Super-Resolution Radiation Biology: From Bio-Dosimetry towards Nano-Studies of DNA Repair Mechanisms

Past efforts in radiobiology, radio-biophysics, epidemiology and clinical research strongly contributed to the current understanding of ionizing radiation effects on biological materials like cells and tissues. It is well accepted that the most dangerous, radiation induced damages of DNA in the cell nucleus are double strand breaks, as their false rearrangements cause dysfunction and tumor cell proliferation. Therefore, cells have developed highly efficient and adapted ways to repair lesions of the DNA double strand. To better understand the mechanisms behind DNA strand repair, a variety of fluorescence microscopy based approaches are routinely used to study radiation responses at the organ, tissue and cellular level. Meanwhile, novel super-resolution fluorescence microscopy techniques have rapidly evolved and become powerful tools to study biological structures and bio-molecular (re-) arrangements at the nano-scale. In fact, recent investigations have increasingly demonstrated how super-resolution microscopy can be applied to the analysis of radiation damage induced chromatin arrangements and DNA repair protein recruitment in order to elucidate how spatial organization of damage sites and repair proteins contribute to the control of repair processes. In this chapter, we would like to start with some fundamental aspects of ionizing radiation, their impact on biological materials, and some standard radiobiology assays. We conclude by introducing the concept behind super-resolution radiobiology using single molecule localization microscopy (SMLM) and present promising results from recent studies that show an organized architecture of damage sites and their environment. Persistent homologies of repair clusters indicate a correlation between repair cluster topology and repair pathway at a given damage locus. This overview over recent investigations may motivate radiobiologists to consider chromatin architecture and spatial repair protein organization for the understanding of DNA repair processes.


Introduction
Past efforts in epidemiological (nuclear power industry, atomic bomb explosions, nuclear reactor accidents, etc.) and clinical (diagnostic imaging, radiation oncology, radiation therapy planning etc.) research strongly contributed to the current understanding of ionizing radiation effects on human organs, tissues, and cells [1,2]. In principle radiation biology is based on effects of instantaneous (10 −18 s) [3], stochastic damaging interactions of ionizing radiation with cells, a main target being the genetic material, i.e. chromatin in the cell nucleus [4]. In this context, radio-sensitivity and radio-resistance as opposing terms describe the extent of individual cellular susceptibility or 'response' upon radiation exposure which are highly dependent on physical (e.g., radiation type, dose, dose rate, etc.), chemical (e.g., hydroxyl radicals, etc.) and biological (e.g., developmental and proliferative state of the affected cell type) factors. As the overall organismal radiation response results from the entirety of all individual radiation responses on the single cell level, deeper understanding of the underlying, complex molecular mechanisms and dynamics of radiation induced DNA damaging and repair on the cellular level is highly relevant for fundamental and applied radiation biology (for review see [1,2,5]).
Hence, cytometric analyses based on fluorescence microscopy have become the method of choice to study damaging effects of ionizing radiation and DNA repair. This has contributed a lot to today's knowledge. However, conventional fluorescence microscopy is limited to average lateral resolutions around 200 nm laterally and 600 nm axially [6] and thus is limited to the bulk analysis of molecular cellular processes and structures. In parallel, super-resolution fluorescence microscopy techniques have rapidly evolved during the last few decades and turned out to be powerful tools to study cellular structures and molecular architectures on the nanoscale (for review see [6][7][8]). Methods based on stochastic reversible photobleaching [9-15] of single molecules called Single Molecule Localization Microscopy (SMLM) [16] reach effective resolutions down to 10 nm and have become popular among modern super-resolution imaging techniques as their realization is highly practical and straightforward using established specimen preparation methods of standard fluorescence microcopy [17]. As such resolutions allow the detection of single molecules, such as nucleosomes [18], proteins [19,20], receptors and junction proteins [21,22], or even single chromatin loops [23] etc., super-resolution microscopy opens new avenues for the research of radiation induced damaging and repair processes [5,24].
With this article, we attempt to introduce the novel SMLM approach to radiation biophysics and radiation biology. We start with a brief summary on the basics of ionizing radiation, induction of DNA damage and damage repair mechanisms, to follow up with some standard radiobiology analysis methods. We further provide an overview of the working principles of selected sub-diffraction microscopy techniques with a focus on SMLM. Finally, the successful application of localization microscopy in radiation biology research is demonstrated along examples of current works.

Effects of ionizing radiation on cells and cell nuclei
Ionizing radiation penetrates through material and deposits enough energy to ionize molecules or atoms by liberating electrons. The effects of ionizing radiation on biological materials are highly dependent on the dose, the dose rate and type of radiation. In living cells, ionizing radiation hits all kinds of biomolecules, such as desoxyribonuleic acids (DNAs), aminoacids (proteins), lipids (membranes), carbohydrates, etc. However, most harmful consequences to living organisms show damages inflicted to their genomic DNA, especially in the form of DNA double-strand breaks (DSBs) [25,26]. Especially follow-up effects of false strand repair may lead to significant dysfunctional development as for instance tumor genesis.

