Navigating contradictions: Salmonella Typhimurium chemotaxis amidst conflicting stimuli of the intestinal environment

Abstract

Motile bacteria sense and avoid deleterious stimuli in their environment through chemorepulsion, a behavior that helps them locate permissive ecological niches. In the gut, indole is a bacteriostatic compound produced by the microbiota and is thought to act as a chemorepellent for invading pathogens, thereby protecting the host against infection. The principal reservoir of intestinal indole is fecal matter, a complex biological material that contains both attractant and repellent stimuli. Whether indole in its natural context is sufficient for pathogen chemorepulsion or host protection has remained unknown. Using an intestinal explant system, we show that while pure indole indeed suppresses an infection advantage mediated through chemotaxis for the enteric pathogen Salmonella enterica serovar Typhimurium, this effect is abolished in the presence of other chemoeffectors present in feces, including the chemoattractant L-Serine (L-Ser), in a manner dependent on the chemoreceptor Tsr. Live imaging reveals that although S. Typhimurium is repelled by pure indole, the pathogen is actually strongly attracted to human fecal matter despite its high indole content, and that this response is mediated by Tsr, which simultaneously senses both indole and L-Ser. Fecal attraction is conserved across diverse Enterobacteriaceae species that harbor Tsr orthologues, including Escherichia coli, Citrobacter koseri, Enterobacter cloacae, and clinical isolates of non-typhoidal Salmonella. In a defined system of fecal chemoeffectors, we find that L-Ser and other fecal chemoattractants override indole chemorepulsion, but the magnitude of bacterial chemoattraction is controlled by indole levels. Together, these findings suggest that indole in its native context is not protective against enteric infection and that indole taxis actually benefits pathogens during infection by locating niches with low competitor density. Our study highlights the limitations of applying single-effector studies in predicting bacterial behavior in natural environments, where chemotaxis is shaped by the integration of multiple, often opposing, chemical signals.

Fig. 1.

Chemotaxis-mediated infection advantages in the presence of fecal effectors. A. Overview of the role of Tsr in coordinating responses to conflicting stimuli. B. Experimental design of colonic explant infections. See Materials & Methods for experimental details such as tissue dimensions. C. Conceptual model of the explant infection system. The effectors from the treated tissue (gray) diffuse into the surrounding buffer solution providing a gradient. Note that the bacteria are not immersed in the effector solution, and experience a local concentration during infection much lower than the effector pretreatment. Quantifications of tissue-associated bacteria reflect the ability of chemotaxis to provide an advantage (black arrow) in accessing the intestinal mucosa (reddish brown). D. Serine (presumed to be nearly 100% L-Ser, see Materials & Methods) and indole content of liquid human fecal treatments, as measured by mass spectrometry. E-I. Competitive indices (CI) of colony-forming units (CFUs) recovered from co-infected swine explant tissue, either from the total homogenate (open box and whiskers plots), or from tissue washed with gentamicin to kill extracellular and attached cells, which we refer to as the “invaded” intracellular population (checkered box and whisker plots), as indicated. Each data point represents a single experiment of a section of tissue infected with bacteria, normalized by tissue weight, and the CI of CFUs recovered from that tissue (n=7–10). Boxes show median values (line) and upper and lower quartiles, and whiskers show max and min values. Effect size (Cohen’s d) and statistical significance are noted for each experiment in relation to competitive advantage, i.e. deviation from a CI of 1 (not significant, ns; * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001). See also Figure S1 for competition between WT and an invasion-inhibited mutant invA, and Figure S2 for disaggregated CFU enumerations for each experimental group prior to CI calculation. Data S1 contains all numerical CFU measurements.

Fig. 2.

Salmonella Typhimurium exhibits attraction toward indole-rich liquid human fecal material. A. Diffusion modeling showing calculated local concentrations in CIRA experiments with liquid human fecal material based on distance from the central injection source. B. Max projections of representative S. Typhimurium IR715 responses to a central source of injected liquid human fecal material. C-E. Bacterial population density over time in response to fecal treatment. The initial uniform population density in these plots is indicated with the blue line (time 0), and the final mean distributions with the red line (time 280 s), with the mean distributions between these displayed as a blue-to-red spectrum at 10 s intervals. F-G. Temporal analyses of area under the curve (AUC) or relative number of bacteria within 150 μm of the source. Effect size (Cohen’s d) comparing responses of WT and tsr attraction at 120 s post-treatment is indicated. Data were collected at 30 °C. Data are means and error bars are standard error of the mean (SEM, n=3–5). See also Movie 1, Table S1, and Figure S3.

