High dose bystander effects in spatially fractionated radiation therapy

Abstract

Traditional radiotherapy of bulky tumors has certain limitations. Spatially fractionated radiation therapy (GRID) and intensity modulated radiotherapy (IMRT) are examples of advanced modulated beam therapies that help in significant reductions in normal tissue damage. GRID refers to the delivery of a single high dose of radiation to a large treatment area that is divided into several smaller fields, while IMRT allows improved dose conformity to the tumor target compared to conventional three-dimensional conformal radiotherapy. In this review, we consider spatially fractionated radiotherapy approaches focusing on GRID and IMRT, and present complementary evidence from different studies which support the role of radiation induced signaling effects in the overall radiobiological rationale for these treatments.

1. Introduction

The success of traditional radiotherapy of bulky or deep-seated tumors is limited by poor blood flow, hypoxia in the tumor, poor depth dose distribution and toxicity to the skin and surrounding normal tissue. Normal tissue cannot tolerate the large radiation doses required to treat the increase in tumor volume, associated with bulky tumors, successfully. Although significant reductions in normal tissue complication have been afforded through the implementation of advanced modulated beam therapies such as intensity modulated radiotherapy (IMRT) in the clinic, emerging evidence suggests additional benefit may be gained by delivering a decreased number of higher dose fractions in some tumor types [1,2]. An additional approach which has the potential to offer further improvement is spatially fractionated radiation therapy (GRID). GRID describes the delivery of a single high dose fraction to a large treatment area which has been divided into several smaller fields with steep dose gradients thus reducing the overall toxicity of the treatment [3,4]. In this review, we consider spatially fractionated radiotherapy approaches focusing on GRID and IMRT, and present complementary evidence from different studies which support the role of radiation induced signaling effects in the overall radiobiological rationale for these treatments.

2. Classification of radiation induced signaling effects

The efficacy of ionizing radiation in cancer therapy stems from its ability to induce cell death as a consequence of DNA damage due to energy deposition in the cellular environment. Cellular radiobiological responses are mediated largely through direct energy deposition in cellular DNA or indirectly through reactive oxygen species (ROS) and other free radicals formed during the radiolysis of water [5].

The classical paradigm in radiation biology which focused on nuclear DNA as the sole target of radiation induced damage has been challenged over the last 25 years with an increasing amount of evidence demonstrating radiobiological effects in cells which are not directly traversed by the radiation field. These effects, termed radiation induced bystander effects (RIBEs) generally describe a range of radiation induced signaling effects that have been observed under different in vitro and in vivo exposure conditions.

RIBEs were first identified by Nagasawa and Little [6] who observed chromosome damage in the form of sister chromatid exchanges in more than 30% of a cell population under conditions in which only 1% of cell nuclei had been targeted using α-particles. Since then, RIBEs have been demonstrated using a range of experimental systems with multiple biological endpoints. Despite increasing evidence in a growing number of model systems, the implications of RIBEs for radiotherapy and cancer risk remain to be fully determined. Whilst conventional approaches to study RIBEs have used techniques somewhat removed from clinical exposure scenarios including media transfer [7] and co-culture models [8,9], characterization of RIBEs occurring in response to advanced clinical exposures such as intensity modulated radiotherapy (IMRT) and GRID will provide additional understanding of their importance in overall radiobiological response.

RIBEs are primarily radiation induced signaling effects that have been shown to be mediated through direct physical cell contact via gap junction intercellular communication (GJIC) [8] or through the secretion of diffusible signaling molecules into the surrounding medium [10,11,12]. The underlying mechanisms mediating response have been extensively studied in a number of model systems and shown to include reactive oxygen and nitrogen species (ROS/NOS) including nitric oxide (NO), cytokines such as transforming growth factor-β (TGF-β) and interleukin-8 (IL-8), which initiate multiple downstream signaling pathways including the mitogen activated protein kinases (MAPKs) and nuclear factor-κB (NF-κB) [13].

Classification of RIBEs is often dependent on the experimental model and exposures conditions which are being investigated. A recent framework for the classification of more general radiation induced signaling effects based on human radiation exposure scenarios was proposed by Blythe and Sykes [14,15] in which effects were classified into three categories; bystander, abscopal and cohort effects. Within this framework, bystander effects are defined for human exposure scenarios as radiation induced, signal mediated effects in unirradiated cells adjacent to a target volume that are exposed to only very low levels of scatter radiation, if any [14,15,16]. These effects are relevant for whole and partial body exposures to very low doses, such as those from background radiation, high altitude flights and ingested radioactive potassium.

