Chromosomal alterations among age-related haematopoietic clones in Japan

Abstract

The extent to which the biology of oncogenesis and ageing are shaped by factors that distinguish human populations is unknown. Haematopoietic clones with acquired mutations become common with advancing age and can lead to blood cancers^1-10. Here we describe shared and population-specific patterns of genomic mutations and clonal selection in haematopoietic cells on the basis of 33,250 autosomal mosaic chromosomal alterations that we detected in 179,417 Japanese participants in the BioBank Japan cohort and compared with analogous data from the UK Biobank. In this long-lived Japanese population, mosaic chromosomal alterations were detected in more than 35.0% (s.e.m., 1.4%) of individuals older than 90 years, which suggests that such clones trend towards inevitability with advancing age. Japanese and European individuals exhibited key differences in the genomic locations of mutations in their respective haematopoietic clones; these differences predicted the relative rates of chronic lymphocytic leukaemia (which is more common among European individuals) and T cell leukaemia (which is more common among Japanese individuals) in these populations. Three different mutational precursors of chronic lymphocytic leukaemia (including trisomy 12, loss of chromosomes 13q and 13q, and copy-neutral loss of heterozygosity) were between two and six times less common among Japanese individuals, which suggests that the Japanese and European populations differ in selective pressures on clones long before the development of clinically apparent chronic lymphocytic leukaemia. Japanese and British populations also exhibited very different rates of clones that arose from B and T cell lineages, which predicted the relative rates of B and T cell cancers in these populations. We identified six previously undescribed loci at which inherited variants predispose to mosaic chromosomal alterations that duplicate or remove the inherited risk alleles, including large-effect rare variants at NBN, MRE11 and CTU2 (odds ratio, 28-91). We suggest that selective pressures on clones are modulated by factors that are specific to human populations. Further genomic characterization of clonal selection and cancer in populations from around the world is therefore warranted.

Extended Data Fig.1|

Age and sex of carriers of mosaic event types. Mean age and sex of carriers of specific mCA types (defined by chromosome and copy number) with at least 100 carriers in the 179,417 subjects. Marker sizes are proportional to mCA frequencies. Error bars, s.e.m. Numeric data are provided in Supplementary Table 7.

Extended Data Fig.2|

Comparable chromosomal coverage by heterozygous genotypes in BBJ and UKB. Average numbers of heterozygous genotyped sites (averaged across individuals) in each 1Mb region of the genome for the BioBank Japan and UK Biobank genotyping arrays.

Extended Data Fig.3|

Similar breakpoint distributions of CN-LOH events in the BBJ and UKB. Relative frequencies of estimated CN-LOH breakpoint locations in BioBank Japan and UK Biobank. Breakpoints were smoothed over +/−2Mb to allow plotting of frequency curves, which were rescaled to 1.

Extended Data Fig.4|

Quantile-quantile plots of mosaic events with significant associations demonstrating no inflation of association statistics. Quantile-quantile plots of results for mosaic events with significant associations are indicated. Analytic results of Fisher’s exact test (two-sided, nominal p-values) using 173,599 subjects are shown. We defined as hit loci 42–49M at chr1 (1p CN-LOH), 88–94M at chr8 (8q CN-LOH), 92–96M at chr11 (11q CN-LOH), 88–90M at chr16 (16q CN-LOH), 23–26M and 100–103M at chr14 (cis association of 14q CN-LOH), 4–6M at chr9 (9p CN-LOH), 0–2M at chr5 (trans association of 14q CN-LOH) and 1–3M at chr7 (trans association of chr15 gain).

