Plasmid DNA Strand Breaks Are Dose Rate Independent at Clinically Relevant Proton Doses and Under Biological Conditions

Abstract

We investigated the effect of proton FLASH radiation on plasmid DNA. Purified supercoiled pBR322 plasmids were irradiated with clinical doses (≤10 Gy) of protons at ultra-high and conventional dose rates using the Paul Scherrer Institute (PSI) isochronous cyclotron. The proton beam in this clinical facility has been validated to produce the FLASH effect in preclinical models. Plasmid samples were irradiated under various oxygen tensions, scavenger levels, pH conditions and Fe (II) concentrations as these biochemical parameters vary across tissues and tumors. Over the range of doses used, plasmid DNA strand breaks were found to be dose rate independent at all conditions investigated. Irradiation within the Bragg peak and spread-out Bragg peak increased clustered strand breaks, except in the presence of scavengers. With this model system, we demonstrate conclusively that plasmid DNA strand breakage is dose rate independent at doses below 10 Gy and does not constitute a high throughput assay endpoint predictive of the biological effect of FLASH.

INTRODUCTION

Plasmids are short circular DNA strands found in bacteria that have been extensively used in classical radiobiology for evaluating radiolytic strand break yields and their processing under a variety of conditions (1, 2). They have been used to investigate relative biological effectiveness (RBE) of various beam types including photon, proton and very high-energy electron (VHEE) (3–5). Plasmids have also been proposed and studied as surrogate models to assess the possible contribution of DNA damage to the FLASH effect (4, 6, 7), defined as the capability of ultra-high dose rates (UHDR) to minimize normal tissue toxicities in vivo without compromising tumor kill (8, 9).

The FLASH effect has been reported to occur in multiple in vivo experimental models and for all radiation modalities used in the clinic, i.e., photons, electrons, protons, and carbon ions (8). While the majority of studies reporting on the FLASH effect have been conducted using intermediate energy electron beams, the clinical translation of FLASH radiotherapy is most readily achievable through the use of protons. Proton beams have been adapted to produce ultra-high dose rates and possess the requisite energy profiles suitable for the treatment of tumors located at any depth (10). Accordingly, the first feasibility FLASH clinical trial was conducted using a proton beam and a second trial is ongoing (11, 12).

Several hypotheses have been proposed and investigated to explain the FLASH effect. The oxygen depletion hypothesis is based on the idea that UHDR can produce depletion of oxygen that elicits a transient state of hypoxia and resultant radioresistance. Evidence in support of this was originally derived from studies in rodents where the sparing of cognitive decline was reversed by oxygen supplementation with both electron and proton FLASH beams (13, 14). Furthermore, when zebrafish embryos (ZFE) are placed in hypoxic oxygen conditions, the sparing effect is enhanced as compared to those irradiated under ambient oxygen conditions (15). Additionally, in lymphocytes irradiated ex vivo (16), DNA damage assessed by the comet assay was also found to be influenced by oxygen tension. The initial appeal for invoking a role for oxygen in the FLASH effect was logical, given its known role as a potent radiosensitizer (17) and its importance in peroxidation cascades impacting biomolecules such as DNA, lipids and proteins (18, 19). As data has accumulated, the oxygen depletion hypothesis has now largely been rejected. Recent experimental results show that radiation-induced oxygen depletion induced by UHDR is too low to explain the multitude of effects observed in vivo (20–22). Indeed, classical radiochemical studies (23–25) have shown that doses of 100Gy are required to deplete oxygen by 3%, doses that far exceed those necessary to observe the in vivo FLASH effect. In 2019, Spitz et al. proposed that differences in the redox metabolism between normal tissue and tumors could explain this isoefficient tumor kill (26), however formal experimental demonstration remains to be done.

The differential production of reactive oxygen species (ROS) has also been investigated in early studies. Yields of H₂O₂, an end product of the water radiolysis cascade used as a surrogate of free radicals, was found to be lower after UHDR than after conventional dose rate (CDR) exposure (7, 13). Additionally, amifostine, a ROS scavenger and well-known radioprotector, was shown to protect ZFE from radiation at CDR but not at UHDR (13), supporting the lower production of ROS after tissue exposure to FLASH. These experimental measurements contrasted with the idea that the density of ROS created was higher, thereby increasing the chances of interaction and recombination of those species (27). Nevertheless, one relevant question is to understand if and how FLASH affects biological targets, in particular DNA, the canonical target of ionizing radiation. One possible hypothesis is that FLASH could induce less DNA breakage, thus contributing to normal tissue sparing.

