Quantification of radiation-induced DNA double strand break repair foci to evaluate and predict biological responses to ionizing radiation

Abstract

Radiation-induced foci (RIF) are nuclear puncta visualized by immunostaining of proteins that regulate DNA double-strand break (DSB) repair after exposure to ionizing radiation. RIF are a standard metric for measuring DSB formation and repair in clinical, environmental and space radiobiology. The time course and dose dependence of their formation has great potential to predict in vivo responses to ionizing radiation, predisposition to cancer and probability of adverse reactions to radiotherapy. However, increasing complexity of experimentally and therapeutically setups (charged particle, FLASH …) is associated with several confounding factors that must be taken into account when interpreting RIF values. In this review, we discuss the spatiotemporal characteristics of RIF development after irradiation, addressing the common confounding factors, including cell proliferation and foci merging. We also describe the relevant endpoints and mathematical models that enable accurate biological interpretation of RIF formation and resolution. Finally, we discuss the use of RIF as a biomarker for quantification and prediction of in vivo radiation responses, including important caveats relating to the choice of the biological endpoint and the detection method. This review intends to help scientific community design radiobiology experiments using RIF as a key metric and to provide suggestions for their biological interpretation.

DNA DOUBLE STRAND BREAKS INDUCED BY IONIZING RADIATION

DNA double strand breaks (DSB) are recognized as the most deleterious form of DNA damage in radiation biology because they cause the loss of genetic information in the form of deletions and/or translocations, which results in genomic instability that can lead to cell death, accumulation of mutations and cancer development (1). DSBs can be induced by exposure to exogenous agents, including ionizing radiation, anticancer drugs and chemicals, but they also occur during normal cellular metabolism, especially in response to oxidative stress and during DNA replication.

MOLECULAR RESPONSE TO DSB

To cope with continuous generation of DNA lesions, cells have evolved checkpoints to inspect genome for damage as well as repair mechanisms to maintain genome integrity. The repair processes, collectively known as DNA damage response (DDR), are highly interconnected signaling pathways consisting of hundreds of proteins that interact with each other in close coordination with cell cycle progression and chromatin remodelling (2–4). Three main DSB repair pathways are described below and illustrated in Figure 1A. Notably, in each pathway, DSB repair begins by sensor proteins detecting the damage site and recruiting transducer proteins, which then provide the signals to effectors.

Figure 1.

(A) General overview of the three main DNA double-strand break repair pathways in eukaryotic cells. Alt-EJ, Alternative –EJ; ATM, Ataxia Telangiectasia Mutated; BRCA1, Breast cancer type 1 susceptibility protein; c-NHEJ, Canonical non-homologous end joining; CtIP, CtBP-Interacting protein; DNA-PKcs, DNA-dependent protein kinase catalytic subunit (DNA-PKcs); HR, Homologous recombination; Lig 1, DNA ligase I; Lig4, DNA ligase IV; MRN, Mre11-Rad50-Nbs1; PALB2, Partner and localizer of BRCA2; PARP1, Poly (ADP-ribose) polymerase 1; Pol δ, DNA polymerase δ; Pol θ, DNA polymerase θ; RPA, replication protein A; XLF, XRCC4-like factor; XRCC4, X-ray repair cross-complementing protein 4. (B) Timeline of the repair pathways. The irradiation time is indicated by the lightning bolt.

• Canonical non-homologous end joining (c-NHEJ)

c-NHEJ is a rapid error-prone repair process that directly ligates the DNA break ends without using a template. It is initiated by the binding of the heterodimer Ku70-Ku80 complex to the DSB site acting as an initial barrier to DNA resection and end processing. This complex provides a platform for further binding of DNA-dependent protein kinase catalytic subunit (DNA-PKcs). This kinase has the ability to autophosphorylate itself promoting the recruitment and activation of downstream transducers that leads to the final DSB seal by a ligation complex, composed of DNA ligase IV (Lig4), X-ray repair cross-complementing protein 4 (XRCC4) and XRCC4-like factor (XLF) proteins (5). XRCC4 is a binding partner of Lig4 that stimulates its enzymatic activity while XLF plays the role of scaffold by enabling the correct positioning of DNA ends prior to ligation (6). However, in some cases DNA ends are not complementary and/or contain modified nucleotides, which necessitates a preliminary processing step prior the ligation. This is ensured by Artemis, a nuclease recruited to DSBs through its interaction with DNA-PKcs (3).

• Homologous recombination (HR)

HR pathway is a slow DSB repair process restricted to cells in S and G2 phases. It uses the undamaged sister chromatid as a template enabling an error-free DNA restoration. Following chromatin decondensation, MRN complex (a tripeptide composed of Mre11-Rad50-Nbs1) recognizes DNA damage and recruits two main sensor proteins: the kinase Ataxia Telangiectasia Mutated protein (ATM) and the CtBP-Interacting protein (CtIP). The interaction between ATM and the NBS1 subunit of the MRN complex promotes ATM autophosphorylation at ser1981 position leading to dissociation of inactive ATM dimers into single protein entities with increased kinase activity (7). This initiates a subsequent cascade of phosphorylation events on various transducer proteins. Meanwhile, CtIP activates the endonuclease activity of the MRE11 subunit of the MRN complex, enabling a partial strand resection and thus the formation of a long 3′-ssDNA tail coated by the replication protein A (RPA) complex in order to prevent DNA hairpin formation (8,9). Next, through the recruitment of Breast cancer type 1 susceptibility protein (BRCA1) by the MRN complex, an accumulation of transducer proteins (PALB2 and BRCA2) to the 3′ssDNA regions enables the exchange between RPA and RAD51 protein. This results in the formation of a RAD51-ssDNA nucleoprotein filaments, which are able to search for a homologous DNA sequence on sister chromatid (4). Then, the action of DNA polymerase δ enables the synthesis of a new DNA sequence using the invaded homologous strand as a template. The DNA break is finally resolved by DNA ligase I (Lig1), an effector protein that ligates the new sequence strand to the other end of the DSB on the original DNA molecule.

• Alternative End Joining (Alt-EJ)

In mammalian cells, DSBs are mainly repaired either via NHEJ or HR pathways. Nevertheless, cells can also use an alternative error-prone pathway called Alt-EJ, which involves the loss of nucleotides and requires stretches of homologous sequences on the same chromosome. Although the mechanism is comparatively poorly understood, it is well accepted that Poly (ADP-ribose) polymerase 1 (PARP1) is the main actor in Alt-EJ process. Similarly to HR, the first step is a short resection stage mediated by the MRN-CtIP association leading to a 5′ DNA end resection of the DNA breaks (10). However, in this case, DSB joining is carried out via annealing of short microhomology sequences (2–20 bp) to the complementary strand spanning the break site (5). PARP1 mediates the alignment of the complementary DNA strands and initiates the formation of long negatively charged poly (ADP-ribose) chains constituting a recruiting platforms for downstream effectors including DNA polymerase θ (Pol θ). By binding to the 3′-ends of micro-homology sequences, Pol θ extends the 3′-ends DNA strands and the break can be sealed either by Lig1 or by the Lig3-XRCC1 complex (11). Although the Alt-EJ activity has been demonstrated throughout all the cell cycle, it strongly fluctuates with a reported maximum in G2-phase and a reduced activity in G1 (12).

The time course of DNA repair depends on the pathways that are involved. As shown in Figure 1B, NHEJ is the fastest pathway leading to complete repair within 1 to 2 h whereas a full HR repair requires at least 8 h (13–15). This is due to the high cellular expression of Ku complex enabling it to reach the DNA lesions in around 1 s post-irradiation during NHEJ (16). In contrast, sensor proteins associated with resection-dependent pathways in HR are recruited after several tens of seconds following DNA damage (16–18).

The proteins involved in all three repair pathways can be divided into two categories: (i) proteins modified near the damage site and (ii) proteins recruited to the damage site (Table 1). In this review, we limit our discussion to one protein per category that is typically used to evaluate DNA damage:

Table 1.

