Severe microbial infections, such as tuberculosis and MRSA, often require treatment with three (or more) drugs. Multidrug therapies are currently discovered by empirically testing large numbers of potentially therapeutic combinations. This labor-intensive method is necessary because there isn’t a general method for predicting how a combination of drugs will affect cell growth from the effects of the drugs individually or in smaller subsets. Establishment of a predictive framework for the effects of multidrug combinations would decrease the number of required experiments and provide a step towards more rational design of multidrug therapies.
In a recent study published in PNAS, Cluzel and colleagues show that inhibition of bacterial growth by a multidrug treatment can be predicted from growth responses to the component drugs singly and in pairs. Further, the authors demonstrated that multidrug growth responses could be predicted using a simple algebraic formula that should be widely applicable in pharmacology.
The authors, headed by senior postdoctoral fellow Kevin Wood, experimentally measured the responses of gram-negative bacteria (Escherichia coli) and gram-positive bacteria (Staphylococcus aureus) to a wide variety of drugs, including protein synthesis inhibitors (macrolides, aminoglycosides, tetracyclines, lincosamides, and chloramphenicol), DNA synthesis inhibitors (fluoroquinolones and quinolones), folic acid synthesis inhibitors (sulfonamides and diaminopyrimidines), inhibitors of cell wall synthesis, polypeptide antibiotics, preservatives, and analgesics, singly and in combinations and at a range of concentrations.
Borrowing a technique from statistical physics, Wood et al. used entropy maximization to show that apparent interactions among combinations of three and four drugs came almost entirely from the pairwise interactions between component drugs. Their predictions of bacteria’s growth in the presence of multiple drugs were accurate even in cases where the measured, apparent interaction of the multidrug combination was qualitatively different from that of the pairwise interactions. Though the authors used a rigorous, formal framework (entropy maximization) to ensure that the contribution of pure three- and four-drug interactions was very small compared to that of pairwise interactions, for practical purposes, other researchers may apply their findings using simple algebraic equations.
Because their predictive framework did not depend on any intracellular mechanisms that might be specific to E. coli or to S. aureus, Cluzel and colleagues hypothesize that it will be extensible to other bacteria, many strains of which have become resistant to previously powerful antibiotics and now require multidrug therapy.
These results have significant practical applications in drug discovery. When testing multidrug therapies, the number of empirical experiments needed increases exponentially. For example, testing the efficacy of all possible combinations of the 19 drugs used in this study at just 3 different dosages each would require about one billion experiments. Testing of the drugs singly and in pairs would only require about 1,600 measurements, which could then be used in the formula to discover the most therapeutically promising combinations.
Read more in HARVARDgazette