Fig. 3.

Non-typhoidal Salmonella exhibit fecal attraction. A-E. Dual-channel imaging of chemotactic responses to solubilized human feces by WT S. Typhimurium IR715 (pink) and isogenic mutants or clinical isolate strains (green), as indicated. Shown are max projections at time 295–300 s post-treatment. Data were collected at 37 °C. Data are means and error bars are standard error of the mean (SEM, n=3–5). See also Movie 2, Movie 3.

Fig. 4.

Representative Enterobacteriaceae exhibit fecal attraction. Shown are max projections from CIRA experiments over 5 s before fecal treatment and after 5 minutes of treatment, as well as quantifications of bacteria within 500 μm of the treatment source at these same time points for E. coli MG1655 (A-B, GFP-reporter), E. coli NCTC 9001 (C-D, phase), and C. koseri CDC 4225–83 (E-F, phase). Data were collected at 37 °C. Data are means and error bars are standard error of the mean (SEM, n=3–5). See also Movie 4, Movie 5, Movie 6.

Fig. 5.

Chemotactic responses to defined fecal effector mixtures. CIRA experiments with S. Typhimurium IR715 were performed with different combinations of fecal effectors (n=3–5). Shown are max projections from experiments over 5 s before fecal treatment and after 5 minutes of treatment as well as quantifications of bacteria within 500 μm of the treatment source at these same time points. Data are means and error bars are standard error of the mean (SEM, n=3–5). To achieve the greatest degree of sensitivity to differences in responses, experiments were performed using the same culture on the same day. The complete fecal effector mixture consists of indole (862 μM), L-Ser (338 μM), D-Glucose (970 μM), D-Galactose (78 μM), ribose (28.6 μM), and L-Asp (13 μM), modified to include or exclude certain effectors as indicated. See also Movie 7, Movie 8, Movie 9, and Movie 10 Data were collected at 30 °C.

Fig. 6.

Tsr mediates indole chemorepulsion in S. Typhimurium. A. Representative max projections of responses at 295–300 s of indole treatment. B-C. Quantification of chemorepulsion as a function of indole concentration (n=3–5). D-F. Comparison of WT and tsr mutant responses to L-Ser or indole. E. Shows a quantification of the relative number of cells in the field of view over time following treatment with 5 mM indole for a competition experiment with WT and tsr (representative image shown in F). Data were collected at 30 °C. G-H. Isothermal titration calorimetry (ITC) experiments with 50 μM S. Typhimurium Tsr ligand-binding domain (LBD) and indole, or with L-Ser in the presence of 500 μM indole. Data are means and error bars are standard error of the mean (SEM, n=3–5). AUC indicates area under the curve.

Fig. 7.

S. Typhimurium mediates distinct chemotactic responses based on the ratio of L-Ser to indole. A-D. Representative max projections of responses to treatments of L-Ser and indole at 295–300 s, as indicated. E. Relative bacterial distribution in response to treatments of 500 μM L-Ser and varying amounts of indole, from panels A-D, with the mean value normalized to 100%. Data were collected at 30 °C. Data are means and error bars are standard error of the mean (SEM, n=3–5). F. Diffusion modeling of local effector concentrations based on sources of 5 mM indole (dark brown), 500 μM L-Ser (blue), 500 μM indole (light brown), and 50 μM indole (yellow) are shown as dashed lines. The approximate local concentration of indole that elicits a transition in chemotactic behavior is highlighted in light blue. G-H. Bacterial growth as a function of L-Ser or indole, at the time point where the untreated culture reaches A₆₀₀ of 0.5. I-J. Bacterial growth +/− pretreatment with 500 μM indole or L-Ser, and increasing concentrations of indole or L-Ser, as indicated at the time point where the untreated culture reaches A₆₀₀ of 0.5. Data are means and error bars are standard error of the mean (SEM, n=8–24). See also Movie 11.