Clustered DNA Damage Leads to Complex Genetic Changes in Irradiated Human Cells¹

The accumulation of mutations that permit unrestrained growth of somatic cells may lead to cancer (1). In human cancer cells, mutations that activate proto-oncogenes or inactivate tumor suppressors are frequently the result of major chromosomal alterations, such as translocations, inversions, large deletions, and partial or complete loss of chromosomes, generically termed gross chromosomal rearrangements (2, 3). These rearrangements may form at many different sites in the large genome of mammalian cells, particularly after damage by carcinogens, and have therefore been difficult to analyze at the molecular level.

Carcinogens vary in the types of damage they cause to DNA. It is now considered that approximately half of the effective dose to humans from ionizing radiations comes from radon gas, a natural decay product of uranium present in the earth’s surface. ²²²Rn decays with a half-life of 3.8 days, emitting 5.5 MeV α-particles that produce densely ionizing radiation tracks when traversing biological material. For example, a single α-particle track has ∼10,000 ionizations (equivalent to a dose of 150 mGy) when traversing a 12-μm human cell nucleus, whereas a γ-ray track has only ∼30 ionizations (dose, 0.4 mGy) in the same nucleus (4). Therefore, there is no such thing as a low dose of α-particles to the cell. Because of the density of ionization, α-particle interaction with DNA will cause heavily clustered sites of damage (5, 6), consisting of a mixture of closely spaced base damage and breaks in the sugar-phosphate backbone. Although it is calculated that domestic exposure to radon gas contributes significantly to cancer incidence (7, 8), little is known about the molecular consequences of such clustered DNA damage.

In particular, we wished to ask whether the clustering of DNA damage after irradiation has its counterpart in mutational response, i.e., complex molecular losses and rearrangements. This is not a trivial task, as noted above, because the possibility of loss or rearrangement of sequence over large genetic regions makes sequence-level analysis difficult, particularly in primary human material. We have used normal diploid human cells to study mutations caused by α-particles from a well-characterized source of ²³⁸Pu, which emits α-particles of initial energy (5.5 MeV) similar to that of radon. Early-passage human cells should respond in ways that most closely reflect the cells of the intact organism, and they are chromosomally stable. However, such cells have a limited life span, which can make it difficult to analyze fully the molecular changes in mutants. We have chosen the X-linked HPRT gene as the target for mutation induction because mutant selection methods are well established and it is relatively tolerant to large genetic changes (see, for example, Ref. 9). Additionally, coherent genomic sequence data are available for HPRT (10) and for some of the surrounding region of Xq26, facilitating mapping of large mutational changes.

HF12 primary male human fibroblast cells (11) at passage 3 were grown in Eagle’s minimum essential medium (Life Technologies), with 2 mm l-glutamine, antibiotics, and 10% FCS. Cells from passages 3–8, having a normal cytogenetic profile, were used in mutant selection experiments. Growth to determine cloning efficiencies and surviving fractions was optimized using feeder layers irradiated with 30-Gy X-rays and respread at 1000 cells/cm². The average cloning efficiency was 49.7 ± 13.3%.

To reduce the frequency of spontaneous mutants present, cell stocks were grown from low cell numbers (1000–5000 cells) to ∼10⁷ cells. Stocks were seeded onto specially constructed 30-mm dishes with very thin (2.5 μm) plastic bases (12). Cells were irradiated through the dish bases at 3 × 10⁵ cells/dish with ²³⁸Pu α-particles (range, ∼20 μm; peak energy, 3.26 MeV; linear energy transfer at the position of the cells, 121.4 keV/μm; Ref. 13). After irradiation, trypsinized cells from three dishes were pooled as a separately irradiated culture. In each experiment, 12–20 irradiated cultures were set up to isolate independently derived mutants. The immediate cell surviving fraction was determined by seeding 750 irradiated cells or 150 unirradiated cells into dishes with feeder cells and incubating these for 2 weeks to form colonies. In an initial experiment, cultures irradiated at either 25 or 50 cGy were pooled to determine mutant yield (mutants α1/25-1 and α1/50-4 came from this experiment). After this, independent mutants induced by 50 cGy were selected from 10 additional experiments. Spontaneous mutants were treated similarly but not irradiated.

