Proton and electron UHDR isodose irradiations produce differences in reactive oxygen species yields

Abstract

Purpose:

Investigations into ultra-high dose rate (UHDR) radiotherapy have dramatically risen because of the observed normal tissue sparing FLASH effect without sacrificing tumor control. The purpose of this study was to provide a direct beamline comparison of protons and electrons to determine where UHDR to conventional dose rates (CDR) changes affect the resultant radiochemistry.

Methods and Materials:

We used well characterized assays of reactive oxygen species (ROS) and oxygen consumption to assess the radiolysis in protein solutions. Three optical reporters related to ROS (CellROX Deep Red, reflects highly reactive radicals; Amplex Red reflects H₂O₂; and Oxyphor reflects partial pressure loss (ΔpO₂)). A Varian ProBeam proton cyclotron and a converted Varian Trilogy electron linac were used for irradiation at both their CDR and UHDR capable level, to assess the assay change per unit dose.

Results:

For both protons and electrons an expected reduction in ROS was noted going from CDR to UHDR, and results interpreted as a reduction in the ratio of UHDR/CDR yield. The CellROX assay showed no difference between beamlines, each showing ~80% reduction in ROS from CDR to UHDR. The Amplex assay showed the largest inter-beamline difference, with ~5% loss using protons vs ~69% loss with electrons, in protein solution. The Oxyphor assay of ΔpO₂ showed a small difference in CDR to UHDR with a 23% loss with protons and 43% loss with electrons.

Conclusion:

Interpretation of ROS assays and oxygen consumption is notoriously challenging. These assays might be interpreted by their most activating species’ lifetime. The assay for highly reactive OH●, appeared independent of beamline, whereas the assays for the longer lived H₂O₂ species and ΔpO₂ assay showed differences between beamlines via the UHDR/CDR ratio. This work can be used for FLASH hypothesis testing by comparing these assays to isodose biological FLASH effects in vivo.

1. Introduction

In the last decade, several UHDR investigations into the FLASH effect have occurred due to the remarkable biological result of normal tissue sparing, when compared to CDR irradiation at iso-dose levels [1]. While electrons have been widely used in many early FLASH studies [2, 3], the FLASH effect has also been observed with protons [4], photons [5], and Carbon ion beams [6]. As varying beam structures begin to achieve UHDR capabilities, there is growing need to compare beamlines so that we fully understand the comparative FLASH effects and if they are even similar. These comparisons are limited by the large differences between proton and electron sources in temporal beam structures, such that single parameter variation studies are not readily possible between them.

Research into the exact physical mechanism of FLASH is still ongoing with primary hypothesis from oxygen effects [7–9] to fast radical recombination[10]. Investigations into these theories has not yet let to a mechanistic conclusion [11], but still studies of the reactive species chemistry that occur using quantitative assays, can be used as surrogates for elucidating differences and probing mechanisms for some of the differences seen between UHDR and CDR isodose irradiation. In this study we use commercially available optical reporters said to reflect primary radicals, assays of downstream reactive oxygen species (ROS) and oxygen consumption via local oxygen partial pressure changes (ΔpO₂), to compare the differences in radiochemistry of UHDR in both protons and electrons. The function of these reporters is based on the understanding of how primary radical species of water radiolysis that are thought to be responsible for damage, include fast production and short lived highly reactive hydroxyl ions (OH●) amongst other ROS species, and longer-lived species such as hydrogen peroxide (H₂O₂). The loss of oxygen via consumption is directly measurable via ΔpO₂, where the consumed oxygen molecules might be included into peroxyl radicals or reduced to superoxide and/or hydrogen peroxide (H₂O₂). The slower reactions to produce H₂O₂ (e.g. by disproportionation of superoxide) are likely less relevant in model solutions, but are still a part of the downstream reactive milieu. Albumin is convenient mimic of the protein content in vivo, providing substrate for ROS reactions. While these assays and methodology are imperfect, this study is a step towards assessment whether there are differences between UHDR proton and electron beams, which is especially relevant given the lack of biological comparisons between their FLASH effects. These assays might be used to compare with biological endpoints of FLASH between electrons and protons to determine if there are differences or not, and what thereby what the dominating chemical reactions for these might be.

2. Methods

2.1. Dosimetry & Delivery

This study compared effects from an UHDR electron beam (Dartmouth Cancer Center, converted Varian Trilogy linac [12]) to those from a isochronous cyclotron generated proton beam (University of Alabama Birmingham, Varian ProBeam). The Varian Trilogy MeV beam was operated at 10 MeV with repetition rates varied to reach corresponding maximum (360 Hz) and minimum dose rates (60 Hz). The electron pulse width was 4 μs, with 3–16 ms between pulses. The ProBeam generated protons at 250 MeV that are quasi-continuous having a pulse width of 2 ns with 13.7 ns between pulses, resulting in a 72.8Mhz rep rate.

