Background

B-cell precursor acute lymphoblastic leukemia (BCP-ALL) is the most common neoplasm in children. Recent genome-wide association studies (GWAS) on pediatric patients with BCP-ALL have identified common single nucleotide polymorphisms (SNPs) associated with disease susceptibility [1,2,3,4,5,6]. SNPs located in intron 3 of the ARID5B gene (i.e. rs7923074 and rs10821936; Fig. 1) are the most significant and recapitulated SNPs in various races, including Asian populations [7,8,9,10]. ARID5B belongs to the AT-rich interactive domain (ARID) family and acts as a transcription coactivator that binds to the 5′-AATA[CT]-3′ core sequence [11, 12]. Although the direct mechanism for leukemogenesis is not fully understood, the risk allele of susceptibility-linked SNPs in intron 3 of the ARID5B gene may alter the transcription network involved in normal lymphopoiesis by disrupting ARID5B expression [13]. Interestingly, further GWAS on pediatric ALL patients revealed that the other SNPs located in intron 2 of the ARID5B gene (i.e. rs4948488, rs2893881, and rs6479778; Fig. 1) were significantly associated with their relapse rate [14]. This clinical observation suggests that the genotype of these relapse-linked SNPs of ARID5B may be associated with the responses to chemotherapeutic agents. Nevertheless, few studies have focused on the association of ARID5B with drug sensitivities in BCP-ALL [15].

Fig. 1
figure 1

Locations of relapse- and susceptibility-linked SNPs of ARID5B

Full size image

Therefore, to address this issue, we analyzed any association of ARID5B genotype with ARID5B gene expression and drug sensitivity in a series of BCP-ALL cell lines. We found that genotypes of the relapse-linked SNPs of ARID5B are associated with resistance to several chemotherapeutic agents.

Materials and methods

Cell lines

We used 72 BCP-ALL cell lines that were established from Japanese patients as described in detail previously [16] (Additional files 1, 2: Tables S1, S2). Among the 72 cell lines, 15 cell lines were MEF2D fusion-positive, 14 cell lines were BCR/ABL1-positive, 13 cell lines were TCF3/PBX1-positive, 12 cell lines were MLL (KMT2A)-rearranged, 4 cell lines were ETV6/RUNX1-positive, 3 cell lines were TCF3/HLF-positive, and 2 cell lines were BCR/ABL1-like. No hyperdiploid cell lines were included. Forty-six cell lines were sequentially established in our laboratory from 1980 to 2011, while 24 cell lines were provided by 10 institutes. Two additional cell lines were purchased from American Type Culture Collection (ATCC). All cell lines were maintained in RPMI1640 media with 10% fetal calf serum (FCS) at 37 °C under a 21% O2 and 5% CO2 atmosphere.

Real-time reverse transcription polymerase chain reaction (RT-PCR)

Total RNA was extracted from each cell line using TRIzol reagent (Invitrogen, Carlsbad, CA), and reverse transcription reactions were performed using a random hexamer (Amersham Bioscience, Buckinghamshire, UK) and Superscript II Reverse Transcriptase (Invitrogen). To remove unreacted mRNA, the samples were treated with RNase (Invitrogen) after the reaction. Real-time reverse transcription polymerase chain reaction (RT-PCR) analyses of ARID5B were performed using a TaqMan probe kit (Hs01382781_m1). Gene expression level of the beta-actin (ACTB) gene was also examined as an internal control using a TaqMan probe kit (Hs01060665_g1).

SNP genotyping

Genomic DNA was extracted from each cell line using a PureLink Genomic DNA Mini Kit (Invitrogen). Genomic regions containing two representative susceptibility-linked SNPs (rs7923074 and rs10821936) and three representative relapse-linked SNPs (rs4948488, rs2893881 and rs6479778) of ARID5B in 72 BCP-ALL cell lines were amplified using primers described in Table 1. Then, genotypes of five SNPs in each cell line were determined after direct sequencing of each genomic PCR product using forward primers for rs7923074, rs4948488, and rs2893881 as well as a reverse primer for rs10821936 and rs6479778.

