Applying Precision Oncology Principles in Radiation Oncology

Abstract

Radiation therapy is a critical component in the curative management of many solid tumor types, and advances in radiation delivery techniques during the past decade have led to improved disease control and quality of life for patients. During the same period, remarkable advances have also been made in understanding the genomic landscape of tumors; however, treatment decisions in radiation oncology continue to depend primarily on clinical and histopathologic characteristics rather than on the genetic features of the tumor or the patient. With the development of novel genomic techniques and their increasing use in clinical practice, radiation oncology is uniquely positioned to leverage these advances to identify novel biomarkers that could inform radiation dose, field, and the use of concurrent systemic agents. Here, we summarize efforts to use genomic techniques to guide radiation decisions, and we highlight some of the current opportunities and challenges that exist in attempting to apply precision oncology principles in radiation oncology.

INTRODUCTION

Radiation plays a central role in cancer management, and it is estimated that more than half of all patients with cancer will receive radiation therapy during their treatment course.¹ Radiation is used in a variety of clinical contexts, including in the definitive management of several solid tumor types as well as in palliation of symptoms associated with advanced disease.²

Many of the changes in radiation oncology in recent decades have been driven by advances in imaging and dosimetry that have resulted in the ability to deliver higher radiation doses to tumor while minimizing the dose to surrounding normal tissue.³ In contrast, advances in understanding tumor biology and genetics have affected radiation oncology practice less to date, particularly when compared with other oncology specialties.^4,5

Currently, genomic biomarkers are rarely used to inform the use of radiation therapy. Instead, clinical-pathologic factors, such as tumor size, histology, lymph node involvement, and surgical margin status, continue to drive radiation oncology standards of practice. Thus, although radiation is a precision treatment modality in a spatial and anatomic sense, the potential to incorporate tumor genomic features as a precision tool in radiation oncology has not yet been realized. Here, we discuss progress toward leveraging genomic insights to inform radiation treatment and highlight areas for future investigation.

GENOMIC DETERMINANTS OF TUMOR RESPONSE TO RADIATION

From the earliest days of its use as a therapeutic modality, there has been an appreciation that different tissue types demonstrate markedly different responses to radiation. Efforts by radiobiologists to understand and model these differences have driven current clinical strategies, such as dose fractionation (ie, delivering a fractional dose of radiation each day over several weeks), that exploit differences in the radiation sensitivity of tumor and normal cells. The development of massively parallel sequencing and other high-throughput techniques has led to an explosion in available tumor genomic data, which provide a unique opportunity to map the landscape of radiation response across tumor types. Nevertheless, defining the underlying genomic determinants of differential radiation response remains challenging for several reasons.

Historically, the tumoricidal effects of radiation were believed to be mediated primarily through DNA damage, but accumulating evidence suggests that radiation has numerous effects on the tumor and microenvironment that vary on the basis of anatomic site, tumor histology, radiation dose and fractionation, and the use of concurrent therapies.^6,7 Therefore, the molecular underpinnings of radiation response may vary within and among tumor types and may be strongly dependent on clinical and treatment factors.

When delivered in the neoadjuvant or definitive settings, radiation is often combined with cytotoxic chemotherapy, and separating the effects of each agent on tumor response is difficult. Conversely, when radiation is used in the adjuvant setting, no measurable tumor is present, and response is defined by lack of tumor recurrence over months or years, which can be affected by factors beyond tumor cell radiosensitivity.

Finally, although comprehensive genomic profiling of thousands of tumors has been performed through efforts such as The Cancer Genome Atlas, these studies often pool cases that represent diverse clinical settings and disease states, and detailed treatment and response data are often not available. Few of these large, publicly available data sets include patients treated with radiation. Furthermore, even when an association between a specific genomic event and treatment response is observed, rigorous experimental work is required to validate the association and establish causality.

Experimental Systems to Study Radiation Sensitivity

Many of the tenets of radiobiology were developed and validated using radiosensitivity assays, including in vitro approaches such as clonogenic cell survival and in vivo approaches using transplantable tumor systems.⁸ Although these assays have been invaluable in establishing the mechanisms of radiation-mediated cell killing and the properties of dose fractionation, the assays are often time consuming, technically challenging, and difficult to scale. Therefore, one of the most important challenges currently facing the field is the development of efficient and reliable techniques that faithfully recapitulate the effects of radiation to yield insights at both the cellular and tissue levels.

In an attempt to comprehensively characterize associations between genomic features and radiation sensitivity, Yard et al⁹ profiled radiation sensitivity across 533 genomically annotated cell lines representing 26 tumor types using a validated high-throughput assay. Perhaps one of the most surprising findings from this study was the large variation in radiation sensitivity observed within and among lineages: many tumor types exhibited a greater than five-fold difference in survival between the most- and least-sensitive cell lines. In most tumor lineages, the extent of somatic copy number alterations (ie, aneuploidy), a marker of genomic instability, was correlated with radiation resistance. Conversely, mutations in several DNA repair genes, such as BRCA1, MLH1, and ATR, were associated with high point mutation frequency, low levels of aneuploidy, and radiation sensitivity. Whole-transcriptome gene expression profiling revealed that upregulation of pathways involved in DNA damage response, cell cycle, and chromatin organization were associated with radiation sensitivity, whereas increased expression of genes in pathways involved in cellular signaling (Janus kinase-signal transducers and activators of transcription [JAK/STAT] and phosphatidylinositol 3-kinase [PI3K] pathways), stem-cell state, and inflammation were associated with radiation resistance. This study is one of the largest efforts to date to investigate the association of tumor genomic features with radiation sensitivity and highlights the potential to leverage large data sets to study cellular radiation response.

