Cytarabine Resistance Leads to a Specific Mutator Phenotype in Leukemic Cells

The mechanism(s) behind chemoresistance in relapsed T-cell acute lymphoblastic leukemia (T-ALL) patients are incompletely understood. To study this question, we utilized a strain that is hypomorphic for the DNA replication factor Mcm2. Mcm2^hypo mice invariably develop T-ALL due to acquired DNA copy number variations (CNV) involving tumor suppressor or oncogenes. Since Mcm2^hypo cells are prone to CNV, we hypothesize that treating Mcm2^hypo cells with specific chemotherapy agents may reveal CNV that lead to chemoresistance, including both gains and losses. This approach may be more informative than classical CRISPR/Cas9 screens, which typically identify loss of function events. We selected six drugs with different mechanisms of action that are known to be active against T-ALL, including cytarabine, etoposide, melphalan, methotrexate, prednisone, and vincristine. We used 4 mcm2 T-ALL cell lines (designated 2883, 2696, 2641 and 2869). After 1 year of treatment, we obtained 12 independent resistant cell lines that could tolerate drug concentrations up to 10,000X that of parental cell lines. These samples were characterized by sparse whole genome sequencing (WGS) to determine CNV, RNA-Seq and whole exome sequencing (WES).
Both vincristine and etoposide resistant cell lines, had copy number gain of a region encompassing several genes, including Abcb1a (also known as Multi-Drug Resistance gene 1a). Two prednisone resistant cell lines had similar copy number loss of Nr3c1, a gene involved in glucocorticoid metabolism. A methotrexate resistant cell line had copy number gain of dihydrofolate reductase (DHFR), a key enzyme for intracellular folate metabolism. These are all known or suspected mechanisms for resistance to the respective classes of drugs and gave us confidence that the screen was working as designed. Two cytarabine resistant cell lines, 2883CR and 2869CR had bi-allelic mutations in Dck, the rate-limiting enzyme in the cytidine salvage pathway. In both cases, one allele was deleted while the second had a splice site/region mutation leading to intron retention and resultant frameshift. Western blotting showed absence of WT Dck protein in the resistant cell lines.
WES revealed acquired single nucleotide variants (SNV) in both the parental (2883 and 2869) and Dck mutant (2883CR and 2869CR) Mcm2^hypo cell lines. The parental Mcm2^hypo cell lines (2883 and 2869) showed a unique mutation signature that is not present in the COSMIC database, characterized by C>T, T>C, and T>G mutations, each in a specific trinucleotide context. Remarkably, the cytarabine resistant clones (2883CR and 2869CR) showed a >10 fold exaggeration of this signature, with up to 100 acquired SNV per million base pairs. In addition to the specific C>T, T>C, and T>G mutations, the 2883CR and 2869CR cell lines also showed C>G mutations in a specific 5’-GCC-3’ context. Further studies of untreated 2883CR and 2869CR cells demonstrated that ongoing cytarabine exposure was not required for the mutational signature, but that untreated 2883CR or 2869CR cells showed ongoing accumulation of mutations in the specific trinucleotide contexts described above. Using CRISPR/Cas9 to inactivate Dck, we found that we could re-create this unique signature (C>T, T>C, T>G, and C>G in specific trinucleotide contexts) in Mcm2^hypo cell lines by Dck inactivation.
In summary, we identified three mechanisms responsible for Dck inactivation associated with cytarabine resistance: copy number loss, splice site mutations, and splice region mutations. Moreover, Dck inactivation results in a specific, previously unrecognized, mutational signature. These findings add to emerging data that exposure to specific chemotherapy agents can produce specific mutational signatures.