Formation of large deletions by illegitimate recombination in the HPRT gene of primary human fibroblasts

Apparent and real recombination frequencies in multicopy plasmids: the need for a novel approach in frequency determination

Recombination studies of bacteria are often carried out with multicopy plasmids, and recombination frequencies are often deduced from the proportion of cells in the population that express a recombinant phenotype. These frequencies should however be called apparent frequencies, since detection of the recombinant cells requires not only the formation of a rearranged plasmid but also its establishment in the cell. The establishment of the recombinant plasmid can possibly be affected by its interaction with the parental plasmids. To test this hypothesis, we have used a plasmid system enabling the study of deletion formation between short direct repeats (18 bp) in Bacillus subtilis and developed a method by which deletion frequencies are measured under conditions under which interaction is abolished. Real deletion frequencies were thus determined and compared with apparent deletion frequencies. Real frequencies were underestimated by a factor ranging from 4- to 500-fold, depending upon the plasmid under study. This implies that a large majority of the recombinant molecules that are formed are generally not detected. We show that apparent deletion frequencies strongly depend upon (i) the parental plasmid copy number, (ii) the ability of the recombinant molecules to form heterodimeric plasmids, and (iii) the fitness of the recombinant molecules relative to that of parental molecules. Finally, we show that under conditions under which all recombinant molecules are scored, transcription can inhibit the deletion process 10-fold.

Targeted base substitutions and small deletions induced by neocarzinostatin at the APRT locus in plateau-phase CHO cells

Treatment of confluence-arrested CHO-D422 cells for 48 h with low concentrations (0.5-3 nM) of the radiomimetic antibiotic neocarzinostatin resulted in an increase in up to 11-fold in the frequency of mutations at the hemizygous APRT locus. Analysis by PCR and DNA sequencing revealed that the mutations were a mixture of base substitutions, small deletions, and large-scale rearrangements. base substitutions occurred preferentially at sequence positions where the drug is known to produce abasic sites with closely opposed strand breaks, e.g., AGT, TGT and AGC, where the abasic site occurs at the underlined base and the strand break occurs opposite the first base in each triplet. These results suggest that the substitutions were produced by replicative bypass of the abasic sites, perhaps during attempted repair of the accompanying strand break. Single-base deletions, which comprised nearly half of all deletions, were targeted to these same sequence positions, suggesting that they may have been generated either by replicative bypass of the abasic sites, or by end-joining repair of double-strand breaks, which are induced the same sites. Quantitative analysis of neocarzinostatin-induced damage to APRT DNA in vitro confirmed the association between lesions involving concommitant damage to both DNA strands, and mutations. The results are consistent the hypothesis that agents which induce such bistranded DNA damage can produce biologically significant levels of mutagenesis even in nondividing cells.

Construction of a vector containing a site-specific DNA double-strand break with 3'-phosphoglycolate termini and analysis of the products of end-joining in CV-1 cells

Previous studies have shown that linearized SV40-based shuttle vectors transfected into mammalian cells are efficiently recircularized by an error-prone end-joining pathway. To determine whether and with what specificity free radical-mediated double-strand breaks are rejoined by this pathway, a structural mimic of such a break was introduced at a specific site in an SV40-based shuttle vector, by ligating purified 3'-phosphoglycolate-terminated oligonucleotides into 3' recessed ends generated in the linearized vector. These terminally blocked linear vectors were efficiently repaired and replicated when transfected into simian CV-1 cells. Sequencing across the repair joints in progeny plasmid indicated that, for a blunt-ended vector, the most frequent mechanism of rejoining was splicing at a terminal 4-base homology; however, a significant fraction of the joints retained all bases from both ends of the break, consistent with a mechanism involving simple 3'-phosphoglycolate removal, followed by blunt-end ligation. For the analogous 3'-hydroxyl terminated break, the fraction of simple blunt-end ligations was considerably higher. For a phosphoglycolate-terminated vector with cohesive ends the most frequent repair mechanism was simple ligation of the annealed cohesive ends, presumably preceded by phosphoglycolate removal. For all these substrates, the remaining repair joints showed small or large deletions from one or both of the ends, usually with apparent annealing at short (1-4-base) homologies. The results suggest that while breaks with 3'-phosphoglycolates can be repaired, these blocked termini represent a significant barrier to DNA end-joining, and can significantly alter its specificity. The presence of cohesive ends appears to improve markedly the fidelity of rejoining for terminally blocked double-strand breaks.

