It is now recognized that a subset of B-cell chronic lymphocytic leukemia (CLL) is familial. The genetic basis of familial CLL is poorly understood, but recently germ line mutations in the Ataxia Telangiectasia (ATM) gene have been proposed to confer susceptibility to CLL. The evidence for this notion is, however, not unequivocal. To examine this proposition further we have screened the ATM gene for mutations in CLLs from 61 individuals in 29 families. Truncating ATM mutations, including a knownATM mutation, were detected in 2 affected individuals, but the mutations did not cosegregate with CLL in the families. In addition, 3 novel ATM missense mutations were detected. Common ATM missense mutations were not overrepresented. The data support previous observations that ATM mutation is associated with B-CLL. However, ATM mutations do not account for familial clustering of the disease.

Leukemia affects between 1% and 2% of Western populations.1 B-cell chronic lymphocytic leukemia (CLL) is the most common form of leukemia, accounting for around 30% of cases.2 The incidence rate of CLL increases logarithmically from age 35 with a median age at diagnosis of 65 years.3 

Epidemiologic studies and published case reports of families indicate that a subset of the CLL is ascribable to an inherited genetic predisposition. A dominantly acting gene or genes with pleiotropic effects appear to be the most likely genetic model of inheritance since CLL appears to segregate with other lymphoproliferative disorders (LPDs) in many families.4 However, no gene has been shown unequivocally to be causative.

Ataxia Telangiectasia (A-T), is one of a group of recessive syndromes characterized by excessive spontaneous chromosomal breakage associated with an elevated risk of hematologic malignancies.5Specifically, in A-T there is an increased risk of lymphomas and leukemias, including those in the B-cell lineage.6 

The Ataxia Telangiectasia (ATM) gene maps to chromosome 11q237 and specifies a 12-kilobase (kb) mRNA encoding an approximately 350-kd protein.8 The 3′ end of the gene has some homology to phosphoinositide 3–kinases and toSaccharomyces cerevisiae telomerase 1 (TEL1) that controls telomere length and maintenance of genome integrity.9 Homozygous germ line mutations in ATMare associated with increased radiosensitivity and genomic instability. The mutation rate is increased, double-strand breaks show reduced rejoining fidelity, there is a high frequency of cytogenetic rearrangements, and homologous recombination is increased and error prone.10-13 

A-T cells also show defects in cell-cycle checkpoints. These properties provide the rationale for proposing a model14 in whichATM has a role as a DNA damage-response gene probably responding to a subset of DNA double-strand breaks.

There is evidence that A-T carriers may display an increased cancer risk.15 Some reports have suggested that A-T heterozygotes display an increased chromosomal instability and hypersensitivity to carcinogens, thereby increasing the probability of acquiring oncogenic somatic mutations.16-18 The relationship between cancer risk and A-T heterozygosity has been most extensively studied in relation to breast cancer.19 CLL has been reported in A-T families,15 suggesting that heterozygosity may also confer an increased risk of this disease. In a retrospective study of cancer incidence in 110 A-T families, the risk of hematologic and lymphoid malignancies was increased in blood relatives of A-T patients and CLL accounted for all but one of the leukemias seen in adult blood relatives. However, these observations did not attain statistical significance. It has recently been shown that the ATM gene is mutant in approximately 20% of CLL samples from unrelated patients and that some patients have heterozygous germ line mutations.20-23 

The prevalence of B-cell malignancy in A-T coupled with the possible increase in risk of leukemia in relatives of A-T patients has led a number of researchers to question whether germ line ATMmutations are involved in familial cases of CLL. In a small study of 32 CLL patients, Stankovic et al20 reported that 6% of cases harbored germ line ATM mutations.20 Bevan et al24 recently examined the role of ATM in familial CLL through a linkage analysis. While there was no evidence for linkage, the study did not preclude ATM underlying a subset of familial CLL.

ATM represents an attractive candidate CLL predisposition gene but direct evidence for its role is lacking. To investigate further the relationship between ATM and CLL we have screened tumor samples from 61 individuals in 29 CLL families for mutations.

