A cancer cell acts differently than a healthy cell, with unique features enabling its quick growth and quality control bypasses. Rice University’s Natasha Kirienko is interested in finding these unique features and exploiting them to advance cancer treatments. In a paper recently published in the journal Leukemia, her team used a new screening method to identify drugs that could target these unique features, or vulnerabilities, of a cancer cell without unduly harming healthy cells.
“We developed a screening method to identify potential drugs against acute myeloid leukemia,” said Kirienko, a professor of biosciences and corresponding author. “This method uses machine learning to essentially work backwards from the traditional drug screening approach. With it, we identified not only new potential drug compounds but a new drug target specific to AML cells.”
To do this, the research team identified three functions found only in successful drugs selected in previous, wet lab-based screens that tested drugs against acute myeloblastic leukemia (AML) cells. Then they used a machine learning-based software and asked it to select compounds most likely to have the three identified functions. The program was able to provide predictions for over four million drug compounds with less than 100 meeting selection criteria.
“We were looking for compounds that caused increased cell death, or apoptosis, increased organellar recycling, or autophagy, and that inhibited a protein called glutathione reductase,” said Bernadetta Meika, a graduate student and co-author on this paper. “Our computer models identified just under a hundred molecules that could do these functions. We picked the best 20 to test in the lab.”
When the team looked further into the glutathione reductase protein, they found that inhibiting it was actually exploiting a specific vulnerability of the cancer cells. Leukemia cells have high energy needs and thus heavily rely on mitochondria, the powerhouses of the cells. As mitochondria produce energy, they also make harmful reactive oxygen species, or ROS, that must be neutralized upon production. Glutathione reductase proteins are a key part of the neutralization process. When they are inhibited, a cell’s ability to neutralize ROS is significantly reduced.
In a healthy cell, secondary neutralization methods can be used to make up for the inhibited glutathione reductase. In AML cells, however, the high energy needs mean the amount of ROS produced is more than the cell’s backup methods can handle. The excess ROS damages then kills the cell.
“In a typical in silico, or computer-based, screening process, the researcher tells the program exactly what to target,” Meika said. “In this process, we told it to look for functions that identified potential drugs and allowed us to define a target.”
When the team tested the top compounds in AML cell lines or in cancer cells donated by AML patients, they found that the compounds selectively targeted AML cells. They also tested the new compounds in combination with existing AML drugs, where they show a synergistic effect; the combination killed up to 95-97% of the cancer cells but only 5% of healthy blood cells.
“Collaborating with chemists, biologists, computational biologists and clinicians enabled us to develop this pipeline and identify a promising new target with potential drug candidates,” Kirienko said. “Now, we can use this pipeline to look for new compounds and targets in other cancers.”
Key collaborators in this work include Scott Gilbertson, a medicinal chemist from University of Houston, and Simona Colla and Steven Kornblau from The University of Texas MD Anderson Cancer Center, who also provided access to patient samples from MD Anderson’s Leukemia Sample Bank.
This work was funded by the National Institutes of Health’s National Institute of General Medical Sciences (R35GM129294, T32GM139801), the National Cancer Institute (R21CA280500, 3R21CA280500-01A1S1) and the Cancer Prevention and Research Institute of Texas’ High-Impact/High-Risk Research program (RP250573).