Extremely high-dose-rate irradiation, referred to as FLASH, has been shown to be less damaging to normal tissues than the same dose administrated at conventional dose rates. These results, typically seen at dose rates exceeding 40 Gy/s (or 2,400 Gy/min), have been widely reported in studies utilizing photon or electron radiation as well as in some proton radiation studies. Here, we report the development of a proton irradiation platform in a clinical proton facility and the dosimetry methods developed. The target is placed in the entry plateau region of a proton beam with a specifically designed double-scattering system. The energy after the double-scattering system is 227.5 MeV for protons that pass through only the first scatterer, and 225.5 MeV for those that also pass through the second scatterer. The double-scattering system was optimized to deliver a homogeneous dose distribution to a field size as large as possible while keeping the dose rate >100 Gy/s and not exceeding a cyclotron current of 300 nA. We were able to obtain a collimated pencil beam (1.6 × 1.2 cm2 ellipse) at a dose rate of ∼120 Gy/s. This beam was used for dose-response studies of partial abdominal irradiation of mice. First results indicate a potential tissue-sparing effect of FLASH.
Radiation therapy is an effective treatment for primary solid cancers. Further improvement in therapeutic efficacy may be achieved by increasing the dose to tumors within the tolerance limits of surrounding normal tissues. Advanced radiotherapy technologies have increased precision and conformality of dose delivery, however, non-negligible side effects and resulting healthy tissue constraints impede further dose escalation. Renewed focus on the unique normal tissue-sparing benefits of FLASH radiotherapy has triggered great research interest in the phenomenon since 2014. Compared with conventional-dose-rate (CDR) irradiation, extreme-dose-rate photon or electron beams have been shown to spare normal tissues, such as brain (1–4 ), lung (5 ), and skin (6 ), without compromising tumor control (5, 7 ). To date, most experiments have been performed using photon and electron irradiations. These studies used beams generated by modified clinical linear accelerators (8 ), synchrotrons (9 ) or some prototype linacs (10 ). Recently, University of Pennsylvania researchers conducted whole-abdominal proton irradiation at a dose rate of 78 ± 9 Gy/s to test the protective effect of FLASH proton beam. Compared with conventional-dose-rate irradiation, the loss of proliferating intestinal crypt cells was significantly reduced by FLASH proton irradiation (11 ).
Before applying FLASH radiotherapy to the clinic, the mechanism underlying its normal tissue-sparing effect needs to be understood to determine its appropriate use and maximize its potential efficacy for clinical practice. It has been suggested that rapid oxygen consumption by FLASH pulses leads to oxygen depletion or instantaneous hypoxia and enhances radioresistance of normal tissues (12 ). In addition, it has been suggested that normal tissues may exhibit a greater capacity to eliminate radiation-induced reactive oxygen species than tumors (13 ). Experiments are ongoing to test these hypotheses.
Compared with photon therapy, proton therapy has a clinical advantage in dose conformity. Both passively scattered proton beams and scanning pencil beams allow the delivery of target doses with much lower integral dose compared to photon techniques (14, 15 ). Combining the advantageous dose distributions delivered to normal tissue by proton therapy with the potential normal tissue-sparing effect of FLASH may provide a new level of treatment optimization in radiation therapy, allowing for further dose escalation, which can improve tumor control, or alternatively reduced normal tissue toxicity, which is particularly important for long-term survivors.
In this article, the development of a FLASH proton platform using the experimental beamline of our existing clinical treatment facility is presented. The steps from initial design to preliminary results assessing the abdominal tissue-sparing capacity of extreme-dose-rate proton beams are shown.
MATERIALS AND METHODS
Design of a Proton Beam Delivery System for FLASH
Irradiations were performed at the Francis H. Burr Proton Therapy Center at Massachusetts General Hospital (Boston, MA). To avoid conflicts with clinical safety features that are defined for conventional dose rates, the experiments were designed for the experimental beamline. The experimental beamline is a parallel beamline coming from the cyclotron, passing the two gantry rooms and ending in the experimental room. We used a C230 isochronous cyclotron (Ion Beam Applications SA, Louvain-la-Neuve, Belgium) to accelerate the proton beam to 228.9 MeV. The cyclotron can provide a current of up to 300 nA in the standard operational mode, up to 450 nA with minor modifications and potentially higher currents with larger modifications such as changing the ion source. The proton beam is delivered with a radiofrequency of 106 MHz, resulting in bunches of ∼3 ns with a spacing between the start of each bunch of 9.4 ns.
