May oxygen depletion explain the FLASH effect? A chemical track structure analysis
Daria Boscolo, Emanuele Scifoni, Marco Durante, Michael Krämer, Martina C. Fuss
1 May Oxygen depletion explain the FLASH effect? A chemical track structure analysis
Daria Boscolo a , Emanuele Scifoni b , Marco Durante a,c *, Michael Krämer a , and Martina C. Fuss a * a GSI Helmholtzzentrum für Schwerionenforschung GmbH, 64291 Darmstadt, Germany. b Trento Institute for Fundamental Physics and Applications (TIFPA), National Institute for Nuclear Physics (INFN), 38123 Trento, Italy. c Institut für Festkörperphysik, Technische Universität Darmstadt, 64289 Darmstadt, Germany * Corresponding authors: Marco Durante, Martina C. Fuss Planckstr. 1 64291 Darmstadt Germany Email: [email protected], [email protected]
Keywords
FLASH effect mechanism, radiation track structure simulation, oxygen depletion, reactive oxygen species
Highlights ▪ Bottom-up radiation chemistry study reproduced observed radiolytic oxygen depletion (ROD) with high accuracy ▪ Dynamical nature of oxygen depletion and its impact on OER taken into account for the first time ▪ Demonstrated negligible impact of ROD on radiosensitivity through transient hypoxia in both in vitro and in vivo conditions of reported experiments ▪ ROD impact on therapeutic window is shown to occur eventually in opposite direction Abstract Background and purpose:
Recent observations in animal models show that ultra-high dose rate (“FLASH”) radiation treatment significantly reduces normal tissue toxicity maintaining an equivalent tumor control. The dependence of this “FLASH” effect on tissue oxygenation has led to the assumption that oxygen “depletion” could be its major driving force.
Materials and Methods:
In a bottom-up approach s tarting from the chemical track evolution of 1 MeV electrons in oxygenated water simulated with the TRAX-CHEM Monte Carlo code, we determine the oxygen consumption and radiolytic reactive oxygen species production following a short radiation pulse. Based on these values, the effective dose weighted by oxygen enhancement ratio (OER) or the in vitro cell survival under dynamic oxygen pressure is calculated and compared to that of conventional exposures, at constant OER. Results:
We find an excellent agreement of our Monte Carlo predictions with the experimental value for radiolytic oxygen removal from oxygenated water. However, the application of the present model to published radiobiological experiment conditions shows that oxygen depletion can only have a negligible impact on radiosensitivity through oxygen enhancement, and eventually a detrimental effect on the therapeutic window.
Conclusion:
We show that the magnitude and dependence of the “oxygen depletion” hypothesis are not consistent with the observed biological effects of FLASH irradiation. While oxygenation plays an undoubted role in mediating the FLASH effect, we conclude that state-of-the-art radiation chemistry models do not support oxygen depletion and radiation-induced transient hypoxia as the main mechanism.
Introduction
From the late 1950’s on, pioneering studies reported an enhanced radioresistance when using ultra-high dose rates both in vitro [ ] and in vivo [ ]. In an environment where the radiobiological oxygen effect had been recently discovered [ , ], the first theories aiming at a mechanistic description of the oxygen-induced radiosensitivity, notably the so-called oxygen fixation hypothesis [ ], were progressively developing. Parallel and sometimes overlapping investigations [e.g. , ] on both topics, together with the observation of “breaking” survival curves (displaying a slope corresponding to anoxia after a threshold) under low oxygenations at ultra-high dose rate (10 Gy/s), rapidly put the focus on oxygen depletion as the suspected mechanism behind the increased radioresistance under FLASH conditions [ ]. In the chemical stage of radiation damage of oxygenated targets, a large part of the hydrated electrons and hydrogen radicals produced by water radiolysis react with the dissolved molecular oxygen through e aq ⁻ + O → O ·⁻ (1a) H· + O → HO ·, (1b) producing the cytotoxic superoxide anion and its protonated form, the perhydroxyl radical. Under ultra-high dose rate irradiation, when oxygen is consumed according to reactions (1) and its rediffusion into the irradiated volume can be excluded, a transient radiation-induced hypoxia has been hypothesized to increase radioresistance. A current revival of experimental investigations into the so-called FLASH effect [ , ] with electron irradiation is successfully demonstrating a (differential) sparing effect in vitro and in vivo when applying ultra-high dose rate (>40Gy/s). Beneficial effects observed include reduced pulmonary fibrosis and pneumonitis [ ] and cutaneous lesions [ , , ] and a decrease / absence of late neurocognitive end points [ ] all of which can potentially widen the therapeutic window between tumor control and normal tissue (NT) complication probability. Discussions about the underlying mechanism [ , ] are, however, ongoing and growing especially in the last two years. Some authors [ ] point towards the early chemical stage of radiation damage as the decisive time frame, guided by the duration of the irradiation itself. Besides radiolytic oxygen depletion (ROD), which is still among the most favored theories [ , ], it was discussed whether the differential effect could be justified on the basis of a lipid hydroperoxide / peroxyl radical effect [ ], with Labarbe et al. [ ] offering a very interesting alternative mechanism based on peroxyl radical recombination. Other tentative explanations include a different inflammatory response via the reduced exposure of circulating blood lymphocytes [ ], a decrease in reactive oxygen species production [ ], differences in redox metabolism in healthy vs. tumor tissue [ ] and the selective sparing of hypoxic stem cell niches in NT [ ]. Here, we tackle the question of radiation-induced oxygen consumption based on first principles. Our novel physico-chemical approach focuses on the chemical evolution of electron tracks (ps-µs after physical interactions) and implies a non-constant oxygen consumption rate. We start from the p O -dependent track chemistry of 1 MeV electrons in water in assuming a short radiation pulse (~ms range) which effectively prevents a rediffusion of oxygen intra-irradiation from outside the irradiated volume. The simulated O consumption yield is used to study the transient variation of target oxygenation. Subsequently, the instantaneous oxygenation under FLASH conditions is combined with the radiobiological oxygen effect through the oxygen enhancement ratio and compared to that in conventional irradiation to identify the dose and oxygenation range where a sparing effect is expected. A tentative application of this general analysis to published pre-clinical [ ] and clinical [ ] FLASH experiment conditions is included in appendix C. Using an alternative method, we compute in vitro cell survival [ ] under dynamic vs. constant oxygen pressure. Materials and Methods
Monte Carlo simulations
Figure 1.
