Single Cell Transcriptomics Predicts Relapse in Infants with Acute Lymphoblastic Leukemia

Care of infants (<1 year of age) diagnosed with MLL (KMT2A)-rearranged acute lymphoblastic leukemia (MLLr-iALL) suffers from two major drawbacks. First, a poor survival rate due to a high rates of early relapse and chemo-resistance. Additionally, the approximately 30-50% of patients that do survive, suffer from life-long, debilitating side-effects of current treatment. While almost all MLLr-iALL patients show an initial promising response to treatment, two-third of the patients relapse, typically within the first year from diagnosis and while still on treatment. Accurate relapse prediction would allow implementation of treatment strategies that take relapse risk into account, with great potential benefit for all patients. Here, we show that Single-cell RNA sequencing (scRNA-seq) can be valuable for risk stratification and that the abundance of chemo-resistant cells within the diagnosis sample might be a powerful indicator of the likelihood of relapse.

We have used scRNA-seq to analyze the response to treatment of leukemic cells in bone marrow biopsies of seven MLLr-iALL patients, expressing either the oncogenic MLL-AF4 or MLL-ENL fusion gene, at the time of initial diagnosis. Three of these patients successfully underwent treatment and remained disease-free during 7 years of follow-up, while in the remaining four cases the disease returned within a year from diagnosis. All samples were subjected to scRNA-seq by FACS index sorting with the aim of identifying differences between early relapsers and long-term survivors.

Quantification of the proportion of cells classified by single cell transcriptomics, categorized as either chemo-resistant or chemo-sensitive, accurately predicts the occurrence of future relapse in individual patients. Strikingly, the single cell-based classification is even consistent with the order of relapse timing. Additionally, leukemic cells associated with high relapse risk are typified by a small phenotype, which coincides with an apparent quiescent gene expression pattern.

This study clearly and, to the best of our knowledge, for the first time shows how disease classification and treatment management can directly benefit from single cell genomics. It demonstrates how classification based on a pivotal functional characteristic of single cells can be performed, despite individual patient variation. Our results shed light on the subpopulation from which leukemic relapse originates, and opens up opportunities for strong, risk-based strategies for future MLLr-iALL treatment regimens.