Copy-choice recombination mediated by DNA polymerase III holoenzyme from Escherichia coli

Formation of deletions by recombination between short direct repeats is thought to involve either a break-join or a copy-choice process. The key step of the latter is slippage of the replication machinery between the repeats. We report that the main replicase of Escherichia coli, DNA polymerase III holoenzyme, slips between two direct repeats of 27 bp that flank an inverted repeat of approximately equal 300bp. Slippage was detected in vitro, on a single-stranded DNA template, in a primer extension assay. It requires the presence of a short (8 bp) G+C-rich sequence at the base of a hairpin that can form by annealing of the inverted repeats. It is stimulated by (i) high salt concentration, which might stabilize the hairpin, and (ii) two proteins that ensure the processivity of the DNA polymerase III holoenzyme: the single-stranded DNA binding protein and the beta subunit of the polymerase. Slippage is rather efficient under optimal reaction conditions because it can take place on >50% of template molecules. This observation supports the copy-choice model for recombination between short direct repeats.

Instability of long inverted repeats within mouse transgenes

Various sequences in the mammalian genomes are unstable. One class of sequence arrangement is long inverted repeats, which are known to be unstable in bacteria and yeast. While in mammals some evidence suggests that short inverted repeats (<10 bp long) may show instability, nothing is known about the stability of long inverted repeats. Here we describe two unrelated multicopy transgenes in the mouse (loci 109 and OX1-5), each of which contains a long inverted repeat that shows substantial mitotic instability. This instability also occurs in the germline so that mutant transgenes appear within pedigrees at a high frequency. The mutation processes acting at these two inverted repeats are complex and can involve insertion or deletion, and can result in stabilization of the transgene. At transgene 109 mutational events range from very small rearrangements at the centre of the inverted repeat to complete transgene deletion. In addition we show that the rates of mutation at the inverted repeat of transgene OX1-5 can vary between the male and female germlines and between inbred strains of mice, suggesting the possibility of a genetic analysis to identify loci that modulate inverted repeat instability.

From targets to genes: a brief history of radiosensitivity

The biological work of Douglas Lea spanned the period from 1934 to his early death in 1947, and during this short period he made important contributions to the theory of radiation action. He interpreted experimental data relating to the effects of radiation on viruses, bacteria, bean roots, etc in terms of the inactivation of discrete targets, which he identified with cellular genes. He thus laid the foundation of much subsequent research. It is now well recognized that mammalian cells differ substantially in radiosensitivity, especially in the low-dose region of the survival curve. The dependence of radiosensitivity on dose rate has been widely studied; this has practical significance for clinical radiotherapy as well as mechanistic implications. Since Lea's time there have been a number of efforts to describe models that can relate cell killing to radiation dose, dose rate, and track structure. So far these have not led to a comprehensive and widely accepted picture. Microdosimetric considerations lead to the concept of differing severity of lesions induced in DNA. Much is known about the sequence of processes that subsequently lead to cell inactivation: this can be divided into phases of induction, processing, and manifestation. Chromosomal events are currently attracting much attention, as they did in Lea's time. Considerable progress has also been made in identifying genes that control the repair of radiation damage. It has been found that mutation is frequently associated with the loss of a large segment of the genome around the damage site and this will have important implications for interactive processes between particle tracks.

High-frequency illegitimate integration of transfected DNA at preintegrated target sites in a mammalian genome

To examine the mechanisms of recombination governing the illegitimate integration of transfected DNA into a mammalian genome, we developed a cell system that selects for integration events in defined genomic regions. Cell lines with chromosomal copies of the 3' portion of the adenine phosphoribosyltransferase (APRT) gene (targets) were established. The 5' portion of the APRT gene, which has no homology to the integrated 3' portion, was then electroporated into the target cell lines, and selection for APRT gene function was applied. The reconstruction of the APRT gene was detected at frequencies ranging from less than 10(-7) to 10(-6) per electroporated cell. Twenty-seven junction sequences between the integrated 5' APRT and its chromosomal target were analyzed. They were found to be randomly distributed in a 2-kb region immediately in front of the 3' portion of the APRT gene. The junctions fell into two main classes: those with short homologies (microhomologies) and those with inserted DNA of uncertain origin. Three long inserts were shown to preexist elsewhere in the genome. Reconstructed cell lines were analyzed for rearrangements at the target site by Southern blotting; a variety of simple and complex rearrangements were detected. Similar analysis of individual clones of the parental cell lines revealed analogous types of rearrangement, indicating that the target sites are unstable. Given the high frequency of integration events at these sites, we speculate that transfected DNA may preferentially integrate at unstable mammalian loci. The results are discussed in relation to possible mechanisms of DNA integration.