Scheme of the physico-chemical electron track model following irradiation of a water target. (B-E) Projection, at different time points, of physical (10 -15 s) and chemical (1 ps, 1 ns, 1 µs) tracks of 600 1 MeV electrons impinging along the z-axis, corresponding to a dose of 1.49 Gy. Simulation performed with the TRAX-CHEM code in a water target in physioxia (oxygen concentration of 91 µM or 7% p O ). The evolution of electron tracks in water through the physical, prechemical and chemical stage is simulated using the TRAX [33] code and its extension TRAX-CHEM [ ]. Implementation of track chemistry in water [ ] and the addition of dissolved oxygen [ ] have been described in detail elsewhere, and a sketch of the simulation procedure is given in fig. 1. In brief, physical particle interactions are simulated with an event-by-event approach and the resulting positions and ionization or excitation levels of the target molecules (10 -15 s, fig. 1B) are handed over to the pre-chemical stage where molecular dissociation and thermalization of fragments and sub-excitation electrons takes place. Then (~1ps, fig. 1C) these radiolytic species generated close to the corresponding physical events are tracked as they diffuse and react with each other [ ], water, and oxygen, and transition from an initially heterogeneous, track-structure-determined dynamic (fig. 1D) to a homogeneous solution readily described by species concentrations at ~1µs (fig. 1E). Inter-track reactions / recombinations (reactions of the radiolytic species of one electron track with those produced by a neighboring one) are not included in the simulations (see Discussion). The total yield of molecules or radicals of each species per deposited energy is recorded throughout the process as time-dependent G-value curve (fig. 1F) and allows to understand the reaction dynamics in the chemical track. Radiolytic yields for net oxygen consumption ( G –O ) and superoxide production ( G O ·⁻+HO · ) are obtained by simulating many 1 MeV electron passages through a defined water volume (2 for physical interactions) containing a specific concentration of O and normalizing the yields originating in that volume by the energy deposited. The estimated statistical uncertainty of the mean values thus obtained is 2%. Dose-dependent species concentrations and OER
Oxygen depletion and superoxide (O ·⁻+HO ·) production in FLASH irradiations can be directly calculated as a function of dose, by numerically integrating the corresponding tabulated p O -dependent yields (units of mol/J equivalent to M/Gy in water) [O ] = [O ] ini – ∫ G –O ([O ]( D )) d D (2a) [O ·⁻+HO ·] = ∫ G O ·⁻+HO · ([O ]( D )) d D . (2b) To that end, we assume sufficiently short total irradiation times (<100 ms) in order to exclude rediffusion of oxygen into the target volume; in practice this gives an upper limit for the oxygen depletion and is equivalent to sealed targets irradiated in conventional mode. Next, OER-weighted dose under dynamic oxygenation is computed as D OER,DYN = ∫ OER([O ]( D )) d D (3) (in contrast to D OER,CONV = D ·OER([O ] ini ) ) and represents the most general approach to include oxygen-enhanced radiosensitivity without the need to consider a specific tissue or cell line. It is particularly useful as a quantity indicating the “effective dose” predicted under varying oxygenation. We use the definition OER( p O ) =D anoxia /D ( p O ) ⃒ same effect (4) and parametrization proposed by Grimes and Partridge [ ] OER( p O ) = 1 + ( Φ O / Φ D )(1–exp(– p O )) (5) ( Φ O / Φ D = 1.63, = 0.2567). In vitro cell survival calculation
In order to assess the impact of the dynamical O at ultra-high dose rates on cell survival [18], dose response curves for all oxygenations fitted to the linear-quadratic model, S ( D ) = exp (– αD – βD ), were taken from the original authors [18]. They hold α constant while β accounts for oxygenation and dose rate regime, β 𝐶𝑂𝑁𝑉 = β [1 − ( pO ) ] + β ( pO ) (6a) β 𝐹𝐿𝐴𝑆𝐻 = β [1 − ( pO ) ] + β ( p,ini20 ) (6b) where β = 0.0025 and β = 0.0194 are the quadratic coefficients under anoxia and normoxia, respectively. To separate the effect of oxygen depletion at FLASH dose rates from other influences in the experiment, we follow this fitting strategy but combine the CONV parameters with the dynamic (instantaneous) oxygen concentration [O ]( D ) according to eq. (2a) and obtain a dynamic quadratic coefficient β 𝐷𝑌𝑁 = β [1 − ( pO (D)20 ) ] + β ( pO (D)20 ) . (7) The survival under dynamic oxygenation is determined as follows. The survival increment Δ S for a dose increment Δ D can be generally expressed as Δ S = S – S = exp (– αD – βD ) – exp (– αD – βD ) (8a) Δ S = S (exp (– α Δ D – Δ D – β( Δ D) ) –1), (8b) so that cell survival at a dynamic p O ( D ) is evaluated iteratively as S DYN ( D n ) = S DYN (D n-1 ) exp((– α Δ D – DYN D n-1 Δ D – β DYN ( Δ D) ). (9) Results
The oxygenation-dependent molecular yield per dose for dissolved oxygen removal for 1 MeV electrons is presented in figure 2. The radiation quality is chosen to be close to the effective energy in a superficial target irradiated with a suitable linear accelerator. In well oxygenated targets, the yield for oxygen removal does not show a dependence on initial oxygenation. In this plateau region the calculated ROD yield of 0.33 µM/Gy (3.23 molecules /100 eV) is in excellent agreement with the literature value of 0.32-0.33 µM/Gy [ , ] and very similar to that obtained for low-LET protons (0.35 µM/Gy [ ]). Below an oxygenation value around 65 µM (5% p O at room temperature), the curve exhibits a steep decrease as the sparse distribution of oxygen in the target slows down the chemical track dynamics sensibly and reactions (1) fail to reach equilibrium within one microsecond [ ]. To avoid any bias induced by time end point selection of the chemical track evolution, all results are presented with a confidence interval (shaded area of figures) obtained by calculating the chemical yields at two alternative time points (0.5 µs, 2 µs) as explained in appendix A. Additional reaction pathways can take place in biological media, such as the reaction of the abundant radiolytic OH· radicals to produce organic radicals R·+H O which subsequently undergo fast reactions with O . In accordance with experimental ROD values in cell culture medium [ ], a tentative oxygen removal yield curve scaled by a factor 4/3 is therefore included. In what follows, the radiolytic yields in water at 1 µs will be regarded as the reference for further calculations, while the shaded area, used to investigate the impact of the time end point and liquid chosen, is propagated as a confidence interval. Figure 2.