Many other groups have studied gene expression patterns in an attempt to define transcriptional signatures of radiation response. One of the most widely applied of such signatures is the radiosensitivity index, a 10-gene signature that was derived by modeling cell survival across a panel of 48 cell lines from the National Cancer Institute–60 cell line panel.^10-12 The radiosensitivity index has been found to associate with radiation response and treatment outcomes in several primary tumor cohorts; however, questions have also been raised about its reproducibility in the remaining National Cancer Institute–60 cell lines and in independent cohorts.^13,14 Similar gene expression–based approaches have focused on specific disease types; for example, a transcriptional signature of radiosensitivity developed in a panel of breast cancer cell lines was independent of intrinsic subtype and was strongly correlated with local recurrence risk after breast radiation.¹⁵ Other signatures have focused specifically on certain pathways, such as DNA repair, and are strongly correlated with response to DNA-damaging agents, including radiation.¹⁶

In addition to mutation- and expression-based features of radiation sensitivity, efforts have also focused on developing functional assays to characterize radiation sensitivity of primary tumor samples. Several studies have identified an association between increased clonogenic survival of primary tumor cells irradiated ex vivo and worse clinical outcomes after radiation-based treatment.^17,18 In addition, immunohistochemical (IHC) readouts of DNA repair function—such as formation of functional DNA repair foci after ex vivo tumor irradiation—have been described as a potential biomarker for predicting radiosensitivity.¹⁹

Finally, recent efforts have focused on developing platforms for establishing patient-derived organoid and xenograft models that can serve as a living biobank to test drugs or other therapy combinations.²⁰ The goal of these approaches is to identify personalized treatment approaches in a time frame that can be useful for individual patients. Organoid and xenograft systems likely reflect more elements of relevant tumor biology than two-dimensional cell culture systems; however, these systems are often more expensive and time consuming to develop and may select for genetic changes that are unique to the experimental system.²¹

Given the expanding role of numerous genomic and experimental platforms for tumor profiling, an important challenge will be developing tools for integrating multi-omic tumor data to derive clinically relevant insights. For radiation oncology, this will also require incorporating data from novel methods for predicting or directly measuring radiation sensitivity. It is likely that the most powerful clinical insights will be achieved by combining analyses of large genomic data sets with functional data derived from experimental systems (Fig 1).

Fig 1.

Opportunities for applying precision medicine tools to guide radiation oncology. IHC, immunohistochemistry.

Candidate Clinical Radiation Biomarkers

DNA repair pathway alterations have been studied extensively as potential biomarkers of radiation sensitivity, and several clinically relevant associations between DNA repair deficiency and tumor biology have been described (Fig 2). Promoter methylation of O6-methylguanine DNA methyltransferase (MGMT) is associated with improved outcomes after radiation with or without temozolomide in glioblastoma multiforme.^22-24 Loss of mismatch repair function confers the microsatellite instability (MSI) phenotype, which is associated with unique clinical properties and predicts response to conventional chemotherapy and immune checkpoint blockade^25,26; however, MSI is not currently a biomarker that directly affects radiation decision making. Somatic mutations in the nucleotide excision repair gene ERCC2 are associated with improved response to cisplatin-based chemotherapy and chemoradiotherapy in muscle-invasive bladder cancer.^27,28

Fig 2.

Genomic biomarkers and potential combination therapies for precision radiation approaches. 5-FU, 5-fluorouracil; AI, aromatase inhibitor; AR, androgen receptor; BET, bromodomain and extra-terminal domain; DNA-PKcs, DNA-dependent protein kinase catalytic subunit; DNMT, DNA methytransferase; EGFR, epidermal growth factor receptor; ER/PR, estrogen receptor/progesterone receptor; FGFR, fibroblast growth factor receptor; HDAC, histone deacetylase; HER2, human epidermal growth factor receptor 2; HIF1, hypoxia-inducible factor 1; LHRH, luteinizing hormone–releasing hormone; MSI, microsatellite instability; PARP, poly (ADP)ribose polymerase; PD-1, programmed cell death–1; PD-L1, programmed cell death–ligand 1; SERD, selective estrogen receptor degrader; SERM, selective estrogen receptor modulator; TMB, tumor mutational burden; VEGF, vascular endothelial growth factor.

Deleterious germline and somatic mutations in homologous recombination (HR) genes such as BRCA1, BRCA2, PALB2, and ATM occur across tumor sites. Although the rates are highest in breast and ovarian cancers, a subset of other tumors also have HR gene alterations or have mutational patterns suggestive of functional HR loss.²⁹ Because HR is a primary repair mechanism for radiation-induced DNA double strand breaks (DSBs), it has been hypothesized that HR-deficient tumors may have heightened radiation sensitivity. However, defining the clinical role of HR gene alterations as a radiation biomarker has been challenging for several reasons.³⁰ Radiation is not part of the standard treatment of ovarian cancer and is typically used in the adjuvant setting in breast cancer, making it more challenging to assess treatment response than in settings where radiation is used to treat gross disease. In addition, a significant fraction of HR gene alterations are germline events, meaning that nontumor cells may also have increased sensitivity to radiation (and thus narrowing the potential therapeutic window).³¹ Some clinical studies of BRCA1/2 mutation carriers treated with lumpectomy and adjuvant radiation for localized breast cancer have reported higher rates of ipsilateral breast event rates (tumor recurrence and/or second primary tumors) among BRCA1/2 carriers, whereas others have found no difference.^32-34 The rate of contralateral breast cancer is significantly higher in BRCA1/2 carriers than in noncarriers,³⁵ and the role of germline BRCA1/2 status in driving clinical management (including prophylactic mastectomy and oophorectomy) is an active area of investigation with important medical and socioeconomic implications.