Patient selection

Families with 2 or more individuals affected with CLL were ascertained through hematologists in the United Kingdom, Norway, Israel, Italy, Germany, Portugal, and Poland. The diagnosis of CLL in all cases was based on standard hematologic and immunologic criteria. Samples were obtained from family members with informed consent and ethical review board approval from the Royal Marsden Hospital's National Health Service Trust. DNA was salt extracted from peripheral EDTA venous blood samples using a standard sucrose lysis method. No separation of tumor B cells from other cells was undertaken prior to DNA extraction.

Mutation detection

The 63 coding exons and exon 2 of ATM were analyzed by polymerase chain reaction (PCR) amplification of DNA and single-strand conformational polymorphism (SSCP) electrophoresis of32P–deoxycytosine triphosphosphate (dCTP)–labeled products through glycerol polyacrylamide gels as described,25 except that primers were redesigned to ensure that PCR products were approximately 250 base pairs or shorter. Table 1 details the nucleotide sequences of each of the 65 sets of PCR primers. Samples with bandshifts were sequenced with the same PCR primers in forward and reverse directions using the Li-Cor method (MWG, Milton Keynes, United Kingdom; http://www.mwg-biotech.com). Identified nucleotide changes were coded according to the genomic nucleotide sequence at GenBank (http://www.ncbi.nlm.nih.gov/Genbank/) accession no.U82828 for introns and NM 000051 for each exon (renumbered with base 190 corresponding to base 1 in order to provide data in the same form as that used in the Virginia Mason ATM Mutation Database).26 

Twenty-nine families with 2 or more affected individuals with CLL were studied. Table 2shows the pedigree structure of each of the families and the ages at diagnosis of CLL in family members.

DNA from blood samples was examined for nucleotide changes in theATM gene by PCR-SSCP. All coding exons were examined using primers in the flanking introns or, where an exon was long, using 2 pairs of primers ensuring the middle 2 primers amplified overlapping segments of the exon.

The ATM nucleotide changes detected in the 61 affected individuals from the 29 CLL families are detailed in Table 2. Allele frequencies estimated for each of the mutations detected in the study are given in Table 3. This table also shows the allele frequencies of each of the mutations reported in published studies together with reported designations of each nucleotide change. In some instances different authors have interpreted the same nucleotide change as a mutation, a polymorphism, or present in a carrier. To be consistent we have classified nucleotide changes as “substitution” or “truncating” (leading to amino acid substitution or to truncation of the ATM polypeptide) or other (outside exons and their splice site consensus sequences or within an exon but not leading to change in amino acid usage).

In the samples analyzed, 11 nucleotide changes were identified that have not previously been reported. Altered SSCP patterns corresponding to these nucleotide changes are shown in Figure1. Three of the nucleotide changes lead to amino acid substitutions. Two of the nucleotide changes lead to truncation of ATM due to either an insertion or deletion of 2 base pairs within respective exons of the gene. Of the 7 other nucleotide changes detected, 2 result in no change in amino acid usage and 5 were intronic nucleotide changes residing at least 27 base pairs away from intron-exon boundaries. No nucleotide changes identified created or destroyed consensus splice sites.

Fig. 1.

SSCP bandshifts corresponding to novel ATM nucleotide changes identified.

Out of 61 samples, 11 novel nucleotide changes were identified. Bandshifts (arrowed) due to these nucleotide changes detected on autoradiograms after SSCP are shown adjacent to normal samples. The nucleotide changes were: IVS4-36insTG (intron 4); 162T>C (exon 5); 378T>A (exon 7); IVS15-48T/C (intron 15); IVS16 + 78G/A (intron 16); 4578C>T (exon 32); 5042T>C (exon 36); 5756delAA (exon 40); IVS40 + 27G/A (intron 40); 6919C>T (exon 49); 7271insGT (exon 51).

Fig. 1.

SSCP bandshifts corresponding to novel ATM nucleotide changes identified.

Out of 61 samples, 11 novel nucleotide changes were identified. Bandshifts (arrowed) due to these nucleotide changes detected on autoradiograms after SSCP are shown adjacent to normal samples. The nucleotide changes were: IVS4-36insTG (intron 4); 162T>C (exon 5); 378T>A (exon 7); IVS15-48T/C (intron 15); IVS16 + 78G/A (intron 16); 4578C>T (exon 32); 5042T>C (exon 36); 5756delAA (exon 40); IVS40 + 27G/A (intron 40); 6919C>T (exon 49); 7271insGT (exon 51).