This study was therefore conducted to assess whether a simple and reliable plasmid assay could mimic the in vivo response to FLASH, and as such, represent a high throughput assay amenable to test the physics parameters required to trigger the FLASH effect in vivo. Here we report our findings using plasmids to probe the effect of UHDR and CDR radiation, a system compatible with controlled manipulations able to mimic defined chemical environments.

MATERIALS AND METHODS

Proton Beam Description and Parameters

Irradiations were conducted at the Paul Scherrer Institute (PSI) in Villigen (CH) with a cyclotron-accelerated quasi-continuous proton beam. This cyclotron is routinely used for clinical treatment and the experimental beamline and conditions used in these experiments were shown to produce the FLASH effect (i.e., to preserve neurocognitive capacity of mice while controlling tumor growth after UHDR exposure) (28). The samples were irradiated in the transmission region (TR) of the depth dose curve with an incoming energy of 250 MeV at both conventional dose rate (CDR, 1 Gy/s) and ultra-high dose rate (UHDR, 1,400 Gy/s). Two additional similar experiments were also performed in the Bragg peak (BP) or spread-out Bragg peak (SOBP) with the same incoming energy and with the following dose rates: CDR = 1 Gy/s. UHDR = 1,300–1,400 Gy/s for BP or 500 Gy/s for SOBP. Beam parameters are shown in Supplementary Tables S1, S2 and S3³ (https://doi.org/10.1667/RADE-24-00118.1.S1). The proton linear energy transfer (LET) at the given energy is estimated at approximately 8–12 keV/μm in BP/SOBP compared with 0.5–0.8 keV/μm in TR (29).

Geometry of Irradiation

For TR and SOBP irradiations, polypropylene (PP) PCR tubes containing 20 μl of the plasmid solution at 40 ng/ μl were placed in Plexiglas cylindrical holders confined in a horizontal Plexiglas block (Fig. 1a and b). PP PCR tubes were used with the proton beam since they are water equivalent (glass vials would have significantly altered the dose prediction and accuracy of our experiments). Furthermore, the quality of the PCR tubes was optimized; to avoid any potential chemical contamination due to external sources, all the PCR tubes were autoclaved before irradiation, ensuring a sterile environment and eliminating any possible contamination. For BP irradiations, 45 ml of solution at the same concentration were placed in a well formed Plexiglas holder (Fig. 1c), to position the whole BP over the sample despite its limited width (Supplementary Fig. S1; https://doi.org/10.1667/RADE-24-00118.1.S1). Oxygen cannot be controlled in the BP setup, as samples are placed in open-air holders and not in tubes; thus only TR and SOBP irradiations were performed with low oxygen. For SOBP irradiations, a 3D printed polylactic acid (PLA) ridge filter, optimized using a procedure from a previous study (30), was placed in front of the beam nozzle to spread the range of the particles (Fig. 1d), in such a way that the peak spread was sufficient to cover the whole PCR tube (Supplementary Fig. S1; https://doi.org/10.1667/RADE-24-00118.1.S1). Additionally, as the ridge filter scatters the beam laterally, a copper collimator is introduced to cut the field to the required size. The collimator width was chosen such that no ripples in the dose distribution due to the ridge filter structure were present at the exit of the collimator. Twenty-two mm Plexiglas build-up caps were placed in front of each sample for better dose uniformity along the sample. For the BP and SOBP irradiations, the energy of the proton beam was reduced to the required range using a gantry internal and an external range shifter. The external range shifter width was chosen to ensure that the highest dose in depth (corresponding to the flat region of the dose distribution for SOBP) was delivered at the sample location.

FIG. 1.

Experiment setup at PSI Gantry 1 for irradiation. Eppendorf 0.2 mL tubes were placed in cylinders (panel b) for TR and SOBP irradiation, while samples were directly pipetted in cylinder wells (panel c) for BP irradiation before inserting the cylinders in the sample holder (panel a). Samples were subsequently positioned in front of the beam nozzle to be irradiated and a Faraday cup was placed behind it as a beam dump for TR irradiation. MicroDiamond was placed in one of the sample holders and Gafchromic film was placed against the sample holders for dosimetry. For SOBP irradiation the ridge filter as well as an external range shifter were placed between the beam nozzle and the sample holder (panel d).