The protons are transported along the beamline to the experimental room, resulting in a spot size sigma of approximately 5 × 7.5 mm at maximum energy. To increase the size and to achieve a homogeneous radiation field across a 12 × 16 mm radiation field (95% falloff), a double-scattering system was designed consisting of a first scatterer with a 0.4064 mm (0.016″) lead foil placed 87.82 cm upstream of the surface of the first irradiation target/mouse, and a second scatter composed of a circular lead sheet of 0.8128 mm thickness (0.032″) with a radius of 1.65 mm placed on a thin mylar sheet positioned 52.29 cm upstream of the surface of the target. The design of the scattering system was optimized using an efficient Monte Carlo code2 and subsequent TOPAS (18 ) Monte Carlo simulations to validate the accuracy of the selected settings.
An aperture was placed 8.56 cm upstream of the target surface with an oval shape with vertical and horizontal axis of 1.2 × 1.6 cm. The aperture consisted of 7.62 cm (3″) thick brass. The mice were irradiated on specifically designed holders of 4 cm width (Fig. 1B). Two mouse holders are placed in parallel on a fixed holder base (Fig. 1C) to allow simultaneous irradiation of two mice to reduce the required beamtime. Each position would have a specific dose level determined by the loss of dose between the two positions.
Dosimetric experiments were conducted to quantify the dose at extreme dose rates and evaluate beam flatness and the size of the radiation field with a combination of thin-gap parallel-plate ion chamber, thimble chamber, Faraday cup, Gafchromic™ film and Monte Carlo simulation.
The saturation of the thin-gap ion chamber was measured using a Faraday cup as baseline. We found that the thin-gap parallel-plate ion chamber had a recombination rate of less than 2% at maximum dose rate by comparing measured doses between the ion chamber and the Faraday cup for a wide range of dose rates from conventional dose rates of a few Gy/min to ∼120 Gy/s with a 300-nA cyclotron current. A thimble chamber was used to test the dosimetry of the scattering apparatus at the center of each of the two mouse positions, in front of the first mouse, in between the two mice, and at the back of the second mouse using a specially-built mouse phantom (Fig. 2A). The thimble chambers were operated only at low dose rates (<0.745 Gy/s) and for low doses (<3 Gy) as an additional test of the absolute doses at different positions along the beamline and for validation of the Monte Carlo simulations. Since the scattering behavior of protons should not change with dose rate, these measurements were used to calibrate dose to monitor unit conversions for different positions. Additionally, the doses were evaluated using EBT3 Gafchromic film at each surface of the two mouse phantoms (front and back) (Fig. 2B).
To measure dose profiles of the small beam spot with high accuracy, we designed a new 2D scanning device constructed from a 3D printer to monitor the flatness of small FLASH proton beams. The vertical and horizontal dose distribution was measured with a DFLR1600 diode providing a 1 × 1 mm2 active area. The diode was attached to a specially-designed holder that was attached in the position of the 3D-printer nozzle carriage, which enabled us to scan the dose profile within a few seconds. Amplified diode signals were acquired using a Lecroy Wavepro 715Zi digital oscilloscope (Chestnut Ridge, NY) and then reconstructed to present the dose profiles (16, 17 ).
Monte Carlo Simulations
Monte Carlo simulations were used for the design of the scattering system and evaluations of doses and dose distributions in the two mice. For the simulations we used the TOPAS Monte Carlo version 3.3 (18 ), which is based on Geant4 version 10.5.p01 (19, 20 ). A schematic view of the simulation setup is shown in Fig. 2C. The source was modeled as an emittance source with a Gaussian spread of 5 mm (7.5 mm) sigma in X (Y) and an energy at the beginning of the nozzle of 228.5 MeV. The mice are represented by water boxes of 2.5 cm thickness and dose distributions are scored by voxelizing the mouse phantom with 1 mm resolution in X and Y and a binning in Z of 0.5 mm. An additional scorer was used covering the central axis of the beamline. A TOPAS parameter file used for the simulations is included in the Supplementary Data (Text S1; https://doi.org/10.1667/RADE-20-00068.1.S1).
