The general tenet when trying to identify drug combinations that are effective against resistant bacteria, is that using three or more antibiotics is unlikely to be any more beneficial than using just one or two, either because interactions between the different drugs would largely cancel out any benefits, or because any additional benefits would be insignificant.
Researchers at the University of California, Los Angeles (UCLA) have now turned this assumption on its head, with the discovery of thousands of four- and five-antibiotic combinations that can kill pathogenic Escherichia coli far more effectively than would be expected. The scientists paired their tests in bacteria with a specially developed software, which includes a mathematical framework that can unpick the complexity underlying how the different antibiotics interact with each other either synergistically or antagonistically.
“There is a tradition of using just one drug, maybe two,” says Pamela Yeh, Ph.D., a UCLA assistant professor of ecology and evolutionary biology, who is co-senior author of the team’s published paper in npj Systems Biology and Applications. “We're offering an alternative that looks very promising. We shouldn't limit ourselves to just single drugs or two-drug combinations in our medical toolbox. We expect several of these combinations, or more, will work much better than existing antibiotics.”
The team reported its results in a paper titled, “Prevalence and patterns of higher-order drug interactions, in Escherichia coli.”
Many fields of research are concerned with how factors interact, whether we consider chemical or biochemical reactions, the effects of environmental stressors on species diversity, social interactions in economics, or drug interactions in pharmacology, the authors explain. What scientists look to identify and understand is whether the different components interact in a way that enhances (synergy) or weakens (antagonism) the individual effects of the different components in the mix.
The overall effects of a four-drug combination, for example, could be determined by pairwise interactions, three-way interactions, and/or four-way interactions. The situation can be further complicated when there are emergent interactions that require all the components to be present. “For instance,” the researchers write, “ when the addition of a small amount of a third drug alters the interaction between two drugs – as opposed to the third drug interacting with either of the individual drugs already present – this is an emergent interaction … Therefore, an approach that explicitly distinguishes emergent interactions—interactions that are not due to the presence of lower-order interactions—from the existence of any (net) interaction is indispensable to fully represent complex system dynamics.”
To try and gain a better insight into these higher-order interactions and emergent processes, the team developed a software that incorporates a computational formula called MAGIC (mathematical analysis for general interactions of components), which can analyze how multiple factors interact. They applied the software to the results of testing three doses of eight different antibiotics, either singly, or as 251 two-drug combinations, 1,512 three-drug combinations, 5,670 four-drug combinations, and 13,608 five-drug combinations, on the growth of a pathogenic strain of E. coli. “This full-factorial design of drug combinations allows characterization of net and emergent interactions for all five-way and lower-order interactions (two-, three-, four-way), and thus represents a staggering amount of data compared with previous studies, allowing us to shed new light on how interactions change as more and more drugs (components) are added,” the researchers state.
The results demonstrated that the number of drug interactions, including emergent interactions, increased significantly as the number of drugs in the combination increased. What was most surprising was the finding that adding more drugs to the mix often had antibacterial benefits. “Notably, we find many interactions that only emerge when multiple drugs are present, and even more surprisingly, we find that the frequency of interactions increases as the number of components in the system increases, contradicting the assumptions and limited findings of many previous studies,” the authors re-assert. This, they claim, was “extremely surprising for emergent interactions because these have largely been ignored in the literature.”
Prior to making their analyses, the researchers had predicted the likely effectiveness of each drug combination against E. coli growth. What they found was that 1,676 of the four-drug combinations were more effective than expected, as were 6,443 groupings of five drugs. “I was blown away by how many effective combinations there are as we increased the number of drugs,” says co-senior study author Van Savage, Ph.D., UCLA professor of ecology and evolutionary biology and of biomathematics. “People may think they know how drug combinations will interact, but they really don't.”
It's likely that some of the multi-drug combinations demonstrated positive synergy because the eight different antibiotic drugs tested use six different antibacterial mechanisms to attack the pathogen. “Some drugs attack the cell walls, others attack the DNA inside,” Dr. Savage adds. “It's like attacking a castle or fortress. Combining different methods of attacking may be more effective than just a single approach.”
The effects of increasing the numbers of drugs did work both ways, however. There were 2,331 four-drug combinations, and 5,199 five-drug combinations that were less effected than expected. Overall, it seemed that emergent interactions tended towards antagonism, whereas net interactions tended towards synergy. “Specifically, as more drugs are added, we observe an elevated frequency of net synergy (effect greater than expected based on independent individual effects) and also increased instances of emergent antagonism (effect less than expected based on lower-order interaction effects),” they write.
The researchers hope that their combined experimental and computational approach will help scientists unpick the complexity of drug combinations, “as well as to inform effective design of multidrug treatments.” They are also developing an open-access version of the software, which will hopefully be made available next year so that scientists can analyze their own drug combination results, not just for antibiotic development, but in other disease areas. The same approach can also be applied to areas other than therapeutics, suggests lead study author Elif Tekin, Ph.D. “We think MAGIC is a generalizable tool that can be applied to other diseases – including cancers—and in many other areas with three or more interacting components, to better understand how a complex system works.”
The authors highlight the more widespread utility of their mathematical framework, pointing out that while their study was focused on a drug-bacteria system, the same approach to uncovering and analyzing higher order emergent interactions could help to unravel the complexities of combinatorial interactions in many other fields. “Intriguingly, the general questions explored here for drug interactions are highly relevant for other fields,” they conclude. ” … we anticipate that the research program presented here will be useful for revealing higher-order emergent properties and patterns in ecological, medical, evolutionary, and social systems.”