Radiation-Chemical Oxygen Depletion Depends on Chemical Environment and Dose Rate: Implications for the FLASH Effect

Abstract

Purpose:

FLASH (dose rates >40 Gy/s) radiation therapy protects normal tissues from radiation damage, compared with conventional radiation therapy (~Gy/m). Radiation-chemical oxygen depletion (ROD) occurs when oxygen reacts with radiation-induced free radicals, so a possible mechanism for FLASH involves radioprotection by the decreased oxygen as ROD occurs. High ROD rates would favor this mechanism, but prior studies have reported low ROD values (~0.35 μM/Gy) in chemical environments such as water and protein/nutrient solutions. We proposed that intracellular ROD might be much larger, possibly promoted by its strongly reducing chemical environment.

Methods and Materials:

ROD was measured, using precision polarographic sensors, from ~100 μM to zero in solutions containing intracellular reducing agents ± glycerol (1M), to simulate intracellular reducing and hydroxyl-radical-scavenging capacity. Cs irradiators and a research proton beamline allowed dose rates from 0.0085 to 100 Gy/s.

Results:

Reducing agents significantly altered ROD values. Most greatly increased ROD but some (eg, ascorbate) actually decreased ROD and additionally imposed an oxygen dependence of ROD at low oxygen concentrations. The highest values of ROD were found at low dose rates, but these montonically decreased with increasing dose rate.

Conclusions:

ROD was greatly augmented by some intracellular reducing agents but others (eg, ascorbate) effectively reversed this effect. Ascorbate had its greatest effect at low oxygen concentrations. ROD decreased with increasing dose rate in most cases.

Introduction

Attempts to differentially modify the response of tumor versus normal tissue to cancer therapy have been underway for many decades. In the case of radiation therapy (RT), dose fractionation has been used to selectively spare normal tissue.¹ Additionally, conformal methods to increase the relative dose to tumor have continuously evolved.² Similarly, the relatively recent use of proton therapy has inherent dose-deposition advantages compared with photons.³ In 2014, a study using a murine model showed that high-dose-rate (electron) RT (>40 Gy/s average), known as FLASH-RT (FLASH), provided a normal tissue-sparing effect while showing no change in tumor response⁴; FLASH requires both aspects. Radioprotection by FLASH has now been demonstrated in multiple models⁵ for various types of normal tissue, and has been extended to proton RT.⁶

Radiation-chemical oxygen depletion (ROD) has been the most-studied mechanism to explain FLASH.⁴ On the one hand, free radicals produced by ionizing radiation combine with oxygen to remove it from solution.⁷ On the other hand, ionizing-radiation effects are diminished as oxygen concentration decreases—the “oxygen effect.”⁸ Thus, ionizing radiation can cause radiation resistance by depletion of oxygen during the dose delivery.⁹ The chances of this occurring are increased with high ROD rates and Spitz and colleagues suggested mechanisms for this, including FLASH itself.¹⁰ However, most reported ROD values are only ~0.35 μM/Gy^11–16 (Fig. E1). The only studies reporting higher ROD values have included the reducing agent glutathione.^17,18

The goal of this study was to undertake a pragmatic survey of chemical environments (emphasizing cellular reducing agents and primary free radical scavengers) that might modify ROD values or that could support tumor-specific changes in the kinetics for ROD that would diminish its effectiveness relative to normal tissue. We used precision polarographic oxygen sensors^19–21 to monitor ROD in solutions containing chemical mixtures with the full complement of free radicals produced by the radiation sources over a 10,000-fold dose-rate range. These conditions approach relevance to cellular radiobiology and are considered critical for understanding and modeling ROD as a possible mechanism for the FLASH effect. The rather complex variety of results warrant much further research to elucidate the radiation-chemistry mechanisms.

Materials and Methods

Radiation sources and dose rates

The radiations sources consisted of 2 Cs irradiators (Mark I, JL Shepherd & Associates, San Fernando, CA; 0.0085 and 0.1 Gy/s) and a collimated (20 × 20 mm², double-scatter) proton beam, 0.4 to 0.5 and 90 to 120 Gy/s (IBA Proteus Plus Louvain-La-Neuve, Belgium); for dosimetry and other method details see E Supplement. All conditions were investigated at least twice, with example results displayed in the figures. Replicate results were within 5% of each other.

