Mechanisms, challenges and opportunities for FLASH radiotherapy in cancer

Abstract

FLASH radiotherapy has the potential to improve both patient quality of life and outcomes by delivering radiation at ultrahigh dose rates to effectively target tumours while sparing healthy tissues. However, the differential sensitivity of healthy tissues versus tumours to FLASH radiotherapy remains unexplained. In this Perspective, we hypothesize that FLASH radiotherapy distinguishes healthy tissues from tumours based on subtle functional and structural biological differences. We identify commonalities present in the various healthy tissues that are spared by FLASH radiotherapy that might be lost during tumorigenesis. We also propose that a specific class of proteins, termed long-lived proteins, define a critical radiolytic target that are present in nearly every healthy tissue that is FLASH radiotherapy resistant yet are absent in tumours. We extend this structural hypothesis further by suggesting that tumour and extracellular matrix rigidity affects sensitivity to changes in radiotherapy dose rate, where more rigid and dense desmoplastic tumours are more sensitive to FLASH radiotherapy than those possessing more elasticity. Substantiating these concepts experimentally may provide a new and generalized mechanism of action of radiation effects and may therefore inform clinical trial designs by identifying those tumour subclasses expected to exhibit optimal responses to FLASH radiotherapy.

Introduction

After the initial discovery of X-rays and radioactivity at the end of the nineteenth century, radiation therapy emerged as a potent anticancer strategy. Since this discovery, more sophisticated methods, tools and devices have been developed to characterize and deliver ionizing radiation in a safer and more effective way. Today, radiation therapy stands as a fundamental and cost-effective pillar in the treatment of cancer, with over 50% of patients with cancer receiving it as treatment for both curative and palliative intents¹. However, despite advancements in imaging and treatment planning, many patients who undergo definitive or adjuvant radiotherapy still experience local relapse or disease progression, termed radioresistance². In addition, radiation induces toxicities both acutely and later in time, which must be addressed as these limit a patient’s quality of life³. Today, the therapeutic progress driven by the implementation of image-guided and conformal radiotherapy has reached a plateau, and improving radiotherapy delivery precision alone is unlikely to cure radiation-resistant tumours or prevent recurrence without toxicity. However, a decade ago, we introduced FLASH radiotherapy, a novel method for expanding the therapeutic index of radiotherapy⁴. Despite its capitalized name, FLASH is not an acronym, and is defined by a combination of physical parameters and biological effects. The fundamental physics principle is the ultrahigh dose rate, which means that the therapeutic dose is delivered within milliseconds or less, at an ultrahigh dose rate above 100 Gy s⁻¹, with electron, proton or X-ray beams (Box 1). This contrasts with the conventional dose rate used for radiation therapy, which requires minutes to deliver the same dose at rates of 1–4 Gy min⁻¹ (Fig. 1). The biological consequences of this ultrahigh dose rate effectively eliminates tumour cells while minimizing damage to healthy tissues. Although the benefits of ultrahigh dose rates on healthy tissues have already been described^5,6, the real potential of FLASH came to the forefront when it was demonstrated to ablate pulmonary tumours in a syngeneic model and delay the growth of xenografted subcutaneous breast and head and neck tumours while sparing healthy lung tissue⁴. The resultant differential biological effect was then named the ‘FLASH effect’. Notably on further experimental applications of FLASH radiotherapy delivered with electrons, protons and X-ray beams, most tumours have been found to be equally sensitive to radiotherapy at FLASH and conventional dose rates, whereas healthy tissues were always spared by FLASH radiotherapy (reviewed in refs. 7,8). Today, some of the physical parameters required to systematically produce the FLASH radiotherapy effect on biological systems have been identified and used in preclinical studies; however, whether they can be scaled up for clinical use is still under investigation. Herein, we will focus on the mechanisms underlying the differential FLASH radiotherapy effect.

Box 1 |. The current technological landscape and developmental trajectory of FLASH radiation-capable accelerators.

The clinical translation of FLASH radiotherapy will require extensive advances in the development of FLASH-capable accelerators⁸. Most preclinical observations have been made using first-generation FLASH-capable experimental electron linear accelerators (named Kinetron and Oriatron), which have intermediate-energy levels (4–6 MeV) and produce pulsed electrons at dose rates ranging from 0.01 Gy s⁻¹ to 10⁶ Gy s⁻¹, meaning that the dose can be delivered over durations ranging from minutes down to microseconds. However, they are not optimal for clinical applications owing to their limited penetration in tissues (~1–2 cm).

The second generation of FLASH-capable irradiators have been adapted from existing radiotherapy devices. These include electron beams commonly used in intraoperative settings where the power supply has been boosted, clinical X-ray accelerators where the filters have been removed to maintain high dose rates and clinical proton beams tuned to produce higher dose rates. These modified clinical devices are generally able to reach ultrahigh dose rates in the range of 100 Gy s⁻¹, thus delivering the dose within milliseconds, enabling early clinical applications. The proton FLASH beams also allow treatment of tumours located in deep anatomical regions (rather than superficial).

The third generation of FLASH-capable devices is currently in development, with the intention to treat large fields (10 cm by 10 cm and larger) and deep-seated tumours. These devices use very-high-energy electron (defined as electrons with an energy above 100 MeV) beams, laser-driven electron and proton beams, superconducting materials or compact synchrotrons to produce X-rays. They can achieve ultrahigh dose rates radiation ranging from 100 Gy s⁻¹ to 10¹¹ Gy s⁻¹, delivering the dose in milliseconds to picoseconds.

Whereas a dose rate above 40 Gy s⁻¹ is often cited as necessary to produce the biological FLASH effect in experimental murine models, the optimal dose rate in patients is not yet defined and therefore requires experimental refinement. Given that shifting the time required to deliver the dose from 1 min to 100 ms can modify the biological response⁵², it becomes challenging to forecast what effect shifting this time from 100 ms to 1 μs or even 1 ps would have on the biological response. Although current experimental evidence suggests that radiation in the millisecond and microsecond range is still cytotoxic to tumours but spares healthy tissues¹¹⁹, much less is known about radiation dosed within the nanosecond or picosecond ranges. This topic has been extensively reviewed elsewhere⁸; however, being able to fully characterize and optimize the physics parameters required to produce the FLASH effect is crucial for building the next generation of FLASH-capable accelerators.

Fig. 1 |. The relative timescale of measured radiation events involving physical, chemical and biological processes.

The exposure of biological tissue to radiation induces changes over discrete timescales that cover physical, physicochemical and chemical processes. The various dose rates introduced by FLASH radiotherapy are shown along the top and can be delivered with various beams including, very-high-energy electron (VHEE), intermediate-energy electron (IEE) and transmission irradiation (TI) with protons. The associated colours align with the elicited effects during dose delivery. Increasing the dose rate alters the temporal dynamics of beam–matter interactions and modifies the cascade of radiochemical reactions that propagate and evolve within the biological system over time. These reactions modify both early and late radiation-induced biological response and ultimately affect healthy tissue turnover, patient outcomes (for example, cognition or fibrosis) and tumour control.

