Locations of radiation-produced DNA double strand breaks along chromosomes: a stochastic cluster process formalism.

Ionizing radiation produces DNA double strand breaks (DSBs) in chromosomes. For densely ionizing radiation, the DSBs are not spaced randomly along a chromosome: recent data for size distributions of DNA fragments indicate break clustering on kbp-Mbp scales. Different DSB clusters on a chromosome are typically made by different, statistically independent, stochastically structured radiation tracks, and the average number of tracks involved can be small. We therefore model DSB positions along a chromosome as a stationary Poisson cluster process, i.e. a stochastic process consisting of secondary point processes whose locations are determined by a primary point process that is Poisson. Each secondary process represents a break cluster, typically consisting of 1-10 DSBs in a comparatively localized stochastic pattern determined by chromatin geometry and radiation track structure. Using this Poisson cluster process model, which we call the randomly located clusters (RLC) formalism, theorems are derived for how the DNA fragment-size distribution depends on radiation dose. The RLC dose-response relations become non-linear when the dose becomes so high that DSB clusters from different tracks overlap or adjoin closely. The RLC formalism generalizes previous models, fits current data adequately and facilitates mechanistically based extrapolations from high-dose experiments to the much lower doses of interest for most applications.     

Clusters of DNA double-strand breaks induced by different doses of nitrogen ions for various LETs: experimental measurements and theoretical analyses

The yields and clustering of DNA double-strand breaks (DSBs) were investigated in normal human skin fibroblasts exposed to gamma rays or to a wide range of doses of nitrogen ions with various linear energy transfers (LETs). Data obtained by pulsed-field gel electrophoresis on the dose and LET dependence of DNA fragmentation were analyzed with the randomly located clusters (RLC) formalism. The formalism considers stochastic clustering of DSBs along a chromosome due to chromatin structure, particle track structure, and multitrack action. The relative biological effectiveness (RBE) for the total DSB yield did not depend strongly on LET, but particles with higher LET produced higher fractions of small DNA fragments, corresponding in the formalism to an increase in the average number of DSBs per DSB cluster. The results are consistent with the idea that DSB clustering along chromosomes is what leads to large RBEs of high-LET radiations for major biological end points. At a given dose, large fragments are less affected by the variability in LET than small fragments, suggesting that the two free ends in large fragments are often produced by two different tracks. The formalism successfully described an extra increase in small DNA fragments as dose increases and a related decrease in large fragments, mainly due to interlacing of DSB clusters produced along a chromosome by different tracks, since interlacing cuts larger DNA fragments into smaller ones.     

Induction and rejoining of DNA double-strand breaks in normal human skin fibroblasts after exposure to radiation of different linear energy transfer: possible roles of track structure and chromatin organization

DNA double-strand breaks are nonrandomly induced by high-LET radiation. Differences in the induction and rejoining of DSBs after irradiation with ions having different LET were detected by fragment analysis. The data obtained indicate that the track structure of the traversing particle and its interaction with the different chromatin structures of the cellular DNA influence the yield as well as the distribution of the induced damage. The induction and rejoining of clustered DSBs induced by the same nitrogen ion fluence at LETs of 80-225 keV/microm were investigated by a detailed analysis of the DNA fragmentation patterns in normal human fibroblasts. The DSBs in the cells were allowed to rejoin during incubations for 0-20 h. Two separate pulsed-field gel electrophoresis protocols were used, optimized for separation of fragments in the size ranges 1-6 Mbp and 5 kbp-1.5 Mbp. A strong influence of LET on the level of DSB induction was evident. The DSB yield increased from 4.5 +/- 0.2 to 10.0 +/- 0.3 DSBs per particle traversal through the cell nucleus when LET increased from 80 to 225 keV/microm. Further, the size distribution of the DNA fragments showed a significant dependence on radiation quality, with an excess of fragments at 50-200 kbp and around 1 Mbp. Differences in repair kinetics were also evident, with slower rejoining for increasing LET, and the initial nonrandom fragment distributions were still present after 1 h of repair.     

