LAUSR

Microbes are capable of complex feats of social organization, working collectively to manipulate their local environment. When their local environment contains competing species or strains, the scene is set for microbial war, mediated by chemical and biological weapons of mass destruction (1, 2). Biomedical science has long been interested in chemical warfare among microbes, as these chemicals can be co-opted as essential medicines in the prevention and treatment of bacterial infections—what we know as antibiotics (3). Palmer and Foster (4) take an evolutionary perspective on microbial warfare to ask a central yet startlingly overlooked question of why chemical weapons vary in their range of species destruction or “spectrum.” The authors combine math models and bioinformatic analyses to identify conditions favoring narrowand broad-spectrum antibiotics, raising avenues for both evolutionary and biomedical research. Biomedical scientists have long recognized that antibiotics vary from narrowto broad-spectrum activity and have historically prized “broad-spectrum” drugs as they increase the chances that an unidentified pathogen will be taken down by the drug. Palmer and Foster (4) use mathematical models to explore whether this clinical logic also works for the microbes that make the drug. Is broader spectrum better when engaged in microbial war? Using a series of mathematical models, the authors identify a key logical limitation of a broad-spectrum approach; the chemical weapon will be wasted whenever it kills a bacterium that is not a competitor to the antibiotic-producing organism (Fig. 1A). In contrast, narrow-spectrum compounds that are specifically tuned to only bind and kill directly competing species or strains will produce a higher return on investment via their targeted removal of key competitors (Fig. 1B). If this logic of precision warfare is correct, why do we see so many examples of broad-spectrum antibiotics? Palmer and Foster (4) provide a potential solution to this question by introducing ecological heterogeneities into their math models. Specifically, they model a scenario where a focal antibiotic-producing species is sometimes rare and sometimes dominant within a community. Under this scenario, periods of ecological dominance drive selection for broad-spectrum antibiotic production, as the loss of the antibiotic when killing rare noncompetitors does not limit the availability of the antibiotic to kill competitors (Fig. 1C). After setting out their math-grounded predictions for conditions favoring narrowvs. broad-spectrum antibiotics, the authors turn to a comparative bioinformatic approach to test their ideas. Their model predicts that broad-spectrum antibiotics will be favored if the producer is capable of at least periodically achieving local ecological dominance. Finding data on the antibiotic spectrum is relatively straightforward, thanks to decades of biomedically driven research on the range of activity of antibiotics (and related chemical weapons known as bacteriocins). In comparison, data on the in situ ecological properties of bacteria “in the wild” are unfortunately very limited, and so, the authors turned to proxy measures that are identifiable from genomic features. Specifically, the authors used markers of density-dependent gene expression (e.g., quorum sensing) to identify species that likely experience periods of high density and found that these markers were indeed associated with the production of broad-spectrum antibiotics. The comparative results support their model logic that broad-spectrum antibiotics are used adaptively when producer strains are ecologically abundant (Fig. 1C). The comparative data are a valuable first step in assessing (and ultimately, exploiting) the perspective that the antibiotic spectrum is an evolutionarily modifiable trait. One intriguing avenue for a more direct test is experimental evolution of antibiotic spectrum in drug-producing organisms. While mention of “antibiotics” and “evolution” typically conjures thoughts of resistance evolution, a few studies have raised the possibility that we can also evolve antibiotic producers themselves (5, 6), opening a new front in drug discovery and development. The work of Palmer and Foster (4) suggests that with appropriate passaging designs (simulating Fig. 1 scenarios), we can also select on drug spectrum, offering the unusual potential to both test evolutionary theory and deliver candidate drugs. We note that we see this as a high-risk venture and therefore, not a critical test of the ideas of Palmer and Foster (4). Experimental evolution has proven to be an efficient tool for the rapid adaptation of existing and clearly defined microbial phenotypes, such as growth rate or antibiotic resistance (7). In the case of drug spectrum, we are arguably selecting for novel functions (e.g., the ability to establish binding to a new cell target), placing this work in the rare event space of “evolutionary innovations,” which are typically beyond the scope of most [but not all (8)] experimental evolution projects. The authors highlight the growing interest in narrowspectrum antibiotics in a therapeutic context due to their limited off-target effects on commensal organisms (9, 10). The most widely used current antibiotics are, in the authors’ terms, broad spectrum, and there has been good reason for