In total, 687 ALL patients and 971 healthy controls were included in the current study. All the individuals were enrolled from Guangzhou Women and Children’s Medical Center and Nanfang Hospital. Briefly, ALL patients < 18 years old confirmed with a clinical and histological diagnosis were recruited. All patients were newly diagnosed, and a detailed medical history was recorded for each case. Thus, 687 children with ALL were recruited from January 2016 to June 2019. During the same period, 971 cancer-free healthy volunteers were also collected as controls, which were matched to the cases on age, sex, and residential region. The controls were randomly selected from children undergoing a routine physical examination. All included subjects were ethnic Han Chinese. In addition, those with other malignant disorders, a history of chemotherapy or radiotherapy were excluded. The written informed consent was obtained from each subject before participation. This study obtained permission from the institutional review boards of both hospitals.

The SNP selection was performed using data from SNPinfo and NCBI dbSNP databases, and the selection strategy was based on four criteria as described previously (26, 27): (1) the minor allele frequency (MAF) reported in HapMap was >5% for Chinese subjects; (2) located in or near the MYBL2 gene (i.e., < 2kb upstream or downstream of MYBL2); (3) affecting transcription factor binding sites (TFBS) activity, splicing activity or protein-coding; (4) not in high linkage disequilibrium (LD, R² < 0.8). Based on the above criteria, three SNPs (rs285162 C>T, rs285207 A>C, and rs2070235 A>G) in MYBL2 gene were retrieved for further analyses.

Peripheral blood samples were collected from each participant at diagnosis and then used for DNA extraction using the TIANamp DNAKit (TianGen, Beijing, China) according to the manufacturer’s instruction. rs285162, rs285207 and rs2070235 were selected for genotyping. In the genotyping assays, the Taqman ProAmp master mix and pre-designed SNP genotyping assay mix containing polymerase chain reaction (PCR) probes and primers (ABI, Massachusetts, USA) were used. The quantitative real-time PCR method was performed to genotype these three SNPs using QuantStudio™ 6 Flex System (ABI, Massachusetts, USA). In each of the 384-well plates, five positive and five negative controls were included to ensure the accuracy of genotyping. About 10% of samples were selected randomly for direct sequencing to ensure quality control (28, 29), and the results were 100% concordant.

The probability that rs285207 A>C might influence the regulation of MYBL2 was evaluated by using the Roadmap Epigenome Browser (30, 31), TFBIND software (32), and the ENCODE Project (33). Briefly, promoter and enhancer were predicted via histone modification and DNase hypersensitivity (DHS) of GM12878 (lymphocyte cell line) in Roadmap Epigenomics data. TFBIND was used to assess whether rs285207 altered any transcription factor binding sites (TFBS), and then ENCODE ChIP-seq experiments of IKZF1 in GM12878 (Experiment Series: ENCSR816OIY) was used to assess the binding signals and motifs overlapping rs285207.

For each SNP in controls, Hardy-Weinberg equilibrium (HWE) was assessed via the goodness-of-fit χ² test. Genotype distribution of each SNP and demographic variables between the case and control group was analyzed using a 2-sided χ² test. To evaluate the strength of the relation between MYBL2 polymorphisms and ALL susceptibility, odds ratios (ORs) and 95% confidence interval (95% CIs) were calculated using logistic regression analyses, adjusting for age and sex. The false-positive report probability (FPRP) was also computed for each significant finding as previously described (34). A prior probability of 0.1 was adopted to detect an OR of 0.67 for protective effects, and an FPRP value reaching the threshold of <0.2 was considered noteworthy. All statistical analyses were conducted with SAS software (version 9.4; SAS Institute, Cary, North Carolina). In this study, all P values were 2-sided, and a P value of <0.05 was considered as statistical significance.

In the present study, a total of 687 cases and 971 controls were included, and the detailed characteristics are summarized in Table 1 . Briefly, no significant differences were observed in age (P =0.494) or sex (P =0.107) distribution between cases and controls. According to immunophenotype-based classification (35–37), 596 (86.75%) cases were diagnosed with B cell ALL (B ALL), including 227 (33.04%) pro B ALL, 200 (29.11%) common B ALL, 166 (24.16%) pre B ALL, and 3 (0.44%) mature B ALL; 61 (8.88%) diagnosed with T cell ALL (T ALL); 30 (4.37%) with no available data (NA). Besides, information about gene infusion, risk level, karyotype, relapse, and the levels of minimum residual disease (MRD) at multiple time points post-therapy were also included in Table 1 .

