). Critical to this development is identification of stable, reproducible, and readily assayable markers that have a high specificity and sensitivity for detection in blood or other body fluids. To this end, Aryankalayil used a whole-body irradiated mouse model to identify significant alterations in the expression patterns of long noncoding RNAs (lncRNA) and of miRNAs and target mRNAs at different time points after various levels of exposure (60, 61). Importantly, these studies demonstrated that to triage victims of a radiological incident, multiple RNA biomarkers are needed to differentiate dose at different time points following exposure. For example, let-7e-3p may have utility within a set of biomarkers, but the varying upregulation and downregulation across doses and time points make it insufficient as a single marker (Fig. 4A). In addition, several other groups reported plasma- and serum-based miRNA signatures that distinguish dose, including high versus low and lethal versus sublethal doses (62–64), further validating the potential of circulating RNA biomarkers for radiation biodosimetry and the importance of multiple biomarkers.

Ongoing work continues to identify circulating RNA biomarkers for a biodosimetry decision tree using mouse, mini-pig, and nonhuman primate models (Fig. 4B). Tissue-specific injury markers are also being identified, which could be most useful for discerning tumor and normal tissue adaptations during radiotherapy. In addition, studies of the microenvironment immune response to local tumor irradiation have demonstrated the role of normal tissue immune regulation on tumor recurrence, metastasis, and response to therapy (65–67). Investigation into molecular biomarkers of this immune response could therefore be informative in guiding more effective treatment regimens. Ultimately, molecular signatures for each application (e.g., biodosimetry, stromal response to radiotherapy) can be implemented into a molecular diagnostic workflow (Fig. 4C). While blood-based RNA is currently the primary molecule of interest, other sources of circulating markers, such as exosome-derived RNA, are also being evaluated.

Clinical application requires a more thorough understanding of the underlying biology and the development of biomarkers or imaging strategies to assess the tumor and normal tissue. Clinical trials could then be designed with biomarkers to validate the preclinical data. Table 1 includes ongoing studies and considerations to further develop clinical targeting of radiotherapy adaptation.

Evolution in the technological delivery of radiotherapy is accompanied by rapid progress in cancer genetics and biology to decipher key molecular signaling and survival pathways and identify druggable targets. In this setting, as described in this review, the biological perturbations induced by radiation could potentially be exploited to enhance precision oncology. This includes pathways that are induced immediately after treatment and that may be sustained throughout radiation fractionation and beyond, for weeks/months after the end of radiotherapy, as a result of what we have described as radiotherapy-induced adaptation. Using a range of “omics,” it is possible to interrogate the time and type of adaptation radiation induces in cancer, when compared with the pretreatment phenotype. We speculate that different pretreatment genotypes will have specified types of adaptation (e.g., p53 normal vs. abnormal) that will help guide the initial treatment as is being studied in the match-type trials (69). A correct interpretation of these results enables rational design of multimodality therapies. Some adaptations take place within the first 5–10 or so fractions (2–4 days) and may persist for a number of days; long-term adaptations occur up to 2 months after treatment completion, and possibly longer. Early adaptation was traditionally called a “stress response” and can now be interrogated on how it is changing the tumor biological phenotype. In each model, the exact timing and how different forms of adaptation occur and how long they persist remain to be defined. Likewise, adaptation may be influenced by the irradiated tumor microenvironment, and the cross-talk between the tumor and the host. Growing evidence shows that the early changes induced by radiation in surviving cancer cells can improve the recognition of the tumor by the adaptive immune system and generate targets for immunotherapy (22, 70).

While little information exists about the mechanisms of late adaptation on the tumor interaction with the immune system, some clinical papers suggest that pretreatment with radiation of chemoradiation may enhance effects of immune checkpoint blockade (71). Research in this area is warranted, particularly because novel immunotherapies are becoming rapidly available.

The PK/PD of radiation does depend on dose and fractionation and on type of radiation (e.g., photons vs. particles), such that radiation can and should be considered “like a drug,” and function as a key partner within the precision medicine paradigm. In multiple models, we found that the initial gene expression pattern did not predict radiation-induced gene response, indicating that investigation of adaptation mechanisms and pathways requires sequential interrogation. As outlined in Table 1, much remains to be done. Nonetheless recognizing that adaptation occurs offers the opportunity for unique approaches to cancer treatment. It is clear that postradiation, the tumor and normal tissue “know” they have been irradiated and display persistent changes that depend on the type of radiation, dose, and fractionation and whether radiation was given alone or with other modifiers, as in combination with chemotherapy, molecular-targeted therapy or immunotherapy.

Defining an individual tumor's adaptation to a radiation regimen permits testing treatments against newly emerged targets. Most importantly, radiotherapy can be used multiple times to enable testing different permutations of combinations. Indeed, drugs and/or immunologic agents may become more effective after the adaptation, potentially inducing susceptibility to a drug based on cellular adaptation rather than a mutation. This could have major impact on return on investment for drug development as a drug deemed unlikely to work based on the initial tumor profile may become effective after adaptation.Complex analysis of biomarkers of both tumor and normal tissue response are warranted. The ability to utilize the focused biology (12) from radiotherapy is a potentially unique enhancement of both local and systemic therapy adding a new dimension to accurate, precision cancer care.

S. Demaria reports grants and personal fees from Lytix Biopharma (research grant paid to institution; personal fees for advisory service), grants from Nanobiotix (research grant paid to institution), and personal fees from EMD Serono, Mersana Therapeutics, AbbVie Inc., and AstraZeneca (all for ad hoc advisory service) outside the submitted work. S.C. Formenti reports grants from Bristol-Myers Squibb, Varian, Eli-Lilly, Janssen, Regeneron, Eisai, and Merck, and other (honorarium) from Bayer, Bristol-Myers Squibb, Varian, ViewRay, Elekta, Janssen, Regeneron, GlaxoSmithKline, Eisai, AstraZeneca, MedImmune, Merck US, EMD Serono/Merck, and Accuray outside the submitted work. No potential conflicts of interest were disclosed by the other authors.

This study was supported by the NIH Intramural Research Program, NCI, and Center for Cancer Research (grant ZIA BC 010670, to C.N. Coleman).

For research collaboration: Murali Cherukuri, Radiation Biology Branch; Deborah Citrin, Radiation Oncology Branch; James Hodge, Laboratory of Tumor Immunology and Biology; Mansoor Ahmed, Radiation Research Program; Laurel MacMillan and Rocco Casagrande, Gryphon Scientific, Takoma Park, MD; and for support of the projects, Kevin Camphausen, Chief Radiation Oncology Branch.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.