Polarographic oxygen sensors and sample vials

Precision oxygen sensors (Table E1) were sealed, via matching standard tapers, into stirred Pyrex vials containing pre-equilibrated solutions with temperature controlled at 25°C as described previously²¹ (Fig. E2). Each sensor produced an oxygen-partial-pressure-dependent current monitored throughout each experiment by an electrometer and recorded by an Arduino Uno microprocessor (1 data point per sensor per minute), the data of which were subsequently transferred to an Excel spreadsheet and analyzed as described in Fig. E3.

Reagents

All reagents were obtained from Sigma-Aldrich; they were grouped generally into primary free-radical scavengers and reducing agents (Table E2). They were purchased at best available purity and were not further purified. Water was Sigma purified DI, #38796. The most often-studied solutions were buffered with phosphate (NaH₂PO₄ 5 mM, pH 7.1 ± 0.2) or HEPES (2-[4-(2-hydroxyethyl)piperazine-1-yl]ethanesulfonic acid, 10 mM, pH 7.1 ± 0.2). The base of most mixtures has been previously referred to as CELL, consisting of HEPES buffer with glycerol (1000 mM), glucose (5 mM) and glutathione (5 mM). Clearly, CELL does not closely approximate the complexity of a cell, but we showed that phage DNA demonstrated intracellular radiation sensitivity and oxygen effect in similar solutions.^22,23

Because our studies typically involved the use of reducing agents and their combination, it was very important to minimize possible metal contamination, thus achieving low levels of autoxidation (see Supplement).

Results

For phosphate buffer containing primary radical scavengers at high concentration, all primary radicals (OH^•, H^•, and ${{\mathrm{e}}^{-}}_{\text{aq}}$) were expected to react with the scavenger. ROD should occur via oxidation of the resulting scavenger radicals. Under these conditions glycerol (1000 mM) or lysozyme (5%) had similar ROD values (0.3–0.4 μM/Gy, 0.1 Gy/s Cs irradiator), independent of oxygen concentration (zero-order). There was little decrease in ROD for either scavenger at 10-fold lower concentration (Fig. 1, Table 1).

Fig. 1.

Examples of radiation-chemical oxygen depletion (Cs source; 0.1 Gy/s) in phosphate buffer with primary radical scavengers glycerol (1000 mM, closed squares, 0.33 μM/Gy; 100 mM, open squares, 0.28 μM/Gy) and lysozyme (5%, closed hexagons, 0.40 μM/Gy; 0.5%, open hexagons, 0.34 μM/Gy).

Table 1.

Summary of ROD values

Figure	Buffer	Substrate	DR (Gy/s)	ROD (μM/Gy)	Order
1	Phosphate	Glycerol (0.1M and 1M)	0.1	0.28 and 0.33	0
1	Phosphate	Lysozyme (0.5% and 5%)	0.1	0.34 and 0.40	0
2	Phosphate	NADPH (1 mM)	0.1	1.05	0
2	Phosphate	sLin (1 mM)	0.1	0.95	0
2	Phosphate	Cysteamine (1 mM)	0.1	0.70	0
2	Phosphate	GSH (1 mM)	0.1	0.55	0
2	Phosphate	Ascorbate (1 mM)	0.1	0.25	0–1
2	Phosphate	UA (1 mM)	0.1	0.10	0–1
2	Phosphate	-	0.1	0.21	0–1
3	Phosphate	NADH (1 mM)	0.1	0.90	0
3	Phosphate	NADH (10 mM)	0.1	3.1	0
3	Phosphate	Cysteamine (1 mM) 10	0.1	0.55	0–1
3	Phosphate	Cysteine (10 mM)	0.1	2.0	0
3	Phosphate	sLin (1 mM)	0.1	0.95	0
3	Phosphate	sLin (3 mM)	0.1	1.60	0
3	Phosphate	GSH (1 mM)	0.1	0.55	0
3	Phosphate	GSH (5 mM)	0.1	0.65	0
3	Phosphate	GSH (25 mM)	0.1	0.73	0
4	HEPES	CELL	0.1	0.75	0
4	HEPES	CELL + NADPH (1 mM)	0.1	1.06	0
4	HEPES	CELL + Cysteamine (5 mM)	0.1	1.06	0
4	HEPES	CELL + additions	0.1	2.64	0
5	DI water	-	0.0085–100	0.25, 0.25, 0.25, 0.29	0
5	Phosphate	-	0.0085–100	0.20, 0.15, 0.19, 0.20	0–1
5	H2SO4 (0.8 N)	Fricke solution	0.0085–100	0.41,0.35, 0.35, 0.35	0
5	HEPES	CELL	0.0085–100	0.85, 0.70, 0.50, 0.40	0
5	Phosphate	NADH (10 mM)	0.0085–100	5.1, 3.1, 1.7, 0.91	0
5	Phosphate	Cysteine (10 mM)	0.0085–100	3.0, 2.0, 1.5,0.63	Mixed
5	HEPES	CELL + additions	0.0085–100	3.6, 2.0, 1.5,0.69	0