According to the SNP selection strategy, three SNPs (rs285162 C>T, rs285207 A>C, and rs2070235 A>G) that overlapped with transcription factor binding site (TFBS), splicing regulating site (SRS), or non-synonymous SNP (nsSNP) were selected ( Table 2 ). The genotype frequencies of MYBL2 gene SNPs in all 687 cases and 971 controls and their association with ALL risk were described in Table 3 . All these three SNPs were in HWE (P _HWE >0.05) among the controls. Of the three SNPs, significant differences were observed for rs285207 A>C (crude P = 0.017) between ALL cases and controls in a dominant model. After adjustment with age and sex, rs285207 C allele was significantly related to a decreased ALL risk in the dominant model (AC/CC vs AA: adjusted OR = 0.78; 95% CI = 0.63-0.97, P = 0.022). The rest two genotypes (rs285162 C>T and rs2070235 A>G), however, were not significantly associated with ALL risk.

To further explore the association between the MYBL2 gene rs285207 A>C polymorphism and ALL susceptibility, subgroup and stratification analyses were performed in terms of age, sex, karyotype, immunophenotype, gene infusion, risk level, and other clinical information. All the results were shown in Table 4 . Compared with the rs285207 AA genotype, the AC/CC genotype decreased ALL risk in females (adjusted OR=0.68; 95% CI=0.47-0.97, P=0.033), and children < 10 years old (adjusted OR=0.78; 95% CI=0.62-0.98, P=0.030). For the immunophenotype, the rs285207 AC/CC genotype decreased ALL risk in patients with B ALL (adjusted OR=0.78; 95% CI=0.62-0.98, P=0.035), particularly in patients with pre B ALL (adjusted OR=0.56; 95% CI=0.38-0.84, P=0.004). The rs285207 AC/CC genotype also decreased ALL risk in children with normal diploid (adjusted OR=0.73; 95% CI=0.57-0.95, P=0.017), with low risk (adjusted OR=0.68; 95% CI=0.49-0.94, P=0.021), with lower WBC (adjusted OR=0.62; 95% CI=0.43-0.87, P=0.007) or lower platelet level (adjusted OR=0.76; 95% CI=0.59-0.96, P=0.023). Furthermore, the rs285207 AC/CC genotype reduced ALL risk in patients without relapse (adjusted OR=0.78; 95% CI=0.61-0.99, P=0.046). The level of MRD is a key prognosis factor for ALL patients, and the negative MRD (<0.01%) usually indicates a good prognosis. In this study, MRD was also analyzed according to different therapy strategies ( Table S1 ). In patients treated with China Children’s Cancer Group (CCCG)-ALL-2015 protocol, no association was observed between MRD level and rs285207 polymorphism. In the group of South China Children Leukemia Group (SCCLG)-ALL-2016 protocol, the rs285207 AC/CC genotype seemed to increase the rates of patients with negative MRD (<0.01%) (adjusted OR= 3.10, 2.67 and 999 at day 19, day 35 and week 12, respectively), although no significant result was found due to the limited sample size.

The FPRP values for these remarkable results at different levels of prior probability and statistical power were shown in Table 5 . In the FPRP analysis, all the statistically significant findings were noteworthy at the prior probability level of 0.25 (FPRP<0.200). At the prior probability of 0.1, the association between the rs285207 AC/CC genotype and decreased ALL risk was still noteworthy (FPRP=0.128, statistical power=0.903), especially in patients with pre B ALL (FPRP=0.191, statistical power=0.469), normal diploid (FPRP=0.190, statistical power=0.738), lower WBC (FPRP=0.135, statistical power=0.327) or lower platelet level (FPRP=0.183, statistical power=0.855), which further strengthen the significant results above.

To explore the potential mechanisms by which rs285207 influences the ALL risk, we evaluated the probability of rs285207 polymorphism altering transcription regulation of MYBL2 . The Roadmap Epigenomics data showed that rs285207 overlapped DHS marks and histone modifications related to both promoter and enhancer in multiple tissue types ( Figure 1A ). This observation was further supported by H3K4me1, H3K4me3, H3K9ac, H3K27ac and DHS ChIP data in GM12878 cells ( Figure 1B ). TFBIND analysis revealed that rs285207 altered the binding affinity of this site to transcription factors including IKZF1, THAP4 and FOXA2. For the rs285207 risk allele A, all of IKZF1, THAP4 and FOXA2 have high TFbind scores (0.85, 0.77 and 0.86, respectively), while none of these three bindings was found for the non-risk allele C. In ENCODE ChIP-seq analysis, however, only IKZF1 was found to bind to DNA motif overlapping rs285207 ( Figure 2A ). IKZF1 has a higher preference for the risk allele A and leave no chance for allele C ( Figure 2B ). The results showed that rs285207 A>C might influence the transcription of MYBL2 via disrupting IKZF1 binding site.