Homologous recombination deficiency was also recently identified as a relatively common feature of metastatic prostate tumors,³⁶ and germline HR alterations are surprisingly frequent among men with metastatic disease.³⁷ HR-deficient prostate cancers have aggressive clinical behavior and respond to poly (ADP ribose) polymerase (PARP) inhibitors^38,39; however, it is not known if HR alterations predict improved response to definitive radiation for men with localized prostate cancer. Similarly, the prognostic and predictive implications of HR pathway alterations in other tumor types in which radiation is used in the first-line setting—such as gastric and pancreatic cancers—have not yet been defined. In pancreatic cancer, patients with a family history of breast, ovarian, or pancreatic cancer had significantly improved survival after platinum-based chemotherapy compared with patients with sporadic disease, suggesting that germline BRCA1/2-mediated HR deficiency may contribute to treatment response.^40-42 Somatic mutations in the DSB checkpoint kinase ataxia-telangiesctasia mutated (ATM) have been associated with durable responses to radiation for recurrent or metastatic lesions,⁴³and increased IHC staining of the DSB nuclease MRE11 is associated with improved outcomes after radiotherapy in muscle-invasive bladder cancer.^44,45

In addition to DNA repair pathway alterations, numerous other tumor genomic features have been correlated with clinical radiation response (Fig 2). Activating KRAS mutations are associated with resistance to radiation as well as to numerous conventional and targeted systemic agents.^46,47 Oncogene activation can result in increased oxidative stress in tumor cells, and mutations in KEAP1, or the KEAP1-binding region of NFE2L2 (which encodes nuclear factor erythroid 2-related factor 2 [Nrf2]), can result in stabilization of Nrf2 and transcriptional upregulation of oxidative stress response genes. KEAP1 mutations are associated with radiation resistance in cell lines and have been shown to drive tumor formation and radiation resistance in a mouse model of lung squamous cell cancer.^48,49 KEAP1 mutations are enriched in patients with lung cancer who develop local recurrence after radiation, consistent with a role in mediating radiation resistance.⁴⁸

The advent of next-generation sequencing techniques has allowed genomic analyses to be performed from small amounts of formalin-fixed paraffin-embedded tumor tissue.⁵⁰ In addition to characterizing features of primary (pretreatment) tumors, these techniques can also be applied to post-treatment residual or recurrent tumors to provide a window into mechanisms of radiation-induced tumor evolution. Studies comparing patient-matched pre- and post-radiation tumors are now emerging and have the potential to provide unique insights regarding clonal dynamics and mechanisms of radioresistance. Sequencing of pre- and post-chemoradiotherapy rectal tumors suggest enrichment of TP53 or co-occurring KRAS/TP53 mutations in poorly responding tumors,^51,52 and enrichment of KRAS/TP53 mutations were also noted in liver metastases that recurred after proton-based stereotactic radiotherapy.⁵³ Analysis of pre- and post-treatment anal cancer showed enrichment or emergence of mutations in known cancer genes, such as TP53, PIK3CA, and FBXW7.⁵⁴ Interestingly, in both the rectal and anal tumor cohorts, numerous subclonal mutations were gained or lost through treatment, but there was no change in overall mutational burden. Additional studies are needed to more accurately define the dynamics of radiation-induced genomic evolution across clinical contexts.

Single-cell whole-exome sequencing approaches are also being applied in the context of radiation, and in one recent study in esophageal squamous cell carcinoma, disappearance and emergence of distinct clonal populations were noted, highlighting the role of tumor heterogeneity in driving therapy response.⁵⁵ Studies such as these will continue to define the genomic features of tumor response to radiation and may identify mechanisms of resistance that can be therapeutically targeted.

Finally, there is immense interest in the use of minimally invasive techniques, such as circulating tumor DNA (ctDNA) and circulating tumor cell (CTC) analyses, to monitor disease status and treatment response. These liquid biopsy approaches have the potential to more accurately reflect intrapatient tumor heterogeneity and can provide an early and sensitive readout of disease activity and mutational dynamics.⁵⁶ Numerous efforts are underway to optimize these techniques for clinical use, and although many of the early efforts have focused on patients with metastatic disease receiving systemic therapy, several published reports now demonstrate the potential to use ctDNA analysis to monitor response after treatment of localized disease.^57,58 In one example, ctDNA was present in the first post-treatment blood sample from nearly all patients with localized lung cancer who experienced recurrence, and the presence of ctDNA preceded radiographic evidence of recurrence by approximately 5 months.⁵⁹ It seems likely that ctDNA and perhaps other circulating genomic biomarkers will become integrated into clinical radiation oncology practice in the coming years.

APPLYING GENOMICS TO INFORM RADIATION TREATMENT DECISIONS

As studies continue to map the genomic landscape of tumors and link genomic events with clinical outcomes, one important challenge in precision oncology will be to separate the prognostic versus predictive value of individual genetic alterations. It is already clear that many of the common genomic features of tumors have important prognostic implications—that is, the genomic event is associated with better (or worse) clinical outcomes independent of the type of treatment. For example, KRAS and TP53 mutations can be associated with poor outcomes independent of how patients are treated. However, to advance precision radiation oncology, it will also be critical to identify genomic features that, although perhaps not strongly prognostic, are predictive of favorable response to radiation. These are patients for whom radiation could significantly improve disease outcomes.⁶⁰

The ultimate goal of these ongoing efforts is to prospectively apply tumor genomic data to individualize treatment decisions. Although tumor mutational status is used to direct systemic therapy choices in select clinical settings, mutational biomarkers are not routinely used to inform radiation decisions. However, despite the lack of mutation-based biomarkers, several gene expression signatures have been validated as prognostic and predictive tools in a variety of radiation settings.

