A germ-line deletion of APOBEC3B does not contribute to subtype-specific childhood acute lymphoblastic leukemia etiology

Approximately 4/100,000 children are diagnosed with acute lymphoblastic leukemia (ALL) in the United States annually. Early life exposures related to immune priming (i.e. vaginal birth, daycare attendance, and high birth order)1 and having fewer infections requiring medical treatment2 have been inversely associated with disease, suggesting an etiological role of infectious agents, perhaps via dysregulation of the immune system.

Patterns of somatic mutation in ALL tumors give further insight into disease etiology. An innate immune enzyme, APOBEC3B (Apolipoprotein B mRNA-editing enzyme, catalytic polypeptide-like 3B), inhibits viral infection by inducing TpC>T nucleotide changes in foreign nucleic acid. This mutation signature has been identified in tumor genomes of several cancer types with known (cervical, head and neck, stomach3) or hypothesized (breast4) infectious etiologies, and is attributed to aberrant enzymatic activity of APOBEC3B. The APOBEC3B point mutation signature is also predominant in ALL,5 but with subtype specificity. The signature is present in ETV6-RUNX1 fusion ALL6 but absent in high hyperdiploid ALL.7 The high expression of APOBEC3B in lymphoblasts further justifies examination of functional APOBEC3B polymorphisms in ALL etiology.

An ~30kb germline deletion polymorphism at the APOBEC3B locus has been associated with increased risk of several cancers that bear the APOBEC3B mutation signature8 and studies have shown that the deletion transcript yields an enzyme with a higher in vitro propensity for collateral genomic DNA damage than its wild-type counterpart.9 In fact, the signature TpC>T point mutation burden is higher in the tumors of ALL and breast cancer patients carrying the germline APOBEC3B deletion compared with those without.9 The deletion is common in populations of Native American ancestry (approx. 60%), and relatively rare in Europeans and Africans (6% and 0.9%, respectively).10 Hispanic children, whose genetic ancestry is typically comprised of a mixture of Native American, European, and African ancestries, are at the greatest risk for developing ALL in the United States.11 While it has been suggested that the APOBEC3B deletion polymorphism contributes to the patterns of somatic mutations observed in ALL tumors,9 it is not known whether the variant contributes to disease risk. Here, we report results from the first association study of germline APOBEC3B variants in childhood ALL risk.

The APOBEC3B deletion genotype was assessed in California Childhood Leukemia Study (CCLS) case and control subjects (see Online Supplementary Methods for enrollment details) with a polymerase chain reaction (PCR)-based assay (n=1126). The deletion was tested for association with childhood ALL status overall and within ETV6-RUNX1 fusion and high hyperdiploid ALL subtypes. Overall, controls tended to be wealthier than cases with a higher proportion self-reporting as white and non-Hispanic (Table 1).

Table 1.CCLS study participants’ characteristics.

APOBEC3B deletion copy number was detected using a validated polymerase chain reaction (PCR) method described previously.10 A total of 518 ALL cases and 608 controls were genotyped using this PCR assay, with copy number (homozygote wild-type, heterozygote, and homozygote deletion) determined from agarose gel electrophoresis results (Online Supplementary Figure S1). A χ test for Hardy-Weinberg equilibrium was performed; the null hypothesis of deletion equilibrium was accepted among controls after stratifying by Hispanic versus non-Hispanic ethnicity (P=0.45 and 0.45, respectively). Ethnic heterogeneity in the CCLS population is supported by the distribution of multidimensional scaling (MDS) components compared to reference populations (Online Supplementary Figure S2).

After adjusting for global genetic ancestry (first 3 MDS components), no association was observed between the APOBEC3B deletion and overall ALL risk for the additive, dominant, or recessive models, nor after stratification by cytogenetic subtype (Table 2). When study subjects were stratified by self-reported Hispanic status, results did not change (Online Supplementary Table S1).

Table 2.Odds ratios for the association between the APOBEC3B deletion polymorphism and risk of childhood acute lymphoblastic leukemia (ALL) overall and stratified by cytogenetic subtype.

Previous studies have identified SNPs within the APOBEC3 region that are associated with cancer risk independent of the APOBEC3B deletion.128 Thus, we tested all SNPs across the APOBEC3 gene region (chr22:39,200,000-39,650,000) that passed quality filtering (8275 SNPs) for association with ALL in 1083 cases and 1137 controls. After controlling for genetic ancestry, no variant reached statistical significance after correcting for multiple testing (Online Supplementary Figure S3). A SNP ~20Kb upstream of APOBEC3B and previously associated with bladder cancer,8 rs1014971, was not associated with ALL risk [Odds Ratio (OR) 0.91, P=0.33]. The top association was seen for rs73424730 (OR 1.35, P=0.004), a SNP ~ 100Kb downstream of the APOBEC3H gene.