Samples at 21% oxygen in pure water were irradiated with doses ranging from 2 to 10 Gy due to a higher radiosensitivity while samples at 1% were irradiated with doses ranging from 5 to 30 Gy, and samples with DMSO were irradiated with doses ranging from 10 to 50 Gy.

Dosimetry

Doses and dose rates were determined using a synthetic single-crystal PTW microDiamond detector (https://www.ptwdosimetry.com/en/products/microdiamond, Freiburg, Germany). For TR, consistency of sample irradiations was checked using a Faraday cup (also acting as a beam dump), while for BP and SOBP the consistency was checked using a PTW TM7862 chamber placed at the nozzle exit (Fig. 1a). Gafchromic films (EBT3 films, Bridgewater, NJ) were used as a postirradiation dosimetry check of the field size stability for TR and BP irradiations. Uncertainties on the delivered dose and dose rate were below 4% and 9%, respectively (details are shown in Supplementary Tables S1, S2 and S3; https://doi.org/10.1667/RADE-24-00118.1.S1). Further details on the set-up and dosimetry are found in the literature (31–33).

Plasmid Preparation

pBR322 (Thermo Fisher) was purified using membrane dialysis, Slide-A-Lyzer 3.5K (Thermo Fisher) and the concentration after dialysis was measured. The purified plasmid solution was diluted in UltraPure^™ RNAase- and DNAase-free distilled water (Thermo Fisher) (40 ng/μl).

Hypoxic samples, water and plasmids were placed at least 24 h before irradiation in a hypoxia hood (CoyLab Inc.) equilibrated at 1% oxygen, on ice to limit evaporation. The PP PCR tubes were also incubated prior to the experiment to ensure the plastic did not hold residual oxygen. Oxygen level stability in the tubes was checked using an oxygen monitor (OxyLite^™).

Ultrapure water had a pH of approximately 5. For the pH specific experiments, the pH of the solution was adjusted using NaOH and HCl. Dimethyl sulfoxide (DMSO), a ROS scavenger, or iron sulfate hydrate Fe (II) were diluted in water at the desired concentration, respectively (14 mM for DMSO and 1.5 μM/5 μM for iron sulfate hydrate).

Quantification of DNA Strand Breaks

After irradiation, DNA breaks were resolved using agarose gel electrophoresis (0.8% agarose in 0.5X TAE; SYBR Safe was used as an intercalating agent). The compact supercoiled form and the two relaxed forms, open circular and linear were quantified using densitometric analysis (UVITEC Cambridge, UK) and an ImageJ gel analyzer (34). Each sample was normalized to the sum of quantified plasmid in the sample itself, meaning together supercoiled, open circular and linear plasmids equaled 100%, ignoring the fragmented plasmid contribution (35).

As supercoiled plasmid does not bind to the intercalating agent as well as the other forms due to its closed structure, a correction factor of 1.31 was applied. The correction factor was determined by computing the difference in intensities from the same concentration of supercoiled and open circular plasmid ran using the protocol described above.

Statistical Analysis

For each dose, mean and standard deviation were computed from multiple data points (≥4) and the McMahon model (35) was fitted to the collected data using the symfit Python package to perform least squares regression using the L-BFGS-D algorithm. Resulting parameters are single-strand break (SSB) and double-strand break (DSB) rates per Gy, denoted as β_S and β_D as well as the standard errors of those parameters. They represent the expected number of breaks per Gy in a single molecule of plasmid. This assay is considered semi-quantitative due to the intrinsic limits of the plasmids model and the gel electrophoresis method. The ratio β_S/β_D is used as an indicator of clustered strand breaks (36).

RESULTS

DNA Strand Breaks are Dose Rate Independent Under Atmospheric Oxygen Conditions

At 21% O₂, DNA strand breaks in pBR322 were dose dependent (Fig. 2a and Table 1) with the supercoiled (non-damaged) form gradually being converted into open circular and linear forms as the number of SSBs and DSBs increased. As SSBs were more frequent than DSBs the open circular form increased faster than the linear form. A dose of 1.8 Gy was sufficient to convert 50% of the plasmid into a circular form and 90% conversion was achieved above 7.5 Gy independent of the dose rate, while the yield of DSBs was lower and reached 6% at 10 Gy.