One of first gene expression signatures to be widely used in clinical practice was the OncotypeDX test, which measures expression of a panel of genes involved in estrogen receptor signaling and cell proliferation.⁶¹ Although the assay was initially validated to predict benefit of adjuvant chemotherapy in estrogen receptor–positive breast cancer, it has now been studied in other contexts and can be used to estimate risk of locoregional recurrence after radiation for invasive cancer⁶² or ductal carcinoma in situ.⁶³ Similar approaches have also been used in the post-mastectomy setting.⁶⁴ Clinical trials are ongoing to investigate if transcriptional signatures can be used to prospectively identify women at low risk of locoregional recurrence for whom post-lumpectomy radiation can be omitted (ClinicalTrials.gov identifier: NCT02653755).⁶⁵

Gene expression–based assays are also being used to inform management of prostate cancer. For example, several commercial genomic classifier tools are available and can be used to predict benefit of post-prostatectomy radiation^66,67 and to predict risk of metastatic disease after definitive radiation.^68,69 When combined with traditional clinical and pathologic variables, genomic features can improve prognostication,⁷⁰ and genomic risk groups are included as stratification variables in several planned and ongoing prostate cancer radiation clinical trials (ClinicalTrials.gov identifiers: NCT03070886, NCT03371719).

Human papillomavirus (HPV) is a strong prognostic biomarker and is associated with improved survival outcomes for patients with oropharyngeal squamous cell carcinoma managed with surgery or definitive chemoradiation.^71,72 Preclinical evidence suggests that the radiation sensitivity of HPV-positive tumors may be driven by impaired double strand break repair.^73-75 Genomic studies have further defined differences in mutational patterns within and among clinical/HPV-defined risk groups, with PI3K pathway mutations enriched in clinically defined low-risk patients and TP53 mutations enriched in high-risk patients.⁷⁶ Together, these data have led to several ongoing randomized trials that are investigating radiation dose de-escalation for HPV-positive head and neck squamous cell carcinomas (ClinicalTrials.gov identifier: NCT01302834, NCT02254278, and others).

Epstein-Barr virus (EBV) infection is associated with nasopharyngeal carcinoma (NPC), and EBV DNA is detectable in the plasma of most patients with NPC. Lin et al⁷⁷ showed that higher pretreatment plasma EBV titers (> 1,500 copies/ml) were associated with worse outcomes after chemotherapy and radiation in patients with stage III and IV NPC. Furthermore, patients with persistently detectable EBV DNA after completion of radiation had inferior survival compared with patients with undetectable EBV DNA. Given the association between the presence of post-treatment EBV DNA and poor outcomes, several clinical trials are investigating adjuvant chemotherapy as a therapy escalation strategy in patients with detectable plasma EBV DNA after radiotherapy (ClinicalTrials.gov identifier: NCT02135042, NCT02363400, NCT00370890).

Another disease in which tumor genomics are being used to guide radiation decisions is pediatric medulloblastoma. Historically, patients with medulloblastoma were defined as standard risk or high risk on the basis of clinical features, and chemotherapy and radiation decisions were driven by clinical risk group. However, elegant work by several groups has identified four consensus molecular subtypes of medulloblastoma with distinct biology and clinical behavior.⁷⁸ Patients with tumors harboring mutations in the wingless-type MMTV1 integration site (WNT) pathway (such as CTNNB1) have extremely good prognosis, patients with group 3 tumors have extremely poor prognosis, and patients with group 4 tumors or tumors with alterations in the sonic hedgehog (SHH) pathway (such as PTCH1, GLI3, MYCN) have intermediate prognoses. Several clinical trials are now investigating tailored treatment approaches on the basis of molecular subtype. For example, the Children’s Oncology Group is examining the role of reduced-dose craniospinal irradiation (CSI) and chemotherapy for WNT-driven medulloblastoma (ClinicalTrials.gov identifier: NCT02724579). Similarly, St Jude’s Hospital is evaluating different radiation and drug combinations across molecular subtypes (ClinicalTrials.gov identifier: NCT01878617): patients with low-risk WNT pathway mutations will receive dose-reduced CSI, and patients with SHH alterations will receive an SHH inhibitor (vismodegib) after four cycles of standard chemotherapy. If successful, these studies have the potential to maximize cure rates while minimizing the known acute and late toxicities of chemotherapy and CSI in pediatric patients.

In current practice, the prescribed radiation dose is defined by tumor histology, with doses reduced when necessary to avoid exceeding normal (nontumor) tissue tolerance. However, preclinical studies and clinical experience demonstrate that tumor radiosensitivity varies significantly both within and among tumor types. Therefore, precision radiation oncology approaches should strive to individualize radiation dose on the basis of specific genomic features,^79,80 and approaches that integrate preclinical modeling with clinical data will be necessary to move toward patient-centric approaches that match radiation dose with predicted tumor radiation sensitivity.

IDENTIFYING GERMLINE BIOMARKERS TO PREDICT NORMAL TISSUE TOXICITY

Acute adverse effects of radiation (ie, those occurring during the treatment course or in the weeks after completion of radiation) are relatively common and are typically reversible. However, a subset of patients develop late complications from radiation, including radiation-induced malignancies, and these late effects can have profound negative impacts on health and quality of life.