Radiation breakage of DNA: a model based on random-walk chromatin structure

Monte Carlo computer software, called DNAbreak, has recently been developed to analyze observed non-random clustering of DNA double strand breaks in chromatin after exposure to densely ionizing radiation. The software models coarse-grained configurations of chromatin and radiation tracks, small-scale details being suppressed in order to obtain statistical results for larger scales, up to the size of a whole chromosome. We here give an analytic counterpart of the numerical model, useful for benchmarks, for elucidating the numerical results, for analyzing the assumptions of a more general but less mechanistic “randomly-located-clusters” formalism, and, potentially, for speeding up the calculations. The equations characterize multi-track DNA fragment-size distributions in terms of one-track action; an important step in extrapolating high-dose laboratory results to the much lower doses of main interest in environmental or occupational risk estimation. The approach can utilize the experimental information on DNA fragment-size distributions to draw inferences about large-scale chromatin geometry during cell-cycle interphase. Received: 2 March 2000 / Revised version: 2 February 2001 / Published online: 21 August 2001     

Spatial distribution of radiation-induced double-strand breaks in plasmid DNA as resolved by atomic force microscopy

Atomic force microscopy (AFM) has been used to directly visualize, size and compare the DNA fragments resulting from exposure to low- and high-LET radiation. Double-stranded pUC-19 plasmid ("naked") DNA samples were irradiated by electron-beam or reactor neutron fluxes with doses ranging from 0.9 to 10 kGy. AFM scanning in the tapping mode was used to image and measure the DNA fragment lengths (ranging from a few bp up to 2864 bp long). Double-strand break (DSB) distributions resulting from high-LET neutron and lower-LET electron irradiation revealed a distinct difference between the effects of these two types of radiation: Low-LET radiation-induced DSBs are distributed more uniformly along the DNA, whereas a much larger proportion of neutron-induced DSBs are distributed locally and densely. Furthermore, comparisons with predictions of a random DSB model of radiation damage show that neutron-induced DSBs deviate more from the model than do electron-induced DSBs. In summary, our high-resolution AFM measurements of radiation-induced DNA fragment-length distributions reveal an increased number of very short fragments and hence clustering of DSBs induced by the high-LET neutron radiation compared with low-LET electron radiation and a random DSB model prediction.     

Image-based modeling reveals dynamic redistribution of DNA damage into nuclear sub-domains

Several proteins involved in the response to DNA double strand breaks (DSB) form microscopically visible nuclear domains, or foci, after exposure to ionizing radiation. Radiation-induced foci (RIF) are believed to be located where DNA damage occurs. To test this assumption, we analyzed the spatial distribution of 53BP1, phosphorylated ATM, and γH2AX RIF in cells irradiated with high linear energy transfer (LET) radiation and low LET. Since energy is randomly deposited along high-LET particle paths, RIF along these paths should also be randomly distributed. The probability to induce DSB can be derived from DNA fragment data measured experimentally by pulsed-field gel electrophoresis. We used this probability in Monte Carlo simulations to predict DSB locations in synthetic nuclei geometrically described by a complete set of human chromosomes, taking into account microscope optics from real experiments. As expected, simulations produced DNA-weighted random (Poisson) distributions. In contrast, the distributions of RIF obtained as early as 5 min after exposure to high LET (1 GeV/amu Fe) were non-random. This deviation from the expected DNA-weighted random pattern can be further characterized by “relative DNA image measurements.” This novel imaging approach shows that RIF were located preferentially at the interface between high and low DNA density regions, and were more frequent than predicted in regions with lower DNA density. The same preferential nuclear location was also measured for RIF induced by 1 Gy of low-LET radiation. This deviation from random behavior was evident only 5 min after irradiation for phosphorylated ATM RIF, while γH2AX and 53BP1 RIF showed pronounced deviations up to 30 min after exposure. These data suggest that DNA damage–induced foci are restricted to certain regions of the nucleus of human epithelial cells. It is possible that DNA lesions are collected in these nuclear sub-domains for more efficient repair.     

DNA double-strand breaks in mammalian cells exposed to Á-rays and very heavy ions

