Clonogenic assays

In radiobiological research, the clonogenic assay has been a reference method for in vitro studies since Puck & Markus introduced it in 1956 (Ref. Reference Puck and Marcus36). Radiation-induced cell death can occur in several ways, including mitotic catastrophe, apoptosis, necrosis, senescence, autophagy and ferroptosis (Ref. Reference Sia37). The advantage of the clonogenic assay is its ability to capture cancer cells ability (or inability) of ‘endless’ division, i.e. including all kinds of cell deaths. Hence, the assay provides an in vitro-surrogate for the complete sterilisation of tumour cells in vivo. Although the assay may have limitations, such as being dependent on cell densities, it should be a reliable method for head-to-head comparisons in dose rate studies, provided all other experimental conditions are equal (Refs Reference Brix38–Reference Brix40).

Interestingly, both early (from the 60's) and recent (here defined as in the 21st century) studies of ultra-high dose rate irradiations in mammalian cells using clonogenic assays have provided inconsistent results. In the early days of ultra-high dose rate irradiation, a ‘hockey-stick’-shaped survival curve for ultra-high dose rate irradiated samples was described. Typically, CONV and FLASH-curves were indistinguishable at lower doses, and then separated at higher doses, i.e. an increased survival fraction after FLASH irradiation was noted at higher irradiation doses (Refs Reference Berry41, Reference Town42). Such survival curves were initially described for normoxic conditions (ambient air, 21%), though other experiments could not reproduce the findings (Refs Reference Berry and Stedeford43, Reference Nias44). Attention was then turned to the dependence of oxygen. It was found that the ‘break’ of the survival curve was influenced by the oxygen tension (Refs Reference Berry and Stedeford43–Reference Michaels46). Increasing the level of hypoxia in cells resulted in a lower total dose required to ‘break’ the survival curve. In addition to these results, similarly shaped survival curves were also described when irradiating bacteria at ultra-high dose rates, with breaks at doses around 60–70 Gy in normoxic conditions (Refs Reference Dewey and Boag47, Reference Epp, Weiss and Santomasso48). During the 90's, two independent studies could not detect any survival difference after FLASH and CONV irradiation in neither normoxic nor anoxic conditions (Refs Reference Cygler49, Reference Zackrisson, Nyström and Ostbergh50).

Recent investigations using clonogenic assays in normoxic conditions have also provided inconsistent results. Despite the fact that the FLASH sparing in vivo is found in the healthy tissue, few of the recent studies have used normal cell lines. Instead, most have investigated the potential sparing of cancer cell lines. An increased survival fraction was found for H454 murine glioblastoma cells after 20 Gy FLASH compared with CONV irradiation in normoxia (Ref. Reference Montay-Gruel8). Congruently, an increased survival fraction was found for 4/7 cell lines after FLASH compared with CONV in our recent investigations (Ref. Reference Adrian51). On the other hand, Venkatesulu et al. found opposing FLASH effects, hence a lower survival fraction after FLASH for two murine pancreas cancer cell lines in normoxia (Ref. Reference Venkatesulu52). Other studies could not distinguish any differences in survival fraction in normoxia for IMR90 normal human lung fibroblasts, DU145 prostate cancer cells, or A549 lung cancer cells (Refs Reference Adrian19, Reference Buonanno, Grilj and Brenner20, Reference Khan22). In hypoxia, we have previously described an oxygen-dependent FLASH sparing for DU145 prostate cancer cells (Ref. Reference Adrian19). Similarly, A549 lung cancer cells irradiated as spheroids with naturally occurring hypoxic cores exhibited a FLASH sparing, in contrast to the results for cells irradiated as a normoxic monolayer (Ref. Reference Khan22). Noteworthy, in the early publications, the dose required to break the survival curve under normoxia was found to be at 7–10 Gy with a distinct inflexion point (Refs Reference Berry41, Reference Town42). Such inflexion points have since only been described for cells in hypoxia and/or for much higher (non-clinical) doses and seem to indicate the dose required to consume all available oxygen (Ref. Reference Michaels46). Later studies have instead successfully used the linear-quadratic (LQ) model to fit the data (Ref. Reference McMahon53).