Included are the dose rate used and the order of the ROD reaction, with respect to oxygen concentration. For zero-order reactions (0), oxygen depletion is linear with dose and independent of oxygen concentration. Some substrates are characterized by zero-order at high oxygen concentrations and firstorder (dependent on oxygen concentration) at low oxygen concentrations (0–1). ROD values shown are for the higher oxygen levels in the case of 0 to 1 kinetics. For the bottom 7 rows of the table, 4 ROD values are shown, one for each dose rate in order of lowest to highest. CELL is the mixture of glycerol (1000 mM), HEPES (10 mM), glucose (5 mM), and glutathione (5 mM). Additions include cysteine (10 mM), dextrin-conjugated linoleic acid (1 mM), and NADPH (1 mM).

Abbreviations: DR = dose rate; GSH=glutathione; ROD = radiation-chemical oxygen depletion; sLin = dextrin-conjugated linoleic acid.

ROD (0.1 Gy/s Cs irradiator) was next measured in phosphate buffer (5 mM) with reducing agents at a standard concentration of 1 mM. Under these conditions, most primary radicals would react with the reducing agent and the resulting radicals may react secondarily with oxygen, additional reducing agents or each other. ROD varied from >1 μM/Gy for NADPH to <0.1 μM/Gy for uric acid (Fig. 2, Table 1). The curves with lowest ROD values (buffer alone, ascorbate and uric acid) also showed a substantial oxygen dependence at low oxygen levels, described as “zero-order to first-order,” whereas the rest demonstrated zero-order oxygen dependence throughout.

Fig. 2.

Examples of radiation-chemical oxygen depletion (ROD) (Cs source; 0.1 Gy/s) in phosphate buffer (solid line at extreme right) with additions (1 mM each) of NADPH (solid diamonds, ROD = 1.05 μM/Gy), dextrin-conjugated linoleic acid (open triangles, ROD = 0.95 μM/Gy), cysteamine (solid circles, ROD = 0.70 μM/Gy), glutathione (open hexagons, ROD = 0.55 μM/Gy), ascorbate (plus signs, ROD = 0.25 μM/Gy), and uric acid (dotted line crossing other curves, ROD = 0.10 μM/Gy). With one exception (uric acid), all curves were characterized by a sharp transition at the time of beam turn-on and remained stable at the time of beam turn-off. In the case of uric acid there was initially no change in sensor reading at the time of beam turn-on and a slight continuation of ROD at the time of beam turn-off. The substrates with high ROD values were characterized by curves that were nearly independent of oxygen concentration (zero-order) (NADPH, dextrin-conjugated linoleic acid, cysteamine, glutathione) but the rest showed substantial oxygen dependence (phosphate buffer, ascorbate, uric acid). Because the individual curves were derived from multiple independent experiments, the time axis (X) is arbitrary and for each curve shows an initial “beam off” time, then “beam on” to monitor ROD, and then another “beam off” time.