First, we examine the early physicochemical hypotheses alongside the biological mechanisms investigated thus far, noting that neither offers a comprehensive explanation of the FLASH effect. Then, we propose new directions and a hypothesis where FLASH radiotherapy acts as a physical probe, able to uncover subtle structural differences between healthy and tumour tissues. We suggest novel mechanisms by which the initial structural and mechanical properties of tissues could drive the differential effect of FLASH radiotherapy, further amplified by distinct metabolic alterations. To support our hypothesis, commonalities present in the various healthy tissues known to be spared by FLASH radiotherapy will be identified. Next, we will investigate whether these elements are lost or absent in tumours and their microenvironment. We will explain our structural hypothesis based on the presence of long-lived proteins (LLPs) in healthy tissues and their absence in tumours and tissue stiffness. Finally, we will propose therapeutic indications based on these hypotheses to pave the way for more effective and less harmful radiotherapy.

Clinical implications

On the basis of the preclinical efficacy of FLASH radiotherapy in multiple animal models and diverse tissues⁹, two case reports and the first clinical study have^10–12 demonstrated that FLASH administration is clinically feasible using available beam lines. One patient with superficial skin lymphoma received a single dose of 15 Gy electron FLASH radiotherapy localized to a single lesion, using an experimental electron beam (Oriatron eRT6; Box 1)¹⁰. Comparing this dose of FLASH radiotherapy versus conventional dose rate radiotherapy applied to other lesions in the same patient showed no difference in acute reactions, no side effects at a 2-year timepoint and overall tumour control¹¹. A feasibility study (FAST-01) was performed, where ten patients with bone metastases in the extremities were successfully treated with a clinical proton beam line optimized to deliver FLASH radiotherapy using a single dose of 8 Gy to the lesion¹². In the two case reports and the first clinical study, the doses used were equal to the ones used in conventional radiotherapy and known to induce minimal toxicity. This was a necessary step for the clinical transfer of FLASH radiotherapy and confirmed that no unexpected toxicity was induced in humans. A follow-up phase I trial, FAST-02 with the same proton FLASH radiotherapy beam line, is ongoing and designed to treat thoracic bone metastases with a single dose of 8 Gy (refs. 13,14). Two trials also started with electron FLASH radiotherapy beams: the IMPulse trial¹⁵, a phase I trial assessing the maximum tolerated dose of single-fraction FLASH radiotherapy for melanoma skin metastases and the LANCE trial¹⁶, a randomized phase II trial comparing FLASH and conventional dose rates for the treatment of cutaneous basal cell and squamous cell carcinomas using single- and multifraction regimens¹⁷.

Ultimately, further clinical trials are needed to demonstrate the efficacy of FLASH radiotherapy, which has the potential to address three unmet clinical needs (Box 2). The first is the possibility to cure radioresistant tumours with poor baseline therapeutic outcome, such as sarcomas and other common carcinomas, by dose escalation. Typically, these tumours respond to ~60–70 Gy doses of radiotherapy, but patients commonly relapse¹⁸. FLASH radiotherapy provides the opportunity to pursue curative intent using higher doses as enhanced dose rate escalation is associated with local control rate¹⁹. A second unmet clinical need is effectively treating recurrent, previously irradiated disease with a second course of radiotherapy as this protocol enhances the risk of severe toxicities²⁰. Recently, re-irradiation with FLASH radiotherapy in tumour-free mouse models showed good tolerance in the skin, gut and bone, supporting the use of FLASH radiotherapy for treating sarcoma and pelvic tumours²¹. A third unmet clinical need is the development of strategies that avoid damaging the remaining organ tissue or surgically removing the entire organ with cancer, such as those surgeries often performed for treating rectal cancer. These improvements would be beneficial for all patients with cancer provided FLASH beams can be shaped to match the three-dimensional shape of the tumour (conformality; Box 3) and fractionated to allow healthy tissues time to repair between doses while increasing tumour cell killing (fractionation; Box 4). Combining the benefit of FLASH radiotherapy with conformality and fractionation is expected to further enhance the therapeutic index.

Box 2 |. A clinical decision tree conceptualizing how FLASH might facilitate clinical workflows and provide alternative treatment options for improving patient care.

From the time of diagnosis, clinicians evaluate patients for a variety of medical conditions that ultimately lead to the selection of a specific radiotherapy treatment plan. Current limitations of conventional treatment plans can be expected, under certain circumstances, to be greatly improved with the opportunities afforded by FLASH in the clinic. Outcomes provided by FLASH are projected that span multiple segments of the clinical and patient experience, resulting in overall improvements in patient care (see the figure). Most notably, FLASH radiotherapy reduces the time required for dosing a single patient on a linear accelerator (LINAC), and thus increases the number of patients that can be treated. This is particularly important for cancer care in countries with limited resources.

Box 3 |. Integrating FLASH radiotherapy with spatial dose optimization.

Modern radiotherapy at conventional dose rates involves technological innovations made to improve the conformality of physical dose delivery specifically to the tumour. It consists of image-guided procedures that allow for precise tumour delineation and enable focused delivery of higher doses of radiation to the tumour while sparing healthy tissues. Techniques for accommodating organ or tissue movement during radiotherapy have also contributed to improved treatment accuracy and safety.

Although none of these advances have been implemented for use with FLASH radiation so far, these technologies are needed as they will further enhance the therapeutic window of FLASH radiotherapy. Scanning with X-ray FLASH beams⁶⁰ and proton FLASH beams^51,120 has been shown to maintain the sparing effect of FLASH radiation, suggesting that sequential beam shift can be implemented and targeting the entire tumour volume with multiple FLASH beamlets can be achieved. Split delivery experiments, where the radiation dose is divided into short intervals to mimic the pauses that typically occur within a single treatment when the radiation beam arm is repositioned, showed that the protective effect gradually decreased when the intervals were extended to 2 min^121,122. However, normal tissue was spared when the doses were spaced out by 1 min⁵⁸ using a proton FLASH beam at dose rates of approximately 100 Gy s⁻¹, suggesting that poses of 1 min or less maintain the sparing effect of FLASH radiotherapy and that this is a threshold to take into consideration when implementing conformal FLASH radiation. Interestingly, when an electron FLASH beam (eRT6) that operates at an instantaneous dose rate above 10⁶ Gy s⁻¹ was used, the normal tissue was spared when the dose was spaced out by 10 min²⁶. It also showed that higher dose rates may well provide more suitable alternatives for conformal developments. Therefore, future FLASH radiotherapy systems will probably utilize beam scanning and rotation to achieve more precise dose conformality and thus improve patient outcomes.

Box 4 |. FLASH and dose fractionation.

Although preclinical studies on FLASH radiotherapy have primarily used single-dose irradiation for practical reasons, there was initially a widely held but mistaken belief that FLASH radiotherapy could not be fractionated. This view, initially supported by modelling studies and the oxygen depletion hypothesis¹²³, suggested limited clinical applicability. However, subsequent experimental evidence convincingly refuted this hypothesis¹²⁴, demonstrating that doses of FLASH can indeed be effectively fractionated.