Table 1 Primers for SNP genotyping
Full size table

AlamarBlue assay

Fifty percent inhibitory concentration (IC50) values of prednisolone (Pred), dexamethasone (Dex), vincristine (VCR), daunorubicin (DNR), L-asparaginase (L-Asp), cytarabine (AraC), methotrexate (MTX), mercaptopurine (6MP), and mafosfamide [Maf; active metabolite of cyclophosphamide (CY)] were determined using the alamarBlue cell viability assay (Bio-Rad Laboratories, Hercules, CA) as previously reported [17]. Cells (1–4 × 105) were placed onto 96-well flat bottom plates in the presence or absence of seven separate concentrations of each drug in triplicate. The cells were cultured for 44 h to determine the DNR, VCR and CY (Maf) sensitivities and for 68 h to determine Pred, Dex, L‐Asp, MTX, and 6MP; 20 µL of alamarBlue was then added. After incubation for an additional 6 h in the presence of alamarBlue, the optimal density was read on a spectrophotometer at 570 nm using 600 nm as a reference wavelength. Cell viability was calculated by the ratio of the optical density of the treated wells to that of the untreated wells as a percentage. The concentration of each agent required to reduce the viability of the treated cells to 50% of the untreated cells (IC50 value) was calculated and the median IC50 value of three independent assays was determined.

Cell cycle analysis

Each cell line at a density of 0.5 × 105 cells/ml was cultured with fresh RPMI1640 media with 10% FCS for 24 h. Then, cell cycle analysis was performed using flow cytometry after PI staining as previously reported [18]. The median percentage of G0/G1 dormancy phase was determined in three independent analyses.

Statistics

We applied Fisher’s exact test for comparison of allele frequencies between cell lines and Japanese population in HapMap project database (https://www.ncbi.nlm.nih.gov/variation/news/NCBI_retiring_HapMap/). Mann–Whitney U test was always applied for comparisons between two groups of cell lines using R (version 3.5.1) statistical software.

Results

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 [16]. 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 (Additional files 1, 2: Table S1, S2). 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 (Additional files 1, 2: Table S1, S2). We determined genotypes of two representative susceptibility-linked SNPs [14] (rs7923074 and rs10821936, Fig. 1) and three representative relapse-linked SNPs [14] (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).

Table 2 Comparison of SNP allele frequencies between BCP-ALL cell lines and the Japanese population
Full size table

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.

Fig. 2
figure 2

Association of relapse- and susceptibility-linked SNP genotypes with ARID5B gene expression. Relative ARID5B gene expression levels were compared between cell lines with homozygous genotype of risk allele and those with heterozygous or homozygous genotypes of non-risk allele in each SNP. P-value in Mann–Whitney U test is indicated at the top of each SNP

Full size image

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 (Additional file 3: Fig. S1a–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 (Additional file 4: Fig. S2a–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.

Fig. 3
figure 3

Association of relapse- and susceptibility-linked SNP genotypes with sensitivities to VCR (a), CY (b), and AraC (c). Vertical axis indicates log-scaled IC50 values of VCR (a), CY (Maf) (b), and AraC (c). The IC50 values of cell lines with homozygous genotype of risk allele and those with heterozygous or homozygous genotypes of non-risk allele in each SNP were compared. P-value in Mann–Whitney U test is indicated at the top of each SNP

Full size image

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 (Additional file 3: Fig. S1a-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 (Additional file 5: Fig. S3). 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 (Additional file 6: Fig. S4a, 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 (Additional file 6: Fig. S4c-e). These observations suggest that lower ARID5B expression may be a genetic marker for MTX resistance in BCP-ALL.

Fig. 4
figure 4

Association of ARID5B gene expression with sensitivities to MTX (a), VCR (b), CY (c), and AraC (d). Vertical axis indicates log-scaled IC50 values of MTX (a), VCR (b), CY (Maf) (c), and AraC (d). The IC50 values of 36 cell lines with higher ARID5B expression and the other 36 cell lines with lower expression were compared. P-value in Mann–Whitney U test is indicated at the top

Full size image

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 (Additional file 7: Fig. S5). 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 (Additional file 8: Fig. S6). 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.