All ROS assays irradiations were irradiated between 8 Gy-30Gy and results are interpreted via the normalized slope per unit Gy absorbed. At the Varian Trilogy UHDR conditions were as follows: 360 Hz. 24 to 42 pulses. Total Dose: 8 to 30 Gy. Average Dose rates: 115–368 Gy/s. For Conventional irradiations the conditions were: 60 Hz. Total Dose: 20Gy, Average Dose rate: 0.1 Gy/s. At the Varian ProBeam the UHDR parameters were: Total Dose: 30 Gy. Average Dose rate: 80 Gy/s. At the Varian ProBeam the Conventional parameters were: Total Dose: 30 Gy. Average Dose rate: 10 Gy/s.

For the oxygen assay exact electron irradiation UHDR parameters were as follows: Pulse frequency: 180. Pulses delivered: 8. Dose: 29.6 Gy. Average dose rate: 660 Gy/s. Electron beam conventional irradiations were as follows: Pulse frequency: 360 Hz Dose: 30 Gy. Average dose rate: 0.1 Gy/s. Proton beam UHDR were as follows: Dose: 15Gy. Average dose rate: 80 Gy/s. Proton beam CDR parameters were as follows: Dose: 15Gy. Average dose rate: 10 Gy/s.

The dosimetry for total dose at both sites was assayed using Gafchromic EBT-XD film to verify delivery doses. The dosimetric validation for dose rate on the Varian ProBeam was Varian’s log file verification which uses the delivery log file to calculate dose-averaged dose rate on a voxel-by-voxel basis via the beam on time, current, and corresponding dose accumulation of the proton as the use of PBS for FLASH research is still developing [13]. For electron deliveries, dose rates were calculated using EBT-XD film readings combined with an FPGA pulse-monitoring system [14] and pulse repetition rate information to determine dose per pulse, instantaneous dose rate, and mean dose rate.

The Varian ProBeam conventional dose rate was 10 Gy/s, which is a typical conventional proton dose rate despite being higher than conventional electron delivery. For proton deliveries, the spot spacing in this setup was 5 mm in a diamond pattern to make a 2×10 cm² field. For electron deliveries, samples were placed at isocenter with a 10×10 cm² field conventionally and a 2×2 cm² during UHDR which has a Gaussian shape that is homogeneous over the small diameter of the cuvette (~0.75cm). Both delivery field parameters were expressly chosen to ensure full coverage of the samples used and achieve maximal dose uniformity throughout the irradiated solutions. No build up was used in these setups as to further ensure uniformity through the cuvette the entrance region of the proton curve and the peak of the electron depth dose curve were used. Due to these conditions the LET of 250 MeV protons is similar to electrons at 10MeV throughout the cuvette (~0.4–1.2 keV )[15].

2.2. Chemical Assays

2.2.1 The fluorescent CellROX deep red reagent assay (ThermoFischer) was used at the manufacturer suggested concentration of 5 µM with both pure water solution, and in a 100X excess of bovine serum albumin (BSA) in water solution. Samples were irradiated in 0.5 mL PCR tubes and read out using a compact Fluorometer (model DS-11 FX, DeNovix, Wilmington DE) using the red excitation at 613–662 nm with NIR emission at 664–740 nm. Measurements were taken immediately before and after the irradiations in each vault and the change in intensity was normalized to dose delivered, after confirming that the response was linear with dose delivered. The dose response analysis of this assay was completed and is shown in Supplemental Information Figure 1 illustrating the linearity of response with increasing dose.

2.2.2 The fluorescent Amplex Red reagent assay (ThermoFischer) was used at the manufacturer recommended concentration of 100 µM in both pure water and in 100X excess BSA in water. Samples were irradiated in 0.5 mL PCR tubes and read out with the same compact fluorometer using green excitation at 490–558 nm with red emission at 565–650nm. Measurements were taken immediately before and after irradiations in each vault and the change in intensity was normalized to the dose delivered. The dose response assay analysis of this assay was completed and is shown in Supplemental information Figure 2 illustrating the linearity of response with increasing dose.