Ionizing radiation
Ionizing radiation (IR) includes all high energy/speed (> 1% speed of light) ions (e.g. carbon ions), atom nuclei (e.g. alpha particles), subatomic particles (e.g. beta particles, protons or neutrons) and high-energy electromagnetic waves (e.g. high energy ultraviolet (UV) rays, X-rays and gamma rays), that carry enough energy to directly or indirectly ionize atoms or molecules by liberating electrons from them, and to break molecular bonds [27].
The most common types of ionizing radiation occurring under environmental circumstances are caused by radioactive decay and can be divided into three groups: alpha, beta and gamma radiation [27]. Alpha radiation is made up of particles comprising two protons and two neutrons (helium nucleus) that carry energies in the range of up to several MeV. Due to its large particle size, alpha radiation has the lowest penetration depth through biological materials and the highest energy deposition per distance traveled. Beta particles are made up of electrons or positrons, thus exerting higher penetration depths and lower energy depositions compared to alpha particles. Gamma rays are high energy electromagnetic waves that exhibit the highest penetration and compared to particles, lowest energy deposition per track unit in biomaterials among these three types of IR. Due to the dispersed and low energy deposition in tissue, gamma and beta radiation are often referred to as low linear energy transfer (LET) radiation, whereas alpha particles belong to high LET radiation [27].
For clinical diagnosis and therapy in radiology or radiation oncology [28, 29], typically artificial radiation sources are applied, as for instance to produce X-rays in the energy range of keV to MeV, electrons and positrons, protons, and heavy ions (carbon or nitrogen). Like alpha particle, protons and heavy ions belong to high LET radiation. The advantage of protons and especially heavy ions is based on the characteristic absorbance with a Bragg peak at the end of the particle track where most of the particle energy is deposited. This energy positioning peak can exactly be localized in the tumor volume so that intact cells and tissues in the tumor surroundings are excluded from radiation damaging [30].

Dose measures
The absorbed dose D of ionizing radiation is quantified by the amount of energy deposited per unit mass of the penetrated material and is measured in units of Joule per kilogram (J/kg) or Gray (Gy) [27]. It describes an universal energy absorption for all types of ionizing radiation and is most commonly used in radiophysical research, whereas a radiation type specific dose also called the equivalent dose H calculated by multiplication with a weighting factor W R (e.g. W R = 1 for gamma radiation and W R = 20 for alpha radiation) is often used in radio-biology, radio-medicine, or radiation protection and safety. The equivalent dose can be further weighted by a tissue weighting factor W T to result in the effective dose E, which describes radiobiological effects considering the used radiation type and the tissue/organ of interest. Both, the equivalent dose and effective dose are quantified in units of Sievert (Sv) and do not represent physically measurable quantities but rather a value based on clinical and epidemiological outcome that is typically used in radiation safety [31].

Damages of DNA induced by ionizing radiation
Among all kinds of ionizing radiation induced biological effects, damages to chromatin especially the DNA molecules in the nucleus of cells are thought to be the DNA -Damages and Repair Mechanisms most severe with respect to cellular survival and carcinogenesis [2,5,32,33]. DNA base oxidation, single strand breaks (SSBs) and double strand breaks (DSBs) are the most common ionizing radiation induced damages to the DNA molecule, that affect genome integrity and DNA biochemistry [34].
DSBs of DNA belong to the most complex and severe types of DNA damages as they directly affect genome integrity and the way of cellular survival [35][36][37]. Single strand breaks (SSBs) induced by ionizing radiation and base damages occur more frequently than double strand breaks [34]. It can be estimated to about 40 DSBs/Gy and about 1,000 SSBs/Gy. SSBs are less severe to genome integrity as an intact template strand is still available for complementarity-aided, error-free repair of the lesion. But DSBs are also simply formed by two or more opposing SSBs in close proximity or combinations of different DNA damage types [26].
Induction of DSBs in native chromatin is rapidly followed up by phosphorylation of nearby histones of the H2A variant H2AX at serine residues at position 139 [38]. This results in the generation of plenty γH2AX molecules around a DSB damage site, where about 2 Mbp of DNA are usually phosphorylated [39]. This leads to the formation of focus structures of sizes in the range of micrometers, which can be visualized under a fluorescence microscope [40]. These phosphorylated histones serve as signal and anchor points for many downstream recruited proteins of certain DNA damage response and repair machineries [41]. As the number of γH2AX foci is quantitative for DNA damage, counting of specifically labeled foci has been established as a measure for dose-efficiency and correlated to cell survival [42].
Single ionizing radiation induced DNA lesions can be caused by direct or indirect hits [43]. Ionizing radiation penetrating through a cell nucleus can hit and ionize atoms in a DNA molecule itself with a certain probability. However, the most prominent primary reaction underlying all ionizing radiation induced DNA damages is the radiolysis mediated formation of reactive oxygen species (ROS), e.g. •OH radicals, O2•-radicals and H2O2, which can further inflict reducing damage and thus lesion to the DNA [44]. Ionizing radiation, especially high LET radiation, is known for its property to efficiently induce highly complex damages to DNA. Such complex DNA damage sites composed of multiple lesions in close proximity on both strands are also termed locally multiple damage sites (LMDS) [45].