As the field gradually shifts towards delivering larger and therefore fewer fractions, hypofractionated FLASH regimens have been investigated more recently. Fractionated regimens delivered with electron FLASH radiotherapy reduced tumour progression^61,114. Moreover, in a mouse model of head and neck cancer, the use of a hypofractioned proton FLASH radiation decreased mucositis and preserved both saliva production and bone integrity as compared with proton radiation at a conventional dose rate¹²⁵. Cognitive sparing in tumour-free animals was also demonstrated following hypofractionated FLASH radiation dosing in both adult (3 × 10 Gy (ref. 126)) and juvenile (2 × 10 Gy (ref. 41)) mice. The latter is particularly relevant for paediatric patients with brain tumours as the juvenile brain is known to be highly radiosensitive and prone to late toxicities⁶².

So far, the effect of a standard of care fractionation consisting of 2–3 Gy per fraction regimen has only been investigated in one publication. In this study, tumour-free animals were given ten fractions of 3 Gy whole-brain irradiation, a regimen used for the treatment of brain metastases in human patients was delivered at FLASH dose rate (10⁶ Gy s⁻¹)¹²⁴. Long-term potentiation, a process involving increased synaptic strength at neuronal connections, was preserved in FLASH-irradiated animals, whereas conventional dose rate irradiation inhibited long-term potentiation. This demonstrates that FLASH radiation can still spare healthy tissues with a standard fractionation protocol. More studies are needed to determine the benefits of various fractionation regimens but, thus far, evidence indicates that the increased therapeutic index of FLASH radiation can be maintained after dose fractionation.

Beyond these unmet clinical needs, a pragmatic approach could involve using FLASH radiotherapy in situations where conventional radiotherapy is used in the adjuvant postoperative setting and achieves local control with moderate toxicity, such as in patients with breast cancer²². For this large population of patients, FLASH radiotherapy could further reduce toxicities and optimize cosmetic outcome²³. Finally, as FLASH radiotherapy can be delivered within a short amount of time (2–5 min with conventional dose radiotherapy versus less than 100 ms with FLASH radiotherapy) and with larger doses per fraction, the overall number of fractions and treatment sessions would be substantially reduced, therefore saving time for patients and clinicians. The ultrashort irradiation time of FLASH radiotherapy may also help to reduce targeting errors caused by organ or tumour movement during treatment. For example, breathing and peristalsis, respectively, interferes with treating thoracic and pelvic tumours with conventional radiotherapy⁹. Therefore, implementing FLASH radiotherapy could streamline clinical workflows and reduce treatment-related costs.

The mechanistic bases of the FLASH effect

The physicochemical hypothesis

One of the earliest hypotheses in the field, drawn from radiation chemistry proposed that the FLASH effect was a physicochemical phenomenon initiated when radiation first interacts with the biological milieu (Fig. 1). The initial physicochemical reactions triggered by FLASH radiotherapy could create primary free radicals at a different rate compared with conventional dose rate irradiation.

This hypothesis was tested in cell-free aqueous-based systems, and the primary radiolytic yields of H₂O₂, used as a surrogate marker for secondary end product of radiolytic free radical reactions, was used to investigate the effect of irradiation with FLASH (~100 Gy s⁻¹) and conventional (0.1 Gy s⁻¹) dose rates using pulsed electrons, protons and gamma rays^24,25. The results showed that the primary radiolytic yields of H₂O₂ remain unaffected by dose rate modulation from 0.1 to 100 Gy s⁻¹. However, when the dose rate was increased above 10⁹ Gy s⁻¹ using very-high-energy electron (VHEE; electron energies >100 MeV) radiotherapy, a lower primary yield of H₂O₂ was found, suggesting a rapid recombination of radicals in overlapping ionization tracks and spurs²⁶. Beyond the primary yield, the kinetics of subsequent chemical reactions leading to H₂O₂ production seems dose-rate sensitive. However, contradictory findings have been reported. Prior results demonstrated an increase of H₂O₂ production when using pulsed electron beams at ultrahigh dose rates (5 × 10⁶ Gy s⁻¹) compared with conventional gamma-ray radiation delivered at lower dose rates (1 × 10⁻³ Gy s⁻¹) in high-oxygen conditions^27,28. More recent results report reduced secondary production of H₂O₂ with FLASH radiotherapy at doses ranging from 100 Gy s⁻¹ to 10¹¹ Gy s⁻¹ under atmospheric oxygen pressure (2.5 × 10⁻⁴ M)^24,25. The implications of these differences and whether or how they translate to biological systems remains uncertain. Moreover, these processes remain difficult to investigate directly owing to multiple confounding factors, including the presence of scavengers and antioxidants, the variable rates of enzymatic reactions and cellular compartmentalization. Notably, however, the effect of dose rate on plasmid DNA, peptides and lipids are available, yet at times contradictory. Some studies have shown that DNA damage induced by doses below 10 Gy in cell-free systems is dose-rate insensitive and remains unchanged when beam characteristics, such as the mean dose rate, the dose rate per electron bunch or pulse, the beam structure, the particle type and the particle energy, were modified^26,29–31. In these studies, modification of the dose rate did not modify the DNA damage induced in plasmids, and this was maintained under biochemical conditions that mimic some aspects of the tumour microenvironment, such as hypoxia and high Fe²⁺ levels²⁸. Other reports showed reduced DNA damage in plasmids using FLASH electron beams (46.6 Gy s⁻¹) compared with conventional dose-rate radiotherapy³², and reduced levels of single-strand breaks with FLASH VHEE versus VHEE radiation³³, but only at doses above 90 Gy, far above clinical doses. These latter results suggest that although the extent of DNA damage is not modified by an ultrahigh dose rate, DNA damage might be reduced when an ultrahigh dose rate is used in combination with high doses.

Moreover, in the context of FLASH radiotherapy, other biological molecules such as lipids and proteins have been less extensively studied than plasmids. Nevertheless, the effect of FLASH on these alternative molecular targets has been investigated in cell-free systems^34,35. For instance, oxidation levels in micelles and liposomes remain near zero after FLASH irradiation, whereas the yields of lipid peroxidation end points increase linearly with dose after conventional dose rate irradiation³⁴. A recent publication quantified the expression of oxylipins, bioactive lipid hydroperoxides derived from the oxygenation of polyunsaturated fatty acids, finding lower levels in the lungs of mice after FLASH irradiation compared with conventional dose rate radiotherapy³⁶. When modelled in vitro, this effect was evident in healthy cells at 37 °C, but not in tumour cell or cells at 20 °C. Additionally, peptide oxidation was reduced upon FLASH irradiation³⁷. In another study, brain sections from mice 24 h post-exposure to FLASH or conventional dose rate irradiation were characterized using an analytical method based on Fourier-transform infrared spectromicroscopy³⁸. Distinct peptide imprints were observed between the two radiation modality groups associated with modifications in the protein backbone, as indicated by changes in the amide I signal. Specifically, samples irradiated at conventional dose rates had increased β-sheet content relative to the α-helical secondary structures, whereas samples irradiated FLASH retained a secondary structure profile similar to the non-irradiated controls. These changes in protein backbones found in the samples irradiated at conventional dose rates were concomitant with nucleic acid fragmentation and condensation, alterations in methylene and methyl group concentrations, and variations in lipid chain length.