The reasons for the inconsistent findings of FLASH versus CONV using clonogenic assays could be several (Table 1). Firstly, we would not expect to see significant oxygen depletion effects for FLASH versus CONV in fully anoxic cell cultures as there is no oxygen available to deplete, nor in normoxic cultures as there is too much oxygen to deplete for clinical doses to have a meaningful impact on the available oxygen. Furthermore, beam characteristics differ between laboratories. Single-pulse versus pulsed delivery, instantaneous dose rate, dose per pulse, pulse repetition frequency, average dose rate, total delivery time and type of irradiation (electron/photon versus ion) could possibly all affect the radiobiological response (Refs Reference Bourhis5, Reference Wilson16). Experimental conditions vary between laboratories. Temperature during irradiation, time outside the incubator, type of cell medium used, the volume of cell medium per flask or dish might influence results (Ref. Reference Potter54). The definition of survival is stated to be a single cell that has proliferated to form a colony of at least 50 cells (Ref. Reference Puck and Marcus36). Laboratories may use different approaches to determine the colony size, i.e. to decide which clones are to be counted as survivors. It has been shown that different clone-size cut-offs influences the results in clonogenic assays (Refs Reference Joshi55–Reference Yohem, Bregman and Meyskens57).

Table 1. Published Clonogenic FLASH data using mammalian cells, separated by electron and other type of irradiation

Challenges also arise from performing the clonogenic assay with irradiation at higher total doses. Depending on the cell line's plating efficiency and radiation sensitivity it may not be possible to achieve more than a few surviving colonies. Thereby, statistical uncertainties arise. In addition, the high inoculation density may cause difficulties to evaluate the samples. Together, clonogenic assays performed at higher dose ranges may be less reliable. The number of plated cells per flask or dish, the cell density, also affects the radiobiological response (Refs Reference Brix38–Reference Brix40). In addition, the clonogenic assay can be performed in two principally different ways; pre-plating or post-plating (Ref. Reference Franken64). In the pre-plating method, cells are plated in appropriate densities in individual flasks, dishes or wells, and each sample is irradiated. The cells are then incubated, without re-trypsinisation, to grow and form colonies. In the post-plating methods, a densely seeded flask is irradiated, trypsinised, cells are counted, plated in appropriate cell densities in dishes and allowed to grow and form colonies. Depending on the method used, different radiation response has been reported (Ref. Reference Kriegs65). To investigate if plating methods could influence some of the reported inconsistent FLASH findings, we have performed pre- and post-plating experiments for the melanoma cell line MM576. Our current results suggest that both plating methods may detect a FLASH sparing, although the magnitude of the sparing and the power to detect statistical differences could differ (Fig. 2). Our lab has investigated clonogenic survival in normoxia after FLASH and CONV for nine human cell lines (including the MM576 in this publication) (Refs Reference Adrian19, Reference Adrian51). Five of the nine cell lines were found to have significantly increased survival fractions after FLASH irradiation. Although it is possible that statistical uncertainty caused the observed discrepancy between cell lines, it is also possible that a difference in biological factors affected the FLASH response. It has been found that the response to FLASH irradiation in vivo is cell line specific (Ref. Reference Chabi66), which is likely to also be the case in vitro. Further studies comparing different experimental set-ups could reveal additional phenomenological insights of the radiation response. For instance, by varying the time before re-plating after irradiation (delayed plating), the effect of ‘potentially lethal damage’ could be assessed (Ref Reference Liu67).

Fig. 2. Clonogenic survival for the melanoma cell line MM576 after irradiation with 9 Gy of CONV (red, measured delivered doses 9.2–9.4 Gy) or FLASH (blue, measured delivered doses 9.4–9.6 Gy) using the pre-plating (left panel) or post-plating method (right panel). The box and whisker plots illustrate median (grey line), interquartile range (box), the lowest/highest observation within ± 1.5 × interquartile range from the box (whiskers), and individual flasks as black dots. Irradiation was performed using a modified clinical accelerator (Ref. Reference Hall and Brenner28) with beam characteristics and the experimental pre-plating protocol as previously described (Ref. Reference Zhang14). For the post-plating, 500 000 cells were plated in T12.5 flasks the day before irradiation, and 1 h after irradiation the cells were trypsinised, counted, re-plated in appropriate densities and then incubated and evaluated as the pre-plating flasks. Statistical comparisons were made in RStudio (version 1.2.5042) using the unpaired Student's T-test after testing for normality using the Shapiro-Wilk's test. ‘Ratio FLASH/CONV’ was calculated as the mean (Survival Fraction_FLASH) divided by the mean (Survival Fraction_CONV). Data from three independent experiments with triplicate-sextuplicate flasks per condition.