Numerous clinical factors—such as patient age, the presence of comorbidities, and the nature of other surgical and medical treatments—can affect the risk of severe late radiation toxicity.^81,82 However, a significant component of this risk is likely genetic.⁸³ Examples of extreme radiation-induced normal tissue toxicity have been described in patients harboring germline mutations in ATM or other DNA repair genes^84,85; however, monogenic associations with severe radiation toxicity are relatively rare, and normal tissue sensitivity in the vast majority of patients is driven by a combination of genetic as well as clinical factors.

No germline factors are currently used to personalize radiation doses. Instead, normal (nontumor) radiation dose limits are typically selected to limit the rate and severity of toxicity to an acceptably low level (often ≤ 5%). In other words, normal tissue dose constraints are defined in the most sensitive patients and applied across the population. Therefore, prospectively identifying the patients at highest risk for normal tissue toxicity could inform mitigation strategies such as dose de-escalation or aggressive supportive care in the most sensitive patients and, conversely, could allow for dose escalation in the majority of the population who are at lower risk for normal tissue toxicity.⁸⁶

Many preclinical and retrospective clinical studies have attempted to identify germline factors that affect radiation sensitivity.^87,88 Candidate gene approaches focusing on genes involved in DNA damage signaling and repair have identified associations between specific germline alleles in DNA repair genes and increased normal tissue sensitivity to radiation, but many of these associations have failed to validate in larger cohorts.⁸⁹ Genome-wide association studies have also identified associations between specific single nucleotide polymorphisms and radiation sensitivity. Although some of these alleles occur in genes known to play a role in cancer biology or tissue injury, many exist in noncoding regions of the genome and/or have no obvious role in damage-related cell signaling. Therefore, additional work will be needed to validate these associations and uncover the relevant biology.

In addition to defining predictive biomarkers of radiation toxicity, efforts have also focused on developing functional assays to profile and predict normal tissue toxicity. For example, increased radiation-induced apoptosis of circulating CD8⁺ T lymphocytes from patients undergoing breast radiotherapy was associated with increased risk of late breast fibrosis.⁹⁰ Ideally, genetic and functional data reflecting the sensitivity of both the tumor and normal tissue to radiation would be used to personalize radiation dose for individual patients.

One of the most feared and life-threatening late toxicities of radiation is radiation-induced second malignancy (SMN). SMNs typically develop in normal tissues adjacent to the treated tumor and have unique genomic features and aggressive clinical behavior.^91-93 The risk of developing an SMN depends on multiple factors, including age at radiation, anatomic site, and radiation dose. Germline mutations in known cancer genes such as TP53 and RB1 can increase the risk of SMN after radiation.⁹⁴ However, accurately predicting SMN risk is challenging in part because radiation-induced tumors often take years to develop; therefore, patients who present with an SMN were often treated using techniques that no longer reflect the standard of care. Ongoing studies to understand the germline factors that drive SMN risk as well as continued efforts to incorporate radiation techniques that minimize normal tissue exposure will be needed to minimize SMN risk for patients.

LEVERAGING PRECISION IMAGING AND NOVEL RADIATION TECHNIQUES

Advances in imaging and dosimetry have led to the development of stereotactic radiotherapy techniques—referred to as stereotactic radiosurgery (SRS; typically referring to treatment of intracranial lesions), stereotactic body radiotherapy (SBRT), or stereotactic ablative radiotherapy (SABR). Radiosurgery (RS)-based treatments are now commonly used in many disease settings and are capable of precisely delivering high doses of radiation in just a single fraction or over a small number of treatments.⁹⁵ RS-based approaches minimize toxicity by using advanced delivery techniques to reduce normal tissue radiation exposure and can also increase patient convenience by decreasing the number of treatment visits.

Currently, the decision to pursue RS-based treatment is dependent on clinical factors such as tumor size and location. However, the biologic response of tumors to RS doses (up to 20 Gy or higher in a single fraction) may be significantly different than the response to conventional radiation (1.8 to 2.0 Gy per fraction). For example, the classic radiobiology tenets of sublethal damage repair and cell cycle redistribution between fractionated radiation treatments are likely less relevant in RS-based therapy.⁹⁶ RS doses may increase the extent of DNA damage or create DNA lesions that are more difficult to repair than those created by conventional doses. In addition, RS doses have been proposed to have effects on tumor vasculature, the innate immune system, and tumor hypoxia^97,98; however, the contribution of each of these factors has not been firmly established, and much of the observed efficacy of RS treatments may be due simply to the ability to deliver a higher biologically relevant dose.^99,100

Tumor genomic features that affect the dose-dependent radiation response have the potential to serve as biomarkers to inform selection of the dose and fractionation scheme. Specifically, RS-based dose escalation may be particularly beneficial in specific populations of cells, such as hypoxic cells or cancer stem cells, that are traditionally considered to be radioresistant. In addition, the potential to induce systemic immune responses using RS is of substantial interest in the setting of combined immune checkpoint blockade.

Radiotherapy with charged particles (such as protons or carbon ions) is becoming increasingly common in some developed countries. Charged particles have distinct physical properties that make them attractive for treatment of specific tumor types.¹⁰¹ In addition to their unique physical properties that can be exploited to minimize dose to normal tissues, charged particles may induce tumor cell death via distinct mechanisms. For example, charged particles may be less dependent on oxidative intermediates to create lethal DNA damage and may therefore be more effective for hypoxic tumors.¹⁰² In addition, emerging evidence suggests that charged particles may elicit a host immune response that is distinct from the immune response elicited by conventional (photon-based) radiation,¹⁰³ and studies reporting associations between specific genomic features and response to particle therapy are now emerging.⁵³