DNA double strand break repair mechanisms
Living organisms developed highly efficient and customized ways to repair the severe damages inflicted to their genome. The DNA DSB sites are rapidly (within seconds to minutes) recognized and marked by proteins of an initial response, which serve as signals and docking sites for more specialized proteins of DNA repair pathways. The fate of repair type depends on the concerted presence of pathway specific damage response proteins [1, 2, [46][47][48]. The main two ways by which cells respond to DNA double-strand breaks are non-homologous end joining (NHEJ; also called canonical NHEJ = cNHEJ) and homologous recombination (HR). NHEJ mediated DSB repair is fast and can be error-prone, but it can be flexibly performed throughout all cell cycle phases. HR works error-free, but is mostly restricted to late S and G2 phases as a homologous sister chromatid is required as a repair template [49][50][51][52]. Recent data, however, have suggested that active genes may employ HR also in G1 phase, by utilizing the nascent RNA as a template for precise repair (reviewed in [53]). As the DNA-end resection is inhibited in G1 cells, an alternative model with cNHEJ taking the advantage of the same principle (RNA-templated repair) has also been proposed. Interestingly, DNA repair by HR is preferred in lower eukaryotic life forms, whereas NHEJ is predominantly observed in mammalian organisms. Alternative low abundant DSB repair pathways are the alternative end joining pathways (a-NHEJ; also called back-up EJ), micro-homology mediated EJ (MMEJ) and single strand annealing (SSA). One main difference between all DSB repair mechanisms is the extent of initial DNA end-resection at the damage site [26, [54][55][56][57][58][59]. The DNA damage response (DDR) against DSBs is subject to intensive radiobiological investigation and fluorescence microscopy of in situ DSB repair proteins serves as state of the art biological dosimetry.

The initial response
After the induction of a DSB, damage response proteins are rapidly recruited and accurately determine the fate of the DSB towards a repair pathway that best deals with the damage site in a certain genomic and cellular context. The chromatin remodeling p53-binding protein (53BP1) protects the break site from extensive end resection [60], thereby promoting repair by non-homologous end joining [61], whereas BRCA1 facilitates extensive end resection for repair by homologous recombination [52,62].

Non-homologous end joining
The NHEJ repair pathway is initiated with binding of the Ku70-Ku80 heterodimer complex to the DNA ends of the DSB site, which serves as a linkage between damage site and further damage response proteins [61,63]. In a second step, DNA-dependent protein kinase catalytic subunit (DNA-PKcs) is recruited to the Ku complex forming the DNA-PK complex. On-going recruitment of X-ray complex (XRCC4)/DNA Ligase IV (X4LIG4) complex and XLF to the DNA-PK complex forms the core NHEJ complex [64]. DNA-PKcs sterically protects the break site for repair and phosphorylates other repair proteins [65,66] and H2AX [41]. Furthermore, DNA-PK auto-phosphorylation results in a conformational change of the core complex, thereby enabling DNA end processing by nucleases and dissociation of the DNA-PKcs subunit [67,68]. Finally, ligation of the DNA ends is mediated by the X-ray complex (XRCC4)/DNA Ligase IV (X4LIG4) complex and XLF [69][70][71]. Artemis endonuclease [72,73], polynucleotide kinase (PNK) [74], DNA polymerase (pol) μ and λ can be additionally involved in NHEJ repair depending on the chemical properties of the DNA damage site [75].

Homologous recombination
To initiate repair by HR, the free damaged DNA ends at the DSB site must be sensed and bound by a protein complex comprised of MRE11, RAD50 and NBS1 (MRN complex) [76]. Next, the protein kinase Ataxia Telangiectasia Mutated (ATM) [77] is recruited to the MRN complex at the damage site [78], which autophosphorylates and phosphorylates components of the neighboring chromatin. Most prominent phosphorylations are those of the histone variant H2AX (γH2AX), one of the earliest and a very sensitive marker of cellular response to DSBs [38]. End resection is initiated by the single-strand endonuclease and exonuclease activity of the Mre11 protein [52,79] of the MRN complex. RAD50 further stimulates Mre11 nuclease activity and Nbs1 interacts with CtIP [80], another protein that is essential for the initiation of MRN complex mediated end resection [81]. Exonuclease 1 (Exo1) and Dna1/BLM are recruited by CtIP to continue end resection [82][83][84] until it gets attenuated by RPA coating of resected ssDNA ends [85]. BRCA2 in combination with BRCA1 and PALB2 dismantles the ssDNA ends from RPA coats enabling binding and forming of the RAD51 nucleoprotein filament, which stimulates homology search and strand invasion [86]. Sister chromatid strand recombination via Holiday junctions is further facilitated by RAD54A and its paralog RAD54B [87,88], finally resulting in conservative repair of the DNA lesion.