The Saccharomyces cerevisiae Ku autoantigen homologue affects radiosensitivity only in the absence of homologous recombination

In mammalian cells, all subunits of the DNA-dependent protein kinase (DNA-PK) have been implicated in the repair of DNA double-strand breaks and in V(D)J recombination. In the yeast Saccharomyces cerevisiae, we have examined the phenotype conferred by a deletion of HDF1, the putative homologue of the 70-kD subunit of the DNA-end binding Ku complex of DNA-PK. The yeast gene does not play a role in radiation-induced cell cycle checkpoint arrest in G1 and G2 or in hydroxyurea-induced checkpoint arrest in S. In cells competent for homologous recombination, we could not detect any sensitivity to ionizing radiation or to methyl methanesulfonate (MMS) conferred by a hdf1 deletion and indeed, the repair of DNA double-strand breaks was not impaired. However, if homologous recombination was disabled (rad52 mutant background), inactivation of HDF1 results in additional sensitization toward ionizing radiation and MMS. These results give further support to the notion that, in contrast to higher eukaryotic cells, homologous recombination is the favored pathway of double-strand break repair in yeast whereas other competing mechanisms such as the suggested pathway of DNA-PK-dependent direct break rejoining are only of minor importance.

Repair of x-ray-induced DNA double-strand breaks in specific Not I restriction fragments in human fibroblasts: joining of correct and incorrect ends

An assay that allows measurement of absolute induction frequencies for DNA double-strand breaks (dsbs) in defined regions of the genome and that quantitates rejoining of correct DNA ends has been used to study repair of dsbs in normal human fibroblasts after x-irradiation. The approach involves hybridization of single-copy DNA probes to Not I restriction fragments separated according to size by pulsed-field gel electrophoresis. Induction of dsbs is quantitated from the decrease in the intensity of the hybridizing restriction fragment and an accumulation of a smear below the band. Rejoining of dsbs results in reconstitution of the intact restriction fragment only if correct DNA ends are joined. By comparing results from this technique with results from a conventional electrophoresis assay that detects all rejoining events, it is possible to quantitate the misrejoining frequency. Three Not I fragments on the long arm of chromosome 21 were investigated with regard to dsb induction, yielding an identical induction rate of 5.8 X 10(-3) break per megabase pair per Gy. Correct dsb rejoining was measured for two of these Not I fragments after initial doses of 80 and 160 Gy. The misrejoining frequency was about 25% for both fragments and was independent of dose. This result appears to be representative for the whole genome as shown by analysis of the entire Not I fragment distribution. The correct rejoining events primarily occurred within the first 2 h, while the misrejoining kinetics included a much slower component, with about half of the events occurring between 2 and 24 h. These misrejoining kinetics are similar to those previously reported for production of exchange aberrations in interphase chromosomes.

Menage à trois: double strand break repair, V(D)J recombination and DNA-PK

All organisms possess mechanisms to repair double strand breaks (dsbs) generated in their DNA by damaging agents. Site-specific dsbs are also introduced during V(D)J recombination. Four complementation groups of radiosensitive rodent mutants are defective in the repair of dsbs, and are unable to carry out V(D)J recombination effectively. The immune defect in Severe Combined Immunodeficient (scid) mice also results from an inability to undergo effective V(D)J recombination, and scid cell lines display a repair defect and belong to one of these complementation groups. These findings indicate a mechanistic overlap between the processes of DNA repair and V(D)J recombination. Recently, two of the genes defined by these complementation groups have been identified and shown to encode components of DNA-dependent protein kinase (DNA-PK). We review here the three fields which have become linked by these findings, and discuss the involvement of DNA-PK in dsb rejoining and in V(D)J recombination.