Non-exhaustive list of the main DNA damage marker induced by ionizing radiation. Protocols for marker labeling can be found in the associated references

Protein name	Molecular weight [kDa]	Gene (chromosome location)	Repair pathway	Reference
Proteins modified at the damage site
Phosphorylated H2AX (Ser 139)	14	H2AFX (11q23.3)	All pathways	(54,157,158)
Phosphorylated ATM (Ser 1981)	370	ATM (11q22.3)	HR & Alt-EJ	(158,159)
Phosphorylated DNA-PKc (Ser 2056)	450	PRKDC (8q11.21)	NHEJ	(160–162)
Proteins recruited to the damage site
53BP1	213	TP53BP1 (15q15.3)	NHEJ	(53–56,157)
BRCA1	207	BRCA1 (17q21.31)	HR	(54,163,164)
CtIP	110	RBBP8 (18q11.2)	HR & Alt-EJ	(162,165,166)
Ku70	70	XRCC6 (22q13.2)	NHEJ	(162,165)
Ku80	86	XRCC5 (2q35)	NHEJ	(162,167,168)
Lig1	102	LIG1 (19q13.33)	HR & Alt-EJ	(169)
Lig4	100	LIG4 (13q33.3)	NHEJ	(168,170)
Mre11	80	MRE11 (11q21)	HR & Alt-EJ	(166,171–173)
Nbs1	85	NBN (8q21-24)	HR & Alt-EJ	(17,166,173,174)
PARP1	113	PARP1 (1q42.12)	Alt-EJ	(167)
Rad50	153	RAD50 (5q31.1)	HR & Alt-EJ	(168,174,175)
Rad51	37	RAD51 (15q15.1)	HR	(162,165,176)
RPA	68	RPA1 (17p13.3)	HR	(167,176)
XRCC4	38	XRCC4 (5q14.2)	NHEJ	(167,170)

• Category I (modified near the damage site): phosphorylated histone 2AX (γ-H2AX). H2AX (14 kDa), a member of the histone H2A family, is rapidly phosphorylated at ser139 position following the damaging of genomic DNA. Since this protein modification can be performed by ATM and DNA-PKc (3), γ-H2AX constitutes a DSB marker that is independent of the repair pathway used by the cell. H2AX phosphorylation triggers a cascade of post-translational modifications enabling an increase in DNA accessibility and the recruitment and accumulation of a multitude of DDR proteins at the site of the DNA lesion. Moreover, this chromatin modification creates an epigenetic signal that prevents further DNA damage by anchoring broken ends together through nucleosome repositioning at damaged sites and the recruitment of cohesins, which keep the ends close together during repair to avoid the loss of large chromosomal regions (19–21). Finally, γ-H2AX modulates checkpoint responses by arresting the cell cycle progression, giving DNA time to repair itself (21,22).

• Category II (recruited to the damage site): p53-binding protein 1 (53BP1). 53BP1 is a large protein (214 kDa) encoded by the TP53BP1 gene located on human chromosome 15 and highly conserved across species. Although this member of Tudor-containing proteins was originally described as a binding partner of p53, the evidence of its accumulation following DNA damage indicates that it is a master regulator of DDR (23,24). 53BP1 is expressed throughout the cell cycle, leading to its diffuse localization in non-irradiated nuclei, whereas upon irradiation, it binds to chromatin epitopes (H2A-type histones ubiquitylated by E3 ligase RING finger 168 (RNF168)) surrounding break sites, resulting in accumulation at the DNA damage sites (25). This local increase in 53BP1 acts as a recruitment platform for other DDR proteins and contributes to the selection of DNA repair pathway, which is determined by the cell cycle phase. Specifically, in G1 phase 53BP1 counteracts the function of BRCA1 (an essential HR actor) and impairs the 5′ end resection by interacting with Rif1, thereby promoting c-NHEJ (26–28). In S/G2 phases, BRCA1 and its interacting partner CtIP prevents Rif1 accumulation at the DSB site and therefore the 53BP1-Rif1 binding (29), which promotes 5′ end resection pathways (HR or Alt-EJ) instead. Interestingly, it was demonstrated that 53BP1 loss in BRCA1 deficient cells restores an error-free repair by HR (26,30). Recently, the role of 53BP1 in DNA repair has expanded since studies have demonstrated that it also impairs 5′ end resection, modifies DSB mobility in the nucleus and keeps DSBs intact until their successful repair, thus avoiding ectopic DSB recombination (31–33). In addition, 53BP1 sequence contains 28 sites, which can be phosphorylated by ATM and/or ATR upon induction of DNA damage (32,34). These phosphorylation sites bind downstream partners such as Rif1, indicating that they are not necessary for the recruitment of 53BP1 to DNA damage sites but play other roles in DNA repair signalling. Interestingly, each phosphorylation site of 53BP1 may have different functions in repair pathways (35).

DETECTION OF DSB

There are numerous strategies for the investigation of DNA damage and repair (36,37), each of which has inherent advantages and drawbacks regulating their use in radiation biology.

Direct detection of DSB

Electrophoresis-based approaches, such as the comet assay and pulsed-field gel electrophoresis (PFGE), are classic DSB assays based on direct detection of broken DNA fragments by electrophoresis. In the comet assay, single cells are embedded in agarose gel on a glass slide for microscopy, and broken DNA fragments migrate in an electric field (38). Since undamaged DNA is identified as a spherical mass (head) and the DNA fragments streams out from this (tail), the structure can be seen as a comet. Although this approach does not require either expensive devices or large amount of cells, the information obtained is relatively limited, since the length of DNA fragment remains unknown. In addition, this method has low sensitivity, since it requires a significant amount of DSB to create a detectable tail and finally, the length of the tail is relative to each experiment, making it difficult to compare across various laboratories.

In PFGE, the ability to quantify the amount of DSB per cell is lost, but accuracy is increased compared to the comet assay. Cells are embedded in wells of an agarose gel prior to being lysed. After cell lysis, DNA fragments migrate in the agarose gel during electrophoresis, whereas undamaged DNA and large molecular fragments remains in the wells (39,40). By using a molecular weight as a standard, (semi-) quantitative results can be obtained on the length of DNA fragments.

Despite being commonly used in radiation research (41,42), all electrophoresis-based strategies require large ionizing radiation doses which are typically above the usual radiobiologically relevant doses (ranging from 10^–3 to couples of Gy) and are difficult to replicate among independent laboratories (43). These limitations are addressed by immunofluorescent labeling of DNA repair proteins as molecular markers of the damage, using highly specific antibodies against 53BP1 and γ-H2AX (44).

Detection of molecular response to DSB

Following irradiation (IR), DDR proteins are rapidly localized and/or modified at DSB sites. This leads to protein concentration near the damage site, which can be visualized by fluorescent microscopy as nuclear foci, i.e. radiation-induced foci (RIF). Thus, RIF quantification is an indirect way to detect DNA damage by evaluating DNA repair (molecular response to altered DNA sequence) which results in foci resolution. RIF measurements typically use fixed samples at discrete time points after IR. Alternative strategies based on fluorescent protein fusion constructs enable the monitoring of foci formation and loss in live cells (45,46).

Regardless of the labeling strategy used, foci quantification can be performed using repair proteins either from category I or category II, but the choice of target protein influences the experimental design. For instance, the use of category I proteins implies the labeling of only the small fraction of a distributed protein that is modified near the damage sites, requiring that the cells be fixed in a manner that preserves this modification of the epitope (47,48). In this section, we discuss two different approaches to quantify RIF and present their advantages and limitations.

Intensity-based approaches

Western blotting and flow cytometry are two commonly used methods of RIF quantification based on measuring the difference in intensity of a fluorescent-tagged antibody bound to one of the RIF cluster proteins. These methods provide a quick quantification of a protein of interest in a large cell population. However, they have two major limitations: the fact that they can only provide relative quantification compared to sham control instead of absolute values of RIF, and the inability to either detect changes in the pattern of protein localization, or acquire discrete counts of RIF. Therefore, these approaches can only be used with category I markers (proteins modified near the damage site, e.g. γ-H2AX). In contrast, proteins such as 53BP1, which show a pan-nuclear distribution in unirradiated cells and redistribute to concentrate at the DSB site following IR, typically do not show measurable differences in total fluorescence intensity despite a different pattern of intensity after irradiation. This inability to identify the portion of 53BP1 bind to chromatin from a large 53BP1 protein pool could be resolved by using a phosphorylated form of 53BP1. For example, phosphorylation of 53BP1 on serine 1778 is a protein modification occurring at sites of initial DSBs in a cell cycle-independent manner in response to exogenous DSBs (34,35,49). Labeling of 53BP1^ser1778 enables to detect DSBs based on total fluorescence intensity by using a marker of category I with the same protein target, representing an interesting DSB marker in case of intensity-based detection methods.