Alternative repair pathways
a-NHEJ or b-NHEJ has been described in slightly different ways which are not well distinguished [56][57][58]. Mostly, in the presence of short micro-homologies (>4 bp) after CtIP-MRN mediated end resection, repair via an alternative end joining (MMEJ) can take place [89]. This is initiated by Poly(ADP-ribose) polymerase 1 (PARP1) and followed up by DNA polymerase θ (pol θ) mediated strand extension starting at the paired micro-homology site. Ligase1 and Ligase2 are supposed to perform the final ligation of DNA ends [90,91].
When the damage site is flanked by larger regions with non-allelic sequence homologies, repair by single-strand annealing is also possible. The absence of Ku proteins and even more extensive end resection to expose the homologous regions as single strands are necessary for SSA repair [92]. Again, RAP binding to the resected ends promotes RAD52 mediated annealing of homologous regions. Nuclease XPF-ERCC1 trims the remaining non-homologous overhangs and DNA Ligase 1 connects the DNA ends [93].
Several studies indicate, that damaged genomic Alu elements use microhomologies for single-strand annealing, thereby often leading to translocations [94,95]. Such nonconventional damage repair processes might explain a significant portion of the observed deletion events associated with malignancies [59]. In fact, in vitro model systems could already demonstrate Alu mediate non-allelic homology dependent DSB repair [96].

Radio-sensitivity and biological dosimetry
Radio-sensitivity can be assessed on different scales ranging from whole organs and tissues over single cells to molecular markers and mechanisms. Based on clinical and experimental findings on DNA damage induction and response mechanisms, the dose effect of ionizing radiation on biological material is commonly described with a linear-quadratic model [97,98]. At low dose ranges (below 1 Gy), radiation damages are supposed to linearly increase with the applied dose, whereas at higher doses the probability for multiple hits increases and complex DNA damage spots dominate.
In this chapter, we summarize some established methods to study cellular and molecular effects of ionizing radiation. In the past, sophisticated assays were developed to detect and quantify radiation induced damages to the cell and nuclear DNA ranging from techniques to assess overall cell survivability, large-scale chromosomal damages and rearrangements over sensitive detection of DNA break sites to modern state of the art technologies that can visualize the formation of damage foci in situ with the help of suitable biomarkers. a. Colony forming assays (CFAs) based on clonogenic survival (also called clonogenic assay) (see [99] and citations therin) as a method to quantify cell survival after radiation exposure was firstly described by Puck and Marcus in 1956 [100]. CFAs measure the ability of cells to divide after treatment with agents that impair cellular reproduction, e.g. radiation (Figure 1). Since then, colony forming assays were improved for many different cell types and are widely used as a "gold standard" in radiobiological studies. In practice, irradiated cells are plated at higher dilutions so that single cells are well separated. Upon incubation colonies form each originating from a single cell. Thereby, colonies comprising 50 cells or more are considered for estimating the survival fraction. Treatments such as exposure to ionizing radiation damages the reproductive survivability of cells and thus results in a lower number of colony formation events at the same number of plated cells [103].
b. A fraction of ionizing radiation induced DNA double strand breaks results in heavy genomic rearrangements that can be detected on metaphase chromosomes. False rearrangements of multiple centromeric regions between chromosomes can lead to dicentric, acentric, centric ring conformations [104,105] that can be visualized under a conventional fluorescence microscope (Figure 2a). The good reproducibility and comparability lets the so called dicentric assay stand among the gold standards of biological dosimetry [106]. Nowadays dicentric assays are further developed towards biological dosimetry in the low dose range (< 500 mGy).
c. The micronucleus test is a method to assess and detect chromosomal breakages in interphase nuclei developed by Schmid et al. in 1975 [107]. Radiation damage can result in major chromosomal aberrations and loss on the centromeric region by wrong rearrangement of DNA double strand breaks (Figure 2a). These heavily damaged acentric chromosomes can form separated globular structures outside the main nucleus in interphase. As micronucleus formation can be readily detected in interphase nuclei, analysis can be performed much faster and serves as an efficient alternative for the analysis of instable chromosome aberrations [108].
d. In 1984, Ostling and Johanson published a micro-electrophoresis technique that could visualize DNA damages in single cells [109]. First, cells are embedded in agarose and lysed with non-ionic detergents under high salt concentrations, so that only nucleoids (supercoiled DNA loops attached to the nuclear matrix) remain. Ionizing radiation-induced breaks relax and locally unwind the supercoiled DNA structure, thereby partly linearizing the strand at the break site. When voltage is applied linearized DNA segments (SSB) protrude from the nucleoid and migrate faster towards the anode while the nucleoid core remains assembled. The nucleoid and its tail resemble a comet, when stained with 4′,6-diamidino-2-phenylindole (DAPI) or other quantitative DNA fluorescence dyes and visualized under the fluorescence microscope, thus leading to the term 'comet assay' [110,111]. Alkaline variations of the comet assay  [101,102]. [110] were introduced, that can detect DNA damages over an extended dose range (0.25 Gy to 2 Gy) than under neutral pH conditions (1 Gy -3 Gy) [111].
Modern approaches extend the method by automatization of experimental procedures and image analysis [112][113][114], thereby enabling statistically robust high-throughput detection of DNA damages for potential clinical applications.
e. DNA damage response proteins like γH2AX, 53BP1, RAD51 etc. accumulate at initial damage sites and rapidly form foci-like structures in the nucleus (see for example Figure 2b). Antibody staining and fluorescence microscopy of such damage response proteins is an established tool to visualize and quantify DNA damage repair foci at single cell resolution. One advantage of this technique is the ability to assess molecular dynamics of DNA damage repair by visual observation of foci formation at different time points after irradiation.