Close modal

Among previously unreported nucleotide changes, intervening sequence (IVS)4-36insTG was the most common in this series, detected in 10% (12/120) of alleles. The abnormal band pattern due to IVS4-36insTG was only detected when autoradiograms were briefly exposed to film (∼15 minutes). The pattern involved a new dark band virtually comigrating with a normal dark band. There were no detectable new bands among the less intense bands. The subtle nature of change in bands associated with IVS4-36insTG may explain why this common polymorphism has not previously been detected. The allelic frequencies of other previously unreported nucleotide changes were less than 4%.

In 61 of the samples analyzed, 9 amino acid substitutions or truncating changes were detected that might affect the function of the ATMprotein. Six of these changes were observed in single cases but for none of these cases was the same change seen in any affected relative. Among these 6 changes, 2 were detected in the same case: 5756delAA and 7271insGT were detected in an affected sibling in family 43. Figure 1 shows the comparative relative intensities of the exon 40 and exon 51 bands corresponding to these mutations. Examination of autoradiograms indicates that 5756delAA constituted around 20% of the sample while 7271insGT constituted about 50% of the sample. Two amino acid substitutions, 378T>A (Asp126Glu) and 5042T>C (Ile1681Thr) were each detected in 2 affected individuals. 378T>A was detected in the index case of family 63, but not in his affected sibling, and in the index case of family 98, but not in his affected sibling. 5042T>C was detected in 2 of the 3 affected siblings in family 5. One mutation, (5557G>A), which leads to an amino acid substitution (Asp1853Asn), was detected in 15 alleles giving a frequency comparable to that previously reported in healthy individuals from the general population.

The hypothesis that ATM represents a predisposition locus for CLL is attractive a priori. Strong evidence for such a notion would be provided by cosegregation in families of microsatellite markers and of mutations. However, this depends on ATMmutations conferring a markedly elevated risk. Furthermore, the power of any set of families to demonstrate linkage will be severely reduced if the disease is genetically heterogeneous. If ATMonly confers modest risks, mutations will not generate multiple-case families and linkage will not be detected, especially if phenocopies are common. There should, however, be evidence of overrepresentation of mutations in cases compared to the general population. The mutation analysis reported here on the ATMgene in tumor samples from 61 individuals in 29 CLL families was undertaken to assess further the question of whether ATMrepresents a predisposition locus for CLL.

In our study, 2 of the 61 patients with CLLs in the 29 families had truncating ATM mutations. One patient with CLL had 8266A>T (Lys2756Xaa). This mutation has been reported in A-T families.33-35 The second patient with CLL had 2 truncating mutations, 5766delAA and 7271insGT. It is highly probable that just one of these mutations is somatic. To assess whether this class of ATM mutation confers an increase in risk of CLL per se, the observed number of mutations needs to be compared with that expected by chance. The number of mutations expected is not simply a function of the number of affected individuals analyzed but is also a function of the familial relationship between affected individuals within each family. The expected number of mutations computed on this basis is 0.46. Therefore, observation of 2 mutations equates to a 4.4-fold overrepresentation, albeit, nonsignificant (95% CI, 0.5-15.9). Stankovic et al20 suggested that germ line mutations in ATM confer susceptibility to CLL following the finding that 2 of 32 sporadic CLL cases harbored constitutional mutations. One of the mutations identified was a truncating mutation and the other was a missense mutation (Pro1054Arg, 3161C>G) which the authors designated as pathogenic. The significance of this observation depends critically on the carrier frequency of pathogenic ATMmutations in the general population. Most estimates based on the prevalence of A-T suggest that the population carrier frequency of pathogenic mutations in ATM is around 1%.36Assuming this to be the case, then the observation of 2 mutations in 32 cases is significant (P = .04). Pro1054Arg has however, been designated as a polymorphism by some workers.37 If this is the case, then the observation of 1 pathogenic constitutional mutation in 32 CLL cases does not constitute an overrepresentation (P = .28). However, combining this data with ours, the observation of 3 truncating mutations indicates overrepresentation at the 0.04 level.

