How type 1 fimbriae help Escherichia coli to evade extracellular antibiotics

To survive antibiotics, bacteria use two different strategies: counteracting antibiotic effects by expression of resistance genes or evading their effects e.g. by persisting inside host cells. Since bacterial adhesins provide access to the shielded, intracellular niche and the adhesin type 1 fimbriae increases bacterial survival chances inside macrophages, we asked if fimbriae also influenced survival by antibiotic evasion. Combined gentamicin survival assays, flow cytometry, single cell microscopy and kinetic modeling of dose response curves showed that type 1 fimbriae increased the adhesion and internalization by macrophages. This was caused by strongly decreased off-rates and affected the number of intracellular bacteria but not the macrophage viability and morphology. Fimbriae thus promote antibiotic evasion which is particularly relevant in the context of chronic infections.

following an exponential binding behavior (Eq. 2). (b) Root mean square error (RMSE) values of the two-step, reversible and one-step, irreversible model for bacterial adhesion. An RMSE value was obtained as an estimate of the fit performance for each adhesion curve and the respective fit from the adapted Michaelis Menten (Eq. 1) and the exponential adhesion model (Eq. 2). Error bars are S.D. fim↑, fimbriae overexpression strain; wt, fimbriae wild type strain; ∆fim, fimbriae knockout strain; αMM, alpha-methyl mannosepyrannoside; LatB, LatrunculinB.

Modeling the adhesion kinetics of bacteria to macrophages a) Michaelis Menten model: Assuming a reversible, non-cooperative two-step internalization process
From the observation of a saturation trend in the adhesion of bacteria to macrophages for the fimbriae overexpression strain (Fig. 2b), we tested if Michaelis Menten-like kinetics could describe the experimentally determined adhesion data. Cooperativity is an important factor in binding kinetics and has been shown e.g. in the infection of epithelial and HeLa cells by Salmonella 1 . Characteristic binding curves of cooperative processes show a sigmoidal dose response curve, a feature that we did not observe in our data. We thus tested a noncooperative Michaelis-Menten model for its ability to fit the experimental dose response data. Accordingly, initial adhesion of bacteria to macrophages was assumed to be reversible, such that physical contact between bacteria and macrophage can result in transient, weak adhesion at a rate kon with unbinding at a rate koff. In a subsequent step, bacteria can stably adhere to the macrophage surface and be taken up in a slower, rate-limiting step at a rate k.
The rate limiting and irreversible step results in the GFP-positive macrophage population which can be detected by flow cytometry. Note that the Ks from the fit is not a rate per se, but a dimensionless ratio of two rates which is inversely proportional to the binding rate in the exponential model, since Ks ~ 1/kon. The maximum amount of macrophages that can be bound was set to an upper limit of 100%.
b) The one-step adhesion model: Assuming an irreversible, non-cooperative internalization process: In an alternative approach on modeling the rate of bacterial adhesion, we tested the assumption that the adhesion of bacteria is irreversible, i.e. contact between receptors and bacterial ligands immediately leads to formation of a stable complex that results in phagocytosis such that dissociation of the receptor-ligand complex is not possible.
with the rate of internalization R(x), the bacteria to macrophage ratio x, the adhesion on-rate k and the time t.
Furthermore, a non-cooperative adhesion process was assumed, since the binding curve obtained from the raw data did not show a sigmoidal behavior as had been observed for example for Salmonella mediated infection of epithelial cells 1 . In this case one would also expect to observe a sigmoidal shape of the binding curve. In analogy to a reaction scheme: The change of the fraction of infected macrophages r depends on the bacteria to macrophage ratio x, and the bacterial adhesion rate k, and the fraction of not yet infected macrophages (1-r): This ordinary differential equation has the solution Since x, but not t was varied in the experimental design, a resulting fit using this exponential expression will give the parameter k. The root mean square error (RMSE) values of both models showed that the deviation between fit and experimental data was lower for the Michaelis Menten model ( Supplementary Fig. S2b).
The fit parameters are shown in the Figure 3d and 3e of the main text. The absolute numbers are not related to experimental parameters since the flow cytometric assay did not quantify the number of bacteria per macrophage, or the number of adhesion attempts per time as only fixed timepoints were available. Under the assumptions we used, the purpose of the model is to estimate differences in adhesion rates for the different proposed steps, information that would otherwise not be experimentally accessible.

Statistical analysis
For an overview on all statistical tests that we applied in this study, the results from the one-way ANOVA and one-tailed t-test are found in the Supplementary tables T1-T8.
Supplementary Supplementary movie captions Supplementary Movie M1: 3D reconstruction from confocal images of surface adherent macrophages incubated with fim↑ at a bacteria-to-macrophage ratio of 10. Bacteria were incubated with macrophages for 0.5 hours before fixing, staining and image acquisition.
Supplementary Movie M2: 3D reconstruction from confocal images of surface adherent macrophages incubated with ∆fim at a bacteria-to-macrophage ratio of 10. Bacteria were incubated with macrophages for 0.5 hours before fixing, staining and image acquisition.