FIG. 2.

Fraction of plasmid in each of the three forms (supercoiled, open circular, linear) in ultrapure water under ambient oxygen conditions (21%) (panel a), ultrapure water with 14 mM DMSO under ambient oxygen conditions (panel b) and ultrapure water under hypoxic oxygen conditions (1%) (panel c). Results from representative agarose gels are shown for each condition (top row is open circular, center row is linear and last row is supercoiled) and data are represented as squares, circles and triangles with standard deviation for the fraction of each form as well as the error on the dose, at different doses irradiated in TR. The McMahon model was fitted to the data as shown as full, dashed or dotted lines. UHDR and CDR are respectively shown in purple and green.

TABLE 1.

β_D and β_S Resulting from Fitting the McMahon Model to the Data Using the Symfit Package Along with the Standard Deviation

Condition	βS	std βS	βD	std βD
1% O2 CDR	0.22	0.01	0.0045	0.0006
1% O2 UHDR	0.21	0.01	0.0039	0.0006
21% O2 CDR	0.34	0.01	0.0040	0.0007
21% O2 UHDR	0.38	0.01	0.0042	0.0006
21% O2 + 14 mM DMSO CDR	0.025	0.001	0.0009	0.0002
21% O2 + 14 mM DMSO UHDR	0.030	0.001	0.0010	0.0003

Note. For ultrapure water under ambient oxygen conditions (21%), ultrapure water with 14 mM DMSO under ambient oxygen conditions and ultrapure water under hypoxic oxygen conditions (1%) for TR irradiations.

DNA Strand Breaks are Dose Rate Independent Under Scavenging Conditions

To investigate the impact of a scavenger, DMSO (a hydroxyl free radical scavenger) was added to mimic the intracellular scavenging capacity and probe the indirect contribution of ROS on plasmid strand breaks after UHDR and CDR exposure. A concentration of 140 mM of DMSO, which corresponds to the intracellular scavenging capacity, fully protected the plasmid from radiation damage (Supplementary Fig. S2; https://doi.org/10.1667/RADE-24-00118.1.S1). Subsequently, the DMSO concentration was decreased down to 14 mM, a concentration 10 times lower than cellular antioxidant levels, while the dose of radiation was increased to 50 Gy (Fig. 2b and Table 1). DMSO significantly protected the plasmid from radiolytic damage, with a dose modifying factor (DMF) of about tenfold as a dose of 23 Gy was required to convert 50% of the plasmid to the open circular form and a dose of 50 Gy produced a yield of 7% DSBs. Again, DNA strand breaks were dose rate independent.

DNA Strand Breaks are Dose Rate Independent Under Hypoxic Conditions

To investigate the impact of lower oxygen concentration, plasmids were irradiated in 1% oxygen (Fig. 2c and Table 1). Under these hypoxic conditions, the plasmid was also protected from radiolytic damage, but the DMF was only twofold, as a dose of 3.9 Gy was required to convert 50% of the plasmid to the open circular form. Once more, DNA strand breaks were dose-rate independent.

Higher LET Increased the Ratio of Clustered Strand Breaks but Does Not Show a Dose Rate Dependency

The experiments described above were repeated by exposing plasmids to higher LET radiation. At ambient oxygen levels, as a reference condition, plasmids were irradiated in the BP and SOBP. As hypoxia cannot be maintained for our BP setup, only SOBP irradiation could be conducted at 1% oxygen. For irradiation with DMSO added, only BP irradiations were conducted. Those experiments led to the same conclusions: DNA strand break yields were dose rate independent, as shown in Fig. 3a and Supplementary Table S4 (https://doi.org/10.1667/RADE- 24–00118.1.S1). For the samples irradiated in pure water, at 1% and 21% oxygen tension, the ratio of SSBs to DSBs (an indicator of clustered strand breaks) decreased, suggesting an increase of clustered strand breaks (Fig. 3b). Clustered strand breaks increased with increasing LET (BP > SOBP > TR). In the presence of DMSO however, this ratio was not significantly modified for BP irradiation compared to TR irradiation.