Although reductionist physicochemical studies are conceptually interesting, they fail to provide a unifying mechanism explaining the differential biological effect of the FLASH effect on tumours versus healthy tissues. FLASH radiotherapy-induced decreases in reactive oxygen species (ROS) levels — which lower the downstream peroxidation cascade and reduce DNA, lipids and protein damages — should protect both tissues equally, which is not what is observed in vivo. This suggests that the main determinant of the FLASH effect must be found in the biological substrate. The recent results showing the role of physiological temperature³⁶ and the ones obtained with Fourier-transform infrared spectromicroscopy³⁸ corroborate the idea that the main factors involved in the differential FLASH effect involve structural differences in healthy tissue and tumours. Findings from a recent MRI-based study performed on the healthy brain of mice 1 year after exposure to FLASH and conventional radiotherapy support this idea further³⁹. Various diffusion metrics, including the apparent diffusion coefficient and mean apparent propagator, were unchanged after FLASH radiotherapy, but were markedly altered in mice that received conventional dose-rate radiotherapy. These measurements support the idea that FLASH does not elicit the same level of structural disruption, suggesting a preservation of synaptic elements and neurotransmission. These data corroborate previous structural and functional outcomes in the brains of FLASH-irradiated mice^40,41, where conventional dose rates caused tissue damages.

Healthy tissues are resistant to FLASH

In classical radiobiology, organs exhibiting sensitivity to early phase radiation effects, termed ‘early responding organs’ are those composed of cells with high proliferative capacity and a rapid turnover. This includes cells in the gut and haematopoietic system, including enterocytes, colonocytes and keratinocytes. In these organs, radiation destroys the resident cells and disrupts cell division and tissue repair^42,43. Notably, zebrafish embryos can also be considered as early responding organs as they are composed of highly proliferative cells⁴². In contrast, organs that do not exhibit sensitivity to radiation until later timepoints, termed ‘late-responding organs’ include the brain, the heart, the dermis and the lungs. These are composed of cells that divide infrequently, such as fibroblasts and endothelial cells, or those that are postmitotic, such as neurons and cardiomyocytes⁴⁴. In these organs, damage occurs in a delayed manner as the result of vascular changes, fibrosis and progressive loss of functionality due to impaired repair mechanisms^45,46. Thus, the sequence and timing of tissue and functional toxicities are a result of the molecular and cellular damage induced by ionizing radiation on a given organ (Fig. 2).

Fig. 2 |. The temporal and biological dynamics of radiation response.

Organs that exhibit sensitivity to radiation in the days and weeks following exposure are termed ‘early responding’ tissues, whereas those that exhibit sensitivity in the months and years following exposure are termed ‘late-responding’ tissues. These distinctions are important for understanding radiation effects, planning radiotherapy and predicting side effects. After exposure to ionizing radiation, immediate damage to biomolecules, and especially to DNA, activates the DNA damage response (DDR) pathways, leading to either DNA repair and cell survival, or to sustained cell injury. In the situation where DNA repair is efficacious and allows cell survival, the potential errors occurring during DNA repair can lead to the maintenance of genetic mutations in tumour suppressor genes or oncogenes that can, decades after radiation exposure, lead to radiation-induced secondary cancers¹¹⁸. In the situation where DDR pathways are overwhelmed and DNA damage is irreparable, early responding organs, characterized by a high cellular turnover, typically exhibit radiation effects within days post-exposure. These effects are primarily characterized by cell death or senescence. In the weeks following exposure, epithelial cell barriers can break down, inflammation can become chronic and the vasculature can exhibit damage. Over months, fibrosis can accumulate. By contrast, in late-responding organs, with slow cellular turnover rate, the damage induced by radiation primarily results from the gradual accumulation of structural and functional defects.

Interestingly, although FLASH radiotherapy does not modify the sequence and timing of the toxicities, it increases the dose threshold required to induce these toxicities in early^47–51 and late-responding organs^4,40,52. Although it is possible that in some organs FLASH does not show any superiority over conventional dose rate radiotherapy (including the liver, as proposed later), the dose-modifying factor (DMF; also called the FLASH-modifying factor)⁵³ is larger in late-responding organs^40,51 than in early responding organs⁴⁹. This observation suggests a biological difference between these organs, where the large DMF values, owing to the slowly proliferating, late-responding tissues, translates into enhanced healthy tissue sparing. However, common sparing patterns persist and might help to identify or exclude intrinsic FLASH radiotherapy-responsive elements. For instance, cell cycle regulation, mitosis, acute cell death, DNA damage and repair, and specific differentiation patterns are divergent in early and late-responding organs⁴³, suggesting that those elements are not critical to mediate the response of healthy tissues to FLASH irradiation. This concept is also supported by results obtained with zebrafish embryos^24,54, where the embryonic cells are mitotically active and able to differentiate into any cell type constituting a zebrafish within the first 24 h of fertilization⁵⁵. When zebrafish embryos are irradiated with electron or proton beam FLASH radiotherapy at 4 h post-fertilization, an apoptotic peak occurs 24 h later and is accompanied by a high rate of proliferation. Conversely, zebrafish embryos irradiated at a conventional dose rate showed the apoptotic peak but not the proliferative response. A 5 days post-irradiation, only FLASH-irradiated zebrafish embryos presented a spared morphogenesis, suggesting that FLASH radiotherapy preserves the repair potential of the embryo²⁴.

Analyses of FLASH radiotherapy-induced transcriptomic changes in tumour-free C57BL/6 mice are becoming available^56–58. These results may help define common genomic patterns in FLASH-irradiated tissues, as well as any similarities or differences between tissues irradiated at similar doses using either FLASH or conventional dose rates. Confounding these analyses are the variable conditions used across such experiments, including differences in the FLASH modalities used (that is, electron versus proton), the dose rates (10²–10⁶ Gy s⁻¹), the doses (10–40 Gy) and the regimen (that is, single- or hypofractionated) delivered. In addition, tissues have been collected for transcriptomic analysis at different timepoints. For example, transcriptomic profiling has been performed on lungs irradiated at 17 Gy collected after 4 days⁵⁶; on bones and skin irradiated at 30 Gy collected after 1 month and 5 days, respectively⁵⁷; on heart irradiated at 40 Gy collected after 2 weeks⁵⁸; and on brain irradiated at 10 Gy or 3 × 10 Gy and collected after 24 h. Various RNA profiling methodologies have also been used including bulk RNA sequencing (on bones, skin and heart^57,58), single-cell RNA sequencing (on lungs⁵⁶) and whole-transcriptome digital spatial profiling. In all organs, classical radiation-induced pathways such as the DNA damage response, p53 signalling, induction of tumour necrosis factor superfamily, cell death pathways, and inflammatory and fibrogenic responses were found to be expressed at lower levels in the FLASH samples as compared with samples irradiated at conventional dose rates^56–58. Conversely, repair pathways, such as vascular and epithelial wound healing were sustained after FLASH irradiation but not after conventional dose rate irradiation^56,57. Heart and skin regenerative pathways were uniquely upregulated in FLASH samples^57,58. Organ-specific differentiation pattern such as interstitial pulmonary macrophage⁵⁶, osteoclast differentiation⁵⁸, endochondral bone morphogenesis⁵⁸ and synaptic plasticity were also sustained after FLASH irradiation. The combined results of these transcriptomic studies suggest that, unlike conventional dose rate irradiation, FLASH irradiation induces less apparent healthy tissue damage and/or promotes a repair response, two factors that could help maintain organ function (Fig. 3). These results raise an important question as to whether the genomic response is specific to FLASH radiotherapy or similar to conventional radiation dose rates but occurring at a higher dose. This question remains unanswered as studies have been performed at isodose but not at isotoxic levels.

