Metabolome-wide Association Study of End-Induction Minimal Residual Disease in Children with Acute Lymphoblastic Leukemia
Introduction
Despite intensive multiagent chemotherapy, a substantial proportion of children with acute lymphoblastic leukemia (ALL) fail to clear their disease by the end of induction chemotherapy (i.e., they are end-induction MRD-positive). End-induction MRD positivity is an important predictor of relapse and death. Individually-tailored therapies may improve rates of disease clearance during induction chemotherapy for children with ALL. Metabolomics, the systematic study of the small molecules in a biological system, can be used to understand ALL biology and drug treatment response. To better understand disease characteristics associated with poor treatment response in this group, we are performing metabolome-wide association studies (MWAS) of end-induction MRD positivity in a cohort of children treated for ALL at Texas Children's Hospital (Houston, TX).
Methods
Three hundred ninety-two bone marrow supernatant samples (263 collected at diagnosis and 126 collected at the end of induction chemotherapy) were obtained from N=265 patients with ALL and sent to Metabolon for untargeted metabolomic profiling using the Metabolon Global Discovery Platform. Demographic and clinical data, including age, sex, race, ethnicity, address at diagnosis, ALL immunophenotype, ALL cytogenetic features, end induction MRD status, relapse, and vital status were extracted from the electronic health record. We autoscaled metabolite abundances and imputed missing values with one-half the minimum observed. We computed crude logistic regression models to estimate the odds ratio and 95% confidence interval of MRD positivity for a one standard deviation increase in the abundance of each metabolite at diagnosis and end-induction, and applied a 5% false discovery rate (FDR) correction to account for multiple testing.
Results
The study sample included N=149 males (56.2%) and 116 females (43.8%). Patients were predominantly White (N=174; 66.2%) and Hispanic (N=153; 58.2%). Median age at diagnosis was 5.8 (range 0.8-19.5) years. Hispanic children had an older median age at diagnosis than non-Hispanic children (6.3 versus 5.4 years, p=0.04) and were marginally less likely than non-Hispanic children to be diagnosed with T-ALL (4.0% versus 10.6%, p=0.11). Overall, 19.4% of children were MRD-positive at end-induction; MRD positivity rates were 22.4% among Hispanic children and 14.7% among non-Hispanic children, although this difference was not statistically significant (p=0.27). Rates were similar for White, Black, and Asian/Pacific Islander children. Metabolomic profiling identified 869 unique compounds with known identities in bone marrow supernatant. However, none differed between MRD-positive and MRD-negative patients at diagnosis or the end-induction after FDR correction using the Benjamini-Hochberg procedure.
Conclusion
Ongoing analyses, including metabolite set enrichment analysis and partial least squares-discriminant analysis, may identify broader metabolomic signatures associated with end-induction MRD positivity. However, preliminary results suggest that the metabolomes of MRD-positive and MRD-negative children are similar at diagnosis and end-induction, indicating that other approaches will be required in order to identify those at risk of poor early treatment response and assign targeted therapies.