Super-resolution radiation biology
Fluorescence microscopy of potent marker labels is a powerful analysis tool to assess cellular effects of ionizing radiation on the single cell level by optical examination. Due to past efforts, a myriad of fluorescent probes exists for the molecular labeling of almost any known biological target structure (e.g. specific antibodies against γH2AX, 53BP1, MRE11, RAD51 or other repair proteins as well as against heterochromatin or euchromatin etc.). This opens the door to analyze molecular mechanisms underlying fundamental biological functions by optical investigation, e.g. DNA damage response and repair dynamics upon ionizing radiation exposure [1, 2, 5].

Super-resolution microscopy
A variety of novel super-resolution microscopy techniques were invented in the last few decades [115]. With the help of novel super-resolution microscopy techniques, the molecular effects of ionizing radiation in single cells can be studied on the nanoscale. Nano-labeled molecular structures can be resolved in biological specimens down to a precision of 10 nm (1/50 of the wavelength of visible light), which is in the range of single nucleosomes, antibodies, receptors, etc. (see for example [20,21]).
In order to improve the resolution in light microscopy, a prerequisite is to circumvent the diffraction limit of light, a physical phenomenon firstly described by  Ernst Karl Abbe and John William Strutt, 3rd Baron Rayleigh, during the late 19th and early 20th century [115]. In diffraction limited fluorescence/light microscopy, the Abbe or Rayleigh criterion (Figure 3) is commonly used to define a resolution measure describing the minimal distance D between two point-like light sources with wavelength λ that can be resolved: Therein λ is the wavelength and NA the Numerical Aperture of the objective lens (NA = n sin(α/2); n = refraction index; α = lens aperture angle). Conventional fluorescence microscopy techniques that use objective lenses with high numerical aperture NA (≥ 1.4) are available today. In confocal laser scanning microscopes, they typically achieve resolutions down to 200 nm in lateral and 600 nm in axial direction. However, modern super-resolution microscopy using the same objective lenses circumvent this physical limit by sophisticated interaction with fluorescence signals so that they can visualize biological specimen down to resolutions in the order of 10 nm, which is in the range of single nucleosomes, antibodies, receptors, etc. [16].
A complete overview of super-resolution microscopy techniques is beyond the aim of this article. However, we want to mention some meanwhile very well established ones: Sophisticated near-field super-resolution methods, e.g. total internal reflection fluorescence (TIRF) microscopy (TIRFM) [118,119] or near-field scanning optical microscopy (SNOM, NSOM) [120,121], belong to the first techniques breaking the diffraction limit by novel techniques working in the optical near field of fine crystal tips probing the specimen without an microscope objective lens. Unfortunately, near-field techniques are technically restricted to the visualization of surfaces of cells, membranes or isolated organelles [122][123][124].
More recently evolved far-field super-resolution fluorescence microscopy techniques use objective lenses available from establishes microscope manufactures and can be separated into two principle approaches. The first is based on the spatially modulated excitation of fluorophores, e.g. by point spread function engineering as in stimulated emission depletion (STED) [125] or by excitation through a series of illumination patterns as in structured illumination microscopy (SIM) [126]. A second group of super-resolution techniques is based on optical isolation of fluorescent molecules through switchable intensities [17] or intrinsic differences in spectral signatures [127]. The latter techniques often referred to as single molecule localization microscopy (SMLM) in general, can be practically implemented with customary microscope parts and standard objective lenses [16]. Spectral precision distance microscopy (SPDM) an early development of the 1990s [128] is the one and only localization microscopy method, that establishes optical isolation of molecular labels through constant differences in absorption and emission spectra of different fluorophores, that are applied in a combinatory labeling strategy [127,129]. Most localization microscopy methods, however, rely on stochastic spectral modulations of single fluorophore molecules, such as photo-activated localization microscopy (PALM) [12], fluorescence PALM (FPALM) [13], stochastic optical reconstruction microscopy (STORM) [15,130], direct STORM (dSTORM) [131], ground state depletion microscopy followed by individual molecule return (GSDIM) [132], SPDM with physically modifiable fluorophores (SPDM phymod ) [14, 17], etc. In the following chapters, we will describe single molecule localization microscopy in more details as being applied in radiation biophysics and we will provide examples indicating wide applications in nano-probing biomolecules and molecular mechanisms.