The experiments were conducted using 10-to-12-week-old female C57BL/6J mice (Jackson Laboratory, Bar Harbor, ME). C57BL/6 mice (n = 42) were randomly divided into eight cohorts (n = 6 animals in each 19 Gy group; n = 5 animals in the remaining dose groups plus one control group with n = 2) for FLASH or conventional-dose-rate proton irradiation (conventional group). All mice in this study were housed in micro-isolation cages (5 animals per cage), fed sterile pelleted chow, and given acidified sterile water ad libitum. All animal care and procedures were performed in accordance with the Public Health Service Policy on Humane Care of Laboratory Animals and approved by the Institutional Animal Care and Use Committee (IACUC) at Massachusetts General Hospital.
Prior MRI and CT images were obtained from 4 and 3 mice, respectively, with mice in the constraining device to determine the optimal position for irradiations as well as the reproducibility and uncertainty in the setup. The positioning setup is composed of a plastic plate, a block with a gap for the tail and two columns (Supplementary Fig. S1; https://doi.org/10.1667/RADE-20-00068.1.S1). Each mouse was fixed on the plate using two clips pulling the front limbs with rubber bands tested to reduce stress and pain for the mice while providing reproducible setup accuracy and tape to constrain the tail such that the hip is fixed against the block. The maximal distance from the proximal boundary of the bladder to the fixed frame was found to be 2.0 cm based on MRI imaging. Thus, the caudal field border was set 2.0 cm away from the frame to exclude the bladder from the field. The orange ellipse in the CT image indicates the planned 1.6 × 1.2 cm2 lateral radiation field (Fig. 2D). The maximal height of the radiation field was set to be 1.1 cm away from the surface of the mouse holder to protect the spinal cord. The mice were positioned in the entry region of the monoenergetic proton beam. This setup provided sufficient conformality within the field for two mice to be irradiated within the same beam.
The collimation of the proton pencil to the small target size resulted in a significant scattering component within the aperture. For the experiment reported here, we reduced the contribution of aperture scattering by placing 2 cm of polymethyl methacrylate plastic in front of the mice leaning against the aperture as the dose drop-off was much higher in the first 2 cm according to both film dosimetry (Fig. 2E and F) and Monte Carlo simulations (Fig. 2G). While a 2-cm plastic absorber was sufficient for single mouse experiments, in the future the amount of plastic will be increased to 4 cm for experiments with two mice in the beam to reduce inhomogeneous dose distribution in the mice and reduce the dose falloff between the mice. In that scenario, the aperture will be shifted upstream by 2 cm.
Our proton irradiation setup enabled us to deliver a dose to two mice at a time, with a reduction in dose of ∼16% between the two mice and an acceptable enlargement of the field size. The radius of the dose profiles from Monte Carlo simulations (Supplementary Fig. S2; https://doi.org/10.1667/RADE-20-00068.1.S1) increased by approximately 3% from the center of the first mouse position to the center of the second mouse position. The 16% difference in dose can be used to irradiate two dose cohorts (e.g., 16 and 18.6 Gy) at the same time, significantly improving irradiation efficiency.
Prior to irradiation, each mouse was anesthetized with a mixture of ketamine (100 mg/kg) and xylazine (10 mg/kg) and restrained on the mouse holder. The beam profile and positioning enabled the irradiation of two mice at a time as indicated in Fig. 1A. All mice were irradiated in the entrance plateau region of the monoenergetic beam. Each FLASH procedure was completed within 180 ms using ∼120 Gy/s proton beams. The sub-structure of our delivery time consisted of a pulsed beam from the cyclotron. A total of 21 C57Bl/6 mice were irradiated at FLASH dose rates. Each group received a dose ranging from 13 to 22 Gy in 3 Gy increments. The absolute dose and dose rate received by each mouse is shown in Supplementary Table S1 (https://doi.org/10.1667/RADE-20-00068.1.S1). In addition, a conventional cohort of C57Bl/6 mice (n = 21), received a conventional dose rates of 1.9 Gy/min to 4.5 Gy/min, keeping the irradiation time constant at approximately 5 min for doses between 13 and 22 Gy.
At day 90 postirradiation, intestinal tissues were collected from the control and irradiated surviving mice and fixed in 10% neutral buffered formalin. Then the Swiss-rolling (21 ) technique was implemented to prepare tissues for embedding. Briefly, intestine and colon were cut longitudinally and then rolled from the proximal end with the luminal side facing up. The tissues were pinned to keep the roll solid. After embedding in paraffin, the tissues were cut at 5-µm thickness, and routinely stained with hematoxylin and eosin (H&E). These sections enabled us to assess the histopathologic features of damage and repair after irradiation within the whole intestinal and colonic tissues.