Increasing the reducing agent concentration greatly increased the ROD in several cases investigated. At 10 mM, the rates of ROD for NADH or cysteine were substantially greater than the total primary radical yield (~0.7 μM/Gy) suggesting chain or serial reactions (Fig. 3, Table 1). ROD for solubilized linoleic acid (dextrin-conjugated) was not measured >3 mM (solubility limit). Surprisingly, ROD for glutathione (1–25 mM) varied minimally (Fig. 3).

Fig. 3.

Dependence of radiation-chemical oxygen depletion (ROD) (Cs source; 0.1 Gy/s) on substrate concentration for several reducing agents, each dissolved in phosphate buffer. NADH (10 mM, solid diamonds, ROD = 3.1 μM/Gy; 1 mM, open diamonds, ROD = 0.9 μM/Gy), cysteine (10 mM, solid circles, ROD = 2.0 μM/Gy; 1 mM, open circles, ROD = 0.55 μM/Gy), dextrin-conjugated linoleic acid (3 mM, solid triangles, ROD = 1.5 μM/Gy; 1 mM, open triangles, ROD = 0.95 μM/Gy), and glutathione (25 mM, solid hexagons, ROD = 0.73 μM/Gy; 5 mM, open hexagons with central dot, ROD = 0.65 μM/Gy; 1 mM, open hexagons, ROD = 0.55 μM/Gy). In cases where there was some curvature, ROD values are reported for the upper part of the curve. This had the largest effect for cysteine (1 mM), which showed a substantial decrease in ROD <50 μM oxygen.

For many combinations investigated, we found that mixing reducing agents having low rates of ROD and substantial oxygen dependence of the ROD process (ascorbate, uric acid) often transferred these characteristics to the mixture (Fig. 4). For example, ROD for CELL was ~0.75 μM/Gy independent of oxygen concentration (Fig. 4A). The addition of ascorbate (1 mM) to CELL caused a major reduction in initial ROD (0.43 μM/Gy) and additionally imposed a significant oxygen dependence at low oxygen concentration (Fig. 4A).

Fig. 4.

Radiation-chemical oxygen depletion (ROD) in solutions containing CELL (glycerol, 1000 mM, glucose, 5 mM, glutathione, 5 mM, HEPES, 10 mM), plus various additions or substitutions (Cs source; 0.1 Gy/s). (A) CELL (open squares) ROD = 0.75 μM/Gy; CELL + ascorbate (1 mM, open squares with “x”), initial ROD = 0.43 μM/Gy. CELL plus additions (cysteine, 10 mM, dextrin-conjugated linoleic acid, 1 mM, NADPH, 1 mM) (open circles) with the further addition of 0.5 mM uric acid (open circles with central dot), ROD = 2.0 μM/Gy for both curves. Addition of ascorbate (circles, left half solid and circles, right half solid, respectively) decreased the initial ROD to 0.64 μM/Gy (both curves), again with no effect of uric acid (these curves are essentially identical so the symbols look like solid circles). (B) Three-step curves (40–90 minutes): first step, glycerol (1000 mM) in HEPES buffer plus NADPH (1 mM, open diamonds) or cysteamine (5 mM, open circles), ROD = 1.06 μM/Gy for both conditions; second step, addition of glucose (5 mM, no change in ROD); third step, addition of glutathione (5 mM, no change in ROD). Addition of ascorbate (open diamonds and circles with central + and x, respectively, for the second set of curves) decreased the initial ROD to 0.64 μM/Gy (both curves) with the effect of ascorbate at low oxygen concentrations much greater for cysteamine than NADPH.

The relative effect of uric acid versus ascorbate was tested directly in an even more complex mixture (Fig. 4A). In view of the very high ROD seen for 10 mM cysteine (Fig. 3) and the known very high cellular values for intracellular protein thiols, CELL was combined with cysteine (10 mM), dextrin-conjugated linoleic acid (1 mM) and NADPH (1 mM), ± uric acid (0.5 mM) (CELL plus additives). Independently of the uric acid presence, ROD was 2.0 μM/Gy, with essentially zero-order characteristics at oxygen concentrations >10 μM (Fig. 4A). The further addition of ascorbate (0.5 mM) led to a >3-fold decrease of ROD and imposed a minor oxygen dependence (Fig. 4A).