Single molecule localization microscopy for radiation biophysics
Single molecule localization microscopy is one the most popular superresolution techniques, because it can be practically realized with standard optical setup and standard specimen preparation methods using commercially available fluorophore labels. Fundamental to all SMLM techniques is the stochastic sampling of signals. The intrinsic blinking nature of a variety of available fluorophores at excitation with high laser powers (in the range of several kW/cm 2 ), enables SMLM with conventional dyes like GFP / YFP, Alexa488, Alexa568, etc. [16,17,133].
Apart from conventional fluorescence, which is based on rapid, repetitive excitation (10-15 s) and red-shifted emission (10-9 s) between the ground state S 0 and excited singlet state S 1 , fluorescent molecules additionally undergo intersystem crossing (ISC) [134] from S 1 to dark triplet states T 1 [135]. From there, fast  [116,117]. relaxation (10-3 s) to the original S 0 ground state enables re-entry to new cycles of normal fluorescence (Figure 3E, left). Further transition from the T 1 state into a second dark state D also occurs, which takes longer (ms to min) to recover to the ground state S 0 (Figure 3E, right) [116,136]. This reversible photobleaching via the long lived dark state D results in a limited number of stochastically blinking fluorochromes at resolvable time scales that can be used in single molecule localization microscopy to determine sub-diffraction positions of single fluorescing molecules [133].
For image acquisition, a time series of raw diffraction limited images (several hundreds to a few thousand) from the same region of interest are registered and efficiently searched for blinking events under a user-defined intensity threshold to discriminate signals from background. Then, the intensity profile of each blinking event is fitted by a Gaussian curve and the barycenter point of the signal source is calculated. Notably, the localization precision of such a point merely depends on its intensity/background ratio [137].
A major advantage of SMLM approaches lies in the data format. The point matrix containing the lateral x and y coordinates of each localization signal allows all kinds of mathematical and statistical analyses (Figure 4). Most prominent are analyses based on Ripley's point-to-point distance information which can be used for the elucidation of signal densities, cluster formation, and spatial organization of labels [139]. Recently, novel mathematical approaches like persistent homology determinations were introduced to investigate topological similarities [138]. Computation of the coordinate matrix into an image with user-defined resolution and visual enhancements is then possible. If provided, multi-color analyses on the single molecule level can be performed to study more complex molecular mechanisms and dynamics.

γH2AX clustering and chromatin arrangements at DNA damage sites
Phosphorylated histone variant γH2AX molecules at the site of DSBs and their accumulation into γH2AX-foci are well-established markers of DNA damage response and repair. Most recent studies performed SMLM of γH2AX specific antibodies in HeLa cells that were exposed to different doses of γ-radiation and fixed at different time points after radiation exposure [140]. Quantitative analysis resulted in a linear quadratic increase in measured γH2AX localization signal points and cluster numbers with increasing doses of radiation exposure (Figure 5A, B). With increasing repair time, the number of γH2AX clusters decreases; thereby successfully demonstrating repair dynamics and cell recovery by γH2AX-cluster relaxation on the molecular level. As dose responses and molecular dynamics for γH2AX clusters and raw γH2AX signal points well correlate with past observations, this study can serve as a benchmark standard for future super-resolution radiobiology experiments.
Similar studies indicated that the γH2AX cluster size remained constant during repair also at later times post irradiation, i.e., at later times only the number of clusters reduced. This typical size was about 400 nm in diameter after photon irradiation and nearly independent from dose or the cell types analyzed [140,141]. For α-particle irradiation from radioactive decay [142], the γH2AX cluster size along the particle track was about 200-300 nm; this size could be also observed for γH2AX clusters induced by environmental stress as for instance the deficiency of folat during long time culturing [143].
After exposure to photon radiation (different doses and energies) the SMLM analysis of heterochromatin around γH2AX clusters using specific antibodies against H3K9me3 methylation sites, revealed a fast relaxation of the chromatin and slower re-condensation after finishing the repair processes [144]. The degree of  relaxation was independent of the dose which is in good relation to the equally sized γH2AX clusters [140,141]. In contrast the euchromatin density increased during repair followed by a decrease after finishing the repair processes [144]. However, in total the chromatin showed an increasing clustering during repair followed by a reduction of clusters dependent on the energy of the damaging photons (unpublished). In general it can be assumed that DNA damaging by ionizing radiation does not only induce a reorganization of chromatin at the damaged sites but may also induce long range chromatin rearrangements for repair processes. Whether such chromatin rearrangements are random or directed to improve repair protein recruitment will be subject of future investigations.