Calculated oxygenation-dependent radiolytic yields for ROD at different time points around the end of the track (heterogeneous) chemistry stage (1 MeV electron radiation). The dotted curve approximating cell culture medium is obtained by scaling the 1µs TRAX-CHEM simulation results by a factor 4/3.
In fig. 3A, the oxygen concentration is shown as a function of dose for different initial oxygenations p O covering a typical range from physioxic NT (7%, 5%) to hypoxic tumors (1%) [ ]. In the dose range depicted, oxygen consumption is proportional to dose, and complete “depletion” cannot be observed for any p O . Full depletion has therefore to be discarded as a hypothetical explanation for increased NT sparing under FLASH irradiation conditions. Figure 3.
Dose dependence of oxygen concentration under FLASH irradiation conditions with 1 MeV electrons for different initial oxygenations (A) and its implication on, OER-weighted dose D OER,DYN (B) for a short pulse, at room temperature. Note that in (B), the lines for 3%, 5% and 7% initial p O overlap. While concentration is given in absolute SI units, the initial target oxygenations corresponding to each line are given in % p O for ease of comparison to experiments. However, as proposed recently [ ], radiation-induced hypoxia reduces radiosensitivity [ , ] as described by OER [ ]. Assuming that the radiobiological oxygen effect is a result of the oxygenation level at the moment of irradiation [ ], the dose weighted by the instantaneous OER( p O ), D OER,DYN , is then calculated. Results in fig. 3B reveal that similar to ROD, any noticeable effect (deviation from initial slope D ·OER( p O ), representative of constant OER) arises at high doses (generally well over 100 Gy, see appendix B). Moreover, the present calculations yield a smaller ratio D OER,DYN / D OER,CONV (at a given absorbed dose) for the low oxygenations typically found in hypoxic tumors, but unrealistic for the surrounding healthy tissue. This translates into a narrowing of the therapeutic window due to selective tumor sparing. To investigate this closer, fig. 4 depicts the relative D OER of FLASH with respect to conventional irradiation as a measure for the FLASH sparing effect. Target oxygenation is here used as a proxy for normal/tumor tissue nature. For typical NT oxygenations (5–7% equiv. to 65–91 µM), the D OER ratio remains close to one (no sparing) for elevated hypothetical doses up to ~150 Gy and displays a slow decrease afterwards. Intermediate oxygenations (2-3% equiv. to 26–39 µM) show a rather flat behavior of D OER,DYN / D OER,CONV for low doses and a ratio of 0.98–0.93 at 100 Gy. Initial oxygenations of 1% (13 µM) and lower lead to steeper, nearly linear slopes starting from 0 Gy. The maximum sparing due to radiation-induced transient hypoxia for a 25 Gy dose is D OER,DYN / D OER,CONV = 0.94, achieved with oxygenations of 0.3–0.5% (3.9–6.5 µM); for even lower oxygen concentration, the corresponding curves become shallower again (less sparing). This is due to a tradeoff effect between oxygen consumption and oxygen availability for further depletion. Figure 4.
Relative OER-weighted dose (ratio) of FLASH vs. conventional dose rate irradiation for different initial oxygenations as a function of dose (1 µs time end point in water). Solid lines indicate the typical range of normal tissue oxygenations, while dashed lines represent different degrees of hypoxia found in tumors.
The relative effect plotted in fig. 4 easily identifies the tissue oxygenations leading to a better FLASH sparing for a given dose. Specifically, in order to afford a better sparing in the more oxygenated volume (NT), it is necessary that the corresponding D OER,DYN / D OER,CONV curves cross each other. In the dose range presented, this occurs only for a few combinations as follows. 2% and 0.1% and 1% and 0.3% curves intersect around 150 Gy and the 1% and 0.1% curves at ~40 Gy. None of these oxygenation levels are reasonable estimates for healthy tissue [ ], and as a consequence no realistic combination can be identified where the FLASH sparing due to radiation-induced transient hypoxia is expected to favor NT. Finally, the conditions of the recent in vitro experiment at ultra-high dose rates of Adrian et al. [18] are reproduced. They reported prostate cancer cell survival for different starting oxygenations in FLASH vs. conventional electron beam irradiation, see fig. 5, which we complement with dynamic O concentration for the different p O . For all of them, the resulting survival S DYN (fig. 5) is slightly higher than the experimental CONV curve but falls short of the experimental FLASH results (solid curves) which indicate a much more pronounced sparing effect. The numerical ratios S DYN / S CONV for 18 Gy as compared to the experimental ratio S FLASH / S CONV [18] are listed in table 1. For cell survival of prostate cancer cells as an endpoint and a dose of 18 Gy, oxygen depletion is responsible for an increase in survival of around 10% (a maximum of 22% in the most favorable combination) but cannot justify the full 30-490% difference observed experimentally [18]. Table 1.
Calculated ratio S ( D ) DYN / S ( D ) CONV obtained at 18 Gy. The stated values are derived using 1 µs yields in water, with the range of possible ratios for the other combinations from fig. 2 given in brackets. p O S DYN / S CONV S FLASH / S CONV exp. [18]
Figure 5.
Impact of oxygen consumption under FLASH conditions on the survival curves of prostate cancer cells. Lines “Exp. fit CONV” and “Exp. fit FLASH” show the surviving fraction
S(D) of prostate cancer cells at different oxygenation levels for conventional (red) and FLASH (blue) irradiations fitted by Adrian et al. [18] to their in vitro data. The experimental data points at 18 Gy are reproduced from [18]. For comparison, the shaded areas (“dynamic p O ”) represent the model prediction of cell survival S DYN, accounting for the oxygen consumption results. Note that in (A), the conventional fit curve and calculated survival S DYN overlap partially. As in the figs. 2 and 3 of the main article, the shaded areas contain the range of results for the different radiolytic yield time points and the underlying target medium.