The spatial distribution of DNA double-strand breaks (DSB) was assessed after treatment of mammalian cells (V79) with densely ionizing radiation. Cells were exposed to beams of heavy charged particles (calcium ions: 6.9 MeV/u, 2.1⋅10³ keV/μm; uranium ions: 9.0 MeV/u, 1.4⋅10⁴ keV/μm) at the linear accelerator UNILAC of GSI, Darmstadt. DNA was isolated in agarose plugs and subjected to pulsed-field gel electrophoresis under conditions that separated DNA fragments of size 50 kbp to 5 Mbp. The measured fragment distributions were compared to those obtained after γ-irradiation and were analyzed by means of a convolution and a deconvolution technique. In contrast to the finding for γ-radiation, the distributions produced by heavy ions do not correspond to the random breakage model. Their marked overdispersion and the observed excess of short fragments reflect spatial clustering of DSB that extends over large regions of the DNA, up to several mega base pairs (Mbp). At fluences of 0.75 and 1.5/μm², calcium ions produce nearly the same shape of fragment spectrum, merely with a difference in the amount of DNA entering the gel; this suggests that the DNA is fragmented by individual calcium ions. At a fluence of 0.8/μm² uranium ions produce a profile that is shifted to smaller fragment sizes in comparison to the profile obtained at a fluence of 0.4/μm²; this suggests cumulative action of two separate ions in the formation of fragments. These observations are not consistent with the expectation that the uranium ions, with their much larger LET, should be more likely to produce single particle action than the calcium ions. However, a consideration of the greater lateral extension of the tracks of the faster uranium ions explains the observed differences; it suggests that the DNA is closely coiled so that even DNA locations several Mbp apart are usually not separated by less than 0.1 or 0.2 μm. Received: 27 January 1998 / Accepted in revised form: 15 April 1998     

DNA fragmentation induced in human fibroblasts by accelerated (56)fe ions of differing energies

DNA fragmentation was studied in the fragment size range 0.023-5.7 Mbp after irradiation of human fibroblasts with iron-ion beams of four different energies, i.e., 200 MeV/nucleon, 500 MeV/nucleon, 1 GeV/nucleon and 5 GeV/nucleon, with gamma rays used as the reference radiation. The double-strand break (DSB) yield (and thus the RBE for DNA DSB induction) of the four iron-ion beams, which have LETs ranging from 135 to 442 keV/mum, does not vary greatly as a function of LET. As a consequence, the variation of the cross section for DSB induction mainly reflects the variation in LET. However, when the fragmentation spectra were analyzed with a simple theoretical tool that we recently introduced, the results showed that spatially correlated DSBs, which are absent after gamma irradiation, increased markedly with LET for the iron-ion beams. This occurred because iron ions produce DNA fragments smaller than 0.75 Mbp with a higher probability than gamma rays (a probability that increases with LET). These sizes include those expected from fragmentation of the chromatin loops with Mbp dimensions. This result does not exclude a correlation at distances smaller than the lower size analyzed here, i.e. 23 kbp. Moreover, the DSB correlation is dependent on dose, decreasing when dose increases; this can be explained with the argument that at increasing dose there is an increasing fraction of fragments produced by DSBs caused by separate, uncorrelated tracks.     

Biological imaging of heavy charged-particle tracks

The immunocytochemical response to DNA damage induced by low-energy bismuth and carbon ions was investigated in normal human fibroblasts. Inside the nuclei, the traversing charged particles lead to the accumulation of proteins related to DNA lesions and repair along the ion trajectories. Irradiation under a standard geometric setup with the beam direction perpendicular to the cell monolayer generates spots of these proteins as described previously for MRE11B (hMre11), CDKN1A (p21) and PCNA (Jakob et al., Int. J. Radiat. Biol. 78, 75-88, 2002). Here we present data obtained with a new irradiation geometry characterized by a small angle between the beam direction and the monolayer of cells. This new irradiation geometry leads to the formation of protein aggregates in the shape of streaks stretching over several micrometers in the x/y plane, thus facilitating the analysis of the fluorescence distributions along the particle trajectories. Measurements of fluorescence intensity along the ion tracks in double- and triple-stained samples revealed a strict spatial correlation for the occurrence of CDKN1A and MRE11B clusters. In addition, immunostained gamma-H2AX is used as a marker of double-strand breaks (DSBs) to visualize the localized induction of these lesions along the particle paths. A clear coincidence of CDKN1A and gamma-H2AX signals within the ion-induced streaks is observed. Also for PCNA, which mainly associates with lesions processed by excision repair, a strict colocalization with the MRE11B aggregations was found along the ion trajectories, despite the higher estimated yield of this type of lesions compared to DSBs. Strikingly similar patterns of protein clusters are generated not only for the various proteins studied but also using different ion species from carbon to bismuth, covering LET values ranging from about 300 to 13600 keV/microm and producing estimated DSB densities differing by a factor around 45. The patterns of protein clustering along the very heavy-ion trajectories appear far more heterogeneous than expected based on idealized DSB distributions arising from model calculations. The results suggest that additional factors like compaction or confined movement of chromatin are responsible for the observed clustering of proteins.     

