The Siren Song of Synergy

Summary:

In ancient Greek mythology, sirens were creatures of stunning beauty whose mystical songs led sailors to sail their boats onto hidden rocks and into total destruction. In this issue, Mason-Osann and colleagues present data in the context of acute myelogenous leukemia to suggest that while synergy may show initial attractions in drug combinations, it may carry with it hazards previously unforeseen.

See related article by Mason-Osann et al., p. 95 (1).

It may be hard to imagine why synergy in a drug combination could possibly be bad. While there are different algorithms for mathematically identifying synergy, synergy can simply be described as a situation where two drugs combined work vastly better than one might expect based on the effect of either used alone. This suggests that one could use low, presumably less toxic, doses of drugs and still get good efficacy. How could that be bad? One way to understand this is to consider that synergy is usually defined in short-term assays that evaluate immediate killing by a combination. However, clinical response is usually determined over many days, roughly a month in the case of most initial treatments for acute myelogenous leukemia (AML). The longer period of time allows for adaptation to or selection for phenotypes that are resistant to the treatment. Over these longer, more clinically relevant periods, therefore, one might posit that properties of combinations related to acquisition of resistance might be more important than those that mediate initial cytotoxic response. Then the question becomes, “Are there theoretical reasons to expect that a cancer treated with a combination based on synergy would more rapidly develop resistance?” Indeed there is one, which is that selection for a resistance mechanism in either drug would eliminate effectiveness of both drugs, because they depend on synergy for their efficacy. Therefore, there are simply more possible mechanisms available to develop resistance simultaneously to both drugs in a synergistic combination. In a combination where the drugs are not interacting, to develop resistance simultaneously to both drugs resistance to both would have to be acquired, not just resistance to either (Fig. 1). By this reasoning, therefore, one might expect resistance to arise more rapidly in combinations with properties of synergy.

Figure 1.

A model comparison of the effects of acquisition of single-drug resistance on overall combination efficacy. A and B are two different hypothetical drugs. In the hypothetical case of a synergistic combination between A and B (blue bars), A and B have little ability to kill cells individually, but together kill 80%. However, when single-drug resistance to either or A or B emerges, the efficacy of the entire combination is lost. In the hypothetical case of two drugs that kill via independent mechanisms (yellow), drug A can kill 30% as a single agent, drug B can kill 50% as a single agent, and the combination can kill 80%. However, now when single agent resistance to either A or B emerges, there is still residual ability for the remaining active drug to kill cancer cells.

In this issue, the Palmer and Mettetal labs put these concepts to an experimental test (1). To evaluate the relationship between early drug combination cytotoxic activity and the rate of development of resistance, they evaluate growth rates of two AML cell lines after treatment with pairwise combinations of four therapeutics. At 72 hours after treatment, Mason-Osann and colleagues identify drug concentrations where the combinations exhibit either synergy or additivity. They also identify resistance in response to the same drug combinations by continuously measuring growth rates over 21 days and determining instances where cell growth is initially inhibited but increases at later points. Importantly, the authors find that resistance measured over 21 days is more likely to develop when 72-hour treatments score as synergistic (38% of synergistic combinations develop resistance) relative to when 72-hour treatments are additive (21% of additive combinations develop resistance).

The authors also test the relationship between synergy at 72 hours and the fitness of the resistant cells. They calculate fitness as the difference between the peak growth rate when resistance emerges and the initial growth rate after the combination treatment. While the authors find a correlation between the fitness of the resistant cells and the synergy score of the drug combination treatment, there are instances where the fitness of resistant cells is high even when drug combinations are additive. Nonetheless, this trend of increased fitness is one reason why drug combination synergy may compromise response durability.

The direct evaluation that drug synergy promotes resistance is potentially more complex in vivo. However, the authors provide supporting evidence that initial tumor shrinkage in vivo is not correlated with the durability of response. First, they retrospectively analyze treatments of patient-derived xenograft (PDX) models showing that the shrinkage rate of these models does not correspond with progression-free survival of the animals. In support of this, Mason-Osann and colleagues, compare in event-free survival in a clinical trial of CPX-351 (a liposomal formulation of cytarabine and daunorubicin found to act synergistically in vitro) to standard cytarabine and daunorubicin treatment (found to act additively in vitro). While CPX-351 shows an initial benefit after 3 months, this is diminished by 12 months which indicates that the benefits of an initial response do not indicate a durable response. While conventional measurements of drug response after 72 hours are more convenient for high-throughput drug combination screening, Mason-Osann and colleagues, make a compelling case that the short-term benefits of some drug combinations may not translate into long-term responses. However, a comprehensive evaluation across more drug combinations, and tumor types are required to fully evaluate this. Nonetheless, the methods and analysis presented in this study are amenable to such an expansion.

Perhaps a synergy fan would maintain, “Sure, maybe sometimes you get unlucky with resistance, but don't you need some synergy to have an effective clinical combination?” This issue was considered previously by Palmer and colleagues. Let us consider rituximab (R), cyclophosphamide (C), doxorubicin (AKA hydroxydaunorubicin, H), vincristine (AKA oncovin, O), and prednisone (P; R-CHOP) chemotherapy, a combination regimen that has cured millions of lymphoma patients over the years, and can be considered the most valuable combination regimen in the history of oncology. Is there a synergistic interaction among any of these R-CHOP drugs? It turns out, the answer is no (2). Similar results pertain to several other combinations used in the clinic (3–5), showing that independent drug action is sufficient, and perhaps even superior to synergy, as a priority in constructing clinically active combinations.

None of this is to say that we should exclude the use of effective combinations that may exhibit synergy. Indeed, for some tumors—particularly in a relapsed or resistant setting— providing any response would be of clinical benefit. Moreover, as Mason-Osann and colleagues discuss, sequential use of combinations that are synergistic and nonsynergistic may be beneficial. Nonetheless, the relationship between synergy and resistance outlined in this study should moderate enthusiasm for the concept of synergy, a concept that is typically uncritically viewed as an unalloyed good.

How can cancer biologists escape the siren song of synergy? Odysseus had his men plug their ears with wax so they could not hear the siren song and had himself lashed to the ship's mast so he could hear the song but not act upon it. While the tying up of certain colleagues is not a prospect to be contemplated wholly without schadenfreude, Odysseus’ remedies are probably too drastic for cancer biologists to employ. Nonetheless, Mason-Osann and colleagues provide us with further evidence that some restraint is needed in the tendency to uncritically seek synergy in deriving clinically active drug combinations. Combinations of drugs acting independently have their place. This and prior studies should serve as a reminder that the classic principles of nonoverlapping toxicities, mechanisms of action, and mechanisms of resistance are still important in designing clinically effective combinations.

Authors’ Disclosures

A. Letai reports personal fees from Flash Therapeutics and Trueline Therapeutics, and personal fees from Zentalis Pharmaceuticals outside the submitted work; in addition, A. Letai is an inventor on several patents relevant to BH3 profiling that are owned by his employer, Dana-Farber Cancer Institute. No disclosures were reported by the other author.