Western blotting (WB) is a qualitative method that confirms the presence of a target by a simple visual assessment of protein bands. It becomes a (semi-)quantitative technique through the normalization of protein intensity to the intensity of an internal control protein, expressed in all cell types, whose expression level is unaffected by changing experimental conditions such as actin or tubulin (50). Unlike flow cytometry, WB is unable to investigate intercellular variability in H2AX phosphorylation within a cell population and to identify cell subpopulations since this method averages protein levels between cells. This leads to misinterpretations in case of the induction of an apoptotic process since the associated total γ-H2AX signal may be greater than the one originating from cells containing discrete foci (51). However, it has been demonstrated that WB analysis is a reliable method for the analysis of γ-H2AX after high doses of ionizing radiation when quantification by other methods saturates (52).

Overall, it must be stressed that intensity-based assays measure total protein expression without taking into account their nuclear spatial distribution and are typically less sensitive than imaging approaches. However, flow cytometry can be used to quantify a large number of cells (>10 000) with additional phenotyping based on up to eight surface markers in the same population, which is difficult to reach with imaging methods.

Imaging-based approaches

Imaging-based foci assays enable an easy DSB assessment by imaging and quantifying nuclear spots (foci) labeled with fluorescent antibodies against 53BP1, γ-H2AX or other DSB repair proteins in fixed or live samples. This method has been extensively used for quantifying RIF in a wide range of established human and mouse cell lines (46,53,54), primary cell cultures (55–57), ex vivo peripheral blood lymphocytes (58,59) and histological sections of tissues (60–62). Unlike intensity-based approaches, foci assays enable the detection of changes in both protein modification and protein localization, making it suitable to quantify multiple DNA repair proteins in both previously described categories. Their major advantage lies in the number of valuable parameters for data analysis, including:

1. The average number of foci per nucleus, which is the main parameter used for data analysis. It is related to the number of DSB produced by ionizing radiation as photons or particles pass through the cell.

2. Foci spatial distribution within an individual nucleus, which is related to the type of ionizing radiation (scattered pattern caused by gamma and X-ray irradiation versus distinct tracks caused by particle irradiation) and can also be variable within tissue, for example, due to differences in radiation responses of normoxic versus hypoxic regions in a tumour.

3. Foci size, which can reflect both the complexity of the DSB and the number of DSB contained within one focus. The interpretation of foci size will be discussed in more detail further in this review.

4. The fraction of foci-positive cells. This parameter gives information regarding the homogeneity of DSB distribution in a cell population and separates the cells that have been directly irradiated from bystanders, so is frequently used for bystander studies.

5. The fluorescence intensity inside an individual focus or per nucleus, which is proportional to the number of DSB. However, the use of fluorescence intensity as a reliable marker is complicated by the asynchronous nature of DNA damage foci formation and repair, as will be discussed further in the manuscript (cfr. Section 4.2.)

Finally, the biggest advantage of foci-based assays lies in the opportunity to co-stain with different repair markers. Although 53BP1 and γ-H2AX are commonly accepted as DSB markers, they do not always co-localize due to multiple reasons described in the next sections. Therefore, co-localization of γ-H2AX/53BP1 is currently considered the most reliable marker of DSB (63) and enables a spatial and temporal dissection of DNA repair processes. In addition, different proteins label DSB that are repaired by different pathways: while γ-H2AX foci is an indicator of any type of DSB, 53BP1 and RAD51 foci respectively indicate DSB repaired by NHEJ and HR pathways. A more extensive list of proteins to be used for DSB labeling based on the pathway of interest is provided in Table 1. Therefore, DSB marker co-localization analysis represents a valuable tool for the assessment of DNA repair pathway type and damage complexity, enabling a more precise insight into the genotoxic mechanism triggering the DDR and downstream effects (64). This is particularly useful for pre-clinical studies of new chemo-radiotherapy combinations for which better understanding of the mechanism of action of the synergy between medications and radiotherapy enables to define clinical indications in patients.

However, in fluorescence-based RIF quantification the manual acquisition of images followed by their visual analysis is time-consuming and the inter-/intra-observer variability is often high, limiting the use of imaging-based approaches in large-scale experiments (65). This limitation has spurred the development of high-throughput multiplexing applications characterized by the automation of both data acquisition and image analysis. They also allow simultaneous analysis of multiple adherent cells, as opposed to flow cytometry which analyses individual cells in suspension. Therefore, image cytometry methods have the advantage of combining the high throughput and strong statistical analyses of intensity-based approaches with the spatial information of imaging-based approaches.

In this review, we would like to emphasize the importance of spot co-localization between different repair proteins and discuss the quantification methods which are critical to identify complex DSB (66). Our laboratory has been developing unbiased high-throughput methods for DNA repair foci quantification over the past 15 years with the goal to increase the reproducibility of results (46,67). While RIF quantification is attractive in its apparent simplicity, the underlying biology and physics complicate the data analysis. In the following sections, we will introduced multiple biophysical concepts and their associated confounding factors.

FOCI AS A TOOL TO QUANTIFY RESPONSES TO IONIZING RADIATION AND RELATED CAVEATS

Although foci quantification is a commonly used method to study the cellular response to irradiation due to its apparent simplicity, the complex underlying biology can result in misinterpretation of the results. In this section, we draw attention to a series of confounding factors that can influence the analysis of the results. The careful breakdown of these factors will underscore how challenging a ‘one size fits all’ approach to RIF is as a tool for radiobiology.

Background foci

DSBs are constantly produced in metabolically active cells (10 to 50 DSBs/cell/day, depending on cell cycle and tissue (68)). This results in a γ-H2AX⁺ and/or 53BP1⁺ signal that represents a major confounding factor in studying the effects of radiation. In particular, several groups have reported a diffuse and high intensity γ-H2AX signal in unirradiated cells undergoing S-phase and mitosis (69–71). It has been hypothesized that S-phase cells are the most vulnerable to DNA damage due to active replication of their DNA, a process associated with chromatin states that make DNA more susceptible to damage (70). The increase in spontaneous foci at low confluence levels also reflects DSBs induced by free radicals that are constantly generated in metabolically active cells, together with a higher proportion of G2 cells in cell population (57).

Similarly, differences in background levels of 53BP1⁺ foci have also been observed, even though 53BP1 immunostaining is not affected by cell cycle (45,71). Specifically, our group has reported a decrease in 53BP1+ foci in unirradiated mouse fibroblasts with increasing cell confluence, indicating that proliferating cells exhibit a higher spontaneous DSB level than confluent ones (57).

To take this into account, flow cytometry can benefit from the possibility to directly correlate, within the same cells, γ-H2AX abundance with other attributes of the cells (cell type, phase in cell cycle and cellular health markers) (70,71). By correlating γ-H2AX with DNA content, Mac Phail et al. (69) showed that the γ-H2AX intensity increases as cells move through their cell cycle with cells in S/G2/M phases having higher fluorescence than the expected 2-fold increase from G1 DNA levels.

Similar approaches should be used to eliminate apoptosis (‘programmed cell death’) as another confounding factor. Cells undergoing apoptosis are characterized by the formation of apoptotic bodies in which γ-H2AX expression is important due to a controlled DNA fragmentation. Although the pattern of γ-H2AX in apoptotic bodies differs from the focal patterns produced by DNA damaging agents, including radiation, an intensity-based detection of γ-H2AX will fail to distinguish between them if the γ-H2AX expression cannot be correlated to apoptosis marker (72). Since γ-H2AX positive apoptotic bodies come in the sub-G1 fraction (73), a multi-parameter cytometry analysis that integrates DNA content/cell cycle distribution and apoptosis markers would address both confounding factors simultaneously.

Similarly, background foci are a major confounding factor for RIF quantification since they are indistinguishable from radiation-induced foci. This is especially important for low dose radiation research. Assuming that low-LET radiation induces 20 to 30 RIF/Gy (46,53,74), applications such as airport security checks or dental radiography that typically deliver radiation in a dose range from 0.001 to 0.1 Gy, induce an increment of only 0.02 to 3 RIF/nucleus compared to the background level, leading to a situation in which the amount of RIF generated by the IR can fall within the range of background variability. To cope with this limitation, the design of studies has to be optimized, for example, by using confluent cultures and cell types known to have low proliferation rates, such as isolated PBMCs (although the variability in metabolic activity and free radical generation at baseline would still need to be addressed). In cycling cells, specific tools have to be used to take into account the cell proliferation as a confounding factor (57).