DNA-incorporated 125I induces more than one double-strand break per decay in mammalian cells

The Auger-electron emitter 125I releases cascades of 20 electrons per decay that deposit a great amount of local energy, and for DNA-incorporated 125I, approximately one DNA double-strand break (DSB) is produced close to the decay site. To investigate the potential of 125I to induce additional DSBs within adjacent chromatin structures in mammalian cells, we applied DNA fragment-size analysis based on pulsed-field gel electrophoresis (PFGE) of hamster V79-379A cells exposed to DNA-incorporated 125IdU. After accumulation of decays at -70 degrees C in the presence of 10% DMSO, there was a non-random distribution of DNA fragments with an excess of fragments <0.5 Mbp and the measured yield was 1.6 DSBs/decay. However, since these experiments were performed under high scavenging conditions (DMSO) that reduce indirect effects, the yield in cells exposed to 125IdU under physiological conditions would most likely be even higher. In contrast, using a conventional low-resolution assay without measurement of smaller DNA fragments, the yield was close to one DSB/decay. We conclude that a large fraction of the DSBs induced by DNA-incorporated 125I are nonrandomly distributed and that significantly more than one DSB/decay is induced in an intact cell. Thus, in addition to DSBs produced close to the decay site, DSBs may also be induced within neighboring chromatin fibers, releasing smaller DNA fragments that are not detected by conventional DSB assays.     

Biophysical modelling of DNA DSB repair following high LET irradiation

One of the quantitative methods used in DNA repair research is a measurement of the size-distribution of DNA fragments at different times following cell irradiation. The aim of the present study was to evaluate the relationship between the experimentally observed size-distributions of DNA fragments and the parameters of doublestrand break (DSB) repair. A biophysical model of DNA DSB repair in chromosomal DNA including DSB clusters repair was proposed. Complex shapes of (1) DNA fragments distribution at different repair times, (2) rejoining kinetics for DNA fragments in different length intervals, (3) total fragments rejoining kinetics were simultaneously described with rates of DSB repair different for active/inactive chromatin compartments.     

Polymer chromosome models and Monte Carlo simulations of radiation breaking DNA

MOTIVATION: Chromatin breakage by ionizing radiation is relevant to studies of carcinogenesis, tumor radiotherapy, biodosimetry and molecular biology. This article focuses on computer analysis of chromosome irradiation in mammlian cells. METHODS: Polymer physics and Monte Carlo numerical methods are used to develop a coarse-grained computational approach. Chromatin is modeled as a random walk on a cubic lattice, and the radiation tracks hitting the chromatin are modeled as straight lines hitting lattice sites. Each track can make a cluster of DSBs on a chromosome. RESULTS: The results obtained replace conjectured DNA fragment-size distribution functions in the recently developed RLC formalism by more mechanistically motivated distributions. The discrete lattice algorithm reproduces features of current radiation experiments relevant to chromatin on large scales. It approximates the continuous formalism and experimental data with adequate precision. It was also found that assuming either fixed chromatin with correlations among different clusters of DSBs or moving chromatin with no such correlations gives virtually identical numerical predictions.     

Spatial distribution and yield of DNA double-strand breaks induced by 3-7 MeV helium ions in human fibroblasts

Accelerated helium ions with mean energies at the target location of 3-7 MeV were used to simulate alpha-particle radiation from radon daughters. The experimental setup and calibration procedure allowed determination of the helium-ion energy distribution and dose in the nuclei of irradiated cells. Using this system, the induction of DNA double-strand breaks and their spatial distributions along DNA were studied in irradiated human fibroblasts. It was found that the apparent number of double-strand breaks as measured by a standard pulsed-field gel assay (FAR assay) decreased with increasing LET in the range 67-120 keV/microm (corresponding to the energy of 7-3 MeV). On the other hand, the generation of small and intermediate-size DNA fragments (0.1-100 kbp) increased with LET, indicating an increased intratrack long-range clustering of breaks. The fragment size distribution was measured in several size classes down to the smallest class of 0.1-2 kbp. When the clustering was taken into account, the actual number of DNA double-strand breaks (separated by at least 0.1 kbp) could be calculated and was found to be in the range 0.010-0.012 breaks/Mbp Gy(-1). This is two- to threefold higher than the apparent yield obtained by the FAR assay. The measured yield of double-strand breaks as a function of LET is compared with theoretical Monte Carlo calculations that simulate the track structure of energy depositions from helium ions as they interact with the 30-nm chromatin fiber. When the calculation is performed to include fragments larger than 0.1 kbp (to correspond to the experimental measurements), there is good agreement between experiment and theory.     