Clustering of repair proteins at DNA damage sites
Beyond γH2AX cluster formation, foci and sub-foci clusters of repair proteins were investigated after photon or particle irradiation [141,142,[145][146][147]. In the following, some typical examples are shown taken from ongoing projects:  [140] with permission from the Royal Society of Chemistry.
1. 53BP1 foci were investigated in differently radio-resistant cell types, the moderately radio-resistant neonatal human dermal fibroblast cell line (NHDF) and highly radio-resistant U87 glioblastoma cell line. Specimens of both cell types were exposed to high-LET 15 N-ion radiation of doses of 1.3 Gy (in a 10° irradiation scheme) and 4.0 Gy (in a 90° irradiation scheme) at the particle irradiation facility of the Joint Institute for Nuclear Research, Dubna, Russia [145,146].
At given time points up to 24 h post irradiation, SMLM of fluorescently tagged 53BP1 molecules was performed and the coordinate data of each labeled molecule were quantitatively evaluated [137,139,140]. Clusters of these tags were determined as sub-units of repair foci (Figure 6a) and the formation and relaxation of these clusters revealed a higher ratio of 53BP1 proteins being recruited into clusters in NHDF cells (less radio-resistant) as compared to U87 cells (more radio-resistant) with different levels of distribution prior to DNA damage induction. This relation of 53BP1 inside and outside particle track clusters (Figure 6b) remained different for both cell types during the repair time observed. This could be seen as a measure of the "just-in-time" availability of 53BP1 proteins but did not reflect the absolute number of 53BP1 proteins available. The speed of cluster formation and relaxation differed for the two cell types (Figure 6c) indicating the recruitment of the existing proteins in the cell nucleus (higher in U87 cells) rather than a de novo production [147]. A certain number of the clusters remained persistent, even longer than 24 h post irradiation (Figure 6b); thereby the number of these remaining clusters varied in each cell line. The heavily damaged cell nuclei maintained repair activity in order to process the complex damage patterns caused by high-LET 15 N-radiation. This long-standing repair activity of 53BP1 proteins was shown in both cell types and the behavior of the cells could causatively be linked to the cell-type specific radio-resistance. The dynamics and cluster formation of tagged 53BP1 molecules showed that these clusters were embedded within a random distribution of points. After irradiation, a fast formation of 53BP1 clusters was observed (Figure 6c). During the early repair time of about 30 min -1 h after radiation exposure some clusters were dispersed while others persisted and the amount of randomly distributed proteins was growing. The latter clusters that were persistent did not disappear until the end of the repair period being studied (24 h).

Another study performed two-color SMLM of immunostained γH2AX and
Mre11 proteins [141] and revealed significantly delayed foci formation by Mre11 compared to γH2AX. While γH2AX clusters are already established at 30 min after radiation exposure (Figure 7, left), Mre11 is still ubiquitously distributed in the nucleus. Mre11 cluster formation is maximal at around 180 min after irradiation with significant association to γH2AX clusters (Figure 7, right).