It is conceivable that ATM may impact more significantly on the risk of CLL if some nontruncating mutations have pathogenic potential. Gatti et al38 have recently proposed that cancer susceptibility may be associated not with truncating ATMmutations, but rather with missense ATM mutations. In the present study, 6 missense mutations were identified; 3 have been reported previously (Asp1853Asn, Phe858Leu, and Leu1420Phe) and 3 are new (Asp126Glu, Ile1681Thr, and Leu2307Phe). Asp1853Asn (5557G>A) was detected in 13 cases of CLL from 8 of the 29 families studied. Some evidence suggests that 5557A may be associated with a modest increase in cancer risk.39 The evidence from our study does not support such a postulate in the context of CLL. Although in 4 families the CLLs from both affected individuals harbored at least one copy of this allele, in the other 4 families the mutation did not segregate with disease. Furthermore, there was no evidence of overrepresentation of the 5557A allele. Overall the frequency of the 5557A allele was comparable to the population estimate of 18%. 2572T>C (Phe858Leu) was detected in 4 unrelated individuals, failing to segregate with CLL. The frequency of the 2572C allele was also similar to the population estimate of 2% in published reports. The recent proposal40that 2572C only confers an elevated cancer risk in the presence of 5557A is not supported by our study as this combination was not observed to segregate with CLL. Asp126Glu (378T>A) was also not observed to segregate with CLL, being detected in 2 unrelated individuals. While the residue is conserved between human and murine ATM protein, 126Glu is unlikely to confer risk as the amino acid change is conservative. Ile1681Thr (5042T>C) was detected in 2 of the 5 affected individuals in family 5. 5042C is unlikely to confer risk as the amino acid change is conservative and Ile1681 is not conserved between human and murine ATM proteins.

Leu1420Phe (4258C>T) and Leu2307Phe (6919C>T) mutations were detected in single CLLs. Leu1420Phe (4258C>T) has been reported as conferring risk of A-T. Leu2307Phe is a nonconservative change and thus may confer risk of CLL. In the absence of robust functional assays for each domain of the multifunctional ATM protein it is difficult to specify pathogenicity of these variants. However, the absence of cosegregation of amino acid changes in the families we have studied implies that the nonconservative amino acid changes will by themselves confer small genotypic risks. If a sequence variant is detected in just one family member we cannot exclude in the present study the possibility that some mutations are somatic. As a result, the ATMsequence variants we report here may overrepresent the germ line variability of ATM in this collection of CLL samples.

As with most common cancers susceptibility to CLL is likely to exhibit genetic heterogeneity. Furthermore, excluding large multiple-case families segregating CLL in a clear mendelian fashion, nuclear families with small numbers of cases may display a degree of within-family heterogeneity whereby, in addition to phenocopies, affected individuals in the same family may be caused by different susceptibility genes.

We report additional data that provide support for previous observations that ATM mutations are associated with CLL. However, on the basis of the data presented here, ATM is unlikely to represent a major susceptibility gene for familial CLL as the familial relative risk ascribable to mutations is unlikely to be more than approximately 1.1. To determine whether specificATM variants represent predisposition mutations, large cohorts of CLL patients and controls will be required. However, the spectrum and the frequency of ATM variants makes the evaluation of specific nucleotide changes as risk factors for CLL inherently difficult, as has been seen in studies seeking to establishATM as a risk factor for breast cancer.

A complete list of the participating clinicians are included in the at the end of this article. We thank Jenny Burrows, Julie Fuller and Andrea Marossy for their assistance. Finally, we are grateful to the 2 anonymous reviewers for their comments.

We thank the following clinicians and their patients: K. Quabeck, Germany; P. Stark, Israel; E. Gaminara, St Albans, United Kingdom; P. Antunovic, Norway; F. Mauro, Italy; H. Sykes, Kingston Upon Thames, United Kingdom; I. Ribeiro, Portugal; A. Bell, Poole, United Kingdom; M. Auger, Sutton in Ashfield, United Kingdom; S. Rassam, Sidcup, United Kingdom; M. Junior, Portugal; I. Ben-Bassat, Israel; R. M. Stewart, Chesterfield, United Kingdom; J. R. Duncan, Brighton, United Kingdom; L.G. Quaglino, Italy; P.M. Chipping, Stoke on Trent, United Kingdom.

Supported by the Kay Kendall Leukemia Trust and the Leukemia Research Fund.

The publication costs of this article were defrayed in part by page charge payment. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 U.S.C. section 1734.