Evidence for the organization of chromatin in megabase pair-sized loops arranged along a random walk path in the human G0/G1 interphase nucleus

We determined the folding of chromosomes in interphase nuclei by measuring the distance between points on the same chromosome. Over 25,000 measurements were made in G0/G1 nuclei between DNA sequences separated by 0.15-190 megabase pairs (Mbp) on three human chromosomes. The DNA sequences were specifically labeled by fluorescence in situ hybridization. The relationship between mean-square interphase distance and genomic separation has two linear phases, with a transition at approximately 2 Mbp. This biphasic relationship indicates the existence of two organizational levels at scales > 100 kbp. On one level, chromatin appears to be arranged in large loops several Mbp in size. Within each loop, chromatin is randomly folded. On the second level, specific loop-attachment sites are arranged to form a supple, backbonelike structure, which also shows characteristic random walk behavior. This random walk/giant loop model is the simplest model that fully describes the observed large-scale spatial relationships. Additional evidence for large loops comes from measurements among probes in Xq28, where interphase distance increases and then locally decreases with increasing genomic separation.     

Computational model of chromosome aberration yield induced by high- and low-LET radiation exposures

We present a computational model for calculating the yield of radiation-induced chromosomal aberrations in human cells based on a stochastic Monte Carlo approach and calibrated using the relative frequencies and distributions of chromosomal aberrations reported in the literature. A previously developed DNA-fragmentation model for high- and low-LET radiation called the NASARadiationTrackImage model was enhanced to simulate a stochastic process of the formation of chromosomal aberrations from DNA fragments. The current version of the model gives predictions of the yields and sizes of translocations, dicentrics, rings, and more complex-type aberrations formed in the G(0)/G(1) cell cycle phase during the first cell division after irradiation. As the model can predict smaller-sized deletions and rings (<3 Mbp) that are below the resolution limits of current cytogenetic analysis techniques, we present predictions of hypothesized small deletions that may be produced as a byproduct of properly repaired DNA double-strand breaks (DSB) by nonhomologous end-joining. Additionally, the model was used to scale chromosomal exchanges in two or three chromosomes that were obtained from whole-chromosome FISH painting analysis techniques to whole-genome equivalent values.     

Simulation of DNA damage after proton irradiation

The biophysical radiation track simulation model PARTRAC was improved by implementing new interaction cross sections for protons in water. Computer-simulated tracks of energy deposition events from protons and their secondary electrons were superimposed on a higher-order DNA target model describing the spatial coordinates of the whole genome inside a human cell. Induction of DNA double-strand breaks was simulated for proton irradiation with LET values between 1.6 and 70 keV/microm and various reference radiation qualities. The yield of DSBs after proton irradiation was found to rise continuously with increasing LET up to about 20 DSBs per Gbp and Gy, corresponding to an RBE up to 2.2. About half of this increase resulted from a higher yield of DSB clusters associated with small fragments below 10 kbp. Exclusion of experimentally unresolved multiple DSBs reduced the maximum DSB yield by 30% and shifted it to an LET of about 40 keV/microm. Simulated fragment size distributions deviated significantly from random breakage distributions over the whole size range after irradiation with protons with an LET above 10 keV/microm. Determination of DSB yields using equations derived for random breakage resulted in an underestimation by up to 20%. The inclusion of background fragments had only a minor influence on the distribution of the DNA fragments induced by radiation. Despite limited numerical agreement, the simulations reproduced the trends in proton-induced DNA DSBs and fragment induction found in recent experiments.     

Discrimination between genotoxicity and cytotoxicity for the induction of DNA double-strand breaks in cells treated with aldehydes and diepoxides

