Genotype of susceptibility-linked and relapse-linked SNPs of ARID5B in BCP-ALL cell lines
We first analyzed ARID5B genotypes in 72 BCP-ALL cell lines established from Japanese patients [28]. Our cell line bank contained 15 MEF2D fusion-positive, 14 BCR/ABL1-positive, 13 TCF3/PBX1-positive, 12 MLL (KMT2A)-rearranged, 4 ETV6/RUNX1-positive, 3 TCF3/HLF-positive, and 2 BCR/ABL1-like cell lines, but no hyperdiploid cell lines (Supplemental Table). Thus, the majority of our cell lines had been established from BCP-ALL with high or intermediate risk karyotypes. Among 65 cell lines with basic records of cell line establishment, 28 and 37 cell lines were established from the samples at diagnosis and those at relapse, respectively (Supplemental Table). We determined genotypes of two representative susceptibility-linked SNPs [26] (rs7923074 and rs10821936, Fig. 1) and three representative relapse-linked SNPs [26] (rs4948488, rs2893881, and rs6479778, Fig. 1) in each cell line after direct sequencing of each genomic PCR product. Allele frequencies of each SNP in BCP-ALL cell lines were in Hardy–Weinberg equilibrium. Due to linkage disequilibrium, genotypes of rs7923074 and rs10821936 were identical in 71 of 72 cell lines. Genotypes of rs2893881 and rs6479778 were also identical in 71 cell lines. In the HapMap project database (Table 2), we compared the allele frequency of each SNP between our cell lines and the Japanese population, but no significant differences were observed in the genotypes of both the susceptibility-linked SNPs and the relapse-linked SNPs of ARID5B. We also compared the allele frequency of each SNP between 28 cell lines established at diagnosis and 37 cell lines established at relapse, but no significant differences were observed (data not shown).
No association of susceptibility or relapse-linked SNPs of ARID5B with ARID5B expression
Since both the susceptibility-linked SNPs and the relapse-linked SNPs of ARID5B are located in intronic regions, we next performed an expression quantitative trait locus (eQTL) analysis. We quantified the ARID5B gene expression level in each cell line by real-time RT-PCR using ACTB gene expression as an internal control. However, in the eQTL analysis of 72 BCP-ALL cell lines, neither genotypes of susceptibility-linked rs7923074 and rs10821936 nor those of relapse-linked rs4948488, rs2893881, and rs6479778 were significantly associated with ARID5B expression level (Fig. 2). These observations demonstrated that genotypes of both susceptibility-linked SNPs and relapse-linked SNPs of ARID5B were not clearly associated with ARID5B expression levels in the BCP-ALL cell lines.
Association of relapse-linked SNPs of ARID5B with drug sensitivity
Next we verified whether the genotypes of relapse-linked SNPs of ARID5B in BCP-ALL cell lines were associated with their sensitivities to chemotherapeutic agents. We performed an alamarBlue assay to determine IC50 values (concentration required to kill 50% of the cells) of nine representative chemotherapeutic agents [Pred, Dex, VCR, DNR, L-Asp, AraC, MTX, 6MP, and CY (Maf)] used for children with ALL. Of note, IC50 values of VCR (median IC50: 39.6 ng/ml) in 12 cell lines with homozygous risk allele (C) genotypes in the relapse-linked rs4948488 were significantly higher (p=0.031 in Mann–Whitney U test) than those (1.04 ng/ml) in 60 cell lines with heterozygous or homozygous genotypes of the non-risk allele (T) (Fig. 3a). In addition to VCR, sensitivities to CY (Maf) (Fig. 3b) and AraC (Fig. 3c) tended to be associated with the genotype of the relapse-linked rs4948488. Similar associations were observed in genotypes of rs2893881 and rs6479778 (Fig. 3a-c). IC50 values of six agents (Dex, Pred, DNR, L-Asp, MTX, and 6MP) were not significantly associated with genotypes of the relapse-linked rs4948488, rs2893881, and rs6479778 (Supplemental Fig. 1a-f). Considering that BCR/ABL1-positive and BCR/ABL1-like ALL are characteristic entities, we analyzed the association in 56 BCP-ALL cell lines excluding 14 BCR/ABL1-positive and 2 BCR/ABL1-like ALL cell lines (Supplemental Fig. 2a-c). Similar associations were observed between genotypes of the relapse-linked SNPs of ARID5B and the sensitivities to VCR, CY (Maf), and Ara-C in BCR/ABL1-negative and BCR/ABL1-like-negative ALL cell lines.
We further analyzed any association of the susceptibility-linked rs7923074 and rs10821936 with drug sensitivities. In contrast to the genotypes of the relapse-linked SNPs, no significant associations were observed in genotypes of rs7923074 and rs10821936 with sensitivities to VCR, CY and AraC (Fig. 3a-c) and the other six chemotherapeutic agents (Supplemental Fig. 1a-f). These observations suggest that the risk allele of relapse-linked SNPs (but not susceptibility-linked SNPs) may be associated with a higher relapse rate in pediatric BCP-ALL patients due to reduced sensitivities to VCR, CY and AraC.
Association of ARID5B gene expression with drug sensitivity
Finally, we verified whether gene expression level of ARID5B was associated with drug sensitivities of BCP-ALL cell lines. To address this issue, we simply divided our 72 BCP-ALL cell lines into two groups—36 cell lines with higher than median value gene expression levels and 36 cell lines with lower than median value gene expression levels—and compared the IC50 values of each drug. Of note, the IC50 values of MTX in 36 cell lines with lower ARID5B expression (median IC50: 37.1 ng/ml) was significantly higher (p=0.023 in Mann–Whitney U test) than those in the other 36 cell lines with lower expression (16.9 ng/ml) (Fig. 4a). A similar trend was observed in 56 BCP-ALL cell lines, excluding 14 BCR/ABL1-positive and 2 BCR/ABL1-like ALL cell lines (Supplemental Fig. 3). In contrast, although the sensitivities to VCR, CY, and AraC were associated with genotypes in the relapse-linked SNPs of ARID5B, no significant differences were observed in the IC50 values of VCR, CY, and AraC between the two groups (Fig. 4b-d). Furthermore, although genotypes in the susceptibility-linked SNPs of ARID5B were associated with sensitivities to Pred and Dex, no significant differences were observed in the IC50 values of Pred and Dex between the two groups (Supplemental Fig. 4a, b). In the IC50 values of the remaining three agents (DNR, L-Asp, and 6MP), there were no statistically significant differences between the two groups (Supplemental Fig. 4c-e). These observations suggest that lower ARID5B expression may be a genetic marker for MTX resistance in BCP-ALL.
Association of ARID5B SNPs with cell cycle progression
Since anti-leukemic activity of chemotherapeutic agents are dependent on cell cycle progression, we analyzed the association of the susceptibility-linked and relapsed-linked SNPs of ARID5B with cell cycle progression in BCP-ALL cell lines. First, we compared cell cycle progression between cell lines with homozygous genotype of risk allele in each relapsed-linked SNP and those with the non-risk allele (Supplemental Fig. 5). No significant associations were observed in genotypes of the susceptibility-linked and relapsed-linked SNPs of ARID5B with cell cycle progression. Next, we compared cell cycle progression between the 36 cell lines with higher than median value ARID5B gene expression levels and the other 36 cell lines with lower than median value gene expression levels (Supplemental Fig. 6). No differences were observed between the two groups. These observations suggest that cell cycle progression may not directly be involved in the association of SNP genotype and gene expression level of ARID5B with drug sensitivities in BCP-ALL cell lines.