The second class of effects are abscopal effects, defined as radiation induced effects in unirradiated tissues occurring distinctly outside of an irradiated volume. Abscopal effects have been observed for more than 60 years as systemic radiation effects in some patients following radiotherapy. They do not appear to be dose dependent, making them particularly relevant to the partial body exposures typically delivered during conformal radiotherapy. Abscopal effects are rarely recognized in the clinic and so their importance in radiotherapy response remains controversial [17].

The final class of effects are defined as cohort effects. These describe the component of overall radiobiological response in irradiated cells which is not a consequence of direct energy deposition in the target cell but rather due to communication between cells within an irradiated volume. Cohort effects are relevant for any exposures where the majority of a cell population is exposed to significant dose and whilst this interpretation is relatively uncommon in the literature, there is increasing evidence that intercellular signaling plays a role in overall radiation response [16].

Although this framework clearly defines different classes of radiation induced signaling effects, which may potentially impact the overall radiobiological responses, it is unlikely that they occur independently in the advanced clinical scenario where patients are exposed to complex spatially and temporally modulated beam profiles with nearby cells receiving vastly different doses. It is consequently difficult to investigate either of these effects in isolation as they may stem from the same or similar cellular signaling origin and may be interpreted as different consequences of the same generalized RIBEs.

3. Observations of RIBEs under modulated beam conditions

3.1. Spatially fractionated radiotherapy (GRID)

Spatially fractionated radiotherapy (GRID) refers to the delivery of a single high dose fraction of radiation by dividing a large treatment area into several smaller fields, thus reducing the overall toxicity of the treatment [3,4]. GRID has been successfully used in the treatment of bulky and deep-seated tumors. GRID may be combined with traditional dose/time fractionated radiation therapy or used along with other treatment modalities, including chemotherapy to achieve better control of bulky tumors and it extends the treatment course minimally.

GRID is not a new technique. In the early 1900s, GRID irradiation was invented and performed originally by Köhler through a perforated screen to successfully deliver higher than normal doses of radiation, safely and without causing complication due to skin toxicity [3]. In the 1950s, GRID was routinely used along with orthovoltage radiation to treat deep seated tumors with minimal skin and subcutaneous tissue toxicity [18]. However, with the development of megavoltage radiation, better depth dose distribution and reduced skin toxicity could be achieved and this method was used in a GRID format to achieve a more efficient treatment of bulky and advanced tumors [19]. GRID involves delivery of high doses of radiation through a specially made GRID collimator or a multileaf collimator, such that the entire target does not receive a uniform radiation dose. Instead, only the target directly under the open areas receives irradiation. Recently, it has also been demonstrated that GRID could be applied to deep seated and irregular geometries using the advanced capabilities of a tomotherapy system [20].

Clinical results of GRID obtained thus far are very encouraging. Most importantly, implementing GRID therapy in a radiation oncology clinic does not require any additional clinical personnel. The already in place clinical staff (physician, medical physicist and dosimetrist) are sufficient to implement a GRID therapy program. In addition, GRID therapy can be easily delivered using a commercially available pre-fabricated cerrobend block with the required opening to opening distance (on/off ratio) and opening diameter. However, even a block such as this is not necessary since the procedure can be delivered with existing multi-leaf collimators on most linear accelerators in a step-and-shoot fashion. With respect to clinical outcomes, Mohiuddin et al. [21,22] constructed a special GRID collimator using 7 cm of lead consisting of 250 equidistant holes. Seventy-one patients with advanced bulky tumors were treated with 15 Gy (6 MV photons) radiation delivered by GRID to various sites including lung, head and neck, gastrointestinal, sarcomas, gynecologic, genitourinary and skin. The patients were treated with GRID alone or GRID followed by fractionated radiation therapy or GRID followed by fractionated radiation therapy and surgery. An overall 75.7% response rate was observed at the end of the study and 78% response rate was observed for palliative treatment and 72.5% for mass effect. A complete response of 16% was observed for all treatments. In a second study 79 patients with bulky tumors were treated spatially fractionated radiation [23]. The overall response in terms of pain and tumor mass management between groups treatment with GRID using a cerrobend block and a multi leaf collimator were found to be comparable. In another study, 27 patients with advanced head and neck cancer were treated with GRID along with radiation therapy (Group 1) or GRID followed by radiation therapy and planned neck dissection (Group 2) [24]. The overall neck control rate was 93% for Group 1 with a disease specific survival of 50% and morbidity was limited to mild soft tissue damage. In Group 2, an 85% pathologic complete response rate was observed, as well as, 92% neck control rate, 85% disease specific survival and surgical morbidity was limited to wound healing complications. Penagaricano et al. [4] used a multileaf collimator based GRID design in a cerrobend alloy 7 cm in thickness. Fourteen patients with advanced head and neck cancers were evaluated. These patients received GRID (6 MV photons) followed by chemo-radiotherapy. The overall control rate of the GRID treated tumor volume was 93%, overall neck and primary tumor control rate was 86%. The overall survival was 64% and disease specific survival was found to be 79% and 57% patients exhibited disease free survival. Other encouraging results have been reported from a clinical comparison of differing collimation approaches by Neuner et al. [23].

