Quantitative dose-response analysis untangles host bottlenecks to enteric infection

Abstract

Host bottlenecks prevent many infections before the onset of disease by eliminating invading pathogens. By monitoring the diversity of a barcoded population of the diarrhea causing bacterium Citrobacter rodentium during colonization of its natural host, mice, we determine the number of cells that found the infection by establishing a replicative niche. In female mice the size of the pathogen's founding population scales with dose and is controlled by a severe yet slow-acting bottleneck. Reducing stomach acid or changing host genotype modestly relaxes the bottleneck without breaking the fractional relationship between dose and founders. In contrast, disrupting the microbiota causes the founding population to no longer scale with the size of the inoculum and allows the pathogen to infect at almost any dose, indicating that the microbiota creates the dominant bottleneck. Further, in the absence of competition with the microbiota, the diversity of the pathogen population slowly contracts as the population is overtaken by bacteria having lost the critical virulence island, the locus of enterocyte effacement (LEE). Collectively, our findings reveal that the mechanisms of protection by colonization bottlenecks are reflected in and can be generally defined by the impact of dose on the pathogen's founding population.

Fig. 1. A constrictive bottleneck impedes C. rodentium enteric colonization.

a Experimental strategy. To quantify the contraction (bottleneck) of the C. rodentium population during colonization a barcoded library of bacteria was orally administered to mice. The number of unique cells from the inoculum (founders) that expand to produce the observable population (burden) was determined from the frequency and diversity of barcodes. b 5 cohoused, adult, female, C57BL/6 J mice were administered 4 × 10⁸ CFU STAMP-CR253 by oral gavage and the C. rodentium population was monitored in the feces for 5 days post inoculation (p. i.). Dose and burden were enumerated by serial dilution and plating; founders were determined by STAMP. Lines and error are geometric means and standard deviation. Bacteria not detected (ND) counted as 0.5, CFU colony forming units, g grams. Validation of STAMP libraries in Supplemental Fig. 1. Source data are provided as a Source Data file.

Fig. 2. The C. rodentium bottleneck is defined by a fractional relationship between dose and founders.

a Models for the relationship between dose and founding population. In the absence of a bottleneck, all bacteria from the inoculum become founders. If the inoculum contracts due to the limited availability of finite nutrients or niches (e.g. iron, sugar, binding sites), the diversity of the population and thus the size of the founding population will remain fixed once those nutrients are saturated. If increasing dose increases the number of founders, then the underlying mechanism is not due to a limited resource; instead, the bottleneck acts proportionally on the inoculum by eliminating potential founders. b, c C57BL/6 J mice were inoculated with doses ranging from 10⁷ to 10¹⁰ CFU of STAMP-CR253 and the C. rodentium population was monitored in the feces (geometric means and standard deviations; ND not detected counted as 0.5). Additional shedding analysis in Supplemental Fig. 2. c The bottleneck impeding B6 colonization is described 5 days post inoculation by comparing dose and founders with a linear regression of the log₁₀-transformed data (regression line with 95% confidence intervals; not detected counted as 0.8; x-intercept “ID₅₀” 10^7.2–10^7.9 CFU). 4–8 animals per dose. Source data are provided as a Source Data file.

Fig. 3. Stomach acid constricts the C. rodentium population by 10- to 100-fold.

a The effect of Loxtidine on stomach acid in C57BL/6 J mice 3–5 h after intraperitoneal administration of 1 mg in 0.1 ml PBS. pH determined post-mortem in aspirated stomach fluid. Boxes arithmetic mean (2.5 for mock and 4.7 for Loxtidine). Two-tailed t test with p-value 0.0002. Animals are 7 (vehicle) and 5 (Loxtidine). b The acid tolerance of STAMP-CR69K in culture measured by diluting cells in LB at pH 2.5 or 4.7 and incubating at 37 °C with shaking. pH 2.5 sterilized 10¹⁰ CFU in 15 min. c, d 3–5 h after intraperitoneal administration of PBS (vehicle) or Loxtidine (1 mg), C57BL/6 J mice were orally gavaged with doses ranging from 10⁷ to 10¹⁰ CFU of STAMP-CR69K. c Bacterial burden monitored in the feces for 5 days following inoculation (geometric means and standard deviations; ND not detected counted as 0.5). d Bottleneck to colonization measured 5 days post inoculation by comparing dose and founders with linear regression of the log₁₀-transformed data (regression line and 95% confidence intervals; significance compares elevation with p-value 5.5 × 10⁻⁷; not detected counted as 0.8). 4 animals per dose per group. Source data are provided as a Source Data file.