Statistical analysis was performed using the Graphpad Prism 7 software (La Jolla, CA) (22 ). Statistical evaluations were performed for the survival data and the weight loss. The comparison of weight loss between the FLASH and conventional groups was implemented using a t test. The log-rank test was used in the analysis of survival fractions. All data are presented as mean ± SEM, and P < 0.05 was considered to be statistically significant.
FLASH Proton Platform Setup and Dosimetry
The highest dose rate measured during the mouse irradiations was 137.88 Gy/s. We found that the delivered dose rate fluctuated between 106 Gy/s and 138 Gy/s during our first irradiations. The system became more stable over time, with much less fluctuations for the final set of mice (between 122 Gy/s and 128 Gy/s); see Supplementary Table S1 (https://doi.org/10.1667/RADE-20-00068.1.S1). In the future, FLASH experiments will be performed after the conventional-dose-rate irradiations to allow for the machine to warm up thereby reducing dose-rate fluctuations.
Reconstructed curves from the 2D dose scanning without aperture showed a field of 1.6 × 1.2 cm2 at 90% of the maximum dose value (Fig. 2H). The beam covered approximately 65% of the volume of the abdomen.
Figure 3A and B shows the survival curves of the eight mouse groups. The mice were followed for 21 days. The first two mice reaching the defined end points were observed at day 6 after 22 Gy FLASH. All conventional and FLASH irradiated mice in the 19 Gy and 22 Gy groups died within 15 days. Compared to the 16 Gy conventional group (two survivors), the surviving fraction of C57BL/6 mice in the FLASH group with corresponding dose level was substantially higher (Fig. 3C, P = 0.049). All mice irradiated with 13 Gy survived. At day 9 postirradiation, the weight loss of 16 Gy conventional irradiated C57BL/6 mice was most pronounced, and significantly differed from the FLASH group in both absolute weight value (Fig. 4A) and percentage weight loss (Fig. 4B), but the difference was significant only at day 9. Two mice survived in the 16 Gy conventional group and their weight recovered completely at day 12 postirradiation after up to 11.8% weight loss. Less time was needed for the weight of the 16 Gy FLASH irradiated C57BL/6 mice to recover their initial weight. The initial weight of the mice can be found in Supplementary Table S1 (https://doi.org/10.1667/RADE-20-00068.1.S1).
Late Intestinal Effect
Nonirradiated mice exhibited no inflammatory characteristics (Fig. 5A and B). Several layers of the intestinal rolls from the 16 Gy FLASH (Fig. 5C and D) and conventional dose-rate (Fig. 5E and F) irradiated mice showed collagen deposition, hyperplastic submucosa and muscularis as well as an infiltration of inflammatory cells. A necrotic lesion was observed in one mouse in the conventional group (Fig. S3; https://doi.org/10.1667/RADE-20-00068.1.S1). Overall, the submucosa and muscle layer in the FLASH group indicated improved repair (thinner layer); however, as only two mice in the 16 Gy conventional group survived, the sample size was not suitable for conducting statistical analysis to assess the late effect difference among these groups.
In this proof-of-principle study, we established a reliable FLASH proton platform for small animal experiments with only minor modification of our clinical facility. Creative usage of the 2D dosimetry scanning system improved the efficiency and precision of measuring the dose profiles. A dose rate of up to 138 Gy/s was achieved using a double-scattering system for irradiating mice at the entrance plateau region of a single pristine Bragg peak. A satisfactory dose distribution was obtained, and doses and dose rates were verified using a combination of radiochromic film, Monte Carlo simulation, Faraday cup, and two different ion chamber measurements. Patriarca et al. also built a FLASH proton irradiation set-up for small animal experiments utilizing the spread-out Bragg peak (SOBP) (23 ); however, their design aimed at much lower dose rates in the order of 40 Gy/s, and included a ridge filter to obtain a spread-out Bragg peak for treatment. The FLASH proton radiation system at the University of Pennsylvania was reported to be able to deliver a proton beam with a dose rate of 60–100 Gy/s and a FLASH effect on intestine was observed (11 ).