Although the GRID dose distribution is non-uniform, the regression of the tumor mass receiving GRID has exhibited uniform regression clinically [4,24]. One plausible explanation might be the enhanced reoxygenation of the tumor following GRID, since the tumor was treated with standard chemoradiation following the GRID dose. A recent study using spatially fractionated microbeam therapy indeed observed marked improvements in tumor oxygenation over the first 2 weeks following exposure [25]. While this suggested that spatial fractionation on the macroscale may also improve tumor physiology, more work needs to be done since the effects of high dose microbeams in terms of direct and bystander influence are likely very different than GRID doses.

Induction of tumor necrosis factor α and ceramide, as well as down regulation of transforming growth factor β1 have been observed in patients, following GRID [26,27]. In addition, GRID therapy leads to increased cytokine production, resulting in broad systemic effects [4]. It is also possible that bystander effects might play a role in killing adjacent non-irradiated or partially irradiated cells. Studies focusing on the type of inter-cellular communication that might exist between the adjacent cells in the open and closed areas of GRID, might provide valuable information on the mechanisms involved in the promising clinical results observed.

3.2. GRID-induced bystander effects

We recently evaluated the ability of spatially fractionated radiation (GRID) to induce bystander effects in murine carcinoma cells following exposure to a single dose of 10 Gy, analogous to the high single doses used in spatial fractionation [28]. The GRID experiments were performed using a Small Animal Conformal Radiation Research System (SACRRS). A programmable robotic arm was used to obtain precise positioning of the target/beam, for both therapy and imaging. During the GRID irradiation, the cells were placed on the “palm” of the robot and aligned with the X-ray beam. The GRID pattern of irradiation was then created by programming the robot platform to move perpendicular to the X-ray beam direction. The cells were then irradiated at 10 Gy using GRID, to create a pattern of nearly 50:50 direct and bystander exposure pattern of 9 circular fields, 12 mm in diameter with a center-to-center distance of 18 mm. Confluent murine mammary carcinoma (SCK) and head and neck sarcoma (SCCVII) cells were irradiated using GRID. The regions that were exposed to 10 Gy irradiation were considered as “directly irradiated” and the adjacent cells which did not receive direct irradiation, but were exposed to indirect radiation (i.e., scattering) which amounted to a valley dose of approximately 1 Gy (shown in Fig. 1), were considered as “bystander cells”. The cells were harvested at various time points and re-plated to determine clonogenic survival. A significant bystander killing in cells adjacent to irradiated regions was observed compared to the sham treated controls. The decrease in survival of cells in the adjacent regions was found to be more than that expected from exposure to only background ‘valley’ or scatter doses, suggesting the existence of true cytotoxic bystander effects following GRID irradiation.

Fig. 1.

Cells were irradiated using spatially fractionated radiation to evaluate bystander effects. The cells were irradiated at a peak dose of 10 Gy using a brass collimator to create a GRID pattern of 9 open circular areas, 12 mm in diameter with a center–center distance of 18 mm. The bystander cells were harvested from the valley dose region along the diagonal lines illustrated, which represents about 10% of the total radiation [28].