Fig. 4. Infection is initiated by related populations of C. rodentium in the cecum and colon.

a–e C. rodentium populations in whole organ homogenates from 5 cohoused (intra-cage) C57BL/6 J mice 5 days post inoculation with 4 × 10⁸ CFU of STAMP-CR253. Within a mouse (intra-mouse) the C. rodentium populations at the primary sites of colonization (cecum, proximal colon, distal colon, feces) share founders (number, identity, and frequency of barcodes). a, b Lines connect intra-mouse samples. c Clustering of barcode populations by principal component analysis (PCA). d, e Relatedness determined by comparing the barcode frequencies by genetic distance (arithmetic means) with zero indicating no difference between populations (identical). e Two-tailed t test with p-value 5.8 × 10⁻⁴⁸. p proximal, m mid, d distal, SI small intestine. Heatmap depicting genetic distance relationships of all intra-cage populations in Supplemental Fig. 3. Source data are provided as a Source Data file. f, g To determine when/where C. rodentium establishes a replicative niche, C57BL/6 J mice were orally gavaged with between 3 × 10⁹ and 6 × 10⁹ CFU STAMP-CR253. Following dissection, the cecum and colon were flushed to separate organ adherent (f) and luminal (g) bacteria. Burden and founders display geometric means and standard deviations. Bacteria not detected (ND) counted as 0.5. Relatedness of populations was determined by comparing the barcode frequencies of colon and cecal populations from within the same animal (intra-mouse) by genetic distance (arithmetic mean and standard deviation). 22 animals. Source data are provided as a Source Data file. h Model depicting how related C. rodentium populations could initiate infection in both the cecum and colon: (1) the inoculum minorly constricts passing through the stomach and SI to deposit diverse populations in the cecum and colon, (2) the populations in the cecum and colon contract separately over the first 24–48 h becoming dissimilar, and then (3) expansion occurs in either the cecum or colon moving to both locations. We depict the movement from cecum to colon as we judge this to be more likely, but the opposite is possible.

Fig. 5. Host genotype impacts the bottleneck to C. rodentium colonization.

C3H/HeOuJ (a) or C57BL/6 J (b) mice were inoculated with doses ranging from 10⁶ to 10¹⁰ CFU of STAMP-CR253. The C. rodentium population was monitored in the feces (geometric means and standard deviations; ND not detected counted as 0.5) and animal health assessed by weight loss (percent compared to pre-inoculation; arithmetic means and standard deviations) and body condition. For survival, lines are percent of initial animals not moribund. c The bottleneck to C3Ou and B6 colonization is described 5 days post inoculation by comparing dose and founders with a linear regression of the log₁₀-transformed data (regression line with 95% confidence intervals; significance compares elevation with p-value 5.5 × 10⁻⁷; not detected counted as 0.8). B6 bottleneck data is repeated from Fig. 2. 4 animals per dose. Source data are provided as a Source Data file.

Fig. 6. Streptomycin treatment ablates most of the bottleneck preventing C. rodentium colonization.

The microbiota of conventional C57BL/6 J mice was reduced by treatment with the antibiotic streptomycin for the 3 days prior to inoculation with a streptomycin resistant library of STAMP-CR69K. a Founding population and bacterial burden monitored in the feces, and (b, c) the bottleneck to colonization measured by comparing dose and founders (geometric means and standard deviations; resolution limit is ~10⁶ founders). (a) Animal health monitored by weight loss (percent compared to pre-inoculation; arithmetic means and standard deviations) and body condition. For survival, lines are percent of initial animals not moribund. Untreated B6 bottleneck data is repeated from Fig. 2 for comparison. 4 animals per dose. Streptomycin treatment does not impact stomach acidity (Supplemental Fig. 5). Source data are provided as a Source Data file.

Fig. 7. The bottleneck to C. rodentium colonization is microbiota dependent.

Germ-free C57BL/6 J mice were orally inoculated with doses ranging from 10² to 10¹⁰ CFU of STAMP-CR69K. 4 animals per dose, cohoused with animals receiving the same dose in sterile cages. Measurement of germ-free stomach acidity in Supplemental Fig. 5. Source data are provided as a Source Data file. a Bacterial burden and founding population monitored in the feces (geometric means and standard deviations; resolution limit is ~10⁶ founders). Animal health monitored by weight loss (percent compared to pre-inoculation; arithmetic means and standard deviations) and body condition. No animals became moribund. b, c Bottleneck to colonization measured by comparing dose and founders (geometric means and standard deviations). SPF B6 bottleneck data is repeated from Fig. 2 for comparison. d To determine if colonization was accompanied by changes in the C. rodentium genome, whole genome sequencing was performed on 3 clones (colonies) from the STAMP-CR69K input library and compared to clones isolated from feces of infected mice (1 colony per mouse). Boxes represent the genome status of the LEE pathogenicity island. No LEE genomic changes wild-type (wt), deletion of the entire LEE region (del), insertion within the LEE (ins). Mice with a conventional microbiota SPF specific pathogen free. Other genomic changes listed in Supplemental table 1. e Depiction of a ~100,000 base-pair region of the C. rodentium genome containing the LEE pathogenicity island. Read depth from STAMP-CR69K inoculum and 5 clones isolated after 20 days passage in otherwise germ-free animals. Large deletions in 3 of 5 clones revealed by lack of specifically mapped reads in regions of up to 97,691 base-pairs. Non-specific reads map to multiple loci in the genome (primarily transposons).