Fig. 3 |. The differential toxicity effects induced by FLASH and conventional dose rate radiotherapy in healthy tissues and tumours.

The therapeutic index (or therapeutic window) of radiotherapy is described by the combination of tumour control probability (TCP) and the normal tissue complication probability (NTCP). The TCP curves often rise steeply with increasing doses and reach high values earlier than NTCP curves, meaning that tumour control can be achieved before normal tissue complications become dominant. Radiotherapy at conventional dose rate delivers radiation to the tumour over minutes, typically 1–4 Gy min⁻¹. The surrounding healthy tissues limit the level of dose that can be delivered to the patient as radiation can induce direct and indirect killing of non-cancerous, resident cells through the production of reactive oxygen species (ROS) as well as damage to the DNA and other biomolecules. Acute and delayed cell death is followed by a complex series of events including stimulation and release of inflammatory factors and thrombotic factors, vascular activation and immune cell recruitment, as well as the activation and differentiation of fibroblasts into myofibroblasts that are able to remodel the extracellular matrix (ECM), inducing further activation. These molecular and cellular events disrupt tissue homeostasis and can cause a loss of organ function, impairing the patient’s quality of life and, in severe cases, their overall survival. FLASH radiotherapy compresses the time needed for radiation delivery to milliseconds or less, typically 100 Gy s⁻¹, thereby increasing the dose threshold that could induce the toxicities associated with conventional dose rate radiotherapy. Although the DNA damage response seems to be similar after exposure of healthy tissue to FLASH or conventional dose rate radiotherapy, FLASH radiotherapy reduces the amount of oxidative stress and inflammation induced; preserves oxidative phosphorylation (OXPHOS), the vasculature and stem cell pool, and the expression of genes related to the tumour necrosis factor superfamily and cell death, as well as reduces inflammatory and fibrogenic responses. In tumours, both conventional and FLASH radiotherapy induce ROS production, damaging DNA and other biomolecules. FLASH and conventional dose rate radiotherapy are isoeffective in most experimental models investigated. They induce similar adaptive immune responses; however, FLASH radiotherapy reduces immunosuppressive signals and M2 polarization of macrophages due to the decreased ROS production and associated reduction of oxidized low-density lipoproteins (oxLDL), repressing the activation of the PPARγ pathway promoting M1 polarization.

Interestingly, the relative resistance of healthy tissues to FLASH radiotherapy is also maintained with age. Juvenile (3 weeks old)^41,59, adult (8–20 weeks old)^40,52,60,61 and old (>55 weeks old) animals maintain a relative cognitive acuity after whole-brain FLASH irradiation, whereas exposure to conventional dose rate irradiation at any age induces irreversible cognitive deficits, with juvenile brains being the most sensitive⁶². These results suggest that ageing-induced physiological changes do not alter the healthy tissue response to FLASH. For example, many biological processes decline with age, including mechanisms involved in the maintenance of genetic and epigenetic programmes⁶³. These processes encompass DNA damage and repair, telomere maintenance, DNA methylation, histone modifications and chromatin remodelling⁶⁴. Additionally, the maintenance of cell signalling pathways that affect proteostasis, mitochondrial function, metabolism, stem cell quantity and stem cell function are also affected⁶⁵. Each of these processes is required for the responses to injury, chronic oxidative stress and inflammation⁶⁴, all of which are associated with age-related diseases⁶⁴. Although changes in epigenetics, chromatin structure and proteostasis have not been investigated after FLASH radiotherapy, other mechanisms have. Compared with conventional dose rate radiotherapy, FLASH radiotherapy has also been found to reduce oxidative stress and inflammation in the brain^40,66–69; preserve the stem cell pool in the brain, lung and gut^{49,52,56,57,60}; and maintain metabolic function in the lung and heart^56,58. The protective effect of FLASH radiotherapy against radiation-induced fibrosis was shown to be lost in Terc-deficient mice, which present with short telomeres⁵⁶, suggesting that telomerase activity could be involved in inducing tissue-protective effects from FLASH radiotherapy. Other mechanisms, such as the DNA damage response, were found to be similar after FLASH radiotherapy versus conventional dose rate irradiation in vivo⁴⁷ and in vitro⁷⁰. Importantly, these studies are correlative, meaning that the end points reflect the consequences of FLASH radiation exposure, but are not necessarily the primary determinants of the FLASH effect. This requires further investigation.

Tumours are sensitive to FLASH

Although FLASH radiotherapy induces fewer healthy tissue complications, the antitumour response remains comparable to conventional dose rate irradiation⁴. This unexpected and counterintuitive differential effect raises two important questions: what makes cancer cells specifically sensitive to FLASH radiotherapy and are all cancer cells equally sensitive to FLASH radiotherapy?

The understanding of tumour radiosensitivity was initially based on the ‘4Rs’, which describe the relevant pathways dictating the response of tumours to ionizing radiation: repair of DNA damage (R1), redistribution of cells in the cycle (R2), re-population (R3) and re-oxygenation of hypoxic regions (R4)⁷¹. Later, more functional and complex outcomes were included in the radiobiology portfolio. They include radiosensitivity defined by the intrinsic response of different tissues and cells (R5), reactivation defined by extrinsic elements making tumour cells more sensitive to radiation (that is, oxygen, drugs and the immune system) (R6) and reinforcement by extrinsic factors from the microenvironment (R7)⁴³. Today, radiotherapy treatments are still rarely designed according to tumour-specific biological features but are instead primarily designed on the assumption that all cancer cells within the same tumour will respond similarly to radiation. Empirically, this assumption is true as the tumour response is primarily dependent on the delivered dose and the fact that higher doses kill better⁷². However, improvements can be expected from better characterization of tumour biology and integration of this mechanistic knowledge.