Discussion With this novel, radiation chemistry-based modelling approach we focus on a situation equivalent to exposing a sealed target volume at conventional dose rate (to impede rediffusion); a reasonable approximation for ultra-high dose rates which serves as an upper limit condition for ROD. Any chemical difference experimentally established between these two conditions would give valuable hints towards the true mechanism at work. Particularly, a possible concentration-dependent homogeneous radical effect (taking place after track dissolution and possibly linked to oxygen, e. g. as described in [ ]) which affects the biological response could be identified in this way. Chemical inter-track effects (in the heterogeneous stage), in contrast, are not likely to take place. Most prominently, they affect OH· radicals and enhance recombination into H O [ , , ]) within a time window of about 1 ns (the half-life of OH· in biological environment [ ]) and a dose in that time window well over 10 Gy [ ]. For typical parameters used in recent FLASH experiments this can be ruled out, see also fig. 1C-1D which is representative of a single pulse given instantaneously . The situation is, however, very different for historical results showing an enhanced radioresistance in vitro [ , ] where dose rates as high as 10 Gy/s were used and irradiations were accomplished in 3 ns. While the present ab initio approach links heterogeneous stage effects to the relevant macroscopic, experimental results [ , ], some limitations should be kept in mind and reduce the applicability of these simulations in water to living cells or tissues. First, water is a convenient target medium for simulations, but a biological target is different in many aspects, such as the existence of cellular targets sensitive to chemical damage, reactions of radiolytic species with target molecules competing or adding up with those in water, the buffered pH condition (different in tumor / NT), the presence of catalytic enzymes to remove cytotoxic species, etc. As pointed out by Wardman [ ], even primary radical interactions (notably of e aq ⁻ and H·) are likely to change in presence of high concentrations of competing biomolecular scavengers [ ]. This would cause the dissolved oxygen to react in the next stage with the organic radicals R· formed and produce ROO· partially in peroxidation chain reactions [ , ]. A meaningful quantification of the latter organic peroxyl radical yield (and associated oxygen consumption), and of possible subsequent radical-radical reactions, needs however to be performed with a simulation based on homogeneous species concentrations. This goes beyond the focus of the present study but could be integrated into our code in future developments. In a biochemically more realistic model, it would also be desirable to extend the radiation chemical simulations until longer time scales, where the oxygen effect becomes effective biologically (a few ms [ ]). In water, a longer end time would entail differences mainly for the tumor oxygenation range. In order to serve as a reference for future modeling work, a larger body of experimental radiation chemical results in complex target media is needed. Finally, OER here used to translate an oxygenation change to a modification in effective dose, is a concept adopted from extensive in vitro evidence focusing mostly on clonogenic cell survival, but might not be equally suited for predicting the late effects in healthy tissue seen in vivo . Global and possibly indirect effects manifesting at longer time scales post-irradiation (see e.g. [51]) may need to complement the mechanistic explanations of the classical oxygen effect which center on a chemical enhancement of localized DNA lesions involving damage fixation or translocation within a defined ms time window directly by reaction with oxygen. Our findings agree with some inconsistencies of the oxygen depletion hypothesis noted earlier [ , ] and warnings about possible larger effects in hypoxic tumors than in physioxic tissues [ ]. In particular, Pratx and Kapp [ ] applied a radiolytic oxygen depletion model with constant depletion rate, but accounting explicitly for diffusion, blood vessel perfusion and metabolic oxygen consumption, to conclude that oxygen depletion yields an OER modification only for hypoxic starting conditions. A recent article by Petersson et al. [ ] presents a simplified mathematical framework to describe oxygen variation including reoxygenation during irradiation which can be adjusted to reproduce some experimental findings under FLASH conditions [ ]. The authors acknowledge, however, that many of the key response parameters remain uncertain; this circumstance compromises that model’s capability to predict the magnitude of the effect in experimental conditions. Finally, oxygen consumption would not show any dependence on intra-pulse dose rate if the total irradiation time is sufficiently short, however a delicate dependence on pulse dose, length and frequency has been established pre-clinically [ ]. Conclusions
Based on the chemical track evolution of electrons in oxygenated water, and allowing for a margin of conditions around the experimentally confirmed [ ] 0.33 µM/Gy depletion in buffered water, we have shown consistently that oxygen removal does not account for a sparing effect at FLASH dose rates and clinically realistic doses. A full depletion can be confidently ruled out, and no substantial normal tissue sparing effect emerges when applying our computational model (cell survival and OER-weighted dose) to typical experiment conditions. Indeed, due to their typical oxygenation, tumors would experience a more pronounced “FLASH effect” than NT. The present results, thus, lend support to the growing awareness that further and more complex mechanisms should be investigated beyond the simple ROD hypothesis.
Acknowledgments
The authors thank K. Petersson for providing numerical cell survival data from [ ]. References D. L. Dewey, J. W. Boag, Modification of the oxygen effect when bacteria are given large pulses of radiation, Nature 183 (1959) 1450-1451 2.
E. R. Epp, H. Weiss, B. Djordjevic, A. Santomasso, The radiosensitivity of cultured mammalian cells exposed to single high intensity pulses of electrons in various concentrations of oxygen, Radiat. Res. 52 (1972) 324-332 3.
H. Weiss, E. Epp, J. Heslin, C. C. Ling, A. Santomasso, Oxygen depletion in cells irradiated at ultra-high dose-rates and at conventional doserates, Int. J. Radiat. Biol. Rel. Stud. 26 (1974) 17-29 4.
H. B. Michaels, E. R. Epp, C. C. Ling, E. C. Peterson, Oxygen sensitization of CHO cells at ultrahigh dose rates: Prelude to oxygen diffusion studies, Radiat. Res. 76 (1978) 510-521 5.
S. Hornsey, D. K. Bewley, Hypoxia in mouse intestine induced by electron irradiation at high dose-rates, Int. J. Radiat. Biol. Relat. Stud. Phys. Chem. Med 19 (1971) 479-483 6.
S. B. Field, D. K. Bewley, Effects of dose-rate on the radiation response of rat skin, Int. J. Radiat. Biol. Relat. Stud. Phys. Chem. Med 26 (1974) 259-267 7.
J. H. Hendry, J. V. Moore, B. W. Hodgson, J. P. Keene, The Constant Low Oxygen Concentration in All the Target Cells for Mouse Tail Radionecrosis, Radiat. Res. 92 (1982) 172-181 8.
