Targeting oncogenic protein interactions in Acute Myeloid Leukemia with nucleophosmin (NPM1c) mutations

Acute Myeloid Leukemia (AML) is the most common acute leukemia in adults. While some patients with AML are cured by treatment, most are not; only ~30% of patients live for 5 years after diagnosis. The most effective current treatments are chemotherapy and/or bone marrow transplants, but both have substantial toxicities. Less toxic, more effective treatments are urgently needed.

Cancer is caused by DNA changes (mutations) in cancer cells; these mutations can alter proteins, telling cells to grow when they should not. Treatments that directly block some mutant proteins have been developed. These treatments can kill cancer cells with fewer effects on normal cells; for patients, this can mean fewer side effects.

Nucleophosmin (NPM1) is a gene that is mutated in ~30% of AML patients. In a recent study, we found that AML cells with NPM1 mutations rely on an overabundance of a protein called GAB2. We found that inactivating GAB2 caused cancer cells to stop growing, whereas healthy cells grew normally. This suggests that blocking GAB2 may help AML patients with NPM1 mutations, perhaps with limited side effects.

In this proposal, we will try to find drugs that block the actions of GAB2. We know that GAB2 interacts with many proteins in the cell, and transmits messages that tell the cell to grow. We will determine which of these messages is most important for AML cell growth. We will test a library of 20,000+ drugs to find one (or more) that blocks the critical interaction(s). If successful, we will find drugs that can be developed into treatments for patients with NPM1-mutant AML.

We will also target the mutant NPM1 protein directly. Historically, it has been challenging to define exactly how two proteins physically interact. Recently, however, artificial intelligence (AI) tools have become more accurate at modeling protein interactions. One problem, however, is that while these tools can model an interaction between any pair of proteins; it is hard to know which interactions truly happen. Our experiments identified 207 proteins that interact specifically with mutant NPM1. We will use AI-based modeling (and validation experiments) to understand how NPM1 interacts with these proteins, and AI-based tools to design molecules that block these NPM1 interactions. These may ultimately be developed into drugs that target mutant NPM1 to benefit AML patients.