Radiomics and Molecular Imaging

Imaging-based analyses are a particularly promising tool for precision radiotherapy because they are noninvasive and can be used longitudinally to assess tumor response during and after treatment. Radiomics is an emerging field of study that aims to characterize and quantify imaging properties of tumors and correlate these features with available genomic or clinical data to define new image-based biomarkers.¹⁰⁴ Quantifying novel anatomic and functional features of tumors can provide relevant clinical insights, and examples of the potential utility of radiomics-based analyses are now being reported. In one large study, computed tomography–based imaging features such as tumor shape, contrast, and texture had independent prognostic power in several cohorts of lung and head and neck cancers.¹⁰⁵ Radiomics features also correlate with genomic properties; for example, arterial phase imaging traits were associated with a doxorubicin resistance transcriptional program in hepatocellular carcinoma.¹⁰⁶ Functional imaging studies also have the potential to be rich sources of radiomics data. Positron emission tomography (PET) tracer uptake (as measured by standard uptake value) has been studied as a potential prognostic and predictive biomarker in several tumor types,¹⁰⁷ and more complex metrics that incorporate both anatomic and functional outputs are now being explored.

Advanced imaging techniques allow improved visualization of both tumor and normal tissue, and many of these imaging modalities are now routinely integrated into radiation oncology practice. Daily imaging with either two-dimensional (x-rays) or three-dimensional (computed tomography) techniques is used to ensure set-up reproducibility and minimize the effect of interfraction changes in anatomy (such as bowel or bladder filling). Intrafraction motion can also be managed using techniques that account for positional uncertainty in the planning phase or directly track the position of an implantable beacon and adjust the radiation field in real time.

In addition to techniques for ensuring reproducible patient position and accounting for slight variations in anatomy, advanced imaging modalities (such as magnetic resonance imaging or PET) can also be used to measure tumor response during the treatment course. Adaptive planning—updating the radiation plan during the treatment course on the basis of tumor response—has the potential to maximize dose delivery to areas of persistent tumor while sparing normal tissue in areas where tumor has regressed.^108-110 For example, an attractive property of particle therapy is the potential to use PET techniques to monitor in vivo dosimetry. PET imaging can be used to detect positron emissions created from inelastic nuclear collisions of protons with elements in tissues, thereby creating a radiation dose distribution from proton^111,112 or carbon ion¹¹³ radiation. Application of on-treatment PET-based dose monitoring in patients undergoing particle therapy may allow for adaptive radiation planning on the basis of delivered dose to the target and adjacent normal tissue.¹¹⁴

Prospective studies are being conducted in multiple disease settings to study the impact of adaptive on-treatment planning on outcome and toxicity. Early results from single-arm studies suggest that dose escalation using adaptive radiation techniques can be performed safely and are associated with promising local control rates^115-117; however, results from randomized trials (such as ClinicalTrials.gov identifier: NCT01507428) will be required to confirm that shrinking the radiation fields does not compromise disease control by sparing regions that appear normal on imaging but harbor residual microscopic disease.

Identifying and Targeting Tumor Hypoxia

Many tumors contain hypoxic regions, and hypoxic conditions are known to induce tumor cell radioresistance.¹¹⁸ A direct measure of tumor hypoxia can be performed using an oxygenation probe, but this method is invasive and cannot readily distinguish between viable hypoxic regions of the tumor and dead/necrotic tumor. Noninvasive functional imaging–based methods are also available, such as ¹⁸F-fluoromisonidazole PET imaging, which uses a hypoxia-sensitive chemical probe.

Measures of hypoxia can also be estimated from harvested tumor specimens. Although challenges with reagent and assay reproducibility exist, several studies have identified associations between increased levels of cellular hypoxia markers and worse outcomes after radiation. For example, increased IHC levels of hypoxia-inducible factor 1 (HIF1) have been associated with worse outcomes after radiotherapy in several clinical contexts,¹¹⁹ and in a randomized trial for muscle-invasive bladder cancer, only tumors with high HIF1 staining benefitted from the addition of the antihypoxia agents carbogen and nicotinamide to radiation.¹²⁰

Gene expression signatures have also been used to identify tumor hypoxia. Most commonly, hypoxic gene signatures are derived from cell line experiments that compare gene expression profiles between cells grown in normoxic versus anoxic conditions, and the derived signatures are validated in retrospective clinical cohorts.^121,122 Treatment strategies to overcome tumor hypoxia include use of altered fractionation schemes and/or charged particles as well as concurrent use of systemic agents designed to reverse tumor hypoxia or preferentially radiosensitize hypoxic cells, such as via administration of a prodrug that is converted to a cytotoxic compound under hypoxic conditions.¹²³

COMBINING RADIATION WITH SYSTEMIC AGENTS

The addition of cytotoxic chemotherapy to radiation has been demonstrated to improve disease control and overall survival for several tumor types and thus represents the standard of care in a number of clinical contexts. The rationale for combining radiation with chemotherapy was outlined by Steel¹²⁴ and others in the 1970s, and despite preceding the advent of modern molecular cancer biology, this contextual framework remains relevant to understanding the interplay of radiation with chemotherapy and targeted agents.

Systemic agents have the potential to interact with radiation to produce additive, synergistic, or antagonistic effects, and the net biologic response is dependent on both tumor and host genomic features (Fig 2). Despite the widespread use of concurrent chemoradiotherapy regimens, the molecular features that drive response and/or toxicity with these combinations remain poorly understood, and treatment decisions continue to be driven primarily by clinical factors, such as patient age, performance status, and tumor histology, rather than by tumor genomic features. However, given the proven efficacy of combined chemoradiotherapy approaches, future efforts should focus not only on discovering novel targets and therapies but also on defining reliable genomic biomarkers that predict response to existing cytotoxic regimens.