2.2.3 Measurement of ΔpO₂ was accomplished via phosphorescence lifetime assay, using the Oxyphor PdG4 probe (Oxygen Enterprises, Philadelphia PA). The phosphorescence lifetime of PtG4 is directly modulated by molecular oxygen, and the probe is designed for use in biological media and in vivo. Oxyphor PdG4 was diluted to 2 µM concentration in 4% BSA solution, and measurements were performed in sealed tubes. Emission lifetime changes were measured on an OxyLED phosphometer (Oxygen Enterprises, Philadelphia PA) with a sampling rate of 50 Hz, to quantify changes in the pO₂ levels. The OxyLED measurement was via [16] a plastic fiber coupled pulsed light source,. The conversion of the phosphorescence lifetime to pO₂ was performed using a probe-specific calibration parameters, providing a Stern-Volmer fit to lifetime. All samples were oxygenated to air equivalence (~150 mmHg) pre-irradiation. The dose response assay analysis of this assay was completed and is shown in Supplemental information Figure 3 illustrating the linearity of response with increasing dose.

3. Results

3.1. ROS testing

The results of the ROS measurements show, despite substantial differences in beam structure, that protons and electrons are similar in terms of reducing radical generation as a result of dose rate changes from CDR to UHDR. Error bars on all graphs are from prior repeated measures, showing the general accuracy of the assay measurement. In the water/BSA solutions CellROX deep red radical measurement (Fig 1) shows highly reduced ROS production at UHDR compared to CDR, with roughly a 80% overall reduction in the ROS measured by CellROX deep red in protein solutions. Measurements in pure water were even larger changes with a 93% and 94% reduction of ROS at UHDR relative to CDR in protons and electrons respectively showing no significant difference.

Figure 1.

Data of CellROX Deep Red assay for highly reactive ROS yields (such as OH●) with change in fluorescent intensity per unit absorbed dose, as well as the ratio of these delivery time responses in UHDR vs CONV to view proportional differences between beam lines.

The assay for Amplex Red assay for H₂O₂ (Fig 2) showed very different results, with negligible change with proton CDR to UHDR change with only 5% reduction in H₂O₂ production, whereas with electrons there was 69% loss in H₂O₂ production at UHDR relative to CDR, a significant difference. In pure water these results were a 22% reduction for protons and 42% reduction for electrons (not significantly different).

Figure 2.

Data of Amplex Red assay for H₂O₂ yield with change in fluorescent intensity per unit absorbed dose, as well as the ratio of these responses in UHDR vs CONV to view proportional differences between beam lines.

Another important difference between the response two assays themselves is the signal quenching of the H₂O₂ assay as a result BSA in solution, meanwhile the CellROX assay had significant increase in signal change with BSA in solution, as compared to pure water.

3.2. Oxygen consumption measurements.

The results of the ΔpO₂ measurement, shown in Figure 3, show strikingly similar results from protons and electrons in ratio of CDR to UHDR responses. In proton irradiation, the relative ΔpO₂ reduced from 0.45 mmHg/Gy at CDR to 0.35 mmHg/Gy with UHDR, for a 22% drop in the oxygen consumption. In electron irradiation, the relative ΔpO₂ reduced from 0.56 mmHg/Gy at CDR to 0.32 mmHg/Gy with UHDR, for a 43% drop in the oxygen consumption. The difference in change with electron beams is significantly higher than with proton beams. However, electrons do have both a significantly lower CDR (0.1 Gy/s vs 10 Gy/s) and a higher UHDR (80 Gy/s vs. 1000 Gy/s) so the larger overall change of pO₂ levels could be explained by this.

Figure 3.

The oxygen consumption yield (ΔpO₂) per unit absorbed dose is shown for proton and electron beams, for both UHDR and CDR irradiation values.

4. Discussion

These investigations are the first to directly compare electron and proton beamlines for their effects using the exact same assays. While the assays are imperfect measures of the radiochemistry going on, the direct comparison of electron and proton beamlines is extremely timely given the intensity of research into FLASH with both, but with essentially no research ongoing about their comparative FLASH efficacies. Unfortunately, because of the complexity of proton and electron beamlines it is not easy to exactly compare them based upon a single dosimetric factor such as dose rate, because they each have different instantaneous dose rates, times of delivery, pulse structures and independent alteration of many of these factors is not possible over comparable ranges. Nonetheless, we are able to compare them at similar total dose levels, using their standard CDR to their highest UHDR values, and examine the ratio of change.