FIG. 3.

Panel a: Fraction of plasmid in each of the three forms (supercoiled, open circular, linear) for ultra-pure water under ambient oxygen conditions (21%), ultra-pure water with 14 mM DMSO under ambient oxygen conditions and ultra-pure water under hypoxic oxygen conditions (1%). Data are represented as squares, circles and triangles with standard deviation, at different doses irradiated in BP or SOBP. The McMahon model fitted to the data is shown as full, dashed or dotted lines. UHDR and CDR are respectively shown in purple and green; panel b: β_SSB/β_DSB ratios of the four plots above and in Fig. 2 are presented as histograms, grouped by irradiation method. Increased ratio corresponds to less clustered strand breaks.

Mimicking the Tumor Chemical Environment Increased the Yield of DNA Strand Breaks

Lastly, we investigated the tumor chemical environment by decreasing the pH and adding labile iron (Fig. 4). An acidic milieu enhanced the yield of DSBs but did not show a clear effect on SSBs, while increasing labile iron concentrations enhanced the yield of SSBs but had no discernable impact on DSB yields as reported in Supplementary Table 5 (https://doi.org/10.1667/RADE-24-00118.1.S1). More importantly, while these chemical modifications altered DNA strand breaks yields, strand break production was again dose rate independent.

FIG. 4.

Fraction of plasmid in each of the three forms (supercoiled, open circular, linear) for ultrapure water with varied pH (4, 5 and 7.2) (top row) and varied Fe (II) concentrations (0 uM, 1.5 uM and 5 uM) (bottom row) under hypoxic oxygen conditions (1%). Data are represented as squares, circles and triangles with standard deviation, at different doses irradiated in BP or SOBP. The McMahon model fitted to the data is shown as full, dashed, or dotted lines. UHDR and CDR are respectively shown in purple and green.

DISCUSSION

Our experiments conducted using a quasi-continuous proton beam in a clinical facility showed that induction of DNA strand breaks in pBR322 plasmids was dose rate independent at ambient oxygen levels, whether in the presence or absence of DMSO or under hypoxia. Additionally, comparison of BP/SOBP irradiations to TR revealed that LET does not impact the observed dose rate independence of DNA strand break induction. When studies were conducted to mimic the tumor chemical environment, i.e., low pH and high Fe (II) concentration (37, 38), DNA strand break induction was enhanced but remained dose rate independent.

The first objective of our study was to perform a systematic and comprehensive evaluation of DNA strand break induction by proton FLASH vs. CDR under various conditions. This was conducted to investigate the possible dose rate dependency of DNA strand break formation and to validate the relevance of this simple model for the investigation of various physical parameters and chemical conditions required to produce the FLASH effect.

First, our results show no influence of the dose rate on DNA strand breaks when this simple system was used at ambient oxygen levels. Neither the addition of scavenger nor hypoxic conditions modified this outcome and DNA break yields remained dose rate independent but were globally reduced as expected. Similar results were reported by Milligan et al. in the 1980s with dose rates ranging from 0.001 Gy/s up to 1.0 Gy/s (2) and more recently when plasmid was irradiated with IEE and VHEE beams (4, 7, 39) under FLASH and CDR conditions. However, two studies have shown a dose rate dependency of DNA breaks in pBR322 plasmid (6, 40). The first study showed conversion of the plasmid into open circular or linear forms at high doses (20–30 Gy) under atmospheric oxygen conditions in a dose rate-dependent manner, while the second study found a difference in single-strand break yields under one condition (10 mM Tris) and above 90 Gy. In contrast, we examined smaller doses (<10 Gy in water) and provided compelling evidence that the formation of DNA strand breaks is dose rate independent at doses relevant for the in vivo FLASH effect. For the aforementioned published articles claiming otherwise, the purification status of the plasmid preparation likely explains these differences. However, it is possible that plasmid conversion is dose rate dependent at high doses (for example above 28 Gy for scavenger-free medium) but dose rate independent at clinical doses (below 10 Gy).