Although the antitumour responses to FLASH and conventional radiotherapy are compared at isodoses⁷, mechanistic differences have only been found in some of the 7Rs. For instance DNA damage (R1) has also been found to be identical after FLASH and conventional dose rate radiotherapy in all solid tumours (lung and ovararian cancer) and tumour cell lines (lung, glioblastoma, breast, prostate, and head and neck cancer) tested so far^4,56,70,72. Clonogenic survival, a functional assay to assess DNA damages and the largely accepted gold standard for assessing intrinsic radiation sensitivity (R5), was also found to be equivalent between FLASH and conventionally irradiated tissues below 10 Gy (ref. 40). However, results obtained on primary human T cell acute lymphoblastic leukaemias (T-ALL) xenografted in immunocompromised mice and exposed to whole-body irradiation with FLASH or conventional dose rate radiotherapy suggest that the response of liquid tumours to FLASH radiotherapy differs⁷³. In this study, out of three patient-derived xenografts of T-ALL, two were found more responsive to FLASH radiotherapy than conventional radiotherapy, and one was found to be completely unresponsive to FLASH radiotherapy but controlled by conventional radiotherapy. This demonstrates individual variability in T-ALL and challenges the general idea of isoefficacy, and may suggest that, in T-ALL, the antitumour mechanisms triggered by FLASH and conventional radiotherapy are distinct. Although genomic profiles were not probed after irradiation, the intrinsic genetic profiles and karyotype abnormalities of the three tumour lines were found to be distinct. Gadd45B, a gene involved in the inhibition of the G2/M transition of the cell cycle, was found to be expressed less in the FLASH-resistant T-ALL. On the contrary, the FLASH-sensitive T-ALL overexpressed Gadd45B, suggesting that the redistribution of cells in the cell cycle (R2) and re-population potential (R3) might play a differential role in response to FLASH versus conventional dose rate radiotherapy. Other findings reported by Leavitt et al.⁷⁴ using subcutaneous xenografts of human glioblastoma cell line (U87-MG) did not show major genomic differences in response to FLASH and conventional radiotherapy at 24 h post-irradiation, unless acute tumour hypoxia was induced by tumour clamping. Under physiological oxygen conditions, the transcriptomic profiling showed a downregulation of cell cycle and DNA repair-related genes, whereas under hypoxic conditions (R4), FLASH radiotherapy uniquely retained antitumour efficacy associated with a high expression of Gadd45B (R2 and R3). These data suggest that FLASH exerts higher antiproliferative and enhanced cytostatic effects (R2 and R3). Moreover, these data support that FLASH can overcome hypoxia, a major factor that contributes to radioresistance (R4)⁷⁵. It is interesting, however, that the only tumour model reported to not respond to FLASH radiotherapy thus far is a liquid tumour, suggesting the importance of the tumour microenvironment and structure. This requires further investigation.

So far, the affect of FLASH radiotherapy on the tumour microenvironment has focused on the interplay between oxygen tension and hypoxia (R4) as well as the immune response (R6 and R7), whereas other potential structural differences between healthy tissues and tumours have not yet been explored. One of the most popular hypotheses put forth to explain the FLASH effect states that the ultrahigh dose rate could deplete oxygen more efficiently than conventional dose rate irradiation⁷⁶. This mechanism is called radiolytic oxygen depletion. However, measurements of oxygen tension performed in vitro^77,78 and in situ⁷⁹ as well as in subcutaneous U87-MG xenografts and in healthy skin, muscle and brain exposed to increasing doses of FLASH radiotherapy⁸⁰, have shown that the local oxygen depletion rate induced by clinically relevant doses (<10 Gy per fraction) was too low to affect tumour response, suggesting that radiolytic oxygen depletion does not account for the entirety of the FLASH effect.

Another popular hypothesis proposed to explain the FLASH effect involves enhanced immunogenicity induced by FLASH versus conventional dose rate radiotherapy, as well as the preservation of circulating immune cells (R6 and R7). Although the latter has been proposed based upon mathematical modelling⁸¹, monitoring of circulating immune cells in rats with orthotopic glioblastoma exposed to proton FLASH radiotherapy showed no substantial protection as compared with protons delivered at conventional dose rates when the same dose is used⁸². Furthermore, recent results also refute the FLASH radiotherapy-enhanced immunogenicity idea, showing similar immune responses after FLASH and conventional dose rate radiotherapy both when single and fractionated isodoses are delivered. These studies were performed in immunocompetent and immunodeficient animals, with electron and proton FLASH beams, and with subcutaneous and orthotopic tumours^82–85. They all showed no enhanced recruitment of immune cells in FLASH radiotherapy-treated animals nor a decrease in immunosuppressive signals, including TGFβ expression. These findings suggest that changes in dose rate do not differentially activate pathways that regulate immune cell recruitment in solid tumours, or that intrinsic tumour-associated immunosuppressive signals are not overridden by FLASH. On the contrary, the FLASH response in normal tissues (lung and skin) of rodents and canines has been associated with less TGFβ production^4,57,86, showing that the affect of FLASH on immunosuppressive signals is not fully understood. This question deserves further investigations as recently highlighted in a publication where a transgenic Math1-Cre;SmoM2^+/– mouse model was used and showed that proton-based FLASH radiotherapy but not proton-based conventional radiotherapy was able to sensitize medulloblastoma to GD2 chimeric antigen receptor-T cell therapy through FLASH radiotherapy-induced effects⁸⁷. FLASH radiotherapy reduced ROS production, leading to the reprogramming of lipid metabolism in macrophages, reversing the immunosuppressive polarization typically induced by radiation at conventional dose rates. The use of a transgenic model in which tumours are constitutively induced and the selection of a non-clinically relevant dose (10 Gy) limit the findings of this study, and it should also be tempered by the cognitive impairment induced by chimeric antigen receptor-T cell therapy recently reported in mice⁸⁸. Nonetheless, this demonstrates that FLASH radiotherapy can alleviate immunosuppressive signals in the tumour microenvironment, and suggest that this could be exploited to improve the efficacy of immunotherapies.

One final hypothesis that deserves more investigation is related to the effect of FLASH radiation on the vascular system. Certain structural changes have been described at the vascular level. In a Lewis lung carcinoma mouse model, FLASH radiation led to less vasculature collapse and decreased expression of phosphorylated myosin light chain, a protein involved in the control of vascular smooth muscle cells contraction, vascular tone and resistance, as well as blood pressure and tissue perfusion, compared with conventional dose rate radiotherapy⁸⁹. This preserved vascularization in tumours could promote the influx of tumoricidal immune cells; however, it was not associated with an enhanced antitumour effect. In the healthy brain, where radiotherapy at a conventional dose rate is known to induce persistent vasculature abnormalities by disrupting the cells of the blood–brain barrier by altering the tight junction proteins⁹⁰, FLASH radiotherapy was found to reduce the levels of apoptosis in the microvessels of neurogenic regions and preserve the expression of tight junction proteins such as occludin and claudin-5 (ref. 69). This microvascular protection may certainly contribute to the preservation of cognition while enabling sustained tumour control.