Our results demonstrate that the FLASH effect on fast-responding abdominal tissues can be generated by a single dose of FLASH proton radiation. C57BL/6 mice showed notably (while statistically limited) better tolerance to FLASH protons than to equivalent doses for conventional dose rates. Preliminary results from Loo et al. showed a significant survival difference between conventional-dose-rate (0.05 Gy/s) and FLASH groups. The LD50 was 16.6 and 18.3 Gy for mice after 70 and 210 Gy/s electron abdominal irradiation, respectively (24 ). Preserved gut function, reduced cell death in crypts and fewer γ-H2AX foci were observed after FLASH electron irradiation (25 ). In our research, approximately 120 Gy/s partial abdominal FLASH proton irradiation decreased body weight loss and mortality rate at 16 Gy. It should be noted, however, that the current experiments were pure proof-of-concept experiments with low statistical power. A different outcome of one mouse in either cohort would result in non-significant results.
The mechanisms responsible for the FLASH tissue-sparing effect are not yet well understood. Extremely rapid depletion of tissue oxygen by a sufficiently high dose administered within milliseconds is a potential explanation for the FLASH effect. The first quantified demonstration of oxygen depletion by high-dose-rate irradiation was conducted in bacteria in 1959 (26 ). Similar studies were subsequently performed in mammalian cells in vitro (27, 28 ). In 1971, Hornsey et al. found that the protective effect of high-dose-rate electron irradiation in mice could be reduced by breathing oxygen-free nitrogen for 25 s (29 ). At a high dose rate, no effect was observed on anoxic rat feet skin (30 ). The results of a later study also suggested that the protective effect of extreme-dose-rate proton irradiation on skin was linked to oxygen consumption (31 ). Additionally, high-concentration oxygen also reduced or eliminated the neurocognitive benefits of 10 Gy FLASH dose (1 ). Significant tissue-sparing effects after 10 Gy irradiations at high dose rates will likely be observed when the oxygen tension is 3–5 mmHg (32 ). If oxygen depletion is the dominant factor, the dose at which the FLASH effect can be observed will depend not only on the initial oxygen concentration but also on the delivered dose and the dose rate.
Another potential hypothesis involves the sparing of circulating lymphocytes. The immune system responds to the unbalanced tissue environment after ionizing radiation exposure. Lymphocytes are recruited to the damaged sites and modulate inflammatory reactions (33, 34 ). High-dose-rate irradiations reduce the volume of blood and the proportion of circulating lymphocytes that experience radiation damage, which may play a role in the FLASH effect. Genome-wide microarray analysis of mice receiving whole-thorax irradiation suggests the involvement of immune modulation pathways, such as dendritic cell maturation, PKC signaling in lymphocytes, TH1 pathway modulation and calcium-induced T-lymphocyte apoptosis, in the FLASH effect (35 ). However, Venkatesulu et al. reported that 35 Gy/s cardiac or splenic irradiation depleted more lymphocytes than conventional-dose-rate irradiation (36 ).
The “FLASH mode” is typically defined by the mean dose rate of the whole irradiation procedure. However, parameters, such as intra-pulse dose rate, pulse frequency and fraction pattern of the total dose, should also be carefully considered. Judging from previously reported studies, both single-pulse and multiple-pulse rays are able to trigger the FLASH effect (1, 3 ). In our study, a train of pulses were delivered to the FLASH cohorts, resulting in a relatively constant dose rate experienced at the cell scale. If our preliminary results can be reproduced with higher statistical experiments, we can assume that the dose can be delivered with different pulse structures as long as the average dose rate is high enough.
We demonstrated the setup of a FLASH platform using an existing proton infrastructure, allowing us to investigate the FLASH effect of protons on different tissues and organs. The results for our first, low statistical experiments indicate that extreme-dose-rate proton beams were less harmful to the normal gastrointestinal tract than the same doses delivered at normal dose rates.
Table S1. The initial body weight as well as the absolute dose and dose rate received by each mouse.
Fig. S1. The specifically designed holder for mouse irradiation.
Fig. S2. The Monte Carlo simulation result indicated ∼16% dose difference between two mice.
Fig. S3. Tissue necrosis was observed after conventional dose rate proton irradiation.
Text S1. The TOPAS parameter file that is used for the simulations.
We thank Iris Yuwen Zhou and Eric McDonald for helping us with the animal imaging. This study was supported by the Damon Runyon-Rachleff Innovation Award (no. DRR 57-19). We also thank the MGH Radiation Oncology Machine Shop for their help and support.
2 Developed by Bernie Gottschalk.
Editor's note. The online version of this article (DOI: https://doi.org/10.1667/RADE-20-00068.1) contains supplementary information that is available to all authorized users.