L. Gray, A. Conger, M. Ebert, S. Hornsey, O. Scott, The concentration of oxygen dissolved in tissues at the time of irradiation as a factor in radiotherapy, Br. J. Radiol. 26 (1953) 638-48 9.
T. Alper and P. Howard-Flanders, Role of oxygen in modifying the radiosensitivity of E. coli B, Nature 178 (1956) 978-979 10.
T. Alper, Variability in the oxygen effect observed with microorganisms part II: Escherichia coli B, Int. J. Radiat. Biol. 3 (1961) 369-377 11.
P. Howard-Flanders and T. Alper, The sensitivity of microorganisms to irradiation under controlled gas conditions, Radiat. Res. 7 (1957) 518-540 12.
P. Howard-Flanders, D. Moore, The time interval after pulsed irradiation within which injury to bacteria can be modified by dissolved oxygen: I. A search for an effect of oxygen 0.02 second after pulsed irradiation, Radiat. Res. 9 (1958) 422-437 13.
B. Michael, G. Adams, H. Hewitt, W. Jones, M. Watts, A posteffect of oxygen in irradiated bacteria: a submillisecond fast mixing study, Radiat. Res. 54 (1973) 239-251 14.
J.-M. V. Favaudon, L. Caplier, V. Monceau, F. Pouzoulet, M. Sayarath, C. Fouillade, M.-F. Poupon, I. Brito, P. Hupé, J. Bourhis, J. Hall, J.-J. Fontaine, M.-C. Vozenin, Ultrahigh dose-rate FLASH irradiation increases the differential response between normal and tumor tissue in mice, Sci. Translat. Med. 6 (2014) 15.
M.-C. Vozenin, P. De Fornel, K. Petersson, V. Favaudon, M. Jaccard, J.-F. Germond, B. Petit, M. Burki, G. Ferrand, D. Patin, H. Bouchaab, M. Ozsahin, F. Bochud, C. Bailat, P. Devauchelle, J. Bourhis, The advantage of FLASH confirmed in mini-pig and cat-cancer patients, Clin. Canc. Res. 25 (2019) 35-42 16.
P. Montay-Gruel, M. M. Acharya, K. Petersson, L. Alikhani, C. Yakkala, B. D. Allen, J. Ollivier, B. Petit, P. G. Jorge, A. R. Syage, T. A. Nguyen, A. A. D. Baddour, C. Lu, P. Singh, R. Moeckli, F. Bochud, J.-F. Germond, P. Froidevaux, C. Bailat, J. Bourhis, M.-C. Vozenin, C. L. Limoli, Long-term neurocognitive benefits of FLASH radiotherapy driven by reduced reactive oxygen species, Proc. Natl. Acad. Sci. USA 116 (2019) 10943-10951 J. Bourhis, W. J. Sozzi, P. G. Jorge, O. Gaide, C. Bailat, F. Duclos, D. Patin, M. Ozsahin, F. Bochud, J.-F. Germond, R. Moeckli, M.-C. Vozenin, Treatment of a first patient with FLASH-radiotherapy, Radiother. Oncol. 139 (2019) 18-22 18.
G. Adrian, E. Konradsson, M. Lempart, S. Bäck, C. Ceberg, K. Petersson, The FLASH effect depends on oxygen concentration, Brit. J. Radiol. 93 (2020) 20190702 19.
M. Durante, E. Bräuer-Krisch, M. Hill, Faster and safer? FLASH ultra-high dose rate in radiotherapy, Brit. J. Radiol. 91 (2018) 20170628 20.
M.-C. Vozenin, J. H. Hendry, C. Limoli, Biological benefits of ultrahigh dose rate FLASH radiotherapy: sleeping beauty awoken, Clin. Oncol. 31 (2019) 407-415 21.
D. R. Spitz, G. R. Buettner, M. S. Petronek, J. J. St-Aubin, R. T. Flynn, T. J. Waldron, C. L. Limoli, An integrated physico-chemical approach for explaining the differential impact of FLASH versus conventional dose rate irradiation on cancer and normal tissue responses, Radiother. Oncol. 139 (2020) 23-27 22.
G. Pratx, D. S. Kapp, A computational model of radiolytic oxygen depletion during FLASH irradiation and its effect on the oxygen enhancement ratio, Phys. Med. Biol. 64 (2019) 185005 23.
C. Koch, Re: Differential impact of FLASH versus conventional dose rate irradiation, Radiother. Oncol. 139 (2020) 62-63 24.
J. Hendry, Taking Care with FLASH Radiation Therapy, Int. J. Radiat. Oncol. Biol. Phys. 107 (2020) P239-242 25.
K. Petersson, G. Adrian, K. Butterworth, S. J. McMahon, A quantitative analysis of the role of oxygen tension in FLASH radiotherapy, Int. J. Radiat. Oncol. Biol. Phys. 107 (2020) 539-547 26.
J.-Y. Jin, A. Gu, W. Wang, N. L. Oleinick, M. Machtay, F.-M. (Spring) Kong, Ultra-High Dose Rate Effect on Circulating Immune Cells: A Potential Mechanism for FLASH Effect?, Radiother. Oncol. 149 (2020) 55-62 27.
R. Labarbe L. Hotoiu, J. Barbier, V. Favaudon, A physicochemical model of reaction kinetics supports peroxyl radical recombination as the main determinant of the FLASH effect, Radiother. Oncol. 153 (2020) 303-310 28.
H. Liew, S. Mein, I. Dokic, J. Debus, A. Abdollahi, A. Mairani, Deciphering time-dependent DNA damage complexity, repair and oxygen tension: a mechanistic model for FLASH-dose-rate radiotherapy, Int. J. Radiat. Oncol. Biol. Phys. (online before print) 29.
D. R. Spitz, G. R. Buettner, C. L. Limoli, Response to letter regarding: “An integrated physico-chemical approach for explaining the differential impact of FLASH versus conventional dose rate irradiation on cancer and normal tissue responses”, Radiother. Oncol. 139 (2020) 64-65 30.
C. C. Ling, G. Josef, To the editors, Radiother. Oncol. 147 (2020) P240 31.
D. R. Spitz, G. R. Buettner, C. L. Limoli, Response to Ling et al. regarding: “An integrated physico-chemical approach for explaining the differential impact of FLASH versus conventional dose rate irradiation on cancer and normal tissue responses”, Radiother. Oncol. 147 (2020) P241-242 32.