In summary, published preclinical data suggest that cancer and healthy tissues differentially respond to FLASH radiotherapy. Although the dose rate matters for the induction and manifestation of healthy tissue injury, all studies comparing the effect of similar doses of FLASH versus conventional dose rate radiotherapy on tumours suggest that the total dose remains the most important determinant of an antitumour effect.

Novel hypotheses for the FLASH effect

LLPs

Some organs contain postmitotic cells, such as neurons in the brain or cardiomyocytes in the heart, that can persist throughout the lifespan of an organism. The presence of LLPs, which can also persist through an organism’s lifespan^91,92, are believed to protect the functionality of these postmitotic cells from transcriptional and translational stress caused by diseases and ageing^93–95. As data indicate that LLPs are resistant to processes that normally degrade proteins^91,92, we hypothesize that they are also resistant to degradation by FLASH radiation. As the presence of LLPs is defined early in development⁹⁴ and as the majority of tumours results from the accumulation of lifelong mutations, tumours are not expected to contain LLPs. Alternatively, should tumours contain LLPs, they would be present at much lower levels than in their corresponding healthy tissue counterparts. This important distinction is fundamental to the hypothesis presented and raises the possibility that the mechanistic basis of the FLASH effect might involve the differential sensitivity of these specialized proteins to changes in dose rate, which would be inconsequential in tumours posited to be devoid of LLPs.

In the context of traditional radiobiology, the idea that cellular components other than DNA, such as proteins, constitute critical targets for radiation effects has been met with considerable scepticism over the years. As proteins are typically continuously turned over in cells, their importance in heritable changes or in persistent functional effects transpiring over an organism’s lifespan has often been overlooked or discounted. However, the differential presence of LLPs in healthy tissues versus tumours challenges this dogma. The Hetzer laboratory has used ¹⁵N isotopic labelling techniques that have identified LLPs in the brain, pancreas and endothelium^92,94,95. Interestingly, mitochondrial proteins belonging to oxidative phosphorylation and mitochondrial contact site and cristae-organizing system complexes, nuclear pores, histones and the myelin sheath were found to contain LLPs, as defined by extremely long half-lives^95,96. More specifically, they have shown that the mitochondrial proteome turns over more slowly than the cellular proteome, where 40% of mitochondrial proteins are LLPs that exhibit longer lifespans than the average cellular protein⁹³. Mitochondrial LLPs are core components of the cristae, where they are thought to support their structure and function^93,97. They are also part of the electron transport chain, specifically the membrane arm of Complex I and the Complex III dimer⁹³. Notably, the majority of mitochondrial LLPs in the membrane arm of Complex I display increased stability in super complexes, which counteracts age-associated oxidative degradation to mitochondrial functionality⁹³.

Given the high energy demands and postmitotic composition of the brain, it is not surprising that this organ has evolved to contain a larger proportion of LLPs than other more rapidly proliferating tissues^92,93. As the brain exhibits robust FLASH sparing (DMF ~1.4) for many functional end points (for example, cognition or long-term potentiation), this further supports that FLASH sparing is related to LLP content, driven in part by their inherent resistance to degradation at FLASH radiation compared with conventional dose rate radiotherapy (Supplementary Fig 1). This hypothesis is also consistent with a more pronounced FLASH effect observed in late-responding versus early responding tissues^7,98. Moreover, the presence of LLPs in nearly every normal tissue spared upon FLASH radiation⁸, and their possible absence in tumours that are FLASH sensitive, further supports this hypothesis.

The inability to manipulate LLPs experimentally confounds approaches that could directly test their functional roles in healthy tissues. Interestingly, a low ^15/14N isotopic ratio was proposed as a diagnostic marker in microbiopsies to distinguish tumours from healthy tissues⁹⁹. This suggests that similar ^15/14N isotopic labelling techniques could be used in early and late-arising spontaneous tumour models to more formally test this diagnostic approach. Such studies could also validate the proposed absence of LLPs in tumours. Clearly, longer term studies are required to overcome inherent limitations in the evaluation of LLPs and would be necessary to accurately measure the dose rate dependence of LLP decay, and test whether certain post-translational modifications on LLPs contribute to their protective effects in healthy tissues after FLASH radiation.

The FLASH effect may also, in part, be explained by changes in protein homeostasis such as through changes in autophagy and/or targeted lysosomal degradation post radiation exposure. Many of these ideas provide testable hypotheses. For example, in this case, FLASH radiation would be expected to preserve the half-life of LLPs whereas conventional dose rate radiotherapy will hasten their degradation, causing a more rapid decay of ¹⁵N-labelled proteins. Similarly, we might expect to find increased yields of LLPs targeted for lysosomal degradation after conventional dose-rate radiotherapy as compared with FLASH radiotherapy, again indicating their resistance to turnover after exposure to ultrahigh dose-rate irradiation. Perhaps the most interesting way to test our hypothesis would be to investigate the liver as, so far, this remains the one healthy tissue where the response to FLASH radiation has not been published, and is the only tissue reported to lack LLPs^94,96. Although we believe LLPs provide a plausible mechanism for the FLASH effect, this will require further experimentation to substantiate.

Biophysical properties of the tissue

Healthy and tumour tissues vary in their respective structures, where healthy tissues are generally well organized, with single epithelial cell layers on a well-defined and relatively elastic basement membrane, and tumour tissues, especially solid tumours, lack cellular organization and a basement membrane, and exhibit a denser, stiffer and disorganized extracellular matrix (ECM)^100,101. As organ systems have evolved to provide optimal homeostatic control over physiological processes, it is interesting to consider whether and how structural changes in tissues might relate to the FLASH effect. It is plausible the intrinsic properties of the targeted tissue such as a dense ECM can influence how tissues respond to FLASH versus conventional dose-rate radiotherapy via the regulation of biomechanical forces such as pressure, shear stress and deformability¹⁰¹. If we examine how ionizing radiation interacts with tissues, most of its energy is converted to heat, leading to an increase in the local temperature of the irradiated tissue. Measurements have shown that the temperature can rise by several thousand Kelvin in the vicinity of the proton or heavy ion track¹⁰². This results in thermal expansion within the tissue and acoustic emission. Interestingly, acoustic measurements have been proposed as dosimetric tools and have been used to successfully measure FLASH radiation doses¹⁰³. However, these acoustic emissions also create a pressure wavefront known to vary linearly with the instantaneous dose rate and the properties of the tissue, especially its density. This concept was developed by Lascaud and Parodi who performed an in silico analysis, suggesting that thermoacoustic emissions produced by FLASH radiation could have an amplitude sufficient to induce acoustic cavitation, acoustic oscillation and possible collapse of gas bubbles that could severely damage tissues and organs¹⁰⁴. In a clinical study of domestic cats treated with FLASH radiotherapy for squamous carcinoma of the nasal planum, they confirmed this concept as the cats developed osteoradionecrosis¹⁰⁵. Their work showed that the tissue heterogeneities within the nasal cavity, such as air cavities present in sinuses, oral cavity and maxillary and nasal bones, could induce a peak of pressure with an amplitude higher than the initial or pre-existing pressure, which could then induce acoustic resonance and tissue collapse¹⁰⁶. Solid tumours, which are typically stiffer, may be more affected by this process. Another modelling study suggested that the density of particles (electron and proton) produced in tissues upon irradiation at ultrahigh dose rates would decrease the reactivity of the resultant free radicals¹⁰⁷. In well-organized healthy tissues, these dense packs of particles delivered with FLASH radiotherapy could create a high level of free radicals locally that enhance free radical recombination probability owing to their proximity, and this recombination would quench the free radical toxicity, preserving the tissue. However, in tumours, the high stiffness of the tissue lowers the diffusion capability of free radicals, thus reducing the recombination process and sustaining toxic effects. Although these hypotheses are interesting, they are difficult to validate functionally in vivo. One option is to use tissues with dense and/or disorganized ECM, such as desmoplastic tumours or tissues with dense matrices. In this case, FLASH radiation is expected to elicit greater biomechanical constraints and thus increase the severity of FLASH radiation-induced damage. According to this concept, desmoplastic or stroma-rich tumours, such as sarcomas and pancreatic cancer, would be more sensitive to FLASH irradiation than tumours in organs with loose and elastic connective tissue.

Examining the ECM composition in the healthy brain, lung and gut, tissues that, despite their functional, structural and radiobiological diversity, have all shown resistance to FLASH radiotherapy, may offer insight into which ECM components provide healthy tissues with protection. Interestingly, fibrillar collagens, which increase tissue stiffness, are low in the brain, whereas components that promote tissue elasticity, such as glycosaminoglycans, elastin, fibronectin, laminin, tenascin-C and matrix metalloproteinases, are common components of the brain¹⁰⁸, lung¹⁰⁹ and gut¹¹⁰ ECM. This could decrease FLASH radiotherapy-induced biomechanical forces.

Dense, fibrous and stiff stroma are known to influence cellular responses to anticancer treatments, including radiation, by altering mechanical signalling pathways and cellular adhesion¹¹¹. They can also sequester stress-associated molecules, including ROS, cytokines and damage-associated molecular patterns, which can indirectly impair tumour oxygenation and radiation response¹¹². This could, therefore, enhance FLASH radiation-induced mechanical constraint and tumour cell killing. Accordingly, many FLASH radiotherapy studies have been successfully executed in subcutaneous tumour grafts, which are routinely surrounded by a dense fibrotic cap^{4,74,83,84,113,114}, as well as in models of pancreatic ductal adenocarcinoma, a desmoplastic and dense tumour²¹. Notably, the only FLASH-resistant tumour described thus far is T-ALL⁷³, which is a non-stromal lymphoid tumour¹¹⁵. Together, these data suggest that elevated tumour stiffness and a dense stroma enhance tumour sensitivity to FLASH radiation Accordingly, as healthy tissues are relatively elastic, all healthy organs, except mature bone, could be spared from FLASH radiotherapy-induced toxicity and cell death.

Conclusions and future perspectives

The functional and preclinical evidence supports the use of FLASH radiotherapy in cancer; however, the key factors that make healthy tissues resistant to ultrafast radiation delivery but vulnerable when delivered at conventional dose rates remain unknown. Although we provide ideas in this Perspective article proposing a role for structural and biomechanical components of tissues such as LLPs and ECM stiffness, these aspects remain to be formally investigated or demonstrated with complex biochemical and biophysical approaches. A related and equally important question pertains to the temporal dynamics involved in the interaction of energetic electrons, photons and charged particles within tissues. These investigations require multidisciplinary teams working at the crossroads of physics, chemistry and biology to understand whether the healthy tissue-sparing effect of FLASH or its antitumour effect reach a plateau at higher dose rates (that is, at mega Gray per second or above). This is necessary to define the optimal temporal conditions to be used in the clinic and maximize FLASH radiotherapy benefits in patients with cancer. Advancements in particle accelerators will be essential to ensure global clinical application of FLASH in the future, as the available technology is currently limited (Box 1). Moreover, producing radiation at ultrahigh dose rates requires high power machines that are large, heavy and expensive, and thus would require the construction of specialized facilities⁸. As of today, FLASH radiotherapy has opened new possibilities to enhance radiotherapy safety and efficacy. Novel technologies that are compatible with spatial clinical constraints are now emerging^116,117. Although patient recruitment may be challenging given the highly competitive landscape of oncology clinical trials, enrolling patients who may benefit from organ-preserving irradiation or those eligible for safe re-irradiation in the recurrent setting is likely to be more achievable. Ultimately, a comprehensive elucidation of the biological mechanisms and physics parameters underlying the FLASH effect will be critical for overcoming the current barriers to its clinical implementation. This will facilitate optimal patient selection and combination strategies to enable effective clinical translation, contributing to improved therapeutic efficacy, reduced healthy tissue toxicity and enhanced patient quality of life.

Supplementary Material

Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41568-025-00878-9.

Glossary

Acoustic cavitation

The formation, growth and implosive collapse of bubbles in a liquid due to pressure fluctuations induced by sound waves, typically ultrasound.

Acoustic oscillation

The periodic variation in pressure, particle velocity and other acoustic quantities in a medium as a sound wave propagates through it.

Desmoplastic tumour

A tumour that develops fibrous connective tissue and adhesions that increase stiffness.

DNA damage response

A series of cellular signalling events following DNA damage to promote repair.

Dosimetric tools

Devices and methods used to measure and quantify the amount of radiation delivered to a specific area, ensuring accuracy and safety in radiotherapy and other radiation applications.

Extracellular matrix

A cellular stromal components composed of macromolecules (that is, proteins, carbohydrates and lipids) and minerals, which provide structural and biochemical support to cells.

Fibrosis

Progressive and chronic alteration of a tissue due to pathological wound healing characterized by the replacement of a normal parenchymal tissue by connective tissue.

Free radical recombination

A chemical reaction in which two free radicals combine to form a stable, non-radical molecule.

Ionizing radiation

Particles or electromagnetic waves with sufficient energy to induce the ionization of atoms it interacts with, and in the context of external beam radiotherapy, these are often X-rays, protons, electrons or carbon ions.

Isodose

An equivalent dose given by two different irradiation modalities.

Proteostasis

The biological processes involved in the synthesis, folding, trafficking and degradation of proteins required for the proper maintenance of cellular functions.

Radiation chemistry

The study of the chemical effects and reactions induced by the absorption of ionizing radiation in matter.

Radiobiology

A branch of science studying the interactions of ionizing radiation with biological tissues and organisms.

Radiolytic oxygen depletion

The consumption of oxygen by radiation-induced chemical reactions in aqueous systems, producing ROS and reducing overall oxygen level.

Radioresistance

Intrinsic or adaptive capacity of cells, tissues, organs or organisms to overcome the detrimental effects of ionizing radiation.

Therapeutic index

The measurement of treatment efficacy relative to its side effects.

Tumour clamping

An experimental procedure used to restrict blood flow to a tumour, inducing hypoxic or ischaemic conditions.