Experiments using real-time PCR arrays specific for mouse DNA damage and cellular stress response pathways were used to determine GRID-induced bystander gene expression changes in mouse head and neck carcinoma (SCCVII) cells, compared to sham-treated controls [28]. The bystander (GRID adjacent) cells exhibited increased expression of genes involved in DNA repair, cell cycle arrest and apoptosis immediately following exposure or 4 h after exposure. In some instances the increase persisted up to 24 h post GRID irradiation. Increased expression of antioxidant, heat shock and chaperone genes immediately following irradiation or 4 h post GRID exposure were observed in the bystander cells. A significant increase in expression of these genes in the directly irradiated cells was not observed. In contrast, it has been reported that p53 related genes exhibited minimal activation in bystander cells, while the genes involved in NFκB were activated to equal degrees in direct and bystander cells [29,30] following alpha particle irradiation of fibroblasts. In these studies, a significantly higher level of the genes encoding Glutathione peroxidase, Gpx1 and superoxide dismutase, Sod1 was observed in the bystander cells compared to sham treated controls [28]. These antioxidant enzymes have been known to be important in the cellular defense to oxidative stress [31,32]. It has been hypothesized that proteins able to withstand freezing and thawing, might be responsible for transmitting the bystander signal from irradiated to naïve bystander cells [11,33,34,35,36]. Reactive oxygen species [34,37,38], growth factors and cytokines [39] have been implicated in the maintenance of the bystander signal. Our results suggest that secreted factors that lead to reactive oxygen species are a very likely candidate for the effects observed in vitro, since we observed the greatest increase in expression of antioxidant genes immediately following GRID treatment [28].

3.3. Intensity modulated radiotherapy (IMRT)

RIBEs have been demonstrated using a wide range of experimental approaches at the single, multi-cellular and whole organism level. Increasingly sophisticated clinical approaches have been driven by technological advancements to more accurately target the tumor volume. Whilst early experimental approaches such as media transfer and cell co-culture have provided clear evidence of radiation induced bystander effects, they have largely been conducted under conditions which do not accurately represent the multi-cellular environment or replicate radiation exposures in vivo.

A number of novel approaches have been taken to investigate RIBEs in vitro under exposure conditions which more accurately replicate those during clinical exposures [10,15,40,41,42]. These approaches have focused on characterizing cellular responses following complex spatial and temporal dose distributions which are delivered during typical advanced radiotherapy such as intensity modulated radiotherapy (IMRT). IMRT allows improved dose conformity to the tumor target compared to conventional three-dimensional conformal radiotherapy (3D-CRT). Clinical evidence for IMRT has shown significant improvements in tumor control afforded by dose escalation along with reduced acute and late toxicity in certain tumor types [43]. Although IMRT improves dose distribution to the tumor target, multiple entry fields results in increased volumes of normal tissue exposed to low dose compared to conventional delivery techniques. These dose baths may have important implications for secondary cancer risk, however, the clinical significance of this remain to be fully determined [44,45].

Since 2005 several groups have attempted to characterize RIBEs under more clinically relevant exposure scenarios with respect to beam energy, delivery time and dose distributions (Table 1). The impact of intercellular contact on cell survival following exposure to reduced radiation dose (cold spot) has been studied [42]. The cell survival in the dose cold spot was found to be significantly lower when the cells exposed to the reduced dose were in contact with cells receiving the complete dose compared to cells that received the lower dose but were not in contact with the cells receiving 100% radiation. In a separate study, using a wedge filter to deliver modulated 6 MV photon beam, Suchowerska et al. [46] demonstrated decreased survival in low dose regions with improved survival in high dose regions with both responses being dependent on intercellular communication between the high and low dose regions. This work was complemented by the subsequent study of Mackonis et al. [37] who investigated cell survival under 3 different beam configurations using a multi-leaf collimator (MLC); a uniform field, a quarter field (25% of cells exposed) and a striped configuration (25% of the cells exposed in 3 parallel strips). The authors showed differential cell survival responses in different regions of the beam profiles compared to uniform exposures. These results showed close agreement with that of Suchowerska et al. [60], demonstrating decreased survival in low dose regions and increased survival in high dose regions.

Table 1.

Summary of experimental investigations characterizing radiation induced bystander effects using modulated beam exposures.

Authors	Endpoint	Beam energy	Beam modulation	Observed effect
Suchowerska et al. [46]	Clonogenic survival	6 MV	Wedge filter	Decreased survival in low dose regions Intercellular communication dependent
Mackonis et al. [47]	Clonogenic survival	6 MV	75% MLC shielded	Decreased survival in low dose regions Increased survival in high dose regions Intercellular communication dependent
Syme et al. [48]	DNA damage	6 MV	100% Shielded	27–48% Increase in DNA damage intensity in shielded region
Butterworth et al. [10]	Clonogenic survival	6 MV	50% MLC shielded	Decreased survival in shielded region Increased survival in-field
Trainor et al. [50]	Clonogenic survival	225 keV	50% Shielded	Decreased survival shielded region Increased survival in-field
McGarry et al. [41]	Clonogenic survival	6 MV	50% MLC shield	Decreased survival in shielded region Increased survival in-field No impact of field complexity
Butterworth et al. [40]	Clonogenic survival	178 MeV protons 101–130 MeV protons	Passive scattering Pencil beam scanning	Decreased survival in shielded region Increased survival in-field
Trainor et al. [50]	DNA damage	225 keV	50% Shielded	Elevated 53BP1 foci in shielded region
Sjostedt et al. [42]	Clonogenic survival	6 MV	Shielded cold spot	Decreased survival in cold spot regions