Dependence of time and dose rate

Studies on DSB repair usually use discrete time points post-IR in fixed samples. This approach only offers a static view of a dynamic process in which time is a major confounding factor due to the delay between irradiation and DSB detection/repair. This leads to a higher amount of RIF that is actually produced throughout the process (cumulative RIF count) than the maximum amount of RIF observed at any time point (RIF per cell count), as reported in (46) and illustrated in Figure 2A.

Figure 2.

Time as a confounding factor of radiation induced foci (RIF) quantification. (A) Kinetics of DSB detection and repair. The average number of RIF per cell observed at a given time post irradiation (solid black line) is lower than the cumulated number of RIF (solid red line) which saturates at a value equal to A D. This comes from the concurrent DSB detection (dashed line) and DSB repair (dotted line). (B) Repaired fraction of DSB at time of next DSB creation is presented depending on the dose rate and the DSB repair half-life, a parameter determined by the repair kinetic analysis. Repaired fractions were calculated on a 35 DSB/Gy basis. Red areas show conditions where DSB are fully repaired when the next one is created (repaired fraction 100%) while blue ones correspond to conditions where DSB are accumulating due to a high DSB creation rate (dose rate) and a low DSB repair rate (high DSB repair half-life).

The number of observed RIF usually reaches a maximum around 30 min post-IR, followed by a gradually decrease as DNA repair occurs (46,52,53). Low RIF levels prior to this time point can be explained by various factors:

• RIF are detected as 0.2–0.8 μm² foci, a larger size than the DSB itself suggesting that multiple copies of the protein of interest are localized at the same site. Therefore, some RIF might remain below the detectable levels at early time points post-IR.

• Some DSBs are repaired by mechanisms that do not require the recruitment of the target protein used in the assay.

• Multiple proteins are involved in DNA damage detection requiring extra time to assemble all the components of DSB repair machinery, especially if the site is less accessible, leading to a delay in DSB being formed and it being detected.

Furthermore, extra DSBs can be created by other radiation-induced phenomena in addition to those created directly by the passage of the ionizing radiation. Bystander effect is one of them and describes a phenomenon in which radiation response is observed in cells not traversed by the ionizing radiation. This response is caused by the release of growth factors, cytokines, oxygen/nitrogen species and exosomes transmitted via Gap junctions, direct intercellular communication, or soluble factors from irradiated cells to neighbouring ones (75,76). Among exchanged molecules, long-lived reactive oxygen species are the main source of formation of DSB in unirradiated cells. For example, significant induction of DNA damage was reported in unirradiated cells neighbouring irradiated cells, while this induction drastically decreased when cells were pre-incubated with an inhibitor of Gap junction (77). Interestingly, this response seems to be cell type/ tissue-specific (78), dependent of the irradiation modality (broad beam versus microbeam) (79) and genetically controlled as higher bystander effects were reported in DNA repair-deficient cells (80).

In addition, the attempted repair of clustered DNA damage sites (e.g. closely spaced single lesions within one or two helical turns of DNA) by the base excision repair oxidative DNA glycosylases can lead to extra DSB formation in human cells. Consequently, Georgakilas et al. detected additional DSBs 2 days post-IR when repair-proficient human monocytes 28SC were exposed to γ-rays (81).

As both bystander effect and base excision repair-mediated DSB formation require time to occur, extra DSB are produced in cells with a delay compared to the initial DSBs caused by IR (82,83). This impacts the kinetics and assessment of DSBs since their repair and the formation of new ones are concurrent.

Studies also suggest that radiation dose rate has a marked influence on RIF frequency. Barnard et al. (84) observed a reduced amount of 53BP1 RIF/nucleus in lymphocytes 24 h post-IR when a total accumulated dose of 1 Gy was delivered at low dose rate (0.01 Gy/min; irradiation time: 1.7 h) rather than at higher dose rate (0.3 Gy/min; irradiation time: 3.3 min). Similarly, Brooks et al. (85) were unable to distinguish γ-H2AX RIF from background foci (unirradiated cells) when a total dose of 5 Gy was delivered at 0.3 mGy/min (irradiation time: 11.5 days). However, the expected increase in RIF level was observed when the same irradiation was performed at 1.8 Gy/min (irradiation time: 2.8 min). This discrepancy could be caused by the fact that during a chronic exposure, DSB formation induced by IR happens simultaneously with the DSB repair, a process that take place on a time scale from minutes to hours. For a given dose of radiation, this leads to a reduced RIF frequency since cells have time to repair a fraction of DNA damage (RIFs are resolved and no longer detectable). Therefore, low dose and low dose rate irradiations might be associated to a RIF frequency of the same order of magnitude as background foci making the signal to noise distinction very challenging. This effect depends on the exposure dose rate as well as the DSB repair rate. Figure 2B highlights that dose rate above 10^–2 Gy/min leads to an accumulation of RIFs since DSBs are not repaired prior to the creation of the next one (repaired fraction 0%) independently of the DSB repair half-life. However, when the dose rate is below 10^–4 to 10^–5 Gy/min, the repaired fraction of DSB at the time of next DSB formation reaches 100% (fully repaired) impacting the detection of RIF post-IR like in Brooks et al. (85).

It is interesting to note that this dose rate effect is used in various clinical applications. While diagnostic applications in nuclear medicine use low dose rate radiation (≈1 mGy/min) to prevent patient side-effects, cancer external radiotherapy is based on overwhelming the DSB repair capacity by using high dose rate IR ranging from 1 Gy/min in intensity-modulated radiation therapy to 10 000 Gy/min in Flash therapy (Figure 3). Thus, dose rate dependency has to be taken into account for RIF quantification in low dose rate IR but can be omitted in high dose rate experiments (86).

Figure 3.

Different biomedical applications of ionizing radiation depending on their dose, dose rate and linear energy transfer; IORT, Intra-Operative Radiation Therapy; PRRT, Peptide Receptor Radiation Therapy.

Due to concurrent DSB generation by IR (especially in low dose rate irradiation experiments), DNA damage recognition by antibodies and DSB repair, time is a major confounding factor in studying DNA repair. In order to draw strong conclusion on the kinetics of repair process, key repair parameters have to be extracted from mathematical models (see Section ‘Modeling of DSB Formation and Repair’) and analysis has to be strengthened by RIF quantification in the same samples and/or experiments at multiple time points.

Dependence of dose

A 1:1 correspondence is often assumed between DSB and RIF. However, based on experimental results, a non-linear dose response behaviour occurs leading to RIF saturation (RIF numbers lower than the expected number of DSB) at high radiation doses, as illustrated in Figure 4A. Several groups reported a maximum of γ-H2AX⁺ RIF/Gy ranging from of 17 to 24 after low dose X-ray exposition (1 Gy) while this number decreased to 13–15 RIF/Gy in the range 1 to 4 Gy (87–89).

Figure 4.

(A) Dose saturation of average 53BP1 RIF/cell in fibroblasts from 15 mouse strains 4 h after X-rays exposure. Black dashed line represents a linear fit and red continuous line indicates nonlinear fit. Data adapted from (55). (B) (Left) Schematic representation of a nucleus exposed to 1 GeV/amu Fe ion. (Right) Human fibroblasts exposed to 1 GeV/amu Fe ion, and immunostained for detection of 53BP1. Each white focus which corresponds to a RIF were surrounded in red or green if the damage is originated from the dose deposition of the core track or the delta rays, respectively. (C) Two different geometrical configuration of cell monolayer during irradiation with arrows representing direction of high-LET beam. Blue circle and green dots correspond to nucleus and radiation-induced-foci respectively. (D) RIF/μm dependence on LET in primary skin fibroblasts of Balb/c mice at 4 h post-irradiation. Black dashed line represents a linear fit and red continuous line indicates nonlinear fit. Data adapted from (55).

We have hypothesized that the observed RIF saturation reflects the fact that as the total dose increases, multiple DSBs start co-localizing within single RIF. For this process to take place, the dose delivery must happen at rate faster than repair (i.e. within a few minutes). Using time-lapse fluorescent microscopy of 53BP1 fused to mCherry, Georgescu et al. (45) demonstrated the RIF dynamics in non-malignant human mammary epithelial cells MCF10A, indicating RIF movement was following a diffusion pattern constrained within a 1.5–2 μm radius. In this work, time-lapse imaging also showed several RIF merging within constrained domains of size of 7.5 μm². Similarly, in a recent study by Penninckx et al. (55), we demonstrated a non-linear dose dependence of RIF in non-immortalized fibroblasts from 15 different mouse strains, with lower RIF number than expected at 1 Gy and 4 Gy 4 h following exposure to X-ray, compared to 0.1 Gy.