Despite potential pitfalls with the clonogenic assay, it is still considered as the ‘gold standard’ and will undoubtedly help verify important findings in future FLASH studies. If care is taken to ascertain identical experimental conditions (e.g. time outside the incubator, cell medium volume, cell density and well-matched CONV and FLASH beam characteristics and irradiation doses), the assay will provide robust and reproducible data. For instance, the importance of beam line characteristics for the FLASH effect, such as pulse repetition frequency, dose per pulse, total delivery time, and average dose rate can be investigated in large-scale experiments. Detailed studies of the FLASH-response in physoxic and/or hypoxic/anoxic conditions (physoxia: oxygen levels of 3–7%, hypoxia: oxygen level ⩽2%, anoxia: 0% oxygen (Ref. Reference McKeown68)) are possible using the clonogenic assay, if care is taken to control the oxygen tension at the time of irradiation. Confirmatory studies in labs with different beam lines, using the same cell line, will be important. Besides varying oxygen concentration, the addition of scavengers such as superoxide dismutase might provide indirect mechanistic evidence for the underlying mechanism of the FLASH effect.

The current data suggest that the clonogenic assay can resolve at least part of the FLASH effect. Due to the clonogenic assay's central role in radiobiological research and the great advantage of large-scale experiments without involving living mammals, the continued use of clonogenic assays to investigate the FLASH effect will be important in our search for the underlying mechanisms.

Antibody-based assays

Radiation changes the physical and chemical environment of the cell, which leads to changes in protein phosphorylation, enzymatic activity, localisation, and the formation of protein-protein complexes that participate in cell cycle arrest, apoptosis, and DNA repair (Refs Reference Levine and Oren69–Reference Prise72). Ionising radiation can induce DNA damage by direct interaction with the DNA or indirectly through the production of free radicals, which can react with the DNA. This causes a variety of DNA lesion types including base damage, single-strand breaks (SSBs) and double-strand breaks (DSBs) (Refs Reference Alloni73, Reference Alloni74). The generation of these DNA lesions, particularly DSBs, triggers sophisticated and highly regulated DNA damage response and repair (DDR) pathways (Ref. Reference Santivasi and Xia75) that can easily be assessed in vitro by antibody-based techniques such as western blotting (WB), enzyme-linked immunosorbent assay (ELISA), flow cytometry, immunocytochemistry (ICC), or immunohistochemistry (IHC).

Both IHC and ICC provide spatial information such as distribution or localisation of specific cellular or subcellular components. IHC is generally applied on biological tissues from in vivo samples either embedded by paraffin or frozen to maintain the morphology, while ICC is generally applied on cells. They are especially useful in studying the DNA damage foci, which are subnuclear foci formed by DDR proteins near the DNA lesion site. IHC has provided some of the recent evidence identifying an in vivo FLASH effect. IHC staining for immune, apoptotic, and DDR markers showed compared to CONV irradiation, FLASH irradiation causes less apoptosis, DNA damage, and immune response in guts (Ref. Reference Levy12), lung (Refs Reference Favaudon1, Reference Fouillade7), and brain (Refs Reference Montay-Gruel8, Reference Simmons9, Reference Alaghband15). WB refers to the transfer of biological samples (mixture of proteins from cells or tissue samples) from a gel to a membrane and their subsequent detection on the surface of the membrane. It allows for the semi-quantification of a protein of interest amid a complex protein mixture, which can be achieved by ELISA, involving the use of standard curves with known protein concentrations. While WB detects the protein of interest from a mixture of cells, flow cytometry enables the study of intracellular and cell surface proteins at single-cell resolution.

