Applying dynamic precision medicine with intra-tumoral bet-hedging for winning drug combinations…

Harvard scientists shared intriguing insights a few years ago in Cell. They suggest that the success of approved cancer drug combos might not be due to the drugs boosting each other's effects. Instead, it seems to be a strategy of diversification, much like an investment portfolio.

After reviewing data from 15 trials, it appears that these drug pairs work independently to improve survival rates, rather than synergistically, when stacking up against single-drug therapies. A nice summary from Peter Sorger, who coined the term, is here.

This is an interpatient way of looking at combinations and hopefully we’re moving past that now to precision medicine strategies….

However, bet-hedging could still be relevant in an intra-patient sense.  Tumors are heterogeneous, and targeted treatments for advanced disease often fail because of heterogeneity for a couple of reasons:

Acquired resistance –

A sensitive cell develops another mutation which allows this cell to become resistant to the drug. This cell now has an evolutionary advantage when under the selection pressure of the drug so it divides rapidly, creating a resistant subclone.

Pre-existing low-level resistance –

This can be a problem for targeted treatment in 2 ways:

1.     Not all cells in the tumor have the target mutation prior to treatment. The targeted treatment effectively targets the clone with the target mutation, but other cells exist in the tumor that don’t have the target mutation. Treatment then leaves more resources (space, nutrients etc) for the clone(s) without the mutation to grow out.

2.     There are initially cells present that have both the target mutation AND a resistance mutation.  Some cells may have the target mutation, but also have a mutation at the site the drug binds which makes them resistant to the drug. It seems far-fetched, but many tumor cells have ways to create DNA damage to increase heterogeneity and ensure survival. These mutations can be in very small pockets of the tumor that are undetectable in testing.

The resolution of our assays mean we can often not distinguish these scenarios because of (a) the resolution of our assays, or (b) our inability to sample and sequence the whole tumor.

Scenarios of pre-existing low level and acquired resistance in cancer. Adapted from Sharma et al, Cell 2017

Combinations to combat acquired resistance in advanced NSCLC

For acquired resistance, why not bet-hedge and combine 2 inhibitors for the same target either in combination or in a short pulsed sequence?

Even though we cannot detect the resistance mutations, we assume they are there and treat alongside another drug, that has the same target, but binds the protein in different way.

An example in lung cancer…

NCCN guidelines preferred first line treatment for non-small cell lung cancer (NSCLC) with EGFR mutations is osimertinib, which head-to-head with other EGFR inhibitors showed the best results. However, in a study conducted at Dana Farber Hospital, 13% of patients on this regimen progress due to emergence of a new EGFR mutation that is where the drug binds to the protein (C797S). As the progression in these patients is much slower than other patients, it’s thought that these mutations may be acquired on treatment.

If osimertinib is combined with afatinib or gefitinib (EGFR mutation inhibitors with different binding sites) maybe you could delay progression in these patients….

Why not bet-hedge and combine 2 inhibitors that have different targets?

In the same study as above, 68% of NSCLC patients more slowly progressed with loss of a target mutation (T790M) along with emerging gene fusions, MET amplifications and KRAS mutations.

If osimertinib is combined with a KRAS inhibitor, or a MET inhibitor, maybe you could delay progression in these patients….

I love Robert Gatenby’s work on how you can use evolutionary models to design sequential combination regimens. If you haven’t checked out his podcast, take a look here.

We will likely never have the techniques available to detect low level resistant clones. There are approximately 100 million cells in a 1cm3 epithelial tumor (about the size of a fingernail and approximately the resolution of a traditional MRI scan). Even in a single resected tumor that small, not every cell will be represented in the sequencing test. In many patients the tumors are even larger, there may be multiple tumors, and some are impossible to biopsy. Bet-hedging, combined with an evolutionary approach to drug scheduling, may be the best option for successful combination therapy.