1
Miller
BA
Ries
LAG
Hankey
BF
Kosary
CL
Harras
A
Devesa
SS
SEER Cancer Statistics Review 1973-90.
1993
National Cancer Institute
Bethesda, MD
2
Gale
RP
Foon
KA
Biology of chronic lymphocytic leukemia.
Semin Hematol.
24
1987
209
229
3
Linet
MS
Blattner
WA
The epidemiology of chronic lymphocytic leukemia.
Chronic Lymphocytic Leukemia.
Polliack
A
Catovsky
D
1988
11
32
Harwood Academic Publishers
Chur, Switzerland
4
Yuille
MR
Matutes
E
Marossy
A
Hilditch
B
Catovsky
D
Houlston
RS
Familial chronic lymphocytic leukemia: a survey and review of published studies.
Br J Haematol.
109
2000
794
799
5
German
J
Chromosome-breakage syndromes: different genes, different treatments, different cancers.
Basic Life Sci.
15
1980
429
439
6
Taylor
AMR
Metcalfe
JA
Thick
J
Mak
Y-F
Leukemia and lymphoma in ataxia telangiectasia.
Blood.
87
1996
423
428
7
Gatti
RA
Berkel
I
Boder
E
et al
Localization of an ataxia-telangiectasia gene to chromosome 11q22–23.
Nature.
336
1988
577
578
8
Savitsky
K
Sfez
S
Tagle
DA
et al
The complete sequence of the coding region of the ATM gene reveals similarity to cell cycle regulators in different species.
Hum Mol Genet.
4
1995
2025
2032
9
Savitsky
K
Bar-Shira
A
Gilad
S
et al
A single ataxia-telangiectasia gene with a product similar to PI-3 kinase.
Science.
268
1995
1749
1752
10
Steingrimsdottir
H
Rowley
G
Waugh
A
et al
Molecular analysis of mutations in the hprt gene in circulating lymphocytes from normal and DNA-repair-deficient donors.
Mutat Res.
294
1993
29
41
11
Stumm
M
Neubauer
S
Keindorff
S
Wegner
RD
Wieacker
P
Sauer
R
High frequency of spontaneous translocations revealed by FISH in cells from patients with the cancer-prone syndromes ataxia telangiectasia and Nijmegen breakage syndrome.
Cytogenet Cell Genet.
92
2001
186
191
12
North
P
Ganesh
A
Thacker
J
The rejoining of double-strand breaks in DNA by human cell extracts.
Nucleic Acids Res.
18
1990
6205
6210
13
Meyn
MS
High spontaneous intrachromosomal recombination rates in ataxia-telangiectasia.
Science.
260
1993
1327
1330
14
Meyn
MS
Ataxia-telangiectasia, cancer and the pathobiology of the ATM gene.
Clin Genet.
55
1999
289
304
15
Cole
J
Arlett
CF
Green
MH
et al
Comparative human cellular radiosensitivity, II: the survival following gamma-irradiation of unstimulated (G0) T-lymphocytes, T-lymphocyte lines, lymphoblastoid cell lines and fibroblasts from normal donors, from ataxia-telangiectasia patients and from ataxia-telangiectasia heterozygotes.
Int J Radiat Biol.
54
1988
929
943
16
West
CM
Elyan
SA
Berry
P
Cowan
R
Scott
D
A comparison of the radiosensitivity of lymphocytes from normal donors, cancer patients, individuals with ataxia-telangiectasia (A-T) and A-T heterozygotes.
Int J Radiat Biol.
68
1995
197
203
17
Pernin
D
Bay
JO
Uhrhammer
N
Bignon
YJ
ATM heterozygote cells exhibit intermediate levels of apoptosis in response to streptonigrin and etoposide.
Eur J Cancer.
35
1999
1130
1135
18
Athma
P
Rappaport
R
Swift
M
Molecular genotyping shows that ataxia-telangiectasia heterozygotes are predisposed to breast cancer.
Cancer Genet Cytogenet.
92
1996
130
134
19
Swift
M
Reitnauer
PJ
Morrell
D
Chase
CL
Breast and other cancers in families with ataxia-telangiectasia.
N Engl J Med.
316
1987
1289
1294
20
Stankovic
T
Weber
P
Stewart
G
et al
Inactivation of ataxia telangectasia mutated gene in B-cell chronic lymphocytic leukemia.
Lancet.
353
