Abstract
Introduction: Despite cure rates in acute lymphoblastic leukemia (ALL) exceeding 90% in clinical trials, morbidity due to drug toxicities is high. Genetic polymorphisms can influence gene expression and activity, impacting pharmacokinetics and causing inter-individual variation in drug levels, which contributes to toxicity if levels are high or relapse if levels are low. We hypothesize that pharmacogenomic testing will identify patient specific variations in genes involved in metabolism of cytotoxic agents. This knowledge will allow clinicians to optimize therapy by providing pharmacogenomics based biomarkers related to increased toxicities. Data has shown that treatment interruptions and omissions due to toxicities affect outcomes and morbidities in children with cancer.
Objective: To correlate pharmacogenomic biomarkers with toxicity phenotypes in children receiving therapy for ALL.
Methods: This cross-sectional study involved subjects at a tertiary academic center (Fig. 1A). Subjects aged 1 year to 26 years with ALL treated after May 2012 were eligible. A total of 75 patients treated between 2012 and 2020 were included. Pharmacogenomic testing was performed on peripheral blood. Genomic DNA was tested for 118 single-nucleotide polymorphisms (SNP) in 55 genes for transport and metabolism of cytarabine, vincristine, methotrexate, dauno/doxorubicin, and mercaptopurine/thioguanine were analyzed using the Sequenom-based genotyping that uses MALDI-TOF based chemistry. SNPs were tested using logistic regression models for association with toxicities in additive, dominant, and recessive modes of inheritance. CTCAE v4.0 was used for grading all toxicities during the first 100 days of therapy. For endocrine (endo) and neurological (neuro) toxicities, 25 patients exhibited between grade 1-3 toxicities. For gastrointestinal (GI) toxicities, 25 patients exhibited between grade 2-3 toxicities. For hematological (heme) toxicities, 11 patients exhibited between grade 2-4 toxicities. Odds ratio and 95% confidence interval were calculated for each test and SNPs with association P-value <0.05 were considered significant. To conduct multivariable SNP combinations analysis, SNPs with univariate association p-value <0.1 were selected for each toxicity, and then SNP combinations (up to 3 SNPs per model) were tested for association with each toxicity. The combination model with the 1000 permutation p-value <0.05 and lowest BIC value was selected to build a SNP score after considering the mode of inheritance and the direction of association with the toxicity for the individual genotypes. A SNP score was generated by summation of genotype scores for the individual SNPs passing the top model. Patients were further classified into either of the 3 following SNP score groups: >0, 0 or <0.
Results: For a GI toxicity score derived from 3 SNPs (TYMS-rs151264360, FPGS-rs1544105, and GSTM5-rs3754446), patients with >0 score had 79% incidence of GI toxicity (N=67) as compared to 10% in patients with score of 0 and 8% in patients with score <0 (p=2.07E-05, Fig 1B). For the SNP score (AKR1C3-rs6621365, ABCB1-rs4148737, and CTPS1-rs12067645) generated for heme toxicity, higher SNP scores were associated with high toxicity (p=0.003, Figure 1C). For neurotoxicity, the 3 SNPs (DCTD-rs6829021, SCL28A3-rs17343066, and CTPS1-rs12067635) involved were all in cytarabine metabolic pathway and predicted greater neurotoxicity (56%) with a score of >0; no toxicity was observed in patients with neurotoxicity score of <0 (p=0.0005, Fig. 1D). For SNP endo toxicity score (AKR1C3-rs1937840, TYMS-rs2853539, and CTS-rs648743), we observed 91% incidence of endo toxicity in patients with toxicity score of >0 (p=4.7E-08, Fig. 1E). None of the patients with a score of <0 experienced endo toxicity.
Discussion: We identified germline SNPs predictive of toxicity phenotypes in a cohort of 75 subjects with ALL. The results of our multivariable SNP combination analysis suggest susceptibility to chemotherapy-induced toxicities is likely multigenic in nature. Instead of a single SNP approach, identification of combinations of mutations in drug pathways increases the robustness of predicting a patient's response to chemotherapy. Our results provide promising SNP models that can help establish clinically relevant biomarkers allowing for individualization of cancer therapy to optimize treatment for each patient.
No relevant conflicts of interest to declare.