Molecularly targeted therapies have transformed the management of diseases such as chronic myelogenous leukemia and non–small-cell lung cancer. As with conventional chemotherapy, there are several potential therapeutic advantages to combining targeted agents with radiation (Table 1).^125,126 In the primary tumor, targeted agents may radiosensitize tumor cells, resulting in improved local control and/or allowing for radiation dose de-escalation. Targeted agents also have the potential to treat micrometastatic disease that cannot be addressed with radiation alone. The sequence of therapies is an important factor that may affect the interactions between radiation and targeted agents at the cellular and systemic levels, and the best outcomes will be achieved when the correct combination is provided in the correct genomic context with the correct timing (Fig 2).

Table 1.

Representative List of Clinical Trials Combining Radiation With Targeted Agents

Given that DNA damage is a major mechanism of radiation-induced cell killing, there has been significant interest in combining targeted inhibitors of DNA repair with radiation.^127,128 As large genomic studies continue to define the landscape of DNA repair pathway deficiencies across tumor types, it will be critical to match DNA repair inhibitors with the tumor genomic contexts in which maximum benefit can be achieved when combined with radiation. Inhibitors of PARP are the most clinically advanced DNA repair–directed agents and are now approved for the treatment of HR-deficient tumors, such as those with BRCA1/2 mutations.¹²⁹ Preclinical evidence suggests that PARP inhibition may synergize with radiation to kill tumor cells, and clinical trials combining PARP inhibitors with radiation are now open. It is unknown whether adding PARP inhibition to radiation will primarily benefit HR-deficient tumors—the setting in which the largest PARP inhibitor monotherapy effect has been observed—or whether PARP inhibition might best be used to sensitize HR-proficient tumors to radiation-induced DNA damage. In addition to PARP inhibitors, combination trials of radiation plus inhibitors of numerous different kinases that are involved in orchestrating the cellular response to DNA damage (including ATM, ATR, CHEK1/2, WEE1, and PRKDC [DNA-PKcs]) are also underway (Table 1).

Inhibition of receptor tyrosine kinases—with either antibodies or small molecules—represents one of the most well-developed and successful targeted therapeutic strategies to date, and inhibitors of several families of receptor tyrosine kinases are US Food and Drug Administration approved. Epidermal growth factor receptor (EGFR) inhibitors are approved in first-line treatment of several tumor types, including lung, head and neck, and colon cancers, and elegant studies have identified mechanisms of resistance and led to development of second- and third-line agents that target common resistance mechanisms.^130,131

Preclinical evidence has supported a radiosensitizing role for EGFR inhibition,¹³² and indeed, the addition of cetuximab to radiation in patients with head and neck squamous cell cancer was shown to improve tumor control and overall survival compared with radiation alone.¹³³ However, although both cetuximab and cisplatin have been shown to improve survival when added to radiation in this context, the use of all three agents simultaneously did not improve outcomes but did demonstrate increased toxicity, highlighting the importance of understanding the mechanism(s) of interaction and potential toxicities.¹³⁴ In addition, the spectrum of EGFR alterations (gene amplification, activating mutations, and so on) varies across tumor types and has different implications for radiosensitivity.^135,136 Therefore, clearly defining the impact of distinct alterations on radiosensitivity and identifying relevant biomarkers in each setting remains a challenge.

ERBB2 (human epidermal growth factor receptor 2 [HER2]) amplification occurs in a subset of breast cancer and is associated with increased rates of local recurrence.¹³⁷ In preclinical models, HER2 amplification is associated with radiation resistance, which can be reversed by treatment with the anti-HER2 antibody trastuzumab.¹³⁸ Clinically, the anti-HER2 antibody trastuzumab has been shown to be safe and active when delivered concurrently with radiation in patients with advanced breast cancer,¹³⁹ and concurrent anti-HER2 plus radiation trials are now underway in breast cancer and other tumor types with frequent HER2 alterations.

The use of androgen deprivation therapy (ADT) in prostate cancer is one of the earliest examples of a targeted therapeutic approach,¹⁴⁰ and the addition of ADT to RT prolongs survival for patients with intermediate- or high-risk prostate cancer.^141,142 The mechanisms underlying the benefit of combined radiation and ADT have remained incompletely understood, and much of the benefit was believed to derive from spatial cooperativity, with radiation eradicating disease in the prostate and ADT treating micrometastatic disease. However, multiple lines of evidence now suggest that direct ADT effects in the prostate may result in radiosensitization via relieving hypoxia or suppressing DNA repair.^143,144 Radiation upregulates androgen receptor (AR) expression and drives an AR transcriptional program that includes DNA repair genes such as PRKDC (encoding DNA-PKcs) to promote radioresistance,^145-147 and this AR-mediated radioresistance can be countered with ADT.

The androgen receptor is also found in a subset of breast tumors, and increased AR transcript levels are associated with radioresistance and increased risk of local recurrence.^148,149 In preclinical systems, blocking AR signaling with enzalutamide led to increased levels of DNA damage and cell death after radiation, and ADT and radiation cooperate to decrease cell growth in orthotopic breast cancer models.⁹ Interestingly, although the interactions between radiation and estrogen signaling in breast cancer have been extensively studied, there is no standard for sequencing of antiestrogen therapy and radiation in the breast-conserving or post-mastectomy setting.¹⁵⁰

Radiation is also being investigated in combination with modulators of intracellular signaling pathways. For example, activating PIK3CA mutations are common in head and neck squamous cell cancers, and increased activity of the PIK3/AKT serine/threonine kinase (AKT)/mammalian target of rapamycin (MTOR) signaling pathway drives cellular growth, proliferation, and a radioresistant phenotype. Accordingly, depletion or pharmacologic inhibition of PI3K impairs DNA repair and confers radiosensitivity in cell lines with activating PIK3CA mutations.^151,152