It is critical to account for time dependence in observing and interpreting the effects of dose saturation. For instance, in our study on mouse fibroblasts 48 h after exposure to X-ray, the 4 Gy dose point was much higher than what one would predict with a linear dependence at 1 or 0.1 Gy, suggesting that after 4 Gy irradiation, RIF were unable to fully resolve. Based on this discrepancy, we hypothesized that as the dose increases, more DSBs are localized per isolated RIF, making it more difficult for the cell to resolve these foci, leaving permanent ‘scars’ of persistent DNA damage in the cell.

In the same study, we also showed that RIF measurements were less variable within cell lines derived from the same strain of mice than between different strains. This result suggests that DSB clustering is modulated by the genetic background of a strain, which may be a phenomenon occurring in all mammalian cells and worth considering during experimental design. On the other hand, another confounding factor for interpreting the differences in RIF resolution may be the variability in DDR pathway efficiency in each mouse strain. For instance, lower RIF levels at early time points can reflect faster DSB repair and/or a higher amount of DSB clustering, which have opposite biological effects. Thus, to accurately assess DNA repair efficiency, it is important to quantify the remaining RIF at later time points as well.

Overall, RIF saturation observed with the increase in IR dose is a major confounding factor in DSB quantification, because it indicates that multiple DSBs can coalesce leading to the formation of ‘repair centres’ where DSBs colocalize within single RIF. Evidence for this hypothesis of non-static DNA damage sites is growing in literature, contradicting the classic ‘contact-first’ model where DSBs are assumed to be immobile, unable to coalesce and repaired at the lesion site. In terms of analysis approaches, RIF saturation results in the loss of 1:1 correspondence between DSB and RIF, requiring the use of a clustering parameter to draw reliable conclusions on DSB quantification, as presented in Section 5 (90).

Charged particles

Since several decades, radiobiology researches using charged particles is growing worldwide ranging from proton and carbon ions in clinical settings to high mass-high energy (HZE) particles in space radiobiology. These ionizing radiations are characterized by a high linear energy transfer (LET). They deposit the majority of their energy nearby their linear path (typically within nanometers, forming a region defined as the core of the track) leading to a concentration of DSB along the particle track inside the nucleus (Figure 4B), accompanied by a low dose deposition induced by a radial projection of secondary ionizing radiations called ‘δ rays’.

These physical properties influence the design of DNA damage experiments. If cell monolayers are irradiated by a high-LET beam oriented perpendicularly to cells (i.e. along the Z axis when the cells are located in the XY plane), RIF frequency will typically match the number of particles crossing the cell and not necessarily the number of individual DSBs induced by these particles (Figure 4C). As we have previously investigated (88), this is due to a combination of factors. First, the cells are much thinner in this orientation, leading to fewer DSBs generated for each track. In addition, the resolution of microscopes along the Z axis is much lower than in the X-Y axis, which makes it more difficult to resolve multiple foci, even by confocal microscopy. Finally, this spatial configuration may increase the amount of DSB clustering, as DSB movement is constrained along the Z axis. For these reasons, we suggest irradiating cells in the XY plane, slightly tilting the plates at ∼5° angle from the horizontal plane in order to get longer particle tracks which are better resolved during imaging (Figure 4C).

Due to their high linear and low radial dose deposition, HZE particle impact and their associated biological response is better described by the concept of fluence (particle/unit area) than average dose. Specifically, since the number of cells being traversed by particles follows a Poisson distribution, the average dose delivered to a cell population is less biologically meaningful, because cells are either traversed by one or more particles, leading to a significant dose deposition, or not traversed by a particle at all, resulting in negligible dose. The relationship between the macroscopic dose deposited in a cell population by a broad beam and the fluence are given by:

(1)

Where φ is the fluence [particles.cm^–²], ρ is the material density [g.cm^–³] and LET is expressed in keV/μm.

This suggests that the average dose deposited in the sample is not sufficient information for drawing strong conclusions on response to ‘low-dose’ HZE particles. Yang et al. (91) observed that DNA damage is detectable at ∼10 times lower dose of protons (70 μGy) than ⁵⁶Fe particles (0.5 mGy). This finding can be explained by the difference in the number of particles passing through the cell. In their experimental design, the dose of 70 μGy protons corresponds to two protons crossing each cell while only 0.02 iron ions cross each cell at 0.5 mGy, e.g. a 100-fold fewer iron ions per cell are needed to induce significant DNA damage than in the case of protons due to their difference in LET. Thus, while most cells received at least one proton at 70 μGy, only a few percent of the cells were actually crossed by an iron particle. This leads to a very heterogeneous situation where the crossed cells received a huge dose, but the large majority of cells received no significant dose (or only a small dose from δ rays) resulting in an average low dose deposition in the cell population. This highlights the importance to distinguish the response of irradiated cell from the one of cells that were not traversed by the main track in case of low dose/low fluence HZE exposure (91–93).

In addition to dose, LET is another confounding factor in studying the biological response to HZE particles. As energy deposition along the track increases with LET, the number of DSBs increases faster than the RIF frequency, as more DSB clustering within a single RIF occurs in cells (94). In Penninckx et al. (55), we showed that the number of RIF/cell at 4 h post-IR is similar after exposition to 104 keV/μm ⁴⁰Ar and 170 keV/μm ⁵⁶Fe for the same fluence, even if theory predicts more DSB/cell for the higher LET at identical fluence. This was observed in >70 primary cell cultures derived from 15 different strains of mice. Moreover, we reported a decrease in RIF/Gy with increasing LET at 4 h post-IR. This finding cannot be explained by DSB repair kinetic, as it would be contradicting the fact that DNA repair is slower as the LET increases due to more complex DSB (95). We therefore hypothesize DSB clustering is the reason for this LET dependence and suggest that DSB move to cluster within specific areas along particle tracks for easier repair.

Based on these considerations, we propose to replace the classic RIF/cell metric previously used for X-ray with a metric which better reflects responses to HZE particles: the average number of RIF per unit length along each track imaged by microscopy, with RIF/μm as a unit. This implies that particle tracks have to be visible to enable accurate RIF quantification and track detection by software. Nevertheless, the metric RIF/μm can be estimated from the number of RIF detected per cell using the following expression (55):

(2)

Where B is the foci background level measured without irradiation, φ is the fluence and V is the average volume of the nucleus.

It has to be noted that, in a similar way to high X-ray doses, we reported a saturation of the RIF/μm as LET of particle increases (Figure 4D).

As an alternative to RIF number per unit length, two other endpoints can provide biologically relevant information after high-LET IR: foci intensity and foci size (94).

• Foci intensity. Even though the number of RIF/μm reaches a plateau at about 200 keV/μm, a linear increase in total foci intensity per unit length was reported with LET increasing above this value (94). This constitutes an additional evidence towards DSB coalescence: since a linear increase in DSB density is expected with LET, but the number of RIF only increases modestly, the average number of DSB per RIF increases, increasing RIF intensity by reflecting the concentration of protein recruited at site.

• Foci size as a surrogate marker of DSB complexity. Oike et al. (96) compared computed LET distribution in radiotherapy treatment to 53BP1+ foci size distribution in fresh biopsy samples of cervical cancer. They reported a bimodal distribution of RIF at the centre of the spread-out Bragg peak which is similar to the predicted LET distribution. The study further suggests that low-LET peaks lead to small foci, similar in size to those induced by X-ray (single DSBs per RIF), while higher-LET peaks induced larger RIFs (multiple DSBs per RIF). In another study, Reindl et al. (97) demonstrated that 53BP1+ and γ-H2AX+ RIF induced by higher LET 500 keV/μm carbon ions are significantly larger (0.54 μm) than ones generated by 2.6 keV/μm proton IR (≈0.4 μm), thus presumably contain more DSBs per RIF. It is also important to note that the size of foci may change with respect to the amount of damage change with time (88).