Topological similarities of repair clusters
The reason to apply topological analyses is to record properties of point patterns, which are invariant under certain deformations of the object. Mathematically these deformations correspond to continuous transformations of the topological space defined by the structures. Here we have considered two properties, the number of "components" (explained below), which are independent from each other in such sense that connections between points only exist within the respective components   [146,147]. and the number of "holes" of the structures inside the components (explained below, Figure 8a). In algebraic topology, these properties are called the Betti numbers for zero and one -dimensional simplicial complexes [148].
SMLM images as for instance of γH2AX foci/clusters are point-sets for which components and holes can be defined. A geometric relationship among the points is defined by growing spheres of radius α around each of them. Whenever two spheres mutually embed each-other's center, these centers of the growing spheres are connected and the connected points belong to the same component. With increasing radii, the number of components is reducing. At the end of the procedure, a single component is remaining, the whole γH2AX cluster. For the definition of holes, a polygon is appropriate. Whenever the edges form a closed area, a hole is counted until another line closes a triangle separated from the original hole [147,148].
The results are presented as "barcodes" to track the formation and disappearance of components and holes with increasing α (Figure 8, left panel). These barcodes offer easy comparison of different sets of barcodes and their similarity can be calculated by the Jaccard index [149]. The Jaccard index results is a value between 0 and 1, where 0 is equal to no overlap of two bars and 1 describes two identical bars. Barcodes of different dimensions are defined as similar, if the averages of the individual similarity indices fulfill the Jaccard index conditions of similarity. Importantly, topological comparisons are independent of the scale so that it is possible to compare variably large clusters.
The barcode transfers the examined structures into a form of visualization that is scale invariant. The formation and dissolution of small scaled complexes is recorded alongside the lifetime of large scaled complexes. Consequently, for γH2AX clusters, the barcodes contain bars representing components and holes in the nanometer but also in the micrometer scale ranges. In Figure 8 (right panel) a representative result of a heat map of Jaccard Indices is shown for SkBr3 breast cancer cells. 200 heterochromatin associated γH2AX clusters and 200 non-heterochromatin associated γH2AX clusters were selected by determining those with the highest and lowest heterochromatic densities in the environment and examined according to their topological similarity. For the average similarity for components and holes, heterochromatin associated γH2AX clusters showed a clear similarity whereas nonheterochromatin associated γH2AX clusters did not. This means that by topological analysis the heterochromatin associated γH2AX clusters could be discriminated as those clusters of high topological similarity [138]. The proximity of γH2AX clusters to heterochromatin seems to have a significant measurable impact on its structure. Interestingly, the non-heterochromatin associated γH2AX clusters and heterochromatin associated γH2AX clusters were more similar than the non-heterochromatin associated γH2AX clusters themselves. It can be clearly seen that the proximity to heterochromatin influences the structure of the clusters.
For particle irradiated NHDF cells and U87 cells, the similarity values obtained by averaging of components and holes values for each 53BP1 cluster were determined and the clusters of the 10° irradiations scheme were compared. The Jaccard indices revealed values between 0.55 and 0.82 for U87 and NHDF cells. The broad frequency distribution did not show a peak for NHDF cells whereas for U87 cells a clear peak at 0.64 was found. If the clusters of the 90° irradiation scheme were compared, the peak was located at 0.63 for U87 cells. This value was the same, if the 10° with the 90° irradiation scheme was compared. For these two comparisons (90° vs. 90°, 10° vs. 90°), NHDF cells revealed a bimodal peak distribution where one peak was located at 0.67 and the other one at 0.72 (Figure 9). Thus, it can be concluded that in case of the more radio-sensitive NHDF cells a higher topological similarity in 53BP1 clustering was identified than the case the less radio-sensitive U87 cells.
SMLM of irradiated breast cancer cells stained by combinatorial fluorescence in situ hybridization (COMBO-FISH) [169,170] with a unique, short 17-mer oligonucleotide specific for genomic Alu elements (Figure 10A) resulted in a negative linear quadratic decline of the dose efficiency curve of localization signal points in the 0.5 Gy to 4 Gy dose range (Figure 10C) [170]. Furthermore, differential association of Alu signals with H3K9me3 heterochromatin between irradiated and non-irradiated cells could be revealed (Figure 10B). The heterochromatin relaxed after irradiation. However, the extension of this relaxation was independent of the dose. Alu dosimetry was also applied to the low dose range (< 0.5 Gy) (Figure 10D) [16], thereby opening new paths to study the molecular mechanisms underlying the controversial low dose radiation effects [171][172][173][174][175], which are difficult to assess due to a lack of appropriate biomarkers for the <0.5 Gy dose range [176].

Conclusion and perspectives
With this article, we have addressed scientists, researchers, and clinicians working in interdisciplinary fields, which are searching for a brief introduction to current radiobiology, its fundamental principles and methodologies. We would further like to have caught the attention of radiation biologists in laboratories, clinics, and industry by demonstrating novel super-resolution microscopy techniques that have the potential to drive radiobiology to a next generation. Single molecule localization allows geometrical and topological analyses on the meso-and nano-scale at the single-cell level in situ with the advantages of easy practice and the applicability to already existing experimental methods (e.g. immunostaining, FISH). As superresolution microscopy techniques are still not a wide-spread routine in molecular biology laboratories, the long history of fluorescence microscopy data from radiobiological studies provides a solid basis for validation. We have shown that radiobiology can be an application of SMLM based nanoscopy and its versatile data analysis method which allow the investigation of new perspectives of DNA damage induction and repair. It can even help to discover novel markers of biological dosimetry as demonstrated by our recent studies assessing dose dependent effects on retrotransposon Alu availability. Nano-scaled analysis of repair foci architecture and dynamics by assessing foci like 53BP1, Mre11, etc. will give further insight into the molecular mechanisms of DNA damage response and fate of repair pathway of individual damage sites in single cells. Indeed, evidence grows that nanostructure and function of chromatin are highly interdependent aspects that govern the fundamentals of molecular genetics, such as cell type differentiation, gene expression, DNA damage repair and reproduction. Thus, super-resolution radiobiology could serve as a general proof of principle for many other molecular biology applications in future. Finally, we believe that single-molecule localization microscopy will develop to a standard application of radiation biology and might even add to the repertoire of diagnostic technologies in clinical facilities in the future.