Numerous drugs targeting epigenetic processes are also being tested in combination with radiation. For example, the histone deacetylase inhibitor vorinostat has shown radiosensitizing effects in preclinical studies.^153,154 A clinical trial combining vorinostat with temozolomide and radiation in newly diagnosed glioblastoma multiforme showed the combination to be safe; however, the primary efficacy end point was not met.^155,156

As the portfolio of targeted therapies continues to expand, and as targeted agents become approved for first-line use in more settings, significant work will be needed to understand the interactions of these agents with radiation. Given the complex and multifaceted cellular response to radiation, the effects of combining targeted agents with radiation are likely to be unique for each agent. Robust preclinical data will be vital to understanding the potential benefits of combinations, as well as the potential risks for mechanistic antagonism or additive toxicities. The identification of reliable predictive biomarkers for both the targeted agent and radiation will be critical to ensuring that combinations are applied in the most rational way to maximize benefit.

Combining Radiation With Immunotherapy

The discovery and clinical development of therapies that modulate the host immune response has transformed oncology practice, and immune checkpoint inhibitors (anti–cytotoxic T-cell lymphocyte-4 [CTLA-4], anti-programmed cell death 1 [PD1], and anti–programmed death-ligand 1 [PD-L1]) are now approved in numerous clinical settings. Given the widespread incorporation of these agents, there is intense interest in combining immunotherapies with radiation, and the topic has been addressed comprehensively in several recent reviews.^157-159

The interaction between the immune system and radiation has been appreciated for decades^160,161; however, the tumor and host genomic features governing the interplay between immune-directed therapies and radiation are incompletely understood. Remarkable case reports of an abscopal effect, in which radiation drives regression of lesions outside the radiation field, have become more common in the era of immune checkpoint inhibitors,^162,163 but the mechanisms remain unclear, and it is currently impossible to predict which patients may experience a systemic response to localized radiation. It is likely that tumor and host genomic features as well as clinical factors such as radiation field size, dose, and fractionation pattern play a role.

In preclinical models, radiation has been shown to have both immune-stimulating and immune-suppressive effects. Radiation can increase major histocompatibility complex (MHC) class I expression, promote chemokine release, and activate tumor-infiltrating T lymphocytes, which have the potential to promote an antitumor immune response.^158,164 However, radiation can also induce immune suppression via activation of regulatory T cells (Tregs) and upregulation of PD-L1 expression on tumor cells.¹⁶⁵ The potential to overcome these immune suppressive effects provides rationale for combining anti–PD-L1 and anti–CTLA-4 therapy with radiation.¹⁶⁶ However, the ideal sequence and timing of immunotherapy with respect to radiation is not known.

Numerous clinical trials are now investigating radiation with immune checkpoint inhibitors or other immune-modulating agents in localized and metastatic disease settings (Table 1). The addition of adjuvant durvalumab (anti–PD-L1) after chemoradiotherapy for stage III non–small-cell lung cancer increased progression-free survival from 5.6 months to 16.8 months and led to the first US Food and Drug Administration approval of an anti–PD1/PD-L1 agent in a population of patients specifically treated with radiotherapy.¹⁶⁷ Continued progress will rely on integration of these emerging clinical results with correlative genomic studies as well as novel insights from preclinical studies. It is likely that regimens combining radiation and immunotherapy will become standard of care, and precision oncology approaches will be required to define tumor and host genomic features that drive synergy (or antagonism) between radiation and immunotherapy.

REALIZING THE PROMISE OF PRECISION MEDICINE IN RADIATION ONCOLOGY

As systemic therapies continue to improve, the importance of durable local disease control and functional organ preservation will be magnified,¹⁶⁸ and radiation will continue to play a critical role in maximizing patient survival and quality of life. Advances in imaging and dosimetry have allowed radiation to become remarkably precise in its physical delivery; however, recent advances in cancer biology and genomics must now be translated into clinical radiation oncology practice to further maximize the therapeutic potential of radiation.

Many oncology centers are now performing routine genomic analysis of primary tumor specimens from patients. Although these assays often identify genomic features that are known to be prognostic, translating these findings into actionable treatment decisions at the point of care remains a significant challenge (Fig 3).¹⁶⁹ The recent advent of noninvasive biomarkers—such as circulating tumor DNA and circulating tumor cells—will likely become integral techniques in monitoring disease response and identifying resistance mechanisms.

Fig 3.

Model for a radiation oncology precision medicine workflow. ctDNA, circulating tumor DNA; CTC, circulating tumor cell; PDX, patient-derived xenograft; SNP, single nucleotide polymorphism.

Incorporating genomic biomarkers to shift radiation oncology practice represents a unique challenge. Radiation has complex, dose-dependent cellular effects and is often combined with systemic agents; therefore, defining reliable signatures of radiation response will require unraveling intricate interactions among signals and effectors. The ultimate objective of the precision radiation oncology approach is to leverage unique tumor and patient genomic features to inform radiation decisions. Realizing this promise will require continued cross-disciplinary efforts and long-term resolve to ensure that advances in genomics, cancer biology, and radiation techniques are translated to improved outcomes for patients.

AUTHOR CONTRIBUTIONS

Conception and design: All authors

Collection and assembly of data: All authors

Data analysis and interpretation: All authors

Manuscript writing: All authors

Final approval of manuscript: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/po/author-center.

Sophia C. Kamran

No relationship to disclose

Kent W. Mouw

Consulting or Advisory Role: Pfizer, EMD Serono