Extended Data Fig.5|

Local plots for cis and trans associations. Associations of inherited variants with a. 8q CN-LOH, b. 11q CN-LOH, c. 16q CN-LOH, d. chr15 gain, e.1p CN-LOH, f.9p CN-LOH, and g-i. 14q CN-LOH are shown for regions containing the NBN, MRE11, CTU2, MAD1L1, MPL, JAK2, NEDD8/TINF2, DLK1 and TERT loci, respectively. a-c and e indicate rare cis associations, f-h indicate common cis associations and d and i indicate trans associations. a-d and g-i are unreported regions. Purple points indicate lead variants. Other variants are color-coded according to LD with lead variants. The TCL1A variant that significantly associated with 14q CN-LOH allelic imbalance is not shown here because it did not significantly associate with 14q CN-LOH risk. Analytic results of Fisher’s exact test (two-sided, nominal p-values) using 173,599 subjects are shown.

Extended Data Fig.6|

Action of CN-LOH events on rare and common inherited variants. Schematic images showing the patterns of selection or elimination of inherited variants by CN-LOH events. Asterisks indicate risk alleles. (For the TCL1A locus, which did not significantly associate with presence of 14q CN-LOH, we depict TCL1A as a gene for which mCAs select an allele.)

Extended Data Fig.7|

Examples of multiple overlapping CN-LOH clones in a single chromosome. We identified 185 individuals carrying multiple CN-LOH clones on a single chromosome. a.Multiple clones were observed in at least one individual for all chromosomes except chromosomes 18, 20 and 22. The plots show phased BAF deviations (y-axis) as a function of chromosome position (x-axis) for the individual with the largest clone per chromosome (among all individuals with multiple CN-LOH clones on that chromosome). Colored horizontal lines of different colors indicate distinct BAF deviations corresponding to overlapping CN-LOH events. b.the number of subjects carrying multiple CN-LOH clones on a single chromosome is shown in each chromosomal arm.

Extended Data Fig.8|

Mortality risk conferred by mosaic chromosomal alterations. a. Risk of mortality from various causes conferred by presence of an mCA at >1% cell fraction. Leukemia, malignant lymphoma, and multiple myeloma are subdivisions of blood cancer. Cardiovascular mortality includes deaths from coronary artery disease (CAD) and ischemic stroke (IS). b. Risk of leukemia mortality conferred by specific mCAs (grouped by chromosomal location and copy-number change) reaching Bonferroni significance. c. Risk of leukemia mortality conferred by mosaic status stratified by mosaic cell fraction. d. Risk of leukemia mortality conferred by mosaic status stratified by mosaic cell fraction and number of mosaic events detected (one vs. two or more). All analyses were restricted to individuals with no previous cancer diagnosis and were corrected for age, sex, smoking status, and genotyping array (Methods). Error bars, 95% CIs. Numeric data are provided in Supplementary Tables 24–27. Results using 86,546 subjects are indicated. Cox Proportional Hazard model (two-sided) is applied for a, b, and d. Cochran-Mantel-Haenszel test is applied for c.

Figure 1.. Genomic locations of 33,250 autosomal mCAs detected in 27,910 unique BioBank Japan participants.

Loss, CN-LOH, and gain events are plotted as blue, orange, and red horizontal lines, respectively. Events with undetermined copy number are plotted in grey. Commonly deleted regions are labeled in blue; loci associated with CN-LOH mutations in cis are labeled in orange.

Figure 2.. Classification of mCAs, frequency as a function of age, and comparison of genomic distributions between BBJ and UKB.

a. Classification of mCAs as loss, CN-LOH, or gain events using log R ratio (LRR, measuring total DNA abundance) and B allele frequency deviation from 0.5 (|ΔBAF|, measuring allelic imbalance; Methods). Unclassified events are indicated in grey. b. Frequency of detectable mosaicism stratified by age and sex. Frequencies (means) and error bars for 95% CIs are indicated among 179,417 subjects analyzed. c,d. Distribution of mCAs by chromosome and copy number in BioBank Japan and UK Biobank. e. Chromosomal coverage of loss and CN-LOH events in BioBank Japan and UK Biobank. Curves indicate frequencies at which each chromosomal position is contained in loss (resp. CN-LOH) events, normalized to 1 on each chromosome. Numeric data are provided in Supplementary Tables 6 and 13.