The relative contribution of components of CELL to ROD showed that glutathione was responsible for the enhancement of ROD above that for glycerol alone (addition experiments similar to Fig. 4B; data not shown). Substituting for glutathione either NADPH (1 mM) or cysteamine (5 mM) to glycerol (1000 mM) in HEPES buffer caused a substantial increase in ROD (1.06 μM/Gy for both solutions; first step of 3-step curves). Further addition of glucose (5 mM) and glutathione (5 mM) had no additional effect (Fig. 4B, second and third step of 3-step curves). A final addition of ascorbate led to some subsequent autoxidation but also a large drop in ROD. Interestingly, the oxygen dependence of ROD in the solutions containing ascorbate was different for the 2 initial reducing agents, having a much greater effect for the solution containing cysteamine (nearly first-order), compared with NADPH (nearly zero-order) (Fig. 4B).

Conditions demonstrating very high rates of ROD implicated the presence of chain or serial reactions and in turn suggested the likelihood of a dose-rate dependence. This was indeed observed and for the low-dose-rate Cs source (0.0085 Gy/s), ROD values were even higher than shown in Figs. 1 to 4 (Table 1). Conversely, for the 2 proton dose rates (particularly 100 Gy/s) much lower ROD values were observed (Fig. 5).

Fig. 5.

The dose-rate dependence of various solutions indicated a nearly constant value for relatively low radiation-chemical oxygen depletion (ROD) solutions (eg, Fricke dosimeter solution [+], phosphate buffer [stars], and deionized water [x]). CELL (solid squares) showed a modestly decreasing ROD as the dose rate changed from 0.0085 to 100 Gy/s. In contrast, solutions with high ROD at the lowest dose rate showed a dramatic decrease as the dose rate increased: NADH (10 mM, solid diamonds); cysteine (10 mM, open circles); and CELL plus additions (cysteine, 10 mM, dextrin-conjugated linoleic acid, 1 mM, NADPH, 1 mM; open squares).

Chain or serial reactions for high ROD values may propagate via peroxyl radicals²⁴ so we also investigated 3 examples in which these should not occur (Fig. 5, Table 1). The first was for phosphate buffer alone, and indeed the low ROD value observed at 0.1 Gy/s Cs was also observed for all dose rates. Similar findings were documented for pure water. The oxygen sensors performed normally in water, despite the extremely low ionic strength (Fig. 5, Table 1). Finally, we observed no dose-rate effect on ROD for Fricke dosimeter solution (0.8 N H₂SO₄ with 1 mM each ferrous ammonium sulfate and NaCl; Fig. 5).

Discussion

The goal of this study was to undertake a pragmatic survey of conditions that might substantially modify ROD values or might be suggestive of tumor-specific changes in the kinetics for ROD that would diminish its effectiveness relative to normal tissue (eg, as suggested in Fig. E1). Five types of substrate were examined: (1) DI water and buffer, (2) Fricke dosimeter solution, (3) primary free-radical scavengers, (4) reducing agents, and (5) mixtures of types 3 and 4.

For phosphate buffer and DI water (type 1), ROD values were <0.3 μM/Gy and demonstrated no dose-rate dependence for the complete range available (0.0085–100 Gy/s) (Figs. 2 and 5). Our result for DI water argues strongly against its use as a “model” for ROD in FLASH, a subject of much recent controversy.²⁵ Our observations are at odds with recently published data (Jansen et al) suggesting a much more complex behavior for ROD in water.²⁶ The Jansen et al study reported substantially higher ROD values in water than reported here and demonstrated an inverse square root dependency of ROD on dose rate. Key experimental differences between the methods used herein and those of Jansen et al include their use of 3-dimensionally printed acrylic-like vials (unstirred), incorporating a phosphorescence-decay oxygen detector, attached by silicone-sealant to the inner base of the vials. The contact time between the water and sealed (nylon screw) vials was very long (24 hours) to allow oxygen equilibration of the whole system from the vial’s exterior gas atmosphere in a reduced oxygen environment glove box. It is possible that these experimental details are relevant to the large differences in ROD and dose-rate effects found by Jansen et al compared with the results reported herein. However, from a radiation chemistry perspective, it is difficult to imagine any mechanism for chain/serial oxidations operating in distilled water. With no oxidizeable substrates, the primary free-radicals would be expected to react with each other producing low ROD values, with final molecular products being hydrogen peroxide and hydrogen.²⁷