Irradiated cells were grown nonselectively for a 7-day expression time and were then respread in dishes at 1000/cm² in 5 μg/ml 6-thioguanine for mutant selection (14) or at 150 per 9-cm dish without selection for cloning efficiency. Mutant frequency is expressed as the number of mutant clones detected divided by the number of cells seeded, corrected for cloning efficiency. Clones arising after 21–28 days in 6-thioguanine medium were isolated using micropipette tips. Picking one mutant per culture ensures independence, but sometimes more than one mutant was picked to offset losses attributable to senescence (in this case a mutant was considered as independently derived if it was found to have a molecular change in HPRT that was clearly different from those in mutants derived from the same culture). Otherwise, mutants were isolated randomly, without regard to colony size.

DNA was prepared from mass cell cultures or direct from cell colonies (15). PCR primer pairs were designed to cover the entire HPRT gene at ∼1-kb intervals (14); after breakpoints were localized, additional primers were designed to isolate and sequence the mutant junctions (not shown). The number of sites tested depended on the nature of the mutation; at least five sites were tested within the gene for each mutant. PCRs were duplicated and included an “outside marker” primer pair to detect genes on the short arm of the X-chromosome (either DMD or STS) to ensure that negative results were not attributable to failed reactions. The absence of PCR products indicated a genetic deletion, and all such reactions were repeated.

RAGE-PCR⁴ (16) was used to amplify unknown regions of genomic DNA associated with HPRT mutations. This method involved the separate digestion of plasmid (pUC18) and human genomic DNA with restriction enzyme, followed by mixing in equal quantities and ligation of the fragments to each other (at 50 μg/ml in 10 μl). Several different restriction enzymes were used to attempt to get products of a size that could be PCR-amplified using nested primers designed to both the plasmid and genomic DNA sequences; this reaction usually gave no visible products in the first set of cycles, but could yield products in the second or third nested sets for at least one restriction enzyme. PCR products were sequenced directly.

Deletion breakpoints were searched for sequence features such as repeats, palindromes, and topoisomerase recognition sites, as described previously (9). Human genomic databases were searched for alignments to unknown sequences at breakpoints by use of BLASTN (17). PCR primers were designed to screen a human monochromosomal library (United Kingdom Human Genome Resource Centre), to analyze the origin of the L1 repeat sequence of mutant α4/3-1, by maximizing the difference at their 3′ ends between this L1 repeat and others in the genomic sequence databases.

Primers were as described previously (14, 18). Sequence data for DXS53 and DXS177 and the plasmid carrying DXS311 were kindly supplied by Dr. J. A. Nicklas (University of Vermont, Burlington). The DXS311 fragment was isolated and sequenced by standard methods to allow primer design. DXS144 was used in Southern analysis as a 1-kb PstI fragment of the probe C11, supplied by Dr. J-L Mandel (CNRS, Strasbourg, France). The molecular distances and orientations cited in the text for these and other X-linked sequences were taken from data given in acedbX⁵ and Lippert et al. (19).

HF12 cells were exposed to α-particle irradiation as described above. Cells were exposed to colcemid (0.1 μg/ml for 4 h) and harvested at 3 and 15 days after irradiation. Chromosomes 1, 2, and 5 were painted with FITC-labeled chromosome-specific probes (Cambio) as described previously (20). To analyze chromosomal changes in α-particle-induced mutants, the chromosomes were initially painted with biotin-labeled total X-chromosome probe (Cambio). Subsequently, M-FISH 24-color chromosome-specific paints were used as described (SpectraVysion assay protocol; Vysis). Pretreatment consisted of RNase A (100 μg/ml) for 1 h at 37°C and pepsin (50 μg/ml in 10 mm HCl) for 2 min at 20°C.

HF12 cells grown on ultrathin plastic surfaces were irradiated with α-particles, and independently arising HPRT gene mutants were detected by their resistance to 6-thioguanine. An ∼7-fold increase in mutant frequency was found, compared with unirradiated cells (Table 1). Mutant clones varied widely in their growth capacity; clonal expansion after selection was sufficiently poor in approximately one-fourth of the mutants to preclude analysis of the HPRT gene mutation.

More than 100 induced mutants (representing a minimum of 64 independently induced mutants) were classified by the presence/absence of PCR products for sites over the 60-kb HPRT genomic region. Approximately half of these mutants had large genetic changes, detectable as losses of part or all of the HPRT gene (Table 1). The remainder were classified as point mutations because all sites tested gave the correct-sized PCR product. On the basis of the same criteria, all 11 spontaneous mutants (a minimum of 8 independent mutants) had point mutations.