The time-dependent dose-response relationships for the induction of DNA double-strand breaks (DSB) assessed by pulsed-field gel electrophoresis (PFGE) and for viability (evaluated by the MTT cytotoxicity test) were investigated in order to discriminate between genotoxic and cytotoxic mechanisms of DNA fragmentation. Cultured human lung epithelial cells (A549) were treated (i) with the aldehydes formaldehyde or glutaraldehyde and (ii) with the DNA-DNA interstrand crosslinkers melphalan, diepoxybutane or diepoxyoctane. Induction of DSB by formaldehyde and glutaraldehyde was seen only after cell viability was reduced to less than about 60% of the control values, indicating that DSB were the consequence of extragenomic damage and viability loss. Melphalan, diepoxybutane and diepoxyoctane induced DSB by a genotoxic mode with concentrations that did not affect cell survival: 8 h after treatment initiation both heat-labile crosslinks and DSB could be detected. Cells were not able to repair the crosslinks induced by diepoxybutane, the crosslinker with the shortest chain length. In contrast, with melphalan and diepoxyoctane, which have a longer crosslinking property considerable repair of crosslinks was observed. The molecular size distribution of the produced DNA fragments supported this mechanistic distinction. The DNA fragments generated by diepoxides were initially large, their concentration decreasing monotonously from 7 Mbp to less than 1 Mbp and were converted to smaller fragments by 72 h in the course of cell death. In contrast, DNA fragments induced by formaldehyde peaked below 1 Mbp, implicating activation of DNA-degrading enzymes.     

Nanoscale analysis of clustered DNA damage after high-LET irradiation by quantitative electron microscopy – The heavy burden to repair

Low- and high-linear energy transfer (LET) ionising radiation are effective cancer therapies, but produce structurally different forms of DNA damage. Isolated DNA damage is repaired efficiently; however, clustered lesions may be more difficult to repair, and are considered as significant biological endpoints. We investigated the formation and repair of DNA double-strand breaks (DSBs) and clustered lesions in human fibroblasts after exposure to sparsely (low-LET; delivered by photons) and densely (high-LET; delivered by carbon ions) ionising radiation. DNA repair factors (pKu70, 53BP1, γH2AX, and pXRCC1) were detected using immunogold-labelling and electron microscopy, and spatiotemporal DNA damage patterns were analysed within the nuclear ultrastructure at the nanoscale level. By labelling activated Ku-heterodimers (pKu70) the number of DSBs was determined in electron-lucent euchromatin and electron-dense heterochromatin. Directly after low-LET exposure (5 min post-irradiation), single pKu70 dimers, which reflect isolated DSBs, were randomly distributed throughout the entire nucleus with a linear dose correlation up to 30 Gy. Most euchromatic DSBs were sensed and repaired within 40 min, whereas heterochromatic DSBs were processed with slower kinetics. Essentially all DNA lesions induced by low-LET irradiation were efficiently rejoined within 24 h post-irradiation. High-LET irradiation caused localised energy deposition within the particle tracks, and generated highly clustered DNA lesions with multiple DSBs in close proximity. The dimensions of these clustered lesions along the particle trajectories depended on the chromatin packing density, with huge DSB clusters predominantly localised in condensed heterochromatin. High-LET irradiation-induced clearly higher DSB yields than low-LET irradiation, with up to ∼500 DSBs per μm³ track volume, and large fractions of these heterochromatic DSBs remained unrepaired. Hence, the spacing and quantity of DSBs in clustered lesions influence DNA repair efficiency, and may determine the radiobiological outcome.     

Chromatin organization contributes to non-randomly distributed double-strand breaks after exposure to high-LET radiation

The influence of higher-order chromatin structure on the non-random distribution of DNA double-strand breaks induced by high-LET radiation was investigated. Five different chromatin structures (intact cells, condensed and decondensed chromatin, nucleoids and naked genomic DNA) from GM5758 cells or K562 cells were irradiated with (137)Cs gamma-ray photons and 125 keV/microm nitrogen ions (16-25 MeV/nucleon). DNA was purified with a modified lysis procedure to avoid release of heat-labile sites, and fragment size distributions and double-strand break yields were analyzed by different pulsed-field gel electrophoresis protocols. Whereas double-strand breaks in photon-irradiated cells were randomly distributed, irradiation of intact K562 cells with high-LET nitrogen ions produced an excess of non-randomly distributed DNA fragments 10 kb-1 Mbp in size. Complete removal of proteins eliminated this non-random component. There was a gradual increase in the yield of double-strand breaks for each chromatin decondensation step, and compared to intact cells, the yields for naked DNA (in buffer without scavengers) increased 83 and 25 times after photon and nitrogen-ion irradiation, respectively. The corresponding relative biological effectiveness decreased from 1.6-1.8 for intact cells to 0.49 for the naked DNA. We conclude that the organization of DNA into chromatin fiber and higher-order structures is responsible for the majority of non-randomly distributed double-strand breaks induced by high-LET radiation. However, our data suggest a complex interaction between track structure and chromatin organization over several levels.