G. Pratx, D. S. Kapp, Ultra-high-dose-rate ash irradiation may spare hypoxic stem cell niches in normal tissues, Int. J. Radiat. Oncol. Biol. Phys. 105 (2019) 190-192. 33.
M. Krämer, G. Kraft, Calculations of heavy-ion track structure, Radiat. Environ. Biophys. 527 (1994) 91–109 34.
D. Boscolo, M. Krämer, M. Durante, M. Fuss, E. Scifoni, TRAX-CHEM: A pre-chemical and chemical stage extension of the particle track structure code TRAX in water targets, Chem. Phys. Lett. 698 (2018) 11-18 D. Boscolo, M. Krämer, M. C. Fuss, M. Durante, E. Scifoni, Impact of target oxygenation on the chemical track evolution of ion and electron radiation, Int. J. Mol. Sci. 21 (2020) 424 36.
D. R. Grimes, M. Partridge, A mechanistic investigation of the oxygen fixation hypothesis and oxygen enhancement ratio, Biomed. Phys. Eng. Express 1 (2015) 045209 37.
D. W. Whillans, A. M. Rauth, An experimental and analytical study of oxygen depletion in stirred cell suspensions, Radiat. Res. 84 (1980) 97-114 38.
N. T. S. Evans, Removal of dissolved oxygen from aqueous media by ionizing radiations, Radiat. Effects 1 (1969) 19-22 39.
H. B. Michaels, Oxygen depletion in irradiated aqueous solutions containing electron affinic hypoxic cell radiosensitizers, Int. J. Rad. Oncol. Biol. Phys. 12 (1986) 1055-1058 40.
B. Palcic, L. Skarsgard, Reduced oxygen enhancement ratio at low doses of ionizing radiation, Radiat. Res. 100 (1984) 328-339 41.
S. McKeown, Defining normoxia, physoxia and hypoxia in tumours - implications for treatment response, Brit. J. Radiol. 87 (2014) 20130676
P. Montay-Gruel, M. M. Acharya, P. Gonçalves Jorge, B. Petit, I. G. Petridis, P. Fuchs, R. Leavitt, K. Petersson, M. Gondre, J. Ollivier, R. Moeckli, F. Bochud, C. Bailat, J. Bourhis, J.-F. Germond, C. L. Limoli and M.-C. Vozenin, Hypo-fractionated FLASH-RT as an effective treatment against glioblastoma that reduces neurocognitive side effects in mice, Clin. Canc. Res. (online before print)
W. Tinganelli, M. Durante, R. Hirayama, M. Krämer, A. Maier, W. Kraft-Weyrather, Y. Furusawa, T. Friedrich, E. Scifoni, Killpainting of hypoxic tumours in charged particle therapy, Scient. Rep. 5 (2015) 1-13 44.
P. Wardman, Time as a variable in radiation biology: The oxygen effect, Radiat. Res. 185 (2016) 1-3 45.
R. Watanabe, K. Saito, Monte Carlo simulation of water radiolysis in oxygenated condition for monoenergetic electrons from 100 eV to 1MeV, Rad. Phys. Chem. 62 (2001) 217-228 46.
M. S. Kreipl, W. Friedland, H. G. Paretzke, Interaction of ion tracks in spatial and temporal proximity, Radiat. Environ. Biophys. 48 (2009) 349-359 47.
J. Ramos-Méndez, N. Domínguez-Kondo, J. Schuemann, A. McNamara, E. Moreno-Barbosa, B. Faddegon, LET-dependent intertrack yields in proton irradiation at ultra-high doce rates relevant for FLASH therapy, Radiat. Res. 194 (2020) 351-362 48.
H. Sies, Strategies of antioxidant defense, Eur. J. Biochem. 215 (1993) 213-219 49.
P. Wardman, Radiotherapy using high-intensity pulsed radiation beams (FLASH): A radiation-chemical perspective, Radiat. Res. 194 (2020) 607-617 50.
H. B. Michaels, J. W. Hunt, A Model for Radiation Damage in Cells by Direct Effect and by Indirect Effect: A Radiation Chemistry Approach, Radiat. Res. 74 (1978) 23-34 51.
C. Liu, Q. Lin, Z. Yun, Cellular and Molecular Mechanisms Underlying Oxygen-Dependent Radiosensitivity, Radiat. Res. 183 (2015) 487-496 52.
P. Montay-Gruel, K. Petersson, M. Jaccard, G. Boivin, J.-F. Germond, B. Petit, R. Doenlen, V. Favaudon, F. Bochud, C. Bailat, J. Bourhis, M.-C. Vozenin, Irradiation in a FLASH: Unique sparing of memory in mice after whole brain irradiation with dose rates above 100 Gy/s, Radiother. Oncol. 124 (2017) 365-369 53.