Inter-individual variability

High variability of baseline and radiation-induced DNA damage has been reported among genetically diverse individuals, both humans and model organisms, suggesting that both genomic and epigenetic factors might influence DNA repair processes. Mutations in the key genes encoding the proteins in both HR and NHEJ pathways and other genes discovered to modify their effects have been implicated both in carcinogenesis (98–101) and in sensitivity to ionizing radiation (102,103). However, the phenotypic effects of these mutations are determined by the type of irradiation and the DDR pathway that is recruited. Thus, HR mutations are crucial for responses to high-LET particles, while NHEJ mutations affect both low and high-LET responses (104,105).

A recent study has indicated major differences in cancer progression after irradiation with high-LET particles as well as gamma rays in a genetically diverse mouse population (106), which is consistent with our previous results demonstrating a high variability in 53BP1⁺ foci formation in ex vivo samples among 15 genetically different mouse lines both at baseline and after irradiation. Notably, in both cases the mouse strains that showed the strongest and the weakest responses to low-LET X-ray and gamma ray irradiation only partially overlapped with the responses to high-LET particle irradiation, indicating both general and LET-specific factors regulating radiation responses. This result was recapitulated by a recent study from our group on the genomic associations with radiosensitivity in 15 mouse strains, which revealed SNPs mapping to distinct pathways that were associated with responses either to all ionizing radiation, or to high-LET particles, but not low-LET X-rays. [Cekanaviciute et al., submitted]

In humans due to genomic and phenotypic diversity the individual differences are even more pronounced. Although we have recently demonstrated the general pattern of dose and LET-dependent increase in 53BP1⁺ RIF formation in human PBMCs ex vivo, the individual average RIF/cell values showed major variability (59). A similar variability in responses to radiation has previously been reported using oxidative stress markers (107) and changes in gene expression (108). Some of the phenotypic factors associated with increased DNA damage foci formation are already known, for example, we have reported a significant increase in baseline DNA damage with age, but no difference based on sex, by assessing spontaneous 53BP1⁺ foci in PBMC from 674 healthy donors (59). At a whole organism level, multiple studies of genomic associations with negative side effects from radiotherapy have identified a number of SNPs that merit further analysis to uncover the associated changes in cellular functions, and specifically, to define whether they affect DNA damage repair pathways (109,110). Finally, other genomic and phenotypic associations with both baseline and radiation-induced DNA damage remain to be investigated at a cellular level in both healthy humans and model organisms, since they might predict individual responses to both therapeutic and space radiation and suggest targets for developing novel therapeutics.

Persistent RIFs

The RIFs observed at late time points (24 h and later post-IR) are usually considered unrepaired DNA damage and suggest a flawed repair process (111). However, they also depend on the method of quantification: γ-H2AX⁺ RIF could reportedly be observed even at time points when direct methods for DSB detection, such as PFGE, suggest no unrepaired DNA damage (112,113). Although this result might be considered as a proof of the greater sensitivity of foci-based assay, it may also indicate that residual RIFs may not be a specific marker of unrepaired IR-induced DSB and may reflect in fact chromatin changes (114). In general, the biological interpretation of residual DNA damage foci is still subject to debate and several theories have been proposed.

Ojima et al. (115) showed that 50–70% of the p-ATM foci observed in the low-dose range 1.2–5 mGy were due to bystander rather than direct radiation effects. Interestingly, several groups have reported that bystander DSB are produced in cells with a delay compared to the IR time and are not readily repaired, maintaining unrepaired bystander DSBs for longer time period. The extent of this delay varies between experimental models, being hours in in vitro cell cultures to days in tissue and animal models (79,82,116,117). For instance, Ojima et al. (82) observed that 48 h after irradiation, 81% of the initial numbers of p-ATM foci are still detected in unirradiated cells co-cultured with cells exposed to 20 mGy. In the light of the previous sections, it has to be noted that these bystander foci induced at very low dose are difficult to distinguish from baseline DNA damage.

Similarly, as mentioned previously in this review, DSBs can be created several hours post-IR by attempted repair of clustered damage sites by DNA glycosylases leading to extra RIF observed 48 h after beam exposure (81,83). This suggests that residual RIFs can be related to the induction of secondary effects of IR that eventually produce new DNA damage foci, which persist for much longer than the direct IR-induced foci.

However, an alternative explanation could be that the reported RIFs do not represent unrepaired DSB, but instead, DNA regions where chromatin decondensation occurs. Consistently with this interpretation, Suzuki et al. (118) observed persistent RIFs in normal human fibroblasts 5 days following 4 Gy X-ray IR, even though by that time all DSBs should have been fully repaired. The residual γ-H2AX RIFs were large and co-localized with phospho-ATM as well as with p53 phosphorylated at serine 15, suggesting ongoing DSB repair. In a follow-up study, the same persistent RIFs were detected on intact metaphase chromosomes that did not contain any DNA fragments 96 h following IR (119). From these results, the authors concluded that persistent foci are related to chromatin alteration, which causes senescence-like growth arrest following IR. Senescence-like growth arrest is a p53-dependent irreversible G1 arrest thought to suppress radiation-induced telomere dysfunction following genomic instability in fibroblasts. Notably, larger initial IR-induced foci, which contain more p53, are particularly good at triggering this G1 arrest (120) and leading to persistent chromatin alteration.

Moreover, chromatin density can modulate the DNA damage response to radiation because chromatin decondensation around the DSB triggers ATM autophosphorylation and subsequent DDR activation (121), leading to a slower DSB repair in heterochromatic regions compared to euchromatic ones (122). However, Lorat et al. (123) observed that persistent 53BP1⁺ foci in heterochromatin did not colocalize with NHEJ proteins, suggesting that radiation-induced DSBs are repaired even in heterochromatic regions, but that these sites remain marked by 53BP1 after the repair is complete. They proposed that the residual 53BP1⁺ foci may reflect irreversible chromatin structural changes, leaving a DSB-induced epigenetic memory of DNA damage that may lead to future perturbation in gene expression (124). However, Chiolo et al. (114) reported that radiation induced-DSBs in heterochromatin of Drosophila cells are typically repaired by HR pathway suggesting alternative explanation of these results.

MODELING OF DSB FORMATION AND REPAIR

Owing to all the aforementioned confounding factors (time, dose, LET and genetic), a careful choice of experimental parameters must be made in the experimental design to ensure the relevance of the results generated. In addition, the analysis of repair foci must be done on a sufficient number of cells to ensure that differences between conditions can be statistically reported if they are present. Given the complexity of RIF scoring, we have developed the ‘tool for enhanced results of RIF in cells’ (terrific) tool: https://radbiolab.shinyapps.io/terrific/ (57). This online open access application provides guidelines to assess the minimum number of cells required to reach statistical significance for RIF detection as a function of irradiation dose, cell confluence and time post-irradiation.

Finally, DSB formation and repair is a dynamic process that have to be analyzed using mathematical models in order to draw strong conclusions (46,53). Following acute exposure to ionizing radiation, one DSB is detected by a target antibody used for immunostaining at a rate k₁ creating one RIF which will be observed by fluorescence. This DSB will be repaired at a rate k₂ leading to disappearance of the fluorescent signal (RIF resolution). By assuming that both processes are irreversible, we can represent the model as followed:

The kinetic evolution of this model can be mathematically expressed as a system of differential equations that define the change in DSBs and the change in RIFs respectively:

The resolution of this system requires two initial conditions: First, the number of DSB created by the irradiation can be decomposed into two parameters A and D corresponding to the number of DSB per Gray of radiation and the dose in Gray delivered to the cell respectively. Second, the number of RIF observed at the beginning of irradiation (t = 0) is equal to 0 since the background (spontaneous foci) are subtracted from the RIF number:

By solving the system of differential equations with the two aforementioned initial conditions, we find two solutions describing the average number of RIF and the amount of unrepaired DSB at a given time post-IR:

(3)

(4)

Equation 3 can be used to fit the number of RIF at a given time (static measure). However, time-lapse imaging can also be used to measure the total number of RIF that have been produced since the beginning of irradiation (cumulative measure since t = 0, see Figure 2A). This can be described mathematically from equation 3 by putting k₂ to 0:

(5)

Although obtaining these fit parameters enables a better understanding of the evolution of DNA repair than a simple foci scoring, it is important to note that the basic mathematical model presented above can be improved by taking into account the aforementioned confounding factors (see Section ‘Foci as a Tool to Quantify Responses to Ionizing Radiation and Related Caveat’). In this context, we have previously developed a method to account for the non-linear dependence between foci and dose by introducing a clustering parameter (C) and not letting the decay in RIF return to zero by adding an offset parameter, as shown below:

(6)

where τ is the repair time constant and ρ the repaired fraction of DSB. β represents the number of DSB generated at irradiation time (t = 0) per unit of X-ray dose and is considered constant. C is the number of DSB per RIF, which reflects the amount of clustering at a given dose (monotonically increasing with dose, with the shape of the curve depending on the cell type, genetics and species; Figure 4A).