The radiation chemistry for the Fricke dosimeter solution (type 2) is known in great detail.^28,29 We found no effect of dose rate for ROD in Fricke solution, expected in view of overwhelming data showing dose-rate independence of this dosimeter.²⁹ The value of ROD determined here (0.35 μM/Gy; Fig. 5, Table 1) is very close to the expected hydrogen radical yield (0.37 μM/Gy).

In the Fricke dosimeter, the only oxygen depleting reaction involves

$${\mathrm{H}}^{•}+{\mathrm{O}}_2\to {\mathrm{H}\mathrm{O}}_2^{•}$$ (1)

which then oxidizes ferrous to ferric while producing hydrogen peroxide. The hydrogen radical yield represents the sum of the endogenous ${\mathrm{H}}^{•}$, plus ${{\mathrm{e}}^{-}}_{\mathrm{a}\mathrm{q}}$, which reacts immediately with H⁺ from the acid.

The third type of substrate studied (primary free-radical scavengers) included glycerol and protein (lysozyme) solutions. These demonstrated ROD values of 0.3 to 0.4 μM/Gy, in keeping with the early literature.^12–14

Using the flash photolyis of hydrogen peroxide to produce pure hydroxyl radicals, Bothe et al³⁰ suggested that hydroxyl radical attack on glycerol would proceed via

(2)

Because the superoxide/hydroperoxy radicals would quickly dismutate to hydrogen peroxide plus oxygen, this would account for only about half of our observed ROD. It seems likely that the other half would arise from electrons, initially associated with glycerol but then transferred to oxygen, with a similar fate for the superoxide produced.

Because the major macromolecular component of cells is protein (~20%), protein solutions would appear at first glance to be the most appropriate model for investigation of intracellular ROD. However, intracellular proteins contain a very high reduced cysteine content (~20 mM—often referred to as protein sulfhydryl or PSH) whereas highly soluble proteins like albumin or lysozyme contain cysteine only in its oxidized (disulfide) form. An additional technical problem involved very high autoxidation rates when bovine serum albumin or serum were combined with reducing agents (data not shown).

ROD in protein solutions like lysozyme is unlikely to be as simple as glycerol (reaction 2, as discussed previously). The reactions of the primary oxidizing radicals are much more diverse for lysozyme and it is very unlikely that all of the initial peroxyl radicals would end up eliminating superoxide.³¹ In contrast, a majority of the primary aqueous electrons would likely produce anion radicals on lysozyme’s disulfides (R’SSR^•−) and these could transfer their electrons to oxygen, producing superoxide. Despite this diversity, the total ROD for lysozyme was only ~30% larger than for glycerol.

A general property of the primary free-radical scavengers investigated (type 3 substrates) was that their rate of ROD was only minimally dependent on concentration, in agreement with recent results using albumin and a different oxygen measurement method.¹⁶

In dramatic contrast, cellular reducing agents (type 4 substrates) demonstrated a wide distribution of ROD values (0.1–1.1) at concentrations of 1 mM and, for a few compounds tested, several fold higher values at 10 mM concentration (Figs. 2 and 3, Table 1). Although high rates of radiation-induced lipid oxidation were known to occur for pure unsaturated lipids at high pH (to allow solubilization), ROD was not directly measured in those studies.²⁴ Similarly, ROD has not previously been investigated for solubilized lipids at pH 7 such as the dextrin-conjugate of linoleic acid used here. Our additional observations regarding high rates of ROD in pyridine nucleotides, cysteine and cysteamine (at low dose rates) are novel, but in all cases examined these decreased monotonically with increasing dose rate. This dose-rate dependence had been explained for linoleic acid (at high pH), where an initial hydrogen abstraction, followed by oxygen addition to produce a peroxyl radical, can initiate a second hydrogen abstraction from another molecule, allowing a chain oxidation process. As the dose rate increases, the chains are terminated by radical-radical recombination.²⁴