To determine whether the observed absence of association between the APOBEC3B deletion and ALL was due to confounding by genetic ancestry, local ancestry at the APOBEC3 megalocus was inferred using a discriminative modeling approach, RFMix.13 The APOBEC3B deletion polymorphism is a highly population-stratified genetic variant.10 Thus, in the admixed CCLS population, it is possible that adjustment for global genetic ancestry was insufficient, resulting in residual confounding or a washing-out of the true effect of interest. After adjusting for regional ancestral proportions for four continental ancestries, there remained no association between the APOBEC3B deletion variant and childhood ALL overall (Online Supplementary Table S2). To ensure that effect heterogeneity was not washing-out true associations, RFmix-assigned genetic ancestries (African, Native American, European, and East Asian) at each SNP in the APOBEC3 locus were tested for independent association with ALL to determine whether local genetic ancestry was associated with disease. African ancestry at the APOBEC3 megalocus was nominally associated with ALL risk, but did not reach statistical significance after correcting for multiple tests (Online Supplementary Figure S4

Despite the previously observed presence of the APOBEC3B-mediated point mutation signature in the tumor genomes of a subset of ALL patients, and a higher burden of point mutations among APOBEC3B deletion carriers, we found no apparent relationship between the germline APOBEC3B deletion and risk of developing the disease. Moreover, we found no evidence of association between any SNPs across the APOBEC3 gene region and childhood ALL, including at a locus previously associated with the APOBEC mutation signature in bladder cancer.14

Evidence from studies of other cancers suggests a complex relationship between germline variation at the APOBEC3 gene region and tumorigenesis. For instance, the deletion is associated with an increased risk of breast cancer9 but appears protective in bladder cancer.8 Further, the precise contribution of germline APOBEC polymorphisms to the presence of the APOBEC mutation signature in an individual tumor has yet to be determined.15 There is evidence that the APOBEC3B deletion polymorphism changes the mutagenic behavior of the enzyme;9 however, the presence of the mutational signature in a tumor genome does not imply that the deletion is present, as the signature could arise by some other means (i.e. increased enzyme expression by other mechanisms).

Childhood ALL is a rare and heterogeneous disease, and more prevalent in admixed populations, making the study of any potential germline genetic risk factors challenging. While it is unlikely that the APOBEC3B deletion polymorphism is a strong independent risk factor for disease overall, our sample size was limited in statistical power to assess small effects, especially following stratification by cytogenetic subtype. Associations observed in other cancers suggest the deletion alters disease risk by 20–30%.8 These cancers have mixed somatic mutation signatures, suggesting mutagenesis occurs from multiple sources.5 Conversely, the APOBEC mutational signature is the only one observed in ALL other than common, spontaneous cytosine deamination;5 therefore, we hypothesized a moderate association with the APOBEC3B deletion would be present in ALL. Further, the CCLS study population is representative of cases occurring in the state of California, and reflects the substantial racial and ethnic diversity to be found there. Thus, residual confounding by genetic admixture remains a challenge in investigating this variant, which differs significantly in frequency across populations based on ancestral origin. However, stratifying by self-reported Hispanic status in this study yielded no apparent association in either group. To further account for potential confounding effects of admixture, local ancestry at the APOBEC3 megalocus was estimated. Adjusting for local ancestral proportions did not change the observed null association of the APOBEC3B deletion polymorphism with ALL risk, nor were any of the four observed regional genetic ancestries associated independently with ALL risk.

The apparent lack of association between the germline deletion of APOBEC3B and risk of ALL does not provide support that this mutagen, active in tumor cells, is a driver of tumorigenesis. In a previous study, our group showed expression of double-stranded (ds)DNA viruses in primary, treatment naïve, childhood ALL tumors.16 dsDNA viruses are a primary target of the APOBEC3B enzyme. Transient expression of these viruses in leukemic cells could thus produce a passenger-type APOBEC3B point mutation signature in ALL. A virus-mediated induction of APOBEC3B expression may be dominant over the impact on APOBEC3B expression by the polymorphism studied here. Alternatively, the APOBEC mutational signature may reflect an infectious etiology of ALL in a subset of cases with heritable predisposition occurring in unrelated genes. Experimental studies in mice have shown that PAX5 mutation, a predisposing risk factor in some childhood ALL cases, can result in development of ALL following exposure to common infection.17 Though there is no apparent relationship between the germline APOBEC3B deletion polymorphism and ALL risk, delineating a potential role for the polymorphism in tumor progression may have treatment implications, warranting further study.