In a similar way, the kinetic model of DSB repair can be improved in case of HZE irradiation by fitting simultaneously different LET and time data using a clustering parameter (C) and the RIF/μm metric:

(7)

Where τ is the repair time constant and ρ the repaired fraction of DSB. α represents the number of DSB generated at irradiation time (t = 0) per unit of LET and is considered constant. C is the number of DSB per RIF, which reflects the amount of clustering at a given LET, depending on the cell type of interest (monotonically increasing with LET with shape of the curve depending on the cell type, genetics and species; Figure 4D).

RIF QUANTIFICATION TO ASSESS AND PREDICT RADIATION RESPONSE

RIF quantification as a method to assess radiation exposure

The number of medical X-ray examinations has increased dramatically during the past decades, triggering concern about the associated health risk. In this context, quantification of radiation-induced DNA damage foci in PBMCs has been proposed as a potent biomarker enabling to evaluate radiation response on individuals and apply necessary countermeasures (125,126). Furthermore, DSB assessment could provide a retrospective estimation of received radiation dose, a key information for monitoring radiotherapy treatments or for patient management following accidental uncontrolled exposure to IR. Foci assays are a particularly advantageous biomarker, since they are characterized by a reproducible RIF–dose relationship, especially in the low dose range where performance of other biodosimetry approaches is limited (127). Furthermore, they remain robust despite high variability between samples: Vandevoorde et al. (63) reported that blinded samples originating from different human donors could be successfully ranked on the basis of ex vivo received dose in a range from 0.01 to 0.1 Gy by assessing the amount of γ-H2AX⁺ foci/cell analyzed 30 min post IR.

In addition, foci assay also provides information regarding exposure homogeneity. Accidental radiation exposures are characterized by a high dose deposition in a fraction of tissue, but a total body dose that remains low. Therefore, using mathematical models to extract dose information from RIF distribution analysis is a helpful tool for patient management (128). However, the dose–RIF response curve is modulated by several factors including life style, genetics and time post-IR, requiring individual and time-specific calibration curves for accurate dose estimation (129).

During the last decade, multiple studies have investigated the potential of RIF as a biodosimeter in case of large-scale uncontrolled exposure to IR. They consistently highlighted that the estimated dose from microscopy analysis was well-correlated with true delivered doses (58,63,130,131). Ainsbury et al. (132) reported that 70% of samples (28/40) were correctly associated with the received dose, and this value rose to 85% (34/40) by considering samples for which dose estimation was either correct or higher than the real one.

On the other hand, even though RIF scoring offers promising results, microscopy analysis remains a time-consuming method with limited usefulness in case of large-scale uncontrolled exposure in which the triage of thousands of samples in very short time is required. Therefore, automated alternatives have emerged, such as Dosikit, a portable dosimeter based on fluorescence spectroscopy that is able to perform high-throughput analysis of total γ-H2AX fluorescence intensity in blood and plucked hairs sampled directly on accident site (133,134). Due to lower sensitivity of spectroscopy compared to microscopy analysis, the dose estimation from Dosikit is less accurate, but enables a triage of individuals into three groups (<2 Gy, 2–5 Gy, >5 Gy) in only 45 min. While the most accurate dose estimation is preferable from the dosimetric point of view, these dose ranges can provide sufficient information to meet the clinical needs of disaster medicine, considering that few casualties with severe clinical response are expected below certain dose thresholds (deterministic effects) (135). Thus, biodosimetry approaches such as Dosikit could be used as an early triage to identify victims who will require priority follow-up in hospitals. Notably, in this setting, the accuracy of dose estimation is difficult to improve, because critical individual information (e.g. foci baseline, DNA repair and clustering abilities) is missing at the time of accident. This lack of data on individual variability coupled to significant inter-laboratory differences in RIF scoring are the main limitations of foci-based dosimetry for applications that require community screening (127).

RIF quantification as a method to predict personalized radiation response

Radiotherapy is considered as the main treatment modality against cancer since it was estimated that 50% of all cancer patients will experience it during their treatment (136). However, the use of combined treatments, such as chemotherapy drugs coupled with radiotherapy sessions, is growing worldwide. In this case, DNA damage quantification constitutes a novel pharmacodynamic marker of drug effects, able to give information on treatment outcome. It is therefore not surprising to observe an increasing interest for DNA damage assessment in cancer clinical trials (Supplementary Table S1).

Currently, the choice of radiation therapy in the treatment plan of a given patient is driven by clinical and pathological features including the primary site localization, the tumor stage and its histology. However, increasing understanding of patient-related factors that influence the radiation response highlights the limitation of this medical framework: each person is fundamentally different from the average of the population. This leads to 5–10% of patients that experienced severe side-effects as a results of the dose prescribed in the treatment (137), while the same dose can be insufficient to achieve a satisfactory tumour control in radioresistant patients. Therefore, the use of personalized approaches to identify the radiosensitivity profile of a patient in advance could optimize radiotherapy efficacy by avoiding severe side-effects in radiosensitive subpopulation and enabling a safe dose escalation in the rest of patients.

Advances in this personalized medicine approach can be envisioned to go hand-in-hand with the development of assays for predicting the individual radiation response. RIF assays are likely to be the most suitable for this purpose. For example, the amount of residual 53BP1⁺ and γ-H2AX⁺ foci at 24 h post-IR was reported to be correlated to cell death (56,138). In addition, by assessing proteins specific to HR pathways, RIF assays can identify HR-deficient tumours known to be more sensitive to proton therapy (139) and combined radiotherapy with platinum drugs or PARP inhibitors (140). This suggests that HR-pathway specific Rad51⁺ foci assessment on ex vivo biopsy materials might be used as an alternative to gene sequencing for patient selection and treatment choice.

RIF assay has also been investigated as a predictive tool for the identification of patients at risk of severe side effects, but showed contradictory results (Table 2). On one hand, several groups demonstrated significant correlation between higher amounts of foci in PBMCs and radiation toxicity, suggesting the potential use of RIF as a predictive marker for radiosensitivity. Lobachevsky et al. (141) demonstrated a positive correlation between the proportion of residual RIFs and the induction of dysphagia and cough in lung cancer patients treated by thoracic radiotherapy. Consistently with this finding, another study reported the association between RIF foci level at 24 h and complications in prostate cancer patients (142). Moreover, this study demonstrated that γ-H2AX⁺ foci scoring could also distinguish patients experiencing grade ≥ 3 severe radiation side-effects from those without radiation toxicities (grade = 0). On the other hand, no significant associations were reported between RIF levels and severity of normal tissue side-effects in four different studies focusing on gynecological (143), head and neck (144), prostate (145) and rectal (146) cancer patients. For example, Brzozowska et al. (145) found no significant differences in the mean RIF number per cell at different time points post-IR by comparing patients with and without side effects. This apparent controversy in results opens the question of RIF assay reliability in assessing individual radiosensitivity. A meta-analysis of the aforementioned studies would be difficult to perform due to large diversity in reported parameters: indicators, techniques (RIF counting versus total fluorescence intensity by flow cytometry), irradiation modalities, received dose and time point post-IR (cfr. Table 2). However, it appears that the very low number of patients involved in some studies can partially explained this discrepancy. Moreover, the large majority of these studies focus on residual RIFs, an endpoint for which the biological interpretation is still under debate as previously discussed.

Table 2.