ROD for cysteine and cysteamine is likely more complex in view of the variety of oxygen addition reactions to the thiyl radical.³² Additionally, thiol anion (RS⁻) can combine with thiyl radical (RS^•) to produce the disulfide anion radical, indicated previously (RSSR^•−). Aqueous electrons also react with thiols, producing R^• and HS⁻ (or other possibilities) and the resulting radical can then be oxidized.^33,34 Note that hydrogen sulfide is now considered an important signalling molecule and so its physiological effect could be affected by rate of production (ie, dose rate).^35,36

$$\begin{array}{l}{{\mathrm{e}}^{-}}_{\mathrm{aq}}+\mathrm{R}\mathrm{S}\mathrm{H}\to {\mathrm{R}}^{•}+\mathrm{H}{\mathrm{S}}^{-}\\ \mathrm{H}{\mathrm{S}}^{-}+{\mathrm{H}}^{+}\leftarrow \to {\mathrm{H}}_2\mathrm{S}\left(\mathrm{p}{\mathrm{K}}_{\mathrm{a}}7.0\right)\end{array}$$ (3)

An unexpected finding related to glutathione was the lack of a concentration effect for ROD by this ubiquitous reducing agent, possibly related to its ability to transfer the sulfur radical (after OH^• or H^• induced hydrogen abstraction from R-SH) to an internal carbon atom.³⁷ These so-called “reverse repair reactions” have now been widely studied for many thiols (for a recent review, see Schoneich³⁸), and determination of their role in ROD may be very important.

From all of this, it is clear that the reactions that occur in irradiated, oxygen-containing solutions of thiols ± other substrates are very complex³⁹ and may depend on the concentrations of O₂, thiol, ascorbate and urate, and likely pH (not studied here). Products that might be formed include thiyl peroxyl radicals as well as some hydrogen sulfide, the latter more likely at high concentrations of thiol or low concentrations of O₂. Reactions of the primary and secondary radicals with the pyridine nucleotides are expected to be just as complex and have not previously been studied in the context of ROD.

Our initial expectation was that all cellular reducing agents would increase ROD. We found the opposite result for ascorbate and uric acid. Indeed, uric acid (1 mM) in phosphate buffer demonstrated the lowest rate of ROD for all conditions tested (0.1 μM/Gy). Often, uric acid or ascorbate imposed an oxygen dependence on the rate of ROD and, especially for the case of ascorbate led to dramatic decreases in ROD for substrates and mixtures that otherwise would have had much more substantial rates of ROD (Fig. 4).

With respect to mixtures (type 5 substrates) our primary focus was with CELL (glycerol, 1000 mM, HEPES, 10 mM, glucose, 5 mM and glutathione, 5 mM; see Methods and Materials) ± additional reducing agents. We found that glutathione was the component that roughly doubled the ROD (0.35 to 0.7) in CELL, compared with glycerol alone (data not shown) but that the subsequent addition of ascorbate (1 mM) caused a decrease in ROD to much lower levels (0.41 μM/Gy), while additionally imposing an oxygen dependence to the ROD process (Fig. 4A). In mixtures using reducing agents with higher inherent ROD values than glutathione (NADPH, cysteamine) the same effect of ascorbate was found. Even larger effects of ascorbate were found for CELL plus additives (combined cysteine, dextrin-conjugated linoleic acid, NADPH) whereas uric acid had no effect in this complex solution. Thus, an important finding in our studies is the strongly inhibitory action of ascorbate on ROD.