Several studies have further characterized radiobiological responses to modulated fields under different beam configurations and energies. Butterworth et al. [49] determined cell survival responses to modulated and non-modulated field configurations delivered using an MLC generated step wedge. This study showed no difference in cell survival response to uniform dose distributions delivered using a uniform field or parallel opposed step wedges and no difference in response to modulated exposures using shallow dose gradients. A further study by Butterworth et al. [10] determined cell survival responses occurring in- and out-of-field following modulated beam exposures in which 50% of the cells were shielded using a multi-leaf collimator (MLC). In agreement with data from other authors, significantly reduced survival was observed outside of the radiation field compared to the level of response predicted on the basis of scattered dose alone, along with indications of enhanced survival in-field when the cell populations were free to communicate. Additionally, out-of-field responses were shown to be dependent, at least in part on nitric oxide signaling and field size.

A number of reports are now available that further characterize out-of-field responses in different cell lines at different energies, dose rates and using clinically relevant treatment plans [40,41,50]. In addition, these effects have also been demonstrated for clinical proton beams delivered using passive scattering and pencil beam scanning techniques [40].

Whilst these studies have provided evidence for RIBEs under clinically relevant beam profiles, they are limited to two dimensions, lacking cellular architecture and physiological context which may be of significant importance pertaining to the tumor microenvironment. RIBEs have been demonstrated at the whole organism level in a number of in vivo model systems under different exposure conditions [51]. Historically, the first in vivo evidence of RIBEs was provided by the identification of clastogenic factors obtained from the serum of irradiated patients which showed cell damaging activity when transferred onto cultures of unirradiated lymphocytes [52,53]. Since then, most experimental models used to investigate RIBEs have involved partial body exposures and are therefore classified as abscopal effects, which have been observed clinically for many years and were originally defined as systemic radiation effects following local radiotherapy [54].

Experimentally, abscopal effects have been demonstrated in a number of models. Camphausen et al. [55], showed significant reduction in the growth of tumors implanted to the dorsal midline when the legs of the animals were irradiated in C57BL/6. In the same model, Koturbash et al. [56] showed induction of DNA damage in skin tissue up to 7 mm away from the irradiated site following partial body exposure. Another important model which has been used to demonstrate the dose and spatial dependency of abscopal effects increasing tumorigenesis is the Patched-1 (Ptch1^+/−) mouse [57].

Although RIBEs have clearly been demonstrated in vivo, the systems in which they have been investigated do not accurately replicate typical exposure conditions during radiotherapy, presenting an important opportunity to determine RIBEs under conditions analogous to clinical protocols. This may be perhaps possible through the application of tumor bearing animals models in combination with advanced small animal radiation research platforms [58] and presents an exciting opportunity to determine the precise implications of RIBEs for radiotherapy and cancer risk following clinically relevant exposures.

RIBEs are not necessarily restricted to exposure to ionizing radiation. Recently, the ability of genotoxic stress producing agents, other than ionizing radiation, to induce bystander effects have been reported. Chemotherapeutic drugs, such as chloroethylnitrosurea [59], paclitaxel [60], mitomycin C [61,62] and phleomycin [61] can induce bystander effects through secretion of media soluble factors. In addition, photosensitizers [63], heat [64] and photodynamic stress [65] have all been reported to cause some type of bystander effect. Therefore, the true nature of bystander effects appears to be a cell-stress related phenomenon and not necessarily a unique by-product of radiation damage or a specific type of treatment.

4. Perspective

Evidence from a range of studies focusing on understanding the role of RIBEs under clinically relevant exposure conditions from both GRID and IMRT have shown remarkable similarity in radiobiological response. Steep dose gradients associated with IMRT and GRID overall suggest significantly greater than expected decreases in survival out-of-field and unexpected increases in survival infield. These observations may have important implications for radiotherapy and a more detailed understanding of the mechanisms mediating response may result in improved tumor control whilst reducing normal tissue toxicity. These benefits may be realized only by defining RIBEs in the context of the tumor microenvironment where, amongst other important factors, oxygen tension is likely to have a significant impact. Understanding the contribution of RIBEs following high radiation dose exposure will facilitate the development of optimized dose delivery and potent chemoradiation approaches.