Targeted therapies are a classical example of a glass that can be seen as half full or half empty.

Since the approval of Gleevec (imatinib) 2 decades ago, oncology has racked up an impressive number of successes in inhibiting driver mutations. With advances in medicinal chemistry, there are now drugs against targets that were formerly poster children for undruggability, such as Lumakras (sotorasib), approved for non-small-cell lung tumors with a G12C mutation in the KRAS GTPase.

Overall, however, targeted therapies are only available for about 25% of patients.

Many driver mutations cannot be directly targeted for different reasons. And even for some mutations that can be directly targeted, some clinicians question whether they should be.

For example, Tibsovo (ivosidenib; Servier Pharmaceuticals), which is approved for the treatment of acute myeloid leukemia (AML), directly inhibits mutated isocitrate dehydrogenase 1 (IDH1) to prevent the buildup of the oncometabolite 2-hydroxyglutarate (2HG). But there is some evidence that like BRCA-mutated tumors, IDH1-mutated tumors are particularly vulnerable to treatment with PARP inhibitors, a phenomenon called synthetic lethality.

Synthetic lethality as an alternative approach to direct inhibition was on display at the 2021 Molecular Targets meeting as well.

The conference opened with an educational session on synthetic lethality.

And at a Saturday session on protein-protein inhibitors, Chris Vakoc, co-director of the Cancer Center at Cold Spring Harbor Laboratories, reported on his team's discovery that mutations in the myoinositol transporter SLC5A3 created a synthetic lethal dependency with ISYNA1, the gene that encodes the rate-limiting enzyme for myoinositol biosynthesis, in AML.

The work that Vakoc described at the meeting, and that was recently published in Cancer Discovery, began with an attempt to see how well interactions now being catalogued in databases such as the Cancer Dependencies Map (DepMap) through the use of cell-based approaches correlate with data obtained in vivo.

The DepMap has safeguards built in to avoid identifying dependencies that have more to do with cells' adaptations to an in vitro state, and less with clinically relevant changes. Vakoc lauded the database, calling it the gold standard for uncovering targetable interactions.

Nevertheless, like everything, DepMap has blind spots.

However, the use of cells does not seem to be a major blind spot for DepMap.

When the team used targeted CRISPR screens in an animal model of AML to look at dependencies identified in DepMap and other in vitro-based databases, "we were struck by the fact that most dependencies we observed in vivo correlated well with in vitro," Vakoc said.

Digging deeper into the dataset, though, they identified SLC5A3 as an in vivo dependency that was both novel, and specific to AML.


To fuel their growth, cancer cells take up many nutrients and building blocks from the extracellular space, a phenomenon called auxotrophy. Recombinant asparaginase is sued in the treatment of AML, where cells are highly dependent on extracellular asparagine.

In their experiments, Vakoc and his team showed that some AML cells are auxotrophs for myoinositol because ISYNA1 expression is epigenetically silenced.

Myoinositol's most studied function is as component of lipid phosphatidylinositols (PIs), which are important for signaling cascades downstream of growth factors. Its phosphorylation state is regulated by kinases that are intimately familiar to cancer researchers -- PI3K and PTEN.

Vakoc and his colleagues identified recurrent ISYNA1 silencing in samples from AML patients, and showed, using gain- and loss-of-function experiments in animals, that cells with epigenetically silenced ISYNA1 were specifically vulnerable to silencing of SLC5A3.

That silencing itself, Vakoc stressed, is a passenger event with no apparent effect on fitness. But it created a synthetic lethal combination with SLC5A3 inhibition.

The Cancer Discovery paper was published online on September 19, 2021, back to back with a report from investigators at Dana-Farber Cancer Institute reporting broadly similar results.

Transporters could make good drug targets, as they tend to have binding pockets that could be targeted with small molecules. Still, important questions remain to be answered about the therapeutic relevance of the approach.

Transporter proteins are rarely specific to one molecule, and SLC5A3 is no exception. It will also transport glucose, and "probably glucose transporters can bring in myoinositol" if its concentrations are high enough.

Then there is the ever-present heterogeneity. "Silencing in cancer is often stochastic," Vakoc said, "and it's probably not an irreversible event."