It was possible to determine with some certainty the localization of the deletion breakpoints for a subset of the α-particle-induced mutants. First, approximately half of the deletions were found to have at least one breakpoint within the HPRT gene (Fig. 1, upper part). Second, where deletion breakpoints extended either proximally or distally into “unknown” sequences adjacent to HPRT, it was found that deletion breakpoints were distributed throughout the Xq26 region (Fig. 1, lower part). The deletion in one mutant, α9/9-1, extended beyond the most proximal and distal markers tested to give a deletion size of at least 3 Mb. Third, the chromosomes of mutants with partial or total HPRT deletions were examined with a total X-chromosome probe. None of the deletion mutants examined showed an X-chromosome rearrangement except for mutant α11/10-1, which showed a translocation located at approximately Xq26. Further analysis by M-FISH, in which each chromosome is colored differently, identified the translocation in mutant α11/10-1 as an X:9 reciprocal translocation and showed that this mutant had further translocations (Fig. 2 A). M-FISH was used to look at other mutants with large HPRT deletions, and in each case other gross chromosomal rearrangements, not involving the X-chromosome, were identified (data not shown).

Deletion mutants that had one or both breakpoints within the HPRT gene were chosen for detailed analysis. Where deletions extended into unknown sequences, we used RAGE-PCR on DNA from both mutant and wild-type cells (see “Materials and Methods”) to define the sequence on both sides of a breakpoint. In the following descriptions, the term “breakpoint” is used to indicate the position at which the break occurred in nonmutant DNA, whereas the term “junction” indicates the rejoined mutant sequence (i.e., there are two breakpoints and one junction in simple deletion mutants).

Mutant α1/50-4 had a simple internal HPRT deletion of 505 bp, including exon 4 (Fig. 1). A 2-bp (CT) direct repeat was present at the breakpoints (Fig. 3), and there was a strong consensus recognition site for DNA topoisomerase II (21) overlapping the upstream breakpoint (Fig. 3, vertical arrow). Mutant α1/25-1 had a deletion extending into unknown sequence upstream of HPRT (Fig. 1), but use of RAGE-PCR to identify the sequence in wild-type DNA showed again that a 2-bp (TG) repeat occurred at the breakpoints (Fig. 3). The downstream breakpoint (in the HPRT gene) was in an Alu repeat, but no corresponding repeat sequence occurred at the upstream breakpoint. Searches for the unknown sequence identified a 100-kb BAC clone (bWDX180) that overlapped with another fully sequenced BAC (bWDX187) carrying the HPRT gene; alignment of these two BAC sequences showed that a deletion of 75,164 bp had occurred in mutant α1/25-1.

Mutant α8/3-1 had a still larger deletion, shown by loss of DXS79 but retention of 299R (∼700 kb upstream; Fig. 1). Sequence of the mutant junction, followed by identification and sequencing of the breakpoints in wild-type DNA by use of RAGE-PCR showed that 3 bp of additional sequence (TTC) occurred at the mutant junction (Fig. 3). The upstream breakpoint occurred in an Alu repeat, but no repeat was found at the downstream breakpoint. Searches of human genomic sequence databases found an exact match for the unknown sequence to a partially sequenced BAC clone (RP11-274K13). This clone mapped to the X-chromosome but was at least 1 Mb downstream of HPRT (Fig. 4 A), indicating that the relative positions of the remaining HPRT sequence and the unknown sequence were inverted in the mutant sequence.

Two other mutants, α12/5-1 and α4/3-1, were initially considered to be simple internal HPRT deletions, each of ∼4 kb (Fig. 1), but PCR primers designed to amplify the HPRT sequence on either side of the mutant junction failed to give products. Use of RAGE-PCR revealed that the HPRT mutant breakpoints were now joined to unknown sequence on each side (giving at least four breakpoints and two junctions in each mutant). The simplest explanation for this finding was that unknown sequence had been inserted at the deletion site in HPRT.

Sequencing across the proximal junction of the α12/5-1 mutant gave unknown sequence, which was found to map to the X-chromosome at a position 3.4 Mb upstream of HPRT (defined by the PAC clone RP1-37M17; Fig. 4,A, upper part). The sequence TTTCC immediately adjacent to the proximal junction has been found previously at some radiation-induced mutant junctions (14, 22), and again the HPRT breakpoint occurred in an Alu repeat (Fig. 3). The HPRT breakpoint at the distal junction occurred in a Donehower repeat and had a 2-bp (CA) direct repeat with the unknown sequence. The unknown downstream breakpoint was within a L1 repeat, which has been mapped to Xq25–26.2.