Discussion

In the present study, using a series of BCP-ALL cell lines, we tried to verify the significance of genotype in the susceptibility-linked and relapsed-linked SNPs of ARID5B with ARID5B gene expression and drug sensitivities. It should be noted that the karyotypes in our cell lines were highly biased in comparison with those in childhood BCP-ALL patients: 14 cell lines were BCR/ABL1-positive, 13 cell lines were TCF3/PBX1-positive, 12 cell lines were MLL (KMT2A)-rearranged, and 3 cell lines were TCF3/HLF-positive. Furthermore, we later discovered that 15 cell lines were positive for MEF2D-fusions [18], which are recently identified fusion genes with a poor therapeutic outcome [19, 20]. In contrast, only four cell lines were ETV6/RUNX1-positive, and no cell lines were hyperdiploid. Thus, the majority of our cell lines were established from BCP-ALL with a poor prognosis.

Using these biased samples, we analyzed associations between genotypes of the relapsed-linked SNPs of ARID5B and the sensitivities to representative drugs for ALL treatment. We analyzed the sensitivities to nine agents, and found that sensitivities to VCR, CY, and AraC were associated with relapsed-linked SNPs; cell lines with homozygous genotypes of risk alleles in the relapsed-linked rs4948488, rs2893881, and rs6479778 were significantly more resistant to VCR, CY, and AraC than cell lines with non-risk alleles. These relapsed-linked SNPs of ARID5B were located in intron 2 of the ARID5B gene. However, no significant association was observed between genotypes of the relapse-linked SNPs of ARID5B and ARID5B gene expression level in BCP-ALL cell lines. Moreover, ARID5B gene expression level was not associated with sensitivities to VCR, CY, and AraC. Thus, genotypes of the relapse-linked SNPs of ARID5B are associated with VCR, CY, and AraC sensitivities of BCP-ALL cell lines independent of their ARID5B expression levels. Additionally, although cell cycle progression is highly associated with the drug sensitivity of leukemia cells, no clear association was observed between cell cycle progression and genotypes of the relapse-linked SNPs of ARID5B in BCP-ALL cell lines. Thus, further analyses are required to clarify the underlying mechanism behind the association between genotypes of the relapse-linked SNPs of ARID5B with the sensitivities to VCR, CY, and AraC in BCP-ALL cell lines.

Regarding the association of ARID5B with drug sensitivities in BCP-ALL cell lines, we also found that lower ARID5B gene expression level was associated with resistance to MTX. This finding seems to be partly consistent with a recent report by Xu et al. [15] who found that ARID5B knockdown in ALL cell lines led to specific resistance to MTX and 6MP. The authors showed that knockdown of ARID5B by using shRNA and CRISPR/Cas9 in ALL cell lines induced partial cell-cycle arrest at G0/G1 phase through upregulation of p21 [15], suggesting that cell-cycle arrest mediated by p21 may be involved in the induction of resistance to MTX and 6MP. However, in the present study, the proportions of G0/G1 phase in cell lines with lower ARID5B expression were similar to those in cell lines with higher ARID5B expression, suggesting that cell cycle progression was not a direct mediator in the association between ARID5B gene expression level and sensitivity to MTX in BCP-ALL cell lines. Further mechanism(s) other than cell cycle progression may be involved in the association of ARID5B expression with MTX sensitivity.

Conclusion

In summary, our observations in 72 BCP-ALL cell lines suggest that the risk allele of the relapse-linked SNPs of ARID5B may be associated with higher relapse rates because of resistance to chemotherapeutic agents such as VCR, CY, and AraC. Moreover, lower ARID5B expression may be associated with MTX resistance. Limitations of the present study include that the study sample was restricted to limited numbers of leukemic cell lines with biased karyotypes and that underlying biological mechanisms for the associations remain unclarified. Since our findings were obtained from leukemia cell lines, further studies are needed before making a firm conclusion.