NPM1 mutations in AML arise from replication errors primed by illegitimate TdT activity.
The involvement of TdT in both NPM1 and FLT3-ITD mutagenesis suggests a significant proportion of AML is a by-product of adaptive immunity.
Nucleophosmin (NPM1) is the most commonly mutated gene in acute myeloid leukemia (AML). AML with mutated NPM1 is recognized as a separate entity in the World Health Organization 2016 classification and carries a relatively favorable prognosis. NPM1 mutations are predominantly 4-bp duplications or insertions in the terminal exon that arise through an unknown mechanism. Here we analyze 2430 NPM1 mutations from 2329 adult and 101 pediatric patients to address their origin. We show that NPM1 mutations display the hallmarks of replication slippage, but lack suitable germline microhomology available for priming. Insertion mutations display G/C-rich N-nucleotide tracts, with a significant bias toward polypurine and polypyrimidine stacking (P < .001). These features suggest terminal deoxynucleotidyl transferase (TdT) primes replication slippage through N-nucleotide addition, with longer syntheses manifesting as N-regions. The recurrent type A, type D, and type B mutations require 1, 2, and 3 N-nucleotide extensions of T, CC, and CAT, respectively, with the last nucleotide used as occult microhomology. This TdT-mutator model successfully predicts the relative incidence of the 256 potential 4-bp insertion/duplication mutations at position c.863_864 over 4 orders of magnitude (ρ = 0.484, P < .0001). Children have a different NPM1 mutation spectrum to adults, including a shift away from type A mutations and toward longer N-regions, consistent with higher TdT activity in pediatric myeloid stem cells. These findings complement our FLT3-ITD data, suggesting illegitimate TdT activity contributes to around one-half of AMLs. AML may therefore reflect the price for adaptive immunity.