The radiochemical assays chosen were selected for those that have been most conventionally adopted and are commercially available and validated. The first assays of varying radical production observed under UHDR were shown by Montay-Gruel et al. [8], and the quantification of oxygen consumption has been shown in vitro and in vivo [7–9]. These studies and others have shown that there are known differences in generally decreased yields of certain ROS species and oxygen consumption with increasing dose rate. Very interestingly, the data here appears to show that the magnitude of change between conventional and UHDR values are similar for the shorter lived and more reactive ROS yields by CellROX assay, whereas there is a larger difference in the change for longer lived species H₂O₂ and likely for the loss of oxygen (ΔpO₂). These data contain information about the mechanisms of what might occur in the FLASH effect, as we compare these results to those of electron and proton FLASH in vivo.

In the ΔpO₂ assay, loss of oxygen is partially associated with formation of peroxyl radicals, which could be a direct indicator of oxygen mediated damage to molecules such as DNA. Although experimental conditions occurred at ambient oxygen levels (~20%) as the FLASH effect is primarily observed in oxygenated normal tissue radiochemical differences should be apparent in this condition. Decreased ΔpO₂ would indicate a decreased formation of peroxyl radicals, i.e. fixation of damage, while the primary damage is due to hydroxyl radicals (OH^●). Through systematic comparison of these in vitro assays with in vivo damage assays, we might be able to parse out which are the dominant or fractional radiochemical species causes of the FLASH effect. Additionally, by comparing across beamlines and by comparing beam dose rate parameters, we can further use these assays to determine how parameters such as pulse structure and total irradiation time affect FLASH. While the theory of induced oxygen depletion was initially a proposed theory behind the FLASH effect, our data contradicts this, and would rather suggest that it is the change in oxygen depletion (i.e. variation in ΔpO₂) that matters, and it is where the oxygen goes that is the most significant factor is seeing differences in tissue damage [8].

An important note on the limited experimental conditions here is that conventional and achievable UHDR values are different for proton and electron sources, with the proton values being 10 and 80 Gy/s respectively and the electron values being ~0.1 and 110–660 Gy/s. Notably while the conventional peak dose rates are similar the major difference is that the peak dose rate achievable with electrons is much higher with electrons (10⁶ Gy/s) as compared to protons (550 Gy/s). It is possible that some of the key differences in yields were from these secondary effects of mean or peak dose rate, rather than from the actual particle radiation type itself. Parsing out the relative variations of dose rates versus particle type will be even further challenging. However, the relative changes seen in UHDR to conventional change with the fast primary radical yields from CellROX did not seem significant. The next important step in this type of study is to determine if electron and proton UHDR beamlines produce similar biological FLASH effects in vivo at isodose levels.

Monte-Carlo modeling of various experimental conditions can aid in interpretation, and be used to examine hypothesis testing of the FLASH effects. Current models in the ROS generation and oxygen content changes are ongoing and have helped explain the limits to oxygen depletion levels in UHDR irradiation [17–19]. A limitation to date is that the majority of these models have examined situations in pure water alongside comparison to experiments with pure water. As models continue to expand and allow for implementation of more complex systems, scaling towards in vivo comparisons to experiments [10] will be more interpretative of in vivo data, to understand the mechanisms behind FLASH effect [20, 21].

5. Conclusions

This study compared yields of key radiochemical factors between UHDR versus CDR for isodose electron and proton beams. While the comparisons do not allow exact variation of single dosimetric parameters, we did see very reasonable similarity in the assay yields per unit dose between electrons and protons. There was a smaller difference in the reduction of oxygen consumption change between the two beams with the largest difference between beams being the longest-lived species, H₂O₂. Assays were done in both pure water and albumin solution, with significant differences seen when protein solution was used. In all cases the shift from conventional to UHDR dose rates decreased the yields of ROS or oxygen consumption. These observations provide a step towards understanding if there are differences between proton and electron FLASH. These results and assays may now be used for hypothesis testing of dosimetric parameter variation in FLASH studies to systematically complete paired in vitro radiochemistry and in vivo FLASH effect studies.

Table 1:

Beamline structures and corresponding dose rates for conventional and UHDR experiments.

Particle		Proton	Electron
Machine		Varian ProBeam	Varian Trilogy
Energy (MeV)		250	10
Rep rate (Hz)		72.8×106	60–360
Pulse on time (s)		2×10−9	4×10−6
Pulse off time (s)		13.7×10−9	2.7–16×103
Avg. Dose rate (Gy/s)	CONV	10	0.1
	FLASH	80	115–650
Inst. Dose rate (Gy/s)	CONV	68	108
	FLASH	550	0.2–1×106
Dosimetric validation		Film & log file simulation	Film & Diode

Data Availability Statement for this Work

All data generated and analyzed during this study are included in this published article (and its supplementary information files).