Breakpoints and junctional regions of intragenic deletions in the HPRT gene in human T-Cells

DNA sequences of the deletion breakpoints of 24 human T-lymphocyte hprt gene mutations are reported. These independent deletions ranged in size from 18 to 15655 base pairs. Seven of the 21 in vivo mutations arose in normal adults, three in normal children, eight in radioimmunotherapy patients and three in platinum chemotherapy patients. One in vitro mutation was isolated after 93cGy radon exposure and two after 300cGy gamma radiation. The breakpoints were found to be non-random and a cluster of small deletions in exon 6 is reported. Ten of the mutations had 2-5bp direct repeats at the breakpoints. There was no excess of "deletion-associated" motifs over that expected by chance. Some breakpoints do occur at consensus topoisomerase II cleavage sites and the centromeric end of a Donehower sequence occurs exactly at a telomeric breakpoint. Three mutants had breakpoints at hairpins expected by the model of Glickman and Ripley.

Deletions at short direct repeats and base substitutions are characteristic mutations for bleomycin-induced double- and single-strand breaks, respectively, in a human shuttle vector system

Using the radiomimetic drug, bleomycin, we have determined the mutagenic potential of DNA strand breaks in the shuttle vector pZ189 in human fibroblasts. The bleomycin treatment conditions used produce strand breaks with 3'-phosphoglycolate termini as > 95% of the detectable dose-dependent lesions. Breaks with this end group represent 50% of the strand break damage produced by ionizing radiation. We report that such strand breaks are mutagenic lesions. The type of mutation produced is largely determined by the type of strand break on the plasmid (i.e. single versus double). Mutagenesis studies with purified DNA forms showed that nicked plasmids (i.e. those containing single-strand breaks) predominantly produce base substitutions, the majority of which are multiples, which presumably originate from error-prone polymerase activity at strand break sites. In contrast, repair of linear plasmids (i.e. those containing double-strand breaks) mainly results in deletions at short direct repeat sequences, indicating the involvement of illegitimate recombination. The data characterize the nature of mutations produced by single- and double-strand breaks in human cells, and suggests that deletions at direct repeats may be a 'signature' mutation for the processing of DNA double-strand breaks.

Radiation-induced genomic instability

Quantitative assessment of the heritable somatic effects of ionizing radiation exposures has relied upon the assumption that radiation-induced lesions were 'fixed' in the DNA prior to the first postirradiation mitosis. Lesion conversion was thought to occur during the initial round of DNA replication or as a consequence of error-prone enzymatic processing of lesions. The standard experimental protocols for the assessment of a variety of radiation-induced endpoints (cell death, specific locus mutations, neoplastic transformation and chromosome aberrations) evaluate these various endpoints at a single snapshot in time. In contrast with the aforementioned approaches, some studies have specifically assessed radiation effects as a function of time following exposure. Evidence has accumulated in support of the hypothesis that radiation exposure induces a persistent destabilization of the genome. This instability has been observed as a delayed expression of lethal mutations, as an enhanced rate of accumulation of non-lethal heritable alterations, and as a progressive intraclonal chromosomal heterogeneity. The genetic controls and biochemical mechanisms underlying radiation-induced genomic instability have not yet been delineated. The aim is to integrate the accumulated evidence that suggests that radiation exposure has a persistent effect on the stability of the mammalian genome.

Different capacities for recombination in closely related human lymphoblastoid cell lines with different mutational responses to X-irradiation

WIL2-NS and TK6 are two distinct human lymphoblast cell lines derived from a single male donor. WIL2-NS cells are significantly more resistant to the cytotoxic effects of X-irradiation but considerably more sensitive to induced mutation. In an effort to determine the mechanistic basis for these differences, we analyzed the physical structures of thymidine kinase (tk)-deficient mutants isolated after X-ray treatment of tk heterozygotes derived from TK6 and the more mutable WIL2-NS. Southern analysis showed that while 84% of TK6-derived mutants had arisen by loss of heterozygosity (LOH), all 106 mutants from WIL2-NS derivatives arose with LOH at tk and all but one showed LOH at other linked loci on chromosome 17. We adapted a fluorescence in situ hybridization technique to distinguish between LOH due to deletion, which results in retention of only one tk allele, and LOH due to a mechanism involving the homologous chromosome (e.g., recombination), which results in the retention of two alleles. Among the LOH mutants derived that were analyzed in this way, 9 of 26 from WIL2-NS and 11 of 17 from TK6 cell lines arose by deletion. The remaining mutants retained two copies of the tk gene and thus arose by a mechanism involving the homologous allele. Since many of these mutants arising by a homologous mechanism retained partial heterozygosity of chromosome 17, they must have arisen by recombination or gene conversion, and not chromosome loss and reduplication. Finally, the recombinational capacities of WIL2-NS and TK6 were compared in transfection assays with plasmid recombination substrates. Intermolecular recombination frequencies were greater in WIL2-NS than in TK6. These data are consistent with a model suggesting that a recombinational repair system is functioning at a higher level in WIL2-NS than in TK6; the greater mutability of the tk locus in WIL2-NS results from more frequent inter- and intramolecular recombination events.