Next, we investigated the impact of LET on pBR322 DNA strand breaks, where the LET of the protons ranged from 0.5 (TR) to 12 keV/μm (BP fall off) (29). Results showed that the LET did not modify the dose rate independence of DNA strand breakage but enhanced clustered strand break yields, consistent with findings reported earlier in the literature for this LET range (36). As expected, the impact of LET on clustered strand breaks was not observed in the presence of the free radical scavenger DMSO, while for a given LET (either in TR or in SOBP), samples irradiated under hypoxic conditions showed more clustered strand breaks than samples irradiated at ambient oxygen concentration. Indeed, at normal oxygen levels, a higher amount of DNA damaging ROS may be produced, which would increase the proportion of SSBs. Conversely, at lower oxygen levels, the production of ROS is reduced, leading to a lower proportion of SSBs. Consequently, when SSBs are reduced, localized and clustered damage becomes apparent. Thus, a decrease in the fraction of SSBs correlates with a relative increase in DSBs.

Lastly, we investigated the impact of low pH and high Fe (II) concentration on pBR322 DNA strand breaks. Our results showed more DNA breaks under these conditions. However, the strand break yield remained dose rate independent. Our study supports the growing evidence that radiation-induced DNA strand break yields in plasmids depend on total dose but are dose rate independent when IEE, VHEE and proton beams are studied over clinically relevant dose ranges.

The second objective of our study was to validate this simple model for use as a surrogate of the in vivo biological effect triggered by FLASH. Our findings show that this is likely an invalid assumption for two reasons. First, we showed no dose rate dependency under any of the described experimental conditions, even when attempting to mimic the chemical milieu of normal tissue and tumors. This is critical, as the FLASH effect was defined in vivo using experimental models able to discriminate the differential impact of dose rate modulation between normal tissue and tumors, as reviewed elsewhere (8, 9). Additionally, under hypoxia we observed less strand breaks in plasmids suggesting similar outcomes in hypoxic tumors both by FLASH and CDR, in marked contrast to our recent studies demonstrating that hypoxic tumors remain sensitive to FLASH but not CDR (41). Second, the role of DNA as a critical target responsible for the FLASH effect is not clear. Indeed, the impact of dose rate on DNA damage and repair remains a matter of debate. In vitro experiments in pulmonary cells showed less γH2AX foci after FLASH exposure (22). These findings were corroborated by COMET assays performed ex vivo on lymphocytes that showed a 4% relative reduction in DNA strand breaks after IEE FLASH under ambient oxygen conditions (21%) a 14% reduction at 1% oxygen and 21% reduction at 0.5% oxygen (16). However, other studies using the HTGTS-JoinT-seq assay (42) showed no impact of the dose rate on chromosome translocations and junction structure and more importantly, in vivo studies showed no differences in residual levels of γH2AX foci after 24 h in murine gut, brain and tumors (43–45). Those latter results combined with ours suggest convincingly that DNA damage is not a primary determinant of the FLASH effect, and that by extension, plasmids might be an irrelevant model for studying the mechanisms of the FLASH effect.

To conclude, plasmid conversion was consistently shown to be independent of dose rate under a variety of carefully controlled conditions using a clinically relevant dose range. Nonetheless, further studies on plasmids might look at higher doses, using atomic force microscopy to further evaluate plasmid DNA strand breaks (46), as gel electrophoresis does not provide the resolution for analyzing heavily damaged plasmids, much less DNA base damage. Gel electrophoresis is also limited in the evaluation of small differences for the same reasons; this is part of the reason this assay is considered semi-quantitative. Another reason is the DNA molecule length (4,361 bp). This was apparent in our study by the necessity to use DMSO concentrations 10× lower than in cells, as the inherent scavenging capacity necessitates doses on the order of 1,000 Gy to obtain observable strand breaks using gel electrophoresis on small target DNA molecules. The use of alternative thiol scavengers able to compete with oxygen for interaction with DNA radicals, thereby reducing the amount of oxygen-fixed damage, could also inform about the radiochemical effects following UHDR radiation. Other non-nucleic acid molecules can also be of interest as suggested by the recent results obtained with fatty acids and peptides (18, 19). A comparison study of different biomolecules and parameters such as temperature and viscosity may prove useful in the future as well (47, 48). However, to date, investigations focused on the physicochemical and chemical impact of UHDR irradiation alone have not yielded sufficient insight to understand the FLASH effect.