The phosphorylation of the histone H2AX (γH2AX) is one of the most well-characterised DNA DSB markers. It generally occurs within minutes at the DSB sites in the nucleus and shows a maximum number of foci around 30 minutes post-irradiation (Ref. Reference Mariotti76). Therefore, γH2AX can be used as a marker for exploring the spatial and temporal dynamics of DNA repair in cells following irradiation (Refs Reference Costes77–Reference Schmid80). In addition, γH2AX has been suggested as a prognostic biomarker to predict the radiotherapy response of patients (Ref. Reference Ivashkevich81). Another classic DDR marker is 53BP1, which becomes hyperphosphorylated and colocalises with γH2AX near the site of DNA DSBs (Ref. Reference Ward82).

WB, ICC, and IHC have previously been used to assess DNA damage level after FLASH or CONV irradiation. Using IHC, acute apoptosis was quantified with caspase-3 cleavage and TUNEL labelling in histological sections of irradiated lungs (Ref. Reference Favaudon1), brain (Ref. Reference Allen83) and intestines (Ref. Reference Levy12). Using flow cytometry, Ehlert et al. showed increased γH2AX signal intensity in Jurkat and Ramos cells with increased radiation dose, using laser-accelerated protons at a dose rate of 10⁸ Gy/s (Ref. Reference Ehlert84). Fouillade et al. (Ref. Reference Fouillade7) investigated γH2AX and recruitment of 53BP1 at sites of DNA damage by immunofluorescence microscopy in two normal lung fibroblast cell lines (MRC5 and IMR90) and A549 lung cancer cells. Though no difference in γH2AX foci per nucleus was found among the cell lines between CONV and FLASH irradiation, less 53BP1 was observed in the two normal lung fibroblasts after FLASH irradiation but not in the cancer cell line. However, Adrian et al. (Ref. Reference Adrian51) found no significant difference in 53BP1-foci number after FLASH and CONV irradiation for three studied cell lines.

Comet assay

Another method that has been implemented in many radiobiology studies to measure DNA damage and repair is the comet assay (Refs Reference Olive85, Reference Zhang86). It was first introduced by Ostling and Johanson using neutral lysis (Ref. Reference Ostling and Johanson87) in 1984 and then modified by Singh et al. to an alkaline version to increase the sensitivity (Ref. Reference Singh88). The alkaline comet assay is an inexpensive, time efficient, highly sensitive method for assessing DNA damage formation and repair at the level of single cells (Refs Reference Collins89, Reference Azqueta90). In its simplest form, it enables the detection of SSBs, DSBs, alkali-labile sites (ALS), as well as SSB sites associated with incomplete DNA excision repair (Ref. Reference Tice91). The comet assay can be further extended to detect DNA-DNA/DNA-protein cross-links, adding to the versatility in its applications (Refs Reference Pfuhler and Wolf92–Reference Fikrova96).

In the alkaline comet assay, cells embedded in low melting point agarose on microscope slides are lysed in a solution containing high salt and detergents to remove cell membranes, cytoplasm, nucleoplasm and histones leaving nucleoids of supercoiled DNA linked to the nuclear matrix (Ref. Reference Azqueta and Collins97). Relaxed in alkaline buffer, broken negatively charged DNA is free to migrate towards the anode upon electrophoresis. These stretched nucleoids resemble comets upon staining with an intercalating DNA dye which may be visualised via fluorescence microscopy (Ref. Reference Collins89). The intensity of the comet tail relative to the amount of DNA residing in the ‘head’ indicates the level of induced strand breaks; with a reduction in DNA migration indicative of DNA crosslinks (Ref. Reference Tice91). Furthermore, the addition of lesion-specific bacterial repair enzymes post-lysis enables the detection of a variety of DNA lesions other than the standard SSBs, DSBs, and ALSs (Refs Reference Ding98–Reference Muruzabal, Collins and Azqueta100). Therefore, the assay is useful in the profiling of DNA damage following different delivery modalities of ionising radiation, for instance different dose rates or beam particles.

