Growth feedback as a basis for persister bistability

A small fraction of cells in many bacterial populations, called persisters, are much less sensitive to antibiotic treatment than the majority. Persisters are in a dormant metabolic state, even while remaining genetically identical to the actively growing cells. Toxin and antitoxin modules in bacteria are believed to be one possible cause of persistence. A two-gene operon, HipBA, is one of many chromosomally encoded toxin and antitoxin modules in Escherichia coli and the HipA7 allelic variant was the first validated high-persistence mutant. Here, we present a stochastic model that can generate bistability of the HipBA system, via the reciprocal coupling of free HipA to the cellular growth rate. The actively growing state and the dormant state each correspond to a stable state of this model. Fluctuations enable transitions from one to the other. This model is fully in agreement with experimental data obtained with synthetic promoter constructs.As far back as the 1940s, it was known that a small fraction of a bacterial population can survive even when exposed to prolonged antibiotic treatment (1, 2). This phenomenon is termed persistence and members of the surviving subpopulation are called persisters. It has been estimated that the frequency of persisters in normal wild-type populations is extremely small, perhaps of order (3). Although the number of persisters is tiny, they are often the main obstacle to attempts to completely eradicate infection.Remarkably, there is no apparent change in the persisters’ DNA sequence; i.e., their survival is not due to mutation (4). Already in 1944, Bigger suggested that persisters are phenotypically different, in a dormant state instead of an actively growing state (1). The dormant state is presumably better able to deal with common antibiotics, which typically target only actively growing cells. Bigger’s assumption was confirmed by a later study (3). In this study, Balaban et al. investigated the persistence of a single cell of Escherichia coli by using a microfluidic device. They showed that individual persisters do not always remain in the dormant state. Instead, they stochastically transit into an actively growing state and these newly transited cells are indistinguishable from other normally growing cells. Conversely, normal cells can transit into the persistent state. Thus, bacterial persistence at the population level is governed by a single-cell “phenotypic switch.” The precise workings of this switch have to date remained unclear.In the 1980s, Moyed and Bertrand identified the first high-persistence mutant, HipA7, having a persister frequency that is near 10⁻² (4). The discovery of HipA7 facilitated the study of bacterial persistence due to its relatively high proportion of persisters. It was found that HipA7 is formed by a two-residue substitution in the HipA protein. This protein acts as a toxin in a toxin–antitoxin (TA) module (5, 6), where the hipB gene is coexpressed with hipA and the corresponding protein binds to and neutralizes HipA toxicity. To date, HipA is one of only a few molecules that are validated tolerance factors (7).There have already been several models proposed for the Hip system and its connection to persistence. Two modeling groups have claimed that fluctuations cause the apparent coexistence of these two phenotypes, growing and dormant, even though there may or may not be any formal bistability. They were partially driven to this conclusion by their inability to find bistability in their assumed dynamics. The pioneering model of Rotem et al. (8) did not consider the dimerization of HipB and the repression by the HipB dimer of the hip promoter. In the alternate formulation of Koh and Dunlop (9), the HipA-dependent reduction of the translation rate and the growth rate is not included. Thus, both these works claim that bistable states are not necessarily the mechanism underlying persister formation. However, models with a single stable state invariably predict fast transitions between persisters and normally growing cells. For example, simulations in ref. 9 show that transitions from persisters to normally growing cells typically happen within 1 h. In contrast, a sizeable number of persisters can survive even when the antibiotic treatment is maintained for longer than 1 d. If cells stay in a persister state only for less than 1 h, and the persister becomes fragile when it transits into the normally growing state, they would not survive much longer than the other normal cells. The correct picture must include a long-time duration of the persister state.One model has indeed suggested that bistability is the key to the formation of persisters (10). This model made some assumptions now known to be inaccurate, for example that free HipA undergoes dimerization and that the binding of the HipA-HipB complex to the hip promoter is independent from the binding between the HipB dimer and the hip promoter. (Actually they compete with each other in binding to the same operator sites.) However, this model does explain an interesting observation, that often persisters are formed much more readily in stationary phase and in fact persisters seen in normal exponential phase are often just the remnant of persisters formed at a different growth stage. This pattern has been called type I persistence (3) and is the type seen in the HipA7 mutant. As we will see, this occurs due to the fact that the range of bistability can depend on the growth condition. A different issue is that this model is fully deterministic and hence cannot address stochastic effects such as transitions between the two stable states.The drawbacks of these models have motivated us to construct a more precise and comprehensive stochastic model for the HipBA system. A recent paper revealing the structure of HipA and its binding has helped guide us to correct the assumptions in the previous bistable model (11). We show that our approach can consistently account for different classes of experimental data and hence can form a framework for continuing analysis of this important survival strategy for wide classes of bacteria.