N. W. Colangelo, E. I. Azzam, The Importance and Clinical Implications of FLASH Ultra-High Dose-Rate Studies for Proton and Heavy Ion Radiotherapy, Radiat. Res. 193 (2020) 1-4 A. Colliaux, B. Gervais, C. Rodriguez-Lafrasse, M. Beuve, Simulation of ion-induced water radiolysis in different conditions of oxygenation, Nucl. Instrum. Meth. Phys. Res. B 365 (2015) 596-605
Appendix
A. Methods details
Monte Carlo simulation time points
It is generally observed and accepted that the chemical track structure of a single incident particle of radiation reaches the spatial dissolution into a homogeneous distribution at ~1µs. For a pure (anoxic) water target, this time end point is also where chemical equilibrium is attained, radiolytic yields (G-values) are stable, and TRAX-CHEM results have been benchmarked against the available experimental data. In presence of oxygen (a target molecule in homogeneous dissolution, according to the given concentration), further reactions with the radiolytic species can still take place and produce changes in species yields even later than 1 µs which may affect the present results. The authors therefore include within the present results two more time points which we consider delimit the meaningful range for biological applications: on the one hand, 0.5 µs is the earliest moment where the radiolytic yields generally adopt their final value in anoxic conditions , thus representing a stable situation after radiolysis where track chemistry has lost its prevalence. On the other hand, the selection of 2µs is an extended time point guaranteeing an even better species homogeneity, especially when adding O , but limits the possible reaction partners to H O and O at a stage when other competing reactions with additional solutes of a realistic biological environment (e.g. a cell) become very important. For simulation over longer times, an inclusion of specific target molecules, especially those likely to suffer chemical damage from the radiolytic radicals as well as radical scavengers / catalyzers capable of counteracting, would be mandatory and give additional interesting insights into the biochemical radiation effects. This functionality is planned but not currently included in TRAX-CHEM or any other tool available to the authors. In all results figures, the shaded areas depict the range of variation of the presented results when using different time points (0.5µs to 2µs) of the underlying radiolytic yields or changing the target medium. Acid-base equilibrium
In the present track structure approach, O ·⁻ is almost exclusively produced by aqueous electrons and HO · by H·, so that the predicted ratio of both therefore closely follows the yield of e aq ⁻ vs. H·. In a biological environment, the relative abundances of this acid-base pair would rapidly equilibrate depending on pH. In order to avoid confusion, we limit the presentation of results and discussion to the sum of both. Concentration units
With the objective to use absolute SI units for oxygenation which are independent of solubility, we opt for µM for oxygen concentration [O ] in graphs and numerical results. However, for a better comparability with experimental data, the oxygenation levels of particular target examples is sometimes also described in partial oxygen pressure (% p O ), and the definition of oxygen enhancement ratio is based on p O following usual practice. The corresponding relative oxygen saturation in experimental targets was generally calculated according to Henry’s law for a temperature of 25°C where 21% p O = 273 µM. Only for the reconstruction of in vivo conditions, Henry’s constant was set to its value at 37°C, H cp =1.0264·10 -5 mol/(m Pa), so that 21% p O corresponds to 218 µM. The chemical diffusion-reaction model itself remained unchanged from Boscolo et al. [35]. OER convention
Note that the OER definition followed here is different from the ratio
D(p O )/ D normox which has been adopted in some studies and rather reflects an “oxygen diminution ratio” (ODR); OER=1/ODR · ( D anoxia / D normox ). B. Species concentrations and their dose dependence
ROS production: superoxide and its protonated form
Since in oxygenated targets a larger impact of oxidative stress is expected, the production of superoxide O ·⁻ and its conjugate acid perhydroxyl HO ·, which are closely related to radiolytic oxygen depletion in water, are obtained as an example of changes in ROS production. Oxygenation-dependent chemical yields per deposited dose for O ·⁻/ HO · production by 1 MeV electrons are presented in figure A1. In well oxygenated targets, the yields for oxygen consumption and production of O ·⁻/ HO do not depend on initial oxygenation. Below an oxygenation value of around 65 µM (5% pO at room temperature), the curve however exhibits a steep decrease as the sparse distribution of oxygen in the target slows down the chemical track dynamics sensibly and reactions from eq. (1, main article) become less frequent. Figure A1.
Calculated oxygenation-dependent radiolytic yield for the production of superoxide + perhydroxyl at different time points around the end of the track (heterogeneous) chemical stage (1 MeV electron radiation).
Radiolytic oxygen depletion for high doses
Even if irradiation with doses exceeding some tens of Gy is usually too slow to comply with the ms dose delivery we assume as a condition, and also clinically irrelevant, we want to present an extended dose dependence analysis until reaching full depletion. Please note that the results of this section represent an approximation and do not account for oxygen rediffusion in the irradiated target, and thus show an upper limit to oxygen consumption. In an extended dose range (fig. A2A), oxygen consumption for different initial oxygenations p O is found to be not proportional to dose any more (as often approximated). The successive deviation from linearity is especially pronounced for the low range of oxygenation typically found in tumors [ ]. Absorbed doses required to deplete virtually all dissolved oxygen range from ~400 Gy ( p O = 7%) to ~150 Gy ( p O = 1%), far exceeding pre-clinically accepted doses in a single fraction even at ultra-high dose rate [ ]. The dynamical oxygen depletion introduced here is fundamental to estimate the high doses which would be necessary. Simultaneously to progressive oxygen consumption, superoxide production is steady until a complete oxygen depletion is approached (fig. A2C). The resulting plateau is observed at 89.6 µM (7 % pO ), 64.0 µM (5 % pO ), 38.4 µM (3 % pO ), and 12.8 µM (1 % pO ), respectively. Figure A2.
Chemical species concentrations under FLASH irradiation conditions with 1 MeV electrons for different initial oxygenations and its implication on the radiobiological outcome for high doses. (A) Radiochemical oxygen consumption with dose, (B) production of superoxide + perhydroxyl, and (C) OER-weighted dose D OER,DYN for a hypothetical short pulse, all at room temperature. The initial target oxygenations corresponding to each line are given in % p O for an easier comparison to experimental data. Relative dose-dependent superoxide production, FLASH vs. conventional irradiation
The production of cytotoxic superoxide is the second end point influenced directly by oxygen pressure in the heterogeneous (track) chemistry stage of radiation damage. In addition to radiolytic oxygen consumption, the reduced superoxide production O ·⁻+HO · of FLASH with respect to conventional irradiation conditions can therefore serve as a proxy for FLASH tissue sparing. The relative effect is presented in figure A3. It shows, indeed, a generally larger sparing effect at the same dose than the effective dose D OER discussed in the main article, indicating higher sensitivity to transient hypoxia at ultra-high dose rate. However, the intrinsic nature of superoxide production dictates that the less oxygen is available, the less superoxide will be produced, and the relative curves never cross to predict a selective NT sparing – hypoxic tissues are still favored over physioxic ones. In water, the ROS produced can undergo spontaneous disproportionation giving hydrogen peroxide at longer time scales in the chemical evolution, so that a decreased production might be linked to a decrease in H O . For a quantitative analysis of that aspect, other competing reactions (e.g. with OH·) need to be included in simulations of time points extending into the homogeneous chemistry stage. Figure A3.