The complexity of mutant junctions was even greater in the final mutant analyzed, α4/3-1. The unknown breakpoints at both proximal and distal junctions were linked to an X-chromosome sequence usually found 2.9 Mb upstream (defined by another recently sequenced PAC clone, RP1-119E23; Fig. 4,A, upper part). However, one of these upstream sequences was present in the mutant in the opposite orientation with respect to HPRT, indicating that a sequence inversion had occurred at this junction. To add to the complexity, the proximal junction had an insertion of unknown sequence between the RP1-119E23 sequence and HPRT. Searches revealed an exact match for the unknown sequence to part of a L1 repeat from a partially sequenced BAC clone (RP11-172E10) mapped to chromosome 8. Because spurious clone assignments can occur, the location of this sequence in nonmutant cells was confirmed by screening a monochromosomal library using PCR primers designed to detect uniquely the inserted sequence (Fig. 4,B). The proximal HPRT breakpoint was part of an Alu repeat, and unusually, a 7-bp sequence common to HPRT and the unknown sequence was present close to the junction (Fig. 3). At the distal breakpoints, 4-bp direct repeats (CATA) occurred, and these repeats were themselves within other short repeats (e.g., ATA/CATAC at the HPRT breakpoint).

Are the α-particle-induced mutations in the HPRT gene, which are mostly not visible with the light microscope, part of a spectrum of transmissible large genetic changes including those seen microscopically? The translocation in mutant α11/10-1 supports this idea, but to answer this question directly is technically difficult because it would require the molecular detection and sequencing of randomly induced sites of visible radiation-induced changes. To strengthen this association, we used FISH to determine whether similar types of visible genetic change (deletion, insertion, inversion) could be found in the chromosomes of large numbers of irradiated HF12 cells without selection for mutation. As seen in Table 2, at 3 days after irradiation, when most cells had recovered from radiation-induced division delay, a high frequency of unstable chromosomal changes (e.g., dicentrics and fragments) were present, but with time in culture their frequency declined sharply. Potentially stable changes, such as translocations, persisted at approximately the same frequency over 15 days of culture, and the occasional inversion was detectable at both 3 and 15 days after irradiation. Complex rearrangements, which by definition include insertions (Table 2), declined by half at 15 days because of the loss of unstable types. Significantly, given our molecular data, at 15 days almost all of these rearrangements were stable interchromosomal insertions of fragments, as depicted in Fig. 2 B.

We have shown that it is possible to carry out detailed molecular analysis of large and complex genetic changes induced by α-particles in the HPRT gene of human diploid fibroblasts, despite their limited life span. To complement these molecular data, we assessed genome-wide changes in the same irradiated primary cells by cytogenetic analysis to show that similar types of transmissible changes may be found, conceptually linking together molecular and visible genetic changes.

Our success in sequencing both sides of each breakpoint in these mutants allows comment on mechanisms of deletion and rearrangement formation. At the radiation dose used (0.5 Gy), knowing the average projected nuclear area of the HF12 human cells in this experimental situation (mode, 225 μm²; Ref. 20), approximately five α-particle tracks would occur in each nucleus traversed. These tracks will yield 30–40 DSBs/nucleus, but the DSBs will be distributed nonrandomly such that most will be part of a cluster of several (≥4) DSBs within megabase intervals, with a few much larger clusters (23). Additionally, each DSB will be associated with base damage in these localized sites. Repair of base damage within a heavily clustered damage site is likely to occur relatively slowly and can lead to further DSBs, either through inopportune endonuclease action at base damage on complementary DNA strands or through interaction of base damage with the replication machinery (24).

The rejoining of DSBs can be accomplished by repair pathways, including homologous recombination and nonhomologous end joining (reviewed in Refs. 25, 26). Although it has been established in principle that interchromosomal homologous recombination can give rise to translocations in mammalian cells (27), breakpoint sequence analysis of the present series of mutants gave little indication that interaction of homologous sequences was involved. Lengthy repeat sequences such as Alu occurred only at one breakpoint in any mutant and were sufficiently common for these to have been associated with the mutation site by chance. Similarly, the occurrence of short sequence motifs, indicating the involvement of specific enzymatic processes in the induction and/or processing of damage, was not common at breakpoints (Fig. 3). A DNA topoisomerase II cleavage site at the upstream breakpoint of mutant α1/50-4 may be significant because topoisomerase II has previously been implicated in deletion formation (28), potentially through enhanced cleavage of DNA at sites of base damage (29, 30). Mutant α8/3-1 alone had a short (3 bp) sequence insertion at the deletion junction and occurred distally to HPRT exon 6. Interestingly, small deletions with breakpoints in exon 6 have been found with inserted bases that are inverted complements of sequences at the breakpoints (31); these deletions commonly have the triplet TTC at the breakpoint.