Different capacities for recombination in closely related human lymphoblastoid cell lines with different mutational responses to X-irradiation

WIL2-NS and TK6 are two distinct human lymphoblast cell lines derived from a single male donor. WIL2-NS cells are significantly more resistant to the cytotoxic effects of X-irradiation but considerably more sensitive to induced mutation. In an effort to determine the mechanistic basis for these differences, we analyzed the physical structures of thymidine kinase (tk)-deficient mutants isolated after X-ray treatment of tk heterozygotes derived from TK6 and the more mutable WIL2-NS. Southern analysis showed that while 84% of TK6-derived mutants had arisen by loss of heterozygosity (LOH), all 106 mutants from WIL2-NS derivatives arose with LOH at tk and all but one showed LOH at other linked loci on chromosome 17. We adapted a fluorescence in situ hybridization technique to distinguish between LOH due to deletion, which results in retention of only one tk allele, and LOH due to a mechanism involving the homologous chromosome (e.g., recombination), which results in the retention of two alleles. Among the LOH mutants derived that were analyzed in this way, 9 of 26 from WIL2-NS and 11 of 17 from TK6 cell lines arose by deletion. The remaining mutants retained two copies of the tk gene and thus arose by a mechanism involving the homologous allele. Since many of these mutants arising by a homologous mechanism retained partial heterozygosity of chromosome 17, they must have arisen by recombination or gene conversion, and not chromosome loss and reduplication. Finally, the recombinational capacities of WIL2-NS and TK6 were compared in transfection assays with plasmid recombination substrates. Intermolecular recombination frequencies were greater in WIL2-NS than in TK6. These data are consistent with a model suggesting that a recombinational repair system is functioning at a higher level in WIL2-NS than in TK6; the greater mutability of the tk locus in WIL2-NS results from more frequent inter- and intramolecular recombination events.

Supravalvular aortic stenosis associated with a deletion disrupting the elastin gene

Supravalvular aortic stenosis (SVAS) is an inherited vascular disease that can cause heart failure and death. SVAS can be inherited as an autosomal dominant trait or as part of a developmental disorder, Williams syndrome (WS). In recent studies we presented evidence suggesting that a translocation disrupting the elastin gene caused SVAS in one family while deletions involving the entire elastin locus caused WS. In this study, pulsed-field, PCR, and Southern analyses showed that a 100-kb deletion of the 3' end of the elastin gene cosegregated with the disease in another SVAS family. DNA sequence analysis localized the breakpoint between elastin exons 27 and 28, the same region disrupted by the SVAS-associated translocation. These data indicate that mutations in the elastin gene cause SVAS and suggest that elastin exons 28-36 may encode critical domains for vascular development.

The role of mitotic recombination in carcinogenesis

Genetic recombination systems are present in all living cells and viruses and generally contribute to their hosts' flexibility with respect to changing environmental conditions. Recombination systems not only help highly developed organisms to protect themselves from microbial attack via an elaborate immune system, but conversely, recombination systems also enable microorganisms to escape from such an immune system. Recombination enzymes act with a high specificity on DNA sequences that either exhibit extended stretches of homology or contain characteristic signal sequences. However, recombination enzymes may rarely act on incorrect alternative target sequences, which may result in the formation of chromosomal deletions, inversions, translocations, or amplifications of defined DNA regions. This review describes the characteristics of several recombination systems and focuses on the implication of aberrant recombination in carcinogenesis. The consequences of mitotic recombination on the inappropriate activation of protooncogenes and on the loss of tumor suppressor genes is discussed. Cases are reported where mitotic recombination clearly has been associated with carcinogenesis in rodents as well as humans. Several test systems able to detect recombinagenic activities of chemical compounds are described.