Relative superoxide (O ·⁻+HO ·) production of FLASH vs. conventional dose rate irradiation for different initial oxygenations as a function of dose (calculated for 1 µs time end point in water). Solid lines correspond to typical normal tissue oxygenations, while dashed lines represent different degrees of hypoxia found in tumors. C. Application to pre-clinical/clinical experiment conditions
Finally, to enable a quantitative comparison, we present numerical values for the specific experimental conditions of recent in vivo studies [14,15,16,17] where significant sparing was reported in FLASH mode. This represents a tentative application of the present model to experiments realized in biological organisms and needs to be interpreted with precaution, c. f. Discussion in main article. Summarizing these studies, Favaudon et al. [14] observed a reduced induction of pulmonary fibrosis, pulmonitis and cutaneous lesions in a syngeneic, orthotopic murine lung tumor model by FLASH irradiation as compared to conventional dose rate. The lung sparing was quantified by a dose modifying factor of 2 for the same average late effect score in the recent review of Vozenin et al. [20], while antitumor efficiency and survival was comparable in both modalities. In a dose escalation study comprising doses between 22-41 Gy, Vozenin et al. [15] showed significant healthy skin sparing (≥ 20% extra dose in FLASH for fibronecrosis end point) in cats and a minipig, in combination with good tumor control, afforded by the possibility of giving higher doses, in the cat patients. Montay-Gruel et al. [15] found a marked decrease (up to absence for some end points) in late neurocognitive effects after a 10 Gy dose when using FLASH vs. conventional dose rate in mice. Furthermore, a first human skin lymphoma was treated with FLASH radiotherapy [17] safely for the first time and yielded a very promising estimated effective FLASH dose to healthy skin of 2/3 of the absorbed dose.
Table A2 presents our model results reproducing the experimental conditions of several parts of the above mentioned experiments. This includes the murine lung tumor and healthy lung from [14] at the 17 Gy dose (data available for both conventional and FLASH with clearly diverging late effect scores [20]), minipig and cat skin sparing ≥ 20% from Vozenin et al. [15] with good tumor response as a secondary endpoint, mouse whole brain irradiations to 10 Gy which showed neurocognitive benefits at FLASH dose rates [15], and the first human patient case [17] treated with 15 Gy (cutaneous lymphoma and healthy skin). The tumor xenograft results (tumor control) also presented in [14] are not modeled since the oxygenation would be very difficult to estimate, highly inhomogeneous, and very imprecise. Since tissue-specific oxygenation is not stated in the original studies, we rely on a review by McKeown et al [41] on average oxygenation values for a range of tumors and the corresponding NT. Target oxygenations were set by selecting the closest available match for each case, from 3.4% to 5.9% in normal tissue and from 1.5% to 2.1% in tumor. Calculated production of the superoxide anion and its protonated form, and OER-weighted dose are given for conventional irradiation together with the corresponding % values (FLASH/ conventional) and final oxygen pressure p O for FLASH mode. Results for normal tissue and tumor are listed beneath each other. It is evident from these normalized values that no significant sparing effect is predicted for healthy tissue. D OER,DYN = 100% for all normal tissues but presents slight drops (maximum of 2.3%) for tumors, smaller and/or opposite to the desired differential effect [14,15,16,17]. Interestingly, superoxide + perhydroxyl production is more sensitive to the transient change in oxygenation and displays a maximum difference FLASH vs. CONV of 3.4% in normal tissue and up to 11.4% in tumor. Even if these magnitudes are more noticeable, they predict, again, a differential sparing of the tumors in disagreement with [14,15]. The present results therefore fail to explain the experimental findings [14,15,16,17] through oxygen depletion or a decrease in ROS production. Quantitative predictions based on oxygen depletion analysis and OER are therefore in sharp contrast to most of the experimental findings at FLASH dose rates so far [14,15,16,17,18] where a differential effect in favor of NT sparing has been shown.
Table A2.
Model results for superoxide production, final oxygenation and OER-weighted dose in conditions employed in different pre-clinical and one clinical experiment. For the FLASH results (right block of columns), the range given in brackets indicates the confidence interval of values when applying the different sets of chemical yields from figs. 2 and A1 (shaded areas).
Conventional dose rate FLASH Experiment Dose (Gy) a p O (%) b [O ·⁻+HO ·] (µM) D OER,CONV (Gy) [O ·⁻+HO ·] (% of conv.) p O (%) D OER,DYN (% of conv.) Mouse whole brain [16] 10 3.4 2.78 26.3 98.2 (97.4-99.3) 3.13 (3.10-3.20) 100 Minipig skin [15] 31 5.3 9.48 81.5 98.1 (96.6-99.4) 4.39 (4.09-4.54) 100 Cat, healthy skin/ mucosa [15] 33 5.9 10.3 86.8 98.1 (96.5-99.4) 4.91 (4.59-5.05) 100 Cat squameous cell carcinoma [15] 33 1.9 7.22 85.5 88.6 (88.6-90.0) 1.27 (1.10-1.48) 98.5 (97.7-99.1) Mouse lung [14] 17 5.6 5.24 44.7 99.2 (98.5-99.7) 5.09 (4.92-5.17) 100 Lung tumor [14] 17 2.1 3.84 44.3 96.1 (96.1-97.1) 1.74 (1.63-1.85) 99.3 (99.3-99.5) Human patient, healthy skin [17] 15 5.3 4.59 39.4 98.9 (98.2-99.7) 4.86 (4.71-4.93) 100 Human skin lymphoma [17] 15 1.5 2.77 38.2 96.8 (96.1-98.0) 1.24 (1.16-1.33) 99.1 (98.6-99.3) a For dose escalation studies, a representative intermediate dose was chosen. b Initial oxygenation is estimated based on the values given for certain tumors and normal tissues in [ ] D. Use of protons or ions for FLASH radiotherapy
Regarding the possibility to use protons or ions for FLASH radiotherapy [53], considering the present results together with [O ]-dependent yields for ions at different LET from [35,54] allows to conclude that high energy protons (corresponding to the beam in the entrance channel) present oxygen consumption characteristics numerically very close to the electrons discussed here. Low energy protons with higher LET (close to stopping in the target volume), however, deplete less oxygen per dose so that the differences found in superoxide production and OER would be shifted to higher doses than for electrons, or low LET radiation in general. Likewise, heavier ions such as C exhibit even smaller yields G –O , G O ·⁻ , and G HO · due to prominent early intra-track species recombination. Consequently, compared to the electrons investigated here, considerably higher doses would be needed for complete oxygen depletion or a significant impact on superoxide production.· due to prominent early intra-track species recombination. Consequently, compared to the electrons investigated here, considerably higher doses would be needed for complete oxygen depletion or a significant impact on superoxide production.