1999
26
29
21
Schaffner
C
Stilgenbauer
S
Rappold
GA
Dohner
H
Lichter
P
Somatic ATM mutations indicate a pathogenic role of ATM in B-cell chronic lymphocytic leukemia.
Blood.
94
1999
748
753
22
Bullrich
F
Rasio
D
Kitada
S
et al
ATM mutations in B-cell chronic lymphocytic leukemia.
Cancer Res.
59
1999
24
27
23
Starostik
P
Manshouri
T
O'Brien
S
et al
Deficiency of the ATM protein expression defines an aggressive subgroup of B-cell chronic lymphocytic leukemia.
Cancer Res.
58
1998
4552
4557
24
Bevan
S
Catovsky
D
Marossy
A
et al
Linkage analysis for ATM in familial chronic lymphocytic leukemia.
Leukemia.
13
1999
1497
1500
25
Vorechovsky
I
Luo
L
Dyer
MJS
et al
Clustering of missense mutations in the ataxia-telangiectasia gene in a sporadic T-cell leukemia.
Nat Genet.
17
1997
96
99
26
Concannon P, Gatti RA. The Ataxia-Telangiectasia Mutation Database. Available at http://www.vmresearch.org/atm.htmAccessed March 2001.
27
Sandoval
N
Platzer
M
Rosenthal
A
et al
Characterization of ATM mutations in 66 ataxia telangiectasia families.
Hum Mol Genet.
8
1999
69
79
28
Vorechovsky
I
Rasio
D
Luo
L
et al
The ATM gene and susceptibility to breast cancer: analysis of 38 breast tumors reveals no evidence for mutation.
Cancer Res.
56
1996
2726
2732
29
Izatt
L
Greenman
J
Hodgson
S
et al
Identification of germline missense mutations and rare allelic variants in the ATM gene in early-onset breast cancer.
Genes Chromosomes Cancer.
26
1999
286
294
30
Castellvi-Bel
S
Sheikhavandi
S
Telatar
M
et al
New mutations, polymorphisms, and rare variants in the ATM gene detected by a novel SSCP strategy.
Hum Mut.
14
1999
156
162
31
Li
A
Swift
M
Mutations at the ataxia-telangiectasia locus and clinical phenotypes of A-T patients.
Am J Med Genet.
92
2000
170
177
32
Maillet
P
Chappuis
PO
Vaudan
G
et al
A polymorphism in the ATM gene modulates the penetrance of hereditary non-polyposis colorectal cancer.
Int J Cancer.
88
2000
928
931
33
Telatar
M
Wang
Z
Udar
N
et al
Ataxia-telangiectasia: mutations in cDNA detected by protein truncation screening.
Am J Hum Genet.
59
1996
40
44
34
Teraoka
SN
Telatar
M
Becker-Catania
S
et al
Splicing defects in the ataxia-telangiectasia gene, ATM: underlying mutations and phenotypic consequences.
Am J Hum Genet.
64
1999
1617
1631
35
Stankovic
T
Kidd
AM
Sutcliffe
A
et al
ATM mutations and phenotypes in ataxia-telangiectasia families in the British Isles: expression of mutant ATM and the risk of leukemia, lymphoma, and breast cancer.
Am J Hum Genet.
62
1998
334
345
36
Easton
DF
Cancer risks in A-T heterozygotes.
Int J Radiat Biol.
66
1994
S177
S182
37
Vorechovsky
I
Luo
L
Ortmann
E
Steinmann
D
Dörk
T
Missense mutations at ATM gene and cancer risk.
Lancet.
353
1999
1276
38
Gatti
RA
Tward
A
Concannon
P
Cancer risk in ATM heterozygotes: a model of phenotypic and mechanistic differences between missense and truncating mutations.
Mol Genet Metab.
68
1999
419
423
39
Maillet
P
Chappuis
PO
Vaudan
G
et al
A polymorphism in the ATM gene modulates the penetrance of hereditary non-polyposis colorectal cancer.
Int J Cancer.
88
2000
928
931
40
Angele
S
Hall
J
The ATM gene and breast cancer: is it really a risk factor?
Mutat Res.
462
2000
167
178

Author notes

Richard S. Houlston, Section of Cancer Genetics, Institute of Cancer Research, Sutton, Surrey, United Kingdom; e-mail:r.houlston@icr.ac.uk.

Sign in via your Institution