It is significant that all of the mutants except α8/3-1 had short sequence repeats (microhomologies) of 2–5 bp present at one pair of breakpoints and that one copy of the repeat was lost in the mutant. The presence of microhomologies at sites of DSB rejoining has been observed in some previous studies with mammalian and yeast cells (32, 33) and in human cell extracts (34). Short inverted repeats may be involved similarly at sites of rejoining, where single-stranded “hairpin” structures are formed, which can be resolved by loss or inversion of the looped-out DNA (35, 36). It is possible to model some of the more complex junctions in these mutants by use of microhomologies as rejoining sites; for example, in mutant α4/3-1, it is striking that inverted repeat sequences (ACCAG/CTGGT) occurred at the proximal and distal HPRT breakpoints (Fig. 3). Starting with the annealing of these inverted repeats in single-stranded DNA, the structure of the α4/3-1 junctions is modeled in Fig. 5. The DNA in the “internal” loop, including HPRT exons 2 and 3, was replaced by external sequences that annealed at the sites of microhomology at or close to their ends, and the structure could be resolved by rejoining at the junctions.

It is not clear why microhomologies are used in end joining, because it can take place without any homology (DNA-protein kinase-dependent nonhomologous end joining; Ref. 37). One possibility would be that certain damage sites are difficult to repair by nonhomologous end joining and may require the activity of other repair proteins. Repair difficulties might arise by virtue of damage location, such as proximity to noncanonical DNA structures (38), or, in the present case, because of clustering of damage.

In addition to intrachromosomal complexity, mutant α4/3-1 carried an insertion into HPRT of a fragment from a L1 element from chromosome 8, showing that interchromosomal interactions are stimulated. We have supported this finding by showing that visible interchromosomal insertions are not uncommon many days after α-particle irradiation (Fig. 2,B; Table 2). Spontaneously occurring insertions are relatively rare events in the human genome; in the germ line these account for ∼1 in 600 mutations and mostly derive from endonuclease-dependent L1 element transposition (39). However, at sites of DNA breakage, truncated L1 insertions can occur by endonuclease-independent transposition, and these are accompanied by deletions of the target site (40). Significantly, where L1 insertions have been found in association with tumorigenesis, they show internal rearrangements, including deletions and inversions (41, 42). Similarly, in mutant α4/3-1, a truncated L1 sequence was inserted into HPRT alongside another X-linked sequence (identified by PAC 119E23), part of which had become inverted relative to the normal sequence. We think that it is unlikely that both parts of the 119E23 sequence inserted separately into HPRT, suggesting that the α4/3-1 rearrangement had undergone further rearrangement.

Recent experiments with both yeast and mammalian cells show that “capture” of DNA sequence to patch DSBs can make use of any linear end to rejoin break sites and that these insertions are derived from the most abundant sources of available ends (43, 44). The rarity of these simple insertions suggests first that end joining normally gives relatively precise ligation of DSBs with few sequence rearrangements. Second, it suggests that the patching of breaks with endogenous DNA fragments is limited by the availability of fragments. We suggest that when irradiation causes dense clusters of DNA damage, repair is inefficient (45) and many short DNA fragments are produced (46), providing more opportunity for patching to occur at damage sites. Therefore, an intriguing prediction of our analysis is that the captured fragments found in complex α-particle-induced mutants derive from collateral α-particle damage elsewhere in the genome. If this is true, then it may be possible to link rearrangements on different chromosomes and derive a molecular inventory of radiation-induced large genetic changes within a damaged cell.

Our data contrast strikingly with similar studies of large genetic rearrangements found after treatment with other DNA-breaking agents, such as bleomycin and neocarzinostatin. These chemically induced rearrangements were highly conservative, with almost no loss of sequence at the sites of DNA breakage (47, 48, 49). This comparison reinforces the conclusion that the heavy clustering of DNA damage from α-particle irradiation leads to the complex genetic changes we observed at the sequence level. Analysis at the visible (chromosomal) level will represent only the tip of the iceberg, considerably underestimating the extent and complexity of radiation-induced rearrangements. An important feature of these molecular changes is that they can be transmitted through mitotic divisions. At the sequence level, we see intra- and interchromosomal insertions and inversions in association with some large deletions, both indicators of genetic instability and potentially of carcinogenesis.

We are grateful to Walter Masson, Paul Simpson, and Cathryn Tambini for help at various stages of these experiments and to David Stevens for the irradiations.