Studies investigating the use of radiation-induced foci (RIF) assay as predictive tool for normal tissue response in radiation therapy. Study reporting no significant association between RIF and clinical outcome are stained in gray

Cancer type	# of patients	Indicator	Clinical observation	Reference
Breast cancer	57	γ-H2AX foci at 24 h post-IR	A higher number of RIF was detected in patients experiencing grade 3 skin reaction compared to patient with grade 0 reaction	Djuzenova et al. (177)
	80	γ-H2AX foci at 6 h post-IR	The residual RIF can be used as predictive tools to identify over-responding patients that will experience acute skin reactions	Mumbrekar et al. (178)
	24	γ-H2AX/53BP1 foci at 24 h post-IR	A positive correlation was observed between residual RIF levels and late normal tissue toxicity	Vandevoorde et al. (179)
Head and neck cancer	31	γ-H2AX foci at various time post-IR	No signification association between RIF and clinical outcome	Werbrouck et al. (144)
	31	γ-H2AX foci at 24 h post-IR	An increased incidence of severe oral mucositis was observed in patients with large amount of residual RIF	Fleckenstein et al. (180)
	54	γ-H2AX foci at 6 h post-IR	An increased skin toxicity was observed in patients with high amount of residual RIF	Goutham et al. (181)
	25	γ-H2AX foci at 24 h post-IR	Patients with severe oral mucositis had higher residual RIF compared to patients experiencing mild oral mucositis	Li et al. (182)
Gynaecological cancer	29	γ-H2AX foci at 24 h post-IR	No signification correlation between a residual excess of foci and side-effect severity	Werbrouck et al. (143)
Lung cancer	45	Unrepaired component (fitting parameter)	A positive correlation was observed between unrepaired component and induction of dysphagia/cough in lung cancer patients	Lobachevsky et al. (141)
Paediatric cancer	47	γ-H2AX foci at 24 h post-IR	ATM -/- homozygote patients that developed acute normal tissue toxicity are characterized by high residual RIF	Rube et al. (183)
Prostate cancer	25	γ-H2AX foci at various time post-IR	No signification association between RIF and clinical outcome	Brzozowska et al. (145)
	30	Background 53BP1 foci	High baseline DNA damage in T Cells is associated to more severe clinical side effects after proton irradiation	Pariset et al. (59)
	61	γ-H2AX foci decay ratio (RIF at 0.5 h/ RIF at 24 h)	Lower decay ratios were observed in over-responding patients compared to other ones	Van Oorschot et al. (184)
	200	γ-H2AX foci decay ratio (RIF at 0.5 h/ RIF at 24 h)	RIF assay can distinguish patients experiencing severe radiation toxicity (grade ≥ 3) from those without toxicity (grade = 0)	Van Oorschot et al. (142)
Rectal cancer	53	γ-H2AX foci at 24 h post-IR	No signification correlation between RIF and clinical outcome	Djuzenova et al. (146)
Various cancer type	22	γ-H2AX foci at 24 h post-IR	The 12 over-responding patients of the cohort are characterized by high residual RIF	Bourton et al. (185)
	28	Unrepaired component (fitting parameter)	Over-responding patients are characterized by a higher unrepaired component than non-over-responding individuals.	Lobachevsky et al. (186)

In a recent study by our group, Pariset et al. (59) observed that levels of spontaneous foci in T lymphocytes of prostate cancer patients (i.e. extracted before irradiation) can be used as a surrogate biomarker of radiation toxicity. The findings indicate that the median baseline of 53BP1⁺ foci/cell increases from 0.76 for non-reactive patients to 1.19 and 1.41 for normally reactive and over-responding individuals respectively. This new approach is particularly promising since it avoids ex vivo PBMC irradiation that introduces additional experimental variabilities and increases the time to obtain the prognosis. However, by using the spontaneous foci level as the metric, we lack information about the kinetics of 53BP1 recruitment because we do not know when the damage occurred. Therefore, background 53BP1⁺ foci represent a steady-state equilibrium between the spontaneous damage constantly generated in metabolically active cells and the efficiency of the repair machinery continuously fixing the damage. Higher spontaneous foci numbers in an individual might indicate inefficient DNA repair that might be overwhelmed following radiation exposure, leading to more severe side-effects. This hypothesis was consistent with our results showing that spontaneous 53BP1⁺ foci are inversely associated with 53BP1⁺ RIF levels in irradiated in vivo and ex vivo samples (59). As a potential confounding factor, it is important to note that radiation causes both direct and bystander, abscopal effects, which differ based on the organ that received therapeutic irradiation as well as on the tissue that was collected for quantification of radiation responses (79). For example, the measured changes in radiotherapy-induced foci in the blood immune cells after prostate cancer treatment may be representative of the abscopal effects of irradiation acting together with the changes in the immune system in response to cancerous and normal tissue injury.

In addition to the prognosis of therapeutic radiation-induced toxicity, predictive tools for radiation-induced carcinogenesis and other late biological consequences (stochastic effects) could be used to support human deep space exploration. Indeed, one of the predominant health concerns associated with deep space exploration missions is the continuous exposure of astronauts to galactic cosmic rays, which are a mixture of HZE particles (147). The estimated 1 Sv radiation dose calculated for a 3-year Mars mission is associated with various health risks including carcinogenesis and dysfunctions of cardiovascular, immune and central nervous systems (148–150). Since these impairments can partially be caused by unrepaired and mis-repaired DSBs that can result in mutations and complex genome rearrangements, RIF assay might constitute a useful predictive tool. In this context, studies have suggested to use foci baseline to predict a carcinogenesis endpoint (151,152), noting that higher spontaneous foci numbers may indicate inefficient DNA repair that promotes mis-repair and mutations responsible for cancer induction.

This finding supports the results of another study by Pariset et al. (56), who observed that the time of DNA repair in mouse fibroblasts exposed to X-rays ex vivo is positively correlated to cancer incidence of several mice tissues in vivo. However, equivalent human data regarding the link between spontaneous RIF and carcinogenesis are currently only obtained from correlation studies since it requires long follow-up (several decades) to observe radiation-related malignancies. Finally, further studies are needed in a full spaceflight environment or its simulation to study the potential synergistic effects of chronic low-dose rate of HZE irradiation with concomitant exposure to other stressors (e.g. microgravity, social isolation and circadian rhythm disruption).

In summary, as illustrated in Figure 5, spontaneous foci might be interpreted as the most comprehensive metric currently in use for predicting radiation effects since it takes into account both the ability to repair DNA (τ), a predictive indicator of stochastic effects, and the level of repaired damage (ρ), an indicator of toxicity in the study of deterministic effects. While current retrospective studies show the benefit of labelling spontaneous foci with 53BP1, similar investigations have to be performed with other DSB markers, such as γ-H2AX, and co-localization analysis of these markers should be performed in order to advance in a translation towards the clinic. The reason for the need of multiple markers is that although 53BP1 is recognized as a reliable marker of DSBs, its binding to the damaged site is regulated by ubiquitination of H2AX-type histones by RNF168 (153,154). Since the latter is mutated in some human populations, for example, in the RIDDLE (human radiosensitivity, immunodeficiency, dysmorphic features and learning difficulties) syndrome, the development of robust clinical markers will likely involve proteins other than 53BP1 and require further research (155,156).

Figure 5.

Schematic representation of the correlation between key DNA damage indicators and in vivo radiation response. The duration of repair (parameter τ) correlates to the amount of misrepaired damage constituting a predictive indicator of stochastic effects such as carcinogenesis risk. The fraction of repaired DNA damage (parameter ρ) is associated with cell toxicity and survival. These two indicators can be taken into account by using spontaneous foci (ρ₀) as a metric since it represents a steady-state equilibrium between the spontaneous damage constantly generated in metabolically active cells and the efficiency of the repair machinery continuously fixing the damage.

CONCLUSION

Since the discovery of γ-H2AX as nuclear foci detectable by immunostaining, an incredible revolution took place in the field of radiation biology. Back in the 1990s, NASA scientists only dreamed of being able to see where a galactic particle traversed a tissue. This is now a reality and scientists have shown incredible creativity in the usage of such biomarkers, from nuclear terrorism triage to clinical predictor and cancer companion diagnostic. However, as the years went by, the story became more complicated, as more proteins were discovered with similar characteristics but distinct roles in repair pathways and cellular functions. In this manuscript, we showed the benefit of using imaging-based analysis of DNA repair foci as a metric of DNA damage kinetics and associated biological effects of ionizing radiation. Selecting the appropriate target protein, understanding the imaging modality caveats and using the right mathematical models to interpret these data are all essential requirements to draw strong biological conclusions. This review underscores how challenging a ‘one size fits all’ approach to RIF is as a tool for radiobiology. It attempts to summarize the relevant information and suggests guidelines reflecting the past 20 years of knowledge accumulated on the topic.

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Cancer Online.

FUNDING

National Aeronautics and Space Administration award #NNJ16HP24I to S.V.C. (Principal Investigator - PI).

Conflict of interest statement. None declared.