The comet assay may also be used as a diagnostic, prognostic, and predictive biomarker in oncology (Refs Reference McKenna, McKeown and McKelvey-Martin101, Reference Vodicka102). It has been used as a tool for predicting an individual's tumour sensitivity to radiation with and without chemotherapeutic regimes (Refs Reference Olive85, Reference Moneef103–Reference Wood105). A key feature of the assay is that only a small cell sample is required, making it possible to analyse cells from biopsy prior to and post treatment (Refs Reference McKenna, McKeown and McKelvey-Martin101, Reference Bowman104), or from lymphocyte cells obtained via finger prick (Refs Reference Garcia and Mandina106, Reference Al-Salmani107). Historically, the assay has been used to identify the hypoxic fraction of tumours in mice and humans obtained from fine-needle aspiration (Refs Reference Olive108, Reference Shibuya109). This method also allows for monitoring changes to the hypoxic fraction of cells during a course of fractionated RT (Ref. Reference Olive108).

The comet assay has recently been used by us to assess the difference in DNA damage formation of human peripheral blood lymphocytes (PBL) irradiated with 6 MeV electrons at FLASH (2 kGy/s) or CONV (0.1 Gy/s) dose rates, under a variety of oxygen concentrations. This was achieved by incubating cells embedded in low melting point agarose gels mounted on glass slides for 2 h in a humidified hypoxia chamber, prior to irradiation. PBL were used as a representative body-wide systemic normal tissue susceptible to irradiation, to assess DNA damage formation following FLASH or CONV irradiations (Refs Reference Louagie110–Reference Heylmann112). We found that the difference in DNA damage was modulated by the oxygen concentration, with a maximum difference of 30–40% seen at 0.25–0.5% oxygen tension. We also utilised the method to show how dose rate and total dose modulated the ex vivo DNA damage sparing effect observed between FLASH and CONV irradiation at a low (0.5%) oxygen tension, with significant sparing observed at average dose rates ⩾ 30 Gy/s and total doses ⩾ 20 Gy (Ref. Reference Cooper113).

Genomic approach

We and other groups have used RNA sequencing to identify potential novel biomarkers relevant to the differential RT response between FLASH and CONV irradiation (Refs Reference Fouillade7, Reference Velalopoulou114, Reference Ruan115). In addition, functional genomic screening using techniques like CRISPR (Ref. Reference He116) or RNA interference (Ref. Reference Mullenders and Bernards117), which provides a large-scale genetic loss-of-function experimental approach, can be used to systematically identify the genes responsible for the FLASH effect. However, such methods are generally very time-consuming, resource demanding, and sometimes challenging to perform as they require high standards in quality control to give robust and reliable results, i.e. it requires identical irradiations (FLASH and CONV) of numerous identically prepared samples. In addition, the results will also need to be carefully validated in subsequent experiments using assays such as the ones described above.

Assessing oxygen content, free radicals and oxidative stress

It has been observed by us and several other groups that the FLASH effect is oxygen dependent (Refs Reference Montay-Gruel8, Reference Adrian19, Reference Hendry29, Reference Tessonnier63, Reference Petersson118). Therefore, being able to measure and control the oxygen concentration is of importance when conducting FLASH experiments. Techniques to measure oxygen tension in tissues in vivo have been previously reviewed (Ref. Reference Springett and Swartz119), some of which may be applied in vitro. Due to the nature of oxygen depletion during FLASH irradiation, the oxygen measurement technique of choice should ideally have a temporal resolution in the order of milliseconds. Quenching of fluorescent or phosphorescent dyes by oxygen is a common technique that has been applied to measure the oxygen consumption during FLASH irradiation. We and other groups have used fibre optic probes coated with oxygen-sensitive fluorescent compounds to measure the local oxygen tension in vitro during FLASH irradiation (Refs Reference Moon, Petersson and Olcina31, Reference Jansen120). Using a water-soluble molecular probe Oxyphor 2P, Cao et al. measured and compared the oxygen consumption during FLASH and CONV irradiation, both in vitro and in vivo (Ref. Reference Cao121). Other oxygen measurement techniques could be used to study the in vitro oxygen dynamic in FLASH irradiation, e.g. cellular tracking of oxygen concentration can be achieved using soluble oxygen probes and fluorescence/phosphorescence lifetime imaging microscopy (Refs Reference Kurokawa122, Reference Penjweini123).