Forni et al showed that ascorbate interacts with thiyl radicals (produced by addition of thiols to a system generating phenoxyl radicals) via electron transfer rather than hydrogen transfer.⁴⁰ The same effect was observed using NADH.⁴¹ Because we see greatly differing effects of adding NADH versus ascorbate in our ROD experiments, the possible effect of electron versus hydrogen transfer is presently unclear. This area requires a great deal more input from the radiation chemistry community. Both ascorbate and uric acid can be oxidized by the superoxide radical, unlike most thiols and pyridine nucleotides.⁴²

Although we have only tested a few combinations at FLASH dose rates (Fig. 5) our data for CELL showed the same trend, over a much larger dose-rate range, as the very recent demonstrations (using the phosphorescence decay method for detection of oxygen).^15,16,18 These showed that, in general, FLASH had ~15% to 25% lower ROD rates than conventional for simple protein solutions and CELL. Confirmation of this change by the present data is important because polarographic sensors have a response time much lower than the rapid oxygen changes caused by the 2 highest dose rates used here (0.5 but particularly 100 Gy/s). This implies that the oxygen concentration after the FLASH dose remained nearly constant, and was the critical parameter in determining rates of ROD. It is important to emphasize that the most dramatic changes in ROD occured at the 2 lowest dose rates where the oxygen sensors could easily keep up with the ROD process.

The inverse dose rate effect for lipid peroxidation was explained kinetically for unsaturated lipids by Raleigh et al²⁴ but the current studies show, for the first time and over a similar dose-rate range, the same effect for pyridine nucleotides and thiols. However, we do not yet know whether the mechanism is the same (propagation via hydrogen abstraction by peroxyl radicals, and termination via radical-radical recombination).

As indicated previously, reducing agents such as ascorbate (and possibly uric acid) have the property of imposing a substantial oxygen dependence to the ROD process (Fig. 4). An oxygen dependence of ROD might provide a direct mechanism to explain the lack of protection by FLASH in murine tumor models where one typically expects lower pO₂’s than for normal tissue at risk.¹⁶ Indeed, direct measurements of pO₂ in normal muscle versus subcutaneous tumor have confirmed these expectations.^15,18 However, this introduces a cautionary note because there are certainly examples of nonhypoxic tumors in humans, and essentially all hypoxic tumors will also contain well oxygenated regions. Additionally, ascorbate and uric acid have been much less-studied than thiols for their effect on radiation response.²⁵

Conclusion

Our data show that ROD is enhanced at low dose rates by glutathione and greatly enhanced by alternative thiols (cysteine and cysteamine), the pyridine nucleotides (NAD(P) H), and unsaturated lipids (only minimally exemplified by dextrin-conjugated linoleic acid). Not all reducing agents behave in this manner, and uric acid, but particularly ascorbate, inhibits ROD when added to mixtures that would otherwise have high rates of ROD. These 2 reducing agents also cause a transition from nearly zero-order to first-order dependence of ROD on oxygen concentration. All conditions having high rates of ROD at low dose rates show a decrease at FLASH dose rates, and this seems to be a proportional response (ie, the greater the ROD at low dose rate, the greater the inhibition by FLASH). FLASH, even in the absence of ascorbate, may impose an additional dependence of ROD on oxygen concentration at low oxygen levels.¹⁸ A decrease in the rate of ROD at low oxygen levels could cause a decrease in effect for tumors, generally considered to have a lower overall oxygen level than their tissue of origin. We note that aqueous electrons, typically ignored in terms of biological radiation effects, can cause the formation of the signalling molecule hydrogen sulfide on reaction with thiols. Thus, its rate of formation could affect tissue physiology. Most importantly, the novel effects of ascorbate observed here emphasize the surprising lack of information about the role of this essential antioxidant in radiation biology.²⁵

For simple solutions such as pure water, low ionic strength phosphate buffer, and the well-characterized Fricke dosimeter solution, no dose-rate dependence of ROD was observed. Thus, ROD values for FLASH in vivo recently reported by Van Slyke et al (~0.4 maximum in extracellular space, oxygen dependent) and roughly 0.8 for CELL may represent reasonable starting points for modeling purposes.¹⁸