Molecular mechanisms of radiation oncogenesis

Resolving the molecular mechanisms of radiation oncogenesis represents an important but daunting challenge in radiation research. This brief review outlines the principal oncogenic mechanisms that need to be considered in the context of radiation effects on the genome, how these might relate to specific gene and chromosomal changes relevant to neoplasia and the possible implications of such knowledge for the modelling of cancer risk. The long-term application of this mechanistic knowledge to the determination of tumour causality and for the assessment of individual cancer risk is also briefly discussed.

Detection of deletion mutations extending beyond the HPRT gene by multiplex PCR analysis

A multiplex PCR assay was developed for the rapid analysis of deletion size at the hypoxanthine phosphoribosyltransferase (hprt) locus. The DNA sequence of mapped DNA segments flanking the hprt gene was determined. These cloned DNAs were derived from the ends of a set of overlapping yeast artificial chromosomes (YAC) defining a contig of 8 Mb at Xq26 and including hprt. We used "bubble" PCR to isolate an additional YAC end-clone. Seven primer pairs were derived from DNA sequence analysis of the clones and incorporated into a multiplex PCR assay. These primer pairs define loci located approximately 750 kb and 350 kb upstream of hprt and 300 kb, 540 kb, 900 kb, 1260 kb, and 1400 kb downstream of hprt. A primer pair for an unlinked and unselected gene sequence (K-ras) was also included in the multiplex reaction to serve as an internal positive control. Using this new assay, hprt mutant DNAs can be screened to determine the extent of deletion. Deletions larger than 2 Mb have been identified and show that large deletions can be tolerated at this hemizygous locus.

D8S7 is consistently deleted in inverted duplications of the short arm of chromosome 8 (inv dup 8p)

Ten patients with inverted duplication of 8p (inv dup 8p) were studied with cytogenetic, biochemical and molecular techniques. The duplication for the region 8p12-p22 was always associated with a deletion of the locus D8S7 (mapped in 8p23.1) as demonstrated with the probe pSW50 by both in situ hybridization and Southern blot. Restriction fragment length polymorphisms detected by probes pSW50 (1 case) and by pG2LPL35 (locus LPL) (two cases) were informative as to a maternal origin of the anomaly. The activity of glutathione reductase, whose gene maps in the duplicated region at 8p21.1, was increased in all patients. The recognizable phenotype of inv dup 8p includes neonatal hypotonia, prominent forehead, large mouth with everted lower lip, abnormally shaped large ears, brain malformations and severe mental retardation. Our findings indicate that the chromosome rearrangement is homogeneous at least for the presence of the deletion and support the hypothesis of a common mechanism of origin.

Weiss Lecture. Effects of radiations of different qualities on cells: molecular mechanisms of damage and repair

Studies of ionizing radiations of different quality are discussed with particular emphasis on damage to DNA of mammalian cells. Three related themes are followed. Firstly, inactivation and mutation experiments with ultrasoft X-rays and slow heavy ions, coupled with theoretical analyses of the structures of the radiation tracks, have emphasized the biological importance of localized track features over nanometre dimensions. This led to the suggestion that the critical physical features of the tracks are the stochastic clusterings of ionizations, directly in or very near to DNA, resulting in clustered initial molecular damage including various combinations of breaks, base damages, cross-links, etc. in the DNA. The quantitative hypotheses imply that final cellular effects from high-LET radiations are dominated by their more severe, and therefore less repairable, clustered damage, and that these are qualitatively different from the dominant low-LET damage. Second, relative effectiveness of different types of radiation led to questions on the mechanisms of induction of chromosome exchanges. The high efficiency of ultrasoft X-rays, despite their very short track lengths, suggested that single sites of DNA damage may lead to exchanges by a molecular process involving interaction with undamaged DNA. Also it is shown that a single site-specific DNA break, introduced by restriction enzymes, sometimes leads to a large deletion when misrepaired by cell extracts. These deletions occur between short DNA repeats, and are therefore a form of 'illegitimate' recombination, but clearly do not involve the interaction of two damage sites. Third, it was shown that cells from patients with the radiosensitive disorder ataxia-telangiectasia (AT) lack a post-irradiation recovery process. The sensitivity of AT cells to high LET radiations was found to be reduced relative to that for normal cells, reinforcing the concept that high LET damage is less easy to repair. AT patients are prone to lymphoreticular cancers, and their cells show characteristic chromosomal rearrangements, which may be associated with misrepair at specific genomic sequences. Similarly, studies of radiation-induced leukaemia in the mouse have implicated rearrangement at specific interstitial chromosome sites, which are rich in telomere-like repeat sequences.