It has been hypothesised that the differential tissue response between FLASH and CONV irradiation is due to difference in damage from radiation-induced free radicals, caused by the higher temporal concentration of radicals produced for FLASH leading to an increase in radical-radical interactions and consequently less indirect damage of DNA (Fig. 1) (Refs Reference Labarbe32, Reference Favaudon, Labarbe and Limoli124). Therefore, detection of free radicals in cells and tissues will be important to decipher if this is the mechanism responsible for the FLASH effect. A comprehensive review on free radical detection has previously been published by Halliwell and Whiteman (Ref. Reference Halliwell and Whiteman125). Direct measurement of reactive oxygen species (ROS, a subset of free radicals containing oxygen) in cells is generally achieved by fluorogenic probes (Ref. Reference Bai126). In addition, cells genetically encoded with fluorescent protein-based redox sensors can be used to capture the ROS dynamics within cells, in real-time (Ref. Reference Bai126). Electron paramagnetic resonance (EPR) is the only technique that can detect specific free radicals directly, and it can also be used to measure oxygen concentration (Ref. Reference Khan127). However, many of the free radicals generated in vitro or in vivo have very short half-life times and will require the use of ‘spin-traps’, chemicals that can form stable radicals, to enable detection by EPR. Incorporation of antibody-based technology with EPR, the ‘immuno-spin trapping,’ can detect DNA radicals with high sensitivity and subcellular information (Ref. Reference Mason128). These techniques allow for spatiotemporal detection of free radicals following in vitro FLASH and CONV irradiation.

Compared to detecting free radicals, the oxidative stress, the damage of cells and tissues caused by ROS, is less technically challenging to detect since the markers are generally more stable. Oxidative stress can be assessed by DNA damage, lipid peroxidation, and protein damage using antibody-based or chromatographic assays (Ref. Reference Katerji, Filippova and Duerksen-Hughes129). In addition, fluorogenic lipophilic probes have been developed to study the spatiotemporal information of lipid metabolism (Ref. Reference Greene, Lincoln and Cosa130).

Novel in vitro assays

The radiobiological assays used for studying the FLASH effect have so far mostly been conducted with 2-dimensional (2D) cell cultures. Recent advancement in 3-dimensional (3D) cultures has offered more ways to model/mimic physiological conditions. Therefore, these technologies may be useful to create second-order events such as oxygen gradient and immune response, which could enhance the FLASH effect. The characteristics of each technology and the potential second-order events each can model to enhance the FLASH effect are summarised in Table 2.

Table 2. Novel cell culture technology and the potential secondary events to enhance the FLASH effect

Concluding remarks/translational implications

The FLASH effect is an intriguing phenomenon that is currently being studied by many research labs around the world. The magnitude of the effect is such that it could be a ‘game changer’ for the treatment of many tumours. However, as we do not yet understand the effect or know the mechanisms behind the effect, we cannot translate it into the clinic in an optimal way. Consequently, more preclinical studies are needed. Specifically, in vitro studies to complement the more challenging and expensive in vivo studies. Here, we have described several in vitro assays that have been used or could be used to finally elucidate the mechanisms behind the FLASH effect. The doses often required to show a significant FLASH effect (⩾10 Gy) make some of the mentioned assays more suitable than others, e.g. the comet assay compared to γH2AX.

FLASH-RT is a very promising new radiotherapy technique that we would like our cancer patients to benefit from as soon as possible. However, a successful clinical translation of the technique hinges on a better understanding of the FLASH sparing effect of normal tissues. If we understand this phenomenon, we can optimise our FLASH treatments to maximise the effect, e.g. by using hypofractionated approaches (Ref. Reference Montay-Gruel17), non-homogeneous irradiation (Ref. Reference Griffin162), and combination with drugs that enhances the effect (Ref. Reference Zhang163). Preclinical assays, such as the ones described above, will be essential tools in identifying the radiobiological mechanisms behind